Association of Acquired and Heritable Factors With Intergenerational Differences in Age at Symptomatic Onset of Alzheimer Disease Between Offspring and Parents With Dementia

This cohort study examines the associations of acquired and heritable factors associated with intergenerational differences in age at symptomatic onset of Alzheimer disease among offspring of parents with Alzheimer disease.

History of tobacco abuse: Present (1) or absent (0). Deemed present when participant or collateral source endorsed a lifetime tobacco use exceeding 30 "pack years" (# of packs per day x number of years of tobacco use).
History of alcohol abuse: Present (1) or absent (0). Deemed present when the participant or collateral source endorsed a history of prior or current alcohol abuse, where alcohol abuse was defined as a pattern of drinking "too much alcohol too often" that interferes with daily life.
Retrospective reporting of age-at-symptomatic onset (AAO): Present (1) or absent (0). Deemed present when the participant's AAO was determined by retrospective report of the participant and collateral source (versus diagnosis / designation during prospective follow-up).
Heritable factors APOE ε4 allele status: Participants were defined according to the number of APOE ε4 alleles: participants with one copy had genotypes ε4/2, ε4/3; participants with two copies had genotypes ε4/4. APOE genotyping was performed as previously described, 1 using DNA extracted from peripheral blood samples using standard procedures.
Polygenic AD risk score: Calculated as detailed in the manuscript Methods.

GWAS data
Knight ADRC participants were genotyped with the Illumina 660K, OmiExpress or Human Core exome. As part of routine quality control steps, single-nucleotide polymorphisms (SNPs) with minor allele frequency <1%, call rates <98%, Hardy-Weinberg equilibrium p-values >10 -6 and individuals with >2% missing genotypes were removed before imputation. Chromosome SNPs were analyzed to verify sex identification. Each genotyping array was imputed, separately, using SHAPEIT/IMPUTE2 with the 1000 Genomes Project as the reference panel. All genotypes with dosage levels <0.9 for all three possible genotypes or with information scores <0.3 were excluded. Variants out of Hardy-Weinberg equilibrium (p<1x10 -6 ) or with a genotyping rate below 95% were also omitted. After imputation all data from the different arrays were combined. Population structure was inferred by principal component (PC) analysis using PLINK v1.9. PLINK v1.9 was also used to find duplicate and related individuals (first cousins or more proximate) who were eliminated from the analyses. APOE ε2, ε3 and ε4 isoforms were determined by genotyping rs7412 and rs429358 using Taqman genotyping technology as previously described. 2

Polygenic Risk Scores
Polygenic risk scores (PRS) were calculated as explained elsewhere. 3 Briefly, to derive the weighted PRS for AD-AAO, the odds ratios were modelled as reported in IGAP 4  . PLINK v1.9 was used to calculate the PRS choosing the score function and the no-mean-imputation option to disallow imputed scores. The resulting mean was corrected by multiplying the allele count (log OR Score).

Whole Exome Sequencing
Exome libraries were prepared using Agilent's SureSelect Human All Exon kits V3 and V5 or Roche VCRome using a HiSeq2000 with paired end reads, with a mean depth of coverage of 50× to 150×. Alignment was conducted against GRCh37.p13 genome reference. Variant calling was performed following GATK's 3.6 Best Practices (https://software.broadinstitute.org/gatk/best-practices/) and restricted to Agilent's V5 kit plus a 100bp of padding added to each capture target end. We used BCFTOOLS (https://samtools.github.io/bcftools/bcftools.html) to decompose multiallelic variants into biallelic prior to variant quality control. Variant Quality Score Recalibration (VQSR) was performed separately for SNPs and INDELs. Only those SNPs and indels that fell above the 99.9 confidence threshold, as indicated by WQSR, were considered for analysis. Variants within low complexity regions were removed. Non-polymorphic variants and those outside the expected ratio of allele balance for heterozygosity calls (ABHet=0.3-0.7) were removed. Additional hard filters implemented included quality depth (QD ≥7 for indels and QD≥2 for SNPs), mapping quality (MQ≥40), fisher strand balance (FS≥200 for indels and FS≥60 for SNPs), Strand Odds Ratio (SOR≥10 for indels and SOR≥3 for SNPs), Inbreeding Coefficient (IC ≥-0.8 for indels) and Rank Sum Test for relative positioning of reference versus alternative alleles within reads (RPRS≥-20 for indels and RPRS≥-8 for SNPs). We used PLINK1.9 (https://www.cog-genomics.org/plink2/ibd) to remove variants that were out of Hardy-Weinberg equilibrium (p-value <1×10 -6 ), with a genotype calling rate below 95%, with differential missingness between cases versus controls, WES vs WGS, or among different sequencing platforms (p-value<1×10 -6 ).
AD biomarker data CSF amyloid-β peptide 42 (Aβ42), total tau (tTau), and phosphorylated tau 181 (pTau) were measured with the corresponding Elecsys immunoassays on the Roche cobas e601 analyzer. Amyloid PET scans ( 11 C-Pittsburgh compound B; PiB) or Florbetapir ( 18 F-AV-45) were acquired on a Siemens Biograph mMR PET/MR scanner and attenuation corrected with a corresponding CT. Data were processed using an ROI approach using FreeSurfer software. Regional PIB or AV45 values were converted to standardized uptake value ratios (SUVRs) using cerebellar grey as a reference and partial volume corrected using a regional spread function approach. 5 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.
CSF biomarker positivity was defined as pTau/Aβ42 >0.0198, 6 which has very high concordance with amyloid PET positivity, defined as a mean cortical standardized uptake value ratio (SUVR) of >1.42 for PIB 7