Improved Differential Diagnosis of Alzheimer's Disease by Integrating ELISA and Mass Spectrometry-Based Cerebrospinal Fluid Biomarkers

Abstract

Background: Alzheimer's disease (AD) is diagnosed based on a clinical evaluation as well as analyses of classical biomarkers: Aβ42, total tau (t-tau), and phosphorylated tau (p-tau) in cerebrospinal fluid (CSF). Although the sensitivities and specificities of the classical biomarkers are fairly good for detection of AD, there is still a need to develop novel biochemical markers for early detection of AD.

Objective: We explored if integration of novel proteins with classical biomarkers in CSF can better discriminate AD from non-AD subjects.

Methods: We applied ELISA, mass spectrometry, and multivariate modeling to investigate classical biomarkers and the CSF proteome in subjects (n = 206) with 76 AD patients, 74 mild cognitive impairment (MCI) patients, 11 frontotemporal dementia (FTD) patients, and 45 non-dementia controls. The MCI patients were followed for 4-9 years and 21 of these converted to AD, whereas 53 remained stable.

Results: By combining classical CSF biomarkers with twelve novel markers, the area of the ROC curves (AUROCS) of distinguishing AD and MCI/AD converters from non-AD were 93% and 96%, respectively. The FTDs and non-dementia controls were identified versus all other groups with AUROCS of 96% and 87%, respectively.

Conclusions: Integration of new and classical CSF biomarkers in a model-based approach can improve the identification of AD, FTD, and non-dementia control subjects.

Keywords: Alzheimer’s disease; ELISA; cerebrospinal fluid; mass spectrometry; mild cognitive impairment; proteomics.

Fig.1

Alzheimer’s disease classification criteria, as reported by Hansson et al. [15]. The dashed lines represent cutoff levels based on Aβ₄₂ <530 (ng/L), t-tau>350 (ng/L), and p-tau> = 60 (ng/L).

Fig.2

Comparison of AUROCs between the classical model (ELISA measurements of Aβ₄₂, t-tau, p-tau) and the integrative model (ELISA measurements of Aβ₄₂, t-tau, p-tau in combination with MS-based measurements of 12 proteins). AD, Alzheimer’s disease; MCI, mild cognitive impairment; FTD, frontotemporal dementia.

Fig.3

Variable importance extracted from the sPLS-DA model trained on a model of proteins (MS) and Aβ₄₂, t-tau, and p-tau. The model selected the proteins with the most influence on the responses resulting in a total of 15 unique variables including Aβ₄₂, t-tau, and p-tau. Aβ₄₂ (VIP = 6.80), t-tau (VIP = 4.29), p-tau (VIP = 3.84), cadherin-2 (VIP = 3.68, Uniprot AC: P19022, Uniprot ID: CADH2), neurosecretory protein VGF (VIP = 3.49, Uniprot AC: O15240, Uniprot ID: VGF), afamin (VIP = 2.41, Uniprot AC: P43652, Uniprot ID: AFAM), plasma protease C1 inhibitor (VIP = 2.38, Uniprot AC: P05155, Uniprot ID: IC1), inter-alpha-trypsin inhibitor heavy chain H4 (VIP = 2.01, Uniprot AC: Q14624, Uniprot ID: ITIH4), apolipoprotein A-I (VIP = 1.75, Uniprot AC: P02647, Uniprot ID: APOA1), secretogranin-2 (VIP = 1.47, Uniprot AC: P13521, Uniprot ID: SCG2), beta-Ala-His dipeptidase (VIP = 1.15, Uniprot AC: Q96KN2, Uniprot ID: CNDP1), alpha-1B-glycoprotein (VIP = 0.58, Uniprot AC: P04217, Uniprot ID: A1BG), chitinase-3-like protein 1 (VIP = 0.5, Uniprot AC: P36222, Uniprot ID: CH3L1, also known as YKL-40), cystatin-C (VIP = 0.43, Uniprot AC: P01034, Uniprot ID: CYTC) and SPARC (VIP = 0.15, Uniprot AC: P09486, Uniprot ID: SPRC).

Fig.4

Boxplots of CSF levels of the analyzed proteins. The levels were compared between the different groups with nonparametric statistical testing. A: Aβ₄₂, B: t-tau, C: p-tau, D: cadherin-2 (Uniprot AC: P19022, Uniprot ID: CADH2), E: neurosecretory protein VGF (Uniprot AC: O15240, Uniprot ID: VGF), F: afamin (Uniprot AC: P43652, Uniprot ID: AFAM), G: plasma protease C1 inhibitor (Uniprot AC: P05155, Uniprot ID: IC1), H: apolipoprotein A-I (Uniprot AC: P02647, Uniprot ID: APOA1), I: beta-Ala-His dipeptidase (Uniprot AC: Q96KN2, Uniprot ID: CNDP1), J: chitinase-3-like protein 1(Uniprot AC: P36222, Uniprot ID: CH3L1, also known as YKL-40), K: cystatin-C (Uniprot AC: P01034, Uniprot ID: CYTC) and L: SPARC (Uniprot AC: P09486, Uniprot ID: SPRC). AD, Alzheimer’s disease; MCI, mild cognitive impairment; FTD, frontotemporal dementia. p-value: ^****0–0.0001, ^***0.0001–0.001, ^**0.001–0.01, ^*0.01–0.05.

Fig.5

Rank based correlations of the selected proteins. Aβ₄₂, t-tau, p-tau, cadherin-2 (Uniprot AC: P19022, Uniprot ID: CADH2), neurosecretory protein VGF (Uniprot AC: O15240, Uniprot ID: VGF), afamin (Uniprot AC: P43652, Uniprot ID: AFAM), plasma protease C1 inhibitor (Uniprot AC: P05155, Uniprot ID: IC1), inter-alpha-trypsin inhibitor heavy chain H4 (Uniprot AC: Q14624, Uniprot ID: ITIH4), apolipoprotein A-I (Uniprot AC: P02647, Uniprot ID: APOA1), secretogranin-2 (Uniprot AC: P13521, Uniprot ID: SCG2), beta-Ala-His dipeptidase (Uniprot AC: Q96KN2, Uniprot ID: CNDP1), alpha-1B-glycoprotein (Uniprot AC: P04217, Uniprot ID: A1BG), chitinase-3-like protein 1(Uniprot AC: P36222, Uniprot ID: CH3L1, also known as YKL-40), cystatin-C (Uniprot AC: P01034, Uniprot ID: CYTC) and SPARC (Uniprot AC: P09486, Uniprot ID: SPRC). p-value: ^****0–0.0001, ^***0.0001–0.001, ^**0.001–0.01, ^*0.01–0.05.