A large-scale resource for tissue-specific CRISPR mutagenesis in Drosophila

Author's response

Fillip Port and Michael Boutros shared

It is interesting that very low expression of Cas9 was sufficient to produce robust knockdown of target genes (figure 2 in preprint). Presumably this is because only one or two successful targeting events are necessary to produce an indel. Is it possible that reducing the expression of Cas9 might also reduce off-target cleavage? (I.e. There may be less probability of gRNA/Cas9 complex recognising a low-affinity off-target binding site).

We observe robust, although no longer optimal, editing with constructs that express Cas9 at levels below the detection limit of immunohistochemistry. This highlights the very low amounts of Cas9 that are necessary to mediate mutations in the genome. And while the frequency of DNA targeting in these conditions might be low, the resulting mutations are permanent and accumulate over time, which can lead to high levels of modified target sites at later stages of development.

In situations where Cas9 is limiting, it is expected that suboptimal target sites, such as off-targets, are no longer efficiently mutated. In line with this assumption it has been shown in other systems that the occurrence of off-target mutations is inversely correlated with the expression levels and duration of CRISPR components in the cell. It is therefore best practice to supply Cas9 not beyond of what is necessary and our new Cas9 tools help to do that independent of the promoter and enhancers used to drive Cas9 expression.

Do any of the protein products of the EGFP uORFs employed in your Cas9 constructs retain any fluorescence or could they otherwise be detectable by antibody recognition for use as visible markers for Cas9 expressing cells?

Only the full-length EGFP uORF that is used in UAS-u^XXLCas9 produces a fluorescent protein. It is possible that the other polypeptides can be recognized with anti-GFP antibodies, but for such applications there are also good antibodies available to detect Cas9 protein directly.

Do you think that it is likely that any given sgRNA line will work equally well in all cell types, or do you think that developmentally regulated changes in the local chromatin environment could change the efficiency of gRNA recognition/Cas9 cleavage?

This is a very interesting question that so far has not been systematically analysed. It is generally known that chromatin can affect the efficiency of gene targeting by CRISPR and since it is known that chromatin is dynamic, it would indeed be expected that the efficiency of any given sgRNA is somewhat context dependent. Systematic studies will be necessary to explore to what extent that is the case. Our resource employs two sgRNAs targeting independent positions in each gene. This should help to somewhat alleviate such effects, as one sgRNA might be active in a context that the other one is not. However, we would expect that differences between tissues exist, which will also be dependent on the strength of the specific GAL4 line.

Given that mosaic tissues are produced using this approach, do you think that false negatives may be prevalent (more so than with alternative approaches) when targeting genes that produce phenotypes with low penetrance? Particularly if only a small population of cells is being assayed (e.g. a subset of neurons).

The fact that somatic CRISPR often induces genetic mosaics of unknown composition is one of the unique features of this system. Using multiple sgRNAs targeting the same genes enriches mosaics for bi-alleleic knockout cells, but with two sgRNAs commonly used in large-scale resources cells with one or two functional alleles are still relatively common. It will be very interesting to see what effect this will have in systematic genetic screens. Our data and data from other labs suggest that somatic CRISPR phenotypes are very robust, but it will be important to explore this in many different contexts. In situations where only few cells are targeted by CRISPR one would expect to observe a higher variability of phenotypes (e.g. animals with complete knock-out phenotypes and ones with no phenotype in the same experiment). So compared to RNAi screening, which gives very homogenous phenotypes, the fact that CRISPR induces mosaics might mean that variability becomes an expected (and potentially usable) feature of a screen.

The Vienna and TRiP RNAi collections have been well used by fly researchers. Do you think that with the advent of resources such as yours these collections will become redundant, or do you think that there will always be a space for these as complementary resources? Can you speculate on whether there would be occasions where it may still be more appropriate to use RNAi rather than CRISPR for a large-scale screen or other genetic experiment?

RNAi and CRISPR work by fundamentally different mechanisms and therefore it is likely that they will produce different outcomes – at least in some cases. RNAi is limited by the fact that the majority of RNAi lines retain substantial residual gene activity. While that means that phenotypes can be weak or not present at all, in some cases hypomorphic phenotypes are more desirable than complete abrogation of gene function, for example for genes that are essential for cell survival.

Mutagenesis mediated by conditional CRISPR can generate genetic null alleles and in some cases will give rise to phenotypes that would have been missed with RNAi. But somatic CRISPR can have it’s own issues, for example phenotypes can only arise once mRNA and protein that are already present in the cell have been sufficiently reduced, which could be problematic for genes that encode very stable mRNAs. Therefore we think it is very likely that in the future RNAi and CRISPR will be used in parallel as complementary strategies to investigate gene function. What will be used as the primary reagent of choice will likely be context dependent.

When/where will the stocks be available for the community, and do you have any plans to expand the collection?

All of the tools we generate will be made freely available to the community. We are in the process of finalizing the details for distribution of our fly lines with one of the public stock centers and hope these will become available during the summer. Our Cas9 expression plasmids will soon be available from the non-profit repository Addgene. We are also working on expanding this resource, so stay tuned.