# paper.pdf ## Metadata - PDFFormatVersion=1.7 - Language=Englisch: Großbritannien - IsLinearized=false - IsAcroFormPresent=false - IsXFAPresent=false - IsCollectionPresent=false - IsSignaturesPresent=false - CreationDate=D:20260630044948+02'00' - Creator=Adobe InDesign 18.3 (Windows) - ModDate=D:20260722042054+02'00' - Producer=Adobe PDF Library 17.0; modified using iText® 5.5.13.4 ©2000-2024 iText Group NV (SPRINGER SBM; licensed version) - Trapped=False ## Contents ### Page 1 © The Author(s) 2026. Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as yougive appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified thelicensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images orother third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to thematerial. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation orexceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit. †I gn a z i o Stefano Piras and Ana Werneck Capuano contributedequally to this work and share first authorship. †Matt J. Huentelman and Emily Rogalski jointly supervised this workand share senior authorship. *Correspondence: Ignazio Stefano Piras Full list of author information is available at the end of the article Abstract Background As longevity increases and the population over age 65 expands, advancing age remains the mostreliable predictor of cognitive decline, highlighting the need to identify biological mechanisms that supportexceptional cognitive aging. We tested whether lower inherited risk of Alzheimer’s disease (AD) dementia predictsSuperAger status (adults ≥ 80 years with episodic memory at least as good as middle-age adults) using prospectivelyenrolled SuperAgers and Cognitively Average Controls (Controls) from the multisite SuperAging Research Initiative. Methods We studied 231 participants (SuperAgers n = 142; Controls n = 89). We confirmed that the genetic ancestrystructure across groups was comparable. We evaluated whether APOE status (ε2, ε3, ε4) and three AD polygenic riskscores (PRS) derived from large contemporary Genome-Wide Association Studies (GWAS) (PRSLambert, PRSWightman,PRSBellenguez) predicted SuperAging status using logistic regression models adjusted for age, sex, and education,considering ancestry interactions. Results APOE allele and genotype distributions did not differ between groups, and neither APOE nor any of thethree PRS predicted SuperAger status. Results were unchanged when accounting for global non-European or Africanancestry or principal components. In this well-characterized cohort, neither APOE nor contemporary PRS explainedSuperAger status. Conclusions These findings suggest that the exceptional late-life memory phenotype that is characteristic ofSuperAging is not explained by common-variant AD genetic risk captured by APOE or contemporary AD PRS,motivating a deeper investigation of potential rare genetic variations and experiential factors contributing toexceptional cognitive aging. SuperAging is not the inverse of common-variant Alzheimer’s risk: evidence acrossgenetic ancestries Ignazio Stefano Piras1,2*†, Ana Werneck Capuano3†, Amanda Cook Maher4, Rhiana Schafer3, Anna Bonfitto1,Serena Song1, Francis Taguinod1, Felicia Goldstein5, Angela Roberts6,7,8, Ozioma Okonkwo9, Adam Martersteck3,The SuperAging Research Initiative, Matt J. Huentelman1,2† and Emily Rogalski3† Piras et al. Alzheimer's Research & Therapy (2026) 18:168 [ ![http://crossmark.crossref.org/dialog/?doi=10.1186/s13195-026-02124-2&domain=pdf&date_stamp=2026-6-30]() ](http://crossmark.crossref.org/dialog/?doi=10.1186/s13195-026-02124-2&domain=pdf&date_stamp=2026-6-30) ![BioMed Central Logo]() ### Page 2 Background Episodic memory typically declines with advancing age,paralleling age-related changes in hippocampal and dis-tributed cortical systems that support memory function.Yet a small subset of older adults, hereafter referred toas SuperAgers, maintain episodic memory performanceinto their eighth decade and beyond at levels at leastcomparable to middle-age adults. Studying this excep-tional phenotype offers a powerful opportunity to inter-rogate mechanisms that support memory function underconditions of highest age-related risk. By focusing onthe extreme right tail of late-life cognition, SuperAg-ing enhances spectrum-wide models of cognitive aging,resilience, and resistance by sharpening inference aboutbiological and experiential factors that sustain mem-ory despite advancing age and accumulating pathology.Critically, the SuperAging phenotype is defined usingprospectively specified, age and performance-basedcognitive criteria, enabling systematic identificationand enrollment of individuals with exceptional late-lifememory. The multisite SuperAging Research Initia-tive was established to harmonize deep phenotypingacross SuperAgers and cognitively average older adults,providing a uniquely rigorous framework for isolatingdeterminants of exceptional cognitive aging and distin-guishing them from factors associated with typical agingand dementia. As such, SuperAging provides a tractablemodel for studying resilience and resistance mechanismsin late life. Genetic factors represent a natural launch point forinterrogating factors contributing to exceptional memoryaging phenotypes. The APOE locus remains the stron-gest common-variant determinant of Alzheimer’s disease(AD) risk, with ε4 increasing risk and lowering age atonset and ε2 generally conferring protection. In contrast,contemporary polygenic risk scores (PRS) summarizethousands of additional variants into aggregate indices ofinherited liability. However, how these liabilities operatein adults over age 80 remains uncertain. Our prior workusing an early, prospectively enrolled SuperAging cohort,predominantly of European ancestry, found no differencein AD polygenic hazard between SuperAgers and age-similar controls, suggesting exceptional late-life mem-ory is not simply explained by reduced common-variantAD risk and may instead reflect resilience or protectivemechanisms [,]. In contrast, large multi-cohort anal-yses report lower ε4 and higher ε2 frequencies amongnon-Hispanic White individuals with superior cogni-tion, with evidence for similar patterns across ancestralgroups []. Other cohorts demonstrate that superiormemory confers reduced progression to mild cognitiveimpairment (MCI) or dementia despite comparableAPOE ε4 or amyloid burden [,]. Together, these find-ings raise unresolved questions about whether inher-ited AD risk distinguishes exceptional cognitive agingin advanced age, or whether genetic effects are modifiedby age, ancestry, or phenotype definition, questions thatrequire rigorously defined, deeply phenotyped cohorts ofthe oldest-old. Here, we leverage the multisite SuperAging ResearchInitiative to test the role of inherited AD genetic liabilityin participants with exceptional late-life memory using aprospectively defined, performance-based cognitive phe-notype. By studying a large, well-characterized cohort ofSuperAgers and similar-age cognitively average controls,and incorporating ancestry-aware analytic approaches,this study provides a rigorous evaluation of whethercommon-variant AD genetic risk, captured by APOEand contemporary polygenic risk scores, distinguishesindividuals with exceptional memory performance inadvanced age. This design enables assessment of geneticeffects across ancestries and clarifies whether excep-tional cognitive aging reflects reduced inherited AD riskor mechanisms of resilience beyond common-variantliability. Methods Participants Participants included prospectively enrolled SuperAgersand cognitively average 80 + -year-old Controls from theoriginal Chicagoland cohort and through the multisiteSuperAging Research Initiative through 2024. Establishedin 2021, the SuperAging Research Initiative is a NationalInstitute on Aging and McKnight Brain Research Foun-dation-funded consortium enrolling participants acrossfive regions in North America (Ann Arbor/Detroit, MI;Atlanta, GA; Chicago, IL; London, ON, Canada; andMadison, WI). The Initiative builds upon and extends theoriginal Chicagoland SuperAging cohort established andoperationalized by Rogalski and colleagues by expand-ing geographic and demographic representation andenabling harmonized, multidomain phenotyping acrosssites [–]. Eligibility criteria at enrollment for the SuperAgingResearch Initiative included age ≥ 80 years, English-speaking adults with the absence of neurological disor-ders known to affect cognition (e.g., Parkinson’s disease,multiple sclerosis), and no significant uncontrolled medi-cal conditions at enrollment. Participants were classifiedas SuperAgers or cognitively average controls based onprospectively specified, a priori cognitive performance Keywords SuperAger, Resilience, Resistance, APOE, Polygenic risk score, Genetic ancestry, Alzheimer’s disease ### Page 3 criteria consistent with our prior SuperAging studies [,]. Briefly, SuperAgers were required to demonstrateexceptional episodic memory performance, defined asscoring at or above the mean for 56–64-year-old adultson the delayed free recall trial of the Rey Auditory VerbalLearning Test (RAVLT); this corresponds to a standardscore ≥ 10 based on Mayo’s Older Americans NormativeStudies (MOANS) norms and is equivalent to recall-ing ≥ 9 words at delay [,]. In addition, SuperAgerswere required to perform within or above age-appropri-ate normative expectations on non-memory cognitivemeasures, including naming (Boston Naming Test, BNT,or Multilingual Naming Test, MiNT), Category FluencyTest (Animals), and executive functioning (Trail Mak-ing Test Part B), using age-, sex-, education-, and race-adjusted normative values as appropriate [–]. Controls were required to perform within the averagenormative range for their age on the RAVLT delayed freerecall and memory measure and within or above nor-mative expectations for the same non-episodic mem-ory cognitive measures used to classify SuperAgers. AllSuperAgers and Controls had a Clinical Dementia Rating(CDR) equal to zero, confirming the absence of clinicalimpairment in both groups. All study procedures were approved by the appropriateInstitutional Review Boards, including Advarra, ClinicalTrials Ontario, Northwestern University, and the Univer-sity of Chicago. Written informed consent was obtainedfrom all participants in accordance with the Declarationof Helsinki. This expanded, multisite cohort provides a rigorousframework for evaluating common genetic influences onexceptional late-life memory using larger, more diversesamples and updated polygenic risk scoring approaches. Genotyping DNA from SuperAgers and Controls participants wasextracted from whole blood using the QIAamp DNABlood Midi Kits (Qiagen, USA) and analyzed via wholeexome sequencing (WES). Samples were collected andsequenced between 2013 and 2025 using multiple librarypreparation kits (TruSeq 62 Mb Exome Enrichment Kitor the Nextera 62 Mb Rapid Capture Expanded ExomeKit (Illumina, Inc., San Diego, CA, USA)), following themanufacturers’ protocols. Sequencing was performedon a HiSeq 2500 System using 100-bp paired-end reads(Illumina, Inc., San Diego, CA, USA). FASTQ files fromSuperAgers and Controls were aligned to the human ref-erence genome GRCh38 using the Parabricks pipelinev4.0.1. Joint genotyping was performed using GLnexusv1.4.1. Variants were then filtered to exclude those with:quality score (QUAL) < 300, depth (DP) < 10, genotypequality (GQ) < 20, single nucleotide polymorphisms(SNPs) located within 10 bp of insertion/deletions(INDELS), and allelic balance for heterozygous geno-types outside 0.2–0.8. Only biallelic SNPs were retained,and INDELS were excluded. The resulting filtered variantcall format (VCF) files were converted to PLINK format[] for further SNP and sample filtering. Specifically, weremoved SNPs with a genotyping rate < 95%, those failingHardy–Weinberg Equilibrium (HWE) with P < 1.0 × 10–06,and with minor allele frequency (MAF) < 1%. Sampleswith genotyping call rates below 20% were excluded. Alower threshold than the standard 95–99% was appliedto minimize sample loss in this limited dataset, as sub-sequent imputation was expected to recover missinggenotypes with high accuracy. Duplicates were identi-fied using the kinship-based inference for Genome-Wide Association Studies (KING) robust estimator [],with duplicate pairs defined as those with KING-kin-ship ≥ 0.354. Heterozygosity was estimated using the –hetcommand from PLINK2, then removing samples belowor above ± 3 standard deviations from the sample average.In total, we removed four samples as follows: low geno-typing rate (n = 1), duplicates (n = 2), and sex mismatches(n = 1), resulting in a total of 142 SuperAgers and 89 Con-trols for imputation. Imputation was conducted usingthe TOPMED server [] with the vr3 reference panel,retaining only imputed variants with r2 ≥ 0.3. Anotherround of SNPs quality controls was conducted on theimputed data (HWE and MAF, using the same cutoffs asabove), after which we obtained a total of 2,812,549 SNPswith MAF ≥ 1%. APOE genotype was determined usingSNPs rs429358 and rs7412 []. Ancestry Genetic ancestry analysis was estimated using ADMIX-TURE [], incorporating reference data from 3,202individuals from the 1000 Genomes Project Phase 3[], representing five major population groups: Ameri-can Mixed (AMR), African (AFR), European (EUR), EastAsian (EAS), and South Asian (SAS). ADMIXTUREanalyses were conducted with K = 5, corresponding tothe five population groups, using fivefold cross-validationto evaluate model performance. Continuous ancestryproportions were derived for each participant and usedas covariates and interaction terms in logistic regres-sion models evaluating associations between APOE, ADpolygenic risk scores (PRS), and SuperAger status. Spe-cifically, AFR and non-European (non-EUR) ancestryproportions were examined as potential effect modifi-ers. Principal component analysis (PCA) was performedto capture major axes of genetic variation and accountfor population structure. Ancestry-based PCA was con-ducted using SNPs with MAF ≥ 5% following linkagedisequilibrium (LD) pruning in PLINK2 (command –indep-pairwise 50 5 0.2). This command performs LD ### Page 4 pruning in sliding windows of 50 SNPs, advancing thewindow by 5 SNPs each time, and removes one SNP fromany pair with r2 ≥ 0.2. The resulting principal componentscapture the major axes of genetic variation and wereincluded as covariates in genetic association models toaccount for population structure. Polygenic risk score The AD polygenic risk scores (AD-PRS) were estimatedusing PRsice-2 []. Three AD-PRSs were generatedbased on summary statistics from widely used genome-wide association study (GWAS) meta-analyses repre-senting successive stages of discovery: PRSLambert [],PRSWightman [] and PRSBellenguez []. These GWAS wereselected to provide complementary coverage of the com-mon-variant genetic architecture of AD across increasingsample sizes and analytic frameworks spanning a decadeof discovery. Lambert et al., 2013 (International Genomics ofAlzheimer’s Project (IGAP) stage-I/II) was the first largeconsortium meta-analysis and established many of thecanonical late-onset AD loci used in early PRS bench-marking; using this score enables continuity with a vastprior literature and enables historical comparability. Thephase-1 GWAS from Lambert et al. includes 17,008 ADcases and 37,154 controls and a dataset of 7,055,881SNPs. Wightman et al., 2021 expanded the sample sizeto > 1.1 million individuals (using clinical and proxy phe-notypes), adding new loci and updated effect sizes; a PRSfrom this study tests whether newer signals discoveredwith greater power improve discrimination beyond theprior published IGAP-based scores. The meta-analysisfrom Wightman et al. includes 90,338 AD cases (46,613proxy) and 1,036,225 (318,246 proxy) controls assessedacross a total of 12,206,525 SNPs. Bellenguez et al.,2022 further increased power (64,498 clinically diag-nosed or ‘proxy’ AD cases and 677,663 controls), cover-ing 21,101,115 SNPs and reported ~ 75 risk loci (42 novelat the time), refining pathway inferences and variantweights; evaluating a PRS from this meta-analysis pro-vides a state-of-the-art benchmark anchored in the cur-rent consensus locus set. Summary statistics genomics coordinates from Lam-bert et al. and Wightman et al. were converted usinggenome liftover with the MungeSumstats R-package [](GRCh37 to GRCh38). SNPs names were reformattedto the structure chromosome:bp:a1:a2, where a1 and a2represent the allelic variants in alphabetical order. Pal-indromic SNPs were removed from the dataset. Targetdatasets were filtered to retain variants with minor allelefrequency ≥ 1%, SNP and sample call rates ≥ 95%, andHWE P ≥ 1 × 10⁻5. SNP names were harmonized to thesummary statistics format using the command –set-all-var-ids '@:#:$1:$2' from PLINK2 []. AD-PRS were computed using PRSice-2 [] acrossa range of GWAS P-value thresholds (Benjamini andHochberg adjusted P: 5 × 10⁻⁸ to 0.05) with LD clump-ing (r2 = 0.1; 250 kb window). SNP weights were derivedfrom GWAS β coefficients. Optimal P-value thresholdsfor SNP inclusion were determined using logistic regres-sion in an independent, neuropathologically confirmedAD cohort (TGEN2; 1,019 AD cases and 591 cognitivelynormal controls), modeling AD status as the outcomeand standardized PRS as the predictor, adjusting for age,sex, and the first three ancestry principal components.Details of genotyping and imputation in the TGEN2cohort have been described previously []. The opti-mized PRS thresholds identified in TGEN2 were subse-quently applied to the SuperAging and Control datasetsfor primary analyses. Statistical analysis Differences in demographic characteristics and raw neu-ropsychological test scores between the SuperAgers andControls were examined using t-tests, Pearson’s χ2 test orMann–Whitney U-tests as appropriate. Allele and geno-type frequencies were compared between the SuperAg-ers and Controls using a Pearson’s χ2 test, quantifying theeffect sizes using Cramer’s V statistics. Associations between APOE and SuperAger statuswere evaluated using logistic regression models. Binaryindicator variables were created to denote the presenceof at least one APOE ε2, ε3, and ε4 allele; mixed-riskgenotypes ε2/ε4 (n = 7) and ε3/ε4 (n = 35) were excludedfrom allele-specific analyses. The comparison betweenSuperAgers and Controls on APOE genotype (APOE ε2,ε3, and ε4 allele) was carried out using separate logisticregression models including terms for age, sex, years ofeducation, and either PRS or APOE. Age and years ofeducation were mean-centered by subtracting the sam-ple mean. Interaction models with APOE also includedself-reported race (Black/White) and genetic ancestryproportions for AFR and non-EUR, estimated using theadmixture method and centered by subtracting the sam-ple mean. Polygenic risk score (PRS) associations with SuperAgerstatus were examined using logistic regression modelsadjusted for age, sex, and education. To evaluate poten-tial effect modification, interaction models additionallyincluded self-reported race (Black/White) and continu-ous genetic ancestry proportions for African (AFR) andnon-European (non-EUR) ancestry, estimated using theadmixture method and mean-centered. Sensitivity analysis was conducted by removing theAPOE region from the summary statistics, as APOE isthe strongest genetic factor associated with AD and thelarge effect size of this region can dominate polygenicrisk score estimations and obscure the contribution of ### Page 5 other variants. The SNPs removed were defined basedon the genomic interval showing significant associationin the original AD GWAS (Benjamini and Hochbergadjusted P < 0.05) and ranged from chr19:43,349,456 bpto chr19:49,932,605 bp. Genetic association analysis was performed using alloverlapping SNPs included in the PRS derived fromthe three GWAS, excluding low frequency variants(MAF < 1%). The same logistic regressions were fitted butnow with a term for each SNP instead of the PRS, andusing PLINK (–glm command). As a conservative thresh-old of significance, we used 5.0 × 10–08, which representsthe standard genome-wide significance level derivedfrom the total number of independent common variantsin the human genome []. For exploratory purposes,we also considered a less stringent and more tailoredthreshold based on the number of independent SNPsin our dataset after including only the PRS SNPs. Thenumber of independent variants was determined usingPLINK LD pruning as for the PCA. P-values from theassociation tests were then adjusted with the Bonferronimethod, accounting for the number of independent SNPsafter pruning (n = 43) and the number of models evalu-ated (n = 5), yielding an adjusted significance thresholdof (P < 2.3 × 10–04). SNPs annotation was conducted usingANNOVAR []. Results The study design is summarized in Fig.A. After SNPand sample quality controls (see for details), weincluded a total of 142 SuperAger and 89 Control partici-pant samples. Mean age was 83.7 (SD = 4.06) in Super-Ager and 84.7 (SD = 4.38) in Control (Table). Globalancestry proportions showed substantial overlap betweenSuperAger and Control groups, indicating comparableancestry structures across the two groups (Fig.B–C). ### Page 6 The distribution of APOE allele and genotype fre-quency is reported in Table S1 and Fig.D. Groupwisecomparisons of APOE status did not differ significantlybetween SuperAgers and Controls. Using allele- andgenotype-level χ2 tests, neither overall allele frequenciesnor genotype distributions showed evidence of group dif-ferences (alleles: χ2 = 2.81, Cramer’s V = 0.078, P = 0.245;genotypes: χ2 = 4.566, Cramer’s V = 0.141, P = 0.335;Table S2). In models adjusted for age, sex, and years ofeducation, APOE ε2, ε3, and ε4 allele status were notassociated with odds of SuperAger classification (ε2:β = 0.094, SE = 9.434, P = 0.828; ε3: β = 1.226, SE = 0.735,P = 0.095; ε4: β = − 0.252, SE = 0.380, P = 0.507; Table).Descriptively, the proportion with at least one ε2 allelewas similar across groups (SuperAger 12.7%, Control13.1%), as was the proportion carrying at least one ε4allele (SuperAger 15.7%, Control 19.0%) (Table S1). Mean AD-PRS values derived from the three discoveryGWAS (Fig S1 to S3) were broadly comparable betweenSuperAgers and Controls (PRSLambert: SA 1.30 × 10⁻3vs CTL 1.15 × 10⁻3; PRSWightman: SA 1.28 × 10⁻3 vs CTL0.91 × 10⁻3; PRSBellenguez: SA 2.57 × 10⁻3 vs CTL 2.47 × 10⁻3;Table, Fig.E). In adjusted logistic regression mod-els (age, sex, and years of education), none of the PRSwere associated with odds of SuperAger classification(PRSLambert: β = 1.565, SE = 23.108, P = 0.946; PRSWightman:β = 4.743, SE = 13.132, P = 0.718; PRSBellenguez: β = 2.238,SE = 70.728, P = 0.975; Table). Next, we tested if therewas a differential association between genetic predictors(PRS, APOE and PRS without APOE ε4) and race/ances-try, represented as (i) a non-EUR global ancestry propor-tion, (ii) AFR ancestry proportion, (iii) the first geneticprincipal component and (iv) self-reported race (indica-tor of Black race 1, White 0). There was no evidence of asignificant interaction (Table). We conducted exploratory SNP-level association anal-ysis using the 768 shared SNP variants across the threeAD-PRS (Table S4). No individual SNP was significantlyassociated with SuperAger status after correction formultiple testing (adjusted P < 2.3 × 10⁻4). A heatmapshowing the top 5 nominal associations for each model isreported in Fig. and all associations are summarized inTable S4. In the primary model (adjusted for age, sex, andyears of education), the top association was in the inter-genic region between PICALM and EED (OR = 0.463;P = 3.1 × 10–04; adj-P = 0.068), followed by other vari-ants within PICALM (OR = 0.560; P = 4.3 × 10–03; adj-P = 0.925); none survived multiple-testing correction.Interaction analysis (SNP x self-reported race), similarlyyielded nominal associations clustered in the MS4A genecluster, including SNPs between MS4A4E and MS4A4A(OR = 6.066; P = 0.017; adj-P = 1.000) and for a secondMS4A4E variant (OR = 5.785; P = 0.017; adj-P = 1.000),but all interaction effects were non-significant after cor-rection (Table S4). Finally, we screened our cohort for the presence of ADrare protective variants, including the APP Icelandic vari-ant (rs63750847; chr21:25,897,620), the PLCG2 P552Rvariant (rs72824905; chr16:81,908,423), and APOEChristchurch variant (rs121918393; chr19:44,908,756).We identified a single heterozygous SuperAger carryingthe PLCG2 P522R variant, but no carriers of the othermutations were identified in SuperAgers or Controls.Given our SuperAgers sample size of N = 142, this corre-sponds to a minor allele frequency (MAF) of 0.0035 forPLCG2 P552R, consistent with the frequencies reportedin gnomAD for European non-Finnish (0.0083) and ### Page 7 African/African American (0.0012). Current analyses donot suggest enrichment of AD-protective variants acrossour SuperAging sample. Discussion In this multisite, harmonized cohort of prospectivelyenrolled adults aged 80 years and older, we testedwhether lower inherited Alzheimer’s disease riskdifferentiates the SuperAger phenotype, defined byyouthful episodic memory, from a cognitively averagephenotype. There was no association between SuperAgerstatus and APOE (ε2/ε3/ε4) or any of the three AD poly-genic risk scores (AD-PRSLambert, AD-PRSWightman, AD-PRSBellenguez) in ancestry-adjusted logistic models. Therewas also no interaction between genetic ancestry (non-European or African ancestry proportions or by ancestryprincipal components) and APOE or the AD polygenicrisk scores. Taken together, these findings indicate thatcommon-variant AD risk, as captured by APOE and con-temporary AD-PRS, does not account for exceptionallate-life memory, but does not preclude contributionsfrom other biological or experiential mechanisms. The three PRS were drawn from successive, largeGWAS that together encompass canonical and newlydiscovered loci, providing state-of-the-art coverage ofcommon-variant liability; yet none predicted Super-Ager status in adjusted models. This divergence impliesthat genetic architectures linked to case–control AD riskmay be only weakly informative for the extreme righttail of late-life cognition, reinforcing the need to studySuperAgers with designs tailored to their phenotyperather than extrapolating from AD genetics alone. Theexpanded, ancestry-diverse SuperAging Research Initia-tive specifically enables such inquiry by re-examiningcommon genetic influences with larger, more representa-tive samples and updated scoring approaches. Importantly, while APOE allele frequencies and ADpolygenic risk scores did not distinguish SuperAgersfrom cognitively average older adults, both groups dif-fered meaningfully from individuals with AD, in whomthese common genetic risk factors are strongly enriched.This pattern indicates that lower inherited AD risk atthe group level is a shared feature of successful cognitiveaging into advanced age, but it is not sufficient to explainthe exceptional memory performance that defines theSuperAging phenotype. Thus, common AD genetic riskappears permissive rather than determinative; that is, itmay enable survival into late life without dementia, butit does not fully account for the preservation of youthfulepisodic memory among SuperAgers. This is not unex-pected, as AD polygenic risk scores are optimized forcase–control discrimination and may have reduced sensi-tivity for resolving variance at the upper extremes of cog-nitive performance among non-demented older adults.However, the present findings empirically demonstratethat approaches beyond AD-risk are required to explainexceptional late-life memory. Consistent with this inter-pretation, screening for rare AD-protective variants alsodid not suggest enrichment in the SuperAging cohort. Together, these findings place an upper bound on thecontribution of common-variant Alzheimer’s diseasegenetic risk to the SuperAging phenotype. Variants of ### Page 8 the magnitude captured by APOE and contemporaryAD polygenic risk scores do not fully account for excep-tional late-life memory relative to cognitively averageolder adults. While lower inherited AD risk appears tobe a shared feature of dementia-free cognitive aging intoadvanced age, it is not sufficient to explain the preserva-tion of youthful episodic memory that defines SuperAg-ing. Thus, common AD genetic risk may be permissive,enabling survival into late life without dementia, but notdeterminative of exceptional cognitive performance. Insum, SuperAging is not simply the inverse of AD risk asindexed by common variants. SuperAging may represent an active preservation ofmemory systems through mechanisms that are orthog-onal to those captured by case–control AD genetics.Importantly, SuperAging is defined phenotypically, andlikely reflects convergence of multiple biological andexperiential pathways rather than a single genetic deter-minant. These mechanisms could include enhancedneural maintenance, compensatory network dynamics,or life-course experiential factors (e.g., vascular health,physical activity, and other modifiable exposures) thatbuffer age-related cognitive decline even in the presenceof typical genetic risk. These factors are actively beinginvestigated through the multisite SuperAging ResearchInitiative. At the individual level, we observed a small number ofSuperAgers with elevated AD genetic risk (e.g., APOE ε4homozygosity or high PRS) who nonetheless maintainedyouthful cognition. Although not the focus of the pres-ent analyses, such individuals may be particularly infor-mative targets for future in-depth study for identifyingresilience mechanisms, analogous to cognitively intact“escapees” observed in autosomal Alzheimer’s disease[–]. The study has several strengths: consistent, a prioricognitive criteria for SuperAging and Controls; central-ized variant processing and imputation; explicit mod-eling of genetic ancestry using both global admixtureand principal components; and external optimizationof PRS thresholds in a neuropathologically confirmedcase–control resource before application to the SuperAg-ing cohort. Limitations include modest power for smallgenetic effects, limited power for ancestry-stratifiedanalyses within subgroups, potential constraints on PRSportability across ancestries, and the inability of cross-sectional analyses to establish temporal precedence.Additionally, the present analyses did not incorporatevascular, lifestyle, or behavioral factors, which may con-tribute to the SuperAging phenotype and represent animportant direction for future work. By demonstrating that common-variant AD geneticrisk does not explain exceptional memory in advanced ### Page 9 age, this work refines the biological scope of SuperAgingand underscores the need for discovery frameworks thatmove beyond disease risk toward mechanisms of cogni-tive preservation, resilience, and resistance across thelifespan. Future work would benefit from expansion ofmulti-ancestry samples, integration of rare-variant anal-yses and gene-environment measures, lifestyle factorsassessment, and measurements of the presence of ADpathology through biomarkers and/or at autopsy. Theassessment of underlying AD pathology (e.g., circulatingpTau) would allow us to investigate whether preservedcognition in SuperAgers reflects AD pathological bur-den, resilience, or both, an important direction currentlyunderway within the SuperAging Research Initiative.Finally, linking genetic profiles to longitudinal cognitivetrajectories will allow us to identify mechanisms thatsupport exceptional memory in very late life. Abbreviations AD Alzheimer’s disease PRS Polygenic risk score MCI Mild Cognitive Impairment RAVLT Rey Auditory Verbal Learning Test MOANS Mayo’s Older Americans Normative Studies BNT Boston Naming Test MiNT Multilingual Naming Test QUAL Quality score DP Depth GQ Genotype quality SNPs Single nucleotide polymorphisms (SNPs) INDELS Insertion/deletions VCF Variant call format HWE Hardy-Weinberg Equilibrium MAF Minimum allele frequency KING Kinship-based inference for Genome-Wide Association Studies AMR American Mixed AFR African EUR European EAS East Asian SAS South Asian PCA Principal component analysis LD Linkage Disequilibrium GWAS Genome-wide association study IGAP International Genomics of Alzheimer’s Project kb Kilobases Supplementary Information The online version contains supplementary material available at. Supplementary Material 1. Supplementary Material 2. Acknowledgements We are deeply grateful to the participants of the SuperAging ResearchInitiative and the original Chicagoland SuperAging cohort, whose time,commitment, and generosity make this research possible. Their willingnessto participate in longitudinal research has been essential to advancingunderstanding of exceptional cognitive aging. We also thank the studycoordinators, clinicians, psychometrists, research staff, and trainees across allparticipating sites for their dedication to rigorous data collection, participantengagement, and study coordination. We are particularly appreciative of thecommunity partners and outreach teams who have supported recruitmentand retention efforts and helped ensure that this work reflects the olderadults aging in our communities. This research was made possible throughthe collaborative efforts of the SuperAging Research Initiative, which wasdesigned to support harmonized, multisite investigation of exceptionalcognitive aging. We also acknowledge the participants and investigators ofthe external datasets used to support analytic validation. The SuperAging Research Initiative Trammell, AR1, Goldstein, F1, Parker, M1, Kaddoura, L1, Harris, G1, Ellison, A1,Saleh, S1, Qui D1, Ahmed H1, Chung, J1, Piras, IS2, Huentelman, MJ2, Bonfitto,A2, Stark, B2, Song, S2, Ayala, I3, Geula, C3, Mesulam, M3, Lim, A4, Swartz, RH4,Martersteck, A5, Seibert, K5, Kharitonova, M5, McDermott, S5, Zolliecoffer, C5,Addison, E5, Rogalski, EJ5, Peirce, H5, Engelmeyer, J5, Timpo, P5, Devine, R5,Schafer, R5, Moore, S5, Whitmore, H5, McCarthy, M5, Hearns, K5, Capuano, A5,Wang, S5, Maher, AC⁶, Rose, E⁶, Lincoln, GA⁶, Paulson, H⁶, Wathen, L⁶, Balog,V⁶, Bross, J⁶, Reader, J⁶, Bhaumik, A⁶, Warner, K⁶, Sturt, Y⁶, McIlroy, B⁷, VanOoteghem, K⁷, Beyer, K⁷, Thai, V⁷, Dunne, C⁷, Doyle-Thomas, K⁸, Okonkwo, OC⁹,Lose, S⁹, Pandos, A⁹, Bruce, M⁹, Uday, V⁹, Roberts, AC1⁰, Finger, E1⁰, Culum, I1⁰,Bartha, R1⁰, Orange, JB1⁰, Coleman, K1⁰, Jesso, S1⁰, Silveira, C1⁰, Narayan, E1⁰, Li,J1⁰, Aulakh, N1⁰, Tuncel, Y1⁰, Haller, B1⁰, Ofori, DO1⁰, Cardaio, A1⁰, El-Khoury, J.11 1Emory University, Atlanta, GA, United States. 2Translational Genomics Research Institute, Phoenix, AZ, United States. 3Northwestern University, Chicago, IL, United States. 4Sunnybrook Health Sciences Centre, Toronto, ON, Canada. 5University of Chicago, Chicago, IL, United States. 6University of Michigan, Ann Arbor, MI, United States. 7University of Waterloo, Waterloo, Ontario, Canada. 8Mohawk College, Hamilton, Ontario, Canada. 9University of Wisconsin, Madison, WI, United States. 10Western University, London, Ontario, Canada. 11MGH Harvard University, Boston, MA, United States. Authors’ contributions MJH, ISP, and ER contributed to the conceptualization of the study. All authorscontributed to data collection and curation. ISP and AWC conducted theformal analyses. ER acquired funding for the study. ISP, AWC, MJH, and ERcontributed to analysis planning and methodological design. ISP, AWC, MJH,and ER drafted the original manuscript. All authors contributed to manuscriptreview and editing and approved the final version. Funding This project is supported in part by the following NIH Research GrantsU19AG073153, R01AG045571, R56AG045571, 5R01AG067781, P30AG13854,P30AG072977, and by the McKnight Brain Research Foundation (MBRF). Thecontent is solely the responsibility of the authors and does not necessarilyrepresent the official views of the National Institutes of Health. Data availability The datasets used analyzed in the current study are available from thecorresponding author on reasonable request. Declarations Ethics approval and consent to participate All study procedures were approved by the appropriate Institutional ReviewBoards, including Advarra, Clinical Trials Ontario, Northwestern University, andthe University of Chicago. Written informed consent was obtained from allparticipants prior to study participation in accordance with the Declaration ofHelsinki. Competing interests The authors declare no competing interests. Author details 1Division of Early Detection and Prevention, The Translational GenomicsResearch Institute, a part of City of Hope, 445 N Fifth Street, Phoenix, AZ,USA 2The Arizona Alzheimer’s Consortium, Phoenix, AZ, USA 3Department of Neurology, The Healthy Aging & Alzheimer’s ResearchCare (HAARC) Center, University of Chicago, Chicago, IL, USA 4Department of Psychiatry, University of Michigan, Ann Arbor, MI, USA ### Page 10 5Department of Neurology, Goizuetta Alzheimer’s Disease ResearchCenter, Emory University School of Medicine, 6 Executive Park Drive,Atlanta, NEGA, USA 6Department of Computer Science, Western University, London, ON,Canada 7School of Communication Sciences and Disorders, Western University,London, ON, Canada 8Canadian Centre for Activity and Aging, London, ON, Canada 9Wisconsin Alzheimer’s Disease Research Center and Department ofMedicine, University of Wisconsin School of Medicine and Public Health,600 Highland Avenue, Madison, WI, USA Received: 5 March 2026 / Accepted: 17 June 2026 References 1. Spencer BE, Banks SJ, Dale AM, Brewer JB, Makowski-Woidan B, WeintraubS, et al. Alzheimer’s polygenic hazard score in SuperAgers: SuperGenes orSuperResilience? Alzheimers Dement Transl Res Clin Interv. 2022;8:e12321.. 2. Desikan RS, Fan CC, Wang Y, Schork AJ, Cabral HJ, Cupples LA, et al. Geneticassessment of age-associated Alzheimer disease risk: Development andvalidation of a polygenic hazard score. PLoS Med. 2017;14:e1002258.. 3. Durant A, Mukherjee S, Lee ML, Choi S-E, Scollard P, Klinedinst BS, et al.Evaluating the association of APOE genotype and cognitive resilience inSuperAgers. medRxiv. 2025;. 4. Dang C, Harrington KD, Lim YY, Ames D, Hassenstab J, Laws SM, et al. SuperiorMemory Reduces 8-year Risk of Mild Cognitive Impairment and DementiaBut Not Amyloid β-Associated Cognitive Decline in Older Adults. Arch ClinNeuropsychol. 2019;34:585–98.. 5. Fulton-Howard B, Goate AM, Adelson RP, Koppel J, Gordon ML, Alzheimer’sDisease Genetics Consortium, et al. Greater effect of polygenic risk score forAlzheimer’s disease among younger cases who are apolipoprotein E-ε4 carri-ers. Neurobiol Aging. 2021;99:101.e1-101.e9.. 6. Harrison TM, Weintraub S, Mesulam M-M, Rogalski E. Superior memory andhigher cortical volumes in unusually successful cognitive aging. J Int Neuro-psychol Soc. 2012;18:1081–5.. 7. Rogalski EJ, Gefen T, Shi J, Samimi M, Bigio E, Weintraub S, et al. Youthfulmemory capacity in old brains: anatomic and genetic clues from the North-western SuperAging Project. J Cogn Neurosci. 2013;25:29–36.. 8. Rogalski EJ, Martersteck A, Roberts AC, Huentelman MJ, Okonkwo OC, TimpoP, et al. The Multisite SuperAging Research Initiative: Enrollment and ScientificProgress. Alzheimer’s & Dementia. Wiley; 2024;20.. 9. Schmidt M. Rey auditory verbal learning test: a handbook. Western Psycho-logical Services; 2024. 10. Ivnik RJ, Malec JF, Smith GE, Tangalos EG, Petersen RC, Kokmen E, et al. Mayo’solder americans normative studies: Updated AVLT norms for ages 56 to 97.Clinical Neuropsychologist. 1992;6:83–104.. 11. Kaplan EF, Goodglass H, Weintraub S. Boston Naming Test. 2nd Edition, Lea &Febiger, Philadelphia, 1983. 12. Jefferson AL, Wong S, Gracer TS, Ozonoff A, Green RC, Stern RA. GeriatricPerformance on an Abbreviated Version of the Boston Naming Test. ApplNeuropsychol. 2007;14:215–23.. 13. Reitan RM. The relation of the Trail Making Test to organic brain damage. JConsult Psychol. 1955;19:393–4.. 14. Morris JC, Heyman A, Mohs RC, Hughes JP, van Belle G, Fillenbaum G, et al.The Consortium to Establish a Registry for Alzheimer’s Disease (CERAD). PartI. Clinical and neuropsychological assessment of Alzheimer’s disease. Neurol-ogy. 1989;39:1159–1159. 15. Heaton RK, Miller SW, Tayor MJ, Grant I. Revised Comprehensive Norms for anExpanded Halstead-Reitan Battery: Demographically Adjusted Neuropsy-chological Norms for African Americans and Caucasian Adults. PsychologicalAssessment Resources, 2004 16. Chang CC, Chow CC, Tellier LCAM, Vattikuti S, Purcell SM, Lee JJ. Second-generation PLINK: Rising to the challenge of larger and richer datasets.Gigascience. 2015.. 17. Manichaikul A, Mychaleckyj JC, Rich SS, Daly K, Sale M, Chen W-M. Robustrelationship inference in genome-wide association studies. Bioinformatics.2010;26:2867–73.. 18. Taliun D, Harris DN, Kessler MD, Carlson J, Szpiech ZA, Torres R, et al. Sequenc-ing of 53,831 diverse genomes from the NHLBI TOPMed Program. Nature.2021;590:290–9.. 19. Zhong L, Xie Y-Z, Cao T-T, Wang Z, Wang T, Li X, et al. A rapid and cost-effective method for genotyping apolipoprotein E gene polymorphism. MolNeurodegener. 2016;11:2.. 20. Alexander DH, Novembre J, Lange K. Fast model-based estimation of ances-try in unrelated individuals. Genome Res. 2009;19:1655–64.. 21. Byrska-Bishop M, Evani US, Zhao X, Basile AO, Abel HJ, Regier AA, et al.High-coverage whole-genome sequencing of the expanded 1000 GenomesProject cohort including 602 trios. Cell. 2022;185:3426-3440.e19.. 22. Choi SW, O’Reilly PF. PRSice-2: Polygenic Risk Score software for biobank-scaledata. Gigascience. 2019;8.. 23. Lambert JC, Ibrahim-Verbaas CA, Harold D, Naj AC, Sims R, Bellenguez C, etal. Meta-analysis of 74,046 individuals identifies 11 new susceptibility loci forAlzheimer’s disease. Nat Genet. 2013;45:1452–8.. 24. Wightman DP, Jansen IE, Savage JE, Shadrin AA, Bahrami S, Holland D, et al. Agenome-wide association study with 1,126,563 individuals identifies new riskloci for Alzheimer’s disease. Nat Genet. 2021;53:1276–82.. 25. Bellenguez C, Küçükali F, Jansen IE, Kleineidam L, Moreno-Grau S, Amin N, etal. New insights into the genetic etiology of Alzheimer’s disease and relateddementias. Nat Genet. 2022;54:412–36.. 26. Murphy AE, Schilder BM, Skene NG. MungeSumstats: a Bioconductor pack-age for the standardization and quality control of many GWAS summarystatistics. Bioinformatics. 2021;37:4593–6.. 27. Corneveaux JJ, Myers AJ, Allen AN, Pruzin JJ, Ramirez M, Engel A, et al.Association of CR1, CLU and PICALM with Alzheimer’s disease in a cohort ofclinically characterized and neuropathologically verified individuals. Hum MolGenet. 2010;19:3295–301.. 28. Fadista J, Manning AK, Florez JC, Groop L. The (in)famous GWAS P-valuethreshold revisited and updated for low-frequency variants. Eur J Hum Genet.2016;24:1202–5.. 29. Wang K, Li M, Hakonarson H. ANNOVAR: functional annotation of geneticvariants from high-throughput sequencing data. Nucleic Acids Res.2010;38:e164.. 30. Bellenguez C, Charbonnier C, Grenier-Boley B, Quenez O, Le Guennec K,Nicolas G, et al. Contribution to Alzheimer’s disease risk of rare variants inTREM2, SORL1, and ABCA7 in 1779 cases and 1273 controls. Neurobiol Aging.2017;59:220.e1-220.e9.. 31. Arboleda-Velasquez JF, Lopera F, O’Hare M, Delgado-Tirado S, Marino C,Chmielewska N, et al. Resistance to autosomal dominant Alzheimer’sdisease in an APOE3 Christchurch homozygote: a case report. Nat Med.2019;25:1680–3.. 32. Jonsson T, Atwal JK, Steinberg S, Snaedal J, Jonsson PV, Bjornsson S, et al. Amutation in APP protects against Alzheimer’s disease and age-related cogni-tive decline. Nature. 2012;488:96–9.. 33. Lopera F, Marino C, Chandrahas AS, O’Hare M, Villalba-Moreno ND, AguillonD, et al. Resilience to autosomal dominant Alzheimer’s disease in a Reelin-COLBOS heterozygous man. Nat Med. 2023;29:1243–52.. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims inpublished maps and institutional affiliations.