RNA-based thermoregulation of a Campylobacter jejuni zinc resistance determinant

RNA thermometers (RNATs) trigger bacterial virulence factor expression in response to the temperature shift on entering a warm-blooded host. At lower temperatures these secondary structures sequester ribosome-binding sites (RBSs) to prevent translation initiation, whereas at elevated temperatures they “melt” allowing translation. Campylobacter jejuni is the leading bacterial cause of human gastroenteritis worldwide yet little is known about how it interacts with the host including host induced gene regulation. Here we demonstrate that an RNAT regulates a C. jejuni gene, Cj1163c or czcD, encoding a member of the Cation Diffusion Facilitator family. The czcD upstream untranslated region contains a predicted stem loop within the mRNA that sequesters the RBS to inhibit translation at temperatures below 37°C. Mutations that disrupt or enhance predicted secondary structure have significant and predictable effects on temperature regulation. We also show that in an RNAT independent manner, CzcD expression is induced by Zn(II). Mutants lacking czcD are hypersensitive to Zn(II) and also over-accumulate Zn(II) relative to wild-type, all consistent with CzcD functioning as a Zn(II) exporter. Importantly, we demonstrate that C. jejuni Zn(II)-tolerance at 32°C, a temperature at which the RNAT limits CzcD production, is increased by RNAT disruption. Finally we show that czcD inactivation attenuates larval killing in a Galleria infection model and that at 32°C disrupting RNAT secondary structure to allow CzcD production can enhance killing. We hypothesise that CzcD regulation by metals and temperature provides a mechanism for C. jejuni to overcome innate immune system-mediated Zn(II) toxicity in warm-blooded animal hosts.


Introduction
The Gram-negative bacterium Campylobacter jejuni is a globally significant cause of gastroenteritis with an estimated annual incidence of 166 million cases [1]. In developing countries disease is often endemic with symptomatic infections in infants but not in older age groups, presumably due to development of acquired immunity through repeated exposure [2]. In developed countries C. jejuni infections are typically symptomatic, occur sporadically across all age groups and are commonly transmitted via food derived from birds and animals where it is considered to be a commensal. The consumption of poultry, in particular chicken, is a major source of infection [3]. Disease symptoms typically comprise a bout of self-limiting inflammatory diarrhoea and stomach cramps although in the young, elderly and immunosuppressed the symptoms can be more severe and even life threatening. There are also rare but serious infection sequelae including the autoimmune mediated peripheral neuropathy Guillain-Barré syndrome [4]. At the cellular level the ability to adhere to and invade gut epithelial cells is an established important first step in human infections and is associated with inflammation and phagocyte infiltration [5][6][7]. The public health significance of campylobacteriosis underlines the urgent need for new strategies to reduce infections. However, the mechanisms employed by C. jejuni to adapt to its intraand extracellular lifestyles and to colonise and infect its hosts are relatively poorly understood.
Zn(II) is an essential co-factor for many enzymes, however, elevated levels of Zn(II) can also be toxic and hence precise intracellular Zn(II) homeostasis is required for bacterial survival [8]. To inhibit bacterial proliferation, host innate immune defences exploit this need and both restrict Zn(II)-availability, in a process termed "nutritional immunity", and poison bacteria with antimicrobial concentrations of Zn(II) [9][10][11]. For example, the proinflammatory cytokine IL-6 causes increased abundance of the Zn(II) transporter Zip14 at the plasma membrane of hepatocytes, resulting in increased hepatic Zn(II) accumulation which likely contributes to the reduced serum Zn(II) levels that occur during the acute phase response to infection [12]. The inflammatory S100 protein calprotectin also contributes to bacterial killing at certain sites of infection by sequestering metals [including Zn(II) and Mn(II)] following its copious release from neutrophils [13]. Nonetheless many bacterial pathogens can overcome this host mediated metal-limitation via Zn(II) uptake systems (for example [14][15][16]), and notably C. jejuni possesses an ATP-binding cassette (ABC) Zn(II)-importer needed for intestinal survival [17]. Direct Zn(II) toxicity has also been documented as a host antimicrobial strategy against several pathogens. For example Mycobacterium tuberculosis and Salmonella enterica serovar Typhimurium are subjected to Zn(II) intoxication in macrophage phagosomes, where Zn(II) toxicity can act synergistically with the antimicrobial effects of copper which is also targeted to this compartment [18,19]. Zn(II) levels are also reported to increase in the various host niches colonised by Streptococcus pneumoniae to inhibit bacterial growth [20] and dietary Zn(II) restriction impacts on the ability of phagocytes to control S. pneumonia infection by Zn(II) intoxication [21]. Furthermore, high levels of both calprotectin and Zn(II) are found at skin sites colonised by group A Streptococcus, with bacterial control being associated with both extracellular Zn(II) chelation by calprotectin and intracellular Zn(II) intoxication following phagocytosis [22]. Consistent with host mediated Zn(II)-toxicity contributing to immune defences, the disruption of genes encoding Zn(II)-exporting P 1B -type ATPases (often designated ZntA) or Cation Diffusion Facilitators (CDFs)/CzcD proteins in these pathogens reduces virulence [11]. To date, nothing is known about the mechanisms that might protect C. jejuni from Zn(II)-toxicity during infection. Here we report that Cj1163c (herein designated CzcD) recently shown to enhance colonisation in a mouse intestinal infection model [23] is a CDF metal-effluxer with a primary role in Zn(II)-resistance. Moreover, we reveal that C. jejuni CzcD is under the control of an RNA thermal switch that restricts its production to the elevated temperatures of its warm-blooded hosts.
RNA thermometers (RNATs) are temperature-sensing elements located within 5' untranslated regions of mRNAs [24]. They typically respond to temperature through sequestering ribosome binding sites within the secondary structure at low temperatures and as the temperature increases the structure "melts" allowing ribosome access and consequent translation. RNATs are increasingly associated with bacterial virulence and are thought to enable rapid production of key virulence proteins on sensing elevated host body temperatures (reviewed in [25]). Examples include virulence associated transcriptional regulators from Listeria monocytogenes [26], Vibrio cholerae [27], Pseudomonas aeruginosa [28] and Yersinia pseudotuberculosis [29]; genes involved in capsule biosynthesis and production of Factor H binding protein in Neisseria meningitidis [30,31]; a complement protein-binding lipoprotein from Leptospira interrogans [32] and a cytotoxin from Y. pseudotuberculosis [33]. The only examples, to date, of bacterial RNAT-regulated genes associated with metal handling are Shigella dysenteriae shuA and shuT involved in heme uptake [34,35].
In this study, as well as functionally characterising CzcD, we have identified and characterised an RNAT in the intergenic untranslated region immediately upstream of czcD that is responsible for increased CzcD expression at elevated host body temperatures (42˚C in chickens and 37˚C in humans) to confer temperature-dependent C. jejuni Zn(II)-resistance. To our knowledge, this is the first RNAT to be described in C. jejuni and the first to be associated with Zn(II) resistance. Finally, we demonstrate that czcD inactivation attenuates killing in a simple Galleria mellonella larvae infection model and that at lower temperatures larval killing is enhanced when the czcD RNAT is mutated to allow increased CzcD production. Our findings are consistent with Zn(II)-export being important for C. jejuni host colonisation and infection by protecting against host mediated Zn(II)-toxicity.

Results
The 5' untranslated region upstream of czcD confers temperature sensitive regulation in a heterologous host To identify C. jejuni genes involved in metal homeostasis we used blastp to search the NCTC 11168 predicted proteome with well characterised proteins from a number of bacterial species.
Following sequence alignment and more detailed scrutiny of potential metal-binding ligands, an uncharacterised potential metal exporter Cj1163c was identified located within a four-gene operon consisting of Cj1164c through to Cj1161c (Fig 1A) with an associated upstream σ 70dependent promoter [36,37]. Cj1164c encodes a 10 kDa Zn(II)-finger-containing protein of unknown function, Cj1162c encodes a CopZ/Atx1 Cu(I)-chaperone like protein and Cj1161c encodes a P 1B -type ATPase CopA that confers copper resistance [38]. Cj1163c encodes a CDF Diagrammatic representation of the C. jejuni NCTC 11168 operon containing the Cj1163c/czcD gene with the shaded arrowhead indicating the associated σ 70 promoter identified by RNAseq analyses [36,37]. Located between Cj1164c and czcD is a 75 bp untranslated intergenic region (SLczcD) that was cloned between reporter genes malE (red) encoding maltose binding protein and gfp (green) on plasmid pMSLG downstream of the inducible tac promoter P tac . b. Predicted SLczcD secondary structure from Mfold analysis with the czcD start codon underlined and ribosome binding site (RBS) indicated. Site directed mutations predicted to enhance (A41G and A36U) or disrupt (CC29,30GG and CC29,30UU) secondary structure are indicated, whilst bases in bold indicate the region removed in the pMSLG ΔSL construct. Boxed base pairs indicate where disrupting and subsequent compensatory mutations were introduced as indicated. c. E. coli pMSLG-containing cells were grown at 25, 30 or 37˚C and standardised whole cell lysates western blotted and probed with anti-MBP (red) and anti-GFP (green) antisera. All GFP signals were standardised to the corresponding internal loading control MBP signal and normalised to the 25˚C value. Relative GFP quantifications are provided below the blots. These data are representative of at least three replicate experiments. d. Similar analyses as (c) with extracts from strains containing variants of pMSLG. The pMSLG ΔSL plasmid had the bases in bold in (b) removed. The CC29,30GG; CC29,30UU; A41G and A36U derivatives are as indicated in (b). All GFP signals were standardised to the corresponding internal loading control MBP signal and normalised to the pMSLG value. Relative GFP quantifications are provided below the blots. These data are representative of at least three replicate experiments. e. Similar analyses as (c) with compensatory mutants (boxed in b) in the predicted stem region of SLczcD on pMSLG. The SLczcD G59:C38 predicted base pair was disrupted by introducing a G59C mutation and base pairing restored with a C38G mutation. In a similar manner the U34:A63 predicted base pair was disrupted with a U34A mutation and restored with an A63U mutation. All GFP signals were standardised to the corresponding internal loading control MBP signal and normalised to the pMSLG value. Relative GFP quantifications are provided below the blots. These data are representative of at least three replicate experiments. https://doi.org/10.1371/journal.ppat.1009008.g001

PLOS PATHOGENS
RNA-based thermoregulation of a Campylobacter jejuni zinc exporter family member containing Pfam motif PF01545 characteristic of integral membrane proteins that enhance tolerance to divalent cations through functioning as efflux pumps and was previously annotated as a cation efflux protein [39]. It possesses features typical of the Zn(II) exporting CDFs [40] and as a putative Cd(II)-Zn(II)-Co(II) transporter protein we have termed it CzcD. Alignment of C. jejuni CzcD and YiiP, a structurally characterised CDF family member from Escherichia coli [41], demonstrated conservation of the four critical aspartate/histidine residues that form the YiiP membrane embedded Zn(II) binding active site (S1 Fig).
Notably, between Cj1164c and czcD is a 75 bp intergenic region containing an annotated 52 bp stem loop (nucleotides 1095030 to 1095081 of the complementary strand-GenBank accession AL111168.1) that we have termed SLczcD. Mfold secondary structure prediction [42] of the transcript from this intergenic region indicated that the putative czcD Shine-Dalgarno (SD) sequence AGGA was sequestered by base-pairing with a UCCU anti-SD sequence in SLczcD (Fig 1B), SD sequestration being a characteristic feature of RNATs involved in temperature-dependent post-transcriptional gene regulation. Searches of the RefSeq genome database indicated that the SLczcD was present in all C. jejuni but not the vast majority of C. coli strains. To test whether this region conferred temperature responsiveness in the heterologous host E. coli, we created plasmid pMSLG in which the SLczcD region including the czcD start codon (reverse complement of nucleotides 1095025 to 1095084) was cloned between reporter genes malE and gfp, with the latter placed such that it used the czcD start codon (Fig 1A). In the resultant strain, EcpMSLG, while MBP levels did not vary with temperature, GFP levels increased as the temperature was raised from 25 to 37˚C ( Fig 1C) suggesting a thermoregulatory role for SLczcD on the downstream gfp gene. Deletion of residues 1 to 24 of the SLczcD (highlighted in bold in Fig 1B), to create the pMSLG ΔSL construct lacking the stem loop, resulted in similar and significantly elevated levels of GFP at both 25˚C and 37˚C ( Fig 1D). We further examined GFP production in site-directed mutants predicted to enhance or disrupt SLczcD secondary structure and thereby have repressing or de-repressing activity respectively (Fig 1D and 1E). The minimum free energy values of the Mfold predicted secondary structures of these variants are consistent with these predictions (S1 Table). The predicted pairing of the anti-SD:SD sequence was disrupted by substituting the two anti-SD C residues with G (to create pMSLG CC29,30GG ), or U residues (to create pMSLG CC29,30UU ) predicted to form weaker non-canonical base pairing. The resultant strains, EcCC29,30GG and EcCC29,30UU, produced significantly increased GFP levels relative to EcpMSLG at both 25 and 37˚C ( Fig 1E) highlighting the critical role of these residues in regulating production of downstream gene products. Furthermore, two repressing mutations (A36U and A41G, Fig 1B), predicted to increase SLczcD secondary structure through replacing small internal loops with base-pairing, resulted in decreased GFP levels in strains EcA36U and EcA41G compared to EcpMSLG (Fig 1D) demonstrating the importance of these unpaired regions in RNAT functioning. Finally, we constructed two sets of SLczcD mutants (G59C and U34A) that disrupted base-paired stem regions and then restored the predicted base pairing by introducing compensatory mutations (C38G and A63U respectively; Fig 1B). As predicted GFP levels increased in the EcG59C and EcU34A strains relative to EcpMSLG and were reduced to levels similar to those for the wild-type SLczcD following the introduction of compensatory mutations in strains EcG59C,C38G and EcU34A,A63U (Fig 1E).
To further confirm temperature responsiveness in E. coli we utilised a lacZ reporter gene placed immediately downstream of the SLczcD on a pMLBAD derivative plasmid termed pSLlacZ (see Materials and Methods). Derivatives of this plasmid with repressing (pSLlac-ZA41G) and derepressing (pSLlacZCC29,30GG) mutations were also constructed in a similar manner to as described above. Reporter activity measurements of cultures containing these plasmids incubated at 30, 37 and 42˚C confirmed temperature responsiveness of SLczcD, and the effects of repressing and derepressing mutants were consistent with SLczcD structural prediction and western blotting data presented above (S3 Fig).
Overall these data demonstrate that the SLczcD acts independently of other C. jejuni factors to confer temperature responsive regulation of downstream genes in a heterologous host. Mutations predicted to disrupt or enhance secondary structure of the SLczcD led to predicted changes in the level of downstream gene product consistent with this region functioning as an RNAT. The ability to act independently of additional endogenous factors in a heterologous host is a notable feature of RNATs due to the conserved nature of core translational machinery that they rely on for function.

SLczcD-dependent temperature regulation of CzcD production in C. jejuni
To investigate CzcD temperature regulation in C. jejuni directly we produced an NCTC 11168 strain that produces hexa-histidine tagged CzcD (CjCzcDhis) from the native locus (see Materials and Methods) and tested for CzcD thermoregulation. As predicted, CzcDhis levels increased with temperature from 32 to 37˚C, and again from 37 to 42˚C (Fig 2A). Importantly, this was not due to a significant increase in transcript abundance as measured by RT-qPCR (Fig 2B), consistent with RNAT-mediated regulation at the translational level. To determine if the SLczcD was responsible for CzcD temperature regulation, we created a derivative of CjCzcDhis in which the SLczcD was removed (CjΔSL). In this strain CzcD temperature regulation was reduced (Fig 2A). A strain with de-repressing mutations CC29,30GG in the SLczcD, termed CjCC29,30GG, retained only a limited degree of temperature induction (Fig 2A). These data demonstrate the role of the cj1164-czcD intergenic region in temperature regulation of CzcD production and are consistent with the predicted stem loop region, SLczcD, within this intergenic region acting as an RNAT.

CzcD is a Zn(II) resistance determinant
To investigate the function of CzcD, it's gene was inactivated by insertional mutagenesis in C. jejuni NCTC 11168 to create strain CjczcDand the ability of this strain to grow in media

PLOS PATHOGENS
RNA-based thermoregulation of a Campylobacter jejuni zinc exporter supplemented with increasing concentrations of a range of metals examined. The CjczcDstrain was hypersensitive to Zn(II) compared to the parental wild-type (>5-fold), and to a lesser extent cobalt, but not to other metals tested ( Fig 3A). Notably, Zn(II) resistance was restored to the CjczcDstrain by reintroducing a single copy of czcD in trans on the chromosome to create the genetically complemented CjczcD + strain ( Fig 3A) and similar complementation with czcD site directed mutants with altered Zn(II) binding sites (S1 Fig) failed to restore Zn(II) tolerance (S4 Fig). We therefore demonstrate a primary role for CzcD in C. jejuni Zn(II) resistance. A likely consequence of loss of a Zn(II) exporter is that cells over-accumulate Zn(II). Indeed total cellular Zn(II) content of CjczcDcells, as measured by inductivelycoupled plasma mass spectroscopy (ICP-MS), was increased compared to wild-type following growth in media both with and without 25 μM Zn(II) supplementation ( Fig 3B), a level which is not inhibitory to either strain ( Fig 3A).

Zn(II) and Co(II) regulation of CzcD
Having established that CzcD is a Zn(II)/Co(II) resistance determinant we investigated whether these metals regulate CzcD production. When CjCzcDhis was grown with varying sub-inhibitory levels of Zn(II), Co(II) or copper, CzcD levels increased with added Zn(II) and Co(II) but not copper (Fig 4A), consistent with a role for CzcD in Zn(II) and Co(II) export. Both qRT-PCR ( Fig 4B) and promoter reporter fusion data (S5 Fig) indicate that this is a transcriptional effect on the promoter driving czcD expression (Fig 1A). In other bacteria, Zn(II)exporting CzcD proteins are typically expressed in response to Zn(II) due to the actions of one of the Zn(II)-dependent transcriptional regulators ZntR, CzrA or ScrA (and their aliases). However, whilst related sequences are present within the C. jejuni genome they lack the characteristic Zn(II)-sensing motifs (reviewed in [43]), thus providing few clues as to the mechanism of CzcD metalloregulation. The czcD gene is transcribed downstream of Cj1164c, encoding a Zn(II) finger-containing protein, which could potentially influence CzcD expression ( Fig 1A). However, disruption of Cj1164c (by insertion of a promoterless kanamycin resistance cassette) did not abolish Zn(II) induction of CzcDhis (S6 Fig). Importantly, removal of the 75 bp Cj1164c/czcD intergenic region in strain CjΔSL (Fig 1A) also did not abolish Zn (II)-dependent induction of CzcDhis (Fig 4A), demonstrating that the Zn(II)-responsive expression of CzcD is independent of the RNAT.

C. jejuni SLczcD-mediated temperature dependent Zn(II) sensitivity
Having demonstrated temperature-dependent CzcD production and the role of CzcD in mediating Zn(II) resistance we predicted that C. jejuni Zn(II) resistance would be increased at the elevated temperatures of a host. The ability of C. jejuni to grow in media titrated with increasing Zn(II) levels was therefore examined over a range of temperatures (32, 37 and 42˚C) that support growth. Note that for these experiments a defined medium, minimal essential medium alpha (MEM-α), was used that allows C. jejuni to tolerate higher levels of Zn(II) supplementation ( Fig 5A) than when grown in Mueller-Hinton (MH) medium ( Fig 3A). Following 8 hours growth the OD 600 of cultures was measured and data normalised to readings obtained without added Zn(II) (Fig 5). For the wild-type strain Zn(II) resistance was elevated at both 37 and 42˚C compared to 32˚C (Fig 5A). Importantly the derived CjΔSL strain was no longer more sensitive to Zn(II) at 32˚C (Fig 5B) precluding a more general effect of differing growth rate of strains on zinc sensitivity. This temperature-mediated difference in Zn(II) sensitivity was also lost on inactivation of czcD (Fig 5C), with the CjczcDstrain showing similar Zn(II) hypersensitivity at 32 and 37˚C. Combined, these findings confirm SLczcD-mediated temperature regulation of CzcD in C. jejuni confers increased Zn(II) resistance at host body temperatures.

CzcD enhances killing in the Galleria mellonella larval infection model
We evaluated the role of CzcD using the G. mellonella larval infection model. Larvae were infected with wild-type C. jejuni NCTC 11168, the czcDand the genetically complemented czcD + strain. Approximately 10 6 colony forming units (cfu) of mid logarithmic phase MH broth grown bacteria were injected into larvae and following incubation at 37˚C larvae were scored as alive or dead at daily intervals. The CjczcDstrain was attenuated compared to the parental wild-type strain and wild-type levels of larval killing were restored in the CjczcD + complemented strain (Fig 6A). These data demonstrate a role for CzcD in C. jejuni larval killing in this simple infection model and suggest that host induced Zn(II) toxicity is a feature of the innate immune response in G. mellonella larvae.
An advantage of the G. mellonella model system is that larvae can be incubated at a range of temperatures and we therefore investigated C. jejuni induced larval killing at 32˚C, a temperature known to inhibit CzcD production in C. jejuni (Fig 2A). Wild-type C. jejuni and derivatives with de-repressing mutations within SLczcD (CjCC29,30GG) or with the SLczcD removed (CjΔSL) were similarly used to inoculate larvae but at a higher dose of 10 7 , rather    6 CFU injected into G. mellonella larvae that were incubated at 37˚C and scored daily for survival. Strains were wild-type NCTC 11168, czcD insertionally inactivated mutant (CjczcD -), the derived genetically complemented strain (CjczcD + ), the derivative lacking SLczcD (Cj wt ΔSL) or with the derepressing mutation CC29,30GG (CjCC29,30GG). Data represent five biological replicates for all strains except CjczcD + with four biological replicates. Each biological replicate consisted of injecting ten larvae (i.e. n = 50 or 40). Pairwise p values were calculated using a Log-rank (Mantel-Cox) test implemented in GraphPad Prism. Data for the CjczcDstrain was statistically significantly different when compared to all four other strains with p values of <0.0001, <0.0001, 0.0001 and 0.0004 for the wild-type, CjczcD + , CjCC29,30GG and CjΔSL comparisons, respectively. No other pairwise comparisons were statistically significant. b. Similar experiment as for (a) except C. jejuni strains were grown and larvae incubated at 32˚C, and larvae were injected with~10 7 CFU. Data represent three biological replicates with ten larvae per experiment per strain (n = 30). P values were calculated as in (a). The results obtained for pairwise comparison of the wild-type with CjCC29,30GG and CjΔSL strains were statistically significant with p values of 0.0009 and <0.0001, respectively. There was no statistically significant difference between the CjCC29,30GG and CjΔSL strains. https://doi.org/10.1371/journal.ppat.1009008.g006

PLOS PATHOGENS
RNA-based thermoregulation of a Campylobacter jejuni zinc exporter than 10 6 cfu, to overcome the delayed killing at this lower temperature. We have shown herein that at 32˚C CzcD levels are elevated in CjCC29,30GG and CjΔSL compared to the wild-type (Fig 2A). Crucially, both the CjCC29,30GG and CjΔSL strains show enhanced killing of G. mellonella larvae compared to wild-type at 32˚C (Fig 6B), but not at 37˚C (Fig 6A). These data demonstrate that CzcD is involved in larval killing regulated by temperature through the activity of the SLczcD RNAT.

Discussion
Several lines of evidence demonstrate a role for the C. jejuni CDF family member CzcD as a Zn(II)/Co(II) exporter: (i) its expression is induced by Zn(II) and Co(II) (Fig 4A); (ii) C. jejuni mutants with inactivated czcD are hypersensitive to Zn(II) and to a lesser extent Co(II) ( Fig  3A); and (iii) czcD mutants over accumulate Zn(II) (Fig 3B). Importantly, we show that CzcD production is controlled at the post transcriptional level by an intergenic RNAT located immediately upstream of czcD (Figs 1 and 2), which confers Zn(II) resistance in response to the environmental cue of elevated host body temperatures (Fig 5). The RNAT allows for temperature-regulated gene expression independently of other C. jejuni factors since it is functional in a heterologous host, E. coli (Figs 1 and S3). Finally, using a Galleria infection model we show that CzcD is involved in larval killing ( Fig 6A) and that disruption of the RNAT, thereby allowing CzcD production at lower temperatures, increases killing at these lower temperatures relative to the wild-type strain (Fig 6B). This is the first identified C. jejuni RNAT and the association with a Zn(II)-resistance determinant has not been previously reported in bacteria. Its role in regulating production of CzcD suggests that Zn(II) export is important during interaction of this pathogen with a host, ultimately allowing for successful colonisation.
Overcoming metal toxicity is emerging as a virulence mechanism of diverse bacterial pathogens with Zn(II) export to overcome host induced Zn(II) toxicity now an established bacterial virulence mechanism [44]. In Helicobacter pylori, a close relative of C. jejuni, a metal efflux pump of the resistance-nodulation-cell division (RND)-type is required for gastric colonization in a gerbil model [45]. Notably in Streptococcus pyogenes, CzcD was also protective against innate immune responses involving exposure of bacteria to elevated levels of Zn(II) in neutrophils [22,46]. A recent global transposon mutagenesis screen demonstrated the significance of C. jejuni metal handling systems for colonisation of a mouse intestinal infection model [23]. Mutants with insertions in the various iron uptake systems were surprisingly not colonisation defective although this may be due to redundancy, whilst insertions within genes encoding deduced copper and Zn(II) handling systems were significantly under-represented in output pools. However, for copper, despite the multi-copper oxidase CueO and Cu(I) exporter Cj1161c/CopA being required for C. jejuni copper tolerance [38] insertions within either of these genes did not significantly affect mouse colonisation whereas a colonisation defect was associated with a second predicted multi-copper oxidase. Notably, insertions within both the znuABC genes encoding a Zn(II) uptake system previously shown to be important for chicken colonisation [17,47] and Zn(II) resistance associated czcD characterised herein were significantly underrepresented in the output pools from the mouse intestine. These data suggest that both Zn(II) limitation and poisoning are encountered by C. jejuni in this model. The identification of CzcD as important for both mouse intestinal colonisation and for host killing in a larval infection model highlights the significance of countering Zn(II) toxicity in these diverse hosts and suggests a conserved Zn(II) poisoning strategy employed in both. Furthermore, in common with the regulation of key virulence factors in major pathogens, the exploitation of RNA-mediated thermosensing to regulate its expression restricts CzcD production to environments with elevated host body temperatures.

RNA-based thermoregulation of a Campylobacter jejuni zinc exporter
The relationship between Zn(II) and Campylobacter infection is therefore complex with a requirement for systems to overcome both Zn(II) deficiency and toxicity for C. jejuni host colonisation [17,47]. Moreover, Zn(II) sequestration by the host antimicrobial protein S100A12 (alias calgranulin C), which accumulates in the gastrointestinal tract of C. jejuni infected patients, is known to inhibit C. jejuni growth [48]. Yet, using a recently developed mouse model of C. jejuni infection, a Zn(II) deficient diet was also seen to promote more severe outcomes of infection including more persistent and bloody diarrhoea, increased weight loss, prolonged inflammation, increased levels of inflammatory biomarkers and altered intestinal morphology [49]. Thus, the fine-tuning of both Zn(II) uptake and Zn(II) efflux are likely to be a requirement for C. jejuni for successful host colonisation and infection, and it will be intriguing to identify where and when these opposing activities are critical. The precise coordination of Zn(II) import and export to protect against host innate immune defences is also a likely requirement for other pathogens. Notably, group A Streptococcus must overcome both Zn(II) starvation and toxicity during infection, with the former being encountered extracellularly and the latter upon phagocytosis by neutrophils, with Zn(II) import and export systems contributing to virulence [22].
The czcD gene is located within the same operon as genes Cj1162c and Cj1161c/CopA involved in copper homeostasis (Fig 1A). Cj1161c/CopA, a copper translocating P 1 -type ATPase, is involved in the export of copper from the cytoplasm to the periplasm [38] and a C. jejuni copA transposon mutant was hypoinvasive for cultured intestinal epithelial cells [50]. Furthermore, insertional inactivation of copA or cueO, also associated with copper-resistance [38], resulted in C. jejuni attenuation in the chicken colonisation model [51]. The presumed transcriptional co-regulation of Zn(II) (CzcD) and copper (Cj1161c/CopA) resistance determinants in C. jejuni suggests a response against the so-called "brass dagger" component of the innate immune response, whereby bacteria are killed through exposure to the combined toxicity of copper and Zn(II) [52]. Indeed, the combinatorial activity of copper and Zn(II) intoxication against Salmonella, with a bacterial requirement for both Zn(II) and copper export, has been demonstrated in infected macrophages [19]. Significantly we have demonstrated that the SLczcD, acts to regulate CzcD at the post transcriptional level, and may therefore enable a further level of differential regulation of the Zn(II) and copper resistance determinants mediated via temperature. It is tempting to speculate that C. jejuni encounters environments, presumably outside of its warm blooded hosts, where Zn(II) export is disadvantageous and yet copper export must be maintained.
The tight control of bacterial cellular Zn(II) pools is needed to ensure a supply for metabolic processes whilst avoiding Zn(II)-poisoning [8]. However, this also provides a potential vulnerability to Campylobacter that can be exploited for the development of antimicrobials and for novel approaches for its control in the food chain. Zn(II) has been used as an antimicrobial, for example in food packaging [53], as well as in animal (including poultry) feeds to promote growth and to restrict the proliferation of pathogenic bacteria in the gut [54,55]. Indeed, feeding piglets high dosage Zn(II) oxide causes a significant reduction in their faecal excretion of Campylobacter coli [56]. Clinical trials have also demonstrated the benefits of adding Zn(II) to oral rehydration solution in reducing the morbidity and mortality associated with severe diarrheal diseases in infants [57,58]. The identification of a temperature regulated C. jejuni Zn(II) resistance determinant herein potentially informs further development of strategies to exploit metal toxicity and reduce Campylobacter levels in food products.
A major finding of our work is that an RNAT functions in C. jejuni to regulate CzcD production. At the sequence level RNATs are poorly conserved among bacterial species, though distinct structural classes such as ROSE (repression of heat shock gene expression) and FourU elements have been proposed. ROSE elements are relatively complex with between two and four stem-loop regions and typically regulate heat shock genes [59], whilst fourU RNATs characteristically contain four contiguous uridine residues base-pairing with the AGGA SD sequence [60]. The czcD RNAT belongs to neither class having a relatively small structure combined with perfect Watson-Crick base pairing of the SD (AGGA) and anti-SD (UCCU) sequence (Fig 1), an arrangement found in the Synechocystis heat shock protein 17 RNAT [61]. Although the predicted canonical pairing between the SD and the anti-SD results in relatively high predicted thermostability (S1 Table), this may be alleviated by the large proportion of weaker A-U pairs including a stretch of five A-U pairs immediately above the RBS region and the presence of two small internal loops (A20�C35 and A15�A40) in the stem region which initiate melting of the structure in response to temperature [62][63][64]. Indeed, mutations that introduced complementary base pairs at these two regions had a significant repressing effect on temperature regulation in E. coli (Fig 1D). Both the stem and loop regions in the czcD RNAT therefore appear important in the stability and flexibility of the structure, thus enabling it to respond in C. jejuni to the small temperature shifts (32 to 42˚C) reported here. The czcD RNAT is also unusual due to being coded for in an intergenic region of a multigene operon (Fig 1A). RNATs are generally found in the 5' UTR of a transcript though in Yersinia an RNAT is located between the yscW and lcrF genes [29] in a dicistronic operon. Hence, the structure and intra-genic location of the C. jejuni RNAT in a four gene operon appear unique.
It is now clear that innate immune defences exploit the combined toxic properties of Zn(II) and copper. Here we reveal that Zn(II) resistance in C. jejuni is triggered post-transcriptionally by temperature, via an RNAT. We speculate that this novel C. jejuni RNAT provides a mechanism for the expression of the co-transcribed Zn(II) and copper exporters to be uncoupled such that Zn(II) export is prevented in the cooler environments outside of its hosts, presumably where Zn(II) is limited and Zn(II) export is disadvantageous, but is rapidly expressed in response to the environmental signal of its warm blooded hosts and where both Zn(II) and copper resistance are required.

Bacterial strains and growth conditions
Primers, strains and plasmids are described in S2, S3 and S4 Tables respectively. E. coli were routinely grown at 37˚C on Luria-Bertani (LB) agar plates or in LB broth with shaking. C. jejuni NCTC 11168 and derived strains were grown at temperatures of 32, 37 or 42˚C in a Variable Atmosphere Incubator (Don Whitley Ltd.) with gas settings of 5% O 2 , 10% CO 2 and 85% N 2 . Strains were maintained on Mueller-Hinton (MH) or Columbia blood agar (CBA) plates, and growth experiments carried out in either MH broth or defined minimal essential medium alpha (MEM-α) with no glutamine or phenol red (Gibco) and supplemented with 20 mM serine and pyruvate. Media were supplemented where appropriate with antibiotics at the following concentrations: ampicillin (100 μg/ml), kanamycin (50 μg/ml), and chloramphenicol (17 μg/ml for C. jejuni or 34 μg/ml for E. coli).

Inactivation of czcD and genetic complementation
The czcD gene was inactivated by insertion of a kanamycin resistance cassette lacking a terminator with the same transcriptional polarity. Approximately 550 bp PCR amplicons, containing 5' and 3' regions of czcD, were produced using primer pairs Cj1163-UF/Cj1163-UR and Cj1163-DF/Cj1163-DR, with HindIII sites incorporated within Cj1163-UR and Cj1163-DF. The two products were joined together in an overlap PCR reaction and ligated to pGEM-T Easy (Promega) to create pGEMCj1163c. The kanamycin resistance cassette was ligated into the HindIII site of pGEMCj1163c, immediately following base 322 of the czcD sequence, and

PLOS PATHOGENS
RNA-based thermoregulation of a Campylobacter jejuni zinc exporter the resulting construct introduced into C. jejuni NCTC 11168 by electroporation. Mutants were selected on CBA plates supplemented with kanamycin and interruption of czcD by insertion of the kanamycin resistance cassette verified by PCR and sequencing.
For complementation of the czcD mutant, a linear DNA fragment was generated by the gene Splicing by Overlap Extension (SOEing) method (as outlined in [65]), comprising the Cj0223c pseudogene sequence flanking a chloramphenicol resistance cassette and czcD. Briefly three PCR products were generated (i) 1313 bp containing Cj0223 5' sequences and the chloramphenicol resistance cassette with associated promoter using primers 0223upF and endcatR with pCJC1 [66] as template, (ii) 726 bp containing czcD using primers Cj1163compF and Cj1163compR with C. jejuni NCTC 11168 genomic DNA as template, and (iii) 973 bp containing Cj0223 3' sequences using primers 0223downf and 0223downR with pCJC1 as template. Following purification, 50, 90 and 50 ng, respectively, of the products were used with Phusion High-fidelity PCR master mix (New England Biolabs) in a reaction with primers 1633F and 1636R at a final concentration of 0.06 μM to generate the full length product (3.012 kb), using the following conditions: 1 cycle of 98˚C for 3 minutes, 61˚C for 1 minute, 72˚C for 10 minutes, 98˚C for 1 minute; 40 cycles of 98˚C for 15 seconds, 55˚C for 20 seconds and 72˚C for 1 minute; followed by a 10 minute final extension at 72˚C. The resulting PCR product, with the chloramphenicol resistance cassette and associated promoter located immediately upstream of czcD and in the same transcriptional orientation so that it can drive czcD expression, was introduced into the czcDstrain by electroporation and transformants selected on chloramphenicol-containing CBA.

Construction of C. jejuni strains expressing histidine-tagged CzcD and variant Cj1164c/czcD intergenic regions
A ligase cycling reaction (LCR) was used to assemble DNA fragments, using single-stranded synthetic bridging oligonucleotides as described by [67], to generate a C. jejuni strain that produces a C-terminal hexa-histidine tagged CzcD (CjCzcDhis) from the native locus. Two DNA fragments were PCR amplified using C. jejuni NCTC 11168 genomic DNA as template: (i) containing czcD with six histidine codons (cat cac cat cac cat cac) at the 3' end plus 545 bp of czcD upstream sequences using primers LCR_A F and LCR_B R, and (ii) 558-bp of czcD downstream sequences using primers LCR_C F and LCR_C R. The fragments were 5 0 -phosphorylated before assembly, using T4 polynucleotide kinase in reactions containing 90 fmol of each DNA fragment, 5 mM ATP, 10 U T4 polynucleotide kinase, and 1× Ampligase reaction buffer (Epicenter Biotechnologies, Madison, WI) for 1 hour at 37˚C followed by 20 minutes incubation at 65˚C. Ligase cycling reactions were subsequently performed with 7.5 U Ampligase thermostable DNA ligase, 3 nM of each phosphorylated DNA fragment, and 30 nM of the bridging oligonucleotide bridge_BC with 1 cycle of 94˚C for 2 minutes followed by 50 cycles of 94˚C for 10 seconds, 55˚C for 30 seconds, and 66˚C for 60 seconds. The resulting fragment, containing hexa histidine-tagged czcD flanked by wild-type 5' and 3' sequences, was introduced into the Zn(II)-sensitive CjczcDstrain by electroporation and mutants selected on CBA supplemented with 300 μM ZnSO 4 (a level that is inhibitory for CjczcD -). Integration of the DNA fragment and replacement of the kanamycin resistance cassette was confirmed by PCR and sequencing. The resulting strain is essentially wild-type other than the introduction of six histidine codons at the 3' end of czcD to code for CzcDhis.
A derivative of CjCzcDhis with de-repressing mutations (CC29,30GG) in the SLczcD was generated in the same way but assembled from three fragments: (i) 545 bp of czcD upstream sequences amplified using primers LCR_A F and LCR_A R, the latter including a CC29,30GG substitution within SLczcD, (ii) containing hexa-histidine tagged czcD amplified using primers

PLOS PATHOGENS
RNA-based thermoregulation of a Campylobacter jejuni zinc exporter LCR_B F and LCR_B R, and (iii) 558 bp of czcD downstream sequences using primers LCR_C F and LCR_C R, and using the bridging oligonucleotide bridge_AB.
Two further derivatives lacking SLczcD, in the CzcDhis producing strain and in wild-type NCTC 11168, were similarly generated but using the SOEing method. Two PCR products were generated using C. jejuni containing czcD-his or wild-type NCTC 11168 genomic DNA as template: (i) 511 bp containing SLczcD 5' sequences using primers LCR_A F and deltaSL1, and (ii) 1512 bp containing SLczcD 3' sequences including hexa-histidine tagged czcD and downstream sequences using primers deltaSL2 and LCR_C R. Following purification, the two PCR products were mixed in a 1:1 ratio and used as template in Phusion High-Fidelity PCR to generate the full length 2.013 kb product with primers LCR_A F and LCR_C R (0.06 μM final concentration), using the following conditions: 1 cycle of 98˚C for 3 minutes, 61˚C for 1 minute, 72˚C for 10 minutes, 98˚C for 1 minute; and 40 cycles of 98˚C for 15 seconds, 65˚C for 20 seconds and 72˚C for 1 minute; followed by a 10 minute final extension at 72˚C. The final product was ligated to pGEM-T Easy and the resulting construct introduced into the czcDstrain of C. jejuni by electroporation and mutants selected on CBA supplemented with 300 μM ZnSO 4 as above.

Determination of metal tolerance and Zn(II) quotas
C. jejuni strains were grown on MH agar plates, sub-cultured into MH broth and following overnight growth diluted to an OD 600 of 0.05 in MH broth or MEM-α supplemented with varying levels of metals (5 ml cultures in wells of six well tissue culture plates). Growth was monitored by measuring OD 600 . Metal quotas of cells were determined essentially as described previously [68], but using cells washed in Dulbecco's PBS without CaCl 2 and MgCl 2 for cell culture (Sigma-Aldrich) and expressed as atoms per absorbance unit at 595 nm (determined using a synergy HT plate reader, BioTek Instruments Inc.). All assays were performed in triplicate on at least three separate occasions.

SDS-PAGE and Western blot analyses
Western blot analyses were performed using whole cell lysates of E. coli and C. jejuni grown as described in individual experiments. Cells were harvested by centrifugation, suspended in PBS to a standardised OD 600 and LDS loading buffer (Thermo Fisher Scientific) added. Samples were incubated at 95˚C for 5 minutes, centrifuged at 12,000 g for 5 minutes and electrophoresed for 90 minutes at 130 V in 10% NuPAGE Bis-Tris gels (Invitrogen) using NuPAGE MES (2-(N-morpholino)ethanesulfonic acid) SDS Running Buffer (Invitrogen). Proteins were transferred to nitrocellulose membrane using an XCell blot module (Invitrogen) in Tris-Glycine transfer buffer [192 mM glycine, 25 mM Tris pH 8.3, 20% (v/v) methanol, 0.01% (w/v) SDS] at 25 V for 90 minutes. The membrane was blocked with 3% (w/v) BSA in PBS overnight at 4˚C, washed twice in PBS with 0.1% (v/v) Tween 20 and once in PBS for 5 minutes. For detection of histidine-tagged proteins, the membrane was incubated for one hour at room temperature with anti-histidine primary antibody (Qiagen), diluted 1:1000 in PBS, followed by a one hour incubation with IRDye 680RD goat anti-mouse IgG (LiCor Biosciences, Lincoln USA) as a secondary antibody. For detection of GFP, rabbit-derived anti-GFP antibody (Invitrogen) diluted 1:2000 in PBS and IRDye 800CW goat-derived anti-rabbit IgG (LiCor Biosciences) were used. For detection of maltose-binding protein (MBP), murine anti-MBP antibody (New England Biolabs) diluted 1:10,000 was used as primary antibody followed by IRDye 680RD goat anti-mouse IgG (LiCor Biosciences). Anti-GroEL and anti-RecA antibodies (Abcam) were diluted 1:1000. After washing membranes in PBS, signals were detected using Odyssey CLx imaging system (Li-Cor Biosciences) and signals quantified using Image Studio software (LiCor Biosciences).

Construction of the malE/gfp and lacZ E. coli reporter plasmids and derivatives
A reporter construct was generated which contained SLczcD and the czcD start codon placed between the two reporter genes malE and gfp such that gfp was placed in the same position as czcD with respect to SLczcD on the C. jejuni NCTC 11168 chromosome. The gfp TCD gene [69] was ligated into NdeI/SalI digested pMAL-c5x (New England Biolabs) downstream of malE to create pMG. A DNA fragment containing SLczcD and the czcD start codon (plus 7 nucleotides at the 5 0 end that contain a stop codon and an NdeI-compatible overhang, and an NcoI-compatible overhang at the 3 0 end) was constructed by annealing two single-stranded complementary synthetic oligonucleotides SL1 and SL2 (S2 Table). These were incubated at equimolar concentrations in annealing buffer (10 mM Tris, 1 mM EDTA, 50 mM NaCl, pH 8.0) at 95˚C for 5 minutes and cooled to room temperature. The annealed fragment was then ligated into NdeI/NcoI digested pMG (note the NcoI site is at the 5'end of the introduced gfp TCD ) and the resultant construct, pMSLG, was confirmed by PCR and sequencing. Q5 Site-Directed Mutagenesis (New England Biolabs) was used to generate pMSLG with the following mutations within SLczcD (refer to Fig 1 for numbering): CC29,30GG; CC29,30UU; A36U; A41G; G59C; U34A; G59C,C38G; and U34A,A63U; plus a derivative lacking residues 22 to 45 of the SLczcD (designated pMSLG ΔSL ).
To create a second lacZ based reporter system a PCR amplified lacZ gene (primer pair lacZNheIF and laczXbaIR-S2 Table) was directionally cloned into plasmid pMLBAD using NheI and XbaI sites to create pBADlacZ. The SLczcD with NcoI and NdeI termini was ligated into pBADlacZ immediately upstream of the NdeI creating pSLlacZ. In this construct transcription of the stem-loop regulated reporter lacZ is driven by the arabinose-inducible and glucose-repressible P BAD promoter. Plasmid pSLlacZ derivatives pSLlacZA41G and pSLlacZCC29,30GG were created by site directed mutagenesis. Strains of E. coli containing each plasmid were grown to mid-log phase at 37˚C, and induced with arabinose at 30, 37, or 42˚C for two hours. Cells were harvested by centrifugation, re-suspended in Z buffer (0.06 M Na 2 HPO 4 �7H 2 O, 0.04 M NaH 2 PO 4 �H 2 O, 0.01 M KCl, 0.001 M MgSO 4 , 0.27% [v/v] β-mercaptoethanol [Sigma-Aldrich]) and permeabilised with 100 μl of chloroform (Sigma-Aldrich) and 50 μl of 0.1% (w/v) sodium dodecyl sulphate (Sigma-Aldrich) per ml of cell suspension. Following addition of ortho-nitrophenyl-β-galactoside reactions were incubated at 28˚C until coloured products were visible and Na 2 CO 3 added. Absorbance values at 420 nm were used to calculate β-galactosidase activity in Miller units.

Construction and analysis of C. jejuni GFP reporter strains
The pCJC1 plasmid [66] was modified by introduction of the Cj1164c associated promoter with gfp placed immediately downstream. These were inserted upstream of the chloramphenicol resistance cassette (cat) to create plasmid pCJC1Pgfp. This was electroporated into C. jejuni NCTC 11168 and recombinants selected on chloramphenicol-containing media. Insertion of the Cj1164c promoter-gfp-cat cassette within the pseudogene Cj0223c to create strain CjPgfp was verified by PCR. Following growth of this strain in MEM-α shaking at 37˚C for 24 hours, cells were pelleted by centrifugation, suspended in PBS to an OD 600 of 1.0 and 200 μl transferred in triplicate into a black-walled 96-well plate with a transparent base. Relative fluorescence units (RFU) with excitation at 485/20 nm and emission at 528/20 nm and OD 600 were measured in a Bio-Tek Synergy HT plate reader.

PLOS PATHOGENS
RNA-based thermoregulation of a Campylobacter jejuni zinc exporter

RNA purification and quantitative RT-PCR
Mid-logarithmic C. jejuni cultures were mixed with RNA Protect Reagent (Qiagen) before total RNA extraction and purification using the RNeasy Mini Kit (Qiagen) following the manufacturer's instructions. Samples were treated with 2.5 μl of DNase I for one hour at 37˚C to eliminate DNA contamination. Reverse transcriptase reactions were carried out on 1 μg total RNA using a QuantiTect Reverse transcription kit (Qiagen) according to manufacturer's instructions. cDNA samples were diluted 1:3 in RNase free water and quantitative RT-PCR was performed using SYBR Green I (Applied Biosystems). The PCR mix consisted of 4 μl of cDNA sample, 238 nM of forward and reverse primers, and 4 μl of Fast SYBR Green Master Mix. Each reaction contained four technical repeats and was carried out in an Applied Biosystems StepOnePlus Real-Time PCR system. RT-negative controls were included for all experiments. The rpoA gene was used as an internal control. Amplification conditions were initial denaturation at 95˚C for 20 seconds, followed by 40 cycles of 95˚C for 3 seconds and 60˚C for 30 seconds. Similar primer pair efficiencies were ensured by performing a standard curve with serially diluted DNA. Relative gene expression values were calculated using the Livak method [70].

Galleria infection
Following growth in MH broth to an OD 600 of between 0.4 and 0.5 (mid-log growth phase), C. jejuni cells were pelleted by centrifugation, washed twice and resuspended to an OD 600 of 1.0 in sterile PBS. For each replicate experiment ten larvae of Galleria mellonella (Livefoods Direct Ltd. Sheffield, UK) were injected using a 0.5 ml 29G 12.7 mm insulin syringe (VWR International Ltd.) with 10 μl of cell suspension through the rear proleg. Each experiment included a control batch of larvae injected with PBS.
Supporting information S1 Fig. Amino acid sequence alignment of C. jejuni CzcD and E. coli YiiP. The four active site residues for metal transport in YiiP [41] are highlighted in blue along with aligned conserved CzcD residues H73, D77, H179 and D183. Alignment was produced using Clustal Omega. Strains of E. coli containing plasmids pSLlacZ, pSLlacZA41G or pSLlacZCC29,30GG as indicated were grown to mid log phase and expression of lacZ induced for two hours. Cells were harvested and β-galactosidase activities measured. Data presented are the mean of three (pSLlacZ), five (pSLlacZA41G) or two (pSLlacZCC29,30GG) biological repeats. Error bars represent standard error of the mean with � indicating a t-test p value of <0.05. (TIF)

PLOS PATHOGENS
RNA-based thermoregulation of a Campylobacter jejuni zinc exporter S4 Fig. C. jejuni CzcD active site residue mutants do not restore Zn(II) tolerance. The CjczcDstrain was genetically complemented with wild-type czcD (CjczcD + ) and with two mutant versions each with two alanine substitutions in the Zn(II) binding active site as identified in S1 Fig