 okay so I am I'm I'm the Walmart act for Gary Churchill and I and I hope I'll do justice unfortunately my talk will be depressing but I'll see how we do right after lunch all right so behold the human being as viewed classically I understand that this is one view however oh I should tell you I worked at the center for human disease modeling and I'm also the chief scientific officer of a by the company rescind the therapeutics so there's my disclosures so the human being but according to sort of the post genome view I think this is a more accurate representation of the human being because as we do more biology more physiology and more genetics we beginning to appreciate the inherent complexity so for for those of you who had to suffer through my talk in Toronto the first portion will be the same but I do have some new data to share and I think it's important in the context of what we're trying to do so today I just want to focus on the theme that has been permeating the meeting and I think I have the sense it will be it will be part of the conversation about what comp 3 4 5 10 and 20 will be which is moving from the function of genes which remains critically important and it was in large part the focus on making this systematic gene knockouts to move to the function of alleles and I sat through the morning and I learned a lot and I really appreciate everybody's talks but I'm here to tell you that it's a really really bad idea to just bet everything on introducing single base pair substitutions in another model organism and interpreting the data accurately and the reason for this is something that I did not discover this has been known for the best part of 70 years but we now have more evidence to bolster this and this is the notion that the effect of variation is species-specific and I really appreciated the slide that Brandon showed on the RNA-seq data because even miss splicing non-associated splicing nonsense media decay and all the extra skipping stuff even that is specific to species on another day I can give you this big treatise actually how making crisper single base pair knockouts in zebrafish embryos is a truly terrible idea because this particular model organism has developed a transcriptional compensatory mechanism that essentially shields the model organism from the from early developmental defects this is now shown but enough of this let me show you the data all right so this is a very old study and this is a gene in my favorite group of disorders the psyliopaths arpegrip 1L the allele is arginine 937 loosing now arpegrip and and this is from the paper we published back then and the the reviewer said well this loosing cannot possibly pathogenic because I see here two species that actually encode loosing as the wild type allele now remember we're not in the gene the gene has already been shown to be pathogenic we have met the burden of evidence the burden of proof so here's the question but the in vivo complementation study that was done in vivo was suggesting that this allele was loss of function and just to highlight that for those of you who are familiar with this kind of evidence please bear with me I just want to make sure we're on the same page this is the spread of phenotype this is convergent extension this and a wind phenotype that is a well-known property of psyliopathies this is suppression and this is the distribution of phenotypes and then this is suppression a morpholino plus wild type human mRNA human human human human can I say human six more times okay human mRNA we have almost complete rescue and then we introduce a single-pace space substitution so the null hypothesis is that when we introduce a single-base space substitution that is benign there will be no difference between injection of wild type and mutant and RNA this hypothesis is soundly defeated there's many zeros to this p-value and in fact this guy here is statistically in in distinguishing from this guy and the interpretation is that that variant is a loss of function allele I cannot tell you if it's a complete loss of function or a 95% loss of function by the loss of function so we know that null for RP grip 1L is incompatible with life in human or in the mouse for that matter and yet we have two species that are homozygous for what the functional test tells us is a pathogenic variant and the question is why why why why why why all right so one of you reminded me that I've dubbed this Tarzan genetics because it is and it's sort of this sort of random thing will go around here we jump around data sets and we try to make conclusions but the bottom line is we actually try to do a more systematic study of this so I got together with Shamil Sunaev and and and he educated me on how we can do some of these things and this the experiment that we did back in 2014 so in 2014 what Daniel Jordan did Shamil student is he took two you know dirty ish databases Hume bar is dirtier than clean bar right both are getting better but both of them are imperfect and he took these two databases and he took all Missins variance and he did an alignment against the sequence of 100 vertebrates we stayed on the vertebrates because once you move to invertebrates evolutionary distances make correct alignment sometimes quite problematic and we were concerned about false positive false negative rates and depending what assumptions that you make about these alignments and I'm happy to talk to you about the assumptions and your tolerance at worst the most stringent thing is that 3% 3% of all pathogenic Missins variance found in humans are fixed in another species and the largest estimate is 12% the most permissive estimate in reality the observation that we tend to be playing with is somewhere in the 5 to 7% range so my statement that I made in Toronto and I will repeat here because I think it's an important statement is that all the Missins mutations that we know 7% of them are fixed in another species I think moving forward this a heavily biased data set because when individuals look at research or clinical exomes and they find a candid variant that is homozygos in the mouse or in the cats or in the opossum or something else they are more likely to discarded and computational algorithms are certainly more likely to consider this alias as benign and that is often time correct but not always okay so why is this there is such a thing something called compensatory deviation and this is very old stuff the the bacteriologists and microbiologists would shoot me if I claimed any novelty here because we've known this for a very long time but you know again perhaps we need to remember some of the literature because there was actually literature from the 50s before the internet imagine that so the the the basic idea was that if you have a pathogenic allele here that is doing something horrible say to the half-life of your protein or the or the folding of your protein you could have a second allele that repairs that defect so this is a highly deleterious variant but this plus that makes it a benign very okay so how do you discover these things well it's actually pretty easy here's a here's our protein this actual data and there's a there's a sparaging to Alan here mutated in human disease and the question that we have here is why is that these four species can tolerate the alanine of course the effect could be in trans could be in some other gene could be regular there's many many things but our narrow hypothesis I guess you know okam would be proud is that there's some other variant in the same sequence in the same haplotype that is somehow shielding the effect of the allele so the experiment that we did is very simple and there was no Illumina instrument anywhere to be found which is just take the sequence align it and then look at the other alleles that seem to be traveling uniquely with a candidate pathogenic variant but are absent from the human sequence if these variants have the capacity to shield the detrimental effect of this allele then if we were to have a human mutant alanine here and that has say the glutamine or the valine or the lysine then that should convert the message from mutant to well type yes and this exactly what happened so I'm showing you this an example bbs for this is our big reponell the one that I showed you earlier so the here is our r2l and it turns out I'm sorry that I'm not sure how well that these colors are projecting at the far end but it turns out in our big reponell we identified 33 candidate changes it's not a small protein so we identified 33 candidate changes that evolution said they might travel and protect against the pathogenic allele it turns out a couple of them were actually protecting very well in fact in fact three of them were converting the mutant to well type why is this important here's a little girl that we saw a duke back in 2012 and she's the one of these fabled diagnostic odyssey kiddos that we all working so hard and I must say I have an enormous admiration for my colleagues who do this we all work so hard to overcome so this kiddo whole exome sequence in trio base like everybody else does she had the essential these days almost obligatory mutation in titan right those are you in the game you know what I'm talking about and she also had two de novo variants one of them in btg2 and the other one is nos2 we have some ideas about nos2 we functionalize both alleles in zebrafish embryos and our index phenotype was microcephaly and to cut a long story short it turns out that microcephaly was likely caused by suppression of btg2 the experiment as you know before suppression we've since also made a CRISPR deletion CRISPR mutant and that reproduced suppression rescue and then rescue with this valine 141 methionine and that alia was acting as a loss of function so from that point of view we had conflicting data the prediction was that the alia was benign exact and nomad was telling us that this gene is intolerant to haplitis efficiency and the zebrafish data were telling us that there was a microcephaly caused here when we looked at a marker of bilateral symmetry of postmortotic neurons we had a clear phenotype and when we look at the number of neurons being born we also had a very very clear phenotype here so there was clearly a defect in neurogenesis but this was again an end of one so there was always some concern and the concern was really this there is the valine in this little girl and here is the rest of the animal kingdom so as you can appreciate we have almost everybody being methionine in fact valine is a derivative allele is not the the ancestral allele so my statement to you is that if you have methionine in this particular gene you essentially a haplitis efficient and that is intolerant to sort of a quantity essentially this child has very severe neurological problems each other that or there's a compensatory event so what we did is we align all these sequences and we identified a number of candid residues that seemed to be traveling specifically with a methionine and then we introduced these residues into the human sequence and we asked the question can we rescue the allele and the answer is yes we can we used the most direct phenotyping method which is the most quantitative which is the counting the number of dots right so each dot is a birth neuron so we see here a reduction your genesis and then a rescue and so forth you get the idea and the answer turns out to be that there's two alleles are 80k and l 128 v and I and I love showing this because it's like losing to valine oh yes who gets out of bed for a losing to valine change well but it's true that if you actually change either of these positions you essentially have full rescue or the functionality of the mRNA and the other thing is that when you just consider these two alleles and nothing else in the context of evolution we you explain about 90% of the species for which this methionine allele can be tolerated now I can also tell you I don't have a result but I can also tell you that finally about two weeks ago we were contact we have this in gene matcher of course and we were cornered by a group who identified to the novel nonsense mutation in two individuals and we have also a phenotypic match which sort of supports the the causality but that's another story okay why this is the thing right it's a cool phenomenon but why so this is just predictions so what is this is this is a thermodynamic mapping of the human protein based on crystal structure so this is what your wild that looks like and then what happens is that when you actually model the valine to methionine at the far right you actually get the tail of the molecule to flip out it turns out I don't have time to show you all the data but it turns out that tail has ubiquitination properties but and then on the bottom what I'm showing you is the human sequence in which we have now coded both the pathogenic mutation in the patient and each of the compensatory alleles and I found it quite striking that just changing one amino acid residue and remember we have no real knowledge of why we can actually restore the three dimensionality of the tail and that potentially explains why the double mutant is essentially functioning just like the wild type but the single mutant is not the reason of a big song and dance is because if you try to model this in the mouse you would be profoundly disappointed but I think would be a really cool experiment I'll come to these is when you humanize the entire locus so I guess my message is here I'll be like one of my colleagues I'll give you the punchline don't make miss sense mutations humanize the genes that you want to study and then introduce the human alleles into those individuals single base pair nicks with with a crisper expedites homologous recombination you drop in the entire locus come on your experts are back recombineering but making single base pair mutations based on crisper you are not a good idea and let me tell you who I really this is not a good idea all right so let me tell you a couple examples now one area that I think as a community we need to better is report our negative data because sometimes our negative data are actually posted data that we don't know right so it has been incredibly difficult to track down individuals and examples who have made knockouts or knock ins and they had an unexpected event aka the mouse look fine I find that intensely interesting provided that you can show that you actually did the knockout you know the technical stuff but from the harvesting literature there's there's not many examples but there's more and I'm sure this group will generate tons more data about comparing knock ins with knockouts so for example you know you have a recessive mutation you take out the gene this and that but the performance is really poor in fact we're finding significant discordance in the penetrance and the expressivity of the phenotype of the knock ins compared to the knockouts now why is that of course I'm going to argue that some of it has to do with this compensation and I have a little bit of data to support that claim so here's a gene that rolls of it was named rolls of the tongue EIF to be five I'm sure you all know it intimately this locus is mutated in a severe disorder this a childhood a taxi with hypomyelination clearly will have a severe effect on the quantity so it turned at the mutation of the initial discovery mutation was this arginine 136 to histidine and the human genetic data for this allele beam pathogenic are actually quite compelling yeah but here's the thing the 136 age histidine mouse it's fine not a problem so there is two potential now Gary will tell me you didn't put in the right background which it's possible but so my no hypothesis is that the cognate locus in the mouse contains a second allele that you know does this thing so this actually a testable hypothesis in this the test so we know how to test cerebellar hypoplasia because when we knock out or knock down a gene that in the in the in the fish we can actually flip the embryo in its back and we can actually measure the entire volume of the cerebellar so we love these assays because they're quite quantitative and here's the experiment each diamond is an individual embryo and everything has been done in triplicate so here's where we are we have our spread of wild type measurements we have the suppression and then we have the performance of the wild type human and mouse mRNA here is the key results in the human our 136 age compensated attempt at rescue showed no rescue if I take this allele and I put it in the urine sequence and I inject the fish with mouse cdna with mouse mRNA it's our 132 age in the mouse sequence that residue behaves as wild type right which suggests that this residue is influenced by the only thing that is different between those two mRNAs is the rest of the cognate locus now we have actually not found yet the compensated allele here it's a little bit more complicated because we have close to a hundred candidates so but the fact remains that this experiment is good enough to predict and this is what I'm hoping we can have a good conversation about this very simple experiment is good enough to predict whether you should make a knock in or whether you should make a replacement of the mirror in locus and then knock in the human allele you want to study right these are cheap experiment these are I mean you know these are not you know this is not a rocket science we would do this every day just to give you a sense with this a two and a half thousand alleles in 850 genes over about eight years so now it also works the other way around let's go back to there's this gene a TDC 21b back to the cellopathies we and others reported many many years ago this allele P209L is hypo-morphic and we have data on this guy from two sources source number one is human genetics we know that a null TDC 21b is incompatible with life we know this for humans because fetuses with no alleles in TDC 21b will have you know really horrible disorders like mechogruper syndrome and and things like that most of them will perish prenatally and also in the mouse when we knock out TDC 21b we get open neural tubes and and most of the embryos perish by E15 to E16 P209L we reported the number of years ago in isolated nephronophysis or syndromic versions of nephronophysis likewise when we expressed in human cells and we measured the length of cilia we actually showed that this allele was affecting a ciliary length and also we were also be able to show that it was reasonably stable and all these things okay so we because we didn't know about the cis-complementation stuff we spend you know a little bit of time and we actually made a 209L knocking mouse because we needed to understand how a hypo-morphic allele was behaving and the problem is yeah the hypo-morphic mouse is not hypo-morph at all in the mouse in the strain that we chose 209L is actually a null allele we haven't formally done the experiment to there's a there's a mouse mutant called alien that is a true null and our animal phenocopies alien now we haven't done the formal experiment to cross them in and see what happens but if it's not a null it's a very very severe we've been trying for the best part of a year we've never seen a single 209L homozygote survive so we asked the question is this an inverse mechanism in which the 209L allele is milder is protected in the human sequence but is more severe in the mouse sequence and it turns out to be exactly the case these are embryos and in this particular case again we're looking at convergent extension because this is a cheap and dirty assay and once again we can actually see that the allele in the context of the mouse sequence is behaving as a null whereas in the context of the human sequence it is behaving as a hypo-morph I find that exciting so we have I don't have time to show you all the data I find a little bit exciting because we've actually identified three candidate compensatory alleles and that is cool because these sites might also be candidates for genetic suppression of the phenotype in patients and what is even more interesting here to consider is that we don't really know yet what the frequency of such events would be in the human population because the compensating alleles themselves are under absolutely no evolutionary pressure they're just under genetic drift okay last but not least everything I've shown you so far is dependent on evolutionary fitness we're dealing with pediatric disorders and the compensatory alleles always tend to come first they tend to be older alleles and they tend to be neutral or mildly pathogenic at worst and then they essentially sensitize they make the genome permissive for the really severe allele to land and nothing happens and then you lose the compensated allele the severe alleles becomes highly pathogenic gets gets removed from a population until get a human baby who gets this allele that suffers a severe disease yeah that's not true I mean it is true but it's not the whole truth here's a disease that we can all agree has nothing to do with reproductive fitness and the disease is Parkinson's disease so alpha-c-nuclein the mouse genome that the most perhaps the most famous allele in alpha-c-nuclein is an I53T allele that was actually discovered in a family from the Greek city of Patras if I'm not mistaken so the mouse genome encodes threonine alpha-cine is a small peptide about 140 150 amino acids and for the longest time the hypothesis was all you know the mouse is a terrible model for PD well that hypothesis was defeated as early as 2003 when an A53T transgene by Liedau was actually shown to reproduce many the fix of the disease both at the level of pathology of the loss of dopaminergic neurons as well as lethality and other things so then the second idea was that maybe it's the level of expression it is true that alpha-c-nuclein is expressed at low levels in mouse compared to human yeah but that's not true in that it had to be the cat and the mouse couldn't it now but it is true that the cat genome also encodes a53t and so does the opossum and I'm sure we can all come up all the blind more I'm sure we can all come up with very interesting stories here the point here is that we have a situation where the mouse sure lives two years but the kitty cat lives 20 years so and again we have no evidence of Parkinsonism in cats we'll talk about the wild type strain okay and you know so and this is an allele a53t in case you're wondering is I don't know if it's fixed in all cats but in the cuts that are the the post in the genome databases the oh homo zagas 53t all right so we looked at this and it turns out to give you sort of the reader's digest of this that there's a53t here and there's three other alleles that are you traveling uniquely with a cat genome and and and the mole rat genome and this is important it turns out that there is probably I'm not gonna go into the detail but the prevailing hypothesis right now is that alpha-c-nuclein acts as a tetramer and if you forbid its tetramerization the monomer is actually toxic and the micro fibrils that are formed and all these things are in response to the cell to try to scoop up all the toxic protein right so it looks it and there's some interesting data that suggests that this region of the protein might actually very important for promoting tetramerization now all this of course is speculation what is not speculation is what I'm about to show you so the experiment that we did is we the the zebrafish genome does not have alpha-c-nuclein however if you express alpha-c-nuclein in zebrafish embryos and you measure dopaminergic neurons also when you take out GBA which is another gene that does cause it go shea and also predispose to Parkinsonism in you need to load the system you will actually start seeing quantitative loss of dopaminergic neurons and the experiment that we did is we actually took a transgenic zebrafish embryo these are that positive neurons and we developed an algorithm we're going to actually count the number of neurons the mid-brain and high-brain and we've controlled this with rotanone and you know all these other things I just want to sort of show the highlights here and the highlights are fundamentally that the cat haplotype when when inserted in the context of the human sequence right so what we did is we've taken the genotype of the different species and we changed these positions for human completely restores fd3t to wild type indistinguishable I do not know the biochemistry where we're having fun doing you know filter trap assays and and things we we're we're trying to learn how to do but this is striking because none of this is under evolutionary constraint this is just a stochastic chance and I actually think this really important for one it will be really awesome to start taking human transgene's point in the mouse reproducing the phenotype and then start to start to making these alleles in the context of human sequence in fact we will be deeply motivated to participate in this part of the activity because we need to understand the the effect of human variation for the second thing somebody mentioned sort of non Nikkei's CRISPR editors and all these things well these editors don't work for all residues at the moment in fact we only have two or three I'm sure it'll get better but in the meantime by identifying these sites we open up more opportunities for people to gene editing for residues that cannot be edited and I'll start there this is it has always been my stick for a long time and finally you know you know my old prof Jim love ski you say you know what without data you just another dude with an opinion well here's some data at first the data was one gene and then it became two and now we are five four five so we look for compensatory evidence biological evidence in five genes I understand the small number but this is no longer Tarzan stuff because it's reproducible and we found compensated alleles in each of these five genes so we must really consider the concept there is absolute evidence that are some subset of alleles and I don't know how many maybe ten percent something like this has a species specific effect it provides new info opens absolute therapeutic options and we must consider a little testing before committing resources to make stable alleles we have suffered out of this and so has the community with false negative so this is it and I'll stop there I think I'm on time thank you