In 2007, I made a link filled comment about autism at Colorado Confidential summing up the key points about what we know and do not know about the causes of the condition. I was going back over some of those links, and one of the papers I'd cited to earlier had been cited to by a number of other articles later.
Of particular interest was a paper in the Proceedings of the National Academy of Sciences (a very reputable scientific body) setting out a hypothesis that would comprehensively explain a large share of all autism cases. In a nutshell, here is what is proposed (except as noted this comes solely from the paper cited as explained by me as an educated layman):
1. People receive genes in pairs, one part of each pair from a father and one from a mother. (This statement defines the term "gene" as I am using it in this post.)
2. There is probably a single primary "autism gene" and this gene is probably not on either the X or the Y chromosome. A single copy of an autism gene in a boy results in a near 100% chance that the boy will have an an autism spectrum disorder (ASD). The exact manifestation rate depends upon the assumptions made and the model used. It is likely is that the chance of developing ASD if the autism gene is present is close to high end of that range, and that there is little or no environmental impact on whether or not a boy develops some form of ASD.
3. Other genes, called "modifier genes" determine, at least to some extent, how severely the autism will manifest in that boy, but do not prevent a boy from having some sort of ASD. Full fledged autism, therefore, as opposed to a more mild ASD, is due both to the presence of an autism gene, and the absence of modifier genes. There isn't enough data in this study to understand the nature of these modifier genes in boys.
4. It is also entirely possible that some modifier genes in boys are environmentally triggered, or the environmental can impact the severity, but the the mere presence or absence of ASD is not greatly impacted by environment. (This is from my own general understanding of this area of genetics.)
5. Sometimes a mutation in a parent's particular sperm or particular egg cell will cause a parent who does not have an autism gene to pass an autism gene onto a child. These mutations in autism gene free parents account for about two-thirds of all autism cases. About 6.7-7.7 in one thousand newborn girls are autism gene carriers with no autism symptoms, about 2.3-3.3 in one thousand newborn girls have an ASD and about 1% of newborn boys have an ASD.
6. These mutations most commonly appear in the sperm of older fathers, with the risk rising greatly by paternal age 40, when they are 5.75 times as likely as for fathers under age 30. Advancing maternal age is not linked to ASD after adjusting for paternal age. Advancing paternal age was also associated with some of the male-female sex ratio in autism. The sex ratio in the offspring with autism of fathers younger than 40 years was noticeably higher than the sex ratio of the offspring with autism of fathers older than 40 years. (This statement integrates data from a separate source.)
7. About 6.6 per 1,000 children whose parents both lack the autism gene (i.e. two thirds of children who either have autism or are carriers for the gene) receive one as a result of a mutation on a gender neutral basis, rather than inheriting it from a parent who has the gene.
8. The number of autism gene mutations rate per 1000 births to fathers in the appropriate age group is about 2.9 for fathers 15-29, about 5.7 per father aged 30-39and about 25.8 for fathers 40+.
These numbers from the the Israeli Israeli data in the "separate study" cited above (including a few details from outside the abstract available here), with some adjustments. I attributed part of each age group's to genes directly inherited from a parent by reducing the autism rate for each age group by the one-third of the total population autism rate. This provides a mutation rate for fathers 30-39 that is double that of fathers aged 15-29, and a mutation rate for fathers age 40+ that is roughly nine times that of fathers aged 15-29. Then, I fitted this data to the American autism incidence rate which is about 25% higher than in the Israeli study. In doing so, I used data from the National Vital Statistics Reports for 1980-2004 and census bureau population figures that shows that currently about 50% of newborns have fathers 15-29, about 40% have fathers 30-39, and about 10% have fathers age 40+.
9. My calculations above in connection with other information in this post, suggest that about a third of autism cases are due to non-age related mutations of in sperm or egg cells of parents who don't have an autism gene, about a third of autism cases are due to age regulated mutations in sperm cells of fathers aged 30 or more who don't have an autism gene, and about a third of autism cases are a result of inheriting an autism gene from a parent, usually a carrier mother or a mild ASD symptom parent of either sex.
10. My calculations also suggest that age based autism gene mutations are becoming more common due to the aging demography of fatherhood. In 1980, this model would suggest that 39% of mutatation based autism was attributable to parental age, in 1990, it had risen to 47%, in 2000 to 52%, and based upon the latest data available (2004 fatherhood rates and 2006 population figures), about 54% now. Put another way, older fatherhood can account for an approximately 33% increase in overall autism incidence rates since 1980 to the present, even disregarding entirely autism caused by girls who become carriers for autism because of a mutation in a parent's gene line and then have autistic children as a result. This is nothing near the epidemic increases in autism incidence that have been reported, but autism was only added to the standard diagnositic pyschiatric manual in 1980, and Asperger's syndrome was only separately added in 1994, so there are serious measurement and diagnosis issues associated with surging autism levels (which isn't to say that real prevalance isn't changing, only that we can't be sure from the data that we have now.)
11. It is possible that autism causing mutations in the sperm of older men reflect vulnerability to a sperm mutation only in certain situations, and that the mutation itself will not happen very frequently, in the absence of some environmental exposure experienced by the father that has the mutation. (This statement is based upon my own general understanding of and reading about how genetics and the environment interact.)
12. Girls can also have an autism gene. But, there is probably a "modifier gene" or "modifier genes" that prevent autism from arising at all in about 70% of girls. Thus, in girls, autism arises from both the presence of an autism gene and the absence of a modifier gene or genes.
13. The main modifier gene that protects girls from autism is probably on the X chromosome, and one probably needs to have two copies of this X based modifier gene to prevent autism from arising at all. About 70% of females (some studies would put the number closer to 77% which in turn would impact the other numbers) have two anti-autism modifier genes. About 84% of men, and 13.5% of women have a single anti-autism modifier gene. About 2.5% of women and 16% of men have no anti-autism modifier genes. (I have created in this paragraph the simplest model of a modifier gene that would fit the modifier gene model and data they described in the paper which was a matter that they did not explore. More complex mechanisms with similar effects are possible alternatives.)
14. We don't know whether or not the modifier genes that prevent autism when present in pairs in girls have any impact on the severity of an ASD in either boys or girls.
15. It wouldn't be surprising if the presence of a single anti-autism modifier gene was quite important in determining how severely an ASD would manifest (e.g. producing Asperger's syndrome rather that full fledged autism.) If true, one would expect only 12-16% of both boys and girls with an ASD to have pure autism cases, and would expect sibling studies and family studies to show patterns of heritability for severe v. less severe versions ASD in families with multiple cases. In fact, the percentage of ASD individuals with pure autism is on the order of 16%-25%, consistent with this theory, and the sibling and family studies do show the sibling and inheritance effects in autism severity when multiple people in a family have an ASD. (This theory is my own, although obvious from the suggestion of the paper.)
16. The anti-autism gene that prevent autism from manifesting in many women is probably not the only modifier gene relevant to ASD severity in either boys or girls.
17. About one-third of all autism cases involve a parent who has the autism gene and passes the gene they have onto a child. Usually, this parent is a mother with no autism symptoms, but more rarely, they could also involve a parent with a ASD condition, usually one that is mild or even subclinical (i.e. autism like symptoms too subtle to diagnose as even a mild ASD spectrum condition). In the usual case, where the mother has one copy of the autism gene, and the father has none, about half of the boys born to the mother (regardless of the father of the particular child) will have an ASD, and on average, about 15% of the girls born to parents with this genetic profile will have an ASD.
While this hypothesis is based only upon a medium sized study (a few thousand children with autism in family relationship), its relatively simply hypothesis is a compelling and plausible explanation consistent with inheritance patterns that are well established in many vaguely similar genetic conditions. We won't have proof positive until a particular gene is found in almost everyone with autism, but this study points a clear finger at what we should be looking for and explains neatly how it can persist despite the fact that it impacts the future reproductive abilities of those who have the gene and manifest its effects.
The first place to look for a cause of rising autism incidence is in paternal age trends. A rise in the percentage of fathers over age 40 would strongly point towards that as an important cause of rising ASD incidence. Moreover, this would happen in two ways. First, older fathers would be having more sons born with ASD than younger fathers. Second, older fathers would be having both more daughters born with ASD than younger fathers, but also more daughters carrying the ASD gene have have sons of their own who are much more likely to have ASD. Modeling the projected incidence rise based upon increasing numbers of older fathers would be quite a precise and easy exercise, that would make it quite easy to determine the magnitude of ASD incident due instead to either changes in diagnostic standards or changing environmental exposures, or both.
Second, if there is an environmental impact on ASD prevalence, exposures to parents and not to children is the place to look. There are multiple known environmental mutagens. If older fathers have had increased exposure to these mutagens, that would be a very plausible culprit.