FLASH X-ray spares intestinal crypts from pyroptosis initiated by cGAS-STING activation upon radioimmunotherapy

As most RT practices are delivered using Mega-Volte (MV) X-rays produced by Linacs, an ideal candidate for the preclinical study would be a precisely tunable MV X-ray FLASH irradiation source. But until recently, most studies into FLASH radiation with low linear energy transfer have used electron beams (18, 25–27), presenting a limitation of clinical translation as the low tissue penetration. As a result, 6-MV high-energy X-ray with ultrahigh dose rate that was more attractive in clinical application than FLASH electrons was undertaken in the present study. The pulse structure of the FLASH X-ray is shown in SI Appendix, Fig. S1A, which revealed a full width at half maximum (FWHM) of the continuous electron micropulse of ∼5 ps. Nevertheless, the FWHM of the X-ray micropulse was around 10 ps due to broadening in the targeting procedure, resulting in an instantaneous dose rate of ∼1.85 × 10⁵ Gy/s of the FLASH X-ray irradiation applied in this study. A 50-mm–thick lead secondary collimator was mounted close upstream of the sample (SI Appendix, Fig. S1B), while a diamond detector was installed at the entrance of the collimator to monitor the X-ray (SI Appendix, Fig. S1C). A monitor film was placed on the abdominal surface of the irradiated mouse (SI Appendix, Fig. S1 D and E). During FLASH irradiation, the delivered mean dose rate (110 to 120 Gy/s) depended on the beam current and source–surface distance (SSD) (SI Appendix, Fig. S1F), while the total given dose was set by regulating the width of the macropulse (SI Appendix, Fig. S1A) at different dose rates. The corresponding FLASH (SI Appendix, Fig. S1G) or CONV (SI Appendix, Fig. S1H) dose distribution of an EBT3 film mounted at 8-mm depth in solid water that simulated the center depth of the mouse abdomen illustrated the horizontal (SI Appendix, Fig. S1I) and vertical (SI Appendix, Fig. S1J) dose profiles of WAI, indicating that the central dose of FLASH IR was 15% higher than the mean dose (the blue curve). The electron beam was monitored in real time by a fast current transformer installed upstream of the target (SI Appendix, Fig. S1F), and a typical beam current curve in a 115-ms macropulse is shown in SI Appendix, Fig. S1K, which is consistent with the amplitude curve of the diamond detector (SI Appendix, Fig. S1L) in trend and length. The good linearity between dose (given by ionization chamber) and monitor units (given by diamond monitor) (SI Appendix, Fig. S1M), as well as the relationship between dose rates measured by the ionization chamber and diamond monitor (SI Appendix, Fig. S1N), confirmed the reliability of the dosimetry system, ensuring the accurate dose delivery of FLASH X-ray irradiation in this study.

PD-L1, also known as CD274 or B7 homolog 1 (B7-H1), is a critical immune checkpoint that mainly suppresses adaptive immune response through the interaction with its receptor PD-1 on the membrane of CD8⁺ T cells (28), which maintains systemic immune tolerance in vivo. Therefore, blocking the function of either PD-L1 or PD-1 could significantly activate the antitumor immune response while aggravating the immune-related adverse events in normal tissues (29), among which dermatological toxicity and gastrointestinal disorders occur most frequently (30). Coincidentally, radiation-induced gastrointestinal toxicity is also a prevailing concern in patients receiving abdominopelvic radiation therapy, with more than 60% of patients experiencing symptoms of abdominal pain, diarrhea, or even intestinal obstruction during RT (31). As a result, mice were irradiated in the whole abdomen in the present study to demonstrate the acute and chronic intestinal radiation damage upon PD-1/PD-L1 blockade.

It is worth noting that PD-L1 was mainly expressed in mouse intestinal crypts (SI Appendix, Fig. S2 A–C) and was evoked by 13-Gy CONV (∼0.03 Gy/s) WAI at 2 to 5 d post-IR (SI Appendix, Fig. S2D), indicating a potential function in negative immune regulation within the intestinal crypts. To abrogate the PD-1/PD-L1 signaling, PD-L1 knockout (KO) mice were studied and the lack of PD-L1 expression was determined by Western blotting analysis of various tissues derived from PD-L1 KO mice and their wild-type (WT) C57BL/6 cohorts (Fig. 1A). Fluorescein isothiocyanate (FITC)–dextran assay indicated that after 13-Gy abdominal CONV X-ray irradiation, the FITC levels in the peripheral blood of PD-L1 KO mice were notably higher than those in the WT counterparts (Fig. 1B), suggesting a much more increased gut permeability induced by PD-L1 checkpoint blockade during CONV irradiation. Surprisingly, FLASH WAI at the same dose notably ameliorated the integrity of mouse intestinal epithelium. As shown in Fig. 1 C and D, histopathological analysis of the proximal small intestines confirmed that classical pathogenic patterns such as severe crypt loss and grievous epithelial atrophy in PD-L1 KO mice exposed to CONV WAI were dramatically relieved in mice irradiated by 13-Gy FLASH X-ray. Furthermore, although 13-Gy FLASH X-ray caused a comparable survival fraction with CONV WAI in WT mice, which was similar to that in a recent study based on proton FLASH abdominal irradiation (32), FLASH WAI notably ameliorated mouse survival (83.3%) in PD-L1 KO mice when compared with CONV irradiation (44%) (Fig. 1E). Interestingly, obvious hair depigmentation delimited to the radiation field was observed in all surviving mice at 6 mo post-IR regardless of dose rates (SI Appendix, Fig. S2E).

Using the same FLASH X-ray device undertaken in the present study, Zhu et al. (33) revealed significantly diminished proinflammatory cytokines (e.g., TNF-α and IL-6) in the peripheral blood of mice being irradiated by FLASH abdominal radiation at the late stage of IR exposure. To further investigate the long-term consequences of FLASH WAI, Masson staining analysis was performed to detect the intestinal fibrosis of irradiated mice. As shown in Fig. 1 F and G, surviving PD-L1 KO mice exhibited higher levels of chronic intestinal fibrosis than WT animals 6 mo after 13-Gy CONV WAI. As expected, FLASH WAI considerably reduced the chronic intestinal inflammation in PD-L1 KO mice (Fig. 1 F and G). Together, these findings suggested that FLASH X-ray drastically spared PD-L1–deficient mice from 13-Gy WAI-evoked severe acute enteritis and chronic intestinal fibrosis.

A recent study based on single-cell RNAseq analysis of isolated mouse crypt cells revealed increased crypt-associated immune cells after CONV WAI (34). By reanalyzing the data of the aforementioned study, it was demonstrated that CD8⁺ T lymphocytes were remarkably increased in number (SI Appendix, Fig. S3A) and proportion (SI Appendix, Fig. S3B) after 12-Gy WAI. To confirm the T cell infiltration in the intestinal crypts induced by IR exposure, the percentage of CD8⁺ T lymphocytes in the mouse crypts was examined by flow cytometer. As shown in SI Appendix, Fig. S3 C and D, the proportion of CD8⁺ T lymphocytes was ∼5% in the intestinal crypts of both PD-L1 WT and KO mice during the physiological state. Under CONV X-ray irradiation, the CTL percentage of intestinal crypts in PD-L1–deficient mice was further increased than in their WT cohorts (from ∼12 to 22.7%) (Fig. 2 A and B). Interestingly, FLASH irradiation notably attenuated the CTL ratio by nearly half in the PD-L1 KO crypts to 11.9% in comparison with CONV WAI (Fig. 2 A and B). To evaluate whether this reduction of CTL percentage after FLASH WAI was attributed to the maintaining of crypt architecture and epithelial cell number by attenuating the direct radiation damage or was attributed to the decrease of T cell recruitment, an in vitro study based on three-dimensional cultured organoids of intestinal crypts was performed to examine the differences of direct intestinal damage upon FLASH and CONV X-ray irradiation. And the dose distribution was shown in the form of profiles (SI Appendix, Fig. S3 E and F). As shown in Fig. 2 C and D, the survival rates of crypt organoids subjected to equal doses of CONV or FLASH irradiation were almost the same, suggesting a much more excessive infiltration of CTLs into crypts under CONV WAI than FLASH X-ray irradiation in PD-L1 KO mice.

CD8⁺ T lymphocytes exert the cell-killing function primarily by secreting perforin and serine proteases such as GzmB. Recent data have indicated that CTL-derived GzmB plays a critical role in driving pyroptotic death in tumor cells by cleaving the newly discovered pyroptosis executive protein GSDME (22) to evoke membrane pore formation and membrane permeabilization (24). Therefore, the expression levels of cleaved GSDME were examined to ascertain the pyroptosis in the intestinal crypts of mice being irradiated by different dose rates. Indeed, IR-induced GSDME cleavage was further enhanced by PD-L1 abrogation at 72 to 120 h after CONV WAI, which was accompanied by a robust GzmB expression and Caspase-3 activation in the intestinal crypts (Fig. 2 E and F), since GzmB could also cleave Caspase-3. Administration of FTY-720, a sphingosine-1-phosphate receptor agonist that blocked the egress of T cells from lymphoid tissues (35, 36), dramatically abrogated CTL infiltration (SI Appendix, Fig. S3 G and H) and GSDME cleavage (Fig. 2 G and H) within intestinal crypts 72 h after CONV WAI, indicating that GSDME-mediated pyroptotic cell death was mainly regulated by excessively infiltrated CTLs. Of note, the expression levels of GSDME-NT in the PD-L1 KO crypts were drastically decreased by FLASH X-ray irradiation (Fig. 2 I and J), indicating a reduced intestinal pyroptosis in FLASH-treated animals. Proinflammatory cell death such as pyroptosis always elicits further immune cell recruitment by releasing damage signals, so-called damage-associated molecular patterns (DAMPs), into the extracellular milieu (37). To clarify whether GSDME-induced pyroptosis could, in turn, recruit more CTLs and thus evoke excessive T cell accumulation within intestinal crypts, GSDME was knocked down by short hairpin RNA (shRNA) coated by adeno-associated virus (AAV) (Fig. 2 K–N), and the T cell proportion was examined. Notably, GSDME knockdown considerably restrained the exceeding CTL accumulation within PD-L1 KO crypts at 3 d after CONV WAI (Fig. 2 O and P), indicating a necessity of GSDME-mediated pyroptosis for excessive CTL recruitment. The histopathological analysis further confirmed a reverse of radiation enteritis in GSDME-abrogated mice being irradiated by CONV WAI (Fig. 2 Q and R). Altogether, our data suggested that FLASH X-ray protected PD-L1 KO mice from GSDME-mediated pyroptosis, in comparison with CONV irradiation, which further alleviated the detrimental CTL infiltration and thus impeded the inflammatory feedback driven by ICD.

To further explore the differential molecular drivers of T cell infiltration initiated by radiation modalities with different dose rates, we next sought to investigate the signaling pathways involved in the recruitment of CD8⁺ T cells. The migratory behavior and recruitment of leukocytes are mainly governed by chemotactic cytokines. In 2019, the cooperation between CCL5 and CXCL9 was reported to be critical for T cell engraftment and immune attack in various tumors such as breast, colon, and lung cancers and gynecological solid tumors (38). Several recent studies also investigated whether chemokines of CCL5, CXCL9, CXCL10, and CXCL11 were strongly associated with CD8⁺ T cell infiltration in pancreatic ductal adenocarcinoma, as well as head and neck squamous cell carcinoma (39, 40). However, whether these chemokines orchestrated the infiltration of T cells in the irradiated crypts was not well understood. In this study, the messenger RNA (mRNA) expression levels of Ccl5, Cxcl9, Cxcl10, and Cxcl11 in mouse intestinal crypts irradiated by CONV and FLASH irradiation were examined by quantitative real-time PCR (qRT-PCR). As shown in Fig. 3A, chemokines in the intestinal crypts had been significantly up-regulated by CONV WAI since 24 h post-IR and were further increased the next day. Importantly, FLASH WAI considerably attenuated the expression levels of these chemokines to nearly half of those in CONV X-ray–irradiated crypts at 48 h post-IR (Fig. 3B), indicating a decreased chemotaxis signal of T lymphocytes within intestinal crypts upon this ultrahigh dose rate radiation modality.

It is well known that genes encoding the aforementioned chemokines are up-regulated by the type I IFN-mediated signaling pathway stimulated by the cGAS-STING axis responding to cytoplasmic dsDNA fragments (5, 11, 40). IR-induced chromosomal fragments are recognized by the cytosolic nucleic acid sensor cGAS, facilitating cGAMP synthesis from ATP and GTP. cGAMP then activates STING that phosphorylates TANK-binding kinase 1 (TBK1) and IFN regulatory factor 3 (IRF3) to trigger expression and secretion of type I IFNs, including IFN-β (41). As a result, the activity of the cGAS-STING signaling pathway within mouse crypts was determined by Western blot assay, which revealed notably increased STING, TBK1, and IRF3 phosphorylations at 6 to 9 h after CONV WAI (Fig. 3 C and D). Of note, FLASH X-ray irradiation dramatically diminished the activity of the cGAS-STING signaling pathway (Fig. 3 E and F) and reduced STING-mediated Ifnb1 (the gene that encoded IFN-β) expression that was provoked by CONV WAI 12 h post-IR (Fig. 3 G and H). To further demonstrate the role the STING-IFN signaling pathway played during intestinal pyroptosis induced by WAI with differential dose rates, linked dimeric amidobenzimidazole (diABZI), an enhanced STING agonist (42), was injected intraperitoneally directly after 13-Gy FLASH WAI to activate STING signal artificially (Fig. 3 I and J). Intriguingly, diABZI administration considerably enhanced the CTL proportion within FLASH-irradiated PD-L1–deficient mouse crypts to ∼23.4% (Fig. 3 K and L), which was almost comparable to that in CONV X-ray–irradiated counterparts. Moreover, FLASH IR-induced pyroptotic cell death (Fig. 3 I and J) along with enteritis (Fig. 3 M and N) in PD-L1 KO mice was also exacerbated by STING stimulation.

Furthermore, immunocompetent WT mice being irradiated by fractionated FLASH irradiation (5 Gy at a time for 5 consecutive days, 25 Gy in total, i.e., 5 Gy × 5) combined with anti–PD-L1 antibody administration, a more clinically relevant scenario in radioimmunotherapy, also revealed a similar aggravation of enteritis induced by diABZI treatment (Fig. 3 O and P). Contrarily, RU.521 that impeded cGAS activity notably ameliorated the intestinal architecture of mice being treated with fractionated CONV WAI combined with an anti–PD-L1 antibody (Fig. 3 Q and R). Intriguingly, RU.521 failed to further improve the integrity of the intestinal epithelium of FLASH X-ray–irradiated mice undergoing checkpoint blockade (Fig. 3 Q and R). These findings further confirmed that the relatively inadequate cGAS-STING activity within intestinal crypts exposed to FLASH radiation contributed to sparing mice from intestinal pyroptosis and enteritis upon radioimmunotherapy.

The activity of the cGAS-STING signal was regulated mainly by the content of cytosolic dsDNA. To further elucidate the mechanism of FLASH IR in regulating the cGAS-STING activity, the levels of cytosolic dsDNA were examined using PicoGreen, a dsDNA-specific vital dye. As shown in Fig. 4A, significantly lower levels of dsDNA were present in the cytoplasmic fraction of intestinal organoids exposed to 8-Gy FLASH X-ray than CONV X-ray at the same dose. Given the cytoplasmic dsDNA fragments were reported to be eliminated by 3′ repair exonuclease-1 (TREX-1) (43), the expression levels of TREX-1 within intestinal organoids exposed to WAI were detected in vitro. It was worth noting that the TREX-1 expression levels in FLASH X-ray–irradiated organoids of intestinal crypts were no higher (even lower) than those in CONV IR-irradiated ones (Fig. 4 B and C), indicating that relatively more activated STING signal that presented upon CONV irradiation than FLASH X-ray (Fig. 4 B and C) was due to more grievous primary DNA damage rather than TREX-1 elimination. Surprisingly, the Western blot assay revealed that the expression levels of both γ-H2AX and phosphorylated 53BP1 were almost the same in the intestinal crypt organoids irradiated by FLASH and CONV X-rays (Fig. 4 B and C), indicating that FLASH X-ray evoked as much damage in the genomic DNA as CONV irradiation did. Immunofluorescent assay further verified that 4-Gy FLASH X-ray elicited equivalent 53BP1 foci in the human intestinal epithelial HIEC-6 cells with CONV X-ray (Fig. 4 D and E). Notably, 53BP1 was found to elicit a differential response between FLASH and CONV electrons in MRC5 and IMR90 human lung fibroblast cells by Favaudon and coworkers (18), which was different from our results. In addition to the cell-specific mechanisms of the recruitment of 53BP1 to the DSB foci, the main reason might be that the dose rate of FLASH X-ray applied in our study was much lower than that of FLASH electron beams undertaken by Favaudon and coworkers (18). Specifically, given the single-macropulse mode applied in our study, an actual dose rate in the pulse was only 110 to 120 Gy/s, which was equal to the mean dose rate itself. However, the dose rate in the pulse used in Favaudon and coworkers’ (18) study was more than 1 × 10⁷ Gy/s (18), a dose rate much more likely to spare normal cells from genomic DNA damage by inducing oxygen depletion (25). Similarly, Levy et al. (44) also demonstrated that 14-Gy abdominal FLASH irradiation with a dose rate in the pulse of 4.0 × 10⁵ Gy/s could induce less genomic DNA damage in the intestinal crypt stem cells than CONV electron beams did, taking γ-H2AX foci as endpoints.

On the other hand, the capability of FLASH in sparing mouse intestine from pyroptosis by minimizing IR-induced cGAS-STING activity raised the concern that this might translate to a loss of antitumor efficacy. Therefore, we sought to investigate the phosphorylations of STING, TBK1, and IRF3 within colorectal tumor cells after being irradiated by different radiation modalities. Intriguingly, when compared with CONV X-ray at the same doses, FLASH IR equally enhanced expression levels of phosphorylated proteins within the STING pathway (SI Appendix, Fig. S4 A and B). Meanwhile, 8-Gy FLASH X-ray induced roughly equal amounts of cytoplasmic dsDNA fragments to CONV IR in MC38 tumor cells (SI Appendix, Fig. S4C). Since PicoGreen also stained mitochondrial DNA, MitoTracker was applied simultaneously to indicate the mitochondria. As shown in Fig. 4F, most of the PicoGreen-stained areas in the cytoplasm of MC38 cells were not overlapped with the areas traced by MitoTracker, indicating the accumulation of cytosolic dsDNA induced by CONV or FLASH X-ray IR (Fig. 4G). Altogether, the differences of FLASH X-ray in activating the cGAS-STING signal between tumor cells and normal cells suggested a different mechanism involved in the interaction between FLASH IR and DNA molecules.

Recently, Eggold et al. (45) indicated that abdominopelvic FLASH irradiation could maintain the ability to increase intratumoral CD8⁺ T cell infiltration and enhance the efficacy of anti–PD-1 therapy in ovarian cancers. To investigate whether FLASH radiation was as efficient as CONV radiation in eliciting the abscopal antitumor immune response, MC38 tumor cells were implanted on the left (primary, irradiated) and right (secondary, nonirradiated) flanks of immunocompetent C57BL/6 mice (Fig. 5A). Ten days after primary tumor implantation, mice were randomized to each group based on tumor volumes and were irradiated by X-ray delivered with hypofractionated regimens (5 Gy × 5), either by FLASH X-ray (SI Appendix, Fig. S5A) or by CONV X-ray (SI Appendix, Fig. S5B). The anti–PD-L1 antibody was administered as indicated in Fig. 5A. The two-dimensional profile of radiation fields revealed a relatively even dose distribution within irradiated tumor sites of both CONV and FLASH IR, as well as virtually absent dose delivery at abscopal tumor sites (SI Appendix, Fig. S5 C–H). To investigate whether FLASH RT could augment antitumor immunity equivalent to CONV IR, intratumoral CD8⁺ T cells were evaluated. As shown in SI Appendix, Fig. S6 and Fig. 5B, anti–PD-L1 as a monotherapy hardly elicited T cell infiltration within both primary tumors and secondary tumors, which obtained an extremely slight effect on restraining MC38 tumor growth (Fig. 5 C and D), resulting in 100% of mice eventually succumbing to either excessive tumor volumes or diameters (Fig. 5E). Notably, both CONV and FLASH RT provoked robust CTL infiltration within anti–PD-L1–treated primary tumors (SI Appendix, Fig. S6A and Fig. 5B) accompanied by almost equivalent eradication of primary tumors 3 wk after treatment (Fig. 5C). Although FLASH RT failed to elicit adequate antitumor immunity in all tumor metastases, it still induced the intratumoral T cell response comparably to CONV radiation (SI Appendix, Fig. S6B and Fig. 5B), along with an equal elimination rate of metastatic burdens (Fig. 5D). Moreover, 57.1% of mice treated with the combination regimen were cured and achieved long-term survival regardless of FLASH RT or CONV RT (Fig. 5E), indicating an equal abscopal response and systemic antitumor immunity induced by FLASH and CONV X-ray when integrated with checkpoint blockade.