SORL1 Rare Variants: a Major Risk Factor for Familial Early-Onset Alzheimer’S Disease

ORIGINAL ARTICLE SORL1 rare variants: a major risk factor for familial early-onset Alzheimer’s disease

G Nicolas1,2,3,16, C Charbonnier2,3,16, D Wallon2,3,4, O Quenez2,3, C Bellenguez5,6,7, B Grenier-Boley5,6,7, S Rousseau3, A-C Richard3, A Rovelet-Lecrux2, K Le Guennec2, D Bacq8, J-G Garnier8, R Olaso8, A Boland8, V Meyer8, J-F Deleuze8,9, P Amouyel5,6,7, HM Munter10, G Bourque10, M Lathrop10, T Frebourg1,2, R Redon11,12, L Letenneur13, J-F Dartigues13, E Génin14, J-C Lambert5,6,7, D Hannequin1,2,3,4, D Campion2,3,15 and The CNR-MAJ collaborators17

The SORL1 protein plays a protective role against the secretion of the amyloid β peptide, a key event in the pathogeny of Alzheimer’s disease. We assessed the impact of SORL1 rare variants in early-onset Alzheimer’s disease (EOAD) in a case–control setting. We conducted a whole exome analysis among 484 French EOAD patients and 498 ethnically matched controls. After collapsing rare variants (minor allele frequency ≤ 1%), we detected an enrichment of disruptive and predicted damaging missense SORL1 variants in cases (odds radio (OR) = 5.03, 95% conﬁdence interval (CI) = (2.02–14.99), P = 7.49.10−5). This enrichment was even stronger when restricting the analysis to the 205 cases with a positive family history (OR = 8.86, 95% CI = (3.35–27.31), P = 3.82.10 − 7). We conclude that predicted damaging rare SORL1 variants are a strong risk factor for EOAD and that the association signal is mainly driven by cases with positive family history.

Molecular Psychiatry (2016) 21, 831–836; doi:10.1038/mp.2015.121; published online 25 August 2015

INTRODUCTION set of families, we next aimed to determine whether predicted SORL1 (Sortilin-related receptor, L(DLR class A) repeats containing, damaging rare SORL1 variants may contribute to EOAD genetic risk. also known as SORLA or LR11) plays a key role in the trafficking of To address this issue we performed a case–control study on rare the amyloid β (Aβ) precursor protein (APP) as an APP neuronal variants using whole-exome sequencing (WES) data from 484 sorting receptor. By binding to the retromer complex, SORL1 unrelated French EOAD patients and 498 ethnically matched redirects APP from the cell membrane to the Golgi apparatus controls. through the recycling pathway.1 Sorting APP through this pathway results in reduced amyloidogenic processing of APP.1–3 Moreover, SORL1 was recently shown to direct nascent Aβ peptides to the MATERIALS AND METHODS lysosomal compartment, a mechanism that may also contribute to Patients the reduction of the amyloid load.4 SORL1 therefore plays a β Patients were recruited over a 20-year period by the French National CNR- protective role against A neuronal secretion, a central event in MAJ consortium, a memory clinical network including 24 expert centers Alzheimer’s disease (AD) pathogenesis. We previously reported dedicated to the EOAD patient study and care. Diagnoses were performed potentially pathogenic SORL1 rare variants in 5/14 early-onset AD according to the NINCDS-ADRDA (National Institute of Neurological and (EOAD, defined by an age of onset before 65 years) patients who Communicative Disorders and Stroke and the Alzheimer's Disease and were negatively screened for mutations in known autosomal Related Disorders Association) criteria.6 All patients underwent a compre- dominant EOAD genes, namely APP, PSEN1 or PSEN2, despite the hensive clinical examination including personal medical and family history fact that they belonged to pedigrees in which inheritance was assessment, neurological examination, neuropsychological assessment and fl consistent with autosomal dominant transmission.5 The identifica- neuroimaging. When available, cerebrospinal uid (CSF) AD biomarkers had to be consistent with an AD profile. The cutoffs used to define a tion of a nonsense variant among one of these patients supported − 1 − 1 biochemical AD signature were: Aβ42 o550 pg ml , Tau 4350 pg ml the hypothesis that these variants may exert a proamyloidogenic 4 − 1 β 5 and P-Tau 60 pg ml . We also calculated the Tau/A 42 ratio and a effect through a loss of SORL1 function. This hypothesis was value of 40.52 was considered abnormal.7 Two criteria differing by recently reinforced by the observation that one of the reported stringency were used to classify CSF samples as supportive of an AD missense variants led to a decrease in SORL1 ability to mediate Aβ diagnosis: either (1) all three biomarkers were abnormal or (2) two out of 4 lysosomal degradation. Beyond these results obtained in a small three biomarkers and the Tau/Aβ42 ratio were abnormal. When patients

1Department of Genetics, Rouen University Hospital, Rouen, France; 2Inserm U1079, Rouen University, IRIB, Normandy University, Rouen, France; 3CNR-MAJ, Rouen University Hospital, Rouen, France; 4Department of Neurology, Rouen University Hospital, Rouen, France; 5Inserm, U1167, Lille, France; 6Institut Pasteur de Lille, Lille, France; 7Université Lille- Nord de France, Lille, France; 8Centre National de Génotypage, Institut de Génomique, CEA, Evry, France; 9Fondation Jean Dausset, Centre d’études du Polymorphisme Humain, Paris, France; 10McGill University and Génome Québec Innovation Centre, Montréal, QC, Canada; 11Inserm UMR 1087, l'institut du Thorax, CHU Nantes, Nantes, France; 12CNRS, UMR 6291, Université de Nantes, Nantes, France; 13Inserm U897, Univ Bordeaux, Bordeaux, France; 14Inserm UMR1078, CHU Brest, Univ Bretagne Occidentale, Brest, France and 15Department of Research, Rouvray Psychiatric Hospital, Sotteville-lès-Rouen, France. Correspondence: Dr D Campion, Inserm U1079, Faculté de Médecine, 22 Boulevard Gambetta, Rouen 76183, France. E-mail: [email protected] 16These two authors contributed equally to this work. 17CNR-MAJ collaborators are listed before references. Received 16 April 2015; revised 13 July 2015; accepted 14 July 2015; published online 25 August 2015 SORL1 rare variants G Nicolas et al 832 had negative CSF results, the diagnosis of AD was not retained and they were not included. All patients gave informed, written consent for genetic analyses. This study was approved by our ethics committee. Blood samples were neuropath extracted using the Qiagen (Hilden, Germany) DNA blood kit. APOE Percentage genotyping was performed by sequencing. Only patients with no PSEN1, PSEN2 or APP mutation (detected either by Sanger sequencing or by WES), rmed AD diagnosis. ﬁ no APP duplication (detected by QMPSF (quantitative multiplex PCR of CSF ﬂ 8

short uorescent fragments) and no C9ORF72 expansion (detected Percentage following the protocol described by DeJesus-Hernandez et al.9) were included in this study. We finally included a total of 485 unrelated French patients with EOAD (age of onset ≤ 65 years). Of note, 52.9% had a disease 0.0213) 54.1% 1% ≤ 0.0162) 51.5% 2% onset at 55 years. All but one passed quality controls (see below): 196 0.0117) NA NA – – men (40.5%) and 288 women (59.5%). We verified that none of them – carried a pathogenic variant in a list of 20 other dementia-causing genes.10 Ten of them (2%) had a definite diagnosis of AD (neuropathological

conﬁrmation) and the other 474 (98%) were diagnosed with probable AD. MAF (95% CI) Among them, 249 had CSF biomarkers indicative of AD (51.4% from all 0.005 (0.0021 0.0073 (0.0025 patients). The remaining 225 (46.5%) had no CSF biomarker available and 0.0083 (0.0042 were carefully selected using neuropsychological assessments (evidence of a progressive hippocampal amnestic syndrome associated with another 24 31 10 = = cognitive dysfunction) and evidence of neuronal injury using imaging = (6.4%) (2.0%) N N N

criteria (AD pattern of a cortical atrophy on magnetic resonance imaging (11.7%) and/or decreased 18flurodesoxyglucose uptake on positron emission tomography). 110 214 Family history was obtained by family interview. When available, medical 102 = = reports of the relatives were examined. A positive family history was defined = (53.7%) (44.2%) (20.5%) N N by the presence of at least one first-degree relative diagnosed with AD or with N a reported phenotype consistent with AD, whatever the age of onset in the affected relative. Among the 484 cases, 205 (42%) had a positive family 60 206 history, 230 (48%) had sporadic AD and the family history was unknown in 315 = = =

ﬁ N (29.3%) (42.6%) (63.3%) N 49 (10%). The 205 cases with positive family history included 2 de nite and N 203 probable AD cases, 111 of whom had CSF biomarkers indicative of AD. Among the 484 patients, the distribution of APOE genotypes was 6 as follows: ε2/ε2(n = 1, 0.2%), ε2/ε3(n = 21, 4.3%), ε3/ε3(n = 206, 42.6%), 4 11 = = ε2/ε4(n = 11, 2.3%), ε3/ε4(n = 214, 44.2%) and ε4/ε4(n = 31, 6.4%) = N N APOE distribution TREM2 p.R47H variant (1.2%) (1.9%) (2.3%) (Table 1). We included only 31 out of 188 APOE ε4ε4 EOAD patients initially N eligible for this study. These patients were either sporadic cases with an age of onset at ≤ 55 years (7 patients) or had a positive family history 6 63 that had to occur in at least two generations (24 patients). 21 = = = N (4.3%) (2.9%) N N s disease; MAF, minor allele frequency; NA, not available; Percentage CSF: percentage of patients with cerebrospinal ’ (12.6%) Controls A total of 500 controls were recruited in 5 different French cities and were 2 1 all of French ancestry (498 passed the quality control, see below). Of these, 1 = = ≥ = N N 301 had an age at inclusion of 55 years and had normal Mini Mental N (0.4%) (0.2%) (0.5%) State Examination score, according to age and education level. The APOE22 APOE23 APOE24 APOE33 APOE34 APOE44 remaining controls were healthy blood donors. Mean age (± s.d.) was: 66.6 ±18.3 years (range: 19–103).

Whole-exome sequencing

Exomes were captured using Agilent SureSelect Human All Exon kits females) uid; EOAD, early-onset Alzheimer

(Agilent Technologies, Santa Clara, CA, USA). Final libraries were sequenced Sex (percentage fl on a HiSeq2000 or 2500 (Illumina, San Diego, CA, USA) with paired ends, 76 or 100 bp reads. Exome samples were all processed through the same bioinformatics pipeline following GATK 3.3-0 Best Practices recommendations.11 Reads were mapped to the 1000 Genomes GRCh37 12 65) 59.5% build using BWA 0.7.5a. Picard Tools 1.101 (http://picard.sourceforge.net) 65) 59.5% 103) 57.1% – – was used to flag duplicate reads. We applied GATK for short insertion and – s disease (patients with CSF not indicative of AD were not included); Percentage Neuropath., percentage of patients with a neuropathologically con deletion (indels) realignment, base quality score recalibration and single- ’ nucleotide variants and indel discovery using the Haplotype Caller across 55.3 (44 all samples simultaneously. The joint variant calling file was annotated with 56.6 (45 refGene gene regions, variant effects and functional effect prediction tools 13 fi using Annovar. We then extracted high-quality (de ned as variants with inclusion (control) (ranges) a Variant Quality Score VQSLOD above − 2) exonic and splice site variants Mean age at onset (cases) or dence interval; CSF, cerebrospinal

(located ± 2 bp around each coding exon) with a minor allele frequency ﬁ (MAF) of o1% in our whole data set within each refGene gene region. Only genotypes satisfying the following quality ﬁlters were retained for 498) 66.6 (19

analysis: genotype read depth 410 and genotype quality 450. Of note, all = n

rare SORL1 variants retained for analysis were detected with high Demographic features of the 484 patients and the 498 controls conﬁdence among all variant carriers: observed read depth ≥ 20 and 484) genotype quality ≥ 88. We considered nonsense, frameshift indels and 205) = = n ﬁ n

splice site variants as disruptive. We classi ed missense variants as strictly uid biomarkers indicative of Alzheimer Controls ( EOAD patients ( Including EOAD patients with positive family history ( ﬂ damaging if they were simultaneously predicted damaging by the Table 1. Abbreviations: CI, con

Molecular Psychiatry (2016), 831 – 836 © 2016 Macmillan Publishers Limited SORL1 rare variants G Nicolas et al 833 following three softwares: Polyphen2 HumDiv, Mutation Taster and SIFT.13 expression ratio was calculated by reporting the ratio between the peak Allele frequencies from the Exome Aggregation Consortium database area of the two mutant alleles and/or rs78274293 in complementary DNA (ExAC) obtained from the open access web resource http://exac.broad to the same peak ratio measured on genomic DNA. institute.org/ were added manually for each SORL1 variant passing the filters. RESULTS Sanger sequencing We conducted WES with high depth of coverage (mean ~ 120 ×) Nonsynonymous SORL1 variants with a MAF o1% in our data set were on 484 well-characterized EOAD French patients and 498 confirmed by Sanger sequencing using previously published primers.5 ethnically matched controls. We extracted all SORL1 exonic and flanking intronic (splice sites, ± 2 bp) high-quality variants. As Quality checks nonsense, frameshift short insertions and deletions (indels) and splice site (thereafter grouped under the name of disruptive) Individuals and variants were analyzed through every recommended fi quality control steps for genetic case–control association studies.14 Most variants are predicted to result in a loss of SORL1 function, we rst checks were carried out with PLINK 1.9 (https://www.cog-genomics.org/ analyzed these variants and identified 9 disruptive variants (5 plink2). All individuals in the sample were (1) checked for concordant sex nonsense, 3 frameshift indels and 1 splice site) in patients and information using Plink–sex-check,15 (2) discarded when identified with none in controls (Table 2). All of them were singletons and none of significantly higher heterozygosity rates (and contamination confirmed if them was present in the ExAC database (accessed February 2015, relevant using the VerifyBamID software, http://csg.sph.umich.edu/kang/ Table 1). RNA was available for three patients carrying disruptive verifyBamID/) and (3) discarded if Plink pi_hat relatedness estimation variants. For each patient, we observed reduced expression of the exceeded 18.5%. Out of 485 initially included cases and 500 controls, 484 allele carrying the disruptive variant (see Supplementary Figure 1 cases and 498 controls passed these quality checks and were retained for 5 association analysis. No individual of divergent ancestry was detected by and Pottier et al. ). To investigate the role of rare missense Plink–neighbor function. In particular, ethnical match between cases and variants, we focused on variants with a MAF o1% in our entire controls was confirmed by principal component analysis on common data set. We identified a total of 50 rare missense variants: 37 in variants (MAF 45%) after exclusion of long-range linkage disequilibrium cases (56 variant carriers) and 17 in controls (36 carriers). To regions and variant pruning on linkage disequilibrium (r2 40.2). reduce the risk of including putatively benign variants in our gene- o 4 None of the SORL1 variants with a MAF 1% (1) were missing in 5% based collapsing test, we restricted our analysis to missense fi – of individuals, (2) showed a signi cant deviation from Hardy Weinberg variants that were predicted as damaging by three softwares equilibrium and (3) presented significantly different missing call rates between cases and controls, as confirmed by Plink–test-missing at a (referred to as strictly damaging variants): Polyphen2 HumDiv, threshold of 1.10 − 3. Mutation Taster and SIFT. On top of the 9 disruptive variants, we identified 15 strictly damaging missense variants in 19 patients (collapsed MAF of disruptive plus strictly damaging variants = Statistics 2.9%, all but one were singletons) and one strictly damaging o – All rare (MAF 1% on the whole case control sample) disruptive and missense variant in 6 controls (collapsed MAF = 0.006%; Table 2). strictly damaging missense SORL1 variants were imported into Variant Association Tools (VAT).16 Unadjusted gene-level collapsing tests were Note that the sole strictly damaging recurrent variant, p.D2065V, computed using the CFisher function grouped by RefGene name that was found in both patients and controls. We checked for potential corresponds here to Fisher’s exact test comparing the proportions of confounding because of fine-scale population stratification by variant carriers among cases and controls. Corresponding odds ratios (ORs) visual inspection of the quantile–quantile plots and computation and confidence intervals (CIs) were computed with the R statistical of genomic inflation factors λ (until λ is not greater than 1, no software (http://www.R-project.org/). The same quality checks and gene- evidence for confounding is found and no adjustment is required). level association tests were applied to every RefGene coding region to The cumulative effect of disruptive and strictly damaging fl λ extract all possible gene-level P-values. The genomic in ation factor was missense variants resulted in an OR of 5.03 (95% CI = 2.02–14.99, then computed with the estlamba function of the R statistical package − 5 λ genABEL. Gene-level collapsing tests adjusted on APOE ε4+ status were P = 7.49.10 , genomic control = 0.87), approaching exome-wide fi − 6 computed by adopting the logistic regression framework of the CollapseBt signi cance (Bonferroni threshold of 2.5.10 accounting for a function in VAT. Corresponding ORs and CIs were computed from the theoretical number of ~ 20 000 gene-level tests). This was the β-coefficient estimate and Wald statistics provided by VAT. Again, we strongest enrichment signal detected at the gene level among the computed the corresponding genomic inflation factor λ. All genomic entire WES data set. The OR remained of the same order of inflation factors were o1, excluding possible confounding from fine-scale magnitude (OR = 5.38, 95% CI = 1.97–17.02, P = 2.95.10 − 4) when population stratification. This was further confirmed by projecting SORL1 fi restricting the analysis to the 259 cases with the highest level of variant carriers onto the map of the rst two principal components evidence of an AD diagnosis (neuropathological diagnosis or computed exome-wide on common variants (MAF 45%) after exclusion of long-range linkage disequilibrium regions and variant pruning on linkage positive CSF AD biomarkers). Interestingly, 8 out of 9 patients with disequilibrium (r2 40.2). No evidence of confounding from fine-scale a disruptive variant had a positive family history (the mother of population stratification could be detected and no adjustment on principal the ninth patient deceased from myocardial infarction at the age components was found necessary. of 60 years, leading to a censoring effect regarding AD), as well as 12/16 patients harboring a strictly damaging variant and for SNaPshot analysis whom available family history information was available (total: Blood was collected into PAXgene Blood RNA tubes (Qiagen). Total RNA 20/25, 80%). To better delineate the role of SORL1 disruptive and was extracted from whole blood using the PAXgene Blood RNA kit strictly damaging missense variants in EOAD risk, we next focused (Qiagen). RNA was then reverse-transcribed using the Verso complemen- on the 205 patients with positive family history and found an tary DNA kit (Fisher Scientific, Illkirch, France) with a blend of random exome-wide significant enrichment in cases versus controls with hexamers and anchored oligo-dT (3:1). Allele-specific expression was an OR of 8.86 (95% CI = 3.35–27.31, P = 3.82.10 − 7, λ = 0.94; measured using the SNaPshot (PE Applied Biosystems, Foster City, CA, fi 5 Figure 1). Again, SORL1 was the only exome-wide signi cant hit, USA) technique, as previously described. The genomic DNA and the and remained by far the strongest enrichment signal as shown by complementary DNA were PCR amplified using primers flanking the 4 Figure 1. These results remained robust after adjustment on APOE mutation. The SORL1 polymorphism rs78274293 (c.2499A T), which is ε – − 5 heterozygous for the control and both patients, and the two disruptive 4+ genotype (OR = 8.78, 95% CI = 3.22 23.93, P = 2.19.10 , mutations, c.162delC and c.3647G4A, were used as markers for allele- λ = 0.68). The OR was of the same order of magnitude (OR = 9.69, − 6 specific expression measurement. Fluorescence was analyzed using an 95% CI = 3.27–32.24, P = 6.55.10 ) when restricting the analysis to automated sequencer (ABI 3100; Applied Biosystems). The allele-specific the 113 patients with a positive family history among the 259

Table 2. Disruptive and strictly damaging missense variants in cases and controls and demographic data

Sample ID Status Agea Sex APOE TREM2 Family history Chr coordinates Variant nomenclatureb Classiﬁcation ExAC dbsnp138, variant rs ID p.R47H

– ROU-0055-001 Case 55 M 34 Absent 1 chr11:121385032 c.1211+2T4G Disruptive (splice) 0 – 836 EXT-0400-001 Case 50 M 34 Absent 1 chr11:121323202 c.162delC, p.D54fs Disruptive (frameshift deletion) 0 – ALZ-0167-015 Case 60 F 44 Absent 1 chr11:121416025 c.1938delA, p.G646fs Disruptive (frameshift deletion) 0 – EXT-0438-001 Case 50 F 34 Absent 1 chr11:121428054 c.2603delC, p.T868fs Disruptive (frameshift deletion) 0 – ROU-0699-001 Case 52 F 34 Absent 1 chr11:121367621 c.802C4T, p.R268X Disruptive (nonsense) 0 – ROU-1409-001 Case 52 F 33 Absent 0 chr11:121454233 c.3647G4A, p.W1216X Disruptive (nonsense) 0 – EXT-0050-001 Case 56 F 34 Absent 1 chr11:121466396 c.4434C4A, p.C1478X Disruptive (nonsense) 0 – EXT-0017-001 Case 60 F 34 Absent 1 chr11:121485623 c.5463G4A, p.W1821X Disruptive (nonsense) 0 – EXT-0087-001 Case 48 M 33 Absent 1 chr11:121492969 c.6163G4T, p.E2055X Disruptive (nonsense) 0 – ALZ-0579-001 Case 63 M 34 Absent 1 chr11:121340802 c.372C4G, p.S124R Missense, strictly damaging 0 – SORL1 ALZ-426-010 Case 46 M 33 Absent 1 chr11:121348846 c.422A4G, p.Y141C Missense, strictly damaging 0 – EXT-0290-001 Case 60 M 34 Absent 1 chr11:121383766 c.994C4T, p.R332W Missense, strictly damaging 0.00005766 – Nicolas G EXT-0833-001 Case 60 M 34 Absent 1 chr11:121393308 c.1418G4C, p.C473S Missense, strictly damaging 0 – variants rare EXT-049-002 Case 63 F 33 Absent 1 chr11:121393633 c.1531G4C, p.G511R Missense, strictly damaging 0 – B00EXRC Case 57 F 34 Absent U chr11:121403204 c.1628G4A, p.G543E Missense, strictly damaging 0 –

ROU-309-001 Case 65 F 33 Absent 1 chr11:121416047 c.1960C4T, p.R654W Missense, strictly damaging 0 – al et EXT-0524-001 Case 62 F 24 Absent 1 chr11:121421298 c.2185C4T, p.R729W Missense, strictly damaging 0.000008274 – B00EXRD Case 48 F 34 Absent U chr11:121424795 c.2416G4A, p.D806N Missense, strictly damaging 0 – B00EXT2 Case 60 F 34 Absent U chr11:121454242 c.3656A4G, p.D1219G Missense, strictly damaging 0 – ROU-1227-001 Case 56 F 34 Absent 1 chr11:121456951 c.3727C4T, p.R1243C Missense, strictly damaging 0.000008237 rs201196740 EXT-0047-001 Case 47 F 33 Present 0 chr11:121460836 c.4166A4T, p.D1389V Missense, strictly damaging 0 – EXT-0205-001 Case 59 F 33 Absent 1 chr11:121477566 c.4961C4T, p.P1654L Missense, strictly damaging 0 – ALZ-0244-011 Case 46 F 24 Absent 1 chr11:121477647 c.5042G4A, p.G1681D Missense, strictly damaging 0 – ALZ-0446-001 Case 52 M 33 Absent 0 chr11:121495816 c.6194A4T, p.D2065V Missense, strictly damaging 0.002823 rs140327834 B00EXT2 Case 60 F 34 Absent U chr11:121495816 c.6194A4T, p.D2065V Missense, strictly damaging 0.002823 rs140327834 EXT-094-001 Case 56 F 34 Absent 1 chr11:121495816 c.6194A4T, p.D2065V Missense, strictly damaging 0.002823 rs140327834 EXT-0691-001 Case 59 M 23 Absent 0 chr11:121495816 c.6194A4T, p.D2065V Missense, strictly damaging 0.002823 rs140327834 ROU-1216-001 Case 54 M 34 Absent 1 chr11:121495816 c.6194A4T, p.D2065V Missense, strictly damaging 0.002823 rs140327834 ROU-1261-001 Case 52 F 23 Absent 0 chr11:121495816 c.6194A4T, p.D2065V Missense, strictly damaging 0.002823 rs140327834 B00FWYF Control 87.8 F 33 Absent NA chr11:121495816 c.6194A4T, p.D2065V Missense, strictly damaging 0.002823 rs140327834 B00FWYY Control 78.4 M 33 Absent NA chr11:121495816 c.6194A4T, p.D2065V Missense, strictly damaging 0.002823 rs140327834 B00G72B Control 22 F 33 Absent NA chr11:121495816 c.6194A4T, p.D2065V Missense, strictly damaging 0.002823 rs140327834 B00G7JR Control 92 F 33 Absent NA chr11:121495816 c.6194A4T, p.D2065V Missense, strictly damaging 0.002823 rs140327834 ROU-430-002 Control 68 M 33 Absent NA chr11:121495816 c.6194A4T, p.D2065V Missense, strictly damaging 0.002823 rs140327834 ROU-601-002 Control 75 M 33 Absent NA chr11:121495816 c.6194A4T, p.D2065V Missense, strictly damaging 0.002823 rs140327834 06McilnPbihr Limited Publishers Macmillan 2016 © Abbreviations: Chr, chromosome; ExAC, Exome Aggregation Consortium; F, female; M, male; NA, not applicable; U, unknown. Family history is noted ‘1’ if at least one ﬁrst-degree relative presented Alzheimer's disease whatever the age of onset. Note that one case carries two strictly damaging variants and that collapsing tests compare the proportion of variant carriers among cases and controls. aAge at onset (cases) or age at inclusion or at last examination (controls). bAccording to NM_003105. Chromosomal coordinates refers to human genome GRCh37/hg19 version. Missense, strictly damaging refers to missense variants which are predicted deleterious by SIFT, probably damaging by Polyphen-2 HDiv, and disease causing by Mutation Taster. SORL1 rare variants G Nicolas et al 835 ACKNOWLEDGMENTS This study was funded by grants from the Clinical Research Hospital Program from the French Ministry of Health (GMAJ, PHRC 2008/067), the CNR-MAJ, the JPND PERADES and France Génomique. This work was supported by Labex GENMED ANR- 10-LABX-0013, the National Foundation for Alzheimer’s disease and related disorders, the Institut Pasteur de Lille, the Centre National de Génotypage, Inserm, FRC (fondation pour la recherche sur le cerveau) and Rotary. This work has been supported by the LABEX (laboratory of excellence program investment for the future) DISTALZ grant (Development of Innovative Strategies for a Transdisciplinary approach to ALZheimer’s disease). The funding sources had no speciﬁc roles.

CNR-MAJ COLLABORATORS Didier Hannequin, Dominique Campion, David Wallon, Olivier Martinaud, Gaël Nicolas (Centre Hospitalo-Universitaire (CHU), Rouen); Olivier Godefroy (CHU Amiens); Frédérique Etcharry-Bouyx, Valérie Chauviré (CHU Angers); Ludivine Chamard, Eric Berger, Eloi Magnin (CHU Besancon); Jean-Francois Dartigues, Sophie Auriacombe (CHU Bordeaux); Vincent de la Sayette, Fausto Viader (CHU Caen); Dominique Castan (CH Castres), Elsa Dionet (CHU Clermont- Ferrand); Francois Sellal (CH Colmar); Olivier Rouaud, Christel Thauvin (CHU Dijon); Olivier Moreaud, Mathilde Sauvée (CHU Grenoble); Adeline Rollin-Sillaire, Stéphanie Bombois, Marie-Anne Mackowiak, Vincent Deramecourt, Florence Pasquier (CHU Lille); Maïté Formaglio, Hélène Mollion, Isabelle Roullet-Solignac, Figure 1. Quantile–quantile plot of gene-level Fisher’s P-values among 205 early-onset Alzheimer’s disease (EOAD) cases with Alain Vighetto, Bernard Croisile (CHU Lyon); Mira Didic, Olivier Félician, Lejla positive family history and 498 controls (n = 13630 tests). Koric, Mathieu Ceccaldi (CHU Marseille); Audrey Gabelle, Cecilia Marelli, Jacques Touchon, Pierre Labauge (CHU Montpellier); Thérèse Jonveaux (CHU Nancy); Martine Vercelletto, Claire Boutoleau-Bretonnière (CHU Nantes); Giovanni cases with the highest level of evidence of an AD diagnosis Castelnovo (CHU Nimes); David Renaud, Philippe Robert (CHU de Nice); Claire (neuropathological diagnosis or positive CSF AD biomarkers). Paquet, Julien Dumurgier, Jacques Hugon (CHU Lariboisière, Paris); Agnès Michon, Isabelle Le Ber, Bruno Dubois, Charles Duyckaerts (CHU La Salpêtrière, Paris); Foucauld De Boisgueheneuc (CHU Poitiers); Serge Belliard (CHU Rennes); DISCUSSION Serge Bakchine (CHU de Reims); Marie-Odile Barrellon, Bernard Laurent (CHU SORL1 association signals with common variants have been Saint-Etienne); Frédéric Blanc, Christine Tranchant (CHU Strasbourg); Jérémie 1,17 previously identified. However, the effect of these noncoding Pariente, Michèle Puel (CHU Toulouse); Caroline Hommet, Karl Mondon (CHU 18 variants, which might influence SORL1 expression, was weak Tours). (OR = 0.77, 95% CI = 0.72–0.82).17 To our knowledge, along with our previous study,5 only one other study reported an enrichment of rare coding SORL1 variants in AD patients, mainly diagnosed REFERENCES with late-onset AD.19 In the latter study, a joint burden of 17 low- 1 Rogaeva E, Meng Y, Lee JH, Gu Y, Kawarai T, Zou F et al. The neuronal sortilin- frequency and rare coding variants was found to be associated related receptor SORL1 is genetically associated with Alzheimer disease. with late-onset AD (adjusted P = 0.0009). Here, we extend the Nat Genet 2007; 39:168–177. conclusions of our previous study5 and demonstrate for the first 2 Dodson SE, Andersen OM, Karmali V, Fritz JJ, Cheng D, Peng J et al. Loss of LR11/ time the effect of rare coding SORL1 variants on EOAD risk. We SORLA enhances early pathology in a mouse model of amyloidosis: evidence for a proximal role in Alzheimer's disease. J Neurosci 2008; 28: 12877–12886. found a strong positive association when considering our whole fi 3 Andersen OM, Reiche J, Schmidt V, Gotthardt M, Spoelgen R, Behlke J et al. data set and identi ed that this effect was mostly driven by cases Neuronal sorting protein-related receptor sorLA/LR11 regulates processing with a positive family history. When restricting our analysis to this of the amyloid precursor protein. Proc Natl Acad Sci USA 2005; 102: subset of patients, exome-wide significance was reached. In our 13461–13466. WES study, SORL1 was the top hit and the sole gene for which 4 Caglayan S, Takagi-Niidome S, Liao F, Carlo AS, Schmidt V, Burgert T et al. exome-wide significance for the association test with collapsed Lysosomal sorting of amyloid-beta by the SORLA receptor is impaired by a familial rare variants was reached. Considering recent lessons on rare Alzheimer's disease mutation. Sci Transl Med 2014; 6: 223ra20. variant association testing,20,21 we draw a specific attention to four 5 Pottier C, Hannequin D, Coutant S, Rovelet-Lecrux A, Wallon D, Rousseau S – et al. High frequency of potentially pathogenic SORL1 mutations in autosomal crucial points: (1) stringent selection of cases, (2) case control 17 – matching on ancestry, (3) high depth sequencing to improve dominant early-onset Alzheimer disease. Mol Psychiatry 2012; : 875 879. fi 6 McKhann GM, Knopman DS, Chertkow H, Hyman BT, Jack CR Jr, Kawas CH et al. variant calling ef ciency and guarantee homogeneity of variant The diagnosis of dementia due to Alzheimer's disease: recommendations from detection across individual samples and (4) use of stringent the National Institute on Aging-Alzheimer's Association workgroups on diagnostic filtering criteria based on in silico deleteriousness predictions. Not guidelines for Alzheimer's disease. Alzheimers Dement 2011; 7:263–269. strictly fulfilling these four criteria might contribute to explain why 7 Duits FH, Prins ND, Lemstra AW, Pijnenburg YA, Bouwman FH, Teunissen CE et al. several association studies based on rare exonic variants failed in Diagnostic impact of CSF biomarkers for Alzheimer's disease in a tertiary detecting significant signals in other neuropsychiatric diseases memory clinic. Alzheimers Dement 2014; 11:523–532. despite larger sample sizes.22,23 The high ORs obtained in our 8 Rovelet-Lecrux A, Hannequin D, Raux G, Le Meur N, Laquerriere A, Vital A et al. sample after restriction to cases with positive family history are APP locus duplication causes autosomal dominant early-onset Alzheimer disease 38 – strong arguments in favor of a role of rare SORL1 variants in EOAD with cerebral amyloid angiopathy. Nat Genet 2006; :24 26. 9 DeJesus-Hernandez M, Mackenzie IR, Boeve BF, Boxer AL, Baker M, Rutherford NJ genetic determinism, but these orders of magnitude should not et al. Expanded GGGGCC hexanucleotide repeat in noncoding region of be expected on random AD samples. C9ORF72 causes chromosome 9p-linked FTD and ALS. Neuron 2011; 72: 245–256. 10 Beck J, Pittman A, Adamson G, Campbell T, Kenny J, Houlden H et al. Validation CONFLICT OF INTEREST of next-generation sequencing technologies in genetic diagnosis of dementia. The authors declare no conflict of interest. Neurobiol Aging 2014; 35:261–265.

© 2016 Macmillan Publishers Limited Molecular Psychiatry (2016), 831 – 836 SORL1 rare variants G Nicolas et al 836 11 McKenna A, Hanna M, Banks E, Sivachenko A, Cibulskis K, Kernytsky A et al. The 18 Young JE, Boulanger-Weill J, Williams DA, Woodruff G, Buen F, Revilla AC et al. Genome Analysis Toolkit: a MapReduce framework for analyzing next-generation Elucidating molecular phenotypes caused by the SORL1 Alzheimer's disease DNA sequencing data. Genome Res 2010; 20:1297–1303. genetic risk factor using human induced pluripotent stem cells. Cell Stem Cell 12 Li H, Durbin R. Fast and accurate long-read alignment with Burrows-Wheeler 2015; 16: 373–385. transform. Bioinformatics 2010; 26:589–595. 19 Vardarajan BN, Zhang Y, Lee JH, Cheng R, Bohm C, Ghani M et al. Coding muta- 13 Wang K, Li M, Hakonarson H. ANNOVAR: functional annotation of genetic variants tions in SORL1 and Alzheimer disease. Ann Neurol 2015; 77:215–227. from high-throughput sequencing data. Nucleic Acids res 2010; 38: e164. 20 Babron MC, de Tayrac M, Rutledge DN, Zeggini E, Genin E. Rare and low frequency 14 Anderson CA, Pettersson FH, Clarke GM, Cardon LR, Morris AP, Zondervan KT. variant stratification in the UK population: description and impact on Data quality control in genetic case-control association studies. Nat Protoc 2010; association tests. PLoS One 2012; 7: e46519. 5: 1564–1573. 21 Do R, Stitziel NO, Won HH, Jorgensen AB, Duga S, Angelica Merlini P et al. Exome 15 Chang CC, Chow CC, Tellier LC, Vattikuti S, Purcell SM, Lee JJ. Second-generation sequencing identifies rare LDLR and APOA5 alleles conferring risk for myocardial PLINK: rising to the challenge of larger and richer datasets. Gigascience 2015; 4:7. infarction. Nature 2015; 518:102–106. 16 Wang GT, Peng B, Leal SM. Variant association tools for quality control and 22 Liu L, Sabo A, Neale BM, Nagaswamy U, Stevens C, Lim E et al. Analysis of rare, analysis of large-scale sequence and genotyping array data. Am J Hum Genet exonic variation amongst subjects with autism spectrum disorders and popula- 2014; 94: 770–783. tion controls. PLoS Genet 2013; 9: e1003443. 17 Lambert JC, Ibrahim-Verbaas CA, Harold D, Naj AC, Sims R, Bellenguez C et al. 23 Purcell SM, Moran JL, Fromer M, Ruderfer D, Solovieff N, Roussos P et al. A Meta-analysis of 74,046 individuals identifies 11 new susceptibility loci for polygenic burden of rare disruptive mutations in schizophrenia. Nature 2014; 506: Alzheimer's disease. Nat Genet 2013; 45: 1452–1458. 185–190.

Supplementary Information accompanies the paper on the Molecular Psychiatry website (http://www.nature.com/mp)