Despite progress made to reduce the global malaria burden, Plasmodium falciparum (Pf) remains one of the leading causes of mortality among children under 5 years of age (Geneva: World Health Organization, 2022). Unfortunately, progress has been impeded by a plateau in malaria control since 2015. Anti-malarial drug resistance (Aydemir et al., 2018; Taylor and Juliano, 2014), and unprecedented logistical challenges during the COVID-19 pandemic that dramatically impacted the distribution of insecticide-impregnated mosquito nets led to an increase in malaria cases in children under 5 years of age from 17.6 million in 2015 to 19.2 million in 2021 in East and Southern African countries (Geneva: World Health Organization, 2022; Ippolito et al., 2021), thus, contributing to reinvigorated prioritization of malaria vaccine initiatives.

After 30 years of development, the first malaria vaccine - RTS,S/AS01E - was approved by the World Health Organization in 2021 for use in children residing in malaria endemic regions. However, the limited efficacy of RTS,S (RTS,S Clinical Trials Partnership, 2015), particularly against severe malaria (RTS,S Clinical Trials Partnership, 2015) has motivated the search for additional vaccine candidate antigens including blood-stage Pf-schizont egress antigen-1 (PfSEA-1A) (Raj et al., 2014) and Pf-glutamic acid-rich protein (PfGARP) (Raj et al., 2020). Antibodies against PfSEA-1A correlate with significantly lower parasite densities in Kenyan adults and adolescents, and substantially reduce schizont replication in vitro (Nixon et al., 2017; Raj et al., 2014). Whereas PfGARP-specific antibodies kill trophozoite-infected erythrocytes in culture and confer partial protection against Pf-challenge in vaccinated non-human primates (Raj et al., 2020). The presence of antibodies against PfGARP was associated with a 2-fold-lower parasite density in Kenyan adults and adolescents (Raj et al., 2020). Due to the complexity of the parasites’ life cycle, there is consensus that a multi-valent vaccine would prove to be more efficacious (Holder, 1999). However, determining which vaccine candidate antigens hold the most promise for inclusion in next-generation malaria vaccines remains a challenge.

Many pathogens and vaccines engender protective antibody responses after a single or few exposures (Amanna et al., 2007) characterized by the production of long-lived plasma cells and memory B cells (Tarlinton and Good-Jacobson, 2013). Affinity maturation takes place in the germinal center (GC) where not only antigen-specific CD4pos T-follicular helper (TFH) cells are required to provide cellular (CD40L) and molecular (IL21, IL4, IL13) signals to trigger B cell proliferation, promote GC maintenance and plasmablast differentiation, but where higher affinity B cells also out-compete B cells with lower affinity for TFH help (Crotty, 2019, 2014). TFH cells are defined by the expression of a combination of markers, starting with chemokine receptor CXCR5 which directs CD4pos T cells from the T cell zone to engagement with follicular B cells (Breitfeld et al., 2000; Crotty, 2011). Once antigen-experienced TFH cells leave the GC, they become circulating TFH (cTFH) cells, and correlate with the generation of long-lasting antibody responses. Expression profiles of CCR6 and CXCR3 categorize cTFH subsets as follow: cTFH1 (CCR6negCXCR3pos), cTFH2 (CCR6negCXCR3neg) and cTFH17 (CCR6posCXCR3neg) (Schmitt et al., 2014); whereas the expression of PD-1, ICOS, CD127 and CCR7 define their functional status: quiescent/central memory cTFH (CCR7highPD-1negICOSnegCD127pos) or activated/effector memory cTFH (CCR7lowPD-1posICOSposCD127low/neg) (Dunham et al., 2008; Gong et al., 2019; Schmitt et al., 2014). In addition, transcription factors (i.e., Bcl6 and cMAF) and cytokines (i.e., IFNγ, IL-4, and IL-21) are crucial to further characterizing the role of each cTFH subset within the context of their interactions with B cells to promote antibody production (Andris et al., 2017; Bélanger and Crotty, 2016; Olatunde et al., 2021; Seth and Craft, 2019). Bcl6 facilitates the production of IL-21 by T cells which aids B cell affinity maturation and antibody production (Liu et al., 2013; Nurieva et al., 2009; Nurieva and Chung, 2010). TFH cells also secrete cytokines that align with their subset classifications (but are not limited to them), such as IFNγ (TFH1), IL-4/IL-13 (TFH2), IL-4/IL-5/IL-13 (TFH13), and IL-17 (TFH17) that direct antibody isotype class switching and mediate effector functions (Bélanger and Crotty, 2016; Crotty, 2014; Gong et al., 2019; Gowthaman et al., 2019; Olatunde et al., 2021; Seth and Craft, 2019).

Several studies have characterized cTFH subsets within the context of both adult and childhood Pf-malaria infections, as well as in healthy malaria-naïve volunteers under controlled malaria infection conditions (Chan et al., 2020). In Mali, where malaria is seasonal, Obeng-Adjei and colleagues showed that TH1-polarized cTFH PD1posCXCR5posCXCR3pos were preferentially activated in children, and were less efficient than the CXCR3neg cTFH in helping autologous B-cells produce antibodies; yet they used U.S. healthy adult cTFH cells for their in vitro assays. Therefore, the authors suggested that promoting Th2-like CXCR3neg cTFH subsets could improve antimalarial vaccine efficacy (Obeng-Adjei et al., 2015). Similarly, a recent study in Papua, Indonesia where malaria is perennial, found that all cTFH subsets from adults showed higher activation and proliferation compared to cTFH cells from children (Oyong et al., 2022). After in vitro stimulation with infected red blood cells, they found that all cTFH subsets were activated in adults accompanied by IL-4 production, whereas only the TH1-polarized cTFH subset responded in children (Oyong et al., 2022). One study in Uganda where malaria is holoendemic with seasonal peaks, was able to show that a shift from TH2-to TH1-polarized cTFH subsets occurred during the first 6 years of life, and was associated with the development of functional antibodies against Pf-malaria; yet appeared to be independent of malaria exposure as this age-associated shift was also observed in a malaria naïve population (Chan et al., 2022). Interestingly, this study also found that a higher proportion of TH17-cTFH cells was associated with a decreased risk of P. falciparum infection the following year, however the authors postulated that this phenomenon could have been driven by previous exposure to the parasite. The observed higher abundance of TH2-like cTFH cells in children younger than 6 years old and the age-associated increase in TH1-cTFH cells achieving the same proportion of TH1-cTFH cells by 6 years of age and into adulthood appeared to be independent of malaria exposure. Of note, these studies were limited by the use of a 2-dimensional gating strategy to classify cTFH subsets and whole parasite activation conditions, either during infections or in vitro stimulation assays, leaving the malaria-antigen specificity of the different cTFH subsets responses undefined. However, these studies highlight the progression in the field of evaluating the role of human TFH cells in anti-malarial immunity.

Antibody levels are an unreliable predictor of malaria vaccine efficacy (Stanisic and McCall, 2021). Thus, establishing a combined immune profile to incorporate other surrogates of protection is warranted. Here, we examined the profile of cTFH subsets using multiparameter spectral flow cytometry (Bentebibel et al., 2015; Wei et al., 2015) against two malaria vaccine candidate antigens (PfSEA-1A (Raj et al., 2014) and PfGARP (Raj et al., 2020)) in a cross-section of children and adults residing in a malaria holoendemic region of Kenya. Our findings revealed significant differences in cTFH subsets between children and adults, where children had more abundant cTFH1-like cells and showed antigen-specific responses from all cTFH types whereas these responses were limited to cTFH17- and 1/17-like in adults. Moreover, this study showed that PfGARP triggered Bcl6 expression across cTFH subsets whereas PfSEA-1A induced more cMAF expression, demonstrating the feasibility of implementing T cell immune correlates to down-select new malaria candidates.


Children had lower anti-PfSEA-1A antibodies compared to adults but similar levels of anti-PfGARP antibodies

This cross-sectional study selected a convenience sample of 7 years old children and adults with a mean age of 22.67 years (ranging from 19 to 30 years old). No statistical difference was observed (p=0.35 after Welch’s test) regarding the absolute lymphocyte count (ALC) between children (mean of 373.1; SD of 118.1) and adults (mean of 335.2; SD of 99.08). Children had significantly lower hematocrit (median of 37%, IQR of 32.7-40.5) compared to adults (median of 44.4%, IQR of 40.5-46), p=0.002 after a two-tailed Mann-Whitney t-test, however these values are within normal ranges adjusting for age.(Pluncevic Gligoroska et al., 2019) Both males and females (sex assigned at birth) were enrolled, with 43% (6/14) and 54% (8/15) being female within the children and adult groups, respectively. Seroprofiles against a panel of commonly used malaria antigens were generated to confirm the history of previous malaria infections for the selected children (Figure 1a). All participants had high IgG levels against AMA-1 and MSP1 (merozoite antigens), confirming at least one malaria infection within their lifetime (O’Flaherty et al., 2021; Yman et al., 2019). IgG antibodies against CSP and CelTOS (liver-stage antigens) were characteristically lower than against blood-stage antigens yet were present in all study children. We observed a clear bimodal distribution in antibody levels against HRP2 with half of the children having “high-HRP2” versus “low-HRP2” IgG levels, possibly a reflection of recent malaria history, since it has been suggested that HRP2-specific antibodies are short-lived and could serve as a surrogate for a recent infection (Turnbull et al., 2022). Serological profiles were then assessed for two Pf-malaria antigens being considered as potential vaccine candidates (Raj et al., 2020, 2014), PfSEA-1A and PfGARP (Figure 1b). We found that children had a significantly lower median level of IgG against PfSEA-1A compared to adults (p<0.0001), whereas median levels to PfGARP were similarly high for adults and children, yet with a broad range of reactivity.

Pf-malaria IgG sero-profiles for children and adults.

(a) IgG antibody levels against AMA1, MSP1, HRP2, CelTos and CSP for children (n=14). The bar plots indicate mean with SD. (b) IgG antibody levels against PfSEA-1A and PfGARP comparing children (n=15) and adults (n=15). The bar plots indicate mean with SD. Net MFI values are the antigen-specific MFI values minus the BSA background. Mann-Whitney tests were performed.

The overall abundance of the CD4posCXCR5pos cells is unaltered by in vitro antigen stimulation

To determine whether in vitro antigen stimulation with PfSEA-1A or PfGARP altered the abundance of total cTFH cells, we compared cTFH cells from adults and children using a FlowSOM unbiased clustering analysis and EMBEDSOM dimensional reduction based on common lineage markers assessed by spectral flow cytometry from 87,812 lived lymphocytes from each sample: CD8pos, CD4pos, CD4posCXCR5pos, and CD4posCD25pos (Figure 2a). As expected after a short stimulation (6 hrs), their overall abundances were similar across conditions for both adults (Figure 2b), and children (Figure 2c). This observation was confirmed by EdgeR statistical analysis (Supplemental Figure 2) and demonstrated that in vitro stimulation did not preferentially expand the T cell populations on which we based our subsequent analyses.

Clusters visualization of immune cell types and their abundance.

(a) A 100 node and 75-meta-clusters FlowSOM tree was generated on live lymphocytes from all our participants (n=29 children and adults), highlighting CD3posCD8pos, CD3posCD4pos, CD4posCXCR5pos and CD4posCD25pos cells. The 5 nodes of the CD4posCXCR5posCD25neg population are circled and will be used for the downstream analysis. The colored scale was based on the median arcsinh-transformed marker expression. UMAP plots showing 5 clusters defined as follow: CD14/CD19pos (orange), CD4pos (blue), CD4posCXCR5pos (red), CD8pos (purple) and CD4posCD25pos (green) from (b) adults (n=15) (c) children (n=14) PBMCs after no-stimulation, PfSEA-1A and PfGARP stimulation, respectively from left to right.

An unbiased clustering analysis identifies twelve distinct cTFH meta-clusters

Numerous markers and a two-dimensional gating strategy have previously been used to determine the frequency of cTFH subsets (Chan et al., 2022; Obeng-Adjei et al., 2015; Oyong et al., 2022). To simultaneously account for the expression of 17 markers required to define cTFH subsets, we used FlowSOM unbiased clustering analysis to determine the frequency of cTFH subsets from a pool of 1,000 CD3posCD4posCXCR5posCD25neg cells from each sample (total of 13,000 CD3posCD4posCXCR5posCD25neg cells in both children and adult). Based on CXCR3 and CCR6 expression as well as the expression of effector/memory/activation markers (CCR7, CD127, PD1, ICOS), cytokines (IFNγ, IL-4, and IL-21) and transcription factors (Bcl6, cMAF), we initially identified 15 meta-clusters within the CD4posCXCR5pos T cells (Figure 3a). First, we found different CXCR5 expression levels between meta-clusters (Figure 3b), suggesting different capabilities of cTFH to migrate to the lymph nodes and, therefore, to interact with B-cells, i.e., cTFH cells with lower CXCR5 expression would be less likely to reach the B-cell zone. Data presented in Figure 3 are from children, however, similar observations were made for adults (Supplemental Figure 3). Because CD45RAposCXCR5pos cells are likely naive cells with transient low expression of CXCR5 yet expressing high CD45RA, we excluded 3 meta-clusters using these criteria (i.e., MC12, MC14, and MC15) (Figure 3c). Then, using the overall expression of CXCR3 and CCR6 across the cTFH subsets (Figure 3d and 3e), we identified the remaining 12 clusters as follows: MC01 and MC02 were cTFH2-like; MC06 and MC07 were cTFH1-like; MC09 and MC11 were cTFH1/17-like; MC10 and MC13 were cTFH17-like;. However, CXCR3 expression was not clearly delineated for some subsets and did not align with the conventional CCR6 vs CXCR3 cytoplot (Figure 3f and 3g, Supplemental Figure 4). Based on the heatmap (Figure 3f), MC03, MC04, MC05 and MC08 clusters appear closer to cTFH2-like MC01 and MC02 than cTFH1-like clusters MC06 and MC07 suggesting that they might be part of the cTFH2-like subset. But, based on their intensity of CXCR3 expression and their distribution across the CCR6 vs CXCR3 cytoplot (Supplemental Figure 4) we defined MC03, MC04, MC05, MC08 as “undetermined” and would require additional cytokine and transcription factor characterization to fully categorize them.

Characterization of TFH subsets by CXCR5, CXCR3, CCR6 and CD45RA expression.

(a) A new 25 nodes and 15 meta-clusters FlowSOM tree was generated from the 5 CD4posCXCR5posCD25neg nodes shown in Figure 2A (CXCR5 expression tree). The colored scale of CXCR5 expression was based on the median arcsinh-transformed with red as the highest and deep blue the absence of CXCR5 expression. (b) Box-plots showing CXCR5 expression across all cTFH-like meta-clusters from children (n=13) with no stimulation (blue), and after in vitro stimulation with PfGARP (green) or PfSEA-1A (pink). Fifty percent of the data points are within the box limits, the solid line indicates the median, the dash line the mean and the whiskers indicate the range of the remaining data with outliers being outside that range. Similar box-plots are shown for (c) CD45RA, (d) CXCR3, and (e) CCR6 expression across all cTFH-like meta-clusters. (f) Clustered heatmap showing the median arcsinh-transformed expression for CCR6, CD45RA, CXCR3 and CXCR5 across meta-clusters, red showing the highest expression and blue the lowest. (g) Cytoplots of CCR6 vs CXCR3 expression where MC01 is in blue, MC07 in pink, MC11 in yellow and MC13 in green.

Heterogeneity of activation/maturation markers within cTFH subsets

By assessing the expression of CCR7, PD1, CD127, and ICOS (Figure 4) and following the three-dimensional expression patterns adapted from Schmitt et. al (Schmitt et al., 2014) (Supplemental Figure 5), we determined the activation state of each cTFH meta-cluster and which markers created novel subsets. Data represented in Figure 4 are from children; however, similar observations were made for adults (Supplemental Figure 6). As expected, none of the extracellular markers showed significant differences in expression patterns after a short 6 hrs stimulation, thus representing the cTFH repertoire present within our study participants. Interestingly, the cTFH1or2 subset (MC03) was the only meta-cluster with high expression of PD1 (Figure 4a) and accompanied by high ICOS (Figure 4b), low CCR7 (Figure 4c), and low expression of CD127 (Figure 4d) suggesting that MC03 was an activated/effector cTFH1or2 subset. Low CCR7 and high ICOS but with low CD127 and PD1 expression indicated that MC04 was a less activated/effector cTFH1or2 subset compared to MC03. Additional nuances in expression pattern were also observed for other cTFH1-like meta-clusters: cTFH1-like MC06 had an activated/effector profile whereas cTFH1-like MC07 had a quiescent/effector profile; cTFH1or2-like MC05 and MC08 were defined as quiescent/memory (overall high CCR7 expression) yet with very low expression of ICOS for MC05. Likewise, cTFH2-like and cTFH17-like meta-clusters also displayed heterogeneity: cTFH2-like MC01 had an activated/effector profile whereas cTFH2-like MC02 seemed to be a quiescent/effector subset; cTFH1/17-like subsets MC09 and MC11 overall had an activated profile even though they had higher expression of CCR7 and CD127 compared to other subsets, suggesting an activated/memory phenotype; and finally cTFH17-like MC10 seemed quiescent whereas cTFH17-like MC13 had an activated/effector profile. Of note, because of our short stimulation time (6H) we were not able to assess CD40L expression (Supplemental Figure 7). However, our analysis methods revealed a higher degree of previously unrecognized heterogeneity within circulating cTFH cells.

cTFH subset activation state determined by PD1, ICOS, CCR7, and CD127.

Box-plots showing (a) PD1, (b) ICOS, (c) CCR7, and (d) CD127 expression across all cTFH-like meta-clusters from children (n=13) with no stimulation (blue), and after in vitro stimulation with PfGARP (green) or PfSEA-1A (pink). Fifty percent of the data points are within the box limits, the solid line indicates the median, the dash line the mean and the whiskers indicate the range of the remaining data with outliers being outside that range.

Activated cTFH1or2-, cTFH1-, and quiescent cTFH1or2-like subsets were more abundant in children

After having deconvoluted cTFH cells into 12 subsets, we next wanted to determine whether their abundance differed by age or after antigen stimulation. Using UMAP visualization, we found that cTFH dimensional reduction was contiguous as meta-clusters merged with each other; in addition, there were notable differences between adults and children (Figure 5a). We found that activated PD1high cTFH1or2-like (MC03), activated cTFH1-like (MC06), and quiescent ICOShigh cTFH1or2-like (MC08) subsets were significantly more abundant in children compared to adults regardless of the stimulation conditions (Figure 5b, 5c and 5d, p<0.05). In contrast, quiescent cTFH2-like cells (MC02) were significantly more abundant in adults compared to children after both PfSEA-1A and PfGARP stimulations (Figure 5c and 5d, p<0.05). Interestingly, following PfGARP stimulation, the activated cTFH1/17-like subset MC09 became more abundant in children compared to adults (Figure 5d, p<0.05 with FDR=0.08); but no additional subsets expanded after PfSEA-1A stimulation (Figure 5c). Of note, activated PD1low cTFH1or2-like (MC04) seemed more abundant in non-stimulated adults and PfGARP stimulated children, but, because these observations were not present in all the participants, they did not achieve statistical significance in EdgeR.

Differences in abundance of antigen-specific cTFH meta-clusters for adults and children.

(a) UMAP plots showing the 12 different cTFH meta-clusters in adults (top three plots, n=13) and in children (bottom three plots, n=13) in the absence of stimulation or after in vitro stimulation with PfSEA-1A or PfGARP, from left to right, respectively. Colored circles highlight the meta-clusters showing differences in their abundance between adults and children for each condition. An EdgeR statistical plot was performed to assess the change in abundance of the 12 meta-clusters between adults and children after (b) no stimulation, (c) PfSEA-1A and (d) PfGARP stimulations. EdgeR plots indicate which meta-clusters are significantly abundant between two groups by using green color dots. The Y-axis is the - log10(p-value) and the X-axis is the log(FC). Green dots were statistically significant (p<0.05). Numbers next to the dots indicate the meta-cluster. (e) An abundance heatmap indicates the percentage (black numbers) of each meta-cluster within the total number of CD3posCD4posCXCR5posCD25neg cells for adults and children (concatenated from 13 participants in each group) under the different conditions: no stimulation (Neg) or stimulation with PfSEA-1A or PfGARP. The color scale ranges from high expression (red) to low/no expression (blue). The star in the heatmap indicates which meta-cluster is significantly abundant in children or adults based on the EdgeR results.

The abundance heatmap (Figure 5e) reiterates the differences observed between children and adults and highlights important considerations when assessing the potential role of each cTFH subsets in assisting with cognate antibody production. Overall, the most common cTFH subset in both children and adults is quiescent cTFH2-like cells (MC02, 24.1% and 41.2% respectively). However, the antigen-specific differences in cTFH subset abundance for children (MC09 for PfGARP) and for adults (MC02 for both PfSEA-1A and PfGARP) suggest that children engage different cTFH cells as they are developing immunity. Of note, only the activated PD1low cTFH1or2-like (MC04) was less abundant in children with low compared to high HRP2 antibody levels (Supplemental Figure 8), suggesting that this subset may be involved in short-term antibody production. Overall, this comprehensive examination of the abundance of cTFH subsets demonstrates important diversity based on age and malaria-antigen specificity.

PfSEA-1A and PfGARP induced IL4, Bcl6 and cMAF from a broad range of cTFH subsets in children

Because the children in this cohort were 7 years of age and resided in a malaria holoendemic area, they had ample time to develop premunition. To assess antigen-specific cytokine and transcription factor expression signatures and further characterize cTFH subsets we generated clustered heatmaps of the MFI of each analyte (IFNγ, IL-4, IL-21, Bcl6 and cMAF) for children (Supplemental Figure 9a) and adults (Supplemental Figure 9b). Next, using these MFI data, we performed Wilcoxon paired tests to compare PfSEA-1A and PfGARP stimulation to unstimulated cells (Figure 6a and 6b, respectively). Significant differences in these expression profiles allowed us to further characterize meta-clusters into three main groups. Group 1: cTFH2-like (activated MC01 and quiescent MC02) and activated cTFH1or2-like (PD1highMC03 and PD1lowMC04); Group 2: activated and quiescent cTFH1-like (MC06 and MC07); Group 3: activated cTFH1/17-like (MC11) and cTFH17-like (quiescent MC10 and activated MC13). For children, PfSEA-1A and PfGARP induced robust IL4 expression in 9 out of 12 cTFH meta-clusters (p-values ≤ 0.0105); although the composition of which cTFH subsets were engaged differed slightly by antigen (quiescent cTFH1or2-like ICOSlow MC05 versus quiescent cTFH2-like MC02, respectively). In contrast to IL-4, we observed no change in expression for IFNγ or IL-21 after in vitro antigen stimulation (except for quiescent cTFH1or2-like ICOShigh MC08, p-value = 0.0391), suggesting these cytokines are not informative to define the development of antigen-specific cTFH subsets in children.

Diverse Pf-malaria antigen-specific expression patterns of cTFH defining cytokines and transcription factors for children.

Heatmaps of Wilcoxon-paired two-tailed t-test p-values are shown for cTFH meta-clusters comparing (a) PfSEA-1A and (b) Pf-GARP versus unstimulated PBMCs from children (n=13) for each cytokine (IFNγ, IL4, and IL21) and transcription factor (Bcl6 and cMAF). The color-scale indicates the significance of the p-value: white (non-significant, p>0.05), yellow (0.05>p>0.02), orange (0.02>p>0.005), and red (highly significant, p<0.005). The down-directed arrow indicates a decrease of expression from unstimulated to stimulated condition whereas no arrow indicates an increase of expression from unstimulated to stimulated condition. The cTFH meta-clusters that co-expressed transcription factors were grouped as follows: Group 1 (blue line), Group 2 (green line), and Group 3 (purple line). Bar plots indicating mean with SD of the median intensity fluorescence (MFI) of (c) IFNγ, (d) Bcl6, and (e) cMAF for the cTFH meta-clusters showing significant statistical differences between PfSEA-1A and PfGARP stimulations. p-values from Wilcoxon-paired two-tailed t-tests are indicated.

We found that PfSEA-1A (Figure 6a) and PfGARP (Figure 6b) induced similar Bcl6 and cMAF expression profiles from some of the same cTFH subsets: both transcription factors were expressed by activated cTFH2-like MC01 and quiescent cTFH17-like MC10, but only cMAF was expressed in activated and quiescent cTFH1-like MC06, MC07 and activated cTFH1/17-like MC11. However, PfSEA-1A induced Bcl6 and cMAF from activated PD1highMC03 whereas PfGARP only induced Bcl6. In contrast, quiescent cTFH2-like MC02 did not seem to respond to PfSEA-1A (only cMAF was significant, p=0.0266, Figure 6a, Supplemental Figures 10 and 11), whereas PfGARP stimulation induced significantly more IL-4, Bcl6, and cMAF compared to unstimulated cells (p=0.0081, p=0.0017, and p=0.0081, respectively, Figure 6b, Supplemental Figures 10 and 11). We then compared the response intensity between PfSEA-1A and PfGARP stimulations, and found significant differences in IFNγ, IL21, Bcl6, and cMAF expression levels (Figure 6c, 6d, 6e and 6f, respectively). In Group 1, Bcl6 expression was significantly higher within activated and quiescent cTFH2-like (MC01 and MC02) as well as activated PD1high cTFH1or2-like (MC03) meta-clusters after PfGARP compared to PfSEA-1A stimulation (p=0.0266, p=0.0134 and p=0.0327, respectively, Figure 6e). Within Group 2, IFNγ and Bcl6 were found to be highly expressed by activated cTFH1-like (MC06) after PfGARP compared to PfSEA-1A stimulation (p=0.0093 and p=0.0024, respectively, Figure 6c and 6e). Finally, within Group 3, activated cTFH1/17-like (MC11) expressed higher cMAF after PfSEA-1A stimulation compared to PfGARP (p=0.0186, Figure 6f). Interestingly, PfGARP induced significantly more IL21 within the quiescent ICOSlow cTFH1or2-like (MC05) meta-cluster compared to PfSEA-1A stimulation (p=0.0186, Figure 6d). Non-significant differences in cytokines and transcription factors expressed by cTFH subsets between conditions are shown in Supplemental Figures 12 and 13, respectively.

PfGARP induced IL4, Bcl6 and cMAF expression in activated cTFH1/17- and cTFH17-like subsets in adults

A similar heatmap was generated for adult expression profiles comparing PfSEA-1A and PfGARP stimulation to unstimulated cells (Figure 7). Here, we found that both PfSEA-1A and PfGARP induced significant expression of both IFNγ and IL-4 for activated (MC01) and quiescent (MC02) cTFH2-like cells (Figure 7a and 7b, Supplemental Figure 12). Whereas, PfGARP also induced IL-4 and IFNγ expression from quiescent and activated cTFH17-like (MC10 and MC13), in addition to Bcl6 and cMAF for MC13 (Figure 7b, Supplemental Figure 13). This observation was surprising because IFNγ-expression is commonly used to categorize the cTFH1 subset (Group 2); however, as shown earlier, quiescent and activated cTFH17-like (MC10 and MC13) did not express CXCR3 (Figure 3d and 3e). In contrast, activated cTFH1/17-like (MC11) only responded to PfGARP (Figure 7b, Supplemental Figure 12), expressing higher levels of IL-4, IL-21, Bcl6, and cMAF. Finally, while assessing the differences between the two malaria antigens, we found that PfGARP induced more Bcl6 expression than PfSEA-1A within quiescent cTFH2-like (MC02, p=0.0327) and quiescent ICOSlow cTFH1or2-like (MC05, p=0.0105), as well as within quiescent cTFH17-like (MC10 p=0.0479) and activated cTFH1/17-like (MC11 p=0.0020) (Figure 7d), the latest expressing higher IL21 as well after PfGARP compared to PfSEA-1A stimulation (p=0.0273, Figure 7c); whereas PfSEA-1A induced cMAF within activated cTFH2-like (MC01, p=0.0134) but Pf-GARP did not (Figure 7e). Overall, the main observation for adults is that PfSEA-1A predominantly induced IL-4 from slightly more than half of the cTFH clusters, whereas PfGARP induced a broader range of cytokines and transcription factors but within the activated cTFH1/17-like (MC11), quiescent and activated cTFH17-like (MC10 and MC13) similar to children. This analysis shows clear differences in cTFH subset specificity by malaria antigen and cTFH subset engagement by age group, with the cTFH repertoire becoming more restricted in adults compared to children.

Limited Pf-malaria antigen-specific expression patterns of cTFH defining cytokines and transcription factors for adults.

Heatmaps of Wilcoxon-paired two-tailed t-test p-values are shown for cTFH meta-clusters comparing (a) PfSEA-1A or (b) Pf-GARP versus unstimulated PBMCs from adults (n=13) for each cytokine (IFNγ, IL4, and IL21) and transcription factors (Bcl6 and cMAF). The color-scale indicates the significance of the p-value: white (non-significant, p>0.05), yellow (0.05>p>0.02), orange (0.02>p>0.005) and red (highly significant, p<0.005). The down-directed arrow indicates a decrease of expression from unstimulated to stimulated condition whereas no arrow indicates an increase of expression from unstimulated to stimulated condition. The only cTFH meta-clusters that expressed transcription factors were in Group 3 (purple box). Bar plots indicating mean with SD of the median intensity fluorescence (MFI) of (c) Bcl6, and (d) cMAF for the cTFH meta-clusters showing significant statistical differences between PfSEA-1A and PfGARP stimulations. p-values from Wilcoxon-paired two-tailed t-tests are indicated.

Activated cTFH1or2-like subset is more abundant in participants with high anti-PfGARP antibodies

As shown in Figure 1b, a broad range of anti-PfGARP IgG antibody levels were found within both children and adults. Thus, we wanted to determine whether the abundance of any of the cTFH subsets was associated with the level of anti-PfGARP IgG antibodies. When stratifying by high versus low anti-PfGARP IgG antibody levels, we found that an activated cTFH1or2-like subset (MC04) was more abundant in participants with high levels of anti-PfGARP IgG for both children (Figure 8a) and adults (Figure 8b) after PfGARP stimulation compared to participants with low levels or an absence of anti-PfGARP IgG. However, even though the p-values were significant for both children and adults (p=0.02 and p=0.004, respectively), the FDR was less than 0.05 only for the adults (FDR=0.018). This suggests that this particular subset might be important for the generation of anti-PfGARP antibodies. Interestingly, activated cTFH1-like (MC06) and quiescent cTFH1-like (MC07) were more abundant for adults with low or no anti-PfGARP IgG antibodies compared to those with high levels (p=0.0024 with FDR=0.0185 and p=0.0034 with FDR=0.0185, respectively) consistent with previous observations describing TFH1 subsets as not efficient help for antibody production (Obeng-Adjei et al., 2015).

Abundance of cTFH meta-clusters stratified by anti-PfGARP IgG antibody levels.

UMAP and EdgeR analysis of (a) children (n=13) and (b) adults (n=13) where significant differences (p<0.05) in the abundance of the cTFH meta-clusters are circled on high versus low PfGARP antibody level (left to right) UMAP and green dots on the volcano plot (far right).


The overall aim of this study was to define cTFH subsets using unbiased clustering analysis and assess their malaria antigen-specific (PfSEA-1A and PfGARP) (Duffy and Patrick Gorres, 2020; Raj et al., 2020, 2014) profiles for adults and children residing in a malaria holoendemic area of Kenya. Contrary to previous publications (Obeng-Adjei et al., 2015; Oyong et al., 2022), our study found that children did not only respond via their cTFH1-like subsets but also engage a broader spectrum of cTFH subsets. In fact, cytokine and transcription factor profiles for children involved cTFH1-, cTFH2-, cTFH17- and cTFH1/17-like subsets (summarized in Figure 9) whereas in adults, the dominant antigen-specific memory response was from cTFH17- and cTFH1/17-like subsets but only for PfGARP. Thus, revealing a potential difference in engaging cTFH help between the two malaria vaccine candidates evaluated in this study.


Illustration of combined findings showing remarkable differences in cTFH subset abundance and Pf-malaria antigen-specific cytokine and transcription factor responses between children and adults residing in a malaria holoendemic region of Kenya. Made with BioRender.

© 2024, BioRender Inc. Any parts of this image created with BioRender are not made available under the same license as the Reviewed Preprint, and are © 2024, BioRender Inc.

Several key points can be made from our results. First, it may be too soon in the field of cTFH biology to select only one or two types of cTFH subset(s) to measure when evaluating malaria vaccine candidates, especially if we miss essential, albeit transient, correlates of protection for children which may differ in adults. Longitudinal studies with clinical outcomes are needed to determine the impact of the changing dynamics of cTFH subset adaptations after repeated malaria infections that lead to premunition and parasite clearance. A Ugandan study by Chan et al (Chan et al., 2022) demonstrated an age-associated change in cTFH subsets independent of malaria and, thus, supports the importance of accounting for age when evaluating antigen-specific cTFH profiles. Second, our study reveals that adults have a robust cTFH17- and cTFH1/17-like subset response to PfGARP but not to PfSEA-1A. In addition to demonstrating differences between malaria antigens, this supports the premise that cTFH17 cells are important for maintaining immunological memory (Gao et al., 2023), which has intriguing implications for evaluating malaria vaccine efficacy. Third, we found a correlation between cTFH1or2-like (MC04) and high PfGARP antibodies for both children and adults, indicating that a cTFH subset accompanied by high antibody levels could serve as a potential biomarker of protection. More studies are needed to explore this association; however, we postulate that a coalition of cTFH subsets might be engaged to develop long-lived antibody responses and, therefore, categorizing subsets as efficient or inefficient might be context dependent (Obeng-Adjei et al., 2015).

As the field of computational immunology and unbiased clustering analyses evolves, it presents an ongoing challenge to meaningfully define cTFH subsets. Here, we intentionally chose to use cTFH1or2-like nomenclature for meta-clusters MC03, MC04, MC05, and MC08 (Figure 9) because of their low CXCR3, but not null, and robust IL4 expression (instead of IFNγ along with Bcl6, and cMAF). CXCR3 expression leans towards a cTFH1-like polarization, whereas IL4 expression indicates a cTFH2-like profile when we follow the most commonly used cTFH classification (Bentebibel et al., 2015; Olatunde et al., 2021; Schmitt et al., 2014; Seth and Craft, 2019). This difference is crucial as cTFH2 cells are described as good promoters for functional GC antibodies (Chan et al., 2020; Oyong et al., 2022) whereas cTFH1 cells are not (Hansen et al., 2017; Obeng-Adjei et al., 2015; Oyong et al., 2022). Therefore, their definitive classification will require a more in depthr investigation. Even though we used malaria antigen-specific stimulation, and not a Pf-lysate or infected red blood cells (iRBCs) as previously described (Chan et al., 2022; Obeng-Adjei et al., 2015; Oyong et al., 2022), adults had significantly more of the quiescent cTFH2-like subset (MC02) compared to children who instead had significantly more of the activated cTFH1-like subset (MC06), the latest being consistent with previous publications (Obeng-Adjei et al., 2015; Oyong et al., 2022), and more PD1high activated and ICOShigh quiescent cTFH1or2-like subsets (MC03 and MC08, respectively). However, even when classifying the cTFH meta-clusters as cTFH1, cTFH2, cTFH17, or cTFH1/17; the most abundant cTFH subset was cTFH2 cells for adults (more than 50%) contradicting a previous study showing only ∼20% of cTFH2 cells for the same age range (Chan et al., 2022) As suggested by Gowthaman’s TFH model (Gowthaman et al., 2019), TFH1 cells are involved in the development of neutralization antibody responses to viruses and bacteria whereas parasites, such as helminths, lead to a TFH2 response. These observations reinforce the need to assess TFH subset profiles stratified by exposure to potentially immune-modulating co-infections.

Transcription factors cMAF and Bcl6 play essential roles in TFH development and function (Imbratta et al., 2020; Liu et al., 2012; Nurieva et al., 2009). cMAF induces the expression of various molecules, such as ICOS, PD1, CXCR5, IL4, and IL21, which are all essential for TFH function (Imbratta et al., 2020). It was, therefore, not surprising to find that cMAF significantly increased after antigen stimulation for most of the cTFH subsets in children. However, again, this expression profile was only observed in the activated cTFH1/17-like subset (MC11) after PfGARP stimulation in cells from adults, supporting a role for TFH17 cells in the maintenance of immunological memory (Gao et al., 2023). Importantly, cMAF cooperates with Bcl6 in TFH development and function and is essential to establish an efficient GC response (Nurieva et al., 2009; Yu et al., 2009). Bcl6 is also described as being expressed by mature TFH cells, inducing the expression of CXCR5 and PD1, but it can also regulate IFNγ and IL17 production (Crotty, 2014; Johnston et al., 2009; Nurieva et al., 2009; Yu et al., 2009). In our study, PfGARP induced a highly significant increase of Bcl6 in cTFH2- like and cTFH17-like subsets in children, suggesting that Pf-GARP may be a better candidate to trigger an efficient humoral response in children compared to PfSEA-1A. Of note, Blimp1, a Bcl6 antagonist (Johnston et al., 2009), was absent from our flow panel and would be of interest to assess in future studies.

There are several limitations of human immune profiling studies. Similar to other such studies, we used peripheral blood and, thus, were only able to provide a snapshot of the cTFH cells. Study participants did not have blood stage malaria infections at the time of blood collection and children were 7 years of age; thus, our profiles were by design meant to reflect cTFH memory recall responses and demonstrate differences between children and adults. As this was not a birth cohort study design, it is possible that the number of cumulative malaria infections could have confounded the association between malaria and cTFH subsets causing a certain cTFH meta-cluster to arise. Most human immunology studies are unable to assess tissue-resident TFH or TFH in the lymph nodes, and, therefore, we are only able to speculate on the composition of TFH cells observed in children that were perhaps short-lived helper and not maintained as T cell memory, or alternatively may have trafficked from the blood into tissues over time (Potter et al., 2021). Using an in vivo non-human primate model, Potter et al., showed an entry rate of lymphocyte subsets into peripheral lymph nodes per hour of 1.54% and 2.17% for CD4pos central memory (CD45negCCR7pos) and effector memory (CD45negCCR7neg) T cells, respectively (Potter et al., 2021), both of which may include central and effector memory cTFH cells. With this in mind, cTFH cells seem to be at a crossroad with multiple possible fates. The first one being the recirculation of cTFH cells from lymph node to lymph node whereby a peripheral blood sampling captures migratory cTFH subsets, which could explain the cTFH subsets we observed still expressing cMAF and Bcl6. A decade ago, it was hypothesized that once a TFH cell leaves the lymph node, it can become a PD1low memory cTFH cell with the possibility of returning to a germinal center TFH stage after a secondary recall response (Crotty, 2014). Other possible fates include becoming a PD1neg memory cTFH with the same path after recall or progressing to a non-TFH cell (Crotty, 2014). More recently, a few studies showed circulating CXCR5negCD4pos cells with TFH functions, such as production of IL21 and the capability to help B cells in systemic lupus erythematosus and HIV-infected individuals (Bocharnikov et al., 2019; Del Alcazar et al., 2019), and importantly they mapped these circulating CXCR5negCD4pos cells to an original lymph node CXCR5pos TFH subset (Del Alcazar et al., 2019), suggesting another possible outcome for TFH cells outside secondary lymphoid organs. To date, no clear destiny has been established for cTFH cells in humans, thus, highlighting the importance of including a complete panel of cTFH subsets to continue to improve our understanding of their respective roles against different pathogens and eliciting long-lived vaccine-induced antibody responses.

Other limitations of this study include not measuring other cytokines (i.e., IL5, IL13, IL17) and transcription factors (i.e., T-bet, BATF, GATA3, RORγt) that have been used in other studies to fully characterize cTFH subsets (Crotty, 2014; Gowthaman et al., 2019). However, even with our small sample size we demonstrated significant age-associated differences in malaria antigen-specific responses from different cTFH subsets. To minimize false positive results that can arise when using algorithms for computational analyses, we ran the statistical tests in triplicate and the clustering algorithm in duplicate, to validate our findings.

In summary, our study provides additional justification for the resources needed to conduct cellular immunological studies that include cTFH cell signatures and provides insight into which types of cTFH subset during a person’s lifespan assist germinal center B cells to produce long-lived plasmocytes and functional antibodies (Moormann et al., 2019) against malaria. This is particularly important when selecting immune correlates of protection that could be used to predict the efficacy of the next generation of Pf-malaria vaccine candidates within various study populations.


Study populations and ethical approvals

Adults and children were recruited from Kisumu County, Kenya which is holoendemic for Pf-malaria. Written informed consent was obtained from each adult participant and children guardian. An abbreviated medical history, physical examination, and blood film were used to ascertain health and malaria infection status at the time of blood sample collection. Participants were also life-long residents of the study area with an assumption that they naturally acquired immunity to malaria. This study was conducted prior to the implementation of any malaria vaccines. Participants were eligible if they were healthy and not experiencing any symptoms of malaria at the time venous blood was collected. For this cross-sectional immunology study, we selected fourteen 7-years old children from a larger age-structured prospective cohort study (enrollment age range 3-7 years) and fifteen Kenyan adults. We selected 7-year olds because of the age-dependent shift in major cTFH subsets occurring after 6-years of age (Chan et al., 2022), and as a comparable age published by other studies (Obeng-Adjei et al., 2015), in order to maximize our ability to measure antigen-specific differences in TFH subsets.

Ethical approvals were obtained from the Scientific and Ethics Review Unit (SERU) at the Kenya Medical Research Institute (KEMRI) reference number 3542, and the Institutional Review Board at the University of Massachusetts Chan Medical School, Worcester, MA, USA, IRB number H00014522. Brown University, Providence, RI, USA signed a reliance agreement with KEMRI.

Plasma and PBMCs isolation

Venous blood was collected in sodium heparin BD Vacutainers and processed within 2 hours at the Center for Global Health Research, KEMRI, Kisumu. Absolute lymphocyte counts (ALC) were determined from whole blood using BC-3000 Plus Auto Hematology Analyzer, 19 parameters (Shenzhen Mindray Bio-Medical Electronics Co.). After 10 mins at 1,000g spin, plasma was removed and placed for storage at -20°C and an equivalent volume of 1X PBS was added to the cell pellet. Peripheral blood mononuclear cells (PBMCs) were then isolated using Ficoll-Hypaque density gradient centrifugation on SepMate (StemCell). PBMCs were frozen at 5x106 cells/ml in a freezing medium (90% heat-inactivated and filter-sterilized fetal bovine serum (FBS) and 10% dimethyl sulfoxide (Sigma)) and chilled overnight in Mr. Frosty™ containers at -80°C prior being transferred to liquid nitrogen. Transport to the USA used an MVE vapor shipper (MVE Biological Solutions) to maintain the cold chain.

In-vitro stimulation assay

PBMCs were thawed in 37°C filtered-complete media (10% FBS, 2mM L-Glutamine, 10mM HEPES, 1X Penicillin/Streptomycin) and spun twice before resting overnight in a 37°C, 5% CO2 incubator. PBMCs were counted using Trypan Blue (0.4%) and hemocytometer, and the cell survival was calculated. Our samples showed a median of 94.6% live cells [25% percentile of 92%; 75% percentile of 97%]. Using a P96 U-bottom plate, 1x106 PBMCs per well were placed in culture with one of the following stimulation conditions: PfSEA-1A (Raj et al., 2014) (5 µg/ml) or PfGARP (Raj et al., 2020) (10 µg/ml) both produced in the Kurtis lab (Brown University); SEB (1 µg/ml; EMD Millipore) as a positive control; sterile water (10µl, the same volume used to reconstitute PfSEA and PfGARP) as a negative control. A pool of anti-CD28/anti-CD49d (BD Fast-Immune Co-Stim following the manufacturer’s instructions), GolgiSTOP (0.7 µg/ml), and GolgiPLUG (0.1 µg/ml) (BD Biosciences) were added to each well before incubating cells at 37°C for 6-hours.

Cell staining and Flow cytometry

A multiparameter spectral flow cytometry panel was used to characterize cTFH cell subsets: CCR6-BV421 (RRID: AB_2561356), CD14-Pacific Blue (RRID: AB_830689), CD19-Pacific Blue (RRID: AB_2073118), CCR7-BV480 (RRID: AB_2739502), IFNγ-BV510 (RRID: AB_2563883), CD127-BV570 (RRID: AB_10900064), CD45RA-BV605 (RRID: AB_2563814), PD1-BV650 (RRID: AB_2738746), CXCR3-BV711 (RRID: AB_2563533), CD25-BV750 (RRID: AB_2871896), CXCR5-BV785 (RRID: AB_2629528), Bcl6-AF488 (RRID: AB_10716202), CD3-Spark Blue 550 (RRID: AB_2819985), CD8-PerCP-Cy5.5 (RRID: AB_2044010), IL21-PE (RRID: AB_2249025), IL4-PE-Dazzle (RRID: AB_2564036), CD4-PE-Cy5 (RRID: AB_314078), ICOS-PE-Cy7 (RRID: AB_10643411), cMAF-eFluor 660 (RRID: AB_2574388), CD40L-AF700 (RRID: AB_2750053), and Zombie NIR (BioLegend cat# 423106) for Live/Dead staining. Data was acquired on a Cytek Aurora with 4-lasers (UMass Chan Flow Core Facility) using SpectroFlo® software (Cytek) and compensation for unmixing and fluorescence-minus-one controls. Quality control of the data was performed using SpectroFlo®, and the multi-parameter analysis was performed with OMIQ data analysis software ( Thus, we assessed the expression of markers commonly used to define the following different cTFH (CD4posCD25negCXCR5pos) subsets: cTFH1-like (CCR6negCXCR3pos), cTFH2-like (CCR6negCXCR3neg), and cTFH17-like (CCR6posCXCR3neg) (Schmitt et al., 2014), as well as quiescent/central memory cTFH (CCR7highPD-1negICOSneg) or activated/effector memory cTFH cells (CCR7lowPD-1posICOSpos) (Gong et al., 2019; Schmitt et al., 2014). Representative cytoplots can be found in Supplemental Figure 1.

Multiplex suspension bead-based serology assay

To measure plasma IgG antibody levels to PfSEA-1A (Raj et al., 2014) and PfGARP (Raj et al., 2020), we used a Luminex bead-based suspension assay as previously published (Cham et al., 2009; Forconi et al., 2018). In addition, previous Pf exposure was determined using recombinant proteins to blood-stage malaria antigens: apical membrane antigen 1 (AMA1) and merozoite surface protein (MSP1), HRP2, CelTos, CSP (gifts from Sheetji Dutta, Evelina Angov, and Elke Bergmann from the Walter Reed Army Institute of Research). Briefly, 100μg of each antigen or BSA (Sigma), as a background control, were coupled to ∼12x106 non-magnetic microspheres (BioRad carboxylated beads) and then incubated with study participant plasma (spun down 10,000g for 10 mins and diluted at 1:100 in the assay dilution buffer) for 2 hrs, followed by incubation with biotinylated anti-human IgG (BD #555785) diluted 1:1000 for 1hr and streptavidin (BD #554061) diluted 1:1000 for 1 hr following the manufacturer’s instructions. The mean fluorescence intensity (MFI) of each conjugated bead (minimum of 50 beads per antigen) was quantified on a FlexMap3D Luminex multianalyte analyzer (Xponent software). Results are reported as antigen-specific MFI after subtracting the BSA value for each individual, since background levels can vary between individuals.

OMIQ analysis

The fcs files were uploaded into the OMIQ platform after passing QC under SpectroFlo® (Cytek) where compensation was re-checked. In OMIQ, we arcsinh-transformed the scale to allow downstream analysis and then gated on singlet live lymphocytes and sub-sampled the data to yield 87,712 live lymphocytes per sample. Using only lineage markers CD3, CD4, CD8, CD14, CD19, CXCR5 and CD25, FlowSOM consensus meta-clustering was run on 100 clusters based on the 87,712 live lymphocytes per sample with a comma-separated k-value of 75, and Euclidean distance metric. Using these 75 meta-clusters, we defined subsets of cells based on lineage markers, such as CD3posCD8pos, CD3negCD14posCD19pos, and CD3posCD4pos, and then distinguished CD4posCXCR5posCD25neg (cTFH) and CD4posCXCR5negCD25pos (T regulatory [Treg] or T follicular regulatory [TFR]) subsets. CD25 was used to exclude Treg and TFR cells which share numerous markers with cTFH cells (Sage et al., 2014; Wing et al., 2018; Zhao et al., 2020). EmbedSOM dimensional reduction was used to visualize the different groups of cells and EdgeR analysis was run to assess the significance of their differences. A clustering heatmap was used to visualize cytokine expression and transcription factor profiles for each subset.

Focusing on the CD4posCXCR5posCD25neg TFH cells, we ran another FlowSOM analysis based on the 1,000 CXCR5pos cells per sample (two samples from the adult’s group and one sample from the children group were excluded from the analysis as they had less than 1,000 CXCR5pos cells), using extracellular markers CXCR5, CXCR3, CCR6, ICOS, CCR7, CD45RA, CD127, CD40L and PD1 allowing the identification of 15 meta-clusters. From there, we performed Uniform Manifold Approximation and Projection (UMAP) dimensional reduction, heatmaps, and EdgeR analyses, the latter allowing statistical analysis of the cTFH abundances. To demonstrate the reproducibility of these results, statistical analysis algorithms were run at least three times downstream of the same clustering algorithm and downstream of repeated clustering algorithms. To assess statistical differences in cytokines and transcription factor expression, we exported the statistics dataset from OMIQ containing MFI values from each marker (IFNγ, IL-4, IL-21, Bcl6 and cMAF) per cluster and for each sample and stimulation condition. To assess cytokines and transcription factors without bias, we chose to use the total MFI expression per meta-cluster with the assumption that cells with an increased production of the desire analyte will trigger an increase of the overall meta-cluster MFI compared to unstimulated cells, and if there is no production of the desire analyte, the overall MFI will not differ.

Statistical analysis

For this cross-sectional immunology study, we selected both male and female, sex defined at birth, fourteen 7-year-old children and fifteen adults. Using GraphPad Prism software (version 7.0), age, sex, absolute lymphocyte count (ALC), and serological data were compared between adults and children. Because the number of participants within each group was too low to verify the normality of the underlying distributions (adults n=15 and children n=14), we chose to use non-parametric tests, including the Mann-Whitney U test (for unpaired analysis) and Wilcoxon signed-rank test (for paired analysis). When data passed the Normality test (D’Agostino and Pearson test), we used Welch’s parametric test. All tests were two-tailed with a p-value < 0.05 for significance. Because of the exploratory nature of the analysis, we did not use any adjustment of the p-value for multiple comparisons. The tests used are indicated in the legend of each figure. Results were expressed using the mean with standard deviation for dot plots, and exact p-values.

To perform a statistical analysis of the cytokine and transcription factor expression from each cTFH subset, the exported data file from OMIQ was integrated into GraphPad Prism, and a non-parametric Wilcoxon paired-t-test, two-tailed, analysis was done (n=13 in each group). As this analysis generated numerous bar plots (all included in the supplemental figures 12 to 15), in order to have a better visualization of the cytokines and transcription factors patterns, the p-values obtained from each analysis are presented using a non-clustering heatmap.

Role of the funding source

Funders had no role in the study design; the collection of the samples, the analysis and the interpretation of the data as well as the writing of the manuscript and its decision to be submitted for publication.



CN, JMO, JK and AMM designed research study; CF and SPT conducted experiments; CF, JK and AMM analyzed data; CF, SPT, JK and AMM contributed to experimental design; HWW, BO, JMO, JK and AMM conducted and supervised the clinical study and the sampling; CF generated figures; HWW and BO verified underlying data; CF and AMM led manuscript preparation with feedback from all the authors. All authors have read and approved the final version of the manuscript.

Data Availability

Deidentified raw data from this manuscript are available from ImmPort platform under the accession study number SDY2534.

Declaration of Competing Interest

Dr Kurtis was the PI on 1R01AI127699-01A1 which supported this study. He holds several patents related to the use of PfSEA-1 and PfGARP as vaccine candidates for P. falciparum and has consulted for and is an equity holder in Ocean Biomedical.

The authors would like to thank the children and their families for participating in this study. We thank the field and lab teams for their work collecting data and processing blood samples. We also thank Dr. Melanie Trombly from UMass Chan for proofreading our manuscript. This manuscript was approved for publication by the Kenya Medical Research Institute.

This study was supported by NIH R01 AI127699 (Kurtis).


Representative cytoplots of the extracellular flow staining.

Panel a shows the CXCR5 vs CD25 cytoplots on unfiltered, CD19pos, CD3pos, CD4pos and CD8pos cells from left to right. Panel b shows CD45RA vs CCR7 cytoplots on CD3pos, CD8pos, CD4pos and CD4posCXCR5pos cells from left to right. Panel c shows CCR6 vs CXCR3 staining on CD4posCXCR5pos cells. Panel d shows CD127 vs CD40L expression within CD4posCXCR5pos cells by unstimulated and SEB stimulated cells. Panel e shows ICOS vs PD1 cytoplots after no stimulation or SEB 6h stimulation.

EdgeR analysis of the CD4pos, CD4posCXCR5pos, CD8pos and CD14/CD19pos populations under different stimulation conditions in adults (n=15) and children (n=14).

Volcano plot from EdgeR analysis comparing CD4pos, CD4posCXCR5pos, CD8pos, and CD14/CD19pos populations from adult PBMCs under the following conditions: (a) unstimulated versus PfSEA-1A, (b) unstimulated versus PfGARP, and (c) PfSEA-1A versus PfGARP. Volcano plot from EdgeR analysis comparing CD4pos, CD4posCXCR5pos, CD8pos,and CD14/CD19pos populations from children PBMCs under the following conditions: (d) unstimulated versus PfSEA-1A, (e) unstimulated versus PfGARP, and (f) PfSEA-1A versus PfGARP. The green dots are statistically significant whereas the black dots are not, here no green dots were observed.

Characterization of TFH subsets using CXCR5, CXCR3, CCR6, and CD45RA expression.

Bar plots showing (a) CXCR5, (b) CD45RA, (c) CCR6, and (d) CXCR3 expression across all cTFH-like meta-clusters from adults (n=13) under no stimulation (blue), PfGARP (green) and PfSEA-1A (pink) stimulations. Fifty percent of the data are within the box limits, the solid line indicates the median, the dash line the mean and the whiskers indicate the range of the remaining data with outliers being outside that range.

CCR6 versus CXCR3 representative cytoplots. a.

CCR6 vs CXCR3 expression from activated cTFH2-like MC01 (in blue), quiescent cTFH1-like MC07 (in pink), activated cTFH1/17-like MC11 (in orange) and activated cTFH17-like MC13 (in green). Undetermined meta-clusters was superposed to the sus-mentioned metacluster: activated cTFH1or2-like MC03 (in black) (b); activated cTFH1or2-like MC04 (in black) (c); quiescent cTFH1or2-like MC05 (in black) (d); quiescent cTFH1or2-like MC08 (in black) (e).

Cartoon of TFH subsets and their state of activation.

State of activation of cTFH subsets from adults determined by PD1, ICOS, CCR7, and CD127.

Bar plots showing (a) PD1, (b) ICOS, (c) CCR7 and (d) CD127 expression across all cTFH-like meta-clusters in adults (n=13) under no stimulation (blue), PfGARP (green) and PfSEA-1A (pink) stimulations. Fifty percent of the data are within the box limits, the solid line indicates the median, the dash line the mean and the whiskers indicate the range of the remaining data with outliers being outside that range.

CD40L expression across stimulation and meta-clusters.

Dot plots of CD40L expression after manual gating based on unstimulated condition from children (n=13) (a) and adults (n=13) (b) cTFH meta-clusters.

Abundance of the cTFH meta-clusters within children with low or high levels of HRP2 antibodies.

(a) UMAP plot showing the clustering of the 13 different cTFH meta-clusters (MC15) in children with low levels of HRP2 (on the left, n=6) and high levels of HRP2 (on the right, n=6) in the absence of stimulation. Each color represents a meta-cluster. The red circles highlight the meta-cluster showing differences in its abundance between the two groups of children. (b) An EdgeR statistical plot was performed to assess the abundance of the 13 meta-clusters between unstimulated PBMCs from both groups of children; the Y-axis being -log10(p-value) and the X-axis shows the log(FC). The green dot (MC04) is statistically significant whereas the black dots are not. (c) An abundance heatmap showing the distribution (in %) of each meta-cluster within the two groups of children (both concatenated unstimulated PBMCs from six participants). The color scale ranges from high abundance (red) to low abundance (blue).

Clustered heatmap of the cytokines and transcription factors expressed from cTFH meta-clusters in adults and children under the different stimulation conditions.

Heatmap of Bcl6, IFNγ, IL21, IL4, and cMAF expression from the 13 cTFH meta-clusters from concatenated data from (a) children (n=13) and (b) adults (n=13), after stimulation by PfSEA-1A or PfGARP, or without any stimulation as indicated in the name of each row. The color scale ranges from high expression (red) to low/no expression (blue).

Bar plots of cytokines expressed under the different conditions of stimulation in children (n=13).

Comparison of mean fluorescent intensity (MFI) of (a) IFNγ, (b) IL4 cytokines expression after PfSEA-1A (pink) or PfGARP (green) stimulation or no stimulation control (blue). Bar plots indicate mean with SD. Wilcoxon-paired two-tailed t-tests were performed and p-values are indicated.

Bar plots of transcription factors expressed under the different conditions of stimulation in children (n=13).

Comparison of mean fluorescent intensity (MFI) of (a) Bcl6 and (b) cMAF expression after PfSEA-1A (pink) or PfGARP (green) stimulation or no stimulation control (blue). Bar plots indicate mean with SD. Wilcoxon-paired two-tailed t-tests were performed and p-values are indicated.

Bar plots of cytokines expressed under the different conditions of stimulation in adults (n=13).

Comparison of mean fluorescent intensity (MFI) of (a) IFNγ, (b) IL4 and (c) IL21 and cytokines expression after PfSEA-1A (pink) or PfGARP (green) stimulation or no stimulation control (blue). Bar plots indicate mean with SD. Wilcoxon-paired two-tailed t-tests were performed and p-values are indicated.

Bar plots of transcription factors expressed under the different conditions of stimulation in adults (n=13).

Comparison of mean fluorescent intensity (MFI) of (a) Bcl6 and (b) cMAF expression after PfSEA-1A (pink) or PfGARP (green) stimulation or no stimulation control (blue). Bar plots indicate mean with SD. Wilcoxon-paired two-tailed t-tests were performed and p-values are indicated.