Alzheimer-Like Changes in Rat Models of Spontaneous Diabetes

Prediabetic male type 1 BB/Wor (n = 10) and type 2 BBZDR/Wor rats (n = 10) and sex- and age-matched non–diabetes-prone BB-control rats (n = 10) were obtained from Biomedical Research Models (Worcester, MA). All diabetic animals were maintained in metabolic cages and cared for in accordance with the guidelines of the Wayne State University Animal Investigation Committee and those of the National Institutes of Health (publication #85-23, 1995). Urine volume and glucose levels were measured daily to ascertain the onset of diabetes and to titrate daily insulin doses for type 1 BB/Wor rats. Blood glucose levels were measured weekly using blood glucose test strips (Bayer, Mishawaka, IN). Diabetes developed spontaneously in BB/Wor rats at 73 ± 2 days of age and in BBZDR/Wor rats at 76 ± 4 days. Type 1 diabetic rats were treated with daily titrated doses (0.5–3.0 IU) of protamine zinc insulin (Blue Ridge Pharmaceuticals, Greensboro, NC) to maintain blood glucose levels between 20 and 25 mmol/l and to prevent ketoacidosis (10). At the time of death (18 h after last insulin injection of type 1 diabetic rats), serum insulin and IGF-1 concentrations were analyzed using radioimmunoassay kits (Linco Research, St. Charles, MO), GHb was measured using a DCA 2000 Analyzer (Bayer, Elkhart, IN). Cholesterol levels were obtained (Biomedical Research Models, Worcester, MA). Both type 1 diabetic BB/Wor and type 2 diabetic BBZDR/Wor rats were killed after 8 months of diabetes, i.e., ∼10.5 months of age.

Four animals per group were prepared for cortical neuronal counts and immunocytochemical studies. Rats were anesthetized with an intraperitoneal overdose of sodium pentobarbital (120 mg/kg body wt). Each rat was perfused transcardially through the left ventricle with 300 ml PBS (pH 7.4), followed by 4% paraformaldehyde in PBS (pH 7.4). After perfusion, the whole brain was dissected and placed in the same fixative overnight.

For protein isolation, six animals per group were used. The animals were anesthetized with isoflurane and decapitated. The brains were rapidly removed and placed on an ice-cooled cutting board. The meninges were removed and the hemispheres separated. The bilateral frontal cortices between Bregma 5.2 and 3.2 mm (25) were dissected coronally, snap frozen in liquid nitrogen, and stored at −70°C.

Coronal segments of the frontal cortex between Bregma 5.2 and 3.2 mm were embedded in paraffin. Ten 6-μm-thick paraffin sections 90 μm apart were deparaffinized, rehydrated in distilled water, and stained with FD cresyl violet solution (FD Neuro Technologies, Baltimore, MD). After washing in distilled water three times, sections were differentiated in 95% ethanol containing 0.1% glacial acetic acid for 1 min and dehydrated in 100% ethanol followed by clearing in xylene. Images of perpendicularly sectioned cortex of ∼1.5 mm in length were captured and analyzed using an Image-Pro Plus 3.0 image analysis software (Media Cybemetics, Silver Spring, MD). The cortex was separated into layers I, II–III, IV, V, VIa, and VIb (25). Only neurons displaying visible nuclei were counted, and their numbers in each layer were expressed per area unit (n per millimeters squared).

Deparaffinized 6-μm sections of the frontal cortex were incubated with the primary antibody for 30 min at room temperature, washed with three changes of PBS, and incubated with the peroxidase-conjugated secondary antibody (Vector Lab., Burlingame, CA) for 30 min at room temperature. The immunoreactive products were visualized with diaminobenzidine as color chromogen. Counterstaining was performed with hematoxylin to show nuclear staining. Primary antibodies used were: mouse anti-BACE1 antibody (Chemicon, Temecula, CA), rabbit anti–β-APP antibody (Zymed Lab., South San Francisco, CA), mouse anti-sAPP (A4, aa 68–81 of APP; Chemicon), rabbit anti–COOH-terminal fragment (CTF) (C-term 751-770; Calbiochem, LoJolla, CA), rabbit anti–phospho-Akt (Thr308) and anti–phospho–glycogen synthase kinase 3-β (GSK-3β) (Ser9) antibodies (Cell Signaling, Danvers, MA), mouse anti-synaptophysin and rabbit anti–glial fibrillary acid protein (GFAP) antibodies (Sigma-Aldrich, St. Louis, MO), monoclonal mouse anti-phosphorylated neurotilament M and neurotilament N (SMI-31; Covance, Berkley, CA), rabbit anti–IGF-I receptor and insulin receptor β-subunit (IR_β) antibodies (Santa Cruz Biotechnology, Santa Cruz, CA), monoclonal mouse anti–β-amyloid antibody 4G8 (against β-amyloid aa 18–22[VFFAE]; Convance), and rabbit anti-phospho τ (phospho-τ) antibody (PS396) (gift from Dr. Koichi Ishiguro, Mitsubishi Kagaku Institute of Life Sciences, Tokyo, Japan). Additional immunoflurosence staining was used for synaptophysin. Sections were deparaffinized, rehydrated, and blocked with PBS containing 10% goat serum at room temperature for 30 min and incubated with the primary antibody for 1 h. After washing three times with PBS, sections were incubated with the Alexa Fluor 499 goat anti-mouse IgG (Invitrogen, Carlsbad, CA) as secondary antibody. The fluorescent immunoreactive products were visualized under a Leica DMLB microscope (Leica Microsystems Wetzlar, Wetzlar, Germany).

Negative controls were performed by omitting the primary antibody. To further ensure specific labeling, secondary antibodies (6 μl/ml) of anti-mouse IgG or anti-rabbit IgG were incubated with 4% normal rat serum in room temperature for 30 min and then applied to the tissue slides (26). Comparisons with nonpreabsorbed secondary antibodies revealed the same results (data not shown). In a second set of control experiments, isotype experiments were performed in which the primary monoclonal antibodies (IgG₁ and IgG_2b) were replaced by nonimmune IgG₁ and IgG_2b control immunoglobulins (Santa Cruz Biotechnology) (27). Tissue slides treated accordingly showed no immunoreactivity (data not shown).

Cortical sections immunostained with synaptophysin were examined densitometrically. The intensity range of immunostaining was assessed by a gray scale paradigm using Image-Pro image analysis (Media Cybemetics). Four randomly chosen areas of cortical layers II–III from each of 10 sections from three animals per group were examined. Sections immunostained with SMI-31 were used to examine the density of dystrophic neurites. Eight areas (each measuring 0.284 mm²) chosen as above were assessed in each of three animals. The number of dystrophic neurites per area was calculated and expressed as number per millimeters squared. The density of GFAP-positive astrocyte somata was calculated in the same fashion and expressed as number per millimeters squared.

Dissected frontal cortices between Bregma 5.2 and 3.2 mm from six animals were used. Protein lysates (40 μg/lane) were resolved by SDS-PAGE electrophoresis under reducing conditions and transferred to polyvinylidene difluoride membranes (Millipore, Bedford, MA) or nitrocellulose membranes (Bio-Rad, Hercules, CA). Membranes were stained with 1% Ponseau S or the expression of actin was used to ensure equal loading. The membranes were blocked with 5% BSA in Tris-buffered saline with Tween (TBST) at room temperature for 1 h and incubated with the primary antibody in the blocking solution for 1 h at room temperature. The antibodies used were the same as those used for immunocytochemistry. The blots were washed three times with TBST and incubated for 1 h with the horseradish peroxidase–conjugated secondary antibody (Santa Cruz Biotechnology), followed by an enhanced chemiluminescence detection system (Amersham Pharmacia Biotech, Piscataway, NJ) and exposed to Kodak X-OMAT blue film (Eastman Kodak, Rochester, NY). The immunoblots were quantitated using Scion Image analysis software (Scion, Frederick, MD).

All data are reported as means ± SD. Comparisons were made using ANOVA. When a P < 0.05 was obtained, group comparisons were made using Scheffes post hoc test. Significance was defined as a P value <0.05. All analyses were performed by personnel being unaware of the animal identities.

After 8 months of diabetes, type 2 diabetic BBZDR/Wor rats were 24% heavier (P < 0.05) than control rats, whereas the weight of type 1 diabetic BB/Wor rats was 77% (P < 0.001) of that of control animals. Both type 1 and type 2 diabetic rats showed significantly (P < 0.001) elevated blood glucose levels and GHb values with no differences between the two diabetic models (Table 1). Plasma insulin levels were markedly decreased in type 1 (P < 0.001) and elevated (P < 0.05) in type 2 diabetic rats. Plasma IGF levels were significantly and similarly decreased (P < 0.001) in both models. Cholesterol levels were elevated in type 2 diabetic BBZDR/Wor rats (P < 0.001) but not in their type 1 counterparts.

Cell counts of neuronal populations in layers I–VIb of the frontal cortex showed significantly decreased densities in all layers in type 2 diabetic BBZDR/Wor rats (P < 0.05 or less), being most severe in layers II–III and IV (P < 0.005) (Fig. 1A), in which the densities were decreased by 40–50%. The neuronal loss in these layers was significantly less (P < 0.005 and P < 0.01, respectively) in type 1 diabetic BB/Wor rats, although these animals also demonstrated significant losses (P < 0.05 or less) in all layers except for layer I (Fig. 1A). The neuronal loss was accompanied by a noticeable astrogliosis as illustrated by GFAP-positive astrocytes. Astrocyte density was increased sixfold in BBZDR/Wor rats (P < 0.001), which was significantly (P < 0.001) greater than the 3.7-fold increase (P < 0.01) in BB/Wor rats (Fig. 1B). Densitometric evaluation of synaptophysin positivity, reflecting synapse density, was decreased by ∼35% in both type 1 and type 2 diabetic rats (both P < 0.001) (Fig. 1C). Degenerated dystrophic neurites visualized by immunostaining for phosphorylated neurofilaments (Fig. 1D) were increased 8.6-fold (P < 0.001) in BBZDR/Wor rats, which was significantly (P < 0.001) greater than the 2.7-fold increase (P < 0.01) in BB/Wor rats (Fig. 1D). Dystrophic neurites occasionally occurred in clusters (Fig. 1E), which were not associated with extracellular deposits of β-amyloid, or as single dystrophic neurites (Fig. 1F).

The protein expression of APP, β-secretase, and β-amyloid was increased in diabetic rats compared with age-matched control rats (Fig. 2A–C). These upregulations were significantly (P < 0.05 or less) more pronounced in type 2 diabetic BBZDR/Wor-rats than in the type 1 diabetic BB/Wor counterparts. The more than twofold increases in the expression of these proteins in BBZDR/Wor rats corresponded immunocytochemically to a more intense staining of APP, β-secretase, and β-amyloid in frontal cortical neurons compared with control rats (Fig. 3A–C). To maximize the capture of β-amyloid, nitrocellulose membranes (0.2 μm) were used and protein lysates were boiled before loading. However, monomeric β-amyloid (4 kd) could not be detected. Instead, the multimeric form of β-amyloid was detected between 6.5 and 1.0 kd (Fig. 2C). In type 2 diabetic rats, most neurons showed intense APP staining, whereas control rats showed less intense staining. The staining pattern in type 1 diabetic rats was intermediate between that of type 2 diabetic and control rats. Similar differences between the three groups of animals were displayed by positive β-secretase staining (Fig. 3B). APP and β-secretase often showed a punctate positive staining of the neurolemma (Fig. 3A, B, and D). β-Amyloid–positive staining was weak in the cytoplasm of control neurons, a staining pattern that was increased in type 1 and more so in type 2 diabetic rats (Fig. 3C). In none of the animals could extracellular deposits of β-amyloid be demonstrated. The expressions of soluable APP (Fig. 2D) and CTF (Fig. 2E) were minimally increased in type 1 diabetic BB/Wor rats but showed a more than twofold increase (P < 0.01 and P < 0.005, respectively) in type 2 BBZDR/Wor rats.

Western blotting for phospho-τ demonstrated several bands corresponding to the multiple phospho-τ isoforms (Fig. 4A). Phospho-τ was significantly more expressed in type 2 diabetic BBZDR/Wor rats than in control (P < 0.01) and BB/Wor rats (P < 0.05) (Fig. 4A), which corresponded to a more intense immunostaining of phospho-τ in cortical neurons in type 2 diabetic rats (Fig. 4B) associated with a punctate staining pattern in the neuropil. The neuronal staining pattern was less evident in type 1 diabetic rats in whom the expression of phospho-τ was not significantly increased. Only faint positivity was evident in age-matched control rats (Fig. 4B). Intracytoplasmic phospho-τ–positive tangle-like inclusions were not observed.

The expression of the insulin receptor β-subunit was significantly (P < 0.01) decreased in the frontal cortex of type 1 diabetic BB/Wor but not in type 2 diabetic BBZDR/Wor rats (Fig. 5A). On the other hand, the expression of the IGF-1 receptor β-subunit was equally suppressed in type 1 and type 2 diabetic rats (P < 0.05) (Fig. 5B). In the insulin signaling pathway, phosphorylation of protein kinase B (Akt) is important for GSK-3β (Ser 9) phosphorylation, which suppresses GSK-3β, an important phosphokinase implicated in τ phosphorylation (28). In the present study, the expression of phosphorylated Akt was significantly (P < 0.01) decreased in frontal cortex of both type 1 and type 2 diabetic rats (Fig. 5C), which was accompanied by an ∼50% reduction (P < 0.05) in phosphorylated GSK-3β (Ser 9) expression in both diabetic models (Fig. 5D). These findings are in keeping with the notion that disinhibition of phosphorylated GSK-3β by GSK-3β (Ser 9) is in part responsible for τ phosphorylation (28).

Alzheimer's disease is characterized by an increased β-amyloid load, accumulation of hyperphosphorylated τ, neurite degeneration, and neuronal loss (19,21,22). In this study, we demonstrate significant loss of frontal cortical neurons and neurite degeneration in chronically diabetic rats, accompanied by increased expression of APP, its cleavage enzyme β-secretase, and increased sAPP and β-amyloid burden as well as CTR, abnormalities that were more expressed in type 2 diabetic BBZDR/Wor rats. Phosphorylation of cleaved τ was markedly increased in type 2 diabetic rats. In this model, phosphor-τ was associated with suppressed expression of insulin signaling intermediaries, despite the fact that the insulin receptor expression was unaltered, although that of the IGF-1 receptor was downregulated. This constellation may reflect increased caveolin-1 activity in cholesterol-rich lipid rafts that enhances the expression of insulin receptor, whereas other growth factor receptors, including that of IGF-1, are suppressed (29,30). The more severe neurite degenerative changes in the type 2 diabetic BBZDR/Wor rat is in keeping with the toxic effects of altered β-amyloid and phospho-τ in this model. These findings indicate that experimental diabetes, particularly type 2 diabetes, develops abnormalities that characterize early Alzheimer's disease.

The differences in the severity between the two types of diabetes cannot be related to differences in hyperglycemic exposures, since these were the same in the two models. In the type 2 diabetic rats, the insulin receptor expression was normal, despite that they demonstrated impaired insulin signaling activities, indicating that the CNS also develops insulin resistance downstream from the receptor and hence impaired insulin signaling and action, as do peripheral tissues in this model (24). This is the first evidence demonstrating perturbed CNS insulin signaling in type 2 experimental diabetes, although diet-induced insulin resistance in the CNS has been shown in transgenic Tg 2576 mice, which model Alzheimer's disease (31). Impaired insulin signaling has also been described in hippocampus after intraventricular injection of streptozotocin (32). Similar abnormalities in insulin signaling intermediaries in type 1 diabetic rats are probably related primarily to suppressed expression of the insulin receptor itself and hence different from the insulin resistance in type 2 diabetic animals. The differences between the two models are similar to those previously demonstrated in hippocampus (33). Despite the fact that insulin signaling appeared to be affected to similar extents in the two models, although for different reasons, the phospho-τ abnormalities were significantly more expressed in the type 2 model, suggesting that additional factors must be considered. Other so-called “τ-kinases” include cyclin-dependent kinases (cdk 5), mitogen-activated protein kinase family members, extracellular regulated kinase 1/2, and p38 and c-jun NH₂-terminal kinase (34). The negative regulation of τ phosphorylation by O-G1cNAcylation due to deficient brain glucose uptake has been invoked in Alzheimer's disease (35). A similar “yin-yang” relationship between O-GlcNAc and phosphorylation exists with respect to nodal ankyrin_G in diabetic rats (36). Since the expression of the IGF-1 receptor, another factor implicated in the pathogenesis of Alzheimer's disease (12), was affected equally in the two models, this is not likely to account for the discrepancies in the expression of Alzheimer's disease–like changes.

Both clinical and experimental studies suggest that overweight is a contributing risk factor in Alzheimer's disease. This linkage is believed to be mediated by insulin resistance and/or dysfunction of the hypothalamic-pituitary-adrenal axis (31,37,38). Although corticosterone measurements were not performed in this study, the type 2 diabetic BBZDR/Wor rats examined here were obese and insulin resistant. Conversely, caloric restriction and leaner body weight, as in the BB/Wor rats, have a protective effect on Alzheimer's disease–like changes in transgenic mouse models (39,40). Therefore, differences in body weight between the two models used here could contribute to the differences in the expression of Alzheimer's disease–like changes.

Hypercholesterolemia has been associated with Alzheimer's disease with formation of caveolae and lipid rafts, invaginations of plasma membranes enriched in cholesterol and sphingolipids (41). Cholesterol levels were significantly elevated in type 2 but not in type 1 diabetic rats. Associated caveolin signaling has been postulated to influence the activity of several growth factor receptors including those of insulin and IGF-1 (42). Furthermore, the formation of lipid rafts are coupled with upregulation of APP and β-secretase (43,44). The punctate stainability of APP and β-amyloid on the neurolemma as demonstrated here could potentially represent their concentration to caveolae. APP is also upregulated by hyperglycemia under hypoxic conditions (9) and by perturbed insulin signaling (13). Cholesterol loading enhances β- and γ-secretase activities, which result in enhanced release of amyloidogenic β-amyloid (44), as β-amyloid is generated by the sequential cleavage of APP by β- and γ-secretase.

Therefore, hypercholesterolemia, present only in type 2 diabetic BBZDR/Wor rats, may in part, through formation of caveolae and lipid rafts, explain the unaltered expression of the insulin receptor and the downregulation of the IGF-1 receptor via caveolin action as well as the significantly greater upregulation of APP and β-secretase, resulting in a greater β-amyloid burden in the type 2 diabetic animals. Despite the fact that β-amyloid was markedly increased in BBZDR/Wor rats, no extracellular deposits of β-amyloid could be detected. It has been demonstrated that Wortmannin, which blocks insulin signaling, hampers the release of soluable APP and β-amyloid from the intra- to the extracellular compartment (45,46), which would be consistent with the present findings of increased expression of soluable APP in BBZDR/Wor rats but not in BB/Wor rats. The abnormal APP metabolism in the type 1 diabetic BB/Wor rat could possibly be related to hyperglycemia and impaired cerebral blood flow (9), with elevated expression of inflammatory mediators such as cytokine transforming growth factor-β1 and interleukin-1β, which promote amyloidogenicity (47).

Activation of several caspases by β-amyloid has been linked to the cleavage of τ, hence linking amyloid deposition and neurofibrillary tangles in Alzheimer's disease (48). We have previously demonstrated significant apoptotic stresses affecting hippocampal neurons in the same diabetic models (33). Isoforms of the τ protein are highly susceptible to phosphorylation by several serine/threonine kinases such as mitogen-activated protein kinase, cdk5, and GSK-3β (32,34,47,48), of which the latter appears to play a prominent role (28,32,48,49). GSK-3β (Ser 9) is phosphorylated in response to insulin and IGF via phospho-Akt downstream in the phosphatidylinositol 3 kinase pathway and inhibits GSK-3β. Here we demonstrated impaired phospho-Akt and GSK-3β (Ser 9), indirectly suggesting disinhibition of GSK-3β, which is a prime candidate for hyperphosphorylation of cleaved τ (28,34,49,50), as demonstrated here in type 2 diabetic rats. However, the relative absence of phospho-τ in type 1 diabetic rats suggest that additional τ-kinases or other factors are operable in type 2 diabetic rats (33,35), since the suppression of phosphoAkt and GSK-3β (Ser 9) was similar in the two models. These possibilities are now actively being pursued in our laboratory.

In summary, this study has demonstrated neuronal loss, neurite degeneration accompanied by perturbations of APP metabolism, hyperphosphorylation of τ isoforms, and impaired signaling of insulin and IGF-1 in two models of diabetes. The changes were more severe in the type 2 model and mirror those characterizing Alzheimer's disease. Although several potential underlying mechanisms remain to be explored, particularly the role of hypercholesterolemia and caveolin signaling, the present study suggests that mechanistic linkages do exist between spontaneous onset of experimental diabetes and Alzheimer's disease–like changes. We conclude that particularly type 2 experimental diabetes replicates the main abnormalities in Alzheimer's disease, such as abnormal amyloid handling, hyperphosphorylation of τ, neurite degeneration, and neuronal loss.

This research was supported in part by the Thomas Foundation, Bloomfield Hills, Michigan (to A.A.F.S.); the Juvenile Diabetes Research Foundation, New York, New York (to A.A.F.S.), and the Morris Hood Diabetes Center, Detroit, Michigan (to Z.-G.L.).