Cutting Edge: Polycomb Repressive Complex 1 Subunit Cbx4 Positively Regulates Effector Responses in CD8 T Cells

CD8 CTLs are an essential component of adaptive immunity, working as key players in the elimination of intracellular pathogens and tumor cells (1). Upon activation, CTLs undergo intense expansion and differentiate into a heterogeneous population composed of terminal effector cells (TEs) that are characterized by the expression of KLRG1 and low expression of IL-7Ra (KLRG1⁺CD127⁻) and memory precursor cells that express the opposite pattern (KLRG1⁻CD127⁺). Complex and dynamically regulated transcriptional programs control the differentiation of these distinct CD8 T cell states, and the identification of the key players in this process is critical for understanding these molecular events and for advancing novel therapeutic strategies (1, 2). Transcription factors cooperate with epigenetic regulators, including chromatin modifiers that add, remove, or recognize post-translational modifications in histones, adding a new level of complexity to this mechanism (3–5). Polycomb repressive complex (PRC)2 plays an essential role in effector cell differentiation through the deposition of its repressive histone mark, H3K27me3, in memory-related genes during acute viral infection (4, 6, 7). Furthermore, H3K27me3 removal by lysine demethylase Kdm6 promotes effector differentiation through derepression of effector genes (8–10), highlighting the importance of H3K27me3 methylation/demethylation dynamic regulation to CTL differentiation. Another polycomb group (PcG) epigenetic complex with repressive function, PRC1, is composed of several distinct subunits, and the canonical configurations (cPRC1) can be recruited to genome sites by H3K27me3. This process is mediated by recognition and binding of chromobox (Cbx) family proteins (Cbx2, Cbx4, Cbx6, Cbx7, and Cbx8) to H3K27me3 through its conserved chromodomain (11, 12). Additionally, Cbx4 is reported to function as a small ubiquitin-like modifier (SUMO) E3 ligase through its two SUMO-interacting motifs (SIM) (13, 14), and Cbx4 E3 SUMO ligase activity has been reported to regulate essential transcriptional regulators, such as Dnmt3a and Hif-1α (15, 16). In this study, we report that Cbx4 deficiency promotes a memory-associated phenotype and transcriptional profile in CTLs, and overexpression of Cbx4 mutants in distinct functional domains showed that this epigenetic regulator induces effector differentiation primarily through its SIM domain.

All mice were on a C57BL/6 background. Experimental mice were 6 to 8 wk old and sex and age matched. P14 (lymphocytic choriomeningitis virus [LCMV] gp33-41-H2-Db–specific) Thy1.1 and Cd4-Cre mice have been described previously (17). Cbx4^fl/fl mice (obtained from Dr. Guoliang Xu, Shanghai, Institute of Biochemistry and Cell Biology, China) were bred to create Cbx4^fl/flCd4^Cre and P14 Thy1.1⁺ Cbx4^fl/flCd4^Cre mice. Mice were housed according to protocols approved by the Rosalind Franklin University of Medicine and Science Institutional Animal Care and Use Committee or according to Conselho Nacional de Experimentação Animal rules and Universidade Federal do Rio de Janeiro Comitê de Ética no Uso de Animais approval (protocols 054/20 and 003/23).

Naive CD8⁺ T cells were purified and cultured to generate effector- and memory-like CD8 T cells in vitro. Cells were activated with anti-CD3 (1 μg/ml for effector-like cells; 50 ng/ml for memory-like cells) and anti-CD28 (1 μg/ml for both) and polarized to effector- or memory-like phenotypes through supplementation with 100 U/ml rhIL-2 or 10 U/ml rhIL-2 and 10 ng/ml rhIL-7 and rhIL-15, respectively (18, 19). Knockdown experiments used ametrine-expressing murine retroviral vectors containing shRNAs targeting CD4 (shCD4) or Cbx4 (shCbx4) mRNA (20). Overexpression experiments used GFP-expressing murine retroviral vectors containing a Cbx4 coding sequence (Cbx4-OE), Cbx4 mutants carrying two point mutations (F11A/W35L) at chromodomain (ΔChromo), deletion of the two SIM domains (ΔSIM1-2), or empty vector (Mock) kindly donated by Dr. Wang (21). Transduction was carried out as previously described (9).

For adoptive transfer experiments, congenic C57BL/6 (Thy1.2) mice received i.v. 5 × 10⁵ in vitro transduced P14 Thy1.1 CTLs, were subsequently infected i.p. with 1.5 × 10⁵ PFUs LCMV clone 13, and were analyzed 8 d postinfection (dpi) as previously described to induce an acute infection response (20). Cbx4^fl/flCd4^Cre mice were infected i.p. with 2 × 10⁵ PFUs LCMV Armstrong and analyzed at 7 dpi and 60 dpi. LCMV strains were initially provided by Dr. Shane Crotty, La Jolla Institute (La Jolla, CA), and expanded with BHK cells as described before (22).

Cell surface, intracellular, and LCMV tetramer (H2Db-gp33-41 [KAVYNFATC] Alexa Fluor 647 or allophycocyanin) staining was performed as previously described (9). Cytokine production was measured from in vitro cultured cells upon restimulation with 10 nM PMA and 1 μM ionomycin for 4 h in the presence of brefeldin A. Samples were run on a FACSAria IIu, LSR II, or LSRFortessa cytometer (BD Biosciences), and data were analyzed with FlowJo (versions 9.9.4 and 10.7.1).

In vitro cultured and transduced P14 CD8 T cells were purified by FACS (ametrine⁺) and cocultured at different ratios with GFP-expressing parental mammary carcinoma cell line EO771 or EO771 cells expressing LCMV-antigen gp33-41 (EO771-GP33), as previously described (9). After 12 h, live GFP-expressing EO771 cells were determined by flow cytometry.

FACS-purified cells from in vivo experiments were used for preparation of RNA-sequencing (RNA-seq) libraries using the SMARTer Stranded RNA-Seq Kit (Clontech Laboratories). Reads were analyzed as previously described (23). Differentially expressed genes (DEGs) were considered when the DESeq2 analysis resulted in a p value <0.05. Gene set enrichment analysis (GSEA) was performed by comparing DEGs with a published dataset (24).

Total RNA was isolated from FACS-purified CD8 T cells using TRIzol (Invitrogen). cDNA synthesis and gene expression analysis were performed as previously described (9). Gene expression was normalized to Rpl22.

Statistical analysis was performed in Prism 7 or 8 (GraphPad Software) using nonpaired one-way ANOVA followed by the Tukey multiple comparisons test, a two-tailed paired or nonpaired Student t test, or two-way ANOVA followed by the Tukey or Šidák multiple comparisons test, as indicated.

To assess the role of Cbx4 in CTL differentiation, we employed an LCMV acute infection model using adoptive transfer of P14 CD8 T cells. Naive CD8 T cells from P14 Thy1.1⁺ mice were activated in vitro and transduced with retroviral vectors expressing shRNA targeting Cbx4 mRNA (shCbx4) or CD4 mRNA as a control (shCD4). Transduced cells were then adoptively transferred to wild-type (WT) Thy1.2⁺ congenic mice that were infected on the same day with LCMV (20) (Fig. 1A). The efficiency of Cbx4 silencing was confirmed by quantitative RT-PCR (Supplemental Fig. 1A).

The differentiation of P14 cells to an effector or memory phenotype was analyzed at 8 dpi by measuring KLRG1 and CD127 expression in transduced (ametrine⁺) and adoptively transferred (Thy1.1⁺) P14 CD8 T cells recovered from the spleen (Fig. 1B–1D, Supplemental Fig. 1B). Transferred LCMV-specific Cbx4-deficient CTLs (P14 shCbx4) had a lower frequency and number of TEs (KLRG1⁺CD127⁻) and, accordingly, a higher frequency and cell number of populations expressing the memory-associated marker CD127, both in memory precursor (KLRG1⁻CD127⁺) and in KLRG1⁺CD127⁺ populations. The frequency of transferred Thy1.1⁺ P14 cells and transduced ametrine⁺ Thy1.1⁺ P14 cells was lower in mice receiving P14 shCbx4 cells in comparison with mice receiving control cells; however, no significant difference was observed in the cell numbers (Supplemental Fig. 1B–1D).

To further investigate the contribution of Cbx4 to the regulation of the CTL transcriptional profile, P14 CTLs from the LCMV acute infection model described above were sort purified at 8 dpi, and total RNA was extracted for RNA-seq. Data analysis showed 836 DEGs when comparing P14 shCbx4 with P14 shCD4 cells (Fig. 1E). Among those genes, several surface markers that positively correlate with the memory cells were upregulated on Cbx4-deficient cells, including Il7r (CD127) and Slamf6 (2, 25). In addition, genes encoding key transcription factors Eomes and Tcf7 (Tcf-1), which promote the generation and persistence of central memory CD8 T cells, were also upregulated in comparison with control cells (26–29). Concordantly, genes positively correlated with the effector phenotype and function, such as Klrg1, Prf1, and Cx3cr1, were downregulated in P14 shCbx4 compared with P14 shCD4 cells. These gene expression alterations, such as the upregulation of Tcf7 and Il7r, were not just due to a higher frequency of memory cells in our Cbx4-deficient population, because sorted Cbx4-deficient KLRG1^hi CTLs also showed an enrichment of memory-associated genes (Supplemental Fig. 1E). To test whether these gene expression alterations lead to a global shift toward memory transcriptional programs, we performed GSEA using a previously published memory precursor (KLRG1^loCD127^hi) transcriptional signature from day 8 acute LCMV infection (24). We found that genes upregulated in P14 shCbx4 cells showed enrichment in a memory precursor transcriptional signature (Fig. 1F), indicating that Cbx4 is involved in the regulation of memory-associated genes. It is also plausible that Cbx4 could indirectly impact the expression of transcription factors related to CTL differentiation by controlling other transcriptional regulators. For example, our RNA-seq data showed that c-Myb, a transcriptional activator of Tcf7 (30), is upregulated in P14 shCbx4 cells. Similarly, Cbx4 was reported to activate the Wnt/β-catenin pathway, a pathway upstream of TCF-1 activation, in human lung adenocarcinoma cells (31).

Because we observed the downregulation of genes related to effector function (i.e., Prf1), we explored if Cbx4 deficiency could impact CD8 T cell cytotoxic function by coculturing activated and retrovirally transduced P14 cells with GP33-expressing GFP⁺ EO771 tumor cells (EO771) (Fig. 1G). We observed that P14 shCbx4 cells had diminished cytotoxicity compared with P14 shCD4, which further supports the observation of defective effector differentiation with concomitant skewing to a memory phenotype in the absence of Cbx4 protein.

Collectively, our findings show that Cbx4 deficiency upregulates a memory precursor transcriptional signature and expression of memory surface markers and decreases effector cytotoxic function, revealing a skewing of CD8 T cells toward a memory phenotype.

To confirm that Cbx4 could control memory CTL formation, we used acute LCMV infection in T cell–specific Cbx4-deficient mice (Cbx4^fl/flCd4^Cre, herein referred to as Cbx4 T knockout [KO] mice), analyzing polyclonal Cbx4-deficient CD8 T cells at 7 or 60 dpi (Fig. 2A; Supplemental Fig. 2). Analysis of lymphocyte populations in the thymus, lymph nodes, and spleens of Cbx4 T KO mice at steady state showed no alteration (data not shown). At 60 dpi, analysis of Cbx4 T KO mouse spleen cells showed a slight increase in the frequency and number of CTLs, as well as a higher number of LCMV-specific H2Db-gp33-41⁺ (GP33⁺) cells (Fig. 2B, 2C). In addition, Cbx4 T KO mice had a higher frequency of the KLRG1⁺CD127⁺ population in GP33⁺ CD8 T cells at both 7 and 60 dpi (Fig. 2D–2F; Supplemental Fig. 2C, 2D). On day 7, we observed an increased frequency of KLRG1⁻CD127⁺ memory precursor cells (Supplemental Fig. 2C, 2D). Similarly, at 60 dpi, we observed a higher number of the memory KLRG1⁻CD127⁺ population, although this was not reflected in population frequency. Accordingly, we found a decreased frequency and cell number of TE (KLRG1⁺CD127⁻) GP33⁺ CTLs (Fig. 2D–2F). Validating the enrichment of memory-associated genes in Cbx4-deficient CTLs, we observed that the T-bet/Eomes ratio was significantly lower in Cbx4 T KO mice, regardless of KLRG1 expression, in line with a global increased memory profile upon Cbx4 deficiency (Fig. 2G). Overall, our data demonstrate a CD8 T cell–intrinsic role of Cbx4 in controlling effector T cell differentiation. However, the results presented in the CD4^Cre mouse system cannot exclude a role of Cbx4 in CD4 T cells. Future studies will focus on the role of this protein in CD4 T cell biology.

To recapitulate this phenotype in an in vitro model system, we polarized CD8 T cells to either effector- or memory-like cells using a previously established protocol (19) (Supplemental Fig. 3A). Consistently, we observed increased levels of the memory markers CD127 and CD62L in Cbx4-deficient cells, regardless of the polarized phenotypes (Supplemental Fig. 3B, 3C). In addition, memory-like shCbx4 CD8 T cells revealed a significant reduction of IFN-γ⁺TNF⁺ population frequency and IL-2 expression upon PMA/ionomycin stimulation (Supplemental Fig. 3D–3J). In summary, deficiency in Cbx4 leads to increased expression of memory-associated markers in CTLs and a concomitant reduction in effector function, both in vivo and in vitro.

We next investigated the mechanism through which Cbx4 impacts CTL differentiation. Because Cbx4 protein can mediate PcG-dependent repression and function in parallel as an E3 ligase enzyme (11–14), we enforced the expression, in either WT or Cbx4-deficient P14 cells, of WT or mutant Cbx4 cDNAs lacking key functional domains: (1) ΔChromo, with an amino acid substitution at the chromodomain that prevents its binding to H3K27me3; and (2) ΔSIM1-2, which lacks both SIMs (Fig. 3A).

Cbx4 KO CD8 T cells activated and differentiated in vitro to a memory-like phenotype displayed accentuated expression of CD62L and CD127 during cell culture, and complementation of Cbx4-deficient cells with WT Cbx4 counteracted this phenotype (Fig. 3B, 3C). On one hand, overexpression of ΔChromo mutant reproduces, to a lesser degree, the effect seen upon WT isoform overexpression, indicating that the Cbx4 H3K27me3 binding function has a partial contribution to its role in the commitment to effector cell differentiation. On the other hand, deletion of SIM sequences not only reversed the WT isoform Cbx4 overexpression impact but also induced the opposite effect, promoting a memory phenotype in both P14 WT and P14 Cbx4 T KO cells. Overexpression of ΔSIM1-2 mutant in P14 WT cells raised the frequency of the CD62L⁺CD127⁺ population to levels observed in mock-transduced P14 Cbx4 KO cells, suggesting that overexpression of the SIM-deficient Cbx4 isoform might be competing with endogenous Cbx4 function and potentially acting as a dominant-negative version of the protein (Fig. 3C). The same patterns were observed for the expression of memory-related markers CD62L and CD127 at the protein (Supplemental Fig. 4A, 4B) and transcriptional levels (Supplemental Fig. 4C, 4D). Expression of CD25, a marker related to the effector phenotype, corroborated these data, showing an increase upon WT Cbx4 overexpression, whereas ΔChromo mutant overexpression partially reproduced the WT Cbx4 effect and ΔSIM1-2 mutant reversed the effect (Supplemental Fig. 4E). A similar pattern has been observed for CXCR3, consistent with the fact that its expression favors the development of short-lived effector cells (Supplemental Fig. 4F) (32).

The observation that both chromodomain and SIM domains are required for effector CTL differentiation is consistent with the fact that it has been shown in mouse embryonic fibroblasts that conjugation of SUMO at the Cbx4 SIM domain is essential for recruitment of cPRC1 to H3K27me3 in genome loci and control of PRC1 repressive activity (33). Moreover, it was reported that Cbx4-mediated SUMOylation of Ezh2 promoted its recruitment and enhanced Ezh2 methyltransferase activity, demonstrating that Cbx4 can regulate PRC2 (34). Further studies investigating Cbx4 SUMOylation, Cbx4 SUMO E3 ligase function targets, and partnerships with other PcG proteins (especially Ezh2) in CTLs are needed to fully define the role of Cbx4 in CTL differentiation. In addition, examination of the potential interaction of Cbx4 with H3K27me3 residues through the chromodomain is needed to understand how this protein controls polycomb-mediated repressive mechanisms during CTL differentiation.

Taken together, our data demonstrate that Cbx4, a PcG protein, participates in the control of CTL differentiation and that both Cbx4 SIM1-2 domains and chromodomain are required for the repression of the memory phenotype; however, SIM1-2 might play a more dominant role in this process.

The authors have no financial conflicts of interest.

We thank the La Jolla Institute Sequencing Facility and Bioinformatics Core for assistance with RNA sequencing, the National Institutes of Health Tetramer Facility for providing the LCMV tetramers, and the Flow Cytometry and Mice Facilities at the Departamento de Imunologia (IMPG/UFRJ) for technical support.