Thread by @JohnDoench

New @biorxivpreprint from the group, setting our sites on Cas12a (the #CRISPR enzyme formerly known as Cpf1) as a simpler system for multiplexed genetic screens. Thread...

Optimization of AsCas12a for combinatorial genetic screens in human cells biorxiv.org/content/10.110…

@biorxivpreprint Cas12a has a major benefit, only needing one promoter for expressing multiple guides, and has a smaller constant region (the tracrRNA-like ‘handle’ for Cas12a, the DR, is only 20nts). So, a second guide is only 43 more nts -- much easier to write (synthesize) and read (sequence)

@biorxivpreprint Anyway, Cas12a has been around for a little while now, but not much uptake for screens. Part of that is likely because SpCas9 already has a lot of infrastructure around it, but probably also because there’s been less work on Cas12a to really bring it up to speed.

@biorxivpreprint First off, protein expression: the use of 2x NLS sequences, combined with the enhanced Cas12a version recently described by @BKleinstiver and @JKeithJoung, really makes a difference. Here are AUCs for a screen with thousands of guides, and can see enCas12a matching Cas9

@biorxivpreprint @BKleinstiver @JKeithJoung Next up, on-target rules. Short version: use Seq-deepCpf1, from
Henry Kim’s group. Validated well on data we generated, and models we trained / tested line up well with their predictions. We extend its use to enCas12a. Over 90% of high-scoring guides are active with enCas12a!

@biorxivpreprint @BKleinstiver @JKeithJoung Off-target rules are next, naturally. We generated a library with tens of thousands of mismatches to come up with a cutting-frequency determination (CFD) matrix for ‘regular’ Cas12a and enCas12a. Since there’s no free lunch, enCas12 shows more off-target activity (darker boxes).

@biorxivpreprint @BKleinstiver @JKeithJoung Hmm, seems bad for enCas12a, right? Fear not! First, we show that the values in the CFD matrix can be combined to be predictive of activity at double-mismatch sites, AND that the predictions for most double-mismatch sites suggest they are unlikely to be cut.

@biorxivpreprint @BKleinstiver @JKeithJoung Next up, optimizing the DR sequence. Repetitive elements in lentiviral vectors is a Bad Idea, so we wanted to come up with variant DRs that maintain activity. We screened tens of thousands to find some that did, using synthetic lethality between BCL2L1 and MCL1 as our assay

@biorxivpreprint @BKleinstiver @JKeithJoung 38 were as good / better than wildtype. What wonderful specificity of the nucleotides tolerated at each position! (this is my favorite figure) So now we should be able to multiplex in lentiviral vectors with variant DRs and not worry too much about recombination

@biorxivpreprint @BKleinstiver @JKeithJoung We do a synthetic lethal screen for some past favorites and other combos in the literature. In its own right, and compared to our previous Big Papi approach, we see Cas12a perform quite well. The ease of reading / writing the guides really recommends it for this type of work.

@biorxivpreprint @BKleinstiver @JKeithJoung Of course, if 20,000^2 isn’t a large enough space for exploration, how about triple knockout? Yup, that works too! Here we target 3 cell surface genes so we can analyze by flow, and see ~70% of the cells with all three knocked out. FYI, 20,000^3 = 8x10^12

@biorxivpreprint @BKleinstiver @JKeithJoung So to sum up, by building off others’ work and adding in some of our own, we show that Cas12a is really a first-class enzyme for genetic screens, and I’d argue that for anything combinatoric, the tool of choice.

@biorxivpreprint @BKleinstiver @JKeithJoung We are very interested in comments / questions, so ask away. Also, and this is important, if we missed a citation, please please let us know (email or DM, doesn’t need to be public “complaining” although it is def. not complaining). Certainly an oversight and best to fix it now

@biorxivpreprint @BKleinstiver @JKeithJoung Thanks to the great GPP team @broadinstitute, including @p_deweirdt @HegdeMudra @ruth_e_hanna. Thanks also to collaborators at Tango, who were great partners in this effort, and very importantly, fully supported sharing all this work.

END TRANSMISSION