Anti-CRISPRs on the march

See allHide authors and affiliations

Science  12 Oct 2018:
Vol. 362, Issue 6411, pp. 156-157
DOI: 10.1126/science.aav2440

Viruses and other mobile genetic elements (MGEs) are ubiquitous to all cellular life-forms, and thus, evolution of life is a perennial host-parasite arms race (1). Nearly all cellular organisms have evolved diverse, multilayer defense systems against parasites (2). Arguably, adaptive immunity is the most elaborate, efficient, and specific form of such defense. In almost all archaea and about one-third of bacteria, adaptive immunity is mediated by CRISPR-Cas systems that incorporate fragments of viral genomes into CRISPR arrays and use processed transcripts of these inserted fragments (spacers) to recognize and destroy the cognate viruses (3). The efficiency of CRISPR-Cas is such that when a virus is recognized by the immune system, it has no chance to survive. And yet, viruses thrive in all microbial communities. Virus genomes mutate fast, so the most straightforward way to avoid adaptive immunity is mutational escape. However, the hosts keep adapting, so in addition to simply attempting to outpace immunity, viruses evolve active antidefense—in particular, multiple anti-CRISPR proteins (Acrs) (see the figure) (4, 5). On pages 240 and 236 of this issue, Marino et al. (6) and Watters et al. (7), respectively, expand the collection of Acrs, which may have application in the regulation of gene editing.

Although CRISPR-Cas are programmable immune systems that can provide resistance to any pathogen, they show remarkable diversity of molecular organization, with two distinct classes, six types, and about 25 subtypes discovered to date (8). The CRISPR-Cas types and subtypes differ primarily in the protein composition and structure of the effector complexes that are involved in interference—that is, recognition and cleavage of target nucleic acids by the CRISPR-Cas systems. Hence, there is a diversity of Acrs that appear to be specific for CRISPR-Cas subtypes or even to lower levels of classification. So far, all discovered Acrs have been shown to target several subtypes of class I or class II CRISPR-Cas systems. Marino et al. and Watters et al. identify previously unknown Acrs against subtypes I-C and V-A. The latter finding is of special interest considering both the distinct structure of the subtype V-A effector (known as Cas12a or Cpf1) and its increasing use as a genome-editing tool (3).

Anti-CRISPR proteins

In the presence of a cognate protospacer, CRISPR-Cas systems protect archaeal and bacterial cells from viruses. Virus-encoded Acrs prevent virus inactivation by CRISPR-Cas systems.


The approaches used to identify Acrs are both straightforward and elegant, seamlessly combining bioinformatics and biochemistry (see the supplementary materials). The clue that helped the original discovery of the Acrs was the ability of certain Pseudomonas bacteriophages (viruses that infect bacteria) to reproduce in bacterial strains that were supposed to be immune because of the presence of CRISPR spacers against those phages. This observation prompted genetic and biochemical experiments that led to the identification of the first Acrs (9). A related observation is “self-targeting,” in which a provirus integrated in a bacterial genome is targeted by one or more CRISPR spacers from the same genome, but the provirus persists in the bacterial population (7, 10). Such tolerance requires inactivation of the CRISPR-Cas system(s), and Acrs are the most likely solution. Indeed, experimental tests for Acr activity of proteins encoded by predicted proviruses (prophages) in the tolerant bacteria has led to the identification of several additional Acrs (7, 10). A remarkable feature that became apparent once the first few Acrs were identified is that viruses often possess multiple acr genes that form clusters in virus genomes (4, 5). Furthermore, although the Acrs themselves characteristically show little sequence conservation, the clusters of acr genes are typically flanked by more highly conserved genes that often encode proteins containing a helix-turn-helix (HTH) DNA-binding domain and are implicated in the regulation of acr gene expression. The presence of such conserved Acr-associated (aca) genes enables the use of the “guilt by association” approach to predict new Acrs (5, 11). Together, these features provide for a consistent and relatively simple strategy for the identification of new Acrs that has been successfully used by Marino et al. and Watters et al.

The discovery of multiple Acrs against type V CRISPR-Cas in predicted MGEs from the bacterium Moraxella bovoculi brings the tally of Acrs to more than 30 families. These Acrs are encoded by diverse bacterial and archaeal viruses and target CRISPR-Cas systems of both classes, three types, and seven subtypes (see supplementary materials). The Acrs only share some generic features, such as gene clustering, small size (typically between 50 and 150 amino acids), and extreme sequence and structural diversity. In a sense, the major common feature of the Acrs is that they have so little in common in sequence and structure. This diversity and the fast evolution of the Acr protein families are not surprising given that the Acrs are locked in the never-ending arms race with CRISPR-Cas. The mechanisms of action of those Acrs that have been studied in detail are also common in that they all directly bind the cognate effector Cas proteins and block their activity (4, 5). However, different Acrs bind the same CRISPR effector, such as Cas9, at distinct, nonoverlapping sites or, in the case of the class I CRISPR-Cas system, bind to different subunits of the effector Cas complex.

Undoubtedly, exploration of Acr diversity is only just beginning. One can confidently predict that the Acrs encoded by bacterial and archaeal viruses are numerous, and multiple Acrs target each of the CRISPR-Cas subtypes. Thus, Acrs that target three types of CRISPR-Cas remain to be discovered, along with additional Acrs that target the widespread class I and II CRISPR-Cas systems. Numerous bacterial and archaeal viruses contain clusters of genes that encode small proteins, sometimes flanked by genes for HTH-containing proteins (see supplementary materials). The proteins encoded in such gene clusters are candidate Acrs, although activity against other host defense systems cannot be ruled out. Given that nearly all archaea have CRISPR-Cas systems, Acrs can be expected to be particularly abundant in archaeal viruses, and indeed, many archaeal virus genomes contain candidate gene clusters (12). Strikingly, such clusters account for up to 40% of some archaeal virus genomes, but their size and content are highly variable, even among closely related viruses. A notable example of this variation is the anti–type I-D Acr (AcrID1) family, one of the largest, relatively highly conserved Acr families, with more than 50 members encoded by archaeal viruses of three families (13). Among the Sulfolobus islandicus rod-shaped viruses (SIRVs ), the number of AcrID1 genes varies from 0 to 12 (among about 50 virus genes), suggesting major, still not understood, differences in virus-host interactions.

The discovery of the Acrs came as a sensation, but in retrospect, this could have been readily anticipated. Indeed, the diversity of the Acrs reflects a general organizational principle of the virosphere: Apart from the structural and replication gene modules, virus genes encode proteins involved in virus-host interactions, of which the primary form is inhibition of host defenses. Perhaps the closest parallel with the Acrs are suppressors of gene silencing that are encoded by most plant viruses and inhibit the plant RNA interference (RNAi) machinery that is functionally analogous, albeit not homologous, to CRISPR-Cas (14). Furthermore, similar to the Acrs, plant virus RNAi suppressors are small, fast-evolving proteins.

Even if the general principles of Acr function and evolution can be considered understood, burning questions abound. Where do the Acrs come from? So far, not a single homolog of any of the Acrs with a different function has been discovered through sequence or structure comparison in either microbial or MGE genomes (apart from the presence of an HTH domain in AcrIIA1), despite extensive effort. Strikingly, all solved Acr structures seem to represent unique protein folds (15). Are there acr genes in MGEs other than viruses, such as plasmids and transposons? Such MGEs could have benefited from Acrs, but so far, none have been discovered. Do archaea and bacteria encode anti-Acrs? Again, from the arms-race perspective, one could predict the existence of such “defense against counter-defense.” If so, how many turns does the arms-race spiral make? So far, all Acrs have been shown to inhibit Cas effector activity; do other types of Acrs exist—in particular, those inhibiting adaptation (spacer acquisition)? In the more practical vein, can Acrs be harnessed as useful regulatory tools for CRISPR-based genome editors? There is no doubt that these and other open questions will fuel fascinating Acr research for years to come.


View Abstract

Navigate This Article