Zinc-Induced Transposition of Insertion Sequence Elements Contributes to Increased Adaptability of Cupriavidus metallidurans

Bacteria can respond to adverse environments by increasing their genomic variability and subsequently facilitating adaptive evolution. To demonstrate this, the contribution of Insertion Sequence (IS) elements to the genetic adaptation of Cupriavidus metallidurans AE126 to toxic zinc concentrations was determined. This derivative of type strain CH34, devoid of its main zinc resistance determinant, is still able to increase its zinc resistance level. Specifically, upon plating on medium supplemented with a toxic zinc concentration, resistant variants arose in which a compromised cnrYX regulatory locus caused derepression of CnrH sigma factor activity and concomitant induction of the corresponding RND-driven cnrCBA efflux system. Late-occurring zinc resistant variants likely arose in response to the selective conditions, as they were enriched in cnrYX disruptions caused by specific IS elements whose transposase expression was found to be zinc-responsive. Interestingly, deletion of cnrH, and consequently the CnrH-dependent adaptation potential, still enabled adaptation by transposition of IS elements (ISRme5 and IS1086) that provided outward-directed promoters driving cnrCBAT transcription. Finally, adaptation to zinc by IS reshuffling can also enhance the adaptation to subsequent environmental challenges. Thus, transposition of IS elements can be induced by stress conditions and play a multifaceted, pivotal role in the adaptation to these and subsequent stress conditions.


INTRODUCTION
Bacteria continuously evolve to survive different environmental challenges and the field of evolutionary biology has long been interested in the interplay between generation of genetic diversity and natural selection favoring the best adapted organism. The central dogma that evolution proceeds through natural selection of heritable mutations was first proposed by Darwin (Darwin, 1910). In addition, however, Darwin also suggested that environmental stress might generate a diversity of genetic variants, upon which subsequent natural selection works (Darwin, 1910). The claim that mutation rate might be influenced by stress is, however, still under debate (Rosenberg and Hastings, 2004;Roth, 2011). The classic experiments of Luria and Delbrück (1943) and Lederberg (1989) demonstrated for the first time that mutations can arise without the influence of selective stress. However, the lethal selections they used could likely not have detected mutations induced by selective conditions, and others have described genetic systems in which selective conditions indeed seemed to increase the mutation rate (Shapiro, 1984;Cairns et al., 1988;Hall, 1990;Foster, 2007;MacLean et al., 2013;Ram and Hadany, 2014). In addition, the coupling of potential mutation-generating systems to local events such as stress-induced transcription of a specific subset of genes (Wright, 2004) might form an example of stress-directed mutation where the mutation is specifically induced by the stress condition that the mutation relieves (Cairns et al., 1988). However, non-selected mutations readily occur during selective conditions disputing the existence of non-random mutations (Foster, 2007).
One frequently encountered type of mutation results from the hopping of transposable elements from the donor to the target site, potentially leading to important phenotypic changes (Mahillon and Chandler, 1998;Mahillon et al., 1999). Insertion sequence (IS) elements, which are widely distributed in the genomes of bacteria as well as higher organisms (Siguier et al., 2006(Siguier et al., , 2014, are of interest as they constitute an important driving force for genome plasticity (Kazazian, 2004). On the one hand, intra-genomic transposition of these small elements (<2.5 kb), which generally encode no other functions than those involved in their mobility, can both inactivate or activate adjacent genes by physically interrupting coding sequences or encoding regulatory sequences, respectively. The activation can be caused by a promoter encoded by the element and directed outward to the adjacent gene (Ciampi et al., 1982;Zhang and Saier, 2009;Wang and Wood, 2011;Saier and Zhang, 2014) or by the formation of a hybrid promoter in which the −35 region at the end of the IS element is correctly situated close to the −10 region of the adjacent gene (Jaurin and Normark, 1983;Prentki et al., 1986). On the other hand, homologous IS elements within the chromosome can become substrates of the cell's recombination machinery, which can result in complex chromosomal rearrangements (e.g., inversions, duplications, or deletions) depending on the orientation of the involved IS elements (Louarn et al., 1985;Schneider et al., 2000a;Parkhill et al., 2003).
Selective conditions can promote the activation and redistribution of IS elements that exist in the chromosome. Indeed, stress factors were found to induce the transposition of mobile genetic elements (MGEs) in plants (McClintock, 1984;Studer et al., 2011), flies (González et al., 2010), yeast (Stanley et al., 2010), and bacteria (Ohtsubo et al., 2005;Drevinek et al., 2010). The acquisition of phenotypic traits caused by transposition of IS elements has been well-documented, particularly in bacteria. Different conditions such as UV light, low frequency magnetic fields or sub-inhibitory concentrations of antibiotics were shown to promote transposition of IS elements in Escherichia coli (Eichenbaum and Livneh, 1998;Del Re et al., 2003;Lartigue et al., 2006). Transposition of IS elements in response to a selective condition has also been described in other gram-negative genera, e.g., Pseudomonas and Burkholderia in response to high temperature, conjugation or oxidative stress (Ohtsubo et al., 2005;Christie-Oleza et al., 2009;Drevinek et al., 2010). Active redistribution of IS elements was also observed in the gram-positive Deinococcus radiodurans after exposure to radiation (Pasternak et al., 2010) and in the gram-positive Bacillus subtilis under competence-inducing conditions (Takahashi et al., 2007).
Cupriavidus metallidurans strains are, next to metalcontaminated soils (Diels and Mergeay, 1990;Brim et al., 1999;Goris et al., 2001), being increasingly recovered from other anthropogenic environments not typified by metal contamination (Van Houdt et al., 2012;Mijnendonckx et al., 2013), including medically-relevant sources (Coenye et al., 2005;Langevin et al., 2011;D'Inzeo et al., 2015). They are studied for their resistance and adaptation to toxic levels of metal ions. Type strain C. metallidurans CH34 counteracts metal ion toxicity via a battery of resistance mechanisms, including transporters belonging to the Resistance Nodulation cell Division (RND), the cation diffusion facilitator (CDF) and the P-type ATPase families, with an important role for its two mega-plasmids as most of the heavy metal resistance mechanisms are encoded on either pMOL28 or pMOL30 (Mergeay et al., 1985(Mergeay et al., , 2003Nies and Silver, 1995;Nies, 1999Nies, , 2003Monchy et al., 2007;von Rozycki and Nies, 2009;Janssen et al., 2010). In addition, C. metallidurans CH34 harbors 21 distinct IS elements belonging to 10 different IS families (Mijnendonckx et al., 2011). Out of the 21 distinct IS elements, active transposition has been observed for ISRme1, ISRme3, IS1086, IS1087B, IS1088, and IS1090 (Dong et al., 1992;Collard et al., 1993;Grass et al., 2000;Ma-e-Talat, 2000;Schneider et al., 2000b;Tibazarwa et al., 2000). In fact, active involvement of IS1087B in metal resistance has been shown for C. metallidurans AE126, which is a derivative of CH34 sensitive to zinc as it only carries plasmid pMOL28 and not plasmid pMOL30 (Collard et al., 1993;Tibazarwa et al., 2000). Spontaneous zinc-resistant AE126 mutants constitutively expressing cnrCBAT coding for the RND-driven efflux system CnrCBAT appear when AE126 is grown in the presence of 0.8 mM Zn 2+ (Collard et al., 1993;Grass et al., 2000;Tibazarwa et al., 2000). The cnrCBAT operon is regulated by the upstream cnrYXH locus coding for the membrane-bound anti-sigma factor CnrY, the sensor protein CnrX, and the ECF (Extracytoplasmic Function) family sigma factor CnrH. The latter is released from the CnrYX transmembrane anti-sigma factor complex in the presence of inducers (Ni 2+ or Co 2+ ) and subsequently promotes transcription of both cnrYXH and cnrCBAT (Grosse et al., 2007;Trepreau et al., 2011Trepreau et al., , 2014Maillard et al., 2015). Analysis of two of such derivatives indicated that the inactivation of the gene coding for the anti-sigma factor CnrY (by IS1087B and a frameshift mutation), and concomitant release of CnrH increased transcription of the structural cnrCBAT cluster and led to increased (non-specific) Zn 2+ efflux (Collard et al., 1993;Grass et al., 2000;Tibazarwa et al., 2000).
These observations indicate that IS transposition may play a dynamic role in the genetic adaptation of C. metallidurans AE126 to toxic zinc concentrations. Therefore, in this study, we thoroughly scrutinized the mutations leading to, and the genetic determinants essential for increased zinc resistance in AE126. The involvement of IS elements was dissected via the identification of the concerned IS elements, their target sites, and the effect elicited by their transposition as well as via analysis of transposase expression upon metal ion exposure.

Determination of Growth and Minimal Inhibitory Concentration
The susceptibility of C. metallidurans strains to different heavy metals was determined via minimal inhibitory concentrations (MICs). The strains were cultivated in triplicate by inoculating 2 ml MM284 supplemented with different concentrations of heavy metals with 20 µL of a stationary phase culture at 30 • C. The MIC is defined as the lowest concentration of heavy metal that will inhibit visible growth of the culture after 2 days of incubation. Growth experiments were performed by 1:100 dilution of stationary phase cultures in fresh MM284 and subsequently for each sample (biological replicates), 200 µL was added to a 96-well white cell culture plate (Nunc Flat-bottom, Thermo Scientific) which was placed into a CLARIOstar R (BMG LABTECH). Plates were incubated at 30 • C with shaking for 48 h and at 30-min time intervals, the optical density at 600 nm was measured. The growth data were fitted by the model of Baranyi and Roberts (1994) using the Microsoft Excel add-in package DMFit (Institute of Food Research, Norwich, United Kingdom). Statistical comparison of the growth curve parameters was performed using a one-way ANOVA analysis, followed by a post-hoc Tukey test.

Isolation of Zinc-Resistant Mutants
C. metallidurans AE126 was cultivated in MM284 at 30 • C up to stationary phase, 10 9 cells were pelleted and cell suspensions (100 µL) of a serial ten-fold dilution in saline were spread on MM284 agar plates containing a final concentration of 0.8 mM Zn 2+ and incubated at 30 • C. Colony forming units (CFU) were counted after day 2 up to day 5 and survival frequency was calculated as viable cell count on MM284 0.8 mM Zn 2+ agar plates divided by viable cell count on MM284 agar plates. The variance-to-mean ratio and mutation rate in a fluctuation assay with 20 independent cultures was calculated according to Luria and Delbrück (1943) and using Fluctuation Analysis CalculatOR (FALCOR) (Hall et al., 2009), respectively.

Analysis of IS Transposition
Transposition of IS elements in the regulatory locus of the cnr operon ( Supplementary Figure 1) was identified by colony PCR (DreamTaq DNA polymerase, Thermo Scientific) with primers CnrFw and CnrRv (Supplementary Table 1) and sequencing. A sacBR IS trap (Gay et al., 1985;Dong et al., 1992) was used to compare the adaptation potential of AE126 R4 , which carries an ISRme5 copy inserted in cnrX in the same transcriptional orientation as cnrYX, with AE126. Sucroseresistant derivatives of C. metallidurans AE126 and AE126 R4 carrying pJV240 (Gay et al., 1985;Dong et al., 1992) were selected on LB plates without NaCl and supplemented with 10% (w/v) sucrose and 1500 µg/ml kanamycin. The insertion of ISRme5 in pJV240 was identified by colony PCR with primer pairs ISRme5_Fw and SacR_Rv or pISRme5_Rv and SacR_Rv, respectively (Supplementary Table 1).

Whole Genome Gene Expression Microarrays
The whole-genome expression profile of AE126 R2 and AE126 R3 were compared to its parental strain AE126 by whole genome oligonucleotide microarrays. The strains were cultivated by inoculating 30 ml of MM284 in biological triplicates with 300 µL of a stationary phase culture at 30 • C. These subcultures were allowed to grow until an OD 600 -value of around 0.6 was reached. Next, cells were harvested by centrifugation for 2 min at 10,000 rpm and the bacterial pellets were flash frozen by immersion into liquid nitrogen. The bacterial pellets were kept frozen at −80 • C until total RNA extraction was performed using the SV Total RNA Isolation system (Promega Corporation). The quantity of extracted RNA was measured using a NanoDrop TM 1000 spectrophotometer (Thermo Scientific). The RNA quality was determined with a Bioanalyzer (Agilent 2100 Electrophoresis Bioanalyzer Agilent Technologies). Only extracted RNA with a RNA integrity number (RIN number) of higher than eight was accepted for further analysis (Schroeder et al., 2006). Twenty micrograms of RNA were reverse transcribed following the instructions provided with the Pronto kit (Promega). Microarray slide spotting, RNA labeling, hybridization, and data analysis were performed as previously described . The full description of the microarray data have been deposited at the Gene Expression Omnibus website (http://www.ncbi.nlm.nih.gov/geo/) under accession number GSE74091.

cnrH Mutant Deletion Construction
The cnrH gene of C. metallidurans AE126 was amplified by PCR (Phusion High-Fidelity DNA polymerase) with primer pairs cnrH_Fw-Rv (Supplementary Table 1), providing HindIII/EcoRI restriction sites. Afterwards, these PCR products were cloned as a HindIII/EcoRI fragment into the mobilizable

Identification of cnrCBAT Transcription Start Site
The transcription start site of cnrCBAT in wild type AE126 and in the two zinc-resistant derivatives AE126 cnrH::tet R1 and AE126 cnrH::tet R2 were identified using the cRACE method (Maruyama et al., 1995;Dallmeier and Neyts, 2013). The strains were cultivated by inoculating 30 ml of MM284 with 300 µL of a stationary phase C. metallidurans AE126, AE126 cnrH::tet R1 , or AE126 cnrH::tet R2 culture at 30 • C. The cultures were allowed to grow until an OD 600 -value of around 0.6 was reached. Next, cells were harvested by centrifugation for 2 min at 10,000 rpm and the bacterial pellets were flash frozen by immersion into liquid nitrogen. The bacterial pellets were kept frozen at −80 • C until total RNA extraction was performed using the SV Total RNA Isolation system (Promega Corporation). The quantity of extracted RNA was measured using a NanoDrop TM 1000 spectrophotometer (Thermo Scientific). The RNA quality was determined with a Bioanalyzer (Agilent 2100 Electrophoresis Bioanalyzer Agilent Technologies). Only extracted RNA with a RNA integrity number (RIN number) of higher than eight was accepted for further analysis (Schroeder et al., 2006). One microgram of RNA was reverse transcribed following the instructions provided with the GoScript TM Reverse Transcription System (Promega) with a gene-specific 5 ′ -phosphorylated oligonucleotide cnrC_R1 (Supplementary Table 1). Subsequently, three volumes of TE buffer (10 mM Tris-HCl and 1 mM ethylenediaminetetraacetic acid, pH 8.0) containing 4 µg/ml RNase A (Promega) was added to the reaction. Successively, 28 µL of the resulting cDNAs was circularized by 20 U of T4 RNA ligase (Thermo Scientific) in the presence of 15% (w/v) polyethylene glycol (PEG4000; Thermo Scientific) in a total volume of 50 µL at 37 • C for 60 min. Next 1.5 U of Pfu DNA polymerase (Thermo Scientific) was added to the reaction at 37 • C for 30 min employing its 3 ′ -5 ′ exonuclease activity to remove unreacted residual first-strand primers and cDNAs. Finally, a 2 µL aliquot was directly used as template for the amplification by PCR (DreamTaq DNA polymerase) with primer pair cnrC_F1-R2 (Supplementary Table 1), the PCR product was cloned by TOPO cloning (Thermo Scientific) and the transcription start site was identified by sequencing.

Monitoring of Transcription
For quantitatively assaying endogenous promoter activity, C. metallidurans AE126 carrying luminescence reporter constructs (

Isolation and Phenotypic Characterization of Zinc-Resistant AE126 Derivatives
The minimal inhibitory concentration (MIC) of Zn 2+ for C. metallidurans AE126, which is type strain CH34 cured from its pMOL30 mega plasmid encoding the czc operon (cadmium/zinc/cobalt resistance), is much less than its parent, being 0.2 and 12 mM, respectively. However, AE126 has the capability to spontaneously acquire additional zinc resistance, i.e., the ability to grow in the presence of 0.8 mM Zn 2+ at a frequency of ∼10 −6 ( Table 2; Collard et al., 1993). Interestingly, using a zinc plate assay two populations of zinc-resistant AE126 derivatives were observed that appeared at different incubation times. The first population, which arose after 2 days of incubation, showed a high variance-to-mean ratio (26 CFU) in a fluctuation assay with 20 independent cultures and a calculated mutation rate to zinc resistance of 0.105 ± 0.022 × 10 −7 per generation. However, new colonies emerged on the selective plate in the ensuing days and we defined derivatives specifically emerging at day 5 as the second population. A higher mutation rate to zinc resistance of 0.307 ± 0.056 × 10 −7 per generation and lower variance-to-mean ratio (3 CFU) was calculated for the second population. The lower mutation rate and high variance-to-mean ratio of the first population, with most plates yielding a variable number of mutant colonies while occasional plates showed a large number of mutants (so-called "jackpots"), suggested that the first population was pre-existing while the second population emerged on the plate itself. Cultured separately in liquid MM284 without zinc demonstrated similar growth rates for AE126 and zinc-resistant derivatives from the first (AE126 R1 ) and second (AE126 R3 ) population (Table 3). Moreover, the zinc-resistant derivatives exhibited a similar lag phase that was shorter than that of AE126 (Table 3). In addition, both derivatives grew equally fast in the presence of 0.8 mM Zn 2+ but displayed a prolonged lag phase compared to MM284 without zinc ( Table 3). Next to increased Zn 2+ resistance (a 10-fold increase of the MIC), the characterized zinc-resistant AE126 derivatives also displayed an increased resistance to Ni 2+ , Co 2+ and Cd 2+ (3, 2.5 and two-fold increase of the MIC, respectively). However, the level of zinc and cadmium resistance of these mutants is still considerably lower than those from the type strain C. metallidurans CH34 ( Table 2).

Genetic Characterization of Zinc-Resistant AE126 Derivatives
In previous work, only two zinc-resistant AE126 derivatives were analyzed (Collard et al., 1993;Grass et al., 2000). Therefore, we scrutinized cnrY and cnrX of 498 randomly selected zincresistant AE126 isolates, of which 296 belong to the first population and 202 belong to the second population, by PCR and sequencing to determine if cnrY and cnrX were indeed the sole targets leading to increased zinc resistance of AE126 derivatives (Supplementary Figure 1). This analysis confirmed that all 498 randomly selected isolates either had an IS element inserted or another mutation (point mutation, small insertions, or deletion) in cnrY or cnrX (Figure 1). Interestingly, however, the first population (46.3%) had a significantly lower contribution of IS elements than the second population (68.8%) as determined by a Fisher Exact test with a p < 0.05 (Figure 1), indicating that transposition of these IS elements might be promoted by the selective zinc condition.

Characterization of IS-Independent Zinc-Resistant AE126 Mutants
Regarding the IS-independent mutations in cnrYX, it is clear that only mutations resulting in inactivation of either CnrY or CnrX are selected in the zinc plate assays. The IS-independent mutations are dispersed throughout the cnrYX locus and include frameshift mutations, non-sense mutations and larger deletions. For instance, a 360-nucleotide deletion was observed including the complete cnrY gene.
Two zinc-resistant IS-independent AE126 derivatives, AE126 R2 (harboring a frameshift mutation caused by a single nucleotide deletion in cnrX) and AE126 R3 (harboring a five nucleotide 5 ′ -CGCGA-3 ′ duplication in cnrY), both belonging to the second population of zinc-resistant AE126 derivatives, were selected for whole-genome expression profiling to further  confirm the cnr-dependency underlying toxic zinc stress and subsequent acquired zinc resistance. Although, IS-independent mutations are discussed here, the five nucleotide duplication in cnrY of AE126 R3 might suggest an insertion and subsequent excision event of an IS element resulting in a direct target repeat in cnrY. Both AE126 R2 and AE126 R3 bear mutations leading to inactivation of the anti-sigma CnrYX complex, and wholegenome expression profiling showed 219 and 86 open reading frames (ORFs) that were significantly differentially expressed (>2-fold with an adjusted p < 0.05) in comparison with the parental AE126 strain, respectively (Supplementary Table 2). Twenty ORFs were commonly differentially expressed in both zinc-resistant derivatives. The mutated regulatory genes cnrY and cnrX were the most upregulated (11-and 15-fold in AE126 R2 , and 33-and 57-fold in AE126 R3 , respectively) followed by the structural locus cnrCBAT (nine-fold average for AE126 R2 and 18-fold average for AE126 R3 ). Transcription of cnrH was threeor six-fold upregulated in AE126 R2 and AE126 R3 , respectively (Supplementary Table 2). As such, it seems that constitutive overexpression of the cnr pump is responsible for the acquired zinc resistance. The high amount of other differentially expressed genes and the fact that most of these genes (93%) were unique for one of both zinc-resistant derivatives indicate that probably different mutations, not involved in the acquired zinc resistance, occurred during the selective exposure to Zn 2+ , which thus may comprise a more general response to stress with mutations randomly happening, supporting zinc-induced mutagenesis for the second population of zinc-resistant derivatives.
To further and unambiguously confirm the essential role of the cnrYXHCBAT locus in the capability to spontaneously acquire zinc resistance, complementation assays were performed in the zinc-sensitive and plasmidless C. metallidurans CH34 derivative, i.e., strain AE104 ( Table 2). The findings of Collard et al. (1993) that strain AE104 can never be evolved to withstand the toxic zinc concentration used in zinc plate assays (0.8 mM Zn 2+ ) was confirmed. Complementation of C. metallidurans AE104 with the complete cnr locus, thus both the regulatory and structural genes, resulted in the ability to acquire zinc resistance (frequency of 0.96 ± 0.38 × 10 −6 after 48 h exposure to 0.8 mM Zn 2+ ), whereas complementation of AE104 with the complete but modified cnr locus from AE126 R1 resulted in full resistance (frequency of 100%; Table 1). To test if the CnrH sigma factor could drive other potential resistance genes as well, the parental regulatory cnrYXH locus, and the mutated cnrYXH locus from AE126 R1 were introduced in AE104 but no zinc resistance could be acquired. Thus, the presence of the cnrCBAT genes encoding the RND-driven efflux system was essential to acquire zinc resistance ( Table 2). These complementation assays showed that the cnrYXHCBAT cluster is necessary for acquiring zinc resistance at high frequency in C. metallidurans AE104 or AE126 and that the inactivation of cnrYX and concomitant CnrH-mediated derepression of cnrCBAT is responsible for this zinc resistance.

Characterization of IS-Dependent Zinc-Resistant AE126 Mutants
Focusing further on the IS-dependent zinc-resistant mutants, ISRme5 appeared to be the main contributor to the obtained insertions both in the first (29.4%) and second (31.7%) population (Figure 1). Other identified IS elements included ISRme3, ISRme15, IS1090, IS1088, IS1086 and IS1087B. Comparing the contribution of specific IS elements in the zincresistant AE126 population to their abundance in the genome showed that some IS elements were clearly more represented than would be expected from their copy number (Figure 2). ISRme5 (four identical copies in the genome) had a significant higher contribution to both populations, and IS1087B (two identical copies in the genome) to the second population as determined by a Fisher Exact test with a p < 0.05. In addition, IS1088 (nine identical copies in the genome) was significantly more represented in the second than in the first population (Fisher Exact test; p < 0.05), but comparing its contribution to the second population to its abundance in the genome revealed no significant higher contribution than would be expected from its copy number (Figure 2).
Next, the identified IS insertion sites were scrutinized. A different use of target genes (i.e., cnrY or cnrX) was detected for the identified IS elements, although this had no influence on zinc resistance as determined by cell survival at 0.8 mM Zn 2+ and by MIC testing (data not shown). Most IS elements inactivated cnrY with the exception of ISRme5 and IS1086. In general, all IS elements, except IS1088, inserted exclusively in the same transcriptional orientation as the disrupted gene. No difference in target site use was detected between mutants from the first or second population. ISRme5 used six different insertion sites throughout the target locus and two were used preferentially (Figure 3).
A position-specific scoring matrix was derived from the multiple sequence alignment of all ISRme5 target sites (Figure 4). This position-specific scoring matrix was used to scan the genome sequence of C. metallidurans CH34 via the RSAT motif scan (Turatsinze et al., 2008) and resulted in numerous hits (Supplementary Table 3). Besides cnrX, other genes involved in regulation are found in the list composed of the top 100 hits, including genes encoding for transcriptional regulators, receptor proteins, a sensor histidine kinase, and a signal transduction protein. In addition, the family of transporters is with 21 genes also overrepresented in the motif scan. The other identified IS elements in cnrYX showed a more random insertion pattern although IS1087B preferentially inserted into the 5 ′ -region of the regulatory locus.

Characterization of cnrH-Independent Zinc-Resistant AE126 Mutants
To further analyze the adaptation capacities of C. metallidurans AE126 to toxic zinc concentrations, a cnrH deletion strain was created to assess if other resistance mechanisms than CnrHmediated expression of cnrCBAT could result in increased zinc resistance. No difference in zinc resistance was observed between the parental and cnrH deletion strain, however, as expected, the cnrH deletion resulted in a drastically decreased cobalt and nickel resistance ( Table 2). Interestingly, zinc-resistant AE126 cnrH::tet derivatives could be isolated after exposure to 0.8 mM Zn 2+ , but only after a minimum exposure time of 96 h and at a frequency 100-fold less compared with AE126 ( Table 2). Cultured separately in liquid MM284 without zinc demonstrated similar growth rates for AE126, AE126 cnrH::tet, and the zinc-resistant derivative AE126 cnrH::tet R1 (Table 3). In addition, again, the zinc-resistant derivate exhibited a lag phase that was shorter than that of AE126 cnrH::tet and AE126 (Table 3). Moreover, AE126 cnrH::tet R1 displayed the same growth rate in the presence of 0.8 mM Zn 2+ than the CnrH-dependent zinc-resistant AE126 R1 and AE126 R3 ( Table 3).
Phenotypic characterization of zinc-resistant AE126 cnrH::tet derivatives further indicated an increased resistance FIGURE 2 | Correlation between the copy number and transposition of IS elements in the cnrYX locus identified in zinc-resistant AE126 derivatives. Red and green bars represent the relative abundance of a specific IS element identified in the pool of IS-mediated zinc-resistant AE126 derivatives isolated after 2 and 5 days of incubation, respectively. Light gray bars represent the relative abundance of a specific IS element in the genome of C. metallidurans AE126. to Ni 2+ (22-fold) and Co 2+ (32-fold; Table 2), indicative for a role of the cnr locus. Genetic characterization of six independent zinc-resistant AE126 cnrH::tet derivatives revealed for five derivatives an insertion of ISRme5 at the same target site 25 bp upstream the cnrC start codon (Figure 5). One representative was further used and designated AE126 cnrH::tet R1 . The other derivative (AE126 cnrH::tet R2 ) carried an IS1086 insertion 29 bp upstream of the cnrC start codon (Figure 5). To exclude any other possible involvement of additional resistance determinants, complementation assays were performed where the complete but modified cnr operon of AE126 cnrH::tet and AE126 cnrH::tet R1 were transferred to C. metallidurans AE104, respectively. These complementation assays demonstrated that cnrYX cnrH::tetCBAT R mediated the increased zinc resistance ( Table 2). Furthermore, transcriptional reporter fusions with cnrC promoter fragments of AE126 cnrH::tet R1 (953 bp fragment harboring also a 3 ′ fragment of ISRme5 without its transposase promoter) and AE126 cnrH::tet R2 (853 bp fragment harboring also a 3 ′ fragment of IS1086 without its transposase promoter) confirmed the constitutive expression of cnrCBAT (Supplementary Figure 2). Therefore, the insertion of these IS elements in the intergenic region between cnrH::tet and cnrC resulted in a constitutively increased transcription of cnrCBAT, which was also confirmed by microarray analysis of AE126 cnrH::tet R1 (Supplementary Table 4). Since readthrough expression from the transposase promoter did not account for this increased expression, the transcription start site of cnrCBAT in AE126 cnrH::tet R1 (ISRme5 insertion) was identified and showed that expression is driven by an outward-directed promoter positioned at the 3 ′ extremity of ISRme5 (Figure 5). This could also be deduced for IS1086 based FIGURE 4 | Position-specific scoring matrix derived from the multiple sequence alignment of all target sites used by ISRme5. Eighty-seven target sites and the flanking DNA sequences have been aligned to generate the matrix shown. The sequences compiled are those adjacent to the left (−1 to −10) and right end (+1 to +10) of ISRme5. The central six nucleotides are those that would be duplicated on insertion. A sequence logo was generated from this matrix with WebLogo (Crooks et al., 2004).

Endogenous Promoter Activity of IS Elements
Based on their relative high abundance in the zinc-resistant derivatives of AE126, especially in the second population, ISRme5, IS1088 and IS1087B were selected for further investigation. To analyze the effect of metal ion stress on the endogenous promoter activity of the transposase of the three selected IS elements, transcriptional fusions between the promoterless tandem bi-cistronic gfp-luxCDABE reporter and fragments of ISRme5, IS1088, and IS1087B comprising the left inverted repeat up until the start codon of the transposase were constructed and analyzed. The promoters of all three IS elements were induced by Zn 2+ and Cd 2+ , but not by Ni 2+ and Co 2+ (Figure 6). Of the three endogenous promoters analyzed, the ISRme5 promoter was induced the most by Zn 2+ , and in lesser extent by Cd 2+ (1.7-fold for Zn 2+ and 1.4-fold for Cd 2+ after 4 h). A dose-dependent response was observed with the highest response for 100 µM Zn 2+ and 10 µM Cd 2+ , respectively (Figure 7). Higher doses of Zn 2+ or Cd 2+ could not be tested as they are detrimental to the growth of C. metallidurans AE126 strains. The IS1088 and IS1087B endogenous transposase promoters were induced 1.14 and 1.25-fold by Zn 2+ (after 4 h), respectively, and to a lesser extent by Cd 2+ .

Analysis of Upstream Flanking Genomic Sequence on ISRme5 Transposase Expression
For ISRme5, the most active element in response to zinc, the effect of the upstream flanking genomic sequence on transposase expression was scrutinized in more detail. C. metallidurans AE126 harbors three copies of the ISRme5 element on the chromosome and one copy on the chromid, which are identical at the sequence level. However, the transposase open reading frame contains no stop codon within the IS element, implying that the different ISRme5 transposases, encoded by the four copies, have different lengths generated by sequence specific insertion. This is not unique as more members of the IS481-family do not display a stop codon for the transposase open reading frame within their respective IS elements (Redenbach et al., 1996;Tauch et al., 1998). The transposase of the corresponding ISRme5a, b, c and d element consist of 333, 463, 321 and 321 amino acids, respectively. Two copies are located in genomic island CMGI-2, which is involved in hydrogenotrophy and the metabolism of aromatic compounds (Van Houdt et al., 2009;Mijnendonckx et al., 2011). Transcriptional fusions between the promoterless tandem bi-cistronic gfp-luxCDABE reporter and sequences starting 118, 166, 206 or 148 bp upstream the left inverted repeat up until the transposase start codon of all four ISRme5 copies were constructed and analyzed. Comparing the basic expression activity, it is clear that the genomic sequence upstream the left inverted repeat of all ISRme5 FIGURE 6 | Endogenous promoter activity of the ISRme5 (A), IS1088 (C), and IS1087B (B) transposase in response to Zn 2+ (100 µM), Ni 2+ (100 µM), Cd 2+ (10 µM) and Co 2+ (50 µM) after 4 h of incubation, calculated by normalizing the reporter signal RLU to cell density ratio (RLU/OD 600 ) to that of the control without added metals. *indicates a significant difference between the tested metal ion and the control, determined by a t-test with a p < 0.05. The average values of three independent experiments with standard deviations are shown.
copies exerts a negative effect on ISRme5 endogenous promoter activity. This is especially true for ISRme5b where almost no expression is detected, i.e., <0.5% (Figure 8). Furthermore, although the reporter construct composed of only the promoter region of ISRme5, thus without additional genomic sequences, is Zn 2+ -inducible, no Zn 2+ induction was detected for the reporter constructs composed of upstream genomic content of ISRme5a, ISRme5b and ISRme5d. The reporter construct of ISRme5c, however, was more than two-fold induced for Zn 2+ (Figure 8).

Enhanced Adaptation Potential to Other Stress Challenges
The effect of IS transposition in the cnrYX genes on the general fitness was analyzed. In this test, the parental AE126 and the zinc-resistant C. metallidurans AE126 R4 strain were used. AE126 R4 carries an additional ISRme5 element inserted in cnrX in the same transcriptional orientation as the cnr operon, yielding a continuous read-through ISRme5 transposase expression as whole-genome expression verified that expression of cnrYX is induced in CnrH-dependent zinc-resistant AE126 derivatives (Supplementary Table 2). The pJV240 plasmid was introduced into AE126 and AE126 R4 , and biological triplicates were subsequently exposed to 10% sucrose. Sucrose-resistant derivatives were isolated and the pJV240-encoded sacBR locus was screened by PCR to determine the involvement of ISRme5. AE126 R4 pJV240 displayed a higher survival frequency (4.78 ± 1.58×10 −6 ) than AE126 pJV240 (1.42 ± 0.30×10 −7 ) and ISRme5 was found to disrupt sacBR in sucrose-resistant AE126 R4 pJV240 derivatives (3/20) opposed to sucrose-resistant AE126 pJV240 derivatives (0/20).

DISCUSSION
MGEs have been shown to provide their hosts with a wide diversity of adaptive traits. Although, transposition of IS elements mostly results in detrimental mutations, a fraction of insertions can be beneficial and lead to increased survival. The simplest effect of IS transposition is the inactivation of a gene at a particular target site. In addition, IS transposition can also lead to altered expression of adjacently located genes by co-transcription or by providing a (hybrid) promoter or other protein binding elements located inside the IS element.
It was previously demonstrated that C. metallidurans AE126 could readily acquire increased zinc resistance (Collard et al., 1993). We observed that zinc-resistant AE126 derivatives arose at different time points after exposure to 0.8 mM Zn 2+ and all 498 tested derivatives had mutations targeting the regulatory locus of the RND-driven CnrCBA efflux system. The first population arose after an exposure time of 48 h and new derivatives kept appearing in the ensuing days with a second population defined as derivatives that arose after an exposure time of 120 h. The first population displayed a high variance-to-mean ratio, with most plates yielding a variable number of mutant colonies while occasional plates showed a large number of mutants (so-called "jackpots"), indicating the involvement of pre-existing mutations (Luria and Delbrück, 1943). Interestingly, the second population displayed a higher mutation rate and lower variance-to-mean ratio. Although, the latter was still three times the theoretical value of one when only stress-induced mutagenesis is considered, the much lower value and higher mutation rate are indicative for the involvement of zinc-induced mutagenesis in the second population. Furthermore, mutants from both populations have the same growth rate on zinc medium suggesting once more that the second population may originate from selection on the selective zinc plate. Adaptive mutation and accumulation of colonies throughout incubation time has been described for the activation of a promoterless phenol degradation operon in starving Pseudomonas putida. Although, phenol-utilizing mutants already appeared on phenol-minimal plates on day 2, the majority of the colonies began to appear on day 3 and continued to accumulate during the next 7 days (Kasak et al., 1997). In addition, mutations to ciprofloxacin resistance also continually occurred in non-dividing E. coli cells during a 7-day exposure to ciprofloxacin in agar and the late-appearing resistant mutants had no growth rate defect compared to wild type or earlyappearing mutants (Riesenfeld et al., 1997). Moreover, E. coli cells could utilize glycerol in the absence of Crp by activation of the silent glpFK glycerol utilization operon by transposition of IS5 upstream of glpFK. The transposed IS5 element harbors a permanent bend and IHF binding site at its 3 ′ -end that promoted the expression of the adjacently located glpFK operon (Zhang and Saier, 2011). When crp − cells of E. coli are incubated on solid glycerol minimal medium, Glp + colonies first appeared after 3 days of incubation and new colonies continued to appear in the ensuing days. However, wild type and crp Glp + cells plated on the same solid medium formed already colonies in less than 2 days, indicating that the Glp + mutants that arose from the crp cells on glycerol minimal medium plates were not present in the cell culture before plating (Zhang and Saier, 2011).
For acquiring zinc resistance at high frequency (10 −6 ), inactivation of anti-sigma factor CnrYX complex is essential to liberate CnrH and increase cnrCBAT expression. Although the efflux pump CnrCBAT is necessary to mediate zinc efflux, the role of the sigma factor CnrH in promoting cnrCBAT expression is not. In the absence of cnrH, expression of cnrCBAT and acquisition of zinc resistance could be mediated by the correct insertion of ISRme5 or IS1086 carrying an outwarddirected promoter positioned at their 3 ′ extremity, albeit to a lower frequency (10 −8 ), and only after a minimum exposure time of 96 h. These mutants displayed the same growth rate on zinc medium than CnrH-dependent zinc-resistant AE126 derivatives from the first population, which were isolated after 48 h, suggesting that these CnrH-independent zinc-resistant derivatives may originate from selection on the selective zinc plate. In addition, using the cRACE method, the wild type promoter transcription start site of the cnrCBAT structural resistance genes was found to be located downstream of cnrH as previously reported by Grass et al. (2000) and not upstream as previously reported by Tibazarwa et al. (2000).
Exposure of AE126 to 0.8 mM Zn 2+ revealed transposition of seven different IS elements in the cnrY or cnrX gene (occurrence of transposition ISRme5 > IS1088 > ISRme3 > IS1087B > IS1090 > IS1086 > ISRme15). Active IS transposition in C. metallidurans was already observed for ISRme1, ISRme3, IS1086, IS1087B, IS1088 and IS1090 (Dong et al., 1992;Collard et al., 1993;Grass et al., 2000;Ma-e-Talat, 2000;Schneider et al., 2000b;Tibazarwa et al., 2000;Mijnendonckx et al., 2011), but not for ISRme5 and ISRme15. Overall, IS elements contributed more to the second population that arose after five days of incubation on plate. ISRme5 was the main IS contributing to both populations, whereas IS1088 and IS1087B showed a higher contribution to the second compared with the first population. Most of the transposed IS elements (>95%) inserted exclusively in the same transcriptional orientation as cnrCBAT, although inactivation of CnrYX is sufficient to acquire zinc resistance. It is unclear why a given site should be used in only one orientation (Lampe et al., 1998). Most transposases did not show insertion site specificity, only for ISRme5 two hotspot loci were identified that were used preferentially and specifically by ISRme5. Scanning the genome of C. metallidurans CH34 with the generated target sequence matrix resulted in numerous hits, with various hits in genes involved in regulation and transport processes, indicating the evolutionary power that ISRme5 might exert as inactivation of those genes mostly results in significant phenotypic alterations. In some cases transposases prefer a particular DNA structure, most notably bent DNA, as target site (Craig, 1997). It seems that this was the case for the IS1087B transposase as it mostly inserted in the 5 ′ -region of cnrY, indicating that the IS1087B transposase does not recognize target DNA in any sequence-specific manner but that certain sequences surrounding the insertion site may allow local structures to form that are conductive for IS1087B insertion. There is precedence for this phenomenon in the non-random use of target sequences of P transposable elements, where the local target preference depends more on DNA structure than on primary sequence (Liao et al., 2000).
Our study showed for the first time that the endogenous promoters of the ISRme5, IS1088 and IS1087B transposases can be induced by metal ions. This increased promoter induction was specific for zinc and cadmium but not nickel and cobalt. The dual zinc and cadmium induction could be explained by the fact that both metal ions have many similar (bio)chemical characteristics. Other examples of transposase transcriptional expression induced by environmental stimuli include Tn4652 and IS1246 in P. putida KT2440 by zinc, cadmium and nickel but not cobalt (Haritha et al., 2009), and ISFtu1 and ISFtu2 by spermine in Francisella tularensis (Carlson et al., 2009). Elevated transpositional activity of several IS elements was also observed at high temperature in Burkholderia multivorans (Ohtsubo et al., 2005). However, the induced transpositional activity in the examples above did not mediate enhanced tolerance to the inducer. Although, the ISRme5 transposase promoter internal to the element is inducible by zinc and cadmium, the genomic region upstream IS integration has a strong regulatory effect on the expression of the IS element. The upstream genomic region resulted in reduced promoter activity of all ISRme5 copies, especially for ISRme5b, and only the ISRme5c copy was found to be metal-inducible. In addition, the transposase of ISRme5 (IS481 family member) has no stop codon within the element. Consequently, the ISRme5b transposase has a prolonged carboxy-terminal end compared with the other copies, generated by its insertion site, which might influence its enzymatic stability and hence, could contribute to the possible diminished transposition activity of this copy compared to the other ISRme5 elements. Therefore, the evolutionary success of ISRme5, as probably for all IS elements, depends not only on the element itself, but also on its genomic integration site. Taking into account that many transposases preferentially act in cis (Mahillon and Chandler, 1998), which reduces the probability that transposase expression from a particular element would activate transposition of related copies elsewhere in the genome, the higher basal expression level and zinc-inducibility of the ISRme5c transpose indicate that most ISRme5-mediated CnrYX inactivation events are probably the result of ISRme5c transposition. Repression or enhancement of promoter activity by the upstream flanking sequence is also observed for other MGEs (Lavie et al., 2004).
Bacteria respond to environmental changes by adapting the expression of key genes, however, such reprogramming requires time and energy. Interestingly, all tested zinc-resistant derivatives showed a considerably shorter lag phase in liquid MM284 without zinc compared to wild type, rendering the evolved population better suited to quickly respond to other environmental challenges. For instance, in conditions where the lag phase is long, adaptations reducing the lag phase are highly beneficial as the evolved population can start their exponential growth earlier and consume most of the environmental nutrients before the wild type exits the lag phase (Oxman et al., 2008). Furthermore, adaptation to zinc by IS reshuffling can have an additional impact on the adaptation to subsequent stress challenges. Insertion of an IS element in cnrYX in the same transcriptional orientation as the cnr operon will probably result in read-through transposase expression as whole-genome expression indeed verified that expression of cnrYX is induced in CnrH-dependent zinc-resistant AE126 derivatives. Thus, redistribution of IS elements yielding IS-dependent zinc-resistant derivatives, hence harboring an additional copy of the IS element in a constitutively transcribed region, might result in an increased pool of the transposase in the cell and subsequent enhance the adaptation potential to new environmental challenges, such as was shown here for ISRme5 and SacB-mediated sucrose hypersensitivity. The higher survival frequency of the ISdependent zinc-resistant derivative compared to wild type can be explained by (i) an increased pool of ISRme5 transposase resulting in ISRme5-mediated sacBR disruption, (ii) the possible insertion and subsequent excision of ISRme5, resulting in various possible frameshift deletions (Ooka et al., 2009) or duplications (Mahillon and Chandler, 1998) inactivating sacBR, however, the presence of these mutations was not investigated, and (iii) the uncharacterized genetic background of the ISdependent zinc-resistant derivative, which likely, consistent with other derivatives from the second population, harbors more mutations.
To summarize, it was shown that zinc induced the expression of IS elements and that these elements mediate the adaptation of C. metallidurans to zinc stress by modulating its transcriptional network or by providing outward-directed mobile promoters.

AUTHOR CONTRIBUTIONS
Conceived and designed the study: MM, AA, and RVH. Performed the experiments: JV. Analyzed the data: all authors. Wrote the paper: JV, AA, and RVH.

FUNDING
This work was supported through a Ph.D. grant for JV by SCK•CEN.

SUPPLEMENTARY MATERIAL
The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmicb.
Black line represent the amplicon generated with the primer pair Cnr_Fw and Cnr_Rv used for screening the cnrYXH structural locus.

Supplementary Figure 2 | Promoter activity of the cnrC promoter of AE126
cnrH::tet R1 and AE126 cnrH::tet R2 . Data are represented as fold change of the report signal normalized to cell density (RLU/OD 600 ) obtained for AE104 pGLR1-P ISRme5cnrC -gfpluxCDABE (953 bp fragment harboring also a 3 ′ fragment of ISRme5 without its transposase promoter) and AE104 pGLR1-P IS1086cnrC -gfpluxCDABE (853 bp fragment harboring also a 3 ′ fragment of IS1086 without its transposase promoter) compared with the control AE104 pGLR1-P cnrC -gfpluxCDABE. Only open reading frames (ORFs) that were significantly expressed (>2-fold with an adjusted p < 0.05) are shown. (A) ORFs differentially expressed in both derivatives, (B) ORFs differentially expressed in AE126 R2 and (C) ORFs differentially expressed in AE126 R3 with ND not determined. The full description of the microarray data have been deposited at the Gene Expression Omnibus website (http://www.ncbi.nlm.nih.gov/geo/) under accession number GSE74091.

Supplementary
Supplementary Table 3 | Scan of the C. metallidurans CH34 genome sequence (+ for forward and -for reverse strand) with the position-specific scoring matrix obtained from the multiple sequence alignment of the ISRme5 target sites in 87 zinc-resistant AE126 derivatives using RSAT. The weight score Ws = log (P(S|M)/P(S|B)) is the probability for the sequence segment S to occur according to the motif model M, whereas P (S|B) is the probability for the same sequence S to occur under the background model B. The weight score is thus the log-ratio of the likelihood of S in these two respective models. Top 100 targets are shown.
Supplementary Table 4 | Transcriptional expression of the cnrCBAT cluster, determined by whole-microarray analysis, in two zinc-resistant AE126 derivatives (AE126 R2 and AE126 R3 ) and in a zinc-resistant AE126 cnrH::tet derivative (AE126 cnrH::tet R1 ). Data are represented as fold change relative to the parental AE126 and AE126 cnrH::tet, respectively. The full description of the microarray data have been deposited at the Gene Expression Omnibus website (http://www.ncbi.nlm.nih.gov/geo/) under accession number GSE74091.