Identification of Novel Protein Targets of Dimethyl Fumarate Modification in Neurons and Astrocytes Reveals Actions Independent of Nrf2 Stabilization*

Dimethyl fumarate (DMF) is a reactive fumarate ester used in the treatment of relapsing remitting multiple sclerosis; however, the neuroprotective mechanisms of DMF action are incompletely understood. The results uncover novel DMF-modified cysteine residues in neurons and astrocytes, including cytoskeletal proteins whose modulation by DMF may alter the response to neurodegenerative cues and myelination. Graphical Abstract Highlights Dimethyl fumarate covalently modifies cysteine residues in neurons and astrocytes. Cofilin-1, tubulin and collapsin response mediator protein 2 (CRMP2) are targets. DMF-modified cofilin-1 reduces actin-severing ability, preserving filamentous actin. The fumarate ester dimethyl fumarate (DMF) has been introduced recently as a treatment for relapsing remitting multiple sclerosis (RRMS), a chronic inflammatory condition that results in neuronal demyelination and axonal loss. DMF is known to act by depleting intracellular glutathione and modifying thiols on Keap1 protein, resulting in the stabilization of the transcription factor Nrf2, which in turn induces the expression of antioxidant response element genes. We have previously shown that DMF reacts with a wide range of protein thiols, suggesting that the complete mechanisms of action of DMF are unknown. Here, we investigated other intracellular thiol residues that may also be irreversibly modified by DMF in neurons and astrocytes. Using mass spectrometry, we identified 24 novel proteins that were modified by DMF in neurons and astrocytes, including cofilin-1, tubulin and collapsin response mediator protein 2 (CRMP2). Using an in vitro functional assay, we demonstrated that DMF-modified cofilin-1 loses its activity and generates less monomeric actin, potentially inhibiting its cytoskeletal remodeling activity, which could be beneficial in the modulation of myelination during RRMS. DMF modification of tubulin did not significantly impact axonal lysosomal trafficking. We found that the oxygen consumption rate of N1E-115 neurons and the levels of proteins related to mitochondrial energy production were only slightly affected by the highest doses of DMF, confirming that DMF treatment does not impair cellular respiratory function. In summary, our work provides new insights into the mechanisms supporting the neuroprotective and remyelination benefits associated with DMF treatment in addition to the antioxidant response by Nrf2.


Identification of Novel Protein Targets of Dimethyl Fumarate Modification in Neurons and Astrocytes Reveals Actions
The fumarate ester dimethyl fumarate (DMF) has been introduced recently as a treatment for relapsing remitting multiple sclerosis (RRMS), a chronic inflammatory condition that results in neuronal demyelination and axonal loss. DMF is known to act by depleting intracellular glutathione and modifying thiols on Keap1 protein, resulting in the stabilization of the transcription factor Nrf2, which in turn induces the expression of antioxidant response element genes. We have previously shown that DMF reacts with a wide range of protein thiols, suggesting that the complete mechanisms of action of DMF are unknown. Here, we investigated other intracellular thiol residues that may also be irreversibly modified by DMF in neurons and astrocytes. Using mass spectrometry, we identified 24 novel proteins that were modified by DMF in neurons and astrocytes, including cofilin-1, tubulin and collapsin response mediator protein 2 (CRMP2). Using an in vitro functional assay, we demonstrated that DMF-modified cofilin-1 loses its activity and generates less monomeric actin, potentially inhibiting its cytoskeletal remodeling activity, which could be beneficial in the modulation of myelination during RRMS. DMF modification of tubulin did not significantly impact axonal lysosomal trafficking. We found that the oxygen consumption rate of N1E-115 neurons and the levels of proteins related to mitochondrial energy production were only slightly affected by the highest doses of DMF, confirming that DMF treatment does not impair cellular respiratory function. In summary, our work provides new insights into the mechanisms supporting the neuroprotective and remyelination benefits associated with DMF treatment in addition to the antioxidant response by Nrf2. Fumarate esters have been utilized for the treatment of autoimmune psoriasis for two decades (1). More recently, a dimethyl fumarate (DMF) 1 formulation has been developed for the treatment of relapsing-remitting multiple sclerosis (RRMS), a chronic inflammatory condition resulting in neuronal demyelination and axonal loss (2). Two randomized, double-blind, placebo-controlled trials (DEFINE and CONFIRM) with DMF demonstrated that treatment of RRMS results in sustained clinical and neuroradiological efficacy, and a reduced progression toward disability (3)(4)(5). As a result of these positive outcomes, this DMF formulation (marketed as Tecfidera®) was approved in the US in March 2013. DMF therapy is associated with beneficial immunomodulatory and neuroprotective effects; however, the complete mechanism of action remains unknown.
In the current study we propose that DMF chemically modifies cysteine residues on a range of intracellular protein targets. Early studies demonstrated that fumarate esters mediated anti-inflammatory effects through the modulation of T helper (Th) cells (6,7), inducing a shift from the Th1 profile toward a favorable Th2 profile and the production of IL-4 and IL-5 (8), as well as an increase in type II dendritic cells (DCs) (9). Mechanistically, the immunomodulatory and neuroprotective effects are mediated in part through the cysteine modification of both reduced glutathione and reactive thiols on Kelch-like ECH-associated protein 1 (Keap1) by the fumarate esters (10,11). The depletion of reduced glutathione and modification of Keap1 results in the stabilization of nuclear factor (erythroid-derived 2)-related factor 2 (Nrf2), a transcription factor regulating the cellular response to oxidative stress via the transcription of antioxidant response element (ARE) genes. This leads to an increase in proteins such as heme oxygenase 1 (HO-1), NADPH-quinone-oxidoreductase-1 (NQO1) and glutamate cysteine ligase (GCL), facilitating the replenishment of glutathione and a sustained defense against oxidative stress (12,13). Although Nrf2 activation remains the primary described mechanism of action of DMF (14 -16), recent studies demonstrate that DMF is therapeutically beneficial for the treatment of multiple sclerosis models in Nrf2 knockout mice (17). Upon ingestion DMF may be rapidly converted to monomethyl fumarate (MMF, generated after removal of a methyl group), and MMF agonism of the hydroxycarboxylic acid receptor 2 (HCAR2) also appears to be responsible for some of the positive immunologic effects of DMF therapy (18). A comprehensive quantitative proteomic approach (isotopic tandem orthogonal proteolysis-activitybased protein profiling, isoTOP-ABPP), focused on the immune system has recently identified ϳ40 DMF-sensitive cysteines in primary human T cells, confirming the reactivity of this potent electrophile with immunomodulatory proteins including inhibitor of ␤ kinase (IKK␤), tumor necrosis factor-␣induced protein 3 (TNFAIP3) and IL-16 (19). DMF decreased cell surface levels of the IL-2 receptor to a similar level in Nrf2 ϩ/ϩ and Nrf2 Ϫ/Ϫ mouse splenocytes, again suggesting that the modulation of T cell activation by DMF involves additional protein targets. Blewett et al. demonstrated that the modification of a CXXC motif on PKC by DMF disrupted its interaction with CD28 at the immunological synapse, preventing T cell activation and IL-2 production (19).
We have recently described the increased succination of protein thiols by endogenously produced fumarate in the brainstem of NDUFS4 knockout mice (20), a model of the mitochondrial disease Leigh Syndrome, prompting us to consider further the impact of succination in neurons. The chemical modification of proteins by fumarate yields S-(2-succino-)cysteine (2SC) (21), and is also increased in adipocytes under diabetic conditions (22)(23)(24)(25) as well as fumarase deficient tumors (26,27), where fumarate also leads to the succination of Keap1 thiols (28). We have used several molecular and chemical experimental approaches to increase intracellular fumarate levels, including DMF treatment (29), and we have described the reactivity of tubulin thiols with DMF in vitro (30). When using DMF to rapidly increase protein succination in cells in vitro, we had predicted that the methyl groups of DMF would be hydrolyzed by esterases upon entry into the cell, and that fumarate would react with protein thiols and be detected using a specific anti-2SC antibody (22). However, the lack of visible protein succination upon DMF or MMF treatment led us to consider that the continued presence of the methyl groups was preventing detection of the modified proteins using the anti-2SC antibody. It is therefore significant to note that in vitro DMF treatment does not replicate the modification generated by endogenous fumarate accumulation, such as that observed in the fumarate deficient tumors described above. Consequently, we designed a brief saponification procedure to facilitate immunodetection and observed a wide range of succinated protein bands by immunoblotting (29). In the current investigation we first confirmed increased modification of proteins by DMF using this procedure, followed by a proteomic approach to identify and confirm the site of modification of novel DMF-modified proteins in neurons and astrocytes. We propose that DMF exerts direct effects in neural cell populations that are independent of the activation of the peripheral immune response to modulate the disease progression in multiple sclerosis. This is supported by evidence that DMF and its metabolites accumulate in the brain and induce differential changes in gene expression in regions such as the cortex, cerebellum and hippocampus (17,31). We further propose that the mechanism of action of DMF in neural cell populations is not centered solely upon the activation of the Keap1/Nrf2 antioxidant system; instead other abundant targets of DMF modification may impact a broad range of cellular functions that also contribute to therapeutic efficacy. We performed select functional analyses on several of the novel protein targets we identified to determine if DMF modification altered well-characterized functions of these proteins; in some cases, the reactive thiols identified on these proteins have been studied in the context of neurodegenerative processes. These observations better define the extensive action of DMF in vitro and provide more insight on neuroprotective mechanisms that may be exploited to improve MS treatment through the design of targeted therapeutics.
Primary Neuron Isolation and Culture-All animal use described in this and other sections was consistent with the guidelines issued by the National Institutes of Health and were approved by the University of South Carolina Institutional Animal Care and Use Committee. Primary neurons from newborn rat brain cortices were isolated and cultured using an adaptation of the method described by Brewer (32). Briefly, postnatal day 1 rats were sacrificed by decapitation, the brains were aseptically dissected and cortices were separated from the rest of the brain in ice-cold Hibernate A medium (Cat # A1247501, Gibco/Thermo Fisher Scientific, Waltham, MA), containing 2% (v/v) B-27 supplement (Cat # 17504044, Gibco/Thermo Fisher Scientific) and 0.5 mM glutamine (Cat # 25030149, Gibco/Thermo Fisher Scientific). The tissue was minced in fragments of about 1 mm 3 with a scalpel and subjected to digestion with 2 mg/ml papain (Cat # LS003120, Worthington Biochemical Corp., Lakewood, NJ) in the supplemented Hibernate A medium for 20 min at 30°C in a shaker incubator set at 100 rpm. After thorough trituration through a fire polished Pasteur pipette, the tissue was allowed to settle for 5 min and the supernatant was carefully layered on top of a discontinuous OptiPrep (Cat # D-1556, Sigma-Aldrich) gradient prepared in Hibernate A medium; the layers contained 35, 25, 20 and 15% OptiPrep. The gradient was centrifuged at 800 g for 15 min at room temperature, and layers 1 (15% OptiPrep) and 2 (20%) were discarded. Layer 3 (25%) was collected and added a 5-fold volume of Neurobasal A medium (Cat # 10888022, Gibco/Thermo Fisher Scientific), containing 0.5 mM glutamine and 2% B-27 supplement. After a centrifugation at 500 ϫ g for 5 min, cells were resuspended in Neurobasal A medium containing 0.5 mM glutamine, 2% B-27 supplement, and 5 ng/ml bFGF (Cat # 13256029, Invitrogen/Thermo Fisher Scientific); counted and plated on 24-well plates pretreated with 0.01% poly-L-lysine (Cat # P4707, Sigma-Aldrich) at a density of 200,000 cells/well. Fifty percent of the media was replaced every third day, with the addition of 5 M AraC (Cat # BP2512100, Fisher Scientific, Thermo Fisher Scientific, Waltham, MA) from DIV 3 to inhibit glial proliferation. On DIV 8, cells were left untreated or treated for 24 h with 10 M or 100 M dimethylfumarate (DMF, Cat # 242926, Sigma-Aldrich) prepared in Dulbecco's PBS (DPBS, Cat # 21316003, Corning Cellgro, Manassas, VA) and filtered. On DIV 9, medium was removed, cells were rinsed 3 times with DPBS and collected after the addition of 250 l radioimmunoprecipitation assay (RIPA) lysis buffer [50 mM Tris-HCl (Cat # BP152-5, Fisher Scientific, Thermo Fisher Scientific), 150 mM NaCl (Cat # S7653, Sigma-Aldrich), 1 mM EDTA (Cat # ED2SS, Sigma-Aldrich), 0.1% Triton X-100 (Cat # T9284, Sigma-Aldrich), 0.1% SDS (Cat # BP166 -5, Fisher Scientific, Thermo Fisher Scientific), 0.5% sodium deoxycholate (Cat # D6750, Sigma-Aldrich), pH 7.4], with the addition of 2 mM diethylenetriaminepentaacetic acid (Cat # D6518, Sigma-Aldrich), and a protease inhibitor mixture (Cat # P8340, Sigma-Aldrich). Homogenization was performed by pulse sonication at 2 watts using a Model 100 sonic dismembrator (Thermo Fisher Scientific) for 30 s before resting on ice for 30 min in lysis buffer as described before (30). Protein in the lysates was precipitated with 9 volumes of cold acetone for 10 min on ice. After centrifugation at 3000 ϫ g for 10 min and removal of the acetone, the protein pellet was resuspended in 150 ìl RIPA buffer. The protein content in the lysates was determined by the Lowry assay (33).
N1E-115 Cell Culture-N1E-115 cells (subclone N1E-115-1 neuroblastoma cells) were obtained from Sigma-Aldrich (Cat # 08062511). The cells were expanded in nondifferentiation medium (NDM): 90% DMEM (Cat # 12430 -054, Gibco/Thermo Fisher Scientific, containing with 25 mM glucose, no pyruvate, 25 mM HEPES, 4 mM glutamine) and 10% FBS. At 80% confluence, the cells were differentiated into neurons in the presence of 2% FBS and 1.25% dimethyl sulfoxide (DMSO, Cat # 32434, Alfa Aesar, Thermo Fisher Scientific) in DMEM for 5 days (34). In addition to the assessment of the neuronal phenotype by light microscopy, the detection of synaptophysin protein levels confirmed successful differentiation. During the final 24 h of differentiation the cells were treated with 10 -100 M DMF and protein was collected as described for primary neurons.
Lentiviral Transduction of 3T3-L1 Fibroblasts-The lentiviral vectors were prepared by the University of South Carolina Viral Vector Facility. Briefly, TRC2 Fh1 shRNA, clone-TRCN0000246831 or MIS-SION TRC2 pLKO.5-puro nonmammalian shRNA control plasmids (Cat # SHC202, Sigma-Aldrich) were used to generate the lentiviral vectors that also contained a puromycin resistance gene. Fifteen microgram vector plasmid, 10 g psPAX2 packaging plasmid (Cat # 12260, Addgene, Cambridge, MA), 5 g pMD2.G envelope plasmid (Cat # 12259, Addgene) and 2.5 g pRSV-Rev plasmid (Cat # 12253, Addgene) were transfected into 293T cells. The filtered conditioned medium was collected and stored at Ϫ80°C. 3T3-L1 fibroblasts (Cat # CL-173 ™ , ATCC, Manassas, VA) were incubated overnight with 150 l of filtered conditioned medium containing Fh1 shRNA or control lentivirus. Successfully transduced fibroblasts were selected using 1 g/ml puromycin (Cat # P9620, Sigma-Aldrich). The selected fibroblasts were propagated in the presence of puromycin until confluent and harvested in RIPA buffer as described above. Successful knockdown of fumarase expression was determined by immunoblotting and fumarate levels were determined by GC-MS as described previously (20).
Saponification of Fumarate Esters-Sixty micrograms of protein from control and DMF treated cell lysates was incubated with 80% DMSO, 6 mM potassium hydroxide (KOH, Cat # 484016, Sigma-Aldrich), and 1 mM EDTA at room temperature for 30 min, with vortexing at 5 min intervals. The pH was adjusted to 7, and the protein was precipitated with 90% acetone (Cat # BDH1101-4LG, VWR, Radnor, PA) before being resuspended in 40 l RIPA buffer.
The pH was again adjusted to 7 before gel electrophoresis and immunoblotting.
One-dimensional PAGE and Western Blotting-Western blotting was performed as described previously, after separation of the proteins by SDS-PAGE (22,23). For protein identification purposes gels were stained with Coomassie Brilliant Blue R (Cat # 27816, Sigma-Aldrich) following electrophoresis to allow band isolation and mass spectrometry (see below). In some cases, membranes were stripped with 62.5 mM Tris, pH 6.8, containing 2% SDS and 0.7% 2-mercapto ethanol (Cat # M6250, Sigma-Aldrich) for 20 min at 65°C before reprobing with a different antibody.
Protein Identification from SDS-PAGE Gel Bands by LC-MS/ MS-To identify the sites of fumarate ester modification, 60 g of protein from primary neurons, 200 g of protein from primary astrocytes or 120 g of protein from differentiated N1E-115 neurons were resolved by SDS-PAGE, and the gels were stained with Coomassie Brilliant Blue R. The visible protein bands were excised from the gels and subjected to in-gel digestion with trypsin (the gel was cut into segments based on the intensity of the Coomassie stain and the protein bands were excised). Briefly, after destaining the proteins were reduced with 10 mM dithiothreitol (Cat # V3155, Promega, Madison, WI) and alkylated with 170 mM 4-vinylpyridine (Cat # V3204, Sigma-Aldrich). Protein digestion was carried out overnight at 37°C in the presence of 500 ng sequencing grade modified trypsin (which cleaves C-terminal to lysine and arginine residues, Cat # V5280, Promega) in 50 mM ammonium bicarbonate (Cat # 09830, Sigma-Aldrich). After gel extraction, the peptides were analyzed in a blinded manner on a Dionex Ultimate 3000-LC system (Thermo Scientific) coupled to a Velos Pro Orbitrap mass spectrometer (Thermo Scientific). The LC solvents were 2% acetonitrile (Cat # 34851, Sigma-Aldrich)/0.1% formic acid (Cat # 85178, Pierce, Thermo Scientific) (Solvent A) and 80% acetonitrile/0.1% formic acid (Solvent B); the water used for these solvents was LC-MS grade (Cat # 39253, Honeywell International Inc., Morris Plains, NJ). Peptides were first trapped on a 2 cm Acclaim PepMap-100 column (Thermo Scientific) with Solvent A at 3 l/min. At 4 min the trap column was placed in line with the analytical column, a 75 m C18 stationary-phase LC Pico-Chip Nanospray column (New Objective, Inc., Woburn, MA). The peptides were eluted with a gradient from 98%A:2%B to 40%A: 60%B over 30 min, followed by a 5 min ramp to 10%A:90%B that was held for 10 min. The Orbitrap was operated in data-dependent acquisition MS/MS analysis mode and excluded all ions below 200 counts. Following a survey scan (MS1, up to 8 precursor ions were selected for MS/MS analysis. All spectra were obtained in the Orbitrap at 7500 resolution. The DDA data were analyzed using Proteome Discover 1.4 software with SEQUEST algorithm against the uniprot_ref_mouse database (2014 -10 Ϫ03 version, 52,474 proteins) or uniprot_ref_rat database (2011-5-11 version, 39765 proteins) with XCorr validation Ͼ2 (ϩ2) or Ͼ2.5 (ϩ3). An allowance was made for 2 missed cleavages following trypsin digestion. No fixed modifications were considered. The variable modifications of methionine oxidation (M OX ), proline hydroxylation (P OX ), cysteine pyridylethylation (C PE , 105.058) or cysteine succination by fumarate (C 2SC , 116.011), cysteine modification by monomethyl fumarate (C MMF , 130.026) or cysteine modification by dimethyl fumarate (C DMF , 144.042) were considered with a mass tolerance of 15 ppm for precursor ions and a mass tolerance of 10 ppm for fragment ions. The results were filtered with a false discovery rate of 0.01 for both proteins and peptides (Percolator node). A minimum of 8 unique peptides was reported for all proteins identified (maximum 106 peptides). For all identifications the spectra were manually inspected to confirm identity of the proposed DMF-modified Cys containing peptides. Any low-quality spectra/incorrect identifications were discarded before performing MS/MS analyses on select masses of interest from the high quality protein identifications.
Select MS/MS Analyses-The samples were reanalyzed to target the expected DMF-modified Cys containing peptide mass from the data obtained in DDA mode, further MS/MS analyses were used to monitor select modified peptide masses of interest. The spectrometer repeatedly acquired an MS1 spectrum followed by the desired MS/MS spectra (CID fragmentation in the ion trap, all spectra in the Orbitrap at 7500 resolution). These MS/MS spectra were then averaged over the entire LC peak yielding a spectrum superior to that obtained by DDA (where the acquisition time was split looking for many peptides at once). The CID-MS/MS data was inspected in Proteome Discover 1.4 software for the masses of interest with either cysteine modification by monomethyl fumarate (C MMF , 130.026), or cysteine modification by dimethyl fumarate (C DMF , 144.042) being considered based on the expected modification from the DDA run. The variable modifications of methionine oxidation (M OX ), proline hydroxylation (P OX ) or cysteine pyridylethylation (C PE , 105.058) were considered. The MS/MS spectra were inspected again using Thermo Xcalibur 2.2 software. Manual sequencing of the spectra was used to confirm the sequence and modification site of the peptides. The mass spectrometry analyses constitute a Tier 3 measurement and the proteomics data has been deposited to the ProteomeXchange Consortium via the PRIDE (35) partner repository with the data set identifier PXD008314. The file names in the PRIDE data reflect the protein identified in the original DDA run from a gel band.
DRG Culture and Fluorescence Time-Lapse Microscopy-Dorsal Root Ganglion (DRG) neurons were isolated from rat sciatic nerve as described previously (36). 48 h after sciatic nerve crush to induce a regenerative response, adult rat DRG neurons were plated for 16 -24 h onto poly-D-lysine (Cat # P6407, Sigma-Aldrich) and laminin (Cat # 11243217001, Sigma-Aldrich)-coated German glass coverslips and maintained in Ham's F12 medium (Cat # 11765-047, Gibco, Thermo Fisher Scientific) supplemented with 10% horse serum (Cat # 16050 -122, Gibco, Thermo Fisher Scientific). Neurons were exposed to 500 nM Lysotracker Red (Cat # L7528, Invitrogen, Thermo Fisher Scientific) for 5 min before imaging. Coverslips were transferred into fresh medium containing 25 mM HEPES (Cat # 15630080, Gibco, Thermo Fisher Scientific), pH 7.4 and 1% OxyFluor (Cat # OF-0005, Oxyrase, Inc., Mansfield, OH) in a water-heated microscope stage warmed to 37°C. Cells expressing a relatively low level of Lysotracker Red were selected for time-lapse imaging using an Axiovert 200 inverted microscope (Carl Zeiss, Inc.). Fluorescent images were acquired every 0.3 s for 2 min using a Plan-Apo 63ϫ/1.2 W/0.17 water objective. Kymographs were generated from time-lapse movies using NIH Im-ageJ software. The kymographs were generated such that the direction toward the cell body was always to the right, so lines that sloped toward the right at any point with a net displacement of Ͼ5 m were categorized as retrograde organelles. Lines that sloped toward the left Ͼ5 m at any time during the recording interval were considered anterograde organelles. Lines that zigzagged were categorized as bidirectional, and lines that showed Ͻ5 m lateral displacements in any direction during the recording interval were categorized as static. A total of fifteen axons per experimental group in three independent experiments were used for quantification purposes, and the number of organelles for each category of movement expressed as % of total organelles was used to construct a box and whiskers graph (Fig. 4B). The data was further analyzed by calculating the speed (in m/s) and run length (in m) for the anterograde and retrograde motile events; results were expressed as mean Ϯ S.E. (Fig. 4C). Finally, the speed and run length for anterograde and retrograde motile events were grouped in 4 different intervals, transformed in % and expressed as mean Ϯ S.E. (Fig. 4D).
Cofilin Activity Assay-Cofilin activity was determined by measuring its ability to sever actin filaments, as described by Yonezawa et al. (37). Briefly, aliquots of 7.5 g human recombinant cofilin (containing Tris, NaCl, sucrose and dextran, Cat # CF01, Cytoskeleton Inc., Denver, CO) were dissolved in water (final concentrations:10 mM Tris-HCl, pH 8.0, 10 mM NaCl, 5% sucrose, 1% dextran), added sodium diethylenetriaminepentaacetate (Cat # D6518, Sigma-Aldrich) to 100 M, and Tris(2-carboxyethyl)phosphine hydrochloride (Cat # 20490, Thermo Fisher Scientific) to 250 M, and incubated overnight at room temperature with a 5-fold molar excess of dimethylfumarate (DMF), or with vehicle. The following day, aliquots of 15 g rabbit muscle actin (Cat # AKL99, Cytoskeleton Inc.) dissolved in ice-cold General Actin Buffer (5 mM Tris-HCl pH 8.0, 0.2 mM CaCl 2 , Cat # BSA01001, Cytoskeleton Inc.) supplemented with 0.2 mM ATP (Cat # BSA04, Cytoskeleton Inc.), were polymerized into F-actin by addition of 1/10 volume of Actin Polymerization Buffer (100 mM Tris-HCl pH 7.5 containing 20 mM MgCl 2 , 500 mM KCl, 10 mM ATP and 50 mM guanidine carbonate, Cat # BSA02, Cytoskeleton Inc.) followed by 1 h incubation at room temperature. At this point, the pH of the F-actin aliquots was adjusted to 8.0 with an excess of 10 mM Tris-HCl before the addition of the cofilin aliquots that were incubated overnight with or without DMF, and the incubation continued for another 30 min at room temperature. The samples were then centrifuged at 150,000 ϫ g for 1 h at 25°C, with deceleration set without brakes. After separation of the supernatants (containing depolymerized actin monomers) from the pellets (containing F-actin), fractions of supernatants and pellets containing the equivalent to 3 g actin and 1.5 g cofilin in the initial mix were resolved by SDS-PAGE, and gels were stained with Coomassie Brilliant Blue R (n ϭ 4 replicates per experimental group). Images of the gels were captured using a Gel Doc XRϩ scanner (Bio-Rad Laboratories, Inc.), and the integrated OD x area value of the bands was analyzed with the Image Lab Software V5.2.1 (Bio-Rad Laboratories, Inc.). For each particular protein (actin and cofilin), results of the band integrated OD in supernatants and pellets were expressed as % of the total integrated OD (supernatant band ϩ pellet band), grouped by treatment, and then reported as the mean Ϯ S.E. Additional control experiments included F-actin samples that did not have added cofilin, or F-actin samples that were adjusted to pH 6.8 (to inhibit cofilin severing activity on F-actin) before cofilin addition.
Ubiquitin C-Terminal hydrolase L1 (Uchl1) Activity Assay-UchL1 deubiquitinase activity was assessed using recombinant UchL1 protein (Cat # 50690-M07E, Sino Biological, Beijing, China) and the substrate Ubiquitin-Rhodamine110-Glycine (Cat # U-555, Boston-Biochem, Cambridge, MA). UchL1 protein (0.1 g) was preincubated with 0, 25, 50, or 100 M DMF for 24 h prior the addition of 0.1 M fluorogenic ubiquitin rhodamine in a deubiquitination buffer (50 mM Tris-HCl, 150 mM NaCl, 5 mM DTT). Fluorescence was measured at 535 nm (following excitation at 485 nm) over a 2 h period using the Tecan SAFIRE Spectrofluorimeter. The initial reaction velocities were analyzed and the results were averaged for each experimental group and expressed as the mean Ϯ S.E. n ϭ 5 per condition analyzed).
Measurement of Oxygen Consumption Rate (OCR)-N1E-115 cells were seeded on V7 cell culture microplates coated with 0.2% gelatin (Cat # G1890, Sigma-Aldrich) at a density of 10,000 cells/well. After 3 days in culture, the cells were differentiated for 5 days as described above, and treated with 0, 10, or 50 M DMF for the last 24 h (n ϭ 6/group). The Seahorse extracellular flux analyzer XF-24 (Agilent Technologies, Inc., Santa Clara, CA), was used to measure the oxygen consumption rate (OCR), using XF Assay Medium (Cat # 102365-100, Agilent Technologies, Inc.) supplemented with 25 mM glucose (Cat # G8270, Sigma-Aldrich) (38). After measurement of basal respiration, oligomycin (5 g/ml, Cat # O4876, Sigma-Aldrich), carbonyl cyanide 4-(trifluoromethoxy)phenylhydrazone (FCCP, 0.5 M, Cat # C2920, Sigma-Aldrich), and rotenone (3 M, Cat # R8875, Sigma-Aldrich) ϩ antimycin A (4 M, Cat # A8674, Sigma-Aldrich) were added sequentially to determine ATP production/proton leak, spare respiratory capacity and nonmitochondrial respiration. After completion of the assay, the medium was removed and the cells were immediately washed 3 times with cold PBS. The plate was stored at Ϫ70°C before the measurement of the total protein content to normalize individual measurements. The results were averaged for each experimental group and expressed as the mean Ϯ S.E.
Experimental Design and Rationale-The purpose of these experiments was to identify novel targets of DMF mediated electrophilic modification of cysteine residues in neurons and astrocytes in vitro, in an attempt to better understand the biological targets and mechanism of action of this drug. The detection of fumarate ester derived succination will only be present in the DMF treated cells. For experiments using primary rodent cultures, at least 3-6 rodent pups were used to generate cells for primary culture. A minimum of three replicate treatment plates/wells (untreated controls versus DMF treated) were used and each separate experiment was reproduced with new cells onefour times. In some cases, immunoblotting for increased heme oxygenase Ϫ1 was used to confirm a biological effect of DMF treatment on the cells. Protein from individual cell culture replicates were separated by electrophoresis and peptides extracted from parallel excised gel bands were pooled in order to obtain enough material for confirmation of the site of modification. At least three technical repeats (LC-MS/MS analyses) were conducted in order to confirm the protein identification and site of modification. All other analyses (including respiration analyses and monitoring of lysosomal trafficking) were performed with a minimum of 3 independent biological replicates per group, with some up to 8 replicates per group (n ϭ 3-8). The Uchl1 deubiquitinase activity was studied with n ϭ 5 per group. Data are summarized throughout as mean Ϯ S.E. and are plotted using SigmaPlot 11 software (Systat Software, Inc. San Jose, CA) and Prism 4 (GraphPad Software, La Jolla, CA). Statistical analyses were performed using SigmaPlot 11 and Prism 4. Differences between more than two groups were analyzed using one-way ANOVA with either the Holm-Sidak or Tukey's post-test. When two groups were compared, the unpaired Student "t" test was used. In all cases, p Ͻ 0.05 was considered statistically significant.

DMF-Induced Protein Modification in Neural Cells-Al-
though the immunomodulatory and neuronal benefits of DMF are often attributed to the succination of Keap1 and activation of the antioxidant response element (ARE), we predicted that a wider range of protein thiols might also be modified by DMF or its primary metabolite monomethyl fumarate (MMF), because we had observed that this occurs in adipocytes in vitro (29). In order to detect succinated proteins in neurons treated with DMF for 24 h, we employed a procedure that we had developed previously using alkaline hydrolysis to remove the ester and permit the immunological detection of protein succination (supplemental Fig. S1B) (29). In the absence of ester hydrolysis there was limited detection of succinated proteins in rat primary neurons using the anti-2SC antibody, with only one band ϳ50 kDa showing a significant increase in intensity following 100 M DMF treatment (Fig. 1A, lanes 7-9). This indicated that at least one or both of the methyl groups had not been removed by intracellular esterases and was preventing interaction with the anti-2SC antibody that recognizes the S-(2-succino)cysteine epitope (22). To ensure that the succinated proteins detected were solely of neuronal origin we differentiated N1E-115 neuroblastoma cells to a neuronal phenotype (confirmed by increased synaptophysin protein content) (20), and treated these with DMF for 24 h. The hydrolysis of the ester in the presence of KOH facilitated the detection of a large number of succinated proteins (Fig. 1B,  last panel, lanes 5 and 6), similar to what we have observed previously in adipocytes following fumarate ester treatment (29). Fibroblasts that had fumarase knocked down using a lentiviral mediated shRNA approach to increase endogenous fumarate levels (39) were used as a positive control for succination (Fig. 1B, lane 2). The intensity of succinated proteins in the DMF treated neurons versus the positive control indicated that fumarate esters readily enter cells and react with a wide range of protein thiols. Because DMF is known to modify thiols in Keap1 leading to the induction of ARE derived proteins such as heme oxygenase 1 (HO-1), we next confirmed that DMF treatment significantly induced HO-1 protein levels (Fig. 1C).
In order to identify the novel protein targets in DMF treated neurons we used LC-MS/MS to analyze distinct gel bands representing electrophoretically separated proteins. The peptides obtained were analyzed to determine if modification of protein thiols by either DMF (C DMF ), or the de-methylated metabolites MMF (C MMF ) and fumarate (C 2SC ) had occurred. This proteomic approach allows targeted confirmation of the exact sites of modification and it does not require the base hydrolysis of the ester that can also result in a partial loss of protein. As noted in Table I, all of the proteins identified were modified by DMF or MMF, rather than 2SC, suggesting that both of these fumarate esters reacted with intracellular proteins more rapidly than they could be converted to fumarate. In primary rat neurons we confirmed the identity of 4 modified proteins, and in N1E-115 neurons we confirmed a total of 15 protein subunits modified by either DMF or MMF. Because the N1E-115 cultures were devoid of any glial cells that may be present in primary neuronal cultures, the identified proteins reflect true neuronal targets of fumarate ester modification. Because fumarate esters were previously shown to induce Nrf2 in glial cells in vivo (13), we also investigated additional protein targets of DMF-mediated protein modification in primary astrocyte cultures (supplemental Fig. S2A confirms the enrichment of Glial Fibrillary Acid Protein (GFAP) positive cells). Table I confirms the detection of 11 modified protein subunits in primary rat astrocytes after a 24 h treatment with up to 100 M DMF. Overall, this targeted approach confirmed the novel identification of 24 distinct protein subunits in both neurons and astrocytes that are directly modified by either DMF or MMF.
DMF Modification Affects Cofilin-1 Functionality-The chemical modification of several cytoskeletal proteins was observed in all cell types examined (Table I). In astrocytes this included abundant cytoskeletal proteins such as Glial Fibrillary Acid Protein (GFAP) and vimentin, as well as cofilin-1, a dynamic regulator of actin polymerization. Cofilin-1 modulates the actin cytoskeleton by depolymerizing filamentous (F) actin, generating monomeric actin that can be used to reorganize the actin cytoskeleton in response to cellular dynamics. To confirm the initial observation that Cys139 of Cofilin-1 is succinated by MMF in astrocytes in vitro, we performed additional selected reaction monitoring and detected the ϩ3   Fig. 2A). Because 100% modification of the peptide by fumarate esters was not expected, we also detected the pyridylethylated version of the same peptide (supplemental Fig. S2B, ϩ3 charge state, 613.6218 m/z). Importantly, this cysteine residue is conserved across rat, mouse and human sequences. To determine if the succination of Cys139 impacted its activity as a regulator of actin dynamics, we examined the pH-sensitive depolymerization activity of cofilin-1 on F-actin. Cofilin binds to F-actin in a 1:1 ratio below neutral pH, but does not sever the actin filaments (37). In contrast, at pH 8.0 the presence of cofilin results in actin severing and the production of monomeric actin. Fig. 2B confirms that F-actin remains polymerized in the absence of cofilin, and in the presence of cofilin at pH 6.8 (F-actin is detected in the pellet fraction (P)). The adjustment of the pH to 8.0 facilitates F-actin depolymerization and the

FIG. 2. Modification of cofilin impairs the depolymerization of actin filaments.
A, MS/MS spectrum from astrocyte protein extracts after DMF treatment showing the MMF succinated Cys139 of cofilin 1 in the peptide HELQANC MMF YEEVKD. B, Confirmation that the cofilin severing effect on F-actin is pH-dependent. Samples of cofilin (7.5 g, lanes 3-6) were incubated with F-actin (15 g) in a buffer at pH 6.8 (lanes 3-4) or 8.0 (lanes 5-6); tubes containing actin only at pH 8.0 (lanes 1-2) were included as negative controls. After centrifugation, SDS/PAGE separation and Coomassie blue staining, the distribution of actin and cofilin was studied in the pellet after centrifugation (P) and the supernatant (S). Only the mix incubated at pH 8.0 showed cofilin severing activity on F-actin (lanes 5-6, note that the distribution of actin and cofilin is similar in P and S fractions); cofilin was inactive at pH 6.8 (lanes 3-4). C, Succination by DMF reduces cofilin severing effect on F-actin. Samples of cofilin (7.5 g) were incubated with vehicle (0, lanes [1][2][3][4][5][6][7][8]  detection of monomeric actin in the supernatant (S). Cofilin preincubated with DMF in vitro (500 M, equivalent to DMF: cofilin cysteines, 5:4) was unable to fully sever F-actin at pH 8.0 (Fig. 2C, 2D and 2E), resulting in a 29.3% decrease in actin severing activity). Supporting this, there was a greater proportion of actin (DMF: 53.49 Ϯ 0.60% versus CONT: 34.20 Ϯ 2.12%, n ϭ 4, p Ͻ 0.01) and cofilin (DMF: 52.94 Ϯ 0.84% versus CONT: 27.23 Ϯ 3.25%, n ϭ 4, p Ͻ 0.01) present in the pellet fraction (P) when the DMF modified cofilin was used (Fig. 2C, 2D and 2E). The specific modification of Cys139 in these in vitro cofilin preparations was also confirmed separately following CID (data available via PRIDE repository). These data suggest that the chemical modification of Cys139 on cofilin decreases its ability to dynamically regulate actin depolymerization.
Modification of Tubulin by DMF Does Not Alter Axonal Trafficking of Lysosomes-Tubulin was also identified as another cytoskeletal target of DMF reactivity (Table I). Tubulin ␣ and ␤ subunits form dimers that polymerize into microtubules, and we have previously described that in vitro treatment of purified porcine tubulin with DMF results in the succination of 11 of the 20 cysteines in the ␣␤ tubulin dimer (30). In the current study we confirmed the DMF mediated succination of tubulin ␣-1A peptide TIQFVDWC DMF PTGFK (ϩ2 charge state, 843.3976 m/z) in N1E-115 neurons; the site of succination was identified as Cys347 as designated by the prominent y 6 and b 8 fragment ions in the annotated spectrum (Fig. 3A). In addition, the pyridylethylated Cys347 containing peptide was also detected in the same N1E-115 neuronal protein preparation (ϩ2 charge state, 823.9054 m/z, supplemental Fig. S3). We have previously described that increased succination by fumarate can affect the detectability of tubulin by antibodies directed against a cysteine containing antigen (30), as succination alters the epitope size and conveys two novel carboxylate groups. We have observed this in adipocytes cultured in high glucose (30 mM) versus normal glucose (5 mM), and on succinated porcine tubulin prepared in vitro (30). We examined if succination of purified porcine tubulin in vitro by DMF (versus fumarate) would also interfere with tubulin detection. Porcine tubulin was incubated with either 100 or 500 M DMF, 500 mM fumarate, or without any addition for up to 24 h. Fig.  3B demonstrates that tubulin succination is pronounced in fumarate treated samples after a short film exposure (2SC short panel), whereas it is undetectable for the tubulin sam- FIG. 3. Tubulin as a target of DMF succination. A, MS/MS spectrum showing the DMF succinated Cys347 of ␣-tubulin in the peptide TIQFVDWC DMF PT-GFK from N1E-115 neuron protein digests. B, Tubulin modification by DMF does not affect detectability by tubulin antibodies. 1 g purified porcine brain tubulin was incubated for 6 or 24 h with 0 (control) or 100 or 500 M DMF, or for 24 h with 50 mM fumarate (Fum) before SDS/PAGE separation and immunoblotting. If no saponification is performed, the anti-2SC antibody only detects succination by fumarate but not by DMF, even after a long exposure (2SC short and long panels; compare with Fig. 1). The ␣-tubulin B.7 antibody shows decreased detectability of tubulin after succination by fumarate, but no change when tubulin is modified by DMF, whereas the DM1A antibody is insensitive to succination (␣-tubulin B.7 and DM1A panels, respectively). Coomassie staining of a duplicate gel run in parallel was used to verify even loading of the lanes. ples treated with DMF or left untreated. A longer exposure (Fig. 3B, 2SC long panel) shows some basal tubulin succination in untreated samples and a very strong signal in fumarate treated samples. However, we found no increases in modification of DMF treated tubulin compared with the untreated samples, which confirms the inability of the anti-2SC antibody to detect succination in the absence of ester hydrolysis, as shown above (Fig. 1A). After stripping the membranes, we probed the same blots with the ␣-tubulin antibody B-7, which is sensitive to tubulin succination (30). As shown in Fig. 3B  (panel ␣-tubulin B-7), only the tubulin samples treated with fumarate showed a decrease in tubulin detectability, and this correlates with the increased succination detected by our 2SC antibody (Fig. 3B, panels 2SC short and long). Interestingly, tubulin samples treated with up to 500 M DMF, though succinated ( Fig. 3A and Table I), did not show decreased detectability by the ␣-tubulin antibody B-7 compared with controls (panel ␣-tubulin B-7), suggesting that succination by DMF, which leads to an uncharged modification of the cysteine residues, does not affect the interaction of succinated tubulin with the antibody. Moreover, pronounced succination of tubulin by either fumarate or DMF did not affect its detectability by the ␣-tubulin antibody DM1A, as already described, and we routinely use and recommend this antibody to detect total levels of tubulin in samples where succination may be present (30). Considering the abundance of commercially available antibodies against diverse epitopes for the same protein, it is important to examine the antigens used in order to establish reliable immunoblotting methods for total protein detection when post-translational modifications are endogenously present or introduced.
Although the succination of tubulin by DMF did not affect tubulin interaction with the tested antibodies, we were interested in determining if tubulin interactions with motor proteins such as dynein or kinesin were altered, potentially impacting microtubule dynamics. Primary cultures of rat dorsal root ganglion (DRG) neurons were prepared and were treated with 50 M DMF for 24 h or left untreated, and axonal lysosomal trafficking was monitored following lysosomal labeling with LysoTracker. Fig. 4A shows representative kymographs for the untreated DRG neurons (control panel) and DRG neurons treated with 50 M DMF (50 M DMF panel). No differences because of DMF treatment were observed in the percentage of lysosomes undergoing either anterograde or retrograde trafficking, or in the percentage of lysosomes changing directions (both), or of static lysosomes per axon (Fig. 4B). The average speed and run length of the motile lysosomes were also unchanged by DMF treatment (Fig. 4C). However, when motile events were separated according to different speeds (SP), DMF treatment changed the distribution of speed for retrograde events (Fig. 4D). Compared with control, there was higher percentage of 1-2 m/sec motile events but lower percentage of 0.5-1 m/sec motile events in DMF treated DRG neurons, with no changes in the run length (RL). The anterograde distribution of speed and run length were unaffected by DMF treatment. Overall the results indicate that lysosomal trafficking was not significantly altered, suggesting that DMF does not adversely affect this aspect of microtubule dynamics.
Collapsin Response Mediator 2 (CRMP2) as a Target of DMF Modification-All three cell preparations confirmed succination of the Collapsin Response Mediator 2 (CRMP2) peptide GLYDGPVCEVSVTPK by either MMF or DMF (Table I). The y 8 and b 8 fragment ion designations in the annotated spectrum shown in Fig. 5 confirm the designation of Cys504 as the site of succination by DMF in primary rat neurons (ϩ2 charge state, 854.4109 m/z). The presence of pyridylethylated Cys504, representative of the unmodified peptide, was also detected in the primary rat neurons (ϩ2 charge state, 834.9182 m/z, supplemental Fig. S4A). In addition, the MMF modification of Cys504 was also confirmed as shown by the annotated spectra of the GLYDGPVC MMF EVSVTPK peptide from DMF-treated primary astrocytes and DMF-treated N1E-115 cells ( Fig. 5B and supplemental Fig. S4B, respectively). The C terminus of CRMP2 contains several GSK3␤ phosphorylation sites (Thr509, Thr514 and Ser 522) that have a role in mediating axonal growth cone retraction. The proximity of Cys504 to these phosphorylation sites, as well as the dependence of phosphorylation on Cys504 oxidation (40), suggest that site-specific succination of CRMP2 may contribute to DMF mediated axonal preservation (41).
Ubiquitin C-Terminal Hydrolase L1 (Uchl1) Activity is Reduced-Ubiquitin C-Terminal hydrolase L1 (Uchl1) is an abundant deubiquitinase that constitutes up to 5% of total neuronal protein (42). Uchl1 cleaves ubiquitin from small peptide substrates and contributes to the regeneration of the intracellular ubiquitin pool (43). The modification of Uchl1 by both DMF and MMF was confirmed by a prominent y 2 ion on the Cys152-containing peptide NEAIQAAHDSVAQEGQCR supplemental Fig. S5A and S5B), and this site has previously been shown to subject to chemical modification by other agents (44 -46). DMF treatment did not alter the intracellular protein abundance of Uchl1 (detected using an antibody that specifically recognizes an N-terminal Uchl1 portion that did not contain any cysteines), nor did it alter the total levels of ubiquitinated protein by immunological detection supplemental Fig. S5C and S5E). UchL1 deubiquitinase activity was reduced in the presence of increasing concentrations of DMF (supplemental Fig. S5D, n ϭ 5, p Ͻ 0.002 for 100 M versus control), suggesting that further analyses are warranted to determine if the intracellular pools of free ubiquitin or Uchl1 structure are negatively impacted following exposure to DMF.
Mitochondrial Respiration is Minimally Altered by DMF-Because DMF treatment also resulted in succination of the abundant mitochondrial outer membrane protein voltage dependent anion channel-1 (VDAC-1, Table I) ergy production, including ETC complex II 30 kDa subunit and succinate dehydrogenase a, complex III core 2 subunit, complex V subunit ␣ and tricarboxylic acid cycle fumarase, were unaffected by the 50 M DMF treatment. Taken together, these results suggest that low dose 10 M treatment with DMF does not affect the energy production machinery or respiration, whereas higher concentrations of DMF (50 M) may lead to a partial decrease in some respiration parameters. DISCUSSION Dimethyl fumarate (DMF) is an approved anti-inflammatory agent for the treatment of relapsing remitting multiple sclerosis and psoriasis. As an electrophile, DMF has been demonstrated to chemically modify cysteine residues in Keap1 and induce Nrf2-driven antioxidant response element gene transcription. Because DMF has been shown to improve the survival of neurons and astrocytes both in vivo and in vitro, we examined novel targets of DMF thiol reactivity in these cells types in order to better explain why DMF also offers benefit in Nrf2-knockout mouse models of inflammatory disease (17).
Using both primary neurons and astrocytes, in addition to a neuronal cell line, we confirmed the identity of 27 uniquely modified cysteine residues, representing 24 distinct protein subunits in these cell types (Table I). We observed that DMF entered the cells and reacted rapidly with specific intracellular proteins, often with both methyl groups intact, indicating that DMF modifies many proteins before being hydrolyzed to the less reactive monomethyl fumarate (MMF) or fumarate itself. Fig. 1 further confirms that DMF is not completely hydrolyzed to fumarate inside the cell, as we could only detect succinated using an anti-2-succinocysteine antibody following base hydrolysis. Our results agree with the observations of Blewett et al., who recently described the reactivity of DMF versus MMF toward primary T cell proteins after 4 h of exposure to DMF in vitro (19). They indirectly quantified ϳ40 DMF-sensitive cysteine residues, following detection of the cysteine residues that no longer reacted with the electrophile iodoacetamidealkyne (isoTOP-ABPP method) (19) These residues were representative of ϳ1% of the 2400 total cysteine residues identified, and confirm that fractional modification of functionally significant thiols is sufficient to contribute to altered T-cell activity (measured by reduced IL-2 production). In contrast to the isoTOP-ABPP, we searched directly for cysteine residues that had been variably modified by DMF, MMF or fumarate, in addition to pyridylethylated cysteines, and found that both DMF and MMF modified cysteine residues were detected across 24 proteins. Because our DMF incubation was for 24 h, it is possible that MMF may have more of an opportunity to be generated by esterase activity and then react with proteins. We did not detect any endogenously succinated residues directly modified by fumarate, however, because we were using positive ion mode mass spectrometry the detection of the negatively charged succinocysteine is more difficult to detect versus the fumarate esters. The longer exposure to DMF likely increased the probability of detecting modified proteins with extended half lives in the cell, such as the cytoskeletal protein subunits that were detected for tubulin, vimentin, glial fibrillary acid protein and cofilin-1. Exogenously applied DMF is more likely to react quickly with cytosolic proteins versus endogenously produced intramitochondrial fumarate that increases in other models such as fumarase deficient cancer cells (28). In the current study we observe the modification of only one mitochondrial protein by exogenous DMF, Voltage-Dependent Anion Channel-1, which is located on the outer mitochondrial membrane.
Cofilin-1 contributes to the remodeling of the cytoskeleton by depolymerizing filamentous (F) actin to provide actin monomers for growing actin filaments. We detected the modification of cofilin by MMF on Cys139 in astrocytes ( Fig. 2A) and confirmed this site of modification when cofilin was incubated with DMF in vitro. The functional assessment of cofilin activity revealed that Cys139 modification resulted in a ϳ30% decrease in cofilin's ability to depolymerize actin versus unmodified cofilin. DMF has been shown to enhance the differ- to a PVDF membrane and blotted with antibodies to the following ETC components: Complex I NDUFB8, Complex II subunit 30 and succinate dehydrogenase a (SDHa), Complex III core 2 subunit, and Complex V ␣-subunit. DMF treatment did not significantly change the levels of the ETC markers, except for CI-NDUFB8, which was decreased with 50 M DMF. Fumarase and ␣-tubulin antibodies, as well as Coomassie staining of the membrane, were used to verify even loading, n ϭ 3 per treatment group. The molecular masses of the proteins are indicated on the left-hand side. entiation of oligodendrocyte precursor cells (OPC) to oligodendrocytes, as evidenced by increased expression of O4 (47). Recently, Zuchero et al. described that early oligodendrocyte differentiation from OPCs is accompanied by the extension of processes containing ordered arrays of actin filaments, requiring polymerized actin and low cofilin activity (48). As the oligodendrocyte begins to mature further the actin cytoskeleton is disassembled, particularly by proteins such as cofilin, therefore differentiation requires dynamic changes in actin structure. Our data suggests that DMF-modified cofilin would be beneficial for actin arborization during early OPC differentiation. DMF may also act in a neuroprotective manner by inducing oligodendrocyte ensheathment of the axon, even if myelination is not dramatically increased. In addition, DMF treatment reduces localized microglial activation in models of neuroinflammation (49,50), and cofilin has been shown to be critical for lipopolysaccharide mediated microglial activation as cofilin knockdown significantly inhibited microglial activity (51). Recently, magnetization transfer ratio measurements in humans have suggested that myelin density may be increased in those receiving delayed-release DMF treatment versus placebo (52), suggesting that DMF may beneficially modulate aspects of myelination during MS treatment. Together with the current literature, our data suggests that the DMF-mediated regulation of cofilin's actin-severing activity in dynamic glial cell populations, including oligodendrocytes, warrants further investigation.
Collapsin Response Mediator Protein 2 (CMRP2) is an intracellular phosphoprotein that contributes to axonal growth by transporting ␣␤ tubulin heterodimers to the plus ends of the growing microtubule. The phosphorylation of CRMP2 leads to the collapse of the growth cone in response to environmental cues such as Semaphorin3A (Sema3A). CRMP2 is phosphorylated by Cdk5 at Ser522, and subsequently by GSK3␤ at Thr509, Thr514 and Ser518, leading to the disruption of the CRMP2:tubulin association and neurite retraction (53). In the current study, DMF contributed to the modification of Cys504 GLYDGPVCEVSVTPK in both primary neurons and N1E-115 derived neurons ( Fig. 5A and supplemental Fig. S4A). Interestingly, the oxidation of Cys504 has been demonstrated to be critical for the Sema3A mediated recruitment of GSK3␤ (40). Sema3A signaling stimulates the production of intracellular hydrogen peroxide, leading to the oxidation and dimerization of adjacent Cys504. Thioredoxin (TRX) interacts to reduce the oxidized proteins and this CRMP2/TRX complex recruits GSK3␤, resulting in CRMP2 phosphorylation and growth cone collapse (40). Because the oxidation of CRMP2 is necessary for GSK3␤ recruitment, it is possible that modification of this cysteine prevents oxidationmediated phosphorylation. Consequently DMF treatment might be expected to preserve axonal integrity and prevent retraction in response to local degenerative cues. Enhanced phosphorylation of another CRMP2 site, Thr555 is abundant in active multiple sclerosis lesions from human autopsy (54). It is also significant that DMF, in combination with interferon ␤ therapy, resulted in significant axonal preservation in a murine MS model (41). Our data suggests that CRMP2 Cys504 succination could be a significant mediator of the beneficial effects of DMF directly in neuronal cells, making it an attractive site for targeted therapies to prevent axonal loss.
We observed a trend toward the stimulation of mitochondrial respiration parameters following 10 M DMF treatment of N1E neurons for 24 h, indicating that this dose does not acutely impair mitochondrial function. In contrast, 50 M DMF tended to reduce ATP production, maximal respiration and spare respiratory capacity, however, this did not appear to be because of a loss of mitochondrial content as components of the electron transport chain (ETC) and tricarboxylic acid cycle examined were not affected (Fig. 6C). The exception to this was the decreased protein levels of the NDUFB8 subunit of Complex I, a 22kDa accessory protein that is required for the complete assembly of Complex I, it is possible that a loss of efficient electron transfer because of impaired Complex I assembly decreased NADH oxidation at this site. This may also explain why basal respiration appears to be unaffected by 50 M DMF (Fig. 6A and 6B); FADH 2 may sustain respiration under basal conditions, but the subsequent chemical challenges to determine maximal respiration and spare capacity put pressure on the ETC to increase electron transfer in cells where Complex I function may be compromised. The decrease in mitochondrial respiration observed with 50 M DMF is like the decreased OCR parameters reported by Ahuja et al. using 20 M DMF in murine embryonic fibroblasts for 24 h (55). Interestingly, Hayashi et al. have recently reported that 10 and 30 M DMF increase mtDNA and maximal respiration after 48 h in human fibroblasts, and have shown tissue specific increases in mtDNA content (including the cerebellum) following 2 weeks of DMF administration to healthy mice (56). Taken together, the data suggests that variable doses and incubation times can have different effects on the models studied, although prolonged exposures in humans need to be monitored (57). In the current study we observed only a mild effect of DMF to alter lysosomal trafficking along microtubules in neurons, however in patients DMF may have more profound effects on other aspects of microtubule dynamics in cells directly exposed to the oral dosage.
In summary, although DMF is a prominent activator of Nrf2 mediated transcription, it also ameliorates experimental autoimmune encephalitis and the inflammatory status in Nrf2knockout models. The novel discovery in this study that DMF and MMF may modify regulatory thiols on proteins such as cofilin-1 and CRMP2 suggest that DMF treatment may directly contribute to axonal preservation and remyelination, which is distinct from other MS treatments that are only known to modulate the inflammatory cell profile. Further investigation of the effects of DMF therapy on these novel neuroprotective targets in vivo will guide the development of more specific compounds for MS therapy.