“Epidemiological studies of autism spectrum disorders have failed to identify definitive evidence of exposures correlating to increased risk of the disorder”—such was the state of the science as described in a 2008 American Journal of Human Genetics editorial.1 The author was not alone in expressing his skepticism that autism may ultimately be linked to nongenetic causes. However, the absence of associations does not constitute proof of no association. Indeed, autism research has slowly (perhaps reluctantly) shifted from a solitary focus on genetics to a growing acceptance that environment must play a role. In this issue, Schmidt and colleagues2 demonstrate a complex interplay between genes and environment in the etiology of autism, adding to an accumulating body of evidence implicating the folate pathway in at least part of what is likely to be a multicausal phenotype. The reported effects, though often imprecise due to small cell sizes, are notable for a number of reasons, including (1) the exposure effects are time-dependent, observed in the periconceptional window only, and (2) the effect of genetic variants (if any) is masked by prenatal vitamin use (ie, the gene effect is apparent only in a nutrient-poor subgroup).
The concept of “critical windows of susceptibility” has a rich history, dating back at least to James Wilson's3 principles of teratology—the second of which states that the susceptibility to teratogens varies with the developmental stage at the time of exposure. Schmidt et al report that prenatal vitamin use in the periconceptional window (3 months before to 1 month after conception) is associated with a strong and dose-dependent reduced odds of autism (reduced by approximately 40% for those who took prenatal supplements at least 4 times per week). Interestingly, exposure to the antiepileptic drug valproic acid, a folic acid antagonist,4 during the prenatal period appears to increase risk of autism or autistic traits in both humans4–6 and rats,7,8 with embryonic day 12 emerging as the critical window of exposure in rats8—corresponding to the period of neural tube closure (in humans, gestational week 3–4). Exposure to valproic acid during early pregnancy also has been shown to increase risk for neural tube defects,4 the rates of which were strikingly reduced in the United States and abroad following folic acid fortification.9
These themes coalesce in a growing body of research suggesting that the fetal epigenome is sensitive to chemicals, which may carry consequences for childhood developmental dysfunction.10,11 Valproic acid, cadmium, arsenic, and organochlorine pesticides are documented examples of environmental agents that induce epigenetic transformations including hyper- or hypomethylation and histone deacetylation inhibition.9,12–16 This can result in a cascade of events that alter gene expression during critical periods of development or diminish glutathione activity, leaving cells vulnerable to oxidative stress.17 Folate is a critical factor in one-carbon metabolism, providing the methyl donors required for most methyltransferases that modify the epigenome.17,18 Thus, an interesting and intricate picture has begun to emerge that argues strongly for a re-examination of the emphasis on genetic risk factors for this complex and enigmatic disorder.
Genetic Effects Masked by Prenatal Vitamin Use
It cannot be disputed that genetics is likely to play an important role in the development of autism.19 However, it also appears unlikely that genetic features will take the form of common variants with large and independent main effects. And, in fact, among the genome-wide scans published to date, very few loci have reached genome-wide significance (and none has been replicated).20 Failed replication likely reflects a complex polygenetic and poly-environment web that contributes various traits to the autism spectrum, not necessarily through one sequential pathway. Thus, the search for genetic causes has become increasingly focused on rare mutations and copy number variants, some of which may be singular within related individuals in selective family studies. Meanwhile, in deft pursuit of ever smaller P values, arguments have been made for a more refined stratification of the autism phenotype in the hope that genetic associations will be enhanced in more selected case groups.1
Schmidt and colleagues2 present a persuasive alternative: common genetic variants may be manifest only in susceptible subgroups in a population-based sample. The authors report superadditive and supermultiplicative joint odds ratios for the combined effect of multiple variants in the folate pathway and lack of periconceptional prenatal vitamin usage, albeit odds ratios that are imprecise and probably somewhat inflated given the small sample size in the doubly exposed group. But just as importantly, there were no effects among those who did take prenatal vitamins in this period. Additionally, the crude main effect of these variants was essentially null. Thus, a screen for genetic main effects (GWAS or candidate gene) would have missed these important findings. Moreover, no special manipulation of the case population was required. The findings applied equally well to autism spectrum disorders as they do to frank autism, enhancing the generalizability of the findings. These results further support the argument that the prenatal environment is subject to both child and maternal influences. The autism literature is sprinkled with reports focused on child's genes; more attention to maternal genotype is clearly warranted when studying the perinatal environment.
THE EMERGENCE OF EPIDEMIOLOGY IN AUTISM
The concept that genes and environment may interact, without obvious genetic main effects, is not novel in epidemiology. However, this concept is surprisingly under-represented in the autism literature, which until recently has existed in silos of journals focused on either genes alone or environment alone. The field of epidemiology provides fertile ground for the challenging, interdisciplinary collaboration needed to move the field forward. Although prospective studies (some of enriched-risk cohorts) are still in their infancy, they will, with enough patience and resources, ultimately provide rich data on the timing of exposures and genetic liability within families.21–24 Two large, population-based studies, the Childhood Risk for Autism from Genes and the Environment (CHARGE) Study22 and the Study to Explore Early Development (SEED),23 are currently collecting extensive data on perinatal medications, exposures, and genetics.22,23 These case-control studies have well-characterized study populations and hold promise for identifying important gene-environment associations deserving more intensive investigation; yet they must rely on retrospective data for perinatal exposures. That said, it is unlikely that any recall error would vary differentially by genotype. Environmental exposures can affect neurodevelopment in multiple ways, either through direct neurotoxicity, or by epigenetic events that change the composition and susceptibility of all developing fetal cells, possibly increasing susceptibility to specific environmental exposures.
To our knowledge, Schmidt and colleagues2 are the first to provide data illustrating the importance of characterizing genetic susceptibility to environmental exposures in the context of autism. The authors present an excellent model for investigating gene-environment interactions in retrospective case-control studies. The large studies currently collecting biologic and environmental data on sizable populations will have considerable opportunity to apply this approach, with the aim of capturing variability in gene expression in the appropriate person at the relevant critical window in neurodevelopment. Good, solid, epidemiology will continue to be a key player in unraveling the mystery of complex disorders.
ABOUT THE AUTHORS
STEPHANIE M. ENGEL is an Associate Professor in the Department of Epidemiology, Gillings School of Global Public Health, University of North Carolina, Chapel Hill, NC. She has been involved in research pertaining to genetic and gene-environment risk factors for pregnancy complications and pediatric neurodevelopmental disorders. JULIE L. DANIELS is an Associate Professor in the Department of Epidemiology, Gillings School of Global Public Health, University of North Carolina, Chapel Hill, NC. She has been involved in autism surveillance and ongoing case-control studies of perinatal risk factors for autism.
1.Stephan DA. Unraveling autism. Am J Hum Genet.
2.Schmidt RJ, Hansen RL, Hartiala J, et al. Prenatal vitamins, one-carbon metabolism gene variants, and risk for autism. Epidemiology.
3.Wilson JG, ed. Environment and Birth Defects. Environmental Science Series
. London: Academic Press; 1973.
4.Ornoy A. Valproic acid in pregnancy: how much are we endangering the embryo and fetus? Reprod Toxicol.
5.Dean JC, Hailey H, Moore SJ, Lloyd DJ, Turnpenny PD, Little J. Long term health and neurodevelopment in children exposed to antiepileptic drugs before birth. J Med Genet.
6.Moore SJ, Turnpenny P, Quinn A, et al. A clinical study of 57 children with fetal anticonvulsant syndromes. J Med Genet.
7.Ingram JL, Peckham SM, Tisdale B, Rodier PM. Prenatal exposure of rats to valproic acid reproduces the cerebellar anomalies associated with autism. Neurotoxicol Teratol.
8.Kim KC, Kim P, Go HS, et al. The critical period of valproate exposure to induce autistic symptoms in Sprague-Dawley rats. Toxicol Lett.
9.Berry RJ, Bailey L, Mulinare J, Bower C, Folic Acid Working Group. Fortification of flour with folic acid. Food Nutr Bull.
10.Perera F, Herbstman J. Prenatal environmental exposures, epigenetics, and disease. Reprod Toxicol.
11.Wilker EH, Baccarelli A, Suh H, Vokonas P, Wright RO, Schwartz J. Black carbon exposures, blood pressure, and interactions with single nucleotide polymorphisms in MicroRNA processing genes. Environ Health Perspect.
12.Kim KY, Kim DS, Lee SK, et al. Association of low-dose exposure to persistent organic pollutants with global DNA hypomethylation in healthy Koreans. Environ Health Perspect.
13.Marinova Z, Leng Y, Leeds P, Chuang DM. Histone deacetylase inhibition alters histone methylation associated with heat shock protein 70 promoter modifications in astrocytes and neurons. Neuropharmacology.
14.Smeester L, Rager JE, Bailey KA, et al. Epigenetic changes in individuals with arsenicosis. Chem Res Toxicol.
15.Wright RO, Baccarelli A. Metals and neurotoxicology. J Nutr.
16.Wright RO, Schwartz J, Wright RJ, et al. Biomarkers of lead exposure and DNA methylation within retrotransponons. Environ Health Prospect
17.Lee DH, Jacobs DR, Jr, Porta M. Hypothesis: a unifying mechanism for nutrition and chemicals as lifelong modulators of DNA hypomethylation. Environ Health Perspect.
18.Loenen WA. S-adenosylmethionine: jack of all trades and master of everything? Biochem Soc Trans.
19.Lichtenstein P, Carlstrom E, Rastam M, Gillberg C, Anckarsater H. The genetics of autism spectrum disorders and related neuropsychiatric disorders in childhood. Am J Psychiatry.
20.Devlin B, Melhem N, Roeder K. Do common variants play a role in risk for autism? Evidence and theoretical musings. Brain Res.
21.Early Autism Risk Longitudinal Investigation (EARLI). Available at: http://www.earlistudy.org/
22.Hertz-Picciotto I, Croen LA, Hansen R, Jones CR, van de Water J, Pessah IN. The CHARGE study: an epidemiologic investigation of genetic and environmental factors contributing to autism. Environ Health Perspect.
23.Study to Explore Early Development (SEED). Available at: http://www.cdc.gov/ncbddd/autism/seed.html
24.Stoltenberg C, Schjolberg S, Bresnahan M, et al. The Autism Birth Cohort: a paradigm for gene-environment-timing research. Mol Psychiatry.