Introduction

Pneumonia is the leading cause of child mortality worldwide. There are approximately 650,000 deaths in children under five years of age annually, mostly in low- and middle-income countries (LMICs)1. Pneumococcal conjugate vaccines (PCVs) have been introduced in 166 countries, but not in some high-burden settings in Africa and Asia due to their high price2,3.

PCV introduction has resulted in a decline in pneumococcal carriage and invasive pneumococcal disease (IPD) caused by vaccine serotypes4. The World Health Organisation (WHO) recommends a three-dose PCV schedule, given either as a 2 + 1 schedule (two-dose primary series with a booster dose) or a 3 + 0 schedule (three-dose primary series with no booster dose)5. However, there is growing interest in transitioning to schedules with fewer doses (reduced-dose schedules) involving a one-dose primary series with a booster dose (1 + 1 schedule) when herd immunity has been established by a three-dose program5,6. Countries with existing PCV programs are moving towards reduced-dose schedules due to financial and health resource considerations. The United Kingdom (UK) was the first country to transition to a 1 + 1 PCV13 schedule in 2020, administered at three and 12 months of age7. The UK’s transition from a 2 + 1 to 1 + 1 schedule was followed by a 14% decrease in total incidence of IPD in England three years after vaccine transition8. Specifically, this reflected a 34% increase in children, mostly aged 2–5 years, and a 17% decrease in adults. There was no difference in breakthrough and vaccine failure rates in children who received the 2 + 1 and 1 + 1 schedules. However, there remains uncertainty regarding the immunogenicity and long-term protection against pneumococcal carriage and disease following the introduction of reduced-dose PCV schedules, particularly in high-burden settings.

Serotype-specific IgG concentrations and opsonophagocytic titres are considered the gold-standard by regulatory agencies for licensure of new PCVs or to demonstrate non-inferiority of different vaccines and schedules. As these immunogenicity measures correlate with protection against IPD, they are commonly used in vaccine trials since it is not feasible to have IPD as a primary endpoint. However, serum antibody concentrations typically peak one-month post-vaccination, before waning substantially by 12 months of age. This limits the ability to predict long-term protection against carriage and disease in the second year of life9,10,11. Moreover, the level of antibody needed to protect against pneumococcal carriage or mucosal (non-invasive) disease is not known but is believed to be higher than what is required for IPD12. In recent years, there has been considerable interest in cellular immunity, such as memory B cells (Bmem), as a more reliable marker of long-term protection following PCV administration. Evidence from a human challenge study suggested that Bmem levels, rather than IgG concentrations, at the time of pneumococcal exposure is associated with protection against carriage in adults13.

Three studies have assessed Bmem responses following three- or four-dose PCV schedules in infants and children. A study in the United States showed that a 3 + 1 PCV13 schedule induced similar Bmem responses post-dose two and three14. In the UK, a 2 + 1 PCV13 schedule resulted in increased Bmem levels with a PCV13 but not PCV10 booster at 12 months of age15. Another trial from the UK found that a PCV13 booster administered at 3.5 years of age, following a 3 + 0 infant PCV7/13 schedule, increased Bmem levels from low pre-boost levels16. However, there are very few studies which have assessed Bmem responses of reduced-dose schedules in LMICs, where the burden of childhood pneumonia is greatest17,18.

We conducted a randomised-controlled trial (RCT) to evaluate the immunogenicity of reduced-dose schedules of PCV10 and PCV13 in a PCV-naïve population in Viet Nam. In this analysis, we report the kinetics of serotype-specific Bmem responses in Vietnamese infants over a 12-month period following vaccination with a 0 + 1 or 1 + 1 reduced-dose schedule of PCV10 or PCV13.

Results

Participant demographics

A total of 2822 participants were screened between 8 March 2017 and 24 July 2018, with 2501 enrolled (Fig. 1). Approximately 400 participants were allocated into each of the four vaccine arms. Of these participants, approximately 200 participants from each arm were randomly allocated into the immunology subgroup, with approximately 50 participants providing peripheral blood mononuclear cell (PBMC) samples for analysis of Bmem levels at each of the four timepoints.

Fig. 1: Participant flow chart.
figure 1

Blood samples from immunology subgroup participants were collected at 12 months (12 m), 12 months + 7 days (12 m + 7d), 12 months + 28 days (12 m + 28d), and 24 months (24 m).

Full size image

Baseline characteristics were similar between the vaccine arms (Table 1), although there were small differences in sex, type of delivery, cigarette smoker in the house, and breastfeeding status at some timepoints (Supplementary Table 1). Just over half of participants were male and the majority were from district 4, followed by districts 8 and 7. Most participants were born in hospitals (95.7%) and through vaginal delivery (50.5%). The majority of participants had cigarette smokers in their house (60.2%) and were breastfed (81.9%). Baseline characteristics of participants in the immunology subgroup were similar to that of the full trial population (Supplementary Table 2).

Table 1 Baseline characteristics of participants in the immunology subgroup
Full size table

Serotype-specific Bmem levels at seven and 28 days post-vaccination

For each vaccine arm, we calculated the serotype-specific Bmem levels prior to vaccination at 12 months (12 m), and at seven days (12 m + 7d), 28 days (12 m + 28d) and 12 months after vaccination (24 m) (Fig. 2 and Supplementary Table 3). In the month following vaccination, Bmem levels were highest at 12 m + 7d for both PCV schedules (peak timepoint), although this was higher for the 1 + 1 schedules and more apparent for PCV13 than PCV10.

Fig. 2: Analysis of serotype-specific memory B cell (Bmem) levels over time, by vaccine arm.
figure 2

A Study overview including vaccine arms, study timepoints and laboratory assays, B serotype-specific Bmem for each vaccine arm. Serotypes 6A and 19A are only found in PCV13, while all other serotypes are shared between PCV10 and PCV13. Serotype-specific Bmem levels are presented as medians and interquartile ranges (IQRs) of antibody secreting cells (ASC) per million cultured peripheral blood mononuclear cells (PBMCs). Note the difference in y-axes scales for individual serotypes. N = 170 (PCV10 0 + 1), 166 (PCV13 0 + 1), 165 (PCV10 1 + 1), 168 (PCV13 1 + 1) (biological replicates); N = 166 (12 m), 165 (12 m + 7d), 168 (12 m + 28d), 170 (24 m) (biological replicates). Source data are provided as a Source Data file.

Full size image

In the 1 + 1 schedules, Bmem levels increased from 12 m to 12 m + 7d for all serotypes. At 12 m + 7d, Bmem levels were highest for serotype 1 (PCV10: median 12.3 ASCs per million cultured PBMCs [IQR: 3.8–40.0]; PCV13: 35.2 [20.0–56.3]). From 12 m to 12 m + 7d, there was a 4.9-fold (PCV10) and 28.2-fold (PCV13) rise for serotype 1. At 12 m + 28d, Bmem levels had decreased for all vaccine serotypes, although these were still above the baseline Bmem levels observed at 12 m. For PCV10, there was no Bmem response detected at 12 m + 7d for the non-PCV10 serotypes 6A and 19A, as expected, but a response was detected at 12 m + 28d for serotype 6A.

In the 0 + 1 schedules, Bmem levels increased from 12 m to 12 m + 7d for most serotypes, except for serotypes 6B, 23F, and 6A with PCV10. At 12 m + 7d, Bmem levels were highest for serotypes 1, 5, and 14 (1.7 [0–3.8] for all three serotypes) with PCV10, and serotype 1 (8.8 [3.8–17.1]) with PCV13. From 12 m to 12 m + 7d, there was a 4.0 and 2.0-fold rise for serotypes 1 and 14 (PCV10), and 20.8-fold rise for serotype 1 (PCV13). The increase for serotype 5 (PCV10) was from nil baseline levels. At 12 m + 28d, Bmem levels for most serotypes decreased or did not change, except for serotypes 23F, 1, and 19A (PCV10) which increased by 2.3, 2.2, and 1.5-fold, and serotypes 6B and 6A (PCV10) which increased from nil baseline levels. Across all vaccine arms, there was a notable dip in Bmem levels for most serotypes at 12 m + 28d, especially with PCV13 and the 1 + 1 schedules.

Serotype-specific Bmem levels at 12 months post-vaccination

In the 1 + 1 schedules, Bmem levels at 24 m had waned for most serotypes, although they were still above the baseline 12 m levels observed. Bmem levels at 24 m were lower than at 12 m + 7d for all serotypes, except for non-PCV10 serotypes 6A and 19A with PCV10. In the 0 + 1 schedules, Bmem levels increased over time by 24 m for most serotypes, except for serotype 1 with PCV13.

Comparison of serotype-specific Bmem levels between PCVs and schedules

To compare PCV10 and PCV13 when given as a 0 + 1 or 1 + 1 schedule, we calculated the difference in medians of serotype-specific Bmem levels at 12 m + 7d (Fig. 3A and Supplementary Table 4). In the 0 + 1 schedules, Bmem levels for both PCVs were comparable for most serotypes, except shared serotypes 1 and 6B, which were higher with PCV13 (difference 7.1 ASCs per million cultured PBMCs [95% CI: 3.5–10.7] and 1.7 [0.1–3.2], respectively). For the non-PCV10 serotypes, Bmem levels were higher with PCV13 compared to PCV10 for serotype 19A (2.9 [0.9–4.9]), as expected, but not for serotype 6A. In the 1 + 1 schedule, Bmem levels were greater for PCV13 compared to PCV10, except for shared serotypes 6B and 14 that had comparable responses. Differences were greatest for serotype 1 (22.9 [6.3–39.6]) and PCV13-only serotype 19A (13.8 [6.8–20.7]). Serotype 6A also had a higher response with PCV13 compared to PCV10 (6.7 [2.1–11.2]), as expected for a non-PCV10 serotype.

Fig. 3: Analysis of serotype-specific memory B cell (Bmem) levels between pneumococcal conjugate vaccines (PCVs) and schedules.
figure 3

Comparison of serotype-specific memory B cell (Bmem) levels at 12 m + 7d between (A) PCV10 and PCV13 when given as a 0 + 1 (N = 336, biological replicates) or 1 + 1 schedule (N = 333, biological replicates) and between (B) 0 + 1 and 1 + 1 for either PCV10 (N = 335, biological replicates) or PCV13 (N = 334, biological replicates). Serotype-specific Bmem levels are presented as differences in medians and 95% CIs of antibody secreting cells (ASCs) per million cultured peripheral blood mononuclear cells (PBMCs). Source data are provided as a Source Data file.

Full size image

To compare the 0 + 1 and 1 + 1 schedules when given PCV10 or PCV13, we calculated the difference in medians of serotype-specific Bmem levels at 12 m + 7d (Fig. 3B and Supplementary Table 4). For PCV10, Bmem levels were higher with the 1 + 1 than the 0 + 1 schedule for serotypes 5 (5.8 [1.9–9.7]) and 6B (3.8 [1.0–6.5]), and comparable for all other serotypes. For PCV13, Bmem levels were higher with the 1 + 1 than the 0 + 1 schedule for all serotypes, with the greatest differences for serotypes 1 (26.5 [16.3–36.6]) and 5 (13.5 [8.1–19.0]).

Waning of serotype-specific Bmem levels and IgG concentrations

We compared the extent of waning of serotype-specific Bmem and IgG responses from the respective peak timepoints (day 7 for Bmem, day 28 for IgG) to 24 months of age for both PCV10 and PCV13 (Fig. 4 and Supplementary Table 5). Given the peak Bmem responses were more pronounced with the 1 + 1 than the 0 + 1 PCV schedule, we restricted this analysis to the 1 + 1 groups. IgG concentrations waned substantially for all vaccine serotypes, with 24 m levels 1.3 to 7.7-fold lower than at 12 m + 28d. By comparison, the degree of waning for Bmem was lower for both PCV10 (0.7–4.5-fold) and PCV13 (1.0–4.3-fold).

Fig. 4: Comparison of the relative waning of serotype-specific memory B cell (Bmem) levels and IgG concentrations from the respective peak timepoints to 24 m of age in the 1 + 1 PCV groups.
figure 4

The fold-decrease of Bmem levels and IgG concentrations between the two timepoints (12 m + 7d to 24 m for Bmem and 12 m + 28d to 24 m for IgG) for PCV10 and PCV13 are presented as geometric mean ratios (GMRs) of their levels/concentrations and 95% CIs. N = 85 (Bmem PCV10), 84 (Bmem PCV13), 220 (IgG PCV10), 219 (IgG PCV13) (biological replicates). Source data are provided as a Source Data file.

Full size image

Discussion

This is the first study to describe serotype-specific Bmem response kinetics over a 12-month period following a 0 + 1 or 1 + 1 infant schedule of PCV10 and PCV13. It is also the first study to evaluate Bmem responses in an LMIC setting as part of a large RCT. We found that Bmem levels peaked at seven days post-vaccination in contrast with IgG levels that peaked at 28 days post-vaccination. Importantly, Bmem levels were similar for both vaccines following a 0 + 1 schedule, but in a 1 + 1 schedule PCV13 induced higher Bmem levels than PCV10. When compared with serotype-specific IgG concentrations, Bmem levels did not wane as rapidly by 12 months post-vaccination with the 1 + 1 schedules for most serotypes.

Very few studies have assessed Bmem responses following a single infant dose of PCV and none of these studies assessed responses at the seven-day post-vaccination timepoint. A study of Kenyan toddlers aged 12-23 months found that a single dose of PCV10 elevated Bmem levels to serotypes 1 and 19F one month post-vaccination17. Our previous RCT in Viet Nam found that after a single dose of PCV10 at 18 months of age, Bmem responses to serotypes 1 and 18 C remained higher than unvaccinated children at six months post-vaccination18. In our current study, the 0 + 1 schedule resulted in stronger Bmem responses for serotypes 1, 6B, and 19A with PCV13 compared to PCV10. We also found that Bmem levels peaked at seven days following a 1 + 1 schedule for both PCV10 and PCV13. The few studies that have examined PCV Bmem responses in infants and young children have only studied the day 28 timepoint post-vaccination due to the feasibility of such studies. Serotype-specific IgG concentrations and opsonophagocytic titres are currently used for licensure of PCVs. There are established IgG correlates of protection for IPD but not for carriage or mucosal diseases including pneumonia. As such, there is a need to identify novel correlates of long-term protection following PCV use, especially for reduced-dose schedules. A three-dose primary PCV schedule without booster has been introduced into many LMICs. This schedule elicits high antibody titres during the first year of life and has led to major reductions in paediatric morbidity and mortality19,20. However, pneumococcal meningitis outbreaks continue to impact older children and adults in the African meningitis belt21, which suggests poor achievement of herd immunity. Therefore, high antibody concentrations may not translate to high vaccine efficacy in these settings11, which is further compounded by the lower efficacy of PCV in some African settings due to higher carriage rates and force of infection22.

Correlates of protection based on IgG response also differ between serotypes. Although the use of the 0.35 µg/mL correlate of protection has enabled the introduction of PCVs in many settings, serotype-specific correlates of protection could provide a more accurate prediction of protection against common serotypes23. Since antibody concentrations may not be a perfect indication of long-term protection, there is a need to assess Bmem and other novel markers of PCV immunity. We found that Bmem levels were strongest against serotype 1 in the PCV13 0 + 1 and PCV10/13 1 + 1 arms. This is of importance as serotype 1 is a leading cause of pneumococcal meningitis in children living the African meningitis belt24. Interestingly, however, the high Bmem levels for serotype 1 with the 0 + 1 schedules did not reflect the results from our opsonophagocytic assays which were low at 28 days post-vaccination9.

A unique aspect of our study was the head-to-head comparison of Bmem responses following PCV10 and PCV13 as part of a RCT. We observed that PCV13 elicited stronger Bmem levels in the 1 + 1 compared to the 0 + 1 arm for all serotypes. PCV10 elicited similar Bmem levels in the 0 + 1 and 1 + 1 arms, although levels for serotypes 5 and 6B were higher in the 1 + 1 arm. In the PCV13 1 + 1 arm, Bmem levels for all serotypes waned from peak timepoint (12 m + 7d) to 12 months post-vaccination (24 m), but did not return to baseline levels (12 m). A UK study found that a 2 + 1 schedule of PCV13 resulted in elevated Bmem levels for serotypes 1, 4, 9V, 3, and 19A one month post-vaccination, whereas no increases were seen with PCV1015. However, Bmem levels waned at 12 months post-vaccination in both arms, apart from serotypes 14 and 19A with PCV10 and serotype 3 with PCV13. Our large RCT was conducted in a setting with high carriage rates and no routine use of PCVs, and showed sustained Bmem levels over time.

It is hypothesised that Bmem may be a more useful marker of long-term protection as their levels do not significantly wane over time as seen with antibody concentrations. Existing data from adult challenge studies have shown the relationship between Bmem and carriage, but whether this is also seen in infants is not known13,25. Our study was able to compare waning of Bmem levels and IgG concentrations over time following vaccination in the 1 + 1 schedules. We found that for all serotypes, the extent of Bmem waning from peak timepoint to 12 months post-vaccination was not as pronounced as seen with IgG concentrations, and that Bmem levels remained detectable and did not return to baseline levels at 12 months post-vaccination. Our previous study in Viet Nam found that serotype-specific Bmem responses to a single dose of PCV10 persisted up to at least six months post-vaccination18. In our current study, Bmem responses had increased by 24 months of age in the 0 + 1 PCV schedules.

Our study found that Bmem levels generally peaked at seven days, before decreasing by 28 days, then increasing by 12 months post-vaccination. The increase at 12 months post-vaccination may suggest an immune response to carriage, which may have boosted Bmem levels by 24 months of age, particularly in the 0 + 1 groups. Although vaccine-type carriage decreased between 12 and 24 months of age in our RCT, both Bmem and IgG levels increased between these two timepoints9. We observed similar patterns of waning and subsequent elevation as a serosurveillance study from Malawi, which found that serotype-specific IgG concentrations waned in the nine months following a PCV13 3 + 0 schedule, but with subsequent elevation to five years of age26. It is likely that overall exposure to pneumococci during this period may not have been detected by our carriage outcomes as only two timepoints (18 m and 24 m) were measured. The elevated levels of Bmem and IgG at 24 months may have resulted from both vaccine response and boosting through carriage. However, while the 0 + 1 PCV schedules were immunogenic, they may not be sufficient to prevent carriage acquisition compared to the 1 + 1 PCV schedules. Assessing pneumococcal Bmem responses in the control group may have provided further insights into natural exposure. However, we were not able to do this since PBMCs were not collected in the control group as this arm was used to determine the efficacy of the PCV schedules against nasopharyngeal carriage9.

The introduction of PCV programs has led to a decline in vaccine-type pneumococcal carriage in many settings27. However, global coverage of the WHO-recommended three-dose PCV schedule was low even prior to the COVID-19 pandemic, with an estimated coverage of 48% in 201928. There is an even lower coverage in the Southeast Asian (23%) and Western Pacific (14%) regions28, where there is a high burden of pneumococcal disease29. Single-dose PCV schedules may be useful for catch-up campaigns in low-coverage settings or in humanitarian settings that face multiple barriers to immunisation. In this context, our data provides new insights into the long-term immunity and potential protection induced following a single dose of PCV.

Transitions to reduced-dose schedules require the maintenance of herd immunity through the control of vaccine-type carriage. The recent transition to a 1 + 1 PCV13 schedule in the UK was followed by an increase in IPD incidence in children aged 2–5 years, although overall incidence including adults, and differences in breakthrough and vaccine failure rates between the 1 + 1 and 2 + 1 schedules, were not significantly different pre- and post-transition8. The increase in IPD incidence was attributed primarily to serotype 3, but also 19F, 19A, and 4. While the immunological correlates of protection for IPD are defined, these are known to vary by serotype and are not available for pneumococcal carriage, which is a prerequisite for IPD. Early markers of pneumococcal immunity, such as Bmem responses, could be integrated with carriage data for association analysis. This could identify whether Bmem levels in the month following vaccination may be a more reliable marker of long-term protection against vaccine-type carriage acquisition or density at a later timepoint.

ELISpot is a conventional method used to enumerate antigen-specific Bmem30. This assay measures antibody secreting cells but other B cell populations such as plasmablasts could also contribute to this response, although it is recognised that plasmablasts secrete lower levels of antibody31. Detailed phenotyping of B cell populations by flow cytometry would be useful to provide comprehensive insights on the adaptive immune response following reduced-dose PCV schedules. There are very few studies using flow cytometry to assess Bmem responses following PCV administration in infants. We are currently conducting studies using PBMCs from this RCT, focusing on a high dimensional spectral flow cytometry assay to measure serotype-specific Bmem levels and other B cell populations, and their correlation with nasopharyngeal carriage acquisition by five years of age.

The main limitation of this analysis was that a different set of participants were tested for Bmem levels at each timepoint. Due to ethical considerations, we could not collect the larger-volume blood samples required for these tests from the same participant at all study visits. However, baseline characteristics of participants were similar across the four vaccine arms and between-group comparisons for Bmem levels were made at individual timepoints. We also note there were small differences in sex, type of delivery, cigarette smoker in the house, and breastfeeding status between the vaccine arms at some timepoints. Although difficult to determine, any potential impact on these results is likely to be minor given the similar overall Bmem response profile observed for the 1 + 1 and 0 + 1 schedules. We also could not measure all serotypes in PCV10/13 due to limited number of PBMCs, but to our knowledge, this study represents the most comprehensive range of serotypes examined in any PCV study. Additionally, we could only access circulating Bmem from blood samples and could not enumerate Bmem populations from other secondary lymphoid tissues.

In conclusion, this study reports serotype-specific Bmem responses for 0 + 1 and 1 + 1 reduced-dose schedules of PCV10 and PCV13. Compared with IgG concentrations, Bmem levels waned to a lesser degree from peak timepoint to 12 months post-vaccination. There is a need to study the use of Bmem and other novel markers of long-term protection against pneumococcal carriage and disease, especially in LMICs and ideally from RCTs. Integrating Bmem data with carriage data could shed light on whether Bmem responses shortly after vaccination can help predict carriage acquisition at a later timepoint. This information is crucial in the current era where reduced-dose schedules are being considered by LMICs to improve access to pneumococcal vaccines and protect infants at high risk of disease.

Methods

Study design

This study reports data from the Viet Nam Pneumococcal Trial II (VPT-II), which was a single-blind, open-label, phase II/III RCT evaluating reduced-dose schedules of PCV10 and PCV13 (trial registration number NCT03098628)32.

Participants

Infants born at 36 weeks of gestation or more, with no significant maternal or perinatal clinical histories, were recruited in community health clinics in districts 4, 7 and 8 of Ho Chi Minh City, Viet Nam. Full details on the eligibility criteria can be found in the published study protocol32. Written informed consent was provided by a parent or guardian. This analysis includes a subgroup of the main trial’s participants whose blood samples were collected to test for immunological endpoints.

Study procedures

In the main trial, participants were randomly assigned to receive PCV10 at 12 months of age (PCV10 0 + 1), PCV13 at 12 months of age (PCV13 0 + 1), PCV10 at 2 and 12 months of age (PCV10 1 + 1), PCV13 at 2 and 12 months of age (PCV13 1 + 1), or a control group (who received a dose of PCV10 at the final study visit at 24 months) in a 4:4:4:4:9 allocation ratio32. Randomisation was performed by an independent statistician using a computer-generated list in block-randomisation scheme (block size 25) and stratified by district. Allocation was concealed within sealed envelopes at the study clinics and participants were allocated by a study doctor by envelope sequence. Participants and those administering the intervention were not blinded, but laboratory staff were blinded during laboratory analysis.

For the immunological endpoints, the first 200 participants in each vaccine arm were selected. In each vaccine arm, 50 participants were randomly assigned to each of four subgroups (a, b, c, or d) which determined the timing of blood sample collection and laboratory assays to be conducted (Table 2). A relatively larger (7.5 mL) blood sample is required for the enumeration of Bmem. For ethical reasons, a different subgroup of participants provided this larger 7.5 mL blood sample at 12 months (12 m, pre-vaccination), seven days post-vaccination (12 m + 7d), 28 days post-vaccination (12 m + 28d), and at 24 months of age (24 m, 12 months post-vaccination). The analysis reported here does not include the control group as no bloods were collected at corresponding timepoints.

Table 2 Study design and timing of blood sample collection for serotype-specific memory B cell (Bmem) and IgG measurements
Full size table

Enumeration of pneumococcal serotype-specific memory B cells (Bmem) by ELISpot assay

The levels of Bmem (measured as antibody secreting cell [ASC]) were assessed for serotypes 1, 5, 6B, 14, 23F, 6A, and 19A using previously described methods33. This was the maximum number of serotypes able to be assessed due to the limited number of PBMCs collected and selection was based on a spectrum of disease-causing and commonly carried serotypes.

Freshly isolated PBMCs were resuspended at a concentration of 2 × 106 cells/mL in RPMI-EDTA plus 10% foetal calf serum (RPMI-FCS). One hundred microlitres (100 µL) of the suspension was added to each well along with 100 µL of an antigen cocktail, consisting of Staphylococcus aureus Cowan strain – Pansorbin cells (SAC; 1:5000, Calbiochem, San Diego, USA), 2.5 µg/mL CpG (Invivogen, San Diego, USA), and 83 ng/mL pokeweed mitogen (Sigma, St Louis, USA). Plates were incubated at 37 °C with 5% CO2 and 95% humidity for five days. Cells were then harvested using gentle resuspension and placed in a 30 mL tube filled with RPMI-FCS. Cells were washed in RPMI-FCS after centrifugation at 800 × g for 20 minutes and twice after centrifugation at 650 × g for 15 minutes. Cells were counted with trypan blue then resuspended in RPMI-FCS at a final concentration of 2 × 106 cells/mL. Cells were then seeded onto antigen-coated ELISpot plates.

Multiscreen hydrophobic polyvinyldene difluoride membrane ELISpot plates (Merck Millipore, USA) were coated with anti-IgG (10 µg/mL), or pneumococcal polysaccharides (BEI Resources, USA) conjugated to methylated human serum albumin, at concentrations between 10 to 20 µg/mL. Plates were sealed and incubated overnight at 4 °C. ELISpot plates were washed three times with phosphate buffered saline (PBS) and blocked with RPMI-FCS for 30 minutes at 37 °C with 5% CO2 and 95% humidity.

Cultured cells were seeded at 2 ×105 cells/well into the antigen-coated ELISpot plates and incubated overnight at 37 °C with 5% CO2 and 95% humidity. Cells were washed with PBS containing 0.05% (vol/vol) Tween 20 (PBS-T) and then incubated with an alkaline phosphatase-conjugated IgG (Calbiochem, San Diego, USA, Cat # 401442) for four hours at room temperature. Bound IgG was then detected. ELISpot plates were washed with PBS-T four times and alkaline phosphatase substrate solution (nitroblue tetrazolium plus 5-bromo-4-chloro-3-indoylphosphate in dimethyl formamide) was added to all wells to develop spots. The reaction was stopped with two washes in distilled water. An automated ELISpot reader and software were used to visualise and count cells. The total frequency of IgG-secreting ASCs was used as the positive control. Bmem levels were measured as ASCs per million cultured PBMCs. Samples with less than 1000 IgG ASCs per million cultured PBMCs (positive control reaction) were removed from analysis.

Serotype-specific IgG enzyme-linked immunosorbent assays (ELISA)

Serum samples were analysed for serotype-specific IgG concentrations to all 13 serotypes in PCV13 using a modified WHO ELISA as previously described34. Medium-binding ELISA plates were coated using purified pneumococcal polysaccharides (BEI Resources, USA) at 37 °C for five hours, then stored at 4 °C overnight. Serum and control samples were diluted to 1:100 in a preabsorption buffer of PBS with 10% (wt/vol) FCS containing cell-wall polysaccharide (CPS; 10 µg/mL) and serotype 22F (30 µg/mL) to remove non-specific antibodies and incubated overnight at 4 °C. ELISA for serotype 22F only preabsorbed serum samples with CPS. Plates were then washed with PBS-T, blocked with PBS/FCS, and incubated at 37 °C for one hour. The human anti-pneumococcal capsule reference serum 007Sp (National Institute for Biological Standards and Control, Potters Bar, Hertfordshire, England) was preabsorbed with CPS because the assigned serotype-specific IgG values were based on ELISA measurement that used preabsorption with CPS but not serotype 22F. After this incubation, serial dilutions of the preabsorbed 007Sp standard, serum samples, and controls were added to the ELISA plates and incubated at 37 °C for two hours. Plates were washed with PBS-T. Horseradish peroxidase-conjugated sheep anti-human IgG was added at 1:5000 (Southern Biotechnology, USA, Cat # 2040-05) and incubated at 37 °C for two hours. Plates were washed again with PBS-T. The reaction was developed by incubating with a 3.3’, 5.5’-tetrametylbenzidine substrate solution for nine minutes and stopped with 1 mol/L phosphoric acid. Optical density at 450 nm (630 nm reference filter) was measured with a microplate reader. Serotype-specific IgG concentrations for each sample were derived from the 007Sp standard values and expressed in micrograms per millilitre. Here we report IgG results for the seven serotypes with Bmem data (1, 5, 6B, 14, 23F, 6A, and 19A).

Statistical analysis

Outcomes were the level of serotype-specific Bmem (ASCs per million PBMCs) and concentration of serotype-specific IgG antibodies to serotypes 1, 5, 6B, 14, 23F, 6A, and 19A. Outcome data were aggregated by vaccine arm (PCV10 0 + 1, PCV13 0 + 1, PCV10 1 + 1, and PCV13 1 + 1) at each timepoint (12 m, 12 m + 7d, 12 m + 28d, and 24 m).

We used Stata/SE 18.0 for statistical analyses and to generate figures. Bmem levels were presented as medians and interquartile ranges due to data being non-normally distributed. To compare the peak Bmem levels between PCV10 and PCV13 and between 0 + 1 and 1 + 1 schedules, difference in medians and 95% confidence intervals of Bmem levels at 12 m + 7d were calculated using the Bonett-Price method. To compare Bmem levels and IgG concentrations at the respective peak timepoints (12 m + 7d for Bmem and 12 m + 28d for IgG as hypothesised a priori) to 24 m in the 1 + 1 schedules, we used linear regression with logarithmic transformed outcome variables (Bmem and IgG values). Regression models for IgG used cluster robust standard errors, as some participants provided samples at both timepoints. Coefficients were exponentiated to obtain geometric mean ratios and associated 95% confidence intervals. Between-group comparisons were made at individual timepoints (Bmem levels at each timepoint and difference in peak Bmem levels at 12 m + 7d), while comparisons of Bmem and IgG kinetics were made between different timepoints.

Ethics and inclusion statement

This trial was a collaboration between researchers from Viet Nam, Australia and Singapore, who were involved in the trial design, implementation, and interpretation, and included as authors on this manuscript. Parents or guardians of the participants provided written informed consent to enrol in the trial, which included the use of blood samples for this analysis. Ethics approvals were obtained from ethics committees in Australia (The Royal Children’s Hospital Melbourne Human Research Ethics Committee) and Viet Nam (Ministry of Health Ethics Committee).

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.