Ischemia caused by time to freezing induces systematic microRNA and mRNA responses in cancer tissue

Abstract

Time to freezing tumor tissue for RNA expression analysis will always vary to some extent. To evaluate the effect of ischemia time, tumor tissue from ten breast cancer patients was collected and aliquots of tissue were snap frozen at different time points after surgery (0, 0.5, 1, 3 and 6 h). Using miRNA and mRNA expression microarrays and statistical analysis, 56 miRNAs and 1788 mRNAs were found to be significantly altered with ischemia time up to six hours. Several of the 56 miRNAs have been reported to play a role in cancer, such as hsa-miR-663 and hsa-miR-125a-3p. Known stress response genes such as GADD45B, JUND and FOSB were among the mRNAs most significantly affected by time to freezing. A novel statistical method for identification of consistently correlated miRNA-mRNA pairs and miRNA-associated biological processes in time course data is presented. Application of this method revealed that several miRNAs, including hsa-miR-1228, hsa-miR-1225-5p and hsa-miR-574-5p, were associated through their correlation to mRNAs to biological processes such as "response to stimulus" and "stress response". These miRNAs also showed enrichment of predicted targets among either their positively or negatively correlated mRNAs. The induced miRNAs may play both direct and indirect roles in biological responses. Caution should be taken when the miRNAs and mRNAs reported to be affected by ischemia time are included in a prognostic or predictive signature.

Figure 1

Analytical strategy to identify consistently correlated miRNA–mRNA pairs and associate miRNAs to biological processes. The analytical strategy is illustrated by an example of eight miRNAs, ten mRNAs (genes) and six patients. Top left: Expression profiles over time are shown for one miRNA and two genes for three of the patients. Step 1: A correlation matrix representing each miRNA (here: miR1), with rows representing mRNAs and columns representing patients. Step 2: All correlation coefficients in Step 1 are ranked to make rank matrices for each miRNA. Note that this is also done for the reverse rank, when analyzing consistent negative correlation. Step 3: Each of the rank matrices in Step 2 is used to calculate a p‐value (Navon et al., 2009) of consistent correlation between each miRNA–mRNA pair. Step 4: Summarizing the results by (a) generating a heatmap indicating which miRNA–mRNA pairs that are consistently time‐correlated and (b) using the p‐values from the consistent correlation analysis to analyze if there is enrichment of certain biological processes among the most correlated mRNAs of each miRNA (Eden et al., 2009).

Figure 2

The most significantly differentially expressed miRNA and mRNA transcripts up to six hours in room temperature after surgery. (A) Top ten most significant miRNAs, (B) expression levels for each patient as a function of time for the four most significant miRNAs, (C) top ten most significant mRNAs, (D) expression levels for each patient as a function of time for the four most significant mRNAs (with known function).

Figure 3

Enriched biological processes among the mRNA transcripts affected by ischemia time. A color‐coded trimmed Directed Acyclic Graph (DAG) of all significantly enriched biological process GO‐terms among the mRNA transcripts that are the most significantly affected by ischemia time, obtained using GOrilla (Eden et al., 2009).

Figure 4

mRNA transcripts consistently correlated to hsa‐miR‐125a‐3p. Heatmaps visualizing the correlation between hsa‐miR‐125a‐3p and its most consistently (A) negatively or (B) positively correlated mRNA transcripts. Each column represents a patient, and the rows are sorted according to consistent correlation.

Figure 5

Consistently time‐correlated miRNA‐mRNA pairs in the ten patients. Significant (FDR < 0.05) miRNA–mRNA pairs among the miRNAs and mRNAs associated with ischemia time are indicated in a heatmap. (miRNAs and mRNAs that were involved in less than five pairs (FDR < 0.01) were excluded). Rows and columns are ordered by hierarchical clustering with Euclidean distance and complete linkage. Note that none of the miRNAs show a decreasing expression with time. Viral miRNAs are indicated by red color. Color‐coding of heatmap corresponds to red: positive correlation, FDR < 0.01; orange: positive correlation, FDR < 0.05; green: negative correlation, FDR < 0.05; blue: negative correlation, FDR < 0.01.

Figure 6

miRNA associated biological processes. Biological processes GO‐terms that were significantly (p < 10−7) enriched among mRNAs that were consistently positively correlated to at least three miRNAs are ordered according to median p‐value of the significant miRNAs. The number of miRNAs with significant enrichment for each biological process, boxplots of p‐values for all miRNAs significantly associated with each biological process, and a heatmap of the ten miRNAs with the lowest p‐value for any of the biological processes are shown. Viral miRNAs are indicated by red color.