Tumor-Infiltrating Immune Cells and PD-L1 as Prognostic Biomarkers in Primary Esophageal Small Cell Carcinoma

Primary esophageal small cell carcinoma (PESCC) is a weakly prevalent but lethal malignancy with early metastasis and a poor prognosis. Currently, neither effective prognostic indicators nor curative therapies are available for PESCC. Immunotherapy has now evolved into one of the most promising therapies for cancer patients. Tumor-infiltrating immune cells which are integral to the tumor immune microenvironment (TIME) are recognized as highly important for prognosis prediction, while the responsiveness to immune checkpoint blockade may be subject to the features of TIME. In this study, we aim to identify the TIME and provide indication for the applicability of immune checkpoint therapy in PESCC. We found that PD-L1 expression was detected in 33.33% (27/81) of all the patients, mostly exhibiting a stroma-only pattern and that it was positively associated with tumor-infiltrating immune cells (CD4+, CD8+, and CD163+). In 74.07% of PD-L1-positive specimens, PD-L1+CD163+ cells were colocalized more with CD4+ than CD8+ T cells. 83.95% (68/81) of all the specimens were infiltrated with more CD4+ than CD8+ T cells. Further analysis showed FoxP3+ Tregs constituted 13-27% of the total CD4+ T cell population. The Kaplan-–Meier analysis indicated several factors that contribute to poor survival, including negative PD-L1 expression, rich CD4 expression, rich FoxP3 expression, a low CD8/CD4 ratio, and a high FoxP3/CD8 ratio. A nomogram model was constructed and showed good performance for survival prediction. These results highlight that a suppressive TIME contributes to poor survival of patients with PESCC. TIME analyses might be a promising approach to evaluate the possibility and effect of immune checkpoint-based immunotherapeutics in PESCC patients.


Introduction
Primary esophageal small cell carcinoma (PESCC) is a rare but fast-growing tumor that exhibits a neuroendocrine phenotype and accounts for 0.5-2.8% of all esophageal malignancies [1]. PESCC is featured with early dissemination and poor clinical outcomes [2,3]. Current treatment options for PESCC are those universal for common malignancies, including surgery, chemotherapy, radiotherapy, and concurrent chemoradiotherapy based on the tumor-nodemetastasis (TNM) staging system [4]. However, the prognosis can vary significantly among patients with the same TNM stage, and there remains no targeted or curative therapy due to a lack of knowledge about the mechanisms underlying the cause and progression of PESCC.
Tumor immune microenvironment (TIME) is an important regulator of the antitumor immune response, since immune factors in the TIME have significant impacts on cancer patient prognosis [3,[5][6][7]. Antibodies against the programmed cell death protein 1 (PD-1) receptor and its ligand PD-L1, among other immune checkpoint inhibitors, have delivered unprecedented clinical benefit in various human malignancies, including melanoma, non-small-cell lung cancer (NSCLC), head and neck squamous cell carcinoma, and bladder cancer [8][9][10][11]. Clinical trials have demonstrated the effectiveness of pembrolizumab and nivolumab (PD-1 inhibitors) in small cell lung cancer (SCLC) patients [12][13][14]. Given that the clinicopathologic features of PESCC are similar to those of SCLC, PD-L1/PD-1-based immunotherapy might be feasible to treat PESCC. Currently, the expression of PD-L1, as detected by immunohistochemical staining, is an important indicator for the use of PD-L1/PD-1 inhibitors in patients with lung cancer and melanoma [15]. However, growing evidence indicates that PD-L1 as a single biomarker is not precise enough to predict the response to PD-L1/PD-1 inhibition. Other factors, such as tumor-infiltrating lymphocyte (TIL) subsets, also need to be taken into consideration [16][17][18]. Most studies of PESCC are case reports focusing on individuals, and no systematic investigation of the PD-L1 expression pattern and immune cells within the TIME of PESCC has been reported.
Here, we performed a holistic assessment of the expression frequency of PD-L1 in tumor and tumor-infiltrating immune cells (CD4 + , CD8 + , and FoxP3 + T cells and CD163 + macrophages) to get a comprehensive landscape of the TIME in PESCC and provide practicable markers for patient enrollment in future clinical trials. Furthermore, we constructed a prognostic nomogram to predict the survival of PESCC patients, aiming to better classify patients for checkpoint inhibitor immunotherapy.

Patient Selection and Specimens.
A total of 81 patients with PESCC who received surgical resection or biopsy in 4 hospitals between 2010 and 2017 were recruited for this research. All recruited patients met the following criteria: (1) tumors which were primary, and recurrences were excluded; (2) pathologically confirmed small cell carcinoma according to the 2017 WHO classification of neuroendocrine carcinomas (NECs); (3) clinical staging of tumors in terms of TNM classification system of the American Joint Committee on Cancer (AJCC, 8th edition); and (4) complete clinicopathologic and follow-up data. This study was approved by the Institution Ethics Committee and Institutional Review Board of all participating institutions, and all procedures were conducted in accordance with the 1975 Declaration of Helsinki. Written informed consent was obtained from all participants.
2.2. PESCC Histologic Diagnoses. Histological diagnoses were given following an extensive review of all specimens by 2 independent pathologists (Wu and Zhang). Morphological features were evaluated on hematoxylin and eosin (H&E) slides: tumor cells are small, usually less than the sizes of 3 small lymphocytes, and are shaped with round, ovoid, or spindled nuclei and scant cytoplasm. Nuclear chromatin is finely granular, and nucleoli are invisible or inconspicuous. Cell borders are hardly seen, and nuclear mitosis is commonplace (at least 20 mitoses per 2 mm 2 ). Densely packed small tumor cells commonly display a sheet-like diffuse architecture. Architectural patterns, such as nesting, trabeculae, peripheral palisading, and rosette formation (as seen in other neuroendocrine tumors), are comparatively rare in PESCC. Comedo necrosis, extensive necrosis, brisk apoptotic activity, and the Azzopardi effect may all be present (Figure S1A-C). Standard immunohistochemical markers were also applied to the diagnosis of each patient: neuroendocrine markers (chromogranin A, synaptophysin, CD56, or neuron-specific enolase), Ki67, and cytokeratin-Pan (AE1/AE3). Samples were diagnosed as NEC if positive for one or more neuroendocrine markers and have more than 20% of the Ki67 labeling index ( Figure S2). Patients with NEC in other organs were excluded.

Distinguishing Tumor and Stroma.
Stroma that sustains cancer cells mainly consists of the basement membrane, fibroblasts, extracellular matrix, immune cells, and vasculature [19]. As mentioned above, the tumor niche was identified according to the specific morphological features of malignant cells; the stroma area without cancer cells could be clearly distinguished. As we and others observed, PESCC shows mild or no immune cell infiltration in the tumor niche. Notably, lymphocytes and plasma cells, among other immune cells in the stroma, tend to be lined or gobletshaped, but lymphoid follicles are rarely formed at the tumor boundary, and capillary infiltration is more likely to be seen ( Figure S1).

Evaluation of Tumor-Infiltrating Immune Cells by H&E Staining.
A full-section H&E slide was screened for tumorinfiltrating immune cells (including lymphocytes, macrophages, and plasma cells) and defined as poor (no immune cells or a mild and patchy immune cell pattern at the tumor margin) or rich (prominent band-like or florid cup-like immune cells at the invasive edges) ( Figure S1D- Briefly, serial 4 μm tissue sections from paraffin blocks were prepared, and formalin-fixed and paraffin-embedded (FFPE) tissue sections were deparaffinized and dehydrated in xylene and graded ethanol solutions. Antigen retrieval was conducted in Tris-EDTA buffer (pH 9.0). A peroxidase-labeled secondary antibody (EnVision/HRP system, DAKO, Carpinteria, CA) was used to visualize antigen, and the DAKO Catalyzed Signal Amplification System for rabbit/mouse antibodies was used for staining and detection (DAKO, Carpinteria, CA). Normal IgG was used as the negative control and tonsil tissue as positive.

Evaluation of PD-L1 Expression and Tumor-Infiltrating
Immune Cells by Immunohistochemistry Staining. IHC staining was double-blinded examined by 2 pathologists (Wu and Zhang). PD-L1 positive samples were identified using combined positive score (CPS)-the ratio of all membranous expression on tumor cells and immune cells in stroma (e.g., lymphocytes and macrophages) to the total number of viable tumor cells. Sample with a CPS ≥ 1% will be considered as PD-L1 positive [13,25].
Analysis of multi-IF staining was performed using the PerkinElmer Vectra (v1.0; PerkinElmer) platform. Singlestained tissue for each reagent was scanned to build spectral libraries to unmix the multispectral images by using Inform Advanced Image Analysis software (Inform v2.1.0; PerkinElmer). Tissue segmentation, cell segmentation, and phenotyping were performed with Inform Advanced Image Analysis software.

Statistical
Analyses. Statistical analyses were conducted in IBM SPSS statistics (version 25) software. The association between PD-L1 protein expression and clinicopathologic characteristics was analyzed using the chi-square test or Fisher's exact test. The Wilcoxon test was applied to compare continuous variables of the two groups. The Kaplan-Meier models were used for survival analysis. To detect independent predictive factors for aggressive features of PESCC, a Cox proportional hazards regression model was applied. Covariates with p values < 0.01 in the univariate analysis were included in the multivariate analysis. Odds ratios (ORs) and their 95% confidence intervals (CIs) were estimated for each factor. p < 0:05 (two-tailed) was considered statistically significant.

Development of the Prognostic Nomogram.
A nomogram that demonstrated prognostic significance of different risk factors in a single figure was established using the rms package in R (version 3.4.2). Included factors for creating the prognostic nomogram were selected in accordance with the Cox proportional hazards regression model using backward stepwise selection.

Patient Demographics and Clinicopathological
Characteristics. The diagnosis of PESCC in this study was confirmed by hematoxylin and eosin (H&E) staining and IHC staining for neuroendocrine markers, epithelial markers, or TTF1 and Ki67 (Figures S1 and S2). Most patients (79 of 81) had pure PESCC that did not have significant components of invasive adenocarcinoma or squamous cell carcinoma, whereas 2 patients contained a minor component (less than 20%) of invasive squamous cell carcinoma and 30.86% (25/81) of patients presented extensive necrosis. The basic characteristics of these patients are summarized in Table S1. The cohort of patients averaged was 63 years old (from 40 to 78 years). Most tumors (74 of 81) were located between the middle and lower esophagus, with only 7 tumors located in the upper esophagus. Frontline treatment varied with histology and stage. In this study, 45 patients received surgical treatment, 3 Journal of Immunology Research among whom 14 underwent postoperative chemotherapy after surgery and 10 received postoperative radiochemotherapy after surgery because of a positive margin. Among the 36 patients who already had distant metastasis or other underlying health conditions that indicated their unsuitability for surgery, 27 received chemotherapy only, while 9 received radiochemotherapy only. No patient received immunotherapy (Table S2).

The PD-L1 Protein Is Expressed Mainly in the Adjacent
Stroma and Is Associated with Overall Survival in PESCC. PD-L1 protein expression was examined by IHC. Three patterns of PD-L1 expression within PESCC tumor tissue were found in these specimens: PD-L1 negative, PD-L1 positive with staining in the stroma only, and PD-L1 positive with staining in both the tumor and stroma (Figure 1(a)). Unexpectedly, unlike findings observed in oral tongue squamous cell carcinoma (OTSCC) and NSCLC [26,27] Table S3). In coincidence with the observation in OTSCC [26], strong PD-L1 staining intensity was observed in areas of tissue necrosis ( Figure S4). The concordance of PD-L1 status between the 2 pathologists who analyzed the specimens (Wu and Zhang) was 97%. When patients were categorized into a PD-L1-positive group and a PD-L1-negative group for the correlation between PD-L1 expression and clinicopathological characteristics, the results showed that PD-L1 expression was not correlated with sex, age, tumor location, tumor depth, lymphatic metastasis, distant metastases, necrosis, or tumor stage (Table 1). Further, the Kaplan-Meier analysis demonstrated that PD-L1-positive patients experienced prolonged overall survival (p < 0:05, log-rank test) (Figure 1(c)). Multivariate Cox regression analysis showed that PD-L1 expression remained significantly correlated with favorable overall survival in PESCC after adjusting for other clinical factors (HR = 0:48, 95% CI: 0.27-0.86; p = 0:014; Table 2). This result indicated that PD-L1 expression was an independent prognostic predictor of PESCC.

A Higher
Frequency of CD4 + TILs Colocalized with PD-L1 + CD163 + TAMs Than with CD8 + TILs. As described above, immune infiltration frequency was associated with PD-L1 expression within the stroma. We further explored the relationship between PD-L1 + CD163 + TAMs and TILs on the basis of a previous report that demonstrated the PD-L1 + TAM is an ideal indicator for PD-1/PD-L1 blockade treatment [28]. We used multi-IF analyses to confirm that PD-L1 + CD163 + TAMs were located mostly within the tumor stroma (Figures 2(a) and 2(b)), and in patients with necrosis, PD-L1 + CD163 + TAMs were observed within necrotic tissue ( Figure S4). Furthermore, in all 27 PD-L1-positive specimens, 20 of them had a higher frequency of CD4 + TIL infiltration. And in these 20 samples, PD-L1 + CD163 + TAMs colocalized more with CD4 + TILs than CD8 + TILs (Figures 2(c) and 2(d) and Figure S5).

CD4 + TILs Are the Most Clinical Outcome-Related
Components among Inflammatory Cells in the Stroma and Positively Correlate with PD-L1. Notably, IHC analysis of CD4 + and CD8 + T cell proportions revealed that most specimens (68/81) had more CD4 + TILs than CD8 + TILs (Figure 3(c)), in agreement with the results of multi-IF staining for CD4 and CD8 (Figures 3(a) and 3(b)). In addition, more CD4 + TILs colocalized with PD-L1 than CD8 + TILs (Figure 3(d)), suggesting that CD4 + TILs are more responsible for PD-L1-induced anergic lymphocytes than CD8 + TILs and indicating the involvement of CD4 + TILs in PD-L1/PD-1 function in PESCC. Furthermore, the Kaplan-Meier analysis of overall survival revealed that patients whose tumors were rich in CD4 + TILs experienced shorter survival (Figure 3(e)). When patients were classified by CD8 + TILs, tumor-infiltrating immune cells, or CD163 + cells (poor or rich), their overall survival showed no significant difference ( Figure S6A-C). However, a higher CD8/CD4 T cell ratio was associated with a more desirable clinical outcome (Figure 3(f)).

A Low FoxP3 + /CD8 + T Cell Ratio Correlates with Good
Clinical Outcomes in PESCC Patients. Since we identified a predominant contribution of CD4 + TILs in this cohort, which was associated with short overall survival, we suspected that CD4 + FoxP3 + regulatory T cells (Tregs) might play an instrumental role in this context. Besides examining all 81 samples with IHC staining for FoxP3 (mean number: 4 Journal of Immunology Research   (Figure 4(a)). The number of Tregs in the tumor, to a large extent, depended on the abundance of CD4 + TILs but not CD8 + TILs (Figure 4(b), Table S5). The Kaplan-Meier analysis showed that a small number of Tregs and a low FoxP3/CD8 T cell ratio were correlated with prolonged overall survival (Figures 4(c) and 4(d)).
3.7. The TIME Is Different between PD-L1-Positive and PD-L1-Negative PESCC. To further analyze the role of PD-L1 in the TIME of PESCC, prognostic analysis was performed. The Kaplan-Meier analysis showed that a small number of Tregs, as well as a low FoxP3/CD8 T cell ratio and a high CD8/CD4 T cell ratio, were correlated with prolonged overall survival in PD-L1-positive PESCC ( Figure S7A-C). When PD-L1-positive patients were classified by tumorinfiltrating immune cell enrichment (or CD8 + TIL, CD163 + TAM, and CD4 + TIL enrichment) and by necrosis harbored in tumor tissue, there was no significant difference in overall survival ( Figure S7D-H). In PD-L1negative PESCC, the prolonged overall survival is related to no necrosis in tumor tissue, a small number of Tregs, and fewer CD4 + TILs ( Figure S8A-C), while the other indicators had no significant correlation with overall survival ( Figure S8D-H). , and FoxP3, we entered TNM stage, CD4, and the FoxP3/CD8 ratio [29] into the multivariate analysis, which revealed one tumor characteristic (TNM stage (p = 0:003)) and three immune variables (PD-L1 (p = 0:014), CD4 (p = 0:030), and FoxP3/CD8 (p = 0:044)) that were independent indicators for overall survival and were thereby included in the following predictive model ( Table 2).

Development of a Nomogram Based on Independent
Although three immune variants showed independent prognostic significance, their intricate interaction within the TIME does not allow any one of these factors to be accurately indicative of survival in PESCC patients. In this sense, it is necessary to identify a comprehensive immunofactor model. A nomogram for overall survival prediction was built using the 4 prognostic factors mentioned above. The nomogram comprises 10 rows, with their representation as follows. The

Discussion
PESCC is a highly metastatic cancer with a poor outcome. Due to a lack of large-scale clinical studies, opinions on the treatment of PESCC remain controversial, although the importance of surgery has been recognized. Patients with PESCC, even those with early-stage disease, are plagued by high recurrence [30,31]. To date, the association between the strong aggressiveness and poor prognosis of PESCC and specific biomarker has not been identified in clinical     Journal of Immunology Research practice. Increasing evidence has indicated that tumor progression is determined by extrinsic immunological factors in the TIME as well as its intrinsic characteristics (for example, TNM stage) [32][33][34]. Our data showed that in PESCC, PD-L1 is predominantly expressed in the adjacent stroma rather than the tumor nest, which contrasts with nonsmall-cell solid cancers, such as head and neck squamous cell carcinoma and NSCLC [26,27]. We found that PD-L1 + patients experienced prolonged overall survival, suggesting that PD-L1 expression by the tumor-infiltrating immune cells is correlated with a more beneficial immune response. Further analysis revealed positive correlations between the expression of PD-L1 and tumor-infiltrating immune cells, including TAMs and TILs. PD-L1 + CD163 + cells were colocalized more with CD4 + TILs than with CD8 + TILs. Besides, we observed a higher frequency of CD4 + TILs than CD8 + TILs, and the proportion of Tregs was positively correlated to CD4 + TILs in these patients. These results elaborate the suppressive TIME status in PESCC patients. In addition to TNM stage, we found that the status of PD-L1, CD4, and the Foxp3/CD8 ratio can predict the outcome of PESCC, and we further constructed a nomogram that included these parameters for survival prediction. Furthermore, the likelihood ratio test showed that this immunofactor model is superior to the model that includes merely TNM stage.
Therapeutic options for PESCC are very limited. The success of immune checkpoint-blocking antibodies in treating melanoma and NSCLC has inspired researchers to introduce this promising immunotherapy into other aggressive solid tumors [13,14,35,36]. Immunohistochemical staining of patient-derived tumor tissues revealed that most patients in our cohort expressed PD-L1 only in the adjacent stroma. Similar to our observation, it has been reported that PD-L1 is negative on tumor cells but positive in the stroma in small cell NECs (including 61 pulmonary and 33 extrapulmonary tumors) [37]. Taken together, these data suggest that small cell carcinomas may display a distinct stromal pattern of PD-L1 compared to non-small-cell carcinomas, in which PD-L1 is expressed mainly on tumor cells. However, investigations using larger sample sizes and with small cell carcinomas from different origins should be conducted to confirm these results. Recent studies verified that the presence of PD-L1 on tumor-infiltrating immune cells has predictive implication for anti-PD-1 therapy [11,13,38]. Since patients with NECs have similar morphological and pathological features [39], it is reasonable to hypothesize that PESCC patients with stromal PD-L1 might respond to anti-PD-1/PD-L1 therapy. Therefore, PD-L1 expression in tumorinfiltrating immune cells should also be taken into consideration as a companion diagnostic in clinical trial designs for anti-PD-1/PD-L1 therapy, especially for PESCC.
PD-L1 coexpression with tumor-infiltrating immune cells may act as an adaptive mechanism for immune escape [40][41][42]. We conducted analyses as per the TIME in the context of PD-L1 expression to evaluate the correlation between PD-L1 and immune cells and to further determine their prognostic values. A study in head and neck squamous cell carcinomas found that Tregs represent only 2 to 15% of CD4 + TILs [26]. In the current study, we found that Tregs impressively ranged between 13% and 27% of CD4 + TILs in PESCC. Furthermore, PD-L1 was colocalized with CD4 + TILs rather than CD8 + TILs, in agreement with that in OTSCCs [26]. We showed elevated expression of both CD4 + TILs and Tregs, and a low CD8/CD4 ratio correlated with poor overall survival in PESCC. These results suggest that CD4 + TILs, especially Tregs, play pivotal roles in the immunosuppressive microenvironment in PESCC. PD-L1 expressed on immune cells of the innate immune system, such as macrophages, can trigger important modulatory effects within the TIME [20,43]. We found that PD-    (d) A high FoxP3/CD8 ratio (red line, n = 28) was associated with shorter OS than a low FoxP3/CD8 ratio (blue line, n = 49).
The Kaplan-Meier analysis revealed different prognostic values for the overall survival of PD-L1-positive and PD-L1negative patients. A high CD8 + /CD4 + ratio, a low Foxp3/CD8 ratio, and low Foxp3 expression were associated with prolonged survival in PD-L1-positive patients, while the presence of necrosis, high CD4 expression, and high Foxp3 expression were associated with poor survival in PD-L1negative patients. Extrinsic (immune-induced) and intrinsic oncogenic activation can regulate PD-L1 expression. We demonstrated that the balance of suppressive and effective TILs is prognostic for the overall survival of PD-L1-positive patients but not for that of PD-L1-negative patients, indicating that the state of the immune system is an important player in the progression of PESCC. Therefore, it is necessary to evaluate the expression of PD-L1 and the proportion of TIL subsets at the same time.
Tumor necrosis is an important hallmark of aggressive cancers. Tumor necrosis induced by hypoxia attracts macrophages to migrate into tumors, which then contribute to angiogenesis and a poor prognosis [44] (Figure S6D). We found that necrosis indicated a poor prognosis in patients with negative PD-L1 expression but not in patients with positive PD-L1 expression. Combining this result with the observation that PD-L1-positive specimens have increased infiltration of CD163 + cells in the necrotic areas, we speculate that it is the recruitment of TAMs into the necrotic areas that contributes to the different prognostic roles of necrosis here. However, the expression of CD163 alone is not significantly correlated with overall survival. Further analyses on the functions of these TAMs in small cell carcinomas, particularly in necrotic areas, are still needed.

Conclusions
In this study, we present a comprehensive picture of the TIME, which is in a suppressive state in PESCC, by examining PD-L1 expression and analyzing its correlation with tumor-infiltrating immune cells. In addition, using prognostic analyses and building a prognostic nomogram based on the independently prognostic immune variables, we provide an alternative model for survival prediction, and we found that the poor survival of PESCC patients was attributed to a suppressive TIME. Future studies remain necessary to compare the predictive accuracy between this model and the TNM staging system in larger PESCC cohorts so as to confirm the predictive value of this model.

Data Availability
The data supporting the conclusions of this article are included in the article.

Ethical Approval
This study was approved by the Institution Ethics Committee and Institutional Review Board of the Tumor Hospital of Shantou University Medical College, Jieyang People's Hospital, Shantou Central Hospital, and the Second Affiliated Hospital of Shantou University Medical College, and all procedures were conducted in accordance with ethical principles.

Conflicts of Interest
The authors declare that they have no potential conflict of interests. To estimate the OS of an individual patient, the value of each factor is acquired on each variable axis, and a straight line is drawn upward to determine the points. The sum of these 4 numbers is located on the total points axis, and then, a line is drawn downward to the survival axes to determine the likelihood of survival.