Figure 1.
The transcription network controlled by Crx, Nrl, and Nr2e3.
A, Venn diagram summarizing all genes dysregulated under high stringency criteria. A complete list of dysregulated genes is given in Table S2. The identity of all genes at points of intersection (including additional intersections not depicted here) are given in Table S7. B, Summary of rod-specific and cone-specific ISH patterns relative to a hematoxylin-eosin (H&E) stained section of retina. The scleral side of the retina is oriented up in all subsequent images. ONL = outer nuclear layer; INL = inner nuclear layer; GCL = ganglion cell layer. C-F, ISH with the indicated probes on four mutant and two wild-type backgrounds. Size bar (in C) = 100 µm G, Summary of transcription factors dysregulated in Crx-/-, Nrl-/-, and/or Nr2e3-/- under high stringency criteria. Green arrows indicate activation and red indicate repression. H and J, ISH with Crx and Prdm1 probes, respectively, on wild-type (CD-1) retinas at the indicated timepoints. J-L, ISH with the indicated probes on four mutant and two wild-type backgrounds. Note that all ISHs were performed on retinas from 6-9 week old animals with the exception of those on Crx-/- which derive from 4-5 week old animals.
Figure 2.
Retinal disease genes in the photoreceptor network.
This figure is a manually curated list of all retinal disease genes which are known or are likely to have a photoreceptor-enriched expression pattern. Those gene symbols which differ between mouse and human are marked with a single asterisk, and the human equivalents are given here: Cngb1 = CNGB1; Gpr98 = USH2C; Grk1 = RHOK; Gucy2e = GUCY2D; Prom1 = PROML1; Rgs9bp = R9AP; Rp1h = RP1; Ush1g = SANS. ‘Known CRE’ indicates the location of published photoreceptor CREs which have experimental support (references are given in MATERIALS AND METHODS). Annotations in this column such as ‘6.5 Kb 5’’ means the tested CRE contained the first 6.5 Kb upstream of the TSS. Known CREs predicted by our algorithm are highlighted in light green. ‘Predicted CRE(s)’ lists candidate CREs predicted by our algorithm which lie within +/− 15 Kb of the TSS. In some cases, two locations are given which represent that CRE prediction closest to the TSS and that with the highest score. Only one location is given when they are the same. Experimentally tested CREs are highlighted in red. The CREs tested for Crx and Elovl4 lie more than 15 Kb downstream of the TSS. The CRE tested for Tulp1 scores below the cutoff threshold of 200. This CRE was selected for testing during the early phases of this project using an earlier version of PhastCons with which it scored above threshold. A third predicted CRE location is given for Opn1mw which corresponds to the locus control region. ‘CRE strength’ indicates the estimated strength of the indicated CRE as tested by electroporation (see Fig. S2). The last four columns of the table show the wild-type-to-mutant ratios of the averaged microarray scores for the given gene. Dark green = downregulated under high stringency (as described in MATERIALS AND METHODS); light green = downregulated under low stringency; red = upregulated under high stringency; orange = upregulated under low stringency. A dash indicates that the gene was not significantly altered in the given mutant. ‘NA’ indicates that this gene is not represented on the Affymetrix Mouse 430 2.0 microarray.
Figure 3.
Computational algorithm to identify photoreceptor cis-regulatory elements.
A, Output of computational algorithm for 12 Kb of genomic DNA around the Rho locus. ‘PhastCons’ indicates the degree of phlyogenetic conservation of a given stretch of DNA as determined by the PhastCons algorithm. ‘Conserved Crx sites’ indicates the occurrence of a phylogenetically conserved Crx site. The height of the bar representing a site indicates the ‘affinity’ of the site as measured as the percentile rank of that site's log odds score relative to the consensus site for that transcription factor = 1 (i.e., 100%ile). ‘CRE score’ indicates the likelihood that a given 500 bp stretch (on which the peak is centered) harbors a photoreceptor CRE (details about scoring are given in MATERIALS AND METHODS). Color code: red if score ≥500, dark blue if 500>score ≥200, light blue if score <200. B, Pattern of phylogenetic conservation around Rho CRE as displayed by UCSC genome browser. C, Sections of mouse retina electroporated in vivo at P0 with the indicated constructs and harvested at P14. The Rhodopsin promoter used here derives from cow. The white arrows highlight CAG-eGFP-positive cells in the INL in which DsRed is not expressed. IS = inner segment.
Figure 4.
Novel photoreceptor-specific cis-regulatory elements.
A, The top panel shows the output of our computational algorithm for the first intron of Rgs9. The red bar indicates the region of genomic DNA assayed for CRE activity. The bottom panels show flatmount and cross-sectional images of a mouse retina electroporated at P0 with the indicated constructs, cultured as an explant, and harvested at P8. The CAG-eGFP is a ‘loading’ control expressed in all retinal cell types. ‘Rgs9-CRE’ consists of the Rgs9 regulatory peak (indicated by the red bar in top panel) fused to DsRed. The cross-sectional image is a composite of images taken in the green and red channels. Size bar = 500 µm for flatmount images and 100 µm for cross-sections B, A control construct containing the indicated region of DNA from the first intron of Rgs9 which does not correspond to a predicted regulatory peak. C–E, Additional examples of novel photoreceptor-specific CREs. Note that some annotations of the Rgs9bp locus suggest that the indicated regulatory peak falls within the 3′ UTR of this gene. F–H, CREs around the Crx, Nrl, and Nr2e3 loci. The regulatory peak upstream of the TSS in F indicates a previously characterized Crx CRE [44]. The Crx-CRE depicted in the bottom panels corresponds to the downstream regulatory peak identified in the present study (indicated by the red bar). The Nrl-CRE shown in G was previously characterized [24] and is included here as a control.
Figure 5.
In silico evolution of functional photoreceptor cis-regulatory elements.
A, Flatmount and cross-sectional images of mouse Rho-CRE fused to DsRed co-electroporated with CAG-eGFP. All constructs were electroporated at P0, cultured as explants, and harvested at P8. The column labeled ‘CRE architecture’ shows the distribution of Crx (top half) and Nrl (bottom half) sites with the indicated ‘affinity’ as described in the legend at the top of the figure. The orientation of a given site is indicated by the direction of the triangle representing that site. The score on the y-axis is the log odds score for that site which reflects its closeness to consensus. Size bar = 500 µm for flatmount images and 100 µm for cross-sections B-D, Images of retinas electroporated with the indicated synthetic CREs. For example, ‘Syn1-G70’ indicates a sequence corresponding to the genome of the ‘fittest’ organism of generation 70 in evolutionary run 1. The CREs in B-D derive from three separate evolutionary runs. E, Graph of first 700 generations of evolutionary run 3. The four ‘organisms’ from this run whose genomes were tested for CRE activity are circled and images of their expression patterns are given in Fig. S4. F, Graph of expression strength of four synthetic CREs from evolutionary run 3.