RNA ligases are widely used as reagents in molecular biology. T4 phage RNA ligases 1 and 2 besides being used for ligation of RNA can be used in rapid amplification of cDNA ends (RLM-RACE) and 3’ RNA labelling [1, 2]. T4 RNA ligase 2 is also capable of sealing nicks in dsRNA or dsRNA/DNA hybrids . Phage TS2126 ligase (CircLigaseTM, ThermoPhageTM) is used for ssDNA circularization . The Mth RNA ligase (MthRnl), from thermophilic archaeal bacteria Methanobacterium thermoautotrophicum, is used for the enzymatic synthesis of 5’ adenylated DNA linkers . Recently RNA ligases have gained popularity in construction of cDNA libraries for next generation sequencing of small RNAs (for overview, workflow and references see [6, 7]).
Capturing small RNAs of unknown sequences is usually achieved by ligation of two linkers with specific sequences on either end of RNA (3’ linker ligation and 5’ linker ligation). This allows specific primers to be used for cDNA synthesis followed by PCR amplification. Modifications of the RNA present challenges for these ligations . For example, some small RNAs are 2’-O-methylated at their 3’ ends . The 5’ end of RNA can contain a triphosphate or cap that blocks 5’ ligation unless it is removed . If the RNA contains a 5’-monophosphate and 3’-OH, it can form unwanted by-products, like minicircles or concatamers, which reduce cloning efficiency and require additional enzymatic treatments and purification steps. Alternatively after 3’ linker ligation RNA can be reverse transcribed into first strand cDNA and ligated to a second 3’ linker (5’ ligation independent cloning) [10, 11]. Current protocols for capturing small RNAs mostly employ phage T4 RNA ligase 1 (T4Rnl1) and a modified, truncated version of T4 RNA ligase 2 (T4Rnl2tr). However, T4 RNA ligases are either unable to (T4Rnl2)  or are slow and inefficient (T4Rnl1) [13, 14] in joining two ssDNA molecules relative to joining RNA counterparts. A 5’-phosphorylated or pre-adenylated ssDNA donor substrate may be used, but the acceptor substrate requirements are specific for RNA . In addition, most single-stranded nucleic acids have an increased tendency to fold at low temperature and maintained secondary structure in the presence of Mg+2. This is thought to contribute to reducing the efficiency of ligation due to sequestration of hybridized 3’-ends, resulting in ligation bias [7, 16]. These structures may lead to a significant reduction in ligation efficiency for certain RNAs. Thermostable phage T2126 ligase might be used in these applications [17, 18], but not in the ATP independent ligation.
Overall RNA ligation comprises three linked enzymatic reactions . In the first step, the ligase reacts with ATP to form the intermediate covalent complex with lysine of conserved motif I (step 1). In the second step, AMP from the ligase-adenylate covalent complex is transferred to the 5’-phosphate of an RNA or ssDNA donor substrate, creating an activated 5’ adenylated intermediate product, designated AppRNA or single-stranded AppDNA (step 2). In the third step of a ligation reaction, the phosphodiester bond is formed between the 3’-OH of the acceptor substrate and the 5’ adenylated phosphate of a donor, releasing AMP (step 3). It is difficult to optimize all three biochemical reactions for efficient ligation, due to the different reaction requirements. For example, ATP is strictly required for RNA ligase adenylation, but often inhibits the step 3 of the ligation.
RNA ligation in vitro, however, can be performed in two stages. The 5’ pre-adenylated activated donor, the product of the second ligation step, can be synthesized by enzymatic or chemical methods. In a second stage, equivalent to ligation step 3 above, this activated donor substrate is ligated to an acceptor in the absence of ATP. This “split” ligation method was originally described for 3’-linker ligation using T4Rnl1 . An improved method used the T4Rnl2 truncated enzyme, and its derivatives, which have reduced activity in the substrate deadenylation reaction catalyzed by T4Rnl1 [6, 21–23]. Ligation of pre-adenylated DNA to RNA in a 3’ linker ligation reaction lacking ATP is also now a preferred approach to ligation of microRNA for next generation sequencing. Since only the 3’ end of the oligonucleotides can be a substrate for this stage, this approach eliminates (i) the need for a dephosphorylation step (removing the 5’-phosphate from small RNAs prior to 3’ linker ligation) to avoid RNA circularization and concatemerization, (ii) thus there is no need for re-phosphorylation for the second ligation (of the 5’ linker) and (iii) eliminates additional purification steps, which cause significant loss of material. Splitting these reactions into two parts increases the efficiency of ligation by enabling optimization of a single step of the ligation reaction and eliminates ATP, which besides producing by-products, if in excess, also can inhibit ligation.
There is a two-fold approach to make the “split” ligation method successful. First, there is a need for a convenient method for synthesis of pre-adenylated linkers. Second, this type of ligation requires non-adenylated ligase (apoenzyme), which can use the pre-adenylated donor as a substrate. The ligase should also be inactive in the reverse step 2 of the ligation reaction, which transfers AMP from activated pre-adenylated substrate back to the enzyme. The pre-adenylated substrates are commercially available or can be efficiently synthesized as described in the Methods . The choice of suitable RNA ligases is limited to the commercially available C-terminal truncated version of T4Rnl2 and its K227Q mutant. This C-terminal deleted version of T4Rnl2 has greatly reduced, but is not completely defective in self-adenylation activity due to pH-optimum shift of the reaction step 1 to alkaline pH . For the same reason truncated T4Rnl2 also retains some deadenylation activity in ligation reaction using a pre-adenylated donor. An additional mutation in motif V was later introduced (K227Q) to increase the fidelity of truncated T4Rnl2 in ligation [6, 21] but resulted in reduced ligation activity.
Taken together, there is a need for single-stranded nucleic acid ligase, which could work efficiently at elevated temperature, reduce structural constraints, and ligate ssDNA and 2’-O-methylated RNA as well as RNA. In order to eliminate side reactions the ligase should accept pre-adenylated substrates in an ATP independent ligation reaction and not have deadenylation activity. To our knowledge no such ligase exists in nature. Thermostable archaeal RNA ligases are attractive choices for enzyme engineering. Their optimum reaction temperature is 60–65°C and they readily accept ssDNA in circularization reactions, indicating acceptance of deoxy-substrates as a donor as well as an acceptor [4, 25–27].
Archaeal MthRnl is a homodimeric, ATP-dependent, thermostable enzyme with optimum activity at 65°C. At this temperature short hybrids should be significantly reduced. MthRnl can carry out intramolecular ligation of either ssDNA or RNA. The circularization of ssDNA is only half as efficient as the circularization of RNA . However, wild type MthRnl has a few shortcomings. While ATP is required for activity, the third step in ligation is inhibited at ATP concentrations as low as 50 μM. It is not suitable for “split” ligation because MthRnl is purified mostly in a stable self-adenylated form, a form unable to carry out the step 3 ligation with pre-adenylated linkers. Recovering deadenylated apoenzyme is inefficient and unpredictable. The apoenzyme also deadenylates pre-adenylated donor, in ligation reactions without ATP.
To overcome these limitations in RNA ligation, we performed a mutagenesis of essential amino acids in the active site of MthRnl to identify and analyze active mutants defective in ATP binding or ATP reactivity, and consequently unable to self-adenylate. These mutants were then characterized for use in the ligation reactions without ATP. Here we present structure-function analysis of these mutants and their activities in single-stranded nucleic acid (RNA and DNA) ligation. Comparison of active MthRnl mutants to T4 RNA ligases revealed significant differences in substrate specificity, which allowed us to define new conditions for single-stranded nucleic acid ligation.