enAsCas12a enables CRISPR-directed evolution to screen for functional drug resistance mutations in sequences inaccessible to SpCas9

Repair of an SpCas9 double-strand break by non-homologous end joining yields both in-frame and out-of-frame mutations, concentrated in the nucleotides immediately upstream of the 5′-NGG PAM rather than downstream of it. Across 223 amplicon-sequenced samples covering 71 to 78 target sites in human HCT116, HEK293T and K562 cells, most variants are 1- to 30-bp deletions or single-nucleotide insertions, and on average 35.7% ± 16.5% of variants at a given target site are in-frame — the subset that produces functional protein variants rather than knockouts. This distribution supports an effective mutagenesis window of roughly 14 bp around the cut site, including the PAM.

Within a 14-bp mutable window across the coding sequences of 190 human anticancer drug target genes, SpCas9 with 5′-NGG PAMs reaches 83.0% ± 9.5% of coding sequence using 65,841 sgRNAs, whereas AsCas12a with 5′-TTTV PAMs reaches only 29.7% ± 13.6% using 10,498 crRNAs. Combining SpCas9 and AsCas12a raises coverage only marginally, because the sequence each can reach overlaps heavily with the other.

enAsCas12a, an engineered AsCas12a variant with increased activity and relaxed PAM recognition, covers 80.9% ± 9.1% of the coding sequence of 190 human drug target genes using 73,802 crRNAs designed against its top 23 compatible PAMs — comparable to SpCas9's own coverage. SpCas9 and enAsCas12a together reach 97.4% ± 1.1%, so unlike SpCas9 and wild-type AsCas12a, these two nucleases access genuinely complementary regions of the genome.

The relaxed-PAM SpCas9 variants xCas9-3.7 (5′-NG, 5′-GAT, 5′-GAA, 5′-CAA) and SpCas9-NG (5′-NG) cover close to 100% of the coding sequence of 190 human drug target genes — 99.8% ± 0.4% and 99.7% ± 0.5% respectively — but require a steep increase in library size, 317,915 and 220,102 sgRNAs against SpCas9's 65,841. Near-complete coverage is bought with library complexity, not for free.

Broader PAM compatibility does not by itself make a better mutagenesis screen: in a tiling screen of NAMPT in HAP1 cells, SpCas9-NG enriched only three KPT-9274-resistance sgRNAs where canonical SpCas9 had previously enriched six, and it failed to re-enrich any of the NGG-PAM sgRNAs previously identified even though they were present in the library. SpCas9-NG did reveal one new resistance hotspot inaccessible to SpCas9 — codons K84, Y87 and F91 in NAMPT exon 3, reached through non-canonical NGA and NGT PAMs — so it extends reach at the cost of overall activity.

In tiling screens of the NAMPT coding sequence in HAP1 cells treated with the NAMPT inhibitor FK866, enAsCas12a enriches nine crRNAs across duplicate screens against AsCas12a's two, and seven of the nine target non-canonical PAMs. All nine enriched guides confer FK866 resistance when individually transfected with enAsCas12a, whereas with AsCas12a only the two canonical TTTV-PAM guides do.

enAsCas12a substantially outperforms AsCas12a at generating drug-resistance mutations under selection: HAP1 cells expressing enAsCas12a yield 19-fold more KPT-9274-resistant colonies than AsCas12a-expressing cells from the same NAMPT crRNA tiling library, with both resistant populations showing a 350- to 500-fold increase in EC50. Twenty-two crRNAs enrich in both enAsCas12a replicate screens against six with AsCas12a, and individual validation confirms 13 of 24 guides with enAsCas12a against only 2 of 24 with AsCas12a.

enAsCas12a tolerates a 5′-TTTT PAM where AsCas12a does not: the NAMPT-targeting crRNAs NAMPT_91, NAMPT_92 and NAMPT_115, all designed against TTTT, confer drug resistance when paired with enAsCas12a but not with AsCas12a, which shows only residual activity at TTTT PAMs.

Mutagenesis scanning of NAMPT with enAsCas12a identifies resistance hotspots against both FK866 and KPT-9274 at two sites in the inhibitor binding pocket — deletions across residues E89 to D93 in exon 3 and G239 to P243 in exon 6 — plus a L387del in exon 9. The exon 3 site, which contains the D93 residue known to confer NAMPT inhibitor resistance, was missed by a previous SpCas9 tiling screen because only a single sgRNA covers its boundary.

A nuclease's PAM-determined coverage gaps translate directly into resistance sites a screen cannot find. Neither AsCas12a nor enAsCas12a nominates NAMPT exon 1 as a resistance hotspot, because AsCas12a covers none of the first 29 codons of NAMPT and enAsCas12a covers only the first 10 before a gap extending to codon 30, whereas SpCas9 covers codons 1 to 13 and 18 onward and did identify the exon 1 Y18 site, whose pyridyl stacking interaction with KPT-9274 and FK866 explains its role in resistance.

enAsCas12a supports multi-gene target deconvolution, not just single-gene scanning: a library of 3,838 crRNAs tiling 10 genes, screened in HAP1 cells against the KIF11 inhibitor ispinesib, enriches 10 crRNAs across duplicate experiments, nine of them targeting KIF11 itself. The validated guides map to exons 4, 5 and 6, which encode loop 5, loop 6 and α helix 3 of the ispinesib binding pocket, and several reach those hotspots only through non-canonical PAMs including TTCC, TCCC, TACC, ATTA, TATA and CTTC.