Intrathecal cytokine profile in neuropathy with anti‐neurofascin 155 antibody

Abstract Objective To characterize the CSF cytokine profile in chronic inflammatory demyelinating polyneuropathy (CIDP) patients with IgG4 anti‐neurofascin 155 (NF155) antibodies (NF155+ CIDP) or those lacking anti‐NF155 antibodies (NF155− CIDP). Methods Twenty‐eight CSF cytokines/chemokines/growth factors were measured by multiplexed fluorescent immunoassay in 35 patients with NF155+ CIDP, 36 with NF155− CIDP, and 28 with non‐inflammatory neurological disease (NIND). Results CSF CXCL8/IL‐8, IL‐13, TNF‐α, CCL11/eotaxin, CCL2/MCP‐1, and IFN‐γ were significantly higher, while IL‐1β, IL‐1ra, and G‐CSF were lower, in NF155+ CIDP than in NIND. Compared with NF155− CIDP, CXCL8/IL‐8 and IL‐13 were significantly higher, and IL‐1β, IL‐1ra, and IL‐6 were lower, in NF155+ CIDP. CXCL8/IL‐8, IL‐13, CCL11/eotaxin, CXCL10/IP‐10, CCL3/MIP‐1α, CCL4/MIP‐1β, and TNF‐α levels were positively correlated with markedly elevated CSF protein, while IL‐13, CCL11/eotaxin, and IL‐17 levels were positively correlated with increased CSF cell counts. IL‐13, CXCL8/IL‐8, CCL4/MIP‐1β, CCL3/MIP‐1α, and CCL5/RANTES were decreased by combined immunotherapies in nine NF155+ CIDP patients examined longitudinally. By contrast, NF155− CIDP had significantly increased IFN‐γ compared with NIND, and exhibited positive correlations of IFN‐γ, CXCL10/IP‐10, and CXCL8/IL‐8 with CSF protein. Canonical discriminant analysis of cytokines/chemokines revealed that NF155+ and NF155− CIDP were separable, and that IL‐4, IL‐10, and IL‐13 were the three most significant discriminators. Interpretation Intrathecal upregulation of type 2 helper T (Th2) cell cytokines is characteristic of IgG4 NF155+ CIDP, while type 1 helper T cell cytokines are increased in CIDP regardless of the presence or absence of anti‐NF155 antibodies, suggesting that overproduction of Th2 cell cytokines is unique to NF155+ CIDP.

Anti-NF155 antibodies found in a fraction of CIDP patients mainly belong to the immunoglobulin G (IgG)4 subclass. 3 IgG4 anti-NF155 antibody-positive CIDP (NF155 + CIDP) demonstrates distinctive features, including younger age at onset, higher frequencies of drop foot, sensory ataxia, and tremor, marked prolongation of distal and F wave latency, extremely high cerebrospinal fluid (CSF) protein amounts, and marked hypertrophy of nerve roots on magnetic resonance neurography. 2,3,5 However, biopsied sural nerves from IgG4 NF155 + CIDP patients lack inflammation and onion bulb formation, with only subperineurial edema and minimal paranodal demyelination. 3 By electron microscopy, detachment of terminal myelin loops is characteristic for NF155 + CIDP, but not for anti-NF155 antibody-negative CIDP (NF155 À CIDP). 9,10 IgG4 cannot activate complement because it does not bind C1q. 11 In vivo, IgG4 is monovalent and bispecific because of half molecular exchange, and does not internalize target antigens. 11 Therefore, IgG4 autoantibodies only block protein-protein interactions, 11 which explains why sural nerve pathology only presents paranodal terminal loop detachment in the absence of inflammation. 9,10 Restoration of nerve conduction by plasma exchange with decreased anti-NF155 antibody titers is compatible with the blocking antibody action of IgG4. 12 By contrast, extensive proximal nerve hypertrophy and pronounced CSF protein elevation, suggesting severe inflammation and/or edema of nerve roots, is unique to this condition, 3 but is difficult to explain solely by IgG4 antibody functions. Recently, a strong association of certain human leukocyte antigen (HLA) class II alleles with NF155 + CIDP was reported in a European series, 13 suggesting HLA class II-restricted T-cell involvement. These observations prompted us to clarify the CSF cytokine profile in patients with IgG4 NF155 + CIDP to elucidate the mechanism.

Subjects and Methods
Subjects Thirty-five consecutive IgG4 NF155 + CIDP and 36 NF155 À CIDP patients were enrolled in the present study. None of the NF155 + or NF155 À CIDP patients had anti-NF186 or anti-CNTN1 antibodies in sera. Among these patients, 44 were thoroughly examined in the Department of Neurology at Kyushu University Hospital between 01  January 2001 and 31 May 2018, while the other patients  were referred to our department for an anti-NF155 antibody assay between 1 November 2014 and 31 March 2018. All CIDP patients fulfilled the definite electrodiagnostic criteria of the European Federation of Neurological Societies/Peripheral Nerve Society for the diagnosis of CIDP, 14 except for one NF155 + CIDP patient who showed no evoked potentials on nerve conduction studies. Clinical features of 11 NF155 + CIDP patients and biopsied sural nerve pathologies of 3 NF155 + CIDP patients showing subperineurial edema without inflammatory cell infiltrates were previously reported elsewhere. 3, 10 Hughes functional grading 15 was used to evaluate clinical severity. Twenty-two IgG4 NF155 + CIDP and 23 NF155 À CIDP patients had received no treatment at the time of lumbar puncture (LP). Furthermore, two or more CSF samples at different time points were available in nine NF155 + CIDP patients. For controls, 28 other non-inflammatory neurological disease (NIND) patients were enrolled, including 12 with amyotrophic lateral sclerosis, five with spondylosis, four with normal pressure hydrocephalus, three with spinocerebellar ataxia, and one each with metabolic neuropathy, hyperthyroidism, psychosomatic disorder, or cerebral venous malformations. The research protocols for the study were approved by the Kyushu University Ethics Committee. An opt-out recruitment method was adopted.

Sample collection
CSF samples were obtained by non-traumatic LP. None of the subjects for CSF analysis had any ongoing recent infection at the time of LP. The periods (disease durations) from onset of symptoms to CSF sample collection were comparable between NF155 + and NF155 À CIDP patients (Table 1). CSF samples were immediately centrifuged at 800 rpm (100g) at 4°C for 5 min, and the supernatants were stored at À80°C until analysis.

Statistical analysis
Cytokines/chemokines/growth factors with detection rates <30% were excluded from analysis as previously reported. [16][17][18][19] The chi-square test was performed to evaluate the statistical significance of detection rates of cytokines/chemokines/growth factors among NF155 + CIDP, NF155 À CIDP, and NIND patients. If the P < 0.05, Fisher's exact probability test was used for comparisons between any two groups, followed by Bonferroni-Dunn's correction. The Kruskal-Wallis test was performed to compare the levels of cytokines/chemokines/growth factors between three groups. If the P < 0.05, the Mann-Whitney U-test was used for comparisons between any two groups, followed by Bonferroni-Dunn's correction. The Wilcoxon signed-rank test was performed to evaluate the longitudinal analysis according to treatment status. Spearman's rank correlation coefficient was used for statistical analyses between cytokines/chemokines/growth factors and CSF protein amounts or cell counts. All of the above statistical analyses were performed with GraphPad Prism ver. 5.0 software (GraphPad, San Diego, CA). To separately analyze the correlations of CSF cytokines/chemokines in patients with NF155 + CIDP and NF155 À CIDP, heatmaps were generated by the "gplots" package in R on the basis of Spearman's rank correlation coefficient and cluster analysis. A canonical discriminant analysis (CDA) was performed by JMP Pro software (ver. 13.0.0; SAS Institute, Cary, NC).

Effects of immunotherapies on CSF cytokine/chemokine/growth factor levels
Longitudinal analyses of CSF cytokine/chemokine/growth levels in 9 patients with CSF samples repeatedly obtained before and after immunotherapy revealed that CXCL8/IL-8, IL-13, CCL4/MIP-1b, CCL3/MIP-1a, and CCL5/ RANTES were decreased after treatment (P = 0.0078, P = 0.0343, P = 0.0117, P = 0.0438, and P = 0.0234, respectively) (Fig. 3). A similar tendency was observed for CCL2/MCP-1 (P = 0.0547). Five patients (Patients 3-7 in Fig. 3) treated with combined immunotherapies such as intravenous immunoglobulin (IVIg), corticosteroids, plasma exchanges, and immunosuppressants, showed improvement in Hughes functional grade scores together with deceases in most of these cytokines. By contrast, a patient treated with IVIg alone (Patient 2 in Fig. 3) showed increases in some of these cytokines and no clinical improvement. Subsequently, CSF cytokine/chemokine/ growth factor levels were compared between 22 untreated and 13 treated samples from 35 NF155 + CIDP patients and between 23 untreated and 13 treated samples from 36 NF155 À CIDP patients, in which one treated or untreated sample was evaluated per patient. CXCL8/IL-8 and CCL2/MCP-1 levels were lower and IL-1ra levels were higher in treated samples than in untreated samples from NF155 + CIDP patients (P = 0.0043, P = 0.0077, and P = 0.0241, respectively) (Fig. S3A). In NF155 À CIDP, IFN-c levels were significantly lower in treated samples than in untreated samples (P = 0.0442) (Fig. S3B).

Subgroup analysis of NF155 À CIDP
Finally, we classified NF155 À CIDP patients into two subgroups by IL-1b levels: those with median (1.06 mg/dl) or higher IL-1b levels and those with lower than median IL-1b levels. As a result, we found that the frequencies of typical CIDP and CSF protein levels were significantly higher in the low IL-1b group than in the high IL-1b group (P = 0.0063 and P < 0.0001, respectively) (Table S2). Sex ratio, CSF cell counts, and age at onset were comparable between the two subgroups.

Discussion
In this study, the characteristic features of IgG4 NF155 + CIDP were marked elevations of Th2-and Th1-related cytokines. By contrast, NF155 À CIDP showed a significant increase in IFN-c only and positive correlations of IFN-c     and CXCL10/IP-10 (chemokine downstream of IFN-c) with CSF protein amounts, suggesting a Th1 shift. Cluster analyses further supported distinct cytokine profiles between NF155 + and NF155 À CIDP. Upregulation of cluster 3 cytokines, including IL-13, CCL11/eotaxin, CCL2/MCP-1, CCL3/MIP-1a, CCL4/MIP-1b, CXCL8/IL-8, and CXCL10/IP-10, was characteristic of NF155 + CIDP. Thus, we consider that co-upregulation of Th2 and Th1 cytokines is a unique feature of NF155 + CIDP. An intrathecal Th1 shift has been proposed in CIDP, based on elevated levels of Th1-related cytokines including IL-12 and CXCL10/IP10 together with increased IFN-c + IL-4 À CD4 + T (Th1) cell percentages in CSF cells, 20 although no reports have described CIDP subtype-specific CSF cytokine profiles. Because NF155 + CIDP comprises only a minority of total CIDP, the cytokine profile of total CIDP patients may have not reflected Th2 cytokine upregulation in NF155 + CIDP patients in previous studies. [20][21][22][23][24] IgG4 usually arises after chronic exposure to antigenic stimuli, and its physiological role is to compete with IgE and block its pathogenic effects in allergic responses. 25 The class switch to IgG4 is facilitated in the presence of Th2 cytokines including IL-13, IL-4, and IL-10. 26 Intriguingly, CDA revealed that IL-13, IL-4, and IL-10 levels were the major discriminants between NF155 + and NF155 À CIDP, suggesting a key role for these Th2 cytokines in NF155 + CIDP. Furthermore, the positive correlations of Th2 cytokines (IL-13 and CCL11/eotaxin) with CSF cell counts and protein amounts and reduction of IL-13 levels during longitudinal follow-up after immunotherapy in NF155 + CIDP suggested the involvement of Th2 cells in spinal root inflammation, in addition to the induction of IgG4 autoantibodies. In Th2 inflammation, IL-13 is a major effector cytokine, while CCL11/eotaxin exhibits potent chemotactic activity. 27,28 The decreases in CXCL8/IL-8, CCL2/MCP-1, CCL3/ MIP-1a, and CCL4/MIP-1b, as downstream chemokines stimulated by IL-13, 29,30 after immunotherapy in parallel with some clinical improvement reflected in Hughes functional grade scores suggest a direct involvement of these chemokines in NF155 + CIDP. The tendency toward positive correlations of CCL3/MIP-1a and CXCL8/IL-8 with Hughes functional grade scores may support an involvement of these chemokines in nerve damage, although further prospective evaluation of neurological impairment is required to confirm the correlations between CSF cytokine levels and disability. CXCL8/IL-8 is secreted by a variety of cells including blood monocytes, fibroblasts, endothelial cells, and epithelial cells upon stimulation by various cytokines, leading to 10-100-fold upregulation of CXCL8/IL-8 expression. 31 As the changes in levels of downstream cytokines, such as CXCL8/IL-8, are much more amplified than those of upstream cytokines, the reduction in CXCL8/IL-8 levels by immunotherapy could be more prominent than those in IL-13 and CCL11/eotaxin levels, suggesting that these downstream chemokines may be potential biomarkers that can reflect disease activity. In addition, IFN-c was significantly increased in NF155 + CIDP patients compared with NIND patients, also suggesting the potential involvement of Th1 cells. Because we found a small but significant increase in CSF cells in pretreated NF155 + CIDP patients, a more detailed analysis of CSF cytology, including helper T cells and eosinophils, is required to elucidate the effector mechanism of root inflammation.
IL-1b and IL-1ra were significantly decreased in NF155 + CIDP patients compared with NF155 À CIDP and NIND patients. Th2 cytokines such as IL-13 were reported to decrease IL-1b production, 32 suggesting the suppression of these cytokines by activated Th2 cells in NF155 + CIDP. Macrophage-mediated demyelination, which is often seen in biopsied sural nerves of NF155 À CIDP patients, 9,33 was not observed in those of NF155 + CIDP patients. 9,33 Because IL-1b activates macrophages, 34 the pronounced suppression of IL-1b in NF155 + CIDP may be related to the absence of macrophage-mediated demyelination. Even in NF155 À CIDP patients, the low IL-b subgroup showed higher frequencies of typical CIDP (nearly 90%) and higher CSF protein levels than the high IL-b subgroup. Thus, in the low IL-b subgroup preferentially presenting typical CIDP, a mechanism other than macrophage-mediated demyelination could be operative. By contrast, a macrophage-mediated mechanism may be operative in the high IL-b subgroup. Although a decrease in IL-1ra was common in both NF155 + and NF155 À CIDP, the decrease was most marked in NF155 + CIDP. IL-1ra is a representative anti-inflammatory cytokine, 34 and thus the pronounced decrease in IL-1ra levels in NF155 + CIDP patients may also contribute to severe spinal root inflammation.
Biopsied sural nerves showed marked subperineurial edema, suggesting disruption of the blood-nerve barrier (BNB), although no onion bulb formation was observed. 3,10 Therefore, distal nerve edema may be a manifestation of widespread BNB breakdown including the nerve roots. However, it is difficult to explain the mechanism for how IgG4 anti-NF155 antibodies targeting the paranodal structure without activating complement can induce disruption of the BNB. Because proinflammatory cytokines/chemokines, such as IL-13, CCL-11/eotaxin, and CXCL8/IL-8, were increased in NF155 + CIDP patients including three biopsied cases and showed significant positive correlations with CSF protein levels in our study, these cytokines/chemokines and the immunocytes producing them may be directly or indirectly involved in BNB disruption. 35,36 financial support from the China Scholarship Council. RY received a grant from JSPS KAKENHI (Grant No. 16K09694). MK received honoraria from Teijin, Nihon Pharmaceutical and Japan Blood Products Organization. SK received grants from Ministry of Health, Labour and Welfare of Japan, the Japan Agency for Medical Research and Development (AMED) and Ministry of Education, Culture, Sports, Science and Technology of Japan, and received consultant fees, speaking fees and/or honoraria from Teijin, Japan Blood Product Organization and Nihon Pharmaceutical. YN received a grant and salary from Mitsubishi Tanabe Pharma, Bayer Yakuhin, Ltd., and Japan Blood Products Organization. TM received speaker honoraria payments from Mitsubishi Tanabe Pharma, Takeda Pharmaceutical Company, and Biogen

Supporting Information
Additional supporting information may be found online in the Supporting Information section at the end of the article. Table S1. Detection rates of cytokines/chemokines/growth factors in CSF. Table S2. Comparisons of clinical and laboratory findings between anti-NF155 antibody-negative CIDP patients with high and low IL-1b levels. Figure S1. CSF protein amounts, cell counts, and cytokine/chemokine/growth factor levels in untreated patients. Figure S2. Correlations of cytokine/chemokine levels with clinical severity and CSF protein amounts and cell counts in untreated patients. Figure S3. Comparisons of CSF cytokine/chemokine levels between untreated and treated patients.