by Brian Tomasik
First written: 11 Feb. 2016; last update: 23 Feb. 2016

## Summary

Gene drives are a rapidly developing technology that will enable humans to alter genes in wild-animal populations. Some potential uses of gene drives to advance human interests seem plausibly good, and others seem plausibly bad. Pushing for the elimination of mosquitoes with gene drives would probably reduce insect suffering and might be cost-effective compared with other ways to help insects, though further research on this point is needed.

David Pearce has proposed researching the potential of using gene drives to improve the wellbeing of wildlife populations in general. While it's almost certain that humans would never be willing to do this, there may be value in further research into the possibilities in this arena, in case some interventions could help lots of wild animals without facing immense opposition from the rest of society.

## Introduction

Gene drives open up the possibility of spreading gene edits throughout a population of sexually reproducing organisms. If used, gene drives would thus have implications for animal suffering in nature. David Pearce eloquently explores the possibilities of using gene drives to reduce wild-animal suffering in "Compassionate Biology: How CRISPR-based 'gene drives' could cheaply, rapidly and sustainably reduce suffering throughout the living world".

In this piece, I offer a few thoughts on

1. whether gene drives in general are likely to benefit or harm wild animals on balance, and
2. whether we should promote the idea of using gene drives specifically to reduce wild-animal suffering.

## Realistic uses of gene drives

Historically, most technologies have been used for human benefit, and the same is likely to be true of gene drives. But anthropocentric interventions in wildlife populations would still have implications for wild-animal suffering. Following are some proposed uses of gene drives and how they would affect wild animals.

 Proposed use Net impact on wild animals Reason Make female mosquitoes "incapable of carrying malaria, dengue or other diseases" plausibly good? It's plausible that humans on the whole prevent more wild-animal suffering than they cause, in which case saving human lives is net good for wild animals. However, this topic deserves further study. Kill all mosquitoes plausibly good? Same reason as above, plus the benefit of reducing mosquito populations. Of course, it's possible other bugs would somewhat take the place of mosquitoes, but perhaps this wouldn't happen completely, at least not in the short run? "weeds could even be engineered to introduce vulnerabilities to completely benign substances, eliminating the need for toxic pesticides" (source) unclear If this change would allow for more narrowly targeting a specific set of plants to be killed, that might allow other plants to grow more, which is in general bad. Or maybe it would kill plants more effectively until resistance evolved, which would be good? Another consideration is that eliminating herbicides would eliminate herbicide run-off into water bodies, which I suspect tends to reduce populations of marine organisms, though I haven't done a full analysis of the pros and cons of that. "Agricultural weeds like horseweed that have evolved resistance to glyphosate, a herbicide that is broken down quickly in the soil, could have their susceptibility to the compound reintroduced, enabling more farmers to adopt no-till practices, which help conserve topsoil." (source) plausibly bad? Topsoil loss is a major cause of reduced plant productivity, so tilling may be net good, although it hurts lots of animals in the short run. If gene drives can better eliminate weeds, that in itself is good because it reduces plant biomass on crop fields, but this may also mean that fewer total herbicides will be used, which is probably bad. Gene drives might "promote more sustainable agriculture by controlling insect pests" (source) possibly bad? If gene drives required fewer broad-spectrum insecticides, this might increase populations of non-target bugs, which could be bad. "Certain invasive species, like mosquitoes in Hawaii or Asian carp in the Great Lakes, could be targeted with Cas9-based gene drives to either reduce their numbers or eliminate them completely." (source) unclear In general, it's unclear whether invasive species reduce or increase animal abundance on the whole. Cases would have to be analyzed one by one.

In general, gene drives enhance the power of humans to influence ecosystems. And given that most humans want to preserve ecosystems, we might expect a priori that gene drives would on the whole increase wild-animal suffering. On the other hand, human technology has historically probably reduced wild-animal suffering, which might lead us to think that's a more likely outcome with gene drives. The upshot is unclear.

### Catastrophic risks

One additional consideration is that the advance of gene drives and associated technologies probably increases catastrophic risks. For instance:

“Just as gene drives can make mosquitoes unfit for hosting and spreading the malaria parasite, they could conceivably be designed with gene drives carrying cargo for delivering lethal bacterial toxins to humans,” said David Gurwitz, a geneticist at Tel Aviv University in Israel.

The sign of net impact here is non-obvious for suffering reducers, but at least it's an additional -- and possibly crucial -- consideration to include in an analysis of gene drives.

## How good would eliminating mosquitoes be?

### Replaced by other insects?

In general I tend to be wary of many interventions to reduce wild-animal suffering that focus on ecosystem consumers rather than primary producers, since if we reduce the population of one animal species, will another animal species just take its place? In particular, mosquitoes eat mostly nectar, which is a common food for other insects. So if we kill off all mosquitoes, we might expect that in the long run, other bugs would eat the same nectar the mosquitoes ate?

However, it's less common for insects to eat the blood of animals, so if mosquitoes were gone, maybe at least the blood portion of their diets wouldn't otherwise feed some other insect.a Just making up numbers for illustration: If 20% of what mosquitoes eat is blood, then we might expect that a reduction of the mosquito population by N would increase other insect populations by 0.8 * N. Of course, the actual ecological effects of mosquito elimination would be much more complex and might dramatically reduce or dramatically increase insect populations overall. But as a first approximation, until we explore those details, it seems reasonable to use the kind of calculation just described as a rough expected value for the overall impact. Moreover, the ecological effects of eliminating those species of mosquitoes that bite humans might be fairly small: "Aedes aegypti are not a big food source for animals, and they don't pollinate plants."

Unfortunately, humans are unlikely to wipe out all mosquitoes:

In reality, as Lindsay and Ranson are quick to point out, the total extinction of all mosquito species would be as senseless as it would be impossible. Of the 3,000 varieties on the planet, only 200 or so bite us; only Aedes aegypti, and perhaps the more common Culex quinquefasciatus, are thought to carry Zika.

Moreover, other mosquitoes might take the place of the eliminated species after a while:

“Ecosystems would adapt,” says Ranson. “There are so many similar mosquito species that don’t transmit disease that something else would fulfil that niche without doing any major harm.”

### How many insects would be affected?

Another reason I tend to be wary of single-species interventions is that the numbers of individuals affected may be small compared with the total number of animals in a given habitat. This study reports that prior to releasing sterile male mosquitoes in one region of Brazil, the adult Aedes aegypti mosquito population was "an average of 418" per hectare. Since this was a region where an experimental population-suppression technique was being applied, it presumably had higher mosquito density than many other places on Earth.b If we assume that the average mosquito density -- of all the ~200 mosquito species that bite humans, and including both young and adult individuals, averaged over all months of the year -- over the whole Earth is, say, ~25 per hectare, then given that the Earth's land surface is ~1.5 * 1010 hectares, the world's total mosquito population might be around ~3.8 * 1011. In contrast, the world population of all bugs is ~1018. So eliminating all human-biting mosquitoes would reduce less than one millionth of all bug suffering, not even counting the fact that mosquitoes might be replaced by other bugs.

### Is pushing for mosquito elimination cost-effective?

Based on my limited analysis, it's plausible that eliminating human-biting mosquitoes is net good. Whether it's cost-competitive with other interventions depends on how expensive it would be to effect some change in its probability of happening. Since the decision about whether to eliminate human-biting mosquitoes is a mainstream issue with lots of public attention, it's doubtful that any one individual could make a huge difference to the outcome.

## Obstacles to using gene drives to reduce wild-animal suffering

What about the interventions that Pearce proposes for directly helping wild animals using gene drives? This section explores a few challenges with this strategy.

### Lack of interest

A main reason gene drives are unlikely to be used to directly reduce wild-animal suffering is that most people don't care enough about wild animals to incur the costs of helping them systematically in this way. Even 50-100 years from now, I expect that the idea of intervening in nature to reduce wild-animal suffering will still be pretty fringe and unpopular.

### Opposition due to ecological risk

Many people are nervous about the ecological side-effects of gene drives. George Church said that use of gene drives "has to have a fairly high pay-off, because it has a risk of irreversibility — and unintended or hard-to-calculate consequences for other species."

If people aren't even certain whether gene drives are worth the risk to save 600,000 children per year from malaria, people almost certainly won't think it's worth the risk to "play god" in nature to reduce "natural" suffering.

Of course, it's possible that sentiments on this topic would change. If a few gene drives are implemented for anthropocentric reasons and they work without major side effects, then perhaps gene drives would be allowed more liberally for interventions that humans consider less essential. Gene drives might turn out to have relatively minor impact compared with other environmental changes:

Micky Eubanks, an insect ecologist at Texas A&M University in College Station, says that the idea of gene drives shocked him at first. “My initial gut reaction was 'Oh my god, this is terrible. It's so scary',” he says. “But when you give it more thought and weigh it against the environmental changes that we have already made and continue to make, it would be a drop in the ocean.”

### Finding the right genes per species

Pearce cites the example of the SCN9A gene, which encodes the Nav1.7 sodium ion channel. Some alleles of this gene, Pearce says, "confer unusually high or unusually low pain-sensitivity without compromising function to any marked degree." If we drove this gene throughout wild-animal populations, then ignoring ecological side-effects, we could dramatically reduce the amount of physical pain in the wild.

However, it would require non-trivial effort to discover which genes to drive in which species to achieve the desired effect. And even if a given gene is present in many species (as is often the case), it may behave differently from species to species. For example, this study explains that Nav1.7 deficiency manifests differently in humans than in mice:

we have studied a Canadian family with four members exhibiting a congenital inability to experience pain.c [...]

Although the affected individuals in the Canadian family have a global deficiency of Nav1.7, they appear otherwise healthy. In contrast, it has been described that a global deficiency of Nav1.7 is lethal in mice shortly after birth (3), apparently caused by a failure to feed. Although it is possible that this difference is due to extensive parental care in humans, other factors could also be responsible. [...]

[We found] an apparent difference in the Nav1.7 expression levels and pattern in rodents versus primates. This difference may explain why a global deficiency of Nav1.7 is lethal in mice but not in humans.

Other researchers were able to create mice that didn't die from Nav1.7 deficiency:

We developed genetic and animal husbandry strategies that overcame the neonatal-lethal phenotype and enabled construction of a global Nav1.7 knockout mouse. Knockouts were anatomically normal, reached adulthood, and had phenotype wholly analogous to human congenital indifference to pain (CIP) [...].

One way in which the researchers helped the mice survive was to out-cross them with an outbred strain, "as outbred mice in general are more robust than inbred." However, the researchers also had to feed the animals extensively by hand. And even with these improvements, the survival rates of Nav1.7-deficient mice were poor:

A total of 130 adult knockouts were produced from the CD1 and BALB/c background hybrid strains from a total of ~2,450 neonatal knockouts, showing high hurdles to survival even with these more vigorous hybrid background strains and with intensive husbandry. With optimization of procedures, we have been able to achieve a ~20% survival rate in the expected knockout population.

So it's doubtful if these mice could survive in the wild. Maybe they could with a lot more tweaking, but at least, doing that would require substantial effort (and animal experimentation). And if we multiply this effort over a huge number of wild-animal species, the total cost of gene-drive interventions appears rather steep.

This study notes that "We speculate that Nav1.7 is an evolutionarily conserved molecular control point for the transmission of high-threshold noxious stimuli into the spinal cord. [...] SCN9A orthologs could govern similar functions in mammals, reptiles, and birds." So maybe for those taxa, one could just edit the SCN9A orthologs. But still, keeping the gene-edited individuals viable would require extensive tweaking across many different groups of animals.

### Would gene drives always work?

Suppression drives reduce the number of organisms in a target population. Naturally occurring suppression drives are always found together with a resistance element that allows the species to survive. It is believed that the emergence of drives in the past may have driven species all the way to extinction when resistance did not develop in time.

This suggests that human attempts to eradicate a given species using gene drives might not work if the species evolves resistance against the suppression. But in that case, maybe people could just try again with a new gene drive, and keep trying until the species finally went extinct?

This also suggests that non-lethal gene edits driven through populations might get undone over time through evolution, requiring deployal of new drives. Maybe this is what Pearce means by "annual maintenance costs" when using gene drives.

### Getting government signoff

Since gene drives are becoming less expensive to create, one could imagine rogue animal welfarists creating a gene drive to help wild animals without official approval. But such actions would probably be illegal and would be crushed by governments, at least after the first few such actions took place. Thus, this doesn't look like a workable long-term strategy for helping wild animals.

Of course, "the deliberate release of gene-drive-engineered organisms into the wild is [...] potentially anonymous." But even if animal welfarists could keep getting away with illicit release of gene edits, such actions would turn public sentiment against the movement to reduce wild-animal suffering, probably causing more harm than good.

Therefore, actually implementing gene drives to help wild animals would require government approbation, perhaps at an international level given that gene drives affect animal populations worldwide. This seems basically impossible in the world's current political climate, and I assume it would still be nigh impossible within several decades.

## Reducing pain sensitivity across all animals would dramatically change ecosystems

Modifying a subset of a wild-animal species to feel less pain would probably lead that subset of the population to be weeded out over time, on the assumption that reduced pain sensitivity is fitness-reducing. (If it weren't, evolution should have already reduced our pain sensitivities.) But gene drives allow for pushing a gene edit throughout a whole population even if that edit is fitness-reducing, as long as it's not so fitness-reducing that individuals with the trait hardly ever get a chance to mate. So it may indeed be possible to use gene drives to spread reduced pain sensitivity population-wide.

Of course, if this change were made only for one species, then that species would probably lose out in competition with other species that hadn't been modified. Eventually, the non-edited species would take over, undoing our efforts. Thus, it seems one would need to simultaneously drive pain-reduction genes throughout all competitor species in an ecosystem at once. And maybe throughout all trophic levels too, so that less pain-sensitive prey wouldn't be increasingly devoured by predators. But driving pain-reducing genes throughout a whole ecosystem all at once seems very challenging, especially since different species would pick up the edits at different speeds depending on generation times.

Moreover, even if we could wave a magic wand and make all animals of all species have reduced pain sensitivity at once, I suspect there would still be massive changes to ecosystem dynamics as a result. It's not possible to change behavioral dispositions so dramatically and not see a significant ecological response. Of course, the net impact of these ecological changes isn't clear -- would they reduce wild-animal populations on balance or increase them? So from the perspective of wild-animal suffering, these ecological side effects aren't obviously good or bad, but they at least add variance to our calculations. And given that most people don't like massive ecological disruption, the specter of these side effects would probably lead people to reject any ecosystem-wide gene-drive proposals.

Of course, spreading reduced pain sensitivity is not the only possible use of gene drives. Maybe there are other proposals that would somehow have much lower ecological impacts but comparable suffering-reduction benefits. It seems good for some people to continue exploring the possibilities in this space -- as long as we don't take any specific scenarios too seriously.

## Cost-effectiveness of reducing wild-animal pain sensitivity

Even though I think spreading reduced pain sensitivity would almost certainly be rejected by humanity, let's look at its cost-effectiveness anyway. Consider running an advocacy campaign to push for Pearce's intervention to spread pain-reduction genes throughout just 1% of sentient wildlife. Suppose that $1 million spent on such a campaign would increase the probability of this intervention being done by some amount P -- say, P = 10-5. Suppose that if it were done, the affected animals would suffer only half as much as before. And suppose the expected population size of wildlife would remain the same, despite signficant upheavals in ecological dynamics. Assume the effects of the campaign would matter for the next 50 years in expectation, after which either the biosphere would be gone or animals would have re-evolved increased pain sensitivity. Given ~1018 insects on Earth, the insect-years of suffering prevented per dollar would be (1018 insect-years per year) * 0.01 * 0.5 * (50 years) * P / ($1 million) = P * 2.5 * 1011 insect-years per dollar,

and with P = 10-5, that becomes 2.5 * 106 insect-years per dollar, which is, surprisingly, in the same ballpark as the mosquito-specific intervention from before.

Further exploration of this topic could probably quickly improve estimates of cost-effectiveness here, such as deciding whether P should be much higher or much lower than 10-5. In practice, before one spent \$1 million on such a campaign, one would want to survey attitudes toward the idea to better assess whether the campaign was worth undertaking.

## Memetic implications of talking about gene drives

Even if it's highly unlikely that gene drives will systematically reduce wild-animal suffering for more than the few species of wild animals that humans choose to eliminate for anthropocentric reasons, might there still be value in citing the possibility of gene drives in discussions of wild-animal suffering? If people think the only way to significantly reduce wild-animal suffering is to destroy habitats, cognitive dissonance may lead people to conclude that reducing wild-animal suffering isn't a good idea, even in principle. In contrast, if people could envision a scenario in which wild-animal suffering could be dramatically reduced without losing the rainforest or coral reefs, perhaps they would be less resistant. In other words, pointing to a gentler and theoreticaly eco-friendly way of reducing wild-animal suffering might win over more supporters.

As Pearce told me (personal communication, 11 Feb. 2016):

One of the reasons otherwise sympathetic people "switch off” if one starts talking about small mammals, let alone insects, is the sense helping them is hopelessly impractical on any systematic scale. Only if we can show that a policy option is technically and financially feasible can we begin to have a debate over whether it’s ethically desirable, or whether the risks are too great -- and our ignorance currently still too deep. Psychologically, too, most people "switch off" if they sense a policy option wouldn’t be feasible in their lifetime (one of the reasons I'd previously mostly stuck to a case study using free-living elephants, etc.).

On the other hand, given that wide-scale use of gene drives would probably cause massive ecological changes in practice, it's unclear if gene drives are actually as eco-friendly as they seem. And if we don't acknowledge this (though Pearce himself does discuss it), then critics will continue to assume, as most already do, "that any proposal to help free-living non-human animals is ecologically illiterate."

Moreover, in my opinion, the gene-drive solution has extremely small probability of actually happening, which makes it psychologically less appealing for those who are risk-averse, since with 99+% probability, efforts to get widespread gene drives deployed to help wild animals will never come to fruition.

It would be a shame if people rested all their hopes on gene drives and didn't pursue other approaches to help wild animals in the short run. But exploring gene drives as one "high-risk stock" within a diversified portfolio of compassionate-biology research topics seems reasonable.

## Footnotes

1. This isn't quite true. If an animal isn't bitten by mosquitoes, it won't lose tiny amounts of blood, so it may reduce its feeding by a tiny amount, since the animal doesn't need to replace lost blood. As a result, the animal might leave a tiny bit of extra vegetation on the ground to be eaten by someone else -- another big animal, an insect, fungi, bacteria, etc. Thus, fewer mosquitoes means a tiny bit more vegetation left over, but in general, only a small fraction of these left-overs will be eaten by other bugs.  (back)
2. The study said:

A dependence on stored water (due to irregular services for piped water) and high human densities provided ideal habitats for Ae. aegypti and thus the area supported a relatively high and stable year-round population that is atypical of fluctuating seasonal populations that are broadly prevalent in the region.

Since this mosquito population exists year-round, the number of insect-years per year is ~double that of a similarly big summer population of mosquitoes in a temperate climate where bugs die in winter.

That said, conventional mosquito prevention was performed in both the control and treatment locations, which implies a somewhat lower mosquito population than would be the case in regions that don't destroy mosquitoes at all:

Throughout the study, conventional local mosquito control was deployed as normal and public heath agents followed standard procedures. Teams of public health agents typically visited homes between 4 and 6 times per year, where they destroyed some breeding sites and treated others with the organophosphate larvicide, temephos. The same team of public health agents were responsible for the whole of Itaberaba suburb, ensuring that underlying conventional mosquito control was similar between the treated and untreated areas of this study.

3. Interestingly, it's not clear if these people remained pain-free for their whole lives. The study says:

Three of the initially affected individuals (X1226, X1230 and X1377) claimed to start noticing pain in their early teens (12–14 years old), whereas the fourth individual (X1387), examined at age three, did not display pain behavior. [...]

The observation that the patients in the Canadian family claim to start to feel pain during adolescence and thereon afterward may be reflective of the emergence of a true pain sensation during adolescence, but could also be a learned behavior. An emergence of pain sensation has not been described for the patients examined in the two other studies (9,10).