The epigenetic immunomodulator, HBI-8000, enhances the response and reverses resistance to checkpoint inhibitors

The MC38 and CT26 syngeneic murine colon carcinomas were obtained from ATCC (Manassas, VA), and the A20 cells were obtained from Covance (Princeton, NJ). Cells were passaged and maintained using the protocols provided by the vendors. HBI-8000 was supplied by HUYA Bioscience International. HBI-8000 (HUYA Bioscience International) was formulated in 10% hydroxypropyl- β-cyclodextrin and 10% propylene glycol in deionized water, pH 2.5. Dosing solutions were prepared fresh weekly and stored at 4 °C. Animals were dosed orally daily with 50 mg/kg HBI-8000 for 21 days.

Monoclonal antibodies (mAbs) to mouse PD-1 (clone RPM-14), PDL-1 (clone (10F.9G2), and CTLA-4 (clone 9H10) were purchased from Bio-X-Cell (West Lebanon, NH). Antibody dosing solutions were prepared in sterile phosphate-buffered saline on each dosing day, and stored at 4 °C. Mice were intraperitoneally injected with the PD-1 antibody (Ab) or PD-L1 Ab (10 mg/kg) twice weekly for 3 weeks. CTLA-4 Ab (2.5 mg/kg) was administered intraperitoneally on days 1, 4, and 7.

All animal research studies were approved and overseen by the Institutional Animal Care and Use Committees of Charles River (MC38, CT26). All mice obtained from Charles River (Morrisville, NC) were female and 8 weeks old when the tumors were implanted. For MC38 tumors, C57BL/6 mice were implanted subcutaneously in the right flank with 1 × 10⁶ MC38 cells (0.1-mL cell suspension). For CT26 tumors, BALB/c mice were injected subcutaneously in the right flank with 3 × 10⁵ CT26 tumor cells (0.1-mL cell suspension). For A20, BALB/c mice were implanted subcutaneously in the right flank with 1 × 10⁶ A20 cells (0.1-mL cell suspension). Tumor growth was monitored until reaching an average volume of 100 mm³, at which time (day 0) the mice were randomized into the various treatment groups. Treatments were initiated on day 1. Tumor volume was calculated using caliper measurements according to the following formula: where w = width and l = length (in mm) of the tumor.

Gene expression data were analyzed using nSolver software provided by NanoString Technologies, Inc. Raw data were normalized to the geometric mean values of the internal synthetic positive controls and geometric means of the housekeeping genes, as recommended by the manufacturer. The NanoString Technologies’ nSolver Analysis Software 4.0 generated cell type scores, pathway scores, heatmaps, and individual gene normalized data from the nCounter PanCancer Immune Profiling Panel Plus dataset. The cell type score quantifies cell populations using marker genes for given cell types; by centering the mean at 0 for each cell type, immune cell type abundance can be compared on the same scale. The same method was used to generate immune-relevant pathway scores; summarizing the data from multiple genes in a pathway into a single score allowed for comparison between treatments for pathway analysis.

Normalized expression data for individual genes was exported from nSolver, annotated with percent of tumor growth inhibition (%TGI), and then imported into GraphPad Prism 7.04. The %TGI was used to group animals into 3 categories, as follows: nonresponders (TGI < 25%), partial responders (TGI 25–75%), and responders (TGI > 75%). Gene expression data for each mouse was color-coded (TGI < 25%, TGI 25–75%, TGI > 75%) to track gene expression with tumor response and used to determine if changes in gene expression associated with the tumor response.

Prism 7.04 (GraphPad, San Diego, CA) was employed for statistical and graphical analyses. Survival was analyzed by the Kaplan-Meier method. The logrank (Mantel-Cox) and Gehan-Breslow-Wilcoxon tests determined the significance of the difference between the overall survival experiences (survival curves) of two groups, based on time to endpoint values. Differences in tumor size among groups were assessed using 2-tailed t-test statistical analyses. The results are reported as nonsignificant (ns) at P > 0.05, significant (*) at 0.01 ≤ P < 0.05, very significant (**) at 0.001 ≤ P < 0.01, and extremely significant (***) at P < 0.001.

Studies were carried out in compliance with the ARRIVE guidelines.

To test whether HBI-8000 augments the antitumor effects of inhibiting the PD-(L)1 immune checkpoint axis, we treated mice bearing MC38 syngeneic tumors with HBI-8000, a mouse PD-1 Ab, or HBI-8000 plus PD-1 Ab (Fig. 1A, B). Therapeutic regimens and intervals are depicted in Fig. 1I. Figure 1B shows tumor growth in individual mice. In addition, we treated MC38 tumor-bearing mice with HBI-8000, a mouse PD-L1 Ab, or HBI-8000 plus PD-L1 Ab (Fig. 1C, D). Treatment with a single agent (HBI-8000, PD-1 Ab, or PD-L1 Ab) did not significantly affect tumor growth. Tumor regression (i.e., absence of detectable tumor) was not seen in any of the single agent cohorts, and all tumors continued to grow throughout the study. In contrast, combining either the PD-1 Ab or the PD-L1 Ab with HBI-8000 produced a statistically significant and reproducibly synergistic decrease or delay in tumor growth and progression (Fig. 1A, C). To corroborate these results, we extended our investigations to 2 other syngeneic tumor models. Mice bearing A20 tumors were treated with the same modalities and similar results were generated. Single-agent HBI-8000, PD-1 Ab, or PD-L1 Ab (A20, Fig. 1E, F) did not significantly affect tumor growth. As seen in the MC38 model, however, the combination of either a PD-1 Ab or PD-L1 Ab with HBI-8000 produced a significant and synergistic decrease or delay in tumor growth and progression, and importantly, an increase in the number of mice with tumor regression. Finally, we tested HBI-8000, a mouse CTLA-4 Ab, or HBI-8000 plus CTLA-4 Ab in the CT26 model (Fig. 1G, H). Similar to ICIs targeting PD-1 and PD-L1, the CTLA-4 Ab alone did not significantly affect tumor growth. Combining HBI-8000 with CTLA-4 Ab produced a highly significant delay in tumor progression, with 20% of tumor-bearing mice experiencing complete regression.

In summary, irrespective of the mouse tumor model or ICI Ab, single-agent therapy did not inhibit/regress tumor growth in any of the models tested. In all treatments combined with HBI-8000, we observed tumor regression after treatment, with subsets of tumors showing a significant delay in progression or outright regression. The data indicate that combining HDACi HBI-8000 with an ICI Ab was very efficacious in multiple animal models.

To investigate the mechanism of action of HBI-8000 in combination therapy with ICIs, larger groups of mice (n = 20/group) were implanted with MC38 tumors to sufficiently power the statistical analysis. A baseline no-treatment tumor-bearing group was terminated 1 day before initiating treatment. Mice in each treatment arm (n = 20) were terminated at days 7, 14, and 17 post treatment initiation. The NanoString nCounter PanCancer Immune Profiling Panel analysis allows for clustering of immune response-related genes into “gene sets” comprising a collection of genes selected as being representative of an element of the immune response (i.e., cell type, pathway), and provides a high-level view of the antitumor response, which is depicted as scatterplots in Fig. 2A. While all the scores were elevated in the PD-1 Ab plus HBI-8000 combination agent cohorts, it is noteworthy that a subset of cell type scores was augmented by either PD-1 Ab or HBI-8000 alone as early as day 7. Scores for exhausted CD8 T cells and neutrophils were predominantly augmented by the PD-1 Ab. In contrast, HBI-8000 augmented the scores corresponding to dendritic cells, macrophages, NK cells, cytotoxic cells, and CD45 cells, demonstrating that HBI-8000 alone had a profound conditioning or priming effect on immune-relevant gene expression within the TME, and suggesting that it reprograms the TME such that ICI therapy is more effective. Interestingly the same trends is observed in Day 14 and Day 17 analysis indicating consistency in HBI-8000 effects on TME as single agent or in combination with PD-1 antibody.

The de novo generation of new tumor-selective T cell clones might be a key factor in the response to the PD-1/PD-L1 checkpoint blockade [11, 36]. Because data from a preliminary study suggested that HBI-8000, alone or in combination with PD-1 Ab, has profound effects on the early or priming phase of the immune response, we investigated changes in the expression of genes associated with dendritic cell functions, antigen processing, and MHC class II antigen presentation. Consistently, gene expression analysis in single agent HBI-8000–treated tumors showed at least partial co-clustering with the response to the HBI-8000 plus ICI Ab combination therapy within these gene sets (Shown by yellow rectangle, Supplemental Figure 1A). HBI-8000 also co-clustered with combination therapy at the level of MHC class I antigen expression and presentation, which is important for effector T cell recognition and killing of tumor cells (Supplemental Figure 1B). Unsupervised hierarchical clustering of the indicated immune cell type scores vs. treatment and tumor response (Supplemental Figure 1A) showed that gene expression changes representative of these scores were most notable in the PD-1 Ab plus HBI-8000 combination cohort, and in responders vs. nonresponders (Supplemental Figure 1B). The analysis also demonstrated segregated clustering of the adaptive vs. innate response cells types. Not surprisingly, PD-1 Ab plus HBI-8000 combination therapy co-clustered with gene expression sets representing high response rates (TGI > 75%), which was observed for both adaptive and innate immune cell types. HBI-8000 also co-clustered with the HBI-8000/PD-1 Ab combination in modulating the expression of gene sets associated with cytokines, chemokines and their receptors, and with adaptive immunity-related genes. The data suggest that the class I/II selective HDACi can epigenetically modulate gene expression patterns within the TME, which contributes to multiple facets of the antitumor immune response, leading to the priming of effector T and B cells, the recognition of tumor cells by T cells with a consequent shift in the expression of relevant cytokines and corresponding receptors, and the augmentation of both innate and adaptive immune responses.

Induction of immune checkpoint receptors and/or ligands is thought to indicate a shift from a T cell-noninflamed (cold) TME to a T cell-inflamed (hot) TME [45]. The changes observed in immune checkpoints in the MC38 TME are shown in Fig. 2B. The data plots are color-coded to represent the tumor growth inhibition response seen for each individual animal, set arbitrarily for the purpose of illustration as tumor growth inhibition > 75, 25% through 75%, or less than 25% to represent responders, stable disease, and progressors, respectively. We observed increased expression of the immune checkpoints PD-1, PD-L1, CTLA-4, and CD86 (CD28L), the expression levels of which associated with antitumor efficacy and tumor regression (Fig. 2B). We also observed statistically significant changes in the expression of immune checkpoints CD276/B7-H3 and CD244 (Fig. 2B), as well as lymphocyte activation gene-3 (LAG-3), T cell immunoreceptor with Ig and ITIM domains (TIGIT), ecto-5′-nucleotidase (NT5E/CD73), signal regulatory protein α (SIRPα), nuclear factor of activated T cells 4 (NFATC4), and poliovirus receptor (CD155; Supplemental Figure 2). We believe modulation in the expression of the above genes indicates a shift from a noninflamed (cold) TME to an inflamed (hot) TME and interrelated with the antitumor response in MC38 tumor-bearing mice [42, 45‐47].

In addition to analyzing the effects of HBI-8000, PD-1 Ab, or their combination on various immune pathways, cell type functional scores, and immune checkpoint markers, we examined the effect of HBI-8000, PD-1 Ab, and their combination on a number of individual genes relevant to either innate or adaptive immunity (Figs. 3, 4 and 5, and Supplemental Figures 2 and 3). Genes modulated predominantly by the PD-1 Ab included CD8a (Fig. 4), inducible T cell costimulator (ICOS/CD278), and CD40 (Supplemental Figure 3). PD-1 Ab was also the driver for changes in the expression of genes involved in T cell recruitment, memory, and the CD8 T cell response, including CXCR6 (Fig. 5), ICOS and CD40, (Supplemental Figure 3). Our analysis of the nCounter data (Fig. 4) showed increased T-effector and interferon-γ gene scores, which, along with increases in granzyme B (GZMB) and perforin-1 (PRF1), are collectively consistent with an enhanced T-effector and interferon-γ gene, reflecting enhanced existing immune competency [41, 46‐48].

We observed the modulation of many genes affecting the TME inflammation in the tumors treated with the combination of HBI-8000 plus PD-1 Ab. Examples included the co-stimulator CD86 (Fig. 2B), chemoattractant receptors C–C chemokine receptor (CCR) 5 (Fig. 3), and CCR1 (Fig. 4), which are important for initial events in effector T-cell differentiation, markers of increased tumor reactive effector cells, e.g., ectonucleoside triphosphate diphosphohydrolase-1 (ENTPD1/CD39; Fig. 4), PRF1 (Fig. 4), and effector T cell memory precursors (interleukin 7 receptor [IL7R] and interferon regulatory factor 4 [IRF4], Figs. 3 and 5, respectively). Because HBI-8000 enhances both CD8 T cell and NK cell activity and functions (10,30), relevant genes modulated predominantly by HBI-8000 are of great interest. Examples of those genes include: 4-1BB/CD137 (Fig. 3), tumor necrosis factor α (TNFα; Fig. 3), interleukin 2 receptor alpha (IL2Rα)/CD25, GZMB (Fig. 4), IRF4 (Fig. 3), chemokine (C-X3-C motif) receptor 1 (CXC3R1), chemokine (CXC motif) receptor (CXCR)6, and CXCR3 (Fig. 5). These genes are relevant to an initial cytokine or CD8 effector response, tumor infiltrating lymphocyte (TIL) recruitment, effector cell differentiation, and effector memory (31–34).

Importantly, many genes were modulated by HBI-8000 alone relatively early (day 7) in the antitumor response (e.g., CD86, 4-1BB/CD137, TNFα, CCR5, chemokine (C–C motif) ligand 2 (CCL2), IL2Rα/CD25 (Figs. 2b, 3, and 4), and CCR1 and GZMB (Fig. 4). Consistent with reports of HBI-8000 having a positive effect on NK cell functions and innate immunity [26], we observed that HBI-8000 alone or combined with PD-1 Ab modulated the expression of GZMB (Fig. 4), killer cell lectin like receptor D1 (KLRD1/CD94; Fig. 3), and killer cell lectin like receptor C2 (NKG2c/KLRC2), natural killer cell granule protein 7 (NKG7), and killer cell lectin like receptor K1 (KLRK1; Supplemental Figure 3). Finally, and consistent with the upward shifts seen in all scores relevant for antigen presentation machinery and supportive of antigen presentation or tumor cell recognition, we observed increases in the expression of several MHC class I (H2-D1, H2-K1) and II genes (H2-Aa, H2-Eb1) (Fig. 5), in HBI-8000 alone (H2-D1, H2-K1) or the combination of HBI-8000 and PD-1 Ab (H2-Aa, H2-Eb1). This is an important observation and relevant to the reversal of known mechanisms of resistance to ICIs, namely the loss of MHC class I and class II molecules, which impede tumor cell recognition by effector CD8 T cells, as well the presentation of tumor antigens, including neoantigens, to naïve de novo antitumor immune cells [12, 13, 49].

Human cancer patients receiving ICI therapy often experience a transient response or stable disease, but eventually develop resistance and progress, a challenge to which major efforts are directed. Because gene expression data showed that HBI-8000 alone induced positive changes in a significant number of immune-related pathway scores and genes, we examined the ability of HBI-8000 to halt or even reverse progression in mice first treated with single-agent ICI therapy, alone or in combination with an ICI. To explore the effect of HBI-8000 plus ICI on acquired resistance, we developed a model based on the repeated observations that tumor-bearing mice treated initially (first-line) with single agent PD-1 Ab or PD-L1 Ab display 4 patterns of growth: i) approximately 20% experience rapid progression; ii) approximately 20% experience complete regression, and iii) & iv) approximately 60% experience stable tumor growth (defined as 3 consecutive tumor volume measurements with no significant change) or slow progression (relative to rapid growth and progression), which somewhat approximates the clinical situation. Using the above model, we treated a large cohort of tumor-bearing mice with PD-1 Ab alone. Once they reached the criteria for stable disease or slow progression, they were randomized into 6 treatment arms as indicated in Fig. 6. We compared the effect of halting treatment (Vehicle, Fig. 6A), continuing to treat with PD-1 Ab, or continuing PD-1 Ab in combination with HBI-8000. We also compared the effect of mAbs directed against the reciprocal target, PD-L1, treating mice with PD-L1 Ab, alone or in combination with HBI-8000. As shown in Fig. 6A In mice failing PD-1 Ab therapy, HBI-8000 was modestly efficacious in tumor growth inhibition, however, the second course of PD-1 Ab failed to significantly affect tumor growth. A second course of anti-PD-1 therapy combined with HBI-8000 produced no delay in tumor growth compared with anti-PD-1 alone and the modest delay seen in overall tumor growth provided by treatment with PD-L1 Ab alone was not significant. In contrast, combination therapy with HBI-8000 and anti-PD-L1 significantly (p < 0.05) inhibited tumor growth, suggesting that mice progressing on one ICI therapy could see benefit from an alternative ICI in combination with HBI-8000 (Fig. 6A). Analysis of survival, based on terminating mice whose tumors reach 1500 mm³, further validated tumor growth curve and suggested a significant delay in tumors regrowth in Group 6 (HBI-8000 plus anti-PD-L1).