ARTICLE OPEN ACCESS HLA and microtubule-associated protein tau H1 haplotype associations in anti-IgLON5 disease Carles Gaig, MD, PhD, Guadalupe Ercilla, MD, PhD, Xavier Daura, PhD, Mario Ezquerra, PhD, Correspondence Dr. Gaig Ruben Fernández-Santiago, PhD, Eduard Palou, MD, PhD, Lidia Sabater, PhD, Romana Höftberger, MD, cgaig@clinic.ub.es Anna Heidbreder, MD, PhD, Birgit Högl, MD, PhD, Alex Iranzo, MD, PhD, Joan Santamaria, MD, PhD, Josep Dalmau, MD, PhD, and Francesc Graus, MD, PhD Neurol Neuroimmunol Neuroinflamm 2019;6:e605. doi:10.1212/NXI.0000000000000605 Abstract Objectives We investigated the associations with HLA and microtubule-associated protein tau (MAPT) H1 haplotype in anti-IgLON5 disease, a recently identified disorder characterized by gait instability, brainstem dysfunction, and a prominent sleep disorder in association with IgLON5 antibodies and pathologic findings of a novel neuronal-specific tauopathy. Methods We compared the HLA alleles and MAPT H1/H1 genotype of 35 patients with anti-IgLON5 with healthy controls. The on-line server tool NetMHCIIpan 3.1 was used to predict the IgLON5 peptide binding to HLA Class II molecules. Results The HLA-DRB1*10:01-DQB1*05:01 haplotype was overrepresented in patients with anti- IgLON5 disease (OR = 54.5; 95% CI: 22.2–133.9, p < 0.0001). In addition, HLA-DQA was genotyped in 27 patients, and 25 (92.6%) of them had DQ molecules composed by DQA1*01 and DQB1*05 chains compared with 148/542 (27.3%) controls (OR = 43.9; 95% CI: 10.4–185.5, p < 0.0001). Patients DRB1*10:01 positive developed more frequently sleep or bulbar symptoms than those carrying other HLA alleles (70.0% vs 26.7%; p = 0.011). Prediction algorithms identified 2 IgLON5 peptides (1 located in the signal sequence) that showed strong binding to HLA-DRB1*10:01 and other HLA-DRB1, but not to HLA-DQA and HLA-DQB molecules. The MAPT H1/H1 homozygous genotype was present in 20/24 (83.3%) anti- IgLON5 Caucasian patients compared with 54/116 (46.5%) healthy controls (p = 0.0007). Conclusions The robust association of anti-IgLON5 disease with distinct HLA Class II molecules supports a primary autoimmune origin. The significant association of MAPT H1 haplotype also suggests that an underlying neurodegenerative process could be involved in anti-IgLON5 disease. From the Service of Neurology (C.G., A.I., J.S.), Hospital Clinic, Barcelona, Spain; Multidisciplinary Sleep Disorders Unit (C.G., A.I., J.S.), Hospital Clinic, Barcelona, Spain; Department of Immunology (G.E., E.P.), Hospital Clı́nic, Barcelona, Spain; Institute of Biotechnology and Biomedicine (X.D.), Universitat Autònoma de Barcelona, Barcelona, Spain; Institució Catalana de Recerca i Estudis Avançats (ICREA) (X.D., J.D.), Barcelona, Spain; Laboratory of Parkinson Disease and Other Neurodegenerative Movement Disorders (M.E., R.F.-S.), Institut d’Investigacions Biomèdiques August Pi i Sunyer (IDIBAPS), Barcelona, Spain; Neuroimmunology Program (L.S., J.D., F.G.), Institut d’Investigacions Biomèdiques August Pi i Sunyer (IDIBAPS), Barcelona, Spain; Institute of Neurology (R.H.), Medical University of Vienna, Austria; Institute for Sleep Medicine and Neuromuscular Disorders (A.H.), University Hospital Muenster, Muenster, Germany; Department of Neurology (B.H.), Medical University of Innsbruck, Innsbruck, Austria; and Department of Neurology (J.D.), University of Pennsylvania, Philadelphia, PA. Go to Neurology.org/NN for full disclosures. Funding information is provided at the end of the article. The Article Processing Charge was funded by Fondo de Investigaciones Sanitarias, FIS 18/00067. This is an open access article distributed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivatives License 4.0 (CC BY-NC-ND), which permits downloading and sharing the work provided it is properly cited. The work cannot be changed in any way or used commercially without permission from the journal. Copyright © 2019 The Author(s). Published by Wolters Kluwer Health, Inc. on behalf of the American Academy of Neurology. 1 Glossary CBD = corticobasal degeneration; MAPT = microtubule-associated protein tau; PSP = progressive supranuclear palsy. The anti-IgLON5 disease is a recently identified neurologic Standard protocol approvals, registrations, disorder, characterized by progressive gait instability, brain- and patient consents stem dysfunction, and a characteristic sleep disorder. The The study was approved by the Hospital Clinic Ethics biological hallmark of this disease is the presence of antibodies Committee. All patients or proxies gave written informed against IgLON5, a neural cell adhesion molecule of unknown consent for the storage and use of serum, CSF, and clinical function.1 Neuropathologic studies demonstrated a novel information for research purposes. neuronal tauopathy with predominant involvement of the hypothalamus and brainstem.2 Although these findings favor HLA genotyping a primary tauopathy, the presence of highly specific antibodies HLA Class I and Class II genotyping was performed by against a neuronal surface protein and the tight association of techniques based on DNA-PCR and polymorphism identifi- the disease with HLAs DRB1*10:01and DQB1*05:01 (pre- cation by reverse hybridization with specific probes and fluo- viously shown in 15 patients) alternatively suggested an rescence labeling of hybridized fragments (polymerase chain immune-mediated pathogenesis.3 reaction [PCR]-sequence-specific oligonucleotide probes) (Immucor GTI Diagnostics Inc., Waukesha) in combination Familial forms of tauopathies are caused by mutations in the with genomic DNA sequencing by Sanger methodology microtubule-associated protein tau (MAPT) gene.4 In addi- (PCR-sequenced-based typing) (SuBiTo, Inno-Train Diag- tion, the frequency of the MAPT H1/H1 homozygous ge- nostik GmbH and the software HiType, Kronberg, Germany), notype has been described to be consistently increased in providing an unambiguous resolution of the 2 fields that define sporadic tauopathies including progressive supranuclear palsy the HLA molecule polymorphism according to the WHO (PSP) and corticobasal degeneration (CBD), but its fre- official current nomenclature (hla.alleles.org/nomenclature/ quency in IgLON5 disease is unknown.5,6 nomenc_updates.html). It is also unknown whether the HLA alleles associated with MAPT genetic studies anti-IgLON5 disease can present IgLON5 peptides to Genotyping of genomic DNA was performed using the prede- generate a specific autoimmune response.7 We report here signed TaqMan assay for MAPT rs1800547 (A/G), which the identification of IgLON5 peptides with potential defines the H1/H2 haplotypes and run on a Step One Plus Real- binding to HLA Class II molecules using in silico techni- time PCR System (Applied Biosystems, Foster City, CA) as ques. In addition, we performed DNA sequencing analyses previously described.8 Twenty-seven patients with DNA avail- of the MAPT gene and determined the association with able and 116 healthy controls were genotyped. All but 3 patients the H1 haplotype and examined whether the most preva- were of European Caucasian origin. In addition, in 7 patients lent HLA allele (HLA-DRB1*10:01) in patients with anti- with anti-IgLON5, we performed a mutational screening of IgLON5 disease segregated with a specific clinical MAPT exons 1, 9, 10, 11, 12, and 13, where pathogenic muta- phenotype. tions have been previously identified in several neurodegener- ative diseases including frontotemporal dementia or PSP.4 We performed a touchdown PCR amplification on a PTC-100 PCR Methods (MJ Research; Watertown, MA) using 20–50 ng of total ge- nomic DNA and using primer pairs as previously described.9 We Patients used the BigDye Terminator v3.1 Cycle Sequencing Kit The study includes 35 patients with IgLON5 antibodies according to the manufacturer’s instructions (Applied Bio- identified in the Neuroimmunology Laboratory of the In- systems, Foster City, CA). Sequences were run on an ABI3100 stitute of Biomedical Research August Pi i Sunyer (IDI- automatic sequencer (Applied Biosystems, Foster City, CA) and BAPS), Hospital Clinic (Barcelona, Spain). Genomic DNA analyzed using the DNA baser v4.36.0 software (dnabaser.com). of 28 patients was available for HLA and MAPT analysis. For the remaining 7 patients, HLA-DR genotyping was per- HLA-IgLON5 peptide binding prediction formed elsewhere and the information used for analysis of The human IgLON5 sequence was retrieved from UNIPROT HLA-DR frequencies and association with specific neuro- database entry A6NGN9 (uniprot.org) encompassing the logic symptoms. Methods of detection of IgLON5 anti- signal peptide (amino acid positions 1–30) and the mature bodies and collection of clinical information have been protein (amino acid positions 31–336). The on-line server previously reported.1,3 The frequency of HLA and MAPT tool NetMHCIIpan 3.1 was used for the prediction of haplotypes in patients was compared with that of healthy IgLON5 peptide binding to HLA Class II molecules (cbs.dtu. Caucasian Spanish controls (542 controls for the HLA study dk/services/NetMHCIIpan/).10 The full protein sequence and 116 for the MAPT genotyping). was submitted as 15-residue overlapping fragments with 1 gap 2 Neurology: Neuroimmunology & Neuroinflammation | Volume 6, Number 6 | November 2019 Neurology.org/NN residue at the N-terminal flanking position. Binding was table 1. The median age at disease onset was 63 years (range: considered strong if the predicted IC50 (half maximal in- 42–81 years), and 51% were female. Thirty-two patients were hibitory concentration) value was <50 nM and ranked <1% of European Caucasian origin. The countries of origin were among 200.000 random peptides. Only strong binders were Spain (12 patients), Austria (8), Germany (7), Brazil (2), taken for analysis.10–12 Because the 9-residue binding core is France (1), Belgium (1), Italy (1), Australia (1), India (1), contained in several 15-mer peptides, only the strongest and Philippines (1). Twenty of the 35 patients (57.1%) car- predicted binder of any series of 15-mer peptides with ried the HLA-DRB1*10:01-DQB1*05:01 haplotype com- a 9-residue or longer overlap was used for calculations. pared with 2.4% of healthy controls (OR = 54.5; 95% CI: 22.2–133.9, p < 0.0001). Another 7 patients (20%) carried the Binding prediction was calculated with the 15 different HLA- DRB1*01:01 allele (2 were homozygous, and 1 heterozygous DRB1 alleles present in patients (*10:01, *01:01, *01:02, *03: with DRB1*10:01) representing a similar frequency seen in 01, *04:01, *04:03, *07:01, *09:01, *11:01, *12:01, *13:01, *14: controls (14.1%) (p = 0.34). The DRB1*03:01 allele was 01, *14:04, *15:01, and *16:01). Because of the strong linkage present in 8 patients (22.9%) (2 were also DRB1*10:01 disequilibrium between DRB1 and DQ alleles, the NetMH- positive) representing a similar frequency observed in con- CIIpan 3.1 algorithm was also applied to the DQ alleles. The 4 trols (20.6%) (p = 0.75). The remaining 3 patients had other possible combinations for each individual were considered, HLA-DRB1 (table 1). HLA-DQA was genotyped in 27 combining each DQA1 with DQB1 alleles present in 1 patient patients, and 25 (92.6%) of them had DQ molecules com- in cis and/or trans, considering both chromosomes. posed by DQA1*01 and DQB1*05 chains compared with 148/542 (27.3%) controls (OR = 43.9; 95% CI: Molecular modeling of HLA Class II-IgLON5 10.4–185.5, p < 0.0001). Although the high frequency of these peptide interactions alleles could be partially explained by its genetic linkage with To confirm the HLA-IgLON5 peptide binding predictions the HLA-DRB1*10:01 allele (allelefrequencies.net), the generated by the NetMHCIIpan 3.1 program, molecular DQA1*01/DQB1*05 molecule was present in 11/13 (84.6%) modeling of complexes including HLA-DRB1*10:01 and *01: patients without DRB1*10:01. 01 and the 2 IgLON5 peptides identified as strong binders was performed using a simulation protocol (appendix e-1, HLA Class I alleles were studied in 27 patients with anti- links.lww.com/NXI/A136). IgLON5 (table 1). Five patients (18.5%) were HLA-B*37: 01 compared with 17 controls (3.1%) (OR = 7.0; 95% CI: Statistical analysis 2.4–20.8; p = 0.0004). The allele HLA-A*01:01 was also OR with 95% CI was calculated to measure the association of overrepresented in patients with anti-IgLON5 compared HLA alleles and MAPT haplotypes with anti-IgLON5 disease. with controls (47.4% vs 20.5%; OR = 2.7; 95% CI: 1.2–5.9; For MAPT haplotypes, we performed a genotypic association p = 0.01). These associations were expected considering analysis using unconditional logistic regression models, as the strong linkage disequilibrium of these 2 alleles implemented in SNPstat software, to compare the frequencies with DRB1*10:01 that conform a common haplotype in of MAPT H1/H1—which is the risk genotype described in Caucasians.13 other tauopaties (CBD or PSP)—with the other 2 genotypes H1/H2 and H2/H2 in patients with anti-IgLON5 and con- Association of the HLA-DRB1*10:01 allele with trols (dominant model). The χ 2 test and Fisher exact prob- neurologic symptoms ability test were used to determine the significance of Next, we investigated whether the presence of the most differences in the distribution of the clinical features between prevalent HLA-DRB1*10:01 allele associated with distinct patients who were HLA-DRB1*10:01 positive or carried other clinical features (table 2). At disease onset, 14 DRB1*10:01- DRB1* alleles. The nonparametric Mann-Whitney U test was positive patients (70%) presented with sleep or bulbar clinical used to compare medians. p Values < 0.05 were considered phenotypes. Conversely, these 2 phenotypes tended to be less significant. The statistical analyses were performed using frequent (26.7%) in patients with other HLA alleles, who commercially available software (SPSS, Version 22.0). often presented with a PSP-like phenotype or cognitive im- pairment. Accordingly, a trend to a higher frequency of par- Data availability asomnia, sleep breathing alterations, and bulbar dysfunction Data from patients reported within the article are available was observed in DRB1*10:01 carriers. Dysautonomia tended and will be shared anonymously by request from any qualified also to be more frequent in DRB1*10:01 carriers, whereas investigator. cognitive impairment occurred more frequently in non- DRB1*10:01 carriers. Improvement with immunotherapy occurred in 5 (26.3%) of 19 patients. DRB1*10:01-negative Results patients showed a trend to a better response to immuno- HLA subtypes in anti-IgLON5 disease therapy (3 of 5 cases improved). IgG subclass predominance General information and the HLA and MAPT genotyping of was IgG4 in all DRB1*10:01-negative patients compared with the 35 patients with anti-IgLON5 disease are summarized in 70% in DRB1*10:01-positive cases (p = 0.027). Neurology.org/NN Neurology: Neuroimmunology & Neuroinflammation | Volume 6, Number 6 | November 2019 3 4 Neurology: Neuroimmunology & Neuroinflammation | Volume 6, Number 6 | November 2019 Table 1 Clinical characteristics, HLA genotyping, and tau haplotype N Age, sex Syndromea HLA-A HLA- HLA-B HLA-B HLA-C HLA-C HLA-DRB HLA-DRB HLA-DQA HLA-DQA HLA-DQB HLA-DQB Tau haplotype 11,3 59, M Sleep 11:01 66:01 27:05 35:01 01:02 04:01 10:01 11:01 01:05 05:05 05:01 03:01 H1/H1 21,3 54, M Sleep 01:01 29:02 27:05 37:01 01:02 06:02 10:01 14:01 01:05 —b 05:01 05:03 H1/H1 1,3 3 54, M Sleep 32:01 68:01 27:05 37:01 02:02 06:02 10:01 13:01 01:03 01:05 05:01 06:01 H1/H1 41,3 76, M Bulbar 01:01 24:02 08:01 15:01 03:03 15:01 10:01 11:01 01:02 05:01 05:01 03:01 H1/H1 53 42, M Bulbar 01:01 02:01 37:01 44:02 05:01 06:02 10:01 04:01 01:01 03:02 05:01 03:01 H1/H1 6 53, F Sleep 01:01 02:11 15:07 40:06 07:01 12:02 10:01 03:01 01:01 05:01 05:01 02:01 H1/H1 3 7 65, F Cognitive 02:01 66:03 07:02 44:03 02:02 04:01 10:01 13:01 01:01 01:03 05:01 06:03 H1/H1 83 53, F Sleep 23:01 24:02 13:02 50:01 06:02 —b 10:01 07:01 01:01 02:01 05:01 02:02 H1/H1 93 49, M Sleep 01:01 02:01 07:02 44:02 05:01 07:01 10:01 01:01 01:01 —b 05:01 —b H1/H1 10 65, M Sleep 01:01 03:01 07:02 37:01 07:01 07:02 10:01 15:01 01:01 01:02 05:01 06:02 H1/H2 113 64, M Bulbar ND ND ND ND ND ND 10:01 UNK ND ND 05:01 UNK ND 123 62, F Bulbar ND ND ND ND ND ND 10:01 03:01 ND ND 05:01 02:01 ND 13 64, M PSP-like ND ND ND ND ND ND 10:01 UNK ND ND 05:01 UNK ND 3 14 61, F Cognitive ND ND ND ND ND ND 10:01 07:01 ND ND 05:01 02:02 ND 153 63, F Sleep ND ND ND ND ND ND 10:01 13:01 ND ND 05:01 06:03 ND 16 68, M Bulbar ND ND ND ND ND ND 10:01 UNK ND ND 05:01 UNK ND 17 57, M PSP-like 02:01 11:01 37:01 44:02 06:02 07:04 10:01 12:01 01:01 05:05 05:01 03:01 H1/H1 18 70, M Bulbar 02:01 30:02 07:05 18:01 05:01 15:05 10:01 13:01 01:03 01:04 05:01 06:03 H1/H2 1927 75, F Cognitive 03:01 33:01 14:02 55:01 03:03 08:02 10:01 04:01 01:04 03:01 05:01 03:02 H1/H1 20 62, F Ataxia 01:01 03:01 18:01 41:01 07:01 17:01 10:01 09:01 01:04 03:02 05:01 03:03 H1/H1 21 80, F Cognitive 02:01 25:01 15:01 44:02 03:03 05:01 01:01 14:01 01:01 —b 05:01 05:03 H2/H2 22 60, M Stiff-person 01:01 02:01 35:01 51:01 04:01 14:02 01:01 —b 01:01 —b 05:01 —b ND 23 62, M Sleep 03:01 26:01 39:01 56:01 01:02 12:03 01:01 11:01 01:01 05:05 05:01 03:01 H1/H1 Neurology.org/NN 24 52, F Ataxic gait 01:01 02:01 37:01 57:01 06:02 — b 01:01 —b 01:01 —b 05:01 —b H1/H2 25 71, F Cognitive ND ND ND ND ND ND 01:01 07:01 ND ND 05:01 03:03 ND Continued Neurology.org/NN Table 1 Clinical characteristics, HLA genotyping, and tau haplotype (continued) N Age, sex Syndromea HLA-A HLA- HLA-B HLA-B HLA-C HLA-C HLA-DRB HLA-DRB HLA-DQA HLA-DQA HLA-DQB HLA-DQB Tau haplotype 26 62, M Cognitive 02:01 29:02 35:01 44:03 04:01 16:01 01:01 07:01 01:01 02:01 05:01 02:02 H1/H1 27 69, M PSP-like 01:01 29:01 07:02 08:01 07 15 03:01 14:01 01:04 05:01 05:03 02:01 H1/H1 3 28 65, M PSP-like 02:05 30:02 18:01 50:01 05:01 06:02 03:01 07:01 02:01 05:01 02:01 03:03 H1/H1 29 71, M Bulbar 01:01 25:01 08:01 18:01 07:01 12:03 03:01 04:01 03:01 05:01 02:01 03:02 H1/H1 30 67, F PSP-like 01:01 24:02 08:01 14:02 07:01 08:02 03:01 01:02 01:01 05:01 05:01 02:01 H1/H1 31 66, F Bulbar 02:01 11:01 14:02 18:01 07:01 08:02 03:01 14:01 01:04 05:01 05:03 02:01 H1/H1 Neurology: Neuroimmunology & Neuroinflammation | Volume 6, Number 6 | November 2019 1,3 32 69, F Sleep 01 32 08 15 03 07 03:01 16:01 ND ND 05:02 02:01 H1/H1 33 81, F PSP-like 24:02 —b 35:02 35:03 04:01 12:03 04:03 14:01 01:01 03:02 05:03 03:01 H1/H1 34 67, F Cognitive 11:01 — b 35:01 51:01 04:01 — b 04:03 14:01 01:01 03:02 05:03 03:01 H1/H1 3 35 69, F Cognitive 11:01 34:01 38:02 39:01 07:02 12:02 09:01 14:04 01:01 03:02 05:03 03:03 H2/H2 Abbreviations: F = female; M = male; ND = not done; PSP = progressive supranuclear palsy; UNK = unknown. Seventeen patients have been previously reported (references 1, 3, and 27). a Syndromes were defined as described.3 Patient 22 presented with a different syndrome characterized by fasciculations, cramps, and stiffness in the lower limbs reminiscent of stiff-person syndrome (sleep, bulbar, gait, and cognitive symptoms were absent). b Indicates that the allele is homozygous. 5 Table 2 Clinical features in patients HLA-DRB1*10:01 positive and negative HLA-DRB1*10:01 HLA-DRB1*non-10:01 (n = 20) (n = 15) p Value Age at onset (y) 62 (42–76) 67 (52–81) 0.069a Male sex 12 (60) 5 (33.3) 0.12b Age at diagnosis (y) 64.5 (46–77) 71 (61–83) 0.023a Chronic presentationd 17 (85) 9 (60) 0.13c Sleep or bulbar profile vs others 14 (70) 4 (26.7) 0.011b Parasomnia 14 (70) 4 (26.7) 0.011b Sleep breathing disorder 19 (95) 8 (53.3) 0.011c Bulbar dysfunction 17 (85.5) 8 (53.3) 0.062c Gait difficulties 12 (60) 9 (60) 1.0b Cognitive impairment 9 (45) 12 (80) 0.036b Oculomotor abnormalities 10 (50) 8 (53.3) 0.84b Dysautonomia 12 (60) 4 (26.7) 0.05b Chorea 3 (15) 2 (13.3) 1.0c CSF pleocytosis (>5 cells)e 5/17 (29.4) 1/12 (8.3) 0.35c Elevated proteins (>45 mg/dL)e 10/17 (58.8) 5/12 (41.7) 0.36b Anti-IgLON5 negative in CSFf 1/18 (5.6) 2/14 (15.4) 0.57c IgG4-predominant IgG subclass 14 (70) 15 (100) 0.027c Tau H1/H1 haplotype 12/14 (85.7) 9/12 (75) 0.63c Response to immunotherapyg 2/14 (14.3) 3/5 (60) 0.084c Disease duration (mo)h 44.5 (9–180) 96 (8–228) 0.24a Death 6 (30) 3 (20) 0.7c Data are presented as median and (range) or in number and (%). a Mann-Whitney U test. b 2 χ test. c Fisher exact test was used. d Time from symptom onset to initial medical consultation >4 mo. e CSF analysis was performed in 29 of 35 patients. f IgLON5 antibodies were tested in CSF in 32 patients; in 3 (9.4%) of them, the antibodies were only detectable in serum. g Immunotherapy was performed in 18 of 35 patients. Clinically significant improvement occurred in 5 (26.3%). h Time from symptom onset until the last visit or death. HLA-IgLON5 peptide binding prediction or trans, combinations of the DQA and DQB molecules. The binding ability of IgLON5 peptide sequences to 15 Similarly, no IgLON5 peptide sequences were binders to HLA-DRB1 alleles using the NetMHCIIpan 3.1 server HLA Class I–associated alleles A*01:01 and B*37:01 using showed that only 3 DRB1 molecules, the DRB1*01:01, the NetMHC 3.4 server.14 DRB1*10:01, and DRB1*09:01, were strong binders for the same 2 IgLON5-derived peptides (table 3). The epitope Given that the signal peptide sequence may vary among with the strongest predicted affinity was LRLLAAAAL, different family members and isoforms of a protein family, belonging to the signal peptide of IgLON5. The second we determined the prediction of binding to HLA Class II epitope, IVHVPARIV, is located in the Ig-like 2 domain of molecules the signal peptide sequences of the other mem- the protein. Seven different HLA-DRB1 molecules also had bers of the IgLON family, IgLON1, 2, 3, and 4. In contrast to a predicted binding capacity for one of these peptides, the IgLON5 peptide LRLLAAAAL, none of the other signal mainly IVHVPARIV. Another peptide (WTSDPRVRL) peptide sequences showed strong predicted binding affinity. showed binding capacity for only HLA-DRB1*03:01 pres- Only sequences from isoforms 1, 2, 5, 7, and 9 of IgLON2 ent in 8 patients (table 3). No IgLON5 peptide sequences (Uniprot accession NP-057606.1, NP_001041674.1, NP_ were defined as strong binders considering all possible, cis 001338930.1, NP_001338932.1, and NP_001338934.1, 6 Neurology: Neuroimmunology & Neuroinflammation | Volume 6, Number 6 | November 2019 Neurology.org/NN Table 3 IgLON5 peptides with strong binding affinity for HLA-DRB1 molecules in anti-IgLON5 disease Affinity Full peptide Core peptide MHC molecule (15-mer) (9-mer) IC50 <50nM Rank DRB1*10:01 GARLRLLAAAALAGL LRLLAAAAL 11.04 0.04 QVYLIVHVPARIVNI IVHVPARIV 30.04 0.80 DRB1*01*01 GARLRLLAAAALAGL LRLLAAAAL 2.71 0.01 QVYLIVHVPARIVNI IVHVPARIV 4.29 0.25 DRB1*09:01 GARLRLLAAAALAGL LRLLAAAAL 28.19 0.30 QVYLIVHVPARIVNI IVHVPARIV 35.39 0.50 DRB1*03:01 GNDRWTSDPRVRLLI WTSDPRVRL 40.89 0.50 DRB1*04:01 GARLRLLAAAALAGL LRLLAAAAL 47.65 0.90 DRB1*07:01 QVYLIVHVPARIVNI IVHVPARIV 12.11 0.17 DRB1*12:01 QVYLIVHVPARIVNI IVHVPARIV 2.74 0.25 DRB1*13:01 QVYLIVHVPARIVNI IVHVPARIV 12.43 0.01 DRB1*14:01 QVYLIVHVPARIVNI IVHVPARIV 24.24 0.12 DRB1*14:04 QVYLIVHVPARIVNI IVHVPARIV 20.03 0.30 DRB1*15:01 QVYLIVHVPARIVNI IVHVPARIV 20.45 0.17 Binding of epitopes of IgLON5 was predicted using NertMHCIIpan 3.1 server in accordance with the instruction of the server. Binding affinity below 50 nM and a rank <1% among 200.000 random peptides was associated with strong binders. In bold, the 3 DRB1 molecules that bind to both IgLON5 peptides. respectively) resulted in weak predicted binding affinities to The computed scores for the binding of the other strong HLA-DRB1 alleles. The second epitope, IVHVPARIV, was binder peptide IVHVPARIV to DRB1*01:01 and DRB1*10: not present in any of the other members of the IgLON 01 were −75 and −57 kJ/mol, respectively. In this case, the family. difference of 18 kJ/mol is significantly larger than that pre- dicted by NetMHCIIpan 3.1. Nevertheless, using any of these Because tau protein deposits have been described in the 2 methods, a higher affinity for DRB1*01:01 is predicted. brain of patients with anti-IgLON5 antibodies who un- Comparing the values obtained for the 2 peptides, we also derwent autopsy, a prediction study was also made on the found that both methods predicted a higher affinity of binding ability of tau protein sequences (Uniprot P10636) LRLLAAAAL than IVHVPARIV for either of the 2 HLA to the indicated HLA-DRB1 and DQA-DQB alleles. No molecules. Considering the binding scores of IVHVPARIV sequences with high affinity for HLA molecules were relative to LRLLAAAAL, and given that the other anchoring identified. residues are significantly conserved, we conclude that Arg (R) in pocket P7 is likely the main responsible for the lower Molecular modeling of HLA Class II-IgLON5 binding affinity of IVHVPARIV relative to LRLLAAAAL for peptide interactions the 2 HLA molecules. We next modeled the interaction of the strong binder IgLON5 peptide LRLLAAAAL with DRB1*01:01 and Genetic analysis of the MAPT gene in anti- DRB1*10:01. The computed structure-based scores for the IgLON5 disease binding of this peptide to DRB1*01:01 and DRB1*10:01 were The MAPT H1/H1 homozygous genotype was signifi- −97 and −91 kJ/mol, respectively. Although based on fun- cantly overrepresented with respect to H1/H2 and H2/H2 damentally different principles, this finding agrees with the in 27 anti–IgLON5-positive patients (81.5% were H1/H1) NetMHCIIpan 3.1 binding affinity predictions, thus sug- compared with 116 healthy controls (46.5% H1/H1), OR gesting a slightly higher affinity of LRLLAAAAL for 5.05 (95% CI: 1.79–14.25), p = 0.0007. Similar results DRB1*01:01 than for DRB1*10:01. Taking the structure into were found when the sample was limited to the 24 anti– account, the clearest differences between the pockets in the IgLON5-positive patients of European Caucasian origin peptide-binding grooves of DRB1*01:01 and DRB1*10:01 are (83% H1/H1) (p = 0.0007). Finally, we did not find probably found in pocket P9, where there is absence of a well- mutations in the commonly mutated MAPT exons 1, 9, 10, defined pocket in DRB1*10:01, whereas a shallow pocket is 11, 12, and 13 in a subset of 7 patients with anti-IgLON5 visible in DRB1*01:01 (figure). disease. Neurology.org/NN Neurology: Neuroimmunology & Neuroinflammation | Volume 6, Number 6 | November 2019 7 as cause of the disease. One of the hypotheses to explain the Figure Modeling of the interaction between Class II mole- link between antibody-mediated diseases and specific HLA- cules and IgLON5 peptides DR or DQ alleles is that these molecules present peptides of the antibody-targeted antigen to autoimmune-prone CD4+ T cells, which in turn induce B-cell proliferation and dif- ferentiation into antibody-secreting cells.15 In anti-IgLON5 disease, there is an overrepresentation of DRB1*10:01; this is an infrequent allele that in all population studies repre- sents less than 5% of individuals (allelefrequencies.net). The DRB1*10:01 allele has been linked to an increased risk of several rheumatological autoimmune diseases.16 Indeed, in HLA-DRB1*10:01–positive patients with rheumatoid ar- thritis and antibodies against citrullinated type II collagen, CD4+ T cells are activated when a peptide from citrullinated type II collagen protein is presented to T cells by DRB1*10: 01 but not by other DRB1 molecules showing the fine specificity of the peptide-binding motifs of the HLA molecules.17 The present study shows that the DRB1 alleles identified in patients with anti-IgLON5 disease, DRB1*01:01 and Detail of the pockets P7 and P9 in the modeled complexes between peptides LRLLAAAAL (A) and IVHVPARIV (B) of IgLON5 and molecules HLA-DRB1*01: DRB1*10:01, are predicted to bind with high affinity 2 01 and HLA-DRB1*10:01. Residues 7–9 (from right to left) of the peptides are peptides of IgLON5. Molecular modeling of the interaction represented as sticks, with the carbons in golden yellow. Protein residues forming the pockets are represented as sticks, with carbon atoms in teal of the 2 IgLON5 peptides (LRLLAAAAL and IVHV- blue. Covering the protein atoms, their Connolly surface is represented to PARIV) with the strongest binding with DRB1*01:01 and facilitate the visualization of the pockets. Noncarbon atoms are colored as follows: N-blue, H-light gray, O-red, and S-yellow. It can be observed that P9 DRB1*10:01 confirmed the high affinity of this binding. (cavity at the left of each panel) is shallower in HLA-DRB1*10:01 than in HLA- LRLLAAAAL corresponds to a sequence of the signal DRB1*01:01. As a result, the aliphatic residues at position 9 of both peptides bind more deeply in HLA-DRB1*01:01. The models also show that the Arg (R) peptide of IgLON5. Unlike other members of the IgLON residue at position 7 in IVHVPARIV is too bulky for the small P7 and has to family, only the IgLON5 signal peptide region showed orient its side chain toward previous pockets, particularly in HLA-DRB1*10: 01. a high binding affinity for DRB1 molecules. Signal peptides contain numerous hydrophobic residues that help to anchor the nascent polypeptide to the membrane.18 Once the protein is anchored in the membrane, the signal sequence is Discussion cleaved and degraded in the cytosol, and the resulting Anti-IgLON5 disease was the first among multiple CNS peptides are known to be strong binders of the T-cell acti- disorders associated with antibodies against synaptic vating MHC Class I molecules. Indeed, signal peptides have receptors or neuronal surface proteins in which an associ- been proposed as adjuvants in vaccination strategies.19 ation with HLA Class II alleles was identified.1 Here, we Thus, although the epitope recognized by the helper T cell confirm the association with HLA-DRB1*10:01-DQB1*05: (e.g., signal recognition peptide) must be linked to that 01 alleles in a larger series of patients and provide several recognized by the B cell (surface peptide of the same anti- novel findings that help to understand the interplay be- gen, recognized by antibodies), the 2 cells do not need to tween the autoimmune and neurodegenerative processes of recognize identical epitopes, a phenomenon called linked the disease. First, the HLA-DRB1*10:01 was strongly as- recognition.20 The in silico techniques to assess potential sociated with anti-IgLON5 disease and was found more binding of specific IgLON5 peptides to HLA molecules prevalent in patients who presented with parasomnias, used in the present study represent an exploratory com- sleep breathing disorder, or bulbar symptoms. Second, 2 putational tool that will require its confirmation in future IgLON5 peptides showed high affinity for the HLA-DRB1 immunologic experiments assessing the proliferative T-cell molecules identified in patients with the disease as pre- response to the selected IgLON5 peptides. dicted by an algorithm using artificial neural networks. These findings were confirmed by molecular modeling Despite the DQA1*01/DQB1*05 haplotype was strongly using the 2 IgLON5 peptides and DRB1*01:01 and associated with anti-IgLON5 disease, no IgLON5 peptide DRB1*10:01, the 2 alleles that showed the strongest affinity could be defined as a strong binder for these HLA alleles. for these peptides. Last, the MAPT H1/H1 genotype was This association can only be partially explained by linkage overrepresented in patients with anti-IgLON5 disease. disequilibrium with the DRB1*10:01 allele, as 84.6% of the patients non-DRB1*10:01 carriers were positive for The robust association between anti-IgLON5 disease and DQA1*01/DQB1*05. No IgLON5 peptide could be defined distinct HLA alleles supports a primary autoimmune event either to be a binder to HLA Class I–associated alleles 8 Neurology: Neuroimmunology & Neuroinflammation | Volume 6, Number 6 | November 2019 Neurology.org/NN (HLA-A*01:01, B*37:01). In this case, the association is event is an inflammatory autoimmune response that results in probably due to linkage disequilibrium of these Class I alleles a stress situation in patients susceptible to neurodegenerative with the DRB1*10:01 allele. Peptide binding to HLA Class I processes, eventually leading to the indicated tauopathy. How- molecules implies antigenic presentation to CD8+ cytotoxic ever, our finding related to MAPT association must be taken T cells,21 but there are no data suggesting a role of CD8+ with caution due to the relative low number of patients and the T cells in anti-IgLON5 disease. potential genetic heterogeneity among different populations. Future validation studies of the MAPT H1/H1 genotype as- Although this series is relatively small, we noted a trend to an sociation with anti-IgLON5 disease are warranted. association between DRB1*10:01 and sleep and bulbar dys- function as predominant symptoms at disease onset.1 Indeed, In summary, the tight association of anti-IgLON5 disease although 78% (14/18) of patients with these clinical pre- with the DRB1*10:01-DQB1*05:01 haplotype and its high sentations had DRB1*10:01, this haplotype was only observed affinity for IgLON5 peptides, particularly the peptide in the in 36% (5/14) of patients who initially presented with PSP- signal sequence of IgLON5, strongly support that the disease like symptoms or cognitive deterioration. With the increasing is autoimmune. Neuronal tau deposits might occur due to number of reports on patients with IgLON5 antibodies, it is a genetic predisposition of the patients, supported by the becoming evident that this disease is heterogeneous and that association with the MAPT H1/H1 genotype. We have the sleep disorder that was characterized in the initial series is previously shown that IgLON5 antibodies cause an irre- not always present.22 Clinical heterogeneity occurs in many versible antibody-mediated internalization of surface diseases, and the cause is probably multifactorial; however, the IgLON5 in cultured hippocampal neurons.31 Therefore, genetic background plays an important role in some disorders. a task for the future is to determine how antibody disruption For example, distinct HLA-DRB1 alleles appear to influence of IgLON5 ultimately leads to neuronal accumulation of the clinical phenotype of inclusion body myositis or the spinal hyperphosphorylated tau. cord lesion load in MS.23,24 Acknowledgment DRB1*10:01 is in strong linkage disequilibrium with The authors thank all physicians who have contributed by DQB1*05 alleles that associate with some IgG4-related dis- providing clinical information of their patients and all patients eases such as MuSK antibody–mediated myasthenia.25 for their generous contribution to research. They acknowl- IgLON5 antibodies are predominantly of the IgG4 subclass, edge support provided by the technician Manel Fernandez. but we found that in up to 30% of DRB1*10:01-positive patients, the percentage of IgG1 was higher than that of IgG4. Study funding Whether the IgG subclass predominance in anti-IgLON5 This study was supported in part by Fondo de Investigaciones disease associates with different HLA haplotypes should also Sanitarias (FIS), FEDER, Spain (FIS 15/00377, FG; FIS 14/ be assessed in future studies. 00203, J. Dalmau, FIS 18/00067, L. Sabater and C. Gaig), NIH RO1NS077851 (J. Dalmau), Fundació Cellex (J. Dal- An unexpected observation in anti-IgLON5 disease was that mau), Centros de Investigación Biomédica en Red de enfer- the brain of the first 3 autopsied patients showed a novel tau- medades neurodegenerativas (CIBERNED) (C. Gaig, opathy with neuronal accumulation of hyperphosphorylated M. Ezquerra, R. Fernández-Santiago, A. Iranzo, and J. Santa- tau preferentially involving the hypothalamus and the teg- maria) Jóvenes Investigadores (JIN) grant of the Spanish mental nuclei of the brainstem.2 A subsequent case report Ministry of Economy and Competitiveness (MINECO) suggested that the tauopathy is not a constant finding and that (grant # SAF2015-73508-JIN) (Ruben Fernández-Santiago), it could be a late event in the evolution of the disease.26 The and Centro de Investigación Biomédica en Red de Enfer- absence of neuronal tau deposits in early stages of the disease medades Raras (CIBERER) (L. Sabater and J. Dalmau). perhaps could explain that some patients improve with im- munotherapy and that brain biopsies showed inflammatory Disclosure infiltrates without tau pathology.27,28 J. Dalmau receives royalties from Athena Diagnostics for the use of Ma2 as autoantibody tests and from Euroimmun Independently of whether autoimmunity or neurodegeneration for the use of NMDAR and GABAB receptor as autoanti- is the first pathogenic event, our current findings suggest that body tests and licensing fees from Euroimmun for the use patients with anti-IgLON5 disease may have a genetic associa- of DPPX, GABAA receptor, and IgLON5 as diagnostic tion with the MAPT H1/H1 genotype. Patients with this ge- tests. F. Graus receives royalties from Euroimmun for the notype are susceptible to several neurodegenerative diseases, use of IgLON5 as a diagnostic test. The rest of the authors particularly tauopathies (PSP and CBD) and synucleinopathies declare that they have no competing interests. Go to such as Parkinson disease and multiple system atrophy.29 There Neurology.org/NN for full disclosures. is robust evidence that inflammation plays a critical role in some neurodegenerative diseases, and it has been postulated that it is Publication history required for the progression of the degenerative process.30 Received by Neurology: Neuroimmunology & Neuroinflammation Thus, it is conceivable that in anti-IgLON5 disease, the initial April 4, 2019. Accepted in final form July 1, 2019. Neurology.org/NN Neurology: Neuroimmunology & Neuroinflammation | Volume 6, Number 6 | November 2019 9 Appendix Authors Appendix (continued) Name Location Role Contribution Name Location Role Contribution Carles Gaig, Hospital Clinic, Author Conception and design Josep Institut Author Conception and design MD Barcelona of the study; Dalmau, MD d’Investigacions of the study; acquisition and Biomèdiques acquisition and analysis of data; and August Pi i Sunyer analysis of data; and drafting of the (IDIBAPS), drafting of the manuscript Barcelona manuscript Guadalupe Hospital Clinic, Author Conception and design Francesc Institut Author Conception and design Ercilla, MD Barcelona of the study; Graus, MD d’Investigacions of the study; acquisition and Biomèdiques acquisition and analysis of data; and August Pi i Sunyer analysis of data; and drafting of the (IDIBAPS), drafting of the manuscript Barcelona manuscript Xavier Universitat Author Acquisition and daura, PhD Autònoma analysis of data and Barcelona drafting of the References manuscript 1. Sabater L, Gaig C, Gelpi E, et al. A novel non-rapid-eye movement and rapid-eye- movement parasomnia with sleep breathing disorder associated with antibodies to Mario Institut Author Acquisition and IgLON5: a case series, characterisation of the antigen, and post-mortem study. Lancet Ezquerra, d’Investigacions analysis of data and Neurol 2014;13:575–586. PhD Biomèdiques drafting of the 2. Gelpi E, Höftberger R, Graus F, et al. Neuropathological criteria of anti-IgLON5- August Pi i Sunyer manuscript related tauopathy. Acta Neuropathol 2016;132:531–543. (IDIBAPS), 3. Gaig C, Graus F, Compta Y, et al. Clinical manifestations of the anti-IgLON5 disease. Barcelona Neurology 2017;88:1736–1743. 4. Ghetti B, Oblak AL, Boeve BF, Johnson KA, Dickerson BC, Goedert M. Invited Ruben Institut Author Acquisition and review: frontotemporal dementia caused by microtubule-associated protein tau gene Fernández- d’Investigacions analysis of data and (MAPT) mutations: a chameleon for neuropathology and neuroimaging. Neuro- Santiago, Biomèdiques drafting of the pathol Appl Neurobiol 2015;41:24–46. PhD August Pi i Sunyer manuscript 5. Baker M, Litvan I, Houlden H, et al. Association of an extended haplotype in the tau (IDIBAPS), gene with progressive supranuclear palsy. Hum Mol Genet 1999;8:711–715. Barcelona 6. Pastor P, Ezquerra M, Tolosa E, et al. Further extension of the H1 haplotype asso- ciated with progressive supranuclear palsy. Mov Disord 2002;17:550–556. Eduard Hospital Clinic, Author Acquisition and 7. Rock KL, Reits E, Neefjes J. Present yourself by MHC class I and MHC class II Palou, MD Barcelona analysis of data and molecules. Trends Immunol 2016;37:724–737. revised the 8. Ezquerra M, Pastor P, Gaig C, et al. Different MAPT haplotypes are associated with manuscript for Parkinson’s disease and progressive supranuclear palsy. Neurobiol Aging 2011;32: intellectual content 547.e11–16. 9. Rizzu P, Van Swieten JC, Joosse M, et al. High prevalence of mutations in the Lidia Institut Author Acquisition and microtubule-associated protein tau in a population study of frontotemporal dementia Sabater, d’Investigacions analysis of data and in the Netherlands. Am J Hum Genet 1999;64:414–421. PhD Biomèdiques revised the manuscript 10. Andreatta M, Karosiene E, Rasmussen M, Stryhn A, Buus S, Nielsen M. Accurate pan- August Pi i Sunyer for intellectual content specific prediction of peptide-MHC class II binding affinity with improved binding (IDIBAPS), core identification. Immunogenetics 2015;67:641–650. Barcelona 11. Karosiene E, Rasmussen M, Blicher T, Lund O, Buus S, Nielsen M. NetMHCIIpan-3.0, a common pan-specific MHC class II prediction method including all three human MHC Romana Medical University Author Acquisition and class II isotypes, HLA-DR, HLA-DP and HLA-DQ. Immunogenetics 2013;65:711–724. Höftberger, of Vienna analysis of data and 12. Jensen KK, Andreatta M, Marcatili P, et al. Improved methods for predicting peptide MD revised the manuscript binding affinity to MHC class II molecules. Immunology 2018;154:394–406. for intellectual content 13. Huang X, Kushekhar K, Nolte I, et al. HLA associations in classical Hodgkin lym- phoma: EBV status matters. PLoS One 2012;7:e39986. Ana University Hospital Author Acquisition and 14. Lundegaard C, Lamberth K, Harndahl M, Buus S, Lund O, Nielsen M. NetMHC-3.0: Heidbreder, Muenster analysis of data and accurate web accessible predictions of human, mouse and monkey MHC class I MD revised the manuscript affinities for peptides of length 8-11. Nucleic Acids Res 2008;36:509–512. for intellectual 15. Sollid LM, Pos W, Wucherpfennig KW. Molecular mechanisms for contribution of content MHC molecules to autoimmune diseases. Curr Opin Immunol 2014;31:24–30. 16. Stevens AM, Kanaan SB, Torok KS, et al. Brief report: HLA-DRB1, DQA1, and Birgit Högl, Medical University Author Acquisition and DQB1 in Juvenile-onset systemic sclerosis. Arthritis Rheumatol 2016;68:2772–2777. MD of Innsbruck analysis of data and 17. Chemin K, Pollastro S, James E, et al. A novel HLA-DRB1*10:01-restricted T cell revised the manuscript epitope from citrullinated type II collagen relevant to rheumatoid arthritis. Arthritis for intellectual Rheumatol 2016;68:1124–1135. content 18. Hegde RS, Bernstein HD. The surprising complexity of signal sequences. Trends Biochem Sci 2006;31:563–571. Alex Iranzo, Hospital Clinic, Author Acquisition and 19. Kovjazin R, Volovitz I, Daon Y, et al. Signal peptides and trans-membrane regions are MD Barcelona analysis of data and broadly immunogenic and have high CD8+ T cell epitope densities: implications for revised the manuscript vaccine development. Mol Immunol 2011;48:1009–1018. for intellectual 20. Snow EC, Noelle RJ, Uhr JW, Vitetta ES. Activation of antigen-enriched B cells. II. content Role of linked recognition in B cell proliferation to thymus-dependent antigens. J Immunol 1983;130:614–618. Joan Hospital Clinic, Author Conception and design 21. Picascia S, Sidney J, Camarca A, et al. Gliadin-specific CD8+ T cell responses re- Santamaria, Barcelona of the study; stricted by HLA class I A*0101 and B*0801 molecules in Celiac disease patients. MD acquisition and J Immunol 2017;198:1838–1845. analysis of data; and 22. Gaig C, Iranzo A, Santamaria J, Graus F. The sleep disorder in anti-lgLON5 disease. revised the manuscript Curr Neurol Neurosci Rep 2018;18:41. . for intellectual 23. Mastaglia FL, Needham M, Scott A, et al. Sporadic inclusion body myositis: HLA- content DRB1 allele interactions influence disease risk and clinical phenotype. Neuromuscul Disord 2009;19:763–765. 10 Neurology: Neuroimmunology & Neuroinflammation | Volume 6, Number 6 | November 2019 Neurology.org/NN 24. DeLuca GC, Alterman R, Martin JL, et al. Casting light on multiple sclerosis het- 28. Honorat JA, Komorowski L, Josephs KA, et al. IgLON5 antibody: neurological erogeneity: the role of HLA-DRB1 on spinal cord pathology. Brain 2013;136: accompaniments and outcomes in 20 patients. Neurol Neuroimmunol Neuroinflamm 1025–1034. 2017;4:e385. doi:10.1212/NXI.0000000000000385. 25. Bartoccioni E, Scuderi F, Augugliaro A, et al. HLA class II allele analysis in MuSK- 29. Caffrey TM, Wade-Martins R. Functional MAPT haplotypes: bridging the gap be- positive myasthenia gravis suggests a role for DQ5. Neurology 2009;72:195–197. tween genotype and neuropathology. Neurobiol Dis 2007;27:1–10. 26. Cagnin A, Mariotto S, Fiorini M, et al. Microglial and neuronal TDP-43 pathology in 30. Fakhoury M. Immune-mediated processes in neurodegeneration: where do we stand? anti-IgLON5-related tauopathy. J Alzheimers Dis 2017;59:13–20. J Neurol 2016;263:1683–1701. 27. Montagna M, Amir R, De Volder I, Lammens M, Huyskens J, Willekens B. IgLON5- 31. Sabater L, Planagumà J, Dalmau J, Graus F. Cellular investigations with human Associated encephalitis with atypical brain magnetic resonance imaging and cere- antibodies associated with the anti-IgLON5 syndrome. J Neuroinflammation 2016; brospinal fluid changes. Front Neurol 2018;9:329. 13:226. Neurology.org/NN Neurology: Neuroimmunology & Neuroinflammation | Volume 6, Number 6 | November 2019 11