Chidamide and Tyrosine Kinase Inhibitor Change the Tumor Immune Environment and Slow Tumor Growth with Immune Checkpoint Therapy in Mice with New or PD-1 Resistant Tumors

Herein, we initially studied the anticancer activity of small-molecule TKI monotherapy and evaluated the potential of TKIs plus chidamide combination therapy to enhance the objective response rate in CT26-bearing mice models. As shown inA–C, treatment with lenvatinib, cabozantinib, regorafenib, or axitinib alone, only partially reduced tumor growth. For the above 4 drugs, cabozantinib or regorafenib when combined with chidamide markedly inhibited tumor growth when compared with lenvatinib/axitinib combined with chidamide, as shown inC. Mice treated with the different regimens showed no significant loss in body weight, as shown inD. Individual tumor sizes (fold-change) and ORR, as shown inE, indicated that the monotherapy groups achieved 25–38% ORR; cabozantinib combined with chidamide therapy achieved 88% ORR, while regorafenib combined with chidamide also achieved 88% ORR. As shown inF, cabozantinib/regorafenib combined with chidamide exerted a good survival rate. Therefore, cabozantinib/regorafenib combined with chidamide treatment achieved great antitumor effect and may exhibit even greater antitumor activity in combination with anti-PD-1 antibody in immunotherapy application. Figure S1 Figure S1 Figure S1 Figure S1 Figure S1

Our previous study suggested that chidamide and celecoxib are potent TME modulators that improve antitumor activity when combined with anti-PD-1 antibody immunotherapy [29]. PGE2 was shown to activate several key immuno-suppressive cells present in the TME, such as Treg, M-MDSC, and TAM. Therefore, celecoxib, inhibiting PGE2 synthesis, may be used as a co-regulator of the tumor microenvironment. While evaluating the anticancer effect of cabozantinib plus chidamide in combination with anti-PD-1 antibody, we also evaluated whether anti-PD-1 antibody combined with cabozantinib plus celecoxib could achieve the same effect. As shown in Figure S2A–C, the anti-PD-1 antibody combined with cabozantinib plus celecoxib or chidamide-k30 regimen achieved excellent tumor growth inhibition when compared with the anti-PD-1 antibody treatment alone. Celecoxib is a selective cyclooxygenase (COX)-2 inhibitor and chidamide is a potent benzamide-based HDACi that selectively inhibits HDACs 1, 2, 3, and 10. We observed no significant loss in body weight in mice treated with the different regimens, as shown in Figure S2D. The individual tumor sizes (fold-change) and ORR, as shown in Figure S2E, indicated that treatment with anti-PD-1 antibody group achieved 25% ORR, anti-PD-1 antibody combined with cabozantinib plus celecoxib achieved 57% ORR, and anti-PD-1 antibody (2.5 mg/kg) combined with cabozantinib plus chidamide-k30 achieved 100% ORR. These findings demonstrated that the mice treated with anti-PD-1 antibody combined with cabozantinib plus chidamide-k30 achieved better ORR than those treated with anti-PD-1 antibody combined with cabozantinib plus celecoxib. This is the first study to report that TKI plus HDACi combined with ICI significantly enhanced the response rate. As shown in Figure S2F, anti-PD-1 antibody combined with cabozantinib plus chidamide-k30 markedly prolonged survival when compared with the other groups, achieving a 100% survival rate by day 60.

As shown in Figure S2, our results revealed that chidamide-k30 combined with anti-PD-1 antibody plus cabozantinib potently inhibited tumor growth. Next, we examined anti-PD-1 antibody therapy combined with regorafenib (like cabozantinib, a potent TKI and second-line treatment for advanced liver cancer approved by the FDA in 2017; third-line treatment for advanced colon cancer approved by the FDA in 2012) plus chidamide-k30 regimens in CT26 tumor-bearing mice model. As shown in Figure 1A–C, anti-PD-1 antibody combined with cabozantinib plus chidamide-k30 regimen more effectively inhibited tumor growth than anti-PD-1 antibody combined with cabozantinib regimens, as expected. This result suggested that, in combination with cabozantinib, chidamide may be potently effective, not inferior to anti-PD-1 antibody, as shown in Figure 1C. Similar results were also observed in the regorafenib treatment groups; anti-PD-1 antibody combined with regorafenib plus chidamide-k30 was more effective in inhibiting tumor growth than anti-PD-1 antibody combined with regorafenib. As shown in Figure 1D, treatment with regorafenib plus chidamide-k30 combined with anti-PD-1 antibody initially reduced body weight, which was eventually recovered after continuous treatment. Individual tumor sizes (fold-change) and ORR, as shown in Figure 1E, indicated that anti-PD-1 antibody combined with cabozantinib plus chidamide-k30 achieved 50% ORR. The group treated with anti-PD-1 antibody combined with regorafenib plus chidamide-k30 achieved 43% ORR. The survival rate was evaluated in the anti-PD-1 antibody combined with cabozantinib/regorafenib plus chidamide-k30 group. As shown in Figure 1F, anti-PD-1 antibody combined with regorafenib plus chidamide-k30 was more potent in prolonging survival than anti-PD-1 antibody combined with cabozantinib plus chidamide-k30. This result suggested that chidamide is a crucial component for improving the anti-PD-1 antibody combined with the cabozantinib/regorafenib regimen to significantly enhance the ORR and survival rate in CT26 tumor-bearing mice. Several clinical trials have examined anti-PD-1/anti-PD-L1 antibody therapy combined with the cabozantinib/regorafenib regimens to boost ORR and OS in advanced cancers, such as RCC, hepatocellular carcinoma (HCC), and metastatic colorectal cancer (mCC). Interestingly, the anti-PD-1 antibody combined with regorafenib plus chidamide-k30 regimen exhibited 43% ORR, but the OS rate approximated 86%. We also observed that, although treatment with this drug combination was discontinued, the tumor persistently shrunk (as shown in Table 1). It was hypothetically suggested that this regimen could robustly modulate tumor immunologic activity. This effect is the most desirable result of cancer immunotherapy, activating the immune system with drugs that produce long-lasting immune memories that continue attacking tumor cells. To test the proposed mechanism, the observation of recurrence and rechallenge studies were performed, as outlined in Figure 1A. Based on the preliminary results of the recurrence rate, as shown in Table 1, anti-PD-1 antibody combined with regorafenib plus chidamide-k30 regimen seemed to be more effective in preventing relapse than anti-PD-1 antibody combined with cabozantinib plus chidamide-k30 regimen. Furthermore, in the rechallenge study, we found that the mice with CR or PR response after treatment with anti-PD-1 antibody combined with cabozantinib or regorafenib plus chidamide-k30 showed resistance to CT26 rechallenge. Based on the results of these studies, it could be suggested that regimens comprising anti-PD-1 antibody combined with multi-kinase inhibitor, such as regorafenib or cabozantinib, plus chidamide-k30, activated a specific immune memory, thereby exerting strong anticancer activity. These findings indicated that modulation of the TME could improve the tumor response rate, avoid recurrence, and possibly induce immune memory.

We next investigated whether treatment with a double regimen of anti-PD-1 antibody combined with cabozantinib/regorafenib, or a triple regimen comprising anti-PD-1 antibody combined with cabozantinib/regorafenib plus chidamide-k30, affected immune cells, including myeloid-cell and T-cell populations, in tumors. Accordingly, we examined tumor samples isolated on day 20 after 9-day treatment and assessed immune cells using flow cytometry (FACS) in Figure S3. The anticancer activities for all treatment regimens were assessed on day 20 before tumors were excised, and are shown in Figure 2B. As shown in Figure 2C–F, populations of CD3+, CD4+ T-cells, and Tregs in tumors were significantly altered following treatment with anti-PD-1 antibody combined with cabozantinib or regorafenib, with or without chidamide-k30; however, anti-PD-1 antibody+ regorafenib+ chidamide-k30 treatment did not significantly decrease Tregs, while markedly increasing CD8+ T-cells. These results suggested that anti-PD-1 antibody combined with cabozantinib or regorafenib effectively reduced Tregs, which was favorable for activating tumor immunity in response to tumor inhibition. In addition, only the PD-1 antibody + regorafenib + chidamide-k30 regimen significantly increased CD8+ T-cell infiltration; this was deemed more conducive for inhibiting tumor growth. In the MDSC population, compared with other regimens (anti-IgG and anti-PD-1 groups), anti-PD-1 antibody combined with regorafenib plus chidamide-k30 treatment significantly decreased PMN-MDSCs and TAMs in tumors, as shown in Figure 2G,H,J. Accordingly, the results of the anti-PD-1 antibody + regorafenib + chidamide-k30 triple combination were likely attributed to the altering of tumor-infiltrating immune cells, increasing CD8+ T-cells and decreasing immune-suppressive PMN-MDSCs and TAMs in tumors. However, the level of tumor M-MDSCs exhibited no correlation with anticancer activity in double or triple combination regimens, when comparing the results of Figure 2B,I. Taken together, in terms of TME regulation, tumor growth could be inhibited as long as a relative regulatory advantage could be achieved, rather than an absolute advantage, achieved.

Gene expression analysis of CT26 tumors revealed that multiple immune-related pathways were induced following treatment with anti-PD-1 antibody combined with cabozantinib/regorafenib, with or without chidamide. The consecutive treatment schedule and tumor size of each treatment group are shown in Figure S4A,B. To clarify molecular mechanisms underlying the double and triple regimen combination in cancer immunotherapy, we performed RNA-sequencing (RNA-seq) on CT26 tumors measured and excised on day 20 after 9-day treatment (Figure S4A,B) to identify regulated genes. On analyzing our RNA-seq data, we identified mRNA transcript levels of genes upregulated and downregulated by the above treatment regimens (Figure 3A). The numbers of genes upregulated and downregulated after each regimen treatment are labeled on the top-right and top-left corners of Figure 3A, respectively. As shown in Figure 3B, GO analysis was performed on upregulated and downregulated genes, given that the minimal fold-change was 2 and p-adjusted was <0.05. Based on DEG analysis, we also performed GSEA enriched analysis, showing that the pathways possibly involved were those including T-cell-mediated cytotoxicity, angiogenesis, chemokine activity, cytokine activity, leukocyte migration, and cell chemotaxis. Leukocyte migration is a major target of triple combination treatment regimens anti-PD-1 antibody combined with cabozantinib/regorafenib plus chidamide-k30. Therefore, we focused on analyzing chemokine and cytokine activity/gene expression. GSEA of DEGs highlighted enrichment of pathways in anti-PD-1 antibody-treated tumors, with upregulated T-cell-mediated cytotoxicity response gene signature; however, with downregulated angiogenesis response gene signature (Figure 3C; p < 0.05). In PD-1 antibody + cabozantinib treated tumors, we noted upregulated enrichment of the response to T-cell-mediated cytotoxicity, chemokine activation, cytokine activity gene signature, with downregulated angiogenesis, leukocyte migration, and cell chemotaxis response gene signature (Figure 3D; p < 0.05). In PD-1 antibody + cabozantinib + chidamide-k30 treated tumors, we noted upregulated enrichment of the response to chemokine activation, cytokine activity, leukocyte migration, cell chemotaxis, and T-cell-mediated cytotoxicity response gene signature, with downregulated angiogenesis response gene signature (Figure 3E; p < 0.05). In PD-1 antibody + regorafenib treated tumors, upregulated enrichment of the response to chemokine activation and cytokine activity was observed, with downregulated leukocyte migration cell chemotaxis and angiogenesis response gene signature (Figure 3F; p < 0.05). In PD-1 antibody + regorafenib + chidamide-k30 treated tumors, we observed the upregulated enrichment of the response to chemokine activation, cytokine activity, and T-cell-mediated cytotoxicity, along with downregulated leukocyte migration, cell chemotaxis and angiogenesis response gene signature (Figure 3G; p < 0.05). Compared with anti-IgG antibody therapy, all treatments significantly downregulated pathways involved in angiogenesis (Figure 3C–G).

Compared with the anti-PD-1 antibody alone, the triple regimen comprising anti-PD-1 antibody + cabozantinib/regorafenib + chidamide-k30 more significantly showed their effect on the upregulation of chemokine, cytokine activity-related and T-cell-mediated cytotoxicity gene expression. Reportedly, the IFN-γ response is critical for activating TILs [30]. Therefore, GSEA was performed using published gene sets for the immune cell signature. The IFN pathway and T-cell gene signature was upregulated in anti-PD-1 antibody-treated tumors (Figure S4C; p < 0.001). In anti-PD-1 antibody + cabozantinib treated tumors, we observed a trend of upregulated IFN pathway gene signature (Figure S4D; p < 0.001). In addition, PD-1 + cabozantinib + chidamide-k30 treated tumors showed significantly upregulated IFN pathway, macrophage gene, neutrophil gene, and T-cell gene signatures but downregulated enrichment of the monocyte gene signature (Figure 4E; p < 0.05). In PD-1 + regorafenib treated tumors, enrichment of the macrophage, monocyte, and T-cell gene signature was significantly downregulated (Figure S4F; p < 0.05). In PD-1 + regorafenib + chidamide-k30 treated tumors, we detected upregulated enrichment of IFN pathway gene signature and downregulated enrichment of the macrophage and T-cell gene signatures (Figure S4G; p < 0.001). These results demonstrated that intratumor heterogeneity of gene expression elicited responses to anti-PD-1 antibody + regorafenib/cabozantinib + chidamide-k30 therapy in CT26-bearing mice models. Anti-PD-1 antibody + cabozantinib did not significantly downregulate or upregulate macrophage gene and monocyte gene signatures; however, the downregulation of monocyte gene signature and upregulation of macrophage gene, neutrophil gene, and T-cell gene signatures were enhanced by further combination with chidamide. Anti-PD-1 antibody + regorafenib significantly downregulated macrophage, monocyte, and T-cell gene sets, but did not induce IFN response gene set. However further addition of chidamide to the anti-PD-1 antibody + regorafenib combination attenuated this downregulation effect on monocytes and induced IFN-γ response gene set. Collectively, our data suggested that anti-PD-1 antibody combined with regorafenib plus chidamide modulated immune cell migration, angiogenesis, and cytokine, or chemokine, gene expression activity in the TME to enhance the tumor response rate in CT26 tumor-bearing mice.

Next, we evaluated the activities of tumor growth inhibition following treatment with different ICIs, such as anti-PD-1/anti-PD-L1/anti-CTLA-4 antibodies, combined with regorafenib/cabozantinib plus chidamide-k30 regimens in CT26 tumor-bearing mice. As shown in Figure 4A–C, the anti-PD-L1 antibody combined with regorafenib plus chidamide-k30 regimen more effectively inhibited tumor growth than the anti-CTLA-4 or anti-PD-1 antibody combination regimens. However, the anti-CTLA4 antibody combined with the cabozantinib plus chidamide-k30 regimen was more effective for inhibiting tumor growth than the anti-PD-L1 or anti-PD-1 antibody combination regimens. As shown in Figure 4D, most triple combination regimens partially reduced body weight on day 20, except for anti-PD-1+ cabozantinib plus chidamide-k30; however, body weight gradually recovered after the last treatment administration on day 26. Individual tumor sizes (fold-change) and ORR, as shown in Figure 4E, indicated that anti-PD-L1 antibody combined with regorafenib plus chidamide-k30 achieved 89% ORR, whereas treatment with anti-CTLA-4 antibody combined with cabozantinib plus chidamide-k30 achieved 90% ORR; these two regimens were superior to the other regimens. The anti-PD-1 antibody combined with regorafenib/cabozantinib plus chidamide-k30 regimen showed relatively poorer ORR than the other regimens. As shown in Figure 4F, different ICI combinations with cabozantinib/regorafenib plus chidamide-k30 exhibited distinct OS rates. The medians of all triple combination regimens exceeded 61 days. Interestingly, two combinations (anti-PD-1 antibody combined with regorafenib plus chidamide-k30 and anti-PD-1 antibody combined with cabozantinib plus chidamide-k30) showed good OS, even with a moderate ORR (~30 to 40%); this finding indicated that these two triple regimens might potently activate the immune system to inhibit cancer cell growth and avoid relapse, as shown in Table 2. Surprisingly, after discontinuing drug administration, the tumor in CT26 tumor-bearing mice continued shrinking; therefore, a second ORR assessment was performed 10 days after the last drug administration. Treatment with anti-PD-1 antibody combined with regorafenib plus chidamide-k30 significantly enhanced ORR from 30% to 60% in the second ORR assessment, indicating that the immune system was potentially induced for the direct or indirect activation of CD8+ T-cells to attack tumor cells. A similar phenomenon was also observed following treatment with anti-CTLA-4 antibody combined with regorafenib plus chidamide-k30, presenting an enhanced ORR (increased from 60% to 80%) and a greater number of mice achieving CR. These data suggested that anti-PD-1/anti-PD-L1/anti-CTLA-4 antibody combined with regorafenib/cabozantinib plus chidamide-k30 possessed markedly potent activity in terms of activating the immune system. Hence, it should be noted that ICI combined with VEGFR-TKIs plus chidamide could continue alleviating tumor growth in mice who initially achieved SD or PR (first assessment). That is, despite drug discontinuation, these triple combination regimens could improve SD to PR or PR to CR in CT26-bearing mice. Finally, immunity was further evaluated using a rechallenge experiment (as shown in the far-right column of Table 2). Overall, the rechallenge experiment revealed that almost all triple combination regimens could significantly activate immunity, thereby effectively suppressing the rechallenge-induced proliferation of CT26 cancer cells.

In the present study, treatment with anti-CTLA-4 antibody combined with regorafenib or cabozantinib plus chidamide-k30 exhibited potent antitumor activity. We next examined whether this therapeutic benefit could also be observed in anti-PD-1 therapy resistant mice. Mice were treated with second-line therapy to mimic the treatment strategy employed for first-line drug-induced resistance known to occur in humans. It is well-established that numerous patients with cancer receiving first-line anti-PD-1 antibody therapy develop resistance, including primary resistance, acquired resistance, or HPD. To evaluate the effectiveness of different second-line treatments after development of drug resistance to first-line anti-PD-1 antibody, a platform with a suitable treatment schedule was designed, as outlined in Figure 5A. As shown in Figure 5A,B, first, 119 mice were treated with anti-PD-1 antibody as the first-line treatment, and 10 mice were treated with anti-IgG antibody as a negative control. Overall, 17 of the 119 mice achieved response to first-line anti-PD-1 antibody treatment, achieving a 14.3% ORR (17/119). Among the 119 mice, 102 showed primary resistance (non-response) to first-line anti-PD-1 antibody treatment, with an occurrence rate of 76.5% (91/119), as shown in Table 3. Regarding the HPD mice (with an incidence of approximately 9.2% [11/119]), defined accordingly, based on tumor volumes >600 mm³, the average tumor volume was 669 mm³, as shown in Figure 5A. These results suggested that primary resistance and HPD to treatment with the first-line anti-PD-1 antibody remains a challenging issue in cancer immunotherapy. Mice with primary resistance and HPD to anti-PD-1 antibody therapy were randomized to nine different treatment groups (n = 5–11 mice/group), as indicated. As our unpublished data demonstrated, CC-02 (chidamide-HCL + celecoxib) combined with an anti-CTLA4 antibody was shown to elicit a significant anticancer activity for second-line treatment; and the chloride salt formulation of chidamide significantly improved aqueous solubility compared with the k30 coated formulation. As shown in Figure 5C–E, both regimens, i.e., anti-CTLA-4 antibody combined with regorafenib plus chidamide-k30 and anti-CTLA-4 antibody combined with cabozantinib plus chidamide-k30, were more potent in inhibiting tumor growth than the positive controls anti-CTLA-4 antibody combined with CC-02 and anti-CTLA-4 antibody alone. These results suggested that anti-CTLA-4 antibody combined with regorafenib/cabozantinib plus chidamide-k30 potently regulated TME activity to overcome primary resistance to first-line anti-PD-1 antibody therapy. As shown in Figure 5F, no significant body weight loss was detected in all treatment groups, except the anti-CTLA-4 antibody + cabozantinib plus chidamide-k30 group, in which body weight was initially partially reduced and then gradually recovered. The HPD mice were treated with anti-CTLA-4 antibody combined with cabozantinib plus chidamide-HCl salt for 16 days, and large tumors surprisingly showed growth inhibition, while some tumors were controlled by persistent shrinkage under this treatment regimen, as shown in Figure 5D. These results suggested that the anti-CTLA-4 antibody combined with chidamide-HCl salt plus cabozantinib regimen could potently rescue HPD (none of the mice achieved PD). Individual results for each drug-resistant mouse are shown in Figure 5G. Treatment with CC-02 achieved 37.5% ORR. Treatment with anti-CTLA-4 antibody combined with CC-02 as positive control showed 37.5% ORR. However, treatment with anti-CTLA-4 antibody combined with regorafenib plus chidamide-k30 achieved 62.5% ORR. Treatment with anti-CTLA-4 antibody combined with cabozantinib plus chidamide-k30 achieved 57.1% ORR. In addition, treatment of HPD mice with anti-CTLA-4 antibody combined with cabozantinib plus chidamide-HCl salt showed 18.1% ORR. Although the tumor response rates of HPD mice were lower than those of other combination groups, most tumor-bearing mice showed inhibited tumor growth after treatment with this second-line regimen. This result was intriguing because HPD tumors grow more rapidly and are difficult to effectively reduce and inhibit. Furthermore, we evaluated the survival rate of mice with primary resistance or HPD to anti-PD-1 antibody treatment. As shown in Figure 5H,I, after discontinuing second-line therapy, treatment with anti-CTLA-4 antibody combined with regorafenib plus chidamide-k30 achieved an OS of 87.5%, while anti-CTLA-4 antibody combined with cabozantinib plus chidamide-k30 group achieved an OS of 71.4%, thus indicating possible immune system activation to inhibit tumor cell growth and avoid relapse, as shown in Table 4. Regarding HPD mice, treatment with anti-CTLA-4 antibody combined with cabozantinib plus chidamide-HCl salt achieved a good OS rate (45.4%). This result was noteworthy because HPD mice only achieved 18.1% ORR but achieved an OS rate of 45.4% due to significant suppression of tumor growth (Table 4). Finally, immunity was further examined using a rechallenge experiment (as shown in the far-right column of Table 4) and revealed that almost all regimens were significantly active in stimulating immunity, thereby effectively inhibiting the rechallenge-induced proliferation of CT26 cancer cells.

Figure S5A presents the consecutive treatment schedule for the second-line treatment group. To clarify TME remodeling induced by double and triple combination regimens in cancer immunotherapy, we performed RNA-seq analysis on CT26 tumors excised on day 28 after 12-day treatment. Using RNA-seq data, we identified upregulated and downregulated mRNA transcript levels in each treatment group compared with IgG (Figure 6A). GO analysis was performed on upregulated and downregulated genes, given the minimal fold-change was 2 and p-adjusted was <0.05, which were associated with the responses to IFN-γ, cellular response to IFN-γ, leukocyte migration, cell chemotaxis, and adaptive immune response, etc. (Figure 6B). GSEA of DEGs highlighted enrichment for pathways in second-line anti-CTLA-4 antibody-treated tumors, presenting upregulation of cell chemotaxis and leukocyte migration gene signatures (Figure 6C; p < 0.05). In second-line CC-02 treated tumors, the responses to IFN-γ, cellular response to IFN-γ, adaptive immune response, leukocyte migration and cell chemotaxis were upregulated, whereas angiogenesis gene signature was downregulated (Figure 6D; p < 0.05). The second-line anti-CTLA-4 + CC-02 treated tumors showed upregulation of the response to IFN-γ, cellular response to IFN-γ, adaptive immune response gene signature, leukocyte migration and cell chemotaxis (Figure 6E; p < 0.05). The second-line triple combinations anti-CTLA-4 + cabozantinib + chidamide-k30 tumor treatment upregulated the response to IFN-γ, cellular response to IFN-γ, adaptive immune response gene signatures, cell chemotaxis and leukocyte migration gene signatures; however, it downregulated angiogenesis gene signature (Figure 6F; p < 0.05). In second-line anti-CTLA-4 + regorafenib + CD-k30 treated tumors, we noted upregulated response to IFN-γ, cellular response to IFN-γ, leukocyte migration and cell chemotaxis, and downregulated adaptive immune response and angiogenesis gene signature (Figure 6G; p < 0.05). Compared with anti-IgG antibody, except for anti-CTLA4 antibody and anti-CTLA-4 + CC-02 triple combination, CC-02 and triple combinations anti-CTLA-4 + regorafenib/ cabozantinib + CD-k30 significantly downregulated pathways involved in angiogenesis (Figure 6C–G). In addition, all treatments significantly upregulated pathways involved in leukocyte migration and cell chemotaxis. These results, when compared to those presented by naïve treatment, suggested that TME remodeling was highly related to downregulation of angiogenesis and differential regulation of leukocyte migration in this animal model. Furthermore, except for the anti-CTLA-4 treatment, all second-line treatment regimens effectively increased levels of the IFN pathway signature (Figure S5B–F). Therefore, GSEA was performed using the published gene set for the TIL signature. In anti-CTLA-4 antibody-treated tumors, the neutrophil gene signature was upregulated, whereas the IFN pathway and NK cell gene signatures were downregulated (Figure S5B; p < 0.05). In CC-02 treated tumors, we observed upregulated IFN pathway, macrophage, NK, and T-cell gene signatures (Figure S5C; p < 0.05). In anti-CTLA-4 + CC-02 treated tumors IFN pathway, macrophage, NK, neutrophil gene and T-cell gene signatures were upregulated (Figure S5D; p < 0.05). In anti-CTLA-4 + cabozantinib + CD-k30 (chidamide-k30) treated tumors the IFN pathway, macrophage, neutrophil, NK, monocyte, and T-cell gene signatures were upregulated (Figure S5E; p < 0.05). Anti-CTLA-4 + regorafenib + CD-k30 treated tumors showed downregulation of macrophage, monocyte, and T-cell gene signatures and upregulation of IFN pathway and neutrophils (Figure S5F; p < 0.05). These findings demonstrated that the intratumor heterogeneity of tumor responses to double and triple combinations could limit the predictive outcomes of these signatures. Although anti-CTLA-4 antibody can activate neutrophil gene expression, it significantly suppresses the IFN pathway and NK cell gene expression, which might exert synergistic antitumor gene expression when in combination with other drug(s). CC-02 significantly activated the IFN pathway, and macrophage, NK and T-cell gene expressions, and had no effect on neutrophil gene expression. In addition, both CC-02 and anti-CTLA-4 antibody + CC-02 regimens significantly activated the IFN pathway, and macrophage, NK and T-cell gene expressions, while also reflecting a similar antitumor activity as shown in Figure 5G. In addition, treatment with anti-CTLA-4 antibody + cabozantinib + chidamide-k30 significantly activated NK and T-cells, and monocytes, as well as macrophage and neutrophil gene expressions, while treatment with anti-CTLA4 antibody + regorafenib + chidamide-k30 activated neutrophil gene expression and significantly suppressed macrophage, monocyte, and T-cell gene expressions, while both regimens reflected a similar antitumor activity, as shown in Figure 5G. In the TME, the antitumor mechanisms underlying these two triple combination regimens may be different and further elucidation in future study is necessary.

Previously [28], we proposed that anti-PD-1 resistance (non-responsive to first-line anti-PD-1 Ab treatment as shown in Figure 5B) is due to loss of antigen presentation, as well as downregulation of cell chemotaxis, the IFN pathway signature, and T-cell-mediated cytotoxicity (data also shown in Figure S6A; p < 0.05). As shown in Figure S6B, our data revealed that CC-02, anti-CTLA-4 antibody + CC-02, and anti-CTLA-4 antibody + cabozantinib + chidamide-k30 treatment could upregulate transcripts associated with antigen processing and presentation of peptide antigens, and activate T-cell-mediated cytotoxicity in PD-1-resistant tumors. These results suggested that overcoming anti-PD-1 resistance may be achieved via enhanced T-cell activation through the recovery of the antigen presentation pathway and TME modulation. Collectively, the regulation of gene expression described above implied that for effectively overcoming first-line anti-PD-1 antibody-induced resistance, TME regulation should involve gene expression in immune cells, in CT26 tumors. Overcoming resistance to anti-PD-1 antibody therapy as first-line therapy is an urgent clinical issue. Our study suggested that chidamide combined with regorafenib/cabozantinib plus anti-CTLA-4 antibody is an important combination therapy to address resistance induced by first-line anti-PD-1 antibody therapy.

The animal study was approved and performed under the guidance of the Taipei Medical University Institutional Animal Care and Use Committee (TMU IACUC, NO: LAC-2020-0103, LAC-2019-0644). Six- to eight-week-old male BALB/c mice (National Laboratory Animal Center, Taiwan) were used to perform experiments. The CT26 cell line was purchased from American Type Culture Collection (ATCC, Manassas, VA, USA). CT26 tumor cell lines were grown in RPMI-1640 supplemented with 10% (v/v) fetal bovine serum at 37 °C and 5% CO₂. For establishing tumors, a 1 × 10⁶ CT26 cell suspension in 100 µL phosphate-buffered saline (PBS) with 50 µL Matrigel (354248, Corning^® BD Biosciences, Corning, NY, USA) was subcutaneously injected into the left flank of the mice, and tumor growth was determined by measuring two perpendicular diameters. Prior to randomization and treatment, tumors were allowed to grow for 8–12 days (tumor size, approximately 110–250 mm³). An anti-IgG antibody (Catalog #BE0089, BioXcell, West Lebanon, NH, USA), anti-PD-1 antibody (BE0146, BioXcell, West Lebanon, NH, USA), anti-PD-L1 antibody (BE0101, BioXcell, West Lebanon, NH, USA), and anti-CTLA-4 antibody (BE0164, BioXcell, West Lebanon, NH, USA) were administered intraperitoneally (i.p.) at 2.5 mg/kg, every 3 days for a total of 6 injections in 16 days. All antibodies were diluted to appropriate concentrations in 100 μL sterile PBS (pH 7.4; Invitrogen Life Technologies, Carlsbad, CA, USA). Axitinib (HY-10065, 30 mg/kg, per oral [p.o.] daily; MedChemExpress, Monmouth Junction, NJ, USA), lenvatinib (HY-10981, 10 mg/kg, p.o. daily; MedChemExpress)cabozantinib (HY-13016, 30 mg/kg, p.o. daily; MedChemExpress), regorafenib (HY-1031, 30 mg/kg, p.o. daily; MedChemExpress), chidamide-k30 or chidamide-HCl salt (50 mg/kg, p.o. daily; produced by GNTbm, Taipei, Taiwan), celecoxib (50 mg/kg, p.o. daily; capsule/Celebrex^®; Pfizer Inc., New York, NY, USA) were administered for 16 days. Axitinib, lenvatinib, cabozantinib, regorafenib, and celecoxib were dissolved in dimethyl sulfoxide and diluted in water before administration. Chidamide-k30 and chidamide-HCl salts were dissolved in water. Animals were euthanized when tumors exceeded 3000 mm³ in volume. The anticancer activity was measured from treatment initiation until the tumor volume reached 3000 mm³. The tumor volume was calculated as length × width² × 0.5. For assessing treatment efficacy, we defined complete response (CR; <0.5-fold tumor growth in tumor-bearing mice at three days after the end of treatment), partial response (PR; tumor size ≥ 0.5-fold time tumor growth, but <1-fold tumor growth in the tumor-bearing mice 3 days after the end of treatment), stable disease (SD; tumor size ≥ 1-fold tumor growth, but <5-fold tumor growth in the tumor-bearing mice 3 days after the end of treatment), and progressive disease (PD; tumor size ≥ 5-fold tumor growth in tumor-bearing mice 3 days after the end of treatment). ORR was defined as percentage of mice with CR or PR assessed after treatment in a treatment group. Recurrence was defined as tumor growth of at least 5-fold in mice with a CR or PR response after the first tumor assessment.

All mice with a PR/CR response after treatment were rechallenged with CT26 cells on the contralateral side. The CT26 rechallenge was performed on day 34 ± 2, i.e., 7 days after the first tumor assessment (day 27 ± 2), by administering an injection of 5 × 10⁶ CT26 cells per mouse. After rechallenge with CT26 cells, tumors were allowed to grow for another 7 days (day 41 ± 2) to determine the tumor baseline as 1-fold. After an additional 10 days (day 51 ± 2), tumor growth was evaluated for rechallenge. If two of the following criteria were met, the response was considered rechallenge-induced recurrence/relapse: (1) the tumor size was >2-fold when compared to that at baseline on day 41 ± 2; (2) the tumor volume on day 51 ± 2 was >300 mm³. A relapse occurs when immunity is insufficiently activated. If the growth of tumor from the second time cancer cell inoculation (the rechallenge) was inhibited, the immune system was activated by the treatment.

After tumor assessment, the tumor volume of mice was measured once every three or four days (twice weekly). Tumor-bearing mice were considered dead when the tumor volume reached 3000 mm³. All treatment groups were recorded and analyzed.

The following antibodies and reagents were used for flow cytometry: CD8a PerCP-Cy5.5 (53–6.7; BioLegend, San Diego, CA, USA), CD4 PE (GK 1.5; BioLegend, San Diego, CA, USA), CD25 PerCP-Cy5.5 (PC61; BioLegend, San Diego, CA, USA), Foxp3 PE (MF14; BioLegend, San Diego, CA, USA), CD3 APC (17A2; BioLegend, San Diego, CA, USA), CD11b APC (M1/70; BioLegend, San Diego, CA, USA), Ly-6C PerCP-Cy5.5 (HK 1.4; BioLegend, San Diego, CA, USA), Ly-6G PE (1A8; BioLegend, San Diego, CA, USA), MHC-II-PE (BM8; BioLegend, San Diego, CA, USA), and CD45 FITC (30-F11; BioLegend, San Diego, CA, USA). Flow cytometry was performed using FACSCaliber (BD Biosciences, San Jose, CA, USA), and data were analyzed using the FACS Diva software (BD Biosciences, San Jose, CA, USA). Subsequently, cells were fixed, permeabilized with BD Cytofix/Cytoperm (BD Biosciences, San Jose, CA, USA), and stained with an antibody against FOXP antibody.

To assess various tumor-infiltrating immune cells in tumors, further assays were performed to analyze intratumoral CD3+, CD8+, CD4+ T-cells, Tregs, polymorphonuclear (PMN)-MDSCs, monocytic-MDSCs (M-MDSCs), and tumor-associated macrophages (TAMs). Cells were first purified from tumor samples excised from mice on day 20 after 9-day treatment of cabozantinib or regorafenib, with or without chidamide-k30 plus anti-PD-1 antibody. Briefly, primary tumor tissues were harvested, weighed, and minced into fine fragments. Collagenase IV (Sigma-Aldrich, St Louis, MO, USA)) at 1 mg/mL in Hanks’ Balanced Salt Solution (HBSS; Invitrogen Life Technologies, Grand Island, NY, USA) was added to each sample at a ratio of 1 mL per 200 mg of tumor tissue. Samples were incubated on an end-over-end shaker for 150 min at 37 °C. The resulting tissue homogenates were filtered through a 0.4-μm filter and washed three times in PBS (BD Biosciences, San Jose, CA, USA), followed by separation using a Percoll gradient to isolate mononuclear cells, and 1 × 10⁶ cells per sample were used for antibody labeling. CD8+ T-cell levels were assessed using previously established phenotypic criteria for CD45+ CD3+ CD8+. Treg levels were assessed using previously established phenotypic criteria for CD45+ CD4+ CD25+ FoxP3+; PMN-MDSC and M-MDSC levels were assessed using previously established phenotypic criteria for CD45+/CD11b+/Ly6G+/Ly6C− and CD45+/CD11b+/Ly6G−/Ly6C+, respectively. Finally, TAM cell levels were assessed using previously established phenotypic criteria for CD45+ CD11b+ MHC-II+ Ly6C+, and total mononuclear cells were used as a common denominator.

Male BALB/c mice bearing subcutaneous CT26 tumors (1 × 10⁶ cells/mouse) were treated with anti-PD-1 antibody as first-line therapy (mean tumor volume: 113 mm³ at treatment initiation), administered i.p. at 2.5 mg/kg, once every 3 days for 3 doses. Treatment was continued for an additional 3 doses (i.e., a total of 6 doses), if tumors responded to anti-PD-1 antibody treatment (tumors shrunk or with a growth of less than 2.5-fold; please also refer to the definition described below). Acquired resistance was defined as tumors that shrunk after the first 3 doses of anti-PD-1 antibody therapy, then grew gradually following continuous anti-PD-1 antibody treatment (i.e., a total of six doses), presenting only partially inhibited tumor growth and subsequently exhibiting further growth. Primary resistance was defined as tumor failure to respond to anti-PD-1 antibody therapy (after three doses), with a 2.5-fold increase in tumor volume but <600 mm³. HPD was defined as tumors that grew >600 mm³ after three doses of first-line anti-PD-1 antibody treatment. For the efficacy study, mice with primary resistance, acquired resistance, and HPD were subsequently re-enrolled for second-line therapy, as shown in Figure 5A. Second-line therapies used were as follows: anti-IgG antibody as a control; anti-PD-1 antibody, and anti-CTLA-4 antibody administered i.p. at 2.5 mg/kg, once every 3 days for 6 doses; chidamide-HCl (50 mg/kg) + celecoxib (50 mg/kg) (CC-02), regorafenib (30 mg/kg) + chidamide-k30 (50 mg/kg), and cabozantinib (30 mg/kg) + chidamide-k30 (50 mg/kg), orally administered once daily for 16 days. Tumor length and width was measured once every three or four days (twice weekly) by using a caliper. The anticancer activity was measured from treatment initiation until the tumor volume reached 3000 mm³. Tumor volume was measured as described above.

Naïve CT26 tumor-bearing mice were randomized and treated with different regimens, and tumors were excised and collected on day 20 after 9-day therapy. In contrast, following first-line anti-PD-1 antibody therapy, drug-resistant mice were randomized and treated with different regimens, and the tumors were excised and harvested on day 28 after initiating second-line therapy. All tumor samples were snap-frozen in liquid nitrogen and homogenized in Trizol (Life Technologies, Carlsbad, CA, USA). RNA purity and quantification were assessed using the SimpliNano™-Biochrom spectrophotometer (Biochrom, Holliston, MA, USA). RNA degradation and integrity were examined using a Qsep 100 DNA/RNA analyzer (BiOptic Inc., new Taipei City, Taiwan).

Total RNA (1 μg) per sample was used as input material for RNA sample preparation. Sequencing libraries were generated using the KAPA mRNA HyperPrep Kit (KAPA Biosystems, Roche, Basel, Switzerland), following the manufacturer’s recommendations, and index codes were added to attribute sequences to each sample. PCR products were purified using the KAPA Pure Beads system, and library quality was assessed using a Qsep 100 DNA/RNA Analyzer (BiOptic Inc., new Taipei City, Taiwan).

The original data obtained by high-throughput sequencing (Illumina NovaSeq 6000 platform) were transformed into raw sequenced reads by CASAVA base calling and stored in the FASTQ format. FastQC and MultiQC were used to examine the quality of the FASTQ files. The obtained raw paired-end reads were filtered using Trimmomatic (v0.38) to discard low-quality reads, trim adaptor sequences, and eliminate poor-quality bases using the following parameters: LEADING:3 TRAILING:3 SLIDINGWINDOW:4:15 MINLEN:30. The obtained high-quality data (clean reads) were used for subsequent analysis. Read pairs from each sample were aligned to the reference genome using HISAT2 software (v2.1.0). FeatureCounts (v1.6.0) were used to count read numbers mapped to individual genes. For gene expression, normalization of “Trimmed Mean of M-values” (TMM) was performed using DEGseq (v1.36.1) without biological duplicates, while normalization of “Relative Log Expression” (RLE) was performed using DESeq2 (v1.22.1) with biological duplicates. Differentially expressed gene (DEG) analysis for the two conditions was performed in R using DEGseq (without biological replicates) and DESeq2 (with biological replicates), which were based on negative binomial distribution and Poisson distribution models, respectively. The resulting p-values were adjusted using Benjamini and Hochberg’s approach for controlling the false discovery rate. Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis of DEGs were conducted using clusterProfiler (v3.10.1). Gene set enrichment analysis (GSEA) was performed with 1000 permutations to identify enriched biological functions and activated pathways from the molecular signature database (MSigDB). MSigDB is a collection of annotated gene sets for use with GSEA software, including hallmark, positional, curated, motif, computational, GO, oncogenic, and immunologic gene sets. In addition, weighted gene co-expression network analysis (WGCNA) was performed using the co-expression network based on the correlation coefficient of expression patterns using the WGCNA (v1.64) package in R.

All statistical analyses were performed using GraphPad Prism 5.0 (GraphPad Software, Inc., La Jolla, CA, USA). Primary tumor growth curves, flow cytometric analyses, and anti-PD-1-treated resistance tumor growth curves were first analyzed with one-way ANOVA, and individual groups were compared using Tukey’s Multiple Comparison Test. Kaplan–Meier survival curves were analyzed with a log-rank test.