Chromosomal targeting by CRISPR-Cas systems can contribute to genome plasticity in bacteria

Abstract

The clustered regularly interspaced short palindromic repeats (CRISPR) and their associated (Cas) proteins form adaptive immune systems in bacteria to combat phage and other foreign genetic elements. Typically, short spacer sequences are acquired from the invader DNA and incorporated into CRISPR arrays in the bacterial genome. Small RNAs are generated that contain these spacer sequences and enable sequence-specific destruction of the foreign nucleic acids. Occasionally, spacers are acquired from the chromosome, which instead leads to targeting of the host genome. Chromosomal targeting is highly toxic to the bacterium, providing a strong selective pressure for a variety of evolutionary routes that enable host cell survival. Mutations that inactivate the CRISPR-Cas functionality, such as within the cas genes, CRISPR repeat, protospacer adjacent motifs (PAM), and target sequence, mediate escape from toxicity. This self-targeting might provide some explanation for the incomplete distribution of CRISPR-Cas systems in less than half of sequenced bacterial genomes. More importantly, self-genome targeting can cause large-scale genomic alterations, including remodeling or deletion of pathogenicity islands and other non-mobile chromosomal regions. While control of horizontal gene transfer is perceived as their main function, our recent work illuminates an alternative role of CRISPR-Cas systems in causing host genomic changes and influencing bacterial evolution.

The CRISPR-Cas System

The antagonistic relationship between host and parasite leads to the rapid adaptive evolution of host defenses, which in turn results in the generation of parasite counter-defenses. Constantly faced with phage infection, bacteria have several genetic systems that limit phage reproduction (for detailed reviews, see refs. 1 and 2). One group of systems that are widespread among bacteria and archaea are the CRISPR-Cas (clustered regularly interspaced short palindromic repeat-CRISPR associated) systems. CRISPR-Cas systems provide a highly adaptable RNA-based interference strategy against a variety of invading foreign genetic elements, such as plasmids, phages, integrative conjugative elements (ICEs) and transposons (for recent reviews, see refs 3–5). In short, these systems provide an adaptive immune system to bacteria via sequence-specific genetic memory of previous encounters with horizontally acquired elements. CRISPRs are an array of nearly identical short repeat sequences separated by similarly sized spacer sequences. Spacers are derived from invading elements and act as templates for sequence-specific degradation of complementary targets (termed protospacers) in a mechanism with some analogy to eukaryotic RNA interference. In contrast, most CRISPR-Cas systems target dsDNA rather than RNA. Upstream of the array is an AT-rich leader sequence that contains elements for transcription and acquisition of new spacers. A cluster of cas genes encode the molecular machinery for the defense process. Together, the CRISPR-Cas system provides a robust and heritable adaptive immune system that bestows on bacteria the ability to withstand phage predation and control the acquisition of foreign genetic elements (for reviews, see refs 3–5).

CRISPR-Cas systems are highly diverse and are classified into three main types based on phylogeny and the conservation of signature Cas proteins.^6,7 The three main types (I–III) are divided into further subtypes. Although components vary across the three types of CRISPR-Cas systems, the general mechanisms of resistance are similar and involve three phases. Upon a first encounter with the invading element, a new spacer from the invader is introduced into the CRISPR array, typically at the leader end.⁴ Adjacent to the protospacer is a short nucleotide motif (known as protospacer adjacent motif or PAM), which is required for incorporation and subsequently for targeting.^8,9 The CRISPR array is transcribed into a precursor CRISPR RNA (pre-crRNA) and processed by specific Cas proteins (and in type II with a small RNA and host RNase III) into crRNAs, each containing one spacer flanked by portions of adjacent repeats.¹⁰ Finally, the crRNA is assembled into a Cas ribonucleoprotein complex that guides this degradative machinery to the invading elements (i.e. DNA for types I,¹¹ II¹² and III-A¹³ or RNA for type III-B¹⁴) to achieve protection.

CRISPR-Cas Chromosomal Targeting is Detrimental to the Host Bacterium

The CRISPR array has a unique ability to retain information of past invasions. This genetic memory enables bacteria to combat previously encountered threats. Several bioinformatic analyses estimate only ~2% of spacers have identity to sequence databases, with the majority matching to phage and plasmid genomes.¹⁵ Intriguingly, a minority of spacers match bacterial genomes in non-prophage regions. Ultimately, this suggests a phenomenon of self-targeting, where the CRISPR-Cas system targets the host rather than the invader in a process akin to autoimmunity.¹⁶ Studies have proposed that chromosomal targeting may result from accidental spacer incorporation and can either be detrimental¹⁶ or regulatory¹⁷; however, until recently the mechanism driving auto-immunity was not well understood. Furthermore, while CRISPR-based interference of phages and conjugative plasmids are proven dogmas,⁵ the significance of self-targeting is only beginning to be elucidated.

Pectobacterium atrosepticum (previously Erwinia carotovora subsp atroseptica) strain SCRI1403 contains a type I-F CRISPR-Cas system composed of Cas1, a Cas2-Cas3 fusion, Csy1, Csy2, Csy3, and Cas6f (previously Csy4) and 3 CRISPR arrays containing 28, 10 and 3 spacers, respectively (Fig. 1A). We previously demonstrated all CRISPR arrays are transcribed and processed into crRNAs by Cas6f.¹⁸ Furthermore, we have shown that a Csy interference complex is formed, composed of Csy1, Csy2, Csy3 and Cas6f.¹⁹ The Csy complex also interacts with Cas2-Cas3,¹⁹ which is likely to be recruited for target DNA unwinding and degradation.¹¹ Cas2-Cas3 forms a complex with Cas1. The recent evidence that Cas1 and Cas2 from the related type I-E system are involved in acquisition⁹ led to the proposal that the Cas1:Cas2-Cas3 complex is important for acquisition of new spacers.¹⁹ Therefore, Cas2-Cas3 is hypothesized to play crucial roles in both acquisition and interference steps.¹⁹

Figure 1. CRISPR-directed chromosomal targeting is toxic and various routes lead to escape. P. atrosepticum possesses a type I-F CRISPR-Cas system consisting of 3 CRISPR arrays (gray) and a core cas operon (blue). (A) Self-targeting spacers (e.g., spacer 6 in CRISPR2 matches the eca0560 gene of HAI2) can be randomly acquired that result in toxicity. (B) To directly test effects of chromosomal targeting, we engineered a tightly controllable vector for expression of CRISPR arrays that target the host chromosome. When induced, chromosomal targeting results in growth inhibition. Mutations in various components of the CRISPR-Cas system can lead to avoidance of self-targeting. (C) Native HAI2-targeting by spacer 6 of the CRISPR2 array is non-functional due to mutations in the protospacer adjacent motif (PAM). (D) When the cas operon is deleted, the engineered self-targeting spacers can no longer cause toxicity. Furthermore, mutations in the repeat sequences inhibited toxicity. (E) Protospacer targets can also be deleted to avoid targeting. These deletions can result in the loss of large genomic regions and can cause ICE expulsion or remodeling.

P. atrosepticum has a single spacer in CRISPR array 2, which is identical to a protospacer in the genome.^18,20 This led us to use this plant pathogen as a model to investigate the effects of chromosomal targeting by CRISPR-Cas.²¹ First, plasmids were engineered that allowed controlled transcription of an array composed of a CRISPR1 leader followed by four repeats alternating with three anti-gene spacers (Fig. 1B). We targeted two non-essential genes: expI, which encodes the N-acyl homoserine lactone synthase and lacZ, encoding β-galactosidase. Expression of these plasmids in P. atrosepticum was highly toxic compared with plasmids containing either non-self targeting spacers (scrambled controls) or no spacers. Tight repression of the engineered arrays alleviated toxicity and no deleterious effects were detected in a cas deletion strain. The toxicity associated with targeting these chromosomal genes demonstrated that chromosomal targeting is detrimental to the host.

Chromosomal targeting caused a 10⁵-fold reduction in the viable count of P. atrosepticum and this toxicity was not reversible upon shutting off crRNA production. DNA is the likely molecular target of the type I-F CRISPR-Cas systems, a hypothesis supported by our engineered arrays, which were designed to produce crRNA that could not basepair with the mRNA, yet could pair with the template strand of the DNA. In addition, fluorescence and transmission electron microscopy revealed that the few remaining viable cells were elongated, which is a phenotype indicative of DNA damage in the closely related bacterium, E. coli.²² Therefore, The strong selective pressure exerted by CRISPR-Cas chromosomal targeting raised the possibility that mutants would arise to avoid self-targeting.

Several Bacterial Adaptations can lead to Chromosomal Targeting Avoidance

The distinct phenotype of toxicity enabled the requirements for chromosomal targeting and the strategies for its avoidance to be examined in detail. Several features were assessed, such as spacer dosage, Cas-dependency and targeting fidelity. One self-targeting spacer within an array was sufficient to cause toxicity. Our genetic experiments however, clearly demonstrated that self-targeting can be avoided in numerous ways, consistent with bioinformatic predictions.¹⁶ First, the fidelity of the PAM sequence was required for interference, as a single mutation (5′-protospacer-GG-3′ to 5′-protospacer-TG-3′) eliminated targeting of expI (Fig. 1C). Furthermore, deletion of this protospacer or removal of the cas operon abolished toxicity (Fig. 1D). Finally, mutations in the seed sequence that governs efficacy of targeting also attenuate toxicity; however, mismatches within this region were tolerated. Manica and colleagues observed that CRISPRs targeting the host β-galactosidase gene in Sulfolobus solfataricus caused reduced growth, which was alleviated via recombination between the chromosomal arrays and the synthetic CRISPR.²³ This suggests that recombination between and within CRISPR arrays that share repeat sequences might be another route to evade toxicity. Recombination frequency is predicted to be similar for arrays with toxic or non-toxic spacers, with the exception that the strong pressure upon self-targeting would select for bacteria that have removed host-toxic spacers.

Overall, these observations demonstrated that a similar mechanism is used for chromosome targeting as for CRISPR-based phage or plasmid interference, and that similar mutations can lead to the targeting of chromosomes and mobile genetic elements (MGEs) being nullified. This indicates that chromosomal targeting will select for CRISPR-Cas mutations and their degeneration might explain the absence of CRISPR-Cas systems in some bacteria. Interestingly, certain Escherichia coli and Salmonella strains possess CRISPR arrays with spacers matching the associated cas genes, which are missing in these isolates. This is consistent with a self-targeting and target (cas) deletion scenario discussed later.²⁴ As mentioned earlier, a single spacer (spacer 6) in the CRISPR2 array perfectly matched with the chromosomal gene eca0560. ECA0560 is encoded by horizontally acquired island 2 (HAI2), one of 17 putative islands in the genome of P. atrosepticum SCRI1043.²⁵ Given the toxicity associated with self-targeting, the presence of this spacer raised the question of how P. atrosepticum had evolved to tolerate self-targeting of the gene. We showed that the deviation of the PAM sequence of spacer 6 from consensus resulted in a non-functional crRNA that was ineffective at targeting. This represents a simple mechanism of evasion that occurs naturally, requiring no large genomic (e.g., cas genes or operon) deletions.

Contribution of CRISPR-Cas to Genome Mosaicism and Plasticity

The demonstration that CRISPR-Cas systems can target the host genome, cause lethal effects and result in a range of mutations for avoidance led us to investigate the significance of chromosomal targeting for bacterial evolution. More specifically, we addressed whether elimination of the protospacer region in response to the strong selection of self-targeting by CRISPR-Cas could also cause large scale genome remodeling.

Spontaneous mutants arising following expression of the chromosomal targeting CRISPR arrays were examined for loss of the target sequence, indicative of protospacer deletion as a route to avoidance. Mutants resulting from targeting either expI or lacZ had lost the entire respective protospacer sequence and in some instances, this loss was associated with deletion of more than 50 kb of DNA flanking the target. These extensive deletions included lacZ, in addition to two large genes encoding non-ribosomal peptide synthetases predicted to be carried on HAI6.²⁵ Deletion of DNA from a putative HAI suggested that self-targeting might have a profound effect on the evolution of a host cell, especially as islands encode many of the characteristics required for adaptation to new and evolving niches.²⁶

In support of the hypothesis that self-targeting could result in significant genomic changes, targeting of eca0560 in HAI2 produced two different groups of survivors that had undergone genome rearrangements. The first group (65% of the survivors) had lost HAI2 (Class I), the deletions spanning the entire region between two direct repeats that delineate the island. HAI2 expulsion may have resulted from site-specific recombination of integration sites after DNA damage caused by chromosomal targeting. However, as HAI2 readily excises from the chromosome,²⁷ some of these mutants could also have arisen from loss of the excised form of the element generated as a consequence of site-specific recombination between the direct repeats in a subpopulation of cells. The remainder of survivors had internal deletions flanking the targeted region ranging from ~40 to 75 kb (Class II) (Fig. 1E), which resulted in ‘mosaic’ islands that were still able to excise from the chromosome. Intriguingly, seemingly non-transferable islands can still be mobilized by transfer of neighboring MGEs. For example, in Streptococcus agalactiae, an HAI unable to transfer by conjugation was mobilized in cis by an unrelated ICE.²⁸ Bacteriophage remnants can also be disseminated effectively to other cells, possibly through inter-prophage interactions.²⁹ Such interactions between bacteriophages, HAIs and other MGEs likely add to the impact of chromosomal deletions brought about by CRISPR-mediated self-targeting.

The abundance of HAI-related spacers implies a major role of chromosomal targeting by the CRISPR-Cas system in genome remodeling.^16,30 Yet, one would expect that such large deletions involving HAIs might not be tolerated in vivo due to the selective advantage conferred by such elements. For example, HAI2 encodes coronafacic acid and a topoisomerase III enzyme, which are both involved in virulence of the host on potato.^25,27 Retention of HAIs in bacterial genomes can also be enhanced by the presence of addiction systems such as toxin-antitoxin (TA) systems.³¹ Therefore, one key question to be addressed is how and why self-targeting spacers become fixed in a population. The answer may lie in the balance between the benefits and costs associated with generation of deletions within the genome or with recently acquired genetic material.

While HAIs and other MGEs can be considered beneficial by increasing fitness in a particular niche, the existence of H-NS (histone-like nucleoid structuring) proteins and functionally related factors devoted to the silencing of horizontally acquired genes suggests that foreign DNA can also be detrimental to host fitness.^32,33 Some H-NS can be carried in phage genomes³⁴ and in E. coli H-NS negatively regulates cas expression,³⁵ thus ensuring efficient transfer of MGE to recipient cells. Furthermore, MGEs appear to have adapted to reduce their effect on fitness of their host. For example, the conjugative plasmid pSf-R27 from Salmonella typhimurium avoids an impact on bacterial fitness by encoding sfh, a paralogue of H-NS, presumably to silence the plasmid genome.^36,37 Transcription of genes on SPI-2, an island in Salmonella, is also under tight regulatory control, as overexpression can attenuate virulence in mammalian cells.³⁸ Confocal imaging of Pseudomonas syringae pv phaseolicola colony development in bean has revealed that strains containing the HAI PPHGI-1 have reduced multiplication.³⁹ It is postulated that this fitness cost might be associated with the island not having been integrated into existing regulatory networks due to its recent acquisition. Such foreign DNA acquisition may lead to an initial loss in fitness of the bacterium due to the host needing time to adapt to the presence of the new island.⁴⁰ Similarly, the retention of another HAI, the ICEclc, is thought to have arisen by regulating the ICEclc integrase that mediates movement of the island, since excessive amounts of integrase are deleterious to Pseudomonas.⁴¹

The loss and acquisition of HAIs and other related MGEs has played an important role in shaping the genetic repertoires associated with adaptation to specific niches.^40,42 Moreover, HAIs are frequently mosaic, possessing features from multiple mobile genetic elements. The island in PAO1 encompassed by PA0821 to PA0826, for instance, includes homologs of three genes of φE125, a Burkholderia thailandensis bacteriophage.⁴³ HAIs may also appear degenerate due to accumulated deletions and rearrangements that result in the loss of one or more of the activities necessary for mobility. Indeed, for many of these elements, a mechanism of transfer cannot be inferred due to insufficient similarity with characterized MGEs. The importance of mosaicism is only now being realized. For example, genome reduction resulting from loss of HAIs appears to have been a key component in the evolution of Pseudomonas aeruginosa and its virulence. The P. aeruginosa genome is a mosaic consisting of relatively conserved genomic sequences with interspersed accessory regions.⁴⁴ Evolution of P. aeruginosa in the lungs of patients with cystic fibrosis is associated with loss of genetic information from within this accessory gene pool. In some cases, gene loss is associated with a gain of function that confers a selective advantage. For example, deletion of the antisigma factor MucA results in mucoidy,^45,46 while elimination of mutS causes a mutator phenotype.^47,48 Interestingly, mutS is located on an island in E. coli, and even though mutS mutants have observable growth defects, they are relatively common in wild populations. MutS is involved in mismatch repair and its loss results in greater homologous recombination in E. coli. It is possible that E. coli has evolved to readily lose mutS to ensure mutS mutants arise, providing a reservoir of variants in any population that can more readily acquire MGEs from divergent strains.⁴² The acquisition of chromosomally-derived spacers might be considered purely accidental. Interestingly, in the absence of interference, approximately 25% of all newly incorporated spacers are derived from the chromosome in E. coli.⁹ This could indicate that acquisition events are relatively frequent, but are usually not observed due to their associated lethality. It is tempting to speculate that as part of a “bet-hedging” strategy, evolution has selected for a particular incorporation error rate that places a selective pressure on a subpopulation of cells. Survival of members of this subpopulation can result in the generation of genomic variants, some of which can be more adapted to certain niches. Or, as Michael Syvanen says, “Evolution has selected for evolution itself.”⁴²

Engineering CRISPR-Cas Systems as a Tool for Molecular Biology

In addition to the fundamental role of CRISPR-Cas in genome evolution, these systems can serve as tools for molecular biology. Recently, several groups have repurposed the Cas9 nuclease to regulate gene expression in a variety of hosts.^49,50 The expulsion of HAI2 as a result of its targeting by an inducible and reconstituted CRISPR array demonstrates that CRISPR-Cas systems can also be engineered to delete both core chromosomal regions and MGEs such as prophages and genomic islands, even those stabilized by toxin-antitoxin loci. Jiang et al. have also recently enabled Cas9 to mediate precise gene mutations in Streptococcus pneumoniae and E. coli.⁵¹ Our current work has shown that CRISPR arrays can be engineered to be highly toxic and kill target cells in a sequence-dependent manner, which might have application for development of novel antimicrobial strategies.

Conclusion

In conclusion, CRISPR-Cas systems have likely played a significant role in the survival and evolution of bacteria. In addition to adaptive protection from foreign genetic elements, self-targeting of the CRISPR-Cas systems can mediate the expulsion of islands and genomic deletions that may influence fitness of the bacteria. At the same time, self-targeting might contribute to genome mosaicism, ensuring sufficient diversity within bacterial populations for rapid niche adaptation.

Glossary

Abbreviations:

CRISPR

clustered regularly interspaced short palindromic repeats

Cas

CRISPR-associated

MGE

mobile genetic element

PAM

protospacer adjacent motif

ICE

integrative conjugative elements

HAI

horizontally acquired island

Citation: L. Dy R, R. Pitman A, Fineran PC. Chromosomal targeting by CRISPR-Cas systems can contribute to genome plasticity in bacteria. Mobile Genetic Elements 2013; 3:e26831; 10.4161/mge.26831

Vercoe RB, Chang JT, Dy RL, Taylor C, Gristwood T, Clulow JS, Richter C, Przybilski R, Pitman AR, Fineran PC. Cytotoxic chromosomal targeting by CRISPR/Cas systems can reshape bacterial genomes and expel or remodel pathogenicity islands. PLoS Genet. 2013;9:e1003454. doi: 10.1371/journal.pgen.1003454.

Disclosure of Potential Conflicts of Interest

The authors have declared no competing interests exist.