Gene Profiling of Postnatal Mfrprd6 Mutant Eyes Reveals Differential Accumulation of Prss56, Visual Cycle and Phototransduction mRNAs

Mutations in the membrane frizzled-related protein (MFRP/Mfrp) gene, specifically expressed in the retinal pigment epithelium (RPE) and ciliary body, cause nanophthalmia or posterior microphthalmia with retinitis pigmentosa in humans, and photoreceptor degeneration in mice. To better understand MFRP function, microarray analysis was performed on eyes of homozygous Mfrprd6 and C57BL/6J mice at postnatal days (P) 0 and P14, prior to photoreceptor loss. Data analysis revealed no changes at P0 but significant differences in RPE and retina-specific transcripts at P14, suggesting a postnatal influence of the Mfrprd6 allele. A subset of these transcripts was validated by quantitative real-time PCR (qRT-PCR). In Mfrprd6 eyes, a significant 1.5- to 2.0-fold decrease was observed among transcripts of genes linked to retinal degeneration, including those involved in visual cycle (Rpe65, Lrat, Rgr), phototransduction (Pde6a, Guca1b, Rgs9), and photoreceptor disc morphogenesis (Rpgrip1 and Fscn2). Levels of RPE65 were significantly decreased by 2.0-fold. Transcripts of Prss56, a gene associated with angle-closure glaucoma, posterior microphthalmia and myopia, were increased in Mfrprd6 eyes by 17-fold. Validation by qRT-PCR indicated a 3.5-, 14- and 70-fold accumulation of Prss56 transcripts relative to controls at P7, P14 and P21, respectively. This trend was not observed in other RPE or photoreceptor mutant mouse models with similar disease progression, suggesting that Prss56 upregulation is a specific attribute of the disruption of Mfrp. Prss56 and Glul in situ hybridization directly identified Müller glia in the inner nuclear layer as the cell type expressing Prss56. In summary, the Mfrprd6 allele causes significant postnatal changes in transcript and protein levels in the retina and RPE. The link between Mfrp deficiency and Prss56 up-regulation, together with the genetic association of human MFRP or PRSS56 variants and ocular size, raises the possibility that these genes are part of a regulatory network influencing postnatal posterior eye development.

Within the eye, MFRP is exclusively localized to the apical surface of the retinal pigment epithelium (RPE) and the ciliary body epithelium [11][12][13]. The protein has been suggested to play a role in the normal development as well as maintenance of photoreceptor outer segments (OS) [12]. MFRP is a type II transmembrane protein and contains multiple domains, including an N-terminal cytoplasmic domain, a transmembrane domain, two extracellular cubulin (CUB) domains, a low-density lipoprotein domain (LDLa) and a C-terminal cysteine-rich domain (CRD) [8,14]. Complement C1q tumor necrosis factor-related protein 5 (C1QTNF5, also known as CTRP5) is expressed from the same dicistronic transcript as Mfrp [8,11,15]. Dicistronic messages often function in common pathways [16]. Therefore, it is notable that MFRP and C1QTNF5 co-localize in the posterior eye [8,11,15] and have been shown to interact directly by two-hybrid and biochemical studies [11]. The functional consequence of this interaction, however, is unknown, and other potentially interacting partners of MFRP remain to be identified.
To obtain insight into the functional role of MFRP in the RPE and retina during normal eye development and disease, geneprofiling studies were carried out in Mfrp rd6 and wildtype postnatal eyes. Microarray analysis revealed a moderate decrease in phototransduction, visual cycle, and gene transcripts associated with retinal degeneration in Mfrp rd6 mutants. Most interestingly, an increased temporal expression of Prss56, a gene linked to posterior microphthalmia [17] and associated with ocular growth defects in myopia [18,19] was observed. Prss56 upregulation appeared to be specific to Mfrp rd6 mice, as it was not observed in other retinal degeneration models with similar disease progression. Broadly, our findings delineate a potential role of MFRP in postnatal development and/or maintenance of the posterior eye, and provide evidence that MFRP and PRSS56 participate in the same functional pathway.

Ethics Statement
Protocols using mice in this study were approved by The Jackson Laboratory Institutional Animal Care and Use Committee (Animal Welfare Assurance Number: A3268-01) in accordance with the "Guide for the Care and Use of Experimental Animals" established by the National Institutes of Health (NIH) (1996, Revised 2011).

Animals
Mouse strains utilized in this study, B6.C3Ga-Mfrp rd6 /J (Mfrp rd6 ) and C57BL/6J were bred and maintained in a vivarium with a 12 hour light and 12 hour dark cycle in the Research Animal Facility at The Jackson Laboratory. Autoclaved NIH31 diet (6% fat) and HCl acidified water (pH 2.8-3.2) were provided ad libitum.

Gene Profiling
GeneChip Mouse Genome 430 v2.0 array (Affymetrix) was performed at The Jackson Laboratory Gene Expression Analysis Service (GES). Total RNA from Mfrp rd6 and C57BL/6J eyes at P0 and P14 (three biological replicates per group) was extracted with Trizol (Life Technologies, Carlsbad, CA, USA) according to the manufacturer's instructions. Two mg of total RNA was used for cDNA synthesis with the One-Cycle Target Labeling cDNA Synthesis Kit (Affymetrix, Santa Clara, CA, USA) and fragmented cRNA was synthesized using the GeneChip IVT Labeling kit (Affymetrix, Santa Clara, CA, USA) according to the manufacturer's protocol (GeneChip Expression analysis Technical manual, Affymetrix 1999Affymetrix -2004. Following cRNA fragmentation, 15 mg of cRNA were hybridized on a GeneChip Mouse Genome 430 v2.0 array at 45uC for 16 h and the microarray was scanned using the GeneChip Scanner 3000 (Affymetrix, Santa Clara, CA, USA). Average signal intensities for each probe set within the arrays were calculated by and exported from Affymetrix's Expression Console (Version 1.2) software using the RMA method, which incorporates convolution background correction, sketch-quantile normalization, and summarization based on a multi-array model fit using the median polish algorithm to generate gene expression data. For this experiment, six pairwise comparisons were used to statistically resolve gene expression differences between sample groups using the R/maanova analysis package. Specifically, differentially accumulated transcripts were detected by using Fs, a modified F-statistic incorporating shrinkage estimate of variance components from within R/maanova. Statistical significance levels of the pairwise comparisons were calculated by permutation analysis (1000 permutations) and adjusted for multiple testing using the false discovery rate (FDR) q-value threshold of 0.05. For each probe set, the raw intensities for all probes were log2-transformed. The log2-transformed intensities were quantile-normalized and a volcano plot was generated for each of the six pairwise comparisons. Differentially accumulated transcripts were classified into various biological pathways using the Ingenuity Pathway Analysis Systems software (Ingenuity Systems, Qiagen, Valencia, CA, USA). Microarray data (MIAME compliant) were submitted to the Gene expression Omnibus (GEO) database (http://www. ncbi.nlm.noh.gov/geo) under GEO Accession Number GSE53411 (http://www.ncbi.nlm.nih.gov/geo/info/linking. html).

Ingenuity Pathway Analysis
Ingenuity pathway analysis software (IPA; Ingenuity Systems, www.ingenuity.com, Summer 2013 release, Qiagen, Valencia, CA, USA) was used to identify potential networks affected by the disruption of Mfrp across one dataset (Mfrp rd6 versus C57BL/6J at P14). The dataset was derived from differential gene expression analysis resulting from the comparison of mRNA from whole eye tissue samples of Mfrp rd6 versus C57BL/6J mice. Data uploaded into IPA consisted of Affymetrix Mouse Gene 430 2.0 probe sets as identifiers. Each identifier was mapped to its corresponding object in the IPA knowledge base (Summer 2013 release). Expression results were limited to genes having a q-value ,0.05. Gene networks were algorithmically generated based on their connectivity to the uploaded data set. Networks pertaining to phototransduction and visual pathways were retained, and expression and significance values were overlaid onto networks of interest so as to identify differential patterns of up-and down-regulated genes. In each network, molecules are represented as nodes, and the biological relationship between two nodes is represented as an edge (line). All edges are supported by at least one reference from the literature, from a textbook, or from canonical information stored in the IPA Knowledge Base. The intensity of the node color indicates the degree of up-(red) or down-(green) regulation with regards to all expression values in the dataset. Nodes are displayed using various shapes that represent the functional class of the gene product; these are defined in the figure legends. Edges are displayed with various labels that describe the nature of the relationship between the nodes (e.g., P for phosphorylation, T for transcription).

Isolation of RPE and Retina
Mice were sacrificed by carbon dioxide asphyxiation and enucleated eyes were placed in chilled phosphate buffered saline (PBS). The connective tissue, muscles and conjunctiva were carefully removed using iris scissors. A circumferential incision was made below the level of ciliary body and the anterior segment consisting of cornea, lens, iris and ciliary body was discarded. The neural retina was peeled from the RPE layer using surgical forceps and the remaining posterior eyecup consisting of the sclera, choroid and RPE was flash frozen in liquid nitrogen and stored at 270uC.

Quantitative Real time PCR
Eyes were collected at different time points (P7, P14 and P21). Total RNA was extracted from whole eyes, retina, and RPE cells using Trizol reagent (Life Technologies, Carlsbad, CA, USA) combined with the RNeasy kit (Qiagen, Valencia, CA, USA) as per manufacturer's instructions. Genomic DNA contamination of RNA was prevented by on-column treatment with DNase I (Qiagen, Valencia, CA, USA) according to the manufacturer's instructions (Qiagen, Valencia, CA, USA). One mg of total RNA was reverse transcribed using the Retroscript kit (Ambion, Life Technologies, Carlsbad, CA, USA). Quantitative real-time PCR (qRT-PCR) was performed with the SYBR Green Master Mix (Life Technologies, Carlsbad, CA, USA) and gene-specific primers using the ViiA7 Real Time PCR Cycler (Life Technologies, Carlsbad, CA, USA). Appropriate controls including no RT control was included in the assay to rule out genomic DNA contamination of the cDNA samples. A relative fold change in gene transcript was calculated using ViiA7 software V1.2.2 (Life Technologies), applying the comparative CT method (DDC T ) and was quantified using 2 -DDC T with b-actin as an internal calibrator. Melting curve analysis was performed to validate accurate amplification of the target gene. The primers used for qRT-PCR were designed using mouse qPrimerDepot (mouseprimerdepot.nci.nih.gov). The primers spanned exon-exon borders that overlapped intron(s) ( Table 1).

Western blot analysis
Whole eyes were dissected from C57BL/6J and Mfrp rd6 mice and placed in ice-cold PBS. The anterior portion of the eye including cornea, lens and ciliary body was excised. The neural retina was gently peeled from the underlying RPE layer, and the RPE layer including the choroid and sclera was flash frozen in liquid nitrogen and stored at 270uC. For preparation of tissue lysates, the aggregate RPE sample was combined with 100 ml of 16 SDS Laemmli sample buffer (Bio-Rad Laboratories, Hercules, CA, USA) containing 5% b-mercaptoethanol, and sonicated three times for 5 min using a probe sonicator 3000 Ultrasonic Liquid Processor (Misonix Incorporated, Farmingdale, NY, USA) and stored in a 270uC freezer. The protein samples were suspended using a vortex mixer. Proteins were resolved by SDS-PAGE using a 10% Tris-glycine gel, (Bio-Rad Laboratories, Hercules, CA, USA) and transferred onto nitrocellulose (Bio-Rad Laboratories, Hercules, CA, USA) by semi-dry electroblotting using the Trans Blot Turbo Transfer System (Bio-Rad Laboratories, Hercules, CA, USA) according to the manufacturer's instructions. The membrane was washed briefly in PBS and blocked with Odyssey blocking buffer, (927-40000; LI-COR Biosciences, Lincoln, NB, USA) for 1 h at room temperature and subsequently incubated with anti-RPE65 antibody (ab13826; Abcam, Cambridge, MA, USA) at 1:1000 dilution, overnight at 4uC. After four 5-min washes in 16 TBS containing 0.05% Tween20 (TBST), the blots were incubated with secondary antibodies, anti-mouse IRDye 700DX conjugated antibody (KFA011; Rockland Immunochemicals Inc, Gilbertsville, PA, USA) in blocking buffer for 45 min. Following four washes for 5 min in 16 TBST, the blots were scanned using the Odyssey Infrared Imaging system (LI-COR Biosciences, Lincoln, Nebraska, USA) at 700 nm. Blots were stripped and incubated with mouse b-actin antibody (A5441; Sigma, St. Louis, MO, USA) at a 1:10,000 dilution overnight at 4uC and processed as previously described. Odyssey Infrared Imaging System software 3.0 was used to quantify the protein bands using a standard curve after normalizing with the b-actin control. The normalized, relative change in protein concentration was expressed in arbitrary units.

In situ hybridization
Eye tissue from P14 C57BL/6J and homozygous Mfrp rd6 mice was freshly harvested and fixed in 4% paraformaldehyde for 24 h at 4uC, immersed in 70% ethanol, dehydrated and embedded in paraffin blocks [12]. Five mm sections were cut and 3-4 sections were placed on positively charged glass slides (Milennium 2000 superfrost Adhesive slides, StatLab, McCkiney, TX, USA), which were then baked at 60uC for 1 h and stored at 220uC prior to in situ hybridization. We performed in situ hybridization using both the QuantiGene ViewRNA ISH Tissue 1-Plex and 2-Plex assay according to the manufacturer's instructions (Affymetrix, Santa Clara, CA, USA). Briefly, PFA-fixed paraffin-embedded tissues were baked, dewaxed using xylene for 5 min with gentle agitation, twice, followed by two ethanol washes for 5 min each. A hydrophobic barrier was drawn on the glass slide using a template and the samples were heated at 90-95uC for 10 min, followed by protease treatment at 40uC using a Thermobrite Controlled Temperature Slide Processing System (Abbott Laboratories, Abbott Park, IL, USA) for 10 min. Both heat and protease treatments were optimized for the eye tissue and to allow probe to access the specific RNA by unmasking the RNA. Mouse Prss56 Type 1 (FastRed) probe set (VB1-15207) and Mouse Glul Type 6 probe (FastBlue) set (VB6-16850) were custom designed (Affymetrix, Santa Clara, CA, USA) for target hybridization. Each probe set contained 20 oligonucleotide pairs that were specific for Prss56 or Glul and had a binding site for TYPE-specific sequences for signal amplification. Samples were incubated with Prss56 probe and/or Glul probe or no probe control at 40uC for 3 h using the slide processer. For signal amplification, a series of sequential hybridization steps were carried out using bDNA technology and signals were detected by addition of Fast Red substrate and/or Fast Blue substrate to the tissue section as per manufacturer's instructions (Affymetrix, Santa Clara, CA, USA). Target RNA was detected at specific sites within the tissue by chromogenic substrate deposition, which was visualized by using an SP5 confocal

Histological analysis and Immunofluorescence
The procedure for histological assays was previously described [20]. Briefly, the mice were euthanized by carbon dioxide asphyxiation. The harvested whole eyes were placed in either 4% paraformaldehyde or ice-cold acetic acid/methanol solution overnight followed by paraffin embedding using standard protocol. Eyes were cut into 6 mm sections, stained using Hematoxylin & Eosin, and visualized by light microscopy. For Immunofluorescence, PFA fixed sections were used. Deparaffinized sections were incubated with Mouse anti-Glutamine monoclonal antibody (MAB302, Chemicon) at a dilution of 1:200 overnight at 4uC. Glutamine synthetase was fluorescently labeled using Donkey Anti-mouse Alexa 488 (A21202, Life technology) at a dilution of 1:200 at RT for 1 h and visualized by fluorescent microscopy (Leica Microsystems Inc).

Statistical Analysis
In this study, GraphPad Prism Version 5 software was used for statistical analyses. Student's t-test was used to calculate statistical significance (p-value) between two groups. P-values ,0.05 were considered statistically significant.

Microarray analysis of Mfrp rd6 mutant mice relative to controls
In Mfrp rd6 mutant mice, the earliest phenotypic changes, observed at P14, are disorganization and shortening of OSs [7,12]. The absence of MFRP in Mfrp rd6 RPE cells appears to cause the OSs to develop abnormally, and, combined with the observed in vivo impairment of RPE phagocytosis [12] leads in turn to progressive retinal degeneration. MFRP, therefore, plays a central role in the normal function of RPE, which is essential for photoreceptor maintenance. To determine the functional role of MFRP in the eye, gene expression profiling of homozygous Mfrp rd6 and C57BL/6J eyes was carried out at P0 and P14. Strain C57BL/6J served as control, as the Mfrp rd6 allele has been introgressed for more than seven generations on this background and genome-wide single nucleotide polymorphism analysis indicates at least 95% shared identity between B6.C3Ga-Mfrprd6/J and C57BL/6J [10].
Gene expression analysis was performed using the Affymetrix Mouse 430v2 microarray. Volcano plots were generated to graphically represent differentially accumulated gene transcripts at a significance level of q,0.05 using six pairwise analyses. Comparison of Mfrp rd6 and C57BL/6J samples at the P0 time point did not yield any significant differences in transcript levels, indicating that the effects of Mfrp mutation occur postnatally (Fig. 1A). In contrast, comparison at the P14 time point resulted in 2,454 differentially expressed probe sets (Fig. 1B). Not surprisingly, comparison of the P0 and P14 time points in Mfrp rd6 mice yielded 28,452 significant differentially accumulated probe sets (Fig. 1C), and a similar result was obtained in C57BL/6J mice (Fig. S1A). Further analysis of differential accumulated gene transcripts between the two strains (Mfrp rd6 and C57BL/6J) irrespective of the time point yielded a relatively small set of significantly accumulated probe sets (Fig. S1B). By contrast, the comparison of the difference in time point (P0 vs P14) irrespective of the strain difference yielded a relatively large set of significant differentially accumulated probe sets (Fig. S1C). Taken together, this analysis suggests that these probe sets likely change as a consequence of ocular development between P0 and P14.

Differentially accumulated probe sets in Mfrp rd6 mutant mice at P14
We further analyzed the differentially accumulated probe sets at P14 in Mfrp rd6 mutants relative to controls. Heatmaps were generated for both up-and down-regulated probe sets in Mfrp rd6 mutants. The heatmap color scale corresponds to the fluorescence (log-2, normalized) intensity level of the probe sets, where light blue represents a low level of hybridization to the probe set and dark blue a high level. Upregulated probe sets (5)

Ingenuity pathway analysis
To determine the molecular networks and biological pathways affected in Mfrp rd6 mutant eye at the P14 time point, we examined the microarray data by Ingenuity Pathway Analysis (IPA). The top five canonical IPA pathways that were altered significantly in Mfrp rd6 mutant eyes were B cell development, allograft rejection signaling, autoimmune thyroid disease signaling, phototransduction, cytotoxic T lymphocyte-mediated apoptosis of target cells, and visual cycle. The genes identified in these pathways and results of statistical tests are given in Table S1. Of particular interest are the phototransduction and visual cycle genes that were perturbed in Mfrp rd6 mutant eyes. The Mfrp rd6 allele, which is a loss of function mutation leading to the absence of MFRP protein from the RPE [8], is likely to have a direct effect on RPE cell function/ maintenance. In accordance with this effect, visual cycle gene transcripts expressed in the RPE, including Rpe65 and Lrat, were decreased significantly in Mfrp rd6 mice (Fig. 2, A, B). Transcripts of Rgr, which encode a visual cycle protein found in both RPE and Müller cells, were also significantly decreased (Fig. 2B). The Mfrp rd6 mutation also affected retina-specific transcript levels, as evidenced by a significant relative fold change (RFC) of 21.2 to 2 2.0 in transcripts expressed in photoreceptor cells (Fig. 2, A, B). These included transcripts from genes specifically expressed in rod cells (Rho, Gnb1 and Gnb5; Fig. 2A), cone cells (Opn1sw, Gnat2, Gnb3 and Gnb5; Fig. 2B), or in both rods and cones (Rgs9, Rgs9bp, Prkaca, Pde6a, Pde6b, Guca1a and Guca1b; Fig. 2, A,  B). Transcripts encoded by genes implicated in maintaining photoreceptor OS morphology, Fscn2 and Rpgrip1, were also significantly decreased in the Mfrp rd6 mutant (Fig. 2, A, B), consistent with the early OS disorganization that is observed in this mutant.
Selective analysis of our microarray data, examining 250 genes listed in the IPA retinal degeneration (RD) pathway, identified additional transcripts that were differentially regulated in Mfrp rd6 mutants compared to B6 controls at P14. Forty RD pathway gene transcripts were significantly decreased in Mfrp rd6 eyes, with an RFC from 21.15 to 22.39 (Table S2). These include the C1qtnf5 and Mertk transcripts, which are known to be decreased in Mfrp rd6 mice [12]. A further 9 RD pathway gene transcripts were significantly increased in Mfrp rd6 eyes (Table S3), while the remaining 201 failed to show significant change and therefore were not considered further. As RD pathway genes are typically expressed in the retina or RPE, these results suggest that the Mfrp rd6 mutation broadly decreases retina-and RPE-specific transcripts in the posterior eye.

Validation of differentially accumulated transcripts by qRT-PCR
To validate the microarray data, qRT-PCR analysis was performed on whole eyes from Mfrp rd6 mutants and C57BL/6J mice. Transcripts that were significantly increased in the microarray analysis are listed in (Table 2). Transcripts that were significantly decreased in the microarray analysis (Table 3),   including those encoding components of the visual cycle (Rpe65, Lrat, Rgr), phototransduction pathway (Rgs9, Pde6a, Guca1b), and involved in disc morphogenesis (Fscn2, Rpgrip1), were also reduced as determined by qRT-PCR (Fig. 3A). However, some of the changes in transcripts including Prph2 (24.1795, q value = 0.06), Optc (22.395, q,0.05), Aqp5 (22.197, q,0.05) ( Table 3) were not validated by qRT-PCR. The failure to validate these transcripts was unlikely to be caused by assay limitations. Amplification primers were targeted to exon-exon junctions to amplify only processed RNA (Table 1); a PCR product of the correct size was verified; the PCR efficiency of the primers for genes of interest and calibrator was the same; C T values in control and Mfrp rd6 mutant samples, indicated robust target amplification in both; and melting curve analysis confirmed the presence of a single amplified product. However, the lack of verification may be due the difference in samples used for the microarray and the qRT-PCR analyses, to the low abundance of some transcripts, or in some cases the differences were not significant (e.g. Prph2, which is highly expressed in the retina, had a FDR of 0.0605).

Quantitative Real Time PCR analysis of visual and phototransduction genes in Tulp1 and Rpe65 mutants
To determine if the decrease in the transcript levels of visual and phototransduction genes in the Mfrp rd6 mutant was a non-specific effect of the disease process that occurs during retinal degeneration, we examined the levels of visual cycle and phototransduction pathway transcripts in two unrelated retinal degeneration models with mutations in the retina-specific gene, Tulp1 or the RPEspecific gene, Rpe65. Homozygous Tulp1 (Tulp1 tvrm124 ) mutant mice model early onset retinal degeneration and have characteristically shorter OSs at P14 [21], comparable to those observed in Mfrp rd6 mutants. Homozygous null mutation of Rpe65 (Rpe65 tvrm148 ) results in slow retinal degeneration and disorganized OS discs at P14 [22], as observed in Mfrp rd6 mice. These three retinal degeneration models show a similar extent of photoreceptor degeneration at P14 (Fig. S4). We reasoned that if the decrease in visual and photoreceptor transcripts observed in Mfrp rd6 mutants were due to non-specific, secondary effects of OS shortening or disorganization, a similar reduction of visual cycle and phototransduction pathway gene transcripts would also be In Rpe65 tvrm148 /Rpe65 tvrm148 mutants, there was only a significant increase in RpGrip1 from transcripts tested, relative to wild-type (B6/J) controls. The data are expressed as relative fold change (RFC) after normalizing to the wild-type control. RFC was calculated using DDC T method after internal calibration to b-Actin control. Each value represents RFC 6 S.E.M. * P,0.05 and ** P,0.001 relative to controls. N = 3-6 per group. doi:10.1371/journal.pone.0110299.g003 observed. No significant decrease in either visual cycle or phototransduction pathway transcripts were found in either Tulp1 (Fig. 3B) or Rpe65 mutant mice (Fig. 3C), suggesting that the observed effects are specific to Mfrp disruption.

Western blot analysis of RPE65 protein in Mfrp rd6 eyes
The parallel decrease in retina-and RPE-specific transcripts revealed by transcript analysis raised the possibility that the Mfrp rd6 mutation might diminish retinal health by reducing the levels of RPE visual cycle proteins, which are critical for photoreceptor maintenance [23]. To test this possibility, we examined levels of the Rpe65 gene product RPE65, a 65 kDa RPE-specific isomerohydrolase that is essential for producing 11cis retinal from all-trans-retinyl esters in the visual cycle [23,24]. Western blot analysis of RPE/choroid/sclera lysates revealed decreased levels of RPE65 protein in Mfrp rd6 mice compared to C57BL/6J controls (Fig. 4A). An Rpe65 mutant (Rpe65 tvrm124 / Rpe65 tvrm124 ) was used to control for antibody specificity. No RPE65 protein was detected in lysates from the Rpe65 mutant mice, whereas a 65 kDa protein was detected in all other samples (Fig. 4A), thus confirming that the detected band is RPE65. Quantitation of the blot indicated a significant 2.0-fold decrease of RPE65 in the Mfrp rd6 mutant eyes (Fig. 4B).

Increased Prss56 transcript levels in Mfrp rd6 eyes
Microarray analysis revealed a number of transcripts that accumulated at higher levels in Mfrp rd6 mutant mice compared to wild-type controls. The highest change (17-fold) was observed in Prss56, a gene encoding a serine protease ( Table 2). Other transcripts involved in hematological (Hemgn, Ahsp, Alas2, Hba-a1/2, Gypa, Rhag and Car1) and immune (Ctse and Ighm) function, were also upregulated (Table 1). However, despite being of sufficient abundance for detection by qRT-PCR, significant differences between wild type and Mfrp rd6 mutants in the latter transcripts were not validated upon further testing.
PRSS56/Prss56 variants in human and mouse are associated with defects in ocular growth [17][18][19]25], a process that is also affected by human MFRP mutations [1][2][3][4][5][6]. Therefore, we focused on characterizing Prss56 expression in greater detail. Microarray data indicated no difference in Prss56 transcript accumulation between Mfrp rd6 mice and controls at P0, suggesting that the increase in Prss56 expression occurs during postnatal development of the Mfrp rd6 eye. To assess the temporal variation of Prss56 expression in the postnatal period, qRT-PCR was performed at three different time points (Fig. 5A). At P7, there was a 3.5-fold increase in Prss56 transcript, which increased to 14fold at P14 and 70-fold at P21 (Fig. 5A). Thus, the Mfrp rd6 mutation causes a progressive accumulation of Prss56 transcript throughout postnatal development.
To test whether the increase in Prss56 expression was a specific attribute of Mfrp rd6 mutant mice, we also examined Prss56 transcript levels by qRT-PCR in wild type and the Tulp1 and Rpe65 mutants. In P14 Tulp1 mutant mice, there was no significant change in Prss56 transcript (Fig. 5B), whereas in the Rpe65 mutant, there was a significant decrease (Fig. 5B). These results suggest that Prss56 upregulation is specific to homozygous Mfrp rd6 mice. Lastly, we also determined the wildtype level of Prss56 at P7, P14 and P21 (Fig. 5C). The wildtype level of Prss56 decreased from P7 to P21 (Fig. 5C). While there was a significant decrease in Prss56 transcript from P7 to P21, the difference from P7 to P14 was not statistically different (Fig. 5C).

Cellular localization of Prss56 and Glul in Mfrp rd6
In the absence of an antibody that could reliably detect murine PRSS56, in situ hybridization was used to determine the cellular localization of the Prss56 transcript. In the no probe control, no Prss56 transcript was observed (data not shown). In the wild-type control at P14, confocal microscopy revealed a few cells in the retinal inner nuclear layer (INL) that showed expression of Prss56 transcript (Fig. 6A, upper panel). By contrast, in Mfrp rd6 eyes at P14, intense specific staining of the transcript was observed in the INL (Fig. 6A, lower panel). This increased staining further validates both the microarray and qRT-PCR results of increased Prss56 transcripts in Mfrp rd6 mutant eyes relative to controls.
Further we performed 2-plex in situ hybridization using both Prss56 type 1 probe and Glul type 6 probe on the same eye sections. The co-localization of the Prss56 probe to the same cell bodies expressing Glul within the INL of retina in wildtype control (Fig. 6B, upper panel) and Mfrp rd6 (Fig. 6B, lower panel) directly identified Müller cells as specifically expressing Prss56.
We also stained eye sections with an antibody to glutamine synthetase, which specifically marks Müller glial cells in the INL. Comparable positive antibody staining of Müller cells was observed in the retinas of both wild type and Mfrp rd6 mice, with a cell body staining pattern in the INL very similar to that observed by in situ hybridization of Prss56 transcripts (Fig. 6C, upper and lower panels). Taken together, these results suggest that Müller cells in the INL express the Prss56 transcript and are primarily responsible for the observed upregulation of the Prss56 gene in Mfrp rd6 mice.

Discussion
Mouse models of retinal degeneration, in which the causative gene has been identified, are important tools for translational vision research [21], as they allow for in-depth study of cellular and molecular changes during development and disease progression. Such studies are especially important when the underlying function of the disrupted protein and molecular basis of the disease or pathology is unknown. Mutations in human MFRP [1][2][3][4][5][6] and mouse Mfrp [7][8][9] lead to retinal degeneration in both species and have been associated with a decrease in axial length in humans. Although nanophthalmia in Mfrp mutants has not been observed, posterior microphthalmia has yet to be assessed. The localization of MFRP to the RPE cell and ciliary body, suggests a potential role in posterior eye homeostasis, however, its function is unclear, and the molecular mechanisms by which mutations of this protein cause disease pathology are unknown. The microarray and validation studies of ocular transcripts in the mouse Mfrp rd6 mutant described in the present study provide new insights into MFRP function.

Reduced accumulation of retina-and RPE-specific gene transcripts
Microarray analysis identified a modest but significant postnatal decrease in a large number of retina-and RPE-specific gene transcripts in P14 Mfrp rd6 mutant eyes, prior to the degenerative decline in ONL thickness that is associated with photoreceptor cell loss [12]. This result indicates an effect of the mutation beyond the RPE where Mfrp is expressed. In our study, the decreased accumulation of posterior eye transcripts was a distinct feature of Mfrp rd6 mice, as qRT-PCR analysis of the Tulp1 [26] and Rpe65 [21] mutant models at P14 revealed no significant change in selected visual cycle and phototransduction pathway transcripts that were decreased in Mfrp rd6 mutants. Importantly, our Tulp1 and Rpe65 mutant models exhibit a similar time course of retinal degeneration as Mfrp rd6 and show similar OS shortening and , when compared to the Mfrp rd6 /Mfrp rd6 mutant, in Tulp1 tvrm124 /Tulp1 tvrm124 mutants, there was no significant change in Prss56 transcript, whereas in Rpe65 tvrm148 /Rpe65 tvrm148 mutants, there was a significant decrease. (C) Wildtype levels of Prss56 transcript revealed a significant decrease between P7 and P21 timpoints, whereas the decrease from P7 to P14 was not statistically significant. Data are expressed as relative fold change (RFC) in the Prss56 transcript after normalizing to the wild-type control (B6). RFC was calculated by the DDC T method using b-actin as an internal calibrator. Each value represents RFC 6 S.E.M. * P,0.05 and ** P,0.001 relative to controls. N = 3 per group. doi:10.1371/journal.pone.0110299.g005 disorganization at P14 without a change in ONL thickness. Decreased accumulation of phototransduction pathway transcripts has been documented in other retinal degeneration models [27][28][29]. However, these studies may not be directly comparable to ours, since the disease models were analyzed at ages when ONL thickness was significantly decreased [27,28] or photoreceptor outer segments were absent [29]. Gene profiling of additional models with a disease progression closely similar to that of Mfrp rd6 may be required to assess whether the decreased accumulation of visual cycle and phototransduction pathway transcripts is a truly distinguishing characteristic of the Mfrp rd6 mutation. Nevertheless, the microarray and qRT-PCR data on Mfrp rd6 and other retinal degeneration mutants suggest a broad and possibly unique role for MFRP protein in postnatal development of the RPE and retina.
A potential explanation for the widespread but modest reduction of visual cycle and phototransduction transcripts in Mfrp rd6 mutants is a defect in the postnatal development of the retina and RPE. The CUB domains, found in MFRP, are prevalent in genes that are developmentally regulated [30]. The CRD domain in MFRP has a high homology to the Frizzled family of proteins, which are normally involved in Wnt signaling and are important in RPE development [31,32]. Finally, human mutations in MFRP have been associated with nanophthalmia or posterior microphthalmia with shortening of the posterior segment of the eye [1][2][3][4][5][6], an expected consequence of reduced retinal and RPE development.
Although no one has actually assessed RPE development in Mfrp mutants, apical microvilli defects have been reported [12]. It is interesting to note that during normal eye development, there is a specific and strong increase in Rpe65 transcription that coincides with the extension of RPE microvilli and the increase in the photoreceptor OS length (reviewed in [24]). Thus, the decrease in Rpe65 transcript observed in Mfrp rd6 mice may contribute to the decrease in OS length and organization [12]. Moreover, in Mfrp rd6 mutants, the significant decrease in transcripts of Fscn2, encoding a protein involved in outer segment morphogenesis, may contribute to the failure to elaborate OS and to the disorganization of OS, followed by photoreceptor degeneration similar to that observed in the Fscn2 haploinsufficient mouse model [33]. In summary, although we do not know currently how MFRP mediates its effects on the visual cycle and phototransduction genes, it is likely that the observed reductions play a role in the pathogenesis of the disease induced by disruptions in Mfrp. Mfrp and Prss56, features of a common pathway?
In this study, we have demonstrated significant upregulation of the retina-specific Prss56 transcript in Mfrp rd6 eyes. Increased expression was observed during postnatal eye development in Müller cells of Mfrp rd6 mutants, but not in other similarly affected retinal degeneration models, suggesting that Prss56 upregulation is unique to Mfrp disruption. Prss56 encodes a trypsin-like serine protease [34] of unknown function and substrate specificity in the eye. Interestingly, PRSS56/Prss56 mutations are associated with autosomal recessive posterior microphthalmia in humans and mice [17,25,35]. Moreover, two different genome-wide association studies (GWAS) involving multi-ethnic cohorts identified Prss56 as significantly associated with refractive errors and myopia [18,19] that relate to a change in axial length. MFRP mutations in humans are also associated with posterior microphthalmia characterized by abnormal posterior segment size leading to hyperopia [2] and cause recessive nanophthalmos [36]. Studies on the postnatal progression of refractive error in nanophthalmos patients having mutations in MFRP suggest a role of MFRP protein in embryonic ocular growth and postnatal emmetropization [36]. As both PRSS56 and MFRP variants affect axial length and potentially the process of emmetropization, it is plausible that they may function through a common biological pathway, yet to be determined.
Like other serine proteases [37], PRSS56 may either directly or indirectly be involved in extracellular matrix (ECM) processing, degradation and remodeling, as suggested previously [17]. Accordingly, upregulation of Prss56 expression in Mfrp rd6 mutants may promote ECM remodeling. Matrix metalloproteinases are also thought to play an important role in eye development and disease [38]. Consistent with enhanced metalloproteinase activity, IPA analysis of microarray data revealed a significant decrease in Timp3 transcripts (Table S2), which encode tissue inhibitor of metalloproteinase-3, and a significant increase in Adamts19 (Table 1) transcripts, which encode a disintegrin and metalloproteinase with thrombospondin motif family member. Taken together, these findings suggest that altered ECM processing may contribute to the progressive loss of photoreceptor cells in the Mfrp rd6 mutant, as observed in other mouse retinal degeneration models [39].
In conclusion, the present study suggests a broad role of MFRP in determining retinal and RPE transcript levels during postnatal development. Most importantly, the upregulation of Prss56 expression in Mfrp rd6 Müller cells suggests a possible interaction between Mfrp and Prss56 in posterior eye maintenance and development during this period. Future studies would be directed toward understanding how MFRP influences transcript accumulation in the postnatal RPE and retina, and also address how MFRP and PRSS56 interact.