ATR-mediated CD47 and PD-L1 upregulation restricts radiotherapy-induced immune priming and abscopal responses in colorectal cancer

RT upregulates CD47 and PD-L1 in CRC cells

To unravel the immune escape mechanisms deployed by CRC cells after radiotherapy, we utilized multiparameter flow cytometry to systematically evaluate the expression levels of various immune checkpoints on two human CRC cell lines, HCT116 and HT29, versus normal CCD-33Co cells treated with or without 8-Gy irradiation. Notably, our results demonstrated that CD47 and PD-L1 were more highly expressed on CRC cells at baseline compared with normal cells, and their expression was further upregulated 24 hours after radiotherapy (fig.S1). In addition, RT-induced CD47 and PD-L1 upregulation was consistently observed in HCT116 and HT29 cells using either photons (8 Gy) or protons (8 cobalt Gy equivalent, CGE) (fig.1A, fig.S2A, fig.S3A), and the enhanced CD47 and PD-L1 expression was dose-dependent and persisted for at least 6 days after RT (fig.1B–C, fig.S2A, fig.S3B). We also observed that RT-induced CD47 and PD-L1 upregulation was more pronounced on the surviving cancer cells compared with those undergoing apoptosis using either 1.8 Gy or 5 Gy per fraction dose regimen (fig.S4). Radiotherapy induces PD-L1 expression in CRC patients (23); however, the clinical relevance of RT-induced CD47 upregulation remains unclear. Thus, we compared pre- vs. post-irradiation CD47 surface expression measured by immunohistochemistry (IHC) in 30 rectal cancer patients who underwent short-course neoadjuvant radiotherapy (5 Gy for 5 fractions) and surgery. The CD47 histoscores on post-irradiation surgical specimens (mean interval from RT to surgery: 8.2 ± 1.2 days) were significantly increased compared with the paired pretreatment biopsy samples (fig.1D). Moreover, this post-irradiation CD47 enhancement was significantly higher in patients with poor responses to RT (tumor regression grade [TRG] 4–5; fig.1E–F and table S1), supporting the hypothesis of CD47 upregulation as a driver of cancer immune evasion after radiotherapy. Next, we screened 12 distinct human solid tumor cell lines treated with or without 8-Gy irradiation and found a significant increase in CD47 and PD-L1 expression across different cancers (fig.1G, fig.S5). This suggests that RT-induced phagocytosis checkpoint upregulation may be an immune escape mechanism shared across an array of cancers.

RT-induced CD47 and PD-L1 upregulation depends on ATR activity

Radiation-induced cell death is primarily attributed to irreparable DNA double-strand breaks (24–26). CRC cells have a strong dependency on ATR-mediated homologous recombination (HR) DNA double-strand break repair for maintaining genomic integrity and survival in response to radiation insults (24–26). Post-irradiation ATR-Chk1 activation stimulates PD-L1 expression through STAT1/3 signaling (27–29). To interrogate whether RT-induced CD47 upregulation was also mediated by the ATR pathway, we utilized short-hairpin RNA (shRNA) to knockdown ATR expression in HCT116 and HT29 cells (fig.1H, fig.S3C) and compared checkpoint expression levels with or without 8-Gy irradiation. Concordant with previous studies (27, 28), we observed that ATR depletion suppressed downstream Chk1 and STAT3 phosphorylation and PD-L1 upregulation in response to DNA damage (Fig.1H, fig.S3C). Knockdown of ATR also diminished the mRNA and protein expression (fig.1H–J, fig.S2C–D) of CD47 after irradiation, suggesting that RT-induced CD47 upregulation depends on upstream ATR signaling as well. To further assess the above finding that ATR activation is required for CD47 and PD-L1 upregulation in response to radiation damage, we utilized an ATR inhibitor (ATRi), VE822, or a Chk1 inhibitor (Chk1i), UCN-01, to suppress the ATR-Chk1 pathway. Our results demonstrated that RT-induced CD47 and PD-L1 upregulation was significantly attenuated with either ATRi or Chk1i (fig.1K–L, fig.S2E–F), indicating that ATR-Chk1 signaling mediates DNA damage-induced phagocytosis checkpoint augmentation. Notably, recent studies have identified STAT3 as a key transcription factor driving CD47 expression (30, 31). Phosphorylated STAT3 augments CD47 transcription in hepatocellular carcinoma exposed to interleukin (IL)-6, and the addition of a STAT3 inhibitor, NSC74859, disrupted IL-6-induced CD47 upregulation in Huh7 and PLC cells (30). Consistent with their findings, our results demonstrated that STAT3 inhibition diminished CD47 enhancement after 8-Gy irradiation (fig.1M, fig.S2G). Taken together, these results suggest that RT-induced CD47 upregulation is likely regulated by the ATR-Chk1-STAT3 pathway.

RT induces cytosolic double-stranded DNA accumulation in cancer cells. Phagocytosis of damaged cells which contain cytosolic dsDNA species can trigger cGAS-STING and downstream interferon regulatory factors (IRF) and NF-κB signaling in APCs fostering innate immune activation (32–34). To determine whether ATR abrogation makes irradiated cancer cells susceptible to macrophage engulfment and promotes IRF and NF-κB signaling in macrophages, we performed phagocytosis assays using differentiated THP1-Dual cells which express luciferase (Lucia) and secreted embryonic alkaline phosphatase (SEAP) reporter genes driven by IRF and NF-κB signaling, respectively. We found that RT combined with ATRi significantly increased phagocytosis of HCT116 cells by THP1-Dual macrophages (fig.1N, fig.S6), reaffirming that ATR activation plays a role in post-irradiation phagocytosis evasion. Furthermore, significantly higher IRF and NF-κB activities were recorded in THP1-Dual cells co-cultured with HCT116 cells treated with RT and ATRi, indicating that improved phagocytosis of damaged tumor cells potentiated innate defense responses in macrophages. The Cancer Genome Atlas Pan-Cancer dataset (TCGA-PANCAN, n = 12839, fig.S3H) also documented a significant positive correlation between ATR expression and CD47 and PD-L1 levels, positing a link between DNA double-strand breakrepair machinery and phagocytosis evasion. Collectively, our results suggest ATR as a crucial factor regulating CD47 and PD-L1 expression in response to DNA damage.

Phagocytosis checkpoint blockade potentiates RT-induced antitumor responses

Given the significant post-RT CD47 (Fig.1D–F, table S1) and PD-L1 (23) enhancement in rectal cancer patients, we next investigated whether this phenomenon was a critical barrier to radiation-induced antitumor immunity. We implanted C57BL/6J mice with subcutaneous MC38-Ovalbumin (OVA) tumors bilaterally and treated with or without (1) focal 8-Gy irradiation tothe ipsilaterla tumors, (2) anti-CD47 or anti-SIRPα (the cognate receptor of CD47) antibody i.p., and (3) anti-PD-1 antibody (fig.2A). Triple therapy with either RT/anti-SIRPα/anti-PD-1 or RT/anti-CD47/anti-PD-1 significantly improved antitumor activity for both irradiated and abscopal tumors and significantly extended overall survival (fig.2B–C, fig.S7). We observed significantly higher bilateral CR rates of 48% and 33% in mice treated with RT/anti-SIRPα/anti-PD-1 and RT/anti-CD47/anti-PD-1, respectively, compared with mono- or dual-therapies (table S2). Furthermore, subcutaneous tumor rechallenge was performed in mice with bilateral CR on day 42 (fig.S8), and the rechallenged tumors were rejected in all mice, indicating the generation of antitumor immune memory. In addition, anti-SIRPα exhibited significantly higher therapeutic efficacy than anti-CD47 in combination with RT and anti-PD-1 (fig.2C).

Since anti-SIRPα had better antitumor efficacy compared to anti-CD47, we focused on the therapeutic effect of anti-SIRPα in combination with RT and anti-PD-1 in our subsequent experiments. Given that OVA is a highly immunogenic foreign antigen, the OVA-expressing tumor models may not have recapitulated the immunogenicity of human CRC in which the majority of TAAs are weakly- or non-immunogenic (35). In addition, colon orthotopic tumors are more resistant to immunotherapy compared with subcutaneous CRC models due to their distinct tumor microenvironment (TME) (36–38). Fractionated radiotherapy regimens are more commonly used in CRC patient as opposed to single-dose irradiation. To address these caveats, we implanted MC38-luciferase cells orthotopically in the cecum and wild-type MC38 cells on the right flank subcutaneously. We then performed fractionated radiotherapy with 5 Gy for 3 fractions to the right flank tumors with or without anti-SIRPα and anti-PD-1 antibodies (fig.2D). We observed that triple therapy led to highly significant improvements in survival and CR rates of both irradiated (subcutaneous) and abscopal (cecal) tumors, compared to those treated with anti-SIRPα and anti-PD-1 or RT alone (fig.2E–G, fig.S9A, table S3). In addition, 4 out of 9 mice treated with RT/anti-SIRPα/anti-PD-1 had CR of both tumors, and rechallenged MC38-luciferase cells were eradicated in these mice, reconfirming the establishment of systemic sterilizing antitumor memory (fig.2E, fig.S9A). Considering that MC38 is a hypermutated/microsatellite instable (MSI) CRC which is likely more responsive to immunotherapies (39, 40), we also studied a microsatellite stable (MSS) CMT-93 tumor model to re-examine the therapeutic efficacy of RT/anti-SIRPα/anti-PD-1(41). We implanted CMT-93 cells on both flanks of C57BL/6J mice treated with or without radiotherapy to unilateral tumors with 5 Gy for 2 fractions, anti-SIRPα, and/or anti-PD-1 antibodies (fig.2H–I, fig.S9B). Concordant with our previous results, we observed significantly lower growth kinetics of both irradiated and abscopal tumors in mice treated with RT/anti-SIRPα/anti-PD-1. Triple therapy significantly increased survival and bilateral CR rates (table S4) and triggered sustained antitumor immune memory against rechallenged CMT-93 cells in cured mice (fig.S9C). Together, these three distinct murine CRC models all demonstrated that radiotherapy combined with SIRPα and PD-1 blockade elicited a robust systemic antitumor response and immune memory.

RT/anti-SIRPα/anti-PD-1 promotes TAA uptake, APC activation, and TAA cross-priming

To verify whether radiotherapy triggered CD47 and PD-L1 upregulation and expression of calreticulin, a pro-phagocytic DAMP, in MC38 cells in vivo, we harvested subcutaneous MC38-OVA tumors for flow cytometry analysis 2–8 days after 8-Gy radiotherapy. Compared with sham controls, significant increases of CD47, PD-L1, and calreticulin were observed in the irradiated tumor cells (fig.3A, fig.S2B, fig.S10). We next performed phagocytosis assays using bone marrow-derived dendritic cells (BM-DCs) to determine whether RT coupled with phagocytosis checkpoint blockade augmented tumor cell engulfment by APCs (fig.3B, fig.S11A). Anti-SIRPα and anti-PD-1 alone failed to stimulate phagocytosis which was likely due to the lack of trigger “eat me” signals such as calreticulin on tumor cells. In contrast, combination therapy with RT/anti-SIRPα/anti-PD-1 significantly enhanced the scavenging of MC38-OVA cells, suggesting that the pro-phagocytic signals resulting from RT are critical for phagocytosis initiation which is then enhanced with phagocytosis checkpoint blockade.

Antigen cross-presentation to T cells requires three crucial components: 1) surface presentation of MHC-peptide complexes, 2) provision of co-stimulatory signals, and 3) supportive release of pro-inflammatory cytokines (42). To examine whether RT/anti-SIRPα/anti-PD-1 facilitated TAA cross-presentation on MHC complexes and mobilizes APCs in vivo, we harvested the irradiated MC38-OVA tumors and spleen for flow cytometry analysis in mice undergoing the aforementioned treatments. Significantly higher expression of OVA-derived SIINFEKL peptide on H-2K^b-complexes was observed on dendritic cells (DCs; fig.3C, fig.S2C) including the CD11c⁺CD11b⁻CD8⁺ type I conventional DCs (cDC1; fig.S2C, fig.S11B–C), monocytes, and macrophages (Mo-Macs; fig.3C, fig.S2C, fig.S11D) in the irradiated TME and spleen in the triple therapy group, suggesting that RT combined with anti-SIRPα and anti-PD-1 promotes TAA cross-presentation in APCs and their migration to secondary lymphoid organs.

We next asked whether radiotherapy combined with anti-SIRPα and anti-PD-1 antibodies would improve the co-stimulatory functions of APCs. We measured a variety of co-stimulatory and –inhibitory molecules on the antigen-presenting myeloid cells in the irradiated TME. Of note, RT alone upregulated multiple immunosuppressive molecules including PD-1, PD-L1, and Arginase I (Arg1) on DCs and Mo-Macs, albeit also with a slight increase in CD86 expression (fig.3D, fig.S2C, fig.S12). Particularly in the H-2K^b-SIINFEKL⁺ populations, substantial augmentation of PD-1, PD-L1, and Arg1 was recorded in the RT alone group (fig.S13), indicating the tolerogenic potential of TAA-presenting APCs in this setting. On the contrary, RT/anti-SIRPα/anti-PD-1 prevented PD-L1 and Arg1 upregulation and profoundly enhanced CD86 expression on DCs and Mo-Macs (fig.3D, fig.S12) with the highest expression of CD86 in the H-2K^b-SIINFEKL⁺ APCs (fig.S13), suggesting that APCs in the triple therapy group would have the greatest co-stimulatory potential. Consistent with existing literature (43), our results showed that RT alone decreased F4/80⁺CD206⁺ M2-macrophages but also triggered CD11b⁺Ly6G⁻Ly6C^hi monocytic myeloid-derived suppressor cells (M-MDSCs) accumulation (fig.3E, fig.S2C, fig.S14). In contrast, RT/anti-SIRPα/anti-PD-1 significantly reduced M2-macrophages but did not increase the proportion of M-MDSCs in the TME. These findings collectively suggest that triple therapy transformed the tolerogenic TME myeloid population landscape following RT into a pro-inflammatory one dominated by immunostimulatory phenotypes.

To compare the global transcriptome alterations in APCs among different treatments, we isolated CD11b⁻CD11c⁺ DCs, CD11b⁺CD11c⁺ DCs and CD11b⁺CD11c⁻ myeloid cells from the irradiated tumors for RNA-seq analysis (fig.3F, fig.S15). The hierarchical clusters of the top 10 enriched gene ontology (GO) pathways showed significant enrichment in the adaptive immune response- and cytokine production-related pathways across different DC subsets in the RT/anti-SIRPα/anti-PD-1 group. Furthermore, triple therapy conferred significant enrichment in innate immune response, immune effector process, and defense response to virus pathways in the CD11b⁺ APC populations, highlighting the profound activation of innate defense mechanisms in the monocytic lineage. In summary, these data suggest that RT coupled with anti-SIRPα and anti-PD-1 antibodies substantially magnifies TAA cross-presentation, co-stimulatory molecule expression, and cytokine production-related signaling in the myeloid compartment, ultimately conferring highly efficient TAA cross-priming capability.

Host STING activation is essential for RT/anti-SIRPα/anti-PD-1-induced immune priming

cGAS-STING-mediated type I IFN production in the host immune system is required to ignite RT-induced adaptive immune responses (8). We found that anti-SIRPα and anti-PD-1antibodies augmented cGAS expression, STING and downstream interferon regulatory factor 3 (IRF3) phosphorylation, and IFNα and IFNβ transcription in bone marrow-derived dendritic cells (BM-DCs) co-cultured with irradiated MC38-OVA cells (fig.4A–B). Furthermore, differentially expressed gene (DEG) analysis consistently documented enrichment in type I IFN production (GO: 0032606) related genes across different myeloid subsets in the RT/anti-SIRPα/anti-PD-1 group in vivo (fig.4C; fig.S15–16), suggesting that increased phagocytosis of irradiated tumor cells potentiated activation of the cGAS-STING innate immune sensor circuit and subsequent type I IFN production in APCs. To explore the role of host STING activity in response to RT/anti-SIRPα/anti-PD-1 in vivo, MC38-OVA cells were subcutaneously implanted in bilateral flanks of wild-type (WT) versus STING-deficient (STING^−/−) mice and treated with ipsilateral 8-Gy RT with or without anti-SIRPα and anti-PD-1 antibodies. As we posited, the systemic tumoricidal effect of RT/anti-SIRPα/anti-PD-1 was significantly compromised in STING^−/− mice, compared with WT mice (fig.4D–F, fig.S17). The survival outcomes and CR rates of irradiated and abscopal tumors in STING^−/− mice treated with RT/anti-SIRPα/anti-PD-1 were similar to WT mice undergoing RT alone, indicating that the therapeutic advantages of adding phagocytosis checkpoint blockade to RT depend on host STING activation. Furthermore, we noted decreased production of H-2K^b-SIINFEKL-tetramer⁺ CD8 T cells and remarkably lower expression of CD44, ICOS, and surface LAMP1 on CD8 T cells in the peripheral blood of STING^−/− mice compared with WT animals indicating dampened activation (fig.4G–H, fig.S18). Similarly, significantly higher CD44⁻CD62L⁺ naïve and lower CD44⁺CD62L⁻ effector CD8 T cell frequencies were recorded in the bloodstream of STING^−/− mice compared to those of WT mice after RT/anti-SIRPα/anti-PD-1 treatment, suggesting that knockout of host STING profoundly hampered CD8 T cells transitioning from naïve to effector/effector-memory phases in response to triple therapy (fig.4I). These findings collectively demonstrated that host STING activation is critical for TAA cross-priming and cytotoxic T cell activation and dictates the systemic antitumor responses induced by RT/anti-SIRPα/anti-PD-1.

RT/anti-SIRPα/anti-PD-1-induced TAA-specific CD8 T cell cross-priming is predominantly driven by DCs, but not macrophages

To functionally characterize the cross-priming capability of different antigen-presenting myeloid subpopulations, we harvested CD11b⁻CD11c⁺ DCs (mostly cDC1), CD11b⁺CD11c⁺ DCs, and CD11b⁺CD11c⁻F4/80⁺ TAMs from irradiated MC38-OVA tumors treated with or without 8-Gy irradiation and/or phagocytosis checkpoint blockade for ex vivo cross-priming of OT-1 cells (fig.S19). We observed the highest levels of granzyme B, IFN-γ, Ki-67, ICOS, CD44, and PD-1 in OT-1 cells co-cultured with CD11b⁻CD11c⁺ DCs, but not with TAMs, sorted from the RT/anti-SIRPα/anti-PD-1 group (fig.5A–D), indicating the superior effector function, proliferative capability, and activation status of TAA-specific CD8 T cells. Conversely, OT-1 cells treated with anti-SIRPα or anti-PD-1 antibody alone in the absence of APCs remained as functionally inert as untreated controls, suggesting that the activation or maturation of TAA-specific CD8 T cells is driven by APC cross-priming in lieu of the direct effect of anti-SIRPα or anti-PD-1 antibodies. Further, cytometric bead array profiling consistently demonstrated the highest production of secreted IFN-γ and IL-17a in OT-1 cells co-cultured with CD11b⁻CD11c⁺ DCs isolated from the RT/anti-SIRPα/anti-PD-1 group (fig.5E, fig.S20), signifying the exceptional cross-priming capability of cDC1 after triple therapy. RT/anti-SIRPα/anti-PD-1 significantly decreased IL-6 and IL-10 and increased TNFα production in both CD11b⁺CD11c⁺ DCs and TAMs (fig.5E, fig.S20, reconfirming that radiotherapy coupled with phagocytosis checkpoint blockade can functionally repolarize immunosuppressive myeloid subsets into pro-inflammatory phenotypes.

RT/anti-SIRPα/anti-PD-1 triggers TAA-specific CD8 T cell expansion and activation

We next investigated whether the improved cross-priming capability induced by RT/anti-SIRPα/anti-PD-1 results in enhanced quantitative and functional priming of TAA-specific CD8 T cells. RT/anti-SIRPα/anti-PD-1 significantly increased generation of H-2K^b-SIINFEKL-tetramer⁺ CD8 T cells in both irradiated and abscopal tumors as well as peripheral blood in mice bearing MC38-OVA tumors (fig.6A). This indicated that triple therapy enables expansion and mobilization of TAA-specific CD8 T cells which can infiltrate non-irradiated, TAA-expressing tumor sites. The accumulation of M2-macrophages/MDSCs in the context of IL-10 secretion potentiates Treg differentiation and tumor growth (44–47). We observed significantly reduced FoxP3⁺ regulatory T (Treg) cells and increased CD8/Treg ratios in the irradiated TME in the RT/anti-SIRPα/anti-PD-1 group (fig.6B, fig.S2C, fig.S21A), suggesting that the repolarized myeloid compartment (decreases in M2-macrophages/M-MDSCs [fig.3E] and IL-10 [fig.5E]) supports a pro-inflammatory T cell profile. In irradiated tumors, the total (fig.6C–D, fig.S21B–C) and H-2K^b-SIINFEKL-tetramer⁺ (fig.S22) CD8 T cells of the RT/anti-SIRPα/anti-PD-1 group demonstrated significantly lower PD-1 mean fluorescent intensity (MFI) with similar percentages of PD-1⁺ cells compared with untreated controls, indicative of a recently activated phenotype. We also observed significantly higher Granzyme B and Ki-67 in total (fig.6D, fig.S21C) and H-2K^b-SIINFEKL-tetramer⁺ (fig.S22) CD8 T cells in both irradiated and abscopal tumors in the RT/anti-SIRPα/anti-PD-1 group, suggesting that phagocytosis checkpoint blockade restores effector functions and proliferative capacity of cytotoxic T lymphocytes. Anti-PD-1 alone without adequate antigen stimulation induces PD-1⁺CD38^hi subprimed CD8 T cells that are associated with anti-PD-1 resistance, and depletion of PD-1⁺CD38^hi CD8 T cells improves treatment outcomes (48). Consistent with their observation, we saw higher PD-1⁺CD38^hi dysfunctional CD8 T cells in mice treated with anti-PD-1 alone (fig.6E, fig.S2D). In contrast, RT/anti-SIRPα/anti-PD-1significantly decreased PD-1⁺CD38^hi subprimed CD8 T cells compared with anti-PD-1 alone and increased ICOS on CD8 T cells (fig.6E, fig.S21D), reconfirming the superior antigen cross-priming of cytotoxic T lymphocytes. To compare the transcriptomic alterations between different treatments, CD8 T cells from irradiated TME were isolated for RNA-seq analysis (fig.S15). Notably, the hierarchical clusters of the top 10 enriched GO pathways revealed significant enhancement of T cell activation, lymphocyte co-stimulation, and T cell receptor (TCR) signaling-related pathways in the RT/anti-SIRPα/anti-PD-1group (fig.6F). DEG analysis consistently demonstrated that triple therapy stimulated T cell activation (GO: 0042110) related genes such as IFNG, ICOS, and TCF7 (fig.6G–H). Taken together, these results indicated that RT coupled with anti-SIRPα and anti-PD-1 antibodies magnifies adaptive antitumor immunity through improving T cell cross-priming, restoring and maintaining effector function, and expanding the frequency of TAA-specific CD8 T cells.

RT/anti-SIRPα/anti-PD-1 expands CD8 T cell clonality and clonal diversity

To further characterize the nature of the RT/anti-SIRPα/anti-PD-1 expanded anti-tumor T cell populations, we performed high-throughput sequencing of TCRβ CDR3 to comprehensively characterize the tumor-infiltrating CD8 T cell repertoire in mice bearing MC38-OVA tumors. Higher numbers of tumor-infiltrating CD8 T lymphocytes better predict the response to neoadjuvant chemoradiotherapy and lead to a lower risk of recurrence in rectal cancer patients (49, 50). Similarly, we saw increased CD8 T cell counts in both irradiated and abscopal tumors in the RT/anti-SIRPα/anti-PD-1 group (fig.7A, fig.S23), suggesting the possible formation of intratumoral niches for cytotoxic T cells. A higher T cell clonality, which is indicative of a larger proportion of the repertoire constituted by a few dominant clonotypes, is associated with superior outcomes in both patients and preclinical models (51–53). Our results showed RT alone enhanced CD8 T cell clonality only in irradiated tumors but not in abscopal sites (fig.7B), which is concordant with published results (54). We observed significantly increased clonality indices in both irradiated and abscopal tumors and higher numbers of unique CD8 T cell clones in the RT/anti-SIRPα/anti-PD-1group (fig.7C–D), suggesting that improved phagocytosis of irradiated tumor cells might amplify the clonality and clonal diversity of tumor-infiltrating T cells systemically. Furthermore, the numbers of expanded CD8 T cell clones with counts greater than 1000 in both tumors were significantly higher in the RT/anti-SIRPα/anti-PD-1 group (fig.7D). This observation reconfirmed our flow cytometry findings on H-2K^b-SIINFEKL-tetramer⁺ CD8 T cells wherein RT/anti-SIRPα/anti-PD-1licensed robust clonal expansion of cytotoxic T lymphocytes and mobilized these cells to systemically eradicate TAA-expressing tumors (fig.6A). Collectively, our study demonstrated that the superior antigen cross-priming capability induced by RT/anti-SIRPα/anti-PD-1 reshaped the landscape of the tumor-infiltrating CD8 T cell repertoire toward a broader spectrum of unique TCRs and a higher abundance of selective tumor-focused clonotypes.

Study Design

This study was undertaken to interrogate the immune escape mechanisms in CRC following radiotherapy. Multiple murine syngeneic tumor models were used, and tumor-bearing mice were randomly assigned to different treatments by tumor volume or bioluminescence intensity. CD47 expression between pretreatment biopsy and post-irradiation surgical specimens in rectal cancer patients who underwent neoadjuvant radiotherapy with 5 Gy for 5 fractions were compared using immunohistochemistry. Flow cytometry, gene knockdown, and RNA and T cell receptor sequencing were utilized to map the relevant immune evasion pathways. The sample sizes were determined empirically to ensure adequate statistical power and specified in each figure legend and in table S6.

Cell lines and lentiviral transduction

Human cell lines including HCT116, HT29, CaSki, NCI-H2052, SUN449, SNU423, PANC1, NCI-H2228, SNU387, Hs746.T, Hs255.T, U87, SNU475, CCD-33Co, and 293T and murine CMT93 cells were purchased from ATCC, whereas MC38 and THP1-Dual cells were obtained from Kerafast and Invivogen, respectively. The cell culture procedures complied with the protocols provided by the vendors and routinely tested for mycoplasma. HCT116 and HT29 cells were lentivirally transduced with mCherry-shATR (target sequences: GGAATATAATACAGTTGTACA, GCGGCTTAAGTCTGATTTGCT, and GCCATTATCAAAGCTGATAAA) or mCherry-scrambled shRNA vectors, whereas MC38 cells were transduced with mCherry-Luciferase lentiviral particles according to manufacturer’s protocols (GeneCopoeia). MC38 cells expressing OVA and GFP were obtained from the lab of Dr. Steven H. Lin (MD Anderson Cancer Center, TX). Lentiviral particles were generated by transfecting 293T cells with 5 μg DNA vectors, 2 μg pMD2.G, and 4 μg pCMV-dR8.91 in 6 cm dishes, and virus-containing supernatant plus polybrene (10 μg/ml) was used to treat cells. After 48–72 hours, cells with high mCherry expression were selected by cell sorters (BD, NJ).

Radiotherapy and drug treatment

Photon radiotherapy was performed using the J.L. Shepherd ¹³⁷Cs γ-irradiator for in vitro experiments. Proton irradiation was delivered using synchrotron-based, passively scattered beams, and the relative biological effectiveness of protons was set at 1.1 for CGE calculation (75, 76). For animal irradiation, cone beam computed tomography-guided focal radiotherapy (beam diameter: 1.5 cm) was performed using the kilovoltage X-RAD SmART irradiator (Precision X-Ray, CT). Anti-CD47 (clone MIAP-410, 400 μg/dose/mouse), anti-SIRPα (clone P84, 400 μg/dose/mouse), anti-PD-1 (clone RMP1–14, 250 μg/dose/mouse) were administered intraperitoneally in mice. The antibodies and reagents used in the current study are summarized in table S5.

Immunohistochemical staining of human CRC samples

With Institutional Review Board approval (202001191B0C601), the pretreatment biopsy and post-radiotherapy surgical specimens in 30 CRC patients were retrospectively retrieved from the Tissue Bank of Chang Gung Memorial Hospital. Paraffin-embedded sections were stained with CD47 (1:100) IHC as previously described methodology (77). The membranous staining intensity for each tumor cell was categorized into negative (0), weak (1+), moderate (2+), or strong (3+) expression, and the histoscore was calculated using the formula: [1 × (% cells 1+) + 2 × (% cells 2+) + 3 × (% cells 3+)] (78). All patients were treatment-naïve at the time of biopsy. Informed consent was obtained from all patients. The patient and treatment characteristics were summarized in table S1.

Mice and syngeneic tumor models

Six-week-old C57BL/6J (#000664) and C57BL/6J-Sting1^gt/J (#017537, STING^−/−) mice were purchased from Jackson Laboratory and maintained at a pathogen-free facility at MD Anderson Cancer Center. Tumor engraftment or femoral bone marrow (BM) harvest was performed at age 6–10 weeks. For subcutaneous tumor models, 1 × 10⁶ tumor cells in 50 μl of 30% matrigel were engrafted into flank regions. For the orthotopic model, we implanted 1 × 10⁶ MC38-luciferase cells in 25μl of 100% high-concentration matrigel into cecum according to the protocol described by Chan et al (79). Subcutaneous tumors were measured using calipers, and the volumes were calculated by the formula: volume = (length × width²)/2. The growth of cecal tumors was monitored by IVIS Spectrum bioluminescence imaging system (PerkinElmer) 10 minutes after intraperitoneal injection of D-luciferin (150 mg/kg). Mice lacking palpable tumors (at either irradiated or abscopal site) or bioluminescent activity were excluded before treatment randomization. Blinding was not performed in animal experiments since the investigators needed to know the treatment arms to conduct the study. Mice with sum tumor volume ≥2000 mm³, ulcer ≥5 mm, hypoactivity, or a hunched posture were humanely euthanized. All animal experiments were performed in compliance with the approved protocol (00001378-RN01/RN02) by the Institutional Animal Care and Use Committee at MD Anderson Cancer Center.

Flow cytometry

Flow cytometry was performed on BD LSR-II or LSRFortessa X-30 cytometer and analyzed using FlowJo v7.6.5 and v10.7. Subcutaneous tumor engraftment with 2 × 10⁶ MC38-OVA cells in 50 μl of 30% matrigel was performed on bilateral flank regions, and tumor-infiltrating leukocytes were separated using Histopaque-1119 after tumor dissociation. Cells were stained with fixable Dead-Cell-Stain-Kit and then blocked with anti-FcR. H-2K^b-SIINFEKL-tetramers were obtained from the MHC Tetramer Core Facility (Baylor College of Medicine, TX) and stained before fixation. Sample fixation was performed using the Foxp3/Transcription-Factor-Staining-Buffer-Set. Fluorophore-conjugated antibodies were stained at 4 °C for 1 hour.

Phagocytosis assays and bone marrow-derived dendritic cells

For the in vitro human cell experiments, THP1-Dual cells were pretreated with 200 nM PMA for 3 days for macrophage differentiation. HCT116 cells were irradiated with 8 or 0 Gy +/− concomitant 300 nM VE822 in culturing medium. At 48 hours post-irradiation, HCT116 cells were labeled with CFSE and co-cultured with differentiated THP1-Dual cells with 1:3 ratio for 4 (for phagocytosis assays) or 18 hours (for IRF and NF-κB assessment) at 37 °C. Phagocytosis was calculated as the percentage of CFSE⁺ cells in CD64⁺ THP1-dual macrophages (fig.1N, fig.S6). 50 μl of QUANTI-Luc or 180 μl of QUANTI-Blue were added to 10 or 20 μl of THP1-Dual cell supernatant of each sample, respectively. The Lucia signal was measured immediately after adding QUANTI-Luc to cell supernatant, whereas the SEAP activity was evaluated at OD655 after 1-hour 37°C incubation of QUANTI-Blue-cell supernatant mixture.

For murine assays, BM cells were cultured in RPMI with 10% FBS and 1% penicillin-streptomycin plus IL-4 (20 ng/ml), GM-CSF (20 ng/ml), and 50 μM β-mercaptoethanol for 6 days with fresh medium replacement every 3 days for BM-DC generation. MC38-OVA cells were labeled with CFSE 48 hours after 8-Gy or sham irradiation. Differentiated BM cells were pretreated +/− anti-SIRPα (20 μg/ml) and anti-PD-1 (12.5 μg/ml) for 30 minutes and co-cultured with MC38-OVA cells (tumor: BM-DC = 1:3) under the same antibody condition for 1 (for immunoblotting and RT-PCR) or 4 hours (for phagocytosis assays) at 37 °C. Phagocytosis was measured as the percentage of CFSE⁺ cells in CD11c⁺ BM-DCs. For immunoblotting and RT-PCR, BM-DCs were isolated using CD11c MicroBeads.

Ex vivo OT-1 cross-priming assay and secreted cytokine measurement

Splenic OT-1 cells were harvested from 6–12 week old C57BL/6-Tg(TcraTcrb)1100Mjb/J mice (#003831) using the CD8α⁺-T-Cell-Isolation-Kit. CD11b⁻CD11c⁺ DCs (mostly cDC1), CD11b⁺CD11c⁺ DCs, or CD11b⁺CD11c⁻ F4/80⁺ TAMs were isolated from the irradiated MC38-OVA tumors by cell sorting (BD FACS-ARIA-III/Jazz) 5 days after 8 Gy or sham irradiation +/− anti-SIRPα and anti-PD-1 (fig.S19). The unstimulated OT-1 cells were co-cultured with different APC subsets (OT-1: APC = 140,000: 70,000) or anti-SIRPα (20 μg/ml) and/or anti-PD-1 (12.5 μg/ml) antibodies alone in RPMI with 10% FBS and 1% penicillin-streptomycin plus 50 μM β-mercaptoethanol for 48 hours in a 96-well plate. Cells were treated with GolgiPlug 6 hours before intracellular cytokine staining. Secreted cytokines in the supernatant were measured using the CBA Mouse-Th1/Th2/Th17-Cytokine-Kit.

Immunoblotting

Cells were washed with cold PBS and lysed with Cell-Lysis-Buffer containing protease and phosphatase inhibitors. After gel electrophoresis (4–15% gel), proteins were transferred onto nitrocellulose membranes and blocked in 5% bovine serum albumin for 1 hour. The membranes were incubated with primary antibody overnight at 4 °C and then with horseradish peroxidase-conjugated secondary antibodies with buffer dilution according to manufacturer’s specification, and the primary antibodies were detected by chemiluminescence. All the antibodies used in this study were listed in table S5, and original immunoblot images were presented in fig.S24.

RT-PCR

RNA was extracted by using RNeasy-Kit (Qiagen) and reversely transcribed using SuperScript-IV-First-strand-Synthesis-Kit (ThermoFisher). Quantitative PCR was performed using ViiA 7 system (ThermoFisher), and gene expression was normalized to GAPDH. Fold change comparisons of CD47, IFNα, and IFNβ were analyzed using the ΔΔCT method.

RNA Sequencing

Leukocytes in the irradiated tumors were collected and dispersed into single-cell suspensions on day 17 from MC38-OVA tumor-bearing mice treated +/− 8-Gy radiotherapy (day 12), anti-SIRPα (day 12, 14, and 16), and/or anti-PD-1 (day 12 and 15). To increase the RNA yields and sample representativeness, we pooled the tumor-infiltrating leukocytes from 10 tumors for each group. CD11b⁻CD11c⁺, CD11b⁺CD11c⁺, CD11b⁺CD11c⁻ myeloid cells, and CD8 T cells were stained with fluorochrome-conjugated antibodies and isolated using cell sorters (BD FACSAria/Fusion Flow Cytometer, fig.S15), and RNA was extracted immediately after cell sorting using the RNeasy Kit. RNA-sequencing was performed by Avera in 200-base pair paired-end runs. Fastq files were aligned to genome using HISAT/StringTie/Ballgown (80). Differential expression between experimental arms and untreated control was determined using the edgeR package v3.34.0 in Bioconductor v3.12 per User’s Guide (81, 82), and upregulated DEGs in type I IFN production (GO: 0032606; Fig.4C) and T cell activation (GO: 0042110; Fig.6G) genesets in the RT/anti-SIRPα/anti-PD-1 group were depicted. Gene Ontology pathway enrichment was analyzed in preranked gene set enrichment analysis (GSEA) with default parameters, and heatmaps were graphed using the gplots package in R program v4.0.3. The TCGA-PANCAN gene expression profiles were acquired using the University of California Santa Cruz Xena database (https://xena.ucsc.edu/).

TCR-seq

CD8 T cells were isolated from the irradiated and abscopal tumors in mice bearing bilateral subcutaneous MC38-OVA tumors 7 days after ipsilateral 8-Gy or sham irradiation +/− anti-SIRPα and anti-PD-1 using cell sorters (BD FACSAria/Fusion Flow Cytometer; fig.S23). The genomic DNA of tumor-infiltrating CD8 T cells was extracted using DNeasy Blood & Tissue Kits and dissolved in 50 μL of TE buffer. High-throughput sequencing of TCR-β CDR3 regions was performed using the ImmunoSEQ platform at Adaptive Biotechnologies. Non-productive out-of-frame TCR rearrangements were omitted in the analyses. CD8 T cell count was determined by the sum of template counts for all productive rearrangements in each sample, and clonality was calculated as 1 - Peilou’s evenness.

Statistics

IBM-SPSS v.24 and GraphPad-PRISM v.9.0 were used for statistical analyses and graph plotting. Two-tailed Student’s t-tests or one-way analyses of variance (ANOVA) were applied for intergroup comparison, and data were presented as mean ± SEM. The patient samples were compared using paired t-tests. The functionality and TCR characteristics of CD8 T cells between irradiated and abscopal tumors were analyzed using two-way ANOVA. Categorical variables were compared using Fisher’s exact tests. Overall survival was analyzed using the Kaplan-Meier method, and tumor growth kinetics were evaluated using two-way ANOVA. Significance was expressed as: *P <0.05, **P <0.01, ***P <0.001, ****P <0.0001.