|Home | About | Journals | Submit | Contact Us | Français|
Obesity, type 2 diabetes mellitus (T2DM), and non-alcoholic steatohepatitis (NASH) can be complicated by cognitive impairment and neurodegeneration. Experimentally, high fat diet (HFD)-induced obesity with T2DM causes mild neurodegeneration with brain insulin resistance. Since ceramides are neurotoxic, cause insulin resistance, and are increased in T2DM, we investigated the potential role of ceramides as mediators of neurodegeneration in the HFD obesity/T2DM model. We pair-fed C57BL/6 mice with a HFD or control diet for 4–20 weeks and examined pro-ceramide gene expression in liver and brain and neurodegeneration in the temporal lobe. HFD feeding gradually increased body weight, but after 16 weeks, liver weight surged (P < 0.001) due to lipid (triglyceride) accumulation (P < 0.001), and brain weight declined (P < 0.0001-Trend analysis). HFD feeding increased ceramide synthase, serine palmitoyl transferase, and sphingomyelinase expression in liver (P < 0.05 – P < 0.001), but not brain. In HFD fed mice, temporal lobe levels of ubiquitin (P < 0.001) and 4-hydroxynonenal (P < 0.05 or P < 0.01) increased, and tau, β-actin, and choline acetyltransferase levels decreased (P < 0.05 – P < 0.001) with development of NASH. In obesity, T2DM, or NASH, neurodegeneration with brain insulin resistance may be mediated by excess hepatic production of neurotoxic ceramides that readily cross the blood-brain barrier.
The prevalence rates of Alzheimer’s disease (AD), obesity, type 2 diabetes mellitus (T2DM), and nonalcoholic fatty liver disease (NAFLD)/non alcoholic steatohepatitis (NASH), which includes metabolic syndrome, have all increased dramatically over the past several decades [1–6]. The probable inter-relatedness among these diseases is suggested by studies demonstrating: 1) increased risk of developing mild cognitive impairment (MCI), dementia, or AD in individuals with T2DM [7,8]or obesity/dyslipidemic disorders ; 2) progressive brain insulin resistance and insulin deficiency in AD [10–13]; 3) cognitive impairment in experimental animal models of T2DM and/or obesity [14, 15]; 4) AD-type neurodegeneration and cognitive impairment in experimentally induced brain insulin resistance and insulin deficiency [16–20]; 5) improved cognitive performance in experimental models of AD  and in human subjects with AD or MCI after treatment with insulin sensitizer agents or intranasal insulin [22–27]; and 6) similar molecular, biochemical, and mechanistic abnormalities in T2DM, NASH, and AD [7,28–33]. Since obesity,MCI, AD, T2DM, and NASH are all associated with insulin resistance, i.e., impaired ability to respond to insulin stimulation, they may share common etiologies. On the other hand, the lack of complete overlap among these disease states suggests that specific organ systems may be differentially afflicted by the same or similar exposures, resulting in dissimilar degrees of insulin resistance with disparate long-term outcomes, e.g., neurodegeneration versus NASH.
While aging is clearly the strongest risk factor for AD, emerging data suggest that T2DM and dyslidemic states can either contribute to, or serve as co-factors in its pathogenesis . This concept is supported by epidemiologic data demonstrating a significant association between T2DM and MCI or dementia, and that T2DM is a significant risk factor for developingAD [7,35–39]. Mechanistically, increased risk of dementia in T2DM and obesity could be linked to chronic hyperglycemia, insulin resistance, oxidative stress, accumulation of advanced glycation end-products, increased production of pro-inflammatory cytokines, and/or microvascular disease . However, most of these features define the core abnormalities in T2DM and NASH [40–43]. This concept led us to hypothesize that toxic/injurious agents produced in the body and associated with increased body mass index (BMI), mediate similar adverse effects in different target organs and tissues. Injurious agents or toxins causing insulin resistance in adipose tissue and skeletal muscle would result in T2DM, whereas the same insult in liver would produce NASH/metabolic syndrome, and in brain, MCI or early AD-type neurode-generation. We now propose that ceramides and related molecules are critical agents involved in the pathogenesis of these disease states because they: 1) can be generated in liver, adipose tissue, or brain [44–47]; 2) cause insulin resistance ; 3) are cytotoxic ; 4) increase in the central nervous system (CNS) with various dementia-associated diseases, including AD [46, 48–50]; and 5) are lipid soluble and therefore likely to readily cross the blood-brain barrier.
By way of review, ceramides comprise a family of lipids generated from fatty acid and sphingosine (see reviews [45,51,52]). Ceramides are largely distributed in cell membranes, and in addition to their structural functions, ceramides have key roles in intracellular signaling and regulate growth, proliferation, cell migration, adhesion, growth arrest, differentiation, senescence, and apoptosis. Ceramides are generated by either de novo biosynthesis, or sphingolipid degradation (Table 1). For de novo synthesis, ceramides are produced from sphingosine or sphinganine plus fatty acyl-CoA through the actions of ceramide synthases [44,47, 53]. Activation of serine palmitoyl transferase [44,45, 54] or dihydroceramide synthase increases ceramide production in the endoplasmic reticulum via dihydroceramide desaturase (Table 1). Glycosphingolipids are composed of ceramide (hydrophobic) plus oligosaccharide (hydrophilic), and generated enzymatically in the Golgi [55–57]. UDP glucoceramide gycosyltransferase mediates the first step in transferring glucose from UDP-glucose to ceramide . Ceramides also can be generated by hydrolysis of sphingomyelin by sphingomyelinase [44,52], or degradation of complex sphingolipids and glycosphingolipids localized in late endosomes and lysosomes . Potential roles for ceramides in diabetes, obesity, inflammation, NASH, and neurodegeneration have already been suggested [45, 46,49,58]. Of note is that pro-inflammatory cytokines, such as tumor necrosis factor-alpha (TNF-α), can induce ceramide synthesis , and TNF-α levels are increased in both T2DM and NASH [59–62]. We investigated the potential role of ceramides as mediators of neurodegenerationand brain insulin resistance utilizing an in vivo model of chronic obesity/T2DM .
Harlan adult male C57BL/6 mice (N = 10 per group), starting at 4 weeks of age, were pair-fed for 4, 8, 12, 16, or 20 weeks with high fat (HFD) chow diets in which 60% of the calories were derived from fat (Research Diets Inc, New Brunswick, NJ), or low fat (LFD) chow diets in which 5% of the calories were from fat (Harlan, Indianapolis, IN) as reported previously [63,64]. Mice were weighed weekly, and at the time of sacrifice, fresh liver and brain weights were obtained. Livers and brains were sectioned for snap freezing and immersion fixation . Fixed tissues were embedded in paraffin, and histological sections (8 microns) were stained with Hematoxylin and eosin (H&E; liver) or Luxol fast blue, H&E (brain). Frozen tissue was stored at −80°C for biochemical and molecular studies. Our experimental protocol was approved by the Institutional Animal Care and Use Committee at the Lifespan-Rhode Island Hospital and conforms to guidelines established by the National Institutes of Health.
We used qRT-PCR to measure pro-ceramide gene expression with previously described methods . Total RNA isolated from liver and temporal lobe was reverse transcribed using random oligodeoxynucleotide primers. The resulting cDNA templates were used in qPCR amplification reactions with gene specific primer pairs (Table 2) . PCR amplifications were performed in 20 µl reactions containing cDNA generated from 2.5 ng of original RNA template, 300 nM each of gene specific forward and reverse primer, and 10 µl of 2x QuantiTect SYBR Green PCR Mix (Qiagen Inc, Valencia, CA). The amplified signals from triplicate reactions were detected and analyzed using the Master-cycler ep realplex instrument and software (Eppendorf AG, Hamburg, Germany). The amplification protocol used was as follows: initial 15-minutes denaturation and enzyme activation at 95°C, 40 cycles of 95°C × 15 sec, 55–60°C × 30 sec, and 72°C × 30 sec. Post-PCR melting points were examined for all samples. Annealing temperatures were optimized using the temperature gradient program provided with the software. The mRNA levels were determined using the equations of the regression lines generated with serial 10-fold dilutions of 20 ng of recombinant plasmid DNA containing the target sequences studied. Relative mRNA abundance was determined from the ng ratios of specific mRNA to18S [65,66]. PCR amplification efficiencies were all greater than 95%. Relative mRNA abundance was calculated from the ng ratios of specific mRNA to 18S rRNA measured in the same samples. Inter-group statistical comparisons were made using the calculated mRNA/18S ratios. 18S rRNA values were used as denominators because the large abundance of 18S provides an excellent loading control that is virtually invariant with disease state or experimental condition. Moreover, expression levels of traditional housekeeping genes, e.g., actin, glyceraldehyde 3 phosphate de-hydrogenase, and cyclophilin shift unpredictably with disease and experimental treatments, rendering them unreliable for normalizing data with respect to specific genes of interest.
Temporal lobes were homogenized in radioimmuno-precipitation assay buffer with protease and phosphatase inhibitors . Protein concentrations were determined using the bicinchoninic acid (BCA) assay (Pierce, Rockford, IL). We used direct ELISAs to measure tau, phospho-tau, 4-hydroxy-2-nonenal (4-HNE), ubiquitin, β-actin, and choline acetyltransferase (ChAT) expression as previously described . Immunoreactivity was detected with horseradish peroxidase (HRP)-conjugated secondary antibody and Amplex Red soluble fluorophore . Fluorescence was measured (Ex 579/Em 595) in a SpectraMax M5 microplate reader (Molecular Devices Corp., Sunnyvale, CA). Parallel negative control assays included incubations in which the primary, secondary, or both antibodies were omitted.
Frozen tissues (50–70 mg each) were homogenized in 100 µl of water, and 10 µl aliquots were used to measure protein concentration with the BCA assay. Lipids were extracted by adding 1 ml chloroform-methanol (2:1) and incubating the samples for 1 h at room temperature with intermittent agitation. After centrifuging at 3000 rpm for 5 minutes at room temperature, the lipid-containing lower phase was transferred to a clean tube and air-dried. Pellets were re-suspended in 100 µl of 100% ethanol. Total lipid content was measured using a Nile Red fluorescence-based assay (Molecular Probes, Eugene, OR) [68–70]. Briefly, 5 µl aliquots of lipid extract were added to 190 µl of phosphate-buffered saline (PBS) in a 96-well polystyrene white plate, then 5 µl of Nile Red solution (1 mg/ml in DMSO) was added to each well. Reactions were light protected with aluminum foil and incubated at room temperature for 10 minutes with constant platform agitation. Fluorescence intensity (Ex 485/Em 572) was measured in a Spectra-Max M5 microplate reader. Triglyceride levels were measured using a Serum Triglyceride Determination kit (Sigma-Aldrich Co., St. Louis, MO), and cholesterol was measured using the Amplex Red Cholesterol Assay Kit (Molecular Probes, Eugene, OR) according to the manufacturers’ protocols.
QuantiTect SYBR Green PCR Mix was obtained from (Qiagen Inc, Valencia, CA). Rabbit or goat generated monoclonal or polyclonal antibodies to ubiquitin, tau, phospho-tau, 4-HNE, ChAT, and β-actin was purchased from Chemicon (Tecumsula, CA), Cal-Biochem (Carlsbad, CA) or Molecular Probes (Eugene, OR). Secondary antibodies were purchased from Pierce Chemical Co. (Rockford, IL). Amplex Red reagent was obtained from Molecular Probes (Eugene,OR). All other fine chemicals were purchased from either Cal-Biochem (Carlsbad, CA) or Sigma-Aldrich (St. Louis, MO).
Data depicted in the graphs represent the means ± S.E.M.’s for each group. Inter-group comparisons were made using Two-way Analysis of Variance (ANOVA) with the Bonferroni post-hoc test, or the post-hoc Deming test for linear trend. Statistical analyses were performed using GraphPad Prism 5 (GraphPad Software, Inc., San Diego, CA). The computer software generated significant P-values are indicated within the graph panels.
The chronic HFD fed mice developed obesity and T2DM associated with fasting hyperglycemia, hyperlipidemia, and increased serum pro-inflammatory cytokine levels as previously reported [63,64]. Chronic HFD feeding caused progressive increases in mean body weight such that after 8 weeks, the HFD-fed mice were significantly heavier than the LFD control group. With increasing duration of HFD feeding, mean body weight continued to climb (P < 0.0001 for linear trend), whereas the mean body weight of the LFD group remained relatively stable (Fig. 1A), consistent with previous findings in this model . During the first 12 weeks of study, HFD feeding had no significant effect on mean liver weight. Over that interval, the mean liver/body weight ratios were significantly lower in the HFD relative to LFD control group due to their progressive increases in body weight (data not shown). In contrast, at the 16 and 20 week time points, HFD-fed mice exhibited striking and significant increases in mean liver weight (Fig. 1B) due to significant increases in hepatic lipid content as demonstrated with the Nile Red assay (Fig. 1C). Correspondingly, the calculated mean liver-body weight ratios were also increased relative to control (data not shown). Biochemical assays detected significantly elevated mean hepatic triglyceride content after 4, 16, and 20 weeks of HFD feeding (Fig. 1D), and significantly increased hepatic cholesterol after 20 weeks on the HFD (Fig. 1E).
Although mean brain weights were similar in the HFD and LFD groups after 4, 8, 12, 16, or 20 weeks of paired feeding (Fig. 1F), a time-dependent trend of declining mean brain weight was observed in the HFD (R2 = 0.475; P < 0.0001), but not the LFD group (Fig. 1G–1H). The combined effects of increasing body mass and declining brain weight resulted in significantly reduced mean brain/body mass ratios in the HFD relative to control at the 8, 12, 16, and 20 week time points (Fig. 1I). The time dependent trend of declining brain weight/body weight ratio was also statistically significant (P < 0.0001). Finally, the sharpest reductions in brain/body weight ratio occurred concurrently with the striking increases in hepatic steatosis.
LFD livers exhibited regular cord-like architecture with minimal or no evidence of inflammation, steatosis, or apoptosis (Fig. 2). In contrast, after 4, 8, or 12 weeks on the HFD, the livers had progressive increases in hepatic micro- and macrosteatosis with associated disorganization of the architecture (Fig. 2A–2F). After 16 weeks of HFD feeding, the livers exhibited more prominent macro- and microsteatosis with small cluster of lymphomononuclear inflammatory cells (Fig. 2G–2H). After 20 weeks on the HFD, the livers had established histopathological features of NASH, consisting of widespread micro- and macro-steatosis in hepatocytes (40–50% versus 5–10% at earlier time points), patchy lymphomononuclearcell inflammation, apoptosis, and necrosis (Fig. 2I–2J), corresponding with previous descriptions of this and closely related models [72–74].
Histopathological studies were focused on analysis of the temporal lobes because they are major targets of AD neurodegeneration. 4, 8, or 12 weeks of HFD feeding produced no detectable effects on brain histopathology. However, 16 weeks of HFD feeding caused subtle histopathological abnormalities consisting of scattered foci of neuronal loss,apoptosis, or nuclear pyknosis,increased irregularity of neuropil spacing among neurons (consistent with cell loss), and mildly increased white matter gliosis, as previously described . Immunohistochemical staining studies failed to detect neurofibrillary tangles, dystrophic neurites, senile plaques, or AβPP-Aβ deposits in plaques or vessels in any of the brains, corresponding with our previous observations . After 20 weeks of HFD feeding, ongoing apoptosis in both the hippocampal formation and temporal neocortex was more conspicuous, as evidenced by cell loss, and increased lipid peroxidation (4-HNE immunoreactivity), and gliosis (GFAP immunoreactivity) in these regions (Fig. 3). In addition, temporal white matter in HFD fed mice exhibited increased gliosis and lipid peroxidation, reflecting fiber degeneration (Fig. 3), as occurs early in the course of AD .
Exploratory studies demonstrated that normal liver expresses Ceramide synthases (CER) 1, 2, 4, and 5, UDP glucose ceramide glucosyltransferase (UGCG), serine palmitoyltransferase (SPTLC) 1 and SPTLC2, and sphingomyelin phosphodiesterase (SMPD) 1 and SMPD3 (see Table 1 for definitions). qRT-PCR analysis revealed significantly higher mean levels of CER2, CER4, SPTLC1, SPTLC2, SMPD1, and SMPD3 mRNA transcripts in HFD compared with LFD livers, after 12 weeks on the diets (Fig. 4). However, the pro-ceramide synthesis mRNA transcripts, i.e., CER2, CER4, SPTLC1, and SPTLC2, were increased mainly at the 12- and 16-week time points, just preceding or coinciding with the early stages of conspicuous hepatic steatohepatitis. Their expression returned to control levels (4-week time point) after 20 weeks on the HFD. UGCG mRNA levels increased in both groups between the 4- and 8-week time points, and then gradually there-after in the control group, in which peak levels were detected at the 16- and 20-week time points. Mice fed with the HFD had earlier peaks in hepatic UGCG (at 12 or 16 weeks), and significantly lower levels than control at the 20-week time point (Fig. 4E). In HFD-fed mice, SMPD1 and SMPD3 mRNA transcripts increased progressively over time, such that their mean levels were significantly higher than control at both the 16 and 20 week time points (Fig. 4). In contrast to CER and SPTLC, SMPD1 and SMPD3 generate ceramide through hydrolysis of sphingolipids [44,51,52]. Corresponding with the qRT-PCR results, preliminary ELISA studies demonstrated higher levels of ceramide immunoreactivity in HFD relative to LFD fed mice at the 12- (21016 ± 1730 versus 16759 ± 208; P = 0.025) and 16-week (15636 ± 472 versus 14151 ± 160; P = 0.037) time points (Mean ± S.E.M. values in arbitrary units; Student T-tests with 4 samples per group). A more detailed analysis of HFD-induced alterations in ceramide levels and characteristics in liver and brain is currently underway and will be presented in a future report.
We also used qRT-PCR to measure expression of the same pro-ceramide genes in temporal lobe samples (Fig. 5). Generally higher levels of CER2, CER5, UGCG, SPTLC2, SMPD1 and SMPD3 were detected in brain at earlier compared to later time points in the study. We interpreted this result to possibly reflect developmental or aging effects on the expression levels of these genes. At the 8 and/or 12 week time points, CER2, UGCG, SPTLC2, SMPD1 and SMPD3 were all expressed at significantly lower levels in HFD compared with LFD brains. In addition, at the 16 and/or 20 week time points, the mean levels of UGCG mRNA were also significantly lower in the HFD compared with LFD group (Fig. 5). Therefore, unlike liver, temporal lobe ceramide gene expression was not increased by chronic HFD feeding.
AD-type neurodegeneration is associated with increased levels of AβPP, oxidative stress, ubiquitin, and phospho-tau, and reduced levels of tau and ChAT mRNA . ELISAs performed with temporal lobe tissue demonstrated similar mean levels of tau, phospho-tau, β-actin, ubiquitin, and 4-HNE in the HFD and LFD groups at the 8 and 12 week time points (Fig. 6). Only ChAT immunoreactivity was significantly reduced in the HFD group during these early time periods (Fig. 6D). At the 16- and/or 20-week time points, the mean levels of tau, β-actin, and ChAT were significantly reduced, whereas the mean levels of ubiquitin and 4-HNE were significantly increased in temporal lobes of HFD fed relative to control mice (Fig. 6).
Previously, we demonstrated that chronic HFD feeding, which results in obesity and T2DM, causes relatively modest AD-type molecular and biochemical abnormalities in brain, including insulin resistance . For the present study, we re-generated the model to examine the time course of altered gene expression in brain, as well as uncover potential mechanisms by which peripheral insulin resistance causes AD-type neurodegeneration. This line of investigation is important because it could lead to the discovery of biomarkers for detecting individuals at risk for developing cognitive impairment or progressing from MCI to AD in the context of obesity/T2DM. We focused our attention on the liver because previous studies showed that: 1) NASH occurs frequently with obesity/T2DM and is a feature of the model used herein; 2) NASH is associated with hepatic insulin resistance; 3) individuals with NASH can exhibit neuropsychiatric dysfunction, including anxiety and depression , which frequently precede cognitive impairment and dementia; 4) peripheral ceramide production is increased in adipose tissue in NASH, and ceramides, as well as long-chain fatty acids, mediate insulin resistance [45,77,78]; and 5) ceramide can be neurotoxic, and its levels are increased in AD as well as other injury or inflammatory disease states [46,48–50,79,80].
This study demonstrated that HFD feeding caused body weight to increase progressively, and eventually achieve levels that were nearly two-fold higher than control. The indices of T2DM, including hyperglycemia, hyperinsulinemia, and hypercholesterolemia also increased progressively over the time course of HFD feeding as previously reported [64, 81]. In contrast, liver weight remained relatively stable during the first 12 weeks of feeding, but then increased sharply after 16 and 20 weeks of HFD feeding, due to striking increases in hepatic lipid content (steatosis), inflammation, apoptosis, and necrosis, i.e., NASH. Other investigators demonstrated that NASH, in this or related models, is associated with increased serum transaminase levels, reflecting hepatocellular injury [72,74,82]. Therefore, it appears that during the earlier phases of HFD feeding, compensatory mechanisms help sustain structural and functional integrity of the liver, but after 16 weeks of HFD feeding, a critical metabolic perturbation disrupts the homeostatic milieu, allowing hepatic steatosis to progress to NASH. One potential mediator of this response is the abrupt surge in serum TNF-α that occurs after 16 weeks of HFD feeding . TNF-α is a potent pro-inflammatory cytokine that plays a key role in peripheral and hepatic insulin resistance [60,62,84,85], and also promotes ceramide biosynthesis . Although a potential source of the increased TNF-α in this model is peripheral adipose tissue with ongoing adipocyte death and remodeling , preliminary data suggest that endogenous hepatic TNF-α expression is increased in the HFD mice (Longato, et al, unpublished).
After 16 weeks of HFD feeding, the livers exhibited histopathological features of NASH, i.e., steatosis with multiple foci of lymphomononuclear cell inflammation, necrosis, and apoptosis, consistent with previous reports . Molecular studies demonstrated that throughout the period of HFD feeding, the mRNA levels of various pro-ceramide genes were significantly increased relative to control, including genes responsible for generating ceramide de novo (CER 2,3,4,5), and also those involved in sphingomyelin degradation (SMPD1 and SMPD3). The gradual reductions in the expression of genes that mediate de novo ceramide synthesis could represent a compensatory response to lipid accumulation in hepatocytes. On the other hand, the persistently increased levels of SMPD3 after 16 or 20 weeks of HFD feeding may reflect an effort to prevent further accumulation of lipids in hepatocytes through increased hydrolysis of sphingomyelin.
Consequences of increased sphingomyelinase activity include increased generation of ceramide through degradation of sphingmyelin, as well as increased production of fatty acids. Since ceramides have key roles in signaling cellular processes such as growth arrest, senescence, apoptosis, and cell death, increased expression of pro-ceramide genes could contribute to the deficits in hepatocellular repair and regenerative functions in NASH [45,77]. Conceivably, ceramides generated by degradation of sphingolipids and glycosphingolipids localized in late endosomes and lysosomes may be more toxic and inhibitory to insulin signaling than ceramides generated via de novo synthesis pathways [45,58,77]. In contrast to the findings in liver, temporal lobe (brain) pro-ceramide gene expression was not significantly increased by the chronic HFD feeding. Therefore, if ceramides have a role in mediating brain insulin resistance, neurodegeneration, and cognitive impairment in obesity, T2DM, and NASH, the source is not likely to be the CNS, and instead would probably be of liver, and possibly adipocyte origin.
Previously, we demonstrated that chronic HFD feeding of C57BL/6 mice causes mild neuropathological lesions but significant impairments in insulin receptor binding and insulin responsive gene expression in temporal lobe tissue . Those abnormalities were associated with reduced expression of ChAT and GAPDH. The time course analysis performed in the present study revealed early impairments in ChAT expression, beginning after only 8 weeks of HFD feeding, and coinciding with the increases in the expression of multiple proceramide genes in liver. The facts that: 1) ceramide inhibits insulin signaling [45,77]; 2) ChAT gene expression is insulin responsive ; and 3) ceramide is lipid soluble and therefore probably readily crosses the blood brain barrier, suggest that the early impairments in ChAT expression in brain may be mediated by increased hepatic ceramide production. However, it is noteworthy that ChAT expression increased sharply between 4 and 8 weeks on the LFD, whereas in mice fed with the HFD, ChAT expression increased gradually over time, such that at the 16- and 20-week time points, the levels more closely approximated to those in the control group. This suggests that maturation of CNS cholinergic function may be delayed but not thoroughly impaired by brain insulin resistance states.
The other indices of neurodegeneration, including increased 4-HNE (marker of lipid peroxidation) and ubiquitin (index of abnormal protein processing and accumulation), and reduced tau (probably reflecting cytoskeletal collapse and insulin/IGF resistance) were primarily detected at the 16 and 20 week time points, coinciding with surges in steatohepatitis and serum TNF-α, and persistent elevation of hepatic SMPD3. These results provide supportive evidence that hepatic/peripheral ceramide and pro-inflammatory cytokine production play key roles in the pathogenesis of CNS oxidative stress, insulin resistance and neuronal cytoskeletal collapse, all of which occur in AD.
Since AD shares many biochemical, molecular, and signal transduction abnormalities in common with T2DM and NASH, but the relevant pathology is fundamentally confined to the CNS in the vast majority of cases, we suggested that AD be regarded as “Type 3 diabetes” [10,11]. It is of particular interest that, with regard to the antecedent discussion, CNS ceramide levels are elevated in both AD  and AD-relevant experimental animal models [46,79]. These relatively recent observations hearken back to a much earlier report showing that white matter atrophy is one of the earliest abnormalities in AD . At the present time, it appears that white matter atrophy in AD could be mediated by brain insulin resistance, which is evident even in the early stages of AD . Insulin resistance in white matter would result in degeneration and loss of oligodendroglia, since oligodendroglia require insulin/IGF-1 for survival signaling and myelinogenesis [87–89]. On the other hand, insulin/IGF resistance would promote oxidative stress and secondarily lead to activation of pro-inflammatory cascades in astrocytes, and attendant myelin degeneration via activation of sphingomyelinases and pro-inflammatory cytokines, including TNF-α, and attendant release of ceramide. This proposed scheme provides a mechanism for producing Type 3 diabetes without the need for hepatic/peripheral sources of ceramides or cytokines, and which would be relevant in the vast majority of sporadic AD cases. With mounting evidence pointing toward CNS insulin/IGF resistance as the mediator if not initiator of AD-type neurodegeneration, our very next goal should be to identify the agents and factors responsible for establishing this cascade.
Although a likely connection between increased CNS ceramide (and probably other toxic lipids as well) and AD has been demonstrated as discussed above, the novelty of the present work is that extra-CNS sources of ceramide, such as liver or adipose tissue, can contribute to the pathogenesis of cognitive impairment and AD-type neurodegeneration. The critical variable mediating this effect may be the degree to which liver disease or adipocyte degeneration and remodeling increase ceramide production. The fact that the effects of HFD feeding were not identical to that which actually happens in AD, lends support to our hypothesis that T2DM, NASH, and obesity do not cause AD, and instead they probably serve as pathogenic co-factors. This phenomenon could account for both the absence of complete overlap and the two- or three-fold increased risk of developing MCI or AD among individuals with T2DM [7–9]. Improved ability to detect increased levels of toxic ceramides and related lipids in peripheral blood and cerebrospinal fluid may help identify individuals at risk for developing cognitive impairment in the context of obesity, T2DM, and/or NASH.
Research supported by AA02666, AA02169, AA11431, AA12908, and AA16126 from the National Institutes of Health, and a Child Health Research Grant from the Hood Foundation