CSF Biomarkers of Alzheimer Disease in Patients With Concomitant α-Synuclein Pathology

Abstract

Background and Objectives

CSF biomarkers β‐amyloid 1-42 (Aβ₄₂), phosphorylated tau 181 (p-tau₁₈₁), total tau (t-tau), and neurogranin (Ng) can diagnose Alzheimer disease (AD) in life. However, it is unknown whether CSF concentrations, and thus their accuracies, are affected by concomitant pathologies common in AD, such as α-synuclein (αSyn). Our primary goal was to test whether biomarkers in patients with AD are altered by concomitant αSyn. We compared CSF Aβ₄₂, p-tau₁₈₁, t-tau, and Ng levels across autopsy-confirmed AD and concomitant AD and αSyn (AD + αSyn). Antemortem CSF levels were related to postmortem accumulations of αSyn. Finally, we tested how concommitant AD + αSyn affected the diagnostic accuracy of 2 CSF-based strategies: the amyloid/tau/neurodegeneration (ATN) framework and the t-tau/Aβ₄₂ ratio.

Methods

Inclusion criteria were neuropathologic diagnoses of AD, mixed AD + αSyn, and αSyn. A convenience sample of nonimpaired controls was selected with available CSF and a Mini-Mental State Examination (MMSE) ≥ 27. αSyn without AD and controls were included as reference groups. Analyses of covariance (ANCOVAs) tested planned comparisons were CSF Aβ₄₂, p-tau₁₈₁, t-tau, and Ng differences across AD and AD + αSyn. Linear models tested how biomarkers were altered by αSyn accumulation in AD, accounting for pathologic β-amyloid and tau. Receiver operating characteristic and area under the curve (AUC), including 95% CI, evaluated diagnostic accuracy.

Results

Participants were 61 patients with AD, 39 patients with mixed AD + αSyn, 20 patients with αSyn, and 61 controls. AD had similar median age (73 [interquartile range {IQR} = 12] years), MMSE (23 [IQR = 9]), and sex distribution (male = 49%) compared with AD + αSyn age (70 [IQR = 13] years; p = 0.3), MMSE (25 [IQR = 9.5]; p = 0.19), and sex distribution (male = 69%; p = 0.077). ANCOVAs showed that AD + αSyn had lower p-tau₁₈₁ (F(1,94) = 17, p < 2.6e-16), t-tau (F(1,93) = 11, p = 0.0004), and Ng levels (F(1,50) = 12, p = 0.0004) than AD; there was no difference in Aβ₄₂ (p = 0.44). Models showed increasing αSyn related to lower p-tau₁₈₁ (β = −0.26, SE = 0.092, p = 0.0065), t-tau (β = −0.19, SE = 0.092, p = 0.041), and Ng levels (β = −0.2, SE = 0.066, p = 0.0046); αSyn was not a significant factor for Aβ₄₂ (p = 1). T-tau/Aβ₄₂ had the highest accuracy when detecting AD, including mixed AD + αSyn cases (AUC = 0.95; CI 0.92–0.98).

Discussion

Findings demonstrate that concomitant αSyn pathology in AD is associated with lower CSF p-tau₁₈₁, t-tau, and Ng levels and can affect diagnostic accuracy in patients with AD.

CSF signatures of Alzheimer disease (AD) neuropathologic change (ADNC) include decreased β-amyloid peptide 1–42 (Aβ₄₂) related to accumulation of β-amyloid plaques,¹ increased tau phosphorylated at threonine 181 (p-tau₁₈₁) associated with tau neurofibrillary tangles,¹ increased total tau (t-tau) associated with neurofibrillary tangles and neurodegeneration,^2,3 and increased neurogranin (Ng) linked to synaptic degeneration.⁴ Ratios, such as p-tau₁₈₁/Aβ₄₂^5,6 and t-tau/Aβ₄₂,^7,8 also indicate ADNC with high accuracy. Because of their sensitivity to AD neuropathologic processes, these CSF biomarkers can be used in life to stratify patients with AD from individuals without AD, including suspected non-AD pathophysiology (SNAP). However, binary stratification of AD from SNAP may be complicated by the presence of mixed pathology, common in AD.^9,10 In particular, Lewy body disease (LBD) with α-synuclein pathology (αSyn) is observed in an estimated 30%–50% of AD cases.^9,11 Likewise, roughly 50% of patients with LBD show significant ADNC.^2,12

It is unknown how concomitant αSyn affects CSF, an issue for any CSF-based strategy to detect AD, including the 2018 amyloid/tau/neurodegeneration (ATN) framework³ or the t-tau/Aβ₄₂ ratio in LBD.⁷ Postmortem work suggests that CSF Aβ₄₂ in LBD correlates with αSyn, independent of β-amyloid plaque burden.⁷ Patients with early Parkinson disease (PD) have lower CSF p-tau₁₈₁, t-tau, and Ng than healthy controls,^13,14 and longitudinal CSF p-tau₁₈₁ and t-tau levels may decline in the first 3 years in PD.¹³ Thus, αSyn may influence CSF biomarkers in a manner distinct from ADNC, thereby affecting diagnostic accuracy and interpretation when AD and αSyn are mixed. Because there are currently no established diagnostic biomarkers that can positively identify αSyn pathology, autopsy work is needed to determine whether diagnostic accuracy is affected when ADNC and αSyn co-occur.

To address this gap, this study compares CSF Aβ₄₂, p-tau₁₈₁, t-tau, and Ng levels in autopsy-confirmed AD and mixed AD and αSyn (AD + αSyn); we include αSyn without ADNC and nonimpaired controls as reference groups. Our primary goal is to test whether AD CSF biomarkers differ across pathologically defined ADNC with (AD + αSyn) or without (AD) concomitant αSyn pathology. Models also test whether CSF levels are affected by the other common age-associated pathology in our sample, transactive response DNA-binding protein of 43 kDa (TDP-43). We next test the direct association of CSF with increasing postmortem αSyn burden in AD brain tissue. Finally, we evaluate the diagnostic accuracy of ATN and t-tau/Aβ₄₂ ratio and test which strategy best detects ADNC in a mixed pathology sample. Post hoc exploratory analyses test biomarkers sensitive to αSyn, including Ng.

Methods

Patient Criteria

Patients were autopsied at the University of Pennsylvania Center for Neurodegenerative Disease Research and were retrospectively selected from the Penn Integrated Neurodegenerative Disease Biobank and Database.¹⁵ Inclusion criteria were a primary pathologic diagnosis of either αSyn (n = 32) or AD (n = 88). Patients had a clinical phenotype of either amnestic AD (n = 86) or LBD clinical spectrum, including PD (n = 3), PD with dementia (n = 16), or dementia with Lewy bodies (DLB; n = 15). The primary pathologic diagnosis was performed by expert neuropathologists (E.B.L. and J.Q.T.) using neuropathologic criteria.^16,17 All patients were assessed for CSF Aβ₄₂, p-tau₁₈₁, and t-tau using the xMAP Luminex platform.¹⁸ A subset of patients (32 patients with AD, 24 patients with AD + αSyn, and 14 patients with αSyn) were assessed for Ng using an in-house ELISA, previously described.⁴ Exclusion criteria were a neuropathologic diagnosis of frontotemporal lobar degeneration¹⁹ or cerebrovascular disease¹⁶ or a nonamnestic variant of AD.²⁰ One outlier for CSF t-tau was excluded for levels of 929 pg/mL (>9 SDs). A subset of these data (n = 22) was included in a previous publication focused on LBD.⁷ Nonimpaired healthy controls (n = 61) with a Mini-Mental State Examination (MMSE)²¹ of 28 or above and available CSF Aβ₄₂, p-tau₁₈₁, and t-tau were included as a comparison group; a subset of these controls had CSF Ng (n = 14). These selection criteria are outlined in a flowchart (eFigure 1, links.lww.com/WNL/C295).

Demographics were recorded as age at onset (earliest reported symptom), age at CSF collection, interval from CSF to death, age at death, disease duration at CSF (time from symptom onset to CSF), global cognition (MMSE), and sex. Two participants in our sample self-reported as Black/African American, both AD + αSyn; the rest self-reported as White. Consent was obtained according to the Declaration of Helsinki and approved by the Penn Institutional Review Board.

Neuropathologic Assessment and Patient Groupings

Tissue samples were processed as previously described, with immunohistochemical staining for p-tau, β-amyloid, TDP-43, and αSyn using well-characterized antibodies.^15,22 Of the total sample, 100 patients met the criteria for intermediate/high ADNC.¹⁶ αSyn was determined by αSyn-positive Lewy bodies in the brainstem, limbic, or neocortical regions.²³ Of the 100 patients with ADNC, 39 had both AD and αSyn neuropathologic diagnoses (AD + αSyn); because several patients had high levels of both pathologies, this grouping made no distinction between primary AD and primary αSyn. Patients with AD (n=61) had no or scant αSyn pathology (i.e., amygdala-only αSyn).²⁴ Patients with αSyn had not/low ADNC (n = 20). In addition to assessments for AD and αSyn neuropathologic alterations,¹⁵ 44 patients had co-occurring TDP-43 proteinopathy consistent with limbic-predominate age-related TDP-43 encephalopathy²⁵ with or without hippocampal sclerosis.²⁶

Pathologic Burden

Pathologic burden for β-amyloid, tau, and αSyn was assessed prospectively in neocortical, limbic, and brainstem regions standardly sampled at autopsy according to criteria¹⁶: middle frontal, angular, superior/middle temporal, occipital, amygdala, cingulate, CA1/subiculum, entorhinal, pons, and medulla regions. Sampling was randomized between left and right hemispheres. Each region was scored for pathologic severity using a semiquantitative 5-point scale (i.e., 0 = none, 0.5 = rare, 1 = low, 2 = intermediate, and 3 = high). An average burden score for β-amyloid, tau, and αSyn pathology was calculated across all regions.⁷

Statistical Analyses

Shapiro-Wilks tests indicated non-normal distribution of demographic and CSF variables. Kruskal-Wallis tests performed groupwise comparisons for continuous variables. Mann-Whitney-Wilcoxon tests performed planned comparisons between AD and AD + αSyn. Chi-square tests compared categorical variables. To perform nonparametric comparisons for analyses of covariance (ANCOVAs) and linear models, all continuous variables were rank transformed. Permutation testing calculated p values based on unique sum of squares (20,000 iterations). As indicated above, only a subset of participants had Ng data available, and 1 case of t-tau was removed as an outlier. In cases of these missing data, participants were dropped from analyses. Statistical tests were performed with a significance threshold of α = 0.05.

ANCOVAs using type II sum of squares tested whether CSF concentrations differed between patients with AD and AD + αSyn; covariates included age at CSF, CSF-to-death interval, sex, TDP-43 copathology, and APOE ε4. The effect size for ANCOVAs (partial η²) was calculated using 20,000 iterations (>0.01 is considered small, >0.09 medium, and >0.25 large). To ensure that CSF differences were not due to phenotype,²⁰ we repeated ANCOVAs in the subset of ADNC with an amnestic phenotype.

Linear models tested CSF levels as a function of αSyn burden, while accounting for β-amyloid and tau burden; CSF to death and sex were also included as covariates. β-Amyloid and tau accumulation were collinear, and therefore, β-amyloid was not included in final models. To ensure that observed effects of αSyn on CSF were not due to lower AD pathology, models were repeated in a subset of patients with ADNC with high ADNC, high β-amyloid (burden >2), and high tau (burden >2).

Receiver operating characteristic (ROC) analyses using bootstrapping (500 iterations) tested the diagnostic accuracy of each analyte (Aβ₄₂, p-tau, t-tau, and Ng) and 2 ratios (p-tau₁₈₁/Aβ₄₂ and t-tau/Aβ₄₂) when discriminating patients with ADNC (AD, AD + αSyn) from αSyn. Area under the curve (AUC) and 95% CI were reported; the Youden index determined best threshold that maximized sensitivity and specificity. We tested classification of AD, AD + αSyn, αSyn, and controls using 2 diagnostic strategies: ATN and t-tau/Aβ₄₂. To not bias results to favor a strategy, thresholds were specific to this sample. Finally, ROC analyses tested the diagnostic accuracy of each analyte and ratios when discriminating αSyn-positive patients (αSyn, AD + αSyn) from AD without αSyn. Analyses were conducted using R statistical software, using Companion to Applied Regression (car),²⁷ effectsize,²⁸ lmPerm,²⁹ and cutpointr³⁰ packages.

Data Availability

Anonymized data will be shared with qualified investigators who have institutional review board approval and a Material Transfer Agreement on request.

Standard Protocol, Approvals, Registrations, and Patient Consents

Written informed consent was obtained from all patients or legal representatives under guidelines established and approved by the Institutional Review Board at the University of Pennsylvania.

Results

Demographic Characteristics

Participants were 61 patients with AD, 39 patients with mixed AD + αSyn, 20 patients with αSyn, and 61 controls. Table 1 compares demographic characteristics across groups. Wilcoxon tests performed planned pairwise comparisons across patients with AD and AD + αSyn, with no difference for onset age (p = 0.29), age at CSF collection (p = 0.3), age at death (p = 0.48), CSF-to-death interval (p = 0.36), or MMSE scores (p = 0.19). Chi-square tests showed no differences in sex (χ²(1) = 3.1, p = 0.077), APOE ε4 alleles (p = 0.73), or presence of TDP-43 (p = 0.73) across AD and AD + αSyn.

Table 1.

Patient Characteristics

CSF Comparisons

Figure 1 illustrates CSF differences across groups. As expected, biomarkers reflected that patients with αSyn had clinically insignificant/absent AD pathology, with higher CSF Aβ₄₂ and lower CSF p-tau₁₈₁ and t-tau than AD and AD + αSyn; Ng was also lower in αSyn than AD, but levels were similar between αSyn and AD + αSyn. Controls had higher Aβ₄₂ and lower p-tau, t-tau, and Ng than AD; controls had higher Aβ₄₂ and lower p-tau₁₈₁ and t-tau than AD + αSyn; controls had higher Aβ₄₂, t-tau, and Ng than αSyn.

Figure 1. Comparisons of CSF Concentrations.

CSF levels of Aβ₄₂, p-tau₁₈₁, t-tau, and Ng across AD, AD + αSyn, αSyn, and controls. Color indicates ADNC from not (dark blue) to high (dark red) or not assessed (white). Asterisks represent p values from Wilcoxon pairwise comparisons (*p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001, or not significant [ns]). Boxplots show median, IQR, and outliers. AD = Alzheimer disease; ADNC = AD neuropathologic change; IQR = interquartile range; Ng = neurogranin; αSyn = α-synuclein.

ANCOVAs tested CSF differences across AD and AD + αSyn, covarying for age, CSF-to-death interval, sex, TDP-43 pathology, and APOE ε4. Concentrations were significantly lower for AD + αSyn than AD for CSF p-tau₁₈₁ (F(1,90) = 17, p <2.6e-16), t-tau (F(1,89) = 9.6, p = 0.0018), and Ng (F(1,48) = 11, p = 0.00055); all differences had a medium effect size (p-tau₁₈₁ η² = 0.16; t-tau η² = 0.097; Ng η² = 0.19). Only CSF Aβ₄₂ showed no difference across AD and AD + αSyn (p = 0.56).

Sex was also a significant factor in CSF concentrations, with significantly lower t-tau (F(1,89) = 9.3, p = 0.0074; η² = 0.095) and Ng (F(1,48) = 11, p = 0.0022; η² = 0.192) in men than women. Sex was not a significant factor for either p-tau₁₈₁ (p = 0.12) or Aβ₄₂ (p = 0.49).

CSF-to-death interval was not significant for Aβ₄₂ (p = 0.82), p-tau₁₈₁ (F(1,90) = 3.5, p = 0.074), t-tau (p = 0.4), or Ng (p = 0.097). Other nonsignificant factors were age, TDP-43, and APOE ε4 (all p > 0.1).

We tested whether observed differences could be in part due to lower ADNC in patients with AD + αSyn than patients with AD.¹² Wilcoxon test comparisons showed that patients with AD + αSyn indeed had lower β-amyloid (median = 2.3; interquartile range [IQR] = 0.53) than AD (median = 2.4; IQR = 0.6; W = 1,568, p = 0.0073). AD + αSyn also had lower tau burden (median = 2.1; IQR = 0.95) than AD (median = 2.5; IQR = 0.35; W = 1,640, p = 0.0015). In accordance with their pathologic diagnosis, AD + αSyn had greater αSyn burden (median = 1.4; IQR = 1.4) than AD (median = 0; IQR = 0.15; W = 89, p = 4.4e-15). Thus, subsequent linear models accounted for differences in burden of β-amyloid and tau across ADNC. We also repeated models in patients with ADNC with only high levels of β-amyloid and tau pathology.

CSF Comparisons in Amnestic ADNC

Because AD + αSyn consisted of patients with heterogeneous phenotypes (Table 1), we repeated ANCOVAs within patients with ADNC with an amnestic phenotype (eSection 5.1, links.lww.com/WNL/C295). Results were consistent except for CSF t-tau. Amnestic AD + αSyn had significantly lower p-tau₁₈₁ (F(1,79) = 7.4, p = 0.0041) and Ng (F(1,39) = 4.9, p = 0.031) than amnestic AD; there was no difference between amnestic AD + αSyn and AD in neither CSF t-tau (F(1,78) = 3.5, p = 0.057) nor Aβ₄₂ (p = 0.63). Moreover, in amnestic ADNC, we observed no differences in β-amyloid (p = 0.17) or tau burden (p = 0.2) according to Wilcoxon tests; as expected, αSyn was higher in amnestic AD + αSyn than AD (W = 89, p = 1.7e-11).

Post Hoc Analysis of Sex Effects

Because sex was a significant factor for CSF t-tau and Ng, we did a sex-based analysis (eSection 5.2, links.lww.com/WNL/C295). Linear models tested for an interaction between sex (male and female) and group (AD and AD + αSyn). There was no significant interaction for t-tau (p = 0.15) or Ng (p = 0.4).

CSF Associations With αSyn Burden in AD and AD + αSyn

Linear models tested CSF p-tau₁₈₁, t-tau, and Ng as dependent variables of postmortem αSyn within patients with ADNC (AD and AD + αSyn) while accounting for differences in tau burden (Figure 2); CSF-to-death interval and sex were also included as covariates. Greater αSyn burden was associated with lower levels of CSF p-tau181 (β = ‐0.26, SE=0.092, p = 0.0065), t-tau (β = -0.19, SE = 0.092, p = 0.041), and Ng (β=-0.21, SE = 0.071, p = 0.0044). Conversely, these models showed that greater pathological tau burden was associated with higher CSF p-tau181 (β = 0.31, SE = 0.093, p = 0.0016) and higher t-tau (β = 0.32, SE = 0.092, p = 0.00045); CSF Ng showed no association with tau burden (p = 0.74).

Figure 2. CSF and Pathology Burden in Patients With ADNC (AD, AD + αSyn).

CSF p-tau₁₈₁ (A), t-tau (B), and Ng (C) levels relate to pathologic tau (red) or amyoloid-β (red) and αSyn burden (blue). AD = Alzheimer disease; ADNC = AD neuropathologic change.

As in previous models, sex was a significant factor for t-tau (β = −14, SE = 5.3, p = 0.0066) and Ng (β = −14, SE = 3.7, p = 0.0008), but not p-tau₁₈₁ (p = 0.21). CSF to death was not significant in any model (all p > 0.1).

CSF Associations With αSyn Burden in High ADNC

To confirm CSF associations with αSyn even in high ADNC, linear models were repeated within the subset of patients with high ADNC, high β-amyloid (burden>2), and high tau (burden>2; eSection 5.3, links.lww.com/WNL/C295). Results were consistent; both CSF p-tau₁₈₁ (β = −0.26, SE = 0.12, p = 0.031) and Ng (β = −0.3, SE = 0.094, p = 0.0036) significantly declined with increasing accumulation of αSyn; t-tau was not significantly associated with αSyn burden in high ADNC (β = −0.22, SE = 0.13, p = 0.084).

Comparing Diagnostic Schemes

ROC analyses tested CSF biomarkers and established ratios when detecting ADNC in a mixed pathologic sample (Table 2). The t-tau/Aβ₄₂ ratio had the highest AUC when discriminating patients with ADNC (AD and AD + αSyn) from αSyn; CSF p-tau₁₈₁ had the worst performance.

Table 2.

ROC Analyses to Detect ADNC

Figure 3 evaluates 2 established diagnostic strategies using sample-specific thresholds (Table 2): the ATN framework and the t-tau/Aβ₄₂ ratio. Section 5.4 in the supplement discusses these classifications in pathologic detail.

Figure 3. ATN and t-tau/Aβ₄₂ Classifications.

Barplot of how each strategy classifies controls, patients with AD, patients with AD + αSyn, and patients with αSyn. ATN classifications (A): normal (A−T−N−), early Alzheimer pathologic change (A+T−N−), AD (A+T+N±), concomitant Alzheimer pathologic change and SNAP (A+T−N+), and SNAP (A−T±N±). t-tau/Aβ₄₂ Classifications (B): AD+ (≥0.31) and AD− (<0.31). AD = Alzheimer disease; αSyn = α-synuclein.

Under ATN, 12% of patients with AD/AD + αSyn were classified as normal or SNAP; 12% of patients with AD/AD + αSyn were classified as Alzheimer pathologic change (A+T−N−) despite significant tau pathology at autopsy; 60% of patients with αSyn were classified as normal; 39% of controls were classified as SNAP or Alzheimer and SNAP. Under t-tau/Aβ₄₂, 13% of patients with AD/AD + αSyn were classified as negative for AD, despite ADNC at autopsy. For αSyn, 5% were classified as positive for AD, despite low/not ADNC at autopsy. Eleven percent of controls were classified as AD positive.

Post Hoc Analyses Detecting Mixed AD and αSyn Pathology With Ng

While analytes tested are biomarkers of AD pathophysiologic processes, linear models demonstrated that they are also influenced by αSyn pathology. Thus, follow-up ROC analyses investigated whether CSF analytes could detect αSyn in a mixed cohort (eSection 5.4.1, links.lww.com/WNL/C295). When discriminating αSyn-positive patients (αSyn, AD + αSyn) from αSyn negative (AD), CSF Ng had the best accuracy (AUC = 0.83; CI 0.74–0.9; sensitivity = 0.8; specificity = 0.7).

Discussion

Increased CSF p-tau₁₈₁, t-tau, and Ng levels, as well as decreased CSF Aβ₄₂, are all established biomarkers of AD pathologic processes.^3,31,32 In this study, we show evidence that concomitant αSyn, common in AD,^9,11 is associated with lower CSF p-tau₁₈₁ and Ng, even after accounting for β-amyloid and tau burden. These data suggest that pathophysiologic processes linked to αSyn can influence CSF biomarkers in a manner distinct from ADNC. This biological influence adversely affected diagnostic sensitivity to ADNC for AD + αSyn using 2 different CSF classification strategies. At the same time, low CSF Ng levels may be associated with αSyn with or without ADNC.

Our finding that αSyn was inversely associated with CSF p-tau₁₈₁, t-tau, and Ng has important implications for diagnostic strategy. Lower CSF levels may be characteristic of LBD.^4,14 Evidence shows that CSF Aβ₄₂, p-tau₁₈₁, and t-tau are lower in patients with Parkinson disease than healthy controls at baseline.¹³ Likewise, we found that patients with αSyn had lower Aβ₄₂, t-tau and Ng than healthy controls. In accordance with lower CSF analyte concentrations associated with αSyn, we found lower p-tau₁₈₁, t-tau, and Ng in AD + αSyn than AD. We note that CSF p-tau₁₈₁ and Ng findings, but not t-tau, were robust in amnestic phenotype, sex, and high ADNC. In particular, lower CSF p-tau₁₈₁ in AD + αSyn has important implications for the ATN framework, which is increasingly applied in the AD field and, more recently, DLB.³³ Under ATN, p-tau₁₈₁ is the established CSF biomarker of T status, and a patient does not meet the definition of AD unless they are A + T+. Even using sample-specific thresholds to define ATN status, 22 (56%) patients with AD + αSyn were classified as T− despite significant tau pathology at autopsy. The t-tau/Aβ₄₂ ratio had previously been identified by our group as the best metric in patients with LBD spectrum to identify concomitant AD,⁷ and it also had the best accuracy in this study, outperforming Aβ₄₂ alone. Even so, t-tau/Aβ₄₂ too classified fewer AD + αSyn as AD positive than AD.

Despite increasing recognition that mixed pathologies are important for diagnosis and prognosis,^34-36 most current biomarker strategies, including ATN, are optimized to identify AD, not SNAP or concomitant AD and SNAP. Although ATN has a designation for concomitant pathologies (A+T−N+), a similar percentage of AD + αSyn and AD were classified as concomitant Alzheimer and SNAP (21% vs 18%), indicating that this designation does not reliably detect αSyn using the current biomarkers. ATN also classified a minority of patients with αSyn as SNAP (30%). These results highlight how alternative applications of ATN may perform better in a mixed pathology sample that include patients with non-AD pathologies. For example, markers of N that are sensitive to non-AD pathologies, such as neurofilament light chain (NfL),³⁷ might greatly improve diagnostic accuracy and precision for patients with αSyn.^38,39These results highlight the need for established biomarkers of αSyn, and there have been promising advancements in αSyn biomarker development using protein misfolding cyclic amplification and other aggregation-based assays.⁴⁰ Future work with emerging seeding amplification assays for αSyn could be incorporated into these efforts and should test ATN incorporating markers sensitive to αSyn pathophysiology. ROC results using Ng to identify αSyn were modest, but indicated that low Ng in AD might indicate mixed αSyn (AUC = 0.83). Ng is a postsynaptic protein, and although both AD and αSyn are associated with synaptic dysfunction, their synaptic CSF profiles may differ.⁴¹ Previous work has shown elevated levels of CSF Ng specific to patients with AD,^4,42 whereas studies on patients with parkinsonian disorders found decreased CSF Ng,¹⁴ hypothetically due to synaptic dysfunction and/or damage.⁴³ Special preparations of human brain tissue highlight a high burden of αSyn pathology at synapses.⁴⁴ In the future, development of in vivo αSyn markers combined with longitudinal tracking might disentangle how CSF Ng levels change with accumulating AD and αSyn pathology.

There are several important caveats to consider when interpreting results. First, we focused on end-stage disease and models accounted for the time interval from CSF to death (median = 6 years [IQR = 4.25]). Models showed no significant effect of CSF to death on analyte levels, in accordance with previous findings that CSF p-tau₁₈₁ and t-tau levels are relatively stable after dementia onset.⁴⁵ Although we statistically accounted for time interval from CSF collection to death, there may be a nonlinear rate of pathology accumulation over time. There was a substantial lag from CSF to death for pathologic alterations to occur, and thus, pathologic findings at autopsy may not fully represent biological state at CSF collection. We cannot rule out the possibility that patients had lower levels of tau at the time of CSF collection. However, we note that patients were symptomatic at CSF collection, and we would thus expect significant pathologic accumulations.⁴⁶ Longitudinal studies of CSF trajectories are needed to test for an interaction between mixed and pure pathology cases that supports our findings here. Second, this study used ordinal ratings to score β-amyloid, tau, and αSyn burden, and future work using digitized histopathologic methods may better capture the full range of pathologic accumulation.²² Likewise, we averaged burden over a limited number of brain regions (n = 10), which were sampled according to standardized methods.¹⁶ However, this sampling may not fully represent the whole-brain pathologic burden in these patients. Third, CSF levels may also differ by other factors like race,⁴⁷ and indeed, non-Whites with AD may be more likely to have concomitant αSyn.¹¹ However, with limited diversity in our sample, we were not able to explore how CSF and pathology differed by race. Likewise, we found that CSF levels differed by sex, a finding seen in other studies.⁴⁸ However, the mechanism for these differences is unknown. We did not observe a significant interaction between sex and group, although this study may be underpowered to disentangle sex effects from αSyn copathology (12 female AD + αSyn; 8 with Ng). Fourth, although we found no evidence that TDP-43 influenced CSF levels, there are different subtypes of TDP-43,¹⁹ and aggregation across multiple subtypes might obscure underlying differences. Fifth, to reduce bias that might favor one diagnostic strategy, we used sample-specific thresholds to compare ATN and t-tau/Aβ₄₂ accuracy. Therefore, diagnostic performance must still be validated in an independent sample. Sixth, we applied t-tau as the suggested CSF marker of N for ATN.³ However, ATN classification in this study would likely be improved by use of alternative N markers sensitive to non-AD degeneration, including neuroimaging markers^3,49,50 or CSF NfL.^37,39 Seventh, we excluded nonamnestic variants of AD to eliminate the confound of clinical variation in AD, which can also alter CSF levels.²⁰ Future work should test the relationship between αSyn and clinical heterogeneity on CSF levels. Finally, although this work used a well-established immunoassays,¹⁸ future work should explore these associations in newer, high-precision, second-generation immunoassays.

In summary, this autopsy study tested how concomitant αSyn pathology affects sensitivity to ADNC for CSF AD biomarkers. We find that αSyn accumulation is associated with lower CSF p-tau₁₈₁ and Ng in AD and that this can inform biomarker interpretation to improve accuracy in stratifying patients with AD from SNAP, including αSyn.

Glossary

AD

Alzheimer disease

ADNC

AD neuropathologic change

ANCOVAs

analyses of covariance

ATN

amyloid/tau/neurodegeneration

AUC

area under the curve

Aβ₄₂

β₄₂amyloid 1-42

DLB

dementia with Lewy bodies

IQR

interquartile range

LBD

Lewy body disease

MMSE

Mini-Mental State Examination

NfL

neurofilament light chain

Ng

neurogranin

PD

Parkinson disease

p-tau₁₈₁

phosphorylated tau

ROC

receiver operating characteristic

SNAP

suspected non-AD pathophysiology

TDP-43

transactive response DNA-binding protein of 43 kDa

t-tau

total tau

αSyn

α-synuclein

Appendix. Authors

Study Funding

This work is supported by funding from the National Institute of Aging (P01-AG066597, U19-AG062418, P30-AG072979, and R01-AG054519) and the Penn Institute on Aging. K.A.Q. Cousins is supported by the Alzheimer's Association Research Fellowship to Promote Diversity (AARF-D-619473) and the Rapid Program in Dementia (RAPID) Funding Grant (AARF-D-619473-RAPID). H. Zetterberg is a Wallenberg Scholar supported by grants from the Swedish Research Council (2018-02532), the European Research Council (681712), Swedish State Support for Clinical Research (ALFGBG-720931), the Alzheimer's Drug Discovery Foundation (ADDF), USA (201809-2016862), the AD Strategic Fund (ADSF-21-831376-C, ADSF-21-831381-C, and ADSF-21-831377-C), the Olav Thon Foundation, the Erling-Persson Family Foundation, Stiftelsen för Gamla Tjänarinnor, Hjärnfonden, Sweden (FO2019-0228), the European Union's Horizon 2020 research and innovation programme under the Marie Skłodowska-Curie grant agreement no 860197 (MIRIADE), and the UK Dementia Research Institute at UCL. K. Blennow is supported by the Swedish Research Council (2017-00915), the Alzheimer's Drug Discovery Foundation (ADDF), USA (RDAPB-201809-2016615), the Swedish Alzheimer Foundation (AF-742881), Hjärnfonden, Sweden (FO2017-0243), the Swedish state under the agreement between the Swedish government and the County Councils, the ALF agreement (ALFGBG-715986), the European Union Joint Program for Neurodegenerative Disorders (JPND2019-466-236), the NIH, USA (grant 1R01AG068398-01), and the Alzheimer's Association 2021 Zenith Award (ZEN-21-848495).

Disclosure

H. Zetterberg has served at scientific advisory boards and/or as a consultant for AbbVie, Alector, Eisai, Denali, Roche Diagnostics, Wave, Samumed, Siemens Healthineers, Pinteon Therapeutics, NervGen, AZTherapies, CogRx, and Red Abbey Labs, has given lectures in symposia sponsored by Cellectricon, Fujirebio, Alzecure, and Biogen, and is a cofounder of Brain Biomarker Solutions in Gothenburg AB (BBS), which is a part of the GU Ventures Incubator Program (outside submitted work). K. Blennow has served as a consultant, at advisory boards, or at data monitoring committees for Abcam, Axon, Biogen, JOMDD/Shimadzu, Julius Clinical, Lilly, MagQu, Novartis, Prothena, Roche Diagnostics, and Siemens Healthineers and is a cofounder of Brain Biomarker Solutions in Gothenburg AB (BBS), which is a part of the GU Ventures Incubator Program, all unrelated to the work presented in this manuscript. All other authors report no relevant disclosures. Go to Neurology.org/N for full disclosures.