Nerve growth factor regulates endothelial cell survival and pathological retinal angiogenesis

Abstract The mechanism underlying vasoproliferative retinopathies like retinopathy of prematurity (ROP) is hypoxia‐triggered neovascularisation. Nerve growth factor (NGF), a neurotrophin supporting survival and differentiation of neuronal cells may also regulate endothelial cell functions. Here we studied the role of NGF in pathological retinal angiogenesis in the course of the ROP mouse model. Topical application of NGF enhanced while intraocular injections of anti‐NGF neutralizing antibody reduced pathological retinal vascularization in mice subjected to the ROP model. The pro‐angiogenic effect of NGF in the retina was mediated by inhibition of retinal endothelial cell apoptosis. In vitro, NGF decreased the intrinsic (mitochondria‐dependent) apoptosis in hypoxia‐treated human retinal microvascular endothelial cells and preserved the mitochondrial membrane potential. The anti‐apoptotic effect of NGF was associated with increased BCL2 and reduced BAX, as well as with enhanced ERK and AKT phosphorylation, and was abolished by inhibition of the AKT pathway. Our findings reveal an anti‐apoptotic role of NGF in the hypoxic retinal endothelium, which is involved in promoting pathological retinal vascularization, thereby pointing to NGF as a potential target for proliferative retinopathies.

The neurotrophin nerve growth factor (NGF) is a key player in survival, growth and differentiation of neuronal cells. 6 Bioactive NGF derives from a precursor molecule, called proNGF, 7 which is processed and cleaved either intracellularly 8 or extracellularly by plasmin and matrix metalloproteinases. 9 NGF binds with high affinity to the Tropomyosin receptor kinase A (TrkA), a receptor with tyrosine kinase activity. 6 Upon NGF binding, TrkA dimerizes and is auto-phosphorylated at its intracellular part, 6,10 resulting in activation of signalling pathways, such as the Ras-Raf-MEK-ERK pathway and the PI3K-AKT pathway, which are involved in neuronal survival. 6,11 NGF can also bind with lower affinity to the p75 receptor, which induces the JNK or the NFκB signalling pathway through direct interaction with intermediates, such as RhoA or TRAF proteins. 6,11 The neuroprotective effect of NGF has also been studied in various neurodegenerative diseases of the eye. 12 Phase 1/2 and pilot clinical trials have used NGF eye-drops in an attempt to rescue or reduce the degeneration of retinal neuronal or ganglion cells in retinitis pigmentosa and glaucoma patients. [12][13][14] In addition, patients with DR have increased NGF concentration in the vitreous, as compared to non-diabetic individuals. 4 Besides neuronal cells, NGF may act on immune cells or endothelial cells (EC). [15][16][17][18][19] For instance, NGF prevents mast cell apoptosis, 20 while it also induces mast cell activation through its interaction with platelets, which in turn induce tissue apoptosis in ischemic stroke. 21,22 Endothelial cells express NGF receptors. [17][18][19] NGF activates through TrkA the PI3K/AKT, MEK/ERK and PLCγ/PKC pathways 17,18,23 and may exert pro-angiogenic actions. 19,24 For instance, NGF enhanced the proliferation of brain capillary EC 19 and of human umbilical vein EC and promoted angiogenesis in chorioallantoic membranes of chicken embryos. 17 Furthermore, NGF promoted vascularization in hindlimb ischemia 25 as well as tumor angiogenesis. 26 These observations prompted us to investigate the effects of NGF in the environment of the retina, where neuronal and EC display close spatial and functional interactions. We studied the role of NGF in the context of proliferative retinopathies by using the mouse ROP model. We found a pro-angiogenic effect of NGF, which was associated with downregulation of retinal EC apoptosis and with preservation of mitochondrial membrane potential especially under hypoxic conditions. Our findings therefore indicate that NGF may represent a potential target for proliferative retinopathies.

| Mice and ROP model
Wild type C57BL/6 mice were obtained from Janvier Labs (Le Genest-Saint-Isle, France). Mice were subjected to the ROP model as described. [27][28][29] Briefly, 7-day old pups were exposed for 5 days to hyperoxia (75% O 2 ) together with their nursing mothers. Thereafter, pups and mothers were returned to room air (21% O 2 ). On postnatal day 14 (p14), anesthetized pups received an intraocular injection of anti-NGF antibody (1 μg/eye; Abcam, UK) in the right eye and the same amount of control rabbit IgG (Abcam) in the left eye; injections were performed under a stereoscope (Stemi C-2000; Zeiss, Oberkochen, Germany), as described. 30 In other experiments, pups received from p13 to p16 twice per day eye drops of NGF (Merck Millipore, Darmstadt, Germany) (3 μL, 200 μg/mL) in the right eye and same volume of PBS in the left eye, as described. 31 At p17, pups were killed and eyes were retrieved for further analysis of the retinas. 32 Animal experiments were approved by the Landesdirektion Sachsen, Germany.

| Periodic acid -Schiff staining
For assessment of abnormal retinal angiogenesis, eyes were isolated at p17 from pups subjected to the ROP model and were fixed overnight in 4% PFA at 4°C. Paraffin-embedded sections (4 μm) that included the head of the optic nerve were prepared.
After de-paraffinization of retinal sections by overnight incubation at 60°C and subsequent incubation in Roticlear ® solution (Carl Roth, Karlsruhe, Germany), serial washes with ethanol followed.
Sections were washed again with PBS and tap water and they were stained by immersing the slides first in periodic acid (VWR Chemicals Prolab, Dresden, Germany) and then in Schiff reagent (SAV-liquid production GmbH, Flintsbach, Germany) followed by washes with PBS and staining with Hematoxylin (SAV-liquid production GmbH). Neovascular cell nuclei anterior to the internal limiting membrane were counted in each retinal section and the mean of at least 10 counted sections was calculated per retina. 32 Moreover, images were acquired with a computerized microscope (Zeiss) and analysed with the ZEN software (Zen Software Ltd, Rochdale, UK).

| Apoptosis assay in retinas
Eyes were isolated from pups at p17, snap-frozen and embedded in OCT compound (Tissue-Tek, CA, USA). Frozen serial sections (10 μm) were obtained and kept at −20°C until further processing.
For visualization of apoptotic EC in the retina, sections were fixed for 10 minutes with acetone at −20°C. The slides were left to dry at room temperature and after three washes with PBS, they were permeabilized and blocked for 1 hour in PBS with 3% goat serum/0.2% Triton X-100/5% BSA followed by overnight incubation with a rabbit  For the experiments regarding BCL2 and BAX protein expression, cells were treated with NGF (100 ng/mL; Merck Millipore) in starvation medium (plain EGM without supplement) for 3.5 hours under hypoxia (1% O 2 ). For studying the ERK and AKT signaling pathway, HRMEC were cultured to confluence and exposed overnight to hypoxia in EGM without SupplementMix with 1% FBS; thereafter, cells were treated with 100 ng/mL NGF in plain EGM with 1% FBS and analysed at different time points under hypoxic conditions (1% O 2 ), as indicated in the figures.
Western blot analysis was performed as previously described. 16,33 After the aforementioned treatments, cells were washed twice with ice-cold PBS and lysed in RIPA buffer (Santa Cruz, Heidelberg, Germany) including phosphatase and protein inhibitors (Roche, Mannheim, Germany) and benzonase (Sigma-Aldrich). Thirty μg of protein were boiled at 94°C for 4 minutes, loaded onto 4%-12% gradient SDS

| In vitro apoptosis assay
HRMEC were cultured on gelatin-coated coverslips placed in 24well-or 96-well plates until confluence. On the next day, the medium was changed into plain EGM and cells were exposed to F I G U R E 2 NGF reduces apoptosis of endothelial cells in the course of ROP and under hypoxic conditions in vitro. (A) Sections of p17 retinas of ROP mice that were treated with NGF eye drops or vehicle control (PBS), as described under Figure 1C and D, were analysed for endothelial cell apoptosis, as described in Section 2. The number of apoptotic endothelial cells was counted per retinal section. Endothelial cell apoptosis in the retina is presented as % of control; apoptosis in vehicle-treated retinas was set as 100%. Data are presented as mean ± SEM (n = 7); **P ≤ 0.01. (B) HRMEC apoptosis was studied under normoxic and hypoxic conditions in the presence of vehicle control (PBS) or 100 ng/mL NGF by performing cleaved caspase-3 staining, as described in Section 2. HRMEC apoptosis is shown relative to control; apoptosis in the presence of PBS control was set as 1. Data are presented as mean ± SEM and are from one experiment performed in triplicate; similar results were observed in at least 3 separate experiments. *P ≤ 0.05. (C) Expression of cleaved PARP was studied under normoxic and hypoxic conditions in the presence of vehicle control (PBS) or 100 ng/mL NGF, as described in Section 2. Expression of cleaved PARP is shown relative to control; expression of cleaved PARP in the presence of PBS control was set as 1 in each experiment. Data are presented as mean ± SEM (n = 4); *P ≤ 0.05. (D) HRMEC were exposed to normoxia or hypoxia in the presence of PBS or 100 ng/mL NGF and endothelial proliferation was assessed by BrdU incorporation, as described in Section 2. Proliferation is shown relative to control; proliferation in the presence of PBS was set as 1 in each experiment. Data are presented as mean ± SEM (n = 5); n.s.: not significant normoxia (21% O 2 ) or hypoxia (1% O 2 ) for 5 hours, followed by medium change to plain EGM with 100 ng/mL NGF or vehicle control and incubation for further 20 hours in normoxia or hypoxia. For the inhibition of the AKT pathway, Wortmannin (10 ng/mL; R&D Systems) was added in plain EGM 1 hour before the aforementioned

| Statistical analysis
All values are presented as the mean ± SEM. Statistical analysis was performed by Mann-Whitney U test or Student's t test with P ≤ 0.05 as a significance level using GraphPad Prism 6.0 Software (GraphPad Software, CA, USA).

| NGF increased pathological neovascularization in mice subjected to ROP model by inhibiting EC apoptosis
We initially investigated the effect of NGF on retinal neovascularization in mice subjected to the ROP model. After exposure to hyperoxia (75% O 2 ) from p7 to p12, pups were returned to normoxia.

| NGF reduced apoptosis in HRMEC under hypoxic conditions
Our in vivo findings prompted us to investigate how NGF affects cell growth and survival of retinal EC under different oxygen concentrations (21% and 1% O 2 ) in vitro. As an appropriate model, we chose HRMEC, EC from human retina previously shown to express TrkA. 37 We initially examined the effect of NGF on HRMEC apoptosis induced by starvation under normoxic or hypoxic conditions. 38,39 Cells were treated with NGF or vehicle control and then they were stained for cleaved caspase-3. Under hypoxic conditions, NGF significantly reduced the apoptosis of HRMEC, as compared to control treated cells, while it had no effect on apoptosis of cells incubated under normoxic conditions ( Figure 2B). The anti-apoptotic effect of NGF was further confirmed by the detection of cleaved PARP, which was significantly reduced by NGF treatment solely under hypoxic conditions ( Figure 2C).
Next, we tested whether NGF affects HRMEC cell proliferation.
To this end, cells were treated with NGF under normoxic (21% O 2 ) or hypoxic (1% O 2 ) conditions and proliferation was assessed by BrdU incorporation. NGF did not influence cell proliferation of HRMEC in normoxia or in hypoxia ( Figure 2D). This result is in accordance with previous studies, which also showed no effect of NGF on retinal EC proliferation. 40

| Signaling pathways involved in the pro-survival effect of NGF in endothelial cells under hypoxia
Next, we sought to reveal which signaling pathways mediate the anti-apoptotic effect of NGF. We exposed HRMEC to hypoxia, treated them with NGF for different time intervals following previously reported experimental settings, 26,39,44 and examined the activation of ERK and AKT pathways. NGF induces ERK and AKT phosphorylation in neuronal and EC within 60 minutes after stimulation. 26,39 NGF stimulated ERK phosphorylation at 15 minutes ( Figure 5A) and AKT phosphorylation at 30 minutes ( Figure 5B). Furthermore, AKT inhibition by Wortmannin abolished the anti-apoptotic effect of NGF on HRMECs under hypoxic conditions ( Figure 5C). These findings suggest that the AKT pathway is a key mediator of the anti-apoptotic effect of NGF in EC.
F I G U R E 5 AKT mediates the anti-apoptotic effect of NGF. (A, B) HRMEC were cultured under hypoxic conditions, treated without or with 100 ng/mL NGF and analysed at different time points, as indicated. (A) Representative cropped blot images (left) of phosphoERK, total ERK and Vinculin and the densitometric quantification of the ratio phosphoERK/ERK (right) are shown. Data are shown as relative to control (0 min); in each experiment the ratio phosphoERK/ERK of the control treatment was set as 1. Data are presented as mean ± SEM (n = 4); *P ≤ 0.05. (B) Representative cropped blot images (left) of phosphoAKT, total AKT and Vinculin and the densitometric quantification of the ratio phosphoAKT/AKT (right) are shown. Data are shown as relative to control (0 min); in each experiment the ratio phosphoAKT/AKT of the control treatment was set as 1. Data are presented as mean ± SEM (n = 6); **P ≤ 0.01. (C) HRMEC cells were pretreated with Wortmannin (an inhibitor of PI3K/AKT) or vehicle control (DMSO) followed by treatment without or with NGF under hypoxic conditions, as described in Section 2. Apoptosis of HRMEC was assessed by immunofluorescence staining for cleaved caspase-3. Apoptosis is shown relative to control; apoptosis in control (DMSO)-treated cells was set as 1. Data (mean ± SEM) are from one experiment performed in six replicates; similar results were observed in 3 additional experiments; **P ≤ 0.01

| DISCUSSION
Pathological neovascularization is an important feature of ROP and DR, which are leading causes of blindness in infants and adults respectively. 45,46 The role of growth factors, such as VEGF, Insulin Growth Factor or Erythropoietin, in pathological retinal vascularization has been extensively described. 47 NGF is expressed in the developing and adult retina and is produced by retinal ganglion cells, bipolar cells, glial cells and retinal pigment epithelial cells. 48,49 Furthermore, retinas of pups subjected to the ROP model display increased NGF expression. 24 The enhanced NGF expression could be a result of increased inflammation in the diseased retinas, 50 as NGF expression is shown to be upregulated under inflammatory conditions. 4,15 In turn, TrkA is expressed in the outer and inner segments of the retina, in photoreceptors, bipolar cells, Müller cells, amacrine cells, ganglion cells and astrocytes, 49 as well as in retinal EC (HRMEC). 37 Experimental evidence from different cell systems and animal models suggested that NGF promotes angiogenesis. 17,19,25 However, little was known about its role in hypoxia-trig- Mechanistically we showed here that the pro-angiogenic effect of NGF is associated with reduced apoptosis rather than with enhanced proliferation of retinal EC. Furthermore, in vitro studies in HRMEC revealed that the anti-apoptotic effect of NGF was only observed in hypoxia and not in normoxia. Although NGF was previously shown to induce angiogenesis-related functions in retinal EC under normoxic conditions, 24,41 we found here that the anti-apoptotic effect of NGF on EC is restricted to hypoxic conditions. Why NGF exhibits its anti-apoptotic effect in retinal EC primarily under hypoxia is still unclear. It was previously shown that the NGF receptor p75 undergoes oxygen-dependent cleavage, which subsequently mediates HIF-1α stabilization. 53 Whether p75 cleavage could regulate the anti-apoptotic effect of NGF is not known and requires further investigation. Furthermore, it is not known whether p75 cleavage could affect its association with TrkA or binding of NGF to TrkA/p75 heterodimers and whether all these functions could collectively contribute to the anti-apoptotic effect of NGF under hypoxia.
These aspects require further investigation.
In addition, we identified AKT and ERK activation to mediate the anti-apoptotic effect of NGF in HRMEC. AKT and ERK activate the transcription factor CREB, which induces BCL2 expression. 16,54,55 Consistently, we found that NGF treatment increased BCL2 expression, while it decreased BAX expression under hypoxic conditions. In keeping with these findings, NGF preserved the mitochondrial membrane potential under pro-apoptotic conditions (starvation) in combination with hypoxia.
Hence, our findings reveal a direct pro-survival effect of NGF in retinal EC under hypoxic conditions, which thereby contributes to elevated pathological neovascularization in retinas of mice subjected to the ROP model. The therapeutic engagement of neuroprotective agents, such as NGF, in the context of retinopathies has been recently considered. 4,5 Our findings however suggest that caution is needed in this scenario, as the pro-angiogenic effects of NGF may act detrimentally in the course of proliferative retinopathies like DR, by worsening pathological neovascularization. Therefore, the potential administration of NGF as a neuroprotective agent in the context of DR should be monitored very closely.

ACKNOWLEDG EMENTS
Supported by grants from the Deutsche Forschungsgemeinschaft (AL1686/3-1 and AL1686/2-2 to V.I.A.). We would like to acknowledge the core facility cellular imaging (CFCI) of the Medical Faculty of the TU Dresden for support with microscopy.

CONFLI CT OF INTEREST
None.