Survival of Patients with Gastrointestinal Cancers Can Be Predicted by a Surrogate microRNA Signature for Cancer Stem-like Cells Marked by DCLK1 Kinase

Abstract

Doublecortin-like kinase 1 (DCLK1) is a gastrointestinal (GI) tuft cell kinase that has been investigated as a biomarker of cancer stem-like cells in colon and pancreatic cancers. However, its utility as a biomarker may be limited in principle by signal instability and dilution in heterogeneous tumors, where the proliferation of diverse tumor cell lineages obscures the direct measurement of DCLK1 activity. To address this issue, we explored the definition of a miRNA signature as a surrogate biomarker for DCLK1 in cancer stem-like cells. Utilizing RNA/miRNA-sequencing datasets from the Cancer Genome Atlas, we identified a surrogate 15-miRNA expression signature for DCLK1 activity across several GI cancers, including colon, pancreatic, and stomach cancers. Notably, Cox regression and Kaplan-Meier analysis demonstrated that this signature could predict the survival of patients with these cancers. Moreover, we identified patient subgroups that predicted the clinical utility of this DCLK1 surrogate biomarker. Our findings greatly strengthen the clinical significance for DCLK1 expression across GI cancers. Further, they provide an initial guidepost toward the development of improved prognostic biomarkers or companion biomarkers for DCLK1-targeted therapies to eradicate cancer stem-like cells in these malignancies. Cancer Res; 76(14); 4090-9. ©2016 AACR.

Figure 1. A 15-microRNA signature correlates to tumor stem cell marker DCLK1 across five gastrointestinal tumor types

A. DCLK1 is strongly correlated to epithelial-mesenchymal transition in all TCGA gastrointestinal cancer datasets, especially those originating in organs with tuft cells present in normal tissue. B. Circos schematic of expression of DCLK1-activity correlated miRNAs with chromosomal location. From inner to outer concentric circles: COAD, ESCA, PAAD, READ, and STAD miRNA expression correlations; representation of HG18 chromosome cytobands; combined miRNA expression correlations for all 5 cancers; consensus significant miRNA expression correlations with labels.

Figure 2. Dysregulated expression of the 15-microRNA signature is present in tumors expressing high levels of DCLK1

Heatmaps demonstrating dysregulated expression of the 15-miRNA signature between DCLK1-low and DCLK1-high patients from the TCGA COAD, ESCA, PAAD, READ, and STAD datasets.

Figure 3. DCLK1 directly regulates expression of the 15-microRNA signature

Boxplots demonstrating increased expression of DCLK1 (A) and increased EMT status (B) between miRNA-signature low (miR Low) and miRNA-signature high (miR high) tumors in all 5 tuft cell-containing GI cancers from the TCGA COAD, ESCA, PAAD, READ, and STAD datasets (p<0.0001 for all comparisons). C–D. SW480 cells express high levels of DCLK1. Downregulation of DCLK1 in this cell line via DCLK1-targeted siRNA results in upregulation of miR-141, miR-200a-b, miR-425, and miR-532 while overexpression of DCLK1 in the AsPC-1 cell line which expresses nearly undetectable levels of DCLK1 results in downregulation of these same markers (E–F)(*p<0.05).

Figure 4. The 15-microRNA signature demonstrates enrichment of key oncogenic pathways and predicts survival in colon, pancreatic, and stomach cancer

A. Heatmap of significant pathways induced by the miRNA-signature as determined by KEGG pathway enrichment analysis using DIANA miRPath including colorectal, pancreatic, and renal cell cancer pathways – all cancers in which DCLK1 activity has been demonstrated to have significant functional activity. B. Kaplan-Meier survival analysis demonstrating that a DCLK1-activity based 15-miRNA signature predicts overall (p<0.05) and recurrence-free survival (p<0.005) in colon cancer patients, C. predicts overall (p<0.05) survival in pancreatic cancer patients, and D. predicts overall (p<0.05) and recurrence-free survival (p<0.01) in stomach cancer patients (green: low miRNA-signature expression, orange: mid miRNA-signature expression, red: high miRNA-signature expression).

Figure 5. Subgroup analysis highlights key clinical attributes related to the prognostic significance of the 15-microRNA signature

A. Subgroup analysis of overall survival in low risk compared to high-risk miRNA-signature tumors in TCGA COAD, PAAD, and STAD datasets demonstrating the prognostic value of the signature in certain subsets of patients. B. Kaplan-Meier analysis of select patient subsets demonstrating the prognostic significance of the signature in Stage I–II colon cancer patients (p<0.01), pancreatic cancer patients under 65 years old (OS: p<0.02; RFS: p<0.008), and stomach cancer patients receiving radiation therapy (Log-Rank: p=0.078; Gehan-Breslow: p=0.022).

Figure 6. The 15-microRNA signature delineates retrospective survival of colon, pancreatic, and stomach cancer patients at 18 months, 3 years, and 5 years

A. Comparison of observed survival by miRNA signature expression among patients with definite outcomes in the TCGA colon, pancreas, and stomach cancer datasets. B. Predicted prognostic receiver operating characteristic (R.O.C) data for colon (COAD), pancreas (PAAD), and stomach (STAD) cancer datasets as well as relevant subgroups as modeled by the prognosticROC statistical package and observed survival in patients with known outcomes at 18 months, 3 years, and 5 years post-diagnosis.

Figure 7. DCLK1 gene expression predicts survival in colon cancer

A. Kaplan-Meier survival analysis demonstrating that colon cancer patients with high DCLK1 gene expression (top 25^th percentile) have significantly reduced overall (p<0.05, HR: 2.214) and B. recurrence-free (p<0.05, HR: 2.433) survival compared to those with low expression (bottom 25^th percentile).