Absence of anti-hypocretin receptor 2 autoantibodies in post pandemrix narcolepsy cases

Background A recent publication suggested molecular mimicry of a nucleoprotein (NP) sequence from A/Puerto Rico/8/1934 (PR8) strain, the backbone used in the construction of the reassortant strain X-179A that was used in Pandemrix® vaccine, and reported on anti-hypocretin (HCRT) receptor 2 (anti-HCRTR2) autoantibodies in narcolepsy, mostly in post Pandemrix® narcolepsy cases (17 of 20 sera). In this study, we re-examined this hypothesis through mass spectrometry (MS) characterization of Pandemrix®, and two other pandemic H1N1 (pH1N1)-2009 vaccines, Arepanrix® and Focetria®, and analyzed anti-HCRTR2 autoantibodies in narcolepsy patients and controls using three independent strategies. Methods MS characterization of Pandemrix® (2 batches), Arepanrix® (4 batches) and Focetria® (1 batch) was conducted with mapping of NP 116I or 116M spectrogram. Two sets of narcolepsy cases and controls were used: 40 post Pandemrix® narcolepsy (PP-N) cases and 18 age-matched post Pandemrix® controls (PP-C), and 48 recent (≤6 months) early onset narcolepsy (EO-N) cases and 70 age-matched other controls (O-C). Anti-HCRTR2 autoantibodies were detected using three strategies: (1) Human embryonic kidney (HEK) 293T cells with transient expression of HCRTR2 were stained with human sera and then analyzed by flow cytometer; (2) In vitro translation of [35S]-radiolabelled HCRTR2 was incubated with human sera and immune complexes of autoantibody and [35S]-radiolabelled HCRTR2 were quantified using a radioligand-binding assay; (3) Optical density (OD) at 450 nm (OD450) of human serum immunoglobulin G (IgG) binding to HCRTR2 stably expressed in Chinese hamster ovary (CHO)-K1 cell line was measured using an in-cell enzyme-linked immunosorbent assay (ELISA). Results NP 116M mutations were predominantly present in all batches of Pandemrix®, Arepanrix® and Focetria®. The wild-type NP109-123 (ILYDKEEIRRIWRQA), a mimic to HCRTR234-45 (YDDEEFLRYLWR), was not found to bind to DQ0602. Three or four subjects were found positive for anti-HCRTR2 autoantibodies using two strategies or the third one, respectively. None of the post Pandemrix® narcolepsy cases (0 of 40 sera) was found positive with all three strategies. Conclusion Anti-HCRTR2 autoantibody is not a significant biological feature of narcolepsy or of post Pandemrix® autoimmune responses.


Introduction
Although there is strong genetic evidence suggesting autoimmunity in type 1 narcolepsy [1][2][3], notably a 97% association with human leukocyte antigen (HLA) class II alleles DQB1 Ã 06:02 and DQA1 Ã 01:02 that together form the heterodimer DQ0602 [4], the nature of the autoimmune antigen leading to hypocretin cell death has been elusive. Renewed interest in this area came from the observation of increased incidence of type 1 narcolepsy following the 2009-2010 winter vaccination campaign against the pandemic H1N1 (pH1N1) "swine" flu (pH1N1-2009) that emerged in 2009 [5]. This effect was mostly obvious in Scandinavia, where vaccine coverage with a specific adjuvant system 03 (AS03) (composed of α-tocopherol, squalene and polysorbate 80) adjuvanted vaccine Pandemrix 1 from GlaxoSmithKline (GSK) was high (~50%), and rapid onset narcolepsy was frequently observed in children within a few months following vaccination [5][6][7][8][9][10]. In Finland, where incidence of narcolepsy was best estimated and most consistent with known population prevalence (0.02%) [11], risk of Pandemrix 1 induced narcolepsy increased~10 fold, although absolute risk of developing narcolepsy still remained low (less than 10 for 100,000 vaccinees versus (vs) 0.7 for 100,000 baseline incidences) [12]. Others also found that the onset of narcolepsy was frequently seasonal [13], occurring in spring or summer, and following winter upper-airway infections such as Streptococcus Pyogenes [14,15]. Observation of seasonality of onset was more obvious in young children where narcolepsy onset is very abrupt and can more easily be timed to the exact month. All together, these data strongly suggest that narcolepsy is an autoimmune disease triggered by upper airway infections, notably influenza and maybe Streptococcus Pyogenes, with stronger susceptibility to pH1N1-2009. Additional surprising or inconsistent observations were made thereafter. First, whereas in China, a clear increase in childhood narcolepsy cases was observed in ecological studies of two sleep centers during the winter of 2009-2010 [13,16,17], a weaker effect was observed in Germany [18] and in the United States [19,20]. Second, increased onset of narcolepsy independent of vaccination was not observed in Scandinavia [21], although this could be due to herd immunity protection against pH1N1 through the high coverage vaccination campaign. Third, Pandemrix 1 vaccination-induced narcolepsy seemed to have occurred more frequently in some areas of Sweden, notably in the south of the country, possibly reflecting differential timing of the infection vs the vaccination campaign, or possible differences between vaccine batches [22,23]. Fourth, other pH1N1 vaccines were not associated with increased risk of developing narcolepsy, notably in the United States, where only sporadic cases have been reported after flu vaccinations. As only unadjuvanted or live attenuated vaccines were used during the 2009-2010 vaccination campaign in the United States, this suggests that adjuvant AS03 played a role in Pandemrix 1 vaccination-induced narcolepsy. Surprisingly, however, two other adjuvanted pH1N1 vaccines used in countries other than Scandinavia had little or no effects: one is MF59C.1 (containing squalene, polysorbate 80 and sorbitan trioleate) adjuvanted Focetria 1 from Novartis used in Europe (mostly in Italy) [24], the other is Arepanrix 1 used in Canada and South America [25], another AS03 adjuvanted vaccine from GSK that is almost identical to Pandemrix 1 except for the method of antigen isolation.
Risk of narcolepsy remained low even with Pandemrix 1 vaccination in 2009-2010 and DQ0602 positive. Indeed, narcolepsy occurred in vaccinated individuals with an incidence of less than 1/3,000 persons after Pandemrix 1 even in DQ0602 positive individuals. The reason for that narcolepsy incidence rate following pH1N1 vaccination varied across vaccines is likely multifactorial, and may involve vaccine composition differences (adjuvant and antigen isolation protocol), prior immune history and genetics of targeted populations. It is notable that some targeted subgroups were prioritized in each vaccination campaign (children vs adults, high risk groups) and that timing of vaccination in relation to the unfolding pandemic flu occurred in close temporal successions, usually within a few weeks in many countries. Nonetheless, one of the most reliable findings may be that Pandemrix 1 had strong effects in most European countries [26][27][28][29][30][31], and that Arepanrix 1 , a closely related vaccine had less or no effects in Quebec and Canada, suggesting that differences in vaccine composition are involved, as suggested by Ahmed et al. [32].
Following on these observations, several authors went on to compare the composition of known pH1N1 monovalent vaccines, with primary focus on Pandemrix 1 and Arepanrix 1 , the two most closely related vaccines with differential risks [33]. The creation of flu vaccine is a long process that involves growing genetically engineered flu strains in Gallus gallus (chicken) eggs, and purifying vaccine particles and antigens for vaccine preparation, with primary focus on Hemagglutinin (HA) and Neuraminidase (NA), two surface proteins that are strongly involved in lymphocyte B cell antibody (Ab) responses to the flu [34][35][36][37]. The pH1N1-like vaccine strains X-179A and X-181 were created in New York Medical College (NYMC) for the 2009-2010 swine flu campaign, using an old H1N1-1918-like strain, A/ Puerto Rico/8/1934 (PR8), as a backbone, and reassorted with A/California/07/2009 [38]. In both X-179A and X-181 strains, HA, NA and polymerase (basic) protein 1 (PB1) were derived from A /California/07/2009, while other proteins were PR8-derived [39,40]. NYMC X-181, characterized by a mutation of HA 146N to 146D and a few other differences (including HA titer, yield of purified viral protein) [41], was created in late 2009 and selected because it produced higher reassortant HA titers than X-179A. It was used toward the end of the season (2009) in some cases [38][39][40]. Both Pandemrix 1 and Arepanrix 1 were produced by GSK using X-179A, while Focetria 1 was created by Novartis using X-181, with their own patented processes. Jacob et al. analyzed the compositions of Pandemrix 1 (2009 batches) and Arepanrix 1 (a 2010 batch) using MS and showed results obtained with five main viral proteins: HA1 and HA2 subunits of HA, NA, NP, matrix protein 1 (M1), and non-viral proteins from chicken growth matrix [33]. Overall, these vaccines were remarkably similar. Pandemrix 1 contained slightly more NP and NA, while Arepanrix 1 displayed a larger diversity of viral and chicken proteins, with the exception of five chicken proteins that were relatively more abundant in Pandemrix 1 [33]. Interestingly, HA1 146N (HA 146N) (amino acid residue 129N in the mature protein) was found to be mutated to 146D in most of the HA antigens of Arepanrix 1 [33], a surprising finding as both Pandemrix 1 (2009 and 2010 batches) and Arepanrix 1 (a 2010 batch) were manufactured using X-179A. This last finding suggested that the HA 146N to 146D mutation emerged during Arepanrix 1 manufacturing prior to 2010, and raised the possibility that this mutation could have played a role in differential vaccine susceptibility. The main limitation of this study was the use of a 2010 batch of Arepanrix 1 , which was produced one year later than the ones used during the 2009-2010 vaccination campaign [33].
In parallel with Jacob et al. findings [33], Vaarala et al. studied post Pandemrix 1 patient and control antibody reactivity against Pandemrix 1 and Arepanrix 1 , and found that Arepanrix 1 reacted poorly with antibodies of post Pandemrix 1 vaccinated children, suggesting antigenic differences in antibody determinants [42]. Increased antibody levels to HA and NP, particularly to structurally altered viral NP, were also seen in post Pandemrix 1 narcoleptic children [42]. The authors went on to hypothesize that the higher amount of denatured NP in Pandemrix 1 during the purification process of viral proteins may explain increased narcolepsy susceptibility with Pandemrix 1 vaccination.
Following on these observations, Ahmed et al. identified a sequence of NP 111-121 (YDKEEIRRIWR) with NP 116I (underlined) in X-179A with significant homology to a sequence of the first extracellular domain of HCRTR2 (HCRTR2 [34][35][36][37][38][39][40][41][42][43][44][45] , YDDEEFLRYLWR), whereas a mutation of NP 116I to 116M in NP 111-121 (YDKEEMRRIWR, 116M was underlined) was found in Focetria 1 , which in any case contained only very limited amount of NP because of its specific subunit compositions [43]. The authors also found anti-HCRTR2 autoantibodies were detectable in post Pandemrix 1 narcolepsy patients (17 of 20 sera), but not in subjects (0 of 12 sera) after Focetria 1 vaccination in 2009. This article raised two interesting points: (1) the possibility of amino acid residue mutations, such as NP 116I to 116M in some vaccine strains, may influence vaccine response and mimicry; (2) a possible role of anti-HCRTR2 autoantibodies in the pathophysiology of narcolepsy [43]. In this study, we further evaluated these hypotheses by (1) examining the prevalence of NP 116I and 116M in Pandemrix 1 , Arepanrix 1 and Focetria 1 using MS characterization and (2) testing for the presence of anti-HCRTR2 autoantibodies using three independent strategies in post Pandemrix 1 narcolepsy (PP-N) patients, recent early onset narcolepsy (EO-N) patients, post Pandemrix 1 controls (PP-C), and other controls (O-C).

Ethics statement
This study was reviewed and approved by the Stanford University Institutional Review Board (Protocol # 14325, Registration # 5136). Informed consent was obtained from each participant.

Mass spectrometry (MS)
Detailed protocols have been published previously [33]. Briefly, MS was performed on Trypsin/Lys-C mix (Cat# V5073, Promega) and chymotrypsin (Cat# V1062, Promega) digests of Pandemrix 1 , Arepanrix 1 and Focetria 1 (6.5 μg) samples. Each vaccine was first precipitated using 4× volume of high-performance liquid chromatography (HPLC) grade acetone at -80˚C, followed by overnight precipitation. Centrifugation at 4˚C, 12000× g for 15 minutes was performed, supernatant removed and protein pellets dried. Pellets were digested (1:25 of protease to protein ratio at 37˚C for 4-18 hours), then quenched by the addition of 10% formic acid. Resulting peptides were cleaned on a micro spin column (Cat# NC9270379, Nest group Inc.) and dried by speed-vacuum. Samples were then reconstituted in 20 μL 2% acetonitrile, 0.1% formic acid, and 97.9% water, and 3 μL of sample was injected onto a self-packed 15 cm C18 reverse phase column (Easy-nLC II, ThermoFisher Scientific) where a linear gradient from 3-40% mobile phase B was used over 2 hours to elute peptides into the mass spectrometer. Sample ionization was done using a spray voltage of 1.7 kilovoltage (kV). The mass spectrometer was an LTQ Orbitrap Velos or Orbitrap Fusion (ThermoFisher Scientific) set to acquire in a data dependent acquisition (DDA) mode, in which the top 12 most intense multiply charged precursor ions were selected for fragmentation by the ion trap. The target ion values (AGC settings) were 7.5×10 5 and 2.5×10 4 for the orbitrap and the ion trap, respectively.

MS analysis of HA and NP proteins
Raw MS data of these vaccines (S1 and S2 Raw Data) were analyzed using a combination of Byonic 1 and Preview 1 software (version1.4, Protein Metrics). Representative MS spectra were exported from Byonic 1 . Data analysis with Preview 1 was completed for each sample using a concatenated FASTA file containing the canonical proteomes for five influenza viral strains (A/California/07/2009, NYMC X-181A (identical to NYMC X-181), NYMC X-179, NYMC X-179A, and A/Puerto Rico/8/1934). NCBI accession numbers of X-179A viral strains we used for peptide mapping were ADE2909 (used in Ahmed et al. [43] study) ( Table 1) and AIE5269, which was our more recently released (May 2014) and contained an NP 116M (Table 1). These two X-179A sequences contained identical alleles of HA 146N. The X-181 sequence we used was AFM7284 (June 2012) ( Table 1), which contained HA 146D and NP 116M [33,43]. All statistical analyses were performed on raw data and were undertaken using Cochran-Mantel-Haenszel chi square tests using RStudio 3.1.0, using mantelhaen.test and cmh.test functions [46].

HA and NP peptides binding to DQ0602 monomers
Complementary deoxyribonucleic acid (cDNA) templates of HLA-DQA1 Ã 01:02 and DQB1 Ã 06:02 were obtained from the Emory University NIH core tetramer facility (http:// tetramer.yerkes.emory.edu/support/faq). Soluble DQ0602 monomers were expressed in "high five" cell line (a gift from K. Christopher Garcia lab, Stanford University) and purified using fast protein liquid chromatography (FPLC). The CLIP peptide, which derives from the major histocompatibility complex (MHC) class II-associated invariant chain (li), was removed by cleavage with thrombin (Cat# 69671, EMD Millipore). CLIP plays a critical role in the assembly of MHC, especially for antigen processing by stabilizing peptide-free DQ0602 complex [47][48][49]. Another nonclassical MHC class II heterodimer molecule, HLA-DM, which regulates and catalyzes antigenic peptide loading onto DQ0602 [50,51], was also expressed in "high five" cell line and purified. Peptides were ordered at GenScript company with >90% purity and dissolved in dimethyl sulfoxide (DMSO) at stock concentration of 10 mM.
with 300 μl/well of wash buffer, DELFIA 1 time-resolved fluorescence (TRF) intensity was detected using a Tecan Infinite 1 M1000 after adding 100 μl/well of enhancement solution (Cat# 1244-105, PerkinElmer). Non-specific binding was removed through extensive wash with wash buffer. Competitor peptide with Eu TRF intensity that was lower than 25% of Bio-EBV epitope alone was considered strong binder, while peptide with 25-50% was weak binder.
Methods for quantitation analysis have been described previously [54,55] and were modified as follows: for each sample, mean fluorescence intensities (MFI) of AF555 channel (MFI AF555 ) within live GFP positive HEK293T cells (HEK293T GFP+ ) and GFP negative HEK293T cells (HEK293T GFP-) were calculated, then ΔMFI AF555 was determined by subtracting MFI AF555 of HEK293T GFPfrom MFI AF555 of HEK293T GFP+ . A staining control with positive antibody at 1:100 dilution was included in each staining experiment. The MFI AF555 index was calculated as follows: 100× (ΔMFI AF555 of each sample)/(ΔMFI AF555 of positive antibody at 1:100 dilution). Cut-off value for a positive MFI AF555 index is determined as the mean of MFI AF555 + 3× standard deviation (SD) of control subjects [56].

Anti-HCRTR2 autoantibody detection using in-cell ELISA
In-cell ELISA colorimetric assay was performed according to the manufacturer's instructions (Cat# 62200, ThermoFisher Scientific). CHO-HCRTR2 cell line was made in a CHO-K1 host cell line (purchased from ATCC, https://www.atcc.org/products/all/CCL-61.aspx) [57,58] and cultured in Kaighn's modification of Ham's F-12 (F-12K) medium (ATCC 1 30-2004, ATCC) supplemented with 10% FBS (Cat# 26140079, ThermoFisher Scientific) and 1% P/S (Cat# 10378016, ThermoFisher Scientific) at 37˚C, 5% CO 2 . Transgenic HCRTR2 expression was detected by flow cytometry (S3 Fig) and western blotting (S4 Fig). Cell number sensitivity, from 625 to 100,000 live CHO-HCRTR2 cells per well, was tested with a series of positive antibody (Cat# ab65093, Abcam) concentrations (S4 and S5 Tables), and 10,000 live CHO-HCRTR2 cells per well were used in all the other in-cell ELISA assays. In order to verify performance of the in-cell ELISA assay, two additional positive antibodies (Cat# 07505-20, United States Biological, Cat# AP55124SU-N, Acris) present in rabbit sera were also tested with titration (S6 Table). Briefly, 10,000 live CHO-HCRTR2 cells/well were plated in a black collagen I-coated 96-well plate (REF# 152035, ThermoFisher Scientific), and cultured overnight at 37˚C, 5% CO 2 . Cells were fixed with 4% formaldehyde (Cat# 28906, ThermoFisher Scientific) for 15 minutes at 4˚C in the dark, followed by blocking overnight at 4˚C with 200 μL/well of blocking buffer (included in kit). 30 concentrations of positive antibodies (Cat# ab65093, Abcam) in duplicate were made to generate standard curve using 4-parameter logistic fit by SoftMax 1 Pro software (version 5.4, Molecular Devices) (S7 and S8 Tables). In order to make all sera (2 samples running out) fall within a standard range, sera were diluted at different ratios in 1× tris-buffered saline (TBS) buffer (included in kit) and added to plates in duplicate for incubation overnight at 4˚C (S7 and S8 Tables). After washing the plates three times with 200 μL/well of 1× wash buffer (prepared according to the manufacturer's instruction), 100 μL/well of diluted horseradish peroxidase (HRP) conjugate was added and incubated for 30 minutes at RT. The plates were then washed three times with 200 μL/well of 1× wash buffer, and 100 μL/well of 3,3',5,5'-tetramethylbenzidine (TMB) substrate (included in kit) was added. The solution was incubated in the dark at RT and the reaction was stopped within 15 minutes by adding 100 μL/well of TMB stop solution (included in kit). The plate signal was read immediately using a SpectraMax 1 M2 plate reader (Molecular Devices) at 450 nm. The original OD450 of each serum was analyzed using the SoftMax 1 Pro software (version 5.4, Molecular Devices) and corrected to that at 1:800 dilution (S7 and S8 Tables). The OD450 shown in related graphs or figures is determined by subtracting OD450 of 1× TBS buffer in the same plate from corrected OD450 of each serum. The cut-off value for a positive sample is determined as the mean + 3× SD of control samples [56].

Statistical analyses
For MS analysis, categorical values were expressed as percentages, linear values as mean ± SD or standard error. Data were analyzed using SAS software (SAS Institute Inc.) with two-sided t-tests. P-value <0.05 was considered statistically significant.
For anti-HCRTR2 autoantibody analysis and peptide binding assay, values were expressed as mean ± SD, and statistical comparisons were calculated using two-tailed t-tests in Microsoft Excel. Data were plotted using GraphPad PRISM 5 (GraphPad Software, Inc.). P-value <0.05 was considered statistically significant.

NP 116M was predominantly present in all vaccines tested
Spectrograms were mapped on a library of peptides generated from the trypsin and chemotrypsin digests of X-179A and X-181 with accession numbers ADE2909 (March 2010), AIE5269 (May 2014), and AFM7284 (June 2012) (Fig 1, Table 1, S5 Fig, S9-S18 Tables). Frequencies of NP 116I and 116M, HA 146N and 146D for Pandemrix 1 , Arepanrix 1 and Focetria 1 were shown in Table 2, with the caveat that much lower NP coverage was obtained with Focetria 1 (consistent with the subunit nature of this vaccine). We found that the occurrences of mutation vs wild type in Pandemrix 1 , Arepanrix 1 , and Focetria 1 were 47 vs 2, 76 vs 3, and 5 vs 0, respectively, indicating that the NP 116M mutation was predominantly present in all the three vaccines (Fig 1, Table 2).
No positive reaction was found in PP-N patients with each of these three different methods (Fig 4, Table 3, S3, S7 and S20 Tables). No significant difference was found either between PP-N patients and matched PP-C controls or between EO-N patients and matched O-C controls with each of the three tests, except for the former pairs in in-cell ELISA (p-value = 0.0399)  Table 3, S3, S7 and S20 Tables). Any other subject which showed positive reaction with one test was found negative with other two methods (Fig 4, Table 3, S3, S7 and S20 Tables). No known factor was positively correlated to these positive subjects (Table 3).

Discussion
This work shows that anti-HCRTR2 autoantibodies were rarely detected (3 or 4 positive reactions of 176 sera), with no positive subject found in post Pandemrix 1 narcolepsy patients (PP-N). We were unable to confirm that anti-HCRTR2 autoantibodies were present in a large portion of the population, extending on the work in 191 samples recently published by Giannoccaro et al. [61], who used HEK293 cells transiently transfected with human HCRTR2 and scored each serum stained with HEK293 cells, and found 3 of 61 patients positive: two type 1 narcolepsy and one type 2 narcolepsy, but none of the control subjects. In our study, to maximize the possibility of finding positives, we tested 40 post Pandemrix 1 narcolepsy cases and 18 matched controls (similar to those used in the Ahmed et al. [43]), plus 48 sera of patients who had a very recent onset ( 6 months) of narcolepsy. The last group was selected because of the highest likelihood the autoimmune process would still have been active in these patients, and thus detection of autoantibody would be the easiest. With CBA detection described by Giannoccaro et al. [61], positive staining was not observed with positive anti-HCRTR2 antibody (S11 Fig). Therefore, detection of anti-HCRTR2 autoantibodies was performed using three other independent techniques, one of which was similar to that used by Ahmed et al. [43] (in-cell ELISA) except for the use of a different cell line, CHO-K1 instead of Chem-1. The use of the CHO-K1 cell line was due to the fact that the stably transfected Chem-1-HCRTR2 cell line could not be made available to us. The second technique, [ 35 S]-radiolabelled HCRTR2 binding assay, had been previously reported by Tanaka et al. [56], who could not find evidence for anti-HCRTR2 autoantibodies. Using this second technique, we confirmed this observation (Fig 4, Table 3, S10 Fig, S3 and S22 Tables). The disadvantage of this method was that in vitro  No anti-HCRTR2 autoantibodies in narcolepsy translated polypeptide chain was unlikely to be conformationally similar to HCRTR2 that was naturally embedded in cytoplasmic membranes. Because of the limitations of these two techniques (different cell line and conformational difference), we also used a third, more novel technique, which involved HEK293T cells transiently transfected with HCRTR2-GFP. Cells that were double labeled with GFP and anti-HCRTR2 antibody could be easily visualized in the upper right quadrant and distinguished from background or cells that have not successfully expressed the transgene. This last method, which has to our knowledge only been used by a few investigators, such as for the detection of MUC1-Tn antibodies [55], or N-methyl-Daspartate receptor (NMDAR) or dopamine-2 receptor (D2P) autoantibodies [54], was found to be more sensitive (see positive controls in Fig 3 and S1 Fig) and yielded a fewer positives in both narcolepsy patient and control subjects (Fig 3, Table 3, S1 Fig, S2 Table). Like the in-cell ELISA, this last technique should be targeting a normally conformed HCRTR2. As noted in Table 3, a few positives were found with one or another technique, but none of the sera reacted positively using all the three techniques.
Although these results were disappointing, it was also notable that the presence of anti-HCRTR2 autoantibodies in some narcolepsy patients and controls reported by Ahmed et al. [43] was unlikely to be causative to the condition. Indeed, in narcolepsy, the cells containing the hypocretin ligand, not those carrying hypocretin receptors, were the targets of the autoimmune process. Hypocretin receptors were expressed on a large number of target neurons in the brain [62], and one would have had to hypothesize some special vulnerability for hypocretin cells to be destroyed by an anti-HCRTR2 process. Further, as discussed by Vassali and Tafti [63], hypocretin receptors are likely not present on hypocretin cells.
This report also allowed for the clarification of the HA sequence at position 146 and NP sequence at position 116 in these vaccines, and the ability of these sequences to bind to DQ0602. Regarding HA 146N, the fact that the pH1N1 wild-type sequence HA 143-155 (HDSNKGVTAACPH) (146N was underlined) was present in Pandemrix 1 , Arepanrix 1 and Focetria 1 was difficult to interpret with respect to narcolepsy risk. Our peptide binding studies confirmed that HA 136-156 (KTSSWPNHDSNKGVTAACPHA) containing 146N (underlined) bound to DQ0602, and that the introduction of the D mutation at 146 position reduced binding affinity (Fig 2 and S19 Table), a result that was predicted by bioinformatics in our prior publication [33]. Surprisingly, however, whereas information provided by Novartis has indicated that Focetria 1 used in the vaccination campaign was primarily derived from X-181, in this study we found that the predominant residue at 146 position was N (Table 2), like Pandemrix 1 , suggesting the strain used to produce this particular batch was X-179A, not X-181. Further examination of EMA records indicated that a switch to X-181 likely operated in the midseason of 2009 as use was originally granted for A/California/07/2009 (H1N1)v-like strain NYMC X-179A, while X-181 use was approved on 11/11/2009 (see EMA information at http:// www.ema.europa.eu/ema/index). It was thus likely that the batch in this study provided by Novartis was an earlier batch, and that most of the pH1N1 vaccination campaign used X-181 and the HA 143-155 (HDSDKGVTAACPH) sequence containing 146D (underlined), differently from the batch examined here. The hypothesis that HA 136-156 (KTSSWPNHDSNKGVTA ACPHA) containing 146N (underlined), a DQ0602 binding epitope predominantly present in Pandemrix 1 and the pH1N1-2009 virus may be related to narcolepsy risk remained a possibility. In spite of multiple efforts, however, we have been unable to find homology for this sequence with known hypocretin neuron proteins [64][65][66].
Regarding the NP 116I, unlike in the wild-type PR8 NP (H1N1-1918-like) sequence, it appeared that this residue has been mutated from I to M in the X-179A strain that has been used for production of not only Pandemrix 1 , but also Arepanrix 1 and Focetria 1 . The fact that NP 111-121 (YDKEEIRRIWR) (116I was underlined) and HCRTR2 [34][35][36][37][38][39][40][41][42][43][44][45] (YDDEEFLRYLWR) did not appear to bind to DQ0602 (unlike what was reported with the Proimmune 1 array by Ahmed et al. [43]) (Fig 2, S19 Table) suggested this epitope was likely irrelevant to DQ0602-associated narcolepsy or differential vaccine risk. Regarding the antibody cross-reactivity previously reported with this epitope by Ahmed et al. [43], and the fact that could not be reproduced here, explanations could be advanced. Weak antibody cross-reactivity between infectious targets and self-proteins may not be as rare as anticipated. For example, a few years ago, in a study of anti-streptococcal antibodies in narcolepsy, we found that anti-streptolysin (ASO) antibodies, observed more frequently in recent onset narcolepsy sera, cross-reacted weakly with the human phosphodiisomerase (PDI) protein [14,67]. This result was initially exciting to us, however, we rapidly found these autoantibodies to be present in both ASO-positive narcolepsy patients and controls tempering our enthusiasm. Similarly, Deloumeau et al. [68] reported autoantibodies against hypocretin instead of the HCRTR2 as immune complexes, observations that were only seen in post chemical treatment sera, a finding that has not been reproduced to date.
In conclusion, in this comprehensive study we confirmed that mutations could accumulate in specific flu vaccines as they were propagated in eggs, deviating from the original strain sequence. The presence of one specific mutation in Arepanrix 1 , HA 146D, was confirmed in multiple batches used in Canada. Unfortunately, we were not able to use those batches that were used in Scandinavia in 2009-2010. Theoretically there was a possibility that they may have differed from the batches we studied. Skowronski et al. [69] similarly found that key antigenic residues were mutated in H3N2 reassortant vaccine strains and that this likely contributed to reduced efficacy in 2012-2013. These results suggest that regular DNA sequencing or MS characterization of viral isolates may be useful as quality control measures across long periods of production. We also found that previously reported results suggesting molecular mimicry between NP and HCRTR2 could not be reproduced and that autoreactivity by autoantibodies to HCRTR2 was unlikely to play a role in the pathophysiology of narcolepsy.  Table. Raw data of anti-HCRTR2 autoantibody analysis using in-cell ELISA for repeat. (XLSX) S9 Table. MS details of peptide fragment around NP 116M from Pandemrix 1 corresponding to S2 Raw Data. (XLSX) S10 Table. MS details of peptide fragment around NP 116M from Arepanrix 1 corresponding to S2 Raw Data. (XLSX) S11 Table. MS details of peptide fragments around NP 116M, HA 146N, and HA 146D from Focetria 1 corresponding to S2 Raw Data. (XLSX) S12 Table. MS details of peptide fragment around NP 116I from Pandemrix 1 corresponding to S2 Raw Data. (XLSX) S13 Table. MS details of peptide fragment around NP 116I from Arepanrix 1 corresponding to S1 Raw Data. (XLSX) S14 Table. MS details of peptide fragments around NP 116I and HA 146N from Pandemrix 1 corresponding to S2 Raw Data. (XLSX) S15 Table. MS details of peptide fragment around HA 146N from Arepanrix 1 corresponding to S1 Raw Data. (XLSX) S16 Table. MS details of peptide fragment around HA 146N from Arepanrix 1 corresponding to S1 Raw Data. (XLSX) S17 Table. MS details of peptide fragment around HA 146D from Pandemrix 1 corresponding to S1 Raw Data. (XLSX) S18 Table. MS details of peptide fragment around HA 146D from Arepanrix 1 corresponding to S1 Raw Data. (XLSX) S19