Naive and memory CD4+ T cell subsets can contribute to the generation of human Tfh cells

Summary CD4+ T follicular helper cells (Tfh) promote B cell maturation and antibody production in secondary lymphoid organs. By using an innovative culture system based on splenocyte stimulation, we studied the dynamics of naive and memory CD4+ T cells during the generation of a Tfh cell response. We found that both naive and memory CD4+ T cells can acquire phenotypic and functional features of Tfh cells. Moreover, we show here that the transition of memory as well as naive CD4+ T cells into the Tfh cell profile is supported by the expression of pro-Tfh genes, including transcription factors known to orchestrate Tfh cell development. Using this culture system, we provide pieces of evidence that HIV infection differentially alters these newly identified pathways of Tfh cell generation. Such diversity in pathways of Tfh cell generation offers a new framework for the understanding of Tfh cell responses in physiological and pathological contexts.


INTRODUCTION
Within germinal centers (GCs), T follicular helper cells (Tfh) shape B cell responses by promoting the development of high-affinity antibodies, isotypic switch, and B cell maturation (Crotty, 2019;Song and Craft, 2019). Tfh cells are classically identified in secondary lymphoid organs by the expression of CXCR5 and PD-1, which drive their positioning in these lymphoid organs (Sayin et al., 2018). The establishment of the Tfh phenotype is orchestrated by the transcription factor Bcl6 . To control B cell maturation and GC maintenance, Tfh cells express costimulatory molecules including CD40L and ICOS and secrete cytokines such as IL-21 and IL-4 (Crotty, 2019). Until recently, Tfh cell generation was mostly considered as a sequential process where Tfh cells arise after naive CD4 + T cell priming by dendritic cells in the T cell zone and acquisition, in the B cell zone, of fully effective functions following cognate interactions with B cells. However, several in vitro experiments have shown that memory CD4 + T cells can acquire Tfh cell features upon stimulation (Jacquemin et al., 2015;Lu et al., 2011;Pattarini et al., 2017). Thus, heterogeneous Tfh cell profiles might result from cellular plasticity in CD4 + T cell populations in lymphoid tissues.
Deciphering the various pathways leading to Tfh cell generation is of particular interest in chronic infectious diseases such as HIV where a paradoxical increase of dysfunctional Tfh cells has been reported (Colineau et al., 2015;Lindqvist et al., 2012;Perreau et al., 2013). As HIV infection is associated with architectural alterations of lymphoid tissues and CD4 + T cell exhaustion, we hypothesized that increase of Tfh could result from the unregulated reprogramming of CD4 + T cells into Tfh in lymphoid organs that sustain viral antigenic stimulation (Jeger-Madiot et al., 2019).
Addressing pathways of Tfh cell generation remains challenging in humans. Until recently, systems relying on lymphoid cell suspensions were mainly used to study the spread of HIV infection and the development of Tfh was not addressed.
Assuming that lymphoid cell cooperation synergizes to generate Tfh, we developed an original culture system based on the stimulation of splenic mononuclear cell suspensions. Using this strategy, we obtained a robust Tfh cell-like response including both non-GC and GC Tfh, as opposed to the use of peripheral blood mononuclear cells (PBMCs) which did not lead to generation of GC Tfh. Thanks to flow and mass cytometry combined with bulk RNA sequencing, we found that naive and memory CD4 + T cell subsets could differentiate toward a Tfh cell profile. Most importantly, the gain of Tfh cell phenotype by both naive and memory CD4 + T cell subsets was associated with specific transcriptional reprogramming. The reprogramming of various CD4 + T cell subsets leads to distinct phenotypes of Tfh, with differential expression of co-stimulatory molecules and cytokine secretion. As the impact of HIV infection on Tfh cell polarization was never addressed in a system involving a global lymphoid microenvironment, we investigated it using our original model. We showed that in vitro HIV infection modulates the acquisition of a Tfh cell profile by naive and memory CD4 + splenocyte subsets. Taken together, our results indicate that the heterogeneity of Tfh cell responses likely reflects the differential contribution of several CD4 + T cell subsets to the Tfh cell pool. Our work provides a framework for a better understanding of human Tfh cell biology under physiological and pathogenic conditions.

Antigen-experienced splenocytes lead to the generation of Tfh
As Tfh differentiate in the specific environment of lymphoid organs, we hypothesized that generation of Tfh would be optimized using splenic mononuclear cell suspensions. Cell suspensions from healthy donors were stimulated using CytoStim (Miltenyi), which acts as a T cell superantigen by cross-linking the T cell receptor and MHC molecules. Cells were then cultured for 10 days with IL-7, IL-12, and activin A ( Figure 1A), which are reported to be enhancers of Tfh cell generation (Carnathan et al., 2020;Durand et al., 2019;Locci et al., 2016). By evaluating the expression of CXCR5 and PD-1 on CD4 + T cells over time, we found that, after 3 days, the proportion of CXCR5 + PD-1 + Tfh among CD4 + T cells was doubled and then started to decline to reach 10% at day 10 ( Figure 1A). Moreover, the proportion of IL-21-producing cells and cells expressing ICOS among induced Tfh follows the same kinetics, suggesting that cells induced after splenocyte stimulation acquire Tfh cell functional features (Figures 1B and 1C). Of note, the proportion of live-dead stained cells among CD4 + T cells did not increase between day 3 (D3) and day 5 (D5), suggesting that the decrease of Tfh results from a return to a resting state rather than from cell death ( Figure S1A). Finally, Tfh were not more prone to cell death compared with other activated CD4+ T cells (CXCR5 À PD-1 + ) (Figure S1B) on day 3.
Then, we evaluated our stimulation protocol on PBMCs to test its capacity to promote Tfh in a nonlymphoid environment. To better characterize induced Tfh we distinguished GC Tfh, which expresses high levels of CXCR5 and PD-1, from non-GC Tfh (Haynes et al., 2007;Sayin et al., 2018;Vella et al., 2019). First, ex vivo circulating Tfh cell staining showed that total Tfh represented 7.1% (G2.2) of CD4 + T cells and were mainly PD-1 neg , while splenocytes showed 26.6% of PD-1 pos Tfh including 1.5% (G1.3) of GC Tfh. By submitting PBMCs to our stimulation protocol, we observed that Tfh expanded from 7.1% to 23.3% between D0 and D3 without giving rise to GC Tfh. These results suggested that GC Tfh cell generation was restricted to splenocyte stimulation ( Figures 1D and 1E) where GC Tfh D3 express similar levels of PD-1 compared with ex vivo GC Tfh and higher expression of CXCR5 ( Figures 1F and 1G). Interestingly, the generation of GC Tfh was reproduced using lymph node mononuclear cell suspensions (not shown). Such Figure 1. Antigenic stimulation of human splenic mononuclear cells mimics the T CD4 + response of GC reaction (A) Splenic mononuclear cells (splenocytes) were stimulated with CytoStim and cultured for 3 days in the presence of cytokines (IL-7, IL-12, activin) (A). CXCR5 and PD-1 expressions among CD4+ T cells were assessed. Representative flow plots showing CXCR5 and PD-1 expression on CD4 + T cells from ex vivo splenocytes and splenocytes cultured for 3, 5, and 10 days (left) and the percentage of Tfh among CD4+ T cells (right). (B) Representative flow plots showing IL-21 production by Tfh (left) and the percentage of IL-21-positive cells among Tfh (right). (C) Representative histogram showing ICOS expression among Tfh (left) and relative expression of ICOS among Tfh (right). (D) Gating strategy allowing the identification of PD-1 neg Tfh, non-GC Tfh, and GC Tfh among total CXCR5 + PD-1 + cells ex vivo or after 3 days of stimulation of splenocytes or PBMCs in polarizing cytokines. (E) Percentage of total CXCR5 + PD-1 + cells including PD-1 neg Tfh, non-GC Tfh, and GC Tfh among CD4 + T cells ex vivo or after 3 days of culture using splenocytes or PBMCs (n = 7-14). (F and G) Mean fluorescence intensity of PD-1 (F) and CXCR5 (G) expression on ex vivo GC Tfh and GC Tfh D3 splenic cells. (H) Representative flow plots showing IL-21 and IFNg production by non-GC Tfh D3 and GC Tfh D3 cells 3 days after splenocyte stimulation. (I) Percentage of IL-21-and/or IFNg-positive cells among non-GC Tfh D3 cells and GC Tfh D3 . (J) Gating strategy for analysis of Bcl6 expression in CD4 + T cells and histograms showing Bcl6 mean fluorescence intensity for GC Tfh D3 , non-GC Tfh D3 , and CXCR5 À PD-1 -CD4 + T cell subsets (n = 14). Each symbol represents an individual donor. A Wilcoxon matched pairs test was performed; *, p < 0.05; **, p < 0.005.

OPEN ACCESS
iScience 25, 103566, January 21, 2022 3 iScience Article an increase of GC Tfh using lymphoid mononuclear cell stimulation strongly suggests that an activated lymphoid environment supports a complete Tfh cell response. Thus, the opportunity to examine Tfh cell biology appears more relevant with the stimulation of mononuclear cells from lymphoid organs. As reported for GC Tfh and non-GC Tfh from tonsils (Brenna et al., 2020), we found that GC Tfh D3 were preferentially associated with IL-21 secretion, while their frequency was reduced among IFN-secreting cells ( Figures 1H and 1I). Consistently, Bcl6 expression was much greater in both GC Tfh D3 cells and non-GC Tfh D3 compared with non-Tfh ( Figure 1J).
To better characterize the signals required for Tfh cell generation in our system, we modified the protocol by variating CytoStim stimulation and cytokines. First, CytoStim was required to induce de novo Tfh (not shown) and to generate GC Tfh ( Figures S1C and S1D), showing that a sustained T cell receptor signal is needed for the generation of GC Tfh, as previously reported (Baumjohann et al., 2013). By modulating the cytokine environment, we found that addition of exogenous cytokines greatly enhanced GC Tfh D3 proportions (Figures S1C and S1D). Moreover, the addition of the cytokine cocktail was required to induce expression of IL-21 and ICOS (Figures S1E and S1F). Hence, using splenocytes, we designed a reproducible experimental design that supports the establishment of fully differentiated Tfh, with GC Tfh cell generation peaking after 3 days of culture.
Finally, we questioned the capacity of GC Tfh D3 to promote the maturation of CD27 hi CD38 hi plasma cells.
To this end, GC Tfh D3 and CXCR5 À PD-1 + CD4 + T D3 cells were sorted and co-cultured with autologous CD19 + B cells to induce their maturation. After 7 days, the proportion of CD27 hi CD38 hi plasma cells revealed that GCTfh D3 sustained plasma cell differentiation more efficiently than activated CD4 + CXCR5 À PD-1 + T D3 cells (Figures S1G and S1H). Altogether, these data indicate that GC Tfh D3 recapitulated phenotypically and functionally the features of activated bona fide Tfh. Thus, our experimental design appears suitable for studying the dynamics of Tfh cell responses to antigenic stimulation in lymphoid tissue.

Ex vivo and induced Tfh display distinct phenotypic landscapes and differentiation trajectories
To further investigate the landscape of ex vivo splenic Tfh and the phenotypic modifications induced after splenocyte stimulation, we performed deep immunophenotyping using mass cytometry with 29 different markers related to CD4 + T cell biology (Table 1). We focused on ex vivo (D0) and D3 time points of splenocyte stimulation, the latter corresponding to the peak of Tfh cell generation. A Uniform Manifold Approximation and Projection (UMAP) representation of CD4 + CXCR5 + cells from two donors was performed. Strikingly, CD4 + CXCR5 + cells from D0 and from stimulated splenocytes (D3) clearly clustered separately, with uniform distribution of cells at each time point for the two donors ( Figure 2A). In accordance with results presented in Figure 1, Tfh D3 displayed an activated phenotype, characterized by higher expression of activation markers (Tim3, PD-1, CD38, CD25, CD95, Ki67) and costimulatory molecules (CD28, ICOS, OX40), as compared with ex vivo Tfh (Tfh D0 ) ( Figure S2A). These observations were confirmed by the analysis of Tfh D3 from two additional donors (not shown). Of note, FoxP3 + expression does not vary significantly between D0 and D3, indicating that the splenocyte stimulation favors the generation of helper rather than regulatory follicular T cells at day 3 after stimulation. Using the k-means algorithm, four clusters of CD4 + CXCR5 + cells could be defined at D0 ( Figure 2B), namely, naive (cluster 3; CD45RA hi CD45RO lo PD-1 lo CD62L hi ) and non-activated (cluster 2; CD45RA int PD-1 int CD127 hi CD62L hi ) CD4 + CXCR5 + cells, together with GC Tfh (cluster 4; PD-1 pos ICOS pos CCR7 lo CXCR4 hi CD272 pos ) and non-GC Tfh (cluster 5; PD-1 int ICO-S pos CCR7 int CD127 pos ) cells, the latter being predominant (15.7% of total cells) ( Figure 2C). Coherently, CXCR5 expression was increased from D0 to D3, together with immune checkpoint molecules (CTLA-4, Tim3), confirming that Tfh D3 are activated after splenocyte stimulation ( Figure S2A). Although the expression intensity of chemokine and cytokine markers as CCR6, CCR5, and CD126 did not vary at D3, interestingly, expression of CXCR3, which has been associated with tonsillar Tfh, was increased (Brenna et al., 2020). Also, four clusters were identified at D3 ( Figure 2B). Two similar clusters exhibited a highly functional phenotype, namely, mature GC Tfh (cluster 7; PD-1 hi ICOS hi CD127 lo CXCR4 pos CD95 hi CD28 hi CXCR3 lo ) and emerging GC Tfh (cluster 8; PD-1 hi ICOS hi CD127 pos CXCR4 hi CD95 hi CD28 hi CXCR3 pos ). The two other clusters corresponded to proliferating Tfh (cluster 6; Ki67 hi PD-1 hi CXCR5 int CD62L hi CD45RO lo ) and quiescent Tfh (cluster 6; PD-1 pos CD27 lo CD62L lo ) ( Figure 2C).
In order to decipher whether induced Tfh originate from a linear differentiation of naive T cells or not, we performed a trajectory inference combined with a pseudotime analysis on total D0 and D3 CD4 + T cells. We  S2E). After pseudotime calculation, the first striking observation was that naive T cell metaclusters were the ''earliest,'' whereas Tfh cell subsets were the ''latest,'' at both time points ( Figures 2D and 2E), confirming that Tfh represent a terminal stage of CD4 + T cell differentiation. At D0, memory non-Tfh PD-1 pos and PD1 neg metaclusters were the closest to Tfh, suggesting that both are transitional subsets between naive T cells and Tfh. Moreover, ex vivo memory non-Tfh PD-1 pos cells did not seem to originate from naive T cells, which is coherent with resting spleens deprived of antigenic stimulation ( Figure 2D). At D3, naive T cell clusters were no longer the most abundant, being replaced by activated cells (i.e., naive activated  iScience Article and memory subsets) ( Figure 2E). At this time point, the Tfh cell metacluster seemed directly derived from both naive T cells and memory non-Tfh PD1 pos clusters, indicating multiple likely trajectories giving rise to early Tfh (cluster 7) and mature Tfh (cluster 10), respectively (Figures 2E and S2D). Although the subset of origin may dictate Tfh D3 phenotype, other subsets branched out from the global linear trajectory such as memory T regulatory cells, giving rise to either memory non-Tfh PD1 pos -derived follicular regulatory T cells (Tfr, cluster 4) or naive activated-derived regulatory T cells (Treg, cluster 14) ( Figures 2E and S2D). Finally, D3 trajectory from naive to Tfh cells remained similar to D0, with the exception of naive activated cells that emerged as an intermediate between naive and memory subsets. Overall, despite the important heterogeneity of stimulation-induced Tfh, D0 and D3 memory Tfh metaclusters shared a core group of differentially expressed markers (CXCR5, PD-1, CD57, CXCR4, CD45RO, CD95, CD126), highlighting that we were able to generate ex vivo-like Tfh. Taken together, these results suggest that our splenocyte stimulation protocol leads to strong induction of activated Tfh D3 cells either directly from naive CD4 + T cells or through memory PD-1 neg/pos CD4 + T intermediates.

Naive and memory CD4 + T cells take different developmental pathways to become Tfh
We optimized our experimental design to monitor the evolution of distinct CD4 + T cell subsets in the lymphoid environment. Since CD4 + T cell subsets exhibit phenotypic plasticity in response to environmental stimuli, we followed the dynamics of naive CD4 + T cells and memory non-Tfh. Two subsets of memory non-Tfh were distinguished. PD-1 pos memory CD4 + T cells (memPD-1 pos ) were defined as activated cells, whereas PD-1 neg (memPD-1 neg ) were defined as non-activated cells. Indeed, transcriptome analysis revealed that PD-1 expression was associated with recent T cell receptor stimulation and that memPD-1 pos cells displayed higher ICOS expression ( Figures S3A and S3B). In addition to studying the fate of naive and memory non-Tfh, we studied the fate of ex vivo Tfh (Tfh D0 ), which are mainly composed of non-GC Tfh (Figure 3A). Each CD4 + T cell subset was isolated from whole splenocytes using flow cytometry, then labeled with Cell Trace Violet (CTV) and re-incorporated into the negative fraction of splenocytes ( Figure 3B). Then, we investigated the ability of CTV-labeled CD4 + T cell subsets to express CXCR5 and PD-1 3 days after splenocyte stimulation (n = 10 donors). Tfh D3 derived from Tfh D0 cells were still positive for expression of CXCR5 and PD-1 markers ( Figures 3C and 3D). Remarkably, 27% G 10.4 of naive CD4 + T cells became Tfh after 3 days of culture. The proportion of Tfh D3 cells derived from memPD-1 pos cells was significantly higher (44.7% G 20.9) than the proportion derived from memPD-1 neg cells (19.4% G 4.8) ( Figures 3C  and 3D). This suggests a differential contribution of memory CD4 + T cell subsets to the global Tfh D3 pool according to their activation status.
Taking advantage of CTV staining, we further analyzed the expression of CXCR5 and PD-1 through the cell division cycles. First of all, Tfh maintained their CXCR5 and PD-1 expression through all division cycles. Second, the percentage of Tfh D3 cells peaked after only one division cycle for memPD-1 pos or Tfh D0 cells, whereas three and two divisions were required for memPD-1 neg cells and naive CD4 + T cells, respectively ( Figure 3E). These data suggest that, compared with other CD4 T cell subsets, the higher yield of Tfh D3 derived from memPD-1 pos results more from their higher capacity to convert into Tfh than from their overproliferation. (B) At day 0, D4 + T cell subsets were sorted according to the gating strategy presented in (A). Isolated CD4 + T cell subsets were stained with cell trace violet (CTV) and mixed back into the negative splenocyte fraction. Stimulation and culture were next performed as previously described ( Figure 1A). (C) Representative flow plots of CTV tracking for each stimulated CD4 + T cell subset 3 days after antigenic stimulation (top) combined with flow cytometry analysis of CXCR5 and PD-1 expression among CTV + cells at day 3 (bottom).
(E) Percentage of Tfh D3 cells according to the divisions of stimulated CD4 + T cell subsets.
(G) RNA sequencing was performed on CD4 + T cell subsets (Day 0) and the corresponding derived Tfh D3 counterparts (n = 2). Multidimensional scaling was used to better visualize transcriptomic proximity of different CD4 + T cells.
(H) Venn diagrams were used to highlight Tfh-associated genes ( iScience Article Moreover, whatever the CD4 + T cell subset, Bcl6 was expressed more in Tfh D3 cells than in non-Tfh (CD4 + CXCR5 -) derived from the same origin ( Figure 3F), suggesting the induction of a transcriptional program promoting Tfh cell differentiation in each non-Tfh CD4 + T cell subset.
To further define whether Tfh cell phenotype acquisition was associated with a Tfh-related transcriptional program, we performed a transcriptomic analysis of ex vivo isolated naive, memPD-1 pos , memPD-1 neg Tfh and of their respective Tfh D3 counterparts (n = 2 donors) ( Figure 3G).
Coherently with mass cytometry analysis, Tfh D0 and Tfh-derived Tfh D3 clustered apart. Indeed, by comparing the secreting capacities of Tfh D0 with those of Tfh D3 , we evidenced a great increase in IL-21 secretion, confirming a transition from a Tfh resting state to an activated one ( Figure S4A). Moreover, addition of polarizing cytokines greatly enhanced the frequency of naive, memPD-1 neg -and memPD-1 pos -derived Tfh D3 cells as compared with the culture without cytokines ( Figure S4B). Furthermore, IL-21 secretion and ICOS expression were potentiated for every Tfh D3 subset, independently of their origin (Figures S4C and S4D). These data support the idea that an antigen-stimulated lymphoid environment complemented with appropriate cytokines known to support Tfh cell development could favor acquisition of Tfh cell functions by any CD4 + T cell subtype.
To investigate whether orientation of each CD4 + T cell subset toward Tfh D3 was sustained by a specific transcriptional program, we used Venn diagram representations to highlight the overlap and specificities between sets of differentially expressed genes that are related to Tfh cell biology (Table 2). A core of multiple genes involved in Tfh cell biology overlapped between the transition from ex vivo CD4 + T cell subsets to their Tfh D3 cell relatives ( Figure 3H). Molecules involved in Tfh cell signaling as STAT1 (Choi et al., 2013); in Tfh cell function as TNFRSF4 (OX40), SLAMF1, and CXCL13 (Crotty, 2019); and in the Tfh cell transcription program as IRF4 and ZBTB7 (encoding for Thpok) (Kwon et al., 2009;Vacchio et al., 2019) were upregulated during the transition from ex vivo CD4 + T cells to Tfh D3 cells. Molecules associated with Tfh cell regulation of PRDM1 and IL2RA (Ditoro et al., 2018;Johnston et al., 2009) were also found to be upregulated, suggesting that establishment of the Tfh cell program is concomitantly associated with expression of regulatory checkpoints. Additional Tfh cell transcription factors were exclusively upregulated in naive and memPD-1 neg -derived Tfh D3 as MAF (Andris et al., 2017). Interestingly, Bcl6 was exclusively upregulated in memPD-1 neg -derived Tfh D3 . We found that transcription factors KLF2 and BACH2, identified as two inhibitors of the Tfh cell development Lahmann et al., 2019;Lee et al., 2015), were downregulated in the transition from each ex vivo CD4 + T cell subsets to Tfh D3 . Naive-derived Tfh D3 were associated with downregulation of PRKD2, which inhibits the transition from naive CD4 + T cells to Tfh (Misawa et al., 2020). CCR7 was downregulated in the transition from naive CD4 + T to Tfh D3 , as previously reported (Haynes et al., 2007). Downregulation of STAT4, which is involved in Tfh/Th1 commitment, was shared between naive and memory-derived Tfh D3 . In sum, transition of each CD4 + T cell subset toward Tfh D3 is characterized by its own ''original'' transcriptional program including regulation of genes implicated in Tfh cell iScience Article differentiation as well as in T cell activation. Altogether, these data suggest that heterogeneous Tfh cell profiles observed ex vivo and in vitro at day 3 could be driven by the differentiation of multiple CD4 + T cell subsets, differing from each other by their maturation and their activation status.
Tfh cell origins sustain Tfh cell heterogeneity at the peak of the antigenic stimulation To better characterize induced Tfh D3 of different origins, we first analyzed the intensity of CXCR5 expression, which mirrors Tfh cell maturation from non-GC to GC status (Kumar et al., 2021). The lowest CXCR5 mean fluorescence intensity was found in naive-derived Tfh D3 cells, and the highest in Tfh-derived Tfh D3 cells ( Figure 4A). Regarding the expression of memory and naive markers at D3, these data are in accordance with the mass cytometry results, where disparity of CXCR5 expression was observed among global Tfh D3 (Figures 2B and S2A). To test whether the level of CXCR5 expression might relate to specific functional profiles, we analyzed IL-21 and IFNg secretion. Naive-derived Tfh D3 were associated with higher IFNg secretion, whereas Tfh-derived Tfh D3 were associated with abundant IL-21 secretion ( Figure 4B). The cytokine secretion profile of memPD-1 pos -derived Tfh D3 was closer to that of Tfh-derived Tfh D3 , whereas memPD-1 neg -derived Tfh D3 harbored a cytokine secretion profile closer to that of naive-derived Tfh D3 . Coherently with a study showing that IFNg secretion is related to CXCR3 expression in tonsillar Tfh (Brenna et al., 2020), we found that CXCR3 expression was higher in naive-derived Tfh D3 ( Figure 4C). This higher expression of Th1 markers by naive-derived Tfh D3 might reflect the hybrid Tfh/Th1 profile, already described at an early stage of Tfh cell differentiation (Song and Craft, 2019). Regarding ICOS expression, naive and memPD-1 neg -derived Tfh D3 were enriched in ICOS + cells compared with memPD-1 pos and Tfh-derived Tfh D3 ( Figure 4C). We next evaluated the expression of CD40L, which is essential to the function of Tfh (Crotty, 2019). The CD40L expression pattern followed that of ICOS, suggesting that naive and memPD-1 neg -derived Tfh D3 provide more costimulatory signals than memPD-1 pos and Tfhderived Tfh D3 (Figures 4D and 4E). Overall, our results suggest that the heterogeneous landscape of Tfh might result from the distinct contribution of naive and memory CD4 + T cells to the global Tfh D3 cell pool.
We next evaluated the functionality of Tfh D3 cells derived from each CD4 + T cell subset, focusing on B cell maturation, total immunoglobulin (Ig) production and B cell survival ( Figure 4F). As compared with their native counterpart, naive-derived Tfh D3 cells exhibited an increased capacity to provide signals required for B cell maturation and survival. Coherently with the increased frequency of CD27 hi CD38 hi plasma cells, Ig production was higher in co-culture with naive-derived Tfh D3 than with their native counterpart ( Figure 4G). Tfh D3 derived from memPD-1 neg and memPD-1 pos CD4 + T cells, which were grouped because of the limited amount of available memory cells, were more efficient in helping B cell survival than their precursors, but did not promote higher B cell maturation and Ig production. Finally, Tfh-derived Tfh D3 showed similar capacities to provide B cell help as compared with Tfh D0 cells ( Figure 4G). As CD4 + T cell subsets differ in their proliferative ability during co-culture, variation in B cell maturation could result from quantitative rather than qualitative interactions. To investigate this, we compared the numbers of CD4 + T cells present at the end of the co-culture with B cells (Figure S5). Although equivalent numbers of naive CD4 + T cells and naive-derived Tfh D3 were found at the end of the B cell co-cultures, the frequency of plasma cells was higher with naive-derived Tfh D3 , showing that B cell maturation resulted more from qualitative help than from a higher frequency of CD4 + T cell partners ( Figure S5). In this line, although Tfh D0 did not proliferate as much as naive CD4 + T cells, they induced more B cell maturation. Therefore, gain of B cell help functions varies according to the origin of Tfh.
As Tfh D3 harbored heterogeneous phenotypic profiles, we hypothesized that distinct Tfh D3 cell subsets could promote distinct B cell responses. Thus, we measured Ig subtypes in the co-culture supernatants.
Although there was great variability between donors (n = 9) and we did not highlight any drastic B cell help specificities in the function of each Tfh D3 subset, some trends seemed to emerge. We found that Tfh and memory-derived Tfh D3 cells induced a slight increase of IgG1 production in comparison with naive-derived Tfh D3 (4-to 5-fold increase), whereas less production of IgG4 was obtained with memoryderived Tfh D3 (2.5-to 6-fold decrease). Conversely, naive-derived Tfh D3 were more associated with the promotion of IgA production (1.8-to 2.8-fold increase) ( Figure 4H). Altogether, these data suggest that naive and memory CD4 + T cell subsets contribute to the pool of Tfh D3 , resulting in the generation of multiple Tfh profiles, which in turn display slightly distinct B cell help properties.  Perreau et al., 2013). Thus, we hypothesized that our culture system provides a good model to assess whether and how the multiple pathways of Tfh generation are modulated by HIV infection. Stimulated splenocytes were exposed to HIV infection using a CCR5-tropic HIV-1 strain (HIV Yu2b ) ( Figure 5A). First, we checked that our protocol led to HIV Yu2b infection of splenocytes by analyzing p24 expression at day 3 after infection ( Figure 5B). After HIV Yu2b exposure splenocytes were infected, ranging from 0.15% to 2.97% of p24 + cells, which were not detected in the presence of reverse transcriptase inhibitors (not shown) and thus resulted from productive infection. We then isolated Tfh D3 generated under HIV Yu2b infection and compared their transcriptome profile to that of uninfected controls (n = 2 donors). A multidimensional scaling representation of the transcriptome revealed that HIV Yu2b exposure strongly impacts the genetic program driving Tfh cell differentiation. Tfh D3 generated under HIV Yu2b infection showed an intermediate transcriptomic profile between ex vivo CD4 + T cell subsets and Tfh D3 uninfected controls ( Figure 5C). A total of 990 genes were not upregulated in naive-derived Tfh D3 upon HIV Yu2b infection compared with infection-free conditions ( Figure 5D). An equivalent number of non-upregulated genes was found for other cell transitions. These genes were exclusive to each transition pathway, suggesting that HIV Yu2b infection selectively impacts the transcriptional program depending on the Tfh D3 cell precursor. To evaluate more precisely whether HIV Yu2b infection affected the Tfh cell differentiation program, we then focused on the expression of Tfh-related genes ( Table 2). We found that Tcf7, which encodes Tcf1 and is involved in early induction of Bcl6 (Choi et al., 2015;Xu et al., 2015), was downregulated under HIV Yu2b infection in all CD4 + T cell transitions toward the Tfh D3 profile ( Figure 5D). During the transition from naive CD4 + T cells to Tfh D3 under HIV Yu2b infection, we observed no downregulation of PRDK2 and BACH2, two inhibitors of the Tfh cell program. Coherently, no MAF upregulation was observed during the transition of naive CD4 + T cells toward Tfh D3 , whereas this factor is implicated in early Tfh cell commitment after immunization (Andris et al., 2017). Similarly, we found no upregulation of Bcl6, PRDM1, and ZBTB7B (Thpok) in memPD-1 negderived Tfh D3 cells under HIV Yu2b infection. These results suggest that naive and memPD-1 neg -derived Tfh D3 might harbor a defective Tfh cell phenotype. Indeed, CD28 and TNSFR4 (OX40) were not upregulated during the transition of memPD-1 pos cells toward Tfh D3 ( Figure 5D), which is in accordance with defective co-stimulatory functions reported in splenic Tfh from chronically infected patients (Colineau et al., 2015). Hence, PDC1 was downregulated in memPD-1 pos -derived Tfh D3 cells. Moreover, STAT5A and PRDM1, two key regulators of Tfh cell development, were not upregulated under HIV Yu2b infection in memPD-1 pos -derived Tfh D3 . Overall, our transcriptional analysis showed that Tfh cell developmental and functional programs were altered by HIV Yu2b infection and that the effect of HIV infection on the Tfh D3 cell subsets varied depending on their precursors. Surprisingly, cytometry analysis of Tfh D3 cell subsets did not indicate a major impact of HIV Yu2b infection on global Tfh D3 cell proportion regarding their origin ( Figure 5E), suggesting that HIV Yu2b infection did not preferentially orient any CD4 + T cell subsets toward Tfh D3 at this time point. Coherently, HIV Yu2b infection did not induce any preferential cell death among the Tfh D3 cell subsets ( Figure S6).

HIV(-1) infection shapes Tfh cell differentiation
Tfh are a major HIV reservoir compartment, which is one of the major obstacles to HIV eradication. Therefore, we investigated the infectious status of Tfh D3 subsets derived from various CD4 + T cell subsets. We found a preferential infection of memPD-1 neg -derived Tfh D3 compared with memPD-1 pos -derived Tfh D3 ( Figure 5F), suggesting that memPD-1 neg -derived Tfh D3 could preferentially contribute to the HIV reservoir. Finally, to evaluate whether HIV Yu2b infection contributes to the altered phenotype of Tfh D3 cell subsets, we performed mass cytometry analysis to examine phenotypic differences between p24 neg and p24 pos Tfh D3 iScience Article ( Figure 5G). Interestingly, as compared with p24 neg Tfh D3 , the percentage of p24 pos Tfh D3 displaying an activated phenotype (CD127, CD27, CD38) was reduced, whereas FAS was overexpressed ( Figure 5G). The defective expression of activation markers by p24 pos Tfh D3 confirmed transcriptomic analysis (Figure 5D). Finally, a trend to more expression of ki67 was observed in p24 pos Tfh D3 . Thus, p24 pos Tfh D3 presented a defective activation status while maintaining higher expansion ability.
Altogether, these results suggest that qualitative alterations observed in the Tfh cell compartment could result from the differential impact of HIV infection on the transition of ex vivo CD4 + T cell subsets toward Tfh D3 and from the capacity of Tfh D3 cells to sustain HIV reservoirs. These data suggest that many parameters, including the pathway of Tfh cell differentiation, could contribute to the accumulation of dysfunctional Tfh and the establishment of HIV reservoirs in lymphoid organs.

DISCUSSION
Most research investigating Tfh cell biology is based on the use of PBMCs, and few recent studies integrate the microenvironment of lymphoid organs for the study of Tfh cell responses. In these models, cell suspensions from lymphoid organs are mainly used to test the immunogenicity of vaccine candidates or drugs (Schmidt et al., 2020;Wagar et al., 2021). Among complementary approaches, lymphoid tissue explant models enable the study of the spread of HIV infection (Grivel and Margolis, 2009) in a situation much closer to in vivo conditions than our approach, although they do not allow the study of the impact of HIV infection on Tfh development. Therefore, we designed an alternative lymphoid cell-based model that makes a valuable contribution to the study of Tfh cell development and the impact of HIV infection on it. We took advantage of antigen-experienced splenocytes to promote functional Tfh including GC Tfh, which present similarities with those generated in vivo (Brenna et al., 2020;Vella et al., 2019). Moreover, these functional induced Tfh were susceptible to HIV infection.
Our data clearly showed that addition of cytokines known to support Tfh development and functions are strong potentiators of Tfh D3 induction and function regardless of their origin. These results are particularly coherent with recently published studies showing that adjuvanting HIV vaccine candidate with activin A promotes Tfh responses in a simian model (Carnathan et al., 2020) and that production of IL-12 and activin A by tonsillar myeloid cells sustains Tfh development (Durand et al., 2019). Considering that a single cytokine may vary in its effect depending on the micro-environment (Touzot et al., 2014), integration of cellular and molecular factors related to lymphoid organs is one strength of our culture system in comparison with co-culture assays.
One can assume that varying molecular and cellular environments might impact the induction of Tfh cell responses. Indeed, using our experimental design, antigen-experienced PBMCs did not lead to the generation of fully differentiated GC Tfh. Recently published data comparing Tfh isolated from tonsils and from blood showed that independent tissues present distinct proportions of Tfh in different maturation stages (Kumar et al., 2021). Here, our data provide new insights suggesting that the lymphoid environment is required to support the generation of fully differentiated GC Tfh. Indeed, many molecular or cellular factors would explain the propensity of splenocytes to support complete Tfh cell differentiation. In comparison with PBMCs, splenocytes are enriched in pro-Tfh subsets such as cDC2 cells and macrophages (Durand et al., 2019) and data not shown). Furthermore, compared with PBMCs, splenocytes are enriched in B cells, which are likely to play a role in the differentiation of Tfh in humans (Chavele et al., 2015).
We have shown that, in our model, Tfh cell induction requires T cell receptor signaling and polarizing cytokines. Thus, the lymphoid microenvironment initiates the Tfh differentiation program concomitantly with CD4 + T cell activation. Antigen-experienced splenocytes led to highly reproducible Tfh responses, which peaked 3 days after stimulation and then declined at day 5, indicating transitory T cell activation. Since Tfh are maintained with chronic exposure to antigen in lymphoid organs, one would expect that a second antigenic stimulation would maintain Tfh cell generation over time. However, in our culture system, the superantigen magnified CD4 + T cell activation and induced deep changes in the cellular composition of the splenocyte cell suspension, thus impeding Tfh cell program maintenance. Consequently, multiple antigenic stimulations would require further settings of our experimental design, such as renewal of lymphoid cells.
We have demonstrated that any subset of CD4 + T cells, including memory CD4 + T cells, can shift to a Tfh profile as early as day 3 after stimulation, leading a gradient of Tfh phenotype and functions. In our experimental conditions, we evidenced specific trajectories linked to the activation status of CD4 T cells ll OPEN ACCESS 14 iScience 25, 103566, January 21, 2022 iScience Article ( Figure 2). Previous in vitro studies have shown that memory CD4 + T cells can acquire Tfh features upon stimulation (Del Alcazar et al., 2019;Jacquemin et al., 2015;Lu et al., 2011;Pattarini et al., 2017). Here, we showed that orientation of memory CD4 + T cells toward the Tfh profile was sustained by a specific Tfh differentiation program. Indeed, analysis of the DEG between D0 and D3 suggested that each transition from ex vivo CD4 + T cell subsets to their Tfh D3 counterparts followed specific pathways of differentiation, even though a core of multiple genes involved in Tfh cell biology was conserved (Figure 3). Our experimental design induced Tfh activation into highly functional GC Tfh D3 , whereas naive-derived Tfh D3 appeared less mature and memory-derived Tfh D3 cells presented an intermediate phenotype (Figures 2 and 3). This could be due to a delay in the acquisition of Tfh cell features by naive CD4 + T cells, which are supposed to require a multistep differentiation pathway (Crotty, 2014). Unfortunately, since the tracking of CD4 + T cells was not possible after 3 days of culture, we could not test the percentage and the phenotype of naive-derived Tfh after 5 and 10 days. Hence, high proportions of Tfh D3 would result from the differentiation of various CD4 + T cell subsets displaying intrinsic capacities to acquire Tfh cell features.
Furthermore, our data suggest that Tfh D3 derived from distinct CD4 + T cell origins provide different B cell help. For instance, even though naive-derived Tfh D3 produced less IL-21 than memory-derived Tfh D3 , they expressed more ICOS and CD40L. We propose here that naive-derived Tfh D3 keep the expression of CD40L on their surface to interact with B cells and complete their differentiation into GC Tfh, whereas CD40L expression is downregulated on more mature Tfh D3 cells to possibly prevent the activation of non-cognate B cells (Yellin et al., 1994). CD127 expression appeared to reflect distinct stages of Tfh differentiation or activation ( Figure 2D). Hence, expression of CD127 was lower in Tfh D3 than in their original counterpart. Coherently with the literature reporting low expression of CD127 on GC Tfh (Iyer et al., 2015;Lim and Kim, 2007), we identified a mature GC cluster as the one that expressed the least CD127 among Tfh D3 ( Figure 2D).
However, such phenotypic heterogeneity does not translate into dramatic differences in the capacities of Tfh D3 to induce B cell maturation and Ig production ( Figure 4). Of note, we performed T-B cell cocultures with memory B cells that are more prompt to mature than naive B cells (Locci et al., 2016;Pattarini et al., 2017;Ugolini et al., 2018). Consequently, memory B cells are probably less sensitive to the various B cell help abilities of distinct CD4 + T cells. In the same line, isotype class switching and maturation of antigen affinity would be of interest in evaluating the activation of Tfh D0 into GC Tfh D3 .
The capacity of memPD-1 pos CD4 + T cells to generate Tfh was higher in comparison with memPD-1 neg CD4 + T cells, showing that Tfh cell generation differs according to T cell activation status. Lymphoid tissues are particularly enriched in memPD-1 pos CD4 + T cells as compared with PBMCs. The propensity of memPD-1 pos CD4 + T cells to orient toward a Tfh cell profile might confer a significant advantage by rapidly sustaining and diversifying B cell responses after antigenic exposure. However, rapid Tfh cell conversion could be deleterious in HIV infection where memPD-1 pos CD4 + T cells accumulate (Del Alcazar et al., 2019) and lymphoid structures are altered (Estes, 2013). Thus, unregulated Tfh cell generation could be induced in this context.
Transcriptomic analysis revealed that HIV infection deeply impacts the transcriptomic program of Tfh cell differentiation. Moreover, deep immunophenotyping evidenced defective activation of p24 pos Tfh D3 . Under in vitro HIV infection, we showed that the percentage of CD127 + cells among p24 pos Tfh D3 is reduced compared with p24 neg Tfh D3 ( Figure 5D). This observation is in accordance with previous work showing that expression of CD127 is lost on a large proportion of peripheral T cells in HIV-1-infected patients with lymphopenia (Chiodi et al., 2017). One can suppose that CD127 loss contributes to the higher susceptibility of p24 pos Tfh D3 to cell death, which is coherent with the higher percentage of FAS + cells among p24 pos Tfh D3 cells as compared with p24 neg Tfh D3 . These results give new insight into the induction of defective Tfh in HIV-infected patients. Interestingly, we found a preferential infection of memPD-1 negderived Tfh D3 compared with memPD-1 pos -derived Tfh D3 . However, as compared with uninfected control, this preferential infection does not result in a differential contribution of the distinct CD4 + T cell precursors to the overall Tfh D3 cell pool. Longer tracking of CD4 + T cell subsets would be helpful to investigate potential variations in the respective contribution of derived Tfh D3 subsets at later time points. Lastly, characterization of Tfh cell pathways focusing on HIV-specific CD4 + T cells would be very interesting. However, the length of in vitro culture is too short to expect the CD4 + T cell priming allowing the study of HIV-specific Tfh. Hence, using splenocytes from HIV-infected patients could be relevant for this purpose (Colineau et al., 2015).

OPEN ACCESS
iScience 25, 103566, January 21, 2022 iScience Article Altogether, our experimental model provides first-order information on the multiple pathways of Tfh development and activation, which so far are unidentified in a human lymphoid environment. The applicability of this model to HIV infection allowed us to confirm functional defects of Tfh in HIV-infected patients in the light of the newly identified pathways of Tfh cell induction.

Limitations of the study
This study mostly relies on the in vitro culture systems. Further studies are required to evaluate if the differences observed for naive-or memory-derived Tfh cells differentiated in vitro recapitulate the biology of Tfh cells in vivo.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:

DECLARATION OF INTERESTS
The authors declare no competing interests.

Spleen processing and freezing
During delivery, splenic tissues were maintained at 15-22 C and RPMI 1640 (Thermo Fisher Scientific) supplemented with penicillin-streptomycin (100 U/mL, Thermo Fisher Scientific) was used as delivery medium. Spleens were comminuted mechanically in a culture dish containing RPMI with penicillin-streptomycin (100 U/mL, Thermo Fisher Scientific), filtered through a cell strainer (70 mm, Sigma Aldrich). After decantation, the cell suspension was transferred into a 50 mL tube (Falcon) containing Pancoll (Pan Biotech) and centrifuged. Then, cells were washed twice with RPMI medium and frozen at À150 C in medium containing fetal bovine serum (FBS) (Sigma Aldrich) with 10% dimethyl sulfoxide (DMSO) (Sigma Aldrich).

Splenocyte cultures
Cells were thawed and washed twice in pre-warmed RPMI and then resuspended in RPMI supplemented with 10% FBS, L-glutamine 2 mM, penicillin 2 units/mL, 1 mg/mL streptomycin (complete RPMI) and DNase (10 ng/mL, Sigma Aldrich) overnight. Cells were washed and stimulated for 3 hours with CytoStim (Miltenyi, 2 mL/million cells) in complete RPMI at 37 C. CytoStim is a bi-specific antibody which binds simultaneously to the TCR and to MHC, thus cross-linking effector and memory CD4 or CD8 T cells and antigen presenting cells. CytoStim provides a strong polyclonal T cell stimulation, without any TCR Vbeta restrictions. After centrifugation, cells were resuspended in a polarizing medium consisting of complete RPMI supplemented with 100 ng/mL activin A, 5 ng/mL IL-12, 4 ng/mL IL-7 (Miltenyi, premium grade). Splenocytes were cultured in 24-well plates (Dutscher) at the concentration of 2 million cells/1 mL/well for 3 days. For the cell tracking experiment, CD4 + T cell subsets of interest were sorted, using a BD FACSARIA II (BD Biosciences). Then, CD4 + T cells were stained with cell trace violet (Thermo Fisher) for 20 min at 37 C and mixed back into the conserved fraction containing all other cells. The same protocol as before was applied.

Assessment of B cell help by T-B co-culture
CD4 + T cells of interest were isolated from ex vivo or stimulated splenocytes using a BD FACSAria II (BD Biosciences), with over 95% purity. Fresh autologous total or memory B cells were sorted (CD19 + CD27 + IgD À ). 20,000 CD4 + T cells and 20,000 B cells were co-cultured for seven days in the presence of CytoStim. The culture medium contained 1 / 2 medium from previous splenocyte stimulation and 1 / 2 complete RPMI. After seven days, maturation of B cells (CD27 and CD38) was assessed by flow cytometry and immunoglobulin concentrations were determined using Luminex with the antibody Isotyping 7-Plex Human Procarta-Plex TM Panel (Thermo Fisher Scientific).

HIV production
Replicative HIV Yu2b lab strain (CCR5 tropic) was generated as previously described (Cardinaud et al., 2017) by transfection of 293T cells with the Calcium Phosphate Transfection Kit (Sigma Aldrich). After transfection, cells were cultured in DMEM (Thermo Fisher Scientific) supplemented with 10% FBS, 2 mM L-glutamine, 2 units/mL penicillin, 1 mg/mL streptomycin. Supernatants containing HIV Yu2b were harvested 3 times every 12 hours from transfection. After centrifugation for removal of cellular debris, supernatants were filtered and then frozen at À80 C. The Gag-p24 content of all viral stocks was measured using an ELISA (PerkinElmer).

Splenocyte infection by HIV
After CytoStim stimulation for 3 hours, splenocytes were exposed to HIV Yu2b at a concentration ranging from 150 to 700 ng/mL of p24 for 4 million splenocytes. Infection was performed at 37 C for 3 hours in the presence of diethylaminomethyl (DEAE)-dextran (Sigma Aldrich) at 4 mg/mL. Then, splenocytes were washed twice and cultured for 3 days in a polarizing medium as before.

Flow cytometry staining and cell sorting
Before flow cytometry and cell sorting, splenocytes were systematically stained for cell viability using Viobility Dye (Miltenyi) at room temperature for 15 minutes. After two wash steps, splenocytes were stained in PBS 1 3 5% FBS with antibodies directed against chemokine receptor (CXCR5; CXCR3) (Miltenyi) for 10 minutes at 4 C. Splenocytes were washed and cell surface staining was performed (CD4, CD19, PD-1, ICOS)