Skip to Content

Genetic Variation Gets More Real All the Time

This study goes firmly into the file marked “You never could have done this one a few years ago, sonny”. We already know that there’s genetic variation in every population and in every individual. And we know that a large number of marketed drugs (about a third of them) target G-protein coupled receptors (GPCRs). But do we know how much genetic variation there is in GPCRs across a large sample of individual patients, and how that might affect their treatment?

We do now. This paper in Cell (open access) analyzed over 64,000 people (data from the Exome Aggregation Consortium) for the sorts of variants most likely to make themselves felt: mis-sense and loss-of-function mutations (frameshift, stop codons, etc.) as well as copy number variation. Previous studies had turned up some of these, and the statistics suggested that there would be more of them as sample sizes were increased. That there are. Looking across 108 GPCRs that are targeted by known drugs, the team found over 14,000 variants, and the distribution was long-tailed indeed: any individual receptor had an average of 3 or 4 common variations across the 64,000 patients (more than one in a thousand minor allele frequency) and 128 rare ones. The estimate is that at least 120 people out of the 60,000 had what looks like a loss-of-function mutation in a GPCR somewhere.

Of course, not all mutations are created equal. What looks like a nasty shift might honestly not matter as much as you’d think if it hits in a part of a protein that is just taking up some space between the really functional regions. So the paper zoomed in a bit more and found that about two thousand of those fourteen thousand fall into what are very likely to be functional sites. In addition to changes in allosteric regions and drug-binding pockets themselves, there seem to be many mutations on the intracellular regulatory surfaces of the receptors, which suggests that there are a good number of patients out there who will not respond in the ways that you’d expect to a given drug.

The paper shows this experimentally with variations in the mu-opioid receptor (which is already known to be rather variable). Looking at the mutant forms that were predicted to have changes in functional regions, they showed that some of them have less effective signaling across the board (agonists, partial agonists, and antagonists alike), but others showed real variations between the different classes. So there are apparently people out there who are more sensitive to (say) morphine than average, or will show unexpected effects between different synthetic opioids, or greater or lesser susceptibility to a dose of an opioid antagonist. Other receptors that show a lot of natural variation include CCKA, dopamine D5, the calcitonin receptor, and somatostatin SSR5. As the authors put it, “a substantial fraction of the population might carry variant receptors and remain healthy but have the potential to display differences in drug response when treated with a drug.” Among the drugs that the paper flags (in table S5) for a high chance of variable response are Belsomna (suvorexant), targeting orexin receptors, all the natural and synthetic cannabinols, and all the GLP-1 targeting compounds, such as Byetta (exenatide).

So perhaps we’re finally heading for that era of personalized medicine that everyone keeps talking about. As it stands, there are no GPCR-targeting drugs with any human genetic variants flagged by regulatory agencies on the label, but that is surely going to change as sequencing gets relentlessly cheaper and more widespread. It’s worth realizing, though, that most of the stories about personalized medicine seem to assume that we’ll sequence someone’s genome and find the drug or combination that will be much more likely to work. But what this paper suggests (and what has always seemed statistically more likely) is that it’s going to be mostly about telling people that the drugs that work for the rest of the population won’t work for them. What then? Avoiding useless or even harmful treatments is definitely a good thing, but that doesn’t seem to be what the general public is picturing as the most likely outcome.

35 comments on “Genetic Variation Gets More Real All the Time”

  1. luysii says:

    Nice to see some of the basis for what docs have always known, the response to a given dose of any drug varies widely among patients.

  2. bhip says:

    The affects of structural variation on agonist/antagonist pharmacology (i.e. the u-opioid data in the paper) makes perfect sense from the perspective of GPCR theory. If you are lucky (?), you will experience a small molecule program in which compounds toggle back & forth btwn agonism & inverse agonism/antagonism across the SAR (if you haven’t- it’s exciting!). The u-opioid data simply exemplifies the opposite scenario- the small molecule is the constant while the structure of the target is the variable.
    Not to be an old fuddy-duddy but the incidence of the mutations is the more relevant factor for drug R&D. It’s very nice work & interesting but I couldn’t find the data which clearly states the incidence of the mutations. Did I miss it?

    1. ScientistSailor says:

      I will consider myself lucky if don’t get mixed up in such a program! It is already hard enough to find quality clinical candidates with out a good portion of your analogs doing the opposite of what you want…

    2. Mol Biologist says:

      Even if it is true about “the mutant forms that were predicted to have changes in functional regions, they showed that some of them have less effective signaling across the board (agonists, partial agonists, and antagonists alike), but others showed real variations between the different classes”, which I doubt. There is very important indication how pathogenic of minor allele frequency, MAF > 1% compared to it’s frequency in general population. Indeed genetics variants gets more real all the time. Especially for well studied disease such as CPVT (Catecholaminergic polymorphic ventricular tachycardia) .
      Analysis of Exome Sequencing Project database (ESP; n= 6503) was systematically searched for previously published missense and nonsense (189 variants in 5 genes (RYR2, CASQ2, CALM1, TRND, and KCNJ2) common for CPVT – associated variants reported in several comprehensive reviews suggested that these variants are not necessarily the monogenic cause of disease.

  3. Biochemist says:

    Could someone more knowledgeable than myself comment on how one would design clinical trials that attempt to take genetic variation into account? I assume statisticians will want to kill themselves with the onslaught of post-hoc analysis (aka data dredging) that comes with all the sequencing data generated in present and future clinical trials?

    1. Barry says:

      we’ve already seen this with Herceptin, restricting the clinical population (and the indication) to those displaying the “right” genotype

    2. MrRogers says:

      The same way you would with any other laboratory value thought to predict (lack of) efficacy. Early on you use inclusion criteria, and later on, stratification. You might also consider exome sequencing, or genotyping for post-hoc identification of responder populations, but realistically if you have to do that, you may not have much of a market. Somatic mutations are a whole other kettle of fish.

  4. Peter S. Shenkin says:

    This also raises the question how many drugs that have failed FDA approval because of lack of efficacy actually are efficacious, but only for a small mutant population that happened to be disproportionally present in small Phase I, II trials.

    1. Dave says:

      PhIII Double Blind Placebo controlled regression to the mean.

    2. MTK says:

      If a drug is efficacious but no one can see that efficacy, did it have any effect at all?

      I ask that tree falls in a forest question only semi-tongue in cheek. A drug whose efficacy can’t be observed isn’t efficacious.

      I think. Maybe.

      1. aairfccha says:

        Factor VII would show a marginal effect if any averaged across the general population, yet is lifesaving for a select subset.

        1. MTK says:

          There’s plenty of examples of that. That wasn’t really my point though.

          Once again, it was a tree falls in the woods analogy. If it falls but no one is there, did it make it sound? Technically, no, since any air vibrations caused by a falling tree are only translated into sound if those vibrations encounter a sensor, such as an ear, that turn that vibration into a sound in our brain. “To be perceived is to be.”

          So my question was can a treatment be efficacious if it can’t be observed to be efficacious. I would say no. If there’s no way to identify specific patient subsets that would respond in a meaningful way above placebo then we can’t hear it fall. It’s not efficacious no matter what it’s activity may be since that efficaciousness can’t be perceived so it doesn’t exist.

          1. AAA says:

            “So my question was can a treatment be efficacious if it can’t be observed to be efficacious.”

            Sarepta says yes.

          2. MTK says:

            Touche, AAA.

  5. Barry says:

    We’ve long accepted that “de gustibus non est disputandem”.But as sequencing gets cheaper, reading an individual’s ca.800 olfactory GPCRs will potentially explain why some of us can’t taste truffles, or pick the cilantro off our tacos, or don’t mind working in a tannery…

  6. Not Agilist says:

    These data also put to rest the idea of the “me-too” drug being a waste of time. Clearly, with so much variability within a given receptor, we need > one compound in order to successfully treat as many patients as possible.

    1. anon says:

      The variability may also occur at a metabolic level, not only due to receptor variability.

    2. MedicineWoman says:

      Therapeutic index and side effect profiles may also be impacted. We often drive SAR from selectivity against closely related receptors with undesirable secondary pharmacology. Should one be unfortunate enough to have a pair of genetic variants in related receptors which decrease efficacy on the target and increase efficacy on a selectivity target, it could easily have an unfavorable outcome.

  7. Student2314 says:

    Education in Chemical biology was probably the biggests croc of crap ever hoisted onto the american public. Places like TSRI, UC, Harvard literally do not care one iota about their students. In fact, they brag about not caring about their students. I never lie, and ive heard it with my own ears– snarky laughing from the PIs office poking fun at grad students even as your hated students do all of your work for you. Its all about sweatshop labor. I kid you not, the last TSRI pamplet they sent out in the mail, had a whole section convincing grad students to leave the field.

    1. Student says:

      ….so why did the tax bill raise so many concearns? Because the selfish PIs love to have a large pool of students to choose from up front. Afterword, who cares, discard them like cheap Christmas trees the day after Christmas. Screw them. There just people right guys, science says people are peices of trash made out of fun stuff like lipids and proteins? Why care?

    2. Scott says:

      The grad-student tax break discussion isn’t in this comment thread.

      You’re students, you should know forum etiquette by now.

      1. johnnyboy says:

        This guy keeps popping up randomly in the comments, to heap trash on Scripps and occasionally a specific lab. Obviously someone who’s had a bad experience – don’t know what he thinks this venting will accomplish. Or maybe he’s a russian troll-bot, hell-bent on destabilizing the med chem community ?

        1. CR7 says:

          … the pharma senior leadership has been doing that for a decade+

          1. Me says:

            Pharma management is all russian trollbots?! The commies have taken over our medical industry!!! This is so much more sinister than McCarthy believed!!!

    3. Wallace Grommet says:

      Writing “hoisted” instead of “foisted” and “croc” rather than “crock” is eyebrow raising.

  8. Scott says:

    I’m apparently one of those mu-Opioid receptor mutants. I will turn an interesting shade of red if given actual morphine or hydromorphone (as in, allergic reaction red), but hydrocodone/oxycodone or tramadol are fine.

    1. NJBiologist says:

      Many (not all) opioids (including some synthetics and peptides, I think) will provoke mast cell degranulation. I don’t believe anyone has looked at this, but I suspect MRGPRs are involved, not muOR.

  9. Brian says:

    An Ignorant chemist’s question here. When doing SAR studies in a program that looks at in vitro affinity for compounds vs human and rodent receptors, does the biology program manager have to be careful about selecting what “human” gene to express in order to make sure that it is the one found most commonly in the population at large and isn’t one of the less commonly expressed versions?

    1. Dr CNS says:

      In the real world (as far as my experience goes), in vitro screens are developed using a “simplified” version of a receptor – often times not even reflecting a meaningful physiological state. Just a way to rule out your compound is not interacting with a receptor.

      As Einstein said, “do things as simple as possible but not simpler”.
      Recapitulation of compound effects done as part of a good screening cascade (or the lack thereof) may indicate whether you crossed the line or not.

    2. bhip says:

      Different companies have different definitions for what mutational frequency is worth assessing. Japan requires its own registration trials so potential Japan-centric mutations are of particular interest
      I witnessed a program in which the dog receptor was used for drug candidate assessments was from a different breed than that used in large animal tox & (of course) the receptor sequences were different in the two breeds. The biologist in charge got crapped on by the crack Head of Research (like she would have caught that in the fog of war- please…).

  10. Dr CNS says:

    Does anyone more educated than me know if this would be expected to be similar in preclinical species?
    Thank you.

    1. NJBiologist says:

      Expect some in preclinical species; a comparison of sequences between C57L and C57BL (two very closely related inbred strains of mice–Leaden vs. BLack) suggests thousands of SNPs, indels, etc.

      Expect this to go up in outbred strains/species.

  11. Magrinho says:

    Very cool! I think a big question here (and in a lot of other work) is understanding how downstream signalling is impacted by mutations. Could it explain, for example, why opioids generally block pain but individuals differ greatly in susceptibility to addiction/reward.

    1. Mol Biologist says:

      Sound great! You should ask an expert!
      Why everything is done in Female Balb/c mice (6–7 weeks of age) and how it could be transferable to human?

    2. NJBiologist says:

      Gavril Pasternak has a long publication record arguing that muOR sequence variants impact pharmacology. However, I don’t think he has convinced the rest of the field yet.

Leave a Reply

Your email address will not be published. Required fields are marked *

Time limit is exhausted. Please reload CAPTCHA.