Neutralization of Schwann Cell-Secreted VEGF Is Protective to In Vitro and In Vivo Experimental Diabetic Neuropathy

The pathogenetic role of vascular endothelial growth factor (VEGF) in long-term retinal and kidney complications of diabetes has been demonstrated. Conversely, little is known in diabetic neuropathy. We examined the modulation of VEGF pathway at mRNA and protein level on dorsal root ganglion (DRG) neurons and Schwann cells (SC) induced by hyperglycaemia. Moreover, we studied the effects of VEGF neutralization on hyperglycemic DRG neurons and streptozotocin-induced diabetic neuropathy. Our findings demonstrated that DRG neurons were not affected by the direct exposition to hyperglycaemia, whereas showed an impairment of neurite outgrowth ability when exposed to the medium of SC cultured in hyperglycaemia. This was mediated by an altered regulation of VEGF and FLT-1 receptors. Hyperglycaemia increased VEGF and FLT-1 mRNA without changing their intracellular protein levels in DRG neurons, decreased intracellular and secreted protein levels without changing mRNA level in SC, while reduced the expression of the soluble receptor sFLT-1 both in DRG neurons and SC. Bevacizumab, a molecule that inhibits VEGF activity preventing the interaction with its receptors, restored neurite outgrowth and normalized FLT-1 mRNA and protein levels in co-cultures. In diabetic rats, it both prevented and restored nerve conduction velocity and nociceptive thresholds. We demonstrated that hyperglycaemia early affected neurite outgrowth through the impairment of SC-derived VEGF/FLT-1 signaling and that the neutralization of SC-secreted VEGF was protective both in vitro and in vivo models of diabetic neuropathy.


Introduction
Neuropathy is a chronic complication of both type 1 and type 2 diabetes that severely affects patients' quality of life and increases morbidity and mortality [1]. Once established, diabetic axonal damage fails to recover due to a number of events, including the loss of innervated targets and the chronic denervation of Schwann cells (SC) [2]. Different mechanisms have been claimed as critical in the pathogenesis of diabetic neuropathy (DN), including abnormal metabolic and neurovascular pathways, growth factor deficiency, and extracellular matrix remodeling [3]. Nevertheless, hyperglycaemia remains the most important trigger for the development of DN and its control is crucial for the course of the disease [3,4]. The complex relationship between axons and SC in nerve degeneration and regeneration [5] likely plays a critical role also in DN. Previous studies showed that hyperglycaemia can directly affect SC inducing in vitro apoptosis [6], altering the secretion of growth factors [7,8] and interfering with proliferation and migration abilities [9], thus suggesting an effect on this cell type. However, little is known on how hyperglycaemia interferes with the supporting role of SC on axonal growth in cultured dorsal root ganglion (DRG) neurons.
Here we describe that SC mediate the impairment of neurite outgrowth caused by hyperglycaemia through increased secretion of vascular endothelial growth factor (VEGF) and altered fmsrelated tyrosine kinase 1 (FLT-1) receptor signaling, and that bevacizumab, a molecule that inhibits VEGF activity preventing the interaction to its receptors, in vitro prevented axonal outgrowth failure, and in vivo both rescued and restored in a dose-dependent fashion DN in rats.

Animal Experimentation
The Statement of Compliance (Assurance) with Standards for Humane Care and Use of Laboratory Animals has been reviewed (10/28/2008) and approved by the National Institutes of Health-Office for Protection from Research Risks (5023-01, expiration 10/31/2013). The IRCCS Foundation ''Carlo Besta'' Neurological Institute adheres to the principles set out in the following laws, regulations, and policies governing the care and use of laboratory animals: Italian Legislative Decree 116 of Jan. 27,1992 Authorization 169/94-A issued Dec. 19,1994

Cell culture
Primary DRG culture were freshly isolated from embryonic age day-15 Sprague-Dawley rats. Dissected embryonic DRG were enzymatically dissociated with 0.25% trypsin in L-15 medium as previously described [10]. Cells were plated in 24-well plates on collagen-coated glass coverslips, pretreated with poly-D-lysine (Sigma-Aldrich, St. Louis, MO). These cultures contain mainly sensory neurons and SC. DRG neuron monocultures were obtained after exposition to ARA-C (10 mM) for 72 hours [11] and maintained in Neurobasal medium (Gibco Invitrogen, Grand Island, NY) containing 25 mM glucose, supplemented with 1xB27 without antioxidants, penicillin (1 U/L), streptomycin (1 U/L), and nerve growth factor (10 ng/ml). All experiments with DRG culture were performed 7 day after initial plating. SC monocultures were obtained from 2 day-old rat sciatic nerves, purified using a modified Brockes' method [12], and maintained in DMEM 10% FBS, neuregulin (20 ng/ml) and forskolin (2 mM). Twenty-four hours before the experiments, medium was changed to neuronal medium. Hyperglycemic condition was obtained (where not otherwise specified) adding 20 mM glucose to achieve a final concentration of 45 mM [13,14]. DRG neuron monocultures were exposed to SC-conditioned medium collected from control and hyperglycemic (45 mM for 24 hours) cultures. Co-cultures and monocultures were exposed to cisplatin or paclitaxel (10 mg/ ml and 250 ng/ml for 24 hours, respectively; Sigma-Aldrich, St.Louis, MO) as positive control [15,16]. Hyperglycemic and control cultures were exposed to bevacizumab (Avastin; Genentech Inc., Roche Group, San Francisco USA), 0.1 and 0.25 mg/ ml for 24 hours to evaluate the protective effect [17].

Assessment of neuronal apoptosis
Cells were fixed and stained with in situ DNA nick labeling (DeadEnd Fluorometric TUNEL System, Promega, Madison, WI), co-stained with DAPI and anti-BIII tubulin antibody (TuJ1, Berkley Antibody Company, Richmond, CA) and anti-glial fibrillary acidic protein (GFAP, Dako, Glostrup, Denmark) antibodies for neuron and SC identification, respectively. Apoptotic cells were counted using a fluorescence microscope. Apoptosis was also assessed by flow-cytometry using Annexin V/ PI assay (Immunostep, Salamanca, Spain) and processed according to manufacturer's instructions. Early and late apoptosis were evaluated on fluorescence 2 (for propidium iodide) versus fluorescence 1 (for annexin) plots. The percentage of cells stained by annexin V alone was recorded as early apoptosis, whereas the percentage of cells stained by both annexin V and propidium iodide was recorded as late apoptosis.

Assessment of mitochondrial membrane potential changes
Mitochondrial membrane potential was measured using the Jaggregate forming lipophilic cation (JC-1; Molecular Probes, Eugene, OR) [18,19]. Dissociated DRG cells were exposed to hyperglycaemia, cisplatin or paclitaxel. After 24 hours, cells were washed and incubated with JC-1 (2.5 mg/ml) for 30 min according to manufacturer's protocol. The fluorescence emission pattern was analyzed by flow-cytometry using two dimensional green versus red fluorescence plots. The red-to-green ratio allows comparative measurements of membrane potential between cell populations.

Assessment of axonal outgrowth
After 24 hours exposure to treatments, cells were fixed in 4% paraformaldehyde and stained with fluorescent anti-TuJ1 antibody. Neurite outgrowth was assessed measuring the longest neurite on each of at least 100 randomly selected neurons per condition, using an image analysis system on fluorescence microscope (Image Pro-Plus, Media Cybernetics, Silver Spring, MD) [20][21][22]. Each condition was assessed in duplicate for at least three times using cultures prepared on separate days. ELISA assay of secreted VEGF and sFLt-1 protein Enzyme-linked immunosorbent assay (ELISA) were performed with commercial ELISA kit for VEGF (R&D Systems, Minneapolis, MN) and sFLT-1 (Novatein Biosciences, Cambridge, MA). Conditioned medium was collected from wells 24 hours after treatment. Assays were performed in duplicate and values were compared with standards provided by the kits.

Real-time PCR assay
Total RNA was isolated 24 hours after treatment with VEGF, glucose and/or bevacizumab from cultured DRG cells or SC using Trizol (Invitrogen, San Giuliano Milanese, Italy) according to manufacturer's instructions. Precipitated RNA was dissolved in RNase-free water and quantified; aliquots were prepared for further analysis. RNA was analyzed using TaqMan qRT-PCR (quantitative real-time PCR) instrument (CFX384 real time system, Bio-Rad Laboratories, Segrate, Italy) by iScriptTM onestepRT-PCR kit for probes (Bio-Rad Laboratories, Segrate, Italy). Specific TaqMan probes (Life Techonologies, Monza, Italy) were VEGFa (Rn01511601_m1) and VEGFR-1 (Flt1) (Rn00570815_m1). Samples were run in 98 wells in triplicate as multiplexed reactions with a normalizing internal control (18s). Data were confirmed by assay of two additional house-keeping genes: actin (Rn00667869_m1) and cyclophilin A (Rn00690933_m1). Comparative cycle threshold method was used to calculate the relative target gene expression. Each condition was assessed in duplicate for at least three times.

Western blot
Total proteins from treated cell cultures were extracted using Trizol according to manufacturer's protocol. Protein quantification was performed by standard Bicinchoninic acid (BCA, Pierce, Rockford, USA) method; 30 mg of protein per lane were applied, separated by electrophoresis on 10% polyacrylamide gels and transferred onto methanol treated PVDF membranes. Membranes were saturated with 5% fat-free dried milk in Tween-triphosphate buffer solution and incubated overnight at 4uC with primary antibodies (anti-VEGF antibody, rabbit policlonal, Abcam, 1:600; anti-VEGF Receptor 1 antibody, rabbit polyclonal, Abcam 1:200; anti-alpha tubulin, mouse monoclonal, 1:800) diluted in the same buffer, washed and exposed to peroxidise conjugated secondary antibodies (donkey anti-rabbit or anti-mouse IgG 1:10000, GE Healthcare Europe, Milan, Italy). Specific signals were revealed with the chemiluminescence detection kit (ECL, GE Healthcare Europe, Milan, Italy). Semiquantitative densitometric analysis was performed by scanning the bands with image-analysis software (Image-J). Results were normalized for TuJ1.

Nociceptive thresholds
Thermal nociceptive threshold to radiant heat was quantified measuring withdrawal latency on hot plate [23], defined as the time laps between placement and hindpaw withdrawal and licking. Each animal was tested twice separated by a 30-minute rest interval. Mechanical nociceptive threshold was quantified by the Randall-Selitto paw withdrawal test [23] using an analgesy meter (Ugo Basile, Comerio, Italy) that generates a linearly increasing mechanical force. Result represents the maximal pressure (grams) tolerated by the animal. Rats were accustomed to the devices 3 days before performing the tests. At each time point, animals underwent three trials and values were averaged.

Electrophysiological techniques
Antidromic sensory tail-nerve conduction velocity (NCV) was assessed by Myto EBNeuro electromyograph (EBNeuro, Firenze, Italy). The latency of potential was recorded in the two sites after nerve stimulation (stimulus duration, 100 msec; filter, 1 Hz to 5 MHz) was determined (peak to peak), and NCV calculated accordingly. Data were collected with a MP100 acquisition system (Biopac Systems, Santa Barbara, CA). All recordings were performed under standard conditions at controlled temperature (room and animals). Core temperature was maintained at 37uC by using heating pads and lamps. Only diabetic rats with urine glucose .15 mM two days after STZ injection were used. Rats were divided into four groups (n = 8) in both the prevention and therapeutic schedules (controls, STZ, STZ+bevacizumab 10 mg/kg, STZ+bevacizumab 20 mg/kg). Control animals were weight, matched and given sodium citrate buffer without STZ. Prevention protocol: diabetic rats were randomized to receive i.p. bevacizumab at either 10 or 20 mg/ kg body weight at week 1 soon after diabetic induction, and at week 4. NCV and nociceptive thresholds were measured at week 8 (4 weeks after last bevacizumab treatment). Therapeutic protocol: diabetic rats underwent NCV and nociceptive threshold assessment at week 7. Therefore, they were randomized to receive bevacizumab at either 10 or 20 mg/kg body weight at week 8 and 12. NCV and nociceptive thresholds were measured at week 16 (4 weeks after last bevacizumab treatment). Body weight was measured weekly in both groups. In the prevention protocol, bevacizumab was administered at week 1 (after confirmation of diabetes), and week 4; rats were sacrificed at week 8. In the therapeutic protocol, bevacizumab was administered at week 8 and 12, and rats were sacrificed at week 16. Nerve conduction velocity (NCV) and nociceptive threshold assessment were measured at the indicated times. doi:10.1371/journal.pone.0108403.g001

Statistic analysis
Data are presented as mean6standard deviation (SD). Differences were analyzed by Student's unpaired t test or 1-way ANOVA. P,0.05 was considered statistically significant.  Hyperglycaemia did not cause early loss of mitochondrial membrane potential differential Mitochondrial potential was assessed in DRG-co-cultures 24 hours after glucose (45 mM) or toxic compound exposure (paclitaxel 250 ng/ml or cisplatin 10 ug/ml). Cisplatin exposure caused an intense decrease of red/green fluorescence ratio (% of cells in the lower-right quadrant 57.40612.82 vs ctrl 10.6563.95, p,0.05) thus revealing mitochondrial depolarization. Conversely, high glucose treated cultures did not show any significant change in fluorescence plot (% of cells in the lower-right quadrant 10.8260.77 vs ctrl 10.6563.95, p.0.05). These findings indicated that hyperglycaemia did not affect mitochondrial membrane potential in our experimental condition (Fig. 3A, B) Fig. 4B). When DRG neuron monocultures were exposed to the medium of SC cultured in hyperglycaemia, neurite outgrowth decreased (axonal length % difference 81.1361.2 p,0.0001 vs ctrl 100) by similar extent to that observed in DRG co-cultures exposed to hyperglycaemia (axonal length % difference, ctrl 100; glucose 45 mM 109.3610.4; Fig. 4C). These findings suggested that SC secreted a molecule able to exert a toxic effect on DRG neurons. To investigate this hypothesis, we performed a cytokine profile array in the medium of SC monoculture exposed and not exposed to hyperglycaemia. We found higher level of VEGF in the medium of SC monoculture exposed to hyperglycaemia than in control medium (pixel density intensity; ctrl 0.69260.005; hg 45 mM 0.94760.050 p,0.05; Fig. 4D and E). No difference was found in the levels of all the others cytokines and chemokines analyzed (see Method section for the list of cytokines and chemokines analyzed). This suggested that DRG neurons neurite length decrease may depend on increased level of SC-derived VEGF.

Hyperglycaemia induced post-translational VEGF regulation in SC
We performed ELISA assay for VEGF level in the medium of DRG neuron and SC. Basal VEGF concentration was 1464 pg/ ml in DRG neuron monoculture medium, 3-fold lower than in SC (4269 pg/ml) and co-culture medium (4866 pg/ml). After 24 hour exposure to 45 mM glucose, VEGF concentration significantly increased in SC monoculture (45 mM glucose 5469 pg/ml vs ctrl 4269 pg/ml, p,0.05) and DRG co-culture (45 mM glucose 5665 pg/ml vs ctrl 4866 pg/ml, p,0.05) medium, whereas it did not significantly change in DRG neuron monoculture medium (45 mM glucose 1364 pg/ml vs ctrl 1464 pg/ml; Fig. 5A). We next performed VEGF mRNA quantification by real-time PCR and VEGF protein quantification by Western blot. After 24 hour exposure to 45 mM glucose, VEGF mRNA expression did not significantly change in SC (relative mRNA expression 144631 vs ctrl 100), whereas intracellular VEGF protein level decreased (relative protein expression 88.4613.8 p,0.05 vs ctrl 100). According to the ELISA assay, DRG neuron monocultures in basal condition expressed very low VEGF mRNA level compared to SC (0.00460.001 vs 0.02060.01; p,0.01). VEGF mRNA level was significantly increased in DRG neuron monocultures by 24 hour exposure to 45 mM glucose (relative mRNA expression normalized to 18s: 194637 vs ctrl 100; relative mRNA expression normalized to actin 188624 vs ctrl 100; relative mRNA expression normalized to cyclophilin A 172629 vs ctrl 100, p, 0.05), whereas intracellular protein level did not change (relative protein expression 11363 vs ctrl 100) (Fig. 5B-D). To test the hypothesis that VEGF decreased neurite outgrowth, DRG neurons were treated for 24 hours with VEGF in a wide range concentration (from 10 pg/ml to 100 ng/ml  Fig. 5E), suggesting that VEGF had a key role in the decrease of neurite length found in DRG co-cultures exposed to hyperglycaemia.

Bevacizumab prevented hyperglycaemia-mediated neurite outgrowth impairment and normalized FLT-1
In DRG co-cultures exposed to hyperglycaemia, bevacizumab

Bevacizumab improved nerve functions in STZ-diabetic neuropathy rats
Bevacizumab did not have significant effects on weight and nonfasted plasma glucose in diabetic rats at both the concentrations administered. In the preventive protocol, bevacizumab prevented the increase of hindpaw thermal nociceptive latency (seconds ctrl

Discussion
In patients, DN typically presents as a dying-back axonopathy, suggesting that hyperglycaemia targets axons rather than DRG neurons [3] which primary impairment would lead to distinct nonlength-dependent clinical and neurophysiological abnormalities [24]. DRG neurons are particularly vulnerable to toxic circulating agents being supplied by fenestrated capillaries instead of the tight blood-brain-barrier [25,26]. Furthermore, they have long axons rich in mitochondria and are particularly susceptible to any interference in energy metabolism, oxidative stress, and axonal transport [27]. SC are involved in endogenous neuroprotective pathways [28] and the reciprocal interaction with axons is mandatory for the normal functions of nerves. Previous works showed that hyperglycaemia can impair the secretion of SC- derived cytokines and growth factors, and affect axonal growth [8].
We used an in vitro experimental approach using co-cultures and monocultures of DRG neurons and SC to investigate the effects of hyperglycaemia on each cell type and their relationship. The use of embryos allows a high content of cells with a higher efficiency in isolating neurons in comparison to post-natal animals and they have been shown to be an adequate and predictive model for studying neuropathies [29]. Even if embryonic sensory neurons have phenotypic differences from adult sensory neurons needing an higher basal concentration of glucose (25 mM) than adult neurons for survival and being dependent on neurotrophic factors [30] for the first week in culture, it has been demonstrated that aged embryonic DRG neurons cultured exhibit properties similar to adult DRG neurons [31]. The equivalent of a 1.8-fold glucose level above control value used was similar to the 1.5-fold increase commonly found in diabetic patients [32]. In this setting, confirming previous works [25], we found that hyperglycaemia affected the ability of neurite outgrowth, whereas it did not induce early or late DRG neuron apoptosis. Although neurite outgrowth is a model of axon elongation rather than degeneration, our findings support what has been clearly established in animal models and patients with DN [3] and points to axons as the primary target of hyperglycaemia-induced toxicity.
Our study demonstrated the pathogenetic role of SC-derived VEGF/FLT-1 signaling, whose normalization after bevacizumab treatment was protective in the in vitro model. We showed that the exposition of DRG neurons monoculture to hyperglycaemia did not affect neurite outgrowth, which did not differ from control monocultures. Conversely, neurite outgrowth significant decreased when DRG neuron monocultures were exposed to the medium of SC cultured in hyperglycaemia. This effect was mediated by the marked increase of VEGF in the medium of hyperglycaemiaconditioned SC monoculture, as confirmed by the dose-dependent impairment of neurite outgrowth after exposition of DRG coculture to VEGF. Previous studies showed that hyperglycaemia directly stimulates the secretion of VEGF in retinal Müller cells [33] and proximal tubular cells [34]. Moreover it has been observed that VEGF protein level increased in DRG neurons and sciatic nerve axons from chronic STZ diabetic rats [35]. These findings strengthened the hypothesis of a key role of VEGF also in the pathogenesis of DN, like already demonstrated in diabetic retinopathy and nephropathy [36,37].
We demonstrated that hyperglycaemia inversely modified FLT-1 protein level in DRG neuron monocultures (where it increased) and SC (where it decreased). FLT-1 is a cell-surface receptor for VEGF and may function as negative regulator limiting the amount of free VEGF and preventing its binding to VEGF receptor-2 (KDR), the best characterized receptor and known to mediate most VEGF cellular responses [38]. We also found that sFLT-1 was decreased. This soluble receptor lacks one transmembrane domain and may function as a decoy for VEGF. We speculated that in hyperglycaemia VEGF overruled sFLT-1 whose scavenger activity could not limit VEGF increase and its toxic effects.
We showed that bevacizumab, a recombinant humanized monoclonal IgG1 antibody that binds VEGF and inhibits its biologic activity preventing the interaction to its receptors, was protective both in vitro and in vivo models of DN. Indeed, it reduced the level of free VEGF in the medium of DRG co-cultures exposed to hyperglycaemia and protected from the impairment of neurite outgrowth in a dose-dependent fashion. This was associated with the normalization of FLT-1 signaling between neurons and SC. Exposition of hyperglycaemia-conditioned DRG neuron monocultures to bevacizumab normalized FLT-1 mRNA with no change at the protein level. Conversely, exposition of hyperglycaemia-conditioned SC monocultures to bevacizumab increased FLT1 protein and reduced FLT-1 mRNA.
Finally, we demonstrated that bevacizumab both protected and reversed neuropathy in STZ rats, confirming the neuroprotective effects of our in vitro studies. Indeed, preventive and therapeutic protocols of bevacizumab administration counteracted in a dosedependent fashion the pathological changes in thermal and mechanical thresholds, and in NCV which are hallmarks of diabetic neuropathy. Modulation of the VEGF/FLT1 signalling axis in vivo have to be further investigate in order to attribute efficacy of bevacizumab to a specific mechanism. Few and contradictory data are available on the role of VEGF in DN. Some works showed neuroprotective effects of VEGF on sensory and motor neurons [39][40][41], whereas others provided convincing evidence of direct toxic effects on nerves [42] which are reversed by bevacizumab [43]. A recent study reported significantly higher levels of serum VEGF in diabetic patients with neuropathy compared to those without neuropathy [44]. In the STZ model of experimental diabetes it has been reported that VEGF expression increased in DRG and sciatic nerves and that insulin and/or nerve growth factor administration could prevent it [35].

Conclusions
We provided evidence of the pathogenetic role of VEGF in experimental DN. Our findings are in keeping with previous studies showing the direct involvement of VEGF in other longterm complications of diabetes, as demonstrated by its detrimental effects on retina [36] and kidney [45], the elevated serum levels [37,46] and, most importantly, the beneficial effects of anti-VEGF treatment in patients with diabetic retinopathy and nephropathy [47][48][49][50]. We suggest that appropriate clinical trials should be designed to test the efficacy of anti-VEGF drugs also in DN patients.