It's not clear to me whether you're understanding several important concepts. I'll try to explain a few things, if only for other people.
First of all, geneticists have a very confusing lingo where when the word "allele" is used, that can refer to either a particular single nucleotide (SNP) on a gene strand, or to a continuous sequence of any number of nucleotides that also includes that SNP. The same name is used in both cases (e.g., "A-allele"). For purposes of this article, we're mostly talking about the latter. It'll be clearer later below.
Second, it's critical to understand that SNPs within the coding region of the gene for an enzyme can potentially not only affect the functioning of the resultant enzyme that's made, but also affect the quantity of mRNA that is expressed and thus the
amount of enzyme that's actually made. For purposes of this article, we're only interested in the latter.
So for example, in the Abstract, where the word "functional" is used, they're not referring to SNPs that have an impact on how well the enzyme functions, rather they're referring to SNPs that affect expression, and therefore how much enzyme is produced. The quantity of enzyme made can be as important, or more important, than how well the enzyme functions.
This particular gene, TPH2, is especially different than others you may have looked at, like MTR, CBS, MTHFR, etc. With this gene, the authors are not interested in SNPs that affect the function of the resulting enzyme (like, e.g., MTHFR C677T), because there are very few of those on this gene. Rather, they're interested in SNPs that affect expression of mRNA, and therefore how much enzyme is produced. There are lots of those on this gene:
Extensive DNA sequencing of the TPH2 gene has revealed that polymorphisms that change the amino-acid sequence of the TPH2 protein are rare.7, 8, 9 The focus of research has therefore, now changed to identifying genetic variants that influence the TPH2 gene expression.
Now here are some additional comments within what you wrote to explain some things:
Actually they're just saying that the "A" allele shows up more often than the "G" allele in the mRNA of corpses who are heterozygous.
This isn't the most important issue, but from this, and what you write later, it appears you might be thinking that the term "expression" is referring to expression of the SNP by itself, when actually it's referring to expression of a multi-nucleotide allele that the SNP is included on, and from which the amount of mRNA it produces provides a means, using allelic expression imbalance (AEI), for determining the relative amounts of enzyme that each SNP/allele is likely to cause to be yielded upon transcription.
This doesn't necessarily indicate any overall increase in TPH2, just the ratio of the two expressed versions in heterozygotes.
Finding the expression ratio of the two different alleles (again, not the SNPs alone) is essentially the entire purpose for doing the AEI assay. It's not a flaw. It gives a way of determining which allele variant is likely to produce more (or less) enzyme. Additionally, note that AEI has been found to have a very high accuracy, precision and sensitivity.
Also, it was a range of ratios (1.05-2.5), with an average of 1.74, not 2.5. While one patient did have 2.5, the rest were necessarily lower, and it's important to reflect that accurately.
I did say "up to 2.5 fold."
The graph at
http://www.nature.com/mp/journal/v12/n5/fig_tab/4001923f5.html#figure-title actually does compare AA to GG+GA and indicates that while AA does result in more mRNA expression on average, the GG+GA range is wider than the AA range on both ends, and there's a pretty big cluster of the GG+GA samples in the same area where all the AA results are - hence it would seem that there is a factor other than those alleles which is likely having a big impact.
That's not exactly the correct conclusion to be drawn. The correct observation is that there are many SNPs on this particular gene that effect expression (and consequently how much enzyme is produced), and so naturally you might expect to see that type of pattern.
As mentioned above, the SNP does not result in a mutation, despite being in a coding region (exon) of the gene.
The researchers weren't interested in looking for mutations that might affect the functioning of the enzyme. Rather they were looking for SNPs that can affect how much enzyme is produced. This is a major focus of research now because many genes in the brain have SNPs that are more likely to affect the quantity of the enzyme produced as opposed to the functioning of the enzyme.
Which would raise the question as to how "A" might be overexpressed in mRNA compared to "G". They looked at how the DNA expresses itself when injected into hamster cells, and there weren't any significant differences between rates of "A" and "G" appearing in mRNA, nor in rates of degradation, which potentially could have indicated that there was more A in the mRNA because it took a bit longer to break down.
Eventually they theorize that the "A" allele creates a nicer splicing site which results in that exon being included more often than when there is a "G" allele. And that might not have any effect on the TPH2 enzyme, if having the proteins from exon 7 added to it isn't having any impact on the enzymes ability to function as required, and at the normal rate.
Again, they're not looking for a potential impact on the function. They're looking to see if more, or less, enzyme is likely to be produced from one allele versus the other:
The goal of this study was to determine whether allele-specific mRNA expression of TPH2 gene occurs and, if so, identify cis-acting genetic elements that predict high or low levels of expression.
And finally I'd point out that they refer to the "A" allele as representing a gain of function over the ancestral version, and they suggest that this mutation has persevered and become more prevalent because it might offer some reproductive benefit. Hence they really don't offer any suggestion that it might result in any counter-productive effects.
The potential counter-productive effect is readily apparent once you understand that "gain of function" is referring, not to a gain of function in the enzyme, but rather a gain of function of mRNA expression and therefore a greater quantity of enzyme being produced.
The enzyme, tryptophan hydroxylase 2, performs the first and key step in the production of serotonin. Having too much serotonin (= e.g., fatigue) can be just as bad as not having enough (= e.g., depression).
Anyhow, pretty interesting research, but no definitive indication that the "A" allele affects the actual functioning of the the gene's product, especially in any negative manner.
It should be clear now that they were not interested in looking for that.
I also wouldn't pay much attention in general to any research correlating SNPs with personality traits or psychological disorders - that stuff usually has tiny effect sizes, too-high p-values, and a lot of contradictory findings.
I wouldn't be so dismissive. Seemingly contradictory findings may be the result of a complex interaction between multiple SNPs on the same and different genes, including SNPs that affect function and others that effect quantity of enzyme (as in this study). As we understand those interactions useful trends can emerge.
Anyway, I think that the bottom line here is that, until proven otherwise (i.e., unless it's shown, e.g., that rs7305115 is not in linkage disequilibrium with another SNP that is the true "functional" SNP), that given two otherwise genetically (and epigenetically) and environmentally identical human beings, this study has demonstrated that an individual who is homozygous AA will have a greater capacity to produce serotonin than an individual who is GG. And this might conceivably, in combination with other SNPs or factors, potentially contribute to tipping the balance to a pathogenic state.
Lastly, I've had a thought that you might possibly need to revisit your analysis of Yasko's CBS C699T. I've never looked at that SNP, but if there's research showing that the homozygous state Yasko claims is bad results in an increased amount of mRNA/enzyme, it's obvious the potential could exist for it to appear good with respect to lowering homocysteine on the one hand, but bad for sending too much material down the transsulfuration pathway on the other.