To identify the molecular mechanism(s) involved in the response to conventional fractionated irradiation, we set out to compare gene expression profiles in irradiated vs. non-irradiated cells. For this, HT29 colorectal carcinoma cells were subjected to a common clinically applied treatment schedule of daily 2 Gy irradiation, 5 days per week for up to 5 weeks (Fig. 1a). Since our previous work showed that clonogenic survival converges to a steady state after 2 weeks of treatment, i.e. 10 fractions (Fig. 1b), we first compared the expression at that timepoint with the expression in non-irradiated cells, cultured identically for 2 weeks. Gene expression analysis by RNA sequencing identified over a thousand differentially expressed genes (adjusted p-value ≤0.05) in irradiated vs. non-irradiated cells (Fig. 1c and Supplementary Tables 2 + 3). Gene ontology (GO) analysis revealed over 250 significantly enriched upregulated biological processes, amongst which the ‘type I IFN-mediated signaling pathway’ was identified as the most significantly enriched biological process (adjusted p-value ≤0.0001) (Supplementary Fig. S1a and Supplementary Table 4). Other identified GO terms were closely related to biological processes such as positive regulation of cell migration, angiogenesis, negative regulation of cell proliferation, amine metabolism, response to virus and nucleosome assembly. Additionally, over a hundred significantly enriched downregulated biological processes were identified, mainly related to translational processes and cell cycle (Supplementary Fig. S1b). Given the current insights in radiotherapy-induced type I interferon signaling [14, 15], as well as previous (pre) clinical trials on the combination of radiotherapy with type I interferons in cancer [16], we further focused our research on this particular response.
The induction of the type I IFN response in HT29 cells could be confirmed by qPCR with a panel of 10 interferon stimulated genes (ISGs) that are linked to this response (Fig. 1d). Moreover, the increased expression of ISGs by 10 × 2 Gy irradiation could be confirmed in multiple cancer cell lines, including high-grade astrocytoma cells (D384) and different colorectal cancer cell lines (SW480, HCT116, COLO320, RKO; Supplementary Fig. S2). To determine the dynamics of the response, the ISG expression was analyzed weekly for up to 5 weeks. This showed a slight induction in expression for most ISGs after 5 fractions, and a peak induction after 2 to 3 weeks of treatment (Fig. 1e and Supplementary Fig. S3a). Continuation of RTx eventually resulted in a decreased expression, although it generally remained above the level of non-irradiated cells. Of note, while single dose irradiation also induced dose-dependent ISG expression, this typically leveled off after 6 Gy (Supplementary Fig. S3b). Collectively, these data show that fractionated RTx induces an intrinsic type I interferon response in vitro which peaks within 2 to 3 weeks of treatment and coincides with the development of a steady state in clonogenic survival.
To extend these findings, HT29 xenograft tumors were locally irradiated using the same clinical schedule as the cultured cells, i.e., 2 Gy per day, 5 days per week for up to 3 weeks (Fig. 2a). Tumor growth showed a delay after 2 weeks of treatment but appeared to recover in week 3 (Fig. 2b and Supplementary Fig. S4a). Next, gene expression profiles of non-irradiated tumors vs. tumors that received 1, 2 and 3 weeks of radiotherapy were obtained using human microarray analysis. After 2 weeks of treatment, 34 differentially expressed genes in irradiated vs. non-irradiated tumor tissues were identified, of which 5 showed decreased expression and 29 showed increased expression (Fig. 2c, Supplementary Table 5). Gene ontology analysis revealed 52 significantly enriched biological processes, amongst which the ‘type I IFN-mediated signaling pathway’ was again identified as the most significantly enriched pathway (p-value ≤0.0001, count 18/61) (Supplementary Fig. S4b and Supplementary Table 6). Interestingly, a less pronounced but similar gene expression profile was observed after 1 and 3 weeks of irradiation, whereas a single dose of 5 Gy resulted in more differentially expressed genes (Supplementary Fig. S4c). Expression analysis of the same ISG signature panel as used before, again confirmed the induction of a type I IFN response (Supplementary Fig. S4d). Moreover, in line with our observations in the cell lines, time course analysis revealed that the expression of the ISGs peaked after 2 to 3 weeks of treatment (Fig. 2d). Altogether, these results show that fractionated RTx induces a potent type I interferon response in tumor cells after 2 to 3 weeks of treatment.
To determine which type I IFN could have triggered the response, we analyzed the mRNA expression of two key family members in vitro, i.e., IFN alpha (IFN-α) and IFN beta (IFN-β). Since type III interferons (IFN lambda; IFN-λ) were recently shown to be induced by RTx in HT29 [17], these cells were included as a positive control. Analysis of fractionally irradiated HT29 tumor cells revealed that the treatment predominantly induced the mRNA expression and protein secretion of IFN-β and IFN-λ (Fig. 3a+b). Other cell lines subjected to fractionated irradiation displayed either a modest increase in mRNA expression of either IFN-α, IFN-β, IFN-λ or a combination (HCT116 and RKO), or no interferon induction at all (SW480 and Colo320) (Fig. 3c). Interestingly, all of these cell lines showed clear induction of ISG expression in response to irradiation, albeit less profound in the cell lines lacking interferon expression (Supplementary Fig. S2). In the xenograft tumors, no changes in the expression of any of the different interferons could be detected (Fig. 3d+e). These findings suggest an uncoupling between the induction of ISGs and the expression of interferons, the latter usually mediating ISG expression. Of note, all the cell lines expressed the appropriate IFN receptors required to be responsive to the different IFNs (Supplementary Fig. S5).
To assess the clinical relevance of these findings, we analyzed whether a type I IFN response occurs in cancer patients, in the context of a clinical pilot study (NCT02072720) in esophageal cancer patients receiving neoadjuvant chemoradiotherapy (CRT) with paclitaxel, carboplatin and concurrent radiotherapy (41.4 Gy in 23 fractions of 1.8 Gy). Tumor biopsies of 20 patients (see Supplementary Table S7 for patient characteristics) were collected at baseline and after 1, 2, 3 or 4 weeks of treatment, in successive cohorts. Subsequent expression analysis revealed that expression levels of 5 out of 7 investigated ISGs were significantly elevated during treatment as compared to baseline (Fig. 3f). Of note, while the number of patients in this small pilot study did not allow us to confirm an association between pre-treatment ISG expression levels and response to treatment [18, 19], we did observe ISG expression levels were highest in patients that had received 2 weeks of treatment (Supplementary Fig. S6a). The latter is in line with our findings in tumor cells and xenograft tumors. Furthermore, a modest induction of all interferons was seen (significant for IFN-β and IFN−λ2/3; Fig. 4g), but again ISG expression appeared to occur independent of type I interferons, as only a weak correlation was observed between induction of IFN-β and 3 out of 7 ISGs (Supplementary Fig. S6b). Thus, both in a preclinical immunocompromised xenograft model as well as in a clinical setting, commonly applied fractionated irradiation triggers a type I interferon response independent of actual type I interferon expression induction.
Since the induction of a type I interferon response during radiotherapy has been linked to cGAS/STING signaling [14, 20, 21], we further evaluated the role of STING as well as of interferon expression on the induction of ISGs during fractionated irradiation. Analysis of both mRNA and protein expression showed low or even absent basal expression of cGAS, STING or both in the majority of cell lines, except for HT29 (Fig. 4a+b). Since all cell lines did show elevated ISG expression during fractionated irradiation, these findings suggest that the radiation-induced type I interferon response does not depend on cGAS/STING signaling. Of note, when cells that were deficient in either cGAS or STING were irradiated, the expression of the absent proteins was not induced (Supplementary Fig. S7a).
To further study the disconnection between the radiation-induced type I interferon response and cGAS/STING activation or type I interferon expression, we irradiated HT29 cells (which express all components of the pathway) in the presence of either anti-IFN-β antibody, anti-IFN-λ antibody or a STING antagonist. Optimal antibody treatment conditions were based on literature [16] and the levels of IFN-β and IFN-λ in cell culture supernatants. In addition, direct effects of treatment on cell viability were excluded (Supplementary Fig. S7b). Also, the inhibitory function of the STING antagonist was confirmed by Western blot showing reduced phosphorylation of the downstream target protein Tank Binding Kinase (pTBK) after 4 Gy irradiation as compared to no irradiation (Supplementary Fig. S7c). In line with our previous observations, neither treatment with anti-IFN antibodies nor treatment with the STING antagonist had any effect on the induction of ISG or IFN expression during fractionated irradiation (Fig. 4c). Moreover, neither treatment affected the clonogenic survival of HT29 cells prior to irradiation (Supplementary Fig. S7d) or during fractionated irradiation (Fig. 5 d). This was not due to lack of treatment efficacy, since anti-IFNβ antibody treatment did neutralize the known inhibitory effect of IFN-β on cell growth (Supplementary Fig. S7e). Again, these data suggest that the type I IFN response that is triggered by fractionated irradiation occurs independent of cGAS/STING signaling or induction of IFN expression.
Since our findings are different from the previously published role of STING in the response to radiotherapy and could be due to minimal undetected levels of STING, we also generated HT29 STING knockout cells using CRISPR/Cas gene editing. In 8 out of 10 single cell clones, knockdown could be confirmed by Western Blot (Supplementary Fig. S8a). Two clones were selected for further analysis and DNA sequencing confirmed gene editing at the expected location in exon 6, causing a frameshift with two adjacent premature stop codons (Supplementary Fig. S8b). Interestingly, one of the STING knockout clones showed a phenotype similar to the wild type cells while the other knockout clone showed reduced cell growth (data not shown) and increased radiosensitivity (Fig. 4e). This clonal difference in growth and radiosensitivity was further illustrated when both clones were subjected to fractionated irradiation for 3 weeks (Fig. 4f+g). Despite these differences in phenotype, both knockout cell lines showed a similar, albeit delayed, induction of ISG expression as compared to the wild type cells (Fig. 4h). The latter suggests that STING, while it contributes to the induction of a type I IFN response during fractionated irradiation, is not essential for this response to occur. Altogether, STING, IFN-β as well as IFN-λ appear to be dispensable for activation of a type I IFN response during fractionated irradiation.