Genotype inference from aggregated chromatin accessibility data reveals genetic regulatory mechanisms

Genome biology · 2025-03-30 · 5 citations

BACKGROUND: Understanding the genetic causes underlying variability in chromatin accessibility can shed light on the molecular mechanisms through which genetic variants may affect complex traits. Thousands of ATAC-seq samples have been collected that hold information about chromatin accessibility across diverse cell types and contexts, but most of these are not paired with genetic information and come from distinct projects and laboratories. RESULTS: We report here joint genotyping, chromatin accessibility peak calling, and discovery of quantitative trait loci which influence chromatin accessibility (caQTLs), demonstrating the capability of performing caQTL analysis on a large scale in a diverse sample set without pre-existing genotype information. Using 10,293 profiling samples representing 1454 unique donor individuals across 653 studies from public databases, we catalog 24,159 caQTLs in total. After joint discovery analysis, we cluster samples based on accessible chromatin profiles to identify context-specific caQTLs. We find that caQTLs are strongly enriched for annotations of gene regulatory elements across diverse cell types and tissues and are often linked with genetic variation associated with changes in expression (eQTLs), indicating that caQTLs can mediate genetic effects on gene expression. We demonstrate sharing of causal variants for chromatin accessibility across human traits, enabling a more complete picture of the genetic mechanisms underlying complex human phenotypes. CONCLUSIONS: Our work provides a proof of principle for caQTL calling from previously ungenotyped samples and represents one of the largest, most diverse caQTL resources currently available, informing mechanisms of genetic regulation of gene expression and contribution to disease.

Comparison of plasma p‐tau217/Aβ42, p‐tau217, and Aβ42/Aβ40 biomarkers by race to detect Alzheimer's disease

Alzheimer s & Dementia · 2025-08-01 · 19 citations

INTRODUCTION: In clinically and racially diverse training and test sets, we evaluated and validated Alzheimer's disease (AD) plasma biomarkers: phosphorylated tau-217 (p-tau217), amyloid beta 1-42/1-40 (Aβ42/Aβ40), and p-tau217/Aβ42. METHODS: F-florbetapir positron emission tomography (PET) data. In the training set (n = 289; 28% pBlack), we used receiver-operating characteristic (ROC) analysis to calculate two cut points (0.95 sensitivity and specificity) to detect amyloid PET positivity. Cut points were validated in an independent test set (n = 846; 12% pBlack). RESULTS: In the test set, plasma p-tau217/Aβ42 had the highest accuracy (pBlack: 0.88, pWhite: 0.91) and lowest proportion of intermediate classifications (≤0.16), with no classification difference by race (odds ratio [OR] = 1.45, 95% confidence interval [CI] = 0.66-2.95, p = 0.33). False negatives were more likely to be cognitively normal (vs mild cognitive impairment [MCI]; OR = 7.79, 95% CI = 2.33-40.81, p = 5.1e-05) than correct classifications. DISCUSSION: Plasma p-tau217/Aβ42 had high accuracy to detect AD, with comparable performance across race. HIGHLIGHTS: , p-tau217, and Aβ42/Aβ40 in diverse training and test sets. In models, race × amyloid PET interactions were not significant. Plasma p-tau217/Aβ42 had the highest accuracy in people who self-identified as Black/African American (pBlack; 0.88) and people who self-identified as White (pWhite; 0.91). Plasma p-tau217/Aβ42 had the lowest proportion of intermediate cases (range = 0.09-0.16). False negatives were more likely to have normal cognition, female sex, and high body mass index.

MRI Correlates of Executive Function in Cognitively Normal Older Subjects

Alzheimer s & Dementia · 2024-12-01

Abstract Background Cerebral small vessel disease (CSVD) is the most prevalent cause of vascular cognitive impairment. Executive function impairment and white matter (WM) lesions occur early in CSVD, the latter typically in periventricular WM (PVWM) ‐ the least perfused brain region. Accounting for vascular risk factors (VRF) and amyloid status in cognitively normal older subjects, we assessed the relationship between Trail‐making test (TMT) performance and MRI measures preceding WM lesions: arterial spin labeling (ASL)‐derived cerebral blood flow (CBF) and diffusion tensor imaging (DTI) indices in normal‐appearing WM (NAWM) and normal‐appearing PVWM (NAPVWM). Method 133 cognitively normal adults (>50y) from Penn Alzheimer’s Disease Research Center were included. TMT:B minus TMT:A (TMT:B‐A) was used as a metric of executive function. We categorized participants into amyloid‐positive (A+) and amyloid‐negative (A‐) groups based on PET. VRFs included hypertension, diabetes, obesity, hypercholesteremia and smoking. WM hyperintensity (WMH) masks were derived from FLAIR MRI using a deep‐learning model to calculate WMH volume and to mask WMH voxels. We extracted mean CBF and DTI indices (fractional anisotropy (FA) and mean diffusivity (MD)) in NAWM and NAPVWM using a perfusion‐based PVWM mask. NAPVWM relative CBF (rCBF) was defined relative to global CBF. Using linear regression, we assessed the associations of TMT:B‐A with VRF presence and imaging metrics in each group adjusting for age, sex and education. The relative importance of these variables was measured through random forest regression. Result The mean (SD) age was 70 (6) with 64% females. 23 subjects were A+ and 79 were A‐. Neither VRF presence nor imaging metrics were significantly associated with TMT:B‐A in A+ group. In A‐ group, after adjustment for covariates, VRF (Beta= 0.319, P=0.009), NAPVWM rCBF (Beta= ‐0.240, P=0.035)), NAWM MD (Beta=0.266, P=0.022) and NAPWM MD (Beta=0.255, P=0.029) were significantly associated with TMT:B‐A. WMH volume was not associated with TMT:B‐A (P= 0.668). In A‐ group, NAPWM MD and NAPWM rCBF were the most important determinants of TMT:B‐A relative to the other variables in the random forest model (Figure 1). Conclusion CBF and MD in NAPVWM are associated with executive function in normal aging and might be an early biomarker of CSVD.

Comparison of plasma p‐tau₂₁₇ and Aβ₄₂/Aβ₄₀ biomarkers by race to detect Alzheimer’s disease

Alzheimer s & Dementia · 2024-12-01 · 1 citations

Abstract Background To guide blood biomarker implementation to detect Alzheimer’s disease (AD), evaluations in clinically and racially diverse samples are needed. Here, we compare plasma biomarkers phosphorylated tau 217 (p‐tau 217 ), β‐amyloid (Aβ) 1‐42/1‐40 (Aβ 42 /Aβ 40 ) ratio, and p‐tau 217 /Aβ 42 ratio in a University of Pennsylvania (UPenn) sample that includes self‐identified White and Black/African American individuals with and without cognitive impairment, and independently validate findings in the first 100 AD Neuroimaging Initiative (ADNI) autopsy participants. Methods UPenn inclusion criteria were plasma p‐tau 217 , Aβ 42 , and Aβ 40 (n = 213) assayed using the automated Fujirebio platform, self‐reported race (156 White [including 1 Hispanic White, 2 multiracial] and 57 Black), and 18 F‐florbetaben PET scan to determine Aβ positivity by visual read (87 Aβ+, 126 Aβ‐). Participants were cognitively Normal (n = 128) or Impaired (n = 85) according to National Alzheimer’s Coordinating Center (NACC) Uniform Dataset 3.0 criteria. Linear models tested if plasma biomarkers (dependent variable, log‐transformed) differed by race and/or Aβ PET status (±). To confirm observed racial differences, linear models tested the interaction of race by global Aβ PET standardized uptake value ratio (SUVR). Models covaried for age at plasma, sex, 2021 national area deprivation index (ADI), and APOE ε4. Receiver operating characteristic (ROC) with bootstrapping (2000 iterations) compared biomarker performance (area under the curve [AUC]) and calculated 2 thresholds at 0.95 sensitivity and 0.95 specificity. Plasma p‐tau 217 cutpoints were validated in ADNI autopsy sample (n = 100); high/intermediate AD neuropathologic change considered AD+. Results P‐tau 217 /Aβ 42 had the highest performance to discriminate Aβ+ from Aβ‐ (Table 1A,1B; all AUC≥0.90), and p‐tau 217 showed excellent accuracy in ADNI sample (0.96; Table 1B). Linear models showed that Black individuals had significantly higher Aβ 42 (β = 0.098, p = 0.0014); no difference was observed for p‐tau 217 or p‐tau 217 /Aβ 42 (both p ≥0.43) (Figure 1). A Race X Aβ PET SUVR interaction confirmed this difference for plasma Aβ 42 (β = ‐0.44, p = 0.00051; Figure 2). Conclusion Findings indicate plasma p‐tau 217 levels do not differ by race, while plasma Aβ 42 /Aβ 40 was higher in Black individuals even after accounting for differences in Aβ PET SUVR and ADI. Future work will test biomarkers in a larger sample, and examine the influence of comorbidities on differences in plasma levels and accuracy.

Tau Magnitude and eXent–Tau‐MaX: Capturing tau spread and local accumulation provides a robust metric of tau burden

Alzheimer s & Dementia · 2024-12-01

Abstract Background Measures of tau burden have typically relied upon measures of magnitude, such as mean standardized uptake value ratio (SUVR), or extent, such as number of tau positive regions. However, heterogenous patterns of tau spread and accumulation present challenges to using these measures in isolation to quantify tau burden. Therefore, we hypothesized that a combined measure of tau magnitude and extent (Tau‐MaX) would outperform either measure in isolation. Method Baseline 18 F‐flortaucipir PET scans from 916 participants from ADNI (n = 741) and Penn Aging Brain Cohort (ABC, n = 175) were used for this study. SUVRs, using an inferior cerebellar reference region, were converted to Tau Pathology Indices (TPI) based on mixed‐Gaussian models (Figure 1). Regions with TPI > 0 (reflecting tau positivity) were weighted by region size and summed to form a measure of Tau Magnitude and eXtent (Tau‐MaX, Figure 1). This process was carried out globally and within two common meta‐ROIs: meta‐temporal and Braak 5/6. For each of these three meta‐ROIs, we compared diagnostic group differences, association with Centiloid values from amyloid PET, correlation with Fujirebio Lumipulse p‐tau 217 measures, and delayed recall performance using Tau‐MaX, mean SUVR, or extent of tau positivity. Comparison between associations was performed using Hittner’s Z. Results Group differences and results from regression analyses are shown in Table 1. There was a stepwise increase in all measures across diagnostic groups, although extent measures were more sensitive for detecting differences between CN Aβ+ and CN Aβ‐ individuals. Similarly, extent measures were more closely associated with Centiloid values in Aβ+ individuals in the meta‐temporal and Braak ROIs, but Tau‐MaX was more sensitive globally. Tau‐MaX had a stronger relationship with plasma p‐tau 217 compared to either SUVR or extent. Finally, Global Tau‐Max and meta‐temporal Tau‐Max and SUVR were the best predictors of delayed recall performance. Conclusion Tau‐MaX provides a robust metric of tau burden by incorporating both magnitude and extent. Further, the region agnostic approach may be particularly useful in more heterogenous cohorts that do not follow canonical Braak patterns of spread and accumulation. Longitudinal analyses of these measures will help identify their utility as markers of therapeutic response.

A trans-ancestral meta-analysis of genome-wide association studies reveals loci associated with childhood obesity

UNC Libraries · 2024-03-07 · 2 citations

Although hundreds of genome-wide association studies-implicated loci have been reported for adult obesity-related traits, less is known about the genetics specific for early-onset obesity and with only a few studies conducted in non-European populations to date. Searching for additional genetic variants associated with childhood obesity, we performed a trans-ancestral meta-analysis of 30 studies consisting of up to 13 005 cases (≥95th percentile of body mass index (BMI) achieved 2-18 years old) and 15 599 controls (consistently <50th percentile of BMI) of European, African, North/South American and East Asian ancestry. Suggestive loci were taken forward for replication in a sample of 1888 cases and 4689 controls from seven cohorts of European and North/South American ancestry. In addition to observing 18 previously implicated BMI or obesity loci, for both early and late onset, we uncovered one completely novel locus in this trans-ancestral analysis (nearest gene, METTL15). The variant was nominally associated with only the European subgroup analysis but had a consistent direction of effect in other ethnicities. We then utilized trans-ancestral Bayesian analysis to narrow down the location of the probable causal variant at each genome-wide significant signal. Of all the fine-mapped loci, we were able to narrow down the causative variant at four known loci to fewer than 10 single nucleotide polymorphisms (SNPs) (FAIM2, GNPDA2, MC4R and SEC16B loci). In conclusion, an ethnically diverse setting has enabled us to both identify an additional pediatric obesity locus and further fine-map existing loci.

Characterization of non-coding variants associated with transcription factor binding through ATAC-seq-defined footprint QTLs in liver

bioRxiv (Cold Spring Harbor Laboratory) · 2024-09-25 · 1 citations

Abstract Non-coding variants discovered by genome-wide association studies (GWAS) are enriched in regulatory elements harboring transcription factor (TF) binding motifs, strongly suggesting a connection between disease association and the disruption of cis-regulatory sequences. Occupancy of a TF inside a region of open chromatin can be detected in ATAC-seq where bound TFs block the transposase Tn5, leaving a pattern of relatively depleted Tn5 insertions known as a “footprint”. Here, we sought to identify variants associated with TF-binding, or “footprint quantitative trait loci” (fpQTLs) in ATAC-seq data generated from 170 human liver samples. We used computational tools to scan the ATAC-seq reads to quantify TF binding likelihood as “footprint scores” at variants derived from whole genome sequencing generated in the same samples. We tested for association between genotype and footprint score and observed 693 fpQTLs associated with footprint-inferred TF binding (FDR &lt; 5%). Given that Tn5 insertion sites are measured with base-pair resolution, we show that fpQTLs can aid GWAS and QTL fine-mapping by precisely pinpointing TF activity within broad trait-associated loci where the underlying causal variant is unknown. Liver fpQTLs were strongly enriched across ChIP-seq peaks, liver expression QTLs (eQTLs), and liver-related GWAS loci, and their inferred effect on TF binding was concordant with their effect on underlying sequence motifs in 80% of cases. We conclude that fpQTLs can reveal causal GWAS variants, define the role of TF binding site disruption in disease and provide functional insights into non-coding variants, ultimately informing novel treatments for common diseases. Graphical Abstract We leverage footprinting methods to infer transcription factor binding likelihood genome-wide across 170 liver ATAC-seq samples and implicate 693 SNPs with a genetic influence on binding. Unlike other comparable approaches, this analytical method is not limited in resolution by the constraints of linkage disequilibrium, and can prioritize likely causal variants at GWAS loci for subsequent experimental validation.

Spatial Pattern of Medial Temporal Lobe Cross‐Sectional and Longitudinal Structural Change in Cognitively Normal Individuals

Alzheimer s & Dementia · 2024-12-01

Abstract Background The medial temporal lobe's (MTL) early involvement in tau pathology makes it a key focus in the development of preclinical Alzheimer’s disease (AD) biomarkers. ROI analyses in prior studies reported significant MTL structural differences in cognitively normal individuals with and without β‐amyloid (A+/‐CN). Pointwise analysis, offering spatial information of early neurodegeneration, has potential to pinpoint “signature regions” of pathological change, but has been underexplored in the MTL. This study employs a specialized pointwise analysis pipeline to examine the spatial pattern of MTL structural change in subgroups dichotomized by both β‐amyloid and tau status in a large cohort of CN individuals. Methods A dataset of 3036 CN (A‐/A+: 1270/1558, Table 1) individuals from ADNI, HABS, A4 and ABC were analyzed. We extracted MTL regional thickness maps from MRI using tailored pipelines, ASHS‐T1 and CRASHS. For participants with prospective longitudinal MRI (five years follow‐up), regional maps of longitudinal atrophy rate were extracted using SkelDBM. Subjects with cross‐sectional tau PET available (N=563) were further divided into A and T subgroups by tracer uptake. General linear modeling was performed on each surface point to investigate cross‐sectional and longitudinal MTL structural group differences (detailed in Figure 1) and their correlation with MTL tau burden in All/A+/A‐ CN. Age and sex were covariates and cluster‐level multiple comparison correction was performed. Results A+CN demonstrated a significantly faster atrophy rate than A‐CN across the whole MTL, primarily driven by A+T+CN individuals (Figure 1‐b). Notably, A‐T+CN showed significantly faster atrophy rate in the entorhinal cortex (ERC) and Brodmann area 35 (BA35), the earliest sites of tau pathology (Figure 1‐b, second column). Figure 2‐b displays an MTL‐wise significant correlation between atrophy rate and tau in All/A+/A‐ CN. In both analyses, cross‐sectional effects are consistently weaker than longitudinal ones, but have some significant clusters in ERC and BA35. Conclusions Pointwise analysis revealed extensive tau‐associated accelerated neurodegeneration in the MTL in preclinical AD. Furthermore, accelerated atrophy was observed in early Braak regions in A‐CN with evidence of tau pathology, potentially driven by primary age‐related tauopathy (PART). These pointwise longitudinal MTL measures provide sensitive measures that may allow for disease monitoring in preclinical AD.

Variant-to-function mapping of late-onset Alzheimer’s disease GWAS signals in human microglial cell models implicates <i>RTFDC1</i> at the <i>CASS4</i> locus

bioRxiv (Cold Spring Harbor Laboratory) · 2024-08-22 · 3 citations

ABSTRACT Late-onset Alzheimer’s disease (LOAD) research has principally focused on neurons over the years due to their known role in the production of amyloid beta plaques and neurofibrillary tangles. In contrast, recent genomic studies of LOAD have implicated microglia as culprits of the prolonged inflammation exacerbating the neurodegeneration observed in patient brains. Indeed, recent LOAD genome-wide association studies (GWAS) have reported multiple loci near genes related to microglial function, including TREM2 , ABI3 , and CR1 . However, GWAS alone cannot pinpoint underlying causal variants or effector genes at such loci, as most signals reside in non-coding regions of the genome and could presumably confer their influence frequently via long-range regulatory interactions. We elected to carry out a combination of ATAC-seq and high-resolution promoter-focused Capture-C in two human microglial cell models (iPSC-derived microglia and HMC3) in order to physically map interactions between LOAD GWAS-implicated candidate causal variants and their corresponding putative effector genes. Notably, we observed consistent evidence that rs6024870 at the GWAS CASS4 locus contacted the promoter of nearby gene, RTFDC1 . We subsequently observed a directionallly consistent decrease in RTFDC1 expression with the the protective minor A allele of rs6024870 via both luciferase assays in HMC3 cells and expression studies in primary human microglia. Through CRISPR-Cas9-mediated deletion of the putative regulatory region harboring rs6024870 in HMC3 cells, we observed increased pro-inflammatory cytokine secretion and decreased DNA double strand break repair related, at least in part, to RTFDC1 expression levels. Our variant-to-function approach therefore reveals that the rs6024870-harboring regulatory element at the LOAD ‘ CASS4’ GWAS locus influences both microglial inflammatory capacity and DNA damage resolution, along with cumulative evidence implicating RTFDC1 as a novel candidate effector gene.

Regional cerebral blood flow reflects both neurodegeneration and microvascular integrity across the Alzheimer's continuum

Alzheimer s & Dementia · 2024-12-03 · 9 citations

INTRODUCTION: Alzheimer's disease (AD) typically involves both neurodegenerative and vascular pathologies, each associated with reductions in cerebral blood flow (CBF). However, it remains unclear whether vascular and neural contributions to regional CBF can be differentiated. METHODS: Using 3D background-suppressed arterial spin labeled perfusion magnetic resonance imaging, we evaluated regional CBF in a cohort of 257 participants across the AD continuum and assessed the impact of risk factors for both AD and small vessel disease (SVD) on regional CBF. RESULTS: Vascular risk factors (VRFs) were associated with reduced CBF in normal-appearing periventricular white matter, while amyloid positivity was associated with reduced CBF in the posterior cingulate cortex and precuneus. Putative SVD-sensitive regions in white matter exhibited diagnosis-related CBF changes comparable to those in typical AD cortical regions. DISCUSSION: Spatial patterns of hypoperfusion may differentiate AD and VRF-related effects on regional CBF. Our findings also support the contribution of SVD in AD pathogenesis. HIGHLIGHTS: We used 3D background-suppressed pCASL MRI to evaluate CBF across the AD continuum. Putative SVD-sensitive regions in white matter exhibited diagnosis-related CBF changes. AD and/or SVD risk correlated with reduced CBF in AD and/or SVD-related regions. VRFs were associated with more widespread CBF reductions than amyloid positivity. Spatial patterns of hypoperfusion may differentiate AD and VRF-related effects.