Highly specific and ultrasensitive plasma test detects Abeta(1–42) and Abeta(1–40) in Alzheimer’s disease

Plasma biomarkers that reflect specific amyloid beta (Abeta) proteoforms provide an insight in the treatment effects of Alzheimer’s disease (AD) therapies. Our aim was to develop and validate ready-to-use Simoa ‘Amyblood’ assays that measure full length Abeta1-42 and Abeta1-40 and compare their performance with two commercial assays. Linearity, intra- and inter-assay %CV were compared between Amyblood, Quanterix Simoa triplex, and Euroimmun ELISA. Sensitivity and selectivity were assessed for Amyblood and the Quanterix triplex. Clinical performance was assessed in CSF biomarker confirmed AD (n = 43, 68 ± 6 years) and controls (n = 42, 62 ± 5 years). Prototype and Amyblood showed similar calibrator curves and differentiation (20 AD vs 20 controls, p < 0.001). Amyblood, Quanterix triplex, and ELISA showed similar linearity (96%-122%) and intra-assay %CVs (≤ 3.1%). A minor non-specific signal was measured with Amyblood of + 2.4 pg/mL Abeta1-42 when incubated with 60 pg/mL Abeta1-40. A substantial non-specific signal of + 24.7 pg/mL Abetax-42 was obtained when 40 pg/mL Abeta3-42 was measured with the Quanterix triplex. Selectivity for Abeta1-42 at physiological Abeta1-42 and Abeta1-40 concentrations was 125% for Amyblood and 163% for Quanterix. Amyblood and Quanterix ratios (p < 0.001) and ELISA Abeta1-42 concentration (p = 0.025) could differentiate AD from controls. We successfully developed and upscaled a prototype to the Amyblood assays with similar technical and clinical performance as the Quanterix triplex and ELISA, but better specificity and selectivity than the Quanterix triplex assay. These results suggest leverage of this specific assay for monitoring treatment response in trials.


Specificity and selectivity analyses
Specificity is the ability of an assay to distinguish between the analyte that the assay is intended to measure, and structurally similar analytes. We tested specificity of the Amyblood Abeta1-42 assay by measuring the assay signal in response to increasing concentrations of 6 pg/mL, 25 pg/mL and 60 pg/mL of the fragments Abeta1-40, Abeta1-43, Abeta2-42 and Abeta3-42 (custom made by Polypeptide, BE) in sample buffer. We tested specificity of the Amyblood Abeta1-40 assay by measuring the assay signal in response to increasing concentrations of 6 pg/mL, 25 pg/mL and 60 pg/mL of the fragments Abeta1-42, Abeta1-38, Abeta1-39 and Abeta11-40 (custom made by Polypeptide, BE) in sample buffer. The AEBs from Abeta1-42 and Abeta1-40 in buffer measured with the Amyblood assays were extrapolated from the calibration curve to calculate the corresponding concentrations. The AEBs from Abeta1-42 and Abeta1-40 in buffer measured with the Quanterix triplex were calculated derived from the linear measurements in 4, 25, and 130 pg/mL for Abeta1-42 and 1, 15, and 40 pg/mL for Abeta1-40. These AEBs were then extrapolated from the calibration curve to calculated the corresponding concentrations.
Selectivity is the ability of an assay to accurately detect an analyte concentration, in the presence of another, independent analyte that is expected to be present in a test sample. We tested selectivity by measuring the signal of one analyte in sample diluent (either Abeta1-42 or Abeta1-40), in the presence of a known concentration of the other analyte. The added concentrations were selected to include the physiological concentrations present in plasma, as well as concentrations lower and higher than the physiological concentrations. For Abeta1-42 selectivity, we spiked either 6 pg/mL, 25 pg/mL or 60 pg/mL Abeta1-42 in sample diluent, combined with 0 pg/mL (reference condition), 3 pg/mL, 15 pg/mL and 60 pg/mL Abeta1-40. For Abeta1-40 selectivity, we spiked either 3 pg/mL, 15 pg/mL or 60 pg/mL Abeta1-40 in sample diluent, combined with 0 pg/mL (reference condition), 6 pg/mL, 25 pg/mL or 60 pg/mL Abeta1-42.

Quanterix and ELISA assay validation
The Quanterix commercial assays were validated at Amsterdam UMC. Linearity was based on three native EDTA plasma samples, six dilutions were assessed that ranged from df4 to df12500. Intra-assay %CV, LLOQs, specificity and sensitivity were assessed with the same method as the Amyblood assays.
The ELISA LLOQ's were based on the limit of blank data provided by Euroimmun and calculated as the average concentration of 24 blank measurements + 10 standard deviations. The linearity of the ELISA assays was based on data provided by Euroimmun on two EDTA plasma pools spiked with recombinant Abeta1-42 or Abeta1-40 at different concentrations and one native EDTA plasma sample.
Ten dilutions were assessed, from df 1 to df 10. We expect the ELISA assay to perform similar to the Amyblood, since both assays use the same antibodies, and we therefore did not include that assay in the specificity and selectivity studies.

Supplementary results
Prototype and Amyblood assay development and transfer eFigure 1. Lot to lot variation tested on calibrator curves of three different capture antibody lots. A. The mean variation in AEB values between the three different antibody lots tested for Abeta1-42 was 19%. B. The mean variation in AEB values between the three different antibody lots tested for Abeta1-40 was 14%. eFigure 2. Calibrator curve comparison between small and upscaled bead batch A. Mean variation in AEB values between the small batch and upscaled batch is 15.9% for Abeta1-42. B. Mean variation in AEB values between the small batch and upscaled batch is 7.0% for Abeta1-40. Upscaling to bead conjugation volumes from 100µL to 2 mL did not affect calibrator AEB curves.

eFigure 3. Two neat plasma samples measured with the Amyblood assays at AUMC and ADx.
A. The mean Abeta1-42/1-40 ratio of plasma sample 1 was 0.17 ADx and 0.19 at AUMC. B. The mean Abeta1-42/40 ratio of plasma sample 2 was 0.13 ADx and 0.14 at Amsterdam UMC. Mean intra-assay %CV of the Abeta1-42/40 ratio of two neat samples was 3.5% and 2.0% at ADx and 3.9% and 2.2% at Amsterdam UMC. Mean inter-assay %CV of the Abeta1-42/40 ratio was 23% and 13% at ADx and 14% for both samples at Amsterdam UMC. Overall mean inter-assay and inter-center variation of the Abeta ratio was 17.2%, combining all results. Abeta: amyloid beta; AUMC: Amsterdam University Medical Centers; ADx; ADx Neurosciences The Amyblood and Quanterix assays were performed using the Quanterix Simoa. The sample volume is for duplo measurement including dead volume. Abeta; amyloid beta, LLOQ; Lower limit of quantification.