Prognostic relevance of programmed cell death 1 ligand 2 (PDCD1LG2/PD-L2) in patients with advanced stage colon carcinoma treated with chemotherapy

Colorectal cancer (CRC) is the leading cause of cancer-related mortality worldwide. Although the role of tumor programmed cell death 1 ligand 1 (PD-L1) in suppressing antitumor immunity has been validated in various malignances, the impact of PD-L2 (PD-L2/PDCD1LG2) within tumors remains elusive. Here, we examined tumor PD-L2 expression by immunohistochemical analysis and assessed its association with clinicopathological characteristics and the infiltration of intratumoral T lymphocytes in colon carcinoma patients (n = 1264). We found that tumor PD-L2 status was correlated with perineural invasion (PNI) and associated with survival outcome in colon carcinoma patients. The level of tumor PD-L2 was positively associated with tumor PD-L1 expression but inversely associated with the density of CD8+ tumor-infiltrating lymphocytes (TILs). Patients with elevated tumor PD-L2 levels had a favorable 5-year overall survival (OS) compared to patients with low PD-L2 levels (57% vs 40%, p < 0.001), especially in advanced stage colon carcinoma patients. Low tumor PD-L2 expression was associated with an increased 5-year OS risk among advanced stage colon carcinoma patients by univariate analysis [hazard ratio (HR) = 1.69, 95% CI 1.324–2.161, p < 0.001] and multivariate analysis [HR = 1.594, 95% CI 1.206–2.106, p = 0.001]. Moreover, tumor PD-L2 expression was inversely associated with the lymphocytic reaction in advanced stage colon carcinoma, suggesting that PD-L2 may be upregulated by a compensatory mechanism to inhibit T cell-mediated anticancer immunity. Taken together, these results show that tumor PD-L2 expression may be an independent prognostic factor for survival outcome in patients with advanced stage colon carcinoma.

Colorectal cancer (CRC) is the leading cause of cancer-related mortality worldwide, representing approximately 10% of cancer cases 1 . Despite advances in treatment for CRC patients, a remarkable proportion of CRC patients still develop tumor relapse, such as local recurrence (LR) and distant metastasis (DM), within 5 years after curative surgery 2,3 . Moreover, the survival of advanced-stage CRC is poor even when patients receive adjuvant chemotherapy. Thus, novel therapeutic strategies, such as immunotherapy strategies, need to be developed to improve the survival outcome of CRC patients. The development and progression of CRC are triggered in a stepwise-manner by many genetic mutations in cancer cells as well as host-tumor interactions within the tumor microenvironment (TME) 4,5 . Anti-cancer immunity is indispensable for the surveillance and destruction of tumors; however, accumulating evidence indicates that immune checkpoint mechanisms downregulate anti-cancer immunity and lead to immune evasion within the TME 6 . Recently, antibodies targeting the immune checkpoint protein programmed cell death receptor 1 (PD1) and its ligand programmed cell death 1 ligand 1 (PD-L1/CD274) have clinically improved the survival outcomes in several malignances [7][8][9] . The therapeutic responses to immune checkpoint blockade are influenced by the molecular characteristics, immune cell profiles and tumor PD-L1 expression within the TME 4, [10][11][12][13] . Although the infiltration of T lymphocytes is correlated with improved outcomes in CRC, the degree of T lymphocyte infiltration is determined by tumor antigens and tumor molecular characteristics such as microsatellite instability (MSI) [13][14][15] . Infiltrating T lymphocytes, such as cytotoxic CD8 + T cells, may induce adaptive immunity resistance by upregulating PD-L1 via effects on IFNγ to inhibit anti-tumor immunity 11,16 . However, the function of PD-L2 (PDCD1 ligand 2, PDCD1LG2), the other ligand for PD1, within the TME remains mostly unknown. A few reports have suggested that PD-L2 may be implicated in the induction of immune tolerance under physiological and pathological conditions 17,18 and promote CD8 + T cell-mediated anti-tumor immunity 19 .
In this study, we examined the relationship between tumor PD-L2 expression and clinicopathological characteristics, such as DNA mismatch repair status as well as immune cell profiles, and elucidated the prognostic value of tumor PD-L2 expression in colon carcinomas. We found that patients with elevated tumor PD-L2 levels had favorable 5-year survival outcomes in advanced stage colon carcinoma compared with patients with low tumor PD-L2 expression, which is consistent with the results from The Cancer Genome Atlas (TCGA) 20 , especially in patients who received chemotherapy. Encouragingly, we found that tumor PD-L2 status was inversely correlated with the density of intratumoral CD8 + TILs, suggesting that the level of PD-L2 was not mainly induced by IFN-γ from CD8 + TILs but by other cytokines within the tumor microenvironment. These findings indicate that PD-L2 may be an independent prognostic factor for advanced stage colon carcinoma patients.  (Table 1). The average age of the patients was 63.1 years, with a range of 23 to 93 years. Postoperative pathohistological analysis was performed before enrollment in this study, and patients were classified according to the tumor-node-metastasis staging system (AJCC 7th edition staging). A postoperative chemotherapeutic regimen was recommended for high-risk stage II patients and lymph node metastasis stage III patients identified in surgical specimens according to the status of the patients (Table S1).

Tissue microarray (TMA) construction and immunohistochemistry (IHC). The resected tumor
tissue and corresponding normal mucosa specimens from primary tumors of colon carcinoma patients before chemotherapy were collected to establish a TMA ( Table 1). The protocol for the establishment of TMA has been previously described 13,21 and was approved by the Institutional Review Board (IRB) of China Medical University Hospital [protocol number: CMUH105-REC2-073]. Briefly, tumor areas were evaluated on hematoxylin/eosin (HE)-stained tissue slides by a pathologist (Dr. Tsung-Wei Chen), and the corresponding area on a formalinfixed, paraffin-embedded (FFPE) tissue block (donor block) was identified for TMA construction. Tissue cylinders (2 mm in diameter) were punched from the central tumor tissue areas from each donor block, and these individual tissue cylinders were placed into one recipient paraffin block. Each TMA spot contained at least 50% of the tumor area.
Immunohistochemical staining was performed on 3-μm-thick TMA sections according to the manufacturer's manual (VECTASTAIN Elite ABC Kit, Vector Laboratories, CA, USA), and the TMAs were incubated with DAB (Vector Laboratories) and then counterstained with hematoxylin 22,23 . Positive staining for CD8 (ab4055, Abcam, Cambridge, UK), CD45RO (ab23, Abcam) and PD1 (ab52587, Abcam) was defined as staining in the cytoplasm or membranous intratumoral TILs by microscopy (Olympus BX53, Tokyo, Japan). Evaluation of TILs in TMA tissues was performed at 400 × magnification by a pathologist. The area with the highest density of TILs within the TME was counted at 400 × magnification [No. of positively stained TILs/high-power field (HPF)]. The average count of TILs in five HPFs was scored: a count of 0 positively stained TILs in one HPF was scored as www.nature.com/scientificreports/ 0, 1-3 positively stained TILs/HPF was scored as 1, 4-10 positively stained TILs/HPF was scored as 2, and > 10 positively stained TILs/HPF was scored as 3 13 . Tumor PD-L2 (ab200377, Abcam) expression was evaluated based the extent and intensity of immunopositivity of tumor cells as assessed by a semiquantitative scale (0-3+) as follows: 0 for absent; 1 for weak; 2 for moderate; and 3 for strong membrane staining. The percentage of positive cells was scored as follows: 0-10% positive tumor cell proportion was scored as 0; 10-25% positive tumor cell proportion was scored as 1; 26-50% positive tumor cell proportion was scored as 2; and 51%-100% positive tumor cell proportion was scored as 3. The product of the two scores served as the tumor PD-L2 immunostaining score. The cutoff for the immunostaining score was 4 13,15 . The percentage of cancer cells with membranous PD-L1 (ab205921) staining was scored as follows: 0 points were assigned when < 5% positive tumor cells were detected, and 1 point was assigned when > 5% of cells showed positive membranous staining 24 . To evaluate the status of mismatch repair (MMR), MMR proficiency was defined the expression of 4 proteins: MLH1 (ab92312, Abcam), MSH2 (ab92372, Abcam), MSH6 (ab92471, Abcam) and PMS2 (ab110638, Abcam), while MMR-deficient tumors were defined as those lacking at least one of these four markers by IHC.
The Cancer Genome Atlas (TCGA) database. A total of 259 patients with stage III (n = 174) and IV (n = 85) colorectal cancer were included, and their PD-L2 mRNA expression data were retrieved from the openaccess Human Pathology Atlas resource, which is part of the Human Protein Atlas (HPA, www.prote inatl as.org/ patho logy) 20,25 ; HPA contains RNA sequencing (RNA-seq) data together with clinical information from TCGA. The PD-L2 mRNA level had the best log-rank p value based on the Kaplan-Meier analysis. The average RNA expression level was utilized as the cutoff (high expression cutoff = 0.21) according to the algorithm on the HPA website 20 . ware, CA, USA) were utilized to perform statistical analysis in the study. All tests were analyzed with two-sided p-values with the significance level set at 0.05, including Student's t-test and Pearson's chi-square test. Univariate and multivariate models were analyzed by Cox regression analysis with hazard ratios (HRs) and 95% confidence intervals (CIs) 26 . The influential factors were included in the Cox models, including sex (male vs female), age (≥ 65 years vs < 65 years), pT stage (pT3-4 vs pT1-2), pN stage (positive vs negative), pTNM stage (stage IV vs stage III), lymphovascular invasion (presence vs absence), perineural invasion (presence vs absence), tumor location (proximal colon vs distal colon), CD8 + TIL density (high vs low), CD45 + TIL density (high vs low), CD45RO + TIL density (high vs low), PD1 + TIL density (high vs low), tumor PD-L1 level (high vs low) and tumor PD-L2 level (high vs low). The 5-year overall survival (OS) was estimated by the Kaplan-Meier method and defined as the duration between the day of diagnosis and the day of death. Log-rank tests were performed for univariate comparisons.

Ethical approval. This study was reviewed and approved by the Internal Review Board (IRB) of China
Medical University Hospital [Protocol number: CMUH105-REC2-073]. The method was carried out in accordance with the committee's approved guidelines.
Informed consent. Informed consents were obtained from all participants in the study.

Results
Immunohistochemical expression of tumor PD-L2, clinicopathological characteristics and the status of MMR in stage I-IV colon carcinoma. To evaluate the clinical impact of tumor PD-L2 in stage I-IV colon carcinoma, we used immunohistochemical analysis to detect the protein expression of tumor PD-L2 (n = 1264, Table 1). We scored tumor PD-L2 expression in the cytoplasm and membrane based on the intensity of staining and the proportion of cells with positive staining 13 Table 1.
We found that 30.9% (390/1264) of colon carcinoma patients overall and 7.6% (96/1264) of MMR-deficient patients had high expression PD-L2 in the tumor bed. The expression of tumor PD-L2 within the TME was significantly associated with the absence of perineural invasion (PNI, p < 0.001) and MMR proficiency (p < 0.001, Table 1).  Table 2). However, KM curves and log-rank tests revealed that high expression of tumor PD-L2 was associated with a remarkably longer 5-year OS than low expression of tumor PD-L2 in advanced stage patients (57% vs 40%, p < 0.001, Fig. 1G). We then analyzed PD-L2 mRNA data  Fig. 1H) 20 . Data on PD-L2 mRNA expression in stage III-IV colorectal cancer samples were retrieved from the Human Protein Atlas (HPA, www.prote inatl as.org/patho logy) 20,25 , which employs RNA sequencing (RNA-seq) data together with clinical information from TCGA. The cutoff for the PD-L2 mRNA level for KM analysis was based on the algorithm on the HPA website 20 . CRC patients with high PD-L2 mRNA had favorable survival outcomes. Moreover, the KM analysis of clinicopathologic characteristics in advanced stage colon carcinoma showed that older age (p = 0.02), larger pathological tumor size (pT stage, p < 0.001), lymphovascular invasion (LVI, p < 0.001), PNI (p < 0.001) and proximal colon location (p = 0.021) were associated with shorter 5-year OS in advanced stage colon carcinoma (Table 3). Classifying advanced stage colon carcinoma patients into subgroups based on stage and postoperative chemotherapy regimen, we found that stage III or IV patients with elevated tumor PD-L2 tended to have better 5-year OS than those with low tumor PD-L2 (Fig. S1A-1B and Table S2). Patients with advanced stage colon cancer who received postoperative chemotherapy with high tumor PD-L2 had more favorable 5-year OS than those with low tumor PD-L2 ( Fig. 2A,B). Moreover, high tumor PD-L2 was clinically associated with improved 5-year OS in stage III (p = 0.0492, Fig. 2C) and stage IV (p = 0.0561, Fig. 2E) colon carcinoma patients who received postoperative chemotherapy. Multiverse analysis adjusted with clinicopathologic parameters suggested that tumor PD-L2 may be an independent prognostic factor for advanced stage Table 3. Correlation between clinicopathologic parameters and 5-year OS (n = 633). a Number of cases may differ due to missing data. *indicated p<0.05.  Fig. 2F).
To further clarify the role of TILs and tumor PD-L2 expression in CRC prognosis, we evaluated the association of patient survival outcomes with the density of cytotoxic CD8 + TILs, the density of memory CD45RO + TILs and tumor PD-L2. Patients with high CD8 + TIL density within the TME had improved 5-year OS (log-rank, n = 630, p = 0.0005, Fig. 4A). We considered intratumoral CD8 + TILs and tumor PD-L2 expression in combination, and we found that tumor PD-L2 status was not associated with 5-year OS in the subgroup of patients with high CD8 + TILs (CD8 + TILs H /PD-L2 H : 69.7% vs CD8 + TILs H /PD-L2 L : 57.1%, p = 0.167, Fig. 4B). Intriguingly, the expression of tumor PD-L2 was significantly associated with 5-year OS in patients with low CD8 + TILs (CD8 + TILs L /  Fig. 4C). Similarly, patients with high CD45RO + TIL infiltration within the TME had improved 5-year OS (log-rank, n = 630, p < 0.0001, Fig. 4D). Tumor PD-L2 was not associated with 5-year OS in patients with high CD45RO + TILs (Fig. 4E, p = 0.3785) but was significantly associated with 5-year OS in patients with low CD45RO + TILs (Fig. 4F, p < 0.001). These results suggest that tumor PD-L2 expression can be a significant prognostic factor for advanced stage colon carcinoma patients, especially for patients with low densities of infiltrating lymphocytes.   www.nature.com/scientificreports/ (Table 5). These results showed that immune factors such as CD8 + TILs, CD45 + TILs, CD45RO + TILs, PD1 + TILs and tumor PD-L2 expression have clinical prognostic relevance for advanced stage colon carcinoma patients. Moreover, when these parameters that were identified as significant by the univariate analysis were subjected to multivariate Cox regression analysis as covariates, tumor PD-L2 expression was found to be an independent predictor for 5-year OS in advanced stage colon carcinoma patients (Table 5). Patients with low tumor PD-L2 levels (HR 1.59, 95% CI 1.206-2.106, p = 0.001) presented an increased risk for poor OS after adjustment for sex, age, pT stage, pN stage, pTNM stage, LVI, PNI, tumor location and immune signatures. Moreover, stage III and IV patients who received postoperative chemotherapy with low tumor PD-L2 levels presented an increased risk for poor OS after adjustment for age, pT stage, LVI, PNI and tumor location (Fig. 2D,F). These results show that the level of tumor PD-L2 could independently predict the prognosis of advanced stage CRC, especially patients who received postoperative chemotherapy (Table 5).

Discussion
In the present study, we found that tumor PD-L2 expression was positively associated with the status of tumor PD-L1 in colon carcinoma. However, tumor PD-L2 was inversely associated with intratumoral TIL densities in colon carcinoma, such as CD8 + TIL and PD1 + TIL densities, suggesting a possible role of tumor PD-L2 in suppressing antitumor immune responses in colon carcinoma. Moreover, tumor PD-L2 expression was remarkably associated with 5-year OS in advanced stage colon carcinoma, suggesting that it can be considered an independent prognostic factor for advanced stage colon carcinoma, especially for patients who receive postoperative chemotherapy.
CRCs are often heterogeneous tumors because of the accumulation of somatic mutations and the alternations in host-tumor interactions within the TME. Lymphocyte alterations such as intraepithelial T cell infiltration within the TME, have been associated with favorable survival outcomes in CRC [13][14][15] . Therefore, a combination of molecular characteristics and immune cell infiltration parameters, defined as "consensus molecular subtypes" (CMSs), has been proposed to classify CRCs 28 . Consistent with this classification, our previous studies found that strong positive correlations between PD-L1 and intratumoral CD8 + infiltration, and the infiltration of CD8 + TILs and tumor PD-L1 status were positively associated with clinical outcome in colorectal cancer 12,29 . Interferon-γ (IFN-γ) secreted by CD8 + T lymphocytes is required for PD-L1 upregulation 24 , serving as a www.nature.com/scientificreports/ compensatory feedback mechanism for the adaptive immune response within the TME 30,31 . Similarly, PD-L2 can be induced by the adaptive immune resistance mechanism, depending on the milieu of inflammatory cytokines in the tumor microenvironment; for example, IFN-γ, IL-4 and colony-stimulating factor 2 (CSF-2) upregulate PD-L2 11 . Although we observed a positive association between PD-L2 and PD-L1 expression, we found that tumor PD-L2 expression was inversely associated with the intratumoral infiltration of CD8 + immune cells, suggesting a unique immunologic role of tumor PD-L2 within the TME. Several studies have reported that tumor PD-L2 is inversely correlated with TIL density and lymphocyteic reaction alterations, leading to inhibition of  35 . These results suggest that tumor PD-L2 expression may inhibit anti-cancer immunity against colon carcinoma, implying that tumor PD-L2 might be upregulated by other mechanisms, such as hypoxia. Pinato et al. revealed that PD-L2 expression was upregulated by hypoxia-induced factor 1 (HIF-1) 36 , which contributed to immune resistance within the TME. Therefore, it is possible that PD-L2 may be regulated by HIF-1 to potentially mediate immune resistance by downregulating lymphocytic reactions to avoid immune responses within the TME 17,36,37 . However, whether PD-L2 causes the immune suppression within the TME and its molecular mechanism require further investigation in the future. The association between PD-L2 expression and survival outcome has been investigated in several malignancies, including CRC, melanoma, pancreatic cancer, and esophageal cancer 33,[38][39][40][41] ; however, discrepancies between these studies have been observed. Several studies reported that high PD-L2 expression was associated with poor patient survival, including studies in esophageal cancer 32 , pancreatic ductal adenocarcinoma 42 and colorectal cancer 38,39 , and high tumor PD-L2 expression was associated with a favorable survival outcome in melanoma 40 and colorectal cancer 41 . However, Masugi et al. reported that there was no significant correlation between tumor PD-L2 expression and survival in CRC patients 33 . This discrepancy may be due to the lack of FDA-approved IHC antibodies and differences in the criteria between these studies, which could have led to variations in PD-L2 expression and survival results in CRC. The IHC results in these studies were dramatically different. Previous studies have shown that PD-L1 is expressed on tumor and stromal cells 43 44 . Therefore, standardization of IHC antibodies and evaluation criteria is urgently needed. In the current study, we found that 30.9% of tumor cells and fewer immune cells expressed PD-L2. Based on our IHC results, we found that increased PD-L2 expression positively correlated with better overall survival in advanced stage colon carcinoma, which is consistent with the results from the TCGA database 20,25 . The RNA-seq results together with clinical information showed that PD-L2 mRNA was an independent prognostic factor for colorectal cancer in data from the TCGA database. These results suggest that tumor PD-L2 is an independent prognostic factor for OS in advanced stage colon carcinoma, especially in patients who receive postoperative chemotherapy. However, there are several limitations in our study. First, for the assessment of the immune signatures and tumor PD-L2 status by multivariate logistic regression analyses, we cannot exclude the possibility that the role of chemokines within the TME may regulate the recruitment of immune cells. We did not measure chemokines in this retrospective study, and potential chemokine confounders need to be considered in future studies. Second, based on our results, we speculate that tumor PD-L2 may inhibit intratumoral T cell infiltration. However, we did not elucidate what intrinsic and extrinsic factors contribute to PD-L2 upregulation. Further studies are necessary to address possible molecular mechanisms for tumor PD-L2 signaling, intrinsic factors, extrinsic factors and immune cell infiltration.
In addition, a recent study reported that not only the PD-L1 status but also the PD-L2 status was useful for predicting the clinical response to anti-PD1 immunotherapy in head and neck carcinoma 45 , suggesting that the status of tumor PD-L2 may have clinical implications for anti-PD1 immunotherapy. However, further studies are necessary to investigate the prognostic value of PD-L2 expression. Although both PD-L1 and PD-L2 proteins share the same receptor, PD-1, other receptors for PD-L2 (RGMb) have been reported 46 , suggesting that the role of PD-L2 in suppressing T cell responses may be different. Nonetheless, our results showed that the status of PD-L2 was not associated with MMR deficiency, strongly suggesting that further investigation to clarify the potential molecular mechanism of PD-L2 expression is necessary 47 . Accordingly, further investigations of PD-L2 stratifying patients according to receipt of anti-PD1/PD-L1 immunotherapy are needed.
Taken together, our results showed that tumor PD-L2 expression was inversely associated with the intratumoral infiltration of CD8 + TILs in advanced-stage colon carcinoma, suggesting a possible influence of PD-L2-expressing tumor cells on adaptive antitumor immunity. Moreover, patients with elevated tumor PD-L2 levels had favorable survival outcomes, suggesting that tumor PD-L2 may be an independent prognostic factor for advanced stage colon carcinoma. Since the inflammatory microenvironment in the gastrointestinal tract plays a role in CRC progression via host-immunity interactions 48 , the upregulation of tumor PD-L2 may be due to adaptive immunity, suggesting that PD-L2 may have prognostic value and can be considered a therapeutic target for immunotherapy in colon carcinoma.