Blog‎ > ‎

Trust, Risk, and Control

posted Jan 14, 2015, 6:45 PM by Kevin Esvelt   [ updated Oct 25, 2016, 9:31 AM ]
As an inherently collective technology, deciding when (if ever) to use gene drives will always be a challenge. It's clear that any decision to deliberately alter the shared environment must be made by society as a whole. That process can only take place if the public can trust researchers to approach this technology with the utmost respect, caution, and humility.

Few events would damage this trust more than an accidental release. Gene drives are true self-propagating biologics: even a single organism that escapes the lab could eventually alter the entire wild population. Without the drive, the risk of this happening is effectively zero: there are vanishingly few if any examples of transgenes that benefit organisms in unaltered wild environments. This is the major reason why working with recombinant DNA in the laboratory does not pose an ecological risk: we can be confident that our altered organisms will not survive and proliferate in the wild. But if we're working with a gene drive, this expectation is flipped: we can be confident that a well-constructed gene drive will survive and proliferate in the wild.

As a scientist, I question whether we have the right to even develop technologies whose risk of spread is so high that a single instance of human error could potentially affect the lives of many other people. We certainly shouldn't be doing it without consulting them. There is a strong argument for peer reviewing many gene drive experiments at the proposal stage, before drive construction even begins, and releasing all findings to the public. 

At the same time, deciding if and when we should utilize a gene drive will require us to understand the potential consequences of our actions. Among many other factors, we need to know how the drive might evolve over time, how long the altered traits will remain in the population, whether related species might be affected, and whether reversal drives can reliably undo a given alteration. Models are highly informative, but won't be sufficient. We will need empirical data. That means testing gene drives in the laboratory.

Thankfully, there are ways to build gene drives capable of answering these questions that wouldn't be able to spread even if millions of organisms escaped the lab. One of the major reasons for publishing our eLife manuscript was to detail and emphasize the importance of these methods. The first form of molecular control involves separating the cas9 gene from the guide RNAs such that the drive can only spread through populations that provide the missing component; natural populations do not. The second form involves building drives that target and exclusively spread using sequences found in transgenic laboratory populations but not in the wild. Alternatively, performing the experiments in a geographic area where the organisms can neither survive nor find suitable mates should ensure that escape does not lead to gene drive spread. This is a form of ecological control.

These are not the only possible control strategies, and they should absolutely be backed up by conventional measures. For example, our collaborators in the Catteruccia lab at Harvard's School of Public Health perform experiments on mosquitoes in a secured-access room with seven doors between them and the outside, three of them equipped with air-blast fans designed to prevent mosquitoes from escaping. But as recent containment breaches at various high-security facilities have demonstrated, human error is both inevitable and surprisingly powerful.

For this reason, I strongly recommend that scientists employ multiple control strategies, especially in species posing a high risk of escape. For example, fruit flies are a terror when it comes to containment because they fly and are found worldwide. Indeed, scientists have documented the spread of a natural form gene drive called the P-element through fruit fly populations around the world in the last half-century. One of my own recurring nightmares involves media coverage of fruit flies that have mysteriously begun to fluoresce all over the world. To avoid this, gene drive experiments in flies should ideally use at least one if not two forms of molecular control and also perform experiments in a flightless background.

In contrast, yeast and the nematode worm C. elegans pose far lower risks of escape and spread. Since yeast reproduce without having sex considerably more often than they mate, the drive's advantage from inheritance-biasing is much reduced compared to species that always reproduce sexually. Hence, gene drives in yeast would need to be both highly efficient and minimally costly to spread in the wild. This is why our experiments with RNA-guided gene drives in yeast (results to be released soon) employed only one form of molecular containment in addition to physical containment precautions.

Worms are similar in that the vast majority of wild C. elegans are self-fertilizing hermaphrodites. I'm also less worried about scientists accidentally creating a nematode gene drive because it's very difficult to express transgenes in the germline due to a remarkable system that “licenses” germline gene expression. We will need to overcome this challenge if nematodes are to become a model organism for studying gene drives, but its presence ensures that accidental drive creation is extraordinarily unlikely.

What of scientists who make transgenic organisms using Cas9 but don't intend to create gene drives? Is there a risk of accidental gene drive creation? Yes, there is. But there's also an easy preventive measure: don't encode the gene encoding Cas9 and guide RNAs on the same piece of DNA, especially if it's linear. That's because some DNA repair pathways could insert a cas9 gene and guide RNA cassette into the site of the Cas9-induced break, effectively creating a potential gene drive organism. We're currently running some experiments to examine just how likely this is.

I can imagine many scientists suggesting that these precautions – especially proposal-stage peer review and public release of findings – are a massive overreaction. After all, the vast majority of laboratory gene drives are highly unlikely to cause any ecological changes whatsoever. That is because they will typically spread themselves and perhaps a marker gene encoding something innocuous like a fluorescent protein.

But we can't know for sure whether a particular protein might have an effect until we've tested it in a wild setting – ideally without a drive to spread it. It's remotely possible that even the cas9 gene used to spread the drive might have an effect, and it would be irresponsible of us to simply assume that it wouldn't.

That means we are morally obligated to take every precaution to avoid accidental gene drive releases – whether or not we intend to create gene drives. While the greatest risks are posed by highly mobile and globally distributed organisms such as fruit flies, all experiments intended to create transgenic organisms should be performed with great care.

Only by performing sensitive experiments with openness, humility, and an abundance of caution can we earn the trust and goodwill of the public. Act recklessly, and we deserve to lose their trust and support.