Public Health England (PHE) requests the referral of all TSS-associated isolates to the national reference laboratory for characterization, including toxin gene profiling. We identified clinician-diagnosed staphylococcal TSS cases from a database of referred S. aureus isolates from England, Wales, and Northern Ireland during January 2008–December 2012 using the search term “toxic shock syndrome.” Clinical and demographic data from the accompanying isolate referral form (Technical Appendix) that had been recorded contemporaneously were scrutinized for accuracy by a clinician (H.S.) before inclusion in the study.

We classified TSS cases in patients <16 years of age as pediatric. We classified cases in female patients 12–60 years of age as mTSS if the infection was associated with menstruation or positive vaginal culture for S. aureus. We classified the remaining cases as nmTSS. All cases had an associated S. aureus isolate.

The average annual incidence of TSS was calculated as cases per 100,000 population using Office for National Statistics UK population estimates (http://www.ons.gov.uk/ons/datasets-and-tables/index.html) and was based on data from 2009 and later (due to changes in reporting practice from November 2008 prompted by national guidance on toxin-producing S. aureus). We used total population for the United Kingdom excluding Scotland as the denominator for all TSS and nmTSS cases; the total female population 12–60 years of age as the denominator for mTSS cases, reflecting the age range of this group; and the number of children <16 years of age as the denominator for pediatric cases. We included data from 2008–2012 in all other analyses.

We made MLST-CC assignments on the basis of sequencing the staphylococcal protein A (spa) gene repeat region (21) and referencing spa server (http://spa.ridom.de/mlst.shtml) and MLST (http://saureus.mlst.net) databases. We performed SCCmec detection, typing, and toxin gene profiling (sea-e, seg-j, tst, and pvl only) by multiplex PCR (22,23).

For isolates from 2008–2011 (n = 148; Technical Appendix Table 1), we determined antimicrobial MICs by agar dilution (24) and interpreted them in accordance with European Committee on Antimicrobial Susceptibility Testing guidelines (http://www.eucast.org). We did not determine antimicrobial susceptibilities for isolates from 2012.

Based on molecular epidemiologic findings, we assessed TSST-1 production in all tst-positive CC30 MSSA isolates from the TSS cohort (n = 81), including TSS isolates associated with bacteremia, skin and soft tissue infections (SSTI), and deep infections. We also assessed TSST-1 production in randomly selected tst-positive CC30 MRSA isolates from non-TSS patients (n = 39, including carriage, bacteremia, and SSTI isolates) that had been submitted to the reference laboratory during the study period (Technical Appendix Table 1). We quantified TSST-1 in cell-free broth-culture supernatants by Western blot by comparison with purified TSST-1 protein standards (Technical Appendix).

We obtained normal-donor peripheral blood mononuclear cells (PBMC) from an approved subcollection of the Imperial College NHS Trust Tissue Bank (ICHTB reference R12023) from anonymized consenting healthy donors. We incubated PBMC (1 × 10⁶ cells/mL) with cell-free RPMI bacterial supernatants (1:1,000 dilution) prepared from tst-positive CC30 MSSA isolates from the TSS cohort (n = 77; 4 of the isolates did not grow in RPMI) and the randomly selected tst-positive CC30 MRSA isolates (n = 39) that were investigated for TSST-1 production. We cultured the PBMC in RPMI medium (Invitrogen, Hemel Hempstead, UK) supplemented with 10% fetal calf serum at 37°C for 48 h in triplicate (25). We measured proliferation after incorporating 1.0 μCi/well of [³H] thymidine and allowing an additional 16 h incubation.

We extracted whole genomic DNA from randomly selected tst-positive CC30 MSSA isolates from the TSS cohort (n = 4) and tst-positive CC30 MRSA isolates (n = 5) (Technical Appendix Table 1) (26). We prepared libraries using the Nextera-XT DNA Sample Prep Kit (Illumina, Cambridge, UK) and subjected them to MiSeq sequencing (Illumina), generating 150 bp reads. We deposited data in the GenBank short read archive (accession no. SRP082305). We mapped reads to MLST-CC matched reference genomes MRSA252 (GenBank accession no. NC_002952.2 (27) or MN8 (accession no. NZ_CM000952) using SMALT (http://www.sanger.ac.uk/resources/software/smalt/) and determined single-nucleotide polymorphisms (SNPs) by SAMtools and bcftools (28). We performed de novo assemblies using Velvet (https://www.ebi.ac.uk/~zerbino/velvet/) and annotated them using Prokka (http://www.vicbioinformatics.com/software.prokka.shtml). We used Artemis (http://www.sanger.ac.uk/science/tools/artemis) to visualize the mapping of sequence reads to the reference strain and manually confirm all polymorphisms. For targeted ccpA sequencing, we amplified and sequenced DNA using forward primer 1: 5′- CACAGTGTCGCGTGTTGTTA-3′ and reverse primer 1: 5′- TAAGCGCATCCCTACTGCAC-3′.

We analyzed data with GraphPad Prism 6.0 (GraphPad Software, La Jolla, California, USA). We tested categorical variables using Fisher exact test or χ² test. We summarized nonparametric data by medians and interquartile ranges (IQR) and compared 2 groups by Mann-Whitney U test. We summarized parametric data by means and SDs and analyzed 2 groups by t-test (2-tailed); we considered p<0.05 significant.

During January 2008–December 2012, a total of 195 TSS case isolates were referred to PHE. We excluded 15 cases from the study (duplicate isolates from the same case, 4 cases; isolates submitted for quality control testing, 2 cases; isolates from cases incorrectly recorded as TSS, 9 cases), leaving 180 microbiologically confirmed TSS cases with isolates. Because of missing clinical data, we were unable to classify 3 isolates as mTSS or nmTSS and could not ascertain the sex of 1 patient with nmTSS.

We considered the apparent rise in cases during 2008–2009 an artifact of increased clinical awareness of severe toxigenic S. aureus disease from late 2008, prompted by national guidance on toxin-producing S. aureus (Figure 1). Beginning in 2009, mTSS referrals declined annually, whereas nmTSS cases remained stable. By 2012, cases of nmTSS outnumbered mTSS. Overall, most cases were nonmenstrual (107, 59.4%). Average annual incidence per 100,000 population was 0.07 (95% CI 0.05–0.10) for all cases, 0.09 (95% CI 0.06–0.14) for menstrual cases, and 0.04 (95% CI 0.02–0.06) for nonmenstrual cases.

Despite an overall preponderance of female case-patients, we found no gender difference among nmTSS cases (Table 1). The median age of the cohort was 19 years; patients with nmTSS were younger than those with mTSS (median 15.0 vs. 21.5 years; p = 0.01).

Of the TSS cases studied, 39% (71/180) occurred in children <16 years of age; one sixth of all TSS case-patients were <1 year of age (Figure 2). The median age of pediatric TSS case-patients was 4 years, with an average annual incidence of 0.14/100,000 children (95% CI 0.08–022). However, among children <1 year of age, the average annual incidence increased to 0.45/100,000 (95% CI 0.26–0.79). Most pediatric nmTSS cases were related to burns (26.8%, 15/56) or SSTIs (25%, 14/56).

Five percent of all patients with TSS had died at the time of referral of the isolate. We found no difference in fatality rate between mTSS and nmTSS cases and no association with age (Technical Appendix Table 2). The infective focus in nmTSS cases was SSTI (n = 41), primary bacteremia (n = 15), burns (n = 15), deep abscess (n = 13), respiratory tract (n = 10), bone and joint (n = 4), unknown (n = 6), and other sites (n = 3). We found no association between site of infection and S. aureus lineage (Technical Appendix Figure 1).

Among 180 TSS S. aureus isolates, we identified 88 spa types associated with 15 different MLST-CCs (Technical Appendix Table 3). The leading cause of both mTSS and nmTSS was CC30 MSSA, accounting for >50% of infections (Figure 3), although we found a stronger association of CC30 with mTSS than with nmTSS (72.9% vs. 36.4%; p<0.0001; Technical Appendix Table 3). CC30 MSSA was also the leading cause of TSS among pediatric cases (31/71). We identified only 7 MRSA TSS isolates (Technical Appendix Table 4).

TSS isolates carried 3 superantigen genes on average (Technical Appendix Table 5). The most common superantigen gene among both mTSS and nmTSS isolates was tst (Table 2; Technical Appendix Figure 2), with the exception of the other 2 prevalent superantigen genes, seg and sei, that are carried on an enterotoxin gene cluster (egc) along with selm/n/o/u in most S. aureus isolates (5). The tst gene was associated with mTSS (Table 2) and strongly associated with the CC30 lineage of S. aureus (Technical Appendix Table 6). The superantigen gene sea combined with tst was also linked to mTSS (Table 2), whereas sea alone was associated with CC30 (Technical Appendix Tables 5, 6); sec was linked to nmTSS (Table 2) and CC45 (Technical Appendix Table 5). Ten nmTSS cases were associated with isolates that lacked any superantigen gene tested; 7 were CC15, highlighting severe disease attributable to this lineage that was unexplained by the presence of major superantigens (Technical Appendix Tables 5, 6).

Most isolates were MSSA (mecA negative). The rate of resistance to erythromycin was 9.2%; to ciprofloxacin, 8.5%; to tetracycline, 3.5%; and to teicoplanin, 1.4%. For 7 mecA-positive MRSA-TSS isolates, the resistance rate to ciprofloxacin was 57.1%; to erythromycin, 42.6%; and to clindamycin, 14.3%.

As MRSA-related TSS is rarely reported we examined these cases in more detail. All 7 MRSA cases were nonmenstrual, affecting mainly male patients; 3 were associated with SSTIs. The median patient age was 34 (IQR 2.3–64.3) years. Five isolates were identified as CC22-SCCmecIV, and 4 carried sec, corresponding to the healthcare-associated MRSA clade dominant in the UK, EMRSA-15; MRSA-TSS cases showed a clear association with this lineage (Technical Appendix Table 4). The remaining CC22 isolate carried tst and belonged to a MRSA lineage frequently identified in the Middle East. Only 1 MRSA-TSS isolate was CC30-SCCmecII, corresponding to the UK HA-MRSA clade EMRSA-16. One isolate was CC6-SCCmecII and lacked all superantigen genes tested.

The strong association of CC30 with TSS was unsurprising because of the presence of tst. We measured TSST-1 in broth-culture supernatants from tst-positive CC30 MSSA isolates from the TSS cohort and, for comparison, randomly selected clinical tst-positive MRSA isolates that belonged to the same lineage (CC30) (20,29) (Technical Appendix Table 1).

Of note, 77/81 tst-positive CC30 MSSA isolates produced detectable TSST-1, compared with 9/39 tst-positive CC30 MRSA isolates. The tst-positive CC30 MSSA isolates produced more TSST-1 than did tst-positive CC30 MRSA isolates, albeit with marked variability (88.5 ± 48.3 vs. 31.4 ± 18.1 ng/mL,;p<0.0001; Figure 4, panel A). Furthermore, the superantigenic activity of isolates, measured by T-cell proliferation in response to broth-culture supernatants, of tst-positive CC30 MSSA strains (164,893 ± 36,191 counts/min) was significantly greater than that of tst-positive CC30 MRSA strains (149,653 ± 30,412 counts/min; p = 0.02; Figure 4, panel B).

To ascertain the basis for the observed variability in TSST-1 production among CC30 S. aureus, we subjected 4 tst-positive CC30 MSSA isolates from the TSS cohort and 5 tst-positive CC30 MRSA clinical isolates to whole-genome sequencing. The tst gene, promoter, and regulator sequences, including SarA, SrrAB, agr, and σ^B, were identical among the 9 sequenced strains and reference isolates (MN8/MRSA252).

We detected mutations in TSST-1 regulator SaeRS in 2/4 tst-positive CC30 MSSA isolates; a synonymous SNP C481T in SaeR in 1 strain and a nonsynonymous SNP in SaeS in another resulted in a change from asparagine to serine at aa residue 218. Because these strains produced abundant TSST-1 (Technical Appendix Table 1), we did not study these mutations further.

We detected a nonsynonymous SNP in the sequence of regulator ccpA in all 5 tst-positive CC30 MRSA isolates but not in any tst-positive CC30 MSSA isolate. This difference translated into a change from threonine (ACA) to isoleucine (ATA) at aa residue 87/329 (Technical Appendix Figure 3).

To determine the prevalence of the ccpA (T87I) variant in CC30, we sequenced ccpA in an additional 34 tst-positive CC30 MRSA and 19 tst-positive CC30 MSSA isolates (Technical Appendix Table 1). Including genome-sequenced isolates, 33/39 tst-positive CC30 MRSA isolates had ccpA (T87I), compared with 0/23 tst-positive CC30 MSSA isolates, confirming an association of ccpA (T87I) with CC30 MRSA strains. Furthermore, ccpA (T87I) was strongly negatively associated with production of TSST-1 in tst-positive CC30 S. aureus: 26/33 ccpA (T87I) isolates did not produce TSST-1, compared with only 1/23 wild-type ccpA isolates (p<0.0001 by Fisher exact test).

We conducted SCCmec typing on a subset of tst-positive CC30 MRSA strains (n = 15; Technical Appendix Table 1). Results demonstrated an association of ccpA (T87I) with SCCmecII; 7/11 SCCmecII isolates had ccpA (T87I), compared with 0/4 SCCmecIV isolates (p = 0.03 by χ² test). This finding highlights the possibility that reduced TSST-1 production might be attributable to either SCCmecII or ccpA (T87I).