Critical Role of Astrocyte NAD⁺ Glycohydrolase in Myelin Injury and Regeneration

Abstract

Western-style diets cause disruptions in myelinating cells and astrocytes within the mouse CNS. Increased CD38 expression is present in the cuprizone and experimental autoimmune encephalomyelitis models of demyelination and CD38 is the main nicotinamide adenine dinucleotide (NAD⁺)-depleting enzyme in the CNS. Altered NAD⁺ metabolism is linked to both high fat consumption and multiple sclerosis (MS). Here, we identify increased CD38 expression in the male mouse spinal cord following chronic high fat consumption, after focal toxin [lysolecithin (LL)]-mediated demyelinating injury, and in reactive astrocytes within active MS lesions. We demonstrate that CD38 catalytically inactive mice are substantially protected from high fat-induced NAD⁺ depletion, oligodendrocyte loss, oxidative damage, and astrogliosis. A CD38 inhibitor, 78c, increased NAD⁺ and attenuated neuroinflammatory changes induced by saturated fat applied to astrocyte cultures. Conditioned media from saturated fat-exposed astrocytes applied to oligodendrocyte cultures impaired myelin protein production, suggesting astrocyte-driven indirect mechanisms of oligodendrogliopathy. In cerebellar organotypic slice cultures subject to LL-demyelination, saturated fat impaired signs of remyelination effects that were mitigated by concomitant 78c treatment. Significantly, oral 78c increased counts of oligodendrocytes and remyelinated axons after focal LL-induced spinal cord demyelination. Using a RiboTag approach, we identified a unique in vivo brain astrocyte translatome profile induced by 78c-mediated CD38 inhibition in mice, including decreased expression of proinflammatory astrocyte markers and increased growth factors. Our findings suggest that a high-fat diet impairs oligodendrocyte survival and differentiation through astrocyte-linked mechanisms mediated by the NAD⁺ase CD38 and highlights CD38 inhibitors as potential therapeutic candidates to improve myelin regeneration.

SIGNIFICANCE STATEMENT Myelin disturbances and oligodendrocyte loss can leave axons vulnerable, leading to permanent neurologic deficits. The results of this study suggest that metabolic disturbances, triggered by consumption of a diet high in fat, promote oligodendrogliopathy and impair myelin regeneration through astrocyte-linked indirect nicotinamide adenine dinucleotide (NAD⁺)-dependent mechanisms. We demonstrate that restoring NAD⁺ levels via genetic inactivation of CD38 can overcome these effects. Moreover, we show that therapeutic inactivation of CD38 can enhance myelin regeneration. Together, these findings point to a new metabolic targeting strategy positioned to improve disease course in multiple sclerosis and other conditions in which the integrity of myelin is a key concern.

Introduction

Obesity is a key risk factor for developing multiple sclerosis (MS; Gianfrancesco et al., 2014; Olsson et al., 2017). Moreover, recent studies suggest that obesity and related metabolic comorbidities also influence disease progression and disability burden in individuals affected by MS (Salem et al., 2014; Castro et al., 2019). Although obesity-induced white matter (WM) abnormalities and loss of oligodendrocytes within the CNS following chronic consumption of high-fat diets (HFDs) has been found in rodents, how diet influences WM integrity in humans and interacts with MS pathophysiology is largely unknown (Yoon et al., 2016; Graham et al., 2019; Langley et al., 2020).

Astrocyte activation occurs in the CNS following chronic consumption of a HFD (Buckman et al., 2015; Kim et al., 2020) and can exert oligotoxic effects (Liddelow et al., 2017). Building and maintenance of myelin membranes requires oligodendrocyte fatty acid synthesis, yet oligodendrocytes can also use lipids from the diet or those synthesized by astrocytes (Dimas et al., 2019). Targeting astrocytes to improve outcomes in MS has already been suggested since some currently prescribed MS disease-modifying therapies impact inflammatory or oxidative stress signaling in astrocytes (Ponath et al., 2018). Prior studies regarding the role of HFDs in demyelination models have used immune-mediated models such as experimental autoimmune encephalitis (EAE) and demonstrated exacerbation of peripheral immune responses or neuroinflammation with HFD consumption (Timmermans et al., 2014; Herrmann et al., 2016; Hasan et al., 2017; Ji et al., 2019). There are few studies regarding the effects of HFD on the capacity for remyelination. 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.

CD38 is a nicotinamide adenine dinucleotide (NAD⁺)-degrading enzyme expressed by astrocytes and was recently shown to be upregulated in peripheral immune- and toxin-mediated demyelination models (Zhang et al., 2014, 2016; Herrmann et al., 2016; Roboon et al., 2019). CD38 is the main NAD⁺ase in many mammalian tissues and regulates tissue and cellular NAD⁺ homeostasis as well as NAD⁺-dependent enzymes such as poly(ADP ribose) polymerases and sirtuins (Aksoy et al., 2006a,b). NAD⁺ is a key mediator of cellular energy status and is involved in mitochondrial function and quality control processes, primarily through the activity of sirtuins that require NAD⁺ as a cofactor (Aksoy et al., 2006a,b; Camacho-Pereira et al., 2016). Exercise increased NAD⁺ levels in various tissues (Chiang et al., 2015) and protected against Western-style diet-induced deficits in myelin and myelinating cells in two recent studies (Yoon et al., 2016; Graham et al., 2019). Global CD38 knckout confers protection against demyelination, T-cell responses, and EAE disease severity, as well as HFD-induced metabolic dysfunction in rodent models (Chiang et al., 2015; Herrmann et al., 2016; Roboon et al., 2019; Guerreiro et al., 2020).

We tested the hypotheses that HFD promotes oligodendrocyte loss in a CD38-dependent manner and that CD38 inhibition will enhance myelin regeneration. First, we established increased CD38 expression following HFD consumption. Our previous studies indicate that HFD leads to a reduction in oligodendrocytes and their progenitors within the mouse CNS (Yoon et al., 2016; Kim et al., 2020; Langley et al., 2020). Next, we demonstrated that CD38 catalytically inactive (CD38ci) mice fed HFD were protected from HFD-mediated loss of oligodendrocytes with concomitant increases in NAD⁺ levels and reduced neuroinflammation and oxidative damage. In vitro approaches identified oligodendrocyte-linked direct and astrocyte-linked indirect mechanisms by which excessive fatty acid impairs myelin protein production. Finally, use of a small-molecule inhibitor of CD38, 78c, was then identified as a novel protective strategy to enhance myelin regeneration in a model of focal demyelinating injury. Together, these findings highlight a novel indirect pathway of myelin injury involving pathogenic changes in astrocytes. Moreover, these changes can be attenuated by inhibition of CD38 in the context of chronic HFD consumption and targeted to improve myelin regeneration in the CNS.

Materials and Methods

Materials.

See Table 1 for key resources and reagents.

Table 1.

Key resources table

Reagent or Resource	Source	Additional information
Antibodies
Mouse anti-CD38	Novus	1g7f4 (1:500)
Rabbit anti-GFAP	Dako	Z0334 (1:5000)
Mouse anti-OLIG2	Millipore	MABN50 (1:200)
Rabbit anti-OLIG2	Millipore	Ab9610; AB_10141047 (1:500)
Rabbit anti-GST3	Abcam	Ab153949 (1:1000)
Rat anti-MBP	Millipore	MAB386; AB_94975 (1:750)
Rabbit anti-PLP	Abcam	A28486 (1:750)
Rabbit anti-IBA1	Wako	019–19741(1:3000)
Rabbit anti-4HNE	Abcam	ab46545 (1:500)
Chicken anti-GFAP	Abcam	ab4647 (1:2000)
Goat anti-C3D	R&D Systems	AF2655 (1:500)
Rabbit anti-S100a10	Proteintech	11250–1-AP (1:7500)
Mouse anti-SERPING1	Santa Cruz Biotechnology	sc377062 (1:200)
Rabbit anti-EMP1	Abcam	ab202975 (1:200)
Rabbit anti-ASPA	Millipore	ABN1698 (1:1000)
Biotinylated goat anti-rat secondary	Jackson ImmunoResearch	112–066-072 (1:100)
Biotinylated donkey anti-rabbit secondary	Jackson ImmunoResearch	711–065-152 (1:100)
Biotinylated donkey anti-mouse secondary	Jackson ImmunoResearch	715–066-151 (1:100)
Goat-anti-rabbit AF488 secondary	Jackson ImmunoResearch	111–545-047 (1:100)
Goat anti-rat Cy3 secondary	Jackson ImmunoResearch	112–166-072 (1:100)
Gat anti-chicken AF647 secondary	Jackson ImmunoResearch	703–606-155 (1:100)
Oligonucleotides (primers/probes)
BDNF	Thermo Fisher Scientific	Mm04230607_s1
SOD2	Thermo Fisher Scientific	Mm01313000_m1; NM_013671.3;
S100a10	IDT	Mm.PT.58.6571055
CD38	Thermo Fisher Scientific	Mm01220906_m1
OLIG2	Thermo Fisher Scientific	NM_016967; Mm.PT.58.42319010
GFAP	IDT	NM_010277.2; GCAGATGAAGCCACCCTGG/ GAGGTCTGGCTTGGCCAC
IL-1β	IDT	Mm.PT.51.17212823
IL-6	Thermo Fisher Scientific	NM_031168.1; Mm00446190_m1
H2D1	IDT	Mm.PT.58.42136026.g
TNFα	Thermo Fisher Scientific	Mm00443258_m1
Rn18s	Thermo Fisher Scientific	NR_003278.3; Mm03928990_g1
GAPDH	Thermo Fisher Scientific	Mm99999915_g1
Chemicals
78c CD38 inhibitor	Millipore	Catalog #538763
NAD+	Sigma-Aldrich	Catalog #10127965001
NMN	Sigma-Aldrich	Catalog #N3501
EX-527	Tocris Bioscience	Catalog #2780
Nicotinamide 1,N-6-etheno adenine dinucleotide	Sigma-Aldrich	Catalog #41628
Flavin mononucleotide	Sigma-Aldrich	Catalog #F8399
Alcohol dehydrogenase	Sigma-Aldrich	Catalog #A3263
Diaphorase	Sigma-Aldrich	Catalog #D5540
Resazurin	Sigma-Aldrich	Catalog #119303
PA	Sigma-Aldrich	Catalog #P0500
DAPI	Thermo Fisher Scientific	Catalog #D1306; AB_2629482
LL	Sigma-Aldrich	Catalog #L-4129
Methyl green	Sigma-Aldrich	Catalog #67060
Commercial kits
Bio-Plex Mouse Cytokine 23-plex Assay	BIO-RAD	Catalog #m60009rdpd
RNeasy Mini kit	Qiagen	Catalog #74104
Software
R	The Comprehensive R Archive Network	https://cran.r-project.org/
Real Time Analysis and CASAVA	Illumina	http://support.illumina.com/sequencing/sequencing_software/casava.html
Prism 7	GraphPad Software	https://www.graphpad.com/scientific-software/prism/
CellSens	Olympus	https://www.olympus-lifescience.com/en/software/cellsens/
Zen	Zeiss	https://www.zeiss.com/microscopy/us/products/microscope-software/zen.html
Fiji ImageJ	NIH	https://imagej.net/Fiji

Details of resources and reagents used in experiments.

Ethics statement.

All animal experiments, diets, and procedures were approved by the Mayo Clinic Institutional Animal Care and Use Committee and were performed in accordance with institutional and regulatory guidelines. The human pathology study was approved by the Mayo Clinic Institutional Review Board (IRB #2067–99).

Animal husbandry and diets.

Mice were provided ad libitum access to diets along with tap water. For initial characterization of CD38 expression and NAD⁺ levels following HFD, male C57BL6/J mice were obtained from The Jackson Laboratory and were randomly assigned to either regular diet [RD (10% kcal fat); catalog #D12450K, Research Diets] or HFD (60% kcal fat; catalog #D12492, Research Diets) beginning at 10 weeks of age, as previously described (Langley et al., 2020).

For studies comparing wild-type (WT) and CD38ci mice fed RD and HFD, mice were bred onsite after initial generation at TransViragen via recombineering technology, as previously described (Partida-Sánchez et al., 2001; Tarragó et al., 2018). Briefly, CD38ci mice have the CD38 gene knocked out and reinserted with a mutation at E230 of the CD38 gene, rendering the CD38-NAD⁺ase catalytic site inactive, while preserving overall expression levels of CD38 (Tarragó et al., 2018). Male mice received a RD (10.4% kcal fat; catalog #TD000357, Teklad) or HFD (42% kcal fat; catalog #TD88137, Teklad) beginning at 50 weeks of age for 30 weeks. At this age, CD38 protein levels and NAD⁺ase activity are found to be significantly elevated (Camacho-Pereira et al., 2016). 78c is a thiazoloquin(az)olin(on)e CD38 inhibitor that increases NAD⁺ levels in various tissues and cell types and is orally bioavailable (Haffner et al., 2015; Tarragó et al., 2018). The same diets were used for lysolecithin (LL) demyelination studies, with or without 600 ppm 78c, a CD38 inhibitor (Calico Life Sciences), and were started 1 week before the spinal cord microinjection procedure. Food intake and body weights were monitored weekly. Diet composition and in vivo experimental design are summarized in Table 2.

Table 2.

Diet composition and in vivo experimental designs

Strain	Age at start	Diet composition	Diet duration	Injury	Age at end	Tissues analyzed	Injection
C57BL6/J	10 weeks	RD (10% kcal fat; catalog #D12450K, Research Diets) HFD (60% kcal fat; catalog #D12492, Research Diets)	12 weeks		22 weeks	Spinal cord
CD38ci, wild type	50 weeks	RD (10.4% kcal fat; catalog #TD000357, Teklad) HFD (42% kcal fat; catalog #TD88137, Teklad)	30 weeks		80 weeks	Spinal cord
C57BL6/J	6 weeks	RD (10.4% kcal fat; catalog #TD000357, Teklad) HFD (42 % kcal fat; catalog #TD88137, Teklad) RD + RD78c (600 ppm) HFD + HFD78c (600 ppm)	8 weeks, 8 weeks, 5 + 3 weeks, 5 + 3 weeks	1% LL to spinal cord	14 weeks	Spinal cord, injured and uninjured sides
RPL22HA/+: ALDH1L1Cre-ERT2	9 weeks	Chow	15 weeks		15 weeks	Corpus callosum	TAM (100 mg/kg, 5 d) 78c (10 mg/kg, i.p., 4 weeks)

Details of diets and treatments administered to mice. TAM, Tamoxifen.

Lysolecithin spinal cord microinjection procedure.

Male mice were selected for LL microinjection procedures to reduce variability because of progesterone-mediated enhancement of myelin repair and modulation of CNS neuroinflammatory responses (Ghoumari et al., 2020). Young adult mice have previously yielded reliable, reproducible lesions and good survival in this model from studies in our laboratory and others (Keough et al., 2015; Yoon et al., 2017b, 2020). In addition, the spinal cord, compared with the corpus callosum (CC), has large WM tracts that are more frequently myelinated axons, which allows for more confident quantification of remyelinated axons (i.e., thin myelin sheaths and unmyelinated axons can be difficult to differentiate from demyelinated axons; Pavelko et al., 1998; Keough et al., 2015; Yoon et al., 2017a, 2020).

For focal demyelination of the ventral WM of the murine spinal cord (laminectomy at thoracic 10th vertebra; mediolateral, 0.3 mm; dorsoventral, −1.4 mm relative to the dorsal surface), 12-week-old male mice were injected with 2 µl of a saline solution containing 1% lysophosphatidyl choline with 0.01% Evans blue dye (catalog #L4129, Sigma-Aldrich) at 0.2 µl/min using a 30–40 µm glass micropipette and stereotaxic microinjection device (Stoelting).

For anesthesia, ketamine (100 mg/kg) and xylazine (10 mg/kg) were administered intraperitoneally (i.p.) before surgery, while postoperative care consisted of an antibiotic (enrofloxacin, 50 mg/kg/d, i.p.), an analgesic (buprenorphine, 0.05 mg/kg, s.c., twice daily), and saline (0.2 ml/d, i.p.) for 3 d. After a 2 week recovery period, mice were deeply anesthetized (ketamine, 150 mg/kg, i.p.; xylazine, 15 mg/kg, i.p.) and transcardially perfused with 4% paraformaldehyde (catalog #T353-500, Thermo Fisher Scientific) for collection of spinal cord tissue.

Spinal cord segments were microdissected, with the rostral 2 mm segment processed for paraffin-embedded blocks and the caudal segment osmicated (1% osmium tetroxide; catalog #0223B, Polysciences) and embedded in Araldite (Ted Pella). To identify myelin sheaths, Araldite blocks were cut into 1 µm sections, stained with p-phenylenediamine (catalog #P6001, Sigma-Aldrich), imaged with 60× objective lens (model BX51 microscope, Olympus), and stitched together (Photoshop, Adobe) to comprise the entire lesion within the stained tissue section (Pavelko et al., 1998; Yoon et al., 2020). The number of remyelinated axons per square millimeter of lesion was reported after blinded counting in FIJI ImageJ (National Institutes of Health; Pavelko et al., 1998; Yoon et al., 2017b, 2020). Immunohistochemical analysis of focal demyelinated lesions used overall neuropathologic changes (hematoxylin and eosin) along with myelin loss [based on myelin basic protein (MBP) immunoreactivity] to determine the lesion core, while the lesion rim was defined as the region encompassing the entire border of neuroinflammatory markers [glial fibrillary acidic protein (GFAP); ionized calcium-binding adapter molecule 1 (IBA1); Pavelko et al., 1998]. Lesion areas between 50,000 and 180,000 µm² were included in the analysis based on our previous studies using this model (Pavelko et al., 1998; Yoon et al., 2017b, 2020). Analyses were performed with the experimenter blinded to the treatment group by assigning each animal a random number.

Immunohistochemistry.

Spinal cords were fixed in 4% paraformaldehyde, processed, embedded in paraffin, and cut to 6 µm for all immunoperoxidase and immunofluorescence procedures, as described previously (Langley et al., 2020; Yoon et al., 2020). For all antibodies except GFAP and MBP, heat-induced antigen retrieval was achieved through incubation with sodium citrate, 10 mm, pH 6.0, in a steamer basket. Sections were blocked (20% v/v normal goat serum in PBS), then incubated in primary antibodies (4°C, overnight), as described in Table 1.

The following day, incubation with appropriate secondary antibodies (biotinylated or fluorochrome conjugated, 1:200; Jackson ImmunoResearch) for immunoperoxidase or immunofluorescence staining, respectively, was performed at 21°C for 1 h. After incubating in an avidin peroxidase solution (Vectastain Elite ABC kit, catalog #PK-6100, Vector Laboratories), color development was achieved by incubation with 3,3′-diaminobenzidine (catalog #D5637, Sigma-Aldrich), and counterstaining was achieved with methyl green (catalog #67060, Sigma-Aldrich). Slides were dehydrated, coverslipped with dibutylphthalate polystyrene xylene (DPX, Sigma-Aldrich), and images acquired using a microscope (model BX51, Olympus) with a camera (model DP72, Olympus). Immunofluorescence slides were counterstained with 4′,6′-diamidino-2-phenylindole dihydrochloride (DAPI), mounted with fluoromount, imaged with a model BX51 microscope, and captured with a camera (model XM10, Olympus). Cell and axon counting, as well as thresholding of images, was performed in ImageJ. The cropping of regions of interest was performed in Photoshop, using the Allen Brain Atlas (Spinal Cord section) to determine the regions of interest boundaries. An average of two sections per animal (60 µm apart) was taken without knowledge of treatment groups. Only darkly stained cells were counted during quantification of oligodendrocyte markers.

Formalin-fixed, paraffin-embedded, 5-µm-thick sections from confirmed MS and normal CNS control autopsy cases were immunostained for CD38 (1:2000; catalog #LS-C817841, LifeSpan BioSciences). Steamed antigen retrieval was performed with high pH target retrieval solution (catalog #K8004, Dako; Chuang et al., 2016).

RNA isolation and qRT-PCR.

To determine gene expression from spinal cord or primary murine cells, total RNA was extracted using RNA STAT60 (Tel-Test) per manufacturer instructions and was used for quantitative real-time PCR (qRT-PCR) using iTaq Universal Probes or primers (Table 1) and SYBR Green One-Step Kits (BIO-RAD). The starting quantity for quantification of the expression of each gene was normalized to Rn18s or glyceraldehyde-3-phosphate dehydrogenase (GAPDH). Detailed primer and probe information are annotated in Table 1.

Mixed glial preparations and primary cell cultures.

Primary mixed glial cultures were prepared from postnatal day 0 (P0) to P3 mice, as previously described (Langley et al., 2020; Yoon et al., 2020). In brief, dissected cerebral cortices were dissociated and grown in DMEM (catalog #11960–044, Thermo Fisher Scientific) with 1 mm sodium pyruvate (catalog #11360070, Corning), 20 mm HEPES (catalog #15630–080, Thermo Fisher Scientific), 100 U/ml penicillin-streptomycin (PenStrep; catalog #15140122, Thermo Fisher Scientific), 10% fetal bovine serum (catalog #S11150, R&D Systems), and 5 µg/ml insulin. Following 10–12 d in the culture, sequential immunopanning was performed to separate astrocyte, microglia, and oligodendrocyte progenitor cells by agitating at 225 rpm on an orbital shaker (model Innova 2000, New Brunswick Scientific). Purity of the cell types cultured was within the ranges reported previously by our group (Burda et al., 2013; Radulovic et al., 2016).

After removing microglia via differential adhesion, purified oligodendrocyte progenitors were plated onto poly-l-lysine-coated glass coverslips and cultured in differentiation medium (Neurobasal A medium; catalog #10888022, Thermo Fisher Scientific) with N2 (catalog #17502048, Thermo Fisher Scientific), B27 (catalog #17504044, Thermo Fisher Scientific), PenStrep, sodium pyruvate, 2 mm GlutaMAX (catalog #35050–061, Thermo Fisher Scientific), 40 ng/ml T3, and 5% bovine serum albumin (catalog #A4503, Sigma-Aldrich) for 2 d. Treatments with a saturated fat, palmitate (PA; 100 µm), NAD⁺ (50 µm), nicotinamide mononucleotide (NMN; 100 µm), a SIRT1 inhibitor (EX-527) (5 µm) and/or a CD38 inhibitor (78c; 3 µm) were added, and cultures were allowed to differentiate for 72 h. Doses used in cell culture experiments were based on previous studies (Valle et al., 2014; Tarragó et al., 2018) and preliminary dose–response experiments. Positive staining for myelin proteins was compared with control levels by immunofluorescence staining and threshold quantification in ImageJ. Briefly, a positive threshold for each stain was decided then applied to each image (five per coverslip) to determine the percentage of field positivity. Mean values were graphed as a percentage of the control group.

In experiments characterizing astrocyte responses and effects of astrocyte-conditioned media (ACM) on oligodendrocytes, astrocytes were plated at a density of 450,000 cells/well on poly-l-lysine-coated six-well plates and grown for 2–3 d in serum-free Neurobasal A medium containing 1% N2, 2% B27, PenStrep, 1 mm sodium pyruvate, 0.45% glucose, 5% bovine serum albumin, and 50 µm β-mercaptoethanol before treating with or without PA (100 µm) and 78c (3 µm). Astrocytes were collected for an NAD⁺ cycling assay or RNA isolation. Supernatants from treated astrocytes were collected 24 h post-treatment, concentrated using Vivaspin sample concentrators (10 and 100 kDa molecular weight cutoffs; Sigma-Aldrich), and stored frozen as a 10× protein concentrate until used for treating oligodendrocyte lineage cells (Liddelow et al., 2017; Yoon et al., 2020). ACM was brought back to 1× concentration in oligodendrocyte differentiation medium and placed on oligodendrocytes to differentiate, as described above.

NAD⁺ cycling and CD38 activity assays.

To determine intracellular NAD⁺ levels, one well of a six-well plate of astrocytes, or ∼20 mg of spinal cord tissue, was homogenized in 300 µl of 10% trichloroacetic acid in water. Samples were extracted using an organic solvent mixture (1,1,2-trichloro-1,2,2-trifluoroethane and triocylamine), diluted, and measured along with a standard curve, as previously described (Tarragó et al., 2018). Briefly, a cycling reagent (0.76% ethanol, 4 µm flavin mononucleotide, 27.2 U/ml alcohol dehydrogenase, 0.44 U/ml diaphorase, and 8 µm resazurin; Sigma-Aldrich) was prepared in phosphate buffer (100 mm NaH₂PO₄/Na₂HPO₄, pH 8.0) and added to an equal volume of prediluted cell or tissue extract. Plates were measured every 30 s for 1 h at 544 nm using a SpectraMax plate reader. Raw values were compared with a standard curve then divided by protein quantification (Bradford Reagent, BIO-RAD) to be expressed in nanomoles per milligram of protein.

For measuring CD38 NAD⁺ase activity, spinal cord tissue was homogenized in NETN buffer [100 mm NaCl, 1 mm EDTA, 20 mm Tris-HCl, and 0.5% Nonidet P-40 (NP-40); Sigma-Aldrich] and supplemented with 5 mm NaF (Sigma-Aldrich) and protease and phosphatase inhibitors (Thermo Fisher Scientific; Tarragó et al., 2018). After centrifugation (12,000 rpm, 10 min), supernatant volumes were adjusted to protein concentrations and a reaction mix containing 50 µm 1,N6-etheno NAD⁺ in 0.25 m sucrose buffer was added at an equal volume just before reading the plate at 300 nm excitation and 410 nm emission every minute for 1 h on a plate reader (Spark Magellan, Tecan).

Bio-Plex assay for cytokines.

Astrocyte supernatants from cell treatments described above were used for the determination of secreted cytokine levels using a Bio-Plex mouse cytokine 23-plex assay (catalog #m60009rdpd, BIO-RAD), per manufacturer instructions. Briefly, post-treatment medium was collected and incubated with primary antibody-conjugated magnetic beads for 30 min. Wells were washed three times and then incubated for 30 min with detection antibodies. After washing, samples were incubated for 10 min with a streptavidin-phycoerythrin solution, and then washed again before resuspension and reading of fluorescence with a Bio-Plex 200 reader and Bio-Plex Manager software (BIO-RAD).

Organotypic cerebellar slice culture and demyelination.

Organotypic cerebellar slice cultures were cultured and demyelinated using previously described methods (Kim et al., 2020). Cerebella were dissected from pups at P7 to P9 and embedded in low-melting point agar (catalog #16520050, Thermo Fisher Scientific), and 350 µm slices were collected into artificial CSF solution using a McIlwain tissue chopper (Mickle Laboratory Engineering Co. Ltd). Four slices were transferred and grown per each Millicell insert (catalog #PIHP03050, Sigma-Aldrich) in 50% minimum essential medium, 22.5% HBSS, 24% heat-inactivated horse serum, 1.5% glucose, 1% GlutaMAX, and 1% penicillin-streptomycin medium.

Demyelination was induced after 7 d in culture by treatment with 0.5 mg/ml LL (Sigma-Aldrich) for 18 h. Organotypic slices were placed back in LL-free medium and collected for immunohistochemistry and confocal imaging after 24 h (demyelination) or 7 d (myelin regeneration) with or without addition of the saturated fat PA (100 µm) and/or 78c (3 µm) added to the growth media. Confocal images were acquired at 20× magnification from five regions randomly selected per slice using an LSM 780 Microscope and Zen Software (Zeiss) in the Mayo Clinic Microscopy and Cell Analysis Core. Images were thresholded for each stain, and stain-positive immunoreactive area or cell counts per total field area were quantified in ImageJ by normalizing to DAPI and neurofilament.

Generation of RPL22^HA/+: ALDH1L1^Cre-ERT2 mice and astrocyte-specific translatome.

Female RiboTag mice [B6J.129(Cg)-RPL22^{tm1.1Psam/SjJ}; catalog #029977, The Jackson Laboratory] were crossed with ALDH1L1^Cre-ERT2 males [B6;FvB-Tg(ALDH1L1^Cre-ERT2)1Khakh/J; catalog #029655, The Jackson Laboratory] to generate the RPL22^HA/+: ALDH1L1^Cre-ERT2 mice used for experiments (Srinivasan et al., 2016). Male ALDH1L1^Cre-ERT2:RPL22^HA-RiboTag mice were given tamoxifen for 5 days (100 mg/kg, i.p., in corn oil) at P60. At P80, a cohort of mice were given 78c (10 mg/kg, i.p.) or vehicle (5% dimethylsulfoxide, 15% polyethylene glycol 400, 80% of 15% hydroxypropyl γ-cyclodextrin in citrate buffer) for 4 weeks (n = 3/group; Tarragó et al., 2018).

At P108, mouse tissues were harvested and hemagglutinin (HA) immunoprecipitation (IP) and RNA isolation procedures were performed, as previously described (Sanz et al., 2019; Wheeler et al., 2020). Briefly, the corpus callosum from each mouse was quickly microdissected on ice, weighed, and homogenized in 5% w/v prechilled homogenization buffer with supplements [containing 1% NP-40, 100 mm KCl, 100 mm Tris-HCl, pH 7.4, and 12 mm MgCl₂, and supplemented with 100 µg/ml cycloheximide, protease inhibitor cocktail (1:100), 1 mg/ml heparin, RNase inhibitors, and 1 mm DTT] in a Dounce homogenizer. Up to 1.2 ml of tissue homogenate was transferred to a microcentrifuge tube and centrifuged at 10,000 rpm for 10 min. Supernatant was collected and portioned for IP, and an aliquot for the input/bulk RNA fraction was stored at −80°C.

For IP, samples were incubated with mouse anti-HA antibody (BioLegend) for 4 h at 4°C with end-over-end inversion. Pre-equilibrated beads were then added to the sample–antibody mixture and incubated overnight at 4°C with inversion. High salt buffer was used for three washes to remove nonspecific binding from the IP sample solution by using a magnetic stand to capture ribosome–RNA complexes bound to magnetic Dynabeads. RNA extraction for IP and input fractions was achieved with RNeasy Kit (Mini, Qiagen) with on-column DNase digestion, per manufacturer instructions.

RNA quantification (fluorimetric RiboGreen Assay, Thermo Fisher Scientific) and integrity analysis (High Sensitivity RNA ScreenTape, TapeStation 3.2, Agilent) were assessed. Samples with an average RNA integrity number of 7.7 were used for the SMARTer Stranded Total RNA-Seq– Pico Mammalian Kit version 2 library preparation (catalog #634414, Takara Bio). Library size distribution was validated by capillary electrophoresis and quantified by PicoGreen before indexed libraries were normalized and pooled for sequencing on a NovaSeq 6000 System using a 150 PE flow cell (20 million reads/sample; performed at the University of Minnesota Genomics Center, Minneapolis, MN). Base call files were generated by Real Time Analysis software (Illumina), while primary analysis and demultiplexing were performed using CASAVA version 1.8.2 software (Illumina) to generate FASTQ files.

For secondary bioinformatics analysis, the MAPRSeq version 3.1.3 secondary workflow including alignment to reference genome (STAR; Dobin et al., 2013), quality control (RSeQC; Wang et al., 2012), expression quantification (Subread; Liao et al., 2013), and variants, as well as differential gene expression analysis (edgeR_3.28.1; Robinson et al., 2010) was performed by the Mayo Clinic Bioinformatics core (Kalari et al., 2014). Differentially expressed genes were defined as p < 0.05, log-fold change (logFC) > |1|, and false discovery rate (FDR) < 0.1 (Benjamini–Hochberg test). Volcano plot and heatmaps with hierarchical clustering were created using R package ggplot2 (The Comprehensive R Archive Network; Wickham et al., 2016). Heatmaps for select genes were generated in Prism 7 (GraphPad Software). Ingenuity pathway analysis (IPA; Qiagen) was used to create network maps and to perform canonical pathway and upstream regulator analyses.

Statistics.

Data were graphed with bars representing the mean ± SEM with the sample size for each experiment indicated in the figure legends. To determine sample size, power analysis using the fpower macro in SAS was used to determine animal numbers for each end point based on 80% power, α = 0.05, and four treatment groups. Differences and SD used to determine δ were used from the analysis of previous similarly designed studies in our laboratory. Cell culture studies used a combination of male and female pups and were repeated twice; assays were run in technical duplicate or triplicate. No statistical outliers were removed, but the ROUT test in Prism was used to check for outliers. Data normality was assessed using the Shapiro-Wilk test. Comparisons between two groups were analyzed by Student's t test, whereas one-way or two-way ANOVAs with Bonferroni post-test were applied for multiple comparisons. Analysis was performed in Prism 7 software, with p values ≤ 0.05 considered to be significant.

Results

CD38 expression in astrocytes of mice fed chronic HFD, in a model of focal demyelination, and in MS lesions

Considering the emerging role of astrocytes in the pathology of various neurologic disorders, in the rodent CNS following chronic HFD consumption, and in interacting metabolically and functionally with oligodendrocytes, we decided to investigate the potential indirect effects of HFD mediated by astrocytes (Fig. 1A). Our recent untargeted metabolomics profiling of spinal cord tissue identified NAD⁺ metabolism as one of the top pathways impacted by HFD consumption (Langley et al., 2020), which prompted us to quantify spinal cord NAD⁺ levels using a cycling assay. Importantly, spinal cord NAD⁺ levels from HFD-fed mice were less than half that of RD-fed mice (p = 0.029, t = 2.639, df = 8; Fig. 1B).

Figure 1.

CD38 expression in astrocytes of chronic HFD-fed mice in a model of focal demyelination, and in MS lesions. A, Schematic representation of previously shown direct effects of HFD on oligodendrocyte lineage cells including apoptosis and impaired differentiation with potential astrocyte-mediated indirect effects, represented by dotted lines, that comprise the hypothesis to be tested. B, Spinal cord tissue of RD and HFD mice showing a depletion of NAD⁺ after 12 weeks (n = 5 mice/group) as determined by cycling assay. C, Increased RNA expression of CD38 (n = 4), assessed by qRT-PCR, in spinal cord tissue of HFD-fed mice when compared with RD-fed mice. D, E, Immunofluorescence staining with double-labeling and corresponding quantification of immunoreactivity and cell counts indicates increased CD38 (green) immunoreactivity in astrocytes (GFAP, red) within the spinal cords of HFD-fed mice (D, n = 7) or in the LL model of experimental demyelination (E, n = 4) versus the uninjected (UI) side. The broken white line designates the lesion boundary. Scale bars: D, 50 µm; E, 100 µm. F–H, High-fat consumption was modeled in primary murine astrocyte cultures by including the saturated fat PA (100 μm, 24 h) in the media. Increases in RNA expression of GFAP (F; n = 4) and CD38 (G; n = 6) were observed in PA-containing astrocyte cultures. H, Depleted NAD⁺ levels were found in PA-treated astrocytes when assessed using a NAD⁺ cycling assay (n = 7–8). I, Representative microscopic images of CD38 staining in control brain tissue and an MS lesion with active demyelinating activity. Scale bar, 50 µm. Bar graphs represent the mean ± SEM. Asterisks indicate levels of statistical significance as determined by Student's t test: *p < 0.05; **p < 0.01; ***p < 0.001.

We followed up by quantifying expression of CD38, an NAD⁺-glycohydrolase, in spinal cord astrocytes to establish a potential link. Quantification by qRT-PCR indicated higher RNA expression of CD38 in HFD-fed mice compared with RD-fed controls (p = 0.001, t = 5.626, df = 6; Fig. 1C). Further, increases in CD38 immunoreactivity were observed in GFAP-positive astrocytes following HFD (p = 0.038, t = 1.937, df = 12; Fig. 1D). HFD-mediated effects mirror the increased CD38 expression following LL-induced demyelination shown in Figure 1E (p = 0.004, t = 5.105, df = 6). Similar to HFD spinal cords, primary murine astrocytes cultured with the saturated fat PA (100 µm, 24 h) showed increased GFAP (p = 0.046, t = 2.507, df = 6; Fig. 1F) and CD38 RNA expression (p = 0.005, t = 3.576, df = 10; Fig. 1G) and depleted NAD⁺ levels (p = 0.011, t = 2.962, df = 13; Fig. 1H). Collectively, these data suggest that HFD can induce astrocytosis in the spinal cord or cultures, which increases the expression of CD38 and subsequently depletes NAD⁺ levels.

Next, to confirm the human relevance of this target, we performed immunohistochemical analysis on brain sections from control and MS lesion tissue. Microscopic imaging demonstrated extensive CD38 immunoreactivity in WM astrocytes within control brain tissue, as well as increased CD38 in some hypertrophic reactive astrocytes within MS lesion tissue with active demyelinating activity (Fig. 1I). The expression and significance of CD38 immunoreactivity among various astrocyte subtypes and disease phenotypes needs to be further characterized but is likely not specific to MS.

Resistance to HFD-induced oligodendrocyte loss in CD38ci mice

To determine the potential therapeutic significance of CD38 upregulation to HFD-induced oligodendrocyte loss, we collected spinal cords from WT and CD38ci mice fed a RD or a HFD for 30 weeks beginning at 50 weeks of age for histologic analysis (Fig. 2A). To confirm that preservation of NAD⁺ levels was achieved in the CD38ci mice, we next measured NAD⁺ levels on spinal cord tissue from the same cohort of mice. As expected based on our data in Figure 1, NAD⁺ levels were significantly reduced following chronic HFD consumption in WT mice (F_(1,24) = 1.785, p = 0.004; Fig. 2B). Moreover, in CD38ci mice, which had diminished CD38 activity (F_(1,16) = 75.13, p < 0.0001; Fig. 2C), NAD⁺ levels were significantly higher overall (F_(1,24) = 33.90, p < 0.0001) and in the CD38ci-HFD mice compared with the WT-HFD mice (p < 0.0001).

Figure 2.

CD38ci mice are resistant to HFD-induced oligodendrocyte loss. Histologic analysis from WT and CD38ci fed a RD or HFD for 30 weeks beginning at 50 weeks of age. A, Schematic depicts experimental design. B, NAD⁺ levels as determined by cycling assay indicate reductions in WT-HFD mice compared with WT-RD mice and increased NAD⁺ levels in the CD38ci mice versus WT mice. C, CD38 activity levels determined by etheno-NAD⁺ assay show impaired activity in CD38ci mice versus WTs and higher activity in WT-HFD mice than WT-RD mice (n = 5 mice/group). D, Outlined regions depict regions imaged, including dorsal column (broken gray line) and ventral WM (broken black line). E, PDGFRα double-immunofluorescent staining with Ki67 or cleaved-caspase-3 (ClCasp3) shows a depletion of oligodendrocyte progenitor cells with the HFD that was significantly attenuated in CD38ci mice versus WT mice. Increased cleaved-caspase-3 and PDGFRα-colabeled cells were found in WT-HFD versus WT-RD mice, while fewer cleaved-caspase-3 and PDGFRα-colabeled cells were observed in the CD38ci-HFD than in the WT-HFD ventral WM. F, G, Olig2-positive oligodendrocyte lineage cells (F) and GST3-positive mature oligodendrocytes (G) were significantly depleted in HFD-fed WT mice, yet CD38ci mice fed a HFD had no reduction in oligodendrocyte numbers. When compared with WT-HFD mice, CD38ci-HFD mice had significantly more oligodendrocytes remaining in the dorsal column (representative images pictured). Scale bar, 50 µm. Gray lines indicate genotype means, while gray broken lines indicate comparison of overall genotype effect. H, Representative images of immunohistochemical staining confirming no significant changes in MBP immunoreactivity across genotype or diet groups. Scale bar, 100 µm. I, PLP immunostaining demonstrating loss of immunoreactivity in the HFD group of WT mice, but not CD38ci mice. Representative images pictured. Scale bar, 100 µm. J, K, Gene expression of myelin proteins MBP (J) and PLP (K) as determined by qRT-PCR. Bar graphs represent the mean ± SEM. Two-way ANOVA followed by Bonferroni post hoc test (n = 7–10 mice/group). Asterisks indicate levels of statistical significance when compared with controls: *p < 0.05; **p < 0.01; ***p < 0.001. L, Representative images and bar graph of fluorescent triple labeling of cells across the oligodendrocyte lineage. Early-stage marker (PDGFRα, red), Olig2 (cyan), and mature oligodendrocyte marker (ASPA, green) with DAPI (blue) nuclear stain and merged image. Colored asterisks indicate significance by two-way ANOVA followed by Bonferroni post hoc test (n = 5 mice/group) for the corresponding color cell marker combination group. Scale bar, 50 µm.

Upon histologic examination of oligodendrocyte lineage cell markers in these mice, there was a significant reduction in the number of platelet-derived growth factor receptor α (PDGFRα)-positive (Fig. 2E: dorsal column: F_(1,16) = 16.61, p = 0.0003; ventral WM: F_(1,16) = 17.77, p = 0.0007) and oligodendrocyte transcription factor 2 (Olig2)-positive oligodendrocyte progenitors (Fig. 2E: dorsal column: F_(1,26) = 9.066, p = 0.006; ventral WM: F_(1,26) = 24.90, p < 0.0001) as well as glutathione S-transferase 3 (GST3)-positive mature oligodendrocytes (Fig. 2F: dorsal column: F_(1,22) = 14.53, p = 0.001; ventral WM: F_(1,25) = 10.23, p = 0.004) in the HFD-fed WT mice compared with RD-fed mice. However, CD38ci mice fed HFD did not have reductions in oligodendrocyte lineage cell counts compared with their RD genotype controls (p = 0.682), yet had significantly higher PDGFRα-positive (dorsal column, p = 0.0397; ventral WM, p = 0.0198) and GST3-positive (dorsal column, p = 0.0001) cell counts versus WT-HFD mice. The improved cell numbers are likely because of decreased apoptosis rather than proliferative changes, as evidenced by decreased cleaved-caspase-3 staining (Fig. 2E: ventral WM: F_(1,16) = 5.48, p = 0.0224) in the CD38ci-HFD versus WT-HFD group and the lack of a significant difference in Ki67-positive cells.

Across spinal cord regions, no significant changes in MBP because of genotype (Fig. 2D: F_(1,28) = 0.002, p = 0.966) or diet group (Fig. 2D: F_(1,28) = 0.683, p = 0.415) were found. On the other hand, proteolipid protein (PLP) was reduced in the WT-HFD mice when compared with WT-RD mice (Fig. 2I), yet CD38ci mice did not have any differences in PLP staining after long-term consumption of HFD. Significantly more MBP (Fig. 2J) and PLP (Fig. 2K) RNA were also present, reflective of the higher number of mature oligodendrocytes remaining. Confirming our immunoperoxidase staining data, fluorescent triple labeling further demonstrated that the CD38ci mice were protected from HFD-induced loss of cells across the oligodendrocyte lineage. This included early-stage progenitors including PDGFRα-positive/Olig2-negative and PDGFRα-positive/Olig2-positive cells (p = 0.0382, and p = 0.0003, respectively; Fig. 2L) as well as more mature cells expressing both aspartoacylase (ASPA) and Olig2 (p = 0.0179). Overall, these data suggest that oligodendrocyte lineage cell survival is preserved in CD38ci mice following chronic HFD consumption.

Diminished neuroinflammatory responses and oxidative damage in CD38ci mice following chronic HFD

To explore the potential impact of blocking CD38 catalytic activity on CNS neuroinflammation following chronic HFD consumption, we next performed immunohistochemical analysis on the spinal cords from WT and CD38ci mice fed RD or HFD for markers of microglia and astrocytes. Quantification of the thresholded area considered positive for IBA1 demonstrated decreased microglia numbers or activity in the ventral gray matter (GM) of spinal cords from CD38ci mice compared with WT mice overall (F_(1,21) = 5.079, p = 0.030; Fig. 3A). In addition, CD38ci-HFD mice showed significantly reduced IBA1 immunoreactivity in the ventral GM and WM of the spinal cord compared with WT-HFD mice. Although the HFD did not significantly increase IBA1 immunoreactivity in this study, this is consistent with both a recent study on Western-style diet from our laboratory as well as another report that older mice may already have residually higher levels (Clarke et al., 2018; Kim et al., 2020).

Figure 3.

Diminished neuroinflammatory responses in CD38ci mice following chronic HFD. Histologic analysis was performed on WT and CD38ci mice fed a RD or HFD for 30 weeks beginning at 50 weeks of age. Scale bar, 50 µm. n = 7–10 mice/group. A, Representative images of the ventral white matter and immunohistochemical analysis of spinal cord regions demonstrates decreased microgliosis (IBA1) in CD38ci mice. When compared with WT-HFD mice, CD38ci-HFD mice had significantly less IBA1-positive immunoreactivity in the spinal cord ventral GM and WM. B, Representative images of the ventral white matter and corresponding quantification confirm reduced astrogliosis (GFAP) in the spinal cords of CD38ci mice overall when compared with WT mice. Decreased GFAP-positive immunoreactivity in the ventral white matter when compared with HFD mice. C, Representative images of the dorsal column and quantification of the spinal cords confirming significant changes in 4HNE immunoreactivity, an oxidative stress marker, between genotypes for HFD mice. D–F, qRT-PCRs data indicating differences in antioxidant (D) and prorepair astrocyte marker genes BDNF (E) and S100a10 (F) in the CD38ci mice versus controls. Bar graphs represent the mean ± SEM. Two-way ANOVA with Bonferroni post-test. Asterisks indicate levels of statistical significance: *p < 0.05; **p < 0.01. Gray lines on graphs indicate genotype means, while gray broken lines with asterisks indicate a significant effect of genotype overall.

Considering that astrocytes have endogenous expression of CD38 as well as elevated CD38 expression following HFD or demyelination, we also investigated astrocyte reactivity in these mice. In the ventral spinal cord, quantification of astrocyte marker GFAP (F_(1,28) = 12.66, p = 0.0014; Fig. 3B) demonstrated significantly reduced positive area in the WM of CD38ci mice compared with WT mice. WT-HFD mice showed significantly more astrocyte reactivity compared with either RD-fed (p = 0.023) or HFD-fed (p = 0.043) CD38ci mice.

We next sought to compare the extent of oxidative damage in the spinal cords of WT and CD38ci mice fed a RD or HFD. More abundant lipid peroxidation marker 4-hydroxynonenal (4HNE) was present in the spinal cord dorsal column of HFD-fed WT mice than in that of the RD-fed WT mice (F_(1,28) = 3.358, p = 0.019; Fig. 3C). Importantly, a significant reduction in 4HNE was found in the HFD-CD38ci mice compared with HFD-WT mice in the dorsal column and ventral spinal cord (GM and WM).

Next, our histologic findings were further substantiated by observed gene expression increases in superoxide dismutase 2 (SOD2; F_(1,8) = 7.721, p = 0.024; Fig. 3D), a mitochondrial antioxidant that can protect from HFD-induced changes in oxidative stress and metabolism (Liu et al., 2018), prorepair astrocyte marker genes brain-derived neurotrophic factor (BDNF; F_(1,8) = 5.685, p = 0.044; Fig. 3E), and S100a10 (F_(1,8) = 24.73, p = 0.0011; Fig. 3F) in the CD38ci mice, as determined by qRT-PCR. Overall, the observed reductions in neuroinflammation and oxidative stress markers and increased NAD⁺ levels in the CD38ci mice likely provide a more suitable microenvironment for oligodendrocyte survival, differentiation, and repair following chronic HFD consumption.

Improvement in myelin regeneration after inhibition of CD38 with 78c following LL demyelination in the ventral spinal cord

To determine whether pharmacologic inhibition of CD38 affects the potential for remyelination, we administered 78c, a CD38 inhibitor, to mice in the RD and HFD groups and used histologic approaches to investigate the numbers of oligodendrocyte lineage cells and remyelinated axons in the LL model of demyelination. The 6-week-old male WT mice were given either RD or HFD for 5 weeks, then were split into additional groups receiving the same diet with or without 78c (600 ppm; Fig. 4A). All mice were subjected to a focal demyelinating injury to the ventral spinal cord by microinjection of 1% lysolecithin.

Figure 4.

CD38 inhibition with 78c improves myelin regeneration following lysolecithin demyelination in the ventral spinal cord. The 8-week-old male C57BL6/J mice were fed a RD or HFD for 5 weeks, then split into the following two additional groups: one receiving the same diet; and the other half receiving the respective diet with 78c added (600 ppm). One week later, a focal demyelinating lesion to the ventral spinal cord was induced by LL (1% with Evans blue dye) and mice were allowed to recover for 2 weeks. Lesion boundary regions were determined by limiting to the area of continuous myelin loss (core) or by including the entire inflammatory lesion (rim) based on immunohistochemical analyses. A, Schematic depicts experimental design and lesion boundaries. B, Representative images and corresponding analysis illustrating the mean number of remyelinated axons after focal lysolecithin-mediated demyelination of the ventral spinal cord was increased in 78c-treated mice at 14 d postinjury when compared with LL-only mice (**p < 0.01, two-way ANOVA with Bonferroni post-test) as assessed by paraphenylenediamine (PPD)-stained semithin sections of Araldite-embedded spinal cord tissue. Scale bar, 20 µm. C, Immunohistochemical staining of the total lesion and lesion rim identified significant changes in MBP immunoreactivity in 78c-treated animals. D, E, Olig2-positive oligodendrocyte lineage cells (OLCs; D) and GST3-positive mature oligodendrocytes (E) were significantly higher in 78c-RD mice when compared with RD controls. Bar graphs represent the mean ± SEM. Two-way ANOVA (n = 7–10 mice/group). *p < 0.05; **p < 0.01. Gray lines on graphs indicate drug treatment group means, while gray broken lines indicate a significant effect of drug treatment overall.

After 2 weeks of recovery, lesions were collected from perfused mice and used for subsequent counting of remyelinated axons and myelinating cells. At the study end point, HFD mice had gained significantly more weight than their RD counterparts (F_(1,36) = 51.91, p < 0.0001, data not shown), regardless of genotype. While 78c treatment did not have a significant effect on body weight, mice given 78c consumed less food (data not shown). The average lesion area (RD alone, 0.107 mm²; HFD alone, 0.092 mm²; RD-78c, 0.132 mm²; HFD-78c, 0.117 mm²; as determined by hematoxylin and eosin staining; data not shown) was not significantly altered by either HFD (F_(1,24) = 2.642, p = 0.117) or 78c (F_(1,24) = 0.886, p = 0.356). Neurofilament immunoreactivity, a marker of axon integrity, was on average 18.09%, 15.76%, 18.29%, and 18.71% of the lesion area, respectively, for the RD, HFD, RD-78c, and HFD-78c groups (data not shown). Neurofilament-positive staining was not significantly changed by drug (F_(1,24) = 0.717, p = 0.405) or diet (F_(1,24) = 0.261, p = 0.614).

Overall, LL mice fed RD or HFD with 78c showed significantly greater numbers of remyelinated axons compared with the LL mice fed RD or HFD without 78c (F_(1,25) = 10.11, p = 0.004; Fig. 4B). In particular, more than double the number of remyelinated axons was present in the RD-78c mice compared with RD-alone mice (p = 0.0013). Further supporting a role of CD38 inhibition in myelin repair, MBP immunoreactivity in the total lesion or the lesion rim was higher in the 78c-treated mice regardless of diet (F_(1,24) = 5.71, p = 0.025; Fig. 4C). Importantly, the numbers of Olig2-positive cells (F_(1,23) = 9.508, p= 0.005; Fig. 4D) and GST3-positive cells (F_(1,24) = 6.172, p = 0.020; Fig. 4E) were significantly higher in the combined 78c group (which includes LL-RD-78c and LL-HFD-78c mice) versus the combined group without 78c treatment (which includes LL-RD and LL-HFD mice). Moreover, the number of Olig2-positive cells in the lesion rim (p = 0.049) and GST3-positive cells in the rim (p = 0.044) or total lesion (p = 0.032) was significantly higher in the LL-RD-78c group versus the LL-RD group. These findings suggest that, in vivo, 78c helps to preserve or replenish myelinating cells and their progenitors, leading to increased markers of myelin repair. Furthermore, these results provide a strong rationale for targeting CD38 and overall NAD⁺ metabolism as a new strategy to attenuate oligodendrocyte loss and aid in myelin regeneration.

Attenuation of HFD-induced oligodendrocyte loss in the spinal cord after pharmacologic inhibition of CD38 with 78c

To evaluate the potential of pharmacologic CD38 inhibition to recapitulate the protective effects that genetic CD38 inactivation conferred on oligodendrocytes in the HFD experiment (Figs. 2, 3), we next looked at the staining of oligodendrocyte markers and myelin proteins on the contralateral, uninjured side of spinal cord from the study presented in Figure 4. Confirming this, the loss of Olig2-positive cells with a HFD was significantly restored in the ventral GM and WM of the HFD-fed mice that were given 78c (Fig. 5A). Moreover, HFD-induced loss of GST3-positive cells in the ventral WM was significantly attenuated with 78c treatment (Fig. 5B). Although PLP staining was significantly reduced in the dorsal column of the HFD-fed mice in the longer experiment presented in Figure 2, this shorter time point of HFD in younger mice did not result in a significant loss of MBP or PLP (Fig. 5C,D, Table 2). However, these data, together with the results from Figure 2, demonstrate that genetic or pharmacologic approaches targeting CD38 can be used to attenuate HFD-induced oligodendrocyte apoptosis in mouse models.

Figure 5.

Pharmacologic inhibition of CD38 with 78c attenuates HFD-induced oligodendrocyte loss in the spinal cord. Histologic analysis from the uninjured side of the ventral spinal cord from wild-type mice fed a RD or HFD with or without 78c added (600 ppm), as depicted in Figure 4A. A, B, Immunostaining of Olig2-positive oligodendrocyte lineage cells (A) and GST3-positive mature oligodendrocytes (B) in the ventral spinal cord shows loss of Olig2-positive oligodendrocytes with HFD and restoration of these cells in the mice given 78c. HFD-induced loss of GST3-positive cells in the ventral WM was also attenuated with 78c. C, D, Representative images of immunohistochemical staining and quantification of thresholded area demonstrate no significant changes in MBP (C) or PLP (D) immunoreactivity across genotype or diet groups. Scale bar, 100 µm. Bar graphs represent the mean ± SEM. Two-way ANOVA with Bonferroni post-test (n = 7–10 mice/group). *p < 0.05; **p < 0.01.

Role of 78c in prorepair astrocyte signatures after LL-induced demyelination

Next, we chose to explore the role of 78c on neuroinflammation markers in the spinal cord following demyelination. Histologic quantification of threshold-positive staining for microglia marker IBA1 indicated a significant interaction between diet and drug treatment on the lesion rim (F_(1,24) = 6.936, p = 0.015; Fig. 6A), yet no significant effect of diet (F_(1,24) = 2.160, p = 0.154) or drug alone (F_(1,24) = 0.2549, p = 0.618). Although there is a trend of decrease for both IBA1 and GFAP in the 78c-treated group, no statistically significant attenuation of GFAP was observed (Fig. 6B). In both the RD- and HFD-fed mice, there was, however, a significant increase in S100a10 immunoreactivity, considered a marker for prorepair astrocytes, with 78c treatment (F_(1,22) = 20.77; RD, p = 0.001; HFD, p = 0.045; Fig. 6C). The proinflammatory astrocyte marker Serping1 was significantly reduced in the LL-78c-treated groups when compared with LL-control groups (RD, p < 0.001; HFD, p < 0.001; Fig. 6D), while another prorepair marker, Emp1, was unchanged by diet or drug treatment (Fig. 6D). There was also a slight decrease in C3d staining, a marker of reactive astrocytes, with 78c treatment, although this did not reach statistical significance (F_(1,19) = 3.681, p = 0.069; data not shown). Together, these data suggest that 78c can skew astrocytes toward prorepair rather than proinflammatory responses following experimental demyelination to the spinal cord.

Figure 6.

78c promotion of prorepair astrocyte signatures after LL-induced demyelination. Histologic analysis of markers of neuroinflammation in animals fed a RD or HFD with or without 78c (CD38 inhibitor, 600 ppm) and injected with LL followed by a 2 week recovery period. A, B, IBA1 (A; microgliosis) and GFAP (B; astrogliosis) were not significantly affected by either diet or 78c alone, although a significant interaction for drug and diet was detected for IBA1 immunoreactivity in the lesion rim. Scale bar, 100 μm. C, D, Changes in prorepair (S100a10, brown; Emp1, red) and proinflammatory (Serping1, green) astrocyte (GFAP, cyan) markers in adjacent sections within the LL lesion area indicate a shift toward prorepair in the 78c-treated mice. Scale bar, 100 µm. Bar graphs represent the mean ± SEM. Two-way ANOVA (n = 5–8 mice/group). Asterisks indicate levels of statistical significance: **p < 0.01, compared with LL-RD by Bonferroni post-test. Gray lines on graphs indicate treatment group means, while gray broken lines with asterisks indicate a significant effect of treatment group overall.

Astrocytes exposed to saturated fat-enriched conditions increase CD38 and proinflammatory cytokine expression and inhibit oligodendrocyte differentiation in an NAD⁺-dependent manner

Astrocytes release various growth factors, cytokines, chemokines, extracellular matrix proteins, glutamate, and reactive oxygen species that can heavily influence oligodendrocyte proliferation and differentiation. Therefore, we next investigated the effects of culturing oligodendrocytes in conditioned media obtained from saturated fat PA-exposed astrocytes with or without 78c cotreatment, as depicted in Figure 7A. As expected, ACM from PA-exposed astrocytes significantly reduced oligodendrocyte PLP expression (F_(2,17) = 9.396, p = 0.006) and MBP expression (F_(3,8) = 8.155, p = 0.003). However, this negative impact on myelin protein production was partially rescued by cotreatment with 78c (3 µm; PLP, p = 0.002; MBP, p = 0.040). Likewise, the addition of exogenous NAD⁺ (50 µm) to the saturated fat-exposed astrocyte cultures (F_(2,9) = 8.502, p = 0.008; Fig. 7B) significantly increased PLP expression, while ACM-treated oligodendrocyte numbers remained unchanged across treatment groups (F_(2,9) = 1.153, p = 0.358). These data indicate the importance of maintaining proper astrocyte NAD⁺ levels to promote oligodendrocyte differentiation.

Figure 7.

Astrocytes exposed to saturated fat-enriched conditions increase CD38 and proinflammatory cytokine expression and inhibit oligodendrocyte differentiation in a NAD⁺-dependent manner. A, Schematic depicting how ACM treatments on primary murine oligodendrocytes were performed. Conditioned media obtained from saturated fat PA-exposed murine astrocytes [ACM (PA)] reduced protein expression of mature oligodendrocyte markers PLP and MBP, while PA-78c-ACM-treated oligodendrocyte lineage cells (OLCs) have enhanced PLP and MBP expression. B, Representational diagram of astrocyte treatments with PA and NAD⁺ to collect ACM for OPC treatments. The addition of NAD⁺ restored PLP levels when compared with ACM-PA but had no effect on Olig2. Scale bar, 100 µm. C–E, High fat consumption was modeled in primary murine astrocyte cultures by including PA in the media (100 μm, 24 h). C, Addition of a small-molecule CD38 inhibitor (78c, 3 μm) was effective in restoring NAD⁺ levels in PA-treated astrocytes when assessed using an NAD⁺ cycling assay. Astrocytes derived from CD38 knock-out mice did not show depleted NAD⁺ levels in response to PA. D, E, RNA levels by qRT-PCR (D) of proinflammatory cytokines, antioxidant enzyme, and an astrocyte activation marker in PA- and 78c-treated astrocytes and cytokine levels (E) detected in ACM by Bio-Plex assay. Bar graphs represent the mean ± SEM. Asterisks indicate levels of statistical significance when compared with controls as determined by one-way ANOVA with Bonferroni post-test: *p < 0.05; **p < 0.01; ***p < 0.001, versus control.

To address whether the possible NAD⁺-driven astrocytic changes indirectly mediate the observed effects on oligodendrocytes, we next used primary murine astrocyte cultures to assess NAD⁺ levels and transcriptional changes. The addition of 78c (3 µm) was effective to restore NAD⁺ levels in saturated fat-exposed astrocytes when quantified using an NAD⁺ cycling assay (F_(4,23) = 6.113, p = 0.004; Fig. 7C). However, astrocytes derived from CD38 knock-out mice did not have depleted NAD⁺ levels in response to PA. Next, we determined whether CD38 regulates astrocyte proinflammatory responses. After treating astrocyte cultures with PA and/or 78c, we used qRT-PCR and a multiplex bioassay for determining cytokine RNA transcripts and release into culture supernatant, respectively. We observed significant increases in interleukin (IL)-1β and IL-6 expression (IL-1β: F_(3,8) = 8.427, p = 0.007; IL-6: F_(3,8) = 11.83, p = 0.007; Fig. 7D) in astrocytes stimulated with PA and 78c significantly attenuated saturated fat-induced RNA expression of IL-1β (p = 0.007), IL-6 (p = 0.008), tumor necrosis factor-α (TNFα; p = 0.017), and reactive astrocyte marker H2D1 (p = 0.028). Additionally, SOD2 was significantly increased in astrocytes treated with 78c (p = 0.022) or both PA and 78c (p = 0.035).

Significant increases in the secretion of cytokines triggered by PA were observed using a Bio-Plex assay including CXCL1 [keratinocytes-derived chemokine (KC); F_(3,8) = 9.296, p = 0.014, MIP1α F_(3,8) = 549.8, p < 0.0001], RANTES (regulated upon activation, normal T cell expressed and presumably secreted; F_(3,8) = 418.5, p < 0.0001), IL-6 (F_(3,8) = 18.00, p = 0.001), MIP1β (F_(3,8) = 533.5, p < 0.0001), and MCP1 (F_(3,8) = 66.96, p < 0.0001; Fig. 7E). Notably, cotreatment with 78c successfully mitigated PA-stimulated secretion of CXCL1 (p = 0.006), MIP1α (p < 0.0001), RANTES (p < 0.0001), IL-6 (p = 0.001), MIP1β (p < 0.0001), and MCP1 (p < 0.0001). Thus, inflammatory- and oxidative stress-related changes in cultured astrocytes may help to explain the reduced differentiation potential of the oligodendrocyte progenitors in the presence of saturated fat-exposed ACM.

Direct effects of NAD⁺ replenishment on myelin proteins in saturated fat-exposed oligodendrocytes are contingent on SIRT1 activity

To examine the effects of replenishing NAD⁺ levels on oligodendrocytes, we next supplemented cultures with exogenous NAD⁺ (50 µm) and monitored PLP expression (Fig. 8A). While saturated fat alone reduces PLP expression in oligodendrocyte progenitors, cotreatment with PA and NAD⁺ significantly enhanced PLP expression (F_(2,16) = 3.458, p = 0.024; Fig. 8B) when compared with exclusive PA treatment.

Figure 8.

Direct effects of NAD⁺ replenishment on myelin proteins in saturated fat-exposed oligodendrocytes are contingent on SIRT1 activity. A, Representational diagram of experimental treatments in primary murine oligodendrocyte lineage cells (OLCs). B, Adding exogenous NAD⁺ (50 μm) to oligodendrocyte cultures significantly restores PLP protein expression in saturated fat PA-exposed OLCs. C, Treatment with NAD⁺ precursor NMN (100 μm) also enhances the expression of mature oligodendrocyte markers in PA-treated cultures, although not when SIRT1 inhibitor EX-527 (5 μm) is also included in the media, as determined by immunostaining quantification for PLP and MBP per DAPI-positive cell. Scale bar, 100 µm. D, Schematic summarizing of direct and indirect effects of the saturated fat PA on oligodendrocytes characterized in vitro. Graphical results represented as the mean ± SEM (n = 4–8/group). Asterisks indicate levels of statistical significance when compared with a designated group, as determined by one-way ANOVA with Bonferroni post-test: *p < 0.05; **p < 0.01; ***p < 0.001.

We hypothesized that the promyelinating effects of increasing NAD⁺ are SIRT1 dependent, so we determined whether the beneficial effects of NAD⁺ in the presence of PA could be blocked with the SIRT1 inhibitor EX-527. Similar to adding exogenous NAD⁺, treatment with the NAD⁺ precursor NMN (100 µm) also restored markers of oligodendrocyte differentiation in saturated fat-exposed cultures (F_(3,25) = 8.449, p = 0.011; Fig. 8C). However, in the presence of EX-527 (5 µm), the increases in PLP and MBP in response to NMN were suppressed (p = 0.003). At the doses used, PA, NMN, and EX-527 did not significantly alter the number of oligodendrocytes. Hence, the impaired ability of oligodendrocytes cultured with excess saturated fat to differentiate is at least partially regulated by NAD⁺-dependent sirtuin activity (Fig. 8D).

Attenuation of saturated fat-mediated impairments in remyelination of cerebellar slice cultures by CD38 inhibition

Since we observed protective effects of genetic CD38 inhibition on oligodendrocyte survival following HFD, as well as increased CD38 expression in a spinal cord demyelination model, we next sought to explore any benefits conferred by pharmacologic inhibition using 78c in a brain model of LL-induced ex vivo demyelination. Murine cerebellar slices from postnatal pups were demyelinated using LL (0.5 mg/ml, 18 h), followed by a period of 1 week to allow for repair (Fig. 9A). Confocal imaging of immunofluorescence staining demonstrated reductions in MBP after LL that were significantly attenuated by 78c cotreatment (F_(2,25) = 4.578, p = 0.020; Fig. 9B).

Figure 9.

Saturated fat-mediated impairments in remyelination in cerebellar slice cultures are attenuated by CD38 inhibition. A, B, Representational diagram (A) of ex vivo remyelination experiments where murine organotypic cerebellar slices were demyelinated with LL (0.5 mg/ml, overnight) followed by subsequent culture in control media for 1 week; confocal microscopic imaging of slices indicates reduced MBP (red; B) expression when normalized to neurofilament (NF; green). LL-demyelinated slices cultured with 78c had significantly increased MBP expression when normalized to NF staining and expressed as a percentage of the control group. C, Slices cultured in media containing both LL-PA (100 μm) and 78c (a CD38 inhibitor; 3 μm) had significant improvements in Olig2 and MBP expression compared with the LL-PA alone group. Bar graphs represent the mean ± SEM (n = 4–8/group). *p < 0.05, **p < 0.01; ***p < 0.001, compared with controls by one-way ANOVA with Bonferroni post-test.

Finally, to evaluate the therapeutic potential of 78c in a brain model of high-fat exposure combined with demyelination, organotypic cerebellar slices were demyelinated as described above and cultured for 7 d in the presence or absence of PA and/or 78c. Cerebellar slices showed fewer Olig2-positive cells (F_(6,21) = 4.964, p = 0.003) with LL insult. Inclusion of PA in the media further diminished Olig2-positive cells in the LL slices. After 7 d of recovery, Olig2-positive cells remained significantly depleted in the 7 d LL group. There was also less MBP immunoreactivity following LL insult (F_(6,21) = 2.899, p = 0.020; Fig. 9C); however, MBP levels were successfully restored after 7 d of recovery. Importantly, 78c and PA cotreatment significantly increased the percentage of Olig2-positive cells (p = 0.049) and MBP expression (p = 0.023) in the demyelinated cerebellar slices. However, neither saturated fat (PA) nor 78c alone, without LL, for 2 weeks significantly altered MBP expression (PA, p = 0.668; 78c, p = 0.566; vs control). These findings demonstrate the capability of 78c to restore myelin protein expression even in combined saturated fat and demyelinating agent-induced conditions.

Unique astrocyte translatome profile in response to pharmacologic CD38 inhibition in adult mice

After substantiating the potential for CD38 inhibition to enhance myelin regeneration in spinal cord and brain myelin injury models, as well as performing some basic mechanistic work in vitro to identify key cell-to-cell interactions, we next aimed to better characterize the effects of pharmacologic CD38 inhibition in astrocytes in vivo. First, to facilitate astrocyte-enriched RNA profiling in vivo, we crossed the RiboTag (RPL22^HA/HA) mice with ALDH1L1^Cre-ERT2 mice. In this system, the expression of a hemagglutinin-tagged ribosomal subunit is controlled through the ALDH1L1 promoter and Cre recombinase expression is tamoxifen-inducible (100 mg/kg, i.p., 5 d). HA IP and RNA isolation procedures used for bulk (input) and astrocyte RNA (HA-IP) fractions harvested from the corpus callosum of male ALDH1L1^Cre-ERT2:RPL22^HA-RiboTag mice are illustrated in Figure 10A.

Figure 10.

Validation of astrocyte-specific RNA from ALDH1L1-RiboTag mice. A, Diagram illustrating HA IP and RNA isolation procedures for bulk (input) and astrocyte RNA (HA IP) harvested from the CC of male ALDH1L1-RiboTag mice (RPL22^HA/+:ALDH1L1^Cre-ERT2, n = 3). B, Heatmap with hierarchical clustering shows differentially expressed genes between astrocyte-enriched and bulk RNA. A total of 3222 genes were significantly increased in the astrocyte fraction (red), while 4129 genes had lower expression in astrocytes when compared with bulk RNA (blue; p < 0.05, FDR < 0.1, and logFC > |1|). C, Volcano plot demonstrates that known astrocyte genes were significantly enriched in the HA IP fraction when compared with bulk RNA, while the following other cell type markers had reduced expression: astrocyte (black text); myeloid/microglia (green); oligodendrocyte (orange); endothelial (pink); neuron (purple). D, Bar graph shows enrichment/de-enrichment of a selection of cell marker genes. E, F, Representative images of the corpus callosum (E) and cortex (F) show HA expression in GFAP-positive astrocytes but not Olig2-positive oligodendrocytes (E) or NeuN-positive neurons (F). Scale bar, 20 µm.

When comparing astrocytes (HA-IP fraction RNA) to all cell types (bulk RNA) in the CC, the astrocyte-enriched RNA exhibited significant differences in gene expression (3222 up and 4129 down) as depicted in the heatmap with hierarchical clustering and volcano plot (Fig. 10B,C; p < 0.05, FDR < 0.1, logFC > |1|). The volcano plot (Fig. 10C) and bar graph (Fig. 10D) demonstrate that known astrocyte genes were significantly enriched in the HA-IP fraction when compared with bulk RNA, while other cell type markers had reduced expression. Further model validation by immunohistochemistry confirms HA expression in GFAP-positive astrocytes in the corpus callosum (Fig. 10E) and cortex (Fig. 10F) but not Olig2-positive oligodendrocytes (Fig. 10E) or NeuN-positive neurons (Fig. 10F). Together, these data verify that astrocytes were indeed enriched following HA IP in the ALDH1L1^Cre-ERT2:RPL22^HA-RiboTag mice when compared with other brain cell type markers.

Next, to determine the impact of CD38 inhibition on astrocyte transcriptomics in vivo, we performed RNA sequencing on the corpus callosum of ALDH1L1^Cre-ERT2:RPL22^HA-RiboTag mice treated with vehicle (5% dimethylsulfoxide, 15% polyethylene glycol 400, 80% of 15% hydroxypropyl-γ-cyclodextrin in citrate buffer) and 78c (10 mg/kg, i.p.) as shown in Figure 11A. Transcriptional changes between vehicle- and 78c-treated mice were found primarily in the astrocyte fraction, but were also present in the bulk RNA (Fig. 11B–D). Of these, 34 genes were overlapping (Fig. 11B), including increased Glp1r (log₂FC = 3.068 in astrocytes), which has recently been shown to inhibit glucose uptake and promote fatty acid oxidation in astrocytes, maintain mitochondrial integrity, and lead to improved metabolism and memory formation in mice (Timper et al., 2020). The number of genes altered by 78c in the astrocyte-enriched fraction (Fig. 11D) far exceeded the changes in bulk RNA (Fig. 11C), highlighting the need for cell-specific approaches to discover key gene differences between treatment groups.

Figure 11.

Unique astrocyte translatome profile in response to pharmacologic CD38 inhibition in adult mice. A, Schematic representation of experimental design where male ALDH1L1-RiboTag mice were administered tamoxifen (100 mg/kg, i.p., 5 d) before being given vehicle solution or 78c (CD38 inhibitor 78c was given daily, i.p., for 4 weeks to RPL22^HA/+:ALDH1L1^Cre-ERT2 male mice, n = 3/group). CC tissue was then harvested and used for the procedure, as described in Figure 10A. B–D, Differences in the astrocyte translatome between vehicle- and 78c-treated mice are demonstrated in a Venn diagram (B) and volcano plots (C, D) where more genes are significantly altered in the astrocyte-enriched fraction (D) versus the bulk CC RNA (C) when comparing between vehicle (Veh)- and 78c-treated groups. E, F, Top 10 canonical pathways altered by 78c in bulk (E) and astrocyte (F) RNA, as generated by IPA. G, Comparison analysis of upstream regulators filtered for growth factors shows predicted activation of a number of growth factors in the astrocytes of 78c-treated mice versus controls (activation z score, >|2|; Benjamini–Hochberg correction, p < 0.05; predicted activation state: orange, activated; blue, inhibited; gray dot, insignificant). H, Network map of BDNF as the highest upstream regulator predicted to be activated by 78c in astrocytes with predicted (orange/blue) or actual (red/green) activation or inhibition of known interacting genes from the dataset. I–K, Heatmaps of growth factors (I) and other astrocyte-related genes [J; pan, proinflammatory (A1), and prorepair (A2) astrocyte genes] reveal significant changes (*p < 0.05, FDR < 0.1) between vehicle- and 78c-treated mice CC astrocyte RNA, while myelin-related genes in the bulk RNA (K) are not significantly affected.

IPA revealed the top canonical pathways altered by 78c in the bulk RNA to include constitutive androstane receptor, integrin-linked kinase, and glycoprotein VI platelet signaling pathways, as well as fatty acid oxidation, leukocyte extravasation, and histamine degradation, among others (Fig. 11E). In contrast, the top canonical pathways changed between vehicle and 78c identified in the astrocyte RNA included CREB signaling, synaptogenesis, stathmin-1, and calcium signaling (Fig. 11F).

Since we observed increased BDNF in the spinal cords of CD38ci mice when compared with WT controls, along with predicted activation of CREB signaling in the 78c-treated astrocyte RNA, we next performed an upstream regulator analysis for growth factors in the astrocyte-enriched RNA. Remarkably, the activation z score for BDNF was 8.06, predicting strong activation of this network in the 78c-treated mouse astrocytes versus vehicle controls (Fig. 11G,H). Confirming the predicted activation status, the log₂FC for BDNF was 2.79, and many downstream cellular targets match the predicted direction of the effects of BDNF such as increased Egr1 and Ryr1 and decreased Grin2C and Gpc4. Other growth factors predicted to be activated include Ngf (z score = 3.98), Nrtn (z score = 2.96), Nrg4 (z score = 2.38), and Ntf3 (z score = 2.14). Additional heatmaps of select growth factors and other astrocyte-related genes (Fig. 11I,J) display significant increases in Fgf9, Cxcl12, and Igf1, while Stat3, Fgf2, Igf2, Fgf1, Hgf, Lif, Vcam1, and Tlr3 were reduced in the 78c-treated group. Several proinflammatory (A1) astrocyte markers, Ugt1a1, Fbln5, Serping1, and Gbp2, were significantly reduced in the 78c astrocyte RNA, and prorepair (A2) astrocyte marker Ptgs2 was increased, suggesting a skewing to a more advantageous phenotype.

Although these 78c-induced changes in the astrocyte translatome did not significantly affect myelin-related gene expression (Fig. 11K) under normal conditions, the potential of inducing less inflammatory prorepair astrocytes to enhance the capacity for myelin regeneration following demyelination or other insults to the CNS is promising and should be explored in future studies. Overall, these data suggest that CD38 inhibition modulates astrocyte signatures under normal conditions and their responses elicited by demyelinating injury or high-fat diet in the CNS, skewing them toward a phenotype that enhances oligodendrocyte survival and differentiation and myelin repair (Fig. 12).

Figure 12.

Schematic representation of proposed hypothesis. Resting astrocytes typically help to provide a supportive environment for oligodendrocyte differentiation and repair, when necessary. HFD and LL increase CD38 expression in reactive astrocytes, thereby depleting NAD⁺ levels. Reductions in NAD⁺ result in oxidative stress and inflammation, which triggers impaired differentiation and survival in oligodendrocyte lineage cells. 78c inhibits CD38 to attenuate the described detrimental effects of astrocytes on oligodendrocytes.

Discussion

Diet and MS

Myelin disturbances and oligodendrocyte loss can leave axons vulnerable, leading to permanent neurologic deficits. Diet-induced obesity increases the risk for developing MS, especially when combined with other known risk factors (Gianfrancesco et al., 2014; Olsson et al., 2017). The results of our current study suggest that metabolic disturbances, triggered by the consumption of a diet high in fat, promote oligodendrocyte loss and impaired differentiation through astrocyte-linked indirect NAD⁺-dependent mechanisms. We demonstrate that restoring NAD⁺ levels via genetic inactivation of the NAD⁺ glycohydrolase activity of CD38 can overcome these effects. Moreover, we show that therapeutic inactivation of CD38 in experimental demyelination models can enhance myelin regeneration. Canonical pathway analysis in ALDH1L1^Cre-ERT2:RPL22^HA-RiboTag mice demonstrates that several metabolic pathways, including fatty acid metabolism, are affected by pharmacologic CD38 inhibition. Increased Glp1r and Igf1 suggest improved lipid metabolism and insulin signaling, respectively. These findings point to a new metabolic targeting strategy positioned to improve disease course in MS and other conditions in which myelin integrity is a key concern.

Neuroinflammation and CD38

Neuroinflammation and astrogliosis are predominant pathologic mediators of neurodegenerative conditions, including MS (Liddelow et al., 2017; Ponath et al., 2017), and astrogliosis is linked to HFD-induced CNS pathology in rodents (Ogrodnik et al., 2019; Kim et al., 2020). In the current study, we found that HFD and focal LL-mediated demyelinating injury each drove increased CD38 expression in astrocytes, while genetic or pharmacologic CD38 inhibition after HFD or demyelination, respectively, attenuated astrocyte activation markers. Highlighting the potential significance of these findings, CD38 was also identified in hypertrophic reactive astrocytes in MS plaques.

Astrocytes and microglia secrete a variety of chemokines, cytokines, and growth factors that regulate myelin injury and regeneration (Alizadeh and Karimi-Abdolrezaee, 2016; Domingues et al., 2016). In saturated fat-stimulated astrocytes, the expression and secretion of IL-6 and other cytokines was increased. This response was blunted when NAD⁺ levels were restored by cotreatment with 78c. Therefore, improving NAD⁺ levels by targeting CD38 shows therapeutic promise across multiple neurologic and non-neurologic disease types (Guerreiro et al., 2020), and here we establish its significance to oligodendrocyte heath and regeneration with strong linkage to its effects on astrocyte metabolism.

We demonstrate that 78c is capable of affecting astrocytes both in the uninjured adult mouse brain as well as the spinal cord of mice following acute focal demyelination. The RiboTag approach determined gene expression changes in astrocytes resulting from pharmacologic CD38 inhibition in the intact CNS. Results mirrored many effects observed in vitro and in the CD38ci mice. Astrocyte signatures were also similarly skewed to prorepair in mice given oral 78c in the context of focal spinal cord demyelination. Moreover, additional pathways were identified that point to the complex role of CD38 in astrocytes in vivo.

Supporting a potential role for currently available MS drugs to affect astrocytes and strengthening a potential relationship between sphingosine 1-phosphate (S1P) signaling and CD38, fingolimod (an orally bioavailable S1P receptor modulator that leads to functional antagonism of S1pr1) reduced CD38 gene expression in an EAE model (Herrmann et al., 2016). Similarly, S1pr1 was reduced in brain astrocytes of 78c-treated ALDH1L1^Cre-ERT2:RPL22^HA-RiboTag mice. Additionally, there was astrocyte-specific activation of growth factors such as BDNF. Collectively, these data point to the ability of 78c to modulate astrocyte signatures toward those that promote myelin regeneration.

Oxidative stress and differentiation

Reactive oxygen species can also inhibit oligodendrocyte survival (Lassmann and van Horssen, 2016). Complementing this, we demonstrate more SOD2 and less IL1β in PA–78c-cotreated astrocyte cultures compared with treatment with PA alone. We also found more SOD2, BDNF, and S100a10 in the CD38ci mouse spinal cord. Blocking downstream CD38 signaling also reduced markers of oxidative damage in a convulsion model (Zou et al., 2017). Our findings suggest that the beneficial effects of CD38 inhibition include protection against astrocyte proinflammatory factors and reactive oxygen species known to negatively impact oligodendrocyte renewal and maturation.

The effects of CD38 activity and NAD⁺ levels may depend heavily on cell type and disease context, with findings here supporting a critical protective role of lowering astrocyte CD38 activity for protection of myelinating cells and for myelin regeneration in the adult CNS. Future studies will be needed to confirm this finding once astrocyte-specific conditional knockouts for CD38 become available. CD38-positive cells can regulate NAD⁺ levels in neighboring CD38-negative cells by influencing NAD⁺ precursor availability (Tarragó et al., 2018; Chini et al., 2019), which may be the case for astrocytes and oligodendrocytes in our studies. Consistent with this model, 78c had no direct effect on oligodendrocyte differentiation in the context of PA in vitro. PA can inhibit nicotinamide phosphoribosyltransferase (NAMPT) activity (Penke et al., 2017), a rate-limiting enzyme in the NAD⁺ salvage pathway, which can prevent differentiation of stem cells toward the oligodendrocyte lineage (Stein and Imai, 2014). These findings may explain the direct inhibitory effect we observe of PA on oligodendrocyte differentiation. Thus, complex alterations in NAD⁺ biosynthesis, salvage, and degradation are positioned to contribute to the effects of PA on oligodendrocytes that can be rescued by NAD⁺ replenishment.

NAD⁺ and neural repair

Several recent studies explored supplementing NAD⁺ precursors or flavonoids known to inhibit NAD⁺ase activity to improve myelin repair and neuroprotection, as well as to decrease immune cellular responses. For example, an NAD⁺/BDNF/TrkB pathway is proposed to be involved in nicotinamide-mediated protection, including improved remyelination in a stroke model (Wang et al., 2017) and precursor supplementation increased BDNF in a Huntington's disease model (Hathorn et al., 2011). Supporting a possible NAD⁺–BDNF pathway for enhancing remyelination, we observed increased BDNF expression in the spinal cord of CD38ci mice and brain astrocytes of 78c-treated ALDH1L1^Cre-ERT2:RPL22^HA-RiboTag mice, as well as improved remyelination with oral 78c administration in the LL model.

Although effects on myelinating cells and astrocytes were not directly examined, mice with CD38 knockout, given the CD38 inhibitor apigenin, or supplemented with NAD⁺, showed reduced EAE disease severity (Ginwala et al., 2016; Herrmann et al., 2016; Wang et al., 2016). NAD⁺ levels can impact oligodendrocyte lineage cell proliferation and survival via cyclins and E2f1, while differentiation is reported to be regulated redundantly by SIRT1 and SIRT2 (Ji et al., 2011; Stein and Imai, 2014; Jablonska et al., 2016). Results here demonstrate that the ability of NAD⁺ supplementation to promote oligodendrocyte differentiation in vitro is indeed at least partially reliant on SIRT1 function. Restoring NAD⁺ levels was also beneficial in the context of PA or high fat-induced inflammation and oxidative stress via SIRT1 in liver, heart, and muscle (Li et al., 2015; Prola et al., 2017; Sadeghi et al., 2017). Improvements in the expression of myelin proteins in the spinal cord of mice consuming a high-fat, high-sucrose diet in conjunction with coordinate wheel running, were also linked to increases in SIRT1 expression (Yoon et al., 2016). Collectively, these findings are particularly encouraging since effective modulation of disease progression will likely include a combination of therapeutic strategies encompassing anti-inflammatory, myelin-regenerative, and neuroprotective interventions.

Considerations and conclusions

The data presented demonstrate an essential role of NAD⁺ levels in maintaining CNS WM integrity, which can be impaired by focal demyelination or HFD consumption. Further, we show that inhibition of the NAD⁺ase CD38 attenuates oligodendrocyte loss and promotes myelin regeneration. Genetic CD38 inhibition attenuated HFD-induced oligodendrocyte loss and dietary supplementation with 78c increased counts of oligodendrocytes and remyelinated axons following focal demyelinating injury. These findings suggest that increases in astrocyte CD38, and the associated depletion of NAD⁺ levels, may contribute indirectly to impairments in oligodendrocyte health and myelin regeneration via inflammation and oxidative stress. These studies should also inspire future research to elucidate the role of diet on astrocyte–oligodendrocyte interactions during development and aging (Yoon et al., 2017a).

Although 78c did improve myelin regeneration in mice consuming a RD, little impact was seen in mice consuming a HFD, suggesting additional factors driving HFD demyelination also need to be mitigated. Alternately, a higher dose of 78c or targeting multiple aspects of astrocyte responses may be required to improve remyelination in the context of HFD. Tarragó et al. (2018) recently established 78c as a specific, reversible, and noncompetitive inhibitor of CD38 NAD⁺ase activity that improves age-related metabolic function (Hattori et al., 2017). Altogether, these findings suggest that genetic or pharmacologic CD38 inhibition provides therapeutic benefits in HFD-induced metabolic dysfunction and downstream signaling events in the CNS, in addition to the already described peripheral benefits on age-related metabolic dysfunction.

Importantly, inhibition of CD38 with 78c can improve myelin regeneration in the adult CNS, including reducing the proinflammatory profile while increasing the prorepair profile of astrocytes. Together, these findings point to the potential translational value of targeting CD38 to improve outcomes in MS and other demyelinating conditions. Monoclonal antibodies targeting CD38 (Lokhorst et al., 2015) include Daratumumab, which is Food and Drug Administration-approved for use in multiple myeloma. Accordingly, the research findings we present should inspire further research into ways to mitigate the negative consequences of HFD on oligodendrocytes alone and in the context of WM injury, which may include dietary intake of NAD⁺ precursors, administration of CD38 inhibitors, or exercise-related rehabilitation strategies.