Cleavage of 3′-terminal adenosine by archaeal ATP-dependent RNA ligase

Methanothermobacter thermoautotrophicus RNA ligase (MthRnl) catalyzes formation of phosphodiester bonds between the 5′-phosphate and 3′-hydroxyl termini of single-stranded RNAs. It can also react with RNA with a 3′-phosphate end to generate a 2′,3′-cyclic phosphate. Here, we show that MthRnl can additionally remove adenosine from the 3′-terminus of the RNA to produce 3′-deadenylated RNA, RNA(3′-rA). This 3′-deadenylation activity is metal-dependent and requires a 2′-hydroxyl at both the terminal adenosine and the penultimate nucleoside. Residues that contact the ATP/AMP in the MthRnl crystal structures are essential for the 3′-deadenylation activity, suggesting that 3′-adenosine may occupy the ATP-binding pocket. The 3′-end of cleaved RNA(3′-rA) consists of 2′,3′-cyclic phosphate which protects RNA(3′-rA) from ligation and further deadenylation. These findings suggest that ATP-dependent RNA ligase may act on a specific set of 3′-adenylated RNAs to regulate their processing and downstream biological events.

physiological/optimal polynucleotide substrate is not known. In this report, we show that MthRnl cleaves the adenosine residue from the 3′-OH end and leaves RNAs with 2′,3′-cyclic phosphates. Notably, cleavage is selective, with only a single adenosine residue removed. These findings raise the possibility that the RNA ligase acts as a surveillance/editing enzyme that selectively converts the reactive 3′-OH into a 2′,3′-cyclic phosphate end, thereby regulating their processing and downstream biological events.

MthRn1 generates a novel RNA. Previously MthRnl was shown to circularize a 24-mer unstructured
single-stranded pRNA that migrates 2-nts faster than the input linear substrate on a high percent PAGE gel 29 ( Fig. 1A; unstructured 24-mer pRNA). Toward identifying the optimal substrate for MthRnl ligation activity, we prepared five different 32 P-labeled 21-mer pRNA substrates that mimic hairpin structured RNAs that are expected to form overhangs at the 5′-and/or 3′-end (Fig. 1A) Incubation of recombinant MthRnl with the 3′-(5nt) overhang, dual overhang, 5′-(5nt) overhang or 5′-(7nt) overhang pRNA resulted in the generation of a novel RNA species (RNA*) that migrates between the input linear pRNA and cRNA (Fig. 1A). In the case of the 3′-(5nt) overhang RNA, the majority was converted to RNA*. In the case of the dual overhang RNA, the ratio of RNA*:cRNA formed was 2:1, and with the 5′-(5nt) overhang RNA, it was 1:3. The 5′-(7nt) overhang RNA was a poor substrate for formation of both RNA* and cRNA. Moreover, the 3′-(7nt) overhang RNA did not circularized as efficiently as the other RNA substrates tested. As discussed below, RNA* is only detected from RNAs that contain adenine at the 3′-terminus. Neither the 3′-(7nt) overhang pRNA substrate nor the unstructured 24-mer pRNA, each of which has uracil at the 3′-terminus, were converted to RNA*.
To verify that the observed RNA* was a product of MthRnl activity and not due to bacterial contamination, we subjected recombinant MthRnl to glycerol sedimentation analysis. The RNA* forming activity co-sedimented with the adenylatranferase, RNA-adenylate and RNA circularization activities (Fig. 1B), suggesting that MthRnl is responsible for generating RNA*. Neither T4 RNA Ligase 1 (Rnl1) 31 nor T4 RNA Ligase 2 (Rnl2) 32 was capable of forming RNA* from any of the RNA substrates tested ( Supplementary Fig. S1). We also cloned and purified the MthRnl homolog Thermococcus kodakarensis RNA ligase (TkoRnl), and demonstrated that the recombinant protein is likewise capable of generating RNA*, specifically from RNA with a 3′-adenine ( Supplementary Fig. S2). Thus, the formation of RNA* is likely conserved in archaeal ATP-dependent RNA ligases. The 3′-(5nt) overhang and the dual overhang pRNA substrates were used to further characterize the RNA* generating activity.
MthRnl removes adenosine from 3′-end of the RNA. We hypothesized that RNA* is either a linear pRNA with a modification at the 3′-end or a closed circular RNA that migrates differently from a conventional cRNA in the gel. To distinguish between these two possibilities, reaction products generated by MthRnl were treated with an alkaline phosphatase. The RNA*s generated from both the 3′-(5nt) overhang and dual overhang pRNA substrates were sensitive to alkaline phosphatase treatment, indicating that the 5′-radiolabeled phosphate on RNA* is exposed (Fig. 1C). These results imply that RNA* is generated by removal of a single nucleoside from the 3′-end of pRNA.
As described above, RNA* was observed on pRNA substrates that contain adenine, but not a uracil at the 3′-terminus (Fig. 1A). This specificity was further tested using dual overhang RNA substrates in which the 3′-terminal adenine is replaced with uracil, cytosine or guanine ( Fig. 2A). RNA* was detected only when a pRNA substrate containing a 3′-terminal adenosine was used, implying that MthRnl removes specifically 3′-adenosine. Thus, henceforth we refer to RNA* as pRNA(-3′A).
The 3′-deadenylation activity is independent from RNA ligation activity. Kinetic analysis revealed that the majority of the input 3′-(5nt) overhang pRNA is converted to pRNA(-3′A) in 60 min (Fig. 2B). The initial rate of 3′-deadenylation is ~10 fmol/min of 3′-(5nt) overhang pRNA per pmol of enzyme, which is comparable to the initial rate of RNA circularization; 16 fmol/min of pRNA circularization per pmol of enzyme 30 . The 3′-deadenylation activity with 3′-(5nt) overhang pRNA substrate was ~ 2-fold higher than that with the dual overhang pRNA substrate (Fig. 2C). In the case of the dual overhang pRNA substrate, approximately 50% was converted to pRNA(-3′A) in 60 min, whereas the remaining substrate was converted to an AppRNA at an early time point and subsequently to cRNA. Because pRNA(-3′A) is generated independent of the product of the ligation pathway, pRNA(-3′A) is not likely to be an intermediate of the ligation reaction.
Like the ligation reaction, the 3′-deadenylation activity requires a divalent cation. This requirement was satisfied by magnesium, manganese, cobalt or, to a lesser extent, calcium (Fig. 3A). MthRnl preferentially deadenylated pRNA under alkaline conditions (Fig. 3B). At pH values between 6.5 and 7.0, the dual overhang pRNA substrate was converted to pRNA(-3′A) or cRNA with similar frequency. With increasing pH, the yield of pRNA(-3′A) increased and cRNA levels declined; pRNA(-3′A) was not detectable under pH 5.5. MthRnl can deadenylate substrates over wide range of temperatures, the optimum being 55 °C. This differs from the ligation reaction, which was optimal above 55 °C (Fig. 3C). The level of pRNA(-3′A) declined steadily as the temperature increased above 55 °C, with the formation of cRNA favored. The two activities displayed distinct temperature and pH optima, implying that the ligation and 3′-deadenylation reactions are independent of one another. Nonetheless, the ATP binding site of MthRnl appears to be essential for 3′-deadenylation. Addition of ATP inhibited the MthRnl 3′-deadenylation activity, similar to the ATP inhibitory effect for the ligation activity ( Fig. 3D and Supplementary Figs S1 and S3). An ATP analog, AMPPNP, which can occupy the ATP binding site, also inhibits 3′-deadenylation. In contrast, AMP or other rNTPs did not significantly affect the formation of 3′-deadeylated or circular RNA. Mutant forms of the enzyme that fail to generate EpA (K97A and E151A) or whose ligation activity is reduced (N99A) were defective for the 3′-deadenylation activity ( Fig. 3E and Supplementary Fig. S4) 30 ; only trace amounts of pRNA(-3′A) were detected in the presence of the K97A and N99A mutant forms of the enzyme. On the other hand, MthRnl variants that catalyze EpA formation but are defective for the step 2 or step 3 ligation Reaction without MthRnl (-). Positions of input pRNA, 5′ adenylated RNA (AppRNA), and circularized RNA (cRNA) are indicated. A novel RNA product that migrates between pRNA and AppRNA is denoted by asterisk (RNA*). (B) Sedimentation analysis of MthRnl activities. MthRnl (100 µg) was centrifuged for 20 hrs at 4 °C in a 15-30% glycerol gradient as described previously 29 . The amount of ligation product formed was gauged by signal intensity as a proportion of total RNA product (y-axis). The amount of EpA formed was gauged by the signal intensity of radiolabeled MthRnl (yy-axis). The peaks of the marker proteins, catalase, BSA and cytochrome c, are indicated. (C) RNA* has an exposed 5′-phosphate. Standard ligation reaction mixtures (20 µl) containing 0.45 µg of MthRnl and 1 pmol of the indicated pRNA was incubated at 55 °C for 15 min. Reaction was terminated by addition of EDTA, after which the samples were incubated with or without 0.2 unit of alkaline phosphatase (AP) at 37 °C for 30 min. Sample without enzyme served as negative control (-). Positions of pRNA substrate, AppRNA, cRNA and RNA* are indicated. reaction (T117A and R118A) were capable of forming pRNA(-3′A). We speculate that adenosine at the 3′-end of the RNA is recognized by the active site that recognizes the ATP in step 1 of the ligation reaction.
Substrate specificity of 3′-RNA deadenylation. We previously showed that MthRnl can circularize single-stranded DNA 29 . To determine whether MthRnl is additionally capable of removing deoxyadenosine, we prepared RNA substrates in which the 3′-terminus was replaced with 2′-deoxyadenosine ( Fig. 4A; rGdA-3′) and cordycepin ( Fig. 4A; rGrA(-H)-3′). MthRnl did not remove 2′-deoxyadenosine (Fig. 4A, lane 4 and lane 8), but did remove cordycepin, albeit less efficiently than from an RNA that contains ribose at the 3′-end (Fig. 4A, lane 10). We also found that MthRnl was unable to deadenylate the substrate when the penultimate nucleotide was replaced with a deoxynuclotide (dGrA-3′; Fig. 4A, lane 6). We next prepared additional 21-mer DNAs; in one, the sequence was identical to that of the dual overhang RNA substrate ( Fig. 4B: dGdA-3′); in two others, a DNA/RNA hybrid substrate consisted of either 20 deoxynucleotides with a single adenosine ribonucleotide at the 3′-end ( Fig. 4B: dGrA-3′) or 19 deoxynucleotides with 2 ribonucleotides at the 3′-end (Fig. 4B: rGrA-3′). MthRnl could only deadenylate substrates in which the penultimate nucleotide is a ribonucleotide. These results imply that the 3′-deadenylation activity requires 2′-OH at both the terminal adenosine and the penultimate nucleotide, but that the 3′-OH on the terminal adenosine is not strictly required for the reaction chemistry.
The fact that MthRnl removes specifically 3′-rA raises the question of whether it can processively deadenylate RNA containing multiple adenosines. We reasoned that if MthRnl is capable of removing two consecutive adenosines, the product should migrate faster than pRNA(-3′A). However, this was not the case. The 3′-(5nt) overhang RNA substrate with two consecutive 3′-adenosines did not generate a shorter 3′-deadenylated species (Fig. 1A). To confirm this finding, we replaced guanosine at the penultimate position, with adenosine in the 3′-(5nt) overhang substrate (Fig. 4C: rGrA-3′, lane 4) and the dual overhang substrate that contains two consecutive adenosines at the 3′-end (Fig. 4C: rArA-3′, lane 2). No difference in the mobility of the 3′-deadenylated product was detected. We therefore conclude that MthRnl can cleave only a single adenosine residue from the 3′-end and, once the adenosine is cleaved, the pRNA(-3′A) cannot be ligated to form the cRNA.

Discussion
During the course of this study to determine the optimal RNA substrate for MthRnl ligation activity, we discovered that MthRnl possesses a 3′-deadenylation activity that selectively cleaves an adenosine residue from the 3′-terminus of an RNA, to yield a 2′,3′-cyclic phosphate end. The 3′-deadenylation activity is specific to RNA, as it requires two ribonucleotides at the 3′-OH end, whereas the rest of the nucleotides can be replaced by DNA. Conversion of the 3′-OH end to a cyclic phosphate prevents ligation to the 5′-PO 4 end. Once the RNA is 3′-deadenylated, subsequent deadenylation is inhibited due to the presence of a 2′,3′-cyclic phosphate end. Similar to the ligation activity, the 3′-deadenylase activity requires a divalent cation, and excess ATP inhibits the reaction. MthRnl mutant proteins that fail to form the ligase-AMP complex are defective for 3′-deadenylation. In contrast, Thr-117 and Arg-118, which do not affect formation of the ligase-AMP complex (step 1) but are required for formation of the phosphodiester bond (step 3), did not affect the 3′-deadenylation activity 30 . The fact that ATP can inhibit the 3′-deadenylation reaction suggests that the 3′-adenosine on the RNA may compete for the same catalytic site for binding of the ATP. We propose that 3′-adenosine is recognized by the ATP binding pocket in the MthRnl and catalyzes the deadenylation through a transesterification reaction, by stimulating the penultimate ribose 2′-OH to attack the adjacent 3′-5′ phosphodiester to form a 2′,3′-cyclic phosphate and releasing adenosine (Fig. 6E). The active site Lys-97 is not likely involved in transfer of AMP to the 3′-OH end of the substrate, because 3′-adenylated intermediate, pRNA(+pA), was not detected in the reaction, and traced amounts of deadenylated RNA can be detected by K97A mutant protein. Lys-97 may stabilize and/or neutralize the charge on the phosphate between the 3′-adenosine and penultimate nucleoside. The role of divalent cation for 3′-deadenlyation is not clear. The divalent cation does not appear to be required for substrate binding and overall protein folding (Supplemental Fig. S6), or activating the water molecule as a nucleophile, but it may be important for coordinating the active site conformation. With respect to formation of a 2′,3′-cyclic phosphate by removal of a specific nucleoside from the 3′-end, MthRnl appears to have properties similar to those of the Usb1/Mpn1 3′-5′ exonuclease, which trims the U6 snRNA tail. However Usb1/Mpn1 is capable of trimming multiple nucleosides from the 3′-end of the phosphate 8 . Consistent with this difference, the crystal structure of human Usb1 identified that enzyme as a member of the LigT-like superfamily of 2 H phosphoesterases, proteins to which MthRn1 has no structural similarity 7,30 .
Any RNA that terminates with an adenine is a potential substrate for MthRnl 3′-deadenylation. MthRnl efficiently removes 3′-adenosine from RNAs that are single-stranded or have a stem-loop structure with a 3′-single strand overhang. However, it is less efficient in the cases of RNAs that bear a recessed 3′-end, consistent with the notion that the 2′-OH on the penultimate nucleoside needs to be exposed to allow nucleophilic attack of the adjacent adenylate. We also note that the 5′-adenylation activity of MthRnl was generally efficient in the context of a stem-loop RNA bearing a 5′-overhang, but not in the case of one with a recessed 5′-end. The rates of 5′-adenylation and 3′-deadenylation were comparable in RNAs with both 5′-and 3′-overhangs. These results suggest that MthRnl prefers a substrate with an unpaired /unstacked single-stranded RNA end. Whether MthRnl adenylates the 5′-PO 4 end for ligation or cleaves the 3′-adenosine likely depends on the accessibility of each RNA terminus.
The present study raises the interesting prospect that RNA ligation might not be the only biochemical pathway for MthRnl. We speculate that MthRnl acts on a specific set of 3′-adenylated RNAs to regulate their processing and downstream biological events. RNAs that are exclusively terminated by 3′-adenosine include tRNAs and mRNAs. In the final step of tRNA maturation, the 3′-end is modified by the addition of a -CCA OH group, to which the amino acid is attached. The -CCA sequence protrudes from the acceptor stem helical structure of the tRNA as a single-stranded motif, and is thus a potential substrate for MthRnl 3′-deadenylation. MthRnl may regulate the amount of charged and uncharged tRNA in the cell, as removal of 3′-adenosine and formation of the cyclic end is expected to prevent aminoacylation. Intriguingly, analysis of tRNAs in humans identified a subset that lack 3′-adenosine and terminate in a 2′,3′-cyclic phosphate 33 . This suggests that a 3′-deadenylase activity could also be present in mammalian cells. In some archaea, including methanogens, heteropolymeric poly(A)-rich tails are synthesized and removed by the exosome 34,35 . As in the case of the mammalian U6 snRNA, the formation of 2′,3′-cyclic phosphates could preclude the elongation of poly(A)-rich tails and protect them from degradation machinery from regulating stability of the mRNA. It is also plausible that MthRnl acts as a surveillance enzyme that prevents undesirable intramolecular RNA circularization and intermolecular ligation with the 5′-PO 4 RNA, by converting the reactive 3′-OH end into a nonreactive 2′,3′-cyclic phosphate. MthRnl may also function in forming a 2′,3′-cyclic phosphate ends that serve as substrates for subsequent ligation with 5′-OH RNA by RtcB for a production of a novel RNA species, or generating an adenylated capped RNA (5′-AppRNA) with a 2′,3′-cyclic phosphate end. In summary, our findings raise possibilities for unexpected mechanisms whereby the MthRn1 could act as a surveillance or editing enzyme, to selectively remove the 3′-adenosine of an RNA to convert the reactive 3′-hydroxyl group into 2′,3′-cyclic phosphate and thereby regulate its metabolism.

Recombinant RNA ligases.
Plasmids that encode wild-type and mutant MthRnl were transformed into E. coli BL21(DE3). MthRnl production was induced with IPTG, and His tagged-MthRnl proteins were purified from soluble bacterial extracts by Ni-agarose chromatography as described previously 30 . His tagged-T4 RNA Ligase 1 32 and T4 RNA Ligase 2 33 were produced in E. coli and purified as described. Protein concentrations were determined with the BioRad dye reagent, using bovine serum albumin (BSA) as the standard.