Dysregulation of Insulin-Linked Metabolic Pathways in Alzheimer’s Disease: Co-Factor Role of Apolipoprotein E ɛ4

INTRODUCTION

Growing evidence supports the concept that Alz-heimer’s disease (AD) is mechanistically linked to impairments in brain energy metabolism [1] marked by reduced glucose uptake and utilization from pre-symptomatic stages of disease [2]. Furthermore, in the prospective Baltimore Longitudinal Study, impairments in brain glucose uptake were correlated with reduced expression of the glucose transporter 3 (GLUT3) and subsequent development of AD [3]. Since glucose uptake and utilization in the brain and neuronal cells are stimulated by insulin [4–7], insulin deficiency or insulin resistance could dysregulate energy metabolism and contribute to the pathogenesis of AD [8–10]. Besides regulating energy metabolism, insulin stimulates working memory, cognition [11–14], and neuronal plasticity, and its receptors are abundantly expressed in brain regions that are most susceptible to AD-type neurodegeneration [15, 16]. Correspondingly, experimental inhibition of insulin-related signaling networks causes neurodegeneration with AD features [17–19]. Altogether, these findings point to dysregulation of insulin signaling networks as a fundamental mediator of neurodegeneration.

Human postmortem [9, 20] and clinical [10, 21, 22] studies have demonstrated brain insulin deficiency and insulin resistance in AD. Insulin deficiency is mainly manifested by reduced insulin levels in brain and cerebrospinal fluid (CSF), whereas insulin resistance is associated with reduced expression, tyrosine phosphorylation, and binding of the insulin receptor, activation of insulin receptor substrate downstream signaling through phosphoinositol-3-kinase (PI3K)-Akt pathways, and brain glucose levels [9, 16, 20, 23]. Disruption of these insulin-related networks adversely impacts neuronal survival, oligodendrocyte function, neuronal plasticity, and energy metabolism, and promotes neuro-inflammation, oxi-dative, nitrosative, and endoplasmic reticular stress, lipid peroxidation and cell death [8, 24–26]. In addition, impairments in brain insulin signaling have been linked to increased tau phosphorylation and amyloid-β 1–42 (Aβ_1–42) accumulation/toxicity [20, 25, 27–29]. Therefore, apart from the specific AD-associated outcomes, the molecular, biochemical, and cytopathological consequences of insulin deficiency/resistance in the brain closely resemble those that occur in diabetes mellitus and other insulin resistance diseases [8, 30].

The steadily increasing prevalence of AD over the past several decades and across all age groups [31] indicates that factors other than genetics can mediate AD neurodegeneration. Furthermore, the parallel increases in rates of diabetes mellitus and other insulin resistant states, the higher rates of cognitive impairment and AD in people with obesity or type 2 diabetes mellitus [31], and the increased risk of developing mild cognitive impairment (MCI) or AD in non-obese, non-diabetic people with elevated blood glucose [32, 33] suggest the drivers and mechanisms of peripheral insulin resistance and AD may be shared. Another way to consider the problem is that perhaps insulin resistant disease states are fundamentally related but differentially manifested due to variation in tissues, organs and systems targeted. For example, atherosclerosis is a single pathologic process that causes different diseases based on compromised flow through specific arteries. If AD is truly one of the progressive insulin resistance/insulin deficiency diseases in which the brain is selective or prominently involved, then therapeutic interventions developed for other related diseases may be extendable to AD. Already this concept has some validity since cognitive impairment in MCI and AD are positively responsive to intranasal insulin, insulin sensitizers, incretins, and lifestyle modifications that enhance insulin responsiveness [8, 11, 14, 34–38]. Furthermore, intranasal insulin has been shown in humans to increase brain energy, including levels of ATP and phosphocreatine using 31-P magnetic resonance spectroscopy to assess cerebral energy metabolism [39]. To delve deeper into the overarching question concerning insulin network dysfunction versus insulin resistance/deficiency as mediators of neurodegeneration, in this study we measured broad indices of insulin-regulated metabolic integrity in postmortem frontal cortex samples from controls and AD human subjects. The primary objective was to assess the presence and characteristics of central nervous system (CNS) dysregulated metabolic networks.

This study examined the contributions of both Braak stage histopathological grade of AD and apolipoprotein E (APOE) genotype on frontal cortex expression of insulin-related polypeptides in postmortem human brains using a commercial multiplex gut hormone panel. Previously, we used this approach to demonstrate insulin-related metabolic abnormalities in CSF and serum from patients with MCI or AD [40–43]. Furthermore, this study extends earlier work characterizing AD grade and APOE genotype (ɛ3/ɛ3, ɛ3/ɛ4, ɛ4/ɛ4) effects on brain expression of insulin degrading enzyme (IDE) and regulator of calcineurin 1 (RCAN1) [44]. That study demonstrated AD severity and APOE ɛ4 dose-dependent reductions in IDE and increases in RCAN1 expression. Those findings are relevant to the present study because IDE is a 110 kD thiol zinc metalloendopeptidase that degrades insulin and other small polypeptides including atrial natriuretic peptide, transforming growth factor-alpha, amylin, bradykinin, kallidin and Aβ, and RCAN1 inhibits calcineurin causing increased glycogen synthase kinase 3β activation with attendant hyperphosphorylation of tau and neurofibrillary tangle formation. The present work broadens the analysis of dysregulated brain metabolic networks and the co-factor role of APOE4 dose as mediators of neurodegeneration in the pathogenesis of AD.

METHODS

RESULTS

Insulin

Insulin promotes glucose uptake into cells, de-creasing its concentrations in peripheral blood (Table 1). Two-way ANOVA revealed significant effects of Braak stage AD severity (F = 8.78; p =0.0007), APOE genotype (F = 7.27; p = 0.002), and interactions between AD severity and APOE (F =2.83; p = 0.036). Post hoc multiple comparison tests revealed significant reductions in frontal cortex insulin immunoreactivity in B5–6 APOE ɛ3/ɛ3, and both B3–4 and B5–6 APOE ɛ4/ɛ4 relative to B0–2 APOE ɛ3/ɛ3 (Fig. 1A) and B0–2 APOE ɛ3/ɛ4 controls (Table 3). In addition, post hoc testing revealed significant inhibitory effects of APOE ɛ4 dose and AD severity on brain insulin expression, and within B5–6, APOE dose-dependent inhibition of insulin expression (Table 3).

Fig. 1

Insulin and C-Peptide: A commercial magnetic bead-based multiplex ELISA was used to measure A) insulin and B) C-peptide immunoreactivity in 72 human postmortem frontal cortical tissue from patients with established APOE genotypes (APOE ɛ3/ɛ3; APOE ɛ3/ɛ4; APOE ɛ4/ɛ4). Standardized formalin fixed paraffin-embedded histological sections were used to assign Braak stage (B) severities of AD: B0–2 corresponds to normal aging; B3–4 represents moderate AD; B5–6 is severe AD. The graphs depict the mean±S.E.M. levels (pg/mL) of immunoreactivity. Two-way ANOVA with post-hoc Tukey multiple comparison tests were used for intergroup comparisons. ^*p < 0.05; ^**p < 0.01; ^***p < 0.005; ^****p < 0.0001 relative to B0–2 APOE ɛ3/ɛ3 controls. Other significant inter-group differences are provided in Tables 3 and 4.

Gastric inhibitory peptide 1 (GIP-1)

GIP-1 is an incretin with potent stimulatory effects on insulin secretion and a modulator of fatty acid metabolism (Table 1). Two-way ANOVA revealed significant effects of AD Braak stage (F = 8.0; p = 0.001) and APOE genotype (F = 3.94; p = 0.026), but no significant AD×APOE interactions. The major response observed with respect to APOE ɛ3/ɛ3 were progressive declines in GIP-1 with AD severity, resulting in significantly reduced GIP-1 in B5–6 compared with B0–2 (Fig. 2A). For APOE ɛ3/ɛ4, although the trends followed those of APOE ɛ3/ɛ3, the differences between B5–6 and B0–2 did not reach statistical significance. AD, AD severity, and APOE4 dose all significantly inhibited GIP-1 expression relative to APOE ɛ3/ɛ3 and APOE ɛ3/ɛ4 controls (Table 5).

Fig. 2

Incretins (GIP-1 and GLP-1): A multiplex ELISA measured A) GIP and B) GLP-1 immunoreactivity in 72 postmortem human frontal cortex samples from patients with established APOE genotypes (APOE ɛ3/ɛ3; APOE ɛ3/ɛ4; APOE ɛ4/ɛ4). Routine histological studies were used to grade Braak stage (B) severities of AD: B0–2 corresponds to normal aging; B3–4 represents moderate AD; B5–6 is severe AD. Graphs depict the mean±S.E.M. levels (pg/mL) of immunoreactivity. Two-way ANOVA with post-hoc Tukey multiple comparison tests were used for intergroup comparisons. ^**p < 0.01l ^***p < 0.005; ^****p < 0.0001 relative to B0–2 APOE ɛ3/ɛ3 controls. Other significant inter-group differences are provided in Tables 5 and 6.

Leptin

Leptin inhibits food intake and promotes energy expenditure (Table 1). Two-way ANOVA revealed significant effects of AD Braak stage (F = 19.84; p < 0001), APOE (F = 7.19; p = 0.002), and Braak stage×APOE interactions (F = 9.0; p < 0.0001). As shown in Fig. 3A, the mean frontal lobe levels of leptin were significantly reduced in all groups relative to B0–2 APOE ɛ3/ɛ3 (p < 0.0001) (also see Table 7). In addition, within the APOE ɛ3/ɛ3, Leptin expression was lower in B5–6 than B3–4 (p = 0.02) reflecting an AD Braak stage severity effect.

Fig. 3

Leptin and Ghrelin: A multiplex ELISA was used to measure A) leptin and B) ghrelin immunoreactivity in 72 postmortem human frontal cortex samples from patients with known APOE genotypes (APOE ɛ3/ɛ3; APOE ɛ3/ɛ4; APOE ɛ4/ɛ4). Formalin fixed paraffin-embedded histological sections were used to assign Braak stage (B) severities of AD: B0–2 corresponds to normal aging; B3–4 represents moderate AD; B5–6 is severe AD. The graphs depict the mean±S.E.M. levels (pg/mL) of immunoreactivity. Two-way ANOVA with post-hoc Tukey multiple comparison tests were used for intergroup comparisons. ^***p < 0.005; ^****p < 0.0001 relative to B0–2 APOE ɛ3/ɛ3 controls. Other significant inter-group differences are provided in Tables 7 and 8.

Glucagon

Glucagon mediates gluconeogenesis, raising blood glucose levels and opposing the actions of insulin. Two-way ANOVA showed that the main significant effect was APOE genotype (F = 6.3; p = 0.004). Post hoc Tukey tests demonstrated a trend reduction in brain glucagon levels in B5–6 APOE ɛ3/ɛ3 relative to B0–2 APOE ɛ3/ɛ3, and significant reductions in B3–4 and B5–6 APOE ɛ4/ɛ4 relative to B0–2 APOE ɛ3/ɛ3 (p = 0.0007 and p = 0.009, respectively) (Fig. 4A, Table 9). Similarly, brain glucagon expression was lower in B3–4 APOE ɛ4/ɛ4 than in B0–2 APOE ɛ3/ɛ4 (p = 0.009), reflecting an AD effect, and lower in B3–4 and B5–6 APOE ɛ4/ɛ4 than in APOE ɛ3/ɛ3 and APOE ɛ3/ɛ4 with the same Braak stage severities of AD, reflecting an APOE dose effect.

Fig. 4

Glucagon and Resistin: A commercial multiplex ELISA was used to measure A) glucagon and B) resistin immunoreactivity in 72 human postmortem frontal cortex tissue samples from patients with known APOE genotypes (APOE ɛ3/ɛ3; APOE ɛ3/ɛ4; APOE ɛ4/ɛ4). Standardized formalin fixed paraffin-embedded histological sections were used to assign Braak stage (B) severities of AD: B0–2 represents normal aging; B3–4 represents moderate AD; B5–6 is severe AD. The graphs depict the mean±S.E.M. levels (pg/mL) of immunoreactivity. Two-way ANOVA with post-hoc Tukey multiple comparison tests were used for intergroup comparisons. ^*p < 0.05; ^**p < 0.01; ^***p < 0.005; ^****p < 0.0001 relative to B0–2 APOE ɛ3/ɛ3 controls. Other significant inter-group differences are provided in Tables 9 and 10.

Plasminogen activator inhibitor-1 (PAI-1)

PAI-1 is a multi-functional serine protease inhi-bitor that regulates fibrinolysis, and therefore, may be critical for preventing vascular occlusions and ischemic injury (Table 1). Two-way ANOVA demonstrated significant AD Braak stage×APOE effects (F = 4.28; p = 0.005). In contrast to polypeptides that modulate insulin or insulin-related actions, PAI-1 expression in cases with an APOE ɛ3/ɛ3 genotype was elevated in B3–4 (p = 0.01) and B5–6; p = 0.008) relative to B0–2. In addition, the levels in B0–2 and B3–4 with an APOE ɛ3/ɛ4 genotype, and B5–6 with an APOE ɛ4/ɛ4 genotype were elevated compared with B0–2 APOE ɛ3/ɛ3 (Fig. 5A, Table 11), reflecting AD stage and APOE genotype effects. For APOE ɛ3/ɛ3 and APOE ɛ3/ɛ4, PAI-1 expression was similarly elevated by B3–4 and B5–6, whereas for APOE ɛ4/ɛ4, PAI-1 was significantly higher in B5–6 than B3–4, reflecting AD x APOE4 dose interactive effects on PAI-1 expression.

Fig. 5

PAI-1 and Visfatin: A multiplex ELISA platform was used to measure PAI-1 and Visfatin in human postmortem frontal cortex tissue from AD and aged control banked cases. Standardized formalin-fixed, paraffin-embedded histological sections were use to stratify Braak stage (B) severity of AD: B0–2 (normal aging), B3–4 (moderate AD), or B5–6 (severe AD). All 72 cases included in the study had known APOE genotypes (APOE ɛ3/ɛ3; APOE ɛ3/ɛ4; APOE ɛ4/ɛ4). Graphs depict the mean±S.E.M. levels (pg/mL) of immunoreactivity. Two-way ANOVA with post-hoc Tukey tests were used for intergroup comparisons. ^*p < 0.05 and ^**p < 0.01 relative to B0–2 APOE ɛ3/ɛ3 controls. Other significant inter-group differences are provided in Table 11.

DISCUSSION

This study utilized human postmortem brains to examine the independent and interactive relationships between AD severity and APOE4 on frontal lobe expression of insulin-related polypeptides and inflammatory mediators that impact insulin responsiveness. The premise was that, although impairments in insulin expression and signaling have been solidly linked to many aspects of AD pathogenesis and progression, growing evidence suggests that the entire network mediating insulin-related metabolic functions is impaired. The significance of this concept is that more than one target should be considered for effective disease remediation. This work builds on our prior investigations that demonstrated significant and early AD-associated abnormalities in CSF and serum levels of insulin-related polypeptides [40–43], and interactive effects of APOE and AD severity on IDE and RCAN1 expression in the brain [44].

ApoE, a 34 kDa protein, has three major isoforms (e2, e3, e4) that differ by single amino acid substitutions at residues 112 and 158 [45]. The inclusion of APOE genotype as a variable in these analyses was prompted by the facts that: 1) the ɛ4 allele is the strongest genetic risk factor for late onset sporadic AD [46]; 2) ɛ4 carriers account for over 50% of AD cases [47]; 3) AD risk is increased by 3- or 4-fold among ɛ4 carriers and 15-fold in homozygous individuals (ɛ4/ɛ4); and 4) besides its role in Aβ accumulation, APOE4 is associated with reduced brain glucose metabolism in the pre-clinical stages of disease [48], ultimately impairing signal transduction through the insulin receptor with attendant reduction of Aβ clearance and increased Aβ aggregation [49]. In addition, APOE ɛ4’s contribution to AD pathogenesis is likely linked to the associated impairments in synaptic plasticity mediated by alterations in endocytic recycling [50]. The mechanism may involve the inhibition of Reelin signaling through APOE and attendant regulation of N-methyl-D-aspartate (NMDA) receptor activity, stabilization of microtubules, and prevention of neurofibrillary tangle-associated tau hyperphosphorylation [51].

Although it is not known how APOE4 impairs insulin signaling, the mechanism may involve alterations in cholesterol homeostasis since APOE is an important regulator of cholesterol synthesis and receptor-mediated uptake [52], and previous studies showed that alterations in neuronal membrane cholesterol content lead to impaired insulin signaling and glucose uptake [53]. Correspondingly, experimental over-loading of neuronal membranes with cholesterol increases amyloid-β protein precursor (AβPP) internalization and Aβ generation [54]. In addition to its roles in lipoprotein metabolism, synaptic maintenance, and clearance of Aβ, the low-density lipoprotein receptor-related protein 1 (LRP1) ApoE receptor interacts with the insulin receptor beta subunit and regulates insulin-mediated signaling and glucose uptake [55]. Therefore, APOE4-associated reduction in brain LRP1 expression [56] is an important potential mechanistic link between APOE4 genotype and brain insulin resistance in AD. Furthermore, the finding that hyperglycemia also suppresses LRP1 expression and further exacerbates insulin resistance, glucose intolerance, and AD pathology [55] helps link diabetes mellitus to AD-type neurodegeneration [8, 57]. Due to its adverse effects on energy metabolism and insulin signaling [58], we hypothesized that APOE4 exacerbates AD-associated impairments in insulin-regulated functions in a dose-dependent manner.

This study included postmortem brain specimens from 72 men and women with histopathological diagnoses of normal aging or very early AD (B0–2), moderate AD (B3–4), or severe AD (B5–6), with an APOE ɛ3/ɛ3, ɛ3/ɛ4, or ɛ4/ɛ4 genotype. There were no normal aged brains from APOE ɛ4/ɛ4 cases, perhaps due to the strong propensity for such individuals to have developed AD by the time they reach 70–90 years of age. It is also noteworthy that a skewed percentage of the APOE ɛ4/ɛ4 cases had B5–6 AD whereas those with an APOE ɛ3/ɛ3 or ɛ3/ɛ4 genotype predominantly (68% or 70%) had B3–4 stage AD. One limitation of this study was the lack or APOE ɛ4/ɛ4 controls for comparison with AD cases of the same genotype. Nonetheless, using two-way ANOVA tests with post hoc comparisons it was possible to ascertain APOE ɛ4 dose effect in relation to AD severity.

Insulin decreases blood glucose, increases cell permeability to monosaccharides, and accelerates glycolysis, the pentose phosphate cycle, and glycogen synthesis. The decline in brain insulin with increasing severity of AD in the APOE ɛ3/ɛ3 group corresponds with previous observations made without knowledge of APOE status [16, 20, 41]. The absence of an APOE ɛ3/ɛ4 effect suggests that APOE4 carrier status does not adversely impact brain insulin expression. In contrast, homozygous APOE ɛ4/ɛ4 was associated with significantly reduced brain insulin expression such that the mean levels in B3–4 and B5–6 were comparable to that measured in B5–6 APOE ɛ3/ɛ3. Therefore, APOE ɛ4/ɛ4 was associated with earlier Braak stage reductions in brain insulin expression, suggesting that a contribution to AD neurodegeneration is exacerbation of brain metabolic dysfunction via insulin deficiency.

Although it has been suggested that AD-associated reductions in brain insulin could be attributed to increased expression of IDE [59], a previous analysis of these cases revealed primarily reduced expression of IDE mRNA and protein [44]. Alternatively, impairments in local synthesis of insulin [16, 20] or its uptake from the peripheral circulation [60] could account for the reduced CNS levels of insulin with AD progression. One argument in favor of AD and APOE ɛ4-mediated impairments in local insulin synthesis was the finding of reduced C-peptide expression in patterns that mimicked changes in insulin. C-peptide is a cleavage product of pro-insulin and generated with insulin. Its longer half-life provides a more reliable in vivo index of insulin production.

Incretins, including GIP-1 and GLP-1, are potent stimulators of glucose-dependent insulin secretion and modulators of fatty acid metabolism [61]. GLP-1 suppresses plasma glucagon, stimulates glucose disposal, and has neuroprotective actions that may benefit individuals with AD [62, 63]. In the APOE ɛ3/ɛ3 cases, GIP expression was reduced with AD severity whereas GLP-1 was not, but with an APOE ɛ4/ɛ4, frontal lobe GIP and GLP-1 were similarly reduced in B3–4 and B5–6 relative to B0–2 controls. Again, the presence of a single ɛ4 allele was not associated with significantly altered incretin expression in either control or AD cases, supporting the concept that APOE ɛ4/ɛ4 is the principal driver of brain metabolic dysfunction, and that its adverse effects include impairments in brain incretin expression and function.

Leptin and ghrelin have opposing effects in that leptin regulates fat depots by inhibiting food intake and regulating energy expenditure while ghrelin stimulates appetite and induces adiposity in addition to inducing growth hormone release from the pituitary gland [64]. This study demonstrated that frontal lobe leptin was strikingly reduced in moderate and severe AD regardless of APOE genotype, as well as in APOE ɛ3/ɛ4 controls. Downregulation of leptin may reflect an adaptive response to insulin resistance since, although leptin inhibits tyrosine phosphorylation of the insulin receptor substrate 1 (IRS-1) protein and interactions between IRS-1 and growth factor receptor-bound protein 2 (GRB2), its positive effects include stimulation of PI3K, which promotes metabolism, cell survival, neuronal plasticity, and other critical neuronal and glial brain functions [65].

Ghrelin is an endogenous ligand of the growth hormone secretagogue receptor 1a that regulates growth hormone secretion and energy balance. Ghrelin is expressed in the brain, including hypothalamus and besides its physiological role in glucose and lipid metabolism, ghrelin has functional roles in memory, reward, sleep, and neurogenesis [66], which are known to be disrupted in AD [67]. Therefore, reductions in ghrelin, either in the CNS or periphery, could reflect a pathophysiological state that mediates neurobehavioral features of neurodegeneration. Furthermore, evidence suggests that ghrelin has an important role in neuroprotection and may help remediate cognitive dysfunction and neurodegeneration [68]. This study showed no significant alteration of ghrelin expression in relation to AD in APOE ɛ3/ɛ3 or ɛ3/ɛ4 cases. However, ghrelin was significantly reduced in both B3–4 (moderate) and B5–6 (severe) APOE ɛ4/ɛ4 AD. This observation is novel and could reflect significantly impaired neuroprotective mechanisms and metabolic dysfunction linked to APOE ɛ4 dose.

Glucagon increases gluconeogenesis and decrea-ses glycolysis, raising plasma glucose in response to insulin-induced hypoglycemia and likely plays an important role in initiating and maintaining hyperglycemic conditions in diabetes mellitus. Only the APOE ɛ4/ɛ4 AD had significantly reduced frontal lobe glucagon expression. This response could not be attributed to GLP-1 suppression since GLP-1 expression was also downregulated in B3–4 and B5–6 APOE ɛ4/ɛ4. One potential explanation for the finding is that GLP-1/glucagon regulation may be uncoupled. A chronic brain diabetic state (Type 3 diabetes) [16, 25, 69] together with impaired capacity for glucose and lipid utilization could potentially lead to brain starvation.

The adipokine, resistin, promotes insulin resistance [64] and activates pro-inflammatory cytokines [70]. Correspondingly, central deficiency of resistance was found to ameliorate hypothalamic inflammation and increase whole body insulin sensitivity [71]. Elevated resistin in APOE ɛ3/ɛ4 B3–4 relative to all other groups except APOE ɛ3/ɛ4 controls suggests that interactive effects of a single APOE4 allele and early or intermediate-stage AD contribute to brain insulin resistance. In contrast, reductions in brain resistin expression with increasing severity of AD among APOE ɛ3/ɛ3 cases, as well as the lower levels in more advanced stages of disease in APOE ɛ3/ɛ4 or APOE ɛ4/ɛ4 cases may reflect compensatory or adaptive responses, reducing neuroinflammation and increasing insulin sensitivity to support brain energy balance.

Visfatin is also an adipokine but, contrary to the effects of resistin, it increases insulin sensitivity [64], but like resistin, it activates pro-inflammatory cytokines [70]. Visfatin mediates its stimulatory ef-fects on insulin responsiveness via activation of the insulin receptor with attendant lowering of blood glucose [70]. Visfatin, also known as ‘nicotinamide phosphoribosyltransferase’ (NAmPRTase or Nampt), is the rate-limiting enzyme in the nicotinamide adenine dinucleotide (NAD+) salvage pathway that converts nicotinamide to nicotinamide mononucleotide and enables NAD+ biosynthesis [70]. Visfatin-mediated increases in NAD+ have anti-aging and neuroprotective actions [72–78]. Although there were no significant effects of AD on visfatin expression among APOE ɛ3/ɛ3 cases, in both APOE ɛ3/ɛ4 and APOE ɛ4/ɛ4, the significant linear trend effects reflect increasing levels with AD severity. Therefore, in addition to the interactive effects of APOE ɛ4 dose and AD grade, the results suggest that compensatory mechanisms may help preserve glucose utilization and energy metabolism vis-à-vis brain insulin resistance and deficiency.

PAI-1 is a serine protease inhibitor of tissue plas-minogen activator, urokinase, Protein C, and matrip-tase-3/TMPRSS7, and functionally inhibits fibrinolysis [79]. Elevated levels of PAI-1 are strongly associated with insulin resistance, including in diabetes mellitus [80, 81], and have been deemed independent risk factors for type 2 diabetes [81]. Elevated PAI-1 is associated with increased risk for thrombotic cardiovascular [82] and cerebrovascular [79] diseases and ischemic stroke [83]. In this study we observed significantly increased PAI-1 expression in AD relative to control in APOE ɛ3/ɛ3 cases, but similarly elevated PAI-1 in B0–1, B3–4, and B5–6 APOE ɛ3/ɛ4, indicating independent effects of AD severity and APOE /ɛ4. The similarly elevated PAI-1 levels in B5–6 APOE ɛ4/ɛ4 compared with B3–4 or B5–6 APOE ɛ3/ɛ3 and APOE ɛ3/ɛ4 cases indicates that in this instance, APOE ɛ4 dosage did not additively affect PAI-1 expression, and instead, APOE and AD independent upregulated PAI-1. Mechanistically, pro-inflammatory cytokine activation has been linked to increased PAI-1 expression [80], and in earlier reports we showed increased pro-inflammatory and reduced anti-inflammatory cytokine/chemokine levels in AD [40, 42].

CONFLICT OF INTEREST

The authors have no conflict of interest to report.