Moreno-Grau et al. Translational Psychiatry (2021)11:142 https://doi.org/10.1038/s41398-020-01145-1 Translational Psychiatry ARTICLE Open Access Long runs of homozygosity are associated with Alzheimer’s disease Sonia Moreno-Grau1,2, Maria Victoria Fernández 3,4, Itziar de Rojas 1,2, Pablo Garcia-González 1, Isabel Hernández1, Fabiana Farias 3,4, John P. Budde 3,4, Inés Quintela5, Laura Madrid 6, Antonio González-Pérez6, Laura Montrreal1, Emilio Alarcón-Martín1, Montserrat Alegret1, Olalla Maroñas5, Juan Antonio Pineda7, Juan Macías7, The GR@ACE study groupDEGESCO consortium, Marta Marquié 1,2, Sergi Valero1,2, Alba Benaque1, Jordi Clarimón2,8, Maria Jesus Bullido 2,9,10, Guillermo García-Ribas11, Pau Pástor 12, Pascual Sánchez-Juan2,13, Victoria Álvarez14,15, Gerard Piñol-Ripoll2,16, Jose María García-Alberca 17, José Luis Royo18, Emilio Franco-Macías19, Pablo Mir 2,20, Miguel Calero 2,21,22, Miguel Medina 2,21, Alberto Rábano2,21,23, Jesús Ávila2,24, Carmen Antúnez25, Luis Miguel Real 7,18, Adelina Orellana1, Ángel Carracedo5,26, María Eugenia Sáez6, Lluís Tárraga1,2, Mercè Boada 1,2, Carlos Cruchaga3,4 and Agustín Ruiz 1,2, for the Alzheimer’s Disease Neuroimaging Initiative Abstract Long runs of homozygosity (ROH) are contiguous stretches of homozygous genotypes, which are a footprint of inbreeding and recessive inheritance. The presence of recessive loci is suggested for Alzheimer’s disease (AD); however, their search has been poorly assessed to date. To investigate homozygosity in AD, here we performed a fine- scale ROH analysis using 10 independent cohorts of European ancestry (11,919 AD cases and 9181 controls.) We 1234567890():,; 1234567890():,; 1234567890():,; 1234567890():,; detected an increase of homozygosity in AD cases compared to controls [βAVROH (CI 95%) = 0.070 (0.037–0.104); P = 3.91 × 10−5; βFROH (CI95%) = 0.043 (0.009–0.076); P = 0.013]. ROHs increasing the risk of AD (OR > 1) were significantly overrepresented compared to ROHs increasing protection (p < 2.20 × 10−16). A significant ROH association with AD risk was detected upstream the HS3ST1 locus (chr4:11,189,482‒11,305,456), (β (CI 95%) = 1.09 (0.48 ‒ 1.48), p value = 9.03 × 10−4), previously related to AD. Next, to search for recessive candidate variants in ROHs, we constructed a homozygosity map of inbred AD cases extracted from an outbred population and explored ROH regions in whole- exome sequencing data (N = 1449). We detected a candidate marker, rs117458494, mapped in the SPON1 locus, which has been previously associated with amyloid metabolism. Here, we provide a research framework to look for recessive variants in AD using outbred populations. Our results showed that AD cases have enriched homozygosity, suggesting that recessive effects may explain a proportion of AD heritability. Introduction AD presents a strong genetic component. Autosomal Alzheimer’s disease (AD) is a neurodegenerative dis- dominant mutations have been linked to familial early order that is the leading cause of dementia worldwide1. onset AD (EOAD) (<65 years): mutations in presenilin 1 (PSEN1)2, presenilin 2 (PSEN2)3, and amyloid precursor protein (APP)4. These findings lead to the role of amyloid Correspondence: Agustín Ruiz (aruiz@fundacioace.org) 1 metabolism as disease-causing mechanism5. Despite that, Research Center and Memory clinic Fundació ACE. Institut Català de Neurociències Aplicades, Universitat Internacional de Catalunya, Barcelona, dominant causes account for a minority of both familial Spain 2 and sporadic EOAD cases, suggesting that autosomal CIBERNED, Center for Networked Biomedical Research on Neurodegenerative recessive loci might cause most EOAD cases (∼90%)6. Diseases, Carlos III Institute of Health, Madrid, Spain Full list of author information is available at the end of the article However, only two recessive mutations in the APP gene Lists of authors and their affiliations appear at the end of the paper. © The Author(s) 2021 Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/. Moreno-Grau et al. Translational Psychiatry (2021)11:142 Page 2 of 12 (A673V and E693Δ) have been described to date7,8, and relative to controls. Next, we measured the degree of this mode of inheritance for AD remains controversial. inbreeding in AD cases. To prioritize regions with The sporadic form of late-onset AD (LOAD) (>65 years) potential recessive loci, we constructed a homozygosity has a polygenic background. Heritability estimation for map of genomic regions overrepresented in detected LOAD is, roughly, 70%6. Although, near to 40 loci has inbred AD cases. Finally, we performed further explora- been associated with LOAD risk9–12, genetic variance tion of several promising candidate ROHs using whole- captured by genome-wide strategies fall in a range of exome sequencing (WES) data. 7 –31%9,13, explaining a limited part of disease heritability. Current genetic findings were made using an additive Patients and methods mode of inheritance, which overlooks the relevance of The overview of the proposed strategy for ROH detec- non-additive genetic components, i.e., the recessive tion and subsequent prioritization is depicted in Fig. 1. model. Despite the fact these components could explain a fraction of disease heritability. Genotyping data It is well known that inbreeding increases the incidence This study includes 10 independent genome-wide data of recessive diseases. The probability of detecting a sets comprising a total sample of 21,100 unrelated indi- recessive locus increases in offspring of consanguineous viduals (11,921 AD cases and 9181 individual controls) of unions14, because the partners share alleles identical-by- European ancestry (Supplementary Table 1). The descent. This recent parental relatedness points to genu- recruitment and phenotyping, has been described ine regions of autozygosity. Long runs of homozygosity previously12. (ROHs)—long stretches of consecutive homozygous Genotype-level data for each cohort was processed by genotypes (>1 Mb)—are a recognized signature of reces- applying identical quality control and imputation proce- sive inheritance and provide a measure of inbreeding in dures, as previously reported12. Next, we generated a studied populations. Thus far, they have been used for merged data set combining imputed genotypes (MAF > homozygosity mapping15. Population history, e.g., histor- 0.05; imputation quality R2 > 0.90) from available data ical bottlenecks or geographical isolation, also influences sets. We calculated identity-by-descendent (IBD) with homozygosity levels in individual genomes16,17. PLINK 1.9 to generate a cohort of unrelated individuals of An excess of homozygosity has been associated with the European ancestry (Supplementary Fig. 1). All possible risk of AD in individuals of Caribbean-Hispanic and pairs had Pi-hat < 0.1875, a Z0 ≥ 0.75 and a Z1 ≤ 0.25. African-American ancestries18–20, suggesting the pre- Imputed markers with call rates >0.95 and MAF > 0.05 in sence of inbreeding and potentially autosomal recessive the merged data set were selected for ROH calling AD (arAD) cases nested in these populations. Conversely, (NSNPs = 2,678,325). this association was not replicated for individuals of European ancestry21,22. Several factors might explain Runs of homozygosity (ROHs) exploration these inconsistencies, among them it has been estimated 1-Identification of individual ROHs that large sample sizes (12,000‒65,000) are required to Individual ROH calling was conducted using the detect an excess of homozygosity in outbred popula- observational genotype-counting approach implemented tions23. Thus, previous studies might be underpowered. in PLINK (v1.09) (https://www.cog-genomics.org/plink/1.9/), The limited number of deeply characterized con- as it outperforms additional methods in ROH detection sanguineous families, the difficulties in finding familial and it is applicable to outbred populations24. ROH information for sporadic AD individuals (mainly due to detection was performed for each individual study and the late onset of the disease) and the reduced size of for the merged data set using imputed genotypes. We intragenerational pedigrees in western countries make the used a sliding window of 50 SNPs of 5000 Kb in length to search for recessive patterns of inheritance in AD com- scan the genome. In order to manage genomic regions plex. Furthermore, follow-up of candidate ROHs in with a small number of genotyping errors and discrete sequencing data might be a necessary step in the definitive missingness, one heterozygote and five missing calls per mapping of an arAD locus, but it has been poorly assessed window were tolerated. These parameters were similar to to date. those described previously25. The minimal number of Assessing the impact of homozygosity in the genetic SNPs in a ROH was set to 100 SNPs26,27. We empirically architecture of AD, and subsequent follow-up of homo- explored two minimal length cut-offs to consider a ROH, zygous regions remains a challenge. To the best of our 1 Mb and 1.5 Mb. ROHs < 1.5 Mb might reflect LD pat- knowledge, this is the largest genomic data set exploring terns of ancient origin rather than the consanguineous the influence of homozygosity in AD (n = 21,100). First, cultural practices and genetic isolation captured with we investigated whether AD individuals from a European ROHs > 1.5 Mb28. SNPs were included in a ROH if >5% outbred population presented an excess of homozygosity of the sliding window was homozygous. The maximum Moreno-Grau et al. Translational Psychiatry (2021)11:142 Page 3 of 12 GWAS data WES data 21,102 individuals 1,449 individuals 6b. Variant Filtering 1. Individual ROHs 6a. Gene Based Strategy ≥ 1.5 Mb Strategy 807 ROHs; 1,722 Genes ≥ 100 SNPs Coding and flanking variants for candidate 2. Homozygosity 3. Consensus ROHs ROH regions Parameters 21,190 SF3-SF4-SF5 FROH > 0.0156 (2nd degree relave) Minor Minor Allele Allele Frequency Frequency < 0.01 < 0.01 5. 958 inbred AD cases; 5,087 4.b. Consensus ROH 4.a. HP vs AD consensus ROHs vs AD SF7 High or Moderate Homozygous variants > 0.1Mb Impact Variants > 100 SNPs ≥ 2 SNPs Beta > 0.03 > 0.1Mb > 3 SNPs 12 ROH 11 ROH Gene-Based Analysis Variants only present 795 ROH 1,006 ROH SKAT-O in AD individuals in/near in/near Non-AD loci Non-AD loci AD loci AD loci P < 0.05 32 Genes > 3 Variants Variants present in 33 ROHs 1,722 Genes in the model ROHs 32 Genes Fig. 1 Schematic of the stepwise for ROH prioritization. 1. Identification of ROH segments per individual; 2. Estimation of: homozygosity parameters, and 3. Consensus ROHs; 4. Association analysis between: a) Homozygosity parameters and AD status, and b) Consensus ROH and AD status; 5. Identification of inbred AD cases and ROH prioritization; 6. Exploration of selected ROH segments in WES data applying: a) Gene-based strategy, and b) Variant filtering strategy. distance between two consecutive SNPs was set to exploratory analyses are shown in Supplementary Figs. 2–3, 1000 Kb apart, and SNP density to at least 1 SNP in Supplementary Tables 2–3, and Supplementary Meth- 50 Kb. ods. According to them, we decided to conduct down- stream analyses with ROH calling at 1.5 Mb in the 2-Exploration of homozygosity parameters merged data. To assess the data quality and genetic architecture of detected ROHs (>1 Mb and >1.5 Mb) in each individual study and in the whole dataset, we calculated: (a) the SROHðKbÞ F ROH ¼ ð1Þ mean of the total length of ROH or sum of ROH (SROH); Autosomal genome ðKbÞ (b) the average ROH length (AVROH); (c) the number of ROHs (NROH); and (d) ROH-based estimates of the inbreeding coefficient, F, (FROH) per individual. AVROH Copy number variants (CNV), particularly hemizygous is the SROH divided by NROH per subject. FROH deletions, are known to cause spurious ROHs. However, represents the proportion of homozygous segments in the prior studies have demonstrated that the impact of per- autosomal genome per individual (Eq. 1). For individuals, forming ROH calling with or without CNVs is only 0.3% this would be the SROH detected divided by a factor of of the total ROH length28. To assess the impact of CNVs 3,020,190 Kb, the total autosomal genome length deletions, we also conducted ROH calling after removing according to the GRCh37.p13 assembly. We further common CNV deletions extracted from the Database of explored whether the effect of homozygosity parameters Genomic Variants (DGV) (http://dgv.tcag.ca/)29. The was similar when: (1) ROH length was set to 1 or 1.5 Mb; same exercise was conducted after removing CNVs and (2) the analysis was performed per data set or in the detected in GR@ACE dataset. Further description of CNV final merged database. Results emerging from these calling is provided in Supplementary material. Moreno-Grau et al. Translational Psychiatry (2021)11:142 Page 4 of 12 3-Identification of consensus ROHs 5-The homozygosity map of inbred AD individuals Consensus ROHs were defined as overlapping segments 5a-Identificationn of inbred individuals We used FROH between individual ROHs observed in different indivi- to detect the subset of inbred individuals within our duals, with DNA segment match of at least 95% for non- dataset. FROH has been previously shown to better missing SNP markers. Consensus ROH calling was per- correlate with the unobserved pedigree inbreeding23,31. formed using PLINK 1.9. To prevent the detection of The cut-off between inbred and non-inbred individuals false-positive ROHs, we extracted those consensus ROHs was set to FROH > 0.015632, which corresponds to a with a length >100 Kb and >3 SNPs. second-degree relation. It was assumed that there are no different biological effects below 0.0156 than in the general population33. The efficient capture of inbred individuals is 4-Analyses shown in Supplementary Fig. 7. Next, to explore whether 4a-Association analysis between homozygosity para- the frequency of consanguinity was higher in AD cases meters and AD risk To assess the quality of the data in than in controls, we calculated the odds ratio and chi each individual study, we explored sample distribution for square p values using the epitools package in R. each of four homozygosity parameters: NROH, SROH, AVROH, and FROH. An exploratory analysis was 5b-ROHs prioritization based on inbred AD cases depicted with violin plots, which combine a box plot ROH detection was conducted in the subset of inbred AD with a kernel density plot, using the ggplot2 package from cases, applying similar criteria to those previously R (Supplementary Figs. 4 and 5). The inverse rank normal described. Briefly, considering the long size of homozygous transformation was performed to generalize homozygos- tracts for inbred individuals, there is a higher probability of ity parameters using “rankNorm” option in the RNOmni finding a consensus ROH by chance within consanguineous package in R. Transformed distributions are shown in AD cases than in the general population. Hence, we applied Supplementary Fig. 6. To test the association of homo- stringent criteria to define consensus ROHs. Consensus zygosity parameters with AD status, we developed a ROHs from inbred AD cases with ROH lengths >100 Kb generalized linear model for a binominal outcome, using and ROH > 100 SNPs were given priority for further R for individual-level data. We tested three models, analysis. Shared overlapping regions between inbred AD adjusting per: (1) cohort and the first four principal cases and the whole data set were also identified (See bash components (PCs) resulting from ancestry analysis. See code in Supplementary Code Material) and selected based Eq. 2; (2) cohort, PCs and age; (3) cohort, PCs, age and on their overrepresentation in AD cases relative to controls gender. We also conducted a sensitivity analysis excluding (β > 0.03). Prioritized regions were then explored in control individuals <60 years old (See the “Results” sequencing data. We also explored the overlapping of section), these regions with previously identified AD loci30. Z ¼ β1 Homozygosity Parameter þ β2 Cohort ð2Þ WES data þβ3 PC1 þ β4 PC2 þ β5 PC3 þ β5 PC4 þ e To meet the objective of exploring most promising ROH candidates in the sequencing data, we used the Knight-ADRC-NIA-LOAD (KANL) cohort34. We exclu- 4b-Association analysis between consensus ROHs and ded autosomal dominant familial cases and sporadic AD AD The association between the phenotype and con- cases harboring well-known disease-causing mutations, as sensus ROHs was explored using a logistic model, for they could explain disease status. Thus, this study com- ROHs in or near to previously identified AD loci prised 986 AD cases and 463 control individuals of Eur- extracted from de Rojas et al.30 and non-AD ROHs. opean ancestry (See Supplementary Table 1 and The model was adjusted per cohort, and the first four Supplementary Fig. 1). Of these, 488 subjects presented PCs as covariates for downstream analysis. Covariate both GWAS and WES data available for this study. models adjusted for age and gender, in addition to Detailed descriptions of cohort characteristics and quality cohort and PCs, were also calculated. Regression-based control for WES data have been provided previously34. results were corrected for multiple testing using a Bonferroni correction. 6-Candidate gene prioritization strategies using WES Next, we sought to estimate whether there was an 6a-Gene-based analysis To prioritize genes in consen- overrepresentation of risk (β > 0) or protective (β < 0) sus ROH regions, we performed gene-based analysis (986 consensus ROHs in our association results at different cases vs 463 controls) (Fig. 1). To generate variant sets, levels of length and SNP number per consensus ROH. We variants were filtered out according to minor allele applied a binominal test using R. frequency (MAF < 0.01) and functional impact. The allele Moreno-Grau et al. Translational Psychiatry (2021)11:142 Page 5 of 12 frequency cut-off was established according to the Exome 21,100 unrelated European individuals from 10 indepen- Aggregation Consortium (ExAC), non-Finnish European dent cohorts (Supplementary Tables 1–2 and Supple- Exome Sequencing project (ESP), and 1000 G. Only those mentary Fig. 4). Relationships between the mean NROH variants predicted to have a high or moderate effect and SROH are shown in Fig. 2. The mean NROH was according to SnpEff were included35. To compute p- 14.6 ± 4.6, the AVROH was 2.11 ± 0.61 Mb, and the values per gene set, SKAT-O model was applied using R. SROH was 31.9 ± 22.2 Mb. These estimations are in The models were adjusted to consider the impact of the accordance with those observed in European indivi- first two PCs and sex. Genes were filtered out from results duals32, except for the NROH parameter, which was if the number of variants included in the model was ≤3. higher than in the previous studies32. Next, we tested the association of the four homozygosity 6b-Variant filtering strategy for inbred AD cases with parameters with AD risk. We found that (i) higher WES and GWAS data available ROH segments emer- inbreeding coefficient (FROH) increased the risk of suffering ging from inbred AD cases are the most promising AD [βFROH (CI95%) = 0.043 (0.009–0.076); p value = 0.013] candidates to harbor autosomal recessive variants. There- (Table 1); (ii) AD patients presented higher average lengths fore, we deeply explored ROHs by applying a variant of ROHs compared to controls [βAVROH (CI95%) = 0.07 filtering strategy. We explored 488 AD cases with (0.037–0.104); p value = 3.91 × 10−5]; (iii) ROH number was complementary GWAS and WES data. Because there is a not associated with AD risk after adjusting for age [βNROH low likelihood to identify any novel or causative mutation (CI 95%) = 0.010 (−0.024–0.044); p value = 0.571] (Table 1). in available databases, variants with MAF > 0.01 were Results per cohort are shown in Supplementary Table 4. excluded. All heterozygous variants were removed. Finally, Notably, a sensitivity analysis conducted excluding: (1) only the variants mapped in individual ROHs were selected. known deletions, i.e., hemizygous segments29; and, (2) dele- To map genes within ROHs, we first extracted all the tions identified in GR@ACE CNV study; provided compar- variants located in ROH regions. Next, we individually able results (Supplementary Table 5). After excluding control annotated each one. individuals <60yo, a stable and significant effect remains for AVROH [βAVROH (CI 95%) = 0.07 (0.031–0.103); p value = 3.51 × 10−5] (Supplementary Table 5). Results ROH parameters are associated with AD risk ROH analysis of AD risk using the whole data set We examined the typical characteristics of the four We identified 21,190 consensus ROHs in the merged ROH parameters (SROH, NROH, AVROH, FROH) in data set (N = 21,100). We observed a significant over- A B Fig. 2 Runs of homozygosity per cohort and per individual. A Mean number of ROHs versus mean total sum of ROHs in Mb for the 10 cohorts explored. B Mean number of ROHs versus mean total sum of ROHs in Mb per individual explored. Red dashed lines represent the threshold for the inbreeding coefficient of 0.0156 (second cousins’ offspring) and 0.0625 (first cousins’ offspring). Moreno-Grau et al. Translational Psychiatry (2021)11:142 Page 6 of 12 Table 1 Effect of genome-wide homozygosity measures in Alzheimer’s disease for the joint analysis. Dataset Model 1 Model 2 Model 3 Beta (CI 95%) P value Beta (CI 95%) P value Beta (CI 95%) P value FROH 0.051 (0.023–0.078) 3.25 × 10−4 0.044 (0.010–0.077) 0.011 0.043 (0.009–0.076) 0.013 −5 AVROH 0.027 (0.000–0.055) 0.051 0.074 (0.040–0.106) 2.16 × 10 0.070 (0.037–0.104) 3.91 × 10−5 NROH 0.043 (0.015–0.071) 2.48 × 10−3 0.010 (−0.024–0.044) 0.559 0.010 (−0.024–0.044) 0.571 Model 1: adjusted per Cohort and PCs; Model 2: Adjusted by cohort, PCs, and age; Model 3: Adjusted by cohort, PCs, age and gender. Results for the association of excess of homozygosity (FROH), average ROH lenght (AVROH), and number of ROH (NROH) with Alzheimer disease status. OR, Odds ratio; with 95% confidence interval (CI 95%) and level of statistical significance (P value). Association between homozygosity parameters and AD status, adjusted per Cohort, PCs, Age and Sex, was conducted in individuals with all available data; N = 19,253. Table 2 Frequency of consensus ROHs with a potential risk or protective effect in Alzheimer’s disease. N ROH Risk associations Protective associations P value Probability of success Whole dataset 21190 11974 9216 < 2.2 × 10−16 0.56 −16 Category A 1017 593 424 < 2.2 × 10 0.58 −6 Category B 926 537 389 1.30 × 10 0.57 −6 Category C 858 499 359 1.98 × 10 0.58 Category D 42 33 9 2.7 × 10−4 0.79 −16 Whole dataset/map of inbreed AD cases 6636 3969 2667 < 2.2 × 10 0.60 Strategy A, ROHs > 100 kb; >3 SNPs. Strategy B, ROHs > 100 kb; >25 SNPs. Strategy C, ROHs > 100 kb; >50 SNPs. Strategy D, ROHs > 100 kb; >3 SNPs, P < 0.05. representation of ROHs increasing the risk of suffering 26,903,551–27,141,292), by its known role in AD7,8,36; AD (p value < 2.20 × 10−16) (Table 2). The same over- detected in 38 AD cases vs 26 controls (26 vs 12 inbred representation of risk associations was detected after fil- individuals, respectively). For non-previously associated tering at several levels based on the length and number of AD regions, none ROH (N = 1006) reached the sig- SNPs per consensus ROH (Table 2). When the test was nificance threshold (Bonferroni correction of p = 4.97 × conducted with results adjusted for cohort, PCs, age, and 10−5). Previous significant consensus ROH (chr8: gender, the over-representation of risk associations 37835460–38143780) associated with AD in Europeans21 remained highly significant (p value < 2.20 × 10−16). was not detected in this study, which is in line of results We then tested the association of 11 consensus ROH from Sims et al.22, failing replication. (≥100 Kb and ≥3 SNPs) located in or near to previously We then explored the genes located in significant risk identified AD loci (N = 38)30, with AD status (Supple- consensus ROHs (p value < 0.05) in gene-based analysis mentary Table 6). For these analyses, Bonferroni cor- from WES data as well (Fig. 1). A total of 33 ROHs rected significance threshold of p = 1.32 × 10−3 was pre- comprising 32 genes were analyzed (included > 3 SNPs in specified. We detected a strong association near to the model; Bonferroni correction p value = 0.0015). The HS3ST1 locus (consensus ROH length = 115.9 Kb; NECAB1 locus (chr8:91,803,921-91,971,630) presented chr4:11,189,482‒11,305,456), (45 AD cases vs 12 controls, the most significant signal (p = 0.01) (Supplementary β (CI 95%) = 1.09 (0.48–1.48), p value = 9.03 × 10−4). Table 8), but none loci reached the multiple test correc- This region survived age and gender adjustments (Sup- tion threshold. plementary Table 6), and was detected across 12 out of 16 datasets (Supplementary Table 7). The replication of this Homozygosity mapping of AD using DNA segments specific locus with AD, using ROH methodology, provides identified in inbred cases new insights of a potential recessive mechanism for this We detected 1621 individuals (958 Cases and 663 dementia locus. Among other ROH regions in or near to Controls) presenting a FROH ≥ 0.0156 among the total known AD loci (Supplementary Tables 6 and 10), we sample (N = 21,100) (Fig. 2) (Supplementary Table 9). highlighted a 237 Kb ROH upstream the APP gene (chr21: Interestingly, inbreeding over the second degree of Moreno-Grau et al. Translational Psychiatry (2021)11:142 Page 7 of 12 rs140897155 SYNE2/ESR2 Fig. 3 Circos plot for the prioritized regions. Histogram for the effect of the 21,190 consensus ROHs identified in the whole sample is shown. Risk ROH associations are shown in red; protective ROH associations are shown in green. Blue regions represent prioritized ROHs from consanguineous AD cases. Orange segments represent prioritized regions harboring potential recessive variants. consanguinity was associated with a higher risk of suf- variants within long ROH segments (Table 3). All ROH fering AD [OR (95%, CI) = 1.12 (1.01–1.25); p value = segments with homozygote variants were detected in 0.027), which is in line with our previous results. This more than 6 cohorts. Two of these individuals were supports the idea that an excess of consanguineous indi- consanguineous (FROH > 0.156). One had a missense var- viduals is present in the AD population. Accordingly, the iant (rs140790046, c.926A > G) that encodes p.Asn309Ser search for recessive loci that play a role in AD can first be change within the MKX locus. Another carried a rare assessed in consanguineous cases. variant (rs116644203) in the ZNF282 locus, which was in After ROH calling in inbred AD cases, we detected 5087 an extremely large region of homozygosity (14.9 Mb) ROHs, and extracted those with ≥100 Kb and ≥100 SNPs. (Table 3). Furthermore, three additional homozygous We then selected only over-represented regions in AD variants were detected: (i) a variant (rs117458494) in the cases relative to controls in the general analysis (Fig. 1). SPON1 locus, previously related with amyloid metabo- We prioritized 807 consensus homozygous segments lism37, and (ii) two potential causative variants, carried from inbred cases (8.6% of the total autosomal genome) only by this individual, within a previously identified AD (Fig. 3 and Supplementary Table 10). Among them, 12 region (TP53INP/NDUFAF6)12. One (rs73263258-ESRP1; ROHs were in or near to a previously identified AD loci in TP53INP/NDUFAF6 region) is a missense variant (Supplementary Table 10). (c.475G > A) that encodes p.Ala159Thr change (Table 3). After exploring genes in identified ROHs by gene-based Further notes and functional effect predictions for these analysis from WES data, none of them remained asso- variants are provided in Supplementary Table 12. ciated after multiple corrections (Ngenes tested = 1136; p value = 3.47 × 10−5) (Supplementary Table 11). Our top Discussion signal was detected in the FRY locus (p value = 0.001) This study represents the largest analysis of homo- (Supplementary Table 11). zygosity conducted for AD. Our estimates of homo- Considering that recessive variants are expected at very zygosity provide a robust evidence supporting that low frequencies, even gene-based analysis would be recessive allelic architecture might define a portion of AD underpowered to detect significant associations. There- heritability. fore, we decided to further prioritize loci by searching Previous AD ROH studies in European populations homozygous mutations within selected consensus ROHs have shown negative results for the association of ROH from inbred AD subjects (Fig. 1). We identified seven AD parameters with AD21,22. First studies had very modest cases that had eight new (or extremely rare) homozygous sample sizes (N < 3000, vs Npresent study = 21,100)21,22, and Moreno-Grau et al. Translational Psychiatry (2021)11:142 Page 8 of 12 likely were underpowered. Then, these studies used a non_coding_exon_ downstream_gene downstream_gene non_coding_exon non_coding_exon non_coding_exon non_coding_exon sequence_feature sequence_feature sequence_feature ROH calling lengths set to 1 Mb21,22. This generates upstream_gene upstream_gene Variant Effect 3_prime_UTR 5_prime_UTR 3_prime_UTR synonymous synonymous substantial inflation in the inbreeding coefficient (FROH) missense missense missense missense intronic intronic and makes undetectable the enrichment in consanguinity due to unspecific noise (Supplementary material). These reasons might explain initial failures. We encourage other MODERATE MODERATE MODERATE MODERATE MODIFIER MODIFIER MODIFIER MODIFIER MODIFIER MODIFIER groups to conduct ROH analysis in new unrelated Impact LOW LOW LOW LOW populations, but with large enough sample sizes and redefining the ROH lengths at least to 1.5 Mb, to better <0.0000 <0.0000 <0.0000 0.0005 capture the recessive component of AD. 0.004 0.002 0.003 0.001 MAF At the present study we identified a study-wise sig- nificant ROH association close to the HS3ST1 gene Alt Allele (~200 Kb). Genetic markers near to this ROH (~300 kb) have been previously associated with AD using additive G A A A A A C T models38,39, and HS3ST1 locus was differentially expres- Ref Allele sed in the brain of AD cases versus controls38. Our finding reinforces the association of this region with AD, and G G G A C C T T further suggests the role of recessiveness in explaining RP11-678B3.2 RP11-21L19.1 Near locus underlying associations. High-resolution mapping across DPY19L4* ZNF282 ESPR1* SPON1 this ROH could help to identify the causative mutation. SYNE2 SPIDR ESR2 MKX This study failed replication of previously detected ROH at chr8:37835460–3814378021. Although, both studies 7:148903805 10:27964291 11:14282085 14:64688393 7:53930748 8:48352954 8:95658495 8:95780656 include TGEN cohort, overlapping to some extent, the Variant Candidate recessive variants after ROH prioritization focused on inbred AD cases. default technical parameters for ROH definition were completely different (ROH calling: 1 Mb vs 1.5 Mb). We rs116644203 rs138998679 rs140790046 rs117458494 rs140897155 assume that technical differences of the present study rs73263258 respect to prior ones, might be critical points impacting replication of ROH findings, in addition to other causes, NA NA Rs e.g. population-specific genetic patterns, or, even, random SNPs chance. MAF < 0.0000; Variants with non reported allele frequency in ExAC, ESP, 1000G reference database. 12426 ROH 4213 1265 4734 6248 4473 2633 A strength of the present study comes from our effort to prioritize consensus ROHs according to the homozygosity ROH lenght map of inbred AD individuals, performed by the first time 14900.679 2029.938 2526.301 5453.066 3898.626 4326.451 2427.07 in AD, and our capacity to explore them in sequencing data. This strategy lets us to find interesting candidate recessive variants in: MKX and ZNF282 genes, identified 150518393 ROH end 55262627 49544784 96735572 28414933 16358722 66333670 in two independent inbred AD cases; TP53INP1/NDU- FAF6 genomic region, previously associated with AD12,40; and SPON1 locus. The SPON1 locus deserves a further 135617715 ROH start 53232690 47018484 91282507 24516308 12032272 63906601 explanation as it is directly related with APP metabolism, a key player in AD physiopathology. APP cleavage through β-secretases produces amyloid-beta (Aβ), which CHR later accumulates in AD brains5. SPON1 has been found 10 11 14 7 7 8 8 to bind to APP, inhibiting its α/β cleavage37, and to APOE SROH (Kb) family of receptors41. Markers in this gene have been 17421.6 55861.4 29868.8 46194.5 54137.7 41392.6 34603.4 related to dementia severity42 and with the rate of cog- nitive decline43. Considering prior findings and the pre- ROH>1500 kb sent result, it would be biologically plausible that the presence of recessive variants in APP7,8, or its biological 0.0058 0.0185 0.0099 0.0153 0.0179 0.0137 0.0115 partners directly influences the amyloid cascade. Thus, we Froh believe that SPON1 could be considered an interesting candidate, which deserves future resequencing efforts. Table 3 Individual Our observations are subject to limitations that need to be considered. Data sets used in this study were 1 2 3 4 5 6 7 Moreno-Grau et al. Translational Psychiatry (2021)11:142 Page 9 of 12 genotyped using different platforms and shared a small conduct future large-scale homozygosity mapping studies proportion of directly genotyped markers. Given that in AD. Furthermore, the exploration of complementary lower SNP density could impact the accuracy of the sequencing data gave an added value to this research, study32, high quality imputed markers were used (r2 > providing a subset of potential candidates harboring 0.90, MAF > 0.05). Second, to reduce dataset hetero- recessive variants. In any case, the proposed candidates geneity we use a set of European individuals; applied the would need confirmation in larger series. Greater efforts same GWAS quality control per study; generated a and larger collections of individuals with GWAS and merged dataset including common variants across data- sequencing data are needed to confirm the present sets; and controlled all our analyses by cohort, to account findings. with potential confounding. Our understanding of the dynamics of population We assumed that differences in the ROH parameters genomics in AD is far from complete, but ROH analyses between the cases and controls are modest. In that sense, provide us with a means to go further and might be an we are not expecting a very large percentage of recessive alternative strategy to uncover the genetic loci AD cases, but we expect a fraction, in the same way, that it underlying AD. occurs for autosomal dominant forms (<1%). Considering that, the reported findings are supporting the hypothesis Acknowledgements We would like to thank patients and controls who participated in this project. of this work, a group of recessive mutation may explain a The Genome Research at Fundació ACE project (GR@ACE) is supported by portion of AD cases. However, we suspect that the exis- Fundación bancaria “La Caixa,” Grifols SA, Fundació ACE, and ISCIII (Ministry of tence of a large non-allelic heterogeneity is preventing its Health, Spain). We also want to thank the private sponsors who support the basic and clinical projects of our institution (Piramal AG, Laboratorios identification. Echevarne, Araclon Biotech S.A., and Fundació ACE). We are indebted to the Our gene-based analysis strategy did not show sig- Trinitat Port-Carbó legacy and her family for their support of Fundació ACE nificant associations. With a decreasing allele frequency research programs. Fundació ACE is a participating center in the Dementia Genetics Spanish Consortium (DEGESCO). A.R. and M.B. receive support from and high locus heterogeneity, the power to detect genes of the European Union/EFPIA Innovative Medicines Initiative Joint Undertaking interest also decreases. Despite our effort to include WES ADAPTED and MOPEAD projects (Grants No. 115975 and 115985, respectively). data in the present study, the available sample size could M.B. and A.R. are also supported by national grants PI13/02434, PI16/01861, PI17/01474, and PI19/01301. Acción Estratégica en Salud is integrated into the be underpowered. Spanish National R + D + I Plan and funded by ISCIII (Instituto de Salud Carlos The potential impact of CNV deletions on ROH analysis III)-Subdirección General de Evaluación and the Fondo Europeo de Desarrollo must be taken into consideration. Thus, we assessed its Regional (FEDER- “Una manera de Hacer Europa”). L.M.R. is supported by Consejería de Salud de la Junta de Andalucía (Grant PI-0001/2017). Control effect on our analyses, but no differences were found samples and data from patients included in this study were provided in part by before and after CNV exclusion (Supplementary Table 5), the National DNA Bank Carlos III (www.bancoadn.org, University of Salamanca, which is in agreement with the previous studies25. Clonal Spain) and Hospital Universitario Virgen de Valme (Sevilla, Spain); they were processed following standard operating procedures with the appropriate mosaicism, due to aging44, could also generate spurious approval of the ethical and scientific committees of these institutions. The ROHs. At the present study an age-dependent increase in present work was performed as part of the Biochemistry, Molecular Biology the NROH was detected in the control group (Supple- and Biomedicine doctoral program of S. Moreno-Grau at Universitat Autònoma de Barcelona (Barcelona, Spain). This work was supported by grants from the mentary Table 13), which partially disappeared after National Institutes of Health (R01AG044546, P01AG003991, RF1AG053303, excluding consensus ROHs associated with age (p < 0.05) R01AG058501, U01AG058922, RF1AG058501, and R01AG057777) and the (See Supplementary Material, and Supplementary Table Alzheimer’s Association (NIRG-11-200110, BAND-14-338165, AARG-16-441560, and BFG-15-362540). The recruitment and clinical characterization of research 13 and 14). We assumed that these DNA segments might participants at Washington University were supported by NIH P50 AG05681, contain somatic alterations, confounding ROH associa- P01 AG03991, and P01 AG026276. This work was supported by access to tions. Among age-related ROH regions, we identified equipment made possible by the Hope Center for Neurological Disorders, and the Departments of Neurology and Psychiatry at Washington University School some loci previously associated with AD, e.g., RORA, of Medicine. For additional information about GR@ACE & DEGESCO consortia, CD2AP, HS3ST1, and amyloid-beta burden, e.g., GLIS345; please contact with aruiz@fundacioace.com The GR@ACE study group is suggesting that some known AD regions could be affected composed by: Abdelnour C, Aguilera N, Alarcon E, Alegret M, Benaque A, Boada M, Buendia M, Cañabate P, Carracedo A, Corbatón A, de Rojas I, Diego S, by this phenomenon. These findings deserve future Espinosa A, Gailhajenet A, García González P, Gil S, Guitart M, González Pérez A, investigations. Despite the existence of ROH segments Hernández I, Ibarria, M, Lafuente A, Macias J, Maroñas O, Martín E, Martínez MT, associated to age and somatic mosaicism phenomena, our Marquié M, Mauleón A1, Monté-Rubio G, Montrreal L, Moreno-Grau S, Moreno M, Orellana A, Ortega G, Pancho A, Pelejà E, Pérez-Cordon A, Pineda JA, most significant findings largely supported adjustments by Preckler S, Quintela I, Real LM3,8, Rodríguez-Gómez O, Rosende-Roca M, Ruiz A, age. Therefore, we feel that the major observations of this Ruiz S, Sáez ME, Sanabria A, Santos-Santos MA, Serrano-Rios M, Sotolongo-Grau study are not affected by age-related instability of the O, Tárraga L, Valero S, Vargas L. The DEGESCO consortium is composed by: Adarmes-Gómez A.D, Alarcón-Martín E, Álvarez I, Álvarez V, Amer-Ferrer G, human genome. Antequera M, Antúnez C, Baquero M, Bernal M, Blesa R, Boada M, Buiza-Rueda In summary, we demonstrated the existence of an D, Bullido M.J, Burguera J.A, Calero M, Carrillo F, Carrión-Claro M, Casajeros MJ, inbreeding effect in AD and efficiently captured a fraction Clarimón J, Cruz-Gamero J.M, de Pancorbo MM, de Rojas I, del Ser T Diez-Fairen M, Fortea J, Franco E, Frank-García A, García-Alberca JM, Garcia Madrona S, of inbred individuals from outbred populations, providing Garcia-Ribas G, Gómez-Garre P, Hernández I, Hevilla S, Jesús S, Labrador an improved strategy to look for recessive alleles, and to Espinosa MA, Lage C, Legaz A, Lleó A, López de Munáin A, López-García S, Moreno-Grau et al. Translational Psychiatry (2021)11:142 Page 10 of 12 Macias D, Manzanares S, Marín M, Marín-Muñoz J, Marín T, Marquié M, Martín https://doi.org/10.7303/syn2580853. We thank the International Genomics of Montes A, Martínez B, Martínez C, Martínez V, Martínez-Lage Álvarez P, Medina Alzheimer’s Project (IGAP) for providing summary results data for these M, Mendioroz Iriarte M, Menéndez-González M, Mir P, Molinuevo J.L, Monté- analyses. The investigators within IGAP contributed to the design and Rubio G, Montrreal L, Moreno-Grau S, Orellana A, Pastor A.B, Pastor P, Pérez Tur implementation of IGAP and/or provided data but did not participate in J, Periñán-Tocino T Piñol Ripoll G, Rábano A, Real de Asúa D, Rodrigo S, analysis or writing of this report. IGAP was made possible by the generous Rodríguez-Rodríguez E, Royo J.L, Ruiz A, Sanchez del Valle Díaz R, Sánchez-Juan participation of the control subjects, the patients, and their families. The P, Sastre I, Sotolongo-Grau O, Tárraga L, Valero S, Vicente M.P, Vivancos L. Data i–Select chip was funded by the French National Foundation on Alzheimer’s collection and sharing for this project was partially funded by the Alzheimer’s Disease and Related Disorders. EADI was supported by the LABEX (laboratory Disease Neuroimaging Initiative (ADNI) (National Institutes of Health Grant U01 of excellence program investment for the future) DISTALZ grant, Inserm, AG024904) and DOD ADNI (Department of Defense award number W81XWH- Institut Pasteur de Lille, Université de Lille 2, and the Lille University Hospital. 12-2-0012). ADNI is funded by the National Institute on Aging and the National GERAD was supported by the Medical Research Council (Grant no 503480), Institute of Biomedical Imaging and Bioengineering, as well as through Alzheimer’s Research UK (Grant no 503176), the Wellcome Trust (Grant no generous contributions from the following: AbbVie; the Alzheimer’s 082604/2/07/Z), and German Federal Ministry of Education and Research Association; the Alzheimer’s Drug Discovery Foundation; Araclon Biotech; (BMBF): Competence Network Dementia (CND) grants 01GI0102, 01GI0711, BioClinica, Inc.; Biogen; Bristol-Myers Squibb Company; CereSpir, Inc.; Cogstate; and 01GI0420. CHARGE was partly supported by the NIH/NIA grant R01 Eisai Inc.; Elan Pharmaceuticals, Inc.; Eli Lilly and Company; EuroImmun; F. AG033193 and the NIA AG081220 and AGES contract N01–AG–12100, the Hoffmann-La Roche Ltd and its affiliated company Genentech, Inc.; Fujirebio; NHLBI grant R01 HL105756, the Icelandic Heart Association, and the Erasmus GE Healthcare; IXICO Ltd.; Janssen Alzheimer Immunotherapy Research & Medical Center and Erasmus University. ADGC was supported by the NIH/NIA Development, LLC.; Johnson & Johnson Pharmaceutical Research & grants: U01 AG032984, U24 AG021886, U01 AG016976, and the Alzheimer’s Development LLC.; Lumosity; Lundbeck; Merck & Co., Inc.; Meso Scale Association grant ADGC–10–196728. Diagnostics, LLC.; NeuroRx Research; Neurotrack Technologies; Novartis Pharmaceuticals Corporation; Pfizer Inc.; Piramal Imaging; Servier; Takeda Author details Pharmaceutical Company; and Transition Therapeutics. The Canadian Institute 1 Research Center and Memory clinic Fundació ACE. Institut Català de of Health Research provides funds to support ADNI clinical sites in Canada. Neurociències Aplicades, Universitat Internacional de Catalunya, Barcelona, Private-sector contributions are facilitated by the Foundation for the National Spain. 2CIBERNED, Center for Networked Biomedical Research on Institutes of Health (www.fnih.org). The grantee organization is the Northern Neurodegenerative Diseases, Carlos III Institute of Health, Madrid, Spain. California Institute for Research and Education, and the study was coordinated 3 Department of Psychiatry, Washington University School of Medicine, St. by the Alzheimer’s Therapeutic Research Institute at the University of Southern Louis, MO, United States of America. 4Hope Center for Neurological Disorders, California. ADNI data are disseminated by the Laboratory for Neuro Imaging at Washington University School of Medicine, St. Louis, MO, United States of the University of Southern California. The AddNeuroMed data are from a America. 5Grupo de Medicina Xenómica, Centro Nacional de Genotipado public-private partnership supported by EFPIA companies and SMEs as part of (CEGEN-PRB3-ISCIII), Universidade de Santiago de Compostela, Santiago de InnoMed (Innovative Medicines in Europe), an integrated project funded by Compostela, Spain. 6CAEBI. Centro Andaluz de Estudios Bioinformáticos, Sevilla, the European Union under the Sixth Framework program priority FP6-2004- Spain. 7Unidad Clínica de Enfermedades Infecciosas y Microbiología. Hospital LIFESCIHEALTH-5. Clinical leads responsible for data collection are Iwona Universitario de Valme, Sevilla, Spain. 8Memory Unit, Neurology Department Kłoszewska (Lodz), Simon Lovestone (London), Patrizia Mecocci (Perugia), and Sant Pau Biomedical Research Institute, Hospital de la Santa Creu i Sant Hilkka Soininen (Kuopio), Magda Tsolaki (Thessaloniki) and Bruno Vellas Pau, Universitat Autònoma de Barcelona, Barcelona, Spain. 9Centro de Biología (Toulouse). Imaging leads are Andy Simmons (London), Lars-Olad Wahlund Molecular Severo Ochoa (C.S.I.C.-U.A.M.), Universidad Autónoma de Madrid, (Stockholm) and Christian Spenger (Zurich). Bioinformatics leads are Richard Madrid, Spain. 10Instituto de Investigación Sanitaria “Hospital la Paz” (IdIPaz), Dobson (London) and Stephen Newhouse (London). Funding support for the Madrid, Spain. 11Hospital Universitario Ramón y Cajal, Madrid, Spain. Alzheimer’s Disease Genetics Consortium (ADGC) was provided through the 12 Fundació per la Recerca Biomèdica i Social Mútua Terrassa, and Memory NIA Division of Neuroscience (U01-AG032984). The genotypic and associated Disorders Unit, Department of Neurology, Hospital Universitari Mútua de phenotypic data used in the study “Multi-Site Collaborative Study for Terrassa, University of Barcelona School of Medicine, Terrassa, Barcelona, Spain. Genotype-Phenotype Associations in Alzheimer’s Disease (GenADA)” were 13 Neurology Service “Marqués de Valdecilla” University Hospital (University of provided by GlaxoSmithKline, R&D Limited. The data sets used for the analyses Cantabria and IDIVAL), Santander, Spain. 14Laboratorio de Genética Hospital described in this manuscript were obtained from dbGaP at http://www.ncbi. Universitario Central de Asturias, Oviedo, Spain. 15Instituto de Investigación nlm.nih.gov/gap through dbGaP accession number phs000219. The Mayo Biosanitaria del Principado de Asturias (ISPA), Oviedo, Spain. 16Unitat Trastorns Clinic Alzheimer’s Disease Genetic Studies, led by Dr. Nilüfer Ertekin-Taner and Cognitius, Hospital Universitari Santa Maria de Lleida, Institut de Recerca Dr. Steven G. Younkin at the Mayo Clinic in Jacksonville, FL, used samples from Biomédica de Lleida (IRBLLeida), Lleida, Spain. 17Alzheimer Research Center & the Mayo Clinic Study of Aging, the Mayo Clinic Alzheimer’s Disease Research Memory Clinic, Andalusian Institute for Neuroscience, Málaga, Spain. 18Dep. of Center, and the Mayo Clinic Brain Bank. Data collection was supported through Surgery, Biochemistry and Molecular Biology, School of Medicine, University of funding from NIA grants P50 AG016574, R01 AG032990, U01 AG046139, R01 Málaga, Málaga, Spain. 19Unidad de Demencias, Servicio de Neurología y AG018023, U01 AG006576, U01 AG006786, R01 AG025711, R01 AG017216 and Neurofisiología. Instituto de Biomedicina de Sevilla (IBiS), Hospital Universitario R01 AG003949, NINDS grant R01 NS080820, the CurePSP Foundation, and the Virgen del Rocío/CSIC/Universidad de Sevilla, Seville, Spain. 20Unidad de Mayo Foundation. The Neocodex-Murcia study was funded by the Fundación Trastornos del Movimiento, Servicio de Neurología y Neurofisiología. Instituto Alzheimur (Murcia), the Ministerio de Educación y Ciencia (Gobierno de de Biomedicina de Sevilla (IBiS), Hospital Universitario Virgen del Rocío/CSIC/ España), Corporación Tecnológica de Andalucía, Agencia IDEA (Consejería de Universidad de Sevilla, Seville, Spain. 21CIEN Foundation, Queen Sofia Innovación, Junta de Andalucía), the Diabetes Research Laboratory, and the Foundation Alzheimer Center, Madrid, Spain. 22Instituto de Salud Carlos III Biomedical Research Foundation. University Hospital Clínico San Carlos has (ISCIII), Madrid, Spain. 23BT-CIEN, Madrid, Spain. 24Department of Molecular been supported by CIBER de Diabetes y Enfermedades Metabólicas Asociadas Neuropathology, Centro de Biología Molecular “Severo Ochoa” (CBMSO), (CIBERDEM); CIBERDEM is an ISCIII Project. The ROS/MAP study data were Consejo Superior de Investigaciones Científicas (CSIC)/Universidad Autónoma provided by the Rush Alzheimer’s Disease Center, Rush University Medical de Madrid (UAM), Madrid, Spain. 25Unidad de Demencias, Hospital Clínico Center, Chicago. Data collection was supported through funding from NIA Universitario Virgen de la Arrixaca, Madrid, Spain. 26Fundación Pública Galega grants P30AG10161, R01AG15819, R01AG17917, R01AG30146, R01AG36836, de Medicina Xenómica- CIBERER-IDIS, Santiago de Compostela, Spain U01AG32984 and U01AG46152, the Illinois Department of Public Health, and the Translational Genomics Research Institute. The TGEN study was supported by Kronos Life Sciences Laboratories, the National Institute on Aging (Arizona Author contributions Alzheimer’s Disease Center grants P30 AG19610 and RO1 AG023193, the Mayo S.M.G. and A.R. contributed to the study design. S.M.G. analyzed GWAS data. Clinic Alzheimer’s Disease Center grant P50 AG16574, and the Intramural M.V.F., F.F., and J.P.B. analyzed WES data. P.G.G. performed CNV calling in Research Program), the National Alzheimer’s Coordinating Center (U01 GR@ACE dataset. I.H., M.A., E.A.M., S.V., M.M., and J.A.P., J.M., and L.M.R. AG016976) and the state of Arizona. The results published here are in part contributed to phenotyping, database curation and patient selection from based on data obtained from the AMP-AD Knowledge Portal accessed at GR@ACE dataset. S.M.G., I.R., L.M., and A.O. carried out DNA extraction and Moreno-Grau et al. Translational Psychiatry (2021)11:142 Page 11 of 12 quality control for GR@ACE dataset. I.Q., O.M., and A.C. carried out genotyping 11. Sims, R.,J. et al. Rare coding variants in PLCG2, ABI3, and TREM2 implicate procedures for GR@ACE dataset. J.C., M.J.B., G.G.R., P.P., P.S.J., V.A., G.P.R., J.M.A., J. microglial-mediated innate immunity in Alzheimer’s disease. Nat. Genet. 49, L.R., E.F.M., P.M., M.C., M.M., A.A., J.A., C.A., and A.R. are members of DEGESCO 1373–1384 (2017). consortium. S.M.G., A.G., L.M., and M.E.S. contributed to data processing 12. Moreno-Grau, S. et al. Genome-wide association analysis of dementia and its methods. A.B., L.T., M.B., C.C., and A.R. participated in funds procurement. S.M.G. clinical endophenotypes reveal novel loci associated with Alzheimer’s disease prepared the first draft of the manuscript. S.M.G., M.V.F., C.C., and A.R. were and three causality networks: the GR@ACE project. Alzheimer’s Dement. 15, responsible for the critic revision of the manuscript. All authors read and 1333–1347 (2019). approved the final manuscript. 13. Ridge, P. G. et al. Assessment of the genetic variance of late-onset Alzheimer’s disease. Neurobiol. Aging 41, 200.e13–200.e20 (2016). Code availability 14. Garrod, A. E. The incidence of alkaptonuria: a study in chemical individuality. Script used to identify shared overlapping regions between inbred AD cases Lancet 1616, 1902–1911 (1902). and the whole data set is available at Supplementary Code Material. Additional 15. Lander, E. & Botstein, D. Homozygosity mapping: a way to map human scripts used to conduct ROH calling are available from the authors upon recessive traits with the DNA of inbred children. Science 236, 1567–1570 (1987). request. 16. Jakkula, E. et al. The Genome-wide patterns of variation expose significant substructure in a founder population. Am. J. Hum. Genet. 83, 787–794 (2008). Data availability 17. Gross, A. et al. Population-genetic comparison of the Sorbian isolate popu- GR@ACE datased analysed during the current study is available into the lation in Germany with the German KORA population using genome-wide European Genome-phenome Archive (EGA) (https://ega-archive.org), which is SNP arrays. BMC Genet. 12, 67 (2011). hosted by the EBI and the CRG, under accession number EGAS00001003424 18. Vardarajan, B. N. et al. Inbreeding among Caribbean Hispanics from the (https://ega-archive.org/studies/EGAS00001003424). Additional GWAS datasets Dominican Republic and its effects on risk of Alzheimer disease. Genet. Med. used in these study are available upon request in dbGaP. WES dataset used in 17, 639–643 (2015). the study can be downloaded from https://dss.niagads.org/studies/sa000004/. 19. Ghani, M. et al. Evidence of recessive Alzheimer disease loci in a Caribbean Data used in preparing this article were obtained from the Alzheimer’s Disease Hispanic data set: genome-wide survey of runs of homozygosity. JAMA Neurol. Neuroimaging Initiative (ADNI) database (adni.loni.usc.edu). As such, the 70, 1261–1267 (2013). investigators within the ADNI contributed to the design and implementation of 20. Ghani, M. et al. Association of long runs of homozygosity with Alzheimer ADNI and/or provided data but did not participate in the analysis or writing of Disease among African American individuals. JAMA Neurol. 72, 1313–1323 this report. A complete listing of ADNI investigators can be found at http://adni. (2015). loni.usc.edu/wp-content/uploads/how_to_apply/ADNI_Acknowledgement_List. 21. Nalls, M. A. et al. Extended tracts of homozygosity identify novel candidate pdf. genes associated with late-onset Alzheimer’s disease. Neurogenetics 10, 183–190 (2009). 22. Sims, R. et al. No evidence that extended tracts of homozygosity are asso- Conflict of interest ciated with Alzheimer’s disease. Am. J. Med. Genet. Part B Neuropsychiatr. Genet. The authors declare that they have no conflict of interest. 156, 764–771 (2011). 23. Keller, M. C., Visscher, P. M. & Goddard, M. E. Quantification of inbreeding due Publisher’s note to distant ancestors and its detection using dense single nucleotide poly- Springer Nature remains neutral with regard to jurisdictional claims in morphism data. Genetics 189, 237–249 (2011). published maps and institutional affiliations. 24. Howrigan, D. P., Simonson, M. A. & Keller, M. C. Detecting autozygosity through runs of homozygosity: a comparison of three autozygosity detection Supplementary information The online version contains supplementary algorithms. BMC Genomics 12, 460 (2011). material available at https://doi.org/10.1038/s41398-020-01145-1. 25. Joshi, P. K. et al. Directional dominance on stature and cognition in diverse human populations. Nature 523, 459–462 (2015). Received: 13 August 2020 Revised: 27 November 2020 Accepted: 4 26. Lencz, T. et al. Runs of homozygosity reveal highly penetrant recessive loci in December 2020 schizophrenia. Proc. Natl Acad. Sci. USA 104, 19942–19947 (2007). 27. Christofidou, P. et al. Runs of homozygosity: association with coronary artery disease and gene expression in monocytes and macrophages. Am. J. Hum. Genet. 97, 228–237 (2015). 28. McQuillan, R. et al. Runs of homozygosity in European populations. Am. J. References Hum. Genet. 83, 359–372 (2008). 1. 2019 Alzheimer’s disease facts and figures. Alzheimer’s Dement. 15, 321–387 29. Sudmant, P. H. et al. An integrated map of structural variation in 2,504 human (2019) genomes. Nature 526, 75–81 (2015). 2. Sherrington, R. et al. Cloning of a gene bearing missense mutations in early- 30. Rojas, I de. et al. Common variants in Alzheimer’s disease: novel association of onset familial Alzheimer’s disease. Nature 375, 754–760 (1995). six genetic variants with AD and risk stratification by polygenic risk scores. 3. Sherrington, R. et al. Alzheimer’s disease associated with mutations in pre- medRxiv. 19012021 (2020) senilin 2 is rare and variably penetrant. Hum. Mol. Genet. 5, 985–988 (1996). 31. Gazal, S. et al. Inbreeding coefficient estimation with dense SNP data: com- 4. Goate, A. et al. Segregation of a missense mutation in the amyloid precursor parison of strategies and application to HapMap III. Hum. Hered. 77, 49–62 protein gene with familial Alzheimer’s disease. Nature 349, 704–706 (1991). (2014). 5. Hardy, J. & Selkoe, D. J. The Amyloid Hypothesis of Alzheimer’s Disease: 32. Ceballos, F. C., Hazelhurst, S. & Ramsay, M. Runs of Homozygosity in sub- progress and problems on the road to therapeutics. Science 297, 353–356 Saharan African populations provide insights into a complex demographic (2002). and health history. Hum. Genet. 138, 1123–1142 (2019). 6. Wingo, T. S., Lah, J. J., Levey, A. I. & Cutler, D. J. Autosomal recessive causes 33. Bittles, A. H. & Black, M. L. Consanguinity, human evolution, and complex likely in early-onset Alzheimer disease. Arch. Neurol. 69, 59–64 (2012). diseases. Proc. Natl Acad. Sci. USA 107, 1779–1786 (2010). 7. Di Fede, G. et al. A recessive mutation in the APP gene with dominant- 34. Fernández, M. V. et al. Analysis of neurodegenerative Mendelian genes in negative effect on Amyloidogenesis. Science 323, 1473–1477 (2009). clinically diagnosed Alzheimer Disease. PLoS Genet. 13, e1007045 (2017). 8. Tomiyama, T. et al. A new amyloid β variant favoring oligomerization in 35. Cingolani, P. et al. A program for annotating and predicting the effects of Alzheimer’s-type dementia. Ann. Neurol. 63, 377–387 (2008). single nucleotide polymorphisms, SnpEff. Fly 6, 80–92 (2012). 9. Kunkle, B. W. et al. Genetic meta-analysis of diagnosed Alzheimer’s disease 36. Shulman, J. M. et al. Genetic susceptibility for Alzheimer disease neuritic pla- identifies new risk loci and implicates Aβ, tau, immunity and lipid processing. que pathology. JAMA Neurol. 70, 1150–1157 (2013). Nat. Genet. 51, 414–430 (2019). 37. Ho, A. & Südhof, T. C. Binding of F-spondin to amyloid-beta precursor protein: 10. Jansen, I. E. et al. Genome-wide meta-analysis identifies new loci and func- a candidate amyloid-beta precursor protein ligand that modulates amyloid- tional pathways influencing Alzheimer’s disease risk. Nat. Genet. 51, 404–413 beta precursor protein cleavage. Proc. Natl Acad. Sci. USA 101, 2548–2553 (2019). (2004). Moreno-Grau et al. Translational Psychiatry (2021)11:142 Page 12 of 12 38. Broce, I. J. et al. Dissecting the genetic relationship between cardiovascular risk 42. Jahanshad, N. et al. Genome-wide scan of healthy human connectome dis- factors and Alzheimer’s disease. Acta Neuropathol. 137, 209–226 (2018). covers SPON1 gene variant influencing dementia severity. Proc. Natl Acad. Sci. 39. Witoelar, A. et al. Meta-analysis of Alzheimer’s disease on 9,751 samples from USA 110, 4768–4773 (2013). Norway and IGAP study identifies four risk loci. Sci. Rep. 8, 18088 (2018). 43. Sherva, R. et al. Genome-wide association study of the rate of cognitive 40. Escott-Price, V. et al. Gene-wide analysis detects two new susceptibility genes decline in Alzheimer’s disease. Alzheimer’s Dement. 10, 45–52 (2014). for Alzheimer’s disease. PLoS ONE 9, e94661 (2014). 44. Laurie, C. C. et al. Detectable clonal mosaicism from birth to old age and its 41. Hoe, H.-S. et al. F-Spondin Interaction with the Apolipoprotein E receptor relationship to cancer. Nat. Genet. 44, 642–650 (2012). ApoEr2 affects processing of amyloid precursor protein. Mol. Cell Biol. 25, 45. Deming, Y. et al. Genome-wide association study identifies four novel loci 9259–9268 (2005). associated with Alzheimer’s endophenotypes and disease modifiers. Acta Neuropathol. 133, 839–856 (2017). Author details The GR@ACE study group C. Abdelnour1,2, N. Aguilera1, E. Alarcón-Martín1, M. Alegret1,2, A. Benaque1, M. Boada1,2, M. Buendía1, P. Cañabate1,2, A. Carracedo5,26, A. Corbatón27, I. de Rojas1, S. Diego1, A. Espinosa1,2, A. Gailhajenet1, P. García González1, S. Gil1, M. Guitart1, A. González Pérez6, I. Hernández1,2, M. Ibarria1, A. Lafuente1, J. Macías7, O. Maroñas4, E. Martín1, M. T. Martínez27, M. Marquié1, A. Mauleón1, G. Monté-Rubio1, L. Montrreal1, S. Moreno-Grau1,2, M. Moreno1, A. Orellana1, G. Ortega1,2, A. Pancho1, E. Pelejà1, A. Pérez-Cordon1, J. A. Pineda8, S. Preckler1, I. Quintela5, L. M. Real7,18, O. Rodríguez-Gómez1,2, M. Rosende-Roca1, A. Ruiz1,2, S. Ruiz1,2, M. E. Sáez6, A. Sanabria1,2, M. A. Santos-Santos1, M. Serrano-Ríos27,28, O. Sotolongo-Grau1, L. Tárraga1,2, S. Valero1,2 and L. Vargas1 27 Centro de Investigación Biomédica en Red de Diabetes y Enfermedades Metabólicas Asociadas, CIBERDEM, Madrid, Spain. 28Hospital Clínico San Carlos, Madrid, Spain DEGESCO consortium A. D. Adarmes-Gómez2,20, E. Alarcón-Martín1, I. Álvarez12, V. Álvarez14,15, G. Amer-Ferrer29, M. Antequera25, C. Antúnez25, M. Baquero30, M. Bernal19, R. Blesa2,8, M. Boada1,2, D. Buiza-Rueda2,20, M. J. Bullido2,9,14, J. A. Burguera30, M. Calero2,21,22, F. Carrillo2,20, M. Carrión-Claro2,20, M. J. Casajeros11, J. Clarimón2,8, J. M. Cruz-Gamero18, M. M. de Pancorbo31, I. de Rojas1,2, T. del Ser10, M. Diez-Fairen12, J. Fortea2,8, E. Franco19, A. Frank-García2,10,32, J. M. García-Alberca17, S. García Madrona22, G. Garcia-Ribas11, P. Gómez-Garre2,20, I. Hernández1,2, S. Hevilla17, S. Jesús2,20, M. A. Labrador Espinosa2,20, C. Lage2,13, A. Legaz, A. Lleó2,8, A. López de Munáin33, S. López-García2,13, D. Macías2,20, S. Manzanares25,34, M. Marín19, J. Marín-Muñoz25, T. Marín17, M. Marquié1,2, A. Martín-Montes2,9,32, B. Martínez25, C. Martínez15,35, V. Martínez25, P. Martínez-Lage Álvarez36, M. Medina2,21, M. Mendioroz Iriarte37, M. Menéndez-González15,38, P. Mir2,20, J. L. Molinuevo39, L. Montrreal1, S. Moreno-Grau1,2, A. Orellana1, A. B. Pastor21, P. Pastor12, J. Pérez-Tur2,40,41, T. Periñán-Tocino2,20, G. Piñol-Ripoll2,16, A. Rábano2,21,23, D. Real de Asúa42, S. Rodrigo19, E. Rodríguez-Rodríguez2,13, J. L. Royo18, A. Ruiz1,2, R. Sanchez del Valle Díaz43, P. Sánchez-Juan2,13, I. Sastre2,19, O. Sotolongo-Grau1, L. Tárraga1,2, S. Valero1,2, M. P. Vicente25 and L. Vivancos25 29 Department of Neurology, Hospital Universitario Son Espases, Palma, Spain. 30Servei de Neurologia, Hospital Universitari i Politècnic La Fe, Valencia, Spain. 31 BIOMICs, País Vasco; Centro de Investigación Lascaray. Universidad del País Vasco UPV/EHU, Vitori-Gasteiz, Spain. 32Neurology Service, Hospital Universitario La Paz (UAM), Madrid, Spain. 33Hospital Donostia de San Sebastián, San Sebastián, Spain. 34Fundación para la Formación e Investigación Sanitarias de la Región de Murcia, Murcia, Spain. 35Servicio de Neurología-Hospital de Cabueñes-Gijón, Gijón, Spain. 36Centro de Investigación y Terapias Avanzadas. Fundación CITA- alzheimer, San Sebastián, Spain. 37Navarrabiomed, Pamplona, Spain. 38Servicio de Neurología -Hospital Universitario Central de Asturias, Oviedo, Spain. 39Barcelona βeta Brain Research Center – Fundació Pasqual Maragall, Barcelona, Spain. 40Unitat de Genètica Molecular. Institut de Biomedicina de València-CSIC, Valencia, Spain. 41 Unidad Mixta de Neurologia Genètica. Instituto de Investigación Sanitaria La Fe, Valencia, Spain. 42Hospital Universitario La Princesa, Madrid, Spain. 43Hospital Clínic Barcelona, Barcelona, Spain