Clustered Regularly Interspaced Short Palindromic Repeats, or CRISPR systems, have been the subject of a substantial quantity of recent inquiry.1–6
These bacterial loci contain up to several hundred repeats of non-variable, somewhat palindromic sequences that separate stretches of variable spacer DNA. Both sequences are tens of base pairs in length and spacer sequences have been found to be homologous to “proto-spacer” sequences associated with certain phages and plasmids. This homology, in turn, is required for CRISPR-mediated interference with, for example, phage infection. CRISPR systems acquire new spacer sequences seemingly in response to phage or plasmid exposure and so, consequently, have been described as a form of bacterial adaptive immunity.
The term adaptation can be used to describe the molecular mechanism(s) of acquisition of CRISPR spacer sequences. In a broader sense, though, CRISPR adaptation entails the conversion of phagesensitive bacteria to phage-resistant ones. Such adaptation necessarily requires two perhaps distinct aspects. One consists of CRISPR adaptation in the sense of spacer acquisition while the other is the means by which bacteria avoid being killed, by phages, so that spacer acquisition can occur. That latter I will call CRISPR facilitation. I thus envisage CRISPR functioning against phages as requiring a progression of three essential steps: (I) facilitation of CRISPR adaptation, (II) actual acquisition of spacer sequences (CRISPR adaptation, as narrowly defined), and then (III) subsequent CRISPR interference with phage infection.
The CRISPR facilitation step might be accomplished by a variety of mechanisms including what I will describe as (1) “anticipatory,” (2) “deficient,” (3) “biased,” (4) “CRISPR-associated” or (5) “innate.” Keep in mind as I discuss these possibilities that they likely differ not just mechanistically but also in terms of their impact on the efficiency of CRISPR adaptation—as broadly defined—that is, the likelihood that a bacterium or bacterial population will succeed in achieving CRISPR-associated immunity against a given phage. This efficiency, in turn, may be relevant in determining what roles CRIPSR systems might play, ecologically, as just one of numerous bacterial antiphage defenses.7,8
- By “anticipatory” I mean that spacer sequences could precede bacterial exposure to non-self, proto-spacer-sequence-containing DNA, just as vertebrate adaptive immunity “anticipates” antigen exposure. However, given (a) the number of nucleotides making up spacer sequences, (b) the requirement for high identity between spacer and proto-spacer sequences for CRISPR functioning and (c) the finite size of CRISPR loci, it is likely that an anticipatory mechanism of adaptation inherently would be inefficient. For example, the odds of obtaining a specific, minimal,5 23-nucleotide CRISPR spacer sequence by completely random processes are one in 423 or approximately once every 1014 attempts. Most spacer sequences, though, are larger than this, ranging up to 72 nucleotides in length,6 and therefore are even less likely to form as a consequence of purely random processes. Nonetheless, this idea of anticipatory adaptation is not entirely infeasible given that bacterial populations can be relatively large, individual bacteria can contain hundreds of spacer sequences, and phage genomes can have many potential proto-spacer sequences. Still, and just as with the vertebrate adaptive immune system, an anticipatory strategy underlying CRISPR adaptation would require substantial amounts of sequence experimentation in conjunction with both positive and negative selection acting on those sequences, that is, selection for functionality along with avoidance of recognizing self.10 In particular, it would be quite rare for individual bacteria to happen upon, through purely random exploration, spacer sequences that are functional against a specific phages.
- By “deficient” I am referring to phage variants that are either genetically defective mutants or which are genetically wild type but nonetheless have become inactivated prior to infection. In both cases deficiencies must be sufficient that phages are no longer bactericidal but, so that their DNA may be exposed to CRISPR systems, not so deficient that phages are no longer infective. This is a mechanism suggested by Horvath and Barrangou,3 that is, their suggestion of a “possible involvement of degenerate infectious particles in building immunity.” Perhaps consistently, the Cas1 protein from E. coli K12 reportedly has an affinity for damaged DNA,6 where “Cas” stands for CRISPR-Associated Sequences (as discussed further below). Though a reasonable scenario, it is one that by necessity should imply that adaptation events would be limited to perhaps only a small fraction of phage infections.
- By “biased,” I mean that CRISPR adaptation may be more likely as a consequence of delays in bacterial killing, due either to specific circumstances or because of constitutively displayed properties associated with non-deficient invading DNA. Means by which bacterial killing may be delayed, or altogether eliminated, include (a) infections initiated by phages that go on to display lysogeny, (b) infection by phages under conditions where they initially display pseudolysogeny—in the Miller and Day11,12 sense of pseudolysogeny, that is, delays in the initiation of infections under starvation conditions; see also reference 12—or (c) bacterial exposure to foreign DNA that inherently displays delayed killing. The latter could include the DNA associated with plasmids since plasmids generally don't directly kill bacteria. Highly virulent phages, by contrast, typically will genetically kill their bacterial hosts quite early in the infection process: during the first few minutes post adsorption13 rather than delaying bacterial killing until the more obvious bacterial metabolic death that is seen at the point of phage-induced lysis.
- In addition to spacer and variable sequences, CRISPR loci also consist of cas genes, where, as noted, cas stands for CRISPR-associated sequences. These cas genes could potentially play a role in facilitation of adaptation. Such a role, however, would raise an important conceptual issue which is why would or indeed how could a bacterium field a CRISPR-associated, pre-adaptation mechanism of phage resistance in order to acquire a second, post-adaptation mechanism also of phage resistance? Logically, though, we could postulate that the first, hypothetically cas-associated mechanism might be inefficient and therefore only occasionally block otherwise bactericidal phage infections. Once adaptation has occurred, by contrast, CRISPR systems would then provide a more robust mechanism of resistance (though note that CRISPR systems, post adaptation, are not always highly efficient in their mediation of anti-phage interference14–18). As with “anticipatory,” “deficient” or “biased” facilitation mechanisms, most interactions between lytic phages and naïve bacteria presumably would still result in the death of the infected bacterium. With sufficient rarity of adaptation, CRISPR might not routinely serve as a first line of defense against specific phage types. Perhaps consistently, adaptive immunity does not generally serve even in vertebrate animals as a first line of defense against specific pathogens that have not been encountered previously.
- What does serve as the first line of defense against these pathogens, for vertebrates, are numerous mechanisms of non-adaptive but broadly acting innate immunity. CRISPR adaptation similarly could follow expression of innate mechanisms of bacterial resistance to phages,6 including ones that are not CRISPR associated. Unlike for potential facilitation mechanisms discussed above, if these innate mechanisms are reasonably robust then most phage-naïve bacteria might survive their initial exposure to phages rather than die. However, and once again, why possess seemingly redundant mechanisms of phage resistance? One possible answer is that even robust strategies of bacterial resistance to phages tend to be fallible.8,19 The principle utility of CRISPR interference thus could be protecting bacteria only should or only when their “primary” (innate) resistance mechanism(s) fail, with bacterial survival ultimately seen perhaps only in that subset of a bacterial population that has, in the meantime, managed to acquire appropriate CRISPR spacer sequences. Thus, one can envisage a temporal progression consisting of (i) exposure of a bacterium to a phage to which it is naïve, (ii) successful display of innate resistance mechanism against that phage, (iii) nonetheless successful CRISPR adaptation, (iv) subsequent re-exposure of the bacterium or its descendants to homologous phages, (v) occasional failure of the innate resistance mechanism upon such re-exposure and, despite this failure, (vi) ensuing bacterial survival due to CRISPR-mediated interference.
Such speculation begs the question of what bacterial function could play this role of innate, primary phage-resistance mechanism that is capable of facilitating CRISPR adaption. One possibility is the bacterial protection provided by restriction- modification (R-M) systems. Indeed, R-M systems, which are present in many bacteria, would seem to be ideal candidates for CRISPR facilitation: (a) They allow foreign DNA into cells, thereby permitting spacer acquisition. (b) When successfully implemented, R-M systems block subsequent DNA functioning, thereby providing phage resistance. (c) And, occasionally, R-M systems fail to protect not only individual bacteria but entire bacterial populations.8,19
There would be utility, therefore, for bacteria to posses some sort of anti-phage back up to R-M systems in the course of repeated bacterial exposure to specific phage types and CRISPR systems, seemingly, could serve in just such a role.
Which of these five general mechanisms actually facilitates CRISPR adaptation will, of course, only be determined via experimental analysis. There is no reason to assume, however, that any one mechanism might preclude others, or that a mechanism that functions in one system will function for all CRISPR-carrying bacteria. Furthermore, no matter how rare CRISPR adaptation might be, so long as CRISPR systems occasionally play a key role in protecting bacterial populations from phage-mediated extinction,20
then—as a consequence of such positive selection— specific CRISPR spacer sequences should still provide an “historical record” of bacterial interactions with specific phage types.9,21,22