Discovery of widespread type I and type V CRISPR-Cas inhibitors

See allHide authors and affiliations

Science  12 Oct 2018:
Vol. 362, Issue 6411, pp. 240-242
DOI: 10.1126/science.aau5174

Cas12 inhibitors join the anti-CRISPR family

Bacteria and their phages continually coevolve in a molecular arms race. For example, phages use anti-CRISPR proteins to inhibit the bacterial type I and II CRISPR systems (see the Perspective by Koonin and Makarova). Watters et al. and Marino et al. used bioinformatic and experimental approaches to identify inhibitors of type V CRISPR-Cas12a. Cas12a has been successfully engineered for gene editing and nucleic acid detection. Some of the anti-Cas12a proteins identified in these studies had broad-spectrum inhibitory effects on Cas12a orthologs and could block Cas12a-mediated genome editing in human cells.

Science, this issue p. 236, p. 240; see also p. 156


Bacterial CRISPR-Cas systems protect their host from bacteriophages and other mobile genetic elements. Mobile elements, in turn, encode various anti-CRISPR (Acr) proteins to inhibit the immune function of CRISPR-Cas. To date, Acr proteins have been discovered for type I (subtypes I-D, I-E, and I-F) and type II (II-A and II-C) but not other CRISPR systems. Here, we report the discovery of 12 acr genes, including inhibitors of type V-A and I-C CRISPR systems. AcrVA1 inhibits a broad spectrum of Cas12a (Cpf1) orthologs—including MbCas12a, Mb3Cas12a, AsCas12a, and LbCas12a—when assayed in human cells. The acr genes reported here provide useful biotechnological tools and mark the discovery of acr loci in many bacteria and phages.

The discovery of bacterial CRISPR-Cas systems that prevent infection by bacterial viruses (phages) has opened a paradigm for bacterial immunity while yielding exciting tools for targeted genome editing. CRISPR systems destroy phage genomes, and in turn, phages express anti-CRISPR (Acr) proteins that directly inhibit Cas effectors (1, 2). Six distinct types (I to VI) of CRISPR systems are spread widely across the bacterial world (3), but Acr proteins have only been discovered for type I and II CRISPR systems (1, 36). Given the prevalence and diversity of CRISPR systems, we predict that Acr proteins against other types await discovery.

Acr proteins do not have conserved sequences or structures and only share their relatively small size, making de novo prediction of acr function challenging (6). However, acr genes often cluster together with other acr genes or are adjacent to highly conserved Acr-associated genes (aca genes) in “acr loci” (7, 8). In this work, we sought to identify acr genes in bacteria and phages that are not homologous to previously identified acr or aca genes.

Acr proteins were first discovered in Pseudomonas aeruginosa, inhibiting type I-F and I-E CRISPR systems (1, 9). P. aeruginosa strains also encode a third CRISPR subtype (type I-C), which lacks known inhibitors (10). We engineered P. aeruginosa to target phage JBD30 with type I-C CRISPR-Cas (fig. S1A) and used it in parallel with existing type I-E (strain SMC4386) and I-F (strain PA14) CRISPR strains to screen for additional acr candidates.

Homologs of aca1 were searched for in Pseudomonas genomes, and seven gene families not previously tested for Acr function were identified upstream of aca1 (Fig. 1A). Three genes inhibited the type I-E CRISPR-Cas system (acrIE5-7), one inhibited type I-F (acrIF11), restoring the plaquing of a targeted phage, and two genes had no inhibitory activity (orf1 and orf2) (Fig. 1B, fig. S1B, and tables S1 and S2). Another gene exhibited dual I-E and I-F inhibition, and domain analysis revealed a chimera of previously identified acrIE4 and acrIF7 (acrIE4-F7). No type I-C inhibitors were identified. The type I-F inhibitor acrIF11 was commonly represented in both the P. aeruginosa mobilome and in more than 50 species of diverse Proteobacteria (fig. S2 and table S2). acrIF11 is often associated with genes that encode DNA-binding motifs, which we have designated aca4 to aca7 (fig. S2 and tables S1, S3, and S4). To confirm that these aca genes can be used to facilitate acr discovery, we used aca4 to discover an additional Pseudomonas Acr, acrIF12 (Fig. 1, A and B).

Fig. 1 The discovery of a widespread type I inhibitor.

(A) Schematic of type I-E and type I-F Acrs with Acr-associated (aca1 and aca4) genes in Pseudomonas sp. mobile genetic elements, with dotted lines indicating the “guilt-by-association” relationships used to discover new acr genes in Pseudomonas sp. and Moraxella sp. from known acr genes (top two rows). (B) Phage plaque assays to assess CRISPR-Cas inhibition. Tenfold serial dilutions of a type I-E or type I-F CRISPR-targeted phage (JBD8 or DMS3m, respectively) titered on lawns of P. aeruginosa with naturally active type I-E or type I-F CRISPR-Cas systems expressing candidate inhibitors. ∆CRISPR strains measure phage replication in the absence of CRISPR immunity (top row).

Given the widespread nature of acrIF11, we next used it to discover Acr proteins against CRISPR systems in which they have not yet been found: type I-C, a minimal class 1 system, and type V-A CRISPR-Cas12a (Cpf1), a class 2 single effector system that has high efficiency in genome editing (1113). To find AcrIC and AcrVA proteins, we first searched for genomes that encode CRISPR spacers that match a target protospacer elsewhere in the same genome (Fig. 2A). The tolerance of this “self-targeting” in viable bacteria indicates potential inhibition of the CRISPR system (4) because genome cleavage would result in bacterial death.

Fig. 2 Type I-C and type V-A Acr proteins identified in Moraxella.

(A) Schematic of M. bovoculi intragenomic self-targeting, in which spacers encoded by CRISPR-Cas12a system and their target protospacers exist within the same genome. (B) Schematic showing type V-A (acrVA1-VA3), type I-C (acrIC1), and type I-F (acrIF11-IF14) inhibitors in Moraxella. orf1 and orf2 are genes of unknown function. Vertical arrows indicate the percent protein sequence identity. (C to E) Phage plaque assays with 10-fold serial dilutions of the indicated phage to assess inhibition of CRISPR-Cas (C) type I-C, (D) type I-F, and (E) type V-A. Bacterial clearance (black) indicates phage replication. (C) “Uninduced” and [(D) and (E)] “no crRNA” indicate full phage titer.

The Gram-negative bovine pathogen Moraxella bovoculi (14, 15) is a Cas12a-containing organism (11) in which four of the seven genomes feature type V-A self-targeting (table S5), and one strain (58069) also features self-targeting by type I-C (table S6). Although no previously described acr or aca genes were present in this strain, an acrIF11 homolog was found in phages that infect the human pathogen Moraxella catarrhalis (16), a close relative of M. bovoculi. Genes adjacent to acrIF11 in M. catarrhalis had homologs in the self-targeting M. bovoculi strains (Fig. 2B), and together, these genes were selected as candidate acr genes. Each gene was first tested against the type I-C and I-F systems introduced above because both subtypes are found in Moraxella. Gene AAX09_07415 (now acrIC1) inhibited the type I-C system, explaining the tolerance of self-targeting in strain 58069 (Fig. 2C). Additionally, gene E9U_08473 (acrIF13) from the M. catarrhalis BC8 prophage completely inhibited I-F function, as did AKI27193.1 (acrIF14), which is found in phage Mcat5 at the same genomic position as that of acrIF11 in BC8 (Fig. 2, B and D). These Acr proteins possess broad spectrum activity; the type I-C and I-F systems in Moraxella and Pseudomonas only share an average pairwise identity of 30 and 36%, respectively (fig. S3)

Because of the limited tools available for the genetic manipulation of Moraxella sp., the remaining genes were tested for type V-A Acr function in P. aeruginosa PAO1 engineered to express MbCas12a and a CRISPR-RNA (crRNA) that targets P. aeruginosa phage JBD30. Two distinct crRNAs were used, showing strong reduction of titer by more than four orders of magnitude (Fig. 2E). The first gene in the M. bovoculi 58069 acr locus, AAX09_07405 (acrVA1), restored phage titers nearly to levels seen with the crRNA-minus control, indicating that it robustly inhibits Cas12a. This is in good agreement with the independent discovery of AcrVA1 reported in a companion paper (17). The adjacent gene, acrVA2, also inhibited targeting, as did its ortholog (acrVA2.1) (Fig. 2E). An additional gene from this locus, acrVA3, possessed subtle anti-Cas12a activity but was toxic to cells and adversely affected JBD30 phage growth independently of Cas12a (fig. S5). We therefore tested an ortholog with 43% sequence identity, B0181_04965 (acrVA3.1), which showed stronger Cas12a inhibition with no toxicity or adverse effects on phage growth (Fig. 2E and fig. S5). Surprisingly, acrVA3.1 also showed partial restoration of phage titer during type I-C targeting, suggesting that it may inhibit the type I-C as well as type V-A system (Fig. 2, C and E). Although these two CRISPR subtypes do not share any protein components, a dual-specificity inhibitor may use distinct protein-interaction interfaces or modulate an undiscovered host process required for CRISPR immunity. We used the Acr “key” acrIF11 to unlock acr loci that encode seven distinct acr genes inhibiting type I-C, I-F, and V-A CRISPR. Below, we focus on the evolutionary analysis of type V-A inhibitors and their function in mammalian cells.

The gene acrVA1 encodes a 170–amino acid protein found only in Moraxella sp. and Eubacterium eligens (fig. S6), both type V-A CRISPR-containing organisms. By contrast, acrVA2 (322 amino acids) and acrVA3 (168 amino acids) orthologs are found broadly distributed throughout multiple classes of bacteria. For example, acrVA2 orthologs are present in Lachnospiraceae and Leptospira (fig. S7), which contain type V-A CRISPR, as well as in Moraxella, Leptospira, and Lactobacillus phages. Distant orthologs of acrVA2 were also identified on plasmids and conjugative elements in Escherichia coli (fig. S7), although the importance of a bacterium that lacks type V-A CRISPR encoding a putative acrVA gene is unknown. Orthologs of acrVA3 were identified in many Proteobacteria and in Eubacterium and Clostridium species, which encode type V-A CRISPR (fig. S8).

Given the inhibitory effect of acrVA1-3.1 on MbCas12a in bacteria, we sought to determine whether AcrVA proteins could block MbCas12a activity in human cells. Human U2-OS–enhanced green fluorescent protein (EGFP) cells (18) transiently expressing MbCas12a, EGFP-targeting crRNA, and human codon-optimized acrVA1-3.1 were assessed for EGFP disruption by using flow cytometry. Coexpression of MbCas12a and crRNA resulted in ~60 to 70% disruption of EGFP expression relative to background (Fig. 3A). AcrVA1 expression reduced EGFP disruption to background levels, indicating inhibition of MbCas12a, whereas the other acrVA genes showed little evidence of activity here (Fig. 3A). We additionally found that acrVA1 inhibited another Cas12a ortholog (Mb3Cas12a) while having no impact on SpyCas9 editing in the same assay (Fig. 3B). Titration of the Acr plasmid relative to the Cas12a expression plasmid revealed comparable dose-dependent responses to inhibition between MbCas12a or Mb3Cas12a with AcrVA1 and SpyCas9 with AcrIIA4 (fig. S9). Furthermore, AcrVA1 was found to be a broad-spectrum inhibitor of other commonly used Cas12a orthologs (11), providing strong inhibition of AsCas12a and LbCas12a and modest inhibition of FnCas12a (Fig. 3C).

Fig. 3 AcrVA1 blocks Cas12a-mediated gene editing in human cells.

(A to C) Human cell U2-OS-EGFP disruption experiments to assess AcrVA-mediated inhibition of Cas12a activities. (A) Inhibition of MbCas12a activity with various AcrVA constructs; the “no filler” condition contained only plasmids for Cas12a and crRNA expression. (B) Comparisons between the inhibitory activities of AcrVA1 and AcrIIA4 against MbCas12a, Mb3Cas12a, and SpyCas9. Controls that use “filler” plasmid in lieu of Acr plasmids were included in order to equalize amounts of DNA. (C) Assessment of AcrVA1 activity against Cas12a orthologs, with AcrIIA4 used as control. Background EGFP disruption is indicated by the red dashed line; error bars indicate SEM for n = 3 independent biological replicates. (D) Inhibition of activity of Cas12a orthologs against endogenous sites in human cells (RUNX1, DNMT1, or FANCF genes). Gene modification assessed by means of T7E1 assay 72 hours after transfection; error bars indicate SEM for n = 3 independent biological replicates.

Last, to determine whether AcrVA1 could inhibit Cas12a-mediated modification of endogenous loci in human cells, U2-OS cells were cotransfected with plasmids expressing acrVA1, Cas12a, and crRNAs that target endogenous genes (RUNX1, DNMT1, or FANCF) and assessed for gene disruption by means of T7 endonuclease I (T7E1) assay. We found that AcrVA1 completely inhibited gene disruption by MbCas12a and Mb3Cas12a, with modest to strong inhibition of As-, Lb-, and FnCas12a orthologs (Fig. 3D and fig. S10).

Here, we report the discovery of a broadly distributed type I-F Acr protein (AcrIF11) that served as a marker for acr loci and led to the identification of type I-C and V-A CRISPR inhibitors. One of these acrVA genes (acrVA1) potently inhibits Cas12a in bacteria and in human cells, providing a new tool for Cas12a regulation. Our findings show that mobile genetic elements can tolerate bacteria with more than one CRISPR-Cas type by possessing multiple Acr proteins in the same locus. The strategy described here enabled the identification of many widespread Acr proteins, which may prove useful in future Acr discovery.

Supplementary Materials

Materials and Methods

Figs. S1 to S10

Tables S1 to S11

References (1926)

References and Notes

Acknowledgments: We thank J. M. Peters and C. A. Gross (University of California, San Francisco) for providing the pTN7C130 entry vector for cloning MbCas12a. We thank D. Loy (University of Nebraska) and M. Clawson (U.S. Department of Agriculture) for providing M. bovoculi strain 58069 and 22581 for genomic DNA extraction. Funding: Work in the Bondy-Denomy laboratory was supported by the University of California, San Francisco Program for Breakthrough in Biomedical Research, funded in part by the Sandler Foundation, and a National Institutes of Health (NIH) Office of the Director Early Independence Award (DP5-OD021344). Acr discovery efforts were specifically supported by DARPA Safe Genes grant HR0011-17-2-0043. Work in the Joung laboratory was supported by the Desmond and Ann Heathwood MGH Research Scholar Award (to J.K.J.) and the NIH awards K99 CA218870 (B.P.K.) and R35 GM118158 (J.K.J.). Author contributions: N.D.M. conducted Moraxella AcrVA and AcrI discovery, characterization, and bioinformatics. J.Y.Z. and A.L.B. conducted Pseudomonas AcrI and aca discovery, characterization, and bioinformatics. L.M.L. built the Pseudomonas type I-C strain, and J.D.B. and N.D.M. built the Pseudomonas type V-A strain. B.J.R. conducted self-targeting analysis. J.B.-D conceptualized the project and supervised all bioinformatics and bacterial experiments. B.P.K., A.A.S., and R.T.W. constructed AcrVA expression plasmids and performed human cell experiments. Funding for this work was procured by J.B.-D., B.P.K., and J.K.J. The manuscript was written by N.D.M., A.L.B., and J.B.-D., with editing and contributions from all authors. Competing interests: A patent has been filed pertaining to AcrVA genes and their applications. J.K.J. has financial interests in Beam Therapeutics, Blink Therapeutics, Editas Medicine, Endcadia, Monitor Biotechnologies (formerly known as Beacon Genomics), Pairwise Plants, Poseida Therapeutics, and Transposagen Biopharmaceuticals. J.K.J.’s interests were reviewed and are managed by Massachusetts General Hospital and Partners HealthCare in accordance with their conflict-of-interest policies. Data and materials availability: All data are available in the main text or the supplementary materials. Plasmids described in this work are available through Addgene. Phages and bacterial strains will be made available upon request to joseph.bondy-denomy{at}

Stay Connected to Science


Navigate This Article