Cbx3/HP1γ floxed mice were generated and provided by Dr. Prim B. Singh (49). These mice were backcrossed to C57BL/6 for twelve generations; they bred and developed normally before and after Cre-mediated deletion. The Cbx3/HP1γ transgenic line was generated by inserting a 1 kb fragment of mouse Cbx3/HP1γ cDNA, derived from activated CD8⁺ T cells, into EcoR1/Sma1 cloning sites of the VA-hCD2 vector provided by Dr. Dimitris Kioussis (50). The purified Cbx3/HP1γ-VA-hCD2 DNA construct was injected into C57BL/6 pronuclei by Dr. Lina Du at the fee-based Dana-Farber/Harvard Cancer Center (DF/HCC) Transgenic Mouse Core. Two founder transgenic lines were produced. The Lef1 floxed mice were generated as described and provided by Dr. Hai-Hui Xue (12). The following mouse lines were purchased from The Jackson Laboratory: Il21r knock out (B6.129-Il21r^tm1Kopf /J) (51), CD8α-Cre transgenic (C57BL/6-Tg(Cd8a-cre)1Itan/J) (52), ROSA26 (B6;129-Gt(ROSA)26Sor^tm2Sho /J) (53), Prf1 knock out (C57BL/6-Prf1^tm1Sdz /J) (54), and Ifng knock out (B6.129S7-Ifng^tm1Ts /J) (55). The B6.SJL line (B6.SJL-Ptprc^a /BoyAiTac) was purchased from Taconic.

The mouse syngeneic ID8 ovarian tumor line was obtained from Dr. Katherine F. Roby under an MTA executed by the University of Kansas Medical Center (56). Dr. Eva M. Schmelz provided the mouse syngeneic MOSE-L_{TIC v} ovarian tumor line (57, 58). The NB9464 neuroblastoma cell line was a kind gift from Dr. Crystal L. MacKall (Stanford University) (59). The B16-F10 melanoma cell line was purchased from ATCC and tested negative for mycoplasma.

All flow cytometry fluorochrome-conjugated antibodies were purchased from BioLegend, eBio-sciences or BD Biosciences. The following Western blot and ChIP-tested antibodies were purchased from Cell Signaling: tri-methyl-histone H3 (Lys9) (D4W1U) rabbit mAb #13969, phospho-Rpb1 CTD (Ser2) (E1Z3G) rabbit mAb #13499, phospho-Rpb1 CTD (Ser5) (D9N5I) rabbit mAb #13523, LEF-1 (C12A5) rabbit mAb #2230, cleaved caspase-3 (Asp175) antibody #9661 and phospho-HP1γ (Ser83) Ab #2600. The Western blot and ChIP-tested mouse anti-HP1γ mAb (clone 42s2) was purchased from Millipore Sigma. The rabbit polyclonal to TBR2/Eomes (ab23345) was purchased from Abcam.

Ovarian tumor. Mice were injected intraperitoneally (IP) with ID8 (5 x 10⁶/mouse in 200 µl PBS) or MOSE-L_TICv (1 x 10⁴/mouse in 200 µl PBS). On day 120 (ID8) or 36 (MOSE-L_TICv), when abdominal distension was visible, mice were euthanized, ascites were collected, and volume measured using syringes fitted with 18-gauge needles.

Melanoma. Mice were implanted subcutaneously with B16 tumor cells (1 x 10⁵/mouse in 100µL PBS). Using a digital caliber, B16 tumor size was measured and calculated on day 14 and at 2-day or 3-day intervals until day 20 depending on the size of the tumor, at which time mice were euthanized for analysis.

Neuroblastoma. Mice were implanted subcutaneously with NB-9464 tumor cells (1 x 10⁶/mouse in 100µL PBS). Using a digital caliber, NB-9464 tumor volume was measured and calculated (W x L x 0.4) starting on day 22 and at 2-day intervals until day 30 depending on the size of the tumor, at which time mice were euthanized for analysis.

Flow cytometry was performed on the Beckman Coulter CytoFLEX LX. Analysis was performed with FlowJo software version 10.8.0 (Tree Star, Inc.). Tumors and ascites were harvested, minced (NBL and melanoma), and cell-suspensions were filtered through 70μm cell-strainers (Fisherbrand). Cells were stained for appropriate surface markers as indicated in figures. Intracellular FOXP3 staining was performed according to manufacturer’s protocol (Biolegend). Briefly, tumor cells (1 x 10⁶) were harvested and washed 2X with staining buffer [PBS, 2% fetal bovine serum (FBS)]. Standard surface staining for intratumoral CD4⁺CD25⁺ T cells was performed. Cells were washed and incubated in fixation/permeabilization buffer (BioLegend) for 30 minutes at room temperature (RT) or 18 hrs at 4°CC. Following incubation, cells were washed 2X with permeabilization buffer (BioLegend) and stained for FOXP3 using anti-mouse FOXP3-PE (BioLegend). Cells were incubated for 30 minutes at RT. Cells were washed 2X with permeabilization buffer and resuspended in FACS buffer for analysis. All antibodies for flow cytometry were purchased from BioLegend: Il-21R-PE/Cy5 (clone 4A9); B220-Pacific Blue (clone RA3-6B2); CD8-PB; CD8-APC/Cy7; CD8-PE/Cy7; CD8-APC (all CD8 antibodies were from clone 53-6.7); CD25-APC/Cy7; CD25-PE (clone PC61 and clone 3C7); NKG2D-PE (clone CX5); CD4-PE; CD4-APC; CD4-FITC (all CD4 antibodies were from clone GK1.5); CD62L-APC (clone MEL-14); CD44-PB; CD44-APC/Cy7 (all CD44 antibodies were from clone IM7); TIM3-APC (clone B8.2C12); Streptavidin-PE; NK1.1-APC (clone PK136); Gr1-APC/Cy7 (clone RB6-8C5); Mac-1-PE (clone M1/70); CXCR5-Biotin (L138D7); FOXP3-PE (MF-14); CD45.2-PE/Cy7 (clone 104).

CD8⁺CD44^– T cells were purified from spleen and peripheral lymph nodes using the MojoSort™ Mouse CD8 Naïve T-Cell Isolation Kit (BioLegend, #480044) followed by CD25-depletion using the mouse CD25 MicroBead Kit (Miltenyl Biotec, #130-091-072), all according to manufacturers’ protocols. Naïve CD8⁺ T cells (1 × 10⁶/mL) were activated with plate-bound anti-CD3 (clone 145-2C11, 0.25 μg/ml, BioLegend) and anti-CD28 (clone 37.51, 0.5 μg/ml, BioLegend) in T-cell medium (high glucose DMEM, 10% FBS, penicillin/streptomycin, non-essential amino acids, HEPES, L-glutamate and sodium pyruvate) at 37°C in 10% CO₂ for 2 days. Cells were then removed from CD3/CD28 activation and re-cultured (5 × 10⁵/ml) in differentiation medium (T-cell medium supplemented with 10 IU/ml rhIL-2 from NCI) at 37°C in 10% CO₂. Cells were then sub-cultured every day in differentiation medium at 5 × 10⁵/ml each sub-culture. On day 5, cells were harvested and used for Western blots, ChIP-qPCR, qPCR or adoptive T-cell transfer.

In vitro-generated donor CD8⁺ effector T cells (CD45.2⁺) were prepared as above. On day 0, B6.SJL (CD45.1⁺) mice were implanted subcutaneously with NB-9464 (1 x 10⁶/mouse in 100μL PBS) or B16 (1 x 10⁵/mouse in 100μL PBS) tumor cells. On day 10 (B16) or 14 (NB-9464) after tumor induction, mice were treated with donor CD8⁺ effector T cells (3 x 10⁶/mouse in 100μL PBS) via tail vein intravenous injection. B16 tumor size and NB-9464 tumor volume were measured and calculated on indicated dates.

Target B16 tumor cells were plated in 12-well plates at a density of 5 x 10⁵ cells per well in 750μl of medium (high glucose DMEM, 10% FBS, penicillin/streptomycin, non-essential amino acids, HEPES, L-glutamate and sodium pyruvate). CD8⁺ or CD4⁺ effector T cells were added. For ratio of 1:1 (effector:target), 5 x 10⁵ CD8⁺ or CD4⁺ T cells: 5 x 10⁵ B16 cells were co-cultured in the same well; ratio 5:1, 2.5 x 10⁶ CD8⁺ or CD4⁺ T cells: 5 x 10⁵ B16 cells. Plates were incubated at 37°CC in 5% CO₂ for 24 hours. Wells were washed to remove non-adherent T cells. Adherent B16 cells were collected and washed with 1 ml cold PBS. Pellets were resuspended in Radio-Immunoprecipitation Assay (RIPA) lysis buffer containing a protease inhibitor cocktail (Roche) and used immediately or stored at –80°CC for further analysis.

Cells (1 x 10⁶) were lysed with RIPA buffer (Boston BioProducts) containing a protease inhibitor cocktail on ice for 30 minutes. Lysates were centrifuged at 13,000 rpm for 10 minutes at 4°CC. Protein extracts were denatured at 95°CC for 10 minutes, separated by SDS-PAGE, and transferred to PVDF membranes (EMD Millipore). Membranes were probes with primary antibodies. Proteins of interest were detected with HRP-conjugated secondary antibodies and the Pierce™ ECL Western Blotting Substrate (Thermo Fisher Scientific). Membranes were exposed with HyBlot CL Autoradiography films (Denville Scientific, Inc.), and developed with the Kodax X-OMAT 2000 Processor.

ChIP was performed using the SimpleChIP Kit #9003 according to manufacturer’s protocol (Cell Signaling). Briefly, 2 x 10⁷ CD8⁺ T cells were used for each ChIP. Cells were fixed in 1% formaldehyde to cross-link proteins to DNA, then lysed with 500μL of lysis buffer containing 1X ChIP buffer and protease inhibitor cocktail. Chromatin was sheared using a Sonic Dismembrator (Fisher Scientific Model 120) at 120W-20kHz power, at 15 seconds per cycle, 45 seconds break in between, for 3 cycles total. Chromatin was subjected to immunoprecipitation using specific antibodies at 4°C overnight with rotation. Following incubation, ChIP grade protein G magnetic beads were added to each ChIP and incubated for 2 hours at 4°C with rotation. Samples were placed in a magnetic separation rack for 2 minutes each time and washed 3X with high and low salt buffers. Chromatin was eluted from antibody/protein G magnetic beads using 1X ChIP elution buffer at 65°C for 30 minutes. Eluted chromatin was collected and subjected to cross-link reversal using 5M NaCl and 20 mg/ml Proteinase K, incubated for 2 hours at 65°C. After reversal of protein-DNA cross-link, the DNA was purified using DNA purification spin columns and eluted with DNA elution buffer provided in the kit. Purified ChIP DNA was immediately used for qPCR. BioRad hard-shell PCR Plates (BioRad) were used. In each well 2μL of purified ChIP DNA was added in triplicates along with 18μl primer mixture, which consisted of 1μL forward primer (5μM), 1μL reverse primer (5μM), 6μL nuclease-free water, and 10μL of 2X QuaniNova SYBR Green PCR Master Mix (Qiagen) or SimpleChIP Universal qPCR Master Mix (Cell Signaling). The plate was centrifuged at 300 RCF for 1 minute and read in the BioRad CFX384 Real-Time System (BioRad). The following qPCR settings were used: initial denaturation 95°C for 3:00 minutes, denatured at 95°C for 0:15 minutes, annealing and extension at 60°C for 1:00 minute, GOTO step denature and extension for a total of 40 cycles. Quantitative PCR result was analyzed, and the IP efficiency was calculated per formula provided in Simple ChIP Kit #9003 (Cell Signaling). Primers are listed in Table S2 .

ChIP-Seq detailed experiments and analyses have been described previously (60).

Total RNA was extracted from tumors or activated/differentiated CD8⁺ T cells using TRIzol (Thermo Fisher Scientific). After DNAse I treatment and RNA cleanup with the RNeasy kit (Qiagen), 1 μg of total RNA was reversed transcribed into cDNA using the Invitrogen SuperScript II reverse transcriptase kit according to manufacturer’s recommendations (Thermo Fisher Scientific). The BioRad hard-shell PCR Plates (BioRad) were used. Typically, 5-10 ng of cDNA was added to each well in triplicates along with 7.5μL of the primer mixture, which consisted of 0.5μL forward primer (10μM), 0.5μL reverse primer (10μM), 1.7μL PCR grade water, and 5μL of Roches LightCycler 480 SYBR Green I Master (Roche). The plate was centrifuged at 300 RCF for 1 minute and read in the BioRad CFX384 Real-Time System (BioRad). The qPCR settings: 95°C for 10:00 minutes, 95°C for 0:10 minutes, 62°C for 0:20 minutes, 72°C for 0:20 minutes (Read Plate), GOTO step 2 for 39 times, 92°C for 0:05 minutes, melt curve 65°C to 97°C, at increment of 0.5°C for 1:00 minute (Read Plate), 40°C for 0:10 minutes. Primers are listed in Table S3 .

Tumor section processing and TUNEL staining were performed by HistoWiz (Brooklyn, NY). Briefly, all IHC staining is automated using BOND Rx. Slides were treated with Dewax, a solvent-based solution (Leica Biosystems). Tissue sections were fixed by adding 10% neutral buffered formalin for 15 minutes. Slides were treated with Proteinase K at 1:500 dilution. Tissue sections were re-fixed using 10% neutral buffered formalin. Tissue sections were then treated with the equilibration buffer and incubated for 12 minutes at room temperature (RT). Subsequently the TdT reaction mix was added and incubated for 60 minutes at 37°C. The TUNEL reaction was stopped by adding 2X SSC. Tissue slides were treated with Peroxide Block (3-4% Hydrogen Peroxide) followed by wash buffer. Streptavidin HRP was added and incubated for 30 minutes at RT. DAB was added onto the slides and incubated for 10 minutes. Counter-staining was done by adding hematoxylin. Slides were covered using the Sakura Tissue Tek strainer and cover slips then scanned at 40x using the Leica AT2 scanner.

Statistical analysis of tumor growth was determined by the Graphpad Two-Way Anova, for survival the Log-rank (Mantel-Cox) test and for all others the Graphpad unpaired student t test (comparison between 2 groups) or One-Way Anova (multiple comparisons of group means) were performed. The number of mice used were determined using the GraphPad StatMate^®, no animals were excluded from the analysis, and no randomization or blinding was applied.

Previously we showed that germline deletion of Cbx3/HP1γ impairs lymphoid-tissue germinal center (GC) reaction and high-affinity antibody response against a thymus (T)-dependent antigen (Ag) in a CD8⁺ T-cell-intrinsic manner (61). Cbx3/HP1γ deficiency releases the effector capacity of CD8⁺ T cells to control neuroblastoma (NBL) growth (60). However, it is not understood why Cbx3/HP1γ-deficient CD8⁺ effector T cells are not subjected to functional senescence and persist in NBL or if they can control tumors with varied mutation loads. To ensure Cbx3/HP1γ deletion was restricted to CD8⁺ T cells, all experiments were performed using the CD8α-Cre strain (52) crossed with our Cbx3/HP1γ-floxed mice (resulting animals were Cbx3/HP1γ^fl/+ and Cbx3/HP1γ^fl/fl; collectively designated as Cbx3/HP1γ-deficient). Cre recombinase from this CD8α-Cre strain was active in CD8⁺CD4^– T cells not in CD4⁺CD8^– or CD11b⁺ cells ( Figure S1A ). Ablation of Cbx3/HP1γ in CD8⁺ T cells resulted in a near or complete loss of protein expression and phosphorylation ( Figure S1B ). Expression of Bcl6 and Tbx21 (T-bet) (62) mRNAs was increased in Cbx3/HP1γ-deficient CD8⁺ effector T cells over control cells ( Figure S1C ). By contrast Eomes transcript and protein levels were reduced ( Figures S1C, D ). We observed an upregulation of Prf1, Gzmb and Ifng mRNA expression in Cbx3/HP1γ-deficient CD8⁺ effector T cells compared to control cells ( Figure S1E ). More cleaved caspase 3 (CC3) was detected in B16 melanoma tumor cells co-cultured with Cbx3/HP1γ-deficient CD8⁺ effector T cells compared to control co-cultures ( Figure S1F ). Cbx3/HP1γ deficiency in CD4⁺ T cells did not enhance tumor cells killing. Thus, after activation, Cbx3/HP1γ-deficient CD8⁺ T cells differentiate into effector-like cells armed with a heightened effector/killing capacity to induce tumor-cell apoptosis in vitro. To determine mechanisms conferring persistence on Cbx3/HP1γ-deficient CD8⁺ effector T cells, ChIP-Seq data was analyzed (60). In wild-type day 5 activated/differentiated CD8⁺ T cells, Cbx3/HP1γ was bound to the 5’ untranslated region (UTR) surrounding transcriptional start sites (TSSs) of Lef1 ( Figures 1A, B ) and Il21r ( Figures 1D, E ) ( Table S1 ). Western immunoblots showed an induction of LEF-1 expression in activated/differentiated control CD8⁺ T cells; however, its expression was further increased when Cbx3/HP1γ was ablated ( Figure 1C ). LEF-1 levels in control and Cbx3/HP1γ-deficient thymocytes were similar ( Figure S1G ). The effects of IL-21 on IL-21R expression are controversial. In humans, IL-21 was shown to induce IL-21R expression on naïve CD8⁺ T cells (63). In mice, IL-21 exposure led to the downregulation of its receptor expression on all T-cell subsets (64). To mitigate IL-21 confounding effects, all experiments were done without added IL-21. In the absence of exogenous IL-21, activated/differentiated Cbx3/HP1γ-deficient CD8⁺ T cells expressed more IL-21R protein (measured as mean fluorescence intensity or MFI) and transcript, compared to control cells ( Figures 1F–H ). There was a steady expansion of Cbx3/HP1γ-deficient CD8⁺IL-21R⁺ T cells after activation/differentiation while the growth of control cells remained low throughout the activation/differentiation period ( Figures 1F, I ). IL-21R was not detected on thymocytes from control or Cbx3/HP1γ-deficient mice ( Figure S1H ). We establish that Cbx3/HP1γ deficiency induces the sustained increase of LEF-1 and IL-21R in CD8⁺ effector T cells as well as enhancing their effector capacity. IL-21R elevated expression likely provides signals mediating Cbx3/HP1γ-deficient CD8⁺ effector T-cell expansion in the absence of exogenous IL-21.

The rate of gene expression is governed in part by RNA Pol II initiation, elongation and/or chromatin remodeling. Pol II is phosphorylated at serine 5 (Pol II S5) during initiation of transcription while phosphorylation at serine 2 (Pol II S2) generally indicates chromatin remodeling concomitant with transcriptional elongation. To test whether alterations in transcriptional initiation, elongation and/or chromatin remodeling caused Lef1 and Il21r increased expression in Cbx3/HP1γ-deficient CD8⁺ effector T cells, ChIP-qPCR experiments were performed using ChIP-tested antibodies specific for Pol II S2 or S5, and H3K9me3. These results revealed augmented levels of Pol II S5 in or around TSSs of Lef1 and Il21r loci in Cbx3/HP1γ-deficient CD8⁺ effector T cells compared to control cells ( Figures 2A, B ). Similar levels of Pol II S2 density were observed in control and Cbx3/HP1γ-deficient CD8⁺ T cells ( Figures 2C, D ). The levels of H3K9me3 deposition at Lef1 locus was unaltered ( Figure 2E ). However, there was a general loss of H3K9me3 around the TSS of Il21r locus in Cbx3/HP1γ-deficient CD8⁺ effector T cells ( Figure 2F ). Thus, genetic deletion of Cbx3/HP1γ in CD8⁺ effector T cells results in enhanced transcription initiation at Lef1 and Il21r loci, with chromatin remodeling activity taking place in an extended region around the TSS of the Il21r locus. These changes likely underpinned the increased and sustained transcriptional activity at Lef1 and Il21r.

The increased/sustained expression of LEF-1 and IL-21R together with the enhanced effector capacity exhibited by Cbx3/HP1γ-deficient CD8⁺ effector T cells suggest they can persist to control tumor development. Thus, the ability of Cbx3/HP1γ-deficient CD8⁺ T cells to eradicate solid tumors was evaluated using the mouse ID8 or MOSE-L_TICv ovarian, B16 melanoma and NB-9464 NBL tumor models. Ovarian and NBL tumors have low mutation rates, no clear defining tumor-associated antigens (TAAs) and are minimally responsive to ICB (65–67). Accordingly, syngeneic ID8 or MOSE-L_{TIC v} ovarian tumor cells were injected intraperitoneally (IP) into control and Cbx3/HP1γ-deficient mice. Mice were monitored until abdominal distension was visible, which indicated increased ascites and mirrored metastasis observed in humans. Tumor growth and production of ascites were inhibited in Cbx3/HP1γ-deficient mice compared to controls or mice ectopically expressing Cbx3/HP1γ driven by the human T-cell-restricted Cd2 promoter (Cbx3/HP1γ^Tg) ( Figures 3A, B and S2A, B ). As a result, Cbx3/HP1γ-deficient mice lived longer than controls ( Figure 3C ). Cbx3/HP1γ-deficient mice were equally effective in reducing B16 melanoma ( Figure 3D ) and NB-9464 NBL tumor burden ( Figure 3F ), leading to their increased survival compared to controls ( Figures 3E, G ). On day 120 after ID8 injection, there was an enrichment of CD8⁺NKG2D⁺ effector T cells and a decrease in CD4⁺CD25⁺FOXP3⁺ regulatory T cells (Tregs) in ascites from Cbx3/HP1γ-deficient mice compared to control or Cbx3/HP1γ^Tg animals ( Figures 3H, I and S2C, D ). Similarly, enrichment of CD8⁺NKG2D⁺ effector T cells and decrease in CD4⁺CD25⁺FOXP3⁺ Tregs were observed in B16 melanoma as well as NBL tumors ( Figures 3J–M and S2E–H ). Comparable frequencies of NK1.1⁺NKG2D⁺ or CD4⁺NKG2D⁺ T cells were recovered from tumors of control, Cbx3/HP1γ^Tg and Cbx3/HP1γ-deficient mice ( Figure S3 ). Expression of Ifng, Gzmb and Prf1 was elevated in B16 melanoma and NBL tumors excised from Cbx3/HP1γ-deficient mice compared to controls or Cbx3/HP1γ^Tg animals ( Figures S4A–F ). TUNEL staining revealed more apoptotic tumor cells (brown) in melanoma and NBL tumors from Cbx3/HP1γ-deficient mice compared to controls ( Figures 3N, O ). Genetic ablation of Prf1 in Cbx3/HP1γ-deficient animals led to uncontrolled B16 melanoma growth and decreased TUNEL positivity indicative of reduced tumor-cell apoptosis ( Figures 4SG, H ) while NBL tumor burden was incompletely inhibited and tumor-cell death was readily detected ( Figures S4I, J ). By contrast, Ifng deletion resulted in uncontrolled NBL tumor growth in Cbx3/HP1γ-deficient mice ( Figure S4K ). Thus, PRF1 and INF-γ likely mediate the killing of tumors in Cbx3/HP1γ-deficient mice. To show that Cbx3/HP1γ-deficient CD8⁺ effector T cells cause tumor rejection in vivo, CD8⁺ T cells (CD45.2⁺) were activated/differentiated in vitro (see Methods). On day 5 after activation/differentiation, CD8⁺ T cells were collected for adoptive T-cell therapy in tumor bearing congenic B6.SJL (CD45.1⁺) recipients ( Figure S5A ). Treatment with Cbx3/HP1γ-deficient CD8⁺ effector T cells alone resulted in a statistically significant decrease in B16 melanoma and NBL tumor burden compared to treatment with control cells ( Figures S5B, C ). Within B16 melanoma tumors, there was an enrichment of transferred Cbx3/HP1γ-deficient CD8⁺NKG2D⁺ effector T cells, not transferred control or endogenous CD8⁺ effector T cells ( Figures S5D, E ). Possible contaminating CD45.2⁺CD4⁺NKG2D⁺ and CD45.2⁺NK1.1⁺NKG2D⁺ T cells were not detected in tumors indicating that they were not the source of tumor killing ( Figures S5F–I ). These results suggested a role for NKG2D in tumor control by Cbx3/HP1γ-deficient CD8⁺ effector T cells. To test this hypothesis, B16 and NB-9464 tumor cells were injected subcutaneously into control and Cbx3/HP1γ-deficient mice. On day 2, a subset of tumor-bearing Cbx3/HP1γ-deficient mice were treated with the blocking/non-depleting anti-mouse NKG2D antibody HMG2D, and thereafter once per week. NKG2D blockade resulted in uncontrolled tumor growth and decreased survival of treated animals ( Figures S6A–D ), accompanied by the reduction of Cbx3/HP1γ-deficient CD8⁺NKG2D⁺ effector T cells in tumors, to a level similar to PBS-treated control mice ( Figures S6E, F ). NKG2D blockade did not affect CD4⁺NKG2D⁺ and NK1.1⁺NKG2D⁺ T-cell frequencies ( Figures S6G, H ). Together, our results demonstrate that Cbx3/HP1γ-deficient CD8⁺ effector T cells expressing NKG2D can persist and cause tumor rejection, irrespective of tumor mutation status. The engagement of NKG2D with its ligands expressed on tumor cells likely facilitates tumor killing by Cbx3/HP1γ-deficient CD8⁺ effector T cells through the enhanced production of PRF1, GrB and INF-γ. Additionally, B16 and NBL tumors have differential sensitivity to killing by PRF1 and IFN-γ, and surveying for their presence in the TME alone may not reveal the precise mechanism whereby tumor cells are eradicated. Furthermore, deletion of even one Cbx3/HP1γ allele is sufficient to control solid tumor growth suggesting that Cbx3/HP1γ is haploinsufficient.

CD8⁺ effector T cells homing to tumors is mediated primarily by the interaction between CCL2/CCR2 and CXCL9/CXCL10/CXCR3 chemokine/receptor pairs while CD4⁺ Tregs trafficking to tumors is induced mostly by CCL28/CCR10 and CXCL12/CXCR4 (68–77). Chemokines are generally produced by tumor-associated myeloid cells including dendritic cells (DCs) and macrophages (MØs) whereas their cognate receptors are expressed on T cells as well as other immune cells. It has been proposed that changes in the tumor chemokine/receptor landscape influence anti-tumor responses. Whether CD8⁺ effector T cells residing in the TME participate in remodeling the tumor chemokine landscape is not clearly established. RT-qPCR assays demonstrated that Ccl2/Ccr2 and Cxcl9/Cxcl10/Cxcr3 levels were higher in B16 melanoma tumors from Cbx3/HP1γ-deficient mice than those from controls ( Figures 4A, B ). In NBL tumors from Cbx3/HP1γ-deficient mice, Ccl2/Ccr2 and Cxl10 remained similar to those of controls whereas Cxcl9/Cxcr3 levels were elevated ( Figures 4C, D ). These results indicate that homing of Cbx3/HP1γ-deficient CD8⁺ effector T cells to B16 melanoma tumors is likely mediated by CCL2/CCR2 and CXCL9/CXCL10/CXCR3 interactions. Cbx3/HP1γ-deficient CD8⁺ effector T cells likely depend on CXCL9/CXCR3 engagement to invade NBL tumors. Compared to controls, Ccr10 and Cxcl12/Cxcr4 levels were decreased in B16 melanoma tumors from Cbx3/HP1γ-deficient mice ( Figures 4E, F ). In NBL tumors from Cbx3/HP1γ-deficient mice, Ccl28/Ccr10 and Cxcl12 expression was reduced ( Figures 4G, H ). Thus, CCR10 and CXCL12/CXCR4 may regulate homing of CD4⁺ Tregs to B16 melanoma tumors while invasion of NBL tumors by CD4⁺ Tregs is likely governed by CCL28/CCR10 and CXCL12. Decreased levels of CD4⁺ Treg-attracting chemokine/receptor pairs may be causing the diminished presence of these cells in the respective tumors. In short, the presence of Cbx3/HP1γ-deficient CD8⁺ effector T cells in the TME induces remodeling of the chemokine/receptor landscape that favors their optimal trafficking into tumors at the expense of CD4⁺ Tregs.

To establish that LEF-1 and IL-21R contribute to the control of tumor growth by Cbx3/HP1γ-deficient CD8⁺ effector T cells, compound mutant mice were created ( Figure S7A ). In the Cbx3-Lef1-deficient mouse, Lef1 and Cbx3/HP1γ deletion was restricted to CD8⁺ T cells; in the Cbx3-Il21r-deficient mouse, Cbx3/HP1γ ablation was restricted to CD8⁺ T cells while Il21r was deleted in all tissues. LEF-1 and IL-21R expression was abolished in deficient CD8⁺ T cells compared to control cells ( Figures 5A, B ). Loss of one Lef1 or Il21r allele showed a decrease in protein levels that were almost completely abrogated upon deletion of both alleles. Progenitor T-cell development proceeded normally in the thymus of compound mutant animals ( Figures S7B, C ). Mesenteric lymph nodes (mLNs) of compound mutant mice displayed normal frequencies of mature CD8⁺ and CD4⁺ T cells ( Figures S7D, E ). Normal ratios of naïve (CD44^–CD62L⁺), effector (CD44⁺CD62L^–) and memory (CD44⁺CD62L⁺) T cells were observed in the mLNs of compound mutant animals ( Figures S7F, G ). In the mLNs, IL-21R expression was restricted to the CD8⁺ effector (CD44⁺CD62L^–) T-cell population and was also dependent on gene dosage ( Figure S7H ). IL-21R was not detected on any CD4⁺ T-cell populations ( Figure S7I ). To determine LEF-1 contribution to tumor control, ID8 (ovarian), B16 (melanoma) or NB-9464 (NBL) tumor cells were injected into Lef1-deficient, Cbx3-Lef1-deficient and control mice. CD8⁺ T-cell-restricted Lef1 ablation resulted in uncontrolled ovarian, melanoma and NBL tumor growth ( Figures 5C, D, F ). Tumor burden in Lef1- and Cbx3-Lef1-deficient animals was higher than that observed for Cbx3/HP1γ-deficient and control mice. Because Il21r was deleted in all tissues, adoptive transfer of activated/differentiated CD8⁺ T cells from Il21r ^-/-, Cbx3-Il21r-deficient and control animals was performed. Melanoma and NBL growth proceeded unchecked in tumor-bearing B6.SJL congenic mice treated with control, Il21r ^-/- or Cbx3-Il21r-deficient CD8⁺ effector T cells compared to animals receiving Cbx3/HP1γ-deficient CD8⁺ effector T cells ( Figures 5E, G ). Our data establish that LEF-1 and IL-21R are necessary for the optimal control of ovarian, melanoma and NBL tumor growth. Moreover, LEF-1 expression in control CD8⁺ effector T cells is induced, yet they are less capable of persisting in tumors, suggesting there may be a threshold of expression level required for LEF-1 to function effectively. IL-21R expression is confined to the CD8⁺ T-cell effector population and its loss following deletion suggest that IL-21R function is essential to this population, which is critical for halting tumor development.

The inability of Lef1-, Il21r-, Cbx3-Lef1- and Cbx3-Il21r-deficient CD8⁺ T cells to halt tumor development suggests that LEF-1 and IL-21R are necessary for their persistence in tumors. Analyses of tumors from Cbx3-Lef1-deficient mice revealed that the frequencies of CD8⁺NKG2D⁺ T cells were reduced to control levels in all three tumors ( Figures 6A, B, D ). CD8⁺NKG2D⁺ T cells were also decreased to control levels in B16 tumors from Lef1-deficient mice ( Figure 6B ). There was no enrichment of CD8⁺NKG2D⁺ T cells in B16 or NBL tumors from animals treated with Cbx3-Il21r-deficient CD8⁺ effector T cells ( Figures 6C, E ). Likewise, CD8⁺NKG2D⁺ T cells were not augmented in B16 tumors from mice receiving Il21r^-/- CD8⁺ effector T cells ( Figure 6C ). CD8⁺TIM3⁺CXCR5⁺ (progenitor exhausted) or CD8⁺TIM3⁺CXCR5^- (terminally exhausted) T-cell populations were not perturbed ( Figure S8 ). The frequency of endogenous NK1.1⁺NKG2D⁺, CD4⁺NKG2D⁺ T and myeloid cells in all tumor types was similar to that of controls ( Figure S9 ). These results demonstrate that Lef1 and/or Il21r genetic ablation results in the exclusive loss of CD8⁺ effector T cells in all three tumor types. Our data underscore the fact that LEF-1 and IL-21R are requisite for the persistence of Cbx3/HP1γ-deficient CD8⁺ effector T cells in tumors with varied mutation loads that are ICB responsive (B16 melanoma) or non-responsive (ovarian and NBL).

To show that uncontrolled tumor growth resulted primarily from the loss of CD8⁺ T-cell effector capacity, RT-qPCR assays were done using tumor RNA samples from compound mutant and control mice. Prf1 and Gzmb levels in melanoma tumors from control and compound mutant mice as well as those treated with Cbx3-Il21r- or Il21r-deficient CD8⁺ effector T cells were low compared to Cbx3/HP1γ-deficient mice ( Figures 7A, B ). Prf1, Gzmb and Ifng expression in NBL tumors from Cbx3-Lef1-deficient mice and those treated with Cbx3-Il21r-deficient CD8⁺ effector T cells was diminished compared to Cbx3/HP1γ-deficient animals ( Figures 7C–E ). Next, we asked whether the loss of effector capacity is inherent to Lef1 and/or Il21r deficiency or is caused by extrinsic factors in the TME. RT-qPCR analyses showed that Prf1, Gzmb and Ifng levels were reduced in in vitro-generated CD8⁺ effector T cells from compound mutant mice compared to those from Cbx3/HP1γ-deficient cells ( Figures 7F–H ). Therefore, our findings demonstrate that LEF-1 and IL-21R are indispensable for maintaining Prf1, Gzmb and Ifng in tumors. The loss of CD8⁺ effector T cells likely contributes to the observed blunted effector activity observed in tumors from Lef1- and Cbx3-Lef1-deficient mice as well as those treated with Il21r- or Cbx3-Il21r-deficient CD8⁺ T cells.