CSF was retrospectively collected from a total of 30 individuals between 2014 and 2019 at the memory and dementia clinic of the 1st and 3rd Department of Neurology, AHEPA and “G. Papanicolaou” Hospitals, School of Medicine, Aristotle University of Thessaloniki, Greece. The study was approved with written informed consent from study individuals and by the Greek Alzheimer Association and Related Disorders (GAARD) scientific and ethics committees, and the Institutional Review Board of the University of Toronto.

The study participants included 10 control individuals with headache, 10 patients with AD and 10 patients with PD. Clinical diagnosis of probable AD was made based on the NINCDS/ADRDA criteria for probable AD with a threshold cut-off for AD at a Mini-Mental State Examination (MMSE) score of 26 ¹² . Clinical diagnosis of PD was made based on the modified Hoehn and Yahr (H-Y) scale ¹³ . Functional Rating Scale for Symptoms of Dementia (FRSSD) was also measured to assess the impact of dementia on patients’ daily activities.

Following confirmation of diagnosis, CSF samples were collected by lumbar puncture in the morning, centrifuged to remove cellular components and stored at -80°C polypropylene tubes. The samples were then shipped to the Lunenfeld Tanenbaum Research Institute, Mount Sinai Hospital, Toronto, Canada and stored at -80°C until further processing.

Total protein was extracted from four regions of the brain: frontal cortex, pons, cerebellum and brain stem. Each tissue was pulverized in liquid nitrogen using a mortar and pestle. The pulverized tissue was further digested with 0.2% RapiGest SF Surfactant (Waters, Milford, MA, USA) in 50 mM ammonium bicarbonate (ABC) for 30 min on ice, while vortexing every 2–5 min. The homogenate was sonicated on ice for three times, 15 s each, and centrifuged at 15,000 g for 20 min at 4°C. The resulting pellet containing debris and insoluble contaminants was removed. Pierce bicinchoninic acid assay (Thermo Fisher Scientific, San Jose, California) was performed to determine total protein concentration. Fractions from each brain region were pooled in equal parts (in terms of total protein contribution).

The experimental protocol has been described elsewhere ¹¹ . Briefly, 50 µL of 10% w/v Protein-G Mag Sepharose Xtra magnetic beads (GE Healthcare) medium slurry was resuspended by vortexing and added to a microcentrifuge tube. The microcentrifuge tube was placed in a magnetic separator, and the storage solution was removed. The magnetic beads were washed with 500 µL PBS. CSF samples were spiked with 100 ng of human kallikrein 6 (HK6) mouse monoclonal antibody, purified in-house with high sensitivity and specificity ¹⁴ , as a positive control and added to the magnetic beads. PBS was added to the mixture to reach a final volume of 300 µL. IgG from the CSF was bound to the beads during a 30 min incubation with gentle rotation. After two washes with 500 µL PBS, 100 µg of the pooled brain lysate was added to the beads, followed by a 2-hour incubation with gentle rotation. Following incubation, the beads were washed three times with 500 µL PBS 0.05% Tween 20, and subsequently washed three more times with 500 µL PBS. The beads were reconstituted in 100 µL PBS.

The reconstituted beads, along with the captured antibodies and antigens, were reduced by adding 100 mM dithiothreitol (DTT) to a final concentration of 5 mM, and incubated at 56°C for 40 min. For alkylation, 500 mM iodoacetamide (IAA) was added to a final concentration of 15 mM and incubated for 30 minutes in the dark with gentle shaking. For digestion, trypsin was added to each sample in a 1:50 enzyme to substrate ratio and incubated at 37°C overnight with gentle shaking. The supernatant was collected using the magnetic separator, and formic acid was added to a final concertation of 1%, reaching a pH of 2, to stop the reaction.

Peptides were purified by extraction using OMIX C18 tips (Agilent Technologies, Santa Clara, CA), eluted with 3 µL acetonitrile buffer solution (0.1% formic acid in 65% acetonitrile) supplemented with 57 µL of 0.1% formic acid. Using an auto-sampler, 18 µL of sample, run in technical duplicates, was injected from a 96-well plate into a C18 Acclaim PepMap 100 (75 µm x 2 cm, C18 3 µm bead, 100 Å pore size) trap column (Thermo Fisher Scientific, San Jose, California) and peptides were eluted into a 50 cm analytical column (PepMap RSLC C18, 75 μm ID, 2 μm bead, 100 Å pore, ES803, Thermo Fisher Scientific). The liquid chromatography, EASY-nLC 1200 system (Thermo Fisher Scientific), was coupled online to a Q Exactive HF-X (Thermo Fisher Scientific) mass spectrometer with the EASY-Spray ionization source (Thermo Fisher Scientific) with a spray voltage of 2 kV and capillary temperature at 320°C. The 60-minute liquid chromatography (LC) was applied at a flow rate of 250 nl/min with an increasing concentration of buffer D (0.1% formic acid in 95% acetonitrile). In a 60-min data-dependent acquisition (DDA) mode, full MS1 scan was acquired from 400 to 1500 m/z at a resolution of 60,000 in profile mode, followed by MS2 scans of the top 28 parent ions at a resolution of 15,000. Dynamic exclusion was set to 20 s, and 1+ and 6+ or more charge state ions were excluded from MS2 fragmentation. MS method parameters were detailed previously ¹¹ .

Raw files were uploaded into the Proteome Discoverer v.1.4 (Thermo Fisher Scientific) and searched with Sequest HT search engine against the Human 5640 Swiss-Prot protein database (January 2018) ( MaxQuant is an open-source alternative to Proteome Discoverer). The search parameters included: trypsin enzyme with two maximum missed cleavages, cysteine carbamidomethylation as a static modification, precursor mass tolerance of 7 ppm, fragment mass tolerance of 0.02 Da, methionine oxidation as a dynamic modification, 1% false-discovery rate (FDR) at the peptide and protein level using the Percolator node.

Abundant serum proteins that may bind non-specifically to the beads, including hemoglobin, haptoglobin, hemopexin, immunoglobins, keratins, apolipoproteins, serum albumin and complement, were removed from the initial candidate selection. Candidate autoantibody biomarkers were identified by antigens that were 1) absent in the patient control group, defined as identification in a maximum of one out of the 10 control individuals and 2) identified in the patient group, defined as presence in at least half of the patients (5 out of 10).

Statistical analyses for clinical descriptions were performed using GraphPad Prism v. 6.0e. A p-value <0.05 was considered significant. Chi-square test (overall and pairwise) was used to compare categorical demographic characteristics of the three groups. Non-parametric Kruskal-Wallis test by ranks was used to compare characteristics on a continuous scale. Dunn’s multiple comparison test was applied for pairwise comparisons.

Patient descriptions are shown in Table 1 and Underlying data ¹⁵ . There were no significant differences in the proportion of males and females. Age was significantly different between groups ( p = 0.0002), and pairwise comparisons revealed lower age in the control group than both AD ( p = 0.0005) and PD ( p = 0.0025) patient groups. MMSE score was significantly different between groups (<0.0001), and multiple comparisons revealed that AD ( p < 0.0001) and PD ( p = 0.0188) groups had lower MMSE scores than the control group. The median (interquartile range) of the H-Y score in the PD group was 2.0 (1.5-2.8). FRSSD was not significantly different between AD and PD groups.

The total protein content from human brain regions, frontal cortex, pons, cerebellum and brain stems, ranged from 1.7 mg/mL to 5.3 mg/mL ( Extended Data: Supplementary Table 1 ¹⁶ ).

Antibodies from the CSF, bound to protein-G beads, captured putative cognate antigens from the brain tissue-mix. Mass spectrometry identification of these cognate autoantigens infer presence of brain-specific autoantibodies from the CSF. Using a 1% FDR for peptide identification, the number of antigens detected in each individual CSF ranged from 461 to 1192, amounting to 1508, 1754, 1452 total antigens identified in the control, AD and PD groups, respectively. After removal of abundant serum proteins, 1342, 1562 and 1281 antigens were remaining in control, AD and PD respectively. The number of antigens that were uniquely found in control, AD and PD groups was 137, 299 and 129, respectively. A Venn diagram of the putative CSF autoantibody-bound antigens detected in each group, after removal of abundant serum antigens is shown in Figure 1. The positive control, human kallikrein 6, (hK6), was abundantly identified in all samples, with nine to 13 unique peptides ( Extended Data: Supplementary Table 2 ¹⁶ ). See Underlying data ¹⁵ for details of all.

In total, 16 putative autoantibodies fulfilled our aforementioned selection criteria in the AD group. No candidates fulfilled the criteria in the PD group. The candidates, along with the number of unique peptides identified for each antigen, are summarized in Table 2. More details on the identity of each identified peptide per antigen, are shown in Extended Data: Supplementary Table 2 ¹⁶ .

For AD, three autoantibodies against brain-specific antigens, glia-derived nexin (SERPINE2), fibromodulin (FMOD) and quinone oxidoreductase (NQO1), were absent in CSF of all patients in the control group and were present in eight, six and five patients with AD, respectively ( Table 2). SERPINE2 was identified with an average of 2.3 unique peptides in each individual ( Extended Data: Supplementary Table 2 ¹⁶ ), totaling seven unique peptides in the whole patient group. FMOD was identified with an average of 1.5 unique peptides per patient with a total of 4 peptides in the whole patient group. Finally, NQO1 was identified in five patients with one unique (the same) peptide per patient ( Extended Data: Supplementary Table 2 ¹⁶ ).

A further 13 putative autoantibodies against brain-specific antigens were found in one out of ten control individuals and at least half of the AD patients. In all cases in the control group, the antigen was identified using only one unique peptide ( Table 2 and Extended Data: Supplementary Table 2 ¹⁶ ). Autoantibodies against cathepsin F (CTSF), cadherin-13 (CDH13), and phospholipase D4 (PLD4) were identified in six AD patients, with an average of 2.3, 2 and 1 unique peptide per patient, respectively ( Table 2 and Extended Data: Supplementary Table 2 ¹⁶ ). The remaining candidates were identified in five AD patients ( Table 2). This included inositol 1,4,5-triphosphate receptor type 1 (ITPR1, or IP3 receptor), sushi repeat-containing protein (SRPX), isoaspartyl peptidase (ASRGL1), heterogeneous nuclear ribonucleoprotein H (hnRNP H), cerebellin-3 (CBLN3), oligodendrocyte-myelin glycoprotein (OMG), metalloproteinase inhibitor 2 (TIMP-2), WD repeat-containing protein 1 (WDR1, or AIP1), 4F2 cell-surface antigen heavy chain (SLC3A2) and sarco/endoplasmic reticulum calcium ATPase 2 (ATP2A2, or SERCA2). The number of unique peptides detected for each antigen ranged from 1 to 1.8 ( Table 2 and Extended Data: Supplementary Table 2 ¹⁶ ).

Harvard Dataverse: Putative autoantibodies in the cerebrospinal fluid of Alzheimer’s disease patients. https://doi.org/10.7910/DVN/DYMEQR ¹⁵ .

This project contains the following underlying data:

Harvard Dataverse: Extended data for "Putative autoantibodies in the cerebrospinal fluid of Alzheimer’s disease patients" https://doi.org/10.7910/DVN/JJW0LG ¹⁶ .

This project contains the following extended data: