ARTICLE OPEN ACCESS

Disease-modifying therapies alter gut microbialcomposition in MSIlana Katz Sand MD Yunjiao Zhu PhD Achilles Ntranos MD Jose C Clemente PhD

Egle Cekanaviciute PhD Rachel Brandstadter MD Elizabeth Crabtree-Hartman MD Sneha Singh BSc

Yadira Bencosme MPH Justine Debelius PhD Rob Knight PhD Bruce AC Cree MD PhD MAS

Sergio E Baranzini PhD and Patrizia Casaccia MD PhD

Neurol Neuroimmunol Neuroinflamm 20196e517 doi101212NXI0000000000000517

Correspondence

Dr Katz Sand

ilanakatzsandmssmedu

AbstractObjectiveTo determine the effects of the disease-modifying therapies glatiramer acetate (GA) anddimethyl fumarate (DMF) on the gut microbiota in patients with MS

MethodsParticipants with relapsing MS who were either treatment-naive or treated with GA or DMFwere recruited Peripheral blood mononuclear cells were immunophenotyped Bacterial DNAwas extracted from stool and amplicons targeting the V4 region of the bacterialarchaeal 16SrRNA gene were sequenced (Illumina MiSeq) Raw reads were clustered into OperationalTaxonomic Units using the GreenGenes database Differential abundance analysis was per-formed using linear discriminant analysis effect size Phylogenetic investigation of communitiesby reconstruction of unobserved states was used to investigate changes to functional pathwaysresulting from differential taxon abundance

ResultsOne hundred sixty-eight participants were included (treatment-naive n = 75 DMF n = 33 andGA n = 60) Disease-modifying therapies were associated with changes in the fecal microbiotacomposition Both therapies were associated with decreased relative abundance of the Lach-nospiraceae and Veillonellaceae families In addition DMFwas associated with decreased relativeabundance of the phyla Firmicutes and Fusobacteria and the order Clostridiales and an increasein the phylum Bacteroidetes Despite the different changes in bacterial taxa there was anoverlap between functional pathways affected by both therapies

InterpretationAdministration of GA or DMF is associated with differences in gut microbial composition inpatients with MS Because those changes affect critical metabolic pathways we hypothesizethat our findings may highlight mechanisms of pathophysiology and potential therapeuticintervention requiring further investigation

These authors are co-first authors and contributed equally to the manuscript

From the Department of Neurology (IKS AN RB YB) Department of Neuroscience (YZ PC) and Department of Genetics amp Genomic Sciences Icahn Institute for Genomics ampMultiscale Biology (JCC) Icahn School of Medicine at Mount Sinai Department of Neurology (EC EC-H SS BACC SEB) Weill Institute for Neurosciences University ofCalifornia San Francisco EC is now with Universities Space Research Association Space Biosciences Division NASA Ames Research Center Moffett Field CA Department ofPediatrics (JD RK) Department of Computer Science amp Engineering (RK) and Center for Microbiome Innovation (RK) University of California San Diego and NeuroscienceInitiative (PC) Advanced Research Science Center at the Graduate Center of the City University of New York

Funding information and disclosures are provided at the end of the article Full disclosure form information provided by the authors is available with the full text of this article atNeurologyorgNN

The Article Processing Charge was funded by Mount Sinai

This is an open access article distributed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivatives License 40 (CC BY-NC-ND) which permits downloadingand sharing the work provided it is properly cited The work cannot be changed in any way or used commercially without permission from the journal

Copyright copy 2018 The Author(s) Published by Wolters Kluwer Health Inc on behalf of the American Academy of Neurology 1

MS is believed to result from interactions between underlyinggenetic predisposition and environmental exposures12 En-vironmental risk factors including vitamin D levelse1 (linkslwwcomNXIA86) viral exposurese2 smokinge3 and obe-sitye4e5 are related to MS onset and disease course Howevermost of the environmental risk factors in MS remain un-explained In consideration of additional potential mediatorsthe gut is a natural site of investigation given that it is a majorlocus of environmental interaction and home to a large por-tion of the human immune system In addition immunologicdevelopment and patterning dysregulated in MS and otherautoimmune diseases is heavily influenced by resident com-mensal microbes the microbiota Gut microbial compositionis shaped by both genetics and environmental exposures thatbegin in utero providing a potential pathogenic link betweenthese factors and autoimmune diseases such as MS Studies ofexperimental allergic encephalomyelitis (EAE) in mice havedemonstrated reduced disease severity with administration oforal antibiotics3 and established that the presence of com-mensal flora is required for EAE induction4

Several groups have begun to study the gut microbiota inpatients with MS to identify a potential disease-associatedsignature5ndash9 However initial studies were limited by smallnumbers of participants and potentially confounded by theeffects of disease-modifying therapies (DMTs) Separatingchanges in microbial composition related to MS disease statefrom changes induced by DMTs is crucial to further un-derstand the role of gut microbiota in MS We recentlyreported structural and functional changes in gut microbiota inuntreated patients with MS compared with healthy controls10in the current study we began to examine the effects of DMT

Therapy-induced changes in microbial composition maycontribute toward efficacy by favoring microbes with anti-inflammatory properties for example by influencing T-celldifferentiation toward regulatory phenotypes or promotingthe production of regulatory cytokines This could manifesteither as a ldquoreversalrdquo of observed differences in baseline gutmicrobiota associated with the MS disease state or in un-related changes that nonetheless promote immune toleranceover inflammation in MS Some changes may be neutral andstill others may be detrimental contributing to incompleteefficacy Changes to individual taxa may be less importantthan the overall reshaping of the microbial community

This cross-sectional study was designed to define the effects of2 commonly prescribed DMTs on gut microbial composition

in patients with MS Glatiramer acetate (GA) is administeredby subcutaneous injection It is a random copolymer of ala-nine lysine glutamic acid and tyrosine with widespread im-munomodulatory effects including effects on antigenpresentation and polarization of naive T cells away from TH1and TH17 and toward TH2 and regulatory phenotypes withaccompanying changes in cytokine profilese6 (linkslwwcomNXIA86) Efficacy in MS was established by several clinicaltrialse7ndashe9 and it has been US Food and Drug Administration(FDA) approved for the treatment of MS since 1996 Di-methyl fumarate (DMF) is an orally administered fumaricacid ester Immunological effects relevant to MS includeantiproliferative effects inhibition of the NF-KB pathwaypromotion of the heme oxygenase pathway and anti-inflammatory effects related to T-cell differentiation and cy-tokine productione10 Efficacy was established by 2 largeclinical trialse11e12 and US FDA approval was granted forMS in 2013

In this study we demonstrate that these 2 immunomodula-tory therapies distinctly alter gut microbial composition inpatients with MS

MethodsParticipants were recruited at the Corinne Dickinson Gold-smith Center for Multiple Sclerosis at Mount Sinai (NewYork NY) between August 2013 and December 2015 and atthe University of California San Francisco Multiple SclerosisCenter (San Francisco CA) between July 2013 and Sep-tember 2016 as part of the MS Microbiome ConsortiumPotential participants aged 18ndash65 years were approached byMS Center clinicians during clinical visits Patients with MSwere eligible to participate if they met the McDonald 2010criteria for relapsing-remitting MS or qualified as havingclinically isolated syndrome along with abnormal brain MRIand were either naive to MS DMT or stable on GA or DMFfor at least 3 months Participants could not have takenantibiotics within 3 months or high-dose corticosteroidswithin 1 month of enrollment Additional exclusions werea diagnosis of diabetes or inflammatory bowel disease recentgastroenteritis and treatment with an immunosuppressantmedication for any condition in the 3 months precedingenrollment

Clinical dataClinical details including demographic information detailedMS disease history medical history and height and weight

GlossaryDMF = dimethyl fumarate DMT = disease-modifying therapy EAE = experimental allergic encephalomyelitis FDA = Foodand Drug Administration GA = glatiramer acetate IFN = interferon LEfSe = linear discriminant analysis effect size LDA =linear discriminant analysis OTU = operational taxonomic unit PICRUSt = phylogenetic investigation of communities byreconstruction of unobserved states

2 Neurology Neuroimmunology amp Neuroinflammation | Volume 6 Number 1 | January 2019 NeurologyorgNN

were recorded The Expanded Disability Status Scale wascompleted by the treating MS clinician or abstracted from theelectronic medical record by one of the study investigators(IKS and RB) Statistics were performed using SPSS 23

Microbiota

Sample collectionResearch coordinators reviewed detailed instructions regardingstool collection with all participants and provided them stoolkits to be completed at home Care was taken to match the kitsand instructions at both institutions using a dry swab techni-que (BD 220135 Becton Dickinson and Company FranklinLakes NJ) Participants were instructed to collect stool samplesusing the first bowel movement of the day on 2 separate dayspreferably 2 days within the same week They were theninstructed to place each completed kit between 2 freezer packsinside a thermal envelope and ship immediately or keep in theirhome freezer if there was to be a short delay in shipping Aprepaid return shipping label for overnight shipping was pro-vided tominimize transit time Participants were also instructedto send samples only in the early part of the week to avoidsamples arriving over the weekend On receipt samples wereimmediately placed at minus80degC until bacterial DNA extraction

16S rRNA amplicon sequencing and data analysisThe workflow for microbiota sequencing and analysis isoutlined in figure 1 DNA was extracted from samples usingthe PowerSoil DNA Isolation Kit (MoBio 12888) accordingto the manufacturerrsquos instructions For each sample PCRtargeting the V4 region of the bacterialarchaeal 16S rRNAgene was completed in triplicate using the 515806 primerpair and amplicons were sequenced using Illumina MiSeq(150 bp paired-end) sequencing primers and procedureswere described in the Earth Microbiome Project standardprotocol11

Raw reads were clustered into Operational Taxonomic Units(OTUs) using closed-reference SortMeRNA12 method at 97identity based on the GreenGenes database v138 using Quan-titative Insights Into Microbial Ecology (QIIME) v1913 Tax-onomy was assigned to the retained OTUs based on theGreenGenes reference sequence and the GreenGenes tree wasused for all downstream phylogenetic community comparisonsSamples with less than 10000 sequences per sample were fil-tered out OTUs were filtered to retain only OTUs present in atleast 5 of the samples and containing at least 100 reads Severalvery rare (relative abundance lt5 times 10minus5) OTUs includingOTUs 4295564 (k__Bacteria p__[Thermi] c__Deinococci

Figure 1 Flow chart explaining the bioinformatic pipeline and statistical methods (see Methods for details)

NeurologyorgNN Neurology Neuroimmunology amp Neuroinflammation | Volume 6 Number 1 | January 2019 3

o__Thermales f__Thermaceae g__Thermus s__) 112194(k__Bacteria p__Proteobacteria c__Alphaproteobacteriao__Rickettsiales f__Rickettsiaceae g__Rickettsia s__) and4319519 (k__Bacteria p__Spirochaetes c__Spirochaeteso__[Borreliales] f__[Borreliaceae] g__Borrelia s__) wereidentified as potential contaminants and were removed Sam-ples collected from the same participant within 1 week werecombined to 1 sample by averaging the OTU tables

After filtering samples were rarefied to 10000 sequences persample and alpha and beta diversities were computed usingQIIME v19 Unweighted UniFrac14 was used as the distancemetric to perform principal coordinate analysis and to generatethe corresponding plots Alpha-diversity differences across the3 treatment groupswere tested by the nonndashparametric Kruskal-Wallis rank test Beta diversity differences were assessed bycomparing the within-group and between-group unweightedUniFrac distances using a nonndashparametric 2-sample t-testbased on 1000 Monte Carlo permutations followed bya Bonferroni correction for multiple testing Differentialabundance analysis was performed using linear discriminantanalysis effect size (LEfSe)15 Briefly the nonndashparametricKruskal-Wallis rank test was first applied to detect differentiallyabundant taxa Significant taxa (p lt 005) were then used tobuild a linear discriminant analysis (LDA) model which esti-mated the effect size associated with a treatment group plt 005and LDA score gt2 were used as the criteria to determine dif-ferentially changed taxa after treatment compared with samplesfrom those who were treatment-naive

To infer metagenomic functions (and predict metabolicfunctions from the predicted gene content) of the bacteriasignificantly altered by DMTs the software package phylo-genetic investigation of communities by reconstruction ofunobserved states (PICRUSt) v11116 was applied to allOTUs (2191 OTUs) after quality filtering as describedabove following the recommended pipeline of normalizingOTUs by copy number (to account for copy number differ-ences of the 16S rRNA gene among taxa) predicting func-tions using the Kyoto Encyclopedia of Genes and Genomesorthologs and assembling predicted Kyoto Encyclopedia ofGenes and Genomes orthologs into functional pathwaysDifferentially changed metabolomic pathways between thetreated and naive groups were identified using the nonndashparametric Kruskal-Wallis rank test followed by theBenjamini-Hochberg correction for multiple testing witha false discovery rate le15 reflecting the exploratory natureof this piece of the study

ImmunophenotypingWhole blood was collected for immunophenotyping (only atMount Sinai because of the requirement for fresh samples)and immediately processed byMount Sinairsquos Human ImmuneMonitoring Core Specimens were stained with a pre-optimized T-cell antibody cocktail and acquired within 3hours using a BD LSR Fortessa (BD San Jose CA) A min-imum of 500000 events were recorded from each sample to

accurately assess minor cell populations Compensation wasperformed with unstained cells and BD compensation beadsFlowJo 94 software (Treestar Inc San Carlos CA) was usedfor postacquisition analysis Statistical analysis was performedusing unpaired 2 tailed t-test (GraphPad Software LaJolla CA)

Standard protocol approvals registrationsand patient consentsThe research was approved by the Institutional ReviewBoards at Mount Sinai and UCSF and written informedconsent was obtained from all participants

Data availability statementAnonymized data are available and will be shared on requestfrom any qualified investigator

ResultsThe study enrolled a total of 186 participants who met theinclusion criteria Of these 168 (903) had at least 2 sampleswhose DNA passed quality control measures and were ulti-mately included This includes 75 participants who weretreatment-naive 33 treated with DMF and 60 treated withGA as outlined in figure 1 Immunophenotyping data wereavailable only for a subset of the participants recruited atMount Sinai (n = 30 naive n = 26 DMF and n = 12 GA)Clinical characteristics of study participants broken down bytreatment group are outlined in table 1 (microbiota) andtable e-1 (linkslwwcomNXIA89 immunophenotyping)

Immunophenotyping profiles validated theexpected effects of DMF and GAPeripheral blood mononuclear cells were collected from 68participants recruited at Mount Sinai (treatment-naive n = 30DMF n = 26 and GA n = 12) and analyzed by FACS toidentify CD3 CD4 and CD8 populations The surfacemarkers CCR4 and CCR6 were used to further classify CD4T cells as Th17 (CCR4+CCR6+)e13 (linkslwwcomNXIA86) or Th2 (CCR4+CCR6minus)e14

DMF treatment was associated with reduced percentage ofCD3 CD4 and CD8 T cells as previously describede15 (linkslwwcomNXIA86) DMF treatment was also associatedwith a reduced percentage of CCR4+ CCR6+ T cells (t-test plt 00001) as well as a milder reduction of CCR4+CCR6minusT cells (t-test p = 002) Treatment with GA did not have anyeffect on total CD3 or CD4 percentage but was associatedwith a reduced percentage of CCR4+ CCR6+ T cells com-pared with treatment-naive patients (t-test p = 002) whichwas not significantly different from the effect of DMF on thosecells (t-test p = 02) (figure e-1 linkslwwcomNXIA87)

Administration of DMF or GA results in distinctalterations to gut microbial composition inpatients with MS16S amplicon sequencing was performed on the IlluminaMiSeq platform to assess changes in the fecal microbiome of

4 Neurology Neuroimmunology amp Neuroinflammation | Volume 6 Number 1 | January 2019 NeurologyorgNN

patients with MS after different treatments In total49679344 raw reads passed the quality filtering in QIIMEand were clustered into OTUs based on the GreenGenesdatabase at a 97 sequence identity We further appliedstringent filters to remove samples with low sequencing readsand rare taxa (see Methods for detail) and then combinedmultiple collections within a short time interval from the sameparticipant to counteract temporal variations (figure 1) Thisresulted in 8542763 high-quality reads (on average 49628 plusmn31746 reads per participant) mapping to 2191 uniqueOTUs At the phylum level the microbiota of all patients withMS was comprised predominately of Firmicutes (459) andBacteroidetes (410) and to a lesser extent of Proteobacteria(87) Actinobacteria (20) and Verrucomicrobia (20)consistent with recent reports on microbiota profiles identi-fied in patients with MS or healthy participants on a Westerndiet561718 (figure 2)

Because our sample collections were made at 2 geographicallydistinct sites we first established that there were nocommunity-level differences between the 2 sites (figure e-2linkslwwcomNXIA88) Because there were more womenthan men in our study and the proportion of women differedby therapy we also performed diversity analyses betweensexes Although no alpha diversity differences were detectedwe detected small within-sex and between-sex beta diversitydifferences (p = 001 effect size = 06) that are not notice-able on a principal coordinate analysis plot (figure e-2) Asshown in figure 2 administration of DMF orGA did not resultin overall changes to microbial community structure asmeasured by alpha diversity (richness of taxa within samplesby observed_species or richness and evenness of taxa withinsamples by chao1) or beta diversity (diversity between dif-ferent samples by unweighted UniFrac However LEfSeanalysis identified 13 differentially abundant bacterial generain patients with MS on DMF and 14 differentially abundantgenera in patients on GA as compared to those who weretreatment-naive (table 2) Of interest GA led to a reduction inthe relative abundance of 7 genera and an increase of 7 genera

whereas all the 13 genera affected by DMF treatment showeddecreased relative abundance (figure 3) Among the taxo-nomic units decreased by DMF the most dramatic effect wasmapped to the order Clostridiales (figure 4) thereby sug-gesting potential importance of this taxon

Inference of functional consequences ofalterations in gut microbial compositioninduced by DMF and GAPICRUSt was used to evaluate potential functional implica-tions of these changes in gut microbial profiles16 We notedsignificant changes in 42 pathways in DMF-treated partic-ipants (30 increased and 12 decreased) and in 63 pathways inGA-treated participants (36 increased and 27 decreased)Despite the different route of administration and the differenteffect on bacterial taxa there was an overlap of 17 pathwaysbetween the 2 therapies GA and DMF had a concordanteffect on 15 of those pathways which included pathways af-fecting vitamin amino acid energy and xenobiotic metabo-lism More specifically among the common concordantpathways affected were retinol (vitamin A) methane andethylbenzene metabolism as well as valine leucine and iso-leucine degradation (figure 5)

DiscussionDefining the effects of DMT on gut microbial composition iscritical for the inclusion of treated MS patients in microbiotastudies and interpretation of any findings This has beenestablished in several other disease states for example inseparating effects of metformin from those of the type 2 di-abetes disease statee16 (linkslwwcomNXIA86) Manyother medications from nonndashsteroidal anti-inflammatorydrugse17 to proton pump inhibitorse18 have been shown toaffect the gut microbiota and in this study we demonstratethat MS therapies do as well Although we acknowledge thatdefinitive effects are best established through longitudinalstudies of samples collected before and subsequent to the start

Table 1 Participantsrsquo demographic and MS disease characteristics

Treatment-naive (N = 75) Glatiramer acetate (N = 60) Dimethyl fumarate (N = 33)

Age 381 plusmn 109 457 plusmn 103 425 plusmn 109

Female 47 (627) 46 (767) 18 (515)

BMI 251 plusmn 53 250 plusmn 46 266 plusmn 55

Disease duration 42 plusmn 58 104 plusmn 71 102 plusmn 73

Treatment duration NA 51 plusmn 37 15 plusmn 059

EDSS

Median 20 15 10

Mean 18 plusmn 12 14 plusmn 11 11 plusmn 12

Abbreviation BMI = body mass index

NeurologyorgNN Neurology Neuroimmunology amp Neuroinflammation | Volume 6 Number 1 | January 2019 5

of DMT cross-sectional studies allow for preliminary evalu-ation of these ideas and generation of hypotheses for furtherstudy

The phylum Bacteroidetes was increased in relative abun-dance in our DMF cohort compared with the treatment-naivecohort This difference was mainly driven by an increase in thegenus Bacteroides that narrowly missed statistical significanceThis result is interesting in the context of other studies thathave looked at this commensal group in MS and suggesteda potential protective effect A small study in pediatric MS

showed relative depletion of Bacteroides compared withhealthy controls8 From a mechanistic standpoint oraladministration of purified polysaccharide A produced byBacteroides fragilis was shown to mitigate EAE in an IL-10-dependent fashion19 Lipid 654 a lipodipeptide generated byseveral commensal Bacteroidetes species reaches the sys-temic circulation and is reduced in the serum of patients withMS compared with healthy controls20 In an adoptive transfermodel of EAE low-level Lipid 654 was administered to inducetoll-like receptor 2 tolerance and was found to attenuate EAEand decrease macrophage activation and TH17 cells in the

Figure 2 Overall community differences in the gut microbiota of treated or untreated patients with MS

(A and B) Rarefaction curves plotting alpha diversity indices (A Observed_species B Chao1) vs the number of sequences for each treatment group ns notstatistically significant by the Kruskal-Wallis sum-rank test (C) Principal coordinate plot based on the unweighted UniFrac distance metric The percentage ofvariability explained by the first 3 principal components was plotted Each colored point represents a participant from one of the therapy groups (Naive DMFor GA) (D) No obvious beta diversity differences were observed across groups as demonstrated by similar within-group and between-group distances nsnot statistically significant by a nonndashparametric 2-sample t-test based on 1000 Monte Carlo permutations followed by a Bonferroni correction for multipletesting (E)Mean relative abundance of prevalentmicrobes (gt1 in any therapy group) at the phylum level DMF =dimethyl fumarate GA = glatiramer acetatek = kingdom p = phylum

6 Neurology Neuroimmunology amp Neuroinflammation | Volume 6 Number 1 | January 2019 NeurologyorgNN

Table 2 Differentially abundant taxa

Differentially abundant taxa score

Mean relative abundance

FC log2 (DMFNA) log10LDA p ValueNaive DMF

Phylum

k__Bacteriap__Fusobacteria 30E-04 12E-05 minus46 26 991E-03

k__Bacteriap__Cyanobacteria 18E-03 18E-04 minus33 26 356E-02

k__Bacteriap__Firmicutes 51E-01 38E-01 minus04 47 125E-02

k__Bacteriap__Bacteroidetes 41E-01 50E-01 03 47 458E-02

Class

k__Bacteriap__Fusobacteriac__Fusobacteria 30E-04 12E-05 minus46 27 991E-03

k__Bacteriap__Proteobacteriac__Epsilonproteobacteria 41E-04 76E-05 minus24 28 139E-02

k__Bacteriap__Firmicutesc__Clostridia 48E-01 36E-01 minus04 47 181E-02

k__Bacteriap__Bacteroidetesc__Bacteroidia 41E-01 50E-01 03 47 458E-02

Order

k__Bacteriap__Fusobacteriac__Fusobacteriiao__Fusobacteriales 30E-04 12E-05 minus46 26 991E-03

k__Bacteriap__Actinobacteriac__Actinobacteriao__Actinomycetales

28E-03 29E-04 minus32 34 426E-03

k__Bacteriap__Proteobacteriac__Epsilonproteobacteriao__Campylobacterales

41E-04 76E-05 minus24 29 139E-02

k__Bacteriap__Firmicutesc__Clostridiao__Clostridiales 48E-01 36E-01 minus04 47 181E-02

k__Bacteriap__Bacteroidetesc__Bacteroidiao__Bacteroidales 41E-01 50E-01 03 47 458E-02

Family

k__Bacteriap__Fusobacteriac__Fusobacteriiao__Fusobacterialesf__Fusobacteriaceae

296E-04 121E-05 minus46 26 991E-03

k__Bacteriap__Firmicutesc__Clostridiao__Clostridialesf__[Tissierellaceae]

128E-02 966E-04 minus37 39 407E-03

k__Bacteriap__Actinobacteriac__Actinobacteriao__Actinomycetalesf__Corynebacteriaceae

219E-03 216E-04 minus33 33 307E-02

k__Bacteriap__Actinobacteriac__Actinobacteriao__Actinomycetalesf__Actinomycetaceae

586E-04 747E-05 minus30 31 326E-02

k__Bacteriap__Actinobacteriac__Actinobacteriao__Actinomycetalesf__Micrococcaceae

238E-05 430E-06 minus25 27 691E-04

k__Bacteriap__Proteobacteriac__Epsilonproteobacteriao__Campylobacteralesf__Campylobacteraceae

414E-04 759E-05 minus24 29 139E-02

k__Bacteriap__Firmicutesc__Clostridiao__Clostridialesf__Veillonellaceae

172E-02 977E-03 minus08 34 436E-02

k__Bacteriap__Firmicutesc__Clostridiao__Clostridialesf__Lachnospiraceae

193E-01 154E-01 minus03 43 137E-02

k__Bacteriap__Firmicutesc__Bacillio__Bacillalesf__Planococcaceae

202E-06 0 NA 34 388E-02

Genus

k__Bacteriap__Actinobacteriac__Actinobacteriao__Actinomycetalesf__Actinomycetaceaeg__Varibaculum

346E-04 125E-05 minus48 23 189E-03

k__Bacteriap__Firmicutesc__Clostridiao__Clostridialesf__[Tissierellaceae]g__WAL_1855D

316E-03 117E-04 minus48 33 591E-03

k__Bacteriap__Fusobacteriac__Fusobacteriiao__Fusobacterialesf__Fusobacteriaceaeg__Fusobacterium

296E-04 121E-05 minus46 27 991E-03

Continued

NeurologyorgNN Neurology Neuroimmunology amp Neuroinflammation | Volume 6 Number 1 | January 2019 7

CNS21 Our results suggest that the observed effects of DMFon gut microbiota might be relevant to its anti-inflammatoryeffects in MS

A decrease in the genus Prevotella in patients with MS wasnoted in several small studies however several of these mayhave been confounded by treatment with DMT given theinclusion of a significant number of treated patients (partic-ularly interferon [IFN])67 One study found an increase inPrevotella associated with treatment (pooled group includingIFN and GA) suggesting a possible mechanistic associationfor this genus with DMTs5 In addition human-derived Pre-votella histicola suppresses EAE in a CD4+FoxP3+Tregndashdependent fashion22 In our cohort we noted no change inthe relative abundance of Prevotella in treated participantscompared with those who were treatment-naive suggestingthat this genus is less affected by either GA or DMF than byIFN Another possible explanation for observed differences inthe effects of DMT on Prevotella in our study compared withothers relates to our use of the V4 hypervariable region ascompared to the use of V3ndashV5 in the aforementioned studiesHowever previous comparisons of these variations in thetechnique have demonstrated relatively consistent resultsmaking this difference unlikely to fully account for thisdiscordance23ndash25

We noted the genus Sutterella to be 24-fold less abundant inparticipants treated with GA as compared to those who were

treatment-naive with no difference in relative abundance withDMF An increase in Sutterella in treated compared withuntreated patients withMSwas previously reported howeverthe treatment group included patients treated with either IFNor GA5 Suggesting that IFN may in fact be the driving factoran additional study has noted an increase in Sutterella inIFN-treated MS patients26 In a spontaneous EAE modelspecimens from mice who received transplanted MS fecalmaterial were noted to be relatively depleted in Sutterella27

Given that the EAE course worsened in mice transplantedwith fecal material from patients with MS compared withthose transplanted from a healthy twin this finding might bepathologically significant However in this experiment 3 ofthe 5 MS twins received treatment with IFN potentiallyconfounding interpretation of these results Further work isneeded to understand whether Sutterella is associated withMSpathology whether there are differential effects of DMTs onthe abundance of this genus and finally whether such po-tential effects on Sutterella abundance have any clinicalsignificance

In our study the relative abundance of members of the orderClostridiales (phylum Firmicutes class Clostridia) was de-creased Members of 2 particular Clostridial families Lach-nospiraceae and Veillonellaceae were decreased by both DMFand GA Of interest we previously noted increased relativeabundance of the genus Megasphaera (family Veillonellaceae)in untreated MS compared with healthy controls10 DMF

Table 2 Differentially abundant taxa (continued)

Differentially abundant taxa score

Mean relative abundance

FC log2 (DMFNA) log10LDA p ValueNaive DMF

k__Bacteriap__Firmicutesc__Clostridiao__Clostridialesf__[Tissierellaceae]g__Peptoniphilus

330E-03 207E-04 minus40 33 631E-03

k__Bacteriap__Firmicutesc__Clostridiao__Clostridialesf__[Tissierellaceae]g__Anaerococcus

252E-03 184E-04 minus38 32 187E-02

k__Bacteriap__Actinobacteriac__Actinobacteriao__Actinomycetalesf__Corynebacteriaceaeg__Corynebacterium

219E-03 216E-04 minus33 33 307E-02

k__Bacteriap__Firmicutesc__Clostridiao__Clostridialesf__Veillonellaceaeg__Megasphaera

139E-03 149E-04 minus32 28 693E-05

k__Bacteriap__Firmicutesc__Clostridiao__Clostridialesf__[Tissierellaceae]g__Finegoldia

220E-03 338E-04 minus27 31 428E-02

k__Bacteriap__Actinobacteriac__Actinobacteriao__Actinomycetalesf__Micrococcaceaeg__Rothia

238E-05 430E-06 minus25 29 691E-04

k__Bacteriap__Proteobacteriac__Epsilonproteobacteriao__Campylobacteralesf__Campylobacteraceaeg__Campylobacter

414E-04 759E-05 minus24 29 139E-02

k__Bacteriap__Firmicutesc__Clostridiao__Clostridialesf__Lachnospiraceaeg__Blautia

414E-02 292E-02 minus05 38 436E-03

k__Bacteriap__Firmicutesc__Clostridiao__Clostridialesf__Lachnospiraceaeg__uid_Lachnospiraceae

801E-02 635E-02 minus03 39 818E-03

k__Bacteriap__Firmicutesc__Bacillio__Bacillalesf__Planococcaceaeg__uid_Planococcaceae

202E-06 0 NA 34 388E-02

Abbreviations DMF = dimethyl fumarate EDSS = Expanded Disability Status Scale LDA = linear discriminant analysis NA = not applicable

8 Neurology Neuroimmunology amp Neuroinflammation | Volume 6 Number 1 | January 2019 NeurologyorgNN

seems to ldquoreverserdquo this Clostridial organisms have receivedsignificant attention in demyelinating disorders An increasedfrequency of immunoreactivity to epsilon toxinndashproducingClostridium perfringens type B was observed in patients withMS suggesting this organism as contributory to MS pathol-ogy given the potentially toxic effects at the blood-brainbarrier28 A subsequent study demonstrated the ability ofepsilon toxin to destroy oligodendrocytes and induce de-myelination29 An overabundance of C perfringens has also

been noted in a small study of patients with neuromyelitisoptica30 In addition to the described potential effects ofClostridial members on inflammation Lachnospiraceae wasdemonstrated to impair oligodendrocyte differentiation incultured cells and was associated with impaired myelination inmice31 Clostridial taxa therefore have potential for pathologicrelevance in MS and the mechanism of action of DMTs bothwith respect to inflammation and neurodegenerationneuroprotection In concordance with our results regarding

Figure 3 Cladogram indicating the effect of disease-modifying therapies on the phylogenetic structure of MS patientsrsquomicrobiota

Upper DMF vs Naive Lower GA vsNaive Only significant taxa (p-valuelt 005 LDA score gt2) identified byLEfSe were displayed Differentialabundant taxa are marked by col-ored shades red indicates enrich-ment in the therapy groups (DMF orGA) and green indicates enrich-ment in the naive group Circlesrepresent phylogenetic levels fromkingdom to genus from inside outEach dot represents a taxon and itsdiameter is proportional to the tax-onrsquos effect size c = class DMF = di-methyl fumarate f = family g =genus GA = glatiramer acetate k =kingdom LEfSe = linear discriminantanalysis effect size LDA = linear dis-criminant analysis o = order p =phylum uid = unidentified accord-ing to the GreenGenes database

NeurologyorgNN Neurology Neuroimmunology amp Neuroinflammation | Volume 6 Number 1 | January 2019 9

other Clostridial family members DMF was noted to inhibitthe log-phase growth of C perfringens in vitro and inves-tigators hypothesized that this may be contributory to thedrugrsquos mechanism of action32 A decrease in species belongingto Clostridia clusters was previously found in patients withMS compared with controls however this majority of par-ticipants in this study were treated with DMTs7 A decrease inthe relative abundance of Clostridial members could becommon to several MS DMTs

DMF resulted in a decrease in the relative abundance of manytaxa with Bacteroidetes members as the only significant in-crease The methodology used here to calculate relativeabundance values precludes conclusions about the absolutenumbers of microbes present However fumarates haveknown antimicrobial properties and in fact DMF was pre-viously used as an antimicrobial preservative for upholstery33

Of interest previous work in EAE suggests that a reduction in

gut bacterial load might be of benefit3 Compared with micewho received no antibiotics or intraperitoneal antibioticsthose who received oral antibiotics before EAE inductiondemonstrated reduced EAE severity that correlated to an in-crease in FoxP3+Treg cells in distal peripheral lymph nodesFurther costimulation of harvested lymphocytes from distalnodes resulted in decreased production of proinflammatorycytokines such as IFN gamma and increased production ofanti-inflammatory cytokines such as IL-10 Additional studiesare needed to evaluate this potential effect of DMF as itsgeneralized antimicrobial properties may be a contributingfactor to efficacy in MS

When considering the ultimate influence of DMT on MS gutmicrobiota the effects of the overall shaping of the microbialcommunity on downstream functional pathways may bemoreimportant than effects on individual genera PICRUSt analysismay help identify the impact of changes in microbial

Figure 4 Relative abundances of the order Clostridiales and its family members are differentially changed by therapy

Relative abundances of the order Clos-tridiales (A) and its family members aredifferentially changed by therapies (B Lach-nospiraceae C Veillonellaceae D Tissier-ellaceare) Asterisk () indicates significantchange compared with naive as determinedby the LEfSe analysis (see also in table 2) Theinner bisecting line indicates the median Thelower and upper edges of the box representthe 25th and the 75th percentiles re-spectively LEfSe = linear discriminant analysiseffect size

10 Neurology Neuroimmunology amp Neuroinflammation | Volume 6 Number 1 | January 2019 NeurologyorgNN

composition on biological and metabolic pathways and sug-gest potential mechanisms that link changes in gut microbiotawith therapeutic effects16 One of the common pathways af-fected by both GA and DMF was retinol (vitamin A) me-tabolism Vitamin A metabolites are known to modulatespecific functional aspects of the immune response such asthe Th1ndashTh2-cell balance and the differentiation of Treg andTh17 cells34 Retinoic acid also modulates lymphocyte traf-ficking by enhancing integrin a4b7 expression on lymphocytesand imprinting them with gut tropism35 Of interest dendriticcells from the gut-associated lymphoid tissue but not fromthe spleen were able to produce retinoic acid from retinolsuggesting that vitamin A plays an important role specificallyin the gut-immune system interaction Another bacterialpathway affected by both therapies was methane metabolismAn increased relative abundance of methane-producing bac-teria and exhaled methane in patients with MS compared withhealthy controls was reported5 The role of these pathways inMS pathophysiology and precise effects of therapies will needto be further explored in future studies

Our study is still limited by the relatively small sample size andcross-sectional nature of sample collection including un-avoidable imperfect matching of groups with respect to age

sex and disease duration Also although recruitment at 2separate locations is a merit particularly because careful at-tention was granted to implement identical methodology andanalyses did not demonstrate community-level differences bysite it is possible that unmeasured differences in study groupsrelated to geography confounded our results in some fashion

Our study results confirm our hypothesis that DMTs for MShave measurable effects on gut microbiota Future directionswill include working with a larger cohort with longitudinalfollow-up of individuals who begin new DMTs to assess theeffects more precisely and to evaluate questions regardingthe relationship of gut microbiota to therapy response andtolerability Bidirectional relationships have been confirmedin other disease states36ndash39 and will need to be investigated inMS as well Effects are likely to be particularly important fororally deliveredmedications In addition lessons learned fromthese studies may inform our understanding of the patho-genesis of the disease and lead to the exploration of newpotential treatment targets

Author contributionsConception and design of the study was completed by I KatzSand Y Zhu A Ntranos R Knight BAC Cree SE Baranzini

Figure 5 Functional implications of gut microbial profile changes

Functional implications of the gut microbialprofile changes were assessed by applying thepackage phylogenetic investigation of commu-nities by reconstruction of unobserved stateson whole microbial communities of patientstaking either GA or DMF Forty-two pathwayswere found to be affected in participants trea-ted with DMF (30 upregulated and 12 down-regulated) and 63 pathways in participantstreated with GA (36 upregulated and 27downregulated) Despite the different route ofadministration and the different effect on bac-terial taxa there was an overlap of 17 pathwaysthat were changed by both therapies 15 ofwhich had a concordant effect These includedpathways of vitamins amino acids energy andxenobiotic metabolism and more specificallythemetabolism of retinol (vitamin A) methaneand ethylbenzene as well as valine leucineand isoleucine degradation DMF = dimethylfumarate GA = glatiramer acetate

NeurologyorgNN Neurology Neuroimmunology amp Neuroinflammation | Volume 6 Number 1 | January 2019 11

and P Casaccia Acquisition and analysis of data were com-pleted by I Katz Sand Y Zhu A Ntranos JC ClementeR Brandstadter Y Bencosme E Cekanaviciute S SinghJ Debelius R Knight SE Baranzini and P Casaccia Themanuscript was drafted by I Katz Sand Y Zhu A Ntranosand P Casaccia with critical editing and approval by allremaining authors

AcknowledgmentThe authors thank all of those who contributed to this projectespecially the MS clinicians (Michelle Fabian Aaron MillerFred Lublin Stephen Krieger Sylvia Klineova Aliza Ben-Zacharia Gretchen Mathewson and Jennifer Graves) andresearch coordinators and technicians (Jessica ZhengTamjeed Sikder Cuquita OrsquoShea Refujia Gomez JohnMorrissey and Patrick Barba) at Mount Sinai and UCSFwho recruited the participants as well as our MS patients andtheir families for their participation

Study fundingThe project was funded by the National MS Society theUnited States Department of Defense and the FriedmanBrain Institute

DisclosureI Katz Sand received research support from United StatesDepartment of Defense the National Multiple Sclerosis So-ciety and the Guthy Jackson Charitable Foundation Y Zhuand A Ntranos report no disclosures JC Clemente is anassociate editor for Microbiome and an editor for mBioE Cekanaviciute received research support from NIHIRACDA R Brandstadter reports no disclosures E Crabtree-Hartman received speaker honoraria from Biogen and Gen-zyme Sanofi consulted for Teva Novartis and Biogen andserved on the speakers bureau for Teva and Biogen S Singhreports no disclosures Y Bencosme has been employed byNovo Nordisk J Debelius received research support from theUniversity of California San Diego R Knight served on thescientific advisory boards of Janssen CommenSe and Pro-metheus received travel funding from Metagenics Gen-entech and Zurich Insurance Co served on the editorialboards of Biology Direct Gut Microbes AEM Env Microand Genome Biology is an editor for the ISME Journal holdsa patent for microbiome bases systems apparatus and meth-ods for monitoring and controlling industrial processes andsystems and microbiome bases systems apparatus andmethods for the exploration and production of hydrocarbonsreceives publishing royalties from Simon and Schuster and StMartinrsquos Press is employed by Biota Technology consultedfor CommenSe and Prometheus received research supportfrom Janssen (Johnson amp Johnson) the NIH NIDDK NIHNIAID DOJ ONR USAMRAA NSF NIAIDNIH NIHNIH-NHLBI NIH-NIDDK the Laureate Institute for BrainResearch the Alfred P Sloan Foundation the Gerber Foun-dation CRISP and the Robert Wood Johnson Foundationand received stock or stock options from Biota TechnologyBAC Cree served on the scientific advisory board of Akili

and consulted for AbbVie Biogen EMD Serono GeNeuroNovartis and Sanofi Genzyme SE Baranzini served on thescientific advisory boards of Novartis EMD Serono andSanofi-Aventis received a gift from Novartis Pharma re-ceived speaker honoraria and travel funding from Novartisserved on the editorial boards of theMultiple Sclerosis JournalNeurology and mSystems has a patent pending for gene ex-pression signature that could identify patients at high risk ofdeveloping MS consulted for Novartis EMD Serono andTeva and received research support from NIHNINDSDOD and NMSS P Casaccia received research support fromthe Department of the Army NMSS Full disclosure forminformation provided by the authors is available with the fulltext of this article at NeurologyorgNN

Received June 6 2018 Accepted in final form September 10 2018

References1 Olsson TT Barcellos LF Alfredsson L Interactions between genetic lifestyle and

environmental risk factors for multiple sclerosis Nat Rev Neurol 20171325ndash362 Hedstrom AK Alfredsson L Olsson T Environmental factors and their interactions

with risk genotypes in MS susceptibility Curr Opin Neurol 201629293ndash2983 Ochoa-Reparaz J Mielcarz DW Ditrio LE et al Role of gut commensal microflora in

the development of experimental autoimmune encephalomyelitis J Immunol 20091836041ndash6050

4 Berer K Mues M Koutrolos M et al Commensal microbiota and myelin autoantigencooperate to trigger autoimmune demyelination Nature 2011479538ndash541

5 Jangi S Gandhi R Cox LM et al Alterations of the human gut microbiome inmultiplesclerosis Nat Commun 2016712015

6 Chen JJ Chia N Kalari KR et al Multiple sclerosis patients have a distinct gutmicrobiota compared to healthy controls Sci Rep 2016628484

7 Miyake SS Kim S Suda W et al Dysbiosis in the gut microbiota of patients withmultiple sclerosis with a striking depletion of species belonging to Clostridia XIVaand IV clusters PLoS One 201510e0137429

8 Tremlett HH Fadrosh DW Faruqi AA et al Gut microbiota in early pediatricmultiple sclerosis a case-control study Eur J Neurol 2016231308ndash1321

9 Cantarel BL Waubant E Chehoud C et al Gut microbiota in multiple sclerosispossible influence of immunomodulators J Investig Med 201563729ndash734

10 Cekanaviciute E Yoo BB Runia TF et al Gut bacteria frommultiple sclerosis patientsmodulate human T cells and exacerbate symptoms in mouse models Proc Natl AcadSci USA 201711410713ndash10718

11 Caporaso JG Lauber CL Walters WA et al Ultra-high-throughput microbial commu-nity analysis on the Illumina HiSeq and MiSeq platforms ISME J 201261621ndash1624

12 Kopylova EE Noe L Touzet H SortMeRNA fast and accurate filtering of ribosomalRNAs in metatranscriptomic data Bioinformatics 2012283211ndash3217

13 Caporaso JG Kuczynski J Stombaugh J et al QIIME allows analysis of high-throughput community sequencing data Nat Methods 20107335ndash336

14 Lozupone C Knight R UniFrac a new phylogenetic method for comparing microbialcommunities Appl Environ Microbiol 2005718228ndash8235

15 Segata NN Izard J Waldron L et al Metagenomic biomarker discovery and expla-nation Genome Biol 201112R60

16 Langille MG Zaneveld J Caporaso JG et al Predictive functional profiling of microbialcommunities using 16S rRNAmarker gene sequences Nat Biotechnol 201331814ndash821

17 Schnorr SL Candela M Rampelli S et al Gut microbiome of the Hadza hunter-gatherers Nat Commun 201453654

18 De Filippo C Cavalieri D Di Paola M et al Impact of diet in shaping gut microbiotarevealed by a comparative study in children from Europe and rural Africa Proc NatlAcad Sci USA 201010714691ndash14696

19 Ochoa-Reparaz J Mielcarz DW Wang Y et al A polysaccharide from the humancommensal Bacteroides fragilis protects against CNS demyelinating disease MucosalImmunol 20103487ndash495

20 Farrokhi V Nemati R Nichols FC et al Bacterial lipodipeptide Lipid 654 is a micro-biome-associated biomarker for multiple sclerosis Clin Transl Immunol 20132e8

21 Anstadt EJ Fujiwara M Wasko N Nichols F Clark RB TLR tolerance as a treatmentfor central nervous system autoimmunity J Immunol 19501972110ndash2118

22 Mangalam AA Shahi SK Luckey D et al Human gut-derived commensal bacteriasuppress CNS inflammatory and demyelinating disease Cel Rep 2017201269ndash1277

23 Liu Z Lozupone C Hamady M Bushman FD Knight R Short pyrosequencing readssuffice for accurate microbial community analysis Nucleic Acids Res 200735e120

24 Liu Z DeSantis TZ Andersen GL Knight R Accurate taxonomy assignments from16S rRNA sequences produced by highly parallel pyrosequencers Nucleic Acids Res200836e120

25 Soergel DA Dey N Knight R Brenner SE Selection of primers for optimal taxonomicclassification of environmental 16S rRNA gene sequences ISME J 201261440ndash1444

12 Neurology Neuroimmunology amp Neuroinflammation | Volume 6 Number 1 | January 2019 NeurologyorgNN

26 Castillo Alvarez F Perez Matute P Colina Lizuain S et al Intestinal microbiota inmultiple sclerosis influence of treatment with interferon B-1b Presented at theEuropean Committee for Treatment and Research in Multiple Sclerosis (ECTRIMS)2016 London United Kingdom

27 Berer K Gerdes LA Cekanaviciute E et al Gut microbiota from multiple sclerosispatients enables spontaneous autoimmune encephalomyelitis in mice Proc Natl AcadSci USA 201711410719ndash10724

28 Rumah KR Linden J Fischetti VA Vartanian T Isolation of Clostridium perfringenstype B in an individual at first clinical presentation of multiple sclerosis provides cluesfor environmental triggers of the disease PLoS One 20138e76359

29 Linden JR Ma Y Zhao B et al Clostridium perfringens epsilon toxin causes selectivedeath of mature oligodendrocytes and central nervous system demyelination MBio20156e02513

30 Cree BA Spencer CM Varrin-Doyer M Baranzini SE Zamvil SS Gut microbiomeanalysis in neuromyelitis optica reveals overabundance of Clostridium perfringensAnn Neurol 201680443ndash447

31 Gacias M Gaspari S Santos PM et al Microbiota-driven transcriptional changes inprefrontal cortex override genetic differences in social behavior ELife 20165piie13442

32 Rumah KR Vartanian TK Fischetti VA Oral multiple sclerosis drugs inhibit the invitro growth of epsilon toxin producing gut bacterium Clostridium perfringens FrontCell Infect Microbiol 2017711

33 Schad K Nobbe S French LE Ballmer-Weber B Sofa dermatitis [in German] J DtschDermatol Ges 20108897ndash899

34 Kim CH Regulation of FoxP3 regulatory T cells and Th17 cells by retinoids Clin DevImmunol 200820081ndash12

35 Iwata M Hirakiyama A Eshima Y Kagechika H Kato C Song SY Retinoic acidimprints gut-homing specificity on T cells Immunity 200421527ndash538

36 Wilson ID Nicholson JK Gut microbiome interactions with drug metabolism effi-cacy and toxicity Transl Res 2017179204ndash222

37 Das A Srinivasan M Ghosh TS Mande SS Xenobiotic metabolism and gut micro-biomes PLoS One 201611e0163099

38 Spanogiannopoulos P Bess EN Carmody RN Turnbaugh PJ The microbial phar-macists within us a metagenomic view of xenobiotic metabolism Nat Rev Microbiol201614273ndash287

39 Taguer M Maurice CF The complex interplay of diet xenobiotics and microbialmetabolism in the gut implications for clinical outcomes Clin Pharmacol Ther 201699588ndash599

NeurologyorgNN Neurology Neuroimmunology amp Neuroinflammation | Volume 6 Number 1 | January 2019 13

DOI 101212NXI000000000000051720196 Neurol Neuroimmunol Neuroinflamm

Ilana Katz Sand Yunjiao Zhu Achilles Ntranos et al Disease-modifying therapies alter gut microbial composition in MS

This information is current as of October 26 2018

ServicesUpdated Information amp

httpnnneurologyorgcontent61e517fullhtmlincluding high resolution figures can be found at

References httpnnneurologyorgcontent61e517fullhtmlref-list-1

This article cites 38 articles 6 of which you can access for free at

Citations httpnnneurologyorgcontent61e517fullhtmlotherarticles

This article has been cited by 1 HighWire-hosted articles

Subspecialty Collections

httpnnneurologyorgcgicollectionmultiple_sclerosisMultiple sclerosis

httpnnneurologyorgcgicollectionall_immunologyAll Immunology

httpnnneurologyorgcgicollectionall_demyelinating_disease_cnsAll Demyelinating disease (CNS)following collection(s) This article along with others on similar topics appears in the

Permissions amp Licensing

httpnnneurologyorgmiscaboutxhtmlpermissionsits entirety can be found online atInformation about reproducing this article in parts (figurestables) or in

Reprints

httpnnneurologyorgmiscaddirxhtmlreprintsusInformation about ordering reprints can be found online

Academy of Neurology All rights reserved Online ISSN 2332-7812Copyright copy 2018 The Author(s) Published by Wolters Kluwer Health Inc on behalf of the AmericanPublished since April 2014 it is an open-access online-only continuous publication journal Copyright

is an official journal of the American Academy of NeurologyNeurol Neuroimmunol Neuroinflamm