Participants and clinical diagnostics

All participants were part of the BioFINDER-1 (NCT01208675; n = 398) or BioFINDER-2 (NCT03174938; n = 485) studies, described previously34,40,41. The inclusion criteria were (1) referred to participating secondary memory clinics due to cognitive symptoms recognized by the patient, caregiver and/or the referring physician; (2) age 40–100 years; and (3) speaks and understands Swedish to the extent that an interpreter is not necessary for the patient to fully understand the study information and cognitive tests. The exclusion criteria were (1) significant unstable systemic illness or organ failure, such as terminal cancer, which makes it difficult to participate in the study; and (2) current significant alcohol or substance misuse. Only participants (1) with MCI or dementia at baseline; (2) with complete biomarker data for Aβ, tau and α-syn (LB pathology); and (3) referred to the participating memory clinics of Skåne University Hospital or the hospital of Ängelholm in Sweden due to cognitive symptoms recognized by the patient, caregiver and/or the referring physician, were included in the present study. All patients were enrolled and underwent baseline examination from 2007 to 2015 (BioFINDER-1) or from 2017 to 2021 (BioFINDER-2). MCI was classified as not fulfilling the criteria for dementia (major neurocognitive disorder according to DSM-5 (ref. 42)) and performing worse than −1.5 × s.d. in at least one of the cognitive domains of memory, attention/executive, verbal or visuospatial function. In BioFINDER-1, this was assessed by a senior neuropsychologist after a thorough neuropsychological battery, as described in detail previously43. In BioFINDER-2, the MCI classification was operationalized as performing worse than −1.5 z scores in any cognitive domain according to regression-based norms accounting for age and education and the test performance in Aβ-negative controls44 (see elsewhere45,46 for a description on the regression-based z scores). Cognitive domain z scores were derived by calculating the mean z score of the tests in each of the following domains: attention/executive function (trail-making test A, trail-making test B and symbol digit modalities test), verbal ability (verbal fluency animals and the 15-word short version of the Boston naming test), memory (ten-word delayed recall from the Alzheimer’s disease assessment scale (ADAS)) and visuospatial function (incomplete letters and cube analysis from the visual object and space perception battery). Dementia was classified according the DSM-5 criteria for major neurocognitive disorders42.

A clinical diagnosis of AD was based on the DSM-5 criteria for mild or major neurocognitive disorder due to AD42. In addition, signs of Aβ positivity were required in agreement with the National Institute of Aging-Alzheimer’s Association13 and International Working Group4 criteria for AD. The biomarker for Aβ positivity was not the same as the Elecsys CSF Aβ42/Aβ40 assay used in the present study (described in detail previously41), which explains why some with a clinical AD diagnosis (Table 1 and Extended Data Table 1) are AD− in the AD/LB biomarker classification. Note that this clinical diagnosis was only used for describing the sample (Table 1 and Extended Data Table 1). The purely biomarker-driven classification for AD pathology (Aβ and tau positivity, regardless of clinical syndrome) was used in the statistical analysis. A clinical diagnosis of DLB was based on the McKeith criteria for probable DLB (MCI16 or dementia17 stage) and PD was based on previously published criteria for PD and PDD47 (where the DSM-5 criteria were used to identify presence of dementia/major neurocognitive disorder42). The PD diagnoses are detailed in Supplementary Table 3, where the criteria used in the present study (Gelb et al.47) are compared to the diagnostic classification according to the Movement Disorder Society (MDS) criteria15. If a participant fulfilled the criteria for both AD and DLB/PD he/she was coded as DLB or PD in the analyses. The other diagnostic entities were diagnosed according to published criteria42,48,49,50,51, as previously described34. Notably, to increase the clinical diagnostic accuracy, the diagnoses were also determined during a longitudinal follow-up of symptoms and advanced investigations at a secondary memory clinic (Extended Data Table 1). Note that α-syn SAA results were not available when the diagnoses were determined.

All participants or their legal representatives provided written informed consent. Ethical approval was given by the Regional Ethical Committee in Lund, Sweden.

Clinical outcomes

All clinical outcomes, except the presence of hallucinations and correctly completed visuospatial task, were z scored according to distribution in Aβ− cognitively unimpaired participants in BioFINDER-1 and BioFINDER-2. The modified Preclinical Alzheimer Cognitive Composite-5 (mPACC5; also referred to as PACC) was used as a measure of global cognition, containing tests of memory, executive, attention and verbal function52. It was calculated based on the previously described PACC5, using the MMSE, symbol digit modality test (SDMT) and animal fluency52. As the memory tests logical memory and the free and cued selective reminding tests were not available in BioFINDER, the ten-word delayed recall task from ADAS-cognition (ADAS-cog)53 was used (weighted twice), as previously applied in several studies54,55. The mPACC5 was thus calculated using z scores based on the distribution in Aβ− cognitively unimpaired in the following way: (MMSE + (ADAS-cog delayed recall × 2) + SDMT + animal fluency) / 5. Memory was measured using the ten-word delayed recall task from ADAS-cog53. Attention/executive function was measured using the SDMT56. If SDMT was not available, the serial 7s task of the MMSE was used (Supplementary Tables 1 and 2 detail missingness)57. Visuospatial function was measured using the incomplete letters task from the visual object and space perception battery and if incomplete letters was not available (per study design only available in BioFINDER-2), the pentagon copying task from the MMSE was used. As pentagon copying is scored as either normal (1) or abnormal (0), the incomplete letters task was binarized at ≤18 of 20 points (mean − 2 × s.d. in Aβ-negative cognitively unimpaired participants in BioFINDER-2).

Presence of hallucinations (yes / no) was assessed using the informant-based cognitive impairment questionnaire (CIMP-QUEST)58, item eight in the associated symptom’s part (‘Does the patient has hallucinations, seeing, hearing or feeling things that don’t actually exist but that he/she has a clear experience of?’). Motor function was measured using the combined score from all motor symptoms in CIMP-QUEST, covering aspects of bradykinesia, reaction, changed the way of walking, poorer balance, clumsier hands, changed facial expressions and dysarthria.

Biomarker of Aβ

The CSF Aβ42/Aβ40 ratio was used to define Aβ positivity, as Aβ-PET was not included at baseline in BioFINDER-1 and was not available in BioFINDER-2 participants with dementia, per study design. Aβ42 and Aβ40 were analyzed on a cobas e 601 analyzer using the Roche NeuroToolKit. The threshold for positivity was defined using mixture modeling statistics59. In BioFINDER-1, the previously established cutoff of CSF Aβ42/Aβ40 < 0.066 was used60. For BioFINDER-2, the cutoff was established in all BioFINDER-2 participants (cognitively unimpaired and impaired) with available CSF Aβ42 and Aβ40 data (n = 1,076), resulting in a cutoff of <0.080 (the difference in cutoffs between cohorts is explained by preanalytical differences in that LoBind tubes were used in BioFINDER-2 but not in BioFINDER-1 (refs. 61,62) for preanalytical protocols and differences).

Biomarker of tau

Tau positivity was defined as either abnormal CSF P-tau217 (BioFINDER-1) or abnormal tau-PET (BioFINDER-2). CSF P-tau217 was measured using the Meso Scale Discovery platform using an assay developed by Eli Lilly and tau-PET was performed using RO948 labeled with radioactive fluorine [18F], as previously described34. SUVR was measured in a temporal meta-ROI using the inferior cerebellar cortex as reference region34. Cutoffs were established at the mean + 2 × s.d. in Aβ− controls as previously described and the cutoff for CSF P-tau217 was >11.42 pg ml−1 and for tau-PET > 1.32 SUVR (ref. 34).

Preparation of recombinant α-synuclein (LB pathology)

The purification of recombinant wild-type α-syn was performed as previously reported8, with minor modifications. Briefly, transformed Escherichia coli BL21 (DE3) bacteria (New England Biolabs) from a glycerol stock were streaked on a selective plate (Kan+, 50 µg ml−1, kanamycin from Sigma) and incubated at 37 °C overnight. A single colony was selected and inoculated into 5 ml Luria Broth (Sigma) with kanamycin and allowed to grow for 4–5 h at 37 °C with continuous agitation at 250 r.p.m. This starter culture was then added to 1 l Luria Broth containing kanamycin and the overnight express auto-induction system (Merk-Millipore 71300-4) in a fully baffled flask. Cells were grown in a shaking incubator at 37 °C, 200 r.p.m. overnight. The next day, the culture was split into four 250-ml flasks and centrifuged at 3,200g for 10 min at 4 °C. The pellet was gently resuspended in 25 ml osmotic shock buffer containing 40% sucrose (Sigma), 2 mM EDTA (Sigma) and 30 mM Tris (Bio-Rad) at pH 7.2 using a serological pipette and incubated for 10 min at room temperature under mild agitation on a rotator mixer. The solution was then centrifuged at 9,000g, 20 min at 20 °C and each pellet was resuspended in 10 ml ice-cold Milli-Q water. The four suspensions were pooled into two 50-ml tubes and 20 µl of saturated MgCl 2 (Sigma). After 3 min incubation under mild rocking on ice, the suspension was centrifuged at 9,000g for 30 min at 4 °C and the supernatant was collected into a 100-ml glass beaker. The pH was reduced to pH 3.5 by adding 400–600 µl HCl 1 M (PanReac AppliChem) and incubated under stirring for 10 min at room temperature. After a second centrifugation at 9,000g for 30 min at 4 °C, the supernatant was collected into a clean 100-ml glass beaker. The pH was adjusted to 7.5 by adding 400–600 µl NaOH 1 M (Sigma). The protein extract was filtered through a 0.22-µm filter (Merk-Millipore), loaded into a Ni–NTA column (Cytiva 17525501) on an NGC chromatography system (Bio-Rad) and washed with 20 mM Tris, pH 7.5 at room temperature. The column was further washed with 50 mM imidazole (Sigma) in Tris 20 mM, pH 7.5, generating a peak that was not collected. A linear gradient up to 500 mM imidazole in 20 mM Tris, pH 7.5 was performed and the peak was collected between 30% and 75% of imidazole buffer (150 and 375 mM, respectively). This peak was loaded onto an anion exchange column Q-HP (Cytiva 17115401) and washed in Tris 20 mM, pH 7.5, followed by another washing in 100 mM NaCl in Tris 20 mM, pH 7.5. Again, a linear gradient up to 500 mM of NaCl in Tris 20 mM pH 7.5 was carried out to collect the peak between 300 and 350 mM NaCl. The fractions were pooled and filtered through a 0.22-µm filter and dialyzed against Milli-Q water overnight at 4 °C using a 3.5 kDa MWCO dialysis membrane (Thermo-Scientific). The next day, the protein was moved into fresh Milli-Q water and dialyzed for 4 h more. The protein concentration was measured by a spectrophotometer using a theoretical extinction coefficient at 280 nm of 0.36 (mg ml−1) − 1 cm−1. Finally, the protein was lyophilized for 6 h and stored in aliquots at a final concentration of 1 mg ml−1 after resuspension into 500 µl phosphate buffer (PB; Sigma) 40 mM, pH 8.0. Lyophilized aliquots were stored at − 80 °C until usage.

α-Syn RT-QuIC analyses

α-Syn RT-QuIC analyses were performed blinded to clinical status and diagnosis of the participant and according to an established protocol8,63,64, with minor modifications. Briefly, six 0.8-mm silica beads (OPS Diagnostics) per well were pre-loaded into black 96-well plates with a clear bottom (Nalgene Nunc International). CSF samples were thawed and vortexed for 10 s before use. Then, 15 μl CSF was added to a 85 μl reaction mix composed of 40 mM PB, pH 8.0, 170 mM NaCl, 10 mM thioflavin-T (Sigma), 0.0015% SDS (Bio-Rad) and 0.1 g l−1 of filtered recombinant α-syn (100-kDa Amicon centrifugal filters, Merck Millipore). Plates were closed with a plate sealer film (Nalgene Nunc International) and incubated into a FLUOstar Omega plate reader (BMG Labtech) at 42 °C with intermittent double orbital shaking at 400 r.p.m. for 1 min, followed by 1 min rest. Fluorescence was measured every 45 min with 450 nm excitation and 480 nm emission filters during the 30 h test run. Samples and controls were run in quadruplicate and considered positive after the first run when at least three out of four replicates reached a threshold arbitrarily set at 30% of the median of Imax values reached by the positive control replicates. To keep the risk of false-positive results at a minimum, we repeated three times the analysis of samples showing seeding activity in only one or two out of four replicates in the first run. We considered a positive result only when at least 4 of the 12 total replicates reached the threshold. We used 30 different batches of α-syn recombinant protein throughout the study. Each batch underwent a quality control test before use. We ran at least one positive and negative control on each plate. The positive controls were chosen from patients with probable or definite DLB or PD whose CSF samples gave four out of four positive replicates during screening. In each validated experiment (plate) included in the final analysis, the positive controls showed at least three out of four positive replicates.

Statistical analyses

In cross-sectional analyses, the AD/LB group, age, sex (assigned, not self-reported), cognitive stage (MCI or dementia) and, for cognitive test outcomes, years of education, were used as independent variables in general linear regression models. Dependent variables were either cognitive function, motor function or presence of hallucinations. Next, binarized Aβ, tau and LB pathology (to facilitate easier comparison of estimates), were used instead of the AD/LB group. Logistic regression models were used for the binary outcomes (visuospatial function and presence of hallucinations). In longitudinal analyses, LME models were used (R packages lme4 and the function lmer for continuous outcomes and glmer for binarized outcomes were used and P values were calculated using the package lmerTest). Cognitive function was used as an outcome and significant results were presented for the AD/LB group × time and pathology × time interactions. The models also included age, sex, cognitive stage (MCI or dementia), years of education and random slopes and intercepts. For models including pathology × time, the interaction between time and all covariates were also included. In a sensitivity analysis, the LME models using AD/LB group × time were also adjusted for baseline cognitive test result (if there was a difference between AD/LB groups at baseline) with change from baseline in cognitive test result as outcome. All available data were used in the statistical analyses. Missing data and number of participants at each visit are described in Supplementary Tables 1 and 2. A two-sided P value <0.05 was considered to indicate statistical significance. Multiple comparison corrections were performed using the false discovery rate method at α = 0.05, applying correction per outcome (six comparisons for AD/LB group comparisons and three for the independent effects of LB, Aβ and tau pathology). The statistical analyses were performed using R v.4.1.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.