Formation of ligation side products by T4 RNA ligases
Attachment of adapters to RNA 3'-ends is a common first step for RNA quantification and discovery by RT-PCR and high-throughput sequencing. Ligations are commonly performed using an RNA ligase in the absence of ATP to prevent the formation of RNA circles and concatemers [7
]. In this case, adapters are 3'- blocked and 5'- adenylated so that they are, in essence, ligation donor reaction intermediates. RNA acceptors are commonly either fragmented from larger RNAs or small RNAs that have 3'-OH and may have 5'-PO4
groups. Therefore, promotion of phosphodiester bond formation (step 3) is the only required enzymatic activity for a strand-joining reaction containing an adenylated adapter. We monitored the formation of ligation products using commercially available RNA ligases and defined [5'-PO4
, 3'-OH] RNA acceptor RNAs and [5'-App, 3'-NH2] DNA donors. In addition to products of the expected size, we detected ligation products that migrated with longer apparent length when treated with Rnl1, Rnl2, or Rnl2tr (Figure ). This finding was surprising since our ligation reaction conditions should have prevented concatamerization of RNA substrates. The absence of ATP in the reactions should prevent enzyme adenylation in all cases, and the truncation of amino acids 250-346 of T4 Rnl2tr should additionally prevent the transfer of the adenylyl group to any available oligonucleotide 5'-PO4
Figure 2 Production of ligation side products by T4 RNA ligases. Intermolecular ligation reactions containing 5'-adenylated DNA adapters, 21-mer 5'-PO4 RNA acceptors and ligase (1 pmol) were incubated at 16°C overnight with 12.5% PEG 8000. Products of (more ...)
These observations led us to test the activity of a number of conservative point mutants of T4 Rnl2 when placed in the context of the truncated enzyme. We chose to examine K227Q, K225R, and R55K since previous studies established that these mutations, in the context of the full-length ligase preserved strand-joining activity (phosphodiester bond formation), but were deficient in enzyme adenylation. We reasoned that preventing enzyme adenylation might further reduce the formation of side-ligation products in the context of the truncated ligase. The use of T4 Rnl2tr K227Q for small RNA cloning has been reported [8
Purification and strand-joining activity of T4 Rnl2tr conservative mutants
The indicated mutations were introduced into T4 Rnl2tr singly or in combination. Mutant ligases were produced as N-terminal Maltose Binding Protein (MBP) fusion proteins in E. coli and purified by amylose and Q Sepharose chromatography (Figure ).
Figure 3 Purification and activity of T4 RNA Ligase 2 truncated mutants. (A) Aliquots of T4 RNA ligase 2 truncated and mutants were separated on 10-20% Tris-glycine SDS polyacrylamide gels and stained with Coomassie blue. The size (in kDa) of marker polypeptides (more ...)
Variant truncated ligases were examined for their ability to promote the intermolecular ligation of 5'-adenylated 17-mer DNA adapters to 5'-FAM-labeled 31-mer RNA oligonucleotides. This experiment measures ligation step 3 - phosphodiester bond formation. All mutants had strand-joining activity, except the R55K K225R ligase (Figure ).
T4 Rnl2tr, T4 Rnl2tr fused to the MBP tag (T4 Rnl2tr+MBP), and R55K displayed the same extent of strand-joining after one hour of incubation, at which ~10% of FAM-RNA was ligated. The extent of strand-joining (~5%) promoted by the K227Q and K225R mutants of the fusion construct was significantly less that of the wild-type T4 Rnl2tr.
The double mutants R55K K227Q and K225R K227Q gave the same or a greater yield of the strand joining reaction product (11% and 5%, respectively) than did their respective single mutants, and the activity of the R55K K227Q mutant was similar to T4 Rnl2tr (Figure ). In contrast, the R55K K225R mutant failed to perform the reaction (< 0.1% of substrate ligated). Surprisingly, the triple mutant retained the ability to ligate its substrates and had a similar efficiency to the K225R K227Q ligase (5% of substrate ligated).
Effect of pH on strand-joining activity
We next examined whether the conservative active site mutations that we had introduced changed the pH optimum for intermolecular strand joining activity (ligation step 3). We performed ligation reactions in buffers from pH 5.0 to pH 9.5 containing 10 mM Mg2+, 1 mM DTT and either 10 mM Tris-HCl or 10 mM Tris-acetate. The substrates for the assays were 5'- FAM-labeled 31-mer RNA and 5'-adenylated 17-mer DNA.
Under multiple turnover conditions, all proteins were active (except the R55K K225R mutant), but their pH optima differed (Figure ). We observed that the optimal pH for T4 Rnl2tr is 7.0, where 1 pmol of enzyme ligated 17% of the FAM-labeled RNA, (equivalent to 0.8 pmol of the final product in one hour). All mutations, with the exception of R55K exhibited reduced yield of ligation products under all pH conditions tested as compared to T4 Rnl2tr at its optimum. Variants containing K227Q, or K225R either singly (Figure ) or in combination with other substitutions (Figure ) had optima shifted toward higher pH. As shown in Figure , the triple mutant had an optimum efficiency between pH 7.0 and 7.5 and the lowest strand-joining activity (5% of input ligated after 1 hour).
Figure 4 Effect of pH on ligase intermolecular strand-joining activity. (A-D) Intermolecular strand-joining reactions were carried out with 10 pmol 5'-adenylated 17mer DNA, 5 pmol 31-mer 5'-FAM-labeled RNA acceptor, and ligase (1 pmol) for 1 hour at 25°C (more ...)
Single turnover assays (in enzyme excess) were performed to mimic common usage conditions for T4 Rnl2tr (Figure ). The assays were identical to the multiple turnover reactions, except for the amount of ligase was increased to 13.8 pmol. T4 Rnl2tr was active over a large pH range from 6.0 to 8.0. This was in contrast to the narrower range observed under multiple turnover conditions (compare Figure to ). T4 Rnl2tr+MBP had the same profile (Figure ). As shown in Figure , the single mutants had differing pH ranges in which they were maximally active: R55K was most efficient between pH 6.0 to 8.0, while K225R was most efficient between pH 7.0-8.0. The K227Q mutant was most active between pH 8.0 and 9.0. As we observed under single turnover conditions, the double and triple mutants had higher strand-joining activity in higher pH conditions as compared to T4 Rnl2tr (Figure ) with the least accumulation of ligation product observed with the triple mutant (Figure ). From a practical standpoint, these data are instructive in that all of the ligases perform the strand-joining reaction efficiently when in high concentration in buffer with pH 7.5-8.0.
Effect of PEG 8000 on single turnover strand-joining reactions
Polyethylene glycol (PEG) is known to stimulate ligation reactions for the T4 RNA ligases [10
]. We examined the activity of the mutants at different PEG 8000 concentrations for intermolecular ligation of 5'-adenylated DNA to 3'-OH RNA (Figure ).
Figure 5 Effect of PEG 8000 on ligase intermolecular strand-joining activity. Strand-joining reactions were carried out with 10 pmol 5'-adenylated 17-mer DNA, 5 pmol 31-mer 5'-FAM-labeled RNA acceptor, ligase (13.8 pmol), and varying amounts of PEG 8000 for 1 (more ...)
In agreement with our previous observations, strand joining of 5'-FAM-labeled 31-mer RNA to 5'-adenylated 17-mer DNA adapter was stimulated as PEG 8000 concentration increased up to 25% for T4 Rnl2tr [11
]. T4 Rnl2tr+MBP showed an identical response up to 25% PEG (Figure ). We did not observe further stimulation when PEG concentration was increased beyond 25%, and handling concentrations of PEG greater than 25% was difficult because of the high viscosity of the reactions.
Overall, the single and multiple mutants displayed stimulated strand joining activity with increased concentration of PEG 8000 (Figure ). Under maximal PEG stimulation, all variants tested were able to convert nearly 100% of the substrate to ligated form.
Strand-joining activity over time
We performed time course experiments to monitor the progression of intermolecular strand-joining reactions under multiple turnover conditions using a 5'-FAM-labeled 31-mer RNA and a 17-mer 5'-adenylated DNA adapter (Figure ). The amount of ligated product was calculated (in % of input) at different time points between 0 and 24 hours. All ligase variants tested, except K225R, ligated ~75% of the input RNA after 24 h. The K225R variant accumulated ~36% of ligated product. Given the input concentration of ligase, these experiments indicate an average of 4 ligation events per enzyme molecule over the course of the reaction for all variants, except K225R which catalyzed 2 events.
Figure 6 Analysis of intermolecular strand-joining over time. Strand-joining reactions were carried out with 10 pmol 5'-adenylated adapter, 5 pmol 31-mer 5'-FAM-labeled RNA acceptor, and ligase (1 pmol) over a span of 24 hours at 25°C to assess the progress (more ...)
Ligation reactions containing K225R accumulated significantly less ligated products than T4 Rnl2tr (Figure and , and Table ). After one hour of ligation, ligated products accumulated to 6.1 +/-1% of maximum when incubated with K225R as compared to 24 +/- 2.3% of maximum when incubated with T4 Rnl2tr (mean +/- SEM, p < 0.01). After 2, 3, 6, 9, 12 and 24 hours, ligated products accumulated to 11 +/- 1%, 14 +/- 2%, 21 +/- 4%, 26 +/- 5%, 34 +/- 3%, and 36 +/- 8% when incubated with K225R as compared to. 39 +/-3%, 48 +/- 3%, 64 +/- 3%, 70 +/- 3%, 76 +/- 3%, 52 +/- 2%, and 76 +/- 3% when incubated with T4 Rnl2tr (mean +/- SEM, p < 0.001).
Differences in mean RNA substrate ligated by T4 Rnl2tr mutants over time.
We noted that K227Q accumulated ligation products significantly more slowly than Rnl2tr, or Rnl2tr+MBP, but accumulated to the same degree after 24 hours of incubation. After one hour, ligated products accumulated to 7.5 +/-0.6% of maximum when incubated with K227Q vs. 24 +/- 2.3% of maximum when incubated with T4 Rnl2tr (mean +/- SEM, p < 0.05). After 2, 3, 6, and 9 hours, ligated products accumulated to 14 +/- 1%, 19 +/- 2%, 32 +/- 4%, and 40 +/- 5% when incubated with K227Q vs. 39 +/-3%, 48 +/- 3%, 64 +/- 3%, 70 +/- 3%, and 76 +/- 3% when incubated with T4 Rnl2tr (mean +/- SEM, p < 0.001). After 12 hours, ligated products accumulated to 52 +/- 2% of maximum when incubated with K227Q, while products accumulated to 76 +/- 3% of maximum for T4 Rnl2tr (mean +/- SEM, p < 0.01). Accumulated ligation products were not significantly different after 24 of incubation 72 +/- 2% vs. 81 +/- 3%, for K227Q vs. T4 Rnl2tr (mean +/- SEM, p > 0.05) (Figure and , and Table ).
Interestingly, combining R55K and K227Q increased the accumulation of ligated RNA at earlier time points, and we could detect no difference at any time point comparing R55K K227Q to Rnl2tr.
Combining K225R with K227Q increased ligation product accumulation at earlier time points and the total product accumulated after 24 h as compared to K225R alone (Figure , Table ). The accumulated product for this mutant was significantly lower after 3 and 6 hours of ligation, but not before of after these time points. After 3, and 6 hours of ligation with the K225R K227Q mutant, ligation products had accumulated to 25 +/- 2%, and 40 +/- 3% if maximum, compared to 48 +/- 3%, and 64 +/- 3% of maximum with T4 Rnl2tr (mean +/- SEM, p < 0.01).
The triple mutant displayed a similar profile to the K227Q mutant for the accumulation of ligation products (Figure andTable ).
Interestingly, T4 Rnl2tr, T4 Rnl2tr+MBP, the R55K and the K225R mutants did not accumulate additional ligation products after twelve hours of reaction. In contrast, the K227Q, the R55K K227Q and the K225R K227Q, and triple mutants continue to accumulate ligated products over the entire course of the experiment. By the end of the experiment, only the K225R mutant had accumulated significantly less ligated substrate than wild-type T4 Rnl2tr (Figure and Table )
Considered together, the results so far established that the introduction of conservative mutations, singly and in combination at positions 55, 225, and 227, in the context of the truncated T4 RNA ligase 2 yielded ligases that could reasonably be used as tools for molecular biology. We next sought to determine whether these ligases had increased performance with respect to the formation of unwanted ligation products.
Concatemer formation by T4 Rnl2tr variants
Ligation side products such as concatemers and circles are problematic for ligase applications such as high-throughput sequencing library construction. We tested the ability of each ligase to produce desired and undesired products in intermolecular ligation reactions.
Ligation reactions were performed using 5'-PO4
22-mer RNA acceptor and 5'-adenylated 17-mer DNA donor. The DNA donor was blocked at the 3'-end by the addition of an -NH2
group. Thus the 5'-PO4
end of RNA could serve as a ligation donor substrate for joining to the 3'-OH of another RNA 22-mer. Reactions were performed in buffer containing 10 mM Tris HCl pH 7.5, 10 mM Mg2+
, 1 mM DTT and 12.5% PEG 8000, to maximize ligation efficiency. Expected products were 39 nt in length, and ligation side products were predicted to be 39+22n or 22+22n (where n is a natural number
[1: ∞]) nt in length.
The two WT ligases (Rnl1 and Rnl2 in Figure ), and the wild-type truncated Rnl2 (Rnl2tr and Rnl2tr+MBP) formed the final ligation product (39-mer band), but also formed higher molecular weight species of ~ 60 and 80 nt. For these ligases, the 5'-PO4 RNA band was completely absent at the end of the reaction indicating that all the substrate was consumed.
Figure 7 Assaying the formation of side products by T4 RNA ligases. Intermolecular strand-joining reactions containing 5'-adenylated adapters, 21-mer 5'-PO4 RNA acceptors, and ligase (1 pmol) were incubated at 16°C overnight in the presence of 12.5% PEG (more ...)
The single K227Q mutation produced only the desired ligation product, whereas ligation side products were observed for R55K and K225R. The R55K+K225R mutant showed low levels of accumulated ligation product of the correct size, consistent with the low activity we observed in earlier experiments. Other multiple mutants, all of which contained the K227Q mutation, formed dramatically reduced levels of undesired ligation products. Together, these observations correlate K227 with the formation of undesired ligation products.
Our ligation conditions did not contain ATP, yet we continually observed the accumulation of ligation side products that could only be explained by the concatemerization of RNA inputs that would require the adenylation of 5'-PO4 ends. In the absence of exogenous ATP, one possible source of adenylyl groups in our experimental system is ligation reaction 2 - transfer of AMP from the ligase active site to the 5'-PO4 of the donor oligonucleotide - running in reverse. That is, the transfer of adenylyl groups from the adenylated donor substrate to the catalytic lysine in the active site of the ligase.
We tested the ability of T4 Rnl2tr and variants to remove the AMP from 5'-adenylated DNA oligonucleotides by incubating these substrates overnight in the absence of acceptor under single turnover ligation reaction conditions containing 12.5% PEG 8000 (Figure ). After incubation, reactions lacking enzyme had only one band corresponding to the 5'-adenylated DNA oligonucleotide indicating that it was stable during the assay. In contrast, when incubated with T4 Rnl1, we observed a single band with lower molecular weight that co-migrated with 5'-PO4 DNA adapter. We interpreted this result to indicate that the AMP was completely removed from the 5'-adenylated DNA adapter.
Figure 8 Deadenylation activity of T4 RNA ligase 2 truncated mutants. 5'-adenylated DNA adapters were incubated with an excess of ligase (13.8 pmol), and 12.5% PEG 8000 at 16°C overnight. Oligonucleotide substrates are depicted schematically above the (more ...)
Adenylated substrates incubated with T4 Rnl2 WT, Rnl2tr and T4 Rnl2tr+MBP migrated as two bands; one corresponding to the input, and the other that migrated at the same position as the 5'-PO4 adapter. We interpret this to indicate that these enzymes were able to remove the adenylyl groups from some of the substrate. R55K and K225R mutants similarly converted the adenylated substrates into 2 species. On the other hand, the adenylated DNA adapters incubated with K227Q migrated as the higher molecular weight intact species. Adenylated DNA adapters incubated overnight with the multiply mutated T4 Rnl2tr variants were largely unchanged. These results correlate the deadenylation activity of T4 Rnl2 with K227.
AMP transfer by variant ligases
The observation that incubation of T4 RNA ligases with adenylated oligonucleotides could result in changes in their migration consistent with deadenylation led us to directly monitor the fate of the AMP group in question. To do so we followed AMP transfer during the ligation reaction using 32P-α-AMP-labeled adenylated DNA substrates and 5'-PO4 RNA.
Intermolecular ligation reactions were carried out using 10 pmol of 5'-α-32P-adenylated DNA adapter and 5 pmol of 5'-PO4 RNA overnight with one pmole of ligase (Figure ). When resolved by urea PAGE, the negative control reaction containing no enzyme displayed a single band that migrated at 22 nucleotides indicating the stability of the α- 32P-AMP attached to the DNA adapter (Ap*p-DNA) over the course of the experiment. The majority of the radioactivity detected was concentrated at the bottom of the gel for all of the reactions containing ligase except for the R55K K225R reaction where the majority of signal was observed to co-migrate with a 22-mer length. This was consistent with the inactivity of this variant observed in other experiments. We interpret the results from the other ligase variants to indicate that, by the end of the reaction, the radioactive AMP had been released as a nucleotide into the reaction mixture.
Figure 9 Following AMP during ligation reactions with T4 RNA ligases. (A) 22-mer DNA adapters were 5'-adenylated with α-32P-labeled ATP (see materials and methods). Intermolecular strand-joining reactions containing 10 pmol radiolabeled DNA adapter, 5 (more ...)
These results are consistent either with ligation of the adenylated DNA to the intended 3'-OH end of the RNA, or with AMP addition to the unintended 5'-PO4 RNA substrate that results in concatemers, followed by phosphodiester bond formation. In both scenarios, α-32P-AMP would be released into the reaction mixture.
We performed deadenylation activity assays of the mutants by performing ligation reactions in the absence of the RNA acceptor (Figure ). In control reactions containing no-enzyme, only the input 22-mer oligonucleotide (Ap*p-DNA) was evident, indicative of the stability of the 5'-adenylated DNA adapter.
Radioactive signal that migrated with the adenylated adapter was absent from the reaction incubated with T4 Rnl1. Instead, the entirety of the signal co-migrated with free AMP. We interpret this to mean that T4 Rnl1 has the ability to both remove the α-32P AMP from the adenylated adapter and to also release it into the reaction mixture. When reactions containing T4 Rnl2, Rnl2tr and Rnl2tr+MBP were resolved, they had reduced signal that co-migrated with the intact adenylated adapter. Full-length T4 Rnl2 had an increased radioactive signal that co-migrated with AMP. We interpret this result to indicate that full-length T4 Rnl2, like T4 Rnl1 has the ability to remove the AMP from the adenylated adapter and to release it into solution. In contrast to T4 Rnl1, in reactions that contained T4 Rnl2, Rnl2tr, and Rnl2tr+MBP, we observed radioactive signal that migrated with higher molecular weight than the intact adenylated adapter, as well as a smaller amount of signal that co-migrated with AMP. The high molecular weight signal that we observed was consistent with the covalent attachment of α-32P AMP to the ligase itself (AMP*-ligase).
Adenylated substrate incubated with T4 Rnl2tr K227Q remained largely unchanged, and we did not observe radioactive signal that co-migrated with AMP or with high molecular weight. Reactions containing R55K and K225R yielded signals co-migrating with the adenylated DNA adapter and with higher molecular weight species. As with the unmutated ligases, we interpret this result to indicate that these variants are able to remove the adenylyl group from the adenylated adapters and that it remains attached to the ligase. As found for K227Q, reactions containing the ligases with multiple mutations, showed the majority of 32P signal co-migrating with the adenylated adapter. Treatment of the ligation reactions with Proteinase K shifted the high molecular weight radioactive signal so that it migrated with lower molecular weight, suggesting that the higher MW band is indeed ligase-32P-AMP since ligases are sensitive to Proteinase K digestion. (Figure ).
AMP transfer from ligation donor to RNA 5'-PO4
To establish that AMP could be transferred from an adenylated DNA adapter to an RNA 5'-end, we incubated radioactively adenylated DNA adapters (Ap*p-DNA) with 5'-PO4 RNA that was blocked at its 3'-end (Figure ). Blocking the 3'-end of the RNA prevented strand-joining. In the absence of enzyme, we observed radioactive signal corresponding to the adenylated adapter. In reactions that contained Rnl1, the entirety of the radioactive signal migrated with low molecular weight corresponding with AMP. In reactions incubated with Rnl2, Rnl2tr, Rnl2tr+MBP, and R55K, we observed radioactive signal that co-migrated with 28-mer 3'-blocked RNA. We additionally observed higher molecular weight AMP*-ligase adducts for Rnl2tr, Rnl2tr+MBP, R55K and K225R. We did not observe radioactive signal corresponding to the 28-mer RNA for the inactive ligase R55K+K225R, or for the active ligases containing K227Q.
Taken together, we interpret these results to directly demonstrate the reversal of ligation reaction step 2 - transfer of adenylyl groups from adenylated oligonucleotide donors to the ligase, and from the ligase to RNA 5'-PO4 ends, or to generate free AMP. Furthermore our observations are consistent with the requirement of a lysine residue at position 227 for this reverse reaction to occur.