Genotyping problems after CRISPRing

Does anyone else have problems with genotyping worms after (successful) CRISPRing?

We get deletions, but then during outcrossing we either lose them (wt band), suddenly get a longer PCR fragment (insertion into deletion…?!) or cannot see ANY band (other loci can be amplified, the CRISPRed one not anymore…)…

The initial deletion was not simply somatic, as the progeny had it too. Is there any modification to the DNA after CRISPR? Is there further cutting by secondarily amplified sgRNAs? Lots of fancy ideas - but does anyone know about this?

Thanks in advance!

This sounds odd, I haven’t seen too much like it. Just to make sure that I understand: you’re generating indels to inactivate a gene through NHEJ-mediated repair. You’re singling F1s, allowing them to lay, and then genotyping the F1. You get a deletion (how scored? PCR? mismatch nuclease?) that you see in the F2, and then all of a sudden you get WT or insertion bands in the F3? If screening F1 are you direct screening or using a co-conversion approach? For gene inactivation, I like to either drop in premature stop codons with a diagnostic restriction site into 3’GG target sites or to replace the entire gene with a selectable marker (Dan Dickinson’s SEC or John Calarco’s recent plasmids seem best).

Yes, that is exactly our setup.

We are using the Co-Conversion protocol (Arribere et al. 2014). Since we often have problems with sgRNA efficiency (of the desired locus), we first PCR for deletions by NHEJ in dpy worms. If there is nothing, I usually don’t test for HR events (anymore).

Actually, we aim for HR events using oligos (>= 100 nt, PAGE purified) and occasionally obtain deletions. However, alleles generated by HR seem stable whereas deletion alleles seem to be lost/ prevent PCR after some generations… In my case, I can’t detect any band using a number of primer sets while other loci can be easily genotyped.
A number of people in our lab have such problems with deletion alleles, so we don’t think this simply happens by coincidence. In some cases, we sequence weird rearrangements of the CRISPRed locus with inversions and repeats, which could explain why there is no band after PCR anymore (if primer binding sites have been inverted as well).

Do you also use the co-conversion protocol? What kind of repair templates do you use and do you use the published concentrations in the injection mix?

Gotcha. So can you ever homozygose these weird deletions in the F2, or do you just get hets? If that’s the case, I could see some kind of weird polq-1 mediated repair if the Cas9-sgRNAs persist and modify the homologous chromosome. Is this just one locus at which you’re seeing this trend? Also, are you identifying indels by changes in migration in agarose gels, PAGE gels, or mismatch nucleases (getting a sense as to what kind of mutations you can detect).

I use pha-1 co-conversion when I don’t need a specific genetic background, and dpy-10 when I need a specific background. For gene inactivations, I like to find 3’GG sgRNA targets (Farboud and Meyer, 2015) as they’re consistently the most active. I then design a repair oligo to replace the NGG with a stop codon and add a BamHI site (or some other restriction site not in the genotyped sequence) and add or delete a base to ensure a frameshift. I use the recommended concentrations from the Arribere paper and my paper, though lately I’ve been playing with reducing the dpy-10 or pha-1 sgRNA concentration. Paix et al had increased knockin efficiency with a 2.5:1 target:dpy-10 sgRNA ratio (both plasmid and RNPs). The idea is that by making the selectable knock-in more difficult to obtain, that one selects for a population of animals with really active Cas9.