Chromosome: *CHR1*; chromid: *CHR2*; arrows show positive → or negative ← gene orientation; \* www.genoscope.cns.fr.

**Figure 1.** Correlation map between Insertions or Deletions (INDELs), Single nucleotide polymorphisms (SNPs), and transcriptional changes found for *C. metallidurans* MSR33. Concentric circles (ring) displayed from the outside inwards: (**A**) In ring 1, chromosome and chromid size scale in megabases (Mb) using a 20 kilobase (kb) window; (**B**) plasmids size scale in kb using a 2 kb window. In ring 2, position of insertion (blue), deletion (pink), and SNPs (mustard). INDELs and SNPs listed in Tables 2 and 3. In ring 3, each dot represents a single gene basal expression, overexpression (log2 *ratio* > 1, green), or repression (log2 *ratio* < −1, red), with a *p*-value < 0.05. Plotted genes listed by function in Table S7. Circos plot created with Circa (http://omgenomics.com/circa).

In addition to the eight insertions and three deletions in the MSR33 genome, sequence analysis revealed the presence of nine single nucleotide polymorphisms (Table 3). Two of those occurred in intergenic regions on the chromosome, without apparent disruption of gene regulatory elements, while the other seven occurred in protein-encoding genes (four on the chromosome and three on the chromid). Except for the "silent" mutation (i.e., no aa change) at position 645608, these Single nucleotide polymorphisms (SNPs) caused aa substitutions in the corresponding gene products (Table 3). No SNPs were detected in any of the plasmids. Remarkably, gene Rmet\_5508 lying on the chromid (CHR2) once again was a target for mutation, displaying two SNPs in the immediate vicinity of the aforementioned triplet insertion and deletion in this gene (Table 3), bringing the full change of this

region from **RFA**QK**P**AY**V** to **FAK**QK**T**AY**E** (changes are in bold and underlined). It is uncertain whether these protein changes would have any effect on the cellular and metabolic functions in MSR33.

Taken together, of the 11 Insertions or Deletions (INDELs) and nine SNPs identified by the whole genome resequencing of the *C. metallidurans* strain MSR33, only the *cnrY* inactivation by IS*1088* and the concomitant derepression of *cnrB* (see above) may be directly linked to the observed augmented heavy metal resistance in this strain, at least for Co2+ and Ni2+ (see above). None of the other genomic changes seemed to play a role in this augmentation. The augmented resistance for Cd2+ in strain MSR33 (Table 1), however, remains a puzzle. The fact that such augmentation for Cd2+ was only noted for MSR33 with an altered genome (i.e., with 17 INDELs and six SNPs), but not in CH34 transformed with plasmid pBBR::*merTPAGB*<sup>1</sup> (Table 1) strongly indicates that the MSR33 genetic background was at play. In basic terms, bacterial resistance to toxic metals depends on two cellular processes, metal binding and metal transport, with the former generally being an intrinsic part of the latter. It is well established that many proteins or peptides that mediate the transport, buffering, or detoxification of metal ions in living cells have metal-binding domains (MBDs) in which certain amino acid residues (e.g., cysteine), as well as their structural layout and relative position to each other, play a key role in metal selectivity and specificity [51,52]. While some of these proteins might be highly metal-specific, other proteins follow a more relaxed, nonspecific mode of metal binding. For instance, divalent metal uptake in *C. metallidurans* is governed by a battery of redundant transporters that display a minimal degree of metal cation selectivity [53]. Depending on the environment, this may lead to a cytoplasmic pool of unsolicited metal ions that at some point, particularly when reaching a toxic threshold, need to be removed by the cell. In *C. metallidurans* this was done by one of three efflux systems: Cation diffusion facilitators (CDF), P-type ATPases, and the earlier mentioned RND-driven transenvelope transporters (HME-RND). Their main task in *C. metallidurans*, because of this bacterium's adaptation to metal-rich environments, was to balance the cytoplasmic and periplasmic concentrations of unwanted transition metals by entering the cellular arena and going into competition for metal cations with the "frivolous" metal uptake systems. Interestingly, all three types of metal efflux systems seemed to possess some degree of frivolity toward metal ions as well, albeit perhaps not as outspoken as for the metal uptake systems. The *C. metallidurans* CzcD exporter (Rmet\_5979), for instance, allowed as a CDF protein Zn2+, Cd2+, or Co2+ as a substrate [54], whereas the DmeF and FieF exporters (Rmet\_0198 and Rmet\_3406) displayed as CDF family members broad metal specificity for Zn2+, Cd2+, Co2+, and Ni2+ [55]. It is worth mentioning that disruption of the *dmeF* gene in strain CH34 dramatically lowered the resistance for Co2+ (but not for Zn2+, Cd2+, and Ni2+), indicating a complex interplay between the DmeF exporter and the CzcCBA and CnrCBA efflux pumps (possibly partially obscured by the action of other metal resistance systems) [55]. Moreover, CDF proteins can play diverse roles and may possess different metal ion selectivity depending on the environmental conditions (i.e., by adjusted Kd values for certain metals) [56]. In addition, the eight metal resistance-related P1B-type ATPases currently identified in strain CH34 can be subdivided into two groups according to their substrate profile [28]: Those that extrude Cu+ and Ag<sup>+</sup> (CupA and CupF) and those that extrude Zn2+, Cd2+, Co2+, or Pb2+ (ZntA, CadA, PbrA, and CzcP). These exporters mainly differ in the presence of unique amino acid sequences in their transmembrane MBDs, hence defining their metal specificity. But even within a subgroup, differences may exist in terms of metal affinity. For example, CzcP encoded by plasmid pMOL30 is unable to mediate Zn resistance on its own but rather augments the metal exportability of the ZntA, CadA, and PbrA exporters [57]. In a similar fashion, the five active HME-RND efflux systems in strain CH34 displayed a limited substrate spectrum, pumping out either the monovalent metal cations Cu<sup>+</sup> and Ag+ (CusA, SilA) or the divalent metal cations Zn2+, Ni2+, and Co2+, with occasionally also Cd2+ (ZniA, CnrA, CzcA) [28] (Figure 2). In such HME-RND systems, two steps of heavy-metal extrusion were discerned, the periplasmic and the transenvelope efflux (Figure 2). Each step involved the interaction of metals with MBDs within the Membrane Fusion Protein (MFP) and RND proteins. Sometimes, the delivery of periplasmic metal ions to the typical C3B6A3-complex is facilitated by a small periplasmic metallochaperone, as is the

case for the *E. coli* CusCBFA system [58] (and likely, based on CusF aa sequence similarities, also the CusCBAF complex of strain CH34). Little is known about the substrate specificity of the metal-binding proteins of HME-RND efflux complexes. Apparently, metal-induced conformational changes in the C3B6A3-complex are required in order to create a proper metal-guiding C3B6A3 channel for metal export to take place [48,59–61].

**Figure 2.** Transcriptional changes in *C. metallidurans* MSR33 with respect to CH34, with both strains grown under equal and nonselective conditions (see methods). Bar graphs show the significantly (*p*-value < 0.05) higher expression (log2 ratio > +1) and lower expression (log2 ratio < −1) of MSR33 genes (with CH34 gene expression levels as reference). Transcriptional changes from genes pertaining to all replicons are depicted. Genes indicated with an \* are part of pMOL30.

As mentioned, it is not inconceivable that the genetic changes in MSR33 instigated cellular conditions or pleiotropic effects that were generally favourable for Cd2+ detoxification and hence led to the observed improvement in Cd2+ resistance. Possibly, this involved the temporal recruitment of one or more metal binding export proteins, from known metal resistant systems or from hitherto unknown export systems, able to bind Cd2+. The transition metals cadmium and mercury belong to Group 12 of chemical elements in the periodic table, together with zinc and copernicium. Although these four metals differ in significant respects, they also have common properties. Particularly, Cd and Hg are similar in their outer shell electron configuration (d10s2) and atomic radius (ca. 150 pm), and their cations both have a high affinity for sulfhydryl groups in cellular compounds and proteins (i.e., in methionine and cysteine residues). From this perspective, competition between Cd2+ and Hg2+ for certain MBDs cannot be excluded. Bacterial evolution and adaptation to new or rapidly changing environments implies a delicate balance between the safeguarding of genome integrity and the tolerance for genome instability. A too-rigid genome inevitably will lead to the demise of innovative power and hence adaptability of the host, whereas a too plastic or "fluid" genome may lead to disadvantageous mutations and cell growth arrest, or even cell death. This balance between beneficiary and perilous change in a bacterial genome also relates to the general fitness and the energy household of its host. Members of the genus *Cupriavidus*, and in particular *C. metallidurans*, appear to be masters in adaptation as they are home to a wide variety of habitats, often in extreme conditions [62–65]. The introduction of the 54 kb plasmid pTP6 into a strain already carrying two large replicons of 3.9 and 2.6 Mb (chromosome and chromid, respectively) and two megasized plasmids of 171 and 234 kb (pMOL28 and pMOL30, respectively) could be seen as a serious additional burden to the host regardless of whether or not phenotypical or physiological changes occur.

Because the charting of genomic changes in MSR33 with respect to its parental strain CH34 did not provide us with any clues or direct evidence on the involvement of certain genetic loci or of any of the known metal resistance determinants on the observed augmented Cd2+ resistance, we decided to compare the basal gene expression data for strains CH34 and MSR33 using RNA microarray technology in an attempt to associate their gene expression profiles with strain-specific physiological behaviour, with a focus on differentially expressed (DE) genes that might be involved in the cellular detoxification of heavy metals such as Cd2+.

### *3.3. Transcriptional Analysis of Strain MSR33*

Strains CH34 and MSR33, which were equally grown in nonselective conditions without any metal-related stress, were investigated for basal gene expression levels, and their expression profiles were compared. A total of 107 DE genes showed statistically significant changes in their expression (Figure 2). Affected genes pertained to the main chromosome (55 genes), the chromid (36 genes), and the plasmids pMOL30 (3 genes) and pMOL28 (13 genes) (Figure 2). In general terms, the products of these 107 genes could be grouped according to their predicted annotated function [12]: Catalytic function (35 genes), transport (24 genes), transcriptional regulation (11 genes), recombination (6 genes), movement and chemotaxis (9 genes), and miscellaneous (22 genes) (Table S7). Most of these genes had a higher expression in strain MSR33, with only 16 genes in strain MSR33 showing a lower expression level. What is immediately striking is that all genes of the *cnrYXHCBAT* operon on plasmid pMOL28 had a significantly higher expression (Figure 2), with log2 fold changes ranging from four to six. As we know from MSR33 whole genome sequence analysis, this high expression of the *cnr* locus in strain MSR33 was the direct result of IS*1088*-mediated *cnrY* inactivation and hence derepression of the *cnr* locus, explaining the increased resistance we observed for strain MSR33 to Co2+ and Ni2+ (see previous sections). Equally noticeable is the complete absence of an altered expression of the *czc* locus on pMOL30, strongly suggesting that the augmented Cd+2 resistance we see for strain MSR33 was independent of this locus. This, in fact, corroborates earlier findings about the pMOL30-less CH34 derivative AE126 (which, like MSR33, also has an IS*1088*-mediated inactivated *cnrY* gene on the remaining plasmid pMOL28): Vandecraen et al. [50] showed, next to a heightened resistance

to Zn2+, Co2+, and Ni2+, a 2-fold increased resistance to Cd2+. Adding another level of complexity, when strains AE126 and AE104 (a CH34 derivative lacking both pMOL28 and pMOL30 plasmids) [35] were transformed with pTP6, these strains, like MSR33, gained an improvement in Cd2+ resistance, albeit to a lesser extent (Rojas LA, personal communication).

This would indicate that the augmentation of Cd2+ resistance in MSR33 by pTP6 conjugation should be seen as a layered process brought about by multiple factors and possibly diverse mechanisms supporting each other. We cannot say at this point what these mechanisms precisely are and how and when they are triggered, as we have no information about the genomic changes in pTP6 conjugants of AE104 and AE126 (as pTP6 conjugation in CH34 causes genomic changes, this would most likely also be the case for pTP6 conjugation in strains AE104 and AE126, but not necessarily involving the same genomic changes). Clearly, further studies are needed to understand the augmented metal resistance in pTP6 conjugants of CH34 and its derivatives, including (1) the resequencing of pTP6-conjugated AE104 and AE126 strains and (2) the extensive RNAseq-based genetic response analyses for a wider range of heavy metals in all three pTP6-conjugated strains. The additional possibility that some of the observed genetic changes were already introduced to the recipient CH34 strain prior to conjugation with pTP6 cannot be entirely excluded. Lastly, the plasmid-curing procedures used to obtain strains AE104 and AE126 (i.e., applying mitomycin C, nalidixic acid, or hydroxyurea to growing CH34 cells [35]) may also have had mutagenic effects or may have induced transposition activity. In this respect, it would be best, in the frame of future studies, to resequence these strains as well.

A very high log2 fold difference in the expression of >4 was also noted in strain MSR33 for the Rmet\_4229 gene, a *dctA* paralogue whose product was functionally annotated as a C4-dicarboxylate transporter and which is unlikely to have any connection to metal detoxification or resistance, and gene Rmet\_2382, originally identified in CH34 as a transposase-encoding *tnpA* gene (IS*1088*) (Figure 2). Intermediate high log2 fold changes of >2 were seen in strain MSR33 for another 17 genes (Table S7), whereas the remaining 72 genes showed a log2 fold change between one and two. None of these genes is thought to be involved in metal binding, metal detoxification, or metal resistance. Among the genes with lowered expression in strain MSR33, we noted the *pelG* gene (Rmet\_4161), which is part of the *pelABCDEFG* operon required to produce an extracellular polysaccharide that has been implicated in biofilm development [66]. Our sequence data confirmed that the IS*1088* element transposed into the *pelF* gene (Table 2), thereby disrupting expression of the *pelG* gene. This could explain the complete lack of biofilm formation in strain MSR33 reported to us by P. Alviz in a personal communication.

In conclusion, the genome of MSR33 underwent eight insertions, three deletions, and nine SNPs. At least seven of the insertions were due to the action of mobile genetic elements, with their presence fully confirmed by sequence data (implicating IS*1088* in six cases), whereas one small insertion and all three deletions in strain MSR33 may have been the result of DNA recombination or transposition events. The *C. metallidurans* genome is known to be ridden with a very high number of mobile genetic elements, with 57 IS elements, 19 other transposable elements, and 16 genomic islands for its type strain CH34 [12,32,34]. In concordance with this genomic fluidity, *C. metallidurans* displays a highly versatile metabolism and an inherent ability to inhabit a variety of harsh environments [9,62–65]. This adaptability has not come about overnight but is the wonderful result of microbial evolution over long periods of time. In a time in which large chunks of DNA were retrieved from the environment (e.g., by plasmid transfer or gene exchanges), adaptation was brought about by DNA mutations and natural selection and molecular inventions took place, steadily moulding the genome into its present large (6.9 Mb) and highly malleable form, providing the bacterium with a vast array of possibilities for rapid genetic responses (hence its well-chosen epithet as "Master Survivalist") [12]. However, tinkering with this hugely evolved and dynamic genome holds intrinsic dangers. Although the plasmid pTP6 was maintained stably in strain CH34 (i.e., MSR33) for over 70 generations under nonselective conditions [31], it has now become clear from our study at hand that the receiving host's genome underwent multiple changes in the form of 11 INDELs and 9 SNPs, affecting the physiology and heavy metal resistance of the host. It would be wrong to point the finger at the extra plasmid as the "usual suspect" for these genetic changes, but rather we hold the actual process of conjugation responsible. Conjugative interaction appears to be a strong stimulus for transposition [67–69], and hence it is easy to envisage that, as a result of conjugation procedures, some elements of the extensive mobilome of *C. metallidurans* (with nearly 100 mobile elements) were triggered into action and "moved around", causing genetic changes that led to clearly perceptible but also less visible (and less understood) effects alike. The take-home message here is that the genetic engineering of bacteria with large complex and dynamic genomes should be carried out with much caution and that a strong preference should be given to the new generation of small broad-host-range cloning vectors and CRISPR-based technologies nowadays available [70–74].

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2073-4425/9/11/551/s1. Figure S1: Genetic map of plasmid pTP6; Table S1: Primers designed for this report; Table S2: Transcriptional changes observed in MSR33 versus CH34 under basal conditions; Table S3: Homology analysis of *mer* gene products present in plasmid pTP6; Table S4: Effect of various mixtures of Hg2+ and Cd2+ on *C. metallidurans* strains MSR33 and CH34 growth; Table S5: Plasmid copy number (PCN) for *C. metallidurans* strains MSR33 and CH34, determined by quantitative PCR; Table S6: *mer* gene occurrences on the replicons of strains CH34 and MSR33; Table S7: Expression changes of *C. metallidurans* MSR33 with respect to *C. metallidurans* CH34, both grown under nonselective conditions.

**Author Contributions:** F.A.M., P.J.J., R.V.H., P.M., and L.A.R. conceived and designed the experiments; F.A.M., A.J., and A.P. performed the experiments; F.A.M., R.V.H., and P.M. analyzed the data; L.A.R. and P.J.J. contributed reagents, materials, and analysis tools; F.A.M., P.J.J., R.V.H., P.M., and L.A.R. wrote the paper.

**Funding:** SCK•CEN EE0630012-09 (to P.J.J., R.V.H., A.J., A.P., and P.M.), CONICYT/FONDECYT 11130117 (L.A.R.), and CONICYT/BC-PhD 72170403 (F.M.)

**Acknowledgments:** Authors acknowledge research funding by SCK•CEN EE0630012-09 (to P.J.J., R.V.H., A.J., A.P., and P.M.), CONICYT/FONDECYT 11130117 (L.A.R.), and CONICYT/BC-PhD 72170403 (F.M.). Genome sequencing was provided by MicrobesNG (http://www.microbesng.uk), supported by BBSRC grant number BB/L024209/1. We thank D. Vallenet and Z. Rouy of Génoscope (Centre National de Séquençage, Evry, France) for implementing additional features in MaGe and their essential advice in genome annotation.

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

### **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/).
