Human germline CRISPR raises major bioethical considerations, but what about technical issues?
Setting aside the many ethical issue about the general idea of human modification itself, could this really work? Yes in theory it could, but there are some very tough technological challenges that could and likely would cause failures or unacceptable outcomes at many steps along the way. These failures or unacceptable outcomes could easily involve real, live people who could be harmed or die. It’s very different than simple in vitro research so the tech side of it has been incredibly refined and robust. That’s not going to be easy.
What are the biggest technological roadblocks to responsible, successful (meaning safe and effective) use of CRISPR-Cas9 for heritable human genetic modification?
I see at least seven biggies.
This post is in part inspired by a section of my book on potential human use of CRISPR called GMO Sapiens.
1. Starting from point ACGT? Unknown initial genomic sequence of embryos
Let’s say we have a father-to-be, Mr. A, with a disease-causing mutation and a spouse Mrs. A who does not have that mutation. Dr. T, the leader of the scientific/medical team, gathers normal eggs from Mrs. A and then the sperm from Mr. A. Half of the sperm will be mutant and half will be the normal “wildtype” so the same will be true of the embryos that result from the IVF. Dr. T’s intention is to use CRISPR-Cas9 to revert the mutant allele back to wildtype in the embryos.
Dr. T would need to know the whole genome sequence of the embryos that the team will be trying to gene target. Here we have a Catch-22 situation. Dr. T’s team needs to know the starting genomic sequencing of the embryos before the intervention to design the CRISPR system components including the repair template with 100% confidence (although the very high level of sequence similarity between humans could allow for use of the so-called “reference” human genome for template design) and later to screen for off-target effects, but how does Dr. T get that information without destroying the one-cell embryos? She can’t.
So perhaps she settles for knowing Mr. and Mrs. A’s individual genomic sequences and then infers what the embryo/fetal sequence might be like? For example, for CRISPR guide RNA and repair template design Dr. T might assume that the mutated gene sequence in question will be from Mr. A in some embryos. They can also bring the reference human genome into play too.
It definitely feels a bit like flying blind. The best-case scenario would be that other than the mutation in question, the parents-to-be have exactly the same sequence in the overall region of interest.
2. Mutating wildtype normal embryos
Because the vast majority of possible clinical human embryo editing scenarios involve one parent with a heterozygous mutation and the other without that mutation like Mr. and Mrs. A, about half of their IVF-created embryos will have no mutation. So, Dr. T cannot know if she is attempting to edit WT or mutant embryos prior to CRISPR-Cas9 injection. Inevitably then they will be CRISPR’ing some normal embryos and probably mutating a subset of those via Indel creation. Is it permissible to genetically modify formerly WT human embryos that they intend to implant in a surrogate mother, potentially creating new disease-causing mutations in the worst-case scenario? I don’t think so. I don’t see how this could be avoided in this scenario.
Keep in mind again that if you sequence a few embryonic cells by PGD (more on that below) later during embryogenesis, a WT sequence showing up at the gene of interest could mean (A) this started out as a WT embryo or (B) it started out as a “mutant” embryo and you successfully corrected the gene mutation. How can you tell the difference at this level?
Let’s say that Dr. T encodes some clever, tiny “bar code” signpost via the CRISPR gene editing beyond the mutation correction to be an indicator during sequencing of a successful gene edit (rather than just a pre-existing WT allele). This could differentiate between correction or just starting with a WT embryo. However, then could the indicator–say something as small as 2 non-codon changing basepair changes–could cause trouble? Who knows?
3. Off-target effects
CRISPR-Cas9 is great and getting better (and we even know have an expanded toolbox with the ‘base editor’ chemical modification tech), but they aren’t perfect. There will always be a risk of it making edits in the wrong place in addition to a chance of making wrong edits such as Indels in the right place. Since Dr. T does not know the embryo’s starting genomic sequences, how does she know if CRISPR-Cas9 created an off-target mutation (e.g. as little as a single bp change that was undesired) or if instead that detected unexpected DNA sequence was just the unique sequence of the embryo to start with at some random place in the genome? There are a lot of naturally occurring variants including SNPs. Again, Dr. T would need to go back to Mr. and Mrs. A.’s sequences and hope she can get some clarity there along with the reference human genome. However, it is easy to imagine scenarios where the team just couldn’t be sure if a different say between a G and T at one place in the billions of genomic basepairs was a variant or an off-target effect.
Another practical confounding issue here is that Dr. T’s team is working with human embryos with few cells to use for screening. Ideally they need to screen for off-target effects so they need to be able to do whole genome sequencing (WGS). Can they accurately and reproducibly do WGS from only 1 or 2 embryonic cells (blastomeres) with accuracy down to the single basepair resolution across the entire genome? Of course, one could wait and get more cells from a fetus later after implanting the embryo, but then if problems arise we are talking about the possibility of an abortion coming into play.
4. PGD would usually miss mosaicism
To do the kinds of sequencing mentioned earlier, Dr. T is going to definitely do PGD and this will also be where she looks for mosaicism (i.e. some of the embryo’s cells have gene edits, while other cells in the same embryo do not and also there is the possibility of mosaic off-target effects). The team needs to know if there’s mosaicism and presumably if there is then they would halt the process as they do not want to create substantially mosaic humans who could have serious health consequences as a result.
Unfortunately if you intend to do full genome sequencing by PGD at the 8-cell stage you are relying on just one or two cells (or perhaps a few more if you do PGD at the blastocyst stage) to be predictive of the whole embryo. PGD of just one or a few cells could give you an entirely wrong view of the genotype of the other cells in the gene edited embryo. You might well incorrectly believe there is no mosaicism when in fact there is variability amongst the cells of the embryo that you just failed to detect. Again by PGD you should also be looking for off-target effects and you don’t want to limit that search only to one or a handful of cells either. PGD is crucial, but only a partial snapshot.
Another hurdle discussed in a previous blog post is coming up with a logical reason for using CRISPR in a clinical way in humans that would make it better than simply using already existing PGD technology by itself (without gene editing) to screen for embryos not possessing an inherited mutation. Why would Mrs. and Mr. A and Dr. T even want to try gene editing instead of using PGD all by itself? It’s difficult to come up with many scenarios where PGD alone wouldn’t work and where gene editing would be substantially better. An excellent Nature Biotechnology piece covers this question of when human modification might make sense. For instance, I thought Robin Lovell-Badge’s response was very cogent. One example is the rare situation with a parent-to-be who is homozygous for a dominant mutation or if both parents have disease-causing mutations.
5. The use of thousands of human embryo for optimization
The old saying goes that “practice makes perfect” and that kind of sentiment applies to CRISPR’ing human embryos even if it can never be 100% guaranteed to be perfect. Most attempts at human embryo editing in the lab are still likely to be informative so a knowledge base will build over time and improve the gene editing technology and methods. It will probably take many thousands of human embryos to optimize the system collectively and every specific lab doing a distinct kind of gene editing may require hundreds of embryos for its own optimization.
Is it acceptable to do such massive scale viable embryo editing simply for advancing knowledge? Also, where are you going to get all these eggs and embryos? I’m a supporter of human embryonic stem cell (hESC) research and embryos remaining from IVF procedures are used to make hESC (or are otherwise generally discarded), but not a tremendously huge number. In contrast, hypothetical yearly use of thousands of potentially newly generated human embryos simply to optimize gene editing and/or for advancing knowledge could start to get ethically and practically complicated. See my recent piece on the Mitalipov lab apparently already making and using hundreds of human embryos for CRISPR.
6. Trapped in a choice of “lesser evils” post-implantation?
Let’s say somehow Dr. T successfully gets further along and the team has got a pregnancy with a genetically modified human embryo. If a problem then arises, what can they do about it? The team involved in this work could well find themselves trapped with a dilemma as to how to handle an adverse situation. The only options might both be problematic: (1) continuing a risky pregnancy of a human embryo/fetus with CRISPR-introduced genetic errors or (2) abort the pregnancy. If mistakes are relatively common in such clinical CRISPR research, is it OK to routinely abort such fetuses if problems arise?
Finally, what if health issues become apparent in Mr. and Mrs. A’s gene edited children only much later on down the road?
7. Unintended consequences.
The genome is a complicated jungle so even if you make the “right” edit with no off-target effects, how do you know you’ll get just the narrowly focused outcome you want?
Possible solution to some problems: gene editing in germ or stem cells?
The above discussion assumes a focus on gene editing conducted in one-cell embryos, but it is also in principle possible to gene edit mutations in germ cells. For example, one might do CRISPR in oocytes or even primordial germ cells (assuming successful working out some of the kinks in producing such sperm and egg-producing cells safely in humans), validate gene correction and lack of off-target effects in the cells prior to fertilization, and then proceed with IVF, implantation, etc. with the gene editing now in the rear-view mirror so to speak. This could resolve some of the issues mentioned earlier. At the same time this approach may have issues of its own such as the risks associated with prolonged manipulations of germ cells in the dish in the lab. It is also possible that the use of cultured primordial germ cells would pose unique risks as the cells change their epigenomes during their growth and manipulation in the lab. Still, this kind of approach is another, interesting option.
Bottom line. Overall to me the big picture at this time at least is one of serious technical hurdles in the way of responsible possible future clinical human genetic modification. Technology will improve and we may come to see solutions to some of these problems, but it seems unlikely that all these issues can be resolved completely. Throw in the numerous thorny ethical and legal issues and it seems even more difficult to imagine a future where there could be responsible, safe human germline genetic modification done with a unique, beneficial purpose. Despite all of this, I do believe that some people will go ahead and try making genetically modified people anyway.
Any responsible discussion of possible heritable human genetic modification needs to include dialogue on these kinds of technical hurdles and problems. When someone is aspirational about CRISPR germline use, it is worth asking them about these sorts of hurdles and also about what specific positive use they had in mind for heritable human gene editing that transcends what embryo screening can already achieve.