**Lon Protease Is Important for Growth under Stressful Conditions and Pathogenicity of the Phytopathogen, Bacterium** *Dickeya solani*

**Donata Figaj 1,\*, Paulina Czaplewska 2, Tomasz Przepióra 1, Patrycja Ambroziak 1, Marta Potrykus <sup>3</sup> and Joanna Skorko-Glonek 1,\***


Received: 10 April 2020; Accepted: 20 May 2020; Published: 23 May 2020

**Abstract:** The Lon protein is a protease implicated in the virulence of many pathogenic bacteria, including some plant pathogens. However, little is known about the role of Lon in bacteria from genus *Dickeya*. This group of bacteria includes important potato pathogens, with the most aggressive species, *D. solani*. To determine the importance of Lon for pathogenicity and response to stress conditions of bacteria, we constructed a *D. solani* Δ*lon* strain. The mutant bacteria showed increased sensitivity to certain stress conditions, in particular osmotic and high-temperature stresses. Furthermore, qPCR analysis showed an increased expression of the *lon* gene in *D. solani* under these conditions. The deletion of the *lon* gene resulted in decreased motility, lower activity of secreted pectinolytic enzymes and finally delayed onset of blackleg symptoms in the potato plants. In the Δ*lon* cells, the altered levels of several proteins, including virulence factors and proteins associated with virulence, were detected by means of Sequential Window Acquisition of All Theoretical Mass Spectra (SWATH-MS) analysis. These included components of the type III secretion system and proteins involved in bacterial motility. Our results indicate that Lon protease is important for *D. solani* to withstand stressful conditions and effectively invade the potato plant.

**Keywords:** protease Lon; *Dickeya solani*; plant pathogen; virulence factors; motility; type III secretion system; pathogenicity; resistance to stress; *lon* expression; pectinolytic enzymes

#### **1. Introduction**

Bacteria from the genus Dickeya, together with Pectobacterium, are classified as soft rot Pectobacteriaceae (SRP) [1]. They cause diseases of many economically important plants leading to significant financial losses all over the world [2,3]. *Dickeya solani* was first identified in 2005 and since then it has spread in Europe reducing yields of its primary host, potato [4,5]. *Dickeya solani* turned out to be a well-adapted and very successful pathogen, and due to its better adaptation in many regions, it has displaced another common potato pathogen*, Dickeya dianthicola*. Briefly, *D. solani* can infect the host plant with as little as 10 cells per tuber and has a broader temperature spectrum for infection compared to other SRP species [2]. SRP causes two types of plant diseases: blackleg and soft rot characterized by the blackening and wilting of a plant stem or tuber rot, respectively [5]. No effective methods to combat these pathogens have been developed so far [5]. The virulence factors that are primarily involved in the development of the disease symptoms caused by *D. solani* are plant cell wall degrading enzymes (PCWDE). This group of proteins encompasses pectinases, cellulases and proteases, whose joint action leads to maceration of plant tissues. Pectinases and cellulases are secreted via the type II secretion system (T2SS) and proteases are secreted via the type I secretion system (T1SS). However, an effective infection process requires also an array of other virulence determinants. These include the ability to actively move (motility) toward wounded tissue (chemotaxis), production of antioxidants, siderophores, secreted effectors and, finally, cellular factors that provide survival under unfavorable environmental conditions inside the host plant. *Dickeya* enters the plant via wounds or natural openings, and then it can penetrate the apoplast. There, the type III secretion system (T3SS), an essential virulence factor, is activated (reviewed in [6]). It consists of a pillus apparatus, so-called injectisome, connecting bacterial and plant cell cytoplasm [7] to directly inject effector proteins into the host. The effectors frequently manipulate host signaling pathways to disable the plant's defense systems to enable successful infection (reviewed in [8]).

For proper functioning of the cell, it is necessary to maintain its homeostasis, which is especially problematic under stressful conditions. Pathogenic bacteria are frequently exposed to unfavorable conditions, both in the host and during transmission between the hosts. In their life cycle, *Dickeya* may encounter several types of stress. These include changes in temperature and pH, exposure to oxidants, as well as osmotically active compounds [6]. All the stressors mentioned above can cause protein damage. To maintain cellular proteostasis, a dedicated set of proteins, termed protein quality control system (PQCS), is employed. It includes proteases and chaperones, that, among others, protect the cell against the accumulation of harmful protein aggregates and participate in regulatory proteolysis [9,10]. Lon and ClpP are two major cytoplasmic proteases, responsible for 70–80% of the ATP-dependent proteolysis in the cytosol [11]. Lon degrades aberrant proteins, which can arise in the cell not only as a consequence of stress but also under physiological conditions. Additionally, native proteins with the degradation tags (in general nonpolar amino acids exposed to the solvent) can undergo proteolysis by the Lon protease. Lon regulates many important cellular processes through the degradation of relevant substrates. For example, in *Escherichia coli,* cell division and synthesis of capsular polysaccharide are regulated by proteolysis of the cell division inhibitor SulA and transcription regulator RcsA, respectively [12,13]. Moreover, in various bacteria, functions of Lon can be implicated in such processes as motility, DNA replication, sporulation and, finally, pathogenicity [14].

Lon is crucial for virulence of many animal and plant pathogenic bacteria [8,14]. The *lon* mutant of *Brucella abortus* displayed decreased pathogenicity in BALB/c mice, however only at an early stage of infection [15]. The more pronounced effect was observed in the case of *Salmonella enterica* serovar Typhimurium *lon* mutants, which were highly attenuated in mice [16]. *Pseudomonas aeruginosa* Lon protease is necessary for effective bacterial infection in the mouse acute lung and amoeba model [17]. *Agrobacterium tumefaciens* depleted of a functional *lon* gene was unable to induce tumors in *Kalanchoe diagremontiana* [18]. The *lon* mutants of *Pseudomonas syringae* pv. *phaseolicola* and *P. syringae* pv. *tomato*, in turn, showed reduced disease symptoms in bean and tomato models, respectively [19].

To determine the role of the Lon protein in *D. solani,* we constructed a *D. solani* Δ*lon* strain using a modified lambda red recombination protocol. This allowed us to provide an insight into the functions played by Lon in *D. solani* pathogenicity and growth under stressful conditions.

#### **2. Results**

#### *2.1. Construction of the D. solani IPO 2222* Δ*lon and the Complemented D. solani IPO 2222* Δ*lon*/*lon Strains*

A *lon* deletion mutant of *D. solani* was constructed using the gene doctoring method, a modified protocol for lambda red recombination dedicated to the pathogenic bacterial strains. The gene encoding the Lon protease was substituted with the kanamycin resistance cassette, amplified from the pDOC-K plasmid. We confirmed the deletion of *lon* by three independent methods: (1) detection of the *lon* sequence on the *D. solani* chromosome by PCR using the primers complementary to the sequences flanking the *lon* gene; (2) detection of the *lon* transcript by real-time PCR; (3) immunodetection of the Lon protein using the anti-*E. coli* Lon rabbit antibodies. As can be seen in Figure 1, the *lon* gene and its product were not found in the strain *D. solani* IPO 2222 Δ*lon* but were present in the WT parental strain.

**Figure 1.** Confirmation of successful deletion of the *lon* gene in *D. solani* IPO 2222: (**A**) PCR analysis of genomic DNA isolated from the *D. solani* WT (wild type) and Δ*lon* mutant. The *lon* gene was replaced by a 1000 bp smaller kanamycin resistance gene. Δlon1 and Δlon2 denote two independent clones (**B**) immunodetection using the anti-Lon *E. coli* primary antibodies. (**C**) The qPCR analysis with the use of the *lon* gene-specific primers revealed that no cDNA amplification product was created within 50 cycles.

Single-copy complementation in the genome of the *D. solani* Δ*lon* strain was obtained using the *E. coli* MFD (Mu-free donor) *pir* conjugation strain. Immediately after the end of the *lon* gene, a marker gene encoding a pink Scarlett fluorescent protein was inserted. We confirmed the Δ*lon*/*lon* complementation at the gene and protein level (Figure S1).

A lack of the *lon* gene did not affect the growth of bacteria at the standard in vitro conditions (LB medium, 30 ◦C), as judged from the growth curves. The growth rates of the Δ*lon*, Δ*lon*/*lon* and WT *D. solani* cultures were comparable (Figure 2). Similar growth patterns of the Δ*lon* and WT strains have significantly facilitated the normalization of bacterial cultures in terms of age and cell density for subsequent stress sensitivity and pathogenicity tests.

**Figure 2.** Growth of *D. solani* Δ*lon.* Curves were determined with the use of a plate reader at 30 ◦C. OD595 (optical density (595 nm)) values in LB medium were averaged for four replicates.

#### *2.2. The Expression of the lon Gene is Upregulated under Certain Stressful Conditions*

Under stress, a cell activates a variety of defense mechanisms that are manifested by increased expression of the key protective proteins [20–22]. To check the importance of Lon in the stress response, we performed the qPCR analysis to measure levels of transcription of the *lon* gene. We chose the common stressors, possible to affect *D. solani* during saprophytic and pathogenic life cycle: elevated temperature, nonionic and ionic osmotica, acidic pH and oxidants [23].

We found that transcription of *lon* was significantly upregulated in the exponentially growing bacteria under stressful conditions such as hyper osmosis, acidic pH and high temperature (Figure 3). The most pronounced effect was exerted by acidic pH and elevated temperature (over three log2 fold increase). A milder effect was caused by the presence of a nonionic osmoticum, sucrose (over two log2 fold increase). In contrast, the expression of the *lon* gene did not change significantly in cells in the stationary phase of growth. The exception was the upregulation (over 5.5 log2 fold increase) of the *lon* gene expression in cells treated with acidic pH. Interestingly, the changes in the *lon* transcript level upon treatment with ionic osmoticum, NaCl, were pronounced regardless of the growth phase, although they were not statistically significant (Figure 3). In contrast, oxidative stress did not affect the transcription of *lon* (Figure 3). Hence, the Lon protease is rather not a component of the oxidation response in *D. solani.*

**Figure 3.** The relative log2 fold change of the expression levels of the *lon* gene in the *D. solani* cells under stressful conditions analyzed by qPCR. The data correspond to the means ± S.D. of three different samples, including three technical replicates. A red horizontal line indicates a relative two-fold increase in expression level. \* indicates statistically significant (95% confidence interval) fold change in expression level according to the REST 2009 software.

#### *2.3. Lon Protease Plays a Protective Role under Ionic and High-Temperature Stresses*

Knowing that expression of *lon* is upregulated in cells in response to stress, we decided to check if the presence of the Lon protease in the cell is necessary for bacterial growth in the presence of selected stressors. Bacteria were exposed to the following stressful conditions: elevated temperature, ionic and nonionic osmotic shock, oxidative stress and low pH. We found that *D. solani* Δ*lon* was characterized by a decreased ability to form single colonies under three of five tested conditions. In particular, elevated temperature and presence of the nonionic osmoticum, sucrose, reduced viable cell counts by five and three orders of magnitude, respectively. Moreover, the Δ*lon* mutant colonies were very small under both tested conditions. The reintroduction of the *lon* gene into the *D. solani* Δ*lon* chromosome restored the wild-type phenotype of bacteria (Figure 4A). Hence, the strong phenotype of the mutant strain resulted from the lack of the Lon protease and not from putative additional suppressor mutations. The pronounced effect was also noticed under ionic osmotic stress: addition of NaCl resulted in a three-log reduction of cell counts of the Δ*lon* mutant with respect to the WT or complemented Δ*lon*/*lon* strain (Figure 4A). Acidic pH and oxidative stress affected all strains similarly (Figure 4A,B).

**Figure 4.** Growth of *D. solani* Δ*lon* under stressful conditions. (**A**) Overnight grown cultures were serially diluted and spotted on the LA (Luria Agar) agar plates, agar plates supplemented with 0.6 M sucrose or 0.3 M NaCl or on the LA medium adjusted to pH 5.0 when indicated. Bacteria grown on the LA agar plates at 30 ◦C refer to control. Disk diffusion assay with 1% hydrogen peroxide. As a negative control, sterile water was used (**B**). All plates were incubated at 30 ◦C except for the elevated temperature stress (39 ◦C).

#### *2.4. Deletion of the lon Gene Delays the Onset of the Infection Symptoms*

To test the importance of the Lon protease for pathogenicity of *D. solani*, we performed in vivo infection of the potato plants under greenhouse conditions. This kind of experiment shows the ability of bacteria to invade plants through the root system and produce blackleg symptoms. Although the deletion of the *lon* gene did not significantly reduce the occurrence of disease, an obvious delay in the development of the disease symptoms was observed. On the seventh day postinfection, only 30% of the potato plants treated with the Δ*lon* mutant bacteria showed blackleg symptoms, compared to 75% of symptomatic plants infected with WT *D. solani*. On the 17th day, the differences were much less pronounced, with 55% and 75% of symptomatic plants infected with Δ*lon* and WT *D. solani*, respectively (Figure 5).

To evaluate the effects of the Δ*lon* mutation on the ability of bacteria to macerate plant tissues, we used three models: potato tubers and leaves of chicory and Chinese cabbage. In no case did we observe differences in the degree of tissue maceration (Figure S2).

**Figure 5.** Pathogenicity of *D. solani* Δ*lon* in the whole potato plant model. Potato plants cv. Vineta were infected with WT and *lon* mutant strains. After 7 and 17 days of incubation at room temperature with a 16/8 photoperiod, the number of plants with blackleg symptoms was counted. Four plants watered with Ringer buffer represented a negative control. The number of infected plants with WT and Δ*lon* mutant was 16 for each strain.

#### *2.5. The Deletion of lon A*ff*ects the Activity of Secreted Pectate Lyases*

PCWDEs are the virulence factors that are directly responsible for the manifestation of disease symptoms. To check if altered pathogenicity of *D. solani* Δ*lon* results from changes in the level or activity of the enzymes that degrade the plant cell wall, we measured the activity of pectate lyases (major pectic enzymes), cellulases and proteases secreted from the mutant and WT *D. solani* cells. The enzymatic activity was assayed using PGA, modified cellulose CMC and casein, which are commonly used substrates for pectinases, cellulases and proteases, respectively. In the case of *D. solani* Δ*lon,* the secreted pectate lyase activity was 85% lower than that of the WT strain (Figure 6A). However, the activities of the remaining hydrolytic enzymes were not affected by the *lon* mutation (Figure 6B,C). The level of other secreted virulence factors, siderophores, also remained unchanged (Figure 6D).

**Figure 6.** The activity of secreted virulence factors. (**A**) Pectinase activity was assayed as described in the Methods section with PGA as a substrate at 30 ◦C. \*\* *p* < 0.01 (*t*-test), *n* = 5. (**B**) Cellulases were assayed on M63Y with CMC. Seven microliters of bacterial cultures (108 CFU/mL) were spotted on the medium and incubated for 72 h. Plates were stained with 2% Congo red solution. (**C**) To monitor protease activity, 7 μL of bacterial cultures (10<sup>8</sup> CFU/mL) were spotted onto LA with skimmed milk and incubated for 48 h. (**D**) Siderophore activity was determined by spotting 10 μL of supernatant from overnight grown bacteria cultures onto chrome azurol S-agar plates. The picture was taken after 1 h of incubation at 30 ◦C. The experiments (**B**–**D**) were performed at least five times. The representative results are shown.

#### *2.6. Lon Protease is Essential for E*ffi*cient Motility*

Motility is one of the key factors for a successful invasion of the plant host. To determine if the observed delay of the blackleg symptoms development in the potato plants infected by *D. solani* Δ*lon* can be associated with altered bacterial motility, we examined two types of motility, swimming and swarming. While both types rely on the rotation of flagella, swimming is characteristic for an individual cell and is enhanced by chemotaxis. In contrast, swarming is common for a group of bacteria [24]. Indeed, the lack of Lon resulted in the altered motile phenotype of bacteria. The mutated strain showed considerable reduction in the swarm (Figure 7A) and 30% reduced swimming motility in the presence of galactose as a chemotactic agent (Figure 7B).

**Figure 7.** Swarming and swimming motility of *D. solani* Δ*lon*. (**A**) Representative pictures of bacteria swarming on 0.8% TSA after 12 h incubation at 30 ◦C. The experiments were performed at least five times. (**B**) Swimming was examined in a 0.3% MMA medium for 24 h at 30 ◦C. The diameter of the bacteria spreading area was measured. Presented data represent values for 5 biological replicates. \*\*\* *p* < 0.001 level (*t*-test).

#### *2.7. Comparison of Proteomic Profiles of the D. solani* Δ*lon and WT Cells under Physiological and Stressful Conditions*

To gain more detailed insight into the properties of the Δ*lon* mutant cells, we compared the proteomes of the mutant and WT strains under physiological, as well as stress conditions, by the means of SWATH-MS (Sequential Window Acquisition of All Theoretical Mass Spectra) analysis. SWATH-MS is an advanced analysis method of proteomic data, recommended for quantification of identified peptides. It allows quantitative comparison of protein levels between different species or treatments due to the construction of a peptide spectral library [25].

It is well known that treatment with severe stressful agents can cause abnormal changes in the level of individual macromolecules [20] so we decided to subject the *lon* mutant to a rather mild stress—a short incubation at 40 ◦C.

The analysis identified a total of 635 proteins, for which at least two peptides per protein were quantified (Table S1). Deletion of the *lon* gene induced global changes in the *D. solani* proteome. We narrowed the number of differentially expressed proteins by applying the following cut-off criteria: *p* < 0.05, as well as fold changes below 0.5 or above 2.0. This resulted in 38 proteins with altered abundance in Δ*lon* compared to WT under physiological conditions and 60 proteins under

stress conditions. Hence, the changes in the mutant proteome were more pronounced following the temperature shift than under control conditions, which may reflect the increased need for the Lon function during stress. In particular, the deletion of the *lon* gene resulted in upregulation of 17 or 41 proteins and downregulation of 13 or 19 proteins under physiological or stress conditions, respectively. Of these, 28 proteins shared a similar pattern of expression under both tested conditions (Figure 8A).

We grouped differentially expressed proteins into eight categories, depending on their physiological functions (Table 1). These include involvement in motility, iron metabolism, stress response, transport, general metabolism, transcription/translation, virulence and others. Percentage of proteins representing particular groups differs between control and induced conditions, however, the most abundant class encompasses proteins associated with general cell metabolism (Figure 8B). Consistent data were obtained for proteins involved in bacterial motility, namely all of them were repressed in Δ*lon* compared to WT. Among them, we identified flagellin, a structural component of bacterial flagella, and proteins responsible for chemotaxis (CheW, a positive regulator of CheA protein activity and CheA, signal transduction histidine kinase CheA). On the contrary, deletion of *lon* caused an increase in the cellular content of a group of proteins associated with virulence. They are all engaged in the T3SS and include hairpins, HrpN, a homolog of HrpW (Various polyols ABC transporter, permease component 2), as well as HrpA, Hrp pili protein. Deletion of *lon* differentially affected levels of proteins related to iron metabolism. We observed the upregulation of the proteins involved in the Fe-S cluster assembly (ISCU) and biosynthesis of achromobactin siderophore. However, the level of proteins involved in the synthesis of another siderophore, enterobactin, was decreased (for example isochorismate synthase enterobactin siderophore). The mutant strain was characterized by an increased content of several proteins engaged in transcription and translation. Among them, we could distinguish ribosomal proteins (50S ribosomal protein L27 and L7/L12), transcription factors (CytR, DksA) and RNA-binding protein Hfq. We obtained a similar trend regarding stress-related proteins, like ClpP protease and cold shock response proteins (CspE and CspG). A total of 77% of proteins associated with transport activity were downregulated, including histidine ABC transporter and efflux pump membrane transporter. The group named "others" comprises uncharacterized proteins or polypeptides which were not assigned to any other category. Among them, we identified putative membrane protein A0A2K8W3L1\_9GAMM whose expression was increased more than 100-fold under both tested conditions. Protein blast indicated very close homology to periplasmic ComEA from many bacterial species, with the closest homology (100% coverage, 99% identity) to *Dickeya fangzhongdai*. ComEA is essential for DNA uptake in naturally competent bacteria, like *Bacillus subtilis* [26]. In Δ*lon*, we also observed an increased cellular level of S-ribosylhomocysteine lyase, which is associated with quorum sensing.

**Figure 8.** Number (**A**) and percentage (**B**) of proteins whose expression level was changed in the Δ*lon* mutant under physiological and heat shock conditions. Data are categorized into 8 groups depending on the protein function. Physiological and heat shock conditions refer to temperatures 30 ◦C and 40 ◦C, respectively.





*Int. J. Mol. Sci.* **2020**, *21*, 3687






**Table1.***Cont.*

#### **3. Discussion**

For successful infection, a pathogen must have the capability to enter the host, overcome the host defense systems, acquire nutrients, multiply and disseminate. All these stages are associated with constant exposure to a variety of potentially harmful conditions, both in and outside the host. Hence, successful pathogens should have well-developed virulence mechanisms but also efficient stress response systems. Proteolytic enzymes were shown to play numerous crucial roles in bacterial virulence. They can directly act as virulence factors, but also can contribute to virulence by regulating the production of virulence factors and/or as components of protein quality control systems to provide cellular proteostasis. One of the latter cases is the Lon protease which is indispensable for stress tolerance and virulence of many bacterial species causing infectious diseases.

Our work showed that the Lon protease is necessary for the bacterium *D. solani* to resist exposure to stress, including ionic- and nonionic osmotic stress, as well as high temperature. This finding is in agreement with data obtained for other bacterial species. Lon has a well-documented role in bacterial viability under heat and salt stress [14]. In *E. coli*, expression of the *lon* gene depends on sigma32 (RpoH) transcription factor [27,28], activated under heat shock and osmotic stress [29,30]. High temperature stimulates expression of *lon* in *E. coli* [31] and *Francisella tularensis* LSV [32]. Ionic osmotic stress is responsible for the elevated level of *lon* expression in *B. subtilis* and *Dickeya dadantii* [33]. Consistently, in *D. solani*, *lon* expression was significantly elevated in exponentially growing bacteria following exposure to elevated temperature; a positive trend was also observed in case of salt stress. Heat shock is more harmful to bacteria in the logarithmic than the stationary growth phase [34], which may explain stronger stimulation of the *lon* gene in the logarithmically growing cells. Higher demand for Lon, suggested by the upregulation of *lon* under certain stress conditions, can explain reduced growth of the Δ*lon* bacteria under thermal and osmotic stress. Additionally, the elevated level of stress-related proteins in proteomes of bacteria treated with 40 ◦C, revealed by the SWATH-MS analysis, indicates the higher stress level in the mutant cells than the WT. Increased expression of RecA and protease HtpX suggests a higher frequency of DNA damage and probably impaired integration of membrane proteins, respectively, according to data published for *E. coli* [35,36]. Finally, the increase in the abundance of the second important cytosolic protease, ClpP, in the deletion strain, reveals the essential role of Lon protease in the quality control proteolysis in the cytoplasm. Most probably, ClpP takes over some of the Lon functions. However, it should be noted that the Δ*lon* strain showed a temperature-sensitive (TS) phenotype, so ClpP cannot substitute for Lon under heat shock conditions.

To our surprise, the Δ*lon* mutants were particularly vulnerable to treatment with sucrose. Nonionic osmotic agents, like sucrose, are considered less harmful for a cell than the ionic ones, (e.g., NaCl) [37]. We did not find data regarding the relationship between Lon and the response to osmotic stress caused by high sucrose in any bacterial species. Hence, this important function of Lon in resistance to nonionic osmotic stress needs to be elucidated. Although the expression of *lon* was strongly upregulated in the response of *D. solani* to low pH, we did not observe differences of growth between the mutant and WT strains. Possibly, the ClpP protease or other component of the protein quality control system takes over the duties of Lon under this type of stress. The involvement of the Lon protease in resistance to acidic stress is rather poorly investigated across different bacterial species. In *E. coli,* Lon is responsible for the degradation of the activator of acidic resistance, GadE, playing a role in the termination of the stress response [38]. Additionally, *S. enterica* serovar Typhimurium requires this protease to successfully cope with low pH [16]; however, the precise mechanism was not provided. Interestingly, the closely related species, *D. dadantii*, showed repression of *lon* expression in the low pH medium (although not statistically significant) [33], which is opposite to our findings. This may reflect interspecies differences but also certain differences in experimental design. As with *D. dadantii* [33], in *D. solani* the expression of *lon* was not affected by oxidative stress-induced with hydrogen peroxide. Hence, in Dickeya, the oxidation response most probably involves other components of PQCS.

Production of functional virulence factors is frequently dependent on specific proteolytic activity in the cell [39,40]. To verify the involvement of Lon in *D. solani* virulence, we checked the activity of

the most abundant secreted virulence factors. We found that activity of the extracellular pectinases was reduced in the case of Δ*lon* mutant. Pectinases constitute a heterogeneous group of proteins. They differ in substrate specificity, abundance and role in virulence but all are secreted via T2SS [41]. At least 10 pectinases produced by *D. solani* have been identified so far [41]. The commonly used tests (including the one used in this work) measure a total pectinase activity and do not allow to distinguish between the individual pectinases. Analysis of the Δ*lon* and WT *D. solani* proteomes did not reveal differences in the cellular content of pectinases. However, we do not know if these proteins were efficiently transported outside the cell. As this is the first report of the function of Lon in the soft rot bacteria, there is no information regarding the relationship between Lon and PCWDE. No literature data is indicating the possibility of regulating the T2SS transport system by Lon. Moreover, the activity of extracellular cellulases, also T2SS dependent, remained unchanged in the *lon* mutant. Hence, the involvement of Lon in the regulation of T2SS is unlikely. Thus, further research is needed to clarify the Lon-dependent protease regulation of the secreted pectinase activity.

We did not observe any changes in the production of siderophores, although the lack of Lon protease lowered the abundance of several proteins engaged in iron metabolism. A higher level of proteins with function in iron-sulfur (Fe-S) protein biogenesis (IscU, ErpA) in the deletion strain may indicate them as potential substrates for the Lon protease. That is true in *Saccharomyces cerevisiae,* where a Lon homolog, Pim1, degrades Isu, a homolog of IscU [42]. The increase in the amount of the negative transcriptional regulator Fur is very interesting. Whether it is degraded by Lon is not known and no such Lon function was found in other bacteria. Almost a 10-fold increase of the Fur level in Δ*lon* may explain the decreased amount of certain Fur-dependent proteins involved in the synthesis of siderophores (e.g., enterobactin synthetase component F in *E. coli* [43]. On the other hand, the other enzymes from the siderophore biogenesis pathway were upregulated (like achromobactin biosynthesis protein AcsD), presumably compensating for the downregulated components to maintain iron homeostasis in the mutant cells.

A lack of Lon exerted a significant impact on the mobility of *D. solani.* We demonstrated that the cells deprived of Lon showed impaired swarming and swimming motility. This can be explained by the reduced levels of flagellin and positive regulators of chemotaxis in *D. solani* Δ*lon*, as revealed by the proteomic analysis. Depending on the bacterial species, the effects of the *lon* mutations on bacterial motility may be radically different. On one hand, the *lon* mutation can cause stimulation of motility. Good examples are *Proteus mirabilis* and *B. subtili*s, in which *lon* mutants showed better swarming [44,45]. In these bacteria, Lon degrades master activators of flagellin biogenesis- FlhD and SwrA, respectively. Hence, in the *lon* backgrounds, these activators became stabilized, leading to a hypermotile phenotype. In contrast, the *lon* mutant of *Erwinia amylovora* was characterized by a nonswarming phenotype [46]. In *E. amylovora,* a mutation in the *lon* gene resulted in the accumulation of RcsA/RcsB that negatively regulates transcription of *flhD*, the master regulator of flagellar biosynthesis. Finally, a lack of Lon may not affect bacterial motility at all, as shown for *S. entrica* serovar Typhimuirum [47]. The results obtained in this work suggest an indirect role of Lon in the motility of *D. solani*, analogous to that of *E. amylovora* Lon, as the *flhD* gene is present in *D. solani* genome and the *lon* mutant was characterized by a decreased flagellin content. Interestingly, *D. solani* Δ*lon* was characterized by the two-fold increased level of the CytR transcription factor, which in *Pectobacterium carotovorum* positively stimulates genes associated with motility: *fldH, fliA, fliC* and *motA* [48]. However, the increased content of the CytR protein in *D. solani* Δ*lon* obviously was not sufficient to compensate for the other Δ*lon* -dependent effects that lead to a reduced flagellin and chemotaxis protein levels, or CytR is not involved in regulation of motility in *D. solani*.

The phenotypes of *D. solani* Δ*lon* reported in this work suggested that the presence of Lon may be necessary for efficient infection of the potato plant. Indeed, we found that the process of development of blackleg symptoms in the plants infected with the mutant strain was markedly delayed in respect to infection with the WT *D. solani.* On the other hand, Lon was not essential for the maceration of plant tissues in vitro. Both types of infection tests differ fundamentally in terms of the availability of plant tissues for bacteria. In the whole plant model, the bacteria were placed in the soil, from where they must have got into the wounded tissue, in this case, roots. In this context, the motility and chemotaxis toward chemical signals (e.g., jasmonic acid) are crucial. Consequently, the nonmotile mutant strains are characterized by a lack or reduced virulence, as they may encounter severe problems with entering and/or spreading in the host [49]. In the slice or leaf model, bacteria were spotted directly into the wounded tissue, so chemotaxis and motility were less important. In the tissue model, the basis of infection's success lies in the production of PCWDE, iron homeostasis, and bacterial fitness under pH, oxidative and osmotic stresses (reviewed in [6]). As we did not observe any difference in the degree of maceration of the tuber or leaf tissues between the WT and mutant strains, we assumed that Lon was not essential for bacterial survival in the plant under experimental conditions. The secreted pectinase activity of the Δ*lon* mutant was reduced but was still enough for efficient plant maceration. Hence, we concluded that the observed delay in the potato plant infection process was most probably due to reduced motility of the Δ*lon* strain.

Lon is known to be engaged in the regulation of T3SS, however, the particular effects of *lon* mutations are species-dependent. In *P. aeruginosa* and *Yersinia*, the deletion of the *lon* gene results in the downregulation of T3SS [17,50]. However, the opposite effect was demonstrated in *E. amylovora* and *P. syringae*[46,51]. We found that in *D. solani,* Lon negatively affected the level of proteins associated with the type III secretion system: HrpN and a homolog of HrpW as well as HrpA (structural protein of T3SS pillus). These results are consistent with data obtained for *E. amylovora* and *P. syringae*. In these bacteria, Lon indirectly downregulates transcription of *hrpL* gene coding for HrpL, the RNA polymerase sigma factor, which is necessary for the initiation of transcription of T3SS genes. In *P. syringae* and *E. amylovora* this is mediated via degradation of the transcriptional activators of the *hrpL* gene, HrpR and HrpS, respectively [46,51]. In addition, Lon also indirectly regulates HrpS levels through RcsA proteolysis in the *E. amylovora* cells. RcsA is a component of RcsA/RcsB regulatory complex, which activates transcription of the *hrpS* gene [46]. Finally, Lon of *E. amylovora* degrades HrpA [46], which may also be true in the case of *D. solani*, as we observed an elevated level of this protein in the mutant strain. Interestingly, in *E. coli*, the Lon substrates, CspG and CspE proteins, positively regulate the *rcsA* expression [52]. In the *D. solani* Δ *lon* strain, CspG and CspE were upregulated. It cannot be ruled out that they can also be degraded by the Lon protease, which would additionally suppress the expression of T3SS.

In addition to the IscU, HrpA and CspG/E proteins discussed above, the proteome analysis revealed one more potential substrate for the *D. solani* Lon protease. We observed elevated levels of the RNA-binding protein Hfq, which is a known substrate for Lon in *P. aeruginosa* [53]. Fernandez and colleagues [53] suggested that the accumulation of Hfq can contribute to reduced motility of the *lon* strain. In the case of *D. solani,* these findings need verification.

The most pronounced effect of the *lon* deletion was observed in the case of the protein A0A2K8W3L1\_9GAMM, a homolog of ComEA, whose level was more than 100-fold higher in Δ*lon*. ComEA is necessary for natural cell competence. However, *D. solani* was not reported to exhibit natural competence. Moreover, the calcium chloride transformation method of *D. solani* is highly inefficient and *D. solani* spp. genomes lack in general large plasmids [54]. Interestingly, in *Vibrio cholerae comEA* expression is activated by the transcriptional regulator CytR [55]. We also observed an increased level of CytR in the deletion strain, which may be the reason for the upregulation of the ComEA protein. However, the relationship between CytR and Lon protease requires further investigation.

Finally, SWATH-MS analysis revealed that the deletion of *lon* affected the balance among proteins involved in cellular metabolism and transport. This is not surprising as the housekeeping proteases, in general, regulate metabolic activities [56] and this is also true for Lon [14].

In light of the data presented in this work, the Lon protease is a protein that plays very important roles in *D. solani* physiology, both under physiological and stressful conditions. Lon was shown to be required for the full virulence of *D. solani* in the whole plant model. Lower pathogenicity of the Δ*lon* bacteria may result from impaired expression/activity of certain virulence factors, including motility and secreted pectinases, but also from decreased ability to withstand stressful conditions. To our knowledge, this is the first report that describes the function of the Lon protein in the bacterial species from the SRP group.

#### **4. Materials and Methods**

#### *4.1. Materials*

Restriction enzymes and dNTPs were purchased from EURx (EURx Sp. z o.o., Gda ´nsk, r, Poland); PrimeSTAR GXL polymerase for construction of deletion strain from Takara Bio Inc. (Shiga, Japan); T4 DNA ligase, T5 Exonuclease and Phusion High-Fidelity polymerase from New England Biolabs (USA) and reagents for media and buffers from Sigma-Aldrich (Saint-Louis, MI, USA), and Chempur (Piekary Sl ˛ ´ askie, Poland). Oligonucleotides were synthesized by Eurofines Scientific (Luxembourg, Luxembourg) or Sigma-Aldrich (Saint Louis, MI, USA).

#### *4.2. Bacterial Growth Conditions*

Bacterial strains and plasmids used in this study are listed in Table 2. Bacteria were grown in the minimal medium M63Y (0.1 M KH2PO4, 15 mM (NH4)2SO4, 9 μM FeSO4, 1 mM MgSO4, 1 mg/L vitamin B1 and 0.3% glycerol, pH = 7.0) [57], LB broth (1% tryptone, 0.5% yeast extract, 1% NaCl) or SOC (2% tryptone, 0.5% yeast extract, 10 mM NaCl, 20 mM glucose, 2.5 mM KCl, 10 mM MgSO4) with shaking at 30 ◦C, unless indicated otherwise. For all analyses, overnight cultures were diluted 1:50 with M63Y or LB and cultured for the next 16 h until they reached the early stationary growth phase. Then, they were used in experiments or diluted again 1:50 with M63Y and grown for 4.5 h to reach a midexponential phase. The overnight *D. solani* Δ *lon* cultures were grown in the medium supplemented with kanamycin (0.1 mM); the cultures directly subjected to experiments were devoid of the antibiotic to provide comparable growth conditions of all bacterial strains.


#### **Table 2.** Bacterial strains and plasmids used in this study.

Please define all abbreviations in table footer, if appropriate.

Growth curves were determined with the use of the EnSpire plate reader (PerkinElmer, Waltham, MA, USA) in a 24-well nontreated plate (#702011 Wuxi NEST Biotechnology Co., Ltd., Wuxi, China). Overnight grown cultures were diluted 1:50 with LB medium to a final volume of 1 mL. Bacteria were grown with orbital shaking (120 rpm) and OD595 measurements were taken every hour. Growth was monitored for 20 h at 30 ◦C. The final OD values were averaged for four biological replicates.

To induce stress, 4 μL aliquots of 10-fold serial dilutions of the stationary bacterial cultures in Ringer buffer (147 mM NaCl, 4 mM KCl, 2.24 mM CaCl2 × 2H2O) were spotted onto the LA agar plates (LB broth with 1.5% agar, control conditions and temperature stress), LA agar plates supplemented with 0.6 M sucrose (nonionic osmotic stress) or 0.3 M NaCl (ionic osmotic stress), or adjusted with malic acid to pH 5.0 and incubated for 20–48 h at 30 ◦C or 39 ◦C (heat shock). The Agar disk diffusion method was used to study the susceptibility of bacteria to oxidative stress. One-hundred microliters of overnight culture diluted 100 times was spread onto the LA medium. A sterile disk (6mm) of Whatman 1M paper was placed on an LA plate and then 8 μL of 1% hydrogen peroxide solution was spotted on it. The water-soaked disk served as a negative control. Plates were incubated at 30 ◦C for 24 h.

To analyze gene expression, the WT strain grown in M63Y to the midexponential or early stationary phase was subjected to the selected stress conditions for 15 min [33]. Briefly, NaCl and sucrose were added to the cultures to a final concentration of 0.3 M and 0.32 M, respectively. The shift of temperature was obtained by incubation of bacteria for 15 min in a water bath at 37 ◦C or 40 ◦C with shaking. As a control, bacteria grown in the absence of a stressor at 30 ◦C were used. To stabilize mRNA, a cold solution of 5% acid phenol (BioShop Canada Inc., Ontario, Canada), in 99.9% ethanol was added to the bacterial culture at the 1:9 ratio and bacteria were immediately put on ice.

#### *4.3. Construction of the lon Deletion Strain*

The deletion of the *lon* gene from the *D. solani* chromosome was performed according to gene doctoring protocol [62]. Briefly, primers with homology to the upstream/ downstream regions of the kanamycin resistance cassette from pDOC-K plasmid and 40 bp flanking region of the *lon* gene were designed (Table 3). Restriction sites for XhoI and KpnI were added at 5 end of lonkan L and lonkan R primers, respectively. PCR reaction was performed with pDOC-K as a template. The PCR product and pDOC-C plasmid were digested with XhoI and KpnI restriction enzymes, then the PCR product was cloned into the backbone of pDOC-C, generating pDFDOC-C-lon. Plasmid pABSCE and pDFDOC-C-lon were electroporated into *D. solani* cells and the transformants were selected for resistance to chloramphenicol and ampicillin. The proper recombinants were selected based on a lack of ability to grow on a medium with 8% filtrated sucrose, as the pDFDOC-C-lon plasmid contains the *sacB* gene. To check this, 1 mL portions of LB medium with 0.5% glucose and appropriate antibiotics were inoculated with single colonies and incubated with shaking at 30 ◦C for 4 h. The culture was centrifuged (1167× *g*, 2 min), pellet resuspended in 1 mL LB medium supplemented with 0.1–2% arabinose and incubated at 30 ◦C with shaking until turbid. Bacteria were spotted on the LA plates supplemented with 8% sucrose (sterilized by filtration) and kanamycin and in parallel on LA with kanamycin. Next, the colonies that did not grow on the sucrose plates were tested for pDFDOC-C-lon and pABSCE plasmid loss by the selection of bacteria unable to grow on ampicillin and chloramphenicol. PCR with primers homologous to the flanking region of a *lon* gene (Table 3) was performed to confirm the deletion of the *lon* gene.



#### *4.4. Single-Copy Complementation*

Complementation strain was obtained by reintroduction of the WT *lon* gene into its native locus on the chromosome in the *D. solani* Δ*lon* cells using conjugation strain *E. coli* MFD *pir*, according to [60]. To do this, a plasmid containing the WT *lon* gene with the *mScarlet* gene coding for a fluorescent pink protein as a marker was obtained by the Gibson assembly approach. Briefly, four insert fragments were amplified (Table 3) and mixed with the allelic exchange pRE112 vector cut with the SmaI restriction enzyme, and reaction mix [65]. pRE112 carries the *sacB* marker gene and chloramphenicol resistance gene. The total amount of DNA in the reaction was 150 ng.

The resulting reaction product was transformed into *E. Coli* DH5α *pir* and the subsequently isolated plasmid was named pLonScar. The pLonScar plasmid was introduced into *E. coli* MFD *pir* in the presence of 0.3 mM diaminopimelic acid (DAP, Sigma-Aldrich, Saint Louis, MI, USA) in the medium to allow the growth of bacteria [60]. The overnight cultures of *E. coli* MFD *pir* (pLonScar) and *D. solani* Δ*lon* were mixed in a 3:1 ratio (total volume 800 μL) and centrifuged (1677× *g*, 2 min). The pellet was suspended in 30 μL of LB and spotted on an acetate cellulose filter placed on the LA solid medium supplemented with chloramphenicol but without DAP to eliminate *E. Coli* MFD *pir*. After 24 h incubation at 30 ◦C, bacteria were recovered by shaking the filter in 1 mL M63Y. Bacteria were spotted onto the LA agar plates with chloramphenicol. Then, individual colonies were tested for loss of the pRE112 plasmid. Briefly, cultures in the middle logarithmic growth phase were serially diluted and spread on LA without NaCl but supplemented with 10% of 0.22 μL filtered sucrose. Only cultures unable to grow on medium with sucrose were subjected for further verification. Ultimately, pink colonies sensitive to chloramphenicol and kanamycin were considered as true recombinants.

#### *4.5. Plasmid and Genomic DNA Purification*

Genomic and plasmid DNA were isolated using Genomic Mini (A&A Biotechnology, Gdynia, Poland) and Plasmid Mini (A&A Biotechnology, Poland) kits, respectively, according to the manufacturer's protocols.

#### *4.6. Preparation of Electrocompetent Cells and Electroporation*

50 mL of SOC medium was inoculated with an overnight culture of *D. solani* at a 1:50 ratio. Bacteria were grown until OD595 of 0.45–0.5. The culture was centrifuged (7 min, 5063× *g*, 20 ◦C), pellet resuspended in 50 mL of deionized water, mixed thoroughly and forwarded to centrifugation (8 min, 5872× *g*, 20 ◦C). The cells were suspended in 25 mL of deionized water, mixed and centrifuged again (as above). The supernatant was precisely discarded, bacteria suspended in 1 mL of deionized water and split into 60 μL portions. Then, the cells were immediately used for electroporation [66]. Briefly, up to 50 ng of DNA was mixed with electrocompetent cells and transferred to 0.1 cm gap electroporation cuvettes (room temperature) for electroporation (1.25 kV). The bacterial suspension was diluted with 1 mL of SOC medium and incubated for up to 3 h for recovery. The 100 μL aliquots and the remaining bacteria (after 1 min 1677× *g* centrifugation and resuspension in 100 μL SOC) were plated onto the LA solid medium supplemented with an appropriate antibiotic and incubated for 24–36 h at 30 ◦C.

#### *4.7. In Vivo Infection of the Potato Plants*

The pot grown potato plants were obtained from sprouts. Briefly, the potato tubers of cultivar Vineta, obtained locally in Gda ´nsk, Poland, were stored in the dark until the development of sprouts (app. 3–4 months). Sprouts of a length of ca. 5 cm were carefully removed from tubers, planted into 7 cm square pots with potting soil (COMPO SANA ca 50% less weight) and placed on the windowsill for rooting and shoot development. After approximately two weeks, the rooted green plants were transferred to the humid growth chamber and grown under the white fluorescent light (48 × 5 W, Mars Hydro Reflector 48 with 16:8 h light: dark photoperiod). The potato plants at least 10 cm high were subjected to pathogenicity tests. Four overnight cultures of WT and four of *D. solani* Δ*lon* grown in LB medium were diluted with Ringer buffer to OD595 ~0.125, corresponding to 10<sup>8</sup> CFU/mL. The roots of each potato plant were wounded with the scalpel about 2 cm from the stem. Plants were watered with 30 mL of bacterial suspensions and left for an hour. Then, the filtrate was discarded. Each bacterial culture was used to infect 4 plants. As a negative control, four plants treated with Ringer buffer were used. The experiment was carried out for 17 days and the percentage of plants with blackleg symptoms was estimated.

#### *4.8. Pathogenicity on Potato Tubers, Chicory and Chinese Cabbage Leaves.*

CFU/mL of each overnight bacterial culture was normalized to 108 with Ringer buffer and then 10-fold serially diluted. Potato tubers were sterilized with a 10% bleach solution for 20 min, then submitted to three washes with sterile deionized water for 20 min each. The tubers were cut into 1 cm thick potato slices; in each slice, a little hole was pierced with a sterile pipette tip. Chicory and Chinese cabbage leaves were washed with sterilized deionized water and incised with a sterile scalpel. Ten-microliter aliquots of bacterial culture of 10<sup>7</sup> CFU/mL were spotted onto the plant material. Ringer buffer was used as a negative control. The infection assays were performed in the humid boxes, at 30 ◦C for up to two days: one day for Chinese cabbage, two days for chicory leaves and for potato slices.

#### *4.9. Determination of Motility*

For swimming, a single bacterial colony (five replicates per strain) was inoculated into the semisolid agar plate with 0.3% MMA medium (40 mM K2HPO4, 22 mM KH2PO4, 0.41 mM MgSO4 × 7H2O, 0.3% agar) supplemented with 1 mM galactose. The plates were incubated under aerobic conditions at 30 ◦C for 48 h. The diameter of the bacterial spreading area was measured.

To monitor swarming motility, a single bacterial colony (five replicates per strain) was inoculated into the plate with 0.5% TSA (tryptone soy broth) medium (Oxoid, Basingstoke, UK) supplemented with 0.5% agar). Plates were incubated under aerobic conditions at 30 ◦C for 12 h. Both tests were repeated two times.

#### *4.10. Determination of Secreted PCWDE Activity*

The measurement of pectinolytic activity was performed as described in [67]. Briefly, bacteria were cultured in the M63Y medium until an early stationary phase and centrifuged (13,148× *g*, 2 min). Then, 260 μL aliquots of supernatant were diluted with equal volumes of distilled water. Briefly, 500 μL samples of the diluted supernatant were mixed with 1.5 mL of PGA (polygalacturonic acid, Sigma-Aldrich, Saint Louis, MI, USA) buffer (100 mM Tris–HCl (pH 8.5), 0.35 mM CaCl2 and 0.24% sodium polygalacturonate) warmed up to 30 ◦C. The reaction consisting in the formation of unsaturated products from polygalacturonate [68] was monitored spectrophotometrically by measurement of increase of absorbance at 232 nm for 2 min, every 30 s. Absorbance values obtained for PGA buffer were subtracted from values obtained for unsaturated product. The spectrophotometer was calibrated with distilled water. Pectynolytic activity was presented as ΔA235/min/mL/OD595. The experiment was repeated two times for each strain with at least three replicates.

The extracellular cellulase activity was assayed as described in [69]. Briefly, bacteria grown in the M63Y medium to the stationary phase were diluted with Ringer buffer to 108 CFU/mL. Seven-microliter aliquots of bacterial cultures were spotted on the agar plates with carboxymethyl cellulose CMC (M63Y medium supplemented with 1.5% agar and 1% CMC). The plates were incubated at 30 ◦C for 48 h and then subjected to staining with 2% Congo red solution for 20 min. The diameters of the arisen halo were measured. The experiment was repeated two times for each strain with five replicates.

To measure the extracellular protease activity, bacteria cultivated in the M63Y medium to the stationary phase were diluted with Ringer buffer to 10<sup>8</sup> CFU/mL. Seven-microliter aliquots of bacterial cultures were spotted on the milk agar plates (the LA medium supplemented with 5% skimmed milk). Plates were incubated at 30 ◦C for 48 h and the diameters of the arisen halo were measured. Each strain was tested two times with five replicates.

#### *4.11. Siderophore Activity Assay*

Ten-microliter aliquots of supernatants from the stationary cultures (grown in M63Y medium for 16 h) were spotted on the chrome azurol S-agar plates [70]. To prepare the chrome azurol S-agar medium the following solutions were prepared: (1) main medium, (2) 10% deferrated casamino acids (CAS), (3) 0.1 M CaCl2, (4) filtered 1 mM FeCl3 × 6H2O in 10 mM HCl, (5) CTAB (cetrimonium bromide, Sigma-Aldrich, Saint Louis, MI, USA). To prepare the main medium solution the components, KH2PO4 (3 g), NaCl (0.5 g), NH4Cl (1.0 g), MgSO4 × 7H2O (0.2 g), sucrose (4.0 g) and agar (15.0 g), were dissolved in 850 mL of 0.5 M Tris–HCl, pH 6.8 and sterilized. The deferrated casamino acids were prepared by the removal of ferrous ions with 3% 8-hydroxyquinoline in chloroform. The sterilized main solution was supplemented by 30 mL of 10% deferrated casamino acids, 10 mL of 0.1 M CaCl2, 50 mL of 0.08 mM CAS, 10 mL of 1 mM filtered FeCl3 × 6H2O in 10 mM HCl and 40 mL of 2 mM CTAB (CAS, FeCl3 and CTAB were mixed before adding to solution). Plates were incubated at 30 ◦C for 1 h and the intensity and diameter of the orange halo were compared. The experiment was performed two times for five replicates for each strain.

#### *4.12. RNA Extraction*

Bacterial RNA was extracted using the Total RNA Mini Plus RNA extraction kit (A&A Biotechnology, Gdynia, Poland) according to the manufacturer's instructions. The quantity and quality of the RNA samples were confirmed by measurement of absorbance at 260 nm and evaluation of A260/A280 (~2) and A260/A230 (>2) ratios, and by agarose gel electrophoresis. Samples of 5 μg of RNA were subjected to DNase treatment (A&A Biotechnology, Poland) by incubation of 20 μL reaction mixtures in the presence of DNase (1U/μL) at 37 ◦C for 25 min followed by incubation at 75 ◦C for 10 min. The samples served as a template for the reverse transcription reaction.

#### *4.13. Reverse Transcription*

cDNA was transcribed from 1.5 μg of RNA with the use of RevertAid First Strand cDNA Synthesis Kit (Thermo Fisher Scientific, Waltham, Massachusetts, USA), according to manufacturer's protocol. Obligatory step of denaturation of RNA with random hexamer primers mixture at 65 ◦C for 5 min was added.

#### *4.14. Quantitative Real-Time PCR (qPCR)*

qPCR analysis was performed as described in [23]. Briefly, diluted cDNA samples in a 1:2 ratio were used as qPCR templates. The qPCR reactions were carried out using the LightCycler 96 instrument (Roche Diagnostics, Rotkreuz, Switzerland). Primer3 software was used to design primers [71] (Table 4). Ten-fold dilution series of genomic DNA templates isolated from *D. solani* IPO 2222 were used to estimate the amplification efficiency of each pair of primers. qPCR reaction was carried out with FastStart Essential DNA Green Master (Roche Diagnostics, Rotkreuz, Switzerland). A 20 μL qPCR reaction mixture contained 0.5 μL of cDNA, 3–4.5 pmol of forward and reverse primers, 10 μL of PCR Mix. Thermal cycling parameters were as follows: preincubation at 95 ◦C for 5 min; 35–50 cycles of amplification and quantitation at 95 ◦C for 15 s, 62 ◦C for 20 s and 72 ◦C for 16 s. At the end of each cycle, melting curve analysis was performed (95 ◦C for 10 s, 65 ◦C for 60 s and 97 ◦C for 1 s). All qPCR reactions were performed for three biological replicates, with three technical repeats, negative no template control (NTC) and no-reverse transcriptase (NRT) controls. Cq (quantification cycle) values were averaged. The 16s rRNA gene was selected for normalization as it showed stability under all tested conditions. Pfaffl-ΔΔCT method with correction for PCR efficiency was used for the

determination of the relative expression of the *lon* gene [72]. Statistical analysis was performed with the use of REST2009 software (v. 2009, Qiagen, Hilden, Germany) [73,74].


**Table 4.** Characteristics of primers used in gene expression analysis.

#### *4.15. Protein Electrophoresis and Immunodetection*

SDS page electrophoresis and Western blotting were performed as described in [75,76]. Then, 7.5% polyacrylamide gels were used. Briefly, the Lon protein was detected with the anti-*Escherichia coli* Lon rabbit antibodies (#40219-T24, Sino Biological Inc., Beijing, China) at dilution 1:2000 followed by incubation with HRP conjugated secondary anti-rabbit antibodies (#31462 Thermo Fisher) diluted 1:50,000. Chemiluminescent signal was developed using a luminol/ p-coumaric acid (Carl Roth GmbH + Co. KG) mix (4 mL of 1.41 mM luminol, 400 μL of 6.7 mM p-coumaric acid in DMSO, 4 μL of 30% H2O2) and was recorded by Azure Biosystems c600 (Dublin, California, USA) imaging system.

#### *4.16. Sample Preparation for Mass Spectrometry*

The stationary growth phase cultures of *D. solani* cultivated in M63Y were subjected to high-temperature stress. Briefly, the cultures were transferred from 30 ◦C to 40 ◦C and incubated for 30 min with shaking. For control conditions, bacteria were cultivated at 30 ◦C. Five biological replicates of each strain were pooled and centrifuged (7000× *g*, 2 min). The pellets were lysed with the solution containing 4% SDS, 100 mM Tris/HCl pH 7.6, 0.1 M DTT (lysis solution) and incubated at 95 ◦C for 10 min. After cooling, cold acetone was added to the solution to precipitate the released proteins. The samples were incubated at -20 ◦C for about 2 h and then centrifuged for 20 min 20,000× *g*. The supernatant was decanted and the precipitate dried. The pellet was then dissolved in 8 M urea in 0.1 M Tris/HCl pH 8.5 [77].

#### *4.17. Protein Digestion*

First, the protein concentration was measured by measuring absorbance at 280 nm (MultiskanTM Thermo, Waltham, Massachusetts, USA) using the μDrop plate. Digestion was carried out according to the standard Filter Aided Sample Preparation (FASP) procedure [77]. Then, 100 μg of protein was used for each digestion and the procedure was carried out using microcons with 10 kDa mass cut-off membrane. Generated tryptic peptides were desalted with StageTips according to the protocol described by Rappsilber [78]. For each desalting step, 10 μg of the peptide was taken and desalted on StageTip containing three layers of 3 M Empore C18 exchange disks.

#### *4.18. Liquid Chromatography and Mass Spectrometry*

LC-MS/MS analysis was performed with the use of a Triple TOF 5600+ mass spectrometer (SCIEX Framingham, MA) coupled with the Ekspert MicroLC 200 Plus System (Eksigent, Redwood City, California, USA). All chromatographic separations were performed on the ChromXP C18CL column (3 μm, 120 Å, 150 × 0.3 mm). The chromatographic gradient for each IDA and SWATH runs was 3.5–20% B (solvent A 0% aqueous solution 0.1% formic acid, solvent B 100% acetonitrile 0.1% formic acid) in 60 min. The whole system was controlled by the SCIEX Analyst TF 1.7.1 software (version 1.7.1, Framingham, MA, USA).

#### *4.19. SWATH Mass Spectrometry Experiments*

All samples were acquired in triplicates. Experiments were performed in a looped product ion mode.

A set of 25 transmission windows (variable wide) was constructed and covered the precursor mass range of 400–1200 *m*/*z*. The collision energy for each window was calculated for +2 to +5 charged ions centered upon the window with a spread of two. The SWATH-MS1 survey scan was acquired in high sensitivity mode in the range of 400–1200 Da in the beginning of each cycle with the accumulation time of 50 ms, and SWATH-MS/MS spectra were collected from 100 to 1800 *m*/*z* followed by 40 ms accumulation time high sensitivity product ion scans, which resulted in the total cycle time of 1.11 s.

#### *4.20. Data Analysis*

Database search was performed with ProteinPilot 4.5 software (Sciex, v.4.5 AB, Framingham, MA, USA) using the Paragon algorithm against the UNIPROT *Dickeya solani* database with an automated false discovery rate, and standard parameters [79,80]. Next, a spectral library was created with the group file data processing in PeakView v. 2.2 (Sciex), with parameters as described in detail by Lewandowska [79]. Files from SWATH experiments for each sample were downloaded to PeakView (Sciex, v.2.2, Framingham, MA, USA) software and processed with the previously established library. Resulting data were exported to the .xml file and exported to Marker View software. All data were normalized using total area sums (TAS) approach, grouped as wild type and tested samples and *t*-test was performed. Samples were compared to each other, coefficient of variation (CV%) was calculated, and proteins with a *p*-value lower than 0.05 with fold change 2 were considered as differentially expressed in examined samples. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE [81] partner repository with the dataset identifier PXD018297.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/1422-0067/21/10/3687/s1, Figure S1: Verification of the constructed Δ*lon*/*lon* strain, Table S1: SWATH-MS raw data, Figure S2: Ability of *D. solani* Δ*lon* to cause maceration of plant tissues: (A) potato tubers (B) chicory leaves and (C) Chinese cabbage leaves.

**Author Contributions:** Conceptualization, J.S.-G.; Data curation, D.F.; Formal analysis, D.F., P.C. and J.S.-G.; Funding acquisition, J.S.-G.; Investigation, D.F., P.C., T.P. and P.A.; Methodology, D.F., P.C. and J.S.-G.; Supervision, J.S.-G.; Visualization, D.F.; Writing—original draft, D.F., P.C. and J.S.-G.; Writing—review and editing, D.F., T.P., P.A., M.P. and J.S.-G. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the National Science Centre, Poland, NCN OPUS-7 UMO-2014/13/B/NZ9/02021.

**Acknowledgments:** We thank Guy Condemine, Erwan Guegen, Nicole Hugouvieux-Cotte-Pattat and Vladimir Shevchik from Claude Bernard University Lyon 1, for the knowledge support in complementation strain construction. The strains *E. coli* DH5α *pir*, *E. coli* MFD *pir* and pRE112 vector, which we used in our research, were donated by Erwan Guegen. We also thank Daniel Stukenberg and Georg Fritz for providing the mScarlet plasmid.

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

#### **Abbreviations**


#### **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/).

## *Article* **A Novel E**ff**ector Protein of Apple Proliferation Phytoplasma Disrupts Cell Integrity of** *Nicotiana* **spp. Protoplasts**

**Cecilia Mittelberger 1, Hagen Stellmach 2, Bettina Hause 2, Christine Kerschbamer 1, Katja Schlink 1, Thomas Letschka <sup>1</sup> and Katrin Janik 1,\***


Received: 7 September 2019; Accepted: 14 September 2019; Published: 18 September 2019

**Abstract:** Effector proteins play an important role in the virulence of plant pathogens such as phytoplasma, which are the causative agents of hundreds of different plant diseases. The plant hosts comprise economically relevant crops such as apples (*Malus* × *domestica*), which can be infected by '*Candidatus* Phytoplasma mali' (P. mali), a highly genetically dynamic plant pathogen. As the result of the genetic and functional analyses in this study, a new putative P. mali effector protein was revealed. The so-called "Protein in *Malus* Expressed 2" (PME2), which is expressed in apples during P. mali infection but not in the insect vector, shows regional genetic differences. In a heterologous expression assay using *Nicotiana benthamiana* and *Nicotiana occidentalis* mesophyll protoplasts, translocation of both PME2 variants in the cell nucleus was observed. Overexpression of the effector protein affected cell integrity in *Nicotiana* spp. protoplasts, indicating a potential role of this protein in pathogenic virulence. Interestingly, the two genetic variants of PME2 differ regarding their potential to manipulate cell integrity. However, the exact function of PME2 during disease manifestation and symptom development remains to be further elucidated. Aside from the first description of the function of a novel effector of P. mali, the results of this study underline the necessity for a more comprehensive description and understanding of the genetic diversity of P. mali as an indispensable basis for a functional understanding of apple proliferation disease.

**Keywords:** phytoplasma; effector protein; apple; pathogenicity; virulence; apple proliferation

#### **1. Introduction**

Phytoplasma are small, biotrophic bacteria that cause hundreds of different plant diseases and are involved in their infection cycle not only in plant hosts, but also in insect vectors. '*Candidatus* Phytoplasma mali' (P. mali), the causal agent of apple proliferation (AP) disease, has caused significant economic losses in apple production in Northern Italy (one of Europe's main production areas) in the last decades [1]. Phytoplasma are obligate plant and insect symbionts that exhibit a biphasic life cycle comprising reproduction in certain phloem-feeding insects as well as in plants [2,3]. Within their plant host, phytoplasma colonize the phloem. By ingestion of phloem sap, insect vectors acquire the phytoplasma, with the colonization of those insects enabling the transmission of the pathogen between host plants [3,4]. Although several concepts of phytoplasma effector biology were able to be unraveled for the '*Candidatus* Phytoplasma asteris' strain Aster Yellow Witches' Broom (AY-WB) in the model plant *Arabidopsis thaliana* [5–12], the understanding of effector-driven changes induced by P. mali remain limited. Genetic and functional homologues of AY-WB phytoplasma protein SAP11 could be identified in P. mali [10,13]. Recently a novel effector was described that exhibits E3 Ubiquitin ligase function and affects the plant's basal defense [14]. Furthermore, the immunodominant membrane protein Imp of P. mali was shown not to be involved in symptom development but is considered to play a role during plant colonization [15]. A role of phytoplasmal HflB proteases and an AAA+ ATPase in AP virulence has been hypothesized but not yet clarified [16–18]. P. mali encodes genes for a Sec-dependent protein secretion system, whereas genes encoding components of other secretion systems, such as the type three secretion system, are mainly lacking [19,20]. Secreted phytoplasma proteins may directly interact with cellular host components and thus manipulate the cell's metabolism [3]. Potential effector proteins may thus be identified by the presence of a characteristic *N*-terminal secretion signal.

The aim of this study was to characterize the function of the phytoplasmal "Protein in *Malus* Expressed 2" (PME2) from P. mali that exhibits genetic features indicating that it acts as an effector protein in plants. To unravel PME2s potential role as an effector, this study analyzed (1) whether it is genetically conserved; (2) whether it is expressed during infection; (3) where it is translocated within the plant cell; and (4) if it induces morphological changes within the expressing plant cells.

To address these questions, we analyzed the expression of PME2 in P. mali-infected *Malus* × *domestica* leaf and root tissue, and in infected *Cacopsylla picta* (i.e., insect species transmitting P. mali). In infected *Malus* × *domestica* we found two distinct genetic variants of *pme2*. In addition, heterologous overexpression of PME2 in mesophyll protoplasts of *Nicotiana* spp. was used to gain insights into the subcellular localization of PME2 as well as its effects on plant cell integrity. These data were complemented by the expression of PME2 in yeast. With the data presented here, the first steps into unraveling the molecular mechanism of PME2 function were taken, but further experiments in the future will be indispensable.

#### **2. Results**

#### *2.1. In Silico Analysis of PME2 Indicates E*ff*ector Potential*

Bioinformatic analysis of conserved hypothetical proteins encoded in the P. mali genome [19] revealed that CAP18323.1, encoded by the gene *atp\_00136,* contains interesting features that might confer effector function. Neural networks and hidden Markov prediction models (Transmembrane Helices Hidden Markov Model; TMHMM) were applied to analyze CAP18323.1 for the presence of a signal peptide and the presence of transmembrane regions (SignalP v. 3.0 [21], TMHMM [22]). Since phytoplasma phylogenetically belong to Gram-positive bacteria [3], a prediction algorithm trained on this bacterial group was applied. The N-terminal amino acid-stretch 1–31 contains a signal peptide that is supposed to confer Sec-dependent secretion of the protein (Figure 1). Further transmembrane regions were not predicted, indicating that CAP18323.1 is not inserted in a membrane. Upon translation, *N*-terminal signal peptides are cleaved [23]. At the C-terminal part of CAP18323.1 an importin α/β-dependent nuclear localization site (NLS) and a nuclear export signal (NES) were predicted [24,25]. The absence of transmembrane regions in the mature protein, the predicted localization in the plant cytoplasma or the nucleus (WoLF PSORT, [26]), and the small size of about 16 kDa (Analysis Tool on the ExPASy Server, [27]) indicate that CAP18323.1 may exhibit an effector function (Figure 1).

#### *2.2.* Atp\_00136 (Pme2) *is Expressed in P. Mali-Infected Malus* × *Domestica but not in the Insect Vector C. Picta*

Subsequently, it was analyzed whether *atp\_00136* was expressed in apple trees infected with P. mali. Leaf and root samples of P. mali-infected and non-infected *Malus* × *domestica* cv. "Golden Delicious" trees were taken in May and October. Expression of *atp\_00136* was analyzed with *atp\_00136*-specific primers and *Malus* × *domestica* cDNA derived from mRNA. Expression of *atp\_00136* was confirmed in P. mali-infected leaf and root tissue by the detection of distinct amplicons at the expected size in the respective samples (Figure 2).

**Figure 1.** Results of the in silico analysis. Sequence analysis of *atp\_00136* revealed the presence of an *N*-terminal signal peptide (indicated in red), as well as a nuclear localization signal (NLS), and a C-terminal nuclear export signal (NES), both indicated in green. Graphs were generated with Geneious Prime 2018 version 11.1.4.

**Figure 2.** Expression of *pme*2 (CAP18323.1) in '*Candidatus* Phytoplasma mali' (P. mali)-infected *Malus* × *domestica*. Transcripts of *pme*2 were detected by PCR using cDNA from *Malus* × *domestica* infected with P. mali. A discrete band of the size indicative for the *pme*2 transcript was detected in P. mali-infected (**Pm**+) but not in non-infected (Pm–) leaves and roots harvested in October. DNA derived from an infected *Malus*×*domestica* served as a positive control (pc) and water as the non-template control (ntc).

Using quantitative PCR (qPCR) the expression levels of *atp\_00136* and P. mali in the samples were quantified. The results show that *atp\_00136* is only expressed in tissue colonized by P. mali (Table 1). Since identified expressed genes were named in a chronological manner, *atp\_00136* was named "Protein in *Malus* Expressed 2" (*pme2*) based on the general recommendations for bacterial gene nomenclature [28].

To analyze if *pme2* was expressed in the transmitting insect vectors during infection, three P. mali-infected *C. picta* individuals were analyzed for the expression of the potential effector. In the RNA/cDNA of all infected individuals, P. mali-specific transcripts of the ribosomal protein *rpl22* were detected, but expression of *pme2* was not detectable.


**Table 1.** Detection of *atp\_00136* in cDNA samples from infected and non-infected leaf tissue from May and October 2011. In May phytoplasma were only detectable in the roots but not in the leaves. *atp\_00136* was only detectable in P. mali-infected and colonized tissue. Cq values are given as the mean value of three repeated qPCR runs.

#### *2.3. Genetic Variability of* Pme2

Cloned amplicon sequencing revealed that the prevalent variant of *pme2* from infected trees in South Tyrol (North-East Italy) differs compared to the *pme2* sequence of the P. mali AT strain from Germany [19]. In a total of 20 samples from naturally infected apple trees in the regions Burggraviato and Val Venosta, a prevalent, conserved sequence of *pme2* was identified (*pme2*ST; accession number MN224214). This conserved variant exhibits a single nucleotide polymorphism (SNP) in the sequence stretch before the NLS, and two SNPs within and one SNP after the NLS compared to the AT strain (Figure 3). All four SNPs in the *pme2*ST variant lead to nonsynonymous missense substitutions at the protein level as compared to the *pme2* sequence published previously [19] (*pme2*AT). The NLS of *pme2*ST has a slightly higher prediction score than the NLS of *pme2*AT. The most striking difference between *pme2*AT and *pme2*ST is a stretch of 120 bp in *pme2*ST which is absent in *pme2*AT. This stretch is a partial duplication of a fragment also present in *pme2*AT (Figure 3). In three *Malus* × *domestica* samples, a very sporadic sequence of *pme2* could be detected that did not contain the *pme2*ST characteristic sequence duplication but showed strong sequence similarity to *pme2*AT. The sporadic sequence contains six SNPs at positions 218 (A > T), 220 (A > G), 322 (A > G), 331 (A > C), 344 (C < T), and 427 (T > G) that lead to nonsynonymous missense mutations (accession number MN224215) compared to *pme2*AT. However, in the trees in which these very sporadic *pme2* sequences were found, *pme2*ST could also be detected, indicating the presence of a mixed population of different P. mali strains.

**Figure 3.** Sequence comparison of PME2ST and PME2AT. The protein variants (**a**) PME2AT and (**b**) PME2ST contain the same *N*-terminal signal peptide sequence (red). PME2ST (**b**) contains a duplicated amino acid stretch (the replicative sequences 1 and 2; marked in blue) of a partial sequence also present in PME2AT (**a**). Both variants show slight differences in and directly before the nuclear localization signal sequences (NLS, green). The nuclear export signal sequence (NES, green) is identical in both protein variants. Amino acid differences of PME2ST to the PME2AT variant are shown in black, whereas similarities are shown in grey. Graphs were generated with Geneious Prime 2018 version 11.1.4.

#### *2.4. PME2ST and PME2AT Translocate to the Nucleus of* Nicotiana *spp. Protoplasts*

To identify the subcellular localization of the PME2 protein in the plant cell, mesophyll protoplasts of *Nicotiana occidentalis* and *N. benthamiana* were transformed, with expression vectors coding for PME2AT and PME2ST tagged with GFP or mCherry-fluorescent protein to allow subcellular tracking. The *N*-terminal signal part was not considered for these studies, since it is removed from the processed, mature CAP18323.1 protein. *N. occidentalis* and *N. benthamiana* can be infected with P. mali. Upon infection, both *Nicotiana* species show disease-specific symptoms and are thus appropriate model plants for P. mali effector studies [15,29]. Confocal microscopy analysis revealed that overexpressed PME2AT and PME2ST are translocated to the nucleus of *Nicotiana* spp. protoplasts. This translocation was independent of the used tag and *Nicotiana* species (Figure 4 and Figures S1–S3). The in vivo results therefore confirm the in silico prediction that PME2ST and PME2AT are translocated to the nucleus of potential host cells.

**Figure 4.** PME2ST and PME2AT are translocated to the nucleus of mesophyll protoplasts. Mesophyll protoplasts of *Nicotiana benthamiana* were transformed with the plasmid pGGZ001 encoding C-terminal GFP-tagged PME2ST (first column), PME2AT (second column), GFP *N*-terminally fused to a NLS sequence (third column), or GFP only as a control for nuclear localization (fourth column). Expression of the transgenes was under the control of a *35S* promoter. The upper panel shows autofluorescence of chloroplasts (Chl), the second panel the signal derived from the GFP, and the third panel the bright field image and the last panel an overlay of all images (merged). Microscopic analysis was performed with a Zeiss LSM 800. Corresponding images after expression of mCherry-tagged PME2 and of use of *Nicotiana occidentalis* mesophyll protoplasts are presented in Figures S1–S3. Bars represent 20 μm.

A leaf infiltration assay using *Agrobacterium* strain EHA105 transformed with PME2 encoding expression vectors did not result in detectable expression or phenotypic alterations of either PME2AT or PME2ST in both *Nicotiana* species. Nonetheless, positive controls expressing the fluorophore tag only and leaves infiltrated with the P. mali SAP11-like effector protein ATP\_00189 [13] as control showed strong signals (Figure S5), indicating that PME2 expression might be somehow blocked or is immediately degraded by the plant.

#### *2.5. PME2ST but not PME2AT A*ff*ect Cell Integrity of* Nicotiana *spp. Protoplasts*

Protoplasts transformed with the PME2ST expression vector often showed shrinkage, and only about 50% of the *N. benthamiana* protoplasts were viable 20 h post-transformation compared to the transformation control expressing the fluorophore only or a GFP with NLS (Figure 5a). The shrunk cells lysed and only the remaining cell debris was microscopically detectable (Figure 4). The effect on protoplast integrity was observed in protoplasts expressing PME2ST:GFP and PME2ST:mCherry, and thus was independent of the fluorophore used as a tag for microscopic analyses. Similar results were obtained using *N. occidentalis* as the heterologous PME2ST expression system. The mCherry-tagged PME2ST induced a weak but significant reduction of viability in *N. occidentalis* protoplasts (Figure 5b). The GFP-tagged PME2ST showed the same tendency but to a stronger extent, i.e., it reduced cell viability by about 50%, which is similar to the effect seen in *N. benthamiana* protoplasts. Cell viability stain with fluorescein diacetate (FDA) showed similar results, i.e., that *N. benthamiana* protoplasts transformed with the PME2ST-expressing vector showed a significantly reduced viability (Figure 6). Shrunk cells were positive for propidium iodide (PI) staining (Figure S4), indicating that these cells were dead.

**Figure 5.** PME2ST overexpression reduces viability of (**a**) *N. benthamiana* and (**b**) *N. occidentalis* mesophyll protoplasts. For each assay, 20,000 mesophyll protoplasts were transformed with the plasmid pGGZ001 encoding PME2ST, PME2AT (tagged with GFP or mCherry), the GFP-tagged control for nuclear localization (NLS), or the mCherry tag (tag only) and viable protoplasts were counted. Overexpression of the transgenes was under the control of a *35S* promoter. Data represent the mean viability +/– SE of 3–4 independent experiments. The respective control (NLS or tag only) was set at 1 to allow comparison between different experiments. Differences between the groups were determined applying a one way-ANOVA analysis. Significant differences (*p* < 0.05) between groups are indicated with an asterisk (\*).

Interestingly, PME2AT did not have an effect on protoplast integrity in *N. benthamiana* nor in *N. occidentalis* protoplasts (Figure 5).

#### *2.6. A Yeast Two-Hybrid Screen Was Unsuitable for the Elucidation of PME2ST Function*

Upon expression of PME2ST, the yeast reporter strain *Saccharomyces cerevisiae* NMY51 showed several macroscopic aberrations in colony growth (Figure 7a). However, at the microscopic level when visualizing the yeast cell wall with calcofluor white, no phenotypic differences between yeast cells expressing PME2AT, PME2ST, and empty bait vector pLexA-N could be detected (Figure 7b). Considering the effect of mere PME2ST expression on growth of the yeast reporter strain, the relevance of any identified interaction in a yeast two-hybrid screen remains highly questionable and the assay was therefore not performed.

**Figure 6.** PME2ST overexpression reduces viability of *N. benthamiana* mesophyll protoplasts. Mesophyll protoplasts were transformed with the plasmid pGGZ001 encoding mCherry-tagged PME2ST (PMEST) or the mCherry tag only (mCherry) and stained with fluorescein diacetate (FDA) to detect viable cells. Data represent the mean percentage of FDA-positive stained cells +/– SE (*n* = 3). The statistical difference between the two groups was determined by using a Student's *t*-test and is indicated with an asterisk (\**p* < 0.05).

**Figure 7.** PME2ST overexpression in the yeast reporter strain NMY51 leads to macroscopic aberrations. *Saccharomyces cerevisiae* strain NMY51 was transformed with the yeast two-hybrid (Y2H) bait vector pLexA-N, which encodes tryptophan auxotrophy, expressing PME2AT, PME2ST, or the empty vector only, and drop-plated onto SD-trp plates. In comparison to the empty pLexA-N vector and the vector expressing PMEAT, colonies expressing PMEST showed reduced growth and remained white (**a**). Yeast cells stained with calcofluor white did not show any phenotypic differences on single-cell level (**b**). Calcofluor white fluorescence was visualized on a confocal microscope (LSM 800, Carl Zeiss AG, Oberkochen, Germany).

#### **3. Discussion**

The results of this study show that PME2ST (a variant of CAP18323.1 previously annotated as "conserved hypothetical protein") affects plant cell integrity. Based on our findings and the definition that effectors are secreted pathogen proteins altering host-cell structure and function [30], we propose defining PME2 as a phytoplasmal effector. Interestingly, two different variants of PME2 were identified and both variants translocate to the nucleus of plant cells, but only the newly described regional variant PME2ST subsequently affects protoplast integrity. The small size of about (at a maximum) 21 kDa (PME2ST: 21 kDa and PME2AT: 16 kDa; both considering the mature protein without the signal peptide) indicates that PME2 can be translocated from the phloem and target adjacent tissues or be distributed systemically in the plant [3]. Subcellular localization using microscopy requires the use of fluorescent tags that are attached to the protein of interest. Tagging can affect subcellular localization of the protein; however, we used two different tags (GFP and mCherry) to analyze whether tagging influences the target localization. In cells expressing the tag only, a localization of the fluorescent signal in the cytoplasm could be observed. PME2 was localized only in the nucleus and since no signal was visible in other cell compartments, it can be assumed that the observed localization is effector-mediated (see also [31]). Only protoplasts transformed with PME2ST showed significant cell disruption as indirectly quantified by counting the remaining viable cells and FDA staining of the protoplasts. Shrunk cells were positive for the PI stain but did not show a GFP signal. The lack of the GFP-signal might be caused by a disruption of the nucleus, protein degradation, and/or leakage of the signal into the surrounding medium. Cells expressing PME2ST were intact, indicating that the effector either exhibits a dose-dependent or delayed effect on cell integrity.

Both variants of PME2 contain an *N*-terminal signal peptide, a nuclear localization signal (NLS), and a nuclear export signal (NES). It is a common feature of nuclear proteins to contain both NLS and NES and these signals coordinate the translocation of the protein between nucleus and cytoplasm [32]. Nuclear targeting of proteins containing a classical NLS is mediated by the importin α/β heterodimer through NLS-dependent binding to the importin α subunit and importin β–mediated attachment to the nuclear pore complex [33,34]. The SNPs in the NLS region of PME2ST lead to a (slightly) higher sequence-based NLS prediction; thus, the differences might show a stronger translocation to the nucleus. The NES signal (which indicates that shuttling of PME2 between nucleus and cytoplasm might occur) is the same in both variants. Even though nucleocytoplasmic distribution is predicted, PME2 was only detected in the nucleus. Many proteins containing NLS and NES appear to be localized in the nucleus because the rate of import to the nucleus is higher than the rate of export to the cytoplasm [35]. It remains thus unclear if PME2 is strictly limited to the nucleus or if a constant shuttling between nucleus and cytosol occurs.

Bacterial effectors that translocate to the nucleus, the so-called nuclear effectors, can affect master switches of the host immune machinery or alter host transcription to the benefit of the pathogen [31]. Effectors from different phytoplasma species target plant–host transcription factors or affect gene expression on the transcriptional level to alter the host metabolism to their own benefit [4,12,13,36,37]. However, none of these effectors have yet been reported to exhibit such detrimental effects during in planta expression. The effector protein BR1 of the phloem colonizing squash leaf curl geminivirus shuttles between the cytoplasm and the nucleus of protoplasts [38]. Upon binding to the second movement protein BL1, BR1 shuttles to the cytoplasma [39] and the concerted action between BR1 and BL1 mediates cell-to-cell movement of the virus within the phloem and to adjacent cells [35,38,40,41]. To unravel BR1 function it was necessary to identify its interaction partner, a general approach to investigate effector function. Yeast two-hybrid (Y2H) screens have been successfully applied to determine phytoplasmal effector targets on the molecular level [10,13]. These screens allow the screening of a protein of interest (effector) against a library containing hundreds of thousands of different potential interaction partners of a certain host species [42,43]. Successful interaction is monitored by a genetic reporter system that complements certain auxotrophies in the recombinant yeast reporter strain. However, a Y2H with PME2ST is not suitable since PME2ST expression strongly

affected the Y2H yeast reporter strain. This effect on yeast cells further supports the finding that PME2ST exhibits a strong effect not only on plant, but also on yeast cells, even though the latter do not have relevance as phytoplasma host cells. Since PME2ST exhibits such a strong effect on the expressing host and non-host cells, alternative approaches must be applied to unravel its molecular function. *Nicotiana* spp. leaf infiltration assays with recombinant *Agrobacterium* strains expressing PME2 failed. It remains furthermore elusive as to whether PME2 exhibits effects on the host plant phenotype. Considering the PME2ST effects on protoplasts it can be assumed that a systemic overexpression would lead to overwhelming deleterious effects in transgenic plants that express this effector. The results show that PME2 is expressed in roots and leaves of infected *Malus* × *domestica*, but not in infected individuals of its insect vector *C. picta*, underlining the hypothesis that PME2 plays a role as an effector protein in plant cells. However, it needs further clarification if expression is fine-tuned in a tempo-spatial manner in the plant host.

A neatly coordinated and local expression during infection might have very local effects and might not lead to cell disruption as seen in heterologous overexpression experiments. It is hypothesized that phytoplasma are able to degrade plant cell walls or generate holes in plant cell membranes to expedite cell-to-cell effector translocation [4]. Infection with P. mali induces cytochemical modifications and injuries of the affected phloem cells [44,45]. It is speculated that plasma membrane integrity is affected by until-now unknown P. mali effector(s) and that plasma membrane disruption is involved in the observed phloem damage induced by virulent P. mali strains [45]. However, since molecular indications are lacking, interpretation of the mode of PME2 action remains speculative. Subsequent approaches to analyze PME2 function should comprise assays that do not depend on functional living cells.

Since PME2 is translocated to the nucleus it is possible that it directly targets the host DNA by mimicking DNA regulatory elements, such as transcription factors or repressors. Some plant pathogen effectors bind host DNA and thus modulate gene expression [46,47]. An example of these effectors are TAL effectors of the plant pathogen *Xanthomonas*. TAL effectors bind promoter elements and regulate plant host expression to the benefit of the pathogen [48–51]. The effector AvrBs3 of *Xanthomonas* translocates to the nucleus where it acts as a transcription factor and affects the size of mesophyll cells [52]. Bioinformatic prediction and sequence comparison did not indicate that PME2 has similarity with currently known transcription factors or other gene expression regulating factors in plants.

Both P. mali strains from which the two different PME2 variants were derived cause infection and typical disease symptoms in *Malus* × *domestica*. Thus, the effect of PME2ST on cell integrity seems to be dispensable for infection and symptom development but might affect strain virulence. However, a direct comparison between the two strains regarding their virulence is missing. It might also be possible that another effector of P. mali strain AT (unknown at the time of this research), mimics and thus complements the function that PME2AT is lacking.

Some P. mali strains strongly differ regarding their virulence potential in *Malus* × *domestica* and several studies addressed the genetic identification of virulence factors or certain genetic determinants that account for these differences [17,18,53–55]. Since phytoplasma cannot be genetically manipulated, determining the importance of an effector during infection often involves tortuous experimental paths. In this study we provide the first characterization of the P. mali effector PME2 and its effect on cells of potential plant hosts. We report an interesting difference between two variants of PME2 that occur in Italy and Germany, claiming that further full genomic sequence analysis is required to better understand how P. mali manipulates its host on the molecular level.

#### **4. Materials and Methods**

#### *4.1. Verification of* Pme2 *Expression in* Malus × Domestica *and* C. Picta

For the verification of *pme2* expression by PCR in infected apple root and leaf samples RNA was extracted from the plant tissue as described in [13]. Extracted RNA was subjected to DNase treatment using DNAfree Turbo reagent (Ambion, Austin, TX, USA) and cDNA synthesis was performed using the SuperScript™ VILO™ cDNA Synthesis Kit (Invitrogen, Waltham, MA, USA). The generated cDNA was diluted 1:200 in nuclease free water and cDNA integrity was checked in all samples by performing a control PCR targeting the house-keeping gene transcript putative tip41-like family (transcript identifier: Mdo.1349) using the primers 5'-ACATGCCGGAGATGGTGTTTGG-3' (forward) and 5'-ACTTCCAGAGTACGGCGTTGTG-3' (reverse). Contamination with genomic DNA was checked by performing a PCR with primers amplifying a fragment within the non-coding region trnL of chloroplast DNA using the primers B49317 and A49855 [56]. No DNA contamination was detected in any of the cDNA samples, and the amplification of the putative tip41-like transcript fragment was positive, thus confirming the integrity of the generated cDNA. PCR reactions to verify *pme2* expression were set up in a total reaction volume of 10 μL, using 2 μL of diluted cDNA (1:200) as template, 0.05 μL GoTaq® DNA Polymerase (Promega, Madison, WI, USA), 2 μL of 5X Green GoTaq® Reaction Buffer (Promega, Madison, WI, USA), 0.2 μL dNTP-mix (40 mM), 1 μL of forward primer ATP00136\_forw\_EcoRI (10 μM, 5'-CCCCCCGAATTCATGTTTCAATTTAAAAAAAATTTA-3'), and 1 μL of reverse primer ATP00136\_rev\_SalI (10 μM, 5'-CCCCCCGTCGACATTATTACTGTTGAGGTTTAA-3'). Cycling conditions were applied as follows: 95 ◦C for 5 min followed by 40 cycles of 95 ◦C for 1 min, 44.9 ◦C for 1 min, 72 ◦C for 1 min, and a final elongation step at 72 ◦C for 5 min. PCR products were visualized on 1% agarose gel. Additionally, *pme2* expression level was detected by qPCR based on SYBR-Green chemistry using the primer pair ATP00136\_GW\_fwd (5'-CACCATGACGAAAAATGATCCAACAAA-3')/ATP00136\_nostopp\_rev (5'-CTGTTGAGGTTTAAAACAT-3') in a total reaction volume of 20 μL using 4.0 μL of diluted cDNA (1:200) as a template together with 10.0 μL 2× SYBR FAST qPCR Kit Master Mix (Kapa Biosystems/αmann-La Roche, Basel, Switzerland), 1.0 μL of each primer (10 μM), and 4.0 μL of nuclease free water. qPCR conditions were as follows: an initial denaturation step at 95 ◦C for 20 s followed by 34 cycles of 95 ◦C for 3 s and 60 ◦C for 30 s and a melting curve ramp from 65 to 95 ◦C, at increments of 0.5 ◦C every 5 s (CFX384 Touch Real-Time PCR Detection System; BioRad, Hercules, CA, USA). Data analysis was performed using the CFX ManagerTM software (BioRad, Hercules, CA, USA).

To control whether *pme2* is expressed in infected individuals of the insect vector *C. picta*, RNA of six potentially infected and two uninfected F1 individuals was extracted with the ZR Tissue & Insect RNA MicroPrepTM kit (ZymoResearch, Irvine, CA, USA) according to the manufacturer's instructions. Extracted RNA was subjected to DNase treatment using DNAfree Turbo reagent (Ambion, Austin, TX, USA) and RNA integrity was controlled with an RNA ScreenTape on a TapeStation 2200 (both Agilent, Santa Clara, CA, USA). cDNA was synthesized with the iScriptTM cDNA Synthesis Kit (BioRad, Hercules, CA, USA). Together with the cDNA synthesis a control was performed lacking the reverse transcriptase (-RT). Here, 2 μL of diluted cDNA (1:200) were used as template in a total qPCR reaction volume of 10 μL, together with 5 μL 2× SYBR FAST qPCR Kit Master Mix (Kapa Biosystems/Hoffmann-La Roche, Basel, Switzerland), 2 μL of nuclease free water, and 0.5 μL of forward and reverse primer (10 μM). The primer combination qPSY-WG-F and qPSY-WG-R, targeting the species-specific *wingless* gene [57], was used to determine cDNA integrity. P. mali infection was detected in three of the six individuals with primer pair rpAP15f-mod and rpAP15r3, targeting the ribosomal protein gene *rpl22* [58]. *Pme2* expression was checked with primer pair ATP00136\_GW\_fwd and ATP00136\_nostopp\_rev using the same qPCR conditions as described for the qPCR detection in *Malus* × *domestica* leaf samples.

#### *4.2. Amplification, Subcloning, and Sequencing of atp\_00136*

DNA was purified from leaves from P. mali infected *Malus x domestica* cv Golden Delicious trees (10 trees from Burggraviato and 10 trees from Val Venosta) using the DNeasy Plant Mini kit (Qiagen, Hilden, Germany) following the manufacturer's instructions. DNA was diluted 1:10 in water and 2 μL template were used in a total PCR reaction volume of 50 μL as follows: *atp\_00136* was amplified using 0.02 U/μL Phusion High-Fidelity DNA Polymerase (Thermo Fisher Scientific, Waltham, MA, USA) using HF-buffer supplied by the manufacturer, 400 μM dNTPs, and 0.5 μM of each primer (forward: 5'-CCCCCCGAATTCATGTTTCAATTTAAAAAAAATTTA-3'; reverse: 5'-CCCCCCGTCGACATTATTACTGTTGAGGTTTAA-3'). DNA was denatured at 98 ◦C for 30 s followed by 30 cycles of denaturation for 10 s at 98 ◦C, amplification for 30 s at 49.3 ◦C, and elongation at 72 ◦C for 30 s. The PCR was finalized by a terminal elongation step at 72 ◦C for 5 min. The PCR product was purified using the Illustra GFX PCR DNA and Gel Band Purification Kit (GE Healthcare, Chicago, IL, USA) and 1 μg of purified PCR product was digested with 4 U EcoRI and SalI following the manufacturer's instructions (Thermo Fisher Scientific, Waltham, MA, USA), ligated into equally digested pUC19 using T4-Ligase (Thermo Fisher Scientific, Waltham, MA, USA) and transformed into MegaX DH10B™ T1R cells (Life Technologies, Carlsbad, CA, USA). At least five clones from each tree were sequenced with pUC19 specific primers (GATC Biotech, Constance, Germany) and analyzed to see different variants of the gene indicating a mixed infection.

#### *4.3. Subcloning of* Pme2 *into GreenGate Expression Vectors*

The genes *pme2*ST and *pme2*AT were subcloned into the GreenGate-entry module pGGC000 [59] using the primer pair ATP00136pP\_CBsaI\_fw (5'-AACAGGTCTCAGGCTCCATGACGAAAAATGATCCAACAAA-3') and ATP00136pP\_DBsaI\_rv (5'-AACAGGTCTCACTGACTGTTGAGGTTTAAAACAT-3'). Using different components from the GreenGate-kit plant, transformation constructs coding for *pme2AT-linker-GFP* or *pme2AT-linker-mCherry* and *pme2ST-linker-GFP* or *pme2ST-linker-mCherry*, driven by the *35S* promoter and flanked at the 3'-end by the *RBCS* terminator, including kanamycin as the plant resistance marker, were designed. The following modules were assembled by GreenGate reaction in a total volume of 15 μL: 150 ng pGGA004 (*35S*), 150 ng pGGB003 (B-dummy), 150 ng pGGC000-*pme2AT* or pGGC00-*pme2ST*, 150 ng pGGD001 (*linker-GFP*) or pGGD003 (*linker-mCherry*), 150 ng pGGE001 (*RBCS*), 150 ng pGGF007 (*pNOS*:*KanR*:*tNOS*), and 100 ng pGGZ001 (empty destination vector). Subsequently, 1.5 μL 10× CutSmart Buffer (New England Biolab, Ipswich, MA, USA), 1.5 μL ATP (10 mM), 1.0 μL T4 DNA Ligase (5 u/μL) (Thermo Fisher Scientific, Waltham, MA, USA), and 1.0 μL BsaI-HF®v2 (20,000 u/mL) (New England Biolab, Ipswich, MA, USA) were added to the module mixture, and 30 cycles of 2 min at 37 ◦C and 2 min at 16 ◦C each, followed by 50 ◦C for 5 min and 80 ◦C for 5 min were performed. Subsequently, 5 μL of the reaction mixture were used for heat-shock transformation of *ccdB*-sensitive One Shot® TOP10 chemically competent *Escherichia coli* (Invitrogen, Carlsbad, CA, USA). For the assembly of positive controls, the modules pGGC012 (*GFP-NLS*) or pGGC014 (*GFP*) or pGGC015 (*mCherry*) were used instead of the above mentioned pGGC000 modules. The correct assembly of the plant transformation constructs was confirmed by sequencing. Plasmid-DNA for protoplast transformation was obtained as described elsewhere [60], using the NucleoSnap® Plasmid Midi preparation kit (Macherey-Nagel, Düren, Germany) and PEG precipitation.

#### *4.4. Protoplast Isolation and Transformation*

Protoplasts of *N. benthamiana* and *N. occidentalis* were isolated from four- to five-week-old plants, cultivated under long photoperiod conditions (16 h/8 h, 24 ◦C/22 ◦C, 70% rH) and transformed as described in [60] using 10 μg plasmid-DNA per 20,000 protoplasts. After 18 h, at least 100 protoplasts of each transformation were checked for the occurrence of GFP or mCherry-fluorescence using a confocal laser scanning microscope (LSM800, Zeiss, Oberkochen, Germany) with an excitation wavelength of 488 nm for GFP and 561 nm for mCherry. The detection wavelength of GFP was set between 410 nm and 575 nm and of mCherry between 575 nm and 650 nm. Autofluorescence of chlorophyll was detected between 650 nm and 700 nm. After 20 h the number of intact protoplasts/mL was determined by counting in a Fuchs-Rosenthal chamber. Protoplast transformation and viability determination was repeated independently four times.

Only experiments in which at least 20% of the protoplasts in the control setup were viable after transformation were considered for further evaluation. Significant outliers were removed from the data set using the GraphPad QuickCalcs Outlier calculator online tool (https://www.graphpad.com/ quickcalcs/Grubbs1.cfm; status of information 16th September 2019). Greisser Greenhouse correction on raw data and one-way-ANOVA with a Tukey Posttest were performed to analyze statistical differences between groups (GraphPad Prism 7.01., GraphPad Software, San Diego, CA, USA). To allow a better visual comparison, data were normalized to each respective control, which was set to 1.

Additionally, protoplast viability was visualized by propidium iodide (PI) and counted by fluorescein diacetate (FDA) staining in three independent repetitions. For the first, 20 μL of protoplasts transformed with GFP tagged expression vectors were mixed with 20 μL of PI solution (10 μg/mL PI in 0.65 M mannitol). FDA staining was done according to [61] using 20 μL of protoplasts transformed with either mCherry tagged PME2AT expression vectors or a vector expressing only mCherry and 20 μL of FDA solution (0.1 mg/mL FDA in 0.65 M mannitol). Fluorescence of PI, mCherry, and GFP was recorded using a LSM800 confocal laser scanning microscope (Zeiss, Oberkochen, Germany) with excitation and detection wavelengths for GFP and mCherry as described above and for PI excitation at 561 nm and detection between 560 nm and 640 nm.

#### *4.5. Nicotiana spp. Leaf Infiltration*

For subcellular localization of PME2, the two GreenGate expression vectors, as well as GFP and GFP-NLS expression vectors as positive controls, were subcloned by electroporation into *Agrobacterium tumefaciens* strain EHA105. As an additional control, we subcloned a GreenGate expression vector expressing the SAP11-like P. mali effector protein ATP\_00189 [13] with an *N*-terminal fused GFP tag into *A. tumefaciens* strain EHA105. The transgenic *A. tumefaciens* clones were cultured for 2 days at 28 ◦C in liquid selective LB medium. Subsequently, 0.5 OD/mL were resuspended in infiltration medium (10 mM MgCl2, 10 mM MES, 200 μM acetosyringone, pH 5.7) and infiltrated with a blunt syringe into leaves from four- to five-week-old *N. occidentalis* and *N. benthamiana*. Fluorescence was recorded after 48 h and 72 h using the confocal laser scanning microscope (LSM800, Zeiss, Oberkochen, Germany) with excitation for GFP at 488 nm and detection between 410 nm and 546 nm and excitation for mCherry at 561 nm and detection between 562 and 624 nm.

#### *4.6. Expression in Yeast*

For a potential Y2H, *pme2AT* and *pme2ST* were subcloned into bait-vector pLexA-N as described in [13,62] with primer pair ATP00136\_forw\_EcoRI/ATP00136\_rev\_SalI. The bait-plasmids pLexA-N-*pme2ST* and pLexA-N-*pme2AT* were transformed into *S. cerevisiae* strain NMY51. Growth aberrations of yeast colonies on selective SD-trp plates were observed and recorded by photographing.

For calcofluor white staining, yeast cells were grown overnight in SD-trp liquid media. Subsequently, 2 mL of the overnight culture were centrifuged, supernatant removed, and the cells resuspended in clear phosphate-buffered saline (PBS) buffer. Then, 10 μL of a 5 mM calcofluor white solution (Biotium, Fremont, California) were added to the cell suspension and incubated for 20 min at room temperature. The yeast cell wall was visualized by a confocal laser scanning microscope (LSM800, Zeiss, Oberkochen, Germany) with excitation at 405 nm and detection wavelength between 400 nm and 560 nm.

#### **5. Conclusions**

In this study we identified and characterized the novel P. mali effector protein PME2. This effector contains an NLS and an NES sequence and translocates to the nucleus of *N. benthamiana* mesophyll protoplasts. Two naturally occurring genetic variants of PME2, namely PME2ST and PME2AT, differ regarding their ability to induce cellular modifications in yeast and plant cells. When overexpressed, the variant PME2ST affects yeast growth and reduces the viability of *Nicotiana* spp. mesophyll protoplasts. These findings indicate that PME2 might play a role for P. mali virulence in plants. Despite the similarities between both PME2 variants, this effect was not observed in yeast or

protoplasts expressing PME2AT. The results of our study show for the first time that a phytoplasmal effector causes detrimental effects when overexpressed in protoplasts.

**Supplementary Materials:** Supplementary materials can be found at http://www.mdpi.com/1422-0067/20/18/ 4613/s1.

**Author Contributions:** Conceptualization, K.J. and K.S.; methodology, K.J., C.M., K.S., H.S., and B.H.; validation, C.M. and K.J.; formal analysis, C.M., K.J., and C.K.; investigation, C.M. and K.J.; resources, K.J., T.L., and K.S.; data curation, C.M. and C.K.; writing—original draft preparation, C.M. and K.J.; writing—review and editing, C.M., H.S., B.H., C.K., K.S., T.L., and K.J.; visualization, C.M. and K.J.; supervision, K.J., T.L., and K.S.; project administration, K.J., T.L., and K.S.; funding acquisition, K.J., K.S., and T.L.

**Funding:** This research was funded by the APPL2.0 and APPLIII project within the Framework agreement in the field of invasive species in fruit growing and major pathologies (PROT. VZL\_BZ 09.05.2018 0002552) and was co-funded by the Autonomous Province of Bozen/Bolzano, Italy and the South Tyrolean Apple Consortium.

**Acknowledgments:** We would like to thank Andreas Putti and Laura Russo of the Food Microbiology Laboratory at the Laimburg Research Centre for helpful and interesting discussions about yeast biology. Furthermore, we would like to thank Mirko Moser from the Fondazione Edmund Mach for discussing phytoplasma sequence data, Vicky Oberkofler for lab assistance, and Amy Kadison for her English proofreading.

**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/).

## *Article* **Proteomics Analysis of SsNsd1-Mediated Compound Appressoria Formation in** *Sclerotinia sclerotiorum*

**Jingtao Li 1,2,†, Xianghui Zhang 1,†, Le Li 1, Jinliang Liu 1, Yanhua Zhang 1,\* and Hongyu Pan 1,\***


Received: 10 September 2018; Accepted: 24 September 2018; Published: 27 September 2018

**Abstract:** *Sclerotinia sclerotiorum* (Lib.) de Bary is a devastating necrotrophic fungal pathogen attacking a broad range of agricultural crops. In this study, although the transcript accumulation of SsNsd1, a GATA-type IVb transcription factor, was much lower during the vegetative hyphae stage, its mutants completely abolished the development of compound appressoria. To further elucidate how SsNsd1 influenced the appressorium formation, we conducted proteomics-based analysis of the wild-type and Δ*SsNsd1* mutant by two-dimensional electrophoresis (2-DE). A total number of 43 differentially expressed proteins (≥3-fold change) were observed. Of them, 77% were downregulated, whereas 14% were upregulated. Four protein spots fully disappeared in the mutants. Further, we evaluated these protein sequences by mass spectrometry analysis of the peptide mass and obtained functionally annotated 40 proteins, among which only 17 proteins (38%) were identified to have known functions including energy production, metabolism, protein fate, stress response, cellular organization, and cell growth and division. However, the remaining 23 proteins (56%) were characterized as hypothetical proteins among which four proteins (17%) were predicted to contain the signal peptides. In conclusion, the differentially expressed proteins identified in this study shed light on the Δ*SsNsd1* mutant-mediated appressorium deficiency and can be used in future investigations to better understand the signaling mechanisms of SsNsd1 in *S. sclerotiorum*.

**Keywords:** *Sclerotinia sclerotiorum*; SsNsd1; compound appressorium; two-dimensional electrophoresis; proteomics analysis; differential expression proteins

#### **1. Introduction**

*Sclerotinia sclerotiorum* (Lib.) de Bary is a destructive and hard-to-control plant necrotrophic fungal pathogen on a broad range of agricultural crops [1,2]. Developmentally, vegetative hyphae gathered together forming hardened, multicellular sclerotia enclosed by a melanized rind layer, which plays an important role in the development and pathogenesis of *S. sclerotiorum* [3,4]. Under suitable environmental conditions, sclerotia germinate to form vegetative hyphae or apothecia, and the latter release numerous ascospores that initiate new disease cycles [5]. Mycelia from sclerotia or ascospores can directly infect the plant tissues by forming compound appressoria (also known as infection cushions) from modified hyphae [2,4] or enter the plant tissue through open stomata by secreting oxalic acid [6]. Therefore, a better understanding of the developmental mechanism of appressorium is also critical to the control of this important plant disease.

The formation of compound appressoria in *S. sclerotiorum* has been reported to require a contact stimulus [7]. Prior to penetration, the tips of hyphae become swollen and extensively branched, and then form modified, multicellular, and melanin-rich compound appressoria [8,9]. The tip of compound appressorium could penetrate the host epidermis and form vesicles of bulbous [8]. Some events are consistent with this development process, such as the production and accumulation of oxalic acid (OA), cell wall-degrading enzymes (CWDE), and effector proteins, which contribute to *S. sclerotiorum* pathogenesis in myriad ways [9–14]. However, despite these important findings, the detailed molecular mechanism underpinning the development and formation of compound appressoria in *S. sclerotiorum* is still largely unclear.

In the past years, many genetic factors have already been characterized to be essential for appressoria development in *S. sclerotiorum*. The disruption of the oxalic acid biosynthesis gene (*Ssoah1*) promotes the compound appressorium development; however, the disruption of Sspks13 only eliminates the pigmentation of the compound appressorium without attenuating its infection and pathogenicity potential [2]. Significant accumulation of the oxalate decarboxylase (OxDC) gene *Ss-odc2* occurs during the compound appressorium development, and Δ*Ss-odc2* mutants were found to have less effective compound appressorium differentiation [11]. In addition, the secretory proteins Ss-Rhs1 and Ss-Caf1 were highly expressed during the hyphal infection process, whereas the silenced strains had decreased appressoria formation [9,15]. Furthermore, *Ss-ggt1* (γ-glutamyl transpeptidase gene), *sac1* (cAMP pathway adenylatecyclase gene), and *rgb1* (type-2A phosphoprotein phosphatase (PP2A) B regulatory subunit gene) have also been identified to be associated with the development of compound appressoria [16–18]. Recently, the type IV GATA zinc finger transcription factor SsNsd1, orthologous to the *Aspergillus nidulans* NsdD (never in sexual development) proteins and *Botrytis cinerea* BcLTF1 [19], was reported to regulate asexual–sexual development and appressoria formation [4]. Its knockout mutants were defective in the transition from hyphae to compound appressorium formation, resulting in a loss of infection-dependent pathogenicity on healthy hosts [4]. However, the signal pathway by which the SsNsd1 regulates the development and pathogenicity remains to be further elucidated.

Life sciences have been deeply influenced by the "omics" technologies in last decade, including genomics, transcriptomics, proteomics, and metabolomics, aiming at a global perspective on biological systems [20]. Proteomics strategies, such as the two-dimensional gel electrophoresis (2-DE) approaches, have been confirmed as efficient, rapid, and powerful means to identify proteins (or genes) followed by mass spectrometry, and matrix-assisted laser desorption/ionization (MALDI) [21]. Large-scale analyses of proteins by 2-DE have been conducted in a number of organisms, such as animals [22], plants [23], yeast [24], and fungi [25,26], which contributes considerably to our understanding of gene functions in the postgenomic era. However, the development and application of such methods in the filamentous plant-pathogenic fungus *S. sclerotiorum* have not yet been reported.

Modern agriculture faces a huge challenge in the prevention from the diseases caused by *S. sclerotiorum*. The transcription factor SsNsd1 was characterized to be essential for appressoria development in our previous study [4]. The possibility to control plant diseases by suppressing the compound appressorium formation would eliminate initial infections. Here, we used the SsNsd1 knockout mutant (Δ*SsNsd1*) to confirm the loss-of-function nature in compound appressorium development and conducted proteomics analysis by 2-DE. Using comparative proteomics analysis of the Δ*SsNsd1* mutant and the wild-type *S. sclerotiorum*, we attempted to identify the Δ*SsNsd1-*mediated differentially expressed proteins combined with peptide mass spectrometry analysis, which would contribute significantly to the SsNsd1-mediated compound appressorium formation.

#### **2. Results**

#### *2.1. Phylogenetic Analysis of SsNsd1 and Other GATA-Type Proteins*

In this study, similar proteins of SsNsd1 and other GATA-type proteins of *S. sclerotiorum* were searched from *Botrytis cinerea*, *Fusarium oxysporum*, *Magnaporthe oryzae*, and *Aspergillus oryzae* by BLAST. The homolog of SsNsd1 was obtained only from the *B. cinerea*. However, all other GATA-type proteins had homologs in *B. cinerea*, *F. oxysporum*, *M. oryzae*, and *A. oryzae*. Phylogenetic analysis of the putative amino acid sequence of these GATA-type proteins showed their genetic relationship in different fungi (Figure 1A). Coincidently, the GATA-type proteins of *S. sclerotiorum* were closely related to *B. cinerea*. Besides, two forms of the type IV zinc finger motif (IVa and IVb) are also depicted in Figure 1A based on the residue loops. Most of the clades contained all five sequences from *S. sclerotiorum*, *B. cinerea*, *F. oxysporum*, *M. oryzae*, and *A. oryzae*. However, the SS1G\_03775 and its homologs were separated as shown by the pink color clade. One branch was also separated from the clade of the SS1G\_03252 and its homologs, which might be due to the poor similarity in different fungi.

**Figure 1.** Phylogenetic analysis and transcription expression of SsNsd1 and other proteins containing GATA-type DNA domains in *S. sclerotiorum*. (**A**) Phylogenetic analysis of the amino acid sequences of SsNsd1 (SS1G\_10366) and other GATA-type proteins in pathogenic fungi (*S. sclerotiorum*, *B. cinerea*, *F. oxysporum*, *M. oryzae*, and *A. oryzae*). A phylogenetic tree was generated by MAGE using the neighbor-joining method. The nine GATA-type proteins were separated by using a different branch color. (**B**) Hierarchical cluster of GATA-type genes and two histone genes in transcript abundance from three developments stages (hyphae, sclerotia, and apothecia) of *S. sclerotiorum*. Each gene is represented by a single row of colored boxes, and a single column indicates different development stages. The gene transcription abundance was evaluated by the fragments per kilobase of exon per million mapped fragments (FPKM) value. (**C**) The transcription level of *SsNsd1*, *Histone H2A*, and *Histone H3* genes at the hyphae developmental stage. The expression level of *SsNsd1* gene was significantly different from those of the histone genes (*n* = 3; \*\* *p* < 0.01).

#### *2.2. Transcript Accumulation of SsNsd1 and Other GATA-Type Proteins*

Digital gene expression (DGE) analysis based on FPKM values was performed using the transcriptomes during three the developmental stages of *S. sclerotiorum* (Figure 1B). The transcript accumulation in the GATA-type proteins was lower than those of the histone genes (*histone H3* and *histone H2A*). Only the SS1G\_03252 protein showed a little higher transcript accumulation in the hyphae development stage, whereas most of the other GATA-type proteins displayed low contents with FPKM values ranging from 1.3 to 64 in all three development stages. Obviously, the SS1G\_08523 and SS1G\_09784 proteins were extremely low abundant with FPKM values under 10 in all three development stages.

As one of the GATA-type proteins, the varied expression patterns of SsNsd1 were previously examined and compared across developmental stages [4]. In this study, to further determine the regulation mechanism of the Ssnsd1 expression, we obtained its expression profile in the vegetative hyphae development stage (Figure 1C). The transcript accumulation of the *SsNsd1* gene was significantly lower than those of the histone genes (*histone H2A* or *histone H3*) during the vegetative mycelial growth prior to the compound appressorium formation. *SsNsd1* was expressed at the lowest level (FPKM value of 39.6), however, *Histone H2A* was expressed higher (FPKM value of 171.8). Moreover, *Histone H3* (FPKM value of 3337.3) displayed the highest expression level, indicating its predominant role, usually as a housekeeping gene. Overall, the transcript accumulation of SsNsd1 and other GATA-type proteins were exceedingly low in the development stages of *S. sclerotiorum*.

#### *2.3.* Δ*SsNsd1 Mutant Suppressed Compound Appressorium Formation*

Although SsNsd1 exhibited a low expression level during the hyphae development, it still played a crucial role in appressorium development (Figure 2). Phenotypically, the Δ*SsNsd1* mutant had inhibited normal production of pigmented compound appressoria from the vegetative hyphae, as established by paraffin film assays (Figure 2A). In determining whether Δ*SsNsd1* affected the compound appressorium formation or only the pigmentation, normal compound appressorium and penetration (invasive mycelium) were observed microscopically only on onion epidermal strips inoculated with the wild type (WT) strain (Figure 2B), but not on the Δ*SsNsd1* mutant strain. The WT strain could colonize onion cells, but no invasive mycelium was observed from Δ*SsNsd1* mutants (Figure 2C). Thus, SsNsd1 abolished the compound appressoria formation from the modified hyphae, resulting in the penetration-dependent loss of pathogenicity.

**Figure 2.** The defective compound appressoria of Δ*SsNsd1* mutant strain led to a loss of penetration into unwounded onion tissue. (**A**) Pigmented compound appressoria of wild type (WT) were observed on parafilm. Pictures were taken four days after transfer (DAT) of 5-mm-diameter mycelial plugs to parafilm. (**B**) Penetration assays with the WT on onion epidermal strips. Invasion mycelium (penetration) of WT strain on onion epidermal strips was observed by light microscopy two days after inoculation (DAI). (**C**) Penetration assays with the Δ*SsNsd1* strain on onion epidermal strips at 2 DAI. No invasion mycelium was observed on onion epidermal strips inoculated by Δ*SsNsd1* strain. (**D**) Penetration assays with the WT and the Δ*SsNsd1* mutant on onion epidermal strips at 4 DAI.

#### *2.4. Diagram of the Identification of Differential Proteins*

In this study on *S. sclerotiorum*, we applied the 2-DE technology. A diagram illustrating the approach of the determination of the proteomics changes is presented in Figure 3. In this diagram, the comprehensive protocol is described by individual steps of the application of this technique, i.e., sample preparation and solubilization, isoelectric focusing (IEF) in IPG strips, running SDS-PAGE gels, image analysis, differential spots identification and mass spectrometry (MS) analysis, and bioinformatic prediction.

**Figure 3.** Schematic diagram illustrating the process of identifying the changes of proteomics between the Δ*SsNsd1* mutant and wild type (WT) during the compound appressorium formation. Generally, tissue samples with different phenotypes were subjected to protein extraction and SDS-PAGE test. Then the proteomics profiles were analyzed by two-dimensional gel electrophoresis (2-DE) to obtain the differential expression spots, which were further identified by mass spectrometry (MS) of the peptides and bioinformatics analyses (2-DE, two-dimensional gel electrophoresis; IEF, isoelectric focusing; SDS-PAGE, sodium dodecyl sulfate polyacrylamide gel electrophoresis).

#### *2.5. Protein Extraction from Enriched Compound Appressoria*

The cellophane induced the formation of abundant compound appressoria from the modified hyphae, which were observed macroscopically when the WT was cultured on cellophane (Figure 4A). No pigmentation was observed due to the lack of compound appressoria when Δ*SsNsd1* were cultured on cellophane (Figure 4A). Thus, this was an effective method to provide enriched compound appressoria tissue with simple sample collection, which was ideally suited for protein extraction.

**Figure 4.** Protein extraction and SDS-PAGE test of accumulated compound appressoria tissue induced on cellophane. (**A**) Induction assay for compound appressoria development on potato dextrose agar (PDA) overlaid with cellophane at 3 DAI. (**B**) Quantification of protein concentration, which was enriched by three different extraction methods: lysate, trichloroacetic acid (TCA)/acetone precipitation, and polyethylene glycol PEG method (*n* = 3; \*\* *p* < 0.01). (**C**) Quality detection of proteomics by SDS-PAGE to optimize the method for protein extraction.

To optimize the best protein extraction method for 2-DE separation, we compared the concentration and quality of the extracted proteins using the three different methods. The highest concentration of protein was extracted by the use of the lysate as compared with those extracted by trichloroacetic acid (TCA)/acetone or polyethylene glycol (PEG) precipitation (Figure 4B); however, the quality of this protein, established by the SDS-PAGE test, was extremely poor (Figure 4C). In contrast, the protein extracted by further TCA/acetone or PEG precipitation exhibited clearer bands, but that extracted by TCA/acetone precipitation was at a higher concentration; the best result was obtained using the SDS-PAGE test (Figure 4C). Furthermore, the 2-DE clearly showed protein spots aggregation and transparent background (Figure 5). Thus, the protein extraction with further TCA-acetone precipitation could lay a foundation for a novel approach in comparative proteomics analysis by 2-DE technology in *S. sclerotiorum*.

**Figure 5.** Silver nitrate-stained 2-DE gel image of the WT and Δ*SsNsd1* strains. The proteins were extracted from the accumulated compound appressoria tissue induced by cellophane. The numbers and arrows correspond to the identified differential expression proteins (≥3-fold change) for further MS analysis of the peptides.

#### *2.6. Identification of the Differential Protein Spots by 2-DE Analysis*

The proteins extracted from the WT and Δ*SsNsd1* culture on cellophane were separated by 2-DE (Figure 5). In this figure, 2-DE displayed well-visible protein spot aggregation and transparent background, while few nonspecific bands or foreign matters were observed. Using three biological replicates for both the wild-type and mutant strains, the clear 2-DE images were compared and analyzed by ImageMaster™ 2D Platinum 6.0 software. More than 2660 protein spots were detected reproducibly on each 2-DE gel image for the WT and Δ*SsNsd1* mutant, within the pH range of 4 to 10 and with relative molecular masses of 8 to 80 kDa. However, only 43 protein spots exhibited changes in the differential abundance (more than three-fold) between the WT and the Δ*SsNsd1* mutant, which are marked with arrows and numbers in Figure 5. Among these selected differential protein spots, 33 protein spots were downregulated, six proteins spots were upregulated, and four proteins spots disappeared in the Δ*SsNsd1* strain compared to the wild type (Figure 6A). Subsequently, all differential expression proteins were excised and subjected to MALDI-TOF analysis.

**Figure 6.** Analysis and evaluation of differentially expressed proteins. (**A**) Differential expression analysis of the identified protein spots in the Δ*SsNsd1* mutant compared to the WT strain. (**B**) Functional categories of differential expression proteins after MS identification and functional annotation. The percentage corresponds to the proportion of the annotated proteins in the classification. (**C**) The population distribution of the protein size evaluated by the amino acid count of the predicted proteins. (**D**) The signal peptide was predicted by running amino acid sequences of predicted hypothetical proteins on the SignalP server.

#### *2.7. Prediction and Characteristics of the Identified Proteins*

After prediction and functional annotation of these excised proteins, 40 proteins were identified. Of them, 17 proteins (38%) were predicted with the known functions (Table 1), and other 23 homologous to unnamed or predicted proteins were collectively designated as "hypothetical proteins", accounting for 56% (Table 2). However, three proteins (spots 7, 13, and 23) were not identified by MALDI for unknown reason, which were designated as "uncharacterized" (6%). The predicted known functional categories (38%) were further sorted into six functional categories, including energy production (18%), metabolisms (7%), protein fate (5%), stress responses (4%), cellular organization (2%), and cell growth and division (2%) (Figure 6B).

In addition to providing functional categories, the global view of the protein sizes was also evaluated by the numbers of the amino acids (Figure 6C). Based on the function annotation, they were divided into hypothetical and annotated proteins. Then, the distribution of their protein size was similar to that of 262 and 247 amino acids, respectively (Figure 6C). Furthermore, we predicted the signal peptide among these hypothetical proteins to obtain additional insights into the putative functions of the hypothetical proteins (Figure 6D, Table 2). Four proteins (17.4%) were predicted to contain the N-terminal signal peptide, which indicated they might be potential secretory proteins during the compound appressorium formation.


**Table 1.** List of the differential expression proteins identified as function known proteins*.*

a: The number of identified protein spots was showed on the two-dimensional gel electrophoresis (2-DE) image; The identified function known proteins were shown in this table; b: Protein name searched by locus tag in NCBI result; c: Locus tag number in NCBI annotation; d: Accession number in NCBI; e: Mascot score (threshold score > 50); <sup>f</sup> : Peptide count; g: Percent sequence coverage (%); ↓: expression level of proteins was down-regulated in the mutant; ↑: expression level of protein was up-regulated in the mutant; -: protein spot was disappeared in the mutant.

**Table 2.** List of the differential expression proteins identified as hypothetical proteins.


a: The number of identified protein spots was showed on the 2-DE gel image; The hypothetical proteins were shown in this table; b: Protein name searched by locus tag in NCBI result; c: Locus tag number in NCBI annotation; d: Accession number in NCBI; e: Mascot score (threshold score > 50); <sup>f</sup> : Peptide count; g: Signal peptide predicted (Yes or No); ↓: expression level of protein was down-regulated in the mutant; ↑: expression level of protein was up-regulated in the mutant; -: protein spot was disappeared in the mutant.

#### *2.8. Functional Analysis of Annotated Proteins*

The 2-DE approach provided a powerful proteomic screening tool to identify the initial candidate differentially expressed proteins. We evaluated these predicted proteins after functional annotation and identified 17 proteins to be functionally known, representing 10 nonredundant unique proteins (Table 1). The fact that identical proteins were available in different protein spots might have been due to the protein modification or other unclear reasons as analyzed in the discussion section.

Here, we provide information concerning the analysis of the predicted functionally known proteins. Nucleoside diphosphate kinase (NDPK) (spots 1, 9, and 17) usually possesses kinase activity exerted by direct response to the G-protein signaling or indirect catalytic GDP–GTP exchange activity, which also plays a major role in the synthesis of nucleoside triphosphates [27–29]. The eukaryotic translation initiation factor 5A-1 (eIF5A-1; spot 18) is involved in the protein fate pathway [30], which activates the 60 s subunits combination, assists in the conformational changes of the 80 s subunits, and participates in the intracellular part; proteins associate with ribosomes cyclically during the elongation phase of the protein synthesis [31]. The elongation factor 1-β (spot 26) plays a central role in the elongation step in eukaryotic protein biosynthesis [32]. The 60 s ribosomal protein L23 (Spot 4) is usually involved in cell growth; its expression was decreased in the mutant (Table 1). Ribosomal protein is involved in regulating gene transcription, translation, and regulation of cell proliferation, differentiation, apoptosis, etc. [33].

The nuclear transport factor 2 (spot 3) mediates the nucleus introduction of GDP-bound RAN (ras-related nuclear) from the cytoplasm, which is of great significance in the cargo receptor-mediated nucleocytoplasmic transport [34]. The GTP-binding nuclear protein (spot 30) is also known as the GDP-bound RAN, which is involved in the nucleocytoplasmic transport processes, nuclear envelope formation, and mitotic spindle formation [35].

The predicted ubiquitin-conjugating (UBC) enzyme E2 (spot 8) belongs to the ubiquitin pathway enzymes, which are involved in protein degradation in eukaryotic cells [36]. The SCF (Skp1/Cul1/F-box) complex submit Skp1 (spot 32) is involved in the assembly of protein complex and joins in ubiquitin depending on the protein catabolism process [37]. Peptidyl-prolyl cis-trans isomerase (spots 12, 25, 16, and 39) was downregulated, whereas spot 29 was upregulated (Table 1). Peptidyl-prolyl *cis*-*trans* isomerase regulates the mitosis-related protease in the cell cycle by protein phosphorylation of the substrate proteins [38] or through other mechanisms such as the ubiquitin-mediated proteasomal degradation [39].

Citrate synthase (spot 35) is localized in the mitochondrial matrix and catalyzes the condensation reaction from acetyl coenzyme A (CoA) and oxaloacetate to form the six-carbon citrate [40]. Oxalic acid biogenesis is realized through the hydrolysis of oxaloacetate, which is a key pathogenicity factor accumulated during the compound appressorium development [10].

The functional analysis of the predicted differential proteins was mainly based on the evidence in mammalian, plant, or yeast cells. However, the proteins displayed their important role on the cell proliferation, differentiation, protein synthesis and degradation, protein transport and modification, etc., which might determine they function as a complex regulatory network during the compound appressorium formation in *S. sclerotiorum*.

#### **3. Discussion**

Novel strategies for prevention and control of the devastating plant pathogenic fungus *S. sclerotiorum* have been intensively investigated [5]. However, still no effective method has been discovered to control the diseases caused by this pathogen. Compound appressoria are formed unless penetration occurs directly via stomata, which could be one of the key targets for disease control. The GATA-family transcription factors are involved in several essential aspects of the life cycle of *M. oryzae*, especially in the regulation of appressorium development and sporulation [41]. In *S. sclerotiorum*, the GATA-type transcription factors SsSFH1 and SsNsd1 were recently reported to be involved in the development of compound appressoria [4,42]. Here, phylogenetic analysis was

performed, and transcription accumulation of all the predicted GATA-type proteins was detected. Importantly, even the transcription accumulation of *SsNsd1* in the vegetative hyphae was significantly higher than that during the sclerotium and apothecium developmental stages [4]. The transcript accumulation remained at a much lower level even during vegetative mycelial growth, compared to that of the histone genes (Figure 1C). The *SsNsd1* gene knockout strain was defective in the development of appressorium, and no penetration was observed into unwounded onion epidermal cells (Figure 2), which confirmed the findings of a previous study [4]. In general, NsdD or its orthologous gene is involved in providing a regulatory balance between asexual and sexual development in ascomycete fungi (e.g., the development of perithecia, fruiting body, conidia, or sclerotia) [43–45]. In addition, NsdD is also involved in pathogenicity. The Δ*bclft1* mutants of *B. cinerea* exhibited only a postpenetration virulence defect without causing significant defects in the compound appressorium development; however, the *S. sclerotiorum* Δ*Ssnsd1* mutant was essentially reversed as established earlier [4]. Due to the particularly different infection defects between *B. cinerea* and *S. sclerotiorum*, in the present study, we focused on the compound appressorium deficiency phenotype and the biological role of SsNsd1, aiming to find the key target of appressorium formation-related genes that would enable the control of this pathogen.

The proteomics analysis as an evaluation of the final level of gene expression started out with techniques based on 2-DE and extended its reach by the use of MS-based techniques that have been increasingly employed in recent years [20]. Although alternative technologies, such as multidimensional protein identification technology (MudPIT), or arrays, have already emerged, thus far, there is no technology that matches 2-DE in its capability to realize routine parallel expression profiling of large mixtures [46]. Furthermore, 2-DE combined with the identification by MS is currently the major approach utilized in most of the undergoing proteome projects to develop a global understanding of the living cell [46]. Compared to the quantitative analysis based on MALDI-TOF techniques, LC–MS is currently in an early stage considering limitations, such as the availability of software, algorithms, etc. [20]. MALDI-TOF is already widely used in fungal proteome research, such as that in yeast [24], *A. fumigatus* [25], and *Cryomyces antarcticus* [26]. Therefore, we applied the 2-DE technology and MALDI-TOF mass spectrometry in this project to investigate the plant pathogenic fungus *S. sclerotiorum* using proteomics analysis for identification of differentially expressed proteins during the compound appressorium formation (Figure 3). Using suitable equipment and experienced laboratory personnel, this system approach can quickly perform the identification of functional proteins.

The 2-DE technology combined with IPGs has already conquered most limitations of carrier ampholyte-based 2-DE with in the respect of reproducibility, handling, resolution, and separation [47]. The efforts to develop the 2-DE technology further have been concentrated on improved solubilization/separation of hydrophobic proteins, show of low abundance, and more reliable quantitation by fluorescent dye technologies in recent years [46]. Despite the obvious advantages of the 2-DE technology, high quality proteins samples are always the bottleneck and precondition to the 2-DE project approach. As the 2-DE was firstly applied in *S. sclerotiorum*, we provided an optimal method for sample preparation after comparing three different pathways for protein extraction (Figure 4). By SDS-PAGE test, the protein extracted by further TCA/acetone precipitation and PEG methods has a better quality and could be used for 2-DE analysis. Both TCA/acetone and PEG methods are useful for minimizing protein degradation and removing interfering compounds, such as salt or polyphenols [48]. However, the amount of protein extracted by the PEG method is difficult to meet the requirements of 2-DE as established based on our results and those of a previous study [49]. In a comprehensive analysis, the TCA/acetone method displayed its advantages and was found to be the best method for protein extraction, which was consistent with the findings of a previous examination on another fungus, *A. fumigatus* [50]. Furthermore, in the 2-DE analysis, we achieved the separation of as many protein spots as needed on the gel, which was a prerequisite for the computerized analysis (Figure 5). In addition, a clear peptide mass spectrum was finally obtained by extraction of the

different expressed protein spots, which laid the foundation of further research on SsNsd1-mediated differentially expressed proteins.

The development of the compound appressoria involves several distinct stages [9] and is tightly regulated by numerous genetic factors. In the present study, a total of 43 differentially expressed proteins were identified with significantly differential expression changes (≥3-fold) by computer analysis (Figures 5 and 6). Most of them were downregulated, which indicated that the SsNsd1 transcript factor might positively regulate them. SsNsd1 might exert a reverse role in the signal pathway of the upregulated protein spots. By MS analysis of peptides and functional annotation, these functionally known proteins were predicted to be involved into energy production, metabolism, protein fate, stress response, cellular organization, and cell growth and division. However, attention had to be paid to the hypothetical proteins as they contained the signal peptide. The secretion and accumulation of effector proteins are usually coincident with the appressorium formation process, which contributes to *S. sclerotiorum* pathogenesis [9,15]. Therefore, these newly identified four proteins might have the effector protein role during the compound appressoria formation, but this notion needs to be further studied (Table 2). Overall, the differentially expressed proteins were finally obtained from the Δ*SsNsd1* mutant, which might play an important role during the compound appressorium formation. Furthermore, losing the capacity to produce compound appressorium could also lead to defective sclerotium development, which is a key factor in the disease cycle of *S. sclerotiorum*, such as the mutation of *Ss-ggt1* (a γ-glutamyltranspeptidase gene) [16], *sac1* (a cAMP pathway adenylatecyclase gene) [17], and *rgb1* (a type-2A phosphoprotein phosphatase (PP2A) B regulatory subunit gene) [18]. Therefore, these identified differential expression proteins were important gene resources involved in the development of *S. sclerotiorum*, which might be associated with the formation of both compound appressoria and sclerotia.

Last but not the least, a section on how to evaluate the 2-DE and the identified differential proteins is included here. After 2-DE, each protein could be theoretically resolved at a unique isoelectric point/molecular size coordinate [51]. Although hundreds of protein spots were also separated on the gel, omission of partial differential proteins can always occur due to unfavorable experimental factors, such as incomplete precipitation and/or dissolution of proteins [48], loss of sample during gel entry, inefficiency transfer of the protein from the first to the second dimension, loss of protein during staining [20], and truly absent spots from the samples [52]. In addition, some spots from 2-DE might result multiple protein identification, however, only the first identified protein with best protein score and most peptide counts was accepted for further study. Moreover, attention had also to be paid to the identical proteins (listed in Table 1), such as the nucleoside diphosphate kinase, peptidyl-prolyl cis-trans isomerase, and the 60 s ribosomal protein. Post-translational protein modifications affect the isoelectric point and, therefore, the focusing behavior of the protein in the first dimension [23], which could lead to the presence of identical proteins in different locations (i.e., spot 25 and spot 29; spot 4 and spot 28) (Table 1 and Figure 5). Post-translational modifications by fatty acid acylation, glycosylation, methylation, acetylation, or phosphorylation largely modulate the activity of most eukaryote proteins [53,54]. For example, certain signaling pathways were found to consist of series of phosphorylation and dephosphorylation events, which defined directionality and allowed different levels of feedback regulation [55]. In addition, the incomplete and insensitive separation can also lead to the appearance of identical proteins in different locations. Besides, attention is to be paid to the silver staining, as Coomassie brilliant blue was used in most of the 2-DE staining [56]. The silver staining has a low dynamic range, which has been criticized for the quantitative analyses of spots. However, the silver staining has very high sensitivity, which allows for a detection of very low protein amounts [20]; the improved and advanced image processing method could become feasible in better quantification of protein spots [57]. Besides, to obtain more data of the exact protein abundance, only the differential proteins with ≥3-fold changes were accepted for further study in this research. Therefore, the 2-DE technology of gradual optimization, further analysis of protein modifications,

and other proteomic analysis methods are still needed to employ, which would present formidable challenges but generate indispensable insight into biological functions in *S. sclerotiorum*.

#### **4. Materials and Methods**

#### *4.1. Fungal Strains and Culture Conditions*

The wild-type (WT) *S. sclerotiorum* isolate 1980 and its derived mutant Δ*SsNsd1* were used in this study based on our previous reports [4]. The strains were routinely grown on potato dextrose agar (PDA) at normal room conditions. The WT and Δ*SsNsd1* stocks were stocked as dry sclerotia or as desiccated mycelia-colonized filter paper at −20 ◦C.

#### *4.2. Phylogenetic Analysis of SsNsd1 and Other GATA-Type Proteins*

From the genome of *S. sclerotiorum*, nine proteins are predicted to containing GATA-type DNA domains: SS1G\_1036, SS1G\_11953, SS1G\_12238, SS1G\_03252, SS1G\_08523, SS1G\_05040, SS1G\_09784, SS1G\_03775, and SS1G\_01151 [4,42]. The BLASTX program at NCBI (http://www.ncbi.nlm.nih. gov/) was employed to search for the homologs of the sequence of the SsNsd1 (SS1G\_1036) and other GATA-type proteins from pathogenic fungi (*Botrytis cinerea*, *Fusarium oxysporum*, *Magnaporthe oryzae*, and *Aspergillus oryzae*). The phylogenetic tree was generated using neighbor-joining method in MEGA5 [58]. Prediction of protein zinc finger domain was performed to classify the different categories. Most fungal GATA factors contain a single zinc finger domain, which can be divided into two distinct categories: the 17-residue loops (CX2CX17CX2C; zinc finger type IVa) and the 18-residue loops (CX2CX18CX2C; zinc finger type IVb) [59,60].

#### *4.3. Digital Gene Expression of SsNsd1 and Other GATA-Type Proteins*

The transcription accumulation of SsNsd1 from three different developmental stages (hyphae, sclerotia, and apothecia) has been characterized by qRT-PCR [4]. To further study the gene expression patterns of SsNsd1 prior to compound appressorium development, we profiled gene expression patterns in *S. sclerotiorum* from hyphae, sclerotia, and apothecia through RNA-seq approach. The transcription level of SsNsd1 and other GATA-type proteins were quantified using the fragments per kilobase of exon per million mapped fragments (FPKM) method [61]. The FPKM means were generated from three technical replicate samples. The FPKM values of *histone H3* (SS1G\_09608.3) and *histone H2A* (SS1G\_02052) were used as endogenous control for quantitative comparison with *SsNsd1* and other GATA-type protein genes. Hierarchical clustering was performed using the MeV program [62]. Clustering was based on the average of FPKM values.

#### *4.4. Compound Appressorium Assays*

Deficiency of compound appressoria was observed macroscopically by placing the freshly colonized mycelial agar plugs (5-mm diameter) on parafilm due to the presence of pigmented appressoria surrounding the agar plug [2]. Yellow onions were purchased from a local grocery store, and the onion epidermal strips were used for inoculation with a colonized PDA agar plug for observing the penetration of compound appressorium using light microscopy. For the enrichment of compound appressoria, colonized agar was cultured on PDA medium covered with cellophane and grown at a temperature of 22 to 25 ◦C as reported previously [2].

#### *4.5. Two-Dimensional Gel Electrophoresis (2-DE) Strategy in This Study*

The 2-DE combines isoelectric focusing separation based on isoelectric point of protein in the first dimension and sodium dodecyl sulfate (SDS) polyacrylamide gel electrophoresis (SDS-PAGE) to separate the complex mixtures of proteins according to the molecular size in the second dimension [23]. The Combined with identification by mass spectrometry (MS), 2-DE is currently the major method used in the majority of the ongoing proteome projects [46]. Besides, 2-DE gels are easy to handle and could

be produced in a highly parallelized way. Furthermore, the corresponding software has also reached a level that allows for routine bioinformatic analysis. Meanwhile, in the presence of a suitable laboratory equipment and experienced personnel, analysis of samples can be theoretically completed through this approach with investments of time and efforts that are much smaller than those needed for the laboratory work [20]. Thus, these mature and coherent techniques are well-suited for comparative proteomics analysis of Δ*SsNsd1* mutant-mediated appressoria deficiency in *S. sclerotiorum*. In detail, tissues derived from two different strains, WT and Δ*SsNsd1*, were harvested, and the proteome was enriched and solubilized. The protein mixture was then applied to a "first dimension" gel strip that separated the proteins based on isoelectric focusing (IEF) in IPG strips. Next, the IPG strip was subjected to equilibration and running of multiple "second dimension" SDS–PAGE gels, where proteins were finally separated by their molecular charge and molecular size. After staining, the visual protein spots were recorded and analyzed by sophisticated software. Then, the differential protein spots were excised for MS analysis. Finally, the differential proteins were subjected to functional annotation and prediction analysis.

#### *4.6. Protein Extraction and Optimization*

The enriched compound appressoria on the cellophane were harvested and frozen in liquid nitrogen, and then ground to a fine powder for protein extraction. To optimize the method for protein extraction from *S. sclerotiorum* and separate the protein by 2-DE, during the compound appressoria production, the protein was extracted by direct lysate and further TCA/acetone or PEG precipitation methods. For the lysate method [63], 0.1 g powdered sample was suspended in 300 μL of precooled lysate (7 M urea, 2 M thiourea, 2% (*w*/*v*) 3-[(3-Cholamidopropyl)dimethylammonio]propanesulfonate (CHAPS), 20 mM Tris-HCl, and 20 mM dithiothreitol), then vortexed for 30 s and centrifuged at 15,000× *g* rpm for 10 min at 4 ◦C. The supernatant was further centrifuged at 15,000× *g* rpm for 30 min. Approximately 250 μL of the supernatant was taken as the crude protein solution and stored in a freezer at −80 ◦C. For further precipitation by trichloroacetic acid (TCA)/acetone [64], 80 μL of crude protein solution was suspended in 5 mL of cold TCA/acetone (10% TCA, 0.07% β-Mercaptoethanol (ME) in acetone) and mixed for 30 s, then precipitated at −20 ◦C for 2 h. Further, the precipitate was centrifuged at 15,000× *g* rpm for 15 min at 4 ◦C. Then, the supernatant was discarded, and the pellet washed three times with 800 μL of cold acetone and finally centrifuged at 15,000× *g* rpm for 15 min at 4 ◦C. Next, the pellet was desiccated using a vacuum dryer and stored at −80 ◦C. For the PEG precipitation [65], 80 μL of crude protein solution was suspended using 40% PEG solution for 30 s and then precipitated at −20 ◦C for 2 h. Further, the pellet was washed with 800 μL of cold acetone and desiccated, as described above. The final protein was resuspended in 80 μL of rehydration buffer (7 M urea, 2 M thiourea, 4% (*w*/*v*) CHAPS, 0.002% (*w*/*v*) bromophenol blue, 2% (*v*/*v*) Bio-Lyte, and 20 mM dithiothreitol). The concentration of the dissolved protein solution was determined according to the method of Bradford [66]. Then, the results were compared by routine SDS-PAGE test to optimize the method for protein extraction, and the extracted protein with the best quality was further separated by 2-DE for image analysis.

#### *4.7. The 2-DE Assay and Image Analysis*

Comparative 2-DE was performed using a EttanTM IPGphor apparatus (GE Healthcare, Pittsburgh, PA, USA) for isoelectric focusing (IEF; first dimension), and an EttanTM DALTsix (GE Healthcare) for the second dimension according to the manufacturer's instructions [48] and the protocol described in a previous report [56]. For IEF, 300 μg hydrated protein was loaded on immobilized pH gradient (IPG) strips (18 cm length, pH 3.0–10.0). The following IEF steps were used: 500 V for 1 h, 1000 V for 1 h, 4000 V for 1 h, 8000 V for 1 h with a linear gradient, holding at 8000 V until a total of at least 40,000 Vh was reached, then holding at 500 V for 20 h. After the strips were balanced twice, the second dimension was performed with 12% SDS-PAGE gel, and the total proteins were stained before further analysis. The stained gel image was captured with an Image Master LabScan (GE Healthcare), and the

images were analyzed using the ImageMasterTM 2D Platinum 6.0 software (GE Healthcare) for spot detection, gel matching, and statistical analysis of the spots [56]. The image analysis remains one of the most labor-intensive parts of the 2-DE approach. In brief, triplicate images from three independent gels for the WT and mutant were obtained, while the normalization of the gels was carried out by the sum of the spot densities on each gel to compare the spots. The abundance of the individual protein spots was determined as vol.%. To identify the protein spots, the silver staining method was applied to the prepared gels. Silver nitrate was added to the solution before use, and then it was quickly admixed into the dyeing tray. The tray was then covered with an opaque cloth to reduce the decomposition of the silver nitrate utilized.

After visualization by staining and gel image analysis, protein spots with at least 3-fold spot volume ratio change (*p* < 0.05) were excised and subjected to mass spectrometry sequence analysis combined with database comparison. Statistical comparisons were conducted using the one-way ANOVA with the Tukey's HSD test.

#### *4.8. MALDI-TOF Analysis and Prediction of Differential Proteins*

In this study, the spots showing statistically significant changes were cut out from the preparative gels and washed twice with ultrapure water. Then, the protein spots were destained with 50% acetonitrile (ACN) in 25 mM NH4HCO3. After removing the destaining buffer, the gel pieces were lyophilized and rehydrated in 30 μL of 50 mM NH4HCO3 containing 50 ng trypsin (Promega, Madison, WI, USA) at 37 ◦C overnight. The supernatant of the resulting peptides was washed with 0.1% trifluoroacetic acid (TFA) in 67% ACN. Extracts were pooled and lyophilized for MS analysis. The MS spectra were obtained using an ABI 4800 MALDI-TOF/MS-MS Proteomics Analyzer (Applied Biosystems, Foster City, CA, USA) as previously described [56]. The positive ion reflector (2 kV accelerating voltage) with 1000 laser shots per spectrum and automatic data acquisition modes were used for data collection, and the TOF spectra were collected over the mass range within 800–4000 Da with a signal-to-noise ratio minimum set to 10 and a local noise window width of *m/z* 250. A maximum of 10 precursors per spot with a minimum signal/noise ratio of 50 were selected for data-dependent MS/MS analysis.

Then, the resultant MS and MS/MS spectra data were analyzed with the GPS Explorer software (Version 2.0, Applied Biosystems). The database search was performed on the Mascot server (http: //www.matrixscience.com) by searching the NCBInr (nonredundant protein sequence) database of *S. sclerotiorum* (http://www.ncbi.nlm.nih.gov/) to identify the proteins. The other important parameters were set as follows trypsin cleavage, two missed cleavage allowed, carbamidomethylation set as fixed modification, oxidation of methionine allowed as variable modification, monoisotopic precursor mass, precursor ion mass tolerance set to ±100 ppm, and fragment mass tolerance set to ±0.5 Da. The protein was correctly identified if a sufficient number of peptides were matched with a high score to a protein in the database. Only the significant hits with a protein score of 100% and the highest peptide counts were recorded and analyzed. Then the predicted protein sequences that matched the sequences in the NCBInr database were further analyzed for functional category denomination [56,62]. Prediction of the signal peptide was done using the online SignalP 4.1 Server (http://www.cbs.dtu.dk/services/SignalP/).

#### *4.9. Data Analysis*

All graphs were exported by the GraphPad Prism 6 software (La Jolla, CA, USA). Statistical comparisons were done using the one-way ANOVA with the Tukey's HSD (Honestly Significant Difference) test in the PASW Statistics 18 (SPSS Inc., Chicago, IL, USA).

#### **5. Conclusions**

In this study, we combined TCA/acetone precipitation for protein extraction, 2-DE, and peptide mass analysis to develop a fast and simple method for studying the proteomics changes of Δ*SsNsd1* mutant during compound appressorium formation. In our approach the results from 2-DE gel analysis are put into a larger context by combining spot data with functional annotations to explore the SsNsd1-mediated compound appressoria formation. Visualizing results, such as differential expression, functional categories, and predicted effector proteins makes it possible to gain new insights from the data accumulated by the "omics" technologies. Thus, this system approach can be effectively used to identify important candidate proteins in response to the SsNsd1-mediated appressorium formation, but it will require subsequent, more detailed studies to determine the precise role of the differentially expressed proteins.

**Author Contributions:** J.L. (Jingtao Li), X.Z., and L.L. performed the experiments and analyzed the data. J.L. (Jingtao Li) wrote the manuscript. H.P. and Y.Z. conceived the study and provided funding. J.L. (Jinliang Liu) provided technical support. All authors commented on the manuscript.

**Funding:** This study was financially supported by the National Natural Science Foundation of China (31772108, 31471730, and 31271991) and the National Research and Development Program of China (2018YFD0201005).

**Acknowledgments:** We gratefully acknowledge Jeffrey A. Rollins (University of Florida) for donating the wild-type strain 1980 and the mutants. We also thank Gang Yu for manuscript proofreading and technical assistance.

**Conflicts of Interest:** The authors declare no conflict 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/).

## **Interspecies Outer Membrane Vesicles (OMVs) Modulate the Sensitivity of Pathogenic Bacteria and Pathogenic Yeasts to Cationic Peptides and Serum Complement**

**Justyna Roszkowiak 1, Paweł Jajor 2, Grzegorz Guła 1, Jerzy Gubernator 3, Andrzej Zak ˙ 4, Zuzanna Drulis-Kawa <sup>1</sup> and Daria Augustyniak 1,\***


Received: 30 September 2019; Accepted: 5 November 2019; Published: 8 November 2019

**Abstract:** The virulence of bacterial outer membrane vesicles (OMVs) contributes to innate microbial defense. Limited data report their role in interspecies reactions. There are no data about the relevance of OMVs in bacterial-yeast communication. We hypothesized that model *Moraxella catarrhalis* OMVs may orchestrate the susceptibility of pathogenic bacteria and yeasts to cationic peptides (polymyxin B) and serum complement. Using growth kinetic curve and time-kill assay we found that OMVs protect *Candida albicans* against polymyxin B-dependent fungicidal action in combination with fluconazole. We showed that OMVs preserve the virulent filamentous phenotype of yeasts in the presence of both antifungal drugs. We demonstrated that bacteria including *Haemophilus influenza*, *Acinetobacter baumannii,* and *Pseudomonas aeruginosa* coincubated with OMVs are protected against membrane targeting agents. The high susceptibility of OMV-associated bacteria to polymyxin B excluded the direct way of protection, suggesting rather the fusion mechanisms. High-performance liquid chromatography-ultraviolet spectroscopy (HPLC-UV) and zeta-potential measurement revealed a high sequestration capacity (up to 95%) of OMVs against model cationic peptide accompanied by an increase in surface electrical charge. We presented the first experimental evidence that bacterial OMVs by sequestering of cationic peptides may protect pathogenic yeast against combined action of antifungal drugs. Our findings identify OMVs as important inter-kingdom players.

**Keywords:** outer membrane vesicles (OMVs); *Candida albicans*; antimicrobial peptides; complement; interspecies interactions; inter-kingdom protection; fungicidal activity; fluconazole; hyphae

#### **1. Introduction**

In the environment, different microbial populations, including bacteria and fungi, co-exist. Mixed microbial populations are also a common feature in many diseases. Hence, understanding which factors may be potentially important players in interspecies dynamics of growth and to what degree these dynamics are mediated by the host is very important. Among these factors are outer membrane vesicles (OMVs) of Gram-negative bacteria classified recently as secretion system type

zero [1]. These proteoliposomal nanoparticles released from the cell play an important role in bacterial physiology and pathogenesis. During pathogenesis, they enhance cellular adherence, cause biofilm formation, and induce apoptosis or inflammation [2–4]. Furthermore, DNA-containing OMVs, through horizontal gene transfer, are important in interspecies communication as well as in host–pathogen interactions [5]. OMVs contribute also to innate bacterial defense through β-lactamase content or trapping and degrading the membrane-active peptides [6–10]. Accordingly, the involvement of OMVs in resistance to antibiotics with various modes of action is growing [11]. Both offensive and defensive functions of OMVs have been reported. In line with this, OMVs can deliver bactericidal toxins or enzymes to other bacteria. Lytic activities have been documented for OMVs derived from *Pseudomonas aeruginosa* or *Myxococcus xanthus* [12–14]. On the other hand, for example, *M. catarrhalis* OMVs carrying β-lactamases confer protection to *M. catarrhalis* and *Streptococcus pneumoniae* against β-lactam antibiotics [15]. Likewise, OMVs from β-lactam-resistant *E.coli* can protect β-lactam-susceptible *E.coli* and fully rescue them from β-lactam antibiotic-induced growth inhibition [16]. Several earlier studies have shown that OMVs are involved in the trapping of antimicrobials, thus providing protection among bacteria, essentially in intraspecies systems [6,8,17]. Nevertheless, limited data are available on the relevance of this phenomenon in various interspecies populations. Among mixed populations, the degree of OMV-dependent protection of individual partners can be completely different. One of the factors that may influence this is the physicochemical nature of the vesicle itself. It includes both the intrinsic antimicrobial binding capacity, which is the result of inherited or acquired resistance [18] but also the ability of vesicle, based on physicochemical properties, to interact with other target cells. There are several factors responsible for the latter phenomenon including electric charge and hydrophobicity of a cellular surface [19,20]. The metabolic activity of OMV-producing bacteria also seems to have a significant impact on this issue. For example, for *P. aeruginosa* it was reported that variation in physicochemical properties was dependent on growth phase (exponential versus stationary), influencing, therefore, the cell association activity [19]. All aforementioned characteristics of OMVs indicate the significant degree of selectivity in OMV–cell interactions. Therefore, considering variety of microbial populations, the question of which bacteria are and which are not protected against membrane-active agents is still very ambiguous. It is also of great importance to ask whether this protection can go beyond the protection against only bacteria, affecting pathogens from other kingdoms such as fungi. Hence, it is of interest to investigate the protective potency of OMVs in the light of their specificity to react with a target cell. In the present study, we used our well-characterized OMVs from *Moraxella catarrhalis* Mc6 [4,21] as model vesicles and characterized their protective potential against model membrane-active agents in various interspecies combinations. First, we examined the protective activity of OMVs in bacterial intra- and interspecies systems and mode of observed protection, showing by HPLC-UV, zeta potential measurement, and o-nitrophenyl-β-D-galactopyranoside ONPG-based permeabilization assay, highly effective sequestration. Next, we examined the protective potential of OMVs in the bacterial-yeast inter-kingdom system. As far as we know, our study is the first report that OMVs can protect *Candida albicans* against antifungals drugs. Finally, we investigated whether OMVs-dependent protection is a signature of only free vesicles or those associated with the cell.

#### **2. Results**

#### *2.1. Mc6 OMVs Characteristics*

As shown in transmission electron microscopy (TEM) image, the diameters of outer membrane vesicles from *M. catarrhalis* 6 had 30–200 nm (Figure 1a). The results of measurements of OMV particle size/zeta potential are shown in Figure 1b. The protein and lipooligosaccharide (LOS) components of OMVs are shown in Figure 1c,d, respectively. As we documented previously [21], the pivotal outer membrane proteins packaged in these vesicles were OmpCD, OmpE, UspA1 (ubiquitous surface protein A1, Hag/MID (*Moraxella* immunoglobulin D-binding protein), CopB, MhuA (hemoglobin-binding protein), TbpA (transferrin-binding protein A), TbpB (transferrin-binding protein B), LbpB (lactoferrin-binding protein B), OMP M35, and MipA (structural protein).

**Figure 1.** Physical characterization of outer membrane vesicles (OMVs) released from Mc6 cells: (**a**) TEM image of OMVs, vesicles are indicated by arrows (magnification, ×50,000); (**b**) the size distribution by volume and the zeta potential of vesicles, as assessed by the Zeta-sizer; each experiment was performed in triplicate; (**c**) the respresentative proteinogram of 12% SDS-PAGE electrophoresis of OMVs; the protein profiles were visualized using Coomassie staining; (**d**) the representative 10% SDS-PAGE electrophoresis of lipooligosaccharide (LOS)-OMVs visualized by silver staining.

#### *2.2. M. catarrhalis OMVs Passively Protect Cross-Pathogens against Polymyxin B-Dependent Killing*

Polymyxin B (PB) was used as a model of cationic peptide. In our study, using 4-h time-kill assay, the minimal bactericidal concentrations (MBC) of PB against 5 <sup>×</sup> 105–106 cfu/mL of prominent human pathogens, including nontypeable *Haemophilus influenzae*, *Pseudomonas aeruginosa,* and *Acinetobacter baumannii*, was in the range 0.5–5 μg/mL, and caused killing effect within 2 h (Figure 2a). When pathogenic bacteria were incubated with a bactericidal concentration of PB in combination with 20 μg/mL OMVs from Mc6, the bacteria showed active growth in contrast to the antibiotic alone, which remained at the level of control. Up to 10 times lower concentrations of OMVs had no effect or the effect was negligible. The amount of OMVs that was required to achieve complete protection against respiratory pathogens referred to 20 μg/mL. The higher concentrations of vesicles did not intensify the growth.

#### *2.3. OMVs Protect Serum-Sensitive Strains and Accelerate Growth of Serum-Resistant Strains against Complement*

To define the influence of OMVs on the survival of cross-pathogens in the presence of active serum (NHS), the 4 h complement bactericidal assays were performed (Figure 2b). Of the three studied cross-pathogens, only nontypeable *H. influenzae* appeared to be serum-sensitive. Two others, *A. baumannii* and *P. aeruginosa,* were serum-resistant. The resistant strains showed either 100% survival or only a slight decrease (within one order) after 4 h of incubation in 50% or 75% NHS. Nontypeable *H. influenzae* strains were sensitive to NHS-dependent killing with a reduction of cfu/mL in the range of six orders after 30 min of incubation (NTHi3) or 120 min of incubation (NTHi6) in the presence of 25% or 50% serum, respectively, compared to the initial inoculum. These strains co-incubated with NHS and 20 μg/mL of OMVs exhibited a time-dependent increase in survival, which was similar to growth of control incubated in the presence of heat-inactivated serum (HiNHS).

**Figure 2.** Mc6 OMVs protect bacteria from other species against polymxin B-dependent and human complement-dependent bactericidal activity: (**a**) Polymyxin B-dependent killing; (**b**) human serum (NHS) complement-dependent killing. Bacteria from log phase were incubated for 4 h in the presence of indicated concentrations of membrane-targeting agents alone or along with OMVs and plated in 0, 30, 120, and 240 min. Data are expressed as mean cfu/mL ±SD from three independent experiments performed in triplicate.

*A. baumannii* incubated only in the presence of active NHS was slightly killed by the lytic complement action within 4 h, whereas after parallel exposure to OMVs, its growth was at the level of HiNHS control. *P. aeruginosa* was highly resistant to 75% action of serum, thus co-incubation with OMVs did not change the growth dynamics (data not shown). When control bacteria were incubated in the presence of heat-inactivated NHS, neither killing nor other growth alterations were observed

for any of the tested strains, indicating that observed lysis was complement-dependent. Overall, these findings indicate that OMVs efficiently protect serum-sensitive gram-negative pathogens against complement action while accelerating the growth of moderately serum-resistant strains.

#### *2.4. Mc 6 OMVs Passively Protect Pathogenic Yeasts against Polymyxin B-Dependent Fungicidal E*ff*ect in Combination with Fluconazole*

It has been previously reported that the antifungal effect of polymyxin B combined with fluconazole can be synergistic or potentiated [22]. We therefore initially examined the sensitivity of *Candida albicans* to combined action of both drugs. Our results showed, that for ~105 cfu/mL of this yeast, the MIC50 for fluconazole (FLC) and MIC100 for polymyxin B (PB) were, respectively, 1 μg/mL and 128 μg/mL. Next, based on checkerboard assay, we assessed the growth of *C. albicans* incubated with polymyxin B alone, fluconazole alone, or their combinations at various concentrations. The results revealed that polymyxin B at concentrations much lower than MIC (1/8 MIC and 1/16 MIC) exerts a potent antifungal effect against *C. albicans* when combined with 1 μg/mL (MIC50) of FLC (Figure 3a).

**Figure 3.** OMVs protect yeast against synergistic fungicidal activity of polymyxin B in combination with fluconazole: (**a**) Synergistic activities determined by checkerboard assay; (**b**) time-kill assay; (**c**) 24 h growth-curve kinetics for *C. albicans* incubated with drug alone and drug in combinations with or without OMVs. The kinetics were measured using Varioskan™ LUX reader with measurements at 1 h intervals. The data for experiments (a,b) show means ± SD from two independent experiments carried out in triplicate; the data for experiment (c) show mean values from two independent repetitions carried out in duplicate; for better readability, SDs were not included.

Next, using time-kill assay we showed that after 24 h of incubation, the combinations of FLC1-PB8 and FLC1-PB16 (μg/mL) are fungicidal, which was expressed by one order and by two orders of decrease in viability, respectively (Figure 3b). In the presence of OMVs, this fungicidal action of both drugs was significantly abolished or weakened in comparison to action of a single drug and depending on PB concentration used. This protective effect of Mc6 OMVs against *C. albicans* was also confirmed in bacterial growth kinetics (Figure 3c). By estimating the profiles of *Candida* growth, as a result of every hour measurements carried out for 24 h at 37 ◦C, we have documented that the OMVs added to both aforementioned drug combinations abolished their inhibitory effect on *Candida* growth rate. Thus, the *Candida* growth in the presence of OMVs remained at the level for a single compound. Collectively, these results indicate that OMVs protect *C. albicans* from PB-dependent fungicidal effect in combination with fluconazole.

#### *2.5. Mc6 OMVs Passively Enhance the Virulence of Pathogenic Yeasts Facilitating the Formation of Filaments*

It was documented that azole drugs are necessary both during growth and induction step to have an effect on transition yeast-to hyphae in *C. albicans* [23]. The presence of FLC in the decreased range between 1–0.125 μg/mL both during growth and induction step caused considerable inhibition of hyphal growth (data not shown). Therefore, to determine the effects of fluconazole in combination with polymyxin B on hyphal growth, the yeasts were initially preincubacted overnight at 37 ◦C with 1/16 MIC of FLC (0.0625 μg/mL). During the 2 h induction of hyphal growth in the presence of 10% fetal bovine serum, 1/16 MIC of FLC was added together with 1/8, 1/16, and 1/32 MIC of PB in the presence of absence of 20 μg/mL of OMVs. Figure 4a,b shows that the number of hyphal cells, shown by the ratio of cells in yeast form compared to that with formed filaments after 2 h of incubation, decreases as the concentration of polymyxin B increases. Under the tested conditions, the combinations FLC0.0625-PB8 and FLC0.0625-PB16 (μg/mL) showed almost complete inhibitory activity against the formation of hyphae. FLC0.0625-PB4 had also clear inhibitory effect on yeast-to-hyphae transition in comparison to FLC alone. The presence of 20 μg/mL OMVs in all tested drug combinations caused a significant increase in the percentage of hyphae-forming cells (Figure 4b), probably as a result of neutralizing PB-dependent potentiating effect of FLC. Furthermore, the increase in yeast-to-hyphal transition was accompanied by a significant extension of filaments (Figure 4c). In summary, these results indicate that in the presence of Mc6 OMVs, the inhibiting effect of FLC and PB, at concentrations significantly lower than MICs, on the formation of filaments is abolished. It shows that in the presence of OMVs, *C. albicans* may retain the virulent filamentous phenotype in the presence of both antifungal drugs, thus increasing its virulence.

#### *2.6. OMV-Dependent Protection against PB and Complement is the Result of PB Sequestration on Free OMVs*

To study how efficiently PB is neutralized by Mc6 OMVs, various biological and biochemical methods were used. Initially, a decrease of free PB was confirmed indirectly using *E.coli* ML35p mutant with constitutive β-galactosidase expression (Figure 5a).

In the experiment, different concentrations of Mc6 OMVs were introduced into the system containing *E. coli* ML-35p mutant at a concentration of ~10<sup>6</sup> cfu/mL and PB at a concentration of 5 μg/mL. By measuring the quantitative intracellular influx and hydrolysis of ONPG (β-galactosidase substrate), as a result of membrane permeabilization, we showed that OMV-dependent protection was dose-dependent and that PB was very quickly depleted from the environment in the presence of at least 20 μg/mL OMVs leading to a decrease in peptide activity (Figure 5a).

**Figure 4.** Effect of OMVs on filamentous growth of *C. albicans*: Yeast cells were preincubated overnight with 1/16 MIC of FLC (0.0625 μg/mL) without shaking and then, after washing, were incubated at OD = 0.4 in 0.4 mL of hyphae inducing medium (10% FBS in PBS) in 24-well microtiter plates with shaking (130 rpm) for 2 h at 37 ◦C in the presence of 0.0625 μg/mL FLC, indicated concentrations of PB, and with or without 20 μg/mL of OMVs. Three independent experiments were performed. (**a**) The filamentation was monitored under inverted microscope using 40× objective (Zeiss) and images were recorded. Scale bars = 100 μm. (**b**) The number of hyphal cells versus yeast cells was determined using ImageJ software. The pool of yeast cells contains yeasts and yeasts whose germ tubes did not exceed 10 μm. The data represent the mean values calculated for ≥ 200 cells for each of the eight tested options. (**c**) Box and Whisker plots of hyphal length: Minimal and maximal values, median, quartiles Q1 and Q3. Statistical analysis was performed by one-way ANOVA (\* *p* < 0.05, \*\* *p* < 0.01).

Next, to directly confirm the role of OMVs in PB sequestration, we determined zeta potential, showing that the initially moderately negative zeta potential of intact OMVs was immediately and significantly neutralized by addition of at least 50 μg/mL of PB, in a concentration-dependent and diluent-dependent manner leading to membrane depolarization (Table 1). The presented results of alteration in Zeta potentials were similar after 30 min and 60 min incubation at 37 ◦C (data not shown). Finally, to quantitatively assess the magnitude of PB sequestration by OMVs, the residual free PB content remaining after ultrafiltration of OMV-PB complexes was measured using HPLC-UV. Figure 6 shows the chromatograms of recovered by ultrafiltration PB that was preincubated for 30 min either alone or with OMVs. The loss of PB following incubation with OMVs at 5 μg/mL and 20 μg/mL was almost 60% (*p* < 0.001) and over 96% (*p* < 0.001), respectively, in reference to standard PB (Table 2).

**Figure 5.** OMVs block the permeabilizing activity of membrane targeting agents. Change in *E. coli* ML-35p membrane permeability was assayed by a time-dependent influx of ONPG in the presence of membrane-targeting agents and different concentrations of Mc6 OMVs. Bacteria at log phase (~10<sup>6</sup> cfu/mL) were incubated on microplate at 37 ◦C for 120 min in the presence of indicated concentrations of OMVs together with (**a**) polymyxin B (PB) and (**b**) normal human serum (NHS). The absorbance was measured at indicated time points. The results are shown as mean ± SD from at least three independent experiments performed in duplicate.


**Table 1.** Alteration in Zeta potential of OMVs treated with polymyxin B.

\*-*p* < 0.001 in reference to control as determined by one-way ANOVA.

Collectively, our results demonstrate that PB is quickly and effectively depleted from the environment via OMV-dependent sequestration.

Similar experiments with *E.coli* ML35p mutant were performed with human active serum (NHS). Similar to PB, the strong permeabilizing potency of membrane attack complex (MAC), present in 10% human serum, against *E.coli* ML35p was considerably decreased or even dumped as the concentration of OMVs increased (Figure 5b). It suggests that by deposition of complement components on the OMV surface, vesicles trigger MAC formation away from target bacteria and thus protect them from MAC-mediated lysis. Accordingly, based on the quantitative formation of soluble non-proteolytic membrane attack complex (SC5b-9), we demonstrated that the Mc6 OMVs activate human serum complement in a concentration-dependent manner (Figure S1).

**Figure 6.** The HPLC spectra of polymyxin B incubated alone or along with OMVs. Chromatograms of (**a**) 50 μg/mL solution of polymyxin B, (**b**) 50 μg/mL of PB after treatment with 5 μg/mL of Mc6 OMVs, (**c**) 50 μg/mL of PB after treatment with 20 μg/mL of Mc6 OMVs. Samples were prepared as described in Materials and Method section. HPLC analyses were performed with Macherey–Nagel Nucleodur C18 Isis column. UV-detection at 215 nm. Double peak with Tr = 3.8 min corresponds mainly to polymyxin PB2 and PB3, and double peak with Tr= 7 min corresponds to polymyxin PB1 and PB1-I. The representative chromatograms are shown.


**Table 2.** Magnitude of PB sequestration on *M. catarrhalis* OMVs.

The PB content was measured by HPLC-UV method as described in material and methods. The results are expressed as mean ± SD. \* *p* < 0.001 in reference to standard PB as determined by one-way ANOVA.

#### *2.7. OMVs Associated with Bacteria do not Protect against PB*

To answer the question of whether Mc6 OMVs protect bacteria against the membrane-active agent by sequestration only indirectly, being far from the bacterial surface or through the direct shield of these cells, we determined the association capability of OMVs using flow cytometry. We used 4-times higher concentration of OMVs (80 μg/mL) to make sure that the entire cell population could be covered (Figure S2). We found that Mc6 OMVs incubated with studied strains strongly interacted within 30 min with the parental strain as well as with other strains from *Moraxella* species. After this time, more than 93% of measured events were fluorescent, indicating that FITC-labelled OMVs were associated with unlabeled bacteria. No increase in fluorescence was observed when Mc6 OMVs were incubated even up to 2 h with bacteria from other species indicating the complete lack of association (Figure 7a,b). These results indicate that OMVs released by *M. catarrhalis* associated only in intraspecies but not in interspecies tested systems, pointing to the specificity of this reaction. To verify whether the covering of *M. catarrhalis* with OMVs is another mode of protection against PB, 4 h time-kill assays (Figure 7c,d), as well as 24 h real-time bacterial growth kinetics (Figure 7e,f) were carried out on OMV-associated and not associated cells in the presence and absence of PB.

Not in the line with expectation, the results showed that for *M. catarrhalis* associated and not associated with OMVs, the lethal effect of PB occurred after 4 h (Figure 7c) and 2 h (Figure 7d), respectively. It indicates that the association does not ensure longer protection but only delays the bactericidal effect. The lack of protection was confirmed in growth kinetic experiments documenting that PB-dependent growth inhibition for OMV-associated *M. catarrhalis* and non-associated control was comparable and preserved for 24 h of incubation (Figure 7e). In the case of bacteria incubated in the presence of free OMVs and PB, the OMV-dependent protection is assured for ~7 h being on the level of control growth. Some modest and stable level of protection is maintained for 24 h (Figure 7f). Compared to the growth profiles of OMV-associated and non-associated *M. catarrhalis*, it was shown that the interaction with vesicles weakens the growth dynamics of the former. Overall, the results are evidence that OMVs associated with bacteria do not protect against PB, assuming of course that the bacterial cell is completely shielded by them. Consequently, it suggests the active fusion between OMVs and target cell membrane rather than only surface association.

**Figure 7.** Mc6 OMVs highly associated with bacteria do not protect against polymyxin B-dependent killing. (**a**) Flow cytometry analysis of fluorescein isothiocyanate (FITC)-labelled OMVs associated with intraspecies bacteria (upper panel) and not associated with interspecies bacteria (lower panel), The fluorescence intensities of OMV-associated bacteria are shown as black histograms whereas control bacteria as dotted histograms. Representative plots are shown. (**b**) Quantification of bacteria associated with FITC-labelled OMVs. Data are expressed as mean fluorescent intensity (MFI) ± SD from three independent experiments performed in duplicates. (**c**) Bactericidal activity of polymyxin B against *M. catarrhalis* after association with OMVs. (**d**) Bactericidal activity of polymyxin B against *M. catarrhalis* incubated with free OMVs. (**e**) 24 h growth-curve kinetics for *M. catarrhalis* after association with OMVs. (**f**) 24 h growth-curve kinetics for *M. catarrhalis* incubated with free OMVs. The kinetics was measured using Varioskan™ LUX reader with measurements at 30 min intervals. In experiments (c,d) results are shown as mean ± SD from at least two independent experiments. In experiments (e,f) results are shown as mean from at least three independent experiments performed at duplicate and for better readability, SDs were not included.

#### **3. Discussion**

Bacteria and yeasts employ a variety of means to protect their envelopes against harmful environmental factors. Some of these factors cause membrane permeabilization, while others can inhibit the synthesis of cellular membrane components affecting the physical properties of the membrane [24,25]. The common strategy of bacteria to avoid membrane targeting factors is to overcome the negative charges of their surface envelope [24,26]. Alternatively, bacteria can release OMVs that act as extracellular decoys for some antimicrobials.

In this study, we report that OMVs may serve as a model of passive interspecies protection of prominent pathogenic bacteria but also pathogenic yeasts. While documenting the lethal effect, we showed that free OMVs block both the bactericidal action of peptide antibiotic polymyxin B and the lytic activity of complement. By quantitative measurement of alteration in electric charge on the OMV surface treated with PB as well as free PB determination by HPLC, we showed that 20 μg/mL of OMVs that had a significant protective effect against interspecies microorganisms caused almost complete depletion of this model cationic antimicrobial agent after co-incubation, thus indicating the immediate sequestration of the peptide on free OMVs. Accordingly, in similar experiments, we showed that OMVs inhibit serum lytic activity against prone pathogens, indicating that by deposition of complement components on the OMV surface, vesicles trigger MAC (SC5b-9) formation away from target bacteria and thus protect them from MAC-mediated lysis. The intensity of OMV-dependent complement activation in vitro was correlated with the number of free vesicles in the environment. Therefore, it is possible that any fluctuation in the number of released vesicles may positively or negatively influence complement activation of this potent innate mechanism.

Next, using the PB-dependent model, we investigated whether OMVs-dependent protection is a signature of only free vesicles or those associated with the cell. Using flow cytometry, we documented that the association of OMVs with bacterial cells may be very potent and species-specific, influencing the cell sensitivity to antimicrobial compound. Our vesicles were able to associate only with representatives of *M. catarrhalis* species while they did not work in interspecies systems. By evidencing the lethal effect of cell-associated OMVs exposed to polymyxin B, we showed that OMVs do not protect against AMPs this way, suggesting a fusion between OMVs and OM rather than only shielding. It is tempting to speculate that the specificity of interaction between OMVs and recipient cell is somehow pivotal in its sensitization at least to AMPs. We are therefore in agreement with Tashiro et al. that elucidating the selectivity in OMV-cell interaction is critical for an improved understanding of the outcome of this reaction. On the other hand, it has been documented that interactions with OMVs for some cross-pathogens can be specific, less specific, or not specific at all [27]. Thus, the OMV-dependent interplay may contribute to the generation of synergistic or antagonistic interactions between pathogens, resulting in more or less harmful outcomes for the host.

Next, we address the question of whether OMV-dependent trapping of cationic antimicrobials can protect pathogens from other kingdoms such as fungi of the species *C. albicans*. Polymyxins alone are effective against *C. albicans* only at relatively high concentrations [28]. Azoles including fluconazole are common antifungal drugs [29]. They affect the integrity of fungal membranes, altering their morphology and inhibiting growth [30,31]. It has been reported that quinolone and other antibiotics may augment the anti-candidal activity of azole and polyene agents [32]. Previous research also documented that PB in lower concentrations exerts a potent antifungal effect when combined with fluconazole [22,33]. Therefore, the elimination of the antibiotic from these systems by means of OMVs should decrease the fungus susceptibility to fluconazole. To test this hypothesis, we incubated *C. albicans* in the presence of both drugs and OMVs using FLC at MIC50 (1 μg/mL) and PB in the range much lower than MIC (1/8–1/16). To our knowledge, for the first time, we have provided evidence that bacterial OMVs can inhibit the fungicidal action of certain combined antifungal agents that are effective against pathogenic yeasts. Next, we documented that sub-MIC concentrations of FLC (1/16) and PB (in the range of 1/8–1/32) applied together weakened or almost completely inhibited the yeast-to-hyphae transition. Furthermore, in the presence of OMVs, the filamentous phenotype was considerably recovered and even strengthened despite exposure to both drugs. It is a very undesirable action of OMVs, since filamentation increases the virulence of *C. albicans,* which in hyphae form is more invasive and can attach in a higher number to epithelial cells than yeast and pseudohyphae forms [34,35]. Using analogies to the role of OMVs contributing to the sequestration of PB, it is tempting to speculate that the same mechanism of action exists in this case. Collectively, both of the aforementioned activities of bacterial vesicles seem to render *C. albicans* less vulnerable to destruction. The documented inter-kingdom OMV-based mutualistic relationship between bacteria and yeasts are, to our knowledge, a novel phenomenon. Therefore, its importance for more complex in vitro and in vivo conditions, as well as pathophysiology, remains to be clarified.

The complex *Candida*–bacteria interactions are not a rare occurrence and may have an important impact on the human disease by causing e.g., the faster biofilm growth or *Candida*-dependent induction of antimicrobial resistance of *Staphylococci* [36,37]. Sometimes, the results investigating cross-kingdom polymicrobial interactions are very contradictory even for the same microbial components. It was shown that direct or indirect (by released soluble molecules) contact of *P. aeruginosa* or *A. baumannii* with *C. albicans* or other fungi may lead to the killing of yeast [38,39]. It can also decrease fungal filamentation, biofilm formation, and conidia biomass [40]. The synergistic collaboration between the two pathogens was also documented. For example, the pre-colonization with *C. albicans,* which compromises the immune system, facilitates the emergence of *A. baumannii* pneumonia [41]. The interactions between microbes are not only affected by the specific combination of microorganisms, but also by the environment such as immunological milieu. Therefore, bacterial OMVs shape the behavior of neighboring microbes and the overall outcome of their interplay for the host. Due to the limited antifungal arsenal, the synergistic effect of PB and fluconazole or human broad-spectrum AMP lactoferrin with amphotericin B or fluconazole, which increases the activity of the antifungals against *Candida* spp, could be an alternative for treatment [22,42]. Likewise, AMPs such as defensins or gramicidin have shown to be a promising alternative to the current antimycotic and antibacterial therapies [43–45]. Furthermore, AMPs display a lower propensity to develop resistance than do conventional antibiotics [45]. In light of our research, however, these promising strategies should consider the unfavorable role of OMVs as a trap for host cationic peptides in mixed bacterial and bacterial–fungal infections. Our results may also have a meaning in medical microbiology. Because AMPs are already used in topical nasal antimicrobials in the treatment of nasal or paranasal cavity infections (sinusitis, maxillary, otitis media) [46,47], it is conceivable that abundantly produced OMVs, by very efficient AMP sequestration, may decrease the pharmacokinetics of these compounds.

Another important issue of our results is the number of OMVs needed to show biological activity. In general, the information about the number of OMVs produced in vivo is still very limited. Although OMV production in the course of infection has been documented [48], the magnitude of this production was not given. During the in vitro part of this study, however, the biologically active concentration of OMVs was 5 μg/mL per 103 cfu/mL. In our study, 20 μg/mL of OMVs was protective against 106 cfu/mL, indicating that OMV amounts used in our study are rational and may resemble the infectious condition.Furthermore, there is many data on how different stress factors may induce bacterial hypervesiculation incuding temperature stress [49], oxidative stress [50], hyperosmotic stress [49], or antibiotic stress [51]. So far, the correlation between OMVs production and pathophysiology of a specific disease was proven both in animal sepsis-like inflammation models [52–54] and in a patient with fatal meningococcal septicaemia [55]. Based on these aforementioned examples, the OMVs concentrations, which seems to be clinically relevant, are in the range 5–20 μg/mL.

Overall, our results on the protective and somehow deleterious for pathogens role of OMVs, in the bacteria–bacteria and bacteria–yeast interspecies systems, underline the enormous potential of these nanostructures as accelerating factors in case of various mixed infections. Furthermore, the OMV-dependent mode of actions may serve as a model of passive resistance of gram-negative bacteria not only to antimicrobials, but also to humoral defense components, which operate to disrupt cell membrane. Likewise, for dimorphic yeasts, the ability of OMVs to sequester membrane active

compounds that augment the antifungal activity of azoles may have an important impact on *Candida* virulence. This work may serve as an important basis for further evaluation of OMVs-dependent interactions within pathogenic bacterial-fungal communities. Our results indicate that OMVs are important players in interspecies and cross-kingdom microbial interactions.

#### **4. Materials and Methods**

#### *4.1. Materials*

#### 4.1.1. Reagents

BHI (Brain Heart Infusion, OXOID, Basingstoke, UK) ); TSB (Tryptone Soya Broth, OXOID, Hampshire, England); TSA (Tryptone Soya Agar, OXOID, Hampshire, England), Bradford reagent (Protein Assay Dye Reagent Concentrate, Bio-Rad, München, Germany); β-nicotinoamide adenine dinucleotide hydrate (NAD, Sigma-Aldrich, Steinheim, Germany); fluconazole (FLC, Sigma-Aldrich, Poznan, Poland), hemin (Sigma-Aldrich, St. Louis, MO, USA); polymyxin B sulfate salt (PB, Sigma-Aldrich, Denmark); fluorescein isothiocyanate (FITC, ThermoScientific, Rockford, IL, USA); Hank's Buffer with Ca2+, Mg2<sup>+</sup> (HBSS, PAN Biotech, UK); RPMI 1640 (Lonza, Walkersville, MD, USA); *o*-nitrophenyl-β-D-galactopyranoside (ONPG, Sigma, Steinheim Germany), heat inactivated (56 ◦C, 1 h) FBS (Fetal bovine serum, Gibco Life Technologies, Grand Island, NY, USA).). HPLC chemicals: Acetonitrile (Sigma, München, Germany) for separation was HPLC far UV/gradient grade (J. T. Baker, Avantor™ Performance Material); 32 mM Na2SO4 solution for chromatographic usage was prepared with 4.5 g anhydrous sodium sulfate (POCh, Avantor™ Performance Material, Gliwice, Poland) and MiliQ (ultrapure water made with Simplicity UV Water Purification System, Merck Millipore, Saint-Quentin, France).

#### 4.1.2. Microbial Strains and Growth Condition

The following microbial strains were used: *Moraxella catarrhalis* (Mc5, Mc6, Mc8), nontypeable *Haemophilus influenzae* (NTHi3, NTHi6), *Acinetobacter baumannii* (ATCC 19606), *Pseudomonas aeruginosa* (PAO1), *Candida albicans* (Ca1), mutant of *Escherichia coli* ML-35p, a lactose permease-deficient strain with constitutive cytoplasmic β-galactosidase. All strains were from the collection of our Institute. *M. catarrhalis* strains were grown on Columbia agar plates or BHI broth. NTHi strains were grown on chocolate agar plates or in BHI broth supplemented with hemin and NAD at final concentrations of 15 μg/mL each. *Moraxella* and *Haemophilus* strains were cultivated at 37 ◦C with 5% CO2. *A. baumannii; E. coli ML-35p* and *P. aeruginosa* were routinely cultured in TSB medium at 37 ◦C. *C. albicans* was cultured in yeast extract-peptone-glucose (YPG) in 37 ◦C.

#### *4.2. Methods*

#### 4.2.1. Outer Membrane Vesicles Isolation

Outer membrane vesicles (OMVs) isolation was performed as described previously [21]. The protein concentartions of purified OMVs preparations was determined by Bradford assay) and the quality of OMVs preparation was confirmed in 12% SDS-PAGE.

#### 4.2.2. Time-Kill Assay

For testing PB or human serum (NHS) activity, the log-phase bacterial suspension (5 <sup>×</sup> <sup>10</sup>5–106 cfu/mL) was incubated with or without PB (in the range 0.5–5 <sup>μ</sup>g/mL) or NHS (in the range 25%–75%) with the presence or absence of free OMVs (2 μg/mL or 20 μg/mL) in the final volume of 200 μL 1% medium (*w*/*v*). The experiments were performed from 0 to 240 min and the 10 μL aliquots of 10-time diluted bacterial suspensions were plated in triplicate on appropriate agar plates at 0, 30, 120, and 240 min time points. The colony counts and cfu/mL were calculated the next day. The controls for

NHS contained heat-inactivated serum (HiNHS), (56 ◦C, 30 min). The bactericidal activity of PB or NHS was expressed in each time point as cfu/mL in reference to cfu/mL in time 0. Analogous experiments with bacteria associated and non-associated with OMVs (20 μg/mL or 80 μg/mL) in the presence of PB (5 μg/mL) were performed. Analogous experiments for selected combinations of antifungals were carried out for *C. albicans* except that: (i) Initial inoculum was ~2 <sup>×</sup> 105–5 <sup>×</sup> 105 cfu/mL, (ii) tested antimicrobials were PB (8 μg/mL or 16 μg/mL) and fluconazole (1 μg/mL), (iii) the medium was 0.5% YPG (*w*/*v*), and (iv) incubation lasts 24 h. All microbicidal assays were performed at least two times in triplicate.

#### 4.2.3. Growth Kinetics Assay

All growth kinetics experiments were performed on the flat-bottomed 96-well microplates (NUNC, Denmark) at 37 ◦C in volume 200 μL. Dynamic of growth was measured using Varioskan™ LUX reader with measurements at 30 min intervals for bacteria at 106 cfu/mL diluted (final concentration) in 1% BHI (*w*/*v*) and 60 min intervals for yeast at 105 cfu/mL (final concentration.) diluted in 0.5% YPG.

OMV-associated bacteria preparation: 0.5 mL of ~108 cfu/mL exponentially growing bacteria were washed by centrifugation with HBSS Ca2<sup>+</sup> Mg2<sup>+</sup> and resuspended in 100 μL OMVs (20 μg/mL or 80 μg/mL) in Eppendorf tube. Association was performed for 30 min at 37 ◦C with gentle mixing. Thereafter, all samples were washed with HBSS Ca2+, Mg<sup>2</sup> by centrifugation (8000 rpm, 10 min, 4 ◦C) to remove free OMVs particles, diluted to 2 <sup>×</sup> 106 cfu/mL with HBSS Ca2+, Mg2, and used in growth kinetics assay.

#### 4.2.4. Checkerboard Microdilution Assay

The microdilution assay was performed on flat-bottom microplate according to the CLSI (formerly NCCLS) standard [56] except that the initial inoculum for *C. albicans* was ~105 cfu/mL and cells were incubated at 37 ◦C without shaking. Synergy/growth potentiation was tested by the checkerboard method including a two-dimensional array of serial concentrations of both drugs. The fluconazole was used in concentrations 1–64 μg/mL whereas polymyxin B in concentrations 1–128 μg/mL. Wells without drugs or yeast inoculation were included as positive and negative controls, respectively. The MIC100 of polymyxin B and the MIC50 of fluconazole was defined as the lowest drug concentration that caused a decrease in absorbance of 100% and 50%, respectively, compared to control in drug-free medium.

#### 4.2.5. Induction of Filamentation

Before the induction of filamentation, *C. albicans* cells were preincubated overnight in YPG supplemented with sub-MIC concentration of FLC at 37 ◦C without shaking. To induce hyphal transition, the suspensions of *C. albicans* (OD = 0.4) were treated with 10% heat inactivated FBS in PBS for 2 h at 37 ◦C in 24-well flat-bottom microplates (NUNC) in volume 0.4 mL, in the presence of sub-MIC concentrations of FLC (1/16) and PB (range 1/8–1/32) and with shaking (130 rpm). The samples were observed under inverted microscope Zeiss Axio Vert. A1 with objective Zeiss LD A-Plan (40 × /0.55 Ph1). The images wer recorded using Industrial Digital Camera 5.1 MP 1 /2.5". The assessment of cell morphology and the length (μm) of hyphae was performed using ImageJ software.

#### 4.2.6. Membrane Permeability Assay with ONPG

To assess the polymyxin B sequestration by *M. catarrhalis* OMVs in vitro, the time-dependent decrease of permeabilizing activity of this peptide against *E. coli* ML-35p with constitutive β-galactosidase activity was measured using the ONPG-mediated β-galactosidase microplate assay as previously described [57]. The final bacterial suspension (~10<sup>6</sup> cfu/mL) in 10 mM sodium phosphate NaPB (pH 7.4) were incubated in the flat-bottom 96-well microplates (NUNC, Denmark) with 5 μg/mL of PB in the presence of OMVs at concentration range 1–20 μg/mL and with 3 mM ONPG as β-galactosidase substrate in the final volume of 150 μL. Microplates were incubated at 37 ◦C for 1.5 h and optical densities at λ = 405 nm were measured every 15 min (spectrophotometer ASYS). All the assays were performed at least 3 times in duplicates.

#### 4.2.7. Complement Activation

This method was described previously [58]. To assess complement activation by *M. catarrhalis* OMVs in vitro, the various concentrations of OMVs (1–20 μg/mL) were pretreated with active normal human serum (NHS) at a volume ratio 1:9 and the soluble terminal complement complex SC5b-9 was measured using ELISA kit according to manufacturer's instructions (Quidel Corporation, San Diego, USA). Unstimulated NHS served as negative controls.

#### 4.2.8. OMV Association Assay and Flow Cytometry

OMVs labeling: Initially, OMVs (80 μg/mL) in PBS were concentrated using 50 kDa Vivaspin centrifugal concentrators (Amicon ultra, Millipore) at 14,000 × *g* for 10 min at 4 ◦C to remove PBS. The collected OMVs were reconstituted with 500 μL of 0.05 M carbonate/bicarbonate buffer (pH 9.5) and washed by centrifugation on vivaspin as described before. The collected OMVs at concentration 80 μg/mL were labeled with 1 mg/mL FITC at carbonate/bicarbonate buffer for 30 min at 37 ◦C with gentle mixing in the dark. The remaining fluorochrome was rinsed of 3 times with a 500 μL of cold carbonate/biscarbonate buffer each time, using 50 kDa Vivaspin. The final FITC-labelled OMVs were resuspended in Hank's buffer Ca2<sup>+</sup>, Mg2<sup>+</sup> at original concentration.

OMVs association with bacteria: 1 mL of each fresh bacterial culture corresponding to OD550 = 0.2 was centrifuged and subsequently washed with 1 mL of HBSS Ca2+, Mg2<sup>+</sup> (8000<sup>×</sup> *g*, 10 min, 4 ◦C). The pellet was resuspended in 200 μL of OMV-FITC conjugate (80 μg/mL OMVs) and incubated for 30 min at 37 ◦C with gentle mixing in the dark. Afterward, samples were washed with HBSS Ca2<sup>+</sup>Mg2 by centrifugation (8000× *g*, 10 min, 4 ◦C) to remove free OMVs particles and finally resuspended in 200 μL of HBSS Ca2<sup>+</sup> Mg2.

Flow cytometric analysis: To detect bacterial cells associated with FITC-labeled OMVs flow cytometry analysis was performed using GUAVA® easyCyte flow cytometer (Millipore, Seattle, WA, USA). Before analysis, the samples were diluted 1:100 to obtain approximately 1–5 <sup>×</sup> 105 cell/mL in HBSS Ca2<sup>+</sup> Mg2<sup>+</sup>. Fluorescence intensity of bacterial cells associated with OMVs was analyzed for green fluorescence in the FL1 channel, by collecting 5000 events. Data were expressed as mean fluorescence intensity (MFI). Data analysis was performed using InCyte Merck Guava software (Millipore, Hayward, CA, USA).

#### 4.2.9. Estimation of Zeta Potential

The Zeta potential of OMVs was measured at room temperature (25 ◦C) by a Zetasizer Nano-ZS 90 (Malvern, UK). The instrument was equipped with a Helium–Neon laser (633 nm) as a source of light. The detection angle of Zetasizer at aqueous media was 173◦. Considering the influence of factors such as conductivity (salt concentration) or pH of the solution on the Zeta potential, to minimize their influence, all Zeta potential measurements were performed with 10 mM NaPB buffer (pH 7.4) or in MiliQ.

#### 4.2.10. HPLC-UV System and Method

HPLC chromatography of polymyxin B was performed as described previously [26,59]. The HPLC system consisted of a Water's 2695 Solvent Manager System with a built-in autosampler and 100 μL sample loop connected to Waters 2996 Photodiode Array Detector. Data collection and peaks integration were realized by computer with Water's Empower 3 Chromatography Data Software. The separation developed with the use of a Macherey-Nagel EC Nucleodur C18 Isis column (50 mm, 4.6 mm ID with 1.8 μm beads) with Hypersil Gold aQ 5 μm (10 mm × 4.6 mm ID) guard precolumn. The mobile phase consisted of 22% acetonitrile and 78% of 32 mM Na2SO4 in water (pH 3.2 achieved with H2SO4) in isocratic separation mode. Eluent flow 0.75 mL/min and detection realized with a UV detector at 215 nm. Column and sample temperature were respectively 30 ◦C and 5 ◦C; separation run time was 15 min. The injection volume was 50 μL. The calibration curve was in the range of 1.56 μg/mL to 100 μg/mL prepared with 10 mg/mL standard stock solution of polymyxin B sulfate salt. Each concentration was injected twice daily for precision into 3 or 2 independent samples, and between day eight samples. The linearity of the method was in a range from 6.25 to 100 μg/mL with 16% of the average coefficient of variation (CV%) for the sum of peaks, and with a coefficient of determination R2 = 0.997. Since the major constituents of polymyxin B are B1 and B2 [59], (polymyxin B sulfate certificate of analysis, Sigma-Aldrich), these two components were analyzed altogether.

To quantify the results, the relative concentration was calculated as the mean PB concentration of sample/mean PB of control × 100%. Results were expressed as mean ± SD.

Sample preparation: 200 μL of reaction mixtures containing OMVs (5 μg/mL or 20 μg/mL) and PB (50 μg/mL) or PB alone (50 μg/mL) were incubated for 30 min at 37 ◦C with gentle mixing. The samples were ultrafiltrated using 50 kDa vivaspin centrifugal concentrators (Amicon ultra, Merck Millipore, Cork, Ireland) at 14 000× *g* for 15 min at RT to remove OMVs. The filtrate was collected and stored at −20 ◦C until use. Tested and control samples were processed identically.

#### 4.2.11. LOS-OMVs Electrophoresis

The concentration of lipooligosaccharide (LOS) from Mc6 OMVs was determined based on the purpald assay [60]. Diluted samples were solubilized in Laemmli sample buffer and heated at 100 ◦C for 5 min. Proteinase K (20 mg mL<sup>−</sup>1) was added per 20 mg of OMV proteins and incubated in a heating block at 60◦C for 1 h. The presence of LOS in proteinase K-treated OMV samples was analyzed by dodecylosulfate gel electrophoresis. The 15 μL of samples were applied to the Glycine-SDS-PAGE (10%) gel path, corresponding to 5 μg LOS. Electrophoresis was carried during 2.5 h at a constant voltage of 80 V at 4 ◦C. The gel was fixed for 1 h at room temperature using a fixing solution (EtOH: Acetic Acid: MiliQ: 40:5:55); POCH, Gliwice, Poland) The fixed gel was stored overnight at 4 ◦C. MiliQ was exchanged for fresh before visualization using a modified Tsai and Frasch [61] silver staining protocol.

#### 4.2.12. TEM

The OMVs preparation for TEM was described previously [4]. OMVs were visualized by standard negative staining using a formvar copper grid (Christine Gröpl Electronenmikroskopie, Tulln, Austria) and 2% (*w*/*v*) aqueous solution of uranyl acetate. The OMVs were imaged with a TEM operating at an acceleration voltage of 150 kV (Hitachi H-800, Japan).

#### 4.2.13. Statistical Analysis

The data were expressed as the mean ± SD, and analyzed for the significant difference by one-way ANOVA or Kruskal–Wallis ANOVA rang using the Statistica (version 13.1) software (StatSoft, Krakow, Poland) Differences were considered statistically significant if *p* < 0.05.

#### **Supplementary Materials:** Supplementary materials can be found at http://www.mdpi.com/1422-0067/20/22/ 5577/s1.

**Author Contributions:** D.A., conceived the study; J.R., P.J., G.G., J.G., DA., performed experiments; D.A., J.R., P.J. and G.G. performed analysis and interpretation of data; A.Z. performed TEM images; D.A. wrote the manuscript ˙ with input from J.R.; D.A., J.R., G.G. reviewed and edited the manuscript. D.A., wrote the revised version of the manuscript; Z.D.-K. and D.A. acquired funding. This work is a part of the PhD thesis by J.R.

**Funding:** This study was supported by research grants of University of Wroclaw (1016/S/IGiM/T-20).

**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/).

## *Article* **Comparative Integrated Omics Analysis of the Hfq Regulon in** *Bordetella pertussis*

## **Ana Dienstbier 1, Fabian Amman 2,3, Daniel Štipl 1, Denisa Petráckov ˇ á <sup>1</sup> and Branislav Veˇcerek 1,\***


Received: 5 June 2019; Accepted: 19 June 2019; Published: 24 June 2019

**Abstract:** *Bordetella pertussis* is a Gram-negative strictly human pathogen of the respiratory tract and the etiological agent of whooping cough (pertussis). Previously, we have shown that RNA chaperone Hfq is required for virulence of *B. pertussis*. Furthermore, microarray analysis revealed that a large number of genes are affected by the lack of Hfq. This study represents the first attempt to characterize the Hfq regulon in bacterial pathogen using an integrative omics approach. Gene expression profiles were analyzed by RNA-seq and protein amounts in cell-associated and cell-free fractions were determined by LC-MS/MS technique. Comparative analysis of transcriptomic and proteomic data revealed solid correlation (r<sup>2</sup> = 0.4) considering the role of Hfq in post-transcriptional control of gene expression. Importantly, our study confirms and further enlightens the role of Hfq in pathogenicity of *B. pertussis* as it shows that Δ*hfq* strain displays strongly impaired secretion of substrates of Type III secretion system (T3SS) and substantially reduced resistance to serum killing. On the other hand, significantly increased production of proteins implicated in transport of important metabolites and essential nutrients observed in the mutant seems to compensate for the physiological defect introduced by the deletion of the *hfq* gene.

**Keywords:** *Bordetella pertussis*; Hfq; omics analysis; T3SS; serum resistance; solute-binding proteins

#### **1. Introduction**

*Bordetella pertussis* is a Gram-negative strictly human pathogen of the respiratory tract and the etiological agent of whooping cough (pertussis) [1]. This highly contagious disease is especially severe in infants and remains a major cause of infant mortality and morbidity worldwide, predominantly in developing countries [2]. Furthermore, pertussis incidence is currently on the rise in industrialized countries with highly vaccinated populations [3,4]. While there are several reasons for this phenomenon [5], there are two major disease-related factors contributing to recent increase in pertussis cases: short-lived immunity induced by current acellular vaccines and pathogen adaptation leading to escape from the immunity by antigenic variation [6–8]. The global reemergence of pertussis clearly suggests that we need to widen our understanding of the molecular mechanisms underlying the pathogenesis of *B. pertussis* [9,10]. In many pathogenic bacteria the RNA chaperone Hfq and small non-coding regulatory RNAs (sRNAs) emerged as critical players in posttranscriptional regulation of virulence and physiological fitness [11–13]. The Hfq protein forms ring-shaped hexamers that possess several RNA binding sites that allow for simultaneous interaction with both sRNA and mRNA molecules and stabilization of their interactions [14–16]. Besides its role in facilitation and stabilization

of RNA duplexes, Hfq can actively remodel the structure of RNAs and also increase the stability of sRNAs [14,16,17].

Recently we have shown that Hfq is required for virulence of *B. pertussis* as the Δ*hfq* mutant was affected both in its ability to efficiently multiply and persist in mouse lungs as well as in its capacity to cause a lethal infection in mouse [18]. Furthermore, our global DNA microarray-based transcriptomic profiling of the *hfq* mutant suggested that Hfq protein significantly affects expression of more than 10% annotated genes [19]. Nevertheless, despite the high sensitivity, transcriptomic profiling does not capture post-transcriptional and post-translational modifications that affect the amounts of produced proteins. On the other hand, mass spectrometry-based proteomics lacks the sensitivity to detect low abundant proteins. Therefore, integrative analysis of both transcriptomic and proteomic datasets enables a more complete understanding of studied biological processes [20,21]. First studies based on such an approach revealed that the overlap between the outcomes of transcriptomic and proteomic analyses is not extensive irrespective of the organism [22–24]. This discrepancy was attributed in part to technological limitations of applied procedures and in part to inherent biological complexity of transcription and translation processes [25,26]. Especially, factors linked to translational efficiency, such as codon usage bias, strength and accessibility of ribosome binding site, secondary structure and stability of the transcript, and post-transcriptional activity of the regulatory proteins, contribute to poor correlation between determined transcript and protein levels [20,27–29]

Hfq is a key player in post-transcriptional control of gene expression in Gram-negative bacteria and therefore, its biological activities should in principle weaken the correlation between the gene expression and protein synthesis profiles. Recently, an integrative analysis of Hfq-specific transcriptomic and proteomic profiles based on high-throughput RNA-seq and LC-MS/MS technologies was performed in soil bacterium *Pseudomonas fluorescence*. It showed that such a multiomics approach allows for dissection of discrete contributions of Hfq to gene regulation at different levels [30]. Therefore, we were interested to perform such a comparative analysis in human pathogen to elucidate the Hfq-related variations at both transcriptomic and proteomic levels per se and also to decipher how the changes in gene expression profiles translate into protein production in *B. pertussis*. Our results indicate that considering the role of Hfq in the post-transcriptional control of gene expression, the correlation between transcriptome and proteome is relatively high. Furthermore, our data corroborate and further clarify the necessity of Hfq for physiological fitness and pathogenicity of *B. pertussis*.

#### **2. Results**

#### *2.1. Identification of the Hfq Regulon by RNA-seq*

Samples of total RNA isolated from biological triplicates of *B. pertussis* Tohama I strain and its isogenic Δ*hfq* strain cultures were analyzed by RNA-seq. RNA-seq analysis yielded on average 16 million reads, which were mapped to the *B. pertussis* genome. The comparison of global expression profiles showed that biological replicates of either wt or Δ*hfq* cells are highly uniform and thereby reproducible (Figure 1A). Principal component analysis (PCA) revealed that samples from wt strain and *hfq* mutant clustered separately along principal component 1 (94%) reflecting global changes in gene expression profiles resulting from deletion of the *hfq* gene (Figure 1B).

Differential expression (DE) analysis identified 653 significantly modulated *B. pertussis* genes (|log2FC| > 1; *q* < 0.05) including 40 non-coding RNAs and 11 transfer RNA genes (Table S1). Among the DE genes, 281 genes were downregulated and 372 upregulated in the Δ*hfq* strain. Remarkably, 56 genes (8.3% of all DE genes) encoding the components of ATP-binding cassette transport system were significantly up- or downregulated in the mutant. In agreement with our previous microarray study, the expression of several genes within the type III secretion (T3SS) *bcs*/*btr* locus including *bsp22*, *bopN*, *bopB*, and *bopD* was substantially decreased in the mutant. Genes involved in iron–sulfur cluster protein biogenesis (*iscU*, *iscA, iscS*) were also highly downregulated in the *hfq* mutant. Among the genes displaying increased expression in the *hfq* mutant were those coding for ribosomal proteins, amino acid biosynthesis and transport, and, surprisingly, genes encoding pertussis toxin subunits and its secretion apparatus (*ptx*/*ptl* locus).

**Figure 1.** Clustering of transcriptomic data. (**A**) Heat map showing hierarchical clustering of the Euclidean sample-to-sample distance between transcriptomic profiles of wt and Δ*hfq* mutant. (**B**) Principal component analysis was applied to transcriptomic profiles of the wt strain (blue circles) and Δ*hfq* mutant (red circles). Each dot represents an independent biological replicate.

To get better insight into the functional profiles of Hfq-dependent genes we performed gene ontology (GO) enrichment analysis using all DE genes. GO term analysis revealed that genes belonging to several biological processes were significantly enriched in both up- and downregulated gene sets. As shown in Figure 2A, within the set of genes which were significantly upregulated in the *hfq* mutant, categories such as "Translation", "Regulation of transcription", and "Transmembrane transport" were highly enriched. On the other side, genes belonging to "Transmembrane transport", "Iron–sulfur cluster assembly", "Oxido-reduction process", "Pathogenesis", and "Protein secretion by the type III secretion system" terms were enriched among the transcripts which were significantly downregulated in the *hfq* mutant (Figure 2B). Apparently, GO term analysis of the DE genes recapitulated many of the observations from our previous studies [19].

**Figure 2.** Gene ontology (GO) term analysis of genes significantly downregulated (**A**) or upregulated (**B**) in the Δ*hfq* mutant. Significantly enriched terms from the domain 'Biological processes' and their catenations are summarized and visualized by REVIGO as an interactive graph. Circle size encodes number of genes associated with respective category and red shades encode the significance level of the enrichment.

#### *2.2. The E*ff*ect of Hfq on Proteome and Secretome Composition of B. pertussis*

LC-MS/MS analysis of the *B. pertussis* proteome and secretome identified 1631 and 733 proteins, respectively, whose label-free quantification (LFQ) intensities passed our detection criteria. Principal

component analysis of both datasets revealed that protein profiles identified in samples of the wt strain clustered separately from those of the Δ*hfq* strain (Figure 3).

**Figure 3.** Principal component analysis (PCA) of proteomic samples. PCA was applied to protein profiles of the wt strain (blue circles) and Δ*hfq* mutant (red circles) determined in corresponding cell-associated (**A**) or secreted (**B**) protein fractions. Each dot represents an independent biological replicate.

For pellet proteins, the production of 489 proteins was found to be significantly modulated between wt and *hfq* mutant strains (Table S2). The abundance of 219 proteins was higher in the mutant including 19 "ON" proteins which were not detected in the wt strain. Among this set of proteins, GO terms such as "Cell cycle", "Peptidoglycan synthesis", and "Aromatic amino acid metabolism" were significantly enriched (Figure 4A). On the other hand, 270 proteins displayed significantly higher LFQ intensities in the wt strain including 10 "OFF" proteins which, in contrast to the wt samples, were not detected in the mutant. GO term analysis revealed that biological processes such as "Proteolysis", "Ion transport", "Pathogenesis", and "Protein secretion by the type III secretion system" were enriched among this group of proteins (Figure 4B).

As for the secreted fraction (Table S3), abundance of 114 proteins was higher in the mutant (including six "ON" proteins) and these proteins clustered into categories such as "Transmembrane transport", "Cell adhesion", and "Amino acid transport" (Figure 4C). On the other hand, 445 proteins displayed significantly higher LFQ intensities in the wt strain (including 136 "OFF" proteins). The GO term analysis of these differentially secreted proteins identified rather broad variety of processes including "Transmembrane transport", "Proteolysis", "Response to oxidative stress", "Protein secretion by the type III secretion system", and several processes linked to translation and amino acid biosynthesis (Figure 4D).

Among the proteins with highly increased abundance in the *hfq* mutant in both proteome and secretome datasets were prevalently members of various transporters including tripartite tricarboxylate transporters (TTT) (BP2066, BP3501, and BP1358), ABC transporters (BP2090, BP2352, and BP0663) and tripartite ATP-independent periplasmic transporters [31] (BP1487 and BP1489), lipoprotein BP2271, adhesin FhaS and all five pertussis toxin subunits. Of note, genes encoding these overproduced proteins also belonged to the set of the most upregulated genes in the mutant strain (Table 1).

**Figure 4.** GO term enrichment analysis of genes either significantly upregulated (**A**,**C**) or downregulated (**B**,**D**) in the Δ*hfq* cells and Δ*hfq* culture supernatants, respectively. Biological processes significantly enriched for genes belonging to corresponding functional category are shown as an interactive graph. Significantly enriched terms from the domain "Biological processes" and their catenations are summarized and visualized by REVIGO. Circle size encodes number of genes associated with respective categories, red shades encode the significance level of the enrichment.


**Table 1.** List of genes showing consistently increased RNA and protein abundance in the *hfq* mutant.

<sup>1</sup> Log2FC values of Δ*hfq*/wt comparison are shown for RNA-seq and proteomic analyses. Values which did not pass the statistical significance are shown in italics. ND: not determined in both strains in the respective analysis. ON: protein was not detected in the wt strain within the corresponding fraction.

A substantial number of proteins displayed significantly diminished levels in the *hfq* mutant in both proteomic analyses. The largest group of such proteins belonged to T3SS structural components and its secreted substrates. Among the 20 proteins displaying the most decreased abundance in the Δ*hfq* cells were nine T3SS-specific proteins. Likewise, in the supernatant fraction, amounts of Bsp22 and BopD proteins were dramatically reduced (log2FC < −9) while BopC, Bcr4, BopB and BopN proteins could not be detected (Table 2).


**Table 2.** List of T3SS genes showing consistently decreased RNA and protein abundance in the *hfq* mutant.

<sup>1</sup> Log2FC values of Δ*hfq*/wt comparison are shown for RNA-seq and proteomic analyses. Values that did not pass the statistical significance are shown in italics. ND: not determined in both strains in the respective analysis. OFF: protein was not detected in the Δ*hfq* strain within the corresponding fraction.

Noticeably, one of the virulence factors displaying consistently reduced expression, production and secretion in the *hfq* mutant was the autotransporter Vag8. The Vag8 protein binds and recruits C1 esterase inhibitor and thereby inhibits complement activity and contributes to serum resistance [32–34]. Therefore, we asked whether the reduced production of Vag8 factor would compromise the capacity of the Δ*hfq* strain to evade complement-mediated killing. When compared to recent isolates, Tohama I strain exhibits high susceptibility to serum killing [34], nevertheless, in line with our assumption, the survival of the *hfq* mutant (0.06% ± 0.01%) was dramatically decreased when compared to the wt strain (1.36% ± 0.38%) (Figure 5).

**Figure 5.** Serum killing assay of Δ*hfq* and Tohama I strains. Serum resistance is expressed as percentage of wt (white bar) and Δ*hfq* (grey bar) cells that survived upon incubation with 10% human serum when compared to controls (bacteria incubated with heat-inactivated serum). The error bars represent the standard deviation of the mean obtained from three biological replicates (\*, *p* < 0.001). The result is representative of three independent experiments.

#### *2.3. Correlation between Transcriptome, Proteome and Secretome Datasets*

Considering previous reports, the correlation between RNA-seq and LC-MS/MS analysis of cell-associated proteins was relatively high (r<sup>2</sup> <sup>=</sup> 0.40, *<sup>p</sup>*-value 1.2 <sup>×</sup> <sup>10</sup><sup>−</sup>174) with 148 proteins displaying same trend in abundance as corresponding genes in transcriptomic profiling (Figure 6A). As shown in Figure 6B, the concordance of RNA-seq data with secretome analysis was much less positive (r2 = 0.24, *p*-value 7.4 <sup>×</sup> 10−34) with only 80 proteins showing similar trend between both analyses. Of note, among genes showing strong correlation with both proteomic datasets were those encoding the T3SS apparatus, pertussis toxin and its transport machinery, ABC, TRAP, and TTT transporters and other proteins involved in the primary metabolism. Apparently, highly modulated genes showed better correlation with abundance of corresponding proteins than those for which the expression was changed only slightly above the thresholds of significance.

**Figure 6.** Correlation analysis between transcriptomic and proteomic datasets. (**A**) Scatterplots representing the pairwise comparisons of Δ*hfq*/wt log2 ratios between transcriptome and either proteome (left) or secretome (right). Only the genes for which levels of corresponding proteins were reliably detected by label-free quantification were used for correlation analysis. Red dots depict genes which are significantly deregulated in both datasets (*p*-value < 0.05, |log2FC| > 1). Line depicts the best fit as predicted by linear regression. (**B**) Venn diagrams showing the number of differentially expressed genes and proteins and the overlap between each dataset. Left: comparison of transcriptome and proteome. Right: comparison of transcriptome and secretome. Zero values indicate intersections which can not materialize (i.e., being up- and downregulated in the same dataset).

#### **3. Discussion**

In this study we present first integrative omics analysis of the Hfq regulon in the human pathogen. Our study had several objectives: (a) to corroborate the outcomes of our previous transcriptomic study, (b) identify novel targets of Hfq-specific regulatory activities using high-throughput omics techniques, and (c) compare and evaluate the general effects caused by an important post-transcriptional regulator at the level of transcriptome, proteome, and secretome. Compared to microarray profiling (368 protein coding differentially expressed (DE) genes), the differential expression RNA sequencing identified almost two-fold higher number of deregulated genes (602 protein coding DE genes) in the *hfq* mutant. This finding is not surprising considering the higher sensitivity and reproducibility of the RNA-seq method compared to DNA microarray technique [35,36].

Our data are in line with previous studies reporting modest correlation between transcriptomic and proteomic analyses. Nevertheless, considering the role of Hfq in the post-transcriptional control of gene expression, the correlation coefficient between transcriptome and proteome is relatively high when compared to other studies [22,23,37]. The comparison of RNA-seq with secretome analysis output yielded lower correlation values. This finding can possibly result from "contamination" of bacterial culture supernatants with abundant cytosolic proteins such as components of transcriptional and translational machineries. We speculate that these proteins were released from lysed cells during cultivation and sample preparation and therefore their levels do not correspond to changes in gene expression profiles between wt and Δ*hfq* strains.

Importantly, we corroborated several Hfq-specific effects on gene expression profiles which were seen in our previous microarray study [19]. We recapitulated the strong requirement of Hfq chaperone for T3SS functionality as the expression of T3SS genes and production as well as secretion of T3SS components were significantly reduced in the *hfq* mutant. Especially the differences observed in culture supernatants were enormous (more than two orders of magnitude). Several regulators were shown to play a role in control of T3SS activity in *B. pertussis*, including BvgAS two-component system and BtrAS regulatory circuit [38,39]. The response regulator BvgA activates the expression of an extracytoplasmic function sigma factor *btrS* (BP2234) as well as of *btrU*, *btrV*, and *btrW* genes [40]. While BtrS was shown to be required for efficient transcription of the *bsc* locus encoding the T3SS injectisome, BtrU, BtrV, and BtrW regulatory proteins encoded within the *btr* locus are required for secretion through the T3SS apparatus. Recently, a secreted antagonist of BtrS factor called BtrA (BP2233) exerting negative control over the expression of *Bordetella* T3SS genes was reported [39,41]. Secretion of the BtrA inhibitor reactivates BtrS and, consequently, activates the expression of the T3SS genes [39]. Nevertheless, we did not observe any significant changes in expression of the *btrAS* regulatory node. Moreover, BtrA protein could be detected only in the pellets and its levels were decreased in the mutant (log2FC of −0.53). Apparently, the reduced expression, production and in particular secretion of T3SS components observed in the *hfq* mutant are independent of *btrAS* circuit. Relatively high LFQ intensities of T3SS secreted substrates detected in the wt strain were rather surprising. *B. pertussis* Tohama I represents a laboratory-adapted strain and was suggested to lose its ability to secrete T3SS components during long-term in vitro passaging [40,42,43]. Nevertheless, the capacity to secrete T3SS substrates in Tohama I can be regained upon contact with the host [19,43] or under nutrient limitation [44,45]. We did not use iron- or glutamate-limited media in our experiments and cells were collected in mid exponential phase of growth. Nevertheless, we cannot completely rule out the possibility that our cultures were partially nutrient-limited at the time of harvest. Of note, when compared to Hfq-specific effects at transcriptional and translational levels, the massive differences in protein abundances seen in culture supernatants suggest that Hfq is indirectly required for efficient secretion process through the T3SS apparatus.

In line with our previous reports, expression and production of autotransporter Vag8, a major player in complement evasion [32–34], was significantly reduced in the *hfq* mutant. In support, the *hfq* mutant displayed strongly reduced resistance to serum killing. Increased serum sensitivity of the *hfq* mutant was described also in *Neisseria meningitidis* [46]. We assume that this phenotype can be ascribed to reduced production of the Vag8 protein, as the amounts of BrkA, FhaB, and BapC factors reported to be involved in diversion of complement-mediated killing [47–49] were comparable (BrkA) or even higher in the mutant (FhaB). BapC autotransporter was not detected by proteomics and expression of *bapC* gene was increased in the mutant.

Similarly to several other *hfq*-deficient bacteria, the Δ*hfq* strain of *B. pertussis* displays growth deficit. Based on the results of our microarray study we hypothesized that *hfq* mutant of *B. pertussis* compensates the slower growth with increased production of translation machinery components and proteins involved in transport of nutrients [19]. In support, our current study reveals that the most upregulated genes and corresponding proteins found in the *hfq* mutant are represented predominantly by different types of transport proteins, namely, TTT, TRAP, and ABC transporter families. These solute-binding protein-dependent transporters allow uptake even at very low concentrations of ligands [50]. Interestingly, TTT family transporter genes called "Bug" genes (Bordetella uptake genes) are highly overrepresented in the *B. pertussis* genome as they encode 81 functional TTT proteins [51]. While ligands for majority of these proteins are unknown, crystal structures of BugD and BugE proteins

identified their ligands as aspartate and glutamate, respectively [52,53]. Intriguingly, expression of *ptx*/*ptl* locus and, consequently, production and secretion of pertussis toxin subunits was significantly increased in the *hfq* mutant. In the light of reduced virulence of the mutant and the importance of this toxin for *B. pertussis* pathogenicity [54] it is rather surprising observation which may be conceived as compensatory response to the lack of Hfq.

With regard to observed high impact of *hfq* deletion on gene expression profiles it is of particular interest that abundance of at least 16 transcriptional regulators and five alternative sigma factors was significantly modulated in the *hfq* mutant. These results are in line with already described roles of Hfq in expression of alternative sigma factors [55–58] and suggest that similarly to other bacteria, a substantial part of the Hfq-specific effects seen in *B. pertussis* represents indirect regulation. For example, the expression and production of the iron transport repressor Fur is increased in the *hfq* mutant of *B. pertussis* and, consequently, the expression of several genes responsible for iron delivery was decreased in the mutant. One of the surprising results of this study was the relatively low impact of Hfq on abundance of non-coding RNAs. Recently we have identified small non-coding RNA RgtA that is involved in the regulation of the transport of glutamate, a key metabolite in the *B. pertussis* physiology and the abundance of which in the *hfq* mutant was strongly reduced [59]. Nevertheless, our data indicate that only 40 non-coding transcripts out of the recently identified 400 candidate sRNAs [60] changed their levels in the absence of Hfq. Similarly, integrative analysis of Hfq regulon in *P. fluorescence* identified only four ncRNAs out of 87 whose abundance was dependent on Hfq [30]. Thus, in the light of observed extensive changes in transcriptomic and proteomic profiles observed in *B. pertussis*, the relatively small impact on sRNA levels suggests that Hfq exerts some of its regulatory activities in the sRNA-independent fashion or does not substantially contribute to sRNA stability in *B. pertussis*.

Collectively, this study reveals that impact of Hfq on the gene and protein expression profiles in *B. pertussis* is very profound. The Hfq regulon is comprised of hundreds of genes/proteins making almost 20 % of its genome and covering broad variety of genes and their products involved in different cellular processes. Obviously, these pleiotropic effects associated with loss of Hfq in *B. pertussis* cannot be completely ascribed to its role in posttranscriptional circuits but instead may be related to other global regulators that are themselves targets of Hfq regulation such as transcriptional factors. We assume that several observed effects are linked to impaired growth of the mutant. Especially increased production of proteins implicated in transport of metabolites and essential elements seems to compensate for the physiological defect introduced by deletion of the *hfq* gene. Finally, our study corroborated and further clarified the necessity of Hfq for physiological fitness and pathogenicity of *B. pertussis*. It will be of our primary interest to characterize the exact mechanism rendering the production and secretion of T3SS components strongly dependent on Hfq. Furthermore, we are currently characterizing function of several identified Hfq-dependent sRNAs in the physiology of *B. pertussis*.

#### **4. Materials and Methods**

#### *4.1. Bacterial Strains and Growth Conditions*

The *Bordetella pertussis* Tohama I strain [61] and its isogenic *hfq* deletion mutant were grown on Bordet-Gengou agar (BGA) plates supplemented with 15% sheep blood for 3 to 4 days at 37 ◦C. For liquid cultures, bacteria were grown in Stainer–Scholte (SS) medium [62] supplemented with 0.1% cyclodextrin and 0.5% casamino acids (Difco) at 37 ◦C. To harvest samples for RNA and protein isolation, the *B. pertussis* cells were grown overnight in SS medium to mid exponential phase of growth (OD600 ≈ 1.0). Three independent cultivation experiments were performed to collect three biological replicates for each of both strains for RNA and protein isolation.

#### *4.2. RNA Isolation*

Total RNA was isolated using TRI Reagent (Sigma, Darmstadt, Germany) according to manufacturer's protocol. Removal of DNA was achieved by treatment of samples with TURBO DNA-free kit (Thermo Fisher Scientific). RNA quality and quantity was determined by agarose gel electrophoresis and using the Nanodrop 2000 machine (Thermo, Carlsbad, CA, USA). Furthermore, the RNA quality was assessed at sequencing facility (Vienna Biocenter Core Facility, NGS unit) on an Agilent 2100 Bioanalyzer device. All samples displayed RNA integrity numbers higher than 9.

#### *4.3. Library Preparation and Deep Sequencing*

Ribosomal RNA was depleted with the Ribo-Zero rRNA Removal Kit for Bacteria (Illumina, San Diego, CA, USA). Libraries were prepared using NEBNext® Ultra™ II DNA Library Prep Kit for Illumina and sequenced on an Illumina HiSeq 2500 platform using HiSeqV4 chemistry with single-end 50-base-pair reads at the Vienna Biocenter Core Facilities Next Generation Sequencing unit. Reads were demultiplexed and quality trimming and adapter removal from the reads was performed using trimmomatic [63]. After quality control and adapter clipping, the reads were mapped to *B. pertussis* Tohama I reference genome using segemehl [64] with default parameters. Reads per gene counts were deduced with htseq-count with default parameters [65]. Differential gene expression analysis was performed with DESeq2 [66]. Genes with a |log2 fold change| > 1 and a *q*-value (*p*-value adjusted for multiple testing correction by the method of Benjamini and Hochberg [67]) < 0.05 were considered as significantly deregulated. RNA-seq data from the sequencing runs were deposited at the European Nucleotide Archive (ENA) under project accession number PRJEB32623.

#### *4.4. Protein Isolation and Sample Preparation for Proteomics*

Cultures of *B. pertussis* were pelleted by centrifugation (10,000× *g*, 4 ◦C, 10 min) to separate cell pellets and culture supernatants. Cells were resuspended in TEAB digestion buffer (100 mM Triethylammonium bicarbonate, pH 8.5, 2% sodium deoxycholate) and lysed by sonication. For analysis of supernatant fractions, supernatants were filtered through 0.22-μm filters and precipitated with 10% (*w*/*v*) trichloracetic acid (Sigma) overnight at 4 ◦C. Precipitated proteins were collected by centrifugation (14,000× *g*, 4 ◦C, 20 min), washed with 80% acetone (*w*/*v*) and finally dissolved in TEAB digestion buffer. Protein concentrations were determined using BCA protein assay kit (Thermo Fischer Scientific) and 20 μg of protein per sample were used for protein analysis. Cysteines were reduced with M Tris(2-carboxyethyl)phosphine (60 ◦C for 60 min) and blocked with 1M methyl methanethiosulfonate (10 min, room temperature). Samples were digested with trypsin (trypsin to protein ratio 1:20) at 37 ◦C overnight. Digestion of samples was stopped by addition of trifluoracetic acid (Sigma) to a final concentration of 1% (*v*/*v*). SDC was removed by extraction with ethylacetate [68] and peptides were desalted on C18 column (Michrom Bio, Auburn, CA, USA).

#### *4.5. Label-Free Proteomic Analysis by LC-MS*/*MS*

A nanoreversed phase column (EASY-Spray column, 50 cm × 75 μm ID, PepMap C18, 2 μm particles, 100 Å pore size) was used for LC-MS analysis. Mobile phase buffer A was composed of water and 0.1% formic acid. Mobile phase B was composed of acetonitrile and 0.1% formic acid. Samples were loaded onto the trap column (Acclaim PepMap300, C18, 5 μm, 300 Å wide pore, 300 μm × 5 mm) at a flow rate of 15 μL/min. Loading buffer was composed of water, 2% acetonitrile, and 0.1% trifluoroacetic acid. Peptides were eluted with gradient of B phase ranging from 4% to 35% over 60 min at a flow rate of 300 nL/min. Eluting peptide cations were converted to gas-phase ions by electrospray ionization and analyzed on a Thermo Orbitrap Fusion (Q-OT-qIT, Thermo Fischer). Survey scans of peptide precursors from 350 to 1400 m/z were performed at 12 resolution (at 200 *m*/*z*) with a <sup>5</sup> <sup>×</sup> 105 ion count target. Tandem MS (MS2) was performed by isolation within 1.5-Th window with the quadrupole, HCD fragmentation with normalized collision energy of 30, and rapid scan MS analysis in

the ion trap. The MS<sup>2</sup> ion count target value was set to 10<sup>4</sup> and the maximal injection time was 35 ms. Only those precursors with charge state 2–6 were sampled for MS2. The dynamic exclusion duration was set to 45 s with a 10 ppm tolerance around the selected precursor and its isotopes. Monoisotopic precursor selection was turned on. The instrument was run in top speed mode with 2 s cycles [69].

Raw data were imported into MaxQuant software (version 1.5.3.8) [70] for identification and label-free quantification of proteins. The false discovery rate (FDR) was set to 1% for peptides and minimum specific length of seven amino acids. The Andromeda search engine [71] was used for the MS/MS spectra search against the Uniprot *Bordetella pertussis* database (downloaded on November 2016), containing 3258 entries. Enzyme specificity was set as C-terminal to Arg and Lys, also allowing cleavage at proline bonds and a maximum of two missed cleavages. Dithiomethylation of cysteine was selected as fixed modification and N- terminal protein acetylation and methionine oxidation as variable modifications. The "match between runs" feature of MaxQuant was used to transfer identifications to other LC-MS/MS runs based on their masses and retention time (maximum deviation 0.7 min) and this was also used in quantification experiments. Protein abundance was calculated from obtained label-free protein intensities using the MaxLFQ algorithm described recently [72]. Proteins with less than four MS/MS spectral counts were removed from the analysis. Statistics and data interpretation were performed using Perseus 1.6.1.3 software [73]. The normalized label free intensities were compared between wt and *hfq* mutant and each abundance ratio was tested for significance with two-group *t*-test (*p*-value < 0.05). The *p*-values were further adjusted for multiple testing correction to control the false discovery rate at cut off of 0.05 using the permutation test (number of randomization 250). Proteins with corrected *p*-value (*q*-value) < 0.05 were considered as significantly modulated. For downstream analyses (e.g., GO term enrichment) only proteins which were detected by at least two unique peptides in at least two of the three biological replicates were considered. Proteins for which label free intensities were not obtained in any of the replicates of either the wt or the Δ*hfq* strain were considered as significantly modulated and defined as "ON/OFF". The proteomics data were deposited to the ProteomeXchange Consortium via the PRIDE [74] partner repository with the dataset identifier PXD013953.

#### *4.6. GO Term Enrichment Analysis*

To gain a comprehensive functional annotation of the reference genome, gene ontology (GO) terms per gene were deduced using blast2go [75]. For the GO term enrichment analysis significantly deregulated genes from the transcriptome and proteome analysis were split into up- and downregulated genes and each gene set was analyzed separately. Each GO term which is associated with more than one gene in the gene set was tested for enrichment in the gene set compared to the whole transcriptome, applying a Fisher's exact test. Afterwards, determined *p*-values were corrected for multiple testing by the method of Benjamini and Hochberg [67]. Enriched GO terms were further summarized and visualized by Revigo [76].

#### *4.7. Transcriptome–Proteome Correlation Analyses*

To correlate the effect of hfq gene deletion on the transcript and protein abundance globally, the log2FC of all genes the products of which were reliably detected (see Chapter 4.5) by the label-free quantification were compared. To this end, the 'lm' function from R was used to fit a linear model between these two datasets. Since the relative errors in log2FC measurements can be expected to be higher for genes with higher *p*-value, each data point was weighted in the course of model fitting by 1 - *q*-value where *q*-value represents the geometric mean of the *q*-values of the proteome and the transcriptome analysis.

#### *4.8. Serum Killing Assay*

Overnight-grown bacterial cultures were diluted in SS medium to 5 <sup>×</sup> 106 bacteria/ ml of culture and supplemented either with intact or heat-inactivated (56 ◦C, 30 min) 10% human serum (Sigma No). Cells were incubated in parallel in the presence of both type of sera for 60 min at 37 ◦C in orbital incubator. Then the bactericidal activity was terminated by addition of 10 mM EDTA, serial dilutions of bacterial samples were plated onto BG agar and colony-forming units (CFU) were counted to assess bacterial survival. Survival was calculated as a percentage of CFU obtained from cultures treated with intact serum compared to CFUs from cultures treated with heat-inactivated serum (control, 100% survival). Mann–Whitney test was applied to assay the statistical significance of observed differences in sensitivity to serum killing.

**Supplementary Materials:** Supplementary materials can be found at http://www.mdpi.com/1422-0067/20/12/ 3073/s1.

**Author Contributions:** A.D. conceived and performed the experiments, analyzed and interpreted the data, and edited the manuscript; F.A. analyzed and interpreted the data and edited the manuscript; D.S. performed the experiment and analyzed the data; D.P. analyzed and interpreted the data and edited the manuscript; B.V. conceived the experiments, supervised the research project, and wrote the original draft of the manuscript. All authors approved the final draft.

**Funding:** This work was supported by grant 19-12338S (to B.V.) from the Czech Science Foundation (www.gacr.cz) and by funding from RVO61388971. This work was also supported by Mobility grant from Czech Academy of Sciences (MSM200201702) to A.D. Acquisition of the proteomic data was supported by the project BIOCEV— Biotechnology and Biomedicine Centre of the Academy of Sciences and Charles University (CZ.1.05/1.1.00/02.0109) from the European Regional Development Fund.

**Acknowledgments:** We thank J. Držmíšek for help with preparing samples for proteomic analysis and Rudy Antoine (Institute Pasteur, Lille) for helpful discussions. We are grateful to Karel Harant and Pavel Talacko from the Mass Spectrometry and Proteomics Service Laboratory, Faculty of Science, Charles University for performing the LC-MS/MS run.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

#### **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/).

## **Targeting** *Pseudomonas aeruginosa* **in the Sputum of Primary Ciliary Dyskinesia Patients with a Combinatorial Strategy Having Antibacterial and Anti-Virulence Potential**

**Giuseppantonio Maisetta 1,\*, Lucia Grassi 1, Semih Esin 1, Esingül Kaya 1, Andrea Morelli 2, Dario Puppi 2, Martina Piras 3, Federica Chiellini 2, Massimo Pi**ff**eri <sup>3</sup> and Giovanna Batoni <sup>1</sup>**


Received: 2 December 2019; Accepted: 17 December 2019; Published: 20 December 2019

**Abstract:** In primary ciliary dyskinesia (PCD) patients, *Pseudomonas aeruginosa* is a major opportunistic pathogen, frequently involved in chronic infections of the lower airways. Infections by this bacterial species correlates with a worsening clinical prognosis and recalcitrance to currently available therapeutics. The antimicrobial peptide, lin-SB056-1, in combination with the cation chelator ethylenediaminetetraacetic acid (EDTA), was previously demonstrated to be bactericidal against *P. aeruginosa* in an artificial sputum medium. The purpose of this study was to validate the anti-*P. aeruginosa* activity of such a combination in PCD sputum and to evaluate the in vitro anti-virulence effects of EDTA. In combination with EDTA, lin-SB056-1 was able to significantly reduce the load of endogenous *P. aeruginosa* ex vivo in the sputum of PCD patients. In addition, EDTA markedly reduced the production of relevant bacterial virulence factors (e.g., pyocyanin, proteases, LasA) in vitro by two representative mucoid strains of *P. aeruginosa* isolated from the sputum of PCD patients. These results indicate that the lin-SB056-1/EDTA combination may exert a dual antimicrobial and anti-virulence action against *P. aeruginosa*, suggesting a therapeutic potential against chronic airway infections sustained by this bacterium.

**Keywords:** antimicrobial peptide; EDTA; *Pseudomonas aeruginosa*; primary ciliary dyskinesia; virulence factor; anti-virulence; sputum; chronic infection

#### **1. Introduction**

Primary ciliary dyskinesia (PCD) is an autosomal recessive disorder characterized by abnormal ciliary ultrastructure and function leading to impaired mucociliary clearance and recurrent respiratory infections [1]. Although *Haemophilus influenzae* is the pathogen most commonly isolated from patients with PCD until adolescence/early adulthood, in adult PCD patients, *P. aeruginosa* plays a major role, especially after the age of 30 [1]. Accordingly, a negative correlation between the abundance of *P. aeruginosa* in the airways of these patients and lung function has been reported [2,3]. The pathogenesis of *P. aeruginosa* infection is at least partially attributable to its ability to synthesize and secrete a number of virulence factors (e.g., pyoverdine, pyocyanin, proteases) and to form biofilms, in which bacterial cells are embedded in an alginate extracellular matrix [4]. Despite intensive antibiotic therapy, once the

patients are stably colonized by *P. aeruginosa*, the eradication of the bacterium is rarely achieved [1,5]. Therefore, there is a critical need for novel antimicrobial drugs that can effectively lower *P. aeruginosa* load in the challenging environment of PCD lung.

Over the last decades, antimicrobial peptides (AMPs) have been intensively investigated as potential antibiotics against multidrug-resistant bacteria [6]. Most AMPs are cationic molecules with an amphipathic structure that selectively target bacterial membranes via electrostatic forces. In contrast to standard antibiotics, AMPs are generally effective against both quiescent and actively growing bacteria, display rapid killing kinetics, and demonstrate low propensity to select resistant mutants in vitro [7,8]. On the other hand, AMPs may display a reduction in their antibacterial potency in the presence of complex biological fluids such as sputum, plasma, or saliva due to the high concentration of salt found in these fluids and/or the presence of anionic proteins and host or bacterial proteases that may neutralize their activity [9,10].

As observed in cystic fibrosis (CF) lungs, the biofilm mode of growth of bacteria together with the lung mucus viscosity reduces the effectiveness of conventional antibiotic therapy in PCD patients [11]. Thus, the usage of adjuvants has been proposed to improve the diffusion of antimicrobials through the mucus and the biofilm matrix and facilitate the targeting of bacterial cells [12]. Previous studies have shown that the divalent cation chelator ethylenediaminetetraacetic acid (EDTA) can destabilize the biofilm structure by interfering with the ionic attractive forces among the biofilm matrix components [13,14]. EDTA is prescribed in a number of clinical conditions demonstrating high tolerability (up to 2 g once a week, intravenously injected) and efficacy [15]. Recently, we demonstrated that the optimized semi-synthetic antimicrobial peptide lin-SB056-1 in combination with EDTA is able to exert a synergistic bactericidal effect against *P. aeruginosa* in an artificial sputum medium resembling CF sputum [16]. Similarities and differences between CF and PDC sputum have been reported. For instance, while both diseases seem to be associated with a similar degree of airways neutrophilia, the concentration of interleukin-8 in sputum is higher in PCD than in CF patients, while neutrophil elastase activity is lower in PCD compared with CF [17]. In order to evaluate the therapeutic potential of the lin-SB056-1/EDTA combination in PCD, in this study, we evaluated its bactericidal activity ex vivo, against endogenous *P. aeruginosa* in the sputum from PCD patients. Importantly, the ex vivo sputum mimics, with good approximation, the lung environment, as it contains both host and bacterial components, including bronchial mucus contaminated by saliva, serum proteins, inflammatory mediators, desquamated epithelial cells, and pathogenic bacteria, as well as bacteria from the normal flora [18].

Previous reports have demonstrated the involvement of different cations (i.e., calcium, magnesium, and zinc) either in the regulation of gene expression or in the production and processing of virulence factors in *P. aeruginosa* [19,20]. Thus, herein, we also evaluated the ability of EDTA to reduce the production of relevant virulence factors of *P. aeruginosa* (e.g., pyoverdin, pyocyanin, proteases, biofilm production). Overall, the results obtained demonstrated that the lin-SB056-1/EDTA combination is able to significantly reduce *P. aeruginosa* load ex vivo and that EDTA is highly active in suppressing the production of relevant bacterial virulence factors, suggesting a dual antibacterial and anti-virulence potential of the combination.

#### **2. Results**

#### *2.1. Killing Activity of lin-SB056-1 in Combination with Ethylenediaminetetraacetic Acid (EDTA) against Endogenous P. aeruginosa*

*P. aeruginosa* strains were isolated from the sputum of six PCD patients known to be chronically infected with the bacterium. All the strains displayed a mucoid phenotype and different antibiotic susceptibility profiles (Supplementary Table S1).

Diluted sputum (5-fold) from each patient was incubated for 1.5 h with the peptide at 25 μg/mL, alone or in combination with EDTA (0.625 or 1.25 mM), and the colony forming unit (CFU) number of *P. aeruginosa* surviving the treatment was detected. During the incubation time, endogenous

*P. aeruginosa* did not grow in PCD sputum. While the peptide and EDTA were almost inactive when used alone, their combination exerted a significant synergistic killing effect against endogenous *P. aeruginosa*, although with different efficacy depending on the sputum sample (Figure 1). When compared to the corresponding controls, the reduction in CFU number caused by the combination ranged from approximately 0.3 Log-units (50% reduction, grey dot) to 3 Log-units (99.9% reduction, blue triangle) (Figure 1).

**Figure 1.** Antibacterial activity of peptide lin-SB056-1, ethylenediaminetetraacetic acid (EDTA), and both in combination against endogenous *P. aeruginosa* in primary ciliary dyskinesia (PCD) sputum. The effect of lin-SB056-1 and/or EDTA after 1.5 h of incubation in six diluted (1:5) sputum samples was assessed against endogenous *P. aeruginosa* strains (PaM1 to PaM6) by colony forming unit (CFU) counting. Lin-SB056-1 was tested at 25 μg/mL in combination with 0.625 mM EDTA against PaM1 and PaM5 strains (triangles), and with 1.25 mM EDTA against PaM2, PaM3, PaM4, and PaM6 strains (dots). Control (CTRL): bacteria incubated in diluted sputum only. Individual sputum samples are identified with different colors. Results represent the mean of 6 sputa done in duplicate. Error bars indicate the standard error of the mean. \*\* *p* < 0.01, \*\*\* *p* < 0.001 (one-way analysis of variance (ANOVA) followed by the Tukey–Kramer post-hoc test).

#### *2.2. E*ff*ects of EDTA and lin-SB056-1 on Virulence Factors' Production by P. aeruginosa PaM1 and PaM5*

Preliminary experiments indicated that PaM1and PaM5 strains are able to produce high levels of most of the virulence factors analyzed; therefore, these strains were selected to evaluate the effect of sub-inhibitory concentrations of EDTA on virulence factors' production. To this aim, we first evaluated the susceptibility of PaM1 and PaM5 strains to EDTA in liquid medium, in terms of minimum inhibitory concentration (MIC). A concentration of 1.25 mM EDTA was able to inhibit visible bacterial growth (MIC), while the concentrations of 0.075 and 0.15 mM were sub-inhibitory and therefore, were selected for the subsequent experiments.

Pyocyanin is a greenish pigment secreted by *P. aeruginosa* that enhances the inflammatory response and causes tissue damage in the host [21]. As shown in Figure 2a, EDTA at the concentrations of 0.075 and 0.15 mM, highly inhibited pyocyanin production by the PaM1 strain at 72 h (by 62% and 70%, respectively) as compared to the untreated cells. Regarding the PaM5 strain, which was a low pyocyanin producer (Figure 2b), EDTA at both concentrations caused a reduction of approximately 40% in the production of such pigment, although the difference did not reach statistical significance compared to the untreated cells.

*P. aeruginosa* produces and secretes a number of proteases, such as LasA, elastase B (LasB), protease IV, and alkaline protease, which are considered important virulence factors as they damage host tissues and interfere with host antibacterial defense mechanisms [22]. The total proteolytic activity of PaM1 was completely abolished in the presence of either 0.075 or 0.15 mM EDTA (Figure 2a). Similarly, EDTA significantly reduced the proteolytic activity of the PaM5 strain but only at the concentration of 0.15 mM (Figure 2b).

LasA is a zinc-dependent metalloprotease secreted by *P. aeruginosa*. It exhibits a staphylolytic activity, enhances the elastolytic activity of LasB in vivo, and induces shedding of syndecans, a family of cell surface heparan sulfate proteoglycans, from host cell surfaces [23]. LasA activity of both the PaM1 and PaM5 strains was significantly inhibited in the presence of 0.075 and 0.15 mM EDTA, with a reduction of approximately 70% and 80%, as compared to the untreated control, respectively (Figure 2a,b). Further experiments were undertaken in order to evaluate whether the reduction of LasA activity was ascribable to the inhibition of protease synthesis or, rather, to the inhibition of the enzyme activity due to zinc chelation by EDTA (Figure S1). To this aim, the assessment of the PaM1 staphylolytic activity was performed in the presence of exogenously added zinc (0.1 mM ZnSO4). When ZnSO4 was added directly in the enzyme assay, at the end of the incubation period (72 h), no significant increase in LasA activity was observed, suggesting that the low levels of LasA activity detected in the presence of EDTA were likely due to inhibition of protein synthesis and not to the chelation of the enzyme cofactor. In contrast, when ZnSO4 was added at the beginning of the incubation of PaM1 with EDTA, LasA activity was restored (Supplementary Figure S1), indicating that the excess of zinc could overcome the inhibitory effect of EDTA. Overall, these data suggest that EDTA may act by interfering with the expression/procession of LasA protease by PaM1 strain rather than by sequestering the zinc cofactor.

Pyoverdin is a chelator involved in iron binding and cellular uptake in a low-iron environment [24]. EDTA at both concentrations tested did not reduce the level of pyoverdin in culture supernatants of both *P. aeruginosa* strains (Figure 2a,b).

Studies on mucoid *P. aeruginosa* isolates have shown that alginate plays a critical role in biofilm establishment and persistence by protecting bacteria against antibiotics and phagocytosis [25,26]. Although EDTA did not inhibit the production of alginate in culture supernatants of both strains (Figure 2a,b), it was able to significantly reduce the viscosity of PaM1 culture supernatants at both concentrations tested (Figure 2a).

Finally, EDTA was tested for its antibiofilm activity against the PaM1 strain. A reduction of 40% and 57% in PaM1 biofilm formation was observed in the presence of 0.075 and 0.15 mM EDTA, respectively (Figure 2a).

The impact of lin-SB056-1 on the production of virulence factors by PaM1 and PaM5 strains was also evaluated. As reported in Table S2, the peptide at sub-inhibiting concentrations did not reduce the production of any of the virulence factors analyzed for both bacterial strains tested.

**Figure 2.** Effects of EDTA on virulence factor production by (**a**) PaM1 and (**b**) PaM5 strains. PaM1 and PaM5 cultures were incubated at 37 ◦C in the presence or absence of EDTA for 72 h. Following incubation, OD600 was measured prior the quantification of the virulence factors in culture supernatants (see the Materials and Methods Section for details). Values obtained were normalized by multiplying them by the ratio between OD600 of the control/OD600 of the corresponding EDTA-treated samples and reported as mean +/- SEM of three independent experiments. CTRL: bacteria incubated without EDTA; \* *p* < 0.05, \*\* *p* < 0.01 (one-way ANOVA followed by the Tukey–Kramer post-hoc test).

#### **3. Discussion**

Similar to CF patients, eradication of chronic *P. aeruginosa* infection in PCD lungs is hardly obtained, and the reduction of bacterial density during chronic colonization or exacerbations is often the aim of the antimicrobial therapy [3]. In previous studies, we have shown that the combination of lin-SB056-1/EDTA possesses antimicrobial activity against *P. aeruginosa* in artificial sputum medium and prevents *P. aeruginosa* biofilm formation in an in vivo-like three-dimensional (3D) lung epithelial cell model [16,27]. Despite resembling the airway mucus, artificial sputum media normally behave like Newtonian fluids lacking many of the intramolecular interactions and covalent cross-links that give respiratory secretions their viscoelastic characteristic [28]. Furthermore, the genotype and physiological state of *P. aeruginosa* cells found in vivo may significantly differ from those of bacteria grown in laboratory media [29]. Hence, in this study, we sought to validate the anti-pseudomonal activity of the lin-SB056-1/EDTA combination in conditions more closely resembling the environment found in vivo. To this aim efficacy of the combination was tested ex vivo, against endogenous *P. aeruginosa* in the sputum of PCD patients. Differently from artificial sputum medium, patients' sputum contains host/bacterial components such as cell-derived factors, normal flora, inflammatory mediators, proteases, or peptidases that may exert an additional inhibitory effect on the peptide's activity. Nevertheless, herein, we showed that the combination of lin-SB056-1 and EDTA at sub-active and non-cytotoxic concentrations [16,27] determined a significant reduction in *P. aeruginosa* load in sputa of PCD chronically infected patients, despite certain differences in the level of reduction among different sputum samples being observed. Such differences are not surprising considering that inter-patient variables (e.g., clinical stage and severity of lung infection, bacterial load, sputum sample composition, and consistency) were not standardized in our experiments, in the attempt to mimic conditions found during the actual antimicrobial therapy. The mechanisms of the synergistic effect of EDTA on the peptide's activity might be multiple. Due to the chelation of divalent cations from their binding sites in lipopolysaccharide (LPS), EDTA may destabilize the bacterial outer membrane, thus increasing the permeability to lin-SB056-1 molecules and their interaction with the bacterial membranes. In addition, at least part of the synergistic effect observed in sputum could be ascribed to the ability of EDTA to reduce sputum viscosity, thus favoring peptide diffusion [16]. Finally, EDTA could also neutralize the inhibitory effect of sputum on the peptide's activity, sequestering cations that may interfere with the electrostatic interactions of the peptide with bacterial surface [9,10]. The possible use of EDTA in the treatment of pulmonary infections and its safety as an adjuvant has been highlighted in previous in vivo studies [30,31]. For instance, Liu and coworkers demonstrated, in the guinea pig model, that EDTA (30 mg/kg intraperitoneally injected) plus ciprofloxacin (4 μg/mL administered by inhalation) significantly reduced the *P. aeruginosa* CFU number per gram of lung tissue as compared to the single treatment groups [31].

Interestingly, the results obtained in this study clearly demonstrated that EDTA could not only favor the activity of lin-SB056-1 in ex vivo conditions, but could also reduce in vitro, at sub-inhibitory concentrations, the production of several *P. aeruginosa* virulence factors (i.e., pyocyanin, total protease and LasA), which are known to play a crucial role in the pathogenesis of *P. aeruginosa* infections. The action of EDTA as an anti-virulence molecule could be ascribed to its capacity of binding divalent cations, many of which are critical for the expression/processing of virulence factors of *P. aeruginosa*. In particular, the reduction of pyocyanin observed in the presence of EDTA is in line with previous reports demonstrating a positive correlation between calcium levels and the expression of proteins involved in the pathway of pyocyanin biosynthesis [32]. Analogously, the reduction of proteases by EDTA is in agreement with previous observations reporting that zinc ions are important for the efficient production and processing of different proteases, such as LasA, LasB, and protease IV [19,33].

On the contrary, EDTA did not inhibit pyoverdin production, in agreement with the observation that neither calcium nor magnesium enhances pyoverdin production [34,35]. Interestingly, although EDTA did not inhibit alginate production, it was able to markedly reduce the viscosity of culture supernatants of the PaM1 strain. It can be hypothesized that this effect may be due to the sequestration of calcium ions that are crucial for alginate cross-linking [36]. A similar mechanism may be involved in the ability of EDTA to reduce the formation of PaM1 biofilm, confirming previous observations in which EDTA significantly reduced biofilm formation by a mucoid strain of *P. aeruginosa* either in vitro or in a guinea-pig model of lung infection [31]. Overall, the ability of EDTA to reduce the accumulation and/or activity of important virulence factors might contribute to limit the pathogenicity of *P. aeruginosa*.

In conclusion, in the present study, we demonstrated that the lin-SB056-1/EDTA combination is able to significantly reduce *P. aeruginosa* load in PCD sputum, and that EDTA decreases the production of relevant virulence factors of mucoid *P. aeruginosa* in vitro. Such results suggest a dual antimicrobial and anti-virulence effect of the lin-SB056-1/EDTA combination and highlight the possible use of EDTA as an adjuvant in the treatment of chronic *P. aeruginosa* lung infections.

#### **4. Materials and Methods**

#### *4.1. Sputum Collection and Treatment*

The sputum samples were collected by spontaneous expectoration from six PCD patients following informed consent. The study was conducted in accordance with the Declaration of Helsinki, and the protocol was approved by the Ethics Committee of Pisa (Protocol number 62532, 11.06.2016). PCD patients included in this study (median age: 34 years) were chronically infected by *P. aeruginosa* and characterized by frequent relapses of infection. A volume of 0.5–1 mL of sputum was collected from each patient after interruption (at least 14 days) of the antibiotic therapy regimen. For easier handling of the samples, the dense and sticky sputa were diluted five-fold in sodium phosphate buffer 10 mM, pH 7.4 (SPB). Samples were plated on cetrimide (Sigma Aldrich, Saint Louis, MO, USA) and MacConkey (Oxoid Basingstoke, Hampshire, UK agar to confirm the presence of *P. aeruginosa* and assess the mucoid phenotype of the colonies, respectively.

#### *4.2. Peptide and EDTA Solutions*

Lin-SB056-1 peptide (KWKIRVRLSA-NH2) was purchased from Peptide Protein Research, Ltd. (Fareham, UK) with a purity of 98%. EDTA (disodium salt) was obtained from Sigma-Aldrich. A stock solution of disodium-EDTA (0.5 M) was prepared in milli-Q water by adjusting the pH to 8.0 with NaOH (Sigma-Aldrich). The stock solution was then diluted in milli-Q water to obtain a working solution of 50 mM that was sterilized and stored at 4 ◦C.

#### *4.3. Susceptibility Testing*

Identification and susceptibility testing of *P. aeruginosa* strains isolated from sputum samples (PaM1, PaM2, PaM3, PaM4, PAM5, PaM6) were performed by MALDI-TOF (Bruker Daltonics, Bremen, Germany) and VITEK 2 automatic instruments (BioMerieux, Lyon, France), respectively (Table S1). Determination of minimum inhibitory concentration (MIC) of EDTA towards PaM1 and PaM5 strains was performed according to the standard microdilution method in Muller–Hinton broth (Oxoid) [37]. The MIC was defined as the lowest concentration of EDTA that completely inhibited visible growth of bacteria after 24 h of incubation.

#### *4.4. Bactericidal Assay in Patients' Sputum*

An aliquot of each PCD sputum was serially diluted and plated on selective cetrimide agar, to assess the CFU number of endogenous *P. aeruginosa* at time 0. After that, a volume of 90 μL of diluted (1:5) sputum of each patient was incubated with sub-bactericidal concentrations of peptide and EDTA, used alone or in combination, for 1.5 h at 37 ◦C. Following incubation, samples were serially diluted and plated on selective cetrimide agar for assessing the *P. aeruginosa* CFU number.

#### *4.5. Assays for Evaluation of Virulence Factors in Culture Supernatants*

Colonies of mucoid strains PaM1 and PaM5 grown on MacConkey agar were suspended in Luria Bertani (LB) broth (Sigma-Aldrich) to obtain an OD600 of 0.1. Cultures were incubated in the presence or in the absence of EDTA (0.075 or 0.15 mM) in static conditions at 37 ◦C for 72 h. Following incubation, the OD600 of the cultures was determined to account for bacterial density. After that, cultures were centrifuged at 10,000× *g* for 20 min at room temperature and culture supernatants were used for the quantification of virulence factors. The same protocol was followed to evaluate the eventual effects of linSB056-1 on the production of virulence factors. To this aim, the peptide was added to PaM1 and PaM5 cultures at the concentration of 6.25 μg/mL and 12.5 μg/mL that were sub-inhibitory for both bacterial strains.

Pyocyanin was extracted from cell-free supernatants with subsequent exposure to chloroform and 0.2 N hydrochloric acid (Sigma-Aldrich) and quantified at OD520 nm, as previously described [38].

Total proteolytic activity was determined using a modified skim milk assay [39]. Briefly, culture supernatants of PaM1 and PaM5 strains (0.5 mL) were incubated with 0.5 mL skim milk (Sigma-Aldrich) (1.25%) at 37 ◦C for 30 min and turbidity was measured at OD600 nm. The decrement in turbidity due to proteolytic activity was expressed as ΔA/min/mL.

Secreted LasA of *P. aeruginosa* has a staphylolytic activity, i.e., it causes a decrement in the OD600 of a culture of *Staphylococcus aureus*. LasA activity was assessed by evaluating the ability of cell-free supernatants from *P. aeruginosa* exposed or not exposed to EDTA to lyse boiled cells (intact) of *S. aureus* American Type Culture Collection (ATCC) 33591 and expressed as ΔA/min/mL [40]. Due to the role of zinc as a cofactor of LasA, in some experiments, the staphylolytic activity in the presence of EDTA was evaluated by adding 0.1 mM ZnSO4 (Sigma-Aldrich) to the enzyme assay.

Pyoverdin was quantified by measuring the OD400 of cell-free supernatants [41].

The quantification of alginate was performed by carbazole-borate assay according to Heidari et al. [42]. Shear viscosity of culture supernatants was assessed by rheometric measurement at 25 ◦C, applying a shear stress of 1 Pa/s on 150 μl of supernatant using a gap between the rheometer plates of 52 μm (Rheometer Scientific RM500, Reologica Instruments AB, Lund, Sweden).

The value obtained for each virulence factor was multiplied by the ratio OD600 of the control/OD600 of the sample, to normalize for small differences in the culture densities between the controls and the EDTA-exposed samples after 72 h of incubation.

#### *4.6. Biofilm Inhibition Assay*

*P. aeruginosa* PaM1 grown in tryptone soy broth (TSB) for 48 h at 37 ◦C was diluted 1:20 in TSB supplemented with 0.25 mM CaCl2. Bacterial suspensions were inoculated into flat-bottom polystyrene 96-well microplates (Corning Costar, Lowell, MA, USA) in the absence (negative control) or in the presence of EDTA at sub-inhibiting concentrations (0.075 and 0.15 mM). Microplates were incubated statically at 37 ◦C for 48 h and biofilm biomass was estimated by crystal violet (CV) (Sigma-Aldrich) staining assay, as previously described [43].

#### *4.7. Statistical Analysis*

Data reported in Figure 1 represent the mean of 6 experiments done in duplicate. Figure 2 and Supplementary Figure S1 depict the data obtained from three independent experiments, while Table S2 reports the mean of two independent experiments. Differences between mean values of groups were evaluated by one-way analysis of variance (ANOVA) followed by the Tukey–Kramer post-hoc test. A *p*-value < 0.05 was considered statistically significant. Data analysis was performed with GraphPad In Stat (GraphPad Software, La Jolla, CA, USA).

*Int. J. Mol. Sci.* **2020**, *21*, 69

**Supplementary Materials:** Supplementary Materials can be found at http://www.mdpi.com/1422-0067/21/1/69/s1. Table S1: Patients' information, colony phenotype, and resistance profile of *P. aeruginosa* strains isolated from PDC sputum; Table S2: Effects of lin-SB056-1 on the production of virulence factors by PaM1 and PaM5 strains; Figure S1: Effect of exogenously added Zinc on LasA staphylolytic activity of PaM1 strain in the presence of EDTA.

**Author Contributions:** Conceptualization, G.M., S.E., M.P. (Massimo Pifferi), and G.B.; methodology, G.M., L.G., M.P. (Martina Piras), E.K., A.M., D.P.; validation, G.M., S.E., M.P. (Massimo Pifferi), and G.B.; formal analysis, G.M., L.G., M.P. (Martina Piras), E.K. and F.C.; writing—original draft preparation, G.M. and G.B.; writing—review and editing, L.G., S.E., E.K., F.C., and M.P. (Martina Piras), supervision, M.P. (Massimo Pifferi), S.E., and G.B.; funding acquisition, G.B. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the University of Pisa, grant number PRA 2017\_18.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

#### **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* **From Gene to Protein—How Bacterial Virulence Factors Manipulate Host Gene Expression During Infection**

#### **Lea Denzer, Horst Schroten and Christian Schwerk \***

Department of Pediatrics, Pediatric Infectious Diseases, Medical Faculty Mannheim, Heidelberg University, 68167 Mannheim, Germany; lea.denzer@medma.uni-heidelberg.de (L.D.); horst.schroten@umm.de (H.S.) **\*** Correspondence: christian.schwerk@medma.uni-heidelberg.de; Tel.: +49-621-383-3466

Received: 30 April 2020; Accepted: 20 May 2020; Published: 25 May 2020

**Abstract:** Bacteria evolved many strategies to survive and persist within host cells. Secretion of bacterial effectors enables bacteria not only to enter the host cell but also to manipulate host gene expression to circumvent clearance by the host immune response. Some effectors were also shown to evade the nucleus to manipulate epigenetic processes as well as transcription and mRNA procession and are therefore classified as nucleomodulins. Others were shown to interfere downstream with gene expression at the level of mRNA stability, favoring either mRNA stabilization or mRNA degradation, translation or protein stability, including mechanisms of protein activation and degradation. Finally, manipulation of innate immune signaling and nutrient supply creates a replicative niche that enables bacterial intracellular persistence and survival. In this review, we want to highlight the divergent strategies applied by intracellular bacteria to evade host immune responses through subversion of host gene expression via bacterial effectors. Since these virulence proteins mimic host cell enzymes or own novel enzymatic functions, characterizing their properties could help to understand the complex interactions between host and pathogen during infections. Additionally, these insights could propose potential targets for medical therapy.

**Keywords:** virulence factors; bacteria; host-pathogen interaction; gene expression; immune response; manipulation; inflammation; persistence; replicative niche

#### **1. Introduction**

Successful defence against extracellular and intracellular bacteria primarily relies on the ability of innate immune cells to sense present bacteria followed by activation of the adequate matching immune response. The identification of bacteria is enabled by a broad array of pathogen recognition receptors (PRRs), which recognize extensively conserved pathogen-associated molecular patterns (PAMPs) as nucleid acids, cell wall components and proteins from viruses, bacteria, fungi and parasites [1]. Toll like receptors (TLRs) and NOD-like receptors (NLRs) represent the two major classes of PRRs, acting at the cell surface or in the cytoplasm, respectively [2]. After activation, PRRs induce multiple signalling pathways aiming at the expression of proinflammatory cytokines, which regulate the innate immune response. The most prominent pathways involved are mediated by mitogen-activated protein kinases (MAPKs) or nuclear factor-κB (NF-κB), which are initiated by path-specific adapter proteins and transferred by downstream phosphorylation cascades [3]. Interestingly, all pathways can synergize to guarantee a specific and appropriate immune response for the present pathogen. This is enabled by the high diversity of PRRs, PAMPs and PRR adapter proteins and their numerous ways to be combined during innate immune response [1,4].

The immune response has to be tightly controlled to ensure a clearance of the bacteria but also to prevent tissue damage and necrosis as result of sepsis. There are several levels to influence the expression of inflammatory genes. A first level of interference is changing of the DNA's structure on the chromatin level. Epigenetic modulation enables remodelling of the chromatin to transfer heterochromatin into euchromatin allowing transcription or vice versa [5–7]. In addition, the affinity of promotors and other regulatory DNA sequences for RNA polymerases and transcription factors (TFs) can be influenced by cytosine or adenine methylation. To induce transcription, TFs and RNA polymerases are recruited to target genes, a step that represents another level to regulate gene expression. Only a minor portion (fewer than 2%) of genes is transcribed into mRNAs, instead the majority is transferred into so called non-coding RNAs (ncRNAs). The long ncRNAs (lncRNAs) as well as some classes of short ncRNAs are also involved in epigenetic regulations but its most studied group, the miRNAs (microRNA), are mostly involved in RNA destruction [8].

Another point for interference with gene expression is during processing of mRNAs, which includes 5 capping, alternative splicing and polyadenylation [9,10]. Primary ncRNAs can be processed by the protein complex DICER (eukaryotic ribonucleases) to generate miRNAs, which can negatively regulate expression of its primary transcript. To guarantee proper cell function, mRNAs need to be degraded after a certain time frame; the RNA stability is, therefore, another switch to modify gene expression. Polyadenylation and 5 capping prolongs RNA stability but several enzymes are able to decap the 5 cap and to remove the polyadenyl tail of the RNA leaving an unprotected mRNA [11–13]. Nevertheless, these structures are crucial for translation initiation and can gain enough time for the mRNA to be translated. This represents a further step for interference with gene expression, as there is the need of several factors to induce and prolong translation [14]. Translation initialisation factors have to be recruited leading to ribosome assembly and binding of the first amino acid loaded tRNA (transfer RNA). To keep translation ongoing, elongation factors and ATP (Adenosine triphosphate) have to be present. The nascent protein chain then needs to fold into its physiological form to be active [15,16]. Therefore, protein-folding and activation catalysed by different chaperones is another important step during gene expression and is also tightly controlled. Finally, a last step to regulate gene expression is represented by the stability and degradation of, in some cases misfolded, proteins [17,18]. An overview of the steps during host gene expression targeted by bacterial pathogens, as well as the bacteria involved, is given in Figure 1.

Host cells fight bacteria with a proinflammatory cytokine response, lysosomal degradation, autophatic clearance and activation of the unfolded protein response, which in the end can lead to apoptosis [19]. Bacteria can hijack all these defence mechanisms by interfering with the host's gene expression at any level. For that purpose, bacteria express several virulence factors, the so-called effectors. These proteins are able to mimic host enzymes, thereby manipulating the host response following invasion favouring intracellular survival, persistence and spreading. Since proteins and enzymes of signal transduction pathways are involved in all defence mechanisms, they are a favoured target of effector proteins [19]. It is worth noting that other virulence proteins own novel enzymatic functions, which allow them to enter the nucleus and directly induce gene expression or repression. Therefore, these bacterial effectors are termed nucleomodulins [20,21]. In the following we will present an overview on the manipulation of host gene expression at different levels by nucleomodulins and other bacterial effectors.

**Figure 1.** Steps of host gene expression manipulated by bacterial pathogens. The figure provides an overview over the main steps of gene expression that are indicated at the left side (I-VI). The numbers in the scheme highlight distinct characteristic processes that are part of each gene expression level and are listed in the legend at the right side. Different bacterial pathogens (indicated at the right) have been described to target the distinct steps and processes during host gene expression to their favor. For detailed information please refer to the text of this review.

#### **2. Bacterial Virulence Factors Manipulating Host Gene Expression**

#### *2.1. Epigenetic Control of Gene Expression*

The expression of genes is dependent on their accessibility for RNA polymerase II (RNA Pol II) and TFs. As approximately 147 base pairs of the DNA are wrapped around histone octamers build by the subunits H2A, H2B, H3 and H4 as well as the scaffold protein H1 to form the nucleosome, those sequences are protected from transcription [5–7]. Therefore, the packaging of the nucleosomes defines the chromatin state into euchromatin and heterochromatin enabling transcription or blocking it. In order to react properly to a certain stimulus, the chromatin state can be remodeled to give access to the required genes, a process called nuclear remodeling or histone modification. Enzymes

posttranslationally modify the amino acids at the N-termini of the histone proteins (called histone tails) by acetylation, phosphorylation, methylation and ubiquitination in a reversible manner to modify the interaction between neighbored nucleosomes favoring an open or closed chromatin state [19,20,22]. Nucleosomes are then allowed to slide along the chromatin fiber in an ATP-dependent manner, to give access to the DNA sequence. This reveals the dual function of chromatin, to provide a natural scaffold and being part of an essential regulatory signaling network processing the incoming data to create a special transient biological output [23]. On top, established posttranslational modifications (PTMs) can be maintained beyond the initial signal and cell divisions inheriting cell type specific gene expression enabling cell lineage specification and cellular identity [23–25].

The enzymes responsible for the modulation of the histone tails are divided into "writers," which attach the chemical units, "readers," which recognize and translate them by recruitment of activating or repressing factors and "erasers," which remove the modifications. The resulting "epigenetic code" is highly dynamic, as each established modification influences the addition or removement of other modifications that in turn influence the own stability and persistence. Moreover, epigenetic mechanisms represent the missing link between more or less stable gene expression and the impact of environmental factors on gene expression that can also cause diseases as cancer [5,26,27]. Therefore, these enzymes represent a central role in the regulation of immune responses as alterations in their activity and expression profiles leading to global changes in the histone modification pattern have been detected as cause of several chronic immune diseases as asthma, chronic obstructive pulmonary disease, colitis, systemic lupus erythematosus and rheumatoid arthritis [28,29].

Additionally, the DNA can be methylated at cytosine or adenosine residues converting them into methyl-cytosine or methyl-adenosine to cause transcriptional repression [30]. Hypermethylation dominantly occurs at CpG islands, cytosine-guanine rich regions at promotor regions, disrupting TFs and RNA polymerase binding to DNA or recruiting other co-repressors. A hypermethylated gene, that was not methylated before is therefore, not suitable for transcription and with the recruitment of further silencing-factors, will finally be silenced. This kind of modification is thought to provide a stable gene silencing that can be inherited to the next generation of cells [31,32].

Moving away from the old definition of epigenetics as hereditable stable changes at chromatin and DNA without changing its sequence, modern opinion changed towards a highly dynamic and reversible mechanism of gene regulation also enabling short term adaptions to changing environments [30]. As consequence, regulation through ncRNAs are also included to the epigenetic regulatory repertoire that can be classified according to their length into short ncRNAs (<200 nucleotides), which include miRNAs or long ncRNAs (>200 nucleotides) [33].

#### 2.1.1. Manipulation at the Level of Histone Modifications

After recognition of bacterial presence by PRRs, signaling cascades activate proinflammatory cytokine expression. To improve accessibility of TFs, such as NF-κB, to the promoters of inflammatory response genes, an activating histone modification as phosphorylation of Serine 10 on histone H3 (H3S10) is established, which itself is mediated by MAPK signalling. It has been shown, that the virulence factor LPS alone is able to induce a global increase of H3S10 leading to promotion of gene expression proving the high sensitivity of the immune reaction [34].

Recent studies revealed that bacteria directly interfere with a host's histone modifications to dampen the expression of proinflammatory cytokines by the secretion of effectors. Presence of *Listeria monocytogenes* induces phosphorylation of H3S10 but the bacterium is able to remove this activating phosphorylation within short time [35–38]. The secreted virulence factor Listeriolysin (LLO) mediates this mechanism and is also responsible for a global deacetylation of H3 and H4. Other bacteria, as *Clostridium perfringens* or *Streptococcus pneumoniae*, produce toxins, such as perfringolysin and pneumolysin, respectively, that belong to the same family as LLO and show also a similar effect on H3S10 phosphorylation [36]. The decreased levels of phosphorylated H3S10 and acetylated H4 at proinflammatory genes resulted in transcriptional downregulation thereby damping the immune

response. As this observation is only dependent on the membrane-binding ability of LLO, it is most likely that LLO modulates the signal transduction to induce alterations in the histone modification pattern [36].

Like *L. monocytogenes*, *Shigella flexneri* is also able to inhibit H3S10 phosphorylation by secretion of phosphothreonine lyase effector OspF, which dephosphorylates MAPKs as p38 or ERK resulting in attenuated NF-κB binding at promotors of inflammatory genes [39]. Together with OspB, another effector of *Shigella,* OspF, interacts with the human retinoblastoma protein Rb that is capable of binding several chromatin-remodeling factors [40,41]. In this constellation, *Shigella* adjusts the chromatin structure at specific genes to downregulate host innate immunity.

*L. monocytogenes* owns another effector, which induces deacetylation on lysine 18 of histone H3 (H3K18). Thereby, Internalin B (InlB) activates the host histone deacetylase sirtuin 2 (SIRT 2), leading to repression of transcriptional start sites through occupation by SIRT 2 and following downregulation of the immune response, which could be attenuated by SIRT 2 inhibition [42]. The listerial virulence factor LntA enters the nucleus after infection of epithelial cells targeting the chromatin silencing complex component BAHD1. Together with heterochromatin protein 1 (HP1), methylated DNA-binding protein 1 (MBD1), histone deacetylases (HDAC1/2) and the KRAB-associated protein 1 (KAP1/TRIM28) that are involved in heterochromatin formation, BADH1 targets interferon-stimulated genes (ISG) for silencing by binding to their promotors [43,44]. This is inhibited by LntA, which is thought to promote chromatin-unwinding and as consequence upregulation of ISG by histone H3 acetylation. The exact mechanisms, how BAHD1 is recruited to its targets and how LntA interferes with this process has still to be investigated [21].

Another prominent histone modification is the methylation or demethylation of lysine residues, mediated by histone N-lysine methyltransferase (HKMT) or histone demethylases (HDM), respectively. Several bacteria express HKMT effectors, which enable them to directly interfere with host gene regulation as they are mimics of host chromatin modifiers. As there are many HKMT homologues in the repertoire of bacterial effectors described this mechanism seems to be a successful strategy to subvert host gene expression [45]. The nuclear effector (NUE), is secreted by *Chlamydia trachomatis* via a type III secretion system (T3SS) to enable its localization to the nucleus, where it might methylate H2B, H3 and H4. The homologous effectors RomA and LegAS4 secreted by *Legionella pneumophila Paris* and *L. pneumophila Philadelphia Lp02* strains, respectively, methylate H3 to alter host transcription but target different residues [45,46]. RomA represses global transcription by methylation of histone 3 lysine 14 (H3K14), a modification that is known to compete with the activating acetylation of H3K14 [46]. Contrary to RomA, LegAS4 increases transcription of ribosomal RNA genes (rRNA) through methylation of histone 3 lysine 4 (H3K4) but if this modification is mediated by LegAS4 alone it is not clear yet [45]. Interestingly all described bacterial methyltransferases own a conserved SET (Suppressor of variegation, Enhancer of zeste and Trithorax) domain, which uses a S-adenosyl-l-methionine (SAM) methyl donor to catalyze methyl group attachment to lysine residues [45,47]. One example is the effector BtSET, secreted by *Burkholderia thailandensis* that localizes to the nucleolus to methylate histone H3K4 promoting transcription of rRNA genes. Some effectors are capable of more unusual modifications, for example, the effector BaSET identified in *Bacillus anthracis* trimethylates histone H1 but none of the core histones. This effector represses the expression of NF-κB target genes after transient overexpression in mammalian cells and its deletion results in the loss of virulence [45–47].

Another modification, which differs from the known mechanisms of histone modification, is represented by dimethylation of histone 3 on arginine 42 (H3R42me2), a residue critical for DNA entry/exit from the nucleosome and not located at the histone N-termini. This modification is involved in the regulation of ROS (reactive oxygen species) production, which represents a crucial host defense mechanism against bacterial pathogens [48]. *Mycobacterium tuberculosis* represses genes involved in ROS production by secreting Rv1988, a methytransferase able to establish H3R42me2 to increase survival in host macrophages [48]. An overview of bacteria and their effectors that are secreted to induce histone modifications is given in Table 1.



*Int. J. Mol. Sci.* **2020** , *21*, 3730

Influencing the expression of histone modifying enzymes is another possibility to affect histone modifications in favor of bacterial survival (see Table 2). Modulation of the histone deacetylase HDAC1 appears to be most targeted by pathogens, to manipulate the key acetylation system enabling protection against eradication. Infection with *Anaplasma phagocytophilum*, an intracellular pathogen causing human granulocytic anaplasmosis, causes upregulation of HDAC1 leading to a globally increased HDAC activity [56]. The recruitment of HDAC1 to AT-rich chromatin sites in promotors of host defense genes is mediated by the effector ankyrin A (AnkA) resulting in the reduction of histone H3 acetylation and the suppression of target genes such as *CYBB* that encodes Cytochrome b-245, beta polypeptide. As this element of the phagocyte NADPH oxidase is involved in the clearance of the pathogen by neutrophils, it is preferentially targeted [57–60]. Furthermore, AnkA functionally mimics SATB1, a protein able to bind AT-rich sequences distributed across distinct chromosomes at attachment regions of the nuclear matrix. Proteins with this ability are involved in nuclear matrix attachment, spatial organization of chromatin and large-scale transcriptional regulation [59,61–63]. AnkA could also perform as global organizer of the neutrophil genome, thereby acting locally (*cis*) and at a distance (*trans*) to a target gene. Moreover, pathogens as *Chlamydia psittaci* secrete nucleomodulins (SinC) that could act like AnkA and influence anchoring factors and lamins that control the dynamics of chromatin looping and organization, as the inner nuclear membrane proteins MAN1 and LAP1 [64].

*Pseudomonas aeruginosa*, an opportunistic pathogen that infects and colonizes inflamed airways and burn wounds, induces HDAC1 expression in human THP-1 monocytes with the help of a molecule usually used for quorum sensing, 2-aminoacetophenone [65,66]. This is followed by global histone H3K18 hypoacetylation and reduced expression of inflammatory cytokines and chemokines (e.g., TNF, IL-1b and MCP-1) resulting in dampened host defense against the bacterium.

Considering that this effect was also dampened by knockdown or inhibition of class I HDACs and the evidence that besides *A. phagocytophilum* and *P. aeruginosa* also *Porphyromonas gingivalis* modulates HDAC1 during infections, HDAC1 family members might play a central role in development of an epigenetic mediated tolerance against the pathogens [67]. In patients with chronic periodontitis, mRNA and protein levels of HDAC1 expression were globally increased compared to healthy individuals and colocalized with TNF expressing cells and tissues. Interestingly, epigenetic regulation mediated by *P. gingivalis* seems to be cell-type specific, since HDAC1 and HDAC2 are downregulated in gingival epithelial cells *in vitro*, while levels of acetylated histone H3 were increased in murine epithelial cells of the gingival tissue [68,69]. In addition, the host acetylation system is also often influenced by short chain fatty acids (SCAFs) produced by commensals or pathogenic bacteria as *P. gingivalis* (for recent reviews please refer to References [70] and [71]).


**Table 2.** Bacteria targeting histone modifying enzymes.

Another strategy followed by bacteria during host infection and manipulation of the epigenetic regulatory mechanisms is to proteolytically degrade histone acetyl transferase (HAT) family members. One example of bacteria using this strategy are enteropathogenic and enterohaemorrhagic *Escherichia coli*, which secrete the effector protein NleC, a zinc-dependent metalloproteinase targeting intracellular signaling to dampen the host inflammatory response [76]. The protein specifically binds and degrades the host HAT p300 in infected cells leading to decreased IL-8 production, an effect that can be restored by p300 overexpression. Thus, HATs and HDACs can both be targeted by pathogenic bacteria to modulate epigenetics and inflammatory gene expression in their benefit.

#### 2.1.2. How to Control Host DNA Methylation

DNA methylation is another way to control gene expression. There are several enzymes called DNA-(cytosine C5)-methyltransferases (DNMTs), which establish methyl residues to cytosine or adenosine residues, respectively [71]. In contrast, the removement of DNA methylation patterns is more complex, as the modified nucleotides or DNA sequences have to be exchanged by DNA-repair mechanisms or the methylation has to be oxidized to form 5-Hydroxymethylcytosine, which can be removed by enzymes [77]. DNA methylation patterns at promotors of tumor suppressor genes had already been discovered, when first hints pointed towards an influence of bacterial inflammation on mechanisms establishing DNA-methylation patterns after *Helicobacter pylori* infection. In this context, among others, genes associated with cell growth (*apc, p14 (ARF), p16 (INK4a)*), cell adherence (*cdh1, flnc, hand1, lox, hrasis, thbd, p14ARC*) and DNA-repair (*brca1, mgmt., hMLH1*) are influenced [52,78–80]. Similar observations of altered DNA-methylation patterns during inflammation were made following uropathogenic *E. coli, Campylobacter rectus* and *Mycobacterium leprae* infections [81–83]. Still, the questions if DNA-methylation is directly induced by bacteria or is a secondary reaction by the host due to persistent inflammations, as well as the underlying mechanisms, are not completely answered yet [84].

However, several *Mycoplasma* species to encode mammalian DNMTs like equivalents that target cytosine-phosphate-guanine (CpG) dinucleotides to establish methylation patterns in the bacterial genome [85–87]. Moreover, their expression in human cells results in their translocation to the nucleus, where they set up unusual methylation patterns on the host DNA. This was shown for the DNMTs Mhy1, Mhy2 and Mhy3 expressed by *Mycoplasma hyorhinis* in combination with up- and downregulation of certain genes resulting in activation of proliferation specific pathways, a process that might contribute to tumor progression [85,88].

*Mycobacterium tuberculosis* owns an effector called Rv2699 that can enter the nucleus of THP1 cells (a monocytic cell line derived from a patient with acute monocytic leukemia) and methylate cytosines outside CpG dinucleotides. Notably, Rv2699 prefers cytosine-phosphate-adenine or cytosine-phosphate-thymine sites to generate a type of methylation that is, with few exceptions, normally not present in mammalian adult differentiated cells [89,90]. However, non-CpG methylation could lead to a more stable type of modification that persists longer in the genome of infected nondividing macrophages, offering an advantage for *M tuberculosis* by establishing an intracellular environment for persistence [90]. A follow up study revealed that THP1 macrophages infected with *M. tuberculosis* strain H37Rv created genome-wide *de novo* methylation patterns at non-CpG dinucleotides that included hyper- and hypomethylated regions [90,91]. Additionally, clinical isolates infecting THP1 cells may downregulate IL-6 receptor expression by hypermethylation of CpG-dinucleotides at the promoter of the IL-6 receptor gene. Still, it has to be mentioned, that the observations of *M. tuberculosis* induced DNA-methylation patterns depend on the infected cell type.

Another interesting bacterial induced modification of gene expression is represented by differentiated Swann cells that adapt the phenotype of progenitor stem-like cells after *M. leprae* infection. This is probably induced by silencing of the *Sox10* gene after bacterial methylation [82]. In contrast to the decreased expression of Sox10, other genes involved in epithelial–mesenchymal transition (EMT) were demethylated and transcribed leading to the transformation of Swann cells into myofibers or smooth muscles in vitro and in vivo [92].

*P. gingivalis* was shown to increase the methylation of the TLR-2 promotor in gingival epithelial cells (GECs) reducing innate immunity activation and causing hyposensibility [69,93]. Besides, coinfection with *Filifactor alocis*, another pathogen associated with periodontitis is suggested to influence the whole cell transcriptome through impact on the nucleosome structure by reduced expression of H1 family members [73,74]. Other histone modifications induced by LPS or short chain fatty acids (SCFAs) produced by *P. gingivalis* are summarized in Tables 1 and 2.

Still, there is not much known about the relation of DNA-methylation and infection and the underlying causalities [71,84]. Considering that many of these modifications are observed in the context of cancer initiation and progression, further investigation may contribute to new therapeutic agents and cancer prophylaxis.

#### 2.1.3. Regulation of Host Gene Expression via lncRNAs

The role of lncRNAs during modulation of gene expression has been discovered in the recent years. Similar to mRNAs, lncRNAs are transcribed by RNA polymerase II or III, followed by splicing, 5 capping and in some cases polyadenylation at the 3 end. Contrary to mRNAs, the expression of lncRNAs is much lower and in a cell-, tissue- and developmental stage-specific manner [94].

Dependent on of their position relative to the neighboring protein-coding gene, lncRNAs are classified as sense, antisense, bidirectional, intronic or intergenic and, despite their enormous number, they were previously considered as "dark matter" or "junk" in the genome [95]. *Au contraire*, lncRNAs are now respected as important physiological regulators during cell homeostasis, growth, differentiation and anti-viral responses [96–99]. In addition, gene imprinting, regulation of the p53 pathway, stem cell self-renewal and differentiation and DNA damage response were reported as lncRNA controlled mechanisms [100–103].

The functionality of lncRNAs is not restricted to the neighbored protein-coding gene (*in cis*), in contrast they are also able to act *in trans* to regulate gene expression across chromosomes. In this context, lncRNAs regulate different processes as chromatin remodeling, transcription and post-transcriptional regulation via their capacity as signals, decoys, guides and scaffolds [104,105]. Interestingly, another origin of lncRNAs is the expression of pseudogenes and gaining Influence over the expression of pseudogenes could, therefore, provide a possibility to control infectious responses [106].

Immune regulation through lncRNAs has already been known after viral infections but recent research indicates its involvement also whilst fighting bacteria [107]. In that context, 76 enhancer RNAs (eRNAs), 40 canonical lncRNAs, 65 antisense lncRNAs and 35 regions of bidirectional transcription are differentially expressed in human monocytes after LPS stimulation [108]. LPS stimulation alone induces a differential expression of about 27 lncRNAs leading to histone trimethylation or acetylation of neighboring genes after de-regulation, pointing towards their regulatory influence during the innate-immune response [109]. The observation, that 44% of total lncRNAs varied in their expression after *Salmonella* infection in HeLa cells could foster these results and substantiate them by a function in the early phase of infection as sensitive markers for pathogen activity [110]. In line with this, the lncRNA HOTAIR that contributes to transcriptional repression of HOX genes also promotes inflammation in mice cardiomyocytes by TNF-α production mediated through phosphorylation of p65 protein and NF-κB activation after LPS induced sepsis [111,112].

Long intergenic non-coding RNAs (lincRNAs) are a subtype of lncRNAs, as they are expressed from intergenic regions. In response to an LPS stimulus, bone-marrow dendritic cells expressed about 20 lincRNAs with the majority being dependent on NF-κB activity, including lincRNA-Cox2, which is also upregulated in bone marrow-derived macrophages following *L. monocytognes* infection [113,114]. Additionally, bacteria sabotage lncRNA activity, as BCG (attenuated strain *M. bovis* bacillus Calmette-Guérin BCG) infected macrophages repress the expression of 11 lncRNAs that are not dampened by infection with heat activated bacteria [115]. Still, possible subversion of lncRNA-mediated inflammatory regulation needs to be further investigated.

#### *2.2. Bacterial E*ff*ectors Manipulating the Host Transcription Machinery*

Proper RNA Pol II complex formation is essential for protein expression and tightly controlled by regulators, who are expressed by approximately 10% of all genes [116]. These are general or specific TFs, which serve as activators and repressors and determine specificity and efficiency of transcription at individual promotors [117,118]. Considering the large number of factors involved in transcriptional regulation, it is not surprising that bacteria target those regulators to drive transcription in their favor. For example, activator protein-1 (AP-1)–dependent gene transcription is inhibited by NleD, an effector of *E. coli* that cleaves and inactivates the MAPKs, JNK and p38. Also, the recently identified nucleomodulin OrfX secreted by *L. monocytogenes* that influences host transcription via its interaction with the Ring1 YY1-binding protein (RYBP), a multifunctional nuclear protein owning a zinc finger motif to interact with several TF components of the polycomb repressive complex 1 [119,120]. Moreover, RYBP promotes gene silencing and transcriptional repression of developmental genes, as it is part of the BCL6 corepressor (BCOR) complex [121,122]. In contrast, it is also involved in the activation of the Cdc6 promoter and mediates interaction of the TFs E2F and YY1 [123]. Furthermore, the TF p53 (a tumor suppression factor) is assumed to be stabilized by RYBP through binding of MDM2, an E3 ligase, preventing p53 from proteosomal degradation. As p53 controls intracellular levels of reactive oxygen (ROS) and nitrogen species (RNS) that are part of the immune defense of macrophages, OrfX targets RYBP for degradation to interrupt P53 activity promoting intracellular bacterial survival. Still, this model needs to be verified [120,124].

*Salmonella* Typhimurium inhibits the expression of NF-κB mediated genes by secretion of PipA, GogA and GtgA via its type II secretion system [125–128]. These proteins belong to the family of zinc metalloproteases and contain the short metal binding-motif HE*XX*H, which consists of two histidine residues coordinating the active-site zinc and a glutamate residue that is essential for catalytic activity [129]. They can cleave NF-κB TF subunits, including p65, RelB and cRel, thereby suppressing their ability to control the transcription of innate immune genes [125–128]. In addition, Jennings et al. predicted suppression of the transcriptional coactivator ribosomal protein S3 (RPS3) by GtgA family members, as it produces p65 (1–40) after cleavage of p65.

Enteropathogenic and enterohemorrhagic *E. coli* possess another zinc metalloprotease that also owns the HE*XX*H motif and is able to cleave p65, RelB and cRel, as well as NF-κB1 (p105/p50) and NF-κB2 (p100/p52) [126–128,130]. Following cleavage, the subunits are left inactive except for the N-terminus of p65, which prevents the nuclear import of the transcriptional coactivator ribosomal protein S3 (RPS3). This in turn inhibits the expression of a specific subset of NF-κB–dependent genes requiring RPS3 for their expression.

The non-pathogenic *E. coli* strain 83972, who is the agent causing persistent asymptomatic bacteriuria (ABU), suppresses host defense in the urinary tract by inhibition of RNA Pol II dependent transcription. While infections with pathogenic strains induce urinary tract infections, these bacteria create an asymptomatic carrier state that reminds of bacterial commensalism and protects patients against infection with more virulent strains. Therefore, therapeutic urinary tract inoculation with the ABU strain is a promising alternative to appease symptoms of therapy resistant, recurrent urinary tract infections [131–133].

Studies with patients and human cells treated with ABU strain 83972 revealed that 24 h after inoculation, over 60% of all genes were suppressed, including regulatory elements as transcriptional repressors, transcriptional activators, regulators of translation and chromatin or DNA organizing factors [134,135]. This phenomenon was observed for many genes of the innate immune response but about 22.5% of the effected genes are involved in Pol II transcription or in regulating Pol II–dependent pathways. After Ingenuity Pathway Analysis, a network incorporating *FOSB*, *HSPA6*, *RN7SK*, *RGS4* and *IFIT1,* inversely regulated genes that control Pol II for instance through TATA box–binding proteins (TBP) [136–138], appeared.

Another fascinating observation revealed that 50% of ABU strains lack virulence genes due to point mutations or deletions resulting in smaller genome sizes. Considering that ABU strains evolved from uropathogenic *E. coli*, this could be a hint for a reductive evolution creating a niche through active adaptation to the host environment [139]. Thereby, the ABU strain generates a commensal like state characterized by a well-balanced immune environment that finally protects the host from colonialization with more virulent strains and destructive immune activation [134,140–144].

Ambite et al. observed in a follow up study that obtaining one single virulence factor was enough to induce virulence of a non-virulent strain causing symptoms in the host, in contrast to the broad repertoire of virulence factors that are normally expressed by pathogens [145]. In this context, reconstitution of the *papG* adhesin gene recreated functional P-fimbriae leading to virulence of the avirulent ABU strain. Considering the high frequency of ABU strains carrying inactive papG genes, the loss of P-fimbriae might induce development of virulence attenuation and evolution towards commensalism [96,134].

*L. pneumophila* is another pathogen that induces global reprogramming of transcription, by interference with transcriptional elongation by Pol II. Its effector AnkH interacts with LARP7, a component of the 7SK small nuclear ribonucleoprotein (snRNP) complex involved in Pol II pausing. Thereby, theβ-hairpin loop of the third ankyrin repeat of AnkH impairs LARP7 interaction with the other 7SK snRNP complex components resulting in promotion of gene wide transcriptional elongation [146]. The nucleomodulin SnlP expressed by *Legionella* also regulates RNA Pol II mediated transcription elongation by inhibition of SUPT5H that is part of the 5,6-Dichlorobenzimidazole 1-β-D-ribofuranoside (DRB, a selective inhibitor of transcriptional elongation by RNA pol II) sensitivity-inducing factor (DSIF) complex [147].

#### *2.3. RNA Processing as Target during Infections*

Following transcription the immature pre-mRNA is processed to mature mRNA, a process that includes 5 capping, 3 polyadenylation and splicing and is essential for normal cell function [148]. Splicing of mRNAs can include omitting or retaining of exons to create a protein with altered structure, function and stability, a process called alternative splicing. In the human genome, more than 90% of all genes are adjusted by alternative splicing, which enables a variation and dynamism in the static genome as protein domains can be easily new combined to form isoforms with unique functional abilities [148,149]. The process is controlled by multi-molecular complexes that assemble at splice junctions, thereby evaluating splicing enhancer/silencer elements flanking splice junctions, which in their combination determine inclusion or exclusion of exons. Those elements are divided into *cis* elements including splicing enhancers and silencers and *trans* elements as snRNPs, hnRNPs, SR proteins (serine-arginine containing proteins) and several other accessory proteins [149,150]. Furthermore, the rate of transcription has a critical influence on alternative splicing, as a paused or decelerated RNA Pol II can use newly transcribed splice junctions that could have been skipped at higher translation speed [151–153].

Recent studies indicate that the host splicing machinery is targeted by pathogens to perturbate immune response. This has been extensively reported for viral infections but quite less is known about bacterial interference. Nevertheless, global alterations of splicing patterns were detected after infection of human macrophages with *M. tuberculosis, Salmonella* or *Listeria* [154–157]. More specifically, hnRNP M interacts with LLO leading to a hampered INF-γ response [158] and co-immunoprecipitation of splicing factors hnRNP U, hnRNP H, hnRNP A2/B1 isoform A2 and SRSF3 with the bacterial protein mtrA was shown in macrophages overexpressing specific secreted proteins that are infected with *M. tuberculosis* [159]. Another hint for the interaction of mycobacterial proteins and host splicing factors is the precipitation of host splicing proteins as SRSF2, SRRM2, SF1, HTATSF1, GCN1L1, CPSF6 and many more by the mycobacterial proteins EsxQ, Apa, Rv1827, LpqN, Rv2074 and Rv1816 [160].

*Mycobacterium avium* subsp. *paratuberculosis*, also induces alternative splicing of 46.2% of all genes, including two genes responsible for monocyte to macrophage differentiation-associated maturation and lysosome function. Since their splice variants lead to failure of macrophage maturation, bacterial intracellular persistence is improved in the early phase of infection by hampered clearance [161]. Furthermore, alternative splicing of RAB8B, a key regulator of phagosome maturation, induced by *M. tuberculosis* infected cells leads to the production of a truncated protein. The alternative splicing event results in nonsense-mediated decay of RAB8B mRNA resulting in lowered protein levels, that dampens phagosome maturation [156].

Analysis of RNAseq data revealed that specific genes are chosen by pathogens for the manipulation of alternative splicing. Indeed, most dominant isoforms of protein kinases produced end up with the loss of critical functional domains including kinase domain or protein–protein interaction domains like SH2, SH3 and PH domains [162]. Considering nonsense-mediated decay of RAB8B mRNA after *M. tuberculosis* infection, it is concluded that this mechanism describes the two strategies host and bacteria developed during their evolutionary concurrence [156]. In this theory, increase of transcription after infection represents the host response, whereas splicing into a truncated isoform, which is destinated for decay, exemplifies bacterial interference. The exact mechanism how bacteria manipulate alternative splicing is not clear yet. Possibly, they activate cryptic or weak splice sites in the host genome to alter the splicing pattern but this has still to be proven [149].

However, another aspect that needs to be considered is the high diversity with that individuals react to the same pathogenic agent. For example, only 5–10% of the individuals in tuberculosis endemic countries that had contact with *M. tuberculosis* develop disease, whereas the majority either eliminates the pathogen or controls it in a metabolically altered latent phase [163,164]. An explanation are single-nucleotide polymorphisms (SNPs) that disrupt splice-site consensus sequences in 15% of human disorders induced by inherited point mutations, whose influence induce strongly fluctuating pathological conditions after varying activation of disease associated genes [165–167]. This was already reported for diseases as diabetes and seems to be also true for infections, as several SNPs were identified that change the host susceptibility to *M. tuberculosis* infections for example, in the intronic region of human ASAP1 (dendritic cell migration). Another polymorphism in IL-7RA helps to protect against tuberculosis due to an impaired IL-7Ra splicing [168,169]. Therefore, alternative splicing gets into focus of possible medical therapy developing splicing inhibitors that are already tested for cancer [149,170,171]. Nevertheless, the knowledge about alternative splicing during bacterial infections and their interplay is very limited and deserves more attention, as this could give more insights in individual susceptibility and immunity.

#### *2.4. The Advantage of Modulating Host RNA Stability and Degradation*

The lifetime of mRNAs has a major impact on the amount of proteins that can be produced; the shorter an mRNA is present, the less it can be transcribed. Since the lifetime of an mRNA is dependent on its stability, there are mechanisms to increase resistance to degrading RNases [11–13]. First, the 5 m7G cap and the 3 poly-A tail that are established post transcriptionally at all mRNAs but especially the length of the poly A tail can vary between mRNAs, determining the duration of resistance against enzymatic degradation [9]. Besides, these structures are involved in virus clearance, as viral mRNAs lack these structures, what marks them for RNA degradation machinery [172]. In addition, the structure of the mRNA alone influences its stability, as hairpin-structures and other secondary structures are formed dependent on the sequence and therefore, increase the stability [11–13]. In the following paragraphs we want to highlight the regulation of mRNA stability and decay mediated by miRNAs during infections and how bacteria interfere with this part of the host immune defense.

#### Manipulation of miRNAs to Favor Bacterial Survival

Their physiological properties enable non-coding RNAs (ncRNAs) to base-pair with their targets and interfere with a twofold effect on gene expression—one single ncRNA can bind multiple targets, thereby influencing several pathways at once and one gene can be regulated by several ncRNA fine tuning gene expression [173]. miRNAs are involved in several cellular processes as cell proliferation or differentiation and after studies with human monocytes treated with LPS, miR-146 was identified as anti-inflammatory miRNA proving miRNA involvement in inflammation [174–176]. Indeed, subsequent studies revealed specific expression of miRNA sets including miR-155, miR-146, let-7 and miR-29 (see Table 3) due to infection with different bacterial pathogens regulated in a time dependent manner [177–181]. Another study with dendritic cells infected with six different bacteria demonstrated a core infectious response in a temporal manner including 49 miRNAs that were always expressed and may, therefore, play essential roles in infectious responses. Additionally expressed miRNAs might hint towards specific variability and signatures arising from the individual pathogens [182]. Interestingly, following infection, the proportion of miRNA variants, the so called isomiRs, varies, which is supposed to effect miRNA identity and regulatory potential but has not been proven yet [182].

The induction of miRNAs is often dependent on PRR and NF-κB pathway induction in response to bacterial stimuli as LPS. Interestingly, there is a dose dependent reaction to the stimulus, as a low dose activates miR-146 that acts as anti-inflammatory regulator promoting tolerance to low doses by targeting two members of the NF-κB pathway, TRAF6 (TNF Receptor-associated factor 6) and IRAK1 (IL-1R-associated kinase 1) [176,183]. In contrast, at high doses of LPS, TNF-α and Interferon β induce miR-155 via TAB2 to maintain the proinflammatory NF-κB activity fighting pathogens and exerting a negative feedback on the immune system. Therefore, both mi-RNAs protect the host from sepsis and overreaction [183,184] but in a dose dependant manner. miR-155 is also involved in T helper cell development or promoting autophagy by inhibition of the mTOR (mammalian/mechanistic target of rapamycin) pathway, indicating that it represents an important part of an efficient immune response [185–187]. Actually, upon *Citrobacter rodentium* or *L. monocytogenes* infection, miR-155 null mice showed slower clearance and an impaired CD8<sup>+</sup> T-cell response, respectively and miR-155 was identified as an essential factor during the vaccination process against *S.* Typhimurium [188–191].


**Table 3.** Activity of miRNAs in the host response. The arrows indicate changes of miRNA expression induced after bacterial infection that result in the described alterations..

Another miRNA family, the let-7 family, is repressed during infection or exposure to LPS, as Lin-28B expression is induced in a NF-κB dependent manner that blocks let-7 maturation [154,192,193,195,199,200]. Additionally, active repression of these miRNAs by bacteria has been reported in several studies (see Table 3). Many other bacteria induce miRNA expression and manipulate expression of these immune regulators in their favor (summarized in Table 4.)


In addition to *H. pylori* with miRNAs (see Table 4), there are two more interactions described, which depend on the effector CagA that activates NF-κB pathway. This effector induces an increased expression of miR-1289 that in turn leads to a decreased gastric acidity, as miR-1289 targets HKα, a component of the gastric H+/K<sup>+</sup> ATPase [213]. Furthermore, Cag A induces cell cycle arrest at G1/S transition, which inhibits the renewal of the gastric epithelium, supporting *H. pylori* colonization [214]. In this context, miR372 and miR-373 expression is suppressed by Cag A, whereas miR-584 and miR-1290 expression is promoted. The latter target FOXA1, a negative regulator of the epithelial-mesenchymal transition, for inhibition, thereby favoring bacterial persistence and survival within the gastric epithelium [215]. Moreover, CagA suppresses miR320 and miR370 expression, who induce MCL1 (an anti-apoptotic gene) or downregulate FoxM1 expression, respectively. FoxM1 downregulation in turn activates p27K1P1 leading to cell cycle inhibition. Together, these factors decrease apoptosis and favor cell proliferation, which can lead to tumor development.

The *M. tuberculosis* effector ESAT-6 is also capable of manipulating miRNA expression to the benefit of the bacterium [212]. ESAT-6 downregulates let-7f expression in macrophages, leading to an enrichment of the deubiquitinating enzyme A20 that negatively regulates the NF-κB pathway. Furthermore, miR-155 expression is stimulated, resulting in BACH1 (a transcription regulator protein) and SHIP1 (SH-2 containing inositol 5 polyphosphatase 1, a multifunctional protein expressed predominantly by hematopoietic cells) repression that in turn induces heme oxygenase 1 expression. Concurrently, serine/threonine kinase AKT is activated fostering bacterial dormancy and survival. This is subsidized by downregulation of SOCS1 (suppressor of cytokine signaling 1) and targeting of Rheb (Ras homolog enriched in brain), which is followed by macrophage apoptosis [199,210,211].

Evidence exists that bacteria are able to also produce their own regulatory RNAs that interfere with the host. In 28 bacterial genomes 68 possible candidate bacterial RNAs were found during an in silico search, which harbor secondary structures that could form miRNAs after host procession and be involved in 47 human diseases [216]. As an example, after *E. coli* ingestion *che-2* and *F42G9.6* gene expression was modulated and probably degraded in *Caenorhabditis elegans* by *E. coli* OxyS and DsrA ncRNAs [217]. Furthermore, *Mycobacterium marinum* expresses a pre-miRNA that associates with the host RISC complex in its mature, 23 nucleotide long form and could effectively downregulate its target mRNA when overexpressed [218].

It has been reported that exosomal transfer of host cell miRNAs is used to spread the host response to other cells and the ratios of miRNAs transported differ in a time and bacterial dependent manner. Hence, the ratios of miR-146a and miR-155 in exosomes can subsequently modulate host cell responses, favoring inflammation or recovery, respectively. The use of exosomes containing miRNAs could give rise for therapeutic possibilities to treat inflammation or to be used during vaccinations [219–226]. Moreover, exosomes containing miRNAs could be used for diagnosis, since they can be detected in many sample types (blood fluids, saliva, tears, urine, amniotic fluid, colostrum, breast milk, stool, etc.) and since there are unique patterns for each pathogen [219–221].

Regulation of gene expression through ncRNAs as miRNAs happens more immediately and flexibly than through transcriptional regulators [33]. The faster response is enabled by the cell and tissue type specific differentially regulated reservoir of ncRNAs, which also allows a precise fine-tuning of gene expression to organize immune defense and damage protection [33]. Taken together, investigation of the host-pathogen crosstalk with a focus on miRNAs and their usage and manipulation by bacteria provides new perspectives to fight bacterial mediated diseases.

#### *2.5. Controlling Host Translation Improves Bacterial Persistance*

Translation is a major regulator of gene expression and immune response. As many factors are needed to induce Ribosome association, start of translation and ongoing elongation, many ways exist to regulate or interfere with the translation machinery. Not surprising, inhibition of translation is a well-known strategy followed by bacteria to circumvent immune defense [227].

In most cases, eukaryotic translation is controlled during initiation, when ribosomes are recruited to the mRNA mediated by eukaryotic initiation factor 4F (eIF4F) that recognizes the 5 cap structure with the help of its cap-binding subunit eIF4E [228]. The reversible association of this subunit with 4E-binding proteins (4E-BPs) inhibits the assembly of eIF4F and its release and activation are in turn mediated by the phosphorylation of the 4E-BPs [229–231]. This phosphorylation is induced by the serine/threonine kinase mTOR complex 1 (mTORC1), thereby requiring the protein Raptor for mTOR substrate binding whereas rapamycin binding inhibits phosphorylation and dissociation [232,233].

*Legionella pneumophila* (*L. pneumophila*) expresses five effectors, Lgt1, Lgt2, Lgt3, SidI and SidL, involved in global protein translation inhibition by interference with the eukaryotic elongation factors eELF1A and eELF1Bγ [227,234–236]. Moreover, *L. pneumophila* was also shown to negatively influence cap-dependent translation initiation mediated by ubiquitination of the mTOR pathway leading to suppression of the eukaryotic initiation factor 4E (eIF4E) through decreased eIF4E assembly into the translation initiation complex eIF4F [237].

Finally, the synthesis of IκB, an inhibitor of the NF-κB TF, is inhibited by *L. pneumophila.* This leads to a prolonged NF-κB activation resulting in the so-called effector-triggered response (ETR) including transcription of target genes, such as *Il23a* and *Csf2* that create a more pro-inflammatory state. Fascinatingly, mutants lacking effectors or the Dot/Icm type IV secretion system transferring them, still inhibit host translation via TLRs and NF-κB activation but not sufficient enough to fully induce ETR [234,235]. In addition, macrophages lacking TLR and Nod signaling still mediated MAPK signaling after exposure to the five *L. pneumophila* effectors that leads to inhibition of host translation. Therefore, translational inhibition does not exclusively rely on ETR but also on effector independent mechanisms that induce mTOR pathway downregulation and cytokine biasing [237]. In this context it is suggested that the host immune system senses not only for PAMPs but also for pathogen-encoded enzymatic activities that disrupt crucial cellular processes [227]. Interestingly, even if host translation is inhibited at the stage of initiation and elongation by *L. pneumophila*, there is still an inflammatory response detectable. The immune response is quite weak compared to its normal potential but few inflammatory cytokines, as IL-1α and IL-1β, circumvent translational inhibition by *L. pneumophila*, which is mediated by MyD88 signaling [227]. This demonstrates that the interference with cap-dependent host translation results in promotion of host defense, as highly abundant transcripts, which often encode proinflammatory cytokines, are favored for translation. Thereby, the bacterial benefit through blockage of host translation may consist of increased availability of amino acid nutrients beneath the dampened immune response, which is impeded by the host [234].

Translational inhibition is also known for *P. aeruginosa* infections in *C. elegans,* where Exotoxin A after its endocytosis into intestinal cells leads to suppression of elongation factor 2 (EF2), followed by selective translation of ZIP-2 and thus, induction of pathogen clearance [238,239]. As the 5 UTR of zip-2 contains several untranslated ORFs (uORFs), it was proposed that the uORFs could mediate selective translation. Moreover, inhibition of translation by pharmacological inhibitors also caused induction of various stress response genes including *Il6*, *Il23*, *Il1*α and *Il1*β [240–242]. As these cytokines are still expressed when the elongation machinery is attacked by *P. aeruginosa* with Exotoxin A, a consideration of similar functionalities of the 5 UTR or the 3 UTR of cytokine genes to uORFs was raised. However, the ADP-ribosylation of elongation factor 2 (EF2) in host cells also triggered a strong immune response that is supposed to be the result of a conserved surveillance mechanism in response to inhibition of the translation elongation machinery [236,243]. Thus, a set of elongation factors can be considered, that are resistant to modification by these effectors or are at least not targeted. Therefore, future research is needed to get more information about the underlying mechanisms and potential therapeutic targets [227].

#### *2.6. Modification of Protein Degradation*/*Activity as Last Possibility to Evade Host Immune Defense?*

#### 2.6.1. Mechanisms to Interfere with Protein Degradation

To maintain cellular homeostasis, the quality of synthesized proteins must be controlled for a proper folding and products with quality issues must be degraded in a controlled manner. This quality control is taking place in the cytosol or the ER lumen, where chaperones and heat shock proteins catalyze and stabilize the protein folding [244]. In the case of physiological stress caused by DNA damage, chemical stimuli or pathogens, the ratio of misfolded or unfolded proteins in the ER lumen increases, causing additional stress. Then, the unfolded protein response (UPR), an evolutionary conserved signaling network, is activated resulting in downregulation of overall protein synthesis, except for chaperones and induction of ER associated protein degradation (ERAD [245]. The main kinases controlling the UPR, the inositol-requiring protein 1 (IRE1), PKR like ER kinase (PERK) and activating TF 6 (ATF6) are located inside the ER membranes. Their luminal domains bind the ER chaperone immunoglobulin binding protein (BiP) in unstressed conditions but in case of stress, BiP is released causing activation of the receptor proteins (IRE, PERKI and ATF6) [246]. Following BiP dissociation, oligomerization and autophosphorylation, the cytosolic RNase domain of IRE1 is activated targeting X-box-binding protein 1 mRNA (XBP1u) and transfers it into its spliced form (XPB1s). This enables transcription of the TF responsible for upregulation of UPR target genes fostering ERAD and enhances overall ER protein folding capacity [240–242]. The ER transmembrane kinase PERK1 also oligomerizes and autophosphorylates after activation, catalyzing the phosphorylation of the α-subunit of the eukaryotic initiation factor 2 (eIF2). This is followed by downregulation of global mRNA translation to reduce ER stress but favoring translation of some mRNAs as ATF4. ATF4 in turn induces several UPR target genes including C/EBP homologous protein (CHOP) [247,248]. In contrast to IRE1 and PERK, ATF6 translocates to the Golgi followed by its activation after proteolytical cleavage and activation of the b-ZIP TF to induce UPR target genes.

Since permanent ER-stress, which cannot be solved by UPR and ERAD, will finally induce apoptosis, intracellular bacteria have evolved strategies to interfere with those pathways [249–251]. Surprisingly, induction of UPR pathways can also promote bacterial replication, as bacterial effectors have been detected that induce UPR [19]. As result, it is difficult to refer an upregulation of UPR to bacteria using the increased ER folding capacity for their benefit or to the defense system of the host. In the case of *L. monocytogenes* infection, the effector LLO activates all three UPR pathways leading to induction of ER stress and reduction of bacterial survival. In contrast, the pharmacological block of UPR during infection reduced the intracellular replication of *Brucella melitensis* and *Brucella abortus* significantly [252].

*B. melitensis* and *B. abortus* both induce the IRE1 branch, a process often mediated by TLRs. The TLR adapter protein myeloid differentiation primary response gene 88 (MyD88) than mediates XBP1u splicing. Interestingly, the bacterial effector protein TcpB, secreted by *B. melitensis*, is able to induce the UPR target genes BiP, CHOP and ER DnaJ-like 4 (ERdj4) in a MyD88 independent manner; instead, the TcpB protein itself is required for UPR induction [253].

The induction of IRE1 by *B. abortus* is mediated by the secreted effector VceC after binding of BiP inside the ER lumen. This is followed by IRE1 dependent activation of Nod1/Nod2 innate immune signaling resulting in proinflammatory cytokine expression [254,255]. The ectopic expression of VceC leads to the structural reorganization of the ER and IL-6 production is stalled after infection with *B. abortus* vceC mutants unable to induce UPR. In mice infected with *B. abortus* vceC mutants necrosis was reduced and survival of the pups was increased [255]. Following infection with *B. abortus* WT, similar effects were observed after treatment with the general UPR inhibitor tauroursodeoxycholic acid, leading to the assumption that pharmacological UPR inhibition could be a promising treatment of *B. abortus* infections.

Several other bacteria are known to inhibit the UPR pathway, as it represents a major host defense mechanism involved in bacterial sensing mediated by TLR signaling. Some examples are summed up in Table 5 but unfortunately, the underlying mechanisms are rarely understood. In some cases, as for *L. pneumophila,* it is known that the observed downregulation of UPR is effector dependent, as mutants lacking functional T4SS were unable to induce those changes but the effectors and its targets have not been identified yet [256,257].


**Table 5.** Inhibition of unfolded protein response (UPR) by bacteria.

Taken together, bacteria follow different strategies of interference with protein folding to ensure their intracellular survival. Manipulation of UPR is one strategy to achieve the best outcome for bacteria by either activation or inhibition of UPR. Activation may be induced to take advantage from increased protein folding capacity and lipid biosynthesis by host cells, whereas UPR blockage should hamper host defense, such as apoptosis or innate immunity [261]. In this context, further investigations are needed to understand the underlying mechanisms, how bacterial pathogens manipulate the UPR and which strategy is favored by the individual pathogens in a spatially and temporally manner.

#### 2.6.2. Control of Protein Activity enables Bacteria to Direct Host Immune Reaction

As already mentioned above, the final level to regulate gene expression, for example, to avoid prolonged inflammatory response, is to control the activity of proteins followed by their degradation. The mechanisms to influence protein activity or to mark a protein for degradation are mediated via attachment of functional groups. To ensure a precise signal transduction, proteins are activated in most cases by addition of phosphate groups, which must be removed when the inducing stimulus is ending. The dephosphorylation of MAPKs and the resulting interruption of host signaling cascades, leading to reduced inflammation and an increase of bacterial replication inside the host, is a common bacterial strategy, for example, used by *S.* Typhimurium by secretion of the effector SpvC, a phosphothreonine lyase [262].

In addition, protein activity, subcellular localization and stability is not only regulated by ubiquitination or phosphorylation but also by reversible acetylation which is proposed for approximately 1700 proteins [263]. These include TFs, structural proteins and signal transduction regulators, indicating that reversible acetylation is critical for cell homeostasis [264]. As there are many examples for this kind of activation of proteins (e.g., histones) mentioned in the chapters above, the focus will here lie on the description of the signaling mechanism by ubiquitination. An ATP-dependent enzymatic cascade establishes covalent ubiquitin attachment to proteins, mediated by enzymes that activate ubiquitin (E1), conjugate ubiquitin (E2) and ligate ubiquitin (E3). Ubiquitin can be ubiquitinated at seven distinct lysine residues and is able to form linear or branched chains; the degree and the linkage determine, whether the substrate is degraded or associated with cell signaling [265]. The established ubiquitin modifications can be removed and modified by deubiquitinases (DUBs) to change the linkage pattern and, as consequence, the destination of the substrate.

Intracellular bacteria are able to mimic enzymes involved in ubiquitination processes, especially DUBs and E3 ubiquitin ligases, to modulate the ubiquitin pathway [266,267]. The SidE effector family (SidE, SdeA, SdeB and SdeC), secreted by intracellular bacteria as *L. pneumophila*, possesses domains conferring multiple enzymatic functions used for ubiquitination into a single effector without the requirement of ATP. Thereby, the mono-ADP-ribosyltransferase domain of SidE family members modifies host ubiquitin posttranslationally by attachment of phosphoribose on arginine 42 to generate ADP-ribosylated ubiquitin (ADPR-Ub) [268–271]. This intermediate is than cleaved by nucleotidase/phosphohydrolase/phosphodiesterase domains into AMP and phosphoribosylated ubiquitin (PR-Ub), which in the next step is attached to host proteins via a noncanonical serine-linked phosphodiester bond. The DUB domain (also found in SidE family members) than removes host ubiquitin modifications but not the SidE-mediated ubiquitination and reduces the level of ubiquitination on the surface of the *L. pneumophila*-containing vacuole (LCV) [271]. In addition, overexpression of SidE effector family members in mammalian or yeast cells generates a pool of PR-Ub that interfere with E1 and E2 enzymes hampering conventional host ubiquitination pathways. This results in interruption of mitophagy, immunity, as shown for TNF-induced NF-κB nuclear translocation, proteasomal degradation (for example of hypoxia inducing factor 1α) and other ubiquitin-regulated processes in the host [268,269]. Furthermore, ER structure and host membrane trafficking are modulated by SidE effector family members to enable LCV biogenesis. To ensure precise temporal control over the signal transduction mediated by ubiquitination, *L. pneumophila* secretes another DUB effector, SidJ, that is able to remove ubiquitin modifications established by the SidE effector family and the mammalian ubiquitination machinery [268,269,272].

*S. flexneri* also owns an effector with E3 ubiquitin ligase activity, IpaH9.8, that disrupts the NF-κB dependent pathway in the cytosol and impairs the activity of U2AF35, an mRNA splicing factor, leading to host inflammatory responses being suppressed [273–275]. The effector owns an N-terminal domain containing Leucine-Rich Repeats (LRR), also known as the LPX-domain, that is responsible for substrate recognition and a C-terminal E3 ubiquitin ligase domain, referred to as NEL domain (novel E3 ligase). Notably, the original structure of this domain differs from known eukaryotic E3 ligases, therefore, IpaH9.8 and its orthologues in other bacteria, for example, SspH 1 of *Salmonella enterica*, are part of a novel family of bacterial E3 ubiquitin ligases [276–281]. SspH 1 targets the host kinase PKN1 for proteasomal degradation, thereby functioning as ubiquitin ligase. In this context, it inhibits NF-κB dependent pro-inflammatory genes and regulates activation and function of neutrophils and macrophages as part of the androgen receptor pathway [278].

Enterohaemorrhagic *E. coli* express the T3SS effector protein NleG5-1, which contains a ubiquitin ligase U-box domain for ubiquitination and degradation of nuclear proteins. One target is a member of the mediator complex, MED15 that is a master regulator of RNA Pol II- dependent transcription [282]. The Ank-family expressed by *Orienta tsutsugamushi* (causative reagent of scrub typhus) is another protein family involved in ubiquitination and degradation, characterized by N-terminal Ank repeats and a C-terminal F-box-like domain termed as PRANC (pox protein repeats of ankyrin C terminus) motif. This family includes the proteins 1U5, 1A, 1B, 1E, 1F, 1U4 and 1U9 that interact with two members of multiprotein E3 ubiquitin ligase complexes, CULLIN-1 and SKP1 [283,284]. These proteins are supposed to act as mediators, as the ANK domain shall specifically bind target substrates, while the F-box recruits SKP1 promoting complex formation and finally inducing substrate degradation. This is also suggested for the degradation of the TF EF1α, probably induced via function of Ank 1U5 [283].

Especially the ubiquitin–proteasome system is a favored target of bacterial pathogens to manipulate the host cell cycle (summarized in Table 6). Two prominent targets are Skp1–Cullin1–F-Box protein (SCF, active throughout the cell cycle) and Anaphase-Promoting Complex/Cyclosome (APC/C, only regulatory active during Mitosis and late G1 phase), two classes of E3 ubiquitin ligase complexes inducing the degradation of cell cycle key regulators, thereby promoting its progression. A RING finger protein within both complexes enables binding to a scaffold cullin-like protein, the ubiquitin conjugating enzyme (E2) and distinct substrate-binding subunits. SCF are rated among the large family of Cullin-Ring E3 ubiquitin ligases (CRLs) as they are regulated via conjugation/deconjugation of the ubiquitin-like protein NEDD8 at the cullin subunit of SCF [285–289].


**Table 6.** Bacteria manipulating the host cell cycle.

Diverse animal pathogens, such as *E. coli*, *Yersinia pseudotuberculosis*, *Burkolderia pseudomallei* and *Photorhabdus* spp., express so-called Cycle inhibiting factors (Cif) that target SCFs and CRLs resulting in cell cycle arrest [292]. This effector expressed by *E. coli*, CifEc, induces cell cycle arrest at G1/S and G2/M transitions by accumulation of cyclin dependent kinase inhibitors p27/Kip1 and p21Waf1/Cip1 and inhibition of key kinases. The appearing cytopathic effect is accompanied by cell enlargement and production of actin stress fibers.

Delay of mitotic progression was observed when IpaB, an effector of *S. flexneri*, induced unscheduled APC/Ccdh1 activation and degradation of its substrates [293]. IpaB has to localize to the nucleus during G2/M phase and has to bind Mad2L2/MAD2B, the mitotic spindle assembly checkpoint protein that inhibits APC/Ccdh1. The benefit of these mechanisms for the bacteria is not completely clarified yet but it is suggested that delay of cell renewal and cell turnover enables the bacteria to persist and further colonize the host tissues.

#### **3. Conclusions**

During the long coevolution of pathogenic bacteria and their host cells, several strategies developed on both sides to fight their counterpart and keep predominance. Thus, intracellular bacteria reach for the establishment of an intracellular niche that allows survival, replication and persistence. This state is achieved through disarming of the host immune defense while keeping it healthy enough to gain permanent nutrition excess. Bacterial effectors are crucial tools during the whole process, thereby targeting the host immune response at each level of gene expression. Even, if there are already many bacterial effectors reported mimicking host enzymes or featuring novel enzymatic functions, the complex interaction mechanisms and networks are still not completely understood. Moreover, it appears, that pathogenic bacteria can target different pathways simultaneously or one pathway with several effectors, thereby creating a species-specific modification pattern. Therefore, understanding the individual strategies gain new insights into the complex host-pathogen interactions during infections. Further investigation might unravel, which bacterial strategies are more efficient or where host cells already found strategies to circumvent attempts of bacterial manipulation. This highlights the importance of further research on bacterial subversion of the host immune response considering each level of gene expression, as new promising targets for successful bacterial clearance during medical therapy might emerge.

**Author Contributions:** Conceptualization, L.D. and C.S.; writing–original draft preparation, L.D.; writing–review and editing, H.S. and C.S. All authors have read and agreed to the published version of the manuscript.

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

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

#### **Abbreviations**



*S. flexneri Shigella flexneri Salmonella* Typhimurium *Salmonella enterica* subsp. *enterica* serotype Typhimurium

#### **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/).
