Neuronal human BACE1 knockin induces systemic diabetes in mice
- 7.4k Downloads
β-Secretase 1 (BACE1) is a key enzyme in Alzheimer’s disease pathogenesis that catalyses the amyloidogenic cleavage of amyloid precursor protein (APP). Recently, global Bace1 deletion was shown to protect against diet-induced obesity and diabetes, suggesting that BACE1 is a potential regulator of glucose homeostasis. Here, we investigated whether increased neuronal BACE1 is sufficient to alter systemic glucose metabolism, using a neuron-specific human BACE1 knockin mouse model (PLB4).
Glucose homeostasis and adiposity were determined by glucose tolerance tests and EchoMRI, lipid species were measured by quantitative lipidomics, and biochemical and molecular alterations were assessed by western blotting, quantitative PCR and ELISAs. Glucose uptake in the brain and upper body was measured via 18FDG-PET imaging.
Physiological and molecular analyses demonstrated that centrally expressed human BACE1 induced systemic glucose intolerance in mice from 4 months of age onward, alongside a fatty liver phenotype and impaired hepatic glycogen storage. This diabetic phenotype was associated with hypothalamic pathology, i.e. deregulation of the melanocortin system, and advanced endoplasmic reticulum (ER) stress indicated by elevated central C/EBP homologous protein (CHOP) signalling and hyperphosphorylation of its regulator eukaryotic translation initiation factor 2α (eIF2α). In vivo 18FDG-PET imaging further confirmed brain glucose hypometabolism in these mice; this corresponded with altered neuronal insulin-related signalling, enhanced protein tyrosine phosphatase 1B (PTP1B) and retinol-binding protein 4 (RBP4) levels, along with upregulation of the ribosomal protein and lipid translation machinery. Increased forebrain and plasma lipid accumulation (i.e. ceramides, triacylglycerols, phospholipids) was identified via lipidomics analysis.
Our data reveal that neuronal BACE1 is a key regulator of metabolic homeostasis and provide a potential mechanism for the high prevalence of metabolic disturbance in Alzheimer’s disease.
Keywords18FDG-PET Ceramide ER stress Glucose homeostasis Insulin signalling Neuronal BACE1
Protein kinase B
Amyloid precursor protein
Brown adipose tissue
C/EBP homologous protein
Central nervous system
Eukaryotic initiation factor 2α
2-[18F]Fluoro-2-deoxy-d-glucose positron emission tomography
Glucose-dependent insulinotropic peptide
Glucose-stimulated insulin secretion
Glucose tolerance test
Human BACE1 gene
Insulin receptor β
Mammalian target of rapamycin
Glycogen synthase kinase-3 β
Ribosomal protein S6
Protein kinase RNA-like ER kinase
Protein tyrosine phosphatase 1B
Retinol-binding protein 4
White adipose tissue
The incidence of type 2 diabetes and Alzheimer’s disease is rising at an alarming rate. Type 2 diabetes increases the risk of Alzheimer’s disease ; however, systemic glucose intolerance and insulin resistance are also reported in dementia patients without a history of diabetes . Alzheimer’s disease brains exhibit defective neuronal insulin signalling  and glucose hypometabolism in 2-[18F]fluoro-2-deoxy-d-glucose positron emission tomography (18FDG-PET) studies . Further, mouse models of both diseases share similar cognitive phenotypes , and increased susceptibility to high-fat diet (HFD)-induced diabetes is observed in Alzheimer mouse models compared with controls . Collectively, the overlapping pathology indicate common pathogenic factors, although their molecular foundations remain unclear.
β-Secretase 1 (BACE1) is implicated in Alzheimer’s disease as the enzyme responsible for the rate-limiting step in β-amyloid (Aβ) production . Brain BACE1 levels increase with age , particularly in Alzheimer’s disease , and following pathological events [10, 11]. Bace1 deletion abolishes Aβ deposition and rescues cognitive deficits in APP mutant mice . We recently confirmed that neuron-specific knockin of human (h)BACE1 induces Aβ accumulation, promotes brain inflammation and recapitulates Alzheimer’s disease-like phenotypes in mice in the absence of mutant APP expression , suggesting that BACE1 represents a molecular risk factor for sporadic Alzheimer’s disease.
Although studied principally for its role in amyloidosis, BACE1 has multiple substrates other than APP , comprising transmembrane proteins involved in intercellular signalling. BACE1 expression is predominantly neuronal, although BACE1 mRNA is also found in the liver, skeletal muscle and pancreas , with pancreatic mRNA encoding an inactive isoform [7, 16]. A potential role for BACE1 in metabolic regulation has only recently emerged, as Bace1 knockout improved glucose metabolism and protected mice from HFD-induced obesity and diabetes . Conversely, the induction of insulin deficiency via systemic streptozotocin (STZ) injection raised central BACE1 levels , and this was associated with endoplasmic reticulum (ER) stress . Thus, BACE1 may contribute to metabolic regulation; however, it remains to be established whether BACE1 mediates the association between type 2 diabetes and Alzheimer’s disease.
To elucidate the contribution of neuronal BACE1 to systemic glucose regulation and lipid metabolism, we characterised central and peripheral metabolic changes in brain-specific hBACE1 (PLB4) knockin mice . Because global deletion of Bace1 protected mice from diet-induced diabetes , we hypothesised that neuronal BACE1 may regulate system metabolism in addition to inducing brain pathologies relevant to Alzheimer’s disease . We provide evidence that BACE1-induced hypothalamic dysregulation causes systemic diabetes, which may explain the high comorbidity of diabetes and Alzheimer’s disease in ageing populations.
All animals were housed and tested in accordance with European (Directive on the Protection of Animals used for Scientific Purposes, 2010/63/EU) and UK Home Office regulations, experiments were approved by the University of Aberdeen Ethics Board and performed in accordance with the Animal (Scientific Procedures) Act 1986 and following Animal Research: Reporting of In Vivo Experiments (ARRIVE) guidelines. The transgenic PLB4 and WT lines were generated and bred as previously described , experimenters were not blinded to genotype. All mice were housed in single sex groups, unless food and water consumption was being measured. Physiological assessments were performed in male mice at 3, 4, 5 and 8 months of age. All mice were killed by neck dislocation. Five-hour-fasted mice aged 3 and 8 months were used for liver assays, and all postmortem molecular analyses of signalling pathways (western blotting and quantitative real-time PCR [qPCR]) were performed using samples from 8-month-old mice. Glucose-stimulated insulin secretion (GSIS) was determined in 4-month-old mice at 0, 15 and 30 min time points during glucose tolerance tests (GTTs). Blood leptin content was determined from baseline reading at 3, 4 and 8 months in a 5 h fasted state. EchoMRI scans (EchoMRI, Houston, TX, USA) were performed on mice at ages 4, 5 and 8 months. Additional 6-month-old mice were used for lipidomic analysis and 18FDG-PET imaging. Body weights were recorded and food and water consumption (in g) were measured for 2 weeks in 5- and 8-month-old male mice.
Tail blood glucose was determined using an AlphaTRAK glucometer (Berkshire, UK). GTTs were performed (as described in [6, 20]) in 5-h fasted male mice at age 3 (WT, n = 4; PLB4 n = 7), 4 (WT, n = 8; PLB4, n = 7), 5 (WT, n = 9; PLB4, n = 15) and 8 (WT, n = 4; PLB4 n = 4) months. Fasting blood glucose was determined at baseline (time 0) before i.p. injection of bolus glucose (20%; 2 g/kg body weight). Blood glucose clearance was assessed at 15, 30, 60 and 90 min post injection in 3-, 5- and 8-month-old mice; and at 0, 15 and 30 min post injection in 4-month-old mice.
Tail- or trunk-derived blood was collected from 5-h fasted mice aged 3 (WT, n = 6; PLB4, n = 8), 4 (WT, n = 7; PLB4, n = 8), 5 (WT, n = 9; PLB4, n = 12) and 8 (WT, n = 8; PLB4, n = 10) months into serum separator microtubes; serum was used for insulin and leptin determination (Insulin ELISA, Millipore, Darmstadt, Germany; leptin ELISA, CrystalChem, Zaandam, the Netherlands). A multiplex assay (customer-designed mouse metabolic hormone Magnetic Multiplex Assay; Merck Millipore, Darmstadt, Germany) was used for the simultaneous quantification of leptin, amylin (active form), C-peptide 2, glucose-dependent insulinotropic peptide (GIP; total), pancreatic polypeptide, peptide tyrosine tyrosine (PYY), IL-6 and resistin following the manufacturer’s instructions. Triacylglycerol and glycogen assays were used as previously described .
Comparative lipidomics plasma and brain analyses
Frozen plasma and forebrain samples from 6-month-old male mice (WT, n = 8; PLB4, n = 8) were used for global lipid analysis using liquid chromatography–mass spectroscopy (LC–MS). Plasma lipids were extracted from samples according to the Folch method. Lipids were solvent extracted in methanol/chloroform (2:1 vol./vol.). The lipid extracts were subsequently analysed by LC–MS in positive and negative ion modes with a C18 column and a water/acetonitrile/isopropanol gradient using an Exactive Orbitrap system (Thermo Scientific, Hemel Hempstead, UK) . Lipidomic datasets were processed using Progenesis QI software (version 2.0, Non-Linear Dynamics, Newcastle upon Tyne, UK) and searched against LIPID MAPS (www.lipidmaps.org/) and the Human Metabolome Database (www.hmdb.ca/).
Brain and upper body PET/CT imaging
Six-month-old female mice (WT, n = 10; PLB4, n = 9) were imaged using an Argus GE dual ring PET/CT scanner (Sedecal, Madrid, Spain) using our published protocol .
All tissues were lysed in RIPA buffer [as described in . Tissue from 5-h fasted mice was immunoblotted using rabbit polyclonal antibodies diluted 1:1000 in TRIS-buffered saline containing Tween-20, 5% BSA and 0.05% sodium azide. Antibodies raised against human BACE1 (C-terminal 485–501; no. 195111, Calbiochem, UK); phospho-protein kinase B (p-AktSer473; no. 4060), total Akt (Akt; no. 9272), phospho-glycogen synthase kinase-3β (p-GSK-3βSer21/9; no. 9331), phospho-ribosomal protein S6 (p-rpS6Ser235–236; no. 4858), total rpS6 (rpS6; no. 2217), phospho-S6 kinase (p-S6KThr389; no. 9234), total S6 kinase (S6K; no. 2708), phospho-mammalian target of rapamycin (p-mTORSer2448; no. 5536), mTOR (no. 2983), phospho-glycogen synthase (p-GSSer641; no. 3891), phospho-eukaryotic translation initiation factor 2α (p-eIF2αSer51; no. 3398), total eIF2α (no. 5324), protein kinase RNA-like ER kinase (PERK; no. 5683; all obtained from Cell Signaling Technology, Leiden, the Netherlands); insulin receptor β (IRβ; C-19, no. Sc-711; 1:500 dilution), apolipoprotein (ApoE; no. Sc-6384), mouse monoclonal protein tyrosine phosphatase 1B (PTP1B; no. Sc-14021), C/EBP homologous protein (CHOP; also known as GAD153, no. Sc-7351; Santa Cruz, Heidelberg, Germany); and retinol-binding protein 4 (RBP4; no. A0040, Dako, Glastrup, Denmark) were used following manufacturer’s instructions and as described before [6, 13]. Coomassie Blue (PhastGel Blue R; GE Healthcare, Watford, UK) was used as a loading control [13, 22].
RNA was extracted from frozen hypothalamic and cortical samples using TRI Reagent, as previously described . Target genes (hBACE1 and mouse Pomc, Mc4R, Npy, LepR, Ptp1b, Chop, Cd11 and Cd68) were amplified by qPCR using GoTaq Master Mix (Promega, Madison, WI, USA). The geometric mean of three commonly used reference mRNAs (Ywhaz, Nono and Actb) was used to normalise data.
Prism 5 (GraphPad, La Jolla, CA, USA) was used for statistical analyses, and data are expressed as the means ± SEM. All molecular and genetic data were adjusted to loading controls or housekeeping genes and calculated relative to WT. Two-tailed t tests with Welch’s correction were used for unbiased comparisons between transgenic mice and controls. Group analyses used two-way ANOVA (with repeated measures where appropriate), followed by Bonferroni post hoc tests.
Systemic glucose homeostasis is impaired in PLB4 mice
Altered plasma metabolic homeostasis and hyperleptinaemia
Levels of the adipocyte-derived hormone, leptin, which strongly correlate with obesity and diabetes, were initially normal in PLB4 mice, but were drastically elevated at 4 and 8 months (Fig. 1p) when adiposity was increased.
Since body mass and long-term energy expenditure are tightly regulated by complex signalling networks between the CNS and peripheral systems, we screened serum samples from 8-month-old PLB4 mice for markers of homeostatic control using a multiplex assay (Fig. 1q). Concentrations of fasting C-peptide were low, indicative of defective proinsulin synthesis, while amylin levels (co-secreted with insulin) were enhanced. GIP was also upregulated in PLB4 mice compared with controls.
Impaired hepatic glycogen synthesis, insulin resistance and fatty liver phenotype
Dysregulated central and plasma lipid composition
Summary of the brain and plasma lipid species altered in PLB4 mice
↑ Increased vs WT
↓ Decreased vs WT
BvsP match, n (%)
BvsP match, n (%)
PC and lysoPC
PE and lysoPE
Reduced brain glucose metabolism
Abnormal neuronal insulin signalling
Altered hypothalamic melanocortin transcription and ER stress
Diabetic complications can lead to cognitive dysfunction and are acknowledged risk factors of Alzheimer’s disease, but little is known about the reverse scenario. Here, we demonstrate that the major amyloidogenic enzyme, BACE1, is sufficient to increase the risk of diabetes development when expressed in neurons only. Severe impairment in systemic glucose homeostasis and insulin sensitivity was evident in PLB4 mice from the age of 4 months onwards and progressively deteriorated with age. We previously reported that PLB4 mice develop mild cognitive deficits between 4 and 6 months. Our current findings therefore indicate that neuronal BACE1 induces global metabolic dysregulation along with brain inflammation and amyloidosis-related cognitive decline. Overall, the diabetic profile of PLB4 mice agrees with the improved glucose clearance and insulin sensitivity in mice lacking murine Bace1 globally [17, 30], but pinpoints neuronal BACE1 as the major driver of this alteration.
Although both Bace1 knockout and our hBACE1 knockin mice were leaner compared with WT controls, there were differential effects on adiposity and leptin signalling. Deletion of Bace1 decreased adiposity and improved sensitivity to leptin, while neuronal hBACE1 expression promoted adipogenesis and induced hyperleptinaemia. Hence, it seems plausible that manipulation of (neuronal) BACE1 affects body weight via changing metabolic efficiency, adipose composition and signalling. Importantly, leptin production and signalling are regulated by neuronal PTP1B . We propose that the elevated central PTP1B expression in PLB4 forebrains may explain the hyperleptinaemic profile and compensatory BAT hyperactivity in these mice, which confirms that leptin has opposing effects on WAT and BAT .
Further, while proinsulin synthesis appeared unchanged in PLB4 mice compared with controls, heightened levels of amylin indicate an attempt to downregulate hyperglycaemia in the insulin-deficient state, and suggests that pancreatic function was at least partially preserved in 8-month-old PLB4 mice. Glucose intolerance and the fatty liver phenotype in PLB4 mice were further associated with drastically elevated levels of serum GIP. Although GIP exhibits insulinotropic properties under physiological conditions, its incretin effects are thought to be blunted in the diabetic state, and elevated levels may promote fatty acid accumulation and induction of proinflammatory cytokines .
The fatty liver phenotype of PLB4 mice corresponded with high plasma triacylglycerol and elevated levels of phospholipids such as phosphatidylcholine and lysophosphatidylcholine, which are typically observed in type 2 diabetic patients . Altered plasma lipid composition was also recently proposed to predict early memory impairments, thus potentially offering a novel approach to identify Alzheimer’s disease [27, 33, 34].
Although lipid accumulation was also affected in neuronal tissue from symptomatic PLB4 mice, there was a poor species match for plasma, suggesting that (1) plasma markers may not be indicative of changes in brain lipid composition and (2) that the increase in neuronal lipids probably originated from the CNS. PLB4 mice had pronounced elevation of several classes of phospholipids such as phosphatidylethanolamine, lysophosphatidylethanolamine and phosphatidylserine, which are major components of neuronal membrane bilayers. Such alterations are reported in human Alzheimer’s disease brains  and are proposed to affect mitochondrial function, signal transduction and receptor activation, hence interfering with neurotransmission and neuronal integrity [36, 37].
Importantly, brain ceramide levels were substantially increased in PLB4 mice compared with controls. This sphingomyelin precursor is of particular interest because it increases naturally with age  and at an early stage of human Alzheimer’s disease . Furthermore, ceramides were previously shown to regulate BACE1 protein expression and promote APP β-site cleavage . Although the mechanisms through which BACE1 may upregulate ceramide accumulation are largely unknown, initial evidence suggests that Aβ peptides may activate sphingomyelinase . It therefore seems plausible that the introduction of BACE1 promotes ceramide biogenesis, and vice versa. In support of this, a recent study demonstrated that intracerebroventricular ceramide infusions induce lipotoxicity and hypothalamic ER stress associated with increased eIF2α and PERK phosphorylation, sympathetic inhibition, reduced weight gain and altered energy balance in rats .
Cerebral hypometabolism revealed via 18FDG-PET imaging is common in early dementia patients as well as diabetic patients with or without mild cognitive impairment , and may ultimately contribute to cognitive pathology. We found that reduced glucose utilisation in PLB4 mice was associated with poor neuronal insulin sensitivity. Altered IR and PTP1B expression occurred in PLB4 mice at an advanced stage of systemic insulin resistance, hepatic dysfunction and Aβ-associated cognitive impairment . Similar elevations in neuronal PTP1B were observed in response to HFD feeding in other Alzheimer’s disease models , while deletion of neuronal PTP1B improved IR signalling and protected against HFD-induced obesity and insulin resistance .
Increased S6K phosphorylation in PLB4 vs WT brains and elevated expression of its substrate, rpS6, suggest an increased demand for protein and lipid synthesis. Additionally, the ribosomal element is regulated by eIF2, offering an alternative route for its modification . Elevated rpS6 phosphorylation along with increased brain RBP4 levels were previously found in Alzheimer’s disease mice on a HFD . Neuronal pathways that mediate RBP4’s action and toxicity are yet to be investigated, but have been linked to proinflammatory cytokines in macrophages and to activation of JNK, a major ER stress kinase .
Recent studies revealed reduced hypothalamic volume and accelerated atrophy of orexin neurons in early Alzheimer’s disease [45, 46]. In contrast to diet-induced obese and diabetic models, PLB4 mice displayed increased Pomc and Mc4r mRNA levels, suggesting a hypothalamic shift toward appetite suppression and increased energy expenditure. The increase in hypothalamic Mc4r and Pomc transcription in PLB4 mice further contrasts with the recently demonstrated Aβ-oligomer-induced elevation in Npy mRNA (but not in Pomc mRNA), which was not associated with changes in circulating leptin levels . Here, elevated melanocortin transcription in the hypothalamus of PLB4 mice may be a downstream effect of persistently increased circulating levels of leptin and systemic hyperglycaemia.
An advanced state of ER stress was confirmed in the hypothalamus of PLB4 mice, resembling that induced by HFD feeding . Pharmacologically induced hypothalamic ER stress resulted in systemic diabetes in mice ; this was also recently illustrated for Aβ oligomer infusions . The pathology of PLB4 mice therefore agrees with a scenario of BACE1-driven elevations of the ER stress marker CHOP and the protein translation regulator eIF2α, which suggest induction of an integrated stress response. Mechanistically, neuronal expression of human BACE1 may promote hypothalamic ER stress via ceramide lipotoxicity, activation of eIF2α and an integrated stress response , in addition to driving Aβ production .
In conclusion, we demonstrate that neuronal expression of human BACE1 causes systemic diabetic complications. We propose that increased levels of central BACE1 promotes metabolic disturbance via inducing hypothalamic impairment, ER stress, and Aβ and lipid accumulation, leading to neuronal damage, insulin resistance, hepatic deficits and global glucose dyshomeostasis. The comorbid phenotype of the PLB4 mouse provides insight into the complex mechanistic interactions between diabetes and Alzheimer’s disease. As an extension to the hypothesis that diabetic complications promote the onset and progression of Alzheimer’s disease, we suggest that the reverse scenario may also apply.
The authors thank S. Tammireddy (Diabetes and Cardiovascular Science, University of the Highlands and Islands, Inverness, UK) for technical support with the lipidomics component.
We would like to thank R. Simcox, Romex Oilfield Chemicals, for financial support for KP, and acknowledge additional contributions from the Scottish Alzheimer’s Research UK network for the lipidomics work. The College of Life Science and Medicine, University of Aberdeen, sponsored the imaging study. MD was funded by British Heart Foundation and Diabetes UK; NM was funded by a British Heart Foundation Intermediate Fellowship; KS was funded by a European Foundation for the Study of Diabetes/Lilly programme grant; and RD was funded by an Institute of Medical Sciences PhD studentship.
Duality of interest statement
The authors declare that there is no duality of interest associated with this manuscript.
BP and MD conceived the study; NM, DK and GR participated in the study design; KP and RD carried out the main experimental work, including western blotting brain and liver samples, physiological glucose experiments and plasma measurements; RD acquired and analysed EchoMRI and liver assay data; KS carried out and analysed the qPCR experiments; PW and MKD conducted the lipidomics study, and analysed brain and plasma lipid composition; MM and AW performed 18FDG-PET imaging and analysed the glucose uptake data; KP, BP and MD performed the final data analysis and interpretation, with significant input from NM, DK and GR; KP, BP and MD co-wrote the paper; and all authors critically revised and approved the final version of the manuscript. BP and MD are responsible for the integrity of this work as a whole.
- 5.Carvalho C, Machado N, Mota PC et al (2013) Type 2 diabetic and Alzheimer’s disease mice present similar behavioral, cognitive, and vascular anomalies. J Alzheimer’s Dis 35:623–635Google Scholar
- 6.Mody N, Agouni A, McIlroy GD, Platt B, Delibegovic M (2011) Susceptibility to diet-induced obesity and glucose intolerance in the APP (SWE)/PSEN1 (A246E) mouse model of Alzheimer’s disease is associated with increased brain levels of protein tyrosine phosphatase 1B (PTP1B) and retinol-binding protein 4 (RBP4), and basal phosphorylation of S6 ribosomal protein. Diabetologia 54:2143–2151CrossRefPubMedGoogle Scholar
- 29.Norseen J, Hosooka T, Hammarstedt A et al (2012) Retinol-binding protein 4 inhibits insulin signaling in adipocytes by inducing proinflammatory cytokines in macrophages through a c-Jun N-terminal kinase- and Toll-like receptor 4-dependent and retinol-independent mechanism. Mol Cell Biol 32:2010–2019CrossRefPubMedPubMedCentralGoogle Scholar
- 45.Loskutova N, Honea RA, Brooks WM, Burns JM (2010) Reduced limbic and hypothalamic volumes correlate with bone density in early Alzheimer’s disease. J Alzheimer’s Dis 20:313–322Google Scholar
Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.