|Home | About | Journals | Submit | Contact Us | Français|
The accessory gene regulator (agr) of Staphylococcus aureus is a global regulator of the staphylococcal virulon, which includes secreted virulence factors and surface proteins. The agr locus is important for virulence in a variety of animal models of infection, and has been assumed by inference to have a major role in human infection. Although most human clinical S. aureus isolates are agr+, there have been several reports of agr-defective mutants isolated from infected patients. Since it is well known that the agr locus is genetically labile in vitro, we have addressed the question of whether the reported agr-defective mutants were involved in the infection or could have arisen during post-isolation handling. We obtained a series of new staphylococcal isolates from local clinical infections and handled these with special care to avoid post-isolation mutations. Among these isolates, we found a number of strains with non-haemolytic phenotypes owing to mutations in the agr locus, and others with mutations elsewhere. We have also obtained isolates in which the population was continuously heterogeneous with respect to agr functionality, with agr+ and agr− variants having otherwise indistinguishable chromosomal backgrounds. This finding suggested that the agr− variants arose by mutation during the course of the infection. Our results indicate that while most clinical isolates are haemolytic and agr+, non-haemolytic and agr− strains are found in S. aureus infections, and that agr+ and agr− variants may have a cooperative interaction in certain types of infections.
agr is a well-studied central transcriptional regulator that controls the expression of virulence-associated protein genes in Staphylococcus aureus. The agr locus (Fig. 1) consists of two divergent transcription units driven by promoters P2 and P3. The P2 operon encodes a two-component signalling module, of which AgrC is the receptor and AgrA is the response regulator. It also encodes two proteins, AgrB and D, which combine to produce and secrete an autoinducing peptide (AIP) that is the ligand for AgrC. AgrA functions to activate transcription from its own promoter and from the agrP3 promoter, which drives the synthesis of RNAIII, the effector of target gene regulation (Novick et al., 1993). RNAIII also encodes δ-haemolysin (Janzon et al., 1989), the expression of which serves as a surrogate for agr functionality.
Although its importance for pathogenesis in animal models is well established (Abdelnour et al., 1993; Arvidson & Tegmark, 2001; Gillaspy et al., 1995; Novick, 2003), and most clinical isolates are agr+, the isolation from clinical material of S. aureus strains expressing agr poorly, or not at all (Fowler et al., 2004; Sakoulas et al., 2002), raises the question of the precise role of agr in human S. aureus disease (Li et al., 1997; Nozohoor et al., 1998). This question is compounded by the fact that agr-defective mutants arise frequently in laboratory cultures (Bjorklind & Arvidson, 1980; McNamara & Iandolo, 1998; Somerville et al., 2002). Consequently, agr-defective genotypes found among clinical isolates could be the result of mutations occurring during subculture after isolation. Accordingly, we studied agr functionality in a series of clinical S. aureus isolates that were handled with a minimum of manipulation to avoid post-isolation mutations. We report that agr-defective mutants occur in clinical material, that these arise and persist during infections, and that they are not the result of post-isolation handling; thus, they represent an important subset of clinical S. aureus.
S. aureus-positive culture specimens were kindly donated by the NYU Tisch Hospital microbiology laboratory. The specimens were catalogued to indicate the age and sex of the patient, days to positive culture, diagnosis, type of specimen, antibiotics used to treat the patient, and antibiotic sensitivities of the isolates (see Supplementary Table S1, available with the online version of this paper). All specimens were obtained from primary cultures: blood specimens were from culture bottles; urine samples were from urine receptacles, etc. Swabs (from wounds, etc.) were placed in 1.0 ml CYGP broth (Novick, 1991) and snap-frozen with dry-ice/ethanol. Before freezing, GL (Novick, 1991) plates were inoculated with the specimen (blood, urine, wound, sputum). Cultures from these plates were tested for agr group (described below), cross-streaked against RN4220 to assess haemolysin activity (described below), and then resuspended in CY broth and frozen in dry-ice/ethanol to create a secondary stock, which was stored at −80 °C. All subsequent studies were performed with bacteria grown from the original specimen. The secondary stock was used to confirm results.
Laboratory bacterial strains and plasmids are listed in Table 1. Strains RN9688, RN9689, RN9690 and RN9691 were used for agr typing as described by Wright et al. (2005b). RN4220 was used to score haemolysin production. CYGP broth cultures, inoculated with bacteria grown overnight on GL agar at a cell density of 15 Klett units (~2×108 cells ml−1), unless otherwise specified, were incubated at 37 °C with shaking at 250 r.p.m.
agr type was determined by a luciferase plate assay as previously described (Wright et al., 2005b).
Haemolytic activities were determined by cross-streaking perpendicularly to RN4220, which produces only β-haemolysin (Traber & Novick, 2006), on a sheep blood agar (SBA) plate. This test can usually identify the three staphylococcal haemolysins active on SBA – α, β and δ (see Fig. 2) – because of the interactions between them: β-haemolysin enhances lysis by δ-haemolysin, but inhibits lysis by α-haemolysin (Elek & Levy, 1950). To determine δ-haemolysin production by single colonies, we prepared SBA plates by spreading 400 μl of a sterile twofold-concentrated RN4220 supernatant before plating a suitable dilution of the strain to be tested. Note that the β-haemolysin produced by RN4220 enables detection of δ-haemolysin (see Fig. 2).
Total cellular RNA was prepared according to a standard protocol (Traber & Novick, 2006), and Northern blot analysis was performed as described by Kornblum et al. (1988). PCR primers used to prepare DNA probes are listed in Supplementary Table S2. In all blots, 16s rRNA was used as a loading control.
A total of 146 S. aureus infection isolates, obtained between 1 August 2001 and 1 June 2002 from 102 patients with documented staphylococcal disease, were kindly provided by the Tisch Hospital bacteriology laboratory. Two or more cultures were obtained from 34 of these patients.
S. aureus strains have been divided into four agr specificity groups (Jarraud et al., 2000; Ji et al., 1997). Of the 102 patients, 51 had an agr-I strain, 34 an agr-II, 14 an agr-III and 1 an agr-IV strain. One patient (no. 75) was infected by strains of two different agr groups: an agr-I strain in two blood cultures, and an agr-II strain in a wound culture. Five strains could not be typed by the luciferase plate test and were typed by direct agrD sequencing (Ji et al., 1997). One strain could not be typed and has not been studied further. The intensity of the bioluminescence produced in the typing assay (Wright et al., 2005b) varied considerably from strain to strain, representing variation in AIP levels. Low levels of bioluminescence likely represented the basal activity of the agr P2 promoter, seen when the agr autoinducing circuit is not activated owing to genotypic defects. Therefore, the production of detectable AIP cannot be taken as an automatic indication of a functioning agr autoinduction circuit. Strains that produced no reaction in the plate test may have had mutations in agrB or agrD.
agr typing by other investigators has generated widely varying distributions of the agr types, presumably representing variations in the local prevalence of strains of the different agr groups (Gilot & van Leeuwen, 2004; Jarraud et al., 2002; Shopsin et al., 2003). The distribution of agr types among our isolates does not represent a significant departure from the overall results of other series, nor have we encountered any significant correlation between clinical features and agr type in this series. The agr typing results and haemolytic activities of the entire set of isolates are summarized in Supplementary Table S1.
Of the 146 cultures collected, 33, from 26 patients, were non-haemolytic on SBA. In 10 different cases, we obtained, from a single patient, separate cultures that had different haemolysin patterns. Many of these contained mixtures of haemolytic and non-haemolytic staphylococci, some of which are described in detail below. Eleven of the pure non-haemolytic strains were analysed to establish the basis for their lack of haemolytic activity. Nine strains had mutations in the agr locus that could have inactivated agr function, Of these nine, three could be complemented and three could not; these are presumed to have other genotypic causes for their non-haemolytic phenotype and are to be studied further. Three were resistant to the antibiotics used to select for the complementing plasmids and were not tested for complementation (see Table 2). Two of the non-haemolytic strains had the wild-type agr sequence corresponding to their agr group. These results are shown in Table 2.
Sixteen of the 33 non-haemolytic and 73 of the 146 haemolytic isolates (Supplementary Table S1) were meticillin-resistant S. aureus (MRSA). Therefore, the SCCmec element was not implicated in the development of agr defects. Loss of agr function has been associated with the development of vancomycin resistance (Sakoulas et al., 2002). However, in this series, there was no correlation between vancomycin treatment and the development of agr defects (Supplementary Table S1); nonetheless, all of the patients were treated with some type of antibiotic, raising the question of whether antibiotics in general could drive selection for agr negatives.
For several non-haemolytic strains, the level of bioluminescence seen in the agr typing test seemed too high to represent the basal level of P2 promoter activity. In these strains, the defect in haemolysin production may be a result of mutations inactivating α- or δ-haemolysin, or impairing RNAIII transcription, timing, function or stability. To differentiate these possibilities we tested the 11 above-mentioned strains for their ability to activate the P3 promoter, transcribe RNAIII and transcribe the RNAIII-regulated exoprotein genes hla and spa, with results shown in Fig. 3(a).
Two strains, RN9771 and RN9774, demonstrated a transcription pattern that closely resembled that of the agr+ strain. RNAIII was produced at wild-type levels, plasmid-carried agrP3-lux was strongly activated, hla was transcribed moderately strongly, and little or no detectable spa transcript was produced. This transcription pattern, coupled with lack of haemolytic activity, is typical of mutants with delayed agr activation, as described elsewhere (Traber & Novick, 2006) and discussed further below. Strain RN9771 had a complementable mutation in agrC while RN9774 had the wild-type agr sequence.
The second pattern was shown by strain RN9764, and is very similar to the pattern seen in the agr null control RN7206. It produced only a trace of RNAIII, activated P3 lux weakly and transcribed hla extremely weakly but had relatively strong spa transcription. RN9764 has a complementable mutation in agrB, indicating that it is agr-defective. It also has a probably unimportant mutation in agrC.
In the third pattern, strains RN9765, RN9772, RN9900 and RN9901, there was little or no detectable RNAIII; there were relatively strong spa bands, as expected, but also high levels of hla transcription. These strains had various agr mutations, some complementable, others not, but more importantly, are likely to have mutations that can bypass the requirement of RNAIII for hla transcription (McNamara et al., 2000). Testing for such mutations is currently in progress.
Finally, the fourth pattern, as seen in strains RN9902, RN9903, RN9904 and RN9906, shows no RNAIII, and a very weak hla transcript, but low to no spa. RN9902 is especially interesting as it had a wild-type agr sequence, suggesting a mutation in an upstream gene required for agr expression. Strains RN9903, RN9904 and RN9906 all have mutations in the agr locus that probably account for their low RNAIII levels, but would not account for low spa transcript levels. RN9903 and RN9906 were not tested for complementation owing to their resistance phenotypes. RN9904, with a clearly inactivating frameshift in agrA, was not complemented by an agrA clone and presumably has a mutation elsewhere that prevents agr expression.
As a further characterization of the phenotypes of these strains, we determined exoprotein profiles (Fig. 3b). Strains of the same agr type group generally had similar exoprotein profiles; however, whenever RNAIII was not detected, there were few or very weak exoprotein bands. These strains did not significantly activate the agrP3-lux fusion (Fig. 3a) and are probably agr-defective. However, the two non-haemolytic strains that transcribed RNAIII and hla strongly activated the agrP3-lux fusion, had robust exoprotein profiles, and were clearly agr+. Thus, using haemolysin as a measure of agr functionality will sometimes result in false negatives.
As relatively late transcription of RNAIII is associated with failure to translate δ-haemolysin and α-haemolysin (Traber & Novick, 2006), we hypothesized that late RNAIII expression could be responsible for the failure of the agr+ strain, RN9774, to produce α- or δ-haemolysin. As shown in Fig. 4(a), transcription of RNAIII in strain RN9774 was indeed delayed by 1–2 h as compared to RN6734. Transcription of hla was also delayed, as well as somewhat lower in intensity than in RN6734. Finally, spa was transcribed in RN9774 and was downregulated concomitantly with RNAIII activation, between 2 and 3 h, as opposed to RN6734, in which spa transcription was inhibited at all time points. Note that in the assay of plasmid-carried agrP3-lux activity, luciferase activity was determined with overnight plate cultures, which are in early stationary phase, and any delay in agr activation would not have been apparent.
We next hypothesized that the late agr activation in RN9774 could be the result of low AIP activity. To test this, we added a 1/10 volume of 6 h culture supernatant from strain RN9774, or from another agr-III strain, RN3984, or from the agr null strain RN7206 or sterile CYGP broth, to early-exponential-phase cultures of RN9774. We grew the cultures for 6 h after the addition of supernatant and analysed exoprotein production. As shown in Fig. 4(b), post-exponential agr-III supernatant from RN3984 or from RN9774 stimulated the production of many exoproteins, generating an exoprotein profile similar to that of the wild-type agr strain RN3984. The results with strain RN9774 indicate that the timing of agr activation may be an important factor in the regulation of virulence factor production in vitro.
As mentioned above, several of the non-haemolytic isolates were obtained from patients from whom haemolytic isolates were also obtained. Among these, five haemolytic and two non-haemolytic blood cultures were isolated from patient 60 and a haemolytic and a non-haemolytic isolate were obtained from patient 28. The cultures from each patient were otherwise identical (see below), suggesting that the primary cultures contained congenic haemolytic and non-haemolytic substrains. As this would not have been detected by cross streaking (see Fig. 2a), we diluted and plated these isolates for single colonies on SBA prespread with RN4220 supernatant (See Methods and Fig. 2).
For these two patients, many of the frozen primary cultures were mixed (Table 3). From patient 28, both wound cultures were mixed. From patient 60 three cultures were mixed, (one each from days 8, 9 and 10), and four were >99% pure (one each from days 5 and 9, both positive; two from day 11, one positive, one negative). The mixed cultures from patient 60 were highly variable in composition, containing 85% δ-haemolysin-positive (1140 of 1342 colonies scored), 3% positive (22 of 777) and 24% positive (330 of 1389) respectively. Each culture was independently plated three times and the relative proportion of haemolytic and non-haemolytic colonies was always similar. The variability of the proportions of haemolytic and non-haemolytic colonies in the cultures from patient 60 probably represents sampling variability owing to the very small number of bacteria that are ordinarily present in any blood sample.
These mixed cultures could have arisen by mutation or could represent co-infection by two distinct lineages. To distinguish between these alternatives, we isolated single colonies from each of the cultures from patients 28 and 60. Since these mixtures had probably resulted from in vivo mutations, we analysed these strains in rather more detail than the non-haemolytic isolates described above. For patient 60, from each of the mixed cultures we examined three haemolytic and three non-haemolytic colonies, and from the homogeneous cultures we also examined three colonies. From patient 28, we examined a haemolytic and a non-haemolytic colony from each culture (Table 4). For each isolate we determined ClaI restriction patterns (Supplementary Fig. S1), and spa types (not shown), which indicated that all of the isolates from patients 28 and 60 were identical, and sharply different from the patient 9 strains, suggesting that a common nosocomial multi-resistant MRSA strain had infected patients 28 and 60.
We sequenced the agr locus in representative non-haemolytic colonies from patients 28 and 60. Of the patient 60 isolates, we sequenced three from each culture containing non-haemolytic organisms, and in the patient 28 isolates, one non-haemolytic isolate from each culture. The non-haemolytic isolates from patient 60 contained four different mutations, three in agrA, and one in agrC (T399P) (Fig. 5, Table 4). Two of the agrA mutations had an insertion or deletion of an adenine in a string of seven adenines near the C terminus. The six-adenine mutation (6A) is predicted to add 21 amino acids and the eight-adenine mutation (8A), three. A second TAATAA stop sequence was brought into frame by the agrA-8A mutation. The run of seven adenines and the additional stop sequence was conserved in all four agr groups (GenBank accession numbers: Group I, CP000046; Group II, BA000018; Group III, BX571857; and Group IV, DQ229853, sequenced for this paper). The third mutation in agrA was a 10 bp central deletion of nucleotides 473–482 that is predicted to truncate the protein at residue 172. Two non-haemolytic variants were present in the patient 28 series – one with a complementable 11 nucleotide deletion in agrC (1105–1115), causing a frameshift in the histidine kinase domain, and one with the wild-type agr sequence.
The various agr mutations from patient 60’s cultures were isolated at different times, as shown in Fig. 6. The agrA-8A mutation was present in at least one colony on days 5, 9, and 11; the other mutants were each present in one culture only. The culture from day 5 contained three different mutants, agrA-8A, agrA-6A and agrA-del, those from days 9 and 11 had only the agrA-8A mutant, while the strain from day 10 had the agrC-T399P mutation and the agrA-8A mutation. Wild-type strains and agrA-8A mutants were present throughout. As noted, these differences in the composition of blood cultures are probably a result of sampling variation. Alternatively, the singly occurring mutants may have been released from the endocarditic vegetations only briefly.
The RNAIII transcription patterns and exoprotein patterns for the agrA-8A and agrA-6A strains have been presented elsewhere (Traber & Novick, 2006). Briefly, transcription of RNAIII in the agrA-8A strain was delayed when compared with the agr wild-type strain. In the agrA-6A strain, RNAIII is not detectably transcribed, suggesting that agrA is inactive. Since the C-terminal end of proteins homologous to AgrA is critical for DNA binding (Koenig et al., 2004) we suggest that the added amino acids may have compromised DNA-binding activity; alternatively, the mutant proteins may be unstable. The agrA-8A mutation had an exoprotein profile that more closely resembled the completely defective agrC-T399P mutant than the agr-wt isolate, perhaps owing to its delayed agr activation, as seen with RN9774.
Having the wild-type parent of naturally occurring agr− mutants enabled a test of the frequency of agr− mutations in a known genetic background and thus a test of the frequency of such mutations occurring during post-isolation handling. This test, using one of the agr+ isolates from patient 60, involved repeated subculturing of 48 h broth cultures, in which the bacteria were in deep stationary phase. Cultures were plated for single colonies on SBA+RN4220 supernatant, and then diluted 100-fold and incubated again for 48 h. Approximately 500 colonies were scored for each subculture. Non-haemolytic colonies began to appear after three subcultures, representing about 20 generations of growth and three stationary-phase challenges. The non-haemolytic portion increased to 20% after four subcultures, and to >99% after five. While this result is consistent with the occurrence and increased fitness in vitro of agr− mutants, it effectively rules out the possibility that the agr− mutants present in the mixed cultures from the patients represent post-isolation mutations, because the number of generations required for the appearance of agr mutations is far greater than could possibly have been experienced by these isolates during handling. Since there is no reason to suspect that the agr locus in the agr+ isolate from patient 60 is different in stability from that in the other isolates, this result supports the conclusion that the mutations identified occurred before or during the infections rather than afterwards during handling. For the four isolates with complementable agr mutations, we tested for the stability of the mutations. For one of the strains, it was possible to obtain a PCR product for the mutated region of the agr locus directly from the frozen clinical specimen. For two of the other three, we were unable to obtain a PCR product and resequenced the agr region using a new primary subculture. For all three strains, the same mutation was identified as that reported above (not shown). For the fourth strain, the primary culture was lost.
In this report, we have shown that non-haemolytic S. aureus variants arise and survive in clinical material and are, in at least one case, almost certainly not the result of post-isolation mutations. We conclude that agr− strains are involved in clinical disease.
In the series reported here, some 15% of isolates were non-haemolytic. These occurred in agr groups I, II and III. A few of the non-haemolytic isolates had complementable mutations in agrA, B and C. Some non-haemolytic strains with agr mutations were not complementable by agr clones, suggesting that genotypic factors outside the agr locus were responsible, and that the observed agr mutations were secondary and adventitious. There were also a few non-haemolytic isolates with a wild-type agr sequence, suggesting the existence of genes upstream of agr in the exoprotein pathway, and required for agr expression. The identity of any such genes remains to be determined, and experiments addressing this are in progress.
Among the non-haemolytic strains, RN9774 had an apparently normally functioning agr locus, as revealed by its ability to activate an agrP3-lux fusion as well as to transcribe RNAIII. Although this strain could have mutations in hla, accounting for its inability to produce α-haemolysin, as has been described previously (O’Reilly et al., 1990), its failure to produce δ-haemolysin is more difficult to explain, since it has the wild-type RNAIII/hld sequence. One possibility is that it expresses agr late in the overall growth curve, which has previously been shown to result in a lack of δ-haemolysin translation (Traber & Novick, 2006). We conclude that haemolysin production, which has been commonly used as an indicator of agr functionality, is usually but not always accurate when examining clinical isolates.
We believe that the mixed cultures we obtained from these blood cultures were representative of a mixture in the patient rather than a post-isolation mutation for the following reasons. First, we obtained the wound isolates directly from the infected material retrieved from the patient, leaving little opportunity for mutations to occur. Indeed, when we monitored the appearance of agr− mutants in an agr-wt strain, although such mutants appeared it took three 48 h passages before they became detectable.
Additional support came from further characterization of isolates from patients for whom we had at least two separate specimens. For patient 28, both cultures were mixed; for patient 60, three were mixed, three were positive and one was negative. The evidence that a mixture was present in patient 60, rather than representing post-isolation mutations, is especially compelling, because in addition to mixed cultures, we obtained a positive and a negative culture on the same day. Finally, we performed single-colony analysis of these specimens multiple times on multiple days. Each specimen contained a reproducible ratio of haemolytic to non-haemolytic organisms. Taken in sum, these data strongly suggest that although there is a low level of instability of the agr locus in these strains, as evidenced by the appearance of agr− mutants during long-term growth in broth, the clinical specimens described here clearly represent the occurrence and persistence of mixed infections.
The results described above confirm recent reports that agr− and other exoprotein-defective variants occur and persist in clinical material (Fowler et al., 2004; Sakoulas et al., 2002), and extend these observations to include skin and soft tissue infections, pneumonia and osteomyelitis. However, since agr-negative variants arise commonly in vivo as well as in vitro, the occurrence of such organisms in clinical material does not mean that they actually initiated the infection. In fact, agr-negative strains have not been shown to initiate infections, and it is suggested that they may be genotypic dead-ends in the ecology of the organism. In this connection, it is noted that agr expression seems to be important during the early stages of an infection (Wright et al., 2005b), although it is expressed poorly, if at all, during later stages, both in experimental models and in material from chronic human infections such as the cystic fibrosis lung, owing either to metabolic depression or to mutations (Goerke et al., 2000). These ideas have important implications for the natural history of staphylococcal infection and are currently under study.