CusRS is a copper-responsive and strain-specific TCS

The two divergently transcribed operons (IHMA87_02165–66 and IHMA87_02164–57) present in the urinary isolate IHMA87 are reminiscent of the copSR-copABGOFCDK locus identified in the soil-based heavy metal-tolerant β-proteobacterium Achromobacter sp. AO22: the gene organization is indeed similar and the sequence identity of the predicted proteins ranges from 84.9 to 99.7% [26] (Fig 1A). These predicted operons are absent from reference strains of P. aeruginosa (PAO1, PA14) and from the taxonomic outlier PA7, the reference strain of a P. aeruginosa clade that has recently been proposed as a new species called P. paraeruginosa [25]. We have chosen the exolysin-secreting clinical isolate IHMA87 as a model several years ago because, unlike PA7, this strain is virulent and amenable to genetic manipulation [27]. The above-mentioned locus is present in a genomic island described hereafter. The regulatory genes IHMA87_02165–66 encode a putative copper-responsive TCS that shares homology with P. aeruginosa CopRS. We named this TCS CusRS due to the higher identity of the predicted proteins with the E. coli CusR and CusS (70.7% and 41.5% amino acid identity, respectively) than with IHMA87 CopR and CopS (69% and 36.2% amino acid identity, respectively). For the IHMA87_02164–57 genes, we used the “pcoA2B2GOFCDK” nomenclature, with “A2 and B2” as copies of pcoA and pcoB genes are already present in the core genome of P. aeruginosa and P. paraeruginosa (www.pseudomonas.com; [28]). The predicted operon codes for several proteins required for metal detoxification, including the multi-copper oxidase PcoA2, a putative P-type ATPase (PcoF) and copper-binding periplasmic proteins likely to play a role as chaperones (PcoC, PcoG and PcoK).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. CusRS is a copper responsive and self-regulated TCS.

(A) Genetic organization of the copper-related gene locus identified in strain IHMA87. Blue: copper-resistance genes; Orange: regulatory genes. (B) The pcoA2-cusR intergenic sequence, framed by the start codon (colored highlights) of each divergent gene. The -10/-35 sequences of the cusR promoter (underlined) were predicted using BPROM. The putative CusR binding site, the palindromic sequence, is in bold and highlighted in gray, with the summit of the peak identified by previous DAPseq [29] identified by an arrowhead. The mutated sequence (BSmut) used in the study is indicated in the inset, with the 4 mutated bases indicated by asterisks. (C) Electrophoretic mobility shift assay of CusR on the pcoA2-cusR intergenic region. Up to 75 nM of recombinant His6-CusR protein were incubated with 0.5 nM Cy5-labeled 60-mer probe for 15 min before electrophoresis. For competition assays, excess unlabeled PcusR probe (200 nM, 400-fold) was added to the reaction. Arrowheads indicate the positions of unbound free probes (bottom) and CusR/probe complexes (top). (D, E) β-galactosidase activities of the wild-type IHMA87 strain (WT) and the ΔcusR strain harboring the indicated PcusR-lacZ (D) and PpcoA2- lacZ (E) transcriptional fusions. The different regions studied are illustrated in the left panels. Black rectangles: putative -10/-35 boxes of cusR promoter; Gray rectangle: palindromic sequence, with the BSmut sequence indicated by an asterisk. Arrow: putative transcription start sites, the arrow is dotted for pcoA2 as it was not predicted by BPROM. Experiments were performed in triplicate and error bars indicate the SEM. Different letters indicate significant differences according to two-way ANOVA followed by Tukey’s multiple comparison test (p-value < 0.05).

https://doi.org/10.1371/journal.pgen.1011325.g001

Previous DAPseq experiments showed that, in vitro, recombinant CusR mainly binds to the 254-bp intergenic sequence of the cusR and pcoA2 genes of the IHMA87 genome [29]. The summit of the enriched DAPseq peak pointed to a perfect palindromic sequence “ATTCATnnATGTAAT” found upstream of a putative cusR promoter identified by BPROM (www.softberry.com) (Fig 1B). This palindromic sequence has similarities to the CopR consensus sequence in P. aeruginosa (CTGACAn₅GTAAT) (PRODORIC, www.prodoric.de). CusR binding to the intergenic sequence was confirmed by Electrophoretic Mobility Shift Assays (EMSA), and its specificity was demonstrated by competition with an unlabeled probe (Fig 1C), suggesting that CusR might regulate its own expression as well as that of the genes transcribed in the opposite orientation.

To assess the copper-responsiveness of the two operons and the involvement of the CusRS system in this response, we analyzed the expression of transcriptional lacZ reporter fusions with 500-bp sequences upstream of the cusR and pcoA2 coding sequences in the wild-type strain and the cusR-deleted mutant. For copper stress conditions, we exposed the bacteria to 0.5 mM CuSO₄ for 2.5 h. This concentration and duration do not affect bacterial viability [19]. Exposure of the wild-type strain to copper led to significant stimulation of expression of both cusR and pcoA2 (Fig 1D,1E). In contrast, no induction of pcoA2 was observed following exposure of the cusR mutant to copper (Fig 1E), and activation of the cusR promoter, while still detected, was reduced about two-fold (Fig 1D). CusR therefore appears to control expression of both operons in a copper-dependent manner, but cusR expression must also be controlled by another copper-responsive regulatory factor.

To assess the importance of promoter elements, we constructed and analyzed a set of lacZ fusions associated with cusR upstream sequences of various lengths. Deletions affecting the palindromic sequence identified or the predicted -10/-35 boxes abolished expression, whereas a short fragment containing the palindrome maintained a copper response, but reduced both basal and activated expression levels (Fig 1D). Interestingly, mutation of a 4-bp sequence within the palindrome (ATTACATnnATAGCCT) led to complete loss of the copper-responsiveness of the promoter. When examining the pcoA2 promoter region (Fig 1E), shortening the sequence reduced the fold-change in expression following exposure to copper, whereas mutations in the palindrome led to the complete loss of copper-mediated PpcoA2 activation. As expression of the pcoA2 promoter is CusR-dependent, we concluded that CusR controls pcoA2 expression by binding to the palindromic sequence identified.

To assess the extent of the pco operon, the mRNA levels of several genes within the locus (pcoB2, pcoF, and pcoK) and that of a downstream gene encoding a putative LysR-type transcription factor were measured by RTqPCR. Transcription was increased for all targets following copper stimulation, although the level of induction measured decreased with the distance between the gene and the pcoA2 promoter (S1 Fig). In all cases, interruption of pcoA2 transcription by introducing an omega interposon abolished the copper-inducing effect down to lysR, indicating that the last gene in the locus, lysR, is part of the same transcriptional unit.

Overall, these data show that the CusRS system in IHMA87 is functional and responds to copper. Following binding to a single site, CusR regulates its own expression and controls the expression of the nine-gene “pcoA2” operon, which contains genes that may play a role in copper resistance.

CusR and the pcoA2 operon contribute to copper resistance

Copper concentrations in urine are known to increase during UTIs, as part of an innate antimicrobial strategy [3,30]. Given the origin of IHMA87, we first questioned whether it had higher copper resistance compared to three reference strains. For these reference strains, we found P. aeruginosa PAO1 and PA14 to be more sensitive to 20 mM CuSO₄ than the P. paraeruginosa strain PA7 (S2 Fig). In these tests, IHMA87 had the highest resistance capacity, which could be provided by the additional, putative copper-resistance machinery, Pco, acquired through HGT. We also studied two others–JT87 and TA19, which like IHMA87 are members of the PA7-like group. These other strains showed variable resistance profiles, with TA19 being the most sensitive of all the strains tested. Thus, high copper resistance does not appear to be a common feature of UTI isolates.

Therefore, we investigated the role of the CusRS system and the pcoA2 operon-encoded proteins in the high copper resistance of IHMA87. First, CusR was compared to the two well-described copper regulators, CopR and CueR. To do so, survival of individual, double, and triple mutants was analyzed on minimal medium in the presence of high copper concentrations (5 mM or 20 mM CuSO₄) and compared to that of the wild-type IHMA87 strain. As shown in Fig 2A, mutation of the cueR gene did not affect copper resistance whatever the genetic background, whereas the absence of CopR considerably reduced survival of the bacteria, revealing a key role of CopR and its regulon in the response to copper stress. When CopR was absent, the additional loss of CusR made the bacteria highly sensitive to copper toxicity. This sensitivity was observable at 5 mM CuS0₄ and accentuated at the highest concentration (Fig 2A), although the magnitude of the effect may vary somewhat between replicates, as shown in Fig 2B. To determine whether this phenotype was caused by the absence of the CusR regulator per se or by the lack of pcoA2-lysR expression, we introduced in several mutants an omega interposon into the pcoA2 gene, stopping transcription of the operon from the first gene. This interposon strongly affected the copper resistance of the ΔcopR pcoA2::Ω strain compared to that of the ΔcopR mutant, with an effect observable from 5 mM CuSO₄ (Fig 2B). Thus, the proteins encoded by the CusR-regulated pcoA2 operon are involved in copper detoxification and resistance. We can also see that pcoA2::Ω further increases the copper sensitivity of ΔcusRΔcopR, suggesting that the basal level of pcoA2 expression in minimal medium and on plate is probably sufficient to confer some degree of copper resistance.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. The CusR regulon plays a role in bacterial copper resistance.

(A, B) Copper sensitivity plate assays of the different strains used in the study. ΔΔΔ corresponds to the ΔcusRΔcueRΔcopR triple mutant and Ω corresponds to the pcoA2::Ω background. 10 μl of 10-fold serial dilutions, as indicated below the images (10⁻¹ to 10⁻⁶), were deposited on M9 plates containing the CuSO₄ concentration indicated. Plates were observed after 24 h (M9) or 48 h (Cu) of growth at 37°C.

https://doi.org/10.1371/journal.pgen.1011325.g002

We further investigated the contribution to copper resistance of the different copper regulatory systems and the proteins encoded by the pcoA2 operon. To do so, we followed the growth of appropriate mutant strains in lysogeny broth (LB) and compared it to IHMA87 wild-type growth. Firstly, increasing the copper concentration from 1 to 4 mM led to an extended lag phase for wild-type growth (S3A Fig). Significant cell death has been reported in the bacterial population after exposure to this high copper concentration for 2 h [19]. As the lag phase is a dynamic and adaptive period preparing bacteria for cell division, its extension indicated that more time is required both to repair altered cells and for physiological recuperation by stressed cells before exponential growth can occur [31,32]. Secondly, we observed that cueR and copR mutant growth was even more dramatically affected than wild-type growth in the presence of 4 mM CuSO₄. For these strains, especially ΔcopR, the lag phase was considerably extended (S3A Fig). This increased delay before division indicated that the copper-resistance mechanisms controlled by the two regulators are important for survival in and adaptation to the stress. Interestingly, growth of ΔcusR and pcoA2::Ω mutants was unaffected, and was similar to IHMA87 growth, indicating that these mutations alone had no effect on the copper resistance. In contrast, the lack of both CopR and Pco proteins completely abolished bacterial growth (S2A Fig). This result further underlined the important role in bacterial copper resistance played by the proteins encoded by the pcoA2 operon. In the conditions used here, this role seems to be masked by proteins from the CopR regulon. Finally, we introduced an arabinose-inducible promoter (PBAD) into the chromosome, upstream of the pcoA2 operon, and compared bacterial growth of the wild-type and modified strains over a range of copper concentrations. Increased expression of the operon conferred a clear growth advantage relative to wild-type bacteria, allowing faster adaptation to exposure to 4 mM CuSO₄, as indicated by the shorter lag phase (S3B Fig).

In conclusion, IHMA87 has a high intrinsic capacity to resist copper stress that can be further increased by overexpressing the pcoA2 operon.

Three TCSs cross-regulate the cusR-pcoA2 intergenic region

As shown in Fig 1D, copper activates cusR expression even in the absence of CusR. As CopR binds in vitro to the pcoA2-cusR intergenic region [29], we assessed its role as well as that of the copper-responsive regulator CueR in the residual activity of PcusR and that of the divergent promoter. To do so, we used transcriptional lacZ fusions in several genetic backgrounds lacking one, two, or all three regulators. We found that CueR was dispensable in all the conditions tested (Fig 3A and 3B). However, induction of cusRS expression was reduced in the absence of CopR compared to the wild-type background, and was lost in the ΔcusRΔcopR mutant. This result confirmed CopR as the additional element responsible for the complete copper induction of cusRS in combination with CusR. Activation of the cusRS promoter by both CusR and CopR was abrogated by mutating the palindrome (Fig 1D), strongly supporting that the same intergenic binding site is required in vivo for activation by both regulators.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 3. Cross-regulation by CusRS and CopRS.

β-galactosidase activities are shown for strains harboring the PcusR-lacZ (A), PpcoA2-lacZ (B), PcopA1-lacZ (C), or PpcoA-lacZ (D) transcriptional fusions. ΔΔΔ corresponds to the ΔcusRΔcueRΔcopR triple mutant. (E) β-galactosidase activities of the wild-type and pcoA2::Ω strains containing the indicated transcriptional fusions. Experiments were performed in triplicate; error bars correspond to the SEM. Different letters indicate significant differences according to two-way ANOVA followed by Tukey’s multiple comparison test (p-value < 0.05).

https://doi.org/10.1371/journal.pgen.1011325.g003

For the expression of the pcoA2 operon, as mentioned above, no copper induction was observed in the ΔcusR strain (Figs 1E and 3B). Interestingly, pcoA2 expression was higher in ΔcopR than in the parental strain upon exposure to copper, suggesting a negative regulatory role of CopR. In contrast, expression of pcoA2 genes was abolished in the ΔcusRΔcopR strain, indicating that the increased activation observed in the presence of copper in ΔcopR was the result of CusR activity (Fig 3B). This observation suggested that CopR binding does not lead to transcriptional activation of the pcoA2 promoter and, by competing for binding on the same palindrome, CopR seems to limit access for CusR, therefore reducing pcoA2 activation.

To further explore cross-regulation between the different pathways, we tested whether CusR affects the expression of copA1 and pcoA, which are respectively regulated by CueR and CopR [19]. Activation of the copA1 promoter was abolished in the absence of CueR, but was unaffected by deletion of either cusR or copR (Fig 3C) as expected based on in vitro binding data [29]. pcoA expression had almost the opposite profile, with no effect observed following cueR deletion, but a complete lack of copper-induced expression in the copR mutant and a slight reduction in the copper-responsiveness in the cusR mutant (Fig 3D). This latter result was unexpected as in vitro results indicated that CusR does not bind to this promoter [29]. Suspecting an indirect effect, we measured expression of PpcoA-lacZ and PpcoA2-lacZ fusions in the pcoA2::Ω strain. Levels of both transcriptional fusions were decreased in the presence of copper compared to levels in the wild-type strain, whereas the level of CueR-regulated copA expression was similar for the two strains (Fig 3E). Based on these observations, the proteins encoded by the pcoA2 operon may be required for the CopRS and CusRS systems to fully develop their copper-dependent activity. The effect is probably due to altered copper compartmentalization in the bacteria. These data also revealed an additional indirect layer of regulation, since the primary target of CusR, the pcoA2 operon, seems important for both TCSs to mount a copper response.

Mining and reanalyzing the DAPseq data [29] indicated that three other RRs strongly targeted the pcoA2-cusR intergenic sequence centered at the same position (S4A Fig). These RRs are the two poorly characterized regulators IrlR and MmnR, and the well-known zinc-responsive protein CzcR that binds to the sequence “GAAACn₆GTAAT” [33]. Additional target genes were found to be shared by CopR, CusR and CzcR (S4B and S4C Fig), presumably due to the similarity of their consensus sequences, especially the second half-site “GTAAT”. Therefore, we analyzed the roles of these additional RRs on the expression of both the cusRS and pcoA2 operons.

The TCS CzcRS is known to respond to several signals, including zinc stress [34,35]. We therefore assessed its effect on pcoA2 and cusRS expression in the presence of copper and/or zinc. As a control of CzcR activity, we introduced a PczcC-lacZ fusion into the wild-type and ΔczcR strains, as a known CzcR target. Fig 4A shows that, as expected, the czcC promoter was upregulated after the addition of zinc, with or without the addition of copper, and that this upregulation was CzcR-dependent. Furthermore, the expression of pcoA2 was activated by zinc to almost the same level as by copper. Exposure to both stimuli was slightly, but significantly, more efficient. Zinc activation was lost in the absence of CzcR, and the pcoA2 expression level dropped to that observed in the wild-type in the presence of copper alone (Fig 4A). Therefore, both CusR and CzcR control and activate expression of the pcoA2 operon to similar extents in response to copper and zinc, respectively. A distinct expression profile was observed for the cusR promoter: while copper strongly activated its expression, the induction following exposure to zinc was lower. In addition, zinc bypassed the copper effect when bacteria were simultaneously exposed to both metals. In the absence of CzcR, this “restraining” effect was lost, and the high-copper-inducible effect was restored, suggesting that CzcR must conceal the CusR binding site. To support the idea that CzcR binds to the same site as CusR, the expression of the two transcriptional fusions carrying the mutated binding site (Fig 1B, 1D and 1E) was analyzed in the presence of copper and/or zinc in the wild-type strain. Not only was the expression of the two operons no longer activated by copper, as observed in Fig 4B, but the activation by zinc was also abolished. This strongly supports that CzcR directly controls the expression of the cusRS and pcoA2 operons by binding to the palindromic site recognized by both CusR and CopR.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 4. CzcR regulates the pcoA2-cusRS intergenic region.

(A,B) β-galactosidase activities of the IHMA87 strain (WT) (A, B) and its isogenic czcR mutant (A) harboring the transcriptional fusions indicated. Activities were measured after 2.5 h of growth in LB or LB supplemented with 0.5 mM ZnCl₂, 0.5 mM CuSO₄, or 0.5 mM ZnCl₂/0.5 mM CuSO₄. Error bars indicate the SEM. Different letters indicate significant differences according to two-way ANOVA followed by Tukey’s multiple comparison test (p-value < 0.05). (C) Model proposed for the action of CusR, CopR and CzcR, that compete for binding to the same site in the pcoA2-cusR intergenic region. The activating molecule (copper or zinc) is indicated for each system. The thickness of the long arrows and the number of « + » signs indicate the strength of transcriptional induction, if any.

https://doi.org/10.1371/journal.pgen.1011325.g004

We then assessed a potential regulatory role of MmnR and IrlR and observed no mmnR or irlR deletion phenotypes in terms of pcoA2 operon and cusR expression, even in the presence of copper (S5 Fig). However, we cannot exclude that this was due to an absence of signal and activation of the sensor in our experimental conditions.

Overall, the data indicate that cusRS and pcoA2 operons are controlled in vivo by at least three TCSs: CusRS, CopRS, and CzcRS. The three RRs bind to the same palindromic site but modulate expression of the divergent operons differently. Indeed, while CusR and CzcR activate both operons, even if the latter is less effective on the pcoA2 promoter, CopR only activates cusRS, repressing the pcoA2 operon by hindering the positive action of CusR (Fig 4C).

Molecular cross-talk between the CopRS and CusRS systems

It has been reported that the P. aeruginosa CopRS system functions in an unusual way [36]. Rather than relying on a conventional “phosphorylation-based mechanism” in which the sensor HK autophosphorylates upon detection of a signal, allowing subsequent phosphorylation of the cognate RR, the copper signal switches off the phosphatase activity of CopS towards its RR CopR. Thus, exposure to copper allows CopR to be phosphorylated and to activate its target genes. The mechanism leading to the phosphorylation of CopR remains unclear, as the phosphodonor has not yet been identified [36]. Our data, shown in Fig 5A, were consistent with the “phosphatase-based mechanism” of the CopRS system in the P. paraeruginosa strain IHMA87. Indeed, in a cusRS background, the absence of CopS led to a high-copper-blind expression of pcoA whereas expression of the sensor CopS_H235A mutant protein with constitutive phosphatase activity [36] abolished pcoA expression. The copper-blind expression in ΔcopSΔcusRS was approximately 5-fold higher than that observed in the copper-induced wild-type, suggesting that phosphatase activity of CopS is not completely switched off by the presence of copper in the wild-type background and that CopR is similarly efficiently phosphorylated in ΔcopSΔcusRS regardless of the presence of metal. The same pattern was observed for cusR expression (Fig 5A). This result also confirmed that CopR can activate expression of this promoter in a ΔcusRS background, as previously suggested by the data presented in Fig 3A.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 5. In vivo cross-talk between CusRS and CopRS.

β-galactosidase activities of the strains harboring the transcriptional fusions indicated, after 2.5 h in LB or LB supplemented with 0.5 mM CuSO₄. Error bars indicate the SEM. Different letters indicate significant differences according to two-way ANOVA followed by Tukey’s multiple comparison test (p-value < 0.05).

https://doi.org/10.1371/journal.pgen.1011325.g005

We then analyzed the functioning of the CusRS system in a genetic background lacking the CopR and CopS proteins to avoid any possible protein interference. As shown in Fig 5B, neither pcoA2 nor cusRS operons were activated in the strain lacking the sensor protein CusS. This result indicates both that this HK is required for CusR activation by phosphorylation, and that no other protein can phosphorylate CusR. We also confirmed that the phosphorylated form of CusR is the active form, because no gene expression was observed when its conserved phosphorylation site (Asp₅₁) was mutated to an alanine (Fig 5B). In addition, mutation of the autophosphorylation site of CusS (His₂₆₆) abolished the expression of the two operons, indicating that it is required for the system to function properly (autophosphorylation prior to phosphotransfer). Altogether, these data show that the CusRS system is a canonical TCS, functioning like the E. coli CusRS system, with the detection of a signal triggering a phosphorylation cascade ultimately leading to target gene activation.

The two TCSs were then studied in different combinations of mutants of their HK gene to obtain information on possible cross-talk between non-cognate HKs and RRs. Focusing on a CopR-only-activated promoter, we found that deletion of copS resulted unexpectedly in an identical copper response pattern of PpcoA-lacZ expression to that observed in the wild-type strain or in a strain lacking the cusS gene (Fig 5C). However, deletion of both copS and cusS led to the strong constitutive activation of PpcoA (copper-blind expression) that was observed in the ΔcopS mutant in the absence of CusRS (Fig 5A). These data suggest that the HK CusS certainly prevents pcoA expression by dephosphorylating CopR in the absence of both copper and CopS, suggesting that CusS can regulate CopR activity and compensate for the absence of the HK CopS. Alternatively, the interaction between CusS and CopR in the absence of CopS could prevent phosphorylation of CopR in the absence of signal. We further observed that the control of the phosphorylation state of CopR by CusS does not require the absence of its cognate HK. Indeed, normal expression of PpcoA was observed when CopS_H235A was produced (Fig 5C), suggesting that CusS allows CopR phosphorylation upon copper sensing, bypassing the constitutive phosphatase activity of CopS_H235A.

To determine whether CopS could also interact with CusRS, we used PcusR-lacZ (Fig 5D) and PpcoA2-lacZ (Fig 5E) transcriptional fusions in several mutant genetic backgrounds. As expected, absence of CopS had little effect on the two promoters: indeed, CusS could control the phosphorylation status of CusR and of CopR, as observed above, and both these proteins activated transcription only in response to the copper signal. However, when cusS was deleted, in the absence or presence of copper, a slight increase in PcusR expression levels was observed (Fig 5D), and the effect on PpcoA2 expression was even more pronounced (Fig 5E). Considering prior observations on Fig 5B, such effect must result from the presence of CopRS. To determine whether this transcriptional activation was due to the activity of phosphorylated CusR, we mutated its phosphorylatable residue (Asp51), as has been done in the CopR protein [36]. Indeed, the expression of cusR and pcoA2 was abolished when CusR_D51A was synthesized in the ΔcusS background mutant. Promoter of cusR is still slightly subject to copper activation by CopR as observed in Fig 5A. Thus, CopS appears to permit CusR phosphorylation in the presence of copper, but also in its absence, and to control its phosphorylation status, although not as tightly as the cognate HK CusS. PcusR and PpcoA2 activation was not observed in the copS_H235AΔcusS strain, suggesting that CopS_H235A maintains CusR dephosphorylated or that it has lost its ability to phosphorylate the RR (Fig 5D). Copper-blind expression of both PcusR and PpcoA2 was observed in the ΔcusSΔcopS mutant. As CusR cannot be phosphorylated in bacteria lacking both sensors (CusS and CopS), we reasoned that the expression pattern observed on PcusR (Fig 5D) can be attributed mainly to dysregulated CopR activity (Fig 5C). The reason for the activation of PpcoA2 is more speculative, but our results indicate that, in such a background, cusR is highly expressed and, consequently, the RR protein is highly produced, which could lead to its binding to its targets despite its unphosphorylated state. Consistent with this is the copper-blind expression of the promoter measured in the strain cusR_D51AΔcopSΔcusS (Fig 5E). Furthermore, PpcoA2 expression levels were lower in ΔcusSΔcopS compared to in ΔcusS, which underscores the negative effect of phosphorylated CopR binding to the promoter (Fig 5E). In the copSH235AΔcusS background, the pcoA2 promoter was not expressed, as no CusR is synthesized (Fig 5D).

The phosphatase activity of HKs towards their cognate RRs has been shown to require a conserved Thr/Asn residue close to the autophosphorylation site [37,38]. Therefore, to better understand the functioning of the two TCSs and the extent of their cross-talk, we mutated Thr₂₃₉ of CopS and Thr₂₇₀ of CusS to potentially abolish their phosphatase activity. As shows Fig 5F, the production of CopS_T239A in the absence of CusRS, but also of CusS alone, resulted in a high and copper-blind expression of the pcoA-lacZ fusion. The same observation was made for the cusR expression in the background copS_T239AΔcusRS (Fig 5G). These data support the model proposed by Novoa-Aponte et al. [36], in which CopS dephosphorylates CopR in the absence of copper. On the other hand, when CopS_T239A was produced in the presence of the second TCS CusRS, pcoA expression was almost equivalent to that observed in the wild-type strain, consistent with the proposed cross-talk model in which CusS can take over and compensate for the absence of CopS (Fig 5C). The Thr₂₇₀ mutation in CusS also appears to inactivate the phosphatase activity of this sensor protein towards CopR, as suggested by the mutant protein that can no longer dephosphorylate CopR in the absence of CopS (cusS_T270AΔcopS), resulting in a copper-blind phenotype. The final evidence that CusS is able to control the phosphorylation state of CopR in vivo in a copper-dependent manner was provided by the analysis of pcoA expression, which showed wild-type regulation in ΔcopSΔcusR (Fig 5F). Differences in the expression levels of the pcoA-lacZ fusion were observed in the different mutants (Fig 5A, 5C and 5F), indicating that the extent of RR phosphorylation/dephosphorylation probably depends on the number of partners present and the protein-protein interactions involved.

We then analyzed the effect of the phosphatase-deficient CopS and CusS proteins on the expression of the cusRS (Fig 5G) and pcoA2 (Fig 5H) operons which are cross-regulated by CusR and CopR. We first observed that CusS_T270A retains its ability to auto-phosphorylate and transfer its phosphate to CusR, as observed with the high copper induction of PcusR-lacZ (Fig 5G) and PpcoA2-lacZ (Fig 5H) fusions. For PpcoA2-lacZ, Fig 5G also shows that the presence of either phosphatase-deficient protein (CusS_T270A or CopS_T239A) did not affect the copper-responsiveness of cusR expression, and the additional absence of the second HK resulted in a copper-blind phenotype, as observed in ΔcopSΔcusS (5D): this is consistent with CusS being able to dephosphorylate CopR in the absence of copper. Analysis of PpcoA2-lacZ expression also showed copper-blind expression in the presence of wild-type and phosphatase-deficient HKs (cusS_T270AΔcopS and copS_T239AΔcusS), as observed in ΔcopSΔcusS (Fig 5E), albeit at a higher level of expression (Fig 5H), consistent with the transcriptional activity of overproduced, unphosphorylated CusR, as proposed above. Finally, the interaction between CopS and CusR was clearly observed in vivo in the absence of CusS and CopR, with a high expression of the PcusR-lacZ and PpcoA2-lacZ fusions in ΔcusSΔcopR, an expression that is halved in the presence of copper. Indeed, CusR can obtain its phosphate from the remaining HK, CopS, to be active (Fig 5G and5H). Therefore, in addition to its phosphatase activity, CopS must have an autokinase activity, which appears to be modulated by copper availability. Interestingly, copper seems to partially decrease rather than increase this activity. Although non-intuitive, unorthodox regulation of the kinase activity may be related to the fact that it does not seem to be relevant for the function of the TCS CopRS and the expression of the CopR regulon.

Finally, we wondered whether the CopRS and CusRS systems respond to and modulate their regulon within the same range of copper concentrations. We compared their individual sensitivities to copper and their sensitivity relative to that of the cytoplasmic detector CueR. To do so, we used lacZ fusions to monitor the expression of a specific target of each transcriptional regulator over time in response to different copper levels. We found CopRS and CusRS to display a similar sensitivity and response to copper, inducing their target in the presence of 0.5 mM CuSO₄. In contrast, CueR activates its target at lower copper concentrations (from 0.01 mM) (S6 Fig).

In conclusion, although they respond to the same signal, the CopS and CusS HKs do not use the same molecular mechanism to control their cognate RR, the former using a phosphatase activity in the absence of the signal, whereas the latter is a phosphodonor in its presence. However, in the absence of one of the HKs, the other can compensate and effectively control the phosphorylation state and therefore the activity of both RRs, with a greater efficiency observed for CusS: the crosstalk thus allows an almost normal expression of the copper resistance proteins.

Horizontal acquisition of the cusRS and pcoA2 operons

As mentioned above, the reference P. paraeruginosa strain PA7 lacks the cusRS and pcoA2 operons. Analysis of the genomic environment surrounding this copper-related locus showed that it is present on a genomic island that presents the features of an ICE (Fig 6A). ICEs are large mobile genetic elements that are crucial for the genetic plasticity of bacteria [39,40]. Consequently, the copper-related locus was probably acquired by HGT. We named this island ICE1_IHMA, and it is one of three predicted ICEs in the IHMA87 genome. It is 96,269 bp long, inserted at a tRNA-Gly site, and has an average GC content of 64%. Among the 100 predicted coding sequences present on ICE1_IHMA, some correspond to core ICE genes, required for the element’s life cycle. These include signature genes coding for an integrase of phage origin (int) and a MOB(H)-type relaxase [41]. In addition, the element possesses a set of mating-pore formation (MPF) genes encoding a T4SS (tfc genes) including an ATPase (tfc16/virB4) and a Type IV coupling protein (tfc6/t4cp2). Among the eight types of MPF system–classed based on diverse features–ICE1_IHMA is a type G (MPF_G). These systems are frequently found in Proteobacteria [42]. ICEs also carry specific cargo genes, which usually provide the host with benefits and diverse phenotypic properties [39]. Based on sequence homology and the organization of its core genes, ICE1_IHMA is highly similar to PAGI-2, a prevalent clc-like family ICE in P. aeruginosa [16] (Fig 6A). By comparison with PAGI-2, we defined two specific cargo modules in ICE1_IHMA, named A and B. Module A comprises four genes, of which three (IHMA87_02154–52) are predicted to encode a CzcCBA-type heavy metal RND efflux pump. To assess whether this pump could be involved in copper efflux, we deleted IHMA87_02154–52. No effect was observed in terms of survival of the IHMA87ΔRND mutant obtained upon exposure to a high copper concentration (S2 Fig). In line with this result, the promoter region of IHMA87_02154–52 was not targeted in vitro by CusR or CopR [29]. The other module, Module B, is composed of 21 cargo genes, among them the cusRS and pcoA2 operons, but also several genes predicted to be involved in resistance to chromate and arsenic compounds. Copper, arsenic, and chromium resistance genes frequently co-occur in mobile genetic elements, favoring the spread of both metal resistance and antibiotic resistance genes. The spread of such resistance is of concern because co-selection promotes the acquisition of antibiotic resistance [10,43]. Although the gene encoding the LysR-type regulator is present on PAGI-2, it was considered to be part of Module B, since we demonstrated that it was part of the pcoA2 operon.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 6. Conservation of lysR-cusS locus in P. aeruginosa and P. paraeruginosa species.

(A) Comparison of the sequences of ICE1_IHMA and PAGI-2. ORFs are color-coded as follows: pcoA2B2GOFCDK, blue; cusRS, orange; genes involved in conjugation, transfer and integration of ICE, black; homologous genes, white; genes without any corresponding counterparts in ICE1_IHMA and PAGI-2, gray. Intensity of the link between ORFs was based on % identity between the orthologous proteins. (B) Distribution of genera across genomes harboring modules carried on ICE1_IHMA, as defined in (A). Genera with a single representative are not shown. Module A: Delftia, Diaphorobacter, Ralstonia and Variovorax; Module B: Aerosticca, Delftia, Tistrella and Variovorax; lysR-cusS locus: Aerosticca, Bordetella, Delftia, Raoultella, Ralstonia, Tistrella, Variovorax and Xanthomonas. (C) Phylogeny of P. aeruginosa and P. paraeruginosa species and distribution of lysR-cusS. Inset shows unrooted tree at scale, with the number of genomes for each clade. The presence of the locus is indicated by blue square, a blue dot signals whether the genes are encoded in a predicted ICE.

https://doi.org/10.1371/journal.pgen.1011325.g006

As ICEs are modular and cargo genes can be exchanged, we assessed how the two cargo modules are distributed in bacterial genomes by performing blast searches against the NCBI database (Fig 6B). Modules A and B from ICE1_IHMA are both present in several bacteria from various genera. The most frequently represented genus was Pseudomonas, and the corresponding predominant species are aeruginosa and paraeruginosa. The two modules were also identified in other genera with distinct distributions, for example in a few Achromobacter and Stenotrophomonas genomes, whereas only Module A was identified in some Comamonas genomes. Interestingly, if the lysR gene located at the 3’ extremity of the pcoA2 operon is excluded from the search for Module B, identical results were obtained. This further emphasizes its co-occurrence with the cusRS and pcoA2 operons, and suggests a possible role in their regulation.

We next focused our analysis on the distribution of the region from lysR to cusS, which comprises the entire sequence of the two operons (Fig 6B). Initially, this search provided more hits than searches using the whole of Module B, suggesting that this region can be acquired or maintained independently of the rest of the module. The genera harboring the locus were similar to those with the module B, but the distribution varied with a higher ratio of Pseudomonas. Among species, the region was mostly represented in P. aeruginosa and P. paraeruginosa, with only two strains of Pseudomonas not belonging to these species. We then examined the distribution and conservation of the lysR-cusS region over a database of 494 complete P. aeruginosa and P. paraeruginosa genomes. The clades were named based on the approach described by Freschi et al. [44] from 1 to 5, and, as proposed by Rudra et al. [25], outlier group 3 was (re)classified as P. paraeruginosa (P.p). Bioinformatics analysis revealed the locus to be present in several strains of P. aeruginosa clades 1 and 2, and (as expected) in the IHMA87 strain, which is classed in clade 3 (Fig 6C). The locus was not detected in the few clade 4 and 5 strains for which genomes are available (www.pseudomonas.com; [28]).The locus was present as part of a putative ICE in over 80% of genomes. In the remaining 20% of genomes, it was part of regions bearing some similarity to an ICE, as determined based on neighboring genes and their predicted functions. The reason these regions were not immediately predicted to be ICEs can be attributed either to ICE degeneracy, or to a detection failure. In all the loci identified, the motif recognized by CusR in the pcoA2-cusRS intergenic sequence is conserved, indicating that the copper-resistance genes can be integrated into the bacterial regulatory network, allowing them to be readily expressed in their host under appropriate conditions.

Taken together, these data show that the lysR-cusS locus involved in copper resistance is mainly found in the Pseudomonas genus where it can be acquired by HGT as part of an ICE. As demonstrated by our research, the integration of this locus into the regulatory network involves cross-regulation and cross-talk with endogenous regulatory actors. These constraints must limit the acquisition of the locus by other genera.

Bacterial strains and growth conditions

The E. coli, P. aeruginosa and P. paraeruginosa strains used in this study are described in S1 Table. Bacteria were grown aerobically in LB at 37°C with agitation. To assess the effect of copper on gene expression, overnight cultures were diluted to an OD₆₀₀ of 0.1 in LB supplemented with the indicated CuSO₄ concentrations, and growth was continued until the OD₆₀₀ reached around 1.0 (2.5 h). Growth in 96-well plates at 37°C with agitation was assessed by following the OD₅₉₅ of 200-μL cultures in LB with the CuSO₄ concentrations indicated. Plates were read by a microplate reader at the indicated times. For copper sensitivity plate assays, 15 g/L agar was added to M9 minimal medium (48 mM Na₂HPO₄, 22 mM KH₂PO₄, 9 mM NaCl, 19 mM NH₄Cl, 2 mM MgSO₄, 0.1 mM CaCl₂, 0.2% Glucose) supplemented with CuSO₄. When LB cultures reached OD₆₀₀ of 1.0, 10 μL of a 10-fold serial dilution of the cultures were spotted on the M9 media and the plates were incubated for 1 to 2 days at 37°C. Antibiotics were added at the following concentrations (in μg/mL): 200 spectinomycin (Sm), 100 ampicillin (Ap), 50 gentamicin (Gm), 10 tetracycline (Tc), and 10 chloramphenicol (Cm) for E. coli; 75 Gm and 75 Tc in presence of 5 irgasan (Irg) in LB for P. aeruginosa.

Plasmids and genetic manipulation

The plasmids used and the primers for PCR are listed in S1 and S2 Tables, respectively. All constructions were verified by sequencing.

To generate P. paraeruginosa deletion mutants, upstream (sF1/sR1) and downstream (sF2/sR2) flanking regions of the different genes were fused and cloned into an SmaI-cut pEXG2 plasmid by sequence- and ligation-independent cloning (SLIC) using the appropriate primer pairs [64].

For chromosomal insertion of the PBAD upstream pcoA2 operon, a pEXG2 plasmid was created by SLIC with upstream and downstream sequence of the pcoA2 promoter sequence, creating a SpeI site 52 bp upstream of the start codon. Then, the 1580-bp “araC-PBAD” fragment was excised from pSW196 using XbaI and cloned into the SpeI site created, to produce pEXG2-PBAD-pcoA2-Sp.

For mutagenesis by insertion of the Ω interposon, the MCS-contained BamHI site was first deleted from pEXG2 (pEXG2ΔB). Then a de novo BamHI site was created while using SLIC to fuse two fragments of pcoA2 in SmaI-cut pEXG2ΔB. Finally, the Ω interposon from pHP45Ω was introduced into BamHI-cut pEXG2ΔB-IHMA_02164.

To replace the His or Thr residues in CusS and CopS with Ala, we used SLIC to generate a fragment with the mutated codon and to create a SacII site (for His to Ala mutation) or BssHII site (for Thr to Ala mutation) to allow subsequent restriction analysis screening of PCR-amplified fragments.

To create the transcriptional lacZ fusions, the miniCTX-lacZ plasmid was first modified to prevent transcriptional interference by integrating the strong rrnB terminator sequence into its XhoI-KpnI sites, as described in Trouillon et al. [65], to produce miniCTX-TrrnB-lacZ. Then, fragments comprising the ATG and around 500 bp upstream of each target analyzed were amplified using the appropriate pairs (sF/sR) for insertion into SmaI-cut miniCTX-TrrnB-lacZ by SLIC. To create mutations in the regulator binding sites, another pair of overlapping mutated primers (sR1/sF2) were used to create two overlapping fragments (sF/sR1 et sF2/sR), also inserted into SmaI-cut miniCTX-TrrnB-lacZ by SLIC.

The pEXG2- and miniCTX-TrrnB-lacZ-derived vectors were transferred into P. paraeruginosa IHMA87 strains by triparental mating using the helper plasmid pRK600. Allelic exchanges for mutagenesis were selected as previously described [66]. To create the mutant encoding the CusR_D51A protein, the mutated sequence was introduced into the ΔcusR mutants to replace the deleted gene.

For E. coli overproduction of CusR protein with an N-terminal 6xHis-Tag, the cusR gene was first amplified by PCR using IHMA87 genomic DNA as template, and SLIC was used to insert the fragment into the pET15b plasmid cut with NdeI and BamHI.

β-galactosidase activity assay

β-galactosidase activity was assessed at least in triplicate as previously described [67,68], after 2.5 h of growth (OD₆₀₀ of ~1.0) in the indicated media. Whole cells (0.5 ml) of each bacterial culture were permeabilized by addition of 20 μl of 0.1% SDS and 20 μl of chloroform, followed by vortexing for 1 min. Next, 0.1 or 0.2 ml of cells were removed and made up to 1ml with Z buffer (0.1 M Na₂HPO₄/NaH₂PO₄, 10 mM KCl, 1 mM MgSO₄, 50 mM 2-mercaptoethanol, pH 7.0), then placed at 28°C. Reaction was initiated by addition of 0.2 ml of o-nitrophenyl-β-D-galactopyranoside at 4 mg/ml and stopped with 0.5 ml of 1 M Na₂CO₃. The OD₄₂₀ was read after sedimentation of cell debris by centrifugation (5 min at 18,000g) and the activities were expressed in Miller units [(OD₄₂₀x1000)(t_minxVol_mlxOD₆₀₀)]. On the bar graphs, activities were expressed in Miller Units (MU) and error bars correspond to the standard error of the mean (SEM). Statistical significance was assessed using ANOVA test with Tukey’s method. S3 and S4 Tables provide all the numerical data and statistics underlying the graphs, respectively.

Reverse Transcription-quantitative Polymerase Chain Reaction (RTqPCR)

Extraction of total RNA, cDNA synthesis and qPCR were carried out as previously described [65] with the following modifications: the cDNA was synthesized using 3 μg of RNA and the SuperScript IV first-strand synthesis system (Invitrogen). For each strain and condition, mRNA expression was measured for three biological replicates. Data were analyzed with the CFX Manager software (Bio-Rad), and expression levels were normalized relative to rpoD reference Cq values using the Pfaffl method. The sequences of the primers used can be found in S2 Table. Statistical analyses were performed by T-test.

Production and purification of CusR

His₆CusR (H₆CusR) was produced in E. coli BL21 Star (DE3) strains harboring pET15b-CusR, grown in LB at 37°C. Expression was induced by adding 1 mM isopropyl β-D-1-thiogalactopyranoside when cultures had reached an OD₆₀₀ of 0.6. Cells were grown for a further 2 h at 37°C and 150 rpm, and then harvested by centrifugation. They were resuspended in 20 mL of IMAC buffer (25 mM Tris-HCl, 500 mM NaCl, pH 8) containing 10 mM Imidazole and supplemented with Protease Inhibitor Mixture (Complete, Roche Applied Science) for lysis at 4°C by sonication. After centrifugation at 200,000 × g for 30 min at 4°C, the soluble fraction was loaded onto a 1-mL anion exchange column (HiTrap 1 mL HP, GE Healthcare) and purified on an Akta purifier system. The protein was eluted by applying a 30-mL gradient ranging from 10 to 200 mM Imidazole in IMAC buffer at a flow rate of 1 mL/min. Eluted protein was diluted 1:20 in Tris-NaCl buffer (50 mM Tris- HCl, 150 mM NaCl, pH 8) for conservation before use.

Electrophoretic Mobility Shift Assays (EMSA)

Probes were generated by annealing complementary primer pairs, one of which was Cy5-labeled, to form the fluorescent probe. 0.5 nM of the resulting 60-bp DNA fragments were incubated for 15 min at 25°C in EMSA Buffer (20 mM HEPES, 80 mM KCl, 20 mM MgCl₂, 0.8 mM EDTA, glycerol 7.5%, 0.1 mM dithiothreitol, pH 7.9). For competition assays, unlabeled DNA probes were added to the reaction at either 0.5 nM (equal amount) or 200 nM (400-fold excess). The proteins were then added at the indicated concentrations in a final reaction volume of 20 μL and incubated for a further 20 min at 25°C. Samples were then loaded onto a native 5% Tris-Borate-EDTA (TBE) polyacrylamide gel and run at 100 V and 4°C in cold 0.5X TBE Buffer. Fluorescence imaging was performed using an ImageQuant 800 imager.

Bioinformatics analyses

ICEs were identified and delineated using the web-based tool ICEfinder [69]. ICEs were typed using ConjScan [70] on the Pasteur Galaxy platform [71]. Sequences were aligned and visualized with clinker [72]. Blast analyses were performed with BLASTn [73]. Complete genome sequences of P. aeruginosa (n = 487) and P. paraeruginosa (n = 7) were downloaded from the Pseudomonas Genome Database [28]. Genomes were annotated with Prokka [74], and core genomes were determined with Roary [75] before alignment with MAFFT using the default parameters [76] on the Galaxy server [77]. The resulting alignment was used to build a maximum-likelihood phylogenetic tree with FASTTREE [78] on the Galaxy server. The phylogenetic tree was finally visualized and annotated using iTOL [79].