CD8+/FOXP3+ ratio and PD-L1 expression associated with survival in pT3N0M0 stage esophageal squamous cell cancer

Data describing relationships between the tumor immune microenvironment and patient outcome are limited for esophageal squamous cell cancer (ESCC). The present study investigated the prognostic values of programmed death-ligand 1 (PD-L1) expression and CD8+ or forkhead box protein 3+ (FOXP3+) tumor-infiltrating lymphocytes (TILs) in 133 pathological T3N0M0 stage ESCC patients who underwent radical resection without neoadjuvant or adjuvant therapy. CD8+ and FOXP3+ TIL densities as well as PD-L1 levels in tumor cells and lymphocytes, were assessed through immunohistochemical staining. Patient survival was not associated with CD8+ or FOXP3+ TILs alone, but PD-L1 expression and the CD8+/FOXP3+ ratio were independent predictors of both disease-free and overall survival. PD-L1 expression correlated with age (p = 0.029), tumor length (p < 0.001), tumor differentiation status (p = 0.002) and reduced intratumoral CD8+ TIL density (p < 0.001). Our results suggest pT3N0M0 ESCC clinical outcomes correlate with CD8+ and FOXP3+ TIL densities and PD-L1 levels. Moreover, an intrinsic mechanism for induction of PD-L1 overexpression may be occurring during early tumor oncogenesis. This information may be useful for stratifying patients and guide the application of checkpoint blockade therapy in ESCC.


INTRODUCTION
Esophageal squamous cell carcinoma (ESCC) remains the predominant type of esophageal cancer (EC) in the "EC Belt", which extends from China via central Asia to northern Iran [1].Despite significant advances in screening, diagnosis and treatment modalities, ESCC patient prognosis remains poor.Clinical outcome for patients with middle range ESCC (generally stage IB, II, and IIIA) can vary significantly, with some achieving long-term survival, while others die following disease recurrence [2,3].Novel ESCC-specific molecular targets are therefore needed to improve prognosis prediction and therapeutic interventions.
Patients with pathological T3N0M0 stage (pT3N0M0) disease represent a considerable fraction of middle-stage patients.Although they have the same TNM stage, they are divided into stages IB, IIA, and IIB based on tumor location and histologic grade in the 7th edition of the AJCC manual [4].However, the classic AJCC/UICC-TNM system only describes tumor burden on a macro scale, and does not provide detailed information on the microenvironment in which ESCC proliferates.Considerable research has focused on the host immune response against cancer and illuminated the prognostic importance of tumor-infiltrating lymphocytes (TILs) in many tumor types [5].However, discussions of TILs with respect to ESCC have been limited and somewhat inconsistent.While one study observed improved prognosis in patients with abundant CD8+ TILs, another found no association between patient survival and CD8+ TILs in residual tumor or scar tissues [6,7].Inconsistent results concerning regulatory T cells (Tregs), defined in this study as forkhead box protein 3 (FOXP3)+ T cells (FOXP3+ TILs), have also been reported in ESCC [6,8].Still, multiple studies associated TILs with clinicopathological factors such as survival, response to neoadjuvant therapy and even tumor stage in
As an immune checkpoint, the programmed death-1 (PD-1)/programmed death-ligand 1 (PD-L1) axis facilitates T-cell exhaustion and enables tumors to avoid immunosurveillance [10].PD-L1-PD-1 pathway blockade improves overall survival (OS) in non-small-cell lung cancer (NSCLC), urothelial carcinoma, melanoma, and renal cell carcinoma patients [11][12][13][14].Thus far, no convincing data has been presented regarding the efficacy of anti-immune checkpoint agents in ESCC.Taube et al. proposed stratifying tumors according to PD-L1 status, and TIL presence or absence in the microenvironment might predict patient response to anti-PD-1 therapy [15].However, interactions between PD-L1 and different TIL subpopulations in ESCC, as well as associated clinical outcomes, are under-studied.In the present study, we investigated CD8+ or FOXP3+ TIL (Treg) densities and PD-L1 expression in a cohort of 133 resected pT3N0M0 ESCC patients without neoadjuvant and adjuvant therapy.We speculate on potential mechanisms for PD-L1 overexpression and assess the impacts of CD8+ and/ or FOXP3+ TIL densities and PD-L1 levels on patient survival.

DISCUSSION
Our study assessed the ESCC tumor microenvironment with respect to tumor cells and TILs in 133 pT3N0M0 patients who had received radical surgery.We demonstrated a strong negative correlation between PD-L1 expression and CD8+ TILs, and correlated PD-L1 expression and CD8+/FOXP3+ TILs ratio with DFS and OS.A patient subgroup with both negative PD-L1 expression and a high CD8+/FOXP3+ TILs ratio tended to show improved survival and reduced recurrence.Previous studies of the ESCC tumor microenvironment have tended to evaluate TILs and tumor PD-L1 expression separately, without regard to combined prognostic values.
A higher CD8+/FOXP3+ TILs ratio has been related to favorable prognosis in aggressive breast cancer, ovarian cancer and osteosarcoma [16][17][18].Similarly, we observed that the prognostic value of this ratio is better than that of FOXP3+ or CD8+ TILs alone, which partly reflects reciprocal interactions between promotive CD8+ TILs and repressive FOXP3+ TILs (Tregs) in tumors [19,20].Thus, assessment of both FOXP3+ and CD8+ TIL densities may provide more detailed information about patient immune state.
The use of different antibody clones for IHC may explain recent contradictory reports associating PD-L1 expression with solid tumor patient prognoses.In our study, we used one of four antibody clones approved by KPS, Karnofsky performance score; PD-L1, programmed death-ligand 1. the US Food and Drug Administration (FDA) for PD-L1 IHC detection (SP142) and validated our results via qRT-PCR in 10 tissue samples [21].We assessed PD-L1 expression in both tumor cells and stromal lymphocytes and found that PD-L1 was correlated with postoperative recurrence and poor patient survival.This is in agreement with previous studies that only evaluated PD-L1 expression in tumor cells [22][23][24].Therefore, these clinical data provide rationale for inhibiting PD-L1 in ESCC patients.In our study, poorly differentiated tumors longer than 4 cm displayed stronger PD-L1 expression than well and moderately differentiated tumors of shorter length.Consistent with our results, other studies have reported a correlation between PD-L1 expression and higher tumor grade in various cancers, although the functions of PD-L1 in tumor differentiation and growth were not addressed [25][26][27][28][29].It might be related to the slightly worse hypoxia conditions in longer tumors, as PD-L1 is reportedly induced under hypoxia conditions [30,31].Additionally, upstream molecules involved in PD-L1 expression may also promote cellular proliferation [27,32].Another possible explanation is that activation of the PD-1/PD-L1 pathway may promote tumor immune escape, allowing tumor cells to proliferate and spread more rapidly [28].Molecular mechanisms underlying immune avoidance appear to be distinct among tumors.To date, two phenotypes have been identified: T cell-inflamed (also known as adaptive immune resistance) and non-T cellinflamed (also known as innate immune resistance) [33].The two phenotypes are differentiated by whether PD-L1 upregulation is driven by constitutive oncogenic signaling pathways or induced in reaction to inflammatory signals that are produced when antitumor immune responses are activated [33].Our study observed a moderate negative association between PD-L1 and CD8+ TILs, which suggested that intrinsic pathways might prevail over inflammatory signals, drive PD-L1 overexpression and block tumor immunity in early ESCC.Evidence suggests that patient response to single-agent PD-1 blockade would be less effective in the setting of innate immune resistance as compared to adaptive immune resistance [34,35].Further studies are required to elucidate PD-L1 overexpression mechanism(s) within the tumor microenvironment and the relevance of such mechanisms to ESCC patient outcome following PD-1 blockade.
Our study had several limitations.First, while pT3N0M0 stage ESCC patient optimal management strategies are still controversial, surgery alone is not standard treatment for such patients.We analyzed tumor original immune states, which are not affected by chemotherapy and radiotherapy.However, preoperative chemoradiotherapy may alter the tumor immune microenvironment, inducing T cell influx and PD-L1 upregulation, and this requires further study [23,36,37].Second, we did not analyze TILs and PD-L1 expression in different subareas of the tumor microenvironment, such as invasive margin, cancer cell nests or peritumoral sites.Third, while we examined five tumor cores per patient from tissue regions that were considered representative, we acknowledge that tissue microarrays have limitations for analysis of tumor heterogeneity.Further studies are warranted to harmonize and standardize testing for CD8, FOXP3 and PD-L1 expression by IHC in larger samples.
Within the context of these limitations, our study identified an inverse relationship between CD8+ TILs and PD-L1 expression, suggesting that an intrinsic induction-type mechanism is active in ESCC.Our findings also suggest that differences in TIL density and PD-L1 expression may partly explain the diversity of clinical outcomes observed in patients with similar ESCC stages.These findings may be suggestive of the optimal treatment choice and checkpoint blockade therapy in patients with pT3N0M0 ESCC.

Patients and specimens
The study cohort included 133 patients with pathological T3N0M0 stage thoracic ESCC who had radical esophagectomies in Shandong Cancer Hospital and Institute since June 2005.No patient exhibited evidence of distant metastases in preoperative examination, none had received any prior anticancer treatments or postoperative adjuvant therapy, and all underwent complete macroscopic tumor removal.All tumor tissues were confirmed as ESCC by hematoxylin and eosin (H&E) staining after surgical resection.Only patients who survived more than three mo post-surgery were included in the study.Tumor stage was determined according to the American Joint Committee on Cancer (AJCC, 2009) TNM (tumor, node, metastases) staging system [4].Each tumor was catergorized histologically according to WHO classification criteria.Follow-ups were completed December 2015.The median follow-up time was 42.6 mo.Overall survival (OS) was defined as the time between surgery and death or last known follow-up.The study design was approved by the Ethics Committee of Shandong Cancer Hospital, and all patients signed informed consent agreements.

Immunohistochemistry
IHC analyses were performed using the streptavidinbiotin-peroxidase method with postoperative specimens from all patients.4-μm thin sections were deparaffinized with xylene, rehydrated with graded alcohol solutions, and then exposed to the antigen retrieval system under high pressure for two min.Sections were stained with anti-CD8 (clone SP16, Beijing Zhongshan Golden Bridge Biotechnology Company, Beijing, China), anti-FOXP3 (clone 236A/E7, ab20034, Abcam, USA) and anti-PD-L1 (clone SP142, Beijing Zhongshan Golden Bridge Biotechnology Company) in a humidified chamber at 37°C for 60 min.Slides were incubated with a secondary anti-rabbit and anti-mouse antibodies (Beijing Zhongshan Golden Bridge Biotechnology Company) at 37°C for 15 min.Diamino-benzidine (DAB) was then added to visualize CD8, FOXP3 and PD-L1 staining and slides were counterstained with hematoxylin.

CD8, FOXP3 and PD-L1 evaluation and quantification
Quantitative evaluations of CD8+ or FOXP3+ TILs were performed by examining at least five nonoverlapping high-power fields (40× objective and 10× eyepiece) with the most abundant TILs in each stained section.PD-L1 expression was considered positive if distinct membranous or cytoplasmic staining was observed in tumor or stromal cells.PD-L1 staining intensity was recorded as follows: negative for no staining, positive for membranous or cytoplasmic staining in tumor cells or lymphocytes.Sections were evaluated independently by two experienced pathologists blinded to clinicopathologic information.Variations of > 5% were reassessed and a consensus was reached.As other cell populations can stain with CD8 and FOXP3 antibodies, CD8+ or FOXP3+ cells with apparent different morphological appearances were excluded from cell counts.FOXP3+ and CD8+ TILs were counted both in the cancer cell nest and tumor stroma.The CD8+/FOXP3+ TILs ratio was defined as the number of CD8+ TILs divided by the number of FOXP3+ TILs (Tregs).The mean numbers of CD8+ TILs and FOXP3+ TILs per field and the ratio of CD8+/FOXP3+ TILs were calculated.ESCCs were classified into high and low groups for CD8+ TILs, FOXP3+ TILs and the CD8+/ FOXP3+ TILs ratio using a cut-off value defined as the median number of infiltrating cells per field, as described previously (Figure 1) [7,8].The results are shown as the mean ± SE number of cells in one field (Table S1).