High-precision plasma β-amyloid 42/40 predicts current and future brain amyloidosis

Participants

The study cohort represents a convenience sample. Participants enrolled in longitudinal studies of memory and aging at Washington University who underwent plasma collection within 18 months of an amyloid PET scan were considered for inclusion based on plasma availability. Because the IPMS assay used 1.6 mL of plasma, samples were selected for which the biorepository had relatively large amounts of plasma available as determined by the biorepository core leader. Participants of all ages and diagnoses were included, but the biorepository had greater availability of plasma from younger and cognitively normal participants. All participants underwent clinical assessments that included the Clinical Dementia Rating (CDR)¹⁴ and Mini-Mental State Examination.¹⁵ APOE genotype was obtained from the Knight Alzheimer Disease Research Center (ADRC) Genetics Core.¹⁶

Standard protocol approvals, registrations, and patient consents

All procedures were approved by the Washington University Human Research Protection Office, and written informed consent was obtained from each participant.

Plasma and CSF collection and processing

CSF was collected as previously described.¹⁷ Participants underwent lumbar puncture at 8 am following overnight fasting. Twenty to thirty milliliters of CSF was collected in a 50-mL polypropylene tube via gravity drip using an atraumatic Sprotte 22-G spinal needle. The tube was inverted gently to disrupt potential gradient effects and centrifuged at low speed to pellet any cellular debris. The CSF was then aliquoted into polypropylene tubes and stored at −80°C. CSF Aβ42, t-tau, and p-tau181 were measured with the corresponding Elecsys immunoassays on the Roche (Basel, Switzerland) cobas e601 analyzer.¹⁸

At the same session as CSF collection, blood was drawn into two 10-mL syringes precoated with 0.5 M EDTA, then transferred to two 15-mL polypropylene tubes containing 120 μL 0.5 M EDTA. The samples were kept on wet ice until centrifugation (<2 hours) to separate plasma from blood cells. The plasma was then transferred to a single 50-mL polypropylene tube, gently mixed, aliquoted into polypropylene tubes, and stored at −80°C.

Immunoprecipitation of Aβ38, Aβ40, and Aβ42

Targeted Aβ isoforms (Aβ38, Aβ40, and Aβ42) were simultaneously immunoprecipitated from 1.6 mL of plasma or 0.5 mL of CSF via a monoclonal anti-Aβ mid-domain antibody (HJ5.1, anti-Aβ13-28) conjugated to M-270 Epoxy Dynabeads (Invitrogen, Carlsbad, CA).¹⁹ Samples were added to 380 μL of a master mix containing 5.26X protease inhibitor cocktail (Roche), 0.263% (w/v) Tween-20, 2.63X phosphate-buffered saline, and 2.63 M guanidine. Plasma samples were spiked with 20 μL of a solution containing 3.75 pg/μL ¹²C¹⁵N-Aβ38, 25 pg/μL ¹²C¹⁵N-Aβ40, and 2.5 pg/μL ¹²C¹⁵N-Aβ42 (labeled peptides from rPeptide, Athens, GA) in 4:1 0.1% ammonium hydroxide:acetonitrile while CSF samples were spiked with 20 μL of a solution containing 75 pg/μL ¹²C¹⁵N-Aβ38, 500 pg/μL ¹²C¹⁵N-Aβ40, and 50 pg/μL ¹²C¹⁵N-Aβ42 in 4:1 0.1% ammonium hydroxide:acetonitrile. All subsequent immunoprecipitation steps were performed as previously described.¹²

Liquid chromatography–mass spectrometry

Plasma analyses were performed as previously described.¹² CSF analyses were performed on a Waters (Milford, MA) Xevo TQ-S triple quadrupole mass spectrometer interfaced with a Waters nanoAcquity chromatography system. For CSF analyses, extracted digests were reconstituted with 50 µL of 20 nM BSA Digest (Pierce, Appleton, WI) in 10% formic acid/10% acetonitrile. A 4.5 μL aliquot of each reconstituted digest was loaded via direct injection onto a Waters 100 × 0.075 mm Acquity M-class HSS T3 column in 10% acetonitrile/2% dimethyl sulfoxide (DMSO)/0.1% formic acid with a flow rate of 600 nL/min for 12 minutes. After loading, peptides were resolved using an 8-minute linear gradient at 400 nL/min from 10% acetonitrile/2% DMSO/0.1% formic acid to 50% acetonitrile/2% DMSO/0.1% formic acid. The initial gradient was followed by a steeper linear gradient to 65% acetonitrile/2% DMSO/0.1% formic acid over 2 minutes at 400 nL/min. The column was then washed with 95% acetonitrile/2%DMSO/0.1% formic acid for 5 minutes at 400 nL/min. Finally, the column was equilibrated back to initial solvent conditions for 5 minutes at 600 nL/min.

Analysis of mass spectrometry data

Peptides derived from human Aβ contained amino acids with the naturally occurring ¹⁴Nitrogen (¹⁴N) isotope, while peptides derived from the exogenous Aβ spiked into samples as a standard contained amino acids that were uniformly labeled with the ¹⁵Nitrogen (¹⁵N) isotope. The precursor/product ion pairs utilized were chosen as previously described^12,19 and the derived integrated peak areas were analyzed using the Skyline software package.²⁰ For each isotopomer (¹⁴N or ¹⁵N) of the Aβ isoforms (Aβ38, Aβ40, or Aβ42), integrated peak areas for selected product ions were summed. The Aβ42 concentration was calculated as follows: the sum of the integrated peak areas for the product ions derived from the ¹⁴N isotopomer for Aβ42 divided by the sum of the integrated peak areas for the product ions derived from the ¹⁵N isotopomer for Aβ42 multiplied by the concentration of spiked Aβ42 ¹⁵N internal standard. The Aβ40 concentration was calculated with the same approach. The final Aβ42/Aβ40 ratio was obtained by dividing the calculated Aβ42 concentration by the calculated Aβ40 concentration.

All mass spectrometry and quality control analyses were performed prior to sample unblinding. Values that failed quality control were not used if they did not meet threshold criteria for sample preparation (missing/mishandled samples), signal intensity, chromatographic properties (peak width/shape), coefficient of variation (technical replicates), and mass spectral noise. Three percent of plasma and CSF samples failed the quality control (QC) protocol and were dropped from the study.

Plasma was collected from a cognitively normal young individual and an older individual known to have brain amyloidosis for use as high and low QC calibrators, respectively. The high and low QC calibrators, along with intermediate mixes of the high and low QC calibrators, were run with every batch of plasma samples. Raw plasma Aβ42/Aβ40 ratio values were normalized to the QC calibrators using linear regression to minimize batch-to-batch variability (see appendix e-1, doi.org/10.5061/dryad.hr45320). This normalization was planned a priori because of batch effects that were observed in unpublished experiments. Although high and low QC calibrators were also run with CSF samples, no significant batch-to-batch variability was noted and therefore no normalization was performed.

Amyloid PET imaging

Amyloid PET was used as the primary reference standard for amyloidosis because it is a well-established biomarker that is widely used in clinical trials for assessment of brain amyloid burden.^5,6,8 Participants underwent a dynamic scan with either ¹¹C Pittsburgh compound B (PiB) or ¹⁸F AV45. PiB PET imaging was performed with a Siemens 962 HR + ECAT PET or Biograph mCT scanner (Siemens/CTI, Knoxville, KY). AV45 PET imaging was performed with a Siemens Biograph mMR scanner (Siemens/CTI). Structural MRI using magnetization-prepared rapid gradient echo T1-weighted images was acquired at 3T and processed using FreeSurfer 5.3²¹ (freesurfer.net/) to derive cortical and subcortical regions of interest.²² Regional data from the 40- to 60-minutes postinjection window for PiB and the 50- to 70-minutes window for AV45 were converted to standardized uptake value ratios (SUVRs) using cerebellar gray as a reference and partial volume-corrected using a regional spread function approach.²³ Values from the left and right lateral orbitofrontal, medial orbitofrontal, precuneus, rostral middle frontal, superior frontal, superior temporal, and middle temporal cortices were averaged together to represent a mean cortical SUVR. Amyloid PET positivity was defined a priori with the established cutoffs of >1.42 for PiB²⁴ and >1.22 for AV45.²⁵ Amyloid PET Centiloid was used to combine PiB and AV45 data on a similar scale.^26,27

Statistical analyses

Characteristics of amyloid PET-positive and PET-negative groups were compared using Student t tests for continuous variables and χ² tests or Fisher exact tests for categorical variables. Receiver operating characteristic (ROC) analyses were performed to evaluate the ability of either plasma or CSF Aβ42/Aβ40 to diagnose amyloid PET status and were implemented with PROC LOGISTIC. Positive percent agreement (PPA) was defined as the percent of amyloid PET-positive individuals who were positive by a given plasma or CSF Aβ42/Aβ40 value. Negative percent agreement (NPA) was defined as the percent of amyloid PET-negative individuals who were negative by a given plasma or CSF Aβ42/Aβ40 value. The Youden Index for each potential plasma or CSF Aβ42/Aβ40 value was calculated as the PPA plus the NPA minus 1 and the value with the maximum Youden Index was selected as the cutoff value. For high areas under the curve (AUCs) (>0.90), we used the Wilson²⁸ score interval for calculation of confidence intervals (CIs).

Because amyloid PET Centiloid values were not normally distributed, Spearman correlations were used to evaluate the relationship between amyloid PET Centiloid and plasma or CSF Aβ42/Aβ40 values. Analysis of covariance with plasma or CSF Aβ42/Aβ40 as the outcome variable and centered age (age = the mean age for the cohort of 63.7 years), APOE ε4 status, and sex as predictors were implemented with PROC GLM. Models predicting last amyloid PET status for initially amyloid PET-negative individuals using baseline plasma or CSF Aβ42/Aβ40 status and follow-up time were implemented in PROC LOGISTIC with exact estimates because there were relatively few amyloid PET converters.²⁹ For individuals with more than one plasma sample, the intraindividual annual rate of change was computed and group differences were compared with one-way analysis of variance (ANOVA) and Tukey multiple comparisons tests.

For calculation of predicted savings in number of amyloid PET scans by screening with plasma Aβ42/Aβ40, the frequency of amyloid PET positivity as a function of age group and APOE ε4 status was estimated based on data from the Anti-Amyloid Treatment in Alzheimer's (A4) Prevention Study.¹¹ The calculations assume that 35% of participants were APOE ε4 carriers, 76% were age 56–75 years old, and 24% were 75–85 years old. The probability of a positive amyloid PET scan for individuals with a positive blood test was based on a logistic regression model generated with data from the present study with the blood test result (positive or negative), age (as a continuous variable), and APOE ε4 status as predictors.

Statistical analyses were conducted using SAS 9.4 (SAS Institute Inc., Cary, NC). Plots were created with GraphPad Prism version 7.04 (GraphPad Software, La Jolla, CA). Heat maps were generated with the R ggplot2 package. A p value <0.05 was considered statistically significant.

Data availability statement

Data in the study will be deposited in the Washington University Knight ADRC dataset, which will be shared by request from any qualified investigator upon approval by the Knight ADRC data request committee.

Participants

A total of 210 plasma samples from 158 individuals were analyzed (see table 1 for participant characteristics) by IPMS. The SD for measurements of both Aβ42 and Aβ40 was 1 pg/mL and the coefficient of variation (CV) for the plasma assay was 5% for Aβ42, 0.6% for Aβ40, and 4% for Aβ42/Aβ40 (see appendix e-2, doi.org/10.5061/dryad.hr45320). A total of 186 available CSF samples collected the same day as plasma from 145 individuals were assayed for Aβ42/Aβ40 by IPMS. Data on CSF Aβ42, t-tau, and p-tau, as measured by Elecsys immunoassays, were available for 152 individuals.

An amyloid PET scan performed within 18 months of the baseline plasma sample was negative for 115 individuals and positive for 43 individuals. The average interval between the plasma collection and the amyloid PET scan was 0.26 ± 0.35 years (mean ± SD) with a range of 0–1.5 years. The baseline age range extended from 46.1 to 86.9 years. Compared to amyloid PET-negative individuals, individuals who were amyloid PET-positive were older (71.4 ± 6.8 vs 60.8 ± 6.7 years, p < 0.0001), were more likely to carry an APOE ε4 allele (63% vs 35%, p = 0.001), were more likely to have cognitive impairment as demonstrated by a CDR greater than 0 (14% vs 3%, p = 0.04), and had lower CSF Aβ42 and higher CSF t-tau and p-tau (p < 0.0001) by Elecsys immunoassays.

Correspondence of baseline plasma and CSF Aβ42/Aβ40 to baseline amyloid PET

Individuals with a positive amyloid PET at baseline had a significantly lower baseline plasma Aβ42/Aβ40 compared to individuals with a negative amyloid PET at baseline (0.115 ± 0.006 vs 0.128 ± 0.009, p < 0.0001) (figure 1A). ROC analysis demonstrated that baseline plasma Aβ42/Aβ40 was a good predictor of baseline amyloid PET status, with an AUC of 0.88 (95% CI 0.82–0.93) (figure 1C). The cohort represented a wide age range, but the performance of the assay was similar in a subcohort of individuals (n = 101) older than 60 years (AUC 0.87, 95% CI 0.80–0.94). A plasma Aβ42/Aβ40 cutoff of <0.1218 was considered positive and had the maximum Youden Index with a PPA of 0.88 (95% CI 0.75–0.96) and an NPA of 0.76 (95% CI 0.67–0.83) with amyloid PET status (figure 1C). Baseline plasma Aβ42/Aβ40 was inversely correlated with amyloid PET on the continuous Centiloid scale (figure 1E), with a Spearman ρ of −0.55 (95% CI −0.65 to −0.43). Baseline plasma Aβ40 was weakly correlated with amyloid PET Centiloid (Spearman ρ of 0.29, 95% CI 0.13–0.43), while plasma Aβ42 was not significantly correlated with amyloid PET Centiloid (figure e-1, A and B, doi.org/10.5061/dryad.hr45320).

CSF p-tau/Aβ42 as measured by the Elecsys platform was chosen as an alternative reference standard for brain amyloidosis because this measure has the highest correspondence with amyloid PET of the established CSF biomarkers and better distinguishes amyloid PET status than Aβ42 alone (figure e-2, doi.org/10.5061/dryad.hr45320).^18,30 For plasma Aβ42/Aβ40, the AUC was 0.85 (95% CI 0.79–0.92) for a CSF Elecsys p-tau/Aβ42 cutoff of 0.0198¹⁸ and 0.85 (95% CI 0.78–0.92) for a cutoff of 0.0220.³⁰

As expected, baseline CSF Aβ42/Aβ40 as measured by IPMS was lower in individuals with a positive amyloid PET at baseline (figure 1B). The concordance between CSF Aβ42/Aβ40 and amyloid PET was nearly perfect (figure 1D), with an AUC of 0.98 (95% CI 0.95–0.99). A CSF Aβ42/Aβ40 cutoff of <0.1094 was considered positive and had the maximum Youden Index with a PPA of 0.98 (95% CI 0.87–1.0) and an NPA of 0.94 (95% CI 0.88–0.98). Baseline CSF Aβ42/Aβ40 was inversely correlated with amyloid PET Centiloid (figure 1F), with a Spearman ρ of −0.66 (95% CI −0.74 to −0.55). Similar inverse correlations between plasma and CSF Aβ42/Aβ40 and amyloid PET were obtained when the 2 tracers used, PiB and AV45, were evaluated separately (figure e-3, doi.org/10.5061/dryad.hr45320).

Baseline plasma and CSF Aβ42/Aβ40 were correlated (Spearman ρ of 0.66, 95% CI 0.56–0.75) (figure 1G). Using the cutoffs described herein, plasma and CSF Aβ42/Aβ40 had concordant predictions for amyloid status in 122 of 145 individuals (84%). All individuals with both a high (negative) CSF and plasma Aβ42/Aβ40 were amyloid PET-negative (n = 81). A total of 35 of 41 individuals (85%) with both a low (positive) plasma and CSF Aβ42/Aβ40 were amyloid PET-positive, and 6 (15%) were PET-negative. A total of 18 of 19 individuals (95%) with a positive plasma Aβ42/Aβ40 but negative CSF Aβ42/Aβ40 were amyloid PET-negative. Four individuals with a negative plasma Aβ42/Aβ40 but positive CSF Aβ42/Aβ40 were amyloid PET-positive. Plasma Aβ42 and Aβ40 were not individually significantly correlated with CSF Aβ42 and Aβ40, respectively (figure e-1, C and D, doi.org/10.5061/dryad.hr45320).

Relationship between plasma or CSF Aβ42/Aβ40 and age, APOE ε4 status, and sex

Baseline plasma Aβ42/Aβ40 was lower with older age (p < 0.0001) and was lower in APOE ε4 carriers (p < 0.0001) and men (p = 0.002) (figure 2A and table 2). Each decade of age, APOE ε4 carrier status, and male sex was associated with lower plasma Aβ42/Aβ40 levels by ∼0.005 (for comparison, the difference between plasma Aβ42/Aβ40 in amyloid PET-positive and PET-negative individuals was ∼0.012). In models for baseline plasma Aβ42/Aβ40, there was no significant interaction between age and APOE ε4 status, age, and sex, or APOE ε4 status and sex. Baseline CSF Aβ42/Aβ40 was lower with older age and was lower in APOE ε4 carriers (both p < 0.0001) (figure 2B and table 2). In contrast to plasma Aβ42/Aβ40, CSF Aβ42/Aβ40 did not vary by sex.

Adding age and APOE ε4 status to a model for prediction of amyloid PET status by plasma Aβ42/Aβ40 improved the AUC from 0.88 (95% CI 0.82–0.93) to 0.94 (95% CI 0.90–0.97) (figure 2C). Both age (p < 0.001) and APOE ε4 status (p < 0.03) were significant predictors in this model. When added to the model, sex was not a significant predictor, likely because the model already correctly classified nearly all participants and sex did not improve classification of the remaining few discordant cases. The combination of plasma Aβ42/Aβ40, age, and APOE ε4 status were used to predict the likelihood of amyloid PET positivity (figure 2D).

Prediction of amyloid PET conversion

A subcohort of 100 individuals underwent at least 1 amyloid PET scan >1.5 years following their baseline plasma sample (for subcohort characteristics, see table e-1, doi.org/10.5061/dryad.hr45320). For all individuals in this subcohort, the average interval between the baseline plasma collection and last amyloid PET scan was 3.9 ± 1.4 years with a range of 1.9–9.0 years. A logistic regression model that included follow-up time from plasma collection to the last amyloid PET scan found that plasma Aβ42/Aβ40 was a good predictor of amyloid PET status at the last amyloid PET scan (AUC 0.88, 95% CI 0.81–0.95). A total of 94 of the 100 individuals in the subcohort with longitudinal amyloid PET data also had matched CSF samples that underwent analysis by IPMS. A similar model found that CSF Aβ42/Aβ40 was an excellent predictor of amyloid PET status at the last amyloid PET scan (AUC 0.96, 95% CI 0.92–0.98).

In the subcohort with longitudinal amyloid PET data, 74 were amyloid PET-negative at baseline; 8 converted to amyloid PET-positive over the follow-up period while 66 remained amyloid PET-negative. The amyloid PET converters had lower baseline plasma Aβ42/Aβ40 than the individuals who remained amyloid PET-negative (0.117 ± 0.008 vs 0.128 ± 0.009, respectively, p < 0.01 by Student t test; see table e-1, doi.org/10.5061/dryad.hr45320, and figure 3, A and C). A logistic regression model that included follow-up time demonstrated that amyloid PET-negative individuals with a positive plasma Aβ42/Aβ40 (<0.1218) had a 15-fold increased risk of conversion to amyloid PET-positive compared to individuals with a negative plasma Aβ42/Aβ40 (p = 0.01 by exact test, figure 3E). Sixty-eight of the 74 individuals with longitudinal amyloid PET data who were amyloid PET-negative at baseline had matched CSF samples with Aβ42/Aβ40 by IPMS. The amyloid PET converters with CSF data (n = 7) had a lower baseline CSF Aβ42/Aβ40 compared to the 61 individuals who remained amyloid PET-negative (0.110 ± 0.014 vs 0.136 ± 0.016, respectively, p < 0.001 by Student t test; see table e-1, doi.org/10.5061/dryad.hr45320, and figure 3, B and D). Amyloid PET-negative individuals with positive CSF Aβ42/Aβ40 (<0.1094) had a 21-fold increased risk of conversion to amyloid PET-positive compared to individuals with a negative plasma Aβ42/Aβ40 (p = 0.03 by exact test, figure 3F).

Amyloid PET converters had a significantly higher baseline amyloid PET Centiloid value compared to individuals who remained amyloid PET-negative (6.9 ± 4.7 vs −0.5 ± 4.0, p < 0.0001), suggesting amyloid PET converters had below-threshold brain amyloidosis. One individual classified as an amyloid PET converter with negative plasma and CSF Aβ42/Aβ40 at both the first and last time points had Elecsys CSF biomarkers that were inconsistent with brain amyloidosis (at the last time point CSF Aβ42 was 1,434 pg/mL, t-tau was 193 pg/mL, and p-tau was 17.5 pg/mL), suggesting the last amyloid PET scan may have been false-positive.

Longitudinal change in plasma and CSF Aβ42/Aβ40

A subcohort of 50 individuals had longitudinal plasma Aβ42/Aβ40 collected within 18 months of a longitudinal amyloid PET scan (figure 4; see table e-2, doi.org/10.5061/dryad.hr45320, for participant characteristics), allowing examination of intraindividual rate of change in plasma Aβ42/Aβ40. For all participants in this subcohort, the average interval between the first and last plasma collections was 3.6 ± 1.2 years with a range of 1.9–7.1 years. Thirty-nine of these individuals also had CSF samples that were analyzed for Aβ42/Aβ40 by IPMS. The intraindividual rate of change for each participant was estimated. There was a significant decline in both plasma (−0.0011/y) and CSF Aβ42/Aβ40 (−0.0023/y) over time (p < 0.001 and p < 0.0001 by 1-sample t test, respectively). There was no difference in the rate of change of plasma Aβ42/Aβ40 by amyloid PET group (one-way ANOVA was not significant; figure 4C). However, amyloid PET converters had a faster decline in CSF Aβ42/Aβ40 compared to individuals who were amyloid PET-positive both at baseline and the last amyloid PET scan (p < 0.05 for one-way ANOVA, p < 0.05 for Tukey post hoc test; figure 4D).

Utility of plasma Aβ42/Aβ40 as a screening test

The value of using plasma Aβ42/Aβ40 to screen for individuals with a high risk of brain amyloidosis was evaluated (table 3). The frequency of amyloid PET positivity as a function of age group and APOE ε4 status was based on data from the A4 Prevention Study, which included cognitively normal individuals aged 65–85 years.¹¹ The probability of a positive amyloid PET scan for individuals with a positive blood test was based on a logistic regression model generated with data from the present study. By screening individuals with a positive plasma Aβ42/Aβ40, fewer confirmatory amyloid PET scans would be required to obtain a cohort of 100 individuals with a positive amyloid PET scan. The percentage of amyloid PET scans saved by first screening participant with plasma Aβ42/Aβ40 was highest in APOE ε4 noncarriers and younger individuals. For a cohort similar to A4, screening participants with plasma Aβ42/Aβ40 could reduce the number of amyloid PET scans required by approximately 62%.