Association of inflammation-related and microRNA gene expression with cancer-specific mortality of colon adenocarcinoma

Abstract

Purpose: Inflammatory genes and microRNAs have roles in colon carcinogenesis; therefore, they may provide useful biomarkers for colon cancer. This study examines the potential clinical utility of an inflammatory gene expression signature as a prognostic biomarker for colon cancer in addition to previously examined miR-21 expression.

Experimental design: Quantitative reverse transcriptase-PCR. was used to measure the expression of 23 inflammatory genes in colon adenocarcinomas and adjacent noncancerous tissues from 196 patients. These data were used to develop models for cancer-specific mortality on a training cohort (n = 57), and this model was tested in both a test (n = 56) and a validation (n = 83) cohort. Expression data for miR-21 were available for these patients and were compared and combined with inflammatory gene expression.

Results: PRG1, IL-10, CD68, IL-23a, and IL-12a expression in noncancerous tissue, and PRG1, ANXA1, IL-23a, IL-17a, FOXP3, and HLA-DRA expression in tumor tissues were associated with poor prognosis based on Cox regression (/Z-score/ >1.5) and were used to generate the inflammatory risk score (IRS). IRS was associated with cancer-specific mortality in the training, test (P = 0.01), and validation (P = 0.02) cohorts. This association was strong for stage II cases (P = 0.002). Expression of miR-21 was associated with IL-6, IL-8, IL-10, IL-12a, and NOS2a, providing evidence that the function of this microRNA and these inflammatory genes are linked. Both IRS and miR-21 expression were independently associated with cancer-specific mortality, including stage II patients alone.

Conclusion: IRS and miR-21 expression are independent predictors of colon cancer prognosis and may provide a clinically useful tool to identify high-risk patients.

Figure 1

Inflammatory genes are consistently altered in colon tumors from both cohorts. a) Unsupervised hierarchical clustering (correlation, average linkage) using 21 inflammatory genes on 113 pairs of cancerous and noncancerous tissues in the Hong Kong Cohort. Clustering separates tissues into two distinct groups; one composed of 97% tumor tissues and one composed of 99% noncancerous tissue. b) Correlation of the tumor/noncancerous tissue expression ratio comparing Hong Kong Cohort with NCI-Maryland cohort indicates consistent changes in inflammatory gene expression in both cohorts.

Figure 2

Strategy for building inflammatory risk scores. Genes were selected to be included in the risk score based on univariate Cox regression on the Hong Kong Training Cohort. Multivariate Cox regression on the training cohort was used to build the risk models. This model was then tested on the Hong Kong test and Maryland validation cohorts.

Figure 3

Inflammatory risk score (IRS) is associated with cancer specific mortality in TNM stage II patients. (a) IRS in Hong Kong training, Hong Kong test and NCI-Maryland validation cohorts separately. (b) Combined analysis of Hong Kong test and NCI-Maryland validation cohorts, stratified by TNM stage. For 1 patient in the Maryland Cohort, it was unclear if that patient had stage III or stage IV colon cancer and therefore is removed from analyses stratifying by TNM stage.

Figure 4

Combined Inflammatory Risk Score (IRS) and miR-21 expression predict colon cancer specific mortality better than either alone. All patients from the Hong Kong test and NCI-Maryland validation cohorts are analyzed. Left is stratified by miR-21 expression; middle is stratified by IRS and miR-21 expression; right is combined IRS and miR-21 expression. miR-21 expression data was available for 115 of 117 patients and only those patients were used for this analysis. The Hong Kong training cohort is also excluded from this analysis.