COL19A1 is a predictive biomarker for the responsiveness of esophageal squamous cell carcinoma patients to immune checkpoint therapy

Abstract Background The use of neoadjuvant immunotherapy plus chemotherapy has revolutionized the management of esophageal squamous cell carcinoma (ESCC) patients. Nevertheless, patients who would maximally benefit from these therapies have not been identified. Methods We collected postoperative specimens from 103 ESCC patients, of which 66 patients comprised a retrospective cohort and 37 comprised a prospective cohort. Patient specimens were subjected to applied multi‐omics analysis to uncover the mechanistic basis for patient responsiveness to cancer immunotherapy. The tumor microenvironment characteristics of these patient specimens was explored and identified by multiplex immunofluorescence and immunohistochemistry. Results Results demonstrated high COL19A1 expression to be a novel biomarker for successful immunotherapy (COL19A1 high, odds ratio [95% confidence interval]: 0.31 [0.10–0.97], p = 0.044). Compared with COL19A1 low patients, COL19A1 high patients benefited more from neoadjuvant immunotherapy (p < 0.01), obtained better major pathological remissions (63.3%, p < 0.01), with a trend toward better recurrence‐free survival (p = 0.013), and overall survival (p = 0.056). Moreover, analysis of an immune‐activation subtype of patients demonstrated increased B cell infiltration to be associated with favorable patient survival and a better response to neoadjuvant immunotherapy plus chemotherapy. Conclusions The findings of this study provide insight into the optimal design of individual treatments for ESCC patients.


INTRODUCTION
Esophageal cancer (EA) is the sixth most frequently diagnosed cancer type and the seventh leading cause of Jian-Hua Liu, Ju-Ze Lin, and Qianhui Qiu contributed to the work equally.
tumor-associated mortality worldwide. 1 As the predominant histological subtype, esophageal squamous cell carcinoma (ESCC) is characterized by an aggressive nature, accounting for approximately 90% of EA cases in China. 2 Due to an absence of specific symptoms for early ESCC, most patients have advanced disease at initial diagnosis. Although significant treatment progress has been accomplished by combination therapy (chemotherapy, targeted therapy, radiotherapy, and surgery), survival of patients with ESCC remains dismal, with a 5-year survival ≤50%. 3 Further, the efficacy of conventional chemotherapy has plateaued due to the development of tumor resistance to the therapeutic agent and the inherent toxicity of the agent. Therefore, there is an urgent need to develop revolutionary treatments for ESCC.
Currently, the use of antiprogrammed death 1 (PD-1)/programmed death-ligand 1 (PD-L1) antibodies has been a striking success. The antibodies are immune checkpoint inhibitors (ICIs) that boost host immunity. The use of these antibodies represents a novel approach to drug development and precision medicine. 4,5 It is well known that ESCC patients can benefit remarkably from first-, second-line, and perioperative immunotherapy. [6][7][8][9][10] For locally advanced ESCC, accumulating evidence demonstrates neoadjuvant immunotherapy combined with chemotherapy and surgery, results in better tumor regression and R0 resection rates than chemotherapy alone. [11][12][13] Our previous study confirmed the promising efficacy and good safety of ICIs in patients with resectable ESCC. 14 Further, there are ongoing clinical trials of these forms of neoadjuvant immunotherapy, such as JCOG1804E and Palace-1. 15,16 However, some ESCC patients fail to benefit from ICIs, with the reason for these failures unclear. Accumulating evidence indicates that the tumor immune microenvironment (TIME) is key to tumor-immune interactions and the response to therapy. [17][18][19] TIME is influenced by many factors including tumor mutations, protein overexpression, cytokines, and inflammation, among others. [18][19][20] ESCC has a high degree of heterogeneity, with the tumor microenvironment (TME) of ESCC and its interaction with immunotherapy poorly understood. It is therefore imperative to explore the immune response within the ESCC TME and also to observe resultant clinical outcomes. At present, biomarkers related to the efficacy of immunotherapy are the expression level of PD-L1, 21 mismatch repair-deficient/high microsatellite instability, 22 and tumor mutation burden (TMB). 23,24 To date, there has been no comprehensive transcriptomic analysis of the intertumor complexity of ESCC and the response of ESCC to neoadjuvant immunotherapy.
Herein, we classified ESCC based on transcriptomics and subtype-specific TME characteristics. COL19A1, a novel biomarker overexpressed in an immune-enriched subtype, was found to be associated with the best overall survival and with a major pathological response (MPR) following neoadjuvant immunotherapy. Based on these results, an immunerelated prediction model for prognosis and response to immunotherapy was successfully established. The model allows for exploration of the immune-activated ESCC landscape. The use of the model will aid development of immune therapies for ESCC patients.

RNA sequencing
Total RNA was extracted according to the manufacturer's instructions. RNA quantity, quality, and integrity was assessed, with an RNA integrity score >6.0 considered satisfactory. RNA was fragmented in accordance with the DV200 value. After evaluation of fragment length, reverse transcription and complementary DNA synthesis of RNA fragments were performed, followed by preparation of a strand-specific library.
Prepared libraries were sequenced with an Illumina Nova Seq 6000 instrument (Illumina) using 2 Â 150 bp paired-end reads based on established experimental quality control parameters. Raw sequencing data were then converted to fastq format (Illumina), as described in our recent study. 14

Consensus clustering
Unsupervised consensus clustering, a class discovery approach, is used to identify unknown possible subtypes that present similar intrinsic features according to differences in gene expression profiles. 25 In this study, after 80% of the specimens were sequentially extracted 100 times, 5000 highly variable expressed genes were applied for hierarchical clustering analysis. Results were based on construction of a cumulative distribution function (CDF) in accordance with consensus matrices.

Differentially expressed gene (DEG) selection and pathway analysis
RNA sequencing (RNA-seq) expression data of the two cohorts were transformed into transcripts per million (TPM) for analysis. DEGs were defined based on specific criteria as follows; specific subtype versus other subtypes, genes with Log 2 FC ≥1, and p < 0.05. Heatmaps and volcano diagrams were constructed to visualize DEGs.
Differential gene expression analysis was performed using the EdgeR package, with major criteria as previously described. 14 Enrichment analyses were performed to determine gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways.

Evaluation of infiltrating immune cells in the TME
Based on the transcriptome data of the two independent cohorts, characteristics of the ESCC TME were identified. The relative enrichment of infiltrating immune cells was evaluated using a single-sample gene set enrichment analysis (ssGSEA) algorithm. Twenty-eight immune cell types and 29 types of immune-related signatures 26 were assessed to calculate the relative proportion of immune cell types.

Immunohistochemistry (IHC) and multiplex immunofluorescence (mIF)
To validate results, IHC and mIF were carried out following recommended protocols. 27 For IHC analysis, formalin-fixed paraffin-embedded (FFPE) tumor specimens were stained with primary antibodies reactive with; Ki-67, PD-L1, and COL19A1. The Akoya Opal 5-color fluorescent detection platform was used to identify distinctive ESCC TME characteristics. Tissues were simultaneously examined with an Opal Polaris 5-Color Automation IHC Kit. Image acquisition and signal quantification (CD8, CD20, CD27, and DAPI) were performed using the Panoramic MIDI Pathology Imaging System. For quality control purposes, each analyzed image was reviewed independently, with histological evaluations performed independently by two blinded pathologists. Detailed antibody information is provided in Appendix S1, Supplementary Materials.

Statistical analysis
The Kaplan-Meier method was used to determine survival outcomes (RFS and OS). Differences in variables were compared using the Kruskal-Wallis test. R package analysis included "Consensus Cluster Profiler (Unsupervised clustering)," "DESeq29 (Differential gene expression)," "ggplot2 (Boxplot)," and "randomForest (Subtype prediction model)." R software (V4.0.5) was used for all bioinformatic analysis. A p ≤ 0.05 for the two-sided Mann-Whitney U test was considered statistically significant.

Study population and baseline characteristics
A total of 103 ESCC patients were included in this study, with (N = 66) for the retrospective cohort and (N = 37) for the prospective cohort. A flowchart of the study is shown in Figure 1. Details of baseline characteristics are summarized in Tables 1 and 2, which have not been previously reported. 14 A total of 19 (28.8%) patients in the retrospective cohort received postoperative adjuvant therapy. A total of 22 patients died during the follow-up period, and all deaths were due to recurrent or progressive disease. All patients in the prospective cohort were treated with the TP regimen (nab-paclitaxel and cisplatin) combined with a PD-1 inhibitor (tislelizumab in 29 cases, pembrolizumab in 8 cases). A total of 51.4% of patients achieved MPR, with no patients lost to follow-up.
The distributions of clinical stage, lymph node (N) stage, tumor differentiation, and other characteristics among the four clusters were compared (Table 1). Results demonstrated molecular subtypes to be significantly associated with clinical stage and pN stage. Patients with the C3 and C4 subtypes tended to have more localized tumor lesions ( Figure S1B-E). Moreover, patients with these two subtypes had a lower incidence of vascular invasion and a slightly higher incidence of neuro-invasion than those with the other subtypes (Figure S1F-G). In contrast, in the prospective cohort, 35.1% (13/37) of patients were classified as the C1 subtype, 51.4% (19/37) as the C2 subtype, and 13.5% (5/37) as the C3/4 subtype (including four cases of the C3 subtype and one case of the C4 subtype). These patients were combined for further analysis due to the similarity of immune profiles.
COL19A1 high expression was associated with improved responsiveness to immunotherapy and better ESCC patient prognosis In the retrospective cohort, survival analysis demonstrated patients with the C3 subtype to have a more favorable OS than those with the other subtypes. Survival curve characterization and cluster analysis have been previously discussed in a separate publication. 14 We observed that six upregulated genes were related to longer RFS for C3 subtype patients of the retrospective cohort (Figure 2a, b). COL19A1 has seldom been investigated in ESCC patients and was selected for further analysis. Patients with high COL19A1 expression had better RFS ( p = 0.013) (Figure 2c) and OS (p = 0.056) (Figure 2d). The mortality of COL19A1 low patients was significantly greater ( p < 0.01) than that of COL19A1 high patients (Figure 2e). IHC staining demonstrated COL19A1, which is expressed by cancerous epithelial cells, to be upregulated in ESCC samples from patients with good outcomes (Figure 2f).

COL19A1 high expression was associated with ESCC immune activation
The TME is associated with the efficacy of immunotherapy. In this study, patients in the retrospective cohort with the C3 subtype had distinct biological features that were related to immunotherapy responsiveness. C3-upregulated genes ( Figure 3a) were enriched for 18 immune-related GO biological process terms, including "B cell activation," "B cell proliferation," "B cell receptor signaling pathway," and "antigen receptor mediated signaling pathway" (Figure 4a). Significant differences in memory B cells ( p < 0.0001), immature B cells (p < 0.05) and activated B cells ( p < 0.05) were evident between the C3 subtype and the other subtypes ( Figure S2).
Upregulated COL19A1 gene expression was related to immune effector cells (memory B cells, immature B cells, and activated B cells) in the retrospective cohort (Figure 4b). In parallel, the infiltration level of B cells ( p < 0.001), checkpoint molecules ( p = 0.022), and T cells F I G U R E 1 Study profile.
T A B L E 1 Clinicopathological characteristics of the retrospective cohort of esophageal squamous cell carcinoma (ESCC) patients.  ( p = 0.044) were greater in the COL19A1 high group than in the COL19A1 low group (Figure 4c). In the prospective cohort, memory B cells, immature B cells, and activated B cells were also significantly related to COL19A1 expression (Figure 4d). We therefore evaluated the baseline expression levels of CD27 and CD20 in tumor tissues by mIF. As

DISCUSSION
Immunotherapy has gradually become a major focus of cancer therapy research. Recently, anti-PD-1/PD-L1 antibodies have been shown to be effective treatments for recurrent and refractory ESCC patients. 15,28,29 With clinical results and biomarker analysis, neoadjuvant chemoimmunotherapy has become a valid approach for ESCC patient treatment. 30 However, tumor heterogeneity within the TME contributes to cancer progression and failure of immunotherapy. [31][32][33] Thus, identification of patients who may benefit most from ICIs is of clinical significance. Herein, we performed a systematic series of analyses to screen a panel of potential predictive biomarkers for successful treatment response. Subsequently, a "biology-centric" method was used to precisely target candidate biomarkers. Based on transcriptome profiles from the retrospective cohort, intertumoral heterogeneity was identified by unsupervised clustering. Based on those results, 66 ESCC patients were classified into four subtypes, with immune subtype (C3) patients exhibiting better long-term survival. Notably, subtype-specific genes were identified as "sensitive" biomarkers, due to the immune-activation characteristics of the C3 subtype. As a result, six genes were found to be remarkably upregulated and significantly related to clinical response. Among those, the COL19A1 gene is rarely expressed by cancer cells but is highly expressed in amyotrophic lateral sclerosis. 34 Li et al. reported high COL19A1 expression in normal esophageal tissues. 35 Further, Brodsky et al. observed that COL19A1 high contributed to prolonged OS in gastric cancer patients. 36 We found that high COL19A1 expression was associated with a better prognosis for ESCC patients. Similarly, in the neoadjuvant immunotherapy cohort, we observed for the first time that COL19A1 high was related to MPR status. As validated by IHC, COL19A1 was predominantly expressed by ESCC malignant epithelial cells in a membrane-anchoring manner. Further, the COL19A1 protein has been shown to participate in the organogenesis of the esophagus as well as the maintenance of its normal structure. 37 The TME plays a critical role in tumor development and in the response to immunotherapy of a variety of solid tumors including lung cancer, 38 triple-negative breast cancer, and colon cancer. 39 Previous studies, such as that by Xu et al. 40 implied that activated B cells are observed only in response to immunotherapy. In this study, a subset (C3) of ESCC patients in the retrospective cohort was shown to be immune-activated and to be specifically characterized by B cell infiltration. As mentioned above, B cell accumulation was found only in patients who achieved MPR. Recent studies have found an association between enhanced B cell infiltration, apart from tertiary lymphoid structures, and an increased response to immunotherapy for various malignancies. [41][42][43] In addition, this study demonstrated B cell markers to have prognostic value, of which COL19A1, as a B cell receptor related gene, 44 was  ). *, **, ***, and **** represent p < 0.05, p < 0.01, p < 0.001, and p < 0.0001, respectively. most related to prognosis. Our study found COL19A1 expression to be positively related to greater immune cell infiltration (CD8+ T cells, B cells, M1 macrophages, and NK cells), which implies that COL19A1 expression is associated with an inflamed ESCC microenvironment. An inflamed TME suggests pre-existing immune-specific activity and is a predictive marker of a positive response to cancer immunotherapy. 45 Further, an inflamed microenvironment has been related to the efficacy of chemotherapy and immunotherapy. 45,46 Consequently, the predictive significance of a COL19A1 expression pathway is reasonable to presume, in that such a pathway would reflect interaction between the immune response and malignancy. COL19A1 may serve as a gatekeeper within the esophagus in that it putatively interacts with B cells. The mechanism of this interaction is unknown and requires further investigation.
This study has several limitations. First, participants in the prospective cohort received "real-world" treatments that were not uniform, and therefore, our outcomes were inevitably biased. Second, this was a single-center study that included a relatively small cohort. A larger patient population is needed to further validate COL19A1 as a predictive biomarker for neoadjuvant chemotherapy responsiveness.
Finally, the precise mechanistic basis for this study's observations has not been established and requires further intensive investigation.
The key strengths of this study are exploration of the molecular subtypes of ESCC patients, evaluation of patient subtype-specific biological characteristics, and the construction of an immune-related prediction model. This study proposed, for the first time, that upregulated COL19A1 is a biomarker that can predict the response to neoadjuvant immunotherapy by ESCC patients. We also found COL19A1 expression to be associated with an inflamed ESCC microenvironment. Altogether, these findings provide valuable insight into the predictive value of COL19A1 as a means to precisely screen and predict disease outcomes in ESCC patients who may benefit from neoadjuvant immunotherapy.