Emergence of key players in eukaryotic chromatin protein complexes

Eukaryotes are distinguished from the two prokaryotic superkingdoms by their dynamic chromatin organized by histones with low complexity tails, which provides a veritable “ecosystem” for several protein-modifying and ATP-dependent remodelers (Allis et al., 2006; Kouzarides, 2007; Aravind et al., 2011; Iyer et al., 2011a). The mysterious origins of several of the unique components of eukaryotic chromatin have begun to considerably clear up with recent genomic data. SWI2/SNF2 ATPases, which had at least six representatives by the time of the last eukaryotic common ancestor (LECA), had already diversified to perform several distinct chromatin remodeling activities, such as sliding/ejection of nucleosomes, exchange of canonical nucleosomes with those containing alternative histones, or altering nucleosomal spacing (Iyer et al., 2008b; Hauk and Bowman, 2011; Hota and Bartholomew, 2011). Phylogenetic, domain architecture, and gene neighborhood analysis revealed that SWI2/SNF2 ATPases are superfamily II DNA helicases, which had their most extensive diversification as part of R-M systems and related systems that are likely to function as a defensive mechanism against bacteriophages (related to the phage growth limitation or Pgl system) (Iyer et al., 2008b) (Figure ​(Figure3).3). In phylogenetic trees, the eukaryotic versions are nested within the radiation of SWI2/SNF2 ATPases from prokaryotic selfish elements and were transferred on at least three independent occasions to eukaryotes (Figure ​(Figure3A).3A). The first of these transfers occurred prior to the LECA, and by the time of the LECA had proliferated to spawn at least six distinct lineages (Iyer et al., 2008b). The remaining two transfers occurred much later in eukaryotic evolution, and gave rise to the Strawberry Notch and HARP-like SWI2/SNF2 ATPases (Figure ​(Figure3)3) (Iyer et al., 2008b). Bacterial R-M systems contributed a second ATP-dependent chromatin remodeling enzyme to eukaryotes, the MORC ATPase, which contains a composite module comprised of gyrase, histidine kinase, and MutL (GHKL) and S5 domains (Iyer et al., 2008a). Analysis of R-M bacterial systems showed that they display a vast radiation of several different types of GHKL-S5 module ATPases, of which the MORCs form one distinct clade (Figure ​(Figure3B).3B). Given that basal excavate lineages, such as parabasalids and diplomonads lack MORCs, they appear to have been acquired by eukaryotes post-LECA, prior to the radiation of the large eukaryotic clade uniting animals, fungi, amoebozoans, and plants (Iyer et al., 2008a) (Figure ​(Figure4).4). Both the MORCs and the SWI2/SNF2 ATPases use ATP hydrolysis to catalyze DNA-unwinding or large-scale looping of DNA in aiding the restriction activity of the REases. This activity has been reused in a biochemically comparable, but functionally distinct, context to remodel protein-DNA contacts or facilitate higher-order looping in eukaryotic chromatin. In a similar vein, R-M systems might also account for the origin of the eukaryotic phosphoesterase enzyme TDP1, which hydrolyzes 3'-phosphotyrosyl bonds between DNA and the active tyrosine of topoisomerase Ib to release DNA from topoisomerase adducts (Gajewski et al., 2012). Sequence relationships of TDP1 suggest that it is likely to have been derived from HKD phosphoesterase domains found fused to SWI2/SNF2 ATPases in bacterial R-M systems (Iyer et al., 2006a).

Open in a separate window

Figure 3

Evolutionary relationships of various families of enzymes illustrating the origin of eukaryotic versions within radiations of systems involved in inter- and intra-genomic conflicts. Reconstructed phylogenetic trees are shown for (A) The bacterial radiation of the SWI2/SNF2 ATPases. (B) MORC-like ATPases and (C) The Double-psi beta barrel containing RNA polymerases. Certain clades with multiple families such as the eukaryotic SWI2/SNF2 ATPases, the Topoisomerase ATPase subunits, the cellular DDRP and eukaryotic RdRPs are collapsed into triangles for clarity. Illustrative domain architectures or gene neighborhoods are shown next to the leaf. Genes in gene neighborhoods are shown in block arrows with the arrow head pointing from the 5′ to the 3′ gene. Proteins and gene neighborhoods are labeled by the gene name and species name separated by underscores. The trees represent only the overall topology because they were obtained by a combination of conventional phylogenetic tree construction and structure-based determination of higher-order relationships.

Open in a separate window

Figure 4

A tree of the eukaryotic relationships illustrating the points of recruitment in eukaryotes in different functional systems of various domains from different biological conflict systems. With a eukaryotic tree as reference, the source and reconstructed point of transfer of various domains recruited from different conflict systems and symbiogenic events are shown. The transfers are shown as dashed arrows with the arrow head pointing to the ancestor in which the transfer is proposed to have taken place. The dashed lines are labeled either with a single gene or a set of genes enclosed in a box. The conflict systems are shown in the key at the bottom left.

Similar studies have shown that the DNA methylases of eukaryotes, which play an important role as encoders of epigenetic information that goes over and beyond the basic genetic information, also largely owe their origin to R-M systems and related methylation systems that protect prokaryotic genomes against restriction attacks by selfish R-M systems (Bestor, 1990; Iyer et al., 2011a). Both DNA cytosine (C5) and adenine (N6) methylases of eukaryotes appear to have been derived from bacterial R-M system and dcm methylases on more than 10 independent occasions (Iyer et al., 2011a). As none of the conserved eukaryotic lineages of DNA methylases can be detected in the parabasalids and diplomonads, it appears that the classical epigenetic DNA modification of cytosine was absent in the LECA. The primary conserved cytosine DNA methylase of eukaryotes, DNMT1, appears to have emerged only just before the time the heterolobosean-kinetoplastid clade branched off from the remaining eukaryotes, and phylogenetic analysis strongly supports its origin from a bacterial R-M system methylase-related to M.NgoFVII (Iyer et al., 2011a). Most other DNA methylases of eukaryotes can be attributed to comparable later acquisitions, primarily from other types of R-M systems. Recent discoveries have indicated that the reversal of cytosine DNA methylation in several eukaryotic lineages occurs via the action of Tet-JBP family of 2-oxoglutarate and iron-dependent dioxygenases (2OGFeDOs), which remove the methyl group through oxidative conversion to hydroxymethylcytosine and further oxidized cytosine derivatives that are then cleared by base excision repair (He et al., 2011; Iyer et al., 2011a). Interestingly, related enzymes, JBP1/2, catalyze the hydroxylation of thymine in the synthesis of base J, an epigenetic modification observed in kinetoplastids (Vainio et al., 2009). Prior studies on the evolution of 2OGFeDOs revealed that the eukaryotic Tet-JBP enzymes were derived from precursors encoded by caudate bacteriophages (Iyer et al., 2011a). Bacteriophages have been known to display a rich variety of DNA modifications, including hydroxymethylated pyrimidines, which enable them to evade restriction by different R-M systems in the host genome (Gommers-Ampt and Borst, 1995). Thus, the bacteriophage Tet-JBP enzymes appear to have first emerged as part of their counter-restriction strategy, and subsequently recruited to generating and erasing epigenetic marks on DNA upon being transferred to eukaryotes. Multiple studies have also revealed that not just enzymatic domains, but also specific DBDs found in eukaryotic chromatin proteins might have been acquired from bacterial R-M systems and replication apparatus of caudate bacteriophages. The SAD/SRA domain, which is a key player in eukaryotic chromatin as an epigenetic “reader” of hemimethylated cytosine marks, has been derived from the DNA-binding domain of REases from R-M systems that discriminate between hemimethylated and fully methylated sites (Iyer et al., 2011a). Likewise, the recently described HARE-HTH domain, which might have an important role in discriminating the DNA modification generated by the cytosine methylases, and the Tet/JBP enzymes has also evolved from bacterial R-M systems, where it is combined with several distinct REase domains (Aravind and Iyer, 2012). On the other hand, another DNA-binding domain, the HIRAN domain, which among other proteins is associated with the eukaryotic chromatin remodeling RAD5-type SWI2/SNF2 ATPases appears to have emerged from the replication apparatus of caudate bacteriophages (Iyer et al., 2006a).

In stark contrast to chromatin remodeling and epigenetic DNA modifications, enzymes catalyzing epigenetic modifications of proteins in eukaryotic chromatin appear to have extensively drawn from very different types of prokaryotic systems involved in inter-organismal conflict. Two key epigenetic modifications are acetylation of lysines and methylation of both lysines and arginines in histones and other proteins in eukaryotic chromatin (Allis et al., 2006; Kouzarides, 2007). Sequence comparisons show that the eukaryotic arginine methylases (PRMT) have been derived from within a bacterial radiation of peptide methylases (Aravind et al., 2011). The closest bacterial sister groups of the eukaryotic PRMTs are encoded in antibiotic-like secondary metabolite biosynthesis operons that also contain genes for peptide dioxygenases, non-ribosomal peptide synthetases and other peptide-oxidizing enzymes such as LSD1-related amine oxidases (Aravind et al., 2011). Bacterial PRMT domains are also incorporated as domains of gigantic antibiotic biosynthesis enzymes, such as anabaenopeptilide synthetase that synthesizes a peptide toxin of the cyanobacterium Anabaena (Rouhiainen et al., 2000; Aravind et al., 2011). Interestingly, the LSD1-like amine oxidases observed in these and other peptide antibiotic/toxin biosynthesis operons are also the precursors of eukaryotic histone demethylases that catalyze oxidative removal of methyl groups from mono- and di-methylated histone H3K4 (Allis et al., 2006; Kouzarides, 2007). All the remaining histone demethylases in eukaryotes belong to one large superfamily of 2-oxoglutarate-dependent dioxygenases known as the Jumonji-related dioxygenases (Iyer et al., 2010). These, along with LSD1, are absent in the earliest-branching eukaryotes such as parabasalids and diplomonads, and first appear as multiple paralogous copies just prior to the divergence of the heterolobosean-kinetoplastid clade from the other eukaryotes (Iyer et al., 2010). However, each of these multiple eukaryotic paralogous lineages have their own bacterial counterparts suggesting that they had already diverged in bacteria before being acquired. In bacteria, like LSD1, they appear in one or more copies in peptide antibiotic/toxin and siderophore biosynthesis operons (Iyer et al., 2010), where they are likely to catalyze multiple oxidative modifications of peptides as previously observed in the biosynthesis of penicillin and its derivatives (Liras and Demain, 2009). Thus, it is plausible that eukaryotes acquired multiple paralogous jumonji-related dioxygenases via the transfer of a single secondary metabolism gene-cluster with multiple versions of these enzymes. In eukaryotes, other than histone demethylation, they also radiated to give rise to enzymes catalyzing the last step in the generation of the eukaryote-specific tRNA^Phe modification, hydroxywybutosine, and protein asparagine hydroxylation (Iyer et al., 2010). In contrast to these, the histone H3K79 methylase Dot1 appears to have emerged from a methylase effector delivered by intra-cellular symbionts and is seen in diverse bacterial endo-symbionts/pathogens of amoeboid protozoans and metazoans, like Parachlamydia and Legionella (Aravind et al., 2011).

Thus, components from R-M and virus-restriction systems, viral replication apparatus, peptide antibiotic/siderophore biosynthesis systems and effectors of intra-cellular bacteria, which are exemplars of intra-genomic, inter-genomic and inter-organismal conflict systems, have been harnessed as progenitors of distinguishing components of eukaryotic chromatin.

Conflict systems and eukaryotic RNA metabolism

Eukaryotes are characterized by the unique RNAi system, which is typified by small RNAs (usually 23–35 nt in length) that perform a number of roles ranging from post-transcription gene regulation to regulation of chromatin structure (Allis et al., 2006; Grewal, 2010). Of these small RNAs, the siRNA-type RNAs are particularly important in gene-silencing, and might be amplified by a distinctive enzyme of this system, the RNA-dependent RNA-polymerase (RdRP), which can be traced back to the LECA (Salgado et al., 2006; Ruprich-Robert and Thuriaux, 2010; Iyer and Aravind, 2011). Sequence-structure analysis of the RdRP revealed that its two catalytic double-ψ−β-barrel (DPBB) domains are related to the catalytic domain found in the two largest subunits of the cellular RNA polymerases from all life forms (Salgado et al., 2006; Ruprich-Robert and Thuriaux, 2010; Iyer and Aravind, 2011). The search for RdRP cognates outside eukaryotes showed that they are prevalent in certain bacteriophages of firmicutes and also a variety of recently identified novel selfish elements in bacterial genomes (Figure ​(Figure3C)3C) (Iyer and Aravind, 2011). In these potential selfish elements they are often encoded alongside genes for different DNase domains such as those belonging to the REase and URI endonuclease fold, which might aid in the mobility of the elements (Figure ​(Figure3C).3C). The RdRPs might also be combined with RNAse H domain in the cyanobacterial versions suggesting that might function in the context of RNA-DNA hybrids (Iyer and Aravind, 2011). Furthermore, structural analysis of the RNA-polymerases with DPBB catalytic domains showed that the RdRP-like enzymes belonged to a radiation of single-subunit RNA polymerases encoded by variety of selfish elements, from within which the cellular multi-subunit versions emerged via fission of the two catalytic domain-containing segments of the single-subunit enzyme (Figure ​(Figure3C).3C). It appears plausible that these RdRP-like enzymes of intra-genomic selfish elements and bacteriophages primarily arose as enzymes that aided their mobility by potentially acting as primases enabling their replication (Iyer and Aravind, 2011). Upon acquisition by the eukaryotic lineage, prior to the LECA, the enzyme appears to have been recruited as a part of the RNAi systems for amplification of small RNAs. Interestingly, the RdRP is not the only nucleic acid polymerase that has been recruited to RNA metabolism from a prokaryotic selfish element. Recent studies on the domain architectures and sequence relationships of the most conserved splicing factor of eukaryotes Prp8, which is part of the spliceosomal catalytic center, has revealed that it has been derived from the polyprotein of a retroelement replete with the reverse transcriptase, “thumb” and RNaseH domains (Dlakic and Mushegian, 2011). However, in Prp8 the active site of the reverse transcriptase domain is disrupted, suggesting that it merely functions in a nucleic acid-binding capacity rather than as an active enzyme (Dlakic and Mushegian, 2011). It is conceivable that this retroelement was associated with the ancestral group-II introns that invaded the genome in the pre-LUCA period to give rise to the spliceosomal introns of eukaryotes.

On several occasions, components of yet another prokaryotic inter-organismal conflict system, namely the recently characterized polymorphic toxin systems, appear to have contributed to eukaryotic RNA-processing and modification systems (Zhang et al., 2011). In eukaryotes, small nucleolar RNAs (snoRNAs) are required for modification and maturation of rRNA in the nucleolus. In several eukaryotes certain snoRNA, like U16 and U86, are directly released from the introns encoding them by the endonucleolytic action of the EndoU RNase (Laneve et al., 2003). Sequence and structure analysis revealed that the EndoU RNase of eukaryotes is nested within a vast radiation of RNase domains that function as toxins in bacterial polymorphic toxin and related secreted toxin systems (Zhang et al., 2011). Thus, acquisition of the EndoU domain appears to have enabled eukaryotes to bypass splicing to directly release snoRNAs from introns. RNA-editing via deamination of cytosine and adenine has considerably expanded in eukaryotes and is observed not just in tRNAs but also in mRNAs and as part of a counter-viral strategy (Iyer et al., 2011b). The origins of certain divergent metal-dependent nucleic acid deaminase domains, such as those of the AID-APOBEC clade and the DYW clade, which catalyzes massive RNA-editing in plant chloroplasts and mitochondria, were rather unclear until recently (Zehrmann et al., 2011). Analysis of the polymorphic toxins revealed that one of the widely used toxin domains was the nucleic acid deaminase that had greatly diversified in such and related secreted toxins (Iyer et al., 2011b). Importantly, the origin of the both the DYW and AID-APOBEC-like deaminases could be placed within specific prokaryotic toxin groups (see below for details).

Prokaryotic conflict systems and protein-modifying enzyme and second messenger in eukaryotic signaling systems

Recent studies on the diversity of catalytic toxin domains deployed in bacterial polymorphic and related secreted toxins systems are also throwing light on the emergence of what were previously considered uniquely eukaryotic signaling systems (Figure ​(Figure4).4). One such is the polyADP-ribosylation system, which modifies aspartate, glutamate and lysine side chains in both cytoplasmic and nuclear proteins including histones, with profound effects on DNA repair, chromatin organization, telomere dynamics, centrosomal and mitotic spindle organization, and endosomal trafficking (Ame et al., 2004). The enzymes catalyzing this modification, polyADP-ribosyl polymerases (PARPs), can be traced back to the LECA, but their emergence in eukaryotes remained a mystery (Citarelli et al., 2010). The closest relatives of the PARPs are found among toxin domains of a toxin used in inter-bacterial conflicts delivered via a distinctive phage-derived, injecting secretory system known as the Photorhabdus virulence cassette (Hurst et al., 2004; Zhang et al., 2012). Related PARP domains are also found as effectors of intra-cellular symbionts/parasites of amoebae and metazoa such as Legionella drancourtii. Recently, a novel family of ADP-ribosyltransferases (ARTs), distinct from the PARPs, was identified, and typified by the Neurl4 protein of humans (De Souza and Aravind, 2012). These ARTs might have an important role in the organization of the eukaryotic centrosome among other processes. They also seem to have been derived from effectors delivered by endoparasitic bacteria, such as Waddlia (Hurst et al., 2004). The use of mono-ADP-ribosyltransferases by diverse bacteria as toxins in intra- and inter-specific conflicts (i.e., polymorphic toxins) and those directed at host proteins is well-known (Koch-Nolte et al., 2008; Laing et al., 2011; De Souza and Aravind, 2012). Indeed, other than the PARPs and Neurl4-like ARTs, the eukaryotes also possess several mono-ARTs which are nested within the radiation of bacterial toxin ARTs. Thus, on more than three occasions eukaryotes appear to have recruited the toxin ART/PARP domains as protein-modifying enzymes, with the event giving rise to the PARPs probably happening before the LECA (Figure ​(Figure4).4). While in bacteria these enzymes appear to largely function as toxins, in eukaryotes they appear to have been utilized to post-translationally modify proteins and provide an additional level of coding information (Koch-Nolte et al., 2008; Laing et al., 2011). Beyond the events spawning pathways that are widespread in eukaryotes, polymorphic and related toxin systems also appear to have contributed to the origin of signaling systems unique to certain lineages, such as metazoans. In addition to ARTs, other bacterial toxin domains utilizing NAD as a substrate have also been recruited to metazoan signaling. The ADP-ribosyl cyclase domain was previously observed only in animals (in the CD38 and CD157 proteins) and generates two messenger molecules, namely cyclic ADP ribose (cADPr) and nicotinic acid adenine dinucleotide phosphate (NAADP), respectively, from NAD and NADP (Guse and Lee, 2008). The former two nucleotides function as messenger molecules that induce calcium signaling pathways via the ryanodine receptors (Guse and Lee, 2008). The discovery of the ADP-ribosyl cyclase as a toxin domain in bacterial polymorphic toxins provides a potential explanation for the sudden origin of this signaling enzyme in animals (Zhang et al., 2012). Additionally, fungi too appear to have independently acquired this domain from bacteria, suggesting that it might have been recruited on more than one occasion in eukaryotic evolution (Zhang et al., 2012).

The Teneurin/Odd Oz proteins found in metazoans and choanoflagellates function as developmental regulators with a potential role in cell-surface adhesion in diverse processes such as cell migration, neuronal path finding and fasciculation, gonad development, and basement membrane integrity (Minet et al., 1999; Silva et al., 2011). These proteins appear to have been derived from a complete bacterial polymorphic toxin, with both the N-terminal RHS/YD repeats, which form a stalk and the C-terminal toxin domain that is a derived version of the HNH/EndoVII fold (Zhang et al., 2012). While the C-terminal toxin domain has lost its active site residues in the animal lineages, it is cleaved and secreted as a potential neuromodulator (Qian et al., 2004). On the other hand the N-terminal RHS repeats appear to play a role in adhesion between different Teneurin/Odd molecules, which is a key aspect of their cell-cell signaling function (Silva et al., 2011). Other than the toxin domains, certain other domains in eukaryotic signaling pathways have also been acquired from bacterial polymorphic toxin systems. The hedgehog signaling pathway is a eukaryotic signaling pathway initiated by the hedgehog proteins, which undergo autoproteolytic cleavage to release signaling messengers (Ingham et al., 2011). The HINT domain, which catalyzes this autoproteolytic cleavage in the eukaryotic hedgehog proteins, is likely to have been derived from the HINT domains commonly found in bacterial polymorphic toxins, where they apparently facilitate the autoproteolytic release of the C-terminal toxin domain into target cells (Zhang et al., 2011). In metazoans, hedgehog activates a down-stream signaling cascade in target cells to activate the transcription factor Gli (Ingham et al., 2011). The Suppressor of Fused (SuFu) protein tethers the Gli in the cytoplasm in the absence of the hedgehog signal to prevent constitutive activation. This SuFu protein of the animal hedgehog pathway also has its origin in bacterial polymorphic toxin systems, where members of the SuFu superfamily function as immunity proteins that neutralize a structurally diverse range of toxin domains (Zhang et al., 2011).

The eukaryotic ubiquitin system: origin and elaboration

One of the most remarkable features of eukaryotes is the ubiquitin system, which comprises of several parallel enzymatic cascades which ligate Ubiquitin or an Ubl protein to target proteins, typically on a lysine residue (Hochstrasser, 2009). These cascades are typified by an E1 enzyme, which activates the Ub/Ubl terminal COOH group by adenylation and trans-thiolation to transfer it to and E2 enzyme. The E2 enzyme may then either directly or via an E3 enzyme transfer the Ub/Ubl to the target protein. In eukaryotes, such modifications often target proteins for degradation via the proteasomal system, where the Ub/Ubl is first cleaved off and released by a JAB domain metallopeptidase (Kerscher et al., 2006). In addition to proteasomal degradation, Ub/Ubl modifications also alter the interactions, localization and biochemistry of the target proteins and are modulated by a series of peptidases (DUBs) that debiquitinate them (Burrows and Johnston, 2012). Until recently it was thought that the Ub-system was a purely eukaryotic innovation. However, multiple studies have shown that the antecedents of the Ub-system first emerged in prokaryotes as part of a dramatic radiation of Ubls and E1-like enzyme in operons for the biosynthesis of cofactors (e.g. thiamin and molybdopterin), cysteine, and peptide secondary metabolites such as siderophores, antibiotics/toxins and small molecule signals (Burroughs et al., 2011, 2012). A subset of these operons is highly mobile (i.e., widespread dispersal across distant lineages) and evolved features characteristic of the eukaryotic Ub-systems, namely the presence of E2 and sometimes E3 enzymes and the deubiquitinating JAB peptidase (Burroughs et al., 2011). The fact that these operons are mobile, and usually tend to couple the ubiquitinating enzymes with deubiquitinating JAB peptidases, presents parallels to the R-M systems (Iyer et al., 2006c). Like them these systems combine opposing actions in the modifying and de-modifying enzymes, and have no links to the metabolic enzymes that are typical of the operons with E1-like enzymes and Ubls that synthesize small molecule. Hence, we posit that these are potential selfish elements that act like the R-Ms, but at the protein level, by possibly destabilizing proteins through transfer of the Ubl and restoring the original protein by removal of the Ubl by the JAB peptidase. Since these Ub-like systems are closest to the eukaryotic versions, it is very likely that they were derived from them. On account of their mobility they are seen in several bacteria and certain archaea (e.g., the Caldiarchaeum) (Iyer et al., 2006c; Burroughs et al., 2011; Nunoura et al., 2011); hence, it is possible that eukaryotes might have acquired the precursor of their Ub-system either from their archaeal precursor or from endosymbiotic bacteria (Figure ​(Figure44).

The only DUB that is consistently observed in prokaryotic Ub-like systems is the JAB peptidase domain (Iyer et al., 2006c; Burroughs et al., 2011). Eukaryotes, however, possess several other DUBs, most of which belong to the papain-like peptidase fold and a few to the Zincin-like metallopeptidase fold (Iyer et al., 2004). Interestingly, papain-like peptidases (e.g., Otu-like peptidase domain) and Zincin-like metallopeptidases are frequently found among the toxin domains of effectors delivered by a range of intra-cellular bacteria (Loureiro and Ploegh, 2006). These were previously thought to be lateral transfers from hosts to their endo-symbionts/parasites, which are used to interfere with the host Ub-system (Lomma et al., 2010; Schmitz-Esser et al., 2010). However, recent studies on polymorphic toxin systems suggest that such peptidase domains are far more widely distributed in bacterial toxins and often among toxins of free-living bacteria deployed in inter-bacterial conflicts (Zhang et al., 2012). Hence, it seems more likely that they first emerged in bacteria as part of the polymorphic and related secreted toxin systems and were acquired by eukaryotes and recruited as DUBs in course of the development of the mitochondrial endosymbiosis (Figure ​(Figure4).4). Not surprisingly, these DUB-like peptidase domains are common among intra-cellular bacteria such as Wolbachia, Rickettsia and Odyssella, which are closely related to the mitochondrial precursors (Figure ​(Figure2).2). Indeed, these DUBs probably originally emerged as part of the strategy utilized by these bacterial endo-symbionts/pathogens that countered the immunity mechanism based on ubiquitination of target proteins. Interestingly, several of these papain-like DUB domains are also related to polyprotein-processing peptidases of eukaryotic RNA viruses and retroviruses (Iyer et al., 2004). It is conceivable that the emergence of the Ub-system in eukaryotes also provided a means for RNA viruses to escape constraints placed by the eukaryotic mRNA cap on internal translation initiation, by simply enabling translation of polyproteins that are then processed by the DUB peptidases. In course of viral evolution many of the DUB domains were probably incorporated into their own polyproteins to allow autoproteolytic processing.

Executers of apoptosis: multiple independent recruitments of domains from prokaryotic conflict systems

One of the simplest counter-pathogen strategies is regulated cell death or apoptosis, in which a cell might sacrifice its own fitness and prevent the pathogen from replicating within it. This typically works in situations where the inclusive fitness accrued from saving kin from infection might contribute to fixation of altruistic behaviors such as apoptosis (Aravind et al., 2009). Such mechanisms are likely to be further enhanced with the emergence of colonial or multicellular organization. Some of the simplest programmed cell death systems seen in bacteria are constituted by intra-genomic selfish elements. For example, in Escherichia coli a defective prophage produces a toxin known as Lit with a zincin-like metallopeptidase domain to cleave the elongation factor Tu and kill the cell when infected by the phage T4, thereby preventing the further spread of T4 to remaining cells in the colony (Snyder, 1995). Likewise, under conditions of starvation, when resources are limiting, chromosomally encoded toxin-antitoxin systems, such as the entericidin locus, mediate cell death in bacteria like E. coli and allow certain cells to survive and grow at the expense of kin that have undergone cell death (Bishop et al., 1998). Thus, the principle of the use of toxins as mediators of programmed cell death appears to be an ancient one (Jensen and Gerdes, 1995; Bishop et al., 1998). Although eukaryotes lack conventional toxin-antitoxin systems, the executioners of apoptosis resemble the prokaryotic toxins from these and other conflict systems in that they cleave or modify specific target proteins or permeabilize membranes in the cell committed to apoptosis. These effectors have been best studied in the animal lineage and include membrane-permeability regulators (the BCL2 superfamily), DNA-cleaving enzymes (e.g., the DNA fragmentation factor/CIDE), DNA-modifying enzymes (e.g., pierisin) and peptidases (e.g., the caspases) (Chou et al., 1999; Lugovskoy et al., 1999; Kanazawa et al., 2001; Riedl and Salvesen, 2007). Investigation into the provenance of these proteins has revealed multiple ancient connections to bacterial toxin systems. The core helical domain of the BCL2 superfamily (the first 6 helices) is specifically related to the translocation (T) domain of several host-directed toxins from distantly related bacteria such as the diphtheria, botulinum, tetanus and Vibrio toxins (Chou et al., 1999). The T-domain undergoes a pH induced conformational change to assume a BCL2-like structure, inserts into the endosomal membrane and transfers the catalytic domain of the toxin into host cytoplasm. Given its sudden emergence in metazoans, it is likely that it was derived from a bacterial toxin and recruited as regulator of the permeability of the mitochondrial membrane. In metazoans these domains diversified into anti-apoptotic versions, which prevent the release of cytochrome C from mitochondria and pro-apoptotic versions which foster its release (Chou et al., 1999; Riedl and Salvesen, 2007). From animals, the BCL2 superfamily was secondarily acquired by large DNA viruses that infect them, such as herpesviruses, poxviruses, iridoviruses and asfarviruses, and used as an anti-apoptotic effector to prevent hosts from using cell death as a defense against them (Iyer et al., 2006b). The T-domain of bacterial toxins appears to have been independently transferred to the fungus Metarhizium, where it appears to be utilized in multiple toxins directed against the insect host.

Among catalytic effectors of apoptosis, in metazoans the DFF/CIDE endonuclease catalyzes the genome fragmentation of DNA that is typical of apoptosis (Lugovskoy et al., 1999; Riedl and Salvesen, 2007). Structural studies had revealed that this domain contains an endonuclease domain of the HNH/EndoVII fold, but its origins remained unclear (Lugovskoy et al., 1999). Recent analysis of the bacterial polymorphic toxins revealed that a subset of them contains a toxin nuclease domain, which shares unique sequence signatures with the DFF/CIDE endonuclease domain to the exclusion of other representatives of HNH/EndoVII fold (Zhang et al., 2012). Here again, the relative abundance of the HNH/EndoVII fold among polymorphic and related toxin domains compared to its lone presence in DFF/CIDE, which is restricted to metazoans, points to an origin for the latter from a representative in the bacterial toxin systems. Pierisin-type ARTs are unusual enzymes that mediate apoptosis (thus far only known from lepidopterans) by ADP-ribosylating the N2 atom of guanine in DNA (Kanazawa et al., 2001). The lepidopteran pierisin-like ARTs are specifically related to the ART toxin domains found in certain bacterial polymorphic toxins and insecticidal toxins of insect pathogens, such as Bacillus sphaericus (Orth et al., 2011). This suggests that they were probably laterally transferred into lepidopterans from a bacterial symbiont or parasite, followed by their reuse as an apoptotic effector. In all the above examples the natural action of the bacterial toxins in disrupting or killing animal cells appears to have been harnessed as a mechanism to execute apoptosis.

Caspase-like peptidases are the central executers of apoptosis throughout eukaryotes and have been demonstrated to play a central role in cell death in animals, fungi, plants, and certain other eukaryotes (Aravind and Koonin, 2002; Riedl and Salvesen, 2007). Prior evolutionary analysis of the caspase-like superfamily revealed that they first diversified in bacteria into several clades such as the metacaspases, paracaspases and numerous other bacteria-specific lineages (Aravind and Koonin, 2002). Metacaspases were transferred to eukaryotes prior the LECA and are found in most eukaryotes (Figure ​(Figure4).4). Subsequently, in the animal lineage and in dictyostelid slime molds metacaspases were displaced by a second acquisition from bacteria, the paracaspases, which then radiated in animal to give rise to the classical caspases (Aravind and Koonin, 2002). This phyletic pattern suggests that paracaspases are effectively functionally comparable to metacaspases, as they have displaced them on more than one occasion. Interestingly, several bacteria, particularly endosymbiotic/parasitic alphaproteobacteria (e.g., Agrobacterium, Labrenzia, Bradyrhizobium) encode metacaspases and paracaspases with N-terminal signal peptides that are likely to be secreted into their hosts (Aravind and Koonin, 2002). Hence, these peptidases were possibly first used in regulating endoparasite/symbiont-host conflicts to modulate the immune response and cell death in favor of the intra-cellular bacterium. Consistent with this, recent studies in humans have shown that the paracaspase modulates the T-cell-dependent immune response by cleaving A20, a deubiquitinating enzyme involved in the process, and is required for prevention of cell death in diffuse large B cell lymphoma (Coornaert et al., 2008; Ferch et al., 2009). This suggests that caspase-like peptidases might have been acquired on multiple occasions in eukaryotic evolution from endosymbiotic bacteria, which were probably utilizing them to regulate the survival of their host cells. On a similar note, the GIMAP/AIG1-like GTPases, which are deployed by certain endo-symbionts/parasites (Figure ​(Figure2),2), could have given rise to the eukaryote representatives of this clade which are known to modulate both apoptosis and the immune response.

Thus, protein domains that originally diversified in prokaryotic conflict systems both as toxin and also as potential modulators of host defensive responses have had a notable effect on the evolution of apoptosis.

Origin of antigen receptor diversification mechanisms and mutagenic immunity mechanisms

Despite the enormous disparities in the immune systems of different eukaryotes, there are a few common strategies that are observed across most of them. These include the use of a relatively small number of families of protein domains as antigen receptors. Diversification of antigen receptors in most eukaryotes is a passive process of sequence divergence, probably under positive selection, within families of lineage-specifically expanded proteins (e.g., LRR proteins). However, in both jawed and jawless vertebrates two distinct and directed mechanisms for their diversification have been observed, namely recombination and active mutagenesis, which result in different populations of lymphocytes expressing different types of antigen receptors (Pancer and Cooper, 2006; Schatz and Swanson, 2011). In both jawed and jawless vertebrates the process of directed mutagenesis by DNA cytosine deaminases of the AID-APOBEC superfamily is utilized (Rogozin et al., 2007). Such mutagenesis is used either as a trigger for antigen gene-conversion, or for hypermutation or for antibody class-switching. Additionally, certain representatives of the AID-APOBEC family of cytosine deaminases are also major line of defense against retroviruses by mutagenizing their genomes by cytosine deamination (Chiu and Greene, 2006). Although AID-APOBEC-like deaminases were, until recently, thought to be restricted to vertebrates, sensitive sequence analysis showed that more divergent members exist in nematodes, cnidarians and several distantly related algal groups. Identification of these sequences helped establish that the fast-evolving AID-APOBEC deaminases have their ultimate origin in the toxin domains of polymorphic and related secreted bacterial toxins (Iyer et al., 2011b). Indeed, effectors with toxin domains most closely related to the AID-APOBEC deaminases are observed in the Wolbachia endosymbiont of the moth Cadre cautella and the plant pathogen Pseudomonas brassicacearum (Iyer et al., 2011b). Thus, these mutagenic deaminase domains, which were originally part of toxins deployed by bacteria, appear to have provided the basis for the unique mechanism for antigen receptor diversification in vertebrates. However, their role in anti-retroviral response suggests that they were probably initially recruited merely as mutagenic enzymes that targeted viruses (i.e., similar to the original toxin role but merely directed at viruses). Interestingly, several filamentous fungi show a lineage-specific expansion of related nucleic acid deaminases that also appear to have been derived from toxin domains of bacterial provenance (Iyer et al., 2011b). It is conceivable that these play a similar role as the counter-retroviral deaminases in potentially mutating cytoplasmic parasitic elements or preventing anastomosis by unrelated hyphae.

In jawed vertebrates, antibody and T-cell receptor diversity is generated by the action (V-D-J and V-J recombination) of a dedicated recombination apparatus comprised of two proteins Rag1 and Rag2, of which Rag1 is the catalytic subunit of the recombinase (Schatz and Swanson, 2011). The origin of the Rag1 recombinase in animals had remained mysterious until it was shown that their recombinase domain is related to the transposase domain of a distinct class of eukaryotic transposons known as the Transib elements (Kapitonov and Jurka, 2005; Panchin and Moroz, 2008). This transposase domain contains a distinctive version of the RNAseH fold and cleaves sites associated with the termini of these transposons, which show sequence relationships to V-D-J and V-J recombination sites. Thus, the Rag1 recombinase appears to have evolved from a “domesticated” selfish element whose recombinase domain and terminal recognition sites were reused as a mechanism to generate diversity. Indeed, domestication of selfish elements for generation of diversity in host-pathogen interfaces is a general phenomenon, which is not restricted to the animal immune system: In certain caudate bacteriophages, the mutagenic reverse transcriptase of an integrated retroelement has been shown to play a role in creating sequence diversity in a tail-fiber-associated protein (Medhekar and Miller, 2007). This allows the bacteriophages to recognize a changing landscape of cell-surface proteins on their hosts.