HDAC1 controls the generation and maintenance of effector-like CD8⁺ T cells during chronic viral infection

During chronic infection and cancer, CD8⁺ T cells display progressive loss of effector function accompanied by enhanced expression of inhibitory receptors and poor memory recall responses (Blank et al., 2019; McLane et al., 2019). These functional impairments are widely known as T cell exhaustion and are postulated as an evolutionarily conserved adaptation mechanism to persistent antigen stimulation, thereby limiting immunopathology or autoreactivity (Kallies et al., 2020; Speiser et al., 2014). Despite their altered state, exhausted T (Tex) cells still provide some protection against viral replication or tumor growth (Jin et al., 1999; Johnston et al., 2014; Schmitz et al., 1999). It is now evident that T cell exhaustion is a unique state of T cell differentiation, which is epigenetically distinct from effector T cells and memory T cells generated upon acute infection or vaccination (Belk et al., 2022; Zebley et al., 2020). It is therefore essential to elucidate the role of epigenetic regulators governing T cell exhaustion.

Tex cells are a heterogeneous population composed of functionally distinct subsets, which reflects the complex state and regulation of T cell exhaustion (Chung et al., 2021; McLane et al., 2019; Sandu and Oxenius, 2023; Zander and Cui, 2023). Tex cells are compartmentalized into at least two major subsets: a self-renewing progenitor subset (Tex^prog subset), defined by high expression levels of the transcription factor (TF) T cell factor 1 (TCF1) as well as the surface markers Ly108 and CXCR5, and a terminally exhausted TCF1^lo subset (Tex^term), which is continuously replenished by Tex^prog cells (He et al., 2016; Im et al., 2016; Utzschneider et al., 2016; Wu et al., 2016). Importantly, Tex^prog cells display a proliferation burst upon immune checkpoint blockades, making them an attractive and promising target for immunotherapies (Kurtulus et al., 2019; Miller et al., 2019; Sade-Feldman et al., 2018). Subsequent studies identified additional Tex subsets, and several groups have independently identified a CX3CR1⁺ effector-like subset. This population, designated Tex^eff-like subset, retains a certain degree of cytolytic activity and thereby plays a central role in controlling chronic viral infection and tumor growth (Beltra et al., 2020; Hudson et al., 2019; Raju et al., 2021; Zander et al., 2019). Of note, immune checkpoint blockades lead to the expansion of CX3CR1⁺ cells and correlate with the better prognosis for patients with metastatic melanoma and non-small cell lung cancer (Yamauchi et al., 2021; Yan et al., 2018). Tex^eff-like cells display a unique transcriptional and epigenetic profile distinguishing them from Tex^prog and Tex^term subsets and, similar to Tex^term cells, are derived from Tex^prog cells. However, although several differentiation trajectories between Tex^eff-like and Tex^term cells have been postulated (Daniel et al., 2022; Giles et al., 2022; Kasmani et al., 2023), the exact developmental paths remain to be clarified. Since Tex^eff-like cells display a unique epigenetic profile (Chen et al., 2021a;,Daniel et al., 2022; Giles et al., 2022), a better understanding of epigenetic pathways controlling their differentiation might reveal novel therapeutic approaches for chronic infection and cancer (Kanev and Zehn, 2021; Zander and Cui, 2023).

One of the best-described epigenetic modifications is reversible acetylation of lysine residues on histones, mediated by histone acetyltransferases and histone deacetylases (HDACs). The HDAC family consists of 18 members, and several HDACs have important functions in T cells (Ellmeier and Seiser, 2018). HDAC1 in T cells is a key driver of autoimmune diseases. Deletion of HDAC1 in T cells (using Cd4-Cre; Hdac1^fl/fl x Cd4-Cre, designated as HDAC1-cKO) conveys resistance to experimental autoimmune encephalomyelitis (Göschl et al., 2018; Hamminger et al., 2021), collagen-induced arthritis (Göschl et al., 2020), and adoptive T cell transfer colitis (Hamminger et al., 2021), while HDAC1 restrains the pathogenic Th2 cells in allergic airways inflammation (Grausenburger et al., 2010; Khan et al., 2025). HDAC1 is also essential for efficient in vivo expansion and activation of CD8⁺ T cells in response to acute lymphocytic choriomeningitis virus (LCMV) infection (Tschismarov et al., 2014), suggesting a role for HDAC1 in antiviral immunity. However, the role of HDAC1 in the differentiation of Tex subsets remains unknown.

Here, we elucidated the role of HDAC1 in CD8⁺ T cell exhaustion. By using a well-established murine model of chronic viral infection, we observed that HDAC1 regulates the generation and maintenance of the CX3CR1⁺ Tex^eff-like cell subset in a T cell–intrinsic manner. Moreover, HDAC1 was essential for controlling viral load. Single-cell RNA sequencing (scRNA-seq) revealed that HDAC1 deletion deviates the differentiation of Tex^prog cells into an alternative cell subset enriched in exhaustion and cytolytic signatures. Mechanistically, as revealed by assay for transposase accessible chromatin with sequencing (ATAC-seq) and cleavage under targets and release using nuclease (CUT&RUN), HDAC1 promoted the generation of Tex^eff-like cell subsets by binding to and opening effector-like signature gene loci in Tex^prog cells to prime their expression. Together, our study demonstrates a specific function of HDAC1 in driving CX3CR1⁺ Tex^eff-like cell subset differentiation, which is essential for viral control during chronic infection. Thus, targeting and modulating HDAC1-regulated pathways in CD8⁺ T cells might be an exciting therapeutic strategy for controlling chronic viral infection.

To dissect the role of HDAC1 in T cell exhaustion, we induced chronic infection using the LCMV model in T cell–specific HDAC1-deficient (Hdac1^f/f, Cd4-Cre) and corresponding WT control (Hdac1^f/f) mice (hereafter referred to as HDAC1-cKO and WT mice, respectively) (Grausenburger et al., 2010). Specifically, we infected HDAC1-cKO and WT mice with the clone 13 strain of LCMV (LCMV Cl13) (Fig. 1 A), which results in chronic viral infection accompanied by a temporal weight loss and hepatitis driven by CD8⁺ T cells (Baazim et al., 2019; Bergthaler et al., 2007; Doherty et al., 1993; Moskophidis et al., 1993; Zinkernagel et al., 1986). Upon infection, HDAC1-cKO mice displayed a similar degree of a transient weight loss over a period of 4 wk in comparison to WT mice (Fig. 1 B). However, alanine aminotransferase levels were reduced in the serum of HDAC1-cKO mice at day 30 post infection (p.i.), and aspartate aminotransferase levels showed a similar tendency. This suggests ameliorated liver inflammation in the absence of HDAC1 (Fig. 1 C). Moreover, an elevated viremia in the serum in the absence of HDAC1 indicated its essential role in T cells for controlling antiviral responses (Fig. 1 D).

Based on the expression of CX3CR1 and Ly108, three exhausted CD8⁺ T cell subsets have been defined in established chronic infection (>3 wk p.i.): Tex^prog cells (Ly108⁺CX3CR1⁻), Tex^eff-like cells (Ly108⁻CX3CR1⁺), and Tex^term cells (Ly108⁻CX3CR1⁻) (Zander et al., 2019). Despite no alteration in the numbers of splenocytes or total and viral glycoprotein 33–41-specific (GP33-tet⁺) CD8⁺ T cells in the absence of HDAC1 (Fig. 1, E and F), a detailed flow cytometry analysis of CD8⁺ T cells revealed that lack of HDAC1 led to a reduction in Tex^eff-like cell frequencies and numbers (Fig. 1, E, G, and H). This was concurrent with an expansion of the Tex^term subset (Fig. 1, E, G, and H). Moreover, given their vital role in virus control (Hudson et al., 2019; Zander et al., 2019), the reduction of Tex^eff-like cells is in line with the observed increase in viremia in HDAC1-cKO mice (Fig. 1 D). Together, these data indicate that HDAC1 is essential for regulating Tex subset distribution and controlling viral load.

The emergence of Tex cells, based on transcriptional and chromatin accessibility profiling, has been observed as early as day 8 p.i. (Daniel et al., 2022; Giles et al., 2022; Yao et al., 2019). Having uncovered that HDAC1 deletion resulted in a reduction of Tex^eff-like cells 30 days p.i., we next determined the time point during infection at which Tex^eff-like cell differentiation is affected by HDAC1 deficiency. Since Tex^term cells (defined as Ly108⁻CX3CR1⁻ cells) were increased in HDAC1-cKO mice (Fig. 1, E, G, and H), we monitored the appearance of Tim3^hiCD101⁺ Tex cells as an alternative approach of identifying Tex^term cells (Hudson et al., 2019). For this, we assessed the expansion kinetics of splenic CD8⁺ T cells as well as Tex subset distribution among virus-specific CD8⁺ T cells in LCMV Cl13–infected HDAC1-cKO and WT mice on day 8, 15, 22, and 29 p.i. (Fig. 2 A). This analysis revealed that lack of HDAC1 led to a reduction in the numbers of CD8⁺ and GP33-tet⁺ CD8⁺ T cells 8 days p.i. (Fig. 2, B–D). Moreover, whereas WT mice displayed a strong induction of CX3CR1⁺ cells among GP33-tet⁺ CD8⁺ T cells from day 8 p.i. on as previously reported (Daniel et al., 2022; Zander et al., 2019), the percentages and numbers of this subset were much lower in HDAC1-cKO mice (Fig. 2, E and F; and Fig. S1, A–C). Furthermore, the number of Tim3^hiCD101⁺ Tex cells, which emerged from day 15 p.i. in WT mice, was reduced at day 15 and 22 p.i. in the mutant mice (Fig. 2, G and H; and Fig. S1 D). Thus, HDAC1 deletion impaired the differentiation of Tex cells (characterized as CX3CR1⁺ or Tim3^hiCD101⁺ cells) from the onset of chronic viral infection. Ly108⁺CX3CR1⁻ cells acquire a certain degree of Tex^prog features as early as 8 days p.i. (Daniel et al., 2022; Giles et al., 2022; Utzschneider et al., 2020; Zander et al., 2019). Moreover, fate-mapping experiments demonstrated that CX3CR1⁺ Tex cells emerging around 8 days p.i. later give rise to a substantial proportion of a Tex^eff-like cell subset (Raju et al., 2021). Hence, we refer to these two subsets on day 8 as early Tex^prog cells and early Tex^eff-like cells, respectively, throughout the manuscript. However, given that virtually no Tex^term cells emerge on day 8 p.i. (Fig. 2, G and H; and Fig. S1 D) (Daniel et al., 2022; Giles et al., 2022; Utzschneider et al., 2020; Zander et al., 2019), we termed the Ly108⁻CX3CR1⁻ subset, with no defined characteristics (nd), early Tex^nd cells.

Having revealed an essential role for HDAC1 in the generation of splenic early Tex^eff-like cells, we further characterized this subset at day 8 p.i. A comparable reduction of the early Tex^eff-like cell subset in HDAC1-deficient CD8⁺ T cells specific for the viral glycoprotein 276–286 (GP276-tet⁺) indicated that the altered differentiation of Tex subsets occurs independently of TCR specificity (Fig. S1, E–G). Moreover, a severely diminished early Tex^eff-like cell population was observed in other lymphoid (i.e., inguinal LNs) as well as nonlymphoid (i.e., liver) organs in HDAC1-cKO mice (Fig. S1, H and I), indicating phenotypic alterations of Tex subsets across many tissues. Finally, HDAC1-cKO mice displayed a higher serum viremia on day 8 p.i. when compared with WT mice (Fig. 2 I). This indicates that early Tex^eff-like cells play an essential role in controlling viral replication, similar to Tex^eff-like cells that emerge at later time points during infection.

In addition to the markers Ly108 and CX3CR1, Tex cells also display specific and distinct expression patterns of inhibitory receptors, effector molecules, and TFs (Beltra et al., 2020; Hudson et al., 2019; Zander et al., 2019). Additional immunophenotyping revealed that the expression levels of inhibitory receptors, such as Tim3 and 2B4, as well as of CD39, an ecto-ATPase linked with T cell exhaustion, were increased in HDAC1-cKO early Tex^eff-like and early Tex^nd cells (Fig. 3, A and B). This was accompanied by an elevated expression of thymocyte selection-associated HMG box (Tox), a central TF initiating T cell exhaustion (Fig. 3, A and B) (Alfei et al., 2019; Khan et al., 2019; Scott et al., 2019; Seo et al., 2019; Yao et al., 2019). Moreover, lack of HDAC1 resulted in elevated eomesodermin (Eomes) expression levels, concurrent with the downmodulation of T-box TF 21 (T-bet) expression (Fig. 3, A and B). Similar expression patterns of Eomes and T-bet are linked to terminal exhaustion (McLane et al., 2021; Zehn et al., 2022). Besides the increased expression of proteins linked to an exhaustion state, HDAC1-deficient Tex cells displayed enhanced expression of killer cell lectin-like receptor G1 (KLRG1), a canonical marker for effector T cells (Fig. 3, A and B) (Chen et al., 2019; Chung et al., 2021; Martin and Badovinac, 2018). In addition, while granzyme B (GzmB) expression was unchanged, perforin expression was higher in HDAC1-deficient early Tex^eff-like and early Tex^nd cells (Fig. 3, A and B), indicating that certain effector features are enhanced upon lack of HDAC1. To gain further insight into the impact of HDAC1 deletion on the heterogeneity and subset composition of Tex cells, we applied t-distributed Stochastic Neighbor Embedding (t-SNE) dimensionality reduction to our flow cytometry data. While a population corresponding to early Tex^eff-like cells was easily detectable in WT mice, deletion of HDAC1 led to an almost complete loss of this subset (Fig. S1 J, encircled purple population). Instead, a unique population (encircled black population) co-expressing inhibitory receptors and KLRG1 at high levels emerged in HDAC1-cKO mice (Fig. S1, J and K). Taken together, these results highlight an essential role of HDAC1 in promoting the generation of early Tex^eff-like cells.

The formation of Tex^eff-like cells requires help from CD4⁺ T cells that secrete IL-21 (Raju et al., 2021; Zander et al., 2019). Since Hdac1 is deleted both in CD4⁺ and CD8⁺ T cells in HDAC1-cKO mice (owing to the expression of Cre recombinase from the late double-negative thymocyte developmental stage onward in the Cd4-Cre deleter line) (Lee et al., 2001), alterations in the subset distribution of Tex cells might be due to a dysfunction of either CD4⁺, CD8⁺ T cells, or even both. In addition, it is possible that HDAC1 deletion leads to developmental defects in thymocytes that result in impaired early Tex^eff-like cell differentiation. Moreover, the increased viremia observed in HDAC1-cKO mice might feedback on early Tex^eff-like cell differentiation. To distinguish among these possibilities, we employed several experimental approaches. Firstly, we performed adoptive transfer experiments with CD8⁺ T cells that express the transgenic P14 TCR, recognizing the viral GP33 peptide presented by H2D^b (Pircher et al., 1989). We crossed WT and HDAC1-cKO mice with P14 TCR transgenic mice and additionally introduced a Rosa26-STOP-EYFP reporter allele (Srinivas et al., 2001) (hereafter referred to as P14-WT and P14-HDAC1-cKO^YFP mice, respectively) (Fig. 4 A). This allows the tracking of YFP⁺ P14-HDAC1-cKO^YFP T cells due to Cre-mediated deletion of the STOP cassette (Fig. 4 B). Naïve P14-WT (CD90.2⁺) and P14-HDAC1-cKO^YFP (CD90.2⁺YFP⁺) cells were mixed at a 1:1 ratio and adoptively co-transferred into WT CD90.1⁺ recipient mice (Fig. 4 C). On the following day, recipient mice were infected with LCMV Cl13, and the differentiation of Tex cells was analyzed 8 days p.i. (Fig. 4 A). Consistent with the reduced number of virus-specific CD8⁺ T cells observed in HDAC1-cKO mice (Fig. 2, C and D), the number of P14-HDAC1-cKO^YFP cells was reduced in this competitive setting (Fig. 4, C and D). Moreover, P14-HDAC1-cKO^YFP cells “phenocopied” the distribution of Tex subsets that has been observed in HDAC1-cKO mice (Fig. 2, E and F), showing reduced numbers of early Tex^eff-like cells (Fig. 4, E and F). Of note, additional co-transfer experiments using congenically marked CD45.1⁺ or CD45.2⁺ WT and HDAC1-deficient donor cells (instead of YFP^– and YFP⁺ cells), revealed the same alterations in the mutant cells (Fig. S2, A–E). This indicates that the phenotypic changes in HDAC1-cKO CD8⁺ T cells were not due to the potential immunogenicity of the YFP protein possibly resulting in rejection of certain cell subsets. Secondly, we also generated mixed bone marrow (BM) chimeric mice, where WT (CD45.2⁺) or HDAC1-cKO (CD45.2⁺) BM cells were mixed at a 1:1 ratio with congenically distinguishable WT (CD45.1⁺) BM cells (Fig. S2 F). Following reconstitution of irradiated recipients (CD45.1⁺) and subsequent infection, we observed reduced frequencies of the early Tex^eff-like cell subset within the reconstituted CD45.2⁺ HDAC1-deficient but not CD45.2⁺ WT and CD45.1⁺ compartment (Fig. S2, G and H). This was congruent with results obtained from the adoptive T cell co-transfer experiments. Finally, to exclude that a potential developmental defect due to HDAC1 deletion contributed to changes in Tex subsets, we specifically deleted HDAC1 in mature CD8⁺ T cells and only after their activation. For this, we generated HDAC1-deficient mice on a Gzmb-Cre background and then introduced a Rosa26-STOP-YFP allele to monitor Cre activity (Hdac1^f/f, Gzmb-Cre, and Rosa26-STOP-YFP; hereafter referred to as HDAC1-cKO^Gzmb-YFP mice). Since Gzmb-Cre deletes in activated T cells (Jacob and Baltimore, 1999), any alterations seen in these mice are due to a role of HDAC1 during or after T cell activation and Tex cell differentiation. We infected HDAC1-cKO^Gzmb-YFP mice and WT^Gzmb-YFP control mice (i.e., Hdac1^+/+, Gzmb-Cre, and Rosa26-STOP-YFP) with LCMV Cl13 and immunophenotyped virus-specific CD8⁺ T cells 8 days p.i. (Fig. S2, I and J). Similar to our observation made in HDAC1-cKO mice, the CX3CR1⁺Ly108^– early Tex^eff-like subset was reduced in HDAC1-cKO^Gzmb-YFP mice (Fig. S2, I and J). Together, these results clearly indicate that HDAC1 controls the generation of early Tex^eff-like subsets in a CD8⁺ T cell–intrinsic manner and independently of antigen levels (i.e., viral titer).

Next, we characterized effector functions of P14-HDAC1-cKO^YFP cells 8 days p.i. in the adoptive T cell co-transfer setting. The examination of immunoregulatory and effector protein expression revealed that donor P14-HDAC1-cKO^YFP cells displayed essentially the same alterations as the ones observed on CD8⁺ T cells in infected HDAC1-cKO mice (Fig. 4, G and H). This underscores the CD8⁺ T cell–intrinsic role of HDAC1 for early Tex cell differentiation. Furthermore, the expression of IFN-γ and TNF-α was unchanged in P14-HDAC1-cKO^YFP cells with the exception of an increased proportion of IFN-γ⁺TNF-α^– cells within the severely diminished early Tex^eff-like cell population (Fig. S3, A–D). Finally, we directly assessed the cytolytic capacity of HDAC1-deficient Tex cells and performed an in vitro CTL assay with early non-Tex^prog cells (i.e., Tex cells other than early Tex^prog cells). We adoptively transferred P14-WT or P14-HDAC1-cKO^YFP cells into recipient mice, which were infected on the following day, and sorted early non-Tex^prog donor cells (i.e., Ly108⁻Tim3⁺ cells) 8 p.i. Subsequently, early non-Tex^prog cells were mixed with relevant or irrelevant peptide-pulsed EL-4 target cells and their specific killing assessed (Fig. S3 E). P14-HDAC1-cKO^YFP cells lysed target cells as efficiently as P14-WT cells, indicating intact cytolytic capacity of HDAC1-deficient early non-Tex^prog cells in vitro (Fig. S3, F and G). Thus, it is likely that the elevated viremia in the mutant mice can be attributed to a significant extent to the reduced numbers of early non-Tex^prog cells. However, it remains possible that HDAC1-deficient early non-Tex^prog cells (which largely consist of early Tex^nd cells) display impaired cytolytic activity in vivo and/or that other alterations in the mutant Tex cells lead to a reduced ability to clear the virus in vivo.

Upon the onset of chronic infection, viral-specific CD8⁺ T cells are activated, clonally expand, and differentiate into Tex cells (Chen et al., 2019; Daniel et al., 2022; Utzschneider et al., 2020). To define at which time point upon infection Tex subset composition is altered in the absence of HDAC1, we examined the kinetics of the appearance of CX3CR1⁺ cells at 48 and 67 h as well as day 5 and 8 p.i. and also determined the activation and expansion of early Tex cells. Using the aforementioned adoptive P14 T cell co-transfer model (Fig. 4 A), the ratio of transferred P14-WT to P14-HDAC1-cKO^YFP cells up to 67 h p.i. (Fig. 5, A and B) was unchanged. In addition, there was no difference in cell size (as revealed by forward scatter values) or expression levels of the early activation marker CD69 between the two groups at the onset of activation (Fig. S3, H and I). Furthermore, both P14-WT and P14-HDAC1-cKO^YFP cells divided at a similar rate based on the “dilution” of division-tracking dye CellTrace Violet (Fig. S3, H and I). These data indicate intact TCR signaling and T cell activation in the absence of HDAC1 at early time points. In contrast, from day 5 p.i. on, P14-HDAC1-cKO^YFP cells displayed a relative reduction compared with the P14-WT cells (Fig. 5, A and B). In addition, while the appearance of CX3CR1⁺ cells first became evident at day 5 p.i. in the WT compartment, the proportion of these cells was severely reduced in the absence of HDAC1 (Fig. 5, C and D). However, the deletion of HDAC1 had no impact on the degree of ongoing proliferation as well as the proportion of apoptotic cells (assessed by intracellular Ki67 and active caspase-3 staining, respectively) over the assessed time period of 8 days p.i. (Fig. S3, J and K). Together, these data indicate that HDAC1 does not regulate early activation, proliferation, or survival of cells during early Tex cell differentiation. Rather, HDAC1 might represent a checkpoint that instructs cell fate specification and differentiation toward early Tex^eff-like population as early as 5 days p.i.

Our data clearly demonstrate a key role of HDAC1 in the generation of early Tex^eff-like cells. Since previous studies have shown that the Tex^eff-like cell pool is maintained throughout chronic viral infection (Daniel et al., 2022; Zander et al., 2019), we next examined whether HDAC1 is also essential for the maintenance of Tex^eff-like cells. We crossed Hdac1^f/f mice onto a Rosa26-CreERT2 background (Hameyer et al., 2007) to inducibly delete Hdac1 by tamoxifen administration during the course of chronic viral infection. We also introduced a Rosa26-STOP-EYFP reporter allele to monitor Cre activity and thus cells that have deleted Hdac1, leading to the generation of Hdac1^f/f, Rosa26-CreERT2/STOP-YFP, and Hdac1^+/+, Rosa26-CreERT2/STOP-YFP mice (hereafter referred to as WT^CreERT-YFP and HDAC1-cKO^CreERT-YFP mice, respectively). In the absence of tamoxifen, YFP expression was not induced, and the distribution of circulating Tex cells was normal 8 days p.i., indicating that the genetic system is tightly controlled as expected (Fig. S3, L and M). Subsequently, mice were treated with tamoxifen on day 9 p.i., and the distribution of splenic Tex cells was analyzed 6 days later (i.e., 15 days p.i.) (Fig. 6 A). After tamoxifen injection, YFP expression was induced in ∼30% of CD8⁺ T cells. However, intracellular anti-HDAC1 staining revealed that YFP expression did not correlate with the deletion of Hdac1 and hence loss of HDAC1 expression, which was detected in ∼40% of CD8⁺ T cells (Fig. 6, B and C). Therefore, we eventually identified HDAC1-deficient cells in HDAC1-cKO^CreERT-YFP mice by “gating” on the HDAC1-negative (HDAC1⁻) population based on the intracellular anti-HDAC1 staining (i.e., cells in which HDAC1 expression was not detected due to deletion of the Hdac1 gene) (Fig. 6 B; red square). Strikingly, the analysis of the distribution of Tex cells revealed a reduction in the proportion of the CX3CR1⁺ cells within the HDAC1⁻ population in HDAC1-cKO^CreERT-YFP mice compared with WT^CreERT-YFP mice as well as to the HDAC1⁺ population of HDAC1-cKO^CreERT-YFP mice (Fig. 6, D and E). Taken together, our results demonstrate a CD8⁺ T cell–intrinsic key role for HDAC1 not only for the generation but also for the maintenance of the Tex^eff-like cell population during chronic viral infection.

To gain further insight into the impact of HDAC1 deletion on Tex subset diversity at early stages during chronic infection, we performed scRNA-seq of naïve and LCMV-specific (GP33-tet⁺) CD8⁺ T cells isolated from WT and HDAC1-cKO mice, encompassing uninfected and day 8 infected conditions, respectively. Uniform manifold approximation and projection followed by Seurat-based clustering led to the identification of eight clusters (Fig. 7, A and B; and Table S1), which were subsequently annotated based on their unique marker genes (Fig. 7 C) as well as their similarity to published signature genes of CD8⁺ T cell subsets upon acute or chronic LCMV infection (Fig. S4, A–C) (Daniel et al., 2022). We found two canonical clusters corresponding to naïve (cluster C1; T^Naïve) and early Tex^prog cells (C2; Tex^prog). Furthermore, two other relatively small clusters were identified that highly expressed genes related to either cell proliferation (e.g., Hist1h1b and Mki67) (C3; Tex^prol) or terminal exhaustion (e.g., Bcl2a1d and Lag3) (C4; Tex^exh) (Fig. 7, A–C and Fig. S4, A–C). The majority of cells (∼77% of WT and 67% of HDAC1-cKO cells) were grouped into four other clusters (C5–C8) (Fig. S4 B). Cluster C5 (Tex^early) was enriched for a gene signature of Tex^eeff cells (early effector exhausted cells where the Tex program is initiated [Daniel et al., 2022]) (Fig. S4, A and C). Cluster C6 (Tex^int) expressed the Cx3cr1 gene and was enriched for a gene signature of Tex^int cells (intermediate exhausted cells harboring potential to become both Tex^term and Tex^eff-like cells [Daniel et al., 2022]) (Fig. 7 C; and Fig. S4, A and C). Cluster C7 (Tex^eff-like) expressed the Cx3cr1 gene at a higher level than the C6 cluster and was enriched for a gene signature of Tex^KLR cells (KLR family member protein-expressing Tex cells, a major subpopulation of Tex^eff-like cells [Daniel et al., 2022]) (Fig. 7 C; and Fig. S4, A and C). Lastly, cluster C8 (Tex^cyt) highly expressed genes encoding cytolytic proteins but also displaying the highest enrichment score of a Tex^term gene signature (Daniel et al., 2022) (Fig. 7 C; and Fig. S4, A and C). Notably, while the C1–C4 clusters (T^Naïve, Tex^prog, Tex^prol, and Tex^exh) were almost equally composed of WT and HDAC1-cKO cells, cluster C6 (Tex^int) and cluster C7 (Tex^eff-like) contained mostly WT cells (Fig. 7, A and B; and Fig. S4 B). In contrast to the C6 and C7 clusters, the vast majority of the C5 (Tex^early) and C8 (Tex^cyt) clusters consisted of HDAC1-deficient cells (Fig. 7, A and B; and Fig. S4 B). A gene ontology (GO) analysis of clusters C5–C8 revealed that several biological processes (e.g., ribonucleoprotein complex biogenesis) were uniquely overrepresented in the HDAC1-cKO–dominant C5 cluster (Tex^early) (Fig. 7 D and Fig. S4 D). In addition, while WT-dominant clusters C6 and C7 shared almost identical enriched terms for GO biological processes, some of these processes (e.g., “leukocyte migration”) were not enriched in the HDAC1-cKO–dominant C5 and C8 (Fig. 7 D and Fig. S4 D). Instead, cluster C8 shared certain processes with C5 (e.g., “oxidation of organic compounds”) (Fig. 7 D and Fig. S4 D). This analysis suggests that the lack of HDAC1 leads to the emergence of transcriptionally distinct subsets within the early non-Tex^prog population (i.e., Tex cells other than early Tex^prog cells), in contrast to its minor impact on the generation of early Tex^prog cells. The specific role of HDAC1 for the formation of early non-Tex^prog cells was further supported by a trajectory inference analysis (Cao et al., 2019), which predicted that Tex^prog cells bifurcate into Tex^eff-like and Tex^cyt clusters, with a major branch point at the Tex^early cell cluster (Fig. 7 E). A comparison of gene expression between WT and HDAC1-cKO early non-Tex^prog cells showed that the lack of HDAC1 led to increased expression of genes associated with cytolytic function (e.g., Gzma, Gzmk, Prf1, and Klre1) and exhaustion (e.g., Havcr2, Tox, Tnfrsf9, and Eomes), concurrent with downmodulation of Cx3cr1 gene expression (Fig. 7 F). This is in line with our immunophenotyping of early Tex cells (Fig. 3, A and B), such as the simultaneous up- and downmodulation of KLRG1 and CX3CR1, respectively, in HDAC1-cKO early non-Tex^prog cells, whose expression levels are positively correlated during Tex cell differentiation in a HDAC1-sufficient setting (Beltra et al., 2020; Zander et al., 2019).

Finally, to investigate the impact of HDAC1 deletion on gene regulatory networks in early Tex^prog and early non-Tex^prog cells, we compared the activity of regulons (i.e., co-expression modules consisting of a TF and its putative direct targets within the same cell) between WT and HDAC1-deficient cells (Fig. S4 E). This analysis identified distinct sets of TFs whose regulon activities were elevated either in WT or in HDAC1-deficient Tex cells. Among these TFs, T-bet activity (encoded by the Tbx21 gene), which is essential for the generation of Tex^eff-like cells (Beltra et al., 2020; Raju et al., 2021; Zander et al., 2019), was upregulated in WT early non-Tex^prog cells. In addition, the regulon activity of Eomes was increased in HDAC1-cKO early Tex^prog and early non-Tex^prog cells. Of note, relative expression levels of Eomes and T-bet are linked with the progression of terminal exhaustion (Beltra et al., 2020; Zander et al., 2019). Since we observed increased (or a trend toward an increase) expression levels of T-bet and Eomes proteins in WT and HDAC1-cKO early Tex cells, respectively (Fig. 3, A and B; and Fig. 4, G and H), it is likely that HDAC1 controls early Tex^eff-like cell differentiation, at least in part, by regulating the expression and/or function of these two TFs. Thus, the deletion of HDAC1 resulted in specific alterations in transcriptional networks controlling early Tex cell differentiation and the enlargement of two Tex clusters with early exhaustion (C5) and exhaustion/cytolysis (C8) signatures.

Transcriptional changes during Tex subset differentiation are associated with dynamic shifts in chromatin accessibility (Beltra et al., 2020; Chen et al., 2021a). Given that our data show no alteration in the activation, proliferation and survival of early Tex cells (Fig. S3, H–K) as well as the intact generation of early Tex^prog cells lacking HDAC1 (Fig. 7, A and B), it is conceivable that HDAC1 controls the transition of early Tex^prog cells into early Tex^eff-like cells, at least in part, by mediating changes in the chromatin landscape. To examine this hypothesis, we conducted bulk ATAC-seq of P14-WT and P14-HDAC1-cKO early Tex^prog (Ly108⁺Tim3⁻) as well as early non-Tex^prog (Ly108⁻Tim3⁺) cells 8 days p.i. A comparison between WT and HDAC1-cKO cells revealed 319 and 2,179 differentially accessible regions (DARs) in early Tex^prog and early non-Tex^prog cells, respectively (Fig. 8, A and B; and Table S2). The vast majority of the DARs were located within introns, intergenic regions, or at promoters in both populations. Lack of HDAC1 generally led to increased numbers of “open” chromatin regions in both populations, in line with the known function of HDACs in promoting a “closed” chromatin structure (Fig. 8, A and B). However, 102 (at 97 gene loci) and 217 DARs (at 201 gene loci) were more open in WT early Tex^prog and early non-Tex^prog cells. We next explored whether there is a correlation between changes in chromatin accessibility and gene expression by integrating the data obtained from bulk ATAC-seq and scRNA-seq. We defined the gene sets harboring DARs either in WT or HDAC1-cKO cells and determined the overall average expression levels (i.e., the module scores) of these gene sets at single-cell level (Fig. 8 C and Fig. S5 A). Moreover, we calculated the average module scores within the individual clusters (i.e., C2–C8) and thereby examined whether DARs-associated genes are preferentially expressed in certain cluster(s) (Fig. 8 D and Fig. S5 B). Notably, this analysis revealed that the 97 genes with 102 DARs more accessible in WT early Tex^prog cells are highly expressed in WT-dominant cluster C6 (Tex^int) and C7 (Tex^eff-like), indicating a role for HDAC1 in priming the expression of early Tex^eff-like cell–associated genes by “opening” their loci in early Tex^prog cells (Fig. 8, C and D). In contrast, for the genes associated with DARs more accessible in HDAC1-deficient early Tex^prog cells, there was no enrichment to a particular cluster (Fig. 8, C and D). Overrepresentation analysis of GO terms showed that genes associated with DARs more accessible in WT early Tex^prog cells are also enriched in GO terms such as “leukocyte cell–cell adhesion” and “lymphocyte differentiation” (Fig. S5 C), similar to the single-cell transcriptome of WT-dominant clusters C6 and C7 (Fig. 7 D). Moreover, gene set enrichment analysis of DAR-associated genes revealed the overrepresentation of an effector T cell signature (i.e., a signature of total CD8⁺ T cells on day 8 p.i. with acute LCMV Armstrong) (Wherry et al., 2007) in WT cells in comparison to the HDAC1-cKO cells (Fig. 8 E). These results further support a role for HDAC1 in priming an early Tex^eff-like program. Of note, genes more accessible in WT or HDAC1-cKO early non-Tex^prog cells were highly expressed in WT- or HDAC1-cKO–dominant clusters (i.e., C6/C7 or C5/C8), respectively, indicating a correlation of accessibility and gene expression in early non-Tex^prog cells (Fig. S5, A and B). Finally, an enrichment analysis of TF-binding motifs revealed shared as well as unique TF-binding motifs enriched in DARs of WT and/or HDAC1-cKO cells, both in early Tex^prog cells as well as in the early non-Tex^prog cell population (Fig. 8 F and Fig. S5 D). This suggests that the same TFs, in particular members of the Runx family, might be recruited to different gene loci and that a different set of TFs might be active, dependent on the presence or absence of HDAC1. Together, the integration of our ATAC-seq analysis with our scRNA-seq data indicates that HDAC1 is essential for an open chromatin state at effector-like signature gene loci in early Tex^prog cells.

To investigate whether the appearance of open ATAC-seq regions correlates with the binding of HDAC1 and whether HDAC1 regulates histone acetylation levels at these gene loci, we performed CUT&RUN for HDAC1, H3K27ac, and H3K27me3 in WT and HDAC1-cKO early Tex cells. Moreover, since DARs that are more open in WT cells showed an enrichment of Runx-binding motifs, we also performed CUT&RUN for Runx3, which plays an essential role in the initiation of CD8⁺ T cell effector programs in acute (Shan et al., 2017; Wang et al., 2018) and chronic viral infections (Chen et al., 2021b; Shan et al., 2021). On a genome-wide level, a total of 42,013 HDAC1-binding sites at 16,276 gene loci in WT early Tex^prog cells and 44,414 HDAC1-binding sites at 15,404 gene loci in WT early non-Tex^prog cells were identified (Fig. 9 A and Fig. S5 E). The heatmaps of H3K27ac and H3K27me3 modifications revealed an inverse correlation, in agreement with the opposite functions of these two histone marks (Millán-Zambrano et al., 2022; Zhao and Garcia, 2015). We also observed a strong enrichment for Runx-binding motifs in the Runx3 CUT&RUN (Fig. 9 A; and Fig. S5, E and F). The deletion of HDAC1 had a minor impact on genome-wide H3K27ac sites in early Tex^prog cells, with 678 upregulated sites (on 571 gene loci) and 241 downregulated sites (on 189 gene loci) of a total of 11,866 sites (Fig. S5 G). 510 of the 571 gene loci were bound by HDAC1, suggesting a direct control of these H3K27ac sites by HDAC1. Although this is in line with the well-known functions of HDACs (i.e., removal of acetylation marks at histones), our data indicate that HDAC1 is not the main epigenetic eraser for H3K27ac marks in early Tex^prog cells. With respect to H3K27me3 modifications, deletion of HDAC1 also had only a minor impact with a down- and upregulation of 954 and 537 sites (of a total 18,979 sites), respectively (Fig. S5 G).

Next, we focused on the 102 ATAC-seq regions that are more open in WT early Tex^prog cells at the 97 gene loci that are later expressed in WT early non-Tex^prog cells (i.e., predominantly consisting of Tex^int and Tex^eff-like clusters). We observed HDAC1 binding at 67 of these 97 gene loci, indicating that ∼69% of these loci are controlled by HDAC1. The majority of HDAC1-binding sites (44 out of 67) overlapped with the open ATAC-seq regions (Fig. 9, B and C), showing an unexpected correlation of HDAC1 binding with open chromatin. On the contrary, from the 217 regions at 201 gene loci that are more open in HDAC1-cKO early Tex^prog cells compared with WT cells, only 75 genes (∼37%) showed HDAC1 binding. From the 75 binding sites, only 21 overlapped with open ATAC-seq regions. This indicates that the frequency of HDAC1 binding to gene loci that are more open in WT early Tex^prog cells than in HDAC1-cKO Tex^prog cells is much higher compared with the gene loci that are more open when HDAC1 is deleted. Furthermore, the appearance of the open 102 ATAC-seq regions in WT early Tex^prog did not lead to an overall increase in H3K27ac levels at these regions, while H3K27ac marks were increased at the 217 ATAC-seq regions that are open in HDAC1-cKO early Tex^prog cells (Fig. 9 D). Finally, we detected increased recruitment of Runx3 in WT early Tex^prog cells (compared with the corresponding HDAC1-cKO subset) at 17 of the 44 open chromatin regions bound by HDAC1 (Fig. 9, B and C; as exemplified by the Sytl2 gene). This suggests that approximately half of the effector-like genes, whose expression is primed by direct HDAC1 recruitment, might be co-regulated by HDAC1 and Runx3. Together, these data suggest an unexpected mode of action for HDAC1 and that HDAC1, in part together with Runx3, primes effector gene loci in Tex^prog cells (Fig. 10 A).

Here, we discovered that HDAC1 is an integral part of the differentiation program endowing Tex cells with effector-like characteristics during chronic viral infection. The molecular mechanisms that direct Tex^prog into Tex^eff-like or Tex^term cell subsets remain largely unknown (Kanev and Zehn, 2021; Zander and Cui, 2023). Based on our data showing a reduction of Tex^eff-like cells in the absence of HDAC1, HDAC1 might promote early Tex^prog toward early Tex^eff-like differentiation or suppress alternative early Tex^prog fates. The deletion of HDAC1 increased the accessibility of certain gene loci in early Tex^prog cells, as one might have expected from the well-known function of HDACs to promote a closed chromatin state (Seto and Yoshida, 2014; Yang and Seto, 2007). However, the integration of our ATAC-seq and scRNA-seq data did not show an enrichment in the expression of genes associated with these open loci in a particular subset, suggesting that HDAC1 does not restrain alternative fates of early Tex^prog cells. Unexpectedly, our analysis revealed that the deletion of HDAC1 resulted in a more closed chromatin state at gene loci in early Tex^prog cells that are later expressed in early Tex^eff-like cells. Since HDAC1 is recruited to many of these effector gene loci in WT early Tex^prog cells, this suggests a novel mode of action for HDAC1 in establishing an open chromatin state in early Tex^prog cells that facilitates differentiation toward early Tex^eff-like cells. Our study is in line with a recently published study (Hu et al., 2025) in which, by analyzing HDAC1-deficient (Gzmb-Cre) mice combined with scRNA-seq and ATAC-seq approaches, they showed that HDAC1 is an early determinant of intermediate-exhausted CD8⁺ T cell fate in chronic viral infection. Our observation that Runx3 is bound to many of the DARs that are open in WT early Tex^prog cells indicates that some of these gene loci might be co-regulated by HDAC1 and Runx3. Since Runx motifs were enriched in open DARs in both genotypes, this might also imply that Runx proteins might have different interaction partners and/or are recruited to different genomic loci in early Tex^prog cells dependent on the presence of HDAC1. Since members of the HDAC family target many nonhistone proteins (Choudhary et al., 2009; Ellmeier and Seiser, 2018), HDAC1 might also control the acetylation status and thus the functional properties of TFs key for the differentiation of Tex^eff-like cells. Our SCENIC analysis identified several regulons whose activities are increased in either WT or HDAC1-cKO Tex subsets. These regulons might represent interesting TFs to study whether their functions are regulated by HDAC1.

Seminal studies supported a linear model of Tex subset generation in which Tex^prog, Tex^eff-like, and Tex^term cell differentiation occurs sequentially (Beltra et al., 2020; Hudson et al., 2019). Subsequent studies using scRNA-seq combined with TCR-seq and trajectory analysis suggested a bifurcation model, where Tex^prog cells first differentiate into CX3CR1⁺ Tex^int cells, which then bifurcate into either Tex^eff-like (also known as Tex^KLR) or Tex^term cells (Daniel et al., 2022; Fagerberg et al., 2025; Kasmani et al., 2023). We identified a CX3CR1-expressing cluster enriched with a Tex^int signature, and our trajectory analysis indicated that this Tex^int cluster differentiates into the Tex^eff-like cluster. Since the generation of the Tex^int cluster was severely impaired in the absence of HDAC1, this might lead to a differentiation defect of the Tex^eff-like cluster. Moreover, since HDAC1-cKO Tex^term cells (i.e., Tim3^hiCD101⁺ cells) were reduced, it is conceivable that this alteration is also a consequence of impaired Tex^int cell subset differentiation. Therefore, we propose a model for HDAC1 function in which HDAC1 plays a key role for the transition of (early) Tex^prog to Tex^int cell subsets. HDAC1 deletion leads to a block in this transition, resulting in the expansion of the Tex^early cluster and their skewed differentiation into the Tex^cyt cluster. We also noticed that the Tex^prog cluster is separated into WT- and HDAC1-cKO–dominant subregions, suggesting the emergence of an early Tex^prog cell subset in the absence of HDAC1. It is therefore possible that these altered characteristics of early Tex^prog cells impact their later differentiation.

Our study has some limitations. Our data indicate an unexpected and novel role of HDAC1 in facilitating open chromatin at effector-like gene loci in early Tex^prog cells. However, HDACs in general not only function as epigenetic regulators, but reversible lysine acetylation also targets many nonhistone proteins. Thus, it will be important to address whether the effect observed on Tex^eff-like cells upon deletion of HDAC1 is (in part) due to changes in posttranslational modifications of factors driving Tex^eff-like differentiation. Moreover, while our data indicate a selective requirement of HDAC1 for Tex^eff-like cell subset differentiation, further studies are needed to elucidate how HDAC1 activity is modulated by multiple environmental cues to promote the formation of Tex^eff-like cell subset (Kanev and Zehn, 2021; Lan et al., 2023; Zander and Cui, 2023).

Pan-HDAC inhibitors are used for the treatment of certain types of cancers, and many show promising results in preclinical models for T cell–mediated and neurological diseases (Falkenberg and Johnstone, 2014). However, based on our findings, a careful assessment is required to evaluate the impact of such inhibitors, as they might impair the generation of Tex^eff-like subset. Conversely, our study suggests that the modulation of HDAC1 activity and the activation of HDAC1-controlled pathways might serve as an attractive strategy to improve the outcome of T cell–based immunotherapies by promoting the generation of Tex^eff-like cells.

Floxed Hdac1 (MGI:4440556) mice were kindly provided by Patrick Matthias (Friedrich Miescher Institute for Biomedical Research, Basel, Switzerland) and Christian Seiser (Medical University of Vienna, Vienna, Austria). Cd4-Cre mice (MGI:2386448) were kindly provided by Chris Wilson (University of Washington, Seattle, WA, USA). Hdac1^fl/fl and Cd4-Cre (HDAC1-cKO) mice were previously described (Grausenburger et al., 2010). Rosa26-STOP-YFP (MGI:2449038) and P14 T cell receptor transgenic (MGI:2665105) mice were kindly provided by Meinrad Busslinger (Research Institute of Molecular Pathology, Vienna, Austria) and Annette Oxenius (ETH Zurich, Zurich, Switzerland), respectively, and were crossed to HDAC1-cKO mice to generate either P14, Hdac1^fl/fl, Cd4-Cre or Rosa26-STOP-YFP, P14, Hdac1^fl/fl, and Cd4-Cre (P14-HDAC1-cKO or P14-HDAC1-cKO^YFP, respectively) mice. Rosa26-CreERT2 mice (MGI:3699244) were kindly provided by Emilio Casanova (Medical University of Vienna, Vienna, Austria) and were crossed to Hdac1^fl/fl mice on a Rosa26-STOP-YFP background to generate Rosa26-STOP-YFP, Hdac1^fl/fl, and Rosa26-CreERT2 (HDAC1-cKO^CreERT-YFP) mice. CD45.1 (MGI:4819849) and CD90.1 (MGI:3579311) congenic mice were kindly provided by Jochen Hühn (Helmholtz Center, Braunschweig, Germany) and Annette Oxenius (ETH Zurich, Zurich, Switzerland), respectively. GzmB-Cre (MGI:2176191) mice were kindly provided by Susan Kaech (NOMIS Center for Immunobiology and Microbial Pathogenesis, Salk Institute, La Jolla, CA, USA) and were crossed to Hdac1^fl/fl mice on a Rosa26-STOP-YFP background to generate Rosa26-STOP-YFP, Hdac1^fl/fl, and GzmB-Cre (HDAC1-cKO^Gzmb-YFP). Mice were analyzed at 8–12 wk of age and were maintained in the Core Facility Laboratory Animal Breeding and Husbandry of the Medical University of Vienna. Animal experiments were assessed by the Ethics Committee of the Medical University of Vienna and approved by the Austrian Federal Ministry for Education, Science and Research (animal protocol number: BMBWF-66.009/0040-V/3b/2019, BMBWF-66.009/0343-V/3b/2019). Animal husbandry and experiments were performed in compliance with national laws and according to the Federation of European Laboratory Animal Science Association and the Animal Research: Reporting of In Vivo Experiments guidelines. Genotyping PCR primers are listed in Table S3.

LCMV Cl13 was propagated in BHK-21 cells (ATCC No. CCL-10) as described (Bergthaler et al., 2010; Welsh and Seedhom, 2008). Mice were infected i.v. through the tail vein with 2 × 10⁶ focus-forming units (FFUs) of LCMV Cl13. Mice were euthanized 8, 15, 22, 29, or 30 days after infection for further analysis. Viral titers were measured using standard immunofocus assays (Battegay et al., 1991) with minor modifications. Briefly, Vero cells (ATCC No. CCL-81) were incubated for 1 h with blood serum collected from infected mice. Afterward, the cells were covered for 48 h with 1x 199 media (Gibco) containing 0.5% medium electroendosmosis agarose (Biozym), 10% FBS (Biowest), 2 mM glutamine (Sigma-Aldrich), and 100 U/ml penicillin and streptomycin (Sigma-Aldrich). Cells were fixed with 25% formaldehyde (Sigma-Aldrich) in PBS (Sigma-Aldrich), permeabilized with 1% Triton X-100/PBS (Promega), and blocked with 5% FBS/PBS. Subsequently, cells were stained with VL-4 rat anti-LCMV nucleoprotein (BioXCell), followed by staining with anti-rat horseradish peroxidase (Jackson Immunoresearch) in 2% FBS/PBS. Infected cells were visualized using 3-amino-9-ethylcarbazole chromogen and substrate (Biolegend). Viral titers were calculated based on the number of colored “spots” on the plate and the dilution factor of the serum samples. Levels of serum alanine aminotransferase and aspartate aminotransferase were measured enzymatically (Roche Diagnostics).

BM cells isolated from either WT or HDAC1-cKO mice (both expressing the congenic marker CD45.2) were mixed at a 1:1 ratio with BM cells from WT CD45.1 mice. Mixed BM cells (2 × 10⁶) were transferred i.v. into lethally irradiated CD45.1⁺ mice. Reconstituted mice were i.v. infected with 2 × 10⁶ FFUs of LCMV Cl13 7–8 wk after transplantation. Eight days later, mice were euthanized, and splenocytes were analyzed by flow cytometry.

Naïve P14-WT and P14-HDAC1-cKO^YFP cells (CD90.2⁺) were isolated as described previously (Gülich et al., 2021) and mixed at a 1:1 ratio. Either 1 × 10⁶ (analyzed at 48 and 67 h p.i.) or 2 × 10⁴ (analyzed on day 5 and day 8 p.i.) of the mixed P14 cells were transferred i.v. into CD90.1⁺ congenic mice. One day after the transfer, recipient mice were infected i.v. with 2 × 10⁶ FFUs of LCMV Cl13. Mice were euthanized at various time points after infection as indicated, and splenocytes were analyzed using flow cytometry. For some co-transfer experiment, naïve P14-WT or P14-HDAC1-cKO cells (CD45.2⁺) were mixed at a 1:1 ratio with P14-WT-CD45.1 cells (CD45.1⁺CD45.2⁺) and adoptively transferred i.v. into CD90.1⁺ congenic mice. One day after the transfer, recipient mice were infected, subsequently analyzing splenocytes using flow cytometry at 8 days p.i.

For timed induction of Hdac1 deletion, HDAC1-cKO^CreERT-YFP mice were treated with 2 mg of tamoxifen (Sigma-Aldrich) intraperitoneally after chronic disease has been established (i.e., 9 days p.i.). 6 days later (i.e., 15 days p.i.), mice were euthanized, and splenocytes were analyzed by flow cytometry.

Single-cell suspensions were prepared by mashing organs through a 70-µm cell strainer (Corning or pluriSelect). Hepatic lymphocytes were enriched via 35% Percoll (Cytiva)/RPMI 1640 (Sigma-Aldrich) gradient centrifugation (700 × g for 20 min at room temperature [RT], no breaks). Erythrocytes were lysed using 1× BD Pharm Lyse Buffer according to the manufacturer’s instruction (BD Biosciences). Cells were incubated with Fc block (1:250; BD Biosciences) and Fixable Viability Dye eFluor 506 (1:1,000; Thermo Fisher Scientific). For the detection of viral-specific CD8⁺ T cells, cells were subsequently incubated with fluorescently labeled H2-D^b/GP33–41 tetramer (1:1,000; kindly provided by the NIH Tetramer Core Facility), followed by staining with antibodies for surface markers. Intracellular stainings were performed using the Foxp3/TF.

Staining Buffer Set (Thermo Fischer Scientific) according to the manufacturer’s instruction. For HDAC1 detection, cells were stained with primary rabbit anti-mouse HDAC1 antibody, followed by incubation with fluorescently labeled secondary goat anti-rabbit antibody. For intracytoplasmic detection of active caspase 3, splenocytes were incubated at 37°C for 5 h in complete RPMI1640 medium, supplemented with 10% FBS, 100 U/ml penicillin-streptomycin (Thermo Fisher Scientific), 2 mM L-glutamin (Thermo Fisher Scientific), 0.1 mM nonessential amino acid (Thermo Fisher Scientific), 1 mM sodium pyruvate (Sigma-Aldrich), and 55 μM of β-mercaptoethanol (Sigma-Aldrich). Subsequently, cells were fixed with BD Cytofix Fixation Buffer and were stained with Anti-Active Caspase-3 (BD Biosciences) in BD Perm/Wash Buffer (BD Biosciences) according to the manufacturer’s instructions (Beltra et al., 2020). For cytokine analysis, cells were stimulated with 1 µg/ml GP33–41 peptide, 0.7 μl/ml GolgiStop (BD Biosciences), and 1 μl/ml GolgiPlug (BD Biosciences) for 4.5 h. Cells were acquired with a BD LSRFortessa (BD Biosciences) or CytoFLEX (Beckman Coulter) and analyzed using FlowJo v10.8.1 software (BD Biosciences). Antibodies used in this study are listed in Table S4.

Dimensionality reduction was performed using the DownSample and t-SNE Plugins in FlowJo v10.8.1 (BD Biosciences). For the generation of t-SNE plots, WT and HDAC1-cKO biological replicates were downsampled to the same number of cells/sample and concatenated for further analysis. t-SNE was run on the concatenated file using the default parameters provided by the software.

Single-cell suspensions from the spleens of either LCMV-infected (infected as described above) or uninfected WT and HDAC1-cKO mice were prepared. Equal numbers of splenocytes from one female and one male mouse were combined for each scRNA-seq sample per condition. Samples were incubated with fluorescently labeled H2-D^b/GP33–41 tetramer, appropriate antibodies against surface markers, oligonucleotide-conjugated antibodies (TotalSeq-A hashtags; Biolegend), and fixable viability dye (Zombie Green; Biolegend). Viable naïve (CD90.2⁺CD8α⁺CD44⁻) and LCMV-specific (CD90.2⁺CD8α⁺CD44⁺GP33-tet⁺) WT and HDAC1-cKO CD8⁺ T cells were sorted using an SH800S Cell Sorter (Sony Biotechnology). These four samples were pooled for multiplex sequencing (∼30,000 cells/run; a total of two runs was performed). Libraries were generated using the Chromium Controller and the Next GEM Single Cell 3′ Reagent Kit (v3.1, 10x Genomics) according to the manufacturer’s instructions. Libraries were sequenced by the Biomedical Sequencing Facility at CeMM using the Illumina NovaSeq 6000 platform. Raw sequencing results were converted to gene-barcode-count matrices using Cell Ranger (v6.1.2, 10x Genomics). The raw count matrices, which recovered 11,813 (first run) and 18,363 (second run) cells with 32,289 genes detected from both replicates, respectively, have been used for further computational analysis.

Analysis was performed in R (version 4.1.2) using the Seurat package (version 4.2.0) as well as tidyverse (version 2.0.0), ggplot2 (version 3.4.2), and EnhancedVolcano (version 1.18.0). Replicates have been demultiplexed with a positive quantile threshold of 0.99 (HTOdemux() function) and resulted in a total of 7,108 negative, 2,398 doublet, and 20,670 singlet cells. Only singlets were used for cell type annotation with SingleR (version 1.8.1). All cells annotated as “T cells” or “NK cells” were kept (20,171 cells) for further analysis. The number of genes (nGene), number of unique molecular identifier (nUMI), percentage of mitochondrial genes (percent.mt), and the novelty score (number of genes detected per UMI [log₁₀GenesPerUMI]) were used for quality control. Following filters were applied: 300 ≤ nGene ≤ 3,000, nUMI ≥ 500, percent.mt ≤ 5, and log₁₀GenesPerUMI > 0.8. Before integrating the data from the two replicates, the 16,393 high-quality cells were used to regress out unwanted variations due to the expression of cell cycle genes (SCTransform() and CellCycleScoring()). Clustering was performed at different resolutions considering all principal components for which the change of variation to the subsequent principal component was larger than 0.1%. A clustering resolution of 0.6 was used for all following analyses and visualizations. Marker genes for all clusters were defined using FindAllMarkers() with the default settings. AddModuleScore() followed by FeaturePlot() was used to visualize the expression of previously published genes. Differentially expressed genes between groups were determined with FindMarkers(), where the minimum log₂ fold change was set to 0 and the minimum fraction of cells in which a gene has to be detected in either of the two groups to be tested was set to 1%. GO term enrichment analyses were conducted with the packages clusterProfiler (version 4.8.1), org.Mm.e.g.,.db (version 3.17.0), and AnnotationHub (version 3.8.0). Top 200 marker genes of each cluster were used, and the resulting top five enriched pathways (in Biological Processes) from each cluster were collapsed to parent terms using Revigo (version 1.8.1) and then visualized with dotplot() from the enrichplot package (version 1.20.0). Trajectory inference was performed with Monocle3 (version 1.3.7), uniform manifold approximation and projection information was retrieved from clustering with Seurat, and cells were colored by pseudotime. Regulon activity analysis was conducted using SCENIC (version 1.3.1) in R using the final Seurat object as input. Regulon activity scores (area under the curve scores) were compared for each regulon between the groups using a Wilcoxon test followed by P value adjustment (false discovery rate [FDR]) using the stats package (version 4.5.0) and visualized as dotplots with the ggplot2 package.

Naïve WT and HDAC1-cKO P14 T cells (CD90.2⁺) were isolated as described previously (Gülich et al., 2021), and 1 × 10⁵ cells were transferred i.v. into CD90.1⁺ congenic mice. One day after the transfer, recipient mice were infected i.v. with 2 × 10⁶ FFUs of LCMV Cl13. Eight days later, mice were euthanized, and splenocytes from 4 to 5 female mice were pooled for staining and subsequent cell sorting. 5 × 10⁴ early Tex^prog (CD90.2⁺CD8α⁺Tim3⁻Ly108⁺) and 5 × 10⁴ early non-Tex^prog (CD90.2⁺CD8α⁺Tim3⁺Ly108⁻) cells were sorted into PBS. High-throughput chromatin accessibility mapping (ATAC-seq) was performed as previously described (Buenrostro et al., 2013; Corces et al., 2016), with minor modifications. After centrifugation at 500 g for 5 min at 4°C, the cell pellets were resuspended in tagmentation-mix (containing Tagment DNA buffer, Tagment DNA Enzyme [llumina], Digitonin [Promega], and Proteinase K [Roche]) and incubated for 30 min at 37°C with shaking (300 rpm). The reaction was stopped by incubating the samples on ice. For DNA isolation, the MinElute PCR Purification Kit (Qiagen) was used according to the manufacturer’s instructions and eluted into 12 μl of elution buffer. 1 μl eluate was used in a quantitative PCR reaction to estimate the optimal number of cycles for library amplification. The remaining sample of each tagmented sample was then amplified corresponding to the C_q value (i.e., the cycle number at which fluorescence has increased above background levels) in the presence of custom Nextera primers. PCR amplification was followed by solid-phase reversible immobilization (SPRI) (Beckman Coulter) size selection to exclude fragments larger than 1,200 bp. The DNA concentration of each library was assessed with a Qubit 2.0 Fluorometric Quantitation system (Life Technologies) before pooling in equimolar amounts. The resulting pools were sequenced on a NovaSeq 6000 instrument (Illumina) in a 50-bp paired-end configuration. Chromatin accessibility mapping by ATAC-seq was done in three biological replicates. next generation sequencing (NGS) reads in unaligned BAM files were converted into FASTQ format with samtools, NGS adapter sequences were removed via fastp (0.23.2, 5′-GTC​TCG​TGG​GCT​CGG-3′), and the reads were aligned to the GRCm38 (UCSC Genome Browser mm10) assembly with Bowtie2 (2.4.5, --very-sensitive, --no-discordant, --maxins 2000) before deduplicating with samblaster (0.1.24). Alignment statistics and mitochondrial read fractions were collected with samtools. Peaks were called with MACS2 (2.2.7.1, --no-model, --keep-dup auto, --extsize 147), and peak annotation was performed with HOMER (4.11). UCSC Track hubs allowed for visual inspection of results. A consensus peak set was compiled, and all peaks in mm10 assembly regions blacklisted by the ENCODE project (Consortium, 2012) were removed. Sequencing and mapping analysis were performed by the Biomedical Sequencing Facility at CeMM. Resulting count matrices and peak files were used for further computational analysis.

DARs in early Tex^prog and early non-Tex^prog were identified with the R (version 4.1.2) package DESeq2 (version 1.40.2) using the default log₂ fold change threshold and an alpha of 0.05. Genes associated with DARs were used to calculate module scores (AddModuleScore() from the Seurat package [version 4.2.0]) to visualize the respective gene expression in the scRNA-seq clusters. For statistical testing, an ANOVA followed by Tukey’s test was performed. The same set of genes was also used to conduct the GO term enrichment analyses for biological processes with the same R packages as described above. The top 10 enriched pathways from each set were collapsed to parent terms using Revigo (version 1.8.1) and then visualized with dotplot() from the enrichplot package (version 1.20.0). Gene set enrichment analysis against a known effector signature (GSE9650_EFFECTOR_VS_EXHAUSTED_CD8_TCELL_UP) from the ImmuneSigDB was carried out using the packages msigdbr (version 7.5.1) and fgsea (version 1.26.0). TF motifs in DARs were determined by de novo motif analysis using findMotifsGenome.pl from HOMER (v4.10.3). Parameters were set to mask repeat regions (-mask), use the exact input region sizes (-size given), and find a maximum of 10 motifs per length (-S 10). P values <10⁻¹¹ were considered true positive.

WT and HDAC1-cKO mice infected with LCMV Cl13 were euthanized 8 days p.i,. and splenocytes from 10 mice per genotype were pooled. CD8⁺ T cells were enriched via REAlease CD8a MicroBead Kit mouse (Miltenyi Biotec) prior to staining and cell sorting. 1.5 × 10⁶ early Tex^prog (CD90.2⁺CD8α⁺Tim3⁻Ly108⁺) and 1.5 × 10⁶ early non-Tex^prog (CD90.2⁺CD8α⁺Tim3⁺Ly108⁻) cells were sorted into FCS. The sorted cells were pelleted at 400 × g for 5 min at 4°C, washed with PBS once, then resuspended in cold nuclear extraction buffer (20 mM HEPES [pH 7.9], 10 mM KCl, 0.1% Triton X-100, 20% glycerol, 1x cOmplete Mini-Tablet [Roche], 1 mM MnCl₂, and 0.5 mM spermidine) and incubated for 10 min on ice. After centrifugation (400 × g, 5 min, 4°C), the nuclear pellet was resuspended in cold nuclear extraction buffer and kept on ice until ConA bead conjugation. ConA beads were prepared by washing twice with bead activation buffer (20 mM HEPES [pH 7.9], 10 mM KCl, 1 mM CaCl₂, and 1 mM MnCl₂) in a volume corresponding to 10x the original ConA bead volume and resuspended with bead activation buffer in the original volume. For CUT&RUN, 250,000 cells in 100 μl nuclear extraction buffer were incubated with 10-μl ConA beads for 10 min at RT. Nuclei-bound beads were then resuspended in 50 μl antibody buffer (20 mM HEPES [pH 7.5], 150 mM NaCl, 0.5 mM spermidine, 0.02% digitonin, 2 mM EDTA, and 1x cOmplete EDTA-free protease inhibitor cocktail [Roche]) containing 1 μl of H3K27ac antibody (Active Motif), rabbit IgG (Cell Signaling), H3K27me3 (Cell Signaling), HDAC1 (Cell Signaling), or Runx3 (Gift from Dr. Yoram Groner). Following overnight incubation at 4°C, beads were washed twice with 200 μl digitonin buffer (20 mM HEPES [pH 7.5], 150 mM NaCl, 0.5 mM spermidine, 0.02% digitonin, and 1x cOmplete EDTA-free protease inhibitor cocktail) and resuspended in 50 μl of digitonin buffer with pAG-MNase (IMP - Research Institute of Molecular Pathology). Beads were incubated for 30 min at RT, washed twice with 200 μl digitonin buffer, and MNase was activated as beads were resuspended in 50 μl digitonin buffer supplemented with 2 mM CaCl₂ and incubated for 2 h at 4°C. The reaction was quenched upon addition of stop buffer (340 mM NaCl, 20 mM EDTA, 4 mM EGTA, 50 µg/ml RNaseA, and 50 mg/ml glycogen). Following incubation at 37°C for 10 min, the supernatant was transferred to a fresh tube. DNA was isolated with NEB Monarch PCR and DNA clean up. Libraries were prepared from 2 ng using NEBNext Ultra II DNA Library Prep Kit, cleaned up using 0.9x volume of SPRI beads, and subjected to PE150 sequencing on the Illumina NovaSeq 6000 platform.

EL4 target cells were pulsed with 1 µg/ml of GP33–41 or OVA257–264 control peptides and labeled with CellTrace Violet to a final concentration of 5 µM or 0.5 µM, respectively. 1 × 10⁴ EL4 target cells were incubated for 4 h. with 0, 1 × 10⁴, 2 × 10⁴, or 4 × 10⁴ early non-Tex^prog (CD90.2⁺CD8α⁺Tim3⁺Ly108⁻) P14-WT or P14-HDAC1-cKO cells and afterward analyzed by flow cytometry.

Raw reads were aligned to the mm10 genome with bowtie2 (version 2.5.0). Blacklisted regions were removed via bedtools intersect (version 2.29.1). Duplicates were marked and, for control samples, removed with Picard tools MarkDuplicates (version 3.3.0-5-g46d23b4e4-SNAPSHOT). For quality control duplication rates, fragment lengths and replicate reproducibility were examined and plotted in R using the packages tidyverse (version 2.0.0) and ggplot2 (version 3.4.2). Reads, which mapped to mm10 as proper pairs, were merged to one BAM file per sample using samtools merge (version 1.21). RPKM normalized coverage files in BigWig format were generated with deeptools bamCoverage (version 3.5.6). Peaks were called with SEACR (version 1.3) for H3K27ac and H3K27me3 samples against IgG background and with MACS3 (version 3.0.3) for HDAC1 against KO background and Runx3 against IgG background samples. Peak calling resulted in BED or narrowPeak file output, respectively. These files together with the BigWig files were used for visualization of regions of interest in the IGV-Web app (version 2.2.7). Heatmaps and profiles were plotted using deeptools ComputeMatrix and plotHeatmap (version 3.5.6). Differential binding analysis was conducted with the R package Diffbind (version 3.18.0), setting summits = T. Numbers of open DARs from ATAC-seq overlapping with HDAC1 and Runx3-binding sites were determined using bedtools intersect (version 2.29.1). Volcano plots were done with the R package EnhancedVolcano (version 1.18.0). HOMER (v4.10.3) was used to annotate identified regions and to determine the most enriched TF motif in Runx3 peaks as described in the ATAC-seq analysis section.

Statistical analysis was performed using Prism 8 or 9 (GraphPad). The values of the geometric mean fluorescence intensity were normalized by dividing the value of each sample by the average of the corresponding WT samples. Unpaired or paired two-tailed Student’s t test was used for comparisons between two groups. One-way ANOVA followed by Tukey’s multiple comparison test was used for the comparison of >2 groups. The P values are defined as follows: *P < 0.05; **P < 0.01; ***P < 0.001; n.s., not significant.

Fig. S1 shows the percentages of Tex cell subsets in WT and HDAC1-cKO mice during LCMV infection and additional characterizations of early Tex cells 8 days p.i., related to Figs. 1, 2, and 3. Fig. S2 shows a series of experiments, related to Fig. 4, revealing a CD8⁺ T cell–intrinsic role for HDAC1 in Tex subset differentiation. Fig. S3 shows a further characterization of P14-WT and P14-HDAC1-cKO^YFP early Tex cells and early Tex subset distribution in WT^CreERT2 and HDAC1-cKO ^CreERT2 mice before tamoxifen administration, related to Figs. 5 and 6. Fig. S4 shows the supplemental data from the scRNA-seq analysis, related to Fig. 7. Fig. S5 shows the supplemental data from ATAC-seq and CUT&RUN analyses, related to Figs. 8 and 9. Table S1 shows the cell clusters and frequencies of early Tex cells as defined by scRNA-seq, related to Fig. 7. Table S2 shows the number of DARs in early Tex^prog and early non-Tex^prog cells as defined by ATAC-seq, related to Fig. 8. Table S3 shows the primer sequences used for genotyping experimental animals. Table S4 shows the list of antibodies used in this study.

Raw and processed scRNA-seq, ATAC-seq, and CUT&RUN (sequence, count, and/or peak) data underlying this study are openly available in the NCBI GEO database with the following accession number: GSE255887. The subseries numbers are as follows: scRNA-seq: GSE255886, ATAC-seq: GSE255884, and CUT&RUN: GSE297018. Used code in form of scripts and relevant metadata underlying the analysis and visualization of the scRNA-seq, ATAC-seq, and CUT&RUN data are available in the author’s GitHub repository (https://github.com/medunivienna-IFI-immunobiology/2025_Rica-Waldherr-Miyakoda_HDAC1-CD8Tcells-chronicLCMV).

We thank the NIH Tetramer Facility (Atlanta, GA, USA) for providing MHC class I tetramers, Ms. Kirkley and Mr. Klimek (CeMM, Vienna, Austria) for their assistance with sorting of cells isolated from LCMV-infected mice, Dr. Groner (Weizmann Institute of Science, Rehovot, Israel) for providing anti-Runx3 antibody, Dr. Cavanagh for proofreading, and the MedUni Wien core facility laboratory animal breeding and husbandry for animal husbandry. Figure graphics were created with https://BioRender.com.

S. Sakaguchi was supported by Austrian Science Fund (FWF) projects: P27747 and P35436. W. Ellmeier was supported by FWF projects: F7001, F7005, P35372 and by the FWF and Medical University of Vienna doctoral programs (DK W1212) “Inflammation and Immunity” and (DOC 32 doc.fund) “TissueHome”. N. Boucheron was supported by FWF projects: F7004 and P30885. C. Bock was supported by FWF project: F7002. A.P. Kutschat was supported by a Marie Skłodowska-Curie Actions Postdoctoral Fellowship: 101061151. D. Seruggia was supported by FWF projects: P36302 and P36069. J. Zhang was supported by China Scholarship Council: 202308340034. Open Access funding provided by Medical University of Vienna.

Author contributions: R. Rica: formal analysis, investigation, methodology, project administration, validation, visualization, and writing—original draft. M. Waldherr: data curation, formal analysis, investigation, software, validation, visualization, and writing—original draft, review, and editing. E. Miyakoda: formal analysis, investigation, and writing—original draft. A.P. Kutschat: investigation. M. Schülein: formal analysis and investigation. J. Zhang: investigation. R.A. Orbegozo-Medina: investigation. L. Sandner: investigation. V. Stolz: data curation. D. Waltenberger: resources. T. Krausgruber: investigation and writing—review and editing. C. Bock: supervision and writing—review and editing. N. Boucheron: supervision. D. Seruggia: supervision. W. Ellmeier: conceptualization, funding acquisition, supervision, and writing—original draft, review, and editing. S. Sakaguchi: conceptualization, formal analysis, funding acquisition, investigation, methodology, supervision, visualization, and writing—original draft.