Factorial Microarray Analysis of Zebrafish Retinal Development

Retinal cells are specified in a zebrafish recessive mutant called young (yng) but they fail to terminally differentiate; i.e. extend neurites and make synaptic contacts. A point mutation in a brahma-related gene 1 (brg1) is responsible for this phenotype. In this microarray study, a three-factor factorial design was utilized to investigate the effects of 1) mutation, 2) change in time (36 vs. 52hpf), and 3) change in tissue (retina vs. whole embryos), and their interactions on gene expression. Significant probesets were inferred by using both specific contrasts of the fitted Analysis of Variance (ANOVA) models and a corresponding 2-fold expression cutoff. The probesets were grouped into three broad categories: 1) Brg1-regulated retinal differentiation genes (731 probsets), 2) Retinal specific genes but independent of Brg1 regulation (3038 probesets) and 3) Genes regulated by Brg1 but outside the retina (107 probesets). Four gene groups/pathways including neurite outgrowth regulators, Delta-Notch signalling molecules, Irx family members and specific cell cycle regulators were identified in the first group, and their relevance for retinal differentiation functionally validated. This study demonstrates that an approach such as ours can identify relevant genes and pathways involved in retinal development as well as the development of other tissues at the same time. Keywords: Development, retina, terminal differentiation, chromatin remodeling, factorial anlaysis The gene expression levels of three independent replicates of retina consisting of ten samples each at 36 and 52hpf retinas from WT (WR36 & WR52) and yng (YR36 & YR52) larvae, and stage-matched whole embryos: WT at 36hpf (WA36), WT at 52hpf (WA52), yng at 36hpf (YA36), and yng at 52hpf (YA52) were measured by Affymetrix Zebrafish Whole Genome Arrays. We analyzed the effects of these factors on gene expression levels: 1) mutation (M), 2) tissue (R), and 3) time (T). Each of these factors have two levels: mutant and WT (M); retinas and whole body (R); and 36 and 52 hpf (T). As a result, there are totally eight conditions for all the comparisons and the design is called a 2x2x2 factorial design. The aim of a factorial analysis is to delineate the effects of each factor and their combinations (interaction) on gene expression. The overall analysis strategy is to first fit a full Analysis of Variance (ANOVA) model with all possible main effects (T, M, R) and their interactions (T*M, M*R, R*M, T*M*R) to the gene expression data for each probeset. The estimates of the model are obtained through a maximum likelihood estimation method. Then a backward elimination strategy is used to remove insignificant main effects and interactions to get the most parsimonious models. Finally, a group of contrasts and their corresponding fold changes are used to infer whether that probeset is significantly associated with a particular biological process.