Differential requirements for Tcf1 long isoforms in CD8⁺ and CD4⁺ T cell responses to acute viral infection

Abstract

In response to acute viral infections, activated naïve T cells give rise to effector T cells that clear the pathogen and memory T cells that persist long-term and provide heightened protection. T cell factor 1 (Tcf1) is essential for several of these differentiation processes. Tcf1 is expressed in multiple isoforms, with all isoforms sharing the same HDAC and DNA-binding domains, and the long isoforms containing a unique N-terminal β-catenin-interacting domain. Here we specifically ablated Tcf1 long isoforms in mice, while retaining expression of Tcf1 short isoforms. During CD8⁺ T cell responses, Tcf1 long isoforms were dispensable for generating cytotoxic CD8⁺ effector T cells and maintaining memory CD8⁺ T cell pool size, but contributed to optimal maturation of central memory CD8⁺ T cells and their optimal secondary expansion in a recall response. On the other hand, Tcf1 long isoforms were required for differentiation of T follicular helper (T_FH) cells but not T helper 1 (T_H1) effectors elicited by viral infection. Whereas Tcf1 short isoforms adequately supported Bcl6 and ICOS expression in T_FH cells, Tcf1 long isoforms remained important for suppressing the expression of Blimp1 and T_H1-associated genes and for positively regulating Id3 to restrain germinal center T_FH cell differentiation. Furthermore, formation of memory T_H1 and memory T_FH cells strongly depended on Tcf1 long isoforms. These data reveal that Tcf1 long and short isoforms have distinct yet complementary functions and may represent an evolutionarily conserved means to ensure proper programming of CD8⁺ and CD4⁺ T cell responses to viral infection.

In response to a viral infection, naïve T cells that recognize their cognate antigens become activated, expand prolifically, and differentiate into effector T cells equipped with diverse functions. Effector CD8⁺ T cells acquire cytotoxic functions and eliminate virus-infected cells (1, 2). On the other hand, activated CD4⁺ T cells predominantly differentiate into two types of effectors, T helper 1 (T_H1) cells that secrete interferon-γ (IFN-γ) and enhance the cytotoxicity of effector CD8⁺ T cells, or T follicular helper (T_FH) cells that secret IL-4 and IL-21 and provide essential help to antibody-producing B cells (3–5). Effector T cells are heterogeneous and contain subsets that have different kinetics of contraction following the peak responses, and hence different potential to give rise to memory T cells. Whereas memory CD8⁺ T cells are more durable than memory CD4⁺ T cells, both populations contribute to enhanced responses upon re-challenge with the same antigen.

Differentiation of effector T cells and their transition to memory T cells are coordinated by transcriptional regulators (6). In activated CD8⁺ T cells, T-bet and Blimp1 transcription factors as well as the Id2 cofactor are potently induced, and critically regulate CD8⁺ effector cell differentiation and acquisition of cytotoxic functions (7–9). On the other hand, Eomes, Bcl6, and Id3 promote the transition and survival of memory CD8⁺ T cells (9–11). In CD4⁺ T cells, T-bet and Bcl6 are the lineage-specifying master regulator for T_H1 and T_FH cells, respectively (3, 4), and induction of Blimp1 and Id2 favors T_H1 differentiation at the expense of T_FH lineage (12, 13). In contrast to advances in elucidating the transcriptional networks in CD4⁺ lineage differentiation at the effector phase, little is known about transcriptional regulation involved in memory CD4⁺ T cell formation and functions.

T cell factor 1 (Tcf1) has been known as a transcription factor acting downstream of the Wnt pathway and can interact with β-catenin coactivator. β-catenin is post-translationally regulated and stabilized by Wnt- or prostaglandin-derived signals. In addition to its essential role for T cell development (14, 15), recent studies have revealed that Tcf1 critically regulates mature T cell responses. Whereas loss of Tcf1 modestly diminished production of effector CD8⁺ T cells, Tcf1 is essential for maturation, longevity, and secondary expansion of memory CD8⁺ T cells (16, 17). In activated CD4⁺ T cells, Tcf1 appears to restrain T_H1 differentiation in vitro (18) but is essential for activating the T_FH program by acting upstream of Bcl6 (12, 19, 20). Due to differential promoter usage and alternative splicing, multiple Tcf1 isoforms can be detected in T cells (21). All isoforms contain a C-terminal HMG DNA binding domain, which can also interact with Groucho/Transducin-like Enhancer-of-split (TLEs) corepressor proteins, and a newly discovered HDAC domain (22). The Tcf1 long isoforms (p45 and p42) contain an N-terminal β-catenin-binding domain, while the Tcf1 short isoforms (p33 and p30) lack this domain and hence cannot interact with β-catenin. Most of the previous loss-of-function studies ablated all Tcf1 isoforms. The specific requirements for the Tcf1 long isoforms in effector and memory T cell responses have not yet been elucidated.

Here we specifically ablated Tcf1 long isoforms in mouse, and coupled this model with MHC-I- and MHC-II-restricted T cell receptor (TCR) transgenes to dissect the roles of Tcf1 isoforms in regulating mature CD8⁺ and CD4⁺ T cell responses. Our data showed that Tcf1 long isoforms were dispensable for generation of cytotoxic CD8⁺ effector T cells and helper T_H1 cells in response to viral infection, while Tcf1 short isoforms were sufficient for maintaining memory CD8⁺ T cell pool size. On the other hand, Tcf1 long isoforms remained critical for T_FH differentiation at the effector phase and for generation of both memory T_H1 and memory T_FH cells. This study reveals a functional complementation among Tcf1 isoforms in programming mature T cell responses, and further suggests an inter-isoform coordination is necessary to constitute a fully functional gene product.

Materials and Methods

Mice

p45^+/GFP (i.e., p45^+/−) mice were generated as previously described (23), and cross-bred with P14 or SMARTA TCR transgenes to generate P14 p45^−/− or SMARTA p45^−/− mice, respectively. Our recent analysis of thymocyte development in the p45-targeted mice showed no detectable differences between p45^+/+ and p45^+/− animals (23), suggesting that heterozygosity of the p45-targeted allele does not significantly impact T cell biology. As such, we used T cells from p45^+/− mice as controls in this study, and the Tcf1-EGFP reporter embedded in the p45^+/− allele was utilized to mark Tcf1 expression in CD4⁺ and CD8⁺ T cells at different stages of responses to viral infection. C57BL/6J and B6.SJL mice were from the Jackson Laboratory. All mice analyzed were 6–12 weeks of age, and both genders are used without randomization or blinding. All mouse experiments were performed under protocols approved by the Institutional Animal Use and Care Committees of the University of Iowa.

Flow cytometry

Single-cell suspensions were prepared from the spleen, lymph nodes (LNs), or peripheral blood lymphocytes (PBLs), and surface or intracellularly stained as described (19). The fluorochrome-conjugated antibodies were as follows: anti-CD8 (53-6.7), anti-CD4 (RM4-5), anti-CD44 (IM7), anti-CD62L (MEL-14), anti-PD-1 (J43), anti-CD45.1 (A20), anti-CD45.2 (104), anti-ICOS (C398.4A), anti-IFN-γ (XMG1.2), anti-TNFα (MP6-XT22), anti-Eomes (Dan11mag), anti-T-bet (eBio4B10), anti-IL-2 (JES6-5H4), anti-IL-7Rα (eBio17B7) and anti-KLRG1 (2F1) from eBiosciences; anti-Bcl6 (K112-91) from BD Biosciences; anti-human granzyme B (FGB12) and corresponding isotype control from Invitrogen/Life Technologies, anti-SLAM (TC15-12F12.2) from BioLegend. For detection of CXCR5, three-step staining protocol was used with unconjugated anti-CXCR5 (2G8; BD Biosciences) (24). For detection of Bcl6, surface-stained cells were fixed and permeabilized with the Foxp3/Transcription Factor Staining Buffer Set (eBiosciences), followed by incubation with a fluorochrome-conjugated antibody or isotype control. Peptide-stimulated cytokine production and detection by intracellular staining were as described (25). Data were collected on a FACSVerse (BD Biosciences) and were analyzed with FlowJo software (TreeStar).

Adoptive transfer and viral infection

Naïve CD45.2⁺ P14 CD8⁺ T cells or SMARTA CD4⁺ T cells were isolated from the LNs of P14- or SMARTA-transgenic p45^+/− or p45^−/− mice. For characterization of CD8⁺ T cell responses, 2 × 10⁴ Vα2⁺ P14 CD8⁺ T cells were intravenously (i.v.) injected into CD45.1⁺ B6.SJL recipient mice and infected intraperitoneally (i.p) with 2 × 10⁵ PFU of lymphocytic choriomeningitis virus, Armstrong strain (LCMV-Arm). For characterization of CD4⁺ T cell responses, 2 × 10⁵ Vα2⁺ SMARTA CD4⁺ T cells were i.v. injected into CD45.1⁺ hosts and infected similarly. For examining CD4⁺ T cell responses under a competitive condition, 5,000 p45^+/− or p45^−/− SMARTA CD4⁺ T cells were mixed with the same number of WT CD45.1⁺CD45.2⁺ SMARTA CD4⁺ T cells and adoptively transferred. At ≥ 40 dpi, the immune mice were i.v. infected with 2 × 10⁶ PFU of LCMV clone 13 (Cl13) to elicit secondary CD8⁺ or CD4⁺ T cell responses.

Immunoblotting

CD44^loCD62L⁺ naïve P14 CD8⁺ T cells or SMARTA CD4⁺ T cells were sort-purified from the spleens of P14- or SMARTA-transgenic p45^+/− or p45^−/− mice. Cell lysates were prepared from the sorted cells, resolved on SDS-PAGE, and immunoblotted with an anti-Tcf1 antibody (C46C7, Cell Signaling Technologies) as previously described (23).

RNA-Seq and quantitative RT-PCR

p45^+/− or p45^−/− SMARTA CD4⁺ T cells were adoptively transferred and recipients infected as described above. On 8 dpi, PD-1^lo CXCR5⁺ cells were sort-purified from CD45.2⁺CD4⁺ splenocytes. The total RNA was extracted and subjected to RNA-Seq analysis as previously described (19). The RNA-Seq data were deposited at the Gene Expression Omnibus (GEO) under the accession number GSE98347 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=gse98347). For validation of key gene expression changes, the RNA was reverse-transcribed and analyzed by quantitative PCR as previously described (19). The primers used for Id2 were 5′-catcagcatcctgtccttgc and 5′-gtgttctcctggtgaaatgg, and those for Id3 were 5′-atctcccgatccagacagc and 5′-gagagagggtcccagagtcc.

Chromatin immunoprecipitation (ChIP)

WT SMARTA CD4⁺ T cells were adoptively transferred and recipients infected as above. On 8 dpi, CXCR5⁻SLAM^hi T_H1 and CXCR5⁺SLAM^lo T_FH cells were sort-purified from CD45.2⁺CD4⁺ splenocytes and cross-linked, and the resulting chromatin fragments were immunoprecipitated with an anti-Tcf1 antibody (C46C7) or normal rabbit IgG, as previously described (19). Enriched Tcf1 binding at the Id3 transcription start site was determined by quantitative PCR.

Statistical analyses

Data from multiple experiments were analyzed with Student’s t-test with two-tailed distribution, assuming equal sample variance when p>0.05 on F-test. When p ≤ 0.05 on F-test comparing the variances, Welch’s correction was applied.

Results

Tcf1 long isoforms are dispensable for CD8⁺ effector T cell differentiation

Tcf1 proteins are encoded by the Tcf7 gene, with long isoforms produced from transcripts initiated at exon 1 and short isoforms from transcripts initiated at exon 3. Previously we have generated a Tcf1-EGFP reporter mouse strain, where we inserted an IRES-EGFP cassette into the first intron of Tcf7 gene (26). In front of the cassette, we placed an En2 splice acceptor that forced the splicing of exon 1 to this cassette instead of exon 2. Mice homozygous for the Tcf1-EGFP reporter alleles thus failed to produce Tcf1 long isoforms, p45 and p42; for simplicity, this strain is called p45^−/− herein. To assess the importance of Tcf1 long isoforms (p45 and p42) in CD8⁺ T cell responses to viral infection, we crossed p45^−/− mice with the P14 TCR transgene, which encodes a TCR specific for the glycoprotein (GP) 33 epitope of LCMV. As expected, p45^−/− P14 CD8⁺ T cells were deficient for Tcf1 long isoforms but retained the expression of Tcf1 short isoforms, albeit at a reduced amount (Fig. 1A). Loss of Tcf1 long isoforms moderately reduced the frequency of P14 CD8⁺ T cells in the LNs, but did not result in aberrant activation of peripheral CD8⁺ T cells, as determined by the CD62L^hi phenotype (Fig. 1B).

Figure 1. Tcf1 long isoforms are dispensable for effector CD8⁺ T cell responses.

(A) Detection of Tcf1 isoforms by immunoblotting in naïve CD8⁺ T cells. CD62L^hiCD44^loVα2⁺CD8⁺ cells were sorted from the LNs of p45^+/− or p45^−/− P14 transgenic mice and immunoblotted with an anti-Tcf1 or β-actin antibody. (B) Characterization of naïve P14 CD8⁺ T cells. LNs from uninfected p45^+/− or p45^−/− P14 transgenic mice were surface-stained to detect the frequency (left) and immuno-phenotype (right) of Vα2⁺CD8⁺ T cells (n = 4 from 4 experiments). (C) Kinetics of CD8⁺ T cell responses in PBLs. CD45.2⁺ p45^+/− or p45^−/− P14 CD8⁺ T cells (2×10⁴ each) were adoptively transferred into CD45.1⁺ recipients followed by infection with LCMV-Arm. The frequency of CD45.2⁺ CD8⁺ effector T cells in CD11a^hiCD8^dim antigen-experienced CD8⁺ T cells were monitored in the PBLs on indicated days post-infection (n = 10–20 from 3 experiments). (D) Numbers of P14 CD8⁺ effector T cells detected in the spleen on 8 dpi (n = 8 from 3 experiments). (E) Detection of Tcf1-EGFP reporter in p45^+/− P14 CD8⁺ effector subsets, and the geometric mean fluorescent intensity (gMFI) of EGFP is marked. Dotted line denotes EGFP expression in naïve CD8⁺ T cells. (F) Detection of P14 CD8⁺ effector subsets. CD45.2⁺ CD8⁺ effector T cells in the spleen (8 dpi) were analyzed for KLRG1 and CD127 expression, and the percentages of KLRG1^hiCD127⁻ and KLRG1^loCD127⁺ subsets are shown in representative contour plots. Cumulative data on the frequency and numbers of the KLRG1^loCD127⁺ subset are summarized in bar graphs (n = 10 from 3 experiments). (G) Cytokine production by P14 CD8⁺ effector T cells. Splenocytes (8 dpi) were incubated with GP33 peptide for 5 hours, and CD45.2⁺CD8⁺ T cells were intracellularly stained to detect IFN-γ and TNF-α. The percentage of each population is shown in representative contour plots from 3 experiments. (H) Granzyme B expression in P14 CD8⁺ effector T cells detected by intracellular staining. gMFI of Granzyme B is shown in representative histogram. Dotted line, isotype control staining. Data in bar graphs are means ± SEM. *, p < 0.05 by Student’s t-test unless specified otherwise.

To facilitate functional characterization on a per cell basis, we adoptively transferred the same numbers (2×10⁴) of p45^+/− or p45^−/− P14 CD8⁺ T cells into congenic recipients followed by infection with LCMV-Arm. As tracked in PBLs, both p45^+/− and p45^−/− CD45.2⁺ CD8⁺ effector T cells showed similar frequency among antigen-experienced CD11a^hiCD8^int T cells (Fig. 1C). In the spleens, similar numbers of p45^+/− and p45^−/− CD45.2⁺CD8⁺ effector T cells were generated at the peak response, i.e., 8 days post-infection (dpi) (Fig. 1D). By comparison, loss of all Tcf1 isoforms diminishes CD8⁺ effector T cells by approximately 50% in both PBLs and spleens (16). CD8⁺ effector T cells are heterogeneous, with KLRG1^hiCD127⁻ cells as terminally differentiated effector CD8⁺ T cells, and the KLRG1^loCD127⁺ subset showing increased potential to give rise to memory CD8⁺ T cells (6). As observed in p45^+/− CD8⁺ T cells, the Tcf1-EGFP reporter expression is potently diminished in KLRG1^hiCD127⁻ cells, but partially retained in the KLRG1^loCD127⁺ subset (Fig. 1E). This is consistent with our previous observations that Tcf1 and Lef1 are required for generation of the KLRG1^loCD127⁺ subset (27). However, the KLRG1^loCD127⁺ subset in p45^−/− CD8⁺ effectors was only modestly reduced in frequency and numbers (Fig. 1F). Both p45^+/− and p45^−/− CD8⁺ effectors exhibited similar capacity to produce IFN-γ, TNF-α, and granzyme B (Fig. 1G–1H), and by comparison, these functional aspects are not affected by loss of all Tcf1 isoforms as shown in our previous study (16). Collectively, these data indicate that Tcf1 long isoforms are dispensable for clonal expansion and cytotoxic functions of CD8⁺ effector T cells.

Tcf1 long isoforms contribute to maturation of central memory CD8⁺ T cells

Following the peak response to LCMV infection, p45^+/− and p45^−/− CD45.2⁺CD8⁺ effector T cells contracted similarly (Fig. 1C), and at ≥ 60 dpi, similar numbers of p45^+/− and p45^−/− memory CD8⁺ T cells were found in the recipient spleens (Fig. 2A). Memory CD8⁺ T cells in the secondary lymphoid organs are heterogeneous, with the CD62L⁻ subset as effector memory (T_EM) cells that reside in non-lymphoid tissues, and the CD62L⁺ subset as central memory (T_CM) cells that are capable of self-renewal and conferring long-term protection (28). Compared with CD8⁺ effector T cells, most of the memory CD8⁺ T cells upregulated Tcf1, and the CD62L⁺ subset expressed higher levels of Tcf1 as determined by the Tcf1-EGFP reporter activity (Fig. 2B, compare with Fig. 1E). Our previous studies showed that deletion of all Tcf1 isoforms almost completely abrogated generation of CD62L⁺ T_CM cells and caused substantial decrease in Eomes expression (16). In p45^−/− memory CD8⁺ T cell pool, CD62L⁺ cells were generated, but at diminished frequency and numbers (Fig. 2C). In addition, p45^−/− memory CD8⁺ T cells showed modest reduction in Eomes expression, accompanied by a modest increase in T-bet expression (Fig. 2D). Regarding functional aspects, fewer p45^−/− memory CD8⁺ T cells (in both CD62L⁻ and CD62L⁺ subsets) were capable of producing IL-2 than p45^+/− cells (Fig. 2E), whereas they showed similar production of IFN-γ and TNF-α (data not shown). Additionally, p45^−/− memory CD8⁺ T cells showed modest increase in granzyme B compared with p45^+/− cells (Fig. 2F). These observations indicate that loss of Tcf1 long isoforms does not affect the pool size of memory CD8⁺ T cells but predisposes them to a T_EM phenotype.

Figure 2. Tcf1 long isoforms contribute to optimal T_CM maturation and secondary expansion.

(A) Numbers of memory P14 CD8⁺ T cells detected in the spleen at ≥ 60 dpi (n = 5–6 from 2 experiments). (B) Detection of Tcf1-EGFP reporter in CD62L⁻ and CD62L⁺ memory CD8⁺ T cell subsets. CD45.2⁺ memory CD8⁺ T cells from p45^+/− mice were analyzed for EGFP expression, and values in the histogram denote EGFP gMFI. (C) Detection of P14 CD8⁺ T_CM cells. CD45.2⁺ memory CD8⁺ T cells in the spleen were analyzed for CD62L expression, and the percentages of CD62L⁺ T_CM cells are shown in representative histograms. Cumulative data on the frequency and numbers of the CD62L⁺ T_CM cells are summarized in bar graphs (n = 5–6 from 2 experiments). (D) T-bet and Eomes expression in memory P14 CD8⁺ T cells detected by intranuclear staining. gMFI of T-bet or Eomes is shown in representative histograms, and cumulative data are in bar graphs (for T-bet, n = 2–3 from 1 experiment; for Eomes, n = 5–6 from 2 experiments). Dotted line, isotype control staining. (E) Cytokine production by memory P14 CD8⁺ T cells. Splenocytes were harvested at ≥ 60 dpi and incubated with GP33 peptide, followed by intracellular detection of IFN-γ and IL-2 in CD45.2⁺CD8⁺ T cells. The percentage of each population is shown in representative contour plots (n = 8 from 3 experiments), and that of IFN-γ– and IL-2-double producers are summarized in bar graphs. (F) Granzyme B expression in memory P14 CD8⁺ T cells detected by intracellular staining. gMFI of Granzyme B is shown in representative histogram, and cumulative data are in bar graphs (n = 8 from 3 experiments). Dotted line, isotype control staining. (G) Numbers of secondary CD8⁺ effector T cells. The immune CD45.1⁺ recipients were infected with LCMV-Cl13, and five days after infection, the numbers of CD45.2⁺CD8⁺ secondary effector T cells were determined in the spleen (n = 5 from 2 experiments). (H) Characterization of P14 CD8⁺ secondary effector T cells. CD45.2⁺CD8⁺ secondary effector T cells in the spleen were surface-stained, and the percentages of KLRG1^hiCD127⁻ and KLRG1^loCD127⁺ subsets are shown in representative contour plots (left panels). Both subsets were analyzed for Tcf1-EGFP reporter expression in p45^+/− recipients, with EGFP gMFI marked (right panel, n = 5–6 from 2 experiments). Data in bar graphs are means ± SEM. *, p < 0.05; **, p<0.01; ***, p < 0.001 by Student’s t-test unless specified otherwise.

To assess the recall response by memory CD8⁺ T cells, we re-challenged the immune recipient mice with LCMV-Cl13. As detected on day 5 post-infection, both p45^+/− and p45^−/− memory CD8⁺ T cells underwent substantial secondary expansion, although the numbers of p45^−/− secondary CD8⁺ effector T cells in the spleens were modestly decreased (Fig. 2G), which is in contrast to severely impaired secondary expansion in the absence of all Tcf1 isoforms (16). Similar to CD8⁺ effector T cells in the primary response, the secondary CD8⁺ effector T cells were heterogeneous, with the KLRG1^loCD127⁺ subset retaining substantially higher Tcf1-EGFP reporter activity than the KLRG1^hiCD127⁻ subset (Fig. 2H). Loss of Tcf1 long isoforms did not detectably perturb the subset composition or IFN-γ and TNF-α production in the secondary effectors (Fig. 2H and data not shown). Collectively, these data suggest that Tcf1 long isoforms contribute to optimal maturation of T_CM cells and optimal secondary expansion in a recall response. It is also noteworthy that reduced expression of Tcf1 short isoforms in p45^−/− CD8⁺ T cells may also be a contributing factor.

Tcf1 long isoforms are required for T_FH differentiation at the effector phase of CD4⁺ T cell responses

We and others have recently demonstrated that deletion of all Tcf1 isoforms impairs T_FH differentiation in response to viral infections (12, 19, 20). To determine the requirements for Tcf1 long isoforms during CD4⁺ T cell responses, we crossed the p45^−/− strain with SMARTA TCR transgene, which encodes a TCR specific for the GP61 epitope of LCMV. By immunoblotting, we validated that p45^−/− SMARTA CD4⁺ T cells were deficient for Tcf1 long isoforms but retained the expression of Tcf1 short isoforms, also at a reduced amount (Fig. 3A). Although the frequency of SMARTA CD4⁺ T cells in the LNs was diminished in p45^−/− mice, the majority of p45^−/− SMARTA CD4⁺ T cells exhibited a CD44^loCD62L^hi phenotype, suggesting that loss of Tcf1 long isoforms did not result in aberrant activation of peripheral CD4⁺ T cells (Fig. 3B).

Figure 3. Tcf1 long isoforms regulate optimal T_FH differentiation.

(A) Detection of Tcf1 isoforms by immunoblotting in CD4⁺ T cells. CD62L^hiCD44^loVα2⁺CD4⁺ cells were sorted from the LNs of p45^+/− or p45^−/− SMARTA-Tg mice and immunoblotted as in Figure 1A. (B) Characterization of naïve SMARTA CD4⁺ T cells. LNs were isolated from p45^+/− or p45^−/− SMARTA-Tg mice and surface-stained to detect the frequency (left) and immunophenotype (right) of Vα2⁺CD4⁺ T cells (n = 8 from 8 experiments). (C) Detection of effector SMARTA CD4⁺ T cells at the peak response. CD45.2⁺ p45^+/− or p45^−/− SMARTA CD4⁺ T cells from LN (2×10⁵ each) were adoptively transferred into CD45.1⁺ recipients followed by infection with LCMV-Arm. On 8 dpi, CD45.2⁺CD4⁺ T cells were detected and enumerated in the spleens of recipient mice (n = 8 from 3 experiments). (D) Detection of T_H1 and T_FH effector CD4⁺ T cells. CD45.2⁺ effector CD4⁺ T cells (as in C) were analyzed for CXCR5 and SLAM expression. The percentages of CXCR5⁻SLAM^hi T_H1 and CXCR5⁺SLAM^lo T_FH cells are shown in representative contour plots (left panels). Cumulative data on the frequency and numbers of each subset are summarized in bar graphs (n = 5–8 from 2–3 experiments). Data are means ± SEM. **, p < 0.01; ***, p < 0.001 by Student’s t-test unless specified otherwise.

To investigate the regulatory roles of Tcf1 long isoforms on a per cell basis, we adoptively transferred the same numbers (2×10⁵) of p45^+/− or p45^−/− SMARTA CD4⁺ T cells from LNs into congenic recipients followed by infection with LCMV-Arm. At the peak of effector CD4⁺ T cell responses (i.e., 8 dpi), p45^+/− and p45^−/− SMARTA CD4⁺ T cells expanded to similar levels, showing similar frequency and numbers of effector CD4⁺ T cells in the spleen of recipient mice (Fig. 3C). Viral infection elicits two major forms of effector CD4⁺ T cells, T_H1 and T_FH cells, which exhibit CXCR5⁻SLAM^hi and CXCR5⁺SLAM^lo phenotypes, respectively. p45^+/− effector CD4⁺ T cells contained approximately equal amounts of T_H1 and T_FH cells; in contrast, p45^−/− effector CD4⁺ T cells were strongly skewed toward the T_H1 phenotype (Fig. 3D). As a result, p45^−/− SMARTA CD4⁺ T cells generated markedly fewer T_FH cells than the p45^+/− counterparts (Fig. 3D), similar to T_FH differentiation defects observed in the absence of all Tcf1 isoforms (12, 19, 20). On the other hand, both p45^−/− and p45^+/− SMARTA CD4⁺ T cells gave rise to similar amounts of effector T_H1 cells (Fig. 3D). These data suggest that Tcf1 long isoforms are important in directing effective T_FH differentiation, but are dispensable for T_H1 responses.

In the experiments above, we transferred a relatively large number of SMARTA CD4⁺ T cells so as to obtain sufficient cells for RNA-Seq and memory phase analysis (see below). To exclude a secondary impact derived from a higher precursor frequency and demonstrate an intrinsic requirement for Tcf1 long isoforms, we transferred a small number of p45^+/− or p45^−/− SMARTA CD4⁺ T cells in a competitive setting. Specifically, we mixed CD45.2⁺ p45^+/− or p45^−/− SMARTA CD4⁺ T cells (as test cells) with the same number of CD45.1⁺CD45.2⁺ wild-type (WT) SMARTA CD4⁺ T cells (as competitors) (Fig. 4A). We then adoptively transferred a total of 10,000 SMARTA CD4⁺ T cells (5,000 each of test cells and competitors) into CD45.1⁺ recipients, followed by LCMV-Arm infection. On 8 dpi, within the T_H1 compartment, the p45^+/−:WT ratio was largely maintained at 1:1, and the p45^−/−:WT ratio was modestly decreased (Fig. 4B–4C). Within the T_FH cells, however, the p45^−/−:WT ratio was reduced by >5 fold compared with the p45^+/−:WT ratio (Fig. 4B–4C). Thus, analysis under the competitive setting at a lower precursor frequency further corroborated a specific, intrinsic requirement for Tcf1 long isoforms in T_FH differentiation.

Figure 4. CD4⁺ T cells lacking Tcf1 long isoforms are less competitive in T_FH differentiation.

(A) Relative abundance of CD45.2⁺ test cells (p45^+/− or p45^−/−) and CD45.1⁺CD45.2⁺ competitors (WT CD4⁺ T cells) in the cell mixture used in competitive transfer experiments. (B) Detection of relative contribution of test and competitor SMARTA CD4⁺ T cells to T_H1 and T_FH responses. CXCR5⁻SLAM^hi T_H1 cells and CXCR5⁺SLAM^lo T_FH cells were detected in the recipients on 8 dpi (left). Within each subset, the percentages of test and competitor cells were determined and shown in representative contour plots. (C) Cumulative data of test/competitor ratio within T_H1 or T_FH subsets at the effector phase (8 dpi). Data are from 2 experiments (n = 5–6). Data are means ± SEM. *, p < 0.05; ***, p < 0.001 by Student’s t-test.

Tcf1 long isoforms regulate a unique subset of genes in the Tfh program

We previously showed that Tcf1 critically regulates activation of the T_FH program in primed CD4⁺ T cells, and deletion of all Tcf1 isoforms diminishes the expression of Bcl6 and ICOS (19). However, p45^+/− and p45^−/− T_FH cells showed similar Bcl6 and ICOS expression (Fig. 5A–5B), suggesting that Tcf1 short isoforms are sufficient for supporting Bcl6 and ICOS induction, in spite of reduced protein expression. T_FH differentiation is known as a step-wise process (4), with T_FH lineage-committed cells undergoing further maturation and becoming germinal center (GC)-T_FH cells, as marked by high PD-1 expression. Although there was an overall reduction of p45^−/− T_FH cells, PD-1^hiCXCR5⁺ GC-T_FH cells were detected at similar frequency and numbers in p45^+/− and p45^−/− effector SMARTA CD4⁺ T cells (Fig. 5C). Thus, loss of Tcf1 long isoforms showed stronger impact on non-GC-T_FH cells, causing subset composition changes within the T_FH population. By comparison, both non-GC T_FH and GC T_FH cells were diminished in the absence of all T cell isoforms (19). Therefore, Tcf1 short isoforms appears to be adequate for promoting GC-T_FH maturation.

Figure 5. Tcf1 long isoforms are essential for regulating a subset of T_FH transcriptional program.

(A–B) Detection of Bcl6 (A) and ICOS (B) in p45^+/− and p45^−/− CXCR5⁺SLAM^lo T_FH cells (8 dpi). gMFI of each protein is shown in representative histograms, and cumulative data on the relative expression between p45^+/− and p45^−/− T_FH cells are in bar graphs (n = 11 from 4 experiments). (C) Detection of PD-1^hiCXCR5⁺ GC-T_FH cells (8 dpi). The percentages of GC-T_FH cells among CD45.2⁺ SMARTA CD4⁺ T cells are shown in representative contour plots, and cumulative data in bar graph are from 2 experiment (n = 6). (D–F) Heatmaps of select differentially expressed genes. p45^+/− and p45^−/− PD-1^loCXCR5⁺ non-GC T_FH cells were sorted from splenic CD45.2⁺ SMARTA CD4⁺ T cells in the recipients on 8 dpi. Two biological replicates were collected for each genotype and analyzed with RNA-Seq. (G) Detection of select gene expression in non-GC T_FH cells by quantitative RT-PCR. The cells were sorted using the same strategy as above. For each gene of interest, its expression in p45^+/− cells was set as 1, and its relative expression in p45^−/− cells was calculated accordingly. Data are from two independent experiments (n = 3–4). (H) Detection of enriched Tcf1 binding. WT T_H1 and T_FH cells were sorted and subjected to ChIP analysis. The relative enrichment at the control Hprt locus and Id3 transcription start site by an anti-Tcf1 antibody was determined by quantitative PCR. Data are from two independent experiments with each sample measured in duplicates or triplicates. Data in bar graphs are means ± SEM. *, p < 0.05; **, p < 0.01 by Student’s t-test unless specified otherwise.

To search for downstream genes that account for defective p45^−/− T_FH differentiation, we specifically sorted p45^+/− and p45^−/− PD-1^lo CXCR5⁺, non-GC-T_FH cells on 8 dpi and performed RNA-Seq analysis. By a setting of ≥ 2 fold expression changes and false discovery rate (FDR) q value <0.05, we found 89 downregulated and 119 upregulated genes in p45^−/− T_FH cells compared with p45^+/− counterparts (Supplemental Table 1). Transcriptomic changes have been profiled in T_FH cells lacking all Tcf1 isoforms or both Tcf1 and Lef1 proteins (19, 20). Whereas differences in specific cell subsets analyzed and profiling platforms utilized among these studies prevented a meaningful and fair comparison for specific genes affected, the number of differentially expressed genes was substantially lower in p45^−/− T_FH cells than that found in Tcf1/Lef1-deficient T_FH cells, suggesting a relatively limited impact of deficiency in Tcf1 long isoforms on T_FH transcriptome. Functional annotation using the DAVID Bioinformatics Resources revealed a partial activation of T_H1 program in p45^−/− T_FH cells, including upregulation of Prdm1, Il2ra, and Ifng (encoding Blimp1 transcription factor, CD25, and IFN-γ, respectively) (Fig. 5D). Increased expression of Prdm1 was validated by quantitative RT-PCR (Fig. 5G). Blimp1 is known to promote activated CD4⁺ T cells to a T_H1 fate at the expense of T_FH differentiation (29), and Tcf1 and Bcl6 co-occupy an intron 3 regulatory region in the Prdm1 gene locus (19, 30). Our data suggest that Tcf1 long isoforms are critical in repressing the T_H1 potential in T_FH cells.

Among the upregulated genes included Maf, Fos and Fosb (Fig. 5E). Maf is necessary for promoting T_FH differentiation and inducing IL-21 (31, 32). AP-1 factors including Fos and FosB may contribute to recruiting Bcl6 to its target genes in T_FH cells (30). Among the downregulated genes, a prominent subset was involved in transcriptional regulation, including Irf6, Nfe2, Sox4, and Id3 (Fig. 5F). Importantly, the first three genes were detected at low abundance in T_FH cells (< 5 FPKMs); in contrast, Id3 was highly abundant in T_FH cells (132 PFKMs on average). As validated in Fig. 5G, Id3 expression was diminished in p45^−/− T_FH cells, whereas Id2 was not detectably altered. By chromatin immunoprecipitation, we further found that Tcf1 exhibited enriched binding to the transcription start site of Id3 gene in T_FH cells but not T_H1 cells (Fig. 5H), establishing Id3 as a direct target gene of Tcf1 in T_FH cells. It has been recently shown that Id3 deficiency results in evident accumulation of GC-T_FH cells (13). These data collectively suggest that another important function of Tcf1 long isoforms appears to be restraining GC-T_FH differentiation by positive regulation of Id3, maybe aided by suppression of Maf and AP-1 factors.

Tcf1 long isoforms are essential for formation of memory T_H1 and T_FH cells

We continued to track the SMARTA CD4⁺ T cell responses at the memory phase, i.e., > 40 dpi, and found significantly fewer p45^−/− memory CD4⁺ T cells than the p45^+/− counterparts (Fig. 6A). Compared with CD4⁺ effector T cells, the number of p45^+/− memory CD4⁺ T cells was a result of approximately 20-fold contraction; in contrast, the number of p45^−/− memory CD4⁺ T cells represented an almost 200-fold contraction of corresponding CD4⁺ effector T cells (Fig. 6B). Phenotypic analysis using SLAM and CXCR5 expression further revealed that p45^−/− memory CD4⁺ T cells were strongly skewed toward T_H1, with very few SLAM^loCXCR5⁺ memory T_FH cells detected (Fig. 6C). In spite of the lineage bias in the absence of Tcf1 long isoforms, p45^−/− memory T_H1 and T_FH cells both showed more potent contraction than their the p45^+/− counterparts (Fig 6C). As detected by the Tcf1-GFP reporter in p45^+/− CD4⁺ T cells, although Tcf1 was downregulated in T_H1 effectors, it was upregulated in memory T_H1 cells, whereas Tcf1-GFP reporter was detected at high levels in both effector and memory T_FH cells (Fig. 6D). We next assessed the recall response by memory CD4⁺ T cells, by re-challenging the immune recipient mice with LCMV-Cl13. As detected on day 5 post-infection, p45^−/− secondary CD4⁺ effectors were greatly diminished in numbers and remained strongly skewed toward T_H1 cells, containing very few T_FH cells (Fig. 6E). These data suggest that Tcf1 long isoforms are essential for generation of memory T_H1 as well as memory T_FH cells and ensuing secondary responses. It should be noted that reduced expression of Tcf1 short isoforms in p45^−/− CD4⁺ T cells may also have contributed to defects in memory CD4⁺ T cells.

Figure 6. Tcf1 long isoforms are critical for formation of memory T_H1 and memory T_FH cells.

(A) Detection of memory SMARTA CD4⁺ T cells. Adoptive transfer and infection were performed as in Figure 3. On >40 dpi, CD45.2⁺CD4⁺ T cells were detected and enumerated in the spleens of recipient mice (n = 7–8 from 3 experiments). (B) Population fold changes due to contraction. p45^+/− and p45^−/− SMARTA CD4⁺ T cells were enumerated at the peak (8 dpi) and memory (> 40 dpi) phases in response to viral infection, and fold reduction between the two phases is calculated (n = 8 from 3 experiments). (C) Detection of memory T_H1 and T_FH T cells. CD45.2⁺CD4⁺ T cells were analyzed for CXCR5 and SLAM expression. The percentages of CXCR5⁻SLAM^hi memory T_H1 and CXCR5⁺SLAM^lo memory T_FH cells are shown in representative contour plots (left panels). Cumulative data on the numbers of T_H1 and T_FH cells at both effector and memory phases are summarized in line graphs (right panels), and the fold reduction during contraction of each cell type is calculated (n = 8 from 3 experiments). (D) Detection of Tcf1-EGFP reporter in CXCR5⁻SLAM^hi T_H1 and CXCR5⁺SLAM^lo T_FH cells. p45^+/− T_H1 and T_FH cells at the effector (8 dpi, left) and memory (>40 dpi, right) phases were analyzed for EGFP expression, and values in the histograms denote EGFP gMFI. (E) Detection of secondary T_H1 and T_FH effector responses. The immune CD45.1⁺ recipients were infected with LCMV-Cl13 and five days later, CD45.2⁺ secondary T_H1 and T_FH effector T cells were detected in the spleen. The frequency of each subset is shown in representative contour plots, and the cumulative data on cell numbers are summarized in bar graphs (n =6–8 from 3 experiments). Data are means ± SEM. For all paired comparison, **, p <0.01; ***, p<0.001 by Student’s t-test.

DISCUSSION

Tcf1 is known for its versatile functions in regulating T cell development and mature CD8⁺ and CD4⁺ T cell responses to infections (15, 33). Most of the previous studies involved deletion of all Tcf1 isoforms. Until recently, Tcf1 short isoforms were generally considered to function as dominant negatives for the full-length molecule, due to their inability to interact with the co-activator β-catenin. Using a new mouse model where generation of transcripts for Tcf1 long isoforms was abrogated, we achieved specific retention of Tcf1 short isoforms only. We have recently showed that Tcf1 short isoforms are sufficient to support developing thymocytes to traverse through maturation stages without causing obvious developmental blocks, whereas Tcf1 long isoforms remain necessary for optimal thymic survival (23). Coupled with MHC-I- or MHC-II-restricted TCR transgenes, we employed the Tcf1 long isoform-targeted mice to reveal isoform-specific requirements for CD8⁺ and CD4⁺ T cell differentiation in response to viral infection.

At the peak of CD8⁺ T cell response to LCMV, the terminally differentiated KLRG1^hiCD127⁻ CD8⁺ effector T cells exhibit strong downregulation of all Tcf1 isoforms. It is therefore not surprising that loss of Tcf1 long isoforms did not show a detectable impact on the numbers or functions of CD8⁺ effector T cells. Recent studies examined early events of CD8⁺ T cell activation in vivo and found that Tcf1 expression is retained at high levels in all daughter cells during the first 3–4 divisions. In the ensuing divisions, due to asymmetrical signaling via the PI3K pathway, most daughter cells obliterate Tcf1 expression and become differentiated CD8⁺ effector T cells, while some daughter cells maintain higher Tcf1 expression, acquire ‘self-renewal’ capacity, and give rise to memory CD8⁺ T cells (34, 35). In this scenario, retention of Tcf1 short isoforms only appears to be adequate to maintain the self-renewing Tcf1^hi population in spite of diminished amount of protein expression because the memory CD8⁺ T cell pool was similar between p45^+/− and p45^−/− genotypes. It is currently unknown if the Tcf1^hi population at the effector phase is responsible for generation of both T_EM and T_CM cells. Loss of all Tcf1 isoforms almost completely abrogated T_CM maturation (16); in contrast, p45^−/− T_CM cells were indeed generated, albeit at a modestly lower frequency. These data suggest that Tcf1 long isoforms remain necessary to promote optimal T_CM maturation.

Similar to the terminally differentiated CD8⁺ effector T cells, CXCR5⁻SLAM^hi T_H1 cells at the effector phase of CD4⁺ T cell responses potently downregulated Tcf1 expression, and loss of Tcf1 long isoforms was rather inconsequential for generation of T_H1 effectors. Unlike memory CD8⁺ T cells which were unaffected in numbers by loss of Tcf1 long isoforms, p45^−/− memory T_H1 cells were greatly diminished. It has also been shown that after 4–5 divisions during early CD4⁺ T cell activation, Tcf1 expression is specifically retained in a subset of activated CD4⁺ T cells due to asymmetrical signaling (36). The Tcf1^hi CD4⁺ cells seem to possess ‘self-renewal’ capacity as well, although whether they are direct precursors of memory CD4⁺ T cells remains to be conclusively demonstrated. Nonetheless, formation of memory T_H1 cells is strongly dependent on Tcf1 long isoforms.

Another major CD4⁺ T cell response to viral infection is differentiation to T_FH lineage, and deletion of all Tcf1 isoforms severely impairs generation of T_FH and GC-T_FH cells due to insufficient induction of Bcl6 and ICOS (12, 19, 20). This study shows that loss Tcf1 long isoforms greatly diminished the total number of T_FH cells but only modestly affected GC-T_FH cells. Transcriptomic analysis, coupled with phenotypic characterization, revealed that ablating Tcf1 long isoforms had a rather limited impact on the gene expression profile of T_FH cells compared with deleting all Tcf1 isoforms, similar to what we have observed in early thymocytes (23). Apparently, Tcf1 short isoforms, even at a reduced amount of protein, appeared to be sufficient to support induction of Bcl6 and ICOS in T_FH cells. Differentiation of non-GC T_FH to GC T_FH is a controlled process, and a recent study identified Id3 as a restraining factor (13). Our data demonstrated that Tcf1 directly binds to the Id3 gene locus and Tcf1 short isoforms alone are not sufficient to support optimal Id3 expression in T_FH cells. Paradoxically, p45^−/− T_FH cells upregulated Maf, another key T_FH regulator, and AP-1 factors which potentially cooperate with Bcl6. These gene expression changes may all indirectly contribute to promoting GC-T_FH differentiation, in addition to direct activation of Bcl6 transcription by Tcf1 itself. p45^−/− T_FH cells also exhibited inappropriate activation of T_H1-associated genes, including Prdm1, Il2ra and Ifng, indicating another unique requirement for Tcf1 long isoforms in protecting the T_FH lineage integrity (37).

Recent studies have provided strong evidence that T_H1 and T_FH cells are largely committed to respective lineages at both effector and memory phases (38, 39). Loss of Tcf1 long isoforms not only impaired T_FH differentiation at the effector phase of CD4⁺ T cell responses, but also caused a stronger contraction, resulting in even fewer memory T_FH cells. It has been suggested that stronger TCR stimulation during CD4⁺ effector differentiation is associated with predisposition to form memory CD4⁺ T cells (40). An increased strength of TCR signaling, resulting from extended dwell time of TCR with peptide-MHC II complex, is considered to favor T_FH differentiation (41). Since both memory CD4⁺ T cells and T_FH differentiation were affected by loss of Tcf1 long isoforms, an interesting possibility is that Tcf1 may contribute to regulation of TCR signaling outcome, albeit not affecting TCR avidity or dwell time per se. In light of our recent finding that Tcf1 has intrinsic HDAC activity (22), Tcf1 may have global impact on the epigenome of T cells and hence affect their response to external stimuli. This scenario merits further investigation.

Collectively, our study identified differential requirements for Tcf1 long and short isoforms in mature CD8⁺ and CD4⁺ T cell responses. It is of interest to note the expression levels of Tcf1 short isoforms in p45^−/− cells were modestly reduced compared with those in control cells. In spite of this, the Tcf1 short isoforms were adequate in supporting the memory CD8⁺ T cell pool and induction of Bcl6 and ICOS in T_FH cells, indicating a positive regulatory role rather than being merely dominant negatives. On the other hand, Tcf1 long isoforms, plus appropriate amount of short isoforms, remained important for supporting T_FH differentiation at the effector phase and memory T_H1 and T_FH formation. It remains unknown if the defects associated with p45^−/− cells are solely ascribed to lack of interaction with β-catenin. Nevertheless, our study supports a notion that generation of multiple isoforms from one gene may represent an evolutionarily conserved, fail-safe mechanism to ensure preservation of key biological functions.

Abbreviations

Tcf1

T cell factor 1

T_FH

follicular helper T cells

T_H1

helper 1 T cells

LCMV

lymphocytic choriomeningitis virus