Arne De Roeck

Mutations in ABCA7 in a Belgian cohort of Alzheimer's disease patients: a targeted resequencing study

BACKGROUND ABCA7 was identified as a risk gene for Alzheimers disease in genome-wide association studies (GWAS). It was one of the genes most strongly associated with risk of Alzheimers disease in a Belgian cohort. Using targeted resequencing, we investigated ABCA7 in this cohort with the aim to directly detect rare and common variations in this gene associated with Alzheimers disease pathogenesis. METHODS We did massive parallel resequencing of ABCA7 after HaloPlex target enrichment of the exons, introns, and regulatory regions in 772 unrelated patients with Alzheimers disease (mean age at onset 74·6 years [SD 8·9]) recruited at two memory clinics in Flanders, Belgium, and 757 geographically matched community-dwelling controls (mean age at inclusion 73·9 years [8·0]). After bioinformatic processing, common variants were analysed with conditional logistic regression and rare variant association analysis was done in Variant Association Tools. To explore an observed founder effect, additional unrelated patients with Alzheimers disease (n=183, mean age at onset 78·8 years [SD 6·0]) and control individuals (n=265, mean age at inclusion 56·9 years [10·8]) from the same cohort who had not been included in massive parallel resequencing because of insufficient biosamples were screened for the ABCA7 frameshift mutation Glu709fs with Sanger sequencing. The effect of loss-of-function mutations on ABCA7 expression was investigated with quantitative real-time PCR in post-mortem brains of patients (n=3) and control individuals (n=4); nonsense mediated mRNA decay was investigated in lymphoblast cell lines from three predicted loss-of-function mutation carriers from the cohort of 772 patients with Alzheimers disease. FINDINGS An intronic low-frequency variant rs78117248 (minor allele frequency 3·8% in 58 patients with Alzheimers disease and in controls 1·8% in 28 controls) showed strongest association with Alzheimers disease (odds ratio 2·07, 95% CI 1·31-3·27; p=0·0016), and remained significant after conditioning for the GWAS top single nucleotide polymorphisms rs3764650, rs4147929, and rs3752246 (2·00, 1·22-3·26; p=0·006). We identified an increased frequency of predicted loss-of-function mutations in the patients compared with the controls (relative risk 4·03, 95% CI 1·75-9·29; p=0·0002). One frameshift mutation (Glu709fs) showed a founder effect in the study population, and was found to segregate with disease in a family with autosomal dominant inheritance of Alzheimers disease. Expression of ABCA7 was reduced in the two carriers of loss-of-function mutations found only in patients with Alzheimers disease (Glu709fs and Trp1214*) compared with four non-carrier controls (relative expression 0·45, 95% CI 0·25-0·84; p=0·002) and in lymphoblast cell lines from three carriers of Glu709fs compared with those from two non-carrier controls. INTERPRETATION We propose that a low-frequency variant can explain the association between ABCA7 and Alzheimers disease, and the evidence of loss-of-function mutations in this risk gene suggests that partial loss-of-function of ABCA7 could be a potential pathogenetic mechanism of Alzheimers disease. FUNDING Belgian Science Policy Office Interuniversity Attraction Poles program P7/16, Alzheimer Research Foundation, King Baudouin Foundation AB Fund, Methusalem Excellence Program initiative of the Flemish Government, Flanders Impulse Program on Networks for Dementia Research, Research Foundation Flanders, Agency for Innovation by Science and Technology Flanders, University of Antwerp Research Fund, and European Unions Seventh Framework Programme for Research, Technological development and Demonstration (AgedBrainSYSBIO).

A 22-single nucleotide polymorphism Alzheimer's disease risk score correlates with family history, onset age, and cerebrospinal fluid Aβ42

The ability to identify individuals at increased genetic risk for Alzheimers disease (AD) may streamline biomarker and drug trials and aid clinical and personal decision making.

Phenotypic characteristics of Alzheimer patients carrying an ABCA7 mutation.

Objective: To generate a clinical and pathologic phenotype of patients carrying rare loss-of-function mutations in ABCA7, identified in a Belgian Alzheimer patient cohort and in an autosomal dominant family. Methods: We performed a retrospective review of available data records, medical records, results of CSF analyses and neuroimaging studies, and neuropathology data. Results: The mean onset age of the mutation carriers (n = 22) was 73.4 ± 8.4 years with a wide age range of 36 (54–90) years, which was independent of APOE genotype and cerebrovascular disease. The mean disease duration was 5.7 ± 3.0 years (range 2–12 years). A positive family history was recorded for 10 carriers (45.5%). All patient carriers except one presented with memory complaints. The 4 autopsied brains showed typical immunohistochemical changes of late-onset Alzheimer disease. Conclusions: All patients carrying a loss-of-function mutation in ABCA7 exhibited a classical Alzheimer disease phenotype, though with a striking wide onset age range, suggesting the influence of unknown modifying factors.

Structural variants identified by Oxford Nanopore PromethION sequencing of the human genome

We sequenced the Yoruban NA19240 genome on the long read sequencing platform Oxford Nanopore PromethION for benchmarking and evaluation of recently published aligners and structural variant calling tools. In this work, we determined the precision and recall, present high confidence and high sensitivity call sets of variants and discuss optimal parameters. The aligner Minimap2 and structural variant caller Sniffles are both the most accurate and the most computationally efficient tools in our study. We describe our scalable workflow for identification, annotation, and characterization of tens of thousands of structural variants from long read genome sequencing of an individual or population. By discussing the results of this genome we provide an approximation of what can be expected in future long read sequencing studies aiming for structural variant identification.

Accurate characterization of expanded tandem repeat length and sequence through whole genome long-read sequencing on PromethION.

Tandem repeats (TRs) can cause disease through their length, sequence motif interruptions, and nucleotide modifications. For many TRs, however, these features are very difficult - if not impossible - to assess, requiring low-throughput and labor-intensive assays. One example is a VNTR in ABCA7 for which we recently discovered that expanded alleles strongly increase risk of Alzheimer’s disease. Here, we investigated the potential of long-read whole genome sequencing to surmount these challenges, using the high-throughput PromethION platform from Oxford Nanopore Technologies. To overcome the limitations of conventional base calling and alignment, we developed an algorithm to study the TR size and sequence directly on raw PromethION current data. We report the long-read sequencing of multiple human genomes (n = 11) using only a single sequencing run and flow cell per individual. With the use of fresh DNA extractions, DNA shearing to approximately 20kb and size selection, we obtained an average output of 70 gigabases (Gb) per flow cell, corresponding to a 21x genome coverage, and a maximum yield of 98 Gb (30x genome coverage). All ABCA7 VNTR alleles, including expansions up to 10,000 bases, were spanned by long sequencing reads, validated by Southern blotting. Classical approaches of TR length estimation suffered from low accuracy, low precision, DNA strand effects and/or inability to call pathogenic repeat expansions. In contrast, our novel NanoSatellite algorithm, which circumvents base calling by using dynamic time warping on raw PromethION current data, achieved more than 90% accuracy and high precision (5.6% relative standard deviation) of TR length estimation, and detected all clinically relevant repeat expansions. In addition, we identified alternative TR sequence motifs with high consistency, allowing determination of TR sequence and distinction of VNTR alleles with homozygous length. In conclusion, we validated the robustness of single-experiment whole genome long-read sequencing on PromethION, a prerequisite for application of long-read sequencing in the clinic. In addition, we outperformed Southern blotting, enabling improved characterization of the role of expanded ABCA7 VNTR alleles in Alzheimer’s disease, and opening new opportunities for TR research.

Human structural variation identified by Oxford Nanopore PromethION

1 Chaisson, Mark J. P., Ashley D. Sanders, Xuefang Zhao, Ankit Malhotra, David Porubsky, Tobias Rausch, Eugene J. Gardner, et al. 2017. “Multi-Platform​​ Discovery​​ Of​​ Haplotype-Resolved Structural​​ Variation​​ In​​ Human​​ Genomes.” bioRxiv. 2 De Coster, Wouter, Svenn D’Hert, Darrin T. Schultz, Marc Cruts, and Christine Van Broeckhoven. 2018. “NanoPack: Visualizing and Processing Long Read Sequencing Data.” Bioinformatics Introduction The majority of the structural variants in the genome, defined as changes in copy number or location of elements > 50 bp, remain hidden with currently dominant technologies. Long read sequencing has the advantage of a higher mappability, the ability to span breakpoints and align uniquely to repetitive sequences. For benchmarking and evaluation of tools we have sequenced the Yoruban reference genome NA19240, part of the HapMap and 1000 genomes project and well characterized using modern technologies1, allowing independent validation of our findings. We reached 258 gigabase or 80x coverage and our data is publicly available on ENA (PRJEB26791). Disclaimer: we have received consumables from ONT for this project and have been reimbursed for presenting at conferences.

An intronic VNTR affects splicing of ABCA7 and increases risk of Alzheimer’s disease

Mutations leading to premature termination codons in ATP-Binding Cassette Subfamily A Member 7 (ABCA7) are high penetrant risk factors of Alzheimer’s disease (AD). The influence of other genetic variants in ABCA7 and downstream functional mechanisms, however, is poorly understood. To address this knowledge gap, we investigated tandem repetitive regions in ABCA7 in a Belgian cohort of 1529 AD patients and control individuals and identified an intronic variable number tandem repeat (VNTR). We observed strong association between VNTR length and a genome-wide associated signal for AD in the ABCA7 locus. Expanded VNTR alleles were highly enriched in AD patients [odds ratio = 4.5 (1.3–24.2)], and VNTR length inversely correlated with amyloid β1–42 in cerebrospinal fluid and ABCA7 expression. In addition, we identified three novel ABCA7 alternative splicing events. One isoform in particular—which is formed through exon 19 skipping—lacks the first nucleotide binding domain of ABCA7 and is abundant in brain tissue. We observed a tight correlation between exon 19 skipping and VNTR length. Our findings underline the importance of studying repetitive DNA in complex disorders and expand the contribution of genetic and transcript variation in ABCA7 to AD.

CONTRIBUTION OF RARE DELETERIOUS ABCA7 MUTATIONS TO A BELGIAN EARLY-ONSET ALZHEIMER’S DISEASE COHORT

onset AD cases for the identification of novel variants and genes implicated on AD. Methods:We have performed whole-exome or whole-genome sequencing on 1,077 clinically diagnosed LOAD and 440 non-demented relatives. All samples were recruited by the Knight ADRC or the NIA-LOAD family study group. Variant discovery was performed following GATK’s best practice followed by stringent quality control. We focused our analysis on nonsynonymous variants with a minor allele frequency <1% in the general population (ExAC). We perform genome-wide scans to examine the presence of causative and low frequency coding variants in the most common neurodegenerative diseases, as well as in novel candidate genes. Results:Known pathogenic mutations were found in 4.42% of the families as well as an enrichment of non-synonymous variants in AD and frontotemporal lobular dementia (FTLD) genes. We have detected potential risk variants that present perfect segregation in several families of 7 members in addition to few novel candidate genes with genome-wide significance associated to disease risk. Conclusions:Known pathogenic and low frequency coding variants in AD, FTD and PD genes can be found in clinical AD cases; indicating that more stringent selecting criteria and genetic testing must be performed before incorporating new samples into analysis. We have identified some preliminary candidate variants and genes that would confer risk to AD; network analysis and replication are underway.

DELETERIOUS ABCA7 MUTATIONS CONTRIBUTE TO EARLY-ONSET ALZHEIMER’S DISEASE AND ARE SUBJECT TO TRANSCRIPT RESCUE MECHANISMS

Background:Genes that are expressed in a cell-specific fashion in the human brain were identified from transcriptome-wide (RNAseq) expression measures of purified cell populations (Zhang Y. et al, Neuron, 2016). Leveraging these findings, we estimated the relative proportions of cell-types in a temporal cortex (TCX), tissue-based, transcriptome dataset (RNAseq: Allen M. et al, Sci Data, 2016) of 268 subjects with a pathological diagnosis of Alzheimer’s disease (AD; N1⁄480), progressive supranuclear palsy (PSP; N1⁄482), pathologic aging (PA; N1⁄430) or normal controls (CON; N1⁄476), to identify disease relevant changes and cell-specific vulnerabilities. Methods:We generated expression residuals, adjusting for appropriate covariates, for cell–specific genes (Zhang. Y et al, Neuron 2016) expressed in our dataset (neurons1⁄4415; astrocytes1⁄4234; oligodendroglia1⁄4117; microglia1⁄4352, endothelia1⁄498 genes). We used principal components analysis to generate a surrogate variable (signed eigenvector 1: EV1) representing cell-specific transcripts for each cell type. Each cell-specific EV1 was used as a quantitative variable to test for association with diagnosis, age, sex or APOEε4 genotype (linear regression, R software). Pairwise correlations were tested between all cell-specific EV1 levels (Pearson correlation, R). Analyses were done on the complete cohort, adjusting for diagnosis, and for each individual diagnostic group. We also performed a genome-wide association study (GWAS; linear regression, PLINK) for each cellspecific EV1 on the complete cohort. Results: Significant positive correlations (r2>0.5; p<2.2E-16) were identified between endothelial EV1 and astrocyte or microglia EV1 levels. Neuronal EV1 was negatively correlated with EV1 levels for all other cell types. These correlations were similar across the diagnostic groups. EV1 for astrocytes, microglia, and endothelia were higher, and for neurons lower, in AD subjects compared to other groups, with similar trends observed also for female sex and APOEε4. GWAS did not identify genome-wide significant variants (p<1E08), although suggestive associations include variants at the TMEM106B locus with astrocytes (p1⁄49.0E-07), and an intergenic locus (Chr2) flanked by NXPH2 and LRP1B(p1⁄42.1E-08) with microglia. Conclusions:We have identified brain cell type variability across different diagnosis groups and genetic associations that may influence these. These findings imply specific vulnerabilities for different cell types. We are performing similar analyses in cerebellum, and additional human and mouse brain transcriptome datasets.

TRANSCRIPTOME ANALYSIS IN BLOOD AND BRAIN IDENTIFIES GENE EXPRESSION REGULATION AND CORRESPONDING QUANTITATIVE TRAIT LOCI IN ALZHEIMER’S DISEASE

Background: TDP-43 (transactive response DNA-binding protein 43kDa) proteinopathy is closely associated with Alzheimer’s disease (AD), and affects clinical outcome in AD. TMEM106B rs1990622is a risk allele for TDP-43 proteinopathy, and its implications in clinical presentations of AD remains to be elucidated. Methods: Our subjects are from Religious Orders Study and the Rush Memory and Aging Project (ROS-MAP), two communitybased longitudinal cohort studies of older adults. Our analyses included 841 subjects with genotyping, cognitive, and pathology data (beta-amyloid, paired helical filament tau, and TDP-43 stage). Cognitive decline was defined as the slope of change in composite cognitive z-score in a general linear mixed model, controlling for age at enrollment, sex, and education. The imputed dosage of TMEM106B rs1990622 was tested for its association with cognitive decline after adjusting for AD pathology. Then, TDP-43 was included in the model, and a subsequent mediation analysis (quasi-Bayesian Monte Carlo method with 10,000 simulations) was done to evaluate whether the TMEM106Bvariant’s effect was mediated by TDP-43 proteinopathy. Of note, all analyses were adjusted for study cohort (ROS vs MAP), genotyping platform, and first three principal components from the genotype covariance matrix. Results:TMEM106B rs1990622was associated with more rapid cognitive decline when controlled for the AD pathology (EE1⁄4-0.0098, 95% confidence interval (CI) -0.018 to -0.0016, p1⁄40.020). When TDP-43 stage was also controlled, the association of rs1990622 with cognitive decline loses statistical significance (p1⁄40.078). A mediation analysis shows that about 15% of the association between rs1990622 and more rapid cognitive decline is mediated by more advanced TDP-43 stages (average causal mediation effect p<0.0001). Conclusions: TMEM106B rs1990622, a known TDP-43 proteinopathy risk variant, is associated with more rapid cognitive decline in people with similar degree of AD pathology, and this association is mediated by more advanced TDP-43 stages. Nonetheless, the association of rs1990622 with more rapid cognitive decline in AD is only partially explained by the link between rs1990622 and more advanced TDP-43 stages.