AGR2 – a Novel Predictor of Neoadjuvant Chemoradiotherapy Response in Esophageal Squamous Cell Carcinoma

Chih-Hung Lin Taichung Veterans General Hospital Han-Ni Chuang Taichung Veterans General Hospital Tzu-Hung Hsiao Taichung Veterans General Hospital V. Bharath Kumar Asia University Chiung-Hung Hsu Hualien Tzu Chi Hospital Buddhist Tzu Chi Medical Foundation Chih-Yang Huang Hualien Tzu Chi Hospital Buddhist Tzu Chi Medical Foundation Li-Wen Lee Taichung Veterans General Hospital Chien-Lin Mao Taichung Veterans General Hospital Jiunn-Liang Ko Chung Shan Medical University Chung-Ping Hsu (  cliffhsu@gmail.com ) Hualien Tzu Chi Medical Center: Hualien Tzu Chi Hospital https://orcid.org/0000-0002-7686-4935


Introduction
Esophageal squamous cell carcinoma (ESCC) is one of the most common human cancers and it is the sixth most common cancer worldwide (1). In Taiwan, it is the fth leading cause of death among men, and most prevalent in South-East and Central Asia. Esophageal cancer is histologically classi ed as squamous cell carcinoma or adenocarcinoma (2). Surgical resection remains the main treatment of earlystage esophageal carcinoma. Although multidisciplinary therapies, including radiotherapy, chemotherapy, and surgery, have been developed to prolong the survival of patients with locally advanced esophageal carcinoma, the prognosis remains poor (3,4). Previous studies have shown that neoadjuvant chemoradiotherapy (nCRT) followed by surgery is a common multidisciplinary treatment for resectable esophageal carcinoma (5)(6)(7)(8)(9)(10)(11)(12). However, it is still disappointing, and > 50% of such patients still exhibit poor response to nCRT (13)(14)(15)(16). As broad tumor pro ling becomes a common component of cancer care, next-generation sequencing (NGS) is increasingly used in many areas of cancer research and in clinical settings. Furthermore, endoscopic biopsies are suitable for targeted NGS, which provides quality sequencing data and accurate mutational information (17)(18)(19). NGS is also a tool that is widely available to gastroenterologists (17)(18)(19). In this study, we aimed to identify potential genes for the prediction of the response to nCRT by using NGS in ESCC patients. Based on function analysis results, we chose AGR2 to perform further investigation. Our results showed that silencing of AGR2 might enhance sensitive to the cytotoxicity effect of cisplatin and 5-uorouracil (5-FU) and may be through up-regulating p21 expression. Collectively, our results showed that, AGR2 might be a potential gene marker for response to nCRT in ESCC. USA) with 1µg of total RNA from all samples following the manufacturer's instructions. The prepared library was sequenced with paired-end runs using Illumina HiSeq 2500 sequencers. RNA reads were mapped to the human reference genome GRCh37 usingHISAT2aligner tools (20), read counts were calculated using feature Counts (21) and gene expression pro les were identi ed using DESeq2 (22). DAVID (the Database for Annotation, Visualization and Integrated Discovery, https://david.ncifcrf.gov/) functional tool was used for functional annotation of differentially expressed genes.
Using siRNA, we knocked down AGR2 gene expression in esophageal cancer cells. The TransIT-X2 Dynamic Delivery System reagent (Mirus Bio) procedure was used to forward-transfect siRNA into the esophageal cancer cells. The esophageal cancer cell cells were plated in 6-well culture plates at a density of 4.0-6.0 ×10 5 cells/well and cultured in 2mL growth medium for 24 h. Cells were transfected with siRNA to a nal concentration of 25nM by using TransIT-X2 transfection reagent. Subsequently, cells were incubated with 5% CO 2 at 37°C for 72 h. Finally, cells were harvested and assayed for the knockdown of target gene expression.

Reverse transcription and quantitative polymerase chain reaction
The total RNA was extracted using the AllPrep DNA/RNA Mini Kit followed the manufacturer's protocol (cat. 80204). Reverse transcription proceeded according to the SS4 (Invitrogen) protocol. Quantitative reverse-transcription polymerase chain reaction employed the Fast Start TaqMan Probes system (Roche, Cat.4913947001) and AGR2 speci c primers, the analysis was performed on a StepOne Plus Real-Time PCR System (Applied Biosystems, Foster, CA). Glyceraldehyde 3-phosphate dehydrogenase was used as an endogenous control, to determine the relative expression levels of target genes using the method.

Reagents
Cisplatin (P4394) and 5-uorouracil (F6627) were purchased from Sigma-Aldrich (Merk). Cisplatin was dissolved in double-distilled water. 5-FU was dissolved in dimethyl sulfoxide (DMSO). The solvent was also used in the control group of the experiment.

MTT assay
To determine the cytotoxicity of the combined effect of AGR2-knock down and chemotherapeutic agents, cells were rst plated in 24-well culture plates for 24 h. Then, cells were transfected with siRNA. After 24 h, we treated the cells with cisplatin (2.0-6.0µM) and 5-FU (3.0-20.0µM) for 72h. Cell viability was evaluated using MTT assay followed by our previous study (23). The medium was removed and the cells were washed twice with 1× phosphate-buffer saline (PBS). Then, 500 µL MTT solutions (1mg/mL) were added to cells, and the solution was incubated at 37°Cfor 30 min. The MTT solution was removed and replaced with 200 µL DMSO, subsequently, the cells were incubated for 5 min to dissolve completely. We transferred 100 µL DMSO of dissolved cells into 96-well enzyme-linked immunosorbent assay (ELISA) plates to measure the absorbance a t 570/670 nm by using ELISA reader. Each experimental data point represents the average value obtained from three replicates.

Western blotting
We employed western blot to determine the levels of AGR2 and associated proteins. Cells were washed with PBS and lysed in RIPA buffer (APOLO) supplemented with protease inhibitor cocktail. Protein concentrations were detected using a protein assay kit (Bio-Rad). Equal amounts of proteins (30µg) were subjected to sodium dodecyl sulfate 8% -12 % polyacrylamide gel electrophoresis. Fractionated proteins were transferred to Hybond-P membranes (Millipore). Membranes were blocked with PBS containing 5% nonfat milk and 0.2% Tween 20. For the detection of human anti-AGR2 (Invitrogen), anti-p21 (Proteintech), and anti β-actin (Sigma) the membranes were incubated overnight at 4°C, followed by the addition of anti-mouse or anti-rabbit IgG Horseradish peroxide -linked antibody (Jackson). Blots were then developed using an enhanced chemiluminescence reagent (Millipore) (24).

Statistical analysis
Statistical analysis of all results was performed using paired two-way analysis of variance with Tukey's test. All results re ect the mean ± standard error of the mean of data obtained from at least three independent experiments. Statistical signi cance was de ned as p < 0.05.

Clinical characteristic
The clinicopathological characteristics of ESCC patients (n = 32) is summarized in Table 1. The standard protocols for operable esophageal cancer patients were brie y described as followings. Chemotherapy was given concomitantly with cisplatin 20 mg/ml iv for 1 hour and uorouracil 800 mg/ml iv for 24 hours daily( day 1 to day 4 (cycle1), and (day 29 to day 32 (cycle 2) of radiotherapy. Radiotherapy was performed 5 days per week, with a daily dose of 180 Gy for a total course of 5-6 weeks. Surgery was performed 4-6 weeks after complete nCRT. The surgical procedure included thoraco-scopic esophagectomy, at least 2-eld lymph node dissection and esophagus reconstruction with gastric tube.
The patients' average age was 59.9 years (range 48-82 years). Patients were divided into two groups according to their response to nCRT; Complete response group (n = 13) and Non-complete response group (n = 19) ( Table 2). Four patients in the Non-complete response group and two patients in the complete response group did not undergo operations after CRT. The response status of these patients was con rmed by clinical evaluation and endoscopic biopsies. 3.2. RNA expression differences between Complete response and Non-complete response groups of patients with ESCC Sequencing reads were mapped against the human genome assembly (Ensembl Build 37) using Tophat (v2.1.1). We identi ed 464 differentially expressed genes (fold change > 2 or < 2, and a DESeq p value of < 0.05), in which 240 genes were up-regulated and 224 were down-regulated among them. The values of fold change among top 20 up-regulated and top 20 down-regulated genes are shown in Table 3 and   Supplemental Table 1, respectively. Unsupervised hierarchical clustering of 20 up-regulated and 20 downregulated genes also revealed the differences in Complete response or Non-complete of nCRT (Fig. 1).The resulting heat map and HCL dendrogram showed a similar separation of the two nCRT types.
The Database for Annotation, Visualization, and Integrated Discovery (DAVID) was used to explore the potential functions of the differentially expressed genes and its biological function (25,26). Up-regulated genes in Non-complete response group was associated with cellular protein metabolic process, glucose homeostasis, TGF-beta receptor signal response pathway, cholesterol homeostasis, cell differentiation and drug response (Fig. 2). We rstly assessed gene expression in up-regulated genes between Complete response and Non-complete response, and further selected four genes that are associated with cell proliferation and cell migration, namely AGR2 (27), PPP1R15A (28), GADD45B (29), and LRG1 (30). RNA sequencing results showed that, these four genes were signi cantly up-regulated in patients with Noncomplete response group compared to complete response group (Fig. 3).

Knockdown of AGR2 in esophageal cancer cells is more sensitive to the cytotoxicity effect of cisplatin and 5-uorouracil
Previous studies have indicated that AGR2 is associated with head and neck squamous cell carcinoma (HNSCC) by regulating cell transformation and epithelial mesenchymal transition (EMT) signaling pathways (31), in esophageal adenocarcinoma it promotes tumor growth (32). However, in this study we found that the AGR2 was up-regulated in the patients with Non-complete response before nCRT, thus we selected AGR2 for further investigation.
We then applied siRNA approach to knock down AGR2 expression in esophageal cancer cell lines (CE146T/VGH, TE2, and CE48T/VGH) and then performed MTT assay. Western blot analysis showed that the protein level of AGR2 was signi cantly reduced in AGR2 siRNA transfected CE146T/VGH, TE2, and CE48T/VGH cells. (Supplementary Fig. 1). MTT assay indicated that cell viability of the AGR2-knockdown TE2 cell line decreased signi cantly following treatment with 2.5µM cisplatin and 3µM 5-FU, whereas CE48T/VGH and CE146T/VGH cells showed no any change (Fig. 4). Whereas in AGR2-transfected cells (CE48T/VGH, CE146T/VGH and TE-2) increased concentration of (6µM) cisplatin and 20µM 5-FU signi cantly decreased cell viability compared to control group (Fig. 4). All these data indicated that AGR2 down-regulated cells were more sensitive to cisplatin and 5-FU treatment.

Cisplatin and 5-FU induced mortality in AGR2-knockdown esophageal cancer cells by up-regulating p21 expression
Esophageal cancer cells were transfected with AGR2-siRNA for 24 h and then exposed to cisplatin and 5-FU for 48 h. InAGR2-knockdown cells, p21 protein level was increased in compared to control group (Fig. 5). These results suggest that the cytotoxicity sensitivity to cisplatin and 5-FU in esophageal cancer was associated with AGR2 expression and might be through the p21 pathway.
DiMaioet al. retrospectively examined 116 specimens of esophageal carcinoma; they demonstrated that the presence of diffused AGR2 expression is highly sensitive to esophageal adenocarcinoma. However, focal expression of AGR2 was found in only approximately one-third (36.59%) of ESCC specimens (42). In this study, we rstly attempted to identify biomarker to predict nCRT responses in esophageal cancer patients (n = 32) by next-generation sequencing (NGS). We found 464 genes were differentially expressed in ESCC that were associated with nCRT response. We then assessed gene expression in up-regulated genes between Complete response and Non-complete response group, and further selected four genes that are associated with cell proliferation and cell migration, namely AGR2 (27), PPP1R15A (28), GADD45B (29), and LRG1 (30). Based on the gene fold change differences, we scrutinized AGR2 with 2.8 fold difference in the patients with Non-complete response than treatment group. Thus we decided to use AGR2 to do further investigation. AGR2 expression was signi cantly correlated with metastasis and poor prognosis in breast cancer and biomarker in prostate cancer (43)(44)(45).
Previous study showed that, AGR2 induces cancer cell proliferation, invasion and survival, chemotherapy resistance, metastasis and tumor growth. Therefore in this present study at rst we knock-downed AGR2 expression in esophageal cancer cell lines (CE146T/VGH, TE2, and CE48T/VGH cells) and then analyzed for AGR2 expression using Western blotting. AGR2 protein expression was greatly decreased in all the three si-AGR2 transfected cancer cells. To further correlate the role of AGR2 with nCRT response, we knockdown AGR2 expression and treated cells with cisplatin and 5-FU and analysed for its cell viability. Cancer cells -CE48T/VGH and CE146T/VGH treated with low concentration of 2.5µM cisplatin and 3µM 5-FU showed no any change in the cell viability percentage compared with TE2 cell lines. This may be due to nature of cell type, contrarily when the treatment dose was increased all the three cells showed signi cant decrease in cell viability compared to control group. This data clearly explains us the signi cant role of AGR2 down-regulation in cisplatin and 5-FU treatment in esophageal cancer cells.
Previous study showed that AGR2 expression and cell cycle proteins showed as good prognosis marker in epithelial ovarian cancer (46), in this study in AGR2-knockdown cells, cisplatin and 5-FU treatment increased p21 protein level compared to control group. These results suggest that the cytotoxicity sensitivity to cisplatin and 5-FU in esophageal cancer was associated with AGR2 expression and might be through the p21 pathway. This was a retrospective study; further prospective studies are required to verify the correlation and precise pathway. Although many studies have con rmed that endoscopic biopsy results in a high concordance between biopsy and resected specimens (18, 47, 48), the bias of gastroscopic biopsy remains a concern.

Conclusions
Using specimen from gastroscopic biopsy and NGS, we obtained su cient specimens con rm the overexpression of genes in patients with ESCC. The AGR2 gene might be one of the predictors of poor response to nCRT. Sensitivity to cytotoxicity may be associated with p21 expression. These ndings might inform alternative personalized therapeutic plans or new directions for targeted therapy in the future. Funding: This study was supported from Taichung Veterans General Hospital (TCVGH-1074703D and TCVGH-1094704D).

Abbreviations
Availability of data and material: The data and material of this study is available for publishing in public.
Ethics approval and consent to participate: Both the data collection and gene expression analysis of tumor tissues were approved by the Institutional Review Board of Taichung Veterans General Hospital (IRB TCVGH No: CE17279A). Consent for publication: Written informed consent was obtained from all participants.
Con icts of interest/Competing interests: The authors have declared that they have no con icts of interest.

Figure 1
Hierachical cluster analysis (HCL) of 20 proteins identi ed in patients with ESCC before neoadjuvant chemoradiotherapy. HCL of the 20 differentially expressed genes identi ed between groups with Complete and Non-complete response to ESCC treatment was performed via unsupervised hierarchical classi cation and distance trees were constructed from all genes. High expression of genes is highlighted in red and low expression in blue; different levels of shade indicate the extent the expression level difference from the median. Up-regulated gene expression pattern according to functional analysis. The Database for Annotation, Visualization, and Integrated Discovery was used to predict the potential functions of the differentially upregulated expressed genes. Clusters with p < 0,05 are presented.