Kdm6b Regulates the Generation of Effector CD8⁺ T Cells by Inducing Chromatin Accessibility in Effector-Associated Genes

Abstract

The transcriptional and epigenetic regulation of CD8⁺ T cell differentiation is critical for balancing pathogen eradication and long-term immunity by effector and memory CTLs, respectively. In this study, we demonstrate that the lysine demethylase 6b (Kdm6b) is essential for the proper generation and function of effector CD8⁺ T cells during acute infection and tumor eradication. We found that cells lacking Kdm6b (by either T cell−specific knockout mice or knockdown using short hairpin RNA strategies) show an enhanced generation of memory precursor and early effector cells upon acute viral infection in a cell-intrinsic manner. We also demonstrate that Kdm6b is indispensable for proper effector functions and tumor protection, and that memory CD8⁺ T cells lacking Kdm6b displayed a defective recall response. Mechanistically, we identified that Kdm6b, through induction of chromatin accessibility in key effector-associated gene loci, allows for the proper generation of effector CTLs. Our results pinpoint the essential function of Kdm6b in allowing chromatin accessibility in effector-associated genes, and identify Kdm6b as a potential target for therapeutics in diseases with dysregulated effector responses.

Cytotoxic T lymphocytes, also known as CD8⁺ T cells, are an essential part of the adaptive immune system designed to eliminate intracellular pathogens and to mediate tumor immune surveillance. Naive CD8⁺ T cells are activated upon recognizing their cognate Ag presented by APCs and rapidly proliferate and adopt distinct fates with various effector and memory potential (1, 2). At the peak of the immune response, CD8⁺ T cells differentiate into a heterogeneous population consisting mainly of two distinct fates: 1) terminally differentiated short-lived effector T cells (SLECs), which are primarily responsible for eliminating the acute infection, and 2) memory precursor effector cells (MPECs) that will give rise to long-lived memory CTLs (1, 3, 4). After resolution of the infection, most Ag-specific terminally differentiated effector CTLs undergo apoptosis because of a lack of Ag stimulation. The remaining CD8⁺ T cells will form the long-lasting memory pool for durable protection upon subsequent reinfections.

The transcriptional regulation involved in the fate commitment of CD8⁺ T cells upon their activation has been widely studied (1). The heterogeneity of CD8⁺ T cells during differentiation is governed and reinforced by several key transcription factors that form a complex regulatory network (2, 5). However, the specific relationship between these fate-governing transcription factors and epigenetic regulators that co-orchestrate CD8⁺ T cell differentiation remains elusive. Effector CTL differentiation is favored by the combined activity of transcription factors Batf, Irf4, T-bet (encoded by Tbx21), Id2, Blimp-1 (encoded by Prdm1), Zeb2, and NFAT1, whereas Eomes, Id3, Zeb1, Bcl-6, Tcf-1 (encoded by Tcf7), Runx3, and NFAT2 are known transcription factors in driving proper memory CD8⁺ T cell formation. The relative ratios or abundances of transcription factors within the same family favor the generation of one population over the other. For instance, T-bet and Eomes both belong to the T-box transcription factor family, and a higher T-bet/ Eomes ratio favors the generation of effector CTLs, whereas a lower ratio promotes memory CTLs (6–8). Similarly, whereas Id2 favors effector cell differentiation, Id3 programs memory CTL generation (9–11). Moreover, we have found that NFAT1 is essential for proper effector cell generation, whereas NFAT2 is vital for memory cells (12). Batf and Irf4 have been shown to act as pioneer transcription factors in Th17 cell generation (13) and have also been demonstrated to regulate early effector CTL gene program expression (14). More recently, Runx3 was shown to be an important regulator of memory CTL generation by establishing chromatin accessibility on cis-regulatory regions (15) as well as also regulating tissue residency in tumors and nonlymphoid tissues (16).

Although recent research has focused primarily on the role of transcription factors in the generation of effector or memory CTLs (1), the cell-intrinsic changes at the epigenetic level are not fully understood (1, 17, 18). Epigenetic regulation includes DNA methylation, long-noncoding RNA interaction, and covalent modifications of histone residues. Histone modifications such as acetylation, methylation, sumoylation, phosphorylation, and ubiquitination play important roles in regulating gene expression (2, 19). For instance, inactive loci are marked by repressive trimethylated histone 3 lysine 27 (H3K27me3) residues (20), whereas H3K4me3 marks are associated with active gene expression. Bivalent deposition of H3K4me3 and H3K27me3 on the same locus is considered to poise gene expression, allowing for adequate temporal regulation of these genes in the presence of the appropriate signals (21). In naive CTLs, many genes critical for lineage commitment, differentiation, metabolism, and cell proliferation possess bivalent H3K4me3 and H3K27me3 marks. Upon naive T cell activation, the repressive H3K27me3 and permissive H3K4me3 marks are extensively changed, facilitating the differentiation of CTLs into effector and memory cells (22–24). Although chromatin dynamics of naive T cell activation are directly associated with transcriptional control of gene expression and T cell lineage commitment (25), the exact molecular mechanisms governing this process are still elusive.

Terminal effector (TE) CTL differentiation requires the repression of pro-memory and pro-survival genes, mediated by polycomb repressive complex 2 (PRC2) catalytic component enhancer of zeste homolog 2 (Ezh2) through trimethylation of H3K27 residues in these loci (22, 26). Dynamic regulation of H3K27me3 levels can be achieved not only through the deposition of this repressive mark by PRC2, but also upon its removal by lysine demethylase 6a (Kdm6a) or 6b (Kdm6b) (27, 28). Demethylation of H3K27 residues is associated with opening of the chromatin structure and deposition of active marks leading to the induction of gene transcription (27). Recently, Yamashita and colleagues have indicated that Kdm6a regulates the generation of effector CD8⁺ T cells during Listeria monocytogenes infection by removing H3K27me3 repressive marks from the Prdm1 gene locus (29). Although Kdm6b and Kdm6a are both H3K27 demethylases and have high homology in their JmjC domains, their N-terminal domains are less conserved. Some reports have suggested deficiency in Kdm6a or Kdm6b can result in opposite phenotypes, indicating they may even have contrasting functions (30, 31). Kdm6a and Kdm6b were found to be necessary for NKT cell development (32), CD4⁺ T cell trafficking, and differentiation and function of Tfh and Th17 cells (33–36). However, the role of Kdm6b during CD8⁺ T cell differentiation is still unclear. In this study, using T cell−specific Kdm6b knockout (KO) mice and short hairpin RNA (shRNA)−mediated knockdown of Kdm6b strategies, we demonstrate that cell-intrinsic Kdm6b-dependent remodeling of chromatin accessibility is indispensable for the proper generation and function of effector CTLs in vivo. We characterized the transcription profile of Kdm6b-deficient CD8⁺ T cells and found impairment in the induction of effector-associated genes, as well as the cytotoxicity program. Kdm6b-deficient cells show reduced killing activity both in vitro and in vivo. We also observed that although Kdm6b deficiency does not impair the early proliferation of Ag-specific CD8⁺ T cells during a primary response, it severely impacts the expansion and effector cell differentiation upon secondary challenge. Our results have important implications in the therapeutic manipulation of effector CTL responses during health and disease.

Materials and Methods

Mice

All mice were on a C57BL/6J background. The experimental mice were 6- to 8-wk-old and sex- and age-matched. P14 Thy1.1 and CD4-Cre mice have been previously described (12). Kdm6b^fl/fl mice were purchased from Jackson Laboratory [stock no. 029615, generated by Dr. Boselut (34)] and bred to create the following genotypes: Kdm6b^fl/+ CD4-Cre, Kdm6b^fl/fl CD4-Cre, P14 Thy1.1 and/or Thy1.2 Kdm6b^fl/fl CD4-Cre. Specific lack of Kdm6b protein expression in thymocytes from Kdm6b^fl/fl CD4-Cre mice was reported by Dr. Boselut’s group (34). We observed 85−90% reduction in Kdm6b mRNA levels from purified CD8⁺ T cells from day 4 (d4) lymphocytic choriomeningitis virus (LCMV)−infected animals (Supplemental Fig. 3F). All mice were maintained in specific pathogen−free barrier facilities and used according to protocols approved by the Rosalind Franklin University of Medicine and Science Institutional Animal Care and Use Committee.

T cells isolation, culture

Naive or total CD8⁺ cells were purified from spleen and lymph nodes using the Mouse Naive CD8⁺ T Cell Isolation Kit (catalog no. 19858) or Mouse CD8⁺ T Cell Isolation Kit (catalog no. 19853) from StemCell Technologies, respectively. DMEM supplemented with 10% heat-inactivated FBS, 2 mM L-glutamine, penicillin-streptomycin, nonessential amino acids, sodium pyruvate, vitamins, 10 mM HEPES, and 50 uM 2-ME were used for T cell culture (37). For T cell activation, naive CD8⁺ T cells were activated with anti-CD3 (clone 2C11) and anti-CD28 (clone 37.51) (1 μg/ml each, both from BioXcell) and 50 ng/ml gentamicin at 1 million cells per milliliter on a 6-well plate that had been precoated with 50 μg/ml goat anti-hamster IgG (Pierce Protein Biology, Life Technologies). Cells were removed from the initial stimulus 48 h after activation and were cultured at 0.5 million/ml with 100 or 10 U/ml of recombinant human IL-2 (38).

Retroviral plasmids and transductions

Retroviral particles were generated by transfecting PlatE cells with Ametrine-expressing murine retroviral vectors containing shRNAs targeting CD4 or Kdm6b as previously described (39). Virus supernatant was filtered through 0.45-μm filters and concentrated by centrifugation at 6000 × g (F14–14 × 50cy rotor) at 6°C overnight. In vitro−activated CD8⁺ T cells (as described above) were transduced with retroviral particles 18−20 h after activation. T cell culture media was carefully replaced with media containing concentrated retrovirus supplied with 8 μg/ml polybrene and centrifuged at 2,000 rpm for 1 h at 37°C and then put into 37°C 10% CO₂ incubator for 4h. For adoptive transfer of transduced P14 cells, T cell cultures were immediately harvested. For in vitro culture experiments, the original T cell media was used to exchange the added media containing retrovirus, and cells were then expanded until day 6 (d6).

LCMV models, Listeria infection and plaque assay

Mice were infected i.p. with 2 × 10⁵ PFUs of LCMV Armstrong (LCMV Arm) and analyzed on d4, day 8 (d8), or day 30 (d30)−day 45 (d45) postinfection (p.i.). For adoptive transfer experiments with transduced CTLs, congenic C57BL/6 (Thy1.2) mice received in vitro−transduced P14 Thy1.1 cells (expressing shCD4 or shKdm6b) and were subsequently infected i.p. with 1.5 × 10⁵ PFUs of LCMV clone13, as previously described (39). LCMV strains were initially provided by Dr. Shane Crotty at La Jolla Institute and expanded with BHK cells as described (40). Serum viral titers were measured by plaque assay as described (41).

For L. monocytogenes−GP33 (LM-gp33) infection, naive P14 wild-type (WT) and Kdm6b KO CD8⁺ T cells were isolated, activated, and cultured in vitro for 6 d using 10 U/ml rhIL-2. A total of 2.5 × 10⁵ cells of each genotype were injected (i.v.) into different groups of 5−6-wk-old WT C57BL/6 recipients. Mice with or without transferred cells were then infected (i.v.) with 5 × 10⁴ CFUs of LM-gp33 the next day. Mice were sacrificed 3 d p.i. Splenocytes were harvested, and serial dilutions were made with a final concentration of Triton X-100 0.5% before plating on brain-heart infusion plates as described (42). For LM-gp33 rechallenge, 1 × 10⁶ CFUs were transferred into mice previously infected with LCMV, and mice were euthanized 4 d after challenge. Splenocytes were harvested and phenotypically characterized by FACS, and bacterial burden determined as described above.

Flow cytometry analysis

Spleens, thymus, lymph nodes, and heparinized blood were used for isolating single-cell suspension. RBCs were lysed from spleens and heparinized blood with ACK lysis buffer. For LCMV tetramer staining, H²D^b-gp_33–41 (KAVYNFATC) Alexa647 or APC were incubated at room temperature before staining for cell surface molecules and intracellular staining. Cytokine production was measured by FACS from ex vivo splenic cells or in vitro cultured cells (in cytotoxicity assay) upon restimulation with 0.2 μg/ml gp33–41 peptide for 4 h in the presence of brefeldin A. The Abs used in staining are listed in Supplemental Table I. Samples were run on LSRII (BD Biosciences), and data were analyzed with FlowJo (Version 9.9.4 and Version 10.7.1).

Cytotoxicity assay and tumor protection model

For cytotoxicity assay, naive P14 CD8⁺ T cells were activated in vitro, transduced with retroviral constructs expressing shRNAs targeting Kdm6b or CD4, and cultured with 10 U/ml rhIL-2 as described above. On d6, Ametrine⁺ cells were purified by FACS and cocultured at different ratios with GFP-expressing parental mammary carcinoma cell line EO771 (negative control to determine nonspecific target lysis), or EO771 cells expressing the cognate Ag gp33–41 (EO771-GFP-gp33–41) (12). After overnight incubation (~18 h), the remaining live GFP-expressing EO771 cells were determined by flow cytometry as a measure of the cytotoxic activity. EO771 cells cultured in the absence of CTLs were used as a baseline for cell death.

For in vivo tumor protection model, 2 × 10⁶ EO771-GFP-gp33–41 cells were s.c. injected into C57BL/6 recipient mice. Three days after tumor inoculation, tumor size was measured using a caliper and mice were assigned to experimental groups to normalize their tumor size. A total of 5 × 10⁵ shRNA-transduced Ametrine⁺−sorted P14 CD8⁺ T cells were injected i.v. into those tumor-bearing mice in different groups. Tumor size then measured every other day using a caliper. The tumor volume was determined using the following formula: Width(W)*Width(W)*Length(L)/2. For TILs analysis, 1 × 10⁶ sorted cells were transferred to tumor-bearing mice (11 days) for 48 h. Tumors were isolated and trenched into pieces smaller than 1 mm² and incubated in digestion solution (1.5 mg/ml collagenase D and 50 U/ml recombinant DNase I in PBS) and incubated in 37°C for 45 mins with mixing in between incubation. RBC lysis was performed using ACK lysis buffer and then cells were counted and characterized by flow cytometry.

Mixed adoptive transfer model and lung tissue analysis

Peripheral naive P14 Thy1.1 WT and Thy1.1/Thy1.2 Kdm6b KO cells CD8⁺ T cells were isolated and mixed in a 1:1 ratio, and 5 × 10³ cells adoptively transferred into C57BL/6-congenic (Thy1.2) recipient mice. Purity of naive cells (CD8⁺ CD44⁻ CD62L⁺) as well as Thy1.1:Thy1.1/Thy1.2 ratio was confirmed by flow cytometry prior to cell transfer. One day later, mice were infected with LCMV Arm virus. To determine tissue-resident memory T cells in lung, and distinguish from vascular-associated CD8⁺ T cells, 1 μg of anti-CD8β PerCP-Cy5.5 Ab per mouse was injected i.v. three minutes prior to euthanasia and lung tissue extraction, as previously described (16). CD8β⁻ cells were considered to be localized within nonlymphoid tissues. Lungs were cut into pieces smaller than 1 mm² and incubated in digestion solution (1.5 mg/ml collagenase D and 50 U/ml recombinant DNase I in PBS) and incubated in 37°C for 30 min with mixing in between incubation. Digestion was neutralized by adding T cell media and tissues further dissociated over a 70-μm nylon cell strainer (VWR). Then lymphocytes were purified using a 40:60% Percoll density gradient separation. Spleens were also harvested from each mouse, and single-cell suspensions were generated. RBC lysis was performed using ACK lysis buffer, after which isolated cells were characterized by flow cytometry.

Mix bone marrow reconstitution and infection

Bone marrow cells were isolated from the tibia and femur. Mixed bone marrow chimera mice were generated by adoptive transfer of 1:1 or 1:5 ratio of bone marrow cells from B6.SJL (CD45.1⁺) mice and bone marrow cells from either Kdm6b T cell−specific (TKO) or WT (CD45.2⁺) mice into lethally irradiated B6.SJL mice (total of 7 × 10⁶ bone marrow cells). After 6 wk, reconstitutions were checked via bleeding and mice were then infected with LCMV Arm strain (2 × 10⁵ PFUs per mouse). Eight days p.i., mice were euthanized, and expression of KLRG1, CD127, and CXCR3 was determined on CD8⁺ CD44^hi CD4⁻ B220⁻ H2D^b-gp_33–41⁺ Ag-specific splenic cells (both in CD45.1⁺ and CD45.2⁺ populations).

RT-PCR and quantitative real-time RT-PCR

Total RNA was isolated from FACS-purified CD8⁺ T cells using TRIzol reagent (Invitrogen) according to manufacturer’s instructions. Superscript reverse transcriptase (Invitrogen) and oligonucleotide primers were used to synthesize cDNA. Gene expression was examined with 7900 Real Time PCR System (Applied Biosystems) using Power SYBR green PCR Master Mix (Thermo Fisher Scientific). Gene expression was normalized to Rpl32 (encodes L32 ribosomal protein) gene expression. The following primers were used: Rpl32 forward, 5′-CGTCTCAGGCCTTCAGTGAG-3′; Rpl32 reverse, 5′-CAAGAGGGAGAGCAAGCCTA-3′; Prdm1 forward, 5′-TGCGG AGAGGCTCCACTA-3′; Prdm1 reverse, 5′-TGCGGAGAGGCTCCACTA-3′; Id3 forward, 5′-TGCTACGAGGCGGTGTGCTG-3′; Id3 reverse, 5′-TGTCGTCCAAGAGGCTAAGAGGCT-3′; Tcf7 forward, 5′-CAAGGCA-GAGAAGGAGGCTAAG-3′; Tcf7 reverse, 5′-GGCAGCGCTCTCCTTGAG-3′; Tbx21 forward, 5′-AGGGGGCTTCCAACAATG-3′; Tbx21 reverse, 5′-AGACGTGTGTGTTAGAAGCACTG-3′; Zeb2(set1) forward, 5′-GGCAA GGCCTTC AAG TACA-3′; Zeb2(set1) reverse, 5′-AAGCGTTTCTTG-CAGTTTGG-3′; Zeb2(set2) forward, 5′-GAGCAGGTAACCGCAAGTTC-3′; Zeb2(set2) reverse, 5′-GATATTGTTTCTCATTCGG-3′; Kdm6b forward, 5′-CCCCCATTTCAGCTGACTAA3′; Kdm6b reverse, 5′-CTGGACCAA GGGGTGTGTT-3′. For RT-PCR, we used the following primer spanning exons 17−18: Kdm6b forward, 5′-CCATCCTGGGCATGAACACC-3′ and Kdm6b reverse, 5′-CAGAAGGCGCTGATGGTCTC-3′.

RNA sequencing

FACS-purified cells from in vivo experiments were washed twice with PBS, and 1 × 10⁴ cells were used for sequencing. cDNA was generated using SMART-seq v4 Ultra Low Input RNA Kit (Takara Bio), followed by Nextera XT (Illumina) library generation according to the manufacturers’ instructions. Multiplexed libraries were pooled and run on NovaSeq S1 or S2 flow cells with 50 × 50 pair-end reads.

The paired-end reads that passed Illumina filters were filtered for reads aligning to tRNA, rRNA, adapter sequences, and spike-in controls. The reads were then aligned to mm10 reference genome using STAR (v2.6.1) (43). DUST scores were calculated with PRINSEQ Lite (v 0.20.3) (44) and low-complexity reads (DUST > 4) were removed from the BAM files. The alignment results were parsed via the SAMtools (45) to generate SAM files. Read counts to each genomic feature were obtained with the featureCounts (v 1.6.5) (46) using the default option along with a minimum quality cut off (Phred > 10). After removing absent features (zero counts in all samples), the raw counts were then imported to R/Bioconductor package DESeq2 (v 1.24.0) (47) to identify differentially expressed genes among samples. The p values for differential expression are calculated using the Wald test for differences between the base means of two conditions. These p values are then adjusted for multiple test correction using Benjamini-Hochberg algorithm (48). We considered genes differentially expressed between two groups of samples when the DESeq2 analysis resulted in an adjusted p value < 0.05. Principal component analysis was performed using the ‘prcomp’ function in R. RNA-sequencing (RNA-seq) differential−expressed genes were plotted on the volcano plot. Further analysis were completed in R. Early effector-versus-effector or TE-versus−memory precursor gene signature (padj < 0.05) data sets were ranked based on their log2FC. Gene symbol or alias from the preranked signature data sets and RNA-seq differential−expressed genes (padj < 0.05) were converted into ENTREZID using clusterProfiler package (Bioconductor) before passing to pathway analysis enrich for biological process or use for gene set enrichment analysis (GSEA) described in clusterProfiler vignette.

Assay for transposase-accessible chromatin sequencing

Assay for transposase-accessible chromatin sequencing (ATAC-seq) libraries were prepared as described (49). Briefly, 5 × 10⁴ FACS-purified cells were washed twice with PBS and treated with lysis buffer (10 mM Tris pH 7.5, 10 mM NaCl, 3mM MgCl2, 0.1% NP-40). Pellets were then resuspended with 50 μl of Tn5 transposase containing 1× TD buffer (Nextera DNA Sample Prep Kit, Illumina, San Diego, CA) and incubated for 30 min at 37°C. Genomic DNA was then purified with QiaQuick MinElute PCR Purification Kit (Qiagen, Valencia, CA), and DNA amplified with KAPA HiFi Real-Time Library Amplification Kit (Kapa, Wilmington, MA) according to manufacturer’s instructions using barcoded primers for 11−13 cycles. Amplified libraries were purified with QiaQuick MinElute PCR Purification Kit, quantified, and pooled. Pooled libraries were pair-end sequenced on S1 100 flow cell on NovaSeq (Illumina, San Diego, CA) with 50 bp cycles in each direction. Raw data from the sequencer were uploaded to Illumina basespace (Illumina, San Diego, CA), and the samples were separated based on their barcodes and FASTQ files for each read obtained from each sample.

Sequencing Reads were mapped to the mouse genome (mm10) using bowtie (50) (-p 15 -m 1 −best - strata -X 2000 -S−fr−chunkmbs 1024). BAM files from replicates of WT and KO were merged and processed with picard MarkDuplicates to identify duplicate reads. Then, duplicate reads, mitochondria DNA, Y chromosome DNA, and fragments larger than 100 bp were excluded by SAMtools (45). Coordinates of reads were then shifted (plus strand +4, minus strain −5). Fragments were used to call peak summits with MACS2 (51) using parameters “−nomodel -q 0.01−keep-dup all−call-summits.” The summits for each peak from all replicates were expanded to regions with a uniform size of 500 bp. We excluded regions that intersected ENCODE-blacklisted regions for mm10 using trackerlayer R package.

We identified the Tn5 insertion site by isolating the first 9bp of each read and computed the number of transposase insertions per peak for each replicate with bedtools coverage (version 2.16.2) (52). Counts of ATAC-seq peaks were generated for all replicates and samples using featureCounts from the Rsubread R package (countMultiMappingReads = FALSE,maxFragLength = 100). Raw ATAC-seq counts in each peak for all replicates of all samples were normalized between replicates with size factors computed with DESeq2 (47) to differential coverage. Pairwise contrasts were performed with DESeq2 and differentially accessible regions were filtered based on a false discovery rate (padj) value of less than 0.05 and −0.5 > log2fc > 0.5. All bedgraph files of ATAC-seq or chromatin immunoprecipitation sequencing (ChIP-seq) are normalized by multiplying counts to a scaling factor calculated by 10⁹/accumulation-factor (accumulation-factor calculation: awk ‘{sum+=(($3-$2)*$4)} END {print sum}’). Bedgraph files were transformed to bigwig by bedgraphtobigwig, and uploaded as UCSC custom tracks. H3K27me3 ChIP-seq was obtained from GSE89036 (26).

Statistics and analysis

Graphs were plotted using Prism 7 and 8 GraphPad. Statistical analysis was performed using nonpaired one-way ANOVA followed by Tukey multiple comparisons, two-tail paired or nonpaired Student t test, or two-way ANOVA followed by Dunnett comparisons.

Results

Kdm6b is required for effector CD8⁺ T cell differentiation throughout acute viral infection

Kdm6a and Kdm6b are essential for proper development, as demonstrated by the embryonic or neonatal lethality of germline KO mice (53, 54). By using T cell−specific deficiency of both Kdm6a and Kdm6b, Bosselut and colleagues have shown the crucial role of Kdm6 members in T cell development (34). To study the role of Kdm6b in CTL differentiation, we bred Kdm6b-floxed with CD4-Cre mice and generated Kdm6b^fl/fl CD4Cre (Kdm6b T cell-specific KO, in this study referred to as Kdm6b TKO) or Kdm6b^fl/+ CD4Cre (T-Het) mice. Our data, similar to previously published reports (33, 34), showed that Kdm6b TKO mice have an increase in the percentage and numbers of both mature CD4 and CD8 single-positive cells compared with WT mice in the thymus (data not shown). Similarly, we also identified a concomitant reduction in peripheral T cells with CD4⁺ T cells being more affected than CD8⁺ T cells, implicating a partial defect of mature T cell egress (data not shown). Moreover, no defects, and in fact somewhat higher regulatory T cells frequencies among CD4⁺ T cells were observed in spleen and lymph nodes of Kdm6b TKO mice compared with WT controls (data not shown). Our data are consistent with previous publications suggesting that Kdm6b contributes to proper T cell development and that this phenotype is independent of housing or possible microbiota differences across institutions (33).

To comprehend the requirement of Kdm6b in the fate decision toward effector and memory CD8⁺ T cell differentiation upon an acute infection, we infected WT, Kdm6b T-Het, and Kdm6b TKO mice with LCMV Armstrong strain. We found that the frequency and numbers of total CD8⁺ T cells as well as Ag-specific CD8⁺ T cells (determined using H2D^b-gp_33–41 tetramers), were similar between all groups on d8 p.i. (Supplemental Fig. 1A–D). These results suggest that Kdm6b deficiency does not impair CD8⁺ T cell expansion, proliferation, and/or survival during a primary immune response. We then assessed the generation of effector and memory CTLs by measuring KLRG1 and CD127 (IL-7Rα) expression in Ag-specific cells. We found that Kdm6b TKO mice showed a significant reduction in TEs, and a concomitant increase in memory precursor and early effector cells (EECs) in terms of frequency and total numbers (Fig. 1A–C). We also observed an increase in CXCR3-expressing Ag-specific cells in Kdm6b TKO mice compared with WT and T-Het control mice, providing further evidence of the enhanced preference toward memory precursor cell generation (Supplemental Fig. 1E–F). We then assessed whether Kdm6b TKO mice were able to properly control the acute viral infection by assessing viral titers in serum. We found all experimental groups displayed no detectable viral plaques (data not shown), suggesting that despite reduced effector CTL generation, Kdm6b TKO mice are able to control a primary acute viral infection.

FIGURE 1.

Kdm6b is required for effector CD8⁺ T cell differentiation at the peak of the T cell expansion phase during an acute viral infection. Kdm6b^fl/fl CD4Cre (TKO), Kdm6b^fl/+ CD4Cre (T-Het), or WT mice were infected with LCMV Arm. Splenocytes were characterized at d8 p.i. (A−C) Representative specific cells determined by the expression of KLRG1 and CD127 (A) or the summary data of both frequency (B) and total numbers (C). (D and E) Expression of transcription factors T-bet, Eomes, Blimp-1, and IRF4 on CD4⁻ CD8⁺ CD44^hi H2D -gp_33–41⁺ was determined by intracellular staining as a representative histogram (D) or the combined normalized expression (E). (F and G) Splenocytes were restimulated with gp33–41 peptide for 4hrs in the presence of brefeldin A, and intracellular cytokine staining was performed. Representative contour plot (F) and combined frequency (G) of IFN-γ and TNF-α expression. The combined data of two independent experiments is shown. Statistical analysis was done with data from two biological replicates using nonpaired one-way ANOVA followed by Tukey multiple comparisons. *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001, ****p ≤ 0.0001. See Supplemental Fig 1.

Given the observed increase in EEC and memory precursor cells in Kdm6b TKO mice compared with WT controls, we then assessed the expression of key transcription factors that regulate effector CD8⁺ T cell differentiation. We found the expression of T-bet, Eomes, and Blimp-1 were significantly lower in CTLs from Kdm6b TKO mice compared with the T-Het or WT control, suggesting their plausible contribution to the reduction in the TE cell population (Fig. 1D, 1E). A similar reduction in T-bet and Eomes expression was observed in distinct populations based on KLRG-1 and CD127 expression (data not shown). To comprehend whether Kdm6b deficiency leads to impaired CTL function, we measured cytokine production upon restimulation of splenocytes with the gp33–41 peptide. Despite decreased numbers of TE cells, Kdm6b-deficient Ag-specific cells showed a higher frequency of IFN-γ⁺ TNF-α⁺−producing cells (Fig. 1F), although there were no differences in total cytokine-producing Ag-specific cell numbers compared with the controls (Supplemental Fig. 1G). Overall, these results suggest that Kdm6b deficiency does not result in defective polyfunctional IFN-γ⁺ and TNF-α⁺ production in CTLs.

Our data, together with previously published reports, show that depletion of Kdm6a or Kdm6b in T cells results in reduced splenic CD4 T cells [data not shown and (34)]. It has also been demonstrated that in the context of chronic viral infection, mice with a deficiency of Kdm6a in T cells show decreased follicular helper T (Tfh) cell development (55). Thus, we next determined if Kdm6b is necessary for the proper generation of Tfh cells in the context of an acute viral infection. Eight days after LCMV Arm infection, we found a reduction of total CD4⁺ T cells as well as Tfh cells (CD4⁺ PD-1⁺ CXCR5⁺) in spleens (Supplemental Fig. 1H). Similarly, mixed bone marrow chimera experiments also showed that Kdm6b-deficient cells differentiated to a lesser extent into Tfh cells compared with WT controls (data not shown). Thus, our results demonstrated a cell-intrinsic role of Kdm6b in driving Tfh cell generation upon acute viral infection. We observed no difference between Kdm6b T-Het mice and WT controls in any of the markers measured, suggesting Kdm6b haploinsufficiency is sufficient for proper T cell differentiation (Fig. 1, Supplemental Fig. 1). For this reason, for the remainder of this work, we excluded any further analysis on Kdm6b T-Het mice.

Kdm6b deficiency results in impaired secondary responses despite sustained memory phenotype

Our results suggest that deficiency in Kdm6b enhances the generation of memory precursor cells at the peak of the acute viral infection (Fig. 1). To determine whether this response is sustained at later time points, we analyzed mice at d45 after LCMV Arm infection (Fig. 2A). Similar to our previous observation, we demonstrate that mice with a T cell−specific deletion of Kdm6b showed significantly enhanced CD127⁺ KLRG1⁻ memory cells and CXCR3⁺ KLRG1⁻ cells, as well as a concomitant decrease in KLRG1⁺ effector memory cells (Fig. 2B, 2C). The total number of splenocytes as well as the number and frequency of total CD8⁺ T cells were similar between experimental groups (data not shown). However, the frequency and number of H2D^b-gp_33–41⁺ Ag-specific cells were significantly higher in Kdm6b TKO mice than the WT group (Supplemental Fig. 2A). We then determined the expression of transcription factors and found a significant reduction in Eomes expression despite observing an increase in the memory subset (Supplemental Fig. 2D). Upon restimulation with gp33–41 peptide, we observed higher frequencies of IFN-γ− and TNF-α−producing cells (Fig. 2D, 2E) but comparable numbers of cytokine-producing cells (Supplemental Fig. 2E), similar to what we have identified at the peak of the immune response in d8 p.i.

FIGURE 2.

Impaired recall response upon Kdm6b deficiency despite enhanced memory cell generation. Kdm6b^fl/fl CD4Cre (TKO) or WT mice were infected with LCMV Arm, and splenic CD8⁺ T cell population characterized on d45 p.i. (A−E). On d45, a subset of mice was challenged with L. monocytogenes expressing gp33–41, and splenocytes characterized 4 d later (F and G). (A) Schematics of the experimental design. (B and C) Representative flow contour plot of CD4⁻ CD8⁺ CD44^hi H2D^b-gp_33–41² Ag-specific cells determined by the expression of KLRG1, CD127, and CXCR3 (B and F) or the summary data of frequency of the indicated populations (C and G). (D and E) Representative contour plot (D) and combined frequency (E) of IFN-γ and TNF-α expression. The combined data of two independent experiments is shown. Student t test was performed to analyze statistical differences between groups from two biological replicates. *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001, ****p ≤ 0.0001. See Supplemental Fig. 2.

To determine whether the enhanced generation of memory cells in Kdm6b TKO mice correlates with enhanced recall responses, d45 LCMV-infected mice were challenged with L. monocytogenes expressing gp33–41 epitope as previously described (56). Four days p.i., splenocytes were harvested and H2D^b-gp_33–41 ⁺ Ag-specific CD8⁺ T cells were further characterized. The frequency and number of Ag-specific cells was higher in Kdm6b TKO mice compared with controls at a memory time point from a primary infection (d45) (Supplemental Fig. 2A). However, upon rechallenge with LM-gp33, we observed a reduction in H2D^b-gp_33–41 ⁺ Ag-specific cells compared with WT counterparts (Supplemental Fig. 2F). Consistently, only the WT mice reacquired KLRG1 expression on Ag-specific cells, indicative of a secondary stimulation of Ag-experienced cells, whereas Kdm6b TKO mice failed to re-express effector-associated marker KLRG1. Instead, Ag-specific CD8⁺ T cells in the Kdm6b TKO mice maintained a higher expression of memory-associated markers CD127 and CXCR3 (Fig. 2F, 2G, Supplemental Fig. 2G), suggesting Kdm6b is essential for the proper expansion and effector differentiation of Ag-specific cells during a recall response.

Cell-intrinsic role of Kdm6b in effector CTL generation and memory population maintenance

To evaluate whether Kdm6b induces effector CTL differentiation in a cell-intrinsic manner, we generated mixed bone marrow chimeras by mixing CD45.1 WT bone marrow cells with either WT or Kdm6b TKO CD45.2 bone marrow cells in a 1:1 ratio and cotransferred into lethally irradiated B6.SJL (CD45.1⁺)-congenic recipient mice (Supplemental Fig. 3A). Mice were infected with LCMV Arm 6 wk after bone marrow transfer, and then cells characterized on d8 p.i. We determined the ratio of CD45.1 versus CD45.2 lymphocytes in the blood (prior to infection) and in spleens (p.i.) and found fewer Kdm6b-deficient CD4⁺ and CD8⁺ T cells compared with controls (around 5:1 ratio of CD45.1 versus CD45.2), but more B cells (around 1:2 ratio of CD45.1 versus CD45.2) in each case, suggesting Kdm6b-deficient T cells were outcompeted by their WT counterparts (Supplemental Fig. 3B, 3C). Upon LCMV infection, we observed a similar decrease in SLECs and an increase in memory precursors and EECs in Kdm6b-deficient cells compared with WT controls (Supplemental Fig. 3D, 3E). Because of the seemingly competitive disadvantage of Kdm6b-deficient T cells compared with WT counterparts, we also performed mixed bone marrow chimeras by mixing Kdm6b TKO (CD45.2⁺) and WT (CD45.1⁺) bone marrow cells in a 5:1 ratio. Six weeks after reconstitution, a comparable ratio of WT to Kdm6b KO CD4⁺ and CD8⁺ T cells was observed in these experiments (data not shown). These data suggest that the competitive advantage of WT cells can be leveled by enhancing the initial frequency of Kdm6b-deficient precursor cells. In these settings, Kdm6b-deficient cells were still defective in committing to an effector fate upon acute viral infection, similar to the results observed in the 1:1 ratio of BM transfer (data not shown).

To further assess if Kdm6b deficiency leads to reinforced memory CD8⁺ cell generation, we used a mixed adoptive transfer model of P14 TCR-transgenic T cells, which recognize the gp33–41 epitope of LCMV. The advantage of this model consists of the fact that both WT and Kdm6b-deficient T cells are exposed to the same viral load and inflammatory condition in the same recipient mice. For this purpose, we isolated naive P14 Thy1.1 WT and P14 Thy1.1/Thy1.2 Kdm6b-deficient CD8⁺ T cells, mixed them in a 1:1 ratio and transferred them into C57BL/6-congenic (Thy1.2) recipient mice, which were subsequently infected with LCMV Arm (Fig. 3A). Although the frequency of Kdm6b-deficient cells were slightly higher on day 0, upon infection, the WT cells outcompeted the Kdm6b-deficient cells (both at the peak of the immune response on d8 or at memory time points past d30) (Fig. 3B, 3C). Together with the mixed bone marrow experiments, our results indicate a two-layer competition disadvantage of Kdm6b-deficient T cells: 1) in the generation of peripheral naive CD8⁺ T cells and 2) in the expansion of activated T cells during acute viral infection.

FIGURE 3.

Cell-intrinsic regulation of central memory CTL differentiation by Kdm6b. Naive WT (Thy1.1) and Kdm6b KO (Thy1.1/Thy1.2) P14 TCR-transgenic CD8+ T cells were mixed 1:1 and transferred into C57BL/6 congenic (Thy1.2) mice, after which mice were infected with LCMV Arm. At different time points, cells were characterized. On d45, mice were challenged with L. monocytogenes expressing gp33–41 (LM-gp33), and mice euthanized 4d later. (A) Schematic representation of the experimental design and execution. (B) Contour plots showing the frequency of WT (Thy1.1⁺) and Kdm6b KO (Thy1.1⁺ Thy1.2⁺) P14 CD8⁺ T cells at different time points. (C) Percentage of adoptive transferred WT (black) or Kdm6b KO (red) CD8⁺ T cells at indicated time points. (D, E, and G) Percentage of WT and Kdm6b KO CD8⁺ T cells based on expression of KLRG1, CD127, CD62L and/or CXCR3 on d8 (D), d30 (E) post LCMV infection, or post rechallenge with LM-gp33 (day 49 [d49]) (G). (E−G) Characterization of adoptive transferred WT or Kdm6b KO CD8⁺ T cells over 30 days p.i. (F and H) Expression (geometric mean fluorescence intensity) of transcription factors T-bet, Eomes, and TCF-1 from adoptive transferred WT or Kdm6b KO CD8⁺ T cells on d30 (F) or d49 (H). (I) Expression of Ki67 on WT or Kdm6b KO cells on d49. Data were collected from two independent experiments and two-tail paired t tests were used for statistical analysis. Lines connecting dots represent an individual mouse. ****p ≤ 0.0001. See Supplemental Fig. 3.

When assessing the generation of effector and memory CTLs in the mixed adoptive transfer experiment, we found that CD8⁺ T cells lacking Kdm6b preferentially committed to a memory fate with reduced effector cell generation both at the peak of the immune response on d8 as well as memory time points (Fig. 3D, 3E, Supplemental Fig. 3G, 3H). We also observed that roughly 20% more CD8⁺ T cells lacking Kdm6b become central memory CTLs (Tcm), compared with WT controls, by measuring expression of CD62L, CD127, and CXCR3 (Fig. 3E, Supplemental Fig. 3H). Similarly, we detected decreased T-bet but increased Eomes and Tcf-1 expression in Kdm6b-deficient splenic CTLs (Fig. 3F). The expression pattern of these transcription factors supports the enhanced memory phenotype, especially the central memory population. We then investigated whether the preferred commitment of Kdm6b-deficient T cells toward Tcm resulted in impaired tissue-resident memory cell generation or maintenance (Supplemental Fig. 3I). For this purpose, we i.v. injected an Ab against CD8β prior to euthanasia to distinguish cells in circulation versus tissue-resident cells, which would not be stained by CD8β. We found decreases in both the frequency of Kdm6b-deficient lung resident memory CD8⁺ T cells (Trm), as well as in the expression of the activation and tissue-resident marker CD69 in those cells compared with WT controls (Supplemental Fig. 3I). Our data suggest that Kdm6b deficiency promotes Tcm generation while reducing the Trm population. Although, it is plausible that the impaired Trm cell generation in Kdm6b-deficient cells might be a secondary effect because Trm cells have the potential to differentiate not only from MPECs but also from SLECs (“ex-KLRG1 cells”) as recently demonstrated (57). Altogether, we identified that Kdm6b supports effector and Trm cell differentiation, and restricts Tcm generation in a cell-intrinsic manner that largely echoes the phenotype observed upon infection of Kdm6b TKO mice.

To further comprehend the role of Kdm6b during recall responses, mice receiving the adoptively transferred cells were rechallenged with LM-gp33 on d45. Four days later, mice were euthanized, and the phenotype of WT and Kdm6b-deficient cells assessed (Fig. 3A). We found, similar to the observations in Kdm6b TKO mice (Fig. 2), a reduction in the expansion of Kdm6b-deficient cells compared with their WT counterparts (day 49, Fig. 3B, 3C). Although WT P14 cells reacquired KLRG1 expression upon secondary challenge, Kdm6b-deficient P14 cells maintained high levels of CD127 and CD62L expression, indicative of enhanced central memory cells (Fig. 3G, Supplemental Fig. 3J). Upon rechallenge, we observed similar increase in T-bet and decrease in Tcf1 expression as observed at memory time points (d30) (Fig. 3H). Consistent with the decrease in expansion of Kdm6b KO cells, we observed decrease Ki67 levels compared with WT controls (Fig. 3I), indicating that Kdm6b-deficient cells have a defective proliferative response upon recall responses. Therefore, these results confirm that Kdm6b, unlike Kdm6a (29), is essential for mounting secondary responses in CD8⁺ T cells in a cell-intrinsic manner.

Acute deletion of Kdm6b results in decreased effector CTL generation and effector-associated gene expression

Kdm6b is deleted during T cell development at the CD4⁺ CD8⁺ double-positive stage in Kdm6b TKO mice, which could result in abnormal differentiation or compensatory mechanisms during T cell development or homeostasis. Thus, we next determined whether acute deletion of Kdm6b upon T cell activation using an shRNA strategy would also impact effector CTL commitment. For this purpose, WT Thy1.1⁺ P14 TCR-transgenic CD8⁺ T cells were activated in vitro with anti-CD3 and anti-CD28 Abs and transduced with retroviral constructs (expressing Ametrine) containing shRNA against Kdm6b (or CD4 as control). Cells were then transferred into congenic C57BL/6 (Thy1.2) WT mice that were subsequently infected with LCMV, as previously described (39). Eight days p.i., spleens were harvested and transduced donor cells (P14 Thy1.1⁺ Ametrine^hi) were analyzed for expression of KLRG1 and CD127 (Fig. 4A). We observed a slight reduction in shKdm6b-expressing cells compared with shCD4 controls, consistent with the competitive disadvantage in the presence of WT cells (Supplemental Fig. 4). Similar to our results using Kdm6b T cell−specific KO mice, we observed a decrease in SLECs but an increase in MPECs and EECs upon Kdm6b deficiency (over 85% knockdown efficiency based on quantitative PCR (qPCR) analysis, Supplemental Fig. 4E), suggesting that acute Kdm6b deficiency during T cell activation affects the proper generation of effector CD8⁺ T cells in vivo (Fig. 4A, 4B). We also observed increased IFN-γ and TNF-α expression, but Granzyme B production in Kdm6b-deficient cells was drastically reduced compared with controls (Fig. 4C), suggesting that the Kdm6b-dependent CTL effector program can be epigenetically regulated through distinct mechanisms (cytotoxicity versus cytokine production).

FIGURE 4.

Acute Kdm6b deficiency in CD8⁺ T cells results in the defective effector-associated transcriptional program. P14 Thy1.1⁺ CD8⁺ T cells were transduced with shCD4 or shKdm6b and adoptively-transferred into congenic C57BL/6 (Thy1.2⁺) mice, which was subsequently infected with LCMV cl13 (1.5 × 10⁵ PFU/mouse). Eight days p.i., splenocytes were harvested and adoptively transferred and transduced CD8⁺ T cells were further gated for phenotypic characterization and sorted for RNA-seq and qPCR analysis. (A and B) Representative flow contour plot of CD8⁺ Thy1.1⁺ Ametrine^hi cells determined by the expression of KLRG1 and CD127 (A) or the summary data of frequency of the indicated populations from three independent experiments (B). (C) Splenocytes were restimulated with gp33–41 peptide for 4hrs in the presence of brefeldin A, and intracellular cytokine staining was performed. A representative histogram is shown on the top panel and the combined frequency in the bottom panel for each cytokine. A representative of three independent experiments is shown. (D and E) RNA-seq was performed on total CD8⁺ P14 Thy1.1⁺ Ametrine^hi cells. (D) Volcano plot displaying the differentially expressed genes with p_adj < 0.05 and Log₂FC > 0.5 cutoff comparing shCD4 control and shKdm6b knockdown from RNA-seq. Differentially expressed genes with the cut-off are labeled at the top. (E) GSEA and normalized enrichment scores (NES) of preranked transcriptional signatures associated with effector or memory cells (25) compared with all differentially expressed genes (P_adj < 0.05) in shCD4 versus shKdm6b using clusterProfiler (67). Student t tests were performed for statistical analysis. *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001, ****p ≤ 0.0001. See Supplemental Fig. 4.

To understand the genes regulated by Kdm6b, we performed RNA-seq on FACS-purified P14 Thy1.1⁺ Ametrine^hi cells. We identified 379 upregulated and 253 downregulated genes in the Kdm6b-deficient CTLs compared with shCD4 control (Fig. 4D). We observed an increase in Cxcr3 expression and decreased Klrg1 expression in Kdm6b-deficient cells, which corroborates our phenotypic assessments. Within the differentially expressed genes, transcription factors Tcf7, Id3, and Eomes, which are crucial for memory cell generation, were upregulated in Kdm6b-deficient CD8⁺ T cells. In contrast, effector CTL-promoting transcription factors such as Id2, T-bet, and Zeb2 were downregulated in the Kdm6b knockdown CTLs compared with controls. To validate these results, we also performed qPCR with the same sorted cells we used for RNA-seq. Additionally, we found Kdm6b-deficient CTLs displayed decreased expression of Prdm1, which is previously known to be regulated by Kdm6a (29) (Supplemental Fig. 4E). Altogether, our data suggest that Kdm6b deficiency derails effector CD8⁺ T cell differentiation and causes cells to preferentially commit to a memory fate. To confirm this, we next performed GSEA using previously published differentially expressed genes in effector and memory subsets. We determined that effector-associated genes are downregulated in Kdm6b KO cells, whereas memory-associated genes are upregulated in Kdm6b KO cells (Fig. 4E). To further understand what pathways may be impacted by Kdm6b deficiency during differentiation, we performed gene ontology analysis. Pathways related to cell migration, chemotaxis, and cytotoxicity were enriched in genes downregulated upon Kdm6b deficiency in CD8⁺ T cells (Supplemental Fig. 4F). Taken together, our shRNA-mediated depletion of Kdm6b as well as our T cell−specific Kdm6b KO data indicate that Kdm6b is needed for proper effector and memory signature gene expression upon T cell activation, and also suggest that the CTL effector program can be epigenetically regulated through distinct mechanisms, with Kdm6b being a central regulator of the cytotoxicity program and not cytokine production.

Kdm6b is required for effector CTL function and tumor control

Our data shows Kdm6b deficiency in CTLs resulted in reduced effector subset differentiation and Granzyme B production but normal IFN-γ and TNF-α polyfunctional cells compared with Kdm6b-sufficient CTLs. Our RNA-seq pathway analysis also suggests that cell migration and cytotoxicity are significantly impacted by Kdm6b depletion. These findings prompted us to investigate whether Kdm6b deficiency affects cytotoxic activity in a cell-intrinsic manner. For this purpose, we used in vitro−activated shCD4 or shKdm6b transduced memory-like cells as previously described (12). We found that, similar to our in vivo findings, Kdm6b-deficient cells displayed decreased Granzyme B expression and similar or even slightly enhanced IFN-γ and TNF-α production (Supplemental Fig. 5A). To assess their cytotoxic function in vitro, WT or Kdm6b-deficient memory-like P14 CD8⁺ T cells were cocultured overnight at different ratios with EO771 mammary carcinoma cell line expressing the cognate Ag gp33–41 and GFP (Fig. 5A). We measured GFP⁺ live cells as a readout for cytotoxicity and found that Kdm6b-deficient cells showed reduced killing capacity upon Ag recognition (Fig. 5B). To confirm that impaired cytotoxicity can be translated in vivo, we used a tumor protection experiment in which we s.c. inoculated EO771-GFP-gp33–41 into naive congenic recipient mice. Three days later, WT (shCD4-expressing) or Kdm6b-deficient P14 T cells, which recognize the gp33–41 epitope expressed by tumor cells, were transferred, and tumor growth was measured every other day (Fig. 5C). We found that whereas cells expressing shCD4 could control tumor growth properly, mice receiving Kdm6b-deficient CTLs were unable to control tumor growth compared with mice that did not receive any cells (Fig. 5D). This lack of tumor control was not caused by impaired T cell infiltration in the tumor, as similar frequencies of WT or Kdm6b-deficient TILs were observed 48 h after T cell transfer (Supplemental Fig. 5B, 5C). Similarly, in vitro−generated Kdm6b-deficient memory-like CTLs were unable to protect mice from L. monocytogenes infection (Supplemental Fig. 5C–F). Thus, these results suggest that Kdm6b is indispensable for proper effect- or cytotoxic functions in CD8⁺ T cells.

FIGURE 5.

Kdm6b is cell-intrinsically indispensable for proper cytotoxic activity in CTLs. (A and B) Naive P14 CD8⁺ T cells were activated in vitro with anti-CD3 and anti-CD28 and transduced with retroviral constructs expressing shCD4 or shKdm6b. Cells were then cultured under memory-like conditions (38). On d6, Ametrine^hi cells were FACS-purified and cultured at indicated ratios with EO771-GFP gp_33–41⁺ target cells for 18 h. The percentage of GFP⁺ live tumor cells was determined as an indication of cytotoxicity. (A) Schematics of the experimental design. (B) Line and dot plot show the cytotoxicity of transduced cells at different dilutions with mean ^+/− SD from four independent experiments. Statistics were performed using a two-way ANOVA followed by Dunnett comparisons. (C and D) Naive P14 CD8⁺ T cells were transduced as in (A) but Ametrine^hi cells were FACS-purified on d3 and further expanded until d6 in vitro. Cells were then transferred into EO771-GFP-gp33–41 tumor-bearing mice. (C) Schematics of experimental design. (D) Tumor growth kinetics of mice that did not receive any cells (blue line), mice that receive shCD4-transduced control cells (black line) or mice that received shKdm6b-expressing cells (redline). Statistics were performed using a one-way ANOVA followed by Tukey multiple comparisons. *p ≤ 0.05, **p ≤ 0.01, ****p ≤ 0.0001. See Supplemental Fig. 5.

Kdm6b is required for proper chromatin accessibility of effector-associated genes, which correlates with the demethylation of H3K27 residues

To understand the mechanism by which Kdm6b regulates effector cell generation, we performed the mixed adoptive transfer experiment described in Fig. 3A. In this instance, mice were euthanized 4 d p.i., and adoptively transferred cells purified to determine the transcriptional and chromatin accessibility changes observed prior to differentiation of cells into effector and memory CTLs. Although we observed higher frequencies of WT P14 cells compared with Kdm6b KO P14 at d8 or d30 p.i. (Fig. 3B, 3C), we found similar or even slightly higher frequencies and numbers of Kdm6b-deficient cells compared with WT controls at d4 p.i. (Fig. 6A, top panel, and Supplemental Fig. 6A, 6B), suggesting that the competitive disadvantage observed from Kdm6b-deficient cells is because of the proliferative burst from effector CTLs after d4. At this time point, few CTLs were able to commit to the terminally differentiated KLRG1⁺ fate, whereas most cells remained EECs with no commitment toward the memory precursor fate as measured by CD127 expression (Fig. 6A–C).

FIGURE 6.

Kdm6b is epigenetically required for promoting effector-associated genes and restraining EEC and memory precursor early during T cell differentiation. Naive WT (Thy1.1) and Kdm6b KO (Thy1.1/Thy1.2) P14 TCR-transgenic CD8⁺ T cells were mixed 1:1 and transferred into C57BL/6 congenic (Thy1.2) mice, after which mice were infected with LCMV Arm. Four days p.i. WT and Kdm6b KO CD8⁺ T cells were phenotypically characterized and sorted for RNA-seq and ATAC-seq. (A) Representative flow plot showing the WT (Thy1.1⁺) and Kdm6b KO (Thy1.1⁺ Thy1.2⁺) P14 T cells on day 0 prior to transfer and on d4 p.i. (B and C). Representative and combined percentage of KLRG1 and CD127 expression that determine SLEC (KLRG1⁺ CD127⁻) and EEC (KLRG1⁻ CD127⁻) populations. (D) Volcano plot showing differentially expressed genes with p_adj < 0.05 and Log₂FC > 0.5 cutoff by comparing P14 WT and Kdm6b KO cells. Genes of interest are labeled, and the number of differentially-expressed genes depicted. (E) GSEA and normalized enrichment scores (NES) of differentially expressed genes (p_adj < 0.05) in preranked transcriptional signatures of EEC, TE or memory precursor datasets obtained from (15). (F) Genomic distribution of differentially accessible regions determined by ATAC-seq (FDR < 0.05). (G) Correlation of differentially expressed genes (y-axis) and chromatin accessible regions of the different gene loci (x-axis) between WT and Kdm6b-deficient cells. Genes of interest are labeled; multiple differential accessible regions are circled out. (H) Representative gene loci showing the ATAC-seq peaks identified in WT and Kdm6b KO cells, and published H3K27me3 ChIP-seq from (26). Data were collected from three independent experiments and two-tail paired t tests were used for statistical analysis. Lines connecting dots represent an individual mouse. Three biological replicates were used for next generation sequencing. ****p ≤ 0.0001. See Supplemental Fig. 6.

To understand the early transcriptional changes regulated by Kdm6b during CD8⁺ T cell activation in vivo, we performed RNA-seq from WT P14 Thy1.1 and Kdm6b KO P14 Thy1.1 Thy1.2 obtained from the same mice. We observed 200 genes were upregulated in Kdm6b-deficient cells compared with WT counterparts, whereas 88 were significantly downregulated in Kdm6b-deficient cells (Fig. 6D). Among the genes that were significantly downregulated upon Kdm6b deficiency, we found Id3, Zeb2, and Prdm1, which have been shown to regulate effector CTL differentiation. Moreover, Il2ra, which codes for CD25 (high-affinity IL-2R subunit) and Il12rb1, the shared subunit between IL-12R and IL-23R, were also downregulated in Kdm6b-deficient cells compared with WT counterparts. It has been previously suggested that IL-2 as well as type I IFNs and IL-12 are important cytokines that drive effector CTL differentiation in part by downregulating expression of TCF-1 among others (8, 38, 58, 59), which could contribute to the phenotypic changes observed in CTLs lacking Kdm6b. To validate that Kdm6b-deficient cells failed to acquire effector-associated signature genes, we performed GSEA. We identified that genes upregulated in Kdm6b-deficient cells were significantly enriched in early effector and memory-associated genes, wheras downregulated genes in Kdm6b KO cells were enriched in effector gene signatures (Fig. 6E). Given that Kdm6b KO cells show reduction in the frequency of KLRG1⁺CD127⁻ population, we cannot rule out that some of the enriched genes are because of the differences in these populations. However, the KLRG1⁺ population represents a minor percentage compared with EECs (KLRG1⁻ CD127⁻), and therefore our data suggests mostly that the observed enrichment reflects an effect of Kdm6b on these effector genes already in the KLRG1⁻CD127⁻ population. Thus, our results suggest that Kdm6b is a critical regulator for promoting the expression of effector-associated genes, including transcription factors and cytokine receptors, among others. Altogether, these results indicate Kdm6b is indispensable during early CD8⁺ T cell differentiation for CTLs to transition from EECs to terminally differentiated effector cells.

We then assessed the chromatin accessibility changes that occur early during the CD8⁺ T cell activation/differentiation between WT and Kdm6b KO cells to understand how deficiency in Kdm6b results in preferential inhibition of effector CTL generation. Kdm6b-deficient CTLs surprisingly showed enrichment of 1113 accessible peaks within 1kb of transcription starting site (TSS), which are more likely to be at the promoter and nearby enhancer regions. In contrast, WT control CTLs have 523 enriched accessible peaks within 1kb of TSS (Fig. 6F). To further examine the relationship between chromatin accessibility changes and RNA expression upon Kdm6b deficiency, we plotted the differential gene expression versus ATAC-seq differential peaks. Only regions assigned to genes differentially expressed by RNA-seq are shown. Among those differentially accessible regions determined by ATAC-seq, we observed a positive correlation between open chromatin and upregulated gene expression in each genotype (Fig 6G). We found that Kdm6b-deficient cells displayed correlation between significantly open chromatin accessibility and increased gene expression such as Tcf7 and Id3. Conversely, their WT counterparts displayed significantly increased accessibility in genes like Klrg1, Zeb2, Il2ra, and Gzmb, which are known to be associated with effector CTLs. Collectively, these results strongly support that Kdm6b is required for remodeling the accessibility of genes associated with effector CD8⁺ T cell differentiation. Using published H3K27me3 ChIP-seq data from naive and effector CTLs, we investigated the correlation between the differentially accessible regions identified by ATAC-seq and their respective H3K27me3 methylation pattern in naive and effector (TE) CD8⁺ T cells. Within the differentially reduced accessible regions identified in Kdm6b-deficient CTLs compared with WT controls, many of them are demethylated upon activation and differentiation into effector CTLs. Examples include the promoter regions and some of the distal regions of Zeb2, Klrg1, Il2ra, Gzmb, and Prf1 loci (Fig 6H, Supplemental Fig. 6C). Conversely, Tcf7 locus, which shows higher accessibility in Kdm6b KO cells compared with their WT counterparts, is associated with increased H3K27me3 marks in effector CTLs compared with naive cells (Supplemental Fig. 6C). Together, our results show that effector-associated gene loci that display demethylated H3K27 residues compared with naive counterparts also show reduced accessibility upon Kdm6b deficiency. Given that we observe enhanced accessibility of memory and early effector genes in Kdm6b KO cells, it is possible that Kdm6b not only directly promotes the accessibility of effector-associated genes, but could also have an indirect effect in the repression of naive- or early effector-associated genes. Overall, these results suggest that Kdm6b is essential for the proper opening of chromatin regions of effector-associated genes from naive CD8⁺ T cells.

Discussion

Understanding CD8⁺ T cell differentiation in response to viral infection and tumor control is critical for novel vaccine development and improving immunotherapy efficacy. The differentiation of CTLs is controlled not only by fate-determining transcription factors but also by epigenetic regulators. Recent studies have demonstrated drastic epigenetic landscape differences among CTL subsets, especially in the deposition of permissive H3K4me3 and repressive H3K27me3 marks, or in DNA methylation status. The epigenetic landscape changes implicate the necessity of epigenetic regulators for proper CD8⁺ T cell differentiation. Many epigenetic regulators, such as Dnmt3a, Suv39h1, Kdm6a, and Ezh2, were found to promote effector CTL differentiation (22, 26, 29, 60–62). In this study, we confirmed previous studies showing that CD4-Cre driven Kdm6b deficiency in T cells resulted in increased mature single-positive CD4⁺ and CD8⁺ T cells (33, 34), and demonstrated that Kdm6b is essential for effector CD8⁺ T cell differentiation and proper effector function. We also identified a cell-intrinsic role of Kdm6b in the generation of Tfh cells, similar to what has been previously described for Kdm6a (55).

Kdm6a and Kdm6b may have nonredundant and additive effects in CD8⁺ T cell responses. Several studies have shown that both Kdm6a and Kdm6b are critical for CD4⁺ T cell egress from the thymus to peripheral lymphoid organs by regulating S1pr1 expression (33, 34). Dual deficiency of Kdm6a and Kdm6b resulted in an exacerbated impediment of CD4⁺ T cell egress, further indicating that Kdm6a and Kdm6b are not redundant and can have additive effects. Kdm6a and Kdm6b contain a Jumonji C catalytic domain on their C-terminal region, but their N-terminal domains are less conserved, suggesting the potential for other nonredundant functions [reviewed in (27)]. Indeed, it has been demonstrated that while Kdm6b modulates H3K27 methylation to promote NOTCH1 induced T cell acute lymphoblastic leukemia, Kdm6a functions as a tumor suppressor and is inactivated in this setting (31). However, the role of these lysine demethylases has not been evaluated jointly in CD8⁺ T cell biology. Our studies demonstrate a Kdm6b-dependent chromatin remodeling of effector-associated gene loci, allowing for the proper generation and function of effector CD8⁺ T cells. Similarly, Yamada et al. have also shown that Kdm6a (UTX) binds to Pdrm1 loci and its deficiency impairs the removal of H3K27me3 from the 59 UTR, first exon and intron (29), highlighting the importance of Kdm6 family members in driving effector CTL differentiation. However, Kdm6a-deficient cells have no defect upon recall responses (29) but we observe an impaired expansion of Kdm6b KO cells upon secondary challenge. Moreover, Kdm6a-deficient cells have normal expression of cytokines and Granzyme B, whereas our data shows that Kdm6b-deficient cells have defective granzyme and perforin expression, which correlates with decreased cytotoxicity function in vitro and in vivo. Yamada et al. were unable to observe any further differences in T cell responses upon treatment of Kdm6a-deficient cells with GSK-J4, an inhibitor targeting both Kdm6a and Kdm6b. This is surprising considering that our data clearly shows an important function of Kdm6b in driving effector CTL commitment, and the differences in phenotype between Kdm6a- and Kdm6b-deficient cells. Also, we have not observed any defects in CD8⁺ T cell expansion upon Kdm6b deficiency at the peak of a primary immune response or during memory maintenance. However, reduced expansion of Ag-specific cells was observed upon secondary exposure to Ag. Further studies are needed to understand the kinetics and requirements of these lysine demethylases, and how they may function in redundant or specific manners to support effector CTL generation during primary and secondary responses.

Our data show that Kdm6b functions early during T cell activation to induce expression of effector-associated genes. We found that deficiency in Kdm6b results in impaired upregulation of effect- or-associated genes including Tbx21, which codes for T-bet, Prdm1, which codes for Blimp1, Zeb2, Granzymes A and B, and several killer cell lectin-like receptors (Klrs). It has been previously suggested that T-bet requires Kdm6b activity to promote expression of Cxcr3 and Ifng in EL4 cells (63). Our results show that Kdm6b deficiency does not result in impaired IFN-γ production and in fact leads to increased levels of CXCR3 in vivo, suggesting that in CD8⁺ T cells, Kdm6b might be dispensable for T-box family member activity. Moreover, despite no defects in CXCR3 or IFN-γ production, Kdm6b-deficient cells do show reduced T-bet expression at the peak of the immune response and at memory time points, suggesting that other T-box member or transcription factors might regulate the expression of these genes. A recently identified T-bet regulatory network shows the unique roles of T-bet in promoting the expression of Zeb2, Gzma, Klrb1c in the TE subset and Bcl2, Crtam, Pou6f1 expression in the memory precursor subset respectively (64). Our RNA-seq data using acute deletion of Kdm6b in CTLs recapitulated the gene expression changes of the T-bet deficiency in the TE subset, suggesting that Kdm6b might hinder TE subset gene expression and function via T-bet regulatory network (Supplemental Fig. 4G). However, we were unable to observe reduced T-bet expression on d4 post infection, suggesting that other pathways may be important in further inducing T-bet expression during the expansion phase of CTLs after d4.

We identified discrepancies in the expression of T-bet and Eomes between the Kdm6b TKO mice model and the adoptive transfer model. The RNA-seq data from acute depletion of Kdm6b with shRNA (Fig. 4) and the P14 T cell adoptive transfer on day 29 (Fig. 3) consistently exhibit reduced T-bet and increased Eomes expression, which is different from the results observed upon infection of Kdm6b TKO mice. One possible explanation is an altered CD4⁺ T helper signal in Kdm6b TKO mice, which might further result in phenotypic changes not observed upon acute deletion or adoptive transfer, both experimental setups that occur in the presence of WT CD4⁺ T cells. In fact, our data showing defects in Tfh cell generation in Kdm6b TKO mice supports these conclusions (36). Moreover, Kdm6b has been shown to bind to T-box protein to positively regulate Eomes expression in endoderm differentiation (65). Our data suggest Kdm6b is required for controlling the balance between T-bet and Eomes expression, resulting in the regulation of CTL differentiation. Further studies assessing the direct interaction of Kdm6b with specific lineage-determining transcription factors could provide more mechanistic insights into these regulatory gene networks.

Our data indicates that Kdm6b is an essential epigenetic regulator that induces chromatin accessibility in effector-associated genes early during infection. We identified that absence of Kdm6b results in increased expression of memory-associated genes, particularly central memory genes on d4 post infection (Fig. 6). Of interest, we observed reduced expression of Il2ra and Il12rb1 on Kdm6b-deficient cells, and IL-2, together with IL-12 and type I IFNs have been shown to repress Tcf-1 expression, allowing for proper differentiation of effector CTLs (8, 38, 58, 59). We found that cells lacking Kdm6b preferentially give rise to Tcm cells, have higher expression of Tcf7 (codes for Tcf-1) based on RNA-seq analysis and also show diminished cytotoxic activity. Consistent with this, Held and colleagues have recently demonstrated that cells expressing high levels of Tcf-1 during the effector phase of an acute viral infection give rise to the central memory pool and lack cytotoxic activity (66). Thus, these results suggest that Kdm6b indirectly represses central memory CD8⁺ T cell generation through epigenetic chromatin remodeling of effector-associated genes. Another interesting finding was that Kdm6b deficiency resulted in defective expression of cytotoxic-related genes while not impacting cytokines such as IFN-γ or TNF-α. This suggests that the Kdm6b-dependent CTL effector program can be epigenetically regulated through distinct mechanisms.

Dynamic regulation of repressive histone marks is essential for the development of effector CTLs. Independent studies have shown that Ezh2, the catalytic subunit of PRC2, represses pro-memory and pro-survival genes through methylation of H3K27 residues, resulting in the proper differentiation of effector CD8⁺ T cells. However, the initial repression of pro-memory genes during differentiation is not likely catalyzed by Ezh2 deposition of trimethylation marks given that Ezh2 deficiency in CTLs does not affect Tcf7 and Bach2 expression on day 4.5 p.i. (22, 26). Despite our observation of reduced initial repression of those pro-memory genes in Kdm6b-deficient CTLs, the exact mechanisms need further investigation. Our work also demonstrates the importance of removing H3K27me3 marks from effector-associated gene loci for the proper generation and function of effector CTLs. Moreover, recently the H3K9 methylase Suv39h1 has also been implicated in the repression of stemness genes to induce the generation of effector CD8 T cells (60). The exact cooperation between these different epigenetic modulators in vivo remains unclear. Future studies are needed to address whether the dynamic H3K9 or H3K27 methylation of distinct gene loci requires both marks simultaneously for gene repression and proper generation of effector CTLs.

In the current study, we demonstrate that Kdm6b acts as an epigenetic modulator of CD8⁺ T cell fate determination by regulating effector-associated gene expression and chromatin accessibility. The exact role and potential cooperative nature between H3K27me3 lysine demethylases Kdm6a and Kdm6b needs further development. Because Kdm6 family members have been the therapeutic target for several cancers, properly understanding their intrinsic role in T cell function is warranted. Thus, our findings provide an important insight for the rational design of therapeutic strategies against infectious diseases, cancer, and autoimmunity.

Abbreviations used in this article:

ATAC-seq

assay for transposase-accessible chromatin sequencing

ChIP-seq

chromatin immunoprecipitation sequencing

d4

day 4

d6

day 6

d8

day 8

d30

day 30

d45

day 45

Ezh2

enhancer of zeste homolog 2

GSEA

gene set enrichment analysis

KO

knockout

LCMV

lymphocytic choriomeningitis virus

LCMV Arm

LCMV Armstrong

LM-gp33

L. monocytogenes−GP33

MPEC

memory precursor effector cell

p.i.

postinfection

PRC2

polycomb repressive complex 2

qPCR

quantitative PCR

RNA-seq

RNA sequencing

shRNA

short hairpin RNA

SLEC

short-lived effector T cell

Tcm

central memory CTL

TE

terminal effector

Tfh

follicular helper T

TKO

T cell−specific KO

Trm

tissue resident memory CD8⁺ T cell

WT

wild-type