Co-expression based regulatory network for lens development

Abstract

Purpose: Although there are >100 mouse lens microarray gene expression datasets representing various developmental stages and mutational states, these have not been integrated to comprehend lens biology on the systems level. We hypothesized that systematic processing and analyses of these datasets can yield novel insights into the regulatory circuitry underlying lens development and maintenance of transparency. Therefore we used a co-expression based approach to derive a condition-dependent network that predicts known- as well as novel-regulators and their interactions in mouse lens development.

Methods: Affymetrix Mouse Genome 430 and Illumina MouseWG-6 v2.0 lens microarray datasets were obtained from NCBI-GEO. Additionally, five new mouse lens microarray datasets were generated to extend representation of post-natal stages. We developed a statistical method to identify key players in lens development and cataract formation. Genes with the highest scores were used to identify stage specific regulators by self-organizing tree (SOTA) clustering, and to generate co-expression network using a weighted correlation approach.

Results: Our analysis revealed four major SOTA clusters containing genes that exhibited distinct co-regulatory patterns in the lens. These included known regulators with experimentally validated spatio-temporal expression patterns like Pax6, Six3, Foxe3 and Hmx1 in a TF-rich cluster representing early lens development. Expectedly, crystallins and other fiber proteins were enriched in a cluster representing late-development. Importantly, this analysis identified several uncharacterized candidates for lens development (Sall4, Dmrta2). Interestingly, Sall4 is implicated in formation of sensory placodes and its mutations in human are associated with lens defects. Finally, this analysis places Sall4 in the lens regulatory circuit with Pax6, Six3, Foxe3, Pitx3 and Maf, further supporting its candidacy in lens development.

Conclusions: In past we developed a strategy termed “whole embryo body in silico subtraction” to generate tissue-enriched gene expression profiles and demonstrated its utility to predict genes associated with cataract. Here, we build on identification of lens-enriched genes and derive a co-expression regulatory network that highlights interactions between known regulators in lens and successfully predicts new nodes (Sall4, Dmrta2) for further investigation in early lens development.