Gene Drive

Part 1: Gene Drive

00:00:07.05 I'm Kevin Esvelt.
00:00:08.13 I'm an assistant professor at the MIT Media Lab and Leader of the Sculpting Evolution group.
00:00:13.23 Now, my group works on engineering organisms, not just in the laboratory but also in the wild.
00:00:20.24 Now, humanity has been engineering organisms for a very long time.
00:00:24.15 We have, of course, selectively bred wolves into a tremendous variety of different dog breeds.
00:00:30.18 All of our crops have been selectively bred.
00:00:33.02 But what we need to realize is that when we alter an organism, we are diverting its resources
00:00:39.00 for our benefit, and away from its need to survive and reproduce in its ancestral habitat.
00:00:45.06 Darwin said "Man selects for his own good, Nature for that of the being which she tends."
00:00:50.23 And this is absolutely true.
00:00:52.26 Now that we have precise genome editing tools, such as CRISPR, the same rules apply.
00:00:58.00 That is, even though we can now use CRISPR to cut a particular site in the genome
00:01:02.13 and insert a desired sequence that allows us to change a particular trait of the organism,
00:01:09.13 we're still altering that organism to benefit us.
00:01:12.13 And that typically reduces its ability to survive and reproduce in the wild.
00:01:16.16 In other words, no matter how we edit an organism
00:01:19.16 -- the traditional selective breeding or modern genome editing --
00:01:23.19 we still expect it to be outcompeted by wild counterparts.
00:01:32.17 That is, our default expectation for any organism altered by humans is that,
00:01:38.07 in the ancestral habitat, natural selection will eliminate the altered genes.
00:01:47.13 But, that's only true for genes that play by the rules.
00:01:51.17 Some genes cheat at inheritance.
00:01:54.10 So, even though they might reduce the odds that the organism will reproduce,
00:02:00.07 they're more likely to be passed on to all of the offspring.
00:02:05.26 When this happens, we call it "gene drive".
00:02:10.01 Gene drive is a naturally occurring phenomenon that happens when a vertically transmitted genetic element
00:02:15.04 -- that is, something transmitted exclusively from parents to offspring --
00:02:20.02 reliably increases in frequency in a population, even though it does not help individual organisms
00:02:26.13 to survive and reproduce.
00:02:29.02 Gene drive is ubiquitous in nature.
00:02:31.27 Evolution invented it hundreds of millions of years ago.
00:02:35.10 Prominent examples include the gene drive responsible for the fact that the cow genome
00:02:40.03 is 25% snake, due to an ancient gene transfer event mediated by a virus in a tick that moved
00:02:48.02 a "jumping gene" gene drive that copies itself within genomes from snakes into cows.
00:02:54.22 Which then replicated until it was a quarter of the cow genome today.
00:03:00.17 More than half of the human genome comprises broken gene drive elements,
00:03:05.04 as well as a few active ones.
00:03:07.12 It's nearly impossible to find the genome of any species that doesn't have any signs
00:03:12.20 of an active or broken gene drive element within it.
00:03:16.02 They are utterly ubiquitous within the natural world.
00:03:19.18 And now, thanks to technologies such as CRISPR, we're on the verge of effectively harnessing them
00:03:25.03 to deliberately change wild populations.
00:03:31.08 CRISPR can be used to build a self-propagating gene drive.
00:03:35.19 Instead of simply delivering DNA encoding the change we want to make in that organism,
00:03:40.07 we also encode the CRISPR system and the instructions for making that change.
00:03:46.19 That is, we deliver into the reproductive cells DNA encoding the change and the CRISPR system.
00:03:53.08 CRISPR then cuts the target site in the genome and the entire DNA cassette is then inserted.
00:03:59.01 So now, this reproductive cell of the organism has the instructions for doing genome editing
00:04:03.16 on its own.
00:04:04.16 So, CRISPR components are produced.
00:04:06.10 They cut the other chromosome and copy over the DNA.
00:04:11.17 So now, there are two copies.
00:04:14.01 Any organism with two copies of a gene in its reproductive cells is guaranteed to
00:04:19.20 pass that gene on to all of its offspring.
00:04:22.01 So, when this engineered gene drive organism mates with a wild counterpart,
00:04:26.12 all of the offspring are guaranteed to inherit the drive system.
00:04:31.06 And in the reproductive cells of those offspring, editing happens again.
00:04:36.11 CRISPR cuts the wild type version inherited from the wild parent, and copies over the edited DNA.
00:04:46.02 That means that those organisms are guaranteed to pass it on to their offspring.
00:04:51.14 And editing happens again and again and again, down through the generations, until, in principle,
00:04:57.26 all organisms within that population have inherited the altered trait.
00:05:03.06 So, the key difference between a normal engineered gene and a standard self-propagating gene drive
00:05:10.24 is that, whereas our default expectation for a standard gene is that natural selection
00:05:15.02 will eliminate it from the wild,
00:05:18.03 our default expectation for a standard gene drive is that it is likely to spread in wild populations.
00:05:25.20 There are some requirements for gene drive systems.
00:05:28.08 The kind... this kind, the kind that we can make with CRISPR, requires sexual reproduction.
00:05:33.23 It doesn't work in bacteria and viruses.
00:05:37.04 Its ability to spread, and how fast it spreads, depends on the generation time of the organism.
00:05:41.23 Again, it's transmitted parents to offspring exclusively.
00:05:45.26 It also requires gene flow between different populations of the organism.
00:05:53.03 Gene drive can be used either to alter or, in some cases, to suppress populations.
00:05:58.17 And these have different potential uses.
00:06:01.00 So, can this really work?
00:06:03.13 Well, CRISPR-based gene drive at this time has been conclusively demonstrated
00:06:08.19 over multiple generations to work in four different species:
00:06:12.05 yeast, fruit flies, and two species of mosquito.
00:06:15.04 But we have every expectation that it will work in most organisms in which
00:06:18.22 CRISPR genome editing works, which is most sexually reproducing organisms.
00:06:23.13 So, how does the one build this kind of gene drive, a standard gene drive?
00:06:28.16 Well, in order to make it stable, you need to identify an important gene,
00:06:32.13 a gene that's important for the fitness of the organism.
00:06:35.12 Then you build a drive construct that recodes the tail end of that gene and inserts guide RNAs
00:06:42.25 that instruct CRISPR to cut the wild type version.
00:06:47.22 You also encode the CRISPR nucleus and whatever genetic cargo you want this gene drive
00:06:52.11 to spread through the population.
00:06:54.18 Then, in the cell, this gene drive system will produce the CRISPR components that then
00:07:03.00 cut the important gene at multiple sites.
00:07:06.17 Then... this creates a double-strand break within that important gene,
00:07:11.01 and then the cell has to repair that before it divides.
00:07:13.17 It can do so using one of two mechanisms.
00:07:16.02 If it goes through homology-directed repair, then it copies the gene drive cassette.
00:07:21.23 We call this homing, when this copying event occurs.
00:07:25.02 And this is how you go from one copy of the drive system to two copies.
00:07:28.03 This is how it cheats inheritance.
00:07:30.07 But there is another repair pathway.
00:07:31.17 The cell can also just jam the broken ends of DNA together.
00:07:34.20 But this is why we're cutting an essential gene at multiple sites.
00:07:38.25 If it jams the ends together, it will delete a section of this important gene.
00:07:44.01 And that means that this product will be costlier than the gene drive.
00:07:48.12 That is, natural selection will favor the correct gene drive over this broken product.
00:07:54.11 And this is required to build a stable gene drive system.
00:07:57.15 This has been backed up by three studies by independent groups,
00:08:01.03 as well as a couple of preprints suggesting that
00:08:04.15 using multiple guides cutting different sites will in fact
00:08:08.15 preclude this form of resistance.
00:08:11.11 So, what can you do?
00:08:13.17 Again, we mentioned that gene drives can alter populations.
00:08:17.08 If that's used to delete existing genes, then the result is evolutionarily stable.
00:08:21.24 No matter what happens, if CRISPR cuts that target gene and replaces it with itself,
00:08:26.23 or if it's deleted, the gene is gone either way.
00:08:29.16 This can be done reliably.
00:08:31.17 It can also be used to add new genes, new cargo genes, or to edit pre-existing genes.
00:08:38.12 But these kinds of changes are not necessarily stable, because they're not linked to
00:08:43.19 the function of the drive system.
00:08:45.01 The drive can spread a broken version just as well as it can an intact version.
00:08:49.02 So there is no selection pressure for maintaining the exact change that we want to spread.
00:08:55.27 Gene drives can also be used to suppress populations.
00:08:59.15 For CRISPR-based drives, two methods were put forward.
00:09:02.24 These were originally posited by Austin Burt in the pre-CRISPR era.
00:09:06.15 In version 1, you ensure that all organisms that inherit the gene drive system
00:09:12.18 develop as one particular sex.
00:09:14.23 So, as the gene drive spreads through a population, all of the organisms become that particular sex
00:09:19.12 until the population crashes.
00:09:22.24 In the second method, the gene drive system is constructed such that inheriting one copy
00:09:28.14 results in a fertile organism that typically will pass on the gene drive system to all of its offspring.
00:09:34.26 But inheriting two copies -- one from each parent -- produces organisms that are
00:09:38.08 either not viable or, better yet, are not fertile.
00:09:42.14 So this means that the gene drive system spreads rapidly when rare.
00:09:45.28 And when it becomes common, most organisms are sterile and the population crashes.
00:09:50.01 Crucially, however, models predict that gene drive organisms can crash populations to low levels
00:09:55.25 but cannot remove them entirely without additional human help.
00:10:00.06 And this is because the gene drive system is not introduced into every subpopulation simultaneously.
00:10:05.17 Even if it succeeds in removing one small subpopulation, that area is likely to be
00:10:10.24 recolonized by organisms that have not yet been affected.
00:10:13.15 So, the net outcome is that the population is maintained at a very low level over time.
00:10:20.08 So, what are the potential benefits of gene drive?
00:10:24.07 That is, why would we want to alter wild organisms?
00:10:29.26 The main reason is our current methods for dealing with wild creatures that pose problems
00:10:36.05 are not very environmentally friendly.
00:10:38.12 We tend to rely very heavily on large-scale environmental engineering, using bulldozers
00:10:43.19 draining swamps, and other sorts of methods.
00:10:46.26 Or we use chemicals that have been tuned to attack entire classes of pests,
00:10:54.09 as well as most organisms reasonably closely related to those pests.
00:10:59.01 With gene drive, we could conceivably alter wild organisms in ways that are very precise,
00:11:05.25 and make the smallest possible change that we think could solve the problem.
00:11:10.06 For example, we might immunize animal reservoirs of human diseases so that they could not serve
00:11:15.11 as reservoirs for those diseases.
00:11:17.10 We could tweak vectors of human disease so they're no longer interested in biting humans,
00:11:21.20 but otherwise go about their ecological role.
00:11:24.05 We could precisely suppress populations of invasive species.
00:11:29.03 We could, similarly, reduce fecundity in animal populations, thereby reducing starvation
00:11:38.11 and predation and animal suffering, especially for organisms that we introduced.
00:11:44.02 And finally, in agriculture, instead of spraying toxic pesticides throughout our fields,
00:11:49.08 we could instead tweak the pests so they no longer like the taste of the crop.
00:11:53.25 This is far away.
00:11:55.05 We still need to learn a great deal more about how, for example, insect olfaction works.
00:12:01.01 But in the long run, this is a much more eco-friendly way to go about solving ecological problems
00:12:06.07 in the wild, if it can be done cautiously and carefully.
00:12:11.25 So, what are the potential risks?
00:12:13.23 What could go wrong?
00:12:14.23 Well, this is very hard to talk about because, ecologically speaking,
00:12:18.16 the effects are going to depend entirely on the nature of the species and the particular ecosystem
00:12:24.20 in which that change happens.
00:12:26.14 What is being done?
00:12:27.22 And how might that affect the relationships of that species with all of the others?
00:12:31.08 It will vary entirely on a case-by-case basis.
00:12:34.22 But there are also social risks with gene drive because, to put it mildly,
00:12:39.20 people are not terribly fond of the idea of genetic engineering,
00:12:43.00 especially not when it comes to wild organisms.
00:12:46.09 So, there is a serious danger of a potential lack of public support and a loss of trust
00:12:52.25 in both science and governance should scientists attempt this in ways that are perceived
00:12:56.25 as unethical, or if anything should go wrong.
00:12:59.26 But one might wonder, why isn't gene drive of biosecurity risk?
00:13:03.27 That is, this is the ability to alter a wild population.
00:13:06.27 And a standard gene drive will spread indefinitely.
00:13:10.17 Isn't that a threat?
00:13:12.19 And the answer is, probably not.
00:13:15.02 Gene drive is a technology that inherently favors defense, for three reasons.
00:13:19.15 First, it's slow.
00:13:20.21 It necessarily takes many generations to spread.
00:13:23.23 Second, it's obvious if you look.
00:13:26.19 That is, if you sequence the genome, in the natural population you should never see eukaryotic
00:13:32.17 expression signals that turn on gene expression adjacent to CRISPR components,
00:13:38.18 because CRISPR comes from bacteria.
00:13:40.05 So, if you ever see a sequencing read that shows two of those elements together,
00:13:46.09 you know it is unnatural.
00:13:48.03 That signature cannot be hidden.
00:13:49.20 So, it's slow.
00:13:51.12 It's obvious if you look.
00:13:52.22 And it can be countered, and reliably so.
00:13:56.19 The first report of a CRISPR-based gene drive was in yeast.
00:14:00.10 And when we did this, we first built a gene drive that disrupted a target gene,
00:14:05.17 thereby changing the traits of the yeast.
00:14:08.22 We also built a second gene drive that cut the first one and restored a functional copy
00:14:15.02 of that broken gene, thereby restoring the broken trait.
00:14:18.27 And the first drive worked, nearly perfectly, at ensuring inheritance of the broken trait.
00:14:25.06 And the second drive, nearly perfectly, restored that broken trait to functionality.
00:14:31.08 So in principle, because CRISPR can target essentially any sequence, any gene drive that
00:14:37.27 one person makes and releases can be overwritten,
00:14:41.22 and its genetic effects undone by a second gene drive.
00:14:45.02 So, this should be able to restore the original phenotype of the organism.
00:14:50.04 But it can't necessarily restore the exact original DNA sequence.
00:14:55.06 Efforts to enable that kind of change are currently underway,
00:14:57.22 but they haven't yet been demonstrated.
00:14:59.23 And it's important to note that just because we can restore the population back to
00:15:03.10 the way it was originally -- in terms of its relationships with other organisms, its behavior and traits --
00:15:07.27 that doesn't necessarily mean that every ecological change that might have occurred
00:15:12.02 will necessarily be reversed, because ecosystems are very complex.
00:15:18.11 In summary, though, fortunately, gene drive does not seem to be a major biosecurity threat,
00:15:23.10 just because anything that is slow and obvious and easily blocked just isn't a major physical threat.
00:15:29.21 But the social risks of gene drive and the potential social threats are very, very high,
00:15:34.20 because people's perception of something doesn't depend on an actual physical change.
00:15:40.26 So, a very relevant question is, how many organisms carrying a standard gene drive
00:15:47.10 have to escape a laboratory or be released in order for that gene drive system to invade a wild population?
00:15:54.09 The answer, somewhat disturbingly, seems to be very, very few.
00:15:58.19 That is, if you have just two organisms released, there's a non-trivial chance that
00:16:03.25 that gene drive system will spread in that wild population.
00:16:07.09 And as the number increases, you can approach near certainty.
00:16:11.07 And the models that demonstrated this, that predicted it to be the case, took into account
00:16:16.17 a large number of factors that had been predicted to block this, including genetic level resistance
00:16:23.08 to the gene drives, whether pre-existing in the population;
00:16:27.09 the presence of inbreeding;
00:16:29.18 other barriers to gene flow.
00:16:31.07 All of these increased the number of organisms needed to invade, but they did not prevent it entirely.
00:16:38.02 Even gene drive systems that are not stable, that get outcompeted by their resistant alleles,
00:16:45.12 that can't spread above 40% of the organisms within one population, can still invade population after population.
00:16:54.00 Never rising above 40% in any one, but ultimately affecting every population.
00:17:00.04 In general, we can plot the relationship between how efficient a gene drive is
00:17:05.06 -- that is, how efficiently is it copied from one chromosome to the other --
00:17:09.06 against the migration rate between two populations
00:17:13.00 -- that is, out of the number of organisms in one population,
00:17:16.06 what fraction of them move to the adjacent population in each generation.
00:17:20.18 And what we find is that highly efficient gene drive systems, such as those
00:17:24.24 demonstrated in mosquitoes that are copied at 95% of the time room, or more often,
00:17:29.16 are predicted to invade all populations connected by gene flow
00:17:32.23 where you have one organism per million moving between those populations.
00:17:37.22 So, what this means is that models predict that a single laboratory accident that releases
00:17:44.26 just a handful of organisms bearing this kind of gene drive into a wild population
00:17:50.16 could devastate public trust in both science and governance.
00:17:54.12 And this would probably delay all applications of gene drive by many years, because one can
00:17:59.09 imagine terrible headlines such as "Scientists accidentally turn entire species into GMOs --
00:18:06.14 is CRISPR to blame?"
00:18:08.15 And that, in turn, would devastate public trust in applications of CRISPR and gene therapy,
00:18:16.00 gene therapy which already has been set back by over a decade due to a tragedy
00:18:21.23 in an unwise and ill-planned clinical trial.
00:18:24.00 An accident involving gene drive would be much worse.
00:18:27.10 Fortunately, the scientific community has been on top of this in agreeing to recommend the use,
00:18:32.15 and demonstrate the use, of laboratory safeguards that allows research on this kind gene drive
00:18:38.22 to be pursued safely.
00:18:40.17 And these include molecular forms of confinement... only program drive systems to cut sequences
00:18:46.00 that are not found in wild populations; separate the components so that not everything necessary
00:18:52.00 for copying is copied when the drive system spreads.
00:18:56.22 And you only do this research in laboratories far from populations of the target organism,
00:19:02.19 such that even if they do escape the lab there's no one to mate with.
00:19:06.15 Above all else, don't rely solely on barrier confinement, because people make mistakes,
00:19:11.22 and sometimes organisms can move.
00:19:15.08 So, what does that mean for scientists who are interested in working with gene drive?
00:19:20.20 It means they should probably only build this kind of standard gene drive system
00:19:25.27 if they're working on a problem for which the goal requires affecting an entire species.
00:19:32.12 And there are only a few such potential applications.
00:19:35.11 And of them, really only one has any kind of chance of international agreement to go ahead and use it.
00:19:43.27 And that one is the eradication of malaria, because malaria kills half a million people every year,
00:19:50.16 most of them children under the age of five.
00:19:53.20 And there are very few possible paths to eradication that don't involve some form of gene drive
00:19:59.13 targeting the African vectors.
00:20:01.23 So, here is a problem for which this kind of standard self-propagating gene drive
00:20:08.03 is almost certainly necessary.
00:20:10.19 And it's one that is eminently worthy.
00:20:12.15 And it's a problem that everyone understands.
00:20:14.20 That is, it's possible that the nations of the African Union will agree to deploy
00:20:19.19 this kind of gene drive against malaria.
00:20:21.25 There are very few other problems for which that is the case.
00:20:26.24 Most other work on gene drive should focus on different kinds of drive systems.
00:20:31.13 That is, don't work on the standard gene drive that will affect most populations of the target species.
00:20:37.27 Instead, work on different forms that are localized.
00:20:41.27 Local drive systems will be needed to accomplish all of the other goals outlined,
00:20:48.09 suppressing populations of invasive species, for example.
00:20:51.15 That is something where you don't want to build a drive system that will spread
00:20:55.09 into the native population, where the species may be playing an important ecological role.
00:21:00.02 More generally, you want to build drive systems that will not spread across international boundaries,
00:21:04.14 because it is much easier to get regulatory approval and popular support for
00:21:09.02 something that is confined within one nation, and ideally within one particular town.
00:21:13.24 So, scientists interested in drive systems and ecological engineering should instead
00:21:19.15 focus on ways of making those technologies localized.
00:21:26.11 But as they do, we need to keep in mind that this technology is very different
00:21:31.09 from conventional biomedical research.
00:21:33.23 Because if we're developing a new drug that can go through regulatory approval,
00:21:38.16 be put on the market, people's doctors can recommend the drug to them.
00:21:43.06 And they're free to say no.
00:21:44.21 They can decline.
00:21:45.21 They can opt out.
00:21:46.21 They can choose not to be affected by our development of this drug.
00:21:50.22 But if we develop a drive system intended to alter the shared environment,
00:21:55.19 people will not be able to opt out of its effects.
00:21:59.12 Even if they get to vote on whether or not it's used in their community, if they're outvoted,
00:22:04.09 they will be affected even though they don't want to be.
00:22:08.06 And that's why ecological engineering is a form of civic governance.
00:22:13.16 Just like passing a law, it will affect everyone within the region.
00:22:17.18 And that means that if we develop these technologies in a closeted manner, behind closed doors,
00:22:24.04 without telling people what we're doing, we're denying people a voice in decisions
00:22:28.27 intended to affect them.
00:22:30.11 And as the National Academies' report says, "The best course of action is to ensure that
00:22:34.08 those who would be affected have an opportunity to have a voice in decisions about it."
00:22:39.05 And that's why current efforts aim to change the scientific incentives governing gene drive research
00:22:45.00 to encourage scientists to pre-register the experiments they intend to perform,
00:22:50.21 and to make those pre-registration forms public so that people know about all gene drive research
00:22:55.13 that is going on, and can suggest improvements and changes that could influence
00:23:02.01 how the particular application ends up affecting their community.
00:23:06.14 This should not only increase the likelihood of public support
00:23:09.23 -- because people will then have a voice,
00:23:12.15 and will be able to shape the technology and the particular application in ways that they want --
00:23:17.05 but it's also the ethical thing to do.
00:23:19.19 More generally, gene drive research is an opportunity to initiate and pioneer
00:23:24.17 a new form of local, open, and responsive science.
00:23:28.14 If we're developing technologies intended to engineer the shared environment,
00:23:32.13 we should ensure that early applications only address problems that are obvious to all citizens.
00:23:38.08 We should openly share our proposals before experiments begin, so that communities can
00:23:42.15 weigh in on the direction the technology is going, and shape the particular application
00:23:47.24 in ways that suit them and their environment.
00:23:50.18 We should actively invite any and all concerns and community guidance, including inviting
00:23:56.03 skeptical voices to point out potential problems.
00:24:00.06 Because, first of all, it's their environment, and so it's their decision.
00:24:05.16 But second, local communities know more about their own local environment than we do.
00:24:11.16 And ecosystems are complex.
00:24:14.04 And when you're engineering a complex system that you don't completely understand,
00:24:17.27 you want to make the smallest possible change that can solve a problem.
00:24:21.16 And you want to start small, and local, with a field trial, before you scale up.
00:24:28.24 And that field trial needs to be evaluated by independent scientists, because try as we might,
00:24:34.04 those of us developing a technology will always be biased in its favor.
00:24:38.07 We need to make it clear to communities when we begin that independent assessment
00:24:43.03 of that technology will be needed.
00:24:45.05 Put this together and we have a chance not just to develop these technologies in an ethical manner
00:24:51.18 that is more likely to win public support and solve real-world problems,
00:24:55.24 but it's also a step towards changing how we do science in general, making it more open and aligned
00:25:01.11 with actual community needs and community wishes.

Part 2: Gene Drive and Local Drive

00:00:07.03 I'm Kevin Esvelt.
00:00:08.03 I'm an assistant professor at the MIT Media Lab, where I lead the Sculpting Evolution group.
00:00:13.10 This is my second iBiology talk on gene drive and other technologies for engineering ecosystems.
00:00:20.08 Now, gene drive is a naturally occurring phenomenon that happens when a vertically transmitted
00:00:26.06 genetic element -- that is, something that's passed on from parents to offspring --
00:00:30.26 will reliably spread in a population, even if it doesn't help individual organisms reproduce.
00:00:38.05 Now, gene drives are ubiquitous in nature, and they operate by many, many different mechanisms.
00:00:45.19 But they all tend to require sexual reproduction, or at least some form of sharing genetic information.
00:00:52.00 The rate at which they spread depends on the generation time of the organism,
00:00:56.23 and also on the rate of gene flow between populations.
00:01:00.12 And different kinds of gene drive can either alter, and some of them can also
00:01:04.23 suppress populations to lower levels.
00:01:08.21 But when we're talking about different kinds of drive systems, it's usually useful to
00:01:13.00 think about them spatially.
00:01:14.08 That is, what happens if you release a particular number of organisms into a local population
00:01:21.19 of the target species?
00:01:23.18 How will that drive system then spread?
00:01:26.09 Will it spread into neighboring populations connected by low levels of gene flow?
00:01:31.06 And will it spread through successive populations?
00:01:33.24 So, we can think of it as releasing a number of organisms at a particular site of release,
00:01:39.20 and then observing how the drive system is predicted to spread through the rest of the population.
00:01:44.10 So, let's start with what we think of as a standard gene drive.
00:01:48.06 And this is what we call a self-propagating drive.
00:01:51.02 Now, as the image indicates, you release a number of these organisms,
00:01:55.25 and a standard gene drive is predicted to alter all of the populations.
00:02:00.16 And the reason is that, whereas a normal gene has a 50% chance of being passed on to offspring,
00:02:07.17 assuming a heterozygote parent, a standard gene drive has up to a 100% chance of being inherited.
00:02:15.19 And the reason is that it cuts the wild type equivalent gene in the reproductive cells
00:02:21.13 and copies itself over, thereby ensuring that up to 100% of gametes carry the gene drive system.
00:02:29.17 And this means that, by copying itself in the reproductive cells of every generation,
00:02:34.10 it can spread through subsequent generations of organisms, eventually affecting an entire population.
00:02:40.16 Mechanistically, we can build these using CRISPR as our genome editing tool of choice.
00:02:46.02 To make them stable, we typically want to identify an important gene, that is,
00:02:50.27 something that is important for the fitness of the organism.
00:02:53.23 And we then build a recoded version of that important gene, wherein we change the codon usage
00:02:58.24 to remove target sites for CRISPR.
00:03:03.02 We then encode guide RNAs targeting the original sites, and a CRISPR nuclease,
00:03:09.17 and whatever gene we want to spread through the population.
00:03:13.08 Then, in the reproductive cells of heterozygotes, that have one copy of each,
00:03:18.14 CRISPR components will be produced.
00:03:21.15 They will cut the important gene at many different sites, causing this double-strand break.
00:03:26.15 The cell will then repair the damage, either by homology-directed repair
00:03:30.21 -- in which case the drive system is copied, so now there are two copies and it is guaranteed inheritance --
00:03:36.00 or it will jam the ends together of the broken DNA.
00:03:41.26 And if this happens, then it will delete this important section of the gene.
00:03:46.22 And that means that the result will be more costly, evolutionarily, than the drive system itself.
00:03:53.05 In other words, using this design, natural selection will favor the drive system,
00:03:58.03 meaning that it should not be blocked by resistant alleles that cannot be cut.
00:04:03.09 And several models and, more recently, experiments have supported this... the effectiveness
00:04:08.27 of this design.
00:04:10.10 So, what needs to be done next to continue to develop standard gene drives?
00:04:14.25 Well, several steps need to be optimized in each organism for a given application.
00:04:21.10 And those include learning how to express the nuclease at a time that maximizes
00:04:26.02 the frequency of homology-directed repair.
00:04:29.02 We also need to ensure that cutting does not take place in the embryo, because homology-directed repair
00:04:35.01 is typically lower, and population suppression will not work if cutting happens
00:04:40.04 in the embryo.
00:04:41.10 Finally, we need to continue to develop alternative CRISPR editing systems, such as Cas12a,
00:04:47.04 that allow us to target multiple sites in regions important for fitness
00:04:51.08 without introducing repetitive DNA sequences throughout the drive system itself.
00:04:56.18 And that's important for stability.
00:04:59.17 So, self-propagating standard gene drive systems have a number of benefits.
00:05:04.25 They're powerful and efficient.
00:05:07.06 And that is likely to be required for the most important application,
00:05:11.06 which is the potential eradication of malaria in Africa.
00:05:14.23 But they also have a number of downsides that rule them out for the vast majority of other applications.
00:05:20.10 Specifically, they're predicted to be highly invasive.
00:05:23.02 And that means that it's likely to be very difficult to run confined field trials,
00:05:27.23 if possible at all.
00:05:29.23 The reason is that standard gene drive systems are capable of copying themselves indefinitely.
00:05:38.14 And a drive system that is copied very efficiently
00:05:43.00 -- that has a high homing efficiency, as we call it, a copying efficiency --
00:05:47.02 is predicted to spread between two populations and invade a second population
00:05:52.07 if one in a million organisms moves between the two populations.
00:05:58.23 We can't say for sure whether the same is true of suppression drives, because no models
00:06:02.20 have been done specifically of population suppression drives.
00:06:05.17 It's possible than in some species the drive system will self-extinguish in a local population,
00:06:12.04 before it can spread to a new one.
00:06:14.13 But if you compare models of different systems... if you compare an alteration drive system
00:06:19.14 that is not designed to be stable, and so will eventually be outcompeted
00:06:24.02 by resistant alleles... it never rises about 40 percent of the population...
00:06:28.00 then this very same system is predicted to be highly invasive.
00:06:33.24 And that's because there are a number of organisms, here, that are carriers,
00:06:38.06 that can move between populations.
00:06:41.01 And if you look at the number of carriers of this alteration drive that gets outcompeted over time,
00:06:44.03 we know that the models say that this is highly invasive.
00:06:48.08 And if we look at a different model of a suppression drive by Austin Burt and coworkers,
00:06:53.06 then this predicts a number of carriers of this suppression drive that are present
00:06:58.08 before the population crashes that is not much less than the highly invasive alteration drive.
00:07:04.28 So, we don't know for sure.
00:07:06.20 Additional modeling needs to be done.
00:07:08.10 But our default expectation is that all self-propagating standard gene drives are likely to move
00:07:14.12 very efficiently between populations.
00:07:16.16 And that's going to be a challenge.
00:07:19.22 We also need to be aware that we can't simply assume that we will be able to develop
00:07:24.15 these drive systems in our laboratories, and even attempt field trials, without outside interference.
00:07:30.26 And just a look at history shows that this is likely to be the case.
00:07:35.16 Australia was interested in using rabbit hemorrhagic fever to control populations of invasive rabbits
00:07:45.11 in Australia.
00:07:46.28 And it escaped from the test island, and was carried, unexpectedly by blowflies,
00:07:51.00 to the mainland, where it began to spread across the continent.
00:07:54.21 This caused New Zealand to decide that they did not want it to be introduced,
00:07:59.07 and in fact made it illegal.
00:08:01.10 But New Zealand farmers decided to smuggle it in anyway, through the tightest biocontrol
00:08:05.22 in the world.
00:08:07.04 And they did this even though the total economic damage caused by rabbits in New Zealand
00:08:11.22 is only somewhere between 7 and 30 million dollars a year.
00:08:15.20 In contrast, rats, which are a major proposed conservation target for suppression gene drive systems,
00:08:22.24 cause over 20 billion dollars a year -- billion, not million -- in the United States alone,
00:08:29.05 in terms of economic damages.
00:08:31.13 So, to imagine that people would not move a suppression gene drive for economic reasons
00:08:39.22 should not even be considered.
00:08:40.26 It will happen.
00:08:42.02 That is how the world works.
00:08:43.07 And we can't just assume that things will happen the way our models predict.
00:08:48.02 Put all this together, and that means that we should probably
00:08:52.19 reserve standard self-propagating gene drive for problems
00:08:56.07 that require us to affect an entire species, and for which an international agreement
00:09:02.06 among the affected countries is likely to be feasible.
00:09:05.15 And the number one such application is indeed the eradication of malaria.
00:09:09.11 Malaria is such an awful disease.
00:09:11.02 It inflicts such a terrible toll.
00:09:13.04 More than half a million people annually, 200 million infections.
00:09:17.13 Most of those who die are children under the age of five.
00:09:21.12 This is a problem that is so severe that it's possible to imagine an international agreement
00:09:26.04 among countries of the African Union to use it, even without field trials.
00:09:31.24 And if they do, then the same technology could be used, potentially, to eradicate other scourges
00:09:37.08 such as schistosomiasis.
00:09:39.24 But we have to keep in mind that a single tragic death in an ill-planned clinical trial
00:09:47.04 set the field of gene therapy back by a full decade.
00:09:51.05 And any accident involving gene drive that prevented the countries of the African Union
00:09:55.27 from deciding to use it to prevent malaria... if that setback the effort by just a decade,
00:10:02.06 the expected cost would be three million lives, at least.
00:10:06.26 So, everyone working with gene drive needs to be aware that lives are literally
00:10:13.26 hanging in the balance.
00:10:15.11 So, when we are developing our safety protocols and taking precautions
00:10:19.22 -- and even deciding, should we be running these experiments in the first place? --
00:10:23.22 that is what is at stake.
00:10:27.27 Instead, most laboratories may want to pursue alternative drive systems that are not self-propagating.
00:10:35.17 The opposite of a self-propagating drive system is one that is self-exhausting.
00:10:40.15 That is, it's one that relies on some form of genetic fuel that is used up over generations
00:10:46.04 as the drive spreads, until eventually it runs out and stops.
00:10:49.23 This means that you get a spreading pattern in which the release of a few organisms
00:10:54.10 leads to some amplification before it eventually stops.
00:10:57.12 If you release more organisms, it spreads further through the population.
00:11:01.28 But it does not spread indefinitely through all connected populations.
00:11:07.14 How do you build such a self-exhausting drive system?
00:11:09.22 Well, my group is developing what we call "daisy drives", which separates the components
00:11:13.28 of the drive system across multiple chromosomes in ways that are linked.
00:11:20.04 So, if you assume that a given daisy drive has three elements, C and B and A,
00:11:25.21 then C has the instructions that cause B to be copied.
00:11:31.15 So, C tells CRISPR to cut the wild type equivalent of B and copy it over.
00:11:37.05 B has the instructions causing A to be copied.
00:11:40.26 But there is nothing causing C to be copied.
00:11:43.06 C is a normal engineered gene.
00:11:46.02 And that means that when we look at the family tree of something like a daisy drive,
00:11:49.20 if we assume that the drive carrier, at the top, is a heterozygote
00:11:53.19 -- it has one copy each of C and B and A --
00:11:57.10 then when that organism mates with a wild type, then, all of the offspring
00:12:01.05 will inherit B and A.
00:12:03.08 But this one on the end here does not inherit C.
00:12:05.17 And in the next generation, that means you get an offspring that does not inherit B.
00:12:12.09 So, this one only carries A.
00:12:15.06 Which means one more generation of mating to wild type, and you finally get
00:12:18.06 a descendant of that original CBA organism that is entirely wild type.
00:12:22.24 So, here's an example of how the drive system literally runs out of genetic fuel.
00:12:26.22 It runs out of daisy elements and stops.
00:12:30.17 Of course, this implies that the more daisy elements you have in the chain, then,
00:12:34.09 the more powerful the drive system, the greater the frequency it will reach in the population
00:12:38.00 for a given release of organisms.
00:12:41.17 How many is that?
00:12:42.17 Well, it depends on how efficient the copying is.
00:12:45.18 So, Daisy drives work by the same CRISPR-based copying mechanism.
00:12:48.26 So, if the copying rate is 90%, and you want to alter most organisms in a population
00:12:57.05 within 20 generations, then you would need to release one daisy drive organism per 50 wild organisms
00:13:02.22 in that population.
00:13:04.02 And 20 generations later, they would all be affected.
00:13:08.01 But if your copying efficiency is 98% percent, as has been seen in some malarial mosquitoes,
00:13:12.11 then you only need to release one daisy drive organism per 500 wild ones.
00:13:17.17 So, this is very efficient.
00:13:21.18 The problem for daisy drive, and indeed for all of these, is that of gene flow.
00:13:27.24 So, let's assume that we have a number of interconnected populations.
00:13:32.13 And we're going to introduce a drive system into one, and it's gonna spread through other populations.
00:13:38.24 And each of these is connected by a comparatively low level of gene flow.
00:13:41.20 Let's look at what happens in the frequency of the different engineered DNA elements over time.
00:13:47.26 For the daisy drive, if you introduce it into the first population at a high enough frequency,
00:13:52.09 it will rise to very high frequency, and then it will eventually... after a couple hundred generations,
00:13:57.27 it will go down to zero.
00:14:00.20 That's in population 1.
00:14:02.01 But in connected population 2, it also reaches a reasonably high frequency before going back down.
00:14:08.02 And there's a little bit that makes it into 3.
00:14:10.12 And basically none into 4 and 5.
00:14:13.15 We can contrast that with a standard gene drive that is self-propagating.
00:14:17.03 And here, once released in population 1, it spreads to virtually all organisms.
00:14:22.17 And of course, it spreads into all of the other populations, 1 through 5.
00:14:28.01 The other useful comparator on the other end is what happens if we simply release
00:14:32.28 a lot of engineered organisms with no drive system at all, that are just engineered to have
00:14:37.18 the relevant trait.
00:14:39.10 If we release enough of them -- what's called an inundative release --
00:14:43.07 such that the frequency of organisms that have the engineered trait is nearly all of them, nearly 1,
00:14:51.14 then we still see a little bit of change in population 2, but virtually none and 3 and 4 and 5.
00:14:59.16 The problem is for many applications, given the sensitivity of genome editing
00:15:06.03 and public views of it, even this little bit of spread of a normal engineered element
00:15:11.11 may not be acceptable.
00:15:13.25 That is, we can't necessarily afford to have any significant gene flow across political boundaries.
00:15:20.23 Ideally, we need to be able to engineer a population within the boundaries of a particular town,
00:15:25.19 and somehow eliminate engineered DNA everywhere else.
00:15:32.28 How can we do that?
00:15:34.27 Well, one option is to turn to a third kind of drive system that is also local,
00:15:40.03 which we a threshold drive.
00:15:41.17 And it's called a threshold drive because it is frequency-dependent.
00:15:44.22 That is, whether or not the drive system spreads or is actively eliminated depends on its frequency
00:15:51.17 in the population.
00:15:52.24 If you introduce few of them, then natural selection will eliminate them all.
00:15:57.20 If you introduce many, then it will spread to become very common, nearly fixed
00:16:03.21 within the one population.
00:16:04.27 But it will not be able to spread effectively into the next, because in that population
00:16:09.20 it will be in the minority.
00:16:12.15 How do you do that?
00:16:14.17 Well, in 1968, Curtis proposed that chromosomal translocations could be used to generate
00:16:20.08 this kind of effect.
00:16:21.22 If you take two wild type chromosomes and you transpose two arms
00:16:27.07 -- that is, you just break off one arm from each chromosome and you swap them --
00:16:33.15 then this organism has two copies of all genes.
00:16:36.27 It's fine.
00:16:38.01 But when it mates with a wild type organism, or indeed any organism, then the offspring
00:16:42.26 will sometimes inherit a wild type chromosome and sometimes a chromosome that has a translocation.
00:16:49.28 If they inherit two wild type, they're fine.
00:16:52.22 If they inherit two with a translocation, then they're balanced.
00:16:56.06 They still have all the relevant genes.
00:16:58.11 In both cases, they're just like one or the other parent.
00:17:00.28 But the other two possibilities involve inheriting one translocated chromosome
00:17:05.26 and one wild type chromosome.
00:17:08.10 And in this case, they're missing critical DNA.
00:17:11.14 And that means they're eliminated.
00:17:12.22 So, this phenomenon is called genetic underdominance.
00:17:17.07 And what it means is that when hybrids... when heterozygote organisms are unfit
00:17:23.26 relative to either homozygous version, then whichever allele
00:17:28.02 -- either, in this case, engineered or wild type --
00:17:31.17 is in the majority will win, assuming they're otherwise equivalent in fitness.
00:17:37.14 And this has actually been experimentally demonstrated by Bruce Hay's group at Caltech
00:17:41.24 in fruit flies.
00:17:42.24 They used CRISPR to generate chromosomal translocations,
00:17:47.07 and introduced flies that had these translocations into cages containing wild type flies.
00:17:52.15 And they did the introduction at different frequencies.
00:17:55.08 And they found that if introduced at above 60% frequency into...
00:17:59.27 into this population of wild flies, then the translocated, engineered version would take over.
00:18:06.20 It would go to 100%.
00:18:07.24 But if it was introduced at any frequency less than 60%, then it would go down to zero.
00:18:13.16 So, here's an actual working example of a threshold-based drive system.
00:18:18.28 Other examples have also been demonstrated using toxin-antitoxin systems, but those are
00:18:22.16 a little bit more complicated and have not been shown to be stable over time.
00:18:26.22 Whereas we know from naturally-occurring translocations that this version is very stable.
00:18:31.02 So, these threshold drives are extremely compelling, precisely because they're so localized.
00:18:37.26 They... if introduced in an area, they will stay within that area, because they will be
00:18:41.22 actively eliminated elsewhere.
00:18:44.04 The problem is that they can't be used to directly suppress populations.
00:18:47.19 They can only alter them.
00:18:49.20 Now, it's true that they can spread alterations that could then be targeted
00:18:53.04 by another kind of drive, say, using CRISPR.
00:18:56.17 But that would still be extra steps.
00:18:59.19 Perhaps the bigger problem is that releasing sufficient organisms to reach a 60% frequency
00:19:06.15 requires releasing more engineered organisms than normally exist in the wild in that area.
00:19:12.09 And for species that are actively causing a problem,
00:19:15.09 whether because they're environmentally destructive or they're pests,
00:19:19.00 that may just not be feasible.
00:19:20.26 So, one potential solution is to combine a self-exhausting drive with a threshold drive.
00:19:28.01 So, my group is working on building what we call daisy threshold systems.
00:19:32.28 Now, a daisy system cannot spread a chromosomal translocation.
00:19:37.24 So instead, we're looking at creating a different form of threshold drive, in which we deliberately
00:19:42.25 swap two haploinsufficient genes.
00:19:46.12 Now, haploinsufficient just means that you need two copies in order to live.
00:19:50.22 If you have one copy, that's just not enough.
00:19:53.08 So, if you take two haploinsufficient genes and you swap their positions
00:19:57.15 -- just the genes, not the whole chromosome arms -- then you get the same effect.
00:20:02.09 When mated with a wild type organism, offspring can either inherit two swapped copies
00:20:06.25 -- and they'll be fine, just like their heterozygote parent --
00:20:10.11 or two wild copies, in which case they'll be wild type, just like their wild parent.
00:20:14.20 But the other two possibilities are inheriting one swapped and one wild type.
00:20:20.04 And these organisms are missing a critical haploinsufficient gene, and so they will perish.
00:20:25.12 So, the net effect is to create a threshold system that can be spread using a daisy.
00:20:31.08 So, here, the idea is you introduce, into a given population of a town that wants to
00:20:38.09 alter its local organisms, daisy threshold organisms in the center of town.
00:20:43.27 The daisy drive spreads the alterations through the town towards the boundary,
00:20:49.22 losing genetic fuel as it goes.
00:20:52.08 And only when it runs out does the threshold switch on.
00:20:55.26 Because while a daisy element is present, all of the offspring are guaranteed to
00:20:59.22 inherit two swapped copies.
00:21:01.17 As soon as you run out of daisy elements, as soon as daisy elements are gone,
00:21:05.27 then you get the threshold effect.
00:21:08.09 So, when the daisy elements run out, say, towards the edge of the town boundary,
00:21:14.02 then within the town the threshold will have been exceeded,
00:21:17.23 and it will actively select for the engineered variant.
00:21:21.05 But in the neighboring town, the engineered allele will be in the minority,
00:21:26.02 and natural selection will actively eliminate it.
00:21:28.06 So, this would allow very few organisms -- comparatively few -- to be released in a town,
00:21:33.03 while still keeping the alterations within that town, and not allowing it to spread across the boundary.
00:21:38.13 So overall, we have two broadly different forms of drive systems.
00:21:44.05 We have standard gene drive systems that are self-propagating and are really only suited
00:21:48.24 for applications where the goal is to affect most or all of a species.
00:21:54.27 On the other hand, we have these local drives that have many, many more potential beneficial applications
00:21:59.22 -- from health to the environment, animal well-being, and eventually, even agriculture --
00:22:04.28 that would allow us to make very small changes that could solve severe ecological problems
00:22:13.02 in the most minimalist way possible.
00:22:15.18 Because whenever we're thinking about engineering a complex system that we don't completely understand,
00:22:19.16 which applies to any biological system, but especially to eco... to wild ecosystems,
00:22:27.02 we should always strive to make the smallest possible change capable of solving the problem.
00:22:31.15 And we should start small and see what happens before scaling up.
00:22:35.28 And that latter test, the ability to run a field trial, requires that we have some form
00:22:40.25 of localization that prevents the alteration from spreading indefinitely beyond the trial site
00:22:46.11 and into the world beyond.
00:22:48.22 However, all forms of drive system face extra safety and regulatory barriers that
00:22:54.10 just don't apply to normal engineered DNA.
00:22:56.24 And the reason is that we... humanity... we have never before engineered something anticipated
00:23:02.20 to evolve outside of our control.
00:23:05.20 Normal engineered organisms only reproduce when deliberately crossed,
00:23:10.08 or at least within environments in which we do control the relevant environmental factors.
00:23:17.16 With a gene drive system, or any drive system, that is just no longer the case.
00:23:22.22 Will a drive system spreading through a wild population that we cannot directly impact...
00:23:27.19 will it evolve to cut the equivalent sequence in a closely related species, say,
00:23:36.02 that can form fertile hybrids with the target organism?
00:23:40.04 That question is gonna depend on... what is the nature of the CRISPR-based drive?
00:23:43.25 Which guide RNAs are used?
00:23:45.13 How different are the two sequences?
00:23:47.18 That's an empirical question.
00:23:49.09 And it's one that is going to need an answer.
00:23:52.16 Will a drive system spread through a local population as anticipated?
00:23:56.07 Is it going to spread further?
00:23:57.09 Is it not going to spread as much?
00:23:59.16 Will it be stable?
00:24:00.16 Will it indeed stay local?
00:24:02.19 How can we know for sure?
00:24:04.20 The problem is that these are evolutionary questions.
00:24:08.15 And evolution is a numbers game.
00:24:11.12 There will always be more organisms present in the wild then we can possibly rear in the lab.
00:24:17.16 And that is a serious problem.
00:24:20.12 My group is attempting to develop a system capable of answering these questions
00:24:24.12 using nematode worms, which reproduce every three days, and can be grown in populations of billions.
00:24:30.16 For example, if we're interested in determining whether a daisy drive system
00:24:35.13 is in fact stable and localized, then we can introduce a handful of daisy drive worms into a large population
00:24:44.00 and observe as the drive spreads over time using a fluorescence marker.
00:24:50.17 But we wouldn't expect that marker to show up in all of the worms, because this is supposed
00:24:53.27 to be a daisy drive.
00:24:54.27 It's supposed to run out of genetic fuel and stop.
00:24:56.27 If it does spread to all of the worms, we'll know that something went wrong.
00:25:00.12 And that drive system is not safe enough to use in a wild population.
00:25:03.18 Similarly, if we're worried about any kind of CRISPR-based drive evolving to spread into
00:25:08.17 a closely related species that might hybridize, we could build worms that have
00:25:14.18 the target sequence in the desired organism, we could build separate worms that have
00:25:20.19 the target sequence of a related species that we don't want it to spread into, that should have
00:25:24.21 several mutations that are different, and then we can introduce worms carrying the drive system
00:25:28.24 into that mixed population.
00:25:29.24 Ideally, it would spread only through the worms carrying the target sequence.
00:25:36.01 But if it ever spread into the related sequence, we would see it, because the fluorescent marker
00:25:41.20 corresponding to the related sequence would disappear.
00:25:44.22 And if that ever happens, even once, then it would spread through all of the related species,
00:25:48.28 and we should definitely see it.
00:25:51.02 So, this is a way of quantifying the relative risk that any kind of drive system
00:25:57.16 becomes unstable and starts to spread in a way that we don't want.
00:26:02.13 Overall, though, it's very important to emphasize that the social barriers in the place of applications
00:26:11.11 of ecological engineering are much greater than the technical challenges.
00:26:16.15 That is, it's possible that technical advances could help address the social barriers
00:26:22.24 -- for example, better localization strategy is exactly that -- but we should never operate
00:26:29.05 under the illusion that the technical barriers are the greatest challenge.
00:26:34.11 They're not.
00:26:35.15 It's more important that we develop this in ways that are cognizant of those social barriers
00:26:41.22 than focusing only on the technical challenges of making something work in a given organism.
00:26:48.03 So, my group is pursuing several different ways of addressing these problems.
00:26:52.08 Number one is to communicate the concept in an understandable manner to very diverse populations.
00:27:00.26 For example, with daisy drive, we're working to build mice in which every daisy element
00:27:06.26 corresponds to a different coat color marker.
00:27:09.28 We're then going to breed a family of mice in which you can see the grandparents,
00:27:15.05 the parents, the kids.
00:27:17.03 And then eventually, the kids grow up and reproduce on their own.
00:27:20.02 And you'll be able to see the genetics in the coat color of each of these mice.
00:27:24.15 And of course, every one will have their name and their own backstory in order to make
00:27:28.11 a compelling story about this particular family of mice.
00:27:34.08 Another problem is to ensure that applications are popularly supported and directed by communities.
00:27:41.12 The Mice Against Ticks project doesn't involve any form of drive at all, by community requests.
00:27:48.03 And the communities in this case are in charge.
00:27:51.05 The people of Nantucket and Martha's Vineyard have been directing this project from inception.
00:27:56.08 They chose among several options, and decided that they did not wish to use any kind of drive system,
00:28:01.23 specifically because they preferred that the project not use any DNA foreign to white-footed mice.
00:28:08.18 And this is a project that aims to address a very important local problem,
00:28:12.04 the spread of tick-borne disease, including Lyme disease.
00:28:17.01 And it intends to address it in a novel way: by immunizing, heritably, the mice responsible
00:28:23.04 for infecting most ticks.
00:28:26.01 If successful on the islands where inundative release can be used, then it's possible
00:28:32.13 that the same approach might be extended to the mainland using something like a daisy threshold system
00:28:37.13 But by starting small, in an island context, without any form of drive system at all,
00:28:43.13 we can ensure that everything works well, initially on mostly uninhabited islands
00:28:50.11 that are small enough that we could actually remove all of the mice and reintroduce them.
00:28:54.14 And only scale up as appropriate once we're sure that things are working well at the previous stage.
00:29:00.08 Lastly, those decisions will not be made by us, but by the communities who are interested,
00:29:06.03 thereby ensuring that the project does actually address any and all community concerns,
00:29:11.05 and does so at the earliest possible stage.
00:29:15.17 Inviting community concerns is absolutely critical when it comes to developing ecotechnologies
00:29:21.08 intended to affect the shared environment.
00:29:24.19 Because decisions that we make as scientists, early on, are very likely to affect the outcome
00:29:31.08 of any given application, including whether or not it goes forwards.
00:29:35.24 So, we need to be aware that if we do or say something that reduces the likelihood
00:29:43.03 that the African Union can agree to use gene drive to help eradicate malaria,
00:29:48.19 a 1% chance of a decade-long delay has an expected cost of 25,000 children's lives.
00:29:59.21 What are our responsibilities to those people?
00:30:03.00 What are our responsibilities to wild organisms that might be aided?
00:30:08.00 Whether... endangered species that might be saved from invasive species that spread
00:30:13.11 using human transport.
00:30:15.12 What of organisms that are suffering, that might be aided by our intervention?
00:30:21.22 Are we responsible for them?
00:30:24.05 What are our responsibilities to the natural world?
00:30:28.00 We as scientists must be cognizant of the fact that the very notion of
00:30:32.12 engineering wild species is deeply disturbing to many of our fellow citizens.
00:30:37.18 And even if we don't share those values, we need to respect that.
00:30:41.27 Because it's their world too.
00:30:44.02 In order to respect the beliefs of people who oppose engineering the shared environment,
00:30:50.12 and also to reduce the potential social backlash should a gene drive ever spread
00:30:56.09 without community support, my group is developing ways of restoring populations
00:31:03.00 to their original wild genetics.
00:31:04.27 That is, we hope to remove all engineered DNA elements from a population,
00:31:10.28 no matter how they got there.
00:31:12.19 And that includes standard self-propagating gene drives.
00:31:16.27 If such a drive system were to be introduced without permission,
00:31:20.10 then we might build a daisy restoration drive to counter it and overwrite it.
00:31:25.08 So, a daisy restoration drive would be introduced into the population that was affected
00:31:32.03 by the unwanted drive system.
00:31:34.13 It would then spread, and overwrite it, and replace it with itself.
00:31:39.14 It would also spread through the wild organisms near the boundary, thereby ensuring that
00:31:44.19 every last copy is overwritten.
00:31:46.25 But all of these daisy restoration elements are linked to a threshold effect,
00:31:54.00 such that introducing additional wild organisms would then ensure that the engineered DNA is
00:32:00.02 below the threshold frequency, and will therefore be eliminated by natural selection.
00:32:04.28 In fact, every last copy.
00:32:07.17 It's our hope that by developing these technologies in the open, focusing on local communities,
00:32:13.27 inviting input, concerns, and criticism, including from skeptics and people whose values
00:32:20.04 oppose the very idea of engineering the shared environment, we hope to come up with a...
00:32:27.08 not a consensus, but still a path forward that will allow us to fulfill our moral responsibilities,
00:32:34.11 not only to one another but also to the organisms with which we share our world.
00:32:39.20 Thank you.

Videos in this Talk

Part 1: Gene Drive
Audience:
- Student
- Researcher
- Educators of H. School / Intro Undergrad
- Educators of Adv. Undergrad / Grad
Part 2: Gene Drive and Local Drive
Audience:
- Student
- Researcher
- Educators of H. School / Intro Undergrad
- Educators of Adv. Undergrad / Grad

Speaker: Kevin Esvelt

Total Duration: 0:58:22

Recorded: July 2018

All Talks in Genetics and Gene Regulation

Talk Overview

Evolution has selected wild organisms to be extremely well adapted to their environment. Because most genetic changes introduced by humans divert the resources of the organism to benefit humans, such mutations are typically eliminated by natural selection in the ancestral habitat. In his first talk, Dr. Kevin Esvelt explains how self-propagating CRISPR-based gene drives can be used to spread genetic alterations through wild populations, potentially impacting all organisms of the target species. Gene drives could be used to benefit public health, the environment, agriculture, and animal well-being. However, real-world use may incur ecological risks, and even research involving self-propagating gene drive systems may risk public trust in science and governance given the possibility of accidental spread. Esvelt explains how to minimize risk and discusses the importance of engaging communities in planning any projects which may affect them.

Esvelt’s second talk focuses on strategies to allow for the safe implementation of localized gene drive technologies that do not spread indefinitely. Daisy drive systems are made up of multiple elements connected like a daisy chain such that each causes the next to be preferentially inherited. They are designed to be self-exhausting by losing elements with each generation, thereby limiting spread. This technique has multiple applications such as removing an invasive species from one area without impacting the same species in its native habitat. Esvelt explains that daisy-drive stability might be tested in a species such as C. elegans where hundreds of generations can be grown in a short period of time. His lab is also developing technologies to reverse any unwanted genetic changes that might be introduced via gene drive. Once again, Esvelt emphasizes the importance of community input into any gene alteration projects. Although it does not currently involve gene drive, he uses the “Mice Against Ticks” project that seeks to prevent tick-borne diseases on the islands of Nantucket and Martha’s Vineyard as an example.

Speaker Bio

Kevin Esvelt

Kevin Esvelt received his B.S. in Chemistry and Biology from Harvey Mudd College and his Ph.D. from Harvard University. As a Technology Development Fellow of the Wyss Institute, he worked with Dr. George Church at Harvard Medical School. He helped to develop CRISPR as a genome editing tool and was the first to identify the… Continue Reading

Part 1: Gene Drive

Part 2: Gene Drive and Local Drive

Talk Overview

Speaker Bio

Kevin Esvelt

Leave a Reply Cancel reply

Find a Video

Topics

For Educators

Educators Recommend

Free Courses

Career Development

Best of iBiology

Top Series

About Us

Talk Overview

Speaker Bio

Kevin Esvelt

Playlist: Environmental Sciences

Related Resources

Reader Interactions

Leave a Reply Cancel reply

Footer

Find a Video

Topics

For Educators

Educators Recommend

Free Courses

Career Development

Best of iBiology

Top Series

About Us