SUSAN E. FOLSTEIN, M.D., JONATHAN HAINES, PH.D., AND SUSAN L. SANTANGELO, SC.D.

      Autism is a neurodevelopmental disorder that is defined, not by its etiology or neural mechanism, but by a unique and strange cluster of behaviors that are usually evident in the first year of life: the inability to form normal, reciprocal social relationships; abnormal (or no) language development, with particular difficulties with social communication (pragmatics) and abstractions; and repetitive, restricted interests and behaviors.
      While abnormal social interaction is most commonly noticeable during the first year of life, as many as 30 percent of cases become manifest after 18-24 months of normal or near-normal development. Intelligence varies from well within the normal range to profoundly retarded, with about 75 percent of affected individuals having measured intelligence in the mentally retarded range (below an IQ of 70). On the Wechsler Scales, autistic persons, regardless of overall IQ, do relatively well on the subtests that tap rote memory (digit span) and some visuo-spacial tasks (block design and object assembly). They do poorly on subtests that tap verbal abstractions and sequencing (picture arrangement and comprehension) .
      Many children improve as they grow older, but social interaction rarely becomes normal. Prognosis is related most closely to the level of language development .Seizures may develop, particularly after puberty in those autistic individuals with moderate to severe mental retardation. Numerous biochemical studies have found increased levels of whole blood and platelet serotonin, and one has suggested decreased central serotonin responsiveness . Other neurotransmitter systems have also been implicated, but have not been consistently replicated.
      A number of the children have larger than average heads, with some in the macrocephalic range (above the 98th percentile of head circumference) .On MRI, this increased head circumference has been shown to be attributable to increased brain volume, more temporal-parietal than frontal . Only a few autopsy studies have been done, but they consistently demonstrate small neurons in several limbic nucleii and decreased numbers of Purkinje cells in the cerebellar hemisphere . Although some MRI studies (but not others) have suggested abnormalities in the cerebellar vermis, this has not been confirmed by neuronal counts of vermis in autopsy specimens . Together, the MRI and autopsy findings suggest abnormal neural development in the limbic cortex and cerebellar hemispheres. The prevalence of autism is between 5/10000 and 10/10000, depending on how broadly or narrowly the diagnostic criteria are interpreted.
      KNOWN ETIOLOGIES
      The majority of cases are of unknown etiology, but there are a few disorders in which autism may occur more commonly than by chance, and about 10 percent of cases have some discernable etiology . However, an accurate measure of the expected level of random chance co-occurrence is somewhat difficult to decide because of the frequent overlap between autism and mental retardation. The fragile X (FRAX-A) mutation is found in about 3 percent of unselected autism cases. Autism is very frequent in tuberous sclerosis cases that are severely mentally retarded . The disorder was said to be commonly associated with untreated phenylketonuria (PKU), but the rate was never firmly established, and the reported cases of autism with PKU were not well enough described to be sure of this claim . Currently, untreated cases of PKU are rarely seen, so that the association can no longer be studied in developed countries. Similar difficulties surround the reported association with congenital rubella.
      Numerous studies have been carried out to investigate the relationship between autism and prenatal and perinatal problems. Compared with IQ-matched controls, autistic individuals have a very small rate of untoward events. Compared with sibling controls, they are often reported to have a slight but significant increase in the occurrence of a less-than-optimal prenatal and perinatal course. This optimality score is based on the unweighted sum of all events that have been reported to be risk factors in other studies, including both minor (for example, primiparity) and major (low Apgar score at 5 minutes, for instance) events.
      Because autistic children are often only children, primiparity is common. When primiparity is omitted from the tally of the optimality score or accounted for in a multivariate analysis , the differences between autistic individuals and their siblings are no longer significant. It is possible that relatively minor pre- and perinatal difficulties may have a negative impact when they occur on a vulnerable genetic background , but it has not yet been possible to test this hypothesis.
      EVIDENCE FOR GENETIC ETIOLOGY
      Among the vast majority of cases with unknown cause, there is clear evidence for a genetic etiology. First, even though the recurrence risk is about 6 percent to 8 percent , this is between 60 and 160 times what would be expected by chance based on the population prevalence of 5-10 per 10,000. Second, population-based twin studies have consistently found a very large difference between the monozygotic (MZ) and dizygotic (DZ) concordance rates. Twin pairs are said to be concordant when both are affected with the disorder or trait under study. Based on summing the results of three population-based twin studies, the MZ concordance rate is about 60 percent while the DZ rate is 0 percent . While the MZ pairwise concordance rate is quite high, the DZ rate would be expected to equal to the sibling recurrence risk of 6 percent to 8 percent. The lack of any concordant DZ pairs is most likely a Type II error since the total number of DZ pairs studied is only 30 and the chance of observing no concordant DZ pairs is 16 percent.
      Thus, heritability is estimated to be very high, accounting for more than 90 percent of the total variance. However, the pattern of transmission in families does not conform to any simple genetic mechanism. Genetic models, based on the results of twin and family studies, suggest that the disorder is most likely caused by the interactive effect of several genes, at least three . However, it is unknown whether all autism is caused by the same set of genes (perhaps say four to six) or whether there may be a larger number that combine in different ways. The latter is thought possible because of the wide variation in intelligence and language. In addition, there is a possibility that in some small subsets of families, autism may be transmitted by a single gene, possibly one that is not expressed in everyone who has the gene (that is, a gene with reduced penetrance)
      . NON-AUTISTIC FAMILY MEMBERS
      Because of the severe social deficits, autistic individuals rarely have children, so that parents of autistic children are rarely autistic themselves. Nevertheless, parent-child pairs are known to exist. Although the majority of autistic individuals also have mental retardation, most studies of parents and siblings of autistic children have not found any increase in the rate of mental retardation among the first-degree relatives, as compared with first-degree relatives of controls .
      There are, however, some traits and disorders that do distinguish the parents and siblings of autistic children from controls. These traits are reminiscent of autism, but generally do not cause impairment or handicap and often have adaptive value . Several studies have found higher rates of developmental reading and language disorders and pragmatic language deficits. Compared with controls, some parents and siblings are socially reticent and others have for-preference fixed routines or difficulties with change in routines.
      Finally, autism parents, more often than controls, have anxiety disorders and probably recurrent mood disorders. In most cases, the onset of these conditions preceded the birth of their autistic child. These findings have been interpreted as possibly representing manifestations of some of the individual genes that are hypothesized to interact to cause autism. There has been one report of an association in parents between whole blood serotonin levels and a questionnaire measure of depression and a questionnaire measure of obsessive compulsive symptoms
      CANDIDATE GENES
      Any number of genes expressed in the central nervous system during fetal life and in the first two years post-natally could be candidates for causing autism. Because of the blood serotonin findings and the trophic role of serotonin in neural development, genes relating to serotonin and its metabolism and receptors are the genes that have been most carefully scrutinized. There are two reports of association between polymorphisms in the serotonin transporter gene and autism. In this gene, there is a common polymorphism in the promotor region that may have functional significance on transcriptional efficiency and serotonin uptake. While both studies reported convincing p-values, one found an association with the longest repeat sequence, and the other with the short sequence, so at this point the findings are not easily interpretable. Studies of some other serotonin-related candidate genes have been negative.
      A French group, in the course of studying several candidate genes involved in monoaminergic metabolism (which yielded negative results), also tested for association between the c-Harvey-ras (H-RAS) gene and autism with positive findings, which they have replicated . While HRAS is best known as an oncogene, it also plays a role in neural development.
      There is a fairly large body of literature, mainly by one research group, demonstrating a range of immunological and immunogenetic differences between autistic individuals and normal, ethnically matched controls. We know of no systematic study that has found a higher rate of autoimmune disorders in autistic children or their families. While some of the immunogenetic findings have been replicated by the same research group, it is not yet clear how they will be related other genetic and etiological factors. A recent report suggests a positive association between elevated blood serotonin and the major histocompatibility haplotypes previously associated with autism . In addition, one Letter to the Editor has suggested the possibility of some maternally transmitted risk factor. In a review of her clinic sample, Lord found that in families with two affected children, the second born almost always had a lower IQ than the first born. No one has published an attempt to replicate this finding, which suggests the possibility of an interaction between a genetic vulnerability and a maternally transmitted risk factor, perhaps immunogenetic in origin.
      CHROMOSOMAL ABNORMALITIES AND GENETIC LINKAGE STUDIES
      Over a number of years, there have been reports of associations between autism and chromosomal rearrangements, visible on karyotype, involving many chromosomes. Most frequently reported are deletions and duplications involving Chromosome 15q . Such findings suggest that there may be one or more gene abnormalities in the area of overlap of these rearrangements. This is being tested using genetic linkage studies of families who have two children with autism. Prior to screening the entire genome for autism-related genes, several groups first investigated markers on chromosome 15. The three collaborating sites of the Collaborative Autism Project (CAP) have independently found evidence suggesting micro-rearrangements of chromosome 15q markers in several families, perhaps as many as 10 percent of those screened. There is some suggestion that these micro-rearrangements are maternally inherited (CAP, unpublished).
      One group in the collaboration has reported a possible genetic linkage using affected sib pairs, to chromosome 15 markers distal to the area implicated in the Prader-Willi and Angelman's syndromes, and just distal to the genes for some of the components of the GABA receptor family . This finding has been confirmed using a different statistical method . These investigators reported multi-allelic disequilibrium in one of the GABA receptor genes using "trios," two parents and one affected child. This gene cluster is just proximal to the area of linkage reported by Pericak-Vance et al. and the markers showing micro-rearrangements found by the CAP group; markers nearer the GABA locus did not give positive lod scores. This region is currently under intense scrutiny.
      The European Collaborative Linkage study of autism has completed their first preliminary genome scan and reported a possible linkage between autism and an area of chromosome 7. The study did not, however, find linkage to markers on 15q. The chromosome 7 finding has not yet been replicated, but several groups are analyzing this area of chromosome 7.
      While the several findings on Chromosome 15q 11-13 do not suggest precisely the same location, this is not surprising or unexpected at this stage of the investigations as it is very difficult to identify precisely the location of genes that have a small effect in an oligogenic disorder. Given the power needed to disentangle oligogenic conditions, the number and size of families studied so far is very small. The convergence of cytogenetic, linkage, and association data makes it highly likely that there is a gene in this area on 15q relevant to autism.
      CONCLUSIONS
      Clearly, autism is most often a strongly inherited disorder, most likely caused by the interactive effects of several genes. While genes related to the serotonin system are likely candidates, the findings to date are unclear and more studies are needed. As the affected sibling pair studies of the complete genome are completed, it is expected that several more loci will be implicated. If some of these are genes of small effect, replication may be difficult. There is converging evidence, based on several different kinds of studies (chromosomal, linkage, and association) of a locus on chromosome 15q 11-13. Currently, several research groups are working to localize the area of interest to an interval that is small enough to study more closely for relevant genes


      The Authors:
      Dr. Susan E. Folstein is Professor, Psychiatry at Tufts University School of Medicine.
      She can be contacted at New England Medical Center, 171 Harrison Ave., Boston, MA 02111, USA. tel: 617-636-5734, fax: 617-636-8318 e-mail: susan.folstein@es.nemc.org
      Dr. Jonathan Haines is Associate Professor, Molecular Physiology & Biophysics, and Director, Program in Human Genetics at Vanderbilt University.
      Dr. Susan Santangelo is Assistant Professor, Psychiatry at Tufts University School of Medicine, and Adjunct Assistant Professor in Epidemiology at Harvard School of Public Health.