Participants with no history of typhoid fever and no typhoid vaccination were enrolled in the Oxford University campus. The National Research Ethics Service (NRES), Oxfordshire Research Ethics Committee A (11/SC/0302) approved the protocol for blood collection in the wild-type S. Typhi challenge model (31). This study was carried out following the ethical standards laid down in the 1964 Declaration of Helsinki and the principles of the International Conference on Harmonization Good Clinical Practice guidelines (32). Participants were informed about the purpose and risks of the study and written informed consent was obtained from the participants before participation in the study. All blood specimens were processed within 4 h of obtaining the samples.

Sixteen healthy participants aged 18–46 years were screened and recruited by the Oxford Vaccine Group, UK as described before (31). Briefly, participants who had previously received typhoid vaccination, or resided for over 6 months in typhoid-endemic areas or were previously diagnosed with typhoid infection were excluded from this study. Participants were first vaccinated with 3 doses of the live oral attenuated typhoid vaccine, Ty21a and then challenged orally with 1–5 × 10⁴ CFU of wt S. Typhi (Quailes strain, an antibiotic susceptible strain) administered after neutralization of gastric acid with NaHCO₃ as previously described (31). As previously explained, following challenge, some participants developed typhoid disease (TD) as determined by blood culture-confirmed S. Typhi bacteremia or development of a fever of ≥38°C for ≥12 h, whilst some participants did not developed typhoid disease (NoTD) (31). Peripheral blood mononuclear cells (PBMC) were obtained from all participants enrolled in this study at various time points as described in Figure 1 . PBMC were isolated from blood by density gradient centrifugation and cryopreserved in liquid nitrogen following standard techniques (33). PBMC collected before and up to 28 days after challenge were evaluated in the studies included in this manuscript ( Figure 1 ).

Using each participant PBMC, autologous Epstein–Barr virus (EBV)-transformed lymphoblastoid cell line (B-EBV cells) were generated as previously described (34, 35). Briefly, B-EBV cells were generated by infection of PBMC with EBV particles [supernatant from the B95-8 cell line (ATCC CRL1612)] and cyclosporine (0.5 μg/ml; Sigma-Aldrich, Saint-Louis, MO, USA) for 15–30 days.

Autologous B-EBV cells were incubated with wt S. Typhi strain ISP1820 at a multiplicity of infection (MOI) of 7:1 (bacteria:target ratio) for 3 h at 37°C in RPMI free of antibiotics. Cells were washed extensively after the infection with cRMPI and cultured overnight in cRPMI supplemented with gentamicin (150 μg/ml). The efficiency of the S. Typhi infection was confirmed by flow cytometry after staining with anti-Salmonella common structural Ag (Kierkegaard & Perry, Gaithersburg, MD, USA) as previously described (36).

As described previously, PBMC were thawed and rested in cRPMI overnight before stimulation with S. Typhi-infected target cells (37–39). For negative and positive controls, uninfected target cells and Staphylococcus enterotoxin B (SEB; 10 μg/ml) were used, respectively. Targets cells were γ-irradiated (6,000 rad) and incubated with PBMC at an effector:stimulator ratio of 7:1 for 2 h in the presence of anti-CD107a (metal conjugated, Fluidigm) monoclonal antibody (mAb). After two hours of incubation, Golgi Stop (0.5 μl; Monensin, BD) and Golgi Plug (0.5 μl, Brefeldin A, BD) were added and the cultures continued overnight at 37°C in 5% CO₂.

After an overnight stimulation, PBMC were stained for mass cytometry analysis as reported before (33, 40, 41). Briefly, cells were first barcoded using CD45 tagged with 141Pr, 154Gd, 156Dy for uninfected, S. Typhi-infected EBV stimulated samples and SEB. The samples were then stained for live/dead cell with cisplatin (194/195 Pt), followed by 30 min-incubation with human Fc receptor blocking IgG. Cells were then stained for surface markers and fixed, permeabilized and intracellular staining performed as previously described using the 28-marker panel of anti-human metal-labeled mAbs shown in Supplementary Table S1 . Finally, within 48 hr of sample labeling, they were stained with an Ir^191/193 DNA intercalator for cell detection and re-suspended in EQ4 normalization beads (Fluidigm). Data acquisition was performed using a Helios mass cytometer (Fluidigm). Mass cytometry experiments were performed at the Flow Cytometry and Mass Cytometry Core Facility of the University of Maryland School of Medicine Center for Innovative Biomedical Resources (CIBR), Baltimore, Maryland.

ELISAs were performed to measure the level of immunoglobulin G (IgG), IgA, and IgM isotype responses to O9:LPS and to H (flagellar antigen) in serum as previously described (30, 42, 43). ELISAs were performed in a set of serum samples obtained at multiple time points (pre-vaccination -D-28-, pre-challenge -D0-, and post-challenge day 28 -D28-) corresponding to the participants (TD n = 8, NoTD n = 8) in whom the cT_FH subsets were evaluated.

Serum bactericidal antibody (SBA) assay was performed as described before (43). SBA assays were performed in a set of serum samples obtained at multiple time points (pre-vaccination -D-28-, pre-challenge -D0-, and post-challenge day 28 -D28-) corresponding to the participants (TD n = 8, NoTD n = 8) in whom the cT_FH subsets were evaluated. Briefly, serum samples were de-complemented by heat inactivation and diluted before addition of 200 CFU of log phase S. Typhi Quailes strain. S. Typhi-specific antibody depleted human complement serum was added to a final complement concentration of 25% and bacteria incubated for 1 h at 37°C with shaking before plating on tryptic soya agar (TSA) plates (Oxoid Ltd., UK) (43). SBA titers were correlated with the mass cytometry measurements for the various cT_FH subsets frequencies.

We determined the role of cT_FH in S. Typhi infection by using specimens obtained following Ty21a vaccination and a typhoid CHIM. We first characterized cT_FH (CXCR5⁺CD45RA^-CD4⁺) as previously reported (8, 10–14) in PBMC isolated from TD and NoTD participants using the gating strategy shown in Figure 2A . Next, we determined and compared the frequencies of total cT_FH between TD and NoTD participants ( Figure 2B ). We observed that at baseline (D-28), the frequencies of total cT_FH (median %) was higher in NoTD than TD participants ( Figure 2B ). Following Ty21a vaccination, the level of cT_FH (median %) remained elevated in NoTD as compared with TD participants at D-14, although these differences were not statistically significant ( Figure 2B ). Following wt S. Typhi challenge (D7), total cT_FH were significantly (p<0.05) higher in NoTD than in TD participants ( Figure 2B ). However, at later time points, the total cT_FH frequencies were not different between NoTD and TD participants ( Figure 2B ).

It is widely accepted that cT_FH can be classified into well defined, distinct subsets, based on expression of CXCR3 and CCR6 as follows: cT_FH1 (CXCR3⁺CCR6⁻), cT_FH2 (CXCR3⁻CCR6⁻), and cT_FH17 (CXCR3⁻CCR6⁺) ( Figure 2A ) with each having distinct capacities in supporting B cells differentiation and maturation to produce antibodies (8, 9). Thus, we explored whether the frequencies of these specialized cT_FH subsets are influenced by Ty21a vaccination and wt S. Typhi challenge. Interestingly, no significant differences were noted for cT_FH1 frequencies between TD and NoTD at baseline (D-28), after Ty21a vaccination (D-14) or following challenge (D7) ( Figure 2C ). In contrast, cT_FH2 frequencies appear to be higher at baseline (D-28) in NoTD than in TD participants and increased following Ty21a immunization (D-14 and D0) ( Figure 2D ). Of note, after challenge (D7), cT_FH2 frequencies from NoTD were significantly (p<0.05) higher than those in TD participants ( Figure 2D ). cT_FH17 frequencies appear to be higher at baseline (D-28), and after Ty21a vaccination (D-14, D0) in the NoTD participants compared with the TD participants ( Figure 2E ). However following challenge (D7), significantly (p<0.05) higher frequencies of cT_FH17 frequencies were observed in NoTD than in TD participants ( Figure 2E ). To confirm these observations, we performed area under the curve analyses (AUC) to determine whether there were any significant increases in frequencies of the subsets during the vaccination and/or challenge phases. The cumulative AUC data for cT_FH1 indicate that there were no significant differences between NoTD and TD participants in the vaccination or challenge phases ( Supplementary Figure S3A ). However, cT_FH2 showed significantly higher (p<0.05) frequencies in NoTD than in TD participants during the vaccination phase, as well as a trend (p ≤ 0.1) to show increases in the challenge phase as measured by AUC ( Supplementary Figure S3B ). Similarly, we noted a trend (p ≤ 0.1) to show significant increases in AUC of cT_FH17 in NoTD as compared to TD participants in the challenge phase ( Supplementary Figure S3C ). These results suggest that cT_FH2 and cT_FH17 might play a role in protection from S. Typhi infection.

To assess the effect of vaccination and challenge, we next compared the frequencies of cT_FH subsets at D-28 to D-14 for vaccination and D0 to D7 for challenge. No significant differences in frequencies of cT_FH1, 2, and 17 were detected following Ty21a vaccination, specifically between D-28 and D-14 ( Supplementary Figures S4A-C ). However, we noted that following challenge (D0 to D7), both cT_FH2 and cT_FH17 frequencies were lower in TD than in NoTD participants but this did not reach statistical significance. When comparing the levels of cT_FH2 and cT_FH17 at D7 after challenge between NoTD and TD, we observed significantly (p<0.05) higher levels of cT_FH2 and cT_FH17 in NoTD participants ( Supplementary Figure S4F-G ). No significant differences in cT_FH1 frequencies were noted following challenge in either TD or NoTD participants ( Supplementary Figure S4E ).

Previous studies have reported that cT_FH1 is less efficient in the induction of B cells than cT_FH2 and cT_FH17 (8). The ratio of cT_FH2+cT_FH17 to cT_FH1 is generally considered an indication of whether there is a shift of cT_FH subsets to those that support antibody responses during infection (52). Thus, we determined the effect of Ty21a vaccination and wt S. Typhi challenge on the induction of these subsets by comparing D-28 to D-14 for vaccination and D0 to D7 for challenge. We observed that following Ty21a immunization, the ratio of cT_FH2+cT_FH17:cT_FH1 did not show any differences between NoTD and TD participants between D-28 and D-14 ( Supplementary Figure S4D ). However, following wt S. Typhi challenge, the ratio of cT_FH2+cT_FH17:cT_FH1 shows a significant (p<0.05) decrease in TD compared with NoTD participants ( Supplementary Figure S4G ). These results suggest that following exposure to wt S. Typhi there was a decrease of the percentages of defined cT_FH subsets (cT_FH2 and 17) in TD participants.

To facilitate the interaction between cT_FH and B cells, CXCR5 is highly expressed on cT_FH to promote the homing of cT_FH to lymphoid follicles. Similarly, for cT_FH to be effective in other tissues, expression of various homing markers such as integrin α4β7 (promoting migration to intestinal mucosa) (53, 54) and CCR7 (promoting migration to secondary lymphoid tissues) (55) are needed on their cell surfaces. Given that S. Typhi invade the intestinal epithelium and disseminate in the lamina propria, we evaluated the capacity of cT_FH subsets to home to the intestine and other secondary lymphoid tissues by determining the expression of integrin α4β7 and CCR7, respectively. Interestingly, we observed that integrin α4β7 expression at baseline (D-28) was similar for all subsets regardless of disease status (TD and NoTD) ( Figure 3A ). Following Ty21a immunization, we observed a dichotomy in the expression of integrin α4β7 with higher levels in TD participants than in NoTD, particularly at D0, with significantly (p<0.05) higher levels of integrin α4β7 in cT_FH17 in TD than in NoTD participants ( Figure 3A ). Moreover, a trend (p ≤ 0.1) to show increases in the percentages of integrin α4β7 was observed in cT_FH1 and cT_FH2 in TD compared with NoTD participants at D0 ( Figure 3A ). Following challenge (D7), expression of integrin α4β7 remained significantly (p<0.05) higher in TD than in NoTD for cT_FH17 and a trend (p ≤ 0.1) to show increases was also observed in TD compared to NoTD participants for the other two cT_FH subsets ( Figure 3A ; Supplementary Tables S2 - 4 ). Further analysis of the data by AUC revealed that during the immunization phase, the expression of integrin α4β7 in all three cT_FH subsets was significantly (p<0.05) higher in TD than in NoTD participants ( Figures 4A-C ). However, in the challenge phase of the study, no significant differences in integrin α4β7 expression were detected in any of the three cT_FH subsets ( Figures 4A-C ). Next, we assessed CCR7 expression on the three cT_FH subsets and found that cT_FH2 and cT_FH17 appear to express higher levels of CCR7 in TD than in NoTD participants at baseline ( Figure 3B ). Following Ty21a immunization, a trend (p ≤ 0.1) to show increases of CCR7 expression on cT_FH2 and cT_FH17 was observed in TD compared with NoTD participants at D0 ( Figure 3B ). Following challenge (D7), trends (p ≤ 0.1) to show increases in the expression of CCR7 were found on cT_FH2 and cT_FH17 ( Figure 3B ). Further analysis of the data by AUC revealed that during the immunization phase, the expression of CCR7 showed a trend (p ≤ 0.1) to increase in cT_FH2 but not cT_FH1 or cT_FH17 in TD compared with NoTD participants ( Supplementary Figures S5A-C ). However, during the challenge phase of the study, trends (p ≤ 0.1) to show increases in the expression of CCR7 were observed in TD compared with NoTD participants for the three cT_FH subsets ( Supplementary Figure S5A-C ).

Having established the frequencies and homing potential of cT_FH subsets in TD and NoTD participants, we next investigated the activation status of these cT_FH subsets following Ty21a vaccination and challenge with wt S. Typhi. While both activation markers, CD69 and CD154 (CD40L) were induced in all cT_FH subsets following Ty21a vaccination and/or wt S. Typhi challenge, no significant differences between TD and NoTD participants were noted at any of the time points ( Figures 3C, D ). To further investigate this phenomenon, we performed AUC analyses. We did not observe any significant differences in the expression of CD69 and CD154 between TD and NoTD participants during the immunization phase (D-28 to D0) for any of three cT_FH subsets ( Supplementary Figures S6A-F ). However, during the challenge phase (D0 to D28), we observed trends (p ≤ 0.1) in TD participants to exhibit higher levels of CD69 (cT_FH2 and cT_FH17) and CD154 (cT_FH1 and cT_FH2) ( Supplementary Figures S6A-F ).

We next determined the level of co-stimulatory molecules (e.g., ICOS and PD1) on the three cT_FH subsets in TD and NoTD participants. We observed a clear dichotomy in the expression of ICOS between TD and NoTD participants ( Figure 3E ). At baseline, no differences were observed in ICOS expression between TD and NoTD in the three subsets ( Figure 3E ). However, following Ty21a vaccination, we observed increases in ICOS expression in all three cT_FH subsets in TD but not in NoTD participants with a trend (p ≤ 0.1) to exhibit increases at D0 ( Figure 3E ). After challenge (D7), a trend (p ≤ 0.1) to show increases of ICOS expression were found in cT_FH2 and cT_FH17 subsets in TD compared with NoTD participants ( Figure 3E ; Supplementary Tables S2 - 4 ). We then performed AUC and observed that during the immunization phase there were significantly (p<0.05) higher levels of ICOS expression in TD than in NoTD participants in all three subsets ( Figures 4D-F ). In the challenge phase, ICOS expression was also higher in TD than in NoTD participants in cT_FH1 (p<0.05), cT_FH2 (p ≤ 0.1) and cT_FH17 (p<0.005) ( Figures 4D-F ). The expression of PD1, however, seems to be higher (not statistically significant) in NoTD than TD in the cT_FH1 (p ≤ 0.1) and cT_FH17 (p ≤ 0.1) subsets at baseline ( Figure 3F ). Following Ty21a vaccination, the level of PD1 decreases in NoTD but remains steady in TD participants at D-14 and D0 ( Figure 3F ). Following challenge, PD1 expression was increased in TD but not in NoTD participants ( Figure 3F ). Analysis by AUC showed that there were no significant differences in PD1 expression between TD and NoTD in either the immunization or challenge phases for any of the three cT_FH subsets, except for cT_FH1 in the challenge phase which displayed a trend (p ≤ 0.1) to show higher responses in TD compared with NoTD participants ( Supplementary Figures S5D-F ). Furthermore, we determined the level of CD27, a costimulatory receptor important in T cell function, and CD62L, homing marker to lymphoid tissues, present on cT_FH subsets. We observed that CD27 is highly expressed (70-90%) on cT_FH subsets at all time points (D-28 to D28) but no significant difference was observed between TD and NoTD participants ( Supplementary Figure S7A ). Similarly, the expression level of CD62L (~50-70%) was high on cT_FH subsets at all time points (D-28 to D28), with no significant differences in CD62L expression noted between TD and NoTD participants ( Supplementary Figure S7B ).

It is well established that cT_FH are driven and skewed by an array of cytokines and chemokines that allow for the control of infectious pathogens (56). For example, IL-21 is highly expressed by T_FH cell subsets and this cytokine has been shown to play a role in accelerating the development of plasmablasts (3). However, each cT_FH subset can be skewed by their secreted set of cytokines. For example, cT_FH1 secrete mostly Th1 cytokines such as IFN-γ while cT_FH2 secrete Th2 cytokines such as IL-4 and T_FH17 secrete mostly IL-17A. Differentiated cT_FH cells produced copious amount of IL-2 but, rather than being induced by it, they are inhibited (57). Since cytokine-skewed cT_FH can influence the magnitude and quality of humoral responses, we determined S. Typhi-specific cytokine responses of cT_FH subsets following Ty21a immunization and wt S. Typhi challenge. For cT_FH1 subsets, we first evaluated IL-21 and IFN-γ responses in TD and NoTD participants and observed that at baseline (D-28), no significant differences were observed in IL-21 and IFN-γ levels between TD and NoTD participants in the cT_FH1 subset following in vitro exposure to S. Typhi-infected autologous targets ( Figure 5A ). Following Ty21a immunization (D-14), cT_FH1 subset secrete significantly (p<0.05) lower levels of IL-21 in TD than in NoTD participants. No significant difference in IFN-γ production was observed between TD and NoTD participants ( Figure 5A ). We also determined the level of S. Typhi-specific macrophage inflammatory protein (MIP)-1β, tumor necrosis factor (TNF)-α, granzyme B and CD107A responses in cT_FH1 ( Supplementary Figures S8A, B ). No significant differences in the production of these cytokines or expression of CD107A responses were noted between TD and NoTD participants in the cT_FH1 subset at baseline (D-28). However, following Ty21a vaccination (D-14), there were significantly (p<0.05) lower levels in MIP-1β ( Supplementary Figure S8A ) and CD107a ( Supplementary Figure S8B ) in TD than in NoTD participants. At D0, there were significantly (p<0.05) lower levels of TNF-α in the cT_FH1 subset in TD than in NoTD participants ( Supplementary Figure S8A ; Supplementary Table S2 ). Following wt S. Typhi challenge (D7), there was an increase in the production of MIP-1β, TNF-α, and granzyme B in the TD group compared with the NoTD group, while CD107a was lower in TD than in NoTD participants ( Supplementary Figures S8A, B ) but none of these responses were statistically significant. Thus, cT_FH1 subsets produced cytokines important for modulating their environment and influencing B cells to produce distinct antibody isotypes. Similarly, we examined cT_FH2 S. Typhi-specific responses and found no significant differences in S. Typhi-specific IL-21 and IL-2 between TD and NoTD participants ( Figure 5B ). Of note, following challenge (D7), the levels of IL-21 were higher in TD than in NoTD participants but did not reach statistical significance ( Figure 5B ). We also determined the levels of S. Typhi-specific MIP-1β, TNF-α, granzyme B and expression of CD107A responses in cT_FH2 of TD and NoTD participants ( Supplementary Figures S8C, D ). No significant differences in the production of these responses were noted between TD and NoTD in cT_FH2 at baseline (D-28) except for a trend (p ≤ 0.1) to exhibit higher CD107a expression in NoTD than in TD participants ( Supplementary Figures S8C. D ). However, following Ty21a vaccination (D-14), there was a significantly (p<0.05) higher production of cT_FH2 S. Typhi-specific MIP-1β,and TNF-α ( Supplementary Figure S8C ) and CD107a ( Supplementary Figure S8D ) in TD than in NoTD participants. At D0, a trend (p ≤ 0.1) to exhibit higher levels of TNFα ( Supplementary Figure S8C ) and CD107a expression ( Supplementary Figure S8D ) were observed in NoTD compared to TD participants. Following wt S. Typhi challenge (D7), we observed trends (p ≤ 0.1) to show increases in the production of MIP-1β, TNF-α, and granzyme B in TD than in NoTD participants ( Supplementary Figures S8C, D ; Supplementary Table S3 ). Next, we investigated cT_FH17 S. Typhi-specific responses and observed that at baseline, significantly (p<0.05) higher production of IL-17A, but not IL-21, was present in NoTD than in TD participants ( Figure 5C ). Following Ty21a vaccination, cT_FH17 produced higher levels of IL-21 (trend; p ≤ 0.1) and IL-17A (significant; p<0.05) in NoTD than in TD participants at D0 ( Figure 5C ). Following wt S. Typhi challenge (D7), we noted an increase in both the production of IL-21 and IL-17A in cT_FH17 in TD participants, as well as an increase in IL-17 production in NoTD participants ( Figure 5C ). Interestingly, there was significantly (p<0.05) higher production of S. Typhi-specific IL-17A in NoTD than TD participants D14 days after challenge ( Figure 5C ). We also determined the level of S. Typhi-specific MIP-1β, TNF-α, granzyme B and CD107A responses in cT_FH17 of TD and NoTD participants ( Supplementary Figures S8E, F ; Supplementary Table S4 ). No significant differences in the production of these responses were noted between TD and NoTD in cT_FH17 at baseline (D-28), except for a significantly (p<0.05) higher production of S. Typhi-specific MIP-1β in NoTD than in the TD group ( Supplementary Figures S8E, F ). However, following Ty21a vaccination (D-14 and D0), there were trends (p ≤ 0.1) to show increases in S. Typhi-specific TNF-α production ( Supplementary Figure S8E ) and significantly (p<0.05) higher expression of cT_FH17 S. Typhi-specific CD107a in TD than in NoTD participants ( Supplementary Figure S8F ). Following wt S. Typhi challenge, there were no significant differences in cytokine production in cT_FH17 between TD and NoTD participants ( Supplementary Figures S8E, F ). Thus, like we observed in the cT_FH1 and cT_FH2 subsets, cT_FH17 subsets produced cytokines specifically when exposed to S. Typhi antigens following immunization and challenge.

To investigate these phenomena in further detail, we focused on the changes of S. Typhi-specific responses in cT_FH subsets following wt S. Typhi challenge by comparing responses at D0 (pre-challenge) and D7 (post challenge) in TD and NoTD participants. Interestingly, in TD participants, there was an increase of IL-21 production in cT_FH1 (significant, p<0.05), cT_FH2 (trend, p ≤ 0.1) and cT_FH17 (significant, p<0.05) from D0 to D7 ( Figure 6A ). No significant differences in IL-21 production were detected from D0 to D7 in any cT_FH subset in NoTD participants ( Figure 6A ). It is worth noting that at D0, IL-21 production in cT_FH subsets was higher in NoTD than TD participants with a trend (p ≤ 0.1) to show increases observed in cT_FH17 ( Figure 6A ). No significant differences were detected in IFN-γ production between D0 and D7 in cT_FH subsets from either TD or NoTD participants ( Figure 6B ). Similarly, we determined and compared S. Typhi-specific IL-2 production between D0 and D7 to evaluate the effect of the S. Typhi challenge on cT_FH subsets. Interestingly, while cT_FH1 produced IL-2, there were no significant differences between D0 and D7 regardless of disease status ( Figure 6C ). However, the production of IL-2 was increased from D0 to D7 in cT_FH2 (significant, p<0.05) and cT_FH17 (trend, p ≤ 0.1) in NoTD participants ( Figure 6C ). We also determined the production of S. Typhi-specific IL-17A at D0 and D7 and found that there were no significant increases in cT_FH1 regardless of disease status ( Figure 6D ). In contrast, production of S. Typhi-specific IL-17A was significantly (p<0.05) higher between D0 and D7 in cT_FH2 of TD participants ( Figure 6D ). Interestingly, we detected similar trends (p ≤ 0.1) in IL-17A between D0 and D7 in cT_FH17 of TD participants ( Figure 6D ). In contrast, the production of IL-17A between D0 and D7 was significantly (p<0.05) decreased in cT_FH17 of NoTD participants ( Figure 6D ). S. Typhi-specific TNF-α production was significantly (p<0.05) higher between D0 and D7 in cT_FH1 and cT_FH2 of TD but not in NoTD ( Figure 6E ). Interestingly, we noted a trend (p ≤ 0.1) to show decreased TNF-α production between D0 and D7 in cT_FH1 of NoTD participants ( Figure 6E ). Finally, the production of granzyme B was determined and compared between D0 and D7. We observed higher levels of S. Typhi-specific granzyme B production in D0 than D7 in cT_FH1 (trend, p ≤ 0.1), cT_FH2 (significant, p<0.05) and cT_FH17 (significant, p<0.05) in NoTD but not in TD participants ( Figure 6F ). In sum, wt S. Typhi challenge alters the production of S. Typhi-specific responses and elicited distinct signatures for each cT_FH subset in relation to disease status (TD vs NoTD).

Of note, we determined whether S. Typhi-specific cT_FH responses (CD107a, GzB, IFNγ, IL-17A and TNFα) were either single-producing cells (S) or multifunctional cells (MF) (simultaneously producing two or more cytokine/chemokine). To address this issue, we used Boolean gating (FlowJo) to determine multifunctionality of the effector responses in CD3⁺ CD4⁺ CD45RA^- CXCR5⁺ cT_FH. The results are displayed in Supplementary Figures S2B-F . Interestingly, we observed that S. Typhi-specific CD107a-associated responses of total cT_FH contain higher levels of MF than S in both TD and NoTD volunteers at all time points ( Supplementary Figure S2B ). Similar observations for S. Typhi-specific MF and S -associated responses were noted for IFNγ, IL-17A, and TNFα ( Supplementary Figures S2D-F ). However, for Granzyme B (GzB) responses, we observed that both S. Typhi-specific S and MF-associated responses display higher levels in TD than NoTD volunteers ( Supplementary Figure S2C ).

To study in further detail the associations between cT_FH subsets and typhoid disease we used unsupervised analysis approaches on the rich datasets generated by mass cytometry, by performing a phenotypic analysis on subpopulations gated on CD4+CD45RA-CXCR5+ cT_FH using concatenated files from TD and NoTD participants at all time points. First, a dimension reduction step was performed using UMAP (Uniform Manifold Approximation and Projection; version 3.1; FlowJo plugin (45)). Next, unsupervised clustering was performed using PhenoGraph (v2.5; FlowJo plugin) (version 2.7) (48) which resulted into 11 clusters. PhenoGraph clusters were then visualized on the initial UMAP to create a reference map of all automatically detected cT_FH subsets with the events numbers in the table below the UMAP for each cluster ( Figure 7A ). We next used this UMAP reference map to visualize and assess the distribution of activation markers (e.g., CD69, CD27, ICOS, PD1 and CD154) ( Figure 7B ), cytokines/chemokines (IL-17A, CD107a expression, IL-2, GzB, IFN-γ, TNFα, IL-21 and MIP-1β) ( Figure 7C ) and homing markers (e.g., CCR4, CXCR3, CD62L, integrin α4β7, CCR7 and CCR6 ( Figure 7D ). We observed that activation markers (CD69, ICOS, PD1, CD154) were distributed in unique patterns across the 11 UMAP clusters, whereas CD27 was expressed in all the clusters ( Figure 7B ). The distribution of cytokines across the 11 clusters also showed distinct patterns ( Figure 7C ). For example, IFN-γ and IL-2 were present in only a few clusters. In contrast, IL-21, MIP-1β, and TNF-α are present on most clusters ( Figure 7C ). Similarly, we noted that homing markers (e.g., integrin α4β7, CCR7, CCR4 and CD62L) are distributed in unique patterns across the 11 clusters. For example, CD62L was expressed in all clusters except for cluster 1 ( Figure 7D ), while CXCR3 and CCR6 markers are expressed in only some cT_FH subsets. These results suggest that cT_FH clusters have unique combinations of markers that could be important in either Ty21a vaccination or wt S. Typhi challenge.

Next, we compared the expression of phenotypic, homing and activation markers of each of the 11 clusters as shown by the red color for maximum expression and blue for minimum expression ( Figure 8A ). For example, cluster 1 was CD62L- CCR6- CXCR3 dim/- and positive for all other markers including IL-17A, MIP-1β ( Figure 8A ). Furthermore, we used the Markers Enrichment Model (MEM) (58) to determine the main phenotypes of the 11 clusters of cT_FH ( Supplementary Table S5 ). We observed that each cluster has a unique phenotype and some of the markers are more abundant than others ( Supplementary Table S5 ). We then determined the frequencies of each cluster in TD and NoTD participants and observed that there were various clusters that were present at higher frequencies in TD than in NoTD participants as shown by the yellow color intensity ( Figure 8A ). For example, clusters 1, 4, 5 and 8 were present at higher in frequencies in TD than in NoTD participants ( Figure 8A ). However, these differences in cluster frequencies includes all time points. Of note, we use individual histogram to show the expression of the various markers in TD and NoTD from a representative cluster (cluster 4 in Figure 8A ) is shown in Supplementary Figure S9 . Thus, we next determined the frequencies of each cluster at each time point (D-28, D-14, D0, D7, TD48, TD96, D14 and D28) ( Figure 8B ). We observed that each cluster is present in higher frequencies at particular time points ( Figure 8B ). For example, following Ty21a vaccination at D-14, clusters 4, 6 and 9 are present at higher frequencies ( Figure 8B ). However, these differences might be masked by the clinical status (TD and NoTD). Thus, to interpret how the clusters varies between TD and NoTD at the various time points, we determined and compared the frequencies of each cluster (1–11) gated on either TD (blue line) or NoTD (red line) at each time point ( Figures 9A-K ). Interestingly, we observed that there were some clusters that were higher in TD than NoTD (e.g., clusters 1, 4, 5, 8), while other clusters (e.g., clusters 6, 7, 9) were higher in NoTD than TD ( Figures 9A-K ). In addition, there were clusters that showed increases following wt S. Typhi challenge (e.g., clusters 7, 10) ( Figures 9A-K ). These results suggest that distinct clusters could be associated with either the development or prevention of typhoid disease. Furthermore, there may be distinct clusters which could be elicited following Ty21a immunization or wt S. Typhi challenge.

In Figure 2A we show, as widely reported in the literature (8, 9, 16), that cT_FH can be classified into four subsets based on the expression of CXCR3 and CCR6, i.e., cT_FH1, cT_FH2, CT_FH17 and cT_FH-DP (double positive). We applied this hierarchical analysis to the clustering analysis and determined that each cluster is present at defined frequencies among cT_FH subsets ( Figure 10A ). For example, cluster 11 was present mostly in cT_FH17 as indicated by yellow coloring ( Figure 10A ). Similarly, clusters 6 and 10 were highly represented in the cT_FH-DP subset ( Figure 10A ). However, these data represent TD and NoTD participants and all time points combined. Thus, we next visualized the cT_FH subsets by embedding them onto the UMAP reference map which shows that cT_FH1 subset is present in clusters 1, 3, 4, 6, 7, 8 and 9, cT_FH2 in clusters 1, 3, 4, 6, 8, 9 and 11, cT_FH17 in clusters 1, 2, 3, 4, 6, 8, 9 and 11 and cT_FH-DP in clusters 1, 2, 5, 6, 7, 9, and 10 ( Figure 10B ). We also observed that some clusters are unique to some of the subsets. For example, cluster 5 is unique to cT_FH-DP while cluster 11 is composed exclusively of cT_FH2 and cT_FH17 ( Figure 10B ). In addition, we observed that the distribution of the 11 clusters varies within each cT_FH subset as shown by the pie chart for a representative volunteer ( Figure 10C ). We next deconvoluted the data that involves each cT_FH subset with the clusters of interest based on the results shown in Figure 10 and apparent differences between the TD and NoTD groups to evaluate the effect of Ty21a vaccination and wt S. Typhi challenge. Note that clusters were excluded if their frequencies (median % events) were less than 1% at any time point following comparison between TD and NoTD participants. We observed that there were 4 clusters of interest (Clusters 3, 4, 7 and 8) for cT_FH1 ( Figures 11A-D ). At baseline (D-28), no significant differences were observed in the frequencies of cT_FH1cluster 3 between TD and NoTD participants, while following Ty21a immunization (D-14 and D0), we observed a trend (p ≤ 0.1) to show increases in the frequencies of cT_FH1 in NoTD compared to TD participants ( Figure 11A ). Following challenge (D7), the frequency of cT_FH1 cluster 3 in NoTD decreases and was not significantly different from TD participants ( Figure 11A ). At D28, however, we again observed a trend (p ≤ 0.1) to show increases in the frequencies of this cT_FH1 cluster 3 in NoTD compared to TD participants ( Figure 11A ). Remarkably, cT_FH1 cluster 4 was higher in frequency (almost twice) in TD than in NoTD participants at all time points ( Figure 11B ), with trends (p ≤ 0.1) observed after Ty21a immunization (D-14 and D0) and after wt S. Typhi challenge (D7 and TD96) ( Figure 11B ). In contrast, cT_FH1 cluster 7 frequencies were higher in NoTD (red line) than in TD (blue line) at all time points ( Figure 11C ), with trends (p ≤ 0.1) observed in NoTD than TD following challenge (D7) ( Figure 11C ). cT_FH1 cluster 8 exhibited higher frequencies at baseline and following Ty21a vaccination in NoTD than in TD participants but differences were not statistically significant ( Figure 11D ). However, following wt S. Typhi challenge (D7), cluster 8 showed a trend (p ≤ 0.1) to be higher in TD than in NoTD participants ( Figure 11D ). We also compared TD and NoTD participants for all remaining cT_FH1 clusters and noted no significant differences in frequencies between the TD and NoTD groups in any phase of the study.

Next, we evaluated cT_FH2 clusters and found that clusters 4, 7 and 11 were of interest ( Figures 11E–G ). cT_FH2 cluster 4 exhibited higher frequencies in TD than in NoTD participants at all time points ( Figure 11E ), with a trend (p ≤ 0.1) to be higher in TD than in NoTD participants following challenge (D7) ( Figure 11E ). In contrast, cT_FH2 cluster 7 frequencies were higher in NoTD than TD at most time points (D-28, D-14, D0, D7 and D28) ( Figure 11F ), showing a trend to be higher in NoTD than in TD participants at baseline (D-28; p ≤ 0.1), after Ty21a vaccination (D-14, p ≤ 0.1) and following challenge (D7, p ≤ 0.1) but with significant differences observed at D0 (p<0.05) ( Figure 11F ). No significant differences were observed in the frequencies of cT_FH2 cluster 11 between TD and NoTD participants at baseline and after Ty21a vaccination ( Figure 11G ). However, following challenge, we found that there was a significant (p<0.05) increase in the frequencies of cT_FH2 cluster 11 in NoTD compared to TD participants ( Figure 11G ). No other trends or statistically significant differences were observed in the remaining cT_FH2 clusters.

We also examined cT_FH17 clusters and found that clusters 4, 7, 8 and 9 were of interest ( Figures 11H-K ). No significant differences in cT_FH17 cluster 4 frequencies between TD and NoTD participants were detected at any time points ( Figure 11H ). cT_FH17 cluster 7 showed trends (p ≤ 0.1) to higher frequencies in NoTD than in TD participants at baseline (D-28) and following Ty21a immunization (D-14) ( Figure 11I ) but not following challenge (D7). A trend (p ≤ 0.1) to show increased frequencies in TD participants was observed in cluster 8 after challenge (D7) ( Figure 11J ). In contrast, a trend (p ≤ 0.1) to show increased frequencies in NoTD participants was observed in cT_FH17 cluster 9 after challenge (D7) ( Figure 11K ). No significant differences were observed in the remaining cT_FH17 clusters.

Finally, we investigated clusters associated with cT_FH CXCR3⁺CCR6⁺ (cT_FH-DP) and found that clusters 4, 5, 7 and 10 were clusters of interest ( Figures 11L-O ). cT_FH-DP cluster 4 showed a trend (p ≤ 0.1) to show higher frequencies in TD than in NoTD participants at baseline (D-28) and following Ty21a immunization (D0), but not after challenge ( Figure 11L ). For cT_FH-DP cluster 5, higher frequencies were observed in TD participants, which became significant (p<0.05) after challenge (D7) ( Figure 11M ). However, for cT_FH-DP cluster 7, we noted that its frequencies were significantly (p<0.05) higher in NoTD participants at baseline (D-28) and following Ty21a immunization (D-14), but with a trend (p ≤ 0.1) observed at D14 following challenge ( Figure 11N ). Similarly, for cT_FH-DP cluster 10, we observed higher frequencies (no statistical significance) in NoTD than in TD participants at most time points, but with a trend (p ≤ 0.1) observed 7 days following challenge (D7) ( Figure 11O ). Thus, each cT_FH subset includes clusters of particular interest that are associated with either in protection or the development of typhoid disease.

The phenotype of the clusters that might be involved in Ty21a vaccination and S. Typhi infection was examined in further depth in Figures 12A, B . Two clusters (4 and 7) have been shown above ( Figures 10 , 11 ) to be present in all four subsets of cT_FH and may play a role in the development of typhoid disease. Interestingly, cluster 4 is present in all four subsets contributing major proportions to cT_FH1 (18-30%) and cT_FH2 (20-30%) and minor components to cT_FH17 (5-20%) and cT_FH-DP (1-2%). We observed that cluster 4 is present at higher frequencies in TD than in NoTD participants at all time points of the study ( Figure 11 ) in the four cT_FH subsets. These results suggest that cluster 4 may be associated with the development of typhoid disease in participants who developed TD following wt S. Typhi challenge. Thus, we examined closely the phenotype of cluster 4 in TD and NoTD participants. Heatmaps showed that there were major differences between TD and NoTD participants ( Figure 12A ). In particular, as denoted by activation markers and cytokine production, we found that cT_FH subsets in TD participants cluster 4 (e.g., cT_FH17 CD69+ PD-1+ ICOS+ IL-2+ TNF-α+ IL-21+ and cT_FH1 CD69- CD154+ Granzyme B(GzB)+ PD-1+ ICOS+) were highly activated in TD as compared to NoTD participants ( Figure 12A ). Of note, the homing marker integrin α4β7 (gut), but not CCR4 (inflammatory) or CCR7 (lymph node), were expressed at higher levels in cT_FH2 and cT_FH17 subsets in TD compared to NoTD participants ( Figure 12A ). Interestingly, the cytokine patterns of cluster 4 in NoTD participants indicate more production of MIP-1β in cT_FH1 and cT_FH2 subsets and, importantly, IL-21 in the cT_FH2 subset ( Figure 12A ).

In contrast, cluster 7, although present in all four cT_FH subsets, the frequencies were higher in cT_FH1 (10-20%) with minor frequencies in cT_FH2 (0.5-2%) and cT_FH17 (0.5-2%) and cT_FH-DP (1-5%). We have described above ( Figure 11 ) that cluster 7 is present at higher in frequencies in NoTD than in TD participants at most time points of the study, suggesting that cluster 7 may be associated with the prevention of typhoid disease. Thus, we examined closely the phenotype of cluster 7 in TD and NoTD participants ( Figure 12B ). Heatmaps showed that there were minor differences between the TD and NoTD groups based on the expression of activation markers CD69, CD154, PD1, ICOS, and CD27 in cluster 7 of cT_FH1 and cT_FH-DP subsets ( Figure 12B ). However, there were some differences in the activation patterns in cluster 7 of cT_FH2, cT_FH17 and cT_FH-DP (e.g., CD69⁺) in NoTD compared to TD participants ( Figure 12B ). The homing markers integrin α4β7 (gut), CCR4 (inflammatory) and CCR7 (lymph nodes) were highly expressed on all cT_FH subsets in both TD and NoTD participants, except for CCR4 in T_FH17 in TD participants ( Figure 12B ). Interestingly, the cytokine patterns on cluster 7 shows that IL-17A, MIP-1β, TNF-α, granzyme B and IL-21 are highly expressed in both TD and NoTD participants ( Figure 12B ). However, IFN-γ expression was absent from all subsets associated with cluster 7 in NoTD participants while IFN-γ associated with cluster 7 in TD participants was expressed in cT_FH1 and cT_FH-DP ( Figure 12B ). Furthermore, IL-2 associated with cluster 7 was only present on cT_FH1 of TD participants while in NoTD participants, cluster 7 in cT_FH17 and cT_FH-DP but not cT_FH1 and cT_FH 2 produced IL-2 ( Figure 12B ). These contrasting phenotypes present in cluster 7 between TD and NoTD participants may account for its association with the absence of typhoid disease.

Cluster 5 was present mostly in cT_FH-DP (10-40%) ( Figure 10 ) and was observed to be at higher frequencies in TD than NoTD participants at all time points but was significantly (p<0.05) higher in cT_FH-DP after challenge (D7) ( Figure 11M ). These results suggested that cluster 5 may play a role in the development of typhoid disease. Thus, we examined closely the phenotype of cluster 5 in TD and NoTD participants using heatmaps ( Figure 13A ). Interestingly, cT_FH-DP (CD69+ CD154+ PD1^low ICOS+ CD27+) cluster 5 seems to be more activated in TD than in NoTD participants as shown by the phenotype of cT_FH-DP (CD69^low CD154^lowPD-1+ ICOS+ CD27+) ( Figure 13B ). The gut homing marker integrin α4β7 was expressed at higher levels on cT_FH-DP cluster 5 in TD than in NoTD participants ( Figure 13B ). Remarkably, we observed higher TNF-α,production and CD107a expression in cluster 5 associated with cT_FH-DP in NoTD which were absent in TD participants ( Figure 13B ).

Cluster 10 is unique to cT_FH-DP and was observed to be higher in frequencies in NoTD than in TD participants at most time points with a trend (p ≤ 0.1) to show increases at D7 following challenge ( Figure 11O ). Cluster 10 seems to be activated and express similarly all the markers in both TD and NoTD participants except for CD154 (absent from TD), IFNγ (lower in NoTD), GzB (lower in NoTD) and ICOS (lower in NoTD) ( Figure 13B ). Cluster 10 expressed α4β7 and CCR7 in NoTD and TD participants while CCR4 was expressed only in NoTD participants ( Figure 13B ). Interestingly, cytokines (IL-17A, IL-2, MIP-1β, TNF-α, IL-21) and granzyme B were highly expressed on both TD and NoTD participants ( Figure 13B ).

Finally, cluster 11 was present in cT_FH2 (2-8%) and cT_FH17 (2-8%) ( Figure 10B ). cT_FH2 cluster 11 was observed to be higher in frequency in NoTD than in TD participants following Ty21a immunization and after challenge (D7) ( Figure 11G ). These results suggest that cluster 11 may play a role in the prevention of typhoid disease. Thus, we examined closely the phenotype of cluster 11 in TD and NoTD participants using a heatmap ( Figure 13C ). Remarkably, activation markers (CD69, CD154, PD1, ICOS and CD27) in TD and NoTD participants were similarly expressed at high levels on cluster 11 of both cT_FH2 and cT_FH17 ( Figure 13C ). Next, we examined homing markers and noted that integrin α4β7, CCR4 and CCR7 were highly expressed on cT_FH2 and cT_FH17 in NoTD while in TD participants, the same pattern was present except for lower expression of integrin α4β7 on cT_FH2 and no expression of CCR4 on cT_FH17 cluster 11 ( Figure 13C ). Production of most cytokines (IL-17A, MIP-1β, TNF-α, IL-21) and granzyme B was present in cluster 11 in both subsets in NoTD and TD participants ( Figure 13C ). However, we observed that IL-2 expression was high in cT_FH2 and cT_FH17 cluster 11 in TD participants but absent from NoTD participants ( Figure 13C ). In addition, IFN-γ was expressed only on cT_FH2 cluster 11 in NoTD ( Figure 13C ). Altogether, these data suggest that defined clusters appear to be important in the prevention or development of typhoid disease.

To understand the contribution of cT_FH subsets in each phase of the study (Baseline, Ty21a immunization and wt S. Typhi challenge) to protection, we focused our analysis on some of the clusters that showed significant differences between TD and NoTD participants. For cT_FH1 subsets, no significant differences in the frequencies of cluster 7 in TD and NoTD were observed at baseline (D-28) ( Figure 14A ). However, following Ty21a vaccination (D-14), cT_FH1 cluster 3 showed a trend (p ≤ 0.1) to exhibit be present at higher levels in NoTD compared to TD participants, while cT_FH1 cluster 4 showed a trend (p ≤ 0.1) to exhibit an increase in TD compared to NoTD participants ( Figure 14A ). In the challenge phase, cT_FH1 cluster 4 continued to show increases in TD compared with NoTD ( Figure 14A ). However, in the challenge phase, cT_FH1 cluster 7 emerged showing a trend (p ≤ 0.1) to exhibit higher frequencies in NoTD than in TD participants ( Figure 14A ). For cT_FH2, we observed that cluster 7 had a trend (p ≤ 0.1) to show increases in NoTD compared to TD participants at baseline (D-28) ( Figure 14B ), which remain higher in the immunization phase ( Figure 14B ). However, in the challenge phase, three clusters emerge as important. Clusters 7 and 11 exhibited trends (p ≤ 0.1) to be higher in NoTD than TD participants, while cluster 4 exhibited a trend (p ≤ 0.1) to be higher in TD than in NoTD participants ( Figure 14B ). For cT_FH17, again we found that cluster 7 exhibited a trend (p ≤ 0.1) to be higher in NoTD than in TD participants at baseline (D-28) ( Figure 14C ). Interestingly, in the immunization phase for cT_FH17, there were 2 clusters of interest. Clusters 2 and 7 exhibited trends (p ≤ 0.1) to be higher in NoTD than in TD participants ( Figure 14C ). In the challenge phase, another cluster (9) appears to be important. Cluster 9 exhibited a trend (p ≤ 0.1) to be higher in NoTD than TD participants ( Figure 14C ). Finally, for cT_FH-DP, we observed 2 clusters of importance at baseline, namely clusters 4 and 7 ( Figure 14D ). Cluster 4 exhibited a trend (p ≤ 0.1) to be higher in TD than in NoTD participants, while cluster 7 was significantly (p<0.05) higher in NoTD than in TD participants at baseline (D-28) ( Figure 14D ). In the immunization phase, we found that significantly (p<0.05) higher frequencies of cT_FH-DP cluster 7 were present in NoTD compared to TD participants ( Figure 14D ). Finally, in the challenge phase, we observed that cT_FH-DP cluster 5 was significantly (p<0.05) higher in TD than in NoTD participants, while for cT_FH-DP cluster 10 we observed a trend (p ≤ 0.1) to be higher in NoTD than in TD participants ( Figure 14D ). Taken together, we observed that various clusters and cT_FH subsets might be of importance in each phase of the study and might play a role in the development or protection of typhoid disease.

Serum bactericidal antibodies are capable of killing bacteria directly through complement activation or opsonization, leading to their clearance from the bloodstream. The correlation between cT_FH frequencies and serum bactericidal antibodies, a representation of functional antibody responses, might be an additional indication of the strength of the immune response against S. Typhi. For example, higher cT_FH subset frequencies may suggest robust B cell activation and antibody production, leading to increased levels of bactericidal antibodies in serum. Conversely, lower cT_FH subset frequencies might indicate diminished B cell help and potentially lower levels of serum bactericidal antibodies, which could compromise the ability to control the bacterial infection. Thus, we examined the association between the frequencies of cT_FH subsets and S. Typhi bactericidal antibodies (SBA) to determine whether they might participate in protection. No significant correlations were observed between total cT_FH and SBA in either TD or NoTD at any time point ( Figure 15D ). However, we observed a strong significant positive correlation (r=0.76; p<0.05) between cT_FH1 and SBA in TD following Ty21a vaccination ( Figure 15D ). No significant correlations were observed between cT_FH2 and SBA in either TD or NoTD at any time point ( Figure 15D ). In contrast, we found a strong significant positive correlation between cT_FH17 and SBA in NoTD at pre-vaccination (r=0.83; p<0.05) and following Ty21a vaccination (r=0.81; p<0.05) ( Figure 15D ). Finally, we observed that there was a strong positive correlation between cT_FHDP and SBA in TD following challenge (r=0.80; p<0.05) (D28). In NoTD participants bactericidal activity a month after challenge (D28) was not measured, as preliminary experiments confirmed complete bactericidal activity consistent with residual ciprofloxacin in the samples (43). This was expected based on its pharmacokinetic profile (D28 coincided with the last day of antibiotic treatment for this cohort) (59). In sum, cT_FH subsets may be used as surrogate’s markers to indicate either susceptibility or protection against typhoid fever.