Combined Analysis of CSF Tau, Aβ42, Aβ1–42% and Aβ1–40ox% in Alzheimer's Disease, Dementia with Lewy Bodies and Parkinson's Disease Dementia

We studied the diagnostic value of CSF Aβ42/tau versus low Aβ1–42% and high Aβ1–40ox% levels for differential diagnosis of Alzheimer's disease (AD) and dementia with Lewy bodies (DLB), respectively. CSF of 45 patients with AD, 15 with DLB, 21 with Parkinson's disease dementia (PDD), and 40 nondemented disease controls (NDC) was analyzed by Aβ-SDS-PAGE/immunoblot and ELISAs (Aβ42 and tau). Aβ42/tau lacked specificity in discriminating AD from DLB and PDD. Best discriminating biomarkers were Aβ1–42% and Aβ1–40ox% for AD and DLB, respectively. AD and DLB could be differentiated by both Aβ1–42% and Aβ1–40ox% with an accuracy of 80% at minimum. Thus, we consider Aβ1–42% and Aβ1–40ox% to be useful biomarkers for AD and DLB, respectively. We propose further studies on the integration of Aβ1–42% and Aβ1–40ox% into conventional assay formats. Moreover, future studies should investigate the combination of Aβ1–40ox% and CSF alpha-synuclein for the diagnosis of DLB.


Introduction
Reduced amyloid-β (Aβ) 42 peptide concentrations and elevated tau levels in cerebrospinal fluid (CSF) represent supportive features of Alzheimer's dementia (AD) diagnosis [1]. These biomarkers have shown their major diagnostic value in comparison of AD to controls, but overlapping values have hampered sufficient diagnostic accuracy in differentiating other kinds of dementia, especially vascular dementias and dementia with Lewy bodies [2]. The specificity of Aβ42 in the differential diagnosis of AD and other dementias could be improved by measuring the relative Aβ1-42 concentration in CSF as compared to the sum of the peptides Aβ1-40, Aβ1-38, Aβ1-37, Aβ1-39, and oxidized Aβ1-40 (Aβ1-40 ox ) as a percentage value (Aβ1-42%) [3]. Moreover, the percentage value of Aβ1-40 ox (Aβ1-40 ox %) has been proposed as a potentially helpful CSF biomarker in diagnosing DLB [3,4].
This study investigates the additional diagnostic value of these novel CSF biomarker candidates as compared to the well-acknowledged combined analysis of tau and Aβ42 in differentiating the dementias AD, DLB, and PDD. For this purpose, CSF levels of tau, Aβ42, Aβ1-42%, and Aβ1-40 ox % were determined in CSF of 45 patients with probable AD, 15 with probable DLB, 21 with PDD and 40 nondemented disease controls (NDC). Their respective diagnostic accuracies for each relevant differential diagnostic quest were analyzed.

Patients.
We investigated 121 CSF samples that were referred to our laboratory between 1999 and 2004. CSF concentrations of tau, Aβ42, Aβ1-42%, and Aβ1-40 were measured. Aliquots of these samples had been studied previously under another objective and focussing a distinct issue of differentially diagnosing dementias [3].
CSF of patients with DLB and PDD, respectively, came from the Paracelsus-Elena Klinik, Kassel, a hospital specialized in the management of movement disorders. CSF samples of AD patients and nondemented disease controls came from the memory clinic and from wards at Goettingen University.
A psychiatrist and a neurologist rendered diagnoses based on thorough clinical examination, neuropsychological assessment, clinical records, and best clinical judgment. The investigators were blinded to the neurochemical outcome measures. Investigations were carried out with the informed consent of patients or their authorized caregiver. The study was conducted under the guidelines of the Declaration of Helsinki [5] and approved by the ethics committee of the University of Goettingen and Hessen.
The nondemented disease controls consisted of two subgroups.

Depressive Cognitive Complainers.
The 25 depressive patients (16 women and 9 men) underwent lumbar puncture for differential diagnosis of cognitive complaints during the course of disease. The diagnosis of depression was made according to the criteria of DSM IV and cognitive impairment was assessed by MMSE at minimum. Patients with persistent cognitive decline for more than six months, MMSE score below 26 or clear focal atrophy in brain imaging (CT or MRI) were excluded. Age of this subgroup was 63.2 ± 10.4 years (mean ± SD).

Patients with
Alzheimer's Disease. 45 patients (27 women and 18 men) fulfilled DSM IV criteria and NINCDS-ADRDA criteria for clinical diagnosis of AD [6]. Structural (CT or MRI) or functional (SPECT or PET) brain imaging displayed global cortical atrophy, or temporal, parietotemporal, or frontotemporal focal atrophy, or marked hypometabolism of these regions.

Patients with Dementia with Lewy Bodies (DLB) and
Parkinson's Disease Dementia (PDD). Dementia with Lewy bodies (DLB, n = 15, 3 women and 12 men) was diagnosed according to the consensus criteria [7]. Patients presented with at least two core features according to the criteria and with parkinsonism less than one year before onset of dementia. Enrolled patients were hospitalized for several days to evaluate fluctuating cognition, extrapyramidal symptoms, and visual hallucinations.
Parkinson's disease dementia (PDD) was diagnosed in 21 patients (6 women and 15 men) according to UK Parkinson's Disease Society Brain Bank clinical diagnostic criteria for idiopathic Parkinson's disease and the consensus criteria [7,8]. All patients presented parkinsonism at least one year before onset of dementia.
The mean age and MMSE score of patient groups are given in Table 1.

Preanalytical Treatment of CSF.
CSF was drawn by lumbar puncture into polypropylene vials and centrifuged (1000 g, 10 min, 4 • C). Aliquots of 200 μL were kept at room temperature for a maximum of 24 hours before storage at −80 • C for subsequent Aβ-SDS-PAGE/immunoblot. The samples were not thawed until analysis. The freezers had an automatic temperature control and alarm system, so that relevant temperature changes during the time of storage can be excluded. CSF for total Aβ and tau ELISA analysis was stored at +4 • C and analyzed within two days. The protocol of preanalytical CSF handling was harmonized between the two centres of Goettingen and Kassel.

Aβ-SDS-PAGE/Immunoblot.
Aβ peptide patterns were analyzed by Aβ-SDS-PAGE/immunoblot. For separation of Aβ peptides and subsequent detection, 10 μl of unconcentrated CSF were boiled in a sample buffer for SDS-PAGE, and Aβ-SDS-PAGE/immunoblot was conducted as published elsewhere [10,11]. CSF samples of each individual patient were run as triplicates. Bands were quantified from individual blots of each patient relative to a four point dilution series of synthetic Aβ peptides using a charge coupled device camera. The detection sensitivity was 0.6 pg (Aβ1-38, Aβ1-40) and 1 pg (Aβ1-37, Aβ1-39, Aβ1-42), respectively. Signal acquisition was linear within a range of 3.8 magnitudes of order [10]. The inter-and intra-assay coefficients of variation for 20 to 80 pg of synthetic Aβ peptides were below 10% [10,11].

Statistical Analysis.
Aβ peptide and tau levels were expressed as absolute values (ng/ml). The data on Aβ peptide levels were obtained from individual blots of each patient. Aβ peptide values were determined in absolute (ng/ml) and percentage values relative to the sum of all investigated Aβ peptides (Aβ1-X%). We have characterized patient groups by mean and standard deviation (SD). The Mann-Whitney U-test was employed for comparisons of diagnostic groups. Multiple comparisons were not performed. Correlations of measured values were estimated by Spearman's Rho. The two-sided level of significance was taken as P < .05. The global diagnostic accuracies were assessed by the area under the curve (AUC) of receiver operating characteristic curve (ROC). Cutoff points were determined at the maximum Youden index [12], providing a sensitivity of ≥80%. The statistical software package SPSS, version 12.0, was used for computations.

Test
Results. The mean age of NDC was significantly younger than each of the dementia groups (P < 5 × 10 −2 ). The dementia groups did not significantly differ from each other in age. The mean MMSE score did not significantly differ between the dementia groups.

Correlations.
Analysis of each diagnostic group gave the following significant correlations. In NDC, Aβ42 and Aβ1-42% were positively correlated to each other. Higher values of Aβ1-40 ox % were correlated with male sex. Negative correlations were observed between Aβ1-42% and age as well as Aβ1-40 ox % and MMSE score. In AD, Aβ42 was positively correlated with Aβ1-42% and male sex, respectively. In PDD, Aβ42 was positively correlated with Aβ1-42%, but negatively with tau levels. No significant correlations were observed in the DLB group.

Estimates.
The results of ROC analysis for each relevant differential diagnostic testing are summarized in Table 2. Figures 1-3 show Receiver operator curves for the most relevant differential diagnostic testings.

Biomarker Patterns in the Different Dementia Groups.
In agreement with numerous previous studies, we found high levels of tau accompanied by low CSF Aβ42 levels in AD in contrast to nondemented disease controls [2,13]. In DLB and PDD, these biomarkers displayed a rather unspecific pattern: tau proteins have been found to be in 1-specificity Aβ1-40 ox % Figure 1: Receiver operator curves for detection of AD among DLB and PDD as a combined group using Aβ42/tau, Aβ1-42% and Aβ1-40 ox %, respectively. a normal range or slightly increased, paralleled by mildly to moderately decreased CSF Aβ1-42 levels [14][15][16][17][18][19]. Rises of CSF tau levels have also been detected in Creutzfeldt-Jakob Disease (CJD), vascular dementias and after acute stroke [13,20,21], indicating tau to be a sensitive biomarker for neurodestruction, but unspecific for the underlying disease process. The range of results for tau levels in DLB and PDD may result from some unexpected variance of values depending on the actual dynamic of neuronal decay at the time of lumbar puncture. Moreover, clinical diagnosis of DLB and PDD may be confounded with AD and vice versa. The selection of control groups varies among the different studies. In the present study, we compare dementia groups to diseased controls that include neurodegenerative disorders, like Parkinson's disease. This may lead to a higher overlap of CSF tau values than in studies in which healthy controls served for comparison. Especially, when taking into account that PDD may be considered as a clinical state of Parkinson's disease.
The decrease of raw CSF Aβ42 concentrations can also be found in dementias other than AD, but then often in the wake of an overall drop of CSF Aβ peptides [3,4]. In contrast, the selective decrease of the Aβ1-42 concentration as compared to constant Aβ-overall concentrations is more specific for AD [3]. In line with previous results, the diagnostic accuracy between AD and other dementias could be clearly improved International Journal of Alzheimer's Disease by scaling Aβ42 as a percentage portion of the sum of all investigated Aβ peptides (Aβ1-42%) [3]. Regarding Aβ1-40 ox %, the present study confirms our previous results of its elevated CSF levels in DLB [4]. Remarkably, Aβ1-40 ox % was only mildly elevated in AD and PDD as compared to controls, leading to a considerably smaller area of overlapping values in comparison to DLB.

Diagnostic
Accuracies for AD and DLB Using Aβ42/Tau, Aβ1-42% and, Aβ1-40 ox %, Respectively. According to the references of The Working Group on "Molecular and Biochemical Markers of Alzheimer's Disease" [22], reasonable diagnostic accuracies of the tau/Aβ1-42 ratio have been reported for detecting AD among nondemented, either healthy or diseased controls [23]. The specificity of this marker combination declined considerably down to 58% when differentiating AD from other neurodegenerative dementias, due to a large overlap of values [9]. We found similar results in the present study. In contrast, disease specific changes of CSF Aβ peptide patterns in AD and DLB enabled higher accuracies for their differential diagnosis, also in discrimination to PDD. With accuracies of 80% at minimum, low CSF levels of Aβ1-42% were the most accurate biomarker for diagnosing AD among PDD alone and in a combined group of DLB and PDD. For the differentiation of AD from DLB, Aβ1-42% and Aβ1-40 ox % yielded comparable accuracies of 80% at minimum. The differential diagnosis of DLB and PDD could be made at a sensitivity and specificity of 80% and 71%, respectively, using Aβ1-40 ox % as the most accurate biomarker. These accuracies fall within the range of the aforementioned requirements or come close to it [22].
The reason for relative Aβ peptide values being superior to raw Aβ levels include: (i) Aβ1-42, but not Aβ1-40/Aβ1-42 showed a U-shaped natural course in normal aging [24]; (ii) in contrast to absolute Aβ peptide values, the relative abundances remained largely stable after different preanalytical procedures [11,25]; (iii) referencing Aβ1-42 to Aβ1-40 avoids false positive and negative AD diagnosis in patients with constitutionally low and high CSF Aβ42 levels, respectively [26]; and (iv) dementias with low Aβ42 levels in the course of an overall decrease of CSF Aβ peptides will be sorted out from the diagnosis of AD [3]. The whole amount of CSF Aβ peptides measurable in the Aβ-SDS-PAGE/immunoblot is closely correlated to CSF Aβ1-40 levels [26]. This makes it possible to insert the ratio Aβ1-42/Aβ1-40 as a substitute for Aβ1-42%. Thus, the above considerations apply to both Aβ peptide ratios and percentage Aβ peptide values.

Conclusions.
We consider CSF Aβ42/tau to be a sensitive biomarker for detection of AD, but not specific enough for excluding other forms of dementia, like DLB and PDD. Yielding contrasts of 80% or greater, decreased CSF Aβ1-42% and elevated Aβ1-40 ox % are promising biomarker candidates for AD and DLB, respectively. However, the pathophysiological meaning of these biomarkers in the development of AD and DLB remains to be clarified.
The further progress of Aβ-peptide patterns as applicable biomarkers requires validation in independent studies on neuropathologically confirmed cases. Under this respect, we recently showed that Aβ1-40 ox % does not differ among clinically and neuropathologically defined cases of DLB [27]. The major component of Lewy bodies, α-synuclein, displayed reduced CSF levels in Parkinson's disease and DLB as compared to AD and controls [28,29]. For future studies on differentially diagnosing DLB, we propose the investigation of combined CSF α-synuclein and Aβ1-40 ox % levels. Furthermore, there is a need for translating the measurement of Aβ1-42% and Aβ1-40 ox % into more common assay formats, like ELISA [30].

Limitations of the Study.
Our results are limited by the reliance on clinical diagnosis results, because of potential misclassification. Another point of concern is the size of patient groups for DLB and PDD.