1. INTRODUCTION

The discovery of CRISPR (clustered regularly interspaced short palindromic repeats)-Cas (CRISPR-associated) adaptive immune systems has revolutionized the study of life sciences. CRISPR-Cas systems, encoded on prokaryotic genomes, consist of a CRISPR array and cas genes. CRISPR arrays form the genetic memory of the prokaryotic adaptive immune system. They are composed of repeating sequences (repeats) that are interspersed by variable sequences (spacers) that match sequences from mobile genetic elements such as viruses (Figure 1). The number and length of CRISPR loci vary among organisms, with the highest number of 23 CRISPR loci found in Methanocaldococcus sp. FS406-22 and the highest number of 600 individual spacers found in Haliangium ochraceum (Anderson et al. 2011). By comparison, a common model organism such as Pseudomonas aeruginosa UCBPP-PA14 has a single set of cas genes flanked by 2 CRISPR loci (with 21 and 14 spacers) (Cady et al. 2011). Cas genes encode the protein machinery that uses the information stored in the CRISPR array to launch an immune response against mobile genetic elements that carry a cognate sequence.

The first identification of a CRISPR locus dates back to 1987 (Ishino et al. 1987), but not until 2005 was it proposed that CRISPR loci act together with cas genes to provide adaptive immunity (Lillestol et al. 2006, Makarova et al. 2006, Mojica et al. 2005, Pourcel et al. 2005). The seminal paper by Barrangou and coworkers (2007) provided the first experimental evidence that confirmed this hypothesis. The authors exposed the lactic acid bacterium Streptococcus thermophilus strain DGCC7710 to phage and found that bacterial clones acquired resistance against phage by incorporating novel phage-derived sequences (spacers) into the CRISPR array. Spacers are thus acquired during the lifetime of an individual clone and the resultant phage resistance phenotype is inherited by future generations (Koonin & Wolf 2009, 2016).

During the past decade many mechanistic details of CRISPR-Cas have been elucidated, which has allowed these systems to be repurposed for sophisticated genome editing and gene regulation technologies. Although clear mechanistic differences exist between variants of the system, the general mode of action is that CRISPR transcripts are processed into small CRISPR RNA (crRNA) molecules that guide Cas proteins to bind and cleave complementary parasitic nucleic acids (see sidebar, The Three Stages of a CRISPR-Cas Immune Response). The mechanism of CRISPR-Cas is generally divided into three stages: adaptation, expression, and interference. During adaptation novel spacers are integrated into the CRISPR array [reviewed by Amitai & Sorek (2016)]. In the expression stage the CRISPR array is transcribed and processed into short crRNA molecules that are loaded onto Cas proteins [reviewed by Charpentier et al. (2015)]. During interference Cas–crRNA ribonucleoprotein complexes bind to complementary nucleic acids, followed by degradation of the target molecule (reviewed in van der Oost et al. 2014).

Although the mechanistic studies of CRISPR-Cas have raced ahead, we are only starting to understand the evolutionary ecology of CRISPR-Cas. In this review we first describe the scenario for the evolution of CRISPR-Cas systems as proposed by Koonin and coworkers (Shmakov et al. 2015). The core of this review focuses on three important but as yet largely unanswered questions: (a) How important are CRISPR-Cas systems in shaping bacteria–phage interactions? (b) When do CRISPR-Cas systems matter in natural populations? (c) What are the long-term consequences of encoding CRISPR-Cas systems in microbial genomes? Finally, we highlight some of the various applications of CRISPR-Cas for evolutionary and ecological research. The mechanism of CRISPR-Cas systems has been thoroughly reviewed elsewhere; however, essential details concerning the mechanistic basis of the different CRISPR-Cas systems are given in the sidebars. For more detailed overviews, we refer to some of the many excellent reviews on this subject (Makarova et al. 2015, van der Oost et al. 2014, Wiedenheft et al. 2012).

2. THE EVOLUTION OF CRISPR-Cas SYSTEMS

CRISPR-Cas systems are extremely diverse (see sidebars) and are currently classified into 2 classes, 6 types, and 19 subtypes (Figure 2) (Makarova et al. 2015, Shmakov et al. 2015). Despite this diversity, all systems share the basic features of a CRISPR locus, consisting of repeats alternated by variable spacers, and a set of associated cas genes. The diverse nature of CRISPR-Cas, which likely results from rapid evolution and extensive horizontal gene transfer (HGT), poses considerable challenges for the evolutionary classification of CRISPR-Cas systems.

Two proteins that are present in almost all CRISPR-Cas systems (but see Makarova et al. 2015 for exceptions to this rule) are Cas1 and Cas2, which are involved in the adaptation phase (see sidebar, The Three Stages of a CRISPR-Cas Immune Response). Cas1 is the most highly conserved Cas protein and hence the most suitable marker to trace the evolutionary history of CRISPR-Cas. The phylogeny of Cas1 generally corresponds well to the subtype classification of CRISPR-Cas (see sidebars, Class 1 CRISPR-Cas Systems and Class 2 CRISPR-Cas Systems). Not all cas1 genes, however, are CRISPR associated. Two clades of orphan cas1 (coined Cas1-solo) were identified that predominantly occur in archaeal lineages, and these have been suggested to represent ancestral genes of CRISPR-associated cas1 genes (Makarova et al. 2013). In some instances Cas1-solo genes are found within predicted transposable elements termed casposons, genomic islands flanked by terminal inverted repeats (Krupovic et al. 2014). Koonin and colleagues hypothesized that casposon-encoded Cas1 functions in the catalysis of casposon integration into the host genome, using a mechanism that may be akin to the spacer integration process, and speculated that casposon-encoded Cas1 may have played a key role in the early evolution of CRISPR-Cas (Krupovic et al. 2014, 2016). Specifically, they suggest that CRISPR-Cas systems may have emerged from a fusion of casposon and a cas10-like innate immunity gene.

Apart from Cas1, almost all CRISPR-Cas systems carry Cas2, which shares structural similarity with VapD, a toxin from Haemophilus influenzae (Kwon et al. 2012). Based on this finding, it has been suggested that Cas2 and other Cas proteins may act as a toxin to induce cell dormancy (He et al. 2014, Makarova et al. 2012), but this hypothesis awaits further experimental testing. Class 1 systems further encode a suite of genes that are essential for the expression and interference stages. Many of these genes belong to the RAMP (repeat associated mysterious protein) family. Makarova and coworkers (2011a, 2013) suggested a series of gene duplication events of an ancestral RAMP that carried a single RNA recognition motif (RRM) may have led to the relatively complex architecture of Class 1 systems of multiple RAMP genes, many of which carry multiple RRMs. Under this scenario, Class 1 systems evolved in thermophilic archaea and subsequently spread to bacteria (Makarova et al. 2011a). Class 2 systems lack much of the complexity observed for Class 1 systems and rely on an evolutionarily unrelated mechanism for target cleavage (see sidebar, Class 2 CRISPR-Cas Systems) in which the whole protein machinery necessary for the Class 1 interference stage is replaced by a single protein (Koonin & Makarova 2013) (cas9 gene in Type II systems; cpf1, c2c1, or c2c3, in Type V systems; and c2c2 in Type VI systems; see Figure 2 and sidebar, New Class 2 CRISPR-Cas Systems) (Makarova et al. 2015, Shmakov et al. 2015, Zetsche et al. 2015). It has been suggested by Shmakov et al. (2015) that the Class 2 effector nucleases may have acted as a stand-alone immune mechanism and may have been co-opted from a mobile genetic element by CRISPR-Cas to replace the typical Class 1 cas genes (Makarova et al. 2011a), resulting in the emergence of the two distinct classes of CRISPR-Cas. The sequence of events outlined in Shmakov et al. (2015) is shown in Figure 3.

3. HOW IMPORTANT ARE CRISPR-Cas SYSTEMS IN NATURE?

Although bacteria have many different immune mechanisms, CRISPR-Cas systems are—at least in terms of their mechanism—arguably their most sophisticated defense, and it is therefore tempting to assume that CRISPR-Cas systems are of key importance for bacteria–virus interactions in nature. What evidence there is suggests that CRISPR loci can evolve rapidly in some environments, consistent with an important role in antagonistic coevolution.

First, CRISPR loci belonging to different populations of the same species are typically highly diverse regarding their spacer content (Andersson & Banfield 2008, DeBoy et al. 2006, Heidelberg et al. 2009, Held et al. 2010, Kunin et al. 2008, Pourcel et al. 2005, Rho et al. 2012, Tyson & Banfield 2008). Second, direct analysis of CRISPR loci over time reveals dynamic spacer content, for example, in microbial metapopulations from Lake Tyrell (Emerson et al. 2013) and in streptococci in saliva samples from healthy human individuals (Pride et al. 2011). Third, CRISPR spacers tend to match phage genomes from the same (sympatric) rather than other (allopatric) environments, demonstrating that CRISPR-based antagonistic coevolution occurs at a timescale that is more rapid than bacterial dispersal (Berg Miller et al. 2012, Emerson et al. 2013, Held et al. 2010, Kunin et al. 2008, Sorokin et al. 2010, Vale & Little 2010). This effect was first reported by Kunin et al. (2008), showing that different isolates of a globally dispersed microbe (Candidatus accumulibacter phosphatis) carry unique CRISPR arrays that target sympatric rather than allopatric viruses. Analysis of available data sets from the Human Microbiome Project showed that CRISPR spacers are rarely shared between individual humans and some but not all spacers were shared between microbiomes resampled from the same individual through time (Rho et al. 2012). Another study found that many individuals shared bacteria with identical spacers, but the most recently acquired spacers were unique and matched co-occurring phage, in at least some cases (Stern et al. 2012). Streptococcal species from saliva were also shown to usually target sympatric phage, although in some instances they were equally likely to target sympatric and allopatric phage (Pride et al. 2012), perhaps due to global distribution of some phages. In the gut many phages are shared, even between geographically separated individuals (Stern et al. 2012), which may result from the fact that gut phages are often lysogenic (Breitbart et al. 2003) and therefore co-migrate with their host.

However, many studies have reported that spacer sequences from metagenomic data rarely match virus genomes (Anderson et al. 2011, Gogleva et al. 2014, Smedile et al. 2013). This finding has led to the suggestion that CRISPR may preferentially target rare viruses (Emerson et al. 2013). However, the lack of spacer matches is perhaps more likely due to low sequence availability of phage (Anderson et al. 2011, Berg Miller et al. 2012, Hatfull & Hendrix 2011, Reyes et al. 2010, Rohwer 2003), as illustrated by the fact that gut virome sequencing from different individuals still yields mostly novel viruses (Minot et al. 2011, 2012; Reyes et al. 2010, 2012). Indeed, when viromes were sequenced with bacterial CRISPR loci, spacers were typically found to match phage sequences from the same sample or geographical location (Andersson & Banfield 2008, Berg Miller et al. 2012, Emerson et al. 2013, Sorokin et al. 2010), with usually only the most recent spacers matching coexisting phage (Andersson & Banfield 2008, Stern et al. 2012).

Together, these studies have demonstrated that CRISPR loci can rapidly evolve in nature and are likely to play an important role in bacteria–phage interactions in these instances. However, from these correlational studies it is less clear when CRISPR-Cas systems are important fitness determinants of bacteria.

4. WHEN DO CRISPR–Cas SYSTEMS MATTER?

Despite the elegant mechanism of CRISPR-Cas–mediated defense, only half of all sequenced bacterial genomes encode CRISPR-Cas systems (Grissa et al. 2007), and different strains of the same species often differ with regard to the presence, number, and length of their CRISPR loci. Why do not all bacteria have CRISPR-Cas systems?

Theory and experimental studies suggest that ecological factors determine the relative benefit of CRISPR-Cas. Hence, these systems may be beneficial in some but not all environments. One example that suggests how ecology can impact the maintenance of CRISPR-Cas in bacterial genomes is provided by Mycoplasma gallisepticum, which lost its CRISPR-Cas system very rapidly following a host switch from chicken to wild finch (Delaney et al. 2012). Whether loss of CRISPR-Cas was adaptive is unclear (it may have hitchhiked along with an unrelated beneficial mutation), but other studies also suggest that the benefit of CRISPR-Cas is contingent on ecological factors (see Sections 4.2–4.6).

If we are to predict and manipulate the evolution of CRISPR-Cas immunity and the associated coevolution of CRISPR-Cas systems, we need to understand the selective forces that drive its evolution. Below we summarize the wide range of ecosystems where CRISPR-Cas systems have been found and then move on to discuss the role of four ecological factors that are predicted—and in some cases confirmed—to be important: (a) virus genetic diversity (Iranzo et al. 2013), (b) the force of infection (Westra et al. 2015), (c) defective phages (Hynes et al. 2014), and (d) the presence of mutualists (Bikard et al. 2012, Gandon & Vale 2014, Jiang et al. 2013, Levin 2010). Finally, we discuss a sequence-specific interaction between nonimmune bacterial hosts and viruses, known as priming, which has received a lot of attention, as it is key for the evolution of CRISPR-based immunity.

4.1. CRISPR-Cas Systems Are Widespread but Enriched in Thermophilic Environments

CRISPR-Cas systems are widespread and have moved extensively by HGT between different species (Cui et al. 2008, Godde & Bickerton 2006, Haft et al. 2005, Heidelberg et al. 2009, Held et al. 2013, Horvath et al. 2009, Tyson & Banfield 2008). Metagenomics studies reveal that CRISPR-Cas systems can be found in bacteria and archaea living in diverse environments, such as the oral cavity (Pride et al. 2011, van der Ploeg 2009), gut (Gogleva et al. 2014, Rho et al. 2012, Stern et al. 2012), rumen of cows (Berg Miller et al. 2012), biofilms in an acid mine drainage (Andersson & Banfield 2008, Tyson & Banfield 2008), ocean (Sorokin et al. 2010) and deep-sea areas (Smedile et al. 2013), hydrothermal vents in the northeast Pacific (Anderson et al. 2011), sludge bioreactors (Kunin et al. 2008), hot springs (Heidelberg et al. 2009, Held & Whitaker 2009, Held et al. 2010, Snyder et al. 2010), hypersaline lakes (Emerson et al. 2013), and low-oxygen cyanobacterial mats in the Middle Island Sinkhole in Lake Huron (Voorhies et al. 2015). However, consistent with the idea that ecology matters, CRISPR-Cas systems are not evenly distributed across these environments. In particular, high-temperature environments are enriched for CRISPR-Cas systems, with thermophiles typically having both more and longer CRISPR arrays (Anderson et al. 2011, Makarova et al. 2006).

4.2. Virus Genetic Diversity and Mutation Rates

Koonin and coworkers (Iranzo et al. 2013, Weinberger et al. 2012b) constructed theoretical models that suggested that differences between mesophiles and thermophiles in the rates of mutation fixation may explain why CRISPR-Cas systems may be more abundant in high-temperature environments. Specifically, both theoretical models predicted that the benefit of CRISPR-Cas adaptive immune systems decreases as virus genetic diversity increases, owing to the sequence specificity of virus recognition (Iranzo et al. 2013, Weinberger et al. 2012b). Virus genetic diversity covaries both with mutation rates and with population sizes (larger populations have more genetic diversity). The authors argue that high-temperature environments typically have lower mutation rates (Weinberger et al. 2012b) and lower microbial densities (Iranzo et al. 2013). Lower host densities directly cause a reduction in the density—and therefore genetic variation—of the virus population (Iranzo et al. 2013). Weinberger et al. (2012b) suggested that if carrying CRISPR-Cas systems was associated with a fitness cost (e.g., due to autoimmunity), the systems might be lost under conditions that correspond to high virus mutation rates as spacers are effective against only a small portion of the virus population. Such a constitutive cost of CRISPR-Cas was recently found to be associated with a Class 2 CRISPR-Cas system (Vale et al. 2015).

4.3. The Force of Infection

An alternative, mutually nonexclusive explanation why CRISPR-Cas systems are less beneficial and therefore less abundant at high microbial densities is that CRISPR-Cas–mediated immunity becomes increasingly costly as the frequency of infection increases (Westra et al. 2015). An inducible cost of resistance was independently observed in both P. aeruginosa (Westra et al. 2015) and S. thermophilus (Vale et al. 2015). These observations are consistent with studies that demonstrated that expression of many CRISPR-Cas adaptive immune systems is strictly regulated and specifically elicited upon infection (Agari et al. 2010, Quax et al. 2013, Young et al. 2012). In the P. aeruginosa system, high frequencies of infection (and therefore high inducible costs of resistance) were found to tip the balance from CRISPR-Cas–mediated immunity to surface modification–based defense, which is associated with a constitutive cost (Westra et al. 2015). Hence, these data suggest that CRISPR-Cas systems are more likely to confer a benefit in ecosystems associated with a low force of infection, such as high-temperature or low-resource environments, owing to the lower microbial and phage densities. At a high force of infection other immune mechanisms may be favored more than CRISPR-Cas.

At present the mechanistic basis of the observed fitness cost associated with CRISPR-Cas is unclear but may be related to autoimmunity (Bikard et al. 2012; Held & Whitaker 2009; Jiang et al. 2013; Paez-Espino et al. 2013, 2015; Stern et al. 2010; Vercoe et al. 2013) (see sidebar, The Protospacer Adjacent Motif) or allocating resources to defense that would otherwise be invested in growth.

4.4. Defective Phages

Recently, Hynes and coworkers (2014) provided compelling evidence that most of the spacer acquisition events in their S. thermophilus strain DGCC7710/phage 2972 experimental system occur in response to defective phages. This response is somewhat analogous to the role of defective viruses in triggering mammalian immune responses (Killip et al. 2015). The rate at which bacteria acquired CRISPR-based immunity also increased when the bacteria encoded both a restriction-modification (R-M)–based innate immune system and CRISPR-Cas (Dupuis et al. 2013, Hynes et al. 2014). In this case, R-M may supply CRISPR-Cas with phage genome cleavage products that can be integrated as novel spacers. These data also suggest that mounting a CRISPR-Cas adaptive immune response from scratch (i.e., bacteria going through the adaptation, expression, and interference stages; see sidebar, The Three Stages of a CRISPR-Cas Immune Response) may be too slow in the face of a highly virulent, rapidly replicating phage.

4.5. The Presence of Mutualists

In the presence of genetic parasites, such as phage, bacteria that carry CRISPR-Cas systems have a clear fitness advantage over strains that lack CRISPR-Cas (Westra et al. 2015). However, it is less clear how the benefits of CRISPR-Cas are affected by mutualistic mobile genetic elements, such as plasmids that can confer antibiotics resistance or prophages that encode virulence factors. Because CRISPR-Cas immunity is adaptive, one would expect the system to discriminate between parasitic and mutualistic DNA elements. Yet, two lines of evidence indicate that CRISPR-Cas systems form a barrier for HGT. First, bacteria evolve CRISPR-based immunity not only against phage but also against mobile genetic elements that are not necessarily parasitic, such as conjugative plasmids and conjugative transposons (Erdmann & Garrett 2012, Lopez-Sanchez et al. 2012). Consistent with this, Staphylococcus epidermidis strain RP62a carries a spacer that targets the nickase gene found in all sequenced staphylococcal conjugative plasmids (Marraffini & Sontheimer 2008). Second, correlational studies indicate that CRISPR-Cas systems indeed limit HGT (but see Gophna et al. 2015). For example, in Enterococcus faecium and Enterococcus faecalis, which are currently among the most important causative agents of hospital infections (Paganelli et al. 2012), the presence of CRISPR-Cas inversely correlates with their antibiotics-resistance gene content (Palmer & Gilmore 2010). These studies therefore suggest that the presence of mutualistic mobile genetic elements and environmental factors that impact their benefit (e.g., heavy metals, antibiotics, etc.) may affect the benefit of CRISPR-Cas. It is interesting to note in this context that a Type III CRISPR-Cas system does not target a prophage unless the target gene is transcribed (Goldberg & Marraffini 2015), which may facilitate a host interaction with a mutualist despite CRISPR-Cas immunity.

4.6. Immune Priming

A key factor that has been experimentally shown as crucial for CRISPR evolution is priming. In the context of CRISPR-Cas, priming refers to the presence of a partial match between a pre-existing spacer and the genome of an invading phage or plasmid (Datsenko et al. 2012, Swarts et al. 2012). Despite the mismatches, the CRISPR-Cas surveillance complex can still recognize the partially complementary target sequence (Blosser et al. 2015), which triggers recruitment of the spacer acquisition machinery (Redding et al. 2015) and results in rapid primed spacer acquisition (see sidebar, The Molecular Characteristics of Priming). The number of mismatches can be relatively high (up to 13 in one study; see Fineran et al. 2014), indicating that priming allows for some level of promiscuity.

Why did the requirement for priming evolve? The efficiency of unprimed (or naive) spacer acquisition is very low, indicating that the requirement for priming imposes a severe constraint on the evolution of CRISPR-based immunity. One possibility is that priming helps to limit autoimmunity problems associated with CRISPR-Cas through self-targeting (Koonin & Wolf 2016). In addition, priming also helps to ensure that only parasitic DNA elements that resemble previous spacers and therefore known parasites are targeted. Hence, although indiscriminate spacer acquisition upon infection may increase the rate at which bacteria acquire immunity, it may trade off with higher levels of self-targeting and greater constraint on the uptake of beneficial DNA.

Taken together, the findings described above make it increasingly clear that CRISPR-Cas may be highly beneficial in some environments but potentially costly or ineffective in other environments. Teasing apart the relative importance of the ecological variables outlined above and their evolutionary consequences will be key to understanding and manipulating bacteria–virus interactions in natural environments.

5. CRISPR-Cas–MEDIATED BACTERIA–PHAGE COEVOLUTION

What are the short-term coevolutionary consequences of CRISPR immunity against viruses? Several experimental studies have demonstrated that phage can overcome CRISPR immunity by mutating the target sequence or the protospacer adjacent motif (PAM) (see sidebar, The Protospacer Adjacent Motif) (Deveau et al. 2008). The high specificity of CRISPR–virus interactions leads to predictions of persistent coevolution (Agrawal & Lively 2002, Iranzo et al. 2013, Vale & Little 2010). Yet, CRISPR-specific models and recent coevolution studies reveal conditions in which CRISPR–virus coevolution is short-lived due to virus extinction (Iranzo et al. 2013, van Houte et al. 2016). What factors determine the type and duration of coevolution? Below we discuss the impact of: (a) spacer diversity, (b) viral mutation and host spacer acquisition rates, (c) the number of different viruses, and (d) anti-CRISPR genes.

5.1. Spacer Diversity

Theory predicts (Childs et al. 2014) and in vitro coevolution experiments show that CRISPR–virus coevolution is short-lived if the bacterial host population generates high levels of spacer diversity, which results in rapid extinction of the virus (van Houte et al. 2016). Crucially, although viruses rapidly evolved infectivity against nearly all individual clones in monoculture, viruses lost this ability when the same bacterial clones were mixed (van Houte et al. 2016). In addition to this evolutionary effect, diversity is also predicted to have an epidemiological effect. First, densities of the matching host genotype are reduced (dilution effect). Second, resistance of nonmatching host genotypes reduces the number of successful secondary infections (Lively 2010). This latter effect is amplified once an escape virus emerges, since the virus drives the frequency of its matching host genotype down (Lively 2010). The propensity to generate spacer diversity is thus an important fitness determinant of CRISPR-Cas and is dependent on priming because this accelerates spacer acquisition.

5.2. Viral Mutation and Host Spacer Acquisition Rates

As explained above, the virus mutation and spacer acquisition rates are predicted to be important for CRISPR–virus coevolution (Iranzo et al. 2013). Specifically, higher rates of mutation fixation are predicted to work to the advantage of the virus. Hence, increased mutation supply rates can move viruses away from extinction toward persistent coevolution or even to host extinction (Iranzo et al. 2013). Spacer acquisition rates work in the opposite direction and could lead to phage extinction (Childs et al. 2014, Iranzo et al. 2013). The reasons why the rate of spacer acquisition is important are twofold. First, it allows the host to renew immunity against escape viruses. Second, it increases spacer diversity in the population, which—as explained above—increases overall population resistance (van Houte et al. 2016).

5.3. Multiple Viruses

In many natural environments, a bacterium is likely to interact with multiple viruses, which may have a strong impact on CRISPR–virus coevolution. Theory and metagenomics data suggest that CRISPR-Cas systems may cause selective sweeps in the host populations if a single host genotype acquires resistance against two phages (Tyson & Banfield 2008, Weinberger et al. 2012a). Another effect of multiple phages, observed in coevolution studies using S. thermophilus, was that a phage may persist longer compared with a single-phage infection (Paez-Espino et al. 2015). Interestingly, it was found that the two phages recombined to escape CRISPR-Cas (Paez-Espino et al. 2015), which is consistent with previous metagenomics analyses (Andersson & Banfield 2008) and may help to explain commonly observed mosaic genomes of phage (Paez-Espino et al. 2015, Pedulla et al. 2003). The dynamics of CRISPR-Cas–mediated, single-host, multiphage coevolution awaits further study.

5.4. Anti-CRISPR Genes

The rapid extinction of viruses by the P. aeruginosa CRISPR-Cas system may have provided strong selective pressure on phages to evolve more sophisticated escape mechanisms. Many Pseudomonas phages encode so-called anti-CRISPR genes (Bondy-Denomy et al. 2013, Pawluk et al. 2014), which bind either the crRNA–Cas complex of P. aeruginosa (known as the Csy complex) to interfere with target DNA recognition or the effector nuclease (Cas3) to block target DNA destruction (Bondy-Denomy et al. 2015). Anti-CRISPR proteins are encoded by an extremely diverse set of genes often located in a conserved locus on phage genomes (Bondy-Denomy et al. 2013, Pawluk et al. 2014) and other mobile genetic elements (Pawluk et al. 2014, van Belkum et al. 2015). The evolution of anti-CRISPR genes likely occurs over much longer timescales compared with phage evolution to escape CRISPR-Cas by point mutation. The dynamics of these long-term coevolutionary processes have not been investigated, and if and how hosts can evolve to overcome the anti-CRISPR genes remains an open question.

6. CRISPR AS A BARRIER OF HORIZONTAL GENE TRANSFER: IMPLICATIONS FOR MICROBIAL ADAPTATION

Apart from short-term coevolutionary consequences, CRISPR-Cas can also impact long-term microbial adaptation. As explained above, CRISPR-Cas systems can form a barrier for HGT. Several correlational studies indicate that CRISPR-Cas limits gene transfer and that this results in important differences between genomes of CRISPR⁺ and CRISPR⁻ strains. For example, strains of the opportunistic pathogen P. aeruginosa, which often causes lung infections in cystic fibrosis patients, have significantly smaller genomes if they encode CRISPR-Cas systems compared with strains that lack CRISPR-Cas (van Belkum et al. 2015). Furthermore, CRISPR-Cas systems are absent in the genomes of species for which gene transfer is an important fitness determinant, such as Streptococcus pneumoniae, the causative agent of pneumonia that relies heavily on natural transformation for capsule switching during infection (Hatoum-Aslan & Marraffini 2014). CRISPR-Cas systems have been proposed to interfere with natural transformation (Bikard et al. 2012). Similarly, genomes of Streptococcus pyogenes, a major human pathogen that can cause, among other symptoms, pharyngitis, sepsis, and necrotizing fasciitis, often carry a range of prophage-encoded virulence factors that are key for S. pyogenes virulence, and a higher number of prophages negatively correlates with CRISPR-Cas (Hatoum-Aslan & Marraffini 2014).

CRISPR-Cas systems may be lost during evolution if they acquire resistance against beneficial mobile genetic elements. For example, when exposed to antibiotics, CRISPR-Cas mutants readily emerged in a clonal population of S. epidermidis that was CRISPR-resistant against an antibiotics-resistant plasmid (Jiang et al. 2013). Similarly, CRISPR-Cas mutants emerged in S. pneumoniae engineered to carry a CRISPR-Cas system from S. pyogenes that targets capsule genes (Bikard et al. 2012). Although in both studies the strains were engineered to target a mobile genetic element that was essential for survival, these studies highlight how selection may favor loss of CRISPR-Cas systems in microbial populations.

Taken together, accumulating evidence shows that CRISPR-Cas can impact microbial adaptation via HGT, and the need for HGT may select against CRISPR-Cas systems.

7. NONCANONICAL FUNCTIONS OF CRISPR-Cas SYSTEMS

In addition to adaptive immunity against invading genetic elements, there is increasing experimental evidence that CRISPR-Cas systems play roles in other cellular functions, including transcriptional control, stress response, and pathogenicity (see Westra et al. 2014 for a review).

The most commonly identified function of CRISPR-Cas systems other than adaptive immunity is the regulation of gene expression. One of the best-characterized examples of a directly selected noncanonical function of a CRISPR-Cas system is that of virulence regulation in the highly infectious zoonotic pathogen Francisella novicida. In F. novicida (strain U112), a Type II-B CRISPR-Cas system is involved in repressing production of the cell surface–associated bacterial lipoprotein (BLP). Because the host innate immune system recognizes BLP and is activated, reduction in BLP enables F. novicida to evade host detection and therefore increases virulence (Sampson et al. 2013). The repression of BLP involves the Cas9 nuclease, trans-activating crRNA (tracrRNA), and small CRISPR-Cas associated RNA, which degrade BLP messenger RNA (mRNA). Mutants that lack any of these CRISPR-Cas components demonstrate marked attenuation in virulence in mouse models (Sampson et al. 2013, 2014), and mutations in these BLP regulators correlate with mutations in the blp gene (Sampson & Weiss 2013).

Type II CRISPR-Cas systems have also been observed to modulate virulence in a number of other bacterial pathogens. Some strains of the enteric pathogen Campylobacter jejuni encode a Type II-C CRISPR-Cas system. Expression of Cas9 protein in C. jejuni strains in which this CRISPR-Cas system is absent leads to an increase in virulence. Further, mutated strains lacking cas9 showed reduced adherence, invasion, and attenuated cytotoxicity toward human gut cell lines (Louwen et al. 2013). Cas9 has also been implicated in the virulence of Neisseria meningitidis, which requires Cas9 for attachment to host lung epithelial cells and for invasion and intracellular replication (Sampson & Weiss 2013). However, the mechanism of action of Cas9 in these bacteria is unknown, but they may act in concert with other CRISPR components, as in F. novicida, to regulate virulence-associated genes.

In Legionella pneumophila, the causative agent of Legionnaires' disease, Cas2 (present within a Type II CRISPR-Cas system) promotes intracellular infection of host amoebae (Gunderson & Cianciotto 2013). Mutants in cas2 are impaired in their ability to cause infection, whereas mutants in any other part of the CRISPR-Cas system display no change. L. pneumophila Cas2 has been shown to have DNase and RNase activity, which is important in the establishment of intracellular infection in amoebae (Gunderson et al. 2015). The introduction of cas2 into a noncarrying L. pneumophila strain increased infectivity, and this introduction has been proposed to provide an advantage for acquisition and maintenance of cas2 within strains. The L. pneumophila Cas2 is currently the only Cas2 protein with a function exclusive from a role in adaptive immunity.

Another interesting case is that of the pathogen Listeria monocytogenes, a facultative intracellular bacterium and the causative agent of listeriosis. All strains of Listeria sequenced to date carry a CRISPR locus termed RliB. Notably, this locus is not adjacent to any cas genes and is also present in L. monocytogenes strains that do not carry any other cas genes. Overexpression of the rliB CRISPR transcript upregulates expression of a ferrous iron transporter (Mandin et al. 2007). Studies show that a mutant strain lacking rliB colonizes the liver of infected mice more effectively than wild-type strains, implicating a role in virulence regulation (Toledo-Arana et al. 2009).

In P. aeruginosa lysogens, disruption of a Type I-F CRISPR-Cas system affects both biofilm formation and swarming motility (Zegans et al. 2009). However, this effect is a by-product of cytotoxicity of CRISPR-mediated self-targeting (cleavage of the lysogenic phage, which is integrated in the host genome) (Heussler et al. 2015). In P. aeruginosa the primary function of this CRISPR-Cas system is in immunity, as shown by, among other indications, high levels of evolved CRISPR immunity in response to phage (Westra et al. 2015) and the evolution of anti-CRISPR genes in P. aeruginosa phage genomes (Bondy-Denomy et al. 2013). In brief, it appears from the examples above and others (Westra et al. 2014) that CRISPR transcripts and Cas proteins are both independently able to influence bacterial virulence via gene regulation. However, the majority of the mechanisms remain to be elucidated. A central question relating to the evolution of noncanonical functions of CRISPR-Cas systems asks if these are simply by-products of their role in immunity or if they are indeed selected functions. From the studies described above it appears that both occur.

8. APPLICATIONS OF CRISPR-Cas IN EVOLUTION AND ECOLOGY

CRISPR-Cas systems have emerged as tools across life sciences, and their applications range from strain typing to engineering genomes or regulating their gene expression. In addition to these applications, CRISPR-Cas are also emerging as model systems to examine host–parasite coevolution (Vale & Little 2010).

8.1. Ecological and Epidemiological Studies

CRISPR diversity results both from rapid acquisition of new spacers (Andersson & Banfield 2008) and from spacer loss (Held et al. 2010, Lopez-Sanchez et al. 2012, Pourcel et al. 2005, Schouls et al. 2003, Tyson & Banfield 2008). Because spacer acquisition is polar (i.e., new spacers are added at one end of the array) (Barrangou et al. 2007, Lillestol et al. 2006, Pourcel et al. 2005, Tyson & Banfield 2008), diversity in CRISPR loci is mostly localized in the area where novel spacers are integrated, the leader-proximal end of the CRISPR array (Horvath et al. 2008, Tyson & Banfield 2008, Weinberger et al. 2012a). As a consequence, trailer ends of CRISPR loci may be conserved between different microbial populations, whereas the middle may be population specific and the leader end unique at the individual level (Tyson & Banfield 2008). The highly variable sequence content of CRISPR loci can and has been exploited to distinguish closely related bacterial strains (Pourcel et al. 2005, Vergnaud et al. 2007). CRISPR-based typing can be done using hybridization-based methods, such as spoligotyping, which is based on hybridization of a CRISPR amplicon to known spacer probes, or using sequencing-based methods (for a review, see Shariat & Dudley 2014). Spoligotyping was developed as an early, rapid, and cost-effective CRISPR-based method to discriminate Mycobacterium tuberculosis strains and has been widely used over the past decades both for strain typing and to examine pathogen evolution and population structuring. M. tuberculosis has ceased to acquire novel spacers, and differences between strains are therefore predominantly due to spacer loss (van Embden et al. 2000). Spoligotyping has also been used for other bacterial pathogens, such as Corynebacterium diphtheriae, L. pneumophila, and Salmonella enterica (reviewed in Shariat & Dudley 2014). Pourcel and colleagues (2005) have directly used the sequence information stored in CRISPR loci for Yersinia pestis typing, and in a follow-up study the authors were able to link different strains to distinct geographical locations (Cui et al. 2008). CRISPR sequences have also been used to distinguish Erwinia amylovora strains, which are plant pathogens that are indistinguishable using other common strain typing techniques, and to study Salmonella outbreaks or phylogeny (reviewed in Shariat & Dudley 2014). CRISPR sequence–based typing coupled to multi-virulence-locus sequence typing (CRISPR-MVLST) has been used to identify outbreak isolates in patients (reviewed in Shariat & Dudley 2014). Finally, sensitive and specific real-time polymerase chain reaction–based methods using primers that anneal to CRISPR arrays have been developed to identify Shiga-toxin-producing Escherichia coli serogroups (Delannoy et al. 2012a, 2012b).

Apart from studying the host, CRISPR sequences can also be used to study mobile genetic elements. One key application of CRISPR spacer sequences has been their use in identifying virus and plasmid sequences in metagenomics sequencing data and linking the phage/virus to a specific host. The Banfield laboratory was the first to use CRISPR from metagenomics sequences of biofilms in an acid mine drainage to identify host–virus interactions (Andersson & Banfield 2008). Over recent years, this method has been used extensively to identify viruses and link them to a specific host (Anderson et al. 2011, Garrett et al. 2010, Minot et al. 2013, Sanguino et al. 2015, Stern et al. 2012). Furthermore, spacer sequences have also been used as probes in microarrays to monitor the abundance of specific viruses in environmental samples (Snyder et al. 2010).

8.2. Genome Editing

The requirement for a single Cas effector protein, rather than multisubunit crRNA–Cas complexes, makes Class 2 systems uniquely suited for genome editing technologies (see sidebar, Class 2 CRISPR-Cas Systems, for details). Nearly all genome editing and related techniques have been developed using the Cas9 enzymes (Class 2 Type II), but as explained in the sidebar, New Class 2 CRISPR-Cas Systems, novel Class 2 enzymes are currently being examined as well (Shmakov et al. 2015, Zetsche et al. 2015). At present, the Cas9 genome editing technology has been used across many different organisms, for example, to make transgenic crops and mutant mouse models, to modify human embryos, and to prevent or treat disease in animal models (reviewed in Sternberg & Doudna 2015). Cas9 cleavage sites are repaired either using the error-prone nonhomologous end joining DNA repair pathway, which results in loss of function due to sequence deletions or insertions, or using the homology-directed repair (HDR) pathway if a homologous DNA molecule with mutations of interest, which is used by the HDR machinery during repair, is supplied. HDR provides much greater control over the mutations that are introduced into the genome. Apart from editing, a catalytically inactive Cas9 mutant can be used to bind promoter regions to inhibit expression of target genes, or Cas9 can also be fused to other proteins, such as transcriptional regulators or fluorescent proteins to image specific genetic loci (reviewed in Sternberg & Doudna 2015). This technology is now widely used to examine gene function, but it can also be applied to examine specific evolutionary questions, such as the adaptive value of specific mutations and their interactions (e.g., epistasis). Furthermore, in the future the application of CRISPR-Cas9 in genome editing may prove to be very important in evolutionary genetics to confirm candidate gene function and to test genes' effects in ecological interactions. As such, CRISPR-Cas9 has clear potential as a tool in evolutionary ecology studies.

8.3. Manipulating Community Composition

Apart from sequence-specific editing of genomes or their transcriptional regulation, CRISPR-Cas can also be used to specifically manipulate species communities. For example, CRISPR-Cas9 has been used to target specific viruses (Kennedy & Cullen 2015), although rapid virus evolution can result in escape from CRISPR-Cas9 (Wang et al. 2016). In a microbial population, a CRISPR-Cas system encoded by a mobile genetic element, such as a virus or conjugative plasmid, can be programmed such that it kills one specific host genotype. In the in vitro laboratory environment, an sgRNA-Cas9–encoding phagemid (plasmid packaged in a phage capsid) could effectively target virulent Staphylococcus aureus genotypes while leaving nonvirulent genotypes unaffected (Bikard et al. 2014). A similar approach of using a virus-encoded Class 1 (Yosef et al. 2015) or Class 2 (Citorik et al. 2014) CRISPR-Cas system was used to target antibiotic resistance genes in E. coli. These proof-of-concept studies clearly demonstrate that CRISPR-Cas–based approaches can in principle be used to manipulate microbial community composition, although this will be much harder to achieve in real environments, where reaching sufficiently high infection rates of the phage is more difficult. Although these studies focused on important human pathogens and their virulence genes, the approach may be extended to examine specific functions of individual strains or species in complex microbial communities.

Apart from manipulating microbial community composition, it has been speculated that CRISPR-Cas can also be used as a tool to manipulate communities of sexually reproducing organisms. Specifically, Esvelt and colleagues (2014) proposed that sgRNA-Cas9 enzymes can be repurposed as gene drives to spread engineered traits through a population, which can in effect be used to alter community composition. The use of gene drives has been considered to control or reduce the spread of insect-borne diseases, but progress on this front has been hindered by technical difficulties associated with genome engineering. Esvelt et al. (2014) built a strong case that the use of sgRNA-Cas9 may overcome many of these problems and potentially open the way to ecological engineering. Indeed, Cas9-based gene drives have been developed that efficiently spread genomic alterations in populations of a number of organisms. Two proof-of-principle studies have demonstrated highly efficient Cas9-based gene drive activity in yeast (DiCarlo et al. 2015) and in fruit flies (DiCarlo et al. 2015, Gantz & Bier 2015), with homing efficiencies (the rate at which the drive gene is copied onto the opposite chromosome) of the gene drives reaching 99% and 97%, respectively. In two recent studies Cas9-based gene drives were successfully employed to spread traits in mosquitos that could potentially limit the spread of vector-borne diseases (Gantz et al. 2015, Hammond et al. 2016). Gantz et al. (2015) used a gene drive to spread antimalarial genes in Anopheles stephensi. Despite the large size of the gene drive used, efficiency reached 99%, although the effect was partially lost in the progeny of females due to instability of the homing element. A different approach to limit malaria parasite transmission was used in a study by Hammond et al. (2016), in which a Cas9 gene drive was used to target genes that are necessary for female reproduction in Anopheles gambiae populations. This gene drive was also efficiently transmitted to offspring but suffered from inactivation of the target genes in heterozygous females, thereby greatly reducing female fertility and hence gene drive transmission (Hammond et al. 2016). However, these studies show that Cas9-based gene drives can be highly efficient systems for ecological engineering of populations (Champer et al. 2016).

8.4. Model to Study Host–Parasite Coevolution

Many studies on host–parasite coevolutionary processes are constrained by a lack of knowledge about the mechanisms of resistance and infectivity. Focusing on the CRISPR-Cas system to study host–parasite coevolution greatly overcomes this problem because we have a deep understanding about the molecular mechanism by which it confers immunity (Vale & Little 2010). As such, CRISPR-Cas can be used as a model system to investigate general questions concerning the evolutionary ecology of host–parasite interactions. For example, bacteria–phage interactions have been used recently to examine when an induced defense, such as CRISPR-Cas, is favored over a constitutive defense, such as surface modification, and vice versa (Westra et al. 2015). Using this tractable experimental system, it was demonstrated that the force of infection determines the relative investment in the two arms of defense, which is in agreement with theoretical predictions (Hamilton et al. 2008). More recently, the impact of host resistance allele diversity on parasite persistence and evolution was examined. This study found that parasites (viruses) rapidly evolved infectivity against monocultures of bacterial hosts that all carried the same CRISPR spacer (targeting the phage), but phage was unable to evolve infectivity on the same clones when they were mixed together (van Houte et al. 2016). This study helps us understand how host diversity can limit the spread of infectious disease. We envisage that CRISPR–virus interactions will further emerge as an important model system for experimental evolution to answer key questions concerning host–parasite interactions.

9. CONCLUSION AND OUTLOOK

The field of CRISPR-Cas biochemistry has raced ahead, leading to groundbreaking applications in genome editing and beyond. However, many questions concerning the ecology and evolution of CRISPR-Cas remain. For example, it is unclear when CRISPR-Cas systems provide a selective advantage. There are many examples in which these sophisticated adaptive immune systems have been lost (Delaney et al. 2012, Sampson & Weiss 2013) or they appear inactive (Touchon & Rocha 2010). It is becoming clear that the benefit of CRISPR-Cas depends on a range of ecological variables, but much more work needs to be done if we are to understand and manipulate the evolution of CRISPR-Cas immunity in the lab, let alone in nature. Moreover, relatively little experimental work has been done on the coevolutionary consequences of CRISPR-Cas–virus interactions and the long-term consequences of CRISPR-Cas on microbial adaptation (reviewed in Hatoum-Aslan & Marraffini 2014). As the mechanistic details of CRISPR-Cas become better characterized, future research is likely to focus on these outstanding questions.

disclosure statement

The authors are not aware of any affiliations, memberships, funding, or financial holdings that might be perceived as affecting the objectivity of this review.