PD-L1 upregulation is associated with activation of the DNA double-strand break repair pathway in patients with colitic cancer

Ulcerative colitis (UC) is a DNA damage-associated chronic inflammatory disease; the DNA double-strand break (DSB) repair pathway participates in UC-associated dysplasia/colitic cancer carcinogenesis. The DSB/interferon regulatory factor-1 (IRF-1) pathway can induce PD-L1 expression transcriptionally. However, the association of PD-L1/DSB/IRF-1 with sporadic colorectal cancer (SCRC), and UC-associated dysplasia/colitic cancer, remains elusive. Therefore, we investigated the significance of the PD-L1/DSB repair pathway using samples from 17 SCRC and 12 UC patients with rare UC-associated dysplasia/colitic cancer cases by immunohistochemical analysis. We compared PD-L1 expression between patients with SCRC and UC-associated dysplasia/colitic cancer and determined the association between PD-L1 and the CD8+ T-cell/DSB/IRF-1 axis in UC-associated dysplasia/colitic cancer. PD-L1 expression in UC and UC-associated dysplasia/colitic cancer was higher than in normal mucosa or SCRC, and in CD8-positive T lymphocytes in UC-associated dysplasia/colitic cancer than in SCRC. Moreover, PD-L1 upregulation was associated with γH2AX (DSB marker) and IRF-1 upregulation in UC-associated dysplasia/colitic cancer. IRF-1 upregulation was associated with γH2AX upregulation in UC-associated dysplasia/colitic cancer but not in SCRC. Multicolour immunofluorescence staining validated γH2AX/IRF-1/PD-L1 co-expression in colitic cancer tissue sections. Thus, immune cell-induced inflammation might activate the DSB/IRF-1 axis, potentially serving as the primary regulatory mechanism of PD-L1 expression in UC-associated carcinogenesis.

www.nature.com/scientificreports/ patients (P = 0.028; Table 2). However, the pattern of PD-L1 expression did not differ between UC mucosa with or without dysplasia/colitic cancer. The clinicopathological characteristics of UC patients with and without UCassociated dysplasia/colitic cancer are summarised in Supplementary Table 1. There was no significant difference between the expression of PD-L1 and p53/Ki-67, which are useful for the diagnosis of UC-associated dysplasia/ colitic cancer, in UC patients with UC-associated dysplasia/colitic cancer (Supplementary Table 2, Supplementary Fig. 2).
Associations among γH2AX, IRF-1, and PD-L1 expression in colitic cancer. Using multicolour immunofluorescence analysis, we were able to validate the co-expression of γH2AX, IRF-1, and membrane  www.nature.com/scientificreports/ PD-L1 in seven UC patients with colitic cancer and dysplasia. Consequently, cancer cells expressing membrane PD-L1 showed nuclear γH2AX and IRF-1 expression, in contrast to those with low PD-L1 expression, which did not show co-expression of γH2AX, IRF-1, and PD-L1 (Fig. 3). Furthermore, in SCRC tissue sections, γH2AX and IRF-1 were observed to be co-expressed in cells expressing PD-L1 as well as those that did not, suggesting that PD-L1 regulation in SCRC may not be dependent on the DNA damage response/PD-L1 signalling axis ( Supplementary Fig. 3). These findings are consistent with the data presented in Table 3.

The association of MMRD with SCRC and UC-associated dysplasia/colitic cancer. The MMRD
status was not significantly different between SCRC and UC-associated dysplasia/colitic cancer (Table 5, Supplementary Table 4).

Discussion
This study showed that PD-L1 expression levels were significantly upregulated in UC and UC-associated dysplasia/colitic cancer tissues compared with that in SCRC tissues and the corresponding non-cancerous mucosa. Moreover, PD-L1 expression was higher in tumoural CD8-positive T lymphocytes in UC-associated dysplasia/ colitic cancer tissues than that in SCRC tissues. PD-L1 upregulation was found to be associated with increased expression of γH2AX (a DSB marker) and IRF-1 (a PD-L1 inducer) in clinical UC-associated dysplasia and colitic cancer tissues but not in SCRC tissues or the corresponding non-cancerous UC mucosa. Colitic cancer is caused by UC-induced persistent chronic inflammation, and the carcinogenic sequence of colitic cancer is suggested to differ from that of SCRC 1 . Chronic inflammation in UC frequently results in DSB through the generation of reactive oxygen species, with a dysfunctional DNA damage repair system suggested to lead to carcinogenesis during UC [6][7][8] . However, PD-L1 is regulated by various inflammation-associated transcription factors, including STAT1/3, IRF-1, and NF-κB 19,22 . Among these, DSB-induced-IRF-1 activates PD-L1 18 . The present findings showed that PD-L1 expression in UC-associated dysplasia/colitic cancer was positively correlated with γH2AX and IRF-1 expression. These results indicated that PD-L1 expression in UC-associated dysplasia and colitic cancer was potentially associated with the activation of the DSB/IRF-1 signalling axis. Moreover, in cancer cells, PD-L1 is associated with the activation of DNA damage repair kinases such as ATR and ATM 23 . Therefore, DNA damage-induced-PD-L1 in dysplasia/colitic cancer may work not only as a mere biomarker for inflammatory DNA damage but also as a regulator of DNA repairment.
ICIs have recently attracted increasing attention as an innovative cancer treatment strategy 24 . However, the beneficial effects of ICI treatment in SCRC patients are often limited to MMRD/MSI-high cancers 25,26 , and it remains unknown whether ICIs are effective in rare cases of UC with UC-associated dysplasia/colitic cancer. Studies have reported that MMRD/MSI-high status is a promising biomarker for predicting the sensitivity to ICIs 27 , with tumours exhibiting high PD-L1/high CD8+ CTLs designated as "immune-inflamed tumours" or "hot tumours" displaying positive responses compared with ICI-resistant uninflamed tumours or 'cold tumours' without PD-L1 expression/CD8+ CTLs 28 . In contrast, enterocolitis is known as one of the most common adverse events associated with ICIs, and it is distinct with clinical and pathological characteristics similar to that of inflammatory bowel disease (IBD) 29,30 . A recent study showed patients with pre-existing IBD to be at an increased risk for several gastrointestinal adverse events associated with ICIs, and the safety of ICI treatment for patients with pre-existing IBD is undetermined 31 . However, TNF-α blockade treatment in patients with IBD strongly inhibits inflammation in the colorectal mucosa by suppressing several pro-inflammatory pathways, including the TNF-α pathway 32 , and immunosuppressive therapy, including TNF-α blockade and vedolizumab, which is an α4β7 integrin inhibitor blocking the migration of gut-specific lymphocytes into the gut, is administered for treating not only IBD, but also ICI-induced colitis 33,34 . Particularly, vedolizumab has been administered concurrently with ICIs in patients with ICI-induced colitis 35 . Furthermore, TNF-α blockade does not inhibit the antitumour effect of ICIs in experimentally induced melanoma 36 . Perez-Ruiz et al 37 reported the therapeutic effect of TNF-α blockade against ICI induced-colitis in tumour-bearing mouse models. Several studies have  www.nature.com/scientificreports/ Crohn's disease flares and ICI-induced colitis. This study showed that PD-L1 and IRF-1 expression levels, as well as CD8+ CTLs in UC-associated dysplasia/colitic cancer increased significantly compared with that in the SCRC samples, suggesting that UC-associated dysplasia/colitic cancer, rather than SCRC, was potentially associated with the "hot tumour" phenotype. Furthermore, patients with melanoma having high IRF-1 expression levels exhibit a significantly higher sensitivity towards ICI treatment than those with low IRF-1 expression 42 . These results suggested that ICI treatment may be more effective in patients with UC-associated dysplasia/colitic cancer with distant metastases after total colectomy who are not at a risk for ICI-induced colitis than in patients with SCRC, because patients with UC-associated dysplasia/colitic cancer exhibit clinical characteristics including favourable ICI responses, "hot tumour" immune phenotypes, including high PD-L1/high CD8+ CTLs, and high IRF-1 expression. Moreover, the combinatorial treatment with ICIs and immunosuppressive therapy such as TNF-α blockade-vedolizumab may suppress ICI-induced colitis in patients with UC-related dysplasia/colitic cancer and exert antitumour effects. Nonetheless, our study has several limitations. First, this was a retrospective study with a small patient cohort because UC-associated dysplasia and colitic cancer are rare types of cancer; this may have introduced a bias in our results. Hence, prospective studies with a larger patient cohort may be challenging for clearly elucidating the UC carcinogenesis mechanisms, considering the rarity of the disease. Currently, biological data on UC with UCassociate dysplasia/colitic cancer have been obtained from small-scale clinical cohort studies. This study provides further valuable insights into the importance of PD-L1 expression in patients with UC with UC-associated dysplasia/colitic cancer. Second, this study discussed the potential use of ICIs for colitic cancer treatment; however, we only enrolled patients that did not receive ICI treatment targeting the PD-1/PD-L1 signalling axis. Therefore, future studies are required to evaluate the potential of DSB/IRF-1/PD-L1 as a biomarker for rare cases of UCassociated dysplasia/colitic cancer to be treated with ICIs. Lastly, only three target markers were assessed, which may not be sufficient to clarify the complexity of the interactions between PD-L1 and the DSB repair response. Therefore, in the future we plan to examine the importance of and relationship among these three markers in an experimental animal model 43 of inflammatory colorectal carcinogenesis.
In conclusion, this study showed that PD-L1 expression was significantly higher in UC and UC-associated dysplasia/colitic cancer than in the normal mucosa and SCRC. Furthermore, PD-L1 expression was significantly associated with high levels of tumoural CD8+ CTLs, γH2AX as a DSB marker, and IRF-1 as a transcriptional factor for PD-L1 in UC-associated dysplasia/colitic cancer compared with SCRC. Therefore, our findings suggested that immune cell induced-chronic inflammation might activate the DSB/IRF-1 signalling axis, which might serve as the primary regulatory mechanism for regulating PD-L1 expression in UC carcinogenesis.

Methods
Patients and samples. Twelve patients (nine male and three female) with UC, who underwent surgical resection for UC-associated dysplasia/colitic cancer at Gunma University Hospital (Maebashi, Gunma, Japan), Maebashi Red Cross Hospital (Maebashi, Gunma, Japan), and Gunma Prefectural Cancer Center (Ohta, Gunma, Japan) between 1999 and 2014, were included in this retrospective study. The median age of the patients was 54 years (range, 37-76 years). One patient only had UC-associated dysplasia. Three patients harboured two www.nature.com/scientificreports/ or more tumours, and all dysplastic and cancerous lesions were evaluated. All dysplastic lesion samples were obtained from patients with high-grade dysplasia, while patients with low-grade dysplasia were not included in the study. Additionally, 17 SCRC patients (12 male and five female) who underwent partial colectomy at the Gunma University Hospital between 1999 and 2014 were randomly selected and included in the study. Ten control UC patients (seven males and three females) who underwent surgical resection for reasons other than UC-associated dysplasia/colitic cancer at Gunma University Hospital were also included in the analysis. Supplementary Tables 1 and 2 summarise the clinical characteristics of the patients. Supplementary Table 3 shows the characteristics of the clinical course in UC patients with UC-associated dysplasia/colitic cancer. We evaluated the severity of UC as previously reported 44 . For an accurate pathological diagnosis of dysplastic and cancerous lesions among patients with UC, all histological cancer tissue sections were evaluated by a specialised pathologist, Dr. Yao T (Department of Human Pathology, Juntendo University Graduate School of Medicine). This study conformed to the tenets of the Declaration of Helsinki and was approved by the Institutional Review Board for Clinical Research at the Gunma University Hospital, Maebashi, Gunma, Japan (approved number: HS2018-092).
Owing to the retrospective nature of this study, need for informed consent was waived, and participants were given the opportunity to decline participate in this study via opt-out method. All methods were performed in accordance with relevant guidelines and regulations. The slides were then stained using rabbit-specific IHC polymer detection kit HRP/DAB (Abcam, ab209101) containing the secondary antibody for PD-L1-stained sections, and with the Histofine Simple Stain MAX-PO (Multi) Kit (Nichirei, Tokyo, Japan) for other markers, in accordance with the manufacturer's instructions. The chromogen 3,3′-diaminobenzidine tetrahydrochloride was applied as a 0.02% solution, containing 0.005% H 2 O 2 in ammonium acetate-citrate acid buffer (50 mM, pH 6.0). Finally, nuclear counterstaining was performed using Mayer's haematoxylin solution. In the negative control sections, primary antibodies were replaced with phosphate-buffered saline in 0.1% bovine serum albumin, thus confirming a lack of staining.

Immunohistochemistry.
Assessment of immunohistochemistry staining. Immunostained sections were evaluated by two experienced researchers blinded to the clinical data. We evaluated tumour cells and non-cancerous cells displaying membrane PD-L1 staining as positive when at least 1% of the cells were stained 46 . The staining intensity for γH2AX was scored as follows: 0, no staining; 1+, weak staining; 2+, moderate staining; and 3+, strong staining. The percentage of nuclear-stained cells was determined by examining three sections with the highest staining intensity. The percentage of nuclear γH2AX staining was scored as follows: 0, no staining; 1+, 1-25%; 2+, 26-50%; and 3+, 51-100%. The final score was defined as the percentage score multiplied by the intensity score (0, 1+, 2+, 3+, 4+, 6+, and 9+). Nuclear immunoreactivity of γH2AX was scored as 0-4+ and 6-9+, which were defined as low and high nuclear expression, respectively 47 . We evaluated nuclear and cytoplasmic staining for IRF-1 using almost the same method as that for γH2AX described above. The staining percentage for IRF-1 was scored as follows: 0, no staining; 1+, 1-25%; 2+, 26-75%; and 3+, 76-100% 48 . The final score was defined as the percentage score multiplied by the intensity score (0, 1+, 2+, 3+, 4+, 6+, and 9+). IRF-1 immunoreactivity was scored as 0-4+ and 6-9+ , which were defined as low and high expression, respectively. Supplementary Fig. 2 shows the representative low or high expression of γH2AX, IRF-1, and PD-L1 in SCRC and UC-associated dysplasia/colitic cancer tissues. Tumoural CD8-positive CTLs, defined as CD8 cells infiltrating the cancer stroma, were scored as low or high when the mean number of cells in a microscopic field at ×200 magnification was < 200 or ≥ 200, respectively 49 . p53-positive cells were defined as previously reported 50 . The Ki-67 labelling index was scored as a percentage of positively stained cells. MMRD was defined as the complete absence of the expression of at least one mismatch repair protein (MLH1, MSH2, MSH6, or PMS2).
Multicolour immunofluorescence staining for PD-L1, γH2AX, and IRF-1. Multicolour immunofluorescence staining was performed for seven patients with UC on a total of seven colitic cancer and two dysplasia tissue sections, and three SCRC tissue sections using a PerkinElmer Opal kit (Catalogue# NEL810001KT, PerkinElmer, Hopkinton, MA, USA) in accordance with the manufacturer's instructions. γH2AX staining was visualised using the Opal 520 Fluorophore, IRF-1 staining with the Opal 690 Fluorophore, and PD-L1 staining with the Opal 570 Fluorophore. All sections were then counterstained with 4′,6-diamidino-2-phenylindole (DAPI) and examined under an FV10i-LIV system (Olympus Life Science, Tokyo, Japan). www.nature.com/scientificreports/ among PD-L1 expression, immune cells, and the DSB repair pathway. All differences were considered statistically significant at P < 0.05.

Data availability
The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.