Plasma GFAP associates with secondary Alzheimer's pathology in Lewy body disease

Abstract Objective Within Lewy body spectrum disorders (LBSD) with α‐synuclein pathology (αSyn), concomitant Alzheimer's disease (AD) pathology is common and is predictive of clinical outcomes, including cognitive impairment and decline. Plasma phosphorylated tau 181 (p‐tau181) is sensitive to AD neuropathologic change (ADNC) in clinical AD, and plasma glial fibrillary acidic protein (GFAP) is associated with the presence of β‐amyloid plaques. While these plasma biomarkers are well tested in clinical and pathological AD, their diagnostic and prognostic performance for concomitant AD in LBSD is unknown. Methods In autopsy‐confirmed αSyn‐positive LBSD, we tested how plasma p‐tau181 and GFAP differed across αSyn with concomitant ADNC (αSyn+AD; n = 19) and αSyn without AD (αSyn; n = 30). Severity of burden was scored on a semiquantitative scale for several pathologies (e.g., β‐amyloid and tau), and scores were averaged across sampled brainstem, limbic, and neocortical regions. Results Linear models showed that plasma GFAP was significantly higher in αSyn+AD compared to αSyn (β = 0.31, 95% CI = 0.065–0.56, and P = 0.015), after covarying for age at plasma, plasma‐to‐death interval, and sex; plasma p‐tau181 was not (P = 0.37). Next, linear models tested associations of AD pathological features with both plasma analytes, covarying for plasma‐to‐death, age at plasma, and sex. GFAP was significantly associated with brain β‐amyloid (β = 15, 95% CI = 6.1–25, and P = 0.0018) and tau burden (β = 12, 95% CI = 2.5–22, and P = 0.015); plasma p‐tau181 was not associated with either (both P > 0.34). Interpretation Findings indicate that plasma GFAP may be sensitive to concomitant AD pathology in LBSD, especially accumulation of β‐amyloid plaques.


Introduction
Lewy body spectrum disorders (LBSD) are a group of movement disorders that include Parkinson's disease (PD), PD with dementia (PDD), and dementia with Lewy bodies (DLB). While α-synuclein (αSyn) is the primary pathology associated with LBSD, concomitant Alzheimer's disease (AD) pathology is common. Nearly 50% of cases have significant accumulations of β-amyloid plaques and tau neurofibrillary tangles (intermediate or high AD neuropathologic change [ADNC]), justifying a secondary ADNC is relatively less severe in LBSD compared to AD, with high proportions of intermediate ADNC. 1,2 Second, cerebrospinal fluid (CSF) biomarkers of AD have an altered profile in LBSD. In particular, CSF tau phosphorylated at threonine 181 (p-tau 181 ) is inversely associated with αSyn pathology, which can reduce its sensitivity to concomitant AD. 10 Another study shows that CSF ptau 181 is significantly lower in early PD than controls, further evincing that concentrations are reduced in LBSD. 11 Thus, as plasma biomarkers for AD are developed, it becomes important to validate their efficacy to detect AD copathology in LBSD.
Two candidate AD plasma biomarkers in LBSD are plasma p-tau 181 and glial fibrillary acidic protein (GFAP). Several studies in autopsy and living participants have shown that plasma p-tau 181 can detect AD 12,13 and is associated with accumulation and spread of brain βamyloid and tau pathology. 13 GFAP is a cytoskeletal filament protein highly expressed in astrocytes, and concentrations of GFAP are raised in the CSF and blood following astrogliosis and the degeneration of astrocytes. 14 Neuroinflammation plays a key role in AD pathogenesis, 15 and plasma GFAP may be an early marker correlated with brain β-amyloid pathology, 16,17 as well as white matter disease. 18,19 Still, the majority of work has studied both analytes in the context of primary AD. Utility of these plasma biomarkers in LBSD is unclear, due to different findings across studies of living LBSD, which define groups using clinical or positron emission tomography (PET) data. [20][21][22] In light of these conflicting findings in biomarker-defined LBSD, it becomes necessary to test plasma biomarkers in autopsy cases with pathologically confirmed diagnoses.
In this autopsy study, we compare plasma p-tau 181 and GFAP in αSyn with concomitant AD (αSyn+AD) versus αSyn without AD; AD without concomitant αSyn is included as a reference group. Multivariable models test how each analyte correlates with brain accumulation of pathological β-amyloid plaques and tau neurofibrillary tangles, as well as αSyn deposition and gliosis. Receiver operating characteristic (ROC) analyses test how accurately analytes detect ADNC in this mixed pathology sample, and we also tested associations with global cognition (mini mental state exam [MMSE]).

General selection criteria
Participants were enrolled at the University of Pennsylvania (Penn) Parkinson's Disease Research Center, Frontotemporal Degeneration Center, or Alzheimer's Disease Research Center, and were selected retrospectively from the Penn Integrated Neurodegenerative Disease Database (INDD) 23 on September 15, 2022, based on eligibility criteria outlined below. The Penn Institutional Review Board approved these studies and written informed consent was obtained from each participant.
Selection criteria were clinically diagnosed LBSD with neuropathologic diagnoses of either αSyn (n = 30) or αSyn+AD (n = 19), and available biomarkers of either plasma GFAP or plasma p-tau 181 . As a neuropathologic reference group, we also examined autopsy-confirmed AD without αSyn (n = 21), with a clinical diagnosis of AD, and available plasma GFAP and p-tau 181 . If individuals had plasma collected at two or more timepoints, the last timepoint was selected to more closely reflect pathology at autopsy. There was a median interval of 2 years (interquartile range [IQR] = 2; max = 11) between plasma collection and death. Finally, we used propensity score matching 24 to select 70 clinically normal individuals as controls, matched for age and sex, with an MMSE ≥ 27; controls did not have autopsy data.

Neuropathologic diagnoses and assessments
Brains were sampled at autopsy and assessed for ADNC and αSyn, as well as for other pathologies according to standardized procedures. 23,25 ADNC was scored according to ABC criteria 26 and high or intermediate ADNC was considered AD positive; low or not ADNC was considered AD negative. Thal phase, 27 Braak stage, 28 and CERAD score 29 are reported using a 4-point scale (0-3). 26 DLB stage was assessed 30 and brainstem predominant, limbic, and neocortical Lewy bodies were all considered αSyn positive; no or amygdala-predominant Lewy bodies were considered αSyn negative. In addition to ADNC and αSyn, the presence of concomitant vascular disease 31 and TDP-43 32 was assessed; only three LBSD patients total had moderate or high vascular disease (one αSyn+AD; two αSyn) 31 and it was therefore not assessed in analyses.
Brain tissue samples were stained using immunohistochemistry as previously described, 23 and gross severity of pathological accumulations of β-amyloid, tau, αSyn, and TDP-43 were scored using a semiquantitative scale (0 = none, 0.5 = rare, 1 = minimal, 2 = moderate, and 3 = severe); in addition, severity of gliosis and cerebral amyloid angiopathy (CAA) were likewise quantified. Burden scores were the average across regions standardly sampled, 26 which included the amygdala, cingulate, CA1/ subiculum, entorhinal cortex, middle frontal gyrus, angular gyrus, superior/middle temporal gyrus, pons, and medulla. Hemisphere was randomized; if both hemispheres were sampled, the average was taken. Missing regional data were dropped to calculate the average.

Plasma analysis
Plasma was collected and assayed for p-tau 181 and GFAP previously described. 33,34 Plasma samples were analyzed on the Quanterix HD-X automated immunoassay platform. Samples were analyzed in duplicate using the Discovery kit reagents for GFAP 35 and using the V2 Advantage kit for p-tau 181 . 33 In our sample, one αSyn and one AD were missing plasma GFAP.

Clinical and demographic features
Demographic features were available through INDD. Where applicable, we examined age at onset (first reported symptom; years), age at plasma collection (years), disease duration at plasma (time from onset to plasma collection; years), interval from plasma-to-CSF (years), interval from plasma-to-MMSE (years), interval from plasma-to-death (years), and age at death (years). Sex and race were determined by self-report.
MMSE was used as a measure of global cognition in LBSD that was available in historical autopsy cases. In the LBSD autopsy sample, 24 αSyn and 16 αSyn+AD patients had MMSE available. There was a median interval of 0.6 years (IQR = 2; max = 8.2) between plasma collection and MMSE.

Statistical analyses
Not all demographic and analyte variables were normally distributed, therefore, nonparametric Kruskal-Wallis and Mann-Whitney-Wilcoxon tests performed unadjusted group comparisons for continuous variables. Fisher's exact tests performed group comparisons for categorical variables; for larger contingency tables where Fisher's exact test was not able to be computed (e.g., DLB type and clinical diagnosis), Chi-square tests were used. Spearman correlations tested within-group associations between continuous variables; given the complex and interrelated relationship of pathological variables, both nominal and Bonferronicorrected P-values are reported. All statistical models used a significance threshold of α = 0.05.
In addition to unadjusted comparisons and correlations, multiple regression was performed to control for possible confounds. 36 For linear models, 95% confidence intervals (CI) for β-estimates were reported. Plasma ptau 181 and GFAP were not normally distributed and were log-transformed in all models. Distributions of pathological and clinical measures varied and were therefore ranktransformed to perform nonparametric analyses. Effect sizes with 95% CI were calculated using generalized η 2 (η 2 G ), 37 using standard interpretation cutoffs (≥0.01 small, ≥0.06 medium, and ≥0.14 large). 38 Statistical analyses were performed and figures were generated using R version 4.1.2 (2021-11-01).
First, linear models tested how plasma biomarkers differed across αSyn and αSyn+AD, covarying for age at plasma collection, interval from plasma-to-death, and sex (Equation 1).
Within LBSD, linear models tested how AD pathological hallmarks β-amyloid plaques (Equation 2) and tau neurofibrillary tangles (Equation 3), associated with plasma GFAP and plasma p-tau 181 , covarying for age, interval from plasma-to-death, and sex. Models were repeated excluding individuals with a plasma-to-death interval >5 years.
Given correlations between β-amyloid, tau, and gliosis burden in LBSD, a post hoc model tested associations of all with plasma GFAP (Equation 4), covarying for age, interval from plasma-to-death, and sex.

Discussion
Accumulating evidence shows that plasma biomarkers, such as p-tau 181 and GFAP, are highly sensitive to AD. 12,13,40 Yet, plasma biomarkers have not been well examined in the context of LBSD with concomitant ADNC. In this autopsy study, we show consistent results that plasma GFAP is sensitive to concomitant ADNC in LBSD with autopsy-confirmed αSyn: antemortem plasma GFAP was significantly higher in αSyn+AD than αSyn, it was associated with higher burden of postmortem βamyloid (even after covarying for gliosis), and was associated with worse antemortem MMSE performance. Surprisingly, we do not find as robust results for plasma ptau 181 . Plasma p-tau 181 was not higher in αSyn+AD than αSyn. Postmortem pathological associations with βamyloid plaques and neurofibrillary tau showed a stronger association with plasma GFAP than p-tau 181 . Our findings emphasize conclusions from previous studies that LBSD-specific strategies may be necessary to detect concomitant AD, 8,10,41 and that plasma biomarkers show unique profiles that may diverge from those seen in CSF. 42,43 Our results here suggest that plasma GFAP is a promising biomarker that is sensitive to concomitant AD pathology in LBSD, and may reflect accumulation of βamyloid pathology.
We note that ADNC is typically less severe in LBSD cases than clinical AD. 1,2 In this study, 74% of αSyn+AD in this sample were intermediate ADNC (26% were high ADNC). In AD, a higher percentage were high ADNC (95%), and likewise AD cases showed significantly higher plasma p-tau 181 than αSyn+AD. When examining ADNC, Thal phase, and Braak stage, plasma p-tau 181 appeared elevated only at the highest levels (i.e., high ADNC, Thal/ Braak = 3); our findings echo other research showing plasma p-tau 181 is elevated only at more severe AD stages, 44 and is significantly lower in intermediate ADNC than high ADNC cases. 45 Likewise, studies that demonstrate good diagnostic accuracy of plasma p-tau 181 in autopsy samples have typically tested discrimination of high ADNC from intermediate/low/not ADNC, showing ROC AUCs of 0.77-0.91. 12,13,46 Longitudinal studies show that this relationship may depend on time between plasma collection and death. 40 An open question is whether other epitopes of p-tau (such as 217 or 231) are more sensitive to intermediate ADNC than p-tau 181 , especially at earlier AD stages, 13,44,45,47 including preclinical AD. 48 If so, p-tau 217 or 231 may be more useful for detecting concomitant AD in LBSD. Poor sensitivity of plasma p-tau 181 to intermediate ADNC may help explain the divergences between previous findings: one study showed no difference in plasma p-tau 181 in LBSD who were PET-Aβ positive versus PET-Aβ negative, 20 although others have found that plasma p-tau 181 does associate with PET-tau status in LBSD 21 and that it associates with cognitive decline in DLB. 22 Another possible factor is that biomarkers, like CSF and PET, can show altered profiles in LBSD compared to AD. 10,41 There is some rare autopsy work examining the effects of concomitant pathology on plasma biomarkers in primary AD 40,45 ; findings demonstrate significantly higher plasma p-tau 181 in primary AD patients with mixed pathology, including concomitant Lewy bodies, than non-AD. Future studies will be needed to disentangle these contributing factors, to test if p-tau 217 and 231 epitopes are more sensitive to intermediate ADNC than p-tau181, and test utility of different p-tau epitopes in LBSD to detect αSyn+AD.
Despite the high proportion of intermediate ADNC in αSyn+AD, our findings provide strong evidence that plasma GFAP is sensitive to AD-copathology in LBSD. A robust astrocytic response to β-amyloid plaques 49,50 may in part explain the strong links found between plasma GFAP and β-amyloid. 16,18,51 In support, we find that plasma GFAP is sensitive to pathological alterations due to concomitant AD, and is most strongly associated with postmortem β-amyloid accumulation. Associations between GFAP and β-amyloid remained robust even after covarying for colinear tau and gliosis burden. These findings are in accordance with previous work showing plasma GFAP is associated with PET-Aβ and CSF Aβ 42 / Aβ 40. 16,17 In LBSD, β-amyloid burden is often high and may have a synergic relationship with αSyn pathology, 52 which might in part explain the superior performance of plasma GFAP in this study compared with plasma ptau 181 . Our analyses also point to the clinical relevance of elevated plasma GFAP, which was associated with impaired cognition (MMSE). While this study focuses on end-stage disease in LBSD, it will be critical for future studies to explore whether our findings generalize to other neurodegenerative disease and other stages of disease: whether plasma GFAP is more sensitive to intermediate ADNC than p-tau 181 in primary AD cases, and whether plasma GFAP is elevated in early/prodromal stages of disease in AD and LBSD. Likewise, it will be important for future studies to test how GFAP changes over disease course and if it predicts future cognitive decline in LBSD.
There are several caveats to consider when interpreting our findings. First, while our focus on LBSD and AD copathology is a strength of this study, it must be noted that pathological associations with plasma biomarkers observed here may not generalize to other conditions, such as primary AD. We also acknowledge that, despite strong associations of plasma GFAP with β-amyloid, we do not find a plasma biomarker strategy that robustly identifies concomitant ADNC in LBSD: the best ROC AUC was 0.71 using plasma GFAP. Future studies should test if plasma GFAP has added value when combined with other biomarker modalities, like CSF or PET. Second, this study focused on end-stage disease, and tested how plasma levels closest to death associated with postmortem pathological accumulations. Because of this, some subjects had a substantial interval between plasma collection and death. To help account for this, models included plasma-to-death interval as a covariate and subanalyses confirmed results after excluding individuals with an interval >5 years. Still, it will be important for future studies to track longitudinal changes in plasma biomarkers within LBSD and to test plasma GFAP and ptau 181 in the context of early stage LBSD. Third, plasma concentrations in LBSD may be confounded by other factors not available in this study, including body mass index (i.e., blood volume) and creatinine (i.e., kidney functioning). 53 Fourth, we measured plasma p-tau 181 concentrations using an established platform that shows excellent performance in AD, 33,54 but did not test other isoforms of p-tau. Future studies should test if other isoforms of plasma p-tau (217 or 231) or measures from different platforms might perform differently in LBSD. Fifth, effects of race were not able to be assessed in this autopsy sample, which was majority white, and thus the generalizability of our findings are limited. It has been shown that CSF p-tau 181 levels are lower in Americans who are Black compared to White 55,56 and is an important factor for plasma as well. 12 Thus, race can be an important factor in plasma levels, and future studies should test how race affects plasma p-tau 181 and GFAP performance in LBSD. Finally, in addition to β-amyloid, plasma GFAP has been previously associated with white matter disease in AD. 18,19 We examined the possible influence of other pathologies but found no evidence that plasma GFAP was associated with CAA; this null association was also observed in a mixed pathology sample. 57 Only three LBSD autopsy patients had significant cerebrovascular disease, and we were not able to assess how GFAP and p-tau 181 levels would be altered by moderate or high vascular disease in this study.
While many studies have tested AD plasma biomarkers p-tau 181 and GFAP in the context of clinical AD, findings have not been validated in LBSD with autopsy-confirmed αSyn and AD neuropathologic diagnoses. This autopsy study focuses on LBSD to evaluate AD plasma biomarkers for detecting concomitant ADNC in αSyn cases. Analyses demonstrated that plasma GFAP was sensitive to concomitant ADNC in LBSD: plasma GFAP was higher in αSyn+AD than αSyn, was sensitive to brain β-amyloid in LBSD, and was associated with global cognition in LBSD. Together, our findings demonstrate that plasma GFAP is associated with β-amyloid and concomitant AD in LBSD.