Crystal Structure and Evolution of a Transfer RNA Splicing Enzyme

See allHide authors and affiliations

Science  10 Apr 1998:
Vol. 280, Issue 5361, pp. 279-284
DOI: 10.1126/science.280.5361.279


The splicing of transfer RNA precursors is similar in Eucarya and Archaea. In both kingdoms an endonuclease recognizes the splice sites and releases the intron, but the mechanism of splice site recognition is different in each kingdom. The crystal structure of the endonuclease from the archaeon Methanococcus jannaschii was determined to a resolution of 2.3 angstroms. The structure indicates that the cleavage reaction is similar to that of ribonuclease A and the arrangement of the active sites is conserved between the archaeal and eucaryal enzymes. These results suggest an evolutionary pathway for splice site recognition.

Introns are found in the tRNA genes of organisms in all three of the great lines of descent: the Eucarya, the Archaea, and the Bacteria. In Bacteria, tRNA introns are self-splicing group I introns and the splicing mechanism is autocatalytic (1). In Eucarya, tRNA introns are small and invariably interrupt the anticodon loop 1 base 3′ to the anticodon. They are removed by the stepwise action of an endonuclease, a ligase, and a phosphotransferase (2). In Archaea, the introns are also small and often reside in the same location as eucaryal tRNA introns (3). Splicing in Archaea is catalyzed by an endonuclease, but the mechanism of ligation is likely different from that in Eucarya as there is no homolog of the eucaryal tRNA splicing ligase in the complete genome sequence of several members of the Archaea (4).

The tRNA splicing endonucleases of both the Eucarya and the Archaea cleave the pre-tRNA substrate leaving 5′-hydroxyl and 2′,3′ cyclic phosphate termini, but these enzymes recognize their substrates differently. The eucaryal enzyme uses a measuring mechanism to determine the position of the universally positioned splice sites relative to the conserved domain of pre-tRNA (Fig.1A) (5). In Archaea, the enzyme recognizes a pseudosymmetric substrate in which two bulged loops of 3 bases are separated by a stem of 4 base pairs (bp) [bulge-helix-bulge (BHB)] (Fig. 1A) (6). These observations suggested that the tRNA splicing mechanisms of the three major kingdoms evolved independently.

Figure 1

(A) Consensus sequence and secondary structure of precursor tRNA substrate for yeast and archaeal endonucleases. Splice sites are indicated by short arrows. O and X = nonconserved bases in regions of conserved and variable secondary structure, respectively; Y = pyrimidines; R = purines. Yeast endonuclease is proposed to interact with the mature domain of the tRNA and measure the distance to the splice sites. In contrast, the archaeal enzyme recognizes two 3-nucleotide loops that are separated by a helix of 4 bp (shaded area). (B) Comparison of endonuclease models in Eucarya (yeast) (7) and Archaea (H. volcanii) (8) (see text).

Recently, however, characterization of eucaryal and archaeal endonucleases has revealed a possible common origin for these enzymes. The yeast endonuclease contains four distinct subunits of 15, 34, 44, and 54 kD (7). The endonuclease from the archaeonHaloferax volcanii is a dimer of identical 37-kD subunits (8). The H. volcanii enzyme is homologous to both the 34- and 44-kD subunits of the yeast enzyme, which suggests that these distinct subunits contain active sites for tRNA splicing. This hypothesis was strongly supported by the finding thatsen2-3, a mutant in the 44-kD subunit, is selectively defective in 5′ splice site cleavage (9), whereas a mutant in the 34-kD subunit, H242A, is selectively defective in 3′ splice site cleavage (7). Thus the 44-kD subunit cleaves the 5′ splice site and the 34-kD subunit is proposed to cleave the 3′ splice site (7). The function of the two nonhomologous subunits (the 15- and the 54-kD subunits) remains unclear, although the basic 54-kD subunit was suggested to embody the molecular ruler (7). InH. volcanii, the homodimeric enzyme is proposed to interact with the pseudosymmetric substrate so that each splice site is cleaved by a symmetrically disposed active site monomer (Fig. 1B). Thus it is likely that the spatial disposition of the active subunits has been conserved throughout evolution.

We believed that a high-resolution structure of the archaeal tRNA endonuclease would shed light on the mechanism of the more complicated but related eucaryal endonuclease. Therefore, we determined the crystal structure of the endonuclease from the archaeon Methanococcus jannaschii, which is a homotetramer of 21 kD (179 amino acids) (10).

The M. jannaschii endonuclease gene with His8and FLAG epitope tag sequences (7) at its 5′ end was cloned by polymerase chain reaction from genomic DNA into the pET11a vector. Soluble protein was obtained by overexpression in Escherichia coli (11). The structure was determined by the multiwavelength anomalous diffraction (MAD) method on crystals derivatized by Au(CN)2. The final model, which includes the tetrameric endonuclease of residues 9 to 179, 53 water oxygen atoms, 2 partially occupied SO4 ions, and 4 gold atoms, was refined to 2.3 Å with good stereochemistry to anR factor of 0.204 and a free R value of 0.268 (Table 1).

Table 1

Statistics for MAD data collection and phase determination from a Au(CN)2-derivatized endonuclease crystal flash cooled to 100 K (11). Crystals belong to the space group P212121, with unit cell dimensions a = 61.8 Å, b = 79.8 Å, andc = 192.8 Å. Three wavelengths near the LIII absorption edge of Au were collected from a single crystal at National Synchrotron Light Source beamline X4A using Fuji image plates: λ1, absorption edge corresponding to a minimum of the real part (f′) of the anomalous scattering factor for Au; λ2, absorption peak corresponding to a maximum of the imaginary part (f") of the scattering factor for Au; λ3, a high-energy remote to maximize f′. The crystal was oriented with itsc* axis inclined slightly from the spindle axis (∼16°) to avoid a blind region. The a* axis initially was placed within the plane of the spindle and the beam direction using the STRATEGY option in the program MOSFLM (version 5.4) (21). A total of 180 consecutive 2° images were collected with every 8° region repeated for all three wavelengths. All data were processed with the DENZO and SCALEPACK programs (22). A previously collected data set at 1.5418 Å on a crystal soaked with the same Au(CN)2 solutions was included as the fourth wavelength. This data set was measured with CuKα x-rays generated by a Rigaku RU200 rotating anode using a SIEMENS multiwire detector and was processed with the program XDS (23) followed by scaling and reducing with ROTAVATA and AGROVATA in CCP4 (24). All reflection data were then brought to the same scale as those of wavelength 1 with the scaling program SCALEIT followed by the Kraut scaling routine FHSCAL (25) as implemented in CCP4. We treated MAD data as a special case of the multiple isomorphous replacement with the inclusion of anomalous signals (26). The wavelength 1 (edge) was taken as the “native” data with intrinsic anomalous signals. The data collected at wavelengths 2 (peak), 3 (remote 1), and 4 (remote 2) were treated as individual “derivative” data sets. Anomalous signals are included in the data set of wavelength 2 where the x-ray absorption of Au is expected to be maximum. Because of the absence of a sharp white line at the Au LIII edge, the observed diffraction ratio at wavelength 2 is small, 0.061, compared with the error signal ratio of 0.036 computed from centric reflections. Two gold atoms in each asymmetric unit could be clearly located by both dispersive difference (λ3 or λ4 with respect to λ1) and anomalous (λ2) Patterson functions with the program HASSP (27). Two partially occupied sites were later located by difference Fourier. Initial phases were computed at 3.0 Å with the program MLPHARE as implemented in the CCP4 program suite, which uses a maximum-likelihood algorithm to refine heavy atom parameters (28). There are four noncrystallographically related monomers per asymmetric unit. We performed fourfold averaging in parallel with solvent flattening and histogram mapping with the program DM (29). The subsequent electron density map was improved markedly and the polypeptide chain could be traced unambiguously. A model representing the monomeric endonuclease containing all but the first 25 amino acids (the 8 His, the 9 FLAG epitope tag amino acids, and the first 8 amino acids in the endonuclease) was built into the electron density map with the interactive graphics program O (30). The assignment of side chains for the M. jannaschii endonuclease was confirmed by a difference Fourier map between the Au(CN)2-derivatized protein and a selenomethionine-substituted protein. The final model was refined to 2.3 Å against the observed anomalous data of wavelength 3 with X-PLOR version 3.8 (31). Noncrystallographic restraints were applied to atoms between dimers A1-A2 and B1-B2 excluding those from loops L3 and L7. The structure was assessed with PROCHECK (32) and VERIFY3D (33) and was found to be consistent with a correct structure. More than 90% of the residues are in the most-favored regions, and no residues are in disallowed regions of a Ramachandran plot (32).

View this table:

The endonuclease monomer folds into two distinct domains, the NH2-terminal domain (residues 9 to 84) and the COOH-terminal domain (residues 85 to 179) (Fig.2A). The NH2-domain consists of a mixed antiparallel/parallel β sheet and three α helices. The first three β strands (β1 to β3) are antiparallel and, together with β4, are packed against two nearly perpendicular α helices, α1 and α2. The third α helix, α3, is associated with α2 via an antiparallel coiled-coil. The connection between β4 and α3 (residues 64 to 67) is disordered, which implies that this loop is flexible. This feature is consistent with the proteolytic sensitivity of this region (10). A short loop consisting of three charged residues (Glu81, Glu82, Arg83) links the NH2- to the COOH-terminal domain. The COOH-terminal domain is of the α/β type and consists of a central four-stranded mixed β sheet β5-β6-β̅7̅-β̅8̅ (overbars denote parallel strands) cradled between helices α4 and α5. Two basic folding motifs lead to this fold: the rare αββ(α4-β5-β6) and the more common βαβ(β7-α5-β8). The last β strand, β9, is partially hydrogen-bonded with β8 and has little involvement in the COOH-terminal domain folding; rather, it participates in dimerization (see below).

Figure 2

(A) Ribbon representation of one subunit of M. jannaschii endonuclease obtained with the RIBBONS program (20). The proposed catalytic triad residues are within 7 Å of each other and are shown in red ball-and-stick models (see text). The electron density in the averagedF o map is contoured at 5σ and is drawn only near the putative catalytic triad. (B) Subunit arrangement and interactions in the M. jannaschii endonuclease. Each subunit is represented by a distinct color and a label. The tetramer is viewed along the true twofold axis relating the A1-A2 and B1-B2 dimers. The main chain hydrogen bonds formed between β9 and β9′ and between loops L8 and L8′ for isologous dimers are drawn as thin lines. Side chains of the hydrophobic residues enclosed at the dimer interface are shown as blue ball-and-stick models. The heterologous interaction between subunits A1 and B2 (or B1 and A2) through the acidic loops L10 and L8 are highlighted by dotted surfaces.

The quaternary structure of the endonuclease is unusual in that it does not assume the D2 point group symmetry observed in most homotetrameric proteins. When viewed along a noncrystallographic twofold axis, the homotetrameric M. jannaschii endonuclease resembles a parallelogram with the four subunits occupying the four corners (Fig. 2B). Subunits A1 and B1, related by a true twofold axis, are closer in space than subunits A2 and B2, which are related by the same twofold axis. Subunits A1 and A2 or B1 and B2 are related by two nonorthogonal pseudo-twofold axes plus a 1.6-Å translation along the rotation axes. The resulting dimers, A1-A2 and B1-B2, lie on two opposite sides of the parallelogram. Each of the pseudo-twofold axes relating monomers within the A1-A2 and B1-B2 dimers is at an angle of 32° with respect to the twofold axis that relates the two dimers. In addition, the three twofold axes do not intersect. Each pseudo-twofold axis is shifted about 10 Å in the opposite direction relative to the twofold axis that relates the two dimers. It is therefore more precise to describe the endonuclease as a dimer of dimers. The α carbons of the A1-A2 dimer can be superimposed with those of the B1-B2 dimer with a root-mean-square difference of 0.47 Å. The α carbons of the subunits within each dimer have a root mean square of 1.2 Å. The largest deviation is observed on loop L7, which harbors the conserved His125 residue. The asymmetrical organization of the four identical subunits suggests that the two subunits in the dimers have nonequivalent roles.

The endonuclease dimers A1-A2 and B1-B2 associate isologously in a tail-to-tail fashion (Fig. 2B). The dimerization interface buries extensive surface area (2422 Å2) (12) and is formed by the interaction of the central COOH-terminal β sheets from each monomer by main chain hydrogen bonding and hydrophobic interactions. The pseudo-twofold axis that relates the two monomers lies between the two β9 strands and between the two L8 loops in each subunit. The NH2 portion of β9 (residues169 to 172) is hydrogen-bonded to β8 by the main chain atoms. The COOH-terminal half of β9 (residues 172 to 177) bends away from the plane of the pleated central β sheet to form main chain hydrogen bonds with the symmetry-related residues of β9 in the isologous subunit (β9′); as a result, a two-stranded β sheet is produced that spans the subunit boundary. The symmetrically related L8 loops form another layer on top of the two-stranded β9 sheet and, together with the β9 sheet, enclose a hydrophobic core at the intersubunit surface. Prealbumin and enterotoxin have similar modes of subunit association (13). The hydrophobic core includes Phe139, Leu141, Leu144, Gly146, Val148, Leu158, Ile160, Met174, and Tyr176 (Fig. 2B, blue ball-and-stick model). These hydrophobic residues are important for stabilizing the dimer interface.

The interface between the two dimers is between subunits A1 and B2 and subunits B1 and A2, with each pair contributing 40% of 5420 Å2 of total buried surface area (12). The remaining 20% of buried surface area can be accounted for by the interface between subunits A1 and B1. Subunits A2 and B2 are not in contact. The principal interaction between dimers is mediated by close contacts between polar residues of opposite charges in the two heterologously associated subunits. A positively charged cleft (9 Å wide, 12 Å long, 14 Å deep) is formed between the two domains of subunit A1 (or B1). It accommodates a protruding negative surface on the subunit B2 (or A2) formed primarily by L10 and partially by L8 (Fig. 2B, dotted surfaces). Residues from L10 and L8 that are deeply buried by this interdomain cleft are mostly acidic (Asp164-Ala165-Asp166-Gly167-Asp168and Glu135-Asp136). Several positively charged residues (Lys88, Lys103, Lys107, Arg113) line the cleft and interact electrostatically with the acidic loop residues. The interactions between L10 and L8 and the interdomain cleft contribute nearly all the buried surface area at the tetramerization interface. In addition, three completely buried water molecules at the base of the cleft form hydrogen bonds with polar residues from both subunits, indicating that this cleft is solvent accessible before tetramerization.

The isologous dimer interaction formed by the COOH-terminal interface is expected to be more stable than the heterologous polar interaction mediated by L10 and L8, which suggests that the tetramer assembles through dimer intermediates. This interpretation is in agreement with protein cross-linking data (10). It is likely that the dimers observed in solution are the isologous dimers A1-A2 or B1-B2 in the crystal structure.

The tRNA splicing endonucleases generate 5′-hydroxyl and 2′,3′ cyclic phosphate termini, as do other ribonucleases (RNases) such as RNase A and T1. Cleavage by RNase A is a two-step reaction that is acid-base catalyzed (14). Three residues in RNase A have been identified as the catalytic triad for this reaction: His12 is the general base in the first step, His119 is the general acid, and Lys41stabilizes the pentacovalent transition state.

Splicing endonucleases contain a conserved histidine residue that has been suggested to be part of the active site on the basis of mutational studies (7, 10, 15). In the M. jannaschii endonuclease structure, His125 is at the boundary between L7 and β7 (Fig. 2A). Although His125 is well ordered in the structure in all four subunits, the rest of L7 (residues 119 to 124) has weak electron density, which is suggestive of flexibility. The average B factor of L7 is 66 Å2 compared with the average value of 44 Å2 for the protein as a whole. Two striking structural features are observed in this region. First, there are primarily positively charged or polar residues in the vicinity of His125, including two other strictly conserved residues, Tyr115 and Lys156. The side chains of His125, Tyr115, and Lys156 are within 4 to 7 Å of each other and form a His-Tyr-Lys triad (5.5 Å between His125 and Lys156, 6.7 Å between Lys156 and Tyr115, 6.1 Å between Tyr156 and His125) (Fig. 2A). Second, a distinct electron density feature (>5σ in the averagedF o map) is found in the middle of the triad. This density is not part of the polypeptide chain and is compatible with the size of a sulfate ion that is present in crystallization and stabilization solutions.

We propose that the side chains of the His-Tyr-Lys triad in M. jannaschii endonuclease form the catalytic triad for the cleavage reaction. This His-Tyr-Lys triad is spatially superimposable with the catalytic triad in RNase A (16) and is consistent with a model in which the endonuclease His125 functions as the general base, Tyr115 functions as the general acid, and Lys156 functions as a stabilizer of the transition state. Although mutagenesis studies do not provide conclusive support for the essentiality of Tyr115, they do indicate an essential role for both His125 and Lys156 (4). More studies are necessary to establish the exact catalytic mechanism and the function of these residues in catalysis. Interestingly, the inactivating Gly292 to Glu mutation in the yeast enzyme occurs within the aligned sequences of the flexible loop L7, which contains both His125 and Tyr115 (7).

Although the M. jannaschii endonuclease tetramer contains four active sites, it is likely that only two of these participate in cleavage. The yeast endonuclease tetramer contains two functionally independent active sites for cleavage. The H. volcanii endonuclease is a homodimer and thus also contains two functional active sites. The two intron-exon junctions of the archaeal tRNA substrate are related by a pseudo-twofold symmetry axis in the consensus BHB substrate. The M. jannaschii endonuclease is therefore predicted to contain only two functional active sites related by twofold symmetry. Consistent with this notion, upon substrate binding the endonuclease is protected from proteolytic cleavage to a maximum of 50% (10). Four pairs of enzyme active sites are related by twofold symmetry; these occur in subunits A1 and B1, A2 and B2, A1 and A2, and B1 and B2 (Fig. 2B). We postulate that the active sites reside in A1 and B1; their His-Tyr-Lys triads are separated by about 27 Å. The residues protected upon RNA binding cluster on A1 and B1 around the two proposed catalytic sites, falling on a path between the two sites. This region of the enzyme, formed by L9 and the carboxyl portions of α5 and β9 in A1 and B1, has a positive electrostatic potential suitable for binding the phosphodiester backbone of the pre-tRNA substrate. By contrast, the distance between the His-Tyr-Lys triads of A2 and B2 is 75 Å and there is a more scattered distribution of substrate protected residues between this pair of active sites (17).

We attempted to dock a model of the substrate with the tetramer. We created the model from a related structure solved by the nuclear magnetic resonance method, that of the human immunodeficiency virus trans-activation response element (TAR) RNA complexed with arginamide (18). TAR RNA is a stem-loop structure with a bulge of three bases in the stem (Fig. 3A). We obtained the substrate model by rotating a section of TAR RNA by 180° to create the pseudosymmetric BHB structure conforming to the archaeal endonuclease consensus substrate (Fig. 3B). This substrate model derived from the arginamide complex of TAR RNA docks well with subunits A1 and B1 in the M. jannaschii endonuclease tetramer (Fig.3C). The two phosphodiester bonds that are cleaved fit precisely into the active sites of A1 and B1 and superimpose with the putative sulfate ion. The 4-bp helix separating the two bulges interacts with the basic face of the enzyme and could contact a number of residues that are protected from protease cleavage upon RNA binding (10).

Figure 3

(A) Sequence and secondary structure of HIV TAR RNA (PDB code 1arj) used to modelM. jannaschii pre-tRNA minimum substrate. (B) A symmetry operation around the twofold axis as indicated is applied to coordinates of the TAR RNA nuclear magnetic resonance structure to obtain a structural model for the second three-base bulge and the stem extension (boxed region). The corresponding splice sites on the final model are indicated by arrows. (C) Modeled complex ofM. jannaschii endonuclease with RNA substrate obtained by manually aligning the phosphate backbone of the 4-bp helix with the positively charged surface between subunits A1 and B1. This view is perpendicular to that of Fig. 2B. Only subunits A1 and B1, which are proposed to participate in the cleavage reaction, are shown. The proposed catalytic triad residues are shown in magenta. The phosphodiester bonds at the splice sites on RNA are depicted in blue and fit nearly perfectly with the putative sulfate density shown by blue dots. The distance from the center of the catalytic triad to the surface where RNA binds is about 7 Å.

As the H. volcanii endonuclease dimer and the M. jannaschii tetramer recognize the same consensus substrate, their active sites must be arrayed similarly in space. This proposal is strongly supported by the fact that the H. volcaniiendonuclease monomer is in fact a tandem repeat of the consensus sequence of the endonuclease gene family (10) (Fig.4B). We propose that the H. volcanii enzyme forms a pseudotetramer of two pseudodimers (Fig.4B). The structure of the pseudodimer is predicted to contain a two-stranded β9-β9′ pleated sheet, an important structural feature of the M. jannaschii endonuclease dimer (Fig. 4A). The tandem repeat must be connected by a stretch of polypeptide (Fig. 4B, dotted line). Within each dimer of the M. jannaschiiendonuclease structure, the NH2-terminus of subunit A2 (or B2) and the COOH-terminus of subunit A1 (or B1) is separated by 28 Å. This requires a span of at least 10 amino acids of an extended polypeptide chain to connect the region between the NH2-terminal and the COOH-terminal tandem repeats in theH. volcanii endonuclease. The H. volcanii enzyme model contains only two of the proposed active site triads in the COOH-terminal repeat of each pseudodimer, and these triads are proposed to occupy a spatial configuration identical to those in the A1 and B1 subunits of the M. jannaschii endonuclease. The pseudodimers are proposed to interact via the conserved L10 sequences. This H. volcanii endonuclease structural model reveals that only two of the active sites are necessary, but to array these in space correctly one must retain features of both the isologous dimer interactions (β9-β9′) and the dimer-dimer interactions mediated by L10 in theM. jannaschii enzyme.

Figure 4

(A) Sequence features and subunit arrangement of the M. jannaschii homotetramer. Several important structural features discussed in the text are indicated: loop L10, the COOH-terminal β9 strands (arrows), and the conserved catalytic residue His125 (pentagon). (B) Sequence features conserved between H. volcanii and M. jannaschii endonucleases and the proposed subunit arrangement of the former. The two tandem repeats are more similar to the M. jannaschii endonuclease sequence than to each other. The NH2-terminal repeat lacks two of the three putative active site residues (white bars). It does, however, contain many of the features of the COOH-terminal domain that are important for structural arrangement of the enzyme, in particular the L10 sequence (yellow bars). The COOH-terminal repeat contains all the sequence features of the M. jannaschii enzyme. Dashed lines represent the polypeptide chain connecting the COOH- and NH2-terminal repeats. (C) Proposed structural model for the yeast endonuclease. Conserved amino acid sequences near the COOH-termini of archaeal enzymes M. jannaschii (M. jann.), H. volcanii NH2-terminal repeat (H. vol. Nt), and yeast Sen54 (Sc. Sen 54p) and Sen15 (Sc. Sen15p) subunits are aligned. Important hydrophobic residues that stabilize the isologous COOH-terminus interaction between M. jannaschii subunits A1 and A2 (or B1 and B2) are highlighted in green and circled on the structural models of heterodimers derived from the two-hybrid matrix analysis (7). The sequences of Sen54 and Sen15 aligned with L10 sequences in M. jannaschii and H. volcanii are highlighted in red. Loops L10 on both Sen54 and Sen15 are labeled on the proposed heterotetramer model of the yeast endonuclease. Abbreviation for amino acids are as follows: L, Leu; I, Ile; A, Ala; V, Val; D, Asp; G, Gly; Y, Tyr; N, Asn; M, Met; E, Glu; T, Thr; F, Phe; W, Trp; S, Ser; K, Lys.

The heterotetrameric yeast endonuclease is proposed to contain two active subunits, Sen2 and Sen34. The other two subunits, Sen15 and Sen54, have a sequence to the M. jannaschii endonuclease at the COOH-terminus (10). This conserved region contains the L10 residues and the hydrophobic residues at the COOH-terminus as observed in β8 and β9 in the crystal structure of the M. jannaschii endonuclease. We propose that the strong Sen2-Sen54 and Sen34-Sen15 interactions observed in the two-hybrid experiment (7) are mediated by COOH-terminal β9-β9′–like interactions (Fig. 4C) and that these dimers are associated to form the tetramer via the L10 sequences of Sen15 and Sen54 (Fig. 4C).

We conclude that the Archaea and the Eucarya have inherited from a common ancestor an endonuclease active site and the means to array two sites in a precise and conserved spatial orientation. This conclusion is supported by the results of Fabbri et al. (19), which demonstrate that both the eucaryal and the archaeal endonucleases can accurately cleave a universal substrate containing the BHB motif. The eucaryal enzyme appears to dispense with the more complex ruler mechanism for tRNA substrate recognition when it cleaves the universal substrate. Thus the precise positioning of two active sites in endonuclease appears to have been conserved in evolution. Subunits A1 and B1 comprise the active site core of all tRNA splicing endonucleases, and subunits A2 and B2 position the two active sites precisely in space. The eucaryal enzyme has evolved a distinct measuring mechanism for splice site recognition by specialization of the A2 and B2 subunits, and it has retained the ability to recognize and cleave the primitive consensus substrate.

  • * To whom correspondence should be addressed. E-mail: abelsonj{at}


View Abstract

Navigate This Article