The effect of SNPs at the same location

Question

[Petras]

Hey all,

I wonder if nucleotide changes at exactly the same location tend to have a 'one-directional' effect on gene expression. Let's say the reference allele is A. So we have 3 possible mutations A->G, A->C, A->T. How common is it that, for example, the first will increase gene expression, the 2nd will have no effect and the 3rd will decrease gene expression? I would expect such situations to be quite rare and more often than not the effect would be in one direction. By 'one-directional' I mean that if one mutation decreases gene expression, the other two will also decrease it or have no effect.

I'm trying to find papers describing something like this but I cannot really find anything. Any pointers would be much appreciated. Thanks!

Answer 1

[Istvan Albert]

I would say that there is no reason or rationale to expect that mutations manifest themselves in the way you describe.

First and foremost, most mutations that affect phenotype are in coding regions and cause differences (or not) in the protein sequence.

Outside of coding regions, mutations that appear in the same position may also indicate neutrality, there is no effect on the phenotype.

Trying to find a one-to-one correspondence i.e. there are three alternative bases and three directions a gene expression might change (same, up , down) hence each base might correspond to a direction sounds like an oversimplification of the biological processes

[bk11]

I m adding to your comment here, pls do not feel it as critics. Many polymorphisms are also in the non-coding (mostly regulatory) regions of DNA that can affect the expression of genes nearby or at distant sites. For example SNP in long non-coding RNA which is an promotor or enhancer can affect the expression of target genes.

[Istvan Albert]

Ah, yes, I fully agree and should have made that point as well.

As you correctly point out, there are many mutations that fall in regulatory regions and many will cause gene expression changes.

The point I was trying to make is that if one were to pick a mutation, the chance that it affects gene expression is probably smaller than that of all the other possible effects that I have enumerated.

Although that might not quite be right either - it was just a sort gut check, now I wonder if that is true.

Now that I thought about it more, if one were to be presented with a mutation and they had to guess its effect, it might go like this:

1. Silent mutations - no effect
2. Synonymous substitution
3. Non-synonymous substitution
4. Regulatory effect (like gene expression)
5. Frameshift mutations
6. Nonsense mutations ( really dislike this naming ... "nonsense" mutation ... should be called "stop" mutation)

perhaps the correct ranking is different - I wonder - and I might be missing some categories here

[bk11]

Further more we should also keep in mind that human DNA in not linear, it has a complex folded structure. For example in a specific folded locus many factors/proteins come to play role for maintaining the folded structure called a topological domain (TAD). The concerted role of TAD will be regulating the expression of genes. Having said that SNP in regulatory (usually non-coding) region can have promotor or enhancer activity. The promotor or enhancer activity of a particular SNP is usually cell type specific.

[LauferVA]

i agree that OPs post is simplistic - however - there are some cases in which it will likely end up being true despite itself, so to speak. consider two examples:

Transcription Factors have consensus binding sites, which depending on the TF and context, could produce a scenario in which substitution of an adenine residue with G, C, and T, could produce the exact pattern that OP describes. I agree with Istvan Albert that this likely accounts for a minority of cases, but there are also reasons in which it is true.

Another (conceptually related) example might be SNVs that affect either miRNAs or the binding sites to which they can anneal.

"most mutations that affect phenotype are in coding regions and cause differences (or not) in the protein sequence."

i agree that protein biology furnishes some of the strongest examples of single-nucleotide based changes producing phenotypic effects. but, i am worried about ascertainment bias, here - so I also want to highlight changes in other regions as well. here are a few examples:

• interestingly, though there are families of diseases that characteristically arise from variation in enhancer regions (for example, autoimmune disease, about 90% of disease risk loci map to noncoding regions)
• SNVs in the near intron that knock out a splice sites can produce effects as severe as frameshift or nonsense mutations. because we finally have the tools to investigate things like perturbation of TADs, enhancer genotypes, epistasis between promoter and enhancer variants, im very excited that going forward we will start to know more about other the phenotypic effects of SNVs outside of protein coding regions.
• i like that you highlighted synonymous variants, at least. for other readers - it is not true that synonymous variants are "silent" in all (or even most) cases. about 8% of cancer driver mutations are thought to be synonymous changes.

---

OP - one thing to do here would be to find a good database for expression quantitative trait loci, or "eQTLs" and see how many of these variants are adenine residues, and then narrow to cases in which expression has been studied for each allele eQTL database will allow for quantification of exactly how frequent a pattern like the one proposed (or, a slightly different pattern as well) is.

if you do this, I think that you will see that the kind of scenario you describe is rare. despite this, there also likely scenarios in which this is true, as well - but here, the trick is to understand why it may be true (i.e., thinking back to the comments on transcription factors and miRNAs, above).