A Longitudinal Study of Total and Phosphorylated α-Synuclein with Other Biomarkers in Cerebrospinal Fluid of Alzheimer’s Disease and Mild Cognitive Impairment

Abstract

Alzheimer’s disease (AD) features a dynamic sequence of amyloid deposition, neurodegeneration, and cognitive impairment. A significant fraction of AD brains also displays Lewy body pathology, suggesting that addition of classically Parkinson’s disease-related proteins to the AD biomarker panel may be of value. To determine whether addition of cerebrospinal fluid (CSF) total α-synuclein and its form phosphorylated at S129 (pS129) to the AD biomarker panel [Amyloid-β1-42 (Aβ₄₂), tau, and phosphorylated tau (p-tau₁₈₁)] improves its performance, we examined CSF samples collected longitudinally up to 7 years as part of the Alzheimer’s Disease Neuroimaging Initiative. From 87 AD, 177 mild cognitive impairment (MCI), and 104 age-matched healthy controls, 792 baseline and longitudinal CSF samples were tested for total α-synuclein, pS129, Aβ₄₂, tau, and p-tau₁₈₁. pS129, but not total α-synuclein, was weakly associated with diagnosis at baseline when t-tau/Aβ₄₂ was included in the statistical model (β=0.0026, p=0.041, 95% CI [(0.0001)–(0.005)]). CSF α-synuclein predicted Alzheimer’s Disease Assessment Scale-Cognitive (β=−0.59, p=0.0015, 95% CI [(−0.96)–(−0.23)]), memory (β=0.4, p=0.00025, 95% CI [(0.16)–(0.59)]) and executive (0.62, <0.0001, 95% CI [(0.31)–(0.93)]) function composite scores, and progression from MCI to AD (β=0.019, p=0.0011, 95% CI [(0.002)–(0.20)]). pS129 was associated with executive function (β=−2.55, p=0.0085, 95% CI [(−4.45)–(−0.66)]). Lower values in the mismatch between α-synuclein and p-tau₁₈₁ predicted faster cognitive decline (β=0.64, p=0.0012, 95% CI [(0.48)–(0.84)]). Longitudinal biomarker changes did not differ between groups, and may not reflect AD progression. The α-synuclein-p-tau₁₈₁-Mismatch could better predict longitudinal cognitive changes than classical AD markers alone, and its pathological correlates should be investigated further.

INTRODUCTION

Alzheimer’s disease (AD), the most common cause of dementia, is a progressive degenerative disorder that affects over 35.5 million individuals worldwide, and this number is expected to double in the next 20 years[1, 2]. The socioeconomic burden on patients, families and healthcare systems highlights the urgent need to understand the disease process, and to develop disease-modifying treatments. The molecular pathology probably starts 10–20 years before patients develop dementia symptoms sufficient to prompt clinical diagnosis[3]. However, diagnosis of both AD and mild cognitive impairment (MCI), a condition with elevated risk for progressing to AD and thought to be a prodromal stage, remains based on clinical examination, neuroimaging, and, to a limited extent, cerebrospinal fluid (CSF) markers[4, 5]. It is likely, therefore, that accurately predicting AD or MCI will require diagnosing patients in earlier stages, when interventions are more likely to preserve neuron function; identification of markers capable of predicting imminent progression of MCI to AD is a question of significant interest.

The most frequently used CSF panel of biomarkers in AD includes amyloid beta 1–42 (Aβ₄₂), a predominant component of amyloid plaques; total tau (t-tau) and hyperphosphorylated tau (p-tau), which are associated with neuronal damage and intracellular neurofibrillary tangles[6]. Low CSF Aβ₄₂ in individuals with MCI likely indicates the presence of AD amyloid pathology, and predicts rapid progression to AD[7, 8]. Other studies have shown that t-tau and p-tau can also predict MCI[9–11]. Although Aβ and tau proteins are postulated to play a central role in AD pathogenesis, the pathophysiology is complex, leading to variability in clinical presentation in both AD and MCI[12, 13]. Understanding these factors is critical for successful early detection and intervention.

Contributing to the variability of AD is the occurrence of comorbid Lewy body (LB) pathology, the hallmark of another family of neurodegenerative diseases, including Parkinson’s disease (PD) and Dementia with Lewy bodies (DLB), in up to 50% of AD patients[9, 14]. α-Synuclein (α-syn), the major component of LBs, has also been found to play an important role in AD[15–17]. As a biomarker, CSF α-syn has been reported as unchanged or slightly increased in AD[18, 19]. Our previous study showed a strong association between CSF α-syn, t-tau and tau phosphorylated at threonine 181 (p-tau₁₈₁), and confirmed the utility of α-syn in improving diagnostic sensitivity and specificity provided by Aβ₄₂, t-tau and p-tau₁₈₁[17]. Intriguingly, our previous study identified a subpopulation of AD patients in which the level of α-syn was lower than expected, given the normally high correlation of α-syn with tau. We hypothesized that this may reflect the subpopulation of AD subjects with concomitant LB pathology, and found that the mismatch (α-syn–p-tau₁₈₁-Mis) between these two markers both correlated with worse cognitive outcomes, and improved the association of the t-tau/Aβ₄₂ ratio with these measures when included in the model [17]. Importantly, most α-syn present in LBs is highly phosphorylated, especially at serine 129 (pS129) [15, 20–22], which is thought to alter aggregation and toxicity of α-syn[23, 24]. Notably, the CSF pS129 was significantly higher in PD patients than in healthy controls, and correlated with PD severity[16, 25], suggesting it may serve as a marker of PD pathology independently of total α-syn, which tends to be lower in the CSF of PD patients[18, 26–29]. However, the roles of these biomarkers in AD have not yet been completely probed, particularly considering pS129 in large, longitudinal cohorts.

The present study was designed to measure the levels of CSF total α-syn and pS129, and to examine their association with CSF t-tau, p-tau₁₈₁ and Aβ₄₂, in a large longitudinal cohort of AD, MCI, and cognitively normal (CN) subjects recruited as part of the Alzheimer’s Disease Neuroimaging Initiative (ADNI) study. This large cohort includes the longest follow-up reported to date, including additional sample collection and clinical assessment continuing up to 7 years. The aim of this study was to investigate whether analysis using a multiple biomarker panel including total α-syn and pS129, along with the standard AD-related CSF biomarkers, could provide a feasible tool for monitoring progression in people with AD and MCI.

MATERIALS and METHODS

Subjects and samples

CSF samples of subjects with clinical AD or MCI, and age-matched, healthy CN controls were obtained from the original ADNI. For the current study, a total of 792 CSF samples were longitudinally collected from 368 subjects, including 87 patients with AD, 177 patients with MCI, and 104 age-matched CN. Clinical data and values for t-tau, p-tau₁₈₁, Aβ₄₂, and APOE status used in the preparation of this article were obtained from the ADNI database (http://adni.loni.usc.edu). The ADNI was launched in 2003 as a public-private partnership, led by Principal Investigator Michael W. Weiner, MD. The primary goal of ADNI has been to test whether serial magnetic resonance imaging (MRI), positron emission tomography (PET), other biological markers, and clinical and neuropsychological assessment can be combined to measure the progression of MCI and early AD. Briefly, all subjects underwent evaluations consisting of medical history, physical, psychiatric and neurological examinations, and APOE genotyping. Cognitive performance was evaluated with Mini-Mental State Examination (MMSE), Alzheimer’s Disease Assessment Scale-Cognitive (ADAS-Cog13), and memory[30] and executive[30] composite scores at screening and yearly follow-up visits (MCI subjects had additional evaluations at 6 and 18 months). Spatial Pattern of Abnormality for Recognition of Early Alzheimer’s disease (SPARE-AD) was used as a composite MRI measure that reflects AD-like brain atrophy. The inclusion and exclusion criteria for healthy controls and patients with AD and MCI have been described previously, and can be found on the ADNI website (http://adni.loni.ucla.edu/wp-content/uploads/2010/09/ADNI GeneralProceduresManual.pdf).

Baseline lumbar puncture was performed within 28 days of the screening visit for all subjects. Follow up samples were collected at 12 (266 cases), 18 (1 case), 24 (81 cases), 30 (1 case), 48 (56 cases), 72 (18 cases) and 84 (1 case) months after screening. Because only a single subject had follow up at 84 months, studies were performed with and without inclusion of this subject; no changes to the outcomes were observed.

Luminex assays

Total α-syn and pS129 levels were detected by Luminex assays as previously described[25, 26]. The sensitivity (lower limit of quantification) was 0.01 ng/ml for the total α-syn assay, and 0.03 ng/ml for pS129. The inter-assay precision (run-to-run or plate-to-plate, determined with an internal standard—an aliquot of a pooled reference CSF—in each plate) was <20% for both assays, and the intra-assay precision was <10%.

Aβ₄₂, t-tau, and p-tau₁₈₁ were measured using the multiplex xMAP Luminex platform (Luminex Corp, Austin, TX) with Innogenetics (INNO-BIA AlzBio3; Ghent, Belgium) immunoassay kit–based reagents, as previously described[31, 32]. The data was obtained from the ADNI database (http://adni.loni.usc.edu).

CSF hemoglobin (Hgb) assays

Hgb levels were measured as previously described[26] with a human Hgb ELISA quantitation kit from Bethyl Lab Inc (Montgomery, TX, USA), according to the manufacturer’s instructions. Only samples with CSF Hgb < 200 ng/mL were used for total α-syn analysis, as previously evaluated and selected based on the association of high CSF Hgb and α-syn values[17].

Statistical analysis

Statistical analyses were performed with Prism 4.0 (GraphPad) and R v. 3.3.3[33]. Biomarker values with a right-skewed distribution underwent logarithmic (t-tau, p-tau₁₈₁ and t-tau/Aβ₄₂ ratio) and square-root (α-syn) transformations. Correlations between biomarkers are reported as Pearson correlation coefficients. For univariate analyses presented in Table 1 the following tests were applied: 1) one way ANOVAs for quantitative normally distributed variables, 2) Kruskall-Wallis and rank-based two way methods for non-normally distributed quantitative variables and 3) Chi-square tests for qualitative variables.

Table 1.

Demographic and clinical characteristics of the ADNI subjects at baseline

Variables	CN (n=104)	MCI (n=177)	AD (n=87)	p-value
Age	75.7 (72–78.5)	74.7 (69.4–79.4)	75.8 (70.1–80.1)	0.53
Gender N (% male)	52 (50%)	117 (66%)	49 (56%)	0.024
APOE ε4 N (% present)	25 (24%)	93 (53%)	60 (69%)	<0.0001
MMSE	29 (29–30)	27 (25–28)	24 (22–25)	<0.0001
ADAS-Cog13	9.67 (6.58–12.67)	18.67 (14.58–23.33)	28.84 (23.08–34)	<0.0001
Aβ1-42	222 (162–254.5)	146 (126.5–207.5)	137 (118.5–159)	<0.0001
T-Tau	61 (49–86)	86 (65–120)	111 (81–154)	<0.0001
P-Tau181	21 (16–29.5)	31 (20–45)	36 (29–49)	<0.0001
α-Syn(ng/ml)	0.52 (0.44–0.68)	0.6 (0.48–0.74)	0.57 (0.48–0.73)	0.16
α-Syn excluding subjects with high Hgb	0.48 (0.41–0.59)	0.54 (0.45–0.65)	0.54 (0.43–0.67)	0.031
pS129-α-syn(ng/ml)	0.075 (0.062–0.083)	0.069 (0.061–0.081)	0.072 (0.065–0.083)	0.080
Hemoglobin	91.33 (44.08–704.25)	86.86 (41.91–836.65)	90.83 (42.17–648.86)	0.91
“α-Syn-p-tau181-Mis excluding subjectswith high Hgb	0.072 [(−0.44)–(0.68)]	0.041 [(−0.62)–(0.60)]	−0.29 [(−0.80)–(0.41)]	0.053
SPARE-AD	−1.46 [(−2.05)–(−0.99)]	0.78 [(0.15)–(1.45)]	1.34 [(0.79)–(1.72)]	<0.0001

Cross-sectional association of CSF biomarkers with diagnosis was determined using a multivariable regression model adjusted for age, diagnosis, gender and APOE ε4 presence, both with and without t-tau/Aβ₄₂ ratio included as a covariable. Distribution of residuals and absence of multicolinearity was tested in these models. Receiver operating characteristic (ROC) curves were generated to compare diagnostic groups based on a logistic regression model, which included age, gender and each of the three α-syn CSF biomarker measures. For prediction of longitudinal clinical changes (ADAS-Cog13, memory[30] and executive[34] function composite scores) of MCI subjects based on baseline CSF biomarker values, a mixed-effects model including age, gender, education and APOE ε4 presence was used with random intercept and slope. An interaction between α-syn and pS129 with time was included in these models, as the analysis main outcome. A significant association would indicate that rates of progression are associated with baseline biomarker values. We also analyzed models that added the t-tau/Aβ₄₂ ratio * time interaction to the previously mentioned covariates to evaluate if these effects were independent and persisted once the tau/Aβ₄₂ ratio was included in the model.

The mismatch between the expected α-syn levels for a given p-tau₁₈₁ level (α-syn-p-tau₁₈₁-Mis) was calculated as the standardized residual of the linear regression model that predicted α-syn based on p-tau₁₈₁, therefore estimating if α-syn is lower or higher than expected based on the p-tau₁₈₁ level.

To evaluate the association of baseline diagnosis with longitudinal changes in α-syn, pS129, and α-syn-p-tau₁₈₁-Mis[17] during the follow up a mixed-effects model including age, gender, and APOE ε4 presence was used with random intercept and slope. An interaction between baseline diagnosis with time was included in these models, as the analysis main outcome. A significant association would indicate that longitudinal changes in α-syn and pS129 are associated with baseline diagnosis. A Cox hazards model, with age, gender, APOE ε4 presence, education and tau/Aβ₄₂ ratio, as covariates, was used to study the conversion of MCI to AD for the different CSF biomarkers studied here. Concentration of α-syn was included in the model as a continuous variable. The studied CSF biomarkers were α-syn, pS129, and α-syn-p-tau₁₈₁-Mis. Statistical tests were two-sided and significance was set at p<0.05.

RESULTS

Demographic and clinical characteristics of the ADNI subjects

A total of 368 ADNI subjects were included in the current study. Clinical and demographic characteristics of the subjects are summarized in Table 1. Gender, MMSE, ADAS-Cog13, t-tau, p-tau₁₈₁, Aβ₄₂ and APOE ε4 differed between the diagnostic groups at baseline. Similar to previous studies of the ADNI cohort[17], 40.5% (n=149) of the baseline samples had Hgb levels exceeding 200 ng/ml. As previous studies have demonstrated, CSF α-syn correlates with Hgb, suggesting a strong influence of blood contamination on the CSF level[26], and subjects with Hgb levels exceeding 200 ng/ml were thus excluded from analyses including α-syn values. During the course of follow up, 53.1% subjects who were classified as MCI at baseline converted to AD diagnosis (median follow-up 3 years, IQR 2.1-6.0 years).

Correlation among CSF biomarkers

CSF α-syn showed a moderate positive correlation with t-tau (r=0.63, p<0.0001), p-tau₁₈₁ (r=0.54, p<0.0001), and a weak inverse correlation with Aβ₄₂ (r=−0.15, p=0.026). CSF pS129 was moderately inversely correlated with p-tau₁₈₁ (r=−0.29, p<0.0001) and α-syn (r=−0.38, p<0.0001) and weakly inversely correlated with t-tau (r=−0.16, p=0.0031) (Fig. 1).

Fig. 1.

Correlation among CSF biomarkers. (A) correlation matrix with Pearson correlation coefficient values between all CSF biomarker pairs (red shading indicates positive correlation and blue shading negative correlation). (B-G) Scatterplots depicting values of the different biomarker pairs. Blue lines represent LOESS lines (LOcal regrESSion).

Association of CSF biomarkers with baseline diagnosis and longitudinal clinical changes

Cross-sectional values of total α-syn, pS129, and α-syn-p-tau₁₈₁-Mis by diagnostic group are presented in Fig. 2. A multivariable regression model was used to examine the cross-sectional association of CSF analytes with baseline clinical diagnostic group. Total α-syn was not associated with diagnosis at baseline, unless APOE4 status was excluded from the model (Supplemental Table 1). pS129 was weakly associated with AD clinical diagnosis at baseline when t-tau/Aβ₄₂ was included in the statistical model (Table 2). No other results differed when APOE4 status was excluded from the model. In the subset of subjects with Hgb <200 ng/ml (the same subjects examined for α-syn), pS129 was weakly associated with MCI when tau/Aβ₄₂ was excluded from the model (Supplemental Table 2). Cross-sectional values overlapped substantially, and performed poorly in differentiating diagnostic groups (Fig. 3).

Fig. 2.

Cross-sectional values of biomarkers by diagnosis group. Baseline biomarker values are reported for (A) total α-syn, (B) pS129, and (C) α-syn-p-tau₁₈₁-Mis for control, MCI, and AD subjects.

Table 2.

Cross-sectional association of CSF biomarkers with clinical diagnosis

	Adjusted for T-tau/Aβ42	MCI		AD
		Coefficients	p-values	Coefficients	p-values
α-Syn	No	0.043 [(−0.02)–(0.10)]	0.17	0.048 [(−0.021)–(0.12)]	0.18
	Yes	−0.039 [(−0.09)–(0.013)]	0.14	−0.056 [(−0.12)–(0.005)]	0.32
pS129- α-syn	No	−0.0012 [(−0.003)–(0.0008)]	0.23	0.00077 [(−0.002)–(0.003)]	0.51
	Yes	0.00001 [(−0.002)–(0.002)]	0.99	0.0026 [(0.0001)–(0.005)]	0.041
“α-Syn-p-tau181-Mis	No	−0.18 [(−0.51)–(0.14)]	0.27	−0.31 [(−0.69)–(0.067)]	0.11
	Yes	−0.19 [(−0.53)–(0.15)]	0.28	−0.32 [(−0.73)–(0.085)]	0.12

Linear regression model is adjusted for age at baseline, gender, and APOE ε4 presence. Models with and without T-tau/Aβ_1-42 ratio are listed as indicated in the second column.

Fig. 3.

ROC curves for separation of subjects by diagnosis for each biomarker. Cross-sectional (A) total α-syn, (B) pS129, and (C) α-syn–p-tau₁₈₁-Mis performed poorly in distinguishing diagnoses.

In the longitudinal analysis using a mixed-effects model, including MCI subjects with up to 7 yearly follow-up visits, CSF α-syn at baseline did not predict any clinical outcome when the t-tau/Aβ₄₂ ratio was excluded from the model. A similar model including all subjects (with all diagnoses), and with diagnosis as a covariate, showed similar results (data not shown). However, when t-tau/Aβ₄₂ ratio was included in the model, CSF α-syn predicted ADAS-Cog13, memory and executive function composite scores and MCI to AD progression. pS129 was associated with executive function composite score when including t-tau/Aβ₄₂ ratio in the model, for both the whole cohort (Table 3), and the subset of subjects with Hgb<200 ng/ml (Supplemental Table 3). No conclusions were altered by excluding APOE4 status from the model (Supplemental Table 4).

Table 3.

Association of baseline CSF biomarker values with longitudinal clinical changes in MCI subjects. Reported as multivariate regression coefficient [95% confidence interval of coefficient] (p-value).

	Adjusted for T-tau/Aβ1-42*time	ADAS-Cog13	Memory composite	Executive function composite	MCI progression to AD
T-tau/Aβ1-42	–	0.32 [(0.26)–(0.38)] (0.0002)	−0.20 [(−0.25)–(−0.15)] (<0.0001)	−0.30 [(−0.37)–(−0.23)] (<0.0001)	3.51 [(1.68)–(7.33)] (0.0008)
α-Syn	No	−0.12 [(−0.46)–(0.22)] (0.48)	0.085 [(−0.12)–(0.29)] (0.42)	0.17 [(−0.13)–(0.47)] (0.26)	0.53 [(0.79)–(3.47)] (0.50)
	Yes	−0.59 [(−0.96)–(−0.23)] 0.0015)	0.4 [(0.16)–(0.59)] (0.00025)	0.62 [(0.31)–(0.93)] (<0.0001)	0.019 [(0.002)–(0.20)] (0.0011)
pS129- α-syn	No	0.42 [(−2.08)–(2.93)] (0.74)	−0.19 [(−1.82)–(1.45)] (0.82)	−1.66 [(−3.69)–(0.36)] (0.11)	0.074 [(0.00)–(27423)] (0.69)
	Yes	1.45 [(−0.89)–(3.80)] (0.22)	−0.93 [(−2.44)–(0.58)] (0.23)	−2.55 [(−4.45)–(−0.66)] (0.0085)	55.76 [(0.00)–(445985)] (0.56)
“α-Syn-p-tau181-Mis	No	−0.070 [(−0.094)–(−0.046)] (0.0018)	0.050 [(0.036)–(0.064)] (0.0002)	0.070 [(0.050)–(0.090)] (0.00021)	0.64 [(0.48)–(0.84)] (0.0049)
	Yes	−0.065 [(−0.11)–(−0.02)] (0.0038)	0.042 [(0.02)–(0.07)] (0.0012)	0.066 [(−0.028)–(0.10)] (0.00061)	0.64 [(0.49)–(0.84)] (0.0012)

Model is adjusted for age at baseline, gender, education, and APOE 4 presence. Models with and without T-tau/Aβ_1-42 ratio are listed as indicated in the second column.

While α-syn correlated significantly with p-tau₁₈₁, some patients have high p-tau₁₈₁ with quite low α-syn. In a previous study[17], we hypothesized that these subjects might represent a subset of AD subjects with concomitant LB pathology. We calculated the α-syn-p-tau₁₈₁-Mis value, in which a value of 0 reflects α-syn values identical to those expected for that subject’s p-tau₁₈₁ level, while positive values reflect higher than expected α-syn and negative values reflect lower than expected α-syn. We tested the association between α-syn-p-tau₁₈₁-Mis and cognitive decline in AD and MCI subjects, and found that lower α-syn–p-tau₁₈₁-Mis values with or without t-tau/Aβ₄₂ as a covariate were associated with a faster decline in ADAS-Cog, memory and executive function composite (Fig. 4 A-C), and predicted faster progression from MCI to AD dementia (Fig. 4D).

Fig. 4.

α-syn–p-tau₁₈₁-Mis is associated with AD progression. A-C) Association of baseline α-syn–p-tau181-Mis with clinical progression during follow-up. Modeled progression by quartile shows faster progression for those with lower (i.e., more negative) α-syn–p-tau₁₈₁-Mis values. D) Association of the α-syn–p-tau181-Mis with progression from MCI to AD. Cox hazard model results were calculated using quantitative values, but for graphical representation we present results for α-syn–p-tau181-Mis tertiles.

Longitudinal associations of CSF biomarkers with clinical measures

Longitudinal changes in CSF α-syn, pS129, and α-syn–p-tau₁₈₁-Mis values did not differ between subjects designated CN, MCI, and AD at baseline. When baseline t-tau/Aβ₄₂ was considered, no interacting effect with longitudinal changes in any of the markers was observed, indicating that the baseline t-tau/Aβ₄₂ ratio does not predict progression of α-syn in CN, MCI, or AD subjects (data not shown). We next considered whether longitudinal changes in CSF total α-syn, pS129, or α-syn–p-tau₁₈₁-Mis might differ between MCI subjects who progressed to clinical AD over the course of follow-up, compared to those whose diagnosis remained stable. Among the 133 subjects included in the pS129 model, 51 subjects were stable and 82 progressed, while 33 of the 77 subjects with Hgb<200 ng/ml who were included in the total α-syn and α-syn–p-tau₁₈₁-Mis analyses remained stable over follow up (data not shown). No differences in the longitudinal progression of total α-syn, pS129 or α-syn–p-tau₁₈₁-Mis were observed between these groups (data not shown). Results did not change when the single subject with 84 months follow-up was excluded (data not shown).

DISCUSSION

Up to 50% of patients with AD dementia also have LB pathology at autopsy[9, 14]. Both the mechanisms and implications of this finding remain incompletely defined, but suggestions that neurodegeneration-related proteins may interact, and that α-syn, Aβ and tau may mutually accelerate their accumulation and aggregation, are abundant[19, 35]. Both AD and PD feature impaired cellular mechanisms for clearing abnormal proteins, contributing to the burden of pathological proteins and their deleterious effects on neurons[36–38]. There is also evidence that intra-axonal α-syn and intraneuronal tau aggregates are associated with axonopathy and cellular dysfunction[39].

Translation of these findings to usefulness as biomarkers has been challenging. Patients with AD or MCI show increased CSF levels of α-syn[12, 17, 18], perhaps due to release from damaged neurons[40, 41], as has been hypothesized for the increased levels of CSF tau in AD. Fewer studies have examined α-syn across the transition from MCI to AD dementia; however, in one longitudinal study, CSF α-syn was higher in MCI subjects with a shorter duration of symptoms, but did not differ across diagnostic groups[42]. Here, we also find a significant difference in CSF α-syn levels between CN, AD and MCI. However, a biomarker that only correlated with existing biomarkers, such as total or p-tau, would be of limited use. Moreover, the cross-sectional measurements of markers overlapped substantially between groups, and perform only poorly in differentiating AD from control. Our previous finding that baseline α-syn–p-tau₁₈₁-Mis was associated with a faster decline in cognitive function provided intriguing potential for a novel marker, which might distinguish subgroups of faster vs slower progressing patients. We have further expanded this analysis, demonstrating that lower α-syn–p-tau₁₈₁-Mis is associated both with faster progression of cognitive decline and conversion from MCI to AD. Notably, low α-syn–p-tau₁₈₁-Mis indicates an α-syn level that is lower than expected, given the high correlation of α-syn and tau in AD. Because CSF α-syn is generally found to be lower in PD patients, in contrast to the unchanged or higher levels in AD patients in most well-controlled studies, we hypothesized that a lower α-syn–p-tau₁₈₁-Mis value may serve as a marker for those AD subjects with concomitant LB pathology. We initially attempted to seek support for this hypothesis by examining the few ADNI cases with pathological analysis available, and found that of the two available at the time, both showed LB pathology and had negative α-syn–p-tau₁₈₁-Mis values. Since then, a small number of additional cases have come to autopsy; however, too few cases exist to perform reliable statistical analyses. Definitive confirmation of this hypothesis will require a complete cohort with matched CSF and histopathological analysis, and thus cannot be performed until such samples become available. Moreover, association of biomarker and histopathology also has limitations as a technique for examining the underlying correlations between brain and biofluid levels of analytes of interest, as substantial time may elapse between biofluid sampling and collection of tissue at autopsy. Thus, even once such samples become available, these studies must be complemented by further strategies such as mechanistic studies using in vitro and in vivo models.

Hyperphosphorylation is a feature of abnormal tau in AD and of α-syn in PD (most α-syn in LBs is pS129) [15, 20, 21, 29]. CSF pS129 has been associated with PD[15, 20–22], but its role in AD remains unclear, despite the frequency of LB pathology in AD brains. CSF total α-syn and pS129 appear to behave differently[43]; indeed, in the current study, pS129 correlated negatively with total α-syn, suggesting that differing, potentially competing processes may determine the compartments in which each form of α-syn is located in AD, including potential differences between pure AD and mixed pathology underlying the clinical symptoms. However, in this study, pS129, unlike total α-syn, predicted clinical outcomes only for executive function impairment. These findings are consistent with the clinical observation of faster cognitive decline in the LB variant of AD or AD with concomitant LBD[44, 45], and suggest that coincident pathologies might trigger cognitive deficits at lower thresholds[46, 47]. Intriguingly, a major distinction between cognitive impairment most often observed in PD compared to AD is the greater frequency of deficits in executive function rather than in memory[48]. That pS129 is associated only with this type of impairment may hint at its role in a mechanism specific to a particular pathway of cognitive decline. The hypothesis that pS129 may exacerbate cognitive deficits in AD through interactions between α-syn, pS129, Aβ and tau is well supported experimentally [49, 50] and clinically, with the finding that AD type pathology is also frequent in patients with PD, PD dementia and DLB [47, 51]. However, while this association is plausible given the known pathophysiology of α-syn protein, the association in our study was not highly significant, and further studies in additional cohorts should be performed to validate this finding.

A major strength of the current study is that prolonged follow up times available for many subjects (up to 7 years), allow us to examine the natural course of progression in total α-syn, pS129 or the α-syn–p-tau₁₈₁-Mis value. However, none of the biomarkers differed in longitudinal progression based on baseline diagnosis. We also considered whether α-syn progression in CSF may be associated with the conversion from MCI to AD. We observed no differences between subjects who remained stably within the MCI diagnosis group compared to those who converted to AD. Thus, changes in α-syn may be unlikely to serve as biomarkers for the progression of AD, whether because they occur earlier in the course of the disease, or because they do not follow a simple trajectory. Nonetheless, as discussed above, α-syn does appear to be altered in AD, and further examination of its changes, perhaps starting even earlier in the disease process, should be examined.

In interpreting this study, several considerations were made to minimize the effects of confounding variables. CSF α-syn is influenced by blood contamination, prompting the use of Hgb to exclude contaminated samples. A relatively large proportion of ADNI CSF samples contained significant Hgb [12, 17]; their exclusion decreases the practical sample size, but ensures the quality of the data set. Cohort heterogeneity may also influence the outcomes, as some of those diagnosed with AD lack biochemical markers of AD pathology (that is, they have t-tau/Aβ₄₂ within the range of CN subjects) [52, 53]. Although the ADNI cohort is better characterized than most, substantial variability among those with AD symptoms remains. However, while this complicates characterization of longitudinal changes in biochemical markers, it also allows the identification of subsets within the clinical category, as we hypothesize in subjects with low α-syn–p-tau₁₈₁-Mis scores. While complete understanding of such subcategories is not possible for studies such as this, they can be useful for hypothesis generation to be tested at later times.

In conclusion, our studies confirm a strong association between CSF α-syn and t-tau and p-tau, as well as the association of α-syn and α-syn–p-tau₁₈₁-Mis with longitudinal clinical outcomes. α-syn-p-tau₁₈₁-Mis subjects represent a subgroup of AD that may show more rapid clinical progression, potentially due to LB pathology in these subjects. pS129 may be useful to predict deficits in executive function, and the distinctiveness of its behavior from that of total α-syn suggests differences in the underlying biology of α-syn isoforms. Further longitudinal studies will be important to uncover and understand additional CSF factors that reflect individual disease progression, and they may aid in the development of novel neuroprotective therapies.