| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
1 Department of Microbiology and Immunology, University of Arkansas for Medical Sciences, Little Rock, AR 72205, USA
2 Department of Pathology and Microbiology, University of Nebraska Medical Center, Omaha, NE 68198, USA
3 Wyeth Vaccines, Pearl River, NY 10965, USA
4 Wyeth Protein Technologies, Cambridge, MA 02140, USA
Correspondence
Mark S. Smeltzer
smeltzermarks{at}uams.edu
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMETHODSRESULTS AND DISCUSSIONREFERENCES |
|---|
The GEO database accession number for the genome-wide study data determined in this work is GSE5466.
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTS AND DISCUSSIONREFERENCES |
|---|
; Cheung & Zhang, 2002; Novick, 2003; Bronner et al., 2004). Understanding these regulatory circuits could facilitate the development of therapeutic agents capable of limiting the ability of S. aureus to cause disease. However, current regulatory models are based on studies done with a limited number of strains, most of which are derivatives of NCTC 8325. The most widely studied lineage is 8325-4, which was generated by curing three prophages from the NCTC 8325 genome (Novick, 1967), and the single most widely studied strain within this lineage is RN6390.
One reason for the focus on NCTC 8325 strains is that they are amenable to genetic manipulation. However, recent data have suggested that regulatory models based on these strains are not representative of the situation observed in clinical isolates. For example, we have confirmed that RN6390 differs from clinical isolates with respect to several clinically relevant phenotypes, including biofilm formation, relative capacity to bind host proteins, and production of exotoxins (Blevins et al., 2002; Beenken et al., 2003). Moreover, all NCTC 8325 derivatives have an 11 bp deletion in rsbU (Kullik et al., 1998). Because rsbU activates sigB expression, these strains are functionally SigB deficient. SigB is the primary stress-response sigma factor of S. aureus, and it has a global impact on expression of multiple genes, including many that contribute to the ability to cause disease (Kullik et al., 1998; Bischoff et al., 2004; Ziebandt et al., 2004). A recent report has also found that all NCTC 8325 derivatives, including RN6390, have a mutation in tcaR, a regulatory locus that plays an important role in biofilm formation (Jefferson et al., 2004), and expression of virulence factors, including sarS and spa (McCallum et al., 2004).
We recently carried out a DNA microarray analysis comparing the genomes of two highly virulent clinical isolates (UAMS-1 and UAMS-601) with RN6390 and seven sequenced strains of S. aureus (SANGER-252, SANGER-476, MW2, COL, NCTC 8325, Mu50 and N315). The results of this comparison confirmed that the UAMS isolates are closely related to each other and to EMRSA-16 (SANGER-252) (Cassat et al., 2005), which is a prominent clinical isolate found in diverse geographical areas worldwide (Aires de Sousa et al., 2005; Johnson et al., 2005; Nimmo et al., 2006; Udo et al., 2006). Our analysis also confirmed that the cluster containing UAMS-1, UAMS-601 and EMRSA-16 is the most distantly related, among the strains we examined, to the prototype laboratory strains NCTC 8325 and RN6390 (Cassat et al., 2005).
Despite recognized differences between clinical isolates and RN6390 (Blevins et al., 2002; Beenken et al., 2003), there is currently no comprehensive picture of gene expression patterns in clinical isolates of S. aureus. To address this, our first objective in this study was to carry out genome-scale transcriptional profiling, comparing the S. aureus clinical isolate UAMS-1 with the prototype laboratory strain RN6390. Furthermore, the only transcriptional profiling studies defining the agr and sarA regulons to date have been done in the 8325 strain RN27 (Dunman et al., 2001), which has both rsbU and tcaR mutations. Based on this, our second objective was to perform transcriptional profiling of UAMS-1 sarA and agr mutants. This is important, because we have previously demonstrated that mutation of sarA or agr in UAMS-1 results in a phenotype that is different from that observed in the corresponding RN6390 mutants with respect to several clinically relevant phenotypes, including biofilm formation (Blevins et al., 2002; Beenken et al., 2003). Additionally, we have previously characterized the biofilm regulon of UAMS-1 (Beenken et al., 2004), and defining the sarA and agr regulons in the same strain allowed us to draw parallels between those genes that are differentially regulated during biofilm growth, and those genes that are differentially regulated by sarA or agr.
|
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTS AND DISCUSSIONREFERENCES |
|---|
. Clinical isolates with the designation ‘UAMS’ refer to primary isolates from patients at the University of Arkansas for Medical Sciences, or the Arkansas Children's Hospital. These include unrelated isolates from osteo-articular infections, skin and soft tissue infections, and septic shock. Two of these, UAMS-1140 and UAMS-1141, are pvl-positive isolates from community-acquired infections.
|
). Complementation of the alsSD mutation in KB1097 was achieved by introducing a plasmid containing the alsSD ORFs and the promoter region of the alsSD operon. Because it was not possible to amplify the entire region containing both the alsSD promoter and the ORFs with a proof-reading polymerase, the complementation plasmid pALS-SD was constructed in a two-step process. First, a 668 bp DNA fragment containing the promoter region upstream of the alsS gene, and the 5' portion of the alsS ORF, was PCR-amplified from NCTC 8325 genomic DNA using primers als-Eco and als-Bam (Yang et al., 2006), and then ligated into the EcoRI and BamHI sites of the plasmid pSK265 (Ranelli et al., 1985) to generate pSJ17. Next, a DNA fragment containing the alsSD ORFs was PCR-amplified from NCTC 8325 genomic DNA using primers alsS-F-HindIII (5'-CCCAAGCTTGGAAATGAATATAAATGACTGAT-3') and alsD-R-XbaI (5'-CCCTCTAGACTTCTCGTAGTAACAGATTG-3'). The resulting fragment was ligated into the HindIII and XbaI sites of pRB474 (Bruckner, 1992). This plasmid, designated pSJ19, was then digested with BamHI and XbaI to liberate a 2.4 kb fragment containing alsSD ORFs. This was subsequently ligated into the BamHI and XbaI sites of pSJ17. This plasmid, pALS-SD, was then electroporated into KB1097 to create KB1098.
To generate a sigB mutant, the sigB mutation in GP266 (Bischoff et al., 2001) was transduced into UAMS-1 by phi11-mediated transduction, as previously described (Blevins et al., 2002). Successful transduction was verified by PCR analysis and DNA sequencing (data not shown), and by demonstrating reduced expression of the sigB-dependent gene aps23 (Fig. 1).
|
For RNA isolation, cultures were grown overnight in TSB, with antibiotic selection where appropriate, diluted to an OD560 of 0.05 in TSB without antibiotic selection, and grown to either the exponential (OD560 1.0) or the post-exponential growth phase (OD560 3.0). OD560 readings were obtained with a Spectronic Instruments Genesys 5 spectrophotometer, with cuvettes of 1 cm path length. All cultures for RNA isolation were grown with a flask : volume ratio of 5 : 1, with constant aeration. Biofilm formation was assessed using the microtitre plate method, as previously described (Beenken et al., 2003). A590 readings were taken after sixfold dilution of the primary eluate.
RNA isolation and cDNA labelling.
For microarray analysis, total bacterial RNA was isolated, processed and labelled, as described by Beenken et al. (2004). Prior to labelling, the absence of contaminating DNA was confirmed using PCR with primers corresponding to the sarA gene (Table 2) [the sarA mutation in UAMS-929 was created by insertion of a kanamycin-resistance cassette (Blevins et al., 2002) in a fashion that did not preclude PCR amplification with the sarA primers listed in Table 2]. For quantitative real-time PCR (qRT-PCR) analysis, RNA was isolated using a modification of the method reported in Beenken et al. (2004), which was optimized to increase the yield from clinical isolates. Briefly, approximately 5x109 cells were harvested from cultures in various growth phases, and resuspended in 500 µl Qiagen RNeasy kit buffer RLT. Resuspended cells were then transferred to Q-Biogene FastPrep Lysing Matrix B tubes. Cells were disrupted in the FastPrep FP120 Cell Disruptor for 20 s at setting 5.0, placed on ice for 5 min, and then disrupted again for 30 s at setting 4.5. Disrupted cells were then centrifuged at maximum speed (13 000 g) for 15 min at 4 °C. The aqueous phase was transferred to a fresh 1.5 ml microcentrifuge tube, and 350 µl buffer RLT was added per 100 µl sample. After centrifugation for 15 s at 8000 g, the supernatant was transferred to a fresh tube, and 250 µl 100 % ethanol was added per 100 µl sample. Samples were then applied to a Qiagen RNeasy mini column, and processed according to the manufacturer's instructions. All RNA samples were analysed by A260/A280 spectrophotometry (Bio-Rad SmartSpec 3000) and gel electrophoresis to assess concentration and integrity. DNase treatment was accomplished and verified as previously described (Beenken et al., 2004).
|
). This GeneChip includes 7723 qualifiers representing the consensus ORF sequences identified in the genomes of the S. aureus strains N315, Mu50, COL, NCTC 8325, EMRSA-16 (strain 252) and MSSA-476, as well as novel GenBank entries, and N315 intergenic regions greater than 50 bp. Hybridizations were performed with 1.5 µg of each labelled cDNA. To ensure reproducibility, hybridizations were performed with two RNA samples isolated from each strain at each growth phase. Samples were collected from each of two separate experiments, and each RNA sample was hybridized to two separate GeneChips. Data from duplicate experiments were normalized, and analysed with GeneSpring version 6.2 (Silicon Genetics). Replicate experiments that were not statistically similar as defined by Student's t test were excluded. Differential expression was defined as a change of more than threefold in transcript levels versus the comparator strain. Differential expression was also defined statistically using Student's t test (P
0.05). For comparison of UAMS-1 and RN6390 growth-phase-dependent transcriptional profiles, we omitted results for genes that were not present in both strains, as defined by our previous comparative genomic hybridization studies utilizing the same Affymetrix GeneChip (Cassat et al., 2005). This included genes that were represented on the array only by defined polymorphic alleles when the allelic forms differed between UAMS-1 and RN6390. We also excluded genes in which Affymetrix algorithms determined that the signal was either below a minimum threshold (signal intensity of 10) or was saturated (signal intensity 
900) in samples from both strains.
qRT-PCR (TaqMan) analysis.
qRT-PCR was performed using the iCycler iQ real-time PCR detection system (Bio-Rad). Briefly, 1 µg DNase-treated RNA was converted to cDNA using the iScript cDNA synthesis kit (Bio-Rad). A master mix was prepared for each reaction using iQ Supermix (Bio-Rad), gene-specific primers, and gene-specific Taqman probes (Table 2
). For each target, a standard curve was created using buffer containing known concentrations of genomic DNA. The negative control in all cases was a reaction mix containing all reagents except template DNA. Each reaction was run in duplicate. Results were normalized based on the corresponding results obtained with gyrB-specific primers and a corresponding Taqman probe (Table 2).
|
RESULTS AND DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTS AND DISCUSSIONREFERENCES |
|---|
; Smeltzer et al., 1997; Blevins et al., 2003), and because previous studies have indicated that UAMS-1 is genotypically very similar to other prominent clinical isolates, including EMRSA-16 (Cassat et al., 2005). UAMS-1 and RN6390 have also been shown to differ with respect to important clinically relevant phenotypes, such as biofilm formation (Blevins et al., 2002; Beenken et al., 2003; Cassat et al., 2005). However, it has been difficult to link these observed discrepant phenotypes to differential expression of sets of genes, as there have been no comprehensive reports of the growth-phase-dependent transcriptional profile of RN6390 in comparison to clinical isolates of S. aureus.
The second objective of this study was to define the sarA and agr regulons in a biofilm-positive clinical isolate. Both sarA and the effector molecule of the agr regulon, RNAIII, play important roles in biofilm formation (Vuong et al., 2000; Beenken et al., 2003; Valle et al., 2003). The only transcriptional profiling study of these two regulatory loci has been performed in a strain possessing the rsbU mutation, which leads to altered expression of both sarA and RNAIII (Dunman et al., 2001). To achieve our objectives, we utilized a custom Affymetrix GeneChip generated based on sequence data from six different strains of S. aureus (Dunman et al., 2004). We also performed qRT-PCR on selected gene targets to determine whether the microarray results observed with UAMS-1 were unique to this strain, or were also observed in other clinical strains isolated from patients with a diverse array of infections. We have previously compared UAMS-1 and RN6390 on a strictly genotypic level using comparative genomic hybridizations to the same Affymetrix GeneChip (Cassat et al., 2005). Results from those previous studies were crucial in that they allowed us to distinguish between absent genes and absent transcripts. For example, an initial assessment of the profiling data indicated that 731 genes were differentially transcribed between UAMS-1 and RN6390 in the exponential growth phase. However, the genes encoding 459 (63 %) of these transcripts were absent in one of the two strains.
Growth-phase-dependent differences in the transcriptional profiles of UAMS-1 and RN6390
Based on at least a threefold difference in the level of transcription and statistical significance (P0.05) in repetitive assays, 272 genes that were present in both UAMS-1 and RN6390 were differentially regulated in the exponential growth phase. Seventy-seven (28.3 %) of these were expressed at higher levels in UAMS-1 than in RN6390, while 195 (71.7 %) were expressed at higher levels in RN6390. In the post-exponential growth phase, 293 genes were differentially expressed in UAMS-1 and RN6390, with 240 (81.9 %) of these being expressed at higher levels in UAMS-1. Interestingly, a recent study has demonstrated that sarA stabilizes mRNA transcripts (Roberts et al., 2006). Although the differences we observed in the relative levels of sarA transcription in RN6390 versus UAMS-1 did not reach our threefold threshold, we did find that the level of sarA transcript was higher in RN6390 than in UAMS-1 in the exponential growth phase (1.7-fold), while the opposite was true in post-exponential growth (2.2-fold). It is unclear whether the growth-phase-dependent differences observed with respect to sarA are just a reflection of the overall expression profile, or whether they contribute to the differences observed between UAMS-1 and RN6390, with respect to overall gene expression levels in exponential versus post-exponential phase. This latter possibility is addressed in more detail below.
Several of the genes upregulated in UAMS-1 relative to RN6390 have been previously reported to be part of the SigB regulon (Bischoff et al., 2004). This is not surprising, given that RN6390 carries a mutation in rsbU, and is functionally SigB deficient (Kullik et al., 1998). We confirmed that our RN6390 strain was functionally SigB deficient, both by sequencing rsbU (data not shown) and by quantitative RT-PCR for asp23 (Fig. 1), expression of which is known to be tightly controlled by SigB (Gertz et al., 1999; Giachino et al., 2001). We also confirmed by sequencing that UAMS-1 did not have the rsbU mutation characteristic of NCTC 8325 strains (data not shown), and microarray analysis confirmed that both asp23 and csbD, a second gene that is tightly regulated by SigB (Gertz et al., 1999), were upregulated in UAMS-1 in comparison to RN6390, in both the exponential and post-exponential phases (Table 3). Expression levels of asp23 were lower in RN6390 than in any of the nine clinical isolates examined. Other genes that were expressed at higher levels in UAMS-1 than in RN6390 have also been shown previously to be positively regulated by sigB (Bischoff et al., 2004). Similarly, there were also examples of genes that have been shown previously to be negatively regulated by sigB, and were expressed at lower levels in UAMS-1 than in RN6390. Included among these were agr (RNAIII), sspABC (the serine proteases) and sak (staphylokinase). For example, GeneChip analysis indicated that RN6390 expressed approximately 50- to 100-fold higher levels of the serine proteases sspA, sspB and sspC during post-exponential-phase growth (Table 3). We have determined using zymogram and azocasein analysis previously that RN6390 has higher proteolytic activity than UAMS-1 (Blevins et al., 2002). The gene encoding staphylokinase was also upregulated approximately 14-fold and 38-fold in RN6390 versus UAMS-1 in the exponential and post-exponential phases, respectively (Table 3).
|
) with the genes differentially expressed in UAMS-1 and RN6390 revealed that the rsbU mutation in RN6390 cannot fully account for the differences observed between UAMS-1 and RN6390. For example, in the exponential growth phase, only 34 of 77 (44.2 %) genes upregulated in UAMS-1 relative to RN6390 were part of the SigB regulon, as defined by Bischoff et al. (2004). Similarly, in the post-exponential growth phase, only 75 of 240 (31.3 %) genes that were upregulated in UAMS-1 versus RN6390 were part of the SigB regulon.
Mutation of sigB has been shown to affect expression of both sarA and agr (Bischoff et al., 2001). We have previously demonstrated that RN6390 expresses RNAIII at elevated levels in comparison to UAMS-1 (Blevins et al., 2002), and our microarray analysis confirmed that RN6390 expressed an increased level of the hld, agrA and agrB transcripts in the exponential growth phase in comparison to UAMS-1 (Table 3) (the agrC and agrD genes were not identified as upregulated in RN6390 versus UAMS-1 because the two strains possess different agr subtypes; Blevins et al., 2002). The hld transcript, which is included within the RNAIII effector molecule of the agr system (Janzon et al., 1989), was upregulated 9.4-fold within RN6390 relative to UAMS-1, in the exponential growth phase. However, this value is likely to underrepresent the actual differences in hld expression between RN6390 and UAMS-1 because the levels of hld/RNAIII exceeded the GeneChip saturation threshold in RN6390, and this precluded accurate assessment of relative RNAIII levels in RN6390 and UAMS-1. Similarly, GeneChip analysis could not accurately measure the levels of hld/RNAIII transcript in the post-exponential growth phase because the expression levels in both RN6390 and UAMS-1 exceeded the GeneChip saturation threshold. We therefore used quantitative RT-PCR (qRT-PCR) to more accurately assess the relative expression levels of RNAIII in RN6390 and UAMS-1. This analysis revealed that RN6390 produces approximately 200-fold more RNAIII than UAMS-1 in the exponential growth phase, and approximately 15-fold more RNAIII in the post-exponential phase (Fig. 2). Given the prominent role for RNAIII in proposed regulatory networks controlling virulence factor expression in S. aureus (Arvidson & Tegmark, 2001; Cheung & Zhang, 2002; Novick, 2003; Bronner et al., 2004), we also examined the expression of RNAIII in nine other clinical isolates by qRT-PCR. Although there was considerable variability in the amount of RNAIII produced by clinical isolates, the amount produced by RN6390 was higher than that of any other clinical isolate in the post-exponential growth phase (Fig. 2). Furthermore, despite variation in RNAIII levels among the ten clinical isolates examined, all of these isolates were able to form a biofilm in vitro (data not shown). The variability observed among clinical isolates, all of which were rsbU positive (data not shown), suggests that factors other than sigB also influence overall levels of RNAIII in clinical isolates of S. aureus. To address this issue more directly, we generated a UAMS-1 sigB mutant, and measured RNAIII expression by qRT-PCR in comparison to wild-type UAMS-1. Although RNAIII expression did increase substantially in a UAMS-1 sigB mutant, RNAIII expression in RN6390 was still approximately twofold greater during post-exponential growth. Additionally, mutation of sigB in UAMS-1 did not result in a reduced capacity to form a biofilm in vitro (Fig. 4). These observations further support the notion that the differences between UAMS-1 and RN6390 cannot solely be explained by the deficit in sigB production by RN6390.
|
|
). However, there are conflicting reports about the role of sigB in the production of SarA (Horsburgh et al., 2002; Bischoff et al., 2004; Ziebandt et al., 2004), and our previous Western blot results failed to demonstrate an obvious difference in the amount of SarA produced by RN6390 and UAMS-1 (Blevins et al., 2002). Nevertheless, as noted above, we did find differences in the overall level of sarA transcript in RN6390 and UAMS-1, and the relative levels were growth-phase dependent in a fashion that appeared to correlate with overall levels of gene expression in the two strains. This suggests that relatively minor differences in sarA expression and levels of SarA may have a dramatic impact on expression of other genes. Additionally, Arvidson & Tegmark (2001) have suggested that SarA and its homologues are repressors, and that the agr-encoded RNAIII effector molecule may function, at least in part, by binding SarA and acting as an anti-repressor. In this case, it would be the relative levels of RNAIII and SarA that are important, rather than the absolute amount of SarA, and, as noted above, RNAIII levels are much higher in RN6390 than in UAMS-1.
We did find differences in the expression levels of other genes that have been reported to have an impact on S. aureus virulence regulatory circuits. For example, transcription of the gene encoding the SarA homologue SarS was upregulated in UAMS-1 relative to RN6390, in both the exponential and post-exponential growth phases (Table 3). In fact, GeneChip analysis indicated that the level of sarS transcript in RN6390 was too low to allow accurate measurement, so we performed qRT-PCR to more conclusively assess the relative levels of sarS transcript in UAMS-1 and RN6390. This confirmed that expression levels of sarS were higher in UAMS-1 relative to RN6390 (Fig. 3), and that sarS levels in other clinical isolates were similar to UAMS-1, in terms of both quantity and temporal pattern of expression (Fig. 3). Although sarT has been reported to induce expression of sarS (Schmidt et al., 2003), the observation that sarS was expressed at lower levels in the sarT-positive strain RN6390, and higher levels in sarT-negative strains (UAMS-1, UAMS-601, UAMS-1138 and EMRSA-16) (Cassat et al., 2005), suggests that sarT may not be a primary determinant of the overall expression level of sarS in clinical isolates.
|
). This is consistent with the observation that expression of spa was also dramatically increased in UAMS-1 relative to RN6390 (Table 3). Indeed, expression levels of spa in UAMS-1 exceeded the GeneChip saturation threshold in both exponential and post-exponential phases, whereas spa expression in RN6390 remained well below the saturation threshold, even in the exponential growth phase. qRT-PCR analysis confirmed that UAMS-1 expressed approximately 300- to 1000-fold more spa than RN6390 in the exponential and post-exponential phases, respectively (data not shown). This is consistent with our earlier report demonstrating increased expression of spa in UAMS-1 relative to RN6390 by Northern blotting (Blevins et al., 2002). Spa production has been associated with increased virulence and modulation of the immune response (Palmqvist et al., 2002, 2005; Gomez et al., 2004; Goodyear & Silverman, 2004). This suggests that the relative levels of spa expression may at least partially account for our previous observation that UAMS-1 is more virulent than RN6390 in our animal models of musculoskeletal infection (Blevins et al., 2003).
Expression of both sarS and spa is repressed by agr (Cheung et al., 2001). The fact that agr is expressed at lower levels in UAMS-1 and other clinical isolates, relative to RN6390, may therefore explain why sarS and spa are produced at elevated levels in these strains. However, it has been shown recently that mutation of tcaR (SACOL2353) in the S. aureus strain COL also results in decreased expression of sarS and spa, and that strains derived from NCTC 8325, including RN6390, are natural truncated mutants of tcaR (McCallum et al., 2004). We have confirmed that UAMS-1 encodes an intact tcaR locus (data not shown), and this suggests that the elevated expression of sarS, and consequently spa, in UAMS-1 relative to RN6390 is also a function of a mutated tcaR in RN6390. Furthermore, expression from the sarS promoter is reported to be SigB dependent (Tegmark et al., 2000; Bischoff et al., 2004), which suggests that the rsbU mutation in RN6390 contributes to lower sarS expression levels. To test this, we measured sarS expression in the UAMS-1 sigB mutant by qRT-PCR. Similar to the results observed with RNAIII expression, mutation of sigB in UAMS-1 led to a decrease in sarS expression, but not to the levels observed in RN6390 (Fig. 3).
Other regulatory elements that were expressed at different levels in UAMS-1 and RN6390 included saeRS, which also encodes a two-component regulatory system (Giraudo et al., 1994; Novick & Jiang, 2003). Specifically, both genes were expressed at higher levels in UAMS-1 versus RN6390 during post-exponential-phase growth (Table 3). However, while the differences we observed between UAMS-1 and RN6390 with respect to saeS were statistically significant (3.6-fold), the differences we observed with respect to saeR fell just below our cutoff (2.8-fold). This was somewhat unexpected, since saeS and saeR are transcribed as an operon (Steinhuber et al., 2003). To further investigate this issue, we also analysed saeR expression by qRT-PCR in UAMS-1 and RN6390, and found that UAMS-1 expressed 2.38-fold and 3.64-fold more saeR transcript than RN6390 in the exponential and post-exponential phases, respectively (data not shown). Taken together, these results confirm that saeS and saeR are expressed at comparable levels in UAMS-1, and that these levels exceed those observed in RN6390. Whether the increased levels observed in UAMS-1 are biologically relevant is difficult to determine. However, expression of the gene encoding thermonuclease (nuc) has been reported to be almost exclusively dependent on the saeSR two-component system (Novick & Jiang, 2003), and the fact that expression of nuc was higher in UAMS-1 than in RN6390 suggests that the differences we observed in the relative expression levels of saeS and saeR were phenotypically relevant (Table 3). This suggestion is supported by the earlier observation that expression of nuc is repressed by sigB (Kullik et al., 1998). It would therefore be expected that nuc would be expressed at higher levels in RN6390 in the absence of the additional regulatory influence of saeSR. This is also potentially important in that sae plays an important role in device-related infection (Goerke et al., 2005).
We found that expression of many heat-shock and stress-response-related genes was upregulated in RN6390 relative to UAMS-1, particularly in the exponential phase of growth. Among the upregulated genes were those encoding the chaperone ClpB, the protease ClpC, the chaperonins GroE and GroS, the molecular chaperones DnaJ and DnaK, the transcriptional repressor of class-III stress genes CtsR, the heat-shock protein GrpE, the heat-inducible transcription repressor HrcA, an Hsp20-family heat-shock protein (COL SA2385), and the universal stress protein UspA (Table 3). Most of these transcripts were subsequently found to be slightly upregulated in UAMS-1 relative to RN6390 in the post-exponential phase, although none met the threefold requirement for differential expression (Table 3). Based on these observations, we hypothesize that the early high-level expression of RNAIII by RN6390 leads to premature activation of a stress response in the exponential growth phase. It is unknown whether these genes are upregulated as a result of the imbalance of global regulatory molecules, such as RNAIII or SigB, in RN6390 relative to clinical isolates such as UAMS-1. However, it is tempting to speculate that the overexpression of genes encoding certain exoproteins in RN6390 (e.g. sspABC) constitutes a stress to the bacterial cell, resulting in upregulation of this group of genes.
Finally, several genes with functions in central metabolic processes were also differentially expressed in UAMS-1 and RN6390. For example, 12 genes encoding proteins in the purine biosynthesis pathway were upregulated in RN6390 relative to UAMS-1 during exponential growth (Table 3). However, expression of these genes was subsequently upregulated in UAMS-1 relative to RN6390 during post-exponential growth (Table 3). Additionally, the arginine deiminase operon (arcCDBA) was also expressed at higher levels in RN6390 than UAMS-1 (Table 3), in both exponential and post-exponential phases. The arc operon has been previously reported to be upregulated by agr in an 8325-4 derivative (Dunman et al., 2001), and our GeneChip data indicated that this was also the case for UAMS-1. These observations suggest that transcripts with functions in central metabolic processes may also be expressed in a strain-dependent manner, and that this is influenced by global regulatory molecules, such as RNAIII. This further calls into question the use of NCTC-8325-derived strains, with respect to both characterization of global regulatory circuits controlling expression of S. aureus virulence factors, and the use of these strains as model organisms for the study of staphylococcal physiology.
Transcriptional profiling of UAMS-1 sarA and agr mutants
The regulatory circuits controlling the production of S. aureus virulence factors are complex, and the number of regulatory loci known to be involved is increasing (Arvidson & Tegmark, 2001; Cheung & Zhang, 2002; Novick, 2003; Bronner et al., 2004). Nevertheless, it seems clear that agr and sarA play central roles in these regulatory circuits. The agr locus encodes a quorum-sensing system that is involved in production of the effector molecule RNAIII, which is postulated to be a central element in the mid-exponential-phase switch from the production of surface and adhesive molecules to the production of toxins and exoproteins (Novick, 2003). SarA is a DNA-binding regulatory protein that influences the expression of multiple genes, including those that contribute to virulence and biofilm formation (Rechtin et al., 1999; Arvidson & Tegmark, 2001; Dunman et al., 2001; Cheung & Zhang, 2002; Beenken et al., 2003; Blevins et al., 2003; Sterba et al., 2003; Koenig et al., 2004). Consensus DNA-binding sites have been proposed for SarA (Chien et al., 1999; Sterba et al., 2003), but these have little predictive value, both because they have considerable ambiguity and because they are very AT rich, which complicates the identification of potential binding sites in the AT-rich genome of S. aureus. Although it does not distinguish between direct and indirect regulatory effects, transcriptional profiling is one method to help identify targets of SarA-mediated regulation. Dunman et al. (2001) have previously reported the transcriptional profiling of both sarA and agr regulatory mutants. However, that work was completed using RN27, which is an NCTC 8325 strain that has the 11 bp deletion in rsbU (Kullik et al., 1998), resulting in altered levels of sarA and RNAIII (Bischoff et al., 2001). Additionally, the GeneChip used in the previous sarA and agr profiling studies represented only 86 % of the COL genome, and did not include any of the unique genes present in clinical isolates such as EMRSA-16. For these reasons, we performed transcriptional profiling of UAMS-1 agr and sarA mutants using a more comprehensive GeneChip (Dunman et al., 2004), in order to better define the sarA and agr regulons in a biofilm-positive clinical isolate of S. aureus.
Transcriptional profiling of the UAMS-1 agr regulon
GeneChip analysis revealed that mutation of agr in UAMS-1 primarily resulted in decreased transcript levels of target genes. Indeed, only four genes were expressed at significantly higher levels in the UAMS-1 agr mutant in comparison to the wild-type UAMS-1. Nine genes were upregulated in the exponential growth phase, and 69 genes were upregulated in the post-exponential growth phase, in UAMS-1 versus its agr mutant. The only genes differentially expressed in UAMS-1 versus its isogenic agr mutant during both exponential and post-exponential growth were the genes of the agr locus itself. Included among the genes that were expressed at higher levels in UAMS-1 than its agr mutant in the post-exponential growth phase were the genes encoding capsule biosynthesis proteins, the gene encoding gamma haemolysin (hlgA), the gene encoding lipase (lip, SACOL2694), and the gene encoding 
-toxin (hla), all of which are consistent with the general model of agr-mediated regulation (Novick, 2003) (Table 4).
|
), and the results of others (Ziebandt et al., 2004), we found that mutation of agr in UAMS-1 did not significantly change the levels of sarA in exponential- or post-exponential-phase growth (data not shown). Our findings of a relatively small number of agr-regulated genes in UAMS-1 may seem to suggest that UAMS-1 is functionally deficient in agr expression. However, UAMS-1 does exhibit the characteristic mid-exponential-phase increase in RNAIII expression (Fig. 2), increasing approximately 100-fold after the transition from exponential phase to post-exponential phase. This observation, together with the fact that the impact of mutating agr was most apparent in the post-exponential growth phase, is consistent with the fact that agr acts as a quorum-sensing system that is induced in vitro as cell density increases, and cultures enter post-exponential growth (Novick, 2003). Moreover, we have previously demonstrated that mutation of agr in UAMS-1 results in a reduced capacity to cause osteomyelitis (Gillaspy et al., 1995). Nevertheless, our results indicate that mutation of agr in clinical isolates has a less dramatic impact on gene expression than that observed in earlier experiments employing an 8325 strain, and this is consistent with the observation that clinical isolates generally produce less RNAIII than RN6390 (see above).
Other S. aureus regulatory loci have also been suggested to undergo agr-mediated regulation. For example, it has been reported that expression of the saeRS locus is activated by agr (Giraudo et al., 2003; Novick & Jiang, 2003). However, we did not observe a significant difference between the levels of saeRS expression between UAMS-1 and its agr mutant (Table 4). This discrepancy could be due to the different genetic backgrounds used in the various studies. Indeed, other researchers have reported that expression of the sae locus is both strain dependent and dependent on growth conditions (Novick & Jiang, 2003). It is interesting to note, however, that expression of saeRS was slightly lower than wild-type levels in the UAMS-1 agr mutant during exponential growth, but slightly higher during post-exponential growth. This does suggest that agr has an impact on saeRS expression, but that it is minimal in comparison to other strains. A similar inverse relationship was observed with other regulatory loci, including srrAB (Table 4). This locus also encodes a two-component system involved in regulating production of S. aureus virulence factors (Yarwood et al., 2001; Pragmann et al., 2004). Interestingly, this same pattern was also observed for certain genes regulated by mgrA (Luong et al., 2006).
We did find associations between certain agr-regulated genes and genes involved in biofilm formation. Previously, we have reported that RN6390 is biofilm deficient unless the agr locus is deleted (Beenken et al., 2003). This suggests that overexpression of RNAIII to the levels seen in RN6390 may have a profound negative impact on biofilm formation. Indeed, others have also observed an inverse relationship between biofilm formation and expression of the RNAIII transcript, which includes the phenol-soluble modulin (PSM) 
-toxin (hld) (Vuong et al., 2000). PSMs are surfactant-like peptides that have pro-inflammatory properties, and a role in the detachment of biofilms, in Staphylococcus epidermidis (Mehlin et al., 1999; Otto et al., 2004; Vuong et al., 2004; Yao et al., 2005). In addition to -toxin (hld) itself, we also found that expression of another gene (SACOL1186) encoding a PSM was reduced approximately 30-fold in a UAMS-1 agr mutant in comparison to the parent strain (Table 4). This suggests that a UAMS-1 agr mutant might have an enhanced capacity to form a biofilm in comparison to the parent strain. However, in contrast to RN6390, UAMS-1 readily forms a biofilm, indicating that the level of PSM production in UAMS-1 does not adversely affect biofilm formation. At the same time, expression of SACOL1186 in RN6390 was elevated approximately eightfold in comparison to UAMS-1 during exponential growth. This suggests that the reduced capacity of RN6390 to form a biofilm in comparison to UAMS-1 may involve PSMs other than the RNAIII-encoded -toxin. Interestingly, we also found that a second PSM-encoding gene, which is adjacent to SACOL1186 in the COL strain, is present in RN6390, but absent in UAMS-1, UAMS-601 and SANGER-252 (Cassat et al., 2005). This suggests that the differences between RN6390 and clinical isolates, with respect to biofilm formation, may involve differences in genetic content, as well as differences in the relative levels of transcription described in this report.
Transcriptional profiling of a UAMS-1 sarA mutant
Genes whose expression was influenced by mutation of sarA in UAMS-1 are shown in Table 5. We have previously reported that mutation of sarA in UAMS-1 and other clinical isolates results in certain phenotypes that are distinct in comparison to an RN6390 sarA mutant. For example, mutation of sarA results in increased haemolysin activity in clinical isolates, but decreased haemolytic activity in RN6390 (Blevins et al., 2002). GeneChip analysis confirmed that expression of hla was significantly increased in a UAMS-1 sarA mutant during exponential-phase growth (Table 5). Expression of hla was also higher in a UAMS-1 sarA mutant during post-exponential-phase growth, although an accurate assessment of the increase was hindered by saturation of the GeneChip (Table 5). These findings are in direct contrast to reports concluding that mutation of sarA results in decreased transcription of hla (Chan & Foster, 1998; Chien et al., 1999); in fact, all current models of global regulatory circuits in S. aureus are based on this conclusion (Arvidson & Tegmark, 2001; Cheung & Zhang, 2002; Novick, 2003; Bronner et al., 2004). This is perhaps due to the fact that these models are based primarily on studies done with 8325-4 strains. Indeed, we have previously demonstrated that the results we observed with UAMS-1 are characteristic of other S. aureus isolates, with the exception of RN6390 (Blevins et al., 2002).
|
). However, it has been reported that EMRSA-16 produces a truncated and inactive form of Hla, due to a nonsense mutation after codon 112 (Holden et al., 2004). Given the high degree of genetic similarity between EMRSA-16 and UAMS-1 (Cassat et al., 2005), we amplified and sequenced hla from UAMS-1 to assess whether this strain also had the nonsense mutation. The results confirmed that both UAMS-1 and UAMS-601 have a nonsense mutation after codon 112, as originally described for EMRSA-16 (Holden et al., 2004), rendering these strains Hla deficient. This clearly indicates that the increased haemolytic activity we reported earlier (Blevins et al., 2002) is not due to increased production of Hla, but to some other haemolytic protein. Nevertheless, the hla mutation in UAMS-1 did not preclude measurement of hla transcript levels by GeneChip analysis. Indeed, Genechip analysis indicated that both UAMS-1 and RN6390 expressed the hla transcript at similar levels in the exponential phase, and both strains showed a characteristic increase in expression after the transition to exponential phase (data not shown).
The fact that UAMS-1 does not make -toxin, yet is virulent in animal models of osteomyelitis (Smeltzer et al., 1997) and septic arthritis (Blevins et al., 2003), is interesting in light of studies indicating that -toxin makes an important contribution to staphylococcal virulence (Jonsson et al., 1985; Nilsson et al., 1999; Dajcs et al., 2002). While this could reflect the use of different animal models, there has also been a report indicating that the loss of haemolytic activity in S. aureus agr mutants is correlated with increased survival in vivo (Schwan et al., 2003). In addition, Bayer et al. (1997) have found that hyperproduction of -toxin results in a paradoxical decrease in virulence in an experimental endocarditis model. Interestingly, a second clinical isolate (UAMS-601) that has the nonsense mutation that precludes the production of -toxin is also highly virulent in the same endocarditis model (Dr Arnold Bayer, personal communication). There is also a single report indicating that -toxin is required for biofilm formation in S. aureus (Caiazza & O'Toole, 2003), but UAMS-1, UAMS-601 and EMRSA-16 are all capable of forming a biofilm (Beenken et al., 2003, 2004; Cassat et al., 2005). While these disparate results do not preclude an important role for -toxin in at least some forms of staphylococcal infection, clinical isolates, such as UAMS-1, UAMS-601 and EMRSA-16, are clearly capable of causing infection both in humans and in experimental models, despite their inability to produce Hla.
Reports have indicated that SarA activates its own transcription (Manna et al., 1998). In contrast to those studies, our GeneChip analysis indicated that sarA transcription increased to saturating levels in a UAMS-1 sarA mutant during exponential growth (Table 5), suggesting that SarA may repress its own expression in at least some strains. Mutation of sarA also influenced the expression of several other putative regulatory genes, including sarY (SACOL2289), the AraC-family regulator adjacent to sarY, a MarR-family regulator (SACOL1060) upstream of the bifunctional autolysin gene atl, and a LysR-family regulator (SACOL0980) (Table 5). Although the phenotypic significance of changes in the expression of these putative regulatory genes is unknown, these results indicate that sarA directly or indirectly influences the expression of several other regulatory molecules. In contrast, we did not find a significant change in transcription of the agr locus or the level of hld (RNAIII) expression in a UAMS-1 sarA mutant. This is in contrast to previous reports indicating that sarA is required for full expression of agr (Cheung et al., 1997; Chien et al., 1999; Dunman et al., 2001). The discrepancy between our results and those of earlier reports may be a function of growth conditions, since the regulatory impact of sarA on agr expression was only evident when S. aureus was grown under relatively low oxygen tension (Chan & Foster, 1998). Our use of a 1 : 5 culture volume : flask ratio, and constant shaking, may therefore have masked the regulatory impact of sarA on expression of agr.
GeneChip analysis also revealed that expression of the serine protease genes (sspABC) was significantly increased in a UAMS-1 sarA mutant (Table 5). This is also consistent with our previous findings that proteolytic activity increases upon mutation of sarA (Blevins et al., 2002). In addition to repression of sspABC expression, we also found that sarA repressed the expression of several other exoprotein-encoding genes, including nuc (thermonuclease), sak (staphylokinase), aur (aureolysin), and lytN (cell wall hydrolase) (Table 5). Because many of these exoproteins are proteolytic, this may contribute to the reduced capacity to bind host proteins in a sarA mutant. We have previously reported that mutation of sarA in UAMS-1 leads to a decreased capacity to bind fibronectin, and that this is primarily a function of the increased production of proteases, rather than a change in expression of fnbA (Blevins et al., 2002). Other investigators have reached the same conclusion (Karlsson et al., 2001; Karlsson & Arvidson, 2002). This is consistent with the observation that we did not find a significant change in the expression of fibronectin-binding proteins in a UAMS-1 sarA mutant (data not shown). Biofilm formation by UAMS-1 and other clinical isolates is dependent on coating the wells with plasma proteins (Beenken et al., 2003), which suggests that the reduced capacity of a UAMS-1 sarA mutant to form a biofilm (Beenken et al., 2003) may be at least partially dependent on the increased production of proteases, and the reduced capacity to bind host proteins, including fibronectin. However, SspA is the primary protease produced by UAMS-1, and while its production is dramatically increased in a sarA mutant (Blevins et al., 2002), mutation of sspA did not restore the capacity of a UAMS-1 sarA mutant to form a biofilm (Fig. 4). Other investigators have also concluded that the reduced capacity of a sarA mutant to form a biofilm is not solely a function of an increase in proteolytic activity (Valle et al., 2003).
In an attempt to identify other genes in the sarA regulon that might influence biofilm formation, we compared the results of our UAMS-1 sarA profiling studies with those of previous experiments detailing the transcriptional response of UAMS-1 during growth within a biofilm (Beenken et al., 2004). We identified two genes, SACOL2198 (alsD) and SACOL2199 (budB or alsS), that were expressed at increased levels in a UAMS-1 biofilm (Beenken et al., 2004), and decreased levels in a UAMS-1 sarA mutant (Table 5). This suggests that the impact of sarA on expression of alsSD may be at least partly responsible for the reduced capacity of a UAMS-1 sarA mutant to form a biofilm (Beenken et al., 2003). To address that possibility, we assessed the relative capacity of UAMS-1 sarA and alsSD mutants to form a biofilm. Yang et al. (2006) have previously generated a UAMS-1 alsSD mutant, and confirmed that it does not produce acetoin. Our results demonstrated that the UAMS-1 alsSD mutant had a reduced capacity to form a biofilm, which was comparable to the deficit observed in a UAMS-1 sarA mutant. Furthermore, complementation of the sarA or alsSD mutations resulted in restoration of biofilm formation (see above). The alsS and alsD genes encode acetolactate decarboxylase and an acetolactate synthase, respectively, which function sequentially to convert pyruvate to acetoin, and ultimately to 2,3-butanediol. Production of acetoin and 2,3-butanediol, rather than the more acidic products of pyruvate metabolism, is important for acid tolerance in a number of bacterial species (Kovacikova et al., 2005). Induction of alsSD in a biofilm is therefore consistent with our hypothesis that a central theme of the adaptation of S. aureus to persistence within a biofilm is survival within the acidic environment associated with anaerobic growth (Beenken et al., 2004). Furthermore, both alsS and alsD have been reported to be upregulated during mild acid treatment of S. aureus (Weinrick et al., 2004). In fact, of the 95 genes found to be upregulated in a mature UAMS-1 biofilm (Beenken et al., 2004), well over half are also upregulated during mild acid treatment (Weinrick et al., 2004).
In summary, we performed genome-scale transcriptional profiling of the S. aureus laboratory strain RN6390 and the biofilm-positive musculoskeletal isolate UAMS-1. These studies confirmed important differences between the two strains with respect to overall gene expression patterns. Specifically, the overall profile in RN6390 was dominated by agr, as reflected by the relatively high expression level of genes encoding exotoxins, and low expression level of genes encoding surface proteins. Conversely, UAMS-1, which has been shown to be virulent in several animal models of musculoskeletal infection (Smeltzer et al., 1997; Elasri et al., 2002; Blevins et al., 2003; Beenken et al., 2004), had the opposite profile. This is consistent with previous studies demonstrating that UAMS-1 has a high binding capacity for host proteins, produces reduced levels of most exoproteins, and has an enhanced capacity to form a biofilm in comparison to RN6390 (Blevins et al., 2002; Beenken et al., 2003). Taken together, these results suggest that the capacity to efficiently bind host proteins makes an important contribution to staphylococcal pathogenesis, and that exotoxin production may be less important in at least some forms of infection. In that regard, it is also interesting to note that UAMS-1, UAMS-601 and EMRSA-16 carry a nonsense mutation in hla and are incapable of producing functional -toxin. Overall, the inverse relationship with respect to the production of different classes of virulence factors in UAMS-1 and RN6390 is consistent with previous studies suggesting that the phenotype of predominant clinical isolates favours the colonization phase of infection (Papakyriacou et al., 2000). We also performed transcriptional profiling of UAMS-1 sarA and agr mutants. These differences involved not only genes implicated in biofilm formation and virulence, but also genes with functions in central metabolic processes. Transcriptional profiling of UAMS-1 and its sarA mutant also allowed us to demonstrate that expression of the alsSD operon is dramatically reduced in a sarA mutant, and we subsequently confirmed that an alsSD mutant has a reduced capacity to form a biofilm that is comparable to that observed in a sarA mutant. In contrast, alsSD was not identified as part of the sarA regulon in the 8325 strain RN27 (Dunman et al., 2001). qRT-PCR analysis also confirmed that the patterns of gene expression observed in other clinical isolates were more similar to those of UAMS-1 than to those observed in RN6390.
Importantly, while we have focused much of our effort on isolates, such as UAMS-1, that cause musculoskeletal infection, the studies we report here also employed clinical isolates from other forms of staphylococcal infection, including isolates from soft tissue infections and septic shock, and pvl-positive isolates characteristic of community-acquired infection. However, while the results obtained with all of these strains were more similar to those of UAMS-1 than to those of RN6390, there was considerable variability among clinical isolates. This further emphasizes the need not only to examine clinical isolates of S. aureus such as UAMS-1, but also to extend analyses to additional clinical strains from other forms of staphylococcal infection.
|
ACKNOWLEDGEMENTS |
|---|
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTION |
|---|
