Characterisation and engineering of a thermophilic RNA ligase from Palaeococcus pacificus

Abstract

RNA ligases are important enzymes in molecular biology and are highly useful for the manipulation and analysis of nucleic acids, including adapter ligation in next-generation sequencing of microRNAs. Thermophilic RNA ligases belonging to the RNA ligase 3 family are gaining attention for their use in molecular biology, for example a thermophilic RNA ligase from Methanobacterium thermoautotrophicum is commercially available for the adenylation of nucleic acids. Here we extensively characterise a newly identified RNA ligase from the thermophilic archaeon Palaeococcus pacificus (PpaRnl). PpaRnl exhibited significant substrate adenylation activity but low ligation activity across a range of oligonucleotide substrates. Mutation of Lys92 in motif I to alanine, resulted in an enzyme that lacked adenylation activity, but demonstrated improved ligation activity with pre-adenylated substrates (ATP-independent ligation). Subsequent structural characterisation revealed that in this mutant enzyme Lys238 was found in two alternate positions for coordination of the phosphate tail of ATP. In contrast mutation of Lys238 in motif V to glycine via structure-guided engineering enhanced ATP-dependent ligation activity via an arginine residue compensating for the absence of Lys238. Ligation activity for both mutations was higher than the wild-type, with activity observed across a range of oligonucleotide substrates with varying sequence and secondary structure.

Graphical Abstract

Graphical Abstract.

Introduction

RNA ligase enzymes are useful tools for the analysis of RNA. They are capable of performing inter- and intra-molecular ligation of RNA and the adenylation of both RNA and DNA, and as such they are highly useful for the manipulation and analysis of nucleic acids. ATP-dependent RNA ligases join the 3′-OH and 5′-PO₄ termini of RNA in a three-step nucleotidyl transfer mechanism that is conserved with DNA ligases, and requires a high-energy cofactor and the presence of a divalent cation (1–8). Ligation is facilitated by highly conserved motifs in the nucleotidyl transferase (NTase) domain: I, III, IIIa, IV and V, that are common to all nucleotidyl transferases (5). In step 1, the conserved lysine in motif I of the RNA ligase reacts with ATP to form a covalent ligase-(lysyl-Nζ)–AMP intermediate, releasing pyrophosphate (PP_i). In step 2, the AMP is rotated away from the catalytic motif I lysine and transferred to the 5′-PO₄ group of the RNA, forming an RNA adenylate intermediate (5′-AppRNA). The orientation and transfer of the AMP at this step are mediated by motif I and motif V lysine residues. In the third and final step, the ligase facilitates the nucleophilic attack of an incoming 3′-OH RNA terminus on the 5′-AppRNA, sealing the two RNA strands through phosphodiester bond formation and the release of AMP.

Specificity for the polynucleotide substrate is dictated by an additional domain that is typically linked to the nucleotidyltransferase domain and differences in structure and substrate specificity form the basis of six RNA ligase families (Rnl1–6). Rnl1 enzymes generally recognise tRNA substrates with a break in the anti-codon loop (7,9,10) with bacteriophage T4 RNA ligase 1 the most well characterized of the Rnl1 family. Whereas the Rnl2 family binds to and repairs nicked duplex RNA substrates, with T4 Rnl2 demonstrating that the C-terminal domain is essential for binding and adenylation of RNA substrates (11). Rnl3 ligases are found primarily in archaea and some thermophilic bacteria and differ from other RNA ligases in that they have homodimeric structures (6,12–17). Rnl4 is a new RNA ligase family found in several bacterial species, and is a complex of two proteins (Pnkp and Hen1) that repairs transfer RNAs that have been cleaved by ribotoxins (18,19). Rnl5 ligases lack a C-terminal domain and recognise nicked RNA duplexes via a different amino-terminal domain (20,21). A fungal RNA ligase (Trl1) recently identified is the founding member of the Rnl6 family of enzymes (22).

RNA ligases are crucial enzymes in the sequencing of microRNA molecules. Sequencing adapters, with known nucleotide sequences, are joined to miRNA molecules' 3′-hydroxyl (3′-OH) and 5′-monophosphate (5′-PO₄) ends using RNA ligases. First, a pre-adenylated DNA adapter (AppDNA) is ligated to the 3′ end of each miRNA, typically by using a variant of T4 RNA ligase 2 (T4 Rnl2) (23,24). Next, an RNA adapter is ligated to each 5′ end using T4 RNA ligase 1 (T4 Rnl1) (24,25). The ligated product, comprising the miRNA flanked by adapters of known sequence, can then be reverse transcribed, amplified, and sequenced. A major challenge in small RNA sequencing arises from the sensitivity of RNA ligases to the sequences and/or secondary structures of their nucleic acid substrates. It is well-established that RNA ligases display biases in their ligation efficiencies towards different miRNA molecules, giving rise to sequencing results that do not reflect the true abundances of different miRNAs in any given biological sample (26–30). For example RNA molecules which fold to form structured stems at their 3′ ends are less frequently observed in miRNA sequencing data, while RNAs featuring at least three unstructured nucleotides at the 3′ ends are more commonly represented (28,29). The presence of secondary structures in RNA and adapter molecules likely conceals 5′-PO₄ and 3′-OH termini, which would hinder the attachment of adapters to miRNA, subsequently leading to an alteration in the perceived abundance of distinct miRNAs.

The recognition of these biases has spurred advancements in enzyme technology and library construction methods (24,29,31–33). PEG, an intramolecular crowding agent, can reduce ligation bias, and has been incorporated into commercially available kits (31,34). Randomised regions can also be incorporated into the adapters, which increases the likelihood of having an adapter suitable for ligation with each miRNA in a sample (24,29). Another approach to tackling bias is to carry out adapter ligations at elevated temperatures. In theory, ligation reactions at elevated temperatures may be effective at melting secondary structures in the miRNAs and adapters, revealing the previously sequestered 5′-PO₄ and 3′-OH termini. However, such a strategy is dependent on RNA ligase enzymes capable of proficient ligation at elevated temperatures.

Recently, the thermophilic Rnl3 family of ATP-dependent RNA ligases have been acknowledged for their potential applications in biotechnology, driven by their unique ability to function at elevated temperatures (35,36). Yet the Rnl3 family of ligases remains comparatively understudied compared to other RNA ligase families, particularly Rnl1 and Rnl2 enzymes. The Rnl3 ligase family was unveiled via structural and functional studies of RNA ligases isolated from two thermophilic archaeal species: Pyrococcus abyssi (PabRnl; PDB ID 2VUG) (12,16) and Methanobacterium thermoautotrophicum (MthRnl; PDB ID 5D1P) (13,15,17). As mentioned above they are unique in that they have a homodimeric structure and can catalyse ligation and circularisation of single-stranded RNA. A third homologous structure from the hyperthermophilic bacterium Aquifex aeolicus (PDB ID 3QWU) was also identified, with bioinformatic analyses indicating Rnl3 ligases are prevalent in bacterial and archaeal thermophiles or hyperthermophiles, which typically thrive in temperature ranges between 50°C and 95°C (6). In vitro, the Rnl3 ligases can adenylate both RNA and DNA substrates, with primarily intra-molecular ligation (also described as RNA circularisation) observed with RNA substrates and display varying activity optima, and efficient activity at temperatures ranging from 50°C to 85°C (14,16,17). A point mutation at the motif I catalytic lysine (K97A) in MthRnl produced an enzyme unable to self-adenylate, and therefore defective in the ligation or adenylation of a 5′-phosphorylated substrate in the presence of ATP, requiring pre-adenylated substrates for activity (15). In contrast, wild-type MthRnl is highly proficient in adenylating RNA and DNA adapters at high ATP concentrations, and as such the enzyme is effectively used to adenylate sequencing adapters for small-RNA sequencing protocols (37).

With the potential of utilising thermophilic RNA ligases for high temperature ligation to mitigate RNA sequencing bias and given the current gaps in our knowledge of the structures and mechanisms of Rnl3 ligases, we conducted a comprehensive biochemical and structural analysis of an RNA ligase from the hyperthermophilic archaeon Palaeococcus pacificus (PpaRnl) that is capable of growth at temperatures up to 90°C (38).

Materials and methods

Gene synthesis, protein expression and purification

P. pacificus Rnl3 gene sequences were commercially synthesised with codons optimised for expression in Escherichia coli (GeneArt, Thermofisher). The genes were cloned into the expression vector pET28b between NdeI and XhoI restriction sites for expression with an N-terminal hexahistidine (His₆) tag (molecular weight of PpaRnl + His-tag, 46.9 kDa and PpaRnl (without tag) 44.6 kDa). The resulting plasmids were used to transform E. coli BL21 (DE3) for protein expression. A 1 L culture of E. coli BL21 (DE3)/pET28b was incubated at 37°C in Luria-Bertani (LB) medium supplemented with 50 μg ml⁻¹ kanamycin until the A₆₀₀ reached 0.5–0.7. Recombinant proteins were expressed for ∼18 h after adding 0.75 mM isopropyl β-d-1-thiogalactopyranoside (IPTG) and transferring to 22°C. Cells were harvested by centrifugation and pelleted bacteria were resuspended in 30 ml of lysis buffer (50 mM Tris pH 7.4, 200 mM NaCl, 20 mM imidazole, 10% (v/v) glycerol). Benzonase nuclease (50 units) was added to the lysate, followed by incubation at room temperature for 30 min. The lysate was sonicated on ice using a QSONICA Q700 sonicator. Insoluble material was removed by centrifugation. The lysate was then incubated at 70°C for 30 min and centrifuged to remove contaminating E. coli proteins. The soluble extract was filtered through a 0.2 μm filter and applied to a 5 ml column of Ni-NTA agarose (GE Life Sciences) pre-equilibrated with lysis buffer. The column was washed with lysis buffer, followed by a high salt buffer (50 mM Tris pH 7.4, 4 M NaCl, 20 mM imidazole, 10% (v/v) glycerol) to remove enzyme bound nucleic acid, then washed with lysis buffer to remove traces of the high salt buffer. Protein was eluted with elution buffer (50 mM Tris pH 7.4, 200 mM NaCl, 1 M imidazole, 10% glycerol) over a gradient of 50 ml.

Fractions containing recombinant PpaRnl were combined and loaded onto a HiLoad 16/60 Superdex Gel Filtration Column (GE Life sciences) pre-equilibrated with gel filtration buffer (50 mM Tris pH 7.4, 200 mM NaCl, 10% (v/v) glycerol). Fractions containing recombinant RNA ligase protein were then collected, combined and stored at 4°C. The components of the column fractions during purification were monitored by SDS-PAGE. Protein concentration was determined using a Nanodrop^TM 2000 (Thermofisher) to measure absorbance at 280 nm, with molar extinction coefficients calculated using protparam. For molecular weight determination, a Superdex 75 10/300 GL Gel Filtration Column (GE Life sciences) column was calibrated with protein molecular-weight standards (Biorad) dissolved in deionised H₂O as per the manufacturer's instructions. Blue dextran 2000 was used to detect the void volume. PpaRnl was loaded separately on the column, and UV peaks of each protein was used to calculate generate a standard curve for elucidation of protein molecular weights.

Thermal shift assays

Thermal shift assays were carried out in a Rotogene Q thermocycler with the fluorescent dye SYPRO Orange (1:1000) and buffers within a pH range of 5.0 to 8.5 (40 mM boric acid, 40 mM phosphoric acid, 40 mM acetic acid). Reactions were prepared in triplicate with a final reaction volume of 25 μl and a protein concentration of ∼8 μM. The fluorescence signal was initially measured at 25°C, then increased to 100°C with a heating rate of 0.5°C/min. Data analysis was carried out using GraphPad Prism. The first derivative was calculated, and the resulting sigmoidal curve underwent non-linear fitting to the Boltzmann equation to find the melting temperature (T_m) that occurs at the midpoint of the unfolding transition.

RNA oligonucleotide T_m determination

Oligonucleotides were added to 1× PBS at a final concentration of 10 μM and transferred to a clean RNAse free quartz cuvette. RNA oligonucleotide unfolding was measured by monitoring absorbance at 260 nm over time while heating to 85°C in a Cary60 spectrophotometer (Agilent). RNA oligonucleotide refolding of the same sample was measured by returning the spectrophotometer to room temperature and measuring absorbance at 260 nm, ensuring changes in absorbance were due to RNA secondary structure and not RNA oligo degradation. Temperature was monitored by a temperature probe and correlated to time readings (seconds) recorded by the spectrophotometer. Data for each oligonucleotide was collected in triplicate. Controls included measuring the absorbance of PBS with temperature and measuring refolding of the RNA with decreasing temperature as described.

Ligation activity assays

Standard reaction mixtures (20 μl) containing NEBuffer 1.1 (New England Biolabs) (1 mM Bis–Tris–propane–HCl, 1 mM MgCl₂, 0.1 mM DTT, pH 7.5), 5 μM 5′-phosphate labelled or pre-adenylated RNA or DNA oligonucleotides, with or without nucleotide cofactors (for a final concentration of 70 μM) were incubated for 90 min in a T100 Thermal Cycler (BioRad) at indicated temperatures. Oligonucleotides used in ligation assays were synthesised at Integrated DNA Technologies (IDT). Non-ligatable oligonucleotides blocked at the 5′- and 3′-ends were synthesised without a 5′ phosphate and with a 3′-amino modifier, respectively. Oligonucleotide adenylation was performed using the 5′-DNA adenylation kit (New England Biolabs) according to the manufacturer's protocol. Sequences with different modifications used for enzyme activity analysis are listed in Supplementary Table S1. Standard reactions used RNA oligonucleotide 1 (5′-phos-GAGCUAGCAUUAACUUGG) as a substrate unless otherwise specified. Enzymatic reactions were stopped with 2× formamide loading dye (0.5 M EDTA, 95% formamide, 2% bromophenol blue, 1% xylene cyanol, 1% orange G). The reaction substrates and products were denatured at 95°C and resolved by denaturing 7 M urea 20% urea-PAGE gels in 1× TBE. Gels were stained with SYBR™ Gold Nucleic Acid Gel Stain (ThermoFisher Scientific) (1:1000) for a maximum of 15 min and visualised using an iBRIGHT^TM gel imager (Invitrogen, ThermoFisher Scientific). The extent of ligation and adenylation was quantified using ImageJ and calculated as a percentage as intensity of product band/full lane band intensity × 100. All assays were performed in triplicate.

Crystallisation and structure determination

Initial protein crystallisation trials employed sitting-drop vapour diffusion in 96-well intelliplates at 18°C. Protein was purified with an additional benzonase incubation at room temperature following HisTrap purification to remove protein-bound nucleic acid. Crystals grew within 1 hour by mixing 100 nl of protein solution (8 mg.ml⁻¹, 50 mM Tris pH 7.4, 200 mM NaCl, 10% (v/v) glycerol) with 100 nl of precipitant solution (Hampton screens HR2-114, HR2-086, HR2-130, HR2-134 and HR2-136). Crystal quality was optimised using hanging-drop vapour diffusion screens in 24-well plates at 18°C. Crystals of PpaRnl grew after 1 hour by mixing 2 μl of protein solution (8 mg.ml⁻¹, 50 mM Tris pH 7.4, 200 NaCl, 10% (v/v) glycerol) with 2 μl of precipitant solution (80 mM strontium chloride hexahydrate, 20 mM magnesium chloride hexahydrate, 40 mM sodium cacodylate trihydrate pH 7.0, 20% v/v (+/–)-2-methyl-2,4-pentanediol, 12 mM spermine tetrahydrochloride). PpaRnl K92A crystals were grown at pH 6.5 with the same precipitant solution, while PpaRnl K238G apo and ATP-soaked crystals were obtained using precipitant containing 80 mM potassium chloride, 20 mM magnesium chloride hexahydrate, 40 mM sodium cacodylate trihydrate pH 6.5, 40% v/v (+/–)-2-methyl-2,4-pentanediol. For co-crystallisation conditions, 12 mM spermine tetrahydrochloride Hampton crystallisation screens with hanging-drop vapour diffusion in 24-well plates, as detailed above, were prepared following incubation with ATP and MgCl₂ at final concentrations of 1 and 2 mM, respectively. PpaRnl K238G ATP and MgCl₂ soaked crystals were grown in a precipitant solution consisting of 200 mM potassium chloride, 50 mM magnesium chloride hexahydrate, 50 mM Tris hydrochloride pH 7.5, 20% w/v polyethylene glycol 4000.

PpaRnl protein crystals were briefly soaked in cryoprotectant (precipitant plus 15% glycerol) and flash-frozen in liquid nitrogen before data collection. X-ray diffraction data were collected at the Australian Synchrotron (Melbourne, Victoria) on the MX2 beamline equipped with an EIGER × 16M detector (dectris). Diffraction data were indexed, scaled and integrated using the CCP4 programme suite crystallographic software (39). The structure of the P. abyssi RNA ligase (PDB ID: 2VUG) was used for molecular replacement using phenix.phaser from the PHENIX suite (40). The wild-type PpaRnl model was used for molecular replacement of mutant datasets. The model was built and refined using phenix.autobuild from the PHENIX suite (40), followed by iterative cycles of manual building and refinement using the COOT program (39), phenix.refine and phenix.autobuild. Figures were generated using PyMOL (The PyMOL Molecular Graphics System, Version 2.3.2). Coordinate files and structure factors structures have been deposited into the Protein Data Bank.

Results and discussion

Biochemical characterisation reveals significant substrate adenylation activity of PpaRnl

We began by producing recombinant wild-type PpaRnl in E. coli. The enzyme purified as a dimer as evidenced by its elution from a gel filtration column (Supplementary Figure S1) and additional SDS-PAGE analysis demonstrated a pure polypeptide (Supplementary Figure S1). To establish the thermostability of PpaRnl we conducted fluorescence-based thermal shift assays at a range of pH values (Supplementary Figure S2). The enzyme was highly thermostable, with a melting temperature (T_m) of 94°C (± 1°C) at pH 7.0, with minimal variation observed across pH 5.5–9. The enzyme appears less stable at pH 5 as evidenced by a slight decrease in the T_m.

Next, we determined the temperature range for PpaRnl activity, using a 5′ phosphorylated 18 nucleotide RNA substrate (oligonucleotide 1; Supplementary Table S1) in the presence of 15 μM ATP (Figure 1A). Minimal ligation and/or adenylation of the oligonucleotide substrate was observed from 30°C to 45°C. As the reaction temperature increased from 50°C to 80°C, the ligase displayed an increase in activity, however, predominantly accumulated adenylated RNA intermediates, with minimal ligated product. Ligation activity reached a maximum between 60°C and 70°C, with adenylation activity reaching a peak at similar temperatures. Reactions above 80°C showed a substantial reduction in both ligation and adenylation activity (Figure 1A). The prevalent substrate adenylation activity of PpaRnl is likely due to abortive ligation. Abortive ligation occurs prior to phosphodiester bond formation during step 3 of the ligation reaction where the adenylated intermediate is released and cannot be rebound by the enzyme when ATP is present as rapid adenylation of the enzyme occurs (11,15,37,41–50). However, despite variations in ATP concentration, PpaRnl continued to present dominant substrate adenylation activity, with an optimal ATP concentration between 15 μM and 100 μM (Figure 1B). The abortive ligation activity at low concentrations of ATP (15 μM) differs to that observed for other Rnl3 ligases, for example abortive ligation activity is only observed at ATP concentrations greater than 500 μM for MthRnl (37). We additionally observed that PpaRnl was able to perform some adenylation and/or ligation in the absence of ATP, indicating that some or all the enzyme was purified in an adenylated form (Figure B and D). It is worth noting that this is not consistently observed.

Figure 1.

Characterisation of PpaRnl ligation and adenylation activity. (A) PpaRnl activity with RNA Oligonucleotide 1 over a temperature gradient (30 - 85°C) in the presence of 15 μM ATP and 1 mM Mg²⁺. The graph represents ligation and adenylation activity. (B) RNA Oligonucleotide 1 substrate adenylation and ligation activity with varying ATP concentrations (0 to 1 mM) in the presence of 1 mM Mg²⁺; C, no ligase control. Adenylation and end-joining activities depicted in the line graph. (C) Substrate ligation and adenylation activity across a test panel of nine-oligonucleotides varying in sequence, secondary structure and T_m in the presence of 70 μM ATP. Adenylation, circularisation and end-joining ligation activity is quantified in the histogram. (D) PpaRnl activity with a single-stranded DNA oligonucleotide substrate (DNA Oligonucleotide 1) in the presence of 70 μM ATP. Reaction products and the input substrate are indicated accordingly; pRNA (input RNA), AppRNA (5′-PO₄ adenylated RNA), RNA circle (circularised RNA) and end to end ligation (end-joining). Control reactions included no PpaRnl enzyme in the reaction mix as indicated. Assays were performed in triplicate, and a representative of the gel image is shown. Graphs represent averages of the triplicate with error bars, representing standard error.

The adenylation activity of PpaRnl, especially at low concentrations of ATP, was higher than that observed for other RNA ligases (13–15,17,37). To test whether the lack of ligation activity at low ATP concentrations was due to the sequence or secondary structure of the oligonucleotide substrate, we designed nine synthetic RNA oligonucleotides that varied in sequence and secondary structure. The RNA oligonucleotides were also designed to have T_m values in the range of 40°C to 85°C, with theoretical T_m values confirmed experimentally to ensure they covered the temperature range (Supplementary Table S1; oligonucleotide name corresponds to the approximate T_m). A small amount of end to end joining ligation activity (intermolecular ligation) of two individual RNA oligonucleotides was observed for oligonucleotides 50 and 80, with more intermolecular activity for oligonucleotide 65 (Figure 1C). As for other Rnl3 family members (12–17,37) more circularisation (intramolecular ligation) was observed compared to end to end joining but across different oligonucleotide substrates (oligonucleotides 75 and 80) with the exception of oligonucleotide 65 that displayed both more intra- and intermolecular ligation activity compared to the other oligonucleotides tested (Figure 1C). However, the enzyme still portrayed abortive ligation activity with the dominant reaction product across all oligonucleotides adenylated oligonucleotide. We also assessed the enzyme's activity using a 5′-phosphorylated, single stranded DNA version of RNA oligonucleotide 1 (Supplementary Table S1) (Figure 1D). Similar to other RNA ligases (14,37), we observed the accumulation of adenylated DNA intermediates, showing that PpaRnl is capable of binding and adenylating DNA substrates. To gain insight into enzyme mechanism, and subsequently facilitate engineering of PpaRnl for improved ligation activity we solved the enzyme structure by X-ray crystallography.

Probing the wild-type structure for insights into substrate adenylation mechanisms

We investigated the preference for substrate adenylation by PpaRnl by crystallising the wild-type RNA ligase through vapor diffusion. The refined structure, determined at 2.3 Å resolution (Table 1), showed an asymmetric homodimer, highly similar to MthRnl and PabRnl (32,34), with an N-terminal domain (residues 1–63), NTase domain (residues 64–241), dimerization domain (residues 242–312) and C-terminal domain (residues 313–380) in each protomer (Figure 2A and B).

Table 1.

Data collection statistics and final model quality statistics for PpaRnl wild-type and mutant structures. Statistics for the highest resolution shell are shown in brackets. Data statistics were generated by AIMLESS and phenix.tableone

	Wild-type	K92A Apo	K238G Apo	K92A ATP/Mg2+	K238G ATP/Mg2+
PDB ID	8UCE	8UCG	8UCF	8UCH	8UCI
Data collection
Wavelength (Å)	0.953725	0.953725	0.953732	0.953659	0.953732
Space group	P 2 21 21	P 2 21 21	C 2 2 21	C 2 2 21	P 2 21 21
Cell dimensions
a/b/c/ (Å)	81.216/98.799/ 142.977	80.793/99.699/ 142.477	81.95/96.859/ 139.454	82.02/97.251/ 140.101	80.849/99.722/ 143.073
α/β/γ (º)	90/90/90	90/90/90	90/90/90	90/90/90	90/90/90
Resolution (Å)	2.12 (2.196–2.12)	1.86 (1.926–1.86)	1.9 (1.968–1.9)	2.14 (2.217–2.14)	2.14 (2.216–2.14)
R-merge	0.1462 (1.275)	0.07396 (0.8492)	0.06594 (0.5275)	0.1088 (0.9675)	0.2477 (1.325)
Completeness (%)	93.20 (84.44)	99.88 (99.89)	99.96 (99.77)	99.97 (99.97)	99.88 (99.95)
Redundancy	13.6 (13.0)	13.8 (14.2)	13.7 (13.7)	13.3 (12.8)	13.7 (14.0)
No. of observations	894 046 (85022)	1 336 529 (135700)	602 974 (59475)	416 722 (39623)	884 932 (89121)
No. of unique reflections	65 953 (6536)	97 092 (9589)	44 026 (4345)	31 266 (3098)	64 482 (6353)
Mean I/σI	14.53 (3.36)	17.65 (2.98)	25.80 (4.73)	16.75 (2.55)	11.12 (2.64)
Refinement
R-work	0.1891 (0.2738)	0.2072 (0.3795)	0.1657 (0.2018)	0.1736 (0.2132)	0.2054 (0.2758)
R-free	0.2363 (0.3294)	0.2352 (0.3956)	0.1931 (0.2427)	0.2066 (0.2536)	2358 (0.3267)
No. of protein residues	761	760	381	381	762
Average B value (Å)	27.64	27.57	25.8	27.97	18.79
RMSD
Bonds lengths (Å)	0.008	0.007	0.019	0.003	0.003
Angles (º)	0.87	0.85	1.16	0.54	0.59
Model contents
Protomers/ASU	2	2	1	1	2
Ligands	21	11	24	12	21
Water	478	662	223	211	477

Figure 2.

Crystal structure of the PpaRnl wild-type enzyme with AMP covalently bound and complexed with Mg²⁺. Two views of the dimer rendered as (A) surface and (B) cartoon models, coloured as shown in panel A. (C) Stereoview of the PpaRnl wild-type AMP-Mg²⁺(H₂O)₅ complex and the second Mg²⁺ binding site with 2Fo – Fc electron density map for AMP (grey). Amino acids and AMP are shown as stick models; the AMP is coloured light pink with phosphorus atoms coloured orange; amino acids in the active site are coloured yellow. The Mg²⁺ ions and interacting waters are depicted as green and red spheres, respectively. Atomic contacts are indicated by dashed lines.

Similar to other nucleotidyltransferase superfamily members (5), PpaRnl has conserved residues in motifs I, III, IIIa, IV, V and VI (Supplementary Figure S3). Electron density in the active site revealed covalent linkage of AMP to Lys92 in motif I of the NTase domain, positioned in a syn conformation (Figure 2C). No nucleotide cofactor was added during crystallisation, confirming the PpaRnl enzyme is purified in an adenylated form. Conserved residues in motifs interacted with the covalently bound AMP without significant changes compared to RNA ligase homologs (12,13) (Supplementary Figure S3). In both protomers A and B, the AMP molecule was covalently bound within the active site, forming interactions identical to those observed in MthRnl and PabRnl. Specifically, the adenine ring is situated between Phe169 and Val227, forming additional contacts with Glu90, Lys229, Val93 and Glu91. Additionally, the ribose ring established contact with Asn97 via its 2′-OH and 3′-OH groups (Figure 2C). The AMP phosphate is coordinated by Lys238, which forms an electrostatic interaction and is positioned 2.8 Å from the α-phosphate. The α-phosphate of the AMP nucleotide interacts with a Mg²⁺ ion (Mg²⁺-A), coordinated by five water molecules, and a direct contact to one of the non-bridging AMP phosphate oxygens (Figure 2C). The Mg²⁺(H₂O)₅ complex makes additional water mediated contacts with Gly95, Asp94, Val93, Glu144 and Glu224 and the 2′-OH of the ribose ring. We also identified a second putative metal ion (Mg²⁺-B) located proximal to Mg²⁺-A. The amount of electron density in this region was more than expected for water, while the octahedral-like shape of the density suggested the presence of a Mg²⁺ ion. Specifically, Mg²⁺-B is octahedrally coordinated between Asp248, Asp94 and Glu155 and four water molecules (Figure 2C, Mg²⁺-B). Since the precise role of this second metal ion remains uncertain, we opted to investigate this metal binding site in subsequent structural-based experiments.

Lys238 plays a crucial role in coordinating and orienting the AMP phosphate between the enzyme-adenylate and the 5′-PO₄ termini of the substrate (5,13,15). Prior efforts replacing the equivalent residue with alanine (K246A) in MthRnl resulted in an enzyme incapable of step 1 of the ligation reaction and no improvement in ligation activity using pre-adenylated RNA oligonucleotide substrates (13). Nevertheless, we chose to revisit mutation at Lys238 because our structure implied that replacement of this residue with glycine would result in more plasticity in the active site and potentially improved activity.

Lysine mutations affect ligation activity, enzyme stability and result in structural alterations in the active site

To optimise the enzyme's ligation efficiency for potential use in miRNA library preparation, we engineered two versions of PpaRnl with the following attributes: (i) reduced substrate adenylation and improved ligation (reduced abortive ligation activity), and (ii) ATP-independent substrate ligation using pre-adenylated substrates. These enzymes could be used for ligation of adapters to the 5′ and 3′ ends of miRNA molecules respectively. The first ligase variant involved mutating Lys238 to glycine in motif V to decrease substrate adenylation and increase ligation activity. We chose glycine, rather than the alanine substitution made previously in MthRnl (13), to maintain maximum active site plasticity. The second enzyme variant featured an alanine mutation at the catalytic lysine (Lys92) in motif I, to render the enzyme inactive in adenylation (13,15,16). Both mutants retained their thermostability (Supplementary Figure S2B–D). Specifically, at pH 7.0, the T_m of PpaRnl(K92A) was 93°C (±1°C), similar to the wild-type enzyme. In contrast, PpaRnl(K238G) showed a slight increase in T_m to 95°C (±1°C) (Supplementary Figure S2D).

To characterise the effects of the mutations on ligase activity, we first tested the activity of PpaRnl(K238G) and PpaRnl(K92A) enzymes with oligonucleotide 1 in the presence of ATP. As anticipated, the K92A mutation resulted in an enzyme incapable of ATP-dependent ligation, indicating mutation of Lys92 eliminates both step 1 and step 2 of the ligation reaction (Supplementary Figure S2E). In contrast, the PpaRnl(K238G) enzyme was capable of ATP-dependent ligation and both intra- and inter-molecular ligation activity was substantially enhanced compared to the wild-type enzyme. Across the panel of oligonucleotides described earlier (Supplementary Table S1), PpaRnl(K238G) still adenylated oligonucleotide substrates (Figure 3A). However, adenylation was no longer the dominant activity across the panel of oligonucleotides with PpaRnl(K238G) displaying increased intermolecular ligation of oligonucleotide 65, and substantial ligated product accumulation for oligonucleotides 1 and 50. Oligonucleotides 55, 60, 70, 75 and 80 exhibited low to moderate ligation (Figure 3A). Interestingly, the reaction product that accumulated varied depending on the substrate, with the enzyme accumulating either intra- or inter-molecular ligated products (Figure 3A). Notably, in each ligation profile, a preference of PpaRnl(K238G) for specific oligonucleotide sequences was evident. No trend was observed in the ligation efficiency of oligonucleotides as a function of their increasing T_m, nor whether the T_m of the oligonucleotide was above or below the assay temperature of 65°C. The improved ligation activity of PpaRnl(K238G) across the ATP-dependent reactions showed that the enzyme could compensate for the lack of Lys238, motivating us to conduct further structural studies.

Figure 3.

Effects of mutations on catalytic activity and active site structure. (A) Ligation efficiencies of PpaRnl(K238G) across a panel of nine oligonucleotides varying in sequence, secondary structure and T_m (Supplementary Table S1) in the presence of ATP. The experiments were performed in triplicate, and a representative of the gel image is shown. The different reaction products and the input substrate are marked accordingly; pRNA (input RNA), AppRNA (adenylated RNA). Superimposition of the wild-type active site (yellow) residues with the (B) K92A ATP-bound Michaelis complex (light green) and (C) covalent AMP-bound K238G structure (light blue), with Fo-Fc electron density map for AMP and ATP nucleotides (grey). Amino acids, AMP and ATP nucleotides (pink) are rendered as stick models. The Mg²⁺ ions (green) and waters (red) are depicted spheres. Panel B highlights the coordination and binding of ATP in active sites. Panel C highlights the difference and coordination of Arg70, hypothesised to compensate for the mutated Lys238 in step 1 and step 2 adenylation reactions.

We crystallised the two mutant enzymes: PpaRnl(K92A) and PpaRnl(K238G). We grew two sets of crystals: one in their apo form devoid of nucleotides or metal cofactors prior to crystallisation, and the other incubated with ATP and Mg²⁺ prior to crystallisation (nucleotide-bound). This led to structures with resolutions of 1.8–2.1 Å across distinct space groups (see Table 1 for data collection, phasing and refinement statistics). Unlike the wild-type structure, both the PpaRnl(K92A) and PpaRnl(K238G) apo structures crystallised with no nucleotide bound in the active site. This implies the deficiency of step 1 ligase adenylation in PpaRnl(K92A) and possibly a decreased affinity for ATP in PpaRnl(K238G). Comparison of the superimposed tertiary structures of the two apo mutants with the wild-type structure revealed no significant overall structural changes (RMSD (α-carbons) 0.643 Å and 0.283 Å for K92A and K238G, respectively). Notably, both enzymes crystallised with bound metal ions within the active site and close to the active site, likely from the crystallisation solution. The density for PpaRnl(K92A) suggests two locations for the Mg²⁺ ion in the region of Mg²⁺-B, which we have modelled as partial occupancy of the ion at two alternate positions in this region, whereas in the PpaRnl(K238G) structure there is full occupancy of the Mg²⁺ ion in the same position as Mg²⁺-B in the PpaRnl wild-type structure. In the PpaRnl(K238G) structure, a Mg²⁺ ion (Mg²⁺-A) remained bound in a position analogous to the catalytic Mg²⁺ ion in the wild-type structure, suggesting metal ion binding prior to lysine adenylation (Supplementary Figure S4C and D). This ion is positioned 2.6 Å away from Lys92 and is coordinated by water molecules in an octahedral coordination near the nitrogen of the catalytic Lys92. Furthermore, comparing Lys92 from the AMP-bound wild-type PpaRnl structure to the PpaRnl(K238G) structure revealed a 3.9 Å shift, likely due to the absence of a nucleotide in the active site (Supplementary Figure S4C).

The superimposition of the nucleotide-bound NTase domains of catalytic motif I of PpaRnl(K92A) shows a similar active site architecture to the wild-type enzyme. Notably, mutation of Lys92 to alanine removes the pertinent side chain interaction, without disrupting other active site interactions with the substrate and cofactors but impedes nucleophilic attack of Lys92 (Figure 3B). The active site of PpaRnl(K92A) is occupied by ATP in a syn-conformation and a catalytic Mg²⁺(H₂O)₅ complex, providing a mimetic of the ligase adenylation reaction, in which the substrate, catalytic metal cofactor and PP_i leaving group are oriented correctly for subsequent catalysis (Figure 3B). The γ- and β-phosphates are coordinated by a Na²⁺ instead of Mg²⁺, likely due to the presence of sodium chloride in the crystal condition. (Figure 3B). Electron density suggests two alternate conformations for Lys238, whereby the side-chain switches from coordinating the α-phosphate, to the γ- and β-phosphates (Figure 3B). The alternate position overlaps with the position of Lys92 in the wild-type structure. This suggests the residue is situated in this position due to the alanine substitution at Lys92. However, this also highlights the critical role of Lys238 in phosphate tail coordination. Overall, minimal disparities in active site configuration are observed between the PpaRnl(K92A) and PpaRnl wild-type structures.

By contrast, the PpaRnl(K238G) mutant, pre-incubated with ATP and Mg²⁺, yielded a structure with a covalently bound Lys92 AMP-Mg²⁺(H₂O)₅ complex (Figure 3C). The configuration of the active site residues of PpaRnl(K238G) remained largely like that of wild-type PpaRnl, except for Arg70. In the absence of Lys238, the side chain of Arg70 is shifted from its typical position by approximately 2.9 Å to establish a new interaction with the α-phosphate of the AMP nucleotide, occupying the original position of Lys238 in the wild-type enzyme (Figure 3C). This suggests that the K238G mutation prompts an alternative interaction involving Arg70 during step 1 and potentially step 2 of catalysis. This alternate interaction likely contributes to the observed increase in ligation efficiency (Figure 3A). Essentially acting as a ‘replacement’ for the mutated Lys238 residue, Arg70 aids in achieving favourable phosphate orientation for lysine adenylation. We hypothesise this substitution may not be as efficient as Lys238, resulting in slowed nucleotide binding (as evident by the absence of a nucleotide in the apo K238G structure), which affects lysine and substrate adenylation. A lower affinity for the nucleotide by PpaRnl(K238G) is supported by the decreased abortive ligation activity observed for PpaRnl(K238G) compared to the wild-type PpaRnl. If nucleotide binding is slower the aborted ligation intermediate can remain bound or be rebound by the enzyme as rapid enzyme adenylation by a new incoming ATP molecule is not occurring, which enhances phosphodiester bond formation in step 3. To the best of our knowledge, we are the first to elucidate such a compensatory mechanism in the active site of a DNA or RNA ligase.

A putative metal ion binding site

Each structure revealed an additional metal binding site formed by three residues: Asp94 from motif I, together with Glu155 and Asp248 (Figure 4). The proximity of the electron density to the bound nucleotide, combined with the octahedral-like density and the presence of MgCl₂ in all five crystal conditions, led us to model the density in each structure as Mg²⁺ (Mg²⁺-B). The electrostatic potential of this triad of residues reveals a negatively charged binding pocket near the catalytic motif I lysine (Lys92), similarly positioned to the negatively charged binding pocket observed in the Thermus filiformis ligase structure (51) (Figure 4B). In the PpaRnl structures, the second metal ion is positioned approximately 6.3 Å from the catalytic Mg²⁺ ion (Mg²⁺-A), and approximately 7.3 Å from the AMP α-phosphate. Asp94 is a key residue in the second metal binding site and is known to facilitate AMP transfer to the substrate's 5′-PO₄ termini in RNA and DNA ligases (13,52–58). Multiple DNA and RNA ligase structures reveal the involvement of the three residues (that form the additional metal binding site in PpaRnl) in contacts with AMP and/or metal ions, including: (i) a crystal structure of the Chlorella virus DNA ligase mutant; ChVLig(D29A)-AMP, where a lutetium ion coordinates with the glutamate residue and AMP (53), (ii) a structure of the DNA ligase LigD from Mycobacterium tuberculosis shows zinc ion coordination involving the glutamate and aspartate residues, and an AMP nucleotide (59), (iii) the structure of MthRnl-AMP-Mg²⁺ demonstrates catalytic Mg²⁺ engagement through water bridges via motif I asparagine and glutamate residues (13) and (iv) the structure of the Chlorella virus DNA-bound DNA ligase; ChVLig-AMP reveals coordination of water molecules by motif I aspartate and motif IV glutamate at the catalytic Mg²⁺ site (60). Additionally, Asp248 is crucial in step 2 catalysis in MthRnl (13). The second metal ion binding site may coordinate substrate binding during steps 2 and 3 of catalysis, suggesting a two-metal mechanism for substrate adenylation and ligation.

Figure 4.

Surface representation with electrostatic potential of the PpaRnl NTase domain and a portion of the dimerisation domain (residues 64–250) (A) without and (B) with residues involved in coordinating Mg²⁺-B. Mg²⁺ ions at the active site are depicted as green spheres. The nucleotide binding site (indicated with arrows) is marked with positive potential (blue), with the AMP moiety (pink) bound in a highly negatively charged pocket (red). The AMP phosphate tail is coordinated by a catalytic Mg²⁺ ion (Mg²⁺-A) in negative charged region. The putative Mg²⁺-B binding site is positioned in the same negatively charged binding pocket as the catalytic Mg²⁺-A, indicated by arrows.

Active site lysine mutations enable pre-adenylated substrate ligation

In miRNA library preparation, pre-adenylated adapters are ligated to the 3′ termini of miRNA molecules, requiring an enzyme that is active in the third (ATP-independent) step of the ligation reaction. We assessed whether the PpaRnl(K92A) and PpaRnl(K238G) mutants retained this activity. Based on oligonucleotides that evaluated ligation bias in bacteriophage T4 RNA ligases (28) we synthesised and pre-adenylated two RNA adapters with similar sequences: one oligo forms a secondary structure (SR1-S) and the other is unstructured (SR1) (Supplementary Table S1). Both adapters were synthesised with a 3′-amino modifier that blocked the 3′-end of the substrate, preventing intra- or inter-molecular ligation to other SR1 and SR1-S substrates to mimic 3′ adapter ligation in miRNA sequencing. In each reaction, we incubated pre-adenylated SR1 and SR1-S adapters (21 nucleotides) with oligonucleotide 1 (18 nucleotides) with and without a 5′ phosphate, respectively. This allowed us to effectively evaluate the ligation between two molecules with different sequences (SR1/SR1-S to the oligonucleotide 1 substrate with and without a 5′ phosphate) and secondary structures (Figure 5A and B). Incubation with the adenylated SR1 and SR1-S adapters resulted in intermolecular (end-to-end) ligated products for both PpaRnl(K92A) and PpaRnl(K238G) mutants, demonstrating adapter ligation, although ligation efficiency is low. Oligonucleotide 1 without a 5′ phosphate resulted in the slightly reduced formation of ligated products with the adenylated SR1-S adapter but not with the SR1 adapter (Figure 5A and B). Given the theoretical T_m of the SR1 and SR1-S adapter sequences are lower than the temperature of the ligation reaction, the majority of the SR1-S adapter should be unfolded, as demonstrated by intermolecular ligation with both SR1 and SR1-S adapters.

Figure 5.

Ligation efficiency of PpaRnl mutants in ATP-independent reactions. (A & B) 3′-blocked SR1/SR1-S RNA, 3′-blocked SR1/SR1-S DNA and blocked/unblocked oligonucleotide 1 for (A) K92A and (B) K238G mutants. (C & D) 3′-blocked SR1 RNA ligation across an oligonucleotide panel varying in sequence, secondary structures, and T_m (Supplementary Table S1) for (C) K92A and (D) K238G mutants. All assay reactions include a no enzyme control (as indicated). The different reaction products and the input substrate are marked accordingly; pRNA/pDNA (input RNA or DNA), AppSR1-RNA or AppSR1-DNA (adenylated SR1 RNA or DNA). Assays performed in triplicate, and a representative of the gel image is shown. Graphs represent the average of the triplicate with errors representative of standard error.

While the results with SR1-S and SR1 were promising, in miRNA sequencing it is pre-adenylated, single-stranded DNA adapters that are attached to the 3′-end of the miRNA (31). Therefore, we repeated our assays with oligonucleotide 1 (with and without a 5′ phosphate) and DNA versions of SR1 and SR1-S (Figure 5A and B). The PpaRnl(K92A) enzyme displayed minimal ligation activity across all conditions, yielding minimal ligated product with oligonucleotide 1 without a 5′ phosphate and the adenylated SR1-S DNA adapter (Figure 5A). In comparison, the PpaRnl(K238G) enzyme accumulated slightly more ligated product across all conditions.

Lastly, as a preliminary test, we aimed to ascertain if there is obvious bias in miRNA capturing using the oligonucleotide panel and the pre-adenylated SR1 RNA adapter (Figure 5C and D). There is some bias in RNA capture that is evident for both mutant enzymes, with PpaRnl(K92A) presenting with greater ligation efficiency compared to PpaRnl(K238G) with slight variations in the amount of ligated product across the different oligonucleotides. The presence of ligated product for oligonucleotides with a T_m greater than the reaction temperature (65°C) suggests that the miRNA’s secondary structure might have minimal influence on capture efficiency. However, sequence-dependent bias could still play a substantial role and further characterisation through sequencing-based experiments is necessary to confirm these findings.

Conclusion

Here, we set out to characterise and subsequently engineer a thermophilic RNA ligase from P. pacificus (PpaRnl), which belongs to the Rnl3 family of RNA ligases. Specifically, we engineered two ligase variants in order to explore their utility in miRNA library construction, by enabling high temperature ligation of oligonucleotide adapters to the 3′ and 5′ of miRNA molecules, respectively, to potentially remove bias due to secondary structures forming in miRNAs at low temperatures.

Similar to other Rnl3 family members (12–17,37), PpaRnl primarily displayed adenylation of both RNA and DNA substrates, and more circularisation activity compared to end to end joining, with the exception of one oligonucleotide in the oligonucleotide panel for which both increased end to end ligation and circularisation activity was observed. Using the crystal structure of PpaRnl with covalently bound AMP we designed the PpaRnl(K238G) enzyme for use in high temperature adapter ligation to the 5′ end of miRNAs with the rationale that glycine in place of Lys238 would enable more flexibility in the active site and improved ligation activity. PpaRnl(K238G) displayed increased activity in the presence of ATP compared to the wild-type enzyme, with differences in ligation efficiency observed ligating substrates of different sequences for example adapter to oligonucleotide ligation, compared to ligating substrates of the same sequence for example oligonucleotide to oligonucleotide. These differences in efficiencies could arise from the forced orientation of substrates in the active site with the adapter to oligonucleotide reactions. i.e. 5′ adenylated adapter to 3′OH of oligonucleotide substrate, is the only orientation possible for successful ligation. Interestingly mutation of the equivalent lysine residue in MthRnl to alanine (K246A) resulted in an enzyme incapable of performing step 1 of the ligation reaction, but capable of performing step 3 when utilising a pre-adenylated RNA oligonucleotide substrate. The differences in activity between the alanine and glycine mutations of the two enzymes could be due to the smaller glycine residue either not interfering with binding of ATP to the active site lysine or increased flexibility of the active site. Supporting the latter, the PpaRnl(K238G) crystal structure revealed a compensatory mechanism adopted by the K238G mutation with Arg70 forming interactions with the ATP cofactor in place of Lys238. Notably this arginine residue is conserved across the Rnl3 ligases and as such is present in MthRnl. We hypothesise the K238G mutation decreases the affinity of the enzyme for ATP, and as such decreases abortive ligation enhancing ligation activity, highlighting the plasticity of the active site in its interactions with nucleotide cofactors.

Based on a previous study (13), we created another enzyme variant, where the active site lysine was mutated to alanine (PpaRnl(K92A)) for high temperature ligation of pre-adenylated adapters to the 3′ end of miRNAs, as is done in miRNA library preparation. As expected PpaRnl(K92A) was defective in steps 1 and 2 of the ligation mechanism and only displayed activity with pre-adenylated oligonucleotides. However, although ligation activity was higher than the wild-type enzyme, ligation activity in general was low to moderate with some sequence-dependent variation. Additionally, we identify a second metal ion with a possible role in catalysis, the functional basis for which would need to be established, as this may play a pivotal role in the enzyme's catalytic mechanism.

The substantive substrate adenylation observed for the PpaRnl wild-type enzyme could be highly applicable in the modification of RNA or DNA termini, especially due to the low concentration of ATP (15 μM) required for successful RNA substrate adenylation compared to other Rnl3 ligases including current commercial enzymes (500 μM ATP for MthRnl (37)). Although the activity of the PpaRnl variants is mild to moderate and as such may not be suitable to replace the current RNA ligases in miRNA library preparation the PpaRnl ligases could be used in combination with T4 RNA ligase variants to increase miRNA capture.

Notes

Present address: Joanna Hicks, School of Health, University of Waikato, Hamilton, New Zealand.

Data availability

The data underlying this article are available in the protein data bank at https://www.rcsb.org/, and can be accessed with accession numbers 8UCI (PpaRnl K238G + AMP), 8UCF (PpaRnl K238G), 8UCH (PpaRnl K92A + ATP), 8UCG (PpaRnl K92A), 8UCE (PpaRnl + AMP).

Supplementary data

Supplementary Data are available at NAR Online.

Funding

New Zealand Ministry of Business, Innovation and Employment via a Smart Ideas grant from the Endeavour Fund [RTVU1807]; A.W. is supported by a Rutherford Discovery Fellowship [20-UOW-004 to A.W.]. Funding for open access charge: MBIE Endevour Fund.

Conflict of interest statement. Some of the work described here is also contained in a patent application, number PCT/NZ2022/050140.