Diabetes‐induced microvascular complications at the level of the spinal cord: a contributing factor in diabetic neuropathic pain

Key points Diabetes is thought to induce neuropathic pain through activation of dorsal horn sensory neurons in the spinal cord. Here we explore the impact of hyperglycaemia on the blood supply supporting the spinal cord and chronic pain development. In streptozotocin‐induced diabetic rats, neuropathic pain is accompanied by a decline in microvascular integrity in the dorsal horn. Hyperglycaemia‐induced degeneration of the endothelium in the dorsal horn was associated with a loss in vascular endothelial growth factor (VEGF)‐A165b expression. VEGF‐A165b treatment prevented diabetic neuropathic pain and degeneration of the endothelium in the spinal cord. Using an endothelial‐specific VEGFR2 knockout transgenic mouse model, the loss of endothelial VEGFR2 signalling led to a decline in vascular integrity in the dorsal horn and the development of hyperalgesia in VEGFR2 knockout mice. This highlights that vascular degeneration in the spinal cord could be a previously unidentified factor in the development of diabetic neuropathic pain. Abstract Abnormalities of neurovascular interactions within the CNS of diabetic patients is associated with the onset of many neurological disease states. However, to date, the link between the neurovascular network within the spinal cord and regulation of nociception has not been investigated despite neuropathic pain being common in diabetes. We hypothesised that hyperglycaemia‐induced endothelial degeneration in the spinal cord, due to suppression of vascular endothelial growth factor (VEGF)‐A/VEGFR2 signalling, induces diabetic neuropathic pain. Nociceptive pain behaviour was investigated in a chemically induced model of type 1 diabetes (streptozotocin induced, insulin supplemented; either vehicle or VEGF‐A165b treated) and an inducible endothelial knockdown of VEGFR2 (tamoxifen induced). Diabetic animals developed mechanical allodynia and heat hyperalgesia. This was associated with a reduction in the number of blood vessels and reduction in Evans blue extravasation in the lumbar spinal cord of diabetic animals versus age‐matched controls. Endothelial markers occludin, CD31 and VE‐cadherin were downregulated in the spinal cord of the diabetic group versus controls, and there was a concurrent reduction of VEGF‐A165b expression. In diabetic animals, VEGF‐A165b treatment (biweekly i.p., 20 ng g−1) restored normal Evans blue extravasation and prevented vascular degeneration, diabetes‐induced central neuron activation and neuropathic pain. Inducible knockdown of VEGFR2 (tamoxifen treated Tie2CreERT2‐vegfr2flfl mice) led to a reduction in blood vessel network volume in the lumbar spinal cord and development of heat hyperalgesia. These findings indicate that hyperglycaemia leads to a reduction in the VEGF‐A/VEGFR2 signalling cascade, resulting in endothelial dysfunction in the spinal cord, which could be an undiscovered contributing factor to diabetic neuropathic pain.


Introduction
Diabetes mellitus leads to an array of health complications that can cause significant morbidity. In people with diabetes, neuropathic pain is common (Tesfaye et al. 2013), characterised by enhanced responses to noxious (painful) stimuli (hyperalgesia) as well as to innocuous stimuli (allodynia). These alterations in pain perception are due to maladaptive changes in the sensory neuronal circuitry. The plasticity of the nociceptive neuronal systems, both peripheral (Reichling & Levine, 2009) and central (Latremoliere, 2009), means that they can respond to disease and/or treatment such as in diabetes (Chen & Levine, 2001;Morgado et al. 2010;Tan et al. 2012). These responses lead to neuronal sensitisation, and in diabetes, chronic pain development. The peripheral sensory nerves are well known to be affected by hyperglycaemia, including degeneration of intra-epidermal nerve fibre innervation patterns  and hyper-excitability (Chen & Levine, 2003). However, pain management in people with diabetes only provides often partial pain relief (Tesfaye et al. 2011(Tesfaye et al. , 2013. There are now focused efforts to investigate how changes in nociceptive processing in the CNS, in particular the spinal cord, are altered in diabetic neuropathic pain (Biessels et al. 2014;Tesfaye et al. 2016). Studies have identified that in diabetic rodents, sensory neurons within the spinal cord elicit exaggerated responses to sensory stimulation (Morgado et al. 2010;Tan et al. 2012). Despite this evidence for the involvement of spinal cord changes in the pathogenesis of diabetic neuropathic pain, there are few investigations into those mechanisms that may underlie the development of central sensitisation in the spinal cord in diabetes (Tan et al. 2012;Lee-Kubli & Calcutt, 2014).
An important component of the nervous system is the supporting blood vessel network. A compromised vascular system is integral to the development of multiple neurological diseases (e.g. stroke, Alzheimer's disease) (Tiehuis et al. 2008;Vandal et al. 2014;Winkler et al. 2015). Hyperglycaemia induces extensive vascular remodelling in the nervous system (Taylor et al. 2015;Hardigan et al. 2016) as well as direct glucose toxicity on sensory neurons (Chowdhury et al. 2014;Hulse et al. 2015), with both of these contributing to neurological complications including increased susceptibility of people with diabetes to cognitive decline, stroke and peripheral ischaemic neuropathies (motor, sensory and autonomic) (Said, 2007;Hardigan et al. 2016). Vascular endothelial growth factor-A (VEGF-A) is strongly implicated in diabetic vascular disease, including driving aberrant vessel growth and increased permeability in diabetic retinopathy (Cai & Boulton, 2002) and is therefore a prime target for diabetic retinopathy treatment (Gupta et al. 2013). The VEGF-A gene gives rise to a variety of VEGF-A splice variants, differing in length and terminal sequence, leading to contrasting functions (Harper & Bates, 2008). The archetypal proangiogenic isoform is VEGF-A 165 a and is typically associated with vascular remodelling such as in diabetic retinopathy and cancer (Perrin et al. 2005). The VEGF-A 165 b isoform is predominantly found in normal tissues, areas with reduced angiogenesis (Pritchard-Jones et al. 2007) and pathologies where angiogenesis is impaired (e.g. systemic sclerosis and peripheral arterial disease) (Manetti et al. 2011;Kikuchi et al. 2014;Ngo et al. 2014). It is known to compete with VEGF-A 165 a for VEGFR2 binding (Bates et al. 2002;Cébe Suarez et al. 2006) and can independently act through VEGFR2, resulting in cytoprotective effects (Beazley-Long et al. 2013). For example, in diabetic nephropathy VEGF-A 165 b treatment leads to a rescue in endothelial cell survival and return to normal kidney function (Oltean et al. 2014), and we have previously found that indicators of altered blood vascular integrity, enhanced in the peripheral nervous system and the dorsal root ganglia (DRG), can be reversed by VEGF-A 165 b, resulting in amelioration of pain behaviours in rats .
Enhanced sensitivity in central nociceptive networks [spinal cord (Tan et al. 2012;Lee-Kubli & Calcutt, 2014), brain (Silva et al. 2013)] have long been attributed to neuroplastic changes. Despite extensive evidence that the cerebral vasculature is altered in diabetic rodents, there has been limited investigation into the neurovascular interactions in the spinal cord, particularly with reference to nociception (Costigan et al. 2009;Beggs et al. 2010).
Here we hypothesised that a decline (endothelial cell loss) in the vascular system within the spinal cord could contribute to the onset of diabetic neuropathic pain. Using an in vivo rat model of type 1 diabetes, and an inducible VEGF receptor-2 (VEGFR2) knockdown transgenic mouse, neurovascular disruption in the spinal cord was associated with changes in neuropathic pain. Administration of the VEGF-A 165 b isoform protected the endothelial component of the CNS, and prevented diabetic neuropathic pain.

Ethical approval, animals used and induction of diabetes
Twenty-four male (21 Evans blue and 3 S1 tissue collection) and 42 female (24 Evans blue and 18 S1 tissue collection) Sprague Dawley rats (ß250 g) were used in this study. Experiments were carried out in accordance with the institution's animal welfare committee (University of Nottingham), and conform to the principles and regulations as described by Grundy (2015). Procedures were carried out in accordance with the UK Home Office Animals (Scientific Procedures) Act 1986 and EU Directive 2010/63/EU after review by the local Animal Welfare and Ethics Review Board (University of Nottingham). Diabetes was induced by I.P. injection of streptozotocin (STZ; 50 mg kg −1 ) . Animals had ad libitum access to standard chow and were housed in groups (n > 2) under 12:12 h light-dark conditions. In male rats, experimental groups (1 week, no insulin supplementation) were naïve (sham injected n = 9) and STZ treated (n = 12). At the end of the study animal weight and blood glucose (>15 mmol) were: naïve = 339 ± 8.26 g and 6.8 ± 0.85 mmol; diabetics = 314 ± 11.86 g and 30.12 ± 0.95 mmol, respectively.

Nociceptive behaviour
Nociceptive behavioural experiments were carried out as previously described (Drake et al. 2014;Hulse et al. 2016) on 8-week-old diabetic animals and age matched sham controls (naïve). Mechanical withdrawal thresholds were measured using von Frey (vF) monofilaments  or a mechanical pincher (Drake et al. 2014;Hirschberg et al. 2017). A range of vF hairs were applied to the hind paw plantar surface (a maximum of 5 s or until paw withdrawal). A total of five vF applications were applied per weighted hair and force response curves were generated and withdrawal values were calculated as the weight at which withdrawal frequency = 50%. A mechanical pincher [equipped with strain gauges and calibrated to force (g)] (Drake et al. 2014) was applied to the hind paw until the animal withdrew to determine mechanical hyperalgesia. Raw data were acquired through a Neurolog power unit and a bridge amp module (Digitimer, Welwyn Garden City, UK), with digital acquisition via CED micro1401v3 and Spike2 v7 J Physiol 596.16 software (Cambridge Electronic Design, Cambridge, UK). Withdrawal to heat was determined using the Hargreaves test (Hargreaves et al. 1988). The experimenter was blinded to treatment.

Evans blue extravasation
In vivo vascular perfusion was evaluated using Evans blue dye, as previously described (Xu et al. 2001). Animals [age matched sham controls (naïve), week 1 and week 8 diabetic rats] were terminally anaesthetised (ketamine medetomidine I.V. 50 mg kg −1 ) and infused (via the external jugular vein) with Evans blue dye I.V. (Sigma-Aldrich, Poole, UK; 45 mg kg −1 ) at 120 mmHg pressure. Two minutes after infusion, 0.2 mL arterial blood was withdrawn, followed by subsequent 0.1 mL withdrawals every 15 min for 2 h. After 2 h, 0.2 mL blood was withdrawn followed by cardiac perfusion of 50 mL saline at 120 mmHg. Lumbar spinal cord and a single brain hemisphere were excised (whole spinal cord including ventral and dorsal horn) and weighed (wet weight). Tissue was dried at 70°C overnight and weighed (dry weight). Dried tissue was incubated in 0.15 mL formamide (Sigma Aldrich) at 70°C overnight. Blood samples were centrifuged (12,000 rpm, 45 min, 4°C), and the supernatant from tissue and blood samples were analysed at 620 nm. Evans blue extravasation was calculated as: solute flux (μg min -1 g -1 ) = Evans blue mass (μg)/tissue dry weight (g) divided by time (120 min).

Spinal cord endothelial cell culture
Whole lumbar spinal cords were dissected from adult male Sprague Dawley rats and dissociated (0.125% collagenase). Endothelial cells were extracted and cultured in endothelial media (M199 media, 60 μg mL -1 endothelial cell growth supplement and 50 μg mL -1 heparin). Endothelial cells were plated (1% gelatin coated) onto either six-well plates (for protein extraction) or 96-well plates (for cell viability studies). For cell viability assays, when 80% confluent endothelial cells were incubated for 24 h in either 5 mM glucose (normal), 50 mM mannitol (osmotic control) or 50 mM glucose (high glucose) ± 2.5 nM VEGF-A 165 b or vehicle. Cell death was determined with neutral red (Sigma-Aldrich).

Flow cytometry
Spleens were isolated from control (CTL) and endothelial cell knockout (ecKO) transgenic mouse experimental groups and placed in media (containing RPMI1640, penicillin/streptomycin, 10% FBS, 1% L-glutamine, 0.1% sodium pyruvate). Tissue was mechanically dissociated through a 40 μm cell strainer and washed through with media. Cell suspension was then centrifuged at 1500 rpm for 3 min. Supernatant was removed and cells were resuspended in 2 mL red cell lysis medium (Sigma), then left for 30 s. Following this 10 mL of medium was added and the cells were centrifuged as previously described. Cells were then fixed in 4% paraformaldehyde for 15 min at room temperature and then subsequently washed three times. Cells were permeabilised with 0.4% Triton X-100 in PBS for 15 min at room temperature. Cells were blocked in 1% FBS in PBS, mixed gently and incubated for 30 min at room temperature. Antibodies [CD11b-APC (1:100; BioLegend, San Diego, CA, USA) and F4/80-PE (1:100; Biolegend)] were added, left overnight in the fridge, and samples were analysed on a MoFlo analyser.

qPCR method
Total RNA was extracted from whole lumbar spinal cord tissue isolated from CTL and ecKO transgenic colony experimental groups using TRIzol reagent (Invitrogen). cDNA synthesis was carried using a PrimeScript RT reagent kit (TaKaRa, Shiga, Japan; RR037A) with a starting amount of 1 μg of RNA. The resulting cDNA was used for quantitative PCR using a LightCycler 480 SYBR Green I Mastermix (Roche, Indianapolis, IN, USA; 04707516001) following the manufacturer's instructions. CD31 and VE-cadherin primers were synthesised by Eurofins. Beta actin and VEGFR2 primers were synthesised by Sigma-Aldrich. Beta actin was used as a reference gene. The primer sequences were as follows (5 -3 ):

Statistical analysis
All data are represented as mean ± SEM unless stated otherwise and the experimenter was blinded where appropriate. Data were acquired/quantified using Microsoft Excel 2010 and Graphpad Prism 6, and Imaris (Bitplane) Spike2 v7 software (CED) was used to digitally acquire mechanical withdrawal thresholds from the pincher and for offline analysis. Immunofluorescence was quantified by obtaining 10 random non-sequential sections (Z stacks) per animal and a mean value was calculated per animal. Dorsal horn spinal cords (Lamina I-V) were imaged with the confocal microscope as described above, and vessels were identified through CD31 and IB 4 immunoreactivity. IB 4 signal fluorescence was vascular in the majority of the spinal cord with C-fibre projections also staining in the peripheral laminae (I and II), but vessels could also be clearly delineated here by CD31 staining as well as by morphology. Stained images were rendered on Imaris 8.11 imaging software. This allows for automated quantification of vessel diameter and blood vessel volume. Neuron number (cleaved caspase-3 and C-Fos quantification) was determined as per laminae of the dorsal horn and therefore expressed according to lamina I-V as previously characterised (Hsieh et al. 2015). Western blot densitometry was quantified using the ImageJ (https://imagej.nih.gov/ij/) gel quantification plugin. Paw mechanical withdrawal thresholds, the number of blood vessels in the dorsal horn of the spinal cord and spinal cord Evans blue extravasation were analysed using a Mann-Whitney test. Evans blue, IB 4 positive blood vessel volume and length, mechanical and heat nociceptive behavior, and western blot densitometry quantification were analysed using a Kruskal-Wallis and appropriate post hoc tests.

Results
Diabetes resulted in increased mechanical hyperalgesiaa reduction in withdrawal threshold to a noxious stimulus (hindpaw pinch) when compared to both before diabetes (P < 0.05) and to vehicle/age matched animals at week 8 after STZ injection (naïve; Fig. 1A, * P < 0.05). This was accompanied by a reduction in the number of blood vessels in the dorsal horn of the lumbar region of the spinal cord ( Fig. 1B-D, * P < 0.05) as well as a reduction in microvessel diameter (labelled with CD31 and IB 4 ) ( Fig. 1E-G). The VEGF-A family is a key regulator of angiogenic processes. Pan-VEGF-A expression was unaltered in the lumbar spinal cord of diabetic rats when compared to age/gender matched control animals ( Fig. 1H and I). However, VEGF-A xxx b expression was significantly reduced in the lumbar spinal cord of diabetic animals ( Fig. 1H and I, * P < 0.05). As hyperglycaemia affected spinal cord microvasculature and reduced VEGF-A 165 b expression, we investigated the direct cytoprotective actions of VEGF-A 165 b upon cultured spinal cord endothelial cells. Isolated spinal cord endothelial cells showed increased cell death when cultured in high (50 mM) versus low glucose (5 mM) conditions ( Fig. 2A). There was no effect of 50 mM mannitol, an osmotic control. VEGF-A 165 b treatment prevented high glucose-induced endothelial cell death ( Fig. 2A)  actions of VEGF-A 165 b on the spinal cord endothelium were investigated in vivo. The vascular network was well defined in naïve animals (Fig. 2B) and was reduced in number in diabetic + vehicle treated animals (Fig. 2C), but this was prevented by VEGF-A 165 b (Fig. 2D). Quantification indicated a reduction in endothelium (reduced total vascular volume) within the dorsal horn of the lumbar region of the spinal cord, which was rescued by VEGF-A 165 b (Fig. 2E). There was also an overall reduction in dorsal horn vessel diameter, again prevented by VEGF-A 165 b (Fig. 2F). There was a non-significant increase in vessel diameter in the diabetic + VEGF-A 165 b treated group when compared with the naïve group (Fig. 2F). Analysis of the frequency distribution of the vessels by size indicated that the reduction in size in the diabetic animals was due to a reduced number of the 'larger microvessels' (8-12 μm) rather than a reduction in the number of small vessels (<8 μm) (Fig. 2G). To determine whether markers of endothelial integrity/activation (including junctional markers) were altered in the lumbar spinal cord in diabetes, lumbar spinal cord protein samples were subjected to immunoblotting for VE-cadherin ( band actin). There was a marked reduction in junctional and adhesion molecules in diabetic + vehicle treated rats, with non-significant reductions in VE-cadherin (Fig. 3C) and significant reductions in CD31 (Fig. 3D) and occludin (Fig. 3E), all of which were prevented by VEGF-A 165 b treatment.
To investigate the possible functional changes in the microvasculature within the spinal cord in diabetes, Evans blue extravasation was measured (Xu et al. 2001). Evans blue vascular leakage is dependent on blood flow, surface area, hydrostatic pressure and vascular permeability. In an acute diabetic rodent model (1 week) there was a pronounced decrease in Evans blue extravasation within the lumbar region of the spinal cord compared with that of the control cohort (Fig. 4A). Diabetic groups were systemically (I.P.) treated with either saline or VEGF-A 165 b (20 ng g −1 ) using a longer term diabetic model (8 weeks). In longer duration diabetes (8 weeks) there was also a significant reduction in Evans blue extravasation compared with the control naive group (Fig. 4B), which was prevented by VEGF-A 165 b (Fig. 4B). Brains were also extracted from the long term (8 week) study (Fig. 4C) although there was no difference in solute flux between groups.
Mechanical hypersensitivity and heat hyperalgesia developed only in the diabetic + vehicle animals. VEGF-A 165 b treatment (timepoint of administration shown via arrow in Fig. 5A and B) not only prevented the spinal cord vascular degeneration (consistent with previous experiments; Hulse et al. 2015;Ved et al. 2017) but also prevented diabetic neuropathic pain behaviours (mechanical allodynia, Fig. 5A; and heat hyperalgesia, Fig. 5B). Within the spinal cord, not only was the vasculature disturbed in the diabetic + vehicle group, but sensory neurons (NeuN) in the dorsal horn also expressed increased cleaved caspase-3 (CC3), an indicator of neuronal damage, when compared with naïve age matched animals (Fig. 5C). There was increased CC3 immunoreactivity in the superficial lamina (I and II) of the dorsal horn of diabetic animals (Fig. 5D). This was blocked by VEGF-A 165 b treatment ( Fig. 5E and F). Sensory neurons within the spinal cord, once activated, express the immediate early gene c-fos (Hunt et al. 1987). c-fos is a marker of neuronal activation in chronic pain states (Kalynovska et al. 2017;Khasabov et al. 2017) and is an indicator of spinal neuronal activation in diabetic neuropathic pain (Morgado et al. 2010). There was an increase in c-fos expression in neurons (NeuN co-labelled) within the lumbar dorsal horn in the spinal cord in diabetic rats ( Fig. 6A  diabetes induced c-fos expression ( Fig. 6A and B). All lamina of the dorsal horn demonstrated increased c-fos expression in neurons in the diabetic + vehicle group compared with the naïve and diabetic + VEGF-A 165 b groups (Fig. 6C). The involvement of VEGF-A/VEGFR2 signalling on spinal cord endothelial cell function and survival and the relationship to the generation of behavioural hypersensitivity was determined by using an endothelial cell-specific inducible VEGFR2 knockdown in vivo. Systemic tamoxifen treatment led to a reduction in VEGFR2 protein expression in endothelial cells from vegfr2 flfl Tie2CreER T2 -positive mice, i.e. VEGFR2 endothelial cell knockout (VEGFR2 ECKO ) compared with control mouse endothelial cells (CTL = vegfr2 flfl Tie2CreER T2 -negative + tamoxifen) isolated from lung ( Fig. 7A-B), as well as a reduction in VE-cadherin expression ( Fig. 7C and D). In the spinal cord there was a reduction in VEGFR2 (Fig. 7E) as well as VE-cadherin (Fig. 7F) and CD31 (Fig. 7G) expression in the VEGFR2 ECKO animals compared with CTL animals. In addition, VE-cadherin protein expression from the spinal cord of VEGFR2 ECKO animals was reduced when compared with the CTL animals ( Fig. 7H and I).
When compared with CTL mice (Fig. 8A-C), the VEGFR2 ECKO animals (Fig. 8A, low power; Fig. 8B, high power) showed a significant decline in vascular integrity within the lumbar spinal cord 8 days after the final drug injection. The endothelium demonstrated a reduced endothelial volume (Fig. 8C) as well as a reduced diameter of spinal cord microvessels in the VEGFR2 ECKO mice treated with tamoxifen when compared with CTL mice (and VEGFR2 fl/fl with Cre but not given tamoxifen 'vehicle') ( Fig. 8D). There was an increase the number of smaller vessels in the VEGFR2 ECKO mice when compared to control mice (CTL and vehicle) (Fig. 8E).
In VEGFR2 ECKO and CTL mice there was no difference in nociceptive behaviour prior to tamoxifen administration (Fig. 8F). The vegfr2 flfl Tie2CreER T2 mice treated with vehicle ('vehicle') and CTL mice treated with either vehicle or tamoxifen demonstrated no change in nociceptive behavioural responses to heat (Fig. 8F). However, the VEGFR2 ECKO mice demonstrated a pronounced heat hypersensitivity compared with all other groups following tamoxifen injection (Fig. 6I). To exclude the involvement of a subset of haemopoietic cells that express Tie2 and VEGFR2 we determined the impact of VEGFR2 knockdown in this cell population. There was no change in the number of F4/80 (macrophage marker) or CD11b marker of leukocytes (including monocytes, neutrophils, natural killer cells, granulocytes and macrophages) positive cell types isolated from the spleen in either the CTL mice treated with tamoxifen or the VEGFR2 ECKO mice (Fig. 9).

N. Ved and others
J Physiol 596.16

Discussion
We show here that the microvasculature in the spinal cord was disrupted in a rodent model of diabetic neuropathic pain, as demonstrated by a reduction in the volume of the blood vessel network in the spinal cord. These findings are associated with neuropathic pain development and spinal neuronal activation. There was a concurrent reduction in spinal cord expression of the VEGF-A 165 b isoform, but no overall change in total VEGF-A expression. In the inducible VEGFR2 ECKO mice, the vasculature in the lumbar spinal cord was also reduced and was accompanied by the development of hyperalgesia. Critically, reversing the diabetes-induced vascular degeneration using systemic VEGF-A 165 b treatment also resulted in reversal of chronic pain. Neuropathic pain is common in diabetic patients and is typically associated with hyperglycaemia impacting upon the peripheral vasculature through alterations in blood flow and solute leakage (Poduslo & Curran, 1992;Tesfaye et al. 2013;Hulse et al. 2015). The peripheral sensory nerves are compromised in these instances through a degeneration of nerve terminals (loss of intraepidermal nerve fibre innervation) in the skin (Narayanaswamy et al. 2012), atrophy of the nerve trunk (shrinkage of nerve fibre diameter) , reduction in nerve conduction velocity (Ali et al. 2014) and excitation of C nociceptor fibres (Chen & Levine, 2001). However, despite these widely diagnosed symptoms, current treatments have poor success rates and low long-term success. Recent advances in understanding diabetic pain have highlighted that neuropathic pain is associated with activation and alteration of nociceptive processing in the CNS. The spinal cord (Morgado et al. 2010) and brain (Silva et al. 2013) are hyperactivated in rodent models of diabetic neuropathic pain, with increased activity of spinal neurons of wide dynamic range (Pertovaara et al. 2001;Morgado et al. 2010;Tan et al. 2012). Microvascular degeneration is disrupted in the brain of diabetic rodents and linked to, for example, cognitive impairment (Taylor et al. 2015;Hardigan et al. 2016), although the impact of hyperglycaemia on the spinal cord microvasculature has not been reported previously. The results we present here support the conclusion that there is a widespread vasculopathy in the spinal cord, which is associated with diabetes and nociceptive processing. The reduced spinal vasculature associated with hyperglycaemia could be in part due to an increase in endothelial cell death. This compromises the function of the spinal cord microcirculation -to provide delivery of solutes as well as appropriate endothelial cell signalling -as evidenced by altered neuronal activation (c-fos staining) in all five lamina of the dorsal horn, and caspase staining in laminae 1 and 2. Whether the effect on sensory neuronal function within the spinal cord is impaired as a result of decreased nutrient/oxygen delivery or through disturbance of an endothelial-neuronal-glial signalling event remains to be determined. Multiple diabetic complications arise due to vascular pathology whereby vessel or cell function is impaired (e.g. retinopathy, nephropathy) (Gupta et al. 2013;Oltean et al. 2014). We have previously shown that enhanced extravasation in diabetes does occur in the peripheral nervous system, including in the DRG, and this is mediated in part through VEGF-A, as treatment with VEGF-A 165 b (which can inhibit endogenous VEGF-A 165 b signalling through VEGFR2, and prevent TRPA1 and TRPV1 activation in cultured DRG cells ex vivo) blocked the enhanced extravasation and the associated pain   Figure 7. Tamoxifen induced VEGFR2 knockout VEGFR2 was inducibly knocked out in endothelial cells by crossing vegfr2 fl/fl with Tie2Cre ERT2 mice and treating with tamoxifen (VEGFR2 ECKO ). A, in lung tissue, VEGFR2 protein was detected in tamoxifen-treated vegfr2 fl/fl mice lacking Cre (CTL), but not in VEGFR2 ECKO mice. B, immunoblot densitometry demonstrated reduced VEGFR2 protein in the VEGFR2 ECKO mice compared with controls ( * P<0.05, n = 6 per group). C and D, VE-cadherin expression was also reduced in lung tissue of the VEGFR2 ECKO mice ( * P < 0.05, n = 6 per group). E, there was a reduction in VEGFR2 in the spinal cord of the VEGFR2 ECKO mice when compared to CTL mice ( * P < 0.05, n = 4 per group. This was accompanied by reductions in endothelial markers (F) VE-cadherin and (G) CD31 ( * * P < 0.01, n = 4 per group). H, western blot of the lumbar spinal cord from VEGFR2 ECKO mice demonstrating (I) a reduction in VE-cadherin expression when compared to CTL mice.
behaviours. In this case we describe a reduction in solute flux in the CNS, which combined with a loss of vessels can only be attributed to a reduction in blood flow. Evans blue extravasation is often taken as an indication of increased vascular permeability, and if blood is still flowing to the vessels where the vascular permeability increases, this is often consistent with the data. However, extravasation requires a provision of solute in the blood, and therefore is highly dependent on blood flow. A reduction in Evans blue extravasation in the spinal cord, where the permeability is already very low compared with most tissues (e.g. heart, skin, DRG; Hulse et al. 2015), can only be reasonably interpreted as an indication of reduced surface area for exchange, or reduced capillary pressure. The loss of endothelial cell staining, reduction of CD31 combined with reduced Evans blue extravasation indicates that diabetic animals have lost functional blood vessels from the spinal cord. This is an interesting comparison with most other vascular beds [including eyes (Ved et al. 2017) and the peripheral nerves ], where diabetes results in increased solute flux, probably through increased vascular permeability (Poduslo & Curran, 1992). In peripheral tissues such as in the sensory nerve and the DRG, increased blood-nerve barrier breakdown could arise due to direct glucose toxicity upon the vessels in the epineurium. However, sensory nerves are activated following exposure to high blood glucose (Chen & Levine, 2001;Chen & Levine, 2003). Such activity would drive peripheral extravasation through release of CGRP or activation of TRPA1; this is activated in sensory nerves of diabetic rodents (Koivisto et al. 2012;Hulse et al. 2015). These systems are not expected to be in play in the spinal cord so alternative mechanisms must be in action.
In this study we find that alterations in VEGF-A/VEGFR2 activation, as evidenced both by inhibition of VEGFR2 by VEGF-A 165 b (Bates et al. 2002;Cébe Suarez et al. 2006) and by VEGFR2 knockout, affect the decline in microvascular function. The VEGF-A family consists of two families of alternative spliced isoforms termed VEGF-A xxx a and VEGF-A xxx b (xxx denote amino acid number). These isoforms differ solely due to exon 8 splicing giving rise to differing C terminus sequences that critically alter isoform function (Harper & Bates, 2008). It has been shown that both bind to VEGFR2 with equal affinity and both possess cytoprotective actions, however VEGF-A xxx a is pro-angiogenic and VEGF-A xxx b is able to inhibit VEGF-A xxx a mediated angiogenesis (Woolard et al. 2004). Therefore VEGF-A 165 b (acting via VEGFR2) may act as a vascular protective agent under normal conditions, preventing loss of blood vessels. Loss of VEGF-A 165 b in diabetic spinal cord removes such protection, resulting in vascular damage, and reducing both vascular integrity and function. This concept, that loss of endogenous endothelial cell VEGFR2-mediated cytoprotection contributes to nociceptive processing, is consistent with the results from the endothelial-specific VEGFR2 KO mice. These mice demonstrated loss of spinal cord endothelium and associated development of hyperalgesia, consistent with the concept that endogenous maintenance of VEGFR2 activity is cytoprotective to the spinal cord endothelium, albeit that we have not directly measured spinal cord neuronal nociceptive processing. One interesting difference between the diabetic and VEGFR2 ECKO animals was that whereas the diabetic rats had a reduction in larger microvessels and no change in smaller vessels, the VEGFR2 ECKO animals had increased numbers of small but reduced numbers of larger vessels. This suggests that in diabetes the loss of vessels was due to either selective loss of larger microvessels (8-12 μm) or a combination of loss of all vessels and a reduction in size of the larger vessels (hypotrophy or atrophy, or vasoconstriction). In contrast, the VEGFR2 ECKO mice showed the latter effect. The functional implications would still be ischaemia, but may have other subtle differences in nutrient or cell delivery. While our understanding of spinal nociceptive neuronal processing in diabetes is progressing, the contribution of microvascular alteration in this context has not been investigated. Furthermore, it has been shown that peripheral sensory neurons become sensitised upon reduced perfusion (So et al. 2016), leading us to speculate that the changes in the spinal cord vascular network, as a result of hyperglycaemia that we report, could alter the microenvironment of the spinal cord sensory neurons and thus alter their level of activation. This would be anticipated to contribute to changes in pain perception and underlie neuropathic pain development.
The VEGFR2 knockdown was restricted to Tie2-positive cell types, which encompass the endothelial and haemopoietic cell populations. The Tie2 promoter-driven transgenic models are widely used to investigate endothelial function (Makino et al. 2014;Moyes et al. 2014), although it has been reported that Tie2 is expressed on CD11b/CD45-positive cells. (De Palma et al. 2005;Tang et al. 2010). In this study there was no impact of VEGFR2 knockout on haematopoietic cell populations. An impact of this knockout system on circulatory macrophages and consequent endothelium would be expected to be minimal as there is a very small number of, if any, Tie2-positive, VEGFR2-positive macrophages (Okubo et al. 2016).
People with diabetes are susceptible to neurological disease as a result of microvascular dysfunction, impacting on motor, autonomic and cognitive systems, and increased susceptibility to stroke and dementia (Kissela et al. 2005;Tiehuis et al. 2008). The VEGF-A angiogenic family has a pivotal role in managing vascular development and function (Bates, 2010), but additional evidence now also supports a fundamental role for this family in neuronal activity and survival (Verheyen et al. 2012;Beazley-Long et al. 2013). As a consequence, VEGF-A treatment has been trialled for peripheral diabetic neuropathy (Schratzberger et al. 2001;Ropper et al. 2009). Experimental hyperglycaemia leads to a reduction in VEGF-A and VEGFR2 expression in the brain; VEGF-A treatment can protect both endothelium and neuronal circuits (Taylor et al. 2015). Dysfunction of the VEGF-A signalling pathway within the hippocampus impacts upon the microvasculature and leads to impaired spatial memory (Reeson et al. 2015;Taylor et al. 2015). A decline in vascular support and diminished vascular response in diabetes and stroke models therefore clearly affects neuronal function. However, VEGF-A 165 a also increases vascular permeability and stimulates abnormal angiogenesis, which is detrimental in the CNS, for instance in diabetic retinopathy (Perrin et al. 2005). The VEGF-A xxx b isoforms, however, do not stimulate angiogenesis or increase solute flux in diabetes, and in fact they can reverse it in rodent models of diabetic retinopathy (Ved et al. 2017). Despite these opposing profiles both families are cytoprotective for endothelial cells and neurons (Beazley-Long J Physiol 596.16 et al. 2013;Oltean et al. 2014;Hulse et al. 2015;Vencappa et al. 2015). For example, reduced VEGF-A 165 b expression in diabetic patients is associated with reduced kidney function (Oltean et al. 2014). Furthermore, treatment of diabetic mice with VEGF-A 165 b prevented endothelial dysfunction in the kidney of these animals (Oltean et al. 2014). Here we show that a reduction in VEGF-A 165 b in the lumbar spinal cord of diabetic animals was associated with neuropathic pain and a degeneration of the spinal vasculature. Systemically reintroducing VEGF-A 165 b prevented vascular degeneration, spinal cord neuron activation and pain. VEGF-A 165 b was reduced in diabetic rats, despite reports that VEGF-A 165 b is associated with inhibition of vessel growth (Woolard et al. 2004). Thus VEGF-A 165 b should be considered both an anti-angiogenic and an endothelial survival factor -or a homeostatic counterpart to its vascular remodelling isoform, VEGF-A 165 a. It must be noted that the actions of VEGF-A 165 b and VEGFR2 on pain may not be restricted to vascular protection but also to effects on the spinal cord neurons, either indirectly, as VEGF-A 165 b has been shown to inhibit peripheral sensory neuron excitability and activation of peripheral nociceptors induces c-fos expression in the dorsal horn , or directly through inhibiting VEGF-A 165 a actions on spinal cord circuitry (Hulse et al. 2016). Accompanying this, neutralising endogenous VEGF-A 165 b as well as pharmacological blockade of VEGFR2 signalling at the level of the spinal cord led to the development of pain (Hulse et al. 2016). In addition, reductions in VEGF-A 165 b expression and no change in total VEGF-A expression would highlight a plausible increase in alternative isoforms such as VEGF-A 165 a, an event associated with pro-nociception and chronic pain (Hulse et al. , 2016, which cannot be ruled out in this instance. Furthermore, some consideration does need to be made when using diabetic models such as STZ. STZ is a reliable rodent model of type 1 diabetic neuropathy displaying comparable pain behaviour, nerve electrophysiological parameters and nerve histology, used widely throughout the field of diabetes research. However, it is an experimental model of islet cell ablation by a toxic agent that may have other effects, so despite its widespread use in many experimental models of diabetic complications (nephropathy, neuropathy, retinopathy and others) it is still critical that these findings are reproduced in type II diabetic models and other models of type I diabetes before being applied to people with diabetes.
These findings lead us to speculate that diabetes-induced alterations of the somatosensory systems, for instance by affecting the vasculature supporting somatosensory processing at the level of the spinal cord, could be a key concept in regulating neuropathic pain, and treatment of such complications (e.g. spinal cord vasculopathy) could provide a key target in treating diabetic neuropathic pain.
In summary, we report for the first time a significant vascular degeneration in the spinal cord of diabetic rats, which is associated with a loss of spinal VEGF-A 165 b. This was accompanied by spinal neuron activation, indicative of altered function of the spinal cord neurons, and also by enhanced nociceptive pain behaviour. Administration of VEGF-A 165 b alleviated both the spinal cord vascular degeneration and neuropathic pain. This work is complementary to previous work in the nervous system demonstrating a neuroprotective action of VEGF-A (Reeson et al. 2015;Taylor et al. 2015), but here we demonstrate central effects in the spinal cord with potential contribution to the control of chronic pain development, rather than actions purely at peripheral sites. These findings provide additional avenues for further understanding diabetic neuropathic pain and the possible mechanisms that underlie sensitisation of nociceptive pathways.