Polynucleotide 3′-terminal Phosphate Modifications by RNA and DNA Ligases

Background: Classic RNA ligases join 5′pRNA to RNA3′OH. Results: Thermophilic RNA ligases were able to modify ssDNA or RNA with a 3′-phosphate. Conclusion: Thermophilic ligases form RNA 2′,3′-cyclic phosphate and ssDNA3′pp5′A. Significance: Thermophilic RNA ligases duplicate the enzymatic function of RtcA and may have an important role in nucleic acid 3′-phosphate biology in addition to conventional ligation of 5′pRNA. RNA and DNA ligases catalyze the formation of a phosphodiester bond between the 5′-phosphate and 3′-hydroxyl ends of nucleic acids. In this work, we describe the ability of the thermophilic RNA ligase MthRnl from Methanobacterium thermoautotrophicum to recognize and modify the 3′-terminal phosphate of RNA and single-stranded DNA (ssDNA). This ligase can use an RNA 3′p substrate to generate an RNA 2′,3′-cyclic phosphate or convert DNA3′p to ssDNA3′pp5′A. An RNA ligase from the Thermus scotoductus bacteriophage TS2126 and a predicted T4 Rnl1-like protein from Thermovibrio ammonificans, TVa, were also able to adenylate ssDNA 3′p. These modifications of RNA and DNA 3′-phosphates are similar to the activities of RtcA, an RNA 3′-phosphate cyclase. The initial step involves adenylation of the enzyme by ATP, which is then transferred to either RNA 3′p or DNA 3′p to generate the adenylated intermediate. For RNA 3′pp5′A, the third step involves attack of the adjacent 2′ hydroxyl to generate the RNA 2′,3′-cyclic phosphate. These steps are analogous to those in classical 5′ phosphate ligation. MthRnl and TS2126 RNA ligases were not able to modify a 3′p in nicked double-stranded DNA. However, T4 DNA ligase and RtcA can use 3′-phosphorylated nicks in double-stranded DNA to produce a 3′-adenylated product. These 3′-terminal phosphate-adenylated intermediates are substrates for deadenylation by yeast 5′Deadenylase. Our findings that classic ligases can duplicate the adenylation and phosphate cyclization activity of RtcA suggests that they have an essential role in metabolism of nucleic acids with 3′-terminal phosphates.

Although 5Ј-phosphorylated oligonucleotides are common intermediates in nucleic acid biochemistry, terminal 3Ј-phosphorylated DNA and RNA can also be produced through a variety of chemical and biochemical reactions. Examples of RNases that generate a 3Ј-phosphate include RNase A, RNase I, and the self-cleaving ribozymes. Alkaline hydrolysis of RNA is a common chemical method used to generate RNA with a 3Ј-phosphate. DNase II, which is involved in the degradation of DNA during apoptosis, generates a DNA 3Јp cleavage product. Com-mon nucleases, such as micrococcal nuclease, generate DNA and RNA breaks with 3Ј-phosphate or 2Ј,3Ј-cyclic phosphate (1). When topoisomerase I is trapped by inhibitors, it can be removed from DNA by tyrosyl-DNA-phosphodiesterase I (Tdp1), leaving a DNA 3Ј-phosphate end. Topoisomerase I has also been shown to cleave at a single ribonucleotide embedded in duplex DNA, giving rise to DNA nicks with a ribo-2Ј,3Јcyclic phosphate end. Tdp1 can convert phosphoglycolate to phosphate termini on 3Ј overhangs of DNA double strand breaks (2). Bleomycin, neocarzinostatin, and ionizing radiation can also induce DNA damage with 3Ј-phosphate ends. After 5-methylcytosine removal, the Arabidopsis DNA glycosylase/ lyase ROS1 cleaves the DNA backbone, and the product has a single-nucleotide gap flanked by 3Ј-and 5Ј-phosphate termini. For a more comprehensive review on 3Ј-terminal phosphate formation and repair, refer to Refs. 3-5. Most RNA and DNA repair reactions, including polymerization or ligation, require a free terminal 3Ј-hydroxyl and cannot act on 3Ј-phosphorylated polynucleotides. Until recently, our understanding of the removal of 3Ј-phosphates and their derivatives from modified 3Ј ends was thought to be performed mostly by nonspecific phosphomonoesterases (phosphatases and enzymes containing phosphatase activity) or by nucleotide excision. Most phosphatases cannot remove a 2Ј,3Ј-cyclic phosphate group from the 3Ј end of RNA. The exceptions are the multifunctional plant and fungal tRNA ligases that convert a RNA 2Ј,3Ј-cyclic phosphate to RNA 2Ј-phosphate, followed by phosphorylation of the 5Ј end and then ligation to generate an unconventional 2Ј-phosphate-3Ј,5Ј-phosphodiester bond at the splice junction (6). Bacteria and archaea contain a 2Ј,5Ј RNA ligase that can seal 2Ј,3Ј-cyclic phosphate and 5Ј-hydroxyl of RNA. Recently, a chloroplast 2Ј,5Ј RNA ligase that can also use cyclic phosphates as a substrates was found (7). T4 polynucleotide kinase, T4 PNK, 2 is another multifunctional enzyme that has 2Ј,3Ј-cyclic phosphodiesterase activity that generates a terminal 3Ј-phosphate that is then removed by its phosphatase. The enzyme is also able to phosphorylate the 5Ј end of the RNA or DNA (8).
However, there are other emerging biochemical pathways for dealing with 3Ј-terminal phosphate: its activation, modification, and direct involvement in ligation. A major discovery was the identification of RtcB ligase as a noncanonical GTP-dependent RNA ligase that joins the 3Ј-phosphate of a donor RNA to 5Ј-hydroxyl of acceptor RNA. In recent studies, the mechanism of 3Ј-phosphate ligation of RNA (9 -11) has been well described (12)(13)(14). The 3Ј-phosphate in this ligation is directly involved in the reaction. It is first activated by the transfer of GMP from the RtcB-GMP complex to form the 3Ј-guanylated intermediate RNA 3Ј pp 5Ј G. This intermediate is then ligated to the 5Ј-hydroxyl of an RNA acceptor to form a phosphodiester bond or to the adjacent 2Ј-OH to form a cyclic phosphate. Moreover, recent studies showed potential for this activity to be involved in DNA repair (15,16).
Terminal 3Ј-phosphates are also substrates for ATP-dependent 3Ј-phosphate cyclase (RtcA) (9,17). RtcA converts 3Ј-terminal phosphate on RNA to 2Ј,3Ј-cyclic phosphate using a tripartite reaction similar to the mechanism of polynucleotide ligases (18). First (Reaction 1), the RtcA protein self-adenylates by reacting with ATP. Second (Reaction 2), the adenyl group is transferred to 3Ј-terminal phosphate of RNA, creating an activated RNA 3Ј pp 5Ј A intermediate. Finally (Reaction 3), the phosphate is internally ligated to the adjacent 2Ј-hydroxyl with loss of AMP.
The resulting 2Ј,3Ј-cyclic phosphate can serve as a substrate for RtcB ligase or certain plant and fungal tRNA ligases that, additionally, have cyclic phosphodiesterase activity (6,19). When the 2Ј-hydroxyl of the 3Ј-terminal ribose is blocked by methylation or the terminal sugar is 2Ј-deoxyribose (Reaction 4), the intermediate nucleic acid-3Ј pp 5Ј A intermediate accumulates (9).
Enz-p5ЈA ϩ DNA 3Јp 3 DNA 3Ј pp5ЈA REACTION 4 Although the chemistry among traditional ATP-dependent ligases is identical, they have diverse specificities with respect to polynucleotide substrates. Most RNA ligases prefer singlestranded RNA, but T4 Rnl2 can also recognize and repair nicks in double-stranded RNA (21). Some thermophilic RNA ligases do not strongly discriminate between RNA and single-stranded DNA (ssDNA) (22,23); they are able to ligate either substrate. Most DNA ligases are specific for dsDNA, although Chlorella virus DNA ligase, in addition, recognizes and ligates DNA/RNA hybrids where nicked DNA was splinted with RNA (24). Recently it has been found that RtcA, in addition to 3Ј-phosphate modification, can also adenylate 5Ј-terminal phosphates of RNA and DNA (25). The discovery that RtcA had activity on either the 3Ј-or 5Ј-phosphate of RNA or DNA prompted us to investigate the possibility that classical RNA ligases, which require a 5Ј-phosphate for ligation, could modify RNA with a 3Јp. We have found previously that the RNA ligase from the thermophile Methanobacterium thermoautotrophicum, MthRnl, is very efficient in adenylation of oligonucleotides with a 5Ј phosphate (26). We discovered that MthRnl has the ability to convert RNA3Јp to RNA 2Ј,3Ј cyclic phosphate and also adenylate ssDNA3Јp. We extended this study to investigate the modification of 3Ј-phosphorylated nicked DNA substrate with RNA and DNA ligases from bacteria and archaea and bacterial and archaeal phages as well as from an eukaryotic virus.
The RtcA gene was amplified from genomic DNA of E. coli K-12 strain NEB Turbo (New England Biolabs) using the PCR primers CATCATATGAAAAGGATGATTGCGC (forward) and TTGGATCCTCATTCAATGCTCACCC (reverse) with NdeI and BamH I restriction sites, respectively (underlined). The PCR product was cloned, expressed, and purified as described above for archaeal RNA ligases.
Substrates-The oligonucleotides used in this study were synthesized by Integrated DNA Technologies. The 3Ј-phosphate donor substrates were as follows: CTGTAGGCACCATCAAT-p (DNA17p); FAM-CTGTAGGCACCATCAAT-p (FAM-DNA17p); FAM-CUGUAGGCACCAUCAAU-p (FAM-RNA17p); and FAM-CUGUAGGCACCAUCAAU-2Ј (OCH 3 ) 3Ј p (FAM-RNA17-(OMe)p). FAM is carboxyfluorescein. The 5Ј-phosphate donor substrate was pCTGTAGGCACCATCAAT-NH 2 (pDNA17-NH 2 ). Double-stranded DNA substrate with a single-strand 3Ј-phosphate break was created by annealing three DNA oligonu it on a heat block that was turned off (Van Waters and Rogers) for 2 h to reach room temperature. Completion of the annealing reaction was confirmed by BanI restriction endonuclease (the recognition site is underlined) digestion at 25°C for 90 min. After boiling for 2 min, the reaction analyzed on a 15% urea-Tris borate-EDTA denaturing polyacrylamide minigel (Invitrogen).

Assays
Terminal phosphate modifications of oligonucleotides with RNA and DNA ligases or RtcA, which are labeled as 1 st enzymes in Fig. 1, were performed in a 10-l volume. The reactions contained 5-10 pmol of either 3Ј-or 5Ј-terminal phosphorylated substrate, 10 pmol of enzyme, 500 M ATP in buffer containing 10 mM BisTris propane HCl (pH 7.0), 10 mM Mg ϩ2 , and 1 mM DTT in 200-l PCR tubes. The reactions were incubated in an S1000 Thermal Cycler (Bio-Rad) for 30 min at 65°C for thermophilic RNA ligases or 25-37°C for other enzymes, followed by heat inactivation at 90°C for 3 min before next step analysis. All parameters that are different from the standard reaction conditions are indicated in the figure legends.

Analysis of Reaction Products
To determine which enzymatic modifications, if any, occurred to an oligonucleotide 3Ј-phosphate, the products were either assayed directly or treated with a second enzyme: 5 units of AnP supplemented with 25 M Zn ϩ2 , 5 units of T4 PNK, or 5 units of yeast 5ЈDeadenylase. All reactions were incubated at 37°C for 60 min. Parameters different from the standard reaction conditions are described in the figure legends. The presence of a 2Ј,3Ј-cyclic phosphate on RNA can be determined by the differential activity of two phosphatases. T4 PNK has 3Ј-phosphatase activity and 2Ј-phosphodiesterase activity and can dephosphorylate 2Ј,3Ј-cyclic phosphate. Antarctic phosphatase has 3Ј-phosphatase activity but is unable to cleave the 2Ј,3Ј-cyclic phosphate. If the 3Ј-phosphate is adenylated, producing RNAppA or DNAppA, they are identified by slower mobility on denaturing polyacrylamide gel. The adenylated oligonucleotides are resistant to both phosphatases but sensitive to yeast 5ЈDeadenylase.
Reactions were stopped by the addition of 5 l of formamide loading buffer to 5 l of a reaction mixture, followed by separation on a 15% urea-Tris borate-EDTA denaturing polyacrylamide (urea-PAGE) minigel (Invitrogen). The FAM-labeled products of the reactions were visualized by fluorescence scanning on a Typhoon 9400 Imager (GE Healthcare). Unlabeled reaction products were stained with SYBR Gold (Invitrogen) and visualized using AlphaImager HP (Alpha Innotech). Radioactive samples were visualized using Storage Phosphor Screen GP (Kodak) and scanned on a Typhoon 9400 Imager (GE Healthcare). Aliquots containing 10 fmol of the FAM-labeled products of various reactions were subjected to fragment analysis using capillary electrophoresis on 3730xl DNA Analyzer (ABI). Chromatograms were aligned using 15-, 20-, and 25-nucleotide dsDNA standards.

RESULTS
Archaeal Mth RNA Ligase Has RNA 2Ј,3Ј-Phosphate Cyclization Activity-We expanded our study of the substrate specificity of MthRnl to include RNA 3Ј-phosphorylated RNA. A 17-nucleotide RNA oligomer labeled with a fluorophore at the 5Ј end (FAM-RNA17p) and a derivative where the 3Ј-terminal ribonucleoside was 2Ј-O-methylated (FAM-RNA17(OMe)p) were used as substrates for MthRnl. The activity of MthRnl was compared with the known 3Ј-phosphate-modifying activity of RtcA. After reaction with ligase and digestion with secondary enzymes, as described under "Experimental Procedures," the oligonucleotide products were analyzed by gel shift analysis on denaturing 15% urea-PAGE (Fig. 1A) or, for better resolution, by capillary electrophoresis (fragment analysis). The reaction products were aligned to internal fragment analysis DNA size standards of 15, 20, and 25 nucleotides (Fig. 1B) to allow comparison between assays. Treatment of 3Ј-phosphorylated RNA with MthRnl in the presence ATP resulted in a small shift in band mobility on PAGE (Fig. 1A, compare lanes 4 and 7), which was clearly evident by fragment analysis (Fig. 1B, compare panels 1 and 4).
Two phosphatases with different specificities were used to characterize the modifications of RNA 3Ј-phosphate by MthRnl. Treatment with T4 PNK produced a slower migrating band on the gel, but treatment with AnP did not (Fig. 1A, lanes  5 and 6). The use of fragment analysis gave a cleaner separation of these reaction products (Fig. 1B, panels 5 and 6). No change in mobility was observed after AnP treatment (Fig. 1B, compare panels 4 and 5), but, after T4 PNK treatment, a slower migrating band was observed (Fig. 1B, compare panels 4 and 6). Dephosphorylation of the starting FAM-RNA17p oligonucleotide with either AnP or T4 PNK produced a slower migrating band on the gel (Fig. 1A, lanes 2 and 3). This dephosphorylated band has the same mobility as FAM-RNA17p reacted with MthRnl and then treated with T4 PNK (Fig. 1B, panel 6). We concluded that this product was the dephosphorylated oligonucleotide FAM-RNA17. Because the product formed by the reaction of MthRnl with FAM-RNA17p was resistant to dephosphorylation by AnP but was converted to the dephosphorylated oligonucleotide by T4 PNK, we concluded that it was RNA 2Ј,3Ј-cyclic phosphate.
To further confirm that the MthRnl product is RNA 2Ј,3Јcyclic phosphate, we performed the same reactions with RNA 3Ј-phosphate cyclase from E. coli, RtcA. These reactions were performed under identical conditions as the MthRnl reaction but at a lower temperature to match the temperature optimum of RtcA (25). Fragment analysis revealed that the product generated by RtcA (Fig. 1B, panel 13) migrated identically to the MthRnl reaction product (Fig. 1B, panel 4). The RtcA reaction product of FAM-RNA17p was resistant to digestion by AnP but sensitive to digestion by PNK. This is identical to what was observed with the MthRnl reaction products (Fig. 1B, compare  panels 5 and 6 and 14 and 15).
When the 3Ј-terminal 2Ј-OMe analog FAM-RNA17(OMe)p was used as a substrate, MthRnl produced a slower migrating band that was approximately one nucleotide larger than the starting material on a 15% denaturing urea-PAGE gel (Fig. 1A,  compare lanes 11 and 14). This product was not sensitive to T4 PNK or AnP phosphatases. The oligo showed no change in mobility by PAGE after treatment with either phosphatase (Fig.  1A, compare lanes 12 and 13 to lane 11), and no change was observed by fragment analysis (Fig. 1B, panels 10 -12). RtcA treatment of the 2Ј-OMe substrate also generated a peak of the same mobility as the sample treated with MthRnl and was resistant to both phosphatases (Fig. 1B, panels 16 -18). These results demonstrate that MthRnl and RtcA have the same 3Ј-phosphate adenylation activity, producing RNA( 2Ј OMe) 3Ј pp 5Ј A as a product (29). MthRnl and RtcA both produced the same 2Ј,3Ј-cyclic phosphate from FAM-RNA17p and the same RNA( 2Ј OMe) 3Јpp 5Ј A product from the 2Ј-O-methylated derivative. These results suggest that MthRnl and RtcA share the same mechanism for the cyclization of 3Ј-phosphorylated RNA oligonucleotides through a RNA 3Ј pp 5Ј A intermediate, followed by internal esterification with the adjacent 2Ј-hydroxyl. MthRnl, Enzymatic Modifications of RNA 3p and DNA 3p NOVEMBER 28, 2014 • VOLUME 289 • NUMBER 48 similar to RtcA, also did not produce any detectable ligation product in the reaction of FAM-RNA17p with free 5Ј-hydroxyl of various RNA substrates in the presence of ATP, GTP, Mg ϩ2 , or Mn ϩ2 (data not shown).
To address the question whether similar 3Ј-phosphate modification of RNA was a common property of other RNA ligases, we tested T4 RNA ligases 1 and 2 for activity on FAM-RNA17p and RNA17( 2Ј OMe)p. These reactions were performed with a molar excess of enzyme as in the RtcA assays. As shown in Fig.  1B, panels 19 -26, under these conditions, neither T4 RNA ligase 1 nor T4 RNA ligase 2 modified RNA with a 3Ј-phosphate. After the reaction, the mobility of the oligonucleotide was unchanged. In a previous report, however, T4 RNA ligase 1 has been shown to catalyze 3Ј-phosphate cyclization at a very slow rate (30).
Thermophilic RNA Ligases Modify Terminal 3Ј-Phosphate of ssDNA-Thermophilic RNA ligases have been shown to catalyze the ligation of 5Ј-phosphorylated ssDNA substrates in addition to RNA substrates (22,26). On the basis of our results in the previous section, it was hypothesized that these ligases could also modify terminal 3Ј-phosphate on ssDNA oligonucleotides through the enzymatic transfer of adenosine monophosphate. This activity has been reported for T4 Rnl1. However, it required several days of incubation (30). We first tested three thermophilic RNA ligases, the archaeal enzyme MthRnl, the bacterial enzyme TVa, and a phage-encoded enzyme, TS2126, with a 3Ј-phosphorylated DNA 17-mer (ssDNA17p) and analyzed the reaction products by gel shift on denaturing 15% PAGE. As shown in Fig. 2A, all three enzymes formed a product with a shift approximately one nucleotide longer than for 60 min, heat-inactivated at 90°C for 5 min, and treated with 10 g of proteinase K at 37°C for 60 min. The products were analyzed on 15% urea-PAGE and visualized with SYBR Gold staining. Positions of a substrate (ssDNA17p) and an adenylated product (ssDNA17ppA) are indicated on the right. Positions of oligonucleotide markers are indicated on the left. C, control. B, pH dependence of ␣-[ 32 P]AMP incorporation into DNA17p using 5 pmol of a DNA17p substrate, 10 pmol of MthRnl under the conditions described for A (except for the indicated variable pH), and supplement with 1 Ci of ␣-[ 32 P]ATP. B I, reaction products were treated and separated as described for A. B II, radioactive products visualized by PhospoScreen scanning. Mr is the oligonucleotide molecular weight marker. C, comparison of 5Ј-and 3Ј-phosphorylated ssDNA adenylation with MthRnl. Reactions were performed as described in A using 5 pmol of pDNA17-NH 2 or DNA17p and variable amounts (5.000 -0.625 pmol) of enzyme. The molar ratio of substrate to enzyme (S/E) is listed at the bottom of the gel. D, comparison of 5Ј-and 3Ј-phosphorylated ssDNA adenylation with RtcA. Reactions were performed and analyzed as described for MthRnl in C but at 37°C with the variable substrate-to-enzyme ratios indicated at the bottom. E, deadenylation of 3Ј-and 5Ј-phosphate-modified ssDNA with yeast 5ЈDeadenylase. 5 pmol of DNA17ppA and AppDNA17-NH 2 , the products of MthRnl modification described in legend for A, were treated with serially diluted 5ЈDeadenylase at 25°C for 30 min. The products were analyzed by gel electrophoresis as described in A. Variable substrate-to-enzyme ratios are indicated at the bottom. F, K97A MthRnl mutant activity assays using 5 pmol ssDNA17p substrate were performed and analyzed as described for wild-type enzyme in C.
the starting oligonucleotide substrate. We further analyzed this activity in detail using MthRnl. The product of MthRnl treatment of DNA17p was resistant to both AnP and PNK phosphatases (data not shown). To confirm the identity of the product as DNA17 3Ј pp 5Ј A, ␣-[ 32 P]ATP was used in the reaction. Fig. 2B II shows the incorporation of radioactivity into the adenylated product with a pH optimum near 7. These results were consistent the product being 3Ј-adenylated DNA, ssDNA 3Ј pp 5Ј A. Even though the ssDNAp substrate used for these assays (ssDNA17p) is not blocked at the 5Ј end and, thus, can serve as a 5Ј-OH acceptor, no ligation products were observed after incubation with the RNA ligases.
The activity of MthRnl was tested on two ssDNA oligonucleotides with identical sequences, one a 5Ј-phosphorylated substrate (with the 3Ј-blocked amino terminus to prevent ligation), pDNA17-NH 2 , and a second substrate the 3Ј-phosphorylated ssDNA17p. Both substrates were converted to adenylated products, and these activities were compared side-by-side in identical conditions with serial dilutions of the enzyme. As shown in Fig. 2C, MthRnl possessed nearly identical 3Ј-and 5Ј-adenylation activities. RtcA is also able to adenylate the 5Ј-phosphate of ssDNA (25). Comparison of adenylation activity of RtcA with 5Ј-and 3Ј-phosphorylated ssDNA with the same sequence showed similar activity (Fig. 2D). When MthRnl activities were compared with RtcA, we found that both enzymes have similar specific activity on the 5ЈpDNA and DNA3Јp substrates (Fig. 2, C and D).
Additional confirmation that the product was ssDNA17 3Ј pp 5Ј A was made by treatment with yeast 5ЈDeadenylase, which can resolve the 5Ј-App group to 5Ј-phosphate (31). The resulting product was approximately one nucleotide shorter (Fig. 2E), migrated identically to the ssDNA17p starting oligonucleotide, and was AnP and PNK phosphatase-sensitive, which suggested that the deadenylated product was ssDNA17p (data not shown). Thus, yeast 5ЈDeadenylase can also convert 3ЈppA groups to 3Ј-phosphates, a newly observed activity for this enzyme. 5ЈDeadenyalse appears to be about five times more active on 5Ј-adenylated DNA compared with 3Ј-adenylated DNA substrates. The mechanism of DNA 3Ј-phosphate adenylation by MthRnl is likely the same as 5Ј-phosphate DNA adenylation because the reaction for both substrates is blocked when the catalytically inactive mutant (K97A) of MthRnl was used (Fig. 2F) (27).
Adenylation of 3Ј-Phosphorylated Nick in Duplex DNA-We next assayed the activity of several DNA and RNA ligases using dsDNA substrates that contained a nick with a terminal 3Ј-phosphorylated nucleotide at the ligation junction. A DNA with a single-stranded break was created by annealing three oligonucleotides as described under "Experimental Procedures" (depicted in Fig. 3A). The 3Ј-phosphorylated 17-deoxy- Enzymatic Modifications of RNA 3p and DNA 3p NOVEMBER 28, 2014 • VOLUME 289 • NUMBER 48 nucleotide oligomer contained a 6-FAM fluorescent label at the 5Ј-end (FAM-DNA17p). Completion of the FAM-DNA17p hybridization into the double-stranded form was confirmed by restriction endonuclease BanI digestion, which produced a sixnucleotide-long product (FAM-DNA6) that was detected by a fluorescent scan of a denaturing PAGE (Fig. 3A). Because of the short length of the DNA duplex it was expected that the duplex would melt at the standard reaction temperatures of the thermophilic ligases. Therefore, all reactions with this substrate were performed at 25°C.
When T4 DNA ligase was tested on the nicked dsDNA substrate, a FAM-labeled product one nucleotide longer than the starting material was observed. This shift is very similar to products produced with MthRnl and RtcA using ssDNA. We concluded that T4 DNA ligase indeed adenylated the 3Ј-phosphorylated nick (Fig. 3C, lane 2). Neither T3, T7, nor PBCV1 DNA ligases have detectable 3Ј-adenylation activity in this assay (Fig. 3C, lanes 3-5). In contrast to thermophilic RNA ligases, none of the tested DNA ligases react with 3Ј-phosphorylated single-stranded DNA (Fig. 3D). DNA ligases are very specific for double-stranded substrates and have very poor activity on ssDNA substrates, indicating the lack of single stranded substrate in the reaction conditions with duplex DNA.
The thermophilic RNA ligases and RtcA were assayed on nicked dsDNA3Јp substrate (Fig. 3C, lanes 6 -9). RtcA was able to adenylate a 3Ј-phosphorylated nick in the double-stranded DNA substrate (Fig. 3C, lane 6), a novel activity for this enzyme. A previous report found that RtcA can adenylate a 5Ј-phosphate at a nick in dsDNA (25). This extends the range of DNA substrates that can be modified by RtcA to include either 5Ј-or 3Ј-phosphates on dsDNA or ssDNA (Fig. 2D). With the 3Ј-phosphorylated nicked dsDNA substrate, neither MthRnl nor TS2126 RNA ligases had significant 3Ј-adenylation activity despite overnight incubation (Fig. 3C, lanes 7 and 9). Although the lack of adenylation activity on this double stranded substrate was most likely due to substrate specificity, we cannot rule out low activity at low temperature, which was 40°C below optimal. However, the putative thermophilic RNA ligase from T. ammonificans, TVa, was active with respect to the adenylation of 3Ј-phosphorylated nicked DNA (Fig. 3C, lane 8). None of the tested ligases, RNA or DNA, displayed any nick-sealing activity on this 3Ј-phosphorylated dsDNA substrate.

DISCUSSION
We are revising our view of the function of RNA ligases with the discovery that they have unexpected diversity in their choice of phosphate donor substrates. We found that traditional RNA ligases that use a 5ЈpRNA as a substrate for ligation can also modify either RNA or ssDNA that has a 3Ј-phosphate. All studied ATP-dependent polynucleotide ligases activate a 5Ј-phosphorylated nucleic acid substrate by transferring an adenosine monophosphate from an active site lysine to the substrate 5Ј-phosphate, producing a 5Ј-adenylated intermediate. We report here that "canonical" ATP-dependent ligases also possess polynucleotide 3Ј-terminal phosphate adenylation activities, in most cases comparable with their 5Ј-adenylation activity.
The archaeal RNA ligase MthRnl converts 3Ј-phosphorylated RNA termini into 2Ј,3Ј-cyclic phosphates using the same tripartite mechanism of the classic ligation reaction. The first step of the reaction is the same as for classic ligation. The enzyme is adenylated with ATP at the ⑀ amino group of a catalytic active site lysine. Second, this AMP group is transferred to 3Ј-terminal phosphate, creating the activated intermediate 3Ј-adenylated RNA (RNA 3Ј pp 5Ј A), where adenosine is joined to the 3Ј end of RNA through a diphosphate linkage. In the third step, activated phosphate is attacked by adjacent 2Ј-hydroxyl, making the intramolecular "ligation" product 2Ј,3Ј-cyclic phosphate. When 2Ј-terminal hydroxyl is blocked by methylation, the reaction stops at the adenylated intermediate in step two of the reaction. The adenylated intermediate is also generated when the 3Ј-phosphorylated substrate is ssDNA, which does not have a 2Ј-hydroxyl acceptor. This mechanism is almost identical to the 3Ј-terminal phosphate cyclization activity described for RtcA (9), except that the catalytic residue in RtcA is histidine and not lysine (18,32). The identity of the reaction products was determined by mobility on polyacrylamide gel or capillary electrophoresis in comparison with known modifications and also by differential sensitivity to alkaline phosphatase, the phosphatase activity of T4 polynucleotide kinase, and yeast 5ЈDeadenylase.
We have shown previously that substitution of catalytic lysine to alanine (K97A) in the active site of MthRnl abolished the reaction of the enzyme with ATP and, consequently, blocked adenylation of the 5Ј-phosphate of RNA or ssDNA (27). We also found that the K97A mutant of MthRnl was unable to adenylate the DNA 3Ј-phosphorylated substrate (Fig.  2E), indicating that the mechanism of 3Ј-adenylation is likely identical to that of 5Ј-adenylation. This observation has interesting implications for the mechanism used by MthRnl to recognize an RNA or ssDNA phosphate donor substrate in either the 3Ј or 5Ј orientation.
The biological role for RNA 2Ј,3Ј-cyclic phosphate is not well understood. The wide conservation of RtcA in all life forms suggests that the enzyme is essential. However, E. col strains lacking RtcA are not impaired in growth (33). This may be partially explained by the fact that both RtcA and RtcB have cyclase activity. In the case of Methanobacteria, there are three enzymes that have RNA-2Ј,3Јphosphate cyclase activity: MthRnl and homologues of RtcA and RtcB. Although they share this biochemical activity, their biological roles may be diverse. RtcA has cyclase but no ligase activity, and RtcB can ligate RNA3Јp but not 5ЈpRNA, whereas the reverse is true for MthRnl; it is only able to ligate RNA with a 5Ј-phosphate.
Although MthRnl is called an RNA ligase, it has the ability to modify both RNA and ssDNA substrates. Its activities include adenylation of either the 5Ј-phosphate or the 3Ј-phosphate of ssDNA, ligation of either 5ЈpDNA or 5ЈpRNA, and RNA 2Ј,3Јphosphate cyclase activity. An element common to all of these substrates is a terminal phosphate, which must be a major determinate in substrate binding. The ability to ligate either 5ЈpRNA or 5Јp-ssDNA and to adenylate either the 5Ј or 3Ј end of ssDNA suggests a remarkable flexibility in the active site of the enzyme. Both TVa ligase and TS2126 RNA ligase have also been found to adenylated ssDNA3Јp. However, neither T4 RNA ligase 1 nor T4 RNA ligase 2 can modify RNA3Јp (Fig. 1B,  panels 19 -24). These phage ligases function in concert with T4 PNK, which has a 3Ј phosphatase activity that removes the RNA 3Ј-phosphate and a kinase activity to generate a 5ЈpRNA, which creates the correct lysine tRNA substrate for ligation (20). It appears that the archaeal Mth RNA ligase has evolved to have a wider range of functions.
We also tested the ability of ligases to adenylate a 3Ј phosphate nick in dsDNA. Of the three thermophilic enzymes tested only the TVa ligase from T. ammonificans was able to adenylate the nicked substrate. The enzyme was identified by sequence similarity to T4 Rnl1 and has not been characterized. We found that the recombinant enzyme has strong 5Ј and 3Ј adenylation activity. However, actual RNA ligation activity is very weak (data not shown). Of the mesophilic dsDNA ligases tested on this substrate, only T4 DNA ligase was able to adenylate the 3Ј-phosphate nick. As expected, neither T4 DNA ligase nor any of the other dsDNA ligases are able to adenylate ssDNA3Јp. This also served as a control that dsDNA did not melt in reaction conditions. Even though adenylation was observed on DNA substrates with a 3Ј-phosphate, none of the enzymes were able to ligate these substrates. Only RtcB has been reported to ligate ssDNA 3Ј-phosphate in a stem loop structure but not a nicked dsDNA substrate (15).
Our study of thermophilic RNA ligases suggests that they have activities beyond RNA ligation. These enzymes are able to modify either the 5Ј-or 3Ј-terminal phosphates of RNA or DNA. Their ability to adenylate the 3Ј-phosphate on DNA should stimulate studies to determine what role, if any, they may have in the repair of damaged DNA. The RNA cyclase activity may have a role in protecting RNA3Јp from exonucleases and allowing their repair by ligation. Additional research is required to elucidate the biological role of these 3Ј-phosphate modifications.