Anti–neurofascin-155 IgG4 antibodies prevent paranodal complex formation in vivo Constance Manso, … , Isabel Illa, Jérôme J. Devaux J Clin Invest. 2019;129(6):2222-2236. https://doi.org/10.1172/JCI124694. Research Article Autoimmunity Neuroscience Neurofascin-155 (Nfasc155) is an essential glial cell adhesion molecule expressed in paranodal septate-like junctions of peripheral and central myelinated axons. The genetic deletion of Nfasc155 results in the loss of septate-like junctions and in conduction slowing. In humans, IgG4 antibodies against Nfasc155 are implicated in the pathogenesis of chronic inflammatory demyelinating polyneuropathy (CIDP). These antibodies are associated with an aggressive onset, a refractoriness to intravenous immunoglobulin, and tremor of possible cerebellar origin. Here, we examined the pathogenic effects of patient-derived anti-Nfasc155 IgG4. These antibodies did not inhibit the ability of Nfasc155 to complex with its axonal partners contactin-1 and CASPR1 or induce target internalization. Passive transfer experiments revealed that IgG4 antibodies targeted Nfasc155 on Schwann cell surfaces, and diminished Nfasc155 protein levels and prevented paranodal complex formation in neonatal animals. In adult animals, chronic intrathecal infusions of antibodies also induced the loss of Nfasc155 and of paranodal specialization and resulted in conduction alterations in motor nerves. These results indicate that anti-Nfasc155 IgG4 antibodies perturb conduction in the absence of demyelination, validating the existence of paranodopathy. These results also shed light on the mechanisms regulating protein insertion at paranodes. Find the latest version: https://jci.me/124694/pdf RESEARCH ARTICLE The Journal of Clinical Investigation   Anti–neurofascin-155 IgG4 antibodies prevent paranodal complex formation in vivo Constance Manso,1,2 Luis Querol,3,4 Cinta Lleixà,3,4 Mallory Poncelet,5 Mourad Mekaouche,1,6 Jean-Michel Vallat,7 Isabel Illa,3,4 and Jérôme J. Devaux1,5 Aix Marseille Université, CNRS, CRN2M-UMR7286, Marseille, France. 2Université de Bordeaux, Interdisciplinary Institute for Neuroscience, UMR5297, Bordeaux, France. 3Neuromuscular Diseases 1 Unit, Hospital de la Santa Creu i Sant Pau, Universitat Autónoma de Barcelona, Barcelona, Spain. 4Centro para la Investigación en Red en Enfermedades Raras (CIBERER), Madrid, Spain. 5Institute for Neurosciences of Montpellier, INSERM U1051, Montpellier University, Hopital Gui de Chauliac, Montpellier, France. 6Aix Marseille Université, CNRS, INP UMR7051, Marseille, France. 7National Reference Center for “rare peripheral neuropathies” and Department of Neurology, University Hospital, Limoges, France. Neurofascin-155 (Nfasc155) is an essential glial cell adhesion molecule expressed in paranodal septate-like junctions of peripheral and central myelinated axons. The genetic deletion of Nfasc155 results in the loss of septate-like junctions and in conduction slowing. In humans, IgG4 antibodies against Nfasc155 are implicated in the pathogenesis of chronic inflammatory demyelinating polyneuropathy (CIDP). These antibodies are associated with an aggressive onset, a refractoriness to intravenous immunoglobulin, and tremor of possible cerebellar origin. Here, we examined the pathogenic effects of patient-derived anti-Nfasc155 IgG4. These antibodies did not inhibit the ability of Nfasc155 to complex with its axonal partners contactin-1 and CASPR1 or induce target internalization. Passive transfer experiments revealed that IgG4 antibodies targeted Nfasc155 on Schwann cell surfaces, and diminished Nfasc155 protein levels and prevented paranodal complex formation in neonatal animals. In adult animals, chronic intrathecal infusions of antibodies also induced the loss of Nfasc155 and of paranodal specialization and resulted in conduction alterations in motor nerves. These results indicate that anti-Nfasc155 IgG4 antibodies perturb conduction in the absence of demyelination, validating the existence of paranodopathy. These results also shed light on the mechanisms regulating protein insertion at paranodes. Introduction sodium channel (Nav) clusters at the nodes of Ranvier where the Chronic inflammatory demyelinating polyneuropathy (CIDP) action potentials are regenerated (13, 14). It was thus speculated is a syndrome affecting peripheral nerves and leading to signifi- that these autoantibodies may affect conduction by altering the cant disability (1, 2). Although it is widely accepted that this dis- axoglial contacts, axonal physiology, or myelin insulation. Indeed, order has an autoimmune pathogenesis, the exact mechanisms Nfasc186 is expressed at the nodal axolemma, where it interacts leading to nerve conduction slowing or conduction loss are still with gliomedin and NrCAM, 2 glial cell adhesion molecules. This elusive. Recent evidence indicates that autoantibodies targeting complex is crucial for the initial clustering of Nav channels at hem- cell adhesion molecules at the nodes of Ranvier are implicated inodes (15), but also seems important for the stabilization of Nav in CIDP pathogenesis in a subset of patients. Autoantibodies to channels at adult nodes of Ranvier and of microvilli structure contactin-1 (CNTN1), neurofascin-155 (Nfasc155), CNTN1-asso- surrounding the nodes of Ranvier (16–18). Recent evidence indi- ciated protein-1 (CASPR1), and neurofascin-186 (Nfasc186) have cates that autoantibodies to Nfasc186 are associated with mor- been described in subgroups of patients showing distinct clinical phological alterations of the microvilli in sural biopsies from CIDP presentations (3–10). In the case of antibodies against Nfasc155, patients and the presence of reversible proximal conduction block the clinical phenotype is clearly distinct from that of typical CIDP (19). This suggests that anti-Nfasc186 antibodies may affect node patients and includes a predominantly distal phenotype, a low- function and lead to conduction failure. However, the exact mech- frequency and high-amplitude tremor with cerebellar features anisms of how these morphological alterations lead to functional (11), and a poor response to intravenous immunoglobulin (12). deficits remain to be demonstrated. These cell adhesion molecules play important roles in nerve phys- At paranodal regions, the association of the axonal protein iology by enabling the formation and stability of the voltage-gated CNTN1 and CASPR1 with the glial protein Nfasc155 forms an axo­ glial complex that is crucial for the formation of atypical adhesive junctions named septate-like junctions (20–22). These junctions Conflict of interest: JJD received a research grant from CSL Behring. II holds a patent spiral around the axons along the paranodes and help myelin insu- for dysferlin detection in monocytes (US20030165937A1), and has consulted for lation by decreasing the extracellular space between the paranodal Grifols, Genzyme, Alexion, and UCB. myelin loop and the axon. The presence of autoantibodies against Copyright: © 2019, American Society for Clinical Investigation. Submitted: September 7, 2018; Accepted: March 12, 2019; Published: April 29, 2019. CNTN1, Nfasc155, or CASPR1 was found to be associated with Reference information: J Clin Invest. 2019;129(6):2222–2236. alterations of the paranodal complex CNTN1/CASPR1/Nfasc155 https://doi.org/10.1172/JCI124694. and loss of septate-like junctions in biopsies from CIDP patients 2222 jci.org   Volume 129   Number 6   June 2019 The Journal of Clinical Investigation   RESEARCH ARTICLE (5, 6, 23–25). These autoantibodies are predominantly of the IgG4 We next investigated whether anti-Nfasc155 IgG4 may affect isotype and are suspected to interfere with the function or struc- the interaction between Nfasc155 and its axonal partners CNTN1 ture of their target antigens to exert their deleterious effects (26, and CASPR1 using a cell aggregation assay. For that purpose, 27). Anti-CNTN1 IgG4 antibodies were recently shown to abolish HEK293 cells were transfected with mCherry-conjugated rat the interaction between CNTN1/CASPR1 and Nfasc155 (28) and Nfasc155, or with CNTN1 and GFP-tagged CASPR1. Cells were to penetrate the paranodal regions (29). Particularly, chronic expo- then mixed together and incubated for 2 hours under gentle agi- sure to these antibodies induces the loss of paranodal specializa- tation to enable cell aggregate formation. The percentage of cell tion and conduction (29). The mechanisms of how anti-Nfasc155 aggregates showing contacts between mCherry-expressing (red) IgG4 antibodies affect conduction are, by contrast, yet unknown. and GFP-expressing (green) cells was then quantified, as well as Here, we purified anti-Nfasc155 IgG4 antibodies from CIDP the percentage of green cells per aggregate. As negative controls, patients and tested their pathogenic function in CNTN1/CASPR1/ Nfasc155-expressing cells were incubated with cells express- Nfasc155 interaction and in animal models. We found that anti- ing GFP alone. Under such conditions, minimal interaction was Nfasc155 IgG4 antibodies act in a different manner compared observed between Nfasc155- and GFP-expressing cells (Figure with antibodies against CNTN1. These autoantibodies do not 1), and most cell aggregates contained either only Nfasc155- determine a functional blocking of Nfasc155, and do not penetrate expressing cells or only GFP-expressing cells. By contrast, the paranodal regions. Instead, these antibodies bind to Nfasc155 Nfasc155-expressing cells formed numerous interactions with on the surface of Schwann cells, induce the selective depletion of CNTN1/CASPR1–expressing cells, and most cell aggregates con- Nfasc155 in the peripheral nerves, and thus preclude the forma- tained 50% green and red cells (Figure 1). To test the efficacy of tion of the paranodal axoglial junction. These data indicate that anti-Nfasc155 IgG4 to dismantle these aggregates, 10 μg of IgG4 the pathogenic mechanisms leading to conduction abnormalities was added to the cell mixture during the 2-hour incubation step. are complex and involve different immune processes. For comparison, anti-CNTN1 IgG4 antibodies were also tested. As previously described (28), anti-CNTN1 IgG4 potently inhibited Results the interaction of Nfasc155-expressing cells with CNTN1/ Anti-Nfasc155 IgG4 does not affect the interaction between CNTN1/ CASPR1. By contrast, anti-Nfasc155 IgG4 from the 3 distinct CASPR1 and Nfasc155. In order to test the pathogenic role of anti- CIDP patients did not decrease cell-cell interaction (Figure 1). Nfasc155 autoantibodies, IgG4 antibodies were purified using Interestingly, cell aggregates incubated with anti-Nfasc155 IgG4 CaptureSelect affinity matrix from plasmapheresis fluids or sera presented a higher percentage of Nfasc155-expressing cells. This of 3 CIDP patients presenting with high titers of anti-Nfasc155 indicated that anti-Nfasc155 IgG4 antibodies do not prevent the IgG4 (Supplemental Figure 1; supplemental material available interaction between Nfasc155 and its axonal partners CNTN1/ online with this article; https://doi.org/10.1172/JCI124694DS1). CASPR1, but may favor Nfasc155 clustering. These patients presented with a severe CIDP with predominant Anti-Nfasc155 IgG4 binds to surface antigens on Schwann cells. distal weakness and refractoriness to intravenous immunoglobu- We previously reported that the function-blocking activity of lin. Patients CIDP1 and CIDP2 also had disabling tremors. Patient anti-CNTN1 IgG4 enables these autoantibodies to dismantle the CIDP3 also had CNS involvement (optic neuropathy and periven- axoglial interaction and to penetrate the paranodal barrier (29). To tricular white matter lesions). The clinical presentation of patient determine whether anti-Nfasc155 IgG4 antibodies have a similar CIDP1 has been previously reported (9). The blood samples from pathogenic potential, rat sciatic nerve segments were incubated these patients showed a predominant IgG4 reactivity against in vitro for 1 hour with 10 μg of anti-Nfasc155 IgG4 purified from human Nfasc155 (27), and IgG4 binding to paranodal regions from the 3 distinct CIDP patients (Supplemental Figure 5). An important rat sciatic nerve fibers (Supplemental Figure 1). We thus focused IgG4 deposition was observed around nodes of Ranvier and at the this study on IgG4 isotype. As control, IgG4 antibodies were puri- surface of Schwann cells with all samples (Supplemental Figure 5). fied from healthy donor plasma. The purity of the IgG4 samples However, no IgG4 deposition was found at paranodes even after was confirmed by immunoblots (Supplemental Figure 2). 3 hours of incubation, and no alterations of the paranodal regions First, we examined the bioactivity and the specificity of the were observed. Incubation with 10 μg of control IgG4 did not gen- purified IgG4 samples against rat Nfasc155 and rat nervous tis- erate any staining (data not shown). To confirm these findings, sues, as our goal was to test the pathogenic function of these anti- intraneural injections of 10 μg of antibodies were performed in the bodies in rodent models. We confirmed that these purified IgG4 sciatic nerves of 1-month-old rats. IgG4 deposition was then mon- fractions reacted against rat Nfasc155 by Western blot against rat itored 1 and 3 days after the injection, and the sciatic nerve fibers brain protein samples (Supplemental Figure 2B), but also by immu- were immunostained for Nav channels and CNTN1 to label nodes nostaining on rat Nfasc155–expressing human embryonic kidney and paranodes, or for β-catenin to label the adherens junctions in 293 (HEK293) cells and on teased rat sciatic nerve fibers (Supple- noncompact myelin. Again, no IgG4 deposition was detected at mental Figures 3 and 4). The reactivity of IgG4 fractions against paranodes at 1 or 3 days after injection, and no alterations of the rat brain samples, rat Nfasc155, or rat sciatic nerves was abrogated paranodal regions were observed. Instead, an important IgG4 was after elimination of anti-Nfasc155 antibodies by immunoadsorp- seen along the adherens junctions on the surface of the Schwann tion against Nfasc155 (Supplemental Figures 2–4). This clearly cells at both 1 and 3 days after injection (arrows in Figure 2, A–C). indicated that the paranodal fixation was due to a specific reactivi- In addition, an important IgG4 deposition was observed around ty against Nfasc155. In keeping, no reactivity was found against rat the nodes of Ranvier (Figure 2, D and E). Confocal microscopy CASPR1/CNTN1 or rat Nfasc186 (Supplemental Figure 3). examination demonstrated that the IgG4 deposition did not colo- jci.org   Volume 129   Number 6   June 2019 2223 RESEARCH ARTICLE The Journal of Clinical Investigation   Figure 1. Antibodies to Nfasc155 do not alter the interaction between Nfasc155 and its axonal partners CNTN1 and CASPR1. (A–C) For aggregation assays, HEK293 cells were cotransfected with mCherry-tagged Nfasc155 or with GFP-tagged CASPR1 and CNTN1. Cells were then incubated under gentle agitation for 2 hours in the presence of 10 μg of control IgG4, anti-CNTN1 IgG4, or anti-Nfasc155 IgG4 from 3 patients (CIDP1–3). As negative controls, Nfasc155-expressing HEK293 cells were incubated with cells expressing GFP alone (top left panel). Cells were examined with a fluorescence microscope at ×10 objective. Representative fields are shown in A (n = 3–4 experiments for each condition). Dashed circles highlight cell aggregates with contacts between red and green cells. The percentage of cell clusters with contacts between green and red cells was quantified (B), as well as the relative frequency of green cells per aggregate (C) (n = 3–4 experiments for each condition). CASPR1/CNTN1– and Nfasc155-expressing cells form clusters. Anti-CNTN1 IgG4 significantly prevented the formation of cell aggregates (**P < 0.005 by unpaired 2-tailed Student’s t tests for 2 samples of equal variance and by 1-way ANOVA followed by Bonferroni’s post hoc tests). By contrast, anti-Nfasc155 IgG4 did not affect the interaction between Nfasc155 and CASPR1/CNTN1. Bars represent mean and SEM. Scale bar: 50 μm. calize with Nav channels at nodal axolemma, but may be located found that anti-Nfasc155 IgG4 can also bind to Nfasc155 expressed on Schwann cell microvilli surrounding the nodes. IgG4 deposition on the surface of cultured oligodendrocytes (4). The acid wash appeared in a disc-shaped manner around the nodes by 1 day, and also completely removed IgG4 deposition from the surface of became juxtanodal by 3 days after injection (Figure 2, D and E). cultured oligodendrocytes (Supplemental Figure 7), further indi- To confirm that this staining pattern was caused by the anti- cating that anti-Nfasc155 IgG4 antibodies do not induce antigen Nfasc155 activity, we incubated sciatic nerve segments with IgG4 internalization. Surface-bound antibodies fully colocalized with fractions immunoadsorbed against Nfasc155. In the 3 samples, the Nfasc155 after permeabilization in cultured rat oligodendrocytes immunoadsorption completely abrogated the surface staining at (Supplemental Figure 7). Altogether, these results indicated that nodes and adherens junctions (Supplemental Figure 5). In order antibodies to Nfasc155 do not penetrate the paranodal domains, to further determine whether this staining pattern corresponded but bind to Nfasc155 expressed on the glial surface notably at to surface-bound antibodies or to internalized antibody/antigen adherens junctions and at the node vicinity. complexes, we performed an acid wash following incubation with Anti-Nfasc155 IgG4 depletes Nfasc155 and diminishes para­ the anti-Nfasc155 IgG4 (Supplemental Figure 6). The acid wash node formation. To determine whether anti-Nfasc155 antibodies completely removed the IgG4 deposition, indicating that IgG4 may impair paranode formation/maintenance, we performed i.p. antibodies were bound to surface antigens. In a previous study, we injections of anti-Nfasc155 IgG4 in neonatal animals before and 2224 jci.org   Volume 129   Number 6   June 2019 The Journal of Clinical Investigation   RESEARCH ARTICLE Figure 2. Anti-Nfasc155 autoantibodies target surface Schwann cell antigens. (A) Sciatic nerve fibers were incubated in vitro with purified anti-Nfasc155 IgG4 from patient CIDP1 for 3 hours, and immunolabeled for IgG4 (green) and CNTN1 (red). (B–E) Sciatic nerves were fixed 1 day (B and D) or 3 days (C and E) after intraneural injections of anti-Nfasc155 IgG4, and immunolabeled for IgG4 (green) and β-catenin (red; B and C) or CNTN1 (red; D and E) and Nav channels (blue; D and E). Note that anti-Nfasc155 IgG4 bound to the surface of the Schwann cells and deposited at the vicinity of the node of Ranvier (dou- ble arrowheads) and at adherens junctions along the internode stained here with β-catenin (arrows). However, no penetration across the paranodal region was observed (images are representative of n = 3 independent experiments). Scale bars: 10 μm. after paranode formation. In the PNS, it has been reported that tion, and 29% of the Nav channel clusters lacked paranodes by P2 the aggregation of Nfasc186 or gliomedin at heminodes border- (vs. 6% in control animals). As a result, the percentage of nodes ing the Schwann cells precedes that of Nfasc155 and CASPR1 at flanked by 2 CASPR1-positive paranodes was drastically dimin- paranodes (30, 31). In keeping, we found that at postnatal day ished at this age (35% vs. 66% in control animals). The remain- 0 (P0), most Nav channel aggregates lacked CASPR1 immuno- ing CASPR1-positive paranodes were shorter in comparison with reactivity at their borders (Figure 3). However, by P2 and P4, control animals (Figure 3D). By P4, the percentage of nodes and more than 90% of the Nav channel aggregates were flanked by heminodes reached normal values and was not significantly dif- CASPR1-positive paranodes. At these ages, the blood-nerve bar- ferent between Nfasc155 IgG4 and control groups. Nonetheless, rier is not functional and remains permeable until P13 (32). We the paranodal length was still significantly shorter in compari- thus performed injections of anti-Nfasc155 IgG4 (100 μg) at P0 son with control animals. By P6, no significant differences were and P4; then the animals were sacrificed 2, 4, and 6 days later. seen between Nfasc155 IgG4 and control groups. When the injec- As control, animals received similar doses of control IgG4. The tion of anti-Nfasc155 IgG4 was performed at P4, no significant percentages of Nav channel clusters with 1, 2, or no CASPR1- effects were observed on paranodal length or paranode forma- positive paranodes were then quantified at all ages tested (Figure tion either 2 or 6 days after injections (Supplemental Figure 9). 3 and Supplemental Figure 9). Because the amount of IgG4 puri- To confirm these findings, anti-Nfasc155 IgG4 antibodies from fied from patients CIDP2 and CIDP3 was limiting, we centered patients CIDP2 and CIDP3 were injected at P0, and the effects this study on the IgG4 fraction from patient CIDP1. In accor- on paranodal formation at P2 were monitored (Supplemental Fig- dance with our hypothesis, the injection of anti-Nfasc155 IgG4 ure 10, A–E). In keeping with the previous findings, the injection from patient CIDP1 at P0 strongly impaired paranode forma- of anti-Nfasc155 IgG4 from patients CIDP2 and CIDP3 potently jci.org   Volume 129   Number 6   June 2019 2225 RESEARCH ARTICLE The Journal of Clinical Investigation   Figure 3. Passive transfer of anti-Nfasc155 IgG4 affects the formation of paranodal axoglial unit during development. (A–D) Newborn rat pups received an i.p. injection of 250 μg of control IgG4 (A) or anti-Nfasc155 IgG4 from patient CIDP1 (B) on the day of birth (n = 4 animals for each condition and age). Sciatic nerve fibers were fixed and immunolabeled for voltage-gated sodium channels (Nav; green) and CASPR1 (red) at postnatal days 0 (P0), 2, 4, and 6. The percentages of Nav clusters with 1 or 2 flanking CASPR1-positive paranodes (double arrowheads) or without CASPR1-positive paran- odes (arrowheads) were quantified at each age (C), as well as the paranodal length (D) (n = 200–300 nodes or paranodes for each condition and age). Injection of anti- Nfasc155 IgG4 importantly delayed the formation of CASPR1-positive paranodes, and a significantly higher percentage of heminodes without flanking paranodes was observed at P2 (**P < 0.005 by 1-way ANOVA followed by Bonferroni’s post hoc tests). Paranodal length was also significantly shorter 2 and 4 days after injection of anti-Nfasc155 IgG4 (*P < 0.05 by 1-way ANOVA followed by Bonferroni’s post hoc tests). Scale bar: 10 μm. (E and F) Sciatic nerve and spinal cord proteins (100 μg) from P2 animals injected with control IgG4 (n = 4) or anti-Nfasc155 IgG4 (n = 4) at P0 were immunoblotted with antibodies recognizing all neurofascin isoforms (Pan-neurofas- cin) or specifically Nfasc155. The level of Nfasc155 was quantified relatively to that of Nfasc186 in both sciatic nerve and spinal cord (F). Nfasc155 level was significantly decreased in sciatic nerves of animals treated with anti-Nfasc155 IgG4, but not in spinal cord (**P < 0.005 by unpaired 2-tailed Student’s t tests for 2 samples of equal variance). Molecular weight markers are shown on the right (in kilodaltons). Bars represent mean and SEM. 2226 jci.org   Volume 129   Number 6   June 2019 The Journal of Clinical Investigation   RESEARCH ARTICLE Figure 4. Anti-Nfasc155 IgG4 deposits at proximity of nodes in neonatal animals. (A–C) Newborn rat pups received daily i.p. injection of 250 μg of control IgG4 (A) or anti-Nfasc155 IgG4 from patient CIDP1 (B) from the day of birth to P6 (images are representative of n = 3 independent experiments for each age). Sciatic nerve fibers were immunolabeled for voltage-gated sodium channels (Nav; blue), CASPR1 (green), and IgG (red) at P2 (A), P4 (B), and P6 (C). At P2 and P4, IgG deposits were mostly detected at the borders of Nav clusters, even in Nav clusters lacking CASPR1-positive paranodes (arrowheads). No colocalization with CASPR1 was found even in Nav clusters flanked by 1 or 2 paranodes (double arrowheads). At P6, IgG deposits appeared fainter and were mostly detected at the node vicinity. A weak IgG labeling can be observed at paranodes in some fibers. Isolated node labeling is shown at a higher magnification in the insets. Scale bars: 10 μm. impaired paranode formation, and resulted in a strong decrease The level of Nfasc155 was also examined in pups treated with of CASPR1-positive paranode length (Supplemental Figure 10E). anti-Nfasc155 IgG4 at P4 and sacrificed at P6. No significant dif- To determine whether paranode formation impairment is ferences were seen between control and Nfasc155 IgG4 groups linked to paranodal protein depletion or hypomyelination, sciatic (Supplemental Figure 9). The late injection of autoantibodies at nerve proteins were extracted at P2 and immunoblotted for neu- P4 thus did not appear to affect paranode stability or Nfasc155 rofascin isoforms, CASPR1, E-cadherin, and neurofilament-200 level. This indicated that anti-Nfasc155 IgG4 antibodies do not (Figure 3 and Supplemental Figure 8). In addition, myelination impact Nfasc155 particles that are stably incorporated at para­ was visualized on semithin nerve cross sections at P2 (Supple- nodes, but induce their elimination before their stabilization at mental Figure 8). No overt myelination defects were observed in paranodes. To further confirm this hypothesis, daily i.p. injec- Nfasc155 IgG4–treated pups on semithin sections. Also, levels of tions of anti-Nfasc155 IgG4 were performed from P0 to P6, and E-cadherin and neurofilament-200 were normal. The levels of paranode formation was examined at P4 and P6 (Supplemental neurofascin isoforms were quantified using pan-neurofascin anti- Figure 10, F–K). The extension of the infusion duration resulted bodies that recognize both Nfasc186 and the 2 forms of Nfasc155, in a significant impairment of paranode formation at P4 and in Nfasc155 high (Nfasc155-H) and low (Nfasc155-L)(33). Nfasc155 a significant shortening of paranodal length (Supplemental Fig- level was monitored using an antibody specific to Nfasc155 ure 10K). By P6, the percentage of Nav clusters flanked by 1 or 2 recognizing both forms (Figure 3). The level of Nfasc186 was CASPR1-positive paranodes and the paranodal length were not unchanged in both control and Nfasc155 IgG4 groups; however, significantly different from those in control animals. This indicat- the Nfasc155 level was significantly decreased in the sciatic nerves ed that a longer antibody infusion time further delayed paranode of anti-Nfasc155 IgG4–treated pups. CASPR1 level appeared nor- formation, but could not counteract the massive myelin formation mal in the sciatic nerves of the Nfasc155 IgG4 group (Supplemen- process taking place at P6. In order to determine whether these tal Figure 8), suggesting that Nfasc155 depletion does not impact alterations are associated with a loss of septate-like junctions, an the level of the other paranodal proteins. Also, Nfasc155 level was electron microscopic observation was done at P4 when the effects unchanged in the spinal cord, presumably because the blood-brain of autoantibodies are important. Despite the young age of the barrier, which is functional at birth, may have precluded antibody animals, many nodes of Ranvier were observed in both groups diffusion into the CNS. of animals. In control animals, numerous septate-like junctions jci.org   Volume 129   Number 6   June 2019 2227 RESEARCH ARTICLE The Journal of Clinical Investigation   Figure 5. The chronic intrathecal infusion of autoantibodies induces gait abnormalities and motor nerve conduction slowing. (A–C) Adult Lewis rats received daily intrathecal infusions (100 μg/d) of control IgG4 (black circles; n = 15 animals) or anti-Nfasc155 IgG4 from patient CIDP1 (gray circles; n = 15 animals) during 3 weeks. The clinical score was monitored daily and averaged. The passive infusion of anti-Nfasc155 IgG4 induced progressive clinical symptoms. (B and C) Footprint analysis revealed abnormal spreading of hind limbs in animals treated with anti-Nfasc155 IgG4 compared with control ani- mals. The footprint angle (gray lines) was significantly increased in animals treated with anti-Nfasc155 IgG4 compared with controls. (D–G) L6 ventral and dorsal roots from animals injected with control IgG4 or anti-Nfasc155 IgG4 were recorded on day 21 after the beginning of the injections (n = 12–14 nerves from 12–14 animals). Representative CAPs from control IgG4–treated (black traces) and anti-Nfasc155 IgG4–treated (gray traces) animals are shown in D and F. The peak amplitude, CAP duration, and conduction velocities at peak amplitude are represented in E and G for ventral and dorsal roots, respectively. The ventral spinal nerves of animals treated with anti-Nfasc155 IgG4 showed a significant decrease in CAP amplitude and conduction velocity compared with controls. This was associated with a significant increase in CAP duration. By contrast, nerve activity was not significantly affected in dorsal root. *P < 0.05, **P < 0.005 by unpaired 2-tailed Student’s t tests for 2 samples of equal variance. Bars represent mean and SD. were observed at paranodal loops bordering the nodes of Ran- paranodal loops with fewer septate-like junctions (Supplemental vier (Supplemental Figure 11). By contrast, animals treated with Figure 11). However, no alterations of the microvilli contacting the anti-Nfasc155 IgG4 presented important alterations at paranodal nodes of Ranvier were found. Altogether, this indicated that the regions (Supplemental Figure 11). Some axons completely lacked loss of CASPR1 labeling at paranodes was correlated with a disap- septate-like junctions at paranodes, whereas others presented pearance of septate-like junctions. 2228 jci.org   Volume 129   Number 6   June 2019 The Journal of Clinical Investigation   RESEARCH ARTICLE Figure 6. Anti-Nfasc155 IgG4 disrupts paranodal specialization in ventral spinal nerve. (A–F) Teased fibers from L6 ventral roots of animals treat- ed with control IgG4 (n = 15; A, C, and E) or anti-Nfasc155 IgG4 from patient CIDP1 (n = 15; B, D, and F) were stained for Nav channels (green), CNTN1 (red), and CASPR1 (blue, A and B), ankyrin-G (Ank-G; blue, C and D), or Nfasc186 (blue, E and F). Most nodes of Ranvier were positive for Nav channel, ankyrin-G, and Nfasc186 clusters and were bordered by paranodes stained for CASPR1 and CNTN1 in control animals. By contrast, many nodes lacked CASPR1 or CNTN1 staining at paranodes (arrowheads) in animals chronically injected with anti-Nfasc155 IgG4. Insets in B and D show representative affected nodes. Scale bars: 10 μm. Because IgG4 deposition onto the Schwann cell surface ing CASPR1-positive paranodes, indicating that IgG4 deposi- was detected after intraneural injection in adult animals, IgG4 tion may precede paranode complex formation, and affect its deposits were also investigated in sciatic nerves from P2, P4, formation. In some axons, the IgG4 deposits were intercalated and P6 animals (Figure 4). Strong IgG4 staining was observed in between CASPR1 and Nav channels (inset in Figure 4A). At at the border of Nav clusters at both P2 and P4 (Figure 4). How- P6, most Nav channel clusters were flanked by 2 paranodes, and ever, no colocalization with CASPR1 was detected, and the the IgG4 deposition was restricted to the nodal vicinity (double regions stained for IgG4 were devoid of CASPR1 labeling. In arrowheads in Figure 4C). Many axons presented a faint IgG4 most axons, IgG4 deposits flanked Nav channel clusters lack- deposition at paranodes that colocalized with CASPR1 at this jci.org   Volume 129   Number 6   June 2019 2229 RESEARCH ARTICLE The Journal of Clinical Investigation   2230 jci.org   Volume 129   Number 6   June 2019 The Journal of Clinical Investigation   RESEARCH ARTICLE Figure 7. Autoantibodies preferentially affect motor axons. (A–D) The (Figure 5, D and E). Compound action potentials (CAPs) from distribution of the axonal population according to the nodal diameters of ventral spinal roots of anti-Nfasc155 IgG4–treated animals had axons was examined in L6 ventral roots (A and B) and dorsal roots (C and D) of animals treated with control IgG4 (n = 1297 nodes for ventral roots a decreased amplitude in comparison with control IgG4 animals and 1061 nodes for dorsal roots from 11 animals) or anti-Nfasc155 IgG4 (Figure 5). This decreased activity was associated with a significant from patient CIDP1 (n = 1376 nodes for ventral roots and 875 nodes for slowing in nerve conduction velocity and a significant increase in dorsal roots from 11 animals). The respective proportion of normal nodes CAP duration suggestive of temporal dispersion. CAP recruitment (open boxes) or nodes lacking paranodes (hatched boxes) is represented on and refractory period were, however, not significantly impacted. the left, and the distribution of the total axonal population is represented on the right. The percentage of nodes showing paranodal alterations was By contrast to ventral roots, dorsal spinal nerves (Figure 5) or sci- calculated in ventral (B) and dorsal roots (D). Scatter plots represent the atic nerves (data not shown) did not exhibit conduction abnormal- percentage of normal nodes in each animal. Paranodes were more signifi- ities. This indicated that anti-Nfasc155 IgG4 induced conduction cantly affected in ventral root axons than in dorsal roots (**P < 0.005 by slowing in motor nerves mimicking a demyelinating phenotype. unpaired 2-tailed Student’s t tests for 2 samples of equal variance). (E and We then examined whether myelin alterations may be respon- F) The mean length of nodes lacking paranodes (hatched boxes) or appear- ing normal (open boxes) was measured in 11 animals for each group, as well sible for these conduction defects. No obvious signs of demyelin- as the mean paranodal length. No significant alterations in node or paran- ation or node disorganization were observed in ventral and dorsal ode length were observed between control and anti-Nfasc155 IgG4–treated spinal nerves from animals treated with anti-Nfasc155 IgG4 (Fig- animals (unpaired 2-tailed Student’s t tests and 1-way ANOVA followed ure 6 and Supplemental Figure 12). All nodes showed normal clus- by Bonferroni’s post hoc tests). In B and D–F, bars represents mean and ters of Nfasc186, Nav channels, and ankyrin-G. However, CNTN1 SEM. In A and C, box bounds represent the first and third quartiles, lines within the box represent the median, and whiskers show the minimal and and CASPR1 staining was often absent from paranodes in animals maximal values. treated with anti-Nfasc155 IgG4 (Figure 6). These alterations were more prominent in the ventral spinal nerves from anti-Nfasc155 IgG4–treated animals (36.1% ± 6.3% vs. 7.4% ± 3.1% in control IgG4; Figure 6) than in dorsal spinal nerves (4.8% ± 1.0% vs. 4.0% age, suggesting that a proportion of the antibody/antigen com- ± 1.0% in control IgG4; Figure 7 and Supplemental Figure 12). plex was incorporated at paranodes. Morphometric analysis did not reveal differences in nod- The chronic infusion of autoantibodies induces paranodal dis- al length between normal-appearing nodes and those lacking mantling and conduction alterations. The existence of protein paranodes. The paranodal length was not significantly different turnover at paranodes has been demonstrated in a tamoxifen- between Nfasc155 IgG4 and control groups. Also, paranodal alter- inducible Cre mouse line. In these animals, Nfasc155 levels were ations were not associated with a specific subset of axons. These shown to decline 20 days after Nfasc gene ablation in Schwann alterations were observed in both small- and large-diameter axons cells (34). We thus explored whether long-term exposure to anti- (Figure 7). The mean diameter of normal nodes and those lacking Nfasc155 IgG4 may alter Nfasc155 levels in adult animals and paranodes was not significantly different (2.1 ± 0.5 μm vs. 1.7 ± 0.4 affect paranode maintenance. For that purpose, chronic infusion μm, respectively; n = 15). of autoantibodies was implemented in naive animals. To bypass Anti-Nfasc155 IgG4 induces the selective depletion of Nfasc155. the blood-nerve barrier that can limit the diffusion of circulating We investigated whether the loss of paranodal specialization may antibodies to nerve fibers, an intrathecal catheter was implanted be associated with protein depletion. For that purpose, proteins into the rachidian canal of Lewis rats to enable infusion of anti- were extracted from ventral and dorsal nerve roots and immuno- body in the cerebrospinal fluid (CSF) bathing the lumbar spi- blotted for Nfasc155 and CASPR1, but also E-cadherin and neuro- nal roots. This procedure did not alter animal behavior or gait filament-200 (Figure 8). Nfasc155 protein level was significantly (Supplemental Video 1). Intrathecal infusion is relevant to CIDP decreased by nearly 40% in the ventral spinal nerves of animals pathology because anti-Nfasc155 IgG4 can be detected in the treated with anti-Nfasc155 IgG4 compared with control animals, CSF and prominent nerve root involvement has been described but was not altered in dorsal spinal nerves. These results matched in patients using magnetic resonance neurography (35, 36). Dai- the morphological alterations seen in ventral spinal nerves. No dif- ly injections of anti-Nfasc155 IgG4 from patient CIDP1 or control ference was observed in CASPR1 protein level in both ventral and IgG4 were performed for a period of 20 days. After 5 injections of dorsal spinal nerves in comparison with control animals. Similar- anti-Nfasc155 IgG4, all animals developed tail weakness or paral- ly, the protein levels of neurofilament-200 and E-cadherin were ysis (Figure 5). The clinical deficits then progressively worsened as unaffected. This was in keeping with the absence of axonal or treatment was pursued, and after 20 days, all animals presented a myelin alterations in these animals. complete tail paralysis, an unsteady gait, and paraparesis (Figure 5 and Supplemental Video 1). In contrast, none of the 15 animals Discussion treated with the same amount of control IgG4 developed signif- The principal finding of this study is that human anti-Nfasc155 icant symptoms. Footprint analysis showed increased outward IgG4 antibodies purified from CIDP patients induce clinical defi- rotation of the hind paws in animals treated with anti-Nfasc155 cits and nerve conduction abnormalities resembling those of CIDP IgG4 (Figure 5, B and C). The hindlimb grip strength test did not patients with anti-Nfasc155 antibodies. We provide evidence that reveal significant differences between Nfasc155 and control IgG4 these autoantibodies bind to the glial surface and cause Nfasc155 groups (data not shown). depletion and affect the formation of paranodal axoglial special- Electrophysiological examination was performed at 21 days ization. Interestingly, our findings indicate that anti-Nfasc155 and revealed important abnormalities in ventral spinal roots IgG4 and anti-CNTN1 IgG4 act through completely different jci.org   Volume 129   Number 6   June 2019 2231 RESEARCH ARTICLE The Journal of Clinical Investigation   Figure 8. Autoantibodies induce the selective depletion of Nfasc155 in ventral spinal nerve. (A) Ventral and dorsal spinal nerve proteins (250 μg) from adult animals injected with control IgG4 (n = 4–6) or anti-Nfasc155 IgG4 from patient CIDP1 (n = 4–6) for 3 weeks were sep- arated on SDS-PAGE gels and immunoblotted with antibodies against Nfasc155, CASPR1, neurofilament-200, and E-cadherin. Arrowheads indicate Nfasc155-H and Nfasc155-L isoforms. (B) The levels of Nfasc155, CASPR1, neurofilament-200, and E-cadherin were evalu- ated in animals treated with anti-Nfasc155 IgG4 relatively to those in control animals in both ventral and dorsal spinal roots. Nfasc155 levels were significantly decreased in ventral spinal nerves, but not in dorsal spinal nerves (*P < 0.05 by unpaired 2-tailed Student’s t tests for 2 samples of equal variance). Molecular weight markers are shown on the left (in kilodaltons). Bars represent mean and SEM. The passive transfer of anti-Nfasc155 IgG4 partially rep- licated the clinical deterioration seen in reactive patients. These latter present a neuropathy with a predominantly distal involvement associated with ataxia and disabling tremor (4, 8, 36, 37). The treated animals presented a severe phenotype akin to that of the patients but no tremor, and the electrophysi- ological and morphological findings suggested a predominant motor pathology. These apparent differences emphasized the limitations of animal models. Indeed, the determination of distal and proximal involvement, ataxia, or tremor relies on clinical tests difficult to perform in small mammals. None- theless, motor nerve conduction study revealed similitude between the patients and the animal model. Both showed conduction slowing and temporal dispersion on motor nerves (9, 24, 25, 36, 40). The passive transfer of anti-Nfasc155 IgG4 thus replicated these electrophysiological features, and sug- gested that paranode alterations could be responsible for the conduction slowing observed in the patients. In keeping with this view, sural nerve biopsies from CIDP patients with anti- pathogenic mechanisms. However, both autoantibodies lead to a Nfasc155 IgG4 revealed specific alterations at paranodal regions common pathogenic consequence: paranodal disorganization. with a loss of septate-like junctions (23–25), but did not show evi- The association of anti-Nfasc155 autoantibodies with CIDP dence of macrophage-mediated demyelination. The pathogenic has now been clearly proven; several independent studies have mechanisms responsible for conduction slowing in patients with reported that these autoantibodies are associated with a subgroup anti-Nfasc155 IgG4 thus appear different from those implicat- of CIDP patients showing homogeneous clinical features (4, 8, ed in other CIDP patients, and further support the existence of 9, 36, 37), pathological findings (23–25), and genetic background nodo-paranodopathy (41). It is possible that paranodal demyelin- (38). As demonstrated in our previous works (4, 9, 27), the antibody ation, as observed in sural nerve biopsies (24, 25), may also partic- isotype in Nfasc155-positive CIDP patients is almost exclusively ipate in conduction slowing or loss in patients with anti-Nfasc155 IgG4, and the other IgG isotypes are not necessarily concomitant- IgG4. Here, we did not observe paranodal demyelination on ly represented. Here, the samples that we used also presented an semithin sections (data not shown), demonstrating that the para­ exclusive IgG4 reactivity toward Nfasc155. We thus focused this nodal junction dismantling itself is sufficient to generate conduc- study on the pathogenic role of this isotype. The passive transfer tion slowing. In fact, the genetic ablation of CASPR1 or CNTN1 of anti-Nfasc155 IgG4 in adult Lewis rats demonstrated that these abrogates paranodal adhesive junctions and results in important antibodies are pathogenic and can mediate conduction slowing conduction slowing in the absence of myelin alterations (20, 21), and locomotor impairments. Two previous studies also reported demonstrating that these subtle changes are sufficient to impact that the passive transfer of mouse IgG against pan-neurofascin conduction velocity. The reason why we did not observe para­ exacerbates the clinical signs of experimental allergic neuritis nodal demyelination or sensory alterations in the animal model is in Lewis rats (8, 39). However, these latter antibodies targeted unclear. One possibility is that a longer exposure time to autoanti- both Nfasc186 and Nfasc155, and may thus more likely reflect body may be required to affect sensory axons or to induce paran- the pathogenic action of anti-Nfasc186/140 IgG (3). Here, anti- odal detachment. Also these features may be more prominent in Nfasc155 IgG4 derived from CIDP patients did not react against distal regions of the axon that could not be examined in our model. Nfasc186 or CASPR1/CNTN1, and only reacted against Nfasc155 The mechanisms of how anti-Nfasc155 IgG4 affects paranodal at paranodal regions. attachment appear to be distinct from those of anti-CNTN1 IgG4. 2232 jci.org   Volume 129   Number 6   June 2019 The Journal of Clinical Investigation   RESEARCH ARTICLE Figure 9. Schematic representation of the pathogenic mechanisms of anti–paranodal protein autoantibodies. (A) Representation of a mature node of Ranvier and myelinated axon. Myelin (yellow) covers the axon except at the node of Ranvier (red). At paranodal junctions (green), Nfasc155 interacts with its axonal partners CASPR1/CNTN1. Nfasc155 is also found at Schmidt-Lanterman incisures (dashed green lines). The Schwann cell nucleus is shown in gray. (B) In neonatal animals, the progressive enwrapping of axons by Schwann cells induces the formation of paranodal regions at node borders. The injection of anti-Nfasc155 IgG4 (blue) during the neonatal period does not affect myelination or node/paranode formation, but induces the depletion of Nfasc155, and thereby alters the formation of paranodal septate-like junctions. (C) At adult age, evidence suggests that the Nfasc155/CASPR1/CNTN1 complex is constant- ly renewed at paranodes possibly through degradation and replenishment mechanisms. The chronic infusion of anti-Nfasc155 IgG4 (blue) may preclude the regeneration of the paranodal axoglial junction by inducing Nfasc155 depletion, and thereby alter paranode structure and conduction. By contrast to anti-CNTN1 IgG4, anti-Nfasc155 IgG4 did not pres- in adult animals. By contrast, chronic antibody infusion induced ent a functional blocking activity and did not inhibit the interac- clinical deficits, Nfasc155 depletion, and paranodal specialization tion of rat Nfasc155 with the CNTN1/CASPR1 complex. Instead, loss, suggesting a blockade in axoglial junction maintenance. Evi- anti-Nfasc155 IgG4 seemed to favor Nfasc155 aggregation and dence suggests that paranodal proteins are permanently renewed induce its depletion. A similar observation was observed with during the lifespan (17, 34, 43). In genetic models, the ablation of human isoforms, and anti-Nfasc155 IgG4 only modestly impaired the Nfasc gene induced a drastic decrease in Nfasc155 protein level human Nfasc155/CASPR1/CNTN1 interaction by comparison within 20 days (34) — a time frame that is in accord with the delay with anti-CNTN1 IgG4. Also, anti-Nfasc155 IgG4 did not disman- in observing important clinical and morphological abnormalities tle the paranodal axoglial contact or penetrate paranodal regions in our model, and with the aggressive onset observed in reactive like anti-CNTN1 IgG4 (29), but bound to the Schwann cell surface. patients. The timeline to reconstitute paranodal axoglial junctions This binding was seen with IgG4 from diverse anti-Nfasc155–pos- is, however, yet unknown. Understanding the mechanisms regu- itive patients and was abolished after immunoadsorption against lating axoglial junction maintenance could be important to better Nfasc155, suggesting the presence of a cell surface Nfasc155 pool delineate the factors regulating functional recovery. at the node vicinity and along the adherens junctions. The pres- Our in vitro assays clearly demonstrated that anti-Nfasc155 ence of Nfasc155 along adherens junctions in Schmidt-Lanterman IgG4 does not have the potency to induce Nfasc155 internalization incisures has already been reported several times (34, 42), but not in HEK293 cells, Schwann cells, or oligodendrocytes. It thus seems its presence at the node vicinity. It is plausible that the high density unlikely that antibody-mediated endocytosis of Nfasc155 may of Nfasc155 at paranodes and incisures may preclude the detection be responsible for paranode alterations. Nevertheless, para­node of this surface protein pool. Indeed, human anti-Nfasc155 IgG4 alterations coincided with a significant and selective Nfasc155 predominantly stained paranodes and incisures after fixation depletion in both the neonatal and adult animal models. The and permeabilization, but staining at the node vicinity or adher- mechanisms responsible for this protein loss are unclear. It seems ens junctions was comparatively difficult to discern. Because the unlikely that depletion of Nfasc155 could be a consequence of antibodies did not react against Nfasc186, it seems unlikely that the loss of its consolidation at paranodes. Indeed, no decrease in this staining was due to binding to nodal Nfasc186. Also, IgG4 Nfasc155 level was reported in animals with a truncated Nfasc155 deposits did not colocalize with nodal axolemma markers, but sur- form that lacks binding capacity to CNTN1/CASPR1 (44). One rounded the node. The significance of the surface Nfasc155 pool is hypothesis is that anti-Nfasc155 IgG4 antibodies favor Nfasc155 not yet clear. Our findings indicate that anti-Nfasc155 IgG4 bind- proteolysis. A very recent study reported that Nfasc155 contains a ing at the node vicinity in neonatal animals is associated with a thrombin cleavage site within its third fibronectin type III domain loss of paranode formation and that anti-Nfasc155 IgG4 induces (45). Interestingly, this domain is located within the antigenic site Nfasc155 depletion. This suggests that antibody binding to the sur- recognized by anti-Nfasc155 IgG4. It could thus be possible that face antigens favors its depletion and precludes CNTN1/CASPR1/ anti-Nfasc155 IgG4 favors Nfasc155 proteolytic cleavage either Nfasc155 complex insertion into paranodes (Figure 9). In addition, by affecting this domain structure or by favoring Nfasc155 aggre- our data suggest that anti-Nfasc155 IgG4 antibodies interfere with gation. Another possibility could be that anti-Nfasc155 IgG4 dis- paranode stabilization in adults (Figure 9). Indeed, acute intraneu- mantles the interaction of Nfasc155 with other partners that are ral antibody injection had no effects once paranodes were formed crucial for its stabilization at the membrane. These latter hypoth- jci.org   Volume 129   Number 6   June 2019 2233 RESEARCH ARTICLE The Journal of Clinical Investigation   eses remain to be demonstrated. Anyway, the loss of Nfasc155 (100 μg) were loaded on a 5% SDS-PAGE gel, transferred onto nitro- was not combined with a depletion in CASPR1 despite the loss cellulose membranes, and immunoblotted with mouse antibodies of CASPR1 accumulation at paranodes. In mice with a truncated against E-cadherin (1:2000; 610181, BD Biosciences) or rabbit anti- Nfasc155 form, CASPR1 levels are also unchanged (44). This fur- sera against pan-neurofascin (1:1000; Ab184378, Abcam), CASPR1 ther confirms that CASPR1 clustering is strongly dependent on (1:10,000; gift of Laurence Goutebroze, INSERM UMR-S839, Paris, Nfasc155 (34, 46), but also indicates that expression and/or stabil- France) (49), Nfasc155 (1:1000; 15035, Cell Signaling Technology), ity of CASPR1 is not dependent on its interaction with Nfasc155. or neurofilament-200 (1:10,000; Ab1987, Abcam) overnight at 4°C. During myelination, CASPR1 first appears in a diffuse manner Immunoreactivity was revealed with the appropriate peroxidase- along unmyelinated axons, then accumulates at the border of the coupled secondary antibodies (1:5000; Jackson ImmunoResearch myelinating Schwann cell in a spiral structure resembling a myelin Laboratories) for 1 hour, washed several times and revealed using a wrap, and is finally consolidated at paranodes (46, 47). Here, the BM chemiluminescence kit (Sigma-Aldrich). reverse phenomenon may occur, and Nfasc155 loss may induce Cell aggregation assay. HEK293 cells were grown in DMEM con- the diffusion of CASPR1 along the axon. taining 10% FCS (Thermo Fisher Scientific) and were transiently In summary, this study demonstrates that Nfasc155 IgG4 anti- transfected using JetPEI (Polyplus-transfection) with mCherry- bodies impair the normal expression of Nfasc155 in the peripheral conjugated rat Nfasc155 or with both rat CNTN1 and GFP-conjugated nerve, and provides strong evidence of their pathogenicity. rat CASPR1. For negative control, cells were solely transfected with peGFP-N1 (Clontech). One day after transfection, cells were washed Methods with PBS, and trypsinated with 0.25% trypsin in PBS for 5 minutes Patients’ antibodies. Plasmapheresis fluids and sera were obtained at 37°C. Cells were centrifuged and suspended in Opti-MEM (Ther- from 3 CIDP patients with high titers of anti-Nfasc155 IgG4 antibodies mo Fisher Scientific). Cells were then mixed together in a 1:1 ratio and 1 CIDP patient with high titer of anti-CNTN1 IgG4 antibodies. All (400,000 cells/ml) in the presence of 10 μg of purified antibodies subjects gave written informed consent. Plasma from 2 healthy donors (final concentration 20 μg/ml), and were agitated at 100 rpm for 2 negative for CNTN1, Nfasc155, or CASPR1 antibodies was used as hours at 37°C. Fifty microliters of cell suspension was then mounted control (Etablissement Français du Sang). IgG4 antibodies were puri- between slides and coverslip, and was immediately observed using an fied using CaptureSelect affinity matrix (Thermo Fisher Scientific) as ApoTome fluorescence microscope at the ×10 objective. Aggregates previously described (29). Fractions were concentrated using a Pierce were defined as clusters of at least 4 cells. Aggregates showing interac- protein concentrator (Thermo Fisher Scientific), then dialyzed to tions between green and red cells were considered as cell clusters with artificial CSF: 126 mM NaCl, 3 mM KCl, 2 mM CaCl2, 2 mM MgSO4, contacts. In addition, the percentage of green cells was quantified in 1.25 mM NaH2PO4, 26 mM NaHCO3, and 10 mM dextrose, pH 7.4–7.5. each cell cluster. Four experiments were performed for each condition, Fractions were finally diluted to a concentration of 10 mg/ml in arti- and a minimum of 40 cell clusters were quantified per experiment. ficial CSF and sterilized by filtration. Fraction purity was verified by Animals and surgery. For in vitro nerve incubation experiments, Western blot and cell-binding assay, but also by ELISA and on rat sci- 6-week-old male Lewis rats (purchased from Janvier Laboratories) atic nerve as previously described (29) (Supplemental Figures 1 and 2). were euthanized, and the sciatic nerves were quickly dissected out, Immunoadsorption. IgG Fc-conjugated human Nfasc155 was puri- desheathed, cut in 1-cm segments, and incubated for 1 hour or 3 hours fied as previously described (48). Five hundred microliters of packed with 10 μg of purified anti-Nfasc155 IgG4 diluted in artificial CSF agarose protein A beads (Sigma-Aldrich) were incubated overnight (final concentration 50 μg/ml). Nerve segments were washed 5 min- at 4°C with a saturating quantity of Fc-Nfasc155 (1 g), and then the utes with artificial CSF, pH 7.4 or pH 2.0, then washed 5 minutes twice beads were extensively washed with PBS. One hundred micrograms with normal artificial CSF, fixed in 2% paraformaldehyde in PBS for of purified anti-Nfasc155 IgG4 was then incubated overnight at 4°C 1 hour at 4°C, and processed for immunolabeling. with 100 μl of Nfasc155 beads for each patient. Beads were sediment- For injections in neonatal animals, newborn pups received a sin- ed by centrifugation, the supernatants were collected, and protein gle i.p. injection of 250 μg of control IgG4 or anti-Nfasc155 IgG4 in concentrations were quantified. This procedure resulted in the deple- artificial CSF (25 μl). Animals were then sacrificed at 2, 4, or 6 days tion of 80%, 62%, and 42% of the proteins in the CIDP1, CIDP2, and after the injection. In some experiments, newborn pups received daily CIDP3 samples, respectively. These samples were used for Western i.p. injection of 250 μg of control IgG4 or anti-Nfasc155 IgG4 in artifi- blot, ELISA, immunohistology on sciatic nerve and HEK293 cells, cial CSF (25 μl) from P0 to P6, and were sacrificed at P4 or P6. and nerve incubation assays. For intraneural injections, anesthesia was induced and main- Western blot. Purified antibody samples were denatured in SDS tained with Vetflurane (Virbac). Animals also received a subcuta- sample buffer for 2 minutes at 92°C, loaded on 7.5% SDS-PAGE gels, neous injection of buprenorphine for pain relief. The right sciatic transferred, and immunoblotted with mouse monoclonal antibodies nerve was exposed at the level of the sciatic notch and injected with against human IgG1 (1:1000; 4E3, Abcam), IgG2 (1:1000; HP6014, 2 μl of antibody (10 μg/μl) using a glass micropipette. One or three Abcam), IgG3 (1:1000; HP-6050, Sigma-Aldrich), or IgG4 (1:5000; days after surgery, injected nerves were processed for immunolabel- HP6035, Abcam). Brain and cerebellum protein samples were pre- ing as detailed below. pared as previously described (4). Sciatic nerves from neonatal or For intrathecal infusion, male Lewis rats (10 weeks old, 300 g) adult rats were directly solubilized in SDS sample buffer and com- were anesthetized with Vetflurane and received a subcutaneous plete protease inhibitor cocktail (Roche), then heated for 2 minutes at injection of buprenorphine for pain relief. Intrathecal polyurethane 90°C. The samples were centrifuged for 10 minutes at 750 g. Protein catheters (Instech, BTPU-027; 0.43 inner diameter, 0.69 mm outer concentration was determined using a BCA kit (Bio-Rad). Samples diameter, 8.0 cm long) were connected to a vascular access button 2234 jci.org   Volume 129   Number 6   June 2019 The Journal of Clinical Investigation   RESEARCH ARTICLE (Instech, VAB62BS/25). The catheters were inserted through a punc- differentiation into oligodendrocytes, cells were maintained in 50% ture of the atlanto-occipital membrane toward the lumbar region as neurobasal medium (Thermo Fisher Scientific), 50% DMEM, 4 mM previously described (50), and the vascular button was fixed under the l-glutamine, 1 mM sodium pyruvate, penicillin/streptomycin, 2× B27 skin on the head of the animal. After recovery for 1 week, the animals supplement, 1× N2 supplement, 100 μg/ml BSA, 40 ng/ml triiodothy- received daily intrathecal injections of 10 μl of purified control IgG4 ronine, 5 μg/ml N-acetylcysteine, and 10 ng/ml biotin. After 10 days in or anti-Nfasc155 IgG4 (10 mg/ml). Injections were performed daily vitro, oligodendrocytes were incubated with CIDP sera (1:200) in neu- for 5 consecutive days followed by a 2-day break for 3 weeks. Animals robasal medium for 1 hour at 37°C, washed 2 minutes with neurobasal were weighed and examined daily for clinical signs. Clinical signs pH 7.4 or pH 2.0, then washed twice 5 minutes with normal neurobasal, were graded as follows: 0, no illness; 1, tail tip hanging; 2, limp tail; 3, and fixed with 2% paraformaldehyde for 20 minutes. Cells were per- tail paralysis; 4, gait ataxia; 5, mild paraparesis; 6, severe paraparesis; meabilized with blocking solution for 20 minutes, incubated for 1 hour 7, paraplegia; 8, tetraparesis; 9, moribund; 10, death. To monitor gait at room temperature with rat antibodies against myelin basic protein abnormalities, animals were allowed to walk freely along a 2-meter- (1:100; MCA409S, Bio-Rad), and revealed with secondary antibodies. long surface (Supplemental Video 1). For footprint analysis, the animal Electrophysiology. Recordings were performed at 21 days as previ- hind paws were dipped in ink and the animals walked down a corri- ously described (48). Briefly, the L6 spinal nerves and sciatic nerves dor on white paper. The angles of the footprints were measured with were dissected out and transferred into artificial CSF. CAP recordings ImageJ version 1.43u software (NIH). The hind-limb grip strength was were made at 35°C in a 3-compartment recording chamber. Conduc- monitored before antibody injection and at 18 days using a BIO-GS3 tion velocities were estimated from latencies and were calculated at grip test (Bioseb). For histopathology and electrophysiology, animals maximal and at half maximal CAP amplitude. were sacrificed after 21 days of injections. Statistics. Statistical significance was assessed by unpaired Immunolabeling and histopathology. Sciatic nerves and L6 ventral 2-tailed Student’s t tests, Kolmogorov-Smirnov tests, or 1-way ANOVA and dorsal spinal nerves were dissected out and fixed in 2% parafor- followed by Bonferroni’s post hoc test using GraphPad Prism (Graph- maldehyde in PBS for 1 hour at 4°C, then rinsed in PBS. Axons were Pad Software). P values less than 0.05 were considered significant. gently teased, dried on glass slides, and stored at –20°C. Teased fibers Study approval. All animal experiments were in line with the Euro- were permeabilized by immersion in –20°C acetone for 10 minutes, pean Community’s guiding principles on the care and use of animals blocked at room temperature for 1 hour with PBS containing 5% fish (2010/63/EU) and were approved by the local ethical committee skin gelatin and 0.1% Triton X-100, then incubated overnight at 4°C (Comité d’Ethique pour l’Expérimentation Animale Languedoc Rous- with primary antibodies: rabbit antisera against Nfasc186 (1:500; gift sillon no. 36, Biocampus, Montpellier, France) and by the “ministére of Alex Gow, Wayne State University, Detroit, Michigan, USA) (51), de l’éducation nationale de l’enseignement supérieur et de la recher- ankyrin-G (1:500; gift of Gisèle Alcaraz, GMGF, INSERM, UMR-S910, che” (APAFIS 3847-2016012610089856v8). Marseille, France) (52), or CASPR1 (1:2000); mouse monoclonal anti- bodies against Nav channels (1:500; K58/35, Sigma-Aldrich) or β-cat- Author contributions enin (MAB2081, Merck Millipore); and goat antibody against CNTN1 CM, LQ, CL, MP, MM, JMV, II, and JJD drafted and revised the (1:2000; R&D Systems). The slides were then washed several times manuscript. JJD, LQ, and II conceived and designed the study. and incubated with Alexa-conjugated donkey antisera against rabbit, CM, LQ, CL, MP, MM, JMV, II, and JJD acquired data. CM, LQ, mouse, human, or goat IgG (1:500; Jackson ImmunoResearch Labo- CL, MP, MM, JMV, II, and JJD analyzed and interpreted data. ratories) or FITC-conjugated mouse monoclonal antibodies against human IgG1 (1:500; clone 4E3, ab99772, Abcam), IgG2 (1:500; clone Acknowledgments HP6014, ab99788, Abcam), IgG3 (1:500; clone HP6050, F4641, Sigma- We thank Laurence Goutebroze, Alex Gow, and Giséle Alcaraz Aldrich), or IgG4 (1:500; clone HP6025, ab99821, Abcam). Slides for gifts of antibodies, and Axel Fernandez for technical assis- were mounted with Mowiol plus 2% DABCO, and examined using an tance. JJD, LQ, and II acknowledge grant support from the Agence ApoTome fluorescence microscope (Carl Zeiss MicroImaging GmbH). Nationale pour la Recherche; the Centro para la Investigación Digital images were manipulated into figures with CorelDraw and Biomédica en Red (CIBERER; ACAMIN project) under the frame Corel Photo-Paint. Teased fibers were analyzed from 11 animals for of E-Rare-2; the ERA-Net for Research on Rare Diseases; the each group. For quantification of paranodal alterations, the percent- Association Française contre les Myopathies (grant 21532 to JJD); age of nodes showing loss of paranodal regions was quantified in 11 the GBS-CIDP Foundation International (to CL); CSL Behring’s animals (~100 axons counted in total). grant in immunology (to JJD); the Pla Estratègic de Recerca i Inno- For histopathological analysis, nerves were fixed in 2.5% glu- vació en Salut (PERIS) 2016–2020 of the Departament de Salut taraldehyde in 0.1 M sodium cacodylate buffer plus 0.16 M sucrose of the Generalitat de Catalunya (SLT006/17/00131 to LQ); and overnight at 4°C and postfixed in 1% OsO4 in 0.1 M sodium cacodylate the Fondo de Investigaciones Sanitarias, Instituto de Salud Carlos buffer for 1 hour. Nerves were dehydrated and embedded in epoxy res- III, Ministry of Health, Spain (PI16/000627, FEDER funds, to LQ; in. Transverse semithin sections were stained with toluidine blue and and FIS 13/00937 to II). examined by light microscopy. Longitudinal ultrathin sections were examined by electron microscopy as previously described (25). Address correspondence to: Jérôme J. Devaux, Institute for Neuro- Oligodendrocyte culture. Oligodendrocyte precursor cells were pre- sciences of Montpellier, INSERM U1051, Montpellier University, pared from rats as previously described (4). Oligodendrocyte precursor Hopital Gui de Chauliac, 80 avenue Augustin Fliche, 34295 Mont- cells were plated into 12-well plates containing poly-l-lysine–coated pellier Cedex 5, France. Phone: 33.4.99.63.60.95; Email: jerome. 12-mm coverslips at a density of 250,000 cells per well. To allow devaux@inserm.fr. jci.org   Volume 129   Number 6   June 2019 2235 RESEARCH ARTICLE The Journal of Clinical Investigation   1. Dalakas MC, Medscape. Advances in the diagno- 19. Vallat JM, et al. Subacute nodopathy with con- 35. Franques J, Chapon F, Devaux J, Mathis S. Teach- sis, pathogenesis and treatment of CIDP. Nat Rev duction blocks and anti-neurofascin 140/186 ing NeuroImages: cranial nerve hypertrophy Neurol. 2011;7(9):507–517. antibodies: an ultrastructural study. Brain. in IgG4 anti-neurofascin 155 antibody-positive 2. Querol L, et al. Antibodies against peripheral 2018;141(7):e56. polyneuropathy. Neurology. 2017;88(7):e52. nerve antigens in chronic inflammatory demy- 20. Bhat MA, et al. Axon-glia interactions and the 36. Ogata H, et al. Characterization of IgG4 anti- elinating polyradiculoneuropathy. Sci Rep. domain organization of myelinated axons neurofascin 155 antibody-positive polyneuropa- 2017;7(1):14411. requires neurexin IV/Caspr/Paranodin. Neuron. thy. Ann Clin Transl Neurol. 2015;2(10):960–971. 3. Delmont E, et al. Autoantibodies to nodal 2001;30(2):369–383. 37. Kadoya M, et al. IgG4 anti-neurofascin155 anti- isoforms of neurofascin in chronic inflamma- 21. Boyle ME, Berglund EO, Murai KK, Weber L, bodies in chronic inflammatory demyelinating tory demyelinating polyneuropathy. Brain. Peles E, Ranscht B. Contactin orchestrates polyradiculoneuropathy: clinical significance and 2017;140(7):1851–1858. assembly of the septate-like junctions at the diagnostic utility of a conventional assay. J Neuro- 4. Devaux JJ, et al. Neurofascin-155 IgG4 in chronic paranode in myelinated peripheral nerve. Neu- immunol. 2016;301:16–22. inflammatory demyelinating polyneuropathy. ron. 2001;30(2):385–397. 38. Martinez-Martinez L, et al. Anti-NF155 chronic Neurology. 2016;86(9):800–807. 22. Sherman DL, et al. Neurofascins are required to inflammatory demyelinating polyradiculoneu- 5. Doppler K, et al. Auto-antibodies to contactin- establish axonal domains for saltatory conduc- ropathy strongly associates to HLA-DRB15. associated protein 1 (Caspr) in two patients tion. Neuron. 2005;48(5):737–742. J Neuroinflammation. 2017;14(1):224. with painful inflammatory neuropathy. Brain. 23. Koike H, et al. Paranodal dissection in chronic 39. Yan W, et al. Antibodies to neurofascin exacer- 2016;139(pt 10):2617–2630. inflammatory demyelinating polyneuropathy bate adoptive transfer experimental autoimmune 6. Doppler K, et al. Destruction of paranodal with anti-neurofascin-155 and anti-contactin-1 neuritis. J Neuroimmunol. 2014;277(1–2):13–17. architecture in inflammatory neuropathy with antibodies. J Neurol Neurosurg Psychiatry. 40. Koike H, et al. Restoration of a conduction anti-contactin-1 autoantibodies. J Neurol Neuro- 2017;88(6):465–473. block after the long-term treatment of CIDP surg Psychiatry. 2015;86(7):720–728. 24. Kuwahara M, et al. Electron microscopic abnor- with anti-neurofascin 155 antibodies: fol- 7. Miura Y, et al. Contactin 1 IgG4 associates to mality and therapeutic efficacy in chronic low-up of a case over 23 years. Intern Med. chronic inflammatory demyelinating polyneu- inflammatory demyelinating polyneuropathy 2018;57(14):2061–2066. ropathy with sensory ataxia. Brain. with anti-neurofascin155 immunoglobulin G4 41. Uncini A, Vallat JM. Autoimmune nodo-para­ 2015;138(pt 6):1484–1491. antibody. Muscle Nerve. 2018;57(3):498–502. nodopathies of peripheral nerve: the concept is 8. Ng JK, et al. Neurofascin as a target for autoan- 25. Vallat JM, et al. Paranodal lesions in chronic gaining ground. J Neurol Neurosurg Psychiatry. tibodies in peripheral neuropathies. Neurology. inflammatory demyelinating polyneuropathy 2018;89(6):627–635. 2012;79(23):2241–2248. associated with anti-Neurofascin 155 antibodies. 42. Tait S, et al. An oligodendrocyte cell adhe- 9. Querol L, et al. Neurofascin IgG4 antibodies in Neuromuscul Disord. 2017;27(3):290–293. sion molecule at the site of assembly of the CIDP associate with disabling tremor and poor 26. Huijbers MG, et al. The expanding field of paranodal axo-glial junction. J Cell Biol. response to IVIg. Neurology. 2014;82(10):879–886. IgG4-mediated neurological autoimmune disor- 2000;150(3):657–666. 10. Querol L, et al. Antibodies to contactin-1 in ders. Eur J Neurol. 2015;22(8):1151–1161. 43. Desmazieres A, Zonta B, Zhang A, Wu LM, Sher- chronic inflammatory demyelinating polyneu- 27. Querol L, Devaux J, Rojas-Garcia R, Illa I. Auto- man DL, Brophy PJ. Differential stability of PNS ropathy. Ann Neurol. 2013;73(3):370–380. antibodies in chronic inflammatory neuropa- and CNS nodal complexes when neuronal neuro- 11. Painous C, López-Pérez MÁ, Illa I, Querol L. thies: diagnostic and therapeutic implications. fascin is lost. J Neurosci. 2014;34(15):5083–5088. Head and voice tremor improving with immuno- Nat Rev Neurol. 2017;13(9):533–547. 44. Thaxton C, et al. In vivo deletion of immuno- therapy in an anti-NF155 positive CIDP patient. 28. Labasque M, Hivert B, Nogales-Gadea G, Querol globulin domains 5 and 6 in neurofascin (Nfasc) Ann Clin Transl Neurol. 2018;5(4):499–501. L, Illa I, Faivre-Sarrailh C. Specific contactin reveals domain-specific requirements in myelin- 12. Querol L, et al. Rituximab in treatment-resis- N-glycans are implicated in neurofascin binding ated axons. J Neurosci. 2010;30(14):4868–4876. tant CIDP with antibodies against paranodal and autoimmune targeting in peripheral neurop- 45. Dutta DJ, et al. Regulation of myelin structure and proteins. Neurol Neuroimmunol Neuroinflamm. athies. J Biol Chem. 2014;289(11):7907–7918. conduction velocity by perinodal astrocytes. Proc 2015;2(5):e149. 29. Manso C, Querol L, Mekaouche M, Illa I, Devaux Natl Acad Sci U S A. 2018;115(46):11832–11837. 13. Faivre-Sarrailh C, Devaux JJ. Neuro-glial inter- JJ. Contactin-1 IgG4 antibodies cause paranode 46. Eisenbach M, et al. Differential clustering of actions at the nodes of Ranvier: implication dismantling and conduction defects. Brain. Caspr by oligodendrocytes and Schwann cells. in health and diseases. Front Cell Neurosci. 2016;139(pt 6):1700–1712. J Neurosci Res. 2009;87(15):3492–3501. 2013;7:196. 30. Eshed Y, et al. Gliomedin mediates Schwann 47. Pedraza L, Huang JK, Colman DR. Organizing 14. Freeman SA, Desmazières A, Fricker D, Lubetzki cell-axon interaction and the molecular principles of the axoglial apparatus. Neuron. C, Sol-Foulon N. Mechanisms of sodium channel assembly of the nodes of Ranvier. Neuron. 2001;30(2):335–344. clustering and its influence on axonal impulse 2005;47(2):215–229. 48. Lonigro A, Devaux JJ. Disruption of neurofascin conduction. Cell Mol Life Sci. 2016;73(4):723–735. 31. Schafer DP, Custer AW, Shrager P, Rasband MN. and gliomedin at nodes of Ranvier precedes 15. Feinberg K, et al. A glial signal consisting of glio- Early events in node of Ranvier formation during demyelination in experimental allergic neuritis. medin and NrCAM clusters axonal Na+ channels myelination and remyelination in the PNS. Neu- Brain. 2009;132(pt 1):260–273. during the formation of nodes of Ranvier. Neu- ron Glia Biol. 2006;2(2):69–79. 49. Menegoz M, et al. Paranodin, a glycoprotein ron. 2010;65(4):490–502. 32. Smith CE, Atchabahian A, Mackinnon SE, Hunter of neuronal paranodal membranes. Neuron. 16. Amor V, et al. Long-term maintenance of Na+ DA. Development of the blood-nerve barrier in 1997;19(2):319–331. channels at nodes of Ranvier depends on glial neonatal rats. Microsurgery. 2001;21(7):290–297. 50. Yaksh TL, Rudy TA. Chronic catheterization of contact mediated by gliomedin and NrCAM. 33. Pomicter AD, et al. Novel forms of neurofascin the spinal subarachnoid space. Physiol Behav. J Neurosci. 2014;34(15):5089–5098. 155 in the central nervous system: alterations in 1976;17(6):1031–1036. 17. Taylor AM, Saifetiarova J, Bhat MA. Postnatal paranodal disruption models and multiple sclero- 51. Southwood C, He C, Garbern J, Kamholz J, Arroyo loss of neuronal and glial neurofascins differen- sis. Brain. 2010;133(pt 2):389–405. E, Gow A. CNS myelin paranodes require Nkx6-2 tially affects node of Ranvier maintenance and 34. Pillai AM, Thaxton C, Pribisko AL, Cheng JG, homeoprotein transcriptional activity for normal myelinated axon function. Front Cell Neurosci. Dupree JL, Bhat MA. Spatiotemporal ablation structure. J Neurosci. 2004;24(50):11215–11225. 2017;11:11. of myelinating glia-specific neurofascin (Nfasc 52. Bouzidi M, et al. Interaction of the Nav1.2a sub- 18. Thaxton C, Pillai AM, Pribisko AL, Dupree JL, NF155) in mice reveals gradual loss of paranodal unit of the voltage-dependent sodium channel Bhat MA. Nodes of Ranvier act as barriers to axoglial junctions and concomitant disorga- with nodal ankyrinG. In vitro mapping of the restrict invasion of flanking paranodal domains in nization of axonal domains. J Neurosci Res. interacting domains and association in synapto- myelinated axons. Neuron. 2011;69(2):244–257. 2009;87(8):1773–1793. somes. J Biol Chem. 2002;277(32):28996–29004. 2236 jci.org   Volume 129   Number 6   June 2019