For Gene Drives, Resistance May Be Inevitable

In the last few years, biologists have realized something remarkable: We might finally be able to beat evolution at its own game.

Using the gene editing tool CRISPR, which snips DNA and inserts a new gene of the designer’s choosing, biologists can arrange it so that a gene will spread exponentially through a population of organisms. While that occasionally happens in nature, with certain caveats, evolution has generally been against us: If a few mosquitoes are released with a gene that prevents them from carrying malaria, the gene will be swamped in the gene pool. The mosquitoes will only pass it on to half their offspring, who will in turn pass it on to only half of theirs, and before long it will drop out of the population. In the meantime, malaria kills more than 1 million people per year, mostly children five years old or younger. And other mosquitoes spread Zika—linked to severe birth defects—and yellow fever and dengue fever, whose numbers have escalated alarmingly in recent decades.

crispr-cas9
An artist's impression of CRISPR in action.

So when, in 2014, researchers began to point out that there was a way to help such genes take over, it caught people’s attention. If a mosquito is given a bit of DNA that not only codes for that gene but for the CRISPR machinery itself, then the CRISPR proteins will go around inserting themselves and their cargo, ensuring that all the insect’s offspring will have them, as will all of their offspring. As the gene (and the CRISPR machinery) are passed down, it will spread exponentially. Soon, it might take over the population or perhaps even the whole species—and, in this example, no mosquitoes would be capable of carrying malaria. The same technology could just as well spread a gene that could cause a mosquito population crash, however, which has yielded many calls for caution and prudence in studying gene drive systems, as these constructs are known.

But there is a chink in gene drives’ armor. Specifically, it is possible for organisms to escape though an age-old trick, one that we’re growing more familiar with by the day—by evolving resistance, just as bacteria evolve resistance to antibiotics. In a recent paper posted to the biorXiv, evolutionary geneticist Rob Unckless of the University of Kansas and colleagues lay out mathematical models of what may happen when gene drives are released, and find that even under the best of conditions, resistance is certain to arise. Exactly what that insight means—whether it is a fatal flaw in the technology, merely an obstacle to be overcome, or perhaps even a tool to be used—is still an unanswered question.

Botched Repairs

It’s been known from the earliest work on CRISPR gene drives that resistance was a strong possibility. That’s because CRISPR will cut DNA only at places with a specific sequence of DNA bases. The recipe for the CRISPR machinery and its cargo gene then have to be pasted into that cut. The cell will usually cooperate because this bit of genetic material is designed to blend in. But with each cut, there is a chance that the cell will instead try to repair the gap by putting in some new DNA base pairs. “It tries to do its best,” Unckless says, “but it makes a lot of errors.”

Once the gap has been repaired, it might well have a different sequence of DNA bases than the CRISPR system can cut. And after that, there’s nothing the gene drive system can do to it. It’s immune. “We think that is going to be the biggest source of resistance,” Unckless says, echoing a general consensus among a number of scientists. “CRISPR breaks the sequence, but instead of having the CRISPR machinery inserted, the repair is done through this [other] process.”

gene-drive-schematic
Here, a mosquito with a gene drive (blue) mates with a mosquito without one (grey). In the offspring, one chromosome will have the drive. The endonuclease then slices into the drive-free DNA. When the strand gets repaired, the cell's machinery uses the drive chromosome as a template, unwittingly copying the drive into the break.

If the repaired area prevents mosquitoes from being modified, depending on the modification, it might suddenly become a very handy thing to have. If all the mosquitoes that receive a gene drive die or become infertile or simply produce less offspring than those that are resistant, the resistant type will eventually take over. By putting our thumb on the scale of evolution on behalf of one gene, we create a situation where an opposing version can take over, if the benefits of having it are great enough.

There are a lot of “ifs” in that paragraph, but what Unckless and his collaborators found in their models—which assign numbers to those unknowns—was that even at very low rates of erroneous self-repair, resistance is a clear specter. If 100 copies of the gene drive are released, and the rate of self-repair is as low as one in 100,000, resistant organisms will still make up half the population after 100 generations have passed. A larger population will have more problems with resistance than a small one—more cuts mean more chances for resistance to arise. And in fact, Unckless and his collaborators made a very conservative assumption for the rate for this kind of repair, to see whether resistance would happen even then. Experiments suggest the rate is much higher in real organisms, perhaps one in 100. “I think [resistance] will block the drive much sooner than their model predicts,” says Kevin Esvelt, an evolutionary engineer at MIT.

By itself, this isn’t terribly surprising, says Omar Akbari, an entomologist and molecular biologist at UC Riverside. “It just highlights the fact that it’s going to be an issue,” he says.

But the question of what effects this kind of resistance will have is something the community has been thinking hard about, says Valentino Gantz, a post-doc at UC San Diego who has made gene drives in fruit flies and mosquitoes. “This is exactly one of my worries from when we started going in to mosquitoes,” he says. “I was hoping that population geneticists would do a nice analysis such as [this] one.”

How big of a problem this will be depends on what exactly a gene drive’s goal is. “If you’re targeting dengue,” says Akbari, “if you’re able to block all disease transmission for a few weeks”—in order to create a situation in which after running its course in infected individuals, the virus cannot spread any further—“that could essentially eliminate the virus. So you don’t need it to last indefinitely.” On the other hand, if the goal is to create a lasting alteration—to permanently keep mosquitoes from transmitting Zika, for instance—it would be a non-starter if in a hundred mosquito generations, around four to eight years, half of all mosquitoes would again be able to spread disease.

Which is why researchers have begun to design ways to thwart resistance. In one strategy, the CRISPR machinery cuts at a gene that is required for life, so that if the self-repair mechanism screws up, it’s lethal to the organism. In another, the CRISPR machinery can cut at five different sites, so that if one gets repaired so as to be immune, the others are still available. Another option would be to release a drive knowing resistance will arise, and, for however long is appropriate, releasing others that target the resistant genes, engaging in a kind of arms race with evolution.

It’s possible, too, that resistance, in certain situations, could be useful. “One of the realizations we’re coming to…is that we think resistance, probably at least in a simple system, is almost inevitable, so maybe we should embrace that,” Unckless suggests. If a drive is gradually out-competed by a resistant version, well, maybe there are situations where that would be preferable. Perhaps resistance could be part of a containment plan, for example in situations where a population is already isolated from others, and the inevitable resistance could catch up with the drive before any pioneering individuals could leave and spread their genetic payload. This is still conjectural, however. And there are likely to be much more predictable tools to accomplish the same task.

Out of the Lab

Still, all these models and ideas and proposals being tossed around eventually have to acknowledge the fact that there have not been very many tests in real organisms or in realistic populations. No one thinks we have enough information yet about how gene drives work in practice to be able to reliably predict their effects.

There are reasons aplenty why they might not function as planned. Anna Lindholm, an ecologist at University of Zurich, studies a naturally occurring gene drive system in mice, which is carried by males and spreads itself to more than 90% of the sire’s offspring. While researchers have never seen genetic resistance to the system evolve, she says, it hasn’t spread as quickly as you might think. That seems to be because of a simple fact of mouse mating behavior: These females mate with many males in quick succession, and the males with the drive often see their sperm out-competed by those of rivals. The drive is out of the game before it ever gets a chance to do its thing.

A gene drive released in a mosquito population could nose-dive because of some unforeseen feature of the natural world—or even of the particular area where it is released. “When you’re doing it with one species of one mosquito in one location, it might work perfectly, exactly as on paper,” Gantz says. “And when you do the same exact experiment in a different location, it might just not work for some other reason.

“I think we’ve gotten to the point where I believe the technology should move toward generating mosquitoes that are field-ready and extensively testing them in the lab and making sure there are no unexpected behaviors of the system in populations in the lab before you go into the field,” he says.

Gantz isn’t worried that a drive will spread too well, taking over populations so fast as to be uncontrollable. Rather, he is worried that a much-anticipated test will fall flat. Others in the community are more preoccupied with the first scenario. Esvelt and his colleagues have proposed a control mechanism that would split up the machinery required for the gene drive system and disperse it throughout the genome, arranging it so that without intervention, it will eventually go defunct. “We like that because it’s less of a blunt instrument,” says Charleston Noble, a graduate student at Harvard University who is working on the idea alongside Esvelt. But until more experiments happen, there is not enough information to tell which scenario is most likely.

As such tests unfold and the data are released, scientists will be looking for signs of problematic resistance, and whether, with these new tools, we really can reliably steer evolution toward our own ends. Gene drives are potentially powerful, but also potentially very delicate—a technology full of contradictions.