1. Immunology and Inflammation
Download icon

A delayed fractionated dose RTS,S AS01 vaccine regimen mediates protection via improved T follicular helper and B cell responses

  1. Suresh Pallikkuth
  2. Sidhartha Chaudhury
  3. Pinyi Lu
  4. Li Pan
  5. Erik Jongert
  6. Ulrike Wille-Reece
  7. Savita Pahwa  Is a corresponding author
  1. Department of Microbiology and Immunology, University of Miami Miller School of Medicine, United States
  2. Biotechnology HPC Software Applications Institute, Telemedicine and Advanced Technology Research Center, U.S. Army Medical Research and Development Command, United States
  3. Henry M. Jackson Foundation for the Advancement of Military Medicine, United States
  4. GSK Vaccine, Belgium
  5. PATH’s Malaria Vaccine Initiative, United States
Research Article
  • Cited 0
  • Views 638
  • Annotations
Cite this article as: eLife 2020;9:e51889 doi: 10.7554/eLife.51889

Abstract

Malaria-071, a controlled human malaria infection trial, demonstrated that administration of three doses of RTS,S/AS01 malaria vaccine given at one-month intervals was inferior to a delayed fractional dose (DFD) schedule (62.5% vs 86.7% protection, respectively). To investigate the underlying immunologic mechanism, we analyzed the B and T peripheral follicular helper cell (pTfh) responses. Here, we show that protection in both study arms was associated with early induction of functional IL-21-secreting circumsporozoite (CSP)-specific pTfh cells, together with induction of CSP-specific memory B cell responses after the second dose that persisted after the third dose. Data integration of key immunologic measures identified a subset of non-protected individuals in the standard (STD) vaccine arm who lost prior protective B cell responses after receiving the third vaccine dose. We conclude that the DFD regimen favors persistence of functional B cells after the third dose.

Introduction

Malaria is a communicable vector-borne disease caused by Plasmodium falciparum, a protozoan parasite that is transmitted to humans via Anopheles mosquitoes. The reported case incidence and associated deaths have declined globally over the years, but malaria is still a major threat to communities in affected areas. In 2017, malaria cases were reported to occur in 87 countries with an estimated 435,000 deaths of children under the age of five (WHO, 2018). Development of a protective vaccine has been challenging and currently there is no licensed malaria vaccine. Among vaccines in development, the RTS,S/AS01 vaccine is the most advanced and is now part of a large-scale pilot implementation program in children in selected African countries. This vaccine includes three main components: a) portions of the circumsporozoite protein (CSP) of Plasmodium falciparum, which is the primary surface antigen (Ag) on the sporozoites; b) hepatitis B surface antigen (HB); and c) a proprietary AS01B adjuvant from GlaxoSmithKline (GSK). The adjuvant is composed of 3-O-desacyl-4’-monophosphoryl lipid A (MPL) from Salmonella minnesota and a saponin molecule (QS-21) purified from an extract from the plant Quillaja saponaria, combined in a liposomal formulation consisting of dioleoyl phosphatidylcholine and cholesterol in phosphate-buffered saline solution (Vekemans et al., 2009). A vaccine delivery regimen in which three doses of RTS,S/AS01B are given on a 0, 1, 2 month schedule was found to provide nearly 50% protection in controlled human malaria infection (CHMI) trials (Kester et al., 2009; Ockenhouse et al., 2015). In the field, however, a Phase 3 trial of RTS,S/AS01 vaccine using the same dosing schedule showed only modest efficacy of 30.1% in 6–12-week-old infants (Agnandji et al., 2012).

In search of better efficacy, a recent CHMI trial of RTS,S/AS01 vaccine (Malaria-071, NCT01857869) was designed to compare the standard (STD) 0, 1, 2 month dosing regimen with a delayed fractional dose (DFD) regimen in which the third dose was one fifth of the standard dose and was administered at 7 months after the second dose (Regules et al., 2016). The rationale for testing the DFD regimen was based on a prior study from 1997 that had found that delaying and reducing the third immunization dose enhanced the immunogenicity of RTS,S vaccine and achieved efficacy to 86% (Stoute et al., 1997). Interestingly, Malaria-071 confirmed the earlier findings and the DFD regimen again showed 86% efficacy, which was significantly greater than the 62.5% protection observed in the STD regimen. To understand the factors associated with protection against Plasmodium infection, a number of immunologic investigations were conducted in a blinded manner.

The present study focused on assessing the role of peripheral T follicular helper (pTfh) cells and B cells, because a prior CHMI trial of the RTS,S vaccine had found an association of anti-CSP antibody titers and CSP-specific CD4 T cells with protection (White et al., 2013). In Malaria-071, antibodies of higher avidity were elicited in the DFD regimen in association with higher somatic hypermutation of B cells, suggesting fundamental changes in the maturation of B cell affinity (Regules et al., 2016). High-affinity antibodies are generated from long-lived plasma cells and memory B cells that are produced after antigen-primed B cells undergo cognate interaction with T follicular helper cells (Tfh). This interaction occurs in the germinal centers (GC) of secondary lymphoid organs (reviewed in Crotty, 2011), causing the B cells to proliferate followed by isotype switching and somatic hypermutation. Many properties of lymphoid Tfh cells, including B cell helper function for antibody (Ab) generation (Crotty, 2011), are also present in a subset of circulating CD4 T cells designated as pTfh cells that are considered as having emigrated from the lymphoid pool into the peripheral circulation (Vella et al., 2019; Pahwa, 2019). To investigate their role in human vaccine trials, pTfh in circulation serve as an attractive alternative to lymphoid Tfh, which require lymph node biopsies (Bentebibel et al., 2016; Bentebibel et al., 2013; Herati et al., 2017; Herati et al., 2014; Pallikkuth et al., 2017; Pallikkuth et al., 2012; Pallikkuth et al., 2019; Boswell et al., 2014; Cubas et al., 2013; Locci et al., 2013; Simpson et al., 2010; Ueno, 2016).

Only a few studies of pTfh have been performed in the context of immunogenicity and the efficacy of malaria vaccines. In a murine model, a nanoparticle-based vaccine presenting recombinant P. vivax CSP led to a protective immune response, characterized by enhanced GC formation with expansion and differentiation of antigen-specific Tfh cells (Moon et al., 2012). A malaria vaccine study in humans involving RTS,S/AS01 alone or co-administered with different viral-vectored vaccines showed that skewing of pTfh cells towards a CXC chemokine receptor 3 (CXCR3+) Th1 phenotype was associated with reduced Ab quantity and quality and lower vaccine efficacy (Bowyer et al., 2018). More recently, in a phase III trial of the GSK malaria vaccine ‘Mosquirix’ in Tanzania and Mozambique, children with increased frequencies of pTfh and plasmablasts at the time of vaccination exhibited higher Ab titers (Hill et al., 2020). An important role was ascribed to antigen-specific pTfh and their cytokine profile in influenza vaccine-induced antibody responses (Pallikkuth et al., 2019). In the present study, investigation of the dynamics of CSP-specific pTfh and B cell responses in the DFD and STD regimens of Malaria-071 (Regules et al., 2016), pre- and post-vaccination, revealed key immune features that were linked with protection after sporozoite challenge and provided insight into the superiority of the DFD regimen.

Results

CSP-specific pTfh responses are elevated in protected subjects

A scheme outlining vaccine timepoints and blood-sample collection for the immunological analyses is shown in Figure 1. Samples were analyzed at eight different timepoints, designated T0-T7: pre-vaccination (T0), day 6 post dose 1 (T1), day 28 post dose 1 (T2), day 6 post dose 2 (T3), day 28 post dose 2 (T4), day 6 post dose 3 (T5), day 21 post dose 3, pre-challenge (T6) and at study end, 159 days post-challenge (T7). The timing of the blood draws on day 6 and day 28 after each vaccine dose was designed to capture important periods for pTfh cell and B cell development post immunization. The distribution of protected (P)/non-protected (NP) participants was 10/6 in the STD regimen, and 26/4 in the DFD regimen (Regules et al., 2016). Given the small number of NP, we pooled data from both study regimens for each vaccine-induced immune response to understand the basis for protection.

Study schema and assay timepoints.

Timings of the first, second and third vaccine doses in either the standard dose regimen or the delayed fractional dose regimen are depicted in blue, green and yellow circles, respectively. Blood draws for immunology studies were performed at 8 timepoints designated T0 to T7: pre-vaccination (T0), day 6 (T1) and day 28 post first vaccination (T2), day 6 post second vaccination (T3), day 28 post second vaccination (T4), day 6 post third vaccination (T5), day 21 post third vaccination (T6, day of challenge) and at study end (T7, day 376; 159 days post-challenge).

Here, we analyzed the quantity and quality of CD4 T cells and pTfh cells to understand their role in vaccine-induced protection after RTS,S/AS01 vaccination. Circulating pTfh cells provide a snapshot of Tfh at the lymphoid inductive sites. Studies in healthy adults have documented the importance of pTfh expansion in response to vaccines as well as in the context of various infectious diseases (Bentebibel et al., 2016; Bentebibel et al., 2013; Herati et al., 2017; Herati et al., 2014; Pallikkuth et al., 2017; Pallikkuth et al., 2012; Pallikkuth et al., 2019; Boswell et al., 2014; Cubas et al., 2013; Locci et al., 2013; Simpson et al., 2010; Ueno, 2016). Data describing the frequencies of CSP-specific pTfh, along with total pTfh and CSP-specific CD4 T cells, in relation to P and NP status and the two vaccination regimens are shown in Figure 2. Detailed gating strategies for the identification of CD4 T cell subsets by flow cytometry are shown in Figure 2—figure supplement 1. We used CXCR5 expression on memory (CD45RO+CD27+) CD4 T cells to identify total pTfh cells. Expression of CD40L, an activation-induced molecule, was used to determine CSP-specific CD4 T cells after 12 hr in-vitro stimulation of peripheral blood mononuclear cells (PBMC) with a CSP peptide pool. These CD40L+ CD4 T cells were then gated for pTfh markers to determine antigen-specific pTfh (CD45RO+CXCR5+) andantigen-specific non-pTfh (CD45RO+CXCR5) cells.

Representative dot plots from P and NP subjects for total pTfh, CSP-specific CD4 T cells, and CSP-specific pTfh are shown in Figure 2A,B and C, respectively. Frequencies of total pTfh were greater at all the timepoints post-vaccination than at T0 in P subjects, and these frequencies showed sustained expansion after vaccination compared to NP subjects at T3, T4 T6 and T7 (Figure 2D). Frequencies of CSP-specific CD4 T cells were significantly increased at T2–T7 compared to those at T0 in P subjects (Figure 2E). CSP-specific pTfh cells showed a strong vaccine-induced expansion in P subjects and were more numerous at all the timepoints post-vaccination than at T0 (Figure 2F). Importantly, neither total pTfh, CSP-specific CD4 nor CSP-specific pTfh cells showed an increase post-vaccination in NP subjects. At T5, T6 and T7, the late timepoints after the two regimens diverged, the frequencies of total pTfh, CSP-specific CD4 and CSP-specific pTfh did not show differences between the STD and DFD regimens (Figure 2G, H and I, respectively), but the frequencies of total pTfh and CSP-specific CD4 T cells in the DFD regimen showed a decline at T7, the last time point that was evaluated (Figure 2G and H). Frequencies of CSP-specific non-pTfh did not show an increase from T0 following vaccination and did not differ between the two groups or the two regimens (Figure 2—figure supplement 2A and E).

Figure 2 with 2 supplements see all
Higher frequencies of total pTfh and CSP-specific CD4 and CSP-specific pTfh cell responses in protected subjects.

Frequencies of total pTfh, CSP-specific CD4 T cells and CSP-specific pTfh cells were identified by flow cytometry after 12 hr of PBMC stimulation with a CSP peptide pool in vaccinated subjects at different timepoints. Longitudinal data at different time points were analyzed for protected (P, n = 35) and non-protected (NP, n = 10) participants. (A–C) Flow cytometry dot plots for total pTfh cells, i.e. CD45RO+CD27+CXCR5+ cells gated from CD4 T cells (A); CSP-specific CD4 T cells, i.e. CD40L+ CD4 T cells (B); and CSP-specific pTfh cells, i.e. CD45RO+CXCR5+ cells gated from CD40L+ CD4 T cells (C). (D–F) Line graphs with error bars indicating mean ± standard error of mean (SEM) for protected (green line) and non-protected (red line) individuals showing frequencies of total pTfh cells (D), CD40L+CD4 T cells (E) and CSP-specific pTfh cells (F). (G–I) Scatter plots of CD4 T cell subsets in DFD and STD regimens at T5, T6 and T7 showing total pTfh cells (F), CSP-specific CD4 T cells (G) and CSP-specific pTfh cells (I) with data for the protected group represented by dark blue open circles for DFD (P DFD) and light blue open circles for the STD regimen (P STD), and the non-protected group represented by red open circles (NP) for both regimens. Statistical analysis was performed using the generalized linear mixed-effects model via Penalized Quasi-Likelihood to accommodate repeated measures over time. P values shown within the graphs refer to significant difference between the P and NP groups at the indicated time points. Statistical significance is shown as *p, <0.05; **, p<0.01; ***, p<0.001.

IL-21+ and ICOS+ pTfh subsets are associated with protection

To investigate the quality of vaccine-induced CD4 T cells in the context of protection, we analyzed (i) CSP antigen-induced intracellular IL-21 (Figure 3D), the signature Tfh cytokine; (ii) expression of inducible co-stimulatory molecule (ICOS) (Figure 3E), which is associated with the follicular recruitment, maintenance and function of Tfh cells, and (iii) Ki67 (Figure 3F), a marker indicative of cellular activation and proliferation. A significant increase in the frequencies (compared to those at T0) of IL-21-expressing (Figure 3G) and ICOS+ (Figure 3H) CSP-specific pTfh cells was evident in P subjects at all timepoints post vaccination and of Ki67+ (Figure 3I) CSP-specific pTfh cells at T3–T7. When comparing CSP-specific pTfh of P to NP subjects, frequencies of IL-21+ and ICOS+ cells showed an increase at T4, T6 and T7 (Figure 3D and E) and a transient increase of Ki67+ CSP-specific pTfh cells at T2 in P subjects (Figure 3F). In the NP subjects, no increase in IL-21+ or ICOS+ or in Ki67+ CSP-specific pTfh was noted post-vaccination, and the levels remained at background levels (Figure 3D, E and F), with ICOS+ cells dipping even lower at T7 (Figure 3E). ICOS+ total pTfh (CD45RO+CD27+CXCR5+ CD4 T cells) were present at higher frequencies in P subjects compared to baseline levels at T3-T7 and at higher frequencies than in NP subjects at T6 and T7 (not shown). IL-21+ total pTfh also showed a trend for higher frequencies in P subjects compared to NP subjects post-vaccination (not shown). Frequencies of CSP-specific IL-21+, ICOS+ and Ki67+ non-pTfh cells did not change post-vaccination and did not differ significantly between P and NP subjects at any timepoint (Figure 2—figure supplement 2B, C and D).

Higher induction of IL-21 and ICOS in CSP-specific pTfh cells from protected subjects.

Representative flow cytometry dot plots showing (A) IL-21, (B) ICOS and (C) Ki67 expression in CSP-specific pTfh cells in protected (P, n = 35) and non-protected (NP, n = 10) subjects. (D–F) Line graphs with error bars indicate mean ± standard error of mean (SEM) for frequencies of IL-21+ CSP-specific pTfh cells (D), ICOS+ CSP-specific pTfh (E) and Ki67+ CSP-specific pTfh (F) in the P (green line) and NP (red line) groups. (G–I) Scatter plots comparing IL-21(G), ICOS+(H) and Ki67+ CSP-specific pTfh (I) in DFD and STD regimens at T5, T6 and T7. Data from the protected group are represented by dark blue open circles for DFD (P DFD) and by light blue open circles for the STD regimen (P STD), whereas data from the non-protected group are shown as red open circles (NP) for both regimens. Statistical analysis was performed using the generalized linear mixed-effects model via the Penalized Quasi-Likelihood to accommodate repeated measurements over of time. P values shown within the graph refer to significant difference between the P and NP groups at the indicated timepoints. Statistical significance is shown as *p, <0.05; **, p<0.01; ***, p<0.001.

Comparing the two regimens, we noticed that IL-21- and ICOS-expressing CSP-specific pTfh were significantly more frequent at T7 in the DFD regimen, but did not show a difference in Ki67 expression (Figure 3G, H and I, respectively). In the STD regimen, the frequencies of IL-21+ CSP-specific pTfh decreased at T7 from those at T5 and T6 (Figure 3G). In the CSP specific non-pTfh compartment, frequencies of IL-21+, ICOS+ and Ki67+ non-pTfh cells did not differ between the DFD and STD regimens at T5, T6 or T7 (Figure 2—figure supplement 2F, G and H). Taken together, these data demonstrate that, as a group, P subjects show vaccination-induced expansion of both total and functional CSP-specific pTfh cells that respond to Ag stimulation with IL-21 production and ICOS and Ki67 expression, whereas NP subjects do not do so.

CSP-responsive B cells emerge after the second dose in the P subjects and are more frequent in the DFD regimen

To test the impact of vaccination on the B cell compartment, we first analyzed alterations in B cell maturation subsets ex vivo. The gating strategy for B cell subsets by flow cytometry is shown in Figure 4—figure supplement 1. Total B cells were identified as CD3CD20+ cells and total memory B cells as CD20+CD27+ cells. On the basis of the expression of CD21, CD27 and IgD, B cell maturation subsets were identified as naïve (CD21hiIgD+CD27), resting memory (RM, CD21hiCD27+), activated memory (AM, CD21lowCD27+) and atypical memory B cells (aMBC, CD21lowCD27). Further, on the basis of the expression of IgD and IgG, switch and unswitch memory B cell subsets were identified as total switch memory (SM), total unswitch memory (UM), switch RM (sRM), unswitch RM (uRM), switch AM (sAM), and unswitch AM (uAM) (Figure 5—figure supplement 1). Neither total B cells nor any of the ex-vivo-derived B cell maturation subsets differed significantly between P vs. NP subjects at any timepoint or between the two vaccine regimens at T5, T6 and T7 (not shown). The expression of CD80, a marker indicative of T-cell-dependent B cell activation, was analyzed ex vivo in total B cells, and in RM and AM subsets (Figure 4—figure supplement 2A), and did not differ significantly at any timepoint between P and NP subjects (Figure 4—figure supplement 2B, D and F, respectively), or between the DFD and STD regimens at the later timepoints (T5, T6 and T7) (Figure 4—figure supplement 2C, E and G, respectively). There was a trend for higher CD80 expression in AM B cells at T6 and T7 in P subjects (Figure 4—figure supplement 2F).

To assess the functional properties of B cells, we cultured PBMC with (i) full-length CSP protein (PF-CSP), (ii) the CS repeat region (R32LR) and (iii) the C-terminal peptide (PF-16) to clarify whether there was a region-specific dominant response to CSP in the B cell compartment. Examination of the B cell phenotype in the antigen-stimulated cultures was performed to assess memory B cell subsets by flow-cytometry and proliferation using Ki67. Changes in antigen-stimulated B cell phenotypes were noted mostly in relation to the regimen and not in the context of protection. Regimen-specific differences emerged post dose 3 (T5 or later) exclusively in the Ag-specific memory B cell compartment, including SM and sAM for PF-CSP (Figure 4A and B) and PF-16 (Figure 5D and E), with larger responses in the DFD arm compared to the STD arm. Ki67+ B cells were also more frequent following PF-CSP and PF-16 antigen stimulation at T5 and T6 in the DFD arm (Figure 5C and F). Frequencies of PF-CSP-specific Ki67+ aMBC B cells were also higher in the DFD arm at T5 and T6 (Figure 4—figure supplement 3). In the context of protection, although the levels of all subsets tended to be higher in P than in NP, none reached significance except for PF-16-stimulated Ki67+ memory B cells, which increased from T0 to T6 (Figure 4—figure supplement 4).

Figure 4 with 4 supplements see all
Frequencies of CSP-specific memory B cell subsets are greater in the DFD regimen than in the STD regimen at later time points.

PBMC were cultured for 5 days in the presence of PF-CSP and PF-16 antigens and analyzed for frequencies of B cell maturation subsets: switched memory (SM: CD20+CD10CD27+IgDIgG+), switched activated memory (sAM: CD20+CD10CD21lowCD27+IgDIgG+) and Ki67 expression on total memory (Mem: CD20+CD10CD27+) B cells in DFD (n = 29) and STD (n = 14) regimens at T5, T6 and T7. (A–C) The scatter plots show PF-CSP-specific SM (A); sAM (B) and Ki67+(C) memory B cells. (D–F) PF-16-specific SM (D); sAM (E) and Ki67+(F) memory B cells. Data for the protected group are represented as dark blue open circles for DFD (P DFD, n = 25) and as light blue open circles for STD regimen (P STD, n = 10) and non-protected as red open circles (NP, n = 10) for both regimens. Statistical analysis was performed using the generalized linear mixed-effects model via Penalized Quasi-Likelihood to accommodate repeated measurements over time. Statistical significance is shown as: *, p<0.05; **, p<0.01; ***, p<0.001.

Figure 5 with 1 supplement see all
Higher CSP-specific plasmablast and memory B cell ASC responses in the P subjects.

Spontaneous antibody secreting cells (ASC) at T0, T1, T3, and T5 were determined in unstimulated PBMC for plasmablast responses by ELISpot against PF-CSP, PF-16 and R32LR antigens. Memory B cell ASC responses were analyzed at T0, T2, T4, T5, T6 and T7 using ELISpot assay in PBMC stimulated with PF-CSP, PF-16, or R32LR antigens. Line graphs with error bars indicate mean ± standard error of mean (SEM). (A, B) Spontaneous ASC/million PBMC for the protected (P, green line, n = 33) and non-protected (NP, red line, n = 10) study groups for PF-CSP (A) and R32LR (B) antigens. (C–E) Memory B cell ASC/million PBMC for PF-CSP (C), PF-16 (D) and R32LR (E) for P and NP subjects. (F–H) Scatter plots showing memory B cells responses as ASC/million PBMC in the DFD and STD regimens at T5, T6 and T7 for PF-CSP (F), PF-16 (G) and R32LR (H) antigens. Data from the protected group are represented as dark blue open circles for DFD (P DFD) and as light blue open circles for the STD regimen (P STD), whereas data for the non-protected group are represented as red open circles (NP) for both regimens. Statistical analysis was performed using the generalized linear mixed-effects model via Penalized Quasi-Likelihood to accommodate repeated measurements over time. P values shown within the graph refer to significant difference between the P and NP groups at the indicated timepoints. Statistical significance shown as: *, p<0.05; **, p<0.01; ***, p<0.001.

Functional assessment of the vaccine-induced plasmablast and memory B cell antibody responses against the test antigens was conducted using antibody secreting cell (ASC) ELISpot assays. Plasmablasts are short-lived ASC that are generated rapidly in response to infection or vaccination, which transiently contribute to serum antibodies (Wrammert et al., 2008; George et al., 2015; Pallikkuth et al., 2011a; Rinaldi et al., 2017). To assess plasmablast responses, we determined the number of spontaneous IgG ASC directed at vaccine antigens on day 6 post-each vaccine dose as compared to the number pre-vaccination (Figure 5). A significant increase in the number of PF-CSP-specific (Figure 5A) and R32LR-specific (Figure 5B) spontaneous ASC were noted at day 6 post dose 2 (T3) in P subjects but not in NP subjects. The spontaneous ASC response did not differ significantly between the DFD and STD regimens at T5 (not shown).

Vaccine-induced antigen-specific IgG secreting memory B cell responses were analyzed at day 28 post-vaccination by memory B cell ELISpot assay following in vitro antigen stimulation (George et al., 2015; Rinaldi et al., 2017; Pallikkuth et al., 2011b). Memory B cells are mainly generated in the GC in secondary lymphoid organs. After leaving the GCs, memory B cells either join the recirculating pool of lymphocytes or home to antigen-draining sites. Kinetics, as well as the CSP epitope specificities of the vaccine-induced functional memory B cell responses, were analyzed (Figure 5). Comparing P to NP subjects, we found that only the P subjects showed an increase in memory B cell response to PF-CSP protein from T0 to T4, T6 and T7, and that at these time points, both the PF-CSP response and the response to the PF16 region were greater in P than in NP subjects (Figure 5C and D). The repeat region R32LR-specific memory B cell response also increased in P subjects only from T0 to T4 and T7, and was greater in the P group than in the NP group at T4 (Figure 5E). Comparing the regimens, in the DFD regimen the memory B cell response to PF-CSP increased from T5 to T6 and T7 (Figure 5F), and the response to PF-16 at T7 (Figure 5G) was larger under the DFD regimen than under the STD regimen.

As an additional measure of memory B cell function, we analyzed IgG secretion by ELISA in the PBMC culture supernatants after 5 days of stimulation with PF-CSP, PF-16, or R32LR antigens (Figure 5—figure supplement 1). Compared to T0, PF-16-specific IgG levels were significantly higher at T4, T6 and T7 in P subjects (Figure 5—figure supplement 1A), whereas PF-CSP- and R32LR-specific IgG did not change significantly (Figure 5—figure supplement 1C and E). IgG responses were significantly higher in the DFD regimen at T5–T7 for PF-16 (Figure 5—figure supplement 1B) and at T7 for R32LR (Figure 5—figure supplement 1F), with a trend of higher response at T7 for PF-CSP (Figure 5—figure supplement 1D), in comparison to responses in the STD regimen. Taken together, our findings indicate that RTS,S/AS01 vaccination elicited strong, functionally competent CSP-specific memory B cell responses in the P subjects, especially at the later timepoints, and that these responses were larger in the DFD regimen and stronger for PF-16 than for R32LR.

Data integration approach for identifying vaccine-induced immune correlates and their association with protection or regimen

In order to identify vaccine-induced immune correlates that are associated with protection and that differentiate the DFD regimen from the STD regimen, we employed a statistical data integration method. We incorporated data obtained for both CSP and HBs antigen-specific immune responses for this analysis, which include frequencies of memory B cell phenotypes, memory B cell ELISpot responses, CD4 and pTfh responses, and IgG levels from PBMC culture supernatants. We identified 676 of 1976 immune measures that were significantly increased from baseline (T0) to different timepoints post-vaccine (Supplementary file 1). By carrying out a correlation analysis to identify groups of correlated immune measures (‘immune clusters’), we were able to group these 1976 immune features into 142 immune clusters, of which 40 clusters had at least one vaccine-antigen-specific immune feature. Analysis of the vaccine-induced responses over the time course of the study revealed that the pTfh response was an early-stage response, emerging as early as T2, and persisting throughout the study. By contrast, the memory B cell response was a later-stage response, peaking between T4 and T5 (Figure 6—figure supplement 1A). 65% to 80% of the immune responses classified as ‘vaccine-induced’ were specific to the vaccine antigens CSP or HBs (Figure 6—figure supplement 1B), and the response was fairly balanced between both antigens.

Individualized predictions using machine learning

In order to assess the extent of regimen- and protection-level differences, we applied a machine-learning approach using random forest statistical modelling that could make individualized predictions of regimen and protection from immune data alone. A general workflow of the data integration approach is shown in (Figure 6—figure supplement 2). This approach allowed us 1) to determine what combination of immune features is most predictive of regimen or protection, and 2) to group subjects according to their pattern of vaccine-induced immune responses. Furthermore, by taking a prediction approach, we were able to determine how early in the vaccination regimen vaccine-induced immune responses would be predictive of protection. In order to assess predictive performance, we carried out a leave-one-out (LOO) analysis, in which each subject was excluded from the data set before the predictive model was trained on the remaining subjects, and then used to predict the outcome (or regimen) of that excluded subject. Accuracy was calculated as the proportion of subjects whose outcome (or regimen) was correctly predicted by the model.

In order to predict vaccine regimen from immune data alone, we performed a random forest analysis using 41 parameters from timepoints prior to challenge (T6) that were shown to be significantly different with respect to regimen in the univariate analysis. The LOO analysis shows that the random forest model, using these 41 parameters, achieved 85% accuracy with a kappa value of 0.63, indicating a strong predictive value. Overall, an average of 39 out of 46 subjects in the vaccine regimens were predicted correctly. Further, we determined the relative importance of each parameter in the random forest (Table 1) and found that antigen-induced B cell characteristics, including proliferation (Ki67+) and frequencies of SM, sAM, and Ki67 expressing aMBC, were most predictive of regimen. Nearly all predictive parameters showed antigen specificity for either CSP (66%) or HBs (27%). We used principal components analysis (PCA) to visualize how well the predictive parameters identified in Table 1 were able to distinguish subjects by regimen (Figure 6A). Overall, we found good separation between DFD and STD regimens using these parameters. We also found that the axis of variation within each regimen was distinct between the two groups, suggesting that these regimens are acting differently on this common set of immune parameters.

Figure 6 with 2 supplements see all
PCA plots showing regimen-specific and protection-status differences.

PCA plots using parameters identified by machine learning as being predictive of (A) regimen differences for DFD (purple) and STD (orange) subjects and (B) protection status for protected (blue) and non-protected (red) subjects.

Table 1
Parameters that are most predictive of vaccine regimen.
Cell typePhenotypeParameterWeight
B cellKi67+memory B cellsBCF.Mem.Ki67.HBs.T5100
BCF.Mem.Ki67.PF.CSP.T673
BCF.Mem.Ki67.MED.T655
BCF.Mem.Ki67.PF.CSP.T553
BCF.Mem.Ki67.PF.16.T645
sAMBCF.sAM.PF.16.T691
BCF.sAM.PF.CSP.T567
BCF.sAM.HBs.T653
SMBCF.SM.PF.CSP.T554
BCF.SM.HBs.T550
aMBCKi67BCF. aMBC.Ki67.PF.CSP.T554
BCF. aMBC.Ki67.PF.16.T548
BCF. aMBC.Ki67.HBs.T544
BCF. aMBC.Ki67.PF.CSP.T642
BCF.sAM.Ki67.PF.CSP.T540
  1. Abbreviations: sAM, switched activated memory; SM, switched memory; aMBCKi67, Ki67+atypical memory B cells; BCF, antigen-specific memory B cell responses by flow cytometry.

In order to predict protection status, we used 36 immune parameters that showed significant protection-level differences prior to challenge (T6 and earlier). We achieved a predictive accuracy of 85% with a kappa of 0.45, indicating low-to-moderate predictive ability, with 18 parameters in the model. Overall, 39 of 46 subjects were predicted correctly. The low sample size and imbalanced data set (78% of subjects were protected) made a more thorough assessment of the predictive ability of this model challenging. After analyzing for variable importance, we found that the parameters that are most predictive of protection (Table 2) include CSP-specific CD40L+ CD4, HBs-specific IL-21+, CSP-specific pTfh, frequencies of total pTfh cells, and CSP-specific antibody-secreting memory B cells. Of note, many of these parameters were from relatively early timepoints such as T2 and T4. We used PCA to visualize how well the predictive parameters identified in Table 2 could distinguish subjects by protection status (Figure 6B), and found that although there was a wide variability in the immune responses for P subjects, NP subjects clustered closely with each other and separately from P subjects. Together, these data suggest that there is a distinct pattern of immune responses associated with vaccine failure in this study.

Table 2
Parameters that are most predictive of protection.
Cell typePhenotypeParameterWeight
B cellMemory B cell ELISpotPF.CSP.T279
PF.16.T464
PF.CSP.T553
T cellCD40L+CD4CD40L.CSP.T4100
CD40L.CSP.T657
CD40L.CSP.T252
Ag.pTfhAg.pTfh.CSP.T271
Ag.pTfh.IL.21.HBs.T460
Ag.pTfh.IL.21.CSP.T656
Ag.pTfh.IL.21.HBs.T655
Ag.pTfh.IL.21.CSP.T449
Total. pTfhTotal.pTfh.MED.T249
Total.pTfh.MED.T443
Total.pTfh.MED.T640
Ag.CD4CSP.T446
CSP.T638
MED.T642
HBs.T452
  1. Abbreviations: Ag.pTfh, antigen specific peripheral T follicular helper cells; Ag.CD4, antigen stimulated total CD4 T cells; ELIspot, memory B cell ELISpot responses; CD40L+CD4, CD40L+ CD4 T cells.

Predicting protection from early-stage responses

Given that many of the immune correlates for protection were found at timepoints before dose 3, we used machine learning to determine whether we could predict if a subject could be protected by early-stage immune responses alone. We trained the model on early-response data alone (post-dose 1 and 2) to predict protection and achieved 87% accuracy with a kappa of 0.46, indicating moderate accuracy in predicting protection using only immune response data prior to dose 3. When we broke down the prediction results by vaccine regimen, we found that the protection status of virtually all DFD subjects is predicted correctly (97% accuracy, kappa = 0.84), whereas the protection status of STD subjects is predicted poorly (69% accuracy, Kappa = 0.26).

We stratified three classes of subjects: subjects whose early-stage immune responses were predictive of protection and who were actually protected, subjects whose early-stage immune responses were predictive of protection but who were not protected, and subjects whose early-stage immune responses predicted non-protection but who were, in fact, not protected. Interestingly, in both the STD regimen and the DFD regimen, approximately 15% of subjects (dark orange) elicited weak early-stage immune responses predictive of non-protection, and these subjects were subsequently found not to be protected following challenge.

In terms of subjects who elicited promising early-stage immune responses, we found that among DFD subjects, virtually all were in fact protected following dose 3 and challenge. By contrast, in the STD regimen, approximately one third of subjects with promising early-stage immune responses were not protected. These findings suggest that the third immunization in the STD regimen may adversely affect the immune response elicited by dose 1 and dose 2, and this may lead to the lack of protection.

On the basis of these individualized predictions of efficacy constructed on the early-stage immune response data, we were able to classify study participants into three groups of outcomes (Figure 7A): 1) ‘weak responders’, approximately 10–15% of subjects in both vaccine regimens (n = 6), who elicited poor early-stage immunogenicity and showed low efficacy (16% efficacy); 2) ‘DFD strong responders’, DFD subjects who showed promising early-stage responses (n = 27) and were almost entirely protected (96% efficacy); and 3) ‘STD strong responders’, STD subjects who showed strong early responses (n = 13) and achieved moderate protection (70% efficacy). To visualize these groups of outcomes, we generated a PCA plot using all parameters that were predictive of either regimen or protection status (Figure 7B). The information-related parameters that were predictive of protection in this model is shown in Supplementary file 1. Although there was some overlap between DFD responders and STD responders, the weak responders clustered close together, suggesting that they are markedly different from subjects that show promising early-stage immune responses. Finally, the difference in efficacy between the DFD and STD regimens seemed to be accounted for entirely by a subset of early strong responders that failed to achieve protection in the STD regimen.

Identification of subjects with promising early-stage immune responses.

(A) Comparison of STD and DFD subjects in terms of their predicted outcome from early-stage (pre-dose 3) immune response data and their actual outcomes. (B) PCA using immune parameters that are predictive of regimen and protection-status as determined by machine learning. Subjects are color-coded on the basis of their classification (on the basis of early-stage [pre-dose 3] immune data) as DFD responders (DFD subjects predicted to be protected), STD responders (STD subjects predicted to be protected), and weak responders (subjects predicted to be not protected). Protection rate is shown as the percentage of subjects in each group that was found to be protected in the study.

Discussion

Malaria is a leading cause of morbidity and mortality in endemic areas, underscoring the need for an effective vaccine. The RTS,S/AS01 vaccine is a promising candidate that has undergone extensive testing to define an optimal dosing and vaccine delivery strategy. In Malaria-071, a CHMI trial, participants in the DFD group who received a delayed and reduced third dose achieved 86% efficacy, which was significantly greater than the 62.5% protection attained under the STD regimen in which three vaccine doses are given at monthly intervals (Regules et al., 2016). To understand whether there was an immunologic basis to explain this difference, we conducted a study to examine T–B cell interactions in PBMC obtained from timed blood samples in the two regimens. For T cells, our focus was on delineating the dynamics of CSP antigen-specific pTfh cells, which were defined by phenotype and function. For B cells, we examined B cell maturation markers to define subsets and evaluated their function. A data integration approach was used to define correlates of vaccine-induced protection and non-protection. We found that protected subjects in both vaccine regimens were characterized by early induction of CSP antigen-specific pTfh responses, followed by functional memory B cell responses preceding the third dose that persisted at later time points. The non-protected subjects in both the DFD and STD regimens failed to mount the early pTfh response or B cell responses, pointing to the importance of pTfh in vaccine-induced protection. A key finding that provided insight into the inferiority of STD regimen was that in some NP subjects an initial ‘protective’ type of immune response was elicited by doses 1 and 2, but was aborted following the third vaccination dose. Understanding the mechanisms by which delaying and reducing the third antigen dose of RTS,S/AS01 after initial priming/boost helps to preserve the B cell immunity will lead to improved vaccination strategies. Our study was performed in a controlled setting with uninfected adult volunteers. Only field trials can tell us how such strategies will translate in endemic areas of the world where malaria exposure is rampant.

We noticed a clear increase in total pTfh cells and CSP-specific pTfh cells as early as 6 days post-first dose of RTS,S/AS01 vaccination that occurred only in participants who were protected following experimental challenge with P. falciparum. Circulating pTfh cells provide a snapshot of Tfh at the lymphoid inductive sites (Vella et al., 2019; Pahwa, 2019). Studies in healthy adults have documented the importance of pTfh expansion at day 7 or day 28 post-influenza and after other vaccines as well as bouts of various infectious diseases (Bentebibel et al., 2016; Bentebibel et al., 2013; Herati et al., 2017; Herati et al., 2014; Pallikkuth et al., 2017; Pallikkuth et al., 2012; Pallikkuth et al., 2019; Boswell et al., 2014; Cubas et al., 2013; Locci et al., 2013; Simpson et al., 2010; Ueno, 2016). The pTfh expansion was noted in both the DFD and STD regimens and was sustained throughout the vaccination schedule. This observation is reminiscent of the early induction of Ebola virus-specific pTfh following a single dose of the rVSV-Zaire Ebolavirus (ZEBOV) vaccine in an endemic population in Guinea, which was associated with protection (Farooq et al., 2016).

We identified pTfh phenotypically by CXCR5 expression in memory CD4 T cells. Cellular markers that have been used to define pTfh cells and their subsets have varied, with CXCR5 being a universally accepted receptor on these cells (Vella et al., 2019; Pahwa, 2019; Bentebibel et al., 2016; Herati et al., 2017; Locci et al., 2013; Pallikkuth et al., 2012; Heit et al., 2017; He et al., 2013; Morita et al., 2011; Moysi et al., 2018). Tfh cells that are located in the GC of secondary lymphoid organs (Vella et al., 2019; Pahwa, 2019; Herati et al., 2017; Herati et al., 2014; Heit et al., 2017) express high levels of PD-1. The frequency of PD-1+ cells is very low in circulating pTfh cells and the use of PD-1 as an essential marker that defines pTfh is likely to limit the frequency of identifiable pTfh. Moreover, the relevance of PD-1-expressing pTfh remains unclear because the molecule was found to be inhibitory for pTfh function in some studies (de Armas et al., 2017a). We have found that the expression of PD-1 in combination with that of the activation markers CD38 and HLADR in pTfh is inhibitory for their function in healthy volunteers given influenza vaccine (Pallikkuth et al., 2019). In light of this information, we opted to focus on the antigen-specificity of pTfh cells to mark functional cells. An important aspect of our analysis was the determination of CD40L expression, intracellular IL-21 production and ICOS upregulation in antigen-stimulated pTfh. Using these criteria, we observed higher CSP-specific pTfh in P subjects throughout the study, indicating a critical role of functional pTfh cells for protection against Plasmodium infection. IL-21 is the signature cytokine of Tfh cells and is required for optimal B cell function (Crotty, 2011; Linterman and Vinuesa, 2010; Vogelzang et al., 2008; Bryant et al., 2007; Pallikkuth and Pahwa, 2013). Likewise, ICOS–ICOSL interactions are known to be important for Tfh–B cell collaboration and also for IL-21 gene transcription in Tfh cells through c-Maf (Bossaller et al., 2006; Bauquet et al., 2009). These results add to the growing body of evidence pointing to the importance of IL-21 (Schultz et al., 2016; Spensieri et al., 2016; de Armas et al., 2017b) and ICOS (Bentebibel et al., 2016; Bentebibel et al., 2013; Herati et al., 2014; Havenar-Daughton et al., 2016) in circulating CD4 T cells or CSP-specific pTfh as biomarkers of vaccine responses. In the data integration analysis, neither the non-pTfh cells nor IFNγ+ CSP-specific CD4 and pTfh responses were identified as variables that were associated with protection or regimen difference, reaffirming the central role of antigen-specific pTfh cells in the CD4 T cell compartment in the response to vaccination. The actual timing, magnitude and duration of the pTfh response that needs to be elicited in order to generate qualitatively superior B cell responses for improved protection in the DFD regimen will require further investigations.

To understand the nature of the B cell response to the vaccine, we examined changes in specific subsets in the context of protection and regimen. The development of strong, functionally competent CSP-specific memory B cells after the second dose in protected subjects suggests the development of highly Ag-experienced functional memory B cells following the early pTfh response induced by vaccination. Development of a CSP-specific memory B cell compartment is in line with previous studies on antibody affinity, B cell somatic hypermutation, and antibody function, and suggests that affinity maturation is altered to some extent in the DFD regimen (Regules et al., 2016; Chaudhury et al., 2017). The stronger vaccine-induced memory B cell responses elicited towards PF-16 (the C-terminus of the CSP protein) over R32LR (the central repeat region) may have played a role in protection, as this region is implicated in the initial entry of the sporozoites into hepatocytes (Ramasamy, 1998). In our study, we found persistence of the vaccine-induced CSP-specific memory B cells up to 159 days post-challenge, the last time point for analysis. In a previous study, Pepper and colleagues reported persistence of Plasmodium-specific memory B cell populations that had been induced by protein immunization up to 340 days post-infection (Krishnamurty et al., 2016). Longer follow-up studies are needed to understand whether the vaccine-induced persistence of pTfh responses favors this B cell response in the DFD regimen. Interestingly, aMBC were also increased in the DFD regimen, compared to the STD regimen, at T5 and T6. These cells, originally described as an exhausted subset of memory B cells in HIV infection (Moir and Fauci, 2009; Moir et al., 2008), have been found in the circulation of Plasmodium-infected individuals from endemic countries (Ly and Hansen, 2019). A role for aMBC in malaria immunity has been suggested on the basis of the accumulation of this sub-population in situations of parasitemia or shorter exposure history (Weiss et al., 2009; Changrob et al., 2018) in children and adults from malaria endemic areas (Weiss et al., 2009; Changrob et al., 2018; Portugal et al., 2015); importantly, aMBC were maintained in situations of persistent parasite exposure (Ayieko et al., 2013) with a decline over 12 months in the absence of transmission. The functional significance of vaccine-induced aMBC expansion and the role of these cells in the development of immunity to malaria needs further investigation.

A major objective of the present study was to determine the mechanism behind the improved protection in the DFD vaccination regimen as compared to the STD regimen. As both the STD and DFD groups received the same regimen during the first two doses, differences between the study arms were expected only after the second dose, when the regimens split into either a reduced third dose at 7 months in the DFD group or a full third dose one month after the second dose in the STD group. To explore how the DFD regimen enhances efficacy, we used machine-learning tools and made individualized predictions of protection on the basis of early immune responses (pre-dose 3) alone. Using this analysis, we were able to identify a group of non-protected participants in the STD regimen who showed a promising immune response after doses 1 and 2, but lost this response after the third dose. These data suggest that dose 3 in the STD dose schedule has an adverse effect on an otherwise promising immune response generated by the first two doses. This effect could be caused by the Ag or adjuvant concentration of the full third dose, or by the one-month spacing between dose 2 and dose 3, which may hinder the selection of high-affinity B and T cell clones, either through overstimulation and anergy, or through weak selection pressure resulting from high Ag availability (Alexander-Miller et al., 1996).

Our data on ex vivo frequencies of dead cells did not differ significantly between the DFD and STD regimens immediately after (day 6) of the third dose (data not shown), and refute the possibility of a higher rate of cell death in the STD regimen. Our results suggest that NP subjects in the STD regimen are a mix of two classes of subjects, true weak responders and strong responders that aren’t protected. This mixed population may explain why it has been difficult to find correlates of protection in RTS,S studies (Ockenhouse et al., 2015). By contrast, in the new DFD regimen, NP subjects are almost entirely of one class — weak responders. On the basis of the overall findings, we suggest that the early Tfh response, induced by the initial two vaccine doses, results in the formation of a strong high-affinity memory B cell pool that is specific to CSP antigen; in the DFD, this leads to the expansion and differentiation of the pre-formed memory B cells to Ab-secreting cells. By contrast, in the STD vaccine regimen, early administration of the booster dose was detrimental to the expansion and differentiation of pre-formed memory B cells. Previous reports found an association of CSP-specific IL-2+, TNF or IFNg+ CD4 T cells or Th1 responses with vaccine responses (Kester et al., 2009; Lumsden et al., 2011). These studies did not examine CSP-specific pTfh within the memory CD4 T cell compartment, and most probably included both pTfh and non-pTfh cells in their analysis. Our data show that within the Ag-specific CD4 T cell compartment, pTfh cells but not non-pTfh show a kinetic and functional response to the malaria vaccine.

It should be noted that our study had limitations. Only a small number of participants (4/30) became infected in the DFD regimen, precluding comparisons of infected and protected subjects within each study regimen. We also did not examine CSP-specific pTfh for their Th1 versus Th2 phenotype or for response to individual CSP peptides in order to fine-map the pTfh responses. In a recent influenza vaccine study, we found that vaccine non-responders were polarized towards an inflammatory Th1/Th17 phenotype with predominant production of inflammatory cytokine TNF and Tfh antagonistic cytokine IL-2, while in responders, pTfh cells showed a Th2 phenotype with ICOS upregulation and IL-21 production (Pallikkuth et al., 2019). Another study has documented a negative impact of CXCR3+, a marker of Th1 type pTfh, on antibody quantity and quality in a vaccine trial involving RTS,S/AS01B (Bowyer et al., 2018). More detailed characterization of the functional and phenotypic heterogeneity of pTfh in future malaria vaccine studies may be informative. Further analyses are needed to ascertain the relationships of the immune parameters investigated herein with the magnitude and breadth of the Ab responses.

We conclude that delaying and reducing the third vaccine dose is advantageous for developing a protective immune response. This is highlighted particularly in those individuals that elicited promising responses after the first two doses, which then seemed to be disrupted when the third dose of RTS,S/AS01 was administered at the standard concentration one month after the second dose. We recognize that the CHMI studies of RTS,S-vaccinated malaria-naive adults represent a controlled setting for the study of immune response in relation to vaccine-induced protection. Whether the DFD regimen can be translated into the field with beneficial effects remains to be seen. The long interval between the second and third doses may be challenging in the face of overwhelming exposure to mosquitoes that can inoculate sporozoites into the host in this period, thereby affecting the development of immunity. However, the amount of CSP delivered via natural infection is much lower than the amount of CSP in RTS,S/AS01 and, thus, it is unlikely that natural infection would disrupt the development of an otherwise protective immune response. Recently, it was shown that children who responded well to RTS,S/AS01 vaccination had increased baseline frequencies of antibody-secreting and Tfh cells (Hill et al., 2020). Nevertheless, emerging data also suggests that malaria infection may induce memory Tfh cells that have impaired B cell helper function, and may inhibit differentiation to fully functional Tfh cells, thus resulting in germinal center dysfunction and suboptimal antibody responses (Hansen et al., 2017). Last, our data indicate the generation of strong CSP-specific pTfh responses that persist even after 159 days post-challenge, suggesting that CSP-specific pTfh could serve as potential biomarkers for vaccine efficacy. Monitoring CSP-specific pTfh should be considered in future malaria vaccine trials in clinical settings or field studies.

Materials and methods

Key resources table
Reagent type
(species) or
resource
DesignationSource or
reference
IdentifiersAdditional
information
Peptide, recombinant proteinStaphylococcal enterotoxin BList Biological laboratories# 1221 µg/ml, PBMC stimulation
Peptide, recombinant proteinCPG ODN 2016In-Vivogen# tlrl2006-11 µg/ml, PBMC stimulation
Peptide, recombinant proteinHBs peptide poolPATH’s Malaria Vaccine Initiative2 µg/ml, PBMC stimulation
Peptide, recombinant proteinCSP peptide poolPATH’s Malaria Vaccine Initiative2 µg/ml, PBMC stimulation
Peptide, recombinant proteinCS repeat region protein (R32LR)GSK2 µg/ml, PBMC stimulation
Peptide, recombinant proteinHBs proteinGSK2 µg/ml, PBMC stimulation
Peptide, recombinant proteinCSP proteinPATH’s Malaria Vaccine Initiative2 µg/ml, PBMC stimulation
Peptide, recombinant proteinC-terminal peptide (PF-16)PATH’s Malaria Vaccine Initiative2 µg/ml, PBMC stimulation
Biological sample (Homo-sapiens)Primary human mononuclear cellsGSKCryopreserved in liquid nitrogen
AntibodyAnti-human CD3 BUV 395, mouse monoclonal, Clone SK7BD BiosciencesRRID:AB_2744382; Cat# 5640015 µl/test, FACS
AntibodyAnti-human ICOS BV 421, regimenenian hamster monoclonal, Clone C398.4ABiolegendRRID:AB_2562545; Cat# 3135240.156 µl/test, FACS
AntibodyAnti-human CXCR5 Alexa 647, rat monoclonal, Clone RF8B2BD BiosciencesRRID:AB_2737606; Cat# 5581130.625 µl/test, FACS
AntibodyAnti-human CD8 Alexa 700, mouse monoclonal, Clone RPA-T8BD BiosciencesRRID:AB_10643765; Cat# 5614532.5 µl/test, FACS
AntibodyAnti-human CCR7 PE-CF594, mouse monoclonal, Clone 150503BD BiosciencesRRID:AB_11153301; Cat# 5623815 µl/test, FACS
AntibodyAnti-human CD28 PE-Cy5, mouse monoclonal, Clone CD28.2BiolegendRRID:AB_314312 Cat# 3029102.5 µl/test, FACS
AntibodyAnti-human CD45RO FITC, mouse monoclonal, Clone UCHL1Beckman CoulterCat# IM1247U8 µl/test, FACS
AntibodyAnti-human CD4 PerCP-Cy5.5, mouse monoclonal, Clone L200BD BiosciencesRRID:AB_394488 Cat# 5528382.5 µl/test, FACS
AntibodyAnti-human CD40L BV 605, mouse monoclonal, Clone 24–31BiolegendRRID:AB_2563832 Cat# 3108260.625 µl/test, FACS
AntibodyAnti-human Ki-67 BV 711, mouse monoclonal, Clone Ki-67BiolegendRRID:AB_ 2563861 Cat# 3505161.25 µl/test, FACS
AntibodyAnti-human CD69 APC-Cy7, mouse monoclonal, Clone FN50BiolegendRRID:AB_314849 Cat# 3109140.156 µl/test, FACS
AntibodyAnti-human IL-21 PE, mouse monoclonal, Clone 12-7219-42eBioscienceRRID:AB_1582260 Cat# 12-7219-420.156 µl/test, FACS
AntibodyAnti-human IFN-g PE-Cy7, mouse monoclonal, Clone B27BD BiosciencesRRID:AB_396760 Cat# 5576432.5 µl/test, FACS
AntibodyAnti-human IgG BV 421, mouse monoclonal, Clone G18-145BD BiosciencesRRID:AB_2737665 Cat# 5625812.5 µl/test, FACS
AntibodyAnti-human CD80 BV 605, mouse monoclonal, Clone 2D10BiolegendRRID:AB_11123909 Cat# 3052250.625 µl/test, FACS
AntibodyAnti-human IL-21R APC, mouse monoclonal, Clone 2G1-K12BiolegendRRID:AB_2123988 Cat# 3478082.5 µl/test, FACS
AntibodyAnti-human CD20 Alexa 700, mouse monoclonal, Clone 2H7BiolegendRRID:AB_493753 Cat# 3023220.625 µl/test, FACS
AntibodyAnti-human CD38 PE, mouse monoclonal, Clone HIT2BD BiosciencesRRID:AB_395853 Cat# 5554602.5 µl/test, FACS
AntibodyAnti-human CD21 PE-Cy5, mouse monoclonal, Clone Bly4BD BiosciencesRRID:AB_394028 Cat# 5510643.75 µl/test, FACS
AntibodyAnti-human CD10 PE-Cy7, mouse monoclonal, Clone HI10ABD BiosciencesRRID:AB_400216 Cat# 3410922.5 µl/test, FACS
AntibodyAnti-human IgD FITC, mouse monoclonal, Clone IA6-2BiolegendRRID:AB_10612567 Cat# 3482060.625 µl/test, FACS
AntibodyAnti-human CD27 BV 650, mouse monoclonal, Clone L128BD BiosciencesRRID:AB_2744352 Cat# 5632282.5 µl/test, FACS
AntibodyAnti-human ICOSL biotin, mouse monoclonal, Clone 2D3BiolegendRRID:AB_528729 Cat# 3094061.25 µl/test, FACS
AntibodyAnti-human streptavidin BV 711, mouse monoclonalBiolegendCat# 4052410.7 µl/test, FACS
AntibodyAnti-human Ki-67 PerCP-Cy5.5, mouse monoclonal, Clone B56BD BiosciencesRRID:AB_10611574 Cat# 5612842.5 µl/test, FACS
AntibodyAnti-human CD27 PerCP-Cy5.5, mouse monoclonal, Clone MT271BioegendRRID:AB_2561906 Cat# 3564080.312 µl/test, FACS
AntibodyAnti-human CD28 unconjugated, mouse monoclonal, Clone L293BD BiosciencesRRID:AB_400197 Cat# 3409751 µg/ml, PBMC stimulation
Commercial assay or kitHuman IgG ELISA Quantitation SetBethyl LaboratoriesCat# E80-104IgG ELISA, PBMC culture supernatants
Chemical compound, drugBrefeldin ASigma AldrichCat# B7651-5mg10 µg/ml, PBMC stimulation
Software, algorithmFlowJoBD Bioscienceshttps://www.flowjo.com
Software, algorithmPrism 8GraphPadhttps://www.graphpad.com/scientific-software/prism/
Software, algorithmBD FACSDivaBD Bioscienceshttps://www.bdbiosciences.com/en-us/instruments/research-instruments/research-software/flow-cytometry-acquisition/facsdiva-software
OtherLIVE/DEAD Fixable Aqua Dead Cell StainInvitrogenL349570.5 µl/ml, FACS

Study timepoints and processing

Request a detailed protocol

Recruitment of participants, vaccine administration and CHMI studies were conducted at Walter Reed Army Institute of Research (WRAIR) (Regules et al., 2016). A schema for vaccine timepoints and blood sample collection for immunology analyses in both vaccine regimens and in the control regimen of the study is shown in Figure 1. Blood draws for this were performed at eight different timepoints designated T0–T7: (i) pre-vaccination (T0), day 6 post dose 1 (T1), day 28 post dose 1 (T2), day 6 post dose 2 (T3), day 28 post dose 2 (T4), day 6 post dose 3 (T5), day 21 post dose 3 and pre-challenge (T6) and at study end (T7), 159 days post-challenge. Blood was processed for peripheral blood mononuclear cells (PBMC) and cryopreserved PBMC samples were shipped to the University of Miami (UM) in liquid nitrogen. All lab testing, analysis, data entry and plotting of graphs were performed in batches in a blinded manner with information related to vaccine regimen and protection status revealed to the UM lab only after all of the samples had been processed. This study was approved by the Institutional Review Boards of UM. De-identified PBMC samples were processed at UM laboratory.

Vaccine antigens, control antigen and monoclonal Abs

Request a detailed protocol

The HBs peptide pool, CSP peptide pool, CS repeat region protein (R32LR) and HBs protein were provided by GSK. The CSP protein (PF-CSP) and the C-terminal peptide (PF-16) were provided by PATH’s Malaria Vaccine Initiative (MVI). Staphylococcal enterotoxin B (List Biological Laboratories) and CPG ODN 2006 (InvivoGen) were used as positive control antigens for T and B cells, respectively. A list of monoclonal Abs used for the flow cytometry studies, including the information about clone and fluorochrome, is included in the Key resources table.

pTfh frequency and antigen-specific intracellular cytokine secretion (ICS) in short-term cultures

Request a detailed protocol

pTfh frequency and function were analyzed at pre-vaccination (T0) and at all timepoints post-vaccination in each regimen. Briefly, PBMC (1.5 × 106/ml/condition) were stimulated with CSP peptide pool (2 µg/ml) and HBs peptide pool (2 µg/ml) along with co-stimulation molecule antiCD28 (1 µg/ml) for 12 hr at 37°C. Brefeldin A (10 µg/ml) was added 7 hr after stimulation to prevent protein transport. SEB (1 µg/ml) was used as a positive control, and medium with co-sitm molecule antiCD28 alone was used as a negative control. After stimulation, cells were stained using a 14-color fluorochrome conjugated monoclonal Ab panel including surface markers that are specific for pTfh identification along with live/dead amine dye (aqua). Cells were then fixed, permeabilized and stained intracellularly for interleukin (IL-21), CD40L, and Ki67, acquired on a BD LSRFortessa, and analyzed by FlowJo (v 9.4.3, Tree Star Inc).

All flow cytometry analyses were performed using optimally titrated Ab concentrations and after applying appropriate fluorescent compensation using DIVA software on BD LSRFortessa at the time of acquisition and fine tuning of compensation using FlowJo at the time of analysis. Gating controls included unstained cells and Fluorescence Minus One (FMO), and for all stimulation experiments, unstimulated cells were used as additional biological controls. For basic pTfh identification, CD4 T central memory (TCM[CD3+CD4+CD45RO+CD27+]) were gated on CXCR5 to determine the frequencies of CXCR5+ subsets in CD4 TCM cells, designated as pTfh cells. For Ag-specific pTfh cells, CD40L+ CD4 T cells were gated on the basis of the expression of CD45RO and CXCR5 (as CD45RO+CXCR5+) as CSP-specific pTfh cells and CD45RO+CXCR5 as CSP-specific non-pTfh cells (Figure 2—figure supplement 1). CSP-specific pTfh cells and non-pTfh cells were further analyzed for the intracellular expression of signature cytokine IL-21, inducible co-stimulator ICOS and proliferation marker Ki67.

Ex vivo B cell maturation subsets

Request a detailed protocol

Thawed PBMC were analyzed for B cell phenotypes without in vitro stimulation by flow cytometry. Total mature B cell were identified as CD3CD10CD20+ cells after excluding immature CD10+ B cells, and total memory B cells were identified as CD20+CD27+ cells. On the basis of the expression of CD21, CD27 and IgD, B cell maturation subsets were identified as naïve (CD21hiIgD+CD27–), resting memory (RM: CD21hiCD27+), activated memory (AM: CD21lowCD27+) and atypical memory B cells (aMBC: CD21lowCD27). Within the total memory, RM and AM B cells IgD+IgG were identified as unswitch and IgDIgG+ as switch memory B cells (Figure 5—figure supplement 1).

Spontaneous Ab-secreting cells (ASC)

Request a detailed protocol

A spontaneous ASC enzyme linked immunospot (ELISpot) assay was performed at T0, T1, T3, and T5 as described previously (George et al., 2015; Rinaldi et al., 2017; Pallikkuth et al., 2011b) against wells coated with PF-CSP, R32LR and PF-16 malaria antigens, as well as HBs-antigen, using unstimulated PBMC. Data are expressed as spontaneous ASC/million PBMC.

Memory B cell analysis

Request a detailed protocol

Memory B cell responses were analyzed at T0, T2, T4, T5, T6 and T7 using memory B cell antibody secreting cell (ASC) ELISpot (George et al., 2015; Rinaldi et al., 2017; Pallikkuth et al., 2011b). PBMC 1.5 × 106/ mL per condition were stimulated for 5 days with 2 µg/ml each of malaria (PF-CSP, PF-16, R32LR) and HBs antigens. CpG oligodeoxynucleotides 2006 (CpG ODN2006: 1 µg/ml) was used as a positive control and medium as a negative control. On day 5, cells were harvested and assayed for Ag-specific induction of IgG-secreting cells by ELISpot assay (Pallikkuth et al., 2011b). The remaining cells were stained with fluorochrome conjugated monoclonal Abs that were specific for B cell maturation subsets (naïve, total memory, RM, AM, and aMBC). Switch and unswitch memory B cells within total memory, RM and AM subsets were identified on the basis of IgG and IgD expression along with proliferation marker Ki67 and analyzed by flow cytometry as described in Figure 5—figure supplement 1. Culture supernatants were stored at −80°C and assayed for IgG by ELISA.

Statistical analysis and data integration

Request a detailed protocol

To analyze this complex immunology dataset, which includes a large number of immune measurements for each subject, we performed an integration approach in which we combined traditional univariate analysis with multivariate machine-learning methods to isolate immune responses that were vaccine induced, to characterize regimen-specific differences, and to identify correlates of protection. A general workflow of the data integration approach is shown in Figure 6—figure supplement 2. Analyses were performed to compare changes in pTfh and B cell related markers for each group between different timepoints pre- and post-vaccination, or between P vs NP, or between regimens DFD and STD, at each timepoint or at selected timepoints. Generalized linear mixed-effects models (GLMM), fitted via Penalized Quasi-Likelihood (PQL) using R ‘MASS’ package was used to accommodate repeated measures of time, with random intercept set by patient ID (PID). P value was adjusted for multiple comparisons by Benjamini and Hochberg correction using R ‘multcomp’ package. A p value of <0.05 was considered to be significant. Immune measures were classified as vaccine-induced responses if they showed a significant difference from the pre-immune (vs. T0) timepoint. Immune measures were classified as regimen-specific and protection-specific differences if they showed a significant difference with respect to vaccine regimen (STD vs. DFD) or protection status (P vs. NP), respectively.

To assess the predictive value of the regimen- and protection-specific differences identified in this study, we used the random forest model, a machine-learning method, to make individualized predictions of regimen or protection status on the basis of the immune data alone. The random forest model was generated using all vaccine-induced immune responses (R caret package). We trained the model using the repeated cv method, subsampling the dataset by five-fold and resampling ten times. The random forest model was tuned using the caret R package. Specifically, the number of branches of the tree (mtry) and the rule for splitting (gini or extratrees) were adjusted to identify the optimal accuracy and kappa values during internal ten-fold cross validation, repeated ten times. The oneSE method was used to select the optimal model. To test the predictive accuracy of the random forest modeling approach, we carried out a leave-one-out analysis, in which one subject was removed from the dataset, after which the model was trained on the remaining subjects and then used to predict the adjuvant condition of the excluded subject on the basis of its immune data. We performed this for all subjects in the dataset, and calculated both the accuracy and kappa value of the prediction model. We used the varImp function to determine the variable importance for each generated model, and reported the average variable importance across all models to assess the relative importance of each vaccine-induced immune measure to predicting regimen or protection status.

Principal component analysis (PCA) was carried out in R using the ir.pca package and visualized using ggbiplot. For the PCA, we used a subset of the immune measures, using only parameters that were found to be predictive of regimen or protection status, as determined by variable importance analysis in the random forest model. Finally, correlation analysis was carried out in R using the cor function to calculate the Pearson correlation coefficient. All immune measures were compared with all other immune measures, and the correlation matrix was used for hierarchical clustering, using the hclust function in R, to identify groups of correlated immune parameters, termed ‘immune clusters’. All immune parameters within an immune cluster have a Pearson correlation coefficient of at least 0.80 to every other parameter in that cluster.

References

  1. 1
  2. 2
  3. 3
  4. 4
  5. 5
  6. 6
  7. 7
  8. 8
  9. 9
  10. 10
  11. 11
  12. 12
  13. 13
  14. 14
  15. 15
  16. 16
  17. 17
  18. 18
  19. 19
  20. 20
  21. 21
  22. 22
  23. 23
  24. 24
  25. 25
  26. 26
  27. 27
  28. 28
  29. 29
  30. 30
  31. 31
  32. 32
  33. 33
  34. 34
  35. 35
  36. 36
  37. 37
  38. 38
  39. 39
  40. 40
  41. 41
  42. 42
  43. 43
  44. 44
  45. 45
  46. 46
  47. 47
  48. 48
  49. 49
  50. 50
  51. 51
  52. 52
  53. 53
  54. 54
  55. 55
  56. 56
  57. 57
  58. 58
  59. 59
  60. 60

Decision letter

  1. Urszula Krzych
    Reviewing Editor; Walter Reed Army Institute of Research, United States
  2. Dominique Soldati-Favre
    Senior Editor; University of Geneva, Switzerland

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Acceptance summary:

RTS,S is a leading Plasmodium falciparum recombinant circumsporozoite protein (CSP) vaccine. Administered according to the standard 1-2-3 month regimen, RTS,S induces but a modicum of protracted protective immunity. By contrast, vaccination according to a delayed and fractional 3rd dose results in a superior protective outcome in controlled human malaria infection challenge studies. Understanding vaccine-induced immune protective mechanisms is crucial for developing of an improved vaccine against malaria. In this study, the authors conducted comparative analyses of RTS,S elicited immune responses in subjects vaccinated according to the standard versus delayed and fractional doses. Analyses of peripheral blood cellular and antibody responses combined with machine relearning approaches revealed that early induction of CSP-specific pTfh CD4 T cells and the persistence of CSP-specific B cell responses are associated with the superior protective outcome in the group that received delayed and fractional dose of RTS,S/AS01.

Decision letter after peer review:

Thank you for sending your article entitled "Cellular biomarkers of antibody response to RTS,S/AS01 malaria vaccine in a controlled human vaccine challenge model" for peer review at eLife. Your article is being evaluated by three peer reviewers, one of whom is a member of our Board of Reviewing Editors, and the evaluation is being overseen Dominique Soldati-Favre as the Senior Editor.

The reviewers considered that the authors indeed posed and important question regarding the mechanism of RTS,S-mediated protection in a very systematic fashion. On the basis of the reviews provided below, the main concerns are as follows: issues with flow cytometry results for both T and B cells; the accompanying statistical analysis; inclusion of fine/distinct CSP protein peptide or region specificities of IgG and pTfh cells; correlative interaction between pTfh and B cells from individuals, which would address the relationship and/or the qualitative difference between these cell types in protected DFD vs. STD; the machine learning portion needs to be properly presented/explained in the result section; revisions in the labeling of certain figures for clarity; and, finally, please address the issue concerning the writing style, particularly in the Results section.

Reviewer #1:

RTS, S is the most advanced vaccine against pre-erythrocytic (sporozoite) stage of malaria. Studies conducted in malaria endemic areas show only a modicum of protective efficacy and short durability. On the basis of the results from an earlier study by Stoute et al., the repeat of the approach of delivering RTS,S as DFD vs. STD immunizing dose has shown much promise in the recently conducted CHMI studies. The current study describes response of peripheral CD4 T cells, pTfh cells, and B cells as correlates of protection. The results add much value as they expand our understanding of the immune responses that form the basis of protection against malaria. Although immunological parameters may be quite different in endemic areas, particularly among older children who have had malaria infection prior to RTS,S vaccination, nonetheless, this study provides new information regarding pTfh cells in protective immunity against CHMI.

As with any studies detailing immune responses in human PBMC samples, this study was admittedly quite challenging; the authors succeeded in performed carefully timed analyses and amassed lots of important results. Applying machine learning analyses, they have deduced predictive value concerning early pTfh cells and memory B cells responses as predictors of protection. The manuscript is well organized and, on the whole, the experiments are well performed; however, there are several critical issues that need to be addressed.

The narrative in the Results section provides data without any benefit of a proper introduction as to what specific question was being asked and why a particular analysis was undertaken. For example, why was ICOS evaluated? Why were B cell responses evaluated? Similarly, what is the significance of measuring cellular expression of Ki67? The exception was an explanation for determining IL-21 for pThf cells.

The entire Results section concerning B cells is very difficult to follow. The main reason of this is that none of the discussed B cell subsets, e.g. MB, are defined phenotypically. I do acknowledge that the phenotypic markers are indicated in the Materials and methods section; however, it is rather clumsy to go back and forth between the pages looking for the specific phenotypes in one section of the paper and then going back to the Results section for a response of a given B cell population. Even for a reader with some background in immunology, it would be difficult to understand the significance of "…vaccine induced alteration in B cell phenotypes and function…", as neither specific phenotypic markers nor functions are indicated. What exactly are recall antigens, R32LR and PF-16?

Speaking of cellular phenotypes, I have not found any gating strategy for B cells. Therefore, it is impossible to evaluate responses shown in Figures 3 and 4. The authors discuss Ig switching, yet apart from ELISPOS for IgG-producing B cells, no other isotypes are shown. It is not clear what criteria the authors have used for the discussion of B cell maturation. Starting with data in Figure 3, how were the B cell gated for the different subsets? The B cell subsets need to be clearly indicated or identified by a set of specific phenotypic markers and Ig isotypes, otherwise, the abbreviations such as sAM, SM, are meaningless for the reader.

The authors mention that Ab ELISA experiments were performed, but the authors do not show any results. Also, the markers for atypical B cells depart from what is currently described in the literature. And what was the rationale for even considering atypical memory B cells?

Another critical issue is the authors' generalization of CSP specificities of CD4 T cells. The authors did use PF CSP peptide pools for activation of CD4 or pTfh cells. The content of these pools is not shown, and it needs to be indicated. Importantly, some peptide pools representing certain regions of PF CSP might have shown more recall activity than other peptide pools. In a previously published study, Schwenk and colleagues (2011, Vaccine 29), interrogated RTS,S induced CD4 T cells responses with PF CSP peptides and they showed that subjects respond to a discrete set of PF CSP peptides. Importantly, once PF CSP peptide-specific early responses are induced, the specificity repertoire does not shift following boost immunization or CHMI. I will add, that the authors looked at inflammatory CD4 T cells. In view of these published results concerning PF CSP peptide specific inflammatory CD4 T cells responses, such information would be quite important if only to understand what specificities induced the early pTfh cells in the current study; is there specificity overlap between CD4 T cells with different functional attributes? Because according to the current observation early CD4 T cell responses do matter; it is imperative to show the results from the individual pool activated T cells. In this regard, it also would have been instructive to determine HLA of P vs. NP or early responders.

It is not entirely clear as to why the authors are surprised that the observed correlates of protection were directed to Ag-specific CD4 and pTfh cells that arose post does 1. To gain some understanding of the T-B cell collaboration for Ab production, it would have been instructive to investigate if the B cell responses in protected persons were T cell dependent. There are markers for such B cell responses, e.g., CD73, CD80 etc. These analyses might need to be conducted to sort our T cell dependent vs. T cell independent B cell responses in the P vs. NP subjects. Pepper et al. did show that Plasmodium induced B cells survive only for a short time as short-lived memory B cells. This may provide some explanation to the authors' query.

The labeling of the figures should be improved, particularly with regard to the time of PBMC sampling versus vaccinations. The authors need to indicate by arrows on the X axis the timing of vaccination vis-a-vis samplings. This will frame the scheduling indicated by T0, T1, etc. Otherwise, one needs to constantly consult the text for clarification.

The writing in general is a bit confusing. For example, (1) In the Discussion – "A major objective of.……was to determine if one regimen, DFD or STD, was superior…" The results of the superiority of DFD over STD in protection was published previously. Please reword the sentence; (2) "…that differences in immunity in the DFD regimen arose…" (3) "…B cell maturation subsets…"; (4) previously conducted studies are not referenced.

In the Abstract the authors state that B cells are rescued by the DFD immunization regimen. It is true that B cells producing IgG responses declined (Figure 3). Are B cell numbers also lost? I do not recall any studies where the authors show apoptosis, or some other form of B cell death induced by the STD immunization regimen.

Reviewer #2:

In this study Pallikkuth et al. examine peripheral T follicular helper (pTfh) cells and memory B cells as potential immunologic mechanisms responsible for the increased protection observed in subjects receiving RTS,S/AS01 in a delayed fractional dose (DFD) compared to the standard dose regimen (STD).

The authors found that subjects protected from malaria demonstrate an expansion of total- and CSP specific pTfh cells following vaccination and that the variation observed in protection between the STD and DFD arms arises mainly from variations in antigen-specific B cells. In addition, the authors apply a machine learning analysis to identify immune correlates associated with protection from malaria, which indicates an important role for early induction of pTfh responses and suggests that class-switched memory B cells can be engendered "saved" by a delayed fractional dose regimen. This study addresses an important question regarding the mechanism of RTS,S-mediated protection in a very systematic fashion, leveraging a unique trial dataset. The reported association with Tfh responses and protection from RTS,S is a significant advance, and the purported differences observed between the dosing regimens of high potential significance. However, my major concerns revolve around the quality of the flow cytometry data, which is difficult to assess given limited info on gating and absence of B cell flow plots. In addition, some aspects of the statistical analyses are not clear. An inherent weakness of the machine learning analyses rests in the fact that it does not use separate training, validation, and test sets. The dataset from which the model is derived appears to be the same dataset from which the authors go back and make predictions, which limits external inference. While the Introduction and Discussion are well written but several aspects of the Results and figures are not clear.

1) It is unclear from Figure 2 how flow gates were set; ideally this could be based on FMO controls, which should be shown in the gating panel, but I could not find any mention of how gates were determined in the Materials and methods. The cutoff for ICOS, IL21, and Ki67 looks rather arbitrary, and the Ki67 stain did not appear to provide good separation based on the (very small) example plot.

2) No flow cytometry plots for B cell subsets are provided. These are critical data and should be shown (with Figure 4 or as Supplemental). This is of particular concern given that the staining for Ki67 on T cells (Figure 2C) was not very convincing, and Ki67+ B cells comprise the majority of the most predictive parameters in Table 1.

3) What do the error bars indicate in Figure 1-4 (SD, SEM, other)? In some cases the error bars are quite short relative to the difference between P and NP values, yet the difference is not significant; while in the puzzling case of Figure 2F (T2) the difference IS statistically significant yet the values are almost the same and the bars widely overlap.

4) The definition of Tfh is somewhat non-conventional and only partially justified (i.e. in Discussion re: PD-1).

5) The reader has to work rather hard to grasp what comparisons were statistically significant. The differences from T0 to other timepoints stand out (but are not the most interesting result). This overshadows the p-values comparing P and NP. Only the latter are mentioned in the Figure Legend. (applies to Figures 1-3).

6) The machine learning portions of the paper (rationale, findings) are nicely explained in the Discussion, but are presented in a very confusing manner in the Results. I did not understand the point of this analysis until the Discussion.

7) Figure 5 was not helpful to this reviewer and could be considered for omission; conversely Figure S1 showing the trial schematic with definition of timepoints T0-T7 was very helpful and should be considered for inclusion in the main text.

8) Was any analysis of how Tfh correlate with B cells in individuals performed? This would be of interest but appears to be missing.

9) The paper would benefit from a discussion of the implications of these findings on RTS,S vaccination in real-world endemic settings, where many children would inevitably be exposed (often repeatedly) to malaria antigens during the "suboptimal" 2-6 month window following priming (see Obeng-Adjei et al., 2015).

Reviewer #3:

The manuscript entitled "Cellular biomarkers of antibody response to RTS, S/AS01 malaria vaccine in a controlled human vaccine challenge model" by Pallikkuth et al., immune response to RTS, S/AS01 malaria vaccine and found that delayed fractional dose (DFD) regimen is superior in inducing higher class-switched Ag specific memory B cells post the third immunization compared to the (STD) standard dose regimen.

Comments as follows:

1) What is the qualitative difference between the protected versus non-protected subjects in terms of the cTfh levels or its functions?

2) The authors indicate that expansion of IL-12 and ICOS+ Ki67+ cells is indicative of vaccine-induced enhancement; however, this happens in both the arms. It did not look like this is happening only in the DFD arm. Results are written as if they are significantly different.

3) Do the authors have any hypothesis for the enhancement of B and Tfh cell response in DFD group compared to the STD regimen?

4) Did they do an antigen-specific IgG subclass response to see whether DFM has wide range of antibody responses compare to STD regimen?

5) It is also surprising to see that non-Tfh component are not looked at properly such as IFN-γ, TNF-α, and IL-2 levels. What is the association of these cells with antigen specific B cell response between P versus NP individuals?

6) In the Figure 7 it is not very clear that what predicts the protection and what is not predicting the protection.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "Cellular biomarkers of antibody response to RTS,S/AS01 malaria vaccine in a controlled human vaccine challenge model" for further consideration by eLife. Your revised article has been evaluated by Dominique Soldati-Favre (Senior Editor) and a Reviewing Editor.

The majority of the requested revisions are editorial in nature, with few exceptions.

1) In their response to a comment on non-pTfh cells, the authors rely on negative data (not shown) derived from the integration analysis. Despite the outcome of this analysis, it might be useful and most informative to assess the non-pTfh cells alongside the pTfh cells. In the event that the PBMCs are no longer available, the authors need to address these potentially crucial components as possible mechanisms for the superior outcome in DFD regimen of immunization. As an example, previously published works have shown an association between IL-2+ and IFNγ+ CD4 T cells and antibody responses in RTS,S vaccinated subjects. In the absence of results, including these points would improve the tenor of the Discussion.

2) Please revisit the legends and confirm whether they indicate that the error bars signify SEM.

3) Figure 3. Unlike IL-21 and ICOS, the expression of Ki67 was rather transitory and this point needs to be acknowledged.

4) A title should be more reflective of the key findings.

5) In the Abstract, there should be a clearer pivot from data that is previously published (trial results) to data presented therein.

6) Subheading “CSP-specific B cell responses emerge in protected subjects after the second dose and higher regimen specific memory B cell response in DFD arm” makes no sense grammatically.

7) In the Discussion, the authors state that "This is the first study demonstrating a protective role of CSP-specific pTfh in RTS,S/AS01 vaccination". Please tone down this statement as the results do not show a protective role of pTfh cells, they simply show an association.

8) Discussion section, final two sentence in the penultimate paragraph should could be removed, as they are unnecessary.

9) Results section requires extensive editing in that a rationale for choosing to ask a certain question or perform a particular experiment needs to be clearly stated. It is insufficient to state, "We next investigated…".

10) Discussion paragraph six is very repetitive to early wording in the Results.

11) The Discussion section on the whole still appears largely as a rehash of the Results section. It fails to provide a cogent picture or a hypothesis that would suggest possible interactions that may be occurring between the pThf cells (and possibly non-pTfh CD4 T cells) and the various B cells subsets and which bring about this enhanced protection seen in subjects receiving DFD regiment.

12) Along these lines, the authors need to relate their finding to a larger picture of RTS,S vaccination in African countries. Specifically, the paper would benefit from a discussion of the implications of these findings on RTS,S vaccination in real-world endemic settings, where many children would inevitably be exposed (often repeatedly) to malaria antigens during the "suboptimal" 2-6 month window following priming. The authors are not requested to provide field data, which they obviously cannot; however, if a DFD regimen were deployed in Africa, many children would get a natural malaria infection that might have the same negative impact as the STD dose #3 and this is worth discussing.

https://doi.org/10.7554/eLife.51889.sa1

Author response

Reviewer #1:

[…]

The narrative in the Results section provides data without any benefit of a proper introduction as to what specific question was being asked and why a particular analysis was undertaken. For example, why was ICOS evaluated? Why were B cell responses evaluated? Similarly, what is the significance of measuring cellular expression of Ki67? The exception was an explanation for determining IL-21 for pThf cells.

We have edited the Results section to address the concerns raised and indicated the relevance of investigating pTfh, ICOS, Ki67, and B cell subsets in this study. (Subsection “CSP-specific pTfh responses are elevated in protected subjects” and “CSP-specific B cell responses emerge in protected subjects after the second dose and higher regimen specific memory B cell response in DFD arm”.)

The entire Results section concerning B cells is very difficult to follow. The main reason of this is that none of the discussed B cell subsets, e.g. MB, are defined phenotypically. I do acknowledge that the phenotypic markers are indicated in the Materials and methods section; however, it is rather clumsy to go back and forth between the pages looking for the specific phenotypes in one section of the paper and then going back to the Results section for a response of a given B cell population. Even for a reader with some background in immunology, it would be difficult to understand the significance of "…vaccine induced alteration in B cell phenotypes and function…", as neither specific phenotypic markers nor functions are indicated. What exactly are recall antigens, R32LR and PF-16?

We have included the gating strategy for the different B cell subsets in Figure 5—figure supplement 1 and have added the description of B cell subsets in the Results and Materials and methods sections of the revised manuscript. We have also revised the statement to make it more specific for measuring the phenotype of B cell subsets by flow cytometry and their function by ELIspot assays. In this prime boost vaccination strategy R32LR peptide corresponds to the repeat region of PF-CSP and the PF-16 peptide corresponds to the C-terminal region of PF-CSP. The vaccine that was administered to CSP-negative adults includes both, the R32LR and the C-term region of PF-CSP.

Speaking of cellular phenotypes, I have not found any gating strategy for B cells. Therefore, it is impossible to evaluate responses shown in Figures 3 and 4. The authors discuss Ig switching, yet apart from ELISPOS for IgG-producing B cells, no other isotypes are shown. It is not clear what criteria the authors have used for the discussion of B cell maturation. Starting with data in Figure 3, how were the B cell gated for the different subsets? The B cell subsets need to be clearly indicated or identified by a set of specific phenotypic markers and Ig isotypes, otherwise, the abbreviations such as sAM, SM, are meaningless for the reader.

Please refer to Comment 2 regarding the inclusion of a detailed gating strategy for B cells in the revised manuscript. Switched and unswitched B cells were identified based on the IgG and IgD expression on CD27+CD21hi or CD27+CD21low B cells with IgD+IgG- as unswitched and IgD-IgG+ as switched B cells. This information is now included in the Materials and methods and Results sections of the revised manuscript.

The authors mention that Ab ELISA experiments were performed, but the authors do not show any results. Also, the markers for atypical B cells depart from what is currently described in the literature. And what was the rationale for even considering atypical memory B cells?

We analyzed the IgG levels in PBMC culture supernatants after 5 days of stimulation with the following vaccine antigens: full-length CSP (PF-CSP), C-term CSP (PF-16), and repeat region (R32LR). Responses to each of these antigens were included in the data integration analysis and none of the IgG responses appeared as variables associated with protection or regimen difference. We have now included the IgG data as a supplementary figure (Figure 5—figure supplement 5) in the revised manuscript.

In this study, we identified atypical memory B cells based on their expression of CD20, CD21, and CD27 (CD20+ CD21lo/neg CD27-B cells). A similar phenotype of atypical memory B cells (aMBC) was reported previously by different investigators. A higher frequency of aMBC have been reported in children and adults with chronic asymptomatic Plasmodium falciparum infection, suggesting an expansion of these B cells in malaria endemic areas (Weiss et al., 2009; Portugal et al., 2015). A decline in the atypical MBC pool was observed during a 12-month period without malaria transmission, further substantiating a role of persistent parasite exposure in the maintenance of this population (Ayieko et al., 2013). In the present study, our goal was to investigate antigen-specific atypical memory B cells following RTS.AS01 vaccination and their association with vaccine-induced protection, Moreover, we wanted to assess effect of DFD vs. STD regimen on the frequencies of vaccine-specific aMBC. We found higher frequencies of PF-CSP-specific proliferating (Ki67+) aMBC at T5 and T6 in the DFD regimen compared to STD regimen. We have discussed this in the Discussion section.

Another critical issue is the authors' generalization of CSP specificities of CD4 T cells. The authors did use PF CSP peptide pools for activation of CD4 or pTfh cells. The content of these pools is not shown, and it needs to be indicated. Importantly, some peptide pools representing certain regions of PF CSP might have shown more recall activity than other peptide pools. In a previously published study, Schwenk and colleagues (2011, Vaccine 29), interrogated RTS,S induced CD4 T cells responses with PF CSP peptides and they showed that subjects respond to a discrete set of PF CSP peptides. Importantly, once PF CSP peptide-specific early responses are induced, the specificity repertoire does not shift following boost immunization or CHMI. I will add, that the authors looked at inflammatory CD4 T cells. In view of these published results concerning PF CSP peptide specific inflammatory CD4 T cells responses, such information would be quite important if only to understand what specificities induced the early pTfh cells in the current study; is there specificity overlap between CD4 T cells with different functional attributes? Becasue according to the current observation early CD4 T cell responses do matter; it is imperative to show the results from the individual pool activated T cells. In this regard, it also would have been instructive to determine HLA of P vs. NP or early responders.

Experiments related to fine-mapping with distinct CSP peptides will not be possible at this point and are out of scope for the current study. We achieved the goals of the present study which were to identify RTS,S/AS01vaccine-induced pTfh and B cells and their relation to malaria infection outcome in a controlled human malaria infection model. Moreover, we don’t have cryopreserved cells available for additional experiments, as blood sample collection was limited to specific objectives of the study. In the context of the comment being raised here, the review by Moris et al. (Hum Vaccin Immunother. 2018;14:17-27) discusses CSP epitope-specific CD4 T cell responses in malaria vaccine trials in prior studies performed by GSK and others. The serum IgG response from the current trial was published previously by our collaborators (Chaudhury et al., 2017).

Importantly, we included CSP antigen-induced intracellular IFNγ responses by total CD4 and pTfh cells, as well as CSP-specific pTfh cells in the data integration analyses and our results show that none of the IFNγ+ T cell variables were associated significantly with protection or regimen level differences. HLA information of protected and non-protected subjects or early responders were not available at the time these analyses were conducted. We mentioned this in the Discussion section as a study limitation.

It is not entirely clear as to why the authors are surprised that the observed correlates of protection were directed to Ag-specific CD4 and pTfh cells that arose post does 1. To gain some understanding of the T-B cell collaboration for Ab production, it would have been instructive to investigate if the B cell responses in protected persons were T cell dependent. There are markers for such B cell responses, e.g., CD73, CD80 etc. These analyses might need to be conducted to sort our T cell dependent vs. T cell independent B cell responses in the P vs. NP subjects. Pepper et al. did show that Plasmodium induced B cells survive only for a short time as short-lived memory B cells. This may provide some explanation to the authors' query.

We included data on ex-vivo expression of CD80, a marker of T dependent B cell responses, on total B cells, resting memory and activated memory B cells as a supplementary figure (Figure 5—figure supplement 2) in the revised manuscript. CD80 expression did not differ significantly between P and NP subjects at any time point, although a trend of higher expression in activated memory B cells at T6 and T7 in the protected subjects was noted.

We also analyzed the data for differences in the frequencies of dead cells between DFD vs. STD regimens at T5 (6 days post dose 3) based on Live/Dead aqua staining (Aqua+/Aqua-) on T cells and B cells as shown in Author response image 1. Mean frequencies of dead cells within the T cell compartment were 5.78 ± 3.6 for DFD and 6.3 ± 3.9 for STD regimen. For the B cells, the mean frequencies of dead cells were 8.6 ± 5.0 for the DFD and 9.2 ± 4.9 STD arms. The differences in the frequencies dead cells for either T or B cells were not significant between the two regimens. We mentioned this information in the Discussion section.

Author response image 1
Frequencies of dead cells in T cell and B cell compartments at T5 (6 days post dose 3).

Dead cells were gated from total CD3+ and CD20+ cells based live dead aqua staining as CD3+aqua+and CD 20+aqua+ cells. Scatter dot plots showing frequencies of dead cells for T cells (top right) and B cell (bottom right) betweeen DFD vs. STD regimen.

The labeling of the figures should be improved, particularly with regard to the time of PBMC sampling versus vaccinations. The authors need to indicate by arrows on the X axis the timing of vaccination vis-a-vis samplings. This will frame the scheduling indicated by T0, T1, etc. Otherwise, one needs to constantly consult the text for clarification.

We revised the figures and included information on the timing of vaccination, PBMC sampling and Plasmodium challenge.

The writing in general is a bit confusing. For example, (1) In the Discussion – "A major objective of.……was to determine if one regimen, DFD or STD, was superior…" The results of the superiority of DFD over STD in protection was published previously. Please reword the sentence; (2) "…that differences in immunity in the DFD regimen arose…" (3) "…B cell maturation subsets…"; (4) previously conducted studies are not referenced.

We revised the manuscript to address the concerns about the writing style.

In the Abstract the authors state that B cells are rescued by the DFD immunization regimen. It is true that B cells producing IgG responses declined (Figure 3). Are B cell numbers also lost? I do not recall any studies where the authors show apoptosis or some other form of B cell death induced by the STD immunization regimen.

We did not include apoptosis markers in the B cell flow cytometry staining panel. Frequencies of dead B cells were estimated from the flow cytometry gating using live/Dead Aqua staining as shown in Figure 1 in the response to comment 6. As discussed above, we did not find significant differences in the frequencies of dead B cells comparing the DFD vs. STD regimens. We are not sure about whether the total B cell numbers decreased, as we don’t have the information for absolute B cell numbers for the DFD and. STD regimens. However, the frequencies of B cells did not change between DFD vs. STD regimens. Our data suggests that the DFD regimen may result in a qualitative improvement within the B cell compartment as compared to the STD regimens.

Reviewer #2:

[…]

1) It is unclear from Figure 2 how flow gates were set; ideally this could be based on FMO controls, which should be shown in the gating panel, but I could not find any mention of how gates were determined in the Materials and methods. The cutoff for ICOS, IL21, and Ki67 looks rather arbitrary, and the Ki67 stain did not appear to provide good separation based on the (very small) example plot.

We revised the gating strategy figures for IL-21, ICOS and Ki67 (now Figure 3 in the revised manuscript). We included the dot plots for fluorescence minus one (FMO) staining controls to clarify the gating issues. The gating strategy write-up was revised in the Materials and methods to include the gating control information. In addition, we included a new supplementary figure (Figure 2—figure supplement 1) showing the gating strategy that was used to identify total pTfh, CSP-specific CD4, and CSP-specific pTfh cells.

2) No flow cytometry plots for B cell subsets are provided. These are critical data and should be shown (with Figure 4 or as Supplemental). This is of particular concern given that the staining for Ki67 on T cells (Figure 2C) was not very convincing, and Ki67+ B cells comprise the majority of the most predictive parameters in Table 1.

A detailed gating strategy for the identification of B cell subsets and Ki67 expression on various B cell subsets is included as Figure 5—figure supplement 1 in the revised manuscript. The definition of B cell subsets based on the markers used, is further clarified in the Materials and methods and Results sections.

3) What do the error bars indicate in Figure 1-4 (SD, SEM, other)? In some cases the error bars are quite short relative to the difference between P and NP values, yet the difference is not significant; while in the puzzling case of Figure 2F (T2) the difference IS statistically significant yet the values are almost the same and the bars widely overlap.

All error bars in the line graphs indicate the standard error of mean (SEM). The data for Figure 2F at T2 was rechecked and corrected as we found an error in the data for the protected subjects. The difference between P and NP subjects at T2 is significant with a p value of <0.01.

4) The definition of Tfh is somewhat non-conventional and only partially justified (i.e. in Discussion re: PD-1).

We identified the phenotype of CSP-specific pTfh as CD4+CD40L+ cells expressing CD45RO and CXCR5 and also investigated their cytokine profile for IL-21 and IFNγ. We did not include markers associated with pTfh heterogeneity such as PD1, CD38, CXCR3 etc. in the 15-color flow cytometry panel that was used to study the Ag.pTfh. Since we were focusing on a less frequent pTfh population within the CSP-specific CD4 T cells compartment, we rationalized that we needed to maintain a broad(er) definition of pTfh cells as we studied their antigen-specific function with the goal of finding novel determinants of malaria vaccine-induced immune responses. In previous studies, we saw that some of the phenotypic markers used for pTfh identity such as PD1 and CD38 are also influenced by age or other infectious diseases (De Armas et al., 2017). Moreover, in an influenza vaccine study, we found that activated pTfh cells (HLA-DR+CD38+) expressing PD1 appear to be detrimental to the antigen-specific pTfh responses (Pallikkuth et al., 2019). This has been discussed in the Discussion section.

5) The reader has to work rather hard to grasp what comparisons were statistically significant. The differences from T0 to other timepoints stand out (but are not the most interesting result). This overshadows the p-values comparing P and NP. Only the latter are mentioned in the Figure Legend. (applies to Figures 1-3).

We revised all the figures and included information on the timing of vaccination, blood sampling and Plasmodium challenge. In order to address the differences between P and NP subjects at each timepoint, we revised the figures by including connecting lines between timepoints in which differences between P vs. NP are significant.

6) The machine learning portions of the paper (rationale, findings) are nicely explained in the Discussion, but are presented in a very confusing manner in the Results. I did not understand the point of this analysis until the Discussion.

The machine learning section of the Results section has been re-organized and significant parts of it have been re-written. One paragraph from the machine learning section in the original draft was moved into a new section of the Results (“Identifying vaccine-induced immune responses in RTS,S”) to make the section more clear. We have revised Results section and Materials and methods section to highlight the changes.

7) Figure 5 was not helpful to this reviewer and could be considered for omission; conversely Figure S1 showing the trial schematic with definition of timepoints T0-T7 was very helpful and should be considered for inclusion in the main text.

As suggested by the reviewer, we have moved the supplementary figure 1 in the original submission showing study schematics as a main figure (Figure 1) and Figure 5 has been included as Figure 6—figure supplement 1 in the revised manuscript.

8) Was any analysis of how Tfh correlate with B cells in individuals performed? This would be of interest but appears to be missing.

We carried out a correlation analysis between all pTfh and all memory B cell parameters at each timepoint and did not find any direct correlations between pTfh parameters and B cell measures. This observation combined with the early pTfh responses in protected subjects, suggest that initial Tfh cell priming may be required for the subsequent development of a qualitatively superior Tfh-induced B cell response induced by RTS,S/AS01B vaccine. The actual timing and magnitude of pTfh response that need to be elicited in order to influence the quality of the later B cell responses and improved protection in the DFD regimen need further investigations. We have discussed the results and implications in the Discussion.

9) The paper would benefit from a discussion of the implications of these findings on RTS,S vaccination in real-world endemic settings, where many children would inevitably be exposed (often repeatedly) to malaria antigens during the "suboptimal" 2-6 month window following priming (see Obeng-Adjei et al., 2015).

While we did not conduct detailed assessments of Tfh and B cell responses following RTS,S vaccination in the field, we are currently in the process of developing a study that will address the role of select T and B cell subpopulations, such as atypical memory B cells, Tfh and gd T cells, in the context of RTSS immunization and immune hypo-responsiveness in Kenyan adults. We have included this information in the Discussion.

Reviewer #3:

[…]

1) What is the qualitative difference between the protected versus non-protected subjects in terms of the cTfh levels or its functions?

In the revised manuscript, we show the ex-vivo frequencies of total pTfh (CD45RO+CD27+CXCR5+) cells in Figure 2A; ICOS expression on ex-vivo total pTfh is mentioned in the Results section. Our data show that frequencies of total pTfh were significantly different between P and NP subjects at T3, T4, T6 and T7 and the ICOS expression on total pTfh was significantly higher in P subjects at T6 and T7 compared to NP subjects. We also analyzed the IL-21 expression in total pTfh after 12 hrs of CSP stimulation, and similar to the observed increase in IL-21 production in the Ag.pTfh cells post-vaccination, there was a trend of higher IL-21 expression in total pTfh cells in P subjects post-vaccination (Author response image 2). Taken together, we consider that pTfh are quantitatively and qualitatively superior in P subjects as compared to NP subjects. We have included this information in the Results section.

Author response image 2
Trend of higher frequencies of total pTfh cells expressing IL-21 after vaccination in P subjects.

Il-21+ total pTfh cells after CSP stimulation by flow cytometry. IL-21+ pTfh cells were identified as CD4+CD45RO+CD27+CXCR5 cells.

2) The authors indicate that expansion of IL-12 and ICOS+ Ki67+ cells is indicative of vaccine-induced enhancement; however, this happens in both the arms. It did not look like this is happening only in the DFD arm. Results are written as if they are significantly different.

Our conclusions are based on the increase in IL-21, ICOS, and Ki67 comparing baseline to post-vaccine timepoints in protected subjects, indicating a vaccine-induced enhancement of these three immune parameters in the CSP-specific pTfh cells. Non-protected subjects did not show an increase in these markers. Moreover, in the DFD arm there was significantly higher IL-21 and ICOS expression at T7 compared to the STD arm, which further supports a regimen-induced enhancement of the pTfh function.

3) Do the authors have any hypothesis for the enhancement of B and Tfh cell response in DFD group compared to the STD regimen?

One possible hypothesis is that early Tfh response induces by the initial vaccine doses result in the formation of a strong high affinity memory B cell pool specific to CSP antigen and the delayed fractional dose lead to expansion and differentiation of the preformed memory B cells to Ab secreting cells. In the case of STD vaccine regimen, early administration of the booster dose was detrimental to the pre-formed memory B cell expansion and differentiation. To support our hypothesis, our data integration analysis showed that nearly 30% of good “early” responders in the STD arm became non-protected after the 3rd dose, indicating that dose 3 within a 0, 1, 2 month dose schedule may have an adverse effect on an otherwise promising immune response generated by dose 1 and 2.

4) Did they do an antigen-specific IgG subclass response to see whether DFM has wide range of antibody responses compare to STD regimen?

We analyzed the total IgG levels in PBMC culture supernatants after 5 days of stimulation with malaria antigens, such as full-length CSP (PF-CSP), C-term CSP (PF-16), and repeat region (R32LR). Responses to each of these antigens were included as a variable in the data integration analysis and none of the IgG responses were selected by the model as being associated with protection or regimen difference. We included this information in the revised manuscript as a supplementary figure (Figure 5—figure supplement 5). Moreover, the serum IgG response specificity/isotype response from this trial was published previously by our collaborators (Chaudhury et al., 2017).

5) It is also surprising to see that non-Tfh component are not looked t properly such as IFN-γ, TNF-α, and IL-2 levels. What is the association of these cells with antigen specific B cell response between P versus NP individuals?

We have analyzed non-Tfh compartment and IFNγ responses, while IL-2 or TNF in the Tfh compartment were not analyzed in this study. Non-Tfh and IFNγ data were included as variables in the data integration analysis and neither was selected as a variable associated with protection or regimen difference in the data integration analysis. We included this information in the Discussion.

6) In the Figure 7 it is not very clear that what predicts the protection and what is not predicting the protection.

We have included the information related parameters predictive of protection in this model as a new Supplementary file 1B.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

1) In their response to a comment on non-pTfh cells, the authors rely on negative data (not shown) derived from the integration analysis. Despite the outcome of this analysis, it might be useful and most informative to assess the non-pTfh cells alongside the pTfh cells. In the event that the PBMCs are no longer available, the authors need to address these potentially crucial components as possible mechanisms for the superior outcome in DFD regimen of immunization. As an example, previously published works have shown an association between IL-2+ and IFNγ+ CD4 T cells and antibody responses in RTS,S vaccinated subjects. In the absence of results, including these points would improve the tenor of the Discussion.

We have now included the non-pTfh data in the Results section of the manuscript as a supplementary figure (Figure 2—figure supplement 2).

2) Please revisit the legends and confirm whether they indicate that the error bars signify SEM.

The information about the error bars signifying SEM is included in all figure legends.

3) Figure 3. Unlike IL-21 and ICOS, the expression of Ki67 was rather transitory and this point needs to be acknowledged.

We have now mentioned the transitory nature of the Ki67 expression in the Results

4) A title should be more reflective of the key findings.

We have changed the manuscript title to be more reflective of our findings. The new title is “A delayed fractionated dose RTS,S AS01 vaccine regimen mediates protection via improved B cell responses”.

5) In the Abstract, there should be a clearer pivot from data that is previously published (trial results) to data presented therein.

Abstract has been revised.

6) Subheading “CSP-specific B cell responses emerge in protected subjects after the second dose and higher regimen specific memory B cell response in DFD arm” makes no sense grammatically.

We have revised the subheading.

7) In the Discussion, the authors state that "This is the first study demonstrating a protective role of CSP-specific pTfh in RTS,S/AS01 vaccination". Please tone down this statement as the results do not show a protective role of pTfh cells, they simply show an association.

We have revised the Discussion section and the suggested statement has been removed.

8) Discussion section, final two sentence in the penultimate paragraph should could be removed, as they are unnecessary.

Deleted as suggested.

9) Results section requires extensive editing in that a rationale for choosing to ask a certain question or perform a particular experiment needs to be clearly stated. It is insufficient to state, "We next investigated…”.

We have extensively revised the Results section to indicate the rationale/question related to each analysis performed.

10) Discussion paragraph six is very repetitive to early wording in the Results.

The Discussion has been revised.

11) The Discussion section on the whole still appears largely as a rehash of the Results section. It fails to provide a cogent picture or a hypothesis that would suggest possible interactions that may be occurring between the pThf cells (and possibly non-pTfh CD4 T cells) and the various B cells subsets and which bring about this enhanced protection seen in subjects receiving DFD regiment.

We have revised the Discussion extensively to discuss the hypothesis that early Tfh response induced by the initial vaccine dose results in the formation of a strong high affinity memory B cell pool specific to CSP antigen, and the DFD leads to expansion and differentiation of the preformed memory B cells to Ab secreting cells.

12) Along these lines, the authors need to relate their finding to a larger picture of RTS,S vaccination in African countries. Specifically, the paper would benefit from a discussion of the implications of these findings on RTS,S vaccination in real-world endemic settings, where many children would inevitably be exposed (often repeatedly) to malaria antigens during the "suboptimal" 2-6 month window following priming. The authors are not requested to provide field data, which they obviously cannot; however if a DFD regimen were deployed in Africa, many children would get a natural malaria infection that might have the same negative impact as the STD dose #3 and this is worth discussing.

We have revised the Discussion to relate our findings to the field. We have updated the section in the Discussion reflecting on the induction of Tfh cells by RTS,S/AS01E vaccination in African children, but balanced by findings that malaria infection can induce dysfunctional Tfh phenotypes. In future DFD trials in the field, the induction of Tfh cells should be investigated.

https://doi.org/10.7554/eLife.51889.sa2

Article and author information

Author details

  1. Suresh Pallikkuth

    Department of Microbiology and Immunology, University of Miami Miller School of Medicine, Miami, United States
    Contribution
    Conceptualization, Data curation, Formal analysis, Supervision, Investigation, Methodology, Writing - original draft, Writing - review and editing
    Competing interests
    No competing interests declared
  2. Sidhartha Chaudhury

    Biotechnology HPC Software Applications Institute, Telemedicine and Advanced Technology Research Center, U.S. Army Medical Research and Development Command, Fort Detrick, United States
    Present address
    Center for Enabling Capabilities, Walter Reed Army Institute of Research, Silver Spring, United States
    Contribution
    Data curation, Software, Methodology, Writing - original draft
    Competing interests
    No competing interests declared
  3. Pinyi Lu

    1. Biotechnology HPC Software Applications Institute, Telemedicine and Advanced Technology Research Center, U.S. Army Medical Research and Development Command, Fort Detrick, United States
    2. Henry M. Jackson Foundation for the Advancement of Military Medicine, Rockville, United States
    Contribution
    Software, Formal analysis
    Competing interests
    No competing interests declared
  4. Li Pan

    Department of Microbiology and Immunology, University of Miami Miller School of Medicine, Miami, United States
    Contribution
    Software, Formal analysis
    Competing interests
    No competing interests declared
  5. Erik Jongert

    GSK Vaccine, Rixensart, Belgium
    Contribution
    Conceptualization, Methodology, Writing - review and editing
    Competing interests
    is an employee of GSK, and owns shares of GSK
  6. Ulrike Wille-Reece

    PATH’s Malaria Vaccine Initiative, Washington DC, United States
    Contribution
    Conceptualization, Resources, Supervision, Methodology, Project administration, Writing - review and editing
    Competing interests
    No competing interests declared
  7. Savita Pahwa

    Department of Microbiology and Immunology, University of Miami Miller School of Medicine, Miami, United States
    Contribution
    Conceptualization, Supervision, Funding acquisition, Methodology, lab facilities, technical support, writing, review and editing
    For correspondence
    spahwa@med.miami.edu
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-4470-4216

Funding

PATH’s Malaria Vaccine Initiative or PATH

  • Savita Pahwa

The funder contributed to conceptualization, data review and manuscript preparation

Acknowledgements

We thank Maria Pallin, Celeste Sanchez and Varghese George from the Pahwa lab for their assistance with laboratory experiments, Dr Ann-Marie Cruz for project management support and scientific input, GSK for PBMC samples, and the study participants. This work was supported by PATH’s Malaria Vaccine Initiative or PATH, the US Army Military Infectious Disease Research Program, and the U.S. Army Medical Research and Development Command. The opinions and assertions contained herein are the private views of the authors and are not to be construed as official or as reflecting the views of the US Army, the US Department of Defense, or The Henry M Jackson Foundation for the Advancement of Military Medicine, Inc. This paper has been approved for public release with unlimited distribution.

Senior Editor

  1. Dominique Soldati-Favre, University of Geneva, Switzerland

Reviewing Editor

  1. Urszula Krzych, Walter Reed Army Institute of Research, United States

Publication history

  1. Received: September 15, 2019
  2. Accepted: April 14, 2020
  3. Accepted Manuscript published: April 28, 2020 (version 1)
  4. Accepted Manuscript updated: April 29, 2020 (version 2)
  5. Version of Record published: May 11, 2020 (version 3)

Copyright

This is an open-access article, free of all copyright, and may be freely reproduced, distributed, transmitted, modified, built upon, or otherwise used by anyone for any lawful purpose. The work is made available under the Creative Commons CC0 public domain dedication.

Metrics

  • 638
    Page views
  • 157
    Downloads
  • 0
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)

Further reading

    1. Immunology and Inflammation
    2. Neuroscience
    Nicola S Orefice et al.
    Research Article

    Cortical lesions represent a hallmark of multiple sclerosis and are proposed as a predictor of disease severity. microRNAs are suggested to be important players in the disease pathogenesis and the experimental autoimmune encephalomyelitis animal model. We implemented a mouse model recapitulating more closely the human pathology as it is characterized by both an autoimmune heterogeneity and the presence of cortical lesions, two parameters missing in experimental autoimmune encephalomyelitis. In our model, mice clustered in two groups displaying high or low clinical scores. Upon cortical cytokine injection, lesions appeared with a specific topography while cortical miRNA profiles were altered. These two features differed according to disease severity. We evidenced changes in miRNA regulators and targets suggesting that miRNA alteration had functional repercussions that could explain the differences in cortical lesions. This model represents a crucial tool for the study of both miRNA involvement and cortical lesion formation in disease pathogenesis.

    1. Immunology and Inflammation
    Jeanette Baran-Gale et al.
    Research Article Updated

    Ageing is characterised by cellular senescence, leading to imbalanced tissue maintenance, cell death and compromised organ function. This is first observed in the thymus, the primary lymphoid organ that generates and selects T cells. However, the molecular and cellular mechanisms underpinning these ageing processes remain unclear. Here, we show that mouse ageing leads to less efficient T cell selection, decreased self-antigen representation and increased T cell receptor repertoire diversity. Using a combination of single-cell RNA-seq and lineage-tracing, we find that progenitor cells are the principal targets of ageing, whereas the function of individual mature thymic epithelial cells is compromised only modestly. Specifically, an early-life precursor cell population, retained in the mouse cortex postnatally, is virtually extinguished at puberty. Concomitantly, a medullary precursor cell quiesces, thereby impairing maintenance of the medullary epithelium. Thus, ageing disrupts thymic progenitor differentiation and impairs the core immunological functions of the thymus.