Characterization of a new tetra-plex assay for the quantification of CSF Aβ42, t-tau, p-tau and a-syn.
The general characteristics of the new tetra-plex assay are summarized in Supplementary Figure 1a (Additional File 2). Intra-coefficient of variation (CV) ranged from 7 to 11% and inter-CV ranged from 12 to 16%. The four biomarker assays showed a large dynamic range comprising more than three orders of magnitude, with fold changes ranging from ~175 to 1100 upper /lower limits of quantification. When tested up to 1:6 dilution from neat CSF, the assays displayed acceptable dilution linearity (101% to 114%), suggesting the absence of matrix-effects. Representative standard curves for each biomarker are shown in Supplementary Figure 1b (Additional File 2).
Influence of pre-analytic parameters on biomarkers measurements in the tetra-plex assay.
Treatment of CSF samples under different pre-analytical conditions revealed no significant differences on Aβ42, t-tau, p-tau and α-syn when CSF was stored at 4ºC up to 14 days (Figure 1a). In contrast, storage at RT led to a decrease on t-tau at 14 days (p<0.01) and on p-tau at 5 days (p<0.05) and 14 days (p<0.001) (Figure 1b). T-tau, p-tau and Aβ42 were stable up to three freeze/thawing (F/T) cycles, while α-syn was significantly lower after three F/T cycles (p<0.05) (Figure 1c). t-tau, p-tau and α-syn were stable up to three tube transfers, in contrast to Aβ42, whose concentrations were significantly lower after two (p<0.05) and three (p<0.01) transfers (Figure 1d).
Diagnostic performance of the tetra-plex assay in the differential diagnostic of neurodegenerative dementia.
In order to investigate the performance of the tetra-plex assay in the differential diagnostic of neurodegenerative dementia 199 CSF cases (ND: n=38, AD: n=35, CJD: n=37, VaD: n=28, DLB/PDD: n=27 and FTD: n=34) were quantified. After controlling for age and sex, significantly lower Ab42 concentrations were detected in AD (p<0.01) and DLB/PDD (p<0.05) compared to ND and in AD (p<0.05) compared to VaD (Figure 2a and 2b). T-tau concentrations were significantly higher in AD (p<0.001), CJD (p<0.001), FTD (p<0.001) and DLB/PDD (p<0.01) compared to ND, in CJD compared to AD (p<0.001), VaD (p<0.001), DLB/PDD (p<0.001) and FTD (p<0.001), and in AD compared to VaD (p<0.01) (Figure 2a and 2c). Compared to ND, p-tau concentrations were significantly higher in AD (p<0.001), CJD (p<0.001), VaD (p<0.01), DLB/PDD (p<0.01) and FTD (p<0.01), in AD compared to CJD (p<0.01), VaD (p<0.001), DLB/PDD (p<0.001) and FTD (p<0.001), and in CJD compared to VaD (p<0.01) and FTD (p<0.01) (Figure 2a and 2d). Significantly higher a-syn concentrations were present in CJD compared to the rest of diagnostic groups (p<0.001) (Figure 2a and 2e). No other significant differences were detected for a-syn.
Correlation analyses between the four biomarkers were studied in the total population. Modest, but significant correlations were detected between Aβ42 and a-syn (rho=0.17, p=0.036), t-tau and p-tau (rho=0.18, p=0.017) and p-tau and a-syn (rho=0.19, p=0.011), while a strong significant correlation was detected between t-tau and a-syn (rho=0.80, p<0.001) as shown in Supplementary Figure 2 (Additional File 2).
Comparison of the diagnostic accuracies obtained with the tetra-plex assay and the single-plex assays.
In order to determine the performance of the fluorimetric tetra-plex assay and compared it to established commercial single-plex assays, AUCs and associated p values were calculated. For Aβ42, significant differences between assays were detected neither in the discrimination of ND from neurodegenerative dementias (Figure 3a), nor in the discrimination of neurodegenerative dementia groups among them (Supplementary Figure 3- Additional File 2). T-tau in the tetra-plex assay was better discriminating ND from VaD (p=0.029), DLB/PDD (p<0.001) and FTD (p=0.002) than single-plex t-tau (Figure 3a), and tetra-plex p-tau was better discriminating ND from all neurodegenerative dementias than single-plex p-tau (ND vs. AD: p=0.003, ND vs. CJD: p=0.009, ND vs. VaD: p=0.049, ND vs. DLB/PDD: p=0.012 and, ND vs. FTD: p<0.001) (Figure 3a). In contrast, neither tetra-plex t-tau nor tetra-plex p-tau displayed better accuracies discriminating neurodegenerative dementias than the corresponding single-plex assays, with the exception of AD vs. CJD (p=0.001) and AD vs. FTD (p=0.018) comparisons for p-tau (Supplementary Figure 3- Additional File 2). Regarding a-syn, the tetra-plex assay displayed significantly lower accuracy in discriminating ND from AD cases compared with the single-plex assay (p=0.038), although both methods showed poor AUC values (0.53 and 0.69, respectively) discriminating these diagnostic groups.
A moderate but significant correlation was observed between tetra-plex and single-plex methods for Ab42 (r=0.61 (95%CI=0.51-0.69), p<0.001) (Figure 3b), while a strong agreement was detected for t-tau (r=0.83 (95% CI=0.78-0.87), p<0.001) (Figure 3c), p-tau (r=0.80 (95% CI=0.74-0.85), p<0.001) (Figure 3d) and a-syn r = 0.72 (95% CI=0.64-0.78), p<0.001 (Figure 3e). Interestingly, for Aβ42, the most discrepant values according to Cook’s distance (Supplementary Figure 4- Additional File 2) were two ND cases with normal Aβ42 concentrations according to tetra-plex assay (654 pg/mL and 745 pg/mL) and low Aβ42 concentrations according to the single-plex assay (125 pg/mL and 235 pg/mL). Thus, in cases with discrepant Aβ42 concentrations, these were associated to its correct diagnosis when using the tetra-plex assay. In spite of significant correlation between concentrations in both types of assay, Passing-Bablok regressions indicated that in all cases there are proportional and systematic differences between tetra-plex and single-plex assays (Figure 3b-e). Therefore, both methods are not interchangeable for the quantification of the four biomarkers.
Next, we aimed to ascertain whether a-syn inclusion added value to the group of core neurodegenerative markers (t-tau, p-tau and Ab42) in the discrimination of diagnostic groups. To address this issue, principal component analysis (PCA) plots were performed in the presence or absence of a-syn (Supplementary Figure 5- Additional File 2). Since biomarkers are best defined to differentiate AD, CJD and ND, and not so much for the rest of the diagnoses, analyses are focused in these three diagnostic groups to simplify the visualization of discriminatory performance in the presence or absence of a-syn. When a-syn was included as a biomarker, a better separation between diagnoses was achieved (Supplementary Figure 5- Additional File 2).
Diagnostic accuracy of the tetra-plex assay using PLS-DA.
We built PLS-DA models to discriminate the 15 pairs of diagnostic groups (Table 1). As expected, ND, AD and CJD were best differentiated, reaching accuracies, sensitivities and specificities over 90%. These groups showed as well good discrimination performance against DLB/PDD, FTD and VaD diagnoses, with diagnostic measures significantly higher than those obtained by random chance (last column in Table 1). These latter groups failed to discriminate one from each other showing accuracies similar to random non-informative data-based accuracies, although moderate accuracies were achieved when compared to the ND group (ND vs. DLB/PDD: 81%, ND vs. FTD 75.5% and ND vs. VaD: 70.5%).
PLS-DA separation performance of each diagnostic pair through the first two components is displayed in Figure 4. All pairs showed a clear separation except for DLB/PDD, FTD and VaD combinations. The combinations of neurodegenerative markers contributing the most (VIP > 1) to the diagnostic accuracy are shown in Table 1. Importantly, all biomarkers contributed to the discrimination performance of the PLS-DA models, and composite markers contributed better than single markers.