Hello all,
Recently I performed a CRISPR experiment where I used a single guide RNA to cut a gene, and supplied a ssDNA repair oligo to introduce a point mutation and restriction site.
In the F1 offspring of the injected P0, I isolated a worm that was homozygous for the repair event. As in Cas9 had cut both the egg and sperm allele, and both were repaired by HDR.
These two events happened independently, and therefore technically are separate alleles, however they are impossible to distinguish and can never be separated.
CRISPR is so new that there may not be a standard for this yet, but can this F1 worm be accepted as a homozygote for one allele, even though they were generated independently?
For now that is how our lab is proceeding. I’m curious on your thoughts.
-Ed
An old-school geneticist would probably insist on outcrossing multiple times before naming anything. This would separate the sperm-derived and egg-derived alleles and would guarantee that you really do have only a single allele, which you would then name.
That’s probably formally the best way to do things, but personally I think it is overkill. If your goal is to protect yourself against second-site mutations, outcrossing will not protect you nearly as much as starting with freshly-thawed N2s as your background, and isolating multiple alleles of your desired mutation. I recognize there’s room for disagreement on that point and would be curious to hear from others.
If you are interested in a compromise approach that avoids the issue altogether, one suggestion would be to deliberately pick a heterozygote in the F1, then let it self to get two copies of the same edit in the F2. This is more work (more PCR), but cleaner because you have only a single edit.
It is not clear to me how picking a homozygous F2 from a heterozygous worm evades potential secondary mutations. Naively I would expect that the profile of secondary mutations
in a homozygous F1 vs. a heterozygous F1 are similar. In this case, isolating a heterozygous worm and then generating a homozygote would select for secondary mutations that are
linked to the edit site, whereas in a homozygous F1, there is no preference. In either case, if the secondary mutations are substantial, then you would have to ultimately outcross the
strain. But fundamentally, is it a risk to take a homozygous F1 and treat both edits as a single allele?
In regards to isolating multiple alleles, I isolated eight F1 worms that are homozygous for the edit. Can these duplicates then safeguard me (treated now as each an independent allele),
or is there still some advantage to selecting a hetero F1?
-Ed
I’m not convinced that it would even be possible to get an animal where the modification happened both in sperm and the oocyte since they are produced asynchronously. The sperm are made when you injected the hermaphrodites, and I don’t see any way that the Crispr/Cas9 could have penetrated the fully formed sperm. If it was an RNA/protein injection, it could have perdured after injection and converted the 2nd allele after fertilization. If it was transgenic, then the conversion could have happened in the germline of the F1 you picked.
I would outcross to wild type, genotype for one of the ‘alleles’ and call it a day. If they are independent F1’s, then you can say they are independent alleles, but I don’t know if you can truly every say which allele was which unless you had additional linked marker criteria.
-Kevin.
I agree with your statements about the sperm allele. I am injecting pre-assembled ribonucleoprotein complexes. I don’t mean to say the injection mix somehow traveled to the sperm itself. I suspect the RNP persists in
the nucleus of the oocyte and modifies the sperm allele post-fertilization. We see these events occur quite frequently. Additional evidence of this model is that we can obtain homozygous
edited males in the progeny of a hermaphrodite that was mated prior to injection.
-Ed
To clarify - waiting until F2 to pick homozygotes doesn’t buffer you against second-site mutations. For that you’d have to outcross or (better still) work with multiple alleles generated on a clean background. But picking homozygotes at F2 will ensure that both alleles derive from a single initial editing event.
Eight alleles is more than enough - we usually keep and analyze four of each edit. Just make sure that your lines each derive from a different injected animal - otherwise they are not independent.
I would add a practical note: after injection of the P0, an F1 that appears homozygous for the targeted change might not be so, and should immediately be outcrossed, even if you wouldn’t bother to outcross the homozygous descendant of an F1 heterozygous for the targeted change.
This is because the CRISPR targeting the site, together with the repair oligo, can do at least 5 different things:
- It can make exactly the change you wanted, that the oligo is designed to cause.
- It can make another small change at the cleavage site (by NHEJ, by local microhomology, or by imperfect incorporation of the oligo or of a flawed oligo)
- It can induce a large deletion
- It can induce a complex rearrangement, the includes the desired, targeted change but also reiterates wild-type sequence
- It might do nothing at one copy of the locus
#4 is unlikely and your PCR screening would probably have detected and rejected it, if not immediately (when it looks like heterozygosity) then later when you tried to homozygose it.
#2 and of course #5 are more likely, but again would be detected and identified as what they are.
The problem is with #3: #3 is not detectable by PCR unless it’s homozygous and the PCR failure is correctly interpreted. And, critically, an animal heterozygous with one copy of type #1 and the other copy of the type #3 is not distinguishable from a #1/#1 homozygote (not by PCR, and perhaps not phenotypically).
For this reason, you should immediately outcross an F1 you think is a #1/#1 homozygote, or should in preference work with F1s you take to be heterozygous.
This suggestion has the additional merit of rendering your otherwise difficult-to-resolve question moot. If, however, I had to answer the original question, I would say that in an actual #1/#1 homozygous F1 the two copies may represent separate events, and may have different things tightly linked to them, but that no-one could ever know the which of these two mutations was present in any resulting strains. For this reason they should be given a single allele number (as there’s no way to knowingly separate them from each other, knowing which is which) but should be well annotated explaining their origin and that different copies of the same allele might have different linked mutations.
I see, Dan. The only way to truly isolate a single allele and know it is one allele is to homozygose an F1 hetero. Can you explain to me how one P0 can give rise to multiple edited
F1 worms that harbor the same allele? That’s a really cool phenomenon that I previously didn’t know could occur with CRISPR. Normally in our protocol, we keep all injected P0s on one
plate, transfer them together at 8 hours post injection, and then again at 14 hours post injection. This 8-14 hour plate is greatly enriched for edits. But if it is true a P0 can produce
multiple F1s of the same allele, then this is potentially a bad idea.
HillelSchwartz, thank you for that well thought out explanation. It is surely possible a worm hetero for a large deletion could masquerade as a homozygous edit and I had not thought of that.
The only way to know would be to genotype a slew of the offspring and make sure that I don’t get 1/4 failed genotyping PCRs. At that point I may as well single from a hetero F2 to homozygose
the allele as that is the same amount of handling.
One other case for avoiding homozygous F1 worrms that occurred to me is that it is possible that in order for the RNP to persist until post-fertilization, there may need to be a high amount in the
nucleus in the syncytium. It is therefore possible that these homo F1 worms will have a higher rate of secondary mutations than in nucleii where the RNP does not persist.
On another note, in general when doing CRISPR with such laser precision nowadays, how important is it to outcross the acquired alleles? Is there any evidence that secondary mutations
are a serious problem in C. elegans experiments? Or is it that the habit of outcrossing has been carried on from the classic mutagenesis and x-ray integration experiments?
-Ed
To my knowledge, no one has yet determined where (physically) in the germline CRISPR HDR events arise. Given the syncytial nature of the gonad, they may arise in several places. If an edit occurs in the mitotic region of the gonad, it’s possible that a single editing event could give rise to multiple gametes, and hence multiple embryos. For this reason we separate injected worms shortly after recovery from injection, and make it standard practice to only keep one line derived from each injected animal.
RE outcrossing, of course it depends on who you ask. Whole-genome sequencing data in worms suggest that second-site mutations in CRIPSR strains are generally not due to CRISPR off-target effects; they are spontaneous mutations that were probably present in the background before injection and/or arose during strain isolation. Of course if you have a mutation in the working stock you started with, then backcrossing to that same strain after injection will not help you. If you have a spontaneous mutation during the CRISPR, it’s likely only going to be present in one line. For these reasons my own opinion is that your work is best rewarded by thawing new worms every 6 months or so, and isolating multiple independent lines. I think that if you do these things, the benefits of outcrossing are minimal. But I realize some people (particularly those trained in classical genetics) would consider this a heretical viewpoint. If you do decide to outcross, you may as well make it worth your while and use a different strain than the one you injected to generate the mutations.
Thank you Dan et al. These are some real gems of worm biology.
It would be interesting to do an experiment to test what proportion of an injected P0’s
offspring are independent alleles. We have done a couple experiments that start to
give some insight (but not with that intent). Twice, for different genes, we have injected an
oligo repair that is degenerate at a single codon. Ie catatgaNNNgctgatg. In both cases, we screened
for a known phenotype and then sequenced to see what amino acid it was. This was done by injecting 10-20
worms and leaving them together on one plate. But I don’t remember getting any codons twice.
It would be interesting to repeat this, but single each injected P0 and see whether each P0
favors any particular codon in its offspring.
I could imagine designing a dpy-10 rol repair in which 2 wobble bases near the edit are changed to
N, but not changing the protein sequence. You could pick 50 rollers from a P0 and see whether the 16
possible combination fall randomly, or if one or two are favored in a given P0.
-EdBack to your question in the beginning, if you sequence the PCR product (that was homozygous tested by restriction enzyme digestion), doesn’t it tell you if the two alleles are identical?
Yes, and I know them to be genetically identical at least inside the PCR product.
Two identical alleles derived twice are classically named as separate alleles.
This is because they may be in different backgrounds, or linked to different other
mutations you don’t know about.
For example, if I used CRISPR to make an unc-119 mutant identical to ed3, I would still assign
it a new allele designation from the lab I am in. That’s because the chromosome around the mutation
is different potentially in my strain than the original ed3 strain. They could possibly behave differently.
The issue is not determining whether they are the same molecularly, rather the nomenclature convention
for this specific event.
-Ed
Great thread with a lot of important points! It’s an interesting question, regarding nomenclature of a homozygote. It’s kind of hard to answer without understanding the nature of the F1 homozygote, as discussed. It might be worth bumping up to the nomenclature folks (Tim Schedl) and see what they think. Playing Devil’s advocate, if you gave the alleles two designations wouldn’t you need to outcross anyways? And they wouldn’t be independent, so you’d likely only proceed with one of the alleles. Otherwise the population would be a mixture of homozygs and hets for the two alleles.
I suppose an alternative, if following Dan’s convention of working with 4 independent alleles is that if you had three alleles recovered as a F1 hets, and one as an F1 homozygote that you could proceed without outcrossing, but explicitly note the allele that was recovered as an F1 homozygote. I got a 3xFLAG knockin as an F1 homozygote, 24 genotyped progeny were F1 homozygotes (so felt that it wasn’t some mega deletion), and at the time gave it a single allele designation, though outcrossing removed the issues associated with the homozygote actually representing two separate alleles.
Also want to second Hillel’s point about large deletions creating false positives for homozygotes. This was a major concern in the recent human embryo CRISPR paper claiming that the paternal allele was being repaired exclusively through the maternal allele (no control for large deletions). And a recent paper about CRISPR in Strongyloides parasitic nematodes reported large deletions that made it impossible to recover genotyping PCR products flanking the edit site. It’s an important consideration to keep in mind.