Bacteria use CRISPR–Cas systems to develop immunity to viruses. Details of how these systems select viral DNA fragments and integrate them into bacterial DNA to create a memory of invaders have now been reported. See Articles p.193 & p.199
Less than a decade ago, immunological memory was regarded as a feature unique to vertebrates — scientists ridiculed the idea that bacteria might be able to 'remember' viruses that attack them. Yet the almost inconceivable concept of bacterial immunological memory has since been shown to exist after all1,2,3. Two papers in this issue, by Heler et al.4 (page 199) and Nuñez et al.5 (page 193), published on Nature's website today, report major advances in our understanding of the molecular mechanism of this phenomenon.
Bacteria remember their viral invaders by sampling short DNA sequences known as protospacers from the viruses' genetic material. These sequences become integrated into the bacterium's own DNA, specifically into an array of repeat sequences called clustered regularly interspaced short palindromic repeats (CRISPRs; Fig. 1); the integrated sequences are called spacers. When a bacterium is subsequently attacked by a recognized virus, the spacers are transcribed from the array and used to guide a complex containing CRISPR-associated (Cas) proteins, which cleave protospacers in viral nucleic-acid molecules1.
Accidental destruction of the CRISPR array could occur if transcribed spacers guide Cas proteins to cleave it, leading to catastrophic degradation of bacterial genetic material. To prevent this potential autoimmunity, some bacteria have CRISPR–Cas systems that cleave DNA targets only if they are flanked by sequences known as protospacer adjacent motifs (PAMs)6. The repeat sequences that flank spacers in such CRISPR arrays lack PAMs and therefore cannot be cleaved (Fig. 1).
The mechanism by which spacers are chosen so that they target only PAM-associated protospacers has remained elusive. Heler and colleagues show that the Cas9 protein in two species of Streptococcus bacterium selects for spacers that have the correct PAM. Before now, the protein's main known role was cleaving targeted DNA.
When the authors exchanged Cas9 proteins for others that had different PAM specificities, they found that the PAM sequence of the acquired spacers changed accordingly. These CRISPR–Cas systems therefore efficiently use Cas9's ability to recognize PAM sequences for memory as well as for cleavage, instead of having dedicated memorizing proteins develop PAM recognition from scratch. In other types of CRISPR–Cas system, such as that found in Escherichia coli, the memorizing proteins have an intrinsic ability to select, at least partially, for PAM-encoding protospacers through an as-yet-unknown mechanism3.
Heler and co-workers went on to show that Cas9 is required not only for determining the PAM sequence of the acquired spacers, but also for integrating spacers into CRISPRs. This feature is peculiar to Cas9 — the proteins that cleave nucleic acids in other CRISPR–Cas systems studied are not required for integration3, but may enhance it under certain conditions7.
These findings add crucial details to the mechanism of molecular memorization revealed by in vivo studies3,7,8,9. Each spacer is integrated into the CRISPR array with a new repeat; it is known that newly integrated repeats maintain the sequence of the existing repeat on the other side of the new spacer3, and that the integration of new spacers probably occurs by separation of the two DNA strands of this repeat9. But more information is needed, and an in vitro system that allows further mechanistic details to be uncovered has long been awaited.
Nuñez and colleagues have established just such a system: it is composed of E. coli memorizing proteins, a supercoiled plasmid DNA as the spacer-acceptor molecule and a double-stranded (ds) DNA that serves as a spacer-donor molecule. The researchers first demonstrated the validity of their system by using it to corroborate many of the in vivo characteristics of the CRISPR memorization process. They went on to analyse high-throughput sequencing of spacers inserted in vitro, and show that the memorizing proteins integrate spacers in the correct orientation by recognizing a specific nucleotide base in the PAM.
Importantly, their system allows the spacer donor to be easily replaced with DNA that has different sequences, end modifications and strand compositions, and thus enables the influence of these features on spacer integration to be studied. In this way, Nuñez et al. show that hydroxyl (OH) groups at the 3′ ends of dsDNA substrates are essential for integration. On the basis of this requirement, and of characterization of intermediates identified in vivo9, the authors propose a highly plausible model for spacer insertion. In this model, the memorizing enzymes catalyse bond formation between the 3′ end of a preferred strand of the spacer and a particular strand at the end of a repeat. This is followed by the formation of another bond between the 3′ end of the complementary strand of the spacer and the complementary strand at the other end of the repeat (see Fig. 5 of the paper).
The strength of in vitro approaches to studying biological systems is that all the components are artificially added to the reaction; the requirements and features of each component can therefore be defined and manipulated. But differences from physiological activity may occur, stemming either from the use of a different chemical environment from that found in vivo or from the absence of regulatory elements. Such elements may not be essential for the generation of the end product of a reaction, but might have a key role in the physiological process.
Nuñez and co-workers report just such a difference. In vivo studies have revealed that spacer integration occurs predominantly at the first repeat of the CRISPR array3,7. By contrast, the authors observe that spacer insertion in their in vitro system is also distributed near other repeats, and even outside the CRISPR array. The researchers suggest that this might represent a physiological way of generating new arrays. This is a valid possibility, but regulatory elements in vivo or in physiological conditions probably often restrict this distribution and direct integration in a specific location.
The authors also report that PAM-encoding spacer donors are not preferred substrates for integration in vitro, as opposed to what has been seen in vivo3. Moreover, they observe that the length of integrated spacers may vary substantially, whereas spacers in naturally occurring arrays have a strictly defined length. These differences might be explained by the fact that the in vitro system simulates only the last stage of spacer integration; earlier steps in the natural process probably account for the PAM preference and for defined spacer lengths observed in vivo.
The differences in the in vivo and in vitro studies nevertheless highlight the cardinal question of what determines the constant length of newly acquired spacers in vivo. Is it dictated by a protein complex that hands the processed spacer to the memorizing enzymes? If so, then what are these proteins? An in vitro system composed of all of the elements that catalyse every step of the reaction is needed to address these issues.
PAMs prevent autoimmunity against the CRISPR array, but autoimmunity could also occur if the CRISPR–Cas system accidentally cleaves other DNA sequences. So how is this prevented? It is known3,10 that foreign DNA molecules are sampled by CRISPR–Cas systems more frequently than the host's chromosome. Future work should investigate the mechanism underlying this selective sampling.
Barrangou, R. et al. Science 315, 1709–1712 (2007).
Brouns, S. J. J. et al. Science 321, 960–964 (2008).
Yosef, I., Goren, M. G. & Qimron, U. Nucleic Acids Res. 40, 5569–5576 (2012).
Heller, R. et al. Nature 519, 199–202 (2015).
Nuñez, J. K., Lee, A. S. Y., Engelman, A. & Doudna, J. A. Nature 519, 193–198 (2015).
Mojica, F. J. M., Díez-Villaseñor, C., García-Martínez, J. & Almendros, C. Microbiology 155, 733–740 (2009).
Datsenko, K. A. et al. Nature Commun. 3, 945; http://dx.doi.org/10.1038/ncomms1937 (2012).
Swarts, D. C., Mosterd, C., van Passel, M. W. & Brouns, S. J. J. PLoS ONE 7, e35888 (2012).
Arslan, Z., Hermanns, V., Wurm, R., Wagner, R. & Pul, U. Nucleic Acids Res. 42, 7884–7893 (2014).
Nuñez, J. K. et al. Nature Struct. Mol. Biol. 21, 528–534 (2014).
About this article
Cite this article
Yosef, I., Qimron, U. How bacteria get spacers from invaders. Nature 519, 166–167 (2015). https://doi.org/10.1038/nature14204
Nucleic Acids Research (2017)
Computational prediction of CRISPR cassettes in gut metagenome samples from Chinese type-2 diabetic patients and healthy controls
BMC Systems Biology (2016)
Molecular Cell (2016)