Molecular (2006) 11, 37–46 & 2006 Nature Publishing Group All rights reserved 1359-4184/06 $30.00 www.nature.com/mp ORIGINAL ARTICLE A genome-wide search for alleles and haplotypes associated with autism and related pervasive developmental disorders on the Faroe Islands MB Lauritsen1, TD Als1, HA Dahl2, TJ Flint1, AG Wang3,4, M Vang4, TA Kruse2, H Ewald1,5 and O Mors1 1Centre for Basic Psychiatric Research, Psychiatric in Aarhus, Aarhus University Hospital, Risskov, ; 2Department of Clinical Biochemistry and Genetics, Odense University Hospital, Odense C, Denmark; 3Department of Psychiatry, Amager Hospital, University Hospital, Copenhagen S, Denmark and 4Department of Psychiatry, Landssju`krahusid (National Hospital), Torshavn, Faroe Islands

The involvement of genetic factors in the etiology of autism has been clearly established. We undertook a genome-wide search for regions containing susceptibility for autism in 12 subjects with childhood autism and related pervasive developmental disorders (PDDs) and 44 controls from the relatively isolated population of the Faroe Islands. In total, 601 microsatellite markers distributed throughout the with an average distance of 5.80 cM were genotyped, including 502 markers in the initial scan. The Faroese population structure and genetic relatedness of cases and controls were also evaluated. Based on a combined approach, including an assumption-free test as implemented in CLUMP, Fisher’s exact test for

specific alleles and haplotypes, and IBD0 probability calculations, we found association between autism and microsatellite markers in regions on 2q, 3p, 6q, 15q, 16p, and 18q. The

most significant finding was on 3p25.3 (PT1 ¼ 0.00003 and PT4 ¼ 0.00007), which was also supported by other genetic studies. Furthermore, no evidence of population substructure was found, and a higher degree of relatedness among cases could not be detected, decreasing the risk of inflated P-values. Our data suggest that markers in these regions are in linkage disequilibrium with genes involved in the etiology of autism, and we hypothesize susceptibility genes for autism and related PDDs to be localized within these regions. Molecular Psychiatry (2006) 11, 37–46. doi:10.1038/sj.mp.4001754; published online 4 October 2005 Keywords: genome-wide scan; association mapping; population isolate; haplotype sharing; susceptibility , autism

Introduction abnormalities, candidate regions of autism have been suggested on chromosome 2q,6–8 Childhood autism is a pervasive neurodevelopmental 7q,9,10 and 15q.11,12 However, no susceptibility genes disorder with onset of symptoms before the age of 3 have yet been identified, and except for a few specific years. The symptoms include severe impairments of etiological factors like the fragile X syndrome, communicative and social skills, together with stereo- 1 tuberous sclerosis, and maternally inherited duplica- typed and repetitive behavior. Related disorders tion of 15q11–13, the etiology of autism is largely falling short of strict diagnostic criteria for childhood unknown.13 autism comprise Asperger’s syndrome, atypical aut- The use of isolated populations is a powerful ism, and pervasive developmental disorder. Based on approach in the search for disease genes and has family and twin studies there is strong evidence of a 2,3 facilitated the identification of candidate chromo- genetic component. The mode of inheritance is some regions involved in monogenic diseases14 and in unknown, but a polygenic model has been suggested some complex diseases, for example, uric acid with estimates of the number of genes involved, 15 4,5 nephrolithiasis. Isolated populations have also been ranging from three to more than 15 genes. Based on used to study susceptibility genes for autism-spec- genome-wide scans, screening of possible candidate trum disorders in Finland, and evidence for a genes, and studies of association between autism and susceptibility locus was found on chromosome 3q25–27,16 in addition to evidence for risk loci for Correspondence: Dr MB Lauritsen, Centre for Basic Psychiatric Asperger’s syndrome on 1q21–22, 3p14–24, and Research, Aarhus University Hospital, Skovagervej 2, DK-8240 13q31–33.17 Risskov, Denmark. E-mail: [email protected] 5In respectful memory of Professor Henrik Ewald 1958–2004. We used the isolated population of the Faroe Received 9 November 2004; revised 22 August 2005; accepted 23 Islands to search for susceptibility genes of autism. August 2005; published online 4 October 2005 The Faroe Islands are situated in the North Atlantic Genome-wide scan for autism susceptibility loci using LD-mapping MB Lauritsen et al 38 Ocean between Norway, Scotland, and Iceland. Based nostic Interview-Revised (ADI-R).27 The observation on historical records the establishment of the popula- (ADOS-G) was videotaped, and the interview (ADI-R) tion was presumably caused by emigration mainly was recorded on audio-tape. ADOS-G and ADI-R are from Bergen and the surrounding area in Norway based on the International Classification of Diseases, approximately 900 AD, but people from the British 10th edition (ICD-10)1 and the Diagnostic and Statis- Isles may also have contributed to the colonization of tical Manual of Mental Disorders, Fourth Edition the islands.18 Based on a genetic study of paternally (DSM-IV).28 The final diagnosis based on ADI-R and inherited Y- from the Faroese popula- ADOS-G was made in collaboration with an indepen- tion, it is likely that the Faroe Islands were colonized dent expert, Lennart Pedersen, National Centre for by males from Norway, Sweden, and Scotland.18 The Autism, Copenhagen, Denmark. Tests of the cognitive population has expanded from a few thousand people level of cases were not performed. to the present level of approximately 45 000 inhabi- In total, 12 individuals were included in the study. tants, covering periods of constant population size as All cases were males, and the mean age of the well as putative periods of population bottlenecks probands at the time of assessment was 17 years with less than 4000 inhabitants due to severe (age range, 4–41 years). Controls were randomly epidemics.19 Migration has been sparse,18 and pre- ascertained and consisted of 22 pairs of parents (44 sumably the expansion has mainly been due to presumably unrelated adults). Controls were without increased reproduction rates. The long distances a history of psychiatric disorders at the time of between the villages, commercial monopoly, and inclusion in the study. legislation, imposed further isolation between differ- The diagnoses of the 12 cases are indicated in the ent geographical areas on the Faroe Islands.19 genealogy shown in Figure 1. Seven cases fulfilled the Increased occurrence of diseases caused by specific criteria for childhood autism, whereas one case mutations, like cystic fibrosis and glycogen fulfilled the criteria for atypical autism because the storage disease IIIA, has been reported on the Faroe specific impairments in areas of social interaction, Islands,20,21 and a locus for primary ciliary dyskinesia communication, and behavior were not present before has been identified in Faroese families.22 The Faroese the age of 3 years. Another case fulfilled the criteria population has, in addition, been used to localize for Asperger’s syndrome, and three autistic cases were interesting chromosome regions for bipolar affective severely mentally retarded. Three diagnostic cate- disorder and schizophrenia.23,24 These findings in- gories were used to define the relevant phenotype of dicate that the population on the Faroe Islands may be autism. The core category includes the seven cases suitable for identification of gene mutations. Fewer, diagnosed with childhood autism, whereas the but more frequent risk loci and alleles presumably narrow category also includes the two cases diag- exist at the population level, compared to an outbred nosed with Asperger’s syndrome and atypical autism. population, due to the action of random genetic drift The broad category comprises all 12 cases including caused by a founder event followed by putative the three severely mentally retarded subjects whose bottlenecks in the size of the Faroese population. cognitive level makes a diagnosis of autism uncertain, Furthermore, this recent founding, isolation, and using the diagnostic instruments. relatively rapid expansion of the Faroese population is reflected in the extensive background linkage Genotyping disequilibrium (LD) observed in the population.25 DNA was prepared from whole blood using a standard In the present study, distantly related individuals sucrose/Triton-lysis protocol with sodium chloride/ with childhood autism and related pervasive devel- isopropanol precipitation. DNA amplification was opmental disorders (PDDs) from the Faroe Islands performed either as single or multiplex PCR with a were genotyped for polymorphic markers distributed maximum of three markers simultaneously. DNA throughout the human genome in the search for allele fragments were subsequently analyzed on an ABI and haplotype association. Association between Prism 310 Genetic Analyzer or an ABI Prism 3100 specific markers and the disorder suggests that the Genetic Analyzer (Applied Biosystems, Foster City, markers are in LD with, and therefore presumably lie CA, USA). close to a susceptibility gene, and may thus contribute Haplotypes for chromosomal segments consisting to the localization of candidate regions for autism. of two and three neighboring markers were deter- mined for cases based on parental genotypes and for controls based on the genotypes of one of their Materials and methods offspring assuming that no recombination occurred Sample in this generation. This is only valid when markers Individuals with autism and related disorders were are in relatively close proximity because the chance of recruited through the National Society for Autism on meiotic crossover from parent to offspring increases the Faroe Islands without prior knowledge of the with increasing distance between markers. This genealogical relationship between cases. Diagnostic method will reconstruct the majority of relatively assessment was performed by one of the authors (MB short haplotypes correctly.23,29 Lauritsen), using the Autism Diagnostic Observation The present scan was performed in two stages. Schedule Generic (ADOS-G)26 and the Autism Diag- First, a genome-wide scan was made, using 502

Molecular Psychiatry Genome-wide scan for autism susceptibility loci using LD-mapping MB Lauritsen et al 39 Childhood autism Atypical autism Asperger's syndrome Mental retardation

BCHG AJ K D F IE L Figure 1 The genealogy of cases. Individual diagnoses are indicated by different shadings of cases. Owing to severe mental retardation in three of the cases, a diagnosis of autism using ADI and ADOS was difficult to apply.

randomly chosen and more or less evenly distributed common ancestor 7.2 generations back in time. This is markers (average distance 6.97 cM, range 0–16.27 cM) assumed to ensure a reasonable size of the shared and a P-value of 0.01 as threshold for interesting chromosomal region surrounding a putative suscept- regions. In the second scan 30 additional microsatel- ibility gene. lites were genotyped and tested around the original When using a case–control design in association 14 interesting markers. As the present scan is part of a studies, it is essential that cases and controls are larger study, also containing samples of bipolar subsamples of the same larger population and share affective disorder and schizophrenia, additional the same genetic history. While the sample of controls markers located around interesting markers for these is essentially randomly chosen with respect to LD, a two disorders were also included in the second scan sample of affected individuals in a population isolate in order to increase the marker density. Specific is, by definition, chosen to be enriched for LD in the additional markers were chosen based on the level of region of disease-related variants, and therefore also heterozygosity in the CEPH database (http:// tends to share alleles of surrounding markers. In such www.cephb.fr/cgi-bin/wdb/ceph/systeme/form) in a scenario cases tend to be more related in general addition to having a position 1–2 cM distal and than the sample of controls, causing an increased proximal to interesting original markers. In total, false-positive error rate and inflated P-values. Instead 601 microsatellite markers with an average distance of of detecting true association between a marker and a 5.80 cM (range 0–16.27 cM) were used to cover the trait, classical case–control analysis might therefore whole genome. The order of the microsatellites was detect differences between cases and controls owing based on the public available database (http://geno- to ignored population substructure or improperly me.cse.ucsc.edu), and linkage map distances, in accounted relatedness among individuals.30,31 In Kosambi centimorgans, were obtained from the order to evaluate whether the population on the Marshfield Center for Medical Genetics (http://re- Faroe Islands is suitable for identifying candidate search.marshfieldclinic.org). genes of autism, we investigated population substruc- ture and the degree of relatedness among individuals. The Faroese population Estimates of genetic relatedness (r) were calculated On the Faroe Islands church records have been kept using the algorithms developed by Queller and since 1727 or 1750 depending on the region, whereas Goodnight32 based on 60 randomly selected unlinked incomplete church records and civic records have markers as implemented in SPAGEDi 1.2.33 Markers been kept for even longer. Using church and civic were chosen at an intermarker distance of 50 cM or records of births, marriages, and deaths, the genealo- more. Pairwise estimates of r can be interpreted as a gical relationship between cases was investigated, measure of similarity between individuals, and it is whereas the genealogical information for controls is used to investigate whether cases were genetically currently unavailable. The average pairwise number more similar than controls. Estimates of pairwise of generations relating two patients to a common genetic distances between individuals were obtained ancestor in the genealogically shortest possible way using the algorithm developed by Rousset34 as through one of their parents was estimated (see Figure implemented in SPAGEDi 1.2. Pairwise genetic dis- 1), and pairs of the 12 individuals with childhood tances were used in a multidimensional scaling autism and related PDDs were on average related to a algorithm (Alscal procedure – as implemented in

Molecular Psychiatry Genome-wide scan for autism susceptibility loci using LD-mapping MB Lauritsen et al 40 SPSS 11.5) to map similarity between individuals be seen for some of the alleles or haplotypes. Allele relative to each other, that is, to inspect potential and haplotype frequencies among cases and controls clustering of cases relative to controls. were compared, either as single markers or as two- or The population structure, including the genetic three-marker segments. CLUMP generates four test differentiation within the case-control sample, was statistics. We chose to report the empirical P-value evaluated by Wright’s F-statistics and Bayesian from the T1 and T4 test statistics as these tests are model-based clustering using multilocus genotype more powerful for evaluating association than the data. The 60 unlinked markers were used to calculate other two test statistics.38 T1 is a w2 test of the raw the genetic distance between the case and control 2 Â N table that evaluates the significance of associa-

group using Wright’s FST. This measure is based on a tion with alleles/haplotypes. For T4 the columns of hierarchical analysis, where the total variance is the raw 2 Â N table are clumped together in a new partitioned into covariance components due to differ- 2 Â 2 table, and each allele/haplotype is reshuffled ences in allele frequencies between populations, subsequently in order to maximize the w2 value. T4 between individuals in populations, and between evaluates the situation where more than one rare alleles in individuals.35 Under the hypothesis of no allele/haplotype is responsible for the deviation. For differentiation between the individuals and popula- both tests, we performed 100 000 simulations in order

tions, a null distribution of FIT, FIS, and FST values was to generate a null distribution of the test statistics and obtained performing 3000 permutations of individual evaluate empirical significance.

genotypes among all individuals (FIT), among indivi- Subsequently, for markers or segments yielding low duals within populations (FIS), and among popula- P-values (Pp0.01), alleles or haplotypes with a high 33 tions (FST) as implemented in SPAGEDi 1.2. frequency among cases compared to controls were In addition, Bayesian model-based clustering using evaluated using Fisher’s two-tailed exact test. Fisher’s multilocus genotype data, as implemented in STRUC- exact test is applied when very few observations are TURE 2.1,36 was applied in order to infer any seen in each cell of a contingency table as is the case potential cryptic population substructure in the data in our study. It is used to evaluate the significance of set. An admixture linkage model was applied without the differences in frequency of the specific allele or using any prior information on population structure. haplotype between cases and controls. In addition to taking into account that individuals Relatives share, on average, relatively large chro- may have mixed ancestry, and that different parts of mosomal segments. Distantly related affected indivi- their genome may actually belong to different clusters duals may, therefore, share haplotypes that are

of the data set, the admixture linkage model also identical by descent by chance (IBD0) from a common considers the possibility of correlations between ancestor without necessarily containing a shared markers due to linkage (‘admixture linkage disequili- disease gene. If affected individuals are more closely brium’).37 This method does not attempt to model the related than the controls in a case–control design, this LD that occurs within populations between very close may result in inflated P-values. In order to determine markers, and markers on the autosomes were ex- whether a potential association between autism and cluded in such a way that all intermarker distances related PDDs and single- or two-marker segments is

were above 1 cM, leaving a total of 529 markers in the due to IBD0, the probability that affected individuals data set. In the analyses partial phase information was share a haplotype that is IBD0 from a known ancestor used. A burning length of 100 000 simulations, through a specific genealogical relationship was followed by 1 000 000 simulations was used to get calculated using the formulas derived by Houwen accurate parameter estimates. Five iterations were et al.39 and Durham & Feingold.40 In these formulas, performed for each of the following number of the average pairwise relationship with a common putative clusters/subpopulations KA[1;3]. STRUC- ancestor of the 12 individuals with autism and related

TURE thus identifies clusters of related individuals PDDs were used. P0 is the probability of the observed within the sample from multilocus genotypes, and number of cases having the most frequent haplotype

should also be able to indicate whether cases are IBD0 from a known ancestor (i.e. false positive 39 indeed more related to each other than to controls. probability) according to Houwen et al. P1 is the The analyses supplement the estimates of average same probability summed over all segments typed on

relatedness and pairwise individual genetic distances the chromosome. P2 is the genome wide probability of in suggesting whether cases and controls fall into the IBD0 according to the method by Durham & Fein- same genealogy or not. gold.40 This was done for segments with a low CLUMP P-value (Pp0.01) for haplotypes with a The case–control study higher frequency among cases compared to controls. Test for association between autism and related PDDs and any of the markers was performed using Monte Results Carlo based tests as implemented in the computer program CLUMP38 (http://www.mds.qmw.ac.uk/stat- In the initial CLUMP analyses using 502 markers and gen/dcurtis). CLUMP is particularly useful for highly the broader category of cases, nine chromosomal polymorphic markers, especially when the sample regions (3p, 4q, 6q, 10p, 10q, 14q, 15q, 17q, and size is small because only few observations may then Xq) showed interesting findings (Pp0.01). When

Molecular Psychiatry Genome-wide scan for autism susceptibility loci using LD-mapping MB Lauritsen et al

À5 À4 41 additional markers were added in the follow-up likely (P0 ¼ 4.0 Â 10 , P1 ¼ 5.4 Â 10 , and P2 ¼ 0.051) analyses, diminished evidence of association was considering the specified genealogy. found in two regions (4q and Xq) whereas evidence of The association analyses as implemented in association was either maintained or increased in the CLUMP, using all 601 genetic markers, were also remaining regions. In addition, interesting regions performed for a subsample of seven cases with a also appeared on 2q, 5q, 8p, 11p, 12q, 16p, 18q, and diagnosis of childhood autism and in a subsample 19q in the follow-up analyses (Table 1). Only analyses excluding the three severely mentally retarded cases. of single- and two-marker segments are shown in However, the results only differed slightly from the Table 1 because only very few interesting three- results of the total sample including 12 cases. marker segments were found, and they are mentioned Average relatedness among cases using Queller and in the text. Goodnight32 did not differ significantly from average

When comparing the distribution of alleles or relatedness of case–control pairs (rcaseÀcase ¼ haplotypes between cases and controls in these inter- À0.025370.0111 vs rcaseÀcontrol ¼À0.022670.0045, esting regions using Fisher’s exact test and the same two-tailed parametric t-test, t12,56 ¼ 0.2459, level of significance, seven chromosome regions (3p, 5q, P ¼ 0.8065), indicating that cases are not more 6q, 12q, 14q, 15q, and 18q) appeared to have markers or related with each other than they are with controls. segments with significant deviation in frequency of a Likewise, average relatedness among controls did not single allele or haplotype. In all cases except on differ significantly from relatedness of case-control chromosome 6q, the P-values of Fisher’s exact test were pairs (rcontrolÀcontrol ¼À0.016470.0029 vs rcaseÀcontrol ¼ lower than the P-values obtained by CLUMP. À0.022670.0045, two-tailed parametric t-test, t44,56 ¼ The largest deviation in frequency of a single allele 1.0889, P ¼ 0.2789). Furthermore, the average within among cases and controls was found on chromosome group relatedness did not differ between cases and

12q for the 239-allele of marker D12S395 (Table 1). controls (rcaseÀcase ¼À0.025370.0111 vs rcontrolÀcontrol ¼ 7 For marker D5S1501 on chromosome 5q and for À0.0164 0.0029, t12,44 ¼ 0.7758, P ¼ 0.4396). Overall marker D3S3594 on chromosome 3p we found a large, within group estimated relatedness did not differ sig- statistically significant, deviation in frequency of nificantly from the average between group relatedness allele 98 and allele 271, respectively, among cases estimate (rwithin group ¼À0.016970.0025 vs rcaseÀcontrol ¼ and controls. Furthermore, a haplotype (271–2) of the À0.022670.0045, t56,56 ¼ 1.1073, P ¼ 0.2706). Multi- two-marker segments D3S3594–D3S3589 at 32.36 cM dimensional scaling of Rousset’s pairwise genetic was close to significantly more frequent among cases distances34 between individuals did not reveal any than controls, and the probability (P0 and P1) that six overall clustering of cases in relation to controls out of 22 haplotypes are shared IBD0 in cases related (results not shown). 7.2 generations ago is as small as 2.7  10À8 and Running STRUCTURE for different numbers of 1.7  10À7, respectively, according to the formula of assumed K clusters indicated that a single cluster 39 À5 Houwen and P2 ¼ 3.7  10 according to Durham (K ¼ 1) was more likely compared to multiple clusters and Feingold.40 The frequency of allele 206 of marker (K ¼ 2,3) given the current data set (Table 2). Applying D15S198 on chromosome 15q also appeared to be a model without admixture gave similar results significantly different among cases compared to (results not shown). Most individuals had the controls. A single haplotype (206–7) of the two- majority of their genome assigned to the same single marker segment D15S198–D15S643, spanning 51.21– cluster, except for two controls who had the majority 52.33 cM, proved to have significantly higher fre- of their genome assigned to another cluster. These two quency among cases compared to controls, and IBD0 individuals also appeared to have a high pairwise calculations showed that IBD0 is less likely. Haplo- relatedness and a short genetic distance (results not type 162–360 of D18S536–D18S970 on chromosome shown). Therefore, the presence of any cryptic 18 occurred on six out of 23 case chromosomes and population substructure could not be clearly demon- on two out of 73 control chromosomes, and calcula- strated. tion of IBD0 probabilities suggests that IBD0 is less In accordance with this, the amount of genetic likely considering the specified genealogy. At marker differentiation (FST) between cases and controls was D14S53 on chromosome 14q and at chromosome 6q at not statistically significant, when using the method marker D6S1652 a single allele was significantly more implemented in SPAGEDi 1.2 using 60 unlinked prevalent in case chromosomes than in control markers (FST ¼ 0.0019, PtwoÀtailed ¼ 0.5262, 3000 per- chromosomes. mutations). Likewise, there was no evidence for One three-marker segment spanning 172.27– inbreeding within individuals neither relative to the

181.87 cM on chromosome 3 (D3S2404-D3S1763- total sample nor relative to subgroups (FIT ¼ 0.0063, D3S3053) appeared to be associated with autism and PtwoÀtailed ¼ 0.5295, FIS ¼ 0.0031, PtwoÀtailed ¼ 0.7241, related disorders (CLUMP PT1 ¼ 0.0065 and 3000 permutations, when cases and controls are PT4 ¼ 0.0223). Furthermore, a single haplotype (1– considered as two subpopulations). There is no sign 272–238) of this three-marker segment proved to have of genetic differentiation among the different sub- significantly higher frequency among cases compared samples, and cases and controls are assumed to be to controls (P ¼ 0.0014), and calculation of IBD0 part of the same panmictic population and share the probabilities suggests that IBD0 by chance is less same genetic history.

Molecular Psychiatry oeua Psychiatry Molecular 42

Table 1 Empirical CLUMP P-values (T1 and T4) for single markers and two-loci segments, Fisher’s exact test for specific alleles or haplotypes, and probabilities of shared haplotypes to be IBD0 by chance in a sample of 12 autistic cases

a b

Region CLUMP Allele/haplotype LD-mapping using loci susceptibility autism for scan Genome-wide

Band Position Single marker Two-marker segment Fisher’s exact test IBD0 by chance

c,d c,d e e cM PT1 PT4 PT1 PT4 Cases Controls PP0 P1 P2

2q31.1 175.91 D2S2381 0.00525 0.01256 — — — 2 3/22 1/86 0.026 — — — 3p25.3 31.13–32.36 D3S3594 0.00007 0.00003 D3S3611–D3S3594 0.00375 0.00730 2–271 2/21 0/86 0.037 0.030 0.583 3.910 271 11/22 10/82 o0.001 — — — 32.36 ———D3S3594–D3S3589 0.00043 0.00046 271–2 6/22 6/85 0.016 2.69*10À 8 1.7*10À 7 3.7*10À 5 5q14.1 85.25 D5S1501 0.00451 0.00477 — — — 98 11/20 16/86 0.003 — — — 6q14.3 92.85 D6S1652 0.01206 0.00539 — — — 9 5/22 3/84 0.009 — — — À 4

6q15 99.01 ———D6S1652–D6S1570 0.01186 0.00521 9–1 3/24 2/87 0.067 6.0*10 0.030 1.249 Lauritsen MB 6q23.3 137.74 D6S1009 0.00741 0.01469 D6S1040–D6S1009 0.29101 0.20938 241 11/24 26/80 0.237 — — — 8p21.2 50.05 D8S1989 0.00428 0.00630 D8S136–D8S1989 0.00461 0.00544 8 3/24 2/88 0.065 — — — 10p11.23–10q11.21 59.03–66.50 ———D10S1426–D10S1669 0.00421 0.00411 168–215 4/23 2/84 0.019 7.6*10À 6 0.002 0.061 11.p15.4 17.19 D11S1999 0.00962 0.01354 — — — 121 7/24 7/56 0.107 — — — al et 11p15.3 17.19–18.26 ———D11S1999–D11S1349 0.00422 0.00277 117–1 3/24 1/78 0.040 0.002 0.015 1.249 113–5 2/24 1/83 0.126 0.040 0.437 3.968 121–4 2/24 0/87 0.045 0.040 0.437 3.968 12q24.23 136.82 D12S395 0.00410 0.00124 — — — 239 11/16 14/80 o0.001 — — — 14q24.3 86.29 D14S53 0.00218 0.00911 — — — 144 4/20 0/78 0.001 — — — 15q14 32.58 D15S1042 0.00917 0.04077 — — — 2 8/24 20/86 0.426 — — — 8 3/24 3/86 0.117 — — — 15q21.3 50.66–51.21 D15S198 0.00529 0.00396 D15S998–D15S198 0.00276 0.00702 2–206 6/24 6/88 0.020 3.8*10À 8 2.8*10À 7 6.6*10À 5 206 9/24 7/88 o0.001 — — — ———D15S198–D15S643 0.03793 0.05353 206–7 3/23 0/88 0.008 0.002 0.020 1.126 16p13.3 22.65 D16S748 0.00117 0.00465 — — — 187 6/20 9/82 0.071 — — — 17q23.2 80.38 D17S1606 0.00786 0.02225 — — — 3 4/24 3/86 0.039 — — — 17q24.3–17q25.1 93.27–97.6 ———GATA31B11–D17S1862 0.00043 0.00061 255–7 3/23 4/82 0.176 8.0*10À 4 0.026 1.126 255–8 3/24 1/82 0.036 9.1*10À 4 0.029 1.249 18q12.1–18q21.1 62.29–68.3 ———D18S536–D18S970 0.0022 0.00141 162–360 6/23 2/73 0.002 2.6*10À 9 3.1*10À 8 5.0*10À 5 19q13.42 88.85–92.56 ———D19S572–D19S418 0.00167 0.00236 2–87 3/24 2/84 0.072 1.0*10À 3 0.021 1.249 4–85 3/24 4/87 0.171 1.0*10À 3 0.021 1.249

aOnly markers with at least one P-value below 0.01 obtained by CLUMP are shown. bMost frequent allele/haplotype among cases. cFrequency of most frequent allele/haplotype. dNote that the total number of alleles or haplotypes in Fisher’s exact test may vary owing to phase uncertainties and missing genotypes. e Owing to the conservative way P1 and P2 are calculated, they may exceed a value of one when the observed value of the haplotype of interest is small. Bold values refer to chromosomal regions, markers, alleles and haplotypes. Genome-wide scan for autism susceptibility loci using LD-mapping MB Lauritsen et al 43 Table 2 Results from STRUCTURE analysis alities.45,48–51 Association between autism and tuber- ous sclerosis has been established,52 and the gene for K lnPr(X|K) Pr(K|X) tuberous sclerosis 2 (TSC2) located in this region53 has been found to be associated with autism.54 Our 1 À89370 B1 data also provide further evidence for a susceptibility 2 À89883 B0 B gene of autism and related disorders on 2q at 176 cM. 3 À94916 0 This region has previously been identified as a candidate region for autism based on genome-wide linkage analysis6,7,55 although other studies point Discussion more distally in the 210 cM region.41,45 As suggested by several genome-wide scans,9 we In contrast to pedigree-based genome-wide scans, we found some support for the localization of a suscept- applied a LD-based approach, using singletons, to ibility gene for autism on 7q (121–129 cM) in the map susceptibility genes of autism. The interpretation follow-up scan for different lengths of segments for all of our study therefore expands to include all affected three data sets (consisting of seven, nine, and all 12 autistic individuals and not only familial cases. The cases) with CLUMP P-values ranging from 0.0016 to present study revealed association for some of the 0.0159. regions previously identified as potentially contain- Based on the size of the Faroese population (45 000 ing candidate genes for autism. inhabitants) and a prevalence of autism of 0.1–0.15%, The most significant finding was at marker we would expect to be able to find more than 40 cases. D3S3594 located 32.36 cM from pter. Shao et al.41 Our random sample of cases included males only, and found LOD scores up to 2.02 for the markers D3S3680 the ascertainment method may thus be biased. The and D3S1259 in the region at 36 cM. At approximately lack of female cases could be due to different clinical 25 cM McCauley et al.42 found a nonparametric allele- presentation56 or that autism is under-recognized in sharing LOD score of 2.22 for marker D3S3691. Faroese females for cultural reasons, or due to early Marker D3S1259 and D3S3691 appeared, however, environmental influences.57 Alternatively, the male– not to be associated with autism in our study. Linkage female ratio could be a chance finding owing to the analyses of complex disorders often suggest relatively limited number of cases included. large chromosome regions of interest,43 which implies Although the results for the subsamples differed that cautiousness should be taken when comparing only slightly from the analyses of the total sample, not with results from more fine-scaled LD-based ap- all cases may be true cases of autism, in particular the proaches like the present study. Chromosome three cases with severe mental retardation. We used 15q11–13 has been suggested to contain a suscept- ADI and ADOS, but it is generally accepted that ADI ibility gene for autism based on several reports of should not be applied to individuals with a mental chromosome abnormalities.11 More distally, at age less than 18 months due to the difficulty in 51.21 cM, we found association with marker distinguishing autism-specific deficits from the gen- D15S198 and for the two-marker segment D15S198- eral absence of directed behaviors in individuals with D15S643, and Shao et al.41 identified a region from developmental levels less than 2 years of age.58 De 43.5 to 62.4 cM. On chromosome 6q marker D6S1652 Bildt et al.,59 however, reported that ADI, and in (at 92.85 cM) appeared to be associated with autism particular ADOS, made satisfactory classifications of and related PDDs. The same region was identified by both childhood autism and PDD, compared to a Buxbaum et al.44 in a sample of autism-affected clinical DSM-IV classification.28 We therefore assume relative pairs with more severe obsessive-compulsive the cases to be diagnosed correctly, although the (OC) behavior with LOD scores from 1.36 to 2.61. Our inclusion of non-PDD cases cannot be completely sample was, however, not characterized by OC ruled out. behavior according to the categories defined to be Care is required when generalizing from findings in OC behavior by Buxbaum et al.44 On 18q, we found isolated populations to outbred populations. New- increased occurrence of a specific haplotype among man et al.60 found similar frequencies of common cases at the 62.29–68.3 cM region spanning markers alleles in both outbred and isolated populations, and D18S536 and D18S970, and nearly the same region it is likely that many of the common disease- was identified by Philippe et al.45 associated variants detected in isolated populations We also found significantly increased occurrence of will appear also to be associated with the disease in alleles and haplotypes among cases on 5q, 12q, and outbred populations. The advantages of using founder 14q, but no clear evidence of involvement of these populations in mapping genes for complex disorders regions in autism has previously been indicated. have been the subject of much debate,61,62 but they are On , 16, and 17 we found association considered to be very useful for several reasons. First with autism and related PDDs, but no single allele or of all, within the isolated population fewer suscept- haplotype supported this. Chromosome 16p has been ibility alleles are likely to exist, each perhaps with a hypothesized to contain a susceptibility gene for relatively larger effect at population level, since autism based on genome-wide scans, association random genetic drift during a founder event and past studies,6,42,45–47 and reports of chromosome abnorm- population bottlenecks may have eliminated some of

Molecular Psychiatry Genome-wide scan for autism susceptibility loci using LD-mapping MB Lauritsen et al 44 the susceptibility alleles present in the source er’s exact tests (Pp0.01) testing for frequency differ- population, thereby reducing the allele and locus ences of a specific allele or haplotype are generally heterogeneity at the population level. Furthermore, lower than the P-values obtained by CLUMP indicat- background LD may due to random genetic drift ing that the deviation in frequency between cases and extend over greater genetic distances in isolated controls may be caused by a single allele or populations, thereby reducing the sample size and haplotype. No correction for type I error was made, the number of markers needed to identify specific despite the large number of tests performed, but a genes.63,64 In addition, isolated populations may be relatively low level of significance (Pp0.01) was more suitable for association mapping because of the chosen. Since many of the markers are in close more homogeneous environmental background.63,65 proximity and in extensive LD,25 the association tests Conflicting evidence has, however, been reported performed cannot be considered independent, and regarding the actual degree of homogeneity of isolated the standard procedure for correction for type I error populations.64 A previous study on the Faroese will therefore be too conservative. There is no population25 indicated extensive LD among the 15 generally accepted method to make the appropriate markers surveyed. Furthermore, the allelic diversity correction, although simulation-based statistics are was smaller, although nonsignificant in the Faroese being explored.66 The significance of the results population compared to outbred populations. Despite obtained in this study is supported by using a lack of evidence for a strong population bottleneck on combined approach including an assumption-free the Faroe Islands, the population is considered test as implemented in CLUMP, Fisher’s exact test, 25 suitable for LD mapping of complex diseases. and IBD0 probability calculations. To avoid bias due to population stratification, it is The number of larger segments found to be essential that cases and controls are subsamples of the associated with autism was rather limited. This could same panmictic population and share a similar in part be explained by the small sample size that

genetic history. Based on Wright’s FST and model- decreases the chance of detecting larger segments. based clustering analyses, this essential assumption Furthermore, the average size of a region among seems valid for our sample of cases and controls. We related cases inherited IBD from a common ancestor found no convincing genetic evidence indicating that is directly proportional to the number of generations cases are actually more related to each other than they that has passed since the common ancestor. The are to controls. Although a few individual pair of number of generations to a common ancestor for a cases might be more related than the average, the pair of cases ranges from three to 12, with an average findings indicate that the control sample somehow of 7.2 generations for a single pair of cases. Based on falls into the same genealogy as shown for cases, and the number of generations to a common ancestor for the bias inflation of P-values introduced is limited. each pair of cases in the genealogy (Figure 1), a rough STRUCTURE using most of the information from the estimate of the expected size of a shared IBD segment current data set available including partial phase among a pair of cases is 14.96 cM (see Wright et al.67 information should potentially be able to identify and Ober and Cox68 for details). Assuming that all 12 clusters if cases are indeed more related to each other individuals had inherited a single copy of a suscept- than they are to controls, which was not the case. ibility locus from a common ancestor, the expected Considering the history of the Faroese population, it size of the surrounding segment shared among all 12 is not surprisingly that controls seem to fall into the cases would be further diminished, suggesting that same genealogy as cases (Figure 1). Going 12 genera- the size of the segment shared between two to 12 tions back in time, each individual has 212 ¼ 4096 cases would be within 1.25–14.96 cM depending on ancestors. The Faroese population size 12 generations the number of cases actually sharing the suscept- ago was somewhere around 4000 individuals. In total, ibility allele. The above calculations, however, are 56 randomly chosen contemporary individuals there- rough estimates and based on the assumption of fore probably have several common ancestors living recombination homogeneity throughout the genome 12 generations ago. which is unlikely.69 In addition, the markers used in The power of this study is currently unknown, and the current study are not evenly distributed through- exact power estimation is complicated by the com- out the genome, and we may have missed some of the plexity of the genealogical structure, the unknown segments shared among cases. mode of disease inheritance, and the unknown Our study does not provide any support for a single number of genes involved in disease etiology as well susceptibility locus significantly increasing the risk of as the magnitude of their effects. The use of micro- developing autism on the Faroe Islands. Instead satellite markers increases the informativeness com- several additional genetic loci, each with a small pared to SNPs, but decreases power due to the highly effect, most likely contribute to the risk of autism. polymorphic nature of microsatellites. This problem This study may not, however, have enough power to created by the marker type used is partially counter- detect disease genes with a weak effect. Despite the acted using CLUMP.38 CLUMP evaluates all alleles or limited number of cases, it was possible to detect haplotypes in one test, and low P-values may be some candidate regions of autism. The replication of caused by differences in frequencies of more than one findings on chromosome 2, 3, 6, 15, 16, and 18 across allele or haplotype. The significant P-values of Fish- different study populations is promising with respect

Molecular Psychiatry Genome-wide scan for autism susceptibility loci using LD-mapping MB Lauritsen et al 45 to the identification of susceptibility genes of autism 9 Badner JA, Gershon ES. Regional meta-analysis of published data in the future. The inclusion of additional cases and supports linkage of autism with markers on chromosome 7. Mol controls, and further analyses with more densely Psychiatry 2002; 7: 56–66. 10 Skaar DA, Shao Y, Haines JL, Stenger JE, Jaworski J, Martin ER et distributed markers in candidate regions, is relevant al. Analysis of the RELN gene as a genetic risk factor for autism. in order to further investigate the significance of the Mol Psychiatry 2005; 10: 563–571. present findings, and to eventually localize suscept- 11 Bolton PF, Dennis NR, Browne CE, Thomas NS, Veltman MW, ibility genes of autism. Thompson RJ et al. The phenotypic manifestations of interstitial duplications of proximal 15q with special reference to the autistic spectrum disorders. Am J Med Genet 2001; 105: 675–685. 12 Simic M, Turk J. Autistic spectrum disorder associated with Acknowledgments partial duplication of chromosome 15; three case reports. Eur Child Adolesc Psychiatry 2004; 13: 389–393. We thank The Faroese association for autism, and the 13 Muhle R, Trentacoste SV, Rapin I. The genetics of autism. late chairman Mrs ‘Margrethe Joensen-Næs for their 2004; 113: e472–e486. participation in this study. We also thank psycholo- 14 Peltonen L, Jalanko A, Varilo T. Molecular genetics of the Finnish gist Lennart Pedersen, National Centre for Autism, disease heritage. Hum Mol Genet 1999; 8: 1913–1923. 15 Gianfrancesco F, Esposito T, Ombra MN, Forabosco P, Maninchedda Copenhagen, for diagnostic assessments, statistician G, Fattorini M et al. Identification of a novel gene and a common Leslie Foldager, Centre for Basic Psychiatric Research, variant associated with uric acid nephrolithiasis in a Sardinian Aarhus University Hospital, Aarhus, for statistical genetic isolate. Am J Hum Genet 2003; 72: 1479–1491. advice, and research assistant Agata El Daoud for help 16 Auranen M, Varilo T, Alen R, Vanhala R, Ayers K, Kempas E et al. with the genotyping. In addition, we thank assistant Evidence for allelic association on chromosome 3q25–27 in families with autism spectrum disorders originating from a professors Thomas Bataillon and Roald Forsberg, The subisolate of Finland. Mol Psychiatry 2003; 8: 879–884. Bioinformatics Research Centre (BiRC), University of 17 Ylisaukko-Oja T, Nieminen-Von Wendt T, Kempas E, Sarenius S, Aarhus and associate professor Jes Søe Petersen, Varilo T, Wendt LL et al. Genome-wide scan for loci of Asperger Department of Population Biology, University of syndrome. Mol Psychiatry 2004; 9: 161–168. 18 Jorgensen TH, Buttenschon HN, Wang AG, Als TD, Borglum AD, Copenhagen, for discussion of population genetics Ewald H. The origin of the isolated population of the Faroe Islands issues. This work was supported by The Danish investigated using Y chromosomal markers. Hum Genet 2004; 115: Medical Research Council (Grant Number 9601887); 19–28. The Lundbeck Foundation (Grants Numbers 39/98 19 West JF. Faroe: The Emergence of a Nation. Hurst: London, 1972. and 59/99); Puljen til Styrkelse af Psykiatrisk For- 20 Schwartz M, Sorensen N, Brandt NJ, Hogdall E, Holm T. High incidence of cystic fibrosis on the Faroe Islands: a molecular and skning i Aarhus Amt; Psykiatrisk forsknings fond; genealogical study. Hum Genet 1995; 95: 703–706. Ministry of Education, Culture and Research, The 21 Santer R, Kinner M, Steuerwald U, Kjaergaard S, Skovby F, Faroese Government; Research Fund of the Faroese Simonsen H et al. Molecular genetic basis and prevalence of Savings Bank, Torshavn, Faroe Islands; Danish Hos- glycogen storage disease type IIIA in the Faroe Islands. Eur J Hum pital Foundation for Medical Research, Region of Genet 2001; 9: 388–391. 22 Jeganathan D, Chodhari R, Meeks M, Faeroe O, Smyth D, Nielsen K Copenhagen, the Faroe Islands and Greenland. et al. Loci for primary ciliary dyskinesia map to chromosome 16p12.1–12.2 and 15q13.1–15.1 in Faroe Islands and Israeli Druze genetic isolates. J Med Genet 2004; 41: 233–240. References 23 Degn B, Lundorf MD, Wang A, Vang M, Mors O, Kruse TA et al. Further evidence for a bipolar risk gene on chromosome 12q24 1 World Health Organization. The ICD-10 Classification of Mental suggested by investigation of haplotype sharing and allelic and Behavioural Disorders. Diagnostic criteria for research. World association in patients from the Faroe Islands. Mol Psychiatry Health Organization: Geneva, 1993. 2001; 6: 450–455. 2 Bolton P, Macdonald H, Pickles A, Rios P, Goode S, Crowson M et 24 Als TD, Dahl HA, Flint TJ, Wang AG, Vang M, Mors O et al. al. A case–control family history study of autism. J Child Psychol Possible evidence for a common risk locus for bipolar affective Psychiatry 1994; 35: 877–900. disorder and schizophrenia on chromosome 4p16 in patients from 3 Bailey A, Le Couteur A, Gottesman I, Bolton P, Simonoff E, Yuzda the Faroe Islands. Mol Psychiatry 2004; 9: 93–98. E et al. Autism as a strongly genetic disorder: evidence from a 25 Jorgensen TH, Degn B, Wang AG, Vang M, Gurling H, Kalsi G et al. British twin study. Psychol Med 1995; 25: 63–77. Linkage disequilibrium and demographic history of the isolated 4 Pickles A, Bolton P, Macdonald H, Bailey A, Le Couteur A, Sim C- population of the Faroe Islands. Eur J Hum Genet 2002; 10: 381– H et al. Latent-class analysis of recurrence risks for complex 387. phenotypes with selection and measurement error: a twin and 26 Lord C, Risi S, Lambrecht L, Cook Jr EH, Leventhal BL, DiLavore family history study of autism. Am J Hum Genet 1995; 57: 717– PC et al. The autism diagnostic observation schedule-generic: a 726. standard measure of social and communication deficits associated 5 Risch N, Spiker D, Lotspeich L, Nouri N, Hinds D, Hallmayer J et with the spectrum of autism. J Autism Dev Disord 2000; 30: 205– al. A genomic screen of autism: evidence for a multilocus etiology. 223. Am J Hum Genet 1999; 65: 493–507. 27 Lord C, Rutter M, Le Couteur A. Autism diagnostic interview- 6 IMGSAC. A genomewide screen for autism: strong evidence for revised: a revised version of a diagnostic interview for caregivers linkage to chromosomes 2q, 7q, and 16p. Am J Hum Genet 2001; of individuals with possible pervasive developmental disorders. J 69: 570–581. Autism Dev Disord 1994; 24: 659–685. 7 Buxbaum JD, Silverman JM, Smith CJ, Kilifarski M, Reichert J, 28 American Psychiatric Association. Diagnostic and Statistical Hollander E et al. Evidence for a susceptibility gene for autism on Manual of Mental Disorders. DSM-IV. 4th edn. American Psychia- chromosome 2 and for genetic heterogeneity. Am J Hum Genet tric Association: Washington, DC, 1994. 2001; 68: 1514–1520. 29 Escamilla MA, McInnes LA, Service SK, Spesny M, Reus VI, 8 Ramoz N, Reichert JG, Smith CJ, Silverman JM, Bespalova IN, Molina J et al. Genome screening for linkage disequilibrium in Davis KL et al. Linkage and association of the mitochondrial a Costa Rican sample of patients with bipolar-I disorder: a aspartate/glutamate carrier SLC25A12 gene with autism. Am J follow-up study on chromosome 18. Am J Med Genet 2001; 105: Psychiatry 2004; 161: 662–669. 207–213.

Molecular Psychiatry Genome-wide scan for autism susceptibility loci using LD-mapping MB Lauritsen et al 46 30 Bourgain C, Hoffjan S, Nicolae R, Newman D, Steiner L, Walker K 50 Hellings JA, Hossain S, Martin JK, Baratang RR. Psychopathology, et al. Novel case-control test in a founder population identifies P- GABA, and the Rubinstein-Taybi syndrome: a review and case selectin as an atopy-susceptibility locus. Am J Hum. Genet 2003; study. Am J Med Genet 2002; 114: 190–195. 73: 612–626. 51 Finelli P, Natacci F, Bonati MT, Gottardi G, Engelen JJ, Die- 31 Bourgain C, Genin E. Complex trait mapping in isolated popula- Smulders CE et al. FISH characterisation of an identical tions: Are specific statistical methods required? Eur J Hum Genet (16)(p11.2p12.2) tandem duplication in two unrelated patients 2005; 13: 698–706. with autistic behaviour. J Med Genet 2004; 41: e90. 32 Queller DC, Goodnight KF. Estimating relatedness using genetic 52 Lauritsen MB, Ewald H. The genetics of autism. Acta Psychiatr markers. Evolution 1989; 43: 258–275. Scand 2001; 103: 411–427. 33 Hardy OJ, Vekemans X. SPAGEDi: a versatile computer program to 53 The European Chromosome 16 Tuberous Sclerosis Consortium. analyse spatial genetic structure at the individual or population Identification and characterization of the tuberous sclerosis gene levels. Mol Ecol Notes 2002; 2: 618–620. on chromosome 16. Cell 1993; 75: 1305–1315. 34 Rousset F. Genetic differentiation between individuals. J Evol 54 Serajee FJ, Nabi R, Zhong H, Mahbubul Huq AH. Association of Biology 2000; 13: 58–62. INPP1, PIK3CG, and TSC2 gene variants with autistic disorder: 35 Weir B. Genetic Data Analysis II. Sinauer Associates: Massachu- implications for phosphatidylinositol signalling in autism. J Med setts, Sunderland, Date: 1996. Genet 2003; 40: e119. 36 Pritchard JK, Stephens M, Donnelly P. Inference of population 55 Barrett S, Beck JC, Bernier R, Bisson E, Braun TA, Casavant TL et structure using multilocus genotype data. Genetics 2000; 155: al. An autosomal genomic screen for autism. Am J Med Genet 945–959. 1999; 88: 609–615. 37 Falush D, Stephens M, Pritchard JK. Inference of population 56 Shastry BS. Molecular genetics of autism spectrum disorders. J structure using multilocus genotype data: linked loci and Hum Genet 2003; 48: 495–501. correlated allele frequencies. Genetics 2003; 164: 1567–1587. 57 Constantino JN, Todd RD. Autistic traits in the general population: 38 Sham PC, Curtis D. Monte Carlo tests for associations between a twin study. Arch Gen Psychiatry 2003; 60: 524–530. disease and alleles at highly polymorphic loci. Ann Hum Genet 58 Lord C, Pickles A, McLennan J, Rutter M, Bregman J, Folstein S et 1995; 59: 97–105. al. Diagnosing autism: analyses of data from the autism diagnostic 39 Houwen RH, Baharloo S, Blankenship K, Raeymaekers P, Juyn J, interview. J Autism Dev Disord 1997; 27: 501–517. Sandkuijl LA et al. Genome screening by searching for shared 59 de Bildt A, Sytema S, Ketelaars C, Kraijer D, Mulder E, Volkmar F segments: mapping a gene for benign recurrent intrahepatic et al. Interrelationship between autism diagnostic observation cholestasis. Nat Genet 1994; 8: 380–386. schedule-generic (ADOS-G), autism diagnostic interview-revised 40 Durham LK, Feingold E. Genome scanning for segments shared (ADI-R), and the diagnostic and statistical manual of mental identical by descent among distant relatives in isolated popula- disorders (DSM-IV-TR) classification in children and adolescents tions. Am J Hum Genet 1997; 61: 830–842. with mental retardation. J Autism Dev Disord 2004; 34: 129–137. 41 Shao Y, Wolpert CM, Raiford KL, Menold MM, Donnelly SL, 60 Newman DL, Hoffjan S, Bourgain C, Abney M, Nicolae RI, Profits Ravan SA et al. Genomic screen and follow-up analysis for autistic ET et al. Are common disease susceptibility alleles the same in disorder. Am J Med Genet 2002; 114: 99–105. outbred and founder populations? Eur J Hum Genet 2004; 12: 584– 42 McCauley JL, Li C, Jiang L, Olson LM, Crockett G, Gainer K et al. 590. Genome-wide and ordered-subset linkage analyses provide sup- 61 Taillon-Miller P, Bauer-Sardina I, Saccone NL, Putzel J, Laitinen T, port for autism loci on 17q and 19p with evidence of phenotypic Cao A et al. Juxtaposed regions of extensive and minimal linkage and interlocus genetic correlates. BMC Med Genet 2005; 6:1. disequilibrium in human Xq25 and Xq28. Nat Genet 2000; 25: 43 Hauser ER, Boehnke M, Guo SW, Risch N. Affected-sib- 324–328. pair interval mapping and exclusion for complex genetic 62 Eaves IA, Merriman TR, Barber RA, Nutland S, Tuomilehto-Wolf traits: sampling considerations. Genet Epidemiol 1996; 13: E, Tuomilehto J et al. The genetically isolated populations of 117–137. Finland and Sardinia may not be a panacea for linkage 44 Buxbaum JD, Silverman J, Keddache M, Smith CJ, Hollander E, disequilibrium mapping of common disease genes. Nat Genet Ramoz N et al. Linkage analysis for autism in a subset families 2000; 25: 320–323. with obsessive-compulsive behaviors: evidence for an autism 63 Shifman S, Darvasi A. The value of isolated populations. Nat susceptibility gene on chromosome 1 and further support for Genet 2001; 28: 309–310. susceptibility genes on chromosome 6 and 19. Mol Psychiatry 64 Lonjou C, Zhang W, Collins A, Tapper WJ, Elahi E, Maniatis N et 2004; 9: 144–150. al. Linkage disequilibrium in human populations. Proc Natl Acad 45 Philippe A, Martinez M, Guilloud-Bataille M, Gillberg C, Rastam Sci USA 2003; 100: 6069–6074. M, Sponheim E et al. Genome-wide scan for autism susceptibility 65 Peltonen L, Palotie A, Lange K. Use of population isolates for genes. Paris Autism Research International Sibpair study. Hum mapping complex traits. Nat Rev Genet 2000; 1: 182–190. Mol Genet 1999; 8: 805–812. 66 Pericak-Vance MA. Linkage disequilibrium and allelic association. 46 International Molecular Genetic Study of Autism Consortium. In: Haines JL, Pericak-Vance MA (eds). Approaches to Gene A full genome screen for autism with evidence for linkage Mapping in Complex Human Diseases. John Wiley & Sons: New to a region on chromosome 7q. Hum Mol Genet 1998; 7: York, Chichester, Weinheim, Brisbane, Singapore, Toronto, 1998; 571–578. 323–334. 47 Barnby G, Abbott A, Sykes N, Morris A, Weeks DE, Mott R 67 Wright AF, Teague PW, Bruford E, Carothers A. Problems in et al. Candidate-gene screening and association analysis at the dealing with linkage heterogeneity in autosomal recessive forms of autism-susceptibility locus on chromosome 16p: evidence of retinitis pigmentosa. In: Pawlowitzki I-H, Edwards JH, Thompson association at GRIN2A and ABAT. Am J Hum Genet 2005; 76: EA (eds). Genetic Mapping of Disease Genes. Academic Press Inc.: 950–966. San Diego, London, NY, 1997; 255–272. 48 Hebebrand J, Martin M, Korner J, Roitzheim B, de Braganca K, 68 Ober C, Cox NJ. The genetics of asthma. Mapping genes for Werner W et al. Partial trisomy 16p in an adolescent with autistic complex traits in founder populations. Clin Exp Allergy 1998; disorder and Tourette’s syndrome. Am J Med Genet 1994; 54:268– 28(Suppl 1): 101–105. 270. 69 Speer MC. Basic concepts in genetics. In: Haines JL, Pericak-Vance 49 Carrasco Juan JL, Cigudosa JC, Otero GA, Acosta Almeida MT, MA (eds). Approaches to Gene Mapping in Complex Human Garcia Miranda JL. De novo trisomy 16p. Am J Med Genet 1997; 68: Diseases. John Wiley & Sons: New York, Chichester, Weinheim, 219–221. Brisbane, Singapore, Toronto, 1998; 17–52.

Molecular Psychiatry