MHC class I upregulation contributes to the therapeutic response to radiotherapy in combination with anti-PD-L1/anti-TGF-β in squamous cell carcinomas with enhanced CD8 T cell memory-driven response

Abstract

Radiation therapy (RT), a mainstay treatment for head and neck squamous cell carcinoma (HNSCC), kills cancer cells and modulates the tumor immune microenvironment. We sought to assess the effect of RT in combination with PD-L1/TGF-β dual blockade in squamous cell carcinomas (SCC) and analyze the underlying mechanisms. We transplanted mouse SCC cells derived from keratin-15 (K15) stem cells harboring Kras^G12D/Smad4^-/- mutations into syngeneic recipients and irradiated tumors followed by PD-L1/TGF-β dual blockade. We identified a responder line and a non-responder line to this combination therapy. Responder hosts eradicated SCCs by the combined therapy and rejected re-transplanted SCC cells 6 months post tumor eradication, which correlated with clonotype expansions of splenic CD8 T cells and effector memory gene expression identified by single cell sequencing of TCR and transcriptomes, respectively. Mechanistically, RT upregulated MHC-I (major histocompatibility complex I) and its transcriptional regulators including NLRC5, in SCCs of the responders but not non-responders. These data are consistent with the TCGA HNSCC database in which NLRC5 correlated to MHC-I genes and CD8 T cell gene expression. Functional contribution of MHC-I to PD-L1/TGF-β blockade response was confirmed by knocking out beta-2-microglobulin in responder cells that attenuated the response to the same therapy. Thus, the therapeutic effectiveness appeared to largely depend on cancer-cell MHC-I expression, triggering CD8 T cell effector memory-driven responses against tumor cell antigens. Identifying the differential RT response to MHC-I induction may serve as a predictive marker for stratifying patients that are most likely to benefit from this combination therapy.

Keywords: Immunotherapy; MHC class I; PD-L1; Radiotherapy; Squamous cell carcinoma; T cell; TGF-β.

Fig. 1.

K15.Kras^12D/Smad4^−/− SCCs tumors derived from A223 or P029 SCC lines exhibited differential therapeutic responses to RT in combination with PD-L1/TGFβ dual inhibition. Mice were implanted with A223 or P029 tumor cells, both characterized as K15.Kras^12D/Smad4^−/−. Once tumors reached ~400–500 mm³, mice began treatment with or without RT. After RT completion, mice were treated with either IgG isotype or bintrafusp alfa (BA) (A) Schematic of therapeutic regimen. ^¥ For BA or IgG without RT, they were given the first dose on day 0 (i.e., 1 week earlier than the combined therapy). (B) Kaplan-Meier plot showing survival after treatment of A223 tumors in each treatment group. Day 0 is defined as the first day of RT.*P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001. (C) Individual tumor volumes of A223 tumors after treatment. (D) Kaplan-Meier plot of P029 tumors in each treatment group. (E) Individual tumor volumes of P029 tumors.

Fig. 2.

Hosts with eradicated A223 tumors by RT in combination with PD-L1/TGFβ dual inhibition harbor spleno-CD8⁺ T-cell clonotype expansion of memory CD8 T cells after re-challenging with A223 cells. Mouse hosts with eradicated A223 tumors by RT + BA were reimplanted with A223 cells 6 months after all tumors were eradicated (“A223-eradicated”). In parallel, naïve mice were implanted with the same batch of A223 cells. Mice were euthanized at 2 timepoints (Early harvest, “EH”; late harvest, “LH”) and paired single cell RNA and TCR sequencing was conducted. (A) Timepoints of host splenocyte harvests of A223 rechallenged and naïve recipients (n = 5/group). Rechallenge was initiated 6 months after tumor eradication shown in Fig. 1A. EH: early harvest (delineated with first dotted line). LH: late harvest (delineated with second dotted line). (B) scTCRseq clonotype analysis reveals that within the splenic CD8⁺ T cell repertoire expanded clonotypes in A223-eradicated mice at EH compared to naïve hosts, and those expansions were no longer detectable at LH time points. Expanded clonotype numbers were grouped for each group at each time point. Bar color codes indicate the number of cells per clonotype. (C–H) scRNAseq analysis was performed from spleno-CD8 cells of EH; A223-eradicated hosts (n = 3) and naïve hosts n = 2. (C) UMAP and unsupervised clustering of all CD8⁺ T cells based on scRNAseq (D) Cluster proportion, based on scRNAseq, differs between A223-eradicated and naïve mice. On the left, each group is subsampled to 1000 cells for equal comparison. On the right, mean±SD is shown for clusters that differed the most between groups. (E) scTCRseq analysis shows differential clonality, as measured by 1 – Pielou’s index, between clusters. 0 represents a perfectly diverse population and 1 represents clonal dominance. (F) ProjecTILs tumor-infiltrating lymphocyte reference atlas projected over the UMAP of all cells, based on scRNAseq. (G) CD8⁺ central memory and effector memory scores projected over the UMAP of all cells, again based on scRNAseq. (H) Differentially expressed genes between expanded clonotypes (n > 2 cells) from spleno-CD8 cells of A223-eradicated vs. naïve hosts.

Fig. 3.

Radiation differentially affected MHC-I levels of A223 and P029 tumor cells in vivo. Tumors derived from transplanting A223 and P029 cells received 4 Gy every other day for 4 treatments, for a total of 16 Gy, when tumor volume reached 400 mm³. Control mice did not receive RT. Tumors were harvested 72h after the 4th RT for analysis. (A–B) Histogram of one representative tumor (A223 or P029) from each treatment group for H2Kb. Counts normalized to mode. (C–D) Mean fluorescence intensity (MFI) of surface MHC-I protein presentation of tumor cells measured by H2Kb (C) or H2Db (D) where each dot represents one tumor.

Fig. 4.

Radiation induced MHC-I protein levels and MHC-I-associated gene expression levels were tumor line dependent. A223 and P029 cells treated with different doses of RT or IFN-γ, showed differential MHC-I upregulation. (A, B) Flow cytometry histograms of cell surface MHC-I protein (H-2Db) expression of live cell populations of A223 or P029 cells post treatment. Isotype: FITC mouse IgG2a, κ (control for FITC anti-mouse H-2Db). (D) Crystal violet staining assay of irradiated or control cells reduced live cell counts at 96 h post-RT. ***P < 0.001 compared A223 and P029 at 2 Gy. N = 6 for each RT dose group. (E) qRT-PCR performed on A223 or P029 cells treated as in A-B. Relative gene expression levels were normalized to GAPDH and baseline gene expression. N = 3 per group.*P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.

Fig. 5.

Knocking out essential MHC-I component β2M attenuated response to PD-L1/TGFβ dual inhibition in A223 tumor bearing mice. A223 β2M-CRISPR knockout cells, or parental A223 cells, were cultured and implanted into mice that were later treated with either BA or IgG. (A) Histograms of flow cytometry data confirms that A223 β2M-CRISPR knockout cell lines lost surface MHC-I (H2) compared to A223 parental cell line. (B) Kaplan-Meier plot of survival after treatment. ****p < 0.0001 between parental and β2M-CRISPR knockout cell line-derived tumors. Log-rank (Mantel-Cox) test rejects the null hypothesis that there is no difference in survival outcomes of different treatment groups (Chi-square 38.66, df = 7, P < 0.0001). (C) Individual tumor volumes after treatment.