The Metallophore Staphylopine Enables Staphylococcus aureus To Compete with the Host for Zinc and Overcome Nutritional Immunity

Examination of the staphylococcal genome identified a single putative Zn-associated ABC permease (NWMN_2306, 1458, and 1459). The transporter was designated AdcABC due to its similarity to other Gram-positive Zn-specific ABC permeases. Bioinformatic analysis of AdcA (NWMN_2306), the metal ion-recruiting component of the permease, revealed that it is comprised of two domains. The N-terminal domain (amino acids 1 to 322) belonged to the cluster A-I subgroup of solute binding proteins (SBPs), which have specific roles in Mn, Fe, and Zn recruitment (36). The C-terminal domain (amino acids 323 to 531) showed high similarity to ZinT, a periplasmic metallochaperone present in some Gram-negative bacteria (37). This SBP composition is unique to the Adc permeases of Gram-positive bacteria such as S. pneumoniae (38). The ABC transporter AdcBC is comprised of a transmembrane component, AdcB (NWMN_1458), and a nucleotide-binding domain, AdcC (NWMN_1459). Unusually, the gene encoding S. aureus AdcA was not clustered with the ABC transporter.

Initially, we examined the role of the putative AdcABC permease in Zn acquisition by assessing its transcriptional response to transition metal limitation. Transcription of adcA was increased in Zn-depleted medium (Fig. 1A) and in a strain in which the Zn-responsive repressor Zur had been deleted (Fig. 1B) (39). In contrast, no change in adcA transcription was observed in Mn- or Fe-depleted medium (Fig. 1A) or in strains that lacked the Mn-responsive regulator MntR or the Fe-responsive regulator Fur (Fig. 1B) (40, 41). Taken together, the S. aureus AdcABC system shows structural and transcriptional features consistent with a bacterial Zn importer (17, 42–44). We then investigated the phenotypic impact of metal restriction on an S. aureus strain that lacks AdcA (the ΔadcA mutant). Limiting Mn and Fe availability had no impact on the growth of the ΔadcA mutant (Fig. 1C). Surprisingly, Zn-depleted medium also had no effect on the growth of the ΔadcA mutant despite the apparent lack of other Zn import pathways in the staphylococcal genome (Fig. 1C and S1A). Collectively, these findings suggested that in addition to the AdcABC permease, S. aureus possesses an import pathway for Zn that is distinct from previously described mechanisms for obtaining this metal.

In Zn-limited medium, in addition to AdcABC, S. aureus expresses the Opp/NikA family transporter CntABCDF (45). However, in these studies loss of CntA did not result in a Zn uptake defect. Loss of CntA did reduce the ability of S. aureus to transport Co and Ni. These results led to the suggestion that Co and Ni are the physiological substrates of the system (45, 46). Building on our observation that the AdcABC permease is not the sole Zn import pathway, we reevaluated the CntABCDF permease and its potential contribution to Zn import. Consistent with prior results (45), the cnt locus was induced in response to Zn limitation (Fig. 1A). Notably, addition of Co and Ni had no impact on cnt transcription (Fig. S1B and C). Removal of Fe or Mn from the growth medium also resulted in a modest increase in expression (Fig. 1A). When Zn, Mn, and Fe were omitted from the medium, expression of the system increased by 21-fold, compared to 4.6-fold in medium that lacked only Zn (Fig. 1A). To explore the regulation of this system further, cnt transcription in strains lacking Zur, Fur, and MntR was assessed. Loss of Zur and Fur, but not MntR, resulted in increased expression of the cnt locus (Fig. 1B). Together, these results show that while the expression of Cnt is responsive to the availability of Fe and Mn, Zn abundance is the dominant factor controlling expression, implying a role in Zn acquisition.

We then evaluated the contribution of the Cnt system to Zn acquisition. To accomplish this goal, we generated a strain lacking CntA (the ΔcntA mutant). We observed that, consistent with prior studies (45, 46), there was no discernible growth defect of the ΔcntA mutant in Zn-depleted medium (Fig. 1C and S1A). Given the potential for overlapping function with the AdcABC system, we assessed the growth of a strain lacking both potential Zn import pathways (the ΔadcA ΔcntA mutant). Growth of the double mutant in medium lacking Zn, Fe, and Mn was severely compromised (Fig. 1C and S1C). Supplementation with either Mn, Fe, Ni, or Co failed to reverse the growth defect of the ΔadcA ΔcntA mutant (Fig. 1C and S1D). Notably, addition of either Mn or Fe further reduced the growth of the ΔadcA ΔcntA mutant. In contrast, supplementation with Zn restored the growth of the ΔadcA ΔcntA mutant to wild-type levels. Ectopic expression of either adcA or cntA also reversed the growth defect of the ΔadcA ΔcntA mutant in Zn-depleted medium (Fig. 1D and E, and Fig. S1E and F). We also observed that in metal-replete medium, loss of AdcA resulted in increased expression of the cnt locus while loss of CntA did not result in increased expression of adcA (Fig. S1G and H). This result suggests that AdcA is the first system used by S. aureus to obtain Zn, with the Cnt system being induced when the bacterium cannot meet the cellular demand for this metal using the Adc system. Collectively, these results show that both AdcA and CntA contribute to the ability of S. aureus to grow in Zn-limited environments.

Whole-cell metal accumulation was then assessed to directly determine the contribution of Adc and Cnt pathways to S. aureus metal uptake. Inductively coupled plasma mass spectrometry (ICP-MS) analysis of wild-type S. aureus and ΔadcA, ΔcntA, and ΔadcA ΔcntA mutants revealed a significant decrease in Zn accumulation by ΔadcA and ΔadcA ΔcntA strains (Fig. 2A). A modest reduction in Mn accumulation was also observed in the double mutant (Fig. 2B). However, this small reduction is unlikely to be contributing to the observed growth phenotype of the double mutant as the addition of Mn further exacerbated the growth defect of the ΔadcA ΔcntA strain (Fig. 1C). No change in the accumulation of Fe, Co, Ni, or Cu was observed with any of the mutant strains (Fig. 2C to F). Collectively, these results directly show that AdcABC and CntABCDF contribute to S. aureus Zn import. Further, CntABCDF, which has not been previously associated with Zn acquisition, does not functionally contribute to the uptake of metal ions other than Zn in S. aureus.

We then sought to delineate the relative contribution of the two systems to resisting CP-imposed Zn starvation. Initially, we assessed how CP influenced transcription of the two Zn uptake systems. In response to CP, both systems were significantly upregulated (Fig. 3A). To assess the relative contributions of each system to Zn acquisition in the presence of CP, we then evaluated the growth of ΔadcA, ΔcntA, and ΔadcA ΔcntA mutants in the presence of CP. Consistent with our earlier observations, the ΔadcA ΔcntA mutant showed greater sensitivity to CP than did wild-type S. aureus at all concentrations tested (Fig. 3B). Loss of CntA diminished the ability of S. aureus to grow relative to the wild type when CP concentrations exceeded 120 µg/ml. In contrast, the ΔadcA mutant was no more sensitive to CP than the wild type (Fig. 3B). Ectopic expression of cntABCDF in ΔcntA and ΔadcA ΔcntA mutants restored wild-type growth in the presence of CP (Fig. 3C). Plasmid-mediated expression of adcA in the ΔadcA ΔcntA mutants permitted growth similar to that of the ΔcntA mutant (Fig. 3D). Additionally, loss of any of the genes associated with the Cnt importer in the methicillin-resistant strain USA300 (JE2) increased the sensitivity of S. aureus to CP (Fig. S2A). To determine if the Cnt system was important for resisting host-imposed Mn and/or Zn starvation, we used CP variants that lack either the Zn site (ΔZn, which binds both Mn and Zn) or the Mn/Zn site (ΔMn/Zn, which binds only Zn) (20). Expression of both the adcA and the cnt loci was induced in the presence of either CP site mutant (Fig. 3E). Both ΔcntA and ΔadcA ΔcntA mutants were sensitive to both the ΔZn and the ΔMn/Zn CP binding site mutants, as both can bind Zn (Fig. 3F), indicating that the increased sensitivity of these strains is due to a reduced ability to compete for Zn. Unexpectedly, we observed that the ΔadcA mutant is more sensitive than wild-type S. aureus to the ΔMn/Zn site mutant. As the ΔadcA mutant is not more sensitive to wild-type CP or ΔZn CP, this result suggests that the Cnt system is a less effective Zn importer when other metals, which could block Zn transport, are freely available, as would occur in the presence of the ΔMn/Zn site mutant, which lacks the ability to restrict Mn availability (Fig. 3F). In summary, these data demonstrate that the Cnt system contributes more to resisting CP-mediated Zn restriction than AdcABC.

Distinct from the direct metal ion recruitment mechanisms of the cluster A-I SBPs, the cluster C SBPs, employed by Opp/NikA ABC permeases, bind metal chelates in a process analogous to siderophore-iron acquisition systems. The cnt locus also encodes CntKLM and CntE, which produce and secrete, respectively, the broad-spectrum metallophore staphylopine (StP). Extracellular StP can then be imported by the CntABCDF transporter (46). Loss of StP, similar to that of CntA (Fig. 1C) (46), does not inhibit the growth of S. aureus in medium that has had the metal content reduced using conventional approaches (Fig. 4A) (46). Our results suggest that the presence of the Adc permease may have obscured a role for StP in Zn uptake. However, the promiscuity of Opp/NikA-type transporters (47–49) raises the possibility that StP may not be the metallophore involved in Zn acquisition. We sought to address this by evaluating the ability of an ΔadcA ΔcntKLM mutant to grow in Zn-depleted medium. Similarly to the ΔadcA ΔcntA strain, this mutant was severely attenuated for growth in Zn-restricted medium and profoundly more sensitive to CP (Fig. 4A and B). The addition of Zn, but not Fe or Mn, reversed the growth defect of the ΔadcA ΔcntKLM mutant (Fig. 4A). The growth defect of the ΔadcA ΔcntKLM mutant was also reversed by ectopic expression of adcA or cntKLM (Fig. S3). Similarly to the ΔcntA mutant, the ΔcntKLM mutant was also more sensitive to CP (Fig. 4B). This sensitivity was abolished by expression of cntKLM from a plasmid in the ΔcntKLM mutant (Fig. S3D). Similar results were also observed with USA300 (JE2) (Fig. S2B). The ΔcntKLM and ΔadcA ΔcntKLM mutants were also more sensitive than wild-type S. aureus to both CP variants (ΔZn CP and ΔMn/Zn CP), indicating that the observed growth defect is attributable to impaired Zn acquisition (Fig. 4C). Collectively, these results show that StP contributes to the ability of S. aureus to overcome host-imposed Zn starvation.

Next, we evaluated if StP is promoting growth in Zn-restricted environments by functioning as a zincophore. If StP is a zincophore, it should function in trans and be dependent on the ABC permease CntABCDF. To evaluate this possibility, supernatants harvested from Zn-starved wild-type S. aureus and ΔcntA and ΔcntKLM mutants were tested for their ability to rescue the growth of various staphylococcal mutants. The supernatants harvested from wild-type S. aureus and the ΔcntA mutant rescued the growth of both ΔcntKLM and ΔadcA ΔcntKLM strains but not strains lacking the SBP CntA (Fig. 4D and S4). These data indicate that CntABCDF and StP function together to promote resistance to CP-mediated Zn starvation. Supernatant harvested from a strain deficient in StP synthesis, the ΔcntKLM mutant, was unable to rescue the growth of any strain tested (Fig. 4D and S4). Collectively, these results demonstrate that StP functions as a zincophore, enhancing the ability of S. aureus to compete with CP for Zn.

To evaluate the contributions of AdcABC, CntABCDF, and StP to staphylococcal infection, a systemic retroorbital infection model was used. For these studies, C57BL/6 mice were infected with either wild-type S. aureus or the ΔadcA, ΔcntA, ΔcntKLM, ΔadcA ΔcntA, or ΔadcA ΔcntKLM mutant (Fig. 5A to D). Mice infected with the double mutants lost significantly less weight than those infected with wild-type S. aureus or the single mutants (Fig. 5A). Consistent with the CP growth assays, mice infected with the ΔadcA strain had bacterial burdens comparable to those infected with the wild type. Compared to the wild type, the ΔcntA mutant showed a significant reduction in bacterial burden in the heart, while the ΔcntKLM mutant had a reduced burden in the liver. Both ΔadcA ΔcntA and ΔadcA ΔcntKLM mutants had reduced bacterial burdens in the liver, heart, and kidneys (Fig. 5B to D). Together, these results indicate that while both Zn import pathways can contribute to the ability of S. aureus to cause disease, the Cnt-StP system is sufficient to facilitate Zn acquisition during infection.

Having established a role for StP in the pathogenesis of S. aureus, we then investigated the distribution of the StP synthesis locus using genome neighborhood network (GNN) analysis. StP is produced by CntK, a histidine racemase; CntL, an enzyme with low similarity to nicotianamine synthase; and CntM, which attaches a pyruvate moiety to the StP precursor. CntM, the only member of a Pfam/InterPro protein family, was used to anchor the analysis (Fig. 6 and Table S1). The CntM InterPro family (IPR016935, 184 nonredundant, nonobsolete, queriable members) contains ~90 unique species. Using this data set, 92% (170/184) of the time immediately adjacent (median gene distance of ±1 open reading frame [ORF]) to CntM was a gene belonging to one of three categories. Fifty percent of the time (85/170), a member of the nicotianamine synthase Pfam (PF03059 ) was immediately adjacent to the CntM homolog. Nine percent (15/170) of the time, a member of the methyltranferase_31 Pfam (PF13847 ) was next to the CntM homolog. The remaining 41% (70/170) of the encoded proteins belonged to no Pfam. BLAST analysis of the latter group revealed that 70% of the sequences shared a high degree of similarity (50% sequence identity or greater) with CntL from S. aureus. Surprisingly, CntK, CntL homologs, and the methyltransferase_31 Pfam family are largely restricted to the Firmicutes. Predicted importer and efflux system were associated with all of the synthesis clusters. While there was variability in the importer associated with the synthesis loci, the putative efflux pump belonged to either the major facilitator superfamily (MFS) or the EamA family of transporters. Notably, the MFS-containing loci were primarily associated with CntABCDF importer homologs and divergent staphylopine synthesis machinery, while the EamA loci were associated with a variety of potential importers but contained conserved core synthesis machinery. Additionally, for at least 20 of the genomes analyzed, additional enzymes were encoded proximal to cntLM that may result in the production of modified StP-like molecules. Collectively, these observations suggest that a diverse collection of StP analogs is present in a variety of Gram-positive and Gram-negative organisms.

All animal experiments were approved by the University of Illinois at Urbana-Champaign Institutional Animal Care and Use Committee (IACUC license number 15059) and performed according to NIH guidelines, the Animal Welfare Act, and U.S. federal law.

For routine overnight cultures, S. aureus strains were inoculated into 5 ml of tryptic soy broth (TSB) in 15-ml conical tubes and grown at 37°C on a roller drum. To preculture bacteria in limited-metal environments, overnight growth was performed in 5 ml of Chelex-treated RPMI medium (NRPMI) supplemented with 1% Casamino Acids, 1 mM MgCl₂, 100 μM CaCl₂, and 1 μM FeCl₂ in 15-ml conical tubes and bacteria were grown at 37°C on a roller drum. S. aureus Newman and derivatives were used for all experiments, unless otherwise noted. The ΔcntA and ΔcntKLM mutants were generated by amplifying the 5′ and 3′ flanking regions of the genes using the indicated primers (see Table S2 in the supplemental material). These fragments were then cloned into pKOR1, and the deletions were generated using allelic replacement, as previously described (71). The adcA::erm mutant was generated by phage transducing the allele from USA300 (JE2) adcA::erm into Newman via Φ85 phage. For complementation constructs, the cntA, adcA, and cntKLM coding sequences were amplified using the indicated primers (Table S2). The cntA and cntKLM coding sequences were cloned into pRMC2, which contains an anhydrotetracycline-inducible promoter (72). The adcA coding sequence was cloned into pOS1 under the control of the lgt promoter. For the fluorescent reporters, the promoters of the cnt operon and adcA were cloned into the yellow fluorescent protein (YFP)-containing vector pAH5 (73). All constructs were verified by sequencing, and all mutants were confirmed to be hemolytic. See Tables S3 and S4 for a full list of the strains and plasmids used in this study, respectively.

For growth assays using NRPMI, following overnight growth in TSB, the cultures were back-diluted 1:50 in 5 ml of TSB for 1 h at 37°C. The cultures were then diluted 1:100 in 96-well round-bottom plates containing 100 μl of NRPMI (12) containing 100 μM CaCl₂ and 1 mM MgCl₂ and combinations of 25 μM ZnSO₄, MnCl₂, FeSO₄, CoCl₂, or NiSO₄. Cultures were incubated at 37°C with shaking at 180 rpm, and growth was assessed by measuring the optical density at 600 nm (OD₆₀₀). For complementation experiments, the cultures were also supplemented with 10 μg/ml of chloramphenicol. Cultures in experiments using pRMC2 were also supplemented with 10 ng/ml of anhydrotetracycline.

CP growth assays were performed as previously described with minor modifications (10, 13, 20). Briefly, overnight cultures were back-diluted 1:50 in 5 ml of TSB for 1 h at 37°C. The cultures were then back-diluted 1:100 in 96-well round-bottom plates containing 100 μl of growth medium, which consisted of 38% TSB and 62% calprotectin buffer (20 mM Tris, pH 7.5, 100 mM NaCl, 3 mM CaCl₂, 10 mM β-mercaptoethanol) and was supplemented with 1 μM MnCl₂ and 1 μM ZnSO₄. Cultures were incubated at 37°C with shaking at 180 rpm, and growth was assessed by measuring the optical density at 600 nm (OD₆₀₀). For complementation experiments, bacteria were back-diluted 1:50 in 5 ml of TSB at 37°C for 2 h and supplemented with 10 μg/ml of chloramphenicol. For the experiments using the USA300 (JE2) ΔcntK, ΔcntL, and ΔcntM mutants, the bacteria were grown overnight in TSB and then washed with phosphate-buffered saline (PBS). For these mutants, the growth medium consisted of 38% defined medium (2.6×) and 62% CP buffer (20 mM Tris, pH 7.5, 100 mM NaCl, 3 mM CaCl₂, 10 mM β-mercaptoethanol) (13). The defined medium (2.6×) consisted of 0.5 g/liter NaCl, 1.0 g/liter NH₄Cl, 2 g/liter KH₂PO₄, 7 g/liter Na₂HPO₄, 0.228 µg/liter biotin, 0.228 mg/liter nicotinic acid, 0.228 mg/liter pyridoxine-HCl, 0.228 mg/liter thiamine-HCl, 0.114 mg/liter riboflavin, 0.684 mg/liter calcium pantothenate, 0.104 g/liter phenylalanine, 0.078 g/liter isoleucine, 0.130 g/liter tyrosine, 0.053 g/liter cysteine, 0.260 g/liter glutamic acid, 0.026 g/liter lysine, 0.182 g/liter methionine, 0.078 g/liter histidine, 0.026 g/liter tryptophan, 0.234 g/liter leucine, 0.234 g/liter aspartic acid, 0.182 g/liter arginine, 0.078 g/liter serine, 0.156 g/liter alanine, 0.078 g/liter threonine, 0.130 g/liter glycine, 0.208 g/liter valine, and 0.026 g/liter proline. This was then supplemented with 1.3% glucose. Prior to bacterial growth, the medium was supplemented with 1 μM MnCl₂, 1 μM ZnSO₄, 1 μM FeSO₄, and 2.3 mM MgSO₄. For supernatant rescue assays, the bacteria were grown in Chelex-treated defined medium. Prior to bacterial growth, the medium was supplemented with 1 μM MnCl₂, 1 μM FeSO₄, and 2.3 mM MgSO₄. Bacterial cultures were grown to late exponential phase, and the supernatants were harvested following centrifugation of the cultures. The supernatant was collected, concentrated approximately 15-fold under reduced pressure using a rotary evaporator, and sterile filtered using a 0.22-μm polyether sulfone (PES) membrane filter. For CP assays using concentrated supernatant, the medium consisted of 19% 2× TSB, 62% CP buffer (20 mM Tris, pH 7.5, 100 mM NaCl, 1 mM CaCl₂, 10 mM β-mercaptoethanol), and 19% supernatant. CP was purified as previously described (10, 12).

The expression of adcA and the cnt locus was assessed using a YFP-promoter fusion (excitation and emission wavelengths of 505 nm and 535 nm, respectively) as previously described (30). For assays using NRPMI, the bacteria were precultured overnight in NRPMI as described above. They were then back-diluted 1:100 in a 96-well plate containing 100 μl of NRPMI supplemented with 100 μM CaCl₂ and 1 mM MgCl₂ and various concentrations of ZnSO₄, MnCl₂, FeSO₄, CoCl₂, and NiSO₄. For assays assessing expression in the presence of CP, bacteria were grown as described above for CP growth assays.

All animal experiments were performed as previously described (8, 10, 12, 13, 30). Nine-week-old female C57BL/6 mice were injected retroorbitally with 100 μl of 1 × 10⁸ CFU/ml of bacteria suspended in sterile PBS. Infection was allowed to proceed for 4 days. On day 4 postinfection, mice were sacrificed and their livers, kidneys, and hearts were harvested. Bacterial burdens were assessed by dilution plating.

Whole-cell metal accumulation of S. aureus strains was performed using the growth parameters described for the CP growth inhibition assay. Bacteria were harvested during log-phase growth (OD₆₀₀ of ~0.1) and then washed by resuspension and centrifugation at 3,700 × g for 10 min, twice with 100 mM EDTA and then twice with sterile double-distilled water (ddH₂O). They were then suspended in 1 ml of sterile water, and a small aliquot was taken to determine the CFU. The bacteria were then centrifuged, and the supernatant was removed. Bacterial pellets were desiccated at 96°C overnight. The dry weight of cells was measured, and the pellets were resuspended in 35% HNO₃ and boiled at 95°C for 1 h prior to removal of debris by centrifugation. Samples were diluted to a final concentration of 3.5% HNO₃ and analyzed by inductively coupled plasma mass spectrometry (ICP-MS) on an Agilent 7500cx ICP-MS (Adelaide Microscopy, University of Adelaide) as described previously (74, 75).

Sequence similarity networks (SSNs) were generated according to the procedure described elsewhere (76, 77). Briefly, the InterPro family IPR016935, of which CntM is a member, was used as input for the EFI-EST web tool (http://efi.igb.illinois.edu/efi-est/ ). The initial SSN was generated with an alignment score threshold corresponding to ~50% sequence identity. To identify the genomic cooccurrence between members of InterPro family IPR016935 and other putative staphylopine biosynthetic genes (cntL and cntK) and transporters, the single SSN cluster was uploaded to the EFI-GNT web tool (http://efi.igb.illinois.edu/efi-gnt/ ) using the default “neighborhood size” (±10 genes) and an “input % co-occurrence lower limit” of 5%. The resulting genome neighborhood network (GNN) quantifies the frequency with which gene products (functionally identified by the Pfam families of the encoded proteins) are encoded proximal to the genes for members of the SSN query cluster. Given the tendency for genes of functionally linked proteins to be proximal to one another in bacterial genomes, the GNN can be used to infer functional relationships between the query and neighboring proteins. Neither CntL nor CntK is a member of Pfam families; therefore, to identify homologs for these sequences, a protein BLAST search was performed separately for CntL and CntK (http://www.uniprot.org/blast/ ; generated using default parameters and BLOSUM62 matrix), and the resulting list was searched against the complete list of genome proximal sequences returned in the IPR016935 GNN. Cytoscape v3.2.0 (78) was used for visualization and analysis of the SSN and GNN.

All statistical analyses were performed using GraphPad Prism version 6. The specific statistical test used is indicated in each figure legend.