Enhanced influenza A H1N1 T cell epitope recognition and cross-reactivity to protein-O-mannosyltransferase 1 in Pandemrix-associated narcolepsy type 1

Narcolepsy type 1 (NT1) is a chronic neurological disorder having a strong association with HLA-DQB1*0602, thereby suggesting an immunological origin. Increased risk of NT1 has been reported among children or adolescents vaccinated with AS03 adjuvant-supplemented pandemic H1N1 influenza A vaccine, Pandemrix. Here we show that pediatric Pandemrix-associated NT1 patients have enhanced T-cell immunity against the viral epitopes, neuraminidase 175–189 (NA175–189) and nucleoprotein 214–228 (NP214–228), but also respond to a NA175–189-mimic, brain self-epitope, protein-O-mannosyltransferase 1 (POMT1675–689). A pathogenic role of influenza virus-specific T-cells and T-cell cross-reactivity in NT1 are supported by the up-regulation of IFN-γ, perforin 1 and granzyme B, and by the converging selection of T-cell receptor TRAV10/TRAJ17 and TRAV10/TRAJ24 clonotypes, in response to stimulation either with peptide NA175–189 or POMT1675–689. Moreover, anti-POMT1 serum autoantibodies are increased in Pandemrix-vaccinated children or adolescents. These results thus identify POMT1 as a potential autoantigen recognized by T- and B-cells in NT1.

Here we demonstrate enhanced T helper 1 (Th1)-cell/cytotoxic T-cell responses in Pandemrix-associated NT1 patients against distinct viral NA and NP peptide epitopes. Patient-derived T-cells also respond against a self-epitope in brain-expressed protein-Omannosyltransferase 1 (POMT1), that represents a peptide mimic of a T-cell epitope of NA strongly recognized in patients. Peripheral blood mononuclear cell (PBMC) gene expression profiles are closely matched, and TCR repertoire analyses detect upregulation of identical clonotypes in response to either NA or POMT1 peptide stimulation, thus supporting molecular mimicry and a pathogenic role of virus-directed, autoreactive T-cell clonotypes in NT1. Finally, vaccinees have increased antibody levels against human POMT1, suggesting that Pandemrix vaccination triggers POMT1 autoimmunity. These findings provide a link between influenza A (H1N1) virus-directed T-cell immunity and the development of autoimmunity in NT1.
Single T-cell peptides from HA, NA or NP that stimulated IFN-γ upregulation either in mice, or in HLA-DQB1*0602 positive individuals, were identified as antigens for subsequent T-cell assays comparing NT1 patients and controls (Supplementary Table 2).
Transcriptomic analyses of PBMC stimulated with influenza A H1N1 NA 175-189 or human POMT1 675-689 indicate activation of T helper 1-cell and cytotoxic T-cell responses with crossreactivity in Pandemrix-associated NT1. To further characterize T-cell responses against the epitopes from influenza A (H1N1) virus NA and human POMT1, and to identify putative pathogenic pathways in NT1, we studied the gene expression levels of viral NA 175-189 -or human POMT1 675-689 -stimulated PBMC from Pandemrix-associated NT1 patients and Pandemrixvaccinated healthy controls. In RNA sequencing, stimulation induced a clear response in NT1 patients versus healthy controls. Altogether 74 genes were differentially expressed, either in response to NA 175-189 (49 genes), or POMT1 675-689 (28 genes; overlap of 3 genes, **p < 0.01; Fig. 6, Supplementary Fig. 6, Supplementary Data 2). The results further demonstrated that the gene expression profiles of PBMC from NT1 patients responding to these two peptides were closely matched (Fig. 6), and that no genes were differentially expressed in PBMC from NT1 patients stimulated with either one of the peptides (p-level < 0.05). These findings confirmed the correlation analyses for IFN-g responses, and supported T-cell cross-reactivity between NA and POMT1. The list of 74 differentially expressed genes contained several genes closely associated with cytotoxic T-cell function, e.g., IFN-γ, chemokines/chemokine ligands known to be associated with IFN-γ, perforin 1, and granzyme B. A focused view onto T-cell effector molecules also showed that T-cells from individual NT1 patients responded similarly to stimulation either with viral NA 175-189 or human POMT1 675-689 , consistent with cross-reactivity of specific, influenza A (H1N1) virusdirected T-cells with the autoantigen POMT1 in NT1 (Fig. 7). This was exemplified most clearly by two strong responders, NT1 patients P003 and P015 (Figs. 6, 7), both characterized clinically by rapid onset of disease with the development of excessive daytime sleepiness and cataplexy within 4 weeks from vaccination with Pandemrix. Fig. 1 Overall experimental study design for T-cell epitope discovery in Pandemrix-associated NT1. As a first step, influenza A (H1N1) virus HA, NA, and NP peptide T cell recognition was tested with spleen cells from Pandemrix-vaccinated HLA-DQ6.2 mice, restimulated with pools of 5 peptides each (15mers). Pools that stimulated IFN-γ or IL-2 expression were broken up, and single peptides tested either with spleen cells from additional Pandemrixvaccinated HLA-DQ6.2 mice, or PBMC from Pandemrix-vaccinated HLA-DQB1*0602 positive individuals. As a second step, recognition of single peptides was then tested with PBMC from pediatric Pandemrix-associated NT1 patients, and healthy Pandemrix-vaccinated controls. As a third step, influenza A (H1N1) T cell peptides that showed increased stimulation of IFN-γ or IL-2 secretion in patients vs. controls were validated and mapped. Cross-reactive T cell self-epitopes were predicted by BLAST against human proteome, and recognition tested with PBMC from pediatric Pandemrix-associated NT1 patients, and healthy Pandemrix-vaccinated controls. In addition, RNA sequencing of FACS-sorted, NA 175-189stimulated CD4+ and CD8+ T-cell populations from a single pediatric, Pandemrix-associated NT1 patient (P003) revealed upregulation of IFN-γ, perforin 1, granzyme B, chemokine (C-C motif) ligand 4, and chemokine (C motif) ligand in CD8+ T-cells, while the gene expression profile of CD4+ T-cells from the same patient was characterized mainly by IFN-γ ( Fig. 6; FACS gating and sort strategy shown in Supplementary Fig. 7). Overall, gene expression in the NA 175-189 -stimulated CD8+ T-cell sample matched the profile obtained from NA 175-189 -stimulated PBMC, suggesting that peptide NA 175-189 might be presented by both HLA class I and II. This finding was also consistent with the notion that cytotoxic T-cells are a key effector cell population mediating immunity against influenza A virus 20 .   . We identified TRA clonotypes that were upregulated in ≥3 NT1 patient samples, stimulated either with NA 175-189 or POMT1 675-689 peptides (Fig. 8). Among these public TRA clonotypes, six clonotypes significantly increased in abundance both when stimulated with NA 175-189 , or POMT1 675-689 (proportion of false positives <0.05), further supporting T-cell cross-reactivity. Most strikingly, clonotypes CVVSAIKAAGNKLTF-TRAV10/ TRAJ17 and CVVSAMTTDSWGKFQF-TRAV10/TRAJ24 were upregulated in 3 of 8 (3 of 8) NA 175-189 -stimulated, and 5 of 10 (4 of 10) POMT1 675-689 -stimulated samples. In NT1 patients P003 and P015, characterized by T helper 1 (Th1)-cell/cytotoxic T-cell gene expression signatures in response to stimulation with NA 175-189 or POMT1 675-689 , both clonotypes were upregulated in response to either of the two peptides. Three additional clonotypes were found, significantly more abundant after stimulation with at least one of the two peptides, and with high similarity to either of the two clonotypes above: CVVITIKAAGNKLTF-TRAV10/TRAJ17, significantly upregulated in POMT1 675-689stimulated samples, including P003; CVVSGMTTDSWGKLQF-TRAV10/TRAJ24, upregulated in NA 175-189 -stimulated samples, including P015; and CVVSGMTTDSWGKFQF-TRAV10/ TRAJ24, upregulated in both NA 175-189 -and POMT1 675-689 -stimulated samples, including both P003 and P015. These results demonstrated converging selection of NA 175-189 -and/or POMT1 675-689 -reactive T-cell clones using TRAV10/TRAJ17 or TRAV10/TRAJ24 gene segments in Pandemrix-associated NT1 patients.
A complete list of both public and private clonotypes using TRAV10/TRAJ17 and TRAV10/TRAJ24, identified in NT1 patients or healthy controls, is provided in Fig. 9a, b. The results demonstrate that in addition to the identified public clonotypes using TRAV10/TRAJ17 (CVVSAIKAAGNKLTF, CVVITI-KAAGNKLTF) and TRAV10/TRAJ24 (CVVSAMTTDSW GKFQF, CVVSGMTTDSWGKLQF, CVVSGMTTDSWGKFQF), several highly similar clonotypes were present in peptidestimulated samples from various NT1 patients, including CVVSMIKAAGNKLTF-TRAV10/TRAJ17, CVVSSMTTDSWG KFQF-TRAV10/TRAJ24, CVVTVMTTDSWGKFQF-TRAV10/ TRAJ24, and CVVSSLTTDSWGKFQF-TRAV10/TRAJ24. The identified public clonotypes were not exclusive to NT1 patients, but occasionally seen in healthy controls, too. In particular, the HLA-DQB1*0602-negative healthy control HC27 shared several of the clonotypes upregulated in NT1 patient samples stimulated with NA 175-189 or POMT1 675-689 . However, in contrast to NT1 patients, HC27 did not show an increase in expression of inflammatory cytokines/ chemokines in response to peptide stimulation (compare Fig. 4a, b; 5a; 7). Therefore, these TRAV10-TRAJ17 and TRAV10-TRAJ24 clonotypes might have contributed to an inflammatory reaction in NT1 patients, but their functional phenotype could be effectively regulated when present in healthy controls.
Discovery of POMT1 autoantibodies in Pandemrix-vaccinated children. Finally, we studied levels of plasma antibodies to POMT1 by radioimmunoassay (RIA), a sensitive and specific liquid-phase method for the detection of autoantibodies recognizing conformational epitopes 21 . Autoantibodies to POMT1 were increased in Pandemrix-vaccinated children (both in NT1 patients and healthy controls) in comparison to unvaccinated control children, suggesting that Pandemrix induced autoimmunity to POMT1 (****p < 0.0001, Fig. 10; Table 1).

Discussion
Genetic studies strongly implicate HLA-DQ6.2-restricted CD4+ T-cells in the pathogenesis of NT1 4,5 . A central role for CD4+ Tcells is seen in many autoimmune disorders, e.g., in celiac disease and type 1 diabetes. Both disorders show exceptionally strong HLA class II associations 22 , to be surpassed still in NT1. HLA class II-associated autoimmune diseases are considered antigen driven (e.g., by wheat gliadin in celiac disease), although tissue pathology is not mediated by CD4+ T-cells only. Other immune cells are involved in tissue destruction, and antigen spreading leads to broad autoimmunity. In Pandemrix-associated NT1, we hypothesized that CD4+ T-cells primed by influenza A (H1N1) virus epitopes are disease drivers, cross-react with CNS autoantigens, and orchestrate an immune-mediated attack against the hypocretin neuronal network. Therefore, we focused our study on influenza A (H1N1) T-cell epitopes that are recognized in the context of HLA-DQ6.2, although HLA-DQ6.2 peptide binding was not directly tested. Due to limitations in the availability of blood samples from pediatric, Pandemrix-associated NT1 patients, we started to search for dominant T-cell epitopes using Pandemrix-primed mice that expressed HLA-DQ6.2 as the only MHC class II allele, and Pandemrix-vaccinated, HLA-DQB1*0602-positive individuals. Choosing this strategy, the selected T-cell peptides were enriched for HLA-DQ6.2-binding, but likely included both MHC class II-and/or class I-binding peptides.
We found that T-cell responses against two NA-and NPderived epitopes were enhanced in children and adolescents with Pandemrix-associated NT1 in comparison to Pandemrixvaccinated controls. Further, our study provided evidence for influenza A (H1N1) virus-directed T-and B-cell cross-reactivity with human POMT1. Homology of a viral NA peptide sequence and a human POMT1 sequence was suggested by basic local alignment. We could show that IFN-γ secretion in response to NA 175-189 /NA 178-192 -and POMT1 675-689 -stimulation was correlated, and that T cell cytokine/chemokine gene expression patterns were closely matched. This was exemplified by two strong responders, patients P003 and P015, who both developed clinical narcolepsy with cataplexy within one month from Pandemrix vaccination. TCR repertoire analyses in NT1 patients revealed increased abundance of seven public clonotypes in response to stimulation both with NA 175-189 and POMT1 675-689 in NT1 patients. In addition, vaccination with Pandemrix induced anti-POMT1 autoantibodies in children. These results were consistent with cross-reactivity of specific, virus-directed T-cells and antibodies with the autoantigen POMT1 in Pandemrixassociated NT1.
The reasons remain unknown for the strongly enhanced T-cell reactivity to different influenza A (H1N1) protein fragments demonstrated in this study in patients who developed NT1 after Pandemrix vaccination. Our earlier study showed that viral proteins contained in the Pandemrix antigen suspension formed complexes, which may have contributed to the enhanced reactivity to both NA-and NP-derived epitopes 16 . Alternatively, and not mutually exclusive, enhanced T-and B-cell reactivity in NT1 patients may be attributed to immunogenetic predispositions [6][7][8]23 , resulting in hyper-responsiveness to proinflammatory signals and AS03 adjuvant-supplemented influenza A (H1N1) priming.
Specific HLA class I alleles, most consistently HLA-A*11:01, account for additional genetic risk contributions in NT1, independent from HLA-DQB1*06:02 24,25 . Cytotoxic CD8+ T cells, likely acting in concert with CD4+ Th1-cells, are frequently implicated in the pathogenesis of autoimmune diseases, including the neurological disorders multiple sclerosis 26 and NT1 25 . In this study of Pandemrix-associated NT1, cytokine and chemokine gene expression by PBMC in response to stimulation with influenza A (H1N1) NA 175-189 or human POMT1 675-689 matched closely, and revealed a cytotoxic T cell signature, characterized by IFN-γ, IFN-γ−associated genes, perforin 1, and granzyme B. This phenotype was consistent with CD8+ T-cells, but also unconventional, cytotoxic CD4+ T-cells, previously observed in influenza A infection 20 . In FACS-sorted T cell fractions, available for study from one Pandemrix-associated NT1 patient only, perforin 1 and granzyme B expression was upregulated in NA 175-189 -stimulated CD8+ T-cells, but not in CD4+ T cells. Thus, NA 175-189 peptide is most likely binding to both HLA class I and II molecules.
Among the identified T-cell epitopes, IAYERMCNILKG (NP 217-228 ; Immune Epitope Database ID 145824), derived from the NYMC X-157 vaccine strain, is shared between different influenza A virus strains. Using HLA-DQ6.2 tetramers, the study by Luo et al. demonstrated the presence of CD4+ T-cells specific for C-terminally amidated, human HCRT 54-66 and HCRT 86-97 peptides, but not for native HCRT peptides, in NT1 patients and controls 18 . Furthermore, TCR sequencing analyses suggested cross-reactivity with HA. The recent study by Jiang et al., also using HLA-DQ6.2 tetramers, detected CD4+ T cells specific for native HCRT 56-69 and HCRT 87-100 peptides in NT1 patients and controls 28 . In addition, Latorre et al. reported increased frequencies of predominantly HLA-DR-restricted CD4+ T cell clones recognizing much more diverse peptide epitopes of HCRT in NT1 patients. These clones did not show reactivity with HA. Notably, T-cell reactivity to HCRT was also detected in type 2 narcolepsy, which is not associated with HCRT-deficiency 30 . Thus, the current literature on T cell reactivity to HCRT in NT1 remains conflicting. In many autoimmune diseases, autoimmunity is directed against multiple autoantigens expressed in the target tissue. The results of our study do not exclude a role of autoimmunity to HCRT, but identify with POMT1 another autoantigen in NT1.
POMT1 is a glycosyltransferase anchored in the endoplasmic reticulum, catalyzing in complex with POMT2 the first step of Omannosyl glycan synthesis. In humans, O-mannose-linked glycosylation is observed mainly in brain, peripheral nerve, and muscle glycoproteins. The best-known substrate of POMT1/ POMT2 is α-dystroglycan, mediating the binding of extracellular matrix components such as laminin to the dystrophin complex. Congenital defects in POMT1 are a cause of muscular dystrophydystroglycanopathy types A1, B1, and C1 31 . Some forms of muscular dystrophies, most clearly myotonic dystrophy type 1, are associated with rapid eye movement (REM) sleep dysregulation 32 . It is currently unknown whether POMT1 or its substrates are involved in neuronal sleep regulation, but this possibility could be explored in experimental studies. Autoantigen discovery in Pandemrix-associated NT1   In this study, T-cell reactivity with POMT1 was detected in Pandemrix-associated NT1 patients. Samples from unvaccinated NT1 patients were not available for comparisons. Currently, it remains unknown whether T cell reactivity against POMT1 can be observed in NT1 patients not exposed to influenza A (H1N1) vaccination or infection. Other questions to be addressed by future studies include the identities of the HLA alleles involved in the restriction of the dominant public TRAV10-TRAJ17, TRAV10-TRAJ24, and TRBV6-1/TRBJ2-7 clonotypes described, their function, antigenic peptide specificity and cross-reactivity, and their value for NT1 diagnosis.
In summary, this study provides evidence for an influenza A (H1N1) virus directed, POMT1 cross-reactive T-cell response in NT1, in the context of HLA-DQ6.2. We identified two dominant T-cell epitopes of influenza A (H1N1) NA and (NYMC X-157 vaccine strain) NP virus proteins in Pandemrix-associated NT1 patients. Patients mounted a vigorous Th1-cell/cytotoxic T-cell response against these epitopes and showed T-cell reactivity against a self-epitope in POMT1, an enzyme expressed in brain. The POMT1 epitope is a peptide mimic of a dominant T-cell epitope of influenza A (H1N1) NA in Pandemrix-associated NT1. IFN-γ, perforin 1, granzyme B, and chemokine gene expression in response to stimulation with T-cell peptides NA 175-189 or POMT1 675-689 was closely matched in individual NT1 patients. TCR repertoire analyses demonstrated converging selection of TRAV10/TRAJ17 and TRAV10/TRAJ24 clonotypes in NT1, upregulated in response to either NA 175-189 -or POMT1 675-689stimulation, thus supporting T-cell cross-reactivity and a pathogenic role of these clonotypes in NT1. Moreover, vaccinees had elevated antibody levels against human POMT1, suggesting that Pandemrix vaccination triggered an autoimmune response to POMT1. The results of this unbiased search identify POMT1 as an autoantigen recognized by T and B cells in NT1. Mice. Homozygous HLA-DQA1*0102, -DQB1*0602, human CD4-transgenic Ab0 NOD mice ("HLA-DQ6.2 mice"; stock number 006023) were purchased from Jackson Laboratories, Bar Harbor, ME, USA, and bred at the University of Helsinki Laboratory Animal Centre 17 . All mice were housed under specific pathogen-free conditions, under a 14-10 h light-dark cycle, at 22 ± 2°C and 50-60% humidity. Pandemrix vaccine was injected under isoflurane anesthesia into the right thigh Fig. 7 T cell-related genes differentially expressed in peptide-stimulated PBMC from Pandemrix-associated NT1 patients (RNA-seq). Dot plots demonstrating the expression levels (normalized read counts) of nine T cell-related genes showing significant differences in expression (**p < 0.01) between NT1 patients and controls either after stimulation with NA 175-189 or POMT1 675-689 peptides (compare Fig. 6). Differentially expressed genes were identified using edgeR, based on a test analogous to Fisher's exact test. The paired method was used for comparisons, calculating 2-sided p-values and adjusting for multiple testing using BH correction. The lines indicate paired data of two strong responders, NT1 patients P003 and P015. muscle (50 μl emulsion volume; male or female mice; 3 months of age), followed by one booster injection after 14 days. After 6-9 weeks from the first injection, mice were exsanguinated by retro-orbital bleeding under ketamine/xylazine anesthesia (Intervet International, Boxmeer, The Netherlands; Bayer HealthCare, Leverkusen, Germany), and spleens were harvested for collection of cells.  Mouse spleen cell stimulation. Spleen cells from 1 to 2 individual mice were pooled, and seeded in triplicates in 96-well plates at 2 × 10 5 cells/well in RPMI 1640 medium (200 μl volume) containing heat-inactivated fetal calf serum (10%), penicillin/streptomycin, glutamine, and HEPES (25 mM), at 37°C and 5% CO 2 (all from Sigma-Aldrich). Cells were stimulated for 6 days with 15-mer peptides (10 μM; resulting in DMSO concentration 0.1%), a combination of anti-CD3 and anti-CD28 antibodies (positive control, 3 μg and 2 μg/ml; clones 145-2C11 and 37.51; eBioscience, San Diego, CA), endotoxin-free ovalbumin (negative control), recombinant influenza A hemagglutinin, neuraminidase or nucleoprotein (all 10 μg/ml). For each stimulation, two plates were run in parallel (from two different spleen cell pools).
Patient and healthy control donor material. The T cell study included 28 pediatric Pandemrix-associated NT1 patients (20 in the discovery cohort, and 14 in the validation cohort; 6 patients were included in both cohorts), and 33 healthy Pandemrix-vaccinated control children or adolescents (Table 1, Supplementary  Table 3A, B). If results from different experiments were pooled for analysis, and two results from a single patient were available, only the result obtained with the sample drawn earlier was used. PBMC from Pandemrix-vaccinated HLA-DQB1*0602-positive individuals (sleep clinic patients without a diagnosis of NT1) were also used for influenza A (H1N1) virus T cell epitope screening (Supplementary Table 1). The plasma study included 37 pediatric Pandemrix-associated NT1 patients, 57 healthy Pandemrix-vaccinated control children or adolescents, and 130 healthy control children or adolescents that had not received Pandemrix vaccine (Table 1). NT1 patients were diagnosed in outpatient clinics at Finnish university hospitals, by pediatric neurologists, pediatricians or neurologists with expertise in sleep medicine. All NT1 patients were diagnosed based on criteria defined in the third edition of the International Classification of Sleep Disorders 33 . All NT1 patients suffered from excessive daytime sleepiness. They also had abnormal Multiple Sleep Latency Tests in polysomnography (sleep latency <8 min and ≥2 sleep onset REM periods), and either cataplexy, or cerebrospinal fluid HCRT 1 levels below 110 pg/ml, as measured using the standardized Phoenix RIA method with Stanford reference.
Human plasma and peripheral blood mononuclear cell (PBMC) separation and storage. Whole blood was drawn and heparinized. Plasma was separated from fresh heparinized blood samples, and stored at −70°C, before isolation of PBMC by Ficoll (GE Healthcare, Uppsala, Sweden) isogradient centrifugation. PBMC were stored in liquid nitrogen until further analysis.
Genotyping of the HLA-DQB1*0602 allele. Genotyping for the HLA-DQB1 locus was performed using a homogeneous assay based on asymmetrical PCR and hybridization with sequence-specific lanthanide labeled oligonucleotide probes, as described earlier 34 . In short, blood dried on sample collection cards was used for asymmetric PCR where the sequence of interest was simultaneously amplified and detected using lanthanide labeled probes with locked nucleic acid additions. Complementary quencher oligonucleotides were also included in the reactions. Two different specific probes, one labeled with europium and another with terbium, were present in each reaction and allowed simultaneous use of two different probes and their individual detection by two signals measured after the amplification. Presence of DQB1*0602 allele was deduced based on a positive signal with DQB1*0602/ *0603 specific probe and a negative with DQB1*0603/*0604 specific probe. All oligonucleotides used in this study are defined in Supplementary Table 4.
Human PBMC stimulation. Frozen PBMCs were thawed and allowed to recover in RPMI 1640 culture medium (Life Technologies, Paisley, United Kingdom) containing 2 mmol L-glutamine, 25 mmol/l HEPES, 25 μg/ml gentamycin (Sigma-Aldrich) and 10% of heat-inactivated AB serum (Innovative Research, Novi, MI, USA), for 1 h in an incubator at 37°C and 5% CO 2 . Cells were seeded in duplicates or triplicates in 96-well round-bottom plates (Nunc, Roskilde, Denmark) at 2 × 10 5 cells/well, using RPMI 1640 culture medium supplemented with 5% of heatinactivated AB serum, 2 mmol L-glutamine, 25 mmol/lHEPES and 25 μg/ml gentamycin (200 µl/well), and stimulated for 6 days with 15-mer peptides (10 μM; resulting DMSO concentration 0.1%), tetanus toxoid (20 μg/ml), PBS (negative control), recombinant influenza A hemagglutinin, neuraminidase or nucleoprotein (all at 10 μg/ml). Generally, samples were stimulated with different T cell peptides selected at random. Peptide testing was prioritized, and proteins were only tested if enough cells were available. For epitope mapping, peptides were not selected at random. Instead, more samples were used for testing the inner three than the outer two peptide epitopes. No patient or control samples that had passed viability control (trypan blue staining, positive response to stimulation with tetanus control) were excluded from the study.
FACS sorting. To study gene expression in CD4+ and CD8+ T cells, PBMCs were stimulated for 6 days with 15-mer peptides, as described in the previous section. Stimulated cells were stained with anti-CD3-Efluor 610 (BioLegend), anti-CD4-FITC (BD Pharmingen), anti-CD8-AF700 (eBioscience) and SYTOX ® Blue (Thermo Fisher Scientific). A BD Influx flow cytometer (BD Biosciences) was used to sort CD4+ and CD8+ T cells. Dead cells and monocytes were excluded from analysis by appropriate gating strategies and SYTOX ® Blue staining (gating and sort strategy shown in Supplementary Fig. 7).
Mouse cytokine RT-qPCR analysis. Total RNA from stimulated murine cells was isolated with the RNeasy Mini kit (Qiagen, Hilden, Germany), and cDNA was synthesized using a high-capacity cDNA reverse transcription kit (Applied Biosystems, Foster City, CA, USA). TaqMan   and with the final extension step of 5 min at 72°C. An aliquot of this PCR product was used for TCR sequencing (below). The PCR products were pooled together in sets of 12 samples containing different indexing oligos and purified with 0.6× HighPrep PCR reagent (MAGBIO, Gaithersburg, USA) according to the manufacturer's instructions. They were eluted in 10 μL of molecular grade water. The 3′-end complementary DNA fragments for sequencing were prepared using the Nextera XT (Illumina) tagmentation reaction. The reaction was performed according to the manufacturer's instructions, except for the use of P5 SMART primer, which replaced the S5xx Nextera primer. Each set of 12 samples that was pooled after the PCR reaction was tagmented with a different Nextera N7xx index. Subsequently, the samples were PCR amplified as follows: 95°C for 30 seconds, 11 cycles of 95°C for 10 seconds, 55°C for 30 seconds, and 72°C for 30 seconds, with the final extension step of 5 minutes at 72°C . Samples were purified twice using 0.6× and 1.0× HighPrep PCR reagent and eluted in 10 μL of molecular-grade water. The concentration of the libraries was measured by using a Qubit 2 fluorometer (Invitrogen) and the Qubit DNA HS Assay Kit (Thermo Fisher Scientific). The quality of the sequencing libraries was assessed using the LabChip GXII Touch HT electrophoresis system (PerkinElmer), with the DNA High Sensitivity Assay (PerkinElmer) and the DNA 5K/RNA/ Charge Variant Assay LabChip (PerkinElmer). Samples were stored at −20°C. The libraries were sequenced on an Illumina NextSeq 500 instrument, with a custom primer (DS Custom Read 1) producing read 1 of 20 base pairs (bp), and read 2 (paired end) of 50 bp. Sequencing was performed at the Functional Genomics Unit of the University of Helsinki, Finland.
Read alignment and generation of digital expression data. Raw sequence data were processed as described before 37 . The pipeline originally suggested by Macosko et al. for processing of Drop-seq data was used 36 . Briefly, reads were filtered to remove polyA tails of length 6 or greater, and then aligned to the human (GRCh38) genome using STAR aligner v2.7.2 38 , with default settings. Uniquely mapped reads were grouped according to the 1-12 barcode, and gene transcripts were counted by their unique molecular identifiers (UMIs) to reduce the bias emerging from the PCR amplification. Digital expression matrices reported the number of transcripts per gene in each sample (according to the distinct unique molecular identifiers sequences counted). Differentially expressed genes were identified using edgeR v3.4.0, based on a test analogous to Fisher's exact test 39 . The paired method was used for comparisons, calculating 2-sided p-values and adjusting for multiple testing using BH correction. RNAseq heatmaps were hierarchically clustered and drawn using the web-based Heatmapper software (accessed 25th of July 2020) 40 .
TCR sequencing. Primers were designed for human TCR constant genes using Primer3 41,42 , and appended to a tailing sequence to produce the reverse primers N7tail-TRAC-2 and N7tail-TRBC-2 (Supplementary Table 4). Forward primers (DS_N5xx_5TSO) contained a sequence complementary to the Template Switch Oligo and an index sequence (I5). TCR sequences were amplified using KAPA HiFi HotStart Readymix, 0.3 μM of each primer and 1 μl pre-amplified cDNA library (see RNA Sequencing) in 10 μl volume with 16 cycles of 95°C for 20 s, 62°C for 20 s and 72°C for 30 s, followed by purification with 1.5× volume of HighPrep PCR reagent. The TCR libraries were dual indexed using the same forward primer as in the previous step, and one of Nextera N7xx primers. PCR was done using KAPA HiFi HotStart Readymix, 0.3 μM of each primer and 3 μl template in 10 μl volume with 10 cycles of 95°C for 20 s, 62°C for 20 seconds and 72°C for 30 s. The libraries were pooled according to concentration and purified first with 1× and then with 1.5× volume of HighPrep PCR reagent. The quality of the libraries was assessed with a LabChip GXII Touch HT electrophoresis instrument and quantitative PCR. Libraries were sequenced with a MiSeq Reagent Kit v3 (600-cycle) using the 300 + 300 bp paired-end protocol, a custom read 1 primer (DS-Cus-tomRead1_5TSO) and 10% PhiX on an Illumina Miseq instrument at Department of Virology, University of Helsinki, Finland.
TCR analysis. TCR sequencing reads were mapped and clonotypes assembled with MiXCR v3.0.12 43,44 . The NA 175-189 -stimulated sample from HC19 did not produce any clonotypes and was excluded. Low-count clones that had an identical CDR3 nucleotide sequence to a high-count clone (over 500-fold difference) were filtered out as probable cross-sample contaminants. Non-productive clones were also filtered out. To normalize the data, counts were downsampled pairwise using VDJtools v1.2.1 45 so that each NA 175-189 -treated and medium-treated sample from the same patient was downsampled to the same total count. The same was done separately for the POMT1 675-689 -treated samples. Further processing was done using R statistical software v3.6.3 (R Foundation for Statistical Computing, Vienna, Austria). Gene usage differences between peptide and media-treated patient samples were examined using the Immunarch package v0.5.5 (Zenodo/CERN, Geneva, Switzerland) and Wilcoxon test with the paired method. P values were corrected for multiple testing with the Benjamini-Hochberg method. For clone-level analysis, low-count clones were filtered out by requiring each clone to have count >10 in at least one sample. One pseudocount was added, the counts were log2 transformed and fold change between the peptide-treated sample and medium control from the same individual was calculated for public clones (present in at least three samples). Statistical significance was tested with Rank Products test implemented in the RankProd package v3.12.0 46,47 , using the paired method. Clones were hierarchically clustered based on the Levenshtein edit distance of their CDR3 sequences.
Plasma POMT1 antibody radioimmunoassay. The full-length human POMT1 cDNA (GenBank entry BC065268.1) was cloned into the pTNT vector (GenScript, Piscataway, NJ, USA) under the control of the SP6 promoter, and multiplied in Invitrogen subcloning efficiency DH5α competent cells (ThermoFisher, Waltham, MA, USA), according to the manufacturer's instructions. 35 S-methionine-labeled POMT1 was produced by in vitro transcription and translation of the purified plasmids using the TNT Coupled Reticulocyte Lysate System according to the manufacturer's instructions (Promega, Madison, WI, USA) in the presence of 35 Smethionine (Perkin Elmer, Waltham, MA, USA). Unincorporated 35 S-methionine was removed by gel chromatography on NAP-5 columns (GE Healthcare, Chicago, IL, USA). Plasma samples were analyzed for POMT1 (auto)-antibodies by RIA, as described for GAD autoantibodies 21 . Briefly, 2 µl of plasma was incubated overnight at 4°C in 96-well plates with 20,000 cpm of 35 S-methionine-labeled POMT1, diluted in 50 µl of TBST buffer (50 mM Tris, 150 mM NaCl, 0.1% Tween-20, pH 7.4). Antibody-antigen complexes were precipitated with protein A-Sepharose, and unbound label was removed by washing several times with TBST buffer (50 mM Tris, 150 mM NaCl, 0.1% Tween-20, pH 7.4). The bound radioactivity was measured with a liquid scintillation counter (1450 MicroBeta Trilux, Perkin Elmer Life Sciences, Turku, Finland), and the results were expressed in relative units (RU), based on a serial dilution in-house standard curve.
Statistics. Statistical comparisons between groups were performed, using Prism version 6.05 (Graph Pad Software, San Diego, CA), and Kruskal-Wallis and Dunn's multiple comparisons tests. Correlation analyses were performed by calculating Spearman's rank correlation coefficient. Additional details on the statistics used in the analysis of RNA and TCR sequencing data are included in the sections above.
Study approval. NT1 patients were Finnish participants in the NARPANORD narcolepsy study. Heathy control children and adolescents were recruited via the Finnish Diabetes Registry. Pandemrix-vaccinated HLA-DQB1*0602-positive individuals without a diagnosis of NT1 were recruited at the Helsinki Sleep Clinic, Vitalmed Research Center, Helsinki, Finland. All participants or their guardians had given written, informed consent. Ethical permissions for clinical studies were obtained from the Ethics Committee of the Hospital District of Helsinki and Uusimaa, Finland. All animal procedures were approved by the Board for Animal Research (ELLA), Southern Finnish State Administrative Agency (ESAVI/1064/04.10.03/2012).
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
The sequencing data have been deposited to the EMBL-EBI European Genome-phenome Archive (EGA) under the accession number EGAS00001004886 (RNA sequencing data, Fig. 6; TCR sequencing data, Fig. 7). Upon reasonable request and subject to a material transfer agreement, data submitted to EGA will be made available by the corresponding data access committee. Source data are provided with this paper.