|Home | About | Journals | Submit | Contact Us | Français|
Staphylococcus aureus is capable of causing a remarkable spectrum of disease, ranging from mild skin eruptions to life-threatening infections. The survival and pathogenic potential of S. aureus depend partly on its ability to sense and respond to changes in its environment. Spx is a thiol/oxidative stress sensor that interacts with the C-terminal domain of the RNA polymerase RpoA subunit, leading to changes in gene expression that help sustain viability under various conditions. Using genetic and deep-sequencing methods, we show that spx is essential in S. aureus and that a previously reported Δspx strain harbored suppressor mutations that allowed it to grow without spx. One of these mutations is a single missense mutation in rpoB (a P-to-L change at position 519 encoded by rpoB [rpoB-P519L]) that conferred high-level resistance to rifampin. This mutation alone was found to be sufficient to bypass the requirement for spx. The generation of rifampin resistance libraries led to the discovery of an additional rpoB mutation, R484H, which supported strains with the spx disruption. Other rifampin resistance mutations either failed to support the Δspx mutant or were recovered at unexpectedly low frequencies in genetic transduction experiments. The amino acid residues encoded by rpoB-P519L and -R484H map in close spatial proximity and comprise a highly conserved region of RpoB. We also discovered that multicopy expression of either trxA (encoding thioredoxin) or trxB (encoding thioredoxin reductase) supports strains with the deletion of spx. Our results reveal intriguing properties, especially of RNA polymerase, that compensate for the loss of an essential gene that is a key mediator of diverse processes in S. aureus, including redox and thiol homeostasis, antibiotic resistance, growth, and metabolism.
IMPORTANCE The survival and pathogenicity of S. aureus depend on complex genetic programs. An objective for combating this insidious organism entails dissecting genetic regulatory circuits and discovering promising new targets for therapeutic intervention. In this study, we discovered that Spx, an RNA polymerase-interacting stress regulator implicated in many stress responses in S. aureus, including responses to oxidative and cell wall antibiotics, is essential. We describe two mechanisms that suppress the lethality of spx disruption. One mechanism highlights how only certain rifampin resistance-encoding alleles of RpoB confer new properties on RNA polymerase, with important mechanistic implications. We describe additional stress conditions where the loss of spx is deleterious, thereby highlighting Spx as a multifaceted regulator and attractive drug discovery target.
Staphylococcus aureus is a major human pathogen that causes a diversity of disease ranging from relatively minor to invasive and systemic disease with significant morbidity and mortality (1,–7). S. aureus is common both in the community and in hospital environments and has the ability to transiently colonize the skin and mucous membranes of most humans, unnoticeably or with mild clinical features. An estimated 20% of the world's population are persistent carriers. The worldwide distribution of S. aureus coupled with its asymptomatic carriage have precluded any effective eradication strategy (8, 9).
The survival of a versatile pathogen like S. aureus requires that it effectively combat environmental dangers, as well as effectors of the innate and adaptive immune response. Among the important environmental sensors responsive to oxidative and thiol stress is Spx (10, 11). Extensive studies revealed that, in Bacillus subtilis, Spx functions as a redox-sensitive thiol disulfide switch by virtue of an N-terminal C10XXC13 motif (11, 13, 15, 17, 19, 21). The interaction of Spx with RNA polymerase (RNAP) is thought to activate transcription through an incompletely understood mechanism that results in enhanced RNA polymerase promoter recognition and open promoter complex formation (17, 19, 21,–23). Until recently, no evidence suggested that Spx was itself a DNA binding protein; however, Spx-DNA contact by cross-linking at an AGCT element situated near position −44 of an spx-regulated promoter was reported (17, 22). Thus, Spx likely acts as a facilitator/appropriator promoting the engagement of the C-terminal domain of the α subunit (α-CTD) of RNAP with so-called UP elements positioned upstream from its −35 promoter region, thereby affecting sigma factor positioning at canonical promoter elements (23, 25,–28).
The adaptor protein YjbH interacts with Spx and helps direct proteolytic turnover of Spx, together with ClpXP (30,–34). The deletion of yjbH, clpP, or clpX or the addition of stress agents like diamide or cell wall-active antibiotics has been shown to stabilize Spx, which is otherwise maintained at very low levels (35, 36). As in B. subtilis, Spx in S. aureus is regulated posttranslationally by YjbH (17, 35, 38, 39). A recent comprehensive study using chromatin immunoprecipitation (ChIP) analysis in B. subtilis also suggests a role for Spx in the basal expression of certain genes even under nonstressed conditions (17, 27).
The Spx regulon has not been as extensively studied in S. aureus. S. aureus Spx is 80% identical at the protein level and 96% similar to its B. subtilis counterpart, suggesting that important biochemical properties are conserved. Disruption of spx in S. aureus led to pronounced pleiotropic effects, suggesting its role as a more global regulator and pointing to multiple functions besides thiol/oxidative homeostasis (36). Recent work demonstrated roles for Spx in modulating resistance to various cell wall-active antibiotics and in the regulation of protein turnover, including that of certain toxins, through its transcriptional regulation of the adaptor protein TrfA, an ortholog of B. subtilis MecA (42, 76).
S. aureus contains multiple redox buffering systems, such as the thioredoxin and thioredoxin reductase system encoded by trxA and trxB and the redoxin and thioredoxin reductase system encoded by nrdH and trxB, which collectively serve to mitigate the extent of damage to oxidized thiols (43). The genes encoding thioredoxin (trxA) and thioredoxin reductase (trxB) are essential in S. aureus (6, 45,–47). Thioredoxin also serves as an electron donor for aerobic ribonucleotide reductase and, thus, serves an important role in the production of deoxynucleotides (48). S. aureus may also accomplish additional thiol protection by the use of l-cysteine pools or coenzyme A (CoA) and CoA reductase (45, 49, 50) and bacillithiol (15, 51, 53,–57). Phylogenomics suggests that additional thiol oxidoreductases are encoded in genomes possessing the biosynthetic machinery for bacillithiol (54, 56). For example, YtxJ may function as a bacilliredoxin, together with bacillithiol, to combat hypochlorite stress (10). Notably, all S. aureus strains derived from strain NCTC8325, a vast majority of standard laboratory strains, lack bacillithiol because of disruption of the bshC (yllA)-encoded cysteine ligase (61, 63).
Spx is not essential in B. subtilis (13, 15), and yet, despite the strong similarity in sequences, posttranslational regulation by YjbH, and role as mediator of thiol/oxidative stress defenses, we discovered that Spx was essential in S. aureus. In the work presented here, we investigated the mechanism underlying the survival of S. aureus strains lacking spx.
The strains used in this study are listed in Table 1 or in Table S3 in the supplemental material. Escherichia coli strains were grown in Luria-Bertani broth, and Staphylococcus aureus strains were grown in Mueller-Hinton broth or agar (MHB or MHA; Difco) or tryptic soy broth or agar (TSB or TSA; Difco) as indicated below. The growth media were supplemented with ampicillin (100 μg/ml), spectinomycin (100 μg/ml), kanamycin (40 μg/ml), tetracycline (1 to 3 μg/ml), erythromycin (3 to 5 μg/ml), or chloramphenicol (15 μg/ml) when appropriate. Recombinant lysostaphin was obtained from AMBI Products LLC (Lawrence, NY). S. aureus strain NE1281 (strain USA300 JE2 with an ermB insertion in ytxJ) was obtained through the Network of Antimicrobial Resistance in Staphylococcus aureus (NARSA) program, supported under NIAID/NIH contract number HHSN272200700055C. Rifampin (product number R3501; Sigma) was prepared as a 20-mg/ml stock solution in dimethyl sulfoxide (DMSO) and diluted in DMSO as necessary prior to agar plate supplementation.
Details on the construction of all plasmids and genetically engineered strains are supplied in the supplemental material.
Genomic DNA of strain AR738 (Δspx) was prepared from an overnight culture of a strain obtained directly from the D. Frees/H. Ingmer laboratory strain archives, University of Copenhagen, which was grown in MHB at 37°C essentially as described previously (35, 76). The strain was sequenced on an Illumina Hi-Seq 2000 instrument (Fasteris SA, Geneva, Switzerland). Paired-end reads were mapped on the NCTC8325 genome sequence (NCBI accession number NC_007795) using the Burrows-Wheeler Alignment tool, giving a raw coverage depth of approximately 2,400-fold. The mapping covered >98.7% of NCTC8325, and 99.95% when corrected for prophage loss. Variant-base calling relative to the sequence of NCTC8325 was initially performed with SAMTOOLS and the CLC Genomics Workbench and identified 125 single nucleotide polymorphisms (SNPs). SNPs supported by less than 99% of the reads were reexamined by capillary sequencing. Subsequent analysis used filtering of potential SNP/indel differences using our previously sequenced strain 8325-4 genome as the reference (7), which reduced the number of SNPs identified to four. Importantly, the 8325-4 genome used for filtering was the immediate precursor of AR738 and was also derived from strain archives in the Frees laboratory. The complete details of the sequence of this strain and its refinement, correction of NCTC8325 reference genome database errors, and revised annotations of single nucleotide changes and insertions/deletions are thoroughly reported in our previously published work (7). Additional confirmatory sequence verification was performed by sequencing the corresponding regions in the Δspx mutant-derived, spx+-complemented strain spx-c (AR739) (36). Insertions/deletions were examined using the Integrated Genomics Viewer (Broad Institute) using tiled data file (tdf) conversion of AR738 Illumina sequence reads. Seven expected indels were detected, which arose from the loss of prophages ϕ11, ϕ12, and ϕ13, the disruption of spx, the previously noted 63-bp deletion in the sarS-spa intergenic region, SA0282 homolog cluster repeat variation, and the excinuclease ABC β subunit region genome assembly artifact (7). Thus, collectively, apart from the Δspx disruption, we detected no additional indels in the Δspx strain (AR738) compared with the sequence of its isogenic 8325-4 precursor (AR1079). The sequence reads obtained for AR738 were also de novo assembled using the string graph assembler (86) running on the Vital-IT high-performance computing platform (www.vital-it.ch). Gap filling of contigs was accomplished by BLASTN alignment and comparison to NCTC8325 (accession number NC_007795) and our 8325-4 reference sequence (7). The assembled AR738 sequence is 2,687,774 bp.
Genomic DNA was prepared and used as a template for PCR amplification of the region surrounding each suspected nucleotide change. Upstream and downstream primers for SNP verification (see Table S1 in the supplemental material) were designed to create convenient fragment sizes (100 to 600 bp) for subsequent agarose gel purification and sequence analysis using either PCR driver primer. Strain construction used similar methods for all SNP verifications.
The trxA (346-bp) and trxB (967-bp) products were amplified by PCR using forward and reverse primers and strain 8325-4 chromosomal DNA as the template. Forward primers carried a KpnI tail, and reverse primers carried an NsiI-BamHI tail. The PCR products were amplified with Pfu DNA polymerase (Promega) and then purified and cloned using KpnI and BamHI in pBluescript II (KS+). Fragments were then ligated using KpnI and NsiI/PstI enzymes in a custom-designed pMK4-based vector containing the PglyS promoter. Following sequence verification, the entire polylinker cassette was excised with Mfe1 and BglII and ligated with the E. coli/S. aureus shuttle vector pMK4 cleaved with EcoR1-BamH1.
Libraries were generated using the same selection strategy as described above but instead starting with an ermB-marked (MV42) rpoB+ strain. Plates grown for 48 h were flooded with MHB, and colonies scraped and combined into four independent pools, each representing >1,100 independent colonies from 10 85-mm petri plates each. The pools were prepared as recipient cells for bacteriophage-mediated transduction of Δspx::kanA from strain CB1400. Transductions were performed on plates with 40 μg/ml kanamycin. Colonies were restreaked and sequenced. We obtained in this manner three independent isolates carrying rpoB-R484H (an allele with an R-to-H change at position 484 encoded by rpoB [rpoB-R484H]) from three independent library pools. To confirm that mutants with this allele could support the Δspx mutation, cells bearing the allele were backcrossed to fresh 8325-4 cells using the rpoB-linked ermB marker and then retested.
MIC assays were performed with rifampin-supplemented MHA plates. Rif MICs were defined essentially as described previously (24), corresponding to the concentration of rifampin on MHA plates at which a spot inoculation (10 μl) of 1 × 10−6 CFU showed no detectable colony formation at 48 h. Each overnight strain culture was first normalized to a McFarland standard of 2 (corresponding to 6 × 108 CFU/ml) using a bioMérieux Densimat instrument and sterile 0.9% saline. Normalized cultures were then further diluted sixfold to 1 × 108 CFU/ml for reference stock solutions.
The 5′ transcriptional start sites of SAOUHSC_01464 and SAOUHSC_00755 were determined essentially as described previously (90), using the SMARTer rapid amplification of cDNA ends (RACE) kit (Clontech). The start sites for these genes were also confirmed by comparison with transcription start sites determined by the exact mapping of transcript ends (EMOTE) method (92), with minor modifications, and unpublished data of P. Redder (University of Geneva).
CFU counts were performed by spotting aliquots of 10-fold serial dilutions onto agar plates and counting after 24 and 48 h of growth at 37°C unless otherwise indicated (see Fig. 5). Serial 10-fold dilutions of overnight cultures adjusted to a McFarland standard of 1 were made using a Densimat apparatus (bioMérieux) and sterile saline. Aliquots (10 μl) of, typically, 10−2, 10−4, 10−6, and 10−7 dilutions were spotted onto plates. After incubation, colonies were counted and CFU titers computed by dividing the test titer by the reference titer. For efficiency of plating (EOP) computation, the control condition was defined as the number of CFU obtained at 37°C on MHA. For Δspx strains compared to their isogenic precursors, EOP was measured within less than 1 log10 difference, typically less than a fivefold fluctuation in titer. For phenotype analysis, the anaerobic culture conditions used mini-Anaerocult A sealed sacs together with the gas generator system as described by the manufacturer (Merck). All sacs used internally fixed BBL dry anaerobic resazurin indicator strips (Becton Dickinson). Nitrate supplementation for anaerobic respiration used 20 mM (final concentration) sodium nitrate in MHA plates. Diamide [bis(N,N-dimethylamide); Sigma Aldrich] was freshly prepared as an aqueous 50 mM stock solution and filter sterilized through Millipore 0.22-μm cartridges prior to use. The stock solutions were diluted as necessary for MHA plates.
Multiple experiments consistently revealed an inability to cross the Δspx allele into other S. aureus strain backgrounds using standard bacteriophage-mediated general transduction procedures. These results and consideration of the difficulty encountered when obtaining the original Δspx disruption in strain 8325-4 (36) led us to test the hypothesis that spx is essential in S. aureus and that the Δspx strain previously described (36) contains a suppressor mutation(s) that permits its viability in the absence of spx.
We designed a test for essentiality based upon genetic linkage and bacteriophage-mediated transduction (95). For this, we used three different S. aureus genetic backgrounds: 8325-4, ISP794, and Newman (recipient strains). For each strain, we engineered control merodiploid derivatives in which a second copy of spx+ under the control of its native promoter sequence had been inserted into a chromosomal site—specifically, in single copy at the geh (lipase) locus (36, 96). The original spx-c (AR739) construction containing this complementing copy of spx+ had been shown to compensate fully or partially for spx disruption phenotypes (36). The Δspx mutation (deletion of amino acids 27 to 131 [Δ27–131]) was originally constructed by markerless recombination, and so, to facilitate genetic analysis, we constructed a donor strain for transducing bacteriophage propagation where we site specifically placed an erythromycin resistance marker 5 kb downstream from the Δspx allele in the spx+-complemented strain termed spx-c (hereinafter called strain AR739) (35, 36). This strain was named AJ850, and the spx locus is depicted schematically in Fig. 1.
The transducing bacteriophage, propagated on AJ850 (the donor strain), carry genomic DNA with the erythromycin resistance marker, ermC, in proximity to Δspx. Infection of the recipient strains described above was performed with selection for erythromycin resistance. We expected (Fig. 1) to measure the cotransduction frequency of the selected erythromycin marker in close proximity to Δspx by screening for the presence of the cotransduced Δspx deletion by PCR assay.
The results are shown in Table 2. For the merodiploid recipient strains, we obtained a measured cotransduction frequency of 50 to 70% (global average of 62% for n = 62) with the Δspx mutation, whereas in no case did we obtain the Δspx mutation in the haploid strains. Because the Δspx mutation could not be crossed by generalized bacteriophage-mediated transduction into any of the three S. aureus strains tested, whereas it could readily be crossed into the identical strains provided they carried a second unlinked copy of spx, we concluded that spx was essential. This finding is further supported by the failure of global transposon insertion analysis to uncover spx disruption (1, 3, 5, 6).
To identify genetic changes that occurred in the Δspx strain (AR738) compared with its parent strain, 8325-4, we prepared genomic DNA and deep sequenced AR738 using Illumina/Solexa technology. Importantly, the Δspx parent strain 8325-4 (AR1089) was archived in one of our laboratories (D. Frees) and represented the immediate isogenic precursor used for the original production of Δspx. The complete genomic 8325-4 sequence was determined in our laboratories in a previous comprehensive study (7) and greatly facilitated the detection of genetic changes occurring specifically in the Δspx strain.
A summary of the sequence analysis results and refinement is shown in Table 3. The four confirmed point mutations found in AR738 compared with the sequence of its isogenic 8325-4 parent are shown in Table 4. The single nucleotide changes represent three transitions and one transversion. Three mutations occur in coding sequences, resulting in two missense mutations and one silent mutation. The fourth mutation occurs in an intergenic region corresponding to the region between the −10 and −35 motifs of the ytxJ promoter upstream from a putative monothiol bacilliredoxin (10). The positions of the missense and intergenic mutations that were detected are depicted on physical maps of the corresponding gene regions (Fig. 2). We did not further consider the FtsY-N98N silent mutation in our analysis.
The rpoB-P519L mutation within the RNA polymerase β subunit was found by sequence alignment to correspond to P444 in Thermus aquaticus RNA polymerase or to P564 in E. coli RNA polymerase. High-resolution X-ray crystal structures are available for the T. aquaticus and E. coli RNA polymerases (12, 14, 16, 18, 20), and this proline occurs in the highly conserved D region that forms the RNA exit channel and in close proximity to amino acids found to complex with rifampin (16). The equivalent mutations of P564 in E. coli and P519 in S. aureus strain 8325-4 have been shown to confer high-level resistance to rifampin (16, 24). The measured MIC for rifampin of strain AR738 was 512 μg/ml, confirming that this strain possesses a high level of resistance to this drug. Importantly, the construction of the Δspx strain (AR738) in S. aureus at no time involved exposure to rifampin and the precursor strain 8325-4 is rifampin sensitive (MIC of 0.008 μg/ml) and does not have a mutation in rpoB (7).
The missense mutation resulting in P37S in SAOUHSC_01464 occurs in a hypothetical protein whose gene position is the first in a three-gene cluster (29). The intergenic-region point mutation occurs in the SAOUHSC_00755 ytxJ promoter, 97 nucleotides upstream from the putative ytxJ start codon. A single transcription start site for ytxJ was determined, and the SNP position places it within the boundaries of the −10 and −35 canonical housekeeping promoter elements (Fig. 2). The open reading frame for ytxJ does not appear to be an operon in S. aureus: 154 nucleotides (nt) separate it from its upstream neighbor and 424 nt from its downstream neighbor.
In order to determine which of the detected mutations, or possibly their combinations, was suppressing the lethality of the Δspx mutation in AR738, we first de novo reengineered each point mutation separately in the chromosome of S. aureus parental strain 8325-4 (see the materials and methods in the supplemental material). As described above, we also engineered merodiploid derivatives of each single mutant strain harboring a second unlinked copy of spx+ at the geh locus.
Using the same strategy outlined above, we conducted bacteriophage-mediated transduction linkage experiments by first selecting for erythromycin resistance in both the haploid and merodiploid strains and then screening colonies for spx+ or Δspx alleles by the position of the crossover point of homologous recombination. The results are shown in Table 5.
Our results showed that strain AJ1009, harboring rpoB-P519L by itself, could support spx disruption just as well as AJ1010, an independent rpoB-P519L strain prepared at the same time as AJ1009. In contrast, strains CB1188 and CB1349, containing the other two point mutations in either SAOUHSC_01464 or the promoter region of SAOUHSC_00755, could not support spx disruption. In control experiments, the Δspx allele was readily transferred to each of the respective merodiploid strains, consistent with the expected cotransduction frequency. Notably, the Δspx allele could not be transferred to AJ1008, the otherwise isogenic parent of AJ1009 and AJ1010; however, the spx disruption could be introduced into the spx merodiploid AJ1008-derived strain AJ1011, as expected.
Although the strategy of cotransduction in principle selects only for the applied drug resistance and subsequently screens for the acquisition of the Δspx allele, we were nevertheless concerned that an unknown mutation(s) besides rpoB-P519L may have arisen during the construction of AJ1009 which could conceivably account for the observed suppression of Δspx allele lethality. To address this possibility, we allele specifically reverted strain AJ1009 (rpoB-P519L) to rpoB+ using bacteriophage-mediated transduction and an exchange of a kanamycin resistance marker for the erythromycin resistance marker placed downstream from the rpoBC locus. The resulting revertant strain was called MV54. We screened for loss of rifampin resistance in strain MV54 and verified reversion to rpoB+ by sequence analysis. Multiple independent experiments using phage transduction of the Δspx::kanA allele from donor strain CB1400 failed to generate viable colonies in MV54. We conclude that strain AJ1009 did not harbor an additional mutation(s) besides rpoB-P519L that was sufficient by itself to support the deletion of spx.
Our results described above predicted that mutations detected by our deep-sequencing analysis of strain AR738 could be eliminated, apart from the P519L mutation, and still yield a viable strain. We used a combination of allelic exchange for SAOUHSC_01464 and transposon disruption, in the case of ytxJ, to create Δspx strains derived from AR738 that harbored reversions of SAOUHSC_01464 with the P37S mutation to the wild type, insertional disruption of ytxJ (ytxJ::ermB), or both. We were able to produce strains with each of the single mutant variants as viable derivatives of AR738 (CB1401 and MV36), as well as the double mutant derivative strain MV38 that contains only the original rpoB-P519L (see Fig. S1 in the supplemental material). Collectively, our results from two complementary genetic methods, mutant reconstruction and sequential SNP elimination/gene disruption, demonstrated that the rpoB-P519L mutation detected in AR738 was, by itself, sufficient to support the otherwise lethal deletion of spx.
We next asked whether the suppression of Δspx was allele specific for rpoB-P519L. As a first step, we selected several random spontaneous rifampin-resistant mutants derived from AJ1008 and sequenced the rpoB region. Three mutants, bearing S486L, D471Y, and A477V mutations, that were found to have a varied rifampin MIC range (4 to 1,024 μg/ml) and whose mutations were spatially located in various regions of the rifampin binding pocket (Fig. 3) were retained for further study. The results (Table 5) revealed that WLK8-4R rpoB-S486L consistently failed to support the acquisition of Δspx, whereas its merodiploid derivative strain MV82 could support the acquisition of Δspx. Multiple experiments with WLK8-1R rpoB-A477V revealed only 5 instances where we obtained stable viable strains harboring Δspx out of 43 (11%) independent erythromycin-resistant transductants tested, considerably below the expected (50 to 60%) experimental cotransduction frequency. In control experiments, the merodiploid WLK8-1R rpoB-A477V-derived strain MV80 readily accepted Δspx. A similar observation was made for WLK8-3R rpoB-D471Y, in that a low number of colonies (4/24, 17%), well below the expected cotransduction frequency, were viable and contained Δspx. We conclude that rifampin resistance per se is not sufficient to support the loss of spx and that certain alleles are required. The rpoB-P519L allele consistently supported spx disruption with the expected cotransduction frequency, whereas rpoB-A477V and rpoB-D471Y apparently support disruption of spx, but at a much lower frequency than expected.
To extend these findings and discover whether additional rifampin resistance alleles could be uncovered which support the deletion of spx, we next performed a genetic selection experiment using CB1400 Δspx::kanA as a donor strain (depicted in Fig. 1) for bacteriophage-mediated transduction.
Multiple independent experiments using recipient cells without prior selection for rifampin resistance consistently failed to give viable colonies following transduction with phage lysates from a Δspx::kanA donor strain, whereas Δspx::kanA could be crossed, as expected, into an spx+ merodiploid strain, CB1348. The failure to obtain viable colonies by direct Δspx::kanA transduction in the absence of prior rifampin exposure suggested that the frequency of spontaneous rifampin-resistant alleles in the unselected population was too low to detect by direct selection methods under our experimental conditions. This finding may also be explained by the fitness cost of rifampin mutation in the absence of selection within a growing population of cells used as the recipient strain in transduction experiments. This finding is also in complete agreement with the original difficulty encountered when constructing the original Δspx strain (36).
In contrast, when we created a library of >4,500 independent rifampin-resistant mutants selected in strain MV42 (rpoB+ with ermB nearby), we readily obtained kanamycin-resistant colonies that remained viable upon subcultivation. Sequence analysis of three independent Δspx::kanA isolates chosen at random from three independent MV42-derived rifampin-resistant-mutant pools contained the same rifampin resistance mutation, rpoB-R484H. We did not sequence all viable colonies obtained, leaving open the possibility that additional suppressor alleles await discovery. Rifampin resistance studied in other organisms has shown that rifampin mutations are likely biased for only a handful of alleles (16), and thus, defining the complete landscape of rifampin resistance mutations that support the viability of Δspx strains was not further pursued in this analysis.
Figure 4 shows an expanded alignment of the D region of RpoB, including the locations of mutations conferring rifampin resistance in various organisms and of rifampin resistance mutations tested in this study. Strikingly, the two mutated amino acids P519L and R484H are located in close spatial proximity (Fig. 3) within a region that has been noted as highly conserved from bacteria to humans (16). R484 makes direct contact with O1 of the naphthol ring of rifampin, and P519 is positioned underneath R484. Of note, an equivalent of the P519L mutation has been described in E. coli (P564L, also historically called ack-1 ), and this amino acid is in close proximity to the secondary channel of the polymerase where nucleoside triphosphates (NTPs) enter. Amino acid S486, located close to R484 (Fig. 3), also makes contact with rifampin naphthol ring O2. Notably, despite the spatial proximity of R484 and S486 and their presumably similar modes of disruption of rifampin binding, mutation of these amino acids results in opposite phenotypes with respect to their support of strains with an spx disruption.
S. aureus strains containing the Δspx disruption are consistently associated with a slow-growth phenotype. We considered the possibility that Spx-dependent gene transcription, such as is known for the essential thioredoxin reductase gene, trxB (36), explains the lethality of spx deletion. However, little is known concerning the regulation of the thioredoxin gene, trxA, in S. aureus. Using reverse transcription-quantitative PCR (qRT-PCR) analysis, we discovered that trxA is also transcriptionally regulated by spx (see Fig. S2 in the supplemental material). Thus, both trxA and trxB are in the spx regulon. In the absence of stress, S. aureus has evolved to survive in the presence of basal levels of Spx, which are kept at extremely low levels by proteolytic degradation orchestrated by YjbH and ClpXP (35, 38). Nevertheless, these basal Spx levels are most likely responsible, in part, for the set points of certain critical promoters, especially those involved in controlling the cell redox state and contributing to the biosynthesis of deoxynucleoside triphosphates (dNTPS) via ribonucleotide reductase. We hypothesized that disruption of spx limited thioredoxin/thioredoxin reductase to such an extent that it compromised cell viability.
To address this hypothesis, we cloned the unlinked genes encoding thioredoxin (trxA) or thioredoxin reductase (trxB) on multicopy plasmids and placed the genes under the control of the T box riboswitch of the glycyl-tRNA synthetase gene promoter PglyS (SAOUHSC_01666) (44, 46). This promoter is not known to be regulated by Spx and would be expected to exert housekeeping levels of gene expression, since S. aureus encodes only one glycyl-tRNA synthetase. As a control, we also cloned emission-enhanced green fluorescent protein (GFPuv4) under the control of the same PglyS promoter. The plasmids were introduced into 8325-4, and the resulting strains (MV120, MV121, and MV122, respectively) used as recipients for transduction of the Δspx::kanA allele.
We repeatedly recovered viable Δspx::kanA strains from strains containing trxA or trxB plasmids but not from the control strain, MV122, expressing GFPuv4 from the same promoter. The Δspx::kanA derivative strains MV126 and MV128 were tested and found to be phenotypically rifampin sensitive (MICs of <0.008 μg/ml) and rpoB+ by direct sequence analysis. We conclude from this analysis that at least one more solution exists for the suppression of spx disruption lethality that is independent of rifampin resistance and a particular rifampin resistance allele. This second mechanism requires heterologous promoter-dependent expression of either thioredoxin or thioredoxin reductase.
Using serial dilutions of spx+ and Δspx strains, we determined CFU counts on agar plates and determined the Δspx-to-spx+ strain efficiency-of-plating (EOP) ratios (see Table S2 in the supplemental material). The results revealed that the suppressor strains (rpoB or trx) and their Δspx derivatives had comparable plating efficiencies (the CFU titer ratios [Δspx-to-spx+ strain ratios] ranged from 0.3 to 0.68). We have also noted that despite slow growth of Δspx strains, the endpoint values for stationary-phase optical density at 600 nm (OD600) of spx+ and Δspx strains were also comparable. Thus, we conclude that the acquisition of the Δspx mutation by the suppressor strains does not engender significant alterations in overall strain viability.
Previous work (36) attributed several diverse phenotypes to the Δspx strain AR738. In light of our findings that this particular strain contained additional mutations in addition to the rpoB-P519L suppressor allele, we wished to examine phenotypes associated with the Δspx mutation in the context of either its rpoB or trxA/-B suppressor to determine whether they were consistent irrespective of the suppressor mechanism. The results are shown in Fig. 5.
We observed that Δspx strains grew aerobically or anaerobically (both respiration and fermentation) but displayed hypersensitivity to the thiol-specific oxidant diamide and to hypersaline conditions (2.2 M NaCl) compared with their spx+ control suppressor strains. Furthermore, we found that Δspx strains have a strong cold sensitivity phenotype. Colonies with the Δspx mutation grew to pinpoint size and displayed an initial efficiency of plating compared to that of the 37°C control of approximately 0.5 to 1; however, these colonies did not continue to grow, even after 10 days of incubation, nor did cells resume growth following replacement at 37°C. This result suggests that the growth of Δspx strains at a cold temperature results in a time-dependent depletion of an essential component(s), culminating in cell death.
The strong phenotypes we have documented for the Δspx strains (Fig. 5) must be considered with the caveat that the successful disruption of spx occurs in the context of a suppressor mechanism that may itself contribute to the phenotypes tested. Notably, as shown by the data presented in Fig. 5, control strains harboring only a suppressor mutation and otherwise wild type for spx did not show these strong phenotypes. Thus, since we observed concordant strong phenotypes arising from spx disruption with both of the suppressor mechanisms we describe herein (rpoB based and trx based), it is likely that the observed effects arise primarily from the loss of spx, with negligible contribution from the suppressor mutation itself. A comprehensive examination of spx-dependent phenotypes and the spx-dependent gene expression that underlies precisely why spx is essential will be the subject of future study.
We have shown that the RNA polymerase-interacting regulator Spx is essential in S. aureus and that the viability of the strain (AR738) in which Δspx was first described (36) could be traced to a single mutation in rpoB, the gene encoding the β subunit of RNA polymerase. Additional analysis showed that this mechanism of spx suppression required certain specific changes in RpoB, most notably P519L or R484H, conferring high-level resistance to rifampin. Another high-level rifampin resistance mutation, RpoB-S486L, failed to suppress the Δspx disruption, and so rifampin resistance mutations in rpoB per se do not underlie the suppression mechanism, which points instead to genetic suppression that is dependent upon specific alleles.
We repeatedly failed to uncover rpoB mutations that could support the viability of a strain with the Δspx mutation without prior selection for rifampin resistance. This suggests that the frequency of rifampin resistance mutations or alternative mutations that could conceivably suppress the deleterious effects of the Δspx mutations and which map outside rpoB and are or are not related to mutations conferring rifampin resistance occur infrequently. With regard to rifampin resistance, this finding is consistent with the fitness cost of rifampin resistance mutations in S. aureus strain 8325-4, where almost all strains with such mutation show impaired growth compared to that of wild-type controls (24, 52).
A second mechanism of suppression was uncovered in our study, which showed that multicopy plasmids harboring a heterologous housekeeping promoter, a PglyS T box riboswitch driving trxA or trxB, could support the viability of strains with a disruption of spx. This result suggests that adequate minimal expression of thioredoxin and thioredoxin reductase is required to create an environment wherein spx can be dispensed with. Both trxA and trxB are spx regulated in S. aureus, as described herein and elsewhere (36, 45), and trxA and trxB are also spx regulated in B. subtilis (15). While we report these two mechanisms in this study, we cannot exclude the possibility that there are additional means to suppress Δspx lethality that await discovery.
A survey of rifampin resistance mutations in other organisms that are synonymous to those we describe herein resulted in the remarkable finding that mutations of both P564 and R529 (equivalent to S. aureus P519 and R484) in Salmonella enterica are very strongly associated with an enhanced ability to survive in aging cultures compared with the ability of wild-type rpoB+ rifampin-sensitive cells to do so (58). Intriguingly, cells with another mutation, S531F (equivalent to S. aureus S486, for which we show no ability to suppress Δspx disruption), showed no enhanced survival advantage. The authors made these findings in the course of an effort to explain an apparently enhanced mutation frequency in aging cultures of E. coli or Salmonella enterica that was ultimately traced to the selective growth advantage of these mutations on acetate rather than altered mutation rates, using rifampin resistance mutation as a proxy (58, 60). P564L in E. coli has also been described as an rpoB allele that helps bypass starvation-induced replication arrest in E. coli cells lacking dksA (62).
Additional consideration of rifampin resistance alleles, especially those equivalent to S. aureus RpoB-P519L, uncovered yet further evidence of linkage to a wide variety of changes in RNAP transcription properties and their physiological consequences. These alterations include transcription slippage, sign epistasis, Rho-dependent terminator read-through, attenuation bypass, global transcriptome changes, and alterations in metabolic profiles (64,–75). It appears that common solutions to confront diverse stress conditions have occurred in various organisms, hinting at a potentially fundamental process.
Of particular note, the rifampin resistance allele rpoB-H481Y in S. aureus is associated with extensive changes in the transcriptome profile compared with that of its isogenic parent strain, and this partially helps to explain the changes in virulence, persistence, and capsule formation in a clinical methicillin-resistant S. aureus (MRSA) strain (59). Further study showed that the rpoB-A477D allele modulates reduced susceptibility to daptomycin, vancomycin, and β-lactam antibiotics and contributes to a variety of pleiotropic effects, including modulation of Spx levels (61). Spx has been linked with both glycopeptide and β-lactam resistance in S. aureus in our own laboratory by virtue of the implication in drug resistance of its negative regulator yjbH (35) and an adaptor gene, trfA, that was shown to be Spx regulated (76). Glycopeptide resistance is multifactorial and often involves mutation of global regulators, such as the two-component systems VraRST, GraRS, and WalKR or the serine threonine kinase/phosphatase Stk/Stp. In this regard, it is noteworthy that rpoB mutations which do not always confer rifampin resistance, are now frequently associated with the intermediate vancomycin resistance phenotype in MRSA strains (77,–82).
Collectively, the evidence suggests that certain rifampin alleles often confer new properties upon RNAP that affect gene expression and physiology, probably by several mechanisms. Pinpointing the precise mechanism that dictates how particular rpoB mutations and not others can suppress the otherwise lethal disruption of spx in S. aureus may therefore prove challenging.
In the absence of Spx in S. aureus, it is reasonable to consider that the suppression mechanism must account for the loss of Spx-dependent transcription regulation, and by extension, postulate a role for compensatory transcription modulation. Regulators like Spx that act through the UP element are thought to direct their promoter specificity by communicating with the sigma factor and core promoter elements via the α-CTD (19, 21, 22, 26, 83, 84). Indeed, a detailed study of Spx-dependent promoter selection in B. subtilis (22) revealed that Spx binds to an upstream sequence in the trxB promoter and helps SigA engage canonical promoter elements that otherwise are poorly recognized (22).
Structural studies of transcription initiation have revealed new insights into promoter recognition and the steps leading to open promoter formation (18, 23, 87, 89). Promoter recognition is thought to be mediated by sigma factor-DNA interaction and core promoter elements comprising −35, −10, and extended −10 regions (12, 23); however, a high-resolution study of an open promoter complex with Thermus thermophilus holoenzyme recently revealed the presence of a new determinant called the core recognition element (CRE) that is thought to provide sequence-specific promoter recognition of the −4 to +2 region of the nontemplate DNA strand (89). Additional studies in E. coli have extended these findings and shown that the CRE can modulate transcription start site selection, elongation, and pausing and that mutation of the CRE impairs transcription (88, 97).
The binding specificity originally attributed to the CRE is thought to be mediated via contacts provided by amino acids within RpoB, particularly those comprising van der Waals contacts that define a pocket for interaction with +2G in the RNA template used for cocrystal formation (89). It awaits further experimentation to ascertain whether determinants throughout the −4 to +2 region are truly sequence specific.
The CRE motif is composed of two discontinuous polypeptide segments of the RpoB β subunit (positions 440 to 455 and 535 to 555 in E. coli coordinates) (88, 89, 91). This motif therefore flanks domain I of region D, comprising the rifampin pocket fold (16). In light of our finding that specific rifampin alleles in a highly conserved region of RpoB are able to compensate for the loss of an essential gene, itself an RNAP-interacting protein mediating promoter selection (22, 27), it is tempting to speculate that the proximity of certain rifampin resistance mutations to the CRE could perturb the local polypeptide chain configurations sufficiently to alter promoter recognition determinants and/or other parts of the transcription initiation process. Experiments are underway to address this hypothesis.
The second mechanism of Δspx suppression we describe in this study was uncovered by considering the possibility that bypass by acquisition of particular rifampin resistance alleles in some way mimics what Spx itself accomplishes when contacting the RNAP α-CTD. In this respect, an spx-dependent gene(s), most likely essential itself, displayed insufficient transcription in the absence of spx to support viability. Since both trxA and trxB are transcriptionally regulated by Spx, indeed the most reasonable scenario explaining the suppression mechanism is that heterologous promoter control of trxA or of trxB compensates for insufficient transcription in the absence of Spx. Although it is conceptually clear that multicopy expression using a heterologous promoter could create conditions whereby spx deletion could be introduced into cells without rpoB mutation by augmenting the gene dosage of certain spx-regulated genes, we cannot at this time definitively say that the two mechanisms are related simply by compensatory transcription levels. For example, S. aureus trxB is known to be regulated both negatively, by SarA and PerR (45, 93, 94), and positively, by Spx (36), and so one possible alternative scenario would be that a particular rifampin resistance mutation perturbs the expression of SarA or PerR such that it relieves negative regulation, thus leading to adequate trxB expression.
The full extent of spx-regulated genes in S. aureus is incompletely understood, although proteomics has shown differential regulation of dozens of proteins (36) and the results of a global ChIP assay in B. subtilis indicate that perhaps several hundred promoter regions bind Spx (27). Furthermore, it cannot be formally excluded that Spx interacts with RNAP at sites outside the α-CTD (22, 98). Future experiments will investigate this.
Our examination of phenotypes arising with spx disruption underscores a critical role for Spx in several cellular processes, particularly low-temperature survival and osmotolerance. Recent work has highlighted an important role for potassium uptake systems in counteracting hypersalinity, mediated by the kdp and ktr systems, together with the finding that cyclic di-AMP is an apparent negative regulator of potassium uptake by virtue of its interaction with KdpD (99,–101). These findings suggest important contributions of Spx and Spx-regulated genes in survival in diverse microenvironments. Furthermore, rifampin resistance can have a profound influence on gene expression in both Gram-negative and Gram-positive organisms, including a significant impact upon antibiotic resistance mechanisms. This suggests a particularly fruitful avenue of research to discover how the acquisition of rifampin resistance globally affects cellular physiology.
We thank Christine Barras and Antoinette Monod for excellent technical assistance, the bioinformatics staff of Fasteris SA (Geneva) for the attention paid to our deep sequencing and bioinformatics needs, and Peter Redder (University of Geneva) for helpful discussion and sharing of unpublished data.
This work was supported by a grant from the Swiss National Science Foundation (310030-146540 to W.L.K.), an M.D.-Ph.D. thesis doctoral training grant from the Swiss Academy of Medical Sciences via the F. Hoffmann-La Roche Research Foundation (D.O.A.), the Danish Council for Independent Research (12-132527 to K.T.B.), the Nordic Joint Committee for Agricultural Research (D.F. and H.I.), and the Canton of Geneva.
Supplemental material for this article may be found at http://dx.doi.org/10.1128/JB.00261-16.