Themis^flox/flox dLck-cre mice (cKO) were generated and kindly provided by Dr. Nicholas Gascoigne at the National University of Singapore. C57BL/6J (#000664), Rag1 ^–/– mice (#034159), CD45.1 (#002014), SMARTA mice (#030450) were purchased from the Jackson Laboratory and maintained in Xiamen University Laboratory Animal Center. All mice were bred and maintained under specific pathogen-free conditions at the animal center at Xiamen University. All experiments were performed using sex-matched littermates ages between 6 and 12 weeks and were approved by the Institutional Animal Care and Use Committee of Xiamen University.

All antibodies used for flow cytometry staining were purchased from BioLegend, eBioscience, Miltenyi Biotec and BD Biosciences: anti-CD16/CD32 (2.4G2), anti-CD4 (RM4-4), anti-CD8 (53-6.7), anti-CD45.1 (A20), anti-CD45.2 (104), anti-CD44 (IM7), anti-CD62L (MEL-14), anti-PD-1 (29F.1A12), anti-CX3CR1 (SA011F11), anti-TCF-7/TCF-1 (S33-966), anti-CD101 (Moushi101), anti-CD19 (D3), anti-B220 (RA3-6B2), anti-CD38 (90), anti-IgM (RMM-1), anti-IgD (IA6-2), anti-CD95 (Jo2), anti-CD138 (281-2), anti-GL7 (GL7), anti-CXCR5(2G8), rat IgG2a (MRG2a-83), anti-SLAM(TC15-12F12.2), anti-ICOS(15F9), anti-Bcl-6(K112-91), biotinylated anti-rat IgG (553894), APC Streptavidin (405207), Annexin V/7-AAD kit (559763), CFSE (C34554), CTV (C34557). The viral peptide gp33-41 (KAVYNFATM) and gp61-80 (GLKGPDIYKGVYQFKSVEFD) were purchased from GenScript. LCMV-specific MHC-I tetramers H-2D^b/gp33–41 were purchased from Reagent platform of Cancer Research Center of Xiamen University. PMA (phorbol 12-myristate 13-acetate) was purchased from Sigma. Ionomycin was purchased from Yeasen.

LCMV Armstrong (LCMV Arm) and LCMV Clone13 (LCMV C13) strains were propagated in BHK-21 cells and titrated on Vero cells. For infection, 2×10⁵ plaque-forming units (PFUs) of LCMV Arm were intraperitoneally injected into mice to establish an acute viral infection, 2×10⁶ plaque-forming units (PFUs) of LCMV C13 were intravenously injected into mice to establish a chronic infection in mice. In adoptive transfer experiments, recipient mice were infected 1 day after cell transfer.

Viral titers were determined in mouse serum by qPCR assay as previously described (42). Briefly, a standard curve was generated by serial dilution of a virus stock with a predetermined titer. RNA was extracted from standard samples and tested samples using a TIANamp virus RNA kit (TIANGEN, no. DP315-R), cDNA was synthesized using the primer gp33 (reverse; 59-CATTCACCTGGACTTTGTCAGACTC-39) and TransScript first-strand cDNA synthesis supermix (TransGen Biotech, AT301-03). Virus titers were detected by quantitative PCR with the primers gp33 (forward; 59-GCAACTGCTGTGTTCCCGAAAC-39) and gp33 (reverse) and then converted to PFU/ml according to the standard curve.

Naïve CD4⁺ T cells were purified using a negative selection strategy. Briefly, single-cell suspensions were freshly isolated from spleens and lymph nodes, followed by red blood cell lysis. A panel of anti-mouse biotinylated antibodies was used to deplete non-CD4⁺ T cells, including anti-B220 (RA3-6b2), anti-CD8 (53-6.7), anti-CD11b (M1-70), anti-CD11c (N418), anti-CD25 (PC61.5), anti-CD44 (IM7), anti-NK1.1 (PK136), anti-Ly-6G/Ly-6C (RB6-8C5), anti-F4/80 (BM8), and anti-TER119 (Ter119). All antibodies were sourced from BioLegend. The cells were incubated with the antibodies for 20 minutes at 4°C, followed by a 10-minute incubation with streptavidin-conjugated magnetic beads (BeaverBeads Streptavidin, Beaver, 22307-10). The cell-bead mixture was placed on a magnetic stand (DynaMag™-2, ThermoFisher, 12321D), allowing unbound cells to be aspirated off, while naïve CD4⁺ T cells were retained and subsequently collected. Purified naïve T cells were phenotypically confirmed as CD4⁺CD62L^hi CD44^lo by flow cytometry, achieving a purity exceeding 95%.

Freshly isolated naïve SMtg CD4⁺ T cells were labeled with CFSE according to the manufacturer’s protocol. Briefly, cells were washed once in prewarmed phosphate-buffered saline (PBS, 37°C) and centrifuged at room temperature to collect the pellet. The cell suspension was adjusted to a concentration of 2–3 × 10⁶ cells/ml in prewarmed PBS. CFSE dye was then added, and the cells were incubated in a 37°C water bath for 10 minutes. Following labeling, excess dye was removed by washing the cells with PBS and culture medium. The labeled cells were resuspended and prepared for downstream experiments.

Naïve SMtg WT and cKO cells, marked with distinct congenic labels, were mixed at a 1:1 ratio, labeled with CTV, and adoptively transferred into preconditioned recipient mice via intravenous injection. Recipient mice, preconditioned by irradiation with 5.5 grays using RS2000 160-keV equipment (Rad Source Technologies), were Rag1 ^–/– hosts. At designated time points post-transfer, donor cells from recipient mice were analyzed by flow cytometry.

Spleens from infected mice were harvested and fixed in PBS-buffered formalin before embedding in paraffin. 5 μm tissue sections were prepared and stained with H&E following the manufacturer’s protocol (Boster, AR1180-100). Microscopic images were captured using a Motic VM1 digital slice scanning microscope.

After treatments as indicated in the figure legends, spleens were fixed in BD Fix/Perm Solution (Cat No: 51-2090KZ) at 4°C overnight. Fixed spleens were washed with PBS for 10 minutes on a shaker, followed by incubation in 30% sucrose (Sigma, V900116) overnight at 4°C. The tissues were embedded in OCT (Sakura, 4583), sectioned at a thickness of 7 μm using a Leica CM1950 cryostat, and permeabilized with 0.3% TritonX-100 in PBS for 10 minutes. Blocking was performed with 5% BSA in PBS for 60 minutes at room temperature. Sections were immunostained with antibodies targeting IgD, CD3, and GL7, incubated in a dark chamber for 24 hours at room temperature. Confocal imaging was conducted using a Leica TCS SP8 microscope.

Single-cell suspensions were prepared from spleens, thymuses, and peripheral lymph nodes, followed by red blood cell lysis. Cells were resuspended in RPMI 1640 medium containing 2% FBS. For surface marker staining, antibodies were added to cells and incubated for 30 minutes at 4°C, followed by washing with flow cytometry staining buffer. CXCR5 staining was performed in three steps: incubation with purified anti-CXCR5, followed by biotinylated anti-rat IgG, and finally APC-labeled streptavidin. Each step was performed at 4°C in CXCR5 staining buffer (PBS with 0.5% BSA, 2% FCS, and 2% normal mouse serum). Intracellular transcription factor staining was conducted after surface marker staining by fixing and permeabilizing cells in Fix/Perm buffer (eBioscience, 00-5521-00) for 30 minutes, followed by incubation with transcription factor antibodies for 1 hour at room temperature.

SMtg WT and cKO cells (0.5 × 10⁴ of each type) were intravenously injected into B6 hosts, which were infected with LCMV C13–24 hours later. At 21 dpi, SMtg WT and cKO cells were sorted from 15 mice. For sorting, CD4⁺ T cells were negatively selected and stained with fluorochrome-conjugated antibodies in sorting buffer (PBS with 1 mM EDTA, 25 mM HEPES, and 1% FBS). Sorting was performed using a BD FACSAria Fusion sorter (BD Biosciences). Flow cytometry data were acquired on a Fortessa or LSRFortessa X-20 (BD Biosciences) and analyzed with FlowJo software (version 10, Treestar).

Serum samples were collected at 8, 21, 30, and 40 dpi. LCMV-specific IgG concentrations were determined using ELISA. Microtiter plates (ThermoFisher, 439454) were coated with lysates from LCMV Cl13-infected baby hamster kidney (BHK) cells. Plates were blocked with 3% BSA in PBS, and serum samples were incubated for 2 hours at room temperature. After incubation with biotin-conjugated anti-IgG (Southern Biotech, 1030-08), plates were treated with horseradish peroxidase-conjugated streptavidin (Southern Biotech, 7105-05) for 1 hour, followed by color development using tetramethylbenzidine substrate (Bioleged, 421101). Absorbance was quantified using a SpectraMax Absorbance Reader (BAOCHENG).

Single-cell RNA sequencing libraries were prepared using the DNBeLAB V2 kit (BGI Genomics), strictly following the manufacturer’s guidelines (43). Briefly, cells were sorted via FACS and washed once with PBS supplemented with 0.04% BSA, then resuspended in PBS containing 0.04% BSA. Following reverse transcription and cell barcoding in droplets, emulsions were broken, and cDNA was purified. PCR amplification was subsequently performed as follows: 98°C for 45 seconds, followed by 14 cycles of 98°C for 20 seconds, 67°C for 30 seconds, and 72°C for 1 minute, with a final extension at 72°C for 1 minute. The resulting amplified cDNA was then utilized for constructing the gene expression library. Quantification of the scRNA-seq libraries was carried out using the Qubit dsDNA HS Assay Kit (Invitrogen) and a High-Sensitivity DNA chip run on a Bioanalyzer 2100 system (Agilent). Sequencing was performed on the DNBSEQ C4 platform (BGI).

The raw sequencing data (GSE284735) were processed using Cell Ranger (10X Genomics) to generate a gene expression matrix containing UMIs (unique molecular identifiers), following the default parameters. The gene expression UMI matrix was further analyzed in Seurat (version 5.1.0) (44). Cells with a number of captured genes falling within two standard deviations of the mean were retained. Cells with mitochondrial RNA content above the 95 percentile were excluded from further analysis to minimize technical biases.

Trajectory analysis of the CD4+ T cell response during LCMV-c13 infection was conducted using Slingshot2 (version 2.12) in R (45). UMAP-based dimensionality reduction was performed on cells from the 21dpi time point (2 samples) after removing cells from infrequent populations. Slingshot2 was then employed to identify the lineage structure and infer pseudotime for the cells, using the naive phenotype cluster as the root node.

Slingshot2 fits principal curves through the data, enabling the identification of differentiation trajectories based on gene expression profiles. These trajectories were visualized using UMAP to illustrate the progression of CD4+ T cell differentiation across time points (21 dpi). Pseudotime was calculated along the inferred curves, mapping the differentiation process in the context of LCMV-c13 infection. This analysis considered dynamic gene expression changes across different stages of differentiation, with pseudotime values used to order the cells along these trajectories.

All summarized data are shown in graphs with mean ± SEMs. Unpaired or paired Student t tests and ANOVA analyses were used as indicated and GraphPad Prism software (version 8.0). Differences with p-values>0.05 were considered non-significant (ns). p-values<0.05 were considered significant (*p<0.05; **p<0.01; ***p<0.001; ****p<0.0001).

To determine whether Themis is involved in CD4⁺ T cell differentiation in chronic LCMV infection, we evaluated the expression of Themis in different CD4⁺ T cell subsets derived from LCMV C13-infected C57BL/6J (B6) mice. In this infection model, CD4⁺ T cells exhibited a differentiation bias toward T_FH cells, which participate in the formation of germinal centers (GC) and promote B cell somatic hypermutation and antibody production (26, 27, 29). We found that at 5 days post infection (dpi), there was a clear effector CD4⁺ T cell (CD44⁺) population emerging while the majority of CD4⁺ T cells were still of naive phenotype (CD44^–) ( Figure 1A , left). The effector CD4⁺ T cells were further subdivided into T_FH (CXCR5⁺) and non-T_FH (CXCR5^–) cells based on their CXCR5 expression ( Figure 1A , left). Under this gating strategy, we evaluated the expression of Themis ( Figure 1A , middle) and found that Themis expression in T_FH and non-T_FH cells was 2.5-fold and 1.6-fold higher than that in naive CD4⁺ T cells, respectively ( Figure 1A , right). To obtain more clues, we also divided effector CD4⁺ T cells into Themis^neg/lo, Themis^int and Themis^hi subsets ( Figure 1B , top row), and examined their CXCR5 and SLAM expression ( Figure 1B , middle row). CXCR5 and SLAM counterstaining more accurately distinguish T_FH cells (CXCR5⁺SLAM^–) (46). This time we found a more evident relationship between Themis expression and T_FH cell differentiation, with the Themis^neg/lo, Themis^int and Themis^hi subsets having the smallest, intermediate and largest proportions of T_FH cells, respectively ( Figure 1B , middle row). Similar results were obtained when we analyzed GC-T_FH (CXCR5⁺PD-1⁺) cells by CXCR5 and PD-1 counterstaining ( Figure 1B , bottom row). Thus, a strong positive correlation was established between Themis expression and T_FH cell differentiation ( Figure 1C ).

Although the above results suggested that Themis may be involved in T_FH cell formation, one caveat is that we analyzed bulk CD4⁺ T cells, not virus-specific CD4⁺ T cells. However, due to the extremely low precursor frequency of endogenous virus-specific CD4⁺ T cells (47), we were unable to obtain consistent results by specific MHC-II tetramer staining at 2–3 dpi, when these cells reportedly initiate T_FH cell differentiation (48–51). To circumvent this issue, we adopted another approach, transferring SMARTA TCR transgenic cells into B6 mice, followed by LCMV C13 infection. SMARTA cells are CD4⁺, MHC-II-restricted, TCR transgenic cells specifically recognizing LCMV glycoprotein epitope gp61–80 presented by H2-IA^b (52). We found that at 3 dpi, Themis expression in T_FH (CD44⁺CXCR5⁺) and non-T_FH (CD44⁺CXCR5⁻) cells in transferred SMARTA cells was 2.2- and 2.1-fold higher than that in host-derived CD4⁺ T cells, respectively ( Figure 1D ). Collectively, these results suggest that Themis is involved in the differentiation of CD4⁺ T cells, including T_FH cells, during the early stage of chronic viral infection.

To ascertain the effect of Themis in T_FH cell differentiation, we infected Themis T-cell conditional knockout mice (Themis^fl/fl; dLck mice, hereafter cKO mice) and wild-type control mice (Themis^fl/fl , hereafter WT mice) (37) with LCMV and examined their GC response. In these cKO mice, Themis was efficiently deleted in both CD8⁺ and CD4⁺ T cells, but a homeostasis defect was only reported in the CD8⁺ T cell compartment (37, 40). We confirmed this result ( Supplementary Figure S1A ) and further found that basal levels of CD4⁺ T cell activation were comparable in cKO and WT mice ( Supplementary Figure S1B ). Interestingly, in a competitive lymphopenia-induced homeostasis model ( Supplementary Figure S1C ), cKO CD4⁺ T cells exhibited a homeostatic proliferation defect at late but not early time points ( Supplementary Figures S1D, E ), but to a much lesser extent than that reported in CD8⁺ T cells (37, 40).

For LCMV C13 infection, at 8 dpi – the peak response – we found that the proportion and number of CXCR5⁺PD-1⁺ GC-T_FH cells in cKO mice was significantly reduced compared with those in WT mice ( Figures 2A, B ), and the generation of CD95⁺GL7⁺ GCB cells was also reduced ( Figures 2C, D ). Furthermore, although the proportion of CD138⁺IgD^– plasma cells was the same in WT and cKO mice, the number of these cells was decreased by 1.8-fold in cKO mice ( Figures 2E, F ). Consistent with the decrease in cell number in cKO mice, we found that the GC structure of cKO mice was very poorly delineated, with no clear shape, while the GC structure of WT mice was very clearly defined ( Figure 2G ).

In contrast to LCMV C13 infection, CD4⁺ T cells and antibodies are not required for viral control during acute LCMV infection, but GC responses still occur in this setting. We found that at 8 dpi, the generation of GC-T_FH cells was similar in WT and cKO mice ( Figures 2H, I ), as was the generation of GCB cells ( Figures 2J, K ). However, the number of plasma cells in cKO mice was significantly increased ( Figures 2L, M ), suggesting that GCB cell outflow may be enhanced in cKO mice. Accordingly, we found that the GC size of cKO mice was larger than that of WT mice ( Figure 2N ). Taken together, these results suggest that Themis deficiency has a more pronounced effect on T_FH cell differentiation and GC responses in chronic viral infection.

Although our above results suggest that Themis-deficient CD4⁺ T cells are defective in T_FH cell formation and in promoting GC responses at the peak of antiviral response, it is noteworthy that most cKO mice developed morbidity around 8 dpi, which may complicate our interpretation of the intrinsic role of Themis-deficient CD4⁺ T cells. To circumvent this, we generated SMARTA cells in cKO (SMtg cKO) and WT background (SMtg WT) and transferred equal numbers of both cell types into B6 recipient mice followed by LCMV C13 infection. Based on the aforementioned defects ( Figures 2A-G ), we expected that SMtg cKO cells might provide little help to B cells compared with SMtg WT cells. Contrary to that notion, we found that at 21 and 30 dpi, the abundance of anti-LCMV IgG antibody in mice receiving SMtg cKO cells were significantly higher than those in mice receiving SMtg WT cells ( Figure 3A ), indicating that the humoral immune response was enhanced in the mice receiving Themis-deficient SMtg T cells. Accordingly, we found that although the viral titers in the sera of mice receiving SMtg cKO or SMtg WT cells were comparable at 8 dpi and 21 dpi, the viral titers in mice receiving SMtg cKO cells were significantly reduced by 40 dpi compared with those in mice receiving SMtg WT cells, and the viral titers in some mice receiving SMtg cKO cells were even below the detection limit ( Figure 3B ).

The above findings prompted us to further analyze GC responses. For this, we chose 21 dpi because at this time point viral titers were comparable in both recipient mice, ruling out a potential effect of different viral load. Moreover, at this time point, the total number and percentage of both WT and cKO SMtg donor cells were roughly comparable ( Figure 3C ). Consistent with enhanced antibody production, we found that the frequencies of CD95⁺CD38^– GCB cells and IgD^–IgM^– isotype-switched B cells were significantly increased in mice receiving SMtg cKO cells, although the absolute numbers of these cells were the same as those in mice receiving SMtg WT cells ( Figures 3D, E ). The number and percentage of CD138⁺IgD^– plasma cells were also markedly increased in mice receiving SMtg cKO cells ( Figure 3F ). Finally, immunohistochemical staining revealed that GL7⁺ GC were enlarged in the spleens of mice receiving SMtg cKO cells compared with those receiving SMtg WT cells ( Figures 3G, H ). Taken together, these results suggested the Themis-deficient LCMV-specific CD4⁺ T cells can enhance the humoral immune responses after transfer into recipient mice.

Next, we wondered whether transferred SMtg cKO cells would enhance the function of endogenous CD8⁺ T cells, which normally undergo exhaustion and dysfunction during chronic LCMV infection. Again, we transferred equal amounts of naive SMtg cKO or SMtg WT cells into B6 mice, followed by LCMV C13 infection. We first examined the kinetics of endogenous total CD8⁺ T cells and LCMV-specific GP33–41 tetramer⁺ T cells in the blood. We found that, while the proportion of total CD8⁺ T cells was comparable in mice receiving SMtg WT and SMtg cKO cells at 7 and 14 dpi, it was slightly higher in mice receiving SMtg cKO cells at 21 dpi ( Figure 4A ). However, the proportion of LCMV-specific GP33–41 tetramer⁺ CD8⁺ T cells was the same at all time points ( Figure 4B ). Notably, we found that the proportions of SMtg cells from both donors were roughly equal and continued to decline synchronously during this period ( Figure 4C ). We next performed a more in-depth analysis at 21 dpi, as CD8⁺ T cells reach complete exhaustion at this stage. We found that while the proportion of total CD8⁺ T cells in the spleen of mice receiving SMtg cKO cells barely increased, their numbers doubled ( Figures 4D, E ). More importantly, the number of LCMV-specific GP33–41⁺ CD8⁺ T cells also doubled ( Figures 4F, G ).

In addition to cell number, we also examined whether the transferred SMtg cKO cells would influence the developmental status of exhausted CD8⁺ T cells. One model proposes that, during the early stage of infection, naive CD8⁺ T cells initially bifurcate into short-lived effector CD8⁺ T cells and stem cell-like T-pex cells. As infection persists, T-pex cells continue to self-renew, generating intermediate or transitionally exhausted CD8⁺ T cells and eventually differentiate into terminally differentiated exhausted CD8⁺ T cells at the late stage of infection (53–57). Consistent with this model, we found that at 21 dpi, effector CD8⁺ T cells (CX3CR1⁺), terminally exhausted CD8⁺ T cells (CD101⁺), and T-pex cells (TCF-1⁺) were clearly present in LCMV C13-infected mice ( Figure 4H , far left). As expected, in naive B6 mice, effector cells were few in number and exhausted cells were absent, whereas TCF-1⁺ cells were predominantly naive cells ( Figure 4H , far right). Furthermore, we observed a clear shift in CD8⁺ T cell subsets when SMtg cells were transferred. That is, the proportion of CX3CR1⁺ effector CD8⁺ T cells was significantly increased, while the proportion of terminally exhausted CD101⁺ CD8⁺ T cells was decreased ( Figure 4H , top row), but no changes were found in TCF-1⁺ T-pex cells ( Figure 4H , bottom row). However, no statistical difference was found between mice receiving SMtg cKO or SMtg WT cells ( Supplementary Figure S2A ). Taken together, these results suggest that, during chronic LCMV infection, transfer of Themis-deficient CD4⁺ cells boosts the total number of endogenous CD8⁺ T cells at the population level but does not affect their exhaustion status at the single-cell level. Nevertheless, the increase in the number of CD8⁺ T cells, especially LCMV-specific CD8⁺ T cells, together with the enhanced antibody response described above ( Figure 3A ), was sufficient to better control the virus ( Figure 3B ).

The benefits of Themis-deficient CD4⁺ T cells during chronic LCMV infection suggest that these cells may be used for cell therapy. However, the expansion of virus-specific CD4⁺ T cells has previously been shown to be associated with immunopathology, including cytokine storm, systemic inflammation, and death (58). To evaluate whether adoptively transferred Themis-deficient CD4⁺ T cells induce immunopathology, we analyzed various tissue sections from mice that received SMtg WT and SMtg cKO cells and found that at 8 dpi, the peak of the T cell response, no obvious damage was observed in the tissues examined ( Supplementary Figure S2B ). To formally test whether Themis-deficient virus-specific CD4⁺ T cells could be used to treat chronic LCMV infection, we injected these cells into previously infected mice and monitored viral load in the blood ( Supplementary Figure S2C , top row). We found that mice transferred with SMtg cKO cells showed a trend toward benefit, as more mice had viral loads below the limit of detection at later time points compared with mice transferred with SMtg WT cells ( Supplementary Figure S2C , bottom row). However, due to individual differences in mice, the statistical results were not significant.

T_FH cell differentiation is a multi-stage process, in which dendritic cells (DCs) are generally required for early T_FH cell differentiation (59), while B cells are required for further differentiation and completion of GC-T_FH cell maturation (60). Interestingly, our above results showed that, on the one hand, Themis deficiency impaired endogenous T_FH cell differentiation at the peak of chronic LCMV infection ( Figures 2A, B ), but on the other hand, transferred Themis-deficient CD4⁺ T cells enhanced GC responses and viral control at the late stage of infection ( Figures 3A, B ). These results suggested that Themis may play a stage-specific role in T_FH cell differentiation and exert opposing effects at different stages. To investigate this possibility and to exclude potential influences from environmental factors, we co-transferred equal numbers of naive SMtg WT and SMtg cKO cells with different congenic markers into B6 mice and then infected with LCMV C13 ( Figure 5A ). In addition, we advanced our analysis to 3 dpi for early T_FH cell differentiation, based on previous studies (50, 61) and our own results that Themis expression is upregulated in T_FH cells at this time point ( Figure 1D ). We first noticed a marked reduction in the proportion of SMtg cKO cells compared with SMtg WT cells ( Figure 5B ). To understand the reasons for this, we examined cell death in transferred SMtg cells. We found that the death rate of SMtg cKO cells was lower than that of SMtg WT cells as determined by Annexin V and 7-AAD staining ( Figure 5C ), ruling out cell death as a contributing factor to the reduction of SMtg cKO cells. Next, we examined their proliferation status by mixing equal numbers of naive SMtg WT and SMtg cKO cells (expressing different congenic markers), labeling them with CFSE, and then transferring them into B6 recipient mice. At 3 dpi, we found that SMtg WT cells exhibited more vigorous proliferation, while SMtg cKO cells displayed weaker proliferation. Specifically, CFSE profile revealed that SMtg cKO cells were predominantly in the third and fourth divisions, whereas WT cells had entered the fifth and sixth divisions ( Figure 5D ), suggesting that Themis deficiency impairs the proliferation of CD4⁺ T cells and is partially responsible for the reduction in cKO cell numbers.

We next analyzed the differentiation of T_FH cells and found that while T_FH lineages detected by CXCR5⁺ PD-1⁺, CXCR5⁺ ICOS⁺, or CXCR5⁺ Bcl-6⁺ all developed normally in SMtg WT cells, their proportions were significantly reduced in SMtg cKO cells ( Figure 5E ). In addition, we analyzed the expression of these markers and found that, except for Bcl-6, their expression in SMtg cKO cells was slightly decreased compared with those in SMtg WT cells ( Supplementary Figure S3A ). Previous studies have shown that CD4⁺ T cell differentiation choice occurs in the first few rounds of cell division, in which CXCR5 expression guides early T_FH cells to migrate to the border of B cell follicles and proceeds further T_FH cell differentiation (48, 49, 60). To investigate how early Themis begins to influence T_FH cell differentiation, we analyzed the relationship between T_FH cell differentiation and cell division by using the CFSE dilution profiles mentioned above. We found that the proportion of T_FH cells was reduced in SMtg cKO cells as early as division 3, and this trend continued until division 5 ( Figure 5F ; Supplementary Figures S3B, C ). Taken together, these findings suggest that Themis promotes virus-specific CD4⁺ T cell expansion and differentiation into the T_FH cell lineage early in infection.

Next, we investigated the effects of Themis on later stages of T_FH cell differentiation. The experimental setup was the same as before ( Figure 5A ), but analysis was performed at 21 dpi, as it was previously observed that transferred SMtg cKO cells enhanced GC response at this time point ( Figure 3 ). Notably, this time we observed an increase in the proportion of SMtg cKO cells compared with that of SMtg WT cells ( Figure 6A ). This ratio of donor cells was completely opposite to what we observed at 3 dpi ( Figure 6B ), confirming that Themis indeed has a stage-dependent effect. Remarkably, we not only observed an overall shift in donor cell contribution, but also found a greatly increased proportion of T_FH cells in SMtg cKO cells, whether they were defined by CXCR5⁺SLAM^–, CXCR5⁺PD-1⁺ or CXCR5⁺Bcl-6⁺ markers ( Figures 6C-E ). Recent studies have identified a population of TCF-1⁺Bcl-6^lo/– Tprog cells that emerge in the late stage of chronic viral infection that are capable of self-renewal and continuous generation of T_H1 effector cells and T_FH cells (31, 32). Based on this model, we suspected that Themis might influence this newly-defined Tprog cell population during chronic LCMV infection. To test this, we examined the dynamics of this population in mice transferred with SMtg cells and subsequently infected with LCMV C13. Consistent with previous reports, we observed that the proportion of TCF-1⁺ Bcl-6⁺ T_FH cells in both SMtg cell types was continuously reduced, whereas the proportion of TCF-1⁺Bcl-6^lo/– Tprog cells was initially stable but then greatly increased. The proportion of TCF-1^– Bcl-6^– T_H1 cells was transiently elevated and then decreased ( Figures 6F, G ). Although the dynamics of all the above cell subsets were comparable regardless of their genotype, there were quantitative differences. Specifically, at 21 dpi, the proportion of T_FH cells in SMtg cKO cells was 1.6-fold higher than that in SMtg WT cells, but the proportion of Tprog cells in SMtg cKO cells was 1.3fold lower than that in SMtg WT cells ( Figure 6G ). No difference was found for T_H1 cells at this stage between two types of SMtg cells ( Figure 6G ). Collectively, these results suggest that Themis plays a crucial role in shaping the long-term differentiation of CD4⁺ T cells during chronic LCMV infection by preserving the Tprog cell pool. In the absence of Themis, CD4⁺ T cells readily differentiate into T_FH cells, while progenitor cells are depleted, leading to enhanced GC responses and antibody production, as well as better protection against persistent viral infection.

Our above results demonstrated that Themis inhibits the differentiation of newly defined TCF-1⁺Bcl-6^lo/– Tprog cells in the late stage of chronic LCMV infection. To obtain an unbiased overview of how Themis affects this process, we sorted the co-transferred SMtg WT and SMtg cKO cells at 21 dpi and performed single-cell RNA-seq (scRNA-seq) ( Supplementary Figure S4A ). Following quality control filtering (see Methods), a total of 11,400 single-cell profiles were obtained, comprising 6,318 cells and 5,082 cells from the SMtg cKO and SMtg WT compartments respectively. Dimensionality reduction and clustering based on gene expression profiles revealed 9 distinct cellular clusters ( Figure 7A ), which were classified and characterized using differentially expressed genes and markers. The distribution of some key gene expressions across these clusters is presented ( Figure 7B , Supplementary Figure S4B ). As expected, the naive cell cluster highly expressed Ccr7, Tcf7, Slmaf6. The effector T cell (Teff) subpopulation was identified by the unique expression of Cxcr6, Id2, Tbx21. The T_FH cluster predominantly expressed high levels of Cxcr5, Bcl6, Pdcd1. Other subsets were also captured and classified in accordance with previous reports (32) ( Supplementary Figure S4B ).

We found that the cellular distribution in SMtg WT cells at 21 dpi was consistent with previous reports (32), confirming our analysis approach. Remarkably, in the SMtg cKO cells, the proportion of T_FH cluster doubled, increasing from approximately 5% to 10% ( Figure 7C ), a result that aligns with our previous phenotypic analysis ( Figures 6C-E ). To further compare T_FH-associated genes within the T_FH cluster between SMtg cKO and SMtg WT cells, we used violin plots to examine their expression. T_FH cluster from SMtg cKO group displayed significantly higher levels of Bcl6, Il21, Cxcr5, and Icos compared to SMtg WT ( Figure 7D ). We also analyzed Teff-related gene expression profiles within the Teff cluster, which showed distinct patterns between the two groups ( Supplementary Figure S4C ). Collectively, these results demonstrate that, during chronic LCMV infection, SMtg cKO cells exhibit an increased propensity to differentiate into T_FH commitment.

Trajectory inference with Slingshot 2 (see Methods) provided deeper insight into the developmental relationships between CD4⁺ T cell clusters, confirming that naïve CD4⁺ T cells transitioned through memory subsets before diverging into T_FH or Teff lineages. This is indeed consistent with previous studies showing that memory-like CD4⁺ T cells serve as progenitors that sustain Teff and T_FH cells during chronic LCMV infection (31, 32). However, a striking difference emerged in the density patterns between SMtg cKO and SMtg WT cells. That is, SMtg WT cells were more prevalent during the early stage of differentiation, whereas SMtg cKO cells exhibited accelerated progression toward terminal T_FH and Teff states ( Figure 7E ). The pseudotime analysis underscored the pivotal role of memory cells in the differentiation into T_FH and Teff lineages. Transcriptional regulators in resting-memory and activated-memory subsets from SMtg cKO and SMtg WT cells were further analyzed. Compared with SMtg WT cells, the expression of cytokine receptors Il6ra and Il6st, which are associated with T_FH cell differentiation, was increased in SMtg cKO memory subsets. In contrast, the expression of cytokine receptors Il2rb and Il2rg, which are more closely associated with Teff cell differentiation, was decreased in SMtg cKO memory subsets. In addition, transcription factors Bcl6 and Tox2, associated with T_FH lineage potential, were upregulated in SMtg cKO memory subsets, while Runx3, Klf2, Id2, Bhlhe40, and Tbx21, linked to Teff differentiation, were downregulated ( Figures 7F, G ). These data suggest that Themis deficiency depletes the Tprog pool to promote the differentiation of CD4⁺ T cells into T_FH cells during chronic LCMV infection. To investigate the molecular basis of these trajectory differences, key transcription factors along pseudotime paths were analyzed. Bcl6 and Tox2, key for T_FH cell differentiation, increased earlier and more prominently in SMtg cKO T_FH lineage, while remaining stable in SMtg WT. In contrast, Teff-associated factors Klf2 and Bhlhe40 were downregulated in SMtg cKO T_FH lineage. ( Supplementary Figure S4D ). These findings underscore a more pronounced and enhanced differentiation of CD4⁺ Tprog cells into T_FH cells in the absence of Themis.