Session 10: Viral Evolution
Transcript of Part 4: Viral Evolution
00:00:01.18 Hello. My name is Harmit Malik, 00:00:03.29 and I'm an evolutionary geneticist 00:00:06.07 studying the molecular arms races 00:00:08.05 between primates and viral genomes. 00:00:10.11 I work at the Basic Sciences Division 00:00:12.10 at the Fred Hutchinson Cancer Research Center, 00:00:14.29 and what we hope to under, simultaneously, 00:00:17.15 is not just the evolutionary rules 00:00:19.23 that govern these interactions between primates 00:00:21.28 and viral genomes, 00:00:23.23 which tells us a lot about how we evolved 00:00:25.21 as well as how the viral pathogens 00:00:27.22 that we interact with evolve, 00:00:29.13 but we also would like to use these rules 00:00:31.13 to design better therapeutic strategies 00:00:33.16 to come up with sort of a better 00:00:35.26 antiviral intervention strategy. 00:00:38.10 So, in the second part of my talk today, 00:00:40.04 I'm going to talk about viral evolution 00:00:42.10 and how viruses might actually adopt 00:00:45.14 completely unexpected pathways 00:00:47.17 in order to evolve in these Darwinian arms races 00:00:49.26 between themselves as well as primate genomes. 00:00:54.12 So, a lot of the work on molecular arms races 00:00:56.20 is actually inspired by the fictional character 00:00:59.04 the Red Queen 00:01:00.22 that was introduced to us by Lewis Carroll in his book 00:01:03.06 "Through the Looking Glass", 00:01:04.26 and he pointed out that it takes all the running you can do 00:01:07.03 to keep in the same place, 00:01:08.28 which is what the Red Queen said to Alice. 00:01:11.04 This was actually recognized as 00:01:13.00 a really powerful evolutionary theorem 00:01:14.23 by the evolutionary geneticist Leigh Van Valen, 00:01:17.27 who formalized this into the Red Queen Hypothesis, 00:01:20.27 and he pointed out that 00:01:23.06 when two systems are basically antagonists of each other, 00:01:25.19 where they're both either competing for the same resources 00:01:28.11 or really taking advantage of each other 00:01:31.00 in order to gain an evolutionary dominant strategy, 00:01:33.22 they're going to be basically be always trying to evolve 00:01:36.21 to get the upper hand 00:01:38.15 in this evolutionary arms race, 00:01:40.11 forcing the other component to evolve. 00:01:42.08 And, essentially, they're just always climbing 00:01:44.12 this staircase over evolutionary adaptation, 00:01:47.00 effectively always staying in the same place. 00:01:49.09 So, there's going to be a temporary winner or a loser, 00:01:51.17 which in the next cycle of adaptation 00:01:53.20 gets switched around. 00:01:55.12 And so, this is a very typical situation 00:01:57.17 as it's seen in host-virus conflicts, 00:02:00.04 because this is exactly the type of situation 00:02:02.04 where we'd expect both the viral genome 00:02:04.04 and the host genome, 00:02:05.22 in order to be in conflict with each other, 00:02:07.22 which is why we actually refer to these 00:02:09.28 as the usual suspects. 00:02:11.06 And so, in this slide cartoon here, 00:02:13.06 we have a situation where the host protein 00:02:15.17 is either recognizing a viral protein 00:02:17.17 such that it can actually degrade it 00:02:19.15 and cleanse the organism of this infection, 00:02:22.00 or a viral protein actually acquiring 00:02:24.16 a single amino acid mutation 00:02:26.29 that allows it to evade detection by the immune system, 00:02:29.01 and thereby basically gains an advantage 00:02:31.15 by virtue of no longer being recognized. 00:02:33.23 And so, this is a situation in which 00:02:36.13 either the host or the virus is always losing this arms race, 00:02:39.26 and therefore there's always going to be 00:02:42.01 an evolutionary advantage to be gained by innovation. 00:02:44.13 Now, in a classical Darwinian sense, 00:02:46.14 this arms race can actually by typified 00:02:48.25 in a cartoon example 00:02:50.22 in a population sense. 00:02:52.03 So, here we have again, a very similar situation, 00:02:53.28 an immune surveillance protein, 00:02:55.24 let's say it's an innate immune defense gene, 00:02:58.06 which is actually surveying a population 00:03:00.17 of viruses represented by these coat proteins. 00:03:03.10 Now, very much like Darwinian selection, 00:03:05.19 a random mutation will arise, 00:03:07.24 which might actually affect 00:03:09.22 one of these coat proteins 00:03:11.11 such that it happens to have a mutation 00:03:13.13 that no longer allows it to be recognized 00:03:15.12 with the immune surveillance system. 00:03:17.12 This mutation could be extremely rare, 00:03:19.12 happening in one in a billion viruses. 00:03:21.28 Nonetheless, because all the other viruses 00:03:24.03 are being recognized 00:03:26.15 and cleansed out by the immune system, 00:03:28.11 very quickly this virus is going to take over the population, 00:03:31.01 and now the host is confronted 00:03:33.17 with a virus that is actually a variant, 00:03:35.23 and specifically a variant in an interaction interface, 00:03:39.19 forcing the host genome 00:03:41.22 to now come up with a counter-evolution strategy 00:03:44.09 which involves changes in the host protein. 00:03:46.29 So, most of the positive selection, 00:03:49.20 or most of the evolutionary adaptation, 00:03:52.02 that we've been considering 00:03:54.14 between hosts and viruses 00:03:56.16 has really been focused on these 00:03:58.19 very rapid amino acid replacements 00:04:00.21 in the host-virus interaction interface, 00:04:03.02 but today I'm actually going to tell you about 00:04:05.02 a completely novel strategy that viruses might use, 00:04:08.17 which actually has been hidden from view 00:04:10.26 from a lot of evolutionary biologists, 00:04:12.11 and we could actually capture that 00:04:14.07 by virtue of laboratory experiments 00:04:16.16 that could actually capture all stages 00:04:18.20 of this adaptation process 00:04:20.11 as it happened in a virus. 00:04:22.11 Before I tell you about that, I need to introduce 00:04:24.20 the particular system I'm going to describe, 00:04:26.08 and that's an antiviral gene 00:04:27.28 called protein kinase R, or PKR, 00:04:30.02 which is a very important innate defense system 00:04:32.06 against viruses. 00:04:33.24 So, PKR is actually expressed as an inactive monomer, 00:04:37.29 which means it's no longer active as a kinase. 00:04:40.15 A kinase is a protein that actually 00:04:42.21 puts phosphate moieties onto other proteins. 00:04:45.00 On interferon production, 00:04:47.00 you make PKR but you don't really mount a response. 00:04:49.09 It actually takes an actual viral infection 00:04:51.19 in that cell 00:04:53.23 in order for PKR to dimerize, 00:04:55.29 using the double-stranded RNA, 00:04:57.25 activate itself as a kinase, 00:04:59.21 and now phosphorylate its substrate eIF2α, 00:05:03.09 or elongation initiation factor 2α, 00:05:06.23 whose phosphorylation 00:05:08.23 will basically block protein production 00:05:10.26 through the ribosome. 00:05:12.18 So, this is a very potent block against viruses. 00:05:14.08 They can no longer go through their life cycle 00:05:16.10 if no protein production is allowed to proceed, 00:05:18.24 and this is basically... 00:05:21.03 they have invented all kinds of strategies 00:05:22.27 in order to block this PKR pathway, 00:05:25.11 including preventing its dimerization, 00:05:27.18 hiding away all the double-stranded RNA, 00:05:29.25 actually reversing this phosphorylation step, 00:05:32.04 as well as a completely eIF2α-independent form 00:05:35.14 of translation initiation. 00:05:37.19 We are very focused in the lab 00:05:40.06 on one particular type of antagonist, 00:05:42.06 which is this K3L antagonist 00:05:44.10 encoded by poxviral genomes, 00:05:46.06 which can essentially break the interaction interface 00:05:49.04 between PKR and eIF2α, 00:05:51.12 and by virtue of that essentially block the PKR pathway. 00:05:55.00 Now, in part one of the seminar, 00:05:56.29 I told you how PKR is actually undergoing very rapid evolution 00:06:00.08 in order to gain one step ahead of K3L. 00:06:03.15 What we also see is this arms race 00:06:05.28 is being played out both on the host side 00:06:08.06 as well as on the virus side, 00:06:10.00 so if you look at K3L among different poxvirus genomes, 00:06:12.26 in this case we compared 00:06:14.28 all the vaccinia proteins to all the smallpox proteins, 00:06:17.21 we have this nice histogram 00:06:19.29 of the rates of protein evolution 00:06:21.24 as they happen along the landscape 00:06:24.03 of the genome of poxviruses. 00:06:26.12 So, a very simple way to look at this histogram, 00:06:29.00 or this bar graph, 00:06:30.21 is that genes on the left-hand side 00:06:32.15 are very slow to evolve at the protein level 00:06:34.13 and genes on the right-hand side 00:06:36.17 are very fast-evolving at the protein level, 00:06:38.14 and K3L happens to be one of the fastest-evolving genes, 00:06:41.14 at the protein level, in poxviral genomes, 00:06:44.05 which means this very intense arms race 00:06:46.11 that has played out between PKR and K3L 00:06:48.20 has not only rapidly changed PKR in primate genomes, 00:06:52.19 but has also changed K3L 00:06:54.18 in [viral] genomes. 00:06:57.12 So, we wanted to actually capture the stages of adaptation, 00:06:59.21 and so to do that we actually turned to an 00:07:01.24 experimental evolution strategy, 00:07:03.12 really a very successful strategy 00:07:05.10 in terms of capturing evolutionary states 00:07:07.24 that might be very transient 00:07:09.21 and very difficult to capture in the wild. 00:07:11.25 This is a very important strategy 00:07:13.21 that's been very successfully used, for instance, 00:07:15.18 in bacterial evolution. 00:07:17.15 So, we took the vaccinia virus 00:07:19.14 and we actually made one change in that virus, 00:07:21.29 which is we knocked out this E3L gene, 00:07:24.20 which I've not introduced to you yet so far. 00:07:27.16 E3L is one of those proteins 00:07:29.13 that actually helps hide away the double-stranded RNA 00:07:32.17 to prevent the PKR activation, 00:07:35.12 so the reason we actually E3L 00:07:37.13 was we wanted to put all the selective pressure 00:07:39.21 to overcome the PKR response 00:07:41.28 onto the K3L gene, 00:07:44.03 and we knew, before we started the study, 00:07:47.07 that the vaccinia K3L gene 00:07:49.07 is actually ineffective at defeating the human PKR, 00:07:52.15 which is why, when you delete the E3L protein, 00:07:55.07 we have this dramatic drop in fitness 00:07:58.10 where the wild type [virus], 00:08:00.08 which contains E3L, 00:08:02.00 is almost 1000-fold better at infecting HeLa, 00:08:04.15 or human cells 00:08:06.04 than is this ΔE3L virus, 00:08:08.19 which has been deleted for the E3L gene. 00:08:10.27 So, what we decided to do 00:08:12.22 was simply take this virus 00:08:14.21 and passage it on a plate of HeLa cells, 00:08:17.16 and what happens when these viruses propagate 00:08:20.25 is that you basically make these small plaques, 00:08:23.28 which is where the virus has actually infected 00:08:25.25 and burst through and made more progeny viruses, 00:08:28.11 and we simply take all of these viruses 00:08:30.16 as they emerge from a plate 00:08:32.10 and transfer them to a new plate... 00:08:34.04 except, in the experimental evolution strategy, 00:08:36.20 we always take a historical record of this adaptation 00:08:39.26 by measuring the replication rate 00:08:42.00 at every step of this evolution, 00:08:43.27 as well as saving a fossil record of these viruses 00:08:46.15 at every stage of their adaptation. 00:08:48.18 So, when we now move these to new plates, 00:08:50.27 what we would hope to see is that the virus is getting better, 00:08:53.29 so we're going to see more and more plaques 00:08:55.19 as this virus learns to adapt to HeLa cells. 00:08:58.18 As a very important aside, 00:09:00.09 vaccinia virus 00:09:02.20 being passaged in chicken cells 00:09:04.13 was the basis for the smallpox vaccine, 00:09:06.17 which was responsible for perhaps saving 00:09:09.02 more lives than any other medical intervention 00:09:11.07 that we know of. 00:09:13.00 And so, what we did was simply passage these 00:09:15.18 in HeLa cells for about 10 passages, 00:09:17.17 and in just 10 passages 00:09:19.29 we observed something quite dramatic. 00:09:22.00 So, remember, the wild type fitness is about here, 00:09:26.20 the ΔE3L virus is about here, 00:09:28.20 and what we see is that, although all these viruses 00:09:30.26 started off really poor at infecting HeLa cells, 00:09:33.19 almost all of them, by 10 passages, 00:09:35.26 have really gained most of the fitness that they had lost 00:09:38.27 in terms of their HeLa cell infectivity. 00:09:42.03 So, we actually have multiple ways to test this. 00:09:44.12 This is actually a virus titer assay, 00:09:46.05 in which we see what the progeny virus count looks like, 00:09:49.28 but we've also replaced the E3L gene 00:09:52.06 with a β-galactosidase reporter gene, 00:09:54.09 and we actually measure levels of that reporter gene 00:09:56.13 as another means of actually assaying 00:09:59.00 how successful the virus is, 00:10:00.27 and both of these assays are very, very consistent 00:10:02.26 with each other, 00:10:04.28 suggesting that you started off with a very poor virus 00:10:06.24 and you've actually gained most of the infectivity back 00:10:09.18 in just 10 passages. 00:10:11.09 And so, what kind of rapid evolution 00:10:13.08 might have actually happened, 00:10:14.27 in the course of just 10 passages, 00:10:16.20 for vaccinia to have regained 00:10:18.12 most of the infectivity that it lost? 00:10:20.15 And so, to actually address the genetic basis 00:10:22.28 of how this happened, 00:10:24.16 we decided to actually take the parental strain 00:10:27.12 and sequence it to completion, 00:10:29.11 which means get very high, in-depth sequence coverage 00:10:33.15 to understand, okay, 00:10:35.12 what is the role that perhaps rare mutations are playing 00:10:37.18 in this adaptation? 00:10:39.09 Then we took these three replicates at passage 10 00:10:41.17 and sequenced them 00:10:43.15 such that we could compare. 00:10:45.00 Why are these three replicates 00:10:46.19 so much better than the parental strain 00:10:48.14 in terms of coming up with the solution 00:10:50.25 to HeLa infectivity? 00:10:52.26 So, when we first actually did this... 00:10:55.20 so, we could actually do this with very high coverage 00:10:57.17 because the vaccinia genome is about 200 kB, 00:11:00.12 and with advances in genome sequencing technology 00:11:03.09 we could essentially get about 1000-fold coverage 00:11:06.00 for every nucleotide of the vaccinia genome, 00:11:09.04 which means for any mutation at the level of 1%, 00:11:12.10 we can be very confident 00:11:14.09 that we are not going to miss it, 00:11:15.22 which is really what we wanted 00:11:17.07 to understand the basis for this evolutionary adaptation, 00:11:19.12 but, actually, the first returns 00:11:21.17 were very disappointing. 00:11:23.09 So, although we did see some really nice mutations in K3L, 00:11:26.20 which I'll return to, 00:11:28.23 we actually saw very low mutation 00:11:31.17 across the entire genome, 00:11:32.29 and that's actually consistent with the idea that vaccinia, 00:11:35.06 unlike other RNA viruses like influenza or polio, 00:11:38.03 is a very slowly-evolving virus. 00:11:41.08 So, we wondered, how is it that the virus, 00:11:43.16 which actually didn't acquire a lot of mutations, 00:11:45.24 and very few of the mutations that actually shared... 00:11:48.08 none, in fact, are shared across all three replicates... 00:11:51.10 how did it acquire this dramatic fitness gain 00:11:54.05 despite actually not having been able to 00:11:57.20 explore a lot of the mutation space, 00:11:59.10 for instance, that a rapidly-evolving RNA virus 00:12:01.10 might be able to do? 00:12:03.03 And so, this is the sort of conundrum 00:12:05.03 that really we were stuck at for a little while, 00:12:07.03 until a couple of people in the lab 00:12:10.13 really recognized 00:12:12.12 that we're actually only looking at some of the data 00:12:14.06 by looking at each individual mutation. 00:12:16.07 We have another readout 00:12:17.18 when we do these kinds of genome sequences, 00:12:19.25 which is we can look at how well is one part of the genome 00:12:23.06 represented across the entire sequence read. 00:12:25.25 So, for instance, 00:12:27.19 what we have here is an average genome coverage, 00:12:30.05 normalized to 1, 00:12:31.16 across the entire vaccinia genome. 00:12:34.00 You will see this very interesting blip right here, 00:12:36.13 and this is where we've actually deleted the E3L gene, 00:12:40.02 and so that's exactly what you'd expect 00:12:42.07 if the E3L gene is now missing 00:12:44.08 from what we are comparing to, 00:12:45.27 which is the reference sequence. 00:12:47.13 What really caught our eye, though, 00:12:49.02 was this dramatic blip upwards, 00:12:51.04 and when we took a closer look at these, 00:12:53.08 what these are are independent expansions 00:12:56.03 of the K3L gene 00:12:57.22 in every single replicate, 00:12:59.27 but not in the parental strain. 00:13:02.07 So, you can see the parental strain, shown here in blue, 00:13:05.11 is completely on the genomic average of 1, 00:13:07.12 exactly like its neighboring regions, 00:13:09.06 whereas every single one of the replicates 00:13:11.05 has an average K3L copy number between 3 or 4, 00:13:15.08 which is sort of a really dramatic example 00:13:18.00 of how each of these three replicates 00:13:20.08 has independently converged on the same evolutionary strategy, 00:13:23.14 which is to amplify K3L. 00:13:25.19 We were then wondering whether, basically, 00:13:28.05 we had viruses in here 00:13:30.10 that each have about 3 copies of K3L, 00:13:32.12 and it's a pretty homogenous population, 00:13:34.16 but now that we had the fossil record, 00:13:37.05 we could ask not only what the basis of this expansion was, 00:13:40.06 but when it occurred over the course of evolution. 00:13:43.18 And so, what we discovered when we did this fossil record 00:13:46.08 was we started with a parental virus that had no K3L expansions, 00:13:50.01 and for about 4 passages, 00:13:52.01 really we didn't see very much, 00:13:54.11 but as we went from passage 4 to 10, 00:13:56.20 we have this very heterogeneous virus population 00:14:00.09 with this accordion-like expansion of the K3L gene, 00:14:03.21 where you started with one gene 00:14:05.17 and now you've been ratcheting it upwards 00:14:07.19 with every passage, 00:14:09.16 increasing the average copy number, 00:14:11.13 but the average copy number is actually hiding the fact 00:14:14.01 that there are some viruses in here 00:14:16.04 who have undergone a 10% genome expansion, 00:14:19.15 which is a dramatic expansion for a virus, 00:14:21.28 where real estate is a really important criteria, 00:14:24.25 and they're doing so only focused on the K3L gene, 00:14:28.08 because that is the evolutionary strategy 00:14:30.12 they have come up with 00:14:32.09 to overcome this PKR response. 00:14:34.21 So, another way to actually describe what we see 00:14:37.05 is that we've been able to molecularly map the breakpoints, 00:14:40.11 they flank this K3L gene shown here, 00:14:42.24 and what we basically have is an accordion-like amplification 00:14:45.21 of this original duplication, 00:14:47.18 now to sometimes 15 copies 00:14:49.27 in these heterogeneous viruses. 00:14:51.26 So, this is a very dramatic 00:14:53.22 and very recurrent expansion. 00:14:55.22 I'm only showing you one of the three replicates we did, 00:14:57.21 but the other two replicates look almost exactly identical. 00:15:01.16 So, the fact that this expansion is so recurrent and so dramatic 00:15:04.29 led us to ask, what are the consequences 00:15:07.13 of this expansion? 00:15:08.27 So, we had multiple genes... 00:15:11.00 are they actually making a lot more protein 00:15:12.26 than what you'd expect? 00:15:14.11 And, indeed, to test that, 00:15:15.29 we actually took these passage 10 viruses 00:15:18.00 and transfected them again back... 00:15:20.04 infected them into HeLa cells, 00:15:22.03 and indeed what we see 00:15:24.01 is they are making a lot more K3L 00:15:26.05 than even wild type virus, 00:15:28.23 and if you actually blow up this picture 00:15:30.28 you can see that the parental E3L gene 00:15:33.09 is making very little K3L 00:15:35.02 compared to what is now being made 00:15:36.29 by virtue of this genomic accordion expansion 00:15:39.14 in these replicate (passage) 10 viruses. 00:15:42.13 We can now ask, okay, 00:15:44.00 we now have this K3L expansion, 00:15:46.01 is this the reason 00:15:47.19 why we're seeing this massive increase in fitness? 00:15:49.16 And, to do that, 00:15:51.05 what we did was a strategy in which 00:15:53.01 we can take small interfering RNAs 00:15:55.05 and essentially get rid of most of the K3L RNA 00:15:58.06 that is being produced in these infected cells. 00:16:00.23 And so, when we do that, 00:16:02.20 we can design RNAs and then infect vaccinia, 00:16:05.17 and when we see that what we can see is 00:16:08.02 that these siRNAs are quite effective. 00:16:10.10 So, here's the non-siRNA-inhibited replicate C, 00:16:15.12 at passage 10, 00:16:16.24 and here are a multitude of different siRNAs 00:16:18.28 that basically, to a different degree, 00:16:21.09 knock down the total levels of proteins, 00:16:23.16 and when we compare the fitness 00:16:25.09 of these knockdowns, 00:16:27.25 versus no knockdown or a scrambled siRNA, 00:16:30.02 we can see that when you knock down the K3L protein production 00:16:34.04 you essentially knock down all the gains of fitness 00:16:36.24 that you've gained over the passage 10. 00:16:38.29 So, this K3L accordion-like expansion 00:16:41.22 is both necessary and sufficient, 00:16:45.17 really, to explain this massive increase in fitness 00:16:47.21 that we saw in our laboratory. 00:16:49.29 So, there is of course the trade-off... 00:16:51.28 I mean, these are viruses that actually 00:16:54.09 usually prefer really compact genomes, 00:16:56.15 and the tradeoff is really apparent 00:16:58.10 when we make a comparison of 00:17:01.00 these passage 10 viruses in HeLa cells 00:17:03.01 versus hamster cells. 00:17:04.24 You can see, in HeLa cells, 00:17:06.15 each of the replicate viruses 00:17:08.08 is doing a lot better 00:17:10.04 than the parental virus, 00:17:11.19 whereas in hamster cells, 00:17:13.10 these viruses are actually doing worse 00:17:15.08 than the parental virus. 00:17:16.18 So, what's going on? 00:17:18.12 It turns out that vaccinia K3L, 00:17:20.05 at the starting point, 00:17:22.00 is ineffective to defeat human PKR, 00:17:24.01 and it needs this massive gene expansion 00:17:26.09 in order to overcome, biochemically, 00:17:28.22 the inhibition encoded by the PKR protein, 00:17:31.17 whereas even a single-copy vaccinia 00:17:33.22 is able to overcome 00:17:35.29 the PKR inhibition encoded by hamsters. 00:17:39.03 And so, what we have now begun to see is, 00:17:41.16 if you take this accordion-expanded virus 00:17:44.06 and now infect BHK, or hamster cells, 00:17:47.12 the accordion has now begun to collapse, 00:17:49.11 by virtue of the fact that 00:17:52.03 the fitter virus is actually the smaller virus. 00:17:54.00 And so, this is an example 00:17:55.26 where we've got this transient expansion in HeLa cells, 00:17:58.01 which is now going the opposite way 00:18:00.01 in hamster cells. 00:18:01.23 So, I'll return to those mutations 00:18:03.18 that we actually first detected, 00:18:05.06 which were so disappointing because they were not recurrent, 00:18:08.03 but one of those mutations was especially interesting 00:18:10.17 to us because it occurred in the K3L gene. 00:18:13.02 It's present at about 3% frequency in replicate A 00:18:16.00 and at 12% frequency in replicate C. 00:18:19.18 This is a mutation in the 47th amino acid, 00:18:22.07 changing a histidine to an arginine. 00:18:24.25 The reason this is really interesting 00:18:26.25 is because a completely independent assay 00:18:29.05 many, many years earlier, 00:18:31.05 from Tom Dever's group, had done a yeast selection experiment 00:18:34.16 in which they wanted to ask, 00:18:36.18 can we do a mutational experiment 00:18:39.10 asking, what mutation in K3L can actually overcome 00:18:42.04 the inhibition encoded by PKR 00:18:44.14 and allow this yeast growth to recover? 00:18:47.10 If you want to learn more about this yeast growth assay, 00:18:49.21 I suggest you watch part 1 of the seminar. 00:18:52.14 What is really interesting is they come up 00:18:54.16 with one mutation, H47R. 00:18:57.05 We have done a completely independent experiment, 00:18:59.22 in vaccinia infections, 00:19:01.25 and vaccinia is basically also telling us 00:19:03.28 that this is the evolutionary solution 00:19:06.05 that vaccinia has come up with 00:19:08.04 in completely different assays, 00:19:10.02 both in yeast and human. 00:19:11.27 So, what that means is, now, 00:19:13.12 you started off with a K3L 00:19:15.06 that was not able to defeat human PKR, 00:19:17.03 and now you've acquired a single amino acid mutation 00:19:19.20 that in a Darwinian sense 00:19:21.26 is able to defeat human PKR. 00:19:23.26 But, because we actually now have 00:19:26.05 two forms of adaptation, 00:19:27.26 this gene accordion model that we've discovered, 00:19:30.04 and this classical Darwinian adaptation model, 00:19:32.06 we could now ask, going back to the fossil record, 00:19:34.23 which occurred first, 00:19:36.10 and did one depend on the other? 00:19:38.08 And, when we basically do that, 00:19:40.08 by looking at when one type of mutation 00:19:42.18 occurred relative to the other, 00:19:44.14 what we find is that the expansion 00:19:47.07 actually already happened by passage 4, 00:19:50.00 whereas this mutation actually 00:19:52.08 only began to occur around passage 5 and 6. 00:19:55.01 Moreover, many of these H47R mutations 00:19:57.28 actually occur in these already expanded accordions, 00:20:01.18 which strongly suggests that after the accordion expansion 00:20:05.11 you actually increase the mutational probability 00:20:08.25 of acquiring an H47R mutation, 00:20:11.06 which now, by virtue of even a single copy, 00:20:14.06 is able to overcome the PKR response. 00:20:18.05 So, that actually leads to a very interesting suggestion 00:20:20.22 in terms of how these Red Queen conflicts 00:20:23.02 might actually play out in evolution. 00:20:25.13 We think of the K3L-PKR interaction 00:20:28.24 as a classical Darwinian arms race, 00:20:31.05 so starting off with a step 00:20:34.05 in which the vaccinia K3L is not able 00:20:36.02 to defeat the human PKR, 00:20:38.02 we basically acquired sort of a transient amplification 00:20:41.09 of the K3L gene, 00:20:43.15 which then allowed for the selection 00:20:45.19 of the H47R mutation, 00:20:47.15 which was able to defeat human PKR. 00:20:50.00 Now, what's really interesting 00:20:51.25 is now we have a single-copy gene 00:20:54.11 that is able to defeat human PKR, 00:20:56.08 and we are very interested in asking, 00:20:58.19 now that you've acquired the right mutation, 00:21:00.12 will the accordion collapse to mitigate 00:21:02.10 all the fitness costs of this gene expansion? 00:21:04.21 So, the reason I put this cartoon up 00:21:06.27 is because this cartoon should remind you 00:21:09.02 a lot about this cartoon of the classical arms race, 00:21:11.09 the way we think about in terms of 00:21:14.07 virus antagonizing humans, 00:21:16.00 but we might actually be missing this very, very important 00:21:18.10 transient step which involves gene amplification, 00:21:21.18 especially in viruses and pathogens 00:21:23.18 that actually don't have the high mutation rates 00:21:26.10 necessary to sample the adaptive landscape. 00:21:29.16 And so, very much like 00:21:32.05 the ability for influenza to undergo chromosome reassortment 00:21:36.04 in order to infect new hosts, 00:21:38.10 we think that gene amplification 00:21:40.09 is one of the critical strategies 00:21:42.06 that large double-stranded DNA viruses 00:21:44.06 actually might be using in order to stay... 00:21:47.05 keep pace with this sort of rapid evolution 00:21:49.12 of the host genes that they're actually antagonizing. 00:21:52.22 So, with that, 00:21:54.12 I'm going to acknowledge the people who did the work. 00:21:56.05 All of this work was actually done by a former postdoc in the lab, 00:21:58.20 Nels Elde, who now heads his own lab at the University of Utah, 00:22:02.22 in collaboration with Emily Baker and Michael Eickbush. 00:22:06.09 We do all of our poxviral work 00:22:08.14 in collaboration with my senior colleague Adam Geballe, 00:22:10.20 and I'd especially like to acknowledge 00:22:12.13 Stephanie Child, from his lab, 00:22:14.20 who really did all of the poxviral infection experiments 00:22:16.20 in collaboration with Nels. 00:22:18.15 I'd like to thank Tom Dever, 00:22:20.08 Welkin Johnson, 00:22:21.20 and Michael Emerman, 00:22:23.10 for a lot of reagents and help, 00:22:24.24 and I'd especially like to thank Jay Shendure 00:22:26.24 and Jacob Kitzman, from his lab, 00:22:28.20 for helping us with the analysis of these poxviral sequences. 00:22:31.25 And thank you, I hope you had a good time listening to this.