We first consider the CRISPR-Cas systems of adaptive immunity. The recent experimental breakthroughs seem to have revealed all the genes that are responsible for DNA/RNA targeting [15
] and spacer integration [44
] by CRISPR-Cas. However, several genes that are stably associated with CRISPR-Cas loci still do not fit into this scheme [45
] (Figure A).
Figure 2 Organization of the genomic loci that encode prokaryotic immune systems including toxin genes. The core genes of CRISPR-Cas, RM, and DND systems in predicted operons are shown by pink arrows; genes with (predicted) toxin activity are shown by different (more ...)
The only two genes that are universal among the CRISPR-cas systems are cas1
both of which are implicated in the first stage of CRISPR-Cas function, adaptation, or spacer integration [44
]. However, despite the genetic evidence of the involvement of both genes, all the enzymatic activities required for adaptation appear to be provided by Cas1 alone [44
]. The Cas2 protein has been shown to possess a sequence-specific endoribonuclease activity [47
] and is a homolog of the VapDHi toxin of the VapDHi/VapX TA system from Haemophilus influenzae
]. Therefore it appears highly likely that like VapDHi, Cas2 is an mRNA interferase that specifically cleaves ribosome-associated mRNAs, an activity that is mechanistically irrelevant for the adaptation stage of the CRISPR-Cas function. Furthermore, certain Cas2 homologs occur independently of an antitoxin in predicted operons or in genomic contexts corresponding to mobile selfish elements that are able to mediate their replication in conjunction with transfer. In these cases, the Cas2-homologs are typically in the neighborhood of a gene encoding a resolvase (e.g. gi: 57504998, from Campylobacter coli
RM2228) or a Mob-type relaxase (gi: 313144877 from Helicobacter cinaedi
). Conceivably, the endoRNase activity of these Cas2 homologs might foster addiction to the respective mobile elements via toxin-like action upon disruption of the selfish element. Furthermore, the combination of the Cas2-like protein with a DNA resolving/cleaving enzyme is reminiscent of the CRISPR-Cas systems where Cas1, the partner of Cas2, plays the key role in the integration of the acquired spacer DNA. This parallel seems to support the potential origin of the Cas1-Cas2 dyad of the CRISPR-Cas systems from an ancient mobile element [47
] similar to the aforementioned elements that combine Cas2-like genes with genes encoding DNA resolving/cleaving enzymes. Therefore, we hypothesize that Cas2 retains its ancestral toxin-like EndoRNase function within the CRISPR-Cas systems but the interferase activity is kept in check through reversible inhibition by the Cas2-Cas1 interaction. This scheme is proposed as a direct analogy to the control mechanism of PrrC which is kept in its latent state via the interaction with the associated RM system.
Under this hypothesis, when CRISPR-Cas fails to contain virus growth due to viral counter-attack and/or the level of genotoxic stress increases due to the accumulation of nucleotide metabolites synthesized by viral enzymes, Cas2 is activated (possibly through degradation of Cas1) and abrogates translation, probably leading to cell suicide or dormancy (Figure ). The requirement of Cas2 for spacer integration might stem from regulation or stabilization of Cas1 via Cas1-Cas2 complex formation that also reversibly inactivates Cas2. Perhaps even more important, Cas2 might act at the initial stage of the CRISPR-Cas response by rendering the infected cell dormant and hence ‘buying time’ to allow the CRISPR-Cas system to integrate virus-specific spacers for effective future use in antivirus response. The strong sequence conservation of Cas2 across the CRISPR-Cas systems, together with its sequence-specific endoribonuclease activity, appear to be better consistent with “self” RNAs being targeted rather than non-self RNAs which would likely select for greater diversity of Cas2 [51
]. Of interest in this regard is the fusion of Cas2 with a 3′-5′ exonuclease RNaseH fold domain (e.g. LSEI_0356 from Lactobacillus casei
ATCC 334), which is present in several Type I-E CRISPR-Cas loci. In these fusion proteins, Cas2 appears to be inactivated as a result of substitution of the catalytic residues so that the ribonuclease activity is most likely supplied by the 3′-5′exonuclease-like domain [52
Figure 3 The immunity-dormancy/suicide coupling hypothesis: Route 1 – toxin action before immunity activation. The coupling between dormancy-suicide and immunity is specifically illustrated by the CRISPR-Cas system that is hypothesized to adapt by inserting (more ...)
Most of the Type I and Type III CRISPR-Cas systems encompass additional domains, besides Cas2, which could potentially function similarly to toxins (Figure A). Typically, these predicted toxins are either fused to or encoded in the same operon with proteins of COG1517 (also known as Csm6 and Csx1) that are common in CRISPR-Cas systems (Figure A). Most COG1517 proteins contain a distinct Rossmann-fold domain [5
] but many in addition contain a third, usually C-terminal domain. The domains fused to COG1517 proteins include REase fold DNases (Pfam Clan: PD-(DE)xK) [49
], as confirmed by a recently solved crystal structure (pdb code 1XMX), and HEPN domains ([53
] and (KSM, VA, EVK, LA, unpublished). The HEPN domain has been predicted to comprise a component of a distinct TA jointly with the minimal nucleotidyltransferase domain [25
]; recently, HEPN has been shown to function as a toxin [54
]. Our recent analysis suggests that the PrrC/RloC ACNase domain and the KEN domain of eukaryotic unfolded protein response/ antiviral RNAses (Ire1/RNAse L) also are distinct versions of the HEPN fold and accordingly that several HEPN domains are RNases with a toxin-type activity (KSM, VA, EVK, LA, unpublished). Additional notable architectures of this CRISPR-CAS associated Rossmann-like domain include fusions to distinct RNase domains with potential toxin activity such as the RelE (e.g. sll7062 from Synechocystis
sp. PCC 6803) and PIN (e.g. APE_2119.1, Aeropyrum pernix
]. Both these toxins have been shown to possess mRNA interferase activity [24
]. Most of the predicted toxin-like genes are located within or in a close proximity of predicted Type III CRISPR-Cas operons [18
Moreover, the widespread Cas4 protein, a REase fold nuclease that is part of most Type I CRISPR-Cas systems but for which no specific function in the CRISPR-Cas response so far has been established also might exhibit a toxin-like activity. In this respect three observations seem to be relevant: first, the aforementioned fusion of the REase fold nuclease domain with the Rossmannoid domain; second, the only CRISPR-Cas Type I-A locus so far detected in which the COG1517 protein (Csa3) does not contain a potential toxin domain encompasses Cas4 (KSM, unpublished); third, the fusion of Cas4 with Cas1 [18
]. The Cas4-Cas1 fusion may be considered in parallel with the fusion of Cas1 with the reverse transcriptase (RT) domain that was described previously [49
] and now appears to be widespread in a variety of bacteria (e.g. alr1468 from Nostoc sp
. PCC 7120, Franean1_1369 from Frankia sp
. EAN1pec, VVA1544 from Vibrio vulnificus
YJ016, etc.). The RT domain fused with Cas1 is related to the RT-like proteins of the AbiA and AbiK families that participate in the abortive infection response [56
]. Recently, it has been shown that AbiK is not a typical RT but apparently catalyzes non-templated synthesis of random sequence DNA that remains covalently attached to the protein and contributes to abortive infection via a still uncharacterized mechanism [57
]. We propose that the Cas1-associated RT functions in a similar fashion. Thus, within the CRISPR-Cas systems, multiple cases of replacement of one type of toxin by another seem to occur.
Similarly to CRISPR-Cas, defense systems that are best classified as innate immunity are also associated with suicide/dormancy genes. In addition to the PrrC-PrrI link discussed above, many other Type I and Type III RM systems encompass domains with potential toxin-like roles including different families of HEPN domains as well as Sir2, ParB and REase fold nucleases that are homologous to abortive infection system subunits and predicted toxins of TA systems [6
] (Figure B). Thus, in the RM system the same phenomenon of repeated toxin displacement as in the CRISPR-Cas systems seems to take place.
Although the specific functions of the proteins involved in DNA phosphorothioation and those that are involved in the restriction by this system are poorly understood, the sets of genes involved in both processes are relatively well-defined [11
] (Figure C). One of the genes (dndB
) that is part of the phosphorothioation operon is a negative regulator of the process and might play a role in discriminating between the modification of self and non-self DNA [13
]. The N-terminal domain of DndB belongs to the AbiU1/AIPR family of abortive infection proteins that are also linked to Type III RM systems [28
]. The AIPR proteins contain an N-terminal HEPN domain that is a toxin and a predicted RNase (see above). Another HEPN domain is present in the DndF protein that is implicated in the restriction process [11
Other, poorly characterized prokaryotic defense systems also might contain toxin domains (Figure D). In particular, it has been predicted that prokaryotic Argonaut protein (pAgo) is a key component of a distinct defense system that can be encoded by a stand-alone gene or within putative operons together with several nucleases [59
]. Notably, one of these nucleases is a close homolog of Cas4 that might function as a toxin in conjunction with CRISPR-Cas systems [59
]. In addition, other REase fold DNases as well as a predicted Sir2 superfamily enzyme are often found in association with pAgo [59
]. Thus, notably, (predicted) Sir2 superfamily enzymes are distinct components of the ABI system AbiH, and are linked both to Type I RM systems and to the putative defense system centered around pAgo (Figure A and D). These enzymes might either function as mono-ADP-ribosylating toxins targeting proteins or DNA [60
] or as nucleases as has been proposed for the Sir2 enzymes found in systems containing HerA-FtsK-like proteins [61
Most likely, there are additional, still uncharacterized innate and/or adaptive immunity defense systems in prokaryotes that are also coupled with toxins. One recently studied example is the phage and chemical stress resistance systems centered on the ter
]. Several operons of these systems combine genes encoding nucleases and helicases similar to those found in RM systems with genes encoding COG1517 family proteins as well as putative toxins related to those in typical TA systems. The latter proteins include Doc domains, which modify proteins by addition of NMP moieties, and predicted RNases homologous to RelE, barnase or the PIN domain. It seems likely that these systems also engender higher level functional cooperation between the restriction DNAses that directly target invading DNA and toxins that promote dormancy or suicide. These operons also share components with the Pgl system of reverse restriction-modification that encompasses many genes whose functions remain unclear although some of these encode domains suggestive of toxin-like activities [6
] (VA, L.M. Iyer, LA, unpublished).
We are now in a position to formulate the immunity-dormancy/suicide coupling hypothesis according to which immune systems and cell dormancy/suicide systems such as TA are functionally coupled and directly regulate each other’s functions (Figure ). Stand-alone suicide/dormancy modules, such as typical TA systems, lack a direct, demonstrable role in immunity and might even decrease the fitness of the host, especially when they exhibit their toxic activity in the context of plasmid addiction or restriction attack upon disruption of RM systems. However, the comparative genomics data summarized here indicate that all major antivirus immunity systems identified to date in prokaryotes are linked to proteins that are homologous to toxin subunits of typical suicide/dormancy modules and are predicted to possess toxin-like activities. These toxin-like proteins might not be mechanistically involved in the immune function but rather could act similarly to their toxin counterparts from stand-alone suicide/dormancy modules.
In most cases, the characterized and predicted toxin proteins in immune systems are highly variable both in terms of (potential) targets (i.e. DNA, tRNA, rRNA or mRNA) and mechanism of action (i.e. protein modification or nucleic acid degradation). However, if our prediction of the toxin-like interferase function for Cas2 is valid, the coupling between immunity and suicidal/dormancy response is an intrinsic feature of all CRISPR-Cas systems. Indeed, such coupling might be critical for the effectiveness of the adaptive CRISPR-Cas systems. Given that antivirus defense by these adaptive immunity systems depends on preliminary infection to integrate virus-specific spacer DNA, a dormancy response through the action of Cas2 is likely to help the host to ‘buy time’ to prime the immunity response. Under the current hypothesis, this is the first route of involvement of dormancy/suicide modules in coupled defense whereby the dormancy-like response precedes and enables the immune response (Figure ). The second route of immunity-suicide coupling follows an even more straightforward biological rationale: when an immunity system fails and/or the level of genotoxic stress increases through the activity of metabolic enzymes encoded by the infecting virus, the cell employs the associated toxins for drastic measures that involve abrogation of key cell processes, typically translation, leading to cell death or dormancy (Figure ).
The immunity-dormancy/suicide coupling hypothesis: Route 2 – toxin action after immunity failure. The coupling between dormancy-suicide and immunity is illustrated by the CRISPR-Cas system as in Figure .
Immunity systems do not encode typical antitoxins. Therefore (and by analogy with the PrrC model), we hypothesize that components of the immunity systems themselves (e.g. Cas1) function as reversible inhibitors of the toxic components, and that they are inactivated by genotoxic stress or viral counter-defenses neutralizing the immunity system and thus unleashing the toxin (Figures and ).