Prognostic value of PD-L1 expression on immune cells or tumor cells for locally advanced esophageal squamous cell carcinoma in patients treated with neoadjuvant chemoradiotherapy

Programmed death-ligand 1 (PD-L1) expression may influence the prognosis of patients with localized esophageal cancer. The current study compared the prognostic value of PD-L1 expression between tumor cells and immune cells. Archival esophageal tumor tissue samples were collected from patients who received paclitaxel and cisplatin-based neoadjuvant chemoradiotherapy (CRT) for locally advanced esophageal squamous cell carcinoma (ESCC) in three prospective phase II trials. PD-L1 expression on tumor and immune cells was examined immunohistochemically by using the SP142 antibody and scored by two independent pathologists. The association of PD-L1 expression with patient’s outcomes was analyzed using a log-rank test and Cox regression multivariate analysis. A total of 100 patients were included. PD-L1 expression on tumor cells was positive (≥ 1%, TC-positive) in 55 patients; PD-L1 expression on immune cells was high (≥ 5%, IC-high) in 30 patients. TC-positive status was associated with poor overall survival (OS) (HR: 1.63, P = 0.035), whereas IC-high status was associated with improved OS (HR: 0.44, P = 0.0024). Multivariate analysis revealed that TC-positive, IC-high, and performance status were independent prognostic factors for progression-free survival and that IC-high and performance status were independent factors for OS. Furthermore, the combination of IC-high and TC-negative status was associated with the optimal OS, whereas that of TC-positive and IC-low status was associated with the worst OS. PD-L1 expression on tumor and immune cells may have different prognostic value for patients with locally advanced ESCC receiving neoadjuvant CRT. A combination of these two indexes may further improve the prognostic prediction.


Introduction
Esophageal squamous cell carcinoma (ESCC) is a prevalent malignant disease in East and Central Asia. At the time of diagnosis, most patients present with locally advanced disease, which can be treated with multimodal therapy with curative intent. However, despite aggressive multimodal therapy, such as neoadjuvant chemoradiotherapy (CRT) plus radical esophagectomy, the long-term survival rate of patients with ESCC is only 30-40% (van Hagen et al. 2012;Huang et al. 2015). Postesophagectomy pathological staging is essential for prognosis (Guo et al. 2015;Oppedijk et al. 2014). Researchers have been investigating prognostic biomarkers based on pretreatment ESCC tumor tissues, but are yet to define them.
Programmed death-ligand 1 (PD-L1) and its receptor program death protein-1 (PD-1) represent an immune checkpoint pathway that regulates T cell activation and function. On binding with PD-L1, PD-1 expressed on T cells transduces inhibitory signaling to help maintain the homeostasis of the adaptive immune system. Dysregulation of PD-L1 and PD-1 signaling occurs in multiple human diseases, including cancers. Since 2014, blockade of the PD-L1/PD-1 immune checkpoint has become an important modality of cancer therapy. As of the end of 2020, multiple anti-PD-1/PD-L1 inhibitors have been approved for the treatment of more than 20 indications across cancers, including advanced ESCC (Kato et al. 2019;Kojima et al. 2020).
PD-L1 expression has been widely investigated as a predictive biomarker to help identify cancer patients who are likely to benefit from anti-PD-L1/PD-1 immune checkpoint blockade (Song et al. 2019;Aguiar et al. 2016). However, PD-L1 expression on immune cells and tumor cells may have distinct biological and clinical significance because of multiple and variable mechanisms underlying PD-L1 expression. The PD-L1 expression on immune cells and tumor cells is believed to be mainly driven by interferon-gamma secreted by infiltrating T cells following activation. However, PD-L1 expression on tumor cells can also be driven by oncogenes and may contribute to cancer invasiveness (Chen et al. 2014), metastasis, antiapoptosis (Azuma et al. 2008;Chen et al. 2016a, b), resistance to chemotherapy and radiation (Black et al. 2016;Chen et al. 2016a, b), epithelial-mesenchymal transition (Alsuliman et al. 2015;Kim et al. 2016), cancer stemness (Almozyan et al. 2017), antiautophagy (Clark et al. 2017), and aerobic glycolysis .
The prognostic value of PD-L1 expression in locally advanced ESCC has been reported (Guo et al. 2018), but the results are inconsistent. Some studies have reported that PD-L1 expression on tumor cells has poor prognostic value (Lim et al. 2016), and others have reported that PD-L1 expression on immune cells has good prognostic value (Hatogai et al. 2016;Zhang et al. 2017). Thus, for patients with ESCC, the difference in the prognostic value of PD-L1 expression between tumor cells and immune cells remains unclear.
In this study, we investigated PD-L1 expression on tumor cells and immune cells obtained from pretreatment esophageal tumor tissues derived from patients with locally advanced ESCC receiving neoadjuvant CRT. We determined the distinct prognostic value of PD-L1 expression on tumor cells and immune cells, and we constructed a combined index to improve prognostic prediction.

Patients
This study included patients with locally advanced ESCC from three prospective clinical trials conducted in our institute from 2000 to 2015 (supplementary Table 1). The study design, protocol treatment, and results have been published before (Lin et al. 2007;Huang et al. 2018). The three trials had similar inclusion and exclusion criteria. In brief, patients with locally advanced ESCC, defined as T3 or N + according to the American Joint Committee on Cancer (AJCC) staging system (6th or 7th edition), with adequate liver, renal, and bone marrow functions and with adequate performance status, were enrolled. All patients received paclitaxel and cisplatin-based neoadjuvant CRT with a total radiation of 40 Gy over 20 fractions. Esophagectomy was performed 4-6 weeks after completing neoadjuvant CRT. Patients who did not undergo esophagectomy received a second course of CRT to an accumulated radiation dose of 66 Gy. Patients with ESCC with archival esophageal tumor tissues available for analysis were retrospectively enrolled in this study, and their clinical stages were reassigned according to the AJCC (7th edition). Other clinical characteristics recorded for analysis included age, sex, ECOG performance status, location of primary esophageal tumor, radical esophagectomy received or not, pathological complete response (pCR) or not, progression-free survival (PFS), and overall survival (OS).

Immunohistochemistry and analysis
All tissues analyzed were obtained by endoscopic biopsy before the start of the treatment. After reviewing the corresponding hematoxylin-eosin stains, sections of formalin-fixed paraffin-embedded tissues were retrieved from the Department of Pathology of our institute. All tissue slides were confirmed to contain suitable tumor lesions representing more than 50% of the tissues based on the hematoxylin-eosin staining. The formalin-fixed paraffinembedded tissue slides were deparaffinized and rehydrated. Antigen retrieval was performed using Tris-EDTA buffer (pH 9.0) in a pressure cooker. Dual endogenous enzyme block (DakoCytomation, Glostrup, Denmark) was used before incubation with primary antibody against PD-L1 (dilution 1:100; clone SP142; Ventana, Arizona, USA) overnight at 4 °C. Incubation was then performed with secondary antibody (Dako real envision HRP rabbit/mouse) for 30 min, followed by the application of diaminobenzidine and hematoxylin. PD-L1 expression on immune and tumor cells was independently scored by two pathologists (Liang and Li) according to a published scoring system, and any discrepancy was resolved through consensus (Fehrenbacher et al. 2016). PD-L1 expression on tumor cells was scored as TC0 (< 1%), TC1 (≥ 1% and < 5%), TC2 (≥ 5% and < 50%), and TC3 (≥ 50%). PD-L1 expression on immune cells was scored as IC0 (< 1%), IC1 (≥ 1% and < 5%), IC2 (≥ 5% and < 10%), and IC3 (≥ 10%). PD-L1 expression on tumor cells was categorized as positive (TC-positive: TC 1-3) or negative (TC-negative: TC0). PD-L1 expression on immune cells was categorized as high (IC-high: IC2/3) or low (IC-low: IC0/1). The optimal cutoff values were determined by survival curves of subgroups of TC and IC ( Figure S1).

Outcomes and definitions
The study outcomes were pCR, PFS, and OS. pCR was defined as the total clearance of cancer cells in the esophagectomy specimens, including dissected esophagus and lymph nodes, after neoadjuvant CRT. PFS was defined as the time from the start of treatment to the time of the first evidence of tumor progression, recurrence, or death, whichever occurred first. OS was defined as the time from the start of treatment to the time of the patient's death.

Statistical analysis
The pCR rates of different groups were compared using a Chi-square test. The survival curves were estimated using the Kaplan-Meier method and compared among patients with different levels of PD-L1 expression by using a logrank test. Cox univariate and stepwise regression models were used for the multivariate analysis.

Study population
Data from 100 patients with locally advanced ESCC were included. The other 93 patients were not eligible because the archival tissues were not available or were inadequate for analysis. The median patient age was 56 years, and most of the patients (89%) were men. Their baseline characteristics are presented in Table 1. The baseline characteristics did not differ significantly between the three cohorts (Table S2). In total, 70 patients underwent esophagectomy following   Fig. 2. PD-L1 expression on tumor cells was TC0, TC1, TC2, and TC3 in 45, 24, 27, and 4 patients, respectively. PD-L1 expression on immune cells was IC0, IC1, IC2, and IC3 in 27, 43, 24, and 6 patients, respectively. No significant correlation was observed between the PD-L1 expressions on tumor cells and immune cells (χ 2 = 1.34, P = 0.51; Table 2). PD-L1 expression on tumor cells or immune cells was not correlated with age, sex, clinical stage, tumor location, or performance status (Table S3).

PD-L1 expression and pCR
Among the 70 patients who underwent esophagectomy, 29 (41%) achieved pCR. The pCR rate was higher in patients with IC-high than in those with IC-low status (57 vs 35%), but the difference was not significant. The pCR rates in patients with TC-negative and TC-positive status were similar (42 vs 41%) (Table 3).

Combined predictor of PD-L1 expression on immune cells and tumor cells
The analyses indicated that PD-L1 expression may have different prognostic effects between tumor and immune cells. Therefore, we combined the PD-L1 expression on tumor and immune cells to improve predictive performance. As presented in Fig. 5, we divided the patients into four subgroups according to their PD-L1 expression on tumor and immune cells. We found that patients with TC-negative and IC-high status had the optimal OS (median: 130 months) and those with TC-positive and IC-low status had the worst OS (median: 15 months; P = 0.0037; Fig. 5).

Discussion
This retrospective study demonstrated the prognostic value of PD-L1 expression on immune cells and tumor cells for patients with locally advanced ESCC who received paclitaxel and cisplatin-based neoadjuvant CRT. Our study demonstrated opposite effects of PD-L1 expression on prognosis between tumor and immune cells: TC-positive status was associated with inferior survivals, and IC-high status was associated with improved survivals. Studies examining the prognostic value of PD-L1 on tumor and immune cells for locally advanced ESCC (Table 6)  High PD-L1 expression on immune cells was associated with good survival in two studies (Hatogai et al. 2016;Zhang et al. 2017). In contrast to previous reports, we included patients with locally advanced ESCC treated with Several recent meta-analyses have indicated the differential prognostic value of PD-L1 expression between tumor and immune cells. A 2017 meta-analysis of 60 clinical studies with 10,310 patients and 15 cancer types reported that high expression of PD-L1 on tumor cells is generally associated with poor prognosis . Another meta-analysis including 2877 patients with ESCC from six studies reported similar results (Guo et al. 2018). However, another meta-analysis of 18 studies enrolling 3674 patients with 12 cancer types evaluated the prognostic value of PD-L1 expression on tumor-infiltrating immune cells (Zhao et al. 2017), and the authors concluded that higher PD-L1 expression on tumor-infiltrating immune cells is related to a lower risk of death. The differential prognostic significance of PD-L1 expression between tumor and immune cells may be attributed to their distinct underlying biologies. PD-L1 expression on tumor cells can be driven by their intrinsic oncogene activation and can increase tumor aggressiveness, drug resistance, and metastasis. These mechanisms may explain the poor prognostic impact of PD-1 expression on tumor cells. However, PD-L1 expression on tumor-infiltrating immune cells likely reflects an existing adaptive anticancer immunity in the tumor microenvironment. Several studies have demonstrated that PD-L1 on immune cells, especially macrophages and other myeloid cells, contributes to immune evasion and the response to PD-1/PD-L1 blockade immunotherapy Noguchi et al. 2017;Lau et al. 2017;Lin et al. 2018).
We demonstrated that combining the PD-L1 expressions on tumor and immune cells may improve the predictive performance. Specifically, patients with TC-negative and IC-high status had the highest OS (median: 130 months) and those with TC-positive and IC-low status had the lowest OS (median: 15 months). If validated, this index may help categorize patients with locally advanced ESCC to receive individualized therapies.
This study has several limitations. First, our sample size was smaller than those of other studies; however, our cohort had relatively homogeneous clinical and treatment characteristics compared to other studies. Our patients were prospectively enrolled in three clinical trials in a single institute. All patients were scheduled to receive neoadjuvant CRT, which was uniform among the cohorts and in line with current standards. However, there is slight heterogeneity of treatment among the three clinical trials. The phase II trial "Phase II Study of Metabolic Response to One-Cycle Chemotherapy in Patients with Locally Advanced Esophageal Squamous Cell Carcinoma" had one cycle of induction chemotherapy before neoadjuvant CRT with the study aim to evaluate the predictive value of metabolic response to the single cycle of chemotherapy. But, adding one cycle of chemotherapy to neoadjuvant CRT did not improve the pCR rate or outcomes of the patients (Huang et al. 2018). Second, we did not use an independent cohort to validate our findings and did not calculate the prognostic sensitivity or specificity of PD-L1 expression on tumor and immune cells. Third, we used the PD-L1 antibody clone SP142, which is not a standard practice for ESCC. Currently, clones 28-8 and 22C3 are used for ESCC treated with nivolumab and pembrolizumab, respectively. However, other clones of PD-L1 antibodies have not been compared for patients with ESCC.
In conclusion, our results revealed that PD-L1 expression on immune cells is a favorable prognostic factor for patients with locally advanced ESCC who have received paclitaxel and cisplatin-based neoadjuvant CRT; moreover, PD-L1 expression on tumor cells is a poor prognostic factor. Combining PD-L1 expressions on tumor and immune cells