RESEARCH ARTICLE MULTIPLE SCLEROSIS Identification of Naturally Occurring Fatty Acids of the Myelin Sheath That Resolve Neuroinflammation Peggy P. Ho,1* Jennifer L. Kanter,1,2* Amanda M. Johnson,1,2 Hrishikesh K. Srinagesh,1 Eun-Ju Chang,2,3 Timothy M. Purdy,2,3 Keith van Haren,1,2,3 William R. Wikoff,4 Tobias Kind,4 Mohsen Khademi,5 Laura Y. Matloff,1 Sirisha Narayana,1,2 Eun Mi Hur,1 Tamsin M. Lindstrom,2,3 Zhigang He,6 Oliver Fiehn,4 Tomas Olsson,5 Xianlin Han,7 May H. Han,1 Lawrence Steinman,1*† William H. Robinson2,3*† Lipids constitute 70% of the myelin sheath, and autoantibodies against lipids may contribute to the demyelination that characterizes multiple sclerosis (MS). We used lipid antigen microarrays and lipid mass spectrometry to identify bona fide lipid targets of the autoimmune response in MS brain, and an animal model of MS to explore the role of the identified lipids in autoimmune demyelination. We found that autoantibodies in MS target a phosphate group in phosphatidylserine and oxidized phosphatidylcholine derivatives. Administration of these lipids ameliorated Downloaded from experimental autoimmune encephalomyelitis by suppressing activation and inducing apoptosis of autoreactive T cells, effects mediated by the lipids’ saturated fatty acid side chains. Thus, phospholipids represent a natural anti-inflammatory class of compounds that have potential as therapeutics for MS. INTRODUCTION tory responses in the CNS and that this protective mechanism is com- http://stm.sciencemag.org/ In multiple sclerosis (MS), aberrant adaptive immune responses target promised in MS, because these guardian lipids are attacked by the and destroy the myelin sheath. Although MS is classically considered a adaptive arm of the immune system. These naturally occurring myelin disease driven by T cells, it is now known that autoantibodies also lipids may have therapeutic potential in MS and other inflammatory contribute to its pathogenesis (1, 2). Several studies on MS demon- brain diseases. strate T cell and antibody reactivity to lipids (3–6), which constitute more than 70% of the myelin sheath. Synthesis of anti-lipid antibodies within the central nervous system (CNS) is associated with an aggres- RESULTS sive disease course in MS (7), and, in an experimental model of MS, by guest on March 14, 2019 anti-lipid antibodies both induced demyelination and prevented re- Anti-lipid antibody reactivity differentiates between MS myelination (8). Despite recent interest in the potential pathogenicity patients and controls of antibodies directed against brain lipids, the specificities of the anti- We printed lipid antigen arrays containing more than 50 brain lipids lipid antibody responses in MS remain undefined. and used these arrays to profile autoantibodies in cerebrospinal fluid Here, we report a “functional lipidomics” approach to discovering (CSF) samples derived from MS and control patients. An anti–IgG (im- autoimmune targets and developing therapeutic strategies for MS. We munoglobulin G) + IgM secondary antibody was used to detect anti- used lipid autoantigen microarrays and lipid mass spectrometry to lipid antibody binding. The Significance Analysis of Microarrays (SAM) identify targets of the adaptive autoimmune response in MS patients. (9) algorithm identified 17 lipids that had significantly greater reac- We then explored these results in an animal model of MS, experi- tivity with autoantibodies in CSF from the 33 individuals with MS mental autoimmune encephalomyelitis (EAE), to define the biological [18 with relapsing-remitting MS (RRMS), 14 with secondary progres- role of the autoantibody-targeted lipids in the pathogenesis of auto- sive MS (SPMS), and 1 with primary progressive MS (PPMS)] versus immune demyelination. Unexpectedly, we found that several of the the 26 controls [21 with other (noninflammatory) neurological diseases autoantibody-targeted lipids—phospholipids naturally present in the (ONDs) and 5 healthy controls (HCs)] [false discovery rate (FDR) = brain—could attenuate EAE. Our findings suggest that phosphatidyl- 0.048]; patient demographics and clinical characteristics are listed in serine and oxidized phosphatidylcholine derivatives containing table S1. We used a hierarchical cluster algorithm (10)todiscernrela- saturated fatty acid side chains serve as natural brakes on inflamma- tionships between patient samples and SAM-identified lipids. Most MS samples clustered together on the basis of the similarity of their 1 Department of Neurology and Neurological Sciences, Stanford University School of anti-lipid autoantibody profiles (Fig. 1A). Specifically, the PPMS sam- Medicine, Stanford, CA 94305, USA. 2Division of Immunology and Rheumatology, Department of Medicine, Stanford University School of Medicine, Stanford, CA 94305, ple, and half of the RRMS and SPMS samples, clustered in the group USA. 3Geriatric Research, Education and Clinical Center, Veterans Affairs Palo Alto Health with the highest anti-lipid autoantibody reactivity, whereas only 3 of Care System, Palo Alto, CA 94304, USA. 4Department of Molecular and Cellular Biology the 21 ONDs and none of the HC samples were represented in this and Genome Center, Metabolomics Group, University of California, Davis, Davis, CA 95616, USA. 5Department of Clinical Neuroscience, Karolinska Institute, 171 76 Stock- group. Most of the controls (15 of 21 ONDs and 3 of 5 HCs) clus- holm, Sweden. 6F.M. Kirby Neurobiology Center, Harvard Medical School, Boston, MA tered in the group with the lowest anti-lipid autoantibody reactivity, 02115, USA. 7Diabetes and Obesity Research Center, Sanford-Burnham Medical Re- whereas only 1 SPMS sample and 4 of the 18 RRMS samples clus- search Institute, Orlando, FL 32827, USA. tered in this group. Enzyme-linked immunosorbent assay (ELISA) *These authors contributed equally to this work. †To whom correspondence should be addressed. E-mail: [email protected] (W.H.R.); analysis showed that levels of total IgG were higher than those of total [email protected] (L.S.) IgM in both RRMS and OND CSF (fig. S1, A and B), and, as expected, www.ScienceTranslationalMedicine.org 6 June 2012 Vol 4 Issue 137 137ra73 1 RESEARCH ARTICLE levels of total IgG were significantly high- A er in RRMS and SPMS CSF than in OND CSF (fig. S1C). 1-Palmitoyl-2-glutaroyl-sn- glycero-3-phosphocholine OND 02-100 reduces EAE severity OND 02-099 OND 02-001 OND 02-096 OND 02-095 OND 02-091 RRMS 02-024 RRMS 02-017 OND 01-085 To determine whether the autoantibody- targeted lipids have a role in autoimmune demyelination, we tested the effect of se- lect lipids on EAE, a mouse model of MS. In our initial antigen array experi- ment(Fig.1A),wescreenedlipidsthat OND 02-099 fell into four categories: (1) brain and myelin lipids (for example, cerebrosides, sulfatides, and gangliosides), (2) mem- B brane lipids (for example, cholesterol, 5 Lipid co-immunization Downloaded from phosphatidylcholine, and sphingomyelin), (3) oxidized lipids [for example, 1-palmitoyl- 4 2-glutaroyl-sn-glycero-3-phosphocholine (PGPC) and its derivatives], and (4) micro- bial lipids [for example, lipopolysaccharide 3 (LPS) and lipoteichoic acid]. From our http://stm.sciencemag.org/ set of 17 lipid hits, we selected several lipids 2 from categories 1 to 3 that exhibited higher Vehicle autoantibody reactivity in MS samples than Mean clinical EAE score PGPC in OND or HC samples. We tested the 1 Sphingomyelin effects of these lipids in T cell proliferation 0 01020 30 40 50 60 Fig. 1. Autoantibody targeting of lipids Days after immunization is higher in MS CSF than in OND CSF by guest on March 14, 2019 and normal CSF, and the autoantibody- targeted lipid PGPC attenuates EAE. (A) Lipid array profiling of IgG + IgM anti- C body reactivity in CSF samples from MS 4 Lipid treatment patients, HCs, and OND controls. Lipid hits with the lowest FDR (q = 0.048) were clustered according to their reactivity 3 profiles. Sample type and ID number are shown above the heatmap, and the lipids targeted are shown to the right ’ 2 of the heatmap. The patients demo- Sphingomyelin graphics and clinical characteristics are presented in table S1. (B) Clinical EAE Vehicle PGPC scores of mice coinjected subcutaneously 1 Mean clinical EAE score with PLP139–151 and 6 mg per injection of PGPC or sphingomyelin (day 0); on days 4 and 7 after immunization, the lipids sol- ubilized in 0.05% Tween 20 in PBS (vehicle) 0 0 10 20 30 40 50 were injected intraperitoneally. Arrows indicate when injections were adminis- Days after immunization tered. Each point represents the mean ± SEM, and results are representative of four independent experiments. +P <0.05 by Mann-Whitney test comparing vehicle-treated (n = 5) and PGPC- vehicle alone, for a total of five intravenous injections. Arrows indi- treated (n =5)mice;*P <0.05byMann-Whitneytestcomparing cate when injections were administered. Each point represents the vehicle-treated (n = 5) and sphingomyelin-treated (n = 5) mice. (C) Clin- mean ± SEM. +P < 0.05 by Mann-Whitney test comparing vehicle-treated ical EAE scores of mice administered PGPC, sphingomyelin, or vehicle (n = 10) and PGPC-treated (n =9)mice;*P < 0.05 by Mann-Whitney at the onset of disease. Upon developing clinical signs of EAE, mice test comparing vehicle-treated (n = 10) and sphingomyelin-treated (n = were intravenously administered 100 mg of PGPC or sphingomyelin, or 9) mice. www.ScienceTranslationalMedicine.org 6 June 2012 Vol 4 Issue 137 137ra73 2 RESEARCH ARTICLE assays and in EAE. We found that cerebrosides and gangliosides were ALDO, azPC ester, and azPC). All eight of the lipids identified as tar- unable to suppress myelin basic protein (MBP)Ac1–11–specific T cell gets with the lowest FDR (0.029) had a phosphate group linked to a proliferation, that administration of cerebrosides did not affect EAE, nitrogen moiety through two carbons (Fig. 2B). The nonpolar portion and that oxidized cholesterol had only a minimal effect on EAE (fig. S2). of the targeted lipids contained either one or two side chains; the sec- We previously tested sulfatides, which worsened EAE (3).