*Cupriavidus metallidurans* **Strains with Different Mobilomes and from Distinct Environments Have Comparable Phenomes**

**Rob Van Houdt 1,\*, Ann Provoost 1, Ado Van Assche 2, Natalie Leys 1, Bart Lievens 2, Kristel Mijnendonckx <sup>1</sup> and Pieter Monsieurs <sup>1</sup>**


Received: 21 September 2018; Accepted: 15 October 2018; Published: 18 October 2018

**Abstract:** *Cupriavidus metallidurans* has been mostly studied because of its resistance to numerous heavy metals and is increasingly being recovered from other environments not typified by metal contamination. They host a large and diverse mobile gene pool, next to their native megaplasmids. Here, we used comparative genomics and global metabolic comparison to assess the impact of the mobilome on growth capabilities, nutrient utilization, and sensitivity to chemicals of type strain CH34 and three isolates (NA1, NA4 and H1130). The latter were isolated from water sources aboard the International Space Station (NA1 and NA4) and from an invasive human infection (H1130). The mobilome was expanded as prophages were predicted in NA4 and H1130, and a genomic island putatively involved in abietane diterpenoids metabolism was identified in H1130. An active CRISPR-Cas system was identified in strain NA4, providing immunity to a plasmid that integrated in CH34 and NA1. No correlation between the mobilome and isolation environment was found. In addition, our comparison indicated that the metal resistance determinants and properties are conserved among these strains and thus maintained in these environments. Furthermore, all strains were highly resistant to a wide variety of chemicals, much broader than metals. Only minor differences were observed in the phenomes (measured by phenotype microarrays), despite the large difference in mobilomes and the variable (shared by two or three strains) and strain-specific genomes.

**Keywords:** phenotype microarray; mobile genetic elements; *Cupriavidus*; metal; resistance

### **1. Introduction**

*Cupriavidus metallidurans* type strain CH34, which was isolated from a decantation basin in the non-ferrous metallurgical factory at Engis, Belgium [1], has been mostly studied because of its resistance to numerous heavy metals [2]. It tolerates high concentrations of metal (oxyan)ions, including Cu+, Cu2+, Ni2+, Zn2+, Co2+, Cd2+, CrO4 <sup>2</sup>−, Pb2+, Ag+, Au+, Au3+, HAsO4 <sup>2</sup>−, AsO2−, Hg2+, Cs+, Bi3+, Tl+, SeO3 <sup>2</sup>−, SeO4 <sup>2</sup><sup>−</sup> and Sr2+ [2,3]. Metal detoxification is encoded by at least 24 gene clusters and many of them are carried by its two megaplasmids pMOL28 and pMOL30 [4]. Resistance to metal ions is mediated by multiple systems, including transporters belonging to the resistance nodulation cell division (RND), the cation diffusion facilitator (CDF) and the P-type ATPase families [2,5].

*Cupriavidus metallidurans* strains have characteristically been isolated from metal-contaminated industrial environments such as soils around metallurgical factories in the Congo (Katanga) and North-Eastern Belgium [6,7], as well as from contaminated soils in Japan [8] and gold mining sites in Queensland (Australia) [9]. Other environments include sewage plants [10], laboratory wastewater (Okayama University, Okayama, Japan) [11] and spacecraft assembly cleanrooms [12]. In addition, *C. metallidurans* strains were also found in the drinking water and dust collected from the International Space Station (ISS) [12,13].

Remarkably, more and more reports describe the isolation of *C. metallidurans* strains from medically-relevant settings and sources such as the pharmaceutical industry, human cerebrospinal fluid and cystic fibrosis patients [14]. It remains to be elucidated if the isolates caused the active infection or only intruded as secondary opportunistic pathogens [14]. Nevertheless, an invasive human infection and four cases of catheter-related infections caused by *C. metallidurans* were recently reported [15,16].

All *Cupriavidus* genomes characteristically carry, next to their chromosome, a second large replicon. This 2 to 3 Mb-sized replicon has recently been coined chromid as it neither fully fits the term chromosome nor plasmid [17,18]. In addition to the chromid, most *Cupriavidus* strains harbor one or more megaplasmids (100 kb or larger in size), which probably mediate the adaptation to certain ecological niches by the particular functions they encode (see [19] for detailed review). For instance, pMOL28 and pMOL30 from *C. metallidurans* CH34 are pivotal in metal ion resistance [4]; hydrogenotrophic and chemolithotrophic metabolism are encoded by pHG1 from *Cupriavidus necator* H16 [20], and pRALTA from *Cupriavidus taiwanensis* LMG19424 codes for nitrogen fixation and legume symbiosis functions [21]. Next to these megaplasmids, other plasmids (mostly broad host range) can be present. One example is pJP4 from *Cupriavidus pinatubonensis* JMP134, which is a broad host range IncP-1β plasmid involved in the degradation of substituted aromatic pollutants [22].

The *C. metallidurans* mobilome is completed with a large diversity of genomic islands (GIs), integrative and conjugative elements, transposons and insertion sequence (IS) elements [7,23–25]. Many mobile genetic elements (MGEs) carry accessory genes beneficial for adaptation to particular niches (resistance, virulence, catabolic genes), but acquired genes may also impact the host by cross-talk to host global regulatory networks [26]. In addition, without accessory genes, MGEs such as IS elements can have an impact on genome plasticity and concomitant adaptability of phenotypic traits, including resistance to antibacterial agents, virulence, pathogenicity and catabolism [27]. Finally, the presence of prophages, until now not identified in *C. metallidurans*, may also affect many different traits and lead to phenotypic changes in the host [28,29].

Recently, we showed that *C. metallidurans* strains share most metal resistance determinants irrespective of their isolation type and place [7]. In contrast, significant differences in the size and diversity of their mobilome was observed. However, our comparison was based on whole-genome hybridization to microarrays containing oligonucleotide probes present on the CH34 microarray. These observations triggered us to further study the diversity of the mobilome, its relation to the environment and impact on the host's global phenome. Therefore, we inventoried the mobilomes and compared the global metabolic capabilities of type strain CH34, strains NA1 and NA4 isolated from water sources aboard ISS [12], and H1130 isolated from an invasive human infection [15]. The global metabolic activities were assessed by employing phenotype microarrays (PMs), which highlight differences in growth requirements, nutrient utilization and sensitivity to chemicals [30].

### **2. Materials and Methods**

### *2.1. Strains, Media and Culture Conditions*

Bacterial strains and plasmids used in this study are summarized in Table 1. *Cupriavidus metallidurans* strains were routinely cultured at 30 ◦C in lysogeny broth (LB) or tris-buffered mineral medium (MM284) supplemented with 0.2% (*w/v*) gluconate [1]. *Escherichia coli* strains were routinely cultured at 37 ◦C in LB. Liquid cultures were grown in the dark on a rotary shaker at 150 rpm. For culturing on agar plates, 1.5% agar (Thermo Scientific, Oxoid, Hampshire, UK) was added. When appropriate, the following chemicals (Sigma-Aldrich (Overijse, Belgium) or Fisher Scientific (Merelbeke, Belgium)) were added to the growth medium at the indicated final concentrations: kanamycin (50 μg/mL for *E. coli* or 1500 μg/mL for *C. metallidurans*), tetracycline (20 μg/mL), 5-bromo-4-chloro-3-indolyl-β-D-galactopyranoside (X-Gal; 40 μg/mL), isopropyl-β-D-thiogalactopyranoside (IPTG; 0.1 mM) and diaminopimelic acid (DAP; 1 mM).



Eurogentec: Seraing, Belgium, KmR: kanamycine resistant, TcR: tetracycline resistant, CmR: chloramphenicol resistant, ApR: ampicillin resistant.

### *2.2. Growth in the Presence of Metals*

*Cupriavidus metallidurans* CH34, NA1, NA4 and H1130 were cultivated in MM284 at 30 ◦C up to stationary phase (10<sup>9</sup> CFU/mL) and 10 μL of a ten-fold serial dilution in 10 mM MgSO4 were spotted on MM284 agar plates containing various metal concentrations (Table S1). Colony forming units (CFU) were counted after 4–5 days. Data are presented as log(N)/log(N0) in function of metal concentration, with N and N0 CFUs in the presence and absence (control) of metal, respectively.

### *2.3. NA4 CRISPR Deletion Construction*

The CRISPR region of *C. metallidurans* NA4 was amplified by PCR (Phusion High-Fidelity DNA polymerase) (Fisher Scientific, Merelbeke, Belgium) with primer pairs CRSPR\_Fw-Rv (Table S2), providing XbaI/HindIII restriction sites. Afterwards, this PCR product was cloned as a XbaI/HindIII fragment into the mobilizable suicide vector pK18mob. The resulting pK18mob\_CRISPR plasmid from an *E. coli* DG1 transformant selected on LB Km50 was further confirmed by sequencing prior to amplifying of the flanking CRISPR sequences by inverse PCR (Phusion High-Fidelity DNA polymerase) with primer pair CRISPR\_tet\_Fw-Rv (Table S2), providing BcuI/BspTI restriction sites. At the same time, the *tet* gene from pACYC184 (Table 1 [34]) was amplified by PCR (Phusion High-Fidelity DNA polymerase) with primer pair Tet\_Fw-Rv (Supplementary Table S1), providing BcuI/BspTI restriction sites. Afterwards, this PCR product was cloned as a BcuI/BspTI fragment into the former inverse PCR product. The resulting pK18mob-CRISPR::*tet* plasmid from an *E. coli* DG1 transformant selected on LB

Tc20 Km50 was further confirmed by sequencing prior to conjugation (with *E. coli* MFDpir as donor host [32]) to *C. metallidurans* NA4. The resulting transformants selected on LB Tc20 were replica plated on LB Tc20 and LB Km1500. NA4 ΔCRISPR::*tet* cells resistant to Tc20 but sensitive to Km1500 were further confirmed by sequencing.

### *2.4. Construction of Plasmids*

PCR amplification of *C. metallidurans* CH34 Rmet\_2825 was performed on genomic DNA from *C. metallidurans* CH34 with primer pair Rmet2825\_Fw-Rv (Table S2). This amplicon was subsequently cloned into pJB3kan1, which was linearized by PCR amplification with the primers pJB3kan1\_Fw-Rv (Table S2), using the GeneArt™ Seamless Cloning and Assembly Enzyme Mix (Fisher Scientific, Merelbeke, Belgium). The resulting pJB3kan1-Rmet2825 plasmid from *E. coli* DG1 transformants selected on LB Km50 was further confirmed by sequencing prior to transformation to *E. coli* MFDpir.

### *2.5. Conjugation Assay for Testing CRISPR-Cas*

Donor (*E. coli* MFDpir pJB3kan1-Rmet2825) and recipient (*C. metallidurans* NA4 or NA4 ΔCRISPR::*tet*) were grown overnight at 37◦ in LB Km50 DAP, and at 30◦ in LB, respectively. Fifty μL of donor and recipient were spotted on a 0.45 μm Supor® membrane disc filter (Pall Life Sciences, Hoegaarden, Belgium) that was put on a LB DAP plate. After overnight incubation at 30 ◦C, cells were resuspended in 1 mL of 10 mM MgSO4 and 10-fold serial diluted on LB Km50 DAP (37 ◦C), LB (30 ◦C) and LB Km1500 plates (30 ◦C) to count CFU of donors, recipients and transconjugants, respectively. Conjugation frequency was measured as the number of transconjugants per donor cell (T/D) and per recipient cell (T/R).

### *2.6. Plasmid Profiling*

The extraction of megaplasmids was based on the method proposed by Andrup et al. [36]. Extracted plasmid DNA was separated by horizontal gel electrophoresis (0.5% Certified Megabase agarose gel (Bio-Rad, Temse, Belgium) in 1X TBE buffer, 100 V, 20 h) in a precooled (4 ◦C) electrophoresis chamber. After GelRed staining (30 min + overnight destaining at 4 ◦C in ultrapure water), DNA was visualized and images captured under UV light transillumination (Fusion Fx, Vilber Lourmat, Collégien, France).

### *2.7. Phenotype Microarray Analysis*

Phenotype microarray (PM) analysis was performed using the OmniLog® automated incubator/reader (Biolog Inc., Hayward, CA, USA) following manufacturer's instruction (PM procedures for *E. coli* and other GN Bacteria version 16-Jan-06 with slight modifications). Briefly, cells were suspended in Biolog's inoculation fluid IF-0a (1x) until an optical density (600 nm) of 0.2 was reached. Subsequently, a 1:50 dilution was made in IF-0a (1x) containing dye mix A. Furthermore, 2 mM sodium succinate and 2 μM ferric citrate (Sigma-Aldrich, Overijse, Belgium) were used as carbon sources in PM 3 till 8. All 20 plates (PM-1 through PM-20) inoculated with bacterial cell suspensions, were incubated at 30 ◦C and cell respiration was measured every 30 min for 144 h. Raw kinetic data were retrieved using the OmniLog—OL\_PM\_FM/Kin 1.30-: File Management/Kinetic Plot Version software of Biolog. Analysis was carried out with the R-library OPM (version 1.3.64) [37,38]. The area under the curve (AUC) threshold to decide whether a strain is or is not growing in a specific well of the PM, was derived by plotting the AUC values of all PM reactions for each strain, showing in all conditions an almost bimodal distribution. The AUC threshold (one value for all four strains) was determined as the value separating both major peaks (threshold value of 8000) (Figure S1). Negative control wells that contained the inoculated Omnilog™ growth medium without any substrate were measured to normalize differences in inocula and redox dye oxidation between samples.

### *2.8. Computational Methods*

The pan-genome analysis was performed via the MaGe platform [39], which uses MicroScope gene families (MICFAM) that are computed with an algorithm implemented in the SiLiX software [40]. The alignment constraints to compute the MICFAM families were 80% amino-acid identity and 80% amino-acid alignment coverage. The MICFAM is part of the core-genome if associated with at least one gene from every compared genome (see Table S3 for complete data set).

A phylogenetic tree of the genomes was constructed via the MaGe platform from the pairwise genome distances using a neighbor-joining algorithm. The pairwise genome distance was calculated with Mash [41].

The CARD (comprehensive antibiotic resistance database) [42] implementation within the MaGe platform [39] was used to identify known resistance determinants and associated antibiotics. All predictions were strict as defined by CARD, meaning a match above the CARD curated bitscore cut-offs [42–44].

A BLAST search against BacMet (antibacterial biocide and metal resistance genes database) was used to inventory genes predicted to confer resistance to metals and/or antibacterial biocides [45]. The alignment constraints were 35% amino-acid identity and 80% amino-acid alignment coverage.

The different constraints used to compute the MICFAM families and CARD/BacMet BLAST hits can result in minor differences in the number of core genome genes from a particular strain that results in a positive CARD/BacMet hit.

### **3. Results and Discussion**

Four *C. metallidurans* strains were selected: type strain CH34 [3], strain NA1 and NA4 isolated from the drinking water systems onboard the International Space Station that were analyzed previously and had mobilomes divergent from that of CH34 [7], and strain H1130, recently isolated from an invasive human infection [15]. This selection allows comparing the type strain with two strains isolated from a similar environment but with different mobilomes (at least based on elements known in CH34 [7]) and an isolate from a human infection.

### *3.1. Comparison of General Genome Features*

The genome of *C. metallidurans* NA1, NA4 and H1130 was previously sequenced [46,47] and estimated to be 6,833,318 bp, 7,370,364 bp and 7,225,099 bp, respectively (with type strain CH34 being 6,913,352 bp [3]). The G + C content of the genomes are very similar to each other, with 63.76%, 63.27%, 63.50% and 63.82% for NA1, NA4, H1130 and CH34, respectively. NA4 contained the most coding sequences (CDSs) (7467), followed by H1130 (7032), NA1 (6815) and CH34 (6757). All strains contained multiple replicons, namely, one chromosome, one chromid and megaplasmids (>100 kb [19]). Strain NA1 carries two megaplasmids. Strain NA4 carries three megaplasmids and one plasmid. Strain H1130 carries only one megaplasmid (Figure 1).

The core genome contains 4697 MICFAM gene families shared by all four strains, which relates to 70.9%, 70.2%, 65.4% and 69.9% of the total CDSs of CH34, NA1, NA4 and H1130, respectively. This means that roughly 30% to 35% of the CDSs belong to the variable (shared by two or three strains) or strain-specific genome (Figure 2). Strains CH34 and NA4 shared the most gene families (Figure 3). Furthermore, the Mash-distance-based phylogeny (Figure 4) indicated that NA4 and CH34 were the most closely related. In addition, NA4 shared more gene families with H1130 and CH34 than with NA1, which corresponded with the phylogenetic distance. These data indicated that NA1 and NA4 were not the two most similar strains, despite their isolation from the same environment.

**Figure 1.** Agarose gel electrophoresis of *Cupriavidus metallidurans* CH34, NA1, NA4 and H1130 (mega)plasmid DNA. The characterized CH34 megaplasmids pMOL30 (234 kb) and pMOL28 (171 kb) serve as reference.

**Figure 2.** Percentage of coding sequences (CDSs) belonging to the core, variable (shared by two or three strains) and strain-specific genome of *Cupriavidus metallidurans* CH34, NA1, NA4 and H1130.

**Figure 3.** Venn diagram displaying the distribution of shared MicroScope gene families (MICFAM) among *C. metallidurans* CH34, NA1, NA4 and H1130.

**Figure 4.** Neighbor-joining phylogenetic tree of *C. metallidurans* CH34, NA1, NA4 and H1130, based on the genome pairwise distance matrix calculated with Mash. *Cupriavidus basilensis*: OR16, *Cupriavidus pinatubonensis*: JMP134, *Cupriavidus necator*: H16 and N-1, and *Cupriavidus taiwanensis*: LMG19424 were included for comparison.

Evidently, with 1827 Microscope gene families shared with *Cupriavidus taiwanensis* LMG19424, *Cupriavidus necator* H16, *Cupriavidus pinatubonensis* JMP134, *Cupriavidus basilensis* OR16 and *Cupriavidus necator* N-1, the *C. metallidurans* strains share more gene families among each other than with strains of different *Cupriavidus* species. Strains CH34, NA1, NA4 and H1130 shared 1977 gene families unique to the *C. metallidurans* species.

The COGnitor module [48] implemented in the MaGe platform was used to compare the CDSs of the core, variable and specific genome assigned to a COG (clusters of orthologous groups) functional category (Figure 5). The latter indicated that for all four strains, COG L (replication, recombination and repair) and U (intracellular trafficking and secretion) are overrepresented on the variable plus specific genome. Other COGs were also significantly overrepresented on the variable plus specific genome for particular strains. For instance, COG D (cell cycle control, division and partitioning) for CH34, NA4 and H1130, and COG V (defense mechanisms) for NA1 and NA4 (see Figure 5 for all significant overrepresentations).

**Figure 5.** Percentage of CDSs assigned to a COG (clusters of orthologous groups) functional class (general categories: cellular processes and signaling: D, M, N, O, T, U, V; information storage and processing: A, B, J, K, L; metabolism; C, E, F, G, H, I, P, Q; poorly characterized: R, S) belonging to the core, variable of strain-specific genome of *C. metallidurans* CH34, NA1, NA4 and H1130. \* Significant (*p* < 0.05; based on hypergeometric distribution) overrepresentation of COG on variable + specific compared to the core genome.

### *3.2. The Mobilome*

Recently, we showed that *C. metallidurans* strains have substantial differences in the diversity and size of their mobile gene pool [7]. However, since this comparison was based on whole-genome hybridization to microarrays containing type strain CH34 oligonucleotide probes, the presence of MGEs other than those in CH34 could not be assessed. Here, the mobilomes of NA1, NA4 and H1130 (including IS elements, transposons, genomic islands and prophages) as well as the presence of CRISPR-Cas systems were scrutinized.

### 3.2.1. Insertion Sequence Elements and Transposons

ISFinder [49] and ISSaga [50] (+ manual curation) were used to create an inventory of the IS elements, which identified 57, 25, 33 and 91 putative IS elements in CH34 [24], NA1, NA4 and H1130, respectively. It must be noted that this list is based on a draft genome assembly for NA1, NA4 and H1130, which could have an impact on the actual number. Possible identical IS elements present in multiple copies will only be represented as one contig in the genome assembly, as such leading to an underestimation of the number of IS elements in the respective genome [51]. Active IS transposition in CH34 was already observed for IS*Rme1*, IS*Rme3*, IS*Rme5*, IS*Rme15*, IS1086, IS1087B, IS1088 and IS1090 [24,52–58]. Transposition activity of IS*Rme5* > IS*1088* > IS*Rme3* > IS*1087B* > IS*1090* > IS*1086* > IS*Rme15*, at least into the *cnr* target after exposure of AE126, a derivative of CH34 cured from plasmid pMOL30 carrying the main zinc resistance determinant, to 0.8 mM Zn2+ [58]. Some of these active IS elements are also carried by NA1 (2 IS*Rme3* copies), NA4 (1 IS*Rme1*, 4 IS*Rme4* and 1 IS*Rme5* copy) and H1130 (16 IS*Rme3* copies) (based on 98% DNA sequence identity cut-off). Next to transposition, IS elements can also cause more extensive/general loss of genetic information by recombination events between identical individual IS copies, e.g., loss of the CH34 genes involved in autotrophy by IS*1071*-mediated excision [24]. Similar observations of IS*1071*-mediated rearrangements affecting the metabolic potential of the host have been described for *Comamonas* sp. strain JS46 [59] and *Cupriavidus pinatubonensis* JMP134 [60]. Thus, these IS elements in CH34, NA1, NA4 and H1130 can play a multifaceted, pivotal role in the adaptation to stress conditions (as shown for CH34) [27,58].

The CH34 genome harbors five distinct transposon families totaling 19 intact transposons. The transposition modules of four transposons are related to those of mercury transposons with Tn*4378*, Tn*4380* and Tn*6050* belonging to the Tn*21*/Tn*501* family, and Tn*6048* to the Tn*5053* family [61]. The transposition module of Tn*6049* could not be categorized. Tn*6048*, Tn*6049* and mercury transposons are also conserved in NA1 (one Tn*6048* copy, one Tn*6049* copy), NA4 (3 mercury transposons, 3 Tn6049 copies) and H1130 (4 mercury transposons). Tn*6050* appeared to be only present in CH34. No other transposons were identified.

### 3.2.2. Genomic Islands

The MaGe platform was used to scrutinize the presence of genomic islands (GIs), including those previously identified in CH34. The largest island (109 kb) on the chromosome of CH34 belongs to the large pKLC102/PAGI-2 family of elements that share a core gene set and are integrated downstream of tRNA genes [62,63]. A similar element is present in NA1 (2 copies), NA4 and H1130 as shown by progressive Mauve alignment [64] (Figure S2). The Tn*4371*-family of integrative and conjugative elements CMGI-2, CMGI-3 and CMGI-4 of CH34 were previously designated ICETn*4371*6054, ICETn*4371*6055 and ΔICETn*4371*6056, respectively [65]. CMGI-2 (ICETn*4371*6054) and CMGI-3 (ICETn*4371*6055) are responsible for CH34's ability to grow on aromatic compounds and to fix carbon dioxide, respectively [7,24]. No Tn*4371*-family genomic island was identified in NA4. One Tn*4731*-family element was identified in NA1, which is highly similar to previously identified elements in *Delftia acidovorans* SPH-1 (DAGI-1; ICETn*4371*60370), *Comamonas testosteroni* KF-1 (CTGI-1; ICETn*4371*6038) and the partial CMGI-4 (ΔICETn*4371*6056) of CH34 [25,65]. The island carries an RND-driven efflux system. In H1130, two Tn*4371*-family genomic islands were identified, one carrying

12 genes (putatively involved in ion transport), while the second could not be correctly defined as the integration/excision and stabilization/maintenance module up to *rlxS* (encoding a relaxase protein) are not located on the same contig as the transfer module (starting from *traR* coding for a transcriptional regulator). Therefore, the accessory genes that are typically located between *rlxS* and *traR* in Tn*4371*-family members could not be properly assessed [65]. All other GIs on CH34's chromosome were not found in the other strains, except CMGI-5 in NA1. CMGI-C and CMGI-E, previously identified on CH34's chromid, are absent in all strains. CMGI-A, -B and -D are conserved in NA4 and H1130, but show limited synteny with NA1. No other genomic islands could be clearly identified in NA1 or NA4. One other genomic island was clearly noticeable in H1130. This 87 kb region, which is absent in CH34, NA1 and NA4, is syntenic with an 80-kb cluster located on the 1.47-Mbp megaplasmid of *Burkholderia xenovorans* LB400. In *B. xenovorans* LB400, this Dit island encodes proteins of abietane diterpenoids metabolism and mediates growth on abietic acid, dehydroabietic acid, palustric acid and 7-oxo-dehydroabietic acid [66] (not included in the phenotypic microarray). Abietane diterpenoids are tricyclic, C-20, carboxylic acid-containing compounds produced by plants and are a key component of the defense systems of coniferous trees [66,67]. This observation also adds evidence to the mobility of this cluster and its distribution among proteobacterial genomes [66]. In addition, two smaller regions (13.6 and 10.3 kb) carrying genes coding for unknown functions and a tyrosine-based site-specific recombinase were identified.

### 3.2.3. Prophages

The presence of prophages was scrutinized via PHASTER [68] and showed no prophages in type strain CH34 (which was already known) and the presence of intact prophages in NA4 and H1130 as well as incomplete/remnants in H1130, NA1 and NA4 (Table 2). Although mitomycin C exposure did not result in prophage induction (data not shown), a derivative of NA4 exposed to uranium lost the 43.6 kb region predicted as an intact prophage (unpublished data).


**Table 2.** Prophage detected in *C. metallidurans* NA1, NA4 and H1130.

<sup>a</sup> Size in kb; <sup>b</sup> Prediction of whether the region contains an intact or incomplete prophage and <sup>c</sup> score based on PHASTER criteria [68]; <sup>d</sup> number of proteins; <sup>e</sup> the phage with the highest number of proteins most similar to those in the region (between parentheses: accession number; number of proteins).

### 3.2.4. CRISPR-Cas

The CRISPR-Cas system is an adaptive immunity system that stores memory of past encounters with foreign DNA in spacers that are inserted between direct repeats in CRISPR arrays [69]. CRISPR-Cas systems were detected with CRISPRfinder [70] and CRISPRDetect [71] (default settings). Only positive hits with both were further examined, resulting in the identification of 1 CRISPR-Cas system in NA4. CRISPRTarget [72] identified 5 spacer sequences related to genomic island CMGI-5 of CH34 (which is also present in NA1). CMGI-5 is probably a plasmid remnant and contains besides hypothetical genes, some typical plasmid-related genes such as *repA*, *traY*, *mobA* and *mobB*. To assess if the identified system is active, the conjugation frequency of plasmid pJB3kan1 carrying

the CMGI-5 *repA* gene (pJB3kan1\_Rmet2825; containing one spacer) was determined for the parental and CRISPR-deleted NA4 strain. CRISPR deletion in NA4 increased the conjugation efficiency 33-fold, indicating an active CRISPR-Cas system in NA4 (Figure 6).

**Figure 6.** Conjugation frequency of pJB3kan1\_Rmet2825 (containing one spacer identified in the NA4 CRISP-Cas system) from donor *E. coli* MFDpir to recipient *C. metallidurans* NA4 and NA4 ΔCRISPR::*tet*, respectively. Median values plus corresponding calculated standard deviations across biological triplicates are shown (T/D = conjugation frequency per donor; T/R = conjugation frequency per recipient).

### *3.3. The Resistome*

### 3.3.1. Antibiotic Resistance

The CARD [42] implementation within the MaGe platform [39] was used to identify known resistance determinants and associated antibiotics. The latter predicted 33, 36, 33 and 39 proteins involved in antibiotic resistance in CH34, NA1, NA4 and H1130, of which 31, 31, 30 and 39 belonged to the core genome, respectively. No marked difference in tolerance to antibiotics was observed.

### 3.3.2. Metal Resistance

The antibacterial biocide and metal resistance genes database (BacMet) was used to create an inventory of genes predicted to confer resistance to metals and/or antibacterial biocides [45]. This showed 302, 282, 337 and 302 proteins involved in biocide and metal resistance in CH34, NA1, NA4 and H1130, respectively. Most genes belonged to the core genome (221, 246, 276 and 251 for CH34, NA1, NA4 and H1130, respectively). The compounds (metal and chemical class) to which these genes confer resistance are very similar for all four strains (Figure 7). Genes conferring resistance to nickel, copper, cobalt and the chemical classes acridine and phenanthridine were the most abundant.

For CH34, the predicted genes contained 68 out of the 174 genes that were previously identified to be related to metal resistance (for an overview see [2,3]). Specific analysis of these 174 proteins showed that almost all are conserved in NA1, NA4 and H1130. Exceptions are (*i*) the accessory cluster related to chromate resistance in H1130, (*ii*) the *hmz* cluster in NA4 and H1130, (*iii*) *cdfX* in NA1, NA4 and H1130, and (*iv*) the *dax*/*gig* cluster in NA1. The latter three are all located on a genomic island. The gene cluster related to chromate resistance on pMOL28 from CH34 contains five additional genes that are strongly induced by chromate in CH34 [73] as well as for the homologous system in *Arthrobacter* sp. FB24 (both at the gene and protein level) [74,75]. The *hmz* cluster is a HME-RND-driven

system, belonging to the HME3b (Heavy Metal Efflux) subfamily of the RND superfamily, with no known substrate and transcriptionally silent in *C. metallidurans* CH34 [5,73,76]. The *cdfX* gene of CH34 encodes a putative permease (211 amino acid residues and six predicted transmembrane α-helices) that shares 87% amino-acid identity with PbtF from *Achromobacter xylosoxidans* A8 [5]. Expression of *pbtF* in *A. xylosoxidans* A8 was induced by Pb2+, Cd2+ and Zn2+, and although PbtF showed measurable Pb2+-efflux activity, it did not confer increased metal tolerance in *E. coli* GG48 [77]. The *dax* cluster [73], which was renamed *gig* for "gold-induced genes" in Wiesemann et al. [78], is induced by Ag<sup>+</sup> and Au3+ [73,79] but not essential for gold resistance [78].

**Figure 7.** Inventory of *C. metallidurans* CH34, NA1, NA4 and H1130 genes conferring resistance to metals/chemical classes based on the BacMet database (antibacterial biocide and metal resistance genes database).

*Genes* **2018**, *9*, 507

In agreement with the conservation of these metal resistance determinants, growth in the presence of increasing metal concentrations showed only minor differences between CH34, NA1, NA4 and H1130 (Figure 8). Moreover, the minor strain-dependent differences (see above) did not mediate differences in metal resistance (Figure 8). Essentially, the most noticeable difference in growth was observed in the presence of Ni2+, with higher concentrations tolerated by NA4 and H1130. Initially, the *nccCBA* locus, which is inactivated in CH34 because of a frame shift mutation, was put forward as a possible explanation [12]. However, the frame shift mutation in *nccB* is present in all four strains. However, NA4 and H1130 carry a second *nccYXHCBAN* locus coding for an RND-driven efflux system involved in Ni2+ and Co2+ resistance. This locus is homologous to that of *C. metallidurans* 31A and KT02, which has been shown to be responsible for resistance to 40 mM Ni2+ [80], and is likely responsible for the observed differences. In addition, although the *nimBAC* locus, coding for an RND-driven efflux system putatively involved in Ni2+ and Co2+ resistance [5], is only inactivated in CH34 (via IS*Rme3* insertion) and not in NA1, NA4 and H1130, growth in the presence of Ni2+ is similar for NA1 and CH34. Other observations are the lower resistance of NA1 to Cd2+ and to lesser extent Co2+ and Ag+. However, based on the current data, no hypotheses can be put forward to explain these observations.

**Figure 8.** Viable count of *C. metallidurans* CH34, NA1, NA4 and H1130 grown in the presence of various metal concentrations (Table S2). Data are presented as log(N)/log(N0) in function of metal concentration, with N and N0 the colony forming units (CFUs) in the presence and absence (control) of metal, respectively.

### *3.4. Phenotypic Microarrays*

In order to scrutinize functional differences between the four *C. metallidurans* strains, phenotypic characterization with OmniLog Phenotypic Microarrays (PMs) was conducted. Area under the curve (AUC) values were calculated and a threshold cut-off (8000) was applied to discriminate a positive (growth) from a negative (no-growth) reaction. This revealed an overall phenotypic similarity among the four strains, with 1744 out of the 1920 assays shared (Figures 9–11).

**Figure 9.** Venn diagram displaying the distribution of OmniLog phenotypic assays shared among *C. metallidurans* CH34, NA1, NA4 and H1130.

*Genes* **2018**, *9*, 507

**Figure 10.** Overview of positive (growth) OmniLog phenotypic assays shared by *C. metallidurans* CH34, NA1, NA4 and H1130 for each PM plate (with 1 being all 96 assays). The assays on each PM plate are detailed in Table S4.

**Figure 11.** Overview of positive (growth) OmniLog phenotypic assays shared by *C. metallidurans* CH34, NA1, NA4 and H1130 for different metabolic and chemical sensitivity tests (with 1 being all assays shared).

### 3.4.1. C, N, P and S Sources

Only around 27% to 28% of the C source reactions was positive, which is related to their inability to assimilate sugars and sugar alcohols (Figure 11) [1,3]. All four strains lack a glucose uptake system. The latter is most likely deleted in all four strains as a N-acetylglucosamine-specific phosphotransferase system (PTS)-type transport system essential for glucose uptake (growth) in *Cupriavidus necator* H16 [81,82] is absent in from a large syntenic region (>110 genes) conserved among *C. necator* H16 and *C. metallidurans* CH34, NA1, NA4 and H1130 (data not shown).

A few marked differences were observed for the use of amino acids as N source, in particular for L-leucine, L-tryptophan and L-Valine (Figure 12). Specific for L-tryptophan, growth was observed for NA1, NA4 and H1130 but not for CH34. Aerobic L-tryptophan degradation in *C. metallidurans* most likely occurs via a three-step pathway to anthrilanate requiring tryptophan 2,3-dioxygenase (*kynA*), kynurenine formamidase (*kynB*) and kynureninase (*kynU*). Experimental verification of the anthranilate pathway was achieved by functional expression of the CH34 *kynBAU* operon in *Escherichia coli* after suppressing the stop codon disrupting *kynB* [83]. This amber mutation is not present in NA1, NA4 and H1130, which could explain the observed differences. Similar differences were also observed when growth was scored for dipeptides (N source), as CH34 grew less or not on L-tryptophan-containing dipeptides compared to NA1, NA4 and H1130. Only minor differences were observed for growth on P and S sources (Figure 11, Table S4).

**Figure 12.** (**a**) Overview of OmniLog phenotypic assays with amino acids as N source (AUC = area under the curve) for *C. metallidurans* CH34, NA1, NA4 and H1130, (**b**) Growth kinetics in the presence of L-leucine, L-tryptophan and L-Valine as N source (AOU = arbitrary OmniLog units).

3.4.2. Osmolytes and pH

The addition of ionic osmolytes had a clear and comparable impact on the growth of strains CH34, NA1, NA4 and H1130, as growth was generally only observed for the lower/lowest concentrations (1% NaCl, 2% Na2SO4, 1% sodium formate, 3% urea and 2% sodium lactate). In contrast, addition of up to 20% of the non-ionic osmolyte ethylene glycol had no impact on growth of CH34, NA1, NA4 and H1130.

The effect of pH over the range 3.5 to 10 growth was comparable for CH34, NA1, NA4 and H1130. Growth was inhibited below pH 5 for all strains. Growth at pH 10 was much more pronounced for H1130 than for the other strains (Figure 13).

**Figure 13.** (**a**) Overview of OmniLog phenotypic assays related to pH (AUC = area under the curve) for *C. metallidurans* CH34, NA1, NA4 and H1130, (**b**) Growth kinetics at pH 10 (AOU = arbitrary OmniLog units).

### 3.4.3. Chemicals

The PM-11 to PM-20 plates carry different chemicals (4 increasing concentrations of each) to test sensitivity, only for eight out of the 240 chemicals tested at least one of the strains was susceptible to the lowest concentration. For more than 50% of the tested chemicals, CH34, NA1, NA4 and H1130 were resistant to the highest concentration included in the phenotypic microarrays (Figure 14).

No growth was observed in the presence of 2,2 -dipyridyl (metal chelator), hydroxyurea (ROS producer) and phenethicillin (a narrow-spectrum, β-lactamase-sensitive penicillin) for all four strains. In contrast to phenethicillin, CH34, NA1, NA4 and H1130 were resistant to (at least one concentration of) all other β-lactam antibiotics tested. Only H1130 grew in the presence of sodium meta- and orthovanadate, and did not grow in the presence of thallium acetate (Figure 15). Strain CH34 and NA4 did not grow in the presence of potassium tellurite (Figure 15). The genetic basis underlying resistance to these metals is poorly understood, therefore, no correlation to the genotype could be established. Strain CH34 and NA1 were susceptible to sodium metaperiodate (oxidizing agent) and tolylfluanid (fungicide), respectively (Figure 15).

**Figure 14.** Percentage of tested chemicals (n = 240) to which *C. metallidurans* CH34, NA1, NA4 and H1130 are resistant. Four increasing concentrations are included in the phenotypic microarrays (0: susceptible to the lowest concentration, 1 to 4: resistant to the lowest up to the highest concentration).

**Figure 15.** Growth kinetic of *C. metallidurans* CH34, NA1, NA4 and H1130 in the presence of different chemicals (lowest concentration in the phenotypic microarrays is shown) (AOU = arbitrary OmniLog units).

### 3.4.4. Trait Prediction

Finally, the prediction of Traitar, an automated software framework for the accurate prediction of 67 phenotypes directly from a genome sequence [84], was evaluated by comparison with the generated phenotypic data (OmniLog Phenotypic Microarray data and previous observations/knowledge). Traitar correctly predicted 85% (45 out of 53 analyzed), 81% (38 out of 47), 80% (37 out of 46) and 80% (37 out of 46) of the CH34, NA1, NA4 and H1130 traits, respectively (Figure 16). Although Weimann and colleagues [84] indicated that the phypat classifier assigned more phenotypes at the price of more false-positive predictions, whereas the phypat + PGL classifier assigned fewer phenotypes with fewer false assignments, it appeared that in the case of the *C. metallidurans* strains, phypat + PGL assigned more false-positive predictions.

**Figure 16.** Overview of the correctly (C) and falsely (F) predicted phenotypic traits of *C. metallidurans* CH34, NA1, NA4 and H1130 by Traitar (two classifiers: phypat and phypat + PGL).

### **4. Conclusions**

The comparison of four *C. metallidurans* strains isolated from different environments indicated that metal resistance determinants and properties are maintained in these environments. As most of the metal determinants are on the native megaplasmids, it could be argued that these environments provided a selective pressure for the conservation of these determinants and plasmids. The previously identified differences in the size and diversity of the mobile gene pool were put in perspective by the identification of intact (and remnant) prophages in NA4 and H1130, and a genomic island putatively involved in abietane diterpenoids metabolism in H1130. The latter indicated that mobilome diversity differed (integrative and conjugative elements/genomic islands versus prophages). Furthermore, the mobilome is apparently not directly related to the isolation environment as the NA1 mobilome is

shaped more like that of H1130 than that of NA4 isolated from the same environment. In addition, an active CRISPR-Cas system was identified in strain NA4, providing immunity to a plasmid that integrated in CH34 and NA1. Despite the large size of the variable and specific genomes, only minor differences were observed in the global phenomes (as measured by phenotype microarrays) and all four strains were highly resistant to a wide variety of chemicals, much broader than metals. The variable and specific genome were probably acquired through later transfer and perhaps carry functions more essential for survival in challenging and fluctuating environments than general metabolic functions.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2073-4425/9/10/507/s1, Figure S1: Density plot of the AUC values of all PM reactions for each strain, Figure S2: Progressive Mauve alignment of CMGI-1 of CH34 with related elements in NA1, NA4 and H1130, Table S1: Metal concentrations used in growth experiments, Table S2: Primers used in this study, Table S3: Pan-genome analysis data set, Table S4: PM area under the curve (AUC) values.

**Author Contributions:** Conceptualization, R.V.H.; Data curation, R.V.H. and P.M.; Formal analysis, R.V.H. and P.M.; Funding acquisition, R.V.H. and N.L.; Investigation, R.V.H., A.P., A.V.A. and K.M.; Methodology, R.V.H. and P.M.; Project administration, R.V.H.; Software, R.V.H. and P.M.; Supervision, R.V.H., N.L. and B.L.; Validation, R.V.H. and P.M.; Visualization, R.V.H.; Writing—original draft, R.V.H.; Writing—review & editing, R.V.H., A.V.A., K.M. and P.M.

**Funding:** This work was funded by the European Space Agency (ESA-PRODEX) and the Belgian Science Policy (Belspo) through the COMICS project (C90356).

**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/).

### *Article* **Unintentional Genomic Changes Endow** *Cupriavidus metallidurans* **with an Augmented Heavy-Metal Resistance**

**Felipe A. Millacura 1, Paul J. Janssen 2, Pieter Monsieurs 2, Ann Janssen 2, Ann Provoost 2, Rob Van Houdt <sup>2</sup> and Luis A. Rojas 3,\***


Received: 6 October 2018; Accepted: 8 November 2018; Published: 13 November 2018

**Abstract:** For the past three decades, *Cupriavidus metallidurans* has been one of the major model organisms for bacterial tolerance to heavy metals. Its type strain CH34 contains at least 24 gene clusters distributed over four replicons, allowing for intricate and multilayered metal responses. To gain organic mercury resistance in CH34, broad-spectrum *mer* genes were introduced in a previous work via conjugation of the IncP-1β plasmid pTP6. However, we recently noted that this CH34-derived strain, MSR33, unexpectedly showed an increased resistance to other metals (i.e., Co2+, Ni2+, and Cd2+). To thoroughly investigate this phenomenon, we resequenced the entire genome of MSR33 and compared its DNA sequence and basal gene expression profile to those of its parental strain CH34. Genome comparison identified 11 insertions or deletions (INDELs) and nine single nucleotide polymorphisms (SNPs), whereas transcriptomic analysis displayed 107 differentially expressed genes. Sequence data implicated the transposition of IS*1088* in higher Co2+ and Ni2+ resistances and altered gene expression, although the precise mechanisms of the augmented Cd2+ resistance in MSR33 remains elusive. Our work indicates that conjugation procedures involving large complex genomes and extensive mobilomes may pose a considerable risk toward the introduction of unwanted, undocumented genetic changes. Special efforts are needed for the applied use and further development of small nonconjugative broad-host plasmid vectors, ideally involving CRISPR-related and advanced biosynthetic technologies.

**Keywords:** *Cupriavidus*; heavy metals; genomic islands; genomic rearrangements; metal resistance genes

### **1. Introduction**

Since life appeared on Earth some 3.7 billion years ago, microorganisms have undergone molecular changes to adapt (i.e., respond to selection) to the harsh conditions of their natural habitats, including extreme temperatures, pH, salinity, UV and ionizing radiation, and heavy metals [1]. In addition, global fluctuations in the composition of the atmosphere, the oceans, and the earth crust have elicited genomic changes in microbes over eons of time and hence contributed to microbial diversity [2]. Certain microorganisms had been adapted to heavy metals and radionuclides prior to human appearance. However, during the past few hundred years, anthropogenic influences have forced microbes to also adapt to pollutants previously nonexistent (xenobiotics) [3–6], as well as to increased concentrations of heavy metals and radionuclides [7–9].

Members of the beta-proteobacterial genus *Cupriavidus* are prime examples of microbial endurance, possessing a variety of genomic islands involved in the resistance to heavy metals or the degradation of aromatics or xenobiotics [10–20]. They all typically display a bipartite chromosomal structure with one chromosomal replicon bearing the marks of a plasmid-type maintenance and replication system henceforth called "chromid". In addition, most *Cupriavidus* strains carry one or two dispensable megaplasmids with a size of 100 kb or more. The model organism for heavy metal resistance, *Cupriavidus metallidurans* strain CH34, carries two megaplasmids, pMOL28 and pMOL30. Together, the four CH34 replicons encode resistance markers for a plethora of heavy metals including copper, nickel, zinc, cobalt, cadmium, chrome, lead, silver, gold, mercury, caesium, selenium, strontium, and uranium [12,21–25]. These resistances are mainly related to a variety of metal reduction and efflux systems [26–28]. Aromatics degradation, on the other hand, is carried out by various bacterial multicomponent mono- and di-oxygenases solely encoded by genes on the chromosome and chromid [29].

In an effort to improve the inorganic and organic mercury resistance of *C. metallidurans* CH34 and thus improve its utility in cleaning up mercury-contaminated environments, the IncP-1β plasmid pTP6, providing additional *mer* genes [30], was introduced in CH34 by biparental mating, leading to strain MSR33 [31]. These extra *mer* genes are part of a transposon, Tn*50580*, which is necessary for broad-spectrum (organomercury) resistance, including two genes, *merG* and *merB*, not native to CH34 (i.e., CH34 only contains two narrow-spectrum mercury resistance *merRTPADE* operons, one on each megaplasmid, conferring resistance to inorganic mercury). MerB plays a key role in methylmercury degradation through its unique ability to cleave the carbon-mercury bond in methylmercury and the subsequent shuttling of ionic mercury to MerA to reduce it to the less harmful elemental mercury.

In comparison to its parental strain CH34, strain MSR33 became 240% more resistant to inorganic mercury and gained resistance to methylmercury by incorporating the previously nonpresent *merBG* genes. Other metal resistances (as tested for chrome and copper) were seemingly unaffected, and the pTP6 plasmid was stably maintained for over 70 generations under nonselective pressure [31]. However, when we recently tested the resistances for both strains to additional metals (i.e., cadmium, cobalt, and nickel), we noted a significant increase of metal resistance for strain MSR33 compared to CH34 (this study). As this was fully unexpected, since the only difference between the two strains should be an additional *mer* gene dosage, implicating only an improved mercury resistance, we set out to investigate the reasons for this phenomenon. Considering the genome plasticity [32–34] and the intricate relationships between metal resistance loci [24] in *C. metallidurans* CH34, we decided to sequence the full genome of strain MSR33 and compare the sequence of its replicons with the corresponding replicons of the parental strain CH34 and with plasmid pTP6. We also performed microarray-based expression analysis on both CH34 and MSR33 gene sets to determine whether genomic differences could be correlated with differences in the expression of individual genes.

### **2. Materials and Methods**

### *2.1. Strains and Culture Conditions*

*Cupriavidus metallidurans* strains CH34 and MSR33, obtained from respectively the SCK•CEN (Mol, Belgium) and the Univesidad Católica del Norte (UCN) (Antofagasta, Chile) culture collections, were cultivated at 30 ◦C and 200 revolutions per minute (rpm) on a shaker in dark, aerobic conditions in a Tris-buffered mineral medium (MM284) [35] with 0.4% (*w*/*v*) succinate as the sole carbon source. *Escherichia coli* JM109 pTP6 (also obtained from the UCN culture collection) was cultured at 37 ◦C on an M9 minimal medium [36] supplemented with 0.4% (*w*/*v*) glucose.

### *2.2. Synthetic Construct Generation*

The *merTPAGB*<sup>1</sup> gene cluster of pTP6 was amplified by PCR using Phusion High-Fidelity DNA polymerase (ThermoFisher, Aalst, Belgium) with primer pair pTP6mer\_Fw-Rv (Table S1). In tandem, the broad-host-range cloning vector pBBR1MCS-2 [37] was linearized by PCR with the same DNA polymerase and primer pair pBBR1MCS-2\_GA\_Fw-Rv (Table S1), providing homologous ends with the amplified *merTPAGB*<sup>1</sup> locus. These compatible PCR products were end-ligated using the Invitrogen GeneArt® Seamless Cloning and Assembly Enzyme Mix (ThermoFisher). After transforming *E. coli* DG1 with the ligation mix and selection on Lysogeny Broth (LB) with 50 μg/mL kanamycin (Km), four randomly chosen transformants were tested by DNA digestion and fragment sizing for correct plasmid construction holding the *merTBAGB*<sup>1</sup> genes. The plasmid gene construct of one transformant was further verified by sequencing its insert using forward and reverse cloning primers (Table S1) prior to electroporation into *C. metallidurans* CH34 on an Eppendorf 2510 Electroporator (Eppendorf, Aarschot, Belgium) using conditions as described previously [38].

### *2.3. Estimation of Bacterial Tolerance to Metals*

Strains CH34 and MSR33 were grown overnight in MM284 liquid media with 0.4% (*w*/*v*) succinate as the sole carbon source and thereafter used as a pre-inoculum (1% *v*/*v*) for a freshly prepared 200 μL culture supplemented with increased concentrations of Hg2+ (from 0.0625 mM to 8 mM), Cd2+ (from 0.25 mM to 8 mM), and increasing steps (1 mM) of Co2+ or Ni2+ (from 5 mM to 20 mM). Cultures for metal contact were grown in microtiter plates at 30 ◦C on a rotary shaker at 120 rpm. Metal ion solutions were prepared from soluble salts of analytical grade (CdCl2, HgCl2, CoCl2·6H2O, and NiCl2·6H2O) in double-deionized water and were filter-sterilized before use. The lowest metal concentration that prevented growth after 48 h (i.e., showing no growth as measured at OD600 by a CLARIOstar microplate reader (BMG LabTech, Offenburg, Germany)), was considered the minimum inhibitory concentration (MIC) (Table 1 and Table S2). All MIC analyses were performed using biological triplicates.

**Table 1.** Minimal inhibitory concentration (MIC) of heavy metals for *Cupriavidus metallidurans* strains MSR33 and CH34.


<sup>1</sup> ND: not determined.

### *2.4. Plasmid Copy Number Determination*

Single-copy (i.e., "replicon-unique") genes were taken as representatives of the chromosome (*cadA*), chromid (*zniA*), and plasmids pMOL30 (*nccA*), pMOL28 (*cnrA*), and pTP6 (*merG*). Primer pairs were designed to amplify 150 bp of each gene (Table S1). Real-time PCR was performed on a 7500 Applied Biosystems Fast Real-Time PCR System (ThermoFisher) using QiaGen RT<sup>2</sup> Sybr Green Rox qPCR Mastermix (ThermoFisher), 20 ng of MSR33 or CH34 genomic DNA as a template, and 0.2 μM of each primer. To reduce nonspecific amplification, we incubated this mixture at 95 ◦C for 10 min as part of a hot start PCR setup. Next, a 40-cycle amplification and quantification protocol (15 s at 95 ◦C, 15 s at 58 ◦C, and 30 s at 60 ◦C) was performed with a single fluorescence measurement for each cycle. Finally, a melting curve program (15 s at 95 ◦C, 60 s at 60 ◦C, 30 s at 95 ◦C, and 15 s at 60 ◦C) was carried out. Plasmid copy numbers for both strains were determined using the absolute method (allowing estimates of both the absolute and relative number of plasmids per cell), following earlier described protocols [39]. Standard curves were created with 7500 Fast Software v2.3 of Applied Biosystems (Foster City, CA, USA), using serial (10-fold) dilutions of genomic DNA in a linear range from 20 ng to 0.2 pg. The qPCR efficiencies were calculated from slopes of the log-linear portion of the calibration curves from the equation *E* = 10(1/*slope*) . Using the linear equation obtained from each calibration curve, log DNA copy numbers were derived by intersecting the obtained *Ct* values. All analyses were done in triplicate.

### *2.5. Illumina Sequencing and Assembly*

MSR33 cells were grown overnight in MM284 liquid medium, and genomic DNA was extracted by using the Qiagen QIAamp DNA Mini Kit (ThermoFisher), following the instructions of the manufacturer. DNA quantity and quality were measured using a DropSense (Trinean, Piscataway, NJ, USA). Genome sequencing was performed on a HiSeq 2500 apparatus (Illumina, San Diego, CA, USA) using 2 × 250 bp paired-end reads. The reads were trimmed using the Trimmomatic tool [40] and their quality assessed using in-house Perl and shell scripts in combination with SAMtools [41], BEDTools [42], and a Burrows–Wheeler aligner with maximum exact matches (bwa-mem) [43].

The entire genome of *C. metallidurans* CH34 was sequenced and largely annotated [12,22]. Sequences for a chromosome (NC\_007973.1), a chromid (NC\_007974.2), and the large plasmids pMOL28 (NC\_007972.2) and pMOL30 (NC\_007971.2) were all obtained from GenBank and used as a reference for the assembly and annotation of trimmed sequences of the MSR33 genome. The DNA sequence of pTP6 plasmid, also obtained from Genbank (AM048832), was used for sequence comparisons between the CH34 and MSR33 sequence data sets. The full *C. metallidurans* MSR33 genome sequence (this study) is available from the NCBI Sequence Read Archive (SRA) under accession number PRJNA493617.

### *2.6. Total RNA Isolation and Microarray*

*Cupriavidus metallidurans* MSR33 and CH34 cells were both cultivated in triplicate on MM284 medium supplemented with 0.4% (*w*/*v*) succinate. Cell samples were taken at the middle exponential phase (OD600 0.6–0.7) and centrifuged at 16,000 × *g* and 4 ◦C for 5 min. Pellets were quick-frozen with liquid nitrogen and kept at −80 ◦C for further analysis. Total RNA extraction was performed as described previously [24] by using an SV Total RNA Isolation System kit (Promega Benelux, Leiden, the Netherlands) according to the manufacturer's recommendations. Samples were cleaned and concentrated using a Nucleospin RNA cleanup XS kit (Macherey-Nagel, Düren, Germany). Concentrated RNA samples (10–20 μg) were retrotranscribed using the Invitrogen SuperscriptTM Direct cDNA Labeling System (ThermoFisher) and labeled by incorporation of Cy3-dCTP (ref PA53021, control condition) and Cy-5dCTP (ref PA55021, experimental condition) by Pronto!TM Long Oligo/cDNA Hybridization Solution (supplied with the Corning® Pronto! Universal Microarray Hybridization kit from Merck/Sigma-Aldrich, Overijse, Belgium), following the manufacturer's instructions. The microarrays we used were designed with 60-*mer* probes for 6205 Open Reading Frames that were spotted in triplicate onto glass slides (UltraGPS, Corning, NY, USA) using a MicroGrid system (BioRobotics, Cambridge, UK) at the microarray platform at SCK•CEN (Mol, Belgium). The spotted slides were cross-linked and placed in the presoaking solutions from the Pronto Kit (Promega, Madison, WI, USA). Analyses were performed on RNAs retrieved from CH34 and MSR33 cells, using respectively Cy3-dCTP and Cy5-dCTP incorporation and determination of Cy3/Cy5 signal intensity ratios. Labeled cDNA was resuspended in the universal hybridization buffer (Pronto kit), mixed, and added to the spotted slide for overnight hybridization at 42 ◦C in a Tecan HS4800 Pro hybridization station (Tecan Group Ltd., Männedorf, Switzerland). Afterwards, the slide was washed according to Pronto kit's protocol. Slides were scanned (at 532 and 635 nm) using the GenePix Personal 4100A microarray scanner (Molecular Devices, San Jose, CA, USA). All post-hybridization analyses were performed as described before [24]. In brief, spot signals were qualified using GenePix Pro v.6.0.1 software, and raw median density data were imported into R version 3.3.2 (https://cran.rstudio.com/) for statistical analysis using the LIMMA package version 2.15.15 (http://bioinf.wehi.edu.au/limma/), as available from Bioconductor (https://bioconductor.org). Background correction, normalizations, *t*-statistics, and *p*-value corrections were done as before [24]. Only log-transformed expression results with a *p*-value > 0.05 were considered for data interpretation (Table S2).

### **3. Results and Discussion**

### *3.1. The Influence of Plasmid pTP6 on Increased Heavy Metal Resistance in MSR33*

The *C. metallidurans* strains CH34 and MSR33 have the same genetic background, but MSR33 has, compared to its parental strain CH34, an extra plasmid 54 kb in size [31]. This plasmid, pTP6, is a broad-host-range IncP-1β plasmid originally isolated from mercury-polluted sediments [30]. It carries a transposon with *mer* genes that are not native to CH34 (i.e., *merG*) that encode an organomercurial transporter, and a pair of duplicate *merB* genes that encode periplasmic organomercurial lyases (Table S3). These additional *mer* genes in strain MSR33, via plasmid pTP6, grant this strain a 2.4-fold increased resistance for Hg2+ and a 16-fold increased resistance for CH3Hg+ [31]. In our hands, we noted a much-improved Hg2+ resistance for strain MSR33, with a 10-fold increase in comparison to its parental strain CH34 (Table 1). As an added note, in contrast to Rojas et al. [31], who performed MIC analyses on solid media, we performed our MIC analyses in a liquid medium, increasing metal bioavailability. Hence, sensitivity (i.e., the MIC (Hg) for CH34) was 0.05 mM in Rojas et al.'s study [31] but 0.01 mM in our study]. Surprisingly, when we tested MICs for the metals cadmium, nickel, and cobalt, we found a 2-fold increased resistance in MSR33 to Cd2+ and Co2+ and a 1.2-fold increased resistance to Ni2+ (Table 1).

The only genes on pTP6 with relevance to metal resistance are mercury resistance genes situated on transposon Tn*50580* as the two clusters *merR*1*TPAGB*<sup>1</sup> and *merR*2*B*2*D*2*E* [30]. All other genes are typical for the backbones of self-transmissible and promiscuous IncP-1 plasmids from subgroups α and β and are involved in replication (*trfA*, *ssb*), plasmid maintenance, partitioning, control (*kfrABC*, *incC*, *korABC*, *kluAB*, *klcAB*, *kleABEF*, and a remnant of the resolvase gene *parA*), conjugal transfer (*traCDEFGHIJKL*), mating pair formation (*trbABCDEFGHIJKLMNO*), and transposition (*tniABQR* as part of Tn*50580*). Two more genes, *upf30.5* and *upf31.0*, are located downstream of *trbP* and encode, respectively, a putative outer membrane protein and a site-specific methylase (Figure S1). Except for the *upf* genes and the plasmid maintenance, partitioning, and control genes, pTP6 genes have at least one counterpart on one of the replicons of CH34. Taken together, we did not expect the pTP6 genes, other than the above-mentioned *mer* genes, to play any significant role in the augmented metal resistance of strain MSR33. Nonetheless, we decided to generate a new synthetic construct by cloning the *merTPAGB*<sup>1</sup> gene cluster in the low copy number broad-host-range cloning vector pBBR1MCS-2 [37]. This small plasmid only contained two genes, *rep* and *mob*, involved in, respectively, plasmid replication and mobilization, as well as a kanamycin resistance marker (KmR). Strain CH34 transformed with this new construct reached the same level of inorganic mercury resistance as the MSR33 strain but did not show an increase in cadmium resistance (Table 1). From this we deduced that the *mer* genes of pTP6 exerted a positive effect on host resistance to mercury. However, neither the concomitant increase of cadmium resistance in MSR33 nor its higher resistance to nickel and cobalt could be readily explained by the presence of the auxiliary, pTP6-associated mercury resistance genes in this strain.

We also determined the effect of various mixtures containing both mercury and cadmium on the growth of strains CH34 and MSR33. In general, the combination of the two metals was expected to be more toxic to cells than the corresponding metals alone. Indeed, strain CH34 was capable of growing in up to 1 mM of Cd2+ when combined with 6.25 μM of Hg2+, and cellular growth was diminished beyond these threshold metal concentrations (Table S4). Instead, strain MSR33 was capable of growing in up to 4 mM of Cd2+ when combined with 6.25 μM of Hg2+ (Table S4). However, even in combination with mercury, cadmium resistance in MSR33 was still twice as high as the cadmium resistance in CH34. Moreover, strain MSR33 showed much higher tolerance to mercury in combination with cadmium, and growth was only affected at the threshold metal concentrations of 125 μM Cd2+ and 100 μM Hg2+ (Table S4).

To exclude the possibility that the increased resistance in strain MSR33 to Hg2+ and Cd2+, either alone or in combination, was a mere effect of gene dosage, we also determined the plasmid copy number for all replicons in strain MSR33 and strain CH34 using quantitative PCR (Table S5). The presence of the pTP6 plasmid in strain MSR33 (plasmid copy number (*PCN*) = 1.8) did not alter

the relative copy numbers for the chromid and plasmids pMOL28 and pMOL30 with respect to the calculated chromosomal copy number taken as a reference. In addition, the Tn*50580* transposon carrying the broad spectrum *mer* gene cluster on plasmid pTP6 did not transpose (verified by genome sequence analysis present in Section 3.2).

The CH34 strain carries on its genome a total of four *mer* gene clusters: *merRTPA* on the chromosome, one complete *merRTPADE* on both plasmids pMOL28 and pMOL30, and one truncated *merRT*Δ*P*, also on pMOL30 [12]. One could argue that the presence of a single copy of pTP6 in strain MSR33 raises the number of *mer* genes in strain MSR33, with one unit for genes *T*, *P*, *A*, *D,* and *E*, and two units for gene *R* (a second *merR* gene is located on the right-hand part of Tn*50580* on pTP6 [30] (Table S6), but the *merG* or *merB* genes of pTP6 were not considered here as they only play a role in organomercurial resistance). Considering the calculated PCN values of all replicons, the theoretical abundance of these genes increased by roughly 50–78% (Table S6). *MerR* and *merD* are transcriptional regulators that compete for the same operator sequence in the *merR–merT* intergenic region. The *merT* product is an inner membrane protein involved in the transport of Hg2+ ions into the cell cytoplasm. The *merP* product is a small periplasmic Hg2+-sequestering protein that shuttles Hg2+ to the mercurial reductase MerA, which converts it into the significantly less toxic Hg(0) that then is allowed to leave the cell by passive diffusion. The *merE* product, finally, is another inner membrane protein and may play a role in the uptake of both CH3-Hg<sup>+</sup> and Hg2+. All these genes appeared to be intact, and there was no reason for us to assume that any of the multiple-copy genes would be dysfunctional or, with respect to each other (from gene to gene or copy to copy), would be differently transcribed or expressed (owing to limitations of gene-specific primer or probe design for multiple copies of these genes in qPCR or hybridization-based microarray procedures, no *mer* gene-specific expression data are available). A plausible explanation for the positive effect of the pTP6 *mer* genes on host mercury resistance could thus lay in the stoichiometry of *mer* gene products, particularly those involved in Hg2+ sequestration (*merP*) and transport (*merT/merE*). Nevertheless, the increased resistance to Cd2+ in MSR33 cannot be readily explained by a stoichiometric change in *mer* gene products. Also, when a single *merTPAGB* was introduced on a small plasmid into strain CH34, the increased Hg2+ resistance was still there, but the increased Cd2+ resistance was no longer seen (Table 1). From this we had to conclude that the MSR33 genetic background, besides the extra plasmid pTP6, may actually have differed from the genetic background of its parental strain CH34. In other words, the MSR33 genome had undergone genetic changes leading to an improved resistance to cadmium and possibly also to other metals. Such a genomic adaptation appears to be common to IncP-1 plasmid backbones [44]. In order to get to the bottom of this we decided to determine the DNA sequence of the entire genome of strain MSR33 and register in detail which genetic changes occurred with respect to the reference genomes of strain CH34 and plasmid pTP6.

### *3.2. The C. metallidurans MSR33 Genome Showed Multiple Insertions or Deletions and Single Nucleotide Polymorphisms*

The whole-genome resequencing of MSR33 (since the known genome sequences of strain CH34 and plasmid pTP6 served as references, we considered this effort as a resequencing project) revealed a total of eight insertions and three deletions (Table 2), and nine single nucleotide polymorphisms (Table 3), all changes being located predominantly across the four replicons of the CH34 backbone, with only one genomic change occurring in plasmid pTP6 (Table 2, Figure 1). Most of the insertions (six out of eight) were found to be related to IS*1088*, an insertion element belonging to the IS*30* family with a typical size range of 1000–1250 bp [45]. It should be noted at this point that the CH34 genome indigenously harboured nine copies of IS*1088*, distributed on its chromosome and chromid but not on its megaplasmids [12], bringing the total of IS*1088* copies in the MSR33 genome to 15.

The majority of the new IS*1088* copies in the MSR33 genome were located on the chromosome and chromid, where all IS*1088* copies indigenous to CH34 resided, but one IS*1088* copy transposed into the *cnrY* gene (Rmet\_6205) of pMOL28 (Table 2). This gene was part of the *cnrYXHCBAT* locus involved in the inducible cobalt and nickel resistance in strain CH34, and encoded the anti-sigma factor CnrY that tethered, in conjunction with the sensor protein CnrX, the sigma factor CnrH, but released it in the presence of Ni2+ or Co2+ [46,47]. The sigma factor CnrH promoted transcription of its own locus *cnrYXH*, but also of the structural locus *cnrCBA*, encoding a resistance nodulation division (RND)-driven efflux system [12,48]. The inactivation of the *cnrY* gene in MSR33 by IS*1088* inevitably led to the constitutive derepression of *cnrCBAT* transcription and explained the increased cobalt and nickel resistance we observed for MSR33 (Table 1) (see also gene expression results in Section 3.3). A similar phenomenon was previously seen for spontaneous mutants of a pMOL30-less CH34 derivative, strain AE126 [35], which showed a significantly increased resistance to cobalt and nickel [49] and which was later acknowledged as being an IS- and frameshift-mediated inactivation of *cnrY* and *cnrX* [50].

The other five genes affected by the insertion of an IS*1088* element were Rmet\_0312 (*nptA*) and Rmet\_2860 (*tauB*), lying on the chromosome; and Rmet\_4160 (*pelF*), Rmet\_4867 (*acrA*), and Rmet\_5682 (*nimB*), lying on the chromid (Table 2). The first three genes encode proteins with general cellular functions, and their inactivation is very unlikely to affect heavy metal resistance in strain MSR33. The fourth gene, *acrA*, encodes a membrane fusion protein and is part of an intact *acrABC* operon whose gene products form a tripartite multidrug efflux system. The last gene, *nimB*, is involved in efflux-mediated heavy metal resistance, encoding also a membrane fusion protein resembling other membrane metal-binding fusion proteins in structure and function (e.g., CzcB, CnrB, CusB, and ZneB) by forming a periplasmic bridge between the cytoplasmic porter and the outer membrane channel [48]. Nonetheless, taking also into account that the *nimA* gene is already inactivated in strain CH34 (and MSR33) by the presence of the insertion sequence element IS*Rme3* [12], it is hard to see how the inactivation of *nimB* would result in the increased metal resistance we observed in strain MSR33. Two insertions were not attributed to IS*1088*. One appeared to be the result of a Tn*3*-related transposition event affecting gene Rmet\_5388, encoding a tentative ApbE-like lipoprotein, while the other concerned an unknown mutational event in gene Rmet\_5508 resulting in the insertion of a nucleotide triplet (+CTT) (Table 2). This gene encodes a long-chain fatty-acid CoA-ligase and also underwent a triplet deletion (-CGG) just a few nucleotides downstream of the triplet insert. As a combined result, the actual change at the protein level remained perfectly in-frame and gave a protein of the same length, but led to an altered peptide sequence at positions 149–153 (i.e., Xxx-Leu-**Arg**-**Phe**-**Ala**-Gln-Xxx in CH34 to Xxx-Leu-**Phe**-**Ala**-**Lys**-Gln-Xxx in MSR33 (amino acidic sequence change from **Arg**-**Phe**-**Ala** to **Phe**-**Ala**-**Lys**). The third deletion in MSR33 occurred in plasmid pTP6, effectively destroying the genes *upf30.5*, *upf31.0*, and *parA*, immediately preceding Tn*50580*. Except for *cnrY* and *nimB*, none of the aforementioned genes are in any way associated with metal resistance.


