Neurochondrin is a neuronal target antigen in autoimmune cerebellar degeneration

Characterization of the patients.

Three patients presented with a pronounced cerebellar and brainstem syndrome of unknown origin (SARA score: patient 1 27/40, patient 2 25/40, patient 32/40; video e-1–e-3; a detailed case description can be found in appendix e-1). Patients 1 and 2 had short disease durations of 21 and 54 months, respectively, whereas patient 3 had clinical symptoms for about 120 months. Symptoms were still progressive in patients 1 and 2 but essentially stable for a couple of years in patient 3. Patient 2 also reported one episode of blurred vision, reduced color vision, and pain upon movement of the right eye for several weeks.

Quantitative high-resolution structural MRI at 3T showed pronounced cerebellar atrophy in all patients (figure e-1, A–C) with pronounced supratentorial symmetric atrophy of cortical more than subcortical gray matter and largely preserved white matter volumes in all patients compared to a cohort of 38 age-matched healthy controls (figures e-1, D–F, and e-2 “pre”). Moreover, infratentorial cerebellar gray more than white matter atrophy was evident in all patients as compared to a cohort of 38 age-matched healthy controls. Diffusion tensor imaging (DTI; figures e-3 and e-4 “pre”) revealed widespread reduction of fractional anisotropy (FA; figures e-3, A–E, and e-4 “pre”) together with increased axial diffusivity (AD; figures e-3F and e-4 “pre”) and radial diffusivity (RD; figures e-2G and e-4 “pre”) in white matter areas to different degrees in all patients as compared to a cohort of 38 age-matched healthy controls.

Despite these widespread structural changes visible on MRI, neuropsychological assessment revealed normal performance in verbal working memory, verbal short-term and long-term memory, as well as executive function (tables e-1 and e-2). Executive function could not be assessed in patient 3 due to severe cerebellar motor symptoms.

All patients had chronic inflammatory changes (i.e., pleocytosis, ≥3 CSF-specific oligoclonal immunoglobulin G [IgG] bands¹⁹) upon routine CSF analysis and were thus assumed to have autoimmune or paraneoplastic cerebellar degeneration. Routine testing of serum and CSF for known antineuronal antibodies was unremarkable. However, binding to an unknown intracellular submembrane neuronal antigen both in serum and CSF of all 3 patients was detected. Repeated tumor searches using whole-body FDG-PET/CT together with gynecologic or urologic examination did not reveal an underlying tumor in any of the patients for more than 2–10 years.

Patients had shown no sustained response to preceding antibody-depleting (IV immunoglobulin [IVIg], immunoadsorption, plasma exchange) or short-term immunosuppressive treatments (cyclophosphamide, rituximab, mitoxantrone; appendix e-1). Due to persistent clinical progression, consequent long-term immunosuppressive treatment using cyclophosphamide pulse therapy (750–1,000 mg/m² body surface area every 4–6 weeks) for 6–12 months was introduced in patients 1 and 2. In patient 3, immunotherapy was continued using azathioprine (150 mg/day) and recurrent IVIg treatments (0.4 g/kg body weight for 5 consecutive days every 3 months), under which clinical stability was present for a couple of years.

Clinical follow-up after 6–12 months showed stable or improved disease in all patients under the respective treatment regimen (SARA score: patient 1 25/40, patient 2 23/40, patient 3 32/40) with definitively halted progression in patients 1 and 2 and continued stability in patient 3. Follow-up quantitative high-resolution structural MRI (figure e-2 “post”) revealed partial regression of cortical and cerebellar gray matter atrophy in all patients. However, DTI (figure e-4 “post”) showed largely unaltered widespread reduction of FA (figure e-4, first row) together with increased AD (figure e-4, second row) and RD (figure e-4, third row) in white matter in all 3 patients consistent with a partial response to long-term immunosuppressive treatment but persistent microstructural damage.

Characterization of the patients' autoantibodies.

Indirect IFA of patients' sera and CSF using permeabilized cryosections of brain tissue showed a fine-granular IgG1 staining of the granular and molecular layer that was stronger on rat hippocampus compared to rat and primate cerebellum (figure 1). Sagittal sections of murine whole brain showed autoantibody reactivity distributed over all parts of the brain (figure e-5). In addition, primate hypothalamus, cerebrum, medulla spinalis, and hippocampus were stained, similarly. The patients had serum and CSF titers of 1:320/1:32 (patient 1), 1:320/1:1 (patient 2), and 1:1,000/1:32 (patient 3) as determined with rat hippocampus corresponding to specific antibody indices (AI) >4 in patients 1 and 3 indicating intrathecal antibody synthesis. Titers were largely unaltered during immunosuppressive treatment. Further monospecific analyses were conducted with recombinant HEK293 cells expressing 30 neural autoantigens: Hu, Yo, Ri, CV2, SOX1, PNMA1, PNMA2, ITPR1, Homer 3, CARP VIII, ARHGAP26, ZIC4, DNER/Tr, GAD65, GAD67, amphiphysin, recoverin, GABA_B receptor, glycine receptor, DPPX, glutamate receptors (types NMDA, AMPA, mGluR1, mGluR5), LGI1, CASPR2, AQP4 (M1 and M23), MOG, ATP1A3. None revealed any specific reactivity.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1 Immunofluorescence staining of central nervous tissues

Cryosections were incubated with patient or control sera (1:100) or CSF (undiluted) in the first step, and with Alexa488-labelled goat anti-human immunoglobulin G in the second step. Nuclei were counterstained by incubation with TO-PRO-3 iodide (blue). A fine granular staining of both granular and molecular layers was obtained with the strongest reaction on the granular cell layer of rat hippocampus. ×200 magnification.

Identification of neurochondrin as the target neuronal autoantigen.

Immunoprecipitates from homogenized rat hippocampus and cerebellum obtained with the patients' sera presented a protein of approximately 75 kDa in sodium dodecyl sulfate polyacrylamide gel electrophoresis, which was absent if the homogenates were incubated with normal control sera (figure 2A). Using matrix-assisted laser desorption/ionization–time of flight, the 75 kDa protein fraction was identified as neurochondrin from Rattus norvegicus (UNIPROT acc. #O35095).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2 Immunoprecipitation and antigen identification

Lysates of rat hippocampus and cerebellum were incubated with patient or control sera (1:33). Immunocomplexes were isolated with protein-G-coated magnetic beads, eluted by sodium dodecyl sulfate, and subjected to sodium dodecyl sulfate polyacrylamide gel electrophoresis analysis followed by (A) staining with colloidal Coomassie or (B) Western blot using polyclonal rabbit antineurochondrin. Frames indicate the position of the immunoprecipitated antigen at about 75 kDa.

Western blot analysis of the immunoprecipitates received with the patients' sera but not controls revealed a strong reaction at 75 kDa using a polyclonal rabbit antineurochondrin antibody (figure 2B). When used in IFA, rabbit antineurochondrin produced a fluorescence pattern on rat hippocampus matching that generated by the patients' sera (figure 3A). In IFA with permeabilized hippocampal neurons, both the patient sera and the rabbit antineurochondrin antibody showed a predominant reactivity against the cytoplasm of the neuronal cell body but also stained the neurites to a weaker extent (figure 3B).

• Download figure
• Open in new tab
• Download powerpoint

Figure 3 Double-staining of hippocampal tissue and cells with patient serum and rabbit antineurochondrin antibody

Immunofluorescence staining of rat hippocampus tissue section (A) or formalin-fixed and TritonX-100 permeabilized rat hippocampal neurons (B) with patient sera (green) and antineurochondrin antibody (red). Nuclei were counterstained by incubation with TO-PRO-3 iodide (blue). (A, B) ×200 magnification.

As a proof for correct antigen identification, the patients' samples were then tested by IFA using transfected HEK293 cells that expressed neurochondrin (figure 4A). All patients' sera and CSFs reacted with the neurochondrin-expressing cells with 10-fold higher endpoint titers than on tissue and, consistently, the antibody subclass was IgG1. All samples also reacted with recombinant neurochondrin in immunoblot using HEK293-neurochondrin lysate. In contrast, mock-transfected cells did not demonstrate any specific antibody binding. Specific antibody indices as determined with HEK293-neurochondrin were also AI >4 in patients 1 and 3.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4 Verification of neurochondrin as the novel autoantigen by indirect immunofluorescence

(A) Indirect immunofluorescence using acetone-fixed neurochondrin or mock-transfected HEK293 cells incubated with 1:1,000 diluted serum or 1:10 diluted CSF of patient 1, patient 2, or a healthy control (green). (B) Neutralization of immunofluorescence reaction on neuronal tissues. Patient serum (green) was preincubated with extracts of HEK293 cells transfected with neurochondrin or with empty vector as control. The extract containing neurochondrin abolished the immune reaction. Nuclei were counterstained by incubation with TO-PRO-3 iodide (blue). (A) Hippocampus tissue section: ×200 magnification. (B) Rat hippocampal neurons: ×400 magnification.

The reaction of the patients' autoantibodies on tissue could be abolished by preincubation with HEK293 lysate containing neurochondrin (figure 4B). Antibody binding was unaffected when a comparable fraction from mock-transfected HEK293 cells was used.

Anti-neurochondrin autoantibodies are not detected in several control cohorts.

Sera from 33 patients with various neural autoantibody-associated neurologic syndromes in part also involving cerebellum and brainstem (anti-NMDA receptor, anti-Hu, anti-Yo, anti-Ri, anti-AQP4, anti-LGI1, anti-CASPR2), 36 treatment-naive patients with RRMS, most of them with infratentorial lesions, 20 patients with MSA-c, 35 patients with hereditary cerebellar ataxias, 150 consecutive patients with various other neurologic disorders, and 37 healthy controls were analyzed by IFA and with HEK293-neurochondrin in parallel to the samples of the patients. None of the control sera produced a similar immunofluorescence pattern as the patients' sera on rat brain tissue, and all were negative when tested on HEK293 cells expressing neurochondrin. Hence, we consider IgG antibodies against neurochondrin specific for patients with autoimmune cerebellar degeneration. Moreover, these autoantibodies do not seem to merely represent an unspecific secondary epiphenomenon of cerebellar and brainstem inflammation or degeneration, although this cannot be fully excluded.

Characterizing the autoimmune response in neurochondrin antibody-associated cerebellitis.

Immunophenotyping of PB and CSF (figure e-6) revealed that inflammatory changes were most pronounced in patient 1, patient 2 was intermediate, and patient 3 was largely within normal limits in line with the clinical disease activity. Before long-term immunosuppression with cyclophosphamide, patient 1 exhibited elevated numbers (figure e-6B) and fractions (figure e-6A) of CD4+ T cells and activated CD4+ HLADR+ T cells in PB and CSF. In contrast, numbers (figure e-6B) and fractions (figure e-6A) of CD8+ T cells and activated CD8+ HLADR+ T cells as well as CD19+ B cells and CD138⁺ CD19+ plasma cells were elevated in CSF but not in PB. Following long-term immunosuppression, these changes were largely regressive (figure e-6). The peripheral and intrathecal accumulation and activation of T cells in addition to B cells might reflect a T-cell response towards neurochondrin-expressing neurons in autoimmune cerebellar degeneration with neurochondrin antibodies.

Indeed, TCR spectratyping in PB (figure e-7) revealed increased fractions of Vβ families with skewed or oligoclonal spectratypes and decreased fractions of Vβ families with normal or shifted spectratypes in CD4+ more than CD8+ T cells in patients 1 and 2 but not patient 3 compared to a cohort of 11 healthy controls. In patients 1 and 2, the degree of TCR repertoire perturbation was comparable or even higher compared to that reported for patients with multiple sclerosis or chronic inflammatory demyelinating polyneuropathy.^20,21 Notably, the CD8+ TCR repertoire showed an unusual high degree of perturbation while the CD4+ TCR repertoire was much more normally distributed in healthy controls (figure e-7) compared to published values.^20,21 These changes largely persisted throughout the disease course and suggest an antigen-driven activation of CD8 T cells and CD4 T cells in addition to the presence of antineurochondrin antibodies. Whether the target of the T-cell immune response is neurochondrin or not is currently unknown.