Decreased Cerebrovascular Brain-Derived Neurotrophic Factor–Mediated Neuroprotection in the Diabetic Brain

All experiments were performed following an institutionally approved protocol in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals. Diabetes was induced with streptozotocin (STZ) (57.5 mg/kg body wt) dissolved in citrate buffer, pH 4.5, through a tail-vein injection. Induction of diabetes was verified after 3 days, and rats with blood glucose concentrations >250 mg/dL were included in these studies. Body weight was measured three times per week, and 3–4 units of neutral protamine Hagedorn insulin (0.5 IU) was administered subcutaneously, as needed, to prevent weight loss and ketosis. The diabetic rats were killed, and brains were collected after perfusion along with age-matched controls. Cerebral cortex was dissected from whole brains for microvessel-enriched fractions and stored at −80°C before fraction preparation. At the time of death, blood was obtained by cardiac puncture for estimation of glycosylated hemoglobin (Glyco-Tek Affinity Kit, Helena Laboratories, Beaumont, TX).

Isolation of microvessel-enriched fractions from rat tissue was performed as described by Wu et al. (24) and Galea and Estrada (25). Briefly, brain tissues were collected from rats after PBS reperfusion. The cortical gray matter was dissected and rolled on filter paper to remove the large blood vessels and then homogenized with PBS and centrifuged. After washing with PBS, the pellet was resuspended in four volumes of 18% dextran and centrifuged at 1,500g for 20 min. The new pellet was saved, and the remaining tissue was reprocessed twice similarly. All three pellets were pooled, washed again with PBS, and lysed in lysis buffer (Cell Signaling Technology, Danvers, MA) with proteinase inhibitors for immunoblotting. Microvessel-enriched fractions in all experiments were not pooled and were derived from a single animal cortex.

For immunohistochemistry, whole brain sections (2-mm thick) were prepared from frozen rat brains. Sections were kept frozen at −80°, after which they were air-dried, fixed in ice-cold acetone, blocked with 5% BSA, and probed with an anti-BDNF rabbit polyclonal antibody (Millipore, Billerica, MA). For immunocytochemistry, cells were seeded onto gelatin-coated glass coverslips and maintained in complete media. Cells were fixed in ice-cold ethanol, rinsed with 1× PBS, blocked with 5% BSA, and probed with an anti-BDNF antibody (Millipore) or anti-CD31 antibody (BD Biosciences, San Diego, CA).

Primary human brain microvessel endothelial cells were purchased from Cell Systems Corporation (Kirkland, WA), mostly derived from a heterogenous mix of rapidly autopsied human brains obtained within a few hours after death. We used cells between passages 5 and 10 at ∼80% confluence for all studies. The endothelial cells were maintained in complete MCDB-131 medium with l-glutamine, 1 g/L d-glucose, and 10% FBS (Vec Technologies, New York, NY), and seeded onto human gelatin-coated plates for all experiments. Primary mouse embryo cortex neurons were seeded on poly-d-lysine–coated plates and maintained with Neurobasal media plus 2% B27, 0.5 μmol/L glutamine, and penicillin/streptomycin (Invitrogen, Carlsbad, CA). Neurons were used for conditioned media-transfer experiments between days 8 and 10. All treatments with BSA, advanced glycation end product (AGE)-BSA, and the MEK/extracellular signal–related kinase (ERK) inhibitor U0126 were performed in 1% serum MCDB media supplemented with l-glutamine and 4.5 g/L d-glucose. AGE-BSA and U0126 were procured from Calbiochem (San Diego, CA).

The ELISA kit for BDNF was obtained from Promega (Madison, WI). MTT assay and LIVE/DEAD viability/cytotoxicity assay were used to assess cell death. For immunoblotting, the primary antibodies were obtained from Cell Signaling Technology (BDNF antibody), Biomol/Enzo Life Sciences (neuron-specific enolase; Plymouth Meeting, PA), and Promega (total and pERK antibodies). All secondary antibodies were obtained from Pharmacia (New York, NY).

Primary endothelial cells were treated with or without AGE-BSA for 48 h, before the conditioned media were transferred to neurons. Conditioned media collected from endothelial cells were transferred to primary mouse cortical neurons cultured overnight in serum-free Neurobasal media minus B27 and antibiotics. Neurons were then placed in a hypoxic chamber (Billups-Rothenberg, Inc., Del Mar, CA). The chamber with cells was perfused with mixed atmosphere (90% nitrogen, 5% carbon dioxide, and 5% hydrogen) for 30 min and then sealed and placed at 37°C for indicated time periods of hypoxia. After 48 h, cells were removed from the chamber and changed to fresh Neurobasal serum-free media, maintained in a regular CO₂ incubator for reoxygenation. Control cultures were incubated under normoxic conditions for equivalent durations.

Quantitative data from all experiments were analyzed using the JMP-IN statistical software 7.0 (SAS Institute Inc., Cary, NC). Student t test was performed for two-group comparisons. One-way ANOVA was performed for multiple group comparisons followed by Tukey-Kramer tests for pairwise comparisons. Differences reaching P < 0.05 were considered significant.

Diabetes was induced in adult male SD rats by STZ injection. After 6 months of diabetes, the diabetic rats showed lower body weight and higher glyco-hemoglobin levels than the age-matched nondiabetic control rats (Table 1).

Diabetic and nondiabetic rat cerebral cortices were extracted, and microvessel-enriched fractions were prepared for Western blotting. These fractions were positive for vascular endothelial cadherin, confirming their endothelial identity (Fig. 1A), whereas staining for neuron-specific enolase and glial fibrillary acidic protein was weak, suggesting that they were not overly contaminated with neurons or astrocytic components (Fig. 1B). A 16-kDa band corresponding to bioactive BDNF protein was detected in all fractions. BDNF levels were significantly reduced in microvessel-enriched fractions from diabetic rats compared with nondiabetic rats (Fig. 1C and D). To confirm our findings, 6-week normal and diabetic rat brains were sectioned and prepared for immunohistochemistry. BDNF staining was observed in the blood vessels of the cortex. BDNF levels appeared to be lower in diabetic endothelium compared with normal vessels (Fig. 2).

Our in vivo data suggested that endothelial BDNF was suppressed in diabetic endothelium. To assess the mechanisms involved, we switched to an in vitro experimental model using AGE-BSA to mimic endothelial stress triggered during diabetes. Human microvascular brain endothelial cells were incubated with AGE-BSA for 48 h in serum-free media. This treatment caused a change in cell morphology, with changes in cell shape and loss of cell–cell contact (Fig. 3A).

After 48 h, BDNF protein levels in the cell lysates were assessed by Western blotting. AGE-BSA clearly induced a dose-dependent reduction in BDNF levels (Fig. 3B). Compared with normal BSA controls, BDNF levels were significantly reduced in the AGE-BSA–treated endothelial cells (Fig. 3C and D). No changes in cell viability were noted, suggesting that these reductions in BDNF were not indirectly caused by nonspecific cytotoxicity (Fig. 3E).

Immunostaining confirmed our Western blot findings. In normal human brain endothelial cells, cytoplasmic BDNF staining was observable in the perinuclear zone and in vesicles being trafficked to the plasma membrane (Fig. 4A). In AGE-BSA–treated endothelial cells, there was a clear reduction in BDNF staining compared with untreated controls (Fig. 4A). Consistent with these immunostaining observations, AGE-BSA affected the secretion of BDNF into extracellular space. Standard enzyme-linked immunosorbent assays confirmed that conditioned media from brain endothelial cells contained BDNF. AGE-BSA–treated endothelial cells showed a significant decrease in secreted BDNF compared with untreated cells or controls treated with normal BSA (Fig. 4B).

AGEs can trigger many intracellular signaling pathways. Because ERK/mitogen-activated protein (MAP) kinase is one of the major mechanisms that allow endothelial cells to respond to extracellular stimuli, we asked whether this signaling pathway might be involved in the phenomenon of AGE-BSA–induced downregulation of BDNF. Human brain endothelial cells were exposed to AGE-BSA for 48 h, and then cell lysates were probed for changes in ERK and phospho-ERK. Levels of phospho-ERK were significantly elevated in AGE-BSA–treated cells with no change in total ERK (Fig. 5A and B). Next, we tested the effects of the potent MEK/ERK inhibitor U0126 to provide pharmacologic evidence of causality. As expected, AGE-BSA decreased BDNF expression in brain endothelial cells. Blockade of ERK signaling with U0126 decreased phospho-ERK levels and significantly prevented this AGE-BSA–induced suppression of BDNF (Fig. 5C and D). Concomitantly, blockade of ERK signaling also interfered with the ability of AGE-BSA to decrease the secretion of BDNF into conditioned media (Fig. 5E). Taken together, these data suggest that the ability of AGE-BSA to suppress endothelial BDNF requires ERK/MAP kinase signaling.

Because AGE-BSA decreased the production of BDNF, we hypothesized that this would mean that “diabetic” endothelium would lose the ability to trophically protect neurons. First, we confirmed that AGE-BSA does not directly kill neurons (Fig. 6A). Next, media transfer experiments were performed whereby brain endothelial-conditioned media were added to primary mouse neurons subjected to hypoxia for 24 h. As expected, hypoxia killed neurons (Fig. 6B and C), and, as previously reported, conditioned media from normal healthy brain endothelial cells were significantly neuroprotective (Fig. 6B and C). In contrast, conditioned media from AGE-BSA–treated endothelial cells were no longer neuroprotective (Fig. 6A and B), consistent with their reduction in BDNF levels. Addition of exogenous BDNF back into the culture media restored protection and decreased neuronal death (Fig. 6B and C). Finally, the role of ERK signaling in the pathway was confirmed by U0126 experiments. Blockade of ERK with U0126 not only prevented endothelial BDNF suppression (Fig. 5D) but also restored neuroprotection (Fig. 6C).

Symptomatic strokes caused by large artery disease have garnered much attention in cerebrovascular investigations. In contrast, small vessel dysfunction in the deeper layers of the brain cortex has received less emphasis, even though this comprises approximately one-third of all symptomatic strokes (26). This problem may be especially important for diabetic patients who present a higher incidence of lacunar stroke with recursive hemorrhagic episodes (6,27). Although the effects of acute hyperglycemia on brain injury are well documented in global and focal ischemia models (28), there is a bigger need to first study the effects of long-term diabetes on the brain. Emerging studies support the notion that preemptive vascular disease in diabetes affects ischemic brain injury before, during, and after stroke by modifying adaptive neovascularization responses and postischemic tissue remodeling (29,30). However, there is little clarity on baseline changes in cerebrovascular function in diabetes and the relative contribution of major diabetic stressors to cerebrovascular disease. Our findings suggest that cellular stress from AGEs leads to reduced cerebrovascular secretion of BDNF, a major mediator of neuroprotection.

A large body of evidence suggests that AGEs are central mediators of almost all diabetic vascular complications (31–34), beyond effects of prolonged hyperglycemia. A marked increase in deposition and accumulation of protein glycation end products in blood vessels has been linked to corresponding neuronal impairment, cognitive dysfunction, and dementia in several streams of investigation. In human brain tissue, increased cerebrovascular carboxymethyl lysine-AGE staining was observed in people with diabetes and clinical dementia, strengthening the argument that vascular accumulation of AGEs correspond with clinical presentations of neuronal injury (35). Notably, an accumulation of AGEs in microvessels along with its receptor, RAGE, is reported to drive microvascular AGE-RAGE–mediated chronic inflammation in Alzheimer pathologies (36). In addition, administration of AGE-BSA significantly increased cerebral infarct size in a rat model of focal cerebral ischemia, suggesting that increased severity of stroke associated with diabetes might be characterized by AGE accumulation (37). In our study, treatment of primary brain microvascular cells with AGE-BSA significantly suppressed the production of BDNF via ERK/MAP kinase signaling pathways, whereas treatment with high glucose alone did not (data not shown). Subsequently, because of the loss of BDNF, endothelial cells exposed to AGE-BSA were no longer able to protect neuronal cultures against hypoxic injury. Although our in vitro model of AGE-BSA treatment cannot be equated to in vivo diabetic conditions, it is useful in assessing the direct contribution of AGEs to neurotrophic reduction observed in vivo.

We chose to study BDNF levels at 6 weeks and 6 months of diabetes as a suitable time window to characterize early chronic injury in diabetes. Several studies have focused on this early duration of diabetes to capture adaptive and maladaptive events in diabetes (28). Degeneration of endothelial cells and smooth muscle cells in cerebral cortical arterioles was observable between 14 and 16 weeks in the STZ model of diabetes (38). Increased capillary basement membrane thickening, pericyte degeneration, and decreased cortical capillary density were observed between 4 and 8 months of diabetes in the rats with STZ-induced diabetes (39). A dramatic loss of neocortical neurons coincident with marked shortening of capillary network in the neocortical tissue was observed after 1 year of diabetes in the rat with STZ-induced diabetes by morphometric quantification (40). Taken together, a progressive depletion of microvascular neurotrophic support seems imminent in the diabetic brain. Reduced cerebrovascular BDNF support, as observed in our study, could be a molecular mechanism underlying neuronal dysfunction observed in previous studies. Our media transfer experiments with cortical embryonic neurons only provide proof of concept on how such a deficit might render neurons more vulnerable to a second insult. This cannot be unequivocally extrapolated, as yet, to BDNF dependence in the adult diabetic brain directly. Nevertheless, it is now accepted that BDNF is not only important for neural development, it (or other trophic factors) contributes to adult brain homeostasis as well. BDNF is required for survival of both developing and adult DRG neurons (41). In vivo blockade of trk receptors leads to widespread neuronal death in adult mouse brains (42), and a conditional BDNF knockout mouse shows disrupted striatal neuron survival and function (43). In this context, our data provide a potential mechanism that may underlie the increased neuronal vulnerability of diabetic brains.

Endothelial cells are a major source of neuroprotective BDNF protein in the brain (44–46). In addition, BDNF has unique systemic and metabotrophic functions (47). Our findings also may be consistent with the clinical disease literature. Cerebral output and plasma levels of BDNF are reduced in individuals with type 2 diabetes (48–50). Because these changes occur in an insulin-dependent manner, it suggests that glycemic memory and BDNF status might be linked (49). In our study, we found that BDNF protein levels were reduced in microvessel-enriched fractions of 6-month diabetic rat cortices and that BDNF protein expression was decreased in the diabetic endothelium. Although these findings point to endothelial dysfunction as a mechanism for loss of neuroprotection in the diabetic brain, they remain reductionist, given that other cell types can secrete BDNF in the brain. The existence of compensatory mechanisms in the parenchyma that make up for the microvascular deficit cannot be ruled out. To dissect the singular contribution of the endothelium, targeted silencing of the BDNF gene by conditional knockout experiments or by small interfering RNA strategy will be required. Nevertheless, there is reason to believe that in the deeper cortical layers of the brain, with a high vessel-to-neuron ratio, the endothelia serve as major life-supporting repositories of trophic growth factors.

Our data suggest that BDNF is a leading candidate of cerebrovascular trophic coupling mechanisms. However, the autocrine and paracrine functions of brain endothelium are likely to include many trophic signals. Further investigation is required to dissect the interactions with other neurotrophic factors. The diabetes-induced decrease of BDNF in cerebral microvessels seems to be mediated by AGEs via activation of the ERK signaling pathway. These findings are consistent with known receptor-mediated effects of AGEs; however, nonreceptor-mediated effects of AGEs cannot be ruled out, and more investigation is required to demonstrate a direct involvement of RAGE in this pathway. Although our in vitro model of AGE-BSA treatment cannot be equated to in vivo diabetic conditions, it is useful in assessing the direct contribution of AGEs to neurotrophic reduction observed in vivo.

In conclusion, our study identified a vascular mechanism that may explain why diabetes increases risk for neurologic injury such as stroke. Strategies to preserve neurovascular trophic coupling mechanisms may lead to the development of novel preventive neuroprotective therapy in diabetes.

This work was supported by National Institutes of Health grants R37-NS-37074, R01-NS-56530, RC2-NS-69335, and P01-NS-55014, and a Bugher award and postdoctoral fellowship from the American Heart Association.

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

D.N. designed and conducted experiments, analyzed data, and wrote the manuscript. S.-z.G. provided technical assistance and reviewed the manuscript. K.H. performed and guided immunochemical procedures. X.W. provided diabetic brain sections for study. C.G. provided 6-month diabetic samples, contributed to discussion, and edited the manuscript. E.H.L. supervised research design, analyzed data, and co-authored the manuscript.