Abstract

Background

   Multiple Sclerosis (MS) results from genetic predisposition and
   environmental variables, including elevated Body Mass Index (BMI) in
   early life. This study addresses the effect of BMI on the epigenome of
   monocytes and disease course in MS.

Methods

   Fifty-four therapy-naive Relapsing Remitting (RR) MS patients with high
   and normal BMI received clinical and MRI evaluation. Blood samples were
   immunophenotyped, and processed for unbiased plasma lipidomic profiling
   and genome-wide DNA methylation analysis of circulating monocytes. The
   main findings at baseline were validated in an independent cohort of 91
   therapy-naïve RRMS patients. Disease course was evaluated by a two-year
   longitudinal follow up and mechanistic hypotheses tested in human cell
   cultures and in animal models of MS.

Findings

   Higher monocytic counts and plasma ceramides, and hypermethylation of
   genes involved in negative regulation of cell proliferation were
   detected in the high BMI group of MS patients compared to normal BMI.
   Ceramide treatment of monocytic cell cultures increased proliferation
   in a dose-dependent manner and was prevented by DNA methylation
   inhibitors. The high BMI group of MS patients showed a negative
   correlation between monocytic counts and brain volume. Those subjects
   at a two-year follow-up showed increased T1 lesion load, increased
   disease activity, and worsened clinical disability. Lastly, the
   relationship between body weight, monocytic infiltration, DNA
   methylation and disease course was validated in mouse models of MS.

Interpretation

   High BMI negatively impacts disease course in Multiple Sclerosis by
   modulating monocyte cell number through ceramide-induced DNA
   methylation of anti-proliferative genes.

Fund

   This work was supported by funds from the Friedman Brain Institute,
   NIH, and Multiple Sclerosis Society.

   Keywords: Obesity, Neurodegeneration, Lipids, Epigenetics, Immunity
     __________________________________________________________________

Research In Context.

Evidence before this study

   Multiple Sclerosis (MS) defines a neurological condition characterized
   by inflammatory demyelination and axonal/neuronal degeneration, which
   is thought to result from the integration of genetic predisposition and
   environmental variables, and clinically progresses towards permanent
   disability. Previous studies on high Body Mass Index (BMI) in childhood
   as a risk factor for the disease suggest that BMI can induce long term
   effects, but the underlying mechanisms and impact on disease course
   remain elusive.

Added value of this study

   This study investigates the effect of BMI on the epigenome of monocytes
   sorted from the blood of recently diagnosed therapy naïve MS patients.
   By conducting blinded immunophenotyping, and genome-wide DNA
   methylation analysis, we identified DNA hypermethylation of genes
   negatively regulating proliferation as part of the mechanism
   responsible for the higher numbers of circulating monocytes in MS
   patients with high BMI. Unbiased lipidomic profiling of plasma samples
   revealed the presence of more abundant ceramide species in MS patients
   with high BMI, which likely reflected the synergy between de novo
   synthesis from dietary fats and recycling of damaged lipids due to the
   disease process. Cell-based approaches in cultured monocytes further
   revealed the nuclear localization of uptaken ceramides and validated
   their role as inducers of DNA hypermethylation on anti-proliferative
   genes, resulting in increased cell counts. Finally, we report an
   association between increased monocyte numbers, reduced brain volume
   and worsened clinical disability at the two-year follow-up in MS
   patients with high BMI; a finding that was reproduced in animal models
   of MS. Overall, these data identify the BMI-dependent increase of
   specific ceramide species as an important determinant of disease course
   in MS, acting, at least in part, by modulating the epigenome of
   monocytes.

Implications of all the available evidence

   The reported DNA methylation changes caused by ceramides in monocytes
   and previous work on neurotoxic ceramides highlight the concept that
   the long-lasting deleterious effect of BMI on MS disease course could
   be hampered by modification of lifestyle variables, such as diet. It is
   suggested that plasma ceramide levels in MS patients are derived not
   only from the degradation of myelin lipids but also from the dietary
   intake of saturated fats. Future studies are needed to evaluate dietary
   guidelines limiting specific saturated fatty acids that can be used for
   ceramide synthesis and possibly define DNA methylation inhibitors as
   potential supplementary therapeutic strategies.

   Alt-text: Unlabelled Box

1. Introduction

   It is becoming increasingly evident that the growing “epidemic” of fast
   food and processed food consumption in the United States and its
   association with increased Body Mass Index (BMI) is impacting the
   overall health status of the population [[49][1], [50][2], [51][3]].
   Alarming statistics have recently demonstrated a link between emergency
   room visits and BMI, not just for obese patients, but also for
   overweight individuals [[52]4,[53]5]. While much of the discussion on
   the exponentially increasing health care costs associated with obesity
   is focused on the prevention of cardiovascular disorders and metabolic
   syndromes [[54]6], a growing body of literature suggests an impact of
   BMI on a wide variety of other health conditions, including
   neurological disorders, such as Alzheimer's disease [[55][7], [56][8],
   [57][9]], Parkinson's disease [[58]10,[59]11] and Multiple Sclerosis
   (MS) [[60][12], [61][13], [62][14]].

   MS is a debilitating neurological disorder characterized by
   inflammatory demyelination and axonal/neuronal degeneration, which
   variably progresses towards permanent disability [[63]15].
   Epidemiological studies have carefully assessed the BMI as risk factor
   for MS [[64]12,[65]13,[66][16], [67][17], [68][18], [69][19], [70][20]]
   and many studies have supported the concept of long-term effects of the
   BMI on the disease, as obesity or high body mass in childhood or
   adolescence will induce disease onset many years later
   [[71]12,[72]13,[73]19,[74]20]. Finally, recent studies have highlighted
   a relationship between BMI and disease severity [[75][21], [76][22],
   [77][23], [78][24], [79][25]] and to some extent, progression
   [[80]26,[81]27], thereby further supporting the need to investigate the
   molecular mechanisms responsible for these long-term effects.

   Increased lipid synthesis in multiple tissues, including adipose
   tissue, liver, skeletal muscle and plasma has been associated with high
   BMI [[82][28], [83][29], [84][30], [85][31]]. This is of high interest
   as lipids are increasingly recognized as important signaling molecules
   regulating a multitude of cellular processes, including proliferation,
   metabolism and migration [[86]32]. Most relevant to MS is the role of
   sphingolipids [[87]33] and more specifically ceramides, whose altered
   levels have been reported both in the brains [[88]34,[89]35],
   cerebrospinal fluid [[90]36] and plasma [[91]37] of MS patients.
   Ceramides are gaining substantial interest as biomarkers for a wide
   variety of metabolic, cardiovascular and neurological disorders
   [[92]38,[93]39] however their precise mechanism of action is not fully
   understood and has often been described as cell-specific [[94]40]. In
   this study we detected intriguing parallels between elevated plasma
   levels of ceramide species, hypermethylation of anti-proliferative
   genes in monocytes, increased monocytic counts and worsened disease
   course in MS patients with high BMI compared to those with normal BMI.
   Given the strong epidemiological data on the long-term effects of BMI,
   we further hypothesized that ceramides act as nuclear signaling lipids
   and epigenomic modulators with the ability to impact DNA methylation
   and affect specific cellular functions, such as monocyte proliferation.

2. Materials and methods

2.1. Human studies

2.1.1. Primary cohort

   All subjects were recruited at the Corinne Goldsmith Dickinson Center
   for Multiple Sclerosis at Mount Sinai. The protocol was approved by
   Mount Sinai's Institutional Review Board and all subjects provided
   written informed consent. Subjects were enrolled from February 2014
   through December 2015. Eligible subjects aged 18–60 met criteria for
   Relapsing Remitting MS [[95]41] or Clinically Isolated Syndrome along
   with a brain MRI consistent with MS as determined by the treating
   physician, and were treatment-naïve with respect to MS
   disease-modifying therapy. Exclusion criteria included the diagnosis of
   another autoimmune disease, treatment with immunosuppressants,
   diagnosis of diabetes, recent high dose corticosteroids (<30 days), or
   recent antibiotics (<90 days). Details of each subject's demographic
   and clinical history and weight were recorded (See Table S1). Expanded
   Disability Status Scale (EDSS) [[96]42] was performed by a study
   co-investigator or in certain cases abstracted from a recent clinical
   visit.

   Follow up clinical information was obtained through review of the
   medical records of all subjects in the spring of 2017. The most recent
   clinical visit was chosen as the follow up time point and EDSS was
   abstracted from this visit. All chart notes from the time of enrollment
   to the follow up time point were reviewed to assess for disease
   activity, which was defined as any new clinical MS relapse or MRI
   activity. An MS relapse was defined as the appearance of a new
   neurological symptom lasting >24 h in the absence of fever or infection
   [[97]41] and was determined by the treating physician (all MS
   specialists) for clinical care purposes independently from the study.
   MRI activity was defined as the appearance of a new T2 lesion or new
   gadolinium-enhancing lesion on clinical MRI scan, again determined by
   the treating MS expert independently from the study. The frequency of
   disease activity in normal and high BMI groups was analyzed using
   chi-squared test on GraphPad Prism v7.0. The distribution of therapies
   in the two groups was also evaluated using chi-squared test on GraphPad
   Prism v7.0. Evaluation of associations between BMI at baseline and EDSS
   changes from follow-up to baseline were assessed by ordinal logistic
   regression on R. Multivariate analysis was conducted for BMI, age,
   gender, and race and partial regressions for BMI were calculated, which
   assessed the independent association between BMI and EDSS change and
   corrected for the other variables. P values <.05 were considered
   significant.

2.1.2. Validation cohort for monocyte counts and brain volume

   The validation cohort was obtained from a prospectively-acquired
   natural history cohort collected under NIH protocol 09-I-0032:
   Comprehensive multimodal analysis of neuroimmunological diseases of the
   CNS ([98]clinicaltrials.gov identifier: [99]NCT00794352). All patients
   signed the informed consent and the study was approved by NIH
   Institutional Review Board (CNS IRB). All patients were assigned
   alphanumeric codes and the personnel generating immunophenotyping or
   volumetric MRI data were blinded to the diagnostic conclusion, clinical
   data or BMI status. Only treatment-naïve subjects fulfilling MS
   diagnostic criteria [[100]41] and classified as RRMS by treating
   clinician were included (See Table S2 for demographic and clinical
   information).

2.1.3. Healthy control cohort for monocyte counts and brain volume

   Healthy subjects were prospectively recruited between June 2013 and May
   2017 as part of a Natural History protocol “Comprehensive Multimodal
   Analysis of Neuroimmunological Diseases of the Central Nervous System”
   ([101]ClinicalTrials.gov identifier: [102]NCT00794352) at the National
   Institutes of Health. The inclusion criteria were age 18 years or older
   and vital signs within normal range at the time of screening visit. The
   exclusion criteria included systemic disorder or central nervous system
   disease of any kind, history of alcohol or substance abuse, MRI
   contraindications and history of auditory disorder (See Table S3 for
   demographic information).

2.1.4. Healthy control cohort for lipidomic analysis

   Forty plasma samples for lipidomic analysis of healthy controls (see
   Table S4 for demographic information). were purchased from BioIVT. They
   included 20 samples from healthy controls with normal BMI (10 males and
   10 females with a BMI range of 21–24) and 20 from high BMI subjects (10
   males and 10 females with a BMI range of 26–40).

2.1.5. Immunophenotyping

   For the primary RRMS cohort, samples were analyzed blinded. Fresh whole
   blood was stained in FACS staining buffer for monocytes
   (CD45^+CD3^−CD14^+ cells), CD8 T cells (CD45^+CD3^+CD8^+ cells), and
   CD4 T cells (CD45^+CD3^+CD4^+ cells) and CD4 T cell subsets, naïve/Th1
   (CD3^+CD4^+CCR4^−CCR6^− cells), Th17 (CD3^+CD4^+CCR4^+CCR6^+ cells),
   and T[reg] (CD3^+CD4^+CD127^−CD25^+ cells). To minimize variability,
   staining and acquisition were performed on the same day of blood draw
   within 3 h using the optimized antibody cocktail. Data were acquired on
   a Becton Dickinson LSR Fortessa (BD) equipped with five lasers (355 nm,
   407 nm, 488 nm, 532 nm, and 633 nm wavelengths) with 22 PMT detectors
   at the Human Immune Monitoring Center at Icahn School of Medicine. Data
   were acquired using DIVA 6.1.2 software (BD). A minimum of 500,000
   cells was recorded from corresponding tubes in order to accurately
   assess minor cell populations. Compensation was performed with
   unstained cells and BD compensation beads particle sets. FlowJo 9.4
   software (Treestar Inc.) was used for post-acquisition analysis. Debris
   and doublets were excluded using light scatter measurements and major
   cell populations were identified based on their forward and side
   scatter properties. Subsequently, cells were gated using CD45 for
   leukocytes. Each cell population was represented as percentages of the
   parent population (i.e. CD4^+T cells as % of CD3^+ cells, T[reg] as %
   of total CD4^+ T cells etc.). Absolute cell counts were calculated by
   multiplying the percentage of each cell type compared to the total
   population by the white blood cell count.

   For the validation RRMS cohort and healthy control cohort, whole blood
   immunophenotyping was performed as previously described [[103]43]. All
   samples were labeled with a prospectively assigned alpha-numeric code
   and personnel performing the studies were blinded to the diagnosis of
   the subject. Specimen collection, handling and processing was performed
   according to written standard operating procedures (SOPs).
   Immunophenotyping of peripheral blood cells was performed on
   anticoagulated blood within 60 min of ex vivo collection after osmotic
   lysis of erythrocytes. A minimum of 10^6 blood cells were stained
   according to a previously established protocol [[104]44], which
   included blocking of Fc receptors by 2% intravenous immunoglobulin
   (IVIg). Monocytes were identified based on forward/side scatter and
   gating on cell lineage markers (CD45^+CD3^−CD14^+ cells). Cells were
   immediately acquired on a BD LSR II with High Throughput Sampler (HTS)
   delivery system and analyzed with FACSDiva 6.1 software (all BD
   Biosciences). Gating was based on isotype controls. Sample acquisition,
   gating and sample exclusion (based on the review of quality of the
   staining and of absolute numbers of acquired events to assess
   reliability of data) was done on coded samples.

   Statistical differences in immune cell counts between normal and high
   BMI categories were adjusted for covariates, age and gender, by One-way
   MANCOVA on SPSS, followed by pair-wise comparisons between the two
   groups with Bonferroni correction to test for statistical significance.
   P values <.05 were considered significant.

2.1.6. Lipidomic Analysis

   All lipid standards were purchased from Avanti Polar Lipids, Nu-Chek
   Prep, Sigma-Aldrich, Cambridge Isotope Laboratory, Cayman Chemical
   Company, Avanti Polar Lipids, or Santa Cruz Biotechnology, Inc. All
   solvents are of HPLC or LC/MS grade and were acquired from
   Sigma-Aldrich, Fisher Scientific, or VWR International.

   Blinded lipidomic analyses of plasma samples randomized as to the BMI
   group, were performed at BERG, LLC. A cocktail of deuterium-labeled and
   odd chain phospholipid standards from diverse lipid classes was added
   to 25 μl plasma. Standards were chosen so that they represented each
   lipid class and were at designated concentrations chosen to provide the
   most accurate quantitation and dynamic range for each lipid species.
   4 ml chloroform:methanol (1:1, v/v) was added to each sample and the
   lipid extraction were performed as previously described
   [[105]45,[106]46]. Lipid extraction was automated using a customized
   sequence on a Hamilton Robotics STARlet system (Hamilton, Reno, NV) to
   meet the high-throughput requirements. Lipid extracts were dried under
   nitrogen and reconstituted in 68 μl chloroform: methanol (1:1, v/v).
   Samples were flushed with nitrogen and stored at −20 °C. Samples were
   diluted 50 times in isopropanol: methanol: acetonitrile: water
   (3:3:3:1, by vol.) with 2 mM ammonium acetate in order to optimize
   ionization efficiency in positive and negative modes. Electrospray
   ionization-MS was performed on a TripleTOF® 5600^+ (SCIEX, Framingham,
   MA), coupled to a customized direct injection loop on an Ekspert
   microLC200 system (SCIEX). 50 μl of sample was injected at a flow-rate
   of 6 μl/min. Lipids were analyzed using a customized data independent
   analysis strategy on the TripleTOF® 5600^+ allowing for MS/MS^ALL high
   resolution and high mass accuracy analysis as previously described
   [[107]47]. Quantification was performed using an in-house library on
   MultiQuant™ software (SCIEX).

   Abundance of lipids was quantified as relative to a lipid standard with
   known concentration and ionization efficiency. Lipid standards
   represented each lipid class so that abundance of lipids within one
   class could be compared. Data was filtered and normalized prior to
   statistical analysis. Lipids that had missing values in over 50%
   samples because they were below the detection limit were excluded. For
   remaining lipids, missing values were imputed by sampling a random
   uniform distribution ranging from the lowest value detected to one half
   the minimum value detected. Lipids were normalized using a variance
   stabilization procedure to reduce the effect of lipid abundance on
   sample variation for each species per lipid class and this was followed
   by median centering and log transformation. Because the MS patients and
   healthy control samples were obtained at distinct time points, they
   were run in separate batches and therefore independently analyzed for
   intra-group differences between low and high BMI. To determine
   statistically significant differences in the abundance of lipids
   between high and normal BMI groups within each lipid class, we used
   multiple t-tests with 5% FDR correction on Graphpad Prism v7.0 and q
   values less than p < .05 were considered significant.

   The relationship between differentially abundant ceramides (only 18:1
   backbone) and brain volume was analyzed by Pearson's correlation. Only
   patients who had lipidomic and MRI information were included in the
   analysis and p values <.05 were considered significant. Graphpad Prism
   v7.0 was used to perform the statistical analysis and create the heat
   map.

2.1.7. DNA methylation analysis

   CD14 cells were positively isolated using magnetic beads (Miltenyi
   Biotec) and DNA was extracted with QIAamp DNA Blood Mini Kit (Qiagen).
   500 ng of genomic DNA from each sample were bisulfite-treated with
   Methylamp One-Step DNA Modification Kit (EpiGentek) and methylation
   levels were measured at ~450,000 CpG sites by using the Illumina 450 K
   Array at the New York Genome Center in two different batches, which
   were randomized and blinded to the personnel running the samples.
   Methylation values for CpG sites were measured as beta values, or the
   ratio of signal intensity between the methylated probe and the sum of
   intensities of both methylated and unmethylated probes, and as M
   values, or the log2 ratio of the signal intensity of the methylated
   probe versus that of the unmethylated one. Because beta values have
   more intuitive value for biological interpretation, while M values are
   more valid for obtaining statistical significance [[108]48], beta value
   differences were used to methylation difference values and M values
   were used to determine statistical significance. Probes with a
   detection p value of >0.01 and bead count <3 in >5% of the samples were
   filtered out. Non-CG probes, those close to SNPs (as defined by Zhou et
   al. [[109]49]) or aligned to multiple locations (as defined by Nordlund
   et al. [[110]50]), or on sex chromosomes, were removed. Beta values and
   M values were normalized with BMIQ v1.6 to correct for Type-I and
   Type-II probe bias. Batch effects were removed by using the combat
   algorithm implemented in the ChAMP package [[111]51]. PCA analysis of
   the normalized data did not reveal any outliers in the samples. As we
   have previously described [[112]52,[113]53], our analysis was focused
   on genomic regions rather than individual CpG sites as regulatory DNA
   modifications generally involve multiple consecutive CpGs. To identify
   differentially methylated regions between overweight/obese MS patients
   (BMI ≥ 25) and normal BMI MS patients (BMI < 25), we utilized LIMMA
   package [[114]54] and used a multilinear regression model, which
   identified significant correlations (p < .05) between M values at each
   individual CpG site and BMI (i.e. partial regression) after controlling
   for age, sex, race, EDSS, disease duration, and batch effect. We then
   used a 1 kb sliding window as implemented in the package DMRcate
   [[115]55] to define genomic regions with closely located CpG sites that
   were either differentially methylated or not. Overlapping regions were
   removed. Based on these regions that were spanning the whole genome, we
   combined each CpG-specific p value from our linear regression model
   within a single region with Stouffer's method [[116]56]. This was
   followed by 1% FDR correction for multiple hypothesis testing.
   Differentially methylated regions (DMRs) were defined as regions with
   an FDR-adjusted combined p value of <0.01. Data analysis was performed
   in R Studio by utilizing the R packages ChAMP [[117]51] for data
   preprocessing and normalization, LIMMA for linear regression [[118]54],
   and DMRCATE for the 1 kb sliding window function and the p value
   combination with the Stouffer method [[119]55] and q value [[120]57]
   for FDR correction.

   Pathway enrichment analysis was performed on hypermethylated CpGs using
   the Kyoto Encylopedia of Genes and Genomes (KEGG) database on Enrichr
   [[121]58,[122]59]. Genes containing individual CpGs, which correlated
   with BMI (p < .05) after adjusting for clinical and demographic
   variables, were used to broaden the gene base. Enrichment scores for
   KEGG terms were calculated as the –log (adjusted p value). Genes with
   the largest methylation differences were determined by averaging the
   beta value differences of each CpG within the DMR.

2.1.8. MRI Measurements

   In a primary cohort of MS patients, brain volumetric analysis was
   conducted on available T1-weighted sequences (TE = 9.5–24.0 ms,
   TR = 359–612 ms) and normalized brain volume (NBV) was measured on
   baseline T1-weighted images using SIENAX [[123]60]. Personnel were
   blinded to the BMI status of the patients. Due to the nature of the
   data, acquired for clinical purposes, further segmentation in gray and
   white matter tissues was not performed. Quantification of T1-weighted
   lesion load was also conducted on available T1-weighted sequences
   (TE = 9.5–24.0 ms, TR = 359–612 ms) at baseline and the two-year
   follow-up. T1 lesions were identified and outlined, for each patient,
   using a semi-automated technique based on user-supervised local
   thresholding (Jim version 7; Xinapse Systems,
   [124]http://www.xinapse.com) at each time point. Changes in lesion load
   were computed as the difference between lesion volume measured on
   follow-up scan and baseline lesion volume. Changes in NBV could not be
   calculated since the available sequences from baseline and follow-up
   were not comparable.

   In a validation cohort of MS patients and healthy controls, the volume
   of brain structures were calculated using LesionTOADS [[125]61] tissue
   segmentation with T1- and T2-weighted MRI images. Total brain volume
   was defined as the sum of white and gray matter and brain parenchymal
   fraction was calculated as the total brain percentage of the sum of
   total brain and total CSF volume.

   Correlations between monocyte counts and brain volume were performed
   using linear regression. Evaluation of changes in T1 lesion volume were
   performed using chi-squared test. All analyses were performed on
   Graphpad Prism v7.0 and p values <.05 were considered significant.

2.2. Cell culture experiments

2.2.1. Monocytic cell line culture

   THP1 cells (TIB-202, ATCC) were derived from a one-year old male infant
   with acute monocytic leukemia. Cells were cultured using complete RPMI
   (RPMI 1640 Medium (Sigma), 10% FBS, 1% Penicillin-Streptomycin, 1%
   L-Glutamine, 1% HEPES, 1% Non-essential amino acids, 1% MEM Sodium
   Pyruvate, 0.05 mM 2-Mercaptoethanol). For visualization of NBD-labeled
   Ceramide C16, cells were treated with 5 uM NBD-Ceramide C16 for 24 h.
   To assess the effects of a single ceramide, cells were treated with
   100 nM Ceramide C16:0 (Avanti Polar Lipids) for 1, 12, or 24 h, or
   untreated, and harvested simultaneously. Cells were treated with a
   cocktail of ceramides at different concentrations – 1. Low: 100 nM
   Ceramide C16:0 (Avanti Polar Lipids), 300 nM Ceramide C22:0 (Avanti
   Polar Lipids), 1 uM Ceramide C24:1 (Cayman Chemicals) and 2. High:
   120 nM Ceramide C16:0 (Avanti Polar Lipids), 420 nM Ceramide C22:0
   (Avanti Polar Lipids), 1.3 uM Ceramide C24:1 (Cayman Chemicals), with
   or without 1 uM 5-aza-2′-deoxycytidine (Sigma) for 1 h followed by a
   23 h wash period or for 24 h. For visualization of NBD-labeled Ceramide
   C16, cells were treated with 5 uM NBD-Ceramide C16 for 24 h.

2.2.2. Immunostaining

   THP1 cells were grown on LabTekII chamber slides during the ceramide
   treatment and then fixed with 4% PFA for 15 min. DNA methylation levels
   were assessed by anti-5-methylcytosine staining as previously described
   [[126]62]. Briefly, cells were permeabilized with 0.4% Triton X-100 for
   15 min followed by denaturation with 2 N HCl for 15 min and
   neutralization with 100 mM TrisHCl for 10 min. Cells were blocked with
   10% goat serum, 3% bovine serum albumin (BSA) in PBS containing 0.1%
   Triton X-100 for 1 h. Cells were incubated in anti-5-methylCytosine
   (Eurogentec, 1:500) and anti-Ki67 (Abcam, 1:1000) to measure
   proliferation at 4C overnight. Cells were stained with anti-mouse
   AlexaFluor 546 (Invitrogen, 1:500) and anti-rabbit AlexaFluor 488
   (Invitrogen, 1:500) the next day for 2 h followed by washing and
   mounting with Fluoromount-G containing DAPI. For visualization of
   NBD-Ceramide C16, cells were blocked with 10% goat serum, and 0.1%
   Triton X-100 in PGBA (0.1 M phosphate buffer, 0.1% gelatin, 1% BSA,
   0.002% sodium azide) for 1 h followed by overnight incubation with
   anti-4-fluoro-7-nitrobenzofurazan (NBD) (BioRad, 1:100) at 4C. Cells
   were stained with anti-rabbit AlexaFluor 488 (Invitrogen, 1:500) the
   next day for 1 h and mounted with Fluoromount-G containing DAPI.

2.2.3. Imaging

   Images were acquired at 20× on the Zeiss LSM800 and 3 images per well
   were taken. For each experiment, 3–4 wells were used per condition, and
   a total of 3 independent experiments were performed. Mean intensity was
   analyzed on ImageJ. Briefly, DAPI^+ nuclei were used to create Regions
   of Interest (ROIs) using the Analyze Particles function and added to
   the ROI manager. Mean gray values of each ROI were calculated for
   5-methylcytosine and Ki67 stainings. An arbitrary threshold for each
   staining was set to determine 5-methylcytosine and Ki67 positive cells
   that was validated by manual counting. Percentage of positive cells
   over total cells was calculated for each image. The percentages for all
   images within one well were averaged followed by averaging of
   percentages in wells for each experiment. For statistical analysis,
   Student's t-test was performed between two groups, and Dunnett's
   multiple comparisons test was used among three groups and was
   calculated on Graphpad Prism v7.0. P values <.05 were considered
   significant. 3D reconstruction of cells treated with NBD-ceramide C16
   was performed on Imaris.

2.2.4. MTT

   The MTT assay was performed as previously described [[127]63]. THP1
   cells were treated with different doses of 5-aza-2′-deoxycytidine (0,
   0.5, 1, 2, 5 uM). After 24 h of treatment, 5 mg/ml MTT was added to the
   cell media to a final concentration of 0.05 mg/ml and allowed to
   incubate for 3 h to allow MTT reduction. The formazan precipitate
   formed by viable cells was solubilized in DMSO, and read
   spectrophotometrically at 540 nm with background subtraction at 655 nm.

2.2.5. Sequenom mass array EpiTyper

   Genomic DNA was isolated using the DNeasy Blood and Tissue Kit
   (Qiagen). DNA purity was assessed by measuring the A260/A280 ratio
   using NanoDrop. 200 ng genomic DNA was sodium bisulfite–treated using
   the Methylamp One-step Modification Kit (Epigentek). Primers were
   designed using EpiDesigner software: NRXN1 intron forward:
   GTGTTATTGAGGAGTTAATGGTTGA, NRXN1 intron reverse:
   ATTTCCCCTCTATATTTAACAATCACA, FZD7 promoter forward:
   TTATTGAGAAGTGATTTGGAAGTGA, FZD7 promoter reverse:
   ACCTCACAAAAATCTAAACAAACCA, TP63 promoter forward:
   GTTTTAAATTGTTAATAATAAGTGTGGTT, TP63 promoter reverse:
   AAAATTTAAACACTTCTCTCCTTCA, TP63 exon forward:
   TTTTTTTGTAAATATGTATGAAGGAGAGA, TP63 exon reverse:
   AAAAAATTATATCACATTCAAACTCACC. Forward primers were designed with a
   10-mer tag (AGGAAGAGAG) and reverse primers designed with a T7-promoter
   tag (CAGTAATACGACTCACTATAGGGAGAAGGCT), as per the manufacturer's
   guidelines. PCR products were then processed as previously described to
   determine methylation levels. CpGs overlapping silent peaks, outside
   the mass spectrometry analytical window (low or high mass), or lacking
   sufficient coverage (methylation level determined for <90% of samples)
   were filtered out before subsequent analysis. Statistical analysis for
   the overall region (taking all individual CpGs into account) was
   performed by Two-way ANOVA on Graphpad Prism v7.0 and p values <.05
   were considered significant.

2.2.6. Quantitative real time PCR

   RNA was extracted from THP1 cells using Trizol (Invitrogen) extraction
   and isopropanol precipitation. RNA samples were resuspended in water
   and further purified with RNeasy columns with on-column DNase treatment
   (Qiagen). RNA purity was assessed by measuring the A260/A280 ratio
   using NanoDrop. For qRT-PCR, RNA was reverse-transcribed with qScript
   cDNA Supermix (Quanta) and performed using PerfeCTa SYBR Green FastMix,
   ROX (Quanta) at the Epigenetics Core at the Advanced Science Research
   Center at CUNY. Primer sequences are as follows: NRXN1 forward:
   TAAGTGGCCTCCTAATGACCG, NRXN1 reverse: TCGCACCAATACGGCTTCTTT, FZD7
   forward: GTGCCAACGGCCTGATGTA. FZD7 reverse: AGGTGAGAACGGTAAAGAGCG TP63
   forward: TCCTCAGGGAGCTGTTATCC, TP63 reverse: ATTCACGGCTCAGCTCATGG,
   RPLP0 forward: GCGACCTGGAAGTCCAAC, RPLP0 reverse: GTCTGCTCCCACAATGAAAC.
   3 technical triplicates per biological replicate, 3 biological
   replicates per experiment, and 3 independent experiments were
   performed. After normalization to the housekeeping gene, RPLP0, CT
   values of technical triplicates were averaged for each biological
   replicate, followed by averaging of biological replicates for each
   independent experiment. Values were normalized to the average value of
   control conditions (i.e. “Untreated” or “Low”). Statistical analysis
   was performed on Graphpad Prism v7.0: Student's t-test was performed
   between two groups, and Dunnett's multiple comparisons test was used
   among three groups. P values <.05 were considered significant.

2.2.7. Generation of NBD-Ceramide C16

   NBD Sphingosine was purchased from Avanti® Polar Lipids, Inc. (Cat. Nr.
   810205). All other reagents and solvents were purchased from commercial
   sources (Sigma-Aldrich, Fisher Scientific, etc.) and were used without
   further purification. Reaction progress was monitored by LC-MS
   analysis. HPLC (1260 infinity II, Agilent) was used for purification of
   NBD-Ceramide C16:0.

   Stock solutions of palmitic acid (0·05 M), TBTU (0·05 M), DIPEA
   (0·05 M), and NBD sphingosine (0·01 M) in EtOAc were prepared. Palmitic
   acid (3·00 ul, 0·15 umol), TBTU (3·00 ul, 0·15 umol), and DIPEA (6·00
   ul, 0·30 umol) were added to a microcentrifuge tube (1·5 ml) and
   incubated for 1 h under mild orbital shaking. NBD sphingosine (10·0 ul,
   0·10 umol) was added and the reaction mixture was incubated for 16 h
   under mild orbital shaking (starting material completely consumed by
   LC-MS). MeOH was added and the reaction mixture was transferred into a
   glass vial. The crude mixture was purified by preparative HPLC (H[2]O &
   1% Formic Acid (10%) in CH[3]CN to CH[3]CN) to yield NBD-Ceramide C16:0
   as a yellow solid with approximately 80% recovery rate (yield estimate
   used for concentration calculations: 57·3 μg, 0·08 umol). HRMS (EI^+):
   m/z calcd. For [C[40]H[70]N[5]O[6]] ^+: 716·5321, found: 716·5326
   ([M + H]^+).

2.3. Animal experiments

2.3.1. Animals

   All experiments were performed according to IACUC-approved protocols
   and mice were maintained in a temperature- and humidity-controlled
   facility on a 12-h light-dark cycle with food and water ad libitum.
   Female C57/bl6 mice (The Jackson Laboratory) were fed a low-fat diet
   (2914, Envigo) or high fat diet (TD.96132, Envigo) starting at
   3–4 weeks of age. Experimental autoimmune encephalitis (EAE) was
   induced after 5 weeks of diet as previously described [[128]64].
   Briefly, mice were injected subcutaneously with 200 μg of MOG[35–55]
   thoroughly emulsified in Complete Freund's Adjuvant (CFA) with 500 μg
   Mycobacterium tuberculosis dissolved in 200 μl of a solution consisting
   of equal amounts of 1.5 mg/ml MOG peptide, 2.5 mg/ml Mycobacterium
   tuberculosis and 50% CFA. 500 ng of pertussis toxin dissolved in 200 μl
   of PBS were administered intraperitoneally at the time of the
   MOG[35–55]injection and 2 days thereafter. Disease progression was
   monitored daily using a standard EAE scoring system (modified from the
   Hooke Lab recommendations) as following: 0 = no visible symptoms,
   1 = flaccid tail, 1·5 = limp tail and hind leg inhibition, 2 = limp
   tail and weak hind legs, 2·5 = limp tail and dragging of hind legs,
   2·75 = limp tail and paralysis of one hindlimb, 3 = limp tail and
   complete hindlimb paralysis, 3·5 = limp tail, complete hindlimb
   paralysis, and weakness of forelimbs, 4 = quadriplegia, and 5 = death.

   For analysis of infiltrating monocytes in the CNS, Cx3cr1-Gfp mice
   (B6.129P-Cx3cr1^tm1Litt/J) and Ccr2-Rfp mice
   (B6.129(Cg)-Ccr2^tm2.1Ifc/J) were purchased from The Jackson Laboratory
   and crossed to generate Cx3cr1^Gfp/+;Ccr2^Rfp/+ mice. Heterozygous
   female mice were used for the experiments and fed a low-fat diet (2914,
   Envigo) or high fat diet (TD.96132, Envigo) starting at 4 weeks of age
   for a period of 5 weeks.

2.3.2. Analysis of disease course

   Changes in weight were calculated after 5 weeks of diet as a measure of
   weight gain and were assessed by Student's t-test in low fat diet and
   high fat diet groups. Blood glucose test strips (Contour) were used to
   measure blood glucose levels as an indication of pre-diabetes and
   differences were assessed by Student's t-test. EAE scores were recorded
   to evaluate disease course. The difference in weight from baseline
   weight (weight at immunization) was calculated for each time point
   throughout disease course. Differences in weight change for individual
   time points were assessed by Sidak's multiple comparisons test.
   Differences in clinical severity on individual time points were
   assessed by Mann-Whitney U test. Area under the curve and maximum EAE
   score were also assessed for each mouse and compared in low fat diet
   and high fat diet fed mice by Mann-Whitney U test. Calculations and
   statistical analysis were performed on Graphpad Prism v7.0 and p values
   <.05 were considered significant.

2.3.3. Immunostaining

   Mice were sacrificed at the peak of EAE for histological analysis. Mice
   were anaesthetized and perfused with 4% (w/v) paraformaldehyde. Spinal
   cords were removed, post-fixed, and cryopreserved in 30% (w/v) sucrose,
   embedded in OCT and sectioned coronally (12 μm). Sections were blocked
   with 10% goat serum, and 0.1% Triton X-100 in PGBA (0.1 M phosphate
   buffer, 0.1% gelatin, 1% BSA, 0.002% sodium azide) for 1 h followed by
   overnight incubation with primary antibodies – NFH (Millipore, 1:400),
   CD45 (BD Biosciences, 1:100), Iba1 (Wako, 1:500), and 5-methylCytosine
   (Zymo Research, 1:100) at 4C overnight. For 5-methylCytosine staining,
   antigen retrieval was performed using sodium citrate buffer at 95C for
   15 min prior to blocking. Sections were washed and incubated with
   secondary antibodies, anti-mouse AlexaFluor 546 (Invitrogen, 1:500),
   anti-rabbit AlexaFluor 488 (Invitrogen, 1:500) and anti-rat AlexaFluor
   647 (Invitrogen, 1:500), followed by counterstaining with DAPI
   (1:20,000) and mounting with Immunomount (Electron Microscopy
   Sciences). Fluoromyelin Green (Invitrogen, 1:300) was applied to
   sections for 20 min at room temperature after incubation with primary
   antibodies.

2.3.4. Imaging

   To assess demyelination, axonal loss, and monocyte infiltration, images
   were acquired at 20× on the Zeiss LSM800. At least three images per
   section were acquired in two sections per animal and at least three
   animals were analyzed for each condition. Images were taken as a
   Z-stack and the maximum projection was acquired using the Zeiss Zen
   software. Areas were calculated on ImageJ. Fluoromyelin and NFH+ areas
   were delineated manually. To calculate CD45^+ area, images were
   thresholded and the area calculated using the “analyze particles”
   function. Regions of interest (ROIs) were defined as CD45^+ areas, and
   Iba1^+ area was calculated within CD45^+ areas using a threshold. Total
   Iba1^+CD45^+ area for each image was calculated as the sum of Iba1^+
   area for all CD45^+ areas or ROIs in each image. ROIs were limited to
   the lesion area. To assess DNA methylation levels, images were acquired
   at 63× on the Zeiss LSM800. 3 images per section were taken and 2
   sections per animal and 3 animals per condition were assessed.
   Intensity was defined as mean gray value and calculated on Image J. 5
   representative nuclei of Iba1^+ cells were selected manually in each
   image and 5mc and DAPI intensity were calculated. For statistical
   analysis, Student's t-test was performed on Graphpad Prism v7.0 and p
   values <.05 were considered significant.

2.3.5. Flow cytometry

   Brains and spinal cords were removed and homogenized. Mononuclear cells
   were separated with a 30%/70% Percoll (GE Healthcare) gradient as
   previously reported [[129]65]. Single-cell suspensions were analyzed on
   a FACSAria II Cell Sorter (BD) equipped with five lasers (355 nm,
   407 nm, 488 nm, 532 nm, and 633 nm wavelengths) running DIVA 6.1.2
   software (BD). Data were analyzed with FlowJo v10.0.7 software (Tree
   Star). Briefly, mononuclear cells were identified by size (forward and
   side scatter) and doublets excluded. Infiltrating monocytes were
   identified as CCR2-RFP^+ and microglia as CX3CR1-GFP^+. Percentages of
   CCR2-RFP^+ and CX3CR1-GFP^+ cells of total singlets were calculated for
   5 mice per group. For statistical analysis, Student's t-test was
   performed on Graphpad Prism v7.0 and p values <.05 were considered
   significant.

2.4. Quantification and statistical analysis

   Analysis of immunophenotyping was performed on SPSS. DNA methylation
   analysis and ordinal logistic regression of BMI and EDSS change were
   analyzed using R. All other statistical analyses were done using
   GraphPad Prism. For all graphs, error bars are mean ± SEM. All
   statistical details for each graph can be found in the figure legends.

2.5. Data and software availability

   Sequencing data deposited in NCBI's Gene Expression Omnibus Series
   accession number [130]GSE103929.

3. Results

3.1. Immune profiling reveals increased monocytic counts in MS patients with
high BMI

   The discovery cohort consisted of fifty-four patients who had been
   recently diagnosed with Relapsing-Remitting MS (RRMS) and had been
   recruited to participate in the study prior to the initiation of
   disease-modifying therapy (“therapy-naïve”). Patients were classified
   by BMI as normal (average BMI = 21·8 ± 1·9, n = 27) or high (average
   BMI = 30·2 ± 5·0, n = 27). At baseline, the clinical disability score,
   as defined by the Kurtzke Expanded Disability Status Scale (EDSS), did
   not differ by BMI (MS with normal BMI: median ± SD = 2·0 ± 1.1; MS with
   high BMI: median ± SD = 1·0 ± 1·0), as patients were recruited at the
   early stages of disease (Table S1).

   Since MS is an immune-mediated demyelinating disease, we sought to
   define whether the peripheral immune profiles differed in patients with
   different BMIs. Immunophenotyping was conducted using flow cytometry
   for the detection of surface markers. The total number of circulating
   CD4 T cells and related subsets, CD8 T cells and monocytes (CD14^+)
   were calculated (Fig. S1). CD8 T cells and total CD4 T cell counts were
   not significantly different among MS patients with differing BMI, nor
   were CD4 naïve/Th1 (CCR4^− and CCR6^−), Th17 (CCR4^+ and CCR6^+) and
   T[reg] (CD25^hi and CD127^−) subsets ([131]Fig. 1A). However, the
   number of CD14^+ monocytes was significantly higher in MS patients
   within the high BMI group, after adjusting for age and sex
   statistically, a finding that was validated in a second independent
   validation cohort of 91 therapy-naïve RRMS patients (normal BMI
   average = 21·4 ± 1·9, n = 42; high BMI average = 29·6 ± 4·4, n = 51)
   and was not detected in 50 healthy controls (normal BMI
   average = 21·5 ± 2·3, n = 13; high BMI average = 30·2 ± 4·5, n = 37)
   ([132]Fig. 1B). This suggested that in out cohorts, the number of
   circulating monocytes was uniquely affected by the interplay between
   high BMI and MS disease state.

Fig. 1.

   [133]Fig. 1
   [134]Open in a new tab

   Immune profiling reveals a BMI-dependent increase in blood monocyte
   counts in MS patients but not healthy individuals. Immune cell
   populations from fresh whole blood were quantified using flow
   cytometry. (a) Differences in T cell counts due to BMI were assessed in
   a primary cohort of MS patients (n = 52). (b) Differences in monocyte
   counts due to BMI were assessed in a primary cohort of MS patients
   (n = 52), a validation cohort of MS patients (n = 91), and a cohort of
   healthy individuals (n = 50). One-way MANCOVA was used to adjust for
   age and sex, followed by pair-wise comparisons with Bonferroni
   correction to determine statistically significant differences in the
   two BMI groups (normal BMI = white dots, high BMI = gray dots)
   (*p < .05).

3.2. Plasma lipidomic analysis reveals increased abundance of ceramides in MS
patients with high BMI

   Because differences in BMI have been associated with lipid changes in
   several tissues [[135]28,[136]29,[137]31], we asked whether MS patients
   with normal or high BMI also differed in circulating lipids. Plasma
   samples were analyzed using an MS/MS^ALL platform which identified over
   2000 lipids (within 13 lipid classes), and the relative abundance for
   each lipid species was calculated in reference to reliable standards
   with known abundance and ionization efficiency for each lipid class.
   Lipids were excluded if values were below the detection limit in over
   50% of samples, yielding 1030 lipid species that were considered
   suitable for further analysis. Differentially abundant lipids were
   identified by multiple t-tests with 5% FDR correction for each lipid
   class, revealing sphingolipids and neutral lipids as more abundant in
   the high BMI group than in normal BMI MS patients (Fig. S2A). The
   largest percentage difference in abundance was detected for the
   ceramides (38%), a subclass of sphingolipids, followed by triglycerides
   (30%) and diacylglycerols (18%) within the neutral lipids class
   ([138]Fig. 2A). To further determine whether the ceramide abundance in
   the high BMI group of MS patients was specific to the disease state, or
   simply reflected the effect of high body mass, we conducted a similar
   analysis on a cohort of 40 plasma samples from healthy controls with
   either a normal (average = 22·6.5 ± 1·1, n = 20) or high BMI
   (average = 29·4 ± 3·4, n = 20) ([139]Fig. 2B, Fig. S2B). Interestingly,
   the ceramide species with differential abundance in the plasma of MS
   patients with high BMI, were not similarly elevated in healthy
   individuals with high BMI, thereby suggesting that, at least within the
   limited size of the cohort included in our study, the specific ceramide
   species detected in patients were the likely result of the process of
   myelin destruction, characteristic of the disease state ([140]Fig. 2C).
   This concept was further supported by an inverse correlation between
   specific ceramides and brain volume (Fig. S3). Overall these data
   suggested that the increased ceramide species detected in MS patients
   with high BMI, were the likely result of the combined de novo synthesis
   of ceramides from dietary fats and their recycling from myelin, damaged
   by the disease process.

Fig. 2.

   [141]Fig. 2
   [142]Open in a new tab

   Lipidomic profiling reveals increased abundance of ceramide species in
   MS patients with high BMI, but not in healthy individuals. Unbiased
   lipidomic analysis was performed on plasma samples from MS patients
   (n = 48) and healthy individuals (n = 40) using an MS/MS^ALL platform.
   Lipids that differed in abundance due to BMI were assessed using
   multiple t-tests with 5% FDR correction, q < 0.05 were considered
   significant (*q < 0.05, **q < 0.01). (a) Pie chart representation of
   numbers and percentage of lipids in MS patients by lipid family and
   class. Red = increased abundance, blue = decreased abundance,
   white = no change in patients with high BMI. (b) Relative abundance of
   ceramide species with 18:1 backbone that were significantly different
   in abundance in MS patients with high BMI (gray dots) compared to those
   with normal BMI (white). (c) The most highly significantly abundant
   ceramide species in MS patients were not statistically different in
   healthy individuals with high BMI (gray dots) compared to those with
   normal BMI (white). (For interpretation of the references to colour in