High fat diet consumption results in mitochondrial dysfunction, oxidative stress, and oligodendrocyte loss in the central nervous system

Abstract

Metabolic syndrome is a key risk factor and co-morbidity in multiple sclerosis (MS) and other neurological conditions, such that a better understanding of how a high fat diet contributes to oligodendrocyte loss and the capacity for myelin regeneration has the potential to highlight new treatment targets. Results demonstrate that modeling metabolic dysfunction in mice with chronic high fat diet (HFD) consumption promotes loss of oligodendrocyte progenitors across the brain and spinal cord. A number of transcriptomic and metabolomic changes in ER stress, mitochondrial dysfunction, and oxidative stress pathways in HFD-fed mouse spinal cords were also identified. Moreover, deficits in TCA cycle intermediates and mitochondrial respiration were observed in the chronic HFD spinal cord tissue. Oligodendrocytes are known to be particularly vulnerable to oxidative damage, and we observed increased markers of oxidative stress in both the brain and spinal cord of HFD-fed mice. We additionally identified that increased apoptotic cell death signaling is underway in oligodendrocytes from mice chronically fed a HFD. When cultured under high saturated fat conditions, oligodendrocytes decreased both mitochondrial function and differentiation. Overall, our findings show that HFD-related changes in metabolic regulators, decreased mitochondrial function, and oxidative stress contribute to a loss of myelinating cells. These studies identify HFD consumption as a key modifiable lifestyle factor for improved myelin integrity in the adult central nervous system and in addition new tractable metabolic targets for myelin protection and repair strategies.

1. Introduction

White matter pathology including myelin sheath abnormalities, demyelination, loss of oligodendrocytes, and impaired myelin repair processes are significant pathophysiological characteristics of diverse neurological conditions, including multiple sclerosis (MS), Alzheimer’s disease, neuropsychiatric disorders, and traumatic brain and spinal cord injury [1–6]. Oligodendrocytes not only serve as insulation by ensheathing axons and allowing for increased speed and temporal control of action potential propagation, but also provide key metabolic and trophic support for neurons [1,7,8]. Development of new strategies to prevent oligodendrocyte loss and guide efficient oligodendrogenesis nd myelination has the potential to therapeutically benefit individuals affected by white matter-related neurological injury and disease. Since oligodendrocyte function is intimately linked to neighboring glial cells and neuronal activity, in addition to microenvironmental factors, the most promising interventions will likely involve a combinatorial approach [1,9].

Obesity during childhood or adolescence is associated with an increased risk of developing MS [10–12]. Recent studies support a causal role for elevated body mass index (BMI) in MS etiology which is further increased when interacting with other known genetic and environmental risk factors such as HLA polymorphisms or Epstein Barr virus positivity [10,13,14]. Interestingly, metabolic dysfunction resulting from excessive dietary fat intake may also influence the initiation, progression, or disability burden of MS [13,15,16]. In terms of conversion from relapsing remitting (RRMS) to secondary progressive (SPMS), elevated BMI is associated with an increased risk of SPMS in smokers, but not non-smokers [15]. There is also a significant positive association between higher BMI or adverse lipid profile and higher Expanded Disability Status Score (EDSS) [16]. Since obesity and metabolic syndrome are common co-morbidities which impact quality of life for individuals with MS, recent studies have evaluated the utility of drugs and interventions to overcome metabolic syndrome as a strategy to improve patients’ physical and mental health or radiological outcomes [16–18].

Disruptions in brain lipid metabolism, serum lipid profiles, and cholesterol levels have all been reported during active disease in MS [16,19]. How to manage dietary fat consumption for optimal CNS health is still controversial since accurate myelin assembly requires large amounts of lipids and fatty acid synthesis, yet overconsumption of saturated fats is detrimental for overall brain function [19,20]. HFD consumption exacerbates both autoreactive immune responses and neuroinflammation in experimental autoimmune encephalitis (EAE) [21,22]. Metabolic derangement resulting from obesogenic diets or excessive saturated fatty acids can lead to oxidative and ER stress responses and ultimate death of vulnerable cell types [23–25]. Elevated levels of oxidative stress and ER stress-related proteins have been observed in MS lesions [26,27]. Oligodendrocytes are particularly sensitive and susceptible to oxidative stress having extremely high energetic and ER demands in order to support myelin membranes [28]. Moreover, our recent findings show that chronic consumption of a Western style diet, containing high fat and high sucrose, causes significant loss of oligodendrocytes and their progenitors in the intact adult murine spinal cord [29]. Therefore, to elucidate mechanisms underlying oligodendrocyte loss observed in the context of a Western diet, here we specifically address whether similar effects occur with consumption of a diet high in saturated fat alone. The brain and spinal cord of mice consuming a regular or a high fat diet for 4 to 12 wk were analyzed using transcriptomic, metabolomic, functional, and immunohistochemical outcome measures to reveal potential mechanisms of oligodendrocyte loss. These approaches identified perturbations in mitochondrial function among the key mechanisms by which HFD-consumption is likely to impair myelinating cell function in vivo. Parallel disruptions in mitochondrial function and differentiation were also identified when HFD was modeled in cultures of purified oligodendrocytes grown under high saturated fat conditions. Altogether these findings highlight new pathogenic mechanisms by which excess dietary fat contributes to myelin injury in the adult brain and spinal cord and uncover new targets for therapy.

2. Materials and methods

2.1. Experimental method details

2.1.1. Animal husbandry and diet

All diets, procedures, and experiments were approved by The Mayo Clinic Institutional Animal Care and Use Committee (IACUC) and performed in accordance with appropriate institutional and regulatory guidelines. Male C57BL6/J mice were obtained from Jackson Labs (Bar Harbor, ME) at 8 wk of age and were randomly assigned to feeding on either a regular diet (RD, 10% kcal fat, D12450K, Research Diets, New Brunswick, NJ) or a high fat diet (HFD, 60% kcal fat, D12492, Research Diets, New Brunswick, NJ) beginning at 10 wk of age. Saturated fat comprised 22.7 and 32% of the total fat content within the diets, respectively. Carbohydrate comprised 70% kcal for RD and 20% HFD, while both diets had 20% kcal protein. Mice were provided ad libitum access to the diets along with tap water and marked by tail tattoo for individual identification. Food intake and body weights were monitored weekly.

2.1.2. Glucose tolerance test

Following a 6 h fast, baseline glucose levels were measured by needle puncture of the tail of mice consuming RD or HFD for 3 wk (Bayer Breeze2, Bayer, Mishawaka, IN). Next, an intraperitoneal injection of glucose (1 g/kg) was administered and blood glucose levels measured again at each time point (0, 15, 30, 60, 90, 120 min) as previously described [30].

2.1.3. Behavioral assessment

Behavioral analysis was performed at baseline and following 4 or 12 wk of RD or HFD consumption. To examine spontaneous movement, mice were placed in a chamber (27 × 27 × 20.3 cm, Med Associates Inc., St. Albans, VT) equipped with infrared beams to record horizontal and vertical movements over a 30 min interval during the light phase. Mice were allowed to acclimate to the room conditions for 1 h prior to recording at 50 ms resolution in Activity Monitor Software. Equipment was cleaned with 70% ethanol between mice. The same testing chamber and period was also used to evaluate anxiety-like behavior by quantifying the amount of time a mouse spent in the center zone (4.5, 4.5 to 12, 12 cm) versus the residual area.

2.1.4. Immunohistochemistry

4% Paraformaldehyde-fixed, paraffin-embedded brain and lumbosacral spinal cord tissues were cut to 6 μm and utilized for all immunostaining procedures. Immunoperoxidase and immunofluorescence was performed as previously described [29]. Briefly, sections were washed, steamed in antigen retrieval solution (sodium citrate, 10 mM, pH 6.0), blocked (20% normal goat serum with 0.25% Triton-x), then incubated in primary antibodies for Olig2 (Millipore, MABN50, 1:200), CC-1 (Millipore, OP80, 1:50), Nkx2.2 (Developmental Studies Hybridoma Bank, University of Iowa, Iowa city, IA), NG2 (Millipore, ab5320), 4-HNE (Abcam, ab46545, 1:500), Bax (Invitrogen, MA514003, 1:200), Bcl-2 (Abcam, AB59348, 1:500), cleaved caspase-3 (Cell Signaling, 9664, 1:1000), GFAP (Abcam, ab4647, 1:2000) for 18 h. For immunoperoxidase staining, subsequent incubation with an appropriate biotinylated secondary antibody (Jackson, 1:200) followed by incubation with an avidin peroxidase solution (Vector) was used to yield a brown reaction product after incubation with 3,3-diamino-benzidine (DAB, Sigma D5637). Immunostained sections were counter stained with methyl green (Sigma, 67060), dehydrated, cover slipped using DPX Mountant (Sigma), and imaged using an Olympus BX51 microscope (Olympus, Center Valley, PA).

Immunofluorescence staining was achieved by incubating with appropriate fluorochrome-conjugated secondary antibodies (Jackson, 1:200) and counterstaining with 4′,6′-diamindino-2-phenylindole (DAPI). Cell counting and thresholding of images from the dorsal column of the spinal cord and corpus callosum of the brain was performed in ImageJ (NIH) without knowledge of treatment groups using cropped regions of interest averaged from 2 to 3 sections per animal with at least n = 4 per group.

2.1.5. Western blotting

One half of the spinal cord (cervical and thoracic) was homogenized in radio-immunoprecipitation buffer and 35 μg protein per sample was resolved by SDS-PAGE (Bio-Rad Laboratories, Hercules, CA) and electroblotted as described previously [29]. Membranes were probed for AMPK (Cell Signaling, 2532S), pAMPK (Cell Signaling, 2531S), SIRT1 (Abcam, Ab121193), PGC1α (Abcam, Ab54481), and β-actin (Novus Biological, NB600–501). Detection was accomplished using appropriate HRP-conjugated secondary antibodies (GE Healthcare), Pierce chemiluminescent developer (ThermoFisher), and ImageLab 2.0 Software (Bio-Rad). Relative optical density (ROD) for each protein was normalized to the actin of the same membrane.

2.1.6. RNA isolation, qPCR, and sequencing

Total RNA was isolated from half of the spinal cords or primary murine cells using RNA Stat60 (Tel-Test, Friendswood, TX) per manufacturer’s instructions and used for qPCR using BioRad iTaq Universal Probes (ThermoFisher) or primers (Integrated DNA Technologies) and SYBR Green One-step kits. RNA expression of each gene was normalized to Rn18s.

RNA libraries were prepared using 200 ng of total RNA according to the manufacturer’s instructions for the TruSeq RNA Sample Prep Kit v2 (Illumina, San Diego, CA). The concentration and size distribution of the completed libraries was determined using an Agilent Bioanalyzer DNA 1000 chip (Santa Clara, CA) and Qubit fluorometry (Invitrogen, Carlsbad, CA). Libraries were sequenced at 100 million reads per sample following Illumina’s standard protocol using the Illumina cBot and HiSeq 3000/4000 PE Cluster Kit. The flow cells were sequenced as 100 × 2 paired end reads on an Illumina HiSeq 4000 using HiSeq 3000/4000 sequencing kit and HCS v3.3.20 collection software. Base-calling was performed using Illumina’s RTA version 2.5.2. One regular diet-fed mouse was excluded because of poor sequencing.

mRNA-seq data was processed by the Mayo Bioinformatics Core Facility to identify genes with differential expression among groups. MAP-RSeq [31] workflow (v1.2.1.3) was used to process mRNA-seq data, including read alignment, quality control, gene expression quantification and finally summarizing the data across samples. Paired-end reads were aligned by TopHat (v2.0.12) against the mouse genome build (mm10) using the bowtie1aligner [32], and the gene counts were generated using Subread package [33] (v1.4.4). The workflow also provides detailed quality control metrics across genes using the RSeQC [34] (v2.3.2).

Differential expression analysis was performed with edgeR [35] package (v2.6.2) to identify genes with altered expression. A cutoff for false discovery rate-adjusted p-value was set at < 0.05 to determine the genes with significant expression changes between conditions. The differential genes were further submitted to Ingenuity Pathway Analysis (IPA®) for pathway enrichment analysis and causal network analysis. For causal network analysis (also described as “master regulator analysis”), only directed and experimentally observed interactions with a false discovery rate of < 0.1 were used. The overall heatmap represents RPKM values representing the entire gene profile differences between groups irrespective of FDR organized by hierarchical clustering with the row distance of one minus Pearson correlation, while the scatterplot depicts only significantly differentially expressed genes based on a FDR of < 0.15. Heatmaps derived from select PantherGo Pathways as well as the scatterplots from 500 gene sets for each cell type [36] were represented regardless of FDR.

2.1.7. Qualitative large-scale metabolite profiling and data analysis

Untargeted metabolomics analysis was performed by the Metabolomics Resource Core Facility at Mayo Clinic in spinal cord tissues of RD or HFD-fed animals by adapting previously described methods [37]. At study endpoints, spinal cords were quickly removed by flushing PBS through the vertebral column, split in half vertically and immediately snap frozen on dry ice for subsequent metabolic and transcriptomic analyses. Briefly, spinal cord samples that were stored at −80 °C were pulverized in liquid nitrogen using a custom mortar and pestle into a fine powder. Accurate weight was obtained using an analytical balance making sure the pulverized sample remained frozen. Pulverized samples were kept at −80 °C. Before metabolite extraction, samples were thawed on ice for 15 min and 1× PBS was added to make a 100 mg/mL homogenate. An aliquot of 50 μl (~5 mg tissue) was used for metabolite extraction. ¹³C₆-phenylalanine (2 μl at 250 ng/μL) was added as internal standard (IS) to all samples and tubes were sonicated in the ice bath for 2 min (30 s × 4 pulses). Samples were then deproteinated with cold acetonitrile: methanol (1:6 ratio), kept on ice with intermittent vortexing and centrifuged at 18,000 ×g for 30 min at 4 °C. The supernatant was divided into 2 aliquots and dried down using a stream of nitrogen gas for analysis on a Quadruple Time-of-Flight Mass Spectrometer (Agilent Technologies 6550 Q-TOF) coupled with an Ultra High Pressure Liquid Chromatograph (1290 Infinity UHPLC Agilent Technologies).

Profiling data was acquired under both positive and negative electrospray ionization conditions over a mass range m/z of 100–1700 at a resolution of 10,000 (separate runs) in scan mode. Metabolite separation was achieved using two columns of differing polarity, a hydrophilic interaction column (HILIC, ethylene-bridged hybrid 2.1 × 150 mm, 1.7 mm; Waters) and a reversed-phase C18 column (high-strength silica 2.1 × 150 mm, 1.8 mm; Waters) with gradient described previously [37]. Run time was 18 min for the HILIC and 27 min for the C18 column using a flow rate of 400 μl/min. A total of four runs per sample were performed to give maximum coverage of metabolites. Samples were injected in duplicate, wherever necessary, and a pooled quality control (QC) sample, made up of all of the samples from each study was injected several times during a run. A separate plasma quality control (QC) sample was analyzed with pooled QC to account for analytical and instrumental variability. Dried samples were stored at −20 °C until analysis. Samples were reconstituted in running buffer and analyzed within 48 h of reconstitution. Auto-MS/MS data was also acquired with pooled QC sample to aid in unknown metabolite identification using fragmentation pattern. The tissue pellet from pulverized samples remaining after the extraction was stored at −80 °C.

Total protein content from the tissue pellet was determined using Pierce BCA protein assay kit using a Spectramax Plus microplate reader. Briefly, the tissue pellet, after metabolite extraction, was solubilized in 0.3 M sodium hydroxide and 10 μl lysate was used to determine total protein concentration. Protein determination was performed in triplicate and bovine serum albumin (BSA) was used to generate the standard curve.

Data alignment, filtering, univariate, multivariate statistical and differential analysis was performed using Mass Profiler Professional (Agilent Inc., USA). Metabolites detected in at least ≥80% of one of two groups were selected for differential expression analyses. Metabolite peak intensities and differential regulation of metabolites between groups were determined as described previously [37]. Each sample was normalized to the protein content and log2 transformed. Unpaired t-test with multiple testing correction p < 0.05 was used to find the differentially expressed metabolites between two groups. Missing values are excluded from the analysis for the calculation of fold change and p-values. Default settings were used with the exception of signal-to-noise ratio threshold (3), mass limit (0.0025 units), and time limit (9 s). Putative identification of each metabolite was done based on accurate mass (m/z) against METLIN database using a detection window of ≤7 ppm. The putatively identified metabolites were annotated as Chemical Abstracts Service (CAS), Kyoto Encyclopedia of Genes and Genomes (KEGG), Human Metabolome Project (HMP) database, and LIPID MAPS identifiers. The differentially expressed metabolites are analyzed for pathway enrichment using MetaCore (Genego, St. Joseph, MI) – this identifies pathways using 3 metabolite identifiers – that is, KEGG, CAS and HMDB.

2.1.8. Targeted TCA cycle metabolomics

Spinal cord levels of tricarboxylic acid (TCA) cycle metabolites were measured using the same samples prepared for untargeted metabolomics after the derivatization with N-methyl-N-(t-butyldimethylsilyl)-trifluoroacetamide + 1% t-butyldimethylchlorosilane by gas chromatography/mass spectrometry under electron impact and selected ion monitoring conditions as previously described [37]. All measurements were performed against 12-point calibration curves that underwent the same derivatization with internal standard.

2.1.9. Mitochondrial functional and structural analyses

Oroboros oxygraph O2k was used for high resolution respirometry readings from isolated spinal cord mitochondria and Seahorse extracellular flux assay (Agilent) was used to assess mitochondrial function in cell cultures as described in more detail in our previous publications [38]. Oxygen consumption rates (OCR) were normalized to BCA protein estimation. In brief, fresh whole spinal cord tissue was placed in mitochondria isolation buffer (225 mM Mannitol, 75 mM sucrose, 20 mM MOPS, 1 mM EGTA, 0.5 mM DTT) containing 0.05% Nargase (w/v)/1 g tissue and homogenized for 10 min. Next, differential centrifugation was used to isolate mitochondria from other fractions, the mitochondria pellet was resuspended in mitochondria isolation buffer with 0.02% Digitonin for 10 min, then centrifuged and resuspended again in fresh buffer (125 μL/100 mg tissue) for running the assay. 50 μL of isolated mitochondria suspension was run for each of two technical duplicates. The oxygraph chamber was pre-washed and filled with respiration buffer (120 mM mannitol, 40 mM MOPS, 5 mM KH₂PO₄, 0.1 mM EGTA, 60 mM KCl, 5 mM MgCl₂) to equilibrate and oxygenate for at least 30 min. Oxygen calibrations for ambient air and zero were acquired for each chamber during each run. Oxygen consumption rates were measured in DatLab (Oroboros) after adding the mitochondria suspension (state I), glutamate (10 mM)/Malate (2 mM)/succinate (10 mM) mixture (state II), ADP (14.25 mM), oligomycin (2 μg/mL), FCCP (0.005 mM) titrations, and antimycin A (2.5 mM) using Hamilton syringes. State 3 mitochondrial respiration was assessed following the addition of glutamate, malate, and succinate and ADP to the chamber, giving the basal respiration rate for complexes I and II. Oligomycin injection then inhibited ATP synthase to induce state 4 respiration, attributable to proton leak across the inner mitochondrial membrane. FCCP addition results in an uncoupled state, allowing for determination of maximum capacity, while non-mitochondrial respiration was assessed after antimycin inhibits complex III. State III respiration measurements were taken from a stable segment following the addition of ADP, while State IV measurements followed oligomycin injection. Maximum respiration was recorded during the highest segment of uncoupled respiration following the FCCP titrations, and non-mitochondrial respiration was measured after antimycin A addition. All measurements were normalized to protein content of the mitochondria fraction.

The Seahorse XF24 (Agilent) extracellular flux assay was used to monitor oxygen consumption rates at various stages of respiration in primary murine oligodendrocytes as previously described [39]. Following cartridge equilibration with injection ports filled with oligomycin (1 μg/mL), FCCP (1 μM), and antimycin A (10 μM), a plate pretreated in serum-free media was placed into the Seahorse analyzer. OCR readouts at each stage were measured in pmol/min and normalized to protein.

For structural analysis of mitochondria, 0.09 × 10⁶ cells primary oligodendrocytes were grown per well on poly-L-lysine (PLL, 10 μg/mL) coated glass coverslips in 24-well plates, treated with 0 or 100 μM palmitate (PA) for 24 h, and then washed twice with fresh media. CMXRos MitoTracker Red (ThermoFisher, 100 nM) was incubated at 37 °C for 30 min. Immunocytochemistry procedures and DAPI counterstaining were then performed. ImageJ was used to quantify mitochondrial parameters as described previously [39,40]. Averaged values are reported from analyzed cells within five separate images per coverslip.

2.1.10. Primary glia isolation, OLC culture

Mixed glial preparations from postnatal day 0–3 murine cerebral cortex were cultured in DMEM (Fisher 11960–044) with 1 mM sodium pyruvate (Corning 11360070), 20 mM HEPES (Gibco 15630–080), 100 U/ml PenStrep (Life Technologies 15140122), and 10% fetal bovine serum (FBS, Atlanta Biologicals S11150) for 10–12 d. Oligodendrocytes, astrocytes, and microglia cultures were separated by differential adhesion/shaking overnight at 225 rpm on an Innova 2000 orbital shaker as previously described [41].

For oligodendrocyte treatments and differentiation, 0.09 × 10⁶ cells were plated on each well of PLL-coated coverslips or 24 well plates and grown in Neurobasal A media (Life Technologies, 10888022) with N2 (17502048), B27 (17504044), PenStrep, sodium pyruvate, 2 mM glutamax (35050–061), and 5% BSA (Sigma A4503) for two d before treating with palmitate (PA, 100 μM, 24 h). This dose was lower than toxic amounts previously reported in other cells types at the same time point [42,43] and did not result in significant lipotoxicity to oligodendrocytes or aNSCs. RNA levels for MBP, PLP, and Olig2 were determined in technical duplicates using and an iCycler iQ5 system (BioRad) with primers identified within the Key Resources Table. Relative starting quantities were normalized to Rn18S and reported as percent of control. PLP protein levels were determined based on threshold of immunoreactivity from five fields per coverslip quantified in ImageJ.

2.1.11. Adult neural stem cell culture

Adult neural stem cells were isolated from the subventricular zone (SVZ) of 8 wk old C57Bl6/J mice as previously described [44]. Briefly, mice were anesthetized with sodium pentobarbital and decapitated to collect coronal brain sections and further dissect stem cells from within the SVZ. Collected cells were dissociated with Accutase and grown in DMEM/F12 with B27, antibiotic-antimycotic, insulin (20 μg/mL), epidermal growth factor (EGF, 20 ng/mL) and basic fibroblast growth factor (bFGF, 20 ng/mL, PeproTech, Rocky Hill, NJ) in T75 tissue culture treated flasks as neurospheres.

For myelin RNA and protein expression experiments, cells were dissociated and plated onto poly-L-ornithine (20 μg/mL) and laminin (10 μg/mL) coated 6 well plates and glass coverslips, respectively. To induce differentiation towards oligodendrocyte lineage, cells were switched to media containing Neurobasal A, antibiotic-antimycotic, B27, insulin (20 μg/mL), T3 (60 ng/ml), glutamax (2 mM) and cultured for an additional 72 h. To identify effects of PA (100 μM) on differentiation to oligodendrocyte lineage, we compared protein and RNA marker expression to controls by immunofluorescence staining and qRT-PCR.

2.1.12. Statistical analysis

Data were tested for normality using the Shapiro-Wilk test, graphed, and analyzed by Student’s t-test unless noted otherwise in GraphPad Prism 7.0 software. Mann-Whitney nonparametric test was used for histological analysis of 4-HNE immunoreactivity in brain and spinal cord tissue, while body weight, food intake, and glucose tolerance test in mice were analyzed by two-way ANOVA with Bonferroni post-test. p-Values ≤ 0.05 were considered significant.

2.2. Key Resources Table

Reagent or resource	Source	Identifier
Antibodies
Mouse anti-Olig2	Millipore	MABN50; AB_10807410
Rabbit anti-BAX	Invitrogen	MA514003; AB_10979735
Rabbit anti-Olig2	Millipore	Ab9610; AB_10141047
Mouse anti-CC-1	Abcam	Ab16794; AB_443473
Rabbit anti-AMPK	Cell Signaling Technology	2532S; AB_330331
Rabbit anti-pAMPK	Cell Signaling Technology	2531S; AB_330330
Mouse anti-SIRT1	Abcam	AB12193; AB_298923
Anti-Nkx2.2	Developmental Studies Hybridoma Bank, University of Iowa, Iowa city, IA	74.5A5; AB_531794
Anti-neural glial antigen-2 (NG2)	Millipore	ab5320; AB_11213678
Rabbit anti-proteolipid protein (PLP)	Abcam	ab28486; AB_776593
Rat anti-myelin basic protein (MBP)	Millipore	MAB386; AB_94975
Rabbit anti-PGC1-α	Abcam	Ab54481; AB_881987
Mouse anti-β-actin	Novus Biological	NB600-501; AB_10077656
Rabbit anti-4-hydroxynonenol	Abcam	ab46545; AB_722490
Rabbit anti-cleaved caspase-3	Cell Signaling Technology	9664; AB_10694088
Rabbit anti-Bcl-2	Abcam	AB59348; AB_2064155
Rabbit anti-neurofilament	Sigma Aldrich	N4142; AB_477272
Biotinylated goat anti-rat secondary	Jackson	112-066-072; AB_2338185
Biotinylated donkey anti-rabbit secondary	Jackson	711-065-152; AB_2340593
Biotinylated donkey anti-mouse secondary	Jackson	715-066-151; AB_2340788
Goat-anti-rabbit A-F488 secondary	Jackson	111-545-047; AB_2338051
Goat anti-rat Cy3 secondary	Jackson	112-166-072; AB_2338257
Goat anti-rabbit A-F488 secondary	Jackson	111-545-041; AB_2341130
Donkey anti-rabbit Cy3 secondary	Jackson	711-166-152; AB_2313568
Experimental models: cell lines
Mouse:C57BL/6 oligodendrocyte lineage cells (OLCs)	Primary isolation (Yoon 2015)	N/A
Mouse:C57BL/6 adult neural stem cells (aNSCs)	Primary isolation (Choi 2017)	N/A
Experimental models: organisms/strains
Mouse: C57BL/6J	The Jackson Laboratory	N/A
Oligonucleotides (primers/probes)
OLIG2	Applied Biosystems	NM_016967; Mm.PT.58.42319010
MBP	Integrated DNA Technologies	NM_001025251; CCAGTAGTCCATTTCTTCAAGAACAT/GCCGATTTATAGTCGGAAGCTC
PLP	Integrated DNA Technologies	NM_011123.2; TCTTTGGCGACTACAAGACCAC/CACAAACTTGTCGGGATGTCCTA
PGC1-α	Applied Biosystems	Mm01208835_m1
FIS1	Integrated DNA Technologies	Mm.PT.56a.21878911
PINK1	Integrated DNA Technologies	Mm.PT.58.23711353
SOD2		Mm01313000_m1
GFAP	Integrated DNA Technologies	NM_010277.2; GCAGATGAAGCCACCCTGG/GAGGTCTGGCTTGGCCAC
SOD2	Applied Biosystems	NM_013671.3; Mm01313000_m1
Rn18s	Applied Biosystems	NR_003278.3; Mm03928990_g1
Chemicals, peptides, and recombinant proteins
Palmitic acid	Sigma Aldrich	Cat#P0500
DAPI	ThermoFisher	Cat#D1306; AB_2629482
Mitotracker Red	ThermoFisher	Cat#m7512
Commercial kits
Seahorse XF Cell MitoStress Test	Agilent	Cat#103015-100
Pierce BCA protein estimation	ThermoFisher	Cat#23225
Software and algorithms
Activity Monitor Software	Med Associates Inc.	https://www.med-associates.com/
GraphPad Prism 7	GraphPad Software	https://www.graphpad.com/scientificsoftware/prism
Fuji ImageJ	NIH	https://imagej.net/Fiji
Other
Regular diet (10 kcal % fat)	Research Diets	D12450K
High fat diet (60 kcal % fat)	Research Diets	D12492

2.3. Contact for reagent and resource sharing

Additional information and requests for resources can be directed to the corresponding author/Lead Contact, Isobel Scarisbrick (Scarisbrick.isobel@mayo.edu).

3. Results

3.1. Chronic high fat diet consumption impairs metabolic parameters and diminishes the number of oligodendrocytes and oligodendrocyte progenitors in the central nervous system

Since our prior studies revealed that a diet high in fat and sucrose results in a loss of myelinating cells and their progenitors in the spinal cord [29], we chose to investigate the differential impact of high fat alone and any temporal effects by feeding 10 wk old male C57Bl/6 mice a regular diet (RD, 10% of kcal from fat) or high fat diet (HFD, 60% kcal from fat) for either 4 or 12 wk (Fig. 1A). As expected, the HFD mice appeared noticeably larger (Fig. 1B) compared to the RD-fed controls and consumed more kcal of food per day (Fig. 1C). Beginning at 3 wk through 12 wk, mice also gained a significant amount of weight (Fig. 1D). After 3 wk, glucose sensitivity was assessed using a glucose tolerance test. As expected, HFD-fed mice demonstrated significantly higher fasting blood glucose levels and impaired glucose tolerance (Fig. 1E, F) [30]. A set of RD and HFD mice were monitored in a Comprehensive Lab Animal Monitoring System (CLAMS) to reveal significant changes in several metabolic parameters including percent fat/lean mass, VO₂, VCO₂, respiratory exchange rate (RER), and metabolic rate (Supplemental Fig. 1). In a separate cohort, a number of movement measurements were also affected by HFD consumption including anxiety-like behavior and decreased vertical motion at 4 wk, and also involving horizontal activity by 12 wk (Supplemental Fig. 2).

Fig. 1.

Chronic high fat diet consumption impairs metabolic parameters and diminishes the number of oligodendrocytes and progenitors in the central nervous system. A, Schematic depicting experimental design in which 10 wk old male mice were fed a regular diet (RD) or a high fat diet (HFD) for 4 or 12 wk. Representative images show mice consuming a HFD gained weight (B) after consuming more kcal food per day (C, n = 3 cages of 5 mice per group). HFD mice gained weight (D, n = 10 per group. RD 25.84 g and HFD 24.97 g at study start; RD 27.53 g and HFD 43.02 g at study end) and had impaired glucose tolerance after 3 wk (E, F; AUC = area under the curve, n = 8 mice per group). Graphical results represented as the mean ± SEM. G, Schematic shows spinal cord regions used in analyses. H, Representative images from immunostained sections of spinal cord dorsal column and counts of the number of Olig2+ cells at 4 wk (n = 5 mice per group). RNA expression of Olig2 (I), MBP (J), and PLP (K) at 12 wk (n = 4 mice per group). Representative images of dorsal column and corresponding counts of Nkx2.2+ (L), NG2+ (M), Olig2+ (N), and CC-1+ (O) cells after 12 wk (n = 7 mice per group). P, Schematic depicting region of corpus callosum examined for immunostaining. Representative images from immunostained sections of the corpus callosum (Q) and counts of Olig2+ cells at 12 wk (n = 5 mice per group). Bar graphs shown represent the mean number of cells per area of spinal cord ± SEM and are expressed as percent of RD-fed control mice. (*, p < 0.05; **, p < 0.01; and ***, p < 0.001). Scale bar = 20 μm (L–O) and 50 μm (Q).

After establishing the systemic and behavioral effects of HFD consumption in mice, we next tested the direct consequences on oligodendrocyte lineage cells (OLCs). At study endpoints, spinal cord tissues were separated into cervical-thoracic and lumbosacral segments for analysis as depicted in (Fig. 1G). The impact of consuming a HFD on myelin producing cells was determined by counting OLC markers in the post-fixed lumbosacral spinal cord of each group. Oligodendrocyte progenitor cells (OPCs) were identified using antibodies recognizing NG2, Nkx2.2, and Olig2 (oligodendrocyte lineage transcription factor 2, marks progenitors and early mature cells), while mature oligodendrocytes were considered CC-1⁺. After 4 wk of HFD, no change in the number of Olig2⁺ cells was observed in the spinal cord dorsal column (Fig. 1H). RNA expression of oligodendrocyte markers Olig2 (Fig. 1I), myelin basic protein (MBP, Fig. 1J), and proteolipid protein (PLP, Fig. 1K) remained unchanged after 12 wk of HFD-consumption. By contrast counts of oligodendrocyte progenitors immunoreactive for Nkx2.2 (Fig. 1L, 40×, and Supplemental Fig. 3A, 20×), NG2 (Fig. 1M, 40×, and Supplemental Fig. 3B, 20×) and Olig2 (Fig. 1N, 40×, and Supplemental Fig. 3C, 20×) were all reduced at 12 wk. Furthermore, counts of CC-1⁺ mature oligodendrocytes (Fig. 1O, 40×, and Supplemental Fig. 3D, 20×) demonstrated a significant depletion. We did not observe a significant change in Olig2 (Supplemental Fig. 4A, B), in MBP isoforms (Supplemental Fig. 1A, C and Supplemental Fig. 5E–H) or PLP (Supplemental Fig. 5A, D) protein expression by Western blot.

Considering we observed a loss of oligodendrocytes and their progenitors in the spinal cord at 12 wk of HFD consumption, we next examined any regional specificity by parallel quantification in the brain. The corpus callosum (Fig. 1P) from mice fed a HFD had fewer Olig2⁺ cells (Fig. 1Q) at 12 wk, but staining for MBP (Supplemental Fig. 3E), a marker of myelin structural integrity, and mature oligodendrocytes (CC-1⁺, Supplemental Fig. 3F) were unaffected. Taken together, these findings suggest that oligodendrocyte progenitors in both the brain and spinal cord are vulnerable to death in response to chronic HFD consumption; however, only spinal cord mature oligodendrocytes appeared susceptible to HFD-induced toxicity at the timepoint examined. Interestingly, remaining oligodendrocytes within both central nervous system (CNS) regions maintained normal PLP and MBP levels at all time points examined.

3.2. Chronic high fat diet consumption differentially regulates the transcriptional profile of the adult mouse spinal cord

To begin to address the mechanisms driving oligodendrocyte depletion in the face of HFD-consumption, we performed RNA sequencing from half of each spinal cord as demonstrated in Fig. 2A. An overall heatmap (Fig. 2B and Supplemental Fig. 4) depicts differential gene expression in the spinal cords of mice consuming a RD (n = 3) and HFD (n = 4) for 12 wk, where blue is the relative row minimum and red is the relative row maximum. Using a false discovery rate (FDR) of P < 0.15, a scatter plot of differentially expressed genes (DEGs, Fig. 2C) showed 75 to be transcriptionally distinct between diet groups (32 up and 43 down). Among these, we observed an upregulation of Hba-a2, Sdf2L1, Hmgcs2, Creld2, Cryab, Hspa5/Grp78, Grn, Kdr, Ddit4l, Fos, Xbp1, Paqr6, Fn14, Gstp1, Ccdc42ep2, and Litaf. Significantly downregulated genes included Lcn2, Ch25h, Sgk1, Slc17a7, Pdyn, Socs3, Plaur, Dbp, Acer2, Pla2g3, Kirrel2, Cp, Fn1, and Klf9.

Fig. 2.

Transcriptional changes in the spinal cord of adult high fat-fed mice. A, One half of the spinal cord from regular diet (RD, n = 3) and high fat diet (HFD, n = 4) fed mice were processed whole genome RNA sequencing. B, Heat map of transcriptional changes observed in the spinal cord of male mice fed RD or HFD for 12 wk representing the entire gene profile differences between groups irrespective of FDR, organized by hierarchical clustering with the row distance of one minus Pearson correlation where blue is relative row minimum and red is the row maximum. C, Scatter plot of differentially expressed genes (DEGs, red dots; 32 up, 43 down with a FDR < 0.15) where the x-axis is false discovery rate (FDR) and y-axis is log fold change (log FC), with annotated gene names for select up (red) and down (blue) regulated genes. PTK6 (D) was predicted to be activated while NFKBIA (E) was predicted as an inhibited upstream transcriptional regulator.

A number of ER stress induced genes including Creld2, Xbp1, and Hspa5 were upregulated in HFD spinal cord. Not surprisingly, ingenuity canonical pathway analysis revealed ER Stress and the Unfolded Protein Response pathways as top pathways affected in spinal cord tissue by HFD consumption (Table 1). Next, to identify transcription factors which may act as master regulators for the observed transcriptional changes, we performed a causal network analysis with only directed and experimentally observed interactions with a FDR < 0.1 (Table 2). PTK6 (Fig. 2D), predicted to be activated, and NFKBIA (Fig. 2E), predicted as an inhibited upstream transcriptional regulator, influence expression of Xbp1 and HSPA5, respectively. Moreover, PDGFRA is a crucial mediator of oligodendrogenesis [45] and was predicted to be inactivated. Loss of Mecp2 in oligodendrocytes down regulates myelin gene expression and was predicted to have lower activation in the HFD mice [46].

Table 1.

Summary of ingenuity canonical pathways significantly affected in spinal cord tissue by HFD consumption. Bold indicates increased.

Master regulator	Exp log ratio	Molecule type	Depth	Predicting activation Z-score	P-value of overlap	Network bias-corrected p-value	Target molecules in dataset
PTK6		Kinase	3	2.67	2.55E–06	0.013	CDC42EP2, CRYAB, GNA12, GSTP1, HBA1/HBA2, HMGCS2, KDR, PDIA6, XBP1, ARL4D, COL8A2, CP, FBN1, FN1, GRN, LCN2, LFNG
MYOCD		Transcription regulator	3	2.12	8.00E–06	0.012	ARC, GSTP1, KDR, XBP1, COL8A2, FBN1, FN1, LCN2
XBP1	0.38	Transcription regulator	1	2	5.69E–04	0.015	HSPA5, PDIA6, SDF2L1, FN1
TAF12	0.04	Transcription regulator	3	2	1.76E–04	0.022	HSPA5, KDR, SGK1, FN1
SRF	0.11	Transcription regulator	2	1.89	3.92E–05	0.026	ARC, GSTP1, KDR, LCN2, COL8A2, FBN1, FN1
NFYC	−0.03	Transcription regulator	2	1.73	1.06E–03	0.042	KDR, SGK1, FN1
CCNE1	0	Transcription regulator	3	1.5	7.52E–05	0.033	ARC, CRYAB, HBA1/HBA2, HMGCS2, HSPA5, KDR, SERPINB1, TNFRSF12A, XBP1, CH25H, COL8A2, DBP, FBN1, GRN, LAMA4, LRG1
CREB3	0.04	Transcription regulator	1	1.41	6.47E–05	0.0032	HSPA5, XBP1
MKL1	−0.07	Transcription regulator	3	1.41	6.39E–06	0.0094	ARC, CDC42EP2, GSTP1, KDR, LCN2, COL8A2, FBN1, FN1
ATF6B	0.10	Transcription regulator	1	1	6.30E–03	0.0318	XBP1
NFYC	−0.03	Transcription regulator	1	1	1.67E–02	0.0449	SGK1
PIK3CA	−0.06	Kinase	3	1	1.32E–05	0.049	ARC, CDC42EP2, CRYAB, GNA12, GSTP1, HMGCS2, HSPA5, PDIA6, XBP1, ARL4D, CIART, CP, DBP, LCN2, LFNG, SGK1
DDIT3	0.13	Transcription regulator	1	0.58	1.60E–04	0.008	HSPA5, XBP1, LCN2
PRKAB1	0.005	Kinase	2	0.45	1.11E–04	0.0435	HMGCS2, DBP, FN1, LCN2, SGK1
PRKAG1	0.002	Kinase	2	0.45	1.11E–04	0.0435	HMGCS2, DBP, FN1, LCN2, SGK1
RPS6KA4	−0.04	Kinase	2	0.33	6.49E–06	0.0083	ARC, GSTP1, HSPA5, CIART, COL8A2, FBN1, FN1, LCN2, SGK1
heme		Chemical - endogenous mammalian	3	0.33	3.15E–05	0.0162	ARC, GSTP1, HSPA5, PDIA6, XBP1, CIART, FN1, LCN2, SGK1
PRKACA	0.07	Kinase	3	0.23	7.34E–06	0.032	ARC, CRYAB, GSTP1, HMGCS2, HSPA5, KDR, SERPINB1, ARL4D, CH25H, CIART, COL8A2, CP, DBP, FBN1, FN1, GRN, LCN2, LFNG, Podxl
MAP3K3	−0.07	Kinase	3	0	3.62E–07	0.0008	ARC, CDC42EP2, CRYAB, GNA12, HSPA5, SDF2L1, SERPINB1, TNFRSF12A, ARL4D, CH25H, CIART, COL8A2, CP, DBP, FBN1, FN1, GRN, LCN2, LFNG, LRG1, PLA2G3, SGK1
CLOCK	0.02	Transcription regulator	1	0	3.33E–04	0.0049	HMGCS2, DBP
ARNTL	0.28	Transcription regulator	1	0	2.07E–3	0.0184	HMGCS2, DBP
PPRC1	−0.05	Transcription regulator	1	0	4.05E–03	0.0319	GNG11, LCN2
SMARCA2	−0.02	Transcription regulator	1	0	2.20E–03	0.0344	KDR, CP
PDGFRA	−0.12	Kinase	3	−0.23	1.19E–06	0.0153	CDC42EP2, CRYAB, GNA12, HBA1/HBA2, HSPA5, KDR, PDIA6, SERPINB1, XBP1, CH25H, COL8A2, CP, FBN1, FN1, GRN, LCN2, LFNG, SGK1
HTT	−0.04	Transcription regulator	3	−0.41	9.99E–07	0.0126	ARC, CDC42EP2, CRYAB, GNA12, GSTP1, HBA1/HBA2, HMGCS2, HSPA5, KDR, PDIA6, SDF2L1, SERPINB1, XBP1, ARL4D, CH25H, CIART, CP, DBP, FN1, GRN, LAMA4, LFNG, PLA2G3, SGK1
MECP2	−0.11	Transcription regulator	1	−0.58	1.25E–04	0.0021	HBA1/HBA2, SGK1, SLC17A7
CREB1	0.05	Transcription regulator	1	−0.82	2.90E–04	0.0314	ARC, HSPA5, CIART, FN1, LCN2, SGK1
FEZF2		Transcription regulator	1	−1	8.39E–03	0.0162	SLC17A7
PER2	−0.19	Transcription regulator	1	−1	1.46E–02	0.0341	DBP
MSGN1		transcription regulator	1	−1	1.67E–02	0.0408	LFNG
PASK	0.05	Kinase	2	−1.13	1.43E–04	0.0373	ARC, HSPA5, CIART, FN1, GRN, LCN2, SGK1
NFKBIA	−0.37	Transcription regulator	1	−1.41	6.19E–07	0.0025	CRYAB, HSPA5, CH25H, CP, FN1, GRN, LCN2, SGK1
NR1I2		Ligand-dependent nuclear receptor	1	−1.73	2.27E–03	0.0453	GSTP1, HMGCS2, LCN2
TAF11	0.07	Transcription regulator	3	−2	1.76E–04	0.0222	HSPA5, KDR, FN1, SGK1

Table 2.

Master regulator analysis shows significantly affected transcriptional regulators in HFD-fed mouse spinal cord. Bold indicates increased.

Ingenuity canonical pathways	−log(p-value)	−log(B-H p-value)	P-value	FDR	#DEGs	Molecules
Endoplasmic reticulum stress pathway	3.45	1.56	0.0004	0.028	2	XBP1, HSPA5
Aldosterone signaling in epithelial cells	2.86	1.29	0.0014	0.051	3	CRYAB, HSPA5, SGK1
Unfolded protein response	2.63	1.29	0.002	0.051	2	XBP1, HSPA5
Glutamate receptor signaling	2.58	1.29	0.003	0.051	2	GNG11, SLC17A7
Huntington’s disease signaling	2.4	1.21	0.004	0.062	3	GNG11, HSPA5, SGK1
Ceramide degradation	2.1	0.99	0.008	0.104	1	ACER2
Sphingosine and sphingosine-1-phosphate metabolism	1.97	0.94	0.011	0.115	1	ACER2
CCR3 signaling in eosinophils	1.89	0.94	0.013	0.115	2	GNG11, PLA2G3
Ketogenesis	1.88	0.94	0.013	0.115	1	HMGCS2
Mevalonate pathway I	1.76	0.87	0.017	0.134	1	HMGCS2

We next focused attention on expression of myelin and cholesterol related genes using PantherGO Pathways (Fig. 3A), where there was surprisingly a trend for increases in a number of genes, which may reflect a compensatory response in remaining oligodendrocytes and explain why frank loss of myelin was not observed at the 12 wk endpoint despite the loss of mature myelinating cells and progenitors. Having characterized overall transcriptomic differences between the spinal cord of mice consuming a RD or HFD, we next used the gene sets described by Zhang et al. [36] to profile transcriptional changes enriched at distinct stages of oligodendrocyte lineage progression (Fig. 3B–E), including oligodendrocyte precursors (OPCs, Fig. 3C), new OLCs (Fig. 3D), and myelinating OLCs (Fig. 3E) [36]. Of the OPC-enriched gene set, Slc6a3 (p = 0.27), Lox (p = 0.18), Cdc25c (p = 0.16), Snora17 (p = 0.015), Bglap2 (p = 0.79), Gsx1 (p = 0.07), Ccnb1 (P = 0.18), Alas2 (p = 0.10), Raet1d, and Troap (p = 0.19) were all increased above 1.5-fold, while Ddx43 (p = 0.002) and Cpz were below 0.5-fold in the spinal cord of mice consuming HFD relative to RD (Fig. 3C, all genes included regardless of FDR). New OL gene changes included increased Krt28 (p = 0.10), S100a14 (p = 0.19), Snora26 (p = 0.25), 1700001P01Rik (p = 0.10), and Sis (p = 0.08) and decreased Erp27 (p = 0.18), Corin (p = 0.09), and Chrng (Fig. 3D). In myelinating OLs, Hist1h2bg (p = 0.40), 1700001P01Rik (p = 0.10), A930003A15Rik (p = 0.09), Hist1h2ae (p = 0.43), Hist1h4i (p = 0.04), and Hist1h2ac (p = 0.34) were increased, while Hist1h2bn (p = 0.006), Erp27 (p = 0.18), and Mpl (p = 0.04) were decreased (Fig. 3E). Together, these results suggest that HFD consumption promotes complex transcriptional changes in the spinal cord and in oligodendrocyte lineage cells that include pathways essential for maintaining homoeostasis and highlight several important new targets for further investigation.

Fig. 3.

Myelin-related transcriptional changes in the spinal cord of adult high fat-fed mice. A, Heat map of myelin-related genes in the spinal cord of male mice fed RD or HFD for 12 wk where blue indicates relative minimum value and red is the row maximum value. Genes within groups are sorted by highest to lowest average HFD relative values compared to RD, with red asterisk indicating genes with p < 0.05 Schematic depicting differentiation of oligodendrocyte lineage cells (B) and scatterplots of RNA expression of genes enriched in oligodendrocyte progenitors (OPCs, C), new oligodendrocytes (OL, D) or myelinating OLs (E) in HFD mice normalized relative to RD controls.

3.3. Functional and metabolic changes in spinal cord mitochondria following HFD consumption are time-dependent

Prior to initiation of cell death, ER stress pathway activation can have a number of effects on metabolic and bioenergetic processes [28,47,48]. Therefore, since the ER Stress Pathway was the top ingenuity canonical pathway identified from transcriptomic profiling of spinal cord tissue, we further explored potential mechanisms for oligodendrocyte loss in this model by comparing metabolite profiles from the spinal cords of mice fed a RD or HFD using untargeted UPLC-ToF-MS-based metabolomics as described previously in detail [37].

HFD consumption significantly altered 238 metabolites analyzed using an unpaired t-test with Benjamini-Hochberg multiple testing correction and identified using the METLIN metabolite database. To determine which pathways were most affected by the metabolites altered due to chronic HFD-consumption, we performed metabolite sets enrichment analysis (MSEA) with MetabaCore depicted with most significant p-values in red and least significant in white (Fig. 4A). These results suggest that protein biosynthesis, TCA cycle, glutathione metabolism, and the mitochondrial electron transport chain were among the significantly enriched pathways. Accordingly, we sought to validate these findings by quantifying TCA cycle metabolites at 4 and 12 wk using a targeted GC–MS-based approach [37]. No significant differences were observed at 4 wk (Fig. 4B), but intermediates from the TCA cycle including oxaloacetate (p = 0.04), citric acid (p = 0.003), and α-ketoglutarate (p = 0.03) were strongly depleted in the 12 wk HFD spinal cord when compared to RD controls (Fig. 4B, C and Supplemental Fig. 7).

Fig. 4.

Functional and metabolic changes in spinal cord mitochondria following HFD consumption are time-dependent. Metabolite Sets Enrichment Overview (A) summarizing significantly affected pathways in HFD-fed mice as revealed by untargeted LC-MS-MS profiling of spinal cord tissue. Heat map (B) summarizes targeted LC-MS-MS profiling of tricarboxylic acid (TCA) cycle metabolites at 4 and 12 wk of RD or HFD consumption with metabolites depleted at 12 wk identified within the schematic by blue arrows (C). Gray indicates metabolite was not detected in samples, red indicated highest fold expression and blue lowest fold abundance. High resolution respirometry by Oroboros oxygraph of isolated spinal cord mitochondria indicates oxygen consumption is increased in 4 wk HFD mice (D, E), but decreased by 12 wk (D, E). Injections of glutamate/malate/succinate (GMS), ADP, oligomycin (O), and trifluoromethoxy carbonylcyanide phenylhydrazone (FCCP) allow for oxygen consumption rates (OCRs) to be determined during various stages of mitochondrial respiration. State 3 mitochondrial respiration was assessed following the addition of glutamate, malate, and succinate to the chamber, giving the basal respiration rate for complexes I and II. Oligomycin injection then inhibited ATP synthase to induce state 4 respiration, attributable to proton leak across the inner mitochondrial membrane. FCCP addition results in an uncoupled state, allowing for determination of maximum capacity, while non-mitochondrial respiration was assessed after antimycin inhibits complex III. OCRs during state 3 (D), state 4 (E), Maximum (F), and non-mitochondrial respiration (G). Graphical results represented as the mean ± SEM (*, p < 0.05; **, p < 0.01; and ***, p < 0.001; n = 4–6/group).

A separate cohort of mice was fed a RD or HFD for 4 or 12 wk to assess the functional impact on spinal cord mitochondria using Oroboros oxygraph [38]. Unexpectedly, we observed a significant increase in State 3 Respiration at 4 wk on HFD (Fig. 4D), while 12 wk HFD spinal cord mitochondria had diminished State 3 oxygen consumption compared to RD (Fig. 4D). Even State 4 respiration was increased at 4 wk, but declined at 12 wk of HFD consumption (Fig. 4E). Maximum (Fig. 4F) and non-mitochondrial respiration (Fig. 4G) were not altered by HFD consumption at 4 or 12 wk. Taken together, these data provide evidence for temporally distinct changes in TCA cycle metabolites and mitochondrial function, where at 12 wk TCA cycle metabolites are depleted and mitochondrial function is impaired.

3.4. Altered mitochondrial quality control processes, oxidative stress, and apoptosis in the spinal cord following 12 wk of high fat diet consumption

HFD consumption was previously shown to alter the metabolic regulators AMPK and SIRT1 in peripheral tissues and to affect mitochondrial function and quality control processes [49,50]. Therefore, we next explored this signaling pathway (Fig. 5A) in the spinal cord of RD and HFD-fed mice. With 12 wk of HFD, immunoblotting demonstrated reductions in AMPK phosphorylation (Fig. 5B, C) and in total AMPK (Fig. 5B, C). SIRT1 protein levels (Fig. 5B, C) were also significantly reduced in the spinal cord of HFD mice compared to controls. PGC1α, a direct target of AMPK and SIRT1 and crucial regulator of mitochondrial biogenesis, showed diminished protein levels (Fig. 5B, C) and RNA expression (Fig. 5D) in the HFD-fed mice compared to RD-fed controls. There was also an upregulation of the fission gene, Fis1 (Fig. 5D), and a reduction in the mitophagy-related gene PINK1 (Fig. 5D), which could indicate an overall accumulation of fragmented mitochondria in the spinal cord of HFD-fed mice.

Fig. 5.

Altered Mitochondrial Quality Control Processes, Oxidative Stress, and Apoptosis in the spinal cord following 12 wk of high fat diet consumption. Schematic (A) and Western blots (B) demonstrate significant reductions in AMPK phosphorylation, SIRT1, and PGC1α. C, Bar graphs show quantification of relative optical density (ROD) highlighting reductions in AMPK phosphorylation and total AMPK, SIRT1, and PGC1-α protein expression in the spinal cord of high fat diet (HFD)consuming mice compared to those fed a regular diet (RD). The first HFD mouse samples protein level changes were less affected than others due to animal variability despite consistent conditions between mice. D, Quantitative PCR demonstrated differences in RNA expression of PGC1α, Fis1, PINK1, and SOD2. E, Heat map summarizes RNAseq data for mitochondria-related transcriptional changes. Representative 4-hydroxynonenol (4-HNE) immunostained sections and corresponding quantification of immunoreactivity normalized to RD control revealing increased lipid peroxidation in the spinal cord (SC) (F) and corpus callosum (G) of HFD-fed mice by 12 wk. Immunofluorescence images and associated bar graphs show increased Bax/Olig2 double positive cells (H) and an increased Bax/Bcl2 ratio as well as more cleaved caspase-3/Olig2 double positive cells (I) in the white matter of spinal cord tissue from HFD-fed mice versus RD controls, indicating oligodendrocyte apoptosis. Graphical results represented as the mean ± SEM (*, p < 0.05, **, p < 0.01, and ***, p < 0.001 (n = 5–7 mice per group)). Scale bar = 50 μm (F), 100 μm (G), and 20 μm (H, I).

Oligodendrocytes are particularly vulnerable to cell death mediated by oxidative damage due to their relatively low antioxidant levels and high amounts of iron storage [51,52]. Notably, gene expression of the oxidative defense enzyme SOD2 (Fig. 5D) was significantly lower in HFD spinal cords compared to RD. Next, in a heatmap of mitochondrialrelated genes (Fig. 5E), we noticed a trend for increased Bax gene expression (p = .03, FDR = 0.63), while anti-apoptosis Bcl-2 appeared less (p = .69, FDR = 0.99) in the spinal cords of HFD mice versus RD controls. Free radical production results in oxidative modification of lipids which can damage biological membranes including the myelin sheath. 4-hydroxynonenal (4-HNE) lipid peroxidation products are found in demyelinating MS lesions and are highly toxic to OLCs [47,53,54]. White matter tracts within both the spinal cord dorsal column (Fig. 5F) and corpus callosum (Fig. 5G) of mice consuming HFD had significantly greater immunoreactivity for 4-HNE compared to RD controls. Furthermore, immunofluorescence staining demonstrated increased levels of pro-apoptotic protein Bax (Fig. 5H) and decreased levels of anti-apoptotic Bcl-2 (Fig. 5I) in the dorsal column of HFD-fed mice. Quantification of the Bax/Bcl-2 ratio (Fig. 5H) and double positive Olig2/cleaved-caspase-3 cells (Fig. 5I) suggest apoptotic cell death signaling is underway in oligodendrocytes chronically fed a HFD. These results point to HFD-related changes in key metabolic regulators such as AMPK and SIRT1, decreased mitochondrial biogenesis and an impaired potential for antioxidant responses as contributing factors to the increases in oxidative stress markers such as 4-HNE and loss of myelinating cells observed in mice consuming HFD.

3.5. Mitochondrial changes and differentiation in oligodendrocytes treated with saturated fat

To model HFD consumption in vitro and study potential direct effects towards oligodendrocytes, we cultured primary murine oligodendrocytes in media containing the saturated fatty acid palmitate (PA, 100 μM) for 24 h. Seahorse extracellular flux analysis (Fig. 6A) was performed to evaluate mitochondrial bioenergetics. PA significantly decreased basal, ATP-linked, and maximal oxygen consumption rates in oligodendrocytes, but had no effect on proton leak. Since we observed transcriptional changes in fission and mitophagy genes in the spinal cord of mice consuming a HFD, we also visualized the integrity and structure of oligodendrocyte mitochondria using MitoTracker. Healthy control oligodendrocytes have a long, filamentous network of mitochondria (Fig. 6B), while PA-treated oligodendroglia have mitochondria that are smaller, more compact, and circular. Overall, mitochondrial function and structure are impaired in oligodendrocytes cultured with high levels of saturated fat, in a manner recapitulating the changes observed in the spinal cord of HFD-fed mice (Fig. 4).

Fig. 6.

Differentiation and mitochondrial changes in oligodendrocyte progenitors cultured in the presence of high saturated fatty acid. High fat consumption was modeled in primary murine oligodendrocyte lineage cell (OLC) cultures by including palmitate (Palmitic acid) in the media (PA, 100 μM, 24 h). Bioenergetics characterization (A) of PA in OLCs was measured using a Seahorse XF24 analyzer. Basal, ATP-linked, and maximum capacity oxygen consumption rates were decreased in OLCs treated with PA. Significant changes in proton leak respiration were not observed. MitoTracker staining (B) showed increased mitochondrial fragmentation in PA-treated OLCs as determined by mitochondrial parameters including significantly increased solidity and circularity and a smaller average size. Proteolipid (PLP) immunoreactivity of OLCs (C) and qPCR quantification of RNA (D) for PLP, myelin basic protein (MBP) and Olig2 illustrate impaired differentiation towards a mature myelinating oligodendrocyte when treated with PA. Similarly PA-treated adult SVZ-derived neural stem cell (aNSC) cultures (M) also express less PLP protein (E) and MBP RNA (F). Graphical results represented as the mean ± SEM (n = 3–6/group; *, p < 0.05, **, p < 0.01, and ***, p < 0.001). Scale bar = 50 μm.

To determine the influence of PA on oligodendrocyte differentiation, PLP-immunoreactivity and RNA transcripts for PLP, MBP and Olig2 were quantified. Addition of PA (100 μM, 24 h) to differentiating primary oligodendrocyte cultures reduced PLP protein (Fig. 6C) to half that of controls. Supporting this, PLP, MBP and Olig2 RNA expression were also depleted (Fig. 6D). In addition to oligodendrocyte progenitors, neural stem cells also contribute to replacement of oligodendrocytes following injury [44,55]. Therefore, we next determined if the oligotoxic effects of PA extend to monolayer cultures of neural stem cells (aNSCs) derived from the adult subventricular zone (SVZ). Supporting deleterious effects of PA across the oligodendrocytes and neural stem cells, the abundance of PLP protein (Fig. 6E) and MBP RNA expression (Fig. 6F) were significantly less in aNSCs cultured in the presence of PA. Taken together, these findings suggest excess PA impairs the ability of OLCs or aNSCs to differentiate into myelinating oligodendrocytes.

4. Discussion

Diet-induced obesity increases the risk for a number of diseases including diabetes, cardiovascular disease, cancer, and cognitive dysfunction and has been steadily growing in prevalence in recent decades [20,56–59]. The role of oligodendrocytes and myelin plasticity is likewise implicated in a variety of neuropsychiatric and cognitive conditions [1,2,6]. Notably, white matter integrity is negatively correlated with BMI and is decreased in the corpus callosum of obese individuals [4,14,60]. Elevated BMI is also linked to an increased incidence and disease burden in MS, especially in combination with other risk factors [10,13,61,62]. Here we explored the potential link between a HFD, myelin integrity and metabolic dysfunction in the CNS. Our findings demonstrate for the first time that a HFD alone depletes oligodendrocyte progenitors in the brain and spinal cord and high saturated fat impedes oligodendrocyte differentiation in vitro. Overall, we observed changes in genes and pathways involving oligodendrogenesis and differentiation, iron homeostasis, oxidative stress, ER stress, insulin signaling, and inflammation all of which may contribute mechanistically alone or collectively to the loss of myelinating cells in the CNS of mice chronically consuming HFD. In the spinal cords of HFD fed mice, RNA sequencing and untargeted metabolomics point towards ER stress and mitochondrial dysfunction as key mechanisms contributing to oligodendrogliopathy. These findings were validated by results showing depletion of TCA cycle intermediates and reduced oxygen consumption rates in the spinal cords of HFD-consuming mice and in isolated oligodendrocytes. Moreover, deficits in AMPK-SIRT1-PGC1α signaling were present in HFD spinal cords, whereas the oxidative stress marker 4HNE was increased. Overall, the vast influence of metabolic dysfunction on oligodendrocytes observed here is of high relevance to demyelinating, neuropsychiatric and cognitive disorders and to considerations for strategies to manage myelin protection and repair.

The results of the current study suggest that consumption of excess high fat alone is sufficient to trigger oligodendrogliopathy and impairments in oligodendrocyte differentiation across the brain and spinal cord and links these to several disturbances in metabolic status. Oligodendrocytes play critical roles in modulating motor [63,64] and psychiatric [5,65] function by improving the speed and integrity of impulse conduction, providing neuronal metabolic support, and regulating synapses [8]. In a recent study, we discovered that a diet high in saturated fat (41%) and sucrose (34%), the so-called Western diet, depletes oligodendrocytes and their precursors after 7 wk of consumption in the adult (4 month) spinal cord [29]. A later study, reported neuroinflammation, cerebrovascular decline, and white matter damage with an increase in Olig2+ progenitors, but a decrease in MBP in the frontal parietal cortex/corpus callosum following long term (10 month) consumption of a “Western-style” diet (16.4% calories from fat with added high-fructose corn syrup) [4]. Importantly, in both prior studies examining the effects of a high fat high sugar diet on myelin, exercise prevented the deleterious effects. The findings of the current study suggest that consumption of high fat alone is sufficient to trigger demyelinating effects in the brain and spinal cord of adult mice. Additional study will be needed to determine if the demyelinating effects of HFD can likewise be prevented by exercise interventions.

Many of the transcriptional changes we observed are increased in the spinal cord of mice consuming a HFD are already established to be differentially expressed in individuals with MS and its experimental models. Fos, Grn, Sdf2l1, Fn14, and Paqr6 RNA accumulation are each separately found in glia of white matter tissue of MS patients, and Grn polymorphisms may influence MS disease course and relapse recovery [66–72]. Down regulated genes in the spinal cord of HFD mice were also affected in MS animal models, include Slc17a7 and Pdyn [73]. Although Cryab was shown to inhibit inflammation and disease in the EAE model, recent studies in the Cuprizone model suggest that Cryab activates astrocytes and exacerbates demyelination [74]. Moreover, activation of the Sgk1 pathway following oxidative stress exacerbates EAE [75]. Among many genes identified by Moyon et al. in a study of transcriptional profiling of Cuprizone-“activated” oligodendrocytes, higher expression of Socs3, Ddit4, and lower expression of Kdr and Cdc42ep2 genes may be required to activate OPCs enabling remyelination [76]. Kruppel like factor 9 (Klf9), down regulated by HFD, is a transcription factor downstream of T3 that regulates oligodendrocyte differentiation and myelin regeneration in the Cuprizone model [77,78]. Ddit3 is expressed by activated astrocytes following Cuprizone exposure, while knockout mice had less oligodendrocyte apoptosis, demyelination, and microglial and astrocyte activation [79]. HFD also exacerbates disease symptoms and neuroinflammation in EAE models [21,22]. Notably, the genes identified as having higher expression in HFD-EAE versus RD-EAE did not overlap with those we observed in the spinal cord of adult mice consuming HFD, indicating unique mechanisms contribute to HFD-triggered pathology depending on the disease context [22].

ER stress and inflammation closely follow lipid dysregulation in diabetic neuropathy and obesity-associated allodynia models and cooperate in a feed forward manner [80–82]. In line with studies showing ER stress in liver, heart, muscle, hypothalamus, and hippocampus in diet-induced obesity models [7,48,83,84], RNA sequencing results in the current study uncovered ER Stress and the Unfolded Protein Response as prominently altered pathways in the adult spinal cord after 12 wk of HFD consumption. Since myelinating cells produce large amounts of plasma membrane, they are particularly susceptible to disruptions in secretory pathways including ER stress and eventual apoptosis if left unchecked [3,26,85]. Thus, controlling ER overload may be beneficial to restoring oligodendrocytes and myelin in various white matter diseases [3,26,85,86]. Xbp1, involved in ER stress responses, is reported to increase in MS lesions as well as in vanishing white matter disease (VWMD) [3,26]. ER chaperones help control cellular responses to ER stress and some were shown to be required for oligodendrocyte survival during EAE [85]. ER and mitochondria interact both physically and functionally, and alterations in these connections have recently been described in MS pathology [27].

Alterations in ER stress are connected to mitochondrial deficits in MS [27]. Mitochondrial morphology and impaired mitochondrial function and antioxidant responses are also reported in experimental models of demyelinating disease [87–89]. Here we document fragmented mitochondria in cultures of oligodendrocytes grown under high fat conditions. Fis1 is involved in mitochondrial fission and in early events in cell-death signaling, and expression was also upregulated in the spinal cord of mice consuming high fat. Similar changes are likewise engaged in OLCs in the context of oxidative stress or inflammation [40,89–91]. Moreover, inhibition of fission is protective against demyelination and neuroinflammation, although has little impact on OLC differentiation in EAE and Cuprizone MS models [89,92]. Impairments in mitochondria also lead to increased reactive oxygen species and downstream effects on proteins, lipids, and nucleic acids [92–94]. Notably, in both the corpus callosum and spinal cord white matter regions where OLC death occurred following chronic HFD consumption, we observed increased lipid peroxidation. Increased lipid peroxidation was also observed in the spinal cord with consumption of high fat and sucrose [29]. In MS pathogenesis and its animal models, oxidative stress mediates apoptosis of oligodendrocytes and is commonly documented by an elevated Bax/Bcl2 ratio and caspase-3 cleavage [95,96]. Our in vivo data suggest that mitochondrial-dependent apoptosis also occurs in oligodendrocytes in relation to HFD-induced mitochondrial dysfunction and increased oxidative damage.

In the current study, we discovered substantial deficits in TCA cycle intermediates and oxygen consumption rates in the spinal cords of HFD mice. Key substrates at all stages of the TCA cycle, from citric acid to oxaloacetate were reduced in the spinal cord of mice consuming HF. In addition, these effects were paralleled by diminished mitochondrial oxidative respiration. Oligodendrocytes have extremely high energy demands and die hours faster than neurons following nutrient deprivation by arterial occlusion [28,97]. Impaired energy metabolism has already been proposed as a contributing factor to oligodendrogliopathy in pattern III MS lesions and is described in animal models of the disease [87,89,93]. Normally, metabolism of pyruvate and other fuel sources through the mitochondrial TCA cycle leads to ATP synthesis via the electron transport chain. Our findings that the dysmyelinating effects of HFD were associated with significant impairments in TCA cycle function are of particular interest given the efficacy of dimethyl fumarate, itself a TCA cycle metabolite, as a first line treatment for relapsing remitting MS [98].

AMPK, SIRT1 and PGC1α were all substantially reduced in the spinal cord of mice consuming HF. AMPK and SIRT1 are crucial metabolic sensors that impact PCG1α and thus mitochondria biogenesis and function. In addition to promoting cell survival, AMPK-SIRT1 activation and mitochondrial function regulate oligodendrocyte differentiation [50,99–102]. Recently, folate was determined to regulate oligodendrocyte survival and differentiation through AMPK activity [99]. SIRT1 and SIRT2 activation redundantly mediate differentiation of neural stem or progenitor cells towards the oligodendrocyte lineage, likely by upregulation of PDGFRα, Sox10, Nkx2.2 and downregulation of p21 [9,103,104]. Moreover, SIRT1 activation through overexpression or agonists is neuroprotective in MS models, while expansion of oligodendrocyte progenitors occurs following SIRT1 inactivation [101,104]. Exercise and other activators of the AMPK-SIRT1-PGC1α system may be beneficial in reducing deleterious effects of obesogenic diets on oligodendrocyte lineage cells and myelin integrity by promoting mitochondrial homeostasis [4,29,49,101]. Interestingly, we observed a predicted decrease in PDGFRα signaling in the transcriptomic data along with decreased AMPK-SIRT1 signaling, leading to a loss of both progenitors and mature oligodendrocytes in the spinal cord following long term HFD. Whether or not the disruptions we observe in TCA cycle metabolite depletion, impaired mitochondrial function and oligodendrocyte loss observed in HFD-fed spinal cords are a direct consequence of the altered AMPK-SIRT signaling, as specified in other tissues and cell types, should be more directly addressed in future studies.

Mature CC-1+ oligodendrocytes along with Olig2+ progenitors are lost in the spinal cord white matter with HFD consumption, but only the later are impacted in the corpus callosum at the same 12 wk time point. Given their increased energy demands for proliferation and myelin membrane production, it is possible that oligodendrocyte progenitors and young oligodendrocytes as labeled by Olig2 are more vulnerable to the stresses associated with HF consumption compared to mature oligodendrocytes. We cannot exclude that mature oligodendroglia in the corpus callosum would be lost with even more chronic HFD feeding. Region specific and possible temporal differences in HF-elicited oligotoxicity may also relate to heterogeneity in the local oligodendrocyte populations, region specific effects on neurons, or differential effects across the astrocyte and microglial compartments that so significantly contribute to the microenvironment [41]. Our studies also demonstrate an anxiety-like phenotype and overall activity level changes in the HFD-fed mice that mirror a number of previous reports [20,58], however future studies will be required to determine if HFD-induced behavioral changes are in fact dependent on the oligodendrocyte loss documented. Although we show that impaired differentiation and mitochondrial fragmentation occurring with HFD are recapitulated across regions and in cultures of oligodendrocytes, a Ribotag approach or single cell sequencing will be beneficial for the identification of cell-specific transcriptional changes in each tissue, and this is another goal for future experiments.

5. Conclusions

The findings of the current study demonstrate significant perturbations in CNS function occur in response to long term consumption of a diet laden with excessive fat, including direct impairment of oligodendrocyte mitochondrial function that ultimately impairs oligodendrocyte survival and differentiation. Previous reports in Western diet-fed mice suggest exercise may mitigate deleterious effects on myelin integrity and behavioral outcomes, and our new data in mice fed a high fat diet suggest that modulation of mitochondrial function, ER and oxidative stress pathways, in addition to the TCA cycle may also be important areas to explore as targets to improve myelin protection and regeneration. Accordingly, this work should inspire new research to evaluate ways to mitigate the negative consequences of a chronic high fat diet on oligodendrocytes alone and in the context of white matter injury.

Abbreviations:

4-HNE

4-hydroxynonenal

aNSCs

adult neural stem cells

CLAMS

Comprehensive Lab Animal Monitoring System

EDSS

expanded disability status score

EAE

experimental autoimmune encephalitis

HFD

high fat diet

MetSyn

metabolic syndrome

MS

multiple sclerosis

MBP

myelin basic protein

OLC

oligodendrocyte lineage cell

OPCs

Oligodendrocyte progenitor cells

Olig2

oligodendrocyte lineage transcription factor 2

PLP

proteolipid protein

RER

respiratory exchange rate

RRMS

relapsing remitting MS

SPMS

secondary progressive MS

TCA

tricarboxylic acid

VWMD

vanishing white matter disease