History of CRISPR

aUnité de Biologie Moléculaire du Gène Chez les Extrêmophiles, Département de Microbiologie, Institut Pasteur, Paris, France

bDepartment of Bioscience and Biotechnology, Faculty of Agriculture, Kyushu University, Fukuoka, Japan

aUnité de Biologie Moléculaire du Gène Chez les Extrêmophiles, Département de Microbiologie, Institut Pasteur, Paris, France

aUnité de Biologie Moléculaire du Gène Chez les Extrêmophiles, Département de Microbiologie, Institut Pasteur, Paris, France

cInstitute of Integrative Cellular Biology, Université Paris Sud, Orsay, France

Clustered regularly interspaced short palindromic repeat (CRISPR)-Cas systems are well-known acquired immunity systems that are widespread in archaea and bacteria. The RNA-guided nucleases from CRISPR-Cas systems are currently regarded as the most reliable tools for genome editing and engineering. The first hint of their existence came in 1987, when an unusual repetitive DNA sequence, which subsequently was defined as a CRISPR, was discovered in the Escherichia coli genome during an analysis of genes involved in phosphate metabolism. Similar sequence patterns were then reported in a range of other bacteria as well as in halophilic archaea, suggesting an important role for such evolutionarily conserved clusters of repeated sequences. A critical step toward functional characterization of the CRISPR-Cas systems was the recognition of a link between CRISPRs and the associated Cas proteins, which were initially hypothesized to be involved in DNA repair in hyperthermophilic archaea. Comparative genomics, structural biology, and advanced biochemistry could then work hand in hand, not only culminating in the explosion of genome editing tools based on CRISPR-Cas9 and other class II CRISPR-Cas systems but also providing insights into the origin and evolution of this system from mobile genetic elements denoted casposons. To celebrate the 30th anniversary of the discovery of CRISPR, this minireview briefly discusses the fascinating history of CRISPR-Cas systems, from the original observation of an enigmatic sequence in E. coli to genome editing in humans.

Clustered regularly interspaced short palindromic repeat (CRISPR)-Cas systems are currently in the spotlight of active research in biology. The first CRISPRs were detected 30 years ago by one of the authors of this review (Y. Ishino) in Escherichia coli in the course of the analysis of the gene responsible for isozyme conversion of alkaline phosphatase (1). The structural features of CRISPRs are shown in Fig. 1. At the time, it was hardly possible to predict the biological function of these unusual repeated sequences due to the lack of sufficient DNA sequence data, especially for mobile genetic elements. The actual function of this unique sequence remained enigmatic until the mid-2000s. In 1993, CRISPRs were for the first time observed in archaea, specifically in Haloferax mediterranei (2), and subsequently detected in an increasing number of bacterial and archaeal genomes, since life science moved into the genomic era. Conservation of these sequences in two of the three domains of life was critical for an appreciation of their importance. In the early 2000s, the discovery of sequence similarity between the spacer regions of CRISPRs and sequences of bacteriophages, archaeal viruses, and plasmids finally shed light on the function of CRISPR as an immune system. This dramatic discovery by Mojica and others (3) was grossly underappreciated at that time and was independently published in 2005 by three research groups (3,–5). In parallel, several genes previously proposed to encode DNA repair proteins specific for hyperthermophilic archaea (6) were identified as being strictly associated with CRISPR and were designated cas (CRISPR-associated) genes (7). Comparative genomic analyses thus suggested that CRISPR and Cas proteins (the cas gene products) actually work together and constitute an acquired immunity system to protect prokaryotic cells against invading viruses and plasmids, analogous to the eukaryotic RNA interference (RNAi) system (8).

The structural features of CRISPR. The repeat sequences with constant length generally have dyad symmetry to form a palindromic structure (shown by arrows). Two examples are shown by the first identified CRISPR from E. coli (bacteria) and H. mediterranei (archaea). The spacer regions also have a constant length but no sequence homology.

This minireview focuses on the contribution of early fundamental microbiological research to the discovery of the CRISPR-Cas system and to our understanding of its function and mode of action (for other recent reviews on the history of the research on the CRISPR-Cas system, see references 9,–14). We also emphasize recent discoveries that shed light on the origins of the system and suggest that more tools remain to be discovered in the microbial world that could still improve our genome editing capacity.

In the mid-1980s, when studying isozyme conversion of alkaline phosphatase (AP), one of us (Y. Ishino), in an attempt to identify the protein responsible for the isozyme conversion of AP in the periplasm of the E. coli K-12 cells, sequenced a 1.7-kbp E. coli DNA fragment spanning the region containing the iap (isozyme of alkaline phosphatase) gene (1). The isozyme of AP was previously detected by biochemical and genetic analyses (15). At that time, for conventional M13 dideoxy sequencing, single-stranded template DNA had to be produced by cloning the target DNA into an M13 vector, whereas the dideoxy chain termination reaction was performed by Klenow fragment of E. coli polymerase I (Pol I). The reaction products were labeled by the incorporation of [α32P]dATP, and the sequence ladder images were obtained by autoradiography. For sequencing, the cloned DNA fragment had to be subcloned into M13 mp18 and mp19 vectors (for the coding and noncoding strands) after digestion into short fragments. During the sequencing of the DNA fragment containing iap, one of us realized that the same sequence appeared many times in different clones. Furthermore, it was difficult to read the repeated sequences precisely using the Klenow fragment at 37°C, because of nonspecific termination of the dideoxynucleotide incorporation reactions for the template DNA, due to secondary structure formation by the palindromic sequence. This is why it took several months to read the sequence of the CRISPR region precisely in 1987 (1). A peculiar repeated sequence was detected downstream of the translation termination codon for the iap gene (Fig. 2). It is remarkable that the exact same region can be sequenced in just 1 day using current technology by amplification of the target region by PCR directly from the genome, followed by fluorescent labeling and cycle sequencing at 72°C (Fig. 3). The feature of the repetitive sequence was so mysterious and unexpected that it is mentioned in the Discussion, even though its function was not understood (1). Notably, the same sequence containing a dyad symmetry of 14 bp was repeated five times with a variable 32-nucleotide sequence interspersed between the repeats (Fig. 2). Well-conserved nucleotide sequences containing a dyad symmetry, named repetitive extragenic palindromic (REP) sequences (16), had been previously found in E. coli and Salmonella enterica serovar Typhimurium and were suggested to stabilize mRNA (17). However, no similarities were found between the REP sequences and the repeated sequences detected downstream of the iap gene. In fact, this sequence was, at the time, unique in sequence databases. As it later turned out, this was the first encounter with a CRISPR sequence. Soon after, similar sequences were detected by Southern blot hybridization analysis in other E. coli strains (C600 and Ymel) and in two other members of the Enterobacteriaceae, Shigella dysenteriae and S. enterica serovar Typhimurium (phylum Proteobacteria) (18). Subsequently, similar repeated sequences were also found in members of the phylum Actinobacteria, such as Mycobacterium tuberculosis (19), but not in the closely related strain Mycobacterium leprae, prompting the use of these highly polymorphic repeated sequences for strain typing (20).

The first CRISPR found in E. coli. As a result of the iap gene analysis from E. coli, a very ordered repeating sequence was found downstream of the iap gene. The conserved sequence unit was repeated 5 times with a constant length of spaces in 1987. It turns out that the repeat was 14 times in total by the subsequent genome analysis. The cas gene cluster was also identified at the downstream region. nt, nucleotides.

The first CRISPR sequence in E. coli. The exact same region, downstream of the iap gene, which was found in 1987 by a conventional dideoxy sequencing, was read by a cycle sequencing with fluorescent labeling recently. The CRISPR units are shown by pink shading.

A major advance was made when similar repeated sequences were identified by Mojica and coworkers in the archaeon Haloferax mediterranei during the research on regulatory mechanisms allowing extremely halophilic archaea to adapt to high-salt environments (2). Transcription of the genomic regions containing the repeated sequences was demonstrated by Northern blotting (2), but compelling evidence for the processing of the transcripts into several different RNA products was shown only more recently (12). In that study, the authors first suggested that these repeated sequences could be involved in the regulation of gene expression, possibly facilitating the conversion of the double-stranded DNA from B to Z form for the specific binding of a regulator protein. It was indeed often suggested at that time that the high GC content of halophilic genomes could facilitate such a B-to-Z transition for regulatory purposes at the high intracellular salt concentration characteristic of haloarchaea. However, such an explanation could not be valid for bacteria. Soon after, the same authors found a similar repeated sequence in Haloferax volcanii and hypothesized that these repeated sequences could be involved in replicon partitioning (21).

In the meantime, the invention of automated sequencing machines and the development of efficient procedures for DNA sequencing during the 90s provided scientists for the first time with access to complete genome sequences. Starting with Haemophilus influenzae (22), followed by Methanocaldococcus jannaschii (23) and Saccharomyces cerevisiae (24), all three domains of life entered into the genomics era. Then, the unusual repeated sequences interspersed with nonconserved sequences, first detected in E. coli and H. mediterranei, were identified in an increasing number of bacterial and archaeal genomes and were described using different names by different authors, such as short regularly spaced repeats (SRSRs) (2), spacers interspersed direct repeats (SPIDRs), and large cluster of tandem repeats (LCTRs) (25). In the hyperthermophilic archaea Pyrococcus abyssi and Pyrococcus horikoshii, two sets of LCTR sequences were located symmetrically on each side of the replication origin, again suggesting a possible role in chromosome partitioning. However, they were more numerous and scrambled in the genome of Pyrococcus furiosus, casting doubt on this interpretation (26).

Mojica et al. were the first to realize that all these bacterial and archaeal sequences were functionally related (27). The term CRISPR was proposed by Jansen et al. in 2002 (7) and became generally accepted by the community working on these sequences, which precluded further confusion caused by many different names for the related repeat sequences. Comparative genomics studies illuminated the common characteristics of the CRISPR, namely that (i) they are located in intergenic regions. (ii) they contain multiple short direct repeats with very little sequence variation, (iii) the repeats are interspersed with nonconserved sequences, and (iv) a common leader sequence of several hundred base pairs is located on one side of the repeat cluster.

The fact that these mysterious sequences were conserved in two different domains of life pointed to a more general role of these sequences. CRISPR sequences were found in nearly all archaeal genomes and in about half of bacterial genomes, rendering them the most widely distributed family of repeated sequences in prokaryotes. As of today, CRISPR sequences have not been found in any eukaryotic genome.

The accumulation of genomic sequences in the beginning of this century enabled scientists to compare the genomic context of CRISPR regions in many organisms, which led to the discovery of four conserved genes regularly present adjacent to the CRISPR regions. The genes were designated CRISPR-associated genes 1 through 4 (cas1 to cas4) (7). No similarity to functional domains of any known protein was identified for Cas1 and Cas2. In contrast, Cas3 contained the seven motifs characteristic of the superfamily 2 helicases, whereas Cas4 was found to be related to RecB exonucleases, which work as part of the RecBCD complex for the terminal resection of the double-strand breaks to start homologous recombination. Therefore, Cas3 and Cas4 were predicted to be involved in DNA metabolism, including DNA repair and recombination, transcriptional regulation, and chromosome segregation. Due to their association with CRISPRs, it was suggested that Cas proteins are involved in the genesis of the CRISPR loci (7).

At about the same time, Makarova and colleagues independently and systematically analyzed the conserved gene contexts in all prokaryotic genomes available at the time and found several clusters of genes corresponding to cas genes (encoding putative DNA polymerase, helicase, and RecB-like nuclease) in the genomes of hyperthermophilic archaea and in the two hyperthermophilic bacteria with available genome sequences, Aquifex and Thermotoga (8). These conserved genes were not found at that time in mesophilic and moderate thermophilic archaea and bacteria. Based on this observation, it was predicted that these proteins could be part of a “mysterious” uncharacterized DNA repair system specific to thermophilic organisms.

In the beginning of the genomic era, most of the archaeal genome sequences were those of thermophilic and hyperthermophilic organisms. Furthermore, thermophilic archaea, in addition to the hyperthermophilic bacteria, such as Aquifex aeolicus and Thermotoga maritima, have more and larger CRISPRs than mesophilic organisms (7). These observations first suggested that the function of a CRISPR may be related to adaptation of organisms to high temperatures. However, with more and more sequences becoming available, it turned out that this correlation was not robust and that many mesophilic organisms also contained CRISPR sequences. The eureka moment came when Francisco Mojica in Alicante and Christine Pourcel in Orsay noticed independently that the spacer regions between the repeat sequences are homologous to sequences of bacteriophages, prophages, and plasmids (3, 4). Importantly, based on the literature review, they pointed out that the phages and plasmids do not infect host strains harboring the homologous spacer sequences in the CRISPR. From these observations, they independently proposed that CRISPR sequences function in the framework of a biological defense system similar to the eukaryotic RNAi system to protect the cells from the entry of these foreign mobile genetic elements. The two groups also suggested that the CRISPRs can somehow trigger the capture of pieces of foreign invading DNA to constitute a memory of past genetic aggressions (3, 4). In a third influential paper from the same year, Bolotin and colleagues confirmed these observations, further noticing a correlation between the number of spacers of phage origin and the degree of resistance to phage infection, and they suggested that CRISPR could be used to produce antisense RNA (5) (for a brief historical account, see Morange [9]).

As mentioned above, these seminal publications were grossly underappreciated at the time and published in specialized journals (12). Interestingly, Morange suggested that lack of adequate recognition of the 2005 papers at that time and in subsequent years in some publications and reviews might be due to both cultural and sociological reasons, based partly on the predominance of experimental molecular biologists over microbiologists and evolutionists (9). In two of the three 2005 papers, the authors acknowledged the previous discovery of the cas genes, suggesting that proteins encoded by these genes should be involved in the functioning of this new putative prokaryotic immune system (4, 5).

The predicted role of Cas proteins as effectors of prokaryotic immunity was emphasized a year after in an exhaustive analytical paper published by Makarova et al. (8). Building on their previous work, Makarova et al. performed a detailed analysis of the Cas protein sequences and attempted to predict their functions in a mechanism similar to the eukaryotic RNAi system (8). Notably, in many cases, these often nontrivial functional predictions, as in the case of Cas1 integrase, were fully confirmed experimentally several years later and continue to guide experimental research on the CRISPR-Cas systems. Importantly, they pinpointed that the CRISPR-Cas system, with its memory component, rather resembles the adaptive immune system of vertebrates, with the crucial difference being that the animal immune system is not inheritable. Considering the diversity of the CRISPR-Cas systems, their erratic distribution suggesting high mobility, and their ubiquity in archaea, Makarova et al. suggested that the CRISPR-Cas system emerged in an ancient ancestor of archaea and spread to bacteria horizontally. They concluded on a practical note, suggesting that CRISPR-Cas systems could be exploited to silence genes in organisms encoding Cas proteins (8).

The function of the CRISPR-Cas system as a prokaryotic acquired immune system was finally experimentally proven in 2007, using the lactic acid bacterium Streptococcus thermophilus in 2007 (28). Insertion of the phage sequence into the spacer region of the CRISPR of S. thermophilus made this strain resistant to the corresponding phage. On the other hand, this bacterial resistance to the phage infection disappeared when the corresponding protospacer sequence was deleted from the phage genome. In addition, it was experimentally demonstrated that CRISPR-Cas restricts the transformation of plasmids carrying sequences matching the CRISPR spacers (29). Then, van der Oost's group reconstituted the immunity system using the E. coli CRISPR, which was originally discovered in 1987. They demonstrated that the processed RNA molecules from the transcription of the CRISPR region function by cooperation with the Cas proteins produced from the genes located next to the CRISPR (30). Around the same time, metagenomic analysis of archaea by Andersson and Banfield indicated dynamic changing of sequences at CRISPR loci on a time scale of months, and new spacer sequences corresponding to phages in the same communities appeared (31). Subsequently, the CRISPR-Cas system of S. thermophilus expressed in E. coli showed heterologous protection against plasmid transformation and phage infection by the reconstituted CRISPR-Cas9 system of S. thermophilus (32). This work also showed that cas9 is, in that case, the sole cas gene necessary for CRISPR-encoded interference. Soon after, it has been proven that the purified Cas9-CRISPR RNA (crRNA) complex is capable of cleaving the target DNA in vitro (33, 34). The CRISPR-Cas system of Streptococcus pyogenes was then applied to perform genome editing in human nerve and mouse kidney cells (35, 36). Thus, CRISPR-Cas came to be widely known as the prokaryotic acquired immunity system (37, 38). The various steps underlying the functioning of this system are shown in Fig. 4.

Process of CRISPR-Cas acquired immune system. (Top) Adaptation. The invading DNA is recognized by Cas proteins, fragmented and incorporated into the spacer region of CRISPR, and stored in the genome. Expression (bottom). Pre-crRNA is generated by transcription of the CRISPR region and is processed into smaller units of RNA, named crRNA. (Bottom) Interference. By taking advantage of the homology of the spacer sequence present in crRNA, foreign DNA is captured, and a complex with Cas protein having nuclease activity cleaves DNA.

Numerous and highly diverse Cas proteins are involved in different stages of CRISPR immunity; they exhibit a variety of predicted nucleic acid-manipulating activities, such as nucleases, helicases, and polymerases, which have been described in detail in several excellent recent reviews (39,–42). In a nutshell, Cas1 and Cas2 are conserved throughout most known types of CRISPR-Cas systems and form a complex that represents the adaptation module required for the insertion of new spacers into the CRISPR arrays. During the expression stage, the CRISPR locus is transcribed and the pre-crRNA transcript is processed by the type-specific Cas endonucleases into the mature crRNAs. During the interference stage, the crRNAs are bound by the effector Cas endonucleases, and the corresponding complexes are recruited to and cleave the target DNA or RNA in a sequence-dependent manner (Fig. 4). Notably, unlike the adaptation module, Cas enzymes involved in the expression and interference stages vary from one CRISPR-Cas type to the other, and the same enzymes may participate in both stages of immunity.

It is striking that closely related strains can vary considerably in their CRISPR content and distribution. For example, in the Mycobacterium genus, CRISPR exists in M. tuberculosis but not in M. leprae. On the other hand, phylogenetically distant E. coli and Mycobacterium avium as well as Methanothermobacter thermautotrophicus and Archaeoglobus fulgidus carry nearly identical CRISPR sequences (7). The number of CRISPR arrays in one genome varies from 1 to 18, and the number of repeat units in one CRISPR array varies from 2 to 374 (43). Based on the CRISPR database (http://crispr.i2bc.paris-saclay.fr), as of May 2017, CRISPRs were identified in the whole-genome sequences of 202 (87%) out of 232 analyzed archaeal species and 3,059 (45%) out of 6,782 bacterial species. Interestingly, a survey of 1,724 draft genomes suggested that CRISPR-Cas systems are much less prevalent in environmental microbial communities (10.4% in bacteria and 10.1% in archaea). This large difference between the prevalence estimated from complete genomes of cultivated microbes compared to that of the uncultivated ones was attributed to the lack of CRISPR-Cas systems across major bacterial lineages that have no cultivated representatives (44).

As shown in Fig. 5, the latest classification of CRISPR-Cas systems includes two classes, 1 and 2, based on the encoded effector proteins (45). Class 1 CRISPR-Cas systems work with multisubunit effector complexes consisting of 4 to 7 Cas proteins present in an uneven stoichiometry. This system is widespread in bacteria and archaea, including in all hyperthermophiles, comprising ∼90% of all identified CRISPR-cas loci. The remaining ∼10% belong to class 2, which use a single multidomain effector protein and are found almost exclusively in bacteria (46).

Genome editing by CRISPR-Cas9. The principle of genome editing is the cleavage of double-stranded DNA at a targeted position on the genome. The type II is the simplest as a targeted nuclease among the CRISPR-Cas systems. The CRISPR RNA (crRNA), having a sequence homologous to the target site, and trans-activating CRISPR RNA (tracrRNA) are enough to bring the Cas9 nuclease to the target site. The artificial linkage of crRNA and tracrRNA into one RNA chain (single-guide RNA [sgRNA]) has no effect on function. Once the Cas9-sgRNA complex cleaves the target gene, it is easy to disrupt the function of the gene by a deletion or insertion mutation. This overwhelmingly simple method is now rapidly spreading as a practical genomic editing technique.

Each class currently includes three types, namely, types I, III, and IV in class 1 and types II, V, and VI in class 2. Types I, II, and III are readily distinguishable by virtue of the presence of unique signature proteins: Cas3 for type I, Cas9 for type II, and Cas10 for type III. The multimeric effector complexes of type I and type III systems, known as the CRISPR-associated complex for antiviral defense (Cascade) and the Csm/Cmr complexes, respectively, are architecturally similar and evolutionarily related (47,–52). Unlike all other known CRISPR-Cas systems, the functionally uncharacterized type IV systems do not contain the adaptation module consisting of nucleases Cas1 and Cas2 (47, 53). Notably, the effector modules of subtype III-B systems are known to utilize spacers produced by type I systems, testifying to the modularity of the CRISPR-Cas systems (54). Although many of the genomes encoding type IV systems do not carry identifiable CRISPR loci, it is not excluded that type IV systems, similar to subtype III-B systems, use crRNAs from different CRISPR arrays once these become available (53).

Finally, each type is classified into multiple subtypes (I-A to F and U and III-A to D in class 1; II-A to C, V-A to E and U, and VI-A to C in class 2) based on additional signature genes and characteristic gene arrangements (45, 51). Figure 6B shows the distribution of CRISPR-Cas systems in archaea and bacteria.

Most recent classification of CRISPR-Cas immune systems. (A) Based on the detailed sequence analyses and gene organization of the Cas proteins, CRISPR-Cas was classified into two major classes depending on whether the effector is a complex composed of multiple Cas proteins or a single effector. In addition to the conventional types I, II, and III, the types IV and V were added to classes 1 and 2, respectively. Types IV and V are those which do not have Cas1 and Cas2, necessary for adaptation process, in the same CRISPR loci. Type VI was added most recently to class 2. (B) Chart showing the proportions of identified CRISPR-cas loci in the total genomes of bacteria and archaea referred from the literature (51, 53). The proportions of loci that encode incomplete systems or that could not be classified unambiguously are not included.

The simple architecture of the effector complexes has made class 2 CRISPR-Cas systems an attractive choice for developing a new generation of genome editing technologies (Fig. 7). Several distinct class 2 effectors have been reported, including Cas9 in type II, Cas12a (formerly Cpf1), Cas12b (C2c1) in type V, and Cas13a (C2c2) and Cas13b (C2c3) in type VI (45, 51). The most common and best studied multidomain effector protein is Cas9, a crRNA-dependent endonuclease, consisting of two unrelated nuclease domains, RuvC and HNH, which are responsible for cleavage of the displaced (nontarget) and target DNA strands, respectively, in the crRNA-target DNA complex. Type II CRISPR-cas loci also encode a trans-activating crRNA (tracrRNA) which might have evolved from the corresponding CRISPR. The tracrRNA molecule is also essential for pre-crRNA processing and target recognition in the type II systems. The molecular mechanism of the target DNA cleavage by the Cas9-crRNA complex, shown in Fig. 7, has been elucidated at the atomic level by the crystal structure analysis of the DNA-Cas9-crRNA complex (55).

Cleavage mechanism of target DNA by crRNA-tracrRNA-Cas9. The Cas9-crRNA-tracrRNA complex binds to foreign DNA containing PAM, where Cas9 binds and starts to unwind the double strand of the foreign DNA to induce duplex formation of crRNA and foreign DNA. Cas9 consists of two regions, called the REC (recognition) lobe and the NUC (nuclease) lobe. The REC lobe is responsible for nucleic acid recognition. The NUC lobe contains the HNH and RuvC nuclease domains and a C-terminal region containing a PAM-interacting (PI) domain. The HNH domain and the RuvC domain cleave the DNA strand, forming a duplex with crRNA and the other DNA strand, respectively, so that double-strand break occurs in the target DNA.

A gene originally denoted as cpf1 is present in several bacterial and archaeal genomes, where it is adjacent to cas1, cas2, and a CRISPR array (45). Cas12a (Cpf1), the prototype of type V effectors, contains two RuvC-like nuclease domains but lacks the HNH domain. However, recent structural analysis of the Cas12a-crRNA-target DNA complex revealed a second nuclease domain with a unique fold that is functionally analogous to the HNH domain of Cas9 (56). Cas12a is a single-RNA-guided nuclease that does not require a tracrRNA, which is indispensable for Cas9 activity (57). The protein also differs from Cas9 in its cleavage pattern and in its protospacer adjacent motif (PAM) recognition, which determines the target strands.

The discovery of two distantly related class 2 effector proteins, Cas9 and Cas12a, suggested that other distinct variants of such systems could exist. Indeed, more recently, Cas12b (type V), and Cas13a and Cas13b (type VI), which are distinct from Cas9 and Cas12a, have been discovered through directed bioinformatics search for class II effectors, and their activities were confirmed (58). Type V effectors, similar to Cas9, need a tracrRNA for the targeted activity. Most of the functionally characterized CRISPR-Cas systems to date have been reported to target DNA, and only the multicomponent type III-A and III-B systems additionally target RNA (59). In contrast, the type VI effectors Cas13a and Cas13b specifically target RNA, thereby mediating RNA interference. Unlike type II and type V effectors, Cas13a and Cas13b lack characteristic RuvC-like nuclease domains and instead contain a pair of higher eukaryote and prokaryote nucleotide-binding (HEPN) domains (60). The discovery of novel class 2 effectors will most likely provide new opportunities for the application of CRISPR systems to genome engineering technology (61).

Analysis of clusters of poorly characterized narrowly spread fast-evolving genes in archaeal genomes, denoted “dark matter islands” (62), revealed several islands encoding Cas1 proteins not associated with CRISPR loci (Cas1-solo) (63). Comprehensive interrogation of the dark matter islands revealed that cas1-solo genes are always located in vicinity of genes encoding family B DNA polymerases and several other conserved genes (64). Furthermore, these gene ensembles were found to be surrounded by long inverted repeats and further flanked by shorter direct repeats, which, respectively, resembled terminal inverted repeats (TIRs) and target site duplications (TSD) characteristic of various transposable elements. However, none of the identified Cas1-solo-encoding genomic loci carried genes for known transposases or integrases. Thus, it was hypothesized that Cas1 is the principal enzyme responsible for the mobility of these novel genetic elements, which were accordingly named casposons (64). Casposons were found to be widespread in the genomes of methanogenic archaea as well as in thaumarchaea, but they are also present in different groups of bacteria. Strong evidence of recent casposon mobility was obtained by comparative genomic analysis of more than 60 strains of the archaeon Methanosarcina mazei, in which casposons are variably inserted in several distinct sites indicative of multiple recent gains and losses (65). Based on the gene content, taxonomic distribution, and phylogeny of the Cas1 proteins, casposons are currently classified into 4 families (66).

Biochemical characterization of the casposon Cas1 (“casposase”) encoded in the genome of a thermophilic archaeon, Aciduliprofundum boonei, has confirmed the predicted integrase activity (67, 68). Integration showed strong target site preference and resulted in the duplication of the target site regenerating the TSD observed in the A. boonei genome (68). Duplication of the TSD segment resembles the duplication of the leader sequence-proximal CRISPR unit upon integration of a protospacer catalyzed by the Cas1-Cas2 adaptation machinery of CRISPR-Cas (69, 70). Remarkably, the sequence features of the casposon target site are functionally similar to those required for directional insertion of new protospacers into CRISPR arrays. In both systems, the functional target site consists of two components: (i) a sequence which gets duplicated upon integration of the incoming DNA duplex (i.e., the TSD segment in the case of casposon and a CRISPR unit during protospacer integration) and (ii) the upstream region which further determines the exact location of the integration (i.e., the leader sequence located upstream of the CRISPR array and the TSD-proximal segment in the A. boonei genome) (68).

Collectively, the comparative genomics and experimental results reinforced the mechanistic similarities and evolutionary connection between the casposons and the adaptation module of the prokaryotic adaptive immunity system, culminating in an evolutionary scenario for the origin of the CRISPR-Cas systems. It has been proposed that casposon insertion near a “solo-effector” innate immunity locus, followed by the immobilization of the ancestral casposon via inactivation of the TIRs, gave rise to the adaptation and effector modules, respectively, whereas the CRISPR repeats and the leader sequence evolved directly from the preexisting casposon target site (71, 72). An outstanding question in the above-mentioned scenario is the switch in substrate specificity of the ancestral casposase from the integration of defined casposon TIRs to the insertion of essentially random short (compared to casposon length) protospacer sequences. It has been suggested that coupling between Cas1 and Cas2 has been critical for this evolutionary transition (72).

Remarkably, casposons are not the only mobile genetic elements that contributed to the origin and evolution of the CRISPR-Cas systems. It has been demonstrated that class 2 effector proteins of types II and V have independently evolved from different groups of small transposons, which donated the corresponding RuvC-like nuclease domains (45, 58).

Microbial engineering directly influences the development of bioindustry. High-throughput genome editing tools are useful for breeding economically valuable strains. It is remarkable how quickly practical application of the CRISPR-Cas system has been adapted to genome editing in eukaryotic cells. Such rapid success of this technology in eukaryotic cells was linked to the fact that eukaryotes employ error-prone nonhomologous end joining (NHEJ) to repair double-strand breaks introduced by the CRISPR-Cas in the target sequence. The use of the CRISPR-Cas technology was not as revolutionary in bacteria, likely because other methods based on homologous recombination (HR) were already available for efficient manipulation of their genomes. Nevertheless, DNA toolkits based on CRISPR-Cas technology for genome editing, gene silencing, and genome-wide screening of essential genes in bacterial and archaeal genomes are gradually emerging and diversifying (73,–77). For instance, a CRISPR-Cas-mediated genome editing technique coupled with heterologous recombineering using linear single-stranded (single-stranded DNA recombineering [SSDR]) or double-stranded (double-stranded DNA recombineering [DSDR]) DNA templates has been developed and successfully applied in E. coli (78). In archaea, gene silencing has been established in Sulfolobus solfataricus, Sulfolobus islandicus, and Haloferax volcanii using the endogenous CRISPR-Cas systems (reviewed in references 77 and 79). More recently, Nayak and Metcalf have harnessed a bacterial Cas9 protein for genome editing in the mesophilic archaeon Methanosarcina acetivorans (80). Hopefully, a thermophilic counterpart of the CRISPR-Cas9 system (or other class 2 systems) will finally be established to perform genome editing in hyperthermophilic species, which are difficult to manipulate genetically. From that perspective, the diversity of CRISPR-Cas systems and mobile genetic elements, which remain to be fully explored, is a treasure trove for future exploitation.

The CRISPR loci are encoded by many bacterial and archaeal organisms and are remarkably diverse; thus, they have been used as genetic markers for species identification and typing, even before the elucidation of the actual function of the CRISPR-Cas, as described above. For example, typing of Mycobacterium tuberculosis is useful for diagnostic and epidemiological purposes (20, 81). Typing using CRISPR has been applied to Yersinia pestis (4, 82), Salmonella spp. (83, 84), and Corynebacterium diphtheriae (85). CRISPR-Cas9 can be used as an antimicrobial agent by cleaving the genomes of pathogenic bacteria, which is a novel mechanism of action. It is expected to be a valuable remedy for the control of antibiotic-resistant bacteria. For example, antibiotic-resistant bacteria, such as Staphylococcus spp., infecting the skin of mice were selectively killed using CRISPR-Cas9 (86). CRISPR-Cas9 also reportedly prevented intestinal infection by pathogenic E. coli (87). Although there are technical challenges, such as delivery methods, which must be overcome before CRISPR-Cas can be used as a safe therapeutic agent, active research in this direction is ongoing and is expected to yield solutions in the near future. Furthermore, imparting phage resistance in specific strains by the CRISPR-Cas system is extremely useful for protecting various beneficial bacteria in the fermented food industry from phage infection during the production process.

Since the HNH nuclease domain and the RuvC nuclease domain are responsible for the DNA cleavage activity of Cas9, Cas9 mutants devoid of cleavage activity (dCas9) were obtained by replacing the amino acids within each active center. The dCas9 protein is a useful tool for molecular biology experiments to regulate gene expression. CRISPR-dCas9 binds to the target DNA sequence but cannot cleave it. This activity of CRISPR-dCas9 is applicable to the labeling of a specific position, by fusing green fluorescent protein (GFP) to dCas9, which binds to the target sequence depending on the single-guide RNA (sgRNA) sequence (88). In addition to this live intracellular site-specific labeling, gene expression can be artificially controlled by linking dCas9 to either the promoter region or the open reading frame of a gene (89,–91). dCas9 can also be fused with a transcription activator or the ω subunit of bacterial RNA polymerase. However, it seems not to be as easy compared to suppression, although ingenious attempts have been made to promote transcription by designing a guide sequence that ensures the binding of dCas9 to a specific promoter.

The dCas9 protein is also useful for the techniques to reduce off-target cleavage in the genomes. An artificial CRISPR-Cas nuclease RFN (RNA-guided FokI nuclease), in which the nuclease domain of FokI is fused to dCas9, like zinc finger nuclease (ZFN) or transcription activator-like effector nuclease (TALEN), was developed by designing the guide RNA so that the nuclease domain can form a dimer at the target site. Since it can be used for double-strand cleavage with different guide RNAs for top and bottom DNA strands, the probability of nonspecific binding decreases (92,–94). The reduction in off-target cleavage was also achieved by using Cas9 nickase (Cas9n). A mutant Cas9, in which the Asp10 active residue in the RuvC domain was substituted with alanine, showed a nickase activity that cleaved only one strand of the target site with an appropriate sgRNA (33, 34). Therefore, nicking of both DNA strands by a pair of Cas9 nickases with different sgRNA leads to site-specific double-strand DNA breaks (DSBs). This paired nickase strategy can reduce off-target activity by 50- to 1,500-fold in cell lines and to facilitate gene knockout in mouse zygotes without sacrificing on-target cleavage efficiency (95).

A method for site-specific mutagenesis of genomic DNA by fusion of dCas9 with a cytidine deaminase has been developed (96). The sgRNA-induced cytidine deaminase causes a base substitution at the target site without cutting DNA. This method significantly reduces cytotoxicity compared to artificial nucleases and Cas9 nuclease and efficiently achieves intended modifications.

Another interesting solution was to split the Cas9 protein into two parts and reconstitute the Cas9 nuclease from the corresponding proteins (97, 98). The photoactivatable Cas9 (paCas9), which is activated by light irradiation, can be used for conditional genome editing. The activity of paCas9 is about 60% compared with the original Cas9, but it can be fully used for cutting the desired double-strand by light irradiation from the outside without changing the culture conditions (99).

Thus, as described above, the genome editing technique using the CRISPR-Cas immune system is not limited to the use of S. pyogenes CRISPR-Cas9, but further variants continue to be developed. These devices will certainly contribute to the improvement of genome editing technologies.

Only 30 years have passed since one of the authors of this review discovered a unique repeated sequence in the E. coli genome at the onset of his postdoctoral career. It was impossible to predict the possible function of this enigmatic sequence at the time; however, genomic revolution in the mid-1990s, coupled with the development of powerful bioinformatics tools, eventually enabled elucidation of the CRISPR functions. CRISPR arrays and Cas proteins, broadly distributed in the genomes of prokaryotes, especially in archaea, are now known to constitute the highly efficient acquired immunity system. Although discovery of the CRISPR-Cas by itself was a great feat of fundamental biology, it also led to the development of next-generation tools for genetic engineering. The development of the genome editing technology by CRISPR-Cas9 reminds of the times when the PCR was born.

When in vitro genetic engineering techniques using restriction endonucleases and nucleic acid modifying enzymes were established, it was still often a complex task to clone a single gene (as in the case of the iap gene). However, this difficulty was alleviated by the invention of PCR using a thermostable DNA polymerase that profoundly boosted the application of genetic engineering techniques in all biological laboratories worldwide. The discovery of a thermostable DNA polymerase was critical for the “PCR revolution,” because it enabled the design of a PCR apparatus for practical use. Similarly, in the case of genome editing, the CRISPR revolution was made possible by identifying the right enzymatic system (Cas9) that could simplify the methodology to exploit the potential of the CRISPR-Cas system. The curiosity of a mysterious repetitive sequence and a sustained inquiry mind for elucidating its function brought grand discoveries.

This perspective paper stemmed from discussions during Y.I.'s stay in P.F.'s laboratory thanks to the generous support from the program of visiting researchers in Institut Pasteur and the two-country exchange program of the Japan Society for the Promotion of Science (JSPS). P.F. is supported by an ERC grant from the European Union's Seventh Framework Program (FP/2007-2013)/Project.

We apologize to the researchers whose work was not cited due to space limitations.

Yoshizumi Ishino is a professor at Kyushu University, Japan. He graduated from Osaka University and earned an M.S. (in pharmaceutical sciences, 1983) and Ph.D. (Institute for Microbial Diseases, 1986) with research on restriction endonucleases and DNA ligase, respectively. He found the first CRISPR sequence in E. coli during his postdoctoral research in 1986. He worked on translation-related enzymes as his postdoctoral research at Yale University (1987 to 1989). He became a principal investigator (PI) at Biotechnology Research Laboratories in Takara Shuzo, Japan, in 1990, and moved to the Biomolecular Engineering Research Institute (BERI), a national project funded by the government (METI) and 18 companies, to manage a research group on nucleic acid-related enzymes. In 2002, he was appointed a full professor of protein chemistry and engineering at Kyushu University. His main research interests have been in molecular mechanisms of DNA replication and recombinational repair in Archaea, the third domain of life.

Mart Krupovic is an assistant professor in the Department of Microbiology at the Institut Pasteur of Paris, France. He received his Ph.D. in general microbiology in 2010 from the University of Helsinki, Finland. He has a profound interest in the origin and evolution of the global virosphere and the evolutionary connections between viruses and other types of mobile genetic elements. His research also focuses on different aspects of virus-host interactions in hyperthermophilic archaea and on a recently discovered superfamily of transposon-like elements called casposons, which might have played a pivotal role in the origin of the CRISPR-Cas immunity.

Patrick Forterre is a professor at the Institut Pasteur and professor emeritus at the University Paris-Saclay. He received his Ph.D. in molecular biology in 1985 from the University Paris VII. He has been a member of the European Academy of Microbiology since 2015. He has a profound interest in the evolution of molecular mechanisms and on the history of life on our planet. His experimental research has focused on various aspects of DNA metabolism and genome evolution in Archaea and their mobile genetic elements. He introduced many new original ideas and concepts and is still active for both experimental and theoretical studies of archaea and their viruses.