We have generated a comprehensive map of retina transcriptome in mouse using high throughput, paired-end RNA-seq. We are making these data publically available and anticipate this could be a highly accessed resource for researchers interested in gene expression in the neural retina. There have been several transcriptome studies in retina, all were done by using previous technologies such as microarray and serial analysis of gene expression (SAGE) [8
]. Dorrell et al. reported 2635 known genes in a microarray study on developing mouse retina [26
]. In this study, postnatal mouse retina development was studied including the postnatal day 21 which is the same sampling day as our study. Using SAGE, Blackshaw et al. reported gene expression during retinal development, but focused on genes expressed by mammalian rods [8
]. A microarray study done by Livesey et al. focused on transcriptional network controlled by Crx
using cDNA microarray [27
]. Zhang et al. also used cDNA microarray to study the mouse retinal development from embryonic day 12.5 to postnatal day 21 [28
]. In the present study, we used RNA-seq to accurately identify transcript abundances (with a broad dynamic range) and size and alternative splicing information in mouse retina at postnatal day 21. Our study has revealed several novel features of the retina transcriptome, including in particular the architecture of known retinal disease genes. We identified 15251 known genes expressed concordantly with precise alternative splicing information and 3655 of those were found to have more than 2 isoforms.
The neural retina has a large number of known disease genes which act in a simple Mendelian fashion (greater than 150 genes) which lead to visual impairment and blindness (http://www.sph.uth.tmc.edu/retnet/sum-dis.htm
). Few other tissues have such a well-characterized and highly specific signature of disease genes. These attributes of the neural retina make this tissue an excellent model to study the transcriptional architecture of neurogenetic disease by genome-wide RNA-seq. Here we have discovered a number of features of the transcriptional architecture of disease genes including disease genes which are among the most abundantly expressed, and are generally larger with more alternative transcripts than non-disease genes. It is possible that the high level of expression of disease genes may be a feature of the retina, but we propose that these attributes may alternatively be fundamental properties of disease genes. One possible interpretation with regard to abundance may be that disease genes are commonly expressed in rod photoreceptors and these cells are among the most abundant cell types in the murine neural retina, representing greater than 70% of cells (Lavail 1980; Morrow 1998). While this observation may explain part of the high abundance of disease genes, the increase in alternative splicing and the larger size of disease-related genes seem likely independent characteristics of the disease gene transcriptional architecture. More work will be necessary in other neural tissues to examine this result in greater detail.
Our data provide interesting findings on alternative splicing in mouse retina. Alternative splicing occurs in 3,655 genes (24%) of the genes identified in both libraries with a large number of genes demonstrating over two transcripts. Our findings are highly consistent with prior estimates of alternatively splicing in human retina where alternative splicing has been estimated to be in 26% of retina genes in a study based on expressed sequence tags (ESTs) [29
]. Alternative splicing events in our study showed high concordance in terms of gene abundances (R2
=0.85, data not shown) yet we anticipate that our results may represent an underestimate of alternative splicing given our strict criteria for identifying transcripts in our biologic replicates. However, we have provided the full analysis of the conservative list of transcripts (discovered in all replicates) and also all transcripts from individual replicates in Supplementary data
Finally, genome-wide profiling of retinal transcriptome has revealed other interesting features of retina biology. For example, we noted that of all the synaptic vesicle genes in the genome, only 30% of such genes are expressed in the retina; that is, the retina uses a very specific signature of synaptic vesicle genes. Further work in different neural tissues will likely elucidate distinct signatures which may suggest specific physiologic properties of neurons in different neuronal circuits. In addition, the observation that neuronal development genes show persistent expression in adult tissues is also compelling. This argues that a majority of such genes may have various functions including in neuronal development and potentially in the maintenance of neuronal health.
In summary, this study presents the most comprehensive view of the transcriptome of the murine neural retina to date using novel, massively-parallel sequencing technologies. Our analysis uses the state-of-the-art tools for calculations of transcript abundance, size and alternative splicing. These data are provided as a resource for the community of researchers studying gene expression in retina.