a potential link between autoimmunity and neurodegeneration in immune-mediated neurological disease

Journal of Neuroimmunology 235 (2011) 56–69

Contents lists available at ScienceDirect

Journal of Neuroimmunology

j ourna l homepage: www.e lsev ie r.com/ locate / jneuro im

A potential link between autoimmunity and neurodegeneration inimmune-mediated neurological disease

Sangmin Lee a,b, Lijing Xu d, Yoojin Shin a,b, Lidia Gardner a,b, Anastasia Hartzes a,b, F. Curtis Dohan c,Cedric Raine e, Ramin Homayouni d, Michael C. Levin a,b,⁎a Veterans Administration Medical Center, 1030 Jefferson Avenue, Memphis, TN 38104, United Statesb Department of Neurology, University of Tennessee Health Science Ctr, 855 Monroe Avenue, Memphis, TN 38163, United Statesc Department of Pathology, University of Tennessee Health Science Ctr, 855 Monroe Avenue, Memphis, TN 38163, United Statesd Bioinformatics Program, University of Memphis, 3774 Walker Ave, Rm 201 LS, Memphis, TN 38152, United Statese Neuropathology Department, Albert Einstein College of Medicine, 1300 Morris Park Avenue, Bronx, NY 10461, United States

⁎ Corresponding author at: University of TennesseBuilding Room 415, 855Monroe Avenue, Memphis, TN 38523 8990X6809, +1 901 448 2243; fax: +1 901 577 748

E-mail address: [email protected] (M.C. Levin).

0165-5728/$ – see front matter. Published by Elsevierdoi:10.1016/j.jneuroim.2011.02.007

a b s t r a c t
a r t i c l e i n f o
Article history:Received 22 November 2010Received in revised form 11 January 2011Accepted 8 February 2011

Keywords:NeurodegenerationAutoimmunityMultiple sclerosisSpastinBioinformatics

Multiple sclerosis (MS) patients make antibodies to heterogeneous nuclear ribonuclear protein A1 (hnRNP-A1),a nucleocytoplasmic protein. We hypothesized this autoimmune reaction might contribute to neurodegenera-tion. Antibodies fromMS patients reacted with hnRNP-A1-‘M9’, its nuclear translocation sequence. Transfectionof anti-M9 antibodies into neurons resulted in neuronal injury and changes in transcripts related to hnRNP-A1function. Importantly, RNA levels for the spinal paraplegia genes (SPGs) decreased. Changes in SPG RNA levelswere confirmed in neurons purified fromMS brains. Also, we showmolecular interactions between spastin (theencoded protein of SPG4) and hnRNP-A1. These data suggest a link between autoimmunity, clinical phenotypeand neurodegeneration in MS.

e Health Science Center, Link163, United States. Tel.: +1 9016.

B.V.

Published by Elsevier B.V.

1. Introduction

Neurodegeneration is a major contributor to neurological disabilityin MS patients (Bjartmar et al., 2003; Dutta and Trapp, 2007; Lassmann,2007; Lassmann et al., 2007). Following the relapse–remitting stage ofMS, patients develop ‘secondary progressive’MS, in which neurologicaldeterioration continues in the absence of relapses (Dutta and Trapp,2007; Lassmann et al., 2007; Noseworthy et al., 2000). Alternatively,others are diagnosed with primary progressive MS in which neurolog-ical dysfunction occurs without relapses from disease onset (Dutta andTrapp, 2007; Lassmann et al., 2007; Noseworthy et al., 2000). Thus, themajority ofMSpatients developprogressive neurological disease (Duttaand Trapp, 2007; Lassmann et al., 2007). Neuroradiological, neuropath-ological and animal studies of MS showmarkers for neurodegenerationmanifest during the progressive phase of the illness. For example, MRIimages of the brain show axonal damage, which correlates withneurological disability (Bjartmar et al., 2000; Lassmann et al., 2007).Also, accumulation of amyloid precursor protein (APP) (Ferguson et al.,1997; Kornek et al., 2000) and staining for non-phosphorylated

neurofilament (SMI-32) (Trapp et al., 1998)—both markers of axonalinjury—showed that axonal damage is amajor component ofMS lesions.Subsequent studies confirmed these observations in MS and in someexperiments using experimental allergic encephalomyelitis (EAE)induced with myelin oligodendrocyte glycoprotein (MOG), a model ofMS characterized by neurodegeneration in which antibodies play asignificant role (Aboul-Enein et al., 2006; Brown and Sawchenko, 2007;Gold et al., 2006; Kornek et al., 2000). Importantly, there are multiplemedications for the treatment of relapsing–remitting MS, but none areefficacious in progressive forms of the disease.

What causes MS is unknown, but evidence suggests that interac-tions between environmental agents, auto-antigens and the immuneresponse in genetically susceptible people contribute to its cause(Dutta and Trapp, 2007; Lassmann et al., 2007; Noseworthy et al.,2000). Because no environmental agents have been rigorously provento cause MS, we use human T-lymphotropic virus type 1 (HTLV-1)associated myelopathy/tropical spastic paraparesis (HAM/TSP) as amodel to study MS (Lee et al., 2005; Levin et al., 2002a). HAM/TSP iscaused by infection with HTLV-1, which allows for direct comparisonbetween an environmental agent and auto-antigens. HAM/TSPpatients develop spastic paraparesis and sensory abnormalities thatcan be clinically indistinguishable from progressive forms of MS,particularly primary progressive MS (Levin et al., 1997; Levin andJacobson, 1997). Furthermore, both diseases are associated withneurodegeneration, particularly of corticospinal tracts and posteriorcolumns (Ganter et al., 1999; Lee et al., 2005; Lovas et al., 2000;

http://dx.doi.org/10.1016/j.jneuroim.2011.02.007

mailto:[email protected]

http://dx.doi.org/10.1016/j.jneuroim.2011.02.007

http://www.sciencedirect.com/science/journal/01655728

57S. Lee et al. / Journal of Neuroimmunology 235 (2011) 56–69

Umehara et al., 2000). HAM/TSP patients make antibodies to HTLV-1that cross-react with hnRNP A1 implicating molecular mimicry in thepathogenesis of HAM/TSP. Molecular mimicry is strongly associatedwith other immune-mediated neurological diseases such as theaxonal form of Guillain–Barre Syndrome and also thought tocontribute to the pathogenesis of MS (Kalume et al., 2004; Lee andLevin, 2008; Lee et al., 2006b, 2005; Levin et al., 1998, 2002a, 2002b;Yuki et al., 2004). hnRNP A1 is an RNA binding protein that plays acrucial role in mRNA transport, metabolism and translation and isrequired for normal cellular functioning (Dreyfuss et al., 2002).Antibodies from HAM/TSP patients reacted specifically with the ‘M9’shuttling domain of hnRNP A1, the sequence required for its transportinto and out of the nucleus (also known as the nuclear exportsequence (NES) and nuclear localization sequence (NLS)) (Lee et al.,2006b; Levin et al., 2002a; Michael et al., 1995). Notably, MS patientsmake antibodies to hnRNP A1 as well as to neuronal and axonalantigens (Norgren et al., 2005; Rawes et al., 1997; Sadatipour et al.,1998; Sueoka et al., 2004). Further, MS patients make antibodies tooligodendrocytes as well as myelin, and these antibodies maycontribute to the pathogenesis of the disease (Abramsky et al.,1977; Edgington and Dalessio, 1970; Genain et al., 1999). Therefore,we hypothesized that antibodies from MS patients would alsorecognize ‘M9’ and that antibodies to ‘M9’ might play a role inneurodegeneration in an in vitro model of antibody-mediatedautoimmunity.

2. Methods

2.1. Sera and tissue

For sera, informed consent is on record from participatinginstitutions. Tissue samples included brain samples from threepatients with MS, two patients with HAM/TSP and one normalcontrol. The clinical and autopsy data are contained within supple-ment 2.

2.2. Samples and IgG purification

IgG was harvested from the serum of MS and HAM/TSP patientsand normal control participants using the Melon Gel IgG PurificationKit (Pierce Biotechnology, Rockford, IL) according to the manufac-turer's instructions. Resulting IgG was dialyzed three times againstphosphate buffered saline (PBS) buffer. Purified IgG was quantifiedusing the BCA method.

2.3. Cell lines and cloning of hnRNP A1

NT-2 and SK-N-SH cells (American Type Culture Collection) werecultured under standard conditions as previously described (Lee et al.,2006b). Cloning of hnRNP A1 and its recombinant fragments wascompleted as previous described (Lee et al., 2006b; Levin et al., 2002a).

2.4. Antibodies and additional reagents

The antibodies to human transportin (ab10303), hnRNP A1(ab4791) and KLH (ab 34766) were purchased from Abcam. Anti-Spastinmouse monoclonal antibodies (54443) were purchased from SantaCruz Biotechnology. Secondary antibodies were as follows. Anti-rabbit–TxRed antibodies (TI-1000) and anti-mouse FITC antibodies(TI-2000) were from Vector laboratories, and anti-goat-FITC anti-bodies (sc2024) were from Santa Cruz Biotechnologies.

2.5. Western blotting

Western blotting was performed as described previously (Leeet al., 2006b; Levin et al., 2002a). Briefly, purified hnRNP A1, GST-

hnRNP A1 fragments, human tissues, human neurons and dNT2 cellswere separated on 8–16% gradient gels to improve resolution of thehighmolecular weight bands and transferred to PVDF (polyvinylidenedifluoride) membranes (Amersham Biosciences), followed by West-ern blotting. For human tissues, neurons and cell lines, 30 mcg ofproteinwas added per lane. Human tissues were extractedwith T-PERtissue extraction reagent containing Halt protease inhibitor mixture(Pierce Technology). Samples were homogenized (4 °C), and tissuedebris was removed by centrifugation (10,000 rpm, 5 min). Thesupernatant was reserved and its protein concentration was deter-mined by the BCA protein assay (Pierce Technology). For epitopeanalyses, 0.2 mcg of each fusion protein was added per lane. Forscreening of multiple IgG samples for immunoreactivity to hnRNPA1-M9, 40 mcg of GST-M9 was added to a single broad center lane(allowing for simultaneous detection of up to 33 samples) using aSURF-BLOT system (Idea Scientific Company). Western blot analysiswas performed using 1:100-diluted MS IgG and normal IgG. Initialepitope analysis was performed using IgG isolated from two MSpatients. Goat anti-human IgG linked to horseradish peroxidase wasutilized as a secondary antibody at a dilution of 1:25,000. Westernblots were visualized using a chemiluminescent substrate (ECLplus,GE Healthcare). The first patient was female and had secondaryprogressive MS, with symptoms for 15 years, an expanded disabilitystatus scale (EDSS) of 4.0 with both pyramidal and posterior columnsigns. The second patient was male and also had secondaryprogressive MS, with symptoms for 41 years, an EDSS of 6.0 withboth pyramidal and posterior column signs. The clinical character-istics of the study population are presented in Supplement 2, Table A.An inhibition assay was performed using MS IgG pre-incubated withhnRNP A1-M9293–304 and hnRNP A1185–196 (control) fragments insequential concentrations of 0 μg/μl, 1 μg/μl, 10 μg/μl, and 50 μg/μl perlane.

2.6. Transfection and detection of antibodies in NT-2 cells

Unlabeled (for microarray experiments) and FITC (for immuno-cytochemistry experiments) anti-hnRNP A1 or anti-KLH antibodieswere transfected using a liposomal-based protein delivery kit, permanufacturer's instructions (Bioporter, Genlantis). Twenty-fourhours after transfection, cells were prepared for immunohistochem-istry using standard procedures.

2.7. Immunohistochemistry

NT-2 cells were grown in Poly-D-lysine covered chambers, treatedwith retinoic acid and mitotic inhibitors prior to the experiment. Cellswere fixed in 4% paraformaldehyde for 10 min at room temperature(RT), then washed in 1× PBS, and permeabilized in 0.2% Triton X-100,for 30 min at RT.

All slides were blocked in 6%Milk-TBST andwashed in PBS prior toantibody application. Primary antibodies were applied for 1.5 h at RT.Each slide was double stained with rabbit polyclonal antibodies tohnRNP A1 and transportin or spastin antibodies. Slides were washed5× in 1× PBS. Secondary antibodies were: FITC (transportin, spastin)and TxRed (hnRNP A1) labeled. Slides were incubated for 1.5 h withsecondary antibodies, then washed and mounted using DAPImounting media (Millipore). Labeling with Fluoro Jade C wasperformed as described (Schmued et al., 2005).

2.8. Microarray and quantitative real-time PCR

Each experimental group (untouched control, anti-hnRNP A1 andanti-KLH) consisted of three separate 75 cm2

flasks of NT2 neurons.Unlabeled antibody (50 mcg/2×107 NT2 cells) was transfected intoeach of the flasks for 24 h per the manufacturer's instructions.Following transfection of antibodies into the NT-2 neurons, total RNA

58 S. Lee et al. / Journal of Neuroimmunology 235 (2011) 56–69

was extracted using the RNeasy Mini Kit (Qiagen). RNA integrity andpurity were confirmed by capillary electrophoresis and used foranalysis of gene expression using the Affymetrix HG U133 Plus 2GeneChip. GeneSpring software was used to perform quality controland statistical analyses. The expression value for each probe set wasnormalized across the chip as well as across the samples in theGeneSpring Software package (Lowess normalization). Of 38,500genes, we found 866 transcripts that were significantly (Pb0.05,ANOVA) altered by M9 specific antibodies (Supplement 3) comparedto the control (anti-KLH and untouched) samples. ANOVA selectedprobe sets were subjected to Benjamini and Hochberg correction forfalse-positive reductions. The resulting 866 gene set was analyzedusing a dual bioinformatic approach: (Geneontology and GeneIn-dexer) to identify transcripts affected by the introduction of hnRNP A1antibodies into NT-2 neurons. To verify the differences in geneexpression that are observed, total RNA (0.2 mg) was reversetranscribed using the High Capacity cDNA Reverse Transcription Kit(Applied Biosystems). Next, qRT-PCR was performed on an ABI 7900(Applied Biosystems) using the TaqMan Universal PCR Master Mix(Applied Biosystems). Primers were designed and supplied byApplied Biosystems. The product size was initially monitored byagarose gel electrophoresis. Melting curves were analyzed to controlfor specificity of PCR reactions. The data on genes that weredifferentially expressed was normalized to the expression of one ofthe housekeeping genes, either GAPDH or actin. The relative unitswere calculated from a standard curve, plotting 3 different concen-trations against the PCR cycle number at which the measuredintensity reaches a fixed value (with a 10-fold increment equivalentto ~3.1 cycles). Fold change comparing anti-hnRNP A1 to anti-KLH ortechnical controls were calculated by comparative quantificationalgorithm—delta delta Ct method (Fold difference=2−ΔΔCt).

2.9. Survival experiments on cultured neurons

NT-2 cells were plated at an equal amount (150,000/ml) in 2-wellpoly-D-lysine coated chambers. Cells were transfected with 0.1 μg/mlof Atto550 (excitation 556 nm and emission 578 nm) labeled hnRNPA1 (Abcam) or rIgG (Millipore) antibodies using Bioporter reagent asdescribed above. Untransfected NT-2 cells were utilized as experi-mental control. The antibody concentration at which there was nocytotoxicity detected was determined in a series of experiments witha set range of 0.1–3 μg/ml. Cells were imaged live every 2 days for theduration of one week on Zeiss Axio Observer A1 inverted microscopeusing 20×/0.3 LD A-Plan objective. The Axiovision image analysissoftware was used to measure the length of the axonal process withinthe field of view. Five to ten field views were taken per condition. Toquantify the morphological response overtime, we defined processesloss (PL) as processes length before transfection (PLb) minusprocesses length after transfection (PLa) PL=PLb−PLa.

2.10. Isolation of neurons

Isolation of neurons was performed as described previously (Levinet al., 1998, 2002a). A cerebral cortex sample was dissected with arazor blade, gently separated and suspended in 2% Ficoll, 50 mM Tris–HCl buffer with 0.32 M sucrose, 50 mM NaCl and 0.5 mM EDTA, pH7.4. The sample was sieved through nylon mesh with the followingprogressively finer pore sizes: 1000 mm, 330 mm, 120 mm, 73 mmand 53 mm. Next, an equal volume of 30% Ficoll, 50 mM Tris–HClbuffer with 0.32 M sucrose, 50 mM NaCl and 0.5 mM EDTA, pH 7.4,was added to the suspension to make a final Ficoll concentration of20%. Neuronal and glial cell fractions were separated by centrifugationthrough a discontinuous density gradient of Ficoll. After centrifuga-tion, the neuronal cell fraction (floating between 30% Ficoll and 40%Ficoll) was obtained. This fraction was diluted with PBS (1:5) andcentrifuged at 2500 rpm for 20 min. to yield a pellet. The supernatant

was discarded and the pellet was suspended in PBS, and used forimmunomagnetic isolation and Western blotting.

2.11. Immunomagnetic separation of neurons

Neurons isolated from human cortex were incubated with tetanustoxin C fragment (Sigma) that binds to cell surface gangliosidespresent on neurons. The cells were reacted with an anti-TTC mouseIgG (Abcam) and were subsequently separated by CELLection BiotinBinder Dynabeads (Invitrogen). The Dynabeads had been coated withrecombinant streptavidin via DNA linker followed by conjugationwith a biotinylated goat anti-mouse IgG (Abcam). After selecting thebound cells in a magnetic particle concentrator and removal from thebeads with DNase I, the cells were used for qPCR and Westernblotting. For quantitative qRT-PCR experiments of neurons andcerebral cortex experiments, RNA quality was assessed using theNanodrop method (Thermo scientific). All samples had a A260/A280ratio N2.0. RNA quantity was normalized to contain equal amounts ofRNA in each sample.

2.12. Immunoprecipitation assay

SK-N-SH neuronal cells were transfected with a Strep Tag IIrecombinant hnRNP A1 using Effectene reagent (Quagen) accordingto the manufacturer's instructions. Forty-eight hours after transfec-tion cells were lysed in immunoprecipitation (IP) buffer (150 mMNaCl, 10 mM NaH2PO4,pH 7.2, 0.2% Triton X-100, 1 mM PMSF) andcentrifuged (14,000 g for 15 min at 4 °C) to remove insoluble cellulardebris. Equal lysate volume (900 ul) was added to Anti-strep tag II andrIgG agarose A beads at 4 °C with gentle rotation overnight. Theimmune complexes were washed 3 times in IP buffer and eluted fromthe beads with 2× sample buffer containing 20 mM dithiothreitol.Resolved proteins and lysate inputs were separated by SDS-PAGEunder denaturing conditions and electro-blotted to PVDF membranefor Western procedure.

3. Results

3.1. IgG isolated from MS patients recognizes the hnRNP A1-M9 epitope

IgG from MS patients was tested for immunoreactivity withhuman tissues by Western blot analysis. Like HAM/TSP IgG (Fig. 1A,Supplement 1A) (Lee et al., 2006b; Levin et al., 2002a), MS IgG reactedwith recombinant hnRNP A1, and proteins derived from human brain,brain nuclear fraction and NT-2 neurons (a human neuronal cell line)at the identical molecular weight. MS IgG did not react with proteinsextracted from systemic organs including lung, heart, kidney andliver. IgG from a normal control did not react with neurons or systemicorgans. Next, to identify the hnRNP A1 epitopes recognized byMS IgG,we designed primers representing biologically important regions ofthe protein (Fig. 1B, Supplement 1B, C) and the overlapping mutantswere cloned, expressed, separated by sodium dodecyl sulfatepolyacrylamide gel electrophoresis (SDS-PAGE) and transferred tomembranes for Western blotting (Lee et al., 2006b). MS IgG reactedwith full-length hnRNP A1 as well as the glycine-rich C-terminalfragment (AA191–320), which contains the M9 sequence (Fig. 1C).There was minimal reactivity with the N-terminal sequence(AA1–197), which contains the RNA binding domains (RBD). IgGimmunoreacted with AA1–256, which contains the RGG domain andthe RBDs, suggesting that the RGG domain contributes to theimmunoreactivity. These data suggest that the core epitope recog-nized byMS IgG parallels HAM/TSP IgG and is contained within the C-.terminal fragment (AA191–320), inclusive of the RGG domain and M9,both of which contribute to hnRNP A1 function (Dreyfuss et al., 2002;Lee et al., 2006b; Michael et al., 1995; Nichols et al., 2000).

Fig. 1. hnRNP A1 immunoreactivity and epitope mapping. A. IgG purified fromHAM/TSP (TSP) andmultiple sclerosis (MS) patients reacted with recombinant hnRNP A1 (rhnRNPA1)and brain, brain nuclear fraction and NT-2 neurons, but not with proteins derived from kidney, spleen, heart, liver or lung. Normal IgG (NL) showed no immunoreactivity. B. Diagramof functional regions of hnRNP A1. The RNA binding domains (RBD) are located in the N-terminal half of the molecule. The glycine-rich C-terminal contains the RGG box (RGG), RGGdomain and the M9 shuttling domain. The epitopes defined by the overlapping fragment experiments were localized to the RGG domain and M9 (see text and supplement 1). C. IgGfrom HAM/TSP and MS patients reacted with fragments representing AA1–256, AA1–320 and AA191–320) but not with AA1–197, suggesting immunoreactive epitopes are contained withthe C-terminal half of hnRNP A1. The doublet shown in the fragments 191–320 lane are likely due to co-purification of GST fusion proteins of different length each containing thetarget fragment (as previously reported (Lee et al., 2006b)). IgG from a normal (NL) patient did not immunoreact with the hnRNP A1 fragments. D. IgG purified fromMS andHAM/TSPpatients reacted with AA191–232, AA263–304 and AA 281–320. These epitopes overlap the RGG domain and M9 shuttling sequence. An antibody to hnRNP A1-M9 (A1 Ab) reacted withAA263–304 and AA281–320. (NL=normal control). E. HAM/TSP IgG reacted with AA191–202 and AA197–208. MS IgG immunoreactivity was slightly broader and also included AA203–214.Epitope analyses localized the epitope to AA191–202. MS and HAM/TSP IgG reacted with AA287–298, AA293–304 and AA299–310. Epitope analyses localized the epitope to AA293–304. Theanti-M9 antibody paralleled this immunoreactivity. F. Increasing concentrations of the M9 peptide abolished MS IgG reactivity to hnRNP A1 (left panel) and NT-2 neurons (rightpanel). There was no change in reactivity when using a control peptide. G. Reactivity ofMS and HAM/TSP IgGwith the hnRNP A1-M9 target epitope (AA293–304). IgG from all of theMSand HAM/TSP patients reacted with M9, as did cerebrospinal fluid (CSF) samples from these patients. There was no reactivity with IgG isolated from normal controls. In a separateexperiment, IgG from Alzheimer's disease patients also did not react with M9. MS and TSP IgG were used as positive controls.


To further delineate the epitope, overlapping primers weredesigned spanning the C-terminal fragment (Supplement 1C) andindividual fusion proteins were prepared for Western blotting. Incontrast to control IgG, MS IgG revealed a pattern suggestive of twoepitopes. The first showed intense immunoreactivity with the N-

terminal of the RGG domain (AA191–232), and no reactivity with eitherthe adjacent fragment (AA209–250) or with AA227–268 (Fig. 1D). Thesedata suggest the epitope lies within AA191–208. A second epitopeincluded AA263–304 and AA281–320, but neither AA245–286 nor AA303–320

(Fig. 1D). This suggests a second epitope is contained within AA287–302,


which overlaps the M9 shuttling domain (AA268–305). Immunoreactiv-ity from HAM/TSP patients was identical and an antibody to hnRNP A1AA293–304 paralleled the immunoreactivity to M9.

To define the fine epitope specificity of hnRNP A1 recognized byMS IgG, overlapping fragments that contain these two core epitopes(AA191–208 and AA287–302) were tested (Supplement 1C, Fig. 1E). Asshown and in previous studies (Lee et al., 2006b), the HAM/TSP IgGimmunoreactivity pattern was: AA191–202 (strongly (+)), AA197–208

(minimal (+)) and AA203–214 (−). Since fragment AA203–214 isnegative, the C-terminal of the epitope is AA202. The fragmentcontaining AA197–208 was mildly positive, suggesting that AA197–202,but not AA203–208 contributed to the epitope. Based on these data, theHAM/TSP IgG epitope is AA 191-SSQRGRSGSGNF-202, located withinthe RGG domain of hnRNP A1. MS IgG closely paralleled these data,except for a mild contribution from AA203–214 (Fig. 1E). This suggeststhe epitope in MS patients is slightly broader compared to HAM/TSPpatients, which may be related to the wider range of clinicalpresentation in MS patients or to intra-molecular epitope spreading(Vanderlugt andMiller, 1996). There was no reactivity with AA185–196,thus confirming the epitope is contained entirely within the RGGdomain. This strategy was used to identify the specificity of theepitope that overlaps M9. HAM/TSP and MS IgG immunoreactivitywere identical and included fragments containing AA287–298,AA293–304 and AA299–310, but not AA281–292 or AA305–320. This indicatesthat the epitope is AA 293-GQYFAKPRNQGG-304 (Lee et al., 2006b),which is contained within the M9 sequence (AA268–305). The hnRNPA1293–304 antibody overlapped the M9, but not the RGG epitope(Fig. 1E).

Next, we confirmed whether the M9 epitope defined by theoverlapping fusion proteins contributed to the immunoreactivity ofMS IgG for hnRNP A1 and neurons. MS IgG was tested forimmunoreactivity with full-length hnRNP A1 and NT-2 neuronsfollowing increasing concentrations of the M9 peptide representingthe AA293–304 epitope. In contrast to a control peptide (AA185–196)(Fig. 1F) (Lee et al., 2006b), increasing concentrations of the M9peptide abolished MS IgG immunoreactivity with these targets,confirming the specificity of the immune reaction. We then screenedMS patients for immunoreactivity to the M9 peptide. (Fig. 1G,Supplement 2—Table A). IgG from all of the MS patients reactedwith M9 compared to controls. Importantly, IgG purified from the CSFof a MS and HAM/TSP patient reacted with M9 (Fig. 1G). In a separateexperiment, IgG isolated from Alzheimer's disease patients did notimmunoreact with hnRNP A1 (Fig. 1G).

3.2. Contribution of anti-hnRNP A1-M9 antibodies to neurodegenerationin a neuronal cell line

We hypothesized antibodies to hnRNP A1-M9 would contribute toneurodegeneration. To test this hypothesis, we transfected M9-specific antibodies into NT-2 neurons. NT-2 neurons maintain aneuronal phenotype, express neuronal antigens and resembleprimary neurons in culture (Pleasure et al., 1992). NT-2 neuronswere transfected with either anti-M9 antibodies, anti-KLH antibodies(experimental control) or no antibodies (‘untouched’ control) (n=3per group). In order to specifically study the contribution ofantibodies to hnRNP A1-M9, we utilized the rabbit anti-M9 antibodiesspecific for hnRNP A1293–304, rather than MS IgG which immunor-eacted with two epitopes (Fig. 1D, E). Immunohistochemistry of thefluorescently labeled antibodies confirmed their presence withinneurons (Fig. 2A). Neuron beta-tubulin III (NBT3, 55 kDa, a neuron-specific marker) showed an intense signal with NT-2 cells byimmunohistochemistry and Western blot (Fig. 2B and C). Further,glial fibrillary acidic protein (GFAP, 52 kDa, an astrocyte marker)immunoreactivity was negative (Fig. 2C). Taken together, thisconfirms the neuronal phenotype of the NT-2 cells. Next, we testedwhether anti-M9 antibodies might contribute to neurodegeneration.

In these experiments, we stained anti-M9 transfected NT-2 neuronswith Fluoro Jade C—a marker of neurodegeneration (Schmued et al.,2005). In EAE, Fluoro Jade C stained degenerating neurons and axons,which correlated with anti-SMI-32 expression (Brown and Saw-chenko, 2007). As shown in Fig. 2D, neuronal processes stained forboth Fluoro Jade C and anti-M9 transfected antibodies, suggestinganti-M9 antibodies’ involvement in neurodegeneration. BecauseFluoro Jade C is a marker for early neurodegeneration, we decidedto see how long it would take for the neurons in culture to lose theirprocesses. Thus, we monitored cultured neurons transfected withequal amounts of anti-hnRNP A1 rabbit polyclonal antibodies or rabbitIgG for one week. We measured the loss of processes every two daysfollowing transfection. As seen in Fig. 2E, one week followingtransfection, hnRNP A1 transfected neurons were severely injuredas shown by the loss of processes and distortion of the cytoplasm.Quantitative assessment showed an increase in the loss of neuronalprocess length over time (Fig. 2F).

3.3. Anti-hnRNP A1-M9 antibodies alter mRNA levels in a neuronal cellline

Following the transfection of neurons with antibodies, RNA wasisolated and used for microarray analyses (see Methods andsupplement 3). Since the M9 epitope is contained within a criticalfunctional domain of hnRNP A1, we hypothesized that antibodies toM9 would alter mRNA transcript levels specifically related to hnRNPA1 function and that correlate with the neural systems damaged inprogressive MS and HAM/TSP. We investigated the functional signifi-cance of the gene expression changes associated with anti-hnRNPA1-M9 antibody transfection of neurons using a dual bioinformaticsapproach: Gene Ontology and GeneIndexer. First, the 866 genes thatwere significantly altered by the anti-M9 antibodies were functionallyclassified by Gene Ontology annotation using GoTree Machine(bioinfo.vanderbilt.edu/webgestalt/) (Supplement 4, which showsall of the functional groups of genes that were significantly altered).Remarkably, the anti-M9 antibodies affected almost all aspects ofhnRNP A1's role in nucleocytoplasmic transport andmRNA processing(Table 1). The exact mechanism of hnRNP A1-M9 nucleocytoplasmictransport is under continuous investigation and recent data indicatesa role for the following processes (Table 1, Fig. 3). Small GTPases,which are required for hnRNP A1 import into and export out of thenucleus (Cook et al., 2007; Lee et al., 2006a), were affected in both thebiological processes and molecular function categories. Specifically,small GTPase mediated signal transduction (20 genes, P=0.00532),small GTPase regulator activity (12 genes, P=0.00617), small GTPasebinding (6 genes, P=0.00972) and Ran GTPase binding genes (3genes, P=0.000913)were all significantly altered. Geneswithin thesecategories, including Ran binding proteins (RBPs), guanine nucleotideexchange factors (GEFs) and importins (also known as β-karyopher-ins), have profound effects on hnRNP A1 function and transport (Contiet al., 2006; Cook et al., 2007; Rebane et al., 2004; Stewart, 2007). Thebiological processes category also indicated significant changes inmRNA processing (14 genes, P=0.00870). Genes in this categoryinfluence mRNA including splicing and polyadenylation, processeswhich are associated with hnRNP A1 function (Dreyfuss et al., 2002).In keeping with hnRNP A1 function in nucleocytoplasmic transport ofmRNA, the anti-M9 antibodies affected genes related to the nucleusand its components. In the cellular component category, the ‘nucleus’(161 genes, P=0.00000896) (Supplement 4), the ‘nuclear part’ (38genes, P=0.000351), the ‘nuclear envelope’ and its subdivisions(9 genes, P=0.00656) and the ‘nucleoplasm’ (20 genes, P=0.00158)all showed significant changes. Like the other categories, genes inthese categories also included RBPs, nucleoporins and importins(Conti et al., 2006; Cook et al., 2007; Lee et al., 2006a; Stewart, 2007).These data indicate the anti-M9 antibodies specifically affected the

Fig. 2. Neurodegeneration produced by transfection of anti-M9 antibodies into NT-2 neurons. A. Nuclei were stained with Hoechst stain (blue) and neuronal processes withmitotracker—(a mitochondria maker, red). Anti-M9 (left panel) and anti-KLH (middle panel) antibodies were detected by immunofluorescence (green) and localized withinneurons. The ‘untouched’ technical control (right panel) did not contain antibodies. B. NT-2 neuronal processes (green) stained with neuron beta-tubulin III (NTB3), a neuronalspecific marker (left panel). Nuclei were stained with Hoechst stain (blue). C. Western blotting showed immunoreactivity with NTB3 but not glial acidic fibrillary protein (GFAP—anastrocyte marker, molecular weight=50 kDa), confirming the neuronal phenotype of NT-2 cells. D. NT-2 cells stained for fluorescently labeled anti-M9 antibodies (left panel, red)and Fluoro-Jade C, a marker for neurodegeneration (middle panel, green). The merged image shows double labeling of the cells indicating that anti-M9 antibodies causedneurodegeneration (right panel, yellow) (arrowheads show examples). E. One week following the transfection of anti-M9 antibodies, neurons showed evidence of severe neuronalinjury including loss of processes and shrinkage of the cytoplasm (upper panels, middle image). Transfection of control antibodies (upper panels, right image) and the ‘untouched’control (upper panels, left image) showed normal neuronal morphology. Antibodies are present within neurons in the anti-hnRNP A1 and control antibody transfected neurons(arrowheads). F. Loss of processes (in length) in cultured neurons. At each time point, the average length of ten neuronal processes were measured and compared to themean lengthof processes at time zero. Processes loss (PL) was calculated as the mean length at time zero (before transfection) minus the mean length at each time point. Thus, higher values inthis graph indicate greater loss of neuronal process length.


function of hnRNP A1, confirming the specificity and biologicalactivity of the anti-hnRNP A1 antibodies in this system.

The second bioinformatics approachwe utilized to gain insights intothe molecular and cellular mechanisms affected by anti-M9 antibodytransfection, relied on a recently developed text-mining system calledGeneIndexer (www.computablegenomix.com). GeneIndexer identifiesgene-to-gene and gene-to-keyword associations contained withinMedline abstracts (Homayouni et al., 2005). In our experience,GeneIndexer gives results that are not revealed when using Gene

Ontology alone. In addition, GeneIndexer is not limited to a defined setof classifications designed and categorized by humans (Homayouniet al., 2005) and allows the user to define the query terms. We usedGeneIndexer to identify genes (from the same 866 gene microarraydata set) that correlate with the neural systems affected and the clinicalsymptoms present in progressive MS and HAM/TSP patients. We foundthat genes associated with the terms ‘paraplegia’, ‘spastic’, ‘weakness’,‘motor’, and ‘sensory’ were significantly enriched in our microarraydata set. To calculate the enrichment P-value, we randomly selected 24

http://www.computablegenomix.com

image of Fig.�2

Table 1Gene Ontology output: statistically significant functional gene groups related to hnRNPA1 function. The table lists functional categories related to hnRNP A1 function, whichwere significantly (pb0.05) enriched in the microarray gene set. (See supplement 3 forthe complete set of genes. Box number refers to functional gene groups presented insupplement 4).

Gene ID Symbol Name

I. Biologic process

Small GTPase mediated signal transduction(20 genes: P=0.00532) (box 6)83871 RAB34 Member RAS oncogene fam396 ARHGDIA Rho GDP dissoc inhib (GDI)α10160 FARP1 FERM, (ARHGEF) and pleckstrin domain prot.19648 GCC2 GRIP and coiled-coil domain 210966 RAB40B RAB40B, member RAS oncogene5865 RAB3B RAB3B, member RAS oncogene9181 ARHGEF2 rho/rac guan nucleotide exchng fac27204 TRIO Triple functional domain5900 RALGDS ral guanine nucleotide dissoc stim9266 PSCD2 Pleckstrin homology coiled-coil dom28498 RANBP3 RAN binding protein 36655 SOS2 Son of sevenless homolog 21399 CRKL v-crk sarcoma virus CT10 oncogene22800 RRAS2 Related RAS viral (r-ras) oncogene 25912 RAP2B RAP2B, member of RAS oncogene2885 GRB2 Growth factor receptor-bound prot 29927 MFN2 Mitofusin 2392 ARHGA1 Rho GTPase activating protein 19462 RASAL2 RAS protein activator like 251560 RAB6B RAB6B, member RAS oncogene

mRNA processing (14 genes: P=0.00870) (box 15)23451 SF3B1 Splicing factor 3b, subunit 11653 DDX1 DEAD (Asp-Glu-Ala-Asp) 122916 NCBP2 Nuclear cap binding protein sub 251362 CDC40 Cell division cycle 40 homo7884 SLBP Stem-loop binding protein3189 HNRPH3 Hetero nuclear ribonucleoprotein H39584 RBM39 RNA binding motif protein 3910492 SYNCRIP Synaptotagmin bind cyto RNA prot9169 SFRS2IP Splicing factor, arg/ser-rich 2 intr prot10658 CUGBP1 CUG triplet repeat, RNA bind prot 110914 PAPOLA Poly(A) polymerase alpha11051 NUDT21 Nudix (type motif 21)79622 C16orf33 Chrom 16 open reading frame 33140890 SFRS12 Splicing factor, arg/ ser-rich 12

II. Molecular function

Ran GTPase binding (3 genes: P=0.00091) (box 22)5903 RANBP2 RAN binding protein 210526 IPO8 Importin 88498 RANBP3 RAN binding protein 3

Small GTPase binding (6 genes: P=0.00971) (box 21)5903 RANBP2 RAN binding protein 210526 IPO8 Importin 89181 ARHGEF2 rho/rac guanine nucleotide exchange factor (GEF) 223062 GGA2 Golgi associated, gamma adaptin ARF protein 28498 RANBP3 RAN binding protein 354843 SYTL2 Synaptotagmin-like 2

Small GTPase regulator activity(12 genes: P=0.00617)(box 20)396 ARHGDIA Rho GDP dissociation inhibitor (GDI)α10160 FARP1 FERM, RhoGEF (ARHGEF) and pleckstrin domain prot 19648 GCC2 GRIP and coiled-coil 29448 MAP4K4 Mitogen-activated prot kin 49181 ARHGEF2 rho/rac guanine nucleotide exchange factor (GEF) 27070 THY1 Thy-1 cell surface antigen7204 TRIO Triple functional domain (PTPRF inter)5900 RALGDS ral guanine nucleotide dissociation stimulator9266 PSCD2 Pleckstrin homology, Sec7 and coiled-coil domains 26655 SOS2 Son of sevenless homolog 2392 ARHGAP1 Rho GTPase activating prot 19462 RASAL2 RAS protein activator like 2

Table 1 (continued)

Gene ID Symbol Name

III. Cellular Component

Nucleoplasm (20 genes; P=0.00158) (box 39)1736 DKC1 Dyskeratosis cong 1, dyskerin5970 RELA v-rel reticuloendotheliosis viral oncogene homolog A,5431 POLR2B Polymerase (RNA) II (DNA directed) polypeptide B,10519 CIB1 Ca+2-integrin bind 1 calmyrin10445 MCRS1 Microspherule protein 19443 CRSP9 Cofactor required for Sp1 transcript activation, sub 9311 ANXA11 Annexin A116595 SMARCA2 SWI/SNF matrix associate actin dependent reg chrom 210622 POLR3G Polymerase (RNA) III (DNA directed) polypeptide G9584 RBM39 RNA binding motif protein 396882 TAF11 TAF11 RNA polymerase II, TATA box binding protein9169 SFRS2IP Splicing factor, arg/ser-rich 2, interacting protein10923 SUB1 SUB1 homolog (S. cerevisiae)4673 NAP1L1 Nucleosome asm prot 1-like 111051 NUDT21 Nudix (type motif 21)4801 NFYB Nuclear transcript fact Y, beta80155 NARG1 NMDA receptor regulated 111201 POLI Polymerase (DNA dir) iota57508 INTS2 Integrator complex subunit 264426 SUDS3 Suppr of defective silencing 3

Nuclear envelope (9 genes: P=0.00656), nuclear membrane (7 genes:P=0.00656), nuclear membrane part (7 genes: P=0.00544)(boxes 37, 38, and 40)5903 RANBP2 RAN binding protein 24001 LMNB1 Lamin B110526 IPO8 Importin 8311 ANXA11 Annexin A115663 PSEN1 Presenilin 17112 TMPO Thymopoietin8498 RANBP3 RAN binding protein 38021 NUP214 Nucleoporin 214 kDa54884 RETSAT Retinol saturase


terms from the GeneIndexer dictionary and queried the subset of 866genes (570 of which had abstracts in GeneIndexer), which weredifferentially expressed in the microarray experiments. A Fisher's exacttestwas applied to compare thenumber of genes above a cosinevalue of0.1 for each query against the average number of genes in the randomquery set. This produced a P-value for the genes that were highlyassociated with each query in Medline abstracts. In total, GeneIndexeridentified 269 genes that were highly associated with the five queryterms (Supplement 5). Therewere 61 genes associatedwith ‘paraplegia’(Pb2.33×10−8), 54 genes associated with ‘spastic’ (Pb8.79×10−7), 45with ‘weakness’ (Pb6.65×10−5), 79with ‘motor’ (Pb8.90×10−13), and32with ‘sensory’ (Pb1.40×10−2) (Supplement 5). Tomaintain thedataset at a reasonable size, we analyzed the top 30 genes from each list(Table 2). Notably, many of the same genes were associated with morethan one query term, suggesting a strong correlation between theclinical terms and the genes identified by GeneIndexer. Next, we askedwhichgeneswere commonbetween at least twoof thefivequery terms.This resulted in a list of 38genes (Fig. 4, Supplement 6). Each coloredboxrepresents whether the query term identified the gene. Four sets ofgenes were clustered. Remarkably, GeneIndexer identified highlyintegrated networks of genes that correlate strongly with neurodegen-eration of the neural systems damaged, and the clinical phenotypeexpressed, by patients with progressive MS and HAM/TSP (Dutta andTrapp, 2007; Lassmann et al., 2007; Lee et al., 2005; Noseworthy et al.,2000). In fact, upon further literature review, we found that 35 of the 38genes contributed to processes linked to neurodegeneration or whenmutated result in abnormal neuronal or axonal function (Fig. 4,Supplement 6). The three genes that did not correlate with theseprocesses (fam13a1, pom121c and 441253) are novel proteinsassociated with either the nuclear pore or contain NLSs (Cohen et al.,2004). Considering the function of hnRNP A1, this again indicates a highdegree of specificity of the anti-M9 antibodies.

Fig. 3. Genes identified by Gene Ontology (GO) related to hnRNP A1 function. Gene categories and some individual genes affected by the anti-M9 antibodies are shown in red type.TheM9 sequence (AA268–305) is bordered by the black lines within hnRNP A1. Amino acids contained within the anti-M9 epitope are not bound to β-karyopherins, thus it is availablefor antibody binding. Anti-M9 antibodies affected the function of many genes directly related to hnRNP A1 function. (Ran in its GTP form is shown as the bright green circle).

Table 2GeneIndexer output: the top 30 genes with the highest cosine values in each clinical category.

Paraplegia Spastic Weakness Motor Sensory(Pb2.33×10−8) (Pb8.79×10–7) (Pb6.65×10–5) (Pb8.90×10–13) (Pb1.40×10–2)

Gene symbol Gene ID Gene symbol Gene ID Gene symbol Gene ID Gene symbol Gene ID Gene symbol Gene ID

spg 20 23111 spg 20 23111 lmna 4000 kif5c 3800 dpysl5 56896spg 4 6683 spg 4 6683 mtmr1 8776 flj20366 55638 etv1 2115spg 7 6687 spg7 6687 lama2 3908 spg20 23111 scn3a 6328dym 54808 psen 1 5663 spg20 23111 mgc3248 84516 unc5b 219699atp 11a 23250 dym 54808 itga7 3679 spg4 6683 myt1l 23040creld 1 48987 presen 55851 banf1 8815 slac2-b 23086 cutl2 23316fam 13a1 10144 creld 1 78987 ak1 203 spg7 6687 sema4f 10505ofd 1 8481 ofd 1 8481 lmnb1 4001 pnma1 9240 runx3 864sema 4f 10505 atp 11a 23250 wbscr20c 260294 neugrin 51335 spg20 23111alms 1 7840 alms 1 7840 obscn 84033 myo5b 4645 ulk2 9706pln 5350 sema4f 10505 tpm2 7169 usp14 9097 neugrin 51335loc340318 340318 mtch 1 23787 dym 54808 sytl2 54843 znrf1 84937loc441253 441253 fam 13a1 10144 spg4 6683 atxn3 4287 loc90410 90410psen 1 5663 mid 2 11043 cugbp1 10658 loc90410 90410 nbl1 4681mid 2 11043 abcd 1 215 surf1 6834 sept2 4735 nrg1 3084dgcr 6 8214 loc340318 340318 slc6a8 6535 rabgap1 23637 ablim1 3983abcd 1 215 loc441253 441253 trim63 84676 syncrip 10492 kif5c 3800mid1ip 1 58526 wbscr 20c 260294 alms1 7840 ulk2 9706 gabrr1 2569surf1 6834 dgcr 6 8214 spg7 6687 pldn 26258 ehd4 30844wbscr 20c 260294 surf 1 6834 mocs3 27304 mid1ip1 58526 olfm1 10439atxn 3 4287 ilk 2 9706 atxn3 4287 nrg1 3084 plxna3 55558ulk2 9706 rp 42 54165 ndufs5 4725 bcan 63827 spg7 6687sqstm1 8878 pnma 2 9240 creld1 78987 slc6a8 6535 cobll1 22837arsd 414 atxn 3 4287 slc25a13 10165 cutl2 23316 pou3f3 5455psenen 55851 mid 1ip 1 58526 trub1 142940 bicd1 636 catsper2 117155ahi 1 54806 sqstm 1 8878 usp14 9097 etv1 2115 ttc10 8100slc6a 8 5635 arsd 414 pnma1 9240 esrrbl1 55081 spg4 6683rab 22a 57403 slc6a8 6535 pom121 9883 ttc10 8100 alms1 7840mtch 1 23787 ahi 1 54806 tmod1 7111 myt1l 23040 tmem4 10330rabep 1 9135 slc25a 13 10165 flna 2316 akap9 10142 slc12a7 10723


image of Fig.�3

Fig. 4. Clustering of genes identified by GeneIndexer. Using GeneIndexer, four clustersof genes were identified (color coded), that were present in at least two clinicalcategories listed in Table 2. Each box indicates that the search term identified the gene.


Quantitative Real-time PCR was used to confirm differences in RNAexpression in anti-M9 compared to anti-KLH transfected NT-2 neurons.In this experiment, we chose seven genes, four from the GeneOntology(RanBP3,RanBP2, IPO8, andhnRNPH3) and three from theGeneIndexer(SPG 20, SPG 4, and SPG 7) data sets (Fig. 5). These data confirmed themicroarray analyses. Specifically, in NT-2 neurons, changes in geneexpression were similar when comparing microarray (Fig. 5A, ‘foldchange’derived fromTables 1 and Supplement 3) to qPCRdata (Fig. 5B).This is particularly evident for the SPG genes (the first three columns ofeach graph). These results indicate that gene expression related toantibody transfection of neurons is reliable because: (1) themicroarraydata showed statistical significance, (2) biological systems that wereaffected were related to hnRNP A1 function and the neural systemsdamaged in MS and HAM/TSP and, (3) real-time qPCR validated thechanges in microarray gene expression, which were well within thepublished limits of the technology (Irizarry et al., 2005).

3.4. Correlation of gene expression in neurons and human CNS tissues

Since the previous experiments were performed in NT-2 neuronalcell lines, it is important to test whether any of the observed transcriptchangesmay also be present in humans. Therefore, we tested if the geneexpression in brain tissues from progressive MS and HAM/TSP patientsparalleled the in vitro data. First, the expression of the genes in corticaltissue derived fromMS andHAM/TSP patients was compared to normalbrain. Similar to the NT-2 neurons, there was decreased expression ofalmost all of themembersof the gene set (Fig. 5C). Because thedissected

tissues were heterogeneous, we also isolated neurons from cortex, andthe relative expression of the target genes in cortex were compared toneurons isolated from the same tissue. These experiments showed thatthe target genes were enriched in human neurons from approximatelyfrom 4 to 21 fold (Fig. 5D). Importantly, neurons from MS patientsshowed decreased expression of the gene set (Fig. 5E), similar to that ofdissected tissues and NT-2 neurons. Neuronal purity was tested byWestern blot in which immunoreactivity was positive for NBT3 andnegative for GFAP (Fig. 5E, inset).

3.5. A potential molecular interaction between spastin (SPG4) andhnRNP A1

Mutations in SPG4 result in hereditary spastic paraparesis (HSP), aneurodegenerative disease whose clinical phenotype is indistinguish-able from some forms of progressive MS and HAM/TSP. Consideringthe strong correlation between HSP and our patients, perhaps themost significant discovery of our study is that SPG4 transcript(‘spastin’) was decreased after anti-M9 antibody transfection. Thissuggests that a close association exists between hnRNP A1 andspastin. To test this hypothesis, we performed the followingexperiments. First, using immunohistochemistry, we found that liketransportin (which binds hnRNP A1 (Cook et al., 2007; Lee et al.,2006a)), spastin co-localized in the nuclei of neurons (Fig. 6A). Thissuggests that a protein:protein interaction may exist between hnRNPA1 and spastin. To test this hypothesis, we transfected neurons withrecombinant hnRNP A1 tagged with Strep Tag II and tested the abilityof the recombinant hnRNP A1 to bind spastin. As shown in Fig. 6B,Western blots of the whole cell neuronal lysates (Lane 1—‘WCL’),showed a signal with antibodies to hnRNP A1 and spastin, confirmingthe presence of these proteins in the lysates. Next, the neuronallysates containing recombinant hnRNP A1 were incubated with anti-Strep II antibodies. The anti-Strep II antibodies bind recombinanthnRNP A1, which consists of a complex with any proteins bound to it.The antibody:recombinant hnRNP A1 complex was eluted from thebeads and tested for immunoreactivity with anti-spastin antibodies.As shown in Fig. 6B—lane 2 (‘Strep Tag II’), anti-spastin antibodiesreacted with the hnRNP A1 complex, indicative of spastin proteinbeing bound within the complex. Of note, the anti-spastin signal isenriched in lane 2 compared to lane 1, consistent with an increasedconcentration of spastin bound to hnRNP A1. There was only minornon-specific signal when performing the identical experiments usinga control anti-rabbit IgG (Fig. 6B—lane 3 (‘rIgG’)).

4. Discussion

These data suggest a link may exist between autoimmunity andneurodegeneration with a strong clinical–pathological correlationwith progressive MS and HAM/TSP. Specifically, we showed that in aneuronal cell line, anti-M9 antibodies targeted hnRNP A1 and alteredtranscripts related to its function. By using clinically relevant searchterms, the anti-M9 antibodies also identified sets of genes thatcorrelate strongly with the clinical phenotype of our patients andmechanisms of neurodegeneration. Finally, our data suggest there is amolecular interaction between spastin and hnRNPA1. Taken together,this implicates the potential for an anti-M9 immune response tocontribute to the expression of clinical phenotype and neurodegen-eration in MS and HAM/TSP.

How antibodies contribute to the pathogenesis of MS and HAM/TSP and whether the transfectionmodel we utilized reflects an in vivoautoimmune reaction to neurons require further study. We utilizedNT-2 neurons because following treatment with retinoic acid andmitotic inhibitors, they express a stable neuronal phenotype thatincludes neuronal processes as well as expression of neuronalcytoskeletal, secretory and surface markers (Pleasure et al., 1992).Importantly, there are several lines of evidence that suggest that

image of Fig.�4

Fig. 5. Validation of transcript changes in neurons and human tissues. A.Changes in transcript levels affected by anti-M9 antibodies compared to anti-KLH antibodies and untouched controlsin the microarray experiments (data derived from Table 1 and Supplement 6). B. Similar changes are noted when using real-time qPCR. C. Changes in gene expression in MS and HAM/TSP(TSP) cortex compared to normal cortex closely paralleled changes in NT-2 neurons. D. Neurons (right) isolated fromhuman cortex (left) showed enrichment of the target genes. E. Changesin gene expression were decreased in neurons isolated from MS brain (right) compared to control neurons (left) and closely paralleled changes in MS cortex and NT-2 neurons. Neuronalpurity was confirmed by western blot (inset) in which NTB3 immunoreactivity was positive, whilst GFAP was negative (expected molecular weight 50 kDa) (Dewji et al., 1990).


image of Fig.�5

Fig. 6. Molecular association between hnRNP A1 and spastin (SPG4). A. Immunohistochemisty of a neuronal cell line. Upper panels show staining of transportin (FITC—green) andhnRNP A1 (texas red). Themerged images show co-localization of the two signals within the nucleus. Lower panels show staining of spastin (FITC—green) and hnRNP A1 (texas red).The merged images show co-localization of the two signals within the nucleus. Nuclei are stained with DAPI and are not included in the merged images. B. An interaction betweenhnRNP A1 and SPG4 (spastin) proteins. Neuronal cells were transfected with recombinant hnRNP A1 tagged with Strep Tag II and the resultant lysate was tested for the presence ofspastin. Lane 1 (‘WCL’): the whole cell neuronal lysate (WCL) shows immunoreactivity with antibodies to spastin (WB: SPG4) and hnRNP A1 (WB: hnRNP A1), confirming theirpresence in the cell line. Lane 2 (‘Strep Tag II’): TheWCLwas incubatedwith anti-Strep Tag II antibodies attached to agarose beads, the complexwas eluted from the beads and tested forimmunoreactivity with anti-spastin antibodies. There was a strong signal with the anti-spastin antibodies (WB: SPG 4) indicative of spastin being bound to hnRNP A1. There was nochange in signalwith the anti-hnRNPA1 antibodies. Lane 3 (rIgG, anti-rabbit IgG):As a control, therewas only aminimal non-specific signalwhenperforming the identical experiment,using anti-rabbit IgG bound to the agarose beads.


intranuclear and intracellular hnRNPs are important antigens for anautoimmune response. For example, in systemic lupus erythematosus(SLE), patients develop autoantibodies to nuclear antigens includingheterogeneous nuclear ribonuclear protein P2 (hnRNP P2) (Radicet al., 2006). Upon apoptosis of target cells, hnRNP P2 is transported tothe cell surface where it is available to generate an autoimmuneresponse (Radic et al., 2006). Significantly, autoantibodies purifiedfrom the autoimmune TAM−/− mouse reacted with hnRNP P2 on thesurface of cell membranes (Radic et al., 2006). Similar data wasreported in the brain of MS patients andmice with cuprizone-induceddemyelination (Kim et al., 2010). In this study, histone deacetylase 1(HDAC1), a nuclear transcription enzyme, was transported fromnucleus to cytoplasm upon TNF-α exposure (Kim et al., 2010).Interestingly, HDAC was detected in the cytosol of axons in MS brain(Kim et al., 2010). HDAC nuclear-cytoplasmic transport was mediatedby CRM-1 (exportin-1), which bound the nuclear export sequence(NES) of HDAC1 (Kim et al., 2010). This parallels our data, since M9 isthe NES of hnRNP A1, and hnRNP A1 is transported from nucleus tocytoplasm upon binding of M9 to transportin (Lee et al., 2006a). Likeexportin-1, transportin is in the β-karphyopherin family of nucleartransport proteins (Lee et al., 2006a). Furthermore, antibodies to RNPsand DNA have been shown to enter viable target cells (Ma et al., 1991;Ruiz-Arguelles and Alarcon-Segovia, 2001; Yanase andMadaio, 2005).This may also be true for the neurologic paraneoplastic syndrome inwhich antibodies to intracellular antigens such as anti-Hu, anti-recoverin and anti-amphiphysin may enter neurons (Adamus et al.,

1998; Geis et al., 2010; Ruiz-Arguelles and Alarcon-Segovia, 2001).This may occur because of the presence of the nuclear antigen on thecell surface or because autoantibodies contain a nuclear localization-like sequence that allows for their entry into the cytoplasm andnucleus (Ruiz-Arguelles and Alarcon-Segovia, 2001; Yanase andMadaio, 2005). Alternatively, following an initial cellular insult inwhich previously sequestered intracellular antigens are exposed tothe systemic immune response, autoantibodies may enter cells andtarget intracellular antigens (El-Fawal et al., 1999; Ruiz-Arguelles andAlarcon-Segovia, 2001). Importantly, recent data supports thehypothesis that antibodies to intra-neuronal targets are not onlymarkers of disease but may also contribute to the pathogenesis of thedisease. For example, antibodies to the intra-neuronal proteinamphiphysin (which is associated with stiff man syndrome) causeda stiff man-like syndrome in rats and were internalized into neuronsby an epitope-specific mechanism where they co-localized withsynaptic vesicles (Geis et al., 2010). Some studies suggest thatcompared to intra-neuronal antigens, only antibodies to the surfaceof neurons are pathogenic (Vincent, 2008, 2010). This remainscontroversial. The studies presented here may help to bridge thecontroversy, since under inflammatory conditions (such as TNF-αexpression), intra-cellular target proteins can be transported from thenucleus to cytoplasm and cell surface, thus being available for animmune response to develop (Kim et al., 2010; Radic et al., 2006).

In MS and HAM/TSP, data indicates that the initiation of diseaseincludes activation of T-lymphocytes, disruption of the blood brain

image of Fig.�6


barrier and T-lymphocyte driven immune attack of target CNSantigens (Frohman et al., 2006; Jacobson et al., 1990; Kawamuraet al., 2008; Levin and Jacobson, 1997; Ratts et al., 2006; Trapp andNave, 2008). In MS, the initial targets are likely myelin antigens(Frohman et al., 2006; Kawamura et al., 2008; Ratts et al., 2006).Interestingly, CNS targets in HAM/TSP are largely unknown, but mayinclude neuronal antigens (Lee and Levin, 2008; Lee et al., 2005; Levinet al., 2002a). Axonal injury is present in relapsing–remitting andprogressive forms of MS as well as in HAM/TSP (Geurts and Barkhof,2008; Kutzelnigg et al., 2005; Trapp and Nave, 2008; Umehara et al.,2000). Several studies suggest axonal damage is the result of twophases of immune-mediated attack of the CNS (Aboul-Enein et al.,2006; Bjartmar et al., 2003; Brown and Sawchenko, 2007; Dutta andTrapp, 2007; Gold et al., 2006; Kornek et al., 2000; Lassmann, 2007;Lassmann et al., 2007). The initial (relapsing–remitting) phasemay becaused by the acute inflammatory response associated with plaqueformation and demyelination (known as the ‘plaque-centered’ view ofMS) (Aboul-Enein et al., 2006; Bjartmar et al., 2003; Brown andSawchenko, 2007; Dutta and Trapp, 2007; Geurts and Barkhof, 2008;Gold et al., 2006; Kornek et al., 2000; Kutzelnigg et al., 2005;Lassmann, 2007; Lassmann et al., 2007; Trapp and Nave, 2008). Inprogressive MS, axonal damage occurs in association with a diffuseleptomeningeal and intra-parenchymal inflammatory response notrelated to plaque location (Kutzelnigg et al., 2005). This is particularlyrelevant when studying the long tracts (corticospinal and posteriorcolumns) of the CNS, which show axonal damage independent ofplaque location and whose damage correlate strongly with progres-sive forms of MS (DeLuca et al., 2004, 2006; Evangelou et al., 2005).The cause of this phase of axonal injury is largely unknown, and ishypothesized to be the result of a lack of tropic support from myelin(Bjartmar et al., 2003), abnormal ion and sodium channel expressionon axons (Dutta and Trapp, 2007; Trapp and Nave, 2008), calciumaccumulation in damaged axons, or Wallerian degeneration (Duttaand Trapp, 2007; Trapp and Nave, 2008). Similar findings are presentin HAM/TSP (Umehara et al., 2000). One possible result of theseproposed mechanisms of axonal and neuronal injury could be thedevelopment of a highly specific antibody response directed atpreviously sequestered CNS antigens including hnRNP A1. As anti-M9 antibodies develop to injured neurons and bind hnRNP A1, theymay disrupt its function, resulting in neuronal dysfunction andneurodegeneration. Notably, neurodegeneration has been found inboth early stages of relapsing–remitting disease and during progres-sive MS (Trapp et al., 1998; Trapp and Nave, 2008), thus hnRNP A1may be available for binding throughout the life of a MS patient.Antibodies to hnRNP A1 are unlikely to initiate disease, but maycontribute to a diffuse neurodegenerative process due to a persistentimmune response to neuronal hnRNP A1.

Several studies support the hypothesis that antibodies contributeto the pathogenesis of MS (Abramsky et al., 1977; Edgington andDalessio, 1970; Franciotta et al., 2008; Genain et al., 1999; Meinl et al.,2006; Racke, 2008; Silber et al., 2002). MS patients develop antibodiesis serum and CSF to other neuronal and axonal antigens such asneurofilaments and gangliosides (Franciotta et al., 2008; Lily et al.,2004; Meinl et al., 2006; Norgren et al., 2005; Rawes et al., 1997;Sadatipour et al., 1998; Silber et al., 2002). The data presented heresuggest that anti-hnRNP-A1-M9 antibodies directed at neuronscontribute to neuronal and axonal injury. hnRNP A1 has beendescribed as being expressed ubiquitously. Our data indicate anautoimmune response specific to hnRNP A1 expressed in humanneurons (Jernigan et al., 2003; Lee and Levin, 2008; Levin et al., 1998,2002a, 2005). This is consistent with other studies that indicate thathnRNP A1 is preferentially expressed in brain (including largeneurons) compared to systemic organs (Clower et al., 2010;Hanamura et al., 1998; Kamma et al., 1995). It is the antibodyresponse that has been tested in our lab and found to contribute to thepathogenesis of HAM/TSP via molecular mimicry to neurons (Kalume

et al., 2004; Lee and Levin, 2008; Lee et al., 2006b, 2005; Levin et al.,1998, 2002a, 2002b). HAM/TSP IgG was used to isolate hnRNP A1,which cross-reacted with HTLV-1-tax (a regulatory and immunodo-minant protein) (Levin et al., 1998, 2002a, 2002b). The immunereaction to tax and hnRNP A1 included functionally important regionsof the proteins, a requirement for molecular mimicry (Lee et al.,2006b; Levin et al., 1998, 2002a, 2002b; Oldstone, 1998). For hnRNPA1 this included M9. MS patients also make antibodies to hnRNP A1(Sueoka et al., 2004), and as shown here, to M9. The M9 sequenceconsists of hnRNP A1268–305, which binds to transportin in the lowerarch at hnRNP A1263–289 (Fig. 3) (Lee et al., 2006a). Antibodies fromour patients bind M9293–304 (Lee et al., 2006b), thus the epitope isavailable for binding since this sequence is not directly bound totransportin. hnRNP A1 plays a critical role in mRNA transport andmetabolism. We altered dozens of transcripts directly related tohnRNP A1, suggesting that the anti-M9 immune response alteredhnRNPA1 function.

In addition to altering transcripts related to hnRNP A1 function,search queries related to the clinical phenotype of our patientsresulted in groups of genes with high correlationswithmechanisms ofneurodegeneration. The spinal paraplegia genes (SPGs) exemplify thecontribution of genes specifically related to mechanisms of neurode-generation (De Vos et al., 2008). The SPGs had the highestGeneIndexer cosine values and were associated with all of the queryterms. Mutations in this family of genes result in HSP, a group ofneurodegenerative diseases characterized by progressive spasticity,weakness of the lower extremities (‘paraplegia’) and sensorysymptoms including difficulty with detecting vibration (Salinaset al., 2008; Soderblom and Blackstone, 2006). These symptomsmay be clinically indistinguishable from progressive MS and HAM/TSP. The most common HSP is SPG 4. Its encoded protein is spastin(Roll-Mecak and Vale, 2008; Salinas et al., 2008). Spastin is involved inmicrotubule regulation, proteosome and endosome function, andsynaptic transmission in neurons (Hazan et al., 1999; Roll-Mecak andVale, 2008; Salinas et al., 2008; Trotta et al., 2004). In addition to themicroarray data which showed down-regulation of spastin mRNA, wefound decreased levels of spastin RNA in MS brains and neurons.Further, isolation of neuronal proteins with anti-hnRNP A1 antibodiesrevealed a molecular interaction between hnRNP A1 and spastin atthe protein level. Previous genetic analysis of MS patients showed noincreased DNA mutations in SPGs (DeLuca et al., 2007). These resultsindicate that the expression of this clinical phenotype (spasticparaparesis) is not related to a genetic predisposition, but rather islikely due to mechanisms affecting RNA or protein function, thussuggesting the anti-M9 immune response may contribute to neuronalinjury and expression of the clinical phenotype. The mechanisms bywhich anti-M9 antibodies cause changes in function or expression ofspastin are not yet know, however several possibilities exist. LikehnRNP A1, spastin is translocated through the nuclear pore via the β-karyopherin nucleocytoplasmic system (Fig. 3) (Beetz et al., 2004;Claudiani et al., 2005). hnRNP A1 uses the ‘non-classical’ (transportinmediated) pathway andM9 acts as both the NLS and NES in its importand export, respectively (Conti et al., 2006; Cook et al., 2007; Lee et al.,2006a; Rebane et al., 2004; Stewart, 2007). Spastin uses the ‘classical(β-importin mediated) pathway’ and it contains separate NLS andNES sequences within its protein structure (Beetz et al., 2004;Claudiani et al., 2005). For export, the spastin NES binds exportin-1(Beetz et al., 2004; Claudiani et al., 2005). For import into the nucleus,the NLS of spastin may bind one of the β-importins, but which one isnot known (Beetz et al., 2004; Claudiani et al., 2005). Both the non-classical and classical nucleocytoplasmic transport pathways aretightly regulated and require the binding of RanGTP to β-karyopherinto function (Fig. 3) (Conti et al., 2006; Cook et al., 2007; Lee et al.,2006a; Rebane et al., 2004; Stewart, 2007). Thus, both pathwaysare regulated by the same system. An interruption in one may affectthe other. Anti-M9 antibodies may potentially alter spastin function


and expression by a number of mechanisms. First, anti-M9 anti-bodies may cross-react with spastin NLS or NES, altering its function.Second, binding of M9 antibodies may disturb the non-classicalpathway, which in turn, may disrupt the classical pathway. Third,because of the protein–protein interaction between the hnRNP A1complex and spastin (Fig. 6), binding of anti-M9 antibodies to hnRNPA1 may sterically hinder spastin-hnRNP A1 interaction. Finally, ifspastin mRNA binds to the RNA binding domains within hnRNPA1, then anti-M9 antibodies may interfere with its transport andsubsequent translation. Further experiments are required to test thesehypotheses.

Other integrated networks of closely related genes that maintainneuronal and axonal homeostasis were also affected such as: (1) theubiquitin proteosome system (SPG 4, ATXN 3, USP 14, andWBSCR20c) which plays a role in Wallerian Degeneration (Ehlers,2004); (2) amyloid precursor protein metabolism (PSEN1, PSEN 2,and PSAP); (3) myelination (NRG 1, MYT1L, SEMA4F, ABCD1, andARSD), (4) autophagy (SQSTM1) and (5) axonal transport (ALMS1,OFD1, IFT20, IFT88, KIFC, and DYM). Future studies are needed todetermine the contribution of anti-M9 antibodies to these mecha-nisms of neurodegeneration. Our study was limited by the utilizationof a neuronal cell line, a transfection reagent was used to introduceantibodies into neurons and our clinical assays were not blinded orpowered to address MS subtype. These data set a stage for futureinvestigations designed to discover contributions of autoimmunity toneurodegeneration in MS. Importantly, our data closely aligns withmicroarray studies fromMS brains, which showed changes in neuron-specific and cellular transport genes including hnRNP A1, importins,and GTPases (Dutta et al., 2006; Lock et al., 2002).

Taken together, these data suggest we may have discovered apotential link between autoimmunity and neurodegeneration inimmune-mediated neurological disease. The target is hnRNPA1 inneurons and the immune response to it is antibodies to M9, itsshuttling domain required for nuclear translocation of mRNA. Anti-M9 antibodies identified integrated networks of genes related tohnRNP A1 function as well as to neurodegeneration and the clinicalphenotype expressed by patients with progressive MS and HAM/TSP.

Supplementarymaterials related to this article can be found onlineat doi:10.1016/j.jneuroim.2011.02.007.

Conflict of interest

Dr. Ramin Homayouni is a co-founder and holds equity inComputable Genomix, LLC, the makers of GeneIndexer. Drs. MichaelLevin and Sangmin Lee have a patent pending titled “Biomarker forneurodegeneration in neurological disease” in collaboration with theUniversity of Tennessee Health Science Center.

Acknowledgements

This work is based upon work supported by the Office of Researchand Development, Medical Research Service, Department of VeteransAffairs. This study was funded by a VA Merit Review Award (to MCL)and NIH RO1 (to RH).

References

Aboul-Enein, F., Weiser, P., Hoftberger, R., Lassmann, H., Bradl, M., 2006. Transient axonalinjury in the absence of demyelination: a correlate of clinical disease in acuteexperimental autoimmuneencephalomyelitis. ActaNeuropathol. (Berl.) 111, 539–547.

Abramsky, O., Lisak, R.P., Silberberg, D.H., Pleasure, D.E., 1977. Antibodies tooligodendroglia in patients with multiple sclerosis. N. Engl. J. Med. 297, 1207–1211.

Adamus, G., Machnicki, M., Elerding, H., Sugden, B., Blocker, Y., Fox, D., 1998. Antibodiesto recoverin induce apoptosis of photoreceptor and bipolar cells in vivo.J. Autoimmun. 11, 523–533 Oct.

Beetz, C., Brodhun, M., Moutzouris, K., Kiehntopf, M., Berndt, A., Lehnert, D., Deufel, T.,Bastmeyer, M., Schickel, J., 2004. Identification of nuclear localisation sequences in

spastin (SPG4) using a novel Tetra-GFP reporter system. Biochem. Biophys. Res.Commun. 318, 1079–1084.

Bjartmar, C., Kidd, G., Mork, S., Rudick, R., Trapp, B.D., 2000. Neurological disabilitycorrelates with spinal cord axonal loss and reduced N-acetyl aspartate in chronicmultiple sclerosis patients. Ann. Neurol. 48, 893–901.

Bjartmar, C., Wujek, J.R., Trapp, B.D., 2003. Axonal loss in the pathology of MS:consequences for understanding the progressive phase of the disease. J. Neurol. Sci.206, 165–171.

Brown, D.A., Sawchenko, P.E., 2007. Time course and distribution of inflammatory andneurodegenerative events suggest structural bases for the pathogenesis ofexperimental autoimmune encephalomyelitis. J. Comp. Neurol. 502, 236–260.

Claudiani, P., Riano, E., Errico, A., Andolfi, G., Rugarli, E.I., 2005. Spastin subcellularlocalization is regulated through usage of different translation start sites and activeexport from the nucleus. Exp. Cell Res. 309, 358–369.

Clower, C.V.0, Chatterjee, D., Wang, Z., Cantley, L.C., Vander Heiden, M.G., Krainer, A.R.,2010. The alternative splicing repressors hnRNP A1/A2 and PTB influence pyruvatekinase isoform expression and cell metabolism. Proc. Natl. Acad. Sci. U. S. A. 107,1894–1899.

Cohen,M., Reichenstein, M., Everts-van derWind, A., Heon-Lee, J., Shani, M., Lewin, H.A.,Weller, J.I., Ron, M., Seroussi, E., 2004. Cloning and characterization of FAM13A1-agene near a milk protein QTL on BTA6: evidence for population-wide linkagedisequilibrium in Israeli Holsteins. Genomics 84, 374–383.

Conti, E., Muller, C.W., Stewart, M., 2006. Karyopherin flexibility in nucleocytoplasmictransport. Curr. Opin. Struct. Biol. 16, 237–244.

Cook, A., Bono, F., Jinek, M., Conti, E., 2007. Structural biology of nucleocytoplasmictransport. Annu. Rev. Biochem. 76, 647–671.

De Vos, K.J., Grierson, A.J., Ackerley, S., Miller, C.C., 2008. Role of axonal transport inneurodegenerative diseases. Annu. Rev. Neurosci. 31, 151–173.

DeLuca, G.C., Ebers, G.C., Esiri, M.M., 2004. Axonal loss in multiple sclerosis: apathological survey of the corticospinal and sensory tracts. Brain 127, 1009–1018.

DeLuca, G.C., Williams, K., Evangelou, N., Ebers, G.C., Esiri, M.M., 2006. The contributionof demyelination to axonal loss in multiple sclerosis. Brain 129, 1507–1516.

DeLuca, G.C., Ramagopalan, S.V., Cader, M.Z., Dyment, D.A., Herrera, B.M., Orton, S.,Degenhardt, A., Pugliatti, M., Sadovnick, A.D., Sotgiu, S., Ebers, G.C., 2007. The role ofhereditary spastic paraplegia related genes in multiple sclerosis. A study of diseasesusceptibility and clinical outcome. J. Neurol. 254, 1221–1226.

Dewji, N.N., Shelton, E.R., Adler, M.J., Chan, H.W., Seegmiller, J.E., Coronel, C., 1990.Processing of Alzheimer's disease-associated beta-amyloid precursor protein.J. Mol. Neurosci. 2, 19–27.

Dreyfuss, G., Kim, V.N., Kataoka, N., 2002. Messenger-RNA-binding proteins and themessages they carry. Nat. Rev. Mol. Cell Biol. 3, 195–205.

Dutta, R., Trapp, B.D., 2007. Pathogenesis of axonal and neuronal damage in multiplesclerosis. Neurology 68, S22–S31 (discussion S43-54).

Dutta, R., McDonough, J., Yin, X., Peterson, J., Chang, A., Torres, T., Gudz, T., Macklin, W.B.,Lewis, D.A., Fox, R.J., Rudick, R., Mirnics, K., Trapp, B.D., 2006. Mitochondrialdysfunction as a cause of axonal degeneration in multiple sclerosis patients. Ann.Neurol. 59, 478–489.

Edgington, T.S., Dalessio, D.J., 1970. The assessment by immunofluorescencemethods ofhumoral anti-myelin antibodies in man. J. Immunol. 105, 248–255.

Ehlers, M.D., 2004. Deconstructing the axon:Wallerian degeneration and the ubiquitin-proteasome system. Trends Neurosci. 27, 3–6.

El-Fawal, H.A., Waterman, S.J., De Feo, A., Shamy, M.Y., 1999. Neuroimmunotoxicology:humoral assessment of neurotoxicity and autoimmune mechanisms. Environ.Health Perspect. 107 (Suppl 5), 767–775.

Evangelou, N., DeLuca, G.C., Owens, T., Esiri, M.M., 2005. Pathological study of spinalcord atrophy in multiple sclerosis suggests limited role of local lesions. Brain 128,29–34.

Ferguson, B., Matyszak, M.K., Esiri, M.M., Perry, V.H., 1997. Axonal damage in acutemultiple sclerosis lesions. Brain 120 (Pt 3), 393–399.

Franciotta, D., Salvetti, M., Lolli, F., Serafini, B., Aloisi, F., 2008. B cells and multiplesclerosis. Lancet Neurol. 7, 852–858.

Frohman, E.M., Racke, M.K., Raine, C.S., 2006. Multiple sclerosis—the plaque and itspathogenesis. N. Engl. J. Med. 354, 942–955.

Ganter, P., Prince, C., Esiri, M.M., 1999. Spinal cord axonal loss in multiple sclerosis: apost-mortem study. Neuropathol. Appl. Neurobiol. 25, 459–467.

Geis, C., Weishaupt, A., Hallermann, S., Grunewald, B., Wessig, C., Wultsch, T., Reif, A.,Byts, N., Beck, M., Jablonka, S., Boettger, M.K., Uceyler, N., Fouquet, W., Gerlach, M.,Meinck, H.M., Siren, A.L., Sigrist, S.J., Toyka, K.V., Heckmann, M., Sommer, C., 2010.Stiff person syndrome-associated autoantibodies to amphiphysin mediate reducedGABAergic inhibition. Brain 133, 3166–3180.

Genain, C.P., Cannella, B., Hauser, S.L., Raine, C.S., 1999. Identification of autoantibodiesassociated with myelin damage in multiple sclerosis. Nat. Med. 5, 170–175.

Geurts, J.J., Barkhof, F., 2008. Grey matter pathology in multiple sclerosis. Lancet Neurol.7, 841–851.

Gold, R., Linington, C., Lassmann, H., 2006. Understanding pathogenesis and therapy ofmultiple sclerosis via animal models: 70 years of merits and culprits inexperimental autoimmune encephalomyelitis research. Brain 129, 1953–1971.

Hanamura, A., Caceres, J.F.,Mayeda, A., Franza Jr., B.R., Krainer, A.R., 1998. Regulated tissue-specific expression of antagonistic pre-mRNA splicing factors. RNA 4, 430–444.

Hazan, J., Fonknechten, N., Mavel, D., Paternotte, C., Samson, D., Artiguenave, F.,Davoine, C.S., Cruaud, C., Durr, A., Wincker, P., Brottier, P., Cattolico, L., Barbe, V.,Burgunder, J.M., Prud'homme, J.F., Brice, A., Fontaine, B., Heilig, B., Weissenbach, J.,1999. Spastin, a new AAA protein, is altered in the most frequent form of autosomaldominant spastic paraplegia. Nat. Genet. 23, 296–303.

Homayouni, R., Heinrich, K., Wei, L., Berry, M.W., 2005. Gene clustering by latentsemantic indexing of MEDLINE abstracts. Bioinformatics 21, 104–115.


Irizarry, R.A., Warren, D., Spencer, F., Kim, I.F., Biswal, S., Frank, B.C., Gabrielson, E.,Garcia, J.G., Geoghegan, J., Germino, G., Griffin, C., Hilmer, S.C., Hoffman, E., Jedlicka,A.E., Kawasaki, E., Martinez-Murillo, F., Morsberger, L., Lee, H., Petersen, D.,Quackenbush, J., Scott, A., Wilson, M., Yang, Y., Ye, S.Q., Yu, W., 2005. Multiple-laboratory comparison of microarray platforms. Nat. Methods 2, 345–350.

Jacobson, S., Shida, H., McFarlin, D.E., Fauci, A.S., Koenig, S., 1990. Circulating CD8+cytotoxic T lymphocytes specific for HTLV-I pX in patients with HTLV-I associatedneurological disease. Nature 348, 245–248.

Jernigan, M., Morcos, Y., Lee, S.M., Dohan Jr., F.C., Raine, C., Levin, M.C., 2003. IgG in braincorrelates with clinicopathological damage in HTLV-1 associated neurologicdisease. Neurology 60, 1320–1327.

Kalume, F., Lee, S.M., Morcos, Y., Callaway, J.C., Levin, M.C., 2004. Molecular mimicry:cross-reactive antibodies from patients with immune-mediated neurologic diseaseinhibit neuronal firing. J. Neurosci. Res. 77, 82–89.

Kamma, H., Portman, D.S., Dreyfuss, G., 1995. Cell type-specific expression of hnRNPproteins. Exp. Cell Res. 221, 187–196.

Kawamura, K., Yao, K., Shukaliak-Quandt, J.A., Huh, J., Baig, M., Quigley, L., Ito, N.,Necker, A., McFarland, H.F., Muraro, P.A., Martin, R., Ito, K., 2008. Differentdevelopment of myelin basic protein agonist- and antagonist-specific human TCRtransgenic T cells in the thymus and periphery. J. Immunol. 181, 5462–5472.

Kim, J.Y., Shen, S., Dietz, K., He, Y., Howell, O., Reynolds, R., Casaccia, P., 2010. HDAC1nuclear export induced by pathological conditions is essential for the onset ofaxonal damage. Nat. Neurosci. 13, 180–189.

Kornek, B., Storch, M.K., Weissert, R., Wallstroem, E., Stefferl, A., Olsson, T., Linington, C.,Schmidbauer, M., Lassmann, H., 2000. Multiple sclerosis and chronic autoimmuneencephalomyelitis: a comparative quantitative study of axonal injury in active,inactive, and remyelinated lesions. Am. J. Pathol. 157, 267–276.

Kutzelnigg, A., Lucchinetti, C.F., Stadelmann, C., Bruck, W., Rauschka, H., Bergmann, M.,Schmidbauer, M., Parisi, J.E., Lassmann, H., 2005. Cortical demyelination and diffusewhite matter injury in multiple sclerosis. Brain 128, 2705–2712.

Lassmann, H., 2007. Multiple sclerosis: is there neurodegeneration independent frominflammation? J. Neurol. Sci. 259, 3–6.

Lassmann, H., Bruck, W., Lucchinetti, C.F., 2007. The immunopathology of multiplesclerosis: an overview. Brain Pathol. 17, 210–218.

Lee, S., Levin, M.C., 2008. Molecular mimicry in neurological disease: what is theevidence? Cell. Mol. Life Sci. 65, 1161–1175.

Lee, S.M., Morocos, Y., Jang, H., Stuart, J.M., Levin, M.C., 2005. HTLV-1 induced molecularmimicry in neurologic disease. In: Oldstone, M. (Ed.), Molecular Mimicry: InfectionInducing Autoimmune Disease. : Current Topics in Microbiology and Immunology.Springer, New York.

Lee, B.J., Cansizoglu, A.E., Suel, K.E., Louis, T.H., Zhang, Z., Chook, Y.M., 2006a. Rules fornuclear localization sequence recognition by karyopherin beta 2. Cell 126,543–558.

Lee, S.M., Dunnavant, F.D., Jang, H., Zunt, J., Levin, M.C., 2006b. Autoantibodies thatrecognize functional domains of hnRNPA1 implicate molecular mimicry in thepathogenesis of neurological disease. Neurosci. Lett. 401, 188–193.

Levin, M.C., Jacobson, S., 1997. HTLV-I associated myelopathy/tropical spastic parapar-esis (HAM/TSP): a chronic progressive neurologic disease associated withimmunologically mediated damage to the central nervous system. J. Neurovirol.3, 126–140.

Levin, M., Lehky, T., Flerlage, N., Katz, D., Kingma, D., Jaffe, E., Heiss, J., Patronas, N.,McFarland, H., Jacobson, S., 1997. Immunopathogenesis of HTLV-1 associatedneurologic disease based on a spinal cord biopsy from a patient with HTLV-1associated myelopathy/tropical spastic paraparesis (HAM/TSP). N. Engl. J. Med.336, 839–845.

Levin, M., Krichavsky, M., Berk, J., Foley, S., Rosenfeld, M., Dalmau, J., Chen, G., Posner, J.,Jacobson, S., 1998. Neuronal molecular mimicry in immune mediated neurologicdisease. Ann. Neurol. 44, 87–98.

Levin, M.C., Lee, S.M., Kalume, F., Morcos, Y., Dohan Jr., F.C., Hasty, K.A., Callaway, J.C.,Zunt, J., Desiderio, D., Stuart, J.M., 2002a. Autoimmunity due to molecular mimicryas a cause of neurological disease. Nat. Med. 8, 509–513.

Levin, M.C., Lee, S.M., Morcos, Y., Brady, J., Stuart, J., 2002b. Cross-reactivity betweenimmunodominant human T lymphotropic virus type I tax and neurons: implica-tions for molecular mimicry. J. Infect. Dis. 186, 1514–1517.

Levin, M.C., Lee, S.M., Morcos, Y., 2005. Autoimmunity to heterogeneous nuclearribonucleoprotein in neurological disease. Ann. Neurol. 57, 931.

Lily, O., Palace, J., Vincent, A., 2004. Serum autoantibodies to cell surface determinantsin multiple sclerosis: a flow cytometric study. Brain 127, 269–279.

Lock, C., Hermans, G., Pedotti, R., Brendolan, A., Schadt, E., Garren, H., Langer-Gould, A.,Strober, S., Cannella, B., Allard, J., Klonowski, P., Austin, A., Lad, N., Kaminski, N.,Galli, S.J., Oksenberg, J.R., Raine, C.S., Heller, R., Steinman, L., 2002. Gene-microarrayanalysis of multiple sclerosis lesions yields new targets validated in autoimmuneencephalomyelitis. Nat. Med. 8, 500–508.

Lovas, G., Szilagyi, N., Majtenyi, K., Palkovits, M., Komoly, S., 2000. Axonal changes inchronic demyelinated cervical spinal cord plaques. Brain 123 (Pt 2), 308–317.

Ma, J., Chapman, G.V., Chen, S.L., Melick, G., Penny, R., Breit, S.N., 1991. Antibodypenetration of viable human cells. I. Increased penetration of human lymphocytesby anti-RNP IgG. Clin. Exp. Immunol. 84, 83–91.

Meinl, E., Krumbholz, M., Hohlfeld, R., 2006. B lineage cells in the inflammatory centralnervous system environment: migration, maintenance, local antibody production,and therapeutic modulation. Ann. Neurol. 59, 880–892.

Michael, W.M., Choi, M., Dreyfuss, G., 1995. A nuclear export signal in hnRNP A1: a signal-mediated, temperature-dependent nuclear protein export pathway. Cell 83, 415–422.

Nichols, R.C.,Wang, X.W., Tang, J., Hamilton, B.J., High, F.A., Herschman, H.R., Rigby,W.F.,2000. The RGG domain in hnRNP A2 affects subcellular localization. Exp. Cell Res.256, 522–532.

Norgren, N., Edelstam, A., Stigbrand, T., 2005. Cerebrospinal fluid levels of neurofila-ment light in chronic experimental autoimmune encephalomyelitis. Brain Res. Bull.67, 264–268.

Noseworthy, J., Lucchinetti, C., Rodriguez, M., Weinshenker, B., 2000. Multiple sclerosis.N. Engl. J. Med. 343, 938–952.

Oldstone, M., 1998. Molecular mimicry and immune mediated disease. FASEB J. 12,1255–1265.

Pleasure, S.J., Page, C., Lee, V.M., 1992. Pure, postmitotic, polarized human neuronsderived from NTera 2 cells provide a system for expressing exogenous proteins interminally differentiated neurons. J. Neurosci. 12, 1802–1815.

Racke, M.K., 2008. The role of B cells in multiple sclerosis: rationale for B-cell-targetedtherapies. Curr. Opin. Neurol. 21 (Suppl 1), S9–S18.

Radic, M.Z., Shah, K., Zhang, W., Lu, Q., Lemke, G., Hilliard, G.M., 2006. Heterogeneousnuclear ribonucleoprotein P2 is an autoantibody target in mice deficient for Mer,Axl, and Tyro3 receptor tyrosine kinases. J. Immunol. 176, 68–74.

Ratts, R.B., Karandikar, N.J., Hussain, R.Z., Choy, J., Northrop, S.C., Lovett-Racke, A.E.,Racke, M.K., 2006. Phenotypic characterization of autoreactive T cells in multiplesclerosis. J. Neuroimmunol. 178, 100–110.

Rawes, J.A., Calabrese, V.P., Khan, O.A., DeVries, G.H., 1997. Antibodies to the axolemma-enriched fraction in the cerebrospinal fluid and serum of patients with multiplesclerosis and other neurological diseases. Mult. Scler. 3, 363–369.

Rebane, A., Aab, A., Steitz, J.A., 2004. Transportins 1 and 2 are redundant nuclear importfactors for hnRNP A1 and HuR. RNA 10, 590–599.

Roll-Mecak, A., Vale, R.D., 2008. Structural basis of microtubule severing by thehereditary spastic paraplegia protein spastin. Nature 451, 363–367.

Ruiz-Arguelles, A., Alarcon-Segovia, D., 2001. Penetration of autoantibodies into livingcells, 2000. Isr. Med. Assoc. J. 3, 121–126.

Sadatipour, B.T., Greer, J.M., Pender, M.P., 1998. Increased circulating antigangliosideantibodies in primary and secondary progressive multiple sclerosis. Ann. Neurol.44, 980–983.

Salinas, S., Proukakis, C., Crosby, A., Warner, T.T., 2008. Hereditary spastic paraplegia:clinical features and pathogenetic mechanisms. Lancet Neurol. 7, 1127–1138.

Schmued, L.C., Stowers, C.C., Scallet, A.C., Xu, L., 2005. Fluoro-Jade C results in ultra highresolution and contrast labeling of degenerating neurons. Brain Res. 1035, 24–31.

Silber, E., Semra, Y.K., Gregson, N.A., Sharief, M.K., 2002. Patients with progressive multiplesclerosis haveelevatedantibodies toneurofilament subunit.Neurology58,1372–1381.

Soderblom, C., Blackstone, C., 2006. Traffic accidents: molecular genetic insights into thepathogenesis of the hereditary spastic paraplegias. Pharmacol. Ther. 109, 42–56.

Stewart, M., 2007. Molecular mechanism of the nuclear protein import cycle. Nat. Rev.Mol. Cell Biol. 8, 195–208.

Sueoka, E., Yukitake, M., Iwanaga, K., Sueoka, N., Aihara, T., Kuroda, Y., 2004.Autoantibodies against heterogeneous nuclear ribonucleoprotein B1 in CSF of MSpatients. Ann. Neurol. 56, 778–786.

Trapp, B.D., Nave, K.A., 2008. Multiple sclerosis: an immune or neurodegenerativedisorder? Annu. Rev. Neurosci. 31, 247–269.

Trapp, B., Peterson, J., Ransohoff, R., Rudick, R., Mork, S., Bo, L., 1998. Axonal transectionin the lesions of multiple sclerosis. N. Engl. J. Med. 338, 278–285.

Trotta, N., Orso, G., Rossetto, M.G., Daga, A., Broadie, K., 2004. The hereditary spasticparaplegia gene, spastin, regulates microtubule stability to modulate synapticstructure and function. Curr. Biol. 14, 1135–1147.

Umehara, F., Abe, M., Koreeda, Y., Izumo, S., Osame, M., 2000. Axonal damage revealedby accumulation of beta-amyloid precursor protein in HTLV-I-associated myelop-athy. J. Neurol. Sci. 176, 95–101.

Vanderlugt, C., Miller, S., 1996. Epitope spreading. Autoimmunity 8, 831–836.Vincent, A., 2008. Stiff, twitchy or wobbly: are GAD antibodies pathogenic? Brain 131,

2536–2537.Vincent, A., 2010. Successful 'passive transfer' of paraneoplastic stiff person syndrome

with antibodies to an intracellular antigen. Brain 133, 3164–3165.Yanase, K., Madaio, M.P., 2005. Nuclear localizing anti-DNA antibodies enter cells via

caveoli and modulate expression of caveolin and p53. J. Autoimmun. 24, 145–151.Yuki, N., Susuki, K., Koga, M., Nishimoto, Y., Odaka, M., Hirata, K., Taguchi, K., Miyatake,

T., Furukawa, K., Kobata, T., Yamada, M., 2004. Carbohydrate mimicry betweenhuman ganglioside GM1 and Campylobacter jejuni lipooligosaccharide causesGuillain–Barre syndrome. Proc. Natl. Acad. Sci. U. S. A. 101, 11404–11409.

a potential link between autoimmunity and neurodegeneration in immune-mediated neurological disease

Documents