He Jiankui’s experiment has brought the world’s first deliberately gene-edited babies into the world, twins nicknamed Nana and Lulu. He edited the DNA of their embryos, hoping to make them resistant to infection by HIV. I see a lot of people writing comments like, “Genome-edited babies – what’s the worry? We’re going to start it sometime. We can start here.”
This misplaced enthusiasm thinks genome editing ‘just works’, but it’s not that simple. In the longer-term broad genome-edit embryos might be inevitable, but what was done in this case is not “OK”.
Balancing this is misplaced worry, not helped by sensationalist headlines.
The devil is in the detail, they say.
Here I’ll give a ‘lite’ take on the biological issues. This one is for every reader. If there is interest I can give geekier a take later. I’ll leave ‘administrative’ concerns out, even though they’re important, real and there are concerns there. [‘administrative’ includes most of the ethical concerns – for a later post.]
Every egg is sacred
(The Pythons satirise Catholic teaching in Every Sperm is Sacred. Unless you have infertility due to a low sperm count, sperm are trivial. One millilitre of normal ejaculate has millions of them. It ought to be every egg that is sacred.)
A woman donating an egg is a big undertaking and you don’t capture many. You start human embryo work (of any kind) with precious few eggs, and the failure rate is high. As a result, you want to get the editing right the first time. To edit embryos you need a high standard.
We’re not there yet. There are many techniques that do different types of editing. We can edit with varying percentages of accuracy. We sometimes get it right, even mostly get it right for some settings with some techniques. And we don’t have a big collection of candidate embryos to screen ‘winners’ from.
For embryos, if it’s not right or you don’t know what they’ll do you don’t take those embryos further.
The edits made by He’s team haven’t been tested, but they didn’t opt out.
There are three main types of errors:
- missing the target and changing stuff elsewhere
- creating mosaics, people with a mix of different genomes
- mistakes on the target, getting stuff you didn’t want
There’s a lot of talk about the first online. It’s easy to think about, and sounds as if things have badly messed up. In practice the second two are likely to be more of an issue.
Changes to the germline are inherited
It’s not just the twin daughters that will have the new gene variants (alleles) but they’ll be passed on to their kids. The edited versions of the genes are inherited.
By contrast gene therapy, which can be used to tackle some rare genetic diseases by changing the DNA of some cells in the person is not passed on.
Off-target changes: erm, Houston, we kinda missed
A concern most people will have heard of is when genome editing makes changes not where it was intended to. They’re called off-target effects.
The method used to edit Lulu and Nana’s embryos was tested in other species. There it showed few off-target effects. The method used on the babies was an adaption of the method tested on primates.
Testing Lulu and Nana’s DNA suggested there are no off-target effects. We’re told that there will be more testing off this from blood samples later.
You also have to bear in mind that (some) animals with many off-targets with no ill effects: just because an off-target change happened, it doesn’t have to have an effect. That said, there may still have been off-target effects that were missed. Time will tell.
Mosaics: blotchy people
If the editing happened in only some of the cells in the embryo, or different cells in the embryo got different edits, the baby will grow up as a mixture of cells with different genetics. Geneticists call people with a mixture of genetics mosaics. (More formally we’d say a mixture of genotypes.)
The classic example of a mosaic is the tortoise-shell or calico cat. Females have two X chromosomes. In each cell one is packaged away and shut down. That’s done randomly. A gene affecting coat colour is on the X chromosome. If the cat has two different alleles for coat colour, it’ll have a patchy coat from randomly shutting down different X chromosomes. (For more on this, see Sea stars and mosaics.)
Something similar can happen in genome editing, where the growing embryo ends up with a mixture of cells with different genetics.
If a child has a mixture of cells that are resistant to HIV and some that are not, then the child is still susceptible to HIV infection.
Some scientists have said that the data indicates that it’s possible that both Lulu and Nana are mosaics. If true, then nothing may have been achieved in terms of resistance to HIV infection.
If this doesn’t harm the kids, then you might says that’s just a wasted effort rather than harm, but it compounds that this wasn’t something that had dramatic medical need in the first instance (see Meeting an ‘unmet medical need’ below).
On-target mistakes: OK… that’s not what we wanted
This is a real concern.
The changes made
He Jiankui’s team aimed to shorten the CCR5 gene, disrupting it. There’s a guy from Berlin with resistance one type of HIV infection (X4). The Berlin patient has a natural mutation where he lost 32 bases of his CCR5 gene. This causes the reading of the gene to stop a little after the deleted portion. He’s team are trying create a similar variant.
The natural ‘Berlin delta-32’ mutation has been studied extensively.
The variants made in Lulu and Nana are not the same as this mutation and they have not been tested. He’s team are hoping that other truncations more-or-less similar to the ‘Berlin’ mutation act the same way.
One of the changes, in Lulu, is completely different.
Despite this implanting the embryos went ahead.
The CCR5 variant in Lulu has just 5 amino acids removed from a part of the CCR5 protein that sits on the outside of the cell. This might interfere with HIV binding to the CCR5 protein. It also might do nothing much. We don’t know.
Nana has two CCR5 variants. Both truncate the protein, and might act the same way has the natural Berlin (delta-32) variant. Or they might not. We don’t know.
The outcome of both are speculations, really. They might not do any harm but they’re still speculations.
That’s not really good enough for this sort of work, doubly so because it’s avoidable.
The researchers could have tested these or aimed for a knock-in.
To test them they could have frozen the embryos, then spent time (years) testing them. Ideally they’d let experts in the CCR5 gene know as they’d have better ideas of what to look for and test. The secrecy of He’s project doesn’t help here. You need to reach out for these sorts of things.
Knock-ins are where you put in a specific variant. What they’ve done are knock-outs. Like boxing knock outs, they’ve tried to slap the gene down. Knock-outs are easier to do: most things that muck the gene up will make it not work. Knock-ins are harder to make.
Given the low success rate they had trying for knock-ins might have nixed the work, but at least you’d be aiming for a natural mutation that has been well studied—and the variant that the thinking of the work was based on.
Can we screen out these?
Yes and no. You hate those non-answers right? But it’s like that actually. It’s one of the subtleties that scientists are seeing, but most others aren’t.
He Jiankui’s team sequenced the DNA of the embryos by (gently!) knocking out a few cells (3-5 or so). A catch is that different DNA sequencing methods are better at detecting different changes.
It’s easy to use that phrase “sequenced the DNA” as if that was that. In practice there are a range of different ways to sequence DNA, and there are many computational methods to pull the results back together and interpret the results. (When you sequence DNA you get a huge pile of little pieces that you get get assemble into chromosomes.)
Since you’ve got a tiny sample—just a few cells—you just can’t do all the tests you might ideally like to.
Some scientists suggest that the sequencing used isn’t the best for identifying large changes. Genome editing can, sometimes, move around or delete large chunks of chromosomes. We’re hearing that these are to be investigated in the children, which is later than you’d like.
I want to call out that last step too. As far as I can tell, He Jiankui’s talk at the Gene Editing Summit didn’t say what computational methods were used at all. As a computational biologist, I’m not happy with that. Those methods matter, too. (To be fair he will have had limited time to speak but it might have been included on a slide.)
Meeting an ‘unmet medical need’
Current ethics encourage efforts towards genome editing (and gene therapy) to be limited to life-altering conditions where there are no reasonable alternative treatments.
A common objection to He Jiankui’s work is that it does not meet this ‘unmet medical need’ because there are ways to reduce the risk of HIV infection in children of parents with HIV.
You can reduce the risk of parent-to-child transmission in the foetus by ‘washing’ to remove the semen (which He Jiankui’s team did), or using Intracytoplasmic sperm injection (ICSI, see About the featured image at the end of this article.) Similarly, there are pharmaceutical treatments to reduce infection risk once the children are older.
In Western eyes at least, this leaves He Jiankui’s treatment as closer to an elective choice rather than something that might be considered essential.
Genes often are re-used in our bodies for more than one purpose, or play a role in more than one pathway. In a sense the CCR5 is an example of this.
Deleting the CCR5 gene reduces infection by some strains of HIV but deleting it increases risk for flavivirus infections, such as West Nile virus. Successful editing of CCR5 would effectively swop the risk of one for the other.
If Lulu and Nana have reduced risk for HIV infection, they’ll also be at higher risk for West Nile virus and other flavivirus infections.
The backfire effect
On concern scientists have is that if an initial attempt looks bad, it might lock up something that has a lot of promise. A real-world example is gene therapy. There failure of the initial cases delayed the field decades. (Or caused people to take stock.) We’re only now starting to see a string of successes.
There’s a stereotype in movies that scientists recklessly try out new things. In my experience individual scientists might do this, but science as a whole can be quite conservative when it comes to practical applications. They’re open to exploring all sorts of things in a laboratory setting, but are usually wary about bringing things to real-world use that haven’t been fully tested.
You see some of this in the protests from scientists about He’s work.
One reason is that they worry about a backlash from early attempts going sour preventing later sound work – like the gene therapy example.
It’s a reason scientists had wanted the first case of genome editing to be very sound, done very cautiously, and for their to be an abundance of supporting evidence. (See also the ‘extra-ordinary claims require extraordinary evidence’ mantra.)
In a more robust world, we might say “look, initial attempts often don’t pan out, but we learn from them”. In the Western world at least, public reactions can block useful applications.
On the other hand
There are two ‘sides’ to everything. (And sometimes—perhaps usually—many ‘sides’. I’m putting sides here in scare quotes, as I don’t like oppositional, adversarial approaches to topics. It’s a key element in the failure to advance things like new agricultural varieties developed using genetic engineering. Once a topic is set up as adversarial by lobby groups, the stage is set for political point-scoring rather than science.)
For balance it’s worth remembering a few things –
- He Jiankui is a scientist, not a complete ignoramus as some characterised him online. His previous work includes testing genome editing of CCR5 in other species, and he will know DNA sequencing well (it’s his speciality). This doesn’t mean he hasn’t over-reached, though.
- The conditions genome editing (and gene therapy) tackle are usually rare.
- We are altering one gene of many.
- We all have mutations, every one of us. Perhaps a hundred changes that are not in either of our parents (de novo mutations), never mind the changes (mutations) that are parents passed on to us. The large majority of them have no real affect on us. The same will likely be true of most off-target changes, where a genome editing enzyme changes something away from the intended target gene. It’s still much better to get it right.
- A lot of people’s concerns are moral, rather than ‘actual’ in the sense of something is a biological concern. That doesn’t mean they should be ignored, but it does mean they are a different kind of concern.
- Some objections to He’s work look to not fully appreciate cultural differences and the perspective of Chinese on what he was trying to achieve. This doesn’t make them ‘right’, but it does help to understand better what was intended.
Other articles on Code for life
For the numbered footnotes, see the second sub-section, The footnotes.
I’ve written a lot of this from recall. It’s part of the reason I’ve focused on conceptual points. Geeky specifics can come later!
If you see any errors, let me know. It’s been a struggle getting this out (it’s a long, boring story), and I’m reluctant to spend time chasing down each fact a second time from my extensive collection of notes.
Right now I’m hoping to write (at least) three posts on this,
- Genome-edited babies – what’s the worry? (#1)
- Genome-edited babies – how we might do better (#2)
- Genome-edited babies – not the same as agricultural GMOs or gene drives (#3)
I’ll likely include differences to somatic gene therapy in the third part, or perhaps a separate post. Likewise I’d hope to quickly cover what was done in the second part, as that’s what you’d compare with to do better. I could also explore the response more – in media, from scientists and everyone in general: there are quite a few issues there, too.
If readers have suggestions or have particular questions they’d like answered, you’re welcome to raise them in the comments.
You can think of my earlier post, Human gene-edited babies: hold the horses, as a sort of prelude to the series.
- Chinese names have the family name first. It’s customary to refer to scientists by their surname once they’ve been introduced in a piece. His surname is tricky for English-language use! (Perhaps doubly so for a few strict Christians.) Apparently in the USA He went by the nickname JK.
- You (can) have big panels of candidates to screen modifications to agricultural species. Hopefully I’ll come to this later in part 3.
- Hopefully more on this in part 2, as it’s about the process they used.
- I can give more details of the edits later if readers would like that.
- Genes are read taking three DNA bases a time. If a deletion is not a multiple of three, it’ll mean that the reading will ‘shift’ to reading groups of three bases different to how the normal gene is read. This usually means that the reading will be garbled for a while until it randomly runs into a three-base sequence (triplet) that says ‘stop reading’. That’s what happens here.
- Years ago I proposed a database or ‘federation’ of experts in particular protein families thinking this would be a useful for experiment researchers to connect to experts, and it would help support curation efforts. (I was an expert in a family of gene regulatory proteins that are used to shape genomes and control what sets of genes are available to be used in different types of cells.) Today collections like these are largely automated. Sometimes I think something has been lost; while more data is available, if anything there are now fewer experts in the protein families themselves.
- It’s not an apples-to-oranges comparison, but you can see some of this in how use of new plant varieties developed using GE have been blocked in some countries.
About the featured image
The photograph is of Intracytoplasmic sperm injection (ICSI). Usually in vitro fertilisation (IVF) would fertilise an egg by mixing the sperm and egg and let fertilisation occur. This requires there to be a large number of sperm. Using ICSI a small number of sperm can be directly injected into the egg, even just one. This can be useful where infertility is because of low sperm counts.
He Jiankui’s team is very unlikely to have used this technique. I’ve included it because ICSI is one of the options for reducing the risk of the baby getting HIV from the father. Several other scientists suggested it as part of an alternative to genome-editing the babies.