Unexpected allele and HDR frequencies in single cell clones

[ilobb]

Hello,
Firstly, thanks very much for providing this excellent tool for the community.

We have been using NGS and CRISPResso2 to try and screen single cell clones for both KO and epitope-tag KI projects. However, we have run into a few teething problems.

Firstly, for KO projects, we are aiming to identify full KO clones; whether they be homozygous for one particular indel, or a compound heterozygote for KO-producing indels.
We can quite easily identify unwanted clones by the presence of high percentages of unmodified reads. However, there are some instances where this is a little bit more ambiguous; say 95% modified reads to 5% unmodified reads, which shouldn't technically be possible in a diploid single cell clone where the CRISPR components were transiently delivered as a RNP.
Additionally, the allele frequency of clones that do appear to be full KOs (>99% modified reads), is unexpectedly diverse in some instances. For example with the clone below.

Again these type of frequency shouldn't really be possible (maybe in some instances where clonality cannot be ensured, but unlikely to be the case across the board).

For the KI project we performed both CRISPResso2 analysis and TIDER on the mixed population, both of which told us that the
HDR rate was ~5%. However, for some samples 100% of the clones which were multiplexed into that sample end up having high HDR rates, whereas other multiplexed samples reflect more closely what we expect given the mixed population CRISPResso2/TIDER results.

If you could offer any insight into these problems that would be most appreciated.
Cheers,
Ian

[kclem]

This allelic diversity does seem unexpected. Do you use PCR-free library prep? (i.e. what are the chances that 10 reads here resulted from a single molecule?)

Some of the diversity can be attributed to possible PCR or sequencing errors (the single base substitutions, especially the T->G subs) but the 1bp insertion is unlikely.

Have you done sequencing of early vs late clones to see if/how much the allelic diversity increases over time?

[ilobb]

The library prep was done using PCR. We did use very high fidelity polymerase (Q5), but to be perfectly honest the cycle number could likely have been reduced over the 2-step reaction we used for amplification and addition of adapters.

I think PCR error might be a good working hypothesis for us to look into.

This is also something I think we need to be wary of for the KI project too, but I'm a little unsure as to whether this PCR error could really explain the unexpectedly high calling of HDR reads for some samples? The KI that we are looking for is a FLAG tag, so it's unlikely that one or two PCR errors may produce something CRISPResso2 might misidentify as a successful HDR in this instance?

I'm not sure I articulated that problem very well earlier, so the table should illustrate it better. The "samples" were amplified in independent PCR reactions with different in-line barcodes, then pooled into a "Tube" and ran with Genewiz's AmpliconEZ service. You can see that for Tube 1, CRISPResso2 calls all of the clones to have HDR (despite the mixed population being at ~5%), whereas in Tube 2 the vast majority have negligible levels, which i think is more believable.

  |