Diabetic Microangiopathy: Impact of Impaired Cerebral Vasoreactivity and Delayed Angiogenesis After Permanent Middle Cerebral Artery Occlusion on Stroke Damage and Cerebral Repair in Mice

Diabetes is an independent risk factor for stroke, tripling its incidence and affecting stroke severity (1) and related mortality (2). Although stroke is classically considered as a macrovascular complication of diabetes due to accelerated atherosclerosis and carotid artery disease, there is increasing evidence that microvasculature of the brain is also severely affected in diabetic patients. Indeed, in addition to diffuse brain atrophy, leukoaraiosis, microbleeds, and asymptomatic lacunar infarcts (<15 mm in diameter, often multiple) have been reported on brain MRI or in postmortem studies (1). Moreover, the prevalence and severity of other microvascular diabetes complications, such as retinopathy and proteinuria, are correlated with increased stroke risk (1). Functional cerebral microcirculation such as decreased baseline regional cerebral blood flow (CBF) and impaired vasoreactivity to CO₂ have been shown in diabetic patients (3). Yet, vessel diameter is a key determinant of stroke extension. Under normal physiological conditions, vessel diameter is regulated by several pathways that ensure sufficient blood flow. After cerebral artery occlusion, the consecutive drop in local CBF is compensated for by collateral artery recruitment and vasodilation, which limit lesion extension, a phenomenon mediated by nitric oxide (NO) (4,5). Later on, angiogenesis initiated during the first 3 days after brain injury participates in brain recovery by triggering neurogenesis and the migration of neural progenitors to the infarct (6).

The superimposition of a secondary injury on preexisting cerebral microvascular disease could have deep effects on neurological function. However, the microvascular “response” during the acute and subacute phases of stroke in patients with diabetes remains unstudied.

Our working hypothesis is that chronic hyperglycemia will result in dysfunctional cerebral microcirculation that in turn will contribute to enhance brain injury after focal cerebral ischemia, during the acute phase by compromising local collateral artery recruitment, an important step in tissue preservation, and during the subacute phase by compromising angiogenesis, an important step in brain repair. We first characterized functional and structural alterations in cerebral microvasculature in an experimental model of type 1 diabetes induced by streptozotocin (STZ) in adult male C57Bl/6J mice, and then evaluated the effects of experimental focal cerebral ischemia by permanent middle cerebral artery occlusion (pMCAo) 1 day and 1 week after stroke.

All experiments using animals were performed according to National Institutes of Health (NIH) guidelines for the care and use of laboratory animals (B 75-10-03). The study was specifically approved by our local institutional ethics committee (CEEALV/2012-05-01). All experiments were performed by investigators blind to the diabetes status of mice.

Male C57Bl/6J mice (JANVIER LABS, Le Genest-Saint-Isle, France) aged 4 weeks with a mean body weight of 18–20 g were divided into two groups: one group received five consecutive daily intraperitoneal injections of STZ (60 mg/kg in 100 μL of citrate buffer) to induce diabetes, and the nondiabetic group received only buffer citrate. Glycemia was tested weekly for 8 weeks. Mice with sustained hyperglycemia (>300 mg/dL) were considered to be diabetic.

All experiments were performed at the age of 12 weeks, 8 weeks following diabetes induction. We first studied functional and structural impairment of microcirculation in chronic diabetic mice before cerebral ischemia. A) We measured vasoreactivity to CO₂ and NO, characterized the expression of NO synthases and brain factors involved in vessel regulation using PCR, Western blotting, and immunohistochemistry, and performed transmission electron microscopy (TEM) of cerebral capillaries and arterioles. Brain MRI was performed before cerebral ischemia in order to detect asymptomatic microvascular lesions. Next, we studied the impact of diabetic microangiopathy on stroke damage and cerebral repair. B) We evaluated infarct volume (cresyl violet staining and brain MRI), neurological deficit, and angiogenesis at D1 and D7. Blood-brain barrier (BBB) protein modification was measured at D1 (Fig. 1).

Systolic blood pressure was measured in unanesthetized mice by the tail-cuff method (Kent Scientific Corporation, Torrington, CT) after daily acclimatization for 2 weeks.

Thermoregulated mice were subjected to ultrasound measurements under 0.5% isoflurane anesthesia using an echocardiograph (Acuson S 3000, Erlangen, Germany) equipped with a 14-MHz linear transducer (14L5 SP) as previously reported (4). Heart rate, peak systolic, end-diastolic, and time-averaged mean blood flow velocities (mBFVs) were measured in the basilar trunk (BT) before cerebral ischemia 1) under air (all mice), 2) 5 min after starting to breathe a gas mixture of 16% O₂, 5% CO₂, and 79% N₂ (n = 7–10), or 3) 5 min after administration of an NO donor (NONOate; Sigma-Aldrich, St. Louis, MO) (n = 4–5) (1 mg/kg, i.p.), in order to discriminate between endothelial and smooth muscle cell (SMC) dysfunction. Indeed, vasodilation is reestablished by the NO donor if endothelium only is dysfunctional. Vasoreactivity was estimated for each mouse as the percentage increase in mBFV recorded under gas mixture or NONOate compared with mBFV recorded under normoxic air; acidosis and hypercapnia were assessed by blood gas measurements.

The procedure was performed as previously described (n = 5–8) (7) using custom-designed primers for all genes of interest (Table 1) and a ready-to-use primer for the housekeeping gene peptidylprolyl isomerase A (cyclophilin A) (Qiagen, Courtaboeuf, France). We normalized the results for each gene to cyclophilin A levels. Results are expressed as arbitrary units (AU).

Animals (n = 5–7) were killed and brains rapidly dissected out on a cold plate. Cytosolic proteins were extracted (5), and equal amounts of proteins (40 µg) were resolved by SDS-PAGE electrophoresis and immunoblotted using the antibodies indicated in Table 1. Blots were scanned and analyzed using ImageJ (NIH, Bethesda, MD).

In brief, as previously described (7), thermoregulated (37 ± 0.5°C) and anesthetized (2% isoflurane in air) mice were subjected to left middle cerebral artery electrocoagulation (pMCAo) using bipolar forceps. The overall mortality was <5% in 24 h. Cortical infarct volume was evaluated on cresyl-violet–stained sections (every eighth 30-µm-thick coronal section) using NIH ImageJ (n = 5–9).

Neurological deficit was assessed 24 h after pMCAo, based on five neurological tests (neurological score, grip and string tests, beam walking, and the pole test), and used to calculate a global neurological score (/21) (n = 7–9). The lower the neurological score, the more severe the deficit (8).

A 2% solution of Evans blue dye (Sigma-Aldrich, St. Louis, MO) in PBS was administered through the tail vein of mice (4 mL/kg) 24 h after stroke (n = 5–6). Two hours later, animals were transcardially perfused under isoflurane anesthesia, with 20 mL of 1% heparinized saline, until a colorless perfusion fluid was obtained. Each hemisphere was placed separately in 2 mL formamide and allowed to soak for 72 h at room temperature. The absorbance of the supernatant solution was measured against a pure formamide standard at 625 nm using a spectrophotometer. The tissue was then dried in the oven at 95°C for 5 days (6), and relative absorbance measured (units/g dry weight). Results are expressed as the percentage of the value measured in nondiabetic mice (100%).

Coronal 30-µm-thick floating sections were incubated with primary antibody overnight at 4°C; anti-Ki67 (1:200; Thermo Scientific, Fremont, CA) and anti-CD31 (1:200; BD Biosciences, San Diego, CA) were used to detect proliferating cells and endothelial cells, respectively, and double immunolabeling to detect angiogenesis (n = 5–6). Appropriate Alexa Fluor 594 or 488–labeled secondary antibodies (1:400; Molecular Probes, Eugene, OR) were applied for 1 h at room temperature. Specificity was checked by omitting the primary antibody. Cell counts and microvessel density measurements were performed at three coronal brain levels (+0.80, −0.80, and −1.20 mm relative to bregma) that consistently contained the infarct area. Angiogenesis was assessed in three regions of interest (ROIs; 0.06 mm² each) located in the peri-infarct area, and expressed as the average number of Ki67⁺/CD31⁺ cells per ROI. Microvascular density was evaluated after CD31 immunolabeling by calculating the integrated pixel density of the images using NIH ImageJ. The average integrated pixel density of one field of view (20×) located in the peri-infarct area was used for analysis.

Mice (n = 4) were transcardially perfused for 12 min with a fixative containing 2% paraformaldehyde, 2.5% glutaraldehyde, and 2 mmol/L CaCl₂ in 0.1 mol/L cacodylate buffer (CB), pH 7.4, at room temperature (n = 4). Brains were removed, and 1-mm³ brain fragments were postfixed in the same fixative for 1 h at 4°C, extensively washed in CB, and fixed in 1% OsO₄ in CB for 45 min at 4°C. After washing in CB, samples were fixed in 1% aqueous uranyl acetate and finally rinsed in water. After dehydration in graded ethanol, followed by propylene oxide, the fragments were embedded in epon. Ultrathin (80 nm) sections were prepared, stained with lead citrate, and photographed with a Jeol S100 TEM equipped with a 2kx2k Orius 200-830 camera (Gatan-Roper Scientific, Evry, France).

Brain MRI was performed before and at D1 and D7 after cerebral ischemia, in order to detect pre- and poststroke (secondary hemorrhagic transformation) hemorrhages in the mouse brain. All MR examinations were performed using a 7.0 Tesla MR unit (PharmaScan; Bruker Biospin, Ettlingen, Germany) equipped with a surface coil with an internal diameter of 10 mm, in thermoregulated and anesthetized (0.8–1% isoflurane in O₂ 30%/N₂O 70%) mice (n = 5–6). Infarct volume was measured on T2-weighted images according to the RARE sequence type parameters; TR = 6,475.1 ms; TE = 8,894 ms; effective TE = 53.36 ms; field of view = 20 × 20 mm; matrix = 128 × 128; slice thickness = 0.5 mm (0.156 × 0.156 = voxel × 0.5 mm³); flip angle = 180°; RARE factor = 16; and number of averages = 2. Maximal gradient slope was 300 mT/m. Gradient-echo T2-weighted images (T2-star) were obtained to evaluate the presence of hemorrhagic spots with the following parameters: TR = 260 ms; TE = 6 ms; field of view = 20 × 20 mm; matrix = 256 × 256; slice thickness = 0.75 mm; flip angle = 55°; and number of averages = 2. The brain of each mouse was imaged before and after an intraperitoneal injection of 0.1 mmol/kg gadopentetate dimeglumine (Gd-DTPA; Magnevist; Berlex Laboratories, Wayne, NJ) with a final set of postcontrast T1-weighted images with the following parameters: spin-echo sequence; TE = 10 ms; TR = 500 ms; FOV = 20 × 20 mm; matrix = 256 × 256; 12 slices; slice thickness = 0.75 mm; and total imaging time = 2 min.

Statistical analyses were performed with Prism 5 software (Prism 5.03; GraphPad, San Diego, CA). Data are expressed as mean ± SD. Comparisons between nondiabetic and diabetic mouse groups were carried out using the nonparametric Mann-Whitney U test. A P value of <0.05 was considered statistically significant.

Mean blood glucose concentration (424 ± 54 vs. 161 ± 10 mg/dL; P < 0.001) was significantly increased and body weight (23.7 ± 1.1 vs. 21.08 ± 2.4 g, P = 0.003) significantly decreased in diabetic mice compared with nondiabetic mice. At 12 weeks, systolic, diastolic, and mean arterial pressures were not significantly different between groups. Blood gas after CO₂ inhalation showed significant acidosis and normoxic hypercapnia in both groups compared with measurements under air (Table 2).

Inhalation of a CO₂ gas mixture resulted in significant hypercapnia (P = 0.017) and acidosis (P = 0.006) in diabetic and nondiabetic mice (Table 2); mBFV in the BT was not significantly modified after CO₂ inhalation in diabetic mice (air 100 ± 14.5% vs. CO₂ 98.8 ± 11.9%; P = 1), whereas mBFV in the BT was significantly increased in nondiabetic mice (air 100 ± 6.4% vs. CO₂ 124.2 ± 24.2%; P = 0.0004) (Fig. 2A). After NO donor injection, the BT mBFV of diabetic mice was not significantly modified (air 100 ± 13.8% vs. NO 90.2 ± 17.4%; P = 0.2), whereas it was significantly increased in nondiabetic mice (air 100 ± 6.2% vs. NO 122.6 ± 8.7%; P = 0.0009) (Fig. 2B).

There was a significant decrease in endothelial NO synthase (eNOS) mRNA (0.63 ± 0.4 vs. 1.44 ± 0.7 AU; P = 0.04) and phospho-eNOS protein expression (0.59 ± 0.32 vs. 1 ± 0.10 AU; P = 0.02) in diabetic compared with nondiabetic mice, although eNOS protein levels were not significantly different (0.99 ± 0.1 vs. 1 ± 0.21 AU; P = 1) (Fig. 3A). The mRNA for neuronal NO synthase (nNOS) (0.59 ± 0.37 vs. 1.32 ± 0.61 AU; P = 0.03) and protein levels of nNOS (0.92 ± 0.06 vs. 1 ± 0.04 AU; P = 0.05) and phospho-nNOS (0.36 ± 0.26 vs. 1 ± 0.12 AU; P = 0.002) were significantly decreased in diabetic compared with nondiabetic mice (Fig. 3B).

mRNA for the adherens junction molecule vascular endothelial cadherin (VE-cadherin) was significantly decreased (0.71 ± 0.49 vs. 1.39 ± 0.61 AU; P = 0.03) in diabetic compared with nondiabetic mice, whereas VE-cadherin protein levels (1.75 ± 0.24 vs. 1.89 ± 0.44 AU; P = 0.5) remained unchanged. There was a significant decrease in mRNA (0.41 ± 0.18 vs. 0.94 ± 0.31 AU; P = 0.001) and protein levels (0.19 ± 0.03 vs. 0.27 ± 0.05 AU; P = 0.01) of the tight junction protein zona occludens 1 (ZO-1) in diabetic compared with nondiabetic mice (Fig. 4A). However, TEM did not show any modification of tight junction structure in the microvessels of diabetic mice (Fig. 4B). Brain MRI did not show any hemorrhages or gadolinium enhancement (data not shown).

Cortical microvessel density was increased (mean gray surface value: 21.5 ± 0.9 vs. 15.61 ± 1.81 AU; P = 0.02) in diabetic versus nondiabetic mice. Vascular endothelial growth factor a (VEGFa) mRNA (0.56 ± 0.29 vs. 1.01 ± 0.39 AU; P = 0.02) and protein expression (0.51 ± 0.09 vs. 1.0 ± 0.36 AU; P = 0.05) were significantly decreased in diabetic compared with nondiabetic mice; angiopoietin 2 (Ang2) mRNA (0.24 ± 0.13 vs. 0.32 ± 0.14 AU; P = 0.3) was not significantly modified, although Ang2 protein levels were decreased (0.56 ± 0.21 vs. 1 ± 0.18 AU; P = 0.001) (Fig. 4C). In addition, mRNAs for Ang1 (0.48 ± 0.33 vs. 0.88 ± 0.39 AU; P = 0.05), transforming growth factor-β (TGF-β) (0.77 ± 0.19 vs. 1.25 ± 0.43 AU; P = 0.02), and platelet-derived growth factor-β (PDGF-β) (0.41 ± 0.16 vs. 0.80 ± 0.24 AU; P = 0.003) were significantly decreased in diabetic compared with nondiabetic mice.

Infarct volumes were significantly larger in diabetic compared with nondiabetic mice at D1 (17.72 ± 2.82 vs. 14.02 ± 0.99 mm³; P = 0.03) and D7 (14.24 ± 4.50 vs. 2.84 ± 0.50 mm³; P = 0.04) (Fig. 5A and B). The neurological score (/21) was significantly worse at D1 (13 ± 3 vs. 18 ± 2; P < 0.001) and D7 (11 ± 5 vs. 19 ± 1; P = 0.008) in diabetic compared with nondiabetic mice, which did not show any neurological deficit (Fig. 5C). No hemorrhagic transformation or suffusion of blood was visible on brain MRI at either time point (Fig. 5D).

Evans blue extravasation was increased in diabetic (114 ± 9%) compared with nondiabetic mice (100 ± 4%; P = 0.02) (Fig. 6A). ZO-1 (0.12 ± 0.02 vs. 0.16 ± 0.04 AU; P = 0.03) and VE-cadherin protein levels (1.16 ± 0.21 vs. 1.87 ± 0.32 AU; P = 0.03) were significantly decreased in diabetic compared with nondiabetic mice (Fig. 6B).

At D1, endothelial cell proliferation (36 ± 11 vs. 46 ± 11; P = 0.2) was not significantly different between diabetic and nondiabetic mice. Similar to prestroke conditions, VEGFa, TGF-β, and PDGF-β mRNA levels at D1 were lower in diabetic compared with nondiabetic mice (data not shown). In contrast, at D7, there was a significant increase in endothelial cell proliferation (16 ± 3 vs. 11 ± 4 cells/ROI; P = 0.05) and a significant increase in VEGFa mRNA (0.67 ± 0.20 vs. 0.42 ± 0.12 AU; P = 0.03) and protein (0.89 ± 0.49 vs. 0.37 ± 0.13 AU; P = 0.05), and Ang2 mRNA (0.96 ± 0.67 vs. 0.36 ± 0.11 AU; P = 0.05) and protein levels (1.91 ± 0.81 vs. 4.0 ± 1.66 AU; P = 0.05) in diabetic compared with nondiabetic mice (Fig. 6C). Ang1, PDGF-β, and TGF-β mRNA levels were unchanged (data not shown).

The current study, performed in mice with STZ-induced type 1 diabetes, shows that cerebral microcirculation is affected after 8 weeks of hyperglycemia; vasoreactivity to inhaled CO₂ is abolished and not reversed by NO donor administration, whereas mRNA and protein levels of phospho-eNOS, nNOS, and phospho-nNOS are decreased, suggesting smooth muscle and endothelial cell dysfunction. In addition, microvessel density is increased, whereas the downregulation of BBB proteins and the pattern of angiogenic factor expression suggest an immature vascular network. After the induction of experimental cerebral ischemia, infarct volume, neurological deficit, and BBB permeability are significantly increased and angiogenesis, which is important for brain repair process, is delayed. Taken together, these findings could explain the greater damage during the acute phase of stroke and delayed brain repair in these mice.

To our knowledge, this is the first study to reproduce impaired cerebral vasoreactivity after CO₂ inhalation in diabetic mice, facilitating the extrapolation of our methodology and results to humans. Our diabetic mice did not exhibit cerebral vasodilation after CO₂ inhalation or NONOate injection, suggesting an SMC dysfunction. SMC alterations have already been reported in diabetes (9) and might be explained by primary reduced NO bioavailability or direct cell toxicity by increased production of reactive oxygen species and peroxynitrite. NO is the main vasodilator involved in the increased CBF induced by experimental hypercapnia in normal brains, since cerebral vasoreactivity is suppressed by NOS inhibitors and reversed by NO agonists (10). NO is produced in the brain by endothelial cells (eNOS) and neurons (nNOS) (11). However, the data concerning their involvement are conflicting (12–15). Although endothelial dysfunction in the cerebral arteries of patients with diabetes is generally accepted (16), there is no modification of eNOS protein level in STZ-induced diabetic rats (17), downregulation in high-fat diet–induced diabetic mice (18), or upregulation in obese diabetic Zucker rats (19). Upregulation was unevenly interpreted as a compensatory mechanism to maintain acceptable NO levels. In our model, although there was no modification in eNOS protein levels, we did observe a decrease in the enzymatically active phosphorylated form of the protein (20). Since the endothelial production of NO requires the phosphorylation of specific eNOS residues by protein kinase B or Akt and since insulin-mediated Akt phosphorylation is impaired in diabetic mice, we can speculate that this might be responsible for the downregulation of eNOS in the brain of diabetic mice. Consistent with these results, mice with the phosphomimetic “SD” mutation of eNOS display reduced infarct size and neurological deficit after MCAo compared with eNOS mice with the unphosphorylatable “SA” mutation (21). The expression of nNOS, whose enzymatic activity similarly depends on its phosphorylation (22), is decreased (17) or unchanged (19) in STZ-induced diabetic rats. We therefore also systematically investigated nNOS phosphorylation in the brain of diabetic mice and demonstrate, to our knowledge for the first time, that both eNOS mRNA and phospho-eNOS protein as well as nNOS mRNA and nNOS and phospho-nNOS protein levels are decreased in diabetic mice, suggesting that a downregulation of both NOS isoforms may contribute to impaired CBF regulation in diabetic mice.

In acute stroke settings, collateral circulation can sustain tissue perfusion following a proximal arterial occlusion. The leptomeningeal collateral network represents the final route for the perfusion of ischemic territory distal to the proximal occlusion (23). After arterial occlusion, the reduction in CBF is more pronounced in the infarct core and less severe at the periphery or penumbra, due to NO-dependent collateral circulation recruitment (24). Furthermore, the inactivation of both nNOS and eNOS activity leads to lesion extension (5). Here, we have shown that a combination of endothelial and SMC dysfunction and eNOS and nNOS downregulation in 8-week diabetic mice prevents the dilatation of cerebral arterioles, compromising the adaptation of blood flow to metabolic rate, and potentially explaining the greater infarct volume and neurological deficit. One should thus encourage the development of therapeutic strategies that aim to improve the early recovery of vasoreactivity, for example by using statins or angiotensin type 1 (AT1) receptor blockers to modify the NO pathway (25).

Another mechanism that might take part in increased stroke severity is increased BBB permeability, as shown by Evans blue extravasation, and as already demonstrated in a model of cerebral artery occlusion and reperfusion (tMCAo) in type 2 diabetic mice (26). We found no hemorrhagic transformation on brain MRI, in accordance with previous reports indicating that hemorrhage is mainly found in tMCAo (27). There was also no gadolinium enhancement, although BBB proteins levels were lower in diabetic mice, suggesting that gadolinium assessment might not be sufficient to detect these changes.

Blood vessel stabilization depends on the equilibrium between proangiogenic factors (VEGFa and Ang2) and maturation and stabilization factors (Ang1, PDGF-β, and TGF-β). We have shown that prior to cerebral ischemia, the brain microvasculature in diabetic mice is immature, with a decrease in VEGFa, Ang2, Ang1, PDGF-β, and TGF-β mRNA levels. This is consistent with another model of diabetes, the Goto-Kakisaki rat, in which there are more prominent unperfused new brain vessels and a decreased pericyte-to-endothelial cell ratio, suggesting a “destabilizing vessel profile” (28). However, this immature vascular phenotype is associated with increased cortical vascular density in Goto-Kakisaki rats, also reported in Lepr^db/db mice by Prakash and colleagues (28,29), and increased VEGFa protein levels in micro- but not macrovessels (28). This is in contrast with our own findings, where increased vessel density was instead associated with a downregulation of VEGFa mRNA and protein levels. In contrast to our experimental conditions, Prakash et al. (28,29) used a rat genetic model of type 2 diabetes with a shorter duration of diabetes and measured VEGFa levels in microvessels and not the whole brain. As the duration of chronic hyperglycemia is longer in our protocol, it is possible that toxic metabolic end products due to the disease might interfere with proangiogenic pathways under chronic conditions (30).

At D1 poststroke, mRNAs for angiogenic factors in diabetic mice were not different from values obtained before stroke, and still lower than those in nondiabetic mice. At D7, PDGF-β, TGF-β, and Ang1 mRNA were not significantly different between diabetic and nondiabetic mice, whereas a significant upregulation of VEGFa and Ang2 mRNA and protein levels was observed, with increased angiogenesis. From these time points, we can deduce that the time course of vessel repair is delayed in our model, since these factors should have returned to basal levels otherwise (31). These results are in contrast to those of Cui et al. (26) who investigated the regulation of Ang1 and Ang2 1 day after stroke in a model of type 2 diabetic mice and found early modifications with decreased Ang1 and increased Ang2 levels. However, protein levels in their study were not compared with nonischemic mice but with the contralateral hemisphere. Since there could be compensatory changes in the contralateral hemisphere, we chose to compare our ischemic diabetic mice to nonoperated diabetic mice. Poststroke angiogenesis results in increased vessel density from D1 that continues until D21 (31), and is maximal between D3 and D7. In a model of tMCAo in CD1 mice, VEGFa and its receptor flk1 (VEGFR2) were maximally upregulated from 3 h until D3 for the ligand and until D7 for the receptor, at which time they returned to baseline. Ang2 follows the same time course, whereas Ang1 expression is delayed with maximal expression at D7. To our knowledge, there is no report as to the time course of proangiogenic factors after stroke. As we have shown previously in C57Bl/6J mice following pMCAo that angiogenesis is also maximal between D3 and D7 (6), we suggest that the time course of expression of proangiogenic factors is also comparable, despite the use of different models. To our knowledge, this is the first time that these factors have been sequentially measured in diabetic mice after stroke. NO is also a positive regulator of neovascularization by contributing to the proangiogenic effect of growth factors (32), and its downregulation in diabetic mice might prevent proper angiogenesis.

Although our experimental data were obtained from two different sets of animals and a causal relationship thus cannot be strictly verified, we believe that cerebral microvessel impairment, already present in nonoperated diabetic mice, is at least partly responsible for worsened stroke outcome. 1) After cerebral ischemia, the magnitude of vasoreactivity, i.e., the recruitment of collateral circulation, is inversely proportional to brain damage (24), and the vasoreactivity impairment in our model suggests that after stroke, compensatory mechanisms, such as collateral opening and patency, are not sufficient to limit brain damage. 2) The magnitude of BBB opening after stroke determines the extent of the infarct (33). This is also observed in the diabetic mice in our study, whose cerebral vessels display an immature phenotype with a relative decrease in proangiogenic factors and BBB proteins, suggesting BBB “fragility.” 3) As shown in a previous work (6), cerebral repair is partly dependent on angiogenesis, and we could speculate that if proangiogenic factors are already decreased before stroke, they will be less effective in initiating an angiogenic response after stroke, as also observed in the diabetic mice in our study. Nonetheless, we cannot exclude the involvement of other factors such as brain-specific basal inflammation in diabetic mice in worsened stroke outcome. Indeed, inflammation is a key contributor to brain damage after stroke (34), and the level of inflammatory cytokines (interleukin-1β, tumor necrosis factor-α, and interleukin-6) has been shown to be increased in the hippocampus of obese and diabetic db/db mice (35). The contribution of inflammation to brain damage STZ-induced diabetic mice following stroke warrants further study.

Alterations of the microvasculature associated with peripheral diabetic neuropathy have also been reported. Demiot et al. (36) have observed an impaired pressure-induced vasodilation response associated with neuropathy in male Swiss mice 8 weeks after they were given a high dose of STZ (200 mg/kg), and Jeong et al. (37) have observed decreased vasa nervorum in the sciatic nerve in a similar model. In human diabetic neuropathy, a thickening of the basement membrane, endothelial fenestration, decrease of periendothelial coverage, reduction of nerve blood flow correlated with the severity of diabetic neuropathy (38–41), and a reduction (40,41) or increase of capillary density (42) have all been reported. The absence of microvascular structural abnormalities in our model at the time of analysis could indicate that such modifications are milder or occur later in the brain.

We chose to develop a chronic induced type 1 diabetic mouse model in order to exclude hypertension and dyslipidemia, which are also responsible for vascular impairment, and other confounding factors, from our analysis. Although type 1 diabetes only represents ∼10% of diabetic patients, though this proportion is greater at younger ages (43), the relative stroke risk is four times greater in type 1 than in type 2 diabetes. Moreover, in most genetic type 2 diabetic models, the lack of the leptin receptor may introduce a bias since leptin is neuroprotective, and the lack of leptin signaling worsens damage after cerebral ischemia (44). The duration and degree of hyperglycemia in diabetes may also be critical for neurovascular outcomes. Most animal studies have been performed after acute or short-term (2–5 weeks) hyperglycemia, which might not be sufficient to cause the microangiopathy that needs to be taken into account in the pathophysiology of diabetic stroke, in particular in repair mechanisms that occur later than purely metabolic effects. However, although STZ induction is a well-accepted model of type 1 diabetes (45), it does not recapitulate the complex pathophysiology observed in type 1 diabetic patients, and the extrapolation of our findings to these patients is thus premature.

Our results provide for the first time direct evidence that diabetes alters cerebral vasoreactivity, microvessel density, permeability, and stability, all of which influence neuronal and vascular damage following ischemic brain injury not only in the short-term but at longer times. These results offer another insight into the crucial role of the neurovascular network in the outcome of stroke. They highlight the fact that there is an urgent need to understand the basic mechanisms that limit or enhance brain repair after stroke and to develop appropriate experimental models to lead to customized stroke therapies.

Acknowledgments. The authors thank Bruno Palmier (Université Paris Descartes) for Western blot guidance, the Centre d’Explorations Fonctionnelles—Imagerie (CEFI, Institut Claude Bernard, www.bichat.inserm.fr/) for MRI image acquisition, and Bruno Saubaméa (Université Paris Descartes) for performing ultrastructural imaging.

Funding. This study was supported by the Institut National de la Santé et de la Recherche Médicale (France). M.P. received a grant from the Ministère de l'Enseignement Supérieur et de la Recherche (France).

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

Author Contributions. M.P. designed and performed experiments and participated in writing and reviewing the manuscript. P.B. performed the inhaled-CO₂ vasoreactivity Doppler analysis and blood gas analyses, reviewed the manuscript, and contributed to the discussion. C.P. and L.R. performed some of the PCR and Western blots for BBB and microcirculation proteins. C.S. and A.D. performed the brain MRI experiments. C.C.-M. helped with the NOS Western blots, provided antibodies, reviewed the manuscript, and contributed to the discussion. N.K. designed the experiments and wrote and edited the manuscript. N.K. 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.