Machine learning predicts new anti-CRISPR proteins

Abstract

The increasing use of CRISPR-Cas9 in medicine, agriculture, and synthetic biology has accelerated the drive to discover new CRISPR-Cas inhibitors as potential mechanisms of control for gene editing applications. Many anti-CRISPRs have been found that inhibit the CRISPR-Cas adaptive immune system. However, comparing all currently known anti-CRISPRs does not reveal a shared set of properties for facile bioinformatic identification of new anti-CRISPR families. Here, we describe AcRanker, a machine learning based method to aid direct identification of new potential anti-CRISPRs using only protein sequence information. Using a training set of known anti-CRISPRs, we built a model based on XGBoost ranking. We then applied AcRanker to predict candidate anti-CRISPRs from predicted prophage regions within self-targeting bacterial genomes and discovered two previously unknown anti-CRISPRs: AcrllA20 (ML1) and AcrIIA21 (ML8). We show that AcrIIA20 strongly inhibits Streptococcus iniae Cas9 (SinCas9) and weakly inhibits Streptococcus pyogenes Cas9 (SpyCas9). We also show that AcrIIA21 inhibits SpyCas9, Streptococcus aureus Cas9 (SauCas9) and SinCas9 with low potency. The addition of AcRanker to the anti-CRISPR discovery toolkit allows researchers to directly rank potential anti-CRISPR candidate genes for increased speed in testing and validation of new anti-CRISPRs. A web server implementation for AcRanker is available online at http://acranker.pythonanywhere.com/.

Figure 1.

Acr candidates selected for biochemical testing. Ten Acr candidates were selected from manual inspection for further biochemical testing (blue fill). Each candidate is shown in its genomic context with its assigned rank from AcRanker noted in red. Homologous proteins share the same color border (green, blue). Homologs of AcrIIA3 (orange border) and AcrIIA1 (red border) are indicated. While testing the ML candidates, ML3 (yellow fill) was identified as a specific inhibitor of LmoCas9 (25).

Figure 2.

Inhibition of SpyCas9 and SauCas9 by newly discovered Acr candidates. (A) In vitro cleavage of dsDNA by SpyCas9 in the absence or presence of a 50-fold excess of AcrIIA4 (positive control) and each Acr candidate. (B) In vitro cleavage of dsDNA by SpyCas9 in the presence of increasing concentrations of (left to right) BSA (negative control), AcrIIA4 (positive control), ML1 and ML8 (Acr:RNP 0.1-, 1-, 2-,10-, 50- and 100-fold excess from left to right). (C) In vitro cleavage of dsDNA by SauCas9 in the absence or presence of a 25-fold excess of each Acr candidate. (D) In vitro cleavage of dsDNA by SauCas9 in the presence of increasing concentrations of (left to right) BSA (negative control), AcrllA5 (positive control, Acr:RNP 0.1-, 1-, 2-, 4-, 8- and 10-fold excess from left to right), ML3 and ML8 (Acr:RNP 0.1-, 1-, 2-,10-, 50- and 100-fold excess from left to right). Uncropped gel images for panels B and D are shown in Supplementary Figures S3 and S4.

Figure 3.

ML1 and ML8 inhibit SinCas9 with ML1 showing high potency. (A) In vitro cleavage of dsDNA by SinCas9 in the absence or presence of a 50-fold excess of each Acr candidate. (B) In vitro cleavage of dsDNA by SinCas9 in the presence of increasing concentrations of ML1. The uncropped gel image for panel B is shown in Supplementary Figure S6.

Figure 4.

ML1 competes with AcrIIA2 to bind to the SinCas9–sgRNA complex. (A) Flowchart for the competition binding experiment between ML1 and AcrIIA2. Binding of the Acr to the SinCas9–sgRNA RNP was reconstituted using size-exclusion chromatography (SEC). (B) Size-exclusion chromatogram of SinCas9-sgRNA in the presence of either ML1, AcrIIA2 or both Acrs with AcrIIA2 added after ML1. (C) Coomassie-stained polyacrylamide gel illustrating the components of the SinCas9-RNP fraction annotated (I), (II) and (III) in panel B.