Study participants and sampling
Mother–child pairs during the first year of life
Mother–child pairs from an urban clinic, The Gambia, participating in a trial of maternal immunization with MenAfriVAC were included (NCT03746665). All samples were obtained from consenting mother–child pairs where the mother had been vaccinated with meningococcal serogroup A conjugate vaccine between 28 weeksʼ and 34 weeksʼ gestation. Participants were vaccinated between December 2018 and October 2019. All participants were included where a paired serum sample from mother at delivery and neonatal cord blood was available.
Household longitudinal cohort study
Participants in the SpyCATS household cohort study were included10,63. A total of 442 participants from 44 households were recruited and visited monthly, with an open cohort approach such that new household members were able to enroll at any monthly visit. At each monthly visit, all enrolled participants were swabbed to determine the presence of Group A beta-hemolytic streptococci (GABHS) in their normal skin and throat. Unscheduled visits took place at any time when participants reported skin sores or a sore throat (disease episodes) to the study team. Suspected disease sites were swabbed for GABHS. At enrollment, a DBS was taken from all participants and a blood sample for serum separation from all participants older than 2 years. A DBS was collected from all participants during each monthly visit and at any presentation with disease.
S.
pyogenes event definitions from the SpyCATS longitudinal cohort study
Disease events were defined as the presence of signs or symptoms of pharyngitis or pyoderma plus a positive culture for GABHS from the disease site. Carriage events were defined as the detection of GABHS from throat or skin swabs without symptoms or signs of disease. All GABHS were considered to be S. pyogenes. Symptomatic pharyngitis was characterized by the presence of a sore throat (or parental reporting of pharyngitis-like symptoms in children younger than 5 years) along with observable tonsillo-pharyngeal redness during examination. Symptomatic pyoderma was characterized by one or more purulent or crusted skin lesions.
We categorized events into two distinct types: response-focused events (RFEs) and protection-focused events (PFEs). This categorization was used to investigate the role of antibodies in both responding to and protecting against S. pyogenes events within the SpyCATS longitudinal cohort study.
RFEs were defined to study how the immune system responds to specific instances of disease or carriage. Disease events in this category were defined as the presence of symptomatic pharyngitis or pyoderma with a corresponding positive culture for GABHS. These events were considered as new events when no related symptomatic disease events occurred in individuals within the prior 42 days. Carriage events were classified based on a positive GABHS culture in the absence of symptoms. RFE carriage events could be defined only in absence of a disease event within the preceding 42 days or the following 14 days. This definition ensured that immune responses to a carriage event in this context were not influenced by overlapping or closely timed disease events.
PFEs were defined previously and were used to analyze risk factors for incident events, excluding weekly swabs from carriage incidence analysis and allowing simultaneous characterization of disease and carriage events10. We, therefore, used PFEs to explore how immune responses influence the incidence of future events. Disease events in the PFE category were defined similarly to those in RFEs but with a key difference: they could be recorded as long as they occurred at least 14 days after a previous event rather than requiring a 6-week gap. Carriage events were also defined as a positive GABHS culture at a monthly visit, in the absence of symptoms, provided there was no carriage at the two consecutive previous visits or no carriage at one preceding visit and the last positive had been more than 28 days earlier or the previous swab was positive but more than 42 days earlier. However, no exclusion was made on the basis of presence of disease, allowing simultaneous disease and carriage events to be defined.
Sample selection for IgG level measurement
For mother–child pairs, samples used were those from mothers at delivery, neonatal cord blood at delivery and samples from infants taken at 6 months and from between 9 months and 11 months of age. From the SpyCATS cohort, a baseline sample of serum (or DBS in children younger than 2 years) was selected from all participants. DBS samples selected around RFEs were taken from the closest sample 14 days prior to the event (pre-event), the time of the event (event) and the closest sample at least 14 days after the event (post-event) as well as 3 months and 6 months after the event where available. To describe IgG level changes around events, the absolute difference in log10-transformed titers between the pre-event and post-event sample were used. We identified a control group of exposed, uninfected household contacts. These individuals were in the same household as an index case, were sampled within 14 days of the index event, had no S. pyogenes event within 90 days of the index case and had a DBS sample taken on the day of the household event or within the 28 days prior. The control group also had a sample selected for testing between 14 days and 42 days from the index event where available. For the cross-sectional analysis of age-stratified IgG levels, a single measurement per participant was taken from baseline, providing there was no S. pyogenes disease event at the time. In the case that both serum and DBS were measured at a single timepoint, a geometric mean IgG level was taken from the two readings. Additionally, in participants younger than 2 years of age, where IgG levels were rising fastest, a DBS sample was tested from every timepoint collected in the study.
Sample preparation
DSB cards (Whatman Protein Saver) stored at −20 °C since collection were punched with four 6-mm punches and eluted in 630 μl of elution buffer. The elution buffer consisted of phosphate buffered saline (PBS) + 0.05% Tween 20 + 0.08% sodium azide at a ratio of 1:10 whole blood to buffer. The samples were first subjected to rotation for 30 minutes at 9×g, followed by overnight incubation at 4 °C. They were then further processed with an additional 15-minute rotation at 300 r.p.m.
IgG measurement in blood for conserved S. pyogenes antigens
Both serum and DBS samples were tested with a characterized Luminex 5-plex assay to measure IgG levels to the conserved S. pyogenes antigens GAC, SLO, SpyAD, SpyCEP and DNAseB28,29. Antigens were coupled to magnetic carboxylated microspheres using carbodiimide chemistry. For GAC, which is polysaccharide in nature, biotinylated GAC was coupled to streptavidin-coated beads. Serum or eluted DBS samples were diluted in PBS at dilutions ranging from 1:300 to 1:60,000 and incubated for 1 hour with 1,000 beads per region per well at room temperature for 60 minutes in the dark, with shaking at 750 r.p.m. After washing, 50 µl per well of R-phycoerythrin-conjugated AffiniPure goat anti-human IgG (F(ab′)2 fragment-specific) secondary antibody (Jackson ImmunoResearch) was applied at a 1:70 dilution in PBS and incubated for 30 minutes in the dark with shaking at 750 r.p.m. A standard curve of Privigen IVIG in three-fold dilutions from 1:990 was added to each plate alongside a single sample of pooled serum to act as a positive control. The same lot of IVIG was used throughout the study. Each plate also included two blank wells containing only PBS and microspheres. Samples were tested in single, at a starting concentration of 1:20,000, and retested at an alternative dilution if the MFI value for any antigen fell outside of limits of standard curve accuracy28. All serial DBS samples from the same individual were tested on the same assay plate. The geometric mean IgG level from all repeats falling within limits of standard curve accuracy was taken as the final IgG level.
Adaptation of Luminex 5-plex assay to measure IgG in DBS samples
It was demonstrated that using Luminex multiplex assays to measure IgG titers to S. pyogenes antigens from DBS samples performs consistently with their measurement in serum29. To confirm this observation, we performed side-by-side measurement on serum and eluted DBS samples from 18 heathy adult healthcare workers. The substudy of ‘COVID-19 Humoral ImmunE RespOnses in front-line health care workersʼ (COVID HERO study) to optimize immunoassay methods for measuring humoral responses to SARS-CoV-2 and other pathogens was approved by the UK Health Research Authority (IRAS 283461, amendment ref: NSA03). DBSs were eluted overnight in PBS + 0.05% Tween 20 + 0.08% sodium azide at a ratio of 1:10 whole blood to buffer (and, therefore, assuming a ratio of serum to buffer of 1:20). log10-transformed RLU ml−1 values were compared with the Spearman method. Very strong correlation of serum with eluted DBS samples was established (Supplementary Fig. 6)29.
M/emm type-specific IgG measurement in a multiplex assay
We further adapted the Luminex assay to measure IgG to 14 M/emm type-specific M peptides, selected as representative of different M/emm clusters21, along with five additional M peptides from the E3 M/emm cluster, which were most common in the study (Supplementary Table 2)10. Peptide sequences for the 50-mer M peptide N-terminal hypervariable regions, corresponding to cluster-representative peptides, were obtained commercially (ProteoGenix). Hypervariable region peptides were constructed with a biotinylated lysine amino acid added to the C terminal of the peptide. The 15 most common emm clusters, based on worldwide population data, were determined21. For each emm cluster, the most frequent emm type globally was selected as the cluster-representative M peptide. M12, from M/emm cluster A-C4, was not available for inclusion in the assay. Fourteen peptides were successfully manufactured and included in a cluster-representative 14-plex assay. Additionally, five M peptides, as well as the cluster-representative M44 from the E3 cluster (the most common emm cluster in our study)10, were included in a 6-plex E3 cluster-specific assay. Microspheres were conjugated to streptavidin-coupled MagPlex beads at a concentration of 5 μg per million beads28.
The assay was optimized to measure M protein antibodies at 1:2,500 dilution of samples. Cluster-representative 14-plex coupled beads were added to the diluted samples at 1,000 beads per region per well. The same incubation conditions were applied as described above. Standard material consisted of 25% IVIG (Gammanorm, Octagen), 25% pooled sera from n = 9 participants who experienced documented emm25, emm18 and emm113 events and 50% pooled human sera from the SpyCATS study final visit (n = 244). This combination was selected to enhance type-specific antibody detection for M peptides with low MFI in commercial IVIG preparations (collected in HICs). Each assay plate included a 10-step serial dilution of the standard, starting from 1:100 in PBS. An additional 6-plex assay including beads conjugated to six peptides from the E3 emm cluster was also employed on specific samples, under the same assay conditions as for cluster-representative M peptide IgG measurement. The cluster-representative 14-plex assay was performed on every sample selected for measurement in the study. The additional E3 6-plex assessment was performed only around events from the E3 emm cluster.
Specificity of multiplex assay to measure IgG in blood to M/emm type-specific hypervariable region peptides
Assay specificity was determined by competitive inhibition. 25 µl of assay standard material (Privigen IVIG supplemented with pooled human sera) was incubated with 25 µl of each non-biotinylated M peptide (acting as an inhibitor) at a final concentration of 2 µg ml−1 in PBS. Sample and inhibitor were incubated for 1 hour before proceeding with the assay as described above. Percentage inhibition, defined as ((MFI Control − MFI Inhibited sample) / MFI control) × 100, was calculated. Homologous inhibition (inhibition of the signal from the matched peptide-coupled bead to the inhibitor) and heterologous inhibition (inhibition of the signal from the unrelated peptide-coupled beads to the inhibitor) were both defined.
Functional immunoassays
Inhibition of SLO-induced hemolysis by sera
The ability of sera to inhibit SLO-induced hemolysis was assessed using a previously characterized assay35. Sera were serially diluted in DPBS in a two-fold series across seven steps, from a 1:10 dilution in a round-bottom 96-well plate. A 75-μl aliquot of each diluted serum sample was incubated with 75 μl of 2,400 U ml−1 SLO in 40 mM DTT. After a 30-minute incubation at room temperature, 50 μl of defibrinated rabbit blood, pre-washed and diluted 1:5 in DPBS, was added to each well. Control wells included IVIG (Privigen), a no-hemolysis control (red blood cells in assay buffer alone) and a maximum hemolysis control (SLO with red blood cells). The plates were then incubated for 30 minutes at 37 °C. After incubation, the plates were centrifuged at 1,000 r.p.m. for 5 minutes at room temperature, and absorbance of 100 μl of supernatant was measured at 540 nm in a flat-bottom 96-well plate. The half-maximal inhibitory concentration (IC50) value, representing the serum dilution required to inhibit 50% of hemolysis, was determined for each sample by plotting optical density against the log-transformed serum dilution and fitting a four-parameter logistic (4PL) curve.
Inhibition of SpyCEP-mediated IL-8 cleavage
The ability of sera to inhibit SpyCEP-induced cleavage of IL-8 was assessed using a previously characterized assay36. Sera were diluted three-fold to achieve 1:12.5 starting dilution in DPBS containing 0.5 mg ml−1 bovine serum albumin (BSA) (Sigma-Aldrich) and incubated with an equal volume of full-length enzymatically active SpyCEP, produced in-house at the GSK Vaccines Institute for Global Health, at a final concentration of 10 ng ml−1 for 5 minutes at 4 °C. Then, 50 μl of IL-8 diluted in DPBS with 0.5 mg ml−1 BSA was added at 20 ng ml−1, and plates were incubated for 2 hours at 37 °C. Controls for maximum IL-8 cleavage (SpyCEP and IL-8 without serum) and minimum IL-8 cleavage (IL-8 only) were included on each plate, alongside a control sample of IVIG (Privigen) for each experiment.
After incubation, uncleaved IL-8 levels were quantified using a sandwich ELISA with a Human IL-8 Immunoassay Kit (Life Technologies, KAC1301) according to the manufacturer’s protocol. Samples were diluted 1:20 in standard dilution buffer, and 100 μl of the diluted sample was added to the ELISA plates preloaded with 100 μl of incubation buffer. A five-point standard curve was generated using known IL-8 concentrations reconstituted in sterile water. After the addition of 50 μl of HRP-conjugated secondary antibody, plates were incubated for 2 hours at 700 r.p.m., protected from light, and washed three times with wash buffer. TMB substrate (100 μl) was added to each well, including four chromogenic blanks, and incubated for 15 minutes at room temperature in the dark. The reaction was terminated with 100 μl of stop solution.
Absorbance was measured at 450 nm, and the chromogenic blank readings were subtracted from each test condition. The optical density for each serum sample was plotted against the log-transformed dilution, and a 4PL curve was fitted using GraphPad Prism version 10 (GraphPad Software). The IC50 was determined as the serum dilution required to achieve an optical density halfway between the maximum and minimum IL-8 cleavage controls.
Measurement of phagocytosis of bacteria and antigen-labeled beads into THP-1 cells
The ability of sera to promote opsonophagocytosis of bacteria and antigen-coupled beads was investigated in a characterized assay37. THP-1 cells (American Type Culture Collection, TIB-202) were cultured in RPMI 1640 HEPES buffer (Thermo Fisher Scientific) supplemented with 20% FCS, 100 IU ml−1 penicillin and 0.1 mg ml−1 streptomycin and maintained at 37 °C with 5% CO2. For bacterial assay preparation, an M/emm1 S. pyogenes strain (051304) was cultured overnight static in Todd Hewitt broth (THB) at 37 °C with 5% CO2, washed and resuspended in 0.05 M (pH 9.6) carbonate buffer containing 0.1 mg ml−1 FITC. After a 30-minute incubation at room temperature with gentle agitation, bacteria were washed and resuspended in phagocytosis buffer (RPMI 1640 HEPES, 0.5% BSA) at an optical density at 600 nm (OD600) of 0.08 (~37.5 × 106 bacteria per milliliter). For antigen-coated bead preparations, FITC at 0.1 mg ml−1 was added to M-280 Streptavidin Dynabeads (Thermo Fisher Scientific), followed by SpyAD and biotinylated GAC antigens at a concentration of 10 μg per 3 × 108 beads and incubated overnight at 4 °C. Beads containing antigens were added at 37.5 × 106 beads per milliliter. Serum samples were tested in three-fold dilution from 1:50 (and repeated at 1:10 if required). 20 μl of sample was added to each well, along with IgG-depleted serum (Molecular Innovations) and Privigen IVIG as controls. Bacteria or antigen-coupled beads in 20 μl were then added to each well and incubated at 37 °C for 30 minutes at 750 r.p.m. THP-1 cells were harvested, washed and resuspended in phagocytosis buffer at 7.5 million cells per milliliter. A 10-μl aliquot was added to each well and incubated for 30 minutes at 37 °C. Phagocytosis was stopped on ice, and cells were fixed with Cytofix (BD Biosciences) for 30 minutes at 4 °C. Cells were then washed, resuspended in PBS and analyzed using an Accuri flow cytometer (BD Biosciences). The geometric mean FITC intensity within the THP-1 gate was determined (Supplementary Fig. 7), and IC50 values were calculated by fitting a 4PL curve to the data, with the lower constraint set to the geometric mean value of wells without serum plus one s.d.
Statistical analyses
To obtain relative IgG levels, a five-parameter logistic (5PL) curve was fitted to the blank-subtracted MFI values obtained for each standard curve point using Bio-Plex manager software. RLU ml−1 values for test samples were obtained by interpolating the blank-subtracted MFI values onto the 5PL curve, and then multiplying this value by the dilution factor. RLU ml−1 values were log10 transformed for statistical analysis. IC50 data during functional assays were produced in GraphPad Prism version 10 by fitting 4PL curves to data. In functional assay analyses, samples with an IC50 below the limit of quantification (LOQ) were assigned an IC50 value of half the LOQ as described in assay characterization35,36,37. All remaining statistical analysis was performed with R (version 4.4.0). Fractional polynomial models were applied to log10-transformed cross-sectional antibody data to determine the 2.5%, 50%, 80% and 97.5% centile for each antigen by age40,64.
After assessment of both IgG levels and absolute antibody level changes around RFEs with Q–Q plots, histograms and the Shapiro–-Wilk test, non-parametric tests were used. After additional testing for homoscedasticity by plotting residuals, the Pearson method was used to determine correlation coefficient with levels and absolute level changes within individuals, given large sample sizes. Functional immunoassay data and IgG levels to M peptides were assessed for correlation with the Spearman method. When IgG levels were compared between two groups, a Mann–Whitney U-test was used, corrected for testing across multiple antigens using the false discovery rate (FDR) method. For multiple comparisons between groups, Kruskall–Wallis followed by Dunn’s test with Bonferroni correction was used. Exploring the impact of age and event type on magnitude of absolute log10 IgG changes around events was done with a mixed-effects linear regression analysis accounting for individuals and households as random effects. P values (adjusted where appropriate) less than 0.05 were considered significant.
Protection associated with IgG level and culture-confirmed events (PFEs) was explored using both logistic regression models and the Anderson–Gill extension of Cox proportional hazards models10,65. IgG measurements included in analysis of protection from conserved antigens were from all baseline samples; all samples in participants aged younger than 2 years; samples before, during and after microbiologically confirmed S. pyogenes events in cases; and samples before and after a microbiologically confirmed event in household controls. IgG levels were assumed to remain constant between measurements. Mixed-effects logistic regression models were used to explore the association of IgG with an event at the next visit so long as the next visit occurred within 45 days, to account for the monthly sampling frame in the SpyCATS study. Random effects for individuals and households were used to account for repeated sampling from individuals and household structure within the study. P values less than 0.05 were considered significant. Models were constructed separately for each antigen given substantial collinearity between conserved antigens. To construct piecewise regression for protection mediated by anti-SLO, anti-SpyAD and anti-SpyCEP, the transition point at which proportion of events within 45 days above each IgG threshold began to diminish was identified visually for each antigen. Next, the AIC values for 0.1 log10 iterations of the transition point were compared to the AIC of non-piecewise logistic regression, ensuring that AIC values were at least 2 lower using piecewise regression, and to confirm the appropriate transition point. AIC values less than 2 were considered similar. In adjusted models, AIC values were used to choose the model to best explain the data and to justify the inclusion of IgG level above the transition point only. Final models selected through this method explored the association between event in the next 45 days and IgG level above the transition point, sex, age group and household size. To explore synergistic effects of antibodies demonstrated to have association with protection, we constructed models with each combination of IgG level above the identified transition points, with and without interaction terms, to select a final model with each antigen, without interaction terms and with covariates included considering AIC values less than 2 to be similar.
To establish putative 50% protective thresholds for IgG levels in blood to SLO, SpyAD and SpyCEP, the IgG level at which the probability of any event in the next 45 days was 50% compared to IgG level at the transition point was estimated using a 10-fold cross-validation approach. To assess whether simultaneous IgG responses above these thresholds to multiple antigens conferred additive protection, we derived a composite variable representing the number of antigen-specific IgG levels above their respective protective thresholds (range, 0–3) at each timepoint during follow-up. This variable was calculated using both the 50% and a more lenient 33% protective threshold definition, and its association with culture-confirmed S. pyogenes events was modeled using mixed-effects logistic regression. Time in the study with IgG levels above any given number (0–3) or protective thresholds was calculated using methodology described for follow-up in this study10.
To explore association between IgG levels to conserved antigens and protection using an orthogonal approach, the Andersen–Gill extension of the Cox proportional hazards model was used to explore the association of antibody level with incident events, as previously described10. Outcomes explored in this model included any incident culture-confirmed event as well as each PFE event category. Multivariable models were selected through AIC criteria assessment, including sex, age group and household size as covariates. Models accounted for both participant ID (for individual clustering allowing for recurrent event inclusion) and household ID (for household clustering).
For analysis of M/emm cluster protection, each culture-confirmed S. pyogenes event was M/emm typed10. Having identified the cases and household controls, we used the M/emm type of the event to allocate a cluster-related IgG z-score to both cases and controls in relation to each event. If a homologous M peptide measurement (matching the event M/emm type) was available, it was selected; otherwise, the cluster-homologous M peptide level (representing the M/emm cluster) was used. This measurement was defined as the cluster-related anti-M IgG z-score. The mean z-score to unrelated M peptides to each event was identified for each timepoint and compared to the cluster-related z-score with Pearson’s correlation. Models incorporating the composite unrelated z-score were compared to those incorporating the cluster-related z-score using AIC criteria. Cluster-related anti-M IgG from before, during and after the event in cases and before and after the event in household controls was used to assess for protection from any microbiologically confirmed event in mixed-effects logistic regression models as described above. Finally, using AIC criteria to select the best model, a mixed-effects logistic regression model was used to explore the association between any event in the next 45 days and IgG level above the transition point for conserved antigens, cluster-homologous anti-M IgG z-score, sex, age group and household size.
Ethics and inclusion statement
The studies received approval from the joint ethics committee of The Gambia Government/Medical Research Council and the London School of Hygiene & Tropical Medicine Research Ethics Committee (ref: 24005 and ref: 1585). Written informed consent was obtained from adult participants as well as from parents or guardians of participants younger than 18 years of age. Additionally, children aged 12–17 years provided assent. The studies are registered on ClinicalTrials.gov (NCT05117528 and NCT03746665).
This study was conducted in close partnership with Gambian researchers and local stakeholders throughout all stages of the research process. This study was co-designed with input from the Sukuta community, including consultation with the local Alkalo (village leader), the Officer in Charge of Sukuta Health Centre and the directorate of health services from the Gambian Ministry of Health. The study’s aims were based on a comprehensive review of the existing literature and health priorities in the region and were determined to be locally relevant. Relevant local and regional literature is cited where appropriate. All members of the Gambian research team who contributed substantially to the design, implementation, data generation or interpretation are included as co-authors. The study supported capacity building by transferring multiplex immunoassays to in-country laboratories at the Medical Research Council Unit The Gambia (MRCG), allowing the majority of data generation and analysis to be performed locally. Training and mentorship for local researchers were integral to the study, and two Gambian scientists, F.C. and O.C., were subsequently awarded PhD fellowships. A.J.K., E.P.A. and O.C. are members, or alumni, of the Wellcome CREATE PhD program, an equity-centred doctoral training initiative focused on fostering partner-led, inclusive global health research practices.
Equitable benefit sharing was considered at every stage of the research. There was no transfer of biological materials out of The Gambia without prior ethics committee approval and participant informed consent, in alignment with national governance and international principles, such as the Nagoya Protocol. To support ethical engagement with participants, preliminary findings from the study were shared with community members and local leaders through a culturally appropriate, locally delivered community engagement event.
Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.
