VEGFC Reduces Glomerular Albumin Permeability and Protects Against Alterations in VEGF Receptor Expression in Diabetic Nephropathy

Systemic endothelial dysfunction is an initiating step in the development of vascular damage in diabetes (1,2) and is associated with microalbuminuria (urinary albumin secretion 30–300 mg/day) (3). It is widely accepted that microalbuminuria indicates disruption of the glomerulus and is the earliest clinically detectable indicator of incipient diabetic nephropathy (DN) (4). Glomerular endothelial cells (GEnCs) restrict the passage of protein across the highly specialized glomerular capillary wall via the luminal facing endothelial glycocalyx (eGLX), which consists of proteoglycans, glycosaminoglycans (GAGs), glycoproteins, and trapped soluble plasma proteins. GEnCs have transcellular fenestrations that form in attenuated areas of GEnC cytoplasm, which constitute 20–50% of the entire endothelial cell surface (5). Fenestrae are necessary for the high permeability of the glomerulus to water and small solutes (5). However, fenestrae are covered by eGLX (6), which is known to regulate vascular permeability (7–10) by limiting the passage of charged macromolecules (11).

Vascular endothelial growth factor (VEGF) A is highly expressed by podocytes in the glomerulus (12) and provides essential maintenance signals for GEnC, including those for survival and regeneration (13), maintenance of fenestrations (14), and regulation of solute flux and protein passage (15,16). Systemic and glomerular vessels become more “leaky” in pathological conditions where VEGFA bioavailability is increased, such as cancer, retinal disease (17), and early DN (18,19). Blockade of VEGFA in these experimental models reduces vascular permeability (20,21), including proteinuria in DN (6,22). However, clinical anti-VEGFA therapies (23–25) can cause proteinuria and hypertension (26), confirming that some VEGFA is necessary for endothelial maintenance. Therefore, strategies that block the adverse effects of excessive VEGFA, but do not remove it completely, would be attractive for DN.

VEGFC, a lymphangiogenic growth factor, is also expressed by podocytes (27), despite an absence of lymphatic vessels in glomeruli. VEGFC typically signals through VEGF receptor (VEGFR) 3, which is predominantly expressed in lymphatic endothelial cells, but can induce “maintenance” signals similar to those of VEGFA in vascular endothelial cells (28) acting through VEGFR3, VEGFR2, or heterodimers of both (29). VEGFC can also induce fenestration formations in the vascular endothelium, but to a lesser extent than VEGFA (30).

VEGFA predominantly signals through the phosphorylation of VEGFR2 (31). We have shown that VEGFA induces transient VEGFR2 phosphorylation after 2 min in human conditionally immortalized GEnCs (ciGEnCs) (32). VEGFC also induced VEGFR2 phosphorylation in these cells, maximally at 30 min, but did not induce VEGFR3 phosphorylation. VEGFC induced a correspondingly slower rise in intracellular calcium than VEGFA and did not result in vascular endothelial cadherin phosphorylation. We hypothesize that such differences in intracellular events underlie our observation that VEGFA increases the permeability of ciGEnC monolayers to protein, whereas VEGFC reduces protein permeability and blocks the effect of VEGFA (32). Additionally, VEGFC enhances the eGLX layer in vitro, in contrast to VEGFA, which increases eGLX shedding (32,33).

Together, these data suggest that VEGFC may be able to ameliorate albuminuria by reducing the permeability of GEnC to protein while retaining high water permeability. Thus, the effect of VEGFC on raised glomerular albumin permeability was assessed under various conditions, and VEGFC receptor signaling was explored. Finally, the mechanisms of VEGFC action on GEnC barrier properties were investigated with a specific focus on the eGLX.

Human ciGEnCs were maintained and propagated in culture as described previously (34). ciGEnCs stably transfected with GFP (green fluorescent protein)-tagged VEGFR2 (ciGEnC/GFP-VEGFR2) (a gift from Dr. H. Mellor, University of Bristol, Bristol, U.K.) were cultured similarly.

Human recombinant VEGFA and VEGFC (R&D Systems) were reconstituted as done previously and were used at a final concentrations of 1 nmol/L (42 ng/mL) and 10 nmol/L (210 ng/mL), respectively, as described previously (32).

All primary and secondary antibodies that were used and their dilutions are described in Supplementary Table 1.

Immunofluorescence was carried out on fresh frozen mouse kidney sections as previously described (35) and on GEnC as previously described (33). The integrated density of glomerular VEGFC was semiquantified using ImageJ (15 glomeruli per animal) and expressed as the fold change to littermate control (LMC) mice. For cellular VEGFR3 quantification, images were acquired using the InCell Analyzer 2200 System (GE Healthcare) as previously described (36). In brief, analysis was performed using the InCell Analyzer 1000 Workstation Multitarget analysis algorithms, using phalloidin to define the cell and DAPI to segment the nuclei. VEGFR3 is reported as the mean fluorescence intensity per cell.

ciGEnCs, tissue, and sieved glomeruli and flow-through were lysed in radioimmunoprecipitation assay buffer containing phosphatase and protease inhibitors. Samples were quantified (Pierce bicinchoninic acid assay; Thermo Fisher Scientific) and Western blotted.

RNA was extracted from isolated mouse glomeruli using an RNeasy Mini Kit (Qiagen) according to manufacturer instructions. Primer sequences were predesigned, selected from http://pga.mgh.harvard.edu/primerbank/, and optimized as previously described (33). Human GAPDH primers were used as previously described (33), with other primer pairs described in Supplementary Table 2. Quantitative PCR was carried out as previously described (33). 2^{−^^CT} (relative fold change) was calculated from the CT values generated.

Mouse lungs and glomeruli were freshly harvested in ice-cold PBS containing 100 μm of Na₃VO₄ and protease inhibitor cocktails (1:100) (Sigma-Aldrich). These tissues were then treated with VEGFA and VEGFC or vehicle at 37°C for 2, 15, or 30 min. PBS was removed, and samples were resuspended in 100 μL of NP-40 lysis buffer then precleared. Equal quantities of protein, to 500 µL, were incubated at 4°C for 18 h with 1.5 µg/100 µg protein of anti-mouse VEGFR3 (ALF4) (eBioscience). Subsequently, 60 μL of equilibrated A/G PLUS-agarose beads (Santa Cruz Biotechnology) were added to each sample and incubated at 4°C for 6 h. Beads were centrifuged, and supernatant samples were taken for immunoblotting. Thirty microliters of immunoprecipitation samples and equal protein concentrations of supernatant and total protein were loaded onto an SDS-PAGE gel and Western blotted.

Podocin reverse tetracycline-controlled transcriptional activator (podrtTA) mice expressing the transgene exclusively in podocytes (37) were a gift from Dr. J. Kopp (National Institutes of Health, Bethesda, MD) (FVB/mixed). Tet-O-VEGFC mice (FVB/N), containing full-length mouse VEGFC, were used as previously described (38). Heterozygous podrtTA mice were crossed with heterozygous Tet-O-VEGFC mice to obtain mice heterozygous for both transgenes (podVEGFC) and LMC mice. Animals were kept according to the Guidelines on the Use of Animals in Research, and the number of animals used was kept to a minimum. Mice were housed in a pathogen-free environment at 21°C, with a 12-h light/dark cycle, and all received a standard laboratory animal diet (Beekey Feeds) and water ad libitum.

Ear notches were digested in 20-μL digestion buffer (1× PCR buffer [supplied with Hotmaster taq]: 0.0045% NP-40, 0.0045% Tween 20, 1 mg/mL proteinase K). Samples were digested for 2 h at 60°C, then for 15 min at 95°C. PCR for podrtTA was carried out as originally described (37). For Tet-O-VEGFC, primers are described in Supplementary Table 2. The PCR reaction (total volume 13 μL) contained 6.5 µL of Mastermix buffer (Amplitaq Gold 360; Applied Biosystems) and 0.5 µmol/L primers. PCR cycle conditions were as follows; 1 cycle at 95°C for 10 min, 42 cycles of 1 cycle at 95°C for 30 s, 68°C 30 s, 72°C for 60 s, and 72°C for 7 min.

Expression of podocyte VEGFC was induced by doxycycline (2 mg/mL) in the drinking water with sucrose (5% w/v) as previously shown (37). Doxycycline was replaced every 3–4 days and protected from light at all times. If mice did not drink the doxycycline water, they received chow supplemented with 625 mg/kg doxycycline (Harlan Laboratories Ltd., Shardlow Derby, U.K.), which was observed to increase VEGFC mRNA expression to a similar extent (0.9-fold change in VEGFC mRNA expression compared with doxycycline-supplemented water; data not shown). Some animals were given sucrose alone. Urine samples were collected weekly in metabolic cages with environmental enrichment up to 6 h at a time.

Twelve- to 14-week-old male C57BLKsJ-db/db (db/db) mice (Harlan Laboratories Ltd.) were used as a model of spontaneous type 2 diabetes. Age-matched, male C57BLKsJ-db/+ (db/+) lean mice were used for comparison. mRNA was extracted from enriched glomeruli from 7-week-old db/db and db/+ mice. Urinary albumin/creatinine ratio (uACR) and glucose levels were measured as below.

Diabetes was induced in 8- to 10-week-old podVEGFC and LMC mice using a low-dose (50 mg/kg/day over 5 consecutive days) streptozotocin (STZ) induction protocol as previously described (39) (as advised by Diabetic Complications Consortium [DiaComp], https://www.diacomp.org). Mice were checked for adverse effects after dosing; animals that demonstrated dehydration (including weight loss, slack skin, or bloating due to constipation, revealed postmortem in the first few diabetic mice) were supplemented with “mash” and DietGel with subcutaneous 0.9% saline injections as necessary. Blood glucose levels were monitored weekly from a tail-tip blood droplet using an Accu-Chek Blood Glucose Meter (Aviva). Animals with blood glucose >15 mmol/L 2 weeks after injections were included in the study. Body weight was monitored regularly. Mice received doxycycline from 0 weeks (prevention) or 4 weeks (rescue) after STZ administration. Urine was collected as previously described, although as animals became more polyuric, the collection time was reduced to minimize water weight loss and stress.

This assay was developed and extensively characterized previously (40). In brief, mice were anesthetized and the kidneys perfused with 4% BSA (RingerBSA), followed by 36.5 µg/mL octadecyl rhodamine B chloride (R18) (Thermo Fisher Scientific) to label cell membranes, then 30 μg/mL Alexa Fluor (AF) 488-BSA (labeled albumin) (Life Technologies). Glomeruli were isolated from kidney cortex by graded sieving. For glomeruli treated ex vivo, glomeruli loaded with AF 488-BSA were held on ice until treated with VEGFA, VEGFC, both VEGFA and VEGFC, or an equal volume of Ringer-BSA (vehicle) for 1 h at 37°C. Individual glomeruli were transferred to a Nikon TIE Inverted Confocal Microscope and trapped before the perfusate was switched from 30 μg/mL AF 488-BSA to 30 μg/mL unlabeled BSA. The rate of decrease in fluorescence intensity within capillary loops was quantified, and from this apparent glomerular albumin solute permeability (Ps’alb) was calculated (40):

P_S’alb = −kR/2

We have shown previously that chondroitinase and hyaluronidase, given acutely i.v. 30 min before sacrifice, raises Ps’alb significantly and causes a reduction in eGLX coverage (40). In the same cohort of animals, glomeruli were isolated and incubated in VEGFC as described above, and Ps’alb level was measured and compared. Further, chondroitinase and hyaluronidase were administered chronically, via an osmotic minipump (0.25 µL/h) (Charles River UK Ltd., Kent, U.K.), and cannulated to the jugular, for a total of 2 weeks (0.087 mU/g chondroitinase and 15 mU/g hyaluronidase, or vehicle over a 24-h period, equivalent to 0.004 mU/h chondroitinase and 0.625 units/h hyaluronidase). In the final week, a VEGFC intervention was given to half of the enzyme-treated mice i.p. daily (100 ng/g body wt as previously described) (41). Glomeruli were isolated, and Ps’alb was measured.

Mice were whole-body perfusion fixed with glutaraldehyde containing Alcian blue. Electron micrographs were taken using a Technai 12 Electron Microscope, and images were analyzed as previously described (39).

Mouse uACRs were quantified as previously described (39) and expressed as the log fold change from baseline.

To achieve a >97% pure population of isolated glomeruli, Dynabeads were cardiac perfused as originally described by Takemoto et al. (42) and placed in RNA later (Invitrogen). Alternatively, to obtain a glomerular enriched population, kidneys were mashed, and glomeruli isolated through graded sieving and collected from 75-µm² pore sieves for protein or mRNA extraction. To demonstrate glomerular enrichment, kidney cortex, flow-through, and sieved glomeruli were collected, and lysates were extracted, quantified, and Western blotted as described above. Blots were probed with anti-nephrin, anti-CD13, and anti-actin antibodies (Supplementary Table 1).

Fresh frozen mouse kidney sections were fixed in 4% paraformaldehyde and permeabilized using 0.1% Triton X-100. Samples were then treated with the Proximity Ligation Assay Kit (Olink Bioscience) according to the manufacturer instructions. In brief, primary antibodies or matched concentration IgG controls in various combinations were incubated together overnight (for details, see Supplementary Table 1), followed by incubation with goat MINUS and rabbit PLUS probes. Nuclei were counterstained using DAPI (1:1,000 diluted in proximity ligation assay (PLA) kit 0.01× wash buffer B). Images were quantified by a blinded analyzer using ImageJ, by counting the number of punctate dots through z-stacks of images and normalizing to the number of nuclei within the field of view. Three representative z-stacks from each sample were analyzed.

Paraffin-embedded mouse kidney sections were stained using a periodic acid-Schiff (PAS) kit (Sigma-Aldrich) according to the manufacturer instructions. Six glomerular images were analyzed per mouse and quantified as previously described (40).

Mouse tissue sections were hydrated followed by counterstaining in a 0.1% (w/v) Direct Red 80 powder (Sigma-Aldrich) in a 1.3% saturated aqueous solution of picric acid (VWR Chemicals) for 90 min at room temperature, then dehydrated and mounted. Images were taken using a bright-field (to indicate collagen I and III) and a polarized lens (to distinguish between type 1 (yellow-orange) and type 3 (green-white) newly laid collagen fibers).

All statistical analysis was carried out on a minimum of three separate experiments. In Ps’alb experiments, data are presented as the means of measurements from glomeruli to show the spread of data points, but for statistical purposes the means from individual animals were used. In experiments with multiple treatments, a one-way ANOVA was used with Bonferroni’s multiple-comparison post hoc test, unless indicated otherwise. A P value of less than 0.05 was considered statistically significant. Error bars indicate the SD.

To test whether VEGFC could protect against the effects of raised VEGFA, an ex vivo isolated glomerular Ps’alb assay was used that we recently published (40). VEGFA significantly increased Ps’alb in isolated glomeruli as anticipated in the FVB/mixed background (Fig. 1A) and db/− (lean) mice (Fig. 1B). Importantly, this effect was prevented by VEGFC, yet VEGFC alone did not change Ps’alb. To understand whether VEGFC could rescue Ps’alb induced by diabetes, glomeruli were isolated from db/db (type 2 diabetic) mice and stimulated with VEGFC ex vivo, as described above. Ps’alb in diabetic glomeruli was significantly raised, as anticipated, whereas it was significantly rescued by VEGFC (Fig. 1C). db/db mice were 12–14 weeks old, had significant proteinuria (Fig. 1D), and were hyperglycemic (Fig. 1E). We confirmed that there was a significant increase in VEGFA mRNA expression in enriched glomeruli from db/db mice (Fig. 1F). We also confirmed that VEGFC was cytoprotective for GEnCs, an important protective effect of VEGFA. Using a Trypan Blue viability assay, we determined that GEnCs were protected from death (serum free, volume reduced media) by VEGFC over 24 h in a manner similar to that by VEGFA (Supplementary Fig. 1A). This was associated with increased Akt phosphorylation over time by VEGFC and VEGFA (Supplementary Fig. 1B), suggesting that VEGFC promotes survival in GEnCs in a manner similar to VEGFA in vitro.

To determine whether VEGFC could be protective in vivo, we generated a podVEGFC mouse (Fig. 2A). First, glomeruli were Dynabead isolated from podVEGFC and LMC mice after receiving doxycycline for 3 weeks. Quantitative PCR revealed a significant increase in normalized VEGFC glomerular mRNA expression in podVEGFC mice (Fig. 2B). This was supported by immunofluorescence showing colocalization of VEGFC with nephrin (Fig. 2C) in podVEGFC mice. Semiquantification demonstrated significantly increased glomerular protein expression in podVEGFC mice (Fig. 2D). Levels of VEGFC mRNA were similar in glomeruli isolated by Dynabead and sieving protocols (data not shown); therefore, sieving was used to isolate glomeruli hereafter.

Glomerular parameters in podVEGFC mice were compared with LMC mice. There was no renal hypertrophy (Fig. 2E), mesangial matrix expansion (Fig. 2Fi and G), or ultrastructural changes in GBM thickness, podocyte foot process width, slit diaphragm length (Fig. 2Fii and H), fenestration density (Fig. 2I), or uACR (Fig. 2J) up to 8 weeks after the induction of VEGFC expression. Together, these results demonstrate that VEGFC does not significantly affect glomerular structure or function over prolonged periods of exposure, in contrast to VEGFA (43).

Glomerular sieving successfully enriched the glomeruli population (Supplementary Fig. 2). In podVEGFC mice, there was a trend toward a reduction in VEGFR2 in the glomerular enriched fraction (Fig. 3A), but no significant reduction in VEGFR3 mRNA expression (Fig. 3B). VEGFA mRNA expression was significantly reduced (Fig. 3C). In nonpermeabilized cells, both VEGFA and VEGFC reduced cell surface VEGFR3 expression at 2 min (Fig. 3D), whereas total VEGFR2 and VEGFR3 expression was unaffected by VEGFA or VEGFC at 2 min (Supplementary Fig. 3). VEGFR3 and VEGFR2 heterodimerization by VEGFC, as reported previously (29,44), was investigated ex vivo in isolated glomeruli using immunoprecipitation. VEGFR2 was not pulled down by VEGFR3 in response to VEGFC (Fig. 3E). However, heterodimers were formed in response to VEGFC in mouse lungs as anticipated (Fig. 3F). Nilsson et al. (29) demonstrated that a PLA was more sensitive to detect VEGFR2/VEGFR3 heterodimerization than immunoprecipitation. Thus, a PLA was used on fresh frozen sections from podVEGFC and LMC mice (Fig. 3G). Optimization, by replacing primary antibodies with matched concentration IgG, demonstrated the specificity of punctate staining (Supplementary Fig. 4). There was a significant increase in punctate fluorescent events, indicating heterodimerization, throughout the z-stacks of each glomerular cell in podVEGFC mice compared with LMC mice (Fig. 3H).

Endothelial dysfunction develops early in diabetes and manifests as microalbuminuria in the kidney. We have shown that VEGFC ameliorates raised Ps’alb associated with type 2 diabetes in isolated glomeruli. To understand whether VEGFC could prevent albuminuria in DN, doxycycline was administered before STZ injections (Fig. 4A). Diabetic mice (D-) were hyperglycemic (Fig. 4B), and had reduced body weight (Fig. 4C) and increased urine output (Fig. 4D), as anticipated. D-podVEGFC mice did not have significant renal hypertrophy (Fig. 4E) and had a significantly reduced fold change in uACR compared with D-LMC mice (Fig. 4F) (please refer to Supplementary Fig. 5A for absolute values), which was high but variable because of the mixed background of the mice. Diabetes did not significantly increase glomerular basement membrane (GBM) thickness or podocyte foot process width (Fig. 4G and H). However, podocyte slit diaphragm narrowing was evident in D-LMC mice (Fig. 4G and H). D-LMC mice had significantly lower endothelial fenestration density than LMC mice, whereas D-podVEGFC mice did not (Fig. 4G and I).

To understand whether VEGFC could rescue DN after diabetes was established, doxycycline was administered 4 weeks after STZ injections (Fig. 5A). Diabetic mice were hyperglycemic (Fig. 5B) and polyuric (Fig. 5C), as anticipated. There was no significant difference in weight gain between groups; however, these diabetic mice were susceptible to intestinal bloating, which was likely to mask body weight reductions (this protocol was carried out chronologically before the one described above, and this issue was mitigated as described in research design and methods) (Fig. 5D). Of note, this cohort of mice also had a lower starting weight than the intervention cohort in Fig. 4, and sham mice did not gain weight as rapidly, which may confound the results. Elevated plasma uric acid level is associated with diabetes and can be caused by hyperglycemia (45). Plasma uric acid was significantly raised in both diabetic groups, as anticipated (Fig. 5E). Importantly, renal hypertrophy (Fig. 5F) was significantly increased in D-LMC mice, but significantly reduced in D-podVEGFC mice. The fold change in uACR (Fig. 5G) (please refer to Supplementary Fig. 5B for absolute values) was significantly increased in D-LMC mice but not in D-podVEGFC mice. Mesangial matrix expansion was significantly reduced inD-podVEGFC mice (Fig. 5H and K), supported by decreased collagen deposition (Fig. 5I and J). There was no significant effect of diabetes on GBM thickness, which is consistent with an early phenotype of DN (Fig. 6A and B). D-LMC mice had significantly increased podocyte foot process width, and both D-LMC and D-podVEGFC mice had significantly reduced slit diaphragm width (Fig. 6B). D-LMC mice also had significantly reduced fenestration density (Fig. 6C), demonstrating the early development of DN. Together, these data from both diabetic cohorts suggest that early VEGFC intervention may reduce the development of early DN.

The induction of VEGFC expression was confirmed in glomerular fractions from D-podVEGFC mice (Fig. 7A) in a manner similar to that from podVEGFC mice. There was no significant change in endogenous glomerular fractions of VEGFC expression between D-LMC and sham-LMC mice (Fig. 7B). VEGFA mRNA from glomerular fractions was increased by diabetes (Fig. 7C), as previously shown (46), with a trend toward an increase in D-podVEGFC mice (Fig. 7C). VEGFR2 and VEGFR3 mRNA from glomerular fractions were significantly increased in D-LMC mice, but not in D-podVEGFC mice (Fig. 7D and E). This was supported by VEGFR2 protein analysis, whereby VEGFR2 expression was significantly reduced in D-podVEGFC mice (Fig. 7F and G). Of note, VEGFR2 protein was undetected in sieved flow-through fractions (Supplementary Fig. 2E), suggesting that VEGFR2 protein changes were glomerular specific. VEGFR3 protein analysis was not significant (Fig. 7F and H). In summary, VEGFC prevented VEGFR changes associated with DN.

We have previously shown that VEGFC increases the synthesis of sulfated and nonsulfated GAGs in GEnC-GLX in vitro (33). Since macromolecular barrier properties in GEnCs are governed by eGLX (47), we investigated the potential for VEGFC to reduce Ps’alb in an eGLX-dependent manner. Previously, we demonstrated that a bolus of hyaluronidase and chondroitinase 30 min before sacrifice significantly reduced eGLX coverage and significantly increased Ps’alb (40). In the same cohort of mice, isolated glomeruli were stimulated with VEGFC for 1 h. In these glomeruli, VEGFC rescued the effect of eGLX disruption on Ps’alb (Fig. 8A). We followed this experiment with chronic, systemic exposure to these enzymes (2 weeks) with VEGFC i.p. injections given in the last week. Enzymes significantly increased Ps’alb in isolated glomeruli. However, this was significantly prevented in glomeruli from mice given the VEGFC intervention (Fig. 8B). Electron micrographs from mice in parallel experiments demonstrated that GAG enzymes significantly reduced GEnC eGLX depth, but not in the presence of VEGFC (Fig. 8C and D). Interestingly, podocyte glycocalyx was relatively unaffected (Fig. 8D). To demonstrate no off-target effects of the GAG enzymes, glomerular ultrastructural parameters were measured. There were no significant effects of enzymes on GBM width (although the combination of enzymes plus VEGFC significantly reduced GBM width), slit diaphragm width, podocyte foot process density (Fig. 8E) (although there was a significant overall effect), or endothelial fenestration density (Fig. 8F). In summary, VEGFC can protect from increased Ps’alb under eGLX dysfunction, suggesting the possibility that VEGFC protection from albuminuria is eGLX dependent.

In this study, we demonstrate for the first time that VEGFC can protect against VEGFA-induced Ps’alb and can restore Ps’alb in type 2 diabetes. Also, it can ameliorate early DN in type 1 diabetes; reducing fibrosis, kidney hypertrophy, loss in glomerular endothelial fenestration density, and albuminuria. There was an 89% reduction in albuminuria when VEGFC was used preventatively and a 67% reduction when VEGFC was used to rescue changes. These were comparable to previously published results demonstrating that glomerular overexpression of angiopoietin-1 in diabetic mice caused a 70% reduction in albuminuria (48) and glomerular overexpression of VEGFA₁₆₅b in diabetic mice caused an 80% reduction in albuminuria (49). That VEGFC reversed the increased Ps’alb in type 2 diabetic glomeruli confirms the effect of VEGFC across mouse strains and diabetes type. This demonstrates that VEGFC can act as an antialbuminuric agent in experimental DN.

Diabetes-induced damage to glomerular ultrastructure included foot process flattening and reduced endothelial fenestration density (39,50,51). Reduced fenestration density is an established feature of DN in humans, which is positively correlated with uACR (51). It has been suggested that loss of GEnC fenestration density reflects both glomerular filtration rate and uACR better than changes in podocyte parameters (52). Of note, we did not see any changes in GBM thickness. This is surprising, but may be a reflection that podocyte damage is not yet severe enough to impact maintenance of the GBM (53) or may simply be a reflection of a mixed strain background. Notably, the D-podVEGFC mice did not have significantly reduced fenestration density. It is well established that GEnC fenestration density is reduced in DN. Fenestrations are too large to restrict albumin passage; however, the loss of fenestration density, or organization, may reflect a loss of filtration control. Also, renal hypertrophy was reduced in D-podVEGFC mice. Although our mice have early DN, they also have evidence of fibrosis, demonstrated by glomerular PAS staining and increased glomerular and tubular collagen I and III deposition, as is predicted in diabetes (54). Kidney weight is partially reduced in the D-podVEGFC mice. This may indicate a reduction in early fibrosis as a result of reduced albumin filtration and therefore reduced tubular toxicity (55).

We suggest that the differing effects between VEGFC and VEGFA may be due to differential activation of VEGFR2. We have previously shown that VEGFC does not induce VEGFR3 phosphorylation in ciGEnCs, yet promotes delayed VEGFR2 phosphorylation at 30 min compared with rapid phosphorylation by VEGFA at 2 min (32). However, we show that VEGFC causes rapid loss of cell surface VEGFR3 expression, suggesting internalization of VEGFR3 by VEGFC. In otherwise healthy glomeruli, chronic glomerular VEGFC expression induced VEGFR2 and VEGFR3 heterodimerization and reduced glomerular VEGFA mRNA expression without inducing glomerular fibrosis or a loss of glomerular function or changes in glomerular ultrastructure. This is important since reduced glomerular VEGFA expression can lead to renal thrombotic microangiopathy and proteinuria (26,43), yet increased glomerular VEGFA expression can lead to collapsing glomerulopathies and proteinuria (43). That VEGFC has no detrimental effect on glomerular function is fundamentally important to future studies aiming to manipulate the VEGFC pathway in order to protect glomeruli. It is known that VEGFC induces VEGFR2/VEGFR3 heterodimerization in lymphatic endothelial cells (44). Here, we demonstrate that chronic exposure to VEGFC induces heterodimerization in glomeruli. Also, glomerular VEGFA expression is repressed, similar to when Ang-1 is overexpressed (56), potentially altering receptor availability to VEGFC and thus facilitating heterodimerization. Interestingly, heterodimerization of VEGFR3 with VEGFR2 leads to reduced phosphorylation of some tyrosine sites in lymphatic endothelial cells (44). These results suggest that VEGFR3 is involved in VEGFC signaling, confirm functional activity of overexpressed VEGFC, and demonstrate for the first time in vivo that mouse VEGFC induces VEGFR2/VEGFR3 heterodimerization. Thus, the change in glomerular VEGFR complex formation may explain the different temporal effects on VEGFR2 phosphorylation by VEGFA and VEGFC. Interestingly, VEGFC also protected from a diabetes-induced increase in VEGF receptor expression in glomeruli. These data support previous observed increases in VEGFR2 expression in DN (57,58), although changes in glomerular VEGFR3 expression are not well documented. VEGFR2 and VEGFR3 expression were both reduced in D-podVEGFC mice; therefore, the potential for heterodimerization remains the same. However, the potential for receptor phosphorylation may be reduced, thereby reducing downstream signaling. Glomerular VEGFA is known to be upregulated in early DN (18) and most glomerular VEGF-receptor complexes are localized to GEnCs (18). This indicates that VEGFA predominantly regulates the endothelial phenotype during diabetic glomerulopathy, although endothelial cells are thought to “signal back” to the podocytes, inducing their injury (43). Hohenstein et al. (18) demonstrated that mildly injured glomeruli in early DN (relating to glomeruli in our study) had increased VEGFA-bound receptor complexes, whereas more severely injured glomeruli had decreased VEGFA-bound receptor complexes. Perhaps the key for VEGFC-mediated protection is to competitively reduce VEGFA receptor signaling, particularly in early DN.

VEGFC induces lymphatic hyperplasia and is associated with lymphangiomas and lymphangiosarcomas (59). Subcutaneous VEGFC overexpression also increases obesity and insulin resistance (60), and VEGFC is implicated in inflammation in late DN (61). Therefore, it is unlikely that VEGFC can be given therapeutically in its native form, although, of note, it was used successfully in mice to treat polycystic kidney disease (41) and to treat renal interstitial fibrosis for 14 days (62), and we gave VEGFC daily i.p. for a total of 7 days with no gross morphological effects. Lymphangiogenesis is increased in the tubular interstitium in DN (61), and this is considered a protective mechanism to aid drainage (63), but may also promote inflammation (64). If this mechanism is saturated, it can lead to edema in the kidney, as suggested by Uchiyama et al. (63). Importantly, in a model of unilateral ureteral obstruction, VEGFC treatment, by osmotic mini-pump, attenuated the infiltration of inflammatory cells (60). We have not yet investigated whether lymphangiogenesis is promoted in podVEGFC mice, but this is an important consideration in diabetes and is another potential reason for reduced hypertrophy in podVEGFC diabetic mice. Whether VEGFC can act as a microvascular protectant in other vascular beds in diabetes remains to be seen. In order to harness the effects of VEGFC therapeutically, it may be more appropriate to target the VEGFC signaling pathway. One such approach would be to target the VEGFR2/VEGFR3 complex as was done previously for VEGFR1/VEGFR2 complexes (65). Another approach would be to target signaling pathways that promote eGLX synthesis. If VEGFC can protect from diabetes-induced eGLX damage, then this is another exciting avenue of treatment. We have shown that VEGFC can protect against increased Ps’alb caused by systemic eGLX damage, both in chronic and acute models. We have also previously shown in vitro that VEGFA induced eGLX shedding in GEnC while VEGFC induced eGLX synthesis (33). As highlighted above, the D-podVEGFC mice did not experience significant fenestration density loss. The restoration of filtration control for albumin is potentially due to altered e-GLX coverage of the fenestrations. Since eGLX dysfunction is considered an early insult in DN (66), and the type 1 diabetes models used show an early phenotype, perhaps VEGFC mediates its protection though eGLX restoration in early DN. This will be a focus of future work.

Importantly, we have demonstrated a novel role for VEGFC in the development of early DN.

Funding. This work was supported by a British Heart Foundation (BHF) Basic Science Intermediate Fellowship (FS/10/017/28249) (to M.G., N.R.B., S.L.B., and R.R.F.) and a BHF studentship and project grant (FS/13/9/29957 and PG/15/81/31740) (to K.L.O.); a Kidney Research UK project grant (RP45/2013) (to R.D.R.) and a studentship grant (ST5/2012) (to S.D.); a European Foundation for the Study of Diabetes grant (EFSD/Novartis Programme, 2014) (to K.B.B.); an Medical Research Council Clinician Scientist award (G0802829) (to A.H.J.S.); and a Diabetes Research Wellness Foundation grant (SCA/OF/12/13) (to R.R.F.).

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

Author Contributions. K.L.O. helped to design the experiments, acquired data, interpreted the results, and approved the final version of the manuscript. M.G., N.R.B., S.L.B., K.B.B., S.D., B.P.D., R.D.R., and L.K.F. acquired data, interpreted the results, and approved the final version of the manuscript. C.R.N. advised on electron microscopy imaging and measurements and approved the final version of the manuscript. P.W.M., L.G., K.A., D.O.B., A.H.J.S., G.I.W., and S.C.S. helped to design the experiments, interpreted the results, and approved the final version of the manuscript. R.R.F. conceived and designed the work, interpreted the results, revised the manuscript, and approved the final version. R.R.F. 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.