BCOR and ZC3H12A suppress a core stemness program in exhausted CD8⁺ T cells

During chronic viral infections, prolonged exposure to antigenic stimulation leads to T cell exhaustion, characterized by heightened expression of inhibitory receptors and diminished proliferative and effector functions (Hashimoto et al., 2018; McLane et al., 2019). While T cell exhaustion serves to prevent immunopathology, it concurrently hinders antiviral immunity (Collier et al., 2021). Precursor-exhausted T cells (T_PEX), crucial for sustaining T cell responses, possess the ability to self-renew, generate TCF1⁻ exhausted effector T cells (T_EX), and respond to PD-1 blockade (He et al., 2016; Im et al., 2016; Leong et al., 2016; Utzschneider et al., 2016; Wu et al., 2016). The defining transcriptional factor of T_PEX, TCF1, plays an indispensable role in their generation and maintenance (Gounari and Khazaie, 2022; Zhao et al., 2022). These cells also express other markers associated with stemness, including CXCR5, Ly108, MYB, and BACH2 (He et al., 2016; Im et al., 2016; Leong et al., 2016; Tsui et al., 2022; Utzschneider et al., 2016, 2020; Wu et al., 2016; Yao et al., 2021). In contrast, TCF⁻ T_EX express effector molecules like GZMB and high levels of exhaustion markers such as PD-1, TIM-3, and CD39, lacking self-renewing capability and relying on replenishment from TCF1⁺ T_PEX (He et al., 2016; Im et al., 2016; Leong et al., 2016; Utzschneider et al., 2016; Wu et al., 2016). While the central role of TCF1⁺ T_PEX in sustaining antiviral and antitumor immunity is well-established (Kallies et al., 2020; Zehn et al., 2022), overexpression of TCF1 only modestly increases T_PEX (Shan et al., 2021), indicating that TCF1 is insufficient to fully program T_PEX. Thus, factors that could potently program T_PEX to reverse T cell exhaustion remain to be determined.

Furthermore, a long-standing dilemma in reversing T cell exhaustion is the seemingly mutual repulsion between the stemness of T_PEX and the functionality of T_EX. Thus, strategies enhancing the effector function of T_EX, like PD-1 blockade, often prove transient due to the consumption of T_PEX (Pauken et al., 2016). Conversely, factors increasing T_PEX, such as TCF1, TOX (Alfei et al., 2019; Khan et al., 2019; Scott et al., 2019; Yao et al., 2019), and MYB (Tsui et al., 2022), typically restrict the effector function of T_EX. Therefore, a thorough and deeper understanding of T_PEX biology, particularly identifying factors robustly dictating T_PEX differentiation and maintenance (Kallies et al., 2020; Zehn et al., 2022), is crucial for designing effective strategies to reinvigorate exhausted T cells against chronic infection and cancer.

We recently reported the remarkable longevity achieved by depleting two factors, BCOR and ZC3H12A, in chimeric antigen receptor T cells (Jin et al., 2024; Wang et al., 2024). However, it was unclear whether BCOR and ZC3H12A play any role in T cell exhaustion during viral infections. The Armstrong and clone 13 strains of lymphocytic choriomeningitis virus (LCMV) share identical dominant T cell epitopes but elicit acute and chronic infections in mice, respectively. These strains have been extensively employed to dissect the mechanisms of CD8⁺ T cell responses in mice (Wherry et al., 2003). We transduced LCMV-specific P14 TCR transgenic CD8⁺ T cells (Cas9⁺) with a retroviral vector co-expressing a Thy1.1 marker and sgRNA-targeting nontargeting control (sgControl), Bcor (sgBcor), Zc3h12a (sgZc3h12a), or Bcor and Zc3h12a simultaneously (sgBcor/Zc3h12a) to knock out these genes in P14 cells (Fig. 1 A; and Fig. S1, A and B). To exclude cell-extrinsic effects, these edited cells were mixed with P14 cells (Cas9⁺) expressing a GFP reporter and sgControl at a 1:1 ratio (Fig. 1 A). The mixture was then transferred into Cas9⁺ B6 mice (to avoid potential rejection of Cas9 protein) infected with LCMV Armstrong or clone 13 1 day before cell transfer (Fig. S1 A). P14 cell proliferation, contraction, and persistence were monitored in the peripheral blood and spleen. Editing of Bcor and Zc3h12a genes in CD8⁺ T cells was validated by PCR and DNA sequencing (Fig. S1 C). For simplicity, we designated P14 cells expressing sgControl, sgBcor, sgZc3h12a, or sgBcor/Zc3h12a as P14_C, P14_B, P14_Z, or P14_BZ, respectively.

In LCMV Armstrong acute infection, P14_Z and P14_BZ, but not P14_B, displayed enhanced expansion at the peak response compared with P14_C (Fig. 1 B), consistent with ZC3H12A acting as a negative regulator of T cells (Uehata et al., 2013; Wei et al., 2019). However, P14_Z and P14_BZ contracted rapidly after the peak response, showing similar percentages to P14_C in the blood 28 days after infection (Fig. 1 B). In LCMV clone 13 chronic infection, P14_B, but not P14_Z, exhibited enhanced expansion and persistence compared with P14_C in the blood (Fig. 1 B). While the kinetics of P14_Z were similar to P14_C, P14_BZ exhibited even more pronounced persistence than P14_B (Fig. 1 B), indicating that, although ZC3H12A deficiency alone has a minimal impact on P14 cells, it amplifies the effect of BCOR deficiency during chronic LCMV infection.

On day 36 after infection, P14_Z and P14_BZ, but not P14_B, displayed reduced numbers in the spleens of mice with LCMV Armstrong acute infection (Fig. 1, C and D), indicating that ZC3H12A, but not BCOR, is required for memory CD8⁺ T cell differentiation in acute infection. In stark contrast, there were ∼43-fold more P14_B and about 234-fold more P14_BZ in the spleens of mice chronically infected with LCMV clone 13 (compared with P14_C in the control group), while P14_Z still exhibited a trend of reduction (Fig. 1, C and D).

To rule out potential off-target effects of the gRNAs, we overexpressed codon-optimized Zc3h12a cDNA in P14_BZ and evaluated their expansion. On day 7 after infection, the enhanced expansion of P14_BZ was reversed upon ZC3H12A overexpression (Fig. S1, D and E), demonstrating that ZC3H12A deficiency is required for the expansion of P14_BZ. However, due to the large size of BCOR (1,759 amino acids), retroviral overexpression in primary T cells is technically unfeasible, preventing us from directly reversing P14_BZ expansion through BCOR. As an alternative, we designed a second set of gRNAs targeting Bcor and Zc3h12a (Fig. S1 F). This approach obtained similar results in P14 cell expansion and persistence to those observed in Fig. 1, A and B; and Fig. S1, G and H. These findings confirm that the phenotypes observed in P14_B, P14_Z, and P14_BZ are not attributable to off-target effects of the gRNAs. Collectively, these results demonstrate that BCOR strongly represses the expansion and persistence of CD8⁺ T cells during chronic LCMV infection, and this repression is further alleviated by ZC3H12A deficiency.

The diminished memory and exhausted ZC3H12A-deficient P14 cells in our infection models contrasts with the enhanced antitumor efficacy of ZC3H12A-deficient CD8⁺ T cells reported in previous studies (Wei et al., 2019; Zheng et al., 2021). To address this discrepancy, we employed the B16-GP33 tumor model to investigate the phenotypes of ZC3H12A-deficient P14 cells in a tumor setting. C57BL/6 mice were subcutaneously inoculated with B16-GP33 tumor cells, and 7 days later, received adoptive transfer of either P14_C or P14_Z. Our results revealed that tumor growth was significantly delayed in mice receiving P14_Z cells compared with those receiving P14_C cells (Fig. S1, I and J). Flow cytometry analysis further demonstrated a markedly higher number of ZC3H12A-deficient P14 cells in both the spleen and tumor tissues compared with control P14 cells (Fig. S1, K and L). These findings align with earlier reports (Wei et al., 2019; Zheng et al., 2021) and suggest that ZC3H12A deficiency drives distinct outcomes in CD8⁺ T cell responses depending on whether the setting is tumor or infection related.

Since both P14_B and P14_BZ persisted in large numbers in mice with chronic LCMV infection (Fig. 1, B–D), we evaluated their stem-like feature through serial transfer experiments. P14 cells from primary recipients were isolated by FACS and transferred into secondary recipient mice, with primary P14 cells as a reference (Fig. 1 E). Serial sampling of peripheral blood revealed that P14_BZ expanded and persisted in secondary recipients after LCMV clone 13 infection, while the secondary expansion/persistence of P14_B was much weaker, albeit still stronger than primary P14_C (Fig. 1 F). Consistently, there were significantly more P14_BZ than P14_B in the spleen 4 wk after secondary transfer (Fig. 1, G and H). These data demonstrate that P14_BZ possess superior stem-like feature than P14_B, aligning with more P14_BZ than P14_B in primary hosts (Fig. 1, B–D).

These findings unveil three critical insights: (1) ZC3H12A promotes CD8⁺ T cell persistence in both acute and chronic LCMV infections, contrary to its known role as a negative regulator of T cells (Uehata et al., 2013; Wei et al., 2019); (2) BCOR, a transcription repressor seldom studied in CD8⁺ T cells (Kotov et al., 2019), emerges as a potent suppressor of CD8⁺ T cell expansion and persistence specifically in chronic, but not acute, LCMV infection; and (3) the combined deficiency of BCOR and ZC3H12A confers robust stemness to CD8⁺ T cells during chronic viral infection compared with BCOR deficiency alone.

The exceptional stemness exhibited by P14_BZ prompted an investigation into whether these cells could effectively suppress viral replication. To exclude the influence of the endogenous CD8⁺ T cell response, Cd8^−/− mice were chosen as recipients. These mice were infected with LCMV clone 13 for 30 days to establish chronic infection, following which either P14_C or P14_BZ were transferred into them (Fig. 1 I). P14_BZ demonstrated significantly enhanced expansion in Cd8^−/− mice compared with P14_C (Fig. 1, J and K), resulting in improved viral control (Fig. 1 L). These findings underscore the capability of P14_BZ to effectively curb viremia when introduced into mice already established with chronic LCMV infection, suggesting promising therapeutic potential.

To further validate the preceding data, we generated conditional knockout mice with T cell–specific ablation of Bcor, Zc3h12a, or both using Cd4^Cre (Fig. S2, A–C and L–N), designated as Cd4^CreBcor^fl/fl, Cd4^CreZc3h12a^fl/fl, and Cd4^CreBcor^fl/flZc3h12a^fl/fl mice, respectively. Since T cell homeostasis in Cd4^CreBcor^fl/fl and Cd4^CreBcor^fl/flZc3h12a^fl/fl mice has not been studied before, we first characterized these mice. T cell development and homeostasis were comparable between Cd4^Cre (as control) and Cd4^CreBcor^fl/fl mice (Fig. S2, D–K), indicating that BCOR is largely dispensable for T cell homeostasis under steady state. Cd4^CreZc3h12a^fl/fl mice exhibited spontaneous T cell activation and began to succumb from 9 wk old (data not shown), consistent with previous reports (Uehata et al., 2013). T cell development in the thymus of Cd4^CreBcor^fl/flZc3h12a^fl/fl mice was not altered compared with control mice (Fig. S2, O and P). Spontaneous T cell activation and expansion were observed in Cd4^CreBcor^fl/flZc3h12a^fl/fl mice (Fig. S2, Q–W), similar to what has been reported in Cd4^CreZc3h12a^fl/fl mice (Uehata et al., 2013).

Under steady state, Cd4^Cre, Cd4^CreBcor^fl/fl, Cd4^CreZc3h12a^fl/fl, and Cd4^CreBcor^fl/flZc3h12a^fl/fl mice survived beyond 8 wk (Fig. S2 X). We infected these mice with LCMV clone 13 at 4 wk old when they were healthy and monitored the endogenous CD8⁺ T cell response and mouse survival (Fig. 2 A). Cd4^CreBcor^fl/flZc3h12a^fl/fl mice succumbed between 5 and 8 days after LCMV clone 13 infection, while mice from other groups survived (Fig. 2 B). FACS analysis of T cells from the spleens of surviving mice showed that H-2D^b-gp33 tetramer⁺ CD8⁺ T cells were increased in Cd4^CreBcor^fl/fl mice compared with control mice and further increased in Cd4^CreBcor^fl/flZc3h12a^fl/fl mice (Fig. 2, C and D). Ex vivo stimulation of splenocytes with gp33 and gp276 peptides revealed a trend of increasing CD8⁺IFNγ⁺ T cells in Cd4^CreBcor^fl/fl mice and a further increase in Cd4^CreBcor^fl/flZc3h12a^fl/fl mice (Fig. 2, E and F). These data indicate that the knockout of both BCOR and ZC3H12A in T cells enhances CD8⁺ T cell expansion and effector function during chronic LCMV infection.

As a control experiment, we infected 4-wk-old control and Cd4^CreBcor^fl/flZc3h12a^fl/fl mice with LCMV Armstrong and monitored these mice (Fig. 2 G). Cd4^CreBcor^fl/flZc3h12a^fl/fl mice did not succumb to LCMV Armstrong acute infection (Fig. 2 H), and the percentages of H-2D^b-gp33 tetramer⁺ CD8⁺ T cells were comparable between control and Cd4^CreBcor^fl/flZc3h12a^fl/fl mice (Fig. 2, I and J). Nevertheless, the absolute cell number of GP33⁺ CD8⁺ T cells was slightly higher in Cd4^CreBcor^fl/flZc3h12a^fl/fl mice due to the increase in total CD8⁺ T cells in the spleens (Fig. 2 K). These data indicate that the deletion of BCOR and ZC3H12A has a limited impact on CD8⁺ T cell response in LCMV Armstrong acute infection, consistent with P14 transfer data (Fig. 1, B–D).

Since the Cd4^Cre strain deletes genes in both CD4⁺ and CD8⁺ T cells, we investigated the role of CD4⁺ T cells in the phenotypes of Cd4^CreBcor^fl/flZc3h12a^fl/fl mice during LCMV infection by generating a Cd8^CreBcor^fl/flZc3h12a^fl/fl mouse model in which Bcor and Zc3h12a were specifically deleted in CD8⁺ T cells using the Cd8^Cre mouse strain. This Cre driver is exclusively expressed in peripheral, mature CD8⁺ T cells but not in CD4⁺ T cells (Maekawa et al., 2008). Specific deletion of Bcor and Zc3h12a in CD8⁺ T cells, but not in CD4⁺ T cells, was confirmed (Fig. S2 Y). Upon infection, Cd8^CreBcor^fl/flZc3h12a^fl/fl mice succumbed to LCMV clone 13 chronic infection but survived LCMV Armstrong acute infection (Fig. 2, L and M). These findings indicate that, while CD4⁺ T cells may contribute to the mortality observed in Cd4^CreBcor^fl/flZc3h12a^fl/fl mice following LCMV clone 13 infection, deletion of Bcor and Zc3h12a in CD8⁺ T cells alone is sufficient to drive these phenotypes.

These findings demonstrate that, in chronic but not acute LCMV infection, BCOR and ZC3H12A collaboratively repress endogenous CD8⁺ T cell expansion and effector production, thereby mitigating immunopathology. These phenotypes are reminiscent of PD-1– or PD-L1–deficient mice, which also rapidly succumbed to chronic but not acute LCMV infection (Barber et al., 2006; Frebel et al., 2012), suggesting analogous functions of these pathways in restraining T cell responses and preventing immunopathology.

To investigate the regulatory role of BCOR and ZC3H12A in T_PEX, we employed the well-established TCF1 and TIM3 markers for identifying T_PEX and T_EX (Fig. 3 A) (He et al., 2016; Im et al., 2016; Leong et al., 2016; Utzschneider et al., 2016; Wu et al., 2016). On day 30 after infection, P14_C, P14_B, P14_Z, and P14_BZ exhibited ∼17%, 56%, 72%, and 90% of TCF1⁺TIM3⁻ T_PEX (Fig. 3, D and E), respectively, while TCF1⁻TIM3⁺ T_EX were proportionally reduced in their respective populations (Fig. 3, D and E). This indicates that both BCOR and ZC3H12A act to inhibit the differentiation of TCF1⁺ T_PEX with an additive effect. However, increased apoptosis of P14_Z resulted in a similar absolute number of TCF1⁺ T_PEX in both P14_Z and P14_C, despite the heightened percentages of TCF1⁺TIM3⁻ cells (Fig. 3, B–E). Conversely, both P14_B and P14_BZ exhibited significantly reduced apoptosis compared with P14_C (Fig. 3, B and C), indicating that BCOR deficiency not only promotes P14_B survival but also counteracts the pro-apoptotic effect of ZC3H12A deficiency observed in P14_Z. Consequently, the absolute number of TCF1⁺ T_PEX dramatically increased in P14_B and P14_BZ compared with P14_C (Fig. 3 E). Although the percentages of TCF1⁻TIM3⁺ T_EX were reduced in P14_B and P14_BZ, the absolute numbers of these cells were increased (Fig. 3, D and E) due to the increased total numbers of P14_B and P14_BZ, indicating that the increase of TCF1⁺ T_PEX in P14_B and P14_BZ is not at the cost of losing TCF1⁻ effector cells.

Recent studies have identified a subset of CD62L⁺TCF1⁺ T_PEX characterized by high stemness, governed by the transcriptional factor MYB (Tsui et al., 2022). To further characterize the stemness features of P14 cells in chronic LCMV infection, we utilized CD62L and the stemness marker Ly108. On day 30 after infection, the percentages of Ly108⁺CD62L⁺ T_PEX increased in P14_Z and further in P14_BZ, while remaining unchanged in P14_B (Fig. 3, F and G). This indicates that ZC3H12A deficiency promotes Ly108⁺CD62L⁺ T_PEX differentiation, an effect further enhanced by BCOR deficiency, although BCOR deficiency alone does not influence the percentages of these cells. When taking total cell numbers into account, the absolute numbers of Ly108⁺CD62L⁺ T_PEX remained unchanged in P14_Z but increased ∼10-fold in P14_B and exhibited a significant ∼100-fold increase in P14_BZ (Fig. 3 G).

Previously, we reported that BCOR/ZC3H12A double knockout chimeric antigen receptor T cells exhibited a distinct Ly108⁺CX3CR1⁺ phenotype (Wang et al., 2024). On day 30 following LCMV clone 13 infection, we observed that deletion of BCOR and/or ZC3H12A did not increase the Ly108⁺CX3CR1⁺ subpopulation in P14_BZ, and the frequency of CX3CR1⁺ subset was reduced in P14_Z and P14_BZ (Fig. 3, H and I). However, despite the reduced frequency, the total number of CX3CR1⁺ P14_BZ cells remained higher than that of P14_C (Fig. 3 I), owing to the dramatic increase in the overall number of P14_BZ.

The expression of proteins associated with T cell stemness, including CXCR5 (He et al., 2016; Leong et al., 2016), LEF1 (Gounari and Khazaie, 2022; Zhao et al., 2022), and the pro-survival factor BCL-2, showed a concurrent strong synergistic upregulation in P14_BZ (Fig. 3, J and K), indicating a potent cooperative effect between BCOR and ZC3H12A in regulating these proteins. Conversely, markers linked to T cell exhaustion, such as PD-1, LAG-3, and CD39, were significantly downregulated by the joint action of BCOR and ZC3H12A (Fig. 3, J and K).

Importantly, although the percentages of IFNγ⁺ cells upon ex vivo stimulation with Gp33 peptide were comparable among P14_C, P14_B, P14_Z, and P14_BZ, the absolute numbers of IFNγ⁺ P14_B and IFNγ⁺ P14_BZ significantly increased due to the increased total numbers of these populations (Fig. 3, L and M). This demonstrates that the ablation of BCOR and ZC3H12A does not compromise the effector functions of P14 cells during chronic viral infection.

Collectively, these findings demonstrate the potent and synergistic repression of the generation and/or maintenance of TCF1⁺ T_PEX and CD62L⁺ T_PEX by BCOR and ZC3H12A.

The inflexibility of the epigenetic program in exhausted T cells poses a considerable challenge for reinvigoration once exhaustion has been established (Abdel-Hakeem et al., 2021; Pauken et al., 2016; Yates et al., 2021). For instance, despite TOX being required and sufficient to orchestrate T_PEX differentiation, inducible deletion of TOX 20 days after LCMV clone 13 chronic infection exhibited negligible impact on T_PEX cells (Alfei et al., 2019). To assess whether depleting BCOR and ZC3H12A could reinvigorate T_PEX after exhaustion had set in, we transferred ERT2^CreP14 or ERT2^CreBcor^fl/flZc3h12a^fl/flP14 cells into B6 mice and infected the mice with LCMV clone 13 (Fig. S3 A). After 21 days, when chronic viral infection was established, mice were treated with tamoxifen to delete Bcor and Zc3h12a in exhausted P14 cells (Fig. S3 A). The percentages of transferred ERT2^CreP14 and ERT2^CreBcor^fl/flZc3h12a^fl/flP14 cells in the blood were comparable before tamoxifen treatment (Fig. S3 B). Following tamoxifen treatment, ERT2^CreBcor^fl/flZc3h12a^fl/flP14 cells outnumbered control P14 cells (Fig. S3, B–D). More importantly, there were significantly more TCF1⁺ or CD62L⁺ T_PEX cells among ERT2^CreBcor^fl/flZc3h12a^fl/flP14 cells than those in ERT2^CreP14 cells (Fig. S3, E–H). These data demonstrate the reinvigoration of T_PEX during chronic LCMV infection by targeting BCOR and ZC3H12A.

To rule out cell-extrinsic effects on the phenotypes of gene-edited P14 cells, we employed a co-transfer system to assess the potential influence of extrinsic factors on the phenotypes of P14_BZ and P14_C cells transferred into the same mouse (Fig. S3 I). Consistent with the data in Fig. 1 B, we observed a higher abundance of P14_BZ cells in the peripheral blood of recipient mice compared with P14_C (Fig. S3 J). On day 30 after LCMV clone 13 infection, the proportion of CD62L⁺Ly108⁺ cells was significantly increased in P14_BZ compared with P14_C (Fig. S3, K and L). These findings demonstrate that BCOR and ZC3H12A regulate T_PEX in a cell-intrinsic manner, confirming that the phenotypic differences between P14_BZ and P14_C are not due to variations in cell-extrinsic factors, such as antigen load.

We delved into the molecular mechanism underpinning the exceptional stemness of P14_BZ. Principal component analysis (PCA) of bulk RNA-sequencing (RNA-seq) analysis performed on day 14 after infection revealed distinct segregation among P14_C, P14_B, P14_Z, and P14_BZ (Fig. S3 M), indicating different cellular states. P14_B and P14_Z exhibited unique gene expression patterns, while P14_BZ displayed features characteristic of both (Fig. S3 N). Notably, BCOR deficiency synergized with ZC3H12A deficiency in inducing genes like Sell, Myb, and Ccr7, and repressing genes like Pdcd1, Tight, and Entpd1 (Fig. S3 N). Antagonistic effects were also observed, exemplified by the regulation of genes, such as Nr4a1 and Fos (Fig. S3 N). Gene set enrichment analysis (GSEA) highlighted shared pathways enriched in P14_B, P14_Z, and P14_BZ (Fig. S3, O and P). These findings suggest that BCOR and ZC3H12A both synergistically and antagonistically regulate gene expression in P14 cells during LCMV clone 13 chronic infection, with features like stemness being additively preserved, while other features, such as apoptosis, are offset by each other.

In the assay for transposase-accessible chromatin with high-throughput sequencing (ATAC-seq) analysis performed on day 14 after infection, P14_C, P14_B, P14_Z, and P14_BZ exhibited distinct epigenetic landscapes (Fig. S4, A–D), indicative of different cellular states. Differential chromatin accessibility regions among the four groups of cells were categorized into five main clusters, C1 to C5 (Fig. S4 B). Loci in C1-associated exhaustion-related genes like Tox, Pdcd1, and Havcr2 showed reduced accessibility in P14_B, P14_Z, and P14_BZ compared with control (Fig. S4, B and E). Loci in C2 associated with proliferation and exhaustion-related genes like Tox2, E2f3, Irf4, etc., which were only increased in P14_B (Fig. S4 B), emphasizing the dominant role of BCOR in their regulation. Loci in C3 associated with T cell activation, such as Cd44, were highly upregulated in P14_Z but only slightly in P14_BZ (Fig. S4 B), indicating a partial offsetting of ZC3H12A’s influence by BCOR. Loci in C4 and C5 associated with stemness and survival genes, like Tcf7, Lef1, Sell, Bcl2, etc., were moderately increased in P14_B or P14_Z but highly enhanced in P14_BZ (Fig. S4, B and E), pointing to a synergistic regulation of stemness by BCOR and ZC3H12A.

To gain detailed insights into the epigenetic landscape of P14_C, P14_B, P14_Z, and P14_BZ, we performed single-cell ATAC-seq (scATAC-seq) analysis on day 14 after LCMV clone 13 infection. Aggregated scATAC-seq data classified the cells into seven major clusters, designated as clusters 0 through 6 (C0 to C6) (Fig. 4 A). To distinguish T_PEX and terminal T_EX subsets, we analyzed differentially accessible peaks between Ly108⁺Tim3⁻ T_PEX and Ly108⁻Tim3⁺ terminal T_EX from published data (GSE123236) (Miller et al., 2019), generating an epigenetic signature specific for T_PEX and terminal T_EX. C0 and C6 were enriched for T_PEX signature, whereas C1 through C5 were enriched for terminal T_EX signature (Fig. 4, B and D). Consistent with flow cytometry results, P14_BZ displayed a significantly higher proportion of the T_PEX signature (Fig. 4, C and D). These findings demonstrate that the double knockout of BCOR and ZC3H12A synergistically enhances the T_PEX signature at the epigenetic level. Together, these analyses reveal the nuanced interplay of cooperation and antagonism between BCOR and ZC3H12A in shaping the epigenetic landscape and gene expression of P14_BZ during chronic LCMV infection, ultimately conferring enhanced stemness.

To further explore the molecular attributes of P14_BZ, we conducted single-cell RNA-seq (scRNA-seq) analysis of four groups of P14 cells on day 14 after infection. Aggregating P14_C, P14_B, P14_Z, and P14_BZ allowed us to classify them into eight major clusters, designed as C0 to C7 (Fig. 4, E–G). Each population, P14_C, P14_B, P14_Z, and P14_BZ, predominantly occupied two or three distinct clusters with limited overlap (Fig. 4, E and F), indicating unique cellular states for each population.

Specifically, P14_C were primarily distributed across C2, C5, and C6 (Fig. 4, E and F). Cells within C2 and C5 exhibited heightened expression levels of Pdcd1, Havcr2, Entpd1 (CD39), and Gzmb (Fig. 4, G and H; and Fig. S5 A), genes associated with T cell exhaustion. Additionally, cells in C5 expressed high levels of genes associated with the cell cycle and proliferation, including Mki67 (Ki-67), Pcna, Ccnb2 (cyclin B2), Birc5, Cdk1, Cdca8, and Ckap2l (Fig. 4, G and H; and Fig. S5 A), indicative of a proliferative cell phenotype. A subset of P14_C was identified in C6, expressing Tcf7, Tox, Id3, Cxcr5 and exhibiting relatively lower levels of Sell (Fig. 4, G and H; and Fig. S5 A), representing the classic TCF1⁺ T_PEX. These observed phenotypes of P14_C align with previously reported scRNA-seq analyses of WT P14 cells during chronic LCMV infection (Giles et al., 2022).

P14_B predominantly inhabited clusters C0, C4, and C5 (Fig. 4, E and F). Approximately half of P14_B resided in C0, a cluster almost absent in P14_C and P14_Z (Fig. 4, E and F), indicating a significant influence of BCOR in this cluster. Cells in C0 displayed intermediate levels of Sell, Tcf7, Cxcr5, Slamf6 (Ly108), and Pdcd1 but high levels of Tox, Eomes, and Id3 (Fig. 4, G and H; and Fig. S5 A), suggesting a combination of stemness and exhaustion features. This gene expression profile partly explains why P14_B persisted in primary hosts but exhibited limited expansion in secondary hosts (Fig. 1, E–H). Importantly, C2, housing cells expressing the highest levels of Pdcd1, Cd7, and Cxcr6, genes associated with terminal exhaustion (Hudson et al., 2019; Sandu et al., 2020), was almost absent in P14_B (Fig. 4, G and H; and Fig. S5 A). Concurrently, C4, composed of cells expressing effector-related genes such as Gzmb, Cx3cr1, Klrg1, and S1pr5, increased in P14_B (Fig. 4, E and F; and Fig. S5 A). These findings indicate that BCOR deficiency prevents the terminal exhaustion of T cells and maintains their effector functions during chronic infection.

P14_Z predominantly localized within cluster C3 (Fig. 4, E and F), displaying heightened expression of Tcf7 and Sell (Fig. 4, G and H), indicative of stemness, consistent with flow cytometry data (Fig. 3, D–G). However, cells in C3 also exhibited elevated levels of Nr4a1 (Fig. 4, G and H), encoding NR4A1 (also known as NUR77), a factor known to promote T cell death and anergy (Calnan et al., 1995; Liu et al., 2019; Thompson and Winoto, 2008). In addition to Nr4a1, these cells expressed other genes associated with TCR stimulation, such as Fos, Jun, etc. (Fig. 4, G and H), suggesting that P14_Z might undergo enhanced TCR stimulation and potentially activation-induced cell death (AICD). Collectively, the gene expression pattern of P14_Z at the single-cell level explains their increased percentages but reduced total number of TCF1⁺ cells during chronic LCMV infection.

Lastly, P14_BZ predominantly occupied cluster C1, which was barely present in P14_C (Fig. 4, E and F), indicating that the simultaneous ablation of BCOR and ZC3H12A induces a distinct state of P14 cells. Cells within C1 exhibited heightened expression of Sell and Tcf7, feature genes of P14_Z-dominated C3, yet without the expression of genes associated with TCR stimulation and AICD found in C3, including Nr4a1, Fos, Jun, etc. (Fig. 4, G and H). This suggests that BCOR deficiency offsets the induction of genes associated with T cell overactivation and cell death, such as Nr4a1 prompted by ZC3H12A deficiency.

Intriguingly, the emergence of C1 in P14_BZ coincided with the concurrent reduction of P14_B-dominated C0 and P14_Z-dominated C3 in P14_BZ (Fig. 4, E and F). Indeed, the gene signature of C1 can be considered a composite of favorable features from both C0 and C3 (Fig. 4, E–H and Fig. S5 A). Supporting this, gene signature analysis indicated an elevated stemness signature in both P14_B and P14_Z, which was additively preserved in P14_BZ (Fig. 4 I). Conversely, the effector signature was compromised in P14_Z but reinstated in P14_BZ (Fig. 4 J). These findings highlight the cooperative enhancement of stemness and the antagonistic effect on cell survival resulting from the combined deficiencies of BCOR and ZC3H12A.

To further delineate the individual and cooperative roles of BCOR and ZC3H12A in T_PEX, we analyzed scRNA-seq data in greater detail. A substantial fraction of proliferating cells (Mki67⁺Pcna⁺) in C5 within P14_B maintained TCF1 expression, whereas most TCF1⁺ T_PEX in P14_C were not in the cell cycle (Fig. 5, A and B), consistent with previous reports (Chen et al., 2019). These findings suggest that BCOR deficiency enhances the self-renewal capacity of TCF1⁺ T_PEX, explaining the increased numbers of P14_B during chronic LCMV infection.

Gene set variation analysis revealed a significant enrichment of AICD and T cell apoptosis pathways in C3, a predominant population in P14_Z (Fig. 5 C). This aligns with the increased apoptosis observed in P14_Z (Fig. 3, B and C). Since cells in C3 expressed high levels of Nr4a1 (Fig. 4, G and H), a gene implicated in T cell death, we investigated the role of Nr4a1 in the death of P14_Z. Depleting Nr4a1 effectively reversed the reduction in P14_Z cell numbers during LCMV clone 13 chronic infection (Fig. 5, D–F). Moreover, overexpression of NR4A1 diminished the cell number advantage of P14_B and P14_BZ (Fig. 5, G–I), aligning with NR4A1’s established role in inducing T cell death. However, NR4A1 depletion alone did not promote the expansion or persistence of P14 cells during LCMV clone 13 chronic infection (Fig. 5, D–F), in agreement with recent findings (Liu et al., 2019).

Collectively, these results demonstrate that upregulation of Nr4a1 accounts for the increased death of P14_Z, while the dramatic expansion and persistence of P14_B and P14_BZ cannot be solely attributed to Nr4a1 downregulation.

The integrated transcriptomic and epigenetic analyses, combined with rescue experiments, reveal that the dual deficiency of BCOR and ZC3H12A synergistically amplifies the beneficial features observed in individual gene deficiencies while alleviating the associated drawbacks. This cooperation results in the emergence of the unique cluster C1 in P14_BZ cells, a subset of T_PEX scarcely represented in P14_C (Fig. 4, E and F).

We thus explored novel genes within this cluster, as none of the known regulators are sufficient to fully reprogram T_PEX (Kallies et al., 2020; Zehn et al., 2022). Approximately 216 feature genes of cells from C1 were identified (Fig. 6 A), exhibiting heightened expression in P14_BZ compared with P14_C, P14_B, and P14_Z (Fig. 6 A). Out of these 216 genes, only 15 were previously recognized for their necessity but not sufficiency in programming T_PEX (Fig. 6 B), including Batf, Sell, Tcf7, Lef1, Foxp1, Socs1, Slamf6, etc. (Fig. 6 B and Fig. S5 B). The upregulation of some feature genes correlated with increased chromatin accessibility at respective loci, including Tcf7, Pdzk1ip1, and Pou2f2 (Fig. 6 C). Aside from the 15 known genes, the majority of the ∼216 feature genes of P14_BZ were not known to play roles in T cell stemness, grouped into various categories, including transcriptional factors, signaling molecules, mitochondrial respiratory components, ribosome subunits, and uncharacterized proteins, such as 2310001H17RRik, Gm3511, Gm10095, and Gm1361 (Fig. 6 B and Fig. S5 B). Since neither BCOR nor ZC3H12A are transcription factors that directly bind DNA, the upregulation of these target genes might be partially mediated by intermediate factors. Consistently, analysis of our bulk ATAC-seq data revealed that binding motifs for transcription factors such as Bcl6, Ctcf, Lef1, and Tcf7l1 were identified in peaks at the Tcf7 locus, while motifs for Bcl6, E2f3, Myb, and Relb were found in peaks at the Pou2f2 locus in P14_BZ (Fig. S4 F). These findings provide insights into the potential regulatory mechanisms driving the expression of Tcf7 and Pou2f2 in P14_BZ cells.

We next investigated potential factors that are both necessary and sufficient for maintaining the core stemness program. From the substantial number of novel factors identified, we selected five top candidates for functional validation: Pdzk1ip1, Ifit3, Pim2, Ltb, and Pou2f2 (Fig. 6, B–D). To assess their roles, we examined the frequency of CD62L⁺ T_PEX in P14_BZ with further depletion of Pdzk1ip1, Ifit3, Pim2, Ltb, or Pou2f2. The percentage of CD62L⁺ T_PEX remained comparable across these triple-knockout P14 cells on day 14 after infection (Fig. S5, C and D), indicating that the deletion of these genes does not significantly impact T_PEX differentiation. Nevertheless, depletion of these factors, similar to Tcf7 depletion, largely reversed the cell number advantages of P14_BZ observed on days 7 and 14 after LCMV clone 13 infection (Fig. 6, E–G; and Fig. S5, E and F). These findings highlight that each of these selected novel genes plays a critical role in supporting the expansion of P14_BZ.

To investigate whether these novel factors are sufficient to confer stemness to CD8⁺ T cells, we individually overexpressed these factors in P14 cells and examined whether they could enhance the expansion and/or persistence of P14 cells during chronic LCMV infection. Overexpression of TCF1, as well as PDZK1IP1, IFIT3, PIM2, or LTB, minimally impacted the expansion of P14 cells in chronic viral infection (Fig. 6, H and I). In contrast, overexpression of POU2F2 was sufficient to promote the expansion and persistence of P14 cells in chronic LCMV infection (Fig. 6, H and I).

Together, these data demonstrate that the ablation of BCOR and ZC3H12A unleashes a core stemness program in P14_BZ. While TCF1 was traditionally recognized as a marker for this program, a group of novel factors, especially POU2F2, contribute to this core stemness program even more.

Previously, POU2F2 was computationally identified as a potential regulator of T_PEX (Pritykin et al., 2021). Given its pronounced impact on CD8⁺ T cell expansion and persistence (Fig. 6, H and I), our focus turned to a detailed analysis of this factor. Both bulk and scATAC-seq analyses revealed that motifs of POU family transcription factors were modestly enriched in P14_B and P14_Z but highly enriched in P14_BZ (Fig. 7, A–C), suggesting a specific gene module regulated by POU transcription factors in P14_BZ. Among all POU family members, Pou2f2 displayed the highest level of upregulation in P14_BZ (Fig. 7 D). Protein-level validation further confirmed the upregulation of POU2F2 in endogenous CD8⁺ T cells during LCMV clone 13 infection (Fig. 7, E and F). Collectively, these findings demonstrate that BCOR and ZC3H12A synergistically repress POU2F2 expression in CD8⁺ T cells during chronic viral infection.

In gain-of-function studies, the overexpression of POU2F2 in P14 cells was confirmed at the protein level (Fig. 7 G). Over an 8-wk monitoring period, POU2F2 overexpression significantly enhanced the expansion and prolonged the persistence of P14 cells in peripheral blood during LCMV clone 13 chronic infection (Fig. 7 H). 8 wk after infection, there were more POU2F2-overexpressing (OE) P14 cells than control P14 cells in spleens (Fig. 7, I and J). Notably, overexpression of POU2F2 increased the percentages of Ly108⁺CD62L⁻ and Ly108⁺CD62L⁺ T_PEX subsets (Fig. 7, K and L), demonstrating its role in promoting T_PEX generation during chronic viral infection. Additionally, POU2F2 overexpression reduced P14 cell apoptosis 8 wk after infection (Fig. 7, M and N). Importantly, mice receiving POU2F2 OE P14 cells exhibited significantly reduced viral loads compared with those receiving control P14 cells (Fig. 7 O). These findings collectively establish POU2F2 as a potent regulator of T_PEX, driving their generation, persistence, and antiviral functionality during chronic LCMV infection.

Lastly, we performed scRNA-seq analysis of POU2F2 OE P14 cells on day 14 after infection. Unsupervised clustering identified nine major clusters within the combined control and POU2F2 OE P14 cells (Fig. 7, P–R). Clusters 0, 4, 5, 7, and 8 expressed high levels of Sell and Tcf7 (Fig. 7, P and R), corresponding to T_PEX. Clusters 1, 2, and 6 exhibited high expression of Havcr2 and Gzmb (Fig. 7, P and R), indicative of terminal T_EX, while cluster 3 contained proliferating cells characterized by high levels of Mki67 (Fig. 7, P and R). Notably, POU2F2 overexpression increased the proportions of cells in clusters 3 and 4 (Fig. 7, P and Q), indicating an enhancement of both T_PEX generation and cell proliferation. These results align with our FACS data, which also demonstrated increased T_PEX upon POU2F2 overexpression (Fig. 7, K and L).

Next, we integrated the scRNA-seq data of POU2F2 OE P14 cells with our previous analysis of P14_C, P14_B, P14_Z, and P14_BZ. POU2F2 OE P14 cells displayed a high proportion of cells within cluster 0, the dominant cluster observed in P14_B (Fig. 7, S and T). Moreover, the transcriptional changes induced by POU2F2 overexpression in P14 cells partially overlapped with those seen in P14_B, P14_Z, and P14_BZ, with the stronger similarity to P14_B (Fig. 7 U). This is consistent with expectations, as the phenotypic changes in P14_BZ are more pronounced than those in POU2F2 OE P14 cells. Importantly, Pou2f2 represents only one of the many target genes downstream of BCOR and ZC3H12A.

Our study demonstrates that the depletion of BCOR and ZC3H12A unleashes a previously unknown core stemness program in T_PEX. In contrast to other key players in T cell stemness, such as TCF1 (Im et al., 2016; Jeannet et al., 2010; Pais Ferreira et al., 2020; Utzschneider et al., 2016; Wu et al., 2016; Zhou et al., 2010) and MYB (Gautam et al., 2019; Tsui et al., 2022), which promote both memory T cells (T_MEM) in acute infection and T_PEX in chronic infection, the depletion of BCOR and ZC3H12A selectively boosts T_PEX but not T_MEM. Furthermore, factors that enhance T_PEX, such as TCF1 (Pais Ferreira et al., 2020), MYB (Tsui et al., 2022), and TOX (Alfei et al., 2019; Khan et al., 2019; Scott et al., 2019; Yao et al., 2019), usually enforce exhaustion and impair the effector function of T_EX. In contrast, depletion of BCOR and ZC3H12A not only boosts T_PEX but also mitigates exhaustion, maintaining the effector function of T_EX. For simplicity, we term T_PEX devoid of BCOR and ZC3H12A as BCOR and ZC3H12A double knockout T_PEX (T_BZEX).

Previous studies have suggested that the T_EX state in chronic infections is epigenetically fixed, making the reversal of T cell exhaustion challenging (Abdel-Hakeem et al., 2021; Pauken et al., 2016; Yates et al., 2021). For instance, although TOX is a key regulator of T_EX establishment and maintenance, inducible deletion of TOX had no effect on established T_EX (Alfei et al., 2019). Our data showed that targeting BCOR/ZC3H12A in established T_EX can partially reinvigorate these cells. Thus, BCOR and ZC3H12A represent promising targets to reinvigorate T_EX in chronic viral infections.

The observation that depletion of BCOR and ZC3H12A specifically boosts T_PEX but not T_MEM suggests that the sustained presence of viral antigens and/or inflammation during chronic viral infection is necessary to induce the core stemness program in T_BZEX. Thus, environmental cues in chronic infections may constrain T_PEX from acquiring robust stemness and functionality through modulation of BCOR and ZC3H12A. This represents a mechanism that limits immunopathology during chronic viral infection, as indicated by the increased mortality of mice with T cell–specific deletion of BCOR and ZC3H12A. Interestingly, PD-1 or PD-L1 knockout mice also exhibited rapid mortality during chronic infection with LCMV clone 13 but not during acute infection with LCMV Armstrong (Barber et al., 2006; Frebel et al., 2012). This phenomenon is generally attributed to immunopathology, although the exact cause of death in PD-1/PD-L1–deficient mice infected with LCMV clone 13 remains unclear. A similar observation applies to our study, where increased immune infiltration into various tissues is anticipated. However, a detailed investigation into the cause of death warrants future studies.

BCOR and ZC3H12A had not been previously studied in CD8⁺ T cells during viral infections. ZC3H12A (MCPIP1/REGNASE-1), known as a negative regulator of T cells (Uehata et al., 2013; Wei et al., 2019), is a ribonuclease-mediating mRNA decay (Fu and Blackshear, 2017), explaining comprehensive changes in mRNA levels in its absence. Ablation of ZC3H12A has been reported to boost CD8⁺ T cell expansion and antitumor efficacy in tumor models (Zheng et al., 2021), which was confirmed in our study in the B16-gp33 tumor model. In contrast, our data show that ablation of ZC3H12A alone inhibits CD8⁺ T cell stemness in both acute and chronic LCMV infections, suggesting context-dependent roles of ZC3H12A in T cells. ZC3H12A deficiency promotes T_PEX differentiation in part by upregulation of TCF1. However, ZC3H12A deficiency also increases T_PEX death by upregulation of NR4A1 during chronic LCMV infection, counteracting its positive role in T_PEX differentiation. ZC3H12A mediates mRNA decay, raising the possibility that its deficiency may extend the half-life of NR4A1 mRNA, leading to increased expression. However, the decay of NR4A1 mRNA was not slowed in the absence of ZC3H12A (data not shown), suggesting that NR4A1 is unlikely to be a direct target of ZC3H12A. Since NR4A1 is a transcription factor induced by TCR signaling, its increased expression in ZC3H12A-deficient T cells may be secondary to their overactivation, warranting further investigation.

BCOR is a transcriptional repressor seldom studied in CD8⁺ T cells (Kotov et al., 2019). It is unexpected that ablation of BCOR dramatically boosts CD8⁺ T cell expansion and persistence, specifically in chronic but not acute LCMV infection. The deficiency of BCOR enhances the self-renewing ability of T_PEX, evidenced by an increase in cycling T_PEX in its absence. BCOR deficiency also mitigates terminal exhaustion of T cells by restraining the expression of Pdcd1, Cd7, and Cxcr6. Surprisingly, BCOR deficiency rescues ZC3H12A-deficient T cells from apoptosis in chronic LCMV infection. Thus, in T_BZEX devoid of both ZC3H12A and BCOR, the stemness program induced by ZC3H12A deficiency and the proliferation program induced by BCOR deficiency were additively preserved, while ZC3H12A deficiency–induced apoptosis was offset by BCOR deficiency. This unique combination of genetic manipulations induces a core stemness program in T_BZEX without impairing the effector function of T_EX.

This core stemness program in T_BZEX is characterized by the heightened expression of ∼216 genes. Unexpectedly, only 15 (<10%) of these feature genes are previously known to play a role in T_PEX, including Tcf7 (Chen et al., 2019), Sell (Tsui et al., 2022), Batf (Chen et al., 2021), Irf4 (Man et al., 2017), Foxp1 (Zhu et al., 2024), Socs1 (Wei et al., 2019), etc. Consistent with an important role of TCF1 in T_PEX, TCF1 is also required for this core stemness program, as its absence largely nullifies T_BZEX induction. Beyond TCF1, we validated that a group of previously unappreciated factors are also essential for inducing and/or maintaining this core stemness program of T_BZEX, including PDZK1IP1, IFIT3, PIM2, LTB, and POU2F2. Intriguingly, while the overexpression of TCF1 minimally boosts T cell responses during chronic LCMV infection, the overexpression of POU2F2, a transcription factor seldom studied in T cells (Clerc et al., 1988; Kang et al., 1992; Schubart et al., 2001), significantly enhances CD8⁺ T cell responses and antiviral immunity in chronic LCMV infection. These findings indicate that these novel factors play even more important roles than TCF1 in T_PEX. Notably, Pou2f2 was previously identified as a potential regulator of T_PEX computationally (Pritykin et al., 2021). Our study experimentally validated a critical role of POU2F2 in T cell exhaustion and further put it into a bigger program downstream of BCOR and ZC3H12A. Nevertheless, these factors only represent a small portion of the ∼216 feature genes of the core stemness program in T_BZEX. Future investigations into other poorly characterized factors within this core stemness program will further broaden and deepen our understanding of T_PEX biology.

In summary, our study reveals a core stemness program in T_PEX, naturally suppressed by BCOR and ZC3H12A, which, when unleashed, potently boosts T_PEX without impairing T_EX functionality, providing multiple novel targets for mitigating, and reversing T cell exhaustion in chronic viral infections.

C57BL/6 (Cat# JAX:000664, RRID:IMSR_JAX:000664), CD45.1 (Cat# JAX:002014, RRID:IMSR_JAX:002014), Cas9 transgenic mice (Cat# JAX:026430, RRID:IMSR_JAX:026430), P14 TCRVa2Vb8 transgenic mice (Cat# JAX: 037394, RRID:IMSR_JAX: 037394), Cd4^Cre mice (Cat# JAX: 022071, RRID:IMSR_JAX: 022071), Cd8^Cre mice (Cat# JAX: 008766, RRID:IMSR_JAX: 008766), and Rosa26-ERT2^Cre mice (Cat# JAX: 008463, RRID:IMSR_JAX: 008463) were originally from The Jackson Laboratory. Cd8^−/− mice were from H. Qi, Tsinghua University, Beijing, China. Zc3h12a flox mice (Strain NO. T010203) and Bcor flox mice (Strain NO. T022468) were purchased from GemPharmatech. Age (4–8 wk old) and sex-matched mice were used for experiments. All mouse studies were conducted in specific pathogen–free facilities at Tsinghua University’s Laboratory Animal Research Center (Beijing, China), accredited by the Beijing Administration Office of Laboratory Animal. Experimental protocols were approved by the Institutional Animal Care and Use Committee.

B16-GP33 melanoma, BHK-21 cell line, and VERO cell line were provided by Yuncai Liu’s lab. BHK-21, VERO, and Phoenix-ECO (Cat# CRL-3214, RRID: CVCL_H717; ATCC) were cultured in DMEM (Gibco) containing 10% FBS, 2 mM glutamine, 100 U/ml penicillin, and 100 mg/ml streptomycin in the incubator. All cell lines were tested for mycoplasma by the TransDect PCR Mycoplasma detection Kit (TRAN, FM311) and were confirmed to be negative.

The LCMV Armstrong and clone 13 strains were gifts from Yuncai Liu’s lab at Tsinghua University, Beijing, China. Both strains were propagated in BHK-21 cells and titrated by plaque assay on VERO cells. Mice were infected either intraperitoneally with Armstrong (2 × 10⁵ PFUs) or intravenously with clone 13 (2 × 10⁶ PFUs). Mice infected with LCMV were housed in accordance with the institutional biosafety regulations of Tsinghua University. The viral loads of LCMV in serum, organs, and tissue samples were quantified by using a quantitative PCR (qPCR) assay (McCausland and Crotty, 2008).

For CRISPR-mediated gene knockout, indicated sgRNA was cloned into pMSCV-sgRNA–Thy1.1 backbone, as previously reported (Zhao et al., 2021). All indicated sgRNA were listed here: sgNontargeting (sgControl), sgZc3h12a, sgBcor, sgTcf7, sgPou2f2, sgNr4a1, sgPdzk1ip1, sgIfit3, sgPim2, and sgLtb. For double knockout of Bcor and Zc3h12a, a mouse-U6–sgZc3h12a cassette was inserted downstream of the gRNA scaffold in the pMSCV-sgBcor–Thy1.1 vector. For triple gene knockout, we replaced the Thy1.1 marker in the pMSCV-sgRNA–Thy1.1 vector with GFP. T cells were then co-infected with the pMSCV-sgBcor/Zc3h12a–Thy1.1 virus and pMSCV-sgGene–GFP virus. Cells that were GFP⁺Thy1.1⁺ expressed three sgRNAs for the triple gene knockout.

For overexpressing mouse TCF1 (p45), POU2F2, PDZK1IP1, PIM2, LTB, IFIT3, and NR4A1, cDNAs of these genes were cloned into pMIG-IRES–GFP vector, and empty vector was used as control.

Retrovirus were packaged by co-transfection of Phoenix-Eco cells with indicated plasmid and helper plasmid pCL-Eco (#12371; Addgene) using ChemifectTM (Fengrbio). The viral supernatant was collected at 48 and 72 h after transfection, filtered via 0.45-µM filters, aliquoted, and frozen at −80°C.

Mouse primary T cells were cultured in T cell medium: RPMI1640 medium (Gibco) supplemented with 5% FBS, 2 mM glutamine, 55 µM β-mercaptoethanol, 1 mM sodium pyruvate, 100 U/ml penicillin, 100 µg/ml streptomycin, and 2 ng/ml IL-2 (Cat# 200-02–1000; PeproTech) in a humidified incubator at 37°C with 5% CO₂.

Single-cell suspension was prepared from spleen and lymph nodes of Cas9⁺ P14 mice or P14 mice. P14 cells were activated for ∼24 h with gp33–41 peptide (Cat#04010023714; ChinaPeptides), and viral transduction was performed by spin infection with 2,000 g at 33°C for 2 h in the presence of 16 µg/ml polybrene (Cat# H9268; Sigma-Aldrich), followed by incubation for another 4 h. Then, cells were washed and cultured in fresh T cell medium with IL-2. 24 h after spin infection, transduced efficiency was determined by examination of reporter (Thy1.1 or GFP)-positive cells by flow cytometry.

For the transduced P14 cells, 24 h after virus transduction, 0.1 million of GFP⁺ or Thy1.1⁺ P14 cells were transferred intravenously into recipient mice that were infected with LCMV Armstrong or clone 13 1 day (or indicated days) before cell transfer. For naive P14 cells, naïve ERT2^CreP14 or ERT2^CreBcor^fl/flZc3h12a^fl/fl P14 cells were FACS sorted from lymph node of donor mice and a total of 5,000 cells were transferred into 5–8-wk-old recipient mice intravenously 1 day before clone 13 infection. For secondary transfer, ∼0.01 million of P14 cells sorted from spleen of primary recipients were transferred into second recipient mice via tail vein, and mice were infected with clone 13 1 day later.

For inducible deletion of Bcor and Zc3h12a in T cells, naïve ERT2^CreP14 or ERT2^CreBcor^fl/flZc3h12a^fl/fl P14 cells were transferred into recipient mice as described above. After 3 wk, tamoxifen (dissolved in coin oil) was intraperitoneally injected to recipient mice at a dose of 1.5 mg per mouse for 5 consecutive days.

B6 mice were subcutaneously inoculated with 0.5 million B16-GP33 tumor cells. On day 7 after inoculation, 0.5 million activated Cas9⁺ P14 cells, infected with either sgControl or sgZc3h12a, were transferred into the tumor-bearing mice. Tumor size was measured every 3 days beginning on day 7 after inoculation using digital calipers. Tumor size was calculated as length × width.

For tumor-infiltrating lymphocyte analysis, tumor tissues were isolated, finely minced using surgical scissors, and digested for 1 h at 37°C in RPMI 1640 medium supplemented with 100 U/ml DNase I (Sigma-Aldrich) and 0.2 mg/ml collagenase IV (Sigma-Aldrich). The samples were mechanically disrupted by pressing through a 70-μm filter and washed with PBS. RBCs were lysed with 1 × RBC lysis buffer for 3 min, and the samples were passed through a 70-μm strainer again. Tumor-infiltrating T cells were then analyzed by FACS.

Single-cell suspension was prepared from blood or indicated organs. Cell surface proteins were stained with indicated antibodies in the presence of Fc block in FACS buffer (PBS containing 1% FBS, 2 mM EDTA, 100 U/ml penicillin, and 100 µg/ml streptomycin) at 4°C for 15 min. Intracellular staining for cytoplasmic and nuclear proteins were performed with Cytofix/Cytoperm Fixation/Permeabilization Solution Kit (BD Bioscience) and Transcription Factor Staining Buffer kit (BD Pharmingen) according to the manufacturer’s instructions. Dead cells were excluded by DAPI (BioLegend) staining or LIVE/DEAD Fixable Near-IR Dead Cell Stain Kit (Invitrogen). Antibodies for flow cytometry were purchased from BD Pharmingen, BioLegend, or Invitrogen, as listed with RRID. Annexin V staining was performed with Annexin V-FITC Apoptosis Detection Kit (Invitrogen) according to the manufacturer’s instructions. Samples were analyzed by a LSR Fortessa cytometer (BD). Flow cytometry data analysis was performed by Flowjo software (https://www.flowjo.com). Cell sorting was performed on an S3e cell sorter (Bio-Rad). All antibodies used in this study were listed in Table S1.

For intracellular cytokine production analysis, splenocytes were stimulated by the 0.2 µg/ml GP33–41 (amino acid sequence: KAVYNFATC) or GP276–286 (amino acid sequence: SGVENPGGYCL) peptides (ChinaPeptides) and brefeldin A for 5 h at 37°C. IFNγ expression was examined by intracellular staining.

Control, BCOR-deficient, ZC3H12A-deficient, and BCOR/ZC3H12A-deficient P14 cells were transferred into Cas9⁺ B6 mice that were infected with LCMV clone 13 1 day before cell transfer. 14 days after transfer, Thy1.1⁺ or GFP⁺ P14 cells from spleen were sorted by an S3e cell sorter with purity >98%. To generate bulk RNA-seq data, 1000 P14 cells were sorted in Buffer TCL (1031576; Qiagen), and RNA was purified by 2.2 × VAHTS RNA Clean Beads (N412-03; Vazyme). The cDNA libraries were generated following a modified SMART-seq2 protocol (Habib et al., 2016; Picelli et al., 2014). Briefly, RNA was eluted into 4 μl elution mix containing 1 μl reverse transcription (RT) primer (10 µM), 1 μl dNTP mix (10 mM each, N0447L; New England Biolabs), 1 μl RNase inhibitor (4 U/μl), and 1 μl H₂O. Eluted samples were incubated at 72°C for 3 min and immediately placed on ice. Each sample was added with 7 μl RT mix containing 0.75 μl H₂O, 0.1 μl Maxima Reverse Transcriptase (EP0741; Thermo Fisher Scientific), 2 μl 5 × RT buffer (EP0741; Thermo Fisher Scientific), 2 μl betaine (5 M, B0300; Sigma-Aldrich), 0.9 μl MgCl₂ (100 mM), 1 μl TSO primer (10 µM), and 0.25 μl RNase inhibitor (40 U/μl). The RT reaction was incubated at 42°C for 90 min, followed by 10 cycles of (50°C for 2 min, 42°C for 2 min), and heat inactivated at 70°C for 15 min. Samples were then amplified with an addition of 14 μl PCR mix containing 1 μl H₂O, 0.5 μl ISPCR primer (10 µM), and 12.5 μl 2 × KAPA HiFi HotStart ReadyMix (KK2602; KAPA Biosystems). The PCR reaction was performed as follows: 98°C for 3 min, 17 cycles of (98°C for 15 s, 67°C for 20 s, 72°C for 6 min), and final extension at 72°C for 5 min. The amplified cDNA product was purified using 0.8 × VAHTS DNA Clean Beads. Sequencing libraries were prepared using TruePrep DNA Library Prep Kit V2 for Illumina (TD503-02; Vazyme). Primer sequences: RT primer, 5′Biotin-AAGCAGTGGTATCAACGCAGAGTACTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTVN-3′; TSO primer, 5′Biotin-AAGCAGTGGTATCAACGCAGAGTACAT/rG//rG//iXNA_G/-3′; and ISPCR primer, 5′Biotin-AAGCAGTGGTATCAACGCAGA*G*T-3′.

Raw reads were trimmed by cutadapt (version 1.18) with parameter -j 3 -m 20 -e 0.1 -O 3 -q 20 --quality-base = 33 to remove low-quality bases and adaptor sequences. The ribosomal RNA (rRNA) reads were detected as those aligned to ribosomal genes using bowtie2 (version 2.3.3.1). The alignment ratio of rRNA was determined as ratio of the number of uniquely mapped reads to rRNA of all reads. For integrative genome browser visualization, sequencing reads were aligned to mm10 using STAR with parameter --outSAMtype BAM SortedByCoordinate --twopassMode Basic --limitBAMsortRAM 30000000000. The sequencing reads were trimmed by cutadapt (version 1.18) with parameter -m 20 -e 0.1 -O 3 -q 20 --quality-base = 33 to remove low-quality bases and adaptor sequences. The trimmed reads were aligned to mouse reference genome (mm10) using STAR (version 2.5.3a) (Dobin et al., 2013) with parameters --runThreadN 8 --readFilesIn --readFilesCommand gunzip -c –outSAMtype BAM SortedByCoordinate --twopassMode Basic --limitBAMsortRAM 30000000000 --outSAMstrandField intronMotif --quantMode TranscriptomeSAM. Expression matrix was generated by RSEM (version 1.3.0) (Li and Dewey, 2011). Deferential expression analysis was executed by DEseq2 (version 1.34.0) (Love et al., 2014). Log-transformed RNA expression changes were calculated by DEseq2 and used to process GSEA performed by ClusterProfiler (version 4.2.2) (Yu et al., 2012). Gene signature of terminal exhausted, progenitor-like, and memory precursor were downloaded from a published paper (Tsui et al., 2022). GSEA results and heatmap were plotted by ggplot2 (3.3.6).

Control, BCOR-deficient, ZC3H12A-deficient, and BCOR/ZC3H12A-deficient P14 cells were transferred into Cas9⁺ B6 mice that were infected with LCMV clone 13 1 day before cell transfer. 14 days after transfer, Thy1.1⁺ or GFP⁺ P14 cells from spleen were sorted by an S3e cell sorter with purity >98%. Cells were resuspended and washed with ice cold PBS for two times before permeabilization and Tn5 tagmentation. To permeabilize cells, the cell pellets were resuspended in 50 μl of ATAC-seq resuspension buffer (RSB; 10 mM Tris-HCl pH 7.4, 10 mM NaCl and 3 mM MgCl₂ in water) containing 0.1% IGEPAL-630, 0.1% Tween-20, and 0.01% digitonin and incubated on ice for 10 min. Afterward, 1 ml of ATAC-seq RSB containing 0.1% Tween-20 (without IGEPAL-630 or digitonin) was added, and the samples mixed by inverting the containing tubes for six times. Nuclei were then centrifuged for 10 min at 1,000 r.c.f. in a pre-chilled (4°C) fixed-angle centrifuge. Supernatant was removed, nuclei were incubated with Tn5 transposases and tagmentation buffer at 37°C for 30 min (N248; Novoprotein). After tagmentation, stop buffer was added directly into the reaction. Samples were incubated at 55°C for 30 min to terminate the tagmentation. Tagmentation products were purified by 2× VAHTS DNA Clean Beads. PCR was performed to amplify the library for five cycles using KAPA HiFi PCR Kit (kk2102; Kapa Biosystems) with the following PCR conditions: 75°C for 5 min, 98°C for 3 min, and thermocycling at 98°C for 20 s, 63°C for 30 s and 72°C for 3 min. The determine the optimal amplification for each sample, 5 µl PCR mix was diluted with 5 µl H₂O and combined with 2× SYBR qPCR supermix (E096; Novoprotein) to perform qPCR with following conditions: 98°C for 3 min and thermocycling for 30 cycles at 98°C for 20 s, 63°C for 30 s, and 72°C for 1 min. The additional number of PCR rounds (N) was determined as the qPCR cycle that reached the 1/3 maximum abundance on qPCR amplification curve. All samples were amplified for another N+3 cycles with the following PCR conditions: 98°C for 3 min; thermocycling at 98°C for 20 s, 63°C for 30 s, and 72°C for 3 min; and 72°C for 5 min for final extension. The library was sequenced on Illumina NovaSeq.

Sequencing reads were trimmed by cutadapt (version 1.18) with parameter -m 20 -e 0.1 -O 3 -q 20 --quality-base = 33 to remove low-quality bases and adaptor sequences. Trimmed reads were aligned to mouse reference genome (mm10) using bowtie2 aligner (version 2.3.3.1) (Langmead and Salzberg, 2012) with parameters –N1-L25 --no-mixed --no-discordant. Reads in blacklist regions defined in the ENCODE blacklist were excluded from downstream analysis. Duplicates were labeled by Picard MarkDuplicates (version 2.20.4) and removed by samtools (version 1.6) view –F 1024. Technical replicates of each batch were merged, and Tn5 accessible sites were identified by MACS2 callpeak (version 2.1.2) (Zhang et al., 2008) with parameter –g mm -q 0.05 –nomodel --nolambda. PCA analysis was performed by diffbind (version 3.4.11), and PCA results were ploted by ggplot2. Differential accessible regions were determined by diffbind with the following criteria: P value <0.01 and a log-transformed fold change (FC) >0.5. To visualize the accessibility intensity change, differential accessible regions identified from these conditions (sgBcor/Zc3h12a versus sgControl, sgBcor/Zc3h12a versus sgBcor, sgBcor/Zc3h12a versus sgZc3h12a, sgBcor versus sgControl, and sgZc3h12a versus sgControl) were consolidated as a single input to deepTools multiBigwigSummary. Matrix of peak intensity was z-scored, and k-means clustered in R. ComplexHeatmap (version 2.10.0) were used to visualize accessibility intensity change. Heatmap of marker gene expression were generated by ggplot2. Integrative genome browser visualization was used to get snapshot of genomic loci of Pdzk1ip1, Lef1, Bcl2, Tox, Havcr2, and Pou2f2.

Control, BCOR-deficient, ZC3H12A-deficient, and BCOR/ZC3H12A-deficient P14 cells were transferred into Cas9⁺ B6 mice that were infected with LCMV clone 13 1 day before cell transfer. 14 days after transfer, Thy1.1⁺ or GFP⁺ P14 cells from spleen were sorted by an S3e cell sorter with purity >98%. Single-cell suspensions (2 × 10⁵ cells/ml) with PBS were loaded onto microwell chip using the Singleron Matrix Single Cell Processing System. Barcoding beads are subsequently collected from the microwell chip, followed by RT of the mRNA captured by the Barcoding beads, and to obtain cDNA and PCR amplification. The amplified cDNA is then fragmented and ligated with sequencing adaptors. The scRNA-seq libraries were constructed according to the protocol of the GEXSCOPE Single Cell RNA Library Kits (Singleron) (Dura et al., 2019). Individual libraries were diluted to 4 nM, pooled, and sequenced on Illumina novaseq 6000 with 150-bp paired-end reads.

Downstream analyses were performed in R using the Seurat (Butler et al., 2018) package. Cells in which <500 genes or >3,000 genes were detected and in which mitochondrially encoded transcripts constituted >5% of the total library were excluded from downstream analysis. Before following analysis, cells that were not αβT cells were identified by scaled expression of marker genes >0.5 (δγT cell: Trgv2 and Tcrg-C2. Antigen-presenting cells: Cd74. Machrophage: S100a8 and Lyz2. Treg: Foxp3) and excluded, yielding an expression matrix of 13,588 cells by 19,875 genes (overexpression experiments, including empty and Pou2f2) and 39,026 cells by 19,804 genes (knock-out experiments, including sgControl, sgBcor, sgZc3h12a, and sgBcor/Zc3h12a). Each gene expression measurement was normalized by total expression in the corresponding cell and multiplied by a scaling factor of 10,000. Top 2,000 variable genes were identified by FindVariableFeatures and used to execute PCA by RunPCA function of Seurat. Principal components were determined to be significant (P < 0.01) using the jackstraw method, and uniform manifold approximation and projection (UMAP) and clustering was performed on these significant principal components. Differential expression analysis was performed between each cluster and all other cells using a Wilcoxon rank-sum test. Differential expression genes (P value <0.01 and log₂ FC >0.25) between progenitor-exhausted T cells and terminal-exhausted T cells were extracted from published results and used as signature for T_PEX versus T_EX (Miller et al., 2019). Expression profiling by array of effector T cell and exhausted T cell were downloaded from NCBI-Gene Expression Omnibus (GSE9650) (Wherry et al., 2007). Downloaded data were normalized by Affy (version 1.72.0) and differential expression analysis were executed by limma (version 3.50.3). Differentially expressed genes with P value <0.01 and Log₂ FC > 0.25 were used as genetic signature of T_EFF versus T_EX. AddModuleScore of Seurat was used to estimate genetic signature on scRNA-seq data.

To integrate scRNA-seq data of Pou2f2 OE experiment (including empty and Pou2f2) and knock-out experiment (including sgControl, sgBcor, sgZc3h12a, and sgBcor/Zc3h12a), both scRNA-seq dataset were normalized by Seurat function SCTransform. Integration anchors between two dataset were identified by Seurat function FindTransferAnchors with parameter dims = 1:30, normalization.method = “SCT,” reference.reduction = “pca,” k.anchor = 10, k.filter = 10, and k.score = 20. After anchors identification, Pou2f2 OE scRNA-seq data were projected to knock-out experiment scRNA-seq UMAP by function MapQuery with parameter transferdata.args = list(k.weight = 10) and integrateembeddings.args = list(k.weight = 10).

Control, BCOR-deficient, ZC3H12A-deficient, and BCOR/ZC3H12A-deficient P14 cells were transferred into Cas9⁺ B6 mice that were infected with LCMV clone 13 1 day before cell transfer. 14 days after transfer, Thy1.1⁺ P14 cells from spleen were sorted by an S3e cell sorter with purity >98% and sent to Novogene for scATAC-seq. scATAC-seq libraries were generated using the 10x Genomics Chromium Cell ATAC Reagent Kit (v1). In brief, sorted CD8⁺ P14 T cells were washed with 0.04% BSA PBS, and then ∼40,000 cells were subjected to the nuclei preparation protocol according to the manufacturer’s instructions. Then, 16,000 nuclei were loaded into a 10x chromium controller. All downstream library preparation steps were performed according to the manufacturer’s instructions. Libraries were sequenced on an Illumina NovaSeq.

For scATAC-seq data analysis, raw fastq data were aligned to mm10 by cellranger-atac count (Satpathy et al., 2019). Data clustering and dimensionality reduction were performed using the R package Siganc following its recommended procedure (Stuart et al., 2021). A unified set of peaks from cellranger–atac results was created to quantify each dataset. Cells with less ATAC counts than 500, more ATAC counts than 50,000, less in-peak reads ratio than 50%, more in-blacklist reads ratio than 5%, more nucleosome signal than 2, or less transcriptional start site (TSS) enrichment fold than 2.5 were excluded from downstream analysis, yielding a matrix of 24,514 cells by 104,107 features (sgControl: 5721, sgBcor: 6163, sgZc3h12a: 6699, and sgBcor/Zc3h12a: 5931). Term frequency-inverse document frequency normalization was performed by function RunTFIDF of Signac. Most frequently observed features were identified by function FindTopFeatures of Signac. Singular value decomposition was run on the term frequency inverse document frequency (TF-IDF) matrix by function RunSVD of Signac. The first seven latent semantic indexing (LSI) components will be used for subsequent data analysis, except for the first and third LSI components, as they are highly correlated with sequencing depth. UMAP dimension reduction and clustering were performed by signac function RunUMAP, FindNeighbors, and FindClusters. To analyze motif activity of different TFs, we ran signac function AddMotifs to add the DNA sequence motif information from JASPAR 2016 dataset (Mathelier et al., 2016). Motif activity was calculated by signac function RunChromVAR.

The statistical information of each experiment, including the statistical methods, the P value, and sample numbers (n) are shown in figure or figure legends. GraphPad Prism 8 was used to plot all graphs and to perform statistical and quantitative assessments. Error bars represent SEM. All experiments were repeated independently at least twice with similar results. Representative flow plots and micrographs were selected from biological replicates.

Fig. S1 shows the method for generating control, BCOR-deficient, ZC3H12A-deficient, and BCOR/ZC3H12A-deficient P14 cells. It also shows the reversal of the cell expansion advantage in BCOR/ZC3H12A-deficient P14 cells through ZC3H12A overexpression. Additionally, ZC3H12A overexpression in P14 cells enhances their antitumor effect. Fig. S2 shows the generation and characterization of mice with T cell–specific deletions of BCOR and ZC3H12A. Fig. S3 shows that the inducible deletion of BCOR and ZC3H12A reinvigorates TPEX during LCMV chronic infection. It also includes the bulk RNA-seq analysis of P14 cells. Fig. S4 shows the bulk ATAC-seq analysis of P14 cells. Fig. S5 shows the scRNA-seq analysis of P14 cells. Table S1 lists all the antibodies used in this study. Table S2 lists all the primers used in this study.

Raw and processed RNA-seq and ATAC-seq data are available in the Gene Expression Omnibus under accession numbers GSE285187 and GSE285186. All other data are presented in the figures and supplementary materials. Reagents generated in this study are available upon request.

We thank Institute for Immunology at Tsinghua University for providing and maintaining equipment. We thank professor Yan Shi (Tsinghua University, Beijing, China) for providing P14 mice, professor Hai Qi for providing Cd8^−/− mice, and professor Yuncai Liu (Tsinghua University, Beijing, China) for providing LCMV Armstrong and clone 13.

This research was supported by the National Natural Science Foundation of China (82350108 and T2495270 to M. Peng), Tsinghua University DUSHI Program (52302102323 to M. Peng), Tsinghua-Peking Center for Life Sciences (to M. Peng), SXMU-Tsinghua Collaborative Innovation Center for Frontier Medicine (to M. Peng). The work in the laboratory of Y. Li was supported by the National Key R&D Program of China 2019YFA0904402 and 2019YFA0906700, Beijing Natural Science Foundation Z210010, the National Natural Science Foundation of China 32171448, and Tsinghua University Initiative Scientific Research Program 2021Z11JCQ020 and 2022Z11QYJ032.

Author contributions: J. Xu: formal analysis, investigation, methodology, project administration, resources, validation, visualization, and writing—original draft, review, and editing. Z. Jia: data curation, formal analysis, investigation, methodology, project administration, resources, software, visualization, and writing—review and editing. X. Zhao: formal analysis, investigation, methodology, project administration, validation, and visualization. L. Wang: methodology. G. Jin: methodology and resources. Z. Li: resources. N. Yin: supervision and writing—review and editing. Y. Li: data curation, formal analysis, funding acquisition, investigation, methodology, project administration, resources, software, supervision, validation, visualization, and writing—review and editing. M. Peng: conceptualization, data curation, formal analysis, funding acquisition, methodology, project administration, resources, supervision, visualization, and writing—original draft, review, and editing.