Epigenetic therapy sensitizes anti-PD-1 refractory head and neck cancers to immunotherapy rechallenge

Abstract

BACKGROUNDImmune checkpoint blockade (ICB) is an effective treatment in a subset of patients diagnosed with head and neck squamous cell carcinoma (HNSCC); however, the majority of patients are refractory.METHODSIn a nonrandomized, open-label Phase 1b clinical trial, participants with recurrent and/or metastatic (R/M) HNSCC were treated with low-dose 5-azacytidine (5-aza) daily for either 5 or 10 days in combination with durvalumab and tremelimumab after progression on ICB. The primary objective was to assess the biologically effective dose of 5-aza as determined by molecular changes in paired baseline and on-treatment tumor biopsies; the secondary objective was safety.RESULTSThirty-eight percent (3 of 8) of participants with evaluable paired tissue samples had a greater-than 2-fold increase from baseline in IFN-γ signature and CD274 (programmed cell death protein 1 ligand, PD-L1) expression within the tumor microenvironment (TME), which was associated with increased CD8+ T cell infiltration and decreased infiltration of CD4+ T regulatory cells. The mean neutrophil-to-lymphocyte ratio (NLR) decreased by greater than 50%, from 14.2 (SD 22.6) to 6.9 (SD 5.2). Median overall survival (OS) was 16.3 months (95% CI 1.9, NA), 2-year OS rate was 24.7% (95% CI: 4.5%, 53.2%), and 58% (7 of 12) of treated participants demonstrated prolonged OS of greater than 12 months.CONCLUSIONOur findings suggest that low-dose 5-aza can reprogram systemic host immune responses and the local TME to increase IFN-γ and PD-L1 expression. The increased expression of these established biomarkers correlated with prolonged OS upon ICB rechallenge.TRIAL REGISTRATIONClinicalTrials.gov NCT03019003.FUNDINGNIH/NCI P01 CA240239.

Keywords: Cancer immunotherapy; Epigenetics; Head and neck cancer; Immunology; Oncology.

Figure 1. Clinical trial schema and multi-omics datasets included in the study.

(A) Participants with locoregional or metastatic head and neck cancer that progressed on prior ICB were enrolled into the clinical trial (www.clinicaltrials.gov; NCT3019003). Participants were treated with escalating doses of a DNA methyltransferase inhibitor, 5-azacytidine (5-aza), and fixed doses of durvalumab (α-PD-L1) and tremelimumab (α-CTLA-4). The primary objective was determining the biologically effective dose (BED) of 5-aza. The secondary outcome was assessing the safety of the combination therapy. (B) Tissue specimens were collected (black arrows) prior to 5-aza treatment (green arrow) and after combination therapy with durvalumab and tremelimumab (purple arrows), which were given at the same time as the second dose of 5-aza (on-treatment).

Figure 2. Swimmer plot of overall survival.

(A) Swimmer plot demonstrates each participant’s (deidentified IDs in the y-axis) response to treatment in months (x-axis). Blue and orange bars denote Dose level 1 (5-days of 5-aza treatment) and Dose level 2 (10-days of 5-aza treatment), respectively. Participants with HPV-mediated disease are labeled with an asterisk (*). (B) The neutrophil-to-lymphocyte ratio at baseline (blue bars) and upon completion of the study (orange bars) is plotted for each participant.

Figure 3. The IFN-γ signature and CD274 expression increased by greater than 2-fold with 5-azacytidine in a subset of patients.

(A) Summary of plotted data. (B) IFN-γ signature score. (C) Expression of CD274 (PD-L1). Participants 2, 5, and 11 (red) showed significant increase in FC compared with others (black) (P = 0.036 for IFN-γ signature score and P = 0.036 for CD274 expression).

Figure 4. Cellular deconvolution indicates a decreased proportion of cancer cells in molecular responders, accompanied by an increased frequency of immune effector cells and decreased frequency of T_regs.

(A) Methylation-based deconvolution shows that the proportion of cancer cells decreased in participants 2, 5, and 11 (responders, red) compared with nonresponders (black), which was associated with an increase infiltration of CD8⁺ T cells (participants 2 and 5) and fibroblasts (participant 11). The proportion of T_regs decreased with 5-aza based on DNA methylation deconvolution. (B) Membrane CD8 and nuclear FOXP3 expression was quantified and analyzed in the baseline and on-treatment tissue sections. *P < 0.05, paired t test. (C) The expression ratio of M1-polarized macrophages versus M2 macrophages between baseline and on-treatment samples also increased in 75% of participants. Scale bars: 50 μm.

Figure 5. Fold change expression of genes relevant to tumor antigenicity after 5-aza treatment.

(A) Participants 5 and 11 had increased expression of HLA and (B) APM compared with participants that did not respond to 5-aza treatment. (C and D) Participant 11 had the greatest number and diversity of cancer testis antigen (CTA), demonstrating at least a 2-fold-change in expression, while participants 2, 6, and 10 did not express any CTAs. (E) Participants 2, 5, and 11 had increased expression of STING pathway genes compared with participants that did not respond to 5-aza treatment.

Figure 6. Gene set enrichment analysis of both methylation and expression datasets demonstrate hypomethylation and upregulation of immune response gene pathways (Participants 2 and 5) and developmental and differentiation pathways (Participant 11).

Participants 2 and 5 show simultaneous hypomethylation and upregulation of immune activation and response pathways (red), while participant 11 shows simultaneous hypomethylation and upregulation of developmental, differentiation, and morphogenesis pathways (blue). The size of each dot represents the number of genes associated with the GOBP terms, and the intensity scale of the dot represents the –log₁₀(P).

Figure 7. 5-aza treatment resulted in hypomethylation of tumor suppressor genes (TSGs).

(A) Treatment with 5-aza resulted in hypomethylation of TSGs commonly hypermethylated in HNSCCs. Specifically, the methylation profile of Participants 5 and 11 reversed to a methylation pattern observed in noncancer tissue samples representative within the HNSCC cohort in the Tumor Cancer Genome Atlas (TCGA). (B) PCA analysis based on the methylation of TSGs demonstrated that the on-treatment samples from participants 5 and 11 were most similar to normal tissue within the TCGA-HNSC cohorts. (C) Participants 5 and 11 had the greatest average decrease in methylation, approximately 12% and approximately 23%, respectively. (D) Participants 5 and 11 also had the greatest number of TSGs with at least a 25% hypomethylation change and 2-fold increase in expression.