Evolution of Aβ42 and Aβ40 levels and Aβ42/Aβ40 ratio in plasma during progression of Alzheimer's disease: A multicenter assessment

Abstract

Objective

To better understand the seemingly contradictory plasma β-amyloid (Aβ) results in Alzheimer's disease (AD) patients by using a newly developed plasma Aβ assay, the INNO-BIA plasma Aβ forms, in a multicenter study.

Methods

A combined retrospective analysis of plasma Aβ isoforms on mild cognitive impairment (MCI) from three large cross-sectional studies involving 643 samples from the participating German and Swedish centers.

Results

Detection modules based on two different amino (N)-terminal specific Aβ monoclonal antibodies demonstrated that Aβ in plasma could be reliable quantified using a sandwich immunoassay technology with high precision, even for low Aβ42 plasma concentrations. Aβ40 and Aβ42 concentrations varied consistently with the ApoE genotype, while the Aβ42/Aβ40 ratio did not. Irrespective of the decrease of the Aβ42/Aβ40 ratio with age and MMSE, this parameter was strongly associated with AD, as defined in this study by elevated hyperphosphorylated (P-tau181P) levels in cerebrospinal fluid (CSF).

Conclusion

A highly robust assay for repeatedly measuring Aβ forms in plasma such as INNO-BIA plasma Aβ forms might be a useful tool in a future risk assessment of AD.

Introduction

For prognostic and diagnostic purposes, as well as for optimizing drug development, there is considerable interest in relating biomarker information to the (future) event of clinical Alzheimer’s disease (AD) diagnosis. To this end, biomarkers obtained from different body fluids have been studied. While highly promising prognostic value has been observed in cerebrospinal fluid (CSF) (1, 2, 3, 4) the results for plasma A often seem contradictory. Differing results may well be related to assay configuration since antibodies are known to recognize different epitopes or conformational forms of Aß (5, 6), or could be related to analytical assay performance, in particular at the low concentrations of Aß42 circulating in plasma. Apart from assay-related issues, disease-related factors could also play an obfuscating role (7, 8, 9). Indeed, two recent studies using a new multiparameter assay format, INNO-BΙA plasma Aß forms, report seemingly divergent results. A first study concentrating on follow-up of MCI patients converting to AD showed no significant differences in Aß plasma levels (10), while in one of the largest MCI follow-up studies (11) in which also early dementia cases were included, highly significant reductions in the plasma Aß42/Aß40 ratio and in Aß42 levels were reported. A major assumption underlying the methodology used in these studies is that (i) the biomarker level obtained at a single point in time is constant, (ii) clinical diagnosis is reliable and (iii) the constant level determines the outcome.

To better understand these apparently contradictory results, a retrospective analysis of the combined results of these latter studies was performed. As a first step, a correlation analysis and an analytical sensitivity assessment of the two detection modules in the multiparameter assay were carried out. The pooled data were subsequently analyzed irrespective of clinical diagnosis in relation to ApoE genotype as risk factor, and in relation to age and Mini-mental State Examination (MMSE) score as indicators for AD progression. Defining an AD surrogate phenotype based on CSF- P-tau181P levels was used to address the association of Aß plasma parameters and AD.

Methods

For practical purposes the analysis was only performed on the subset of data with complete biomarker and covariate information, thereby excluding a limited number of subjects (see results).

The multiple linear regression analysis was performed using a backward elimination strategy starting from a model with study, age, gender and MMSE and all second order interactions with study. The elimination step used a significance threshold of p= 0.05 and study was always kept in the model as a stratification factor. Subsequently interactions were tested between the predictors retained in the model. Final models were verified in detail for normality of residuals, homogeneity of variance and influential values.

All analyses were performed in the statistical package R version 2.6.1 (The R foundation for statistical computing, 2007).

Interlaboratory performance

The INNO-BΙA plasma Aß forms is a multiplex microsphere-based Luminex xMAP® technique that allows simultaneous quantification of Aß1-40 and Aß1-42 using C-terminal antibodies, with 21F12 (42-specific) and 2G3 (40-specific) employed as capturing antibodies on the beads. Two different N-terminal specific-antibodies, 3D6 (full-size Aß, module A) and, 4G8 (N-truncated Aß, Module B; Figure 1) characterize the two modules. A “non-Aβ”- binding antibody, AT120, is added to account for the presence of heterophilic antibodies (Hansson et al. 2008). All plasma samples are diluted 1/3 with a buffer containing detergent before analysis. Synthetic Aß peptides obtained from Bachem (Heidelberg, Germany) were used as calibrators in the concentration range of 15-600 pg/ml for Aß42 and from 30-3000 pg/ml for Aß40. Assay runs were analyzed using a Luminex 100 IS instrument and a 4-parameter sigmoidal curve-fitting method which converted median fluorescence intensity (MFI) values into concentrations.

Figure 1.

Principle of the INNΟ-BΙA plasma Aß forms in which C-terminal antibodies, 21F12 (42-specific) and 2G3 (40-specific) are used as capturing antibodies on the beads and N-terminal antibodies as detectors. Module A, using 3D6, an end N-terminal specific antibody, determines full-size Aß1-42 and Aß1-40 and module B, using 4G8 and internal epitope, determines N-truncated AßN-42 and AßN-40 forms. A bead coated with AT120 antibody is used to monitor potential matrix effects. The sequence of human Aß peptides with the corresponding epitopes of the antibodies used in the assay is also shown

To determine imprecision in the three centers, two human plasma samples, collected, shipped and stored identically were assayed in 6 to 13 runs. The coefficients of variation ranged from 1.9 to 7.2% for Aß1-42, from 3.0 to 7.3% for Aß1-40 and from 4.2 to 6.6% for Aß1-42/ Aß1-40 ratio.

Results

Analysis of the pooled data

A total of 643 plasma Aß determinations from three cross-sectional studies were pooled: 155 samples originated from the Memory disorder clinic at Malmö University (=Malmö) (1), 148 from the Gothenburg MCI study (=DTS, Demensjukdom i Tidigt Skede)) (12), and 361 from the Competence Net Dementias (http://www.kompetenznetz-demenzen.de/index.html, =KND) (11;13;14). Incomplete data (53 samples) with respect to ApoE genotyping (n=44), MMSE (n=7), and invalid Aß1-42 values (n=2) were not considered for the analysis.

Highly significant correlations were found between full-length Aß42 and Aß40 (module A) and N-truncated Aß forms (module B) (Aβ42, r² = 0.75, p<0.001; Aß40, r²=0.72), as well as for the Aß42/Aß40 ratio (r²=0.63, p<0.001). Figure 2 shows the distribution of the raw Aß plasma data. A small number of plasma samples (n=18) showed Aß levels (both Aß42 and Aß40) below the lowest standard point (15 pg/ml for Aß42 and 30 pg/ml for Aß40) in the standard curve in module A and these were not considered for further analysis. In addition, a few subjects had outlying plasma Aß42/A 40 ratios (Grubbs test for outliers) in both modules and these were also not considered in subsequent analysis. Thus, the final number of 572 cases (Table 1 for demographics per center) was considered for covariate analysis of the module A pooled data.

Figure 2.

Distribution of Aß42, Aß40 concentrations and the Aß42/Aß40 ratio in the pooled data in module A and B of the INNO-BΙA plasma Aß forms assay. The dotted lines in the upper module A panel represents the value outside of which data are excluded in the pooled analysis

Table 1.

Demographics of the pooled data, excluding incomplete data, outlying Aß42/Aß40 ratios and Aß concentrations below the lowest standard point in the curve, broken down by center

	DTS	KND	Malmö
N	127	297	148
Age	64 (58–69)	68 (62–74)	72 (65–78)
MMSE	29 (28–30)	26 (23–28)	27 (26–29)
ApoE4	56/127 (44%)	155/297 (52%)	76/148 (51%)
AD/MCI-AD	13/127 (10%)	184/297 (62%)	46/148 (31%)

Covariate and endpoint analysis

As shown in Figure 3A the relation between age and the Aß1-42/Aß1-40 ratio was consistent over the three centers. A decrease of the ratio with age (0.7% per year, p=0.005) and with MMSE (0.3% per MMSE point, p=0.025) was found in the entire study population. The consistent increase of Aß42 and Aß40 with ApoE ϵ4 genotype was also present in the three studies (see Figure 3B), while the ratio was not affected by ApoE genotype. There was a trend for a consistent increase of Aß1-40 with MMSE over the three studies (0.8% per MMSE point, p=0.079), while for age, increased Aß1-42 and Aß1-40 were only found in the study’s German arm (KND) study (1.1 % per year for Aß1-42, p=0.012 and 1.5% per year for Aß1-40, p=0.008).

Figure 3.

Covariate analysis of the pooled data A. Relation of A 1-42/A 40 ratio to age in the different centers. Symbols represent the MMSE score of individual samples. B. Summary of the covariate analysis of all A plasma parameters with respect to age, ApoE ϵe4 genotype, and MMSE

Since the ratio was affected by the MMSE score in this study, suggesting that the Aß42/Aß40 ratio and Aß40 levels are not constant in AD, we defined an AD surrogate phenotype based on CSF-P-tau181P levels. CSF-P-tau181P levels can predict AD at least 2-5 years before onset (1, 2, 15, 16) and are constant once the AD patient becomes demented (17, 18, 19). A CSF-P-tau181P level above a cut-off level of 70 pg/ml (20) defines an AD surrogate phenotype, while < 70 pg/ml defines a non-AD (NAD) phenotype. In the combined Malmö and KND centers, 243 subjects had an AD surrogate phenotype, while the remaining 202 subjects were considered as NAD. A highly significant decrease was found in the Aß42/Aß40 ratio for the AD surrogate (0.149 ± 0.043 (SD) and NAD surrogate: 0.166 ± 0.037), that cannot be explained by the differences in age and MMSE score between the two groups. Despite a considerable overlap, 39% (95% CI 35-49%) of the population could be assigned to two risk groups (Figure 4): a low-risk AD group defined by an Aß42/Aß40 ratio above 0.180 showed a 68% (59-76%) agreement with the AD/NAD surrogate phenotype. The high-risk group with an Aß42/Aß40 ratio below 0.120 demonstrated a 65% (53-75%) agreement with the AD/NAD phenotype.

Figure 4.

Distribution of Aß42/Aß40 ratio in the AD surrogate phenotype (black) and in the non-AD (NAD) surrogate phenotype (white). A ratio above 0.180 defines a low-risk AD group and has a 68% agreement with the surrogate classification, while a ratio below 0.120 defines a high-risk AD group with a 65% agreement. Both risk groups represent 39% of the total population

Discussion

To our knowledge this is the first multicenter study on human plasma Aß determinations with on-site testing that shows an intercenter precision below 10% on plasma samples. This needs further confirmation in larger clinical trials using additional plasma samples covering the complete range of Aß concentrations in human plasma. The high correlation obtained for the respective Aß parameters by using two different N-terminal antibodies also indicate that circulating plasma Aß can be reliably quantified using a sandwich immunoassay approach. It is probably the affinity of the detector antibodies that may determine the degree to which low plasma Aß42 levels can be accurately quantified.

While most of the plasma Aß exploratory studies use a clinical diagnosis as an endpoint, such a diagnosis might be associated with some degree of uncertainty. This involves especially in the preclinical stages of AD before dementia is present and in later stages characterized by the overlap of AD with other type of dementia. Therefore the covariate analysis was done prior to the endpoint analysis, such that the endpoint could be defined in alternative ways. In this study we have used a cut-off of 70 pg/ml CSF-P-tau181P level.

In the pooled analysis of the plasma Aß parameters, a consistent effect of age and MMSE, but not ApoE genotype on the Aß42/Aß40 ratio could be demonstrated. In contrast, individual Aß40 and Aß42 levels are affected by ApoE genotype, while the effect of age is inconsistent over the studies. MMSE exerts only a marginal effect on Aß40 levels.

The highly significant decrease in the Aß42/Aß40 ratio in the AD/NAD surrogate phenotype groups accords with several independent studies (9, 21, 22, 23). When taking into account that repeated measurements in plasma of normal individuals are constant over time (8, 24) and that the Aß42/Aß40 ratio may change during progression of AD, repeated measurements of plasma Aß parameters in AD patients may be more informative as a risk predictor of AD development. Given the rather small effects of age and MMSE on the Aß42/Aß40 ratio, follow-up periods of months or even years probably need to be considered.

In conclusion, a highly robust assay measuring repeatedly Aß forms in plasma might be useful in future risk assessments of AD.

Acknowledgements : The KND has been supported by a grant from the German Federal Ministry of Education and Research (BMBF): Kompetenznetz Demenzen (01GI0420) and the Swedish Research Council grant (09946), Sahlgrenska University Hospital, Swedish Brain Power, Alzheimerfonden, Stiftelsen Gamla Tjänarinnor. The authors thank Fred Shapiro (Innogenetics) for useful comments and suggestions.