You're right about all that.  It should not be happening.  The thing I would add is that the PCR test itself might be at fault.  If your cell is dividing, the extrachromosomal fragment should be lost over time.  But maybe if your PCR randomly turned out to be ridiculously sensitive, you could pick it up for a while. 

The other thing might be that longer fragments amplify less efficiently than shorter fragments.  There can be special cases where this effect is extreme, for example if the deleted region contains some kind of secondary structure.  So you might be looking at heterozygotes, but with your particular PCR situation, the KO band amplifies so much more efficiently that the WT band never shows up.  (I assume you verified that the PCR does give the WT band when used on a homozygous WT, yes?) 

Do you have reason to expect homozygotes?  If you're working with a pooled transfection reaction, it's certainly impossible to eliminate the WT band from that pool altogether.  No knockout is 100% efficient, so if you're seeing only knockout with the cell pool, then you already know the PCR result is false.  You can have a true WT-negative result only if you picked clones and grew them up separately, and even then, the majority of clones should be KO/WT, with only a few % lucky KO/KO in there. 

Thank you John. 

I do verify the knockout bands against the homozygous WT DNA. When I ran the pooled DNA, I could clearly see two bands, one similar to the WT, and one shorter deleted band, showing that the pool contains a mixture of deleted and intact alleles.

When I run the DNA of the single cell clones, I see a few of them showing both bands, pertaining to a mono-allelic knockout, and a few that contain only the shorter deleted band, but no long WT-like band (these few are what I assume to be bi-allelic KOs). But, for all of these single clone DNAs, I see amplification of the deleted region. 

Your observation that the PCR might actually be picking up the KO bands very sensitively is very helpful, but if certain clones show both bands, then I would think that the PCR is able to pick up both? 

Another thing I would like to add is that this DNA is extracted using a quikc and dirty method from cells growing in 96-well plates. The single cells growing there have probably gone through about 5 or 6 division cycles. I checked in silico for similar regions in the genome to find any signs on plausible recombination of the deleted fragment into another region of the genome, but could not find significant similarities.

Any insight on how long extra chromosomal DNA fragments hang around inside the cell?

OK makes sense.  Considering all that, I think it should be contamination from environmental WT DNA.  Have you ever seen your deleted-region PCR being negative?  For example, if you run some control DNA from another species or some other DNA that never had the deleted fragment, but that was processed in the same environment all your samples are processed in?  If a PCR is sensitive enough, it can be very difficult to get the negative controls to be negative sometimes. 

If you can rule all that out, you might be on to something interesting.  Maybe a probe / immunofluorescence method could be used to visualize where the fragment is inside the cells.  And then you could see what happens to it.  Is it really passed on to daughters?  Perhaps you found a new mammalian episomal control system or "origin of replication", and that might be very useful and valuable in other fields, like recombinant protein expression :)

I realize this is an old thread, but I wanted to offer some insight for anyone else with the same question.

Just because you use two gRNAs does not automatically guarantee that you will get your 500bp deletion. There are many other scenarios that may occur following introduction of your gRNAs into the cells.

First, you may get cleavage and NHEJ at one of the gRNA target sites before there is cleavage at the other gRNA site. This will result in indels at both sites, but no deletion and the WT band you saw. This may still yield a non-functional gene product and a knockout clone, though genetically it will be different that a bi-allelic deletion.

Second, you may get simultaneous cleavage and and inversion of the 500bp sequence inbetween your gRNA target sites. This will give you the WT band but will also result in a non-functional transcript.

Third, you may only get cleavage at one locus and not the other which would give rise to an indel and potentially a non-functional gene. This would also give you the WT band but would appear as a knockout on a western blot.

Hi Timothy,

Following up on your post, the second scenario of simultaneous cleavage and inversion that you suggest seems quite interesting because I have been struggling in a situation where I can see gDNA band is intact size but the protein expression is gone. So I have been wondering how to completely validate my KO clones? When I do phenotypic assays to validate the clone the protein KO works perfectly well.

Seemana,

Any of the three scenarios I mentioned will result in a wild-type band size. You can test for inversion with PCR by using a primer which is contained within the deletion region and one which flanks the deleted region. This will not indicate if you have either the first scenario or the third scenario. I did develop a PCR method using a primer which lays across one of the sgRNA target sites to detect indels. It will be published soon and I will attach a link when it is. However, the only way to definitively confirm indel formation without deletion is to sequence that region.

Thanks for the suggestions Timothy

Here is my recent paper which outlines what I have been talking about:

Article CRISPR deletion of MIEN1 in breast cancer cells