|Home | About | Journals | Submit | Contact Us | Français|
Studying rare extreme forms of Alzheimer disease (AD) may prove to be a useful strategy in identifying new genes involved in monogenic determinism of AD. Amyloid precursor protein (APP), PSEN1, and PSEN2 mutations account for only 85% of autosomal dominant early-onset AD (ADEOAD) families. We hypothesised that rare copy number variants (CNVs) could be involved in ADEOAD families without mutations in known genes, as well as in rare sporadic young-onset AD cases. Using high-resolution array comparative genomic hybridisation, we assessed the presence of rare CNVs in 21 unrelated ADEOAD cases, having no alteration on known genes, and 12 sporadic AD cases, with an age of onset younger than 55 years. The analysis revealed the presence of 7 singleton CNVs (4 in ADEOAD and 3 in sporadic cases) absent in 1078 controls and 912 late-onset AD cases. Strikingly, 4 out of 7 rearrangements target genes (KLK6, SLC30A3, MEOX2, and FPR2) encoding proteins that are tightly related to amyloid-β peptide metabolism or signalling. Although these variants are individually rare and restricted to particular subgroups of patients, these findings support the causal role, in human pathology, of a set of genes coding for molecules suspected for a long time to modify Aβ metabolism or signalling, and for which animal or cellular models have already been developed.
The identification of major genes (PSEN1, PSEN2 and APP) responsible for autosomal dominant early-onset Alzheimer disease (ADEOAD, OMIM number #104300) has proved crucial to understanding the pathophysiology of the disease. It has allowed for the definition of the amyloid cascade hypothesis,1, 2 according to which accumulation of highly aggregable forms of the amyloid beta (Aβ) peptide, that results from the amyloid precursor protein (APP) proteolytic cleavage, is the primum movens of this devastating neurodegenerative disease.
ADEOAD is a rare condition with a prevalence rate estimated at 5.3 per 100000 persons at risk.2 Mutation screening of more than 150 ADEOAD families, ascertained in France by the National Centre for early-onset AD (CNR-MAJ), had shown that mutations of known genes account for 85% of ADEOAD families.3 We hypothesised that rare copy number variants (CNVs) could be involved in ADEOAD families without mutations in known genes. Indeed, a previous analysis has shown that APP locus duplications were present in a subset of these families.4 To further explore this issue, we conducted a genome-wide screen for CNVs, using high-resolution oligonucleotide array-based comparative genomic hybridisation (a-CGH) in still unexplained ADEOAD families. In addition, as rare pathogenic CNVs often occur de novo,5 we decided to investigate by a-CGH another subgroup of patients consisting of 12 sporadic cases with particularly young disease onset (age of onset before 55 years). We reasoned that this group is likely to be enriched in de novo cases, even if, due to the pedigree structure, no parental DNA was available for confirmation. Our aim was to identify causative CNVs exclusively associated with these two disease groups and therefore, expected to be absent in a set of 1078 controls. Moreover, to assess the specificity of our findings with respect to these particular AD phenotypes, we also looked for the presence of the CNVs with exclusive association to these two disease groups in a third cohort of 912 AD cases not harbouring these extreme phenotypes. Note that in these population of unselected AD cases, previous studies have failed to detect CNV burden.6, 7
We used a-CGH to search for pathogenic CNVs in two discovery samples corresponding to two highly selected disease groups. The first disease group included 23 unrelated probands from ADEOAD families, in which a previous screen had failed to identify any mutation on the PSEN1, PSEN2, and APP (exons 16 and 17) genes or any APP gene-dosage alteration. The pattern of inheritance was consistent with ADEOAD (ie, several AD cases with onset <60 years in at least two generations), but absence of DNA samples for a sufficient number of relatives precluded linkage analysis in these kindreds (Supplementary Figure S1). The second disease group included 12 sporadic AD cases with onset before age 55 (range 44–54 years, mean 49). These patients had been negatively screened for the same genes as ADEOAD patients. In addition, we retained only the non-carriers of an APOE4 allele to exclude a potential causal role of this potent risk factor. Patients were Caucasian of French origin, except for two patients of Italian origin and another of Moroccan origin. The 912 other AD patients (506 males, 406 females, age of onset 68.2±8.5 years, range 45–83) used to assess the specificity of our findings with respect to the two disease groups were Caucasian of French origin ascertained in Rouen (West of France). All patients fulfilled the NINCDS-ADRDA criteria for probable AD.8 The 1078 controls were mainly patients' spouses. Ethical approval was obtained from the Paris Necker and Nord-Ouest 1 Ethics Committees, and all participants gave signed informed consent.
Total DNA was prepared from peripheral blood lymphocytes using the Flexigen extraction kit (Qiagen, Hilden, Germany). High-resolution a-CGH analysis was performed using the Human High-Resolution Discovery Microarray Kit 1 × 1M (Agilent Technologies, Santa Clara, CA, USA), using standard recommended protocols, except for patient ALZ466, for whom a-CGH analysis was performed using Agilent SurePrint G3 Human 4 × 180K catalog array (Agilent Technologies). A non-commercial genomic DNA pool of 10 control individuals was used as a reference sample. Hybridisation results were extracted with the Feature Extraction Software (10.5.1.1, Agilent Technologies, Santa Clara, CA, USA) and analyzed using Agilent's DNA-analytics software (version 4.0.81, Agilent Technologies). The data were processed using the ADM-2 algorithm, with the threshold set at 6.0 SD. A rearrangement was defined by the deviation of at least five consecutive probes.
Quantitative multiplex PCR of short fluorescent fragments (QMPSF) was performed as previously described.9 Briefly, short genomic fragments (between 100 and 320bp) of the genes subjected to CNVs were simultaneously amplified within one multiplex PCR, using 6-FAM-labelled primer pairs (Supplementary Table S1). PCR reactions were performed in 25μl containing 100ng of genomic DNA, 0.12–0.2μM of each pair of primers, and 1unit of Thermoprime plus DNA polymerase (Abgene, Courtaboeuf, France). An amplicon of the HMBS gene was used as control. DNA fragments generated by QMPSF were separated on an ABI Prism 3100 DNA sequencer (Applied Biosystems, Courtaboeuf, France), and the resulting fluorescence profiles were analysed using the Genescan 3.7 Software (Applied Biosystems). Electropherogram of the patient (in red) was superimposed to that of a normal control (in blue) by adjusting the peaks to the same level obtained for the control amplicon (Figure 1).
For each singleton, we computed the probability to observe no CNV in the control sample given that one CNV was observed in cases. If we assume that cases and controls were random samples from a same population in which the CNV is present at frequency q, then the number Xc of CNV alleles expected in the control sample follows a binomial distribution B(2Nc, q), with Nc being the number of controls. The probability to observe 0 CNV allele is then P0=P (Xc=0)=(1−q)2Nc. An estimate of q with the corresponding 95% confidence interval (CI) can be obtained from the sample of Nd cases: E(q)=1/2Nd, with Nd being the number of cases and V(q)=(1/2Nd)*(1−(1/2Nd))/2Nd.
On average, 45 CNVs (27 deletions and 18 duplications) were detected per individual. Their size ranged from 2.5 to 7.9Mb, with a median size of 29.8kb. CNVs present in the database of genomic variants (http://projects.tcag.ca/variation/), excluding BAC array-based studies, or CNVs present in nongenic or intronic regions were filtered out. At this stage, 18 CNVs (5 deletions and 13 duplications) were retained. These CNVs were then all confirmed by targeted QMPSF analyses, and their frequencies were determined in a sample of 1078 controls using QMPSF. CNVs present in the control sample were then excluded. At the completion of this analysis, four singleton CNVs, absent in controls, were retained in the ADEOAD group and three in the sporadic cases group (Table 1). For each CNV observed once among the 21 ADEOAD cases (q=0.024; 95% CI=(0.001; 0.123)), the probability to have not observed it in the 1078 controls is 2.73 × 10−23 and does not exceed 0.07 at the lower bound of the CI. For each CNV observed once among the 12 sporadic cases (q=0.042, 95% CI=90.002; 0.202)), the probability is 1.41 × 10−40 and 0.01 at the lower bound of the CI.
All retained CNVs were duplications, with a size ranging from 15 to 362kb (median size of 103kb). In index cases of ADEOAD families, we observed (i) a duplication including NLRP8, as well as a large part of the NLRP5 gene, (ii) a partial duplication of MAGI1, (iii) a duplication of KLK6, and (iv) a duplication located upstream of the MEOX2 gene and including exon 1 of the gene. In sporadic early-onset AD cases, we observed (v) a partial duplication of SVOP and USP30 in a patient with onset at age 45, (vi) a large duplication including part of the HAS1 gene, as well as the FPR1, FPR2, and FPR3 genes, in a patient with onset at age 47 and (vii) a duplication including SLC30A3, DNAJC5G, and part of the TRIM54 gene, in a patient with onset at age 48 (Figure 1). QMPSF analysis showed that these 7 CNVs were not present in the 912 other AD cases. The only positive QMPSF signal detected in these patients corresponded to the gain of one copy of an NLRP8 amplicon in a subject with onset at age 80 years. Subsequent a-CGH analysis revealed that this patient bore a small duplication of NLRP8 exons 1–3. This rearrangement, which was reported in the database of genomic variants, was thus only partially overlapping with that found in the ADEOAD case (Supplementary Figure S2). More generally, note that other partially overlapping rearrangements, different from those found in our patients and including short intronic indels, have been reported at some of these loci in the Database of Genomic variants (Supplementary Table S2).
Three CNVs contained genes whose relationship with AD pathology, if any, remains elusive. MAGI1 (3p14.1) and SVOP (12q24.11) encode synaptic proteins currently without any known link to AD pathological processes. USP30 (12q24.11) encodes a deubiquitinating enzyme present in the mitochondrial outer membrane. NLRP8 (19q13.42) encodes a member of the cytosolic Nod-like receptors that have a role in regulation of innate and adaptative immunity. Although the exact functions and signalling pathways of NLRP8 remain obscure, it should be noted that several NLR proteins are components of inflammasome, a multiprotein complex, which is activated in AD brains by Aβ oligomers.10 Thus, a possible link between NLRP8, inflammation, and AD can be envisioned.
The four remaining CNVs also contain several genes with currently no known relevance to AD pathophysiology: HAS1 (19q13.33) encodes a hyaluronan synthase involved in the biosynthesis of glycosaminoglycans of extracellular matrix; DNAJC5G (2p23.3) is an uncharacterized homologue of the hsp40 chaperon; and TRIM54 (2p23.3) encodes a muscle-specific ring finger protein not expressed in the brain. Besides these genes, however, four other genes included in these CNVs encode proteins whose function is firmly related to the Aβ biological network:
In complex diseases, focusing on subgroups of patients with an extreme phenotype is a useful strategy in revealing rare genetic causal factors.19 Using this approach, we have identified seven CNVs, exclusive to ADEOAD or sporadic cases with APOE 33 genotype and disease onset <55 years. None of these CNVs were retrieved in other AD patients, demonstrating a strong specificity for these extreme phenotypes. For each CNV, sequence analysis of the boundary regions did not reveal any interstitial duplication or Alu sequences, suggesting that these rearrangements are not mediated by a non-homologous allelic rearrangement mechanism, which is consistent with their lack of recurrence.
We cannot exclude that the CNVs found exclusively in our two selected disease groups correspond to private non-pathogenic variations. However, arguing strongly against this interpretation is the fact that among the seven CNVs retained at the final stage from this large genome scan, four included genes involved in Aβ-related biological pathways and a fifth was possibly involved in the same network. If we consider these CNVs as ultra low-frequency benign variants, it seems highly unlikely that at least four of the seven haphazardly impact genes strongly related to Aβ metabolism or signalling. Three of these genes (KLK6, SLC30A3, and FPR2) are entirely duplicated. The direction of the observed gene dosage alteration is consistent with overexpression of the protein and is in agreement with the deleterious effect predicted from animal or cellular models. The MEOX2 gene is affected by a partial duplication at the 5′-end of the gene. Although the functional consequence of this kind of rearrangement is more difficult to predict, it can produce non-functional fusion transcripts or alter gene expression by disrupting regulatory elements located in this region, in accordance with the animal model, predicting a deleterious loss of function.13
In conclusion, focusing on rare DNA alterations found exclusively in highly selected populations of AD patients, we have shown that a set of genes coding for molecules long suspected of modifying Aβ metabolism (and for which animal or cellular models had already been developed), most likely has a role in AD pathophysiology. These results provide novel support for the amyloid cascade hypothesis.1 However, it should be stressed that definitive proof of the causal involvement of these genes requires segregation analysis, which was not possible in our families, or identification of similar alterations in other cases with these extreme phenotypes. Nevertheless, particular attention should be devoted to these genes in forthcoming next-generation sequencing studies.
We are indebted to the banque d'ADN et de cellules Pitié Salpêtrière. We thank Dr Mario Tosi and Tracey Avequin for critical reading of the manuscript and Emmanuelle Genin for statistical support. This study was funded by PHRC GMAJ 2008/067 (Rouen University Hospital).
The investigators of the French GMAJ project include Didier Hannequin, Dominique Campion, Olivier Martinaud, Lucie Guyant-Maréchal and David Wallon (Centre Hospitalo Universitaire (CHU), Rouen); Olivier Godefroy and Candice Picard (CHU Amiens); Frédérique Etcharry-Bouyx (CHU Angers); Eric Berger (CHU Besancon); Jean-Francois Dartigues and Sophie Auriacombe (CHU Bordeaux); Vincent de la Sayette (CHU Caen); Francois Sellal (CH Colmar); Olivier Rouaud and Christelle Thauvin (CHU Dijon); Olivier Moreaud (CHU Grenoble); Stéphanie Bombois, Adeline Rollin-Sillaire, Marie-Anne Mackowiak and Florence Pasquier (CHU Lille); Isabelle Roullet-Solignac and Alain Vighetto (CHU Lyon); Mira Didic, Olivier Félician and Mathieu Ceccaldi (CHU Marseille); Audrey Gabelle and Jacques Touchon (CHU Montpellier); Martine Vercelletto and Claire Boutoleau-Bretonnière (CHU Nantes); Pierre Labauge and Giovanni Castelnovo (CHU Nimes); Claire Paquet and Jacques Hugon (CHU Lariboisière); Agnès Michon, Isabelle Le Ber and Bruno Dubois (CHU La Salpêtrière, Paris); Catherine Thomas-Antérion (CHU Saint-Etienne); Frédéric Blanc and Christine Tranchant (CHU Strasbourg); Jérémie Pariente, Michèle Puel and Jean-Francois Demonet (CHU Toulouse); Caroline Hommet and Karl Mondon (CHU Tours); Hélène Mollion and Bernard Croisile (CMRR CHU Lyon); Mathilde Sauvée (CHU Nancy); Gaelle Godenèche and Foucauld De Boisgueheneuc (CHU Poitiers).
The authors declare no conflict of interest.
Supplementary Information accompanies the paper on European Journal of Human Genetics website (http://www.nature.com/ejhg)