4 key reasons Mitalipov paper doesn’t herald safe CRISPR human genetic modification

CRISPR human genetic modification dart board

We can be confident that human genetic modification via CRISPR’ing of embryos soon will be safe and effective after that new exciting Mitalipov team paper, right?CRISPR human genetic modification dart board

Wrong.

The reality is far more complicated and interesting on the tech side.

In a nutshell, I see the paper as a significant scientific, but not necessarily medical advance.

The media coverage overall has painted too rosy and simplistic a picture of the Mitalipov paper. Setting aside tough bioethical and societal issues regarding human genetic modification for the moment in this post, there are 4 major, mostly overlooked technical reasons why the Mitalipov CRISPR human embryo paper is unlikely to herald safe human genetic modification.

1. I dream of Gene-y. The “best” disease-associated gene to target with CRISPR in humans would be one that CRISPR always “hits” perfectly. The Mitalipov team probably gave deep thought to picking the gene they did for CRISPR targeting. It’s a gene going by the name MYBPC3 that is associated with a fatal type of heart defect called familial hypertrophic cardiomyopathy (HCM). That disease association is an important reason for trying CRISPR on the mutation in this gene. However, most likely at least one big technical reason they chose MYBPC3 was because it had so few predicted possible CRISPR off-target sites (meaning other places in the genome found by computer algorithms that CRISPR might stray to and make damage). If I’m right about this, then they were just being smart in that choice and I would have probably done the same thing in their shoes. But the targeting of this hand-picked gene means that the upbeat findings on accuracy with reverting mutant MYBPC3 are probably unlikely to be representative of efforts to “gene edit” disease-causing mutations more generally. By analogy would you like to throw a dart at a dartboard where the bull’s-eye takes up fully two-thirds of the dartboard (MYBPC3?) or where the bull’s-eye is just one thirtieth (some other disease-causing mutations)? For many diseases you may in effect have no choice but to go for the far tougher bull’s-eye because of the nature of the particular gene and its disease-associated mutation.

Isn’t it possible for all major disease-associated gene mutations that CRISPR will work as well (or even better) than MYBPC3? Nope, that’s not the way the real world works unfortunately.

In fact, one of the authors (Jin-Soo Kim of the Institute for Basic Science in Daejeon, South Korea) specifically emphasized to Nature News the low predicted off-target rate of this gene:

“Even so, Kim notes that the CRISPR–Cas9 error rate can vary depending on which DNA sequence is being targeted. The MYBPC3 mutation, in particular, was predicted to produce relatively few opportunities for off-target cutting.”

It is also possible the team picked MYBPC3 because its mutation is very small (only 4 base pairs and in theory easier to repair) or they had the human sperm donor lined up with this mutation.

A combination of factors most likely guided the team in gene choice.

Other gene mutations are going to be far tougher because they will be prone to dramatically more off-targets (see below) and perhaps more Indels (see below). Many mutations are large and complicated as well.

2. Off targets there, but not oft found? There’s also the likely possibility that the team unintentionally missed finding some off target effects of CRISPR that were in their modified embryos, but not found because of how they did the sequencing. The very next quote in that Nature News piece is from Keith Joung on this concern:

“Just because the team did not find off-target changes does not mean that the changes aren’t there, cautions Keith Joung, who studies gene editing at the Massachusetts General Hospital in Boston. “Although this is likely the widest examination of off-target effects in genome-edited human embryos performed to date,” he says, “these investigators would need to do much more work if they wanted to define with certainty whether off-target effects do or do not occur in this context.”

Give the Mitalipov team credit for the screening they did do, which was relatively a lot, but much more is needed to be even close to sure about this, especially if one has clinical hopes as this team does.

3. Indel pain in the neck. The metaphor behind the language and concept of “gene editing”, the preferable phrase to “genetic modification” within the scientific politics of today (admittedly I sometimes use this phrase myself), suggests precision changes as do other metaphors like “genome surgery”. However, even if CRISPR-Cas9 avoids off-targets and sticks to the gene of interest, it can often make these things called “Indels” short for insertions and deletions right in sweet spots in genes. Indels often functionally kill genes entirely rather than precisely changing them. The Mitalipov team found Indels more than 1/4 of the time in embryos even under their most optimized conditions. The rate of Indels needs to be at or very close to zero to begin to have any reasonable chance of clinical safety of using CRISPR in the human germline. Plus, at other mutant genes that may be targeted in human embryos, Indels may be much more commonly created by CRISPR than at MYBPC3. We just don’t know.

4. Mosaic monitor. Mosaicism with CRISPR is where the cells of the embryos after introduction of CRISPR-Cas9 machinery don’t all have the same genome any more. For example, some cells in the same embryo may be normal and some mutant. There’s a genomic gemisch. That’s generally not good for ultimate health so mosaicism would be unsafe for hypothetical clinical applications of CRISPR in humans. One of the most impressive things about the Mitalipov CRISPR embryo paper was that they reportedly got rid of most (just 1 mosaic found) mosaicism in CRISPR’d human embryos. However, this was essentially just a very narrow test case study with one male sperm donor and one or a few women who donated eggs. Thus, the embryos used were very similar. More broadly, there is likely to be substantial variability in propensity to embryo mosaicism in part related to the unknown characteristics of specific gamete donors.

PGD reminder. Beyond the technical challenges, the fact is that almost anything CRISPR could do of medical use heritably in humans is already achievable using embryo screening including by the common, proven method called PGD. Think of it this way by analogy. Let’s say you have 8 books with 4 having errors and 4 not having errors. You have a very reliable way to know which books are which, and you only need 1 correct book. Do you try to correct the 4 errant books knowing that you could easily make more errors yourself in trying to fix the error, or just pick from one of the easily identifiable perfect ones?

Bottom line. For all these reasons, we should all be more cautious in making meaning from this one paper. There’s a long tough road ahead with a marathon of challenges (and the authors rightly acknowledged many of these so kudos to them) if one has clinical aspirations for CRISPR in the human germline.

2 Comments


  1. Very well explained, Paul. I was in a discussion at the BEINGS bioethics meeting a couple of years ago (http://www.beings2015.org/) about how necessary it would be to use CRISPR editing of embryos. George Church was arguing that diseases like Huntington’s disease could be treated with CRISPR, and I took another perspective- maybe just to be contrary to George, but also because I am a very practical person.
    I pointed out that preimplantation genetic diagnosis (PGD) could wipe out many genetic diseases in a generation, and that IVF with PGD should be covered by insurance in the US, as it is in many other first-world countries.
    To decide whether or not to use IVF, It would be necessary to know if, even though you are healthy, you are a carrier for recessive disease, like cystic fibrosis. Luckily, there are easy genetic tests for those- even 23andme can provide the data to find out. Check with me if you have done the 23andme test and want to know more than the information they are allowed to provide. I can give you a link to another site where you can analyze your results for more information.


  2. Great update Paul, room for a 5th? Am I right in thinking that the repair was directed by the maternal strand? Does this suggest the injected DNA template was faulty/inefficient? This presumably limits the types of edits you can do

Leave a Reply