Use of multiple time points to model parotid differentiation

Abstract

In order to understand the process of terminal differentiation in salivary acinar cells, mRNA and microRNA expression was measured across the month long process of differentiation in the parotid gland of the rat. Acinar cells were isolated at either nine time points (mRNA) or four time points (microRNA) in triplicate using laser capture microdissection (LCM). One of the values of this dataset comes from the high quality RNA (RIN > 7) that was used in this study, which can be prohibitively difficult to obtain from such an RNaseI-rich tissue. Global mRNA expression was measured by rat genome microarray hybridization (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE65586), and expression of microRNAs by qPCR array (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE65324). Comparing expression at different ages, 2656 mRNAs and 64 microRNAs were identified as differentially expressed. Because mRNA expression was sampled at many time points, clustering and regression analysis were able to identify dynamic expression patterns that had not been implicated in acinar differentiation before. Integration of the two datasets allowed the identification of microRNA target genes, and a gene regulatory network. Bioinformatics R code and additional details of experimental methods and data analysis are provided.

Keywords: Differentiation; MicroRNAs; Microarray; Network analysis; Salivary gland.

Fig. 1

Raw and normalized expression measurements. (A) Box plots generated from raw and RMA normalized microarray measurements. X-axis: individual replicate samples grouped by age; Y-axis: Log₂ intensity. Independent samples of the same age group are color coded. (B) Box plots generated from raw and median normalized CT values from a qPCR array measuring microRNA expression. X-axis: individual replicate samples grouped by age; Y-axis: Log₂ intensity. Samples of the same age group are color coded.

Fig. 2

Hierarchical clustering of mRNA and microRNA expression across acinar differentiation. (A) Heatmap of RMA normalized mRNA expression of transcription factors (538) in the filtered dataset from rat genome microarrays. Columns: Average expression of triplicates at that age. Rows: Affymetrix probe sets (not listed). Samples were clustered in R, and gene expression was row normalized. (B) Heatmap of differentially expressed microRNAs (64). Columns: individual samples. Rows: microRNAs. − ΔCT values of median normalized and processed expression were used to cluster samples in R, gene expression was row normalized.

Fig. 3

Expression patterns of differentially expressed mRNAs. The 2569 differentially expressed mRNAs identified by ANOVA were clustered by k-means clustering using the per-gene scaled expression data. Eight clusters were identified. For each cluster, average expression at each age is plotted in red, and each individual microRNA is plotted in gray. Y-axis: scaled Log₂ intensity (gene expression for each gene was scaled to mean = 0 and stdev = 1).

Fig. 4

Regression analysis and quadratic trends in mRNA expression. Regression analysis was used to identify mRNAs with a significant quadratic trend. These genes were then clustered. Eight clusters were identified. For each cluster, average expression at each age is plotted in red, along with two example genes that are members of the cluster. Y-axis: scaled Log₂ Intensity (gene expression was scaled to mean = 0 and stdev = 1 within each gene).

Fig. 5

Regression analysis and linear trends in microRNA expression. Regression analysis was used to identify microRNAs with a significant linear trend. These genes were then clustered into six clusters. For each cluster, average expression is plotted in red, and each individual microRNA is plotted in gray. This emphasizes patterns of progressive increases or decreases. Y-axis: scaled Log₂ intensity (gene expression for each gene was scaled to mean = 0 and stdev = 1).

Fig. 6

Gene regulatory network during parotid acinar cell differentiation. Regulatory network derived from the network analysis of differentially expressed mRNAs and microRNAs. Edges between each node are either activating (green arrow) or repressing (red lines). Beside each node, a small graph depicts its relative expression across the time points. The network can be divided into three broad arms, on the right a pro-stemness arm, in the middle a genetic switch that reduces repression of Xbp1, and a pro-differentiation arm on the left.