Locally advanced rectal cancer transcriptomic-based secretome analysis reveals novel biomarkers useful to identify patients according to neoadjuvant chemoradiotherapy response

Abstract

Most patients with locally advanced rectal cancer (LARC) present incomplete pathological response (pIR) to neoadjuvant chemoradiotherapy (nCRT). Despite the efforts to predict treatment response using tumor-molecular features, as differentially expressed genes, no molecule has proved to be a strong biomarker. The tumor secretome analysis is a promising strategy for biomarkers identification, which can be assessed using transcriptomic data. We performed transcriptomic-based secretome analysis to select potentially secreted proteins using an in silico approach. The tumor expression profile of 28 LARC biopsies collected before nCRT was compared with normal rectal tissues (NT). The expression profile showed no significant differences between complete (pCR) and incomplete responders to nCRT. Genes with increased expression (pCR = 106 and pIR = 357) were used for secretome analysis based on public databases (Vesiclepedia, Human Cancer Secretome, and Plasma Proteome). Seventeen potentially secreted candidates (pCR = 1, pIR = 13 and 3 in both groups) were further investigated in two independent datasets (TCGA and GSE68204) confirming their over-expression in LARC and association with nCRT response (GSE68204). The expression of circulating amphiregulin and cMET proteins was confirmed in serum from 14 LARC patients. Future studies in liquid biopsies could confirm the utility of these proteins for personalized treatment in LARC patients.

Figure 1

Unsupervised hierarchical clustering analysis of 28 locally advanced rectal carcinomas (LARC) based on the expression values of probes with standard deviation > 0.2 compared with clinical and histopathological features of each patient (bars below in the heatmap). The pink and blue bars represent two clusters (1 and 2, respectively), with no clear differences among gender, clinical stage (cStage), response to therapy, and pathological lymph node status after treatment (ypN). pCR: Pathological Complete Response, TRG: Tumor Regression Grade.

Figure 2

Strategies used for the identification of potentially secreted proteins that could be used as biomarkers of response to neoadjuvant chemoradiotherapy (nCRT) in locally advanced rectal cancer patients (LARC). (A) The discovery set was composed of five normal rectal tissues (NT, autopsies) and LARC biopsies from 11 patients with pathological complete response (pCR) and 17 with incomplete response (pIR) to nCRT. The over-expressed genes were used for in silico analysis of secreted proteins with the prediction tools available in the CBS server (TargetP, SignalP, and SecretomeP predictions). Several databases (Human Cancer Secretome Database - HCS, Vesiclepedia/Exocarta) were investigated to verify the presence of the potential biomarkers in the secretome of colorectal cancer samples from previously published studies. Plasma Proteome database confirmed the presence of the candidates in the blood circulation. The Venn diagrams illustrate the proteins identified in pCR and pIR using all databases. Four proteins potentially secreted by pCR cases (ERBB3, MMP1, XPO1, and WNT5A) and 16 by pIR (AREG, BACE2, CD44, CD47, CEMIP, CXCL3, DPEP1, ERBB3, GDF15, LIF, MET, MMP1, PDCD5, PHF6, UB2C, and XPO1) were identified in all consulted secretome databases. (B) Gene expression data from TCGA-READ and GSE68204 were used for in silico validation. Similar analysis used in our dataset was applied for the validation cohort. Among the biomarkers found in the discovery analysis, CD47 was not detected among the differentially expressed genes in the TCGA cohort. ANT: adjacent normal tissue; LARC: locally advanced rectal cancer; TCGA-READ: The Cancer Genome Atlas-Rectum Adenocarcinoma; FC: Fold Change; FDR: False Discovery Rate; *pCR potentially secreted biomarker.