Deletion of Placental Growth Factor Prevents Diabetic Retinopathy and Is Associated With Akt Activation and HIF1α-VEGF Pathway Inhibition

Diabetic retinopathy (DR) is a leading cause of visual impairment and blindness. DR is characterized by increased vascular permeability; breakdown of the blood-retinal barrier (BRB); apoptotic cell death of retinal neurons, endothelial cells (ECs), and pericytes; and the appearance of microaneurysms and acellular capillaries. Several biological pathways have been implicated in the pathogenesis of DR, including vascular endothelial growth factor (VEGF) (1), tumor necrosis factor-α (2), the kallikrein-kinin system (3), and altered O-GlcNAc signaling (4). The exact mechanisms and the relationship of pathogenic changes to biochemical pathways in DR need further elucidation.

Placental growth factor (PlGF), a member of the VEGF family, was first described from a human placental cDNA library in 1991 (5). PlGF is expressed by a wide variety of cell types, including ECs and retinal pigment epithelial cells (RPEs) in response to hypoxia (6,7). PlGF is a homolog of VEGF and binds to fms-like tyrosine kinase-1 (FLT1; also known as VEGFR-1) and soluble FLT1, a circulating form of FLT1 that lacks the transmembrane and intracellular domains. Activation of FLT1 by PlGF augments the effects of VEGF signaling with VEGFR-2, suggesting synergistic effects of PlGF and VEGF (8,9). PlGF can also form heterodimers with VEGF (8). PlGF exerts a strong effect upon blood vessel growth and maturation and has direct proangiogenic effects on ECs (6).

Recent studies suggest that PlGF plays a role in the pathogenesis of DR (10). Diabetic human retinas have higher expression of PlGF mRNA compared with nondiabetic retinas (11). In proliferative DR, immunoreactivity for PlGF was localized to endothelial and perivascular regions of neovascular membranes (12). PlGF was produced in a cultured human RPE line in response to hypoxia (13). Other studies have shown higher concentrations of PlGF in aqueous and vitreous humor of eyes with DR compared with controls (14–16).

Mice with inactivation of the gene expressing PlGF (plgf^−/− or pgf^−/−) show little effect of PlGF absence on vascular development, but in response to hypoxia, PlGF^−/− mice have reduced angiogenesis in the retina from oxygen-induced retinopathy (6). The findings from clinical studies, in vitro studies, and PlGF^−/− mice demonstrate the PlGF may have a role in the pathogenesis of DR. To gain further insights, we compared molecular and histological changes in wild-type (WT) mice, Akita diabetic mice, and PlGF^−/− mice and generated a novel Akita.PlGF^−/− mouse strain. We hypothesized that DR would be less severe in Akita diabetic mice that were deficient in PlGF.

Animal use was in accordance with the approved protocols by the Institutional Animal Care and Use Committee of The Johns Hopkins University. Akita mice were crossed with PlGF^−/− mice (17) in a C57BL/6J background for two generations to give birth to the progeny with the genotype of Akita.PlGF^−/−. Additional mating of Akita.PlGF^−/− and PlGF^−/− was performed to fix the PlGF mutation to homozygosity while maintaining the Akita mutation in the heterozygous state. The genotyping of PlGF gene was performed as described in the literature (6). The genotyping for Pde6b^rd1 and Crb1^rd8 genes was performed as described in the literature (18,19).

Western blot (WB) was performed as described previously (20). The anti-PlGF, anti–hypoxia-inducible factor (HIF)-1α, anti-VEGFR1, anti-VEGFR2, anti–intercellular adhesion molecule (ICAM)-1, anti–vascular cell adhesion molecule (VCAM)-1, anti-ZO-1, and anti-VE-cadherin antibodies are described in more detail elsewhere (21,22). Dilutions of 1:200 and 1:1,000 were used for anti-VEGF antibody (Santa Cruz, Dallas, TX) and anti-VEGFR3 antibody (eBioscience, San Diego, CA), respectively. The anti-Akt and anti-phospho (p)-Akt antibodies were bought from Cell Signaling (Indiana, IN) and used as 1:1,000 dilutions.

The optical density (OD) of protein bands in WB images was determined by ImageJ (National Institutes of Health). After normalization by dividing the OD of protein of interest with the housekeeping gene β-actin of the same sample, the percentage of the normalized OD was calculated for each protein of interest based on this equation: percentage of OD (4) = 100*OD (4) / [OD (1) + OD (2) + OD (3) + OD (4)] (each represents one retinal protein sample in the same WB membrane). The average percentage of normalized OD was designated as the relative protein expression (n = 6).

The methods of TUNEL assay were described previously (2). Apoptotic DNA cleavage was analyzed with the cell death ELISA kit (Roche, Indianapolis, IN), which was performed according to the manufacturer’s instructions with some modifications. In brief, retina was incubated with lysis buffer (0.1 mol/L citric acid buffer containing 0.5% Tween 20) for 30 min. Retinal lysates were centrifuged to sediment the cell debris and nuclei (>1,500 rpm, for 10 min). The 20 μL supernatants were subjected to apoptotic DNA cleavage ELISA assay. The OD (absorbance at 405 nm with a reference at 490 nm) was normalized with retinal mass. The average OD/retinal mass indicated the level of cytoplasmic nucleosome containing fragmented DNA (n = 8).

The transcardial perfusion of fluorescein isothiocyanate (FITC)-dextran and confocal microscopy were performed as described previously (21). The z-series of optical sections from confocal images were made of the superficial, intermediate, and deep retinal vessel bed to examine the localization of FITC-dextran (70 kDa, TdB Consultancy, Sweden). Retinal leukostasis was performed as previously described (2).

The quantitative BRB assay was performed as previously described (2,21). The count per minute (CPM) was normalized with tissue weight (CPM/mg retina, lung, or kidney) or with serum volume (CPM/μL/h). The ratio of retina to lung, kidney, or serum was designated as retinal vascular leakage ratio.

Immunofluorescence (IF) staining was performed as described previously (22). The primary antibodies used for this study included anti-CD31/PECAM-1 (BD Pharmingen, San Jose, CA); anti-VEGFR1 (ImClone, Somerville, NJ); and anti-p-VEGFR1 (Y1213; R&D Systems), anti-NG2, anti-GFAP, anti-p-Akt (Ser473), anti-p-VEGFR2 (Tyr1175), and anti-p-endothelial nitric oxide synthase (eNOS; Ser1177; Cell Signaling). For double labeling, two primary antibodies from the distinct host species were mixed for the same tissue section and two appropriate secondary antibodies conjugated with different chromophores (e.g., rhodamine or fluorescein) were used for detection and visualization. For the quantitative comparison of IF images, the specimens and images were prepared as described previously (23). The immunoreactive intensity and area were quantified by ImageJ and Image Pro, respectively.

The procedures of real-time PCR were described previously (20,24). For quantifying the relative expression of genes of interest (GOI), the threshold cycles (Ct) were normalized by subtracting the Ct of housekeeping gene actin of the same sample (ΔCt = GOI Ct-actin Ct). The value of 2^−ΔCt was calculated for each sample and average for each mouse strain. The percentage of the average 2^−ΔCt from each mouse strain to the sum 2^−ΔCt from the four mouse strains was designated as the relative expression of GOI (n = 6).

Isolation of retinal vasculature and quantification of acellular capillary were performed as previously described (25,26). The enucleated eyes were fixed with 2% paraformaldehyde for 24 h. After removing the cornea, the eye cups were fixed with 2% paraformaldehyde for 24 h and dissected to isolate the retina. The retinas were rinsed in tap water overnight and then were incubated with 40 unit/mL elastase (Calbiochem, MA) in 100 mmol/L sodium phosphate buffer with 150 mmol/L sodium chloride and 5.0 mmol/L EDTA pH 6.5 at 37°C for 2 h. Retinas were transferred to a 100 mmol/L Tris-HCl (8.5) solution at room temperature for 3 h. Nonvascular tissues were removed by gentle brushing, and the isolated vasculature was mounted on the slide. After drying overnight, the retinal blood vessels were subjected to the periodic acid–Schiff and hematoxylin stain. The retinal blood vessels of 20–25 random fields (200× magnification) were analyzed in a masked fashion for the quantification of acellular capillaries, which were identified as capillary-sized vessel tubes without nuclei along their length. The numbers were normalized by the counting area (degenerated vessels/mm²). The number of ECs and pericytes were determined by counting their respective nuclei, the differences of which were distinguished by their morphology and position in the blood vessels, as described previously (27). The six midzone fields of retinal digest (400× magnification) were analyzed for each retina in a masked fashion. The average number of ECs per mm² vascular bed was calculated and expressed as the density of EC. The average ratio of EC to pericyte was used to demonstrate the pericyte coverage or death.

The nonparametric Mann-Whitney U test was performed to determine the significance level between two samples: diabetic retina versus nondiabetic retina and Akita versus Akita.PlGF^−/−. P < 0.05 was designated as being statistically significant.

We generated a genetic model of diabetes by crossing the PlGF^−/− mouse and the Akita diabetic mouse. PCR genotyping analysis confirmed the deletion of the PlGF gene in Akita.PlGF^−/− mice (Fig. 1A). We further performed genotyping for Pde6b^rd1 and Crb1^rd8 genes, since the two are common retinal degeneration gene mutations that are found in the C57BL/6 mouse strains (18). The results showed that neither was present in the four mouse strains used in this study. Further histopathological analysis indicated there was no obvious retinal degeneration, as indicated by the 10–12 nuclear layers in photoreceptors, and other retinal abnormalities (Supplementary Fig. 1). The level of blood glucose was equal between the offspring of WT/Akita (PlGF^+/+/Akita.PlGF^−/−) and PlGF^−/−/Akita.PlGF^−/− breeding mice (Fig. 1B and C). The Akita.PlGF^−/− mice were significantly lighter in body weight than the same-aged Akita mice (Fig. 1D and Supplementary Table 1). Despite the loss of body weight, the double mutant Akita.PlGF^−/− mice were viable, fertile, and healthy. The blood assays and histopathological assessments also suggested less severe pathology in the liver and kidney of Akita.PlGF^−/− mice than Akita mice (data not shown).

By comparing the two diabetic mouse strains (Akita and Akita.PlGF^−/−) with their respective nondiabetic mouse strains of the same genetic background (C57BL6 [WT] and PlGF^−/− mice), we found that PlGF deficiency prevented diabetes-caused retinal cell death in the Akita.PlGF^−/− mouse retina. The number of diabetes-induced TUNEL(+) cells was significantly reduced in Akita.PlGF^−/− mice compared with Akita mice of the same age (Fig. 2A–E). The number of activated caspase-3(+) cells was also significantly reduced due to PlGF deficiency in the diabetic retina (data not shown). Furthermore, apoptotic cell death was examined by cell death ELISA analysis, in which cytoplasmic nucleosomes were measured. The results showed that DNA fragmented nucleosomes were increased in Akita diabetic mice compared with nondiabetic mice, but not in Akita.PlGF^−/− mice (Fig. 2F).

Retinal digests were performed as described in 2Research Design and Methods. Intact retinal vasculatures were used for quantitative analyses of acellular capillary, EC density, and pericyte death. Firstly, acellular capillaries or capillary degeneration were significantly increased in Akita mice compared with WT controls. Genetic ablation of the PlGF gene abrogated the diabetes-induced capillary degeneration: the mean number of acellular capillary/mm² retinal vasculature was significantly increased in Akita mice compared with WT and Akita.PlGF^−/− mice. Secondly, there was not a significant change in the density of vascular ECs between the mouse strains (Supplementary Fig. 2 and Supplementary Table 2). Lastly, pericyte death was significantly increased in Akita mice compared with WT mice, but PlGF deletion abrogated diabetes-induced pericyte loss (Fig. 3). Consistently, double immunostaining of NG2 and CD31 showed the pericyte coverage was decreased by diabetes (5-month diabetes duration), but PlGF deletion abrogated the pericyte loss: the average ratio of NG2(+) to CD31(+) vessel areas was significantly reduced in Akita diabetic retina compared with the nondiabetic control retina, but that of Akita.PlGF^−/− retina was not (Supplementary Fig. 3).

Although it is possible that PlGF, like its homolog VEGF, is produced by a variety of retinal cell types, its cellular targets appear to be vascular ECs and pericytes in diabetic retina. We investigated this question with the vascular EC markers CD31 and the pericyte marker NG2. IF staining also showed an increased expression of PlGF protein and its receptor VEGFR1 by diabetes: PlGF or VEGFR1 immunoreactivity was not present or, if present, was barely detectable in nondiabetic mouse retina but was dramatically increased in the retinas of Akita diabetic mice (3 months old with an ∼2-month diabetes duration). Double labeling of IF staining showed that the diabetes-induced increased PlGF protein colocalizes well with its receptor VEGFR1, the EC marker CD31, and the pericyte marker NG2 but does not colocalize with the astrocyte/glial marker GFAP (Supplementary Fig. 4). These results suggest vascular EC and pericytes are likely the cellular targets of PlGF, where it mediates DR by transmitting signaling via VEGFR1. We recently observed that diabetes-induced inflammation and vascular leakage were significantly inhibited due to VEGFR1 blockade (J.H., D. Kim, and H.H., unpublished data).

Vascular leakage was determined by a quantitative BRB method, in which ³H-mannitol (molecular weight 182 Da) acted as a tracer. The results showed diabetes caused a significant vascular leakage in Akita mice compared with the nondiabetic control mice, which is consistent with our previous results (2). However, deletion of PlGF in Akita mice prevented diabetes-related vascular leakage (Fig. 4). Vascular leakage was further examined by the extravascularization of FITC-dextran (70 kDa) in the superficial (ganglion cell layer), intermediate (inner plexiform layer), and deep (outer plexiform layer [OPL]) retinal vasculatures. The z-series of images by scanning confocal microscopy showed that the leakage of the perfused dextran could be observed in some areas of all three layers of Akita mouse vasculature but not in those of nondiabetic mice or Akita.PlGF^−/− mice (Supplementary Fig. 5).

We examined the expression of the tight junction protein ZO-1 and the adhesion protein VE-cadherin, both of which play critical roles in maintaining the integrity of BRB complex. The WB results showed that expression of the two proteins was significantly decreased in the Akita diabetic mice compared with the nondiabetic control mice, suggesting impaired BRB function. Protein expression was significantly increased in the Akita.PlGF^−/− diabetic mouse retina compared with the nondiabetic PlGF^−/− control mice, suggesting an enhanced BRB function (Fig. 5A–D). Furthermore, double labeling showed that ZO-1 colocalized with the vascular marker CD31/PECAM-1, similar to the nondiabetic PlGF^−/− mice. However, the signal intensity and area of ZO-1 immunoreactivity were evidently reduced in Akita mice compared with those of CD31 (Supplementary Fig. 6). The mRNA expression of the two factors was consistent with their proteins: decreased in Akita mice, but increased in Akita.PlGF^−/− mice (Fig. 5E and F). These results suggested PlGF deletion in Akita mice protects against diabetes-caused degradation of ZO-1 and VE-cadherin proteins and also upregulates both genes.

To give insight into the mechanisms regulating the increased expression of ZO-1 and VE-cadherin in these diabetic mouse retina, we examined mRNA expression of sonic hedgehog (SHH), angiopoietin (Ang)-1, activator protein-1, Sry-related HMG-box (Sox7), T-cell acute lymphocytic leukemia-1 (TAL1 or SCL), transcription factor 3 (TCL3 or E47), and LM02, all of which were known to regulate the expression of ZO-1 and/or VE-cadherin (28–30). We found that the expression of Ang-1 and SHH was concomitantly related with that of ZO-1 and VE-cadherin: decreased in Akita diabetic mouse retina, but increased in Akita.PlGF^−/− compared with their respective nondiabetic control mice (Fig. 5G and H). The corelated expression patterns suggested that Ang-1 and SHH are likely involved in the regulation of ZO-1 and VE-cadherin expression in the diabetic conditions.

Akt is a key survival factor that protects retinal cells from various impairments, including diabetes (31). We examined the extent and cellular localization of p-Akt, which is the activated form of Akt protein, by WB and IF staining. The WB results showed that Akt phosphorylation expressed in the WT mouse retina was abolished by diabetes (5-month diabetes duration) but was significantly increased in the retinas of Akita.PlGF^−/− mice compared with its nondiabetic PlGF^−/− control mouse retina (Fig. 6A). The p-Akt was not detected by IF staining in the retinas of the two nondiabetic mice or the Akita diabetic mice but was significantly increased in the retinas of Akita.PlGF^−/− mice. p-Akt immunoreactivity was largely localized to the photoreceptor inner segment layer, outer nuclear layer and OPL in the PlGF^−/− diabetic mouse retina (Fig. 6B–E). Interestingly, p-Akt immuoreactivity appeared to be present in the processes of Müller glial cells (Fig. 6E and F). The induction of Akt phosphorylation in these cell types suggests both autocrine and paracrine protective mechanisms may exist in the retinas of Akita.PlGF^−/− mice.

To investigate whether the HIF1α-VEGF signaling pathway is inhibited in Akita.PlGF^−/− mice, we examined the expression of HIF1α, several VEGF family proteins (PlGF, VEGF-A, VEGF-B), and the three tyrosine kinase receptors (VEGFR1–3) in the Akita and Akita.PlGF^−/− diabetic mice strains and their nondiabetic mouse controls. The WB results (Fig. 7A and B) showed that 1) PlGF protein expression was negative in PlGF^−/− and Akita.PlGF^−/−, further confirming the deletion of the PlGF gene in the PlGF^−/− and Akita.PlGF^−/− mouse strains; 2) protein expression of HIF1α, PlGF, VEGF-A, and VEGFR1–3 was significantly upregulated in Akita mouse retina compared with the nondiabetic WT mouse retina, implicating the HIF1α-VEGF pathway in the pathogenesis of DR; 3) protein expression of HIF1α, VEGF, and VEGFR1–3 was significantly downregulated in Akita.PlGF^−/− mice compared with nondiabetic PlGF^−/− mice, suggesting the inhibition of HIF1 α-VEGF pathway is likely one of the protective mechanisms; and 4) VEGF-B protein expression was of the same level for all four mice stains, suggesting VEGF-B may not be involved in this type 1 diabetic model, although there is evidence supporting that VEGF-B is important in the type 2 diabetic model with obesity (32,33).

In addition, we examined the extent and cellular localization of the three important elements in the VEGF signaling pathway: p-VEGFR1, p-VEGFR2, and p-eNOS. Immunoreactivity of p-VEGFR1 and p-VEGFR2 prominently localized in the OPL; the signal intensity was increased in the retinas of Akita diabetic mice compared with the other three mouse strains. Double labeling showed that p-VEGFR1-immunopositive(+) cells are either CD31(+) or CD31(−). The p-VEGFR1(+)/CD31(−) cells were present in all the four mouse strains, suggesting the constitutive expression, but the p-VEGFR1(+)/CD31(+) cells were only present in Akita diabetic mice, indicating the induced expression by diabetes (Fig. 7C–F). The expression pattern and cellular localization of p-VEGFR2 were similar to that of p-VEGFR1 (Fig. 7G–J). p-eNOS was detected only in the Akita mice, not in the others; its immunoreactivity colocalized with CD31(+) cells or CD31(−) cells. The p-eNOS(+)/CD31(+) cells and p-eNOS(+)/CD31(−) cells were localized in the neural fiber layer/ganglion cell layer (Fig. 7K–N).

ICAM-1, VCAM-1, CD11b, and CD18 play important roles in regulating leukocyte-EC adhesion. We examined the protein and/or mRNA expression of these factors. The results showed that protein and/or mRNA expression of ICAM-1, VCAM-1, CD11b, and CD18 were increased in the retinas of both Akita and Akita.PlGF^−/− mice (Fig. 8). The number of leukocytes adhering to the retinal vasculature was equal between Akita and Akita.PlGF^−/− mice, both of which were significantly greater than WT or PlGF^−/− nondiabetic mouse retina (Supplementary Fig. 7). These results suggest protection from diabetes complications due to PlGF deficiency is likely independent of suppression of leukocyte-EC adhesion or retinal leukostasis.

In the current study, we developed a novel Akita.PlGF^−/− diabetic mouse strain. Investigation of the retinas from Akita.PlGF^−/− mice suggests that PlGF deficiency prevents the diabetes complications of acellular capillary, retinal cell death, and BRB breakdown and also confers additional protection, as shown by the increased expression of ZO-1 and VE-cadherin. To gain further insight into the protective mechanisms, we investigated several important pathophysiological pathways and events in DR. First, we found that PlGF deletion in Akita mice led to an increased Akt phosphorylation, which is an important survival factor and may contribute to the protective mechanisms. Second, we found that the expression of SHH and Ang-1 is corelated with that of VE-cadherin and ZO-1. The findings are in agreement with the previous reports that demonstrated SHH and Ang-1 play important roles in preserving the integrity of blood-brain barrier (34) and BRB (35), respectively. Third, we found that the HIF1 α-VEGF signaling pathway, which is the key mediator of diabetic retinal complications, was inhibited due to PlGF deficiency. Last, we found that leukocyte-EC adhesion, or retinal leukostasis, which is a relatively early event and contributes to the development of DR (36), was not suppressed by knocking out PlGF, and therefore, it is possible that the protection is independent of this inflammatory event. It is important to note that although retinal leukostasis has been shown to occur in the early stage in the streptozotocin-induced diabetic model (37), it also happens in the late stages in the Akita model.

The comparisons in this study were considered fairly robust since diabetic mice were always compared with nondiabetic control mice (Akita vs. WT and Akita.PlGF^−/− vs. PlGF^−/−) to control for baseline differences owing to different genetic background, and nondiabetic control mice were always the male siblings of Akita diabetic mice. The four respective mouse strains were maintained by the breeding of Akita/C57 and Akita.PlGF^−/−/PlGF^−/−, and each breeding pair gave rise to offspring of which half were WT or PlGF^−/− and half were diabetic Akita and Akita.PlGF^−/−; only the male diabetic mice were used for experimental analyses. To further ascertain the effects of PlGF deletion on diabetic retinal complications, we characterized the retinal vasculatures of the two nondiabetic mouse strains. The vascular cell density and pericyte number are comparative between PlGF^−/− mice and WT mice (without statistically significant difference). Similarly, histopathological assessment demonstrated the gross anatomy of the retina looks similar among the mouse strains, without showing any retinal degeneration or other abnormities. Despite the similarities, it appears that there are differences in molecular composition and function. For example, the protein expression of VEGFR1–3 was increased in the nondiabetic PlGF^−/− mice compared with the WT mice. The permeability of ³H-mannitol (molecular weight 182 Da) was increased in PlGF^−/− mice compared with the WT mice (approximately two- to threefold), using the lung and kidney as a reference (but not serum). However, the permeability of FITC-dextran (70 kDa) did not show any increase in the knockout mice. It is possible that the differences are attributed to the effect of long-term diabetes on the lung and kidney and the different molecular weight of the two tracers.

Currently, agents targeting VEGF are widely used for the treatment of vascular disorders, including DR. However, there is concern about potential side effects after a long-standing VEGF blockade because new evidence is emerging that VEGF is a survival factor for the choriocapillaris, retinal neurons, and RPE (38–40). One possibility is to target the upstream regulators of VEGF such as prolyl hydroxylase and HIF (41). Because prolyl hydroxylase and HIF regulate a broad spectrum of gene expression, inhibition of their activities could potentially protect the vasculature. Another possible therapeutic intervention would be inhibition of PlGF. In a study on PlGF^−/− mice, Carmeliet and colleagues (6,42) showed that PlGF regulates angiogenesis and vascular permeability in pathological conditions. The results from the current study showed that deletion of PlGF led to not only the suppression of diabetes complications, but also an increase in junction and adhesion factors. These findings suggest that inhibition of PlGF signaling most likely produces the equivalent efficacy of blocking VEGF, but activation of protective factors in PlGF deficiency may also confer additional benefits.

Future work is needed to determine whether activation of Akt survival signaling and inhibition of VEGF signaling pathway are the causes of protection due to PlGF deficiency; how PlGF coordinates the interaction of VEGF, VEGFR1, and VEGFR2 in the background of DR; and the effect of PlGF deficiency upon other biological pathways involved in protection, such as tumor necrosis factor-α, the kallikrein-kinin system, and altered O-GlcNAc signaling. The Akita.PlGF^−/− mouse is a good tool to use in the study of the in vivo effect of these kinds. From this study, it is now apparent PlGF plays a critical role in the pathogenesis of DR.

Acknowledgments. The authors thank Dr. Zhenhua Xu, Wilmer Eye Institute, for her helpful discussions about the WB experiments, and Dr. Junsong Gong, Wilmer Eye Institute, for his technical assistance.

Funding. This study was supported by research grants from the National Institutes of Health EY017164 (H.H.) and EY022383 (E.J.D.), Research to Prevent Blindness, and Wilmer Pooled Professor Funds (H.H.).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. H.H. designed the study, performed experiments, interpreted data, and wrote the manuscript. J.H., D.J., Y.W., Y.L., S.W., and E.J.D. performed experiments and interpreted data. G.A.L. and R.D.S. discussed the study and reviewed and edited the manuscript. P.C. generated and provided the PlGF knockout mice. H.H. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.