I have an extremely rare mutation

[sue1234]

(Some may know me from the POTS forum!). I have a 30 year history of low blood sugar, that showed up in my late 20s. Fastforward to getting POTS in my late 40s, and have lived with it for the last 9 years. At the same time my low blood sugar has intensified, and I have been visiting doctors for both POTS and hypoglycemia.

I did my 23andme testing a couple of years ago, and looked through the usual SNPs that are talked about here. Nothing totally unusual, but am heterozygous MTHFR. Every time I read a Pubmed or other scientific article that is related to my health conditions AND mentions some gene, I immediately look up my genetic results on my 23andme results just to make sure nothing is overlooked.

This week I looked up GLPR2(glucagon-like peptide receptor) after reading something about blood sugar issues. Anytime I look up where I fall, I always compare it to HapMap-CEU for European ancestry. This time 2 of my SNPs didn't even show up! I was GG on one and AA on another, and they were literally 0.000%. I also had one SNP that was only 0.9%. Does anyone know enough about all this to know if this is significant? I thought it strange that I have the long history of low blood sugar and this gene regulates how your body responds to glucose after eating!

Thanks for any help!

 

[Valentijn]

@sue1234 - What are the rs IDs for those SNPs?

 

[sue1234]

Hi Valentijn! For the GLPR2, the SNPs are:

rs17207885 CC (me) 0.9%

rs4791882 GG 0.0%

rs7217767 AA 0.0%

 

[Valentijn]

@sue1234 - It looks like all three of those are in very tight linkage disequilibrium, especially in Europeans. And none of the three are in coding sections, so would probably just have a mild impact at most.

Those variations are also very common in other ethnic groups (up to 20%), which makes it very unlikely that they're significant.

 

[Critterina]

Sue,
I'm traveling and can't look up my SNPs, but I wanted to say that I've had low blood sugar since puberty, often reactive to moderate amounts of food, mitigated by physical activity, and severe when my adrenal function declined, so I'm leaving this as a note to check when I get back.
I also found no correspondence for some of my SNPs in my presumed ancestry, also European. Turns out I have completely non-European mitochondrial DNA.

 

[sue1234]

Thanks so much for your input, Valentinj. I needed to hear that from someone! Critterina, interesting about the difference in the mitochondrial DNA. This is so over my understanding, for sure.

 

[Eeyore]

@Valentijn

It's sometimes not the best idea to look at SNP's in other reference populations. A SNP in one gene in one population might indicate a very different haplotype from the same SNP in another population. Often they are just markers for other more important changes that are in tight LD. e.g. In my own genome, there is a somewhat uncommon genotype I have (homozygous) at a SNP that is found in about 20% of caucasians (I am caucasian). The same SNP is associated with serious disease in Asians, but is not generally a problem for Europeans. This is because they don't share a common founder, and in one population serve as markers for a risk haplotype, whereas in another, serve as markers for something benign.

This is one of the problems with SNP's versus WGS or WES. Essentially we are trying to impute haplotype from a particular base, which doesn't always work.

While you're right that the SNP itself is not likely to cause much of a problem if in some populations it is common, it may, in some populations, be in tight LD with something less benign.

 

[nandixon]

Yes, I've been meaning to mention this for a while, and have some additional thoughts, especially since @Valentijn may be planning to set up a new platform.

A good example of what can go wrong, i.e., false negative results, from not using an ethnically appropriate dataset, is the inadvertent exclusion of a SNP, rs1004819, in a recent analysis of the IL23R gene that was looked at on @Eeyore's thread in the chart comparison in this post.

Rs1004819 is a well studied and important SNP in European ethnicities with respect to, e.g., Crohn's disease.

It has a minor allele frequency (MAF) of about 0.3 for European ethnicities, yet the MAF rises to 0.4 in the full 1000 Genomes database - because that database contains about 50% Asian genomes, and the risk allele in rs1004819 for Europeans is actually the major allele for many Asian ethnicities. So with a threshold cutoff for the MAFs for IL23R set at 1/3 in the chart comparison/analysis shown on the other thread, that SNP was left out.

Under other circumstances that might have been a critical false negative result.

Studies in the literature that explore the relationship between disease and genetics always use ethnically homogeneous cohorts, unless they are specifically looking to see if a cross-ethnic effect is present. In English language and European journals, for example, these cohorts are going to have been screened to ensure a predominantly European ethnic background.

The reason is that the effects that many/most (non-congenitally pathological) SNPs might have, though statistically significant, are small, and are often easily lost across different ethnic groups.

The allelic frequency differences between European ethnicities and Asian ethnicities, for example, are often large, occasionally to the extent that the minor and major alleles are reversed for the two groups - as seen for rs1004819. This can easily obscure meaningful disease relationships that have been found for SNPs based on studies that were themselves done, for good reason, on strictly European ancestry cohorts.

In simplest understandable terms, a SNP that is (potentially) problematic in one ethnic group may not be so in another ethnic group due to ameliorating SNPs and/or epigenetics, including the effects of traditional/cultural diets, etc., that the other ethnic group may have.

A European ethnicity dataset must be used for these sorts of analyses. Even the HapMap-CEU dataset, although it initially appears somewhat narrow as being only Utah residents, has northern and western European ancestry genomes and so is far superior to the full 1000 Genomes dataset, which doesn't have an ethnicity genome weighting nearly appropriate enough for the general Phoenix Rising population.

Using, e.g., CEU, will obviously increase the false positive rate, but it's much better to turn over too many stones than to miss what you're looking for entirely.

 

[Eeyore]

As the 1000 genome project continues to grow, it will be better and better. I think they have about 2500 phased haplotypes right now.

I usually use the CEU HapMap, as my ancestry is all European and mostly northern and eastern, with some southern as well (Italian primarily).

Sometimes it can be very important to use the right population to interpret a SNP.

e.g. Say that a founder had a mutation at position A in chromosome B before humans left Africa. Say it's harmless - so it confers neither an advantage nor a disadvantage to its carriers. Then, much later, a descendent who becomes an ancestor to many europeans has a mutation in A+10 that is actually beneficial to its carriers, and this is passed on. Around the same time, an ancestor of modern Chinese people has a random mutation at A+50 that is detrimental to its carriers.

Over time, both mutations propagate through the population - let's assume that selection doesn't have time to remove any mutations.

If we then type for SNP A, we'll find that in Europeans, the ancestral allele is "normal risk" but the mutation is beneficial, due to the fact that it's in linkage disequilibrium with the new mutation at A+10. Similarly, in Chinese people, SNP A will be detrimental, due to linkage with the mutation at A+50.

If a Chinese person uses a study on caucasians of that SNP, he will think it's good news - but it probably isn't for him.

I was a bit worried looking at one of my SNP's that supposedly caused a serious disease, but then I realized that 20% of euros had it, and it was normal in that population (and there's no way 20% of euros have the disease, or probably even .20% - I've never heard of it outside promethesae). On closer inspection, it was found in Asians with this disease, but the SNP is also VERY rare in Asians, and signifies a different haplotype.

I guess this is a problem we'll have to contend with until such time as we actually sequence genes rather than look at SNP's - right now, SNP's are cheap though, and whole genome sequencing is not. I'm considering it though, and my PCP would order it I think if I asked her - I figure I'd have to pay out of pocked though.

My point is mostly that SNP's increase or decrease risk often not because of the SNP itself but because of the associated haplotype. That is one reason that non-coding SNP's can be relevant (although they can also affect gene expression, promoters, etc.)

Sometimes it's more obvious what's going on - e.g. say a SNP is a nonsense mutation in a known gene, and the studies show a connection with high or complete penetrance to a disease that would obviously be connected to loss or gain of function in that gene. In these cases, the SNP probably means increased risk in anyone, regardless of ethnicity.

The other unfortunate thing about promethease right now is that it will flag 10 different SNP's and give warnings for each, when they are all just SNP's in varying degrees of LD with the actual risk locus. The average person may not understand that this is one risk reported many times, not many risks that have to be compounded.

 

[Valentijn]

@Eeyore, @nandixon -
I think it's not really a good idea to put too much faith in MAF's from specific ethnic groups. To start with, these are usually pretty small samples, with a maximum of around 113 people in them (or 226 alleles). It's very likely that such a small group will frequently have random results that make a SNP look rare (or common), when it isn't, even for the group which they represent.

The other problem is that the ethnic groups are very small ethnic groups. CEU isn't "European" ... it's a very specific subset of Americans who emigrated from north and western Europe into Utah. To be a little more blunt, it's Mormons, and like most religious minorities who have faced some persecution at times, there is likely a relatively high amount of intermarrying within the group. It is not a diverse European population, but rather a small and very distinct section of it.

This is the same for many (all?) of the ethnic groups where data is provided - they are usually isolated and very distinct groups. They might get labeled as "European" or "Asian" but they are also labeled as being selected from those specific subgroups (CEU, African tribes, etc). Hence if someone does not have significant origins traced back to Utah, the CEU group data is of very limited value.

The other issue is false positives in genetic research. These are almost always looking for correlations with disease, rather than actual enzyme function, so are quite vulnerable to inaccurate results. This is confounded by looking at large numbers of SNPs at the same time, but failing to correct for the increased chance of false positives. It is extremely common for different research groups to get contradictory results, even in similar populations. This problem gets even worse with haplotypes - they are usually having a small impact there is typically not very convincing evidence that they do result in differences in enzyme function.

Frankly I'm not sure that there's good evidence that haplotypes are a valid concept beyond the more directly observable interactions of multiple missense mutations on the same gene. At any rate, haplotypes seem to have the weakest evidence and the smallest impacts of pretty much any type of genetic variants.

It's also important to keep in mind that there is very little functional difference between any ethnic groups. While there are some differences in SNP prevalence, every ethnic group on Earth is nearly identical to every other group. We have very little variation as a species, in contrast to the variation we have when compared to the species most closely related to us. Hence when weak associations from different research teams involving different ethnic groups is contradictory, the more likely explanation is that neither set of results are reliable in any ethnic group. And because the correlations between those SNPs and the disease is miniscule anyhow, it doesn't particularly matter one way or the other.

At this point there are probably millions of SNPs which researchers have found "significant" but with tiny and often contradictory correlations. Considering how many of these fail to hold up consistently when replicated, I think it's far more likely that many people researching SNPs are doing a very sloppy job of it and as a result they are frequently generating false positives.

As I've said before, there are plenty of SNPs having a large and consistent impact to worry about. If getting excited about tiny and contradictory correlations, or unresearched SNPs "maybe" reflecting a significant SNP, we could spend 100 lifetimes exploring those with little to show for it, and certainly not anything resembling a helpful treatment.

 

[Eeyore]

@Valentijn -

I generally agree with most of what you are saying. There is definitely a problem with multiple tests - you have to correct for it. Say you look at every single nucleotide in the human genome, or even just the 600k or so on many SNP chips. You are very likely to get some false positives by chance unless your population is enormous (and even then if you do 6 billion tests - one per bp - you will quite possibly get chance associations that look significant). I think a lot will get sorted out as the studies get larger and larger - more people = more power, and you can start finding things with much, much lower p-values. Even then you have to replicate it in another population.

You're also largely correct on genetic diversity. Africans are FAR more genetically diverse than other races actually, because the human race lived most of its existence in Africa. Much more recently we spread throughout the world, but most SNP variations date back to those times, and were represented in the groups of people who migrated to Europe, Asia, North America, etc. It's not always true though - for example, one mutation I have that is very significant for me is found only in Europeans. On the NCBI site, I found it in the CEU population, the Finnish population, the Tuscan population, and several other European populations, but not at all in any non-European populations. It's generally believed to be of mediterranean origin.

You don't actually have to trace back to Utah, but you do have to share ancestry with people who are in the Utah pool. I never thought of it until now but it makes sense they'd be mormons - mormons are very into genealogy, it's part of the religion. However, being mormom isn't like being Jewish. Jewish genetics are relatively isolated from anyone else and have been for many thousands of years - which is also why Ashkenazi Jews have a number of unusual genetic diseases that no one else really has (e.g. Tay-Sachs). Even then, 5000 yrs or so isn't THAT much and their genetic data would still be useful in phasing our genomes, but not as useful.

When I'm talking haplotypes, I'm not talking about maternal/paternal haplogroups as 23andme reports. I think I maybe didn't express that clearly. By haplotype, I just mean a piece of a chromosome that tends to get inherited together. Haplotypes are the reason that linkage disequilibrium exists, because genes that are close are usually passed together. You can phase your genome vs a large panel of haplotypes. I have done this for stretches of my genome, using both my parents and my sister as well, and it gives me a lot more information and allows me to impute missing snps in many cases with certainty greater than one gets from an actual snp test. Some SNPs are in very high LD, but you can't necessarily tell if you don't have a phased genome, because you don't know which things are traveling together. That's why they report haplotypes on the NCBI genome browser. For example, knowing that you have 2 haplotypes, A G G T and G A T G is much more useful than just knowing you are A/G G/G G/G T/T. (By convention, a / denotes unphased genotypes, and a | represents phased genotypes). The unphased is often ambiguous when being resolved to haplotypes, but can be done with very good accuracy using family information + the available haplotypes (about 2500) of the 1000 genome project + a good imputation program. I usually use impute2, although shapeit is better for entire chromosomes. Interestingly, you get better results adding in unrelated people in the mix - impute2 is smart enough to throw things out that aren't useful - but you definitely need the ones that are more closely related.

I've been able to figure out all the important no-calls from my report. For one, I had to phase the genome, and then use the genome browser and D' measures to determine the ancestral haplotype from which the SNP arose, then I was able to exclude certain haplotype combinations because they simply don't occur (you can do this when you have flanking mutations with greater frequency than the one you are looking at, and where the D' is 1.0). That leaves you only 3 haplotypes that are possible instead of the usual 4. e.g. say the ancestral haplotype had flanking G's on both sides at nearby SNP's that are polymorphic (say G and A). In that case, if the mutation arises later in between them, everyone with the mutation will have the G's at the flanking positions, so you can rule out a mutation if you have A's. You can't do this if your data is unphased though, because you can't tell what is linked to what.

This is why geneticists and bioinformaticists generally use phased DNA data whenever possible. It's MUCH easier to phase data and it can be done with greater accuracy if you have your parents' DNA (or even one parent). So, if you have parents alive still, get them to test now. Even 23andme is plenty of data to phase a full genome if you later do that. I can now look down chromosomes and see where my parents' had crossover events to make me and my sister, and I can see how one of my haplotypes at any point on any chromosome will match mom's, and one will match dad's. I can also often figure out which grandparents I inherited genes from, which is just interesting.

Basically we are all just Frankensteins glued together from bits and pieces of millions of ancestors. Other people are put together out of the same parts in different combinations, so we can use their data to put ourselves together. It kind of makes you feel more connected to the rest of humanity when you think of that way - or at least it does for me.

 

[Valentijn]

@Eeyore - "Haplogroups" was an extended typo in my post. I'd meant to refer to "haplotypes" in each case

 

[sue1234]

So @Eeyore and @nandixon , what is your opinion on my original post? Are my findings something that actually might be need to looked into further? If so, do I bring this to my endocrinologist? Or, do you randomly go to a geneticist? I have been tested for every conceivable issue regarding hypoglycemia, and do not fit into a "nice neat box". I could not find much info on the GLPR2 gene, so it seems like it has not been studied like most other glucose-regulating genes.

 

[Eeyore]

In general, if you don't find any disease associations in the literature associated with those polymorphisms, you probably just have an unusual haplotype. It can't hurt to watch for further research, but you can't get too excited over every SNP where you have something unusual unless there is some connection to something that has been studied, or you have VERY good reason to suspect a problem (e.g. you have a known genetic disease associated with gene Y and a SNP there that is uncommon/rare).

You'll find, as I think Valentijn was saying earlier, that our genomes are full of rare variants of little significance, especially those in non-coding regions or that are synonymous. My point was that in a study they can still be associated with a risk, but that risk could be in a completely different gene through linkage. It doesn't appear that we have enough knowledge to interpret your data at present, but genetic knowledge is increasing rapidly.

 

[nandixon]

So @Eeyore and @nandixon , what is your opinion on my original post? Are my findings something that actually might be need to looked into further? If so, do I bring this to my endocrinologist? Or, do you randomly go to a geneticist? I have been tested for every conceivable issue regarding hypoglycemia, and do not fit into a "nice neat box". I could not find much info on the GLPR2 gene, so it seems like it has not been studied like most other glucose-regulating genes.

Your chances of these particular SNPs being of consequence seem low, but some intronic SNPs can certainly sometimes cause disease, by interfering with the gene expression process. And, interestingly, SNPs that cause problems in various genes that have been found to be related to impaired glycemia states often seem to be intronic.

So if you've reached a dead end and feel like your research might support a problem in the glucagon-like peptide-2 receptor gene (GLP2R), then the simplest thing to try is to contact a current researcher in that specific area, such as the individual in this 2014 paper (their email address is given under "Author information"):

The CNS glucagon-like peptide-2 receptor in the control of energy balance and glucose homeostasis

and ask their opinion. They're usually happy to reply and often looking for new research prospects.

You can also use this link for searching on PubMed to find other researchers as well:

http://www.ncbi.nlm.nih.gov/pubmed/?term=glp2r

 

[Eeyore]

Well, if I were really concerned about it and had a high degree of suspicion, I'd look into modeling the protein and mapping the mutation to a particular functional domain of the protein (if it's coding). If non-coding, I suppose I'd look for a hypothetical reason why up or down regulation of the gene's transcription and/or translation might matter or might affect the phenotype in question, and I'd look for other SNP's in high linkage disequilibrium with that particular protein. I'd try to map out the promoter regions and splice sites for the introns, and compare to known binding motifs for transcription factors that might be relevant in a particular case.

It's not a simple process though. I do not believe you'll get anywhere with researchers, but you could check.

Another approach would be mendelian genetics - make a family tree and map where the SNP came from and went to, and see if you observe a particular pattern of inheritance or similar symptoms in others who have this SNP.

This is, of course, assuming you are very motivated and have a high degree of suspicion. Usually, it's not worth it, and unstudied snp's are more often than not best ignored - even when they are rare, and especially if they are intronic or not even intragenic.

 

[nandixon]

It's not a simple process though. I do not believe you'll get anywhere with researchers, but you could check.

This sort of thing is definitely low odds. I have an extremely rare homozygous mutation in the PEX13 gene (in my signature). I used an online tool to determine that the SNP, which is located in the 3'-UTR of the gene, was predicted to be a microRNA binding site target - and in particular for one of several micro RNAs found in a recent ME/CFS study.

An Australian researcher was interested in obtaining fibroblast skin cells from me to study the mutation. But he's half way around the world and doing a skin biopsy and making arrangements were just going to be too much for me for such a low odds proposition.

 

[Eeyore]

Yeah - if I had a rare mutation and someone was studying it and wanted a sample, I'd probably do it. I'm always excited at the prospect that someone doing real, hard science might actually be interested in helping me, on any level, figure out what is up with my personal biochemistry.

BTW - that SNP isn't on the more recent 23andme chip. Might mean it gets reported less going forward. I can't see my genotype for it - although I could probably impute it.

 

[sue1234]

I need to quit looking at my 23andme results! I just recently looked up my results for the GYS1 gene (glycogen synthase 1), and turned up AA at rs8103451. When I compared it to the HAPMAP CEU, there is no AA percentage. There is also none for Asian and African results. I did find that ONLY the HAPMAP GHI category showed a 1%. (some kind of American Indian group---I am not aware of any Indian genetics in my family).

This gene apparently has something to do with blood sugar, which I mentioned in a post before that I have a history of hypoglycemia.

Any thoughts?

 

[Eeyore]

@sue1234

I wouldn't be overly concerned about it w/o more reason to suspect something.

See here: http://www.ncbi.nlm.nih.gov/projects/SNP/snp_ref.cgi?rs=8103451

It has an allelic frequency of 3.06% in the 1000 genome project, which while not extremely common is also not that rare.

Additionally, it's intronic (this means it occurs in the part of the gene that does not code for proteins). Intronic mutations can have some importance in gene expression, or, rarely, if on the borders, they can affect splicing. This one appears to be right in the middle of an intron - so even less likely to matter.

I've done the 23andme v4 and the ancestry.com chips, which either do not type for it or I was a no call.