Selective in vivo removal of pathogenic anti-MAGautoantibodies, an antigen-specific treatment optionfor anti-MAG neuropathyRuben Herrendorffa, Pascal Hänggia, Hélène Pfistera, Fan Yanga, Delphine Demeesterea, Fabienne Hunzikera,Samuel Freya, Nicole Schaeren-Wiemersb, Andreas J. Steckb,c, and Beat Ernsta,1

aInstitute of Molecular Pharmacy, Pharmacenter, University of Basel, 4056 Basel, Switzerland; bDepartment of Biomedicine, University Hospital Basel,University of Basel, 4031 Basel, Switzerland; and cClinic of Neurology, Department of Medicine, University Hospital Basel, University of Basel, 4031 Basel,Switzerland

Edited by Gabriel A. Rabinovich, University of Buenos Aires, Buenos Aires, Argentina, and approved March 24, 2017 (received for review November 23, 2016)

Anti-MAG (myelin-associated glycoprotein) neuropathy is a dis-abling autoimmune peripheral neuropathy caused by monoclonalIgM autoantibodies that recognize the carbohydrate epitope HNK-1(human natural killer-1). This glycoepitope is highly expressed onadhesion molecules, such as MAG, present in myelinated nerve fi-bers. Because the pathogenicity and demyelinating properties ofanti-MAG autoantibodies are well established, current treatmentsare aimed at reducing autoantibody levels. However, current ther-apies are primarily immunosuppressive and lack selectivity and ef-ficacy. We therefore hypothesized that a significant improvement inthe disease condition could be achieved by selectively neutralizingthe pathogenic anti-MAG antibodies with carbohydrate-based li-gands mimicking the natural HNK-1 glycoepitope 1. In an inhibitionassay, a mimetic (2, mimHNK-1) of the natural HNK-1 epitopeblocked the interaction of MAG with pathogenic IgM antibodiesfrom patient sera but with only micromolar affinity. Therefore, con-sidering the multivalent nature of the MAG–IgM interaction, poly-lysine polymers of different sizes were substituted with mimetic 2.With the most promising polylysine glycopolymer PL84(mimHNK-1)45 the inhibitory effect on patient sera could be improved by afactor of up to 230,000 per epitope, consequently leading to a low-nanomolar inhibitory potency. Because clinical studies indicate acorrelation between the reduction of anti-MAG IgM levels and clin-ical improvement, an immunological surrogate mouse model foranti-MAG neuropathy producing high levels of anti-MAG IgM wasdeveloped. The observed efficient removal of these antibodies withthe glycopolymer PL84(mimHNK-1)45 represents an important steptoward an antigen-specific therapy for anti-MAG neuropathy.

demyelinating peripheral neuropathy | IgM autoantibodies | myelin-associated glycoprotein | HNK-1 glycoepitope | glycosylated polylysine

Anti-MAG (myelin-associated glycoprotein) neuropathy is adisabling demyelinating peripheral neuropathy with an au-

toimmune etiology and a prevalence of about 1 in 100,000 (1). Itis slowly progressive, affecting sensory and motor nerves (2–4).Cardinal clinical symptoms are sensory ataxia with impaired gait,paresthesias, distal muscle weakness, and tremor (5). Monoclo-nal IgM autoantibodies recognize the HNK-1 (human naturalkiller-1) trisaccharide epitope 1 that is present on MAG as wellas on other glycoconjugates of the peripheral nervous system(PNS) (6). There is strong evidence that these IgM antibodieshave a pathogenic role in the development of demyelination andneuropathy (4, 5). Histopathological studies of nerve biopsiesfrom patients showed demyelinating lesions and widening ofmyelin lamellae as well as deposits of anti-MAG IgM on myelinsheaths (3, 7). Moreover, localization of anti-MAG IgM anti-bodies to areas of widened myelin lamellae indicates their role inmyelin disintegration (8). In addition to IgM deposits, somestudies report the presence of the complement factor C3d onmyelin, suggesting an inflammatory element in demyelination (7,9, 10). Strong evidence for the pathogenic role of anti-MAG

antibodies is provided by the damage of peripheral nerve mye-lin observed in experimental animals after the passive transfer ofpatients’ anti-MAG antibodies (11, 12). Additionally, active im-munization of cats with the HNK-1–containing glycolipid sulfo-glucuronyl paragloboside (SGPG) induced autoantibodies againstthe HNK-1 epitope and caused an ataxic sensory neuropathy re-sembling the human disease (13).The term “HNK-1 epitope” denotes the sulfated trisaccharide

SO3-3-GlcA(β1–3)Gal(β1–4)GlcNAc(1) (14), present in the PNSon the glycoprotein MAG, protein zero (P0), peripheral myelinprotein 22 (PMP22), and on the glycolipids SGPG and SGLPG(sulfoglucuronyl-lactosaminyl-paragloboside) (Fig. 1A) (6).MAG belongs to the family of sialic acid-binding Ig-like lectins(siglecs) (15, 16) and is located mainly in periaxonal membranesof oligodendroglial cells in the CNS (17). In Schwann cells of thePNS, MAG is localized mainly in paranodal loops and Schmidt–Lantermann incisures, where the anti-MAG autoantibodies ofpatients could be detected (7, 8, 18, 19). MAG is involved inadhesion and signaling processes at the axon–glia interface andexhibits a regulatory effect on axonal caliber (6). In addition tocomplement activation, anti-MAG antibodies are thought tointerfere directly with MAG’s adhesion and signaling function

Significance

Anti-MAG (myelin-associated glycoprotein) neuropathy is a rarebut disabling autoimmune disorder affecting the peripheral ner-vous system. The pathogenicity of anti-MAG IgM autoantibodiesthat target the HNK-1 glycoepitope is well established. Currenttherapies are mostly immunosuppressive but so far are neitherapproved nor sufficiently effective. Therefore we designed aglycopolymer that acts as an autoantibody scavenger by mim-icking the natural HNK-1 glycoepitope and demonstrated that theglycopolymer neutralizes disease-causing antibodies in patientsera. Moreover, pathogenic antibodies were removed efficientlyin an immunological mouse model of anti-MAG neuropathy.Because clinical improvement of patients’ neuropathic symptomscorrelates with reduced serum levels of anti-MAG antibodies, theglycopolymer represents a promising antigen-specific therapeuticoption for the treatment of this neuropathy.

Author contributions: R.H., P.H., N.S.-W., A.J.S., and B.E. designed research; R.H., P.H.,H.P., F.Y., D.D., F.H., and S.F. performed research; R.H., P.H., H.P., F.Y., D.D., F.H., S.F.,and B.E. analyzed data; and R.H., H.P., D.D., and B.E. wrote the paper.

Conflict of interest statement. R.H., P.H., A.J.S., and B.E. are co-founders of a University ofBasel spin-off, Polyneuron Pharmaceuticals AG, whose activity is related to the subjectmatter of this article. A.J.S. and B.E. are members of the advisory board, and B.E. is also amember of the board of directors. R.H., H.P., F.Y., A.J.S., and B.E. are named as co-inventors on relevant patent applications.

This article is a PNAS Direct Submission.1To whom correspondence should be addressed. mail: beat.ernst@unibas.ch.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1619386114/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1619386114 PNAS Early Edition | 1 of 10

NEU

ROSC

IENCE

PNASPL

US

Dow

nloa

ded

by g

uest

on

June

23,

202

0

Fig. 1. The HNK-1 carbohydrate epitope and its synthetic mimetics. (A) The HNK-1 glycoepitope 1 in the PNS is highly expressed by myelin glycoproteins, suchas MAG (with up to eight HNK-1 epitopes), and glycolipids, such as SGPG and SGLPG. This trisaccharide epitope with its characteristic sulfated glucuronic acidat the nonreducing end is subject to an autoimmune attack in anti-MAG neuropathy and binds to monoclonal anti-MAG IgM autoantibodies. (B) Four HNK-1–related disaccharides were prepared. GlcAβ1–3Gal, sulfated in the 3′ position and equipped with an aromatic aglycone, represents a mimetic of the naturalHNK-1 trisaccharide epitope (2, mimHNK-1). Compound 3 represents the desulfated version thereof, whereas compound 4 represents the minimal epitoperecognized by IgM autoantibodies. For the multivalent presentation of the HNK-1 epitope mimetic 2, derivative 5 modified with an amino linker for polymercoupling was prepared. (C, a) TMSOTf, 4 Å MS, DCM, 0 °C → rt, 86%. (b) LiOH, THF/H2O, 89%. (c, 1) Ac2O, 80 °C. (2) DMAP, Pyr. (d) NaOAc, MeOH, 73% overtwo steps. (e) SO3·Pyr, DMF, 91%. (f) LiOH, THF/H2O. (g) H2, Pd(OH)2/C, MeOH/H2O, 78% over two steps. (h) γ-Thiobutyrolactone, DTT, Et3N, DMF, 85 °C, 59%.(i) (ClCH2CO)2O, 2,6-lutidine/DMF, 96%; (j, 1) DBU, DMF/H2O. (2) Thioglycerol, Et3N, 70%.

2 of 10 | www.pnas.org/cgi/doi/10.1073/pnas.1619386114 Herrendorff et al.

Dow

nloa

ded

by g

uest

on

June

23,

202

0

(4). Several observations suggest that MAG is the main target forthe IgM antibodies: (i) deposits of patients’ antibodies to PNSsites are colocalized with MAG (19); (ii) MAG is selectively lostfrom PNS myelin in patients with high antibody titers (20); (iii)the pathology of MAG-knockout mice (21, 22) shows similaritiesto the human nerve pathology (23, 24); and (iv) human MAG hasa higher HNK-1 epitope density (25) than any other myelinglycoconjugate, leading to strong IgM antibody binding (26).Based on in vitro studies, it was postulated that patients’ IgMautoantibodies increase the permeability of the microvascularendothelial structure, presumably through the binding of endo-thelial sulfoglucuronyl glycosphingolipids, and thereby gain ac-cess to the PNS parenchyma via a leaky blood–nerve barrier (27).The goals of current therapies are to reduce pathogenic au-

toantibodies, decrease expanded autoantibody-producing B-cellclones, and/or interfere with antibody-effector mechanisms (4).Corticosteroids (e.g., prednisone), cytostatics (e.g., cyclophos-phamide), IFN alpha-2a, plasma exchange, i.v. Ig, and theCD20+ B-cell–depleting agent rituximab have been used for thetherapy of anti-MAG neuropathy (4, 5, 28). However, theseapproaches lack efficiency and selectivity, and thus so far nosatisfactory treatment is available (28). Indeed, with the mostpromising agent, rituximab (4, 5), acute exacerbations of symp-toms were observed in some patients (29, 30). Because clinicalimprovement of patients’ neuropathic symptoms correlates withreduced serum levels of anti-MAG antibodies (4, 31–34), anddisease worsening is associated with increasing anti-MAG levelsduring treatment follow-up (5, 35), a more efficient and safertherapy might be achieved with antigen-specific agents that se-lectively target anti-MAG IgM antibodies or antibody-producingB cells.In this study, biodegradable poly-L-lysine–based glycopolymers

containing a disaccharide glycomimetic of the natural HNK-1glycoepitope were synthesized. It could be shown that theseglycopolymers prevent binding of anti-MAG IgM antibodies toMAG at low-nanomolar epitope concentrations in vitro. Moreimportantly, in vivo data clearly demonstrated the efficient re-moval of pathogenic anti-MAG antibodies in an immunologicalmouse model for anti-MAG neuropathy. In summary, multivalentHNK-1 glycomimetics exhibit a considerable therapeutic potentialfor an antigen-specific treatment of anti-MAG neuropathy.

ResultsMimetics of the HNK-1 Glycoepitope 1 Block Anti-MAG Autoantibodiesin Patients’ Sera. Tokuda et al. (36) reported binding of IgM an-tibodies from patients with anti-MAG neuropathy and also mousemonoclonal HNK-1 antibody (20) to derivatives of the glycolipidSGPG containing only the terminal disaccharide SO3-3-GlcA(β1–3)Gal. Therefore, we synthesized three different inhibitors, thedisaccharides 2, 3, and 4 (Fig. 1B), and tested their potential toinhibit the binding of patients’ anti-MAG antibodies to MAG inan ELISA. In disaccharide 2, the GlcNAc moiety at the reducingend of the natural HNK-1 trisaccharide 1 is replaced by a para-methoxyphenyl aglycone, and compound 3 represents the unsul-fated version thereof. Finally, the sulfated disaccharide 4 with amethyl aglycone represents the minimal epitope recognized by

patients’ anti-MAG IgM as well as by the mouse monoclonalHNK-1 antibody (36). With disaccharide 2, inhibition of autoan-tibody binding to MAG was achieved at micromolar concentra-tions, whereas the derivatives 3 (missing the 3′ sulfate) and 4 (witha methyl instead of a para-methoxyphenyl aglycone) inhibitedbinding only at high-micromolar to millimolar concentrations(Table 1). For a further characterization of the disaccharides 2–4,the binding to serum samples with confirmed high anti-MAG IgMantibody titers from four patients (MK, DP, KH, and SJ) di-agnosed with an anti-MAG neuropathy was studied. Under stan-dard assay conditions, the sera were diluted to achieve OD450values close to 1. Although serum from patient MK exhibited amuch higher binding to the sulfated ligand 2 than to the unsulfateddisaccharide 3 (∼230-fold), serum from patient SJ showed only a12.6-fold stronger binding to the sulfated ligand 2 (Table 1).Moreover, the para-methoxyphenyl aglycone in disaccharide2 improves binding to pathogenic IgM autoantibodies. Its re-placement by a methyl aglycone (→ disaccharide 4) led to an af-finity drop for sera from patients MK and SJ by a factor of 6.9 and1.6, respectively (Table 1). In summary, disaccharide 2 turned outto be the most potent autoantibody ligand and therefore the mostsuitable mimetic of the natural HNK-1 glycoepitope. Conse-quently, in further studies, it was used as the mimetic of the HNK-1 epitope (mimHNK-1).

Synthesis of Glycopolymers Comprising the MimHNK-1 Epitope 5.Given the multivalent nature of the IgM–MAG interaction (26),we hypothesized that a multivalent presentation of the mimHNK-1 epitope 2 might increase antibody binding affinity substantially.To enable coupling to the polymer, the mimHNK-1 epitope 2 wasequipped with a tyramine-based thiol-linker (→ 5) (Fig. 1C). Forits synthesis acceptor 7 was glycosylated with the glucuronic acidderivative 6 to yield disaccharide 8. To obtain the 3′-unprotectedalcohol 9, a three-step procedure according to Kornilov et al. (37)was applied. The glucuronic acid-[3,6]-lactone was formed by sa-ponification followed by reflux in acetic anhydride. After acety-lation of the remaining hydroxyls followed by methanolysis of thelactone, the desired alcohol 9 was obtained. Finally, sulfation ofthe 3′-hydroxyl group (→ 10), saponification, hydrogenolysis, andazide reduction yielded the unprotected disaccharide 11. Nucle-ophilic opening of γ-thiobutyrolactone with amine 11 yielded thiol5, ready for coupling in substoichiometric amount to the chlor-oacetylated poly-L-lysine 13. To improve the water solubility of theglycopolymer, the remaining chloroacetyl groups were cappedwith thioglycerol. In the final glycopolymer PLy(mimHNK-1)x, ydefines the degree of polymerization of the backbone in kilodaltons,and x stands for the percentage of epitope loading, as determined by1H NMR spectroscopy.To evaluate two important polymer parameters, namely epitope

loading and degree of polymerization, two series of glycopolymerswere prepared. The inhibitory activities of these polymers weredetermined in a MAG-binding inhibition assay with the mousemonoclonal anti–HNK-1 IgM antibody (20) as competitor (Fig. 2A and B and Table 2). Inhibitory activity increased gradually when10–45% of the lysine side chains of a 40- to 60-kDa poly-L-lysinepolymer (PL40–60) were loaded with epitope 5. However, when the

Table 1. IC50 values of compounds 2, 3, and 4 and the polymer PL40–60(mimHNK-1)45 for MAG binding by the sera offour patients

Patient serum Compound 2, IC50, μM Compound 3, IC50, μM Compound 4, IC50, μM PL40–60kDa (mimHNK-1)45, IC50, μM

MK 124.0 ± 9.5 28,967.3 ± 533.0 860.0 ± 58.2 0.0046 ± 0.0001DP 536.1 ± 23.5 N.d. N.d. N.d.KH 614.2 ± 20.1 N.d. N.d. 0.0245 ± 0.0025SJ 793.1 ± 24.0 9,981.1 ± 1002.2 1,237.0 ± 56.0 0.0316 ± 0.0067

N.d., not determined.

Herrendorff et al. PNAS Early Edition | 3 of 10

NEU

ROSC

IENCE

PNASPL

US

Dow

nloa

ded

by g

uest

on

June

23,

202

0

loading was increased further to 50 and 75%, a drop of inhibitoryactivity was observed. With the most potent polymer (45% load-ing) an IC50 of 53.7 ± 10.8 nM was obtained. Consequently, aloading of 45% was used for the evaluation of the degree of po-lymerization. Poly-L-lysines of different size, i.e., molecular massesranging from 4–15 kDa (average of 45 lysines) up to 150–300 kDa(average of 1,075 lysines) (Fig. 2B and Table 2) were used. Thetwo most potent polymers had a molecular mass of 75–150 kDa(Sigma Aldrich) and 84 kDa (Alamanda Polymers). These gly-copolymers, PL75–150(mimHNK-1)45 and PL84(mimHNK-1)45,exhibited IC50 values of 5.9 ± 5.1 nM and 5.4 ± 1.2 nM, re-spectively. Because PL84(mimHNK-1)45 presents a narrower mo-lecular mass distribution (45% epitope loading, 55% thioglycerolcapping leading to a calculated molecular mass of 217 kDa), it waschosen as the lead candidate for further studies.

In Vitro Validation of PL84(mimHNK-1)45 with Patients’ Sera. To de-termine the inhibitory activity of the lead glycopolymerPL84(mimHNK-1)45 by ELISA, serum samples from five patients(MK, DP, KH, SJ, and HF) with a confirmed clinical diagnosis ofanti-MAG neuropathy and high anti-MAG antibody titers wereused. These serum samples showed strong MAG binding at a1:1,000 dilution, whereas sera from patients with other neuro-logical diseases (n = 5), two of them with a confirmed mono-clonal gammopathy without anti-MAG antibodies, showed nobinding to MAG at the same dilution (Fig. 2C). For inhibitionassays, the patient serum samples were diluted to yield OD450values close to 1. With the sera of five anti-MAG neuropathypatients, the glycopolymer PL84(mimHNK-1)45 exhibited a meanIC50 value of 3.6 ± 0.4 nM, comparable to the value (IC50 = 5.4 ±1.2 nM) we obtained with the mouse monoclonal anti–HNK-1 antibody (Fig. 2D). The 100% thioglycerol-capped, epitope-free control polymer PL40–60(mimHNK-1)0 did not inhibitbinding of anti-MAG antibodies of sera from patients KH, MK,and SJ to MAG (Fig. 2F). To evaluate the therapeutic applica-bility of PL84(mimHNK-1)45, i.e., whether the response is broador restricted to sera from specific patients, MAG-binding wasdetermined by ELISA with an additional set of 10 serum samplesfrom anti-MAG patients at a uniform dilution of 1:1,000 and aglycopolymer epitope concentration of 1 μM. Compared withPBS-treated controls, PL84(mimHNK-1)45 significantly reducedMAG binding in all samples (Fig. 2E).

Generation of an Immunological MouseModel for Anti-MAG Neuropathy.An extract enriched with the two PNS glycolipids, SGPG and itshigher homolog SGLPG, was isolated from bovine cauda equina(38). Both glycolipids share the HNK-1 glycoepitope 1 (Fig. 1A).The purity of the extracts was analyzed by TLC and staining of theseparated glycolipids. Furthermore, the recognition of the HNK-1glycoepitope 1 (Fig. 1A), present in the two enriched glycolipids,

Fig. 2. In vitro inhibitory activity of mimHNK-1 glycopolymers on MAGbinding of anti-MAG IgM. (A) Ten to seventy-five percent of the side chainsof a poly-L-lysine (hydrobromide salt) polymer with a molecular mass of 40–60 kDa were loaded with the mimHNK-1 epitope 5. A MAG-binding in-hibition ELISA was used to determine the inhibitory capacity of the differentglycopolymers. With a loading of 45% the strongest inhibition of MAG-binding by a mouse monoclonal anti–HNK-1 IgM diluted 1:1,000 was ob-served [IC50 for PL40–60(mimHNK-1)45 = 53.7 ± 10.8 nM]. (B) Inhibitory activityof a polymer series with an epitope loading of 45% and different backbonesizes was analyzed by the MAG-binding inhibition ELISA. A polymer with apoly-L-lysine backbone with a molecular mass of 70–150 kDa (Sigma Aldrich),

equivalent to 360–720 L-lysines, and a polymer with a backbone with amolecular mass of 84 kDa (Alamanda Polymers), equivalent to 400 L-lysines,showed comparable affinities: IC50 for PL75–150(mimHNK-1)45 = 5.9 ± 5.1 nM;IC50 for PL84(mimHNK-1)45 = 5.4 ± 1.2 nM. (C) Serum samples (1:1,000) fromanti-MAG neuropathy patients KH, SJ, HF, MK, and DP showed high MAGbinding as determined by ELISA, whereas control serum samples (1:1,000)from five patients with other neurological disorders showed no MAG-binding. (D) The binding of serum anti-MAG IgM from patients KH, SJ, HF,MK, and DP (1:7,500–1:45,000) to MAG was inhibited by PL84(mimHNK-1)45with an average IC50 of 3.6 ± 0.4 nM. (E) Anti-MAG IgM antibodies in pa-tients’ sera (n = 15; 1:1,000) were inhibited from binding to MAG byPL84(mimHNK-1)45 with an epitope concentration of 1 μM. (F) A controlpolymer PL40–60(mimHNK-1)0 (100% thioglycerol-capped control polymer)showed no inhibition of MAG binding up to a thioglycerol-lysine concen-tration of 10 mM per unit. Results in A, B, and D are shown as mean ± SD;results in C are shown as single values with the mean; values in E are shownas median + 95% CI; and values in F are shown as mean + SD.

4 of 10 | www.pnas.org/cgi/doi/10.1073/pnas.1619386114 Herrendorff et al.

Dow

nloa

ded

by g

uest

on

June

23,

202

0

was confirmed by TLC immunostaining with the mouse anti–HNK-1 IgM antibody (Fig. 3A). For the immunization of 6- to8-wk-old BALB/c wild-type mice the SGPG/SGLPG extracttogether with adjuvants was used. In the immunized animals, anti-MAG IgM antibody levels reached a plateau 50–70 d post-immunization (Fig. 3B). In the control group immunized withadjuvants in the absence of SGPG/SGLPG extract (vehicle), noMAG antibody reactivity could be detected. The isotype of theanti-MAG antibodies and their glycoepitope specificity were an-alyzed in plasma samples of four mice. Both IgG and IgM anti-bodies were detected, but the latter isotype was predominant (Fig.3C). In addition, mouse plasma reacted specifically with the HNK-1glycoepitope 1 present on MAG and SGPG but not with otherneuronal/myelin glycoepitopes (i.e., GM1, GM2, GD1a, GD1b,and GQ1b) (Fig. 3C). A comparison of the mouse plasma sampleswith serum samples from patients with anti-MAG neuropathy(n = 4) showed a similar epitope reactivity pattern (Fig. 3D). Asexpected, no IgG antibody reactivity could be determined in pa-tient serum samples. Moreover, Western blot analysis of antibodyreactivity against MAG in a human CNS myelin extract showedcomparable reactivity in patient serum (n = 2) and mouse plasma(n = 2) with an upper band at 100 kDa corresponding to MAGand a lower band at 90 kDa corresponding to dMAG, a proteasebreakdown product consisting of only the soluble extracellulardomain of MAG (Fig. 3E) (3, 6). In a control experiment withrabbit anti–L-MAG antibody recognizing the C-terminal intracel-lular part of L-MAG (39), the extracellular dMAG could not bedetected.

Evaluation of PL84(mimHNK-1)45 in a Surrogate Anti-MAG NeuropathyMouse Model. Subsequently, the generated mouse model was usedto explore the therapeutic potential of PL84(mimHNK-1)45. Ini-tially, two important pharmacokinetic parameters were of interest.The first was the thermodynamic solubility of PL84(mimHNK-1)45in PBS, which was determined to be at least 150 mg/mL, con-firming a favorable high water solubility. The second was theplasma half-life after a single i.v. dose of 10 mg/kg (n = 5), whichturned out to be in the range of 17 min (t1/2, 16.9 ± 5.5 min) (Fig.4A). Thereafter, the therapeutic potential of the glycopolymerPL84(mimHNK-1)45 was explored by administering a dose of10 mg/kg (in PBS) into the tail vein of immunized mice (n = 5); thecontrol groups were treated with PBS or with only control polymerPL40–60(mimHNK-1)0 (n = 5 in both experiments) (Fig. 4 B and C).Three hours before and after administration of the glycopolymer,blood samples were taken and analyzed, showing a significant de-crease of anti-MAG IgM levels upon treatment (n = 5). Moreover,with one administration of 10 mg/kg PL84(mimHNK-1)45, signifi-cantly decreased levels were observed for up to 5 d, and at day7 the decrease was still present, although nonsignificant (Fig. 4D).Importantly, both anti-MAG IgM and anti-MAG IgG levelswere significantly reduced 3 h after treatment with 10 mg/kg of

PL84(mimHNK-1)45 (n = 4) (Fig. 4E). With a dose of 1 mg/kg ofthe glycopolymer (Fig. 4F), anti-MAG IgM antibody levels weresignificantly reduced 3 and 24 h after treatment but returned topretreatment levels after 3 d (Fig. 4F). Weekly treatment with10 mg/kg PL84(mimHNK-1)45 for 5 wk progressively reduced therebound of anti-MAG IgM antibody levels at the end of eachtreatment week (n = 6). Anti-MAG IgM levels were significantlyreduced compared with pretreatment only for 3 d after the firstdrug administration, but after the fifth administration antibodylevels remained significantly lowered for 14 additional days (Fig.4G). With Western blot analysis of mouse plasma samples taken3 h after treatment with 1 mg/kg of glycopolymer, antibody re-activity against MAG using a human CNS myelin extract could nolonger be detected (Fig. 4H).When wild-type BALB/c mice were treated daily with 10 mg/kg

PL84(mimHNK-1)45 for 10 consecutive days, plasma samples didnot show any formation of antidrug antibodies (ADAs) of theIgG or IgM isotype for a follow-up period of 50 d. An ADA-positive control group was obtained when wild-type BALB/cmice were immunized s.c. with PL84(mimHNK-1)45, immuno-genic Keyhole limpet hemocyanin (KLH), and TiterMax Gold.Interestingly, immunization induced a much stronger response toIgG ADA than to IgM ADA (Fig. 4I). These IgG ADA did notshow reactivity with the backbone of the glycopolymer or withMAG but reacted only with PL84(mimHNK-1)45.

DiscussionThe pathogenesis of the anti-MAG neuropathy as an antibody-mediated autoimmune disease is widely accepted, and therapiesare aiming at a reduction of levels of autoantibodies and/orautoantibody-producing B-cell clones through immunomodulationand immunosuppression (4, 5). Nonetheless, there is a strong needfor personalized, new disease-modifying agents devoid of non-specific immunosuppression. To develop an antibody-specifictreatment, Page et al. raised an anti-idiotypic antibody againstthe variable regions of a monoclonal IgM antibody derived from apatient with anti-MAG neuropathy (40). It successfully inhibitedthe binding of the individual patient’s IgM antibodies to MAG invitro but turned out to be ineffective with four other patients’ IgMantibodies, indicating idiotypic heterogeneity among patients’autoantibodies. Despite this microheterogeneity, antibodies fromdifferent patients bind to the HNK-1 epitope, suggesting that allanti-MAG antibodies can be targeted by mimetics of their sharedglycoepitope (41). The monoclonal anti-MAG IgMs of neuropa-thy patients bind specifically to the HNK-1 carbohydrate epitopepresent on MAG and on the PNS-specific glycolipids SGPG andSGLPG (5). The minimal carbohydrate epitope recognized bythe autoantibodies is the SO3-3-GlcA(β1–3)Gal disaccharide (36,37). After MAG deglycosylation, the patients’ autoantibodies losetheir MAG reactivity (42). Saturation transfer difference (STD)NMR experiments revealed that the terminal trisaccharide

Table 2. IC50 values of PLy(mimHNK-1)x polymers with different degrees of polymerization (y, kDa) and epitopeloadings (x, % of epitope loading) for MAG binding by mouse monoclonal anti–HNK-1 IgM antibody

PLy(mimHNK-1)x

IC50, nM

10 25 31 45 ± 1 50 75

4–15 29,406.2 ± 11,149.8*15–30 1,404.0 ± 577.130–70 345.7 ± 53.540–60 3,892.0 ± 644.2 324.7 ± 41.0 124.8 ± 24.1 53.7 ± 10.8 561.9 ± 59.6 935.5 ± 53.970–150 5.9 ± 5.1150–300 14.1 ± 5.284 5.4 ± 1.2

*In the polymer series with different backbone sizes epitope loading was 45% for all polymers except for the smallest PL4–15(mimHNK-1)x polymer, where epitope loading was only 38%.

Herrendorff et al. PNAS Early Edition | 5 of 10

NEU

ROSC

IENCE

PNASPL

US

Dow

nloa

ded

by g

uest

on

June

23,

202

0

SO3-3-GlcA(β1–3)Gal(β1–4)GlcNAc, i.e., the HNK-1 epitope, ofthe glycolipid SGPG interacts with the mouse monoclonal anti–HNK-1 IgM and that the most significant contribution is made bythe terminal disaccharide SO3-3-GlcA(β1–3)Gal (43).In this study, we synthesized a glycomimetic of the natural

HNK-1 trisaccharide epitope. Our data suggest that the aromaticaglycone of 2 (mimHNK-1) mimics the “hydrophobic face” (44)

of the reducing end GlcNAc moiety and that the sulfate group inthe 3′ position of the GlcA moiety is crucial for antibody binding(Table 1).Ogino et al. (26) have shown that decavalent IgM antibodies

can establish multivalent interactions with MAG, which presentsup to eight HNK-1 epitopes on its extracellular domain (45, 46).Thus, to improve the low affinity of the sulfated disaccharide 2, amultivalent presentation on a poly-L-lysine backbone was explored(45). Poly-L-lysine is biodegradable and therefore is suitable fortherapeutic applications (47). A careful optimization of the degreeof polymerization and epitope density yielded the tailor-madeglycopolymer PL84(mimHNK-1)45, exhibiting maximal inhibitoryactivity. MAG binding by anti-MAG IgM antibodies was inhibitedby this glycopolymer in all tested patient sera. Compared with themonomer 2, the inhibition of MAG binding was improved by afactor of 38,000–230,000, strongly supporting the multivalent natureof the antigen–antibody interaction.Next, the potential of PL84(mimHNK-1)45 to deplete anti-MAG

IgM was studied in a mouse model for anti-MAG antibodies.Several animal models for anti-MAG neuropathy have beenreported and are based on passive transfer of patients’ IgM anti-bodies into healthy experimental animals, including cats (11, 48)and chickens (12), and, more recently, on active immunization ofcats with SGPG leading to a sensory ataxic neuropathy (13). Al-though these animal models partly mimic characteristics of thehuman myelin and nerve pathology of anti-MAG neuropathy, ro-dents do not display a clear neuropathic phenotype after immuni-zation with SGPG (49, 50). This lack of neurological symptoms is aconsequence of the following findings: (i) during early neurogenesis,HNK-1 expression in neural crest cells is very low in rats and isabsent in mice (51); (ii) MAG in rodents lacks the HNK-1 epitope(27, 42); and (iii) SGPG/SGLPG expression is about one order ofmagnitude lower in the peripheral nerves of rats and mice than inhigher mammals, such as cats and dogs (52). To obtain a surrogateimmunological model for anti-MAG neuropathy, BALB/c mice,frequently used for immunological studies, were selected for activeimmunization. Because the clinical improvement in patients corre-lates with the reduction of their anti-MAG IgM antibody levels (4,31–33), this mouse model is a suitable surrogate model for studyingthe depletion of anti-MAG IgM antibodies by therapeutic agents.Importantly, the mouse IgM antibodies are comparable with theirhuman counterparts, because they recognized both human MAGand SGPG, although reactivity toward other nerve and myelinglycoepitopes is absent (Fig. 3 C–E).The i.v. administration of 1 and 10 mg/kg of PL84(mimHNK-1)45

to anti-MAG IgM–positive mice effectively reduced antibody levelsin a dose-dependent manner. In addition to anti–HNK-1 IgM, anti–HNK-1 IgG also could be removed efficiently. Upon repetitiveweekly administration of 10 mg/kg of the glycopolymer, the de-creasing rebound of anti-MAG antibody levels suggests the existenceof a potential immunomodulatory treatment effect. Furthermore,the relatively short plasma half-life of PL84(mimHNK-1)45 of∼17 min after i.v. injection in mice implies a total elimination within2 h. Thus, the reduced level of anti-MAG antibodies for up to 2 wkafter treatment (Fig. 4G) results from the removal rather thanthe neutralization of anti-MAG antibodies (53, 54). Further-more, the short half-life is comparable to that of a previouslydescribed glycopolymer (47, 54) and might be crucial for lowimmunogenicity. Indeed, daily treatment of nonimmunized micewith PL84(mimHNK-1)45 did not trigger the formation of IgMand IgG ADAs (Fig. 4I), suggesting low immunogenicity ofPL84(mimHNK-1)45. In addition, no signs of intolerance wereobserved upon repetitive administration, a result that is in accor-dance with a polymer developed by Duthaler et al. (47), whichefficiently eliminated both anti-αGal IgM and IgG in nonhumanprimates and showed neither immunogenicity nor toxicity. Re-garding off-target effects, the glycopolymer may interact with HNK-1–binding molecules, such as the proinflammatory cytokines IL-6

Fig. 3. Generation and analysis of an immunological mouse model for anti-MAG neuropathy. (A) A mixture of the glycolipids SGPG/SGLPG was isolatedfrom bovine cauda equina. The analysis of the extracts E1 and E2, by bothTLC mostain staining (Left) and TLC immunostaining with 1:400 dilutedmouse HNK-1 IgM antibody (Right) is shown. The SGPG/SGLPG ratio varied indifferent extracts, and some contamination was still observed after purifi-cation. E1 was subjected to an additional purification step with silica columnchromatography; E2 was not. (B) Five BALB/c wild-type mice (6–8 wk old)were injected s.c. with E1 in PBS together with the immunogenic KLH andTiterMax Gold as adjuvant at days 0, 14, and 28. A control group (n = 5) wastreated without E1 at the same time points. Anti-MAG IgM levels werefollowed up over time by ELISA and increased until a plateau was reachedafter about day 70. (C) Sera (diluted 1:100) of four immunized BALB/c miceshowed both anti-MAG and anti-SGPG IgM and IgG antibodies bindingspecifically to the HNK-1 glycoepitope on MAG and SGPG but not to fivegangliosides that are relevant myelin/nerve glycoepitopes. (D) Serum sam-ples (1:1,000) from four patients with anti-MAG neuropathy showed a highspecificity for the HNK-1 glycoepitope on MAG and SGPG. In contrast to theimmunized BALB/c mice, which developed both anti-MAG IgM and IgG,patients exhibit only anti-MAG IgM antibodies. (E, Left) Western blot anal-ysis of MAG reactivity in serum and plasma samples was performed using ahuman CNS myelin extract. Serum samples (1:800) from two patients (MKand DP) and plasma samples (1:400) from two mice at day 70 after immu-nization tested positive for anti-MAG IgM, whereas plasma samples (1:400)from the same two mice did not show any MAG reactivity before immuni-zation. (Right) A rabbit anti-human L-MAG antibody was used in a controlexperiment (1:1,000). Results in B are shown as mean ± SD, and results in Cand D are shown as mean + SD.

6 of 10 | www.pnas.org/cgi/doi/10.1073/pnas.1619386114 Herrendorff et al.

Dow

nloa

ded

by g

uest

on

June

23,

202

0

Fig. 4. Treatment of SGPG-immunized mice with the PL84(mimHNK-1)45 glycopolymer. (A) PL84(mimHNK-1)45 was determined in plasma samples of five miceby ELISA after a single i.v. bolus injection of 10 mg/kg and revealed a drug half-life of ∼17 min (t1/2, 16.9 ± 5.5 min). (B) An i.v. bolus injection of 10 mg/kgPL84(mimHNK-1)45 resulted in a significant (93%) decrease of anti-MAG IgM antibody levels 3 h after administration of PL84(mimHNK-1)45 compared withpretreatment levels (n = 5; **P ≤ 0.01). (C) Treatment of immunized mice (n = 5) with PBS and treatment with the control polymer PL40–60(mimHNK-1)0 (n = 5)did not result in any changes in anti-MAG antibody levels. (D) A dose of 10 mg/kg PL84(mimHNK-1)45 led to a sustained decrease in anti-MAG antibodies,which was significant up to 5 d after injection (n = 5; ***P ≤ 0.001). (E) An i.v. bolus injection (10 mg/kg) of PL84(mimHNK-1)45 significantly depleted anti-MAGIgG in the mice (n = 4) 3 h posttreatment (***P ≤ 0.001). (F) A lower dose of PL84(mimHNK-1)45 (1 mg/kg) also depleted anti-MAG IgM antibodies (n = 5;***P ≤ 0.001), which recovered 72 h after injection. (G) Weekly treatment with 10 mg/kg PL84(mimHNK-1)45 for 5 wk resulted in a decreasing anti-MAG IgMantibody rebound toward the end of each week. Compared with pretreatment levels, anti-MAG IgM levels remained significantly lowered for up to 14 d afterthe fifth administration on day 28 (n = 6; *P ≤ 0.05). (H) Western blot analysis of MAG reactivity in a human CNS myelin extract with plasma samples (1:400)taken from mice (n = 2) before immunization, on day 82 after immunization (postimmunization), and 3 h after treatment with 1 mg/kg PL84(mimHNK-1)45 onday 82. Plasma of treated mice showed no MAG reactivity 3 h after treatment compared with plasma before treatment (postimmunization). Sera (1:800) fromtwo patients with anti-MAG neuropathy (MK and DP) and a rabbit anti-human L-MAG antibody (1:1,000) were used in control experiments. (I) Nonimmunizedmice (n = 4) underwent daily i.v. treatment with 10 mg/kg PL84(mimHNK-1)45 for 10 consecutive days (drug treated). ADAs were detected in the treatedgroup. In the ADA-positive control group (immunized ctrl.), which was immunized s.c. with PL84(mimHNK-1)45 together with the immunogenic KLH andTiterMax Gold (n = 4), the IgG ADA response was more pronounced than the IgM ADA response. Except for A and I, in which results are shown as mean ± SD,results are shown as median + 95% CI.

Herrendorff et al. PNAS Early Edition | 7 of 10

NEU

ROSC

IENCE

PNASPL

US

Dow

nloa

ded

by g

uest

on

June

23,

202

0

(55) and HMGB1, the extracellular matrix components laminin-1 and -2, and lecticans (56). However, PL84(mimHNK-1)45 is notexpected to interact with the HNK-1 receptors in the nervous sys-tem, where HNK-1 is mainly expressed; because of the large sizeand high charge of the glycopolymer, it is not anticipated to pen-etrate the blood–nerve or blood–brain barrier.The concept of anti-glycan antibody removal has been ex-

plored previously both in vivo, with anti-idiotypic antibodies (40,57), peptide glycomimetics (58), monomeric glycans (59), andglycopolymers (47, 54, 59), and ex vivo using immunoaffinitycolumns (59, 60). Glycopolymers have been used thus far onlyfor in vivo removal of natural anti-glycan antibodies, e.g., poly-mers of αGal or ABH blood group glycoantigens (47, 54, 59).Here we describe the selective in vivo removal of an anti-glycanautoantibody using a carbohydrate-based therapeutic.The PL84(mimHNK-1)45 glycopolymer potentially enables an

antigen-specific therapy for anti-MAG neuropathy and thereforeis of considerable clinical interest. Furthermore, glycopolymersmight be generally effective as therapeutic agents for the de-pletion of pathogenic anti-carbohydrate autoantibodies in otherantibody-mediated diseases such as multifocal motor neuropathyor Guillain–Barré syndrome.

Materials and MethodsStatistical Analysis. Unless otherwise stated, results are given as mean ± SD ormedian + 95% CI of three independent experiments. Comparisons betweentwo conditions were performed using either Student’s t test or one-wayANOVA with Dunnett’s multiple comparison posttest with a 0.05 confi-dence level accepted for statistical significance (*P ≤ 0.05, **P ≤ 0.01,***P ≤ 0.001).

Synthesis of the Glycomimetics and Glycopolymers. See SI Appendix for thepreparation and analysis of compounds 2–5 and the glycopolymers.

Patient and Control Serum Samples. Sera from patients with positive anti-MAG IgM titers and control sera from patients with neurological disordersother than anti-MAG neuropathy (no anti-MAG IgM reactivity) were obtainedfrom the University Hospital of Basel. Patients MK, DP, KH, SJ, and HF had aclinical diagnosis of amonoclonal IgM gammopathy and anti-MAG neuropathywith high antibody titers (≥70,000 Bühlmann titer units) as measured by anti-MAG ELISA (Bühlmann Laboratories AG). The additional 10 patient sera usedin our study also tested positive by anti-MAG ELISA for high levels of anti-MAGIgM antibodies, which are highly indicative of anti-MAG neuropathy. Controlsera tested negative for anti-MAG IgM antibodies. Two of the five control seraoriginated from patients with a monoclonal IgM gammopathy without anti-MAG reactivity. The use of patient sera was approved by the Ethics Committeeof Northwestern and Central Switzerland (EKNZ). Informed consent wasobtained from all nonanonymized participants.

ELISA. The potential of HNK-1 mimetics (compound 2–4 and polyvalent de-rivatives of 5) to inhibit binding of anti-MAG IgM (mouse or human) toimmobilized MAG was determined in an anti-MAG ELISA. For that purpose,96-well plates coated with human MAG (EK-MAG; Bühlmann LaboratoriesAG) were used. The assay was performed according to the manufacturer’sprotocol. Briefly, test compound, sera from patients or mouse monoclonalanti-HNK-1 IgM antibody (20), and incubation buffer (provided with the kit)were added to a final volume of 50 μL per well. The assay was run in trip-licate. In case of MAG-binding assays, antibodies (in the absence of testcompounds) were incubated in the same volume. If not indicated otherwise,human serum and mouse plasma samples (single or duplicate) were diluted1:1,000 and 1:100, respectively. For detection of human anti-MAG IgM orIgG, anti-human IgM or IgG secondary antibodies conjugated to HRP wereused. The mouse HNK-1 IgM antibody (20) and anti-MAG IgM from plasmaof immunized mice were detected with goat anti-mouse IgM HRP conjugate(A8786; Sigma Aldrich) diluted 1:10,000. Anti-MAG IgG in mouse plasma wasdetected with goat anti-mouse IgG HRP conjugate (A4416; Sigma Aldrich)diluted 1:10,000. The OD of the colorimetric signal was measured at 450 nmon a microplate reader (Spectramax 190; Molecular Devices). The IC50 valuesof the tested compounds were calculated using Prism 5.0 software(GraphPad Software, Inc.). IC50 values of polymers were based not on theaverage molecular mass (MW) of the polymers but on the equivalent weightper glycoepitope. The equivalent weight was calculated by the formula

[(MWglycoepitope-lysine-unit)·x + (MWthioglycerol-lysine-unit)·(1 − x)]/x. It is indepen-dent of the degree of polymerization (n) but is dependent on the fractionx of glycosylated lysine units (as measured by 1H NMR). This calcula-tion allowed a direct comparison between the inhibitory activity of themimHNK-1 epitope monomer and the samemolecule as part of a polymer. Forthe 100% thioglycerol-capped control polymer, only the MWthioglycerol-lysine-unit

was used for concentration calculation. For the evaluation of glycoepitopespecificity of mouse and patient antibodies, both an anti-SGPG ELISA andGanglioCombi MAG ELISA (EK-GCM; Bühlmann Laboratories AG) were per-formed according to the manufacturer’s instructions. In our assay, 50 μL ofpatient sera at 1:1,000 (patients MK, SJ, SP, and KH) and mouse plasma at1:100 dilution (plasma samples from three mice at day 43 and one mouse atday 62 after first immunization) were incubated.

Purification of SGPG/SGLPG from Bovine Cauda Equina. An extract of glyco-lipids enriched with SGPG and SGLPG was isolated from bovine cauda equinasimilar to the protocol of Burger et al. (38). In contrast to Burger’s protocol,ion exchange chromatography of the isolated glycolipid pellet was per-formed using a DEAE-Sephadex A-25 column (6 g, 3 × 6 cm) that was runwith 75 mL of each of the following eluents: 0.05 M NaOAc, 0.1 M NaOAc,0.2 M NaOAc, and 0.5 M NaOAc in MeOH (flow rate 1 mL/min). SGPG andSGLPG were recovered in the 0.5 M NaOAc fraction, which was evaporatedto dryness. The residue was suspended in 10 mL of water (pH 2.5) anddesalted on a RP-18 silica column using a CombiFlash companion (TeledyneIsco) (10 mL per fraction, flow rate 8 mL/min). After removal of NaOAc with80 mL of water (pH 2.5), the glycolipids were eluted with 120 mL of MeOHand were recovered in fractions containing the two glycolipids as analyzedby TLC. MeOH was evaporated, and the extract was lyophilized. To removeimpurities, an additional chromatography on silica was performed. A CH2Cl2/MeOH gradient (30 min 15–45% MeOH; 10 min 45% MeOH, 10 min 100%MeOH) was used (10 mL per fraction, 8 mL/min) and fractions with purifiedSGPG/SGLPG were collected and analyzed by TLC. Again, the solvent wasevaporated, and the extract was lyophilized.

TLC and TLC Immunostaining. TLC was performed on Silica gel 60 F254-coatedglass plates (Merck) for 25–30 min using chloroform/methanol/0.2% KCl(50:40:10) as the eluent. Glycolipids were visualized by charring at 156 °Cwith mostain (0.02 M solution of ammonium cerium sulfate dihydrate andammonium molybdate tetrahydrate in aqueous 10% H2SO4). TLC immu-nostaining was performed as described by Ilyas et al. (50). The plate wasblocked in 0.1% poly(isobutyl methacrylate) in hexane for 60 s, then wasblocked for 30 min in blocking buffer (1% BSA, 0.1% Tween 20 in PBS), andfinally was incubated for 2 h with 1:400 diluted HNK-1 antibody (20) inblocking buffer. After two washing steps with PBS, the plate was incubatedfor 1 h with 1:4,000 diluted anti-mouse IgM alkaline phosphatase (AP)(μ-chain) (A3437; Sigma Aldrich) conjugate in blocking buffer. The plate waswashed twice with PBS and incubated with 3 mL of 1:200 diluted nitro bluetetrazolium/5-bromo-4-chloro-3-indolyl-phosphate (NBT/BCIP) in develop-ment buffer [0.1 M Tris base (pH 8.8), 0.1 M NaCl, 5 mM MgCl]. The colorreaction was stopped after 20 min, and the plates were visualized with a GelDoc XR+ Reader (Bio-Rad Laboratories).

Animals. Immunization was performed based on the protocol of Ilyas et al.(13). Groups of four to six male BALB/c wild-type mice at the age of 6–8 wkwere injected s.c. at multiple sites on the lower back with a total of 100 μg ofpurified SGPG/SGLPG (in PBS) mixed with KLH (1.4 mg/mL final concentration)and emulsified with an equal volume of TiterMax Gold. Two booster injectionswere performed after 2 and 4 wk with 20 μg of purified SGPG/SGLPG mixedwith KLH and TiterMax Gold. Control mice were injected with the same mix-ture but without SGPG/SGLPG. The same immunization protocol was appliedfor the induction of ADAs (immunogenicity assay), but mice were immunizedwith PL84(mimHNK-1)45, together with KLH and TiterMax Gold. Blood sampleswere taken by puncture of the tail vein, were transferred to tubes containing1 μL of 0.5 M EDTA, and were centrifuged for 15 min at 1,800 × g. The su-pernatant (plasma) was transferred to new tubes and stored at −55 °C. Theglycopolymer (or control polymer), dissolved in PBS, was administered by i.v.injection of the tail vein. Animal experiments were conducted in accordancewith Permit no. 2778 of the Animal Research Authorities of the Canton Basel-Stadt, Switzerland.

In Vivo Pharmacokinetics. A single dose (10 mg/kg, in PBS) of PL84(mimHNK-1)45 was administered via the tail vein in control-immunized BALB/c miceadministered KLH and TiterMax Gold without SGPG/SGLPG antigen. Bloodsamples were taken by tail vein puncture at multiple time points after in-jection (i.e., 5, 15, 30 min, 1, 2, 4, 8, 12, and 24 h) and were transferred into

8 of 10 | www.pnas.org/cgi/doi/10.1073/pnas.1619386114 Herrendorff et al.

Dow

nloa

ded

by g

uest

on

June

23,

202

0

tubes containing 1 μL of 0.5 M EDTA. The plasma was isolated by centrifu-gation at 1,800 × g for 15 min and was stored at −55 °C. To detect the gly-copolymer by sandwich ELISA, we used two recombinant monoclonal antibody(Fab) fragments targeting the fully thioglycerol capped poly-L-lysine backbone[PL40–60(mimHNK-1)0], i.e., a capture Fab fragment (AbD27381; Bio-Rad Labo-ratories) and a detection Fab fragment (AbD27389; Bio-Rad Laboratories).MaxiSorp plates (442404; Thermo Fisher,) were coated overnight at 4 °Cwith 50 μL (5 ng/μL) of the capture Fab fragment in PBS. Plates were blockedfor 2 h at room temperature with 5% BSA in PBS containing 0.1% Tween20(PBST). Plasma samples were diluted in PBST to obtain a signal in the linearrange (1:5–1:16,000) and were incubated at room temperature for 2 h (50-μLworking volume). PL84(mimHNK-1)45 in PBST (0.001–8 ng/μL) was used as astandard. The detection Fab fragment, which was conjugated to HRP with theLYNX rapid HRP antibody conjugation kit (LNK006P; Bio-Rad Laboratories),was diluted in 0.5% BSA/PBST at a concentration of 2 ng/μL and incubated atroom temperature for 1 h (50-μL working volume). After each step, the platewas washed five times with PBST (250 μL). Finally, 50 μL of the substrate3,3′,5,5′-tetramethylbenzidine (TMB) (34028;Thermo Fisher Scientific) wasadded, and after 5 min incubation the color reaction was stopped with 1 MH2SO4. OD450 was measured on a microplate reader (SpectraMax 190; Molec-ular Devices). Data analysis was performed with GraphPad Prism 5.0 software(GraphPad Software, Inc.) and PK Solver (61). The plasma half-life was calcu-lated with PK Solver using the one-compartment (i.v. bolus) model.

Immunogenicity Assay (ADA Detection ELISA). Nonimmunized mice underwentdaily treatment with 10 mg/kg of PL84(mimHNK-1)45 (i.v.) for 10 consecutivedays. Blood samples were taken weekly via the tail vein starting 24 h after thelast injection, and plasma was prepared as described earlier for further anal-ysis. As a positive control, blood samples were taken from mice immunizedwith PL84(mimHNK-1)45 and adjuvants, according to the immunization pro-tocol described above. For the detection of ADAs, MaxiSorp plates (442404;Thermo Fisher) were coated overnight at 4 °C with 50 μL (0.1 μg/mL) ofPL84(mimHNK-1)45 in PBS. They were washedwith 0.1% PBST and blocked with5% BSA in 0.1% PBST for 2 h at room temperature. Plasma samples were di-luted 1:100 in 2.5% BSA/PBST and were incubated (50 μL) on the plate for 2 hat room temperature immediately after blocking (without washing). Next,ADA detection was done by a 2-h incubation at room temperature with a goatanti-mouse IgM HRP conjugate (A8786l; Sigma Aldrich) or with a goat anti-mouse IgG HRP conjugate (A4416; Sigma Aldrich) diluted 1:10,000 in 1% BSA/PBST. Finally, plates were washed and incubated with 50 μL of TMB substratefor 30 min, after which the color reaction was stopped with 1 M of H2SO4. ODwas measured at 450 nm with a microplate reader, and data analysis wasperformed with Prism software.

Myelin Preparation. Freshly frozen human postmortem corpus callosum (whitematter) was homogenized with a 12-mm Polytron (2 × 10 s) in ice-cold 0.25 M

sucrose, 10 mM Hepes (pH 7.4), 2 mM EGTA with a protease inhibitor mixture(1 μg/mL aprotinin, 2 μg/mL leupeptin, 1 μg/mL pepstatin, 100 μg/mL PMSF).The suspension was brought up to an end concentration of 1.4 M sucrose,10 mM Hepes (pH 7.4), and 2 mM EGTA. A four-step discontinuous gradientwas set up in a 14 × 95 mm Polyallomer Beckmann tube with 1.1 mL of 2 Msucrose, 10 mM Hepes (pH 7.4), 2 mM EGTA solution at the bottom of thetube; then with 8.5 mL of 1.4 M sucrose, 10 mM Hepes (pH 7.4), and 2 mMEGTA brain suspension layered carefully above; then 2.2 mL 0.85 M sucrose,10 mM Hepes (pH 7.4), and 2 mM EGTA; and finally 0.75 mL of 0.25 M sucrose,10 mM Hepes (pH 7.4), and 2 mM EGTA on the top. After centrifugation for20 h at 25,000 rpm and 4 °C (Beckmann Coulter L-70K free-swing rotor ultra-centrifuge), myelin membranes were enriched in the 0.25 M sucrose fraction,and plasma membranes were located at the interface to the 0.85-M sucrosefraction. Myelin membranes were collected and homogenized in 15 volumesof cold 10 mM Hepes (pH 7.4), 2 mM EGTA and were centrifuged (25,000 rpmfor 2 h at 4 °C). The supernatant was discarded, and the pellet was resus-pended in sterile water and stored at −80 °C.

SDS/PAGE and Western Blotting. The protein concentration of the human CNSmyelin extract was determined with a bicinchoninic acid assay (SigmaAldrich). Samples (15 μg) were separated by Tris/glycine SDS/PAGE on 8%gels and were analyzed by Western blotting using Protran BA85 nitrocel-lulose membranes (GE Healthcare). Membranes were blocked for 1.5 h inblocking buffer [3% BSA, 20 mM Tris base (pH 7.4), 0.1 M NaCl, 0.05% so-dium azide]. The same buffer was used to dilute patient serum samples,mouse plasma samples, and the control antibody. Patient sera (patients MKand DP) were diluted to 1:800, and plasma from immunized mice was di-luted to 1:400. A rabbit anti-human L-MAG antibody (1:1,000) was used ascontrol (39). The membrane was incubated with the antibody overnightat 4 °C. The secondary antibodies, the AP-conjugated anti-human IgM(1:20,000; A3437; Sigma Aldrich), goat anti-mouse IgM (1:20,000; A9688;Sigma Aldrich), and goat anti-rabbit IgG (1:40,000; A3687; Sigma Aldrich),were incubated with the membranes for 1 h at room temperature. Mem-branes were developed for 15–30 min in NBT/BCIP diluted 1:200 in devel-opment buffer. The membranes were imaged with a Gel Doc XR+ reader(Bio-Rad Laboratories).

ACKNOWLEDGMENTS. We thank Axel Regeniter (University Hospital Basel)for collecting and providing patient and control serum samples; MichaelSinnreich and Jochen Kinter (Department of Biomedicine, University Hospi-tal Basel, University of Basel) for supporting us with the setup of the animalexperiments; and Renato Cotti (Bühlmann Laboratories AG) for providing uswith ELISA kits. This work was funded by the Swiss Commission for Technol-ogy and Innovation, the Neuromuscular Research Association Basel, and theGebert Rüf Stiftung.

1. Talamo G, Mir MA, Pandey MK, Sivik JK, Raheja D (2015) IgM MGUS associated withanti-MAG neuropathy: A single institution experience. Ann Hematol 94:1011–1016.

2. Braun PE, Frail DE, Latov N (1982) Myelin-associated glycoprotein is the antigen for amonoclonal IgM in polyneuropathy. J Neurochem 39:1261–1265.

3. Steck AJ, Murray N, Meier C, Page N, Perruisseau G (1983) Demyelinating neuropathyand monoclonal IgM antibody to myelin-associated glycoprotein. Neurology 33:19–23.

4. Steck AJ, Stalder AK, Renaud S (2006) Anti-myelin-associated glycoprotein neuropa-thy. Curr Opin Neurol 19:458–463.

5. Dalakas MC (2010) Pathogenesis and treatment of anti-MAG neuropathy. Curr TreatOptions Neurol 12:71–83.

6. Quarles RH (2007) Myelin-associated glycoprotein (MAG): Past, present and beyond.J Neurochem 100:1431–1448.

7. Lombardi R, et al. (2005) IgM deposits on skin nerves in anti-myelin-associated gly-coprotein neuropathy. Ann Neurol 57:180–187.

8. Ritz MF, et al. (1999) Anti-MAG IgM penetration into myelinated fibers correlateswith the extent of myelin widening. Muscle Nerve 22:1030–1037.

9. Hays AP, Lee SSL, Latov N (1988) Immune reactive C3d on the surface of myelinsheaths in neuropathy. J Neuroimmunol 18:231–244.

10. Monaco S, et al. (1990) Complement-mediated demyelination in patients with IgMmonoclonal gammopathy and polyneuropathy. N Engl J Med 322:649–652.

11. Hays AP, Latov N, Takatsu M, Sherman WH (1987) Experimental demyelination of nerveinduced by serum of patients with neuropathy and an anti-MAG IgM M-protein.Neurology 37:242–256.

12. Tatum AH (1993) Experimental paraprotein neuropathy, demyelination by passivetransfer of human IgM anti-myelin-associated glycoprotein. Ann Neurol 33:502–506.

13. Ilyas AA, Gu Y, Dalakas MC, Quarles RH, Bhatt S (2008) Induction of experimentalataxic sensory neuronopathy in cats by immunization with purified SGPG. J Neuroimmunol193:87–93.

14. Voshol H, van Zuylen CW, Orberger G, Vliegenthart JFG, Schachner M (1996) Structureof the HNK-1 carbohydrate epitope on bovine peripheral myelin glycoprotein P0.J Biol Chem 271:22957–22960.

15. Kelm S, et al. (1994) Sialoadhesin, myelin-associated glycoprotein and CD22 define anew family of sialic acid-dependent adhesion molecules of the immunoglobulin su-perfamily. Curr Biol 4:965–972.

16. Crocker PR, Paulson JC, Varki A (2007) Siglecs and their roles in the immune system.Nat Rev Immunol 7:255–266.

17. Sternberger NH, Quarles RH, Itoyama Y, Webster HD (1979) Myelin-associated gly-coprotein demonstrated immunocytochemically in myelin and myelin-forming cellsof developing rat. Proc Natl Acad Sci USA 76:1510–1514.

18. Erb M, et al. (2006) Unraveling the differential expression of the two isoforms ofmyelin-associated glycoprotein in a mouse expressing GFP-tagged S-MAG specificallyregulated and targeted into the different myelin compartments. Mol Cell Neurosci31:613–627.

19. Gabriel JM, et al. (1998) Confocal microscopic localization of anti-myelin-associatedglycoprotein autoantibodies in a patient with peripheral neuropathy initially lackinga detectable IgM gammopathy. Acta Neuropathol 95:540–546.

20. Gabriel JM, et al. (1996) Selective loss of myelin-associated glycoprotein from myelincorrelates with anti-MAG antibody titre in demyelinating paraproteinaemic poly-neuropathy. Brain 119:775–787.

21. Yin X, et al. (1998) Myelin-associated glycoprotein is a myelin signal that modulatesthe caliber of myelinated axons. J Neurosci 18:1953–1962.

22. Weiss MD, Luciano CA, Quarles RH (2001) Nerve conduction abnormalities in agingmice deficient for myelin-associated glycoprotein. Muscle Nerve 24:1380–1387.

23. Cai Z, et al. (2001) Focal myelin swellings and tomacula in anti-MAG IgM para-proteinaemic neuropathy: Novel teased nerve fiber studies. J Peripher Nerv Syst 6:95–101.

24. Lunn MPT, Crawford TO, Hughes RAC, Griffin JW, Sheikh KA (2002) Anti-myelin-associated glycoprotein antibodies alter neurofilament spacing. Brain 125:904–911.

25. Burger D, Pidoux L, Steck AJ (1993) Identification of the glycosylated sequons ofhuman myelin-associated glycoprotein. Biochem Biophys Res Commun 197:457–464.

26. Ogino M, Tatum AH, Latov N (1994) Affinity studies of human anti-MAG antibodiesin neuropathy. J Neuroimmunol 52:41–46.

Herrendorff et al. PNAS Early Edition | 9 of 10

NEU

ROSC

IENCE

PNASPL

US

Dow

nloa

ded

by g

uest

on

June

23,

202

0

27. Ariga T (2011) The role of sulfoglucuronosyl glycosphingolipids in the pathogenesis ofmonoclonal IgM paraproteinemia and peripheral neuropathy. Proc Jpn Acad Ser BPhys Biol Sci 87:386–404.

28. Lunn MPT, Nobile-Orazio E (2012) Immunotherapy for IgM anti-myelin-associatedglycoprotein paraprotein-associated peripheral neuropathies. Cochrane DatabaseSyst Rev (5):CD002827.

29. Broglio L, Lauria G (2005) Worsening after rituximab treatment in anti-mag neu-ropathy. Muscle Nerve 32:378–379.

30. Stork ACJ, Notermans NC, Vrancken AFJE, Cornblath DR, van der Pol WL (2013) Rapidworsening of IgM anti-MAG demyelinating polyneuropathy during rituximab treat-ment. J Peripher Nerv Syst 18:189–191.

31. Benedetti L, et al. (2007) Predictors of response to rituximab in patients with neu-ropathy and anti-myelin associated glycoprotein immunoglobulin M. J Peripher NervSyst 12:102–107.

32. Nobile-Orazio E, et al. (1988) Treatment of patients with neuropathy and anti-MAGIgM M-proteins. Ann Neurol 24:93–97.

33. Renaud S, et al. (2006) High-dose rituximab and anti-MAG-associated polyneuro-pathy. Neurology 66:742–744.

34. Pestronk A, et al. (2003) Treatment of IgM antibody associated polyneuropathiesusing rituximab. J Neurol Neurosurg Psychiatry 74:485–489.

35. Benedetti L, et al. (2008) Long-term effect of rituximab in anti-mag polyneuropathy.Neurology 71:1742–1744.

36. Tokuda A, et al. (1998) On the specificity of anti-sulfoglucuronosyl glycolipid anti-bodies. J Carbohydr Chem 17:535–546.

37. Kornilov AV, Sherman AA, Kononov LO, Shashkov AS, Nifant’ev NE (2000) Synthesis of3-O-sulfoglucuronyl lacto-N-neotetraose 2-aminoethyl glycoside and biotinylatedneoglycoconjugates thereof. Carbohydr Res 329:717–730.

38. Burger D, Perruisseau G, Steck AJ (1991) Anti-myelin-associated glycoprotein anti-bodies in patients with a monoclonal IgM gammopathy and polyneuropathy, and asimplified method for the preparation of glycolipid antigens. J Immunol Methods140:31–36.

39. Miescher GC, et al. (1997) Reciprocal expression of myelin-associated glycoproteinsplice variants in the adult human peripheral and central nervous systems. Brain ResMol Brain Res 52:299–306.

40. Page N, Murray N, Perruisseau G, Steck AJ (1985) A monoclonal anti-idiotypic anti-body against a human monoclonal IgM with specificity for myelin-associated glyco-protein. J Immunol 134:3094–3099.

41. Fluri F, Ferracin F, Erne B, Steck AJ (2003) Microheterogeneity of anti-myelin-associated glycoprotein antibodies. J Neurol Sci 207:43–49.

42. Ilyas AA, et al. (1984) IgM in a human neuropathy related to paraproteinemia binds toa carbohydrate determinant in the myelin-associated glycoprotein and to a gangli-oside. Proc Natl Acad Sci USA 81:1225–1229.

43. Tsvetkov YE, et al. (2012) Synthesis and molecular recognition studies of the HNK-1trisaccharide and related oligosaccharides. The specificity of monoclonal anti-HNK-1antibodies as assessed by surface plasmon resonance and STD NMR. J Am Chem Soc134:426–435.

44. Davis BG (1999) Recent developments in glycoconjugates. J Chem Soc Perk T1:3215–3237.

45. Sato S, et al. (1989) cDNA cloning and amino acid sequence for human myelin-associated glycoprotein. Biochem Biophys Res Commun 163:1473–1480.

46. Spagnol G, et al. (1989) Molecular cloning of human myelin-associated glycoprotein.J Neurosci Res 24:137–142.

47. Duthaler RO, et al. (2010) In vivo neutralization of naturally existing antibodiesagainst linear alpha(1,3)-galactosidic carbohydrate epitopes by multivalent antigenpresentation: A solution for the first hurdle of pig-to-human xenotransplantation.Chimia (Aarau) 64:23–28.

48. Willison HJ, et al. (1988) Demyelination induced by intraneural injection of humanantimyelin-associated glycoprotein antibodies. Muscle Nerve 11:1169–1176.

49. Yamawaki M, et al. (1996) Generation and characterization of anti-sulfoglucuronosylparagloboside monoclonal antibody NGR50 and its immunoreactivity with peripheralnerve. J Neurosci Res 44:586–593.

50. Ilyas AA, Chen ZW, Prineas JW (2002) Generation and characterization of antibodiesto sulfated glucuronyl glycolipids in Lewis rats. J Neuroimmunol 127:54–58.

51. Tucker GC, Delarue M, Zada S, Boucaut JC, Thiery JP (1988) Expression of the HNK-1/NC-1 epitope in early vertebrate neurogenesis. Cell Tissue Res 251:457–465.

52. Ilyas AA, Dalakas MC, Brady RO, Quarles RH (1986) Sulfated glucuronyl glycolipidsreacting with anti-myelin-associated glycoprotein monoclonal antibodies includingIgM paraproteins in neuropathy: Species distribution and partial characterizationof epitopes. Brain Res 385:1–9.

53. Byrne GW, et al. (2002) Evaluation of different alpha-Galactosyl glycoconjugates foruse in xenotransplantation. Bioconjug Chem 13:571–581.

54. Katopodis AG, et al. (2002) Removal of anti-Galalpha1,3Gal xenoantibodies with aninjectable polymer. J Clin Invest 110:1869–1877.

55. Cebo C, et al. (2002) Function and molecular modeling of the interaction betweenhuman interleukin 6 and its HNK-1 oligosaccharide ligands. J Biol Chem 277:12246–12252.

56. Kleene R, Schachner M (2004) Glycans and neural cell interactions. Nat Rev Neurosci5:195–208.

57. Usuki S, Taguchi K, Thompson SA, Chapman PB, Yu RK (2010) Novel anti-idiotypeantibody therapy for lipooligosaccharide-induced experimental autoimmune neuri-tis: Use relevant to Guillain-Barré syndrome. J Neurosci Res 88:1651–1663.

58. Usuki S, Taguchi K, Gu YH, Thompson SA, Yu RK (2010) Development of a noveltherapy for Lipo-oligosaccharide-induced experimental neuritis: Use of peptide gly-comimics. J Neurochem 113:351–362.

59. Holgersson J, Gustafsson A, Breimer ME (2005) Characteristics of protein-carbohydrateinteractions as a basis for developing novel carbohydrate-based antirejection therapies.Immunol Cell Biol 83:694–708.

60. Willison HJ, et al. (2004) Synthetic disialylgalactose immunoadsorbents deplete anti-GQ1b antibodies from autoimmune neuropathy sera. Brain 127:680–691.

61. Zhang Y, Huo M, Zhou J, Xie S (2010) PKSolver: An add-in program for pharmaco-kinetic and pharmacodynamic data analysis in Microsoft Excel. Comput MethodsPrograms Biomed 99:306–314.

10 of 10 | www.pnas.org/cgi/doi/10.1073/pnas.1619386114 Herrendorff et al.

Dow

nloa

ded

by g

uest

on

June

23,

202

0