M.A. Spence

Response to NIH Questions

Question 1: What is the appropriate framework for studying autism? Are there multiple genetic loci and would this present insurmountable obstacles to research? The consensus is that there must be heterogeneity (different genes being responsible in different families) within the autism spectrum disorders and this will undoubtedly make the search for etiologic factors more difficult. However, the goal is obtainable. Even if numerous loci contribute to autism and the sibling relative risk is therefore much lower than reported, it would take possibly 400 sib pairs and 300-400 marker loci to map the gene. That is a sample which could be achieved with international cooperation and an NIH-directed effort.

Question 2: Are there genetic models and/or genetic techniques that have been used successfully with other developmental disorders than may be applicable to autism? Family studies of affected pairs of relatives is definitely the method of choice for the time being as it avoids having to workup and classify borderline and problematic cases. Actually locating the responsible genetic loci after mapping to a general gene region is a very difficult task. This work would be aided by association tests and linkage disequilibrium and therefore these tests are also applicable but must be applied in special populations to be most informative. It is essential any time family data are collected to consider, ahead of time, the restrictions imposed by genetic methodology and whenever possible take those into account when designing the studies. Epidemiology has provided sufficiently accurate estimates of the prevalence for the genetic analyses but also contributes by providing quality data on subtypes and comorbid conditions. Note also the query under Unresolved Issues regarding evaluating the need for a large twin study. (See recommendations for specific research below.)

Question 3: Do genetic and environmental factors act through common mechanisms to trigger the pathophysiology associated with autism? It is not premature to investigate gene-environment interactions. In fact, there was a strong consensus at the meeting that there must be relevant environmental factors even in the face of the genetic evidence. Even monozygotic (MZ, identical) twin pairs are not always concordant for autism (do not both always have or not have autism). Immune irregularities also suggest a role for pathogens, and findings of minor physical anomalies suggest a delay or disruption in early development. Given the complexity of autism from a clinical neurologic perspective, it appears highly likely that there is a common pathophysiologic sequence that is triggered in various ways by epigenetic and/or environmental factors. Clear identification of subtypes and rigorous studies on defining comorbid conditions will be a major first step in research in this area. Longitudinal studies are also an essential means of obtaining critical information regarding gene-environment interactions. Additional research on environmental causes or precipitants is clearly warranted.

The natural course of the autism spectrum disorders and the early predicators of later diagnosis are not yet well identified and understood. There is a sense that studies in these areas are making progress and should be encouraged for a variety of reasons. An important contribution of these studies will be a better understanding of the developmental stages and critical times in the course of the disorders. This will be invaluable in understanding gene-environment interactions.

Question 4: How can the genetic basis for autism be confirmed and further identified? We are definitely ready to test gene linkage hypotheses by initiating a formal genome search focused primarily on multiplex families (see below). However, there is no reason why these genetic family studies could not also serve as the primary vehicle for obtaining all the essential clinical and treatment data possible on the affected individuals and common relevant variables on the relatives. This procedure would avoid the expense of mounting both genetic and other studies and also improve measurably the quality of the genetic analyses. This multidisciplinary approach is exactly the one that has proven so effective in other complex diseases (such as breast cancer) which have seen quantum leaps in knowledge in the past couple of years.

Recommendations of the Working Group on Etiology

The search for the etiology (underlying causes) of the autism spectrum disorders is intertwined with research on diagnosis, pathophysiology, and treatment. Information from each of these areas helps to point the way toward possible causes. In turn, each of these other areas awaits discovery of the biological marker(s) for autism needed to expand and confirm its own findings. The working group on etiology recommends the following research priorities.

1. Genetic Analyses. There was remarkable consensus at the meeting that autism is a genetic condition. Mapping studies should be undertaken to identify the genetic loci that contribute directly to the disorder. The familial relative risks are sufficiently large to indicate the action of genetic factors and estimates of the number of loci involved are on the order of 3-6. For those reasons it was suggested that studies be initiated using affected pairs of relatives methodology (probably sib pairs, i.e., pairs of siblings). The information from the Human Genome Project, namely, the human fine-resolution genetic map, is exactly the required information to plan and carry out a successful genome search for loci contributing to the autism spectrum disorders. In addition, the parallel development of designated experimental organism maps and their sequencing will also contribute if/when appropriate animal models for specific aspects of the spectrum are developed. However, there are several important concerns and issues to be addressed if this research is to be sufficiently rigorous to have a reasonable chance of success, especially in view of the expected genetic heterogeneity (different genes being responsible in different families): (a) A very strictly applied set of diagnostic and sociodemographic criteria is essential in selecting individuals for these studies. Research on standardization of screening and diagnostic techniques and definitions of subtypes is directly relevant here. (b) The best strategy for ascertainment (identification of subjects) is to focus on multiplex families (more than one affected individual in the same family who independently meet criteria for diagnosis) to minimize the problems with uncertain or borderline cases. (c) From the beginning, the available sample should be split into two subsamples, one for detecting the loci and another for validating the results. These families are sufficiently rare that if most are inadvertently used to detect the linkage it will not be possible to confirm the results without an undue time delay. (d) Careful consideration must be given to the design of the study because of the vast amount of work necessary to have sufficient genetic markers to complete a thorough genome search (which is required in the absence of good candidate genes). Therefore, the design will need to carefully weigh the three points above as well as the possibility of collecting parents and single cases (trios) for haplotype relative risk analyses which will be essential in finding the specific loci responsible after identifying a region of the genome through linkage analyses.
     
2. Family Studies. It is important to emphasize that studies of family members have roles in addition to the linkage studies discussed above. For example, the gender ratio difference in autism is striking and families of females with autism may provide clues for understanding this difference if carefully studied. The following are other possibilities, but not an exhaustive list: (a) Geneticists should be included in planning any family studies since some (not all) of the more rigorous genetic analyses require that families be identified in a manner that can be specified in the likelihood equations. This ascertainment must be defined before the families are selected for study. By including this prior planning, families will be eligible for inclusion in the genetic studies as well as providing data for other purposes. (b) Family studies provide a unique opportunity to test whether or not the defined subtypes (e.g., clinical, drug response, language acquisition) also point to detectable differences in siblings and parents, recurrence risk (genetic subgroups), or other important features. Several areas of research, including neurochemistry and language studies, already have data indicating that some but not all families have nonautistic members who also display detectable differences when studied.
     
3. Epidemiological Studies. Current estimates of prevalence of autism are sufficient for the genetic analyses and no further precision is required at this time. However, the epidemiology approach would provide invaluable information in the definition of subtypes, comorbid conditions, and documentation of the range of variable expression of all of the spectrum disorders through the correct sampling and statistical analyses of the required data. Possible environmental causes or precipitants of autism may also be revealed. UNOC-CAP data would be useful in this regard.
     
4. Statistical Issues. Throughout the conference, there were discussions about the need for statistical rigor in diagnosis, defining subtypes, identifying risk factors, designing studies, and determining sample sizes. All the points raised are essential for the quality of the data and directly affect the genetic studies. These issues are discussed elsewhere in the report (cf. Statistics below), but should also be considered in any discussion of genetics.
      
5. Animal Models. Research in animal models, as with all good research, should be hypothesis driven (i.e., designed to answer a specific, testable question). However, there are now several good reasons why time should be spent considering appropriate animal models with the possibility of using them to move the research forward more quickly. For many of the biological variables, and quite possibly for the behavioral variables (such as cognition), study in animal models permits rapid breeding schemes which lead directly to estimates of heritability and number of involved loci. Added to that now is the direct comparability of the genetic maps among organisms (e.g., mouse and man) which facilitates the identification of genes in an experimental organism and their immediate location in man. There are two recent examples of this approach. First is the cloning of an obesity gene in the mouse and the identification of the human homolog the same day by computer search. Although no one knows the role of this locus in human obesity, there is now a specific candidate gene for etiology and pathophysiology studies. Second is the recent request for applications for studies to develop the genetic map in the rat issued because the investigators in hypertension are very close to mapping a number of loci that have significant effects in the different forms of the disease. The hope again is to move directly from the rat results to test for the importance of the homologous regions in the human genome.
      
6. Unresolved Issues. There are several additional areas of possible future research that have been identified in the discussions of directions for genetic studies: (a) The role of immune factors in the autism spectrum disorders is not resolved and warrants sufficient studies to clarify the situation. There may be indications that serotonin level and the immune response are correlated and this should be confirmed or denied as soon as possible. (b) MZ (identical) twins provide a unique experiment since they must share all their genes but are well documented to differ in important environmental factors including prenatal and developmental stochastic processes. The comparison of twins either discordant (one with autism, one without) or concordant (both with autism) could provide insight into both genetic and environmental factors of importance. An issue for discussion is whether or not it is cost-effective and scientifically useful to mount a large epidemiologic twin study to identify a sample of MZ and DZ twins for study. There are very few twin pairs in the literature, most from European studies and single case reports (U.S.). Should the study be done in the United States or is an international study needed? (c) Numerous cases of autism are reported with chromosomal variants (trisomy, translocations, etc.) but many of these were identified before the high resolution banding studies were available and before breakpoints could be cloned and uniquely identified. These cases should be collected and the newer studies performed to assess whether or not there are specific chromosome changes involved with the clinical features of this disorder. (d) Linkage disequilibrium is utilized via haplotype sharing and association studies to more precisely pinpoint the location of genes than is possible through linkage studies. To successfully apply this approach it is necessary to utilize populations that are genetic isolates or those known to have descended from a few founders (i.e., Finland). Are there identifiable populations that fit these requirements in which autism spectrum disorders are documented to occur, and where the prevalence of the disorders would warrant initiating such studies?

 first | previous | next | last