Differential Gene Expression in Ndph-Knockout Mice in Retinal Development

Abstract

purpose. Mutations in the NDP gene impair angiogenesis in the eyes of patients diagnosed with a type of blindness belonging to the group of exudative vitreoretinopathies. This study was conducted to investigate the differential gene expression caused by the absence of Norrin (the NDP protein) in the developing mouse retina and to elucidate early pathogenic events.

methods. A comparative gene expression analysis was performed on postnatal day (p)7 retinas from a knockout mouse model for Norrie disease using gene microarrays. Subsequently, results were verified by quantitative real-time PCR analyses. Immunohistochemistry was performed for the vascular permeability marker plasmalemma vesicle associated protein (Plvap).

results. Our study identified expression differences in Ndph ^y/– versus wild-type mice retinas at p7. Gene transcription of the neutral amino acid transporter Slc38a5, apolipoprotein D (ApoD), and angiotensin II receptor-like 1 (Agtrl1) was decreased in the knockout mouse, whereas transcript levels of adrenomedullin (Adm) and of the plasmalemma vesicle associated protein (Plvap) were increased in comparison to the wild-type. In addition, ectopic expression of Plvap was found in the developing retinal vasculature of Norrin-knockout mice on the protein level.

conclusions. These data provide molecular evidence for a role of Norrin in the development of the retinal vasculature. Expression of two genes, Plvap and Slc38a5, is considerably altered in retinal development of Norrin-knockout mice and may reflect or contribute to the pathogenesis of the disease. In particular, ectopic expression of Plvap is consistent with hallmark disease symptoms in mice and humans.

Norrie Disease (ND; OMIM 310600; Online Mendelian Inheritance in Man; http://www.ncbi.nlm.nih.gov/Omim/ provided in the public domain by the National Center for Biotechnology Information, Bethesda, MD) is an X-linked, recessive neurologic disorder that presents with congenital blindness, progressive deafness, and mental retardation ¹ and is caused by mutations in the Norrie disease pseudoglioma (NDP) gene. ^{2 3 4} The high phenotypic variability, even within one family carrying the same mutation, ^{5 6 7} suggests the involvement of modifier genes or other factors that might explain why no clear genotype–phenotype correlation has been described so far. In addition, mutations in this gene have been associated with a variety of other recessive and sporadic vitreoretinal diseases, including exudative vitreoretinopathy (EVR; OMIM 133780), retinopathy of prematurity (ROP, stages 4b and 5) and Coats’ disease (OMIM 300216). ^{8 9 10 11} These allelic clinical entities show remarkable similarities in their ocular phenotype, especially with regard to abnormalities in the retinal vasculature. 

The human disease phenotype is strongly replicated by the Norrie disease pseudoglioma homologue (Ndph) knockout mouse. ^{12 13} Although the retinal vasculature in humans develops around mid gestation but only postnatally in mice, the timing of events is comparable. In mice, retinal blood vessels start to develop around birth at the optic disc and spread radially inside the nerve fiber layer across the retina, until they reach the periphery at approximately postnatal day (p)9. In addition, vessels start to sprout into the deeper layers of the retina at p7 to form two additional networks in the plexiform layers, parallel to the superficial plexus. Contrary to the development of the retinal vasculature, the hyaloid vessel system, a transient developmental vasculature nourishing the developing lens, regresses. This process starts after its peak extension around p5 and lasts until p15, around which time the mice open their eyes, and eye development is more or less complete. ¹⁴ In the Ndph-knockout mouse, the inner retinal vessel development is severely impaired. The outgrowth of the superficial retinal vessel plexus is delayed and remains sparse, deep retinal vessels do not develop, and the regression of the hyaloid vasculature is delayed and incomplete. Furthermore, disorganization of the retinal ganglion cell layer and a reduction of retinal ganglion cells have been described. ^{12 13 15} However, proliferation of fibrovascular material in the vitreous cavity is not as massive as the pseudoglioma described in patients who have severe ND. 

Different hypotheses have been postulated about the etiology of these symptoms. It has been proposed that the extended presence of hyaloid vessels may impair the development of the retinal vasculature. ^{16 17} However, blockage of the placental growth factor (PlGF) during early postnatal stages in the eye leads to delayed regression of the hyaloid vessel system, but not to changes in the retinal vasculature, ¹⁸ indicating that persistence of the hyaloidea does not necessarily lead to defects in or lack of retinal vascularization. Another hypothesis was provided by a detailed analysis of the early development of the retinal vasculature in Ndph-knockout mice, which suggested two phases of disease progression in vivo ¹⁵ : In the early phase, the absence of functional Norrin (Ndph protein) causes a defect in sprouting angiogenesis, which leads to a delayed outgrowth of the superficial vessels and prevents the formation of deep capillary networks in the retina. Then, in the later phase, the developmental lack of the deep vasculature leads to inner retinal hypoxia. This pathologic hypoxia may explain the observed clinical features of Norrie disease and may also be responsible for the similarities of the clinical phenotypes of the aforementioned diseases. ¹⁵ In another study, the knockout phenotype could be rescued by ectopic overexpression of Norrin in the lens. ¹⁹ The results in these studies suggest a direct effect of Norrin on vascular proliferation, since proliferation of microvascular endothelial cells was increased after cocultivation with the Norrin-expressing lenses. ¹⁹  

Further, autosomal dominant and recessive forms of FEVR (familial exudative vitreoretinopathy), which are caused by mutations in Frizzled-4 (FZD4) and LDL-related protein 5 (LRP5) in approximately 35% of the patients, ^{20 21} resemble the clinical pictures of the X-linked diseases. ^{12 13 22} Prompted by these findings and the phenotypic similarities between Ndph and Fzd4-knockout mice, Xu et al. ²³ could provide the first hint for a possible cellular function of Norrin. They showed that Norrin is a high-affinity ligand of Frizzled-4 and activates the canonical Wnt/β-catenin pathway in the presence of LRP5 in cell culture. This pathway eventually leads to transcriptional regulation of target genes under the control of TCF/Lef-binding sites. In summary, these data suggest that Norrin has an influence on transcriptional regulation of Wnt target genes, is required for early angiogenic sprouting in the retina and the regression of the hyaloid vessel system, and is necessary for the formation of deep retinal capillary networks around p7. ¹⁵  

The main purpose of the experiment reported herein was to find in retinas of p7 Ndph-knockout mice differentially expressed genes that are involved in the process of deep plexus formation and that could confirm Norrin’s hypothesized role on Wnt target genes in vivo. Our findings support a role of Norrin in blood vessel development and provide evidence of a potential regulation of new target genes that were not considered so far. However, involvement of the Wnt/β-catenin pathway could not be demonstrated by our experiments. One gene, the plasmalemma vesicle associated protein (Plvap), has been characterized in more detail. This gene was described as a major structural component of fenestrated blood vessels, ²⁴ and here was found to be upregulated in the retinal vasculature in the absence of Norrin. Thus, we discuss the question of whether Norrin could also be involved in blood vessel integrity and the formation or maintenance of the blood–retina/blood–brain barrier. 

Materials and Methods

Animals

The Ndph-knockout mouse line has been described before by Berger et al. ¹² Briefly, exon 2 of the Ndph gene has been partially replaced by homologous recombination with a reverse-oriented neomycin cassette, resulting in a loss of about half the coding sequence, including the signal peptide for protein secretion. The mutation is kept on a C57BL/6 background. Genotyping was performed by PCR analysis of tail DNA. ¹² The research was performed in accordance with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research and was approved by the Veterinary Service of the State of Zurich (Switzerland). 

Tissue Isolation and RNA Preparation

Wild-type (wt) and Ndph ^{y / −} mice of different developmental stages (p5, p7, p10, p15, and p21) were killed by cervical dislocation. Five animals of each genotype were prepared, and the retinas were frozen in liquid nitrogen. Total RNA was extracted with a silica-column based kit (RNeasy-Kit; Qiagen, Hilden, Germany) according to the manufacturer’s instructions. DNase digestion was applied either directly on the column (p7 samples, RNase-free DNase Kit, cat. no. 79254; Qiagen), or separately after RNA extraction (RQ1 RNase-free DNase, p/n M610A; Promega, Madison, WI). Both retinas of one animal were pooled into one tube of lysis buffer and homogenized with a high-performance disperser (Ultra-Turrax; Ika-Werke, Staufen, Germany). Then RNA quality was determined (NanoDrop ND 1000; NanoDrop Technologies, Wilmington, DE, and a Bioanalyzer 2100; Agilent Technologies, Palo Alto, CA). The RNA integrity number for all samples was greater than or equal to 9.3. 

Microarray Experiment

Gene chip expression analysis was performed with RNA from p7 retinas (wt: n = 5; ko: n = 5), using a mouse gene microarray (GeneChip Mouse Genome 430 2.0; Affymetrix Inc., Santa Clara, CA). On this chip, more than 34,000 genes and expressed sequence tags (ESTs) are represented by ∼45,000 probe sets. The microarray experiment, from cDNA preparation to raw data processing, was performed at the Functional Genomics Center, Zurich. Briefly, total RNA samples (2 μg each) were reverse transcribed without additional amplification (one-cycle protocol), purified, labeled, and hybridized to the chip according to the manufacturer’s instructions. A gene chip scanner (GeneChip Scanner 3000; Affymetrix) was used to measure the fluorescence intensity emitted by the labeled target. 

Statistical Analysis.

Raw data processing was performed with the microarray software (GCOS, ver. 1.2; Affymetrix, Inc.). After hybridization and scanning, probe cell intensities were calculated and summarized for the respective probe sets by means of the MAS5 algorithm. ²⁵ To compare the expression values of the genes from chip to chip, global scaling was performed, which resulted in the normalization of the trimmed mean of each chip to a target intensity (TGT value) of 500 (as detailed in the statistical algorithms description document of Affymetrix [2002]). Quality control measures were considered before the statistical analysis was performed. Differentially expressed genes were calculated on computer (GeneSpring 7.2 software; Agilent Technologies), filtered for presence (per Affymetrix presence/absence flags) in four of five samples in either one condition (wt or ko), and then subjected to a nonparametric test for differential expression (Mann-Whitney test without multiple testing correction [P < 0.01], or with Benjamini-Hochberg multiple-testing correction [P < 0.1]). 

Pathway Analysis.

Pathway analysis was conducted with the online tool Pathway Express, which is part of the Onto-Tools, hosted by Wayne State University (Detroit, MI), using the default options. ²⁶  

Gene Ontology Annotations.

Gene ontology (GO) annotations are used to describe, for example, the biological process or cellular component of a given gene product. Annotations for the top differentially expressed genes (see 1 2 Table 3 ) were obtained from the Database for Annotation, Visualization, and Integrated Discovery (DAVID) from the National Institutes of Health. ²⁷  

Quantitative Reverse Transcription-PCR

Quantitative RT-PCR was performed, not only to investigate differential expression during development, but also to verify the microarray results. For p7, the same samples as in the array experiment were used. Generation of cDNA was accomplished as follows: 1250 ng total RNA was reverse transcribed in a 20-μL reaction, using random hexamers pd(n)₆ (GE Healthcare Europe, Otelfingen, Switzerland) and reverse transcriptase (SuperScript III; Invitrogen, Carlsbad, CA). For expression analysis of Slc38a5 and Mdm2, quantitative RT-PCR was conducted with 0.5 μM of a forward/reverse primer pair (binding in the area that is recognized by the Affymetrix probes; Table 1 ) and PCR master mix (SYBR-Green; Applied Biosystems, Inc. [ABI] Foster City, CA). Commercial probes and PCR Master mix (TaqMan, Applied Biosystems, Foster City, CA) were used for quantification of Aass, Centd3, ApoD, Cldn5, Agtrl1, Adm, and Plvap expression (Table 1) . In each reaction, 10 ng of transcribed total RNA was used (except: Slc38a5/100 ng). 

Quantitative RT-PCR was performed in five replicates (Prism 7900HT; ABI) using the recommended standard cycling conditions: 45 cycles of denaturation (15 seconds/95°C) and annealing/extension (1 minute/60°C). Gapdh and 18S (TaqMan; ABI) or 28S (SYBR-Green; ABI) rRNA were used as endogenous controls. Confidence intervals for wt and ko gene expression were calculated with Student’s t-test on the basis of the mean ^ΔΔC_t values, at a significance level of 95%. 

Immunohistochemical Staining and Histology

Cryosections were prepared from male wt and Ndph ^{y / −} mice aged p3, p5, p7, p10, p15, or p21. The eyes were dissected, fixed for 1 hour in PBS/4% para-formaldehyde, washed in PBS, and cryoprotected in sucrose (30 minutes in 10%, then overnight in 20%) at 4°C. They were embedded in OCT compound (Tissue-Tek; Sakura Finetek, Zoeterwoude, The Netherlands), submersed in liquid nitrogen, cut into 8-μm sections with a cryostat (CM 3050S; Leica, Heerbrugg, Switzerland), and transferred onto microscope glass slides. After the sections were dried for at least 3 hours, unspecific binding sites were blocked by incubation with 5% normal sheep serum (Sigma-Aldrich, Steinheim, Germany) in PBST for 1 hour. Slides were incubated with primary antibodies in blocking solution (overnight at 4°C) in a wet chamber. Antiserum concentrations were 1 μg/mL for anti-Plvap/MECA-32 (rat anti-mouse; Developmental Studies Hybridoma Bank, Iowa City, IA) and 7 μg/mL for anti-collagen IV (polyclonal Col IV antibody, rabbit anti-mouse, no. 2150-1470; AbD Serotec/MorphoSys, Oxford, UK). The sections were washed and incubated with secondary antibodies in PBS (90 minutes, room temperature [RT]). Cy3-conjugated anti-rabbit IgG serum (Jackson ImmunoResearch Europe, Newmarket, Suffolk, UK) was used at a concentration of 5 μg/mL, Alexa Fluor 488-labeled anti-rat IgG/M (no. A11006; Invitrogen-Molecular Probes, Leiden, The Netherlands) was used at 2 μg/mL. Slides were washed a last time and mounted in a DAPI-containing medium (Vectashield Mounting Medium for Fluorescence with DAPI H-1200; Vector Laboratories, Burlingame, CA). Images were taken with a microscope (Axioplan 2; Carl Zeiss MicroImaging, Jena, Germany) equipped with a digital camera (AxioCam HRc; Carl Zeiss MicroImaging). Contrast in Plvap/AF 488 images was enhanced after acquisition with a photoediting program. 

For retinal flatmounts, the eyes were dissected and fixed for 15 minutes in PBS/4% para-formaldehyde. The retinas were removed and postfixed for an additional 15 minutes. After the retinas were washed in PBS, unspecific binding was blocked by 10% normal sheep serum in PBST for 1 hour. Samples were then incubated overnight with the collagen IV antibody. They were washed six times for 1 hour in PBST and incubated overnight with the secondary Cy3-labeled antibody. After repetition of the washing procedure, the retinas were finally flatmounted on microscope slides. 

Sections (8 μm) of paraffin-embedded eyes were hematoxylin-eosin–stained as described elsewhere ¹⁵ and observed under bright-light illumination. 

Results

In this study, we used the Ndph-knockout mouse to investigate molecular pathways involved in early angiogenic sprouting in the retina and the formation of deep retinal capillary networks at around p7. ¹⁵ We hypothesized that Norrin is required for these processes, probably by transcriptional regulation of Wnt/β-catenin target genes and/or other pathways. 

Retinal Morphology at p7

At p7, conventional microscopy of HE-stained retinal sections did not reveal apparent differences in the retinal morphology of wt and Ndph-knockout mice (Figs. 1A 1B) . In contrast, Col IV immunohistochemistry revealed that the retinal vasculature of the Ndph knockout was strikingly altered, showing an intermediate phenotype to what we earlier observed at p5 and p10. ¹⁵ The superficial vessel system appeared much sparser than in the wt, its outgrowth was delayed, and missing vascular sprouts indicated a lacking development of deep retinal vessels (Figs. 1C 1D 1E) . Since we argue that the vascular phenotype is the earliest and most obvious retinal defect caused by the absence of Ndph, we used this developmental stage to investigate the initial pathologic events on the molecular level (i.e., to identify genes involved in the initiation of the deep retinal capillary network by sprouting angiogenesis), which may be regulated by Norrin directly or indirectly. 

Microarray Gene Expression Analysis

To identify differentially expressed genes in the retina of Ndph-knockout mice, we took a global gene expression approach with microarrays. The gene chip experiment resulted in 26,817 expressed probe sets in p7 retinas. Differences in the overall expression levels between Ndph ^y/− and wt retinas were rather moderate, both in absolute number and change in expression. In microarray studies, usually an arbitrary change cutoff of 2.0-fold was applied, ²⁸ still resulting in dozens or hundreds of differentially expressed genes. Yet, expression changes of highly expressed genes may still be biologically relevant despite being lower, while weakly expressed transcripts may show a change higher than twofold because of their greater inherent measurement error. ²⁸ In this study, we reasoned that a cutoff of 1.6-fold might still be valid considering our use of five biological replicates (instead of the standard three) while giving us a decent list of differentially expressed genes. 

The list of expressed probe sets was subjected to a nonparametric Mann-Whitney test with multiple testing correction (Benjamini-Hochberg, P < 0.1). This resulted in a list of 450 probe sets (with 10% = 45 false positives), 34 of which showed a change of at least 1.6-fold. Each of the 34 represents a different gene (Table 2) , including five currently unknown cDNAs. Sixteen transcripts showed an elevated expression in the ko mouse, whereas levels of 18 genes were decreased. 

Since only one of these 34 genes (Apcdd1) has been associated with the Wnt-signaling pathway so far, we attempted to use a pathway analysis to find evidence of its involvement. Because the list of 34 genes was too small for this purpose, we subjected the list of expressed probe sets to a nonparametric Mann-Whitney test (P < 0.01), but this time without multiple testing correction and change cutoff. This resulted in a list of 872 probe sets (730 genes or ESTs) to be considered differentially expressed between wt and ko. We then performed a pathway analysis based on these 730 genes by using the online tool Pathway Express. ²⁶ Also, the involvement of Wnt signaling was not striking. The five most prominent affected pathways were: the MAP-kinase pathway, focal adhesion, calcium-signaling, and tight junction, and only then, Wnt signaling (Table 3) . Because of the low statistical power of this analysis, these results must be treated with caution. Although they may provide clues for future research activities, they have to be verified first by independent experiments. In the study presented herein, we decided to concentrate on the list of the 34 most differentially expressed genes (Table 2) . 

As expected for a ko, the Ndph gene appeared in the list of lower expressed genes. However, its expression was only decreased by a factor of two. A possible explanation could be the general low retinal Ndph expression in the wt, so the indicated change could be the result of a measurement error. Second, the ko is not a complete null allele, but rather lacks most of exon 2, resulting in the loss of about half its coding region. Since the probes of the microarray (Mouse Genome 430 2.0; Affymetrix) predominantly target the 3′-region and a signal was detected in the ko mice, it is possible that a transcript including this region is present in the retina of Ndph ^{y /−} mice. In fact, a ko-allele–derived, low-expression artificial transcript has been isolated from the brain of Ndph ^y/− mice. Partial sequencing of the transcript revealed that no functional Norrin can be translated from this RNA, as almost half of the open reading frame, including the signal peptide, is missing and no start codon is present in the relevant transcript area (data not shown ²⁹ ). 

Verification of Differential Expression by qRT-PCR

To validate the microarray results, qRT-PCR was performed for three of the most upregulated and six of the most downregulated transcripts in the same p7 retina RNA samples that were used for the array experiment. Relative quantification values were determined with 18S or 28S rRNA and Gapdh (not shown) as endogenous controls. Expression levels in general seemed to correlate with the microarray results (Fig. 2) . Five genes were proven to be significantly differentially expressed. Three genes showed a decreased expression in the ko: the neutral amino acid transporter Slc38a5 (24× less), apolipoprotein D (ApoD; 2.4× less), and angiotensin II receptor-like 1 (Agtrl1; 2× less). Increased expression was found for the plasmalemma vesicle associated protein (Plvap; 4.2× higher) and adrenomedullin (Adm; 2.7× higher). 

Despite the low change threshold of 1.6-fold for our microarray statistical analysis, we were able to confirm differential expression of more than half the genes examined. 

Expression during Development

We further investigated expression for the two most differentially expressed verified genes, Plvap and Slc38a5, during postnatal retinal development. In addition to p7, qRT-PCR was conducted on retinal cDNA from p5, p10, p15, and p21 (Fig. 3) . 

Plvap expression in the ko was significantly increased at every developmental stage in comparison to the wt. The difference increased about one order of magnitude in later stages (p15/p21) when compared to the earlier stages (p5–p10; Fig. 3A ). 

Slc38a5 expression was significantly decreased at every developmental stage in the Ndph knockout (Fig. 3B) . The change was about 10 times lower at p5 and p15, about 25 times lower at p7 and p10 and about 3 times lower at p21. Since the Slc38a5 locus is located within 8.6 Mb from the Ndph locus, we investigated whether the observed difference was caused by a different genetic background, because the knockout Ndph allele is derived from the 129P2 mouse strain, which has then been backcrossed to the C57BL/6 background. We could not exclude a background effect by analysis of a genetic marker inside the Slc38a5 locus (rs13483703), since it indicated the 129P2 allele at this position. Thus, we studied the expression of Slc38a5 by additional qRT-PCR on p7 retinas of 129P2 wt males and found similar levels as in the C57BL/6 wt group (data not shown). Consequently, we excluded an influence of the genetic background on the expressivity of Slc38a5.  

In summary, these developmental studies, which were performed on array-experiment independent RNA samples, added additional proof to the obtained microarray gene expression data. 

Immunohistological Staining of Plvap

As gene expression differences do not always reflect consequences at the protein level, immunohistochemistry was also performed for Plvap at all developmental stages that have been investigated on the transcript level (plus p3 in addition). Eye cryosections of Ndph ^y/− and wt mice were coimmunolabeled with antibodies directed against Plvap, a highly specific marker for endothelial cells (MECA-32), ³⁰ and collagen type IV (Col IV) to visualize the vascular beds ²³ (Fig. 4) . In all stages investigated (p3-p21), Col IV staining has been observed in the choroid and the superficial retinal vasculature of both genotypes. Vessels of the inner retina, located in the outer and inner plexiform layers, were detected in later stages in the wt, but not the ko retina, as has been described before. ^{13 15}  

Plvap was localized similarly to the choroidal vessel system in both wt and ko animals at all investigated developmental stages. However, the Plvap staining pattern of the retinal vasculature differed considerably. In the wt, labeling was very faint at most. Traces were observed in the superficial network at p3, but not in later stages, and the deep vessel system appeared to be slightly stained only at p10 (Fig. 4 , arrows). In contrast, Plvap staining in the ko was detected as early as p3, but even more obvious and with increasing intensity at all later stages. 

Taken together, these data show that Plvap expression in Ndph ^y/– retinas is increased not only on transcript, but also on protein level, which in particular seems to be confined to the retinal vasculature. 

Discussion

Norrin and the Role of Wnt Signaling

When Norrin has been shown to bind to the Wnt receptors Frizzled-4 and LRP-5, ²³ it was suggested that signaling occurs through the canonical Wnt/β-catenin pathway. Surprisingly, our study suggested only one gene among the most differentially transcribed genes besides Ndph to be associated with this pathway: adenomatosis polyposis coli downregulated 1 (Apcdd1). This does not necessarily mean that Norrin is not exerting its function over binding to Wnt receptors, but it may imply that Norrin does not work primarily as a transcriptional regulator via the canonical Wnt/β-catenin signaling. However, this conclusion relies heavily on the number of currently known Wnt target genes and on the assumption that this signaling actually occurs in p7 retinas, so it is still possible that this pathway is more prominent at another developmental stage. Furthermore, since we were not able to detect expression differences of two major angiogenesis-related genes, albeit they have been shown to be regulated by Wnt signaling (Tie-2/Tek ³¹ and Vegfa ³² ), it is possible that Norrin has a direct role in controlling expression of Wnt target genes in a spatially and temporally restricted manner which was not detectable by our experimental approach. 

On the other hand, we were able to confirm differential expression of five genes that so far were not shown to be Wnt targets. 

Plvap as an Early Indicator for Vascular Permeability

Our results suggest an early involvement of Plvap (plasmalemma vesicle associated protein) in the pathogenesis of NDP-related retinopathies. We found increased mRNA expression of Plvap throughout development (Fig. 3A)and also conclusive differences between immunostained cryosections of wt and Ndph ^y/− mice (Fig. 4) . Plvap is endothelial cell specific and has been described as integral membrane glycoprotein associated with the stomatal diaphragms of caveolae, transendothelial channels, vesiculovacuolar organelles, and the diaphragms of endothelial fenestrae. ^{24 33} Enhanced prevalence of endothelial fenestrae may result in reduced vascular integrity. Both have been observed in retinas of Ndph ^y/− mice as early as p14. ^{13 15} Plvap expression further has been shown to be negatively correlated with the differentiation of the blood–brain barrier. ³⁰ Hence, elevated expression of Plvap may indicate enhanced vascular permeability and break-down of the blood–retina barrier. It has been proposed that blood vessel leakiness is a result of hypoxia-driven VEGF (vascular endothelial growth factor A) upregulation ¹⁵ and thus is a secondary effect due to low oxygen supply. Of interest, Plvap itself has been suggested to be a VEGF target, ³⁴ rendering a correlation between vessel leakiness and Plvap more likely. Thus, elevated expression of Plvap in later developmental stages (p15 onward; Fig. 3A , Fig. 4 ), when a malformed/underdeveloped vasculature causes a pathologic hypoxic condition in the retina, was not surprising. Development of the superficial vessel system, however, is driven by a physiological hypoxia through induction of a periphery-to-center gradient of VEGF. ¹⁴ Noteworthy, this hypoxia-driven VEGF expression does not, or only marginally, yield visible Plvap expression in the wt situation that is detectable by immunohistochemistry. Therefore, we think that an increased Plvap expression in the early developmental stages (p3–p10; Fig. 3A ; Fig. 4 ) of Ndph ^y/− mice is not secondary to low oxygen supply and subsequent VEGF upregulation, but rather to a lack of Norrin that may have an important influence on vascular development. This theory is supported by the observation that Plvap upregulation is not solely dependent on hypoxia or VEGF, but is also triggered by other activation of the Erk1/2 MAPK pathway. ³⁵ In addition, fenestrae induction has been shown to be VEGF-independent, as it is regulated through cytoskeletal remodeling by actin depolymerization. ³⁶ However, we could not detect differences in total- and phosphorylated ERK1/2 MAP-kinase levels on Western blot analysis with retinal protein extracts from p7 Ndph ^y/− wt and ko animals so far (data not shown). In addition, differences in Plvap expression may be due to a mechanism of endothelial cells that compensates for the impaired blood vessel development, because evidence has been provided before that Plvap has a role in angiogenesis. ³⁷  

Thus, it remains to be shown whether a possible mode of action for Norrin, besides transcriptional regulation via the Wnt/β-catenin pathway, may be a direct modulation of the microtubule cytoskeleton ³⁸ or Plvap itself. Although we are aware of these alternatives, we hypothesize that transient Plvap expression and fenestration occur during maturation of the retinal vasculature and that Norrin signaling is possibly required for its suppression. This transient effect may be indicated by the faint staining of the superficial and deep vasculature in the wt, which was observed only during early development of the respective network (p3 and p10), and the retinal expression pattern in the Ndph ^y/− ko. Since mRNA and protein levels in the ko have been higher at each developmental stage examined, Norrin may act as a suppressor of Plvap expression, and thus may be involved in sprouting angiogenesis as well as in the formation and maintenance of the blood–retina or blood–brain barrier. 

Differential Gene Expression and Impaired Blood Vessel Development

Besides Plvap, we were able to confirm differential expression of four more genes. All but one of them seem to emphasize Norrin’s hypothetical role in blood vessel development. Agtrl1 (angiotensin II receptor-like 1) is a venous marker ³⁹ that has been shown to be important in retinal angiogenesis. ⁴⁰ Therefore, its twofold reduction correlates well with the observed retinal phenotype of the Ndph ko mouse. However, we cannot exclude that this decreased expression is merely an indicator for the already reduced overall vessel density or, considering the venous/capillary nature of the deep vessel system, ¹⁴ a lack of development of this particular network. The same applies to the 2.4-fold reduction of ApoD (apolipoprotein D). Nevertheless, ApoD could be implicated in the observed angiogenic defect of Ndph ^y/− mice, because it has been reported to stimulate proliferation and migration of vascular smooth muscle cells, ⁴¹ probably through modulation of the cellular response to Pdgf-bb. ⁴² ApoD expression was suggested to be induced by PDGF-BB itself. ⁴³ Of note, Pdgfb and Pdgfrb expression has been shown to be decreased in Ndph ^y/− mice in early stages (p5/p10), ¹⁵ which could indicate reduced angiogenic activity. Although Pdgfb has been shown to be regulated by VEGF, ⁴³ Vegfa levels were not different at p5 and even higher at p10, ¹⁵ suggesting an alternative, VEGF-independent transcriptional regulation of Pdgfb, and thus ApoD, in Ndph-knockout mice. The possibility of a functional relationship between Norrin and ApoD may be supported by the finding of coexpression of these two genes in disease-affected organs. APOD has been shown to be upregulated in human endometrium during implantation, ⁴⁴ a process that is disturbed in female homozygous Norrin-knockout mice. ⁴⁵ Further, ApoD expression was also reported in the inner ear, another affected organ in patients with Norrie disease and mice, ⁴⁶ where it has been suggested to be implicated in cochlear fluid homeostasis. ⁴⁷ Taken together, Norrin may influence angiogenesis either by direct transcriptional regulation of ApoD, or indirectly via Vegfa-independent regulation of Pdgf-β. 

Transcript levels of adrenomedullin (Adm) elevated nearly threefold were observed in the ko mice by quantitative RT-PCR. It has been described to be a hypoxia-induced vasodilating peptide ⁴⁸ that may have a vascular protective function. ⁴⁹ Adm probably exerts its protective role through a reduction of oxidative stress ⁵⁰ and has further been shown to inhibit vascular remodeling. ⁵¹ It is tempting to speculate that elevated expression of Adm itself could lead to a developmental defect due to its antiangiogenic properties, and that improper transcriptional regulation may be the cause of it. Hence, also differential expression of Adm correlates well with the Ndph ^y/− phenotype. 

Surprisingly, and in contrast to the other genes discussed, the gene most differentially expressed has not been linked directly to angiogenesis. Slc38a5 (solute carrier family 38, member 5), which showed a 25-fold decreased expression at p7 in Ndph ^y/– mice (Fig. 3B) , has been reported to be a main glutamine transporter in retinal Müller cells (system N2). ⁵² It therefore may play an important role in neuronal signal transduction through regulation of the glutamate/glutamine household. Its decreased expression in Ndph ^y/– mice may indicate an involvement of Müller glia cells as a target of Norrin signaling. This effect may be reflected by the reduced synaptic activity that has been observed by ERG measurements in later stages ⁵³ and is consistent with our observation that Slc38a5 expression is decreased at every developmental stage investigated, including the rather mature p21 (Fig. 3B) . Besides their importance in neural function, Müller cells also play a role as guidance structures for the developing deep retinal vasculature. ¹⁴ As such, one could speculate that communication between Müller cells and blood vessels may be disturbed in Norrin-knockout mice. Of interest, it has been reported that blood vessels of the ganglion cell layer are frequently not surrounded by glia in p14 Ndph ^y/– mice. ¹³ However, another possible explanation for reduced Slc38a5 expression could be a secondary, neuroprotective response of Müller cells to oxidative stress. Although this seems to be a likely mechanism at later, hypoxic stages in development (p15/p21), we think a 10-fold decreased expression of Slc38a5 at p5 is unlikely to be exclusively the result of hypoxia-induced neuroprotection (Fig. 3B) . Besides this, no evidence for hypoxia at p5 has been provided so far. Western blot analysis for the glial stress marker GFAP and the hypoxia-inducible factor 1-alpha (Hif1a) were shown to be similar up to p15 and p10, respectively, and no significant expression differences could be found for VEGF ¹⁵ at this stage. But Slc38a5 was not only expressed in the retina: It is also abundant in astrocytic end feet in the brain, the highest levels being found in neocortex, hippocampus, striatum, and the cerebellum. ⁵⁴  

Thus, differential expression of this molecule may be a hint of a pathogenic mechanism in the brain that could be associated with the mental retardation phenotype of ND. Noteworthy in this context, a microdeletion including SLC38A5 and the neighboring FTSJ1 has been found in three brothers with moderate to severe mental retardation. ⁵⁵  

Conclusion

Consistent with data obtained from other developmental stages, ¹⁵ our examination of p7 Ndph-knockout mice showed insufficient retinal capillarization and lack of deep vessel formation. For the first time, vascular permeability was observed not only in later, nearly adult stages, but was noted throughout development, as indicated by elevated mRNA and protein expression data for Plvap. 

Because of the early developmental stage investigated in our study and the lack of neovascularization in later stages, we tend to consider a role of VEGF itself in the early pathogenesis of NDP-related diseases rather unlikely and ask whether Norrin instead might modulate the cellular response to VEGF. Norrin’s hypothetical role as a transcriptional regulator of Wnt/β-catenin target genes is not obvious from the array experiment. However, our recent and previous studies suggest that Norrin is important for retinal angiogenesis. In the present study, we identified several genes most likely involved in blood vessel development that may be transcriptionally regulated by Norrin, directly or indirectly. These genes include Agtrl1, ApoD, and Adm, whereas Slc38a5 may represent an interesting candidate with regard to future research of aberrant neuronal development and the massive reactive gliosis that may be the basis for the often described pseudogliomas in ND patients. ⁵⁶ Finally, our findings suggest an important contribution of Plvap to the pathogenesis of Norrin-associated diseases, because of its proposed role in angiogenesis and blood vessel integrity. ^{24 37}  

²  Present affiliation: Division of Molecular Therapy, University College London Institute of Ophthalmology, London, United Kingdom.

 Supported by Grant 3100A0–109890 (WB) from the Swiss National Science Foundation.

 Submitted for publication January 14, 2008; revised June 27 and September 2, 2008; accepted December 4, 2008.

 Disclosure: N. F. Schäfer, None; U.F.O. Luhmann, None; S. Feil, None; W. Berger, None

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

 

Table 1.

 

Primers for SYBR-Green Assays and Part Number of TaqMan Assays Used in This Study

Gene	5′ Primer	3′ Primer
Mdm2	AAG ACA GGC TCT CAC TAT TAG CTA TGG	TGG GAG TTA AAG GTC TGC CTG AT
Slc38a5	CGA CCT TTG GAT ACC TCA CCT T	TGG GTG TAC ATT TCC AGC ATC T
28S rRNA	TTG AAA ATC CGC GGG AGA G	ACA TTG TTC CAA CAT GCC AG
Gapdh	AAC GAC CCC TTC ATT GAC	TCC ACG ACA TAC TCA GCA C

Gene	TaqMan Probe	Gene	TaqMan Probe
Aass	Mm00497118_ml	Adm	Mm00437438_gl
Agtrl1	Mm00442191_sl	Plvap	Mm00453379_ml
ApoD	Mm00431817_ml	18S rRNA	p/n 4308329
Centd3	Mm00551866_ml	Gapdh	p/n 4352339E
Cldn5	Mm00727012_sl

Figure 1.

 

Retinas of p7 Ndph ^y/– mice in comparison to wt. (A, B) Hematoxylin-eosin (HE)–stained central sections of paraffin-embedded retinas. No obvious difference in retinal organization was detected. (C, D) Immunofluorescence images from retinal flatmounts. Images compiled from a z-stack of pictures (acquired with an ApoTome; Carl Zeiss Microimaging, Jena, Germany). To visualize the retinal vasculature, an antibody against collagen IV was used. In the wt p7 retina, deeper networks started to develop from veins and capillaries (arrowheads), but not arteries, as has been shown in other work. ¹⁴ In contrast, outgrowth of the superficial capillary network in the ko mouse was delayed and appeared to be much sparser, and no developing deeper network was detected. (E, F) Higher magnification of the images from (C, D). Red: superficial vasculature; green: deep vasculature. GCL, ganglion cell layer; IPL, inner plexiform layer; INL, inner nuclear layer; OPL, outer plexiform layer; ONL, outer nuclear layer; A, artery; V, vein.

Table 2.

 

Genes with the Most Change in Expression

Increase	Symbol	Name	Gene Ontology (GO)			Affymetrix ID
							Biological Process (BP)	Cellular Component (CC)
3.06	Mdm2	Transformed mouse 3T3 cell double minute 2	Traversing start control point of mitotic cell cycle, protein ubiquitination	Cytoplasm, nucleus	1457929_at
2.79	Wdr67	WD repeat domain 67			1442286_at
2.47	Seh1l	SEH1-like (S. cerevisiae)			1441857_x_at
2.03	Adm	Adrenomedullin		Extracellular space	1447839_x_at
2.00	Lilrb4	Glycoprotein 49 B			1420394_s_at
1.95	Arhgap10	Rho GTPase activating protein 10	Cytoskeleton organization and biogenesis		1442967_at
1.88	Iqsec2	IQ motif and Sec7 domain 2			1448063_at
1.86	Plvap	Plasmalemma vesicle–associated protein		Integral to membrane	1418090_at
1.84	Foxp1	Forkhead box P1	Regulation of transcription	Nucleus	1446280_at
1.81	A2m	Alpha-2-macroglobulin	Protease inhibitor activity (MF)		1434719_at
1.80	Gabarapl2	GABA-A receptor–associated protein-like 2	Intra-Golgi vesicle–mediated transport	Golgi, cytoskeleton	1459477_at
1.76	Sync	Syncoilin	Intermediate filament-based process	Cytoplasm, intermediate filament, Z-disc, synapse	1432350_at
1.67	Acta2	Actin, alpha 2, smooth muscle, aorta	Cytoskeleton organization and biogenesis	Actin filament, cytoskeleton	1416454_s_at
		Unknown genes
2.30		A130078K24Rik			1444482_at
2.04		E330034G19Rik			1433798_s_at
1.66		AW492805			1457729_at

Decrease	Symbol	Name	Gene Ontology (GO)			Affymetrix ID
							Biological Process (BP)	Cellular Component (CC)
14.03	Slc38a5	Solute carrier family 38, member 5	Amino acid transport, oxygen transport	Integral to membrane	1454622_at
2.57	Panx2	Pannexin 2		Gap junction, integral to membrane	1440350_at
2.44	Mfsd2	Major facilitator superfamily domain containing 2		Integral to membrane	1428223_at
2.13	Aass	Aminoadipate-semialdehyde synthase	Response to oxidative stress, electron transport	Mitochondrion	1423523_at
2.06	Apod	Apolipoprotein D	Transport	Extracellular space	1416371_at
2.05	Centd3	Centaurin, delta 3	Regulation of cell shape and GTPase activity, negative regulation of cell migration and Rho protein signal transduction	Cytoplasm, lamellipodium, ruffle, nucleus	1419833_s_at
2.03	Cldn5	Claudin 5	Cell–cell adhesion	Tight junction, integral to membrane	1417839_at
2.01	Ndph	Norrie disease homolog		Extracellular space	1449251_at
1.92	Agtrl1	Angiotensin receptor-like 1	G-protein coupled receptor protein signaling pathway	Integral to membrane	1438651_s_at
1.85	Plekhh1	Pleckstrin homology domain containing, family H (with MyTH4 domain) member 1	Cell adhesion	Cytoskeleton	1435053_s_at
1.85	Sh2d3c	SH2 domain containing 3C	Intracellular signaling cascade	Cytoplasm	1430098_at
1.76	Zfp206	Zinc finger protein 206			1438787_at
1.72	Eltd1	EGF, latrophilin seven transmembrane domain containing 1	G-protein-coupled receptor protein signaling pathway, neuropeptide signaling pathway	Integral to membrane	1418059_at
1.67	Vwa1	von Willebrand factor A domain containing 1		Extracellular space	1426399_at
1.66	Apcdd1	Adenomatosis polyposis coli down-regulated 1			1418383_at
1.62	Slc7a1	Solute carrier family 7 (cationic amino acid transporter, y+ system), member 1	Amino acid transport	Integral to membrane	1454992_at
		Unknown genes
1.89		6230424C14Rik			1441972_at
1.88		2310015A05Rik			1436194_s_at

 GO terms listed may not include all annotations for each gene, but rather only parent or main annotations to facilitate a first impression of possible gene function.

Table 3.

 

Top Five Involved Pathways

Rank	Pathway Name	Impact Factor	Genes in Pathway (n)	Input Genes in Pathway (n)	Pathway Genes on Chip (n)	Input Genes in Pathway (%)
1	MAPK signaling pathway	23.79	233	11	231	1.51
2	Focal adhesion	20.00	156	7	143	0.96
3	Calcium signaling pathway	18.93	180	8	172	1.10
4	Tight junction	13.67	125	7	114	0.96
5	Wnt-signaling pathway	12.77	141	6	139	0.82
	Genes mapped to pathways	75
	cDNAs	868

 Only 10% of the input genes could be mapped to a pathway.

Figure 2.

 

Transcriptional analysis of selected genes in retinas of Ndph ^y/– knockout (n = 5) and wt (n = 5) mice. The average expression value of the wt group was set to 1. Error bars, 95% confidence interval (CI; P < 0.05). Quantitative RT-PCR results from p7 retinas for genes that showed increased or decreased expression in the array experiment are depicted in (A) or (B), respectively. (A) Expression of Adm (2.7×) and Plvap (4.2×) was significantly increased at p7. (B) The neutral amino acid transporter Slc38a5 (25×), ApoD (2.4×), and Agtrl1 (2×) show significantly decreased expression at p7. Significant difference in Claudin 5 (Cldn5) expression is borderline. 18S/28S rRNA were used as endogenous controls.

Figure 3.

 

Transcriptional analysis of the two most differentially expressed verified genes in retinas of Ndph ^y/– knockout (n =5) and wt (n = 5) mice. The average expression value of the wt group was set to 1. Error bars, 95% CI (P < 0.05). (A) Plvap expression is significantly increased at every developmental stage from p5 to p21. The difference between wt and ko increased about one order of magnitude in the later stages (p15/p21) when compared to the earlier stages (p5–p10). (A, inset) expression data for p5 to p10 in a different scale. (B) Slc38a5 expression is significantly decreased at every developmental stage. Variability of wt expression is higher in later stages (p10–p21) than in early stages (p5, p7). The following endogenous controls were used: (A) 18S rRNA; (B) Gapdh.

Figure 4.

 

Immunofluorescence images of retinal cryosections of Ndph ^y/– mice in comparison to wt. For each one of the indicated developmental stages (p3, p5, p7, p10, p15, and p21), co-immunolabeling with antibodies directed against Plvap (AF488; green) and collagen IV (Cy3; red) was performed. Collagen IV localizes to the extracellular matrix of endothelial cells and serves as a marker for blood vessels. In all stages investigated, staining was observed in the choroid (CV, below the RPE) and the superficial retinal vasculature (RV, inside the RGC/nerve fiber layer). From p10 onward, blood vessels of the secondary, deep retinal vasculature were detected in the OPL in the wt, but not the ko retina. In addition, at p15 and p21, the tertiary, intermediate vessel system in the IPL was seen in the wt. Plvap localized to the choroid vessel system (CV) in both wt and ko at all developmental stages. Of interest, Plvap expression was also detected in the retinal vasculature (RV) of the knockout animals beginning at p3 and more intense over time. In contrast, the retinal vasculature of the wt was not, or was only very weakly, labeled by the Plvap antibody. In the superficial vessel network, staining in the wt was detected only at p3 (arrows), but not in later stages. At p10, slight staining of the deep vessel system was observed (arrows). RV, superficial/deep retinal vasculature; CV, choroidal vasculature; RGC, retinal ganglion cell layer; IPL, inner plexiform layer; INL, inner nuclear layer; OPL, outer plexiform layer; ONL, outer nuclear layer; RPE, retinal pigment epithelium.

The authors thank Marzanna Künzli from the Functional Genomics Center Zurich for conducting the microarray experiment, from reverse transcription to raw data processing. The a-Plvap/MECA-32 monoclonal antibody, developed by Eugene C. Butcher, was obtained from the Developmental Studies Hybridoma Bank developed under the auspices of the National Institute of Child Health and Human Development (NICHD) and maintained by the Department of Biological Sciences, University of Iowa (Iowa City, IA).