N-CAM Exhibits a Regulatory Function in Pathological Angiogenesis in Oxygen Induced Retinopathy

Background Diabetic retinopathy and retinopathy of prematurity are diseases caused by pathological angiogenesis in the retina as a consequence of local hypoxia. The underlying mechanism for epiretinal neovascularization (tuft formation), which contributes to blindness, has yet to be identified. Neural cell adhesion molecule (N-CAM) is expressed by Müller cells and astrocytes, which are in close contact with the retinal vasculature, during normal developmental angiogenesis. Methodology/Principal Findings Notably, during oxygen induced retinopathy (OIR) N-CAM accumulated on astrocytes surrounding the epiretinal tufts. Here, we show that N-CAM ablation results in reduced vascular tuft formation due to reduced endothelial cell proliferation despite an elevation in VEGFA mRNA expression, whereas retinal developmental angiogenesis was unaffected. Conclusion/Significance We conclude that N-CAM exhibits a regulatory function in pathological angiogenesis in OIR. This is a novel finding that can be of clinical relevance in diseases associated with proliferative vasculopathy.


Introduction
Proliferative vascular malformations are the cause of sight threatening complications in diseases such as diabetic retinopathy, the dominant cause of blindness in people ,60 in developed countries [1], and retinopathy of prematurity (ROP), the primary cause of blindness in infancy [2]. Common to these complications, initial retinal ischemia is considered to trigger a series of events including epiretinal neovascularization, vitreous hemorrhages and traction retinal detachment, which eventually lead to blindness. The oxygen induced retinopathy (OIR) model mimics blood vessel pathologies and has been used to study proliferative retinopathy in mice [3]. In OIR, mice at postnatal day seven (P7) are exposed to hyperoxia (75% O 2 ) causing the retinal vasculature to regress centrally. When returning the mice to normal oxygen level at P12, the local hypoxia within the capillary free zone induces revascularization and pathological intravitreous neovascularization in the form of epiretinal tuft formation [3,4,5].
Vascular endothelial growth factor A (VEGFA) is a potent mitogen and chemoattractant for endothelial cells [6,7,8]. VEGFA induces retinal blood vessel development both under normal conditions [4] and in pathological proliferative retinopathies [5]. During development of the retinal vasculature, VEGFA is expressed in two layers of the retina in response to local hypoxia [4]. After birth, the superficial blood vessel layer develops peripherally from the optic disc. Astrocytes in the ganglion cell layer express VEGFA to support endothelial tip cell guidance and migration [9]. As the retina thickness increases during the first week, hypoxic conditions stimulate VEGFA expression by cells in the inner nuclear layer (presumably Müller cells) [4], leading to development of the deep blood vessel plexus. Two weeks after birth the retinal vasculature is largely completed. Crucial for vasculature formation is VEGFA binding to the extra cellular matrix (ECM) to build up a gradient around the migrating endothelial tip cells [9]. This gradient guides the tip cells which are probing their way with filopodia extensions. Disruption of this gradient leads to disturbed guidance and defective vascular development [9]. Although both normal and pathological retinal angiogenesis stems from local ischemia, there is an essential difference in the direction of blood vessel growth. In the former, the blood vessels are first restricted to the superficial ganglion cell layer after which they infiltrate the deeper retinal layers. However, during pathological angiogenesis blood vessels penetrate the exceeding inner limiting membrane to form intravitreous epiretinal tufts [3,4,5]. Importantly, the mechanism behind formation of epiretinal tufts remains unknown, whereas revascularization of the avascular area proceeds as in normal blood vessel development.
The neural cell adhesion molecule (N-CAM) is expressed throughout the retina and has pronounced expression in Müller cells and astrocytes [10,11,12]. Müller cells have been reported to be in close contact with the vasculature, producing matrix molecules and having important functions in regulating signal transduction in the retina [11,13,14,15,16]. Astrocytes form the pre-existing cell layer on which the superficial blood vessel plexus develops [17,18]. Recently, we showed that N-CAM regulates pathological angiogenesis during tumor progression [19], suggesting that N-CAM may also be involved in pathologic angiogenesis of proliferative retinopathy. To address whether N-CAM plays such a role, we analyzed the consequence of N-CAM ablation on OIR. Here, we show that during OIR, N-CAM accumulates in astrocytes closely associated with the tufts. Furthermore, N-CAMdeficiency resulted in reduced vascular tuft formation and endothelial cell proliferation, despite an elevation in VEGFA mRNA expression.

N-CAM ablation does not perturb normal development of the retinal vasculature
To examine whether N-CAM is required for developmental angiogenesis in the retina, we studied normal development of the retinal vasculature in N-CAM-deficient mice. At P10, when the superficial blood vessel plexus is almost fully developed, N-CAM expression was localized in the vascular region of the ganglion cell layer in wild type (wt) mice ( Figure 1A). At P7, N-CAM expression was localized around the GFAP + astrocytes in close contact with the vasculature, but even more pronounced in the deeper layers of the retina ( Figure S1A). Ablation of N-CAM resulted in no effect on radial development or tip cell sprouting, ( Figure 1B-D), and there was no significant difference between wt and N-CAM 2/2 when comparing the vessel diameter in the capillary network in P10 retinas (wt: 7.560.3, N-CAM 2/2 : 8.060.2, n both groups = 3.

Epiretinal tufts are surrounded by N-CAM expressing astrocytes
Notably, N-CAM, which normally is distributed between blood vessels on astrocytes (S1A), was accumulated around the retinal tufts during OIR (Figure 2A-D). In OIR, the tufts are surrounded by astrocytes and can be visualized as GFAP 2 spots in areas of tuft formation ( Figure 2E-H). During blood vessel development in the retina, the blood vessels send out thin filopodia along the preexisting astrocyte network. This process is not disturbed in the N-CAM deficient mice ( Figure 2I-L). In OIR the other cell type in contact with the tufts are pericytes, but since they cover the entire surface of the tuft ( Figure 2M, N) and N-CAM distribution was limited to the edges, matching the astrocyte location, the probable source for N-CAM is astrocytes.

N-CAM ablation decreased pathological angiogenesis without affecting the avascular area in OIR
To examine whether N-CAM affected pathological angiogenesis, including retinal revascularization and tuft formation, during OIR, we analyzed N-CAM-deficient mice after five days in normoxia (P17). Heterozygous and homozygous N-CAM mutant animals exhibited a gene-dosage dependent reduction of tuft formation ( Figure 3A-D). The number of tufts was markedly reduced in N-CAM knock outs compared to wt ( Figure S2A). Also by plotting the distribution of tufts in a size dependent manner, it can be seen that there were more tufts of all sizes in wt compared to N-CAM homozygous knock outs ( Figure S2B). However, there was no difference in the size of the avascular area compared to wt littermates ( Figure 3D).

N-CAM ablation does not affect blood vessel leakage
We have earlier reported that N-CAM ablation leads to increased leakage on tumor blood vessel [19]. To analyze whether N-CAM has the same effect on the vascular integrity in pathological angiogenesis during OIR the vasculature of mice after OIR was perfused with FITC-labeled dextran. The perfusion revealed no difference in blood vessel leakage in the N-CAM mutants compared to wt (Figure S1B).
Despite increased VEGFA level in N-CAM deficient retinas, the proliferation rate in tuft endothelium was decreased To analyze whether the reduced size and number of tufts could be explained by diminished endothelial cell proliferation, we performed isolectin and BrdU-labeling. Indeed, N-CAM deficient retinas exhibited a dramatic reduction in endothelial cell proliferation within tufts (wt = 3.860.3 and N-CAM 2/2 = 1.860.3, p = 0.005), whereas the remaining vasculature was unaffected ( Figure 4A-D). The proliferation rate was normal in N-CAM deficient retinas both in the larger vessels (wt = 2.360.6 and N-CAM 2/2 = 3.260.8, p = 0.39) and in the capillary bed (wt = 12.561.6 and N-CAM 2/2 = 16.762.1, p = 0.19). It can, however, not be ruled out that the proliferation of non-endothelial cells within the tufts and the retinal vasculature also could be affected by N-CAM deletion.
The blood vessel structure and sprouting of the revascularization front appeared similar between N-CAM deficient and control retinas ( Figures 4E-F). The tufts formed by pathologic intravitreous neovascularization also display filopodia ( Figure 4G), even though they appear shorter and point in all directions. No change in either kind of sprouting was observed in N-CAM deficient mice during OIR ( Figure 4F, H).
To investigate whether the decrease in pathological neovascularization upon N-CAM deficiency was linked to VEGFA expression changes, VEGFA mRNA levels were quantified at P17 when VEGFA levels are the highest [20]. In contrast to what was expected, there was an increase in VEGFA mRNA expression, despite reduced endothelial cell proliferation within the retinal tufts ( Figure 4A-D, I).

Normal perivascular ECM distribution both in the developing retina and OIR tufts despite N-CAM ablation
To investigate whether N-CAM exhibited similar effects on the perivascular ECM deposition during pathological angiogenesis in the retina as during tumor angiogenesis, expression of the basement membrane proteins collagen IV, fibronectin and laminin a1/c1 was studied. However, N-CAM ablation resulted in no change in the expression of these ECM molecules during normal blood vessel development ( Figure

N-CAM ablation affects TGF-b and FGFR mRNA expression in the retina
In an attempt to elucidate the mechanism behind the tuft specific decrease in proliferation rate, genes known to control endothelial cell proliferation, including FGF1, FGF2, TGF-b and IGF-1 together with genes known to be involved in signaling together with N-CAM, including FGFR and EGFR, were analyzed using quantitative PCR (QPCR) on OIR retinas.
Both FGF1 (aFGF) and FGF2 (bFGF) are known to stimulate endothelial cell proliferation [21,22] but we did not detect any statistical significant altered mRNA levels in the N-CAM mutants ( Figure 6). TGF-b is known to affect endothelial cells in various ways and have been implicated both to stimulate and to inhibit endothelial cell proliferation [23]. For all three isoforms we detected a statistical significant up regulation in N-CAM mutant retinas compared to the heterozygote's, whereas compared to the wt the up regulation did not reach significance ( Figure 6).
IGF-1 signaling has previously been shown to stimulate endothelial cell proliferation in OIR [24,25] but we could not observe any statistically significant mRNA expression changes in the N-CAM mutants ( Figure 6). Both the EGFR and the FGFRs have been shown to functionally interact with N-CAM [26,27]. FGFR4 was statistically significant upregulated in the N-CAM 2/2 mutants compared to the N-CAM +/2 and wt. The expression level of EGFR was not altered by N-CAM ( Figure 6).

Discussion
In OIR, hyperoxia primarily results in blood vessel regression. Returning to normoxia creates a hypoxic milieu in the avascular area, which stimulates revascularization and pathologic neovascularization. The normal direction of retinal blood vessel infiltration due to hypoxia is from the superficial layer down through the deeper layer of the retina. However, it is not known why the vessels penetrate the exceeding inner limiting membrane to form epiretinal tufts under pathological conditions. Although the tufts appear malfunctional, they provide the underlying retina with physiological oxygen levels, which presumably lower VEGFA expression by the astrocytes [28].
In the present study we show that N-CAM ablation does not affect the normal developmental retinal angiogenesis or the revascularization of the retina after OIR but decreases, in a gene-dosage dependent way, the epiretinal tuft formation.
During normal retinal development N-CAM is expressed in a pattern suggesting that astrocytes are the cell of origin close to the vasculature. After OIR N-CAM is accumulated around the epiretinal tufts, also indicating astrocytic origin. Even though astrocytes seem to be the probable source of N-CAM we cannot rule out that other cell types, as Müller cell and microglia, also express N-CAM.
In agreement with our observations that the revascularization was unaffected whereas the tuft formation was decreased, the reduced endothelial proliferation rate was restricted to the tufts.
The migrating endothelial front uses tip cell filopodia extensions to guide their movement by following the pre-existing VEGFA expressing astrocyte network [9]. Deficient sprout establishment could potentially lead to reduced endothelial cell migration and proliferation [9]. In the N-CAM deficient retinas, tip cell sprouting was unaffected both during normal development and OIR. In OIR the epiretinal tufts also produce short and randomly directed sprouts of which neither was affected in the N-CAM deficient retinas.
A VEGFA gradient has been shown to be crucial for endothelial migration and proliferation in the retina [9,29]. Whereas VEGFA is essential for developmental angiogenesis within the retina [4], it has also been associated with pathologic angiogenesis in ocular diseases [30,31,32,33]. VEGFA is encoded by one gene, which after alternative splicing generates at least three major isoforms; VEGFA 188 , VEGFA 164 and VEGFA 120 [34]. Guidance of tip cell filopodia is dependent on the correct relationship between the expression levels of all three isoforms. Therefore, disturbed expression of any of the isoforms may cause defective filopodia guidance, endothelial migration and proliferation [9]. It was therefore of interest to analyze the effect of N-CAM ablation on the expression level of each isoform. In agreement with previous studies, the 164 isoform was the most abundant, followed by 120 and 188 (data not shown) [35]. Statistical analysis failed to reveal significant differences in VEGFA isoform expression between N-CAM mutant and wt retinas in OIR, suggesting that the reduced EC proliferation is not due to altered VEGFA levels. N-CAM may thus regulate endothelial cell proliferation by VEGFA-independent pathways. One such pathway is endothelial nitric oxide synthase (eNOS)-mediated signaling [36]. Nitric oxide (NO) is a free radical produced by NOS of which there are three variants. eNOS, predominantly expressed in the plasma membrane of vascular endothelial cells [37,38], has been implicated to play a proangiogenic role [39,40]. Mice deficient in eNOS express an OIR phenotype, which is very similar to the one presented in this work, with less pathologic angiogenesis and elevated VEGFA levels [36]. To investigate whether eNOS was transcriptionally regulated in N-CAM deficient OIR retinas we measured eNOS mRNA expression levels. No difference in eNOS mRNA expression was detected between N-CAM mutant and wt retinas. To further elucidate the mechanism behind the reduced proliferation in the vascular tufts, the expression of genes known to play a role in endothelial cell proliferation or N-CAM signaling were analyzed by QPCR on OIR retinas. Among the genes analyzed, all three isoforms of TGF-b and FGFR4 were up regulated in N-CAM 2/2 . However, since N-CAM only affects pathological angiogenesis, RNA extraction from whole retina might not be sensitive enough to pick up changes in mRNA expression only from the pathological vasculature. Both TGF-b and FGFR4 could play a role in the mechanism by which N-CAM regulates tuft formation during OIR.
N-CAM deficiency may also directly or indirectly affect the distribution of growth factors bound to the ECM. It has previously been shown that if the basement membrane of blood vessel is injured or degraded, e.g. by defective ECM deposition, the underlying endothelial and mural cells get intimately exposed to surrounding growth factors and respond by starting to proliferate [41,42]. Consistently, ECM molecules have been implicated in endothelial cell proliferation during angiogenesis [43]. We previously showed that N-CAM-deficiency during tumor progression resulted in diminished expression of ECM components and defective perivascular ECM deposition [19]. Further, N-CAM has been shown to inhibit expression of matrix degrading enzymes (matrix metalloproteinases, MMPs) [44] which could indicate a proliferation promoting property. In this study, the proliferation in the pathological retinal tufts was down regulated in N-CAMdeficient mice which rather would expect ECM-inhibiting or MMP-promoting properties of N-CAM. We found no evidence that N-CAM deficiency has any effect on ECM components in the retina during normal development or OIR. It can, however, not be ruled out that other ECM components than the ones examined in this study might be affected in the N-CAM mutants. Further studies are needed to reveal whether there are mechanistically differences in N-CAM's role during OIR and tumor angiogenesis.
In conclusion, we show that N-CAM is involved in hypoxia regulated endothelial cell proliferation in vivo. Hypoxia-exposure in the OIR model resulted in a gene-dosage dependent decrease in pathological angiogenesis, as revealed by reduced formation of epiretinal tufts, in N-CAM heterozygous and homozygous mice, whereas normal retinal vasculature development appeared unaffected. These observations are consistent with our recent finding that N-CAM deletion limits pathological angiogenesis during tumor progression [19]. Altogether, these findings emphasize a novel role of N-CAM in pathological angiogenesis.
However, the underlying mechanism for N-CAM's role in pathological angiogenesis during tumorigenesis and OIR appears to differ. Whereas N-CAM maintains vessel integrity by promoting pericyte-endothelial cell-cell interactions during tumor angiogenesis, N-CAM is required for endothelial cell proliferation during formation of the epiretinal tufts. The distinct accumulation of N-CAM around the epiretinal tufts suggests a potential role in their formation during OIR. Indeed, N-CAM ablation resulted in a decrease in endothelial cell proliferation specifically within tufts, which presumably explains the reduction in tuft formation. This finding would be consistent with reduced expression of the major proliferation growth factor VEGFA. However, instead, we found that VEGFA mRNA expression was increased in N-CAM 2/2 mice relative to the controls. The increase in VEGFA levels could either be a compensatory effect of the defective proliferative response, or a consequence of reduced formation of epiretinal tufts, which normally lower VEGFA expression in astrocytes after providing the underlying retina with higher oxygen levels [28].
By screening several genes involved in endothelial cell proliferation or N-CAM signaling we identified TGF-b and FGFR4 to be potential mechanistically targets of N-CAM in vascular tuft formation.
Based on our findings, we propose that N-CAM may be a potential target of clinical relevance for diseases caused by proliferative retinopathy in humans.

Ethical statement/Mice
N-CAM +/2 (with C57/BL6 background) [45] mice were bred to generate wt, N-CAM +/2 and N-CAM 2/2 littermates. They were housed and bred in accordance with regulations for the protection of laboratory animals, after approval from the local ethical committee. The animals were housed on a 12:12-hour light-dark cycle and food and water were available ad libitum.

Tissues
Oxygen induced retinopathy (OIR); Mice at postnatal day 7 (P7) were exposed to 75% oxygen for 5 days together with their mothers in an oxygen chamber [3], the oxygen concentration was checked daily. The mice were then transferred back to normal air (21% oxygen) for 5 days after which the mice were sacrificed and the eyes were dissected for whole mount immunostainings.
Others; Mice at P5, P7 and P10, respectively, were sacrificed and eyes were dissected for immunostainings.
The FITC dextran perfusions were carried out as previously reported [46].

Immunohistochemistry
Retinas were prepared and stained according to earlier published protocol [9]. Shortly, the mice were sacrificed and the eyes collected and placed in PBS on ice. After a short (2 minutes) fixation in 4% PFA the retina was dissected. To avoid damage to the epiretinal tufts during this procedure the hyaloid vessels were gently removed. The retina was then fixed for an additional 2 hours. The blood vessels were visualized by staining with ImmunoResearch Laboratories were used. Flat mounted retinas or sections were analyzed by fluorescence microscopy using a Nikon E1000 or a Zeiss Axioplan 2 microscope and by confocal laser scanning microscopy using a Leica LCS NT or Zeiss LSM 780. Images were processed using Adobe PhotoshopH and or Imaris.

QPCR
Retinas from mice exposed to the OIR treatment were dissected in ice cold PBS and total RNA was prepared using the RNeasy mini kit (Qiagen) with the RNase free DNase (Qiagen) treatment according to the manufacturer's instructions. Reverse transcription (iScript, BioRad) and QPCR assays using SYBR Green I as detection chemistry was performed as described [47]. Primer sequences are shown in Table S1. PCR products were checked by agarose gel electrophoresis and melting curve analysis. Expression data were normalized against Tubb5. The expression level of VEGFA 188 was found to be significantly lower (,10 fold) than the expression of VEGFA 188+164 , consequently VEGFA 188+164 ,VEGFA 164 . The ratio between VEGFA 188+164 and VEGFA 188+164+120 (,VEGFA 164 : VEGFA 164+120 ) gives the relative difference in expression level between the VEGFA 164 and VEGFA 120 isoforms.

Blood vessel area quantification
Pictures of whole mount retinas after OIR were taken at 2006. For measurement of tuft-, avascular-, and total retina-area the Velocity software was used, and by modulating intensity settings the different areas could be quantified. By letting the software identify only the parts of the vasculature with highest intensity of the isolectin staining, the tufts were marked and the area for each tuft was quantified. By letting the software identify the area with any staining at all, including the very low background, the total Figure 6. N-CAM affects TGF-b and FGFR4 expression levels in retinas during OIR. QPCR was used to quantify the expression of vascular related genes in retinas after OIR. All isoforms of TGF-b (n wt = 6, n N-CAM+/2 = 8, n N-CAM2/2 = 7), were upregulated in N-CAM 2/2 , FGFR4 (n wt = 6, n N-CAM+/2 = 8, n N-CAM2/2 = 7) was significant upregulated in the N-CAM 2/2 mutants compared to the N-CAM +/2 and wt whereas the other genes were not statistically significantly altered, eNOS ( n wt = 5, n N-CAM+/2 = 3, n N-CAM2/2 = 4), IGF-1, FGF1, FGF2, FGFR2, FGFR3and EGFR (n wt = 6, n N-CAM+/2 = 8, n N-CAM2/2 = 7). Statistical method used was Student t-test. * = p,0.05. doi:10.1371/journal.pone.0026026.g006 retina was marked and the area was quantified. By setting the intensity level very low, the avascular part in center of the retina was marked, and the area was quantified.

Blood vessel diameter measurement
Pictures of P10 whole mount isolectin stained retinas were taken att 2006. The vessel diameter (measured in the computer software Photoshop CS) was measured on 30 blood vessels of the capillary network in 3 wt and N-CAM 2/2 mice. The data is presented in length units as average 6SEM.

ECM quantification
P5 retina samples were stained for PECAM and fibronectin or collagen IV or laminin c1. One field/retina (206 objective) in the capillary bed was scanned using Zeiss LSM 780 confocal.
Using Imaris the blood vessel volumes (PECAM) and the ECM volumes were quantified. The ECM volume/blood vessel volume were compared between wt and N-CAM 2/2 . The data was normalized to wt and presented as average 6SEM. n all groups = 5.

Quantification of endothelial cell proliferation retinas after OIR
Pictures of whole mount retinas after OIR were taken at 2006. Tufts. BrdU labeled cells were counted in 338 tufts from 3 wild type retinas and 123 tufts from N-CAM 2/2 retinas. The area of each tuft was measured with the same technique as mention above in the section ''blood vessel quantification''. The number of BrdU labeled cells per tuft area unit was calculated and the average was calculated for each retina. Student t-test was performed on n = 3. The data is presented as average 6SEM BrdU-labeled cells/area unit.
Capillaries. BrdU labeled cells were counted in 34 fields of the capillary network of 3 wt retinas and 29 fields of the capillary network of N-CAM 2/2 retinas. The area was measured in Photoshop and the number of BrdU labeled cells per area unit was calculated. The average was calculated for each retina and Student t-test was performed on n = 3. The data is presented as average 6SEM BrdU-labeled cells/area unit.
Large vessels. BrdU labeled cells were counted in 24 large vessels of 3 wt retinas and 28 large vessels of N-CAM 2/2 retinas. The length of each vessel was measured in Photoshop and the number of BrdU labeled cells per length unit was calculated. The average was calculated for each retina and Student t-test was performed on n = 3. The data is presented as average 6SEM BrdU-labeled cells/length unit. Figure S1 N-CAM is expressed in the mouse retinal blood vessel layers and co-localize with astrocytes and N-CAM ablation does not affect blood vessel leakage in retinas after OIR. (A) Retinal sections from wt P7 were stained for N-CAM (green), PECAM (blue) and GFAP (red). N-CAM was expressed in the blood vessel layer and co-localize with GFAP expressing astrocytes but is also expressed in deeper retinal layers. Astrocytes enter the retina from the optic nerve (on) and first form a superficial plexus which is seen as the GFAP + rim close to the vitreous body. As insets high magnification optical sections are shown. Scale bar = 100 mm. (B) FITC-dextran (green) perfusion of retinas after OIR revealed almost no leakage of the retinal vasculature and there was no difference between N-CAM 2/2 and wt. Isolectin staining (red), FITC-labeled dextran (green). (TIF) Figure S2 N-CAM ablation decreases the number of tufts. (A) The total number of tufts per retina was 2.8 times higher in WT compared to NCAM 2/2 (Student t-test, p,0.05, n = 3 for WT and NCAM 2/2 ). (B) The tuft areas were lognormally distributed in both WT and NCAM2/2 and the number of tufts was lower for NCAM 2/2 than WT for all sizes of area. (Tuft areas were pooled for 3 WT and 3 NCAM 2/2 mice, respectively). (TIF) Figure S3 ECM volume quantification. The blood vessel volumes (PECAM) and ECM (fibronectin, collagen IV and laminin c1 respectively) volumes were quantified in p5 retina samples using confocal microscopy and Imaris software. The ECM volume/blood vessel volume were compared between wt and N-CAM deficient retinas, no significant difference could be detected, collagen IV p = 0,8026, fibronectin p = 0,2705 and laminin c1 p = 0,8002. (TIF)