Neoadjuvant Pembrolizumab and Chemotherapy in Resectable Esophageal Cancer: An Open-Label, Multicenter, Single-Arm Study

Background: Multimodal therapies, such as perioperative chemotherapy and neoadjuvant chemoradiotherapy together with surgical advances, have improved the outcomes of patients with resectableesophageal cancer (EC).However, the 5-year survival of these patients remains below 50%. The advent of immunotherapy has resulted in novel approaches to advanced or metastatic disease, but the role of immunotherapy in the neoadjuvant settings has not yet to be established. In this single-arm study, the ecacy and safetyofneoadjuvant pembrolizumab plus chemotherapy were evaluatedin patients with resectableesophageal squamous cell carcinoma (ESCC). Methods: This study was conducted at 2 hospitals(in Xi'an and Tianjin)and included patients with ESCC of clinical stage II–IVAwho underwent surgery within 4 to 6 weeks after completing treatment withpembrolizumab(200mg) combined with a conventional chemotherapy regimen (3 cycles). The ecacy and safety of this combination treatment were evaluated as primary endpointsof the study. This study was retrospectively registered (Registration number: ChiCTR2100048917). Results:A total of 22 patients (including 18 men) were enrolled, of whom 17 patients progressed to surgery,4 patients refused surgery due to tumor regression and symptomatic relief, and 1 patient had progressive disease. As determinedby the Response Evaluation Criteria in Solid Tumors(RECIST,version 1.1), 7 patients (31.8%) had a partial response and 4 (18.2%) patients hada complete response, which translated to an objective response rate of 50%. The symptom remission rate was 90.9% according to the Stooler classication. Five patients (22.7%) experienced serious treatment-related adverse events of grade 3–4 (including anorexia, vomiting, fatigue, alopecia and hypoadrenalism). Postoperative pathology revealed a major pathological response in 14 cases (82.3%) and a complete pathological response in 8 cases (47.1%). Programmed death-ligand 1expression in pretreatment specimens was not signicantly associated with the percentage of residual viable tumor (RVT)(r=-0.55, p=0.08). Changes in counts of CD68+ macrophage between pre- and post-treatment specimens wereweakly correlated with RVT(r=0.71; p=0.07), while a positive correlation was observed between postoperative Foxp3+T cells/CD4+Tcellsratios and RVT(r=0.84, p=0.03). Conclusion: The combination of neoadjuvantimmunotherapy and chemotherapy for ESCC is associated with a high pathological response and immunologic effects in the tumor microenvironment. It has acceptable toxicity and great ecacy, suggesting a strong rationale for its further evaluation in randomized clinical trials. survival parameters are intriguing. In the TME, tumor cells, blood vessels, immune cells, lymphocytes, cancer stem cells, and cancer-associated broblasts mix, and there is considerable immune cell activity that may be stage- and context-dependent 16 . Macrophages are a key component of the TME. As the principal cells of antigen recognition and presentation, they secrete TNF-α, interleukin-1β, and other cytokines, and impact the magnitude and type of T-cell response. Studies report that a high macrophage count was associated with poor OS 17 , and this conclusion may also account for the positive correlation between RVT and changed CD68 expression in the post-treatment pathologic tissues compared with the pre-treatment samples. Accordingly, CD68 expression may be predictive of a poor response to immunotherapy, and this requires further study. Immunosuppressive cells (i.e., regulatory T cells [Tregs, Foxp3 + T]) are a part of inltrating CD4 + T-cell in the TME, which signicantly inhibit the T-cell-mediated anti-tumor effect and may be associated with T-cell dysfunction 18,19 . In other translational studies, Foxp3 + T cells in the TME of non-small cell lung cancer were associated with poor OS 20 , and in our study, a positive correlation between post-treatment Foxp3 + T cells/ (CD4 + T cells) ratios and RVT was conrmed in the context of ESCC, suggesting a prognostic role of post-treatment Foxp3 + T cells/ (CD4 + T cells) ratios. No direct correlation was found between the counts of other T cells and RVT. These results may be attributable to the time and space heterogeneity of immunotherapy, although the specic mechanism requires further study. TGF-β1, TNF-α and other cytokines in the TME also play key roles in regulating the response to immunotherapy 21 . Among these cytokines, immunoregulatory TGF-β1 suppresses the proliferation of B-cell, cytotoxic T-cell, and natural killer cell and antagonizes the biological effects of TNF-α 22

Conclusion: The combination of neoadjuvantimmunotherapy and chemotherapy for ESCC is associated with a high pathological response and immunologic effects in the tumor microenvironment. It has acceptable toxicity and great e cacy, suggesting a strong rationale for its further evaluation in randomized clinical trials.
Trial registration: ChiCTR, ChiCTR2100048917. Registered 19 July 2021 -Retrospectively registered, https://www.chictr.org.cn/edit.aspx? pid=130073&htm=4 Background Esophageal cancer (EC) is the 7th most common cancer-related death globally. In China, it is the 6 th most common malignancy, with esophageal squamous cell carcinoma (ESCC) being the dominant subtype 1 . The medianoverall survival (OS) of patients with advanced or metastatic esophageal cancer is extremely poor. For patients who have underwent surgeries alone, overall survival (OS) rates are improving; nonetheless, the ve-year survival ratedoes not exceed 50% (49.1%) 2 .
According to the current National Comprehensive Cancer Network (NCCN) guidelines, multimodal therapy with neoadjuvant chemoradiotherapy is the recommended standard therapy for patients with T2-4aNxM0 resectable ESCC.The CROSS Study is the de nitive modern randomized clinical trial (RCT),with an OS of 48.6 months versus 24.0 months in the multimodal and surgery-only cohorts, respectively 3 . However, a major limitationfor this treatment may be aheightened risk of major respiratory complications including pneumonia, acute respiratory distress syndrom, respiratory failure and pulmonary embolism and mortality postoperatively 4 .Consequently,with the advent of a greater understanding of EC tumor biology and genomics, novel approaches whichcombine e cacy and safety are being explored.
In this regard, there is currently enormous interest in therapies that target the immune cells within the tumor microenvironment (TME).
Programmed cell death protein-1 (PD-1) inhibitors have been evaluated in multiple clinical trials.In the KEYNOTE-181 study,pembrolizumab VS chemotherapy was evaluated as a second-line treatment foradvanced (unresectable or metastatic)EC. For patients with PD-L1 combined positive score (CPS)≥10, the 12-month OS rate was43% in the pembrolizumab group and 20% in the chemotherapy group 5 . And in the KEYNOTE-590 trail,combination of pembrolizumab and chemotherapy VS chemotherapy was evaluated as a rst-line treatment for the unresectable or metastaticEC.The survival rate at 12 monthsof ESCC was higher with chemoimmunotherapy versus chemotherapy (51% VS38%) 6 . In the neoadjuvant treatment ofnon-small cell lung cancer (NSCLC), PD-1 inhibitors have produced excellent results.In the NCT02716038 study, NADIM study and our recent trial, neoadjuvant chemoimmunotherapy in resectableNSCLC reported encouragingdata of pathologicalresponse, with MPR (57%, 83%, 50% respectively)and pCR (33%, 63%, 30% respectively) 7,8,9 . Given these encouraging trends for neoadjuvant regimens including anti-PD-1 therapy, the present study aimed to explore the safety and e cacy of anti-PD-1 therapy combined with chemotherapy for resectable ESCC in the neoadjuvant settings.We also preliminarily explore the correlations between pathological response and immunological parameters of the tumor microenvironment (TME). edition]) were eligible for inclusion. The key exclusion criteria were:patients withactive autoimmune disease; patients withactive concurrent malignancy; andpatients receiving ongoing systemic steroids(>10mg daily prednisone equivalents) or other immunosuppressive therapy.
All of the includedpatients were scheduled to receivethe following drugs intravenously:pembrolizumab (200mg) combined with conventional chemotherapy for three 21-day cyclesprior to surgical resection (Table 1).Surgery was plannedwithin 4-6 weeks afterthe completion of the induction regimen. McKeown or IvorLewis esophagectomy, including two-eld lymphadenectomy with totalmediastinal lymph node dissection was performed according to standard institutional procedures.

E cacy
E cacy was measured according to the following criteria: (1)Pathological complete response (pCR), de ned as the complete absence of tumor cells, or major pathological response (MPR), de ned as <10% residual viable tumor (RVT), or incomplete pathological response, de ned as ≥10% RVT (non-MPR/non-PCR) 10 ;(2)Symptom remission, according to the Stooler classi cation 11 ;(3)Treatment radiographic response,as determined using the Response Evaluation Criteria in Solid Tumors (RECISTversion1.1 .

Statistical analysis
Demographic and safety data, aswell as clinical, pathologic, radiographic, and molecular response data, were recorded using descriptive statistics.The associations between RVT and pretreatment PD-L1 expression were analysed using Spearman's correlation analysis. Furthermore, the associations of RVT and other pretreatment and posttreatment immune parameters combined with their changes were analysed using Spearman's correlation analysis, including the expression of TNF-α, TGF-β1 and the counts of Foxp3+ CD4+, CD8+ T cells and CD68+ macrophages. Similarly, the associations of RVT and pretreatment and posttreatment Foxp3+T cells/ (CD4+T cells) ratios and Foxp3+T cells/(CD8+T cells) ratios combined with their changes were analysed using Spearman's correlation analysis. Additionally, the differences in pathological complete response between patients whose PD-L1 combined positive score (CPS≥10) and CPS 10 were analysed with χ² test.All p values reported are 2-sided, with the signi cance level set at 0.05. Statistical analyses were performed using SPSS 19.0.

2.1E cacy
Patients' characteristics and pathological response data are summarized in Table 1. This study enrolled 22 patients (18 men and 4 women), of whom 17 patients progressed to surgery, 4 patients refused surgery due to signi cant tumor regression and symptomatic relief, and 1 patient experienced disease progression (patient 22) and was given de nitive chemoradiotherapy. According to RECIST 1.1, 7 patients (31.8%) had a partial response (PR) and 4 patients (18.2%) attained a complete response (CR), which translated to an objective response rate (ORR) of 50% (Fig. 1). According to the Stooler classi cation, prior to neoadjuvant treatment, 5 patients had stenosis of grade 3 and 17 patients had stenosis of grade 2. When after therapy, 18 patients had grade 0 symptoms, 2 patients had grade 1 symptoms, and 2 patients had grade 2 symptoms, with a

Pathologic assessment and genomic analyses
The data suggested that RVT was not signi cantly associated with pre-treatment PD-L1 expression (r=-0.55; p = 0.08) (Fig. 4). Additionally, no signi cant differences in pathological complete response were identi ed between patients whose PD-L1 combined positive score (CPS ≥ 10) and CPS 10.
Further analysis showed that the changes in counts of CD68 + macrophage were found to be positively correlated with RVT (r = 0.71; p = 0.07) (Fig. 5A, D). To explore the relationship between in ammatory cytokines and RVT, immunohistochemical methods were adopted to examine preand post-treatment expression of TNF-α and TGF-β1 in the pathologic specimens (Fig. 5B, C). In this study, the post-treatment expression of TGF-β1 was increased compared to the preoperative expression, and the changes in TGF-β1 expression was positively correlated with RVT (r = 0.65, p = 0.11) and possibly indicated a poor prognosis (Fig. 5E). However, the available data showed no signi cant correlation between the changes in TNF-α expression and RVT (Fig. 5F). The correlations of RVT and the parameters of lymphocyte populations stated above were further explored ( Fig. 6A, B), with a positive correlation observed between postoperative Foxp3 + T cells/ (CD4 + T cells) ratios and RVT (r = 0.84, p = 0.03) (Fig. 6C), positive correlation observed between changes in Foxp3 + T cells/ (CD4 + T cells) ratios and RVT (r = 0.61, p = 0.15) (Fig. 6D) and negative correlation observed between the counts of postoperative CD8 + T cells and RVT (r=-0.61, p = 0.14) (Fig. 6E). However, no direct correlation was found between other types of immune cells and RVT.

Discussion
Different from PALACE-1, which combines immunotherapy and Chemoradiotherapy for resectable esophageal cancer in the neoadjuvant settings, our study is the rst to report on pembrolizumab combined with chemotherapy alone in the neoadjuvant treatment of EC 12 . Our study were done at two centers involving 22 patients, small sample size though, important observations were made.
Firstly, the pCR of 47.1% was high, not dissimilar to the rates observed with chemoradiation in the CROSS trial (in ESCC, 49%) and the NEOCRTEC 5010 RCT (43.2%) 3,13 . This level of pathological response has not been observed in previous studies with chemotherapy alone in neoadjuvant settings, of which the pCR were typically less than 20% (12.8%) 14 . Moreover, pCR and MPR are veri ed to confer a survival advantage and to prolong median disease-free survival in EC and in many other cancers 10

Consent for publication
Not applicable Availability of data and material All data generated or analysed during this study are included in this published article.
also unrelated to the regimen, having been caused by microbiologically proven bacterial pneumonia and sepsis. This complication is distinct from reported immunotherapy-related pneumonitis, which can be ameliorated by steroids, but which had no impact on this patient 15 .
The associations between residual viable tumor (RVT) and the immunologic parameters are intriguing. In the TME, tumor cells, blood vessels, immune cells, lymphocytes, cancer stem cells, and cancer-associated broblasts mix, and there is considerable immune cell activity that may be stage-and context-dependent 16 . Macrophages are a key component of the TME. As the principal cells of antigen recognition and presentation, they secrete TNF-α, interleukin-1β, and other cytokines, and impact the magnitude and type of T-cell response. Studies report that a high macrophage count was associated with poor OS 17 , and this conclusion may also account for the positive correlation between RVT and changed CD68 expression in the post-treatment pathologic tissues compared with the pre-treatment samples. Accordingly, CD68 expression may be predictive of a poor response to immunotherapy, and this requires further study. Immunosuppressive cells (i.e., regulatory T cells [Tregs, Foxp3 + T]) are a part of in ltrating CD4 + T-cell in the TME, which signi cantly inhibit the T-cell-mediated anti-tumor effect and may be associated with Tcell dysfunction 18,19 . In other translational studies, Foxp3 + T cells in the TME of non-small cell lung cancer were associated with poor OS 20 , and in our study, a positive correlation between post-treatment Foxp3 + T cells/ (CD4 + T cells) ratios and RVT was con rmed in the context of ESCC, suggesting a prognostic role of post-treatment Foxp3 + T cells/ (CD4 + T cells) ratios. No direct correlation was found between the counts of other T cells and RVT. These results may be attributable to the time and space heterogeneity of immunotherapy, although the speci c mechanism requires further study. TGF-β1, TNF-α and other cytokines in the TME also play key roles in regulating the response to immunotherapy 21 . Among these cytokines, immunoregulatory TGF-β1 suppresses the proliferation of B-cell, cytotoxic T-cell, and natural killer cell and antagonizes the biological effects of TNF-α 22 . In other studies, it was reported that TGF-β signalling may counteract anti-tumour immunity by restricting the movement of T-cells in the TME 23 . We hypothesized that the increases in TGF-β may predict poor pathological response. This hypothesis was supported by our ndings that changes in TGF-β1 expression was positively correlated with RVT. It may be di cult that the small sample size in this study precluded a full analysis of the relationship between TNF-α, TGF-β1, and RVT, and further study is clearly required.
There are some limitations to this study that should be noted. Firstly, other markers of relevance including the genomic pro le, tumor mutational burden, and the in ammatory factor interferon-gamma were not evaluated. A further limitation is the study's small sample size; however, since it is the rst clinical trial of its kind, it still has important clinical observational signi cance to serve as the backbone for larger analyses. Another important aspect is that the OS of patients was not extensively explored, as the main endpoint of the study was safety. However, previous studies have shown that high pCR are closely related to patient OS, which will be the focus of our next study.