Following bisulphite sequencing, raw reads were first cleaned with SOAPnuke132 (version 2.0.5) to remove residual adaptor sequences and reads with low-quality scores. Cleaned reads were mapped to the reference genome and duplicated reads were removed using Bismark133 (version 0.20.1). The depth and coverage on chromosomes were calculated with samtools85 (version 1.4) and bedtools134 (version 2.26.0). The methylation call for every cytosine was evaluated by Bismark and the methylation ratio was calculated as the number of reads supporting methylated Cs divided by the total unique reads covering the cytosine position (Supplementary Data 8).
In summary, our RNA-seq analyses identified graded expression changes of shared gene sets in seven available mouse models for CRX-associated disease. Using the data presented and referenced in this paper [2, 21, 24], these models can be ranked as illustrated in Fig. 7, with the lightest bars representing the model with most severely affected rod and cone gene expression. This order correlates with phenotype severity by morphological and electrophysiological standards, and with changes in gene expression in rods and cones. This correlation was seen not only for down-regulated rod and cone genes encoding phototransduction components, but also for those up- and down-regulated gene sets that highlight the partial rod to cone conversion of the developing photoreceptors. The schematic also predicts the level of photoreceptor identity and function in various models. Most importantly, the different gene expression changes between some phenotypically distinct Crx mutant models (such as E168d2/+ versus E168d2neo/+) were rather modest, in contrast to their substantial impact on the disease phenotype. 2b1af7f3a8