Characterizing the emergence of amyloid and tau burden in Down syndrome

Abstract INTRODUCTION Almost all individuals with Down syndrome (DS) will develop neuropathological features of Alzheimer's disease (AD). Understanding AD biomarker trajectories is necessary for DS‐specific clinical interventions and interpretation of drug‐related changes in the disease trajectory. METHODS A total of 177 adults with DS from the Alzheimer's Biomarker Consortium‐Down Syndrome (ABC‐DS) underwent positron emission tomography (PET) and MR imaging. Amyloid‐beta (Aβ) trajectories were modeled to provide individual‐level estimates of Aβ‐positive (A+) chronicity, which were compared against longitudinal tau change. RESULTS Elevated tau was observed in all NFT regions following A+ and longitudinal tau increased with respect to A+ chronicity. Tau increases in NFT regions I‐III was observed 0–2.5 years following A+. Nearly all A+ individuals had tau increases in the medial temporal lobe. DISCUSSION These findings highlight the rapid accumulation of amyloid and early onset of tau relative to amyloid in DS and provide a strategy for temporally characterizing AD neuropathology progression that is specific to the DS population and independent of chronological age. Highlights Longitudinal amyloid trajectories reveal rapid Aβ accumulation in Down syndrome NFT stage tau was strongly associated with A+ chronicity Early longitudinal tau increases were observed 2.5–5 years after reaching A+

Repository for Alzheimer Disease and Related Dementias, Grant/Award Number: U24 AG21886 characterizing AD neuropathology progression that is specific to the DS population and independent of chronological age.

K E Y W O R D S
amyloid, amyloid chronicity, Down syndrome, longitudinal, PET, Tau, trajectory modeling

Highlights
• Longitudinal amyloid trajectories reveal rapid Aβ accumulation in Down syndrome • NFT stage tau was strongly associated with A+ chronicity • Early longitudinal tau increases were observed 2.5-5 years after reaching A+

BACKGROUND
The Down syndrome (DS) population is genetically predisposed to develop Alzheimer's disease (AD) in part due to the triplication of chromosome 21 and the resulting increase in amyloid precursor protein (APP). 1,2Increased dementia prevalence is observed after age 50 in DS 3 with an average onset age of circa 55. 4,5 Amyloid-beta (Aβ) plaque density has been characterized using [ 11 C] Pittsburgh Compound-B (PiB) positron emission tomography (PET) imaging.7][28][29] Comparing Aβ and tau in DS revealed higher tau deposition in limbic and neocortical regions with higher Aβ burden 28 and significant tau PET signal was observed in Braak/NFT stage regions I-III for individuals at the threshold of A+. 29 Annual rates of Aβ change in these individuals were on the order of 1%-2%, suggesting a short latency period exists between emergences of these biomarkers.In neurotypical populations, tau has been associated with both Aβ and cognitive decline, 30 and the associations between tau and cognition have been observed with a stronger effect as compared to Aβ and cognition. 31Early studies of tau PET in DS revealed that higher tau burden corresponded with mild cognitive impairment 16 and that tau PET was significantly associated with cortical atrophy 32 and accelerated longitudinal decline in cognition. 33For non-demented individuals with DS, episodic memory decline was most associated with high tau burden, indicating that evaluating these biomarkers in tandem is a sensitive measure to detect early AD progression. 34 interest is the use of longitudinal data to characterize biomarker change throughout the time course of AD.One method applied groupbased trajectory modeling (GBTM) to longitudinal Aβ PET data in cases of sporadic AD to model Aβ change with respect to an A+ threshold. 35Characterizing the estimated years to symptom onset (EYO) has been carried out in ADAD by subtracting an individual's age from the expected age of clinical symptom onset. 37Studies have taken a similar approach to define "EYO" in DS, 5,38,39 and comparing these groups revealed an earlier onset of Aβ PET signal in ADAD. 38These studies used an estimated age of dementia onset of ∼55 years for DS, but previous findings identified a wide age range of dementia onset spanning 25+ years in DS, 5,[40][41][42][43] adding uncertainty to the EYO measure.While the expected age of symptom onset in ADAD is consistent based on specific mutations, 44 a recent study has demonstrated that variability in predicting symptom onset is similar between DS and ADAD, 5 suggesting that the EYO measure can be further refined.In the current work, our goal was to minimize the uncertainty in the measurement of EYO in DS by directly linking the time estimates and tau burden to well-characterized Aβ PET data using the SILA algorithm 36 across participants from the Alzheimer's Biomarker Consortium-Down Syndrome (ABC-DS) study. 45

Participants
The

Aβ PET quantification
PiB PET images were spatially normalized to the Montreal Neurological Institute 152 space (MNI152) via a DS-specific PET template for PiB as previously described. 11Using gray matter cerebellum as a reference tissue, standardized uptake value ratio (SUVR) images were generated through voxel normalization of summed PET images.Global Aβ burden was determined using the amyloid load metric (Aβ L ), 47 derived from all Aβ-carrying voxels in an SUVR image by linear least squares fitting of DS-specific template images representing Aβ carrying capacity and PiB nonspecific binding, as described previously. 23Aβ L was chosen as the primary outcome measure for Aβ burden as it suppresses the nonspecific binding component of voxels in the image and shows increased power to detect early changes in Aβ compared to SUVR/Centiloids (CL). 47

Tau PET quantification
Baseline and follow-up AV-1451 PET images were registered to the baseline T1w-MRI images for each participant, and SUVR images were

Statistical analyses
Rates of Aβ accumulation are presented as mean with standard deviation (SD).Comparisons of estimated A+ onset age between groups were performed using the Student's t-tests.Associations between NFT stage tau, chronological age, Aβ burden, and A+ chronicity were evaluated using Pearson's correlation coefficients with 95% confidence intervals (CIs).Effect sizes: Significant increases in tau change between baseline and follow-up scans were evaluated using Cohen's d effect sizes 49 with 95% CIs.

Modeling Aβ trajectories identifies A+ chronicity time and estimated A+ onset age in DS
Using the SILA algorithm, a longitudinal Aβ trajectory curve was created to represent Aβ burden across a ∼30-year span (Figure 1), consistent with the estimated disease timeline in neurotypical populations as described elsewhere. 50,51Modeled Aβ L and equivalent CL data points for each A+ chronicity time are displayed in Table S1.Aligning

NFT stage tau is associated with chronological age
Increases in tau were observed across all NFT stage regions with respect to chronological age (Figure 3

NFT stage tau is highly associated with A+ chronicity
Nearly all individuals with DS revealed increased tau burden in NFT regions I-IV following A+ onset (Figure 3 Similar to Aβ L , Pearson analysis shows that a higher association was observed between A+ chronicity and NFT stage tau burden compared to chronological age.As A+ chronicity is dependent on the Aβ burden, correlations between Aβ L and tau, and between A+ chronicity and tau were similar and not statistically different.

Longitudinal tau increases rapidly emerge following A+
To characterize the emergence of tau relative to amyloid, longitudinal tau change was measured across all six NFT stage regions and categorized into different bins of A+ chronicity encompassing the early and late stages of Aβ accumulation.A+ chronicity bins were selected to have a similar number of individuals in each bin for A+ chronicity ≥ 0 years.For the participants with longitudinal tau data available, the rate of tau change in SUVR/year was compared against A+ chronicity (Figure 4).Table 2

DISCUSSION
We present the first longitudinal tau PET analysis in the DS population.Using the measure of A+ chronicity, a timeline of the progression   44,53 and also varies in the DS population. 4,5[42][43] This may be because EYO in ADAD is continuously refined based on individuals with close genetic relationships in a given family; thus, progressively narrowing the range of other genetic variability in the group that determines the EYO value.Measuring cognitive decline in individuals with DS can also be challenging due to their lower and varied baseline intellectual functioning level compared to neurotypical populations, [55][56][57] adding further uncertainty to age-based EYO classification.However, recent studies identified that measuring cognitive decline in the domain of episodic memory is a powerful tool to study AD progression in DS, as significant changes in episodic memory scores were observed in association with higher tau burden 34 and PET measures of neurodegeneration. 58 Note: For an A+ chronicity of 0-2.5 years, medium to large effect sizes were observed for NFT regions I-III.For an A+ chronicity of 2.5-5 years, medium to large effect sizes were observed in all NFT stage regions.the observation that medial temporal tau accumulation plateaus and begins to decline as other regions accelerate has been observed in sporadic AD. 59 A+ chronicity has previously been associated with entorhinal tau at the cross-sectional level in sporadic AD, where tau burden increases were estimated to occur 5-10 years following A+ onset, but with considerable inter-individual heterogeneity in a cohort of mostly initially unimpaired, at-risk adults. 35There was heterogeneity in the length of A+ chronicity required to see elevated tau burden in sporadic AD, and many individuals showed no entorhinal tau following 10 years of A+. 35 This finding is in contrast with the observations in our DS cohort, showing an early and far more homogeneous tau increase in relation to A+ chronicity across all NFT stage regions.In sporadic AD, there is a sharp increase in tau prevalence at an Aβ threshold of 50 CL. 60At 50 CL in cognitively unimpaired individuals, longitudinal rates of tau increase are ∼2% per year, 61 and annual rates of Aβ increase are ∼5 CL/year. 62In DS at 30 CL, we observe 2% annual increases in tau burden (Table 2) and an Aβ change rate of ∼5 CL/year.The Aβ change rate in DS increases to ∼6 CL/year at a threshold of 50 CL and continues to accelerate between 50 and 100 CL, after which the rate of increase (but not the level of Aβ) tends to plateau (Figure 2).Comparing DS and sporadic AD, we find that longitudinal tau increases emerge when Aβ rates increase at 5 CL/year, regardless of the baseline CL value.The mean rate of Aβ accumulation measured with PiB in our DS cohort was 6.9 (2.1) CL/year, compared to rates of 4-5 CL/year in sporadic AD as measured with PiB. 62,63The accelerated deposition of Aβ observed in DS may contribute to the earlier onset of tau accumulation compared to sporadic AD.This may also suggest that, similar to ADAD, 64

1 . 2 . 3 .TA B L E 1
Our previous work comparing both Aβ L and CL in DS identified the following linear relation between metrics 29 : Systematic review: Positron emission tomography (PET) imaging studies in individuals with Down syndrome (DS) have revealed early and aggressive accumulation of amyloid-beta (Aβ).At the cross-sectional level, early tau deposition has been observed in individuals shortly after reaching Aβ-positivity (A+).To date, a longitudinal tau PET analysis has yet to be performed in DS.Interpretation: Our findings present the first longitudinal evaluation of tau change in AD with DS.By relating tau change to duration of A+ (A+ chronicity), it was observed that tau increases emerged within 2.5-5 years following A+.Future directions: A+ chronicity is a powerful tool that can accurately characterize the progression of AD biomarker change in DS.A direct comparison of tau change in DS and late-onset AD relative to A+ chronicity should be performed to better define potential AD treatments for individuals with DS.Demographics for the participants with Down syndrome.
generated by voxel normalization to gray matter cerebellum 80-100 min post-injection.No erosion or elimination of regions of focal uptake of AV-1451 was performed on the cerebellar gray matter ROI used for signal normalization.Using FreeSurfer v5.3.0,ROI masks from multiple brain regions were delineated from the T1w-MRI.The ROI masks were then combined to create ROIs consistent with regions used for NFT stages I-VI as described previously,48 which were used to extract NFT stage SUVRs.For all longitudinal AV-1451 scans, the NFT stage ROI masks from the baseline MRI time point were used for SUVR extraction.
chronicity of 0 years corresponds to the A+ cutoff of 13.3 Aβ L (18.0 CL).Numerical integration was performed with a 0.25-year step size across 200 iterations (corresponding to 50 years of disease time).The algorithm was restricted to integrating individual-level data for participants with two or more time points and terminated if the Aβ slope was negative or if the maximum number of iterations was reached.For participants with only a single Aβ scan, their A+ chronicity was numerically estimated by solving the fit of the nonparametric Aβ versus time curve at their given value of Aβ burden.For all other participants with longitudinal data, their Aβ trajectories were aligned to the curve based on the scan at which they became A+.Thus, the A+ chronicity estimate for these individuals was centered based on their true A+ time point during the study period.For individuals that did not reach the A+ threshold, their most recent scan was aligned to the Aβ versus time curve.If an individual's Aβ burden did not fall within the modeled range, their A+ chronicity was truncated to the earliest modeled value on the Aβ versus time curve.For each participant, an estimated age of A+ onset was determined by subtracting the estimated A+ chronicity at their reference scan from their chronological age at that scan.
each participant's observed longitudinal data to this trajectory based on their first A+ timepoint suggested reasonable model fits for later time points after an A+ chronicity time of 0 years (13.3Aβ L or 18.0 CL).

3
displays the annualized percent change in tau for each NFT region across the different bins of A+ chronicity, while Table 3 displays the Cohen's d effect size of annualized longitudinal change between the baseline and follow-up scans.For A+ chronicity < 0 years, small effect sizes (Cohen's d < 0.5) were observed between baseline and follow-up tau burden.Between A+ chronicity of 0-2.5 years, NFT regions I and III displayed large effect sizes (Cohen's d > 0.8), while NFT regions II and IV displayed medium effect sizes (Cohen's d > 0.5).The lower effect size in NFT region II compared to NFT regions I and III is likely an effect of both choroid plexus signal spillover and hippocampal atrophy.For an estimated A+ chronicity of 2.5-5 years, NFT regions I-II and IV-VI display medium effect sizes, while NFT region III displays a large effect size with respect to tau increase.Longitudinal tau increase is represented by medium to large effect sizes for all NFT stage regions after an A+ chronicity of 5 years, and beyond 10 years, no change is observed in Tau burden (presented as AV-1451 standardized uptake value ratios [SUVR]) within each neurofibrillary tau (NFT) stage with respect to chronological age and Aβ-positive (A+) chronicity.NFT regions I-II.When visualizing the annualized percent change in tau burden relative to A+ chronicity (Figure 5), tau burden in NFT regions I-III has the greatest magnitude of increase with lower A+ chronicity.As A+ chronicity increases, tau change in NFT regions I-II becomes static while NFT regions III-VI continue to rapidly increase.These findings reveal that the rates of tau change in DS are consistent with the proposed spatiotemporal tau deposition patterns outlined by Braak and Braak, and that the A+ chronicity measure can provide a timeline of tau progression throughout the development of AD.

4 TA B L E 2
Longitudinal tau change (presented as standardized uptake value ratio change per year [SUVR/year]) across each neurofibrillary tau (NFT) stage region with respect to Aβ-positive (A+) chronicity.Dashed lines represent zero tau change (horizontal) and an A+ chronicity of 0 years (vertical).Each colored point represents a single individual with Down syndrome (DS).Longitudinal rates of tau change (annualized % change presented as mean [95% CI]) across each NFT stage region of tau pathology with respect to A+ chronicity (years).

5
Annualized % change in tau burden within each neurofibrillary tau (NFT) stage region categorized by bin of Aβ-positive (A+) chronicity.Error bars represent the 95% confidence interval. of AD progression can be achieved compared to age or other indirect measures of disease stage.Comparing longitudinal tau PET change to the A+ chronicity metric identified a short latency period between A+ onset and early-stage tau deposition in the DS population.For an A+ chronicity greater than 0 years, nearly all individuals with DS had elevated tau and increases in tau burden in NFT stage regions I-VI.Significant increases were observed within the first 2.5 years of A+ in NFT regions I-IV, while NFT regions V-VI showed significant increases within 5 years of A+ onset.These findings recapitulate the Braak staging pattern of tau 27 and provide new evidence of a spatiotemporal pattern of tau progression in DS.Many individuals displayed very early increases in the AV-1451 signal in NFT region II near an A+ chronicity of 0 years.These early signal increases were likely attributed to signal spill-in from off-target choroid plexus binding, but this was not tested.With increased A+ chronicity, several individuals with DS displayed decreases in tau PET signal in the entorhinal cortex and hippocampus, and these changes were likely attributed to partial volume effects, atrophy, and ventricular enlargement observed with AD progression in DS.Similarly, Here, we present findings that resolve some of the challenges associated with age-based measures of EYO by directly linking the time course of AD progression to well-characterized biomarker measures.By using A+ chronicity, or the estimated years from Aβ onset, a more disease-specific timeline TA B L E 3 Effect size (Cohen's d [95% CI]) between baseline and follow-up AV-1451 SUVR for each NFT stage region.
50 may have more aggressive Aβ and tau phenotypes compared to late-onset sporadic AD.While a plateau emerged in the rate of Aβ accumulation in DS, none of the participants in our cohort displayed a plateau in Aβ burden.In sporadic AD, it has been shown that Aβ levels plateau late in the disease stage.50Thelack of observed Aβ plateau in our cohort may be a result of the young age of our participants (mean age = 39.2 (SD = 8.50) years), who are approximately 15 years from AD onset.Future longitudinal studies with ABC-DS will explore the Aβ burden in older adults with DS as they age, which may provide insight on whether an Aβ plateau is reached in this population.
The measure of A+ chronicity can be a powerful tool to implement in AD treatment and prevention trials.Because A+ chronicity is modeled from well-characterized biomarker data, it resolves many of the challenges associated with age-based recruitment and minimizes the