A Controlled Human Infection Model of Group A Streptococcus Pharyngitis: Which Strain and Why?

GAS (Streptococcus pyogenes) is a leading global cause of infection-related morbidity and mortality. A modern CHIM of GAS pharyngitis could help to accelerate vaccine development and drive pathogenesis research. Challenge strain selection is critical to the safety and success of any CHIM and especially so for an organism such as GAS, with its wide strain diversity and potential to cause severe life-threatening acute infections (e.g., toxic shock syndrome and necrotizing fasciitis) and postinfectious complications (e.g., acute rheumatic fever, rheumatic heart disease, and acute poststreptococcal glomerulonephritis). In this paper, we outline the rationale for selecting an emm75 strain for initial use in a GAS pharyngitis CHIM in healthy adult volunteers, drawing on the findings of a broad characterization effort spanning molecular epidemiology, in vitro assays, whole-genome sequencing, and animal model studies.

G roup A Streptococcus (GAS; Streptococcus pyogenes) is a major contributor to global infection-related mortality and morbidity. It causes a diverse spectrum of human disease syndromes, from superficial infections (e.g., pharyngitis and impetigo) to invasive disease (e.g., necrotizing fasciitis and toxic shock syndrome) and autoimmune complications (acute rheumatic fever, rheumatic heart disease, and glomerulonephritis) (1). Development of a GAS vaccine has been impeded by scientific, regulatory, and commercial obstacles (2). Controlled human infection models (CHIM) are increasingly assuming an important role for vaccine development (3,4). Drawing on the record of historical CHIM studies that included 172 participants (5-7), a modern pharyngitis CHIM in healthy adult volunteers has been proposed as part of a reenergized global effort to accelerate GAS vaccine development (8). Selection of a thoroughly characterized strain is central for development of a GAS CHIM.
A successful CHIM requires that infection and/or symptomatic disease endpoints are reached reliably and safely and bear sufficient resemblance to a natural state to suggest generalizability. The diverse clinical and microbiological profile of GAS presents challenges for CHIM study design, especially strain selection. There are more than 200 different GAS emm types. This widely used classification system is based on one part of the gene encoding a single GAS antigen, the M protein. No other antigen has been as closely studied, and the concept of M protein type-specific immunity has been a cornerstone of GAS research. GAS is a highly adapted human pathogen, and the limitations of in vitro assays and animal models have been well described. After more than a century of research, fundamental aspects of pathogenesis and human immune protection against GAS remain unknown. These knowledge gaps are simultaneously an argument for building a CHIM and a source of uncertainty in conceiving its design.
A thorough and explicitly stated rationale for strain selection is an important step in minimizing potential harm to participants and maximizing scientific impact. We considered desirable characteristics in selecting an initial strain to establish a GAS pharyngitis CHIM and surveyed available collections for suitable strains, focusing on an emm75 strain (GAS M75 611024, termed M75) isolated in 2011 from the throat of a 5-year-old girl with acute symptomatic pharyngitis in Melbourne (Table 1; see also  Table S1 in the supplemental material) (9). In this paper, we present a multifaceted characterization of the preferred CHIM candidate M75 strain and compared it to three others: GAS M12 611025 (M12), an alternative challenge candidate; M1T1 5448 (5448), representative of the M1T1 clone recently responsible for most pharyngitis and invasive disease globally (10); and CDC SS-496 (SS-496), an M1 strain administered to 88 subjects in 1970s pharyngitis CHIM studies (5,7).

RESULTS
Growth in an animal-free medium. Compared to that in THY broth, no detrimental effect on growth of M75, M12, and 5448 was observed in the animal-free medium (Fig. 1A). Eight-hour growth curves for M75 clones tested after 7 days of repeated in vitro passage were similar to those of the nonpassaged parent (data not shown). Attachment properties. M75 had the highest adherence to D562 (75%) and HaCaT (81%) cells (Fig. 1B). M12 (53%) and 5448 (51%) were similarly adherent to D562 cells. M12 preferentially adhered to D562 over HaCaT cells (P ϭ 0.005), whereas M75 and M1T1 showed no preference. The affinity of M12 to pharyngeal over skin cells matches its designation as an A-C pattern strain, associated with throat tissue tropism (11). Invasiveness of M75 and M12 was low for both cell lines (Յ0.45%). Invasion by 5448 of HaCaT cells (10%) was greater than that for D562 cells (0.2%) (Fig. 1B).
Delivery characteristics and viability. The Dacron swab was considered most suitable for delivery of the challenge inoculum (Fig. S1). Mean broth uptake by Dacron (105 mg) and Rayon S (108 mg) swabs was comparable to that of cotton (129 mg), and  Table S1 for a detailed and referenced version of this table. ARF, acute rheumatic fever; APSGN, acute poststreptococcal glomerulonephritis; RHD, rheumatic heart disease.
Group A Streptococcus Human Challenge Strain uptake variance was lowest (7.7 mg) for the Dacron swab. Superior release was noted for the Dacron swab with a mean of 1.8 ϫ 10 3 CFU of M75 recovered after swab dipping (Fig. 2B). Recovery from M75 vials frozen for 4 months did not fall below 95% of the original inoculum at T ϭ 0. Adherence and invasion were similarly unaffected by storage (data not shown).  (Continued on next page) Group A Streptococcus Human Challenge Strain Antibiotic susceptibility. M75 was susceptible to all tested antibiotics, while M12 was resistant to macrolides and fluoroquinolones ( Table 2). All strains were susceptible to clindamycin, and inducible resistance was not detected.
Whole-genome sequencing and phylogenetic analyses. The complete genome of M75 611024 is comprised of a single chromosome of 1,852,894 bp ( Fig. 2A). M75 has the multilocus sequence type (MLST) ST150 and contains the emm75.0 allele and mrp24 and enn334 alleles, corresponding to the emm-like genes mrp and enn (P. Smeesters, personal communication, July 2018). Three putative prophage sequences were identified in M75 harboring the endonuclease streptodornase 3 (spd3), pyrogenic exotoxins speL and speM, and the endonuclease sdn. M75 shared a hypothetical ancestral relationship with a UK emm75 cluster (Fig. 2B), yet it represents a distinct evolving lineage, suggesting an ancestral relationship to modern-day ST150 emm75 clones.
One single-nucleotide polymorphism (SNP) was found for each of three M75 clones sequenced after 7 days of repeated in vitro passage compared to sequence of the nonpassaged parent strain. Each SNP was intergenic and different, suggestive of random mutations of unlikely functional consequence (data not shown).
M12 611025 belongs to MLST ST36 and carries the emm12.0 allele. It shares a high degree of genome conservation with other emm12 genome sequences, varying in prophage and integrative conjugative element content relative to the reference genomes HKU16 and MGAS9429 (Fig. 2C). Phylogenetic analysis alongside 141 extant emm12 isolates showed an evolutionary relationship with other modern ST36 strains ( Fig. 2D), including recent scarlet fever outbreak strains (13).
The historical challenge strain SS-496 shares a higher degree of genetic and evolutionary similarity with the ancestral M1 reference strain SF370 relative to the modern M1T1 strains MGAS5005 and 5448 ( Fig. 2E and F) (14). SS-496 contains the historical SF370-like purA to nadC genomic region encoding streptolysin O.
Virulence factors and vaccine antigens. M75, M12, and SS-496 carry genes for an array of adhesion and invasion factors common to many emm types (Table 3). M75 contains a frameshift mutation in the fibronectin binding protein Sfb1 within the FCT locus. M12 carries the streptococcal superantigen A (ssa) gene recently reported in  scarlet fever-associated isolates in China and the United Kingdom (15). M12 does not carry the multidrug-resistant integrative conjugative element ICE-emm12 or the ssacarrying prophage ⌽HKU.vir, linked to the emergence of scarlet fever clades (13). The virulence profile of SS-496 is similar to that of pre-1980 M1 strains such as SF370, with speH and speI exotoxins and the absence of the speA exotoxin typical of modern isolates such as 5448 (Table 2). M75, M12, SS-496, and 5448 all possess wild-type covRS and ropB two-component virulence regulators.
High carriage (Ͼ60%) of protein and peptide candidate vaccine antigens was observed for M75 and M12 using a homology-based genome approach (Table 4). T-pilus antigen Fibrinogen-binding protein Putative ABC transporter periplasmic binding protein Putative ABC transporter ATP-binding protein Collagen adhesion protein ✓ psaA Manganese-binding protein Trigger factor ✓ ✓ ✓ ✓ Human neutrophil killing assay. M75 was most susceptible to in vitro killing when incubated with human neutrophils, although killing was observed for all strains (Fig. 1D).

DISCUSSION
We have described the rationale for selecting M75 for initial use in a new GAS pharyngitis CHIM in healthy adults, including results of diverse preclinical studies assessing its fitness for purpose. For context and comparison, we have presented results for three other strains (M12, 5448, and SS-496).
M75 is compatible with critical protocol points: reliable growth in an animal-free medium, retention of growth and attachment properties after prolonged storage at Ϫ80°C, consistent delivery using a commercially available swab, and susceptibility to antibiotics used to treat GAS pharyngitis. M75 looks to have an acceptable virulence profile, with the capacity to cause pharyngitis and low potential for invasive disease. M75 has attractive attachment properties for immortalized human pharyngeal and skin cell lines, with limited cellular invasion. M75 was highly susceptible to in vitro killing by human neutrophils, possibly due to its minimal capsular HA production. In a humanized mouse model of invasive infection M75 was avirulent, whereas M12 and both M1 strains were lethal. Whole-genome sequencing placed the strains in the context of epidemiologically related phylotypes and found broad representation of candidate vaccine antigens and a relatively restricted array of virulence factor genes in M75.
CHIM strain selection has been guided by varied general and pathogen-specific considerations, all with the goal of safely and reliably reproducing relevant and generalizable asymptomatic (infection/carriage) or symptomatic (disease) study endpoints (3,4,16). Suitable well-characterized strains may already exist, such as the Salmonella enterica serovar Typhi Quailes strain (17). Patients with mild to moderately severe uncomplicated disease may be a source of naturally attenuated new strains. Multiple strains, sometimes from different locations, may be required to represent natural strain diversity and/or enable heterologous rechallenge (18). If mechanisms of severe infection and/or complications are known, the implicated virulence factor(s) may be avoided (e.g., Shiga toxin-producing Escherichia coli [19] and Campylobacter jejuni inducing cross-reactive antibodies to GM1 and GQ1b gangliosides [20]). Pathogens may be modified for use, such as the propagation of single-sex Schistosoma mansoni cercariae to prevent chronic schistosomiasis (21). For vaccine studies, target antigen(s) must be present in the challenge strain(s). In every instance, strains must be characterized and be compatible with protocols for manufacturing and inoculation and with techniques to measure organism and host responses. The limitations of this characterization effort are inherent in the rationale for pursuing a GAS CHIM. In vitro assays, genomics, and animal models do not fully capture or predict the dynamic elements and sequelae of human infection by GAS, a highly adapted and human-restricted pathogen. Even advanced nonhuman primate models produce a pharyngitis syndrome with important differences from human disease. A single contemporary clone, represented here by 5448, is simultaneously the most common cause in urbanized settings of both the mildest and most severe disease syndromes, with the basis for tissue tropism and bacterial-human genotype-phenotype relationships still relatively obscure. These uncertainties dictate a cautious approach extending beyond strain selection, including strain manufacture following principles of Good Manufacturing Practice, initial inclusion of healthy adults only without risk factors for severe GAS disease, a dose-ranging study to establish attack rate and safety, inpatient admission at a trials facility supported by a tertiary hospital, universal antibiotic treatment, outpatient follow-up, and echocardiography at screening and final visits.
A generic limitation of CHIM studies is the uncertain degree to which data from healthy adults experiencing a single syndrome (pharyngitis) caused by one strain (M75) can be generalized to other subjects, syndromes, strains, and settings (e.g., children with GAS skin infections due to other emm types in low-and middle-income countries). While inclusion of other strains and even a skin infection CHIM are conceivable extensions, model findings must be interpreted alongside knowledge derived from more naturalistic studies. For vaccine development, a GAS pharyngitis CHIM has dual scientific and strategic purposes, aiming to serve as a bridge to field trials with a more natural distribution of subjects, syndromes, and strains.
With a view to the very high priority given to participant safety and risk minimization, findings from these strain characterization studies reinforce the appropriateness of M75 for initial use in a GAS pharyngitis CHIM.

MATERIALS AND METHODS
Bacterial isolates. M75 611024 and M12 611025 were isolated in 2011 from throat swabs collected from children with acute pharyngitis in Melbourne, Australia, and stored at the Murdoch Children's Research Institute (9). Mark Walker at the University of Queensland supplied 5448 (10,22). The U.S. Centers for Disease Control and Prevention (CDC) Streptococcus Laboratory provided the SS-496 strain, submitted in 1958 from Duke University.
Growth and viability. For administration to human volunteers, an animal-free medium must sustain sufficient strain growth. A chemically defined medium was developed (VR broth) consisting of RPMI 1640 (Gibco) and 2% (wt/vol) Veggietone genetically modified organism-free soya-peptone (Oxoid). Eighthour growth assays were done comparing growth of M75 in this medium to that in Todd-Hewitt broth (Oxoid) with 1% (wt/vol) yeast extract (Bacto) (THY). Bacteria were grown in 125-ml Erlenmeyer flasks containing 25 ml of VR or THY broth and agitated gently at 75 rpm. To simulate manufacturing processes, M75 growth in VR broth was examined after 7 days of repeated in vitro passage, using frozen cultures of three postpassage clones and the prepassage parent isolate.
HA capsule assay. The hyaluronic acid (HA) capsule is a GAS virulence factor that resists opsonophagocytosis (23). Capsular HA levels were quantified using a test kit (Corgenix), as previously described (10). Attachment properties. Cell culture lines have been used to study GAS adherence (24,25). We used Detroit 562 (D562) human pharyngeal cells and skin HaCaT cells, simulating natural sites of infection. As previously described, total cell-associated GAS (percentage of original inoculum) and invasiveness (intracellular fraction of total cell-associated GAS) were determined using GAS grown to mid-exponential phase (optical density at 600 nm [OD 600 ] of ϳ0.5) in VR broth and diluted in 500 l of assay medium (MEM with 5% fetal bovine serum; Gibco) to a multiplicity of infection of 5:1 (GAS:cells) (10). Inoculated trays were centrifuged for 5 min at 200 ϫ g and incubated for 1 h at 37°C in 5% CO 2 and then washed three times with phosphate-buffered saline (PBS) to remove nonadherent bacteria. Cell-associated GAS (adherent plus invasive) were detached using 200 l 0.25% trypsin, lysed with 0.025% Triton-X (Sigma) in distilled water, and enumerated by track dilution on horse blood agar. To measure invasive bacteria, cells were washed once after incubation in assay medium for 1 h and then incubated for another hour in medium containing 100 g/ml gentamicin and enumerated as before.
Delivery characteristics. To assess M75 viability following storage at Ϫ80°C, bacteria were grown in VR broth (OD 600 of 0.5), centrifuged, and suspended in broth containing 10% (vol/vol) glycerol (Sanofi). Vials containing 10 5 , 10 6 , 10 7 , and 10 8 CFU/ml were thawed at intervals and immediately tested without washing (mimicking the challenge protocol) for (i) growth in solid and liquid media, (ii) viability by enumeration, and (iii) attachment properties, as described above.
For the challenge procedure, swab uptake and release of the GAS inoculum should be consistent. We simulated direct oropharyngeal application using four Copan swabs: FLOQSwab (nylon), Dacron (polyester), and small (S) and large (L) rayon swabs. For uptake, vials of broth were weighed before and after dipping of swabs for 10 s. Swab release of GAS was measured by dipping swabs in 1-ml vials containing 1 ϫ 10 5 to 3 ϫ 10 5 CFU of M75 for 10 s, followed by transferring to 1 ml of PBS for 10 s and then enumerated by spread plate dilutions.
Antibiotic susceptibility testing. MICs were determined by Etest, and double disk diffusion (d-zone test) was used to detect inducible clindamycin resistance. Interpretive breakpoints of the Clinical and Laboratory Standards Institute (CLSI; penicillin, erythromycin, azithromycin, clindamycin, and levofloxacin) and the European Committee on Antimicrobial Susceptibility Testing (EUCAST; rifampin) were used (26,27).
Whole-genome sequencing and phylogenetic analyses. The complete M75 611024 genome sequence was determined using long-read single-molecule real-time sequencing on the Pacific Biosciences RS II platform. Filtering of the long reads identified 104,694 reads with an average polymerase read length of 4.1 kb. A single circular assembly was generated using SMRT analysis, v2.3.0 (Pacific Biosciences), and HGAP, v3, and polished using Quiver at an average read depth of 96-fold. To aid in assembly validation, M75 was also sequenced on an Illumina Next-seq 500 to produce paired-end reads with a read length of 150 bases. The M75 611024 genome sequence has been submitted to GenBank (accession number CP033621). The genomes of M12 611025 and CDC SS-496 were sequenced by Illumina Next-seq 500 with a paired-end read length of 150 bases. Draft genome assemblies were generated using SPAdes v3.12.0. Illumina short reads of M12 611025 (accession number SRR8217179) and CDC SS-496 (SRR8217180) have been submitted to the Short Read Archive (PRJNA504701).
To study M75 genomic stability, three clones were sequenced by Illumina Next-seq 500, with 150-bp paired-end reads, after 7 days of repeated in vitro passage. These sequences were aligned with the prepassage parent M75 reference sequence to identify single-nucleotide polymorphisms (SNPs).
Phylogenetic analysis of a global data set of emm75 isolates was determined by mapping short read sequences of 131 global emm75 genomes from the United Kingdom (n ϭ 124), United States (n ϭ 4), and France (n ϭ 3) (28)(29)(30) to the M75 611024 reference genome with BWA MEM (v0.7.16). SNPs with a Phred quality score of Ն30 were identified in each isolate using SAMtools pileup with a minimum coverage of 30ϫ. Prophage sequences within M75 611024 were identified using the Phaster server, with SNPs located within these prophage excluded, as they represent evolutionary confounders. A maximum likelihood phylogenetic tree was built from 1,046 concatenated SNP sites using RAxML, v8.2.8, with the general time-reversible model and gamma correction with 100 bootstrap resamplings to assess phylogenetic support.
Phylogeny of 141 emm12 genomes, including the Illumina reads of M12 611025 and sequences from Australia, the United States, Hong Kong, and mainland China, was analyzed by mapping to the reference genome HKU16 (strain QMH11M0907901 [GenBank accession no. AFRY01000001]) from 1,452 vertically inherited SNPs as previously described (13). Illumina reads of the M1 genome sequence CDC SS-496 were mapped to MGAS5005 (GenBank accession no. NC_007297) and other M1 reference genomes with phylogeny inferred form 780 vertically inherited SNPs.
Virulence factors and vaccine antigens. Virulence gene carriage was determined for M75, M12, 5448, and SS-496 by blastN screening assemblies against the virulence factor database (VFDB) (31). Gene presence was defined by an 80% nucleotide cutoff over 80% of the gene length.
Protection in animal models has been shown for more than twenty-five candidate GAS vaccine protein antigens and several peptide-based antigens (32). For protein antigens, presence was defined by an 80% nucleotide cutoff over 80% of the gene length. For sequence-constrained peptide-based vaccine epitopes J8.0 (SREAKKQVEKAL) (33) as well as the StreptInCor sequence (KGLRRDLDASREAKKQLEAEQQ KLEEQNKISEASRKGLRRDLDASREAKKQVEKA) (34) and associated T-cell (KGLRRDLDASREAKKQLEAEQQ), B-cell (ASRKGLRRDLDASREAKKQVEKA), and common B-T-cell (KGLRRDLDASREAKKQ) epitopes, a 100% nucleotide sequence match was taken to define presence, although 100% homology may not be required to induce production of broadly cross-reactive antibodies and vaccine protection.
Neutrophil killing assay. Survival of M75, M12, and 5448 incubated with human neutrophils in vitro was assayed as previously described (35). Experiments were performed in triplicate using mid-exponential-phase GAS at a multiplicity of infection of 10:1. Differences in neutrophil survival were analyzed using 1-way analysis of variance (GraphPad Prism).
Ethics statement. Animal procedures followed the Australian Code for the Care and Use of Animals for Scientific Purposes and were approved by the University of Queensland Animal Ethics Committee (SCMB/140/16/NHMRC) (39). An initial dose-ranging CHIM study has been approved by The Alfred Hospital Ethics Committee (500/17) and is registered at ClinicalTrials.gov (NCT03361163).
Data availability. The M75 611024 genome sequence has been submitted to GenBank (accession number CP033621). Illumina short reads of M12 611025 (accession number SRR8217179) and CDC SS-496 (SRR8217180) have been submitted to the Short Read Archive (PRJNA504701). M75 was also sequenced on an Illumina Next-seq 500 to produce paired-end reads with a read length of 150 bases (accession number SRR8217178).