A CNS-Specific Hypomorphic Pdgfr-Beta Mutant Model of Diabetic Retinopathy

Abstract

Purpose.: A mouse mutant identified during a recessive N-ethyl-N-nitrosourea (ENU) mutagenesis screen exhibited ocular hemorrhaging resulting in a blood-filled orbit, and hence was named “redeye.” We aimed to identify the causal mutation in redeye, and evaluate it as a model for diabetic retinopathy (DR).

Methods.: The causative gene mutation in redeye was identified by haplotype mapping followed by exome sequencing. Glucose tolerance tests, detailed histologic and immunofluorescence analyses, and vascular permeability assays were performed to determine the affect of redeye on glucose metabolism, pericyte recruitment, and the development of the retinal vasculature and blood–retinal barrier (BRB).

Results.: A mutation was identified in the Pdgfrb gene at position +2 of intron 6. We show that this change causes partial loss of normal splicing resulting in a frameshift and premature termination, and, therefore, a substantial reduction in normal Pdgfrb transcript. The animals exhibit defective pericyte recruitment restricted to the central nervous system (CNS) causing basement membrane and vascular patterning defects, impaired vascular permeability, and aberrant BRB development, resulting in vascular leakage and retinal ganglion cell apoptosis. Despite exhibiting classic features of diabetic retinopathy, redeye glucose tolerance is normal.

Conclusions.: The Pdgfrb^{redeye/redeye} mice exhibit all of the features of nonproliferative DR, including retinal neurodegeneration. In addition, the perinatal onset of the CNS-specific vascular phenotype negates the need to age animals or manage diabetic complications in other organs. Therefore, they are a more useful model for diseases involving pericyte deficiencies, such as DR, than those currently being used.

Introduction

Platelet derived growth factors (Pdgfs) are powerful mitogens, and important regulators of embryologic development, cell proliferation, migration, and survival. First described as stimulating cell growth and proliferation, ^1,2 Pdgfs have two chains, A and B. Their receptors, Pdgfrα and Pdgfrβ, are transmembrane tyrosine kinases. ^3–7 Pdgfrα is able to bind PdgfA and B isoforms, whereas Pdgfrβ can bind only PdgfB with high affinity. ^4,5  

Pericytes are perivascular cells located on the abluminal surface of endothelial cells and are embedded within the basement membrane. ⁸ PdgfB is secreted by endothelial tip cells at the front of the extending vessel, which recruits Pdgfrβ-expressing mural cells, such as vascular smooth muscle cells and pericytes, to developing blood vessels during angiogenesis. ^9,10 These mural cells have an important role in the remodeling and stabilization of blood vessels. ^11,12  

The functions of pericytes include roles in vascular development, immune and phagocytic functions, and hemostasis (reviewed previously ^8,13 ). More recently, pericytes have been shown to induce the formation of tight junctions between central nervous system (CNS) endothelial cells to form the blood–brain barrier (BBB) and blood–retinal barrier (BRB). ^14,15  

PdgfB/Pdgfrβ signaling is crucial for pericyte recruitment and survival, ¹⁶ with mutations in these molecules resulting in pericyte recruitment deficiencies. ¹² PdgfRβ null mice usually are lethal due to severe hemorrhaging either in utero or at birth ¹⁷ ; however, mice harboring specific mutations in Pdgfrβ are viable. ^18,19  

A major complication of diabetes mellitus is diabetic retinopathy (DR), which is one of the leading causes of blindness worldwide. DR is characterized by “pericyte drop-out,” which has severe irreversible pathologic consequences for the retina characterized by microvascular abnormalities, such as endothelial dysfunction and hemorrhage, and neuronal abnormalities, including retinal ganglion cell (RGC) death. ^20,21 Prolonged hyperglycemia in the retina activates PKC-δ and SHP-1, inhibiting Pdgfrβ signaling resulting in pericyte apoptosis. ²² Some of the features of DR also have been identified in mice with diminished Pdgfrβ signaling. ^23–25 Indeed, PdgfB has been implicated in DR for many years. ²⁶  

Here, we present a mouse mutant, redeye, with decreased pericyte coverage in the CNS caused by an N-ethyl-N-nitrosourea (ENU)-induced mutation in Pdgfrb. Most other DR mouse models currently in use do not exhibit all of the features of DR, namely pericyte and RGC loss, BRB breakdown, vascular leakage, and acellular capillaries. ^20,27–31 The redeye strain exhibits all the features of DR within weeks of birth, negating the need to age experimental animals and without the multiple organ involvement ^{12,16,18,23,24,26,32–34} seen in other PdgfB/Pdgfrβ mutants. This enables easier isolation of the vascular defects in the redeye mice and makes them a less complicated disease model. Of particular note is that this strain and DR have the same underlying cause for pericyte loss, as hyperglycemia can result in decreased signaling from Pdgfrβ causing subsequent pericyte apoptosis. ²² Furthermore, the defect in BRB development makes them a useful tool for dissecting the role of pericytes in BRB formation and vascular development. 

Materials and Methods

Experimental Animals

The redeye strain, referred to as Pdgfrb^redeye in this report, was maintained on a sighted C3H background. ³⁵ The following primers were used for sequencing for genotyping: 5′-CATTGTGATGGGCAATGATG-3′ and 5′-ATAGGTGCCCGAATCACTCA-3.′ All experiments complied with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research and the relevant local animal welfare conditions. 

Mutation Identification

Initial mapping by SNP array analysis (University of Edinburgh, Wellcome Trust Clinical Research Facility) identified an 8 Mb region of interest. All exons and splice junctions in this region were captured on a custom oligonucleotide-array, amplified, and sequenced, identifying a mutation in the Pdgfrb gene, which was confirmed by Sanger sequencing. To characterize the mutation we extracted RNA from enucleated eyes and performed quantitative real-time PCR using the Mouse Universal ProbeLibrary Set (Roche, West Sussex, UK) with probes for Pdgfrβ and TBP on a Roche Lightcycler LC480 (Roche). The Lightcycler LC480 software (Lightcycler LC480[C] 1.5.0 SP4 [1.5.0.39]; Roche) was used to normalize results to TBP controls and calculate relative expression values. 

Tissue Preparation and Immunofluorescence

Mice were killed at the appropriate age, and whole mount retinae were prepared and stained as described. ³⁶ We stained mutant and control retinae in the same well to control for changes in staining efficiency, and retinae were distinguished by different numbers of radial incisions. 

Embryonic hindbrains were dissected as described previously, ³⁷ and all other organs were dissected from animals of the appropriate age and fixed in 4% paraformaldehyde (PFA) overnight at 4°C. Postnatal day 5 (P5) brains were embedded in 4% agarose in PBS and 200 μm sections were cut on a Vibratome Series 1000 (Technical Products International, Inc., St. Louis, MO). Immunofluorescence for hindbrains and brain sections was performed as for retinae. 

We embedded the kidneys in paraffin for sectioning on a Leica RM 2235 microtome (Leica Microsystems, Inc., Buffalo Grove, IL). Placentae and hearts were cryopreserved in 30% sucrose and embedded in OCT compound (VWR International, Leicestershire, UK) for sectioning on a Leica CM30505 cryostat (Leica Microsystems, Inc.). Citrate buffer antigen retrieval was used for paraffin sections. We blocked all the sections in 10% heat inactivated donkey serum in PBS with Tween-20 (PBST) for an hour, and performed all antibody incubations at room temperature for an hour in block (antibody details are given in Supplementary Table S1). Hematoxylin and eosin staining was performed according to standard procedures. 

Tissue from at least five mice of each genotype from at least three different litters was used for all analysis. All tissues were mounted in Vectashield (Vector Laboratories Ltd., Peterborough, UK), imaged by confocal microscopy (Nikon A1R; Nikon Instruments, Inc., Melville, NY), and maximum intensity projections of z-stacks were created using NisElements AR Version 4.0 software (Nikon UK, Kingston Upon Thames, UK). All images are representative of at least three animals. 

Immunoblot

We extracted protein from all tissues using 1× radio-immunoprecipitation assay (RIPA) buffer (Cell Signaling Technology, Denvers, MA) containing protease and phosphatase inhibitors (Roche), and followed manufacturer's instructions. The NuPage gel electrophoresis and XCell II blotting systems (Invitrogen, Paisley, UK) were used for protein separation and immunoblotting, respectively. Blots were blocked in 5% milk/tris-buffered saline with Tween-20 (TBST) and all antibodies (details given in Supplementary Table S2) were incubated in block for 1 hour at room temperature. Immunolabeling was detected using the Amersham ECL plus Western blotting kit (GE Healthcare Life Sciences, Buckinghamshire, UK). Quantification was performed using ImageQuant TL software (ImageQuant LAS 4000 Version 1.0 Build 1.0.0.52; GE Healthcare Life Sciences). 

Pericyte and Branchpoint Quantification

The MetaMorph Angiogenesis Tube Formation application (Molecular Devices, Berkshire, UK) was used for quantification. Confocal images were used to determine the total area covered by vessels and pericytes to calculate percentage pericyte coverage of vessels and the total number of branchpoints/mm². We imaged three areas of each retina; three different regions of the central retina each encompassing an artery and vein, and two images each of peripheral arteries and peripheral veins; a total of five images/retina. For pericyte quantification of the cerebral cortex capillaries, sections from the frontal, parietal, and occipital regions of the brain were used, and four images were taken from the cerebral cortex of each section. Threshold values were kept the same for analysis of samples of the same stage. 

Avascular Region Measurement

As a measure of vascular branching, we counted the number and area of the avascular regions between capillaries for each image, again for the three regions of each retina. Image analysis was performed using a custom macro written for the freeware image analysis package Fiji, ³⁸ which is based on ImageJ. ³⁹ The macro requires Gabriel Landini's “Morphological Operators for ImageJ,” available in the public domain at http://www.dentistry.bham.ac.uk/landinig/software/software.html. 

Permeability Assay

Fluorescence angiography was performed as described previously ⁴⁰ with modifications. Mice were injected intraperitoneally at P10 with 40 μL of 50 mg/mL Fitc conjugated Dextran 20000S (Sigma-Aldrich Company Ltd., Dorset, UK) wt/vol in PBS, killed an hour later, and the retinae dissected and imaged using a Zeiss Axioplan II fluorescence microscope (Carl Zeiss Ltd., Cambridge, UK) and a Coolsnap HQ CCD camera (Photometrics, Tucson, AZ). IPLab Spectrum software (Scanalytics Corp., Fairfax, VA) was used to analyze retinal hemorrhages. 

Retinal Ganglion Cell and Macrophage Count

Volocity 3D Image analysis software (PerkinElmer, Waltham, MA) was used for RGC and macrophage counts. Cells stained positively for the RGC marker Brn3 were counted in the center of fully mature retinae (P28), which our preliminary data (not shown) suggested was the region most affected in our mutants. 

Statistical Analysis

Statistical tests were performed using the “R” statistics package, an open source software package based on the “S” programming language (available in the public domain at http://www.R-project.org). The appropriate parametric or nonparametric tests used are detailed in Supplementary Table S3. 

Glucose Tolerance Tests

Eight-week-old animals were fasted for five hours before intraperitoneal injection of 2 g glucose (Sigma-Aldrich Company Ltd.) per kilogram of body weight. Blood was drawn from the tail vein at 0, 30, 60, 90, and 120 minutes after injection, and glucose values were monitored using an automatic glucometer (Accuchek; Roche). Numbers of mice tested were females (4 wild type [WT], 6 homozygote animals [HOM]) and males (6 WT, 5 HOM). 

Results

Redeye Is a Hypomorphic Mutation in Pdgfrb

The redeye strain was identified during an ENU-induced mutagenesis screen for recessive eye mutations at MRC Harwell and presented initially with severe hemorrhaging in the eye (not shown). The severity of the hemorrhaging was reduced upon crossing the strain onto a congenic C3H strain lacking the Pde6b^rd1 mutation. ³⁵ Mutants showed normal responses during optokinetic response drum testing (data not shown). No further defects were identified during phenotyping at MRC Harwell and the redeye mice had a normal lifespan. 

We identified a T-to-C mutation in the Pdgfrb gene, encoding Pdgfrβ, affecting the splice donor site at position +2 of intron 6, changing the consensus GT to the less efficiently used GC. ⁴¹ To investigate the impact of this change on splicing, we extracted RNA from P5 eyes and performed RT-PCR. In addition to the PCR product from the normally-spliced transcript, we detected a product of increased size in HOMs, sequencing of which confirmed the retention of intron 6 (Figs. 1A, 1B) in mutant mRNA. This additional sequence introduces 23 novel amino acids followed by premature termination. The normally spliced Pdgfrb transcript was present but quantitative RT-PCR showed that mutant eyes contain only approximately 25% of WT levels of Pdgfrb mRNA (Fig. 1C). This corresponds to a decrease in Pdgfrβ protein seen by immunoblotting and immunofluorescence of HOM retinae (Figs. 1D, 2), confirming Pdgfrb^redeye is a hypomorphic mutant allele of Pdgfrb. 

Figure 1

 

The redeye is a hypomorphic mutant of Pdgfrb. A donor splice site mutation was identified in intron 6 of Pdgfrb, causing reduced splicing of the intron. RT-PCR between exons 6 and 7 results in a larger, unspliced RT-PCR product in HOM animals of 243 base pairs (bp), as well as the normally spliced product of 190 bp (A). Sequencing of these RT-PCR products shows that the mutant transcript contains intron 6, whereas the WT transcript does not (B). Quantitative analysis shows a decrease in Pdgfrb mRNA by quantitative RT-PCR (C) and protein by immunoblot (D). Error bars: standard deviation of the mean.

Figure 2

 

Pdgfrb^{redeye/redeye} mutants have reduced pericyte coverage in the retina. Representative images of P5 retinal vasculature from WT (A–D) and Pdgfrb^{redeye/redeye} mice (E–H). Vessels are stained with IB4, and pericytes with Pdgfrβ (C) and alpha smooth muscle actin (Asma, [D]). Staining is not detected in Pdgfrb^{redeye/redeye} mice (G, H). Another marker of pericytes, NG2, also was used on P5 retinae, which gives more robust staining. WT vessels (I) have more complete coverage by pericytes than mutant vessels (J). IB4- and NG2-stained retinae were used to quantify the percentage of the vessel covered by pericytes. Pericyte coverage in Pdgfrb^{redeye/redeye} (HOM) is significantly lower than WT and Pdgfrb^redeye/+ (Het) at P5 and P10 (SRH test P < 0.001, Mann-Whitney U test *P < 0.05, **P < 0.005; [K, L]). Scale bars: 10 μm. Error bars: standard deviation of the mean.

The Redeye Mutation Causes Decreased Pericyte Coverage Restricted to the CNS

Pdgfrβ is required for pericyte recruitment, so we performed immunofluorescence on perinatal whole mount retinae to determine whether decreased Pdgfrβ caused a corresponding decrease in pericyte coverage. As there is no definitive marker of pericytes, a panel of markers is required to identify these cells correctly. ⁴² This confirmed a decrease in pericytes as shown by a reduction in Pdgfrβ (Fig. 2G), alpha smooth muscle actin (Fig. 2H), proteoglycan NG2 (NG2, Fig. 2J), endosialin, and desmin (Supplementary Fig. S1). We found that NG2 is the most robust of these markers in the mouse retina and, therefore, used it as a specific pericyte marker for our further investigations in the retina. 

To quantify this reduction in pericytes, we stained whole mount retinae for NG2 to mark pericytes and isolectin B4 (IB4) to mark the endothelium. We analyzed retinae at P5, when the primary vascular plexus has grown halfway between the optic nerve head and the periphery of the retina, and at P10 when the primary vascular plexus is fully mature. There was a significant decrease in pericyte coverage in all three retinal regions (central, peripheral vein, and peripheral artery) in Pdgfrb^{redeye/redeye} versus WT animals at P5 and P10 (Figs. 2K, 2L). 

Other Pdgfrb mutants have defects in the recruitment of mesangial cells, ²⁶ the pericyte equivalent in the kidney, with subsequent developmental defects. Therefore, we harvested kidneys at embryonic day 17.5 (E17.5) when mesangial cells begin to invade the kidney glomerulus. Staining for podocytes and mesangial cells showed no apparent defect in mesangial cell recruitment (Figs. 3A–C, 3E–G). We also analyzed adult Pdgfrb^{redeye/redeye} kidneys that showed no histologic defects (Figs. 3D, 3H). Next, we examined E18.5 placentae and hearts for the developmental abnormalities exhibited in other Pdgfrβ signaling mutants at this stage, ²³ and to determine whether pericyte recruitment was affected in these tissues. Our results showed that pericyte recruitment, and histology of the embryonic heart and placenta are normal (Supplementary Fig. S2), suggesting that the defect in pericyte coverage may be restricted to the CNS. 

Figure 3

 

The pericyte recruitment deficiency is restricted to the CNS. The pericyte equivalent in the kidneys are mesangial cells, which are recruited to the developing glomerulus at approximately E17.5. Paraffin sections of E17.5 WT (A–C) and Pdgfrb^{redeye/redeye} kidneys (E–G) stained for podocytes with podocalyxin (B, F), and mesangial cells with NG2 (C, G) show similar numbers of mesangial cells being recruited. Paraffin sections of adult kidneys stained for vessels with collagen IV and podocytes with podocalyxin (Pod) also show normal histology (D, H). E12.5 embryonic hindbrains also were stained with IB4 for vessels (J, M) and NG2 for pericytes (K, N) in the developing CNS. WT hindbrains recruit pericytes normally (I–K), but Pdgfrb^{redeye/redeye} hindbrains show decreased pericyte recruitment (L–N). Scale bars: 100 μm (D, H), 50 μm (all other images).

To test this, we studied another model of angiogenesis, the mouse hindbrain, which is vascularized between E9.5 and E12.5. Flat mounted E12.5 hindbrains from Pdgfrb^{redeye/redeye} mutants showed reduced NG2 staining (Figs. 3L–N). We then examined pericyte coverage of cerebral vessels in P5 animals. As NG2 stains oligodendrocytes as well as pericytes in the brain, we used desmin as a marker for pericytes (Supplementary Figs. S3A–F). Repeated analysis of P5 retinae with desmin staining showed pericyte coverage similar to that obtained with NG2 staining (Supplementary Fig. S3G). We found a statistically significant decrease in pericyte coverage in Pdgfrb^{redeye/redeye} mutants in the cerebral cortex (Supplementary Fig. S3G) confirming that the reduction in pericyte coverage was widespread in, but limited to the CNS. 

The Pdgfrb^{redeye/redeye} Mice Exhibit Basement Membrane Deposition and Vascular Patterning Defects

During this analysis, we noted that there was a defect in vessel patterning (Figs. 4A, 4D), with increased vessel tortuosity in the mutants (Figs. 4B, 4E). Pericytes secrete their own basement membrane, so a reduction in pericytes would likely result in decreased basement membrane coverage. Immunofluorescence on the mutant retinae, indeed, shows less collagen IV coverage of the vessels with additional weaker expression on perivascular cells, which may be pericytes that have failed to adhere due to reduced Pdgfrβ signaling (Figs. 4C, 4F). Furthermore, artery–vein crossover events (Figs. 4G, 4I) and acellular capillaries or “ghost vessels” (Figs. 4H, 4J) are present in mutant retinae. 

Figure 4

 

Loss of pericytes results in vascular patterning defects. Adult whole mount retinae stained with vessel markers IB4 and collagen IV show regular patterning of vessels in a Pdgfrb^redeye/+ retina with straight vessels (A, B). In the Pdgfrb^{redeye/redeye} retina the vessels are irregular and the arteries in particular seem not to extend straight toward the periphery (arrowheads, [D]). At higher magnification these vessels appear to be tortuous (asterisk, [E]). A normal level of collagen IV is expressed by pericytes as well as endothelial cells in Pdgfrb^redeye/+ retinae (C), and is decreased in Pdgfrb^{redeye/redeye} retinae (F) arteries. WT retinae exhibit evenly spaced arteries (a) and veins (v, [G]) but Pdgfrb^{redeye/redeye} retinae show artery–vein crossover events (I) as well as acellular capillaries (arrows, [J]) that are characterized by positive collagen IV staining and IB4 staining, which are seen in WT retinae (H), but with the absence of nuclei, stained with DAPI. Scale bars: 200 μm (A, D), 50 μm (B, C, E–J).

Analysis of Pdgfrb^{redeye/redeye} retinae also revealed a defect in vessel branching (Fig. 5). To quantify this, we used image analysis software to count the number of avascular regions in each image (Figs. 5A–D) that showed a significant decrease in the number of these regions in the Pdgfrb^{redeye/redeye} animals (Fig. 5E). There also is a significant reduction in the number of branchpoints in Pdgfrb^{redeye/redeye} animals (Fig. 5F). 

Figure 5

 

Vascular branching is defective in Pdgfrb^{redeye/redeye} mutants. To characterize vessel patterning, ImageJ software was used to quantify the avascular regions outlined in yellow (B, D) between vessels per mm² of P5 retinae (E). MetaMorph software was used to quantify branchpoint number per mm² of P5 retinae (F). There are a greater number of smaller avascular regions and branchpoints (central and peripheral) in WT and Pdgfrb^redeye/+ (Het) retinae than in Pdgfrb^{redeye/redeye} (HOM) retinae (ANOVA P < 0.001 in both cases, TukeyHSD ***P < 0.001; [E, F]). Scale bars: 100 μm (A–D). Error bars: standard deviation of means.

Vascular Permeability and BRB Development Are Impaired in Pdgfrb^{redeye/redeye} Mice

Initial phenotyping had shown ocular hemorrhage, consequently we investigated the vascular leakage further by using fluorescently labeled 2 kDa dextran, which should not pass through the BRB. We found leakage from Pdgfrb^{redeye/redeye} retinal vessels into the vitreous within an hour of injection (Fig. 6D). No hemorrhage or leakage was observed in control organs, for example, kidney (not shown), or in the retinae of WT mice (Fig. 6A). As pericytes have been linked to BRB induction, ^14,15 we next looked at the expression of BRB markers. The tight junction molecules claudin 5 and zona occludens 1 (ZO1) are expressed in mutant and control retinae (Figs. 6B, 6C, 6E, 6F). BRB dysfunction can cause neurodegeneration; therefore, we counted RGCs in the Pdgfrb^redeye mice (Figs. 6H, 6K). Our results showed that there is a significant decrease in the number of RGCs in the central retina of Pdgfrb^{redeye/redeye} animals, relative to WTs by P28 (Figs. 6H, 6K, 6M). This is due to RGC apoptosis as seen by positive cleaved caspase 3 staining (Figs. 6I, 6L), confirming neurodegeneration in the Pdgfrb^{redeye/redeye} mutants. 

Figure 6

 

Pdgfrb^{redeye/redeye} mutants exhibit BRB dysfunction. Vascular permeability was tested by intraperitoneal injection of 2kDa Fitc-Dextran at P10, which does not leak from WT vessels in the retina due to the BRB (A), but Pdgfrb^{redeye/redeye} retinae exhibit vascular leakage (arrows, [D]). Claudin 5 and ZO1 are expressed in the tight junctions between endothelial cells in P5 WT (B, C) and Pdgfrb^{redeye/redeye} mutants (E, F). To determine if BRB dysfunction causes neurodegeneration, P28 retinae were stained with IB4 (vessels and macrophages), Brn3 (RGCs), and the apoptosis marker Cleaved caspase 3 (Casp3) showing apoptotic RGCs (circles) being engulfed by macrophages in Pdgfrb^{redeye/redeye} (asterisk, [J]), but not WT retinae (G). RGCs stained with Brn3 were counted in the center of the retina in P28 WT (H), Pdgfrb^redeye/+ (Het), and Pdgfrb^{redeye/redeye} (HOM, [K]) animals. Pdgfrb^{redeye/redeye} retinae had fewer ganglion cells than both WT and Pdgfrb^redeye/+ retinae (Kruskal-Wallis test P = 0.002, Mann-Whitney U test *P < 0.05, [M]). Scale bars: 100 μm (A, D, G–L), 10 μm (B, C, E, F). Error bars: standard deviation of the mean.

The Pdgfrb^{redeye/redeye} Mice Have Normal Glucose Tolerance

The features we describe in Pdgfrb^{redeye/redeye} mice are symptomatic of DR, including decreased pericyte coverage, increased vascular permeability, and BRB loss. Therefore, it is important to exclude diabetes as the cause of this retinopathy. We performed glucose tolerance tests, which showed that the Pdgfrb^{redeye/redeye} mice were able to clear injected glucose (Supplementary Fig. S4). The mutants also expressed insulin and glucagon at WT levels (not shown) and, therefore, were not diabetic. 

Discussion

Redeye Mutants Exhibit a CNS-Specific Reduction in Pericytes

Pdgfrb^redeye is a Pdgfrb partial loss of function mutant expressing approximately 58% of the WT levels of Pdgfrβ. Much work has shown previously the requirement of Pdgfrβ signaling for the recruitment of pericytes to the developing vasculature. ^{9,10,17,18,33,37,43–45} We confirm that a reduction in Pdgfrβ results in decreased pericyte coverage in retinae at P5 and P10 (Fig. 2) due to its requirement for the initial expansion and propagation of pericytes. 

Pdgfrb null mice die at/shortly after birth from severe hemorrhaging due to pericyte deficiency.^16–18 However, the Pdgfrb^{redeye/redeye} mice are viable and occur at mendelian ratios (not shown) with only the CNS vasculature exhibiting a reduction in pericytes (Figs. 2, 3, Supplementary Figs. S1–S3). Previous reports describe similar variations in pericyte deficiency between different organs in other Pdgfrb mutants.¹³ This could be due to Pdgfrα compensating for the loss of Pdgfrβ in these tissues.¹⁸ However, our data showed that Pdgfrα is not upregulated in the retina in response to decreased Pdgfrβ expression (Supplementary Fig. S1). It is more likely that the CNS-restricted phenotype is due to the higher pericyte-to-endothelial cell ratio in the CNS, which is as high as 1:3 in the brain, compared to 1:100 in striated muscle, and reportedly is even higher in the retina.^46,47 Our results support previous findings that even within the CNS, pericyte coverage of capillaries can vary.⁴⁸ We found approximately 40% of vessels are covered by pericytes in WT P5 cerebral cortex compared to approximately 80% coverage in the WT retina (Supplementary Fig. S3G). 

This increased pericyte density results in higher expression of Pdgfrβ in WT adult CNS tissues. Pdgfrβ expression in the brain is much higher than in other organs with a dramatic decrease in the expression of Pdgfrβ in Pdgfrb^{redeye/redeye} mutants in the CNS (Supplementary Fig. S5). Interestingly, Pdgfrβ is expressed at the same level in Pdgfrb^{redeye/redeye} mutant and WT kidneys (Supplementary Fig. S5A), which may explain why Pdgfrb^{redeye/redeye} kidneys are unaffected. The reduced expression of Pdgfrβ in the Pdgfrb^{redeye/redeye} hypomorphs may be sufficient to support a subpopulation of pericytes that are less sensitive to decreased Pdgfrβ signaling and enable vascular function outside the CNS. Also, as pericytes have a specific role in BRB formation, ^14,15 the CNS may be more sensitive to changes in Pdgfrβ. 

Pericytes Are Crucial for Vessel Patterning and Endothelial Maturation

Our work confirms that a decrease in pericyte numbers affects vessel patterning. It has been established that during angiogenesis the initial vascular plexus is remodeled to form a fully functional vascular network. ^9,49 Pericytes prevent excessive remodeling by conferring vascular stability. ^11,12,49 Thus, a delay or reduction in pericyte recruitment as seen in the Pdgfrb^{redeye/redeye} mice would enable increased vascular pruning to occur resulting in a sparser vascular network with fewer branchpoints and fewer, but larger avascular regions between the vessels (Fig. 5). This further highlights the importance of pericytes during angiogenesis. 

Redeye Mutants Model Diabetic Retinopathy

Previous findings show that loss of pericytes in the CNS and the resultant vascular permeability is due to an increase in transcytosis. ¹⁴ We can confirm this in the Pdgfrb^{redeye/redeye} mutants in which BRB dysfunction manifests as increased vascular permeability with normal tight junction formation (Figs. 6B, 6C, 6E, 6F). Pdgfrb^{redeye/redeye} mutants also exhibit the classic DR features of acellular capillaries (Fig. 4J) and neurodegeneration in the form of a reduced number of RGCs (Figs. 6G–K), though pericyte ghosts are not seen. Acellular capillaries arise as a consequence of endothelial cell apoptosis due to the loss of pericyte-derived survival signals ⁵⁰ or a loss of the BRB, ⁵¹ leaving an acellular basement membrane (Fig. 4J). As is the case in diabetes, the remaining endothelium is stressed oxidatively and produces less neurotrophic factors, ⁵² resulting in neurodegeneration (Figs. 6G–K). RGCs are particularly sensitive to retinal ischemia and neurotoxicity, ^53,54 and often are damaged in ischemic retinopathies, such as DR. RGC apoptosis is seen much later than the vascular defects and we conclude that vascular dysfunction precedes neurodegeneration in this model. This is a particularly important finding given the controversy regarding whether diabetic mouse models exhibit RGC loss. ^20,27,55  

The early perinatal onset of these vascular changes is particularly useful as most other DR models currently in use do not exhibit vascular pathology until at least 8 weeks of age, ^27–31 and do not show the full range of vascular and neuronal defects exhibited in this model. Though the Pdgfrb^redeye mutants do not exhibit hyperglycemia, they have the same cause for reduced pericytes, that is diminished Pdgfrβ signaling, resulting in the same phenotype. As the Pdgfrb^{redeye/redeye} mice do not exhibit neovascularization we propose that the Pdgfrb^redeye strain is a useful model for nonproliferative DR. The insights gained by studying this mutant not only will be valuable for understanding how pericytes are involved in maintaining a normal retinal vasculature, but also for their role in BBB/BRB formation. 

Supplementary Materials

pdf S1 

pdf S2 

pdf S3 

pdf S4 

pdf S5 

pdf S6 

pdf S7 

pdf S8 

Acknowledgments

The authors thank David Black and animal staff for animal husbandry, Matthew Pearson and Paul Perry for imaging assistance, Allyson Ross for histology assistance, and Alexi Balmuth at the Genepool for exome sequencing. Clare Isacke, BBCRC, London, United Kingdom, provided the kind gift of the anti-Endosialin antibody. 

Supported by Eumodic; the European Mouse Disease Clinic, an EU Integrated Research Programme, and MRC Core Support to the MRC Human Genetics Unit and the MRC Mammalian Genetics Unit. 

Disclosure: S. Jadeja, None; R.L. Mort, None; M. Keighren, None; A.W. Hart, None; R. Joynson, None; S. Wells, None; P.K. Potter, None; I.J. Jackson, None