Application of RtcB ligase to monitor self-cleaving ribozyme activity by RNA-seq

Adapters are efficiently ligated to structured ribozyme fragments by the thermostable RtcB ligase and Archease from P. horikoshii

Previously, EcoRtcB has been used for a high-throughput sequencing approach, where a DNA adapter with a sequence of eight randomized RNA nucleotides at its 3ʹ end is ligated to RNAs with 5ʹ OH termini (Peach et al. 2015). With this method, novel 5ʹ OH fragments derived from mRNAs were identified. However, in contrast to the adapter ligation to mRNAs, ligation to highly structured ribozyme fragments appears more challenging. Therefore, we investigated and optimized the ligation of an RNA adapter to 3ʹ ribozyme cleavage fragments with diverse structures using EcoRtcB and PhoRtcB with Archease (Figure 1). In addition to previously published reaction conditions, we added 15% PEG8000.

Figure 1:

Ligation of an RNA adapter with 3ʹ phosphate (adapter_p) to the 5ʹ OH end of different 3ʹ ribozyme cleavage fragments using PhoRtcB or EcoRtcB.

For both ligases, the same concentration (5 µM) was used. Archease was added to a concentration of 2 µM to reactions with PhoRtcB. Ligation product is abbreviated with LP. Adapter ligation with the thermostable PhoRtcB was successful for all 3ʹ ribozyme fragments with efficiencies of 61% (twister type P5), 59% (twister type P1), 85% (hatchet), 48% (HDV). In contrast, EcoRtcB ligates the adapter to the twister type P5 and the hatchet ribozyme fragment with reduced efficiencies (38% or 11%), but only ligates the adapter to twister type P1 and HDV ribozyme fragments very poorly.

Our results demonstrate that EcoRtcB was able to join the RNA adapter to the 3ʹ twister type P5 or 3ʹ hatchet ribozyme fragment, but nearly no ligation products of the 3ʹ twister type P1 and 3ʹ HDV ribozyme fragments were detectable. One reason for the failed ligation could be the complex structures of these ribozymes even after cleavage. For example, the 3ʹ cleavage product of the twister ribozyme type P1 can still form both pseudoknots (Roth et al. 2014) and the 3ʹ HDV ribozyme fragment that comprises most of the ribozyme can still fold into its complex structure after cleavage.

In contrast to the EcoRtcB, the PhoRtcB ligase was able to ligate the RNA adapter to all of the used 3ʹ ribozyme fragments. Firstly, the more efficient ligation of the PhoRtcB ligase is possibly achieved by its conversion into an enzyme catalyzing multiple turnovers in presence of the cofactor Archease, which increases the ligation activity (Desai et al. 2014). Because EcoRtcB ligation is a single-turnover reaction, a higher concentration of this enzyme would have to be used to get a comparable ligation efficiency. Secondly, PhoRtcB is thermostable and therefore active at temperatures well above 37 °C. Due to the higher incubation temperature (usually 70 °C) possible with PhoRtcB, RNA’s secondary and tertiary structures unfold and thus, the relevant RNA ends are more accessible for ligation. Taken together, using a thermostable RtcB ligase together with Archease enables efficient ligation of adapters to structured RNAs such as ribozyme fragments.

RtcB ligases do not efficiently ligate DNA adapters to single-stranded DNA or RNA molecules

Some self-cleaving ribozymes like HDV and hatchet cleave themselves near the 5ʹ end (Ferré-D’Amaré et al. 1998; Zheng et al. 2019). Because these ribozymes can still fold into their active structure after cleavage, they could be capable of cutting off RNA adapters immediately or in subsequent reactions. To prove this, we PAGE-purified RtcB-mediated adapter-hatchet ligation product and incubated it for 2 h in Mg²⁺-containing buffer used for the second adapter ligation. In this experiment, we observed – apart from the ligation product – the reappearance of hatchet ribozyme fragment that was likely liberated from the RNA adapter by self-cleavage (Supplementary Figure S1). As the removal of an adapter from the ribozyme fragment is problematic for ribozyme analysis by RNA-seq, we subsequently investigated the ligation of DNA and different 2ʹ modified RNA adapters to 3ʹ ribozyme fragments.

When applying a DNA instead of an RNA adapter, the catalysis of self-cleavage is impossible due to the lack of the 2ʹ O nucleophile. Therefore, we investigated the ability of ligating DNA adapters that are 5ʹ-labeled with [γ³²P]-ATP to RNA and DNA substrates by EcoRtcB ligase as well as PhoRtcB (Figure 2). Wherever PhoRtcB is indicated, the reactions included the RtcB enzyme and its Archease. For the initial analysis, we used a simplified model adapter/substrate pair capable of base-pairing similarly to previously used ligation substrates (Tanaka and Shuman 2011). These nucleic acid molecules closely mimic the natural RtcB substrate of a tRNA anticodon stem-loop and should enable optimal ligation. In the past, ligation experiments with EcoRtcB demonstrated that only at least partly complementary DNAs were ligated resulting in a hairpin (Das et al. 2013). Our experiments confirm that EcoRtcB was able to ligate the DNA adapter to DNA or RNA substrate only when both strands are part of a common stem-loop structure. However, this ligation is not as efficient as if joining two RNAs. In contrast to EcoRtcB, the analogous ligation of a DNA adapter to DNA or RNA substrate by PhoRtcB was not successful. Attempts at joining a DNA adapter with a 3ʹ ribozyme fragment using EcoRtcB failed (data not shown).

Figure 2:

Ligation of model DNA adapter and DNA or RNA substrate with EcoRtcB or PhoRtcB/Archease.

(a) Sequence and structure of model DNA adapter (19 nt) hybridized to substrate DNA (20 nt). In the model RNA substrate, all thymidines are exchanged for uridines. Model adapter and substrate are designed similarly to molecules first published in (Tanaka and Shuman 2011). The model adapter carries a 5ʹ and 3ʹ phosphate (P) and model substrate a 5ʹ and 3ʹ OH. (b) Analysis of ligation efficiency of model DNA adapter (D19) to model DNA or RNA substrates (D20 or R20). DNA adapter was 5ʹ labeled with [γ³²P]-ATP. DNA and RNA substrates are non-radioactive. 1 µM DNA adapter and 10 µM (+) or 50 µM (++) DNA substrate (or 50 µM RNA substrate) were initially incubated at 70 °C for 2 min and cooled down for 5 min at room temperature. Then the ligation was performed as described in the methods section. LP marks the ligation product. For EcoRtcB, one LP carries a 5ʹ P and the other a 5ʹ ppG-cap creating the visible double band. (c) Investigation of the ability to guanylate 5ʹ P by EcoRtcB ligase. DNA adapter (D19) was ordered with a 3ʹ amino protection group and 5ʹ labeled with [γ³²P]-ATP. DNA substrates (D20) carry a 5ʹ OH and a 3ʹ OH or 3ʹ amino protection group. For the ligation reaction, 0.1 µM DNA adapter, 10 µM DNA substrate and 7 µM EcoRtcB ligase were used.

Unlike the ligation reaction, the guanylation of the DNA adapter was very efficient for both ligases (Figure 2B and C). This stable DNA guanylation was already described in earlier publications and is due to the 2ʹ H (Das and Shuman 2013). In contrast, for guanylated RNA it is assumed that the released RNAppG would rapidly form 2ʹ,3ʹ cP in solution via a nucleophilic attack by the terminal 2ʹ OH (Desai et al. 2014). Interestingly, it seems that even a second GMP was transferred to the DNA adapter by EcoRtcB because a shift by one additional nucleotide was detectable in the gel (Figure 2B). For PhoRtcB, only a faint band is visible, likely corresponding to GppDNAppG. Therefore, we analyzed if GMP can be transferred to the 5ʹ P of DNAs (Figure 2C). In this experiment, we used a 5ʹ labeled DNA adapter with a 3ʹ amino protection group and observed that GMP was transferred to the 5ʹ P. Nevertheless, this 5ʹ guanylated DNA adapter was not joined to a DNA substrate with 5ʹ or 3ʹ OH end. Hence, EcoRtcB cannot only transfer GMP to the 3ʹ P, but also to the 5ʹ P of DNA adapters. Previously, an ability of RtcB ligase to transfer GMP to the 5ʹ P and 3ʹ P of a DNA strand has been observed only for an RtcB paralog (RtcB3) from Myxococcus xanthus (Maughan and Shuman 2015). For EcoRtcB a 5ʹ P guanylation was not detected in previous experiments, which used a 5ʹ P DNA oligonucleotide that perfectly annealed at its 5ʹ end with the second DNA strand (Das et al. 2013). We assume that under these conditions the 5ʹ P might not be accessible for guanylation by RtcB and thus, that at least one non-complementary base might be necessary for EcoRtcB to enable the GMP-transfer to a 5ʹ P DNA end.

The ability of a ligase to activate an RNA at either end (the 5ʹ P and the 3ʹ P) is not entirely unprecedented. It was already observed for other enzymes like RtcA cyclase and T4 RNL. In E. coli, RtcA is encoded in one operon with RtcB and it usually converts 3ʹ P ends to 2ʹ,3ʹ cP by adenylating the 3ʹ P (Genschik et al. 1998). However, RtcA has also been shown to adenylate 5ʹ P ends of DNA as well as RNA substrates (Chakravarty and Shuman 2011). The T4 RNL normally adenylates 5ʹ P ends to form 3ʹ–5ʹ phosphodiester bonds between a nucleic acid with a 3ʹ OH and another with a 5ʹ P. This enzyme was also found to produce unusual 3ʹ adenylated products (Hinton et al. 1982).

RtcB ligases do not efficiently ligate 2ʹ modified RNA adapters to ribozyme fragments

One explanation for the low ligation efficiency of DNA to RNA by RtcB could lie in an inability of the enzyme to accommodate binding of a single-stranded DNA. Accordingly, we investigated the ability of RtcB enzymes to ligate RNA adapters that merely contained different 2ʹ modifications instead of the 2ʹ OH at their 3ʹ terminal nucleotide (Supplementary Figure S2). Because the 2ʹ end is modified, no nucleophilic attack can occur and thus even ribozymes such as hatchet or HDV would not be able to support the liberation of the ligated adapter by cleavage.

We used an RNA adapter with either (i) a 3ʹ terminal DNA nucleotide (2ʹ H), (ii) a 2ʹ O-methyl group (2ʹ OMe), (iii) a 2ʹ fluoro group or (iv) a 3ʹ terminal locked nucleic acid (LNA) nucleotide (Supplementary Figure S2). We employed the 3ʹ fragment of a hatchet and twister type P5 ribozyme, which were previously proven to be suitable ligation substrates (Figure 1). High amounts of guanylated RNA adapter with 2ʹ H (PhoRtcB) and 2ʹ F (both ligases) were detectable, especially in comparison to the unmodified RNA adapter (Supplementary Figure S2). In contrast to PhoRtcB, EcoRtcB is able to guanylate all 2ʹ modified adapters.

Moreover, we found that none of the 2ʹ modified RNA adapters were efficiently ligated to ribozyme fragments. Therefore, the presence of a 2ʹ OH group appears to be crucial for ligation. However, the specific role the 2ʹ OH could play in the ligation mechanism is currently unclear. The most likely ligation mechanism relies on the opening of the 2ʹ,3ʹ cP followed by the activation of the resulting 3ʹ P by guanylation (Chakravarty et al. 2012; Tanaka et al. 2011a, 2011b). Then the 5ʹ OH drives a nucleophilic attack resulting in the connection of the two RNA ends in a 3ʹ–5ʹ phosphodiester bond. To our knowledge the only RtcB-related reaction described that involves the 2ʹ OH is the reversal of the cyclic phosphodiesterase reaction observed when RNA ligation is prevented by a 5ʹ P (Tanaka et al. 2011a, 2011b). In this case, the 2ʹ OH acts as a nucleophile for the attack on the 3ʹ guanylated RNA and a 2ʹ,3ʹ cP is formed again. An alternative ligation mechanism having this reaction at its core (Desai and Raines 2012) was soon shown to be invalid (Chakravarty and Shuman 2012). The more efficient RNA ligation underscores that the 2ʹ OH increases the ligation rate directly or indirectly and an influence on the conformation of the terminal pentose sugar has been suggested (Chakravarty et al. 2012). This however leaves to be determined why an RNA with a terminal LNA nucleotide also does not support ligation, as in this case the pentose is locked in a C3ʹ endo conformation analogous to ribose in RNA (Supplementary Figure S2). Possibly a loss of hydrogen-bonding between RtcB and the 2ʹ OH could cause the decline in ligation efficiency. This increases the need for solving a crystal structure of RtcB that gives insight into substrate binding at the 3ʹ end and the exact role of the 2ʹ OH. Additionally, such information might help to predict an altered 2′ group that permits ligation.

The 3ʹ twister ribozyme fragment was captured from E. coli total RNA by RtcB-mediated ligation

Since we showed that only RNA adapters are suitable for the ligation to the 5ʹ OH end of ribozymes, we used this adapter in further experiments. We then applied the RtcB ligation to capture a spiked 3ʹ twister type P5 ribozyme fragment from E. coli total RNA and compared it to the previously described cyPhyRNA-seq method. cyPhyRNA-seq includes two parallel strategies, each designed to select either the 5ʹ or 3ʹ ribozyme fragment. To find the 3ʹ fragment, an alternative method is applied in which the adapter is not ligated directly to the 5ʹ OH end. Instead, 5ʹ OH-RNAs in total RNA are first enriched by degrading RNAs with 5ʹ tri- and monophosphates using RppH and Xrn1. Then, the 5ʹ adenylated adapter is ligated to the 3ʹ OH end of the enriched 5ʹ OH containing RNAs by T4 RNL2 TKQ (Olzog et al. 2021). In contrast, RtcB ligation can serve as an alternative approach, in which the adapter ligation occurs directly at the 5ʹ OH end.

In the RtcB ligation method, E. coli total RNA spiked with the 3ʹ twister type P5 ribozyme fragment was enriched for short RNAs by precipitating large RNAs with PEG8000 and NaCl (Lu et al. 2007). The samples were then used either directly for adapter ligation by RtcB or for ’tRNA blocking’ (Olzog et al. 2021). Using these approaches, large rRNAs and mature tRNAs, which are the most abundant RNA types in total RNA samples (O’Neil et al. 2013; Westermann et al. 2012), can be efficiently removed. The enriched RNA is then used for producing high-throughput sequencing libraries (Figure 3A). First, the RNA adapter is ligated to the 3ʹ ribozyme fragment by PhoRtcB ligase. Next, the second, DNA adapter with a unique molecular identifier (UMI) (Kivioja et al. 2011), which is pre-adenylated by TS2126 RNA ligase 1, is ligated to the 3ʹ OH of the ribozyme fragment with T4 RNL2 TKQ. After adapter ligation to both ends of the ribozyme fragment, the ligation product is reverse transcribed by SuperScript IV (SSIV) reverse transcriptase, subsequently PAGE-purified and used as template for PCR amplification by Phusion DNA Polymerase (Figure 3A). The produced libraries were used for Amplicon-EZ sequencing with ∼50,000 to 100,000 reads (Azenta Life Sciences). After adapter trimming and mapping, the reads were deduplicated, which means that only reads with different UMIs and thus, single ligation events, were considered. Furthermore, we only investigated uniquely mapping reads. Based on the annotation, the deduplicated and uniquely mapping reads were assigned to general RNA classes: rRNAs, tRNAs, mRNAs, other non-coding RNAs and the spiked twister type P5 ribozyme fragments (Figure 3B, Supplementary Table S1).

Figure 3:

Library preparation to capture 3ʹ ribozyme cleavage fragments from total RNA samples.

(a) Twister type P5 ribozymes cleave themselves site-specifically resulting in a 3ʹ fragment with a 5ʹ OH end and a 5ʹ fragment with a 2ʹ,3ʹ cP end. RNA adapter with 3ʹ P is guanylated and then directly ligated to the 5ʹ OH end of the twister type P5 fragment by PhoRtcB ligase. Subsequently, libraries were prepared (see text for detailed description). (b) Percentage of sequencing reads captured for selected RNA types in E. coli spiked with the 3ʹ twister type P5 ribozyme fragment. ‘Other ncRNAs’ refers to non-coding RNAs other than rRNAs, tRNAs and ribozyme fragment added. Strategy was compared to the capture method previously described for cyPhyRNA-seq (see text for further information). Both methods were carried out with and without ‘tRNA blocking’ before adapter ligation.

In all prepared libraries, we were able to detect the spiked 3ʹ twister type P5 ribozyme fragment. The ‘tRNA blocking’ results in a significant reduction of tRNA reads, which we also detected in our RtcB-based method (Figure 3B). When tRNAs are not blocked, 55% of the reads belong to tRNAs, while only 8% of all reads correspond to tRNAs when tRNA blocking was performed. Looking at the percentage of RNA types found, it appears that more rRNAs are observed when ‘tRNA blocking’ is performed. However, the actual number of reads changes only slightly, while the total number of unique and deduplicated reads decreases because fewer tRNA-based reads are generated (Supplementary Table S1). This means, using ‘tRNA blocking’, rarer RNAs can potentially be uncovered as fewer reads are siphoned off by tRNAs. Compared to the RNA enrichment followed by T4 RNL2 TKQ-mediated ligation as performed in the cyPhyRNA-seq method, more rRNA and mRNA reads were detected in the RtcB-mediated ligation (Supplementary Table 1). In order to reduce these mRNA and rRNA reads, the enzymatic enrichment of 5ʹ OH RNAs by RppH and Xrn1 in contrast to the currently used precipitation of large RNAs could be investigated in future. Furthermore, the higher number of detected rRNA and mRNA reads could also be due to increased RNA background degradation during the RtcB ligation reaction that is carried out at 70 °C in 0.5 mM MnCl₂. Therefore, further conditions could be adjusted for the use of PhoRtcB in RNA-seq-based methods, e.g. incubation time, reaction buffer (MOPS instead of Tris as published in Peach et al. 2015) or concentration of MnCl₂. Furthermore, ‘tRNA blocking’ does not lead to an increased number of captured ribozyme fragments, because the total number of captured 3ʹ twister type P5 ribozyme fragments is the same in both RtcB-based libraries (Supplementary Table 1). Using the cyPhyRNA-seq method with the T4 RNL2 TKQ, a higher number of 3ʹ twister ribozyme fragments was observed (16%) compared to the RtcB-based strategy (2%) (Figure 3B).

Use of RtcB-mediated ligations for self-cleaving ribozyme discovery and validation

Previously published methods for capturing ribozyme fragments with 5ʹ OH ends are based on RNA enrichment by enzymatic degradation of ribonucleic acids with 5ʹ PPP and 5ʹ P (Chen et al. 2021; Olzog et al. 2021). In contrast, RtcB ligase can be used to directly ligate adapters to the RNA’s 5ʹ OH end. By using EcoRtcB for RNA sequencing, 5ʹ OH-containing tRNA and mRNA fragments have been detected previously (Peach et al. 2015). Here, we apply the RtcB-mediated ligation to capture ribozyme cleavage fragments. We determined that the adapter ligation to highly structured RNAs, such as ribozyme fragments, is much more efficient with the thermostable PhoRtcB together with its Archease enzyme. The Archease is a cofactor of the PhoRtcB that facilitates a multiple turnover reaction and thus, enhances the ligation efficiency (Desai et al. 2014). However, we were not able to use DNA or 2ʹ modified RNA adapters for the ligation to ribozyme fragments. Some self-cleaving ribozyme classes retain their cleavage ability after self-scission and thus can liberate the RtcB-ligated RNA adapters. This deficit renders the RtcB-dependent ligation method less favorable for the application to self-cleaving ribozyme validation and discovery as some classes would escape detection. Nevertheless, using the PhoRtcB-based ligation, we captured spiked ribozyme fragments from total RNA. However, we demonstrate that the latest cyPhyRNA-seq method utilizing enzymatic enrichment is more efficient in capturing ribozyme fragments with 5ʹ OH, even though it does not rely on a direct ligation to RNAs with 5ʹ OH termini.

Use of RtcB-mediated ligations for screening 5ʹ OH RNAs

Even if the use of RtcB is not ideal for ribozymes, the use of PhoRtcB for the ligation-directed capture of 5ʹ OH RNAs could be useful for studying other RNAs. Identical termini are produced in a number of cellular processes (Shigematsu et al. 2018) or treatments (Lindell et al. 2002; Twittenhoff et al. 2020), for example by certain endoribonucleases (e.g. RNase A or T2) (Luhtala and Parker 2010), in stress-induced cleavage of tRNAs (e.g. oxidative stress) (Thompson et al. 2008) and in spontaneous degradation. Additionally, it could be a great complementary approach for further applications where RNAs with 2ʹ,3ʹ cP are captured by AtRNL AA (Olzog et al. 2021). Although, EcoRtcB has been previously used for transcriptome-wide detection of RNAs with 5ʹ OH (Peach et al. 2015), the use of PhoRtcB offers three advantages. First, elevated incubation temperatures enable ligation of even highly structured RNAs, second, the Archease cofactor enables PhoRtcB to perform a multiple turnover reaction increasing ligation efficiency and third, both of these factors lead to drastically shortened incubation times (20 min vs. 2 h). Therefore, the PhoRtcB ligation strategy could be used to detect highly structured 5ʹ OH RNAs in vivo on a genome-wide scale.

Expression and purification of E. coli RtcB ligase

The expression and purification of EcoRtcB was performed as described previously (Olzog et al. 2021).

Expression and purification of P. horikoshii RtcB ligase and Archease

The plasmid pET32 with the sequence of P. horikoshii Archease or RtcB ligase (gift from Ronald Raines) was used for transformation of E. coli BL21 (DE3) cells. The overexpressions of PhoRtcB ligase and Archease were carried out in the same way: 4 ml of overnight culture was used to inoculate 800 ml TB medium containing 100 μg/ml ampicillin. The main culture was incubated at 37 °C and 180 rpm until an OD₆₀₀ of 0.6 was reached. Then, 0.5 mM isopropyl-β-d-thiogalactoside (IPTG) was added and the incubation was continued for 3 h. Cells were harvested by centrifugation for 15 min at 4 °C and 4000g and stored at −80 °C.

Pyrococcus horikoshii RtcB ligase was purified as previously described (Desai and Raines 2012). Pellet was resuspended in Buffer A [50 mM Tris-HCl (pH 7.7), 300 mM NaCl, 0.5 mM DTT and 25 mM imidazole] (6 ml per g of wet pellet). Cells were disrupted by a Fastprep-24 Homogenisator (MP Biomedicals GmbH, Eschwege, Germany) for 6 m/s and 30 s. After centrifugation for 30 min at 30,600g and 4 °C, supernatant was sterile filtered and loaded onto HisTrap FastFlow column (1 ml, GE Healthcare, Freiburg, Germany) equilibrated with Buffer A. The column was washed with 40 mM imidazole (15% Buffer B). Elution was performed with Buffer B [50 mM Tris-HCl (pH 7.7), 300 mM NaCl, 0.5 mM DTT and 250 mM imidazole]. Peak fractions were pooled and used for size exclusion chromatography (SEC) using HiLoad 16/60 Superdex 75 pg (GE Healthcare) column equilibrated in Buffer C [20 mM Tris-HCl (pH 7.4), 300 mM NaCl]. Peak fractions were combined and concentrated using Vivaspin 6 (10 kDa cutoff). Aliquots of 50 µl were frozen in liquid nitrogen and stored at −80 °C.

Pyrococcus horikoshii Archease was purified as previously described (Desai et al. 2015). Pellet was resuspended in Buffer D [50 mM Tris-HCl (pH 7.7), 300 mM NaCl and 20 mM imidazole] (6 ml per g of wet pellet) and cells were disrupted by a Fastprep-24 Homogenisator for 6 m/s and 30 s. After centrifugation for 30 min at 30,600g and 4 °C, supernatant was used for a heat-kill step at 70 °C for 25 min. After another centrifugation at 30,600g and 4 °C for 20 min, supernatant was sterile filtered and loaded onto HisTrap FastFlow column (5 ml, GE Healthcare) equilibrated with Buffer D. The column was washed with 45 mM imidazole (18% Buffer E). Elution was performed with Buffer E [50 mM Tris-HCl (pH 7.7), 300 mM NaCl and 250 mM imidazole]. Peak fractions were pooled and dialyzed in Buffer F [10 mM HEPES-NaOH (pH 7.5) and 200 mM NaCl] overnight at 4 °C. Aliquots of 50 µl were frozen in liquid nitrogen and stored at −80 °C.

RNA synthesis and preparation of oligonucleotides

Oligonucleotides were ordered from biomers.net (Ulm, Germany) or Microsynth AG (Balgach, Switzerland). Templates for ribozymes were generated by PCR amplification using a 50 µl PCR reaction containing 1× Pfu DNA Polymerase Buffer [20 mM Tris-HCl (pH 8.8), 10 mM KCl, 10 mM (NH₄)₂SO₄, 0.1% Triton X-100, 0.1 mg/ml BSA, 2 mM MgSO₄ × 7 H₂O], 200 µM dNTPs, 4 or 0.4 µM forward and reverse primer (Supplementary Table S2) and 1 µl of the Pfu DNA Polymerase (own laboratory purification). Initial denaturation step was performed at 94 °C for 5 min. Subsequent PCR cycles include a denaturation step at 94 °C for 10 s, annealing step at the respective annealing temperature for 10 s, elongation step at 68 °C for 15–30 s and were performed for 25 cycles. Final elongation step was carried out at 68 °C for 2 min. PCR reactions were purified using phenol/chloroform extraction and ethanol precipitation (Sambrook and Russell 2006).

Adapters were radioactively labeled using [γ³²P]-ATP (Hartmann Analytic) and T4 Polynucleotide Kinase D167N (Das and Shuman 2013). Model substrates and ribozyme cleavage fragments (Supplementary Table S3) were generated by T7-based in vitro transcription using purified PCR products (Ling et al. 1989) and [α³²P]-ATP (Hartmann Analytic, Braunschweig, Germany). Then, ribozyme fragments or oligonucleotides were PAGE-purified as described previously (Lihanova and Weinberg 2021).

Ligations using EcoRtcB

First, 1 µM adapter with 3ʹ P, 0.5 µM 3ʹ ribozyme fragment with 5ʹ OH (Supplementary Table S3) and diethylpyrocarbonate (DEPC)-H₂O were incubated at 65 °C for 5 min and cooled down on ice. Then, 50 mM Tris-HCl (pH 7.4), 2 mM MnCl₂, 100 µM GTP, 15% polyethylene glycol 8000 (PEG8000), 5% dimethyl sulfoxide (DMSO) and 5 µM EcoRtcB were added. Ligation reactions were incubated at 37 °C for 2 h. Reactions were stopped by adding 3× RNA loading dye [10 mM Tris-HCl (pH 7.6 at 23 °C), 80% formamide, 0.25% bromophenol blue and 0.25% xylene cyanol]. The ligation was analyzed by polyacrylamide (PAA) gel electrophoresis using a 15–20% denaturing PAA gel. Ligation efficiency was calculated from measuring band intensities using ImageQuant.

Ligations using PhoRtcB

Adapter with 3ʹ P (1 µM), 3ʹ ribozyme fragment with 5ʹ OH (0.5 µM, Supplementary Table S3) and DEPC-H₂O were incubated 5 min at 65 °C and cooled down on ice. Afterwards, 50 mM Tris-HCl (pH 7.4), 0.5 mM MnCl₂, 100 µM GTP, 300 mM NaCl, 15% PEG8000, 2 µM Archease and 5 µM RtcB ligase were added to the adapter/substrate-mix. After the incubation at 70 °C for 20 min, ligation reactions were stopped by adding 3× RNA loading dye . The ligation was analyzed by polyacrylamide (PAA) gel electrophoresis using a 15–20% denaturing PAA gel. The ligation efficiency was calculated from measuring band intensities using ImageQuant.

Spike-in experiment

Analysis of reads generated by amplicon sequencing

Libraries of spike-in experiments were analyzed based on Amplicon-EZ sequencing (Amplicon-EZ, Azenta Life Sciences, Supplementary Table S2). The reads were treated as follows: (1) reads were preprocessed with UMItools (Smith et al. 2017) to identify and trim UMI sequences, (2) trimming of remaining adapters with TrimGalore (Felix Krueger), (3) mapping of reads and (4) counting. To map the reads, an artificial genome was created by adding the spiked-in 3ʹ twister type P5 ribozyme fragment sequence (Supplementary Table S3) to the E. coli genome retrieved from https://www.ncbi.nlm.nih.gov/genome/167?genome_assembly_id=753562. Then, the artificial genome was indexed and reads were mapped with STAR (Dobin et al. 2013). Command-line calls for these workflow steps are listed in the Supplementary text. Reads mapping only at one position in the genome were extracted and counted with featureCounts (Liao et al. 2014). A framework for the semi-automated analysis of the data presented herein is available at https://github.com/jfallmann/MONSDA (manuscript in preparation). Last, reads were manually inspected and deduplicated for reads mapping to spiked twister ribozyme fragment.