Dual role of the RNA substrate in selectivity and catalysis by terminal uridylyl transferases

Abstract

Terminal RNA uridylyltransferases (TUTases) catalyze template-independent UMP addition to the 3′ hydroxyl of RNA. TUTases belong to the DNA polymerase β superfamily of nucleotidyltransferases that share a conserved catalytic domain bearing three metal-binding carboxylate residues. We have previously determined crystal structures of the UTP-bound and apo forms of the minimal trypanosomal TUTase, TbTUT4, which is composed solely of the N-terminal catalytic and C-terminal base-recognition domains. Here we report crystal structures of TbTUT4 with bound CTP, GTP, and ATP, demonstrating nearly perfect superposition of the triphosphate moieties with that of the UTP substrate. Consequently, at physiological nucleoside 5′-triphosphate concentrations, the protein–uracil base interactions alone are not sufficient to confer UTP selectivity. To resolve this ambiguity, we determined the crystal structure of a prereaction ternary complex composed of UTP, TbTUT4, and UMP, which mimics an RNA substrate, and the postreaction complex of TbTUT4 with UpU dinucleotide. The UMP pyrimidine ring stacks against the uracil base of the bound UTP, which on its other face also stacks with an essential tyrosine. In contrast, the different orientation of the purine bases observed in cocrystals with ATP and GTP prevents this triple stacking, precluding productive binding of the RNA. The 3′ hydroxyl of the bound UMP is poised for in-line nucleophilic attack while contributing to the formation of a binding site for a second catalytic metal ion. We propose a dual role for RNA substrates in TUTase-catalyzed reactions: contribution to selective incorporation of the cognate nucleoside and shaping of the catalytic metal binding site.

Terminal RNA uridylyltransferases (TUTases) are phylogenetically widespread and functionally divergent enzymes that catalyze template-independent transfer of UMP residues to the 3′ hydroxyl group of RNA (1). TUTases belong to the polymerase β nucleotidyltransferase superfamily, which is characterized by the presence of the signature motif hG[G/S]X9-13Dh[D/E]h (X, any; h, hydrophobic amino acids) (2) and also includes poly(A) polymerases (PAPs), ATP(CTP):tRNA nucleotidyltransferases, terminal deoxy nucleotidyltransferases, protein nucleotidyltransferases, 2′–5′ oligo(A) synthetases, and antibiotic resistance nucleotidyltransferases (3). TUTase activities have been reported in mammalian cells, plants, and parasitic protists from the order Kinetoplastida, Trypanosoma brucei, and Leishmania ssp. (1). Extensive uridine insertion/deletion editing in mitochondria of trypanosomatids requires 3′ uridylylation of guide RNAs by RNA editing TUTase 1 (RET1) (4) and insertion of Us into messenger RNAs by RNA editing TUTase 2 (RET2) (5, 6). In addition to mitochondrial RNA editing TUTases, TUT3 (7) and TUT4 (8), cytoplasmic uridylyl transferases have been reported in T. brucei. Human cells apparently possess several distinct TUTase activities, of which the U6 small nuclear RNA-specific TUTase (9) and a UTP-specific enzyme of unknown function (10) have been identified. Cid1, a member of a multifunctional protein family in fission yeast, was characterized as a cytoplasmic PAP (11) that also possesses a robust TUTase activity (10). An enzymatic activity screening of proteins homologous to the animal Gld-2 cytoplasmic PAP (12) revealed several novel TUTases, also referred to as poly(U) polymerases, encoded in the human, Caenorhabditis elegans, and Arabidopsis thaliana genomes (13).

Recent crystallographic studies of TbRET2 (14) and TbTUT4 (8) attributed the specificity of UTP incorporation by trypanosomal TUTases to the largely conserved, high-affinity binding site formed at the interface of the N-terminal domain, which bears a polymerase β signature sequence with three catalytic aspartate residues, and the C-terminal domain, which resembles an ATP cone-like fold also found in 2′–5′ oligoadenylate synthetases (15) and is responsible for base-specific contacts. By combining kinetic analysis and UTP cross-linking studies of mutant TbTUT4 proteins, we have established the importance of the uracil base interactions for UTP binding (8). Whereas the stacking interaction most likely contributes to the selectivity of TbTUT4 toward pyrimidines, base-specific recognition of UTP is apparently achieved by endocyclic N3 donating a hydrogen bond to one water molecule and carbonyl O4 receiving a hydrogen bond from another water molecule, which would require reversal of this hydrogen-bonding pattern at these two positions for effective CTP binding (8).

The spacious UTP binding site formed at the interface of the N-terminal and C-terminal domains, remarkable sequence similarity between the catalytic domains of TUTases and noncanonical PAPs, and conservation of key residues involved in UTP binding [supporting information (SI) Fig. 7] (1, 16) prompted us to investigate whether UTP–protein contacts alone are sufficient to confer U-specificity at nucleoside 5′-triphosphate (NTP) concentrations approaching physiological levels. Surprisingly, we were able to generate cocrystals of TbTUT4 with CTP, ATP, and GTP under the same conditions used for TbTUT4:UTP binary complex crystallization (8). The x-ray structures of these complexes demonstrated nearly perfect superposition of the triphosphate moieties while revealing a major reduction in the observed coplanarity of purine bases with respect to the phenyl ring of conserved Y189, relative to that of pyrimidines. This shift of base positioning diminishes the stacking interaction with Y189 but does not fully explain the enzyme's ability to discriminate against purine NTPs. To further examine this phenomenon, we have established that TbTUT4 can use uridylyl monophosphate in lieu of an RNA primer and elucidated structures of a TbTUT4:UTP:UMP ternary complex (precatalysis), as well as the structure of TbTUT4 with the bound adduct, UpU (postcatalysis, but lacking the 5′ phosphate). We discovered that the terminal base of the RNA substrate forms a “triple-stacked sandwich” with the bound UTP base and the phenol ring of Y189, which positions its 3′ hydroxyl group for in-line attack on the α-phosphorus atom of the UTP. In contrast, the observed positions of purine bases in ATP- and GTP-bound structures display reduced stacking interactions and likely destabilize RNA binding. In addition, the 3′ hydroxyl group of UMP (mimicking the 3′ residue of the RNA substrate) completes a binding site for a second catalytic metal ion, which is typically required for the nucleoside transfer reaction (17) but has not been observed in previously reported TbTUT4:UTP and TbRET2:UTP binary complexes [Protein Data Bank (PDB) ID codes 2IKF and 2B56, respectively] (8, 14). Modeling of additional RNA residues into the ternary complex provides a rationale for the pronounced selectivity of trypanosomal TUTases toward the specific base at the 3′ end of the RNA substrate. Thus, our studies establish a structural model for UTP/ATP recognition by a conserved catalytic module shared among RNA uridylyltransferases and noncanonical PAPs, a group of enzymes involved in a wide variety of RNA processing events in eukaryotes (1, 16).

Results

Nucleoside Triphosphate Binding by TbTUT4.

Although TbTUT4 greatly favors UTP as its substrate, it has been shown that the enzyme is also capable of using, to various extents, CTP, ATP, and GTP. Kinetic parameters of the TbTUT4-catalyzed reaction in the presence of different ribonucleoside triphosphates demonstrated that non-UTP substrates are characterized by similar values for the apparent K_m, which are ≈20-fold higher than that observed for the cognate substrate (Table 1). Surprisingly, the catalytic rate remained nearly identical for pyrimidines while decreasing significantly for purines. This finding indicated that, at physiological concentrations of NTPs in vivo (>0.5 mM), UMP incorporation by TUTases may require, in addition to specific UTP–protein contacts (8, 14), a supplementary component ensuring a selective reaction. The structures of TbTUT4 cocrystals with UTP, CTP, ATP, and GTP demonstrated nearly identical binding of the triphosphate moiety. However, deviations become evident in the positioning of the ribose sugar pucker and, more prominently, the degree of coplanarity of the pyrimidine vs. purine bases with respect to the phenyl ring of Y189 (Fig. 1). Detailed analysis of the various TbTUT4:NTP cocrystal structures illustrates four primary differences in the binding of purines relative to that of pyrimidines: (i) a second metal ion of unknown function is present in ATP and GTP complex structures; (ii) N147, which forms hydrogen bonds with the 2′ hydroxyl group of the ribose and a carbonyl oxygen of the pyrimidine bases, is involved in hydrogen bonding with both ribose hydroxyl groups in the purine-bound structures; (iii) a hydrogen bond is donated from the hydroxyl group of residue T187 to N7 of the purine rings of ATP and GTP, whereas this residue appears to play no role in pyrimidine binding; and (iv) the pose of the purine rings results in substantially reduced π-electron stacking with the aromatic ring of Y189, an interaction previously shown to be essential for catalysis (8). These observations provide a structural explanation for the fact that enzyme–heterocyclic base contacts alone are not sufficient to confer UTP selectivity. This is consistent with the conservation of the catalytic/base recognition bidomain scaffold among TUTases and several divergent PAPs, such as trypanosomal mitochondrial PAP (R. D. Etheridge, I.A., and R.A., unpublished data), animal cytoplasmic Gld-2 (13, 18), Saccharomyces cerevisiae Tr4/5 PAPs (19–21), and the Cid1–13 family of proteins from Schizosaccharomyces pombe (11, 22) (SI Fig. 7). Our findings suggest that NTP binding by TUTase-like nucleotidyl transferases is relatively promiscuous and that specificity is achieved by additional mechanisms.

Table 1.

Kinetic constants for NTP incorporation by TbTUT4

NTP	Km, μM	kcat, min−1	kcat/Km, min−1·M−1	[kcat/Km]UTP/[kcat/Km]NTP
UTP	1 ± 0.9	0.5 ± 0.06	5 × 105	1.0
ATP	47.5 ± 19.8	0.01 ± 0.001	2 × 102	0.0004
CTP	17.1 ± 2.6	0.5 ± 0.03	2.9 × 104	0.06
GTP	18.0 ± 1.8	0.004 ± 0.002	5 × 102	0.0001

Data for UTP are from ref. 8. Reactions were performed with 6[U] RNA as the substrate.

Fig. 1.

Primary interactions in NTP binding by TbTUT4. Ordered water molecules and Mg²⁺ ions are depicted in cyan and black, respectively. (A) The triphosphate moiety contacts. (B) Sugar/base interactions. (C) The relative positions of all four bound triphosphate ribonucleosides derived from superpositioning the structures of TbTUT4:UTP, TbTUT4:CTP, TbTUT4:ATP, and TbTUT4:GTP. Composite annealed omit maps for each bound ligand observed in the various ligand complexes of TbTUT4 are shown in SI Fig. 8.

RNA Substrate Specificity.

RNA substrate specificity differs markedly among trypanosomal TUTases with RET1 (4, 5), TUT3 (7), and TUT4 (8) preferentially acting on single-stranded substrates and RET2 capable of using double-stranded RNAs (5, 6), (Fig. 2A). Putative editing intermediates are represented by model RNA substrates for a “precleaved” U-insertion assay (Fig. 2A). TbTUT4 is inactive on base-paired RNA substrates but readily extends the single-stranded “5′ fragment” (Fig. 2A Left, lane 2) and 24-mer 6[U] (Fig. 2B Left) substrates. Unexpectedly, RNA with six adenosines at the 3′ end, 6[A], displayed much lower efficiency as a primer than 6[U] RNA (Fig. 2B Center). Replacing only the 3′ uridylyl residue with an adenosyl nucleoside had an inhibitory effect (Fig. 2B Right). We hypothesized that the uracil base at the 3′ terminus of the RNA primer is involved in specific interactions with the active site. Because no structure of the homoribonucleotidyl transferase with bound RNA is available and our attempts to cocrystallize TbTUT4 with RNA and a nonhydrolyzable UTP analog have so far been unsuccessful, we investigated whether uridine monophosphate could serve as minimal RNA substrate to elucidate RNA positioning adjacent to the UTP binding site.

Fig. 2.

RNA substrate specificity of TbTUT4. (A) Double-stranded RNA substrates for precleaved in vitro insertion assays (4) were assembled before reaction from the labeled 5′ fragment, 3′ fragment, and guide RNA, which allows for insertion of three, two, and zero nucleotides. RNAs (0.5 μM) were incubated with 50 nM recombinant TbTUT4 or TbRET2 in the presence of 100 μM UTP for 30 min. Lane 1, control RNA; lane 2, 5′ fragment alone; lane 3, 5′ fragment hybridized with a “bridge”; lanes 4, 5, and 6, fully assembled substrates with a 0-, 2-, or 3-nt gap. Products were separated on 15% polyacrylamide/8 M urea gels. (B) Single-stranded 5′ radiolabeled RNA substrates ending with six Us, six As, or terminal A after five Us were incubated with 50 nM purified TbTUT4 in the presence of 1, 10, 100, and 500 μM UTP for 30 min. c, control RNA. (C) TbTUT4 transfers uridylyl residues to UMP and UpU acting as RNA primers. In gels 1–3, the concentration of the 5′ radiolabeled uridylyl-3′,5′-uridine (³²pUpU) was kept at 100 μM and UTP as indicated. Reactions were performed in the presence of 50 nM TbTUT4 for 30 min. Gel 1, the enzyme was preincubated with ³²pUpU for 10 min, and the reaction was started by addition of UTP; gel 2, the enzyme was preincubated with UTP for 10 min, and the reaction was started by addition of ³²pUpU; gel 3, UMP and ³²pUpU were premixed, and the reaction was started by addition of the enzyme; gel 4, 0.1 μM [α-³²P]UTP and 2 mM UMP were premixed, and the reaction was started by addition of the enzyme. Products were separated on 20% polyacrylamide/urea gels. (D) Effect of the order of substrate addition on the TbTUT4-catalyzed reaction. Reactions were carried out with 50 nM TbTUT4 in the presence of 0.5 μM radiolabeled 6[U] RNA and indicated concentrations of UTP for 1, 5, and 30 min. Products were separated on 15% polyacrylamide/urea gels.

In contrast to our earlier findings of primer-independent UTP polymerization by RET1 (23), no such reaction could be detected for TbTUT4 at various UTP concentrations (data not shown). Incubation of TbTUT4 with UMP and radiolabeled UTP led to an accumulation of the pUpU adduct, indicating that UMP indeed binds at the active site and is capable of assuming the role of an RNA substrate (Fig. 2C, gel 4). Increasing the size of the RNA substrate to a 2-mer (pUpU) led to an increase in efficiency, but only if the reaction was initiated by the addition of the enzyme to a mixture of the UTP and pUpU substrates (Fig. 2C, gels 1–3). Preincubating the enzyme with either substrate before addition of the second reactant significantly decreased the efficiency of nucleoside transfer. Similar, but less pronounced, effects were observed with a 24-mer 6[U] RNA indicating potential competition of substrates for the binding site.

The crystal structure of TbTUT4 with bound UTP and bound UMP (as the minimal RNA substrate) illustrates triple coplanar aromatic stacking of the UTP uracil base sandwiched between the phenyl ring of the essential Y189 (8) and the uracil base of the minimal RNA, UMP (Fig. 3A), an arrangement that closely resembles an RNA strand in a double helical conformation. The direct contacts of the “terminal” RNA residue with the enzyme consist of three hydrogen bonds with R121, D68, and D136 and hydrophobic interactions of the base with V122 (Fig. 3B). A significant increase in the apparent K_m for RNA due to the R121A mutation has been demonstrated previously (8). Taken together, these observations suggest that hydrogen bonding of the ribose and uracil base with conserved protein residues, the stacking interaction with UTP, and coordination of the 3′ hydroxyl by a second magnesium ion (Mg²⁺) are all essential for correct positioning of the RNA and, therefore, catalysis. To further analyze the binding of the RNA substrate by TbTUT4, an additional UMP residue was modeled by hand into the ternary TbTUT4:UTP:UMP crystal structure to form a 2-mer RNA molecule in the active site (Fig. 3C). The pose of this modeled penultimate RNA residue suggests hydrogen bond interactions between the hydroxyl groups of the 5′ ribose and R126 (K149 in TbRET2), a hydrogen-bonding pattern similar to that observed between N147 and the ribose hydroxyls of the bound UTP (8, 14). These interactions based on modeling are consistent with the fact that the R126A mutant is virtually inactive because of loss of RNA binding, but not UTP binding (8). Furthermore, the hydrogen bond received by O4 of the 5′ uracil base suggests that the enzyme prefers a U residue in the penultimate position of the RNA substrate. This could explain in part why TbTUT4 functions more efficiently on a 5[U]A RNA primer than on a 6[A] primer (Fig. 2B).

Fig. 3.

The prereaction complex of UTP and UMP (“RNA”) in the active site of TbTUT4. Mg²⁺ ions (black) are labeled Mg1 and Mg2, where Mg1 is the binding site previously observed in the TbTUT4:UTP structure (8). (A) Triphosphate coordination by Mg1 and formation of a binding site for a second metal ion (Mg2) upon UMP binding. (B) Direct protein–UMP hydrogen bond contacts. (C) Hydrogen bond contacts with UMP (“RNA”) at the terminal (gray) and modeled penultimate (green) UMP residues.

RNA Contribution to Nucleotide Selection and Catalysis.

In addition to the primary role of RNA as a nucleophile in the transferase reaction, its terminal residue base-stacking interaction with the bound NTP suggests a role of the RNA in nucleoside incorporation selectivity by uridylyl transferases. In the case of purine nucleoside triphosphates, coplanarity between the purine ring and Y189 is significantly reduced (60° versus 45° in the case of pyrimidines). This, accompanied by a relative translation for purine bases, would result in virtually no stacking of bound ATP or GTP with the 3′ base of the bound RNA substrate (Fig. 4). The resulting reduction in binding affinity for the RNA substrate may adversely affect the positioning of its 3′ hydroxyl group, thereby explaining the reduction in the observed catalytic rates for purine incorporation into RNA (Table 1). Discrimination between UTP and CTP, on the other hand, does not involve a selective RNA binding factor and relies on a highly coordinated water molecule, which accepts a hydrogen bond from the donor-only N3 of the uracil base in the case of UTP binding (8, 14).

Fig. 4.

Triple-stacking interaction is required for productive RNA binding. Stereo views of the superposition of TbTUT4:ATP and TbTUT4:UTP:UMP, respectively, illustrate the various degrees of stacking of the aromatic rings of Y189, the NTP base, and UMP (RNA) for purine NTPs (adenine, shown in green) versus pyrimidine NTPs (uracil, shown in yellow). Upon superposition, there is virtually no base stacking observed between the pyrimidine ring of UMP and the purine ring of ATP.

The nucleotidyltransferase reaction involves nucleophilic attack by the 3′ hydroxyl of the RNA substrate on the α-phosphorus of the bound NTP, which results in nucleoside incorporation into the RNA and liberation of pyrophosphate. Three conserved catalytic aspartates coordinate the divalent metal ions essential for catalysis. The crystal structure of TbTUT4 with bound UTP (8) revealed a single Mg²⁺ (Mg1) coordinated by two aspartates (D66 and D68) and the triphosphate moiety of UTP. Similar observations have been reported for TbRET2 (14). However, upon binding of the minimal RNA substrate, UMP, a second Mg²⁺ (Mg2) was observed, which coordinates the two reactants of the transferase reaction by positioning the 3′ hydroxyl of the RNA ≈4.6 Å away from the α-phosphorus of the bound UTP (Fig. 5A). In line with a universal two-metal-ion mechanism for nucleotidyl transfer (24), Mg2 is expected to facilitate deprotonation of the RNA 3′ hydroxyl to stimulate the nucleophilic substitution reaction, whereas Mg1 is thought to stabilize the pyrophosphate leaving group. The importance of triple base stacking during catalysis is further confirmed by the crystal structure of TbTUT4 with bound UpU, which depicts a pseudoproduct state of the transferase reaction between UTP and UMP (Fig. 5B). The crystal structure reveals that the UpU binds such that one uracil base takes the place of the UTP uracil and the other takes the place of the RNA (UMP) uracil. The possible bases within range to deprotonate the ribose 3′ hydroxyl, a necessary step for catalysis, are D68, D136, and Wat1 (Fig. 5A). However, the direct interactions of D68 and Wat1 with divalent cations would likely make them highly acidic, suggesting that D136 functions as the catalytic base. This would explain previous observations that, although the universally conserved D136 does not contribute directly to NTP or Mg²⁺ binding, its mutation to alanine renders the RET1 and TUT4 TUTases inactive (8, 23).

Fig. 5.

Structures of the active sites of the pre- and postreaction complexes. (A) Reactant state with TbTUT4:UTP:UMP. Mg2 (black sphere) positions the 3′ hydroxyl of UMP (terminal RNA residue) for in-line nucleophilic attack on the α-phosphorus atom of the bound UTP at a distance of 4.6 Å. (B) Product state with TbTUT4:UpU.

Upon UTP binding, TbTUT4 undergoes significant conformational changes that bring the N-terminal and C-terminal domains closer together into a more compact structure (8). On the other hand, the TbTUT4 protein conformations in the UTP-, UTP:UMP-, and UpU-bound structures are virtually identical, suggesting that neither RNA binding to the TbTUT4:UTP binary complex nor catalysis is accompanied by major conformational changes of the protein. With the UpU adduct, or any +1 RNA, occupying the active center as described above (Fig. 5B), a subsequent round of U-incorporation would require a protein conformational change to translocate the new 3′ OH of the RNA into position for nucleophilic attack. An alternative scenario is the dissociation of the entire RNA molecule or its 3′ end from the protein and reassociation, concurrent with or subsequent to UTP binding. Also possible is active displacement of the RNA back to the primer (UMP) binding site due to competition with the incoming UTP molecule, which is likely to have a higher affinity for this site as a result of extensive triphosphate contacts. Preincubation of the enzyme with either UTP or ³²pUpU before addition of the second substrate has a strong inhibitory effect (Fig. 2C, gels 1 and 2) as compared with reactions initiated by the addition of enzyme to a mixture of both substrates (Fig. 2C, gel 3). Longer RNA substrates, which are more physiologically relevant, may induce conformational changes upon catalysis that facilitate reformation of the apo state. However, the “order of addition” effects observed with a longer RNA substrate resemble those observed with pUpU (Fig. 2 C and D), supporting a model in which the 3′ end of +1 RNA partially dissociates from the active site and reassociates in concert with UTP binding. The base-specific contacts of the 3′ and penultimate nucleosides (Fig. 5) are most likely crucial for such reassociation. Indeed, for a relatively small protein, such as TbTUT4, which does appear to possess an extended RNA binding surface adjacent to the active site, processivity may be achieved by relatively few contacts with 3′ terminal RNA residues.

Structural Bases of Terminal Nucleoside Selectivity.

Although the physiological RNA substrates for TbTUT4 are currently unknown, it is clear that this enzyme is selective for RNAs that contain Us at the 3′ terminal and penultimate positions (Fig. 1). Conversely, recombinant TbRET2 is inactive with an RNA primer bearing six Us at the 3′ end, whereas an RNA with a terminal adenosine, 5[U]A, is extended as efficiently as a substrate bearing six adenosines, 6[A] (Fig. 6A). RNAs that have been extended by a single uridylyl residue are no longer active in the TbRET2-catalyzed reaction. Thus, for TbRET2, as well as for TbTUT4, the terminal nucleoside provides a significant contribution to productive RNA binding due to base-specific contacts. Such specificity likely reflects the purine-rich nature of preedited mitochondrial mRNAs, which are extended by RET2 upon endonucleolytic cleavage (for a review see ref. 25). Structural analysis of the TbTUT4:UTP:UMP ternary complex reveals an interaction between R121 and O4 of the UMP uracil base, thereby explaining both the essential role of this residue for RNA binding (8) and the enzyme's preference for a terminal uridylyl residue. Superposition of this structure with the crystal structure of TbRET2:UTP (14) (PDB ID code 2B51) exposes a potential clash at the O4 position with E424 (RET2) (Fig. 6B), a key residue involved in UTP binding (14). However, modeling an AMP in the same relative position as the bound UMP (as observed in the structure of TbTUT4:UTP:UMP), assuming that the terminal RNA residue is fixed because of maximized stacking with the uracil base and the constraints for the position of the 3′ hydroxyl, suggests an interaction between E424 (TbRET2) and the amino group of the adenine base (Fig. 6C). Furthermore, clashes are evident between active site residues of TbTUT4 and the modeled AMP, suggesting that binding of RNA substrates ending in A is less energetically favorable for this enzyme. Therefore, in addition to coordinating the UTP uracil base, two additional roles for E424 in RET2 are likely: (i) contributing to RNA substrate specificity by energetically favoring an adenosine as the terminal RNA residue and (ii) limiting nucleoside incorporation by RET2 into single-stranded RNA to one U residue.

Fig. 6.

RNA specificity of trypanosomal TUTases. Potential stabilizing interactions are shown as blue dotted lines, and destabilizing electrostatic repulsion and steric hindrance are shown as solid and dotted black arrows, respectively. (A) TbRET2 is highly selective for the terminal nucleoside in a single-stranded RNA substrate. (B) Superposition of the TbTUT4:UTP:UMP and TbRET2:UTP (PDB ID code 2B51) crystal structures. Key residues of TbTUT4 (purple) and TbRET2 (cyan) are indicated. (C) Same as in B but with UMP replaced by a modeled AMP molecule in the same position.

Discussion

The fundamental question of how specific nucleoside triphosphates are recognized by template-independent nucleotidyl transferases has been investigated for several types of enzymes. Structural analysis of the apo and substrate-bound CCA-adding enzymes revealed that both class I and II proteins use hydrogen bonding complementarity between the Watson–Crick edge of the incoming NTP and conserved amino acid residues. In the case of class I CCA-adding enzyme from Archaeoglobus fulgidus (AfCCA), base recognition is also aided by stacking of the incoming NTP on the growing RNA primer, as well as contacts with RNA phosphate groups (26). The eukaryotic nuclear PAP, which is structurally homologous to class I archaeal CCA-adding enzymes, has been cocrystallized with ATP analogs under a variety of conditions (27–29) and subjected to extensive mutational analysis (29, 30). However, a coherent mechanism of ATP selection remains to be established. Furthermore, a highly organized ATP binding site has been reported for vaccinia virus PAP (31). Terminal uridylyltransferases belong to the same, apparently monophyletic group as nuclear PAPs and archaeal CCA-adding enzymes (15) but are most closely related to divergent, “noncanonical” PAPs, such as animal Gld-2 type cytoplasmic PAP, Trf4/5 nuclear surveillance PAPs in S. cerevisiae, and the Cid1-like protein family in S. pombe (1). The high-affinity NTP binding site of trypanosomal TbTUT4 is capable of selective UTP binding versus ATP binding at submicromolar NTP concentrations (8). Because of the conservation of protein–UTP contacts between TbTUT4 and TbRET2, this conclusion is likely to be applicable to other TUTases.

In the present work we demonstrate that, at higher, more physiologically relevant concentrations, all four NTPs can occupy the active site of TbTUT4 with nearly perfect superposition of the phosphate groups. The implication of this finding is twofold: (i) other mechanisms, in addition to NTP binding affinity, are required to discriminate non-UTP substrates by TUTases, and (ii) TUTase-like bidomain core modules are quite promiscuous in NTP binding; a minimal number of mutations is likely to suffice to convert this module into an ATP-specific [noncanonical poly(A)] polymerase or vice versa. The necessity for continuous stacking interactions between a conserved tyrosine side chain, the bound NTP, and the terminal nucleoside base of the RNA primer poses a constraint on the positioning of the NTP base. This may reflect principal differences in NTP selection between TUTases and CCA-adding enzymes. In AfCCA, tRNA forms part of the nucleotide binding site; the apo enzyme is unable to discriminate correct substrates (CTP and ATP) from incorrect ones (UTP and GTP). In TUTases, the enzyme has an intrinsic selectivity for UTP in the absence of RNA, and base-specific contacts are essential for UTP binding (8), suggesting that protein–UTP affinity is required but not sufficient. In the case of ATP or GTP, steric constraints prevent the purine base from participating in these stacking interactions, which is thought to interfere with productive RNA binding. The significant drop in catalytic rates for purine NTPs, which is consistent with the loss of optimal substrate orientation, points to a kinetic component in the selection process. The structure of ATP-bound TbTUT4 suggests that, in TUTase-like PAPs, the aspartate at position 297, an essential residue for uracil-specific hydrogen bonding of TbTUT4 (8), would play no role in polyadenylation, whereas S148, Y189, and N147 remain important for catalysis (Fig. 1).

Selectivity of TbTUT4 and TbRET2 toward the terminal RNA base is likely to be dictated by stabilizing effects of direct hydrogen bonding with conserved residues, as well as destabilizing electrostatic repulsion and steric hindrance effects (Fig. 6). Processive TUTases, such as RET1 (23), TUT3 (7), and TUT4 (8), favor oligouridylyl RNA primers while displaying a lesser specificity toward UTP than RET2 (5, 6). This may be explained by our modeling studies (Fig. 6): in TbTUT4, the arginine at position 141, which is conserved among processive TUTases (1), is locked in a salt bridge with E300, thus precluding participation of the glutamate in the coordination of the crucial water molecule that makes a uracil base-specific contact (8, 14). Furthermore, R141 likely disfavors adenosine binding because of a clash with the exocyclic amino group. In TbRET2, the equivalent position (271) is occupied by a valine instead of an arginine, providing a more spacious binding site and resulting in a change of position of the side chain of the aforementioned glutamate, E424 (E300 in TbTUT4). E424 in TbRET2 thus plays a role in favoring adenosine over uridine as the terminal RNA nucleoside and allows for exquisite UTP specificity. Arginine-121 is essential for TbTUT4 activity (8) and acts as a positive determinant for terminal RNA uracil binding while discriminating against adenosine. The phenylalanine in position 52 of TbTUT4 is conserved for RET1 and TUT3 among all kinetoplastids but is replaced with a smaller cysteine in RET2, perhaps further enhancing binding of a purine base. Our model provides a rationale as to why RET2, acting on a single-stranded RNA substrate, adds only a single uridylyl residue. However, in vivo, some substrates of RET2 are likely to be two double-stranded RNAs linked by purine nucleotides (Fig. 2A). Here the extended 5′ fragment reanneals with guide RNA, thus restoring an optimal RNA substrate for each round of addition. Consequently, a mismatched addition product becomes a single-stranded RNA with U at the 3′ end and therefore a suboptimal substrate for RET2. This self-limiting mechanism may have a role in the overall fidelity of U-insertion RNA editing by minimizing the number of nonguided U-insertions.

Materials and Methods

Uridylyltransferase Activity Assays.

Reactions of purified recombinant proteins with synthetic RNA substrates were performed as described previously for TbRET2 (5) and TbTUT4 (8). The SigmaPlot Enzyme Kinetics software package was used for calculations of K_m, V_max, and standard deviations. Reactions with UMP and UpU were separated on a 20% acrylamide gel and exposed immediately to a phosphor storage screen for 30 min. UMP and UpU were obtained from Sigma (St. Louis, MO). Single-stranded RNA substrates were 6[U] (GCUAUGUCUGUCAACUUGUUUUUU), 6[A], (GCUAUGUCUGUCAACUUGAAAAAA), and 5[U]A (GCUAUGUCUGUCAACUUGUUUUUA). Double-stranded RNAs were used as described in ref. 4.

Crystallization, Data Reduction, and Refinement.

Purified TbTUT4 was concentrated to 5 mg/ml in 10 mM Hepes buffer (pH 7.6), 70 mM KCl, and 0.5 mM DTT and crystallized in the presence of 4 mM MgCl₂ and 25 μM of the respective NTP. Crystallization conditions were 100 mM sodium cacodylate (pH 6.5), 200 mM calcium acetate, and 18% PEG-8000 (Crystal Screen solution 46; Hampton Research, Aliso Viejo, CA) and were carried out at 4°C using the vapor diffusion method. A TbTUT4-UTP cocrystal was soaked in mother liquor containing 10 mM UMP and another in 10 mM UpU for 30 min at 4°C, which were used to derive the structures of the TbTUT4:UTP:UMP ternary complex and TbTUT4:UpU, respectively. All crystals were flash-cooled with liquid N₂ in mother liquor supplemented with 25% glycerol as a cryoprotectant. All x-ray data were collected at the Stanford Linear Accelerator Center and processed and scaled by using d*TREK (32). The structures were solved by refining the original TbTUT4:UTP structure (PDB ID code 2IKF) against the structure factors obtained from each data set while taking care to conserve the test set. Model building and refinement were carried out by using the programs COOT (33) and REFMAC5 (34), respectively.

Abbreviations

TUTase

terminal RNA uridylyltransferase

NTP

nucleoside 5′-triphosphate

PAP

poly(A) polymerase

PDB

Protein Data Bank.