Microarray and Morphological Analysis of Early Postnatal CRB2 Mutant Retinas on a Pure C57BL/6J Genetic Background

In humans, the Crumbs homologue-1 (CRB1) gene is mutated in progressive types of autosomal recessive retinitis pigmentosa and Leber congenital amaurosis. The severity of the phenotype due to human CRB1 or mouse Crb1 mutations is dependent on the genetic background. Mice on C57BL/6J background with Crb1 mutations show late onset of retinal spotting phenotype or no phenotype. Recently, we showed that conditional deletion of mouse Crb2 in the retina results in early retinal disorganization leading to severe and progressive retinal degeneration with concomitant visual loss that mimics retinitis pigmentosa due to mutations in the CRB1 gene. Recent studies in the fruit fly and zebrafish suggest roles of the Crumbs (CRB) complex members in the regulation of cellular signalling pathways including the Notch1, mechanistic target of rapamycin complex 1 (mTORC1) and the Hippo pathway. Here, we demonstrate that mice backcrossed to C57BL/6J background with loss of CRB2 in the retina show a progressive disorganization and degeneration phenotype during late retinal development. We used microarray gene profiling to study the transcriptome of retinas lacking CRB2 during late retinal development. Unexpectedly, the retinas of newborn mice lacking CRB2 showed no changes in the transcriptome during retinal development. These findings suggest that loss of CRB2 in the developing retina results in retinal disorganization and subsequent degeneration without major changes in the transcriptome of the retina. These mice might be an interesting model to study the onset of retinal degeneration upon loss of CRB proteins.


Introduction
Retinal cell generation and differentiation in the mouse occurs from embryonic day (E) 11 to postnatal day (P) 10. Six major neuronal and one glia cell type are generated from multipotential retinal progenitors in a characteristic sequence during development [1,2]. Cell adhesion and cell polarity protein complexes, such as the Crumbs (CRB) and adherens junctions complexes, play a critical role in maintenance of the proliferation of the progenitor cells [3]. Changes in these complexes disturb the spatiotemporal aspects of retinogenesis, leading to retinal degeneration resulting in mild or severe impairment of retinal function and vision [4][5][6][7][8]. Recent studies demonstrated that the CRB complex members are able to regulate several important signalling pathways including the Notch1 [9][10][11], mechanistic target of rapamycin complex 1 (mTORC1) [12,13] and the Hippo pathway [3,[14][15][16][17].
The apical CRB complex is located at the subapical region adjacent to adherens junctions between the retinal progenitor cells in the developing retina [4] and, after differentiation, at the subapical regions of Müller glia and photoreceptor cells [7,18]. In mammals, the CRB family consists of CRB1, CRB2, CRB3A and CRB3B [19].
In humans, mutations in the CRB1 gene are responsible for retinal diseases such as Leber congenital amaurosis and retinitis pigmentosa [19][20][21]. The lack of a clear genotypephenotype correlation suggests that other components of the CRB complex have a function influencing the severity of the retinal disease. We recently showed that conditional deletion of the Crb gene family member Crb2, specifically in early progenitors, results in disorganization during late retinal development leading to severe and progressive retinal degeneration with concomitant visual loss that mimics retinitis pigmentosa due to mutations in the CRB1 gene [4]. Variation in the genetic background may influence the severity of the retinal phenotype, as described before for the Crb1 knockout (Crb1 -/-) and Crb1Crb2 conditional knockout mice [7,22,23].
Microarray analysis has been used to study changes in gene expression during retinal development [24][25][26][27][28][29]. Many of the previous studies using this technique focused on a specific cell type, for example using isolated single cells [30] or cell populations [31], or retinas from cell type specific mutants [32][33][34][35][36]. Backcrossing Crb1 mutant retinas from mixed to C57BL/6J genetic background strongly suppressed the morphological phenotype [22]. Here, we show that backcrossing Crb2 null retinas from mixed (50% OLA129 and 50% C57BL/6J) to 99.9% C57BL/6J background did not suppress the severe morphological phenotype. To elucidate the molecular events that precede and lead to the morphological phenotype in the Crb2 retina-specific conditional knockout mice on 99.9% C57BL/6J background, we applied microarray-based mRNA profiling in retinal tissue of postnatal stages P0, P3, P6, and P10. The morphological phenotype did however not result in a significantly altered transcriptome at any stage of retinal development analysed.

Animals
All procedures concerning animals were performed with permission of the animal experimentation committee (DEC) of the Royal Netherlands Academy of Arts and Sciences (KNAW), permit number NIN06-46. Crb2 conditional knockout mice were generated and genotyped as described in [4], and were backcrossed to 99.9% C57BL/6J genetic background. Mice had no mutations in the phosphodiesterase 6b (pde6b) or Crb1 (rd8) genes. mT/mG reporter mice were obtained from the Jackson laboratory (Gt(ROSA)26Sortm4(ACTB-tdTomato,-EGFP)) [37]. Animals were maintained on a 12 h day/night cycle and supplied with food and water ad libitum.

RNA isolation and amplification
Retinae were thawed in RLT buffer (RNeasy Mini Kit, catalogue number: 74106; Qiagen Benelux, Venlo, The Netherlands) and homogenized with a pellet pestle. Total RNA was isolated using RNeasy Mini Kit according to the manufacturer's instructions. RNA extract was eluted in RNase-free water (30 μL). RNA yield and purity were determined using a NanoDrop ND-1000 spectrophotometer (NanoDrop Technologies, Wilmington, DE, USA). RNA integrity was determined by the RNA integrity number (RIN), as measured by the Agilent 2100 bioanalyser (Agilent Technologies, Palo Alto, CA, USA). Overall, the isolated RNA was of high integrity (average RIN = 9.3, range 8.3-10.0, Table S1 and Table S2).

Sample labelling and microarray hybridization
For microarray analysis, Agilent 4x44K v2 whole mouse gene expression microarrays (Agilent Technologies, catalogue number G4846A) were used. Sample labelling and microarray hybridization were performed according to manufacturer's instructions. Briefly, for each sample, 200 ng of RNA were linearly amplified and fluorescently labelled with either Cy3-CTP or Cy5-CTP (Perkin Elmer) using the Agilent Low RNA Input Fluorescent Linear Amplification Kit (Agilent Technologies). Prior to hybridization, equal amounts (825 ng) of Cy3-and Cy5-labelled RNA were hydrolyzed for 30 min at 60°C in 1x fragmentation buffer (Agilent Technologies). The fragmented targets were hybridized to a microarray by incubating for 17 h at 60°C in 1x target solution (Agilent Technologies) in a rotating hybridization chamber. Specifically, each hybridization consisted of two individual samples (one control and one knockout), one labelled with Cy3 and one with Cy5. None of the samples were pooled. A full description of all hybridizations can be found in Table S3. After hybridization, the arrays were washed at room temperature for 5 min in 6x salinesodium phosphate-EDTA (SSPE)/0.005% N-Lauroylsarcosine (Sigma-Aldrich, St Louis, MO, USA), 1 min in 0.06x SSPE/ 0.005% N-Lauroylsarcosine and 30 s in acetonitrile (Sigma-Aldrich), then dried in a nitrogen flow. Microarrays were scanned using an Agilent DNA Microarray Scanner at 5 µm resolution and 10% (low) and 100% (high) photomultiplier tube setting. Microarray scans were quantified using Agilent Feature Extraction software (version 9.5.3.1).

Microarray normalization and single gene analysis
Raw expression data were analyzed in R statistical processing software (version 2.6.0) using the LIMMA package [38] in Bioconductor (www.bioconductor.org). All features that were flagged as saturated or as a nonuniformity outlier by the feature extraction software on 1 or more arrays were excluded from further analysis. This was applicable for 980 features, resulting in 43,020 features which passed these criteria. Array 20 was excluded from the studies due to a technical artefact. Data were normalized using LIMMA by applying a background correction (using the 'normexp' algorithm) followed by normalization of intensity distributions within and between arrays (using the 'quantile' algorithm). It was recently demonstrated that the analysis of the separate intensity channels (the individual Cy3 and Cy5 signals) yields more reproducible results than the standard ratio-based approach (the ratio between the Cy3 and Cy5 channels) for dual-color microarray datasets [39], so we also applied this approach for the present dataset. Thus, the 2 log-transformed intensity measurements per sample were extracted from the normalized ratio data and used in all following analyses.
To detect genes with a significant interaction between age and genotype, we performed a 2-way analysis of variance (ANOVA) using age and genotype as grouping factors. To compare gene expression between Crb2 cKO and control retinas at each timepoint separately t-tests were used. The Benjamini-Hochberg method was used to correct for multiple testing. P-values < 0.05 after correction were considered significant.
A second and independent analysis of the data was done using Partek version 6.6 (Partek Inc., St. Louis, MO). The Feature Extraction result text files were imported into Partek, log 2 transformed and quantile normalized. To visualize the clustering of the samples, principal component analysis (PCA) was used. Differentially expressed genes were selected using a 2-way ANOVA using the factors genotype and age. The cutoff value for significantly expressed genes was a false discovery rate (FDR) [40], of 0.05 or less.
The data discussed in this publication have been deposited in NCBI's Gene Expression Omnibus [41] and are accessible through GEO Series accession number GSE50845 (http:// www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE50845).

Quantitative real-time PCR
RNA was isolated, from P3 and P10 retinas, 5-6 control and Crb2 cKO, as indicated before. After genomic DNA degradation with RNase-free DNase I (New England Biolabs), 0.5 µg of total RNA was reverse transcribed into first-strand cDNA with Superscript III Plus RNase H-Reverse Transcriptase (Life Technologies) and 50 ng random hexamer primers, during 50 min at 50°C in a total volume of 20 µl. To the resulting cDNA sample, 14 µl of 10 mM Tris, 1 mM EDTA was added. From all samples, a 1:20 dilution was made and used for qPCR analysis. For this analysis, primer pairs were designed with a melting temperature of 60 °C, giving rise to an amplicon of 60-206 bp. Real-time qPCR was based on the real-time monitoring of SYBR Green I dye fluorescence on a ABI Prism 7300 Sequence Detection System (Applied Biosystems, Nieuwekerk a/d IJssel, The Netherlands). The PCR conditions were as follows: 12.5 µL SYBR Green PCR 2x mastermix (Applied Biosystems), 20 pmol of primers (see Table S4), and 2 µl of the diluted cDNA (ca 3 ng total RNA input). An initial step of 50°C for 2 min was used for AmpErase incubation followed by 15 min at 95°C to inactivate AmpErase and to activate the AmpliTaq. Cycling conditions were as follows: melting step at 95°C for 1 min, annealing at 58°C for 1 min and elongation at 72°C, for 40 cycles. At the end of the PCR run, a dissociation curve was determined by ramping the temperature of the sample from 60 to 95°C while continuously collecting fluorescence data. Non template controls were included for each primer pair to check for any significant levels of contaminants. Values were normalized by the mean of the 3 reference genes Hypoxanthine-guanine phosphoribosyltransferase, elongation factor 1-a and ribosomal protein S27a.

Quantification of Müller glia cells
Müller glia cells were counted at P10 using a SOX9 antibody.

Statistical analysis
Normality of the distribution was tested by Kolmogorov-Smirnov test. Statistical analysis by Student's t test or by Mann Whitney U test in case of a non-normal distribution. Values of *P<0.05, **P<0.01, ***P<0.001 were considered to be statistically significant. Values are expressed as means ± SEM. Calculation were made using the SPSS statistical package version 17.0.

Gene expression profile
The CRB2 protein is expressed during retinal development in radial glial progenitor cells, and after differentiation in cone and rod photoreceptors and Müller glia cells [4]. Since these cells span the entire retina, we chose to use whole neural retinas (without the retinal pigment epithelium) in gene expression profile studies. To reduce variation and to standardize the genetic background of Crb2 cKO mice, the animals were backcrossed from mixed to 99.9% C57BL/6J background. Variation in the genetic background may influence the gene expression, as described before for the Crb1 knockout (Crb1 -/-) mice [36,42]. Approximately 6 µg of total RNA was extracted from each retina (Table S1 and Table S2). The values of the A260/A280 ratio were in between 1.8 to 2.0 (Table S1 and  Table S2), which indicated the absence of contaminating proteins. The A260/A230 ratios were larger than 2.0, which indicated the absence of organic compounds that can interfere with the labelling reaction, such as guanidinium isothiocyanate, alcohol and phenol as well as cellular contaminants such as carbohydrates. The total RNA quality was assessed using the Agilent 2100 Bioanalyser and the respective software, which provides a RNA Integrity Number (RIN) of maximally 10. The samples had on average a RIN of 9.3 indicating high quality of the total RNA.
We used a whole mouse gene expression microarray that contains over 44,000 probes. To elucidate the molecular events that precede and lead to the morphological phenotype in the Crb2 retina-specific conditional knockout mice, retinal tissue was collected at postnatal day 0 (P0), P3, P6 as well as P10. Five different retinas from Crb2Chx10 cKO (Crb2F/F/ Chx10Cre +/-) and the control (Crb2F /+ /Chx10Cre +/-) from each time point were used to cover the onset and first stages of the morphological phenotype observed in the mutant retinas. Equal amounts (825 ng) of Cy3-and Cy5-labelled RNA were hydrolysed and hybridized to a microarray, the hybridization pairs are indicated in Table S3. Five microarrays were processed for each time point. To detect differentially expressed genes a 2-way analysis of variance (ANOVA) using age and genotype as grouping factors was performed. The Benjamini-Hochberg method was used to correct for multiple testing. P-values < 0.05 after correction were considered significant. Although we observed substantial changes in gene expression between time points (see section hierarchical clustering), our analysis did not reveal significant differences in gene expression between mutant and control retinas. A second analysis, where we used t-tests to compare gene expression between Crb2 cKO and control retinas at each time point separately, also did not reveal significant changes in gene expression. A description of the top 100 genes with the lowest P-values for each time point can be found in the Table S5,  Table S6, Table S7 and Table S8.
The Crb2 cKO mice were constructed in such way that the full length CRB2 protein could not be produced in the homozygote mutant retinas. The 5' loxP site was inserted in intron 9, the 3' loxP site was inserted in exon 13 behind the stop codon. The deleted area contains part of the CRB2 extracellular domain, and the entire transmembrane domain and the 37 amino acids of the C-terminal intracellular domain. Cre mediated recombination deleted coding exons 10-13 and resulted in a nonsense mutation with premature truncation of the CRB2 protein at amino acid 871. The unique 60 base pair Crb2 probe (ID: 235570) in the array is located in between the 3' loxP site and the polyadenylation sequence of the gene (Figure 1). In the microarray study, the Crb2 transcript was not downregulated in the Crb2 cKO retinas.
To study in more detail the levels of Crb2 transcript in the Crb2 cKO retinas we performed quantitative real timepolymerase chain reaction (qRT-PCR) using two different sets of primers, one located outside of the recombined area (exon 7) and another set in the recombined area (exons 11/12) (Figure 1). At P3, we observed a decrease of 70% in the levels of Crb2 transcript compared to the control, using primers located in the recombined area (between exons 10-13) (P = 0.004) (Figure 2A). However, we did not detect changes in the Crb2 levels using the primers located outside the recombined area (exon 7) (Figure 2A). The data suggest that recombination and deletion of exons 10-13 occurred in the knockout retinas, what suggest again the existence of a truncated transcript. At P10, we did not observe differences in the levels of Crb2 transcript, using primers located in the recombined area (P = 0.381) (Figure 2A), what suggest that the proportion of Crb2 null cells is decreased compared to non-recombined cells, which might be due to increased apoptosis of Crb2 mutant photoreceptors [4].
We previously showed using an antibody against the intracellular C-terminal 37 amino acids domain of CRB2 that the protein was removed from the outer limiting membrane from embryonic day 12.5 onwards [4]. So, the truncated transcript is recognized by the probe on the microarray, and expression of the truncated transcript is not changed compared to the full length transcript.
These findings may be explained in part by the mosaic genetic nature of the mutant retinas which contain mutant next to wild type cells due to insufficient levels of Cre expression in a subset of cells. To evaluate the mosaicism of the Chx10Cre mouse line we crossed the Chx10Cre transgenic mouse line with a mT/mG (membrane-targeted tandem-dimer-Tomato RFP / membrane-targeted Enhanced GFP) reporter mouse line [37]. This reporter mouse line expresses red fluorescent protein before and green fluorescent protein after Cre-mediated recombination. While in the mT/mG retinas only red fluorescent protein could be detected ( Figure 2B1), in mT/mG::Chx10Cre (heterozygous mT/mG, heterozygous Chx10Cre) retinas expression of red and green fluorescent proteins was found ( Figure 2B2), suggesting the presence of mutant adjacent to wild type cells.
In mammals, the Crb gene family is composed of Crb1, Crb2 and Crb3, the lack of changes in the transcriptome after loss of Crb2 gene function might be due to compensation by Crb1 or Crb3. To test this, we used qRT-PCR to determine the levels of the Crb1 and Crb3 transcripts, but whereas we detected a decrease in Crb2 transcripts, we did not observe changes in the levels of Crb1 or Crb3 transcripts between control and knockout retinas (Figure 2A).

Crb2 and signalling pathways
We investigated by qRT-PCR the role of Crb2 in the regulation of Notch1, mTORC1, Hippo, Wnt and sonic hedgehog pathways in Crb2 cKO retinas at P3 and P10. We did not observe major changes in these pathways (Figure 3). Only the Hey1 transcript was significantly changed at P10, with an increase of nearly 2.5-fold in the knockout retinas compared to the control (P = 0.041). The qPCR data confirmed the data obtained by microarray analysis.

Hierarchical clustering
Analysis of cluster gene expression was performed at distinct retinal development time points. Figure 4 and Figure S1 show the correlative gene expression from P0 to P10. Two phases are visible, P0-P3 and P6-P10 based on their correlative distance of gene expression. This is in accordance with the findings described by Zhang et al, 2006. They described two phases as a "developmental phase" (from E12.5 to P5) and a "functional phase" (from P7 on) [24]. This analysis also revealed that the gene expression variation induced by differences in development time points was much larger that the differences between Crb2 cKO and control retinas.

The retinal phenotype of Crb2 cKO on C57BL/6J background
The phenotype of retinas lacking CRB2 from early retinal development was previously described [4]. Those animals were kept on a mixed genetic background (50% OLA129 and 50% C57BL/6J). In order to reduce variation and to standardize the genetic background of these mice, the animals were backcrossed to 99.9% C57BL/6J background. Variation in the genetic background may influence the severity of the retinal phenotype, as described before for the Crb1 knockout (Crb1 -/-) mice [7,22]. So, to study the effect of the genetic background   on the morphological phenotype of Crb2 null retinas, retinas of P10, and 1-and 3 months-of-age were analysed.
The phenotype of retinas lacking CRB2 on 99.9% C57BL/6J genetic background appeared grossly the same as on 50% C57BL/6J genetic background [4] at the time points analysed ( Figure 5 and Figure S2). At P10, disruptions of the outer limiting membrane and ectopic photoreceptor nuclei near to the retinal pigmented epithelium were observed ( Figure 5B and Figure S2A).
Adult retinas on 99.9% or 50% C57BL/6J genetic background showed similar progression of the morphological phenotype. In 1 month-old retinas, protrusions of photoreceptors into the subretinal space were observed ( Figure  5D and Figure S2B). Three month-old retinas also showed protrusion of photoreceptor nuclei into the subretinal space, in these retinas the retinal thickness decreased to only a few (4-6) photoreceptor nuclei in a row, the outer plexiform layer was thinner and protrusion of nuclei from the inner nuclear layer into the outer nuclear layer were found ( Figure 5F and Figure S2C).
In conclusion, the CRB2 null retinas on 99.9% C57BL6/J background showed a similar progression of the phenotype than on 50% C57BL6/J background.

Early morphological retinal phenotype of the Crb2 cKO on C57BL/6J background
As in CRB2 null retinas on mixed background, the CRB2 null retinas on 99.9% C57BL6/J background showed at postnatal day 10 a consistent disorganized morphology with many photoreceptors misplaced in the subretinal space adjacent to the retinal pigmented epithelium, thus we examined 10-days old retinae in detail using immunohistochemistry. At P10, the outer limiting membrane was disrupted throughout the mutant retina, and although CRB2 was lost in the entire retina ( Figure  6B), loss of the CRB complex proteins, CRB1 ( Figure 6D), PALS1 ( Figure 6F), MUPP1 ( Figure 6J) and of the PAR complex member PAR3 ( Figure 6H) occurred only at sites of protrusions of photoreceptor nuclei into the subretinal space. Similar disruptions of adherens junction markers, β-Catenin ( Figure 6L), Catenin pp120 ( Figure 6N), N-Cadherin ( Figure  6P), ZO-1 ( Figure 6R) and Nectin1 ( Figure 6T) were observed.
In CRB2 null retinas several ectopic photoreceptor nuclei were present in the subretinal space immediately adjacent to the retinal pigment epithelium ( Figure 5C, Figure 6J, T and Figure 7B, D, F, H). Most of these cell bodies were positive for recoverin ( Figure 7B) and rhodopsin ( Figure 7D). The lamination of cone photoreceptor cells, stained by cone arrestin (Figure 7D), M-opsin ( Figure 7F) was also affected. Many cone nuclei were misplaced in the subretinal space. Some of the misplaced photoreceptors showed polarity defects with a reverse apical-basal orientation. The cone outer segments were present in both retinas ( Figure 7G, H), however in the mutant retinas the outer segments were located between the correct localized outer nuclear layer and the ectopic photoreceptors nuclei in the subretinal space ( Figure 7H), instead of been located immediately adjacent to the retinal pigmented epithelium.
Some of the ectopic photoreceptor cells were not able to establish synapses in the outer plexiform layer, demonstrated by ectopic localization of PSD-95 ( Figure 8B) and MPP4 (data not shown).
Most of bipolar cells stained by anti-PKCα ( Figure 8B), amacrine calretinin-positive cells ( Figure 8D), and amacrine and ganglion cells (PAX6-positive, Figure 8F) were correctly localized, showing no major defects in the lamination of these cell types, however in some areas some misplaced bipolar ( Figure 8B) and amacrine cells ( Figure 8F, arrowheads) were observed in the outer plexiform layer.
Also, no misplaced SOX9-positive Müller glia nuclei were observed in the CRB2 null retinas ( Figure 9B), however in the mutant retinas disruptions at the apical end feet of the Müller glia cell were observed ( Figure 9B, arrowheads), these disruptions coincide with the photoreceptor protrusions, suggesting lack of adhesion between Müller and photoreceptor cells. A mild increase in glial fibrillary acidic protein (GFAP) was observed in the outer nuclear layer of the knockout ( Figure   9D) compared to the control retinas ( Figure 9C), suggesting Müller glia activation.
As in CRB2 null retinas on mixed background, the  Gliosis was more evident in 2-months-old mutant retinas, where a prominent increase of GFAP was observed ( Figure  10B, arrowheads), mainly at foci of cellular mislocalization. At these sites, ectopic SOX9-positive Müller glia nuclei could be detected in the outer nuclear layer ( Figure 10D, arrowheads). Similar as in P10 retinas, an increase in CD45 and CD11b was observed in 2-month-old retinas, these proteins were detected in the outer nuclear layer and adjacent to ectopic photoreceptor nuclei in the subretinal space ( Figure 10F, H).
In conclusion, the absence of differences in transcriptomes between control and CRB2 null retinas is not due to suppression of the CRB2 retinal phenotype on C57BL/6J background.

Discussion
At the time points of retinal development analysed, the retinas lacking CRB2 showed discrete but consistent phenotypes mainly at the periphery of the retina, with the exception of the P10 time point where the phenotype became severe with the accumulation of misplaced photoreceptor cells in the subretinal space throughout the entire retina. However, no differences were found in the mutant retinas in our gene profiling studies, with exception of Hey1 at P10. This finding was unexpected since several studies demonstrated that the CRB complex members are able to regulate several important signalling pathways including the Notch1 [9][10][11], mTORC1 [12,13] and the Hippo pathway [3,[14][15][16][17]. Moreover, the knockout retinas showed alteration in several cellular parameters, as for example, increase in apoptosis, number of retinal progenitor cells and/or in the number of late born cells such as Müller glia and rod photoreceptor cells. The mislocalized photoreceptors showed an immature morphology and polarity defects. The findings may be explained in part by the mosaic genetic nature of the mutant retinas which contain mutant next to wild type cells due to insufficient levels of Cre expression in a subset of cells. The use of a Cre mouse line that expresses Cre in all retinal progenitor cells might circumvent this problem. In the current study, retinas with loss of CRB2 in a mix of mutant radial glial progenitor cells, rod and cone photoreceptors, and Müller glia cells were analysed in the gene expression studies. In future studies, it might be worthwhile to ablate expression of CRB2 by different Credrivers specifically in photoreceptors and/or Müller glia cells and perform gene expression profiles on purified photoreceptors and/or Müller glia cells purified by cell sorting.     In previous studies using Crb1 mutant mice on mixed genetic background (50% OLA129 and 50% C57BL/6J), only a limited number of five genes showed changes in gene expression in 3months-old white-light-exposed Crb1 -/-mice [36]. Three of these genes are involved in inhibition of chromatid separation (Pttg1), chromatid cohesion (Esco1), or chromosome stabilization (similar to histone H2b). Interestingly, downregulation of Pttg1 was also detected in white light exposed Crb1 C249W/-retinas that expressed a CRB1 protein with an amino acid substitution in the extracellular domain [42].
In Crb2 null retinas, the CRB2 protein is lost from the entire subapical region. However, loss of CRB2 leads to loss of localization of other apical proteins, as for example members of the CRB, PAR and adherens junction complexes, only at foci where mislocalized cells were found.
In conclusion, the results suggest that loss of CRB2 in the developing retina had no major influence on the gene expression profile.   Table S5. Differential gene expression between control and knockout neuroretinas in fold differences, at postnatal day 0. Top 100 genes ranked on their P value given by the students' t-test (P value) before applying Benjamini-Hochberg (P value bh) method for correct to multiple testing. The expression value to the individual genes for control (CONT) and knockout (CKO) groups (log 2 intensity), and the fold differences between control and knockout (FC) are also described in the table.

Supporting Information
(DOCX) Table S6. Differential gene expression between control and knockout neuroretinas in fold differences, at postnatal day 3. Top 100 genes ranked on their P value given by the students' t-test (P value) before applying Benjamini-Hochberg (P value bh) method for correct to multiple testing. The expression value to the individual genes for control (CONT) and knockout (CKO) groups (log 2 intensity), and the fold differences between control and knockout (FC) are also described in the table.
(DOCX) Table S7. Differential gene expression between control and knockout neuroretinas in fold differences, at postnatal day 6. Top 100 genes ranked on their P value given by the students' t-test (P value) before applying Benjamini-Hochberg (P value bh) method for correct to multiple testing. The expression value to the individual genes for control (CONT) and knockout (CKO) groups (log 2 intensity), and the fold differences between control and knockout (FC) are also described in the table. (DOCX) Table S8. Differential gene expression between control and knockout neuroretinas in fold differences, at postnatal day 10. Top 100 genes ranked on their P value given by the students' t-test (P value) before applying Benjamini-Hochberg (P value bh) method for correct to multiple testing. The expression value to the individual genes for control (CONT) and knockout (CKO) groups (log 2 intensity), and the fold differences between control and knockout (FC) are also described in the table.