CD8+ T Cell Densities and Stromal PD-L1 but not Tumoral PD-L1 Associated with Favorable Prognosis in Schistosomal Colorectal Cancer

Backgroud: It was known that the expression of programmed cell death-ligand 1 (PD-L1) was correlated with CD8+ T cells, which could produce IFNγ. The effect of infection of Schistosoma japonicum on CD8+ T cells and then on PD-L1 expression was unknown and the utility of CD8+ TILs as a biomarker for schistosomal CRC (SCRC) has rarely been reported. Methods: A total of 338 patients with CRC were enrolled. Immunohistochemical analysis was conducted to evaluate the expression of PD-L1, inltration by CD8+ T cells. Results: In the total cohort, results showed that CD8 + TIL density was positively correlated with tumoral and stromal PD-L1 expression (p(cid:0)0.05). But there were no correlation between schistosomiasis and CD8+ TILs and PD-L1. Furthermore, CD8 + TIL density, schistosomiasis, TNM stage, lymph nodes positive for CRC and gender were independent predictive factors for overall survival (OS) (p(cid:0)0.05). Stromal PD-L1(sPD-L1) but not tumoral PD-L1(tPD-L1) expression was correlated with OS but it was not an independent predictor (p=0.046). In patients without schistosomiasis, CD8+T cells and sPD-L1 were associated with better OS (p(cid:0)0.05). In patients with schistosomiasis, CD8+T cells were independent prognosis factor (p=0.05). Conclusions: The study showed that CD8+ TILs was an independent predictive factor for OS in CRC and SCRC patients. The expression of PD-L1 was positively associated with CD8+TILs density. There were no correlation between schistosomiasis and CD8+TILs and PD-L1. Stromal PD-L1 but not tPD-L1 was signicantly associated with OS, whereas it was not an independent prognostic factor. response levels patients who have PD-L1-positive TILs blocking of PD-1/PD-L1 pathway.


Introduction
Colorectal cancer is one of the most common malignant disease worldwide. Although a variety of anticancer drugs have been developed, the death rates of CRC have not been obviously decreased 1,2 . Antibody blockade of the programmed death-1 receptor/programmed death-ligand 1(PD-1/PD-L1) signaling pathway have been used for the treatment of malignant cancers, showing promising outcomes 3,4 . Expression of PD-L1 in intratumoral compartment has been suggested to in uence immune response 5 . Binding of PD-L1 to its receptors PD-1 and B7.1 suppresses T cell migration, proliferation and secretion of cytotoxic mediators, ultimately restricting tumour killing by diminishing effector T cell functioning 6, 7 , showing immunosuppressive effect. However, PD-L1 is not solely considered as a result of an increased immune inhibiting PD/PD-L1 interplay but rather is viewed as a re ection of adaptive antitumor immunity, where tumor-in ltrating lymphocytes are activated in response to tumor antigens 8 . Thus, the role of PD-L1 in tumor microenvironment remains obscure. Recent study showed that tumoral PD-L1 is a favorable prognostic factor in early stage non-small cell carcinoma 9 . It was also reported that there were differences in outcome in triplenegative breast cancer depending on the expression of PD-L1 in the tumour cell membrane, cytoplasm and stromal cellular compartments 10 . Yaqi Li et al reported that tumoral PD-L1 correlates with better prognosis of CRC patients 11 . Whereas some studies found that PD-L1 was associated with deleterious effect on survival 12,13 , but these studied did not distinguish PD-L1 expression in tumoral or stromal cells. Therefore, PD-L1 expression used as a predictor factor is also controversial.
On the other hand, studies reported that CD8 + TIL induces PD-L1 expression in tumour cells by producing IFNγ 14,15,16 . CD8 + T cells are thought to have antitumor functions during tumor development in a tumor microenvironment. Evidence has shown that activated CD8 + cytotoxic T lymphocytes were correlated with favorable survival of CRC patients and gastric cancer patients 17,18,19,20 . Therefore, further detailed analysis is needed to con rm the prognostic signi cance of PD-L1 and CD8 + TILs in CRC and to investigate the relationship between PD-L1 and CD8 + T cells.
Intriguingly, we discovered schistosome eggs in almost 38% of cases in HE-stained slides under the microscope. Qingpu District of Shanghai in China was one of endemic areas. Schistosomiasis, which is an infectious disease 21 , is considered a risk factor for CRC 22 . Schistosomiasis is correlated with in ammation [23][24][25] . CD8 + TILs are the main force involved in in ammatory response. In addition, PD-L1 was involved in immune microenvironment and upregulated by CD8 + TILs. With these considerations, we wonder to investigate the relationship between schistosomiasis and CD8 + TILs and PD-L1.
In short, this study aimed primarily to investigate the effect of schistosoma infection on CD8 + TILs and PD-L1 expression and the relationship between schistosomiasis and CD8 + TILs and PD-L1 expression. Besides, we proposed to further to evaluate the prognostic role of PD-L1 and CD8 + TILs in CRC, especially in schistosomal CRC.

Methods And Materials
Patients This retrospective analysis includes 338 patients with resected primary CRC at Qingpu Branch of Zhongshan Hospital a liated to Fudan University, from January 2008 to August 2016. All of the operations followed the principle described previously 26 . Brie y, adequate resection margins, all circumferential margins were cleared.
The number of positive lymph nodes and total number of retrieved lymph nodes were recorded. The inpatient medical records and pathological reports were reviewed from the pathological system and Qingpu District Center for Disease Control and Prevention, and the patients were followed up by telephone. OS is de ned as the interval from the surgical operation date to the last follow-up or death caused by CRC. Inclusion criteria included the following: (i) patients with CRC as primary focus, (ii) none of these patients had received any prior anti-tumor therapy, and (iii) patients were diagnosed as adenocarcinoma by pathology after resection of CRC. Exclusion criteria included the following: (i) Tis tumors, (ii) patients who lacked complete information, (iii) patients with synchronous malignancy, and (iv) patients with survival time less than 1 month. Two expert pathologists reviewed HE-stained slides to determine the diagnosis and to restage the tumors according to the eighth edition of American Joint Committee on Cancer (AJCC). This study is approved by the medical ethics committee of Fudan University, in accordance with the Helsinki Declaration of 1975. Prior written informed consent was obtained from all patients.

Tissue microarrays(TMA)
The TMA blocks were manufactured from the most representative areas of individual para n blocks, as previously described 27 . Brie y, reviewed HE-stained slides and marked the represented areas in tumor tissues, and the single core (2 mm wide and 6 mm long) for each case was precisely arrayed in to a new recipient para n block. The cores containing more than 20% tumor cells were considered as valid cores.

Immunohistochemical (IHC)
All the tissue slides were stained by the fully automated Bond-system (Leica Microsystems, Newcastleupon-Tyne, UK) according to the manufacturer`s instructions. The following primary antibodies were used: PD-L1 (MXR003; 1:750; MXB Biotechnologies, Fuzhou, China) and CD8 (clone NCL-L-CD8-4B11; 1 : 100; DAKO, Minneapolis, MN, USA). In brief, IHC labeling was performed on 4-mm-thick unstained sections. Slides were depara nized with serial xylene treatments and subjected to antigen retrieval using onboard heatinduced antigen retrieval with epitope retrieval solution (pH 9.0) at 100°C for 10 minutes, and incubated with the antibody for 30 min at room temperature. This automated system used a Re ne polymer detection kit with horseradish peroxidase-polymer as a secondary antibody and DAB, and incubation with a secondary antibody was performed for 30 min at room temperature.
Pathological assessment of PD-L1 expression and CD8 + T cell density PD-L1 IHC was analyzed independently by two experienced pathologists, who were unware of the clinical data. The results were evaluated according to the percentage of the stained cells. Scoring was assessed in both the tumoral and stromal compartments: tumoral membranous and stromal immune cell membranous compartments. Tumors were classi ed as PD-L1 positive if there was ≥ 1% tumoral membranous PD-L1 expression (tPD-L1 + ), or ≥ 1% stromal PD-L1 expression (sPD-L1 + ).
The TMA slides were scanned using a scanner system (PRECICE 500B) at 40×magni cation. For CD8, the densities of positively stained cells were evaluated on whole section slides using an image analysis system (Image J software, USA ) (cells per square millimeter) (Fig. 1C). At least half of the core area was selected randomly, and the results of the calculated densities were extracted and put into an Excel le. Measurements were recorded as the mean number of positive cells per tissue unit in square millimetres as well as the number of positive cells among each 1-mm 2 tissue units.

Detection of schistosome ova and assessment of tumor budding
Schistosome ova were observed in all of original HE stained formalin-xed para n-embedded (FFPE) sections (usually 4-6 slides), which were examined at × 10 and × 40 magni cation elds using a conventional light microscope by two pathologists who were blinded to the clinical data. The diagnosis of schistosomiasis was done by nding schistosome eggs in HE-stained slides.
Tumor budding was de ned as the presence of dedifferentiated single cells or small clusters of up to 5 cells at the invasive front of CRC 28 . The assessment of tumor budding was conducted as described previously 26,29 . Brie y, the 10-HPF method was used 30 , the invasive front is rst scanned at low magni cation (×4 to ×10) to identify areas of highest budding density. Tumor buds are then counted under high magni cation (×40), and the tumor budding count is reported. The evaluation of tumor budding was conducted by two pathologists who were blinded to the clinical data. Five tumor budding counts were used as breakthrough point. In brief, tumor bud counts greater than or equal to 5 were de ned as the high group, otherwise as the low group.

Statistical analysis.
Data were analyzed using SPSS (version 20.0; IBM Corp.) and Graphpad 5.0. Every variable was analyzed using univariate analysis to identify all potentially important predictors and then variables with P ≤ 0.05 in the univariate analysis were included in a multivariate analysis. Clinically relevant variables that may have impacted outcomes, such as age, gender, TNM stage, lymph node metastasis, histological type and so on. Finally, multivariate Cox regression analysis was performed to identify predictive factors for OS. P < 0.05 was considered to indicate a statistically signi cant difference.

Patients characteristics
The clinical characteristics of the 338 patients are shown in Table 1. The median age of the patients at diagnosis was 67 years (range, 33-91 years). Two hundred and fourteen (61%) of the patients were male. By anatomic site, 27% tumors were in the rectum, 33% in the left colon, and 40% in the right colon. Lymph node metastasis was observed in 41% of patients, and 46% of patients were at late-stage disease. According to AJCC Staging Manual (seventh edition), there were very few highly differentiated cases in the follow-up data.
Seventy-six percent cases were well/moderate differentiated, and 24% were poorly differentiated. Vessel invasion was observed in about 36% patients, but only 1.2% patients was lymph nodes positive. Intriguingly, schistosoma infection was observed in 38% (128 out of 338) CRC patients( Supplementary Fig. 1).

Staining Results of Each Marker
It has been reported that PD-L1 on either tumor cells or host immune cells contributes to tumor escape, and the relative contributions of PD-L1 on these cells seem to be context-dependent 31 . Thus, the immunostaining for PD-L1 was observed in the membrane of the tumor cells and stromal lymphocytes. Figure 1 shows representative PD-L1-stained images on both tumor cells and tumor-in ltrating mononuclear cells. Among 338 cases analyzed, 41% of cases showed tumoral PD-L1 expression (tPD-L1 + : de ned as ≥ 1%), and 64% showed PD-L1 expression within the immune stroma (sPD-L1 + : de ned as ≥ 1%) ( Table 1 and Fig. 1A-B). The median of CD8 + density was 405 cell/mm 2 (range, 0-2466 cell/mm 2 ) ( Table 1 and Fig. 1C).

Relationship between Schistosomiasis and CD8 + TIL Density and PD-L1 Expression
Previous studies indicated that schistosomasis was a risk factor for CRC 26, 32, 33 . And chronic schistosomal infection was related with in ammation. Thus, we further made subgroups according to whether the patient has schistosomiasis: patients with schistosomiasis (SCRC) and without schistosomiasis (NSCRC). As shown in Fig. 2A, there were no signi cant correlation between CD8 + TILs density and schistosomiasis (p 0.05).
We next compared the density of CD8 + T cells in CRC with or without expression of PD-L1 on the tumour cells or in the immune stroma, respectively. As shown in Fig. 2B and Table 2, CD8 + T cell density was signi cantly higher within sPD-L1 + group than sPD-L1 − group (sPD-L1 − group versus sPD-L1 + group, median 347 versus 460 cell/mm 2 , p < 0.0001), and it was also obviously higher within the tPD-L1 + group than in the tPD-L1 − group (tPD-L1 + group versus tPD-L1 − group, median 371 versus 454 cell/mm 2 , p = 0.0102).

Correlation between PD-L1 Expression and Patient Characteristics
The relationships between tPD-L1 and sPD-L1 expression with clinicopathologic features are detailed in Table 3. A ROC analysis showed that the optimal cut-off value of tPD-L1 and sPD-L1 were both 2% for CRC OS. One hundred and thirty-eight patients (41%) and 142 (42%) were placed in the tPD-L1 high (expression level ≥ 2%) and sPD-L1 high group (expression level ≥ 2%) based on the optimum cutoff point, respectively. Stromal PD-L1 positivity were signi cantly associated with less aggressive tumor features, including early pathological T stage (p 0.001), absence of lymph node metastasis (p = 0.031), absence of lymph nodes positive for CRC (p = 0.012), early TNM Stage (p = 0.034), less tumor budding(p = 0.039) and less bowel perforation (p 0.001). Meanwhile, tumoral PD-L1 positivity were signi cantly associated with early TNM Stage (p = 0.020), but there were no correlation between tPD-L1 expression and other clinicopathological factors (Table 2). It was suggested that higher PD-L1 expression lever both in tumoral cells and stromal cells were along with less aggressive features of tumors.
With regard to CD8 + T cell, the optimum cutoff value of CD8 + T cell density were determined by X-tile program, which were 279 cell/mm 2 ( Supplementary Fig. 2). Patients were divided into 2 groups for further analysis (CD8 low 279 and CD8 high ≥ 279 cell/mm 2 ). Tumours with higher CD8 + T cell density had better OS compared with tumours with lower CD8 + T cell densities (p < 0.0001, respectively, Fig. 3C).
The univariate Cox regression model indicated that age, gender, pathological T stage, lymph node metastasis, TNM stages, tumor differentiation, vessel invasion, lymph nodes positive for CRC, tumor budding, Schistosomiasis, CD8 + T cells and sPD-L1 were signi cantly associated with CRC OS (p < 0.05, Table 3), whereas none of the other factors examined were signi cantly associated with OS. Multivariate analysis after adjustment indicated that gender, TNM stage, lymph nodes positive for CRC, Schistosomiasis, CD8 + T cells were independent prognostic factors for OS of CRC patients (p < 0.05, Table 3), but sPD-L1 was not an independent predictor.

Survival analysis based on subgroups
Kaplan-Meier analysis demonstrated that merely sPD-L1 expression level was associated with favorable OS in the NSCRC group (p = 0.0040) (Fig. 4A), sPD-L1 expression level in the SCRC group and tPD-L1 in the both groups were not correlated with OS (p 0.05) (Fig. 4B, C and D). In the NSCRC set, the univariate Cox regression model revealed that gender, TNM stage, pathological T stage, lymph node metastasis, tumor differentiation, tumor budding, vessel invasion, lymph nodes positive for CRC, sPD-L1 expression level and CD8 + T cells density were associated with OS (p < 0.05) ( Table 4), and the multivariate Cox regression analysis showed that gender, lymph node metastasis, vessel invasion, lymph nodes positive for CRC, sPD-L1 expression level and CD8 + T cells density were independent prognosis factors (p 0.05) ( Table 4). In the SCRC set, the univariate analysis demonstrated that lymph node metastasis, TNM stage, tumor differentiation, lymph nodes positive for CRC and CD8 + T cell density were associated with OS (p < 0.05), and multivariate analysis results showed that only TNM stage, lymph nodes positive for CRC and CD8 + T cell density were independent factors for OS (p < 0.05). -:Data is non-signi cant ; Abbreviation: CI = con dence interval; HR = Hazard ratio; P < 0.05 was de ned as the criterion for variable deletion when performing backward stepwise selection

Discussion
Various tumor entities with elevated immune response, including MSI-H CRC, have dense CD8 pos T-cell in ltrates in common, which are responsible for a local production of interferon gamma (IFNγ) 34,35 . IFNγ, in turn, provokes the adaptive upregulation of PD-L1 on nearby tumor cells via NFκB 36 , thereby mediating a negative feedback mechanism that ultimately leads to T-cell exhaustion in tumor-in ltrating lymphocytes.
Emerging data in other tumor types suggest that negative immune checkpoint proteins are usually upregulated in tumor tissues with a "T cell in amed phenotype" and that in ltration of tumours by effector T cells is necessary to drive upregulation of immune checkpoints 37 . Our results showed that PD-L1 expression in tumoral cells and stromal cells were positively correlated with CD8 + TILs density.
In this study, the expression of PD-L1 in tumour cells and immune stroma were associated with less aggressive tumor features and translated into favorable OS in patients with CRC cancer. however, only the association with immune stroma cell expression was statistically signi cant. Consistent with our ndings, J Wyss et al showed that merely stromal PD-L1 were associated with less aggressive features of colon cancer and with better OS in colon cancer, although they had excluded rectal cancer patients given disease`s different tumor biology, treatment, and prognosis 8 . The association of PD-L1 expression with bene cial clinical outcome has been reported in a diverse set of tumour types, such as NSCLC 38 , melanoma 39 , breast cancer 40,41 and including CRC 42 . This might seem inconsistent with the immunosuppressive function of PD-L1. However, this might be explained that PD-L1 expression within tumor microenvironment is not only as a immunosuppression factor, but rather acts as a re ection of adaptive antitumor immunity, where tumorin ltrating lymphocytes are activated in response to tumor antigens. Contrary to our ndings, Thompson et al 43 showed that in patients with locally advanced gastric cancer, both tumoral and stromal PD-L1 expression and CD8 + TILs were associated with unfavorable outcome. These opposite results might explained by the interaction between tumor and tumor-associated stroma and TILs might be different among different tumor types.
Our results showed that CD8 density was also associated with a good OS and it was also an independent predictor for CRC patients. CD8, which is predominantly expressed on cytotoxic T cells, is a crucial component of the cellular immune system and pivotal for cell-mediated anti-tumor immune response 44,45 .
Previous studies of association between CD8 and prognosis have reported that patients whose tumors contained in ltrating CD8 + TIL showed better survival in non-small cell lung cancer (NSCLC) 46-53 . These results further suggest that PD-L1 expression may re ects an association with a TIL-mediated antitumour in ammatory response, rather than always being associated with tumour immune evasion 54 . Anti-PD-L1 antibody MPDL3280A elicited a response in patients with tumours expressing high levels of PD-L1 and tumor-in ltrating immune cells 55 , suggesting that patients who have PD-L1-positive tumors with CD8 + TILs might achieve a better outcome through blocking of PD-1/PD-L1 pathway. These results further con rmed that CD8 plays crucial role in the immune microenvironment, and the association of CD8 + TIL density with PD-L1 expression may be more important than PD-L1 expression alone predicting survival. Unexpectedly, there were no correlation between CD8 + TILs and PD-L1 and schistosomiasis. It was possible that the patients in the cohort with schistosomiasis are obviously older than patients without schistosomiasis. And the vigour of immunity of older people is weak 56 . In order to con rm this speculation, we excluded patients younger than 60 years old, then to analyze the relationship between schistosomiasis and CD8 + TILs. However, the small percentage of SCRC patients did not allow us to perform further analysis strati ed by age. Thus, further work in larger cohort are still needed to investigate the impact of s. japonicum on CD8 + TILs density and PD-L1 expression.
Our retrospective study had several limitations. First, we do recognized the limitation of utilizing a TMA approach to assess expression of a biomarker that may only be locally present in samples, raising the possibility of false negatives, which could possibly change the signi cance of PD-L1 expression in CRC. Second, we speculated that IFNγ which secreted by CD8 + T cells upregulated the expression of PD-L1. However, further studies needed to clarify the association between PD-L1 expression and CD8 + TILs, and to determine whether this combination has predictive relevance as a biomarker for selecting individual patients for treatment involving PD-1/PD-L1 blockade or for selection of certain tumour types for development. Third, determination of PD-L1 expression in tumour samples was generally performed by immunohistochemistry using various antibodies. Fifth, the threshold for positivity was not formally assessed.
In conclusion, results in present study demonstrated that stomal PD-L1 expression, but not tumoral PD-L1 expression in the whole cohort and in the NSCRC set were associated with less aggressive tumor feature and translated into better OS. And the expression of PD-L1 was positively associated with CD8 + TILs density. Consent for publication Written informed consent was obtained from each participant.
Availability of data and materials: The datasets used and/or analyzed during the current study are available from the corresponding authors on reasonable request.
Con ict of interest: The authors declare that they have no competing interests. A. The relationship between schistosomial infection and CD8+ TILs density. B. PD-L1 expression by both immune stroma (p= 0.0001) and tumour cells (p= 0.0102) increased with increasing CD8 density in each location; Correlation between CD8 density and PD-L1 expression by location were examine using the unpaired t test. The optimum cutoff value of CD8+T cell density were determined by X-tile program, which were 279 cell/mm2. CD8low group was de ned as CD8+ T cells density 279, and CD8high group was de ned as CD8+ T cells density≥ 279 cell/mm2.