Identifying Autism Susceptibility Genes

Neuron, Vol. 28, 19–24, October, 2000, Copyright 2000 by Cell Press

Identifying Autism Susceptibility Genes Elena Maestrini,‡§ Alina Paul,* Anthony P. Monaco,† and Anthony Bailey* * MRC Child Psychiatry Unit Institute of Psychiatry De Crespigny Park London, SE5 8AF † Wellcome Trust Centre for Human Genetics University of Oxford Roosevelt Drive Oxford, OX3 7BN United Kingdom ‡ Department of Biology University of Bologna via Selmi 3 40126 Bologna Italy

Introduction The current interest in identifying susceptibility loci for autism reflects the degree to which molecular genetic findings seem likely to transform our understanding of the underlying pathology, improve diagnosis and genetic counseling, and eventually lead to new strategies of prevention and treatment for this handicapping disease. Autism represents the prototypical pervasive developmental disorder (PDD): a group of neurodevelopmental conditions characterized by impaired social interaction and communication, accompanied by unusually restricted and stereotyped patterns of behaviors and interests and an onset in the first 3 years of life (for review, see Bailey et al., 1996). The diagnostic features of autism are outlined in Table 1. The clinical presentation can vary considerably between affected individuals and is influenced by intellectual level and the possible presence of nonspecific behaviors such as hyperactivity and self-injury. Several reports of abnormalities in the hippocampus, amygdala, cerebellum, and cerebral cortex may provide some clues about the neuroanatomical basis of autism; however, there is still no consensus about how these abnormalities relate to developmental psychopathology, general cognitive impairment, and the development of epilepsy. Importance of Hereditary Factors in the Etiology of Autism Autism is etiologically heterogeneous; known medical conditions are implicated in perhaps 10%–25% of cases, the strongest associations being with tuberous sclerosis and the Fragile X syndrome. Family and twin studies show evidence for a substantial genetic predisposition in most idiopathic cases. The prevalence of core autism has been estimated to be 5 in 10,000 (Fombonne, 1999), whereas the rate of similarly defined autism among siblings of affected probands is between 2% and 3% (Szatmari et al., 1998). All three epidemiolog§ To whom correspondence should be addressed (e-mail: maestrin@

alma.unibo.it).

Review

ical same-sex twin studies suggest that the elevated familial recurrence risk has a genetic basis; concordance for autism in monozygotic (MZ) twin pairs is substantial (36%–90%), whereas in dizygotic (DZ) pairs the rate is similar to the rate of autism among siblings of affected children. Twin studies also suggest that the genetic predisposition extends to related, but milder, communication and/or social deficits. In the largest study (Bailey et al., 1995), 92% of MZ pairs were concordant for a broader spectrum of social/cognitive abnormalities—the most severely affected cotwins showing other PDDs; this was in marked contrast to the 10% concordance rate in DZ pairs. A similar range and pattern of difficulties is observed among a minority of relatives in family studies (Bailey et al., 1998). The genetic mechanisms predisposing to autism and related milder phenotypes are not known, but neither the level of familial risk nor the very different concordance rates in MZ and DZ twin pairs is compatible with a simple monogenic model. It is possible that Mendelian, monogenic inheritance is implicated in a minority of individuals, but the majority of cases seem likely to arise on the basis of multiple susceptibility genes; otherwise known as polygenic or oligogenic inheritance. In such genetically complex disorders, functional variants in susceptibility genes have only a weak or moderate effect: they confer an increased risk of developing the disorder but each locus is insufficient alone to cause the full clinical phenotype. There may also be heterogeneity with susceptibility attributable to several, possibly overlapping, sets of interacting genes. Models using family and twin data for autism and related phenotypes have been generated suggesting that between two and ten loci may be implicated (Pickles et al., 1995), a model with three interacting loci providing the best fit. Whether individual loci predispose to specific components of the behavioral phenotype is presently unclear. In addition, other genetic mechanisms, such as expanding triplet repeats or imprinting may contribute to the complexities of inheritance and phenotypic variability. Clues from Chromosomal Abnormalities The strong genetic predisposition and the early age of onset make autism an obvious candidate for molecular genetic approaches to identify the underlying biological mechanisms. An association between a disease and chromosomal abnormalities—such as deletions, translocations, and duplications—can point to areas of potential interest for mapping. A great variety of chromosomal abnormalities have been reported in individuals with autism (reviewed in Gillberg, 1998), but one potentially specific association is with abnormalities of an unstable region of chromosome 15 (15q11-q13). The most common abnormalities are interstitial duplications or a supernumerary pseudodicentric chromosome 15 (inv-dup[15]) in patients with autism spectrum disorders—often associated with seizures, hypotonia, and severe mental retardation (for ex-

Neuron 20

Table 1. Clinical Features of Autism Qualitative impairments in communication: —Delay in or lack of language development together with a paucity of social usage of whatever language skills are present. —Impairment in spontaneous make-believe play and social imitative play. —Poor flexibility and relative lack of creativity in language expression. —Lack of accompanying gesture to emphasize or aid meaning in communication. —Impaired emotional response to other people’s verbal and nonverbal overtures. Qualitative impairments in social interaction: —Lack of social reciprocity due to inadequate appreciation of socioemotional cues. This is evident in the impaired response to other people’s emotions and/or poor modulation of behavior according to social context. —Impaired usage of social, emotional, and communicative behaviors including eye contact and facial expression. —Impairment in the development of peer relationships. Restrictive repetitive and stereotyped behavior, interests, and activities: —Tendency to impose rigidity and routine on a range of aspects of day-to-day functioning. This applies to novel activities as well as to familiar habits and play patterns. —May have stereotyped preoccupations with interests such as dates, routes, or timetables. —Motor stereotypes such as hand or finger flapping and twisting.

ample, see Cook et al., 1997a). The 15q11-q13 region has been shown to contain several genes subject to imprinting, a phenomenon resulting in differential expression of alleles at a locus, depending on the parent of origin (Reik and Walter, 1998). Problems with imprinted genes at 15q11-q13 give rise to the Prader-Willi syndrome (PWS, MIM 176,260) and the Angelman syndrome (AS, MIM 105,830), both disorders that are reportedly associated with autistic-like symptoms in a subset of cases. Interestingly, when parental origin has been investigated, all of the 15q duplications associated with autism were derived from the mother, raising the possibility that any putative susceptibility locus for autism may be imprinted. Rarely sex chromosome abnormalities have been reported in association with autism; a finding of potential interest since three-quarters of individuals with autism are male. Skuse et al. (1997) described an increased incidence of autism among females with 45,X (Turner syndrome). They also found that maternal origin of the single X chromosome was associated with significantly more deficits in social cognition and verbal IQ than paternal origin. Whether this particular apparent imprinting effect is relevant in idiopathic cases of autism is unclear, as fathers seem to show a higher rate of milder autismrelated phenotypes than mothers (Bailey et al., 1998). Mapping Disease Genes: Linkage and Association Genetic mapping relies on tracking the cosegregation of polymorphic DNA markers (naturally occurring variations in DNA sequence) with a disorder. A statistically significant association between the disorder and markers at an identified genetic location implies a nearby susceptibility locus, ultimately permitting identification of the gene without a priori knowledge of its biological function. There are two different approaches to genetic mapping: linkage and association (Figure 1). Linkage is based on identifying markers that cosegregate with the disease within families and classically uses multigenerational pedigrees—containing several affected individuals—in which it is possible to infer the mode of inheritance of the disease locus and specify the relevant parameters. In genetically complex disorders, the mode of inheritance is generally unknown and large affected pedigrees are unavailable. Therefore, alternative

“model-free” linkage methods are used that do not require specification of an inheritance model but instead rely on assessing allele sharing between two or more affected individuals within multiple nuclear families (Figure 1). The goal is to identify markers at which affected pairs share alleles more often than expected by chance, as such markers may be linked to a susceptibility gene. Although these model-free linkage methods are robust, they have relatively low power to detect genes of small effect; a situation in which association studies are more powerful. Association strategies are based on identifying alleles that occur at significantly different frequencies in samples of affected and control individuals. Case-control studies utilize unrelated controls, and careful matching is required to avoid false positive results due to cases and controls being drawn from different gene pools. Family-based association designs, such as the Transmission Disequilibrium Test (TDT, Figure 1), overcome this potential difficulty by using as internal controls the frequencies of parental alleles that are not transmitted to affected offspring (Spielman and Ewens, 1996). Association of a particular allele with a disease can occur if the allele is implicated in the etiology of the disease, or if it is in linkage disequilibrium with the susceptibility locus. Association is usually only detectable if the allele is very close to or within the susceptibility gene, so the approach is predominantly used to test polymorphisms within known candidate genes, or as a tool for fine mapping a region of interest identified by linkage studies. Genome-Wide Linkage Studies In the absence of strong clues to the location or nature of susceptibility loci for autism, many groups have embarked on a whole genome screen using an affected relative-pair design. This typically involves genotyping 300–400 highly polymorphic microsatellite markers, evenly spaced throughout the genome. All the available linkage information is then used to calculate the maximum lod score (MLS) value at each point of the genome to generate multipoint MLS profiles along the 23 chromosomes. The peaks in these profiles represent the possible locations of susceptibility genes. Five genome-wide scans have been completed to date (see Table 2). In interpreting their findings, some

Review 21

Figure 1. Linkage and Association as Tools to Disease Gene Mapping

general difficulties posed by this approach to the study of complex traits must be considered. Quite what represents significant linkage in a whole genome screen is unclear, as the procedure involves multiple statistical tests and the number of true susceptibility loci is not known. Some authors have proposed a genome-wide lod score threshold of 3.6 for affected sib-pairs (resulting in a 5% false positive rate in a whole genome scan). Nevertheless, that threshold will not always be attainable in an average size study (100–200 sib-pairs) when the loci are of moderate effect, and adopting a high threshold for declaring linkage may produce a considerable decrease in power. One strategy is to report all “hits” suggestive of linkage in a genome scan, accepting that false positives will be generated along with the true results. Supplementary evidence can then be sought to distinguish between the true and false results. Independent replication of positive findings is generally accepted as an indicator of a true positive result, but that can also be difficult to achieve as the number of families required to replicate linkage in a multilocus disorder is substantially larger than the number needed to establish linkage (Suarez et al., 1994). A further difficulty is that in multilocus disorders location estimates for relatively weak loci may be several centimorgans (cM) away from the true disease locus (Roberts et al., 1999). Colocalization of a suggestive linkage with a region identified from a cytogenetic abnormality associated with the disorder provides a further means of distinguishing true from false positives. Other factors that may complicate comparison of the results from the genome scans reviewed in Table 2 include differences in the ascertainment of cases and inclusion–exclusion criteria, the use of different analytical approaches and marker maps, and the varying size of family collections. Although all five research groups followed somewhat similar diagnostic and inclusion criteria, there are still identifiable differ-

ences between samples in terms of exclusion of other disorders, age at assessment, sex ratio, and IQ distribution; there are also likely to be other phenotypic differences. The International Molecular Genetic Study of Autism Consortium (IMGSAC, 1998), the Paris Autism Research International Sibpair Study (PARIS) (Philippe et al., 1999), and Stanford study (Risch et al., 1999) employed model-free likelihood methods, while the Collaborative Linkage Study of Autism (CLSA, 1999) and the Duke group (Ashley-Koch et al., 1999; Bass et al., 2000) used a combination of model-free and parametric lod score methods. A two-stage search was carried out by the IMGSAC and Stanford research groups, while the others analyzed a single set of families. The size of family collections ranged from 51 (PARIS) to 139 multiplex families (Stanford study). Although none of the genome scans found a locus with conclusive evidence for linkage (MLS ⬎ 3.6), two or more studies found suggestive evidence of linkage in overlapping regions, which are likely to represent true linkage findings. The most consistent results are for a region on chromosome 7q. Linkage to this region was first reported by the IMGSAC, and it represents the highest score obtained in their genome scan (IMGSAC, 1998), providing a multipoint MLS of 2.53 in 99 multiplex families. Although the other groups each reported a different locus as their most significant result (Table 2), they all detected some degree of increased sharing in 7q. This is a very exciting outcome given the general difficulties in replicating linkage claims in psychiatric disorders. However, a closer look at the individual studies reveals a complex scenario, with some of the positive findings pointing to genomic regions several cM away from each other. Further work including combinations of linkage, association and candidate genes approaches, is now in progress to better define this potential susceptibility locus. The PARIS group (Philippe et

1.78 D15S217 3.00 0.54

2.20 0.80

1.21 1.00 D17S1876 D18S878

cM, position in Kosambi centiMorgans based on the Genethon linkage map. 1 Results from follow-up studies.

D19S49 D22S264 48 5

0.99 1.39

D18S68 D19S226 94 41

D16S407 17

1.51

D15S118 D16S3075 32 21

1.10 0.74

0.84 D10S217 167 D13S193 85

0.59

0.62 1.37

55

11 100

D13S800

0.93 D7S1804 138 0.83 D7S486

199 5 109

125 2.53 1.36 D7S530 D10S197 136 52

1.01 D7S2564 42 0.88 0.84 2.23 D4S1535 D5S417 D6S283

0.64 D2S364 192 1.55 D4S412 5

0.68

104 150

55 32

D13S800 ACTC

D7S1813 D7S1824

Nearest Marker cM 2.15

MLS Nearest Marker

D1S1675 149

cM MLS Nearest Marker cM MLS Nearest Marker cM Chrom

1p 2q 4p 4q 5p 6q 7p 7q 7q 10p 10q 13q 15q 16p 17p 18q 19q 22q

CLSA STANFORD PARIS IMGSAC

Table 2. Summary of Results from Published Genome Scans

15

D7S640 139

cM MMLS/Het

DUKE1

Nearest Marker

2.01

NPL Genehunter

Neuron 22

al., 1999) reported an MLS of 0.83 at the marker D7S486, and the Stanford group identified an MLS of 0.93 at D7S1804 (Risch et al., 1999); both markers are within 15 cM of each other and the IMGSAC peak. Yet, in the CLSA study (1998), there were two distinct peaks separated by over 45 cM; the most significant is ⵑ30 cM proximal to the IMGSAC peak, whereas the other is distal. The two peaks reached a score of 2.20 and 0.80, respectively, using the MMLS/het statistic; this method maximizes the lod score over different models (dominant, recessive) and over an admixture parameter to allow for locus heterogeneity. The authors suggested these two peaks on 7q might relate to a single locus that is possibly coincident with that identified by others. The Duke group have published a follow-up study of the 7q31–7q35 region (Ashley-Koch et al., 1999), which was investigated because of the results from the other scans, their initial screen findings and access to a family segregating a paracentric inversion involving the chromosome 7 candidate region, inv(7)(q22-q31.2). The inversion was inherited from the unaffected mother by three siblings: two brothers have autism, while their sister has expressive language impairment. Precise mapping of the breakpoints was not performed; however, the distal breakpoint of the inversion is located in 7q31.2, proximal to the IMGSAC linkage region. Linkage analysis in 76 multiplex families gave positive lod scores with markers distal to the inversion breakpoint overlapping the IMGSAC candidate region (Genehunter NPL score of 2.01 at marker D7S640). Sib-pair analysis revealed an increased paternal contribution to IBD sharing and an increased recombination rate in autism families compared to CEPH controls was also observed. TDT analysis demonstrated a weak transmission distortion at the two distal markers D7S495 (p ⫽ 0.03) and D7S1824 (p ⫽ 0.01), again with a greater contribution from paternal transmissions, suggesting a possible imprinting effect. The linkage, association, cytogenetic, and parental origin data in the Duke study are not entirely consistent, as the positive linkage results are telomeric to the breakpoint in 7q31.2 and the inversion is maternally transmitted, in contrast to the paternal influence on the linkage and association findings. The second emerging area of interest is on chromosome 15q11-q13, overlapping with the region involved in the cytogenetic rearrangements associated with autism. A cluster of genes encoding ␥-aminobutyric acid receptor (GABAA) subunits, considered potential candidate genes for autism, also map to this region. The Duke group investigated this region in detail (Bass et al., 2000). Parametric and nonparametric analysis on 63 multiplex families gave a peak at D15S217 (p ⫽ 0.03), distal to the GABAA receptor genes cluster. An increased recombination rate in autism families was also observed for this region. Other linkage results pointing to this region are a MLS ⫽ 1.10 from the PARIS genome scan, and a very modest increased sharing detected by the CLSA group. In the IMGSAC and Stanford families, this region showed no evidence of linkage. Other potential areas of interest showing increased sharing in more then one study include regions on chromosome 13q and 16p. The 13q region was the most significant result in the CLSA genome scan, with a MMLS/het score of 3 and an estimated proportion of

Review 23

families linked to this locus of 35%. Weak increased sharing in the same 13q region also appeared in the Stanford genome screen (MLS ⫽ 0.68); in the IMGSAC study, a modestly positive lod score (MLS ⫽ 0.59) was reported ⵑ30 cM distal to the CLSA highest peak. The 16p region was the second most significant result detected by the IMGSAC (MLS ⫽ 1.51); the PARIS study detected a MLS ⫽ 0.74 in 16p, while no increased sharing was present in the remaining genome scans. To date none of the genome scans have found linkage to chromosome X. Association Studies Positive associations using case-control designs have been reported for polymorphisms at several loci, including the c-Harvey-ras-1 (HRAS1) gene, the EN2 homeogene, and specific HLA haplotypes. These associations may be spurious due to either population stratification or the relatively small sample sizes and need confirmation. The most promising candidate genes are claimed to be those regulating neurotransmitter activity, particularly the serotonergic system. Several association studies have investigated specific variants of the serotonin transporter (5-HTT) gene for association with autism. Cook et al. (1997b) reported an association with the short allele of the 5-HTT promoter polymorphism in a sample from the USA (p ⫽ 0.03), a German TDT study reported preferential transmission of the long variant (Klauck et al., 1997), and neither association has been replicated in other independent family samples (for example, see Maestrini et al., 1999). Despite these contradictory findings, the serotonergic system is likely to remain a focus of study. Cook et al. (1998) have also examined several markers in the region 15q11-q13 and reported linkage disequilibrium with a marker within the GABRA3 gene (GABRA3–155CA-2) (p ⫽ 0.0014), although others have not replicated these findings. Conclusions Recent molecular genetic studies indicate that the goal of identifying autism susceptibility genes may soon be attainable. Taken together, the linkage findings suggest that there is a high probability that a region on chromosome 7q contains a susceptibility locus for autism. Nevertheless, the isolation of a putative gene in this region still represents a considerable challenge. The reported linkage peaks cover a very large genomic region (45 cM), and, although more than one locus might be implicated, the variation in location estimates could have arisen by chance or be due to other factors that may confound genetic mapping, such as an increased recombination rate. Clearly a metaanalysis offers one approach to extracting the maximum information from the different studies. At present, the 15q region shows weaker evidence for linkage, but the relatively high incidence of chromosomal abnormalities supports a role for this region in autism etiology, although the effect of this putative locus may be restricted to a smaller subgroup of autism cases. Interestingly, both the 15q and 7q regions contain imprinted genes suggesting that abnormal imprinting may be implicated in the pathogenesis of autism. Over the next few years, we can expect new genome

screens and the enlargement of existing family collections. Larger sample sizes will help all groups to increasingly distinguish between true and spurious linkages and further overlapping areas of linkage are likely to emerge. We can also anticipate the use of more sophisticated analytical approaches to extract the full genetic information from family data by accurately measuring milder autism phenotypes and dimensionalizing the different phenotypic components for use in Quantitative Trait Loci analysis. The identification of the first autism susceptibility gene is likely to accelerate the discovery of other genes and will hopefully provide pointers to the mechanisms underlying abnormal brain development and function. That leap in our knowledge will be crucial for developing new preventative and treatment strategies. References Ashley-Koch, A., Wolpert, C.M., Menold, M.M., Zaeem, L., Basu, S., Donnelly, S.L., Ravan, S.A., Powell, C.M., Qumsiyeh, M.B., Aylsworth, A.S., et al. (1999). Genetic studies of autistic disorder and chromosome 7. Genomics 61, 227–236. Bailey, A., Le Couteur, A., Gottesman, I., Bolton, P., Simonoff, E., Yuzda, E., and Rutter, M. (1995). Autism as a strongly genetic disorder: evidence from a British twin study. Psychol. Med. 25, 63–77. Bailey, A., Philips, W., and Rutter, M. (1996). Autism: towards an integration of clinical, genetic, neuropsychological, and neurobiological perspectives. J. Psychol. Psych. 37, 89–126. Bailey, A., Palferman, S., Heavey, L., and Le Couteur, A. (1998). Autism: the phenotype in relatives. J. Aut. Dev. Dis. 28, 439–445. Bass, M.P., Menold, M.M., Wolpert, C.M., Donnelly, S.L., Ravan, S.A., Hauser, E.R., Maddox, L.O., Vance, J.M., Abramson, R.K., Wright, H.H., et al. (2000). Genetic studies in autistic disorder and chromosome 15. Neurogenetics 2, 219–226. Collaborative Linkage Study of Autism (CLSA): Barrett, S., Beck, J.C., Bernier, R., et al. (1999). An autosomal genome screen for autism. Am. J. Med. Genet. 88, 609–615. Cook, E.H., Lindgren, V., Leventhal, B.L., Courchesne, R., Lincoln, A., Shulman, C., Lord, C., and Courchesne, E. (1997a). Autism or atypical autism in maternally but not paternally derived proximal 15q duplication. Am. J. Hum. Genet. 60, 928–934. Cook, E.H., Courchesne, R., Lord, C., Cox, N.J., Yan, S., Lincoln, A., Haas, R., Courchesne, E., and Leventhal, B.L. (1997b). Evidence of linkage between the serotonin transporter and autistic disorder. Mol. Psychiatry 2, 247–250. Cook, E.H., Courchesne, R.Y., Cox, N.J., Lord, C., Gonen, D., Guter, S.J., Lincoln, A., Nix, K., Haas, R., Leventhal, B.L., and Courchesne, E. (1998). Linkage-disequilibrium mapping of autistic disorder, with 15q11–13 markers. Am. J. Hum. Genet. 62, 1077–1083. Fombonne, E. (1999). The epidemiology of autism: a review. Psychol. Med. 29, 769–786. Gillberg, C. (1998). Chromosomal disorders and autism. J. Aut. Dev. Dis. 28, 415–425. International Molecular Genetic Study of Autism Consortium (IMGSAC) (1998). A full genome screen for autism with evidence for linkage to a region on chromosome 7q. Hum. Mol. Genet. 7, 571–578. Klauck, S.M., Poustka, F., Benner, A., Lesch, K.P., and Poustka, A. (1997). Serotonin transporter (5-HTT) gene variants associated with autism? Hum. Mol. Genet. 6, 2233–2238. Maestrini, E., Lai, C., Marlow, A., Matthews, N., Wallace, S., Bailey, A., Cook, E.H., Weeks, D.E., Monaco, A.P., and IMGSAC (1999). Serotonin transporter (5-HTT) and ␥-aminobutyric acid receptor subunit ␤-3 (GABRB3) gene polymorphisms are not associated with autism in the IMGSA families. Am. J. Med. Genet. 88, 492–496. Philippe, A., Martinez, M., Bataille-Guillot, M., Gillberg, C., Rastam, M., Sponheim, E., Coleman, M., Zappella, M., Aschauer, H., van

Neuron 24

Malldergerme, L., et al. (1999). Genome-wide scan for autism susceptibility genes. Hum. Mol. Genet. 8, 805–812. Pickles, A., Bolton, P., Macdonald, H., Bailey, A., Le Couteur, A., Sim, C.H., and Rutter, M. (1995). Latent-class analysis of recurrence risk for complex phenotypes with selection and measurement error: a twin and family history study of autism. Am. J. Hum. Genet. 57, 717–726. Reik, W., and Walter, J. (1998). Imprinting mechanisms in mammals. Curr. Opin. Genet. Dev. 8, 154–164. Risch, N., Spiker, D., Lotspeich, L., Nouri, N., Hinds, D., Hallmeyer, J., Kalaydjieva, L., McCague, P., Dimiceli, S., Pitts, T., et al. (1999). A genomic screen of autism: evidence for a multilocus etiology. Am. J. Hum. Genet. 65, 493–507. Roberts, S.B., MacLean, C.J., Neale, M.C., Eaves, L.J., and Kendler, K.S. (1999). Replication of linkage studies of complex traits: an examination of variation in location estimates. Am. J. Hum. Genet. 65, 876–884. Skuse, D.H., James, R.S., Bishop, D.V., Coppin, B., Dalton, P., Aamodt-Leeper, G., Bacarese-Hamilton, M., Creswell, C., McGurk, R., and Jacobs, P.A. (1997). Evidence from Turner’s syndrome of an imprinted X-linked locus affecting cognitive function. Nature 387, 705–708. Spielman, R.S., and Ewens, W.J. (1996). The TDT and other familybased tests for linkage-disequilibrium and association. Am. J. Hum. Genet. 59, 983–989. Suarez, B.K., Hampe, C.L., and Van Eerdewegh, P. (1994). Problems of replicating linkage claims in psychiatry. In Genetic Approaches to Mental Disorder, E.S. Gershon and C. R. Cloninger, eds. (Washington, DC: American Psychiatric Press). Szatmari, P., Jones, M.B., Zwaigenbaum, L., and MacLean, J.E. (1998). Genetics of autism:overview and new directions. J. Aut. Dev. Dis. 28, 351–368.