Natural killers

Three days ago, as I pondered my family's history of autoimmune disorders--MS, thyroid, asthma, allergies, arthritis--I began to wonder about a link between the heritable factors of autoimmunity and autism. One question I had was whether or not anyone had ever investigated an epidemiological link between autoimmune disorders in a family tree and autism. Some people have theorized that vaccines trigger an autoimmune response that leads to autism, but that leads to questions of why children who are not vaccinated develop it. It also would lead me to wonder why children who have the actual diseases for which we are vaccinated would not have a similar response. Perhaps it might be timing. But I digress into pure speculation.

There are some studies available on the link between autoimmunity and autism, such as this one and especially this one. The group out of UC Davis has just come out with another study (and it's OPEN ACCESS!), a broad genomic look at the genetic complements that associate with the presence of autism. They found that there were about 11 genes involved in immunity that were far busier in children with autism than in children without the diagnosis. They also found a division of subsets of some genes: in children who appeared to show onset at about 18 months typified by regression (autism with regression), there were almost 500 genes that differed in their activation compared to children who exhibited "early onset" autism. The researchers suggest that this finding indicates that there are two groups of children with autism whose disorder arises from different pathologies. Across the board, however, all of the children with autism shared the enhanced expression of those 11 immunity genes.

Briefly, the authors studied 17 children with early onset autism, 18 with autism with regression, 14 who were "ASD but did not meet the criteria for autism," and 12 general population children. Participants were age and gender matched. They found that 55 genes were upregulated in children with autism vs. the general population; 150 in children with "early onset" autism vs. the general population; 20 genes for late-onset autism vs. the general population; and 494 genes for early- vs. late-onset autism. They also found no significant differences in regulation between the ASD and general population children. They conclude that "the gene expression data support emerging evidence for abnormalities in peripheral blood leukocytes in autism that could represent a genetic and/or environmental predisposition to the disorder."

The study stands out because it relies on the idea of autism phenotypes to subdivide the study groups. In addition, its choice of an ASD-without-autism-dx is interesting. I find their reported results to be a bit confusing, but not for any reasons worth going into here. The big handwaver of a finding here would be the big 11 natural-killer-cell-related genes that were upregulated in both groups of children with autism relative to children in the general population.

The big 11 are genes involved in the activity of natural killer immune cells. T helper cells and other cells of T origin also have been linked to autism. It comes as no surprise, perhaps to some, that these cells also happen to be those involved in autoimmune disorders. The job of a natural killer (NK) cell is to go around poking holes in infected cells (infected with virus, bacteria, or parasite) or in tumor cells, leading to suicide of the affected cell. People who have deficiencies in their own "self' labeling molecules may end up having their cells targeted by these microscopic killers, leading the person's own cells to commit suicide. The role of NKs in autoimmunity may vary. In multiple sclerosis, for example, they may either be doing their job quite well during times of remission, or not doing their jobs very well at all in association with flare up. Some MS drug development targets boosting NK cell activity.

There are some caveats associated with the current study. First, it was small (61 children among four groups) and is considered preliminary, like pretty much any genomics study would be. As someone who has done such studies, I can tell you that significant changes in gene expression can lead to many apparent dead ends when we home in on the genes and try to identify a mechanism of involvement. Sometimes, genes upregulated for reasons that are not even remotely proximal to mechanisms of disease. So it goes with genomics studies.

This group selected the whole blood on purpose, interested in examining the genes of the various cell types that occur in the peripheral blood. It is notable that the NK genes stood out so clearly. That implies a path worth pursuing.

However, the findings of 500 genes upregulated or downregulated in children with "later-onset" autism strike me as more equivocal. There could be any number mechanisms involving neither cause nor effect to explain the difference, if there is one; see below. Developmental timing of gene activation could differ between the two groups, leading to this difference in a good-sized suite of genes without necessarily producing notable differences in outcome.

My concern is their cutoff for -fold regulation change in the genes. They used a cutoff of 1.5. In studies I've reviewed (as a peer-reviewer) and that I've participated in myself, the standard cutoff is usually 2 or slightly more, although this CAN vary as low as 1.5. Indeed, in the paper itself, some of the genes they highlight did not exhibit even a -fold change as high as 1.5 (see Table 1, where two highlighted genes show a -fold change of just over 1.3). The only genes that show the 2-fold cutoff are among those 11 NK genes. From my understanding, generally, a 2-fold cutoff is used unless there appears to be a suite of genes from the same class exhibiting a change in regulation of at least 1.5-fold. It's cool to drop the cutoff just to see what's out there--kind of an interesting fishing expedition--but genes upregulated at that level are by no means necessarily key players in the mechanisms of interest. They applied the statistics necessary to make this defensible. Duh, it is Genomics, after all.

A close look at the Supplementary tables and data shows that in the upregulation category between children with early onset vs. regression-type autism (Supplementary table 4), only three genes meet the >/= 2-fold cutoff for change in regulation, and that is for upregulation only. The researchers found no significant downregulation of genes in a comparison of these groups in this context, with the highest fold change they present at 0.66 (reverse for 1.5 cutoff) between these two groups and a comparatively short list of genes. I find that kind of strange.

I can buy the idea of roughly drawn "autism phenotypes" in terms of timing of early onset and regression (at least as far as our fuzzy perception can tell), but just as development can time us to respond to specific internal switches at certain timepoints--puberty comes to mind as an obvious example--these distinctions may arise from temporal differences in flipping the switch, rather than some exogenous influence. The switch is still no different. Some girls experience menarche when they're 10 (even before environmental hormones), while others do so when they're 17. Some children may manifest autism symptoms "from the beginning" while others don't experience the flipping of the switch until later. I remain skeptical that we're as good as we hope we are even at identifying the earliest signs of autism.

I'll be looking for links between the NK cells and distal effects on dopamine pathways in the brain. Yes, you read it here, possibly first? Dunno. What I won't be looking for is that somehow, a vaccine makes NKs go more cacacoocoo over cells than an actual disease does. At least, I don't expect to see that. These 11 NK genes were present in both "phenotypic" autism groups, early onset or regressed. If there's a link, then obviously, vaccines don't provide an in utero environmental trigger. I muse, however, on the possibility of a pathogen-related trigger. But I only muse. It would have to be one sneaky pathogen.

There are a lot of data here to digest, a lot of potentially involved genes to characterize, a lot of pathways and mechanisms to elucidate. This is a great start, although genomics studies are naturally unwieldy and complex to interpret. It's going to be awhile before researchers wade through this mountain of genomics data and characterize the pathways that matter.