Cellular organisms in the three domains of life are under constant onslaught of invading mobile genetic elements (MGE), such as transposons, viruses and plasmids. Many, if not most, of these diverse selfish elements insert into the chromosomes of the cellular hosts, either as an obligate part of their life cycles or at least occasionally, and in multicellular eukaryotes constitute a substantial proportion of the host genome. For example, sequencing of the human genome has shown that transposons or relics thereof amount to 35% to 50% of the genome [1, 2] and subsequent analyses have only revised these estimates upward [3, 4]. Even more strikingly, in some green plants, MGE-derived DNA seems to represent more than 70% of the genome [5, 6]. Although not as abundant as in eukaryotes, proviruses and other MGE constitute up to 30% of some bacterial genomes [7, 8]. The effects of MGE integration vary from beneficial (gain of new phenotypic traits, such as antibiotic resistance or toxin production) to deleterious (disruption or inactivation of essential cellular genes upon MGE insertion) [7, 9–11]. For most prokaryotic plasmids and viruses, the circular form of the MGE genome is inserted into specific loci (site-specific integration) of the cellular chromosome with the aid of MGE-encoded enzymes known as integrases [12]. The integrases are grouped into two major families on the basis of sequence conservation and mechanistic relatedness: (1) tyrosine recombinases use a catalytic tyrosine residue which attacks the DNA and becomes covalently linked to it during strand exchange; and (2) serine recombinases for the same purpose use a nucleophilic serine residue [12].

Transposons are DNA segments that move from one location in the host genome to another. Although several classification schemes have been proposed, based on the nature of the transposition intermediate, transposons can be generally grouped into two classes [10, 13] or types [14]. Class I (or type 2) elements—also known as retrotransposons—transpose via an RNA intermediate which prior to integration is copied back to the DNA form by the element-encoded reverse transcriptase. Class II (or type 1) DNA transposons move in the genome by the so-called ‘cut-and-paste’ mechanism whereby the transposon is excised from its initial location and inserted into a new genomic locus. Most of the class II transposons possess characteristic terminal inverted repeats (TIR) but differ widely in terms of the transposases they encode, the specific mechanisms of transposition, the element size and gene content [10, 13, 15]. Although most transposases belong to the DDE superfamily (named after two aspartate and one glutamate residues that form the catalytic triad of these enzymes) [10, 13, 16], some transposons encode transposases homologous to the rolling-circle replication initiation endonucleases found in numerous viruses and plasmids [17–19], to phage integrase-like tyrosine recombinases [20, 21] or to the serine integrases/invertases [22]. Furthermore, some bacterial and eukaryotic viruses encode transposases that are involved in the integration of the viral genome into the host chromosome, thereby partially blurring the distinction between different MGE types [23–25].

A distinct group of MGE consists of large (15 to 20 kb), self-synthesizing DNA transposons, called Mavericks or Polintons [26, 27]. The defining feature of Polintons/Mavericks is that they encode their own protein-primed type B DNA polymerase which is most likely involved in the transposon replication (hence ‘self-synthesizing’ transposons) [26]. In addition, these transposons encode several hallmark viral proteins, the genome packaging ATPase and protease. Recently, we have shown that Polintons/Mavericks also encode major and minor capsid proteins, suggesting that these elements combine features of bona fide viruses and transposons [24]. Polintons/Mavericks are widespread in diverse unicellular and multicellular eukaryotes [26, 27]. In contrast, no such self-synthesizing DNA transposons have been described in prokaryotes.

To survive the proliferation of various MGE and to maintain genetic integrity, cellular organisms have evolved numerous defense lines, including a variety of innate and adaptive immunity mechanisms [28–30]. Although once considered to be characteristic exclusively of animals, adaptive immunity has been recently discovered in bacteria and archaea [30–33]. This system consists of arrays of clustered regularly interspaced short palindromic repeats (CRISPR) and CRISPR-associated proteins (Cas) and elicits interference against foreign nucleic acids by degrading them in a sequence-specific fashion. The specificity is ensured by the unique spacers homologous to viral or plasmid DNA and integrated into the CRISPR loci. The action of the CRISPR-Cas system can be divided into three stages. The first stage, called adaptation, involves insertion of foreign DNA spacers into the CRISPR repeats. This step is mediated by the two most conserved core proteins of the CRISPR-Cas system, Cas1 and Cas2 [34–36]. Although the mechanistic details of adaptation remain poorly understood, it has been demonstrated that Cas1 is the endonuclease responsible for the excision of the protospacer from the foreign DNA and its insertion into the CRISPR cassette [36–40]. During the second stage, expression and processing, the CRISPR locus containing the arrays of spacers is transcribed, producing a long pre-crRNA (CRISPR RNA), which is subsequently processed by Cas proteins into short guide crRNAs. The final stage is called interference and involves degradation of the alien DNA or RNA by the Cas enzymatic machinery guided by the bound crRNA [30, 31, 35]. Phylogenomic analyses of the Cas proteins from diverse archaea and bacteria yielded a wealth of information on the diversity and evolution of the CRISPR-Cas immunity [33]. However, it remains unclear how this sophisticated defense system emerged in the first place.

Here, we describe the discovery and characterization of a new superfamily of MGE that possess several features resembling the eukaryotic self-synthesizing DNA transposons but are integrated in the genomes of various archaea and some bacteria. Along with family B DNA polymerases (PolB), that are related either to viral protein-primed PolBs or to typical archaeal PolBs, these elements, which we denote ‘Casposons’, encode Cas1 proteins of a distinct subfamily. We propose that, different from other known MGE, casposons utilize Cas1 endonucleases for integration into the host genomes via a mechanism resembling that of spacer integration by CRISPR-Cas systems. Given that Cas1 is a key enzyme of the CRISPR-Cas immunity and considering the deep branching of casposon homologs in the Cas1 phylogeny, casposons appear to have played a pivotal role in the origin of the adaptive immune system in prokaryotes.

Transposons as a type of MGE are polyphyletic with respect to the enzymes mediating their transposition [10, 13, 15]. Here, we described a new type of mobile elements, the casposons, which appear to rely on Cas1-like endonucleases for genome integration. Structures of several Cas1 proteins have been solved [37, 39, 40] showing that Cas1 proteins adopt a novel structural fold, unrelated to the folds of any of the transposases described to date. Nevertheless, casposons share a number of features with known DNA transposons. On the one hand, they appear to behave as site-specific transposons, akin to the bacterial transposon Tn7 [43, 44]. On the other hand, the molecular structure of casposons is highly reminiscent of the eukaryotic self-synthesizing DNA transposons of the Polinton/Maverick superfamily [26, 27]. Similar to the Polintons/Mavericks, casposons possess TIRs and encode their own DNA polymerase genes which in family 1 casposons belong to the same, protein-primed class as the polymerase of Polintons/Mavericks.

Based on the model previously proposed for Polinton/Maverick transposons [26], we hypothesize that casposon DNA replication proceeds via a single-stranded (ss) DNA intermediate and primarily depends on the casposon-encoded PolB (Figure 4a). First, during cellular DNA replication, the casposon sequence is likely to loop-out on the lagging strand due to the formation of a double-stranded stem involving the TIR sequences. The next stage involves Cas1-catalyzed excision of the casposon. Importantly, Cas1 from E. coli has been shown to act efficiently on different branched DNA substrates [37], including the splayed-arm duplex DNA which is similar to the looped-out casposon intermediate depicted in Figure 4a. The TIRs of the excised ssDNA casposon would form the panhandle structures which serve as the replication origin in various viruses and plasmids encoding protein-primed PolBs [59, 60]. In the latter systems, replication is primed by a MGE-encoded protein that covalently binds to the 5′-terminus of the nascent strand. This is likely to also be the case for the group 1 casposons which encode protein-primed PolBs (Figure 3). Terminal proteins are known to be highly divergent but are typically encoded immediately upstream of the polB genes or as N-terminal fusions of the PolB proteins [59–61]. All family 1 casposons share an appropriately positioned conserved gene which could encode a terminal protein (Figure 2). Family 2 and 3 casposons encode PolBs that are more closely related to the typical archaeal RNA-primed PolBs, suggesting that their replication is primed by the host primase. Eventually, a new double-stranded casposon copy is synthesized through the concerted activities of PolB and accessory host- and/or casposon-encoded replication proteins, such as helicases (Figure 4a).

Given the presence of a PolB gene in all casposons, we propose to classify these elements as the second superfamily within the class of self-synthesizing DNA transposons, in addition to the Polintons/Mavericks [14]. Most of the polintons show a distinct virus-like character, with two conserved genes encoding major and minor capsid proteins, suggesting that these elements form virus particles under some circumstances and prompting their proposed re-classification as polintoviruses [24]. The casposons, however, do not encode any detectable homologs of capsid proteins and accordingly are likely to adhere to the transposon lifestyle. We define casposons as self-synthesizing MGE which rely on Cas1 for integration. Under this definition, casposons may encode distinct DNA polymerases (as is indeed the case for family 1 compared to families 2 and 3) but should possess the ability to self-synthesize.

The Cas1 endonuclease is the key player in the adaptation step of the CRISPR-Cas immunity. Consistent with its importance, Cas1 is the most stable and conserved component of functional CRISPR-Cas systems and is considered a signature gene for these defense systems [32, 33, 35]. A model of spacer acquisition, which integrates the available experimental data, has been proposed [30, 34–36]. This model helps to predict the specific path of the Cas1-mediated integration of casposons which is also consistent with the detailed analysis of the terminal sequences of integrated casposons [see Additional file 1: Figure S2]. Figures 4b and 4c depict the parallel flows of events underlying the insertion of new CRISPR spacers and casposons, respectively. In both cases, staggered nicks are introduced into the target sequence which in the case of CRISPR-Cas corresponds to the first repeat proximal to the leader sequence. In the case of some casposons, the TIR-containing termini are processed to produce short overhangs complementary to the tips of the nicked target site (see above). In the next step, the ends of the protospacer/casposon are joined to those of the nicked target site. The observation that in CRISPR-Cas systems Cas1 is the only protein whose enzymatic activity is essential for integration of new spacers [62], a process that involves cutting and rejoining of the cellular DNA within the CRISPR repeat arrays, suggests that the casposon Cas1 also possesses both DNA cutting and joining activities. However, the latter activity of Cas1 remains to be demonstrated experimentally. Finally, the target site is fill-in repaired, completing the casposon/spacer insertion and resulting in the TSD (for casposons) or repeat duplication (in CRISPR-Cas) (Figure 4b, c).

The discovery of casposons has important evolutionary implications. The deep branching of casposon Cas1 homologs within the global Cas1 phylogeny (Figure 1) is compatible with the possibility that the Cas1 family of endonucleases emerged in the context of mobile elements and only later was adapted for cellular defense. Consequently, we propose that casposons played a pivotal role in the origin of prokaryotic CRISPR-Cas immunity. The origin of Cas1 appears not to be the only contribution of transposable elements to the evolution of CRISPR-Cas. Indeed, recent comparative genomic analysis of the type II CRISPR-Cas systems has shown that Cas9, the key protein of these systems involved in the RNA processing and interference stages, most likely, also evolved from a distinct class of transposon proteins [63].

It has been previously hypothesized that the CRISPR-Cas system originated in archaea [32], and the present observations on the likely archaeal origin of casposons appear compatible with this hypothesis. However, it cannot be ruled out that similar to some other MGE, casposons are even more ancient and antedate advanced cellular life forms [23].

Strikingly, transposons can also be placed at the root of adaptive immunity in eukaryotes. The RAG1 protein, which plays a central role during the V(D)J recombination, was derived from the DDE transposase of Transib transposons [64]. The parallel contribution of transposons to the origin of adaptive immunity in prokaryotes and eukaryotes emphasizes that MGE are the molecular architects behind some of the major evolutionary innovations of their hosts, in particular, the cellular defense systems [23, 65]. More specifically, given the mechanistic similarity between MGE transposition and integration, on the one hand, and insertion of spacers by the CRISPR-Cas system and immunoglobulin gene rearrangement, on the other hand, integrases and transposases appear to be ready-made tools that can be recruited and utilized by adaptive immunity systems.

The diversity of MGE is astounding and is far from being fully explored. This state of affairs is well illustrated by the discovery of casposons described here. Casposons constitute the second superfamily of self-synthesizing transposon-like MGE, beside the eukaryotic Polinton/Maverick transposons, and are the first representatives of this class of elements in prokaryotes. Different MGE have evolved a number of unrelated molecular mechanisms to perform similar tasks that ensure their propagation within the host cells. The casposons, so far, are unique as the only group of MGE that apparently rely on Cas1 endonucleases, a key component of the prokaryotic CRISPR-Cas defense system, for insertion into and excision from the host genome. The perennial arms race between cellular organisms and various MGE seems to be one of the major driving forces underlying the evolution of both interacting parties and it is becoming increasingly clear that cells and MGE exchange molecular inventions that emerge in the process of this evolutionary struggle. The adaptive immunity of both prokaryotes and eukaryotes apparently evolved via recruitment of recombinases from distinct MGE, the casposons and the Transib family transposons, respectively. Although this route of evolution seems paradoxical given that MGE are the primary targets of the immunity systems, it is becoming clear that throughout the course of evolution, MGE have served as a rich source of naturally evolved tools for cellular genome engineering that had a major impact on the diversification of cellular organisms.

Casposon protein sequences were analyzed using PSI-BLAST [66], CD-Search [67], and HHpred [68]. Inverted and direct repeats flanking the casposons were analyzed using Unipro UGENE [69]. The palindromic repeats within the casposon TIR sequences were identified using Mfold [70]. Insertion sequences were analyzed using ISfinfer [71]. Multiple sequences alignments were built using Promals3D [72] and Muscle [73]. The Polinton/Maverick PolB sequences were recovered from the Repbase Update database [74]. For phylogenetic analysis, gapped columns (more than 30% of gaps) and columns with low information content were removed from the alignment [75]. Phylogenetic analysis was carried out by using FastTree [76], with the Jones–Taylor–Thornton model of amino acid evolution and γ-CAT estimation of evolutionary rates across sites. The trees were visualized using MEGA6 [77]. For the Cas1 phylogeny, Cas1 protein sequences representing all major types and subtypes of the CRISPR-Cas systems were obtained from [33] and supplemented with the casposon-encoded Cas1 protein sequences. The Cas1 sequence alignment used to generate the tree is provided in Additional file 2.