Unusual evolutionary history of the tRNA splicing endonuclease EndA: Relationship to the LAGLIDADG and PD-(D/E)XK deoxyribonucleases

Abstract

The tRNA splicing endoribonuclease EndA from Methanococcus jannaschii is a homotetramer formed via heterologous interaction between the two pairs of homodimers. Each monomer consists of two α/β domains, the N-terminal domain (NTD) and the C-terminal domain (CTD) containing the RNase A-like active site. Comparison of the EndA coordinates with the publicly available protein structure database revealed the similarity of both domains to site-specific deoxyribonucleases: the NTD to the LAGLIDADG family and the CTD to the PD-(D/E)XK family. Superposition of the NTD on the catalytic domain of LAGLIDADG homing endonucleases allowed a suggestion to be made about which amino acid residues of the tRNA splicing nuclease might participate in formation of a presumptive cryptic deoxyribonuclease active site. On the other hand, the CTD and PD-(D/E)XK endonucleases, represented by restriction enzymes and a phage λ exonuclease, were shown to share extensive similarities of the structural framework, to which entirely different active sites might be attached in two alternative locations. These findings suggest that EndA evolved from a fusion protein with at least two distinct endonuclease activities: the ribonuclease, which made it an essential "antitoxin" for the cells whose RNA genes were interrupted by introns, and the deoxyribonuclease, which provided the means for homing-like mobility. The residues of the noncatalytic CTDs from the positions corresponding to the catalytic side chains in PD-(D/E)XK deoxyribonucleases map to the surface at the opposite side to the tRNA binding site, for which no function has been implicated. Many restriction enzymes from the PD-(D/E)XK superfamily might have the potential to maintain an additional active or binding site at the face opposite the deoxyribonuclease active site, a property that can be utilized in protein engineering.

RNA maturation involves RNA splicing in all three domains of life. Three different mechanisms are involved in removal of group I introns (Cech 1990); group II, group III, and spliceosomal introns (Michel and Ferat 1995); and Archaeal introns and Eukaryotic tRNA introns (Phizicky and Greer 1993). The first two mechanisms involve two consecutive transestrification steps, which are essentially RNA catalyzed, suggesting that group I, II, III, and spliceosomal introns have their origin in a hypothetical "RNA world" (Newman 1997). Splicing small tRNA introns in Eukaryota and Euryarchaea, and tRNA and rRNA introns in Crenarchaea, differs fundamentally from the other classes, because they are removed by the stepwise action of three protein enzymes: an endonuclease, a ligase, and a phosphotransferase (Lykke-Andersen et al. 1997).

The tRNA splicing endonuclease cleaves the pre-tRNA substrate, leaving 5′-hydroxyl and 2′,3′-cyclic phosphate termini. Then the exons are joined by a nucleoside triphosphate-dependent RNA ligase. Finally, the splice junction 2′-phosphate is removed by an NAD-dependent phosphotransferase (Abelson et al. 1998).

The crystal structure of the tRNA splicing endoribonuclease (RNase) EndA from Methanococcus jannaschii was recently determined (Li et al. 1998). The EndA monomer consists of two distinct α/β domains: the N-terminal domain (NTD) takes the shape of a saddle formed by a five-stranded mixed β-sheet flanked on one side by three α-helices. The C-terminal domain (CTD) folds into a topologically different five-stranded mixed β-sheet cradled between two α-helices. Two pairs of subunits in the EndA tetramer are nonequivalent, suggesting that only two of the four proposed active sites are functional. Three amino acid residues, corresponding to Y¹¹⁵, H¹²⁵, and K¹⁵⁶ in M. jannaschii enzyme, are conserved in catalytically active EndA/Sen family members (Trotta et al. 1997). The corresponding side chains can be spatially superimposed with the catalytic triad of RNase A, an enzyme that is unrelated to EndA, although it seems to carry out a mechanistically similar reaction, arguing for a case of convergent evolution (Li et al. 1998).

LAGLIDADG and PD-(D/E)XK are mutually unrelated families of deoxyribonucleases (DNases), of which biological functions have been regarded as overlapping to only a limited degree, and which show remarkable dissimilarities of function at the molecular level (Belfort and Roberts 1997; Aggarwal and Wah 1998). The LAGLIDADG family comprises relatively closely related proteins, of which some are composed of two tandemly fused versions of the conserved domain (Dalgaard et al. 1997). In contrast, the PD-(D/E)XK superfamily groups together extremely diverged enzymes and exhibit strong variations in sequences, structures, and biological functions (Bujnicki 2000).

Traditionally, the LAGLIDADG family has been grouped together with functionally similar, although evolutionarily unrelated HNH/His-Cys (vel ββαMe) and GIYYIG families, and juxtaposed against the largest and best characterized functional division of the PD-(D/E)XK superfamily, namely type II restriction endonucleases (ENases) (Belfort and Roberts 1997; Jurica and Stoddard 1999). In the presence of divalent metal ions, all the abovementioned enzymes perform a site-specific cleavage of phosphodiester bonds in two strands of DNA at fixed locations within a specific nucleotide sequence, yielding 5′-phosphate and 3′-hydroxyl termini. In contrast to type II restriction enzymes, which recognize 4–8 base pair palindromic sequences (Pingoud and Jeltsch 1997), the homing ENases recognize extended sequences of up to 20 bp, which are generally nonpalindromic (Jurica and Stoddard 1999). This is indicative of their different molecular associations: The homing ENases are usually encoded by introns and inteins, and by making a double-strand (ds) break in the intronless (or inteinless) allele, they promote a gene conversion process that duplicates the intron or intein in the so-called homing event. This event is analogous to the transposition of a transposon, in which, however, in contrast to the transposase, the homing ENase does not interact with its parental DNA, but acts only at the target site (Jurica and Stoddard 1999). Homing ENase genes abound in all three domains of life, with the largest number encoded in organellar genomes. These elements are highly invasive. It is believed that they colonized self-splicing introns and inteins (genetically silent loci, whose disruption is not usually dangerous for the cell), providing them with the potential for genetic mobility (Belfort et al. 1995; Derbyshire et al. 1997).

Traditionally, restriction enzymes have been implicated in the defense against invading genetic elements like plasmids or phages because they might function as weapons-destroying foreign DNA by cleaving it at multiple sites (Bickle and Kruger 1993). They are usually accompanied by DNA methyltransferases (MTases) of similar specificity, which methylate the DNA of the host, thereby rendering it resistant to the cleavage (Wilson and Murray 1991). Although these ENase-MTase pairs (termed restriction-modification or RM systems) might be advantageous to their Prokaryotic hosts, there is growing evidence that many of them exhibit "selfish" behavior (Kusano et al. 1995). RM systems commonly undergo horizontal transfer using plasmids, phages, and other mobile genetic elements as carriers (Jeltsch and Pingoud 1996) and act as toxin-antitoxin systems to become fixed in the new hosts (Kobayashi et al. 1999). It is worth emphasizing that the ds breaks introduced in the unmodified DNA by the restriction enzyme (for example, in the yet not fully methylated chromosome of the new host) might serve a recombinogenic role similar to that of the ds breaks introduced by the intron-encoded homing ENase, promoting integration of the DNA encoding the selfish element into the cut site at some frequency (Eddy and Gold 1992).

Results: Sequence and structure analysis

We conducted extensive database searches using the sequence and structure of EndA to obtain better insight into the lack of its overall similarity to RNase A and to learn more about the evolutionary origin of the family of tRNA splicing enzymes. Iterative psi-blast (Altschul et al. 1997) searches of publicly available databases failed to reveal any significant similarities of the EndA sequence to proteins other than the limited group of close homologs from Archaea and Eukaryota (data not shown). However, comparison of atomic coordinates of the individual NTD and CTD domains with the protein structures from the Protein Data Bank or PDB (Bernstein et al. 1977) using vast (Gibrat et al. 1996) and dali (Holm and Sander 1993) revealed them to be similar to the catalytic domains of the LAGLIDADG and PD-(D/E)XK DNases, respectively. It is noteworthy that none of the folds of other RNase families has been found to be even remotely similar, supporting the notion of convergent evolution between the EndA and RNase A families. None of the structural similarities reported here could be detected using any of the popular threading servers available (data not shown).

Both algorithms recognized structural similarity of the EndA CTD (residues 83–179) to various PD-(D/E)XK enzymes, including the phage λ exonuclease (λ-exo, PDB entry: 1avq) and the cleavage domain of the FokI restriction enzyme (2fok) among the best hits reported (vast score 11; dali z-score 5.0; rmsd for the fit of the C-α atoms is 2.9 å over 78 residues for λ-exo and vast score 10.1; dali z-score 5.1; rmsd 2.6 å over 66 residues for FokI). The NTD (residues 9–82) showed similarity to the LAGLIDADG enzymes, with I-CreI (1af5) reported as the most similar structure by both methods (vast score 9.7; dali z-score 3.1; rmsd 2.2 å over 42 residues). To the best of our knowledge, these similarities have not been analyzed previously.

In the EndA-NTD, the core of the open curved β-sheet structure used by I-CreI and presumably all LAGLIDADG ENase as the DNA-binding interface (Jurica et al. 1998), is perfectly conserved (Fig. 1A ▶). However, the NTD of EndA does not have the long variable loops that extend the recognition sites of the homing ENases. This suggests that the putative DNA-binding site, even if it were present in EndA, would be minimal compared to the typical LAGLIDADG ENases, unless the second domain contributed to the binding surface, as has been shown for the protein splicing domain of PI-SceI intein (Duan et al. 1997; He et al. 1998). Considering the alpha helices of EndA-NTD that shield the opposite face of the β-sheet, the orientations of α1 and α2 (nomenclature adapted from Li et al. 1998) are somewhat different from their equivalents in LAGLIDADG Enases. α3 has a spatial counterpart only in I-CreI, so the reported similarity might be dismissed as a chance convergence on a small folding motif. However, close examination of the structures reveals that EndA-NTD possesses a triad of residues (Thr¹¹, Asp²³, and Lys⁶²), which superimposes quite well with the catalytic triads of dimeric I-CreI (Asp²⁰, Gln⁴⁷, and Lys⁹⁸) (Jurica et al. 1998) and pseudodimeric PI-SceI (Asp²¹⁸, Asp²²⁹, and Lys³⁰¹; and Asp³²⁶, Thr³⁴¹, and Lys⁴⁰³) (Duan et al. 1997; Christ et al. 1999). The only exception is Asp²²⁹ in one of the two PI-SceI active sites. We suspect that this residue takes part in the catalysis using Ser²²⁷ as a part of a charge relay system (Fig. 1A ▶). It is therefore possible that EndA possesses cryptic ENase activity. However, it is not obvious whether (and how) the NTD could dimerize to put the two intramolecular putative active sites together. Remarkably, the counterpart of the LAGLIDADG motif, which forms a dimerization interface in the structurally characterized family members (Heath et al. 1997), and which is not conserved in the tRNA splicing enzymes, is partially disordered in the crystal structure of EndA (Li et al. 1998). It would be interesting to determine if EndA (or the separately expressed NTD) binds DNA and exhibits any kind of DNase activity, and if it shows any sequence specificity or preference, because it might indicate that the function of the NTD is (or was before it degenerated) to mobilize the EndA gene or the intron itself.

Fig. 1.

Comparison between the three-dimensional architecture of the EndA domains and their structural homologs shown in cartoon representation: (A) the structural NTD (left) and the I-CreI endonuclease (right); (B) the catalytic CTD (left) and the conserved core of λ-exo (right). Common secondary structure elements are colored, with blue and red indicating N- and C-terminals, respectively. Helices of the NTD are labeled according to Li et al. (1998). A black arrow indicates the N-terminal α-helix of I-CreI responsible for dimerization; the corresponding region is disordered in the crystal structure of EndA (broken line). For clarity, other nonconserved peripheral elements of both domain pairs are not shown and insertions are not colored. The catalytically important side chains of I-CreI, λ-exo, the CTD, and the putative "cryptic" DNase active site of the NTD are shown in wireframe representation and labeled.

The more closely related RNase (EndA-CTD) and DNase (PD-[D/E]XK ENase) domains show no evidence of mutual sequence conservation, and their respective catalytic residues do not overlap in the structure-based multiple alignment (data not shown). Structure superposition reveals that these two families of site-specific nucleases use different parts of the identical framework to create active sites that are suitable for performing two mechanistically distinct reactions: one leaving tRNA half-molecules with 5′-OH and 2′,3′ cyclic phosphate termini, and the other yielding DNA with 5′-phoshate and 3′-OH termini (Fig. 1B ▶). Although we were not able to identify any candidate for an evolutionary intermediate in the current protein sequence database, which would possess amino acids indicative of the two active sites simultaneously present at both sides of the putative catalytic domain, the extensive structural similarity by itself suggests that PD-(D/E)XK ENases and the EndA RNases originated from a common ancestor or from one another.

The closest structural similarities of EndA-CTD are to λ-exo, which represents an outgroup, branching early in the evolutionary history of the PD-(D/E)XK ENase superfamily, and to FokI, which is also believed to be quite ancient (Bujnicki 2000), suggesting that the EndA family is very old. In contrast, whereas the PD-(D/E)XK ENases are highly divergent at the sequence and structural levels, the sequences of EndA family members are relatively similar to one another (Trotta et al. 1997), suggesting that their divergence began relatively late. The question of mutual relationship of the two nuclease families with alternative active sites must wait a more extensive phylogenetic study of all proteins assuming the PD-(D/E)XK fold, which would considerably exceed the limits of this paper.

Discussion

It is remarkable to find that two domains, structurally similar to the catalytic domains of the LAGLIDADG and PD-(D/E)XK ENases, make up a nuclease implicated in yet another kind of toxin-antitoxin system, which parasitized Archaeal and Eukaryotic genomes by virtue of disrupting essential tRNA genes with an intron and providing a specific splicing apparatus as an instant remedy. It can be hypothesized that whereas group I and group II introns were equipped with intrinsic autosplicing activity and needed only the maturase to make this process efficient, tRNA introns had to make an alliance with an independent enzyme to be able to splice. Whether or not experiments show that the present-day EndA family members possess a genuine homing function, it can be envisaged that when the tRNA introns were actively spreading among Archaea and Eukaryota, the probability that the intron (the tRNA-inactivating toxin) and the splicing RNase (the antitoxin responsible for the first step of the intron removal) were transmitted concurrently could be increased greatly if the intrinsic homing ENase activity of the splicing enzyme increased the mobilization of the splicing RNase gene or the intron itself. Nevertheless, some of the Crenarchaeal tRNA-like rRNA introns were shown to encode bona fide homing endonucleases from the LAGLIDADG family (Dalgaard et al. 1993), indicating that the BHB-containing introns might generally provide their own means for mobility, even if they require external EndA to splice (Lykke-Andersen et al. 1997).

Association of homing and splicing activities within a single protein has precedent. It has been shown in vitro and in vivo that several LAGLIDADG DNases encoded by self-splicing group I introns also promote the splicing process as maturases (Schafer et al. 1994), and that both activities are closely associated (Ho et al. 1997). Much is known about the structure/function relationships in the LAGLIDADG family, but maturase activity is relatively poorly studied. Group II introns also encode proteins with multiple activities, including reverse transcriptase, DNase, and maturase (Moran et al. 1992). Therefore, it is possible that the EndA protein, which is composed of domains implicated in a variety of nucleolytic and nucleic acid binding activities in more or less selfish contexts, might also exert functions other than tRNA splicing, and are probably connected with genetic mobility.

The residues of the noncatalytic CTDs from the positions corresponding to the catalytic side chains in PD-(D/E)XK ENases map to the surface on the side opposite to the tRNA binding site, for which no function has yet been implicated. This raises an intriguing possibility that historically, these sites participated in some kind of DNase activity not necessarily connected with the earlier proposed homing-like function of the NTD. Analogous to our suggestion that the EndA protein might possess several cryptic activities, it can be concluded that many restriction enzymes from the PD-(D/E)XK superfamily, which have a three-dimensional fold essentially identical to that of EndA, might have the potential to develop an additional active or binding site on the face opposite the ENase active site.

Because the evolution of the sequence specificity of restriction enzymes is believed to be strongly influenced by their potential to develop different quaternary structures (Newman et al. 1998), the possibility of introducing new sites for intermolecular or intramolecular interactions at their surface might be utilized in protein engineering.