Histone Deacetylase Inhibition Sensitizes PD1 Blockade–Resistant B-cell Lymphomas

Non-Hodgkin lymphomas (NHL) are a heterogeneous group of malignancies affecting lymphocytes. Collectively, NHL are the fifth most common cancers in the United States, and more than 90% of NHL are of B-cell origin (1). The standard treatment of NHL includes cytotoxic chemotherapy (cyclophosphamide, doxorubicin, vincristine) plus one steroid (prednisone) combined with mAb against CD20 (rituximab; R-CHOP; refs. 2, 3). For patients with diffuse large B-cell lymphoma (DLBCL), the most frequent type of NHL, the 10-year progression-free survival was 36.5% for R-CHOP; the rest of patients endure relapsed and/or refractory disease (4). PD1 blockade is effective in relapsed or refractory classical Hodgkin lymphomas and a subset of DLBCL (5, 6), however, most patients with B-cell lymphoma do not respond well to anti-PD1 (7). There are no good experimental models to study PD1 blockade resistance observed in human B-cell lymphomas. To address these issues, we employed a mouse mature B-cell lymphoma model, G1XP (8), which resembles human B-cell lymphomas, to test therapeutic strategies.

One mechanism mediating PD1 blockade resistance may be downregulation of MHC class I and II in tumors. MHC class I and II present peptides derived from self or foreign antigens to CD8⁺ and CD4⁺ T cells, respectively, thus, are essential for cancer cells to be recognized by antigen-specific CD8⁺ and CD4⁺ T cells. As such, down regulating MHCs reduces the immunogenicity of lymphomas. DLBCLs frequently harbor mutations or deletions inactivating β2-microglobulin (β2M; 29% of cases; ref. 9), an essential component of MHC class I complex, thereby preventing MHC class I expression on lymphomas. Apart from inactivating β₂M, additional undefined mechanisms result in the loss of MHC class I expression in >60% of DLBCL, suggesting that its selective loss during lymphomagenesis may mediate escape from immune surveillance (9). Diminished MHC class II expression also occurs frequently in DLBCLs and other types of mature B-cell lymphomas (10–15). MHC class II loss or low expression correlates with inferior survival in patients of DLBCL or classical Hodgkin lymphoma (10–12), or with poor prognosis in patients with DLBCL and primary mediastinal B-cell lymphoma (PMBCL) following chemotherapies (10, 16–18). Whether MHC expression influences the sensitivity of B-cell lymphomas to PD1 blockade remains unclear.

MHC class I downregulation also occurs in other types of cancers including colon, lung, and breast cancers (19–23). MHC class II loss correlated with a decrease of tumor-infiltrating T cells and an increase of metastatic potential of colorectal cancers (24). MHC class II expression serves as a favorable prognostic marker in colorectal carcinomas (25). Melanoma-specific MHC class II expression predicts response to anti-PD1/PD-L1 therapy (26). In contrast, mutations in the MHC class I pathway correlated with immune checkpoint inhibitor (ICI) resistance (27). Thus, restoring MHC class I and II expression on tumors has been suggested to benefit chemotherapy and immunotherapy (14, 22, 27). Reversible downregulation of MHC class I and class II, so-called “soft lesions,” can be mediated by epigenetic modifications (23); moreover, “soft lesions” dominate MHC class I defects in human cancers (23, 28, 29). Most “soft lesions” of MHC class II are mediated by decreased histone acetylation rather than DNA hypermethylation (23, 28, 29).

Histone acetylation status is regulated by a dynamic equilibrium between histone acetyl transferases and histone deacetylases (HDAC). HDAC expression is often dysregulated in solid tumors and hematologic cancers, including B-cell lymphomas (30–33). Four FDA-approved HDAC inhibitors (HDACi), vorinostat, romidepsin, panobinostat, and belinostat, are used for treating hematologic cancers (34–36). Despite the success of these HDACis, each has liabilities including poor isoform selectivity, marginal potencies toward relevant isoforms, narrow therapeutic indices, and/or nonoral delivery routes. These drawbacks have spurred a search for alternatives with improved biological, physiochemical, and therapeutic properties. A natural product, largazole, was reported as a potent class I HDACi (37, 38); however, largazole has relatively poor physiochemical properties and is not amendable to large-scale chemical manufacturing. Thus, we initiated a lead optimization program resulting in the discovery of next-generation largazole derivatives OKI-005 and OKI-179 (39).

The efficacy of combining ICI with HDACis was tested previously in animal models or early-stage clinical studies (40–42). However, there are challenges in translating these observations into clinically useful therapeutics (43, 44). It remains poorly understood why combined ICI and HDACi worked in certain tumors but failed in others and what biomarkers can predict treatment efficacy. In addition, it remains poorly understood how HDACis affect the immunogenicity of cancer cells (44). To address these questions, we need syngeneic immunocompetent mouse models of B-cell lymphomas with altered immunogenicity. HDACis have been shown to recover or enhance expression of MHC class I and II in different types of cancers (28, 41, 45–49). However, it remains unknown whether HDACi sensitizes B-cell lymphomas to PD1 blockade by regulating MHC expression. Finally, although HDACi are known to modulate antitumor immunity, their net effects are dependent on the specific inhibitors used and the HDACs they target (50). It remains unknown what types of HDACi are suitable for combination with ICI.

In this study, we employed an HDACi (OKI-179) and G1XP lymphoma (8), created by lineage-specific deletion of a DNA repair gene, Xrcc4, and Trp53 in germinal center (GC) B cells, because most human mature B-cell lymphomas are derived from GC or post-GC B cells (51). We show that G1XP lymphomas downregulated their MHC expression and resisted anti-PD1. We dissected the mechanisms by which the HDACi OKI-179 sensitizes the resistant lymphomas to anti-PD1.

Littermate controls of G1XP or BALB/c mice (6–8 weeks) were injected subcutaneously at both flanks with 1 × 10⁶ G1XP or A20 lymphoma cells. When tumor size reached 200 to 350 mm³, recipient mice were randomized into 4 groups and treated three times every other day with vehicle control (intraperitoneal injection of PBS and oral gavage of 0.1 mol/L citrate buffer), anti-PD1 (10 mg/kg/dose; BioXcell) via intraperitoneal injection, OKI-179 (60 mg/kg/dose) via oral gavage, or both anti-PD1 and OKI-179. When tumor size reached 2 cm in any dimension or other humane endpoints were met (e.g., necrotic tumors), mice were euthanized in accordance with institutional guidelines. Mice were maintained under specific pathogen–free conditions in the vivarium facility of University of Colorado Anschutz Medical Campus (Aurora, CO). Animal work was approved by the Institutional Animal Care and Use Committee (IACUC) of University of Colorado Anschutz Medical Campus (Aurora, CO).

Tumors were harvested from tumor-bearing mice. Tumor weight was measured before dissociation and tumors were processed into single-cell suspension. Tumor-infiltrating lymphocytes (TIL) were stained with antibodies against CD45, B220, CD3, TCRβ, CD4, CD8, and CD69. Antibodies used for flow cytometry were listed in Supplementary Table S1. Dead cells were excluded by Live/Dead Fixable Green Dead Cell Stain Kit (Invitrogen). BD Fix/Permeabilization buffer was used for intracellular staining of IFNγ and granzyme B in TILs. Equal numbers of tumors were cultured in vitro for 6 hours in the presence of 50 ng/mL phorbol 12-myristate 13-acetate (Sigma-Aldrich), 1 μg/mL ionomycin (Sigma-Aldrich), and 5 μg/mL BFA (BioLegend). Data were acquired on BD Fortessa or BD FACSCalibur and analyzed with FlowJo software V10.

G1XP lymphomas were generated, established, and cultured as described previously (8). A20 lymphoma cells were obtained from cell line vendor ATCC in 2017. OCI-LY1, OCI-LY3, OCI-LY7, and SU-DHL-16 were obtained from Dr. Wing C. (John) Chan (City of Hope Medical Center, Duarte, CA) in 2016 and cultured as described previously (52). The cell line authentication and Mycoplasma testing were performed by Molecular Biology Service Center at the Barbara Davis Center (University of Colorado, Anschutz Medical Campus, Aurora, CO) in 2019. The cells were tested and reauthenticated by PCR assays as described (http://www.barbaradaviscenter.org/). The number of passages between thawing and use in the described experiments ranged from two to five.

Lymphoma cells were cultured at 0.5 × 10⁶/mL and treated with vehicle control, OKI-005, or OKI-179 at indicated concentrations for 24 or 48 hours. Cells in triplicates were treated as described above, fixed in 70% ethanol, and stained with PI and anti-pH3 as described previously (52). Cell cycles were determined by flow cytometry (FL1-H/FL2-A). Splenic T cells were isolated from wt B6-naïve mice by Negative Selection Kit (Stemcell Technologies), cultured with mouse T-Activator CD3/CD28 Dynabeads (Thermo Fisher Scientific), and collected 3 days after culture for flow cytometry analysis. Murine primary B cells were isolated from spleens of syngeneic mice by negative selection kit (Stemcell Technologies), cultured with mouse IL4 (20 ng/mL) and anti-CD40 (1 μg/mL) as described previously (53) and collected 4 days after culture for flow cytometry or OKI-179 treatment. Human primary B cells were isolated from human cord blood by negative selection kit (Stemcell Technologies), cultured with IL4 (10 ng/mL), IL21 (10 ng/mL), and anti-CD40 (1 μg/mL) and collected 5 days after culture for flow cytometry.

G1XP, OCI-Ly7, and murine primary B cells were treated with 500 nmol/L of OKI-179 for 16 hours, respectively, and total RNA was isolated by TRIzol (Roche). Isolated RNA was subject to library preparation and sequenced with Illumina NovaSEQ 6000 platform by the genomic core facility at University of Colorado Anschutz Medical Campus (Aurora, CO). Primary antibodies used in Western blot analysis were listed in Supplementary Table S1. Secondary horseradish peroxidase–conjugated anti-mouse and anti-rabbit were purchased from Jackson ImmunoResearch. Protein bands were developed by ECL Western blotting detection reagents (GE Healthcare) according to kit instructions. Cell cycles were determined as described previously (52).

Guide RNA (gRNA) oligos targeting mouse B2M gene were designed by online software (http://crispr.mit.edu/). Four gRNA oligos were synthesized and annealed to generate two DNA fragments (1-For 5′-caccgAGTATACTCACGCCACCCAC-3′, 1-Rev 5′-aaacGTGGGTGGCGTGAGTATACTc-3′; 2-For 5′-caccgCGTATGTATCAGTCTCAGTG-3′, 2-Rev 5′-aaacCACTGAGACTGATACATACGc-3′). The two DNA fragments were cloned into pX458 vector, respectively, according to the protocol described previously (54). Both newly constructed pX458 plasmids were cotransfected into G1XP or A20 lymphomas with Nucleofector Kit V (Lonza). Single clones lacking MHC class I were screened by flow cytometry with an anti-mouse MHC class I antibodies and confirmed by PCR with primers (SF 5′-TTATCCAGAGTAGAAATGGAACAGGG-3′, SR 5′-GTATTCTCTAACAATCTCAGTATGC-3′). Deleted fragments were cloned into pGEM-T vector (Promega) and sequenced by MCLAB. G1XP B2M^−/− lymphoma harbored a deletion on each allele of B2M gene, respectively (deletion on allele 1: 3211–3219, 3381–3401; allele 2: 3214–3222, 3386–3406 of B2M genomic DNA, NC_000068.7: 122147687-122153082, total length of 5396 bp). A20 B2M^−/− lymphoma cells harbored the same deletion on both alleles of the B2M gene (deletion location: 3394–3571; NC_000068.7:122147687-122153082, total length of 5396 bp).

Total RNA was isolated from G1XP lymphomas, and used for cDNA synthesis with a Reverse Transcription Kit (Promega). H2-K1 cDNA was amplified by PCR with primers (For 5′-GACAGAATTCATGGTACCGTGCACGCTGC-3′; Rev 5′-GCAGTCTAGATCACG CTAGAGAATGAGGGT-3′). The amplified H2-K1 fragment was subcloned into pcDNA3.1(+) vector. The newly constructed pcDNA3.1(+) plasmid was transfected into G1XP lymphomas using Mouse B Cell Nucleofector Kit (Lonza). After culturing for 2 days, transfected cells were selected with 600 μg/mL neomycin until neomycin-resistant foci were identified. Single G1XP clones with varied MHC class I or II expression were identified by flow cytometry.

Our efforts toward the discovery of a largazole-derived HDACi were prompted by the lackluster physiochemical properties and synthetic challenges of largazole (55, 56). These efforts resulted in the invention of OKI-005 and OKI-179 (39), first- and second-generation prodrug candidates, respectively, derived from the same parent thiol congener. IC₅₀ values were obtained with OKI-179 thiol (the active parent form of OKI-179 and OKI-005) under cell-free, biochemical assay conditions using purified enzymes, compared favorably with other HDACi (Supplementary Table S2). OKI-179 thiol inhibits class I, IIb, and IV HDACs (Supplementary Table S2).

To determine the pharmacokinetics of OKI-179 in vivo, OKI-179 was administered to mice via oral gavage. OKI-179 converted to its active metabolite rapidly in murine plasma, with similar pharmacokinetic profiles observed in both female and male mice (Supplementary Fig. S1A–S1D). Maximum concentration (C_max) in male and female mice ranged from 337 to 4,980 ng/mL (Supplementary Fig. S1A). The time to reach C_max in vivo (T_max) was 0.083 hours (for dose of 25 mg/kg) and 0.25 hours (for dose of 50 mg/kg and 100 mg/kg). The half-life [t_1/2(h)] is shown in Supplementary Fig. S1D. On the basis of C_max in vivo, we decided to use the following concentrations of OKI for in vitro treatment: 250 nmol/L, 500 nmol/L, and 1 μmol/L.

HDACi treatment is cytotoxic to cancer cells (57). To examine whether OKI-005 or OKI-179 exhibits cytotoxic effects on tumors, G1XP lymphomas were treated with OKI-005 or OKI-179 in vitro. Both OKI-005 and OKI-179 inhibited the growth of G1XP lymphomas in a dose-dependent manner (Fig. 1A). OKI-179 treatment arrested tumor cells at G₀–G₁ phases and reduced the percentage of cells at S-phase (Supplementary Fig. S2A and S2B). To test the effects of OKI on mitotic entry, we assessed the phosphorylation of histone3 (pH3), which serves as a marker for M-phase entry because chromosome condensation requires H3 phosphorylation. We found that pH3 was reduced upon treatment of OKI-179 (Fig. 1B and C) and OKI-005 (Supplementary Fig. S2C and S2D), demonstrating OKI-treated lymphomas were inhibited for M-phase progression. Finally, cleaved caspase-3 production, a marker for apoptosis, was induced in G1XP lymphomas upon treatment of OKI-179 (Fig. 1D–F) and OKI-005 (Supplementary Fig. S2E and S2F). Overall, OKI-005 and OKI-179 inhibit tumor cell growth by inducing cell-cycle arrest and apoptosis.

OKI-179 and OKI-005 are congeners because they have identical drug structures but different prodrug structures. OKI-179 is designed to have better drug exposure in vivo and better pharmaceutical quality in terms of chemistry, manufacturing, and controls. OKI-179 was developed on the basis of OKI-005 and is more suitable for in vivo studies. OKI-179 received IND approval from FDA and entered phase I clinical trials in 2019. Thus, we employed OKI-179 for subsequent analyses because it will be more translatable to clinical settings.

We transplanted G1XP lymphomas (8) into syngeneic wild-type (WT) recipient mice that developed secondary tumors approximately 2 weeks after inoculation. We harvested secondary tumors and analyzed TILs. Both CD8⁺ and CD4⁺ TILs upregulated PD1 compared with control splenic T cells (Fig. 1G). PD1 is a coinhibitory receptor associated with T-cell exhaustion, suggesting that G1XP lymphomas might be sensitive to anti-PD1 treatment. We inoculated WT recipients with G1XP lymphomas, when tumor size reached about 250 to 350 mm³, we randomized recipient mice into four groups and treated them with vehicle control, anti-PD1, OKI-179, or OKI-179 plus anti-PD1 (combo). We harvested tumors on day 21 after inoculation for TIL analysis (short-term; Fig. 1H and I) or observed the cohorts for tumor growth (long-term; Fig. 1J and K). In both scenarios, anti-PD1 failed to suppress G1XP lymphoma growth (Fig. 1H–K). OKI-179 alone showed an inhibitory effect on G1XP lymphoma growth at early time points, whereas combo treatment inhibited tumor growth (Fig. 1H–K). When data were analyzed together from three independent experiments in which treated recipients were monitored up to 168 days after tumor inoculation, approximately 25% of the recipients from combo group survived, whereas all other control groups were terminated (according to IACUC guideline) or died due to tumor growth (Fig. 1L). These data show that G1XP lymphomas resist anti-PD1, and OKI-179 treatment sensitizes lymphomas to anti-PD1.

To reveal the mechanisms of PD1 blockade resistance, we tested whether G1XP lymphomas downregulated their MHCs. We established several G1XP lymphoma lines and found that G1XP lymphomas downregulated MHC class I expression compared with activated primary B cells (Fig. 2A, left), which were used as controls because they mimic the developmental stage of G1XP lymphomas. MHC class II expression on G1XP lymphomas was also reduced compared with activated primary B cells (Fig. 2A, right). G1XP lymphomas expressed less MHC class I or class II than other mouse B-cell lymphoma lines, such as CH12 and A20 (Fig. 2B). We conclude that G1XP lymphomas provide a model for studying tumors with reduced immunogenicity.

MHC downregulation can be reversible or irreversible. We tested our HDACis, OKI-179 and OKI-005, for their ability to affect MHC expression on G1XP lymphomas. Lymphomas were cultured in the presence of OKI-179, OKI-005, or vehicle control, and both OKI-179 and OKI-005 upregulated MHC class I and class II expression in a dose-dependent manner in vitro (Fig. 2C; Supplementary Fig. S3). To test whether OKI-179 treatment affects MHC class I and II expression on G1XP lymphomas in vivo, we inoculated G1XP lymphomas into syngeneic recipient mice. Tumor-bearing recipients were randomized into four groups and treated as described above. OKI-179 or OKI-179/anti-PD1 treatment caused lymphomas to upregulate MHC class I and class II in vivo (Fig. 2D–F). These data indicate that OKI-179 increases the immunogenicity of cancer cells. Downregulation of MHC class I and II in G1XP lymphomas was reversible; thus, it likely occurs via epigenetic mechanisms.

Because of the increased MHC expression on G1XP lymphomas upon OKI-179 treatment, we tested whether the treatment influences CD8⁺ and CD4⁺ TILs. Tumors were harvested from recipients treated with vehicle control, OKI-179, anti-PD1 or combo, respectively. We isolated tumors and TILs and performed flow cytometry analysis 21 days after tumor inoculation. This time point was chosen because control group usually had to be terminated due to the size of tumors exceeding institutional guideline. We also cultured the TILs for 6 hours, then examined IFNγ production in CD4⁺ and CD8⁺ TILs and granzyme B in CD8⁺ TILs. Combo treatment increased the percentage and number of CD4⁺ (Fig. 3A–C) and CD8⁺ TILs (Fig. 3D–F). Furthermore, combo treatment activated TILs and increased IFNγ production in both CD4⁺ (Fig. 3G and H) and CD8⁺ TILs (Fig. 3I and J) and granzyme B production in CD8⁺ TILs (Fig. 3K and L). OKI-179 treatment alone also increased IFNγ production in both CD4⁺ and CD8⁺ TILs (Fig. 3G and I). Combo treatment also induced CD69 expression in both CD4⁺ (Supplementary Fig. S4A and S4B) and CD8⁺ TILs (Supplementary Fig. S4C and S4D). Thus, combined OKI-179/anti-PD1 treatment promoted CD4⁺ and CD8⁺ TIL activation.

To determine whether tumor-derived MHC class I is required for efficacy of combo treatment, we employed the CRISPR/Cas9 technique to delete the B2M gene in G1XP lymphoma. Deletion of the B2M gene was confirmed by PCR and Sanger sequencing (Fig. 4A). Consistently, MHC class-I was absent in G1XP B2M^−/− lymphoma (Fig. 4B). WT G1XP lymphoma clone was transplanted into recipient mice that were treated as described above and exhibited sensitivity to single OKI-179 or combo treatment (Fig. 4C), similar to what was observed in parental G1XP lymphomas. However, single OKI-179 or combo treatment did not inhibit the growth of G1XP B2M^−/− lymphomas (Fig. 4D). These data show that tumor-derived MHC class I is essential for the efficacy of combo and single OKI-179 treatment.

Next, we investigated whether OKI-179-mediated MHC class I upregulation is sufficient, namely, whether OKI-179 renders lymphomas sensitive to anti-PD1 just by upregulating MHC class I in lymphomas. We ectopically re-expressed MHC class I in G1XP lymphomas and tested whether tumor's sensitivity to anti-PD1 is affected by the amount of reintroduced MHC class I. We choose to ectopically reexpress H2-K1, an MHC class I α chain gene, in G1XP lymphomas, because H2-K1 is one of the most abundantly expressed MHC class I alleles and OKI-179 significantly upregulated H2-K1 expression in G1XP lymphomas (Supplementary Fig. S5A). Stable clones were selected that expressed more or less MHC class I (Fig. 4E).

Comparing with MHC class I^low clone (E8), MHC class I^high clone (C10) expressed more MHC class I but showed similar MHC class II expression in vitro and in vivo (Fig. 4E; Supplementary Fig. S5B). However, clone E8 and C10 were equally insensitive to anti-PD1 (Fig. 4F and G). These data suggest that upregulation of MHC class I is not sufficient for rendering tumors sensitive to PD1 blockade. In addition, we compared the MHC class II expression among different G1XP lymphoma clones, and found that clone D1 presented more MHC class II expression than E8 and C10, but low MHC class I expression in vitro and in vivo (Fig. 4E; Supplementary Fig. S5B). Anti-PD1 treatment significantly inhibited the growth of clone D1 (Fig. 4H). Our data indicate that more MHC class II correlates with greater tumor sensitivity to PD1 blockade.

To further test whether treatment efficacy of OKI-179 or anti-PD1 is dependent on MHC class I, we employed the A20 lymphoma model known to express high amounts of both MHC class I and II (Fig. 2B). We transplanted A20 lymphomas into syngeneic recipient mice (WT Balb/c) and found that CD8⁺ and CD4⁺ TILs upregulated PD1 compared with control splenic T cells (Fig. 5A). Single OKI-179 or anti-PD1 treatment significantly increased recipient survival (Fig. 5B) and inhibited tumor growth (Fig. 5C). Although combo treatment appeared to be more effective, no significant differences were detected between single and combo treatment (Fig. 5B and C). Thus, our data suggest that the high immunogenicity of A20 lymphomas correlates with high sensitivity to monotherapies or combo treatment.

To test the role of MHC class I in A20 lymphoma model, we generated B2M-KO A20 lymphomas using CRISPR/Cas9 approach (Supplementary Fig. S6A). A20 B2M^−/− lymphoma cells harbored the same deletion on both alleles of the B2M gene (Supplementary Fig. S6A). Because of saturated expression of MHC class I and II on A20 lymphoma, OKI-179 had no effect on MHC class I and II expression (Supplementary Fig. S6B). OKI-179 was unable to induce the expression of MHC class I on A20 B2M^−/− lymphoma (Supplementary Fig. S6C). A20 B2M^−/− lymphomas were transplanted into syngeneic recipient mice that were treated as described above. In contrast to A20 WT lymphomas, A20 B2M^−/− lymphomas were resistant to single or combined treatment of OKI-179 and anti-PD1 (Fig. 5D and E). These data demonstrate that the sensitivity of tumor cells to OKI-179 or anti-PD1 treatment requires tumor-derived MHC class I expression. OKI-179 treatment inhibited proliferation and mitotic entry of A20 WT and B2M^−/− lymphomas (Supplementary Fig. S7A–S7F), and induced apoptosis of A20 WT lymphomas (Supplementary Fig. S7G). However, the treatment effects of OKI-179 were abolished in A20 B2M^−/− lymphomas (Fig. 5E), demonstrating that OKI-179′s efficacy depends on its immunoregulatory effects instead of direct cytotoxic effects.

To test whether different HDACi have varied effects on T-cell proliferation and activation, we purified T cells and stimulated them with anti-CD3/anti-CD28 beads in the presence of vehicle control or increasing concentration of HDACi for 3 days in vitro. OKI-179 and vorinostat did not affect the proliferation of CD4⁺ or CD8⁺ T cells (Fig. 6A and B; Supplementary Fig. S8A). However, panobinostat inhibited both CD4⁺ and CD8⁺ T-cell proliferation (Fig. 6C and D). OKI-179 and vorinostat had no effect on IFNγ production by CD4⁺ or CD8⁺ T cells (Supplementary Fig. S8B and S8C).

To determine whether other HDACi also induce MHC class I, we compared vorinostat with OKI-179 at the same dose range, and found that no induction of MHC class I occurred with vorinostat up to 1 μmol/L, whereas OKI-179 had a steady dose-dependent increase (Fig. 6E). Vorinostat induced less expression of MHC class II than OKI-179 (Fig. 6F). This is probably due to the fact that vorinostat is not as potent in HDAC inhibition as OKI-179 (Supplementary Table S2). Panobinostat has a comparable class I HDAC inhibition potency to OKI-179 (Supplementary Table S2) and it can induce MHC class I and II expressions similarly to OKI-179 (Supplementary Fig. S8D). We conclude that different HDACi exhibit variable effects on T cells and tumors, suggesting that not all HDACis are suitable for combination with PD1 blockade.

PD-L1⁺ cancers might be more sensitive to PD1 blockade (58, 59), such as Hodgkin lymphomas with PD-L1 amplification (6, 60, 61), suggesting that PD-L1 upregulation may serve as a biomarker to predict PD1 blockade sensitivity. Our RNA-Seq data showed that OKI-179 treatment also upregulated PD-L1 expression in both G1XP and OCI-Ly7 lymphomas (Fig. 7A and B). However, OKI-179 had no effect on PD-L1 expression in activated primary mouse B cells (Fig. 7C). Furthermore, PD-L1 expression was upregulated in G1XP lymphomas by OKI-179 in a dose-dependent manner (Fig. 7D). In contrast, the expression of costimulatory factors, including CD83, CD86, OX40L and CD137L, were not altered by OKI-005 or OKI-179 (Supplementary Fig. S9A–S9D). OKI-179 inhibited the growth of human B-cell lymphoma lines in a dose-dependent manner (Supplementary Fig. S10).

To establish a general concept of OKI-179-mediated MHC upregulation, we treated various human B-cell lymphoma lines with OKI-179. Human OCI-Ly7 lymphomas downregulated their HLA-A, B, and C but not HLA-DP, DQ, and DR compared with human primary B cells (Fig. 7E). OKI-179 treatment enhanced the expression of HLA-A, B, and C in OCI-Ly3 and OCI-Ly7, whereas OKI-179 had no effect on OCI-Ly1 and little effect on SU-DHL-16 cell line (Fig. 7F). OKI-179 upregulated HLA-DP, DQ, and DR in OCI-Ly3 cells, but not in OCI-Ly1, OCI-Ly7, and SU-DHL-16 cells (Fig. 7G). Thus, we conclude that different human B-cell lymphoma lines respond differently to OKI-179 treatment in terms of MHC upregulation.

We tested the therapeutic efficacy of combined OKI-179/anti-PD1 treatment and found that: (i) OKI-179 sensitizes G1XP lymphomas to anti-PD1 by enhancing tumor immunogenicity; (ii) sensitivity to single or combined treatment required tumor-derived MHC class I and positively correlated with MHC class II expression in tumors; (iii) the durable antitumor effects of OKI-179 depend on antigen-specific CD8⁺ T-cell-mediated immune responses; and (iv) different HDACis exhibited distinct effects on tumors and T cells, yet the same HDACi could differentially affect HLA expression in different human B-cell lymphomas. Thus, our studies highlight immunologic effects of HDACis on antitumor responses and suggest that optimal treatment efficacy requires personalized design and rational combination based on prognostic biomarkers (e.g., MHCs) and individual profiles of HDACis.

Human B-cell lymphomas often downregulate MHC expression (9–15) and resist anti-PD1 (7). Lacking proper models hinders our effort to understand the underlying mechanisms of these observations. G1XP lymphoma is a syngeneic mouse model for mature B-cell lymphoma that downregulates MHCs and resists anti-PD1. This model provides an opportunity to elucidate how altering lymphoma immunogenicity influences cancer immunotherapy outcome. Different lymphomas responded differentially to single anti-PD1 treatment, with a correlation to their MHC expression. For instance, MHC high–expressing A20 lymphomas are sensitive to anti-PD1, whereas MHC low–expressing G1XP lymphomas resist anti-PD1. Consistently, acquired resistance to PD1 blockade associates with deletional mutation in the B2M gene in a relapse sample of a patient with melanoma (62). PD1 blockade–resistant tumors downregulated MHC class I in a murine lung cancer model (63). Although tumor-derived MHC class I is essential for sensitivity to single or combo treatment, increasing MHC class I alone did not render tumors sensitive to anti-PD1. In contrast, higher expression of tumor-derived MHC class II correlated with increased sensitivity to anti-PD1. Studies show that therapeutic effects of PD1 blockade correlate with the expression of MHC class II but not MHC class I in patients with melanoma (26, 64). Collectively, these studies suggest that sensitivity to PD1 blockade needs high MHC class II expression.

MHC downregulation in G1XP lymphomas is reversible and rescued by epigenetic agents, such as HDACi. However, OKI-179 is unable to restore irreversible MHC class I downregulation in B-cell lymphomas. When tumor-derived MHC class I is removed genetically, the therapeutic effects of single OKI-179 or combo treatment were abolished. Thus, we consider OKI-179 to be a personalized drug suitable for treating tumors with reversible MHC downregulation. We suggest that MHC expression might serve as a predictive biomarker for treatment efficacy of OKI-179. Aside from classical MHC class I, B2M is also required for expression of other nonclassical MHC molecules such as H2-M5 (65), involvement of which we cannot rule out. However, our RNA-seq data did not detect changes in the B2M-associated molecules upon OKI-179 treatment, suggesting that these nonclassical MHC class I molecules are not necessary for mediating responses to OKI-179.

HDACis modulate antitumor immunity (43). For instance, anticancer effects of vorinostat require immune system (66, 67). However, not all HDACis are created equal and their net effects are dependent on the specific inhibitors used and the HDACs they target (44, 50). Typically, inhibitors are developed on the basis of on-target activity; however, off-target activity could result in undesirable side effects or toxicity. For example, panobinostat and romidepsin are potent HDACis but are also toxic. The narrow window for therapeutic success makes such HDACis unsuitable for combined therapies with ICIs. Pan-HDACi such as vorinostat target all 11 HDACs, which is problematic because class IIa HDACs have opposing functions to class I HDACs in modulating immune responses (e.g., Treg; ref. 68). In addition, not all HDACis are equally potent at reprogramming cancer epigenome. For example, vorinostat did not upregulate MHC class I in G1XP tumors. A large dose of vorinostat will be needed to achieve MHC class I induction in vivo, quite a challenge given its poor drug metabolism and pharmacokinetics profile. Thus, when interpreting the immunologic effects of HDACis, we must be aware of the dose, cellular context, and selectivity of agents, and the impact of HDACi on both tumors and T cells.

OKI-179 has favorable properties compared with previously reported HDACis, with its selectivity, oral availability with a broad therapeutic window, direct inhibition of B-cell lymphoma proliferation, and no toxic effects on T-cell proliferation and activation, providing improved therapeutic effects due to its potent immune-enhancing activity. OKI-179 can also effectively inhibit HDAC3, which has been implicated in regulating MHC class II expression in B-cell lymphomas (69). Thus, OKI-179 is a promising agent for treating cancers with reversible MHC downregulation, such as B-cell lymphomas.

OKI-179 upregulates PD-L1 in G1XP and human B-cell lymphomas. PD-L1 interacts with PD1 expressed on CD4⁺ and CD8⁺ TILs, thereby leading to inhibition of effector functions and exhaustion of T cells (70). To overcome the detrimental effects of OKI-179–mediated PD-L1 upregulation on antitumor immunity, anti-PD1 needs to be employed together with OKI-179. Indeed, combo treatment more effectively inhibits B-cell lymphomas, providing a rationale for developing combinatorial therapy using epigenetic agents and ICIs. HDACis also upregulate PD-L1 and augment therapeutic efficiency of PD1 blockade in melanoma (71) or lung cancers (42), suggesting the applicability of such combined strategies to other types of cancers.

PD-L1⁺ cancers might be more sensitive to PD1 blockade (58, 59). PMBCL and Hodgkin lymphomas harbor recurrent chromosomal translocations or amplifications of PD-L1 and PD-L2, leading to overexpression of PD-L1 and PD-L2 (60, 61). Consistently, Hodgkin lymphomas are very sensitive to PD1 blockade. Clinical trials of PD1 therapy showed promising results in PMBCL (6, 72). These data suggest that PD-L1 upregulation may serve as a biomarker to predict combo treatment sensitivity. Prior studies show that tumor-derived PD-L1 is not required for the efficacy of anti-PD-L1 treatment because host cells still express PD-L1 (73). It remains to be determined whether altering tumor-derived PD-L1 will affect the efficacy of combo treatment in our B-cell lymphoma model. On the other hand, HDACi-mediated PD-L1 upregulation may explain why these agents generally fail to treat cancers as a single agent. Our studies may provide insights into why HDACis alone failed and why there should be a renewed emphasis on the effects of HDACi on PD-L1 upregulation.

G. Zhang is a board member, has ownership interest (including stock, patents, etc.), and is a consultant/advisory board member for OnKure. A.D. Piscopio has ownership interest (including stock, patents, etc.) in OnKure. X. Liu is an officer, has ownership interest (including stock, patents, etc.), and is a consultant/advisory board member for OnKure. No potential conflicts of interest were disclosed by the other authors.

Conception and design: X. Wang, G. Zhang, X. Liu, J.H. Wang

Development of methodology: X. Wang, J.H. Wang

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): X. Wang, B.C. Waschke, R.A. Woolaver, G. Zhang, J.H. Wang

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): X. Wang, B.C. Waschke, R.A. Woolaver, A.D. Piscopio, J.H. Wang

Writing, review, and/or revision of the manuscript: X. Wang, B.C. Waschke, Z. Chen, A.D. Piscopio, X. Liu, J.H. Wang

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Z. Chen, A.D. Piscopio

Study supervision: Z. Chen, J.H. Wang

Other (supplied test article): A.D. Piscopio

We thank Stephanie Cung, Amanda M. Perras, and Erin Kitten for technical help. We apologize to those whose work was not cited due to length restrictions. This work was supported by University of Colorado School of Medicine and Cancer Center Startup Funds (to J.H. Wang), R21-CA184707, R21-AI110777, R01-CA166325, R21-AI133110, and R01-CA229174 (to J.H. Wang), a fund from Cancer League of Colorado and American Cancer Society (ACS IRG #16-184-56; to Z. Chen), R01GM113141 and R01AR068254 (to X. Liu). R.A. Woolaver is supported by a NIH F31 Fellowship (F31DE027854). X. Wang is supported by an AAI Careers in Immunology Fellowship.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.