A surprising METamorphosis: Autism genetics finds a common functional variant
Autism is a devastating neurodevelopmental syndrome characterized by fundamental impairments in reciprocal social interactions, delayed and deviant language development, and restricted interests and/or highly repetitive stereotyped behaviors. Approximately 13 in 10,000 individuals are affected. There are several related clinical disorders, including Asperger's syndrome and pervasive developmental disorder not otherwise specified, which, along with autism, are collectively known as autism spectrum disorders (ASDs) and increase the prevalence estimates to ≈60 in 10,000 (1). Although ASDs have been shown to have among the greatest heritability of any neuropsychiatric syndrome (1), specific genes causing or increasing the risk for social disability have been extremely difficult to identify. Nonetheless, over the past several years, a transformation has begun: After two decades of frustration, rare functional variants contributing to a small number of ASD cases have been identified (2), a convergence of linkage data from different studies and patient samples has begun to emerge (3), and promising results have recently been reported from association studies of common genetic polymorphisms (4). In addition, there have been dramatic advances in the understanding of the neurobiology of single-gene syndromes, such as Fragile X, in which affected individuals may also display an autism behavioral phenotype (5). In this issue of PNAS, Campbell et al. (6) now take the next step in solving the puzzle of ASD genetics. These investigators report a large-scale, methodologically rigorous candidate gene-association study and provide exciting genetic and molecular biological evidence pointing to a common functional variant in the promoter of the MET gene as the contributing risk factor. The identification of an association of a specific common functional allele, if replicated, would be an unprecidented advance in autism genetics.
The present study is notable for a variety of reasons, not the least of which is the rigor the authors have applied to addressing the range of methodological issues that have plagued gene identification in complex disorders. Across all of medicine, the major approaches for identifying loci that contribute to disease risk in a complex fashion have proven less powerful than originally anticipated. Candidate gene-association studies have been particularly vexing. The methodology is practically and intuitively attractive: Some version of a case-control study is performed with sequence variations in and around a gene of interest representing the “exposure” and affected status representing the outcome. If one investigates common genetic variations, relatively small risks can theoretically be identified with reasonable sample sizes regardless of the mode of inheritance. However, despite these advantages, in practice, the replication of findings from such studies has been the exception rather than the rule (7). There are likely many reasons for this fact, including but not limited to difficulties matching cases and controls for ethnicity, challenges in conceptualizing appropriate corrections for multiple comparisons, insufficient sample sizes in the face of tremendous heterogeneity, and varying power to detect associations related to the particular characteristics of the unknown functional allele compared with the known marker(s) being tested (reviewed in ref. 8). In psychiatric genetics, the additional challenges of diagnostic uncertainty, shifting syndrome definitions, and the general absence of biologically based diagnostic indicators further complicates matters.
Candidate gene-association studies have been particularly vexing.
In their work, Pat Levitt and colleagues (6) address many of these challenges: They start with a relatively large-scale study of association, build in an internal replication, take advantage of the data on linkage disequilibrium (LD) in the genomic interval to evaluate LD blocks as well as single-nucleotide polymorphisms, use appropriate methods to control for population stratification (the presence of occult differences in ethnicity in cases vs. controls that can lead to spurious associations), and consider the contribution of varying diagnostic categories and family types (i.e., multiply vs. singly affected) to their findings. Finally, and perhaps most importantly, the authors provide evidence narrowing in on the risk allele itself: a single-nucleotide substitution just 5′ to the transcription start site of MET (Fig. 1).
Campbell et al. (6) report the association of a G>C common promoter variant (green vertical bar) 20 bp 5′ to the transcription start site of MET. In vitro analyses show that the substitution leads to a 2-fold decrease in transcription and reduced binding affinity for SP1 and PC4 transcription factors. The MET transcript codes for a receptor tyrosine kinase, composed of α- and β-subunits, which form a homodimer capable of binding to the ligand, hepatocyte growth factor/scatter factor (HGF/SF). HGF binding results in phosphorylation of multiple tyrosine residues located within and outside the intracellular kinase domains. The downstream signaling cascade has been shown to involve a wide range of docking molecules, and MET activation results in pleotrophic effects, including those listed on the figure. The manner in which decreased expression of MET mRNA may lead to autism spectrum disorders (the cardinal features of which are shown in the blue circles) is not known.
At least at first blush, the identity of this gene, a receptor tyrosine kinase known for its role in metastasis, would seem something of a surprise. Indeed, sorting out how an allele that, in vitro, confers a subtle decrease in the expression of MET mRNA might lead to a range of pervasive developmental disorders promises to be a most exciting scientific journey. Previous work has demonstrated abnormalities in brain development resulting from disrupted processing of the MET receptor's single known ligand, hepatic growth factor/scatter factor (HGF/SF), leading to decreased MET signaling (9–11). The attendant alterations in cerebellar and cortical architecture bear a tantalizing resemblance to hypotheses about the neuropathology of autism. However, the MET receptor is also highly pleiotropic and is known to serve as a docking site for multiple protein substrates (reviewed in ref. 12) that influence a wide range of biological processes, including immune function and gastrointestinal (GI) repair. In fact, Campbell et al. (6) point to these findings as a rationale for intensively studying the gene, given that there are some data suggesting increased difficulties in these areas in patients with autism (13, 14).
This possibility, that the MET variant might lead to immune dysfunction and GI disturbance along with ASDs, is an important question to pursue and one that will likely lead to some debate. These phenomena, along with developmental regression, have been at the heart of the hypothesized relationship between measles, mumps, and rubella (MMR) vaccine and autism. In the first description of this purported link, the authors focused on a group of children with inflammatory GI difficulties and a regressive form of autism in which symptom onset was tied to MMR immunization (15). Although subsequent evidence has not supported a relationship between this vaccine and autism (16–18), and the majority of original authors have retracted the interpretation of their report (19), there remains some public uncertainty about the safety of the vaccine. Moreover, the very important question of whether and how gut disturbance, regression, and immunological issues may be related has been obscured in part by the controversy. Parents of children with autism have long reported these types of problems, and, recently, investigations have provided evidence supporting these perceptions, at least with respect to GI symptoms and regression (20). However, not all studies have reached the same conclusions (21), and the relationship between these symptoms and immunological status remains an open issue (14, 20). Hopefully, the work by Campbell et al. (6) will lead to additional rigorous investigations of these questions without fueling unnecessary concern regarding MMR.
Finally, although this report is cause for considerable optimism, as with any initial association study, it is not yet time to declare victory. To be considered definitive, this finding will have to be replicated by independent investigators. One cannot help but imagine, however, the possibilities as the work is confirmed: The common functional variant may be used to stratify samples in neuroimaging, pharmacogenetic, and other intervention studies; the natural history of at-risk individuals can be evaluated with respect to GI symptoms, immunological status, regression, and a host of putative risk factors; and evaluation of the −20G>C allele in samples that have previously shown linkage to chromosome 7 may help clarify the status of this interval as it pertains to ASD. Finally, as the investigators elaborate the autism-related biology of the MET pathway, the prospect that these findings could identify novel molecular therapeutic targets offers an exciting glimpse of the next epoch in autism research.
Footnotes
- *E-mail: matthew.state{at}yale.edu
-
Author contributions: M.W.S. wrote the paper.
-
The authors declare no conflict of interest.
-
See companion article on page 16834.
- © 2006 by The National Academy of Sciences of the USA
