**Bacterial Actin-Specific Endoproteases Grimelysin and Protealysin as Virulence Factors Contributing to the Invasive Activities of** *Serratia*

#### **Sofia Khaitlina \*, Ekaterina Bozhokina, Olga Tsaplina and Tatiana Efremova**

Institute of Cytology, Russian Academy of Sciences, 194064 St. Petersburg, Russia; bozhokina@yahoo.com (E.B.); olga566@mail.ru (O.T.); tefi45@mail.ru (T.E.)

**\*** Correspondence: skhspb@gmail.com

Received: 30 April 2020; Accepted: 2 June 2020; Published: 4 June 2020

**Abstract:** The article reviews the discovery, properties and functional activities of new bacterial enzymes, proteases grimelysin (ECP 32) of *Serratia grimesii* and protealysin of *Serratia proteamaculans*, characterized by both a highly specific "actinase" activity and their ability to stimulate bacterial invasion. Grimelysin cleaves the only polypeptide bond Gly42-Val43 in actin. This bond is not cleaved by any other proteases and leads to a reversible loss of actin polymerization. Similar properties were characteristic for another bacterial protease, protealysin. These properties made grimelysin and protealysin a unique tool to study the functional properties of actin. Furthermore, bacteria *Serratia grimesii* and *Serratia proteamaculans*, producing grimelysin and protealysin, invade eukaryotic cells, and the recombinant *Escherichia coli* expressing the grimelysin or protealysins gene become invasive. Participation of the cellular c-Src and RhoA/ROCK signaling pathways in the invasion of eukaryotic cells by *S. grimesii* was shown, and involvement of E-cadherin in the invasion has been suggested. Moreover, membrane vesicles produced by *S. grimesii* were found to contain grimelysin, penetrate into eukaryotic cells and increase the invasion of bacteria into eukaryotic cells. These data indicate that the protease is a virulence factor, and actin can be a target for the protease upon its translocation into the host cell.

**Keywords:** actin proteolysis; metalloproteinases; protease ECP 32; grimelysin; protealysin; bacterial invasion

#### **1. Introduction**

Invasion of opportunistic bacterial pathogenic into eukaryotic cells is a process of interaction of bacteria with eukaryotic cells. [1,2]. To invade host cells, invasive bacteria should initiate the host cell signaling system by the proteins that interact with cell receptors and modify cytoskeleton through direct interaction with actin and actin binding proteins. Bacterial effectors can mimic natural activators of small GTPases or directly stimulate the host signaling pathways [3,4]. On the other hand, efficiency of bacterial invasion depends on the physiological state of host cells and is determined by the processes associated with changes in the distribution of cell surface receptors and cytoskeleton rearrangements [5]. This implies the presence of specific bacterial virulence factors capable of interacting with or modifying eukaryotic cell receptors, as well as components of the signal transduction system and actin cytoskeleton.

Studying the mechanisms of actin polymerization, we discovered and characterized a new enzyme-bacterial protease that exhibited highly specific "actinase" activity [6]. The protease turned out to be an intracellular enzyme of bacteria originally identified by standard morphological, biochemical, and cultural properties as an atypical lactose-negative strain of *Escherichia coli* A2 [7]. The actin-specific protease isolated from these bacteria was identified as a single 32 kDa polypeptide, which gave the protease the name ECP 32 (Escherichia coli protease, 32 kDa) [7,8]. Protease ECP 32 cleaves actin

at a single site between Gly-42 and Val-43 within the DNase I-binding loop on the top of the actin monomer [9,10] that is involved in extensive interactions with the neighboring subunits within the actin filament [11,12]. The high specificity of actin proteolysis made ECP 32 a unique tool to study actin properties and interactions [13–17]. On the other hand, the enzyme by itself turned out to be not unique.

Although ECP 32 has been described as an *E. coli* A2 enzyme the *N*-terminal amino acid sequence of ECP 32 AKTSSAGVVIRDIF could not be identified in published *Escherichia coli* genomic sequences [8,18]. Therefore, the systematic position of the ECP 32-producing bacterial strain was reinvestigated. Using about 50 biochemical reactions of the Vitek-2 system (bioMeґrieux, Marcy l'Etoile, France) and partial sequencing of the 16S rRNA gene, the former *Escherichia coli* A2 strain was identified as *Serratia grimesii* A2 [18]. Then, the presumptive gene coding for the reference *S. grimesii* 30063 or *Serratia grimesii* A2 was cloned using published sequences of a similar protease protealysin identified in *S. proteamaculans* [19]. Amino acid sequences of two corresponding proteins from the *S. grimesii* A2 (former *E. coli* A2) and the reference strain of *S. grimesii* 30,063 were identical and contained the *N*-terminal 14 amino acids of protease ECP 32 as previously determined [8,18]. The same specific actin-hydrolyzing activity, characteristic of protease ECP 32 [9,10], was also revealed in bacterial extracts of the reference *S. grimesii* strain and *Escherichia coli* transformed by the presumptive gene encoding grimelysin (ECP32) in *S. grimesii* A2 [18]. Taken together, these data suggested that ECP 32 and grimelysin is the same enzyme named grimelysin [18], slightly different from protealysin [19]. While *S.grimesii* and *S. proteamaculans* belong to a cluster of bacteria within the *Serratia liquefaciens* group, a similar protease could be expected to be synthesized by other members of this group. Our preliminary data indicate that protease with the actinase activity is present in *Serratia marcescens*.

Along with the high similarity of the bacteria *S. grimesii* and *S. proteamaculans*, their proteases, grimelysin and protealysin, are highly homologous and differ by only 8 amino acid residues [18,19]. Moreover, similarly to the grimelysin-producing bacteria [20,21], *S. proteamaculans* 94 turned out to be one more bacterial strain where the actinase activity of metalloproteinase protealysin is coupled with bacterial invasion [21,22]. These results are consistent with the idea of the actin-specific metalloproteases being a factor that can be involved in bacterial invasion of eukaryotic cells. This paper describes the properties of grimelysin and protealysin in vitro and in vivo in the context of this idea. Our data indicate that the protease is a virulence factor and actin can be a target for the protease upon its translocation into the host cell.

#### **2. Basic Properties and Substrate Specificity of Grimelysin and Protealysin**

*Grimelysin (ECP 32)*, discovered, purified and initially characterized as protease ECP 32 [6–8], was later shown to be identical to grimelysin [18]. Therefore, the properties of the enzyme identified for ECP 32 could be applied to grimelysin. However, here we retain the name grimelysin (ECP 32) and ECP-cleaved actin to comply with the published data where the protease was named ECP 32. Grimelysin (ECP 32), purified from a bacterial extract using sequential chromatography steps, is a single 32 kDa polypeptide, whose *N*-terminal sequence was determined to be AKTSSAGVVIRDIFL [8]. The optimum of the protease activity was observed in the range of pH 7–8 when actin and melittin were used as substrates [8,23]. The proteolytic activity increased with increasing ionic strength: in 50–100 mM NaCl the activity of grimelysin (ECP 32) towards melittin was shown to be nearly twice higher than in a low ionic strength solution [23,24]. It was also enhanced in the presence of millimolar ATP concentrations, though hydrolysis of melittin was not accompanied by ATP hydrolysis at a rate comparable with the cleavage rate. This implies that protease grimelysin (ECP 32) is not an ATP-dependent enzyme [23], which is important for the experiments involving actin because actin contains ATP as a tightly-bound nucleotide. The protease activity is inhibited by EDTA, EGTA, o-phenanthroline and zincone, and the EDTA-inactivated enzyme can be reactivated by cobalt, nickel and zinc ions [2,3]. Based on these data, grimelysin (ECP 32) was classified as a neutral metalloproteinase (EC 3.4.24) [8].

Limited proteolysis of skeletal muscle actin between Gly-42 and Val-43 [10] was observed at enzyme: substrate mass ratios of 1:25 to 1:3000 [8]. Two more sites, between Ala-29 and Val-30 and between Ser-33 and Ile-34, were cleaved by ECP 32 in heat- or EDTA-inactivated actin, apparently due to conformational changes around residues 28–34 buried in intact actin [8]. Besides actin, only melittin [18,19], histone H5, bacterial DNA-binding protein HU and chaperone DnaK [25] were found to be protease substrates. In agreement with this high substrate specificity, ECP 32 did not hydrolyze tropomyosin, troponin, α-actinin, casein, histone H2B, ovalbumin, bovine serum IgG, bovine serum albumin, bovine pancreatic ribonuclease A, trypsin, human heat shock protein HSP70, chicken egg lysozyme [7], insulin [24], DNAse I [9,13], gelsolin [14] and profilin [26]. The amino acid residues recognized by grimelysin (ECP 32) in actin and melittin are hydrophobic. This specificity is characteristic for thermolysin-like metalloproteinases [27]. However, high specificity of the enzyme seems to be determined predominantly by conformation at the actin cleavage site rather than its primary structure.

*Grimelysin* was obtained as a recombinant protein. This has been achieved by cloning the putative gene encoding grimelysin in *S. grimesii* A2 and in the reference *S. grimesii* 30063 [18] using published protealysin sequences identified in *S. proteamaculans* [19]. Grimelysin shared all properties characteristic for ECP 32 including a molecular weight of 32 kDa, an *N*-terminal 14 amino acid sequence, optimum activity in the range of pH 7–8 and inhibition with o-phenanthroline and EGTA [18].

*Protealysin* is a neutral zink-containing metalloprotease of *Serratia proteamaculans*. The protealysin gene was cloned from a genomic library of *S. proteamaculans* strain 94 isolated from spoiled meat. This protein was expressed in *Escherichia coli* and purified as described earlier [19]. Similarly to other thermolysin-like proteases [27,28], protealysin is synthesized as a precursor containing a propeptide of about 50 amino acids that is removed during formation of mature active protein [29]. The propeptide is much shorter than the propeptides of the thermolysin-like proteases and has no significant structural similarity to the propeptides of most thermolysin-like proteases [30–32]. A similar propeptide of 50 amino acids was also detected in the primary structure of the recombinant grimelysin. According to SDS-electrophoresis, recombinant proteins with or without propeptide had an apparent molecular weight of 37 and 32 kDa, respectively [19].

The molecular weight of the active recombinant protealysin 32 kDa and the *N*-terminal amino acid sequence AKTSTGGEVI are identical to those of grimelysin [8,19]. The optimal pH for azocasein hydrolysis is 7, and protealysin is completely inhibited by *o*-phenanthroline [19], i.e., has the same properties as grimelysin [8,18]. Protealysin and grimelysin (ECP 32) are also similar in their unique property of being able to digest actin specifically [8–10,22,33].

#### **3. Specific Actinase Activity of Grimelysin (ECP 32) and Protealysin**

The ECP 32-like limited proteolytic activity towards actin appears in bacterial lysates of *S. grimesii* and *S. proteamaculans* only at the late stationary phase of bacterial growth [7,18,22]. The bacterial lysates as well as the purified grimelysin (ECP 32) or recombinant grimelysin and protealysin cleave actin at a single site giving rise to two fragments of 36 and 8 kDa (Figure 1A) [10,13,22]. The fragments remain associated in the presence of the tightly bound calcium or magnesium cation needed to maintain the native actin conformation and dissociate after removal of the tightly bound cation with EDTA [13]. The *N*-terminal sequence of the 36 kDa actin fragment produced by ECP 32 was determined as Val-Met-Val-Gly-Met [10]. The same sequence was determined for the 36 kDa *N*-terminal actin fragment produced by the lysates of recombinant bacteria expressing the grimelysin or protealysin gene [18,22]. According to the amino acid sequence of actin [34], this peptide corresponds to the cleavage site between Gly-42 and Val-43 [10], which is not attacked by any known proteases. However, in the case of protealysin, this cleavage pattern was observed only at the protealysin/actin mass ratio of 1:50 or lower. At a higher protealysin to actin ratio, the 36 kDa fragment was further cleaved yielding two closely situated bands with an apparent molecular weight of 33 kDa [22]. The *N*-terminal sequences of these fragments are Leu-Lys-Tyr-Pro-Ile-Glu and Ile-Leu-Thr-Leu-Lys-Tyr, corresponding to the cleavage of the bonds Thr66–Leu67 and Gly63–Ile64, respectively [22]. The difference in activity of the purified bacterial grimelysin and recombinant protealysin may be due to their origin, both in terms of different bacteria and different purification protocols. Therefore, comparison of the actinase activities of recombinant grimelysin and protealysin would be of interest.

**Figure 1.** The actinase activity of grimelysin (ECP 32). (**A**) Three-dimensional structure of skeletal muscle actin cleaved with grimelysin (ECP 32) at a single site, Gly42-Val43, within the DNase I-binding loop [35]. Inset shows SDS electrophoresis of actin (lane 1), with the actin 36 and 8 kDa fragments produced by the grimelysin (ECP32) cleavage; the upper band is actin not completely cleaved in this experiment (lane 2) and isolated grimelysin (ECP 32) (lane 3) [8,10]. (**B**) Polymerization of ECP-cleaved skeletal muscle actin compared to that of intact actin. (**C**) Invasion of human larynx carcinoma Hep-2 cells by *Serratia grimesii* 30063. Intracellular bacteria are marked with arrows. Scale bar, 1 μm. Arrows indicate intracellular bacteria Reproduced from [13,21,35] with permissions from Elsevier Licences 4820990651384 and 1032953-1, 2020 (**A**,**B**) and from Wiley Licence 4821290007049, 2020 (**C**).

#### **4. Specific Properties of Protease-Cleaved Actin**

Although actin cleaved with grimelysin (ECP 32) between Gly42 and Val43 preserves its native conformation [10,13], the cleaved actin completely loses its ability to polymerize in the presence of Ca2<sup>+</sup> [5,8] (Figure 1B). The ability of the cleaved actin to polymerize is partially restored upon substitution of the tightly bound Ca2<sup>+</sup> with Mg2<sup>+</sup> [13]. However, the degree of this polymerization is lower than that of intact actin, while the critical concentration of polymerization of ECP-Mg-actin and the exchange of subunits in the ECP-Mg-actin polymer increase 30 and 20–30 times, respectively (Figure 1B) [13]. This effect is due to a more open conformation of the cleaved actin monomers, which decreases the monomer nucleation step of cleaved actin polymerization or/and increases the actin filament dynamics, i.e., the dissociation/association kinetics of actin monomers at the filament ends [15]. The open conformation of the ECP-cleaved actin monomers is also preserved in the ECP-actin polymer [15]. Similar properties are inherent to actin cleaved with protealysin [36]. Thus, cleavage in globular actin of the only amino acid bond between Gly-42 and Val-43 causes local conformational changes that weaken the intermonomer contacts during actin polymerization and lead to enhanced polymer dynamics. As it turned out later, these properties correlated with the ability of the protease-producing bacteria to invade eukaryotic cells [20–22] (Figure 1C)

Similar instability arises if the protease cleaves the bond between Gly-42 and Val-43 in the subunits of the actin polymer. Incubation of filamentous actin (F-actin) with purified *S. proteamaculans* protealysin or with the lysates of the recombinant *Escherichia coli* producing protealysin showed that 20–40% of F-actin was digested at enzyme/actin mass ratios rising from 1:50 up to 1:5, respectively [33]. This process was accompanied by the increased steady-state ATPase activity (dynamics) of F-actin [33], which was also shown to be characteristic of the polymers forming from grimelysin (ECP 32)-cleaved actin [13]. The cleavage-produced increase in the dynamics of the modified filaments can be reversed with phalloidin [13], aluminium and sodium fluorides [22,36], as well as with actin-binding proteins gelsolin [14], myosin subfragment 1 [37] and tropomyosin [17], indicating that actin-binding proteins can restore integrity of actin filaments damaged by proteolysis.

This assumption was verified in vivo using microinjection of the purified protease grimelysin (ECP 32) preparation into *Amoeba proteus* cells, which revealed dynamics of the cytoskeleton structures by alterations in amoeba locomotion [38]. After injection of the protease solution, pseudopodia formation was ceased, and the cytoplasm motility slowed down and finally stopped [38]. Injected amoebae remained spread and immobile until the locomotion was slowly restored to the control level. No changes in the locomotion were observed when protease grimelysin (ECP 32) was injected in the presence of ECP 32-specific antibodies [38]. These results indicate that the protease-produced modifications of actin cytoskeleton can be reversible and might be used by bacteria to regulate their intracellular activity within eukaryotic and probably bacterial cells. Specifically, actin-like, bacterial proteins that form the bacterial cytoskeleton could be protease substrates.

#### **5. Actin-Like Proteins of Bacteria**

Many functions of bacterial cells are performed by filamentous structures similar to those of the cytoskeleton filaments of eukaryotic cells [39,40]. These structures are formed by globular proteins, which are regarded as actin homologues, actin-like proteins or bacterial actins [39] because three-dimensional structures of their monomers are similar to that of skeletal actin [39,40]. Also similarly to the eukaryotic globular actin, globular bacterial actin-like proteins assemble into two-stranded filaments consisting of two protofilaments coming together in various ways [39,40], although this sets them apart from the eukaryotic actin filaments assembled only in the form of two parallel strands. The most studied bacterial cytoskeleton proteins, MreB, ParM, MamK and Ftz, perform such functions as shape determination in rod-shaped bacteria, cell division, plasmid segregation and organelle positioning [39,40]. It would be tempting, therefore, to suggest that the actin-specific proteases, grimelysin and protealysin, could regulate these activities. However, all these bacterial proteins, including MreB, most closely related to actin [40–42] differ from eukaryotic actin due to the absence of the DNase I-binding loop inserted in eukaryotic actin subdomain 2 [41,42], where eukaryotic actin is cleaved by grimelysin or protealysin. Thus, in contrast to eukaryotic actin, bacterial actin-like proteins do not contain any specific cleavage site to be attacked by grimelysin or protealysin. This suggests that the specific actinase activity of grimelysin and protealysin against eukaryotic actin could be important if the protease is delivered by bacteria into eukaryotic cells.

#### **6. Invasive Activity of Bacteria** *Serratia grimesii* **and** *Serratia proteamaculans*

Capability of the actinase-producing bacteria to invade eukaryotic cells was first found upon incubation of human larynx carcinoma Hep-2 cells with bacteria *Serratia grimesii* A2 (*E. coli* A2) [20,43]. After 2 h of infection, bacteria were detected within the cells, mostly in vacuoles but also free in cytoplasm (Figure 1C). The invasion was accompanied by a change in the shape of the eukaryotic cells, disappearance of the actin stress fibers inside the cells and appearance of protrusions on the cell surface [43]. No changes were observed in the cells infected with the wild-type *Escherichia coli* CCM 5172 that do not produce actinases [43]. Correlation between synthesis of the specific actin hydrolyzing protease and the ability of bacteria to invade eukaryotic cells was also detected in *Shigella flexneri* mutant obtained by exposure of bacteria to furazolidone [44,45]. This treatment resulted in the formation of *Shigella flexneri* L forms whose revertants were not pathogenic but able to invade eukaryotic cells, while their extracts are capable of cleaving actin as the bacterial extracts containing grimelysin or protealysin (Figure 2) [44,45].

It was also shown that the transformed cells are more sensitive to invasion by actinase-producing bacteria than non-transformed or poorly transformed cells [20,46]. Specifically, the grimelysin (ECP 32)-producing bacteria invaded the transformed epithelial and fibroblasts cells A431, HeLa and 3T3-SV40, but they were not found in embryonic fibroblasts, primary human keratinocytes and in cells of poorly transformed 3T3 cell lines [20]. Later, a quantitative analysis of invasion confirmed rather a low susceptibility of 3T3 cells to *Serratia grimesii* invasion, compared to that of 3T3-SV40 cells [46]. The higher susceptibility of the immortal CaCo-2 and HeLa cells compared to their untransformed

counterparts was also shown for invasion of *Listeria monocytogenes* [47]. Moreover, transfection of resistant *Listeria monocytogenes* cells by SV40 large T antigen was shown to induce highly transformed continuous cell lines with a susceptibility to bacteria phenotypes [48]. This difference may be produced by the different set of cell surface receptors contacting with bacteria in the "normal" and transformed cells, as is shown, for example, for the G-protein-coupled receptors [49].

**Figure 2.** Apathogenic *Shigella flexneri* 5a2c mutant treated with furazolidone can invade eukaryotic cells. (**A**) Detection of the actinase activity in the lysates of *Shigella flexneri* 5a2c acquired upon the furazolidone treatment. Numbers indicate the time points of bacteria growth when their extracts were tested for the actinase activity. A, actin; 36, the 36 kDa actin fragment. (**B**) Incubation of human larynx carcinoma Hep-2 cells with the furazolidone-treated non-pathogenic *Shigella flexneri* 2a 4115. (**C**) Invasion of the pathogenic furazolidon-treated *Shigella flexneri* 5a2c into human larynx carcinoma Hep-2 cells [39]. Arrows indicate extracellular bacterium (**B**) and intracellular bacteria lying in vacuoles (**C**). Scale bar, 0.5 μm. Reproduced from [45] with permission from Springer Nature License 4823180887079, 2020.

To find out if grimelysin and protealysin are actively involved in the entry of *Serratia grimesii* and *Serratia proteamaculans* into host cells, human larynx carcinoma Hep-2 cells were infected with recombinant *Escherichia coli* expressing grimelysin or protealysin gene [21]. The results of this work showed that the extracts of the recombinant bacteria cleave actin at the only site corresponding to the cleavage with grimelysin or protealysin [21]. Recombinant *Escherichia coli* carrying the grimelysin or protealysin gene were found in the eukaryotic cells, both in vacuoles (Figure 3) and free in cytoplasm [21]. At the same time, no invasion was observed if the Hep-2 cells were incubated with the non-invasive *Escherichia coli*-carrying plasmids that did not contain the protease gene. Internalization of the non-invasive *Escherichia coli* was also not observed if protealysin was added to the culture medium [21]. These results showed the direct participation of grimelysin and protealysin in the invasion of the host cells by the protease-producing bacteria [21].

Using electron microscopy, two modes of *Serratia grimesii* invasion were revealed. In most cases, interaction of *Serratia grimesii* with eukaryotic HeLa M cells starts with formation of a tight contact between bacteria and the host cell, which is followed by bacteria internalization, apparently due to a specific interaction of bacterial adhesins with a specific cell surface receptor (Figure 4A,B) [50]. This process corresponds to the "zipper" mechanism of invasion involving activation of the signal system of the host cell followed by moderate cytoskeleton rearrangements [51–53]. However, in some cases, bacteria looked like they were trapped by filopodia (Figure 4C,D), probably induced by injected bacterial proteins triggering the bacterial uptake, as described in the "trigger mechanism" of invasion [54–57]. Recently, coexistence of both the trigger and zipper invasion mechanisms

was postulated for *Salmonella* invasion [58–61]. These data provide a frame for revealing bacterial and cellular factors involved in *Serratia grimesii* and *Serratia proteamaculans* invasion.

**Figure 3.** Actinase activity of recombinant *E. coli* transformed with the grimelysin or protealysin gene. (**A**,**B**) Detection of the actinase activity in the lysates of the recombinant *E. coli* expressing grimelysin or protealysin. Numbers indicate the time points of bacteria culturing when their extracts were tested for actinase activity [21]. (**C**,**D**) Invasion of human larynx carcinoma Hep-2 cells by recombinant *E. coli* expressing the grimelysin or protealysin gene, observed by electron (**C**) and confocal (**D**) microscopy [21]. Arrows indicate extracellular and internalized bacteria. By electron microscopy, bacteria are in vacuoles (**C**) that are not visible with confocal microscopy. Scale bars, 1 μm. (**D**) The samples were examined under a Leica TCS SL confocal scanning microscope using a dual argon ion (488 nm; green fluorescence) and helium/neon (543 nm; red fluorescence) laser system to visualize the FITC-stained bacteria and rhodamine phalloidin stained cytoskeleton, respectively. Scale bars, 1 μm (**C**) and 10 μm (**D**). Modified from [21] with permission from Wiley license 4827670496420, 2020.

**Figure 4.** Electron microscopic identification of the initial steps in the invasion of eukaryotic cells by *Serratia grimesii*. (**A**) Tight contact of bacteria with HeLa M cells. (**B**) Extensive contact of bacteria with the surface of the host cell followed by the entry of bacteria into the host cell, which corresponds to the zipper mechanism of invasion. (**C**,**D**) Appearance of host cell filopodia, whose fusing traps bacterium and places it inside the vacuole, which corresponds to the trigger mechanism. (**E**) Bacteria live in the vacuoles. (**F**) Bacteria leave the vacuole to live free in the host cell cytoplasm. Scale bars, 1 μm. (Modified from [50] with permission from Pleiades Publishing, Ltd. 117342, 2020.).

#### **7. Bacterial Virulence Factors Involved in** *Serratia* **Invasion**

Transformation of the non-invasive *Escherichia coli* with the protealysin gene led to the appearance in the bacteria of the invasive activity [21]. However, inactivation of this gene did not abolish the ability of *Serratia proteamaculans* to invade eukaryotic cells [62]. Moreover, invasive activity of *Serratia proteamaculans* was five times higher than that of *Serratia grimesii* [63]. This indicates involvement of other factors regulating invasion.

Bacteria of *Serratia* genus are known to be facultative pathogens able to induce nosocomial infections or infections in immunocompromised patients [64–66]. A most pathogenic strain of *Serratia* genus is *Serratia marcescens*, pathogenicity of which is determined by a secreted pore-forming toxin (hemolysin) ShlA [67,68] and extracellular proteases [69]. Probing of *Serratia grimesii* and *Serratia proteamaculans* for these factors revealed the presence and high activities of pore-forming hemolysin ShlA and extracellular metalloprotease serralysin in *Serratia proteamaculans.* At the same time, in *Serratia grimesii,* the activity of the toxin ShlA was not detected, and the serralysin activity of the bacterial growth medium was very low [63]. It was also shown that iron depletion strongly enhanced the invasive activity of *S. proteamaculans*, increasing activities of hemolysin ShlA and serralysin, but did not affect these properties of *Serratia grimesii* [63]. These results showed that, along with protealysin, the invasive activity of *S. proteamaculans* is also determined by hemolysin and serralysin [63]. At the same time, grimelysin remains the only known virulent factor of *S. grimesii*.

The first step in bacterial invasion is the contact of bacteria with the surface of eukaryotic cells, often performed by bacterial outer membrane protein OmpX [70–72]. Indeed, transformation of *Escherichia coli* by a plasmid carrying the OmpX gene of *Serratia proteamaculans* caused a 3-fold increase in the adhesion of bacteria to the surface of eukaryotic cells Hep G2 and DF-2, without producing any effect on *Escherichia coli* invasion [73]. On the other hand, our preliminary data indicate that OmpX is a substrate for protealysin capable of enhancing the invasion of *Serratia proteamaculans* carrying the inactivated protealysin gene [73]. These data indicate that along with its direct participation in bacterial adhesion, OmpX could be involved in bacterial invasion, while, in turn, protealysin could regulate bacterial adhesion.

#### **8. Cellular Factors Involved in** *Serratia* **Invasion**

The penetration of bacteria into eukaryotic cells includes the contact of bacteria with eukaryotic cells, activation of the cell signaling system, reorganization of the cytoskeleton leading to bacterial uptake and spreading of the invaded bacteria within and between the cells [55,74,75]. The efficiency of this process is determined not only by the activity of bacterial virulent factors, but also by the sensitivity of eukaryotic cells, which, in turn, correlates with the degree of their transformation [20,47] and depends on the cell environment [76,77]. These factors were employed to find out whether bacterial invasion is sensitive to the cell type and the presence of antioxidant *N*-acethylcysteine (NAC) in the culture medium. The results of these experiments confirmed different effects of NAC on 3T3, 3T3-SV40 and HeLa cells [78]. Incubation of HeLa cells with NAC increased penetration of grimelysin-producing bacteria by 1.5–2 times for wild-type *Serratia grimesii* and by 3–3.5 times for recombinant *Escherichia coli* expressing the grimelysin gene [79]. These effects did not correlate with the cytoskeleton rearrangements but might be due to the NAC-induced upregulation of cell surface receptors playing a primary role in cell adhesion and cell–cell junctions [79].

Specifically, one of the cell surface receptors whose expression is regulated by *N*-acethylcysteine is E-cadherin, a transmembrane receptor known to be upregulated by NAC [77] and one of the two cell surface proteins that mediate adhesion and internalization of *Listeria monocytogenes* within epithelial cells [80,81]. Binding of bacteria internalin InIA to E-cadherin triggers the c-Src kinase-mediated phosphorylation of E-cadherin, followed by ubiquitination and E-cadherin internalization via the clathrin-mediated pathway [80–82]. In line with these data, incubation of 3T3 and 3T3-SV40 cells with NAC led both to increased expression of E-cadherin and increased sensitivity of these cells to invasion [46]. Co-localization of *S. grimesii* with E-cadherin of the 3T3 and 3T3-SV40 cells was also shown [46]. In addition, inhibitory analysis with the ROCK and c-Src kinase inhibitors revealed a correlation between c-Src and ROCK protein kinase activities, expression of E-cadherin and invasive activity of *Serratia grimesii* [83]. These data allow us to suggest that E-cadherin is at least one of the receptors involved in *Serratia grimesii* invasion.

On the other hand, participation of OmpX in *Serratia proteamaculans* invasion indicates that other cell surface receptors, including integrins and the epidermal growth factor receptor (EGFR), could be involved in this process, as it has been recently shown for the invasion of *Salmonella* [59–61]. In line with these data, infection of epithelial M-HeLa cells by bacteria *Serratia proteamaculans* led to the changes in localization of EGFR and fibronectin receptors α5-, β1-integrins. Accumulation of α5-, β1-integrins on the cell surface was accompanied by intensive attachment of bacteria to the sites of α5-integrin localization [84], indicating involvement of the outer membrane protein OmpX–fibronectin–integrin in the penetration of *Serratia proteamaculans* into eukaryotic cells.

#### **9. Conclusions**

Though *Serratia* are facultative pathogens able to cause nosocomial infections or infections in immunocompromised patients, hospital infection by *S. grimesii* or *S. proteamaculans* is low [64–66]. Consistent with this, the capability of *S. grimesii* to invade eukaryotic cells is also rather low, with only about 10% of the cells being invaded either by the wild-type or recombinant bacteria when cultured cells are infected in vitro [21]. This implies that production of grimelysin or protealysin does not make the bacteria pathogenic but rather provides them with an opportunity to be rendered pathogenic under specific conditions. Therefore, *Serratia grimesii* and *Serratia proteamaculans* invasion combines specific bacterial mechanisms to contact eukaryotic cells and the specificity of the cells to be invaded.

To penetrate eukaryotic cells, invasive bacteria should initiate their uptake by activating the signal transduction system and cytoskeleton rearrangements [1–4]. Thereby, pathogenic bacteria use specialized secretion systems to deliver bacterial virulence factors directly into the cytoplasm of the host cell to manipulate the intracellular pathways, thus following the trigger invasion mechanism [53–57,85]. In contrast, facultative bacterial pathogens use specific surface proteins that interact with the host cell receptors, according to the zipper invasion mechanism. These interactions activate receptor-mediated cell signaling cascades and lead to bacterial uptake [2,4,60,61,86]. Recent data showed however that the virulence components transported via the type III secretion system of pathogenic bacteria can be also translocated into the eukaryotic cells by the bacterial outer membrane vesicles (OMVs) [87]. Importantly, OMVs can be produced by opportunistic bacteria as well as by pathogenic ones [88–92], thus smoothing the difference between pathogenic bacteria and facultative bacterial pathogens.

Penetration of opportunistic bacteria *Serratia grimesii* and *S. proteamaculans* into eukaryotic cells [21] correlates with the presence in these bacteria of homologous metalloproteases grimelysin and protealysin, respectively [18,22]. Protealysin was found in the cytoplasm of 3T3-SV40 cells infected with *S. proteamaculans* or recombinant *Escherichia coli* expressing the protealysin gene [33], suggesting that protease can be transported into the cytoplasm of eukaryotic cells. Consistent with this suggestion, our preliminary data showed that the outer membrane vesicles produced by *Serratia grimesii* were able to transfer protease grimelysin into the cytoplasm of eukaryotic cells. This transfer enhanced the penetration of bacteria into eukaryotic cells if these cells were incubated with the vesicles prior to infection [93,94]. At least a part of this effect could be produced by the involvement of the OMV-delivered protease in the cytoskeleton dynamics. We have previously shown that the cleavage of actin with grimelysin (ECP32) enhanced turnover of actin monomers within actin filaments [15,38]. This, in turn, could contribute to the rearrangements of the actin cytoskeleton leading to bacteria internalization. The invasive activity of *Serratia* can also be regulated by activities of the intracellular factors, being, for example, diminished upon transfection of eukaryotic cells with anti-RhoA siRNA [83]. The cells transfected with anti-RhoA siRNA exhibited cell rounding, disassembly of actin cytoskeleton and formation of protrusions at the cell periphery [83], indicating an active involvement of the cytoskeleton in *S. grimesii* uptake.

Thus, the transition of opportunistic bacteria to the status of pathogenic is associated both with the activity of bacterial virulence factors and the sensitivity of eukaryotic cells to bacteria. In particular, the sensitivity of eukaryotic cells to bacterial invasion correlates with the degree of their transformation [20,46,47], the effects of antioxidants [78,79] and the activity of cell signaling components like c-SRC and ROCK kinases [83]. Further work is needed to integrate these individual cellular factors and bacterial virulence factors into the *Serratia* invasion mechanism.

**Funding:** The publication was partially supported by the "Director's Foundation of the Institute of Cytology". **Conflicts of Interest:** The authors declare no conflicts of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Review* **Virulence Factors of Meningitis-Causing Bacteria: Enabling Brain Entry across the Blood–Brain Barrier**

#### **Rosanna Herold, Horst Schroten and Christian Schwerk \***

Department of Pediatrics, Pediatric Infectious Diseases, Medical Faculty Mannheim, Heidelberg University, 68167 Mannheim, Germany; rosanna.herold@medma.uni-heidelberg.de (R.H.); horst.schroten@umm.de (H.S.)

**\*** Correspondence: christian.schwerk@medma.uni-heidelberg.de; Tel.: +49-621-383-3466

Received: 26 September 2019; Accepted: 25 October 2019; Published: 29 October 2019

**Abstract:** Infections of the central nervous system (CNS) are still a major cause of morbidity and mortality worldwide. Traversal of the barriers protecting the brain by pathogens is a prerequisite for the development of meningitis. Bacteria have developed a variety of different strategies to cross these barriers and reach the CNS. To this end, they use a variety of different virulence factors that enable them to attach to and traverse these barriers. These virulence factors mediate adhesion to and invasion into host cells, intracellular survival, induction of host cell signaling and inflammatory response, and affect barrier function. While some of these mechanisms differ, others are shared by multiple pathogens. Further understanding of these processes, with special emphasis on the difference between the blood–brain barrier and the blood–cerebrospinal fluid barrier, as well as virulence factors used by the pathogens, is still needed.

**Keywords:** bacteria; blood–brain barrier; blood–cerebrospinal fluid barrier; meningitis; virulence factor

#### **1. Introduction**

Bacterial meningitis, as are bacterial encephalitis and meningoencephalitis, is an inflammatory disease of the central nervous system (CNS). It can be diagnosed by the presence of bacteria in the CNS. Despite advances in treatment, it is still a disease of global dimension, which can end fatally or leave long-term neurological sequelae in survivors [1].

The brain is well protected from invading pathogens by cellular barriers, with the two major barriers being the blood–brain barrier (BBB) and the blood–cerebrospinal fluid barrier (BCSFB) [2,3]. To reach the CNS, blood-borne bacteria must interact with and cross these barriers of the brain. During the course of infection attachment to the host cells is initiated, followed by the hijacking of several host cell pathways by the pathogens. This process can be used by the bacteria to enter host cells with subsequent intracellular survival and, for some pathogens, multiplication. Finally, after crossing into the brain, the bacteria elicit an immune response from the cells within the CNS that might contribute to the inflammatory events leading to disease, with potential disruption of barrier integrity [4].

The traversal or breach of these barriers by meningitis-causing pathogens is defined by a complex interplay between host cells and the pathogens, which use an array of virulence factors to facilitate this interaction, resulting in high morbidity and mortality [1]. These virulence factors are involved during protection in the bloodstream, such as the capsule of both gram-positive and gram-negative bacteria, mediate adhesion to and invasion into host cells, and are responsible for intracellular survival.

#### **2. Barriers of the Central Nervous System**

The barriers protecting the brain are essential for its function and the stability of its internal microenvironment [5]. Important morphological components found in most of these barriers are specialized intercellular tight junctions between the cells, which are the basis for the barrier property. Various transporters, such as ATP-binding cassette (ABC) transporters, as well as efflux pumps control the movement of molecules across these interfaces, making them another component of barrier function [6]. These mechanisms prevent substances such as toxins, pharmacologically active agents and pathogens from accessing the CNS [4,7,8].

#### *2.1. Blood–Brain Barrier*

The BBB is formed by brain microvascular endothelial cells (BMECs), astrocytes and pericytes. It presents a structural and functional barrier and acts as an interface between the CNS and the peripheral circulation [2,9]. It has an additional protective function, restricting the free movement of substances across and inhibiting the entry of pathogens and toxins into the CNS. By regulating the passage of molecules, it is furthermore responsible for maintaining the CNS homeostasis.

The structure of the BBB is defined by BMECs covering the inner surface of the capillaries by forming a continuous sheet of cells interconnected by tight junctions (TJs) [10]. Specific TJ proteins, including occludin and, at the BBB, claudin-5, and claudin-12, but not claudin-3, are responsible for limiting the paracellular diffusion of ion and solutes across the barrier [11,12]. BMECs are characterized by less cytoplasmic vesicles, a larger number of mitochondria and a high amount of intercellular junctions, resulting in a high transendothelial resistance and low paracellular flux [13]. Transporter systems of the BBB include solute carrier-mediated transport, receptor-mediated transport, as well as active efflux and ion transport [14]. The BMECs are supported in shaping the BBB by associated cells such as astrocytes and pericytes [10]. The mechanisms of microbial CNS invasion have been elucidated by the use of in vivo and in vitro studies. To model infections of the CNS, mice and rats are most commonly used. To model the BBB in in vitro cell culture systems, primary human, rodent, and bovine brain microvascular endothelial cells have been used. Extensive studies of the human BBB have been facilitated by the availability of immortalized brain endothelial cells such as human brain microvascular endothelial cells (HBMECs) and hCMEC/D3 [1].

#### *2.2. Blood–Cerebrospinal Fluid Barrier*

Another major barrier protecting the brain is the BCSFB, which can be separated into a barrier to the inner CSF in the ventricles at the choroid plexus (CP) and barriers to the outer CSF located at the arachnoidea and blood vessels present in the subarachnoidal space [15]. Importantly, the choroidal BCSFB is essential in the trafficking of immune cells into the CNS [3,16]. The epithelial cells displaying polarity and forming tight junctions, which specifically express claudin-1, -2, -3, and -11, are the morphological correlates of the BCSFB at the CP [3,11,12]. The presence of specific transporter systems, such as ABC transporters, and low pinocytotic activity regulating the crossing of substances are key characteristics of the CP [17,18]. Furthermore, the BCSFB is crucial for the protection of the CNS from pathogens [19].

To cause diseases of the CNS bacterial pathogens need to overcome these barriers [1,17]. During this process, several different steps are influenced by bacterial virulence factors. To analyze the role of the CP during infectious diseases of the CNS, animal models can be utilized. Important tools for studying the BCSFB in vitro are primary models of CP epithelial cells, such as primary porcine CP epithelial cells (PCPEC). Another functional model of the BCSFB is a human CP epithelial papilloma cell line (HIBCPP cells) which displays apical/basolateral polarity and a good barrier function [17].

#### **3. Stages during the Pathogenesis of BACTERIAL Meningitis**

Bacterial meningitis can be caused by a variety of different gram-positive and gram-negative pathogens. There are mechanisms and steps of disease progression that are similar among them as shown in Figure 1. Colonization of the mucosal surfaces is often the first step of disease progression, followed by different strategies of dissemination into the bloodstream [4]. Once the pathogens have entered the bloodstream and successfully counteracted innate and adaptive immune defenses, they have to translocate across the barriers of the brain to invade the CNS and cause severe inflammation. Attachment to the brain endothelium or the epithelial cells of the choroid plexus and subsequent invasion mark the initial step of translocation into the CNS [4].

**Figure 1.** CNS entry pathways and stages during the pathogenesis of bacterial meningitis. Bacterial pathogens can cross the BBB or BCSFB paracellularly between neighboring cells, in a "Trojan horse" fashion inside infected host macrophages, or transcellularly by invading epithelial or endothelial cells. Cellular entry can be launched by the zipper mechanism involving binding to host cell receptors or by the trigger mechanism (see main text for details). During this process, activation of signal transduction pathways can cause initiation of actin rearrangements. Activation of signal transduction can also be triggered during the transcellular pathway or the "Trojan horse" mechanism, but is not indicated in the figure for reasons of clarity. Once in the cytoplasm, the pathogens need to survive inside the cells for further disease progress. Activation of host genes causes an inflammatory response that can lead to disruption of tight junctions.

#### *3.1. Attachment and Invasion*

To facilitate adhesion and invasion of the barriers protecting the brain, a threshold level of bacteremia has been shown to be required [20–22], which is correlated with the severity of infection and likeliness of developing meningitis [23]. However, direct invasion from neighboring infected tissues can occur as well.

Bacterial adhesion to host cell surfaces is a complex process. It involves multiple adhesion molecules of the pathogen interacting with a variety of target receptors. These interactions, which can involve several adhesins of one microbe, can occur in a sequential manner. Hereby, the initial interactions can trigger the expression of further host receptors, which are then targeted by other bacterial adhesins [24].

Many pathogens have been shown to bind to extracellular matrix proteins to facilitate initial attachment to the host cells. Furthermore, binding of bacterial adhesins to specific host cell receptors can in turn induce different signal transduction pathways, resulting in tight attachment or internalization of the bacteria into the host cells [4]. The use of pili or fibrils for invasion of HBMECs was observed for a multitude of meningitis-causing pathogens [8], making them highly important virulence factors for invasion of the CNS.

Endocytosis of pathogens into non-phagocytotic cells is initiated by one of two mechanisms: the "zipper" and the "trigger mechanism". Some pathogens express surface proteins capable of interacting with transmembrane receptors on the hosts cells which are connected to the cytoskeleton. The "zipper mechanism", in particular, is defined by an interaction of a bacterial ligand and a host-specific membrane receptor initiating signaling events that lead to internalization of the pathogen through endocytosis [25]. The "trigger mechanism" is a micropinocytosis-related process that involves formation of actin-rich membrane ruffles. These are formed by localized changes in actin dynamics and membrane remodeling triggered by delivery of active effectors, which can be injected by the needle-like structure of a type three secretion system (T3SS), into the cytosol of host cells initiating signaling cascades [26,27].

#### 3.1.1. CNS Entry Routes

Meningitis-causing pathogens most commonly cross host barriers in a transcellular or paracellular manner [2]. These processes are associated with protein interactions between pathogens and the host's cells. Transcellular traversal is characterized by pathogens crossing the barrier cells without evidence of TJ disruption or traversal between cells [2]. This is accomplished by intracellular invasion of the barrier cells and exploitation of signaling pathways. Paracellular traversal, on the other hand, involves penetration of pathogens between the host's cells and can occur with and without permanent disruption of TJs [2,4]. Furthermore, the release of bacterial toxins can lead to disruption of barrier function and promote paracellular traversal. Another means of entry is the "Trojan-horse" mechanism, which describes penetration of the barrier by transmigration within infected phagocytes [2]. It has been suggested that the infected phagocytes adhere to the luminal side of brain capillaries. This can occur with and without the activation of BMECs and is followed by either transcellular or paracellular traversal of the BBB [28].

#### 3.1.2. Signal-Transduction Mechanisms and Cytoskeletal Rearrangements

Meningitis-causing bacterial pathogens have been shown to utilize host cell signaling molecules to facilitate infection. The mechanisms deployed can vary between the different pathogens, as well as the host tissues.

To both enter and leave the host cells, several pathogens have developed mechanisms to use the actin polymerization machinery of the host [29,30]. Pathogens not only use this process to spread throughout the cells, but have also developed mechanisms to subvert these regulatory mechanisms [31]. Furthermore, meningitis-causing pathogens use different signal-transduction mechanisms that result in rearrangements of the actin cytoskeleton [2]. This allows the pathogens to initiate attachment and entry of host cells, movement within and among cells as well as vacuole formation. This complex process of actin cytoskeletal remodeling can involve many factors, such as Rho-family GTPases and a variety of actin-binding proteins [30,32].

The innate immune system of the host can be triggered by various molecules which are characteristic for the bacteria. These pathogen-associated molecular patterns (PAMPs) are then recognized by eukaryotic pattern recognition receptors (PRRs), which in turn induce signaling cascades such as the nuclear factor κB (NF-κB) and mitogen-activated protein kinase (MAPK) pathways [33,34]. Activation of these signaling cascades triggers pro-inflammatory responses like the up-regulation of cytokines [35].

#### *3.2. Intracellular Survival*

#### 3.2.1. Multiplication and Intracellular Survival

Extracellular multiplication is the most common process for propagation by pathogenic bacteria. Furthermore, replication and persistence inside host cells has been demonstrated for a variety of meningitis-causing pathogens as well [36]. Entry into the host´s cells and replication within these protects the pathogens from clearance by the complement system and circulating antibodies. However, these pathogens have to overcome several cellular defense mechanisms such as the upregulation and secretion of neutrophil-specific factors in HBMECs. This response by the BBB is assumed to serve to

recognize pathogens resulting in their clearance. However, overactivation of the cellular response through continued exposure to the pathogens could result in increased inflammation and compromised barrier integrity [8].

#### 3.2.2. Disruption of Barrier Integrity and Inflammatory Response

Characteristically, bacterial meningitis is accompanied by a severe inflammatory response leading to neuronal damage. Activation of the transcription factor NF-κB upon invasion of the brain barrier tissues results in high levels of inflammatory cytokines in the blood and CSF [37]. This proinflammatory response can be triggered by bacterial cell wall components. Examples would include the lipopolysaccharide (LPS) for gram-negative bacteria, and lipoteichoic acid (LTA) for gram-positive bacteria. Furthermore, increased permeability of the barriers can be triggered by both bacterial toxins, as well as the initiation of host inflammatory pathways in response to the infection [38]. Tissue damage during bacterial meningitis arises from the initiated inflammatory cascade involving cytokines and chemokines, as well as proteolytic enzymes and oxidants [4]. Consequences of the release of these inflammatory substances, caused by multiplication of pathogens in the CNS, are damage of neurons and edema [39]. The hosts' immune response is therefore unable to embank infection of the CNS, and may even contribute to adverse events during bacterial meningitis [4].

The release of proinflammatory molecules not only increases permeability of the BBB but also attracts leukocytes to the CNS [40]. Cell death can be caused by cytokines, reactive oxygen species, reactive nitrogen species and matrix metalloproteinases [41–45].

#### **4. Roles of Bacterial Virulence Factors During Invasion Through the Barriers of the CNS**

To be able to enter the CNS, bacterial pathogens use several virulence factors, which are involved in the different steps of pathogenesis. Of major importance is the capsule of both gram-positive and gram-negative bacteria. The capsular polysaccharide has a protective function in bloodstream survival [46]. However, it was observed to attenuate invasion of the BBB and BCSFB [47–49]. This could result from electrostatic repulsion or from the masking of bacterial surface structures [8]. The necessity of the capsule for the pathogens survival in the blood but simultaneous hindrance of invasion of the host's tissues indicates a need for the regulation of capsule expression [50].

Besides the capsule, a multitude of further virulence factors are used by gram-positive and gram-negative bacteria [51]. These include, among others, adhesins and internalins, pore-forming toxins, and factors involved in intracellular movement as well as cell-to cell spread and are summarized in Figure 2.

**Figure 2.** Multiple virulence factors are involved in the different steps of pathogenesis during bacterial meningitis. Expression of a capsule can support bloodstream survival of both gram-positive and gram-negative bacteria. It has been described that down-regulation of capsule expression occurs during adhesion to and invasion into host cells, which is mediated by adhesins, internalins, pili, and pore-forming toxins. Pore-forming toxins can also be involved during escape from vacuoles inside of host cells, as well as intracellular survival. These steps are further supported by pore-forming toxins and the capsule. Several virulence factors, including internalins and pore-forming toxins, activate host cell signal transduction and mediate gene activation causing an inflammatory response.

In this review, we will focus particularly on virulence factors, for which evidence for an involvement during brain entry across the blood–brain barriers has been proposed. A summary of these virulence factors for gram-positive and gram-negative bacteria is given in Table 1 at the end of this section.

#### *4.1. Gram-Positive Bacteria*

#### 4.1.1. *Listeria monocytogenes*

*Listeria monocytogenes* (*L. monocytogenes*) is a facultative intracellular gram-positive bacterium. It can traverse several physiological barriers and finally enter the brain via the BBB or BCSFB, especially in immunocompromised individuals [52]. It is ingested through highly contaminated food by the host. Once ingested, *L. monocytogenes* traverses the intestinal epithelial barrier into the lamina propria followed by dissemination of the pathogen via the lymph and blood [53]. *L. monocytogenes* has multiple target organs, including the liver and spleen, and can enter the CNS across the barriers of the brain [53]. In addition to direct traversal of the BBB and BCSFB via the transcellular route, transportation across the BBB within leukocytes and retrograde migration within axons of cranial nerves have been described [54,55].

*L. monocytogenes* can enter non-phagocytotic cells by hijacking the host's receptor-mediated endocytosis machinery using the zipper mechanism. The two major invasion proteins of *L. monocytogenes* are internalin (InlA) and InlB, which bind to eukaryotic cell membrane members E-cadherin and tyrosine kinase receptor protein Met, respectively. These interactions induce receptor-mediated endocytosis of the pathogen. *L. monocytogenes* has been demonstrated to use one or both internalins to mediate invasion of the BBB and BCSFB [25,56,57]. A recent study has further demonstrated the importance of the bacterial surface protein InlF, showing that interaction with surface vimentin was required for an optimal colonialization of the brain [58].

The MAPK signaling cascade is activated during the invasion of *L. monocytogenes* [35,59,60]. In a model system of the BCSFB consisting of choroid plexus epithelial cells, the requirement of MAPK activation for listerial entry was demonstrated. Both extracellular signal-regulated kinases (ERK) 1 and 2 and p38 inhibition resulted in decreased bacterial invasion into this model system suggesting their involvement in the pathogens traversal of the BCSFB [34].

It was previously described that ubiquitination of E-cadherin and Met leads to the recruitment of the clathrin-mediated endocytosis machinery. This in turn results in the polymerization of the actin cytoskeleton. During this process, dynamin recruits several factors that result in two waves of actin rearrangements and subsequently result in the entry of the pathogen inside of vacuoles [61–63]. Accordingly, an in vitro study using a model of the BCSFB based on HIBCPP cells, revealed that *L. monocytogenes* invasion is inhibited if dynamin-mediated endocytosis is blocked [34].

Another essential virulence factor of *L. monocytogenes* is the pore-forming cytolysin Listeriolysin O (LLO). Activation of the NF-κB signaling pathway by LLO was reported in the human embryonic kidney HEK-293 cell line [64], as well as MAPK signaling [65,66]. It is secreted by *L. monocytogenes* and promotes the pathogens intracellular survival. After entering the host cell, lysis of the vacuole is initiated through LLO and the bacterial phospholipases PlcA and PlcB, and followed by intracellular spread in the cytoplasm [61].

Once *L. monocytogenes* has reached the cytoplasm of the host´s cells, it has been demonstrated to move around and enter neighboring cells using actin comet tails and membrane protrusions to facilitate its spread [61,67]. This F-actin-based intracellular motility is dependent on the expression of another essential listerial virulence factor, ActA [68].

Activation of the NF-κB signaling pathway is, as previously described, achieved through LLO. Another mechanism involving NF-κB is its activation by InlC, which is secreted intracellularly. It can directly interact with the subunit of the IκB kinase complex, IKKα. By phosphorylating IκB, this complex is critical for the activation of NF-κB, a major regulator of innate immune response. InlC was shown to impair phosphorylation of IκB, thereby scaling down the hosts immune response [69], and is also involved in cell-to-cell spread [70].

#### 4.1.2. *Streptococcus suis*

*Streptococcus suis* (*S. suis*) is a zoonotic gram-positive bacterium and one of the most important porcine bacterial pathogens. Serotype 2 of *S. suis* has been described to be a major cause of meningitis, especially in South and East Asia [71]. To reach the CNS, *S. suis* has to colonize the host and traverse epithelial barriers in order to reach the bloodstream, where it needs to survive. *S. suis* has been demonstrated to cross the BBB and the BCFSB in human in vitro models as well as in porcine models [48,71–73].

The presence of a capsule is essential for survival in the bloodstream. However, it was demonstrated to attenuate invasion for *S. suis* in epithelial cells [48,72,74]. A link between capsule expression and carbohydrate metabolism has been described, indicating adaptation of *S. suis* to different environments. High concentrations of nutrients, as found in the bloodstream, coincided with high expression of the capsule, whereas in the CNS, which is low in nutrients, expression was reduced [50,75]. Attachment of *S. suis* to BMECs has been demonstrated in human and porcine in vitro models of the BBB [76,77]. Invasion has been reported in porcine models but at very low rates [73,78]. During the adhesion process, in these in vitro model systems, the capsule had no effect on adherence [76]. In contrast, in porcine and human models of the BCSFB, both attachment and significant invasion of *S. suis* strains were demonstrated. The use of unencapsulated mutants further increased invasion rates, indicating a role of the capsule and regulation of its expression [48,72]. Other important virulence factors for invasion of the barriers of the brain are cell wall components such as lipoteichoic acid, LPXTG-anchored proteins as well as lipoproteins [79].

Also, enolase has been identified as a virulence factor of *S. suis.* Previously thought to act only as a glycolytic enzyme, this protein, with a highly conserved sequence, has been implicated in the invasion process of various pathogens. For *S. suis*, enolase has been shown to increase BBB permeability as well as promoting the release of interleukin IL-8 [80]. Transmigration might further be promoted by the thiol-activated cytolysin suilysin, which induces pore formation in membranes. Furthermore, the use of bacterial mutants lacking suilysin has shown that it is not essential for invasion of the host [81]. It has, however, been demonstrated to promote association with epithelial cells, making it another major virulence factor of the pathogen [82].

The upregulation of different cytokines and chemokines in response to *S. suis* infection in BMECs has been reported. Exemplary is the induction of IL-6 and IL-8, stimulated by *S. suis* [83]. An inflammatory response was also described in a porcine model of the BCSFB. Here, induction of tumor necrosis factor (TNF) α and matrix metalloproteinase (MMP)-3 gene expression were described. This was paralleled by rearrangements of the tight junction proteins ZO-1, occludin and claudin-1, and loss of actin at the apical cell pole as well as the induction of stress fiber formation at the basolateral side of the barrier [84]. The expression of TNFα after stimulation by *S. suis* in porcine choroid plexus epithelial cells further promoted adhesion and transmigration of polymorphonuclear neutrophils (PMN) through the barrier, which is a critical step during bacterial meningitis. Interestingly, some PMNs contained internalized *S. suis*, indicating the possibility of the pathogen exploiting the Trojan-horse mechanism [85].

#### 4.1.3. *Streptococcus pneumoniae*

*Streptococcus pneumoniae* (*S. pneumoniae*) is a gram-positive pathogen which is the major cause of bacterial meningitis in the developing world [86]. Close to 30% of individuals carry *S. pneumoniae* asymptomatically. Nasopharyngeal colonization is followed by systemic invasion and access to the bloodstream. Invasion of the CNS via the barriers of the brain is the major cause for meningitis [87]. Furthermore, olfactory neuron invasion was observed to be an entry route for *S. pneumoniae* [88]. To facilitate invasion of the CNS, *S. pneumoniae* utilizes several virulence factors such as the pneumococcal capsule and surface proteins as well as secreted proteins.

Bacterial interactions with the host and the subsequent development of bacterial meningitis are promoted by a high level of bacteremia [4]. Accordingly, attachment of *S. pneumoniae* to the choroid plexus in an in vivo mouse model was observed only during late stages of infection with high levels of bacteremia [87]. Further studies of the interaction with the BCSFB would be needed to determine if *S. pneumoniae* can use it as entry gate to the CNS.

The capacity of *S. pneumoniae* to invade host tissues is majorly determined by its capsule. Similar to other pathogens capable of causing bacterial meningitis, survival in the bloodstream is dependent on maximum capsule expression [89,90], whereas attachment to host tissues is hindered by its presence [91]. As the binding of various pneumococcal surface proteins is hindered by the capsule, altered expression of the capsule through quorum sensing and phase variation has been described [92,93].

Pneumococci have been demonstrated to use multiple virulence factors to initiate attachment to HBMECs in several in vitro studies. The platelet endothelial cell adhesion molecule (PECAM-1) has been implicated as receptor for neuraminidase A (Nad A) in pneumococcal attachment [94,95]. Importantly, NadA has been described as an important virulence factor, anchored in the cell wall of *S. pneumoniae*that can cleave sialic acid of the host substrates. It has been implicated in triggering the transforming growth factor-β (TGF-β) signaling cascade during the interaction with the BBB, resulting in decreased barrier integrity and an increase in invasion [96]. Another important receptor involved in pneumococcal BBB invasion is the polymeric immunoglobulin receptor (pIgR) as the major adhesin of the pneumococcal pilus-1, RrgA, was observed to bind to it as well as PECAM-1. Human platelet-activating factor receptor (PAFR) was demonstrated to be further involved in attachment, together with the before mentioned PECAM-1 and pIgR [97]. Additionally, a second type of pneumococcal pili has been described to mediate adhesion to host cells [98]. Furthermore, choline binding protein PspC was shown to bind only to pIgR [97], and interaction with the laminin receptor by *S. pneumoniae* is initiated by choline-binding protein A (CbpA) [99]. A proteome-based approach in a mouse meningitis model, addressing adaptive capabilities of the pathogens to a defined host compartment, has highlighted a crucial role for two highly expressed pneumococcal proteins; ComDE, a regulatory two-component system, and AliB, a substrate-binding protein of an oligopeptide transporter, in pneumococcal meningitis [100].

A major virulence factor of *S. pneumoniae* is pneumolysin, a pore-forming toxin which acts in a cholesterol-dependent manner. Pneumolysin has recently been demonstrated to induce the expression of CERB-binding protein (CBP), a coactivator of transcription. This results in the release of TNF-α and IL-6, which in turn lead to increased permeability of the BBB through increased apoptosis of the cells both in vivo and in vitro [101]. It has furthermore been implicated in paracellular traversal of the BBB by *S. pneumoniae* as a result of reduced barrier integrity [102].

In addition to the disruption of the BBB by pneumolysin, generation of H2O2 through α-glycerophosphate oxidase (GlpO) was observed to have a cytotoxic effect on HBMECs [103]. A study using rat brain tissues investigated the effect of *S. pneumoniae* infection on nucleotide-binding oligomerization domain 2 (NOD2) and inflammatory factors, suggesting that NOD2 may hold a role in the activation of inflammatory pathways and subsequent BBB damage [104].

A study by Coutinho et al. analyzed the CSF of patients with pneumococcal meningitis, detecting high levels of pro-inflammatory cytokines such as TNF-α, IL-1β, and IL-6 and anti-inflammatory cytokines IL-10 and TGF-β. Furthermore, chemokines IL-8, MIP-1a and MCP-1 were detected [45]. Furthermore, an in vivo study in neonatal Wistar rats demonstrated an increase in cytokines prior to BBB breakdown after induction of pneumococcal meningitis [105]. Entry of the CNS by the pneumococci is followed by rapid multiplication. Components released by the pathogen during this process are then recognized by PPRs resulting in a strong inflammatory response and subsequent BBB impairment [4].

#### 4.1.4. *Group B streptococcus*

*Group B streptococcus* (GBS, *Streptococcus agalactiae*) is a β-haemolytic, gram-positive pathogen and the leading cause of meningitis in human neonates [106]. Classification of GBS strains is done by sequence type based on an allelic profile of seven loci [107]. The development of GBS meningitis is dependent on bloodstream survival and development of a high level of bacteremia of the pathogen, followed by the disruption of the BBB or possibly the BCSFB. This is followed by the multiplication of GBS in the CNS, culminating in severe inflammation and neural damage [4]. The necessity of high-level bacteremia indicates that bloodstream survival is an important virulence factor of the pathogen along with the sialylated GBS capsular polysaccharide [108]. Additionally, reduction of capsule expression by GBS was demonstrated to increase virulence and intracellular persistence [49]

Direct interaction of GBS with the BBB endothelium has been demonstrated, resulting in traversal of the barrier and subsequent infection of the CNS [49,109]. Both direct invasion of the BBB and/or brain invasion as a direct result of increased permeability of the barrier have been observed [4]. The use of in vitro models has further shown transcellular crossing of the pathogen [109]. To this end, a variety of virulence factors have been described. One of these factors is a surface anchored novel protein specific for the GBS ST-17 clone, which is associated with meningitis in infants after the first week of life, called hypervirulent GBS adhesin (HvgA) and is required for GBS hypervirulence. Increased adherence of strains expressing HvgA was detected for intestinal epithelial cells, choroid plexus epithelial cells, and microvascular endothelial cells of the BBB [110]. Another necessary determinant of the GBS interaction with the BBB is the expression of cell-wall anchored pili [111]. Interestingly, these pili displayed similar function in adhesion and invasion of GBS, given that the role of pili has been best described for gram-negative bacteria. Two proteins involved in the formation of the pili, encoded by *pilA* and *pilB,* were shown to facilitate the interaction with the BBB, wherein PilA is promoting attachment of GBS and PilB is mediating internalization of the bacterium [112]. Interaction of PilA with collagen promotes its interaction with the α2β<sup>1</sup> integrin, initiating the integrin signaling machinery [38]. Interaction

with HBMECs is further enabled by the GBS fibronectin-binding protein streptococcal fibronectin binding protein A (SfbA) [113]. Additionally, GBS serine-rich repeat (Srr) glycoprotein was suggested to both promote bloodstream survival, facilitated by the adherence to fibrinogen, as well as adhesion to HBMECs [114]. An antigen I/II family adhesin, BspC, was recently demonstrated to interact with host cell vimentin during the pathogenesis of GBS meningitis, thereby promoting adherence of the pathogen in vitro as well as contributing to the development of GBS meningitis in vivo [115].

During invasion of type III GBS in HBMECs, tyrosine phosphorylation of the focal adhesion kinase (FAK) was demonstrated. Not only was the phosphorylation of the FAK required for invasion of GBS, it further induced association with PI3-kinase and paxillin, an actin filament adaptor protein [116]. BBB penetration of GBS was shown to also involve the *invasion associated gene A* (*iagA*). Mice challenged with a mutant version of this gene, encoding a glycosyltransferase homolog, developed bacteremia comparable to the wild type (WT) mice but had significantly lower mortality. In addition, the *IagA* gene encodes an enzyme, the glycolipid diglucosyldiacylglycerol, which functions as a cell membrane anchor for LTA, indicating that proper LTA anchoring is necessary for invasion of GBS into the BBB [117].

Similarly to *S. pneumoniae*, GBS can secrete a pore-forming toxin to disrupt barrier function in infected BMECs [109]. In addition, a further consequence of the interaction of PilA with α2β<sup>1</sup> integrin is the activation of host chemokine expression as well as neutrophil recruitment, which was correlated with increased permeability of the BBB [38]. Furthermore, GBS hyaluronidase HylB was demonstrated to induce BBB opening in a dose-dependent manner [118]. Another factor influencing BBB dysfunction was described during infection of induced pluripotent stem cell-derived brain endothelial cells with GBS, which resulted in the inhibition of P-glycoprotein, an important efflux transporter for the maintenance of brain homeostasis [119]. Barrier disruption and subsequent bacterial passage was further demonstrated to be promoted by reduced expression of tight junction components ZO-1, Claudin-5 and Occludin in HBMECs. This was facilitated by the induction of transcriptional repressor Snail1 and sufficient to promote tight junction disruption. This process, which was shown to be dependent on ERK 1/2 MAPK signaling as well as bacterial cell wall components, marks another mechanism of BBB disruption by GBS [120].

#### *4.2. Gram-Negative Bacteria*

#### 4.2.1. *Escherichia coli*

*E. coli,* a gram-negative bacillary organism, is a common cause of meningitis and still an important cause of mortality and morbidity throughout the world. Circulating *E. coli* have been shown to traverse the BBB and the BCSFB as a result of hematogenous spread [22,121]. Expression of the K1 capsule and *O*-lipopolysaccharide are critical determinants of *E. coli* meningitis, especially in neonates [122–124]. Several factors have been demonstrated to influence *E. coli* invasion of the CNS such as a high level of bacteremia for the invasion of the blood–brain barrier, as well as a variety of virulence factors that initiate binding and translocation into the CNS [2].

The first step in *E. coli* invasion of the CNS is attachment to the cells of the brain barriers. The two major virulence factors associated with attachment to the blood–brain barrier are type 1 fimbriae and OmpA [125,126]. The virulence factor IbeA has been associated with the subsequent invasion process [127], as well as cytotoxic necrotizing factor 1 (CNF1) [128]. Deletion of *ompA* and *ibeA* reduced infection of choroid plexus epithelial cells in a human model of the BCSFB, whereas deletion of *fimH* enhanced invasion but simultaneously decreased adhesion of *E. coli* strains in the same model [121]. In an in vitro model of the BBB, Ecgp was identified as receptor for OmpA [129].

To promote internalization, *E. coli* induces rearrangements of the actin cytoskeleton. In HBMECs, this has been demonstrated to trigger a zipper-like mechanism that envelops the bacterium and initiates internalization into the cell. This process is dependent on both the actin cytoskeleton as well as microtubules [4]. Induction of tyrosine phosphorylation of FAK and cytoskeletal proteins by *E. coli* was demonstrated in an in vitro model of brain endothelial cells [130]. In addition, phosphatidylinositol

3-kinase (PI3K) interaction with FAK and PI3-kinase signaling was shown to be necessary for successful invasion of *E. coli* in HBMECs [131]. In turn, PI3K activates phospholipase PLCγ, resulting in increased Ca2<sup>+</sup> levels in the cells [132]. The cell adhesion molecule ICAM-1 was selectively upregulated during invasion of *E. coli* in brain endothelial cells through the interaction of OmpA and its receptor Ecgp. This upregulation was dependent on the previously described PI3K signaling pathways, as well as protein kinase C (PKC)-α and NK-κB signaling [133].

Traversal of *E. coli* across the BBB has been studied extensively using HBMECs as in vitro models [22]. Transmission electron microscopy revealed the pathogen crossing the HBMECs in membrane-bound vacuoles without intracellular replication [134]. Infection of HBMECs by *E. coli* K1 was further demonstrated to activate caveolin-1, resulting in the uptake of the pathogen via the caveolae. Furthermore, caveolin-1 interacts with phosphorylated protein kinase Cα at the site of *E. coli* attachment [135]. To promote intracellular survival within the vacuoles, the *E. coli* K1 capsule is essential. It can modulate the maturation of the vacuoles and prevent fusion with lysosomes, thereby enabling traversal of live bacteria to the CNS [136]. There are gene clusters, essential for the production of the precursors that are the basis of the K1 capsule, called the *neuDB* genes. They were identified as an essential virulence factor promoting intracellular survival in HBMEC [122,136]. Furthermore, *neuDB* was also shown to be of importance during infection of an in vitro model of the BCSFB [121].

The breakdown of barrier function and neuroinflammation are considered key mechanisms in the invasion of the brain by pathogenic *E. coli.* In brain endothelial cells, upregulation of platelet-derived growth factor-B (PDGF-B) and ICAM-1 was demonstrated after infection with *E. coli* [137]. Further in vivo and in vitro studies suggest an involvement of PDGF-B in BBB permeability, mediating breakdown of tight junction proteins. Upregulation of ICAM-1 on the other hand was shown to initiate the inflammatory response of the CNS, mediating neutrophils or monocyte recruitment during infection [138]. The activation of PKC–α and its association with vascular-endothelial cadherins at the TJs of HBMECs resulted in increased cellular permeability and decrease in transendothelial electrical resistance by releasing β-catenin from the junctions. Notably, only *E. coli* strains expressing *ompA* could induce this increase in barrier permeability [139]. The infection of HBMECs with *E. coli* was further demonstrated to promote the production of nitric oxide (NO) by activating inducible nitric oxide synthase and, as a result, displayed enhanced invasion rates and increased permeability of HBMEC monolayers [140]. NO production was hypothesized to additionally be triggered by the modulation of pterin synthesis, which is involved in cell differentiation, pain modulation as well as mRNA stability. Infection of HBMECs was demonstrated to induce the rate-limiting enzyme in pterin synthesis, guanosine triphosphate cyclohydrolase (GCH1), indicating its role in the invasion process. GCH1 further interacts with Ecgp96, the receptor for OmpA [141]. These findings indicate an essential role for NO during *E. coli* invasion of the BBB. The response of HBMECs to meningitic and non-meningitic *E. coli* infections highlighted the role for macrophage migration inhibitory factor (MIF) during infection. MIF, a proinflammatory cytokine that has been described as a major factor during infection and septic shock, was demonstrated to have a role in BBB damage, as evidenced by the induction of a significant decrease in ZO-1 and occludin, as well as inflammation [142].

A further study described the role of *E. coli* K1 virulence factor *cglD* in polymorphonuclear leukocyte transendothelial migration [37]. A follow up analysis demonstrated a contribution of *cglD* to NF-κB pathway activation in HBMECs, promoting PMN adhesion and transendothelial migration across the BBB [143].

#### 4.2.2. *Neisseria meningitidis*

*Neisseria meningitidis* (*N. meningitidis*) is a human-specific gram-negative bacterium. It can colonize the nasopharynx extracellularly and is often non-pathogenic and commensal. Some strains can cause life-threatening diseases such as meningitis. To reach the barriers protecting the CNS, *N. meningitidis* has to overcome the mucosal epithelium and enter the bloodstream. To survive in the bloodstream and subsequently enter the CNS, *N. meningitidis* utilizes different virulence factors. These protect the bacterium from being killed by the hosts complement system or other effectors. They include the polysaccharide capsule and other surface structures, such as pili and other adhesins [144]. Factor H-binding protein is another important factor for bloodstream survival and evading the host innate immune system by binding factor H, which is a negative regulator of complement activation and alternative pathway and is bound to the surface of the pathogen [145,146]. *N. meningitidis* has been proposed to cross both the BBB and the BCSFB to the inner and the outer CSF [48,147–149].

The polysaccharide capsule of *N. meningitidis* is a major contributor to meningococcal disease and has been described as its main virulence factor. It can undergo genetic regulation and has the capability to mask the function of non-pilus adhesins [150]. While the capsule is essential for bloodstream survival, adhesion and invasion of host tissues are inhibited by the capsule [151]. Attenuated invasion was further described for capsulated strains of *N. meningitidis* in an in vitro model of the BCSFB [48]. The loss of the bacterial capsule for members of group B and group C meningococcal strains lead to increased uptake into HBMECs [152], and capsule and pili of *N. meningitidis* are downregulated upon contact with epithelial cells [153].

Adherence of *N. meningitidis* to host cells is facilitated by different virulence factors such as pili and surface exposed proteins like Opa and Opc, which further contribute to meningococcal disease. The type IV pili are a crucial adhesin expressed by *N. meningitidis*. They are involved in the attachment of capsulated virulence strains to host surfaces, extending from the bacterial surface through the capsule [154]. These long filamentous structures, made up of the major pilin protein and three minor pilins, promote adhesion of *N. meningitidis* to human endothelial cells via their interaction with CD147 [155]. In addition to the major pilin, type IV pili adhesion is also dependent on the expression of the outer membrane protein PilC [156–158]. Although affinity of pilin monomers to CD147 is weak, their assembly into the type IV pili as well as high expression of CD147 on the BBB supports the interaction [155]. Another group of adhesin-like structures, termed minor adhesion proteins, is expressed at low levels in vitro but may be of importance in in vivo environments which have been shown to alter the transcriptome of *N. meningitidis* [159].

While host, as well as tissue specificity, is determined by the pili and Opa proteins, invasion is facilitated by both Opa and Opc [160]. Opa proteins have been shown to bind to the carcinoembryonic antigen-related cellular adhesion molecule (CEACAM) receptor family, as well as the extracellular matrix proteins fibronectin and/or vitronectin [161,162]. An increase in adhesion and entry into HBMECs has been reported for *N. meningitidis* expressing Opc. This effect is facilitated by binding of Opc to serum vitronectin or fibronectin [163]. Also, activation of the acid sphingomyelidase/ceramide system, which involves clustering of ErbB2, an important receptor involved in bacterial uptake, by Opc-expressing *N. meningitidis* played a major role in determining the pathogens invasiveness [164]. Furthermore, the bacterial adhesins PilQ and PorA were shown to bind to the laminin receptor, thereby initiating contact with the BBB [99].

Attachment and invasion of human cells by three hypervirulent serogroup B strains of *N. meningitidis* is furthermore promoted by the *Neisseria* adhesin A (NadA), a phase-variable, surface-exposed protein [165–167]. For its interaction with the host, different membrane proteins have been suggested. NadA was shown to target human β1 integrin subunits [150]. Also, a genome-wide microarray analysis provided evidence for an interaction between NadA and the endothelial low-density oxidized lipoprotein receptor 1 (LOX-1) [168]. Following initial binding to the cell surface, the formation of microcolonies was observed, which in turn promoted the formation of specific molecular complexes called "cortical plaques". These structures contain accumulated ezrin, moesin, tyrosine-phosphorylated proteins, ICAM-1/-2, CD44 and epidermal growth factor receptors (EGFR) as well as localized polymerized cortical actin, resulting in major reorganization of host cell morphology. Changes in cell surface morphology are a prerequisite for bacterial uptake and the formation of microvilli-like cellular projections which protect the microcolonies from shear stress of the bloodstream [169].

Following attachment, *N. meningitidis* induces host cell signaling events to facilitate invasion into host tissues. These include the recruitment of ezrin as well as the activation of Src kinase and cortactin [170–172]. Furthermore, the FAK is essential for integrin-mediated internalization of *N. meningitidis* in HBMECs, enabling endocytosis through the interplay between FAK, Src and cortactin. This is facilitated by *N. meningitidis* using the integrin signaling pathways to mediate signaling from activated integrins upon attachment to the cytoskeleton [173]. These events promote the reorganization of the actin cytoskeleton resulting in the formation of membrane protrusions that take up the pathogens by surrounding them for internalization [144]. Genes involved in cytoskeletal reorganization were shown to be differentially expressed after infection of HBMECs [174]. Analysis of whole-cell lysates of human endothelial and epithelial cells have revealed an interaction of the Opc protein with alpha actinin, a modulator of various signaling pathways and cytoskeletal functions [175].

Disruption of the BBB integrity was observed in in vitro studies as a consequence of opening intracellular junctions [176]. Furthermore, cell detachment is initiated through activation of MMP 8, promoting cleavage of the TJ protein occludin [177]. Alterations of intracellular junctions were observed in a human brain microvascular endothelial cell line, potentially opening up a paracellular route of crossing the BBB into the CNS through recruitment of the Par3/Par6/PKCζ polarity complex [178].

Meningococcal disease is accompanied by an acute inflammatory response [179]. The transcription factor (TF) NF-κB, associated with the release of proinflammatory cytokines and chemokines during the inflammatory response, was shown to be active in an in vitro CP epithelial model after infection with *N. meningitidis*. Its activation was believed to arise from heterodimerization of TLR2 and TLR6 [33]. In HBMEC the p38 MAPK had an impact on the release of IL-6 and IL-8, whereas the c-JUN N-terminal kinases 1 and 2 (JNK1 and JNK2) were important for invasion [180].

#### 4.2.3. *Haemophilus influenzae*

*Haemophilus influenzae* (*H. influenzae*) is a gram-negative bacterium capable of colonizing the upper respiratory tract. The makeup of the polysaccharide capsule determines the pathogens classification into serotypes a to f, of which *H. influenzae* serotype b (*Hib*) is responsible for the most severe infections, such as meningitis, especially in children under the age of 5 years [181]. Although there is widespread vaccination against *Hib*, some populations remain vulnerable and the vaccines do not protect against other serotypes [182]. Furthermore, in areas where Hib vaccines are used, nontypeable *H. influenzae* (NTHi) strains can now be the most common cause of meningitis [183]. Similar to the before described pathogens, *H. influenzae* has to cross the epithelial barrier of the upper respiratory tract and, after dissemination and survival in the bloodstream, cross the barriers of the brain to enter the CNS [184]. Traversal of both the BBB and BCSFB has been demonstrated for *H. influenzae* [47,51].

Attachment to host cells is enabled by the bacterial capsule and fimbriae which are subject to reversible phase variation [185]. *H. influenzae* shares a common strategy to enter endothelial cells with *S. pneumoniae* and *N. meningitidis*, which involves binding to the PAFR, mediated by phosphorylcholine (ChoP) [186,187]. This interaction results in the entry of the pathogens into the BBB through activation of β-arrestin–mediated uptake [188]. Binding of PAFR by lipooligosaccharide (LOS) glycoforms containing ChoP was also demonstrated during invasion NTHi. This binding resulted in the activation of host cell signaling by coupling with pertussis toxin-sensitive (PTX) heterotrimeric G protein complexes and invasion of the pathogen. It was further suggested that this mechanism was more efficient than micropinocytosis [187]. Binding to laminin receptor, another shared mechanism of these pathogens, is initiated by OmpP2, facilitating the interaction with the brain endothelium by *H. influenzae* [99].

In an in vitro model of the BCSFB, *Hib* as well as clinical isolates of *Hib* and *H. influenzae* serotype f (*Hif)* were shown to adhere and invade the HIBCPP cells as intracellular bacteria. Both fimbriae and the capsule lead to attenuated invasion [47]. Also, a study using *H. influenzae* serotype a (*Hia*) and a co-culture of HBMECs and pericytes demonstrated an activation of stimulated A2A and A2B adenosine receptors after infection. This in turn led to the release of Vascular Endothelial Growth Factor (VEGF) by the pericytes leading to pericyte detachment and endothelial cell proliferation resulting in overall BBB impairment [189].

Using rat models of meningitis, a dose-dependent increase of BBB permeability was observed after inoculation with *Hib* LPS [190]. Later studies showed that the permeability of the BBB was also increased in rats after inoculation with *H. influenzae* outer membrane vesicles (OMV), suggesting a role for these vesicles in transporting *Hib* LPS to the CSF during meningitis [191]. Furthermore, it was shown that zyxin, a cytoskeletal protein implicated in the protection of TJs in the BBB, is critical for the integrity of the BBB and, as a consequence, for protecting against invading pathogens such as *H. influenzae* [192]. Overall, the inflammatory response of patients during the infection with *H. influenzae* is determined by several virulence factors including the capsule, adhesion proteins, pili and outer membrane proteins as well as LPS and IgA1 protease [193].


**1.**Evidencefortheinvolvementofgram-positiveandgram-negativebacteriaduringbrainentryattheBBBand

255



#### **5. Conclusions**

A variety of studies have focused on how pathogens cross the blood-CNS barriers, mostly focusing on the BBB. Although significant progress has been made in identifying mechanisms of host–pathogen interactions during bacterial meningitis, additional efforts towards identifying bacterial and host cell targets are needed. The diversity of mechanisms used by these pathogens presents the need for further research. To this end, the identification of common mechanisms used by multiple pathogens will be of great significance and further assist the development of effective therapies. Of high importance is the distinction between the mechanisms the pathogens use for crossing the BCSFB and the BBB. Pathogens exploiting epitopes of both barriers would present interesting targets for the development of therapeutics.

**Author Contributions:** Conceptualization, R.H. and C.S.; writing—original draft preparation, R.H.; writing—review and editing, H.S. and C.S.

**Funding:** This research received no external funding.

**Acknowledgments:** We acknowledge financial support by Deutsche Forschungsgemeinschaft within the funding programme Open Access Publishing, by the Baden-Württemberg Ministry of Science, Research and the Arts and by Ruprecht-Karls-Universität Heidelberg.

**Conflicts of Interest:** The authors declare no conflict of interest

#### **Abbreviations**



#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Review* **Coagulase-Negative Staphylococci Pathogenomics**

## **Xavier Argemi 1,2,\*, Yves Hansmann 1,2, Kevin Prola <sup>2</sup> and Gilles Prévost <sup>2</sup>**


Received: 22 January 2019; Accepted: 7 March 2019; Published: 11 March 2019

**Abstract:** Coagulase-negative Staphylococci (CoNS) are skin commensal bacteria. Besides their role in maintaining homeostasis, CoNS have emerged as major pathogens in nosocomial settings. Several studies have investigated the molecular basis for this emergence and identified multiple putative virulence factors with regards to *Staphylococcus aureus* pathogenicity. In the last decade, numerous CoNS whole-genome sequences have been released, leading to the identification of numerous putative virulence factors. Koch's postulates and the molecular rendition of these postulates, established by Stanley Falkow in 1988, do not explain the microbial pathogenicity of CoNS. However, whole-genome sequence data has shed new light on CoNS pathogenicity. In this review, we analyzed the contribution of genomics in defining CoNS virulence, focusing on the most frequent and pathogenic CoNS species: *S. epidermidis*, *S. haemolyticus, S. saprophyticus*, *S. capitis*, and *S. lugdunensis*.

**Keywords:** pathogenomics; coagulase-negative staphylococci; virulence factors; whole genome sequencing

#### **1. Introduction**

Bacterial virulence is a complex concept that must be considered from the clinical, molecular, and genomic perspectives. Clinically, the virulence of a pathogen, of a specific species or even of a clonal strain, can refer to its inherent capacity to provoke specific clinical manifestations that can be linked to the production of a virulence factor, often a protein. Microbiologically, Robert Koch postulates illustrated bacterial virulence, which Stanley Falkow redefined in the late 1980s to provide a molecular version of it [1]. Molecularly, the virulence factor is found in the pathogenic strains of a species and not in non-pathogenic ones. Secondly, the inactivation of the encoding gene in an animal model should attenuate the virulence of the strain. Thirdly, the reintroduction/reactivation of the gene should restore virulence. Nevertheless, in the past 20 years, neither Koch's nor Falkow's hypotheses have been able to completely define virulence factors. The development of rapid and cost-effective genome sequencing technologies has provided the opportunity, among others, to discover a wide range of genes that could be linked to bacterial pathogenicity. This ability to determine the pathogenic capacities of a bacterium from its genome is known as pathogenomics [2]. Virulence factor databases allow for fast and easy identification of putative virulence factors in whole-genome sequence data based only on sequence homologies. The Virulence Factor Database was released in 2005 and is an example of a regularly-updated database, of which a fourth version was released in 2018 [3]. Other updated databases exist, such as PHI-base and Victors [4,5]. Thus, some virulence factors retain the definition of Stanley Falkow, with a direct causative link with bacterial pathogenicity and some possible clinical manifestations. Conversely, some other factors, such as those that are produced by human commensal bacteria, do not act as virulence factors under normal conditions, but these factors will take part in bacterial pathogenicity under specific conditions, such as during cutaneous barrier breach. Staphylococci are examples of such dual characterizations of virulence factors.

Staphylococci comprise *Staphylococcus aureus*, which is a human commensal bacterium that colonizes nearly 30% of the human skin and mucosae, and coagulase-negative staphylococci (CoNS), which include species such as *Staphylococcus epidermidis*, which colonizes nearly all human skin [6]. *S. aureus* is usually considered separately from CoNS and as a virulent species. Indeed, it produces a wide range of toxins which can act as virulence factors leading to specific clinical manifestations with a direct causative link. For example, the Panton-Valentine leucocidin (LukSF-PV) is a bicomponent toxin that may lead to necrotizing pneumonia [7]. *S. aureus* produces other toxins such as epidermolysins (ETA-B-D) and superantigens, e.g., toxic shock syndrome toxin (TSST-1), that lead to specific clinical diseases and turn some particular strains into pathogens that fully meet the Koch and Falkow postulates [8,9]. *S. aureus* might also produce numerous adhesion factors, capsule, and biofilm-associated proteins, hemolysins, and immune evasion proteins whose main purpose is to persist on human skin as a commensal and probably beneficial bacterium [10]. Many of those proteins are cell wall-anchored and comprise a group of covalently-attached proteins with specific motifs that bind to the extracellular matrix and are usually identified as microbial surface components recognizing adhesive matrix molecules (MSCRAMMs) [11]. Many other proteins are cell-wall-attached but display distinct functions such as iron-regulated surface proteins, and some genes encode for putative MSCRAMMs but without any molecular and structural confirmation, and can be identified more generally as surface adhesins, a terminology used in this review. Those adhesion factors play a crucial role in the pathogenicity *of S. aureus*. The occurrence of a cutaneous barrier breach will allow its penetration in a new environment, where those factors will behave as real virulence factors, allowing its adherence, persistence, and multiplication in a normally sterile environment. Whole-genome sequence analysis of *S. aureus* has revealed the existence of a very large repertoire of such toxins, adhesion molecules, and biofilm-associated genes, sometimes without any molecular confirmation but based on sequence homologies that lead to the characterization of *S. aureus* as a major pathogenic species [12].

CoNS are often considered as simple commensal bacteria due to their rarity in clinical pathology and the absence of virulence factors like those produced by *S. aureus.* However, the emergence of nosocomial infections with CoNS has led clinicians and researchers to reconsider the status of these bacteria [13]. The publication of several CoNS genomes has been an essential tool to better understand CoNS pathogenicity. The identification of virulence factors, such as toxins with clinical impacts, remains exceptional or even controversial, and it appears that CoNS have a very large repertoire of genes that encode for adhesion factors, biofilm production, hemolysins, exoenzymes, and superantigens. Most CoNS also have similar regulatory systems to *S. aureus*, e.g., the *agr* system, for which homologs have been identified in numerous CoNS species [13]. The cutaneous barrier breach is a critical step to turning the CoNS species into pathogens, turning factors that are mainly implicated in the bacterial life cycle on the skin into virulence factors leading to pathological manifestations. In this situation, once again, the concept of virulence factors does not apply to the definitions provided by Robert Koch and Stanley Falkow. Considering the great species diversity of CoNS, *S. epidermidis* is the most frequent CoNS implicated in clinical diseases, and it remains the best-studied species at a molecular and genomic level [14]. Nevertheless, none of the virulence factors that were discovered could clearly categorized a specific strain as pathogenic or commensal, even if the strains belonging to the clonal type ST2 have been over-represented in the specific context of *S. epidermidis* neonatal sepsis [15]. ST2 clonal type always bears the insertion sequence *IS256* and *ica* genes, which are implicated in biofilm production, but the direct correlation of those loci with *S. epidermidis* invasiveness has never been confirmed [16]. Some other species also came up as significant pathogens, such as *S. lugdunensis* that has been widely investigated due to the severity and the specificity of clinical manifestations of infection due to this species [17]. *S. haemolyticus*, *S. saprophyticus,* and *S. capitis* are three other species for which the rate and the type of infection led to molecular and genomic

investigations in the search of virulence factors [18,19]. *S. caprae* is a CoNS that usually belongs to the animal skin flora, but it can also colonize human skin. This CoNS has been described in veterinary infections (mastitis principally) and in human osteoarticular infections. Interestingly, this CoNS belongs to a group of CoNS that is comprised of *S. epidermidis*, *S. capitis*, and *S. saccharolyticus* [13,20–22].

The identification of virulence factors in staphylococci has resulted in the identification of mobile genetic elements. Indeed, the majority of *S. aureus* virulence genes are located on mobile genetic elements such as pathogenicity islands, plasmids, and phages, which represent up to 25% of the *S. aureus* genome and play crucial roles in host adaptation and the modulation of virulence [23–25]. Several bioinformatic tools have been developed to identify such genomic regions based on comparative genomics [26–28]. Thus, the identification of putative virulence factors can be driven by the identification of these genomic elements, but it has appeared that the presence of such mobile genetic elements is very common in CoNS, but that they do not bear any virulence factors which are similar to *S. aureus.* More generally, this large repertoire of mobile genetic elements in staphylococci species explains why, despite a limited core genome, those species usually displayed an open pan genome with a very high number of existing genes, sometimes shared by a very limited number of clones [29–31]. Worthy of note is the fact that *S. lugdunensis* has a closed pan genome which contains several barriers to horizontal gene transfers, which could explain this unique genomic characteristic amongst CoNS [32].

In this review, we aimed to describe the occurrence of virulence factors, or more accurately, of pathogenicity factors, based on whole-genome sequence data. We focused this systematic review on the major CoNS pathogens for which such data have been published: *S. epidermidis*, *S. lugdunensis*, *S. saprophyticus, S. haemolyticus*, *S. caprae*, and *S. capitis.* A comprehensive review of all CoNS pathogenicity factors has been described elsewhere and is covered in references [13,33–39].

#### **2. Research Method and Results**

This systematic review was based on PubMed published articles in the English language. We used the following MeSH terms: "Virulence", "Virulence factors", "Pathogenicity", "Whole genome sequencing", and "Genomics". In particular, we searched for the following species: "*Staphylococcus epidermidis*", "*Staphylococcus lugdunensis*", "*Staphylococcus capitis*", "*Staphylococcus caprae*", "*Staphylococcus haemolyticus*", and "*Staphylococcus saprophyticus*". We excluded studies that were not based on whole-genome data, and those for which genome sequences were not deposited in public databases (e.g., GenBank, European Nucleotide Archive).

We identified eight studies for *S. epidermidis* [29,30,40–45], four for *S. capitis* [46–49], three for *S. lugdunensis* [32,50,51], two for *S. haemolyticus* [52,53], and one for *S. caprae* [54] and *S. saprophyticus* [55].

#### *2.1. Staphylococcus epidermidis*

*S. epidermidis* is by far the most prevalent CoNS in microbiological samples and the primary cause of CoNS-related infections, particularly in nosocomial settings [56]. This species has been widely studied at a molecular level and has provided the largest amount of data regarding the presence of virulence factors that might be identified as pathogenicity factors [16]. Biofilm formation is the main path by which *S. epidermidis* colonizes and infects prosthetic and medical devices as reviewed by Otto et al. [57]. The biofilm life cycle is a complex process which implicates MSCRAMMs, the proliferation of the exopolysaccharide matrix, and finally, biofilm dispersion, mostly due to phenol-soluble modulin (PSM) peptides [57].

Nearly 500 genomes have been made available on GenBank, mainly draft genomes, but also 12 complete genomes. *S. epidermidis* whole-genome-based studies to determine virulence determinants are detailed in Table 1. In 2003, Zhang et al. provided the first genome-based analysis of *S. epidermidis* virulence [40]. The authors provided the first sequence of the reference strain ATCC 12228, a non-biofilm forming, non-infection associated strain, and they found several putative pathogenicity factors, including multiple MSCRAMMs encoding genes and putative exoenzymes and toxins such

as metalloproteases and Delta/Beta-hemolysins. Interestingly, a revised version of this genome was released in 2017 which was determined using PacBio (PacBio, Pacific Biosciences, Menlo Park, CA, USA) long read sequencing; it provided a slightly different genome from the one released by Zhang et al., and established a new reference genome for comparative genomic studies [58]. In 2005, Gill et al. published a new complete genome from *S. epidermidis* strain RP62a and performed comparative genomic analyses with four *S. aureus* genomes, and included the genome from *S. epidermidis* ATCC 12228 [41]. As expected, the authors found multiple virulence factors in the *S. aureus* genome. Nearly one-half of those factors were carried by seven pathogenicity islands that were not found on *S. epidermidis*. Nevertheless, some other mobile genetic elements were identified in this species (one genomic island and two integrated plasmids), some bearing PSM and putative MSCRAMMs. Some other mobile genetic elements were found in *S. epidermidis* genomes such as prophages, insertion sequences, and *SSCmec-like* cassettes. The authors also identified several other putative virulence factors such as proteases (serine and cysteine protease), lipases, and hemolysins (*Beta/Delta haemolysin*) loci. The presence of mobile genetic elements and virulence factors led the authors to consider that horizontal gene transfer between staphylococci had to be considered as a source of variability and, in this case, as a cause of their pathogenicity.

Conlan et al. provided in 2012 the first large comparative genomic analysis of 30 *S. epidermidis* whole genomes, showing that *S. epidermidis* had an open pan genome, as expected, but more interestingly, that whole-genome-based phylogenetic trees could distinguish between commensal and pathogenic strains. The presence of the *formate dehydrogenase* gene (*fdh*) could explain this observation [42]. Meric et al. found in a second study based on whole-genome comparative analyses that *S. epidermidis* and *S. aureus*, in hospital settings, share some genes involved in pathogenicity, suggesting the existence of horizontal gene transfer between these species [29]. The added value of whole-genome analysis was confirmed by Virginia Post et al., who compared whole-genome sequences of 104 *S. epidermidis* isolates from patients with orthopedic-device-related infections. This study found a correlation between patient outcome and some loci [30]. Patients with multiple surgeries due to treatment failure were more likely to be infected with the biofilm-associated gene *bhp*, the antiseptic resistance gene *qacA*, the cassette genes *ccrA* and *ccrB*, and the *IS256-like* transposase gene. This whole-genome-based study identified, for the first time and directly from whole-genome data, loci that might be involved in *S. epidermidis* pathogenicity based on well-documented clinical data.

While such studies have yielded novel and sometimes revolutionary insights into *S. epidermidis* virulence, the presence of such loci does not provide any information regarding their expression, regulation, and their impact on clinical implications. Yet, whole-genome data can be used as a complementary tool when a putative virulence factor is identified, sometimes starting from clinical considerations. In 2013, Fournier et al. provided observations of a patient with *S. epidermidis* endocarditis, and the authors found several clues that could explain the virulence of this strain [43]. Besides known virulence factors such as MSCRAMMs, exoenzymes, and hemolysins, the authors identified a previously unreported prophage, a new toxin/antitoxin module, and a complete *icaABCD* operon, which was usually not observed in non-pathogenic strains. As observed by Conlan et al., this pathogenic strain also lacked the *fdh* gene.

The existence of enterotoxins and pathogenicity islands in *S. epidermidis* has been controversial, despite observations by Madhusoodanan et al. of such a genetic element in the clinical strain *S. epidermidis* FRI909, which carried two functional enterotoxin genes *sec3* and *sell* [59]. This unique observation was initially considered as a possible genetic "accident". In 2016, we described two strains isolated from septic shock patients that produced a staphylococcal enterotoxin C with homology to *S. aureus* enterotoxin C3 [60]. By using whole-genome data, we thereafter provided evidence that the *sec3* from our strains was located on a pathogenicity island very similar to the one described by Madhusoodanan et al., named SePI-1/SeCI-1 [44]. We also identified several plasmids carrying resistance genes and sharing homologies with *S. aureus*. This result confirmed that some mobile genetic elements from *S. epidermidis* might come from *S. aureus*, with the transfer of virulence factors.





Recently, Xu et al. performed a whole-genome analysis of a multi-resistant *S. epidermidis* strain isolated in the environment, i.e., in a hotel room, and identified multiple loci dedicated to antibiotic resistance but also to numerous virulence genes, as described previously [45].

*S. epidermidis* is a major cause of nosocomial infections within other CoNS. The existence of several factors that modulate its commensal life cycle on the skin also turns this species into a pathogen when entering sterile sites. Whole-genome studies have provided specific and unique data regarding the general evolution of its genome and interactions with other staphylococci, such as *S. aureus*. However, whole-genome data also appear as crucial and unique complementary analyses when a specific strain is isolated clinically, or when a putative virulence factor is molecularly identified.

#### *2.2. Staphylococcus lugdunensis*

*S. lugdunensis* is considered a significant pathogen in human infection [17]. Clinically, its virulence is most likely lower than that of *S. aureus*. Clinical observations have emphasized the frequency and severity of skin and soft tissue infections, endocarditis, and osteoarticular infections [61–63]. Microbiologically, this CoNS produces a coagulase, which is a common feature with *S. aureus* [64]. Its pathogenicity has been noted by several authors, and multiple virulence factors have been identified molecularly, including adhesion molecules like a fibrinogen-binding protein; cytotoxins such as δ-like-hemolysin; a complete biofilm synthesis system, including an *agr* regulating operon; and as seen in *S. aureus*, a large system dedicated to iron metabolism [65–67]. Interestingly, Heilbronner et al. showed that duplication of the *isd* locus provided a selective advantage to *S. lugdunensis* in the case of iron limitation, possibly improving bacteria survival and pathogenicity. Recently, we identified and purified a novel metalloprotease named lugdulysin that could be implicated in osteoarticular infections [61].

*S. lugdunensis* whole-genome-based studies to determine virulence determinants are listed in Table 2. The first complete genome of *S. lugdunensis* was published by Tse et al. in 2010 [68]. Since then, 17 complete genomes have been published, and eight additional draft genomes are available in GenBank. Nevertheless, the first exploitation of a complete genome was provided by Heilbronner et al. in 2011 [50]. The authors published the complete genome of strain N920143 and identified several putative virulence factors based on comparative genomics and homologies searches. They identified, for the first time, a complete operon implicated in iron metabolism (*isd*: *isdB-J-C-K-E-F-G*), several genes coding for MSCRAMMs such as fibrinogen and Von Willebrand adhesion factors, a *streptolysin S-like toxin*, an *IcaABCD* operon for biofilm synthesis, *agr* regulation genes, and even an *esx* locus with homology to the esx-ESAT 6 secretion systems that are implicated in the secretion of virulence factors in *Mycobacterium tuberculosis*. Due to the frequent localization of virulence genes on mobile genetic elements in *S. aureus*, in 2017 we also published the complete genome of seven strains of *S. lugdunensis* and identified multiple mobile genetic elements [51]. We identified a very unusual number of plasmids and prophages by using computational methods, including four new prophages and five plasmids, in which three were previously described in other CoNS and one in *S. aureus*. This study suggested that horizontal gene exchange could occur between staphylococci, including *S. aureus*. Descriptions of plasmids and prophages in CoNS remain rare, but, interestingly, they have been performed mainly in *S. epidermidis* and *S. haemolyticus*.


*Staphylococcuslugdunensis*whole-genome-basedstudiestodeterminevirulence

*S. lugdunensis* is characterized by its unusual antimicrobial susceptibility, even if some rare strains are oxacillin-resistant. In 2017, Chang et al. fully characterized, for the first time, two novel variants of staphylococcal cassette chromosome *mec* elements in two oxacillin-resistant *S. lugdunensis* strains by using whole-genome sequence data [69]. The authors also identified several homologies with *S. aureus*, *S. haemolyticus*, and *S. epidermidis* cassette chromosome regions. If this result suggested once again the existence of horizontal gene exchanges between staphylococci, including *S. aureus*, the very unusual antimicrobial susceptibly of *S. lugdunensis* remained unexplained in comparison to other CoNS, which displayed a very high rate of resistance, particularly in nosocomial settings [19]. We recently published a comparative genomic analysis of *S. lugdunensis* whole-genome with *S. aureus* and *S. epidermidis* [32]. *S. epidermidis* and *S. aureus* are characterized by an open pan genome, which means that those species display a virtually unlimited number of new genes when adding different genomes, even if they conserve a limited number of genes through the strains called the core genome. Conversely, *S. lugdunensis* display a closed pan genome, and all published genomes were very similar, with only a very limited number of new genes among the strains. This unexpected characteristic could explain the very well conserved antimicrobial susceptibility of *S. lugdunensis* that rarely acquires resistance genes. We also found that this observation could rely on the presence of numerous barriers to horizontal gene transfer in this species, including restriction-modification, CRISPR/Cas, and toxin/antitoxin systems.

*S. lugdunensis* is a very particular CoNS and might be closer to *S. aureus* than other CoNS in terms of virulence. Whole-genome sequence studies have emphasized its originality on a genomic scale, including the presence of multiple unusual factors implicated in its pathogenicity and that are uncommon for CoNS, like its iron metabolism system. Overall, its genome encodes for several factors that take part in its survival on the skin as a commensal bacterium, such as adhesion proteins and biofilm capacities. Once again, those factors probably act as virulence factors once the cutaneous barrier is broken. Comparative genomic studies have also revealed that it displays a unique profile in terms of genomic plasticity, which could explain some of its microbiological characteristics, such as its highly-conserved antimicrobial susceptibility.

#### *2.3. Staphylococcus capitis*

*S. capitis* is the third CoNS that has been described in clinical infections and for which whole-genome studies focusing on its virulence are available [13]. *S. capitis* infections have been described in various clinical situations implicating biofilm production as endocarditis, catheter-related bacteremia, and prosthetic joint infections [70,71]. This species also causes neonatal sepsis in neonatal intensive care units [72,73].

*S. capitis* whole-genome-based studies to determine virulence determinants are listed in Table 3. The first draft genome of this species was released in 2009, and the first complete genome was published in 2015 by Cameron et al. to determine the genetic determinants that could support its pathogenicity [46]. The authors identified several loci that could encode for virulence factors and performed a comparative genomic analysis with the *S. epidermidis* genome. They identified putative virulence regulators (as *agr* homologs), biofilm production loci, genes encoding for exoenzymes (as metalloproteases and hemolysins), PSMs, and MSCRAMMs. This genome-scale analysis led to a functional analysis of biofilm and PSMs production.


*Staphylococcuscapitis*whole-genome-basedstudiestodeterminevirulencedeterminants.

At a larger scale, it appears that some clonal strains might be implicated in the context of neonatal sepsis as suggested by Butin et al., who observed the worldwide endemicity in 17 countries of a multidrug-resistant strain by using pulsed-field gel electrophoresis patterns but also whole-genome sequence data [74]. Multidrug resistance has emerged as a crucial issue in *S. capitis* infections, with nearly half of the strains being oxacillin-resistant in hospital settings; occasionally, some strains with vancomycin reduced susceptibility have been identified [72,73]. Li et al. provided a complete whole-genome analysis of a multi-resistant strain causing bacteremia, and found a very high and unusual number of genes implicated in such a resistant profile [47]. Recently, Simoes et al. published the first whole-genome sequence of *S. capitis* strain NRCS-A, which is a multi-resistant clone implicated in neonatal infections worldwide, by using PacBio (PacBio, Pacific Biosciences, Menlo Park, CA, USA) long read sequencing, and provided comparative genomic analysis with three other NRCS-A sequenced clones coming from different countries [48]. The authors identified, as expected, multiple resistant genes, but interestingly, some of them were located on mobile genetic elements. The authors identified several virulence genes which had been described previously by Cameron et al., but more specifically, they identified which of them was clone-specific as a putative restriction-modification system and a nisin resistance gene, which could explain the persistence of NRCS-A clones in patients' gut microbiota. In 2017, Kumar et al. used whole-genome sequence data to reveal the existence of four loci encoding for antimicrobial peptides [49]. The authors were then able to perform a functional analysis to confirm the functionality (antimicrobial activity against Gram-positive bacteria, including *S. aureus*) of those genes, and formulate the hypothesis that such antimicrobial peptides could explain the adaptation and the persistence of *S. capitis* on the human skin.

*S. capitis* is considered a serious pathogen, particularly in neonatal settings, and whole-genome studies have provided useful and unique data regarding the spread of its antimicrobial resistance, but have also emphasized its ability to colonize and persist on human skin, which is directly linked to its pathogenicity regarding catheter-associated infections.

#### *2.4. Staphylococcus caprae*

*S. caprae* is a normal component of the animal skin flora and can cause mastitis, principally in goats [13]. It also belongs to the human skin flora and is an observed cause of osteoarticular infections and also material-associated infections and bacteremia [20]. Until recently, only six draft genomes of *S. caprae* were available from humans and goats. *S. caprae* whole-genome-based studies to determine virulence determinants are listed in Table 4. In 2018, Watanabe et al. published a comparative genomic analysis of three human strains by performing a whole-genome assembly which leads to three complete chromosomes [54]. The authors compared the conserved genome parts of *S. epidermidis*, *S. capitis*, and *S. caprae*, as all previous phylogeny studies found all three species belonged to the same group (*S. epidermidis* group) [22]. They found that all three species shared a biofilm- and capsule-associated loci, and they also shared PSM genes. They also identified several putative MSCRAMMs genes. We observed that the three species belonged to the *S. epidermidis* group and shared very common features for skin resident bacteria. The similarity in clinical presentation of infection caused by these species in human disease has been illuminated.


**4.***Staphylococcushaemolyticus,*and*saprophyticus*whole-genome-basedstudiestodeterminevirulence

*Int. J. Mol. Sci.* **2019**, *20*, 1215

#### *2.5. Staphylococcus haemolyticus*

*S. haemolyticus* is a commensal bacterium but is also a frequent nosocomial pathogen that has been described mainly in catheter-related bacteremia, with one of the highest degrees of methicillin resistance among CoNS [19,75,76]. *S. haemolyticus* whole-genome-based studies to determine virulence determinants are listed in Table 4. *S. haemolyticus* was one of the first CoNS with a complete genome published by Takeuchi et al. in 2005 [52]. The authors identified several putative virulence factors in this pioneering study, such as hemolysins and bacterial capsule. Until now, no publication has completely explored the presence of virulence genes in this species, warranting future studies. Some previous studies have described the presence of cytotoxins, enterotoxins, and PSMs at the molecular level, which is exceptional among CoNS [77,78]. This species is also a biofilm-producing species, but interestingly, in 2016, Hong et al. published the first complete sequence of *S. haemolyticus* by using PacBio (PacBio, Pacific Biosciences, Menlo Park, CA, USA) long read sequencing and found no *ica* operon, but an *agr* regulation system homolog, suggesting that alternative mechanisms were implicated in biofilm formation [53]. Thus, *S. haemolyticus* is a significant pathogen in humans, but it appears that a complete whole-genome analysis of those virulence factors is lacking, whereas, several complete genomes are already available with this species bearing very unusual virulence factors for a CoNS.

#### *2.6. Staphylococcus saprophyticus*

*S. saprophyticus* is a commensal CoNS and can lead to lower urinary tract infections in young women [79,80]. Colonization of the genital area and the gut has been linked to this atypical type of infection, but most importantly, it was shown in vitro that this CoNS has an unusual capacity to adhere to urothelial cells and to produce urease. Nevertheless, those characteristics are not unique in CoNS and do not fully explain the capacity of this CoNS to cause urinary tract infections. *S. saprophyticus* whole-genome-based studies to determine virulence determinants are listed in Table 4. In 2005, Kuroda et al. established the first complete genome of *S. saprophyticus*, and until now, only six complete genomes for this species have been made available in GenBank [55]. The authors identified three factors that could explain the specificity of *S. saprophyticus* pathogenicity by using whole-genome data and a comparative genomic approach. They identified a novel—and unique among CoNS—cell wall-anchored protein, UafA, associated at a molecular level with a high capacity to adhere to cells from the urinary tract. They also identified a unique uro-adaptative transport system and urease production as the two other factors that could be linked to the specific pathogenicity of *S. saprophyticus*. Since 2005, whole-genome virulence studies have been lacking, but adherence and persistence onto the urinary tract have been confirmed as the main factors that could be linked to pathogenicity [81–83]. Overall, *S. saprophyticus*, lacks many of the adhesion proteins and other virulence factors that have been identified in CoNS from the *S. epidermidis* group, *S. caprae*, and *S. lugdunensis*, which probably explains the differences that are observed at a clinical level.

#### *2.7. Phylogenetic Relationship among Staphylococcus Species*

In 2012, Lamers et al. proposed a refined classification of staphylococci using a combination of Bayesian and maximum likelihood analysis of multilocus data based on four loci (non coding 16S rRNA, *dnaJ*, *rpoB*, and *tuf*) [22]. The authors identified six species groups and 15 cluster groups among staphylococci, a classification that is considered as a current standard for staphylococci phylogenetic classification [13]. Interestingly, *S. lugdunensis* appeared in a unique cluster group, a particularity that could be link to the very low allelic polymorphism of this species and the existence of a closed pangenome, a unique characteristic among CoNS [32,84]. In addition, *S. epidermidis*, *S. capare*, and *S. capitis* which are described here as pathogenic species in nosocomial settings belong to the same cluster group, along with *S. saccharolyticus*. *S. saprophyticus* and *S. haemolyticus* belong to distinct cluster groups in this classification. This classification does not clarify the mechanisms by which staphylococci

acquire genes, including virulence factors, but only emphasize the uniqueness of *S. lugdunensis*, even if more generally, all staphylococci have strong barriers preventing horizontal gene transfers, making their transformation extremely difficult even in vitro [85].

#### **3. Conclusions**

The emergence of nosocomial infections with CoNS has illuminated the role of numerous virulence factors. In the context of infections and breaches of the skin barrier, these factors allow virulent adherence, persistence, and multiplication of CoNS. The use of whole-genome sequence data has evidenced the multiplicity of such factors in CoNS. It would be more appropriate to characterize these elements as pathogenicity factors, even if some of them can appear exceptionally as virulence factors that may lead to specific clinical conditions, as seen with *S. epidermidis*-producing strains.

The increasing data that are now available at the clinical, molecular, and genomic levels even make possible the development of innovative approaches to characterize bacterial pathogenicity. Deneke et al. proposed such a novel approach by using a machine learning workflow that determines the pathogenicity of a novel bacterial species based on genomic data only, an eventuality that is not rare with the increasing use of metagenomic analyses in various environments and microbiota [86].

Nevertheless, if sequencing technologies have become cheaper and whole-genome sequence data easier to produce, they will remain as complementary tools to molecular and clinical studies that have to be coordinated in a structured framework [87]. Good practices and quality controls have become crucial to ensure the quality of the released genomes and the validity of analyses [88,89].

**Author Contributions:** Conceptualization: X.A. and G.P.; Methodology: X.A. and G.P.; Analysis of the studies: X.A. and Y.H.; Writing–Original Draft Preparation: X.A. and K.P.; Writing–Review and Editing: G.P. and Y.H.; Supervision: G.P.

**Funding:** No funding.

**Acknowledgments:** The authors would like to thank Enago (www.enago.com) for the English language review.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Review* **Pathogenicity and Virulence of** *Trueperella pyogenes***: A Review**

### **Magdalena Rzewuska \*, Ewelina Kwiecie ´n, Dorota Chrobak-Chmiel, Magdalena Kizerwetter-Swida, Ilona Stefa ´ ´ nska and Małgorzata Giery ´nska**

Department of Preclinical Sciences, Faculty of Veterinary Medicine, Warsaw University of Life Sciences, Ciszewskiego 8 St., 02-786 Warsaw, Poland; ewelina1708@gmail.com (E.K.); dorota.chrobak@wp.pl (D.C.-C.); magdakiz@wp.pl (M.K.-S.); i.stefanska@gmail.com (I.S.); malgorzata\_gierynska@sggw.pl (M.G.) ´

**\*** Correspondence: magdalena\_rzewuska@sggw.pl; Tel.: +48-225936032

Received: 16 May 2019; Accepted: 31 May 2019; Published: 4 June 2019

**Abstract:** Bacteria from the species *Trueperella pyogenes* are a part of the biota of skin and mucous membranes of the upper respiratory, gastrointestinal, or urogenital tracts of animals, but also, opportunistic pathogens. *T. pyogenes* causes a variety of purulent infections, such as metritis, mastitis, pneumonia, and abscesses, which, in livestock breeding, generate significant economic losses. Although this species has been known for a long time, many questions concerning the mechanisms of infection pathogenesis, as well as reservoirs and routes of transmission of bacteria, remain poorly understood. Pyolysin is a major known virulence factor of *T. pyogenes* that belongs to the family of cholesterol-dependent cytolysins. Its cytolytic activity is associated with transmembrane pore formation. Other putative virulence factors, including neuraminidases, extracellular matrix-binding proteins, fimbriae, and biofilm formation ability, contribute to the adhesion and colonization of the host tissues. However, data about the pathogen–host interactions that may be involved in the development of *T. pyogenes* infection are still limited. The aim of this review is to present the current knowledge about the pathogenic potential and virulence of *T. pyogenes*.

**Keywords:** *Trueperella pyogenes*; virulence; pyolysin; infection; pathogenicity; immune response; *Actinomycetales*

#### **1. Introduction**

The species *Trueperella pyogenes* [1], previously classified as *Arcanobacterium pyogenes* [2], *Actinomyces pyogenes* [3,4], and formerly as *Corynebacterium pyogenes* [5], belongs to the family *Actinomycetaceae*, in the order *Actinomycetales* of the class *Actinobacteria*, the so-called actinomycetes [6]. A dendrogram representing the phylogenetic relationship between this bacterium and some other pathogenic *Actinomycetales* is shown in Figure 1. *T. pyogenes* is a Gram-positive, pleomorphic, non-spore-forming, non-motile, non-capsulated, facultatively anaerobic rod, which is characterized by a fermentative metabolism and strong proteolytic activity [4]. Its growth requirements are not excessive, but media enriched with blood or serum need to be used for the culture. The preliminary recognition of *T. pyogenes* isolates is based on the cell morphology; the features of colonies, which are surrounded by a zone of beta-haemolysis on blood agar; and a negative catalase assay. Then, the biochemical properties can be tested for species determination [7]. Sometimes, additional bacteriological methods other than the conventional ones are necessary for the differentiation and appropriate identification of isolates. New techniques, such as loop-mediated isothermal amplification (LAMP) assay, matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF) mass spectrometry, Fourier transform infrared (FT-IR) spectroscopy, or 16S rRNA gene sequencing may be useful for the diagnostics of *T. pyogenes* infections [8–13]. Those methods enable the recognition of the closely related taxa of the order *Actinomycetales*, and sometimes the reclassification of some of them [14].

**Figure 1.** Neighbor-joining [15,16] phylogenetic tree based on 16S rRNA gene sequences (a total of 1422 positions in the final dataset) of *Trueperella pyogenes*, other *Trueperella* species, and related taxa. Only bootstrap values of 70%, based on 1000 replicates, are shown next to the branches [17]. The scale bar represents the number of substitutions per site. The analysis involved 64 nucleotide sequences derived from the GenBank® database, and the evolutionary relationships were calculated in MEGA X [18].

*T. pyogenes* is considered to be a part of the biota of skin and mucous membranes of the upper respiratory and urogenital tracts of animals [19–21]. Moreover, this bacterium was also isolated from the wall of bovine rumen and swine stomachs, as a gastrointestinal microbiota [22,23], and from the udders of clinically healthy cows [24]. However, *T. pyogenes* is also an important opportunistic pathogen. This species, like other well-known actinomycetes, such as *Corynebacterium pseudotuberculosis* or *Rhodococcus equi*, is an etiological agent of common suppurative infections in animals. In the case of *C. pseudotuberculosis*, the highest susceptibility to the infection is observed in small ruminants, and abscesses are located mainly in the lymph nodes [25,26]. *R. equi* is a cause of pyogenic infections, mostly in horses, and lesions are found mainly in the respiratory tract [27,28]. Whereas *T. pyogenes* is pathogenic for a variety of animal species, and purulent or necrotic lesions may occur in different host tissues. Interestingly, there were no relationships found between the virulence gene profiles of *T. pyogenes* strains and their origin, a type of infection and a host [13,21,29–31]. However, Ashrafi Tamai et al. [32] reported a significant association between the virulence genotypes and clonal types of *T. pyogenes* isolates, and the severity of the clinical symptoms in postpartum cows with metritis.

Although *T. pyogenes* is a bacterium that has been known of for a long time, many questions concerning the mechanisms of infection pathogenesis, as well as its reservoirs and routes of transmission, still remain poorly understood. The aim of this review is to present the current knowledge about the pathogenicity and virulence potential of this opportunistic animal pathogen.

#### **2. Pathogenicity**

*T. pyogenes* infections occur in both domestic and wild animals worldwide, but are rare in humans. The prevalence of *T. pyogenes* isolation may differ, depending on a host species and a geographic region. The majority of published data concerns *T. pyogenes* infections in food animals and comes from Europe [21,33–38], China [39–43], Japan [44,45], Brazil [30,46], and the United States [47–50]. In livestock, the diseases caused by *T. pyogenes* generate significant economic losses, mainly in cattle and swine breeding, causing a reduction of meat and milk yield, as well as decreased reproductive efficiency and sometimes the necessity to cull diseased animals. The clinical course of these suppurative infections may be severe, with different mortality rates, which increase in the case of misdiagnosis or inappropriate treatment. Beta-lactams, tetracyclines, and macrolides are the antibiotics most often used to treat *T. pyogenes* infections. A question of the therapy efficacy is especially important, as the antimicrobial resistance in *T. pyogenes* becomes an emerging problem because of the common use of these drugs in agriculture [29,32,34,39,40,42,44,47,49,51,52].

Generally, the infections caused by *T. pyogenes* have an opportunistic nature, in which adverse environmental and host-related factors play a relevant role in the disease establishment [45,50,53,54]. Nevertheless, the risk factors of the infection development are sometimes difficult to estimate. Curiously, there are no obvious differences observed between the virulence genotypes of commensal and clinical *T. pyogenes* isolates, although, in some investigations, a gene encoding one of the virulence determinants, fimbria A, was found more frequently in the isolates obtained from infected cows than from healthy ones [49,55]. However, the in vitro study of Ibrahim et al. [56] showed significant differences in the expression level of eight known virulence genes in *T. pyogenes* isolated from the uterus of a cow with clinical endometritis, and the isolate from the uterus of a healthy cow. This indicates the importance of the regulatory mechanisms of the virulence gene expression in the infection development, which can be also demonstrated in an example of the pyolysin gene expression [57]. In other study, a relationship between the clonal types of isolates and their origin, from a diseased or healthy host, was noted [21]. All of these findings suggest that other unknown bacterial factors may also contribute to the establishment and development of the *T. pyogenes* infection.

*T. pyogenes* may cause infection as a primary etiological agent, but more frequently, this species is involved in polymicrobial diseases, such as mastitis [45], uterine infections [55], interdigital phlegmon [58], or liver abscesses [59]. This bacterium may be recovered from a mix infection of various bacterial species, but especially frequently with Gram-negative anaerobes, such as *Fusobacterium necrophorum*, *Bacteroides* spp., or *Peptoniphilus* (formerly *Peptostreptococcus*) *indolicus*. In these cases, purulent and necrotic lesions are usually observed, leading to systemic signs and resulting in animal death. A particularly strong synergistic interaction occurs between *T. pyogenes* and *F. necrophorum* [60]. It consists in the stimulation of *F. necrophorum* growth by a diffusible and heat-labile product of *T. pyogenes*, which probably decreases the oxygen pressure and the oxidation-reduction potential in a site of infection, generating conditions advantageous for this anaerobe [61]. On the other hand, a leukotoxin produced by *F. necrophorum* protects *T. pyogenes* against phagocytosis, because of its ability to lysis leukocytes or to induce their apoptosis, depending on its concentration [59]. Moreover, lactic acid, which is a metabolic product of *T. pyogenes*, can be used by *F. necrophorum* as an energy substrate. Other bacteria, especially *Escherichia coli*, are also often associated with *T. pyogenes* co-infections, mostly postpartum uterine infections [55,62,63]. However, the synergistic effect between both of these bacteria is not evident. The findings of Zhao et al. [64] indicated that *N*-acyl homoserine lactones from *E. coli* and *Pseudomonas aeruginosa*, which act as the quorum-sensing (QS) signal molecules, can inhibit the growth and virulence of *T. pyogenes* in vitro. The recent in vivo study of Huang et al. [65], conducted in a mouse model, confirmed these observations.

#### *2.1. T. pyogenes Infections in Animals*

In livestock, *T. pyogenes* infections occur mainly in cattle, swine, sheep, and goats, rarely in horses or birds, and are often associated with heavy economic losses.

In cattle, *T. pyogenes* mainly causes infections of the reproductive tract and the mammary gland, as well as pneumonia and liver abscessation. The most prevalent diseases in dairy cows related with this bacterium are metritis and endometritis, which may develop in a clinical form in about 23–52% of animals after parturition [29,32,38,39,66]. *T. pyogenes* together with many other bacteria comprises a vaginal biota of healthy cows [67], and may also colonize and persist in the uterus of dairy cows with normal puerperium [21,38]. However, as an opportunistic pathogen, this bacterium can invade the distant parts of the reproductive tract, especially after parturition, when the protective epithelium of the endometrium is disrupted, and it can also increase the influx of inflammatory cells in these tissues [50,68]. The presence of *T. pyogenes* in the endometrium is correlated with the damage of the tissue because of the cytolytic activity of the pyolysin against the endometrial stromal cells, which are particularly sensitive to this cholesterol-dependent toxin [68,69]. The ability of *T. pyogenes* to produce inflammatory lesions in the endometrium was confirmed by the findings of Lima et al. [70]. They noted that after the intrauterine infusion of *T. pyogenes* suspension containing 109 colony-forming units/mL, moderate to severe endometrial inflammation developed in the studied cows, and additionally, premature luteolysis was observed in some animals. The uterine inflammation, usually later postpartum, associated with *T. pyogenes*, has a form of subclinical or clinical endometritis, and may reduce the reproductive performance and milk yield [63,71,72]. Bonnett et al. [73] noted that cows with *T. pyogenes* infection took significantly longer to conceive. Moreover, the study of Boer et al. [63] demonstrated that *T. pyogenes* isolation at day 21 postpartum was associated with the subsequent diagnosis of purulent vaginal discharge, and this observation was confirmed by Sheldon et al. [74]. *T. pyogenes* can also have a lethal impact on the oviductal epithelial cells, as it was demonstrated by Mesgaran et al. [75] in the in vitro study. In general, in the cows infected by *T. pyogenes*, an increased prevalence of clinical endometritis is observed [50]. Frequently, uterine disorders are from co-infections with *T. pyogenes* and other bacteria, such as *E. coli*, *Streptococcus* spp., *Staphylococcus* spp., *Fusobacterium* spp., *Prevotella* spp., and *Clostridium* spp. [39,55,62,71,76–78]. Such polymicrobial uterine infections, especially those with anaerobes, result in an increased purulent secretion and higher severity of lesions [76,79]. The problems regarding parturition, the subsequent negative energy balance, or hyperketonemia are considered to be important risk factors for these diseases [63,80].

Another important and common *T. pyogenes* infection in cattle is mastitis, which may affect lactating and dry cows, as well as heifers [30,37,43,45,46,81–84]. *T. pyogenes* is well known as one of the crucial agents of polymicrobial infection, called summer mastitis, which occurs mainly in pastured cows during the summer, and is associated with pathogen transmission by an insect, *Hydrotaea irritans* [85,86]. However, Madsen et al. [87] demonstrated no differences in the rates of *T. pyogenes* isolation from the mastitis cases in stabled and pastured cattle. Similarly, Ribeiro et al. [46] and Ishiyama et al. [45] did

not observe a seasonality of the *T. pyogenes* mastitis occurrence during their long-term survey of the disease. Moreover, it was reported that *T. pyogenes* alone can cause clinical mastitis called pyogenes mastitis, even with a high severity of symptoms [45,85]. The mammary gland inflammation caused by this bacterium is characterized by severe pyogenic lesions in the mammary tissue, and malodorous and purulent milk, especially in case of co-infection with anaerobes, decreasing the milk yield and the low recovery rate [45]. The anaerobes most often involved in mastitis together with *T. pyogenes* are *P. indolicus*, *F. necrophorum*, and *Prevotella melaninogenica* (formerly *Bacteroides melaninogenicus*). All of these bacteria can colonize the mucous membranes and skin of clinically healthy cattle, but *P. indolicus* and *T. pyogenes* were the most frequently isolated from a teat skin [88]. Regardless of the many investigations on *T. pyogenes* mastitis, little is still known about the factors involved in the establishment and persistence of that infection.

*T. pyogenes* also contributes to many other disorders in cattle, among them, liver abscesses and interdigital phlegmon have a more significant economic impact. Those infections are mixed with anaerobes, mainly *F. necrophorum*, characterized by the synergistic interaction of both of the bacteria mentioned previously [59]. Liver abscesses related to *T. pyogenes* infection occur mostly in feedlot cattle, with a varying frequency from 2% to 80% [59,89–91]. The infection is associated with *T. pyogenes* presence in the rumen wall, in which the primary lesions can form and then bacteria can penetrate by a hepatic portal venous system to a liver, where abscesses form as secondary infection foci [60]. In the case of abscesses rupturing, systemic signs can develop, resulting in animal death. The development of liver abscesses is probably associated with ruminal acute or subacute acidosis, which can induce primary damage in the protective surface of the rumen wall.

Bovine interdigital phlegmon incidence in dairy cows is usually 2–5% per lactation, but some cases of outbreaks were also reported [35,58,92,93]. The role of *T. pyogenes* in the pathogenesis of this polymicrobial disease is not well defined. Kontturi et al. [58] reported that this bacterium is rather a secondary pathogen associated with the healing stage of the infection, while Bay et al. [94] did not detect *T. pyogenes* at all in the interdigital phlegmon investigation using 16S rRNA gene sequencing.

Moreover, *T. pyogenes* may cause a variety of other purulent infections in cattle such as pneumonia, encephalitis, pyelonephritis and kidney abscesses, lymphadenitis, endocarditis, and abscesses of various localization [30,31,36,44,46,95,96]. This pathogen was also isolated from cases of septicaemia and abortion [36,44,46,97].

In swine, *T. pyogenes* is a common agent of pneumonia, pleuritis, endocarditis, osteoarthritis, polyarthritis, mastitis, reproductive tract infections, and septicaemia [13,29,35,44,98–102]. Abscesses—superficial, muscular, or located in different organs—occur frequently, and may lead to the development of systemic purulent infection and inflammation of lungs, liver, kidneys, muscles, bones, joints, or other tissues [13,46,103]. In many cases, these are infection mixed with different microorganisms, as is observed in *T. pyogenes* infections in cattle. Diseases of swine associated with *T. pyogenes* are an emerging clinical, epidemiological, and economic problem, because they usually result in the necessity for the elimination of infected animals from a herd, and the discard of carcasses with suppurative lesions at slaughterhouses [103]. It has been reported that a number of factors can predispose to the development of these disorders, among other viral infections, that cause immunosuppression, for example with porcine reproductive and respiratory syndrome virus (PRRSV) [104].

In small ruminants, *T. pyogenes* is mostly a cause of abscesses formation in different tissues and localized in various parts of the body, including bone marrow and foot (footrot) abscesses [29,31,105–110]. Those lesions differ from that observed in the case of caseous lymphadenitis caused by *C. pseudotuberculosis* [111]. Moreover, *T. pyogenes* can be associated with purulent, mainly polymicrobial disorders in sheep and goats, such as pneumonia, lymphadenitis, arthritis, osteomyelitis, reproductive tract infections, mastitis, and septicaemia [30,35,36,46,112–114].

In other domestic animals, infections related to *T. pyogenes* are rare. It may be connected with the fact that this bacterium does not belong to their normal biota. Therefore, there are only a few data on the occurrence of *T. pyogenes* infections in companion animals. The first reported cases referred to otitis externa in a cat and cystitis in a dog [115], lung abscess in a dog [116], and feline pyothorax [117]. Moreover, *T. pyogenes* was isolated from cases of wound infection, abscesses, vaginitis, pneumonia, and encephalitis in dogs [30,35,36,46]. Recently, an interesting case of *T. pyogenes* and *Brucella abortus* co-infection in a cat and a dog was described by Wareth et al. [118]. Both animals lived on a dairy cattle farm, where cases of abortion and mastitis in cows were noted. The bacteria were isolated in a mixed culture from the uterine discharge of a bitch after abortion and a cat with pyometra.

Infections related to *T. pyogenes* were noted sporadically in horses, and included single cases of metritis, orchitis, mastitis, septicaemia, umbilical infection in foals, abscesses, and wound infection [30,36,46,119].

Furthermore, two cases of suppurative disorders in rabbits caused by *T. pyogenes* were published. In the first case, this bacterium was isolated from lung lesions, and in the second one, from necrotic foci in liver, spleen, lung, and brain [36,120].

The incidence of *T. pyogenes* infections in birds seems to be very low. Some cases of clinical lameness and osteomyelitis in turkeys were reported [121,122]. A unique case of liver abscesses in pigeons was described by Priya et al. [123].

In wildlife, *T. pyogenes* may be associated with many types of purulent disorders occurring in free-living and captive animals. *T. pyogenes* infections were reported the most frequently in ruminants and other herbivores, in which the bacteria were also found as a resident microbiota of the skin and mucous membranes of respiratory and urogenital tracts [20,54,124].

In the United States and Canada, *T. pyogenes* is an emerging pathogen of captive or free-ranging white-tailed deer (*Odocoileus virginianus*), which cause intracranial abscessation—suppurative meningoencephalitis disease complex (the intracranial abscess disease) that occurs with a variable frequency, depending on animal populations [54,124–127]. *T. pyogenes* may be also involved in pneumonia and necrobacillosis occurring in this animal species [128,129].

Furthermore, *T. pyogenes* was also isolated as an etiological agent of chronic purulent infections, including keratoconjunctivitis, brain and foot abscesses, in other cervids, such as Key deer (*Odocoileus virginianus clavium*) [130], fallow deer (*Dama dama*) [53], roebuck (*Capreolus capreolus*) [131], red deer (*Cervus elaphus*) [132], and mule deer (*Odocoileus hemionus*) [133]. *T. pyogenes* is a prevalent cause of variably located abscesses in forest musk deer (*Moschus berezovskii*), noted mainly in farm animals [40,134].

In European bison (*Bison bonasus*), *T. pyogenes* is associated with abscesses of the liver, spleen, lungs, lymph nodes, skin, and especially with urogenital tract infections [135]. In female bison, abscesses of various localizations were observed. In male bison, *T. pyogenes* is considered to be one of the etiological agents of a chronic necrotizing and ulcerative inflammation of the prepuce and penis (balanoposthitis) [135].

Additionally, *T. pyogenes* infections were reported in antelopes [29,136,137], an okapi (*Okapia Johnstoni*) [138], bison (*Bison bison*) [139], Bighorn sheep (*Ovis canadensis*) [140], a chamois (*Rupicapra pyrenaica*) [112], camels [141–144], an elk [36], and reindeer [145,146]. Moreover, some other sporadic cases of infectious diseases associated with *T. pyogenes* were described in Grey Slender lorises (*Loris lydekkerianus nordicus*) [147,148], macaws, and elephants [48], as well as in reptiles, a bearded dragon, and a gecko [149].

#### *2.2. T. pyogenes Infections in Humans*

Infections caused by *T. pyogenes* in humans are sporadic, mostly occur in immunosuppressed patients, and are connected with occupational exposure, especially relating to contact with farm animals and their environment [150]. Published data concerning *T. pyogenes* infections in humans are limited, and include, among others, reports on endocarditis [151–155], endemic leg ulcers [156,157], pneumonia [158], arthritis [159,160], sepsis [161], and various purulent lesions and abscesses [162]. As *T. pyogenes* was never demonstrated as a commensal microbiota of humans, those infections should

be considered as zoonotic diseases [48]. Although, a probability of animal-to-human transmission of this pathogen has not been well estimated and confirmed.

#### **3. Reservoir, Transmission, and Routes of Infection**

Most actinomycetes are widespread in the natural environment, being found in various ecological niches as saprophytes, but they may also constitute a commensal biota of humans and animals, and some of them can be opportunistic pathogens [163,164]. Curiously, there is no published characterization of environmental *T. pyogenes* isolates, for example from soil, which might be considered non-pathogenic, as it is known, for instance, in case of environmental strains of *R. equi*, a typical soil opportunistic pathogen [165]. Although there are no data on *T. pyogenes* ability to replicate in soil or water, certainly, it can persist for some period in such environmental conditions. For example, *C. pseudotuberculosis*, the related actinomycete, can survive for several weeks in the environment [166]. Considering the nutritional requirements of *T. pyogenes*, it should be assumed that the bacterium can replicate only in the environment rich in peptides, fermentable carbohydrates, inositol, and hemin, and at the temperature range of 20 to 40 ◦C [4]. Therefore, it seems that a main reservoir and source of this bacterium are animals of various species.

Little is known about the dissemination of *T. pyogenes* infections, and about the transmission of the pathogen. It is suggested that the majority of infections have an endogenous character, as the bacteria are a common component of the skin and mucous membrane biota [48]. However, the possibility of exogenous infections should also be considered, because bacteria may be transmitted, for instance, by contaminated husbandry utensils and equipment, or directly from animal to animal [30,46]. The natural environment contaminated by *T. pyogenes* is suggested to be an important source of the bacteria, especially in the case of mammary or foot disorders in domestic and wild animals [46,53,109]. In addition, climate conditions, such as a high humidity and mild temperature, are factors favouring the infection occurrence [53]. In the case of summer mastitis, *T. pyogenes*, exceptionally, may be transmitted between animals by biting flies [86]. However, a possibility of tissue penetration by these bacteria through a micro-trauma caused by ectoparasites, such as ticks, should be also considered. The preliminary results of the study on ticks collected from the skin of European bison indicated a potential contribution of those arthropods to *T. pyogenes* transmission [167].

The observations from different studies indicate a potential threat of *T. pyogenes* transmission from wild to domestic animals, or the other way around [53,135]. It is very possible, regarding the frequent co-occurrence of wildlife species and livestock animals on the same agricultural areas, such as pastures and meadows, which can serve as a reservoir for this pathogen.

The data on the genetic relationship of *T. pyogenes* isolates occurring in a particular host population, and on the dissemination of strains amongst animals and the environment are lacking. The results of a few epidemiological studies on bovine and swine isolates are available [10,13,21,32,148]. In some of them, an association between the clonal types and the development of clinical infection by a strain was shown [21,32,148], while in other investigations, such a correlation was not found [10,13]. A variety of molecular methods, characterized by various discriminatory powers, were used for *T. pyogenes* differentiation, which enabled the phylogenetic analysis of strains to be performed, for example the 16S rRNA gene sequencing [135,149], single enzyme amplified fragments length polymorphism (SE-AFLP) [13], BOX-PCR [21,32], the superoxide dismutase A gene (*sodA*) sequencing [36], random amplification polymorphic DNA (RAPD-PCR), and multilocus sequence analysis (MLSA) [148]. The usefulness of some of them was proved by Nagib et al. [148] during the study on *T. pyogenes* isolates from lorises, which showed a close relationship, indicating their clonal origin. Unfortunately, the gold standard method used for genomic fingerprinting, pulse field gel electrophoresis (PFGE), was never reported as a technique applied for *T. pyogenes* phylogenetic investigation.

The routes of *T. pyogenes* infections are frequently difficult to establish. The *T. pyogenes* infections develop mainly as a consequence of the mechanical injuries of skin and mucous membranes [48]. Wounds and abrasions of skin are common routes of infection, for example, udder injuries in the case

of *T. pyogenes* mastitis, or hoof injuries in the case of interdigital phlegmon. In the case of mucous membranes damage, such as in metritis, the bacteria can invade deeper tissues after endometrial epithelium, disruption during parturition. Another described route of infection is associated with improperly performed surgical procedures, such as castration or intramuscular injection [103]. Bacteria can also invade tissues through micro-injuries of the skin caused by some arthropoda, like flies or ticks, as it has been mentioned above. In deer, behaviour may influence the pathogen transmission and susceptibility to cranial/intracranial abscess infection, which occur more frequently in adult males [125,127]. Interestingly, Belser et al. [54] noted a higher prevalence of *T. pyogenes* along the forehead of the males of white-tailed deer than the females, when the rate of the pathogen carriage on the nasopharyngeal mucosa was the same for both sexes. Therefore, it is suggested that the disease development may be a result of cuts or abrasions arising from antler rubbing on trees or from sparring between males.

#### **4. Pathogenesis of** *T. pyogenes* **Infection**

The pathogenesis of microbial infection is determined by a variety of bacterial and host-related factors. In the case of *T. pyogenes*, the pathogenicity is attributed to the determinants, which induce the formation of abscesses, empyemas, and pyogranulomatous lesions. However, pathogen–host interactions in *T. pyogenes* infection are still poorly understood. Some known and putative bacterial factors contribute to the development of *T. pyogenes* infection, but their role in the infection pathogenesis remains insufficiently explained. Moreover, many researchers highlighted a high phenotypic and genotypic diversity among the studied *T. pyogenes* isolates of various origins, also taking into consideration virulence genotypes [13,21,29–31]. This finding indicates the genetic heterogeneity in the studied *T. pyogenes* populations. On the other hand, Cohen et al. [124] detected significantly more virulence genes in *T. pyogenes* isolates from populations of white-tailed deer males, in which a high incidence of the intracranial abscess disease was observed, compared with those detected in the isolates from apparently healthy animals in populations in which the disease was not noted. This finding suggests that the essential differences in the pathogenic potential may exist among *T. pyogenes* strains, depending on their origin.

#### *4.1. T. pyogenes Virulence*

Only few virulence factors in *T. pyogenes* are recognized to date (Figure 2). They include pyolysin (PLO), the only known toxin of this bacterium; some adhesive factors, such as fimbriae neuraminidases and extracellular matrix-binding proteins; different exoenzymes, such as serine proteases with gelatinase and caseinase activity, or DNAses; as well as the ability to invade host cells and to create biofilm formation [48]. The significance of some of them in the pathogenicity is not clear enough. In general, profiles of genes encoding particular virulence determinants are not correlated with a type of infection and host species. However, a few cases of association between the presence of some determinants and a type of infection were reported [49,55].

**Figure 2.** Schematic diagram of known virulence factors of *T. pyogenes*. Abbreviations: CpbA—collagen-binding protein; Fbp—fibrinogen-binding protein; Fnbp—fibronectin-binding protein; NanH—neuraminidase H; NanP—neuraminidase P; PLO—pyolysin.

#### 4.1.1. Pyolysin

PLO is considered to be both a major virulence factor of *T. pyogenes* and a host-protective antigen [168–170], and, until now, the gene (*plo*) encoding this protein was detected in all wild-type *T. pyogenes* strains. This is an exotoxin belonging to the cholesterol-dependent (also called cholesterol-binding) cytolysins (CDCs), which are a family of 51- to 60-kDa single-chain proteins produced by many species of Gram-positive bacteria, for example *Streptococcus pneumoniae* (pneumolysin), *Streptococcus suis* (suilysin), *Streptococcus pyogenes* (streptolysin O), *Paenibacillus alvei* (alveolysin), *Clostridium perfringens* (perfringolysin O), or *Listeria monocytogenes* (listeriolysin O). [171,172]. PLO is the most divergent protein of this group, presenting only 38% to 45 % identity and 58% to 64% similarity to other toxins from the CDC family, when the conserved core of the CDCs (corresponding to amino acid fragment 38–500 of perfringolysin O) is taken into consideration during the analysis [172]. The highest identity of PLO is shown to suilysin, novyilysin, botulinolysin B, and tetanolysin O, which range from 45% to 44% [172]. This protein was purified and characterized for the first time by Ding and Lämmler in 1996 [173], however, its toxic and haemolytic activities were previously reported by Lovell [174]. As a member of the CDC family, PLO displays a cytotoxic effect on a variety of host cells, for example, erythrocytes, polymorphonuclear neutrophils (PMNs), macrophages, epithelial cells, fibroblasts, and endometrial stromal cells. [68,168,171,172]. The cytolytic activity of PLO is associated with its ability to bind to the plasma membrane and to form transmembrane pores, which is a common feature of CDCs [168,172]. The in vitro study showed that PLO is able to cause the lysis of human, sheep, horse, rabbit, and guinea pig erythrocytes [4]. Specific antibodies against PLO completely neutralize its haemolytic activity, suggesting that PLO is the sole haemolysin produced by *T. pyogenes* [168].

The significance of PLO as a primary virulence factor of *T. pyogenes* was confirmed in many in vitro and in vivo experiments, with use of PLO-deficient mutants or recombinant proteins [168,175,176]. PLO has a lethal effect on mice and rabbits, and a dermonecrotic effect on guinea pigs after intravenous and intraperitoneal injection [168,174,177–179]. Jost et al. [168] observed a loss of haemolytic activity in the PLO-1 mutant strain as a result of the insertional inactivation of the plo gene, as well as its 1.8-log10 reduction in virulence for mice in comparison to the wild-type strain. They also, for the first time, documented the importance of PLO for the in vivo survival of bacteria, studying the effect of the co-challenge of a wild-type strain, and a mutant strain with the inactivated *plo* gene in a mouse model. Similarly, Zhao et al. [179] demonstrated that the *T. pyogenes* strain with an in vitro higher expression of the *plo* gene was more virulent for mice in the in vivo study than the strain with a lower *plo* expression, designated as avirulent.

Billington et al. [180] cloned and sequenced the *plo* gene encoding PLO, which was located in an open reading frame (ORF) of 1605 bp in the *T. pyogenes* chromosom. The *plo* gene sequence was also described by Ikegami et al. [181], for which the deduced pyolysin sequence showed a 99.4% similarity and a 97.6% identity to the PLO sequence previously reported by Billington et al. [180]. A consensus ribosome binding site, two promoter sequences (similar to the *E. coli* σ<sup>70</sup> promoter), and three direct repeats (DR1 to D3 with sequence ATTTTTG(C)TGG) were found upstream of the *plo* gene, and a transcriptional terminator region was found downstream of this gene [180]. Moreover, Rudnick et al. [170] found that the *plo* gene, along with the mentioned sequences and ORF (*orf121*), encoded a 13.4 kDa protein of an unknown function, form a genomic islet of 2.7 kb, characterized by a reduced content of G+C (50.2%). This islet, flanked by two housekeeping genes, *smc* and *ftsY* (62.5% G+C), is located in the *T. pyogenes* chromosome (Figure 3) [170]. Interestingly, the codon usage of *plo* and *orf121* differ from that of the genes flanking the islet. The genes *smc* and *ftsY* are common in many other bacteria, mainly Gram-positive, and are essential for *T. pyogenes* viability [170]. Downstream of the *ftsY* gene are the ff*h* gene, and then the ORF, *orf353*. The *ftsY* and ff*h* genes encode the FtsY and Ffh proteins, respectively, which have a high similarity to the signal recognition particles from the other bacterial species [170]. The *ftsY*, ff*h*, and *orf353* genes in *T. pyogenes* are located in close proximity, and form an operon-like arrangement that is unusual in most other bacteria. It seems that, as in other species, these genes are needed for bacterial growth.

**Figure 3.** Schematic presentation of the *T. pyogenes* chromosomal region containing the *plo* gene and the surrounding genes. P1 and P2 indicate the positions of two promoters; DR 1–3 indicate the positions of three direct repeats. The scale is not designated.

The differences between the islet and the flanking genes suggest a possibility of horizontal transfer of the *plo* gene, though any integrase or transposon sequences were not found in this chromosomal region. Taking into consideration the fact that the *plo* gene is present in all wild-type *T. pyogenes* strains, including commensal isolates, it seems probable that *plo* was inserted into the intergenic region between *smc* and *ftsY* by homologous recombination; it has to be retained, as both of these genes are obligatorily required for bacterial growth. In addition, it was confirmed by Rudnick et al. [170] that the *plo* islet is a conserved region of the *T. pyogenes* chromosome found in all of the studied isolates of various origins.

*T. pyogenes* is an opportunistic pathogen that, being a part of the host microbiota, has no damaging effect on its organism. Therefore, it is suggested that the regulation of the PLO expression may be critical for the establishment of the status of this bacterium as a pathogen or a commensal. Rudnick et al. [170] reported that the *plo* gene expression is not regulated in vitro by *orf121* located upstream of this gene. In their subsequent investigation, two promoters, P1 and P2, located in the *plo* islet, were identified as the regulatory sequences controlling the transcription of *plo* [57]. It was observed that the in vitro haemolytic activity of *T. pyogenes* was in growth phase depending on a peak in the early stationary phase [57,173]. It indicated that *plo* is up-regulated during that phase, and indeed a significant increase in the *plo* specific mRNA was noted. Thus, the suggestion that the PLO expression is controlled at the transcriptional level was confirmed [57]. It was also shown that the P2 promoter is predominant and highly active during the stationary phase of the in vitro growth of bacteria, while the P1 activity is weak. Besides the P1 and P2 promoters, direct repeats (DRs) located in the *plo* promoter region may be involved in the regulation of the *plo* gene transcription. The study of Rudnick et al. [57]

showed that DRs, especially DR3, may function as binding sites for a soluble *T. pyogenes* factor being a transcriptional regulator. However, regarding AT-rich sequences of DRs, which can easily bend, it was supposed that DRs may also act as activators of transcription, by promoting RNA polymerase binding to the promoter.

The predicted, on the basis of the *plo* gene sequence, length of the PLO molecule is 534 amino acids (aa), and then, this protein should have a molecular mass of 57.9 kDa [180]. A signal peptidase cleavage site was found between 27 and 28 aa, indicating that the mature PLO molecule has a weight of 55.1 kDa. According to the Funk et al. [182] study, PLO is an oxygen-stable protein, heat-labile (destroyed at 56 and 100 ◦C), destroyed by pH 3 and 11, and id sensitive to treatment of protease, trypsin, and amylase.

In contrast to the majority of CDCs, the PLO activity is insensitive to thiols or other reducing agents, in other words, PLO does not require thiol activation [180]. This difference is associated with the divergence within the amino acid sequence of undecapeptide (491–501 aa) of PLO. The undecapeptide is a highly conserved region of each CDC, located near the C terminus of the protein. In some CDCs, including PLO, amino acid substitutions occur in this region [180,183]. In case of PLO undecapeptide (EATGLAWDPWW), a unique cysteine residue (C491), responsible for the thiol-activated nature of CDCs, is replaced with an alanine (A492). This substitution has no essential effect on the haemolytic activity of PLO, but is related to the oxygen stability of this cytotoxin [184]. Moreover, other changes in the PLO undecapeptide were noted, such as the insertion of a proline residue (P499) and the deletion of an arginine residue at the end of the CDCs undecapeptide. It seems that these changes have important effects on the conformation and charge of the undecapeptide [180]. The proline insertion has the greatest effect on the PLO activity, as the deletion of this amino acid resulted in a lack of haemolytic activity [183]. The same effect on the PLO haemolytic activity was observed in the case of the removal or substitution of any of the three tryptophan residues [183]. Therefore, as it was previously demonstrated by Billington et al. [180], the amino acid sequence of the PLO undecapeptide is required for full cytolytic activity.

The cytolytic activity of PLO, like other CDCs, is suppressed by free cholesterol in a concentration of 1μg/mL [180]. However, the presence of cholesterol in the target plasma membrane is absolutely necessary for pore formation by CDCs. The role of the undecapeptide in the initial plasma membrane binding seems to be crucial, in particular, the first tryptophan residue (W497) is important. Although, Billington et al. [183] suggested that other fragments of PLO might also participate in recognition and binding to the host cell membranes.

The known secondary and tertiary structures of some CDCs shows a significant similarity [172]. A crystal structure of the PLO molecule presumably is homologous to the structure of other CDCs. The tertiary structure model of PLO, based on the perfringolysin O (PFO)—PLO sequence alignment—was proposed by Pokrajec et al. [185]. The monomeric PLO molecule is rich in β-sheet elements and consists of four domains (D1 to D4), of which D2 and D3 are packed against each other, while D1 is located at the *N*-terminal and D4 at the C-terminal end of the molecule (Figure 4).

Domain 1 contains a-helix and β-sheet elements. The precise role of D1 in the pore-forming process remains undefined. Zhang et al. [186], studying an effect of the replacement of aspartic acid (D238) in this domain with arginine, showed an important role of D1 in maintaining the pore-forming activity of PLO. The investigation of Imaizumi et al. [177] indicated that domain 1, especially the region of 55–74 amino acids, is required for the haemolytic activity of PLO. Yan et al. [187] also confirmed the importance of this region of domain 1, as they demonstrated that the substitution of each amino acid from 58 to 62 in the PLO molecule, particularly isoleucine 61, resulted in the complete loss of its haemolytic activity. Moreover, six B-cell linear epitopes in D1 were identified by Yang et al. [188].

Domain 2 of the molecule forms a long, curved single layer of anti-parallel β-sheet, while the structure of domain 3 is more complex and consists of five strands of anti-parallel β-sheet surrounded by helical layers and short α-helices at either side of the central β-sheet. Domains 2 and 3 are involved in the monomer oligomerization and insertion into the plasma membrane. The separation of D2 and D3, as well as the structural collapse of D2, are required during these processes [189]. The conversion of the D3 short α-helices to two transmembrane β-hairpines starts the molecule insertion into the bilayer, and is a crucial step of the transmembrane β-barrel formation [172].

**Figure 4.** Simplified model of the pyolysin molecule, based on the perfringolysin O-structure.

Domain 4 of the PLO molecule is the only contiguous domain characterized by a β-sandwich structure, connected by a single peptide with the rest of the protein. As it has been mentioned above, the conserved undecapeptide sequence, rich in tryptophan residues, is located at the tip of this domain. This part of the PLO is responsible for host plasma membrane recognition and cholesterol binding [185,190]. However, Pokrajac et al. [185] also demonstrated the ability of D4 to self-oligomerization, but not pore formation. Their findings indicate a greater functional role of this PLO domain in cytolysin activity, compared with the previously described CDCs.

The molecular mechanism of pore-forming by PLO and other CDCs seems to be very similar, and is best described for PFO [172,185]. PLO molecules are secreted to the extracellular environment, probably via the general secretion system, as in the most CDCs [172]. In a form of water-soluble monomers, they bind to the cholesterol-containing areas in the plasma membrane of the eukaryotic host cells [183]. Binding to cholesterol by the PLO domain 4 initiates the oligomerization of the monomers and the formation of incomplete arc-shaped structures, called th pre-pore. Subsequently, the PLO oligomers are inserted into the bilayer by the transmembrane β-hairpins of domain 3, which leads to the formation of a large transmembrane β-barrel pore, which is <50 nm in diameter and protrudes <7 nm above the plasma membrane surface [191]. As a consequence, this process leads to th eleakage of ions and other cytoplasmatic molecules, and finally to the lysis of the host cell.

It seems that a predominant assignment of the cytolytic activity of PLO is to enable access to free iron and other growth factors essential for the bacterial replication contained in the host cells. Furthermore, the ability to lysis phagocytes protects the bacteria against the host immune response. On the other hand, at lower concentrations, PLO can modulate the host immune response.

Despite a wide knowledge about the molecular structure and function of PLO, the precise role of this toxin in the pathogenesis of particular *T. pyogenes* infections, characterized by different course and clinical manifestations, remains unclear.

#### 4.1.2. Fimbriae

The occurrence of fimbriae in Gram-positive bacteria is a rarity. *T. pyogenes* is just one of the exceptions [48]. *T. pyogenes* fimbriae have a filamentous structure, at 200–700 nm in length and 2.5–4.5 nm in width, and probably occur in a limited number on bacterial cells (less than 10 per cell) [48,192]. In *T. pyogenes*, five fimbriae were described, FimA, FimB, FimC, FimE, and FimG [48,192]. The presence of other fimbrial subunits can be deduced on the base of the *T. pyogenes* genome sequences, for example FimJ [192,193]. FimA, 45.7 kDa protein, contains a signal peptide, an E box sequence, and is encoded by the *fimA* gene [48]. FimB, 90.5 kDa protein, has a similar structure, but additionally contains a cell wall-sorting domain and a fibronectin-binding domain, and is encoded by the *fimB* gene [48]. Both genes, *fimA* and *fimB*, are located in a fimbrial gene operon, together with the *srtA* gene encoding a putative sortase. The other fimbriae, FimC, FimE, and FimG, are encoded by the genes *fimC*, *fimE*, and *fimG*, respectively.

The expression of the fimbriae in vitro is poor; therefore, it is difficult to characterize their properties, thus only their genes or mRNAs are better known. Liu et al. [192] investigated the in vitro expression of FimA, FimC, and FimE. However, even though different culture conditions were used, only the FimE expression could be detected. Zhao et al. [179] studied the expression of the virulence factors of *T. pyogenes*, including FimA and FimC, in vitro and in vivo on a mouse model. Their results showed that in in vivo conditions, the *fimA* gene was earlier expressed than *fimC*, which suggests that FimA is a dominating fimbria in *T. pyogenes*. This is in accordance with the findings of many studies on fimbrial genes distribution that demonstrated a high prevalence of *fimA* among various isolates, whereas the genes of other fimbriae were detected with different frequencies [13,21,29–31].

A role of fimbriae in the *T. pyogenes* infection is associated with the bacterial adherence to host cells, but the precise mechanisms of this interaction are not well understood. Taking into consideration that *T. pyogenes* can be a commensal as well as a pathogen, it seems that fimbriae are equally important for bacteria in both of these forms. Likewise, in the case of infections caused by many other bacteria, *T. pyogenes* fimbriae are required for cell adhesion and the colonization of host tissues [49,55]. Liu et al. [192], based on their study results, speculated that the poor expression of FimA can limit the formation of other fimbriae, and that the up-regulation of this fimbria production may be associated with *T. pyogenes* pathogenicity.

#### 4.1.3. Extracellular Matrix-Binding Proteins

*T. pyogenes* also expresses other proteins, which determine the adhesive properties, and which are associated with the cell wall of the bacteria. These proteins probably also promote adhesion and enhance colonisation through the ability to bind extracellular matrix (ECM) compounds [48]. In the case of Gram-positive bacteria, these proteins belong to the microbial surface components recognizing the adhesive matrix molecules (MSCRAMM) family [194]. Firstly, *T. pyogenes* shows a fibrinogen-binding activity, however, it was inhibited by proteases [195]. In addition, the 20 kDa fibronectin-binding protein associated with the *T. pyogenes* cell wall was also detected, but not well characterized [48].

Collagen-binding protein A (CbpA) represents the MSCRAMM family, and is one of the better known extracellular matrix-binding proteins produced by *T. pyogenes* [48,194,196]. This 121.9 kDa surface protein is encoded by the chromosomal *cbpA* gene, and displays a 50.4% similarity to Cna, the collagen adhesin of *Staphylococcus aureus*. CbpA binds almost all types of collagen, but not other ECM. However, Pietrocola et al. [194] reported that CbpA can bind fibronectin, but in a distinct subsite than that recognized by collagen. This protein is involved in the adherence of bacteria to epithelial and fibroblast cell lines [196]. It may be supposed that the *T. pyogenes* strain, which produces CbpA, has higher potential to colonize collagen-rich tissues. Although, the *cbpA* gene was found in many isolates of various origins, with different frequencies [13,21,29–31].

#### 4.1.4. Neuraminidases

Neuraminidases (sialidases) are one of the most important enzymes of the sialic acid catabolism. Neuraminidases cleave the terminal sialic acid residues from complex glycoproteins, glycolipids, and carbohydrates from the host cell receptors. Neuraminidases are produced by many species of Gram-negative and Gram-positive bacteria or viruses [197]. Bacterial neuraminidases are the main factors for promoting adhesion by exposing the cryptic host cell receptors. In addition, neuraminidases can promote tissue colonisation by reducing mucus viscosity. These compounds also weaken the host's immune response by exposing IgA particles, making them more susceptible to bacterial proteases [48,197–199]. Bacterial neuraminidases are secreted as extracellular enzymes or are anchored to the cell wall. While in Gram-negative bacteria, neuraminidases are usually secreted, neuraminidases of the Gram-positive bacteria are commonly in a cell-associated form [200–204].

The first note about the 50 kDa extracellular neuraminidase of *T. pyogenes* was reported by Schaufuss and Lämmler [205]. Afterwards, two neuraminidases expressed by *T. pyogenes* were better characterized [206,207]. Neuraminidase H (NanH) and neuraminidase P (NanP) are 107 kDa and 186,8 kDa proteins, respectively. These neuraminidases are attached to the *T. pyogenes* cell wall, which results from the presence of specific anchoring motifs in their molecules (LPxTG—for NanH; LAWTG—for NanP). NanH and NanP proteins contain the conserved catalytic RIP/RLP motif (Arg-Ile/Leu-Pro) and five copies of the Asp box motif (Ser-X-Asp-X-Gly-X-Thr-Trp), commonly occurring in the bacterial neuraminidases [206,207]. NanH is the most similar to the neuraminidase from *Actinomyces viscosus* (61.8% similarity and 31.2% identity) [206]. NanP shows a similarity to the neuraminidase from *Micromonospora viridifaciens* (61.6% similarity and 45.3% identity) [207]. However, some similarity between NanH and NanP (53.8% similarity and 38.8% identity) was also demonstrated [48].

*T. pyogenes* strains can produce neuraminidases H and P, encoded by the *nan*H and *nan*P genes, respectively [206,207]. The occurrence of these genes is characteristic of the majority of *T. pyogenes* isolates from cattle [21,29–31,36,206–208] and swine [29,30,36]. However, the *nan*H and *nan*P genes were found at high rates in the bovine isolates studied by Zastempowska and Lassa [37]. The genes encoding neuraminidases are detected with various frequencies in the *T. pyogenes* isolates from small ruminants, such as goats and sheep [29–31,36], and in the forest musk deer isolates [134], while the presence of these genes among the isolates from European bison is much less common [29,135]. In several studies, the *nan*P gene was detected more frequently than the *nan*H gene in the tested isolates [30,135,208]. However, Rogovskyy et al. [31] reported that most of *T. pyogenes* isolates from small ruminants carried mainly the *nanH* gene.

*T. pyogenes* neuraminidases play an important role in adhesion to the host cells, as it was shown in the study with epithelial cells [207]. Moreover, *T. pyogenes* neuraminidases are probably necessary for the colonization of host tissues during the early phase of infection, but they are not essential during further stages of a disease development [48]. The different prevalence of the genes encoding neuraminidases in the *T. pyogenes* isolates from various origins, suggests that other adhesion factors may be also produced by this species.

#### 4.1.5. Biofilm

The ability to perform biofilm formation is a well-known feature of microorganisms, which promotes their resistance to many disadvantageous factors, for instance antimicrobials, as well as increases the adherence properties and enables better protection against the immune response of the host [48]. *T. pyogenes*, as with many other bacteria that are resident on the mucous membranes of a host, is able to perform biofilm formation. Biofilm production was noted in a majority of the studied *T. pyogenes* isolates from different animal species and various types of infections [43,134]. Biofilm formation by *T. pyogenes* is controlled by a two-component regulatory system PloS/PloR, where PloR up-regulates the expression of biofilm [64,179,209]. In the case of polymicrobial infection, a biofilm formation by *T. pyogenes* may be inhibited by the QS molecules produced by other bacteria, for example *E. coli* [64,210]. As in some cases biofilm production may promote the infection development, it seems important to find a therapeutical option for reducing this bacterial property. Da Silva Duarte et al. [211] evaluated the use of the *E. coli* phage UFV13 for the disruption of the *T. pyogenes* biofilm, and they obtained a significant decrease of biofilm formation caused by this phage. This finding indicates a necessity of further studies on this issue.

#### 4.1.6. Regulation of Virulence Factor Expression

Jost and Billington [48] suggested the existence of a two-component system comprising a sensor histidine kinase (HK) and response regulator (RR) proteins in *T. pyogenes*. These signalling systems are common among bacteria, and they enable the detection of an environmental or cellular signal that leads to an appropriate cellular response [212,213]. Zhao et al. [64] determined a LuxR-type two-component regulatory system in *T. pyogenes*, named PloS/PloR, using a comparative transcriptome analysis between two *T. pyogenes* isolates from musk deer with different haemolytic activities. The results of the transcriptome analysis showed the presence of ten typical DNA-binding response regulators (RRs) in the *T. pyogenes* genome. However, a protein crystal structure analysis revealed that only one among the identified DNA-binding RRs had a structure similar to that described previously by Jost and Billington [48]. This RR, called PloR, had a CheY-type receiver domain and a C-terminal LuxR-type HTH DNA-binding domain, while a cognate sensor HK was named PloS. Zhao et al. [64] demonstrated that the expression level of the *ploR* gene is correlated with the expression level of the *plo* gene, indicating PloR as an activator of *plo*, which up-regulates PLO production. On the other hand, it was observed that PloR can down-regulate the expression of *T. pyogenes* proteases [48]. In addition, it seems that PloR is able to up-regulate the biofilm formation and probably the expression of the fimbriae in *T. pyogenes*.

The function of the two-component system in *T. pyogenes* was also investigated in vitro using the *N*-acyl homoserine lactones of *P. aeruginosa* and *E. coli* as QS signal molecules, which can bind to the upstream sensor HK, PloS, and via the PloS/PloR system, may regulate the virulence of bacteria by the inhibition of their growth and biofilm production [64]. Huang et al. [65], studying an effect of *N*-acyl homoserine lactones on the *T. pyogenes* virulence in a mouse model, confirmed that those QS molecules inhibited the expression of *plo*, *ploR*, and *ploS*, and increased the survival rate of mice. In conclusion, the function of the two-component system PloS/PloR and the QS system seems to have the essential significance for the modulation of the *T. pyogenes* virulence.

#### *4.2. Induction of Host Defense Mechanisms by T. pyogenes*

The pathogenicity of *T. pyogenes* is assigned mainly to PLO, a cholesterol-dependent cytolysin, which promotes cell lysis and is responsible for altering the host cytokine profile. This toxin also exerts cytolytic, dermonecrotic, and lethal effects on numerous types of cells, including immunocompetent and non-immune cells [176,186]. It is responsible for the complement cascade induction; modulation of the host cytokine profile; inhibition of respiratory burst; and bactericidal activities in neutrophils, monocytes, and macrophages. These features lead to the modulation of the immune response and to evasion from a host's immune mechanisms. Because of the presence of neuraminidases NanH and NanP, *T. pyogenes* is able to adhere to the alimentary and respiratory epithelial mucosa, increasing the mucus viscosity that facilitates bacterial colonization and diminishes the host immune response, including damage to a first line of defence—destruction of IgA by bacterial proteases. Those serine proteases, as it was mentioned above, are involved in the invasion and destruction of tissues [48,196].

For many years, the assessment of PLO involvement in the development of the disease was conducted using in vitro (established cell lines or primary cell cultures) and in vivo (mouse and large animal) models [48,68,69,75,168,186].

One of the first antibacterial mechanisms of immune response is inflammation, resulting in the recruitment of neutrophils, monocytes, and macrophages to a site of infection. The involvement of those cell populations leads to the phagocytosis and killing of bacteria, clearance of infection, and stimulation of adaptive immune response. Still, *T. pyogenes*interactions with host phagocytic cells strongly influence the outcome of the disease. Using macrophage established cell line J774A.1, it was clearly shown that this pathogen was easily phagocytosed, and could survive within macrophages up to 72 h, with the survival rate diminishing over time. Probably, this is caused by the presence of antibiotics in a medium used for culturing infected macrophages. Once bacteria exit the infected cell, they are killed by the antibiotic present in the medium. Similar results were obtained when the macrophage cell line RAW264.7 was infected with either *T. pyogenes* ATCC 19411 strain or *T. pyogenes* European bison isolate. In this study, phagosome formation, intracellular bacteria survival, and changes in the mitochondrial network and distribution were established at 2, 6, and 24 h post infection (pi) [214]. The formation of phagosomes was noticed as early as 2 h pi, and they were still present within the cells up to 24 h pi, with bacteria either within the phagosome or in the cytoplasm of the infected cells, and the recovery of viable bacteria was possible up to 24 h pi (Figure 5). However, the mitochondrial network morphology and distribution was not visibly altered in the RAW264.7 cells infected with either the ATCC 19411 strain or European bison isolate of *T. pyogenes* when compared with the uninfected control cells. Further investigations are required in order to elucidate the mechanism of *T. pyogenes* pathogenesis. Nonetheless, these data are in agreement with the observations done on bovine polymorphonuclear neutrophils (PMNs) infected with *T. pyogenes*, in which increased phagocytosis was noticed. As this pathogen can survive within phagocytic cells, increased phagocytosis is strongly connected with bacterial dissemination throughout the host [195]. Studies done in a mouse model clearly showed that PLO is responsible for peritonitis [168].

One of the common, most recognized infections in dairy and beef cattle are uterine infections caused by *T. pyogenes*. They result from the recruitment of neutrophils, from monocytes, and from the induction of a proinflammatory response, accompanied with formation of mucopurulent discharges within the infected uterus. This was confirmed on a large animal model. The uterine endometrium of healthy cows was scratched with a cytobrush (damage to the epithelial layer), and subsequently, an intrauterine administration of 5 <sup>×</sup> <sup>10</sup><sup>8</sup> cfu of *T. pyogenes* was performed. Mucopurulent uterine discharge was evident in the challenged animals, and was associated with the isolation of this pathogen [68].

However, the ability of cows to defend against *T. pyogenes* infection depends on the interaction between the invading pathogen and the host's innate immunity. An application of bovine endometrial cell culture in an in vitro model elucidated the PLO involvement in tissue damage and the development of the endometrium pathology of the postpartum period. Recombinant PLO (rPLO) and bacteria-free filtrates (BFF) of *T. pyogenes* had a similar haemolytic activity on the target cells. Still, the stromal cells and epithelial cells expressed a different level of susceptibility concerning treatment with those factors. The endometrial stromal cells were markedly more sensitive than the epithelial cells. This haemolytic effect was abrogated when the endometrial cells were treated with BFF derived from *T. pyogenes* with a deleted *plo* gene (*T. pyogenes*Δ*plo*), and when anti-PLO antibodies were applied. The difference in susceptibility or resistance to PLO between the stromal and epithelial cells can be explained by a lower level of cholesterol expression in the epithelial cells. It has biological implications to PLO-mediated uterine tissue damage. The columnar epithelium of the endometrium establishes a first line of defence by forming a physical barrier between the uterine lumen and the underlying stroma. As it expresses less sensitivity to PLO, it protects the tissue from colonization by *T. pyogenes*. Once this barrier is disrupted, the bacteria can colonize the endometrium, and the development of the disease is induced. However, PLO did not induce inflammation, as the treatment of bovine endometrial culture cells with rPLO did not result in increased levels of IL-1β, IL-6, or CXCL8 in the culture supernatant. Nevertheless, during clinical infection, *T. pyogenes* often causes tissue damage in association with other species of bacteria, expressing together with *T. pyogenes* pathogens, the associated molecular patterns (PAMPs) and damage associated molecular patterns (DAMPs) responsible for the induction of inflammatory response.

The described sensitivity of stromal cells to PLO-mediated cytolysis is one of the explanations for how *T. pyogenes* acts as an opportunistic pathogen, responsible for endometrial damage once the protective epithelium is lost after parturition [68].

It has to be remembered that the development of a disease depends on the ability of an infectious agent to adhere, to colonize a host's target tissue, and develop mechanism(s) that are responsible for avoiding a host's defence and the involvement of its innate immunity. In the case of *T. pyogenes*, bovine uterine infections also interplay, as follows: the host–virulent strain of *T. pyogenes* or host–opportunistic strain of *T. pyogenes*, should be considered. The model presented by Ibrahim et al. [56] considers all of these factors. They focused on in vitro investigating the interactions between a bovine endometrial epithelial cell culture with either the *T. pyogenes* strain isolated from the uterus of a postpartum cow developing clinical endometritis, or *T. pyogenes* collected from the uterus of a healthy cow. The assessment of the presence of strain-specific factors participating in the development of the bovine clinical endometritis was also performed. The outcome of the analysis on a genetic level revealed that both strains presented genes encoding virulence factors; however, there were differences between the expression levels of *plo*, *cbpA*, *nanH*, *nanP*, *fimA*, *fimC*, *fimE*, and *fimG* genes, favouring the virulent strain. The proteins encoded by those genes participate in adhesion to the host cells [48,207], so a virulent strain had a higher ability to attach to the epithelium, because of the elevated level of the cell-wall associated molecules. Both strains expressed a cytolytic effect on cultured cells, which means both strains secreted PLO, a major virulence factor. However, this effect was abrogated when the bovine endometrial epithelial cells were treated with heat inactivated bacteria or bacteria-free filtrates. These findings are in line with the results obtained by Amos et al. [68], showing that BFF and rPLO were not responsible for the stimulation of a proinflammatory response in endometrial cells.

To mimic the interplay between the host target tissue and *T. pyogenes* in the chronic proinflammatory environment, endometrial endothelial cells treated with bacteria were co-cultured in the presence/absence of peripheral blood mononuclear cells (PBMCs). The presence of PBMCs did not influence the viability of the endometrial epithelial cells more than 16 h, upon the treatment with live bacteria, irrespective of a strain. However, the early detection (4 h) of genes encoding proinflammatory mediators (*PTGS2*, *CXCL3*, *IL6*, and *CXCL8*) was noticed in the endometrial endothelial cells treated with a virulent strain of *T. pyogenes*, in the presence of PBMCs. This in vitro interaction suggests that the leukocytes found within the endometrium may be responsible for sensitizing the epithelial cells to the initial bacterial infection through the enhancement of uterine innate immunity induction (Figure 6). These data suggest that the development of endometritis in dairy cows after parturition may be caused by a specific characteristic of certain strains of *T. pyogenes*. Moreover, the presence of immune cells can be the reason for the amplification of the proinflammatory response in the endometrial epithelial tissue towards *T. pyogenes* pathogenic strains.

**Figure 6.** Schematic presentation of the pathogenesis of uterine infection with *T. pyogenes*. Recruitment of monocytes (M), neutrophils (N), dendritic cells (DC), and other proinflammatory cells—secretion of proinflammatory mediators.

The research of Zhang et al. [186] indicated that the inflammation-inducing effect of PLO depends on its cytolytic activity, but not on PAMPs activity. These findings are in contrast to the results of Amos et al. [68] and Ibrahim et al. [56].

Studies done on bovine oviductal epithelial cells (BOECs), in an in vitro model, suggested a specific mechanism employed by *T. pyogenes* to avoid an immune response. In the course of *T. pyogenes* infection, there was no induction of any clear proinflammatory response in the BOECs at either the transcriptional or protein level. The BOECs co-cultured with *T. pyogenes* remained viable during the first 24 h of incubation when treated with MOI 0.01. However, a higher MOI caused the death of those cells within 24 h of co-culturing. What is interesting, is that the BOECs co-cultured with *T. pyogenes* expressed a similar level of *IL1A*, *IL1B*, and *TNFA* mRNA as the control cells did. Similar results were obtained when the level of mRNA concerning *CXCL8*, *CXCL1*/*2*, and *PTGS2* was assessed. Although, at the same time, the BOECs co-cultured with *Bacillus pumilus* responded to stimulation, with an increased level of cytokines and chemokines mRNA. It seems that the lack of responsiveness in the BOECs induced by *T. pyogenes* is one of the escape routes from immune surveillance developed by this pathogen, as the expression of the genes encoding the proinflammatory mediators was similar to the control levels [75].

#### **5.** *T. pyogenes* **Genome**

The first complete sequence of the *T. pyogenes* genome was described in 2014 by Machado and Bicalho [215]. However, to date, 19 genome assemblies of this species are available in the GenBank nucleotide database (NCBI). Complete genome sequences were reported for 10 strains, isolated as follows: four from swine, three from cattle, one from goat, one from forest musk deer, and one from water buffalo. Additionally, the vast majority of the sequenced genome of *T. pyogenes* were derived from strains, which were isolated from animals in China. Other available genome assemblies are scaffolds (*n* = 3) and contigs (*n* = 6). The main features of complete *T. pyogenes* genomes are summarized in Table 1.

The *T. pyogenes* genome was found to consist of a single circular chromosome comprising from 2.25 to 2.43 mega base pairs, depending on the strain. This species is characterized by high G+C% content. Moreover, the differences in the rate of G+C between the *T. pyogenes* strains are not significant (59.4–59.8%). tRNA is the most abundant RNA type in these bacteria. The share of rRNA in the whole nucleic acid content ranges from three to nine. In addition, there are also several non-coding RNAs (one or three, depending on the strain). The *T. pyogenes* genomes are indicated by the relatively high numbers of pseudogenes (from 27 to 191).


**Table 1.** General features of the complete *T. pyogenes* genomes available in the GenBank nucleotide database (NCBI).

The *T. pyogenes* genome encodes many genes, for which the functions were assigned by the homologies to known proteins. Apart from the virulence-associated genes, the genome of this species can contain several antibiotic resistance genes. Furthermore, the *lux*S gene regulating the formation of biofilms was also found in the genome sequence [209]. In addition, the presence of a member of the *T4virus* (vB\_EcoM-UFV13) indicates the efficiency in controlling the biofilm formation by *T. pyogenes* [216]. Moreover, the *T. pyogenes* genome contains four housekeeping genes (*met*G, *tuf*, *gyr*A, and *fus*A). The *tuf* and *fus*A genes encode the translation elongation factors Tu and G, respectively. Two other genes confer the DNA gyrase subunit A (*gyr*A) and methionine—tRNA ligase (*met*G). These genes were successfully used in the multilocus sequence analysis (MLSA) of this species [148]. The genetic analysis showed that *T. pyogenes* had a more complex system of amino acid and lipid metabolism, and more genes involved in the pathogenicity than are found in *Arcanobacterium hemolyticum*. These differences may affect the ability to cause infections in different host types by *T. pyogenes* isolates. Moreover, all of the groups of the family *Actinomycetaceae* have phosphotransferase systems that are probably essential for the colonization of a wide range of hosts and for the initiation of abscess formation. Some genes were lost or acquired as a result of lateral evolution, which helps in the adaptation of bacteria to a new environment [6]. The *T. pyogenes* species was reclassified based on 16S rRNA gene sequences, the menaquinone structure, and the phospholipid composition from the genus *Arcanobacterium* to the genus *Trueperella*, together with four other species, such as *Trueperella abortisuis*, *Trueperella bernardiae*, *Trueperella bonsai*, and *Trueperella bialowiezense*. Furthermore, 16S rRNA gene sequence similarities between strains from the genus *Trueperella* ranged from 95.3% to 98.6% [1]. There was also a draft genome sequence of one human *T. bernardiae* isolate deposited in the NCBI [217].

The genomes of *T. pyogenes* also include plasmids. Until now, two plasmids have been described for this species [178,218]. First, a native pAP1 (U83788) is a circular plasmid of 2439 bp, containing three open reading frames (ORFs). Plasmid pAP1 encodes the *rep* gene, which is required for the initiation of replication. Two other genes, *orf* 112 and *orf* 129, encode the hypothetical proteins of an unknown function. However, pAP1 does not include any antibiotic resistance genes [218]. The other, pAP2 (AY255627), is a circular plasmid of 9304 bp, and it contains seven ORFs. In pAP2, *rep*A is probably required for the replication of this plasmid. Moreover, there are other genes, namely: *tet*R(33), which encodes the repressor protein and *gcr*Y, and *orf* 95 with unknown functions. In addition, pAP2 is characterized by the presence of two genes encoding antibiotic resistance determinants—*erm*(X) and *tet*A(33). In *T. pyogenes*, *erm*(X) determines the resistance to macrolide antibiotics, while *tet*A(33) is associated with low-level tetracycline resistance [178]. Furthermore, plasmid pAP2 contains the insertion sequence IS*6100*, which is found either in Gram-positive or in Gram-negative bacteria [219].

The knowledge of the complete genome sequence may allow for the identification of new genes that may contribute to the pathogenicity and antibiotic resistance of *T. pyogenes*. In addition, genome sequencing can help to understand the basics of the metabolism and evolution of the bacterial species.

#### **6. Immunoprotection and Perspectives**

The antimicrobial treatment of *T. pyogenes* infections is often ineffective, because of the increasing resistance of bacteria or the limited distribution of drugs to the site of infection, for example, to abscesses. Hence, vaccination should be considered as a primary method of *T. pyogenes* infection prevention. It seems that some of the virulence factors of *T. pyogenes* could be promising candidates for vaccine antigens.

Most studies concerning the stimulation of protective immunity against *T. pyogenes* infection were done on a mouse model, and only a few were done using a ruminant or swine model. The first trials to protect ruminants against infection with *T. pyogenes* were focused on animal treatment with whole cells of *T. pyogenes* or a culture supernatant, and were mostly unsuccessful; however, some therapeutic effects on mastitis development were observed [48,220,221]. In swine, the vaccination of pregnant sows with an autovaccine was an attempt to control the losses of newborn animals. Studies done by Kostro et al. [222] revealed that the administration of *T. pyogenes* cells treated with a phenol solution to pregnant sows, six and three weeks before anticipated delivery, significantly increased the number of CD4<sup>+</sup> T cells, CD8<sup>+</sup> T cells, and CD25<sup>+</sup> T cells, as well as enhanced the levels of IFN-γ, TNF-α, and IL-10 in a colostrum and milk [222]. The recent trials were concentrated on the protection of heifers against post partum uterine diseases. Pregnant heifers were vaccinated using different routes of immunization (subcutaneously and intravaginally), and different compositions of the vaccine, as follows: (i) inactivated bacterial whole cells (*E. coli*, *T. pyogenes*, and *F. necrophorum*) together with recombinant proteins, FimH, PLO, and leukotoxin (LKT)); (ii) recombinant proteins FimH, PLO, and LKT only; (iii) inactivated bacterial whole cells (*E. coli*, *T. pyogenes*, and *F. necrophorum*) only [223]. The obtained results demonstrated that subcutaneous vaccination significantly decreased the incidents of puerperal metritis, but intravaginal vaccination failed in preventing the disease.

Mice used for the assessment of protective immunity induction against infection with *T. pyogenes* provide a reliable model that can serve for further comprehensive studies on other animal species. The administration of formalin-inactivated recombinant PLO (rPLO) induced the protection of mice against intraperitoneal challenge with 10<sup>8</sup> of *T. pyogenes* cells, and PLO-specific antibodies were detected in the sera of the immunized mice [168,180].

The current trend in vaccine design is concentrated on the development of such a product that will induce immunity against multiple pathogens. Using a mouse model, Zhang et al. [224] and Hu et al. [176] induced specific immunity that protected animals against challenge with *T. pyogenes* and *C. perfringens*. However, their approach to solve the problem concerning the stimulation of protective immunity was quite different. Hu and colleagues [176] designed a chimeric protein called rPC-PD4, which was composed of *C. perfringens* truncated regions of C-domain of phospholipase C (rPLC), and D4 domain of *T. pyogenes* PLO (rPLO), and was encoded by chimeric genes incorporated into a plasmid vector. The production of specific antibodies, and the presence and level of proinflammatory cytokines (TNF-α, IL-1β), chemokines (CXCL8), and IL-10 in sera of immunized mice were assessed; however, only partial immunity was observed after challenge with *T. pyogenes* or *C. perfringens*. Zhang's group [224] concentrated not only on PLC and PLO as major immunogens, but also on the involvement of formaldehyde inactivated cultures of *T. pyogenes* and *C. perfringens* as vaccine antigens, combined with aluminium hydroxide gel. The outcome of their proposal is promising, as the survival rate of immunized mice after challenge with *T. pyogenes* and *C. perfringens* was 80% and 100%, respectively. These studies [176,224] present a very interesting approach for the development of a new, safe vaccine against polymicrobial infections.

Nonetheless, the latest trials to stimulate immunoprotection against *T. pyogens* infection are the most promising so far. The studies of Huang's group [225,226] done on a mouse model, concentrated on genetic immunization. Two pathways to stimulate protective immunity against *T. pyogenes* infection were used. The first one applied the modification of the cytokine environment of immunized animals together with delivering the gene encoding PLO [225], the second one focused on immunization with a DNA vaccine containing genes encoding four different virulence factors of *T. pyogenes*, and simultaneous vaccine protection against destruction by the host environment [226].

In the first study, concurrent immunization with two plasmids carrying the *plo* gene and the *IL1*β gene, respectively, resulted in the protection of mice from intraperitoneal challenge, with 3.7 <sup>×</sup> 108 cfu of *T. pyogenes* [225]. PLO-specific antibodies were detected in the sera of immunized mice with a higher titer of the IgG2a subtype than the IgG1 subtype. The T-cell profile analysis indicated an increased level of CD8<sup>+</sup> T-cells and CD4<sup>+</sup> T-cells. These results correlated with the detection and assessment of the cytokine levels evaluated in the supernatant collected from the spleen lymphocytes culture (increased levels of IFN-γ, and IL-2, and slightly diminished IL-4). In this model, IL-1β enhanced the immunogenicity of the vaccine by influencing the activity of the macrophages, the major population of the phagocytic cell, as well as antigen presenting cells (APCs), and also the B-cells, which act as APCs during early phases of immune response development, later transforming into plasma cells secreting antigen-specific antibodies, hence humoral immunity.

In the second research of Huang's group [226], the *plo*, *cbpA*, *fimA*, and *nanH* genes were paralleled as a single chimeric gene with an attached CpG ODN 1826 motif, so as to overcome low immunogenicity and susceptibility to host endonucleases. Finally, the chimeric gene was incorporated into the expression plasmid vector, and then this construct was encapsulated in chitosan nanoparticles to protect the DNA. The chimeric gene DNA vaccine introduced intramuscularly initiated production of antibodies specific for the target epitopes of PLO, CbpA, NanH, and FimA in mice. Moreover, the level of IgG2a was higher than the IgG1 in the sera of immunized mice, suggesting that immunization with this vaccine mainly primed the Th1 profile.

Compared with the subunit and conventional vaccines, DNA vaccines express some advantages as future stimulants of immunity. They are easy to manufacture, stable, and confer potential safety. Moreover, many genes encoding specific antigen proteins can be inserted in a carrier genome, and once introduced to the host, stimulate protective immunity against a variety of infections. In this context, the studies done by Huang et al. [225,226] are very important, not only in the induction of effective protection against *T. pyogenes* infection. They also proved that the formulated chitosan-CpG ODN nanoparticles are stable in eukaryotic cells, are able to protect the chimeric gene in the encapsulated plasmid from degradation, promote its expression, and also participate in the induction and enhancement of the immune response, so they could serve as a safe and efficient drug release carrier system.

#### **7. Conclusions**

*T. pyogenes* has significant importance as an opportunistic animal pathogen causing a variety of purulent infections. These infections pose a huge economic problem in livestock breeding. Therefore, it is crucial to understand the mechanisms of pathogenicity of this bacterium and the routes of its transmission for the development of an effective prevention strategy.

Although some of virulence determinants in *T. pyogenes* are well recognized, little is known about the dissemination of infections. It is interesting, there is still limited data about the pathogen–host interactions that may be involved in the development of *T. pyogenes* infection. Especially, the determination of factors that have a key role in transformation from a commensal to a pathogenic bacterium is needed. Although it seems that animals are the main reservoir of this bacterium, the epidemiological relationships between the particular isolates remain unclear.

**Author Contributions:** Conceptualization, M.R.; writing (original draft preparation), M.R., E.K., D.C.-C., M.K.-S., ´ I.S., and M.G.; writing (review), M.R., E.K., and M.G.

**Funding:** This work was supported by KNOW (the Leading National Research Centre) Scientific Consortium "Healthy Animal–Safe Food", decision of Polish Ministry of Science and Higher Education, no. 05-1/KNOW2/2015.

**Acknowledgments:** The authors thank Dorota Rzewuska for the technical support during editing the manuscript.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Review* **Divergent Approaches to Virulence in** *C. albicans* **and** *C. glabrata***: Two Sides of the Same Coin**

**Mónica Galocha 1,2, Pedro Pais 1,2, Mafalda Cavalheiro 1,2, Diana Pereira 1,2, Romeu Viana 1,2 and Miguel C. Teixeira 1,2,\***


Received: 10 April 2019; Accepted: 8 May 2019; Published: 11 May 2019

**Abstract:** *Candida albicans* and *Candida glabrata* are the two most prevalent etiologic agents of candidiasis worldwide. Although both are recognized as pathogenic, their choice of virulence traits is highly divergent. Indeed, it appears that these different approaches to fungal virulence may be equally successful in causing human candidiasis. In this review, the virulence mechanisms employed by *C. albicans* and *C. glabrata* are analyzed, with emphasis on the differences between the two systems. Pathogenesis features considered in this paper include dimorphic growth, secreted enzymes and signaling molecules, and stress resistance mechanisms. The consequences of these traits in tissue invasion, biofilm formation, immune system evasion, and macrophage escape, in a species dependent manner, are discussed. This review highlights the observation that *C. albicans* and *C. glabrata* follow different paths leading to a similar outcome. It also highlights the lack of knowledge on some of the specific mechanisms underlying *C. glabrata* pathogenesis, which deserve future scrutiny.

**Keywords:** *Candida*; host-pathogen interaction; virulence; biofilm formation; morphology; immune evasion

#### **1. Introduction**

Infections caused by fungi affect millions of people worldwide, with the overall mortality rate estimated to be roughly 1,350,000 deaths per year [1]. Among pathogenic fungi, *Candida* species are responsible for the most common invasive fungal disease in developed countries—the candidiasis [2]. *Candida* species live as commensals on mucosal surfaces where they are constituents of the normal microbiota of the oral cavity and gastrointestinal and vaginal tracts. However, they can opportunistically become pathogenic under suitable conditions, such as host-disrupted microbiota or immunocompromised hosts, being responsible for clinical manifestations from mucocutaneous overgrowth to bloodstream infections [3–5]. Of the various *Candida* species, *Candida albicans* and *Candida glabrata* not only account for 60% of *Candida* species present in the human body, but also constitute the most prevalent of the pathogenic *Candida* species, being responsible for more than 400,000 life-threatening infections worldwide every year [3,6].

*C. albicans* and *C. glabrata* are the two most common pathogenic yeasts of humans, yet they are phylogenetically, genetically, and phenotypically very different. On one hand, *C. albicans* diploid genome carries several gene families that are associated with virulence [7]. These include the ALS (agglutinin-like sequence) adhesins, required for host adhesion, secreted aspartyl proteases (SAPs) and phospholipases, which allow for the degradation of host barriers and the invasion of surrounding tissue, and proteins involved in oligopeptide and iron transfer [8–10]. On the other hand, *C. glabrata* mechanisms of tissue invasion are mostly unknown, although it is hypothesized to possibly occur by endocytosis induction of host cells [11]. Its haploid genome encodes a large group of glycosylphosphatidylinositol (GPI)-anchored cell wall proteins, such as the adhesins from the *EPA* gene family, implicated in fungus–host interactions or biofilm formation, and a family of aspartic proteases (yapsins) which are mainly associated with cell wall remodeling and possible immune evasion [12,13]. Moreover, key virulence attributes of *C. albicans,* which are known to be the basis of its pathogenicity, are absent in *C. glabrata* [14]. Switching from yeast to hyphal growth not only allows consistent biofilm production but also enables *C. albicans* to be highly invasive and escape macrophage engulfment [15–19]. Nevertheless, both species are known to use biofilms to colonize the surface of several medical devices based on different materials [20]. Unlike *C. albicans*, it has been demonstrated that *C. glabrata* lets itself be taken up by macrophages, where it persists and divides for long periods of time eventually leading to cell lysis due to fungal load [21,22]. It has the ability of detoxifying oxidative radical species and disrupting normal phagosomal maturation, leading to the inhibition of phagolysosome formation and phagosome acidification [21,23].

The interaction between *Candida* and its host cells is characterized by a complex interplay between the expression of fungal virulence factors and the host immune system, and the presence of other microorganisms affects this interplay. This review aims to explore and compare the remarkably distinct paths toward virulence trailed by the two most common causative agents of candidiasis worldwide. On one hand, *C. albicans* is known for its ability to evade host defenses and form bulk biofilms due to its ability to undergo filamentous growth, while on the other hand, *C. glabrata* is an unusually stress-tolerant organism able to survive and replicate inside the immune system cells. Despite having such distinct virulence features, *C. glabrata* and *C. albicans* are frequently co-isolated [11].

#### **2. Host Damage and Invasion**

There is a variety of defense mechanisms through which the human host is able to prevent invasion by pathogenic microorganisms, such as *C. albicans* and *C. glabrata*. These mechanisms consist not only of physical but also of chemical barriers. For instance, epithelial cells, which in most cases are the first line of contact between host and pathogen, function as the prime physical barrier restraining *Candida* from invasion of the underlying tissue. On one hand, these cells are interconnected through "tight junctions" preventing the entry of microorganisms into interepithelial space and eventually into the bloodstream [24,25]. On the other hand, some types of epithelial cells, such as those in the intestinal or vaginal epithelium, are able to produce a mucus layer by secreting mucins [24,26]. This layer impairs *Candida* invasion by preventing contact with the epithelium surface. Likewise, in the oral cavity the flow of saliva plays an important role as both a physical and a chemical barrier as it not only prevents the adhesion to mucosa and dental surfaces but also contains several antimicrobial agents that impair the contact of *Candida* with the oral epithelium [27,28]. Another chemical barrier against *Candida* establishment is the presence of gastric acid and bile in the digestive system which creates a harsh environment for fungal growth. Nevertheless, these human pathogens are known to have a remarkable ability to adapt to these adverse conditions and proliferate.

*C. albicans* relies on two distinct invasion mechanisms to gain entry into host cells: (i) induced endocytosis and (ii) active penetration of hyphal forms through physical forces of hyphae production associated with lytic enzyme secretion [29] (Figure 1). Nonetheless, depending on the host cell, these invasion mechanisms are thought to be exploited to a different extent. For instance, it was demonstrated that while invasion into oral epithelial cells occurs via both routes, invasion into intestinal epithelial cells occurs only via active penetration under normal conditions [29,30].

**Figure 1.** Schematic representation of *C. albicans* and *C. glabrata* host damage and invasion. *C. albicans* enter host cells through induced endocytosis or by active penetration (inter- and intra-cellular) of hyphal forms associated with the release of hydrolytic enzymes, resulting in the damage of cells and loss of epithelial integrity. Induced endocytosis of host cells is thought to be the mechanism behind *C. glabrata* tissue invasion.

Living within host cells is a profitable strategy since it enables cells to deal with the host immune system and antimicrobial treatment, and there are plenty of nutrients available and an absence of competition with other commensal microorganisms. Additionally, dissemination into deeper tissues and ultimately into the bloodstream is easier from within the cell. Fungal invasion via induced endocytosis is dependent on dynamic microfilaments of the host. In *C. albicans*, the GPI-anchored hypha-associated protein Als3 interacts with mammalian cadherins, mimicking the establishment of adherence junctions. This process leads to rearrangements in the actin cytoskeleton that ultimately lead to fungal cell internalization [31]. Similarly, Ssa1, a member of the HPS70 heat shock protein family, was reported to play the same role as Als3, being essential for maximal host cell damage and subsequently fungal cell endocytosis [32]. However, *EED1* was the first fungal gene reported as being required for epithelial escape and interepithelial dissemination and not for initial invasion into epithelial cells [33]. Moreover, it is thought that the contact with the epithelium, among several other stimuli, is a highly potent inducer of *C. albicans* filamentation. In turn, the contact between hyphal cells and epithelial cells induces host defense mechanisms such as the formation of epithelial cell protrusions surrounding the hyphae, and membrane ruffling, which is characteristic for induced endocytosis [34].

The ability to switch between yeast and hyphal growth forms is one of the most discussed and best-investigated virulence attributes of *C. albicans*. This morphology switch is activated by well-established kinase-based signal transduction pathways and is triggered by diverse host environmental cues, including temperature, pH, serum, and CO2 and is linked to several steps during host invasion [35,36]. The extracellular signals are transmitted via Ras to both protein kinase A and the MAP kinase cascade [37–39], inducing hyphal differentiation through the activation of a number of transcription factors such as Efg1 [40] and Ume6 [41]. The transcriptional repressor Nrg1 is inactivated and removed from the hyphal-specific gene promoters, thereby allowing the induction of hyphal morphogenesis [42,43].

Invasion into epithelial cells via active penetration relies on a combination of physical pressure employed by the growing hyphae and the secretion of hydrolytic enzymes. Moreover, hyphal cells are capable of directional growth in response to contact with a solid surface (thigmotropism) which enables *C. albicans* to specifically identify and invade intercellular junctions, thereby damaging the epithelium compact structure [44]. Interestingly, it was very recently discovered that *C. albicans* release hydroxyphenylacetic acid (HPA) during hyphal growth [45]. Its production seems to occur through the same pathway and the same precursors as tyrosol, which is able to stimulate hypha induction in *C. albicans* [46], therefore their biological functions are likely to be the same. Additionally, along with the active penetration by hyphal cells, there is also the secretion of hydrolytic enzymes, such as SAPs and phospholipases, which can digest epithelial cell surface components enabling the entrance into or between host cells [47]. SAPs are the best-characterized members of the *C. albicans* hydrolytic enzymes, as is well reviewed by Hube and Naglik [48]. Interestingly, these tissue damaging

enzymes not only have distinct optimal pH requirements but also are growth stage and infection site related [47]. For instance, *SAP1–SAP3* are yeast growth-associated and have optimum activity at lower pH values, while *SAP4–SAP6* are hyphal growth-associated and have optimum activity at higher pH values [49]. Similarly, it was demonstrated that *SAP1, SAP3,* and *SAP8* are preferentially expressed in vaginal, rather than oral, *C. albicans* infections [47]. One possible explanation for the existence of a significant number of different *SAP* genes may be the necessity for specific and optimized proteinases during the different stages of an infection [48]. Additionally, the cytolytic peptide toxin of *C. albicans* candidalysin, encoded by the hypha-associated gene *ECE1*, was recently found to be essential for damage of enterocytes and is a key factor in subsequent fungal translocation, suggesting that transcellular translocation of *C. albicans* through intestinal layers is mediated by candidalysin [25]. Moreover, phytase activity in *C. albicans* was demonstrated to be important for virulence. Phytate is a major storage form of phosphorus in plants and is abundant in the human diet and intestinal tract [50]. Recently, it was reported that decreased phytase activity leads to a reduced ability to form hyphae, attenuated in vitro adhesion, and reduced ability to penetrate human epithelium [51].

While the transition from yeast to hyphae has been extensively studied in *C. albicans*, the switch from hyphae to yeast still remains poorly understood [35]. Nevertheless, both yeast and hyphae forms are found in infection sites, which suggests that both forms are implicit in the infection process. Interestingly, it was demonstrated that depending on the infected organ, one or the other morphology predominates [52]. It is thought that the yeast form is important for dissemination upon infection, whereas hyphae forms are more relevant to attachment, host invasion, and tissue damage [53].

Unlike *C. albicans*, and despite the existence of some reports demonstrating that *C. glabrata* forms pseudohyphae [54,55], the pathogenicity of *C. glabrata* seems to be independent of morphology. The most common route for this pathogen to reach the bloodstream is through the iatrogenic breach of natural barriers, such as the use of catheters, trauma, or surgery.

In 2000, Csank and Haynes [54] reported for the first time that*C. glabrata* can undergo morphological change and grow as a pseudohyphae on solid nitrogen starvation media. This invasive growth mode could be a possible mechanism of host invasion, however, this phenomenon has not yet been reported in vivo. Despite lacking this prime virulence feature, this opportunistic pathogen is still able to reach the human bloodstream and cause infection. In some cases, *C. glabrata* can involuntarily reach the bloodstream through nosocomial conditions, namely surgery, catheter, parenteral nutrition, and burn injury [56]. However, even when these external factors are abrogated, *C. glabrata* is able to invade the host and colonize different tissues, as shown in an intragastrointestinal mouse model of infection [57] and in a chorioallantoic membrane chicken embryo model of infection [58]. Thus, *C. glabrata* must have other invasion mechanisms.

Co-infection with other microorganisms may be a possible explanation to the invasive capacity of *C. glabrata* since this yeast is often co-isolated in infections with *C. albicans* [59–61] or even other pathogens such as *Clostridium di*ffi*cile* [62]. In the co-infected environment, *C. glabrata* cells may exploit the tissue invasion and destruction caused by *C. albicans* to access nutrients and reach the bloodstream. In fact, Tati et al. (2016) [63] demonstrated that when mice are infected with *C. glabrata* alone, oropharyngeal candidiasis is negligible, however, when co-infected with *C. albicans*, an increased colonization by *C. glabrata* was observed. This effect was attributed to the binding of *C. glabrata* to *C. albicans* hyphae [63] and similar results were reported by Alves et al. (2014) [64] using a reconstituted human vaginal epithelium. Nonetheless, the intracytoplasmic presence of *C. glabrata* was detected in oral epithelial cells [65] and vaginal epithelial cells [66] and it has been shown that when endocytosis is inhibited, the internalization of *C. glabrata* is prevented [65]. This suggests that induced endocytosis by host cells could be the most likely mechanism of *C. glabrata* internalization (Figure 1).

As referred to before, the tissue/cell damaging ability of *C. glabrata* is lower compared to *C. albicans*. In *C. albicans*, secreted hydrolytic enzymes are considered to be important destructive factors that damage host tissues, providing nutrients for its propagation. However, the production of these hydrolytic enzymes is very low or even null in *C. glabrata,* wherefore its importance for virulence does

not seem to be as relevant as it is in *C. albicans* [67–73]. The proteinase enzyme is responsible for protein degradation resulting in tissue invasion. Among the proteases, SAPs are considered crucial for the pathogenicity of *C. albicans* [74]. Although *C. glabrata* does not express SAPs [75], its genome contains 11 non-secreted GPI-linked aspartyl proteases (*YPS* genes), which are surface-exposed aspartic proteases required for virulence, known as yapsins [13]. These yapsins are important for cell wall maintenance, remodeling, cell to cell interactions, and resistance to cell wall stress, however, their direct link with virulence is not very well characterized yet. Otherwise, phospholipases promote the destruction of cell membrane phospholipids, causing cell damage and lysis which allows a greater invasive capacity [76]. Some of the *C. glabrata* strains are able to produce these enzymes [67,68,70,71,77], which appear to play a role in *C. glabrata*-associated persistent candidemia [9]. However, phospholipase production in *C. glabrata* is lower than in *C. albicans,* and in some cases inexistent [69,72,73], therefore its relation to *C. glabrata* virulence is not clear and needs further analysis.

#### **3. Adhesion and Biofilm Formation**

The ability to infect and prevail in the human host is related to different pathogenesis factors, of which biofilm formation excels [19,78–80]. *Candida* species ability to form biofilms on medical devices increases mortality rates associated with infections, while often forcing the treatment to include the removal of the medical device [81]. A lot of efforts have been put into understanding the molecular basis of *Candida* species biofilm formation [20].

Adhesion is one of the most relevant and advantageous capacities of the yeast cell wall. It allows cells to colonize mucosal surfaces and prevail in a nutritional environment, being the first critical step for biofilm formation, which serves as a shield against adverse conditions, as well as a highly drug-resistant reservoir of infective cells [82,83]. *C. albicans* and *C. glabrata* pathogenesis has been strongly related to adhesion, which is considered a crucial virulence factor in these yeasts [84–86]. In this regard, both *Candida* species are able to not only attach to mammalian host cells (epithelial, endothelial, and immune cells) but also to other microbes (bacteria and other *Candida* species) and abiotic surfaces, such as medical devices [85,87]. Several studies have tried to understand the nature of adherence to plastic surfaces. For instance, cell surface hydrophobicity (CSH) seems to have a positive correlation with adhesion in both *C. albicans* and *C. glabrata* species, thus, adhesion is mediated by van der Waals forces. Moreover, compared to *C. albicans*, the relative CSH of *C. glabrata* has been shown to be significantly higher [88–90].

Both *Candida* species have a set of proteins that enable attachment, known as adhesins. The most important *C. albicans* adhesins are agglutinin-like sequence (Als) proteins (Als1–7 and Als9) [91] and hypha-associated GPI-linked protein (Hwp1), known to be required for adhesion and virulence in vivo, and also being associated to biofilm formation through interaction with Als1 and Als3 adhesins [92–94]. Other adhesins required for adhesion and biofilm formation include Hwp2 [95] and Eap1 [96–98]. *C. glabrata* expresses a large group of adhesins, belonging to the epithelial adhesin (Epa) family, which is encoded by 17 to 23 genes, depending on the strain [78]. Among these, Epa1 is a major virulence player in *C. glabrata*, and mediates 95% of in vitro adhesion to epithelial cells [99]. This adhesin is highly heterogeneous among *C. glabrata* clinical isolates, being an important virulence factor [100]. Epa6 and Epa7 are involved in kidney and bladder colonization in vivo and boost biofilm formation [101–104]. Transcriptomic and proteomic studies have revealed that besides *EPA* genes, *C. glabrata* holds other biofilm-related adhesin families, such as Pwp (encoded by seven members *PWP1–7*), Aed (*AED1* and *AED2*), and Awp (encoded by 12 members *AWP1–6* and *AWP8–13*), which are usually found in significantly higher numbers in clinical isolates, consistently with an important role in pathogenesis [12,105–107].

The successful pathogenicity of these yeasts relies on its flexibility, which allows for adaptation and proliferation under both nutrient-rich and nutrient-poor conditions. Several studies have reported the importance of host and antifungal selective pressure on virulence traits as adhesion [108,109]. A recent study conducted by Vale-Silva et al. (2017) [110] used the PacBio technology to compare the genomes of two sequential *C. glabrata* clinical isolates and observed a significant increase in the number of adhesin-encoding genes (101 and 107) when compared to the CBS138 genome (63), despite the limited variation between the two studied isolates. The same authors, along with Ni et al. (2018) [111], further linked the increased expression of the adhesin gene *EPA1* with gain-of-function (GOF) mutations in the *PDR1* gene in drug-resistant clinical isolates [112]. *EPA1* has been strongly related to adhesion of *C. glabrata* cells to mammalian epithelial cells both in vitro [99] and in vivo [113]. Furthermore, Salazar et al. (2018) [114] also observed a GOF mutation in the *PDR1* gene, which led to changes in the transcriptome when compared to the CBS138 strain. Among the genes identified as having the highest number of non-synonymous SNPs, there were several genes encoding adhesins and, agreeing with Vale-Silva et al. (2013) [115], the number of adhesin-expressed genes varied when compared to other GOF *PDR1* mutations [114]. This reinforces the idea that antifungal treatment deploys a tight selective pressure which results in changes at the genomic and transcriptional levels, particularly affecting adhesin-encoding genes [116]. The transcription factor Cst6 was found to also play a role in *C. glabrata* adhesion and biofilm formation, negatively regulating the expression of *EPA6* [103]. In *C. albicans*, a transcriptional regulatory network comprising nine regulators (Bcr1, Brg1, Efg1, Flo8, Gal4, Ndt80, Rob1, Rfx2, and Tec1) was identified in in vitro and in vivo studies as underlying the biofilm formation phenomenon in this pathogenic yeast [117–121].

Hyphae formation, which is exclusive to *C. albicans*, when compared to *C. glabrata* is also an important trait in biofilm development. Various studies have described hyphae as exhibiting improved adhesion to the human epithelium, with these cells displaying increased expression of Als1, Als3, and Hwp1 [122–125]. Interestingly, Tati et al. (2016) [63] characterized the co-colonization of *C. glabrata* and *C. albicans* in a murine model of oropharyngeal candidiasis (OPC) and demonstrated that this is driven by specific adhesins in both species. Namely, the *C. albicans* Als3 and Als1 adhesins are crucial for in vitro binding of *C. glabrata* cells to *C. albicans* hyphae and for further in vivo establishment of infection. Considering *C. glabrata* cells, incubation with *C. albicans* hyphae led to the overexpression of the adhesins *EPA8*, *EPA19*, *AWP2*, *AWP7,* and *CAGL0F00181* [63].

After adhesion, *Candida* species are known to be able to form a 3D-structure of cells embedded in a gel-like matrix [20,126,127]. In order to achieve this, surface-adhered cells begin to adhere to other *Candida* cells, initiating the formation of discrete colonies, which correspond to an early phase of biofilm formation. At this point, the intermediate phase begins with the cellular production and secretion of important molecules, known as extracellular polymeric substances (EPS), that will constitute the extracellular matrix of the biofilm, protecting the cells and ensuring a more developed 3D-structure. The final structure is reached after the maturation phase, where more cells and the matrix are originated. Mature biofilms might also suffer the detachment of some cells that can spread to form new biofilms on other niches of the host, a process called the dispersal phase [20,126,128,129]. Although this process is true for every *Candida* species, there are differences between *C. albicans* and *C. glabrata* biofilms, regarding their dimensions and structure, cell morphology, EPS produced and secreted, response to environmental cues, and resistance to antifungal drugs (Figure 2).

A very clear difference between *C. albicans* and *C. glabrata* mature biofilms is the dimension and total biomass of each biofilm. *C. glabrata* in vivo biofilm formation leads to a thickness of 75–90 ± 5 μm, which is half of the normal thickness of *C. albicans* biofilms [130], with much less biomass in the end of biofilm formation compared to *C. albicans* [131]. The organization of the biofilm structure also differs between these two species. *C. albicans* biofilms are arranged in a three-dimensional structure with different morphologies and empty spaces between cells [132], where microchannels are formed [133]. On the other hand, *C. glabrata* biofilms are thinner, but display a higher density of cells, tightly packed together [132]. Although *C. glabrata*'s biofilms are composed by yeast cells only [131,132], the same is not true for *C. albicans* biofilms, where different morphologies arise. *C. albicans* mature biofilms are composed by a dense network of pseudohyphae, hyphae, and yeast cells [134]. This filamentation process in biofilm formation is controlled by the transcription factor Efg1, without which *C. albicans* only forms scarce monolayers of elongated yeast cells on polyurethane catheters and polystyrene [135,136].

**Figure 2.** Schematic representation of *C. albicans* and *C. glabrata* biofilm formation. *C. albicans* forms thicker biofilms, with much more biomass in the end of biofilm formation and produces more extracellular matrix than *C. glabrata*. *C. albicans* mature biofilms are composed by a dense network of pseudohyphae, hyphae, and yeast cells, whereas *C. glabrata* biofilms are composed by compact yeast cells only, forming a thin but dense biofilm.

Although*C. albicans* produces more extracellular matrix than*C. glabrata* [137], the main components of the matrix, proteins, and carbohydrates, are the same in both biofilms [131,138–140]. Interestingly, *C. glabrata* has a very high content of proteins and carbohydrates, which is five times higher than that found on the biofilms of other non-*albicans Candida* species [131]. *C. albicans'* matrix is also composed by other lipids (mainly neutral glycerolipids, polar glycerolipids, and a small percentage of sphingolipids) [139], phosphorus, and uronic acid [140]. Biofilms of both species also have a small content of extracellular DNA [131,138–140].

Interestingly, *C. albicans* and *C. glabrata* also behave differently on different surfaces when it comes to initiating biofilm formation. Cleary, each species has a different propensity to form biofilm on a given surface. For instance, *C. albicans* is known to adhere better to latex and silicone elastomer while showing less biofilm formation on polyvinyl chloride, polyurethane, or 100% silicone [134]. *C. albicans* also adheres and forms biofilm on different surfaces of denture base materials, having higher biofilm formation on the surface of alloy and lower biofilm formation on methacrylate-based denture material [141]. Moreover, polyetherurethane treated with 6% of polyethylene oxide was found to reduce the metabolic activity of cells and the total biomass of *C. albicans* biofilms [142]. On the other hand, while all other pathogenic *Candida* species have greater biofilm formation on Teflon, *C. glabrata* prefers polyvinyl chloride to form biofilm [143]. Interestingly, other components of the environment might alter the ability of *Candida* species to adhere to a given surface. For instance, the presence of saliva has been shown to decrease the ability of *C. albicans* to form biofilm in vitro [144,145].

Depending on the antifungal drug, *C. albicans* and *C. glabrata* biofilms might be able to resist the therapeutic action of the drug. For instance, Choi and colleagues (2007) [146] measured the in vitro susceptibilities of biofilms of *C. glabrata* and *C. albicans* bloodstream isolates, showing that both biofilms were resistant to fluconazole and only moderately resistant to amphotericin B, while exposure to 0.25 to 1 μg/mL of caspofungin and micafungin lead to an 80% reduction of the biofilms. Moreover, voriconazole is also able to reduce *C. albicans* and *C. glabrata* biofilms, being present at 0.25 mg/L or being used as a surface coating at different concentrations [147]. Nevertheless, when growing on an RPMI 1640 medium, *C. glabrata* mature biofilms have shown to be less susceptible to caspofungin and anidulafungin than *C. albicans* mature biofilms on a polystyrene surface [148], showing that each species biofilms might resist differently to the same antifungal drug. Although reacting differently to antifungal drug exposure, the strategy to achieve resistance seems to be very similar between the two species. The mechanisms known to underlie resistance to antifungal drugs in *Candida* biofilms are believed to be related to alterations in the metabolic activity, the role of the extracellular matrix as a barrier for diffusion, the role of its EPS components, and the presence of persister cells within the biofilm [20,149]. Both species suffer the upregulation of drug efflux pump-encoding genes [138,150], as well as seem to rely on the β-1,3-glucans present on the extracellular matrix [151,152].

Biofilms also allow for the survival of *Candida* species by protecting them against the host immune system. The extracellular matrix is essential for the protection against the action of neutrophils by inhibiting the release of neutrophil extracellular traps (NETs) [153]. Moreover, the β-glucans present on the matrix avoid the activation of neutrophils, actually inhibiting the reactive oxygen response, thereby being an important distracting mechanism to evade the innate immune system [154]. Biofilms are also believed to resist well the action of the innate immune system due to its heterogenicity, given the different types of cells and different metabolic activity the cells might present on biofilms [155].

The combination of *C. albicans* and *C. glabrata* to form biofilm has been well described and the two species are usually found together in niches of candidiasis patients [60]. A recent study has shown that a ratio of *C. albicans* to *C. glabrata* of 1:3, significantly increases the total biofilm biomass comparatively to a *C. albicans* monoculture or ratio of *C. albicans* to *C. glabrata* of 1:1. This co-culture biofilm exhibited a high heterogenicity with *C. albicans* hyphae and *C. glabrata* cells clustered together in a 3D-structure. Interestingly, an upregulation of *HWP1* and *ALS3* genes is observed in this mixed-species biofilm, as well as an increased resistance to caspofungin [3]. *C. albicans* and *C. glabrata* are also known to form biofilms with bacteria from different host niches, usually relying on quorum-sensing mechanisms for the establishment of an interaction [20,155]. All the possible interactions between species increase the complexity of this vast field, pointing out the big clinical challenge of biofilm formation.

#### **4. Host Immune System Evasion**

Throughout infection, when the first line of defense has been breached by invasion into deeper tissues, *Candida* pathogens have to cope with cells of the host innate immune system. Interaction with the host immune system, and the ability to overcome it, is one of the main virulence features for fungal pathogens.

At early stages of infection, upon an interaction between *Candida* pathogens and epithelial cells, the former are recognized as invasive microorganisms by Pattern Recognition Receptors (PRRs) localized at host epithelial cell surfaces. PRRs interact with Pathogen Associated Molecular Patterns (PAMPs), such as β-1,3-glucan or chitin, present on microbial cells, thereby inducing a host response [156,157]. Epithelial cells, that are part of the innate immunity, not only secrete antimicrobial peptides, such as β-defensins and LL-37 [158–160], to try to control fungal infection, but also release proinflammatory mediators, such as chemokines and cytokines, triggering the recruitment of phagocytic cells, such as neutrophils, macrophages, and dendritic cells. These innate immune system cells also have PRRs in their surfaces, such as the C-type lectin receptor Dectin-1 [161], allowing the recognition of the invading pathogens and thereby inducing phagocytosis [162]. After phagocytosis, dendritic cells are responsible for the link between innate and adaptative antifungal immunity, presenting the *Candida*-specific antigens to naïve T-helper cells [163]. Therefore, in an immunocompetent host, this host–*Candida* interaction ultimately leads to the elimination of the pathogen. Otherwise, in immunocompromised individuals, a persistent infection, such as chronic mucocutaneous candidiasis, candidemia and/or persistent visceral candidiasis might be established.

Immune interaction can be translated in distinct spectrums, from avoidance of recognition by host immune cells to escaping or surviving immune attack. Masking PAMPs on the cell wall to avoid recognition, macrophage activation, and consequent phagocytosis is a common strategy of fungal pathogens during interaction with immune cells [164]. Generally, yeasts' cell wall is composed of a carbohydrate-rich inner layer and a protein-rich outer layer of heavily mannosylated proteins and phospholipomannan [165]. The outer layer acts as a shield of immunostimulatory components of the inner layer, such as β-1,3-glucan or chitin, playing a key role in protection and evasion from immune recognition [166]. β-1,3-glucan is the main polysaccharide present in the cell wall of *C. albicans*, *C. glabrata*, and other pathogenic *Candida* species, and is a key PAMP recognized by the host immune system [167]. Recognition of β-1,3-glucan by Dectin-1 receptor prompts phagocytosis by macrophages and neutrophils [161].

In order to avoid immune recognition, *C. albicans* resorts to cell wall remodeling, effectively masking β-glucan from the cell surface. The first reported case of active PAMP masking by *Candida* species was reported by Ballou et al. (2016) [167]. *C. albicans* was shown to mask β-glucan in response to lactate [167], which is a relevant physiological metabolite present in *Candida* niches, such as the vaginal tract and blood, or produced by the host microbiota [168], with which *Candida* interact. Lactate-mediated β-glucan masking is modulated by a signaling pathway associated with the G-protein coupled receptor Gpr1 and the transcription factor Crz1 [167]. This pathway reduces *C. albicans* uptake by macrophages and decreases the inflammatory response (TNFα and MIP1α) and neutrophil recruitment [167]. The participation of lactate was also observed later in low oxygen environments [169]. *C. albicans* was found to mask β-glucan upon oxygen deprivation, hindering recognition by Dectin-1 of polymorphonuclear leukocytes (PMNs). This was seen to modulate PMN responses, crippling phagocytosis, action of extracellular DNA traps, and reactive oxygen species (ROS) production [169]. Interestingly, β-glucan masking was prolonged by the build-up of lactate levels produced by PMNs [169]. Later, another study reported that hypoxia promotes β-glucan masking in *C. albicans* [170]. Hypoxia-induced masking is dependent on mitochondrial function and cAMP-protein kinase A (PKA) signaling, leading to reduced macrophage phagocytosis and cytokine (IL-10, RANTES, and TNF-α) production [170].

Changes in carbon source result in cell wall modifications with correspondent changes in virulence and immune properties [108,171]. As mentioned before, the cell wall is a complex structure with not only glucan, but also mannans, phosphomannans, and chitin [157], although distinct *Candida* species display different glucan exposure and mannan complexity [166]. β-glucans and chitin are located in the inner-most layer, while mannans are present in the outer layer [172–174]. Because of such structure, mannan plays an important role in reducing immunogenic exposure of β-glucan [175], but coordinated chitin and glucan exposure has also been reported to occur in *C. albicans* [176–178]. Moreover, cell wall structure and mannans affect virulence in different ways in *C. albicans* and *C. glabrata* [179–181]. Recently, mannan structure was found to affect glucan exposure in both *C. albicans* and *C. glabrata*, albeit in distinct ways. Deletion of mannosyltransferase family genes was associated with loss of negatively-charged acid-labile mannan and less efficient glucan masking in *C. albicans* (e.g., Δ*cgmnn2*), while in *C. glabrata* increased glucan exposure density was associated with mutants displaying shorter backbones (e.g., Δ*cgmnn1* and Δ*cganp1*) [166]. Previously, another study had shown that the β-1,6-mannosyltransferase encoded by *C. albicans MNN10* is required for backbone synthesis and influences immune recognition [182]. Absence of Mnn10 results in reduced *C. albicans* virulence, enhanced antifungal immunity by T helper cells, and increased recruitment of monocytes and neutrophils. Reinforcing the notion of a complex interplay among cell wall polysaccharydes, mannosyltransferase activity was also associated with β-1,3-glucan masking from Dectin-1 recognition and modulatory action of cytokine production by macrophages [182]. Another study showed how *C. albicans* cell wall responds to immune attacks by NETs [176]. β-glucan exposure and enhanced Dectin-1 recognition is dependent on fungal-pathogen crosstalk, as this response is dependent on neutrophil NET-mediated damage and fungal signaling cascades based on the MAP kinase Hog1. Cell wall structure in response to a neutrophil attack was found to affect more than one component, as Hog1 signaling leads to chitin deposition via the chitin synthase Chs3 and posterior cell wall remodeling via Sur7 and Phr1. Accordingly, with the enhanced immune recognition by Dectin-1 after a NET-mediated attack, macrophage cytokine response was also increased [176].

Much like *C. albicans*, *C. glabrata* resorts to cell wall remodeling in order to avoid the host's immune system, although the underlying mechanisms are mostly unknown. As indicated by increased TNFα secretion and increased efficacy of pathogen killing by macrophages, deletion mutants with disturbed cell wall integrity and altered accessibility of PAMPs caused a stronger inflammatory response. *C. glabrata* deletion mutants lacking cell surface-associated proteases (yapsins) or mutants with defective protein glycosylation were related with a stronger inflammatory response by macrophages [13,181]. Nevertheless, mutations affecting mannan, but not those affecting glucan or chitin, were found to

reduce the uptake of *C. glabrata* cells by murine macrophages, suggesting that mannose side chains or mannosylated proteins are ligands recognized by macrophages [183].

Knowledge regarding the PRRs responsible for *C. glabrata* recognition by macrophages is limited. Notwithstanding, as in *C. albicans* infections, C-type lectin receptors are thought to be involved in *C. glabrata* recognition by the host immune system. Specifically, dectin-1 and dectin-2, which recognize cell wall β-glucan, and mannan and β-glucan respectively, have been reported to be involved in the recognition of this pathogen [184,185].

Unlike *C. albicans* that put effort in escaping the immune system, it is hypothesized that inducing the recruitment of macrophages to the site of infection in vivo is part of the *C. glabrata* immune system evasion strategy [21]. *C. glabrata* infection did not substantially activate any MAPK pathway, including Erk1/2 (Extracellular signal-regulated kinases), SAPK/JNK (Stress-activated protein kinases/Jun amino-terminal kinases), and NF-κB signaling. Accordingly, it was found that upon infection of macrophages, the only cytokine significantly induced is GM-CSF, whereas the induction of other proinflammatory cytokines (TNF-α, IL-1β, IL-6, IL-8, and IFN-γ) is low [21,65,186]. GM-CSF is a potent activator of macrophages and induces differentiation of precursor cells as well as the recruitment of macrophages to sites of infection. This may explain the enhanced tissue infiltration of mononuclear cells, but not neutrophils, observed in vivo [186]. Considering the ability of *C. glabrata* to survive and replicate within macrophages, it is therefore speculated that persistence within macrophages is a possible strategy of immune evasion in this pathogen [21].

Immune evasion by pathogens also entails escaping from the complement system, another pathway of the innate immune system that facilitates phagocytosis. To evade immune response, several pathogens were shown to sequester or bind complement regulators, such as factor H (FH) [187–190]. *C. albicans* expresses the glucose transporter Hgt1 that binds FH, therefore reducing complement regulatory activity and limiting phagocytosis and killing by neutrophils [191]. SAP proteases produced by *C. albicans* not only cause tissue damage [192], but also contribute to immune evasion, as Sap2 is able to cleave antimicrobial peptides and complement proteins [193,194]. *C. albicans* Sap2 also cleaves FH and its receptors (CR3 and CR4) on macrophages, thus limiting their activation [195]. Additionally, *C. albicans* was also seen to bind yet another complement regulator (vitronectin) to modulate the hosts innate response [196]. The *C. albicans* pH-regulated protein Pra1 was recently implicated in the modulation of immune response [197,198]. It was seen to cleave the complement component C3, blocking the complement effector function and interfering with killing by neutrophils [197]. Pra1 is also implicated in adaptive immune response, as it binds to mouse CD4<sup>+</sup> T cells and reduces cytokine (IFN and TNF) secretion and antigen stimulation [198].

Ultimately, many yeast cells are engulfed by macrophages, hence survival and replication or subsequent escape remain important features. Upon phagocytosis, the phagosome carrying the ingested microorganism, fusion with a lysosome is one central antimicrobial mechanism of macrophages [164]. In the phagolysosome, fungal pathogens have to survive the harsh environment, typically characterized by carbon source limitation, production of reactive oxygen and nitrogen species, and acidification of the phagosomal compartment [199–201]. Mature phagolysosomes are normally strongly acidified, inducing antimicrobial effector mechanisms such as the activity of hydrolytic enzymes. Nonetheless, both *C. albicans* and *C. glabrata* are able to actively limit phagosome maturation in macrophages to prevent acidification and limit hydrolytic attack [21,202]. Environmental alkalinization by amino acid use as carbon sources, which results in ammonia extrusion, has been acknowledged as a strategy of these pathogens to actively raise phagosome pH [199,203,204]. Moreover, extracellular pH-raising triggers the yeast–hyphal switch in *C. albicans* [199,204]. Regarding carbon source availability, macrophages actively deprive pathogens of glucose, and therefore alternative carbon sources must support fungal growth. Accordingly, genes coding for enzymes of glyoxylate cycle, gluconeogenesis, and β-oxidation of fatty acids were found to be upregulated in both *C. glabrata* and *C. albicans* cells ingested by macrophages [13,18,205]. Very recently, the glyoxylate cycle gene *ICL1* was demonstrated to be crucial for the survival of *C. glabrata* in response to macrophage engulfment [206]. Disruption of *ICL1* rendered

*C. glabrata* cells unable to utilize acetate, ethanol, or oleic acid and conferred a severe attenuation of virulence in the mouse model of invasive candidiasis [206]. Further, genes of the methylcitrate cycle, which is important for the degradation of fatty acid chains and which allows the use of lipids as alternative carbon sources, were also found to be upregulated in *C. glabrata* [13]. Downregulation of protein synthesis as well as upregulation of amino acid biosynthetic pathways and amino acid and ammonium transport genes are features of the nitrogen deprivation faced by fungal cells inside the macrophages [13].

Although being mainly known by macrophage escape, strategies to counteract oxidative and nitrosative stress have been described in *C. albicans*. For instance, flavodoxin-like proteins are part of the antioxidant response of this species by reducing ubiquinone, which acts as a membrane antioxidant [207]. Expression of four flavodoxin genes (*PST1, PST2, PST3, YCP4*) is required for *C. albicans* virulence and resist neutrophil attack [207]. Other than oxidative burst modulation, *C. albicans* was also reported to modulate the nitrosative stress exerted by macrophages [208]. Nitric oxide production is dependent on the enzyme nitric oxide synthase, which utilizes arginine as a substrate. By increasing chitin exposure, *C. albicans* induces the host arginase-1 enzyme, which competes with the nitric oxide synthase enzyme and prevents the conversion of L-arginine to nitric oxide [208]. Interestingly, the use of amino acids as carbon source is much more prominent in *C. albicans* than other fungi [199] and amino acid metabolism has been associated with more than one mechanism of phagocyte escape/survival. Arginine was also found to play a role in hyphal development upon phagocytosis by macrophages, thus contributing to escaping the phagosome [209]. Additionally, in poor glucose conditions, *C. albicans* excretes amino acid-derived ammonia that increases external pH and interferes with the acidification of the phagosome [204,210], as mentioned before. Moreover, the transcription factor Stp2 involved in amino acid acquisition, is required to prevent phagosome acidification [204,209] based on the SPS amino acid sensing system [211]. Regarding the contribution of nitrosative stress to macrophage defense against *C. glabrata*, it is known that this pathogen induces only low NO production by murine macrophages [13].

*C. glabrata* shows increased tolerance to oxidative stress when compared to other yeasts, including *Saccharomyces cerevisiae* and *C. albicans* [212], mostly associated with the activity of the catalase Cta1, the superoxide dismutases Sod1/Sod2 and the glutathione and thioredoxin pathways [212–214]. Despite this, it is speculated that ROS play a minor role in killing *C. glabrata* cells, since experimental inhibition of ROS production in macrophages did not result in increased fungal survival [215]. In contrast to *C. albicans*, mobilization of intracellular resources via autophagy is an important virulence factor that supports the viability of *C. glabrata* in the phagosomal compartment of innate immune cells [23]. Phagocytosis induces peroxisome production in*C. glabrata* cells, which are then degraded via pexophagy, a specialized form of autophagy [23].

Moreover, besides carbon and nitrogen, trace elements such as iron are important for yeast growth. Macrophage-engulfed *C. albicans* upregulate a set of genes involved in iron homeostasis, for example, the ferric reductase genes *FRE3* and *FRE7*, as well as uptake systems for other trace metals, such as copper (*CTR1*) and zinc (*ZRT2*) [18]. *C. glabrata* is also able to sense and respond to iron limitation, although it has not been shown to use host iron-binding proteins as iron sources and is unable to use heme or hemoglobin. Instead, *C. glabrata* uses the siderophore-iron transporter Sit1, which is essential for utilization of ferrichrome as an iron source under iron-deficient conditions, and for iron-dependent survival in macrophages [216,217]. This iron acquisition system improves the fitness of *C. glabrata* when it is subsequently exposed to macrophages [216]. Recently, PI3K-kinase (encoded by *VPS34*) signaling was revealed to play a central role in *C. glabrata* iron metabolism and host colonization. However, the strategies by which *C. glabrata* gains iron within macrophages remain unknown [218].

While *C. glabrata* is most known for a persistence strategy and survival inside macrophages, due to its high-intrinsic stress tolerance, *C. albicans* is best known for active escape via hyphal growth and phagocyte piercing. In 2014, the model of macrophage piercing due to polarized growth of hyphae was challenged by two studies [17,219]. The study by Wellington et al. (2014) [219] showed that *C. albicans* escape is not exclusively due to disruption by hyphae. The pyroptosis pathway, a proinflammatory programmed cell-death process that is dependent on caspase-1, leads to interleukin production and macrophage lysis [220] and occurs concurrently with hyphae-mediated damage. Pyroptosis has been described to occur in response to intracellular bacteria, and thus is hypothesized to achieve the goal of destroying macrophages themselves in order to eradicate phagocyted pathogens [221]. *C. albicans* yeast-to-hyphae transition induces macrophage pyroptosis, therefore indicating that phagocyte damage caused by this pathogen is a more complex mechanism than originally postulated [17,219]. Accordingly, the study by Uwamahoro et al. (2014) [17] added more evidence of this escape mechanism. The triggering of pyroptosis is necessary for full macrophage damage upon hyphal formation and is activated in early phagocytosis, followed by a more robust hyphal formation that is indeed the main mechanism of macrophage killing in a later phase [17].

*Candida* thrives on multiple carbon sources to survive inside macrophages, but these depend on glucose for viability. Recently, Tucey et al. (2018) [222] demonstrated that *C. albicans* exploits this limitation by depleting glucose, and triggering rapid macrophage death, in vitro. Additionally, they showed that *C. albicans* infection promotes the disruption of host glucose homeostasis in vivo and verified that glucose supplementation improves host outcomes under systemic fungal infection [222]. Thus, depriving host immune cells for glucose seems to be one mechanism of *C. albicans* to induce phagocyte cell death and actively escape from those. Moreover, it was recently discovered that candidalysin is both a central trigger for Nlrp3 inflammasome-dependent caspase-1 activation via potassium efflux, and a key driver of inflammasome-independent cytolysis of macrophages and dendritic cells upon infection with *C. albicans* [223]. This study suggests that candidalysin-induced cell damage is a third mechanism by which *C. albicans* induces phagocyte cell death in addition to damage caused by pyroptosis and the growth of glucose-consuming hyphae [223]. *C. albicans*-induced activation of the Nlrp3 inflammasome, leading to secretion of IL-1β cytokine, is a crucial myeloid cell immune response needed for antifungal host defense [224]. Very recently, Rogiers and co-workers (2019) [225] identified candidalysin as the fungal trigger for Nlrp3 inflammasome-mediated maturation and secretion of IL-1β from primary macrophages. Therefore, the expression of candidalysin is speculated to be one of the molecular mechanisms by which hyphal transformation equips *C. albicans* with its proinflammatory capacity to prompt the release of bioactive IL-1β from macrophages [225].

A more controversial evasion mechanism, based on quorum-sensing stimulation of immune recognition, has been reported in both *C. albicans* and *C. glabrata*. In the case of *C. albicans*, white cells specifically (not opaque cells) secrete E,E-Farnesol, a stimulator of macrophage chemokinesis [226]. Farnesol secretion was found to increase macrophage migration and tissue infiltration [226]. This strategy has been associated with immune evasion by the concealing of the pathogens inside immune cells themselves, an environment where *Candida* species can survive.

Conversely to *C. albicans*, *C. glabrata* is incapable of hyphal differentiation, failing at this virulence trait. Although never observed for clinical isolates, interestingly, work by Brunke et al. (2014) [227] has shown that *C. glabrata* cells co-incubated during six months with macrophages, were able to produce pseudohyphae structures and evolve into a hypervirulent phenotype characterized by higher macrophage damage and faster escape.

Overall, the two more prevalent pathogenic yeasts, *C. albicans* and *C. glabrata*, follow two main different strategies to achieve the same ultimate goal: survive host immune response. On one hand, *C. albicans* hyphal forms actively pierce the membrane of macrophages as a mechanism of killing and escape. On the other hand, *C. glabrata*, that lacks morphological plasticity, survives and replicates within macrophages due to its remarkable ability to surpass its harsh environment, ultimately leading to macrophage lysis after several days due to fungal cells overload [21] (Figure 3).

**Figure 3.** Schematic representation of *C. albicans* and *C. glabrata* host immune system evasion. *C. albicans* actively escape from host immune system cells through hyphal growth and phagocyte piercing. *C. glabrata* is most known for a persistence strategy, surviving and thriving inside macrophages, ultimately leading to immune cells lysis due to fungal load.

#### **5. Conclusion and Perspectives**

The pathogenic yeasts *C. albicans* and *C. glabrata* are the two most prevalent *Candida* species isolated from candidiasis patients worldwide, yet they are phylogenetically, genetically, and phenotypically very different. Indeed, each species displays divergent virulence traits, indicating differential adaptation to the human host.

A greater number of virulence mechanisms has been described in *C. albicans*, most of which are not known to occur in *C. glabrata*. Yeast-to-hyphae dimorphism is one of the most striking divergent features. Hyphal morphology is associated with several key *C. albicans* traits (tissue invasion, biofilm formation, immune evasion), but is absent in *C. glabrata*. The result is that *C. glabrata* must have acquired other molecular mechanisms to reach the same goals. While tissue invasion in *C. albicans* occurs by proteolytic enzyme secretion and hyphal penetration of host tissues, *C. glabrata*-induced tissue damage is quite negligent in comparison and is thought to occur via endocytosis induction by host cells.

A similar observation can be made regarding biofilm formation, where *C. albicans* expresses hyphal-specific adhesins and regulators required for adhesion, while *C. glabrata* biofilms are much less "bulky". Morphological dimorphism also supports noticeable differences in immune evasion, as hyphal development and phagosome piercing is the main phagocyte escape mechanism employed by *C. albicans*, while such a strategy is absent in *C. glabrata*, which rather survives (and thrives) in the phagosome.

Dissimilar traits between *C. albicans* and *C. glabrata* have been identified, and such attributes have been extensively studied in *C. albicans*. However, the molecular mechanisms specific to *C. glabrata* that allow such a disparate species to cause human candidiasis demand further scrutiny. Indeed, the up rise of *C. glabrata* as a key fungal pathogen can only be prevented with specific therapeutic options that match its specific virulence traits.

**Author Contributions:** M.G., P.P., M.C., D.P., R.V. and M.C.T. contributed to the literature review and manuscript drafting, each author in their specific field of expertise. M.G. wrote the manuscript and prepared the illustrative figures. M.C.T. conceived the manuscript structure, supervised its writing, and reviewed all contents thoroughly.

**Funding:** Work conducted in this field was supported by "Fundação para a Ciência e a Tecnologia" (FCT) (Contracts PTDC/BBB-BIO/4004/2014 and PTDC/BII-BIO/28216/2017, PhD grants to MG, PP, and MC). Funding received by iBB-Institute for Bioengineering and Biosciences from FCT-Portuguese Foundation for Science and Technology (UID/BIO/04565/2013) and from Programa Operacional Regional de Lisboa 2020 (Project N. 007317) is acknowledged.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Review* **Small Noncoding Regulatory RNAs from** *Pseudomonas aeruginosa* **and** *Burkholderia cepacia* **Complex**

#### **Tiago Pita, Joana R. Feliciano and Jorge H. Leitão \***

iBB-Institute for Bioengineering and Biosciences, Departmento de Bioengenharia, Instituto Superior Técnico, Universidade de Lisboa, Av. Rovisco Pais, Torre Sul, Piso 6, 1049-001 Lisbon, Portugal; tiagopita@tecnico.ulisboa.pt (T.P.); joana.feliciano@tecnico.ulisboa.pt (J.R.F.)

**\*** Correspondence: jorgeleitao@tecnico.ulisboa.pt; Tel.: +351-218417688

Received: 31 October 2018; Accepted: 23 November 2018; Published: 27 November 2018

**Abstract:** Cystic fibrosis (CF) is the most life-limiting autosomal recessive disorder in Caucasians. CF is characterized by abnormal viscous secretions that impair the function of several tissues, with chronic bacterial airway infections representing the major cause of early decease of these patients. *Pseudomonas aeruginosa* and bacteria from the *Burkholderia cepacia* complex (Bcc) are the leading pathogens of CF patients' airways. A wide array of virulence factors is responsible for the success of infections caused by these bacteria, which have tightly regulated responses to the host environment. Small noncoding RNAs (sRNAs) are major regulatory molecules in these bacteria. Several approaches have been developed to study *P. aeruginosa* sRNAs, many of which were characterized as being involved in the virulence. On the other hand, the knowledge on Bcc sRNAs remains far behind. The purpose of this review is to update the knowledge on characterized sRNAs involved in *P. aeruginosa* virulence, as well as to compile data so far achieved on sRNAs from the Bcc and their possible roles on bacteria virulence.

**Keywords:** cystic fibrosis; *Pseudomonas aeruginosa*; *Burkholderia cepacia* complex; small noncoding regulatory RNAs; pathogenicity

#### **1. Introduction: What are Bacterial sRNAs?**

Regulation of gene expression in bacteria is achieved by a diversified system of multiple regulators, from proteins to RNA molecules. Riboswitches, T Boxes, and small noncoding regulatory RNAs (sRNAs) are among the RNA molecules or sequences that regulate the expression of genes. sRNAs were first identified in *Escherichia coli* in the 1960s as part of a large group of transcriptional regulators [1,2]. Since then, sRNAs have been shown as key regulators in bacteria [3], being capable of driving the fastest responses in the bacterial cell [4]. Ranging from 50 to 400 nucleotides long in size, sRNAs have been found as encoded widespread in the bacterial genomes, mainly in intergenic regions. For example, in enterobacterial genomes, sRNAs are estimated to represent approximately 200 to 300 genes, corresponding to ~5% of the number of protein-encoding genes [4,5].

sRNAs can be multipurpose regulators, targeting a wide range of molecular structures such as DNA/chromatin, proteins, other RNA molecules, and metabolites [6]. These regulatory sRNAs are therefore involved in the regulation of diverse cellular processes, including DNA assembly, plasmid replication, phage development, transcription, translation, peptidoglycan synthesis, and bacterial virulence [1,6]. In addition, sRNAs are also known for playing a role in cellular metabolism, like carbon and amino acid metabolism, iron homeostasis, quorum sensing (QS), and biofilm formation; stress responses including acid, osmotic, and oxidative stress responses; adaptation to growth conditions like high temperature and changes in molecular oxygen availability; as well as in mechanisms related to bacterial pathogenesis [7].

Although the vast majority of the so far characterized sRNAs do not encode proteins, a few can be partially translated originating small peptides, collectively known as dual-function small regulatory RNAs. These sRNAs can exert their regulation by acting as an RNA molecule or as a peptide. The peptide can play a similar regulatory role or a distinct one, capable of acting in a distinct metabolic pathway [8].

The best described sRNAs act by antisense base pairing with a specific target mRNA. These sRNAs can be cis- or trans-encoded, depending on their genome location in relation to the mRNA target [6]. Cis-encoded antisense RNAs share a full complementarity with their target mRNAs, often leading to the formation of a near perfect sRNA-mRNA duplex. Distinctly, trans-encoded antisense sRNAs only establish a partial complementarity with their target mRNAs, forming a partial duplex. Due to their lower specificity, trans-encoded sRNAs usually have multiple targets, being part of a more complex regulatory network in the cell. In both cases, the interaction between the antisense sRNA and target mRNA can affect translation negatively or positively [9].

The base pairing of the sRNAs can occur at different regions of the target mRNA, such as the 3 untranslated region (3 -UTR), the coding region, or, like most of the cases, at the 5 untranslated region (5 -UTR) where the ribosome accommodation usually occurs. The result of the interactions between the sRNA and the target mRNA can also differ, often leading to the suppression of gene translation by inhibition of ribosome interactions and/or by the induction of degradation of the target mRNA. However, activation of gene translation has also been reported. In this case, the interaction of the sRNA with the translation initiation region of the mRNA leads to a restructure of the mRNA conformation, exposing the previously occluded ribosome biding site (RBS) [6,10]. sRNAs can also act as stabilizers of target RNAs, avoiding their degradation. In addition, sRNAs can actually lead to the cleavage of mRNAs, resulting in one or two stable coding mRNAs [9].

Sometimes a sRNA can also become a target, when another RNA molecule (usually a mRNA) binds to the sRNA, trapping it. Known cases involve a sRNA that is constitutively expressed, and it is the expression of the trap-mRNA that will impair the regulatory function of the sRNA on its target [11,12]. Nevertheless, it has also been observed that some sRNAs bind to other sRNA, like the SroC of *Salmonella* that can sequester a given sRNA, impairing its activity [13].

The interaction between sRNAs and proteins usually leads to protein sequestration [9]. One of the best characterized target protein is CrsA, which has the ability to directly bind to the 5 regions of mRNAs, repressing their translation [14]. Some sRNAs act like a sponge, sequestering this protein. This interaction activates the regulation of multiple pathways, related to carbon starvation and glycogen biosynthesis, affecting the physiology and virulence of several pathogenic bacteria [14]. Moreover, sRNAs can also bind to proteins in a more complex way, producing more complex outcomes, such as the modulation of enzymatic activity by inhibiting, activating or modifying the protein activity. The best known example is *E. coli* 6S sRNA, which binds to the housekeeping form of RNA polymerase (σ70-RNAP), modifying its activity [3]. Nevertheless, the best characterized proteins that interact with sRNAs are the Hfq-like RNA chaperones, which play a major role in the regulatory mechanism of sRNAs base pair interactions, by stabilizing the RNA molecules and mediating the interaction between sRNAs and their targets.

Proteins of the Hfq family are conserved among bacterial genomes, being present in approximately one half of the sequenced organisms, a fact that certainly reflects its importance for the bacterial cell [15,16]. Hfq is a member of the (L)Sm protein superfamily and is present among the three domains of life. In bacteria, Hfq assembles as a homohexamer with a donut-like shape containing three distinct RNA binding sites on its surface, the proximal face, the distal face, and the convex circumferential rim of the ring. The proximal face of the ring recognizes uridine-rich sequences, a typical feature of 3 terminal tails of sRNAs. The distal face preferentially binds adenosine-rich sequences or repeated ARN motifs (composed of an adenosine, a purine, and any nucleotide), a characteristic

usually found in mRNAs, allowing the recognition of the sRNA-target regions on the 5 UTR of mRNAs [17,18]. The rim, which is rich in arginine residues, seems to promote RNA annealing, probably by guiding and facilitating the base pairing between the complementary strands of the sRNA and the target mRNA [19]. The mechanisms underlying the Hfq-mediated sRNA–mRNA interaction are quite diverse. Hfq increases the chance of the RNA molecules to meet by reducing their motility and flexibility; the protein can induce changes in RNA secondary structure by exposing the complementary regions of the RNA molecules, allowing the formation of a stable RNA double strand by base paring; the chaperone increases the local concentration and proximity of the pairs sRNA/mRNA by the specificity of its binding sites [20]. Moreover, Hfq is capable of accommodating multiple RNA molecules and interacting with proteins involved in RNA metabolism and translation, greatly improving the performance of sRNAs as regulators [17]. Hfq is now recognized as a major component of post-transcription regulatory mechanisms in bacterial cells. The output of Hfq-mediated sRNA-mRNA interactions is quite similar to those without mediation, occurring either an activation or a repression of gene expression, with or without mediated degradation by RNAse E [18,21]. Deletion of the gene encoding Hfq in several pathogens like *Salmonella* Typhimurium and *Pseudomonas aeruginosa* has resulted in a strong attenuation of virulence [22]. This evidences the major role played by Hfq-dependent sRNAs in virulence.

A recently characterized RNA chaperone, ProQ, is a RNA-binding protein of the ProQ/FinO domain superfamily, conserved and abundant among α-, β-, and γ-proteobacteria [23]. A study in *Salmonella* Thyphimurium revealed that the sRNAs associated to this chaperone possess extensive secondary structures, distinct from those that associate to Hfq. In addition, only 2% of the identified ProQ-interacting sRNAs also bind to Hfq [24]. The only full description of the mechanism of interaction between ProQ and a sRNA was done by Smirnov et al. [23], who described the ability of ProQ to stabilize the trans-encoded sRNA RaiZ, as well as the ability of the chaperone to mediate the interaction of RaiZ with its mRNA target *hupA*.

#### **2. Cystic Fibrosis Lung Infections**

Cystic Fibrosis (CF) is the most common and life-limiting autosomal recessive disorder among Caucasians, affecting ~70,000 individuals worldwide [25,26]. The CF disease is characterized by abnormal viscous secretions in several tissues with epithelia, like airways, pancreas, small intestine, liver, reproductive tract and sweat glands. The disease is due to mutations in the gene encoding the Cystic Fibrosis transmembrane conductance regulator (CFTR). Despite the potential damaging effect of *cftr* gene mutations on the affected organs, nowadays the great concern is centered on chronic bacterial airway infections that are responsible for 80 to 90% of mortality [26–28]. The expression of CFTR is much more pronounced in the airway's tissues, where a dysfunctional CFTR causes deficient cAMP-dependent chloride and bicarbonate secretion into airway secretions. Mucins are also secreted to the epithelial surface and then the pH decreases, a factor that impairs the host defense mechanisms. An inefficient inflammatory response, along with a defective mucus clearance, are the main causes of microbial infections in the airways of CF patients [29]. A wide array of bacteria can colonize the airways of CF patients, like *Staphylococcus aureus*, *Haemophilus influenzae*, *Stenotrophomonas maltophilia*, and *Mycobacteria* sp other than *M. tuberculosis*. Fungi like *Aspergillus fumigatus* have also been found in patients' sputum. All these microbes can contribute to the decline of lung functions [30]. However, the major threats to CF patient's survival are *P. aeruginosa*, bacteria of the *Burkholderia cepacia* complex (Bcc), and *Achromobacter xylosoxidans.* These bacteria can persist in the airways and provoke chronic infections that are the main cause of morbidity and early mortality of CF patients. *P. aeruginosa* is the leading cause of CF patients infections, being responsible for chronic infections in approximately 80% of CF adults [30,31]. Progresses recently achieved in the development of aggressive early eradication of *P. aeruginosa* led to an enhancement on prognosis [32]. Bacteria of the Bcc are the second major microbial threat to CF patients. Despite their lower incidence (only 3%–4% of CF patients are infected by Bcc), these bacteria are especially feared due to the easy transmission among the patients, the extensive

antibiotic resistance, and the risk of cepacia syndrome, a fulminating pneumonia that leads to patient death in a short period of time [33,34].

This review focuses on the current knowledge of sRNAs from the CF major pathogens *P. aeruginosa* and Bcc. Since sRNAs are major regulatory molecules involved in bacteria virulence, we review in this work some sRNAs involved in virulence of *P. aeruginosa*, as well as the present scarce knowledge on Bcc sRNAs and their possible roles on bacteria virulence [35].

#### **3. sRNAs of** *Pseudomonas aeruginosa***, the Major CF Pathogen**

*P. aeruginosa* is a Gram-negative species recovered from a wide range of environmental niches. The species is an opportunistic pathogen to animals and humans and a serious threat to public health [36]. *P. aeruginosa* is responsible for severe infections not only among CF patients, but also in other immunocompromised patients and patients with severe burn injuries [37]. It is still to be fully elucidated how an environmental species has become a major cause of nosocomial infections worldwide. However, it is known that the overuse and misuse of antibiotics have led to the selection of multiple antibiotic resistant strains, against which very few therapeutic options are available [32,36]. In a review, Potron et al. [38] showed that some markers of acquired resistance to antibiotics were found in strains isolated all over the world. The ability of this bacterium to produce a wide range of opportunistic infections and to originate strains resistant to various antibiotics can be attributed to its large genome (typically with more than 6 Mbp), containing a particularly high proportion of regulatory genes, as well as a large number of genes involved in the catabolism, transport, and efflux of organic compounds [32]. Although *P. aeruginosa* infections can be eradicated during acute lung infections in most of the cases, when a chronic infection is established, its eradication is hardly achieved [39]. The most striking examples of infections by *P. aeruginosa* are chronic lung infections in CF patients, which occur in over 60% of adult patients [40]. When invading the CF lung, *P. aeruginosa* has to surpass a heterogeneous, hostile, and stressful environment, being exposed to osmotic stress due to the viscous mucus, oxidative and nitrosative stresses due to host defenses, sublethal concentrations of antibiotics, and the presence of other microorganisms [32].

Acute infections by *P. aeruginosa* are among the best characterized; it is known that the expression of their virulence genes is controlled by extremely complex, interweaving regulatory circuits and multiple signaling systems. To achieve a successful acute infection, *P. aeruginosa* produces an impressive array of virulence factors, such as the Type 3 Secretion Systems (T3SS) to secrete various exotoxins; the quorum sensing (QS) systems to control numerous important secreted components including pyocyanin, elastase, cyanide, and rhamnolipids, exhibiting a motile phenotype [32]. On the other hand, *P. aeruginosa* virulence factors that are important in chronic infections are still to be fully elucidated. It is known that in chronic infections, *P. aeruginosa* strains became less inflammatory and less cytotoxic, and also lose structures responsible for adherence and motility, like the flagellum and pili. Other characteristics include the appearance of mucoidy, due to alginate production, the emergence of highly antibiotic resistant variants, changes in lipopolysaccharide structure including altered lipid A and loss of O-antigen, and alterations in quorum sensing [39]. The GAC system network was pointed as controlling the reversible transition from acute to chronic infections, incorporating the two-component regulatory system GacA/GacS, two other sensor kinases RetS/LadS, the small regulatory protein RsmA, and sRNAs RsmZ and RsmY [32]

To successfully infect their hosts, *P. aeruginosa* must achieve fast and precise regulation of its metabolism. sRNAs are one of the best means to achieve such a fast and tight regulation, and this is most probably why these bacteria possess a so vast array of those posttranscriptional regulators. A total of 573 sRNAs are known to be expressed by the strain PAO1 and 233 sRNAs by the strain PA14 [41]. PAO1 is a moderately virulent strain isolated from a wound, widely used as a reference strain in comparison to the highly virulent isolate PA14, a representative of the most common clonal group found worldwide [41]. 126 sRNAs are common to both strains. However, an in silico study evidenced that the strains share more sRNAs in their genomes than the ones detected [42]. This means

that, although some sRNAs are encoded in the genome of both strains their expression could be strain-specific or environmental-dependent [42]. For the PAO1 strains, 149 sRNAs were already annotated and 117 were experimentally validated, the remaining predicted by similarity searching. In the case of the PA14 strains, 136 sRNAs are annotated and 66 were experimentally validated (data obtained from the sRNAs database BSRD [43]).

Although the number of characterized sRNAs is remarkably reduced compared to the total of predicted sRNAs in *P. aeruginosa*, it is still a huge number when compared to other organisms. A boost in the characterization of sRNAs from *P. aeruginosa* strains PAO1 and PA14 took place in recent years. At least one half of the characterized sRNAs were found to be dependent or mediated by the Hfq chaperone (Table 1). Among the sRNAs characterized are the RsmY and RsmZ, both were involved in the switch from the motile to the sessile modes of living, modulating the expression of T3SS and T6SS [44]. However, only the role of RsmZ in cell motility and biofilm formation was proved. Like RsmY and RsmZ, RsmW is also involved in the RsmA regulation, being implied in the regulation of biofilm formation [45,46]. Moreover, ErsA and SrbA are two sRNAs that also play a role in the regulation in biofilm formation. ErsA is part of the response to envelope stress and exerts its regulatory role by modulating the activity of the checkpoint bifunctional enzyme AlgC with phosphomannomutase/phosphoglucomutase activities and expression of the AmrZ regulon [47]. SrbA has also a role in the pathogenesis of *P. aeruginosa* PA14, although the mechanisms underlying such regulation are still to be elucidated [48]. RgsA and NrsZ are both important in swarming motility. Both are regulated by two-component regulatory systems (TCS). NrsZ is regulated by the TCS NtrB/C and perform its role through its involvement in the modulation of *rhlA* expression, and thus controlling the production of rhamnolipid surfactants [49]. RgsA is regulated by the GacS/GacA TCS, having as targets *fis* and *acpP*, leading to the regulation of pyocyanin synthesis [37].

Las, RhL and PQS are the best known *P. aeruginosa* quorum sensing systems. The Pseudomonas quinolone signal (PQS) plays a central role in the quorum sensing network in *P. aeruginosa*, linking the three systems. sRNAs PrrF1/2, ReaL, and PhrS are major players in PQS modulation, all of them playing a role in the expression of virulence genes. PhrS is involved in the regulation of expression of PqsR, a transcriptional regulator of genes involved in PQS and pyocyanin synthesis [50]. ReaL acts by positively regulating *pqsC* expression, in a response modulated by the Las system [51]. PrrF1/2 works as an iron sensor, leading to an interruption of expression of genes encoding iron-binding proteins in low iron conditions. The transcription activator *antR* is also repressed, avoiding the degradation of anthranilate, a precursor of PQS synthesis. [52,53]. The sRNA PrrH that was not linked to *P. aeruginosa* virulence also plays an important role in iron availability, by heme homeostasis [54]. In this context, the CrcZ sRNA also seems to be involved in iron regulation. Like RsmY/Z/W, CrcZ also sequesters proteins, in this case the Hfq chaperone. Although it was initially characterized as a regulator of carbon catabolite repression, recent evidence points out CrcZ as a PrrF regulator, interfering with the Hfq-mediated repression of *antR* [53,55]. Two other sRNAs were also characterized recently, PaiI plays a pivotal role in anaerobic growth conditions and in denitrification, and PesA seems to play a role in *P. aeruginosa* pathogenesis, being involved in Pyocin S3 modulation and resistance to UV radiation [56,57]. Table 1 summarizes the *P. aeruginosa* sRNAs already characterized according to the strain from which they were studied. A full description of the best characterized sRNAs playing a role in *P. aeruginosa* virulence is described below.



*Int. J. Mol. Sci.* **2018**, *19*, 3759

location related to the mRNA target.

#### *3.1. RsmY and RsmZ*

Chronic infection seems to be the preferred mode of infection by pathogens like *P. aeruginosa*, a premise for its survival and persistence as a population. Since the cellular mechanisms involved in acute or chronic infection are different, the cell is equipped with a complex regulatory network to switch its state. In the case of *P. aeruginosa*, the Csr/Rsm system plays a major role in this switch [52]. This system controls a large variety of physiological processes, like carbon metabolism, virulence, motility, quorum sensing, siderophore production, and stress responses [45]. The system comprises a major translational regulator, the protein RsmA, which negatively regulates mRNA targets by binding to sites containing critical GGA motifs present in the 5 -UTR of mRNAs targets [45] This binding represses the translation of regulons necessary for establishing chronic infections, as the T6SS, quorum sensing systems, exopolysaccharide production, biofilm formation, and iron homeostasis [70,71]. On the other hand, RsmA exerts a positive and indirect regulation of the mechanisms linked to acute infection, like the expression of genes associated with surface motility, T3SS, type IV pili, as well as systems that operate through the cAMP/virulence factor regulator (Vfr) route [72]. RsmA appears to control T3SS gene expression by increasing *exsA* translation through an undetermined mechanism. ExsA is the master transcription factor of all T3SS genes [73]. So, RsmA regulates the switching from planktonic (acute infection) to biofilm (chronic infection) phenotype. RsmA activity is negatively regulated by the two-component regulatory system GacA/GacS that induces the expression of sRNAs antagonist of RsmA, including the RsmY and RsmZ in *P. aeruginosa* (Figure 1) [74]. These two last sRNAs have a secondary structure with several unpaired GGA motifs that sequester RsmA proteins, preventing biding to their targets, and thus enhancing the expression of specific RsmA regulons [45].

Although GacA is a major regulator of *rsmY* and *rsmZ* expression, these sRNAs can also be regulated by alternative regulators, and despite the redundancy, they can have different regulators. It was observed in *P. aeruginosa* that *rsmY* transcription steadly increases during cell growth, whereas *rsmZ* is induced harshly during the late exponential phase of growth [45]. Using swarming motility as a model, Jean-Pierre et al. [61] showed that these sRNAs are differentially regulated depending on the selected growth conditions (planktonic versus surface–grown cells), and that *rsmZ* regulation does not implicate the response regulator GacA in swarming cells. Furthermore, these authors observed that RsmY/Z expression influences swarming motility via the protein HptB, which acts as a negative regulator of these sRNAs and that they do not strictly converge to RsmA [61]. Before a biofilm can be formed, RsmZ is eliminated by the action of ribonuclease G, activated by the TCS BfiSR [45]. Other mechanisms of regulation of RsmY and RsmZ have been discovered recently. AlgR, a TCS response regulator important for *P. aeruginosa* pathogenesis in both acute and chronic infections, seems to affect the antagonizing action of RsmY and RsmZ on RsmA [75]. SuhB regulates the motile–sessile switch in *P. aeruginosa* through the Gac/Rsm pathway and c-di-GMP signaling, regulating multiple virulence factors like the T3SS, swimming motility, T6SS, and biofilm formation [44]. Li et al. [44] reported that this regulation is mediated by the GacA-RsmY/Z-RsmA. MgtE, a magnesium transporter, was demonstrated to play a role in the inhibition of the T3SS transcriptional activator ExsA. The negative regulation is mediated by expression of RsmY and RsmZ, most likely though the TCS GacA/GacS [73]. The polynucleotide phosphorylase (PNPase), which is involved in the regulation of multiple virulence factors like T3SS, T6SS and pili biosynthesis, plays an important role in RsmY/Z stability. This stabilizing ability is due to the PNPase KH-S1 domains that bind to sRNAs stabilizing them, providing a way of pathogenicity regulation by PNPase [76]. Recently, Miller et al. [45] characterized a new sRNA in *P. aeruginosa*, the RsmW. Unlike RsmY and RsmZ, this sRNA is not transcriptionally activated by GacA. RsmW was shown to be upregulated in nutrient-limiting conditions, biofilms, and at higher temperatures [45].

**Figure 1.** *P. aeruginosa* functionally characterized sRNAs and their involvement on virulence regulatory pathways. Grey box—characterized sRNAs (with red border- Hfq biding); Green box—indirect regulator, Light Blue box—direct regulator; Coral box—direct target; Yellow box—indirect target; Dark Blue—virulence factors; ↓: positive control; ⊥: negative control; dashed lines: indirect control.

#### *3.2. ReaL*

ReaL is a highly conserved *P. aeruginosa* sRNA recently characterized by Carloni et al. [51]. ReaL was shown to be a relevant element for *P. aeruginosa* pathogenesis. In the infection model *Galleria mellonella*, deletion of the ReaL encoding gene impaired *P. aeruginosa* PA14 virulence, while the sRNA overexpression resulted in a hypervirulent phenotype. ReaL is involved in the quorum sensing regulatory network architecture, interlinking the Las and PQS QS systems. Both Las and PQS are part of a major network played at least by two additional systems, the integrated quorum sensing system (IQS) and the Rhl QS system [77]. This entire network is closely related with the expression of virulence factors in *P. aeruginosa*. The Las system is composed by the transcriptional regulator LasR, its cognate autoinducer molecule N-3-oxo-dodecanoyl-homoserine lactone (3-oxo-C12-HSL), and the 3-oxo-C12-HSL synthase LasI [51]. PQS, positively regulated by the Las QS system, is an essential mediator of the shaping of the population structure of *P. aeruginosa* and of its responses and survival in hostile environmental conditions [78,79]. PQS synthesis proceeds through the condensation of an anthranilate molecule with a fatty acid to produce the 2-heptyl-4-quinolone (HHQ), which is then hydroxylated by the PqsH enzyme, whose expression is positively regulated by the LasR3-oxo-C12HSL to produce PQS [77,80]. It is also known that ReaL plays a role in the interconnection between both systems. Actually, ReaL is negatively regulated by LasR, and because it is a positive post-transcriptional regulator of the *pqsC* gene, a decrease of PQS will occur. ReaL seems to contribute to the fine co-modulation of HHQ/PQS synthesis [51]. The PQS system plays a vital role in biofilm formation and production of virulence factors such as pyocyanin, elastase, PA-IL lectin, and rhamnolipids [77]. Therefore, ReaL expression triggers the production of some virulence factors, like biofilm formation and pyocyanin synthesis, by a Las-independent way. In addition, ReaL plays a negative role in swarming motility. The upregulation of ReaL expression is observed in the conditions found when colonizing the human lung, like a temperature of 37 ◦C and low oxygen concentration [51].

#### *3.3. ErsA*

ErsA is the first sRNA from *P. aeruginosa* that appears to be embedded in the envelope stress response, a critical transduction pathway that impacts pathogenesis in *P. aeruginosa*. ErsA expression depends of the envelope stress responsive sigma factor σ22, the major player in the

envelope stress-signaling pathway like σ<sup>E</sup> in *S.* Typhimurium and other enterobacteria. ErsA directly exerts a negative post-transcriptional regulation on the virulence associated *algC* gene encoding the bifunctional enzyme AlgC [68]. AlgC plays a central role in exopolysaccharide biosynthesis, generating the sugar precursors mannose 1-phosphate and glucose 1-phosphate, necessary for the synthesis of polysaccharides like alginate, Pel, Psl, LPS, and rhamnolipids, key components of the biofilm matrix [81]. A fine-tuned regulation exerted by ErsA on AlgC is thought to influence the dynamics of exopolysaccharide biosynthesis, underlying the development of biofilm matrix. This regulation is crucial when the bacterial cell is under envelope stress. The repressive role of ErsA on *algC* expression occurs mainly at the translational level by base pairing with the RBS in a Hfq dependent way [68]. Furthermore, Zhang et al. [82] found that ErsA represses the expression of the OprD porin, the major channel for the uptake of basic amino acids, peptides, and the carbapenem antibiotic, evidencing a role played by this sRNA in antibiotic resistance.

#### *3.4. PrrF*

PrrF sRNAs are encoded in *P. aeruginosa* genome by two highly homologous genes, *prrf1* and *prrf2*. These sRNAs are encoded in tandem sequences, separated by 95 nt, a characteristic exclusive of this species. The entire *prrF* region also encodes another sRNA, PrrH, transcribed from the 5 end of *prrf1* to the 3 end of *prrf2* [83]. Infection experiments in a murine model using a deletion mutant on the *prrF* locus showed that this genetic locus plays a pivotal role in *P. aeruginosa* virulence in acute murine lung infection. The encoded sRNAs are also important for iron and heme homeostasis [64]. PrrH seems to play a role in heme homeostasis and on the expression of virulence factors. However, PrrH is not vital for acute murine lung infection, as shown by Reinhart et al. [54]. Iron acquisition is essential for *P. aeruginosa* virulence and biofilm formation, highlighting the role of regulatory systems on the maintenance of iron homeostasis, with this sRNA playing a critical role in bacteria survival in the host [64]. PrrF is responsible for the iron-sparing response, an effect observed under conditions of iron limitation. PrrF exerts its regulatory activity by repressing mRNAs encoding iron-containing proteins like the iron cofactored superoxide dismutase SodB, a putative bacterioferritin, a heme-cofactored katalase and succinate dehydrogenase. Iron "sparing" is essential when the intracellular iron concentration is low [54]. Although iron is an essential metallonutrient, when it accumulates it becomes toxic, due to its ability to catalyze the formation of reactive oxygen species via Fenton reactions. The regulation of such a homeostasis is mediated by the ferric uptake repressor protein Fur [84]. Fur has the ability to bind to iron in iron-enriched environments, becoming an active repressor of PrrFs, leading to a more extensive use of iron by the proteins otherwise downregulated by PrrFs [64]. PrrF sRNAs also repress the expression of AntR, a transcription activator of genes involved in the conversion of anthranilate to catechol. By this way, the anthranilate is channeled into the biosynthetic pathway of the PQS, a system that leads to the expression of several virulence factors as already mentioned above [64,85]. Furthermore, Sonnleitner et al. [53] added another sRNA to the already extensive armory of *P. aeruginosa*, the CrcZ sRNA, which impairs the binding of PrrFs to *antR*. Since the riboregulation played by PrrFs is mediated by Hfq, the CrcZ functions like a sponge, binding to Hfq with a higher affinity compared to *antR*, inhibiting its binding. This mechanism leads to the de-repression of *antR*.

#### *3.5. PesA*

PesA, characterized by Ferrara et al. [57], is encoded in the pathogenicity island PAPI-1, being expressed by the highly virulent strain PA14, as well as by several clinical isolates obtained from CF patients. The *P. aeruginosa* strain most commonly used for research, PAO1, does not encode PesA. The sRNA is expressed under temperature and oxygenation conditions resembling those found in the CF lung, namely 37 ◦C and low oxygen. PesA was shown to be important for *P. aeruginosa* pathogenicity in CF bronchial cells. PesA is involved in the post-transcriptional regulation of genes involved in S-type pyocin production, like pyocin S3. Pyocin S3, like other soluble pyocins, has a killing activity

that causes cell death by DNA cleavage. Pyocins are produced by more than 90% of the *P. aeruginosa* strains, helping *P. aeruginosa* in niche establishment, protecting the bacteria from competition [57,86]. By deleting *pesA*, an increased susceptibility to UV irradiation and to the fluoroquinolone antibiotic ciprofloxacin was observed, correlated with a decreased level of pyocin expression. A similar effect was observed when deleting the gene encoding pyocin S3. Although contradictory, PesA seems to play an important role in the response to DNA damage by promoting the synthesis of pyocin S3 [57]. The mechanism underlying this response is not yet clear.

#### **4. The** *Burkholderia cepacia* **Complex and the Emerging Knowledge on Its sRNAs**

When compared to *P. aeruginosa*, respiratory infections caused by Bcc species have a low prevalence on CF patients. However, bacteria of this complex are still one of the most feared pathogens and represent a higher risk for those patients.

*Burkholderia cepacia* complex (Bcc) is a group of opportunistic human pathogens presently composed of 23 closely-related species [87]. Apart from being a major threat for CF patients, Bcc bacteria can also cause lethal infections among chronic granulomatous disease (CGD) patients, immunocompromised patients (e.g., HIV infected), cancer patients, and other chronic patients [88]. *B. cenocepacia* and *B. multivorans* remain the predominant Bcc species worldwide that are responsible for infections in CF patients [89,90]. Bcc species possess genomes larger than those of *P. aeruginosa* strains, usually with a length of more than 7 Mbp. This large genome is thought to confer the bacterium an enormous ability to overcome antimicrobial therapies, as well as the host immune response, being in constant evolution inside the host lung [91].

The defective mucus clearance of the CF lung is the perfect niche for a persistent infection by *P. aeruginosa*, as well as by Bcc bacteria that tends to occur after *P. aeruginosa* colonization, usually superseding it [92,93]. Moreover, Bcc infections are easily propagated among CF patients. Bcc bacteria possess an extraordinary resistance to antibiotics [94,95]. About twenty percent of CF patients infected with Bcc bacteria develop the cepacia syndrome, a necrotizing pneumonia, often accompanied by septicemia, leading to a rapid, irreversible and deadly decline of the respiratory function [96,97].

Bcc virulence resembles several traits of *P. aeruginosa* virulence, due both to the fragility of the host and to the bacterium extraordinary evolutionary development. Once the colonization of the respiratory tract is established and the bacteria are attached to epithelial cells, they have to overcome a strong immunological response [98]. The successful invasion of host internal system is achieved by mechanisms of penetration like paracytosis and invasion as a biofilm [99,100]. Among the virulence factors expressed by Bcc species the extracellular lipase, metalloproteases and serine proteases, and some structures of the bacterial surface such as flagella, pili, and the lipopolysaccharide (LPS) can be highlighted [96,100]. Exopolysaccharide (EPS) production, namely Cepacian, also represents a major virulence factor, important for biofilm formation and to protect Bcc from the host defense machinery, as well as from antimicrobial therapy [101–104].

Like in *P. aeruginosa*, several noncoding RNAs have been associated with regulatory mechanisms involved in Bcc virulence, as is the case of the modulation of the adaptation to the host changing environment, the adherence to, and invasion of host cells, as well as the replication. However, the Bcc noncoding virulome (sRNAs playing a role in virulence) is still poorly understood compared with *P. aeruginosa*.

#### *4.1. Discovering Bcc Noncoding Transcriptome*

A pioneer systematic search for sRNAs in Bcc was first described by Coenye et al. in 2007 [105], who predicted an array of putative noncoding RNAs (ncRNAs) from *B. cenocepacia* J2315, using *R. solanacearum* GMI1000 genome as a reference. Two-hundred-and-thirteen ncRNA genes, ranging in size from 53 to 1243 nt, were predicted and described. The expression of only four of these sRNAs was confirmed by microarray experiments using total RNA from cells grown on a 10% (*w*/*v*) CF

sputum [105]. The expression of one sRNA has confirmed by real-time PCR; none of these putative sRNAs were functionally characterized.

In order to understand the mechanisms underlying the early stage of Bcc infections, Drevinek et al. [106] performed a microarray-based transcriptomic analysis of *B. cenocepacia* grown from sputum recovered from CF patients. The strain was found out to be indistinguishable from *B. cenocepacia* J2315, a strain involved in multiple fatalities. The genes found to be upregulated were mainly involved in antibiotic resistance, the antioxidant response to reactive species, iron metabolism, and virulence factors such as flagella and metalloproteases. A thorough analysis of transcripts originating from intergenic regions was carried out, leading to the identification of 88 upregulated and 126 downregulated sequences. However, no further characterization of these regions was performed to identify putative sRNAs.

In 2009, Yoder-Himes et al. [107] used Illumina RNA-seq techniques to understand the mechanisms of transcriptional regulation associated to *B. cenocepacia* environmental adaptation. Two strains, one isolated from soil (*B. cenocepacia* HI2424) and another from a CF patient (*B. cenocepacia* AU 1054), were grown under conditions mimicking each environment, and their transcriptomes were analyzed. In both conditions a higher number of genes was transcribed in strain HI2424 than in strain AU 1054. When grown under CF-like conditions, the strains exhibited differences in the expression of genes mainly from chromosome 1, while growth under soil conditions led to differences in the expression of genes mainly encoded in chromosomes 2 and 3. Regarding sRNAs, 13 were identified (ncRNA1 to ncRNA13), but only one seems to be induced under CF-mimicking conditions. Those sRNAs have been predicted to be highly structured. The expression of 4 sRNAs was confirmed by northern blot, allowing the authors to assume that the sequences were indeed expressed in vivo. The majority of these molecules are conserved among the *B. cenocepacia* strains and the Bcc (Table S1).

Effective response to reactive oxygen species by bacteria is an important feature regarding the adoption of disinfection measures, as well as during exposure to oxidative burst by the host defenses. To evaluate the response to reactive oxygen species (ROS), Peeters et al. [108] used microarray transcriptomic analysis and qPCR to analyze whole cell responses of *B. cenocepacia* J2315 sessile cells exposed to H2O2 and NaOCl. The profile of genes up- and downregulated was found to be similar to that of planktonic *P. aeruginosa* PAO1 cells exposed to similar conditions. Many genes involved in prevention (counteracting or repairing of the damage from oxidative stress) were found to be upregulated, as well as some genes involved in the synthesis and assembly of flagella. Moreover, several transcripts from intergenic regions were found, 39 and 56 were upregulated upon exposure to H2O2 and NaOCl, respectively, whereas 54 and 68 were found to be downregulated. To avoid false positives, the authors selected as putative noncoding RNAs the regions whose expression pattern is different from the flanking genes. Eleven and 20 intergenic regions were selected as putative noncoding RNAs upregulated with H2O2 and NaOCl, respectively. Transcripts from intergenic regions IG1\_2935724 and IG1\_3008003 were upregulated in both conditions and their expression was confirmed by qPCR. The ncRNA4 and ncRNA6, found by Yoder-Himes et al. [107], are located in those intergenic regions, thus confirming their expression. A match was found between the IG1\_2935724 and the secondary structure of 6S RNA consensus structure. Seven out of the 11 sRNAs upregulated upon exposure to H2O2 were also found significantly altered in *B. cenocepacia* J2315 cells of CF sputum [106].

Coenye et al. [109] used microarrays to study transcriptomic differences between planktonic and sessile cells of *B. cenocapacia* J2315 exposed to chlorhexidine. The authors found that sessile cells are more resistant to chlorhexidine (0.015%) than planktonic cells. Furthermore, sessile cells highly expressed genes encoding efflux systems related to drug resistance, as well as membrane-associated proteins and regulators. The results suggest that sessile cells engaged a global expression program to evade the biofilm since an adhesin was downregulated and genes encoding chemotaxis and motility-related proteins were upregulated. This study also allowed the identification of several intergenic regions upregulated in the presence of chlorhexidine. After discarding the regions with an expression level similar to the closest coding sequences (CDS) and the ones whose probes overlapped

the coding sequences, 19 intergenic regions were suggested as putative sRNAs. As presented in Supplementary Table S1, these regions are largely conserved among *B. cenocepacia* strains, less conserved among the Bcc members, and only one of them is conserved in the *B. pseudomallei* group. The majority of the predicted sRNAs seems to have a stable secondary structure, however no significant match was found in the Rfam database [109].

Previous work from our research group led the experimental identification of 24 sRNAs from *B. cenocepacia*, based on co-purification of total RNA and the chaperone Hfq [110]. Expression of these sRNAs was confirmed by northern blot and characterized in silico. The only consistent results were related to Bc7, for which the *hemB* was predicted as mRNA target, and the *Salmonella* Paratyphi RybB as homolog. In addition a cis-encoded sRNA, named h2cR, was also previously reported by our research group as being involved in the regulation of the *hfq2* gene that encodes an Hfq-like chaperone in Bcc [111].

The genome of *B. cenocepacia* KC-01, a strain isolated from the coastal saline soil, was recently sequenced by Ghosh et al. [112]. Several potential sRNAs were identified by in silico analysis of the genome, and the expression of seven putative sRNAs was confirmed (Bc\_KC\_sr1-7). Bc\_KC\_sr1 and Bc\_KC\_sr2 were upregulated in response to iron depletion by 2,2 -bipyridyl. Bc\_KC\_sr3 and Bc\_KC\_sr4 were induced under the presence of 60 μM H2O2 in the culture medium. Alterations on the temperature and incubation time also induced the expression of Bc\_KC\_sr2, 3, and 4. Searches within the RFAM and BSRD databases led to the identification of candidate738 of *B. pseudomallei* D286, tmRNA and 6S RNA as homologs of Bc\_Kc\_sr4, 5, and 6, respectively. Interestingly, this group of sRNAs is extensively conserved among members of the Bcc and *B. pseudomallei* groups (Supplementary Table S1). Several targets were predicted for these sRNAs, like Fe-S cluster and siderophore biosynthesis, ROS homeostasis, porins, transcription, and translation regulators.

#### *4.2. A Deeper Approach: Bcc sRNAs Expressed under Biofilm Formation Conditions*

Based on the importance of biofilm formation to antibiotic resistance, Sass et al. [113] analyzed the transcriptome of *B. cenocepacia* J2315 grown in biofilms by differential RNA-sequencing (dRNA-seq). dRNA-seq differs from RNA-seq on a selective sequencing of primary transcripts. The techniques, used for the first time for a Bcc organism, allow mapping of the transcription start sites (TSS) based on the difference of primary and processed transcripts ends. Primary transcripts carry a 5 triphosphate end, while processed transcripts like tRNA and rRNA, carry a 5 monophosphate. Using the 5 P-dependent terminator exonuclease (TEX) to degrade the 5 monophosphate, it is possible to distinguish between primary and processed transcripts [114]. As a complementary approach, Sass et al. also carried out a global RNA-seq (gRNA-seq) to obtain a large coverage of the whole transcript length and the 3 end of transcripts, usually lost by dRNA-seq of longer transcripts. The additional use of 5 RACE allowed the authors to end up with 2089 genes annotated as expressed under biofilm conditions and to identify alternative start codons for some genes and novel protein sequences. A total nine sRNAs with homologs present in the Rfam database and other seven match putative sRNAs that were experimentally identified in previous studies. The sRNAs with homologs in the Rfam database were the 6S RNA; two phage-related regulatory RNAs located on genomic island BcenGI9; two conserved regulatory motifs; the SAH riboswitch located upstream of BCAL0145; an adenosylhomocysteinase; the *sucA* RNA motif located upstream of *sucA* (BCAL1515), an enzyme of the citric acid cycle; and four sRNAs from the family named "toxic small RNAs", whose expression in *B. cenocepacia* have already been confirmed by northern blot but with unknown functions. From the transcript sequences containing a TSS in intergenic regions that did not match with a gene and without a hit in the Rfam database, the authors selected only those obeying to the following criteria: strong transcription initiation with a coverage >300 reads in dRNA-Seq data, a defined 3 end in dRNA-Seq data or a transcript appearing short (<500 nt), and truncated or missing in gRNA-Seq data. Upon selection, the sequences were compared with the experimentally confirmed sRNAs already published by other authors. Six putative sRNAs matched those found by Yoder-Himes et al. [107],

five of them upregulated on soil conditions and one on cystic fibrosis sputum. Another transcript matched one sRNA previously identified by using co-purification with Hfq [110]. The expression of these sRNAs in other studies, together with their overexpression under conditions of biofilm formation, suggests a role for these sRNAs on biofilm regulation.

Sass et al. [115] refined their analysis on putative sRNAs found in association with biofilm formation. From a total of 148 TSS found in intergenic regions, 41 transcripts were classified as rho-independent (RIT), and 82 transcripts as derived from 5 UTR of the downstream gene. Fifteen sRNAs were selected based on the following criteria, the number of starts at TSS (≥250), the z-score (<−1), and the conservation on Bcc (≥13 strains). 3 RACE and Northern blot analysis were used to confirm the length of sRNAs. The expression of 14 sRNAs was quantified by qPCR in different experimental conditions, all compared with planktonic cultures. Twelve sRNAs were upregulated in biofilms. Some sRNAs were slightly upregulated by osmotic or pH stress. Nutrient starvation also induced the expression of 12 sRNAs, as well as the expression of Hfq chaperone. Five of these sRNAs were also more expressed under glucose-rich medium.

Further work performed by these authors led to the identification of ncS63 as highly expressed under conditions of low-iron. This sRNA is located upstream of BCAL2297, which encodes for the hemin-uptake protein HemP, whose expression is under the regulatory control of the Fur repressor. The predicted targets of ncS63 are related to iron homeostasis like *sdhA*, a target of RyhB from *E. coli*. RyhB is a sRNA involved in iron homeostasis regulated by the Fur repressor, and a possible functional analogue of ncS63. Recently, RyhB was hypothesized to play an important role as a mediator of bacterial resistance to multiple antibiotics and stresses [116]. Further investigation is needed to check if ncS63 is a RyhB functional homolog. The overall prediction of sRNAs targets using CopraRNA and RNApredator led to similar clues about the mechanisms that can be under regulation of these sRNAs, like transcriptional regulators, carbon compound transport and metabolism, and cell envelope components such as outer membrane proteins and porins. The transcriptional regulator BCAL1948 has been predicted to be a target of nine sRNAs, exhibiting particularly extensive complementarity to ncS11. sRNAs containing double-hairpin were predicted to target genes involved on the metabolism and transport of amino acids, carbohydrates, and aromatic compounds. The ncS16 has several hits of genes related to outer membrane and cell envelope components, and genes for transport of inorganic compounds, whereas the best hits for ncS63 were genes involved in energy production and detoxication of reactive oxygen species. Regarding the putative sRNAs derived from 5 UTRs, most likely they have a cis-regulatory function, since some 3 ends of adjacent genes are homologs of genes that possess cis-regulatory structures in other species. A possible trans-regulatory function remains to be investigated.

#### *4.3. ncS35, a Functionally Characterized B. cenocepacia sRNA*

While in other organisms sRNAs are being characterized for decades, their characterization in Bcc is starting. ncS35 was identified by Sass et al. [115] in the course of *B. cenocepacia* J2315 biofilm transcriptomic analysis, being the first trans-encoded characterized sRNA from *B. cenocepacia*. The expression of ncS35 was found to be upregulated in cells grown as biofilm and in minimal medium compared to planktonic cells grown in rich medium. Increased aggregation, higher cell metabolic activity, higher growth rate, and increased susceptibility to tobramycin were described for deletion mutant cells. Moreover, an upregulation of the phenylacetic acid and tryptophan degradation pathways was observed in the mutant cells, and the first gene of the tryptophan degradation pathway was predicted to be a putative target of ncS35. This sRNA seems to lead to an attenuation of the metabolic and growth rates, which can be a way of cells to protect themselves against stress conditions. A slow growth rate is observed on *P. aeruginosa* biofilms in CF patient's sputum, and it is also known that bacteria on slow growth rate, even in planktonic cultures, have an increased resistance to antibiotics, including tobramycin [117,118]. Slow growth rates and metabolic activity are characteristics of persister cells and probably the cause of drug resistance. Since the effect of antibiotics is mainly

due to the inhibitory action of some metabolic pathways, slow-growing cells are somehow protected from antibiotics and are more likely to develop drug resistance [119]. The occurrence of persister cells in Bcc infections is known, and it has also been observed in *B. cenocepacia* J2315 after treatment with tobramycin [120]. A better characterization is required to fully understand the roles of ncS35 on *B. cenocepacia* pathogenicity, antibiotic resistance, and possibly on the formation of persister cells.

#### *4.4. Compiling the Bcc Predicted sRNAs*

Although a few sRNAs from Bcc organisms are functionally characterized, a large number of putative sRNAs have been described by several research groups. This information was gathered, and we present on Supplementary Table S1 all predicted and confirmed sRNAs, as well as the intergenic regions, which can include candidate noncoding RNAs described so far for *B. cenocepacia*. Predicted sRNAs for which data is somehow confusing and uncertain, or the methods applied were not accurate enough, were not considered. *B. cenocepacia* was chosen because it is the one of the Bcc species with the noncoding transcriptome better characterized. A total of 167 putative sRNAs were included on Table S1, seven from *B. cenocepacia* KC-01, 13 from *B. cenocepacia* AU 1054, and 147 from *B. cenocepacia* J2315, the majority of them identified by Sass et al. [115]. Putative sRNAs are unevenly distributed throughout the three chromosomes: 67.5% are located on chromosome 1, 28.1% on chromosome 2, and 4.4% on chromosome 3. The length of the replicons on the J2315 strain is 3,870,082, 3,217,062, and 875,977 bp, respectively [121]. The chromosomes 1 and 2 are approximately equally sized; however, chromosome 1 accommodates twice the sRNAs compared with the second chromosome. This suggests that chromosome 1 sRNAs are more relevant for core functions of bacterial metabolism (e.g., cell division). On the other hand, since the regulation of the other two replicons encode genes more related to accessory functions, the regulatory noncoding RNAs encoded in those replicons are probably expressed in response to specific conditions that may have not been assessed yet [121]. Wong et al. [122] used a transposon mutant pool and identified 383 possible essential genes on J2315 strain, 90% of which was present on chromosome 1. Such a strategy can not only enable the discovery of crucial coding genes expressed on different conditions, but might also unveil the Bcc noncoding transcriptome, allowing the discovery of important genes related to some features like virulence and antibiotic resistance.

Regarding the sequence conservation, 93% of the putative sRNAs listed in Table S1 are conserved or semiconserved among the species, 70% among the Bcc bacteria and 24% also in the *B. pseudomallei* group. Although the *B. pseudomallei* group is phylogenetically distant and those bacteria display a distinct pathogenicity and epidemiology, a high level of sequence homology is found in almost 25% of the putative sRNAs identified so far [123]. It is also interesting to access the conservation of the putative sRNAs in each chromosome, since a higher conservation can be observed on chromosome 1, much more extensive than chromosome 2 and 3. A huge difference is observed on the conservation of sRNAs on the Pseudomallei group, with a conservation (and semiconservation) of ~25% on sRNAs from chromosome 1, whereas sRNAs from chromosome 2 and 3 exhibited less than 4% conservation. This is also evidence that Bcc comprises more closely related species than the Pseudomallei group, mainly regarding the accessory functions.

Several sRNAs were identified by more than one experimental approach and given different names. For instance, ncRNA4, ncS17, and Bc\_KC\_sr6 are the same sRNA, which was found on the intergenic region IG1\_2935724. This sRNA shares homology with the 6S RNA, a regulator of RNA polymerase (RNAP) that is widespread among members of the Bacteria domain. This sRNA was found as expressed in distinct situations like soil conditions, biofilm formation and response to ROS [124]. Moreover, an upregulation of the 6S RNA was found in response to oxidative stress, further implying it with a broader role than just on core metabolism, such as on the protection of the cell in the host environment [108]. The possible role of 6S RNA in pathogenesis or in host survival has been described for several bacteria [124]. The homologs ncRNA6 and ncRI12 are both located on the intergenic region IG1\_3008003 and are probably the same sRNA. This sRNA was overexpressed during biofilm

formation and upregulated under soil conditions, exposure to chlorhexidine, and ROS, suggesting a role on protection against stress conditions. ncS06, homolog of ncRNA7, was found to be expressed under soil conditions and biofilm formation. Prediction of ncS06 mRNA targets revealed several peptidase-encoding genes [115]. ncS11 is a homolog of ncRNA13 and was found to be upregulated on soil conditions, as well as on biofilms, rich medium and starvation conditions [115]. ncS11 has an extensive complementarity to BCAL1948, whose expression was found to be downregulated under biofilm formation. BCAL1948 is a protein of the LysR-type transcriptional regulator (LTTR) family, whose global transcriptional regulators can be found in diverse organisms. The regulated genes are hypothesized to be involved in a wide range of processes like cell division, metabolism, nitrogen fixation, oxidative stress responses, motility, quorum sensing, attachment, secretion, virulence, and toxin production [125].

#### **5. Conclusions**

There is a clear contrast between the knowledge currently available about sRNAs characterization, regulatory pathways and interactome in *P. aeruginosa* and Bcc bacteria. In *P. aeruginosa*, fourteen sRNAs have already been functional characterized. Some *P. aeruginosa* sRNAs are involved in the pathogenicity and virulence of the organism, like biofilm formation, iron metabolism, quorum sensing regulation, response to stress, and host cell invasion. On the other hand, in Bcc bacteria the first biological function of a sRNA is now being described. This sRNA is ncS35, a growth regulator that seems to do not interfere with the virulence of Bcc bacteria. Despite the scarce characterization of sRNAs in Bcc, over the last 10 years at least 167 putative sRNAs were identified in *B. cenocepacia* as being expressed on CF sputum, biofilm formation, response to oxidative stress, among others specific conditions.

In the peculiar environment of the CF lungs, it is interesting to note that both *P. aeruginosa* and Bcc bacteria can cause infections presenting a high degree of intrinsic and acquired resistance to antibiotics. Both opportunistic pathogens have two of the largest genomes known among prokaryotes, which were predicted to encode a huge number of sRNAs. The functional characterization of these sRNAs is only beginning in those CF pathogens. However, the already established or predicted roles of some sRNAs in modulating cellular processes linked to pathogenesis, suggests their yet unexplored importance for the physiology and pathology of these pathogens, as well as their likely influence in the outcome of *P. aeruginosa* and Bcc infections. The assessment of the functional roles of sRNAs in the host–pathogen interaction is expected to provide additional fundamental knowledge for the development of next-generation antibiotics inspired by sRNAs and their targets.

**Supplementary Materials:** Supplementary materials can be found at http://www.mdpi.com/1422-0067/19/12/ 3759/s1.

**Author Contributions:** Conceptualization, J.H.L.; Writing—Original draft preparation, T.P. and J.R.F.; Writing—Review and editing, J.H.L., T.P. and J.R.F.

**Funding:** This work was funded by Fundação para a Ciência e Tecnologia (Portugal) through Projects PTDC/BIA-MIC/1615/2014 and UID/BIO/04565/2013.

**Acknowledgments:** Tiago Pita. acknowledges a doctoral grant (PD/BD/135137/2017) and Joana R. Feliciano. acknowledges a postdoctoral grant (BL109/17 IST-ID) from FCT.

**Conflicts of Interest:** The authors declare no conflicts of interest.

#### **References**


© 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Review* **Catching a SPY: Using the SpyCatcher-SpyTag and Related Systems for Labeling and Localizing Bacterial Proteins**

#### **Daniel Hatlem, Thomas Trunk, Dirk Linke and Jack C. Leo \***

Bacterial Cell Surface Group, Section for Evolution and Genetics, Department of Biosciences, University of Oslo, 0316 Oslo, Norway; daniel.hatlem@ibv.uio.no (D.H.); thomas.trunk@ibv.uio.no (T.T.); dirk.linke@ibv.uio.no (D.L.)

**\*** Correspondence: j.c.leo@ibv.uio.no; Tel.: +47-228-59027

Received: 2 April 2019; Accepted: 26 April 2019; Published: 30 April 2019

**Abstract:** The SpyCatcher-SpyTag system was developed seven years ago as a method for protein ligation. It is based on a modified domain from a *Streptococcus pyogenes* surface protein (SpyCatcher), which recognizes a cognate 13-amino-acid peptide (SpyTag). Upon recognition, the two form a covalent isopeptide bond between the side chains of a lysine in SpyCatcher and an aspartate in SpyTag. This technology has been used, among other applications, to create covalently stabilized multi-protein complexes, for modular vaccine production, and to label proteins (e.g., for microscopy). The SpyTag system is versatile as the tag is a short, unfolded peptide that can be genetically fused to exposed positions in target proteins; similarly, SpyCatcher can be fused to reporter proteins such as GFP, and to epitope or purification tags. Additionally, an orthogonal system called SnoopTag-SnoopCatcher has been developed from an *S. pneumoniae* pilin that can be combined with SpyCatcher-SpyTag to produce protein fusions with multiple components. Furthermore, tripartite applications have been produced from both systems allowing the fusion of two peptides by a separate, catalytically active protein unit, SpyLigase or SnoopLigase. Here, we review the current state of the SpyCatcher-SpyTag and related technologies, with a particular emphasis on their use in vaccine development and in determining outer membrane protein localization and topology of surface proteins in bacteria.

**Keywords:** autotransporter; covalent labeling; bacterial surface protein; SpyCatcher; topology mapping; virulence factor

#### **1. Introduction**

The peptide bond is essential to biology. This secondary amide bond is produced by the condensation of the carboxyl group of one amino acid residue with the α-amino group of the following residue and results in the polymerization of amino acids to produce polypeptides. This central reaction is catalyzed by the ribosome and the sequence of the polypeptide chain is defined by the genetic code. In proteins, the peptide bond is sometimes referred to as the eupeptide bond. In addition, peptide bonds can be found in non-ribosomal peptides, which are secondary metabolites such as antibiotics produced by non-ribosomal peptide synthases [1]. A minority of proteins contain a second class of peptide bonds, referred to as isopeptide bonds. This is an amide bond similar to the peptide bond, but it is produced by the condensation of an amine group with a carboxyl group or a primary amide group located in amino acid side chains, rather than the main chain-building amino or carboxyl groups (Figure 1A).

In eukaryotes, the most prevalent example of isopeptide bond formation is the ubiquitination of proteins targeted for degradation by the 26S proteasome [2]. The isopeptide is formed by the C-terminal carboxyl group of ubiquitin and the side-chain amine of a lysine residue in the target protein. This reaction is catalyzed by E3 enzymes, or ubiquitin-protein ligases, which also impart specificity to the system [3]. Similarly, members of the small ubiquitin-like modifiers (SUMO) family, which regulate a number of cellular processes, are ligated to their target proteins via an isopeptide bond [4].

In bacteria, isopeptides are mostly known from surface proteins of Gram-positive bacteria. Two classes of proteins have notable isopeptide bonds: the microbial surface components recognizing adhesive matrix molecules (MSCRAMMs) and pili [5]. In pili, pilin subunits are covalently bound to each other via an intermolecular isopeptide bond between the carboxyl group of a C-terminal threonine and the ε-amine of a lysine in the next subunit, a reaction catalyzed by a pilin-specific sortase [6]. In MSCRAMMs and some pilins, isopeptides are formed spontaneously between residues within the same domain. These domains belong to the immunoglobulin fold superfamily. Isopeptide bond formation increases the stability of the proteins or pili and may contribute to stronger binding to ligands by MSCRAMMs, since in some cases the isopeptide only forms upon ligand binding to lock the protein in a bound conformation [7].

The SpyCatcher-SpyTag system was developed by the Howarth laboratory based on the internal isopeptide bond of the CnaB2 domain of FbaB, a fibronectin-binding MSCRAMM and virulence factor of *Streptococcus pyogenes*[8,9]. An internal isopeptide bond forms spontaneously in this domain between the ε-amine of lysine K31 and the side chain carboxyl of aspartic acid D117. The reaction is catalyzed by the spatially adjacent glutamate E77 (Figure 1B). The resulting isopeptide bond confers high stability to the CnaB2 domain [10]. The CnaB2 domain can be stably split into two components: a larger, incomplete immunoglobulin-like domain (termed SpyCatcher) of 138 residues (15 kDa) and a shorter peptide (SpyTag) of 13 residues (see Table 1 for peptide sequences). SpyCatcher contains the reactive lysine and catalytic glutamate, whereas SpyTag includes the reactive aspartate. The two components can still recognize each other with high affinity (0.2 μM) and the isopeptide can form between SpyCatcher and SpyTag to form a covalently bound complex (Figure 1C). Under experimental conditions relevant to life science research (room temperature, dilute protein concentrations), the reaction rates allow the bonds to form at high efficiency within minutes [9]. SpyTag in particular is equivalent in size to a number of epitope tags and can be genetically fused to a number of proteins and is able to react with SpyCatcher when inserted at the N- or C-termini of target proteins, as well as internal sites [9,11]. SpyCatcher itself can also be produced as a fusion protein, allowing the formation of covalently bonded protein partners that might otherwise be difficult to produce as protein fusions [12–14].

A similar system has been developed based on another Gram-positive surface protein, the pilus adhesin RrgA of *S. pneumoniae* [15,16]. The D4 domain of this protein is stabilized by an isopeptide forming between a lysine (K742) and an asparagine (N854), catalyzed by the spatially adjacent E803 [15] (Figure 1D). This domain was split into a scaffold protein called SnoopCatcher and a 12-residue peptide termed SnoopTag, which can spontaneously form a covalent isopeptide bond upon mixing [16] (Figure 1E). In contrast to SpyCatcher-SpyTag, the reactive lysine is present in SnoopTag and the asparagine in SnoopCatcher. This system is orthogonal to SpyCatcher-SpyTag; that is, SnoopCatcher does not react with SpyTag and SpyCatcher does not react with SnoopTag. This allows the use of both systems simultaneously to produce "polyproteams," programmed modular polyproteins, for use in biotechnological applications [16].

The Howarth lab further modified these technologies by making tripartite systems, where the isopeptide-forming lysine and aspartate/asparagine are located on separate peptides, and the catalytic glutamate is present on a larger scaffold protein [17]. This was first attempted with the SpyCatcher system to produce SpyLigase. Here, a second peptide containing the reactive lysine (KTag) was separated from SpyCatcher, which itself was modified to produce the stable SpyLigase protein containing the catalytic glutamate [18]. When SpyLigase was mixed with two proteins containing one of the reactive peptides each, SpyLigase was able to catalyze the fusion of the two tags. However, although SpyLigase could mediate the fusion of KTag and SpyTag located at both N- and C-terminal and even internal positions, the reactions had a ~50% efficiency at best and were dependent on specific buffer conditions and low temperature [16]. In contrast, the recently developed SnoopLigase system appears

more robust and efficient [19]. SnoopLigase was engineered similarly to SpyLigase, and catalyzes the isopeptide formation between the lysine of a modified SnoopTag (SnoopTagJr) and the asparagine in a second peptide termed DogTag (Figure 1F). This system can have efficiencies over 95% that are less sensitive to temperature and reaction conditions than SpyLigase. Furthermore, as SnoopTagJr and DogTag have relatively high affinity for SnoopLigase, immobilizing SnoopLigase allows washing away unconjugated reactants followed by elution of essentially pure fusion products [19]. The various Catcher-Tag systems and their development are summarized in Table 1.

Below, we review the various applications of the SpyCatcher/SnoopCatcher systems in biotechnology. Our emphasis will be on using this methodology to label surface-exposed proteins, especially for mapping the topology of outer membrane-embedded virulence factors of Gram-negative bacteria.

**Figure 1.** The SpyCatcher-SpyTag, SnoopCatcher-SnoopTag, and SnoopLigase systems. (**A**) Formation of an isopeptide bond. The primary amine of a lysine side chain condenses with either the side-chain amide of an asparagine (upper reaction) or the side-chain carboxyl of an aspartate (lower reaction) to produce the isopeptide bond, releasing either ammonia or water, respectively; (**B**) the isopeptide bond between SpyCatcher-SpyTag. The isopeptide formed by the reactive lysine (K31) in SpyCatcher and aspartate (D117) in SpyTag is shown in green, and the catalytic glutamate (E77) in cyan; (**C**) crystal structure of SpyCatcher-SpyTag. SpyCatcher is in blue and SpyTag is in magenta. The structures shown in panels b and c are based on the Protein Data Bank (PDB) entry 4MLI [20]; (**D**) the isopeptide bond between SnoopCatcher-SnoopTag, formed by lysine (K742) in SnoopTag and the asparagine (N854) in SnoopCatcher, is shown in red, and the catalytic glutamate (E803) is in blue; (**E**) crystal structure of SnoopCatcher-SnoopTag. SnoopCatcher is in yellow and SnoopTag in green. The structures shown in panels d and e are based on the RrgA D4 domain structure (PDB ID: 2WW8) [15]; (**F**) schematic of the function of SnoopLigase. A protein (in brown) containing SnoopTagJr (ST, in orange) and another protein (purple) with DogTag (DT, green) are mixed in the presence of SnoopLigase (yellow). The tags bind to SnoopLigase, which contains the catalytic glutamate (blue dot) and catalyzes the formation of an isopeptide between SnoopTagJr and DogTag (red dot). The ligated proteins can then be eluted from SnoopLigase.


<sup>1</sup> The reactive residues are shown in bold. <sup>2</sup> SpyLigase and SnoopLigase have two target peptides.

#### **2. Applications of the SpyCatcher-SpyTag System**

Being a quick and reliable coupling tool forirreversible peptide-proteinligation, the SpyTag-SpyCatcher system is ideal for a wide range of applications, ranging from increasing protein stability to antigen delivery during vaccination [14,23]. Before focusing on the use of the SpyTag-SpyCatcher system in the investigation of bacterial virulence factors, we give a short overview of those applications. For more detailed information, a list of publications and patents using the SpyTag-SpyCatcher and related technologies is accessible at the SpyInfo web page (available online: https://www.bioch.ox.ac.uk/howarth/info.htm) and the corresponding sequences and expression routes are listed in the SpyBank database [24].

A major application of the SpyTag-SpyCatcher system is the formation of so-called SpyRings in order to increase the intrinsic resilience of proteins to denaturation. SpyRings are generated by circularization of a single protein, which is accomplished by fusing an N-terminal SpyTag with a C-terminal SpyCatcher, or vice versa (Figure 2A). Enzyme circularization increases resistance to hyperthermal denaturation and aggregation, as well as alkali tolerance of individual enzymes without a loss in enzymatic activity [25–29]. Enzymes with a short distance between termini (<15 Å), and an active site which is not in direct proximity to one of the termini, are considered ideal candidates for protein circularization [30]. However, efficient SpyRing circularization and increased thermal stability have been shown for proteins where the termini are even farther apart [28].

Another frequently used application for the SpyCatcher-SpyTag system is the decoration of protein hydrogels (Figure 2B). Hydrogels have found application, for example as an artificial extracellular matrix material in medicine [31], tissue engineering [32,33], and cell culturing [34]. Hydrogels are polymeric materials engineered to resemble the extracellular environment of specific tissues with defined functional and structural properties. They have been used for decades as molecule delivery devices and as carriers for cells in tissue engineering due to their ability to mimic aspects of the native cellular environment [35]. However, hydrogels fail to fully imitate the complexity of biological systems. Fusing SpyCatcher to the polymeric material used in hydrogel synthesis allows the decoration with SpyTagged proteins post-hydrogelation to mimic specific microenvironments [36–39]. Thus, the SpyTag-SpyCatcher system is used as a quick and simple molecular tool for the simultaneous incorporation and presentation of different target molecules into and on hydrogels, thereby avoiding an otherwise laborious engineering process.

The SpyTag-SpyCatcher system has also been used for the modular assembly of proteins onto nanoparticles [40] (Figure 2C) and bacterial outer membrane vesicles [41] (Figure 2D; see also Section 3.3). The same system can be used for the in vivo encapsulation of enzymes fused to phage capsid proteins in order to create a protein nanocompartment [42] (Figure 2E) and for the decoration of virus-like

particles (VLPs) for antigen delivery to the immune system [43,44] (Figure 2F; see also Section 3.3), while successfully preserving the structure and the function of the assembled proteins.

**Figure 2.** Applications of the SpyCatcher-SpyTag system. (**A**) SpyRing: SpyCatcher and SpyTag are fused to the terminal ends of the protein of interest resulting in protein cyclization and thereby conferring an increased resilience to denaturation; (**B**) post-hydrogelation decoration of protein hydrogels: SpyCatcher is fused to the polymeric material used in hydrogel synthesis, which allows the posthydrogelation decoration with proteins of interest fused to SpyTag; (**C**–**F**): (**C**) bioconjugation of target proteins to nanoparticles, (**D**) outer membrane vesicles, (**E**) phage capsid proteins to create a proteinaceous nanocompartment, and (**F**) virus-like particles; (**G**) SpyCLIP: SpyTagged RNA-binding protein interacts with RNA and is covalently attached to beads with fused SpyCatcher for use in pull-down assays; (**H**) fluorescent protein labeling for use, for example, in microscopy (ST represents SpyTag sequence); (**I**) artificial multi-enzyme nanodevices for increased efficiency, stability, and reusability. Schematic of mesh-like nanodevice is shown on the left and an enzymatic reaction scheme on the right. S represents substrate; P represents product; A and B represent cofactors.

Recently, the SpyTag-SpyCatcher system was used for the characterization of protein–RNA interactions as an alternative for ultraviolet (UV) crosslinking and immunoprecipitation (CLIP) [45]. The CLIP method relies on the limited specificity of antibody–antigen interactions which cannot withstand harsh washing conditions resulting in insufficient purity of ribonucleoprotein complexes after immunoprecipitation. Therefore, additional gel purification steps are necessary to further purify these complexes, which results in loss of protein. The disadvantages of CLIP are the labor-intensive gel purification steps as well as a high number of false-positive signals due to non-specific interactions. In comparison, the SpyTag-SpyCatcher technology allows the method to be performed with beads, skips the gel purification steps altogether, and withstands harsh washing steps, thus reducing non-specific interactions. The improved method using the Spy technology was termed SpyCLIP and requires SpyCatcher being fused to beads for immunoprecipitation and fusion proteins of the protein of interest, in this case the RNA-binding protein, to SpyTag [45] (Figure 2G). UV-crosslinking and affinity purification (uvCLAP) [46] as well as gel-omitted ligation-dependent CLIP (GoldCLIP) [47], using the covalent HaloTag-HaloLink purification system for immunoprecipitation [48], are both improvements to the traditional CLIP protocol, omitting the labor-intensive gel purification steps. The advantage of the SpyCatcher system over those methods is the ability of SpyTag-SpyCatcher to withstand even harsher washing conditions than the biotin-streptavidin coupling used during uvCLAP, and the small size of SpyTag compared to the 33 kDa protein tag used in GoldCLIP.

The system has also found increasing usage in conventional and super-resolution microscopy [49–51] by providing an easy and reliable way to label a SpyTagged protein with SpyCatcher coupled to a fluorescent dye (Figure 2H). Additionally, the limited accessibility of large antibodies, classically used for detection of epitope-tagged molecules, can be avoided by using the relatively small SpyCatcher protein during the detection of SpyTagged targets.

The Spy technology can also be used for the targeting of chemically synthesized voltage-sensitive dyes to specific cells for optical measurement of voltage dynamics in living cells. To this end, SpyTag is linked via a polyethylene linker to voltage-sensitive dyes while expressing SpyCatcher on the target cells. These SpyTag–dye conjugates, termed Voltage Spy, display improved targeting, good voltage sensitivity, and fast-response kinetics [52].

The system can also be used for the spatial and functional coordination of enzyme functions, creating a microenvironment normally associated with biological compartmentalization and conferring some of the same benefits (e.g., a local increase in enzyme and metabolite concentrations). Linkage of SpyTag with SpyCatcher is utilized for enzyme organization, allowing construction of artificial multi-enzyme nanodevices with a controlled spatial arrangement for increased efficiency of enzyme cascades while maintaining or even increasing stability and reusability of biocatalysts [53] (Figure 2I).

#### **3. Using SpyCatcher-SpyTag to Investigate Bacterial Virulence Factors**

The SpyCatcher-SpyTag system has recently been used for investigating bacterial virulence factors. So far, these have been limited to studying surface proteins of Gram-negative bacteria, especially autotransporter proteins. Autotransporters, also called type V secretion systems, are a widespread family of secreted proteins from Gram-negative bacteria, and many of these mediate virulence-related functions [54]. These proteins, which are divided into several subclasses (type Va through type Ve), have an outer membrane-embedded β-barrel domain and an extracellular region or passenger, which harbors the specific activity of each protein. Although SpyCatcher-SpyTag has been used mostly for studying autotransporters, the methods described below would be applicable to other surface-exposed proteins, both in Gram-negative and Gram-positive bacteria.

The cell surface of Gram-negative bacteria is represented by the outer membrane (OM), which is a complex asymmetric lipid bilayer. The extracellular face of the OM is mainly composed of lipopolysaccharides and outermembrane proteins thatinclude several common virulence factors responsible for adhesion, mobility, and secretion, such as autotransporter adhesins, flagella, and type I-VIII secretion systems, among others [55,56]. Studying the expression, secretion, migration, and interactions of OM proteins often requires labeling by reporter proteins or other fluorophores. Labeling OM proteins, however, is challenging. Many fluorescent reporter proteins fail to mature in the periplasm [57]; thus, genetic fusions with fluorescent proteins are limited to a handful of options [58–61]. An alternative approach is non-covalent affinity-based labeling by using antibody–reporter protein fusions, affinity tags, or other high-affinity interactions, for example, colicins ColE9 and ColIa for labeling of the vitamin B12 transporter BtuB [62]. Alternatively, labeling methods using small organic molecules include amine-reactive fluorescence labeling [63,64], cysteine-reactive labeling [65], or site-specific labeling with unnatural amino acids, and tag-specific labeling [66,67]. Each of these techniques, however, has shortcomings. Non-covalent labeling requires high (nM) affinities for the label to remain associated with its target for a significant amount of time, which often makes such labeling unsuitable for time-resolved imaging on the minute scale. Small-molecule labeling often requires harsh treatment of the cells with reactive molecules, and site-specific orthogonal labeling requires co-expression of additional OM proteins that can disturb the system. In this section, we present a few recent examples showing how the SpyCatcher system can overcome many of these challenges by covalently labeling OM-bound virulence factors with high specificity, both with the purpose of studying the virulence factors and of exploiting their properties in vaccine development.

#### *3.1. Using SpyCatcher to Investigate Membrane Protein Topology and Secretion*

Trimeric autotransporter adhesins (TAAs), or type Vc secretion systems, constitute a group of surface-displayed virulence factors in Gram-negative bacteria that are responsible for adhesion to organic and inorganic surfaces. The prototypical TAA is *Yersinia* adhesin A (YadA), which consists of a C-terminal transmembrane β-barrel domain, where three protomers form a 12-stranded β-barrel, and an N-terminal trimeric passenger that is translocated through the β-barrel to the bacterial surface [68]. Once the passenger is translocated, it adopts a lollipop-like structure, forming a fibrous trimeric coiled-coil stalk that ends in an N-terminal globular head domain, responsible for adhesion and autoaggregation (Figure 3A). Autotransporters belonging to the type Va and Ve secretion systems initiate their translocation process via a hairpin intermediate, meaning that the part of the passenger most proximal to the β-barrel is initially inserted into the pore and then exported in a hairpin-like loop from the β-barrel until the entire passenger is outside the cell [69,70].

**Figure 3.** Using SpyCatcher to determine the surface topology of the *Yersinia* adhesin YadA. (**A**) Model of full-length YadA [71] and solid-state NMR structure of YadAM (PDB ID: 2LME); (**B**) schematic of the trimeric SpyTag-YadAMA354P construct with two secreted chains. The secreted chains fused to SpyTag are located on the bacterial surface and are able to bind to the SpyCatcher-sfGFP, while the chain located in the periplasm is inaccessible to the SpyCatcher fusion.

In a recent study, we investigated whether the hairpin model also holds true for TAAs using YadA as a model system, and aided by the SpyCatcher-SpyTag technology [13]. Earlier studies had identified a flexible region at the start of the coiled coil embedded within the lumen of the β-barrel, termed the ASSA region, due to the sequence consisting of alanines and serines. The ASSA region was hypothesized to be important for the formation of the hairpin [72]. To test this hypothesis, a truncated YadA construct was used, containing only the β-barrel and a small part of the passenger, termed YadAM (for YadA membrane anchor; see Figure 3A). Single proline substitutions were introduced to the ASSA region to probe its importance for translocation. Prolines are known secondary structure disruptors due to their rigid backbone, and we expected the substitutions to disrupt the flexible ASSA region sufficiently to stall the autotransporter process and trap the N-termini in the periplasm. To test this, a representative mutant, YadAMA354P, was chosen for further studies, and a SpyTag was inserted in the wild-type YadAM and YadAMA354P constructs at the N-terminus of the mature proteins, following the signal peptide required for secretion. For easy detection, we fused purified SpyCatcher fused to sfGFP (superfolder green fluorescent protein). *Escherichia coli* cells expressing the different constructs were incubated with SpyCatcher-sfGFP (that is too large to penetrate the OM), allowing the SpyCatcher domain to bind to the exported SpyTags presented on the bacterial surface, but not to any stalled intermediates where the SpyTags would still be located in the periplasm. Bacterial cells producing the wild-type construct displayed higher fluorescence compared with cells producing the A354P mutant, indicating that the mutation was interfering with secretion of the passenger domain [13]. The effect of the A354P mutation on the surface topology of YadA was further investigated by isolating OM fractions of the two constructs after SpyCatcher-sfGFP treatment and analyzing them by semi-native SDS-PAGE (i.e., without heating the samples beforehand) (Figure 4A). GFP is stable in SDS at ambient temperatures [73], allowing instant visualization of GFP-containing bands by in-gel fluorescence under blue light. The gel showed a ladder-like pattern of SpyTag-YadAM complexes bound to 1–3 SpyCatcher-GFP molecules, corresponding to the number of strands exported to the bacterial surface (Figures 3B and 4A). Comparison of the band intensities indicates that the majority of YadAM exports all three passengers, whereas YadAMA354P mainly exports 1–2 SpyTags. These experiments demonstrated that the proline substitution is a partially stalled intermediate where one or two passenger strands are exported, but not a fully blocked autotransport intermediate.

As an alternative to using reporter molecules such as GFP to study the surface topology of OM proteins, the binding of SpyCatcher alone can be utilized for assessing surface exposure. Because SpyCatcher forms a covalent bond with SpyTag, the added molecular weight of SpyCatcher causes a change in electrophoretic mobility, giving a shift of 16 kDa that can be observed either by Coomassie staining or Western blotting. As an example, we studied the surface exposure of the full-length, SpyTagged YadA construct. YadA was treated with SpyCatcher alone (i.e., with no fusion partner) and compared with YadA treated with an inactive SpyCatcher, where the catalytic glutamate was changed to glutamine (SpyCatcherEQ), by SDS-PAGE. In this example, unlabeled YadA remains trimeric even when heated to 95 ◦C and is seen as a single band at 200 kDa [74]. By contrast, the sample treated with active SpyCatcher forms a ladder-like pattern corresponding to YadA alone and YadA plus 1–3 bound SpyCatchers (Figure 4B). The experiments with truncated and full-length YadA demonstrate how SpyCatcher can readily bind to SpyTags located in close proximity to the bacterial surface, as shown for the YadAM constructs, as well as for the more exposed SpyTag protruding far from the surface located on the full-length protein. Future work will determine whether this technology is also suitable for detecting OM proteins without protruding domains, for example, where SpyTag is inserted into the loops of transmembrane β-barrel domains.

**Figure 4.** SpyCatcher assays with YadA. (**A**) Semi-native SDS-PAGE of OM preparations of wild-type SpyTag-YadAM and SpyTag-YadAMA354P after treatment with SpyCatcher-sfGFP, visualized under blue light. The ladder-like pattern indicates 1–3 bound SpyCatcher-sfGFP molecules. The gel shown is based on results from [13]. The wild-type YadAM binds mainly three SpyCatcher-sfGFP molecules, demonstrating full surface exposure of the N-terminal SpyTags, whereas the A354P mutant mainly binds to 1–2 SpyCatcher-sfGFP molecules. Note that YadAMA354P migrates anomalously compared with the wild-type. A schematic of the species visible in the gel is given on the side: SpyCatcher-sfGFP in green and YadAM monomers in red, blue, and yellow. vc represents vector control; (**B**) SDS-PAGE of OM preparations of full-length, trimeric SpyTag-YadA (~200 kDa) treated with the inactive SpyCatcherEQ and normal SpyCatcher, stained with Coomassie blue. The SpyCatcher-treated SpyTag-YadA shows a similar ladder-like pattern as in panel a, corresponding to the change in electrophoretic mobility as 1–3 SpyCatchers are bound.

#### *3.2. Using SpyCatcher to Investigate Membrane Dynamics*

The OM of Gram-negative bacteria consists of a mixture of OM proteins, lipopolysaccharides, and phospholipids. Both OM proteins and lipopolysaccharides are inserted into the OM in localized patches around the cell center where they diffuse to form a uniform distribution across the bacterial surface [75]. When the bacterium grows in preparation for cell division, the outer membrane proteins migrate towards the poles as a result of new membrane material and peptidoglycan being incorporated near the cell center in a process that is still not well understood [62].

In a recent paper, Keeble et al. used the SpyCatcher system to study the OM dynamics in *E. coli*. To this end, they used a novel phage display method to develop an improved SpyCatcher-SpyTag pair, termed SpyTag002 and SpyCatcher002 [21]. The rate constant of the improved 002 pair was increased by an order of magnitude compared to the original SpyCatcher–SpyTag interaction (2.0 <sup>×</sup> 104 <sup>M</sup>−<sup>1</sup> <sup>s</sup>−<sup>1</sup> vs. 1.7 <sup>×</sup> 103 M−<sup>1</sup> s−1). The efficiency of the newly developed pair was demonstrated by labeling the peptidoglycan-binding autotransporter intimin in order to study the membrane dynamics during cell growth and division. Intimin, a virulence factor of enterohemorrhagic *E. coli*, is a surface-exposed protein belonging to the type Ve secretion systems, also called inverse autotransporters [76]. Intimin consists of three major parts: an N-terminal periplasmic domain that binds to peptidoglycan under acidic conditions [77], a transmembrane β-barrel anchor [78], and the C-terminal passenger that is translocated to the bacterial surface through the β-barrel [79]. Keeble et al. prepared an intimin–SpyCatcher002 construct with SpyCatcher002 fused to the C-terminus of a truncated intimin variant (Figure 5). Upon overexpression in *E. coli* cells, the SpyCatcher002 domain is secreted together with the truncated passenger and presented on the bacterial surface, where it can be labeled using SpyTag002 fused to the fluorescent reporter protein mClover3 [80], in a similar fashion to that described for YadAM in Section 3.1. By imaging

the cells using wide-field fluorescence microscopy after labeling, the movement of the mClover-tagged intimins toward the poles could be tracked to study the dynamics of the OM. Intimin was initially uniformly spread across the bacterial surface, but migrated towards the poles during cell division [21], suggesting that the polar movement results from incorporation of new peptidoglycan and OM material in preparation for cell division. This notion was further tested by repeating the experiment in the presence of cefalexin, a β-lactam antibiotic that prevents bacterial division by inhibiting septum formation and peptidoglycan synthesis. This time, the cells became elongated in preparation for division but were unable to divide. After 45 min, the fluorescent signal was localized in patches around the bacterium, thus demonstrating how inhibited peptidoglycan synthesis prevents the polar migration of OM proteins. Keeble et al. demonstrated here how an intimin–SpyCatcher fusion can be used to study the effect of peptidoglycan synthesis during division upon outer membrane dynamics [21]. This method could be used for tracking the migration of other OM proteins in the membrane during different stages of the cell cycle.

**Figure 5.** Intimin–SpyCatcher002 fusion for studying membrane dynamics. Structural model of the intimin construct used in [21]. The model is based on the crystal structures of the intimin transmembrane domain (PDB ID: 4E1S) and SpyCatcher (PDB ID: 4MLI), the solution structure of the peptidoglycan-binding LysM domain (PDB ID: 2MPW), and the homology model structures of the extracellular D00 and D0 domains [81].

#### *3.3. Exploiting Virulence Factors and Virus-Like Particles for Vaccine Development using SpyCatcher-SpyTag*

Outer membrane vesicles (OMVs) are ubiquitously produced by Gram-negative bacteria and are often responsible for delivering virulence factors to the host cells during infection [82]. Derived from the outer membrane, OMVs contain many naturally occurring immunogenic components such as lipopolysaccharides, lipoproteins, outer membrane proteins, peptidoglycan, and other periplasmic components with intrinsic adjuvant properties [83]. As opposed to live and attenuated bacteria, OMVs are non-replicating and therefore pose no risk of infection after vaccination of immunocompromised individuals. Consequently, OMVs are promising candidates for modern vaccine development, both as adjuvants and as delivery vehicles for antigens. Vaccines utilizing OMVs as adjuvants have already been on the market for more than two decades; however, the development of heterologous antigen-presenting OMV-based vaccines using recombinant technology is still in its infancy [83,84]. Several strategies have been employed to present antigens on OMV surfaces, most notably as heterologous fusions with autotransporters [85–87]. The use of autotransporters for heterologous antigen presentation on OMVs has recently been reviewed in detail elsewhere [88], so we will only provide a short overview here.

The Luirink group has pioneered the use of recombinant OMVs for vaccine development. As a scaffold, they earlier utilized a classical (type Va) autotransporter, the *E. coli* hemoglobin protease (Hbp). Hbp consists of two major parts: a C-terminal transmembrane β-barrel that is inserted into the outer membrane, and an N-terminal passenger that is translocated through the β-barrel to the bacterial surface where it is released to the environment through autoproteolysis (Figure 6) [89,90]. The Luirink group developed an Hbp variant suitable for heterologous antigen export and display by introducing a point mutation that prevents autoproteolysis into a truncated Hbp passenger domain (HbpD(Δd1); from now referred to only as Hbp) [86]. By combining the new Hbp variant with a hypervesiculating *Salmonella* Typhimurium Δ*tolRA* strain, they developed a method to isolate large amounts of Hbp-presenting OMVs [87]. To utilize Hbp as an antigen-presenting tool, several loops in the passenger were identified that allowed the insertion of heterologous protein antigens without impeding the autotransport process. However, even though the autotransporter accepts substantial changes to its passenger, there are still considerable limitations to the size and folding properties of passenger fusions [91].

**Figure 6.** Hbp constructs used for vaccine display on outer membrane vesicles. (**A**) Structural model of full-length HbpD(Δd1) based on the crystal structures of the Hbp transmembrane domain (PDB ID: 3AEH) and passenger (PDB ID: 1WXR); (**B**) and (**C**) schematic drawings of Hbp fusions used in [92] to determine binding efficiencies between the SpyTag-SpyCatcher and SnoopTag-SnoopCatcher pairs on Hbp-fusion-presenting OMVs.

In a recent paper, this problem was circumvented by using the SpyCatcher-SpyTag technology to decorate the OMVs after expression of Hbp [92]. In the process, the investigators comprehensively tested the applicability of the SpyCatcher and SnoopCatcher systems for modular vaccine building, and we will only cover the highlights here. To fully test the range of the new technology, the authors prepared OMVs presenting Hbp N-terminal fusions with SpyTag, SpyCatcher, SnoopTag, and SnoopCatcher (Figure 6). By treating the different constructs with their respective binding partners, and comparing the changes in electrophoretic mobility upon ligation by SDS-PAGE, they showed that all four Hbp-variants exported their passenger to the surface and were amenable to ligation. As a proof-of-concept for vaccine development, they proceeded to treat Hbp-SpyTag-presenting OMVs with a relatively large (24 kDa) domain of the antigenic *S. pneumoniae* surface protein, PspAα, fused to both SpyCatcher and SnoopTag (SpyCatcher–PspAα–SnoopTag) (Figure 7A). Earlier attempts to fuse PspAα with the Hbp passenger

required the antigen to be divided into two smaller domains, which still resulted in significantly reduced export and lower antigen presentation on the OMV surface [93]. The SpyCatcher–PspAα—SnoopTag, on the other hand, was ligated to the Hbp-presenting OMVs with similar efficiency as the earlier controls, demonstrating that the SpyCatcher system is a robust tool suitable for coating OMVs with recombinant antigens, completely circumventing the passenger export problem.

**Figure 7.** Hbp-antigen complexes made through Spy/SnoopCatcher fusions. (**A**) Schematic of Hbp bound to SpyCatcher–PspAα–SnoopTag; (**B**) schematic of Hbp bipartite binding adduct SpyCatcher–PspAα–SnoopTag and SnoopCatcher–SP1690–SpyTag; (**C**) schematic of Hbp tripartite binding adduct with two SpyCatcher–PspAα–SnoopTags flanking SnoopCatcher–SP1690–SpyTag.

The next step was to check whether multiple antigens could be coupled to the OMV surface in an iterative fashion by using the alternating combinations of the SpyCatcher and SnoopCatcher systems flanking the antigen proteins. Combinations of antigens to form multivalent OMV-based vaccines can significantly improve their efficiency, either by eliciting a stronger immune response to a particular pathogen, or by providing a broader response by combining antigens from multiple pathogens [94,95]. For this approach, they treated the OMVs with PspAα fusion and a SnoopCatcher–SP1690–SpyTag

fusion with wash steps in between, successfully forming bipartite (Figure 7B) and tripartite (Figure 7C) coupling adducts. This work demonstrated how various recombinant antigens can be coupled to the OMV surface using the SpyCatcher system.

A similar application for the SpyCatcher system in vaccine development has been used for VLPs. VLPs are virus-derived proteins that self-organize into noninfectious virus-like structures that are capable of eliciting strong immune reactions [96]. Much effort has been put into designing VLP–antigen fusions; however, this approach has proven to be labor-intensive and time-consuming, since the fusions often misalign, or are unable to form stable capsids [97]. Brune et al. from the Howarth group circumvented this problem by using the SpyCatcher technology in an analogous fashion to the OMV-based vaccines [44]. They were able to create stable SpyCatcher-presenting VLPs by using a fusion between the bacteriophage AP205-derived capsid and SpyCatcher. They further demonstrated how the VLPs could be decorated using SpyTag-fused malarial antigens (Figure 2F) in a plug-and-play fashion, and how this product elicited an antibody response after a single injection in mice. In similar work, Thrane et al. fused either SpyTag or SpyCatcher to the C- and/or N-termini of the AP205 capsid protein, and subsequently treated the resulting VLPs with 11 different antigen–SpyCatcher/SpyTag fusions [43]. In both works, the authors confirmed the immunogenic properties of the novel vaccines by immunizing mice, thus confirming the functional activity of the novel vaccines. The results from the OMV and VLP experiments demonstrated the versatility of the SpyCatcher system in a novel approach for designing vaccines. By presenting one of the binding SpyCatcher/SpyTag partners on the particle surface, linking antigens in a plug-and-play fashion circumvents the need for designing complicated fusion constructs, and could soon pave the road for modular vaccines that can easily be tailored for any need. For further information, Brune and Howarth recently published a more comprehensive review on the design of VLP-based vaccines, including the use of the SpyCatcher platform for this purpose [14].

#### **4. Conclusions and Future Perspectives**

The SpyCatcher system and its close relatives have been successfully used for protein engineering purposes, including the cyclic polymerization of enzymes, or the decoration of hydrogels. In our hands, the system has been extremely useful for showing cell surface localization, and to elucidate the topology of membrane proteins. Other authors have used the technology for decorating vesicles and virus particles with antigens. In all cases, the benefits of the system are the same: irreversible, covalent coupling and the fact that the two partners of the system are relatively small entities that do not interfere strongly with the native systems that they label.

It is somewhat ironic that the system, derived from a bacterial surface protein, has been widely used for studying and modifying other bacterial or viral surface proteins. We strongly believe that this system will have many more applications in basic science, and that this is only limited by the fact that many researchers are not aware of its existence. This prompted us to write this review.

Some of the approaches related to vaccines will need more testing before they can potentially be used in patients. For example, it remains unclear whether SpyCatcher (or its relatives) elicits a strong immune response by itself, and whether that would limit its usefulness in vaccines. The work on VLPs suggests that this could be problematic for some applications, where the SpyCatcher-VLPs did elicit an immune response of their own that was then later masked after decorating the particles with the target antigen [44].

Our own experiments were until now limited to surface-localized proteins in bacteria, but it is conceivable that the system could also be used to label proteins in other cellular localization (e.g., after membrane permeabilization) as an alternative to antibody-based approaches. We also believe that the system can be used in versatile ways to modify and functionalize artificial or biological surfaces of almost any kind (e.g., to develop ELISA-like assays that require covalent immobilization), and possibly to functionalize surfaces used in lab-on-a-chip approaches. Another conceivable use of the system includes in situ labeling of proteins for added density to identify individual proteins

(e.g., in cryo-electron micrographs or tomograms). As this review shows, the SpyCatcher-SpyTag system has proven to be an extremely versatile tool for both basic research and applied science, and it will undoubtedly spur further innovations in the future.

**Author Contributions:** Conceptualization, J.C.L. and D.L.; writing—original draft preparation, D.H., T.T., D.L., and J.C.L.; writing—review and editing, D.L. and J.C.L.

**Funding:** This research was funded by the Research Council of Norway, grant numbers 249793 (to J.C.L.) and 240483 (to D.L.).

**Acknowledgments:** The authors wish to thank Nandini Chauhan for providing the gel image shown in Figure 4B. They also acknowledge the members of the Leo and Linke groups for fruitful scientific discussions.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **Abbreviations**


#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

MDPI St. Alban-Anlage 66 4052 Basel Switzerland Tel. +41 61 683 77 34 Fax +41 61 302 89 18 www.mdpi.com

*International Journal of Molecular Sciences* Editorial Office E-mail: ijms@mdpi.com www.mdpi.com/journal/ijms

MDPI St. Alban-Anlage 66 4052 Basel Switzerland

Tel: +41 61 683 77 34 Fax: +41 61 302 89 18