 Hi everybody. Welcome all to this virtual computational biology seminars from CIP. Today we have the pleasure to have Jeffrey Jensen from the population genetics group at DEPFL. So Jeff earned a Bachelor of Science, Bachelor of Arts for the University of Arizona in 2002 in ecology and evolutionary biology and biological anthropology. He earned then his PhD in Molecular Biology and Genetics at the Conrad University in 2006 and then he moved on his post-doc work as a U.S. National Science Foundation Biological Informatics Fellow at UC San Diego and UC Berkeley. And then he founded his lab as an assistant professor at the University of Massachusetts Medical School in 2009 and he relocated his lab here at DEPFL in 2011 and also became group leader at the CIP, the Swiss Institute of Bioinformatics. The group research focuses on statistical and computational aspect of evolutionary theory and population genetics. The lab members of Jeff's lab work on both applied and theoretical problems in fields ranging from population genomics to medical genetics. The group has also developed several software tools. So as you probably don't know, Jeff is going to relocate his lab back in the U.S. at the Arizona State University in the School of Life Science and the Center for Evolution and Medicine in 2017. So today Jeff will share with us a few results of interest regarding the population genetics of creatures that live in humans. So Jeff, I thank you again for accepting this invitation and the floor is yours. Right, thanks everybody for coming. I'll try and stay close to this microphone, I assume. Yeah, okay. Are people actually online? Do I need to do that? They're really there. Okay, fine. And I'll stay by the computer. Okay, well thanks for coming. Feel free to interrupt for a relatively small-ish group, small enough to be interrupted with questions I think. So I'll have to scroll along. I'm going to do three short stories today that are mainly just related by the fact that they're critters that live in humans. That's pretty much the unifying theme. So they're really three quite separate stories. We'll have natural stop points for questions as we go. Okay, so let me start with the basic notion that most of you are probably familiar or more than familiar. Some of you with the sort of generalities of human demographic history, by which I mean there were multiple movements out of Africa in the genus Homo. Modern humans certainly originated there, moved out various waves of migration. We're attaching numbers to some of these things. These numbers over the last couple of years have been changing sort of rapidly with the advent of ancient DNA. So a lot of this history is being revised a bit as I'll return to later. But that's the general kind of picture. And this history of humans is characterized by recurrent events of population size change, population structure with migration, periods of isolation and followed by admixture, positive selection associated with the colonization of novel environments. We have at least a handful of pretty convincing examples in humans of this, whether it's high altitude adaptation or lap case tolerance, et cetera. And of course as with essentially any population for a basic purifying selection to get rid of deleterious things. So that's pretty much the story of humans. What I'm going to talk about today, as I said, is the story of populations that live in humans. And they're characterized, of course, by the same processes, mutation, migration, drift and selection to varying degrees. And the three stories I'm going to tell you today are helical factor, influenza virus, and human cytomegalovirus, actually not in that order. I know this now. It's confusing me, but it doesn't matter. Okay, so let's start with a short story in helical factor. So I'll start each of these little vignettes with just a little background on the critter. So helical factor is a bacterium. It infects mucosa of human stomachs outside of Africa. It's associated with some serious gastric pathologies, including cancer and ulcers. At least half of humans have it, almost certainly more than that. Almost all previous work has focused on seven housekeeping genes. Recently, last year, this work led by a former MBO fellow in the lab, Valeria, looked at this. We analyzed 60 whole genomes from globally distributed strains to characterize patterns of local adaptation in the demographic history of helical factor living in the human host. So let's just get to some quick results here. So the first observation of note is that there's lots of structure in helical factor, by which I mean if you sample populations in South America, they're more alike than a South American population compared to an African population. Okay, that's probably not terribly surprising to you. There's a number of groups. So there's at least two very distinct African groupings here, shown in blue and black, and then there's a sort of North African European, the Asian and Asian-American, so it breaks up at K equals four, where these two things are grouped into K equals five, to get a distinct Native American grouping. While this is interesting, is the following. The estimated demographic history of helical factor itself really recapitulates the demographic history of humans. So you have this quite ancient split in Africa, you have a split off into the Middle East, a split off into East Asia, a split off into the Americas. So this really supports the very ancient association of helical factor in humans, something that wasn't very widely appreciated, certainly before the SAM split. And really intriguingly, as I'll return to in a minute, helical factor itself is offering insights into human demographic history. And that's all again, second, in a way that's often more insightful than you get from analyzing human samples. Okay, so let me just say a few words about selection. So now that we have a demographic model as a null, we can start asking what does local adaptation look like in helical factor. The majority of evidence is for local adaptation, almost entirely in the African population, so the larger and more stable population, the more ancient population. People are all excited in most organisms about the sort of out-of-Africa selection, colonizing all the environments, but of course from population genetics, our expectation a priori is that there should be more positive selection and the larger and more stable population, which is exactly what we're seeing here. I try to avoid this sort of go term storytelling, but I'll just mention a few things that are sort of interesting. Multiple genes were identified involved in heavy metal metabolism, likely important for adherence to the stomach mucosa. And in European populations only, we see antibiotic resistance genes, particularly MRCA, likely because these are the only populations that really saw antibiotics in the recent history, or in any history for that matter. Okay, so I would say these results are interesting in the following ways. One, because this population is very old in the African population. And two, as I mentioned at the start, despite the fact that it has an extremely high prevalence in Africa, there's essentially no gastric pathology associated with it. So it's there, but it's not doing bad things to its host. For those of you who think about viruses at least, this might not be a huge surprise and really suggests a model that's potentially appealing that we're looking at more thoroughly now, which is this long association between helicobacter and humans in Africa has actually been accompanied by selection for reduced pathogenicity. And where the strong bottleneck out of Africa, where the population size is reduced, selective pressures are lessened, has actually freed helicobacter from this sort of constraint, which is now more pathogenic outside of Africa than in its original, where it originated. Okay, I'm going to end the helicobacter story on another thing that we're looking at now, which is completely different, but I think you'll find related. So this is a paper from a couple of years ago that we worked on with Ann and Alex Penis. And this is in humans, and this is just a structure plot. Okay, so hand population, chemo population, so Native American, Polynesians, et cetera. Okay, you see something that looks normal for human structure. These two individuals are two ancient DNA samples, sampled from a tribe that's, as far as we know, extinct in South America now, the Botokugo. Okay, these samples are pre-European contact. All right. And what you see when you sequence the ancient DNA from these guys is that they look like modern Polynesians. So just to say that again, because it's amazing, and not that many of you look amazed, so I'm going to try again. If you take ancient South Americans and sequence them, they look like modern Polynesians. At least these two. A few of you are baking surprise. Thank you. So this is pretty amazing to me, right? People that are old from here look like people that are modern from here. This hypothesis has actually been around for a while. Does anyone recognize this quote? Go on, Tiki. Thank you. So there was an anthropologist in the mid-20th century, Thor Hyerdahl, who also had this idea based on cultural information. He thought that there was a lot of similarities culturally between Polynesian indigenous Polynesian groups and indigenous South American groups and proposed that there could have been an oceanic route of colonization of the Americas potentially. People thought he was a loon. But in a move that was not particularly un-crazy, because it's still sort of a crazy thing to do, he built a balsa wood raft and showed that he could float across in around 100 days and survive. So a sort of experimental anthropological evidence for migration. But the ancient DNA is looking pretty interesting for this idea now. And the point of all of this is that based on the fact that species like Helicobacter are really well-recapitulate human history, for things like this that are very controversial, I would say that Helicobacter is actually a really interesting avenue to pursue. And we're starting to get samples now from indigenous populations from the Polynesian islands in South America to try and ask if this is a better way of looking at this question of whether we should be adding this arrow here and this sort of map of human migration. Okay, Helicobacter questions briefly before we move on to our next story. What year did those two individuals pass? They date to something between 900 AD and 1100 AD. The question was, I'm sorry, was how old are these two individuals? Probably between 900 and 1100 AD. Certainly before we have very strong evidence of European contact. The reason that's important is because there's a well-known Polynesian slave trade in South America beginning around the 1450s. So Europeans started capturing Polynesian islanders, bringing them to South America to work plantations. So if these guys dated from then, that would not be at all surprising. So the dating here is actually the key finding of the whole thing, which is why the dating is probably pretty good, but I wouldn't inhale completely deeply on that, of course, until we have other lines of evidence. I'm still putting a question mark here. This is a suggestive result, but nothing more of this time. John? Well, so Thor went the other way. Yeah. I do have it going that way. That's true because that's in principle what this observation is, right? We have people who look like they're from here who are sampled here. You know astutely that Thor actually went this way and hit the Polynesian islands. Yep, that's the way I had the arrow because that's sort of the result, but in principle that could be bi-directional at least. Or that was just, you know, a 100-day vacation round trip for Polynesian islanders in the 800s. Head over to Brazil for a few weeks and then come back. Okay. Let's move on to our next story. Sorry, I'm forgetting where to get the questions. Yeah, I'm supposed to repeat the questions for the microphone. I always forget that. Okay, our next story then is the human cytomegalovirus, another thing that lives in humans, the samovirus, HCMB for brevity from here on out. Just a few fun facts about CMB, so cytomegaloviruses exist essentially in all primates where we've looked from humans to green monkeys, but CMBs are strongly species-specific. So chimp CMB, what we call CCMB for obvious reasons, and human CMB, HCMB, is a process barrier as far as anyone can tell or anyone's ever heard. So they're really specialists. It's the herpes virus, like helicobacter, at least half of the global population has it. Almost certainly that's an underestimate. It's a DNA virus. It's about 235KB, the genome. We've estimated 200 open reading frames. That's a very conservative estimate. Primary infection usually occurs via mucosal surfaces, so if you're not born with it, as I'll address in a moment, you go to daycare or your first day of school and kids are spitting on you and peeing on you, and you have their CMB that transfers very easily with really any bodily fluid. Yeah, and once it's in the body, it remains throughout life and can be activated. So all the results I'm going to show you are from the last really five years of work, by an excellent postdoc in the lab, Nick Renzetti. Okay, so why is a population geneticist am I excited about CMB? Well, this is one very good reason. So I'm going to focus mainly on congenital infections, that is infections in utero, because this is a virus that actually can cross the placenta and invade tissues throughout the fetus. And in fact, CMB is actually the leading cause of infection-related birth effects in the world, as far as we know. So what happens is you have your population of mom here circulating in her blood and plasma. It crosses the placenta, infects the fetus, and does something very specific, which is it compartmentalizes. So there's multiple compartments here. I'm showing you here a sort of plasma compartment and a kidney compartment. The samples that we actually have information from are usually the kidney compartment because we can sample urine, the salivary gland compartment because we can sample saliva, and the plasma compartment because we can sample blood. We don't really know how many, either certain new brain compartments and heart compartments, but for some reason those are harder samples to get from newborns. So these are the three things that we get. Okay. So I'm going to extend this analogy a little bit that I started off with and say that the demographic history of HEMV within one person is not terribly unlike the demographic of history of humans around the world, by which I mean there's a large ancestral population, in this case, mom, in this case, Africa. There's a population size change associated with colonization. So just like all humans don't get up and leave Africa at one time, all viral population from mom doesn't get up and colonize the fetus at one time. There's subsequent colonization either throughout the new world or in the compartments of the fetus. Colonized populations may subsequently adapt to their new habitat, which is what I'm going to tell you a lot about now. There are very specific pressures whether you're living in northern Sweden or Central America just as there are very different pressures living in the kidney versus the salivary gland compartment if you're HEMV. And there's migrants exchanged between these populations. So they're not randomly mating, just as with humans, but there is some migration via plasma. Okay, so this analogy falls apart if you push too hard, but in general terms this is sort of an intuitive way to think about infection of a virus than a new host. So I'll tell you one simple result first and show you something more interesting, which is if you look at different patient samples so each of these columns is a patient, and this is sampled from saliva, from urine, from plasma, you already see that there's some very compartment specific effects by which I mean if you sample viral population from plasma in different individuals they group together very strongly. That is there seems to be a plasma specific environmental effect, just like there's a salivary gland specific effect whereas the urine seems to be a sort of filter for all compartments which is not a big surprise. So you observe many different types of things in urine but other compartments tend to be quite specific. Because of this and I'll explain the reasons why we think this is true in the moment there are many different levels of diversity within one patient for different compartments so this is just nucleotide diversity for a urine population and a plasma population in a patient. What's more exciting is the following. So this is FST. This is a major of differentiation between populations. This is one individual B-1-1 there's samples of two different kind points and you see through time it's diverging a little bit it's fixing mutations as you would expect but nothing too different. What's really interesting is if you look on this side so this is one patient this is a newborn B-103 sample essentially at birth one week and sample from two locations the urine and the plasma so the CMV population living in blood the CMV population living in kidney and what you see is that even by birth those two populations were very very differentiated and FST nearly 0.5 so very very different from another within one kidney. As different I should say as if you sample those two compartments from completely unrelated individuals so plasma in the kidney compartment between me and you is as different as within myself between plasma and kidney that's kind of wacky I think with this kind of data and following on my analogy with the colonization of the world you can use population genetics to try and estimate the demographic history of this infection. What we see when we do this is the following so you have your ancestral population here which is mom's viral population in the plasma which is what crosses the placenta we estimate this colonization bottleneck here which is around four months in utero there's a second bottleneck that you see in the urine compartment around five and a half months and then you can estimate rates of gene flow between these two you can estimate population growth population size all the same tricks we do in human, versatile etc we're doing here. What's sort of unique about this is virologists haven't really been able to piece this together using their tool sets so this is really the first clinically relevant insight into the timing of fetal infection when it has even occurred during pregnancy conditional on the fact that we've known for a long time that kids are already born with a fetus crossing the placenta sometime now we have a very good idea when those times are and how much gene flow there is between these compartments this is another way of looking at that selection result which I find interesting so this is just a normal looking tree where these are two patients B103M103 and you see they're plasma groups together whereas different patients here they're urinal groups together that's another way of saying that the urine environment is very very specific it's so specific in fact that on a sequence level on a nucleotide level the sequence is more similar between two urine compartments from two random individuals than between urine and plasma in one individual that's weird that's like saying that if I sample a high altitude human and the MLA's and the Andes they're more closer later to cross their full genome on their nucleotide level that's not true by the way I'm distinctly not saying that but this alteration is similar to that right the selective pressure of this environment is so strong that it drives these unrelated populations to be so similar on a sequence level so I would say a compelling suggestive example of very very strong parallel adaptation per host for each human these guys are infecting yes if you look across the whole genome there is a lot of if we saw this with the new president so you know if the variation are having any possible we can actually detect where these are and are there some more that's a real possible so the question is whether or not the variation is distributed uniformly or non-uniformly across the genome and it is distinctly not uniformly distributed so we have a nice figure that I don't include here where we just plot the genome and plot the level of divergence and variation and it's there are definite hot spots of divergence and well there's certainly mutationary variation also but once you account for that there's very clear hot spots because we don't have a ton of information functionally about what these open reading frames are doing we can say that these are clustering to certain places but we just don't have a very strong functional like to stand on in terms of what those places are actually doing I was thinking of seeing if you had some project on being in the ability to buy a data of those ones thinking more of a biochemical way in representation that would be an extremely helpful direction to go at this point do you know for sure that there is a single origin from ancestor of the one sample that you're in because these were two different combinations and that could explain the thing that we do know is that they were impacted from mom because that's the only contact they had when we sampled them at birth what we don't know which I think is probably more to the point of your question is whether mom is passing populations from her plasma and her ura and her kidney and then those are sort of sorting themselves out in fetus that's possible the reason that we don't think that's probably true is there's actually even a placenta specific population and so this really looks like this population with each step it's taking it's having to really adapt to that compartment and then move on to the next one and so there's a placental population also in the handful of cases where we actually can get our hands on that sample that looks very different from other compartments too so I think since that is the route of infection and that population is already pretty differentiated from the other compartments that's probably not what's going on probably they are uniquely moving across the fitness landscape as it were once they enter a new host so everything that we've looked at so far and that I've shown you has been in general infections I think we've had some important insights here there really is no good drug treatment for CMB but there are some experimental drugs that we're going to look at now in a framework not unlike what I'm about to tell you about in influenza with a real goal being to at least prevent that initial infection size that passes from mom we don't have a good way to stop it but we think we have some promising ways to at least reduce the size of the population during the initial infection and we're also trying to collect samples now from cohorts at daycares where there's horizontal infection so when they're not born with it but when they get it they take care and start asking about selection demographic sort of trajectories and histories when infection is horizontal rather than vertical okay any questions on CMB other than what we've had yeah just wondering so what function was under selection that's a tough question right it's a guess a lot of it it's a guess an amount of selection that makes me comfortable as a general neutralist very many locations that's right it's not just changing an open reading frame or two right it suggests that it's almost genome-wide I mean that's I mean it's weird I don't like it either but I just don't have another great explanation for this observation right now when you just when you build a tree and all urine compartments group together and there is more exposure later than the sequence level no matter which patients from where in the world do you pull I don't have an awesome better explanation than lots of strong positive selection for that right now that's important because these virus also stay in place and they have as large a very standard and that's probably the fourth virus in the 80s and 70s and the things that nobody agrees with is the the old aging expression in terms of aging expression which are also stratified depending on where the virus is in for example the epidermis in different tissue and they are programmed to be expressed in certain stratified processes in the production one thing we can see there is a synchronization I mean it would be nice to get it in a model where we could ask these sorts of questions these are essentially all natural population samples right now we don't have this in a cell culture like we don't influence where we could ask that question a little more specifically you could actually I mean people are using primate models for this actually which is just not work uncomfortable doing actually so we don't do that people are doing that Roman which compartment you're asking maybe there's a handful of regions that have to change for each compartment maybe there's 20 regions that are involved in compartment specific adaptation parallel all I can say is our changes there are hotspots of divergence but these changes are occurring largely genome-wide for these compartments like there aren't a small handful of regions that we see changing for a urine compartment versus a kidney compartment which would be reassuring if we did see that but we just were not John that may be a little bit of a question I don't know if it's possible to evaluate this but are you planning to go into this at autonomous versus autonomous sites or is it possible to just go to autonomous sites? so the question is whether there's difference divergence in synonymous versus non-synonymous sites so divergence at non-synonymous sites is about four times higher than divergence at synonymous sites genome-wide so this virus is just cranking along there is a all these viruses seem really quiet in this case which means that they will actually preferentially insert a third nucleotide at random that's the more wild case you might want to that's news to me actually that wouldn't explain parallel fixation potentially parallel fixation but we actually were looking at the prevalence of trial and failure so do you think it goes that way which is more of a one more than we'll do influence it so pathway was the question whether or not there's specific pathways being hit we just don't have great information on that we don't have the information like we have in Versailles to ask these sorts of questions we have a bunch of predicted open reading frames and a small handful of coding changes that we understand functioning what they're doing but we don't have the great functional information that we have in other organisms to ask on a pathway level at the nucleotide level we can see that and actually the fundamental observation that on the nucleotide level they're becoming more similar between compartments so it looks like many nucleotide changes have to happen to be a population living in a kidney compartment okay let's do our final story in influenza are we doing okay? yeah we're doing okay okay so this is I imagine the one of the three that you're most familiar with so it's a human pathogen it's an 8th segment RNA virus so it does not recombine so HMV recombines very heavily influenza does not but it re-assorts so it doesn't recombine within a segment but it can move these segments around okay in a way that looks to be under fairly strong selection actually the most common treatments are near-amididase inhibitors most commonly also Tamavir which is also known as Tamaflu which is what I'll talk about largely today but resistance has been observed in natural populations to evolve very quickly and the work that I'll show has come from these handful of papers over the last couple of years and it's been worked on by a number of members in the group okay so here we're taking an experimental evolution approach so it'll be different from what I've been telling you this is not a natural population sample but these are passages so the experiment just briefly looks like this so they're passages in MBCK cells so this is passage 1, passage 2, passage 3 at each passage we take whole genome population level sequencing so we have 13 time points because we have 13 passages so at each of these time points we have whole population, whole genome sequencing and the experiment is split at passage 4 where we have one population that's maintained as a control with no Tamaflu treatment and one population that's maintained with Tamaflu treatment and this is done in duplicate okay so just one word on inference here this is something that we think a lot about but I'm only showing on one slide today so fundamentally if you think about making inference from time sample data that is polymorphism data taking at multiple time points this is the information you have you have this is frequency and this is time so you have some frequency at time 1 and some frequency at time 2 and some frequency at time 3 you have that per site, that's fundamentally what you're trying to draw inference from the idea here that we proposed which I think seems to work pretty well is to use this distribution of this variance in allele trajectories across the whole genome so at every polymorphic site through your time sample to get a null distribution of this FS prime statistic which just captures that and if you remember introductory population genetics this expected variance through time is really a measure of effective population size right and thus if you have a measure of effective population size that essentially tells you how much another drift or how much variance you would expect through time if the mutation was completely neutral okay so we use every site in the genome to construct this null distribution and then we can look for site specific outliers that is are there sites where their change in allele frequency is too fast to be consistent with genetic drift too fast to be consistent with being a purely neutral mutation and we implement this in an approximate Bayesian framework which we call WF ABC right Fisher ABC okay so you do that and this is just a basic way of summarizing the result which is across the whole genome here each dot is an observed polymorphism above some cutoff frequency and the red line is statistical significance so you see that the great majority of things in this particular replicate of this experiment but this is true across all of them are completely consistent with genetic drift that is their change in allele frequency through time fits very well with our expectation of neutrality but you have a small handful of mutations that are going to be too fast to be consistent with that expectation and those mutations tend to cluster in H&A hemogutinin in your minidase as I told you Tamiflu also Tamivir is a near-medace inhibitor so we actually understand a fair amount about these mutations and I'll show you in a moment this is showing you the trajectories of those identified mutations so this is a comically small legend figure this is frequency this is time across our passages and I'm just showing you our mutations so these are the actual mutational frequencies that were significant using this ABC approach and replicate one with drug and replicate two with drug there is one overlap between the two it's this HG74Y which is a very well known Tamiflu resistant mutation from nature so that's a sort of nice sanity check that we re-identify the thing that we know does the job but we also identify a number of other interesting possible mutations particularly in your minidase that seem to have the same phenotypic effect so as population genocists we like to ask questions about the DFB and I think this is one really nice way sort of visualizing the larger scope of this experiment so here I'm showing you the distribution of fitness effects of beneficial mutations only so the beneficial tail of the DFB that is the selective effect of each observed mutation in our genome and what you see is in the control so when there's no drug treatment so zero here is neutral it's wild type like you see a very condensed distribution where you have many things the things that look beneficial are very close to neutrality that is they're only very weekly beneficial in the presence of drug you see a distribution that looks largely the same but you get this heavy tail so you have mutations that are 10%, 20%, even 30% that's advantage these are obviously resistance mutations these are the mutations that we identified using our approach I just overlaid some distributions here for any fissures geometric model aficionados in the audience who read a lot of Alan Orr distributions that people have proposed to capture these sorts of distributions but fundamentally what you see is a so called heavy tail distribution that is when you stress your population when you bring it further from optimum it's popping up to deal with this challenge this has been seen in yeast and pseudomonas in many of the organisms and it's very clear in our data set as well okay so what's the fundamental story there so also Tamabir resistance spreads quickly because it's only one or maybe two mutational steps away this is why global populations have a pretty easy time dealing with tamaflu so I just want to tell you a few results from something that's in review now which we think looks really cool which is looking at an alternative drug treatment called phobopuribir has a completely different mechanism of action so it interacts with the viral RNA dependent RNA polymerase and what it does in practice is it turns up the mutation rate in the virus population by a lot at least by an order of magnitude and more depending on how much phobopuribir you give it as I'll show you in a moment again that's obviously a very different mechanism from how also Tamabir is working this is not approved on patients though Japan has started stop piling last year, phobopuribir and I'm going to take the exact same experimental setup I showed you and statistical analysis I showed you and just repeat it except for phobopuribir okay so that's what I'm going to show I won't re-explain all the experimental design because they're the same okay here's one result so this is sort of a complicated figure there's a lot of information so this is through time these are passages this light colored line is drug concentration okay so drug concentration we're changing we're modulating throughout the experiment and these are essentially the growth rate of the viral population so what we see is as we start increasing drug concentration our populations are declining if we remove drug at this time the population recovers okay so we turn that mutation rate for a while we get rid of drug mutation rate goes back to normal and the population can recover it can start urging all of these mutations that have been introduced effectively so that's a withdrawal experiment a constant experiment we increase drug concentration to a certain point this is actually a relatively low concentration and then we just keep it uniform for the rest of the experiment this is an interesting replica like we see here the population is declining as you're turning up drug but there are some resistance mutations or rising in some replicates when you keep the concentration low that lets the population size start recovering so the population can repair its polymerase to some extent as long as you keep concentration very low so this is actually the first evidence that a population even can adapt to phobopiric treatment this will be bad news for some people the company ties in the world but we see it very clearly we see it in two replicates there is some adaptation to this treatment there's a lot of ways to show the big result I just chose the simplest which is as you turn up concentration the population crashes so I was showing you on the slide before quite low concentrations but you can see for populations raised at different drug concentrations there's a really pretty good relationship and once you're above 160 the population just crashes it's just overwhelmed by this input of deleteration mutations as I'll explain in a second and the population simply can't recover in all experimental populations that we've looked at anything over 160 dies it goes extinct well that's pretty good news in principle if you wanted to think about treating influenza virus with new therapeutics but this is really just an observation that I've shown you which is the population seems to crash when you add a lot of drugs so I just wanted to end on what we think are the actual evolutionary mechanisms that are going on here to give you hopefully some insight on what's actually happening in these populations so there's this term in literature mutational meltdown and this refers to the accumulation of deleterious mutations usually in small populations which leads to a subsequent decrease in fitness and population size which then allows more deleterious mutations population size is getting smaller and this is a sort of meltdown that this snowballs on you until you eventually hit extinction a mutational meltdown is really an effect it's an effect of these things happening and those things happening is driven by two processes which I'll just briefly summarize here Hill Robertson interference and Muller's ratchet so Hill Robertson interference is really quite simply on non-recombining chromosomes or segments in our case the efficacy of selection is reduced so beneficial mutations have a lower probability of fixing because they're linked to deleterious mutations and they can't recombine off of the same haplotype the population can't get rid of deleterious mutations as efficiently because they may be linked to beneficial mutations so all of these things are essentially just interfering with one another just like it sounds so the population is just less able to keep good things less able to get rid of bad things and in this way the population will decrease load of deleterious mutations over time and a decrease rate of beneficial fixation so as we turn that mutation rate and pop appear here what we're really doing is just throwing in more of these mutations the great majority of them are going to be deleterious rather than beneficial just because that's the shape of the distribution of fitness effects most mutations are bad rather than good and so you have more and more interference and you just can't keep your good things and you can no longer get rid of your bad things that's essentially what's happening here the other concept which is important is mother's ratchet so this is an idea that under models of mutation and drift your most fit segments or individual if you want to think of a totally non-recombining non-resorting genome can be lost and so what I'm showing you here is mutational classes so you can think of these as segments with zero deleterious mutations one deleterious mutation two deleterious three deleterious mutations and the arrows indicate that selection is preferring obviously these classes that have very few deleterious mutations on them and selection is trying to get rid of these that have many deleterious mutations on them so that's one process going on but mutation is always pushing you to the left so it's always pushing you off of your least loaded class off of your best class and then the idea is by mutation and drift you'll periodically lose your best class that is there will no longer be a zero class in your population and now your most fit individuals have one deleterious mutation rather than zero and this is called a click of the ratchet and a ratchet is the analogy because you can only turn it one direction right there's no going back once you've lost it so with each click of this ratchet here our population is getting less and less fit and so fitness decreases, population size decreases and because of that the ratchet starts clicking faster and this is another mechanism so again what we're doing is turning up mutation rate and so we're pushing these classes left faster than they were used to being pushed so just to summarize that little line of argument briefly in the presence of high concentrations of polypurevir we observe increasing rates of mutation in deleterious segregation we see decreasing effective population sizes which we can experimentally measure and population fitness and eventually extinction in all of our replicates and this is really as I've argued Hill-Robertson interference plus smaller ratchet is equally mutational meltdown in this case okay so just a general overall summary so I guess maybe one unifying theme is that whole genome time sample data is really changing a lot the kind of population dynamics that we can do in these critters helicobacter well I guess the main take home there is that the demographic history of helicobacter might actually be really interesting for thinking about questions of the demographic history of humans that we haven't been able to satisfactorily address thus far HMV in terms of population genetics is one of the best systems I know for studying models of subdivision selection and migration something that theoreticians like a lot and this is actually a really good empirical example of that and population genetics is getting actually unique clinical insights as well at the moment and in influenza this sort of framework that develops jointly, experimentally and statistically is really giving us pretty good insight on the fitness landscape with drug resistance and it's a really high throughput framework for testing additional drugs like Hovokiribir which are starting to look extremely promising with that this thing funding so statistical and theoretical development is funded by the FNS and the ERC and the virus work is funded by the Department of Defense and the Department of the Army and Army fan in the back and our lab is currently three PhD students and seven post-docs I mainly point out to people along the way but Valeria has really led the helical back to work Nick the CMB work and two former members Claudia Bank and Matthew Foll led the influenza analyses that I showed today and with that I'll just end on a brief advertisement as Diana said in 2017 the lab moving to ASU to the Center for Evolution and Medicine there's more information on our website if you like if you have interests all overlapping with mine and like I showed today the center is actually going to be a really unique and exciting place in terms of combining theoretical population with neurologists and anthropologists all united by this sort of common interest and Arizona's right grew up and it's actually a really pretty place that you've not been and nothing else it's 28 degrees there today so that's pretty good okay I'll take any additional questions