Rolling circle RNA synthesis catalyzed by RNA
Abstract
RNA-catalyzed RNA replication is widely considered a key step in the emergence of life’s first genetic system. However, RNA replication can be impeded by the extraordinary stability of duplex RNA products, which must be dissociated for re-initiation of the next replication cycle. Here, we have explored rolling circle synthesis (RCS) as a potential solution to this strand separation problem. We observe sustained RCS by a triplet polymerase ribozyme beyond full-length circle synthesis with strand displacement yielding concatemeric RNA products. Furthermore, we show RCS of a circular Hammerhead ribozyme capable of self-cleavage and re-circularization. Thus, all steps of a viroid-like RNA replication pathway can be catalyzed by RNA alone. Finally, we explore potential RCS mechanisms by molecular dynamics simulations, which indicate a progressive build-up of conformational strain upon RCS with destabilization of nascent strand 5′- and 3′-ends. Our results have implications for the emergence of RNA replication and for understanding the potential of RNA to support complex genetic processes.
Editor's evaluation
This paper is of interest to scientists from the field of origin of life or RNA synthesis in general, especially those interested in the "RNA world" scenario. The data analysis is rigorous and the conclusions are justified by the data. The key claims of the manuscript are directly related to, and support, previous findings.
https://doi.org/10.7554/eLife.75186.sa0eLife digest
Many organisms today rely on a trio of molecules for their survival: DNA, to store their genetic information; proteins, to conduct the biological processes required for growth or replication; and RNA, to mainly act as an intermediary between DNA and proteins. Yet, how these inanimate molecules first came together to form a living system remains unclear.
Circumstantial evidence suggests that the first lifeforms relied to a much greater exrtent on RNA to conduct all necessary biological processes. There is no trace of this ‘RNA world’ today, but molecular ‘fossils’ may exist in current biology. Viroids, for example, are agents which can infect and replicate inside plant cells. They are formed of nothing but a circular strand of RNA that serves not only as genetic storage but also as ribozymes (RNA-based enzymes). Viroids need proteins from the host plant to replicate, but scientists have been able to engineer ribozymes that can copy complex RNA strands. This suggests that viroid-like replication could be achieved using only RNA.
Kristoffersen et al. put this idea to the test and showed that it is possible to use RNA enzymatic activity alone to carry out all the steps of a viroid-like copying mechanism. This process included copying a viroid-like RNA circle with RNA, followed by trimming the copy to the right size and reforming the circle. These two latter steps could be carried out by a ribozyme that could itself be encoded on the RNA circle. A computer simulation indicated that RNA synthesis on the circle caused increasing tension that could ease some of the barriers to replication.
These results increase our understanding of how RNA copying by RNA could be possible. This may lead to developing molecular models of a primordial RNA-based replication, which could be used to investigate early genetic systems and may have potential applications in synthetic biology.
Introduction
The versatility of RNA functions underpins hypotheses regarding the origin and early evolution of life. Such hypotheses of an ‘RNA world’—a primordial biology centered on RNA as the main biomolecule—are in accord with the essential role of RNA catalysis in present-day biology (Cech, 2000; Goldman and Kacar, 2021; Nissen et al., 2000; Wilkinson et al., 2020) and the discovery of multiple prebiotic synthetic pathways to several of the RNA (and DNA) nucleotides (Becker et al., 2019; Kim et al., 2020; Patel et al., 2015; Powner et al., 2009; Xu et al., 2020). In addition, progress in both non-enzymatic (Deck et al., 2011; Hassenkam et al., 2020; Prywes et al., 2016; Rajamani et al., 2008; Sosson et al., 2019; Sponer et al., 2021; Wachowius and Holliger, 2019; Zhang et al., 2020; Zhou et al., 2020) and RNA-catalyzed polymerization of RNA and some of its analogs (Attwater et al., 2018; Attwater et al., 2013; Cojocaru and Unrau, 2021; Ekland and Bartel, 1996; Horning and Joyce, 2016; Johnston et al., 2001; Mutschler et al., 2018; Shechner et al., 2009; Tagami et al., 2017; Tjhung et al., 2020) is beginning to map out a plausible path to RNA self-replication; a cornerstone of the RNA world hypothesis.
RNA in vitro evolution and engineering have led to the discovery of RNA polymerase ribozymes (RPRs) able to perform templated RNA synthesis of up to ~200 nucleotides (nt) (Attwater et al., 2013), synthesizing active ribozymes including the catalytic class I ligase core (Horning and Joyce, 2016; Tjhung et al., 2020) at the heart of the most efficient RPRs, as well as initiate processive RNA synthesis using a mechanism with analogies to sigma-dependent transcription initiation (Cojocaru and Unrau, 2021). An RPR capable of utilizing trinucleotide triphosphates (triplets) as substrates (a triplet polymerase ribozyme [TPR]) has been shown to display an enhanced capacity to copy highly structured RNA templates including segments of its own sequence (Attwater et al., 2018).
Nevertheless, there remain a number of fundamental obstacles to be overcome before an autonomous self-replication system can be established. A central problem among these is the so-called ‘strand separation problem,’ a form of product inhibition due to the accumulation of highly stable dead-end RNA duplexes, which cannot be dissociated (efficiently) under replication conditions (Le Vay and Mutschler, 2019). The strand separation problem has been overcome by PCR-like thermocycling (or thermophoresis) (Horning and Joyce, 2016; Salditt et al., 2020), but this approach may be limited to short RNA oligomers (even in the presence of high concentrations of denaturing agents) as the melting temperatures of longer RNA duplexes approach or even exceed the boiling point of water (Freier et al., 1986; Szostak, 2012).
While RNA duplexes occur by necessity as intermediates of RNA replication, the extent of the strand separation problem can be modulated by genome topology. Circular rather than linear genomes are widespread in biology including eukaryotes, prokaryotes, and viruses (Moller et al., 2018; Moss et al., 2020; Shulman and Davidson, 2017). Circular RNAs (circRNAs) are found as products of RNA splicing (Kristensen et al., 2019) and RNA-based self-circularization is known in multiple ribozymes (Hieronymus and Müller, 2019; Lasda and Parker, 2014; Petkovic and Müller, 2015). Continuous templated RNA synthesis on circular templates (rolling circle synthesis [RCS]) is also widespread and found in the replication of the RNA genomes in some viruses and in viroids. Indeed, viroid RNA replication has been proposed to resemble an ancient mechanism for replication (Diener, 1989; Flores et al., 2014).
In an idealized RCS mechanism, both strand invasion and displacement processes are isoenergetic and coordinated to nascent strand extension (Blanco et al., 1989; Daubendiek et al., 1995), with rotation of the single-stranded RNA (ssRNA) alleviating the build-up of topological tension (Kuhn et al., 2002). Thus, RCS is a potentially open-ended process leading to the synthesis of single-stranded multiple repeat products (concatemers) with an internally energized strand displacement (Tupper and Higgs, 2021). RCS as a replication mode has therefore potentially unique properties with regards to the strand separation problem. Specifically in the context of triplet-based RNA replication on a circular template, duplex dissociation, and strand separation may in principle be driven by trinucleotide (triplet) hybridization and ligation, leading to extension of the nascent strand 3′-end and an equal displacement of the 5′-end in triplet increments (Figure 1A). Triplet binding to the template strand and dissociation of an equal trinucleotide stretch from the 5′-end are both equilibrium processes and nearly isoenergetic. However, extension (i.e., ligation of the bound triplet to the growing 3′-end) is an irreversible step. Thus, in this scenario, RCS would be expected to proceed in ratchet-like fashion with strand displacement driven by triphosphate hydrolysis and triplet ligation.

Primer extension on circular RNA (circRNA) templates.
(A) Schematic illustration of rolling circle synthesis (RCS). Red product RNA strand is extended by a triplet at the 3′-end while three base pairs dissociate at the 5′-end keeping the total hybridization energy constant. Topological relaxation is allowed by rotation of the single stranded part of the circular template (swiveling arrow). (B) Linear or circularized RNA is treated with or without endo- or exonucleases (RNase A/T1 mix or Exonuclease T (ExoT)). Only circRNA is resistant to ExoT digestion. (C) Schematic representation of primer triplet extension on a linear or circRNA template and of the TPR hetero-dimer comprising the catalytic subunit (5TU (green)) and the non-catalytic subunit (t1 (red)) (Attwater et al., 2018) (TPR sequence: Supplementary file 1). (D) PAGE gel of TPR primer extension, P9 (5′-FAM-GAAGAAGAA) is the unextended primer, bands 1–9 denote extension of P9 by +1 to + 9 triplets (full length). RNA template used was sc12GAA-p (36 nt, 12 repeat UUC template). Experiment was done under standard conditions in the Tris buffer system described in Materials and Methods except that 800 pmol triplets were used. (E) Extension efficiency of formation of bands 1–9 in (D) (see Materials and methods) is plotted against triplet position. (F) Schematic model of the sc12GAA-p illustrating the different accessibility of in- or outside facing triplet junctions on the scRNA template (blue), with P9 primer (red), and the product strand (light gray). Original gel images and numerical values are supplied in Figure 1—source data 1.
-
Figure 1—source data 1
Gel images and numerical values.
- https://cdn.elifesciences.org/articles/75186/elife-75186-fig1-data1-v2.zip
Here, we have explored triplet-based RCS of scRNA templates as a potential solution to the strand separation problem in RNA-catalyzed RNA replication. We show that RCS can be catalyzed by the triplet polymerase ribozyme [TPR] (Attwater et al., 2018). We show that the TPR is able to perform continuous templated extension of circular RNA templates beyond full-length circle synthesis with strand displacement yielding concatemeric RNA products. We also investigated the mechanistic basis for RCS and strand displacement by molecular dynamics (MD) simulations of scRNA in explicit solvent. Finally, we consider the potential of a full viroid-like replication cycle catalyzed by RNA alone by design and synthesis of a circular Hammerhead ribozyme capable of both product cleavage and self-circularization.
Results
RNA-catalyzed primer extension using small circular RNA templates
We first set out to investigate whether templated RNA synthesis on scRNAs as templates could be catalyzed by an RNA catalyst. To extend the RNA nascent strand beyond the full circle and initiate RCS requires duplex invasion and displacement of the original RNA primer and product strand. However, most RPRs are inhibited by duplex RNA both in the form of template secondary structures and as linear duplex RNA. We therefore explored the potential of a recently described TPR (Attwater et al., 2018), which is able to utilize trinucleotide triphosphates (triplets (pppNNN)) as substrates for polymerization. Due to increased binding of the triplets to the template (compared, e.g., to the canonical mononucleotide triphosphates (pppN, NTPs)), triplets are able to invade and cooperatively ‘open up’ template secondary structures for replication (Attwater et al., 2018). We hypothesized that this ability might also promote the continuous invasion and displacement of the nascent strand 5′-end and facilitate the RCS mechanism (Figure 1A).
As described previously, RNA synthesis by the TPR is most efficient in the eutectic phase of water ice, due to its beneficial reaction conditions for ribozyme catalysis (Attwater et al., 2018). Specifically, eutectic ice phases aid TPR activity by the reduced degree of RNA hydrolysis under low-temperature conditions, reduced water activity, and the high concentrations of reactants (ribozyme, scRNA template, triplet substrates, and Mg2+ ions) present in the eutectic brine phase that arises by excluding solutes from growing water ice crystals and remains liquid at subzero temperatures (Attwater et al., 2010). Thus, all RCS experiments were carried out under eutectic conditions.
We prepared scRNA templates (34–58 nt in length) by in vitro transcription and ligation. Circularity of the scRNA templates was verified based on gel mobility and resistance to exonuclease (exoT) degradation in contrast to the linear, non-ligated counterparts (Figure 1B, Figure 1—figure supplement 1/sequences for all oligonucleotides in Supplementary file 1). We first investigated primer extensions using a single triplet (pppGAA) on the scRNA template as this provides an even banding pattern of incorporation. This facilitates analysis and allows primer extension efficiencies of linear and circular templates to be more readily compared (Figure 1C). Primer extension experiments using a purified 36 nt scRNA as template resulted in full-length extension around the circle (Figure 1D), but with reduced efficiency compared to a linear RNA template. Furthermore, we observed a periodic banding pattern of triplet extension efficiency matching the helical pitch of double-stranded RNA (dsRNA) (11.3 base pairs (bp)/turn Bhattacharyya et al., 1990; Figure 1E). Presumably, triplet junctions located on the inside of the scRNA ring are less easily accessible and therefore less efficiently ligated than in linear RNA, which is freely accessible from all sides (Figure 1F). This may explain both the observed periodicity and reduced synthetic efficiency of RNA synthesis by the TPR on scRNAs.
Despite the reduced extension efficiency in scRNA, we obtained full circle extension products for multiple templates (34–58 nt in size, Figure 2A and B) with a clear trend toward increasing mean extension efficiency for circular templates with increasing size predicting parity with the linear template at around 120 nt (Figure 2C). Note, in these experiments extension beyond full circle was not intended or possible (lanes 1–6 in Figure 2B) as the specific triplet substrates needed for displacing the primer were not present in the reaction.

Full-length and beyond full-length RNA-catalyzed RNA synthesis on circular RNA templates.
(A) Product strand of primer extension experiments with primer P10 (red) and eight triplet scRNA template strands. Potential beyond full-length circle synthesis triplets are shaded opaque. (B) Various scRNA template sizes allow full-length primer extension as indicated (with eight triplet sites) (gray), GAA triplets (black), and primers P10 (5′-FAM-CUGCCAACCG) or P10 +3 (5′-FAM-GAAGAAGAA-CUGCCAACCG) (red). PAGE of primer extensions (under standard conditions) with full-length synthesis for different scRNA templates (34–58, termed scGAA8-16, Supplementary file 1) marked by a black line. (C) Mean extension efficiency plotted as a function of scRNA size calculated from extension experiments including (B) (error bars indicate standard deviation, n=5), with mean extension efficiency for a linear RNA template (red dashed line). (D) PAGE of time-course of primer extension of primer P9 on scRNA template, sc12GAA-p (optimized conditions). Full-length circle synthesis is marked by a dashed black line (after +9). Bands +10 and beyond (see Enhanced contrast section) indicate beyond full-length synthesis. (E, F) Mean extension efficiency (from (D)) plotted against time (E) or triplet position (F), showing the respective amounts of product at full-length (black) and beyond full-length circle (red) synthesis as well as the drop in efficiency at full length, which recovers once beyond full-length synthesis is initiated. Vext and Vinv denote the calculated velocity of formation of bands 9 and 10, respectively. Original gel images and numerical values are supplied in Figure 2—source data 1.
-
Figure 2—source data 1
Gel images and numeric values.
- https://cdn.elifesciences.org/articles/75186/elife-75186-fig2-data1-v2.zip
Having established full-length synthesis on scRNA templates, we next tested if primer extension could proceed beyond full circle requiring duplex invasion and displacement of the primer/product strand. We first tested this using primer P10 +3, comprising a 5′ extension of three GAA repeats, thus covering the last three UUC triplet binding sites on the circular template (Figure 2B, top right). We observed an extension of up to +3 triplets (+9) above the full circle mark (Figure 2B, lanes 7–12), indicating displacement of the primer 5′-end upon incorporation of three additional pppGAA triplets. This indicated that synthesis beyond full circle including strand displacement is possible on scRNA templates, boding well for the implementation of RCS. To that effect, we next optimized buffer and extension conditions for more efficient extension above the full circle mark (Figure 2—figure supplement 1). Interestingly, greater dilution of reaction mixtures prior to freezing resulted in more efficient stand displacement. While greater pre-freezing dilution does not alter the final solute concentrations within the eutectic phase (Attwater et al., 2010), it increases the eutectic phase/ice interface area. This suggests that strand invasion may be aided by surface effects, as previously suggested for RNA refolding (Mutschler et al., 2015). Under these optimized buffer and extension conditions, we observed progressive accumulation of progressively longer RCS products, over prolonged reaction times (up to 6 weeks) (Figure 2D) with reaction speed decreasing after ca. 4 weeks incubation, indicating continued RCS over extended periods of time (Figure 2E and F).
Molecular dynamics simulations of 36 nt scRNA
To better understand the structural and topological constraints of RCS on scRNAs, we performed atomistic MD simulations over 400 ns of the different RCS stages, comprising the starting scRNA template as circular ssRNA and scRNA with a progressively extended dsRNA segments (Figure 3). For simplicity, a 36 nt circular RNA sequence of (UUC)12 was chosen as a template strand (similar to the scRNA template [sc12GAA-p] in Figures 1 and 2D) for direct comparison with the experimental system. The complementary strand comprising GAA triplets starting from 9 bp dsRNA (corresponding to binding of primer P9) was extended from 18 to 30 bp (after +3 triplet incorporation) in triplet increments of dsRNA corresponding to extension products of bands +3, +4, +5, +6, and +7 (see gel in Figure 1D), using the most representative structure of the previous simulation as a starting point for the next one.

Molecular dynamics simulation of small circular RNA.
(A) Main conformations (and zoom-in to relevant regions [squares]) observed from simulations in 100 mM MgCl2 on scRNA exploring consecutive states of primer extension, from 9 to 30 bp dsRNA with pyrimidine (template) strand (UUC)12 (khaki), purine (product) strand (GAA) (light blue), 5′-end and unpaired bases (dark blue) and 3′-end unpaired bases (purple) and matching melted bases from the template strand (dark green (5′-end)/light green (3′-end)). (B) Percentage of frames from the last 100 ns of the simulations presenting canonical hydrogen bond pairing for each bp. (C) Counterion-density maps (in red) around RNA molecules that show an occupancy ~10 times or greater than the bulk concentration.
The simulation trajectories revealed the high energy barrier of dsRNA for bending and accommodating a circular shape (Figure 3A). Instead, we observe that, as dsRNA is elongated, the remaining ssRNA segment of the scRNA becomes increasingly extended. As the dsRNA part reaches 27 bp (corresponding to band 6 in Figure 1D), the ssRNA segment was fully extended and torsional strain was relieved by dissociation (peeling off) of the dsRNA 5′- and 3′-ends rather than by bending or the introduction of kinks into the dsRNA segment (Figure 3B). Subsequently, multiple peeling off and rebinding events were observed during the trajectories indicating that the dynamics of this process are fast (Videos 1 and 2 and Figure 3—figure supplements 1 and 2).
Movie of the RCS simulation where dsRNA is 27-bp long.
We observe fraying and annealing of 5′- and 3′-ends demonstrating the quick timescales of these transitions.
Movie of the RCS simulation where dsRNA is 30-bp long.
We observe again fraying and annealing of 5′- and 3′-ends demonstrating the quick timescales of these transitions.
In the experimental data, we also observed an inhibitory effect for insertion of the final triplets (+8, +9, and +10 (beyond full length)/extension to 33, 36, and 39 nt of RNA in Figure 2D) into the corresponding scRNA template. This may indeed reflect the onset of the 3′- and 5′-ends destabilization observed in the MD simulations (Figure 3), which would likely attenuate primer extension by the ribozyme. Note however that the extension efficiency recovered beyond full length (+11/extension to 41 nt, Figure 2F), although at lower speed (Figure 2E).
As a control for the observed dsRNA end destabilization mechanism, we also ran an MD simulation of a linear RNA molecule containing four consecutive triplets with a nick between two of them, but observed neither base opening nor dissociation at either 5′- or 3′-strand ends (Figure 3—figure supplement 3). Groove dimensions and local helical parameters (roll, twist, and slide) for the RCS simulations on circular RNA did not show any major adjustment compared with the linear RNA control (Figure 3—figure supplement 4). We observed an oscillation of high/low values of bending along the molecule in phase with RNA-turn periodicity in an attempt to create an overall curvature (Velasco-Berrelleza et al., 2020), although with moderate success (~60° on an arc length of 30 bp of dsRNA) and no formation of kinks, internal loops or other disruption of the canonical A-form typical of the RNA duplex (Figure 3—figure supplement 4).
To mirror the experimental eutectic phase conditions, simulations were run at relatively high Mg2+ concentrations (100 mM) and compared with the presence of monovalent ions like K+ (200 mM) and high concentration of Mg2+ (500 mM), but simulations did not show any major differences in terms of dsRNA strand dissociation or bending (Figure 3—figure supplements 3 and 4). However, Mg2+ ions—compared to K+—appear to interact more strongly with different parts of the RNA and, consequently, may increase the probability of distorted conformations facilitating the exposition of nucleobases at the 5′- and 3′-ends. In contrast, K+ ions are mainly positioned as counterions along the major and minor groove, allowing the bases to orient toward the inside of the dsRNA helix for base-pairing interactions (Figure 3C and Figure 3—figure supplement 5). The role of Mg2+ in the stabilization of complex RNA folding has been observed repeatedly in several structures (Sponer et al., 2018), like the ribosome (Klein et al., 2004) and the Hepatitis delta virus ribozyme (Nakano et al., 2001). However, increasing MgCl2 concentration to 500 mM does not seem to bring extra stabilization, as the system appears to be saturated already at 100 mM Mg2+ (Figure 3—figure supplements 5 and 6).
In summary, our simulations support the notion that progressive RNA synthesis on a scRNA template (in the presence of Mg2+ ions) leads to increased dynamics of nucleobase exposure, RNA duplex destabilization, and 5′- and 3′-ends melting. The simulations also clearly illustrate the implausibility of a small circular fully dsRNA molecule (as schematically illustrated in Figure 1E), due to the prohibitive energetic cost of dsRNA bending. Instead, the system appears to relieve internal strain by extending the ssRNA segment of the circle (partially shielding the dsRNA segment) and peeling of both dsRNA 5′- and 3′-ends (Figure 3), consistent with the helical period of triplet extension observed (Figures 1 and 2) (with ligation junctions facing into the ssRNA center being less accessible) and the observed reduction in RCS efficiency. Dynamic destabilization of dsRNA 5′-ends clearly has the potential to facilitate strand displacement during RNA replication on a scRNA template (and may aid continuous RCS), but at the same time may reduce the efficiency of primer extension and triplet incorporation by reducing the availability of the primer 3′-end and the downstream template bases. These effects would be predicted to manifest themselves in RNA circles up to 200 bp as suggested by RNA persistence length (Abels et al., 2005).
Templated rolling circle RNA synthesis
Having validated RNA synthesis on scRNA templates (Figures 1 and 2), we next sought to show RCS beyond a single full-length circle synthesis involving displacement of the primer/nascent strand. To this end, we designed barcoded templates that would allow us to distinguish RNA synthesis products arising from template-instructed RCS from those arising from non-templated terminal transferase (TT) activity of the TPR by sequencing. The barcoded scRNA templates (termed A–D) were prepared either as circular or linear RNAs comprising different internal triplet ‘barcodes’ of variable GC-content (at positions 3, 6, and 9) for individual identification (Figure 4A and Figure 4—figure supplement 1). We performed one-pot primer extension experiments, in which all four templates (either A–D linear or A–D circular) were mixed in equal proportions. After extension, products were separated by gel electrophoresis and the gel sections above full-length extension products were excised and their RNA content recovered, and sequenced (Figure 4—figure supplement 1). Diagram in Figure 4B represents (in %) which triplets were identified at the noted positions. Boxes mark expected triplet according to the template sequence.

RNA-catalyzed RNA synthesis beyond full length for circular templates.
(A) Product strands of primer extension experiments with linear and scRNA templates A–D with primer P91. Opaque sequence illustrate potential beyond full-length synthesis on scRNA. Barcode triplets at positions 3 (A/U rich) (cyan), 6 (mix) (blue), and 9 (G/C rich) (purple) allow identification of product RNAs. Barcode triplet at position 12 is the final triplet of primer P91 and at position 15 is the same as that of position 3 but after beyond full-length circle synthesis on the scRNA template. (B) Fidelity heatmap of the sequences derived from the one-pot experiments with linear (left) or circular (right) templates. Red color indicates high prevalence of a given triplet (vertical axis) at the position noted (3–18). n: denotes the number of sequence reads (3E5=3×105) used to calculate the fidelity for each triplet at the given position. Transparent gray boxes cover positions with n≤5. (C) Plot shows ratio (fold difference) of the probability of a product of reaching positions 4–12 on circular compared to linear templates. Fold difference was calculated based on fidelity data presented in (B). (D) Model illustrating (1.) beyond full-length extension on a circular template (templated RCS) and (2.) on a linear template (non-templated). Full analysis of the data in B is supplied in Figure 4—source data 1.
-
Figure 4—source data 1
Full analysis of sequencing data used in Figure 4B.
- https://cdn.elifesciences.org/articles/75186/elife-75186-fig4-data1-v2.zip
Analysis of the sequencing products from the one-pot reaction showed template-dependent high-fidelity RNA synthesis up to full length (position 9) for all templates (linear and circular) (Figure 4B). Further, all templates yielded longer than full-length products indicative of continued RNA synthesis by the TPR beyond full length (positions>9). However, the fidelity dropped after full length was reached, indicative of significant non-templated terminal transferase-like (TT) activity in this regime (Figure 4B). For example, the average fidelity for insertion of the expected triplet (pppGAA) for position 10 (full length +1) for circular templates was 10.9% whereas for position 9 (full length) it was 89.9%. For linear templates, the fidelity for full length +1 dropped to 0.7% compared to full length 78.8%. Note, that fidelity at full length +1 dropped much more for linear than for circular templates. For this reason, the probability of a product extending to longer than full length (positions 10–12) with the correct sequence was 10–100-fold higher for circular compared to linear RNA templates (Figure 4C). A few events of blunt-end ligation with other template/product strands (see, e.g., position 15 for linear templates C and D) (Figure 4B) were also observed for linear templates.
On all circular templates (with the exception of template B, where too few reads were obtained) extension beyond full length (while containing a significant TT component) continued to insert the barcode triplets correctly, indicating continuous RCS at least up to position 18 (63 nt, more than 1.5 times full-length circle synthesis on the scRNA template). Control experiments, with individually incubated templates (in contrast to the one-pot experiments) mixed after gel purification, showed essentially identical results (Figure 4—figure supplement 2A). Interestingly, non-diluted samples had a decreased fidelity at position 10 (the point of strand invasion) compared to diluted samples (Figure 4—figure supplement 2B) suggesting that dilution appears to aid not only extension efficiency (Figure 2—figure supplement 1D), but also strand invasion fidelity and continued templated synthesis. In summary, these results are consistent with RNA-catalyzed RCS on scRNA templates beyond the full circle.
Multiple repeat rolling circle products
Next, we sought to test if RCS efficiency could be increased by double priming on the circular RNA template, an approach known as branched RCS (Berr and Schubert, 2006). Indeed, we observed a higher degree of RCS with strand displacement with the 36 nt scRNA template termed sc8211 having two identical primer sites leading to two different products being formed (product I or II) (Figure 5A and B). In order to test the primer site functionality individually, we used different triplet combinations (Figure 5B). When only the pppGAA triplet was present, primers were extended only by two triplets as expected (lane 1 in Figure 5B) (with a small amount of non-templated TT incorporation of a third triplet). When pppGAA and pppAUA or pppGCG triplets were added, respectively (lane 2 or 3 in Figure 5B), products extended up to five triplet-incorporations with extension stopping at triplet 6 (coding for CUG) showing that both primer sites were functioning. Finally, when all triplets (pppGAA, pppAUA, pppCGC, and pppCUG) were present, extension continued to beyond full circle (positions≥10) (Figure 5B) and bands corresponding to extension products exceeding two times full-length circle synthesis (>triplet 21 (63 nt)) of replication were observed (Figure 5C).

RNA-catalyzed branched RCS.
(A) Product strand of primer extension experiments with scRNA template containing two priming sites (I and II) for primer P91. Depending on the priming site two different products will be made (I or II). (B, C) Scheme and PAGE of primer extension experiments with only the noted triplets added with (C) long electrophoretic separation to achieve optimal resolution of long products. Cycling (Cyc.) indicates that the samples had been exposed to four thermal and freeze-thaw cycles (80°C 2 min, 17°C 10 min, −70°C 5–15 min, and −7°C 7 days) leading to increased efficiency. (D) Sequencing of longer than full-length branched RCS products on the double primer site scRNA (without cycling). Products I and II both reaching almost three full rounds of replication of the circular RNA template (up to 96 nt, 32 triplet incorporations). Original gel images and full analysis of the data in D are supplied in Figure 5—source data 1.
-
Figure 5—source data 1
Gel images and full data analysis of sequencing data used in Figure 5D.
- https://cdn.elifesciences.org/articles/75186/elife-75186-fig5-data1-v2.zip
Sequencing of the long, branched RCS RNA products (excised from gel band corresponding to ≥15 triplets, Figure 5—figure supplement 1) identified a range of long reads (from both products I and II) including many reads of the product with 15 correct triplet incorporations (Figure 5E) representing ~1.5 times full-length circle synthesis (n: 7×103 and 1×105 reads of products I and II, respectively). However, much longer sequences were present in decreasing numbers of reads, with the longest products comprising 29 correct triplet incorporations (96 nt) (n=2) representing RCS of >2.5 times full-length circle synthesis and the longest reported product synthesised by the TPR. Thus, RNA-catalyzed RCS has the potential to yield extended RNA concatemer products under isothermal conditions. Freeze-thaw (FT) cycles have been previously shown to enhance ribozyme activity by affecting RNA refolding (Mutschler et al., 2015) and indeed inclusion of four FT cycles lead to more efficient production of longer RCS RNA products (Figure 5B and C). In summary, isothermal branched RCS yields long concatemeric products comprising >2.5 tandem repeats of the scRNA template with RCS efficiency further enhanced by FT cycling.
Proto-viroid like self-circularizing ribozyme
A number of biological systems including viroids use an RCS strategy for genome replication. However, RCS synthesis of RNA concatemers is only one part of the viroid replication cycle, which also involves resection (i.e., cleavage) of the concatemer into individual units and circularization of unit length RNAs by ligation to recreate the original circular RNA genome. As both RNA cleavage and RNA ligation can be efficiently catalyzed by RNA, we sought to investigate, if a viroid-like replication cycle might be catalyzed by RNA alone. To this end, we designed a proto-viroid RNA comprising a 39-nt scRNA template encoding a designed micro-hammerhead ribozyme (µHHz) as well as its substrate for cleavage (Figure 6A). We envisaged a viroid-like replication cycle (schematically illustrated in Figure 6B) comprising primer extension on a circular template (step 1) reaching full-length circle synthesis and beyond (RCA, step 2) resulting in formation of the µHHz and its 3′-end substrate. When separated from its template, the µHHz exists in (at least) two conformations (step 4) of which one is the active Hammerhead ribozyme cleaving the 3′-end substrate yielding a 2′,3′-cyclic phosphate (>p) (step 5). Finally, the ribozyme re-ligates (chemical mechanism illustrated in Figure 6B) to form a circular product strand (steps 6 and 7). The µHHz could be synthesized by the TPR (Figure 6C). Furthermore, the µHHz could catalyze both self-cleavage (forming a 2′,3′-cyclic phosphate (>p)) and re-ligation leading to circularization (under RCS reaction conditions at −7°C in eutectic ice) (Figure 6D, lane 2). A similar equilibrium between cleavage and ligation in eutectic ice had previously been observed for the unrelated hairpin ribozyme (Mutschler et al., 2015).

Steps of a viroid-like replication cycle catalyzed by RNA alone.
(A) Illustration of the µHHz (−) and (+) strand. (B) Schematic illustration of the RNA-catalyzed viroid-like replication with steps comprising RNA-catalyzed combined RCS (steps 1–3), resection (steps 4 and 5), and self-circularization (steps 6 and 7). Panel between steps 5 and 6 illustrates the presumed chemical structures of the transition state of cleavage and re-ligation between the µHHz 3′- and 5′-ends. Blue arrow denotes the re-ligation reaction by bond formation with the 5′-nucleotide (forming the circle) and transesterification of the 2,3-cyclic phosphate bond (>p). Red arrows denote cleavage of the phosphate diester backbone (cleaving the circle) and formation of the 2′,3′-cyclic phosphate (>p). X and Y denote acids/bases involved in the transesterification reaction. (C) PAGE gel showing primer extension of RCS synthesis of the µHHz with substrate overhang to allow self-cleavage. scHHz temp was used as template and PHHz was used as primer (Pri). (D) PAGE gel showing cleavage and circularization of a µHHz, but only when incubated at –7°C, allowing eutectic phase to form, and with a free 5′-OH, needed for circularization, but not for cleavage. Original gel images are supplied in Figure 6—source data 1.
-
Figure 6—source data 1
Gel images.
- https://cdn.elifesciences.org/articles/75186/elife-75186-fig6-data1-v2.zip
When the µHHz had been 5′-phosphorylated (Figure 6D, lane 3) only cleavage but no circularization was seen, as phosphorylation blocks the 5′-hydroxyl required as the nucleophile for re-ligation (see steps 5 and 6, Figure 6B). To the best of our knowledge, the µHHz is the smallest (39 nt) self-cleaving and -circularizing RNA system reported to date and the first-time self-circularization has been shown in a Hammerhead ribozyme. Kinetic analysis of the cleavage and circularization reaction show a slow but accumulating amount of cleavage product as a function of time (black points in Figure 6—figure supplement 1A). Analyzing the ratio between linear (cleaved) and circular (ligated) products (Figure 6—figure supplement 1) indicated that the proportion of circle was initially very high (approx. 40% after 12 hr). Based on this, it is likely that all or most µHHz molecules are transiently circular at some point immediately after cleavage, but become progressively trapped in a state unable to re-ligate, most likely due to hydrolysis of the >p or misfolding. While these steps only validate half of a full replication cycle (formation of a (+)-strand scRNA from a (−)-strand scRNA template), these results outline the potential for a full viroid-like rolling circle RNA replication cycle based on RNA-catalysis alone.
Discussion
Viroids are transcriptional parasites composed entirely of a circular RNA genome and are considered the simplest infectious pathogens known in nature. They lack protein-coding regions in their genome and can be completely replicated in ribosome-free conditions (Daròs et al., 1994; Diener, 2003; Fadda et al., 2003; Flores et al., 2009). They can comprise a circular RNA genome of as little as ~300 nt in, for example, Avsunviroidae encoding a Hammerhead ribozyme responsible for maturation by resecting the replicated RNA genome (Flores et al., 2000). The resected viroid genome is then ligated (circularized) by a host protein (e.g., tRNA ligase) (Nohales et al., 2012). Due to the simplicity of this replication strategy, viroids have been suggested to represent possible ‘relics’ from a primordial RNA-based biology (Diener, 1989; Flores et al., 2014). Indeed, circular RNA genomes would present a number of potential advantages for prebiotic RNA replication, including increased stability by end protection (Litke and Jaffrey, 2019) and a reduced requirement for specific primer oligonucleotides to sustain replication (Attwater et al., 2018; Szostak, 2012). A circular RNA genome also resolves the end replication problem, that is, the potential loss of genetic information from incomplete replication in linear genomes. Circular RNA structures can self-assemble from RNA mononucleotides through wet-dry cycling (Hassenkam et al., 2020) providing potential initiation points for primordial RNA replication. Thus, viroid-like replication systems are plausible candidates to have emerged as the simplest genetic systems.
Here, we have explored to what extent such a potentially prebiotic replication strategy can be carried out by RNA alone. Our data show that RNA can indeed perform RCS on scRNA templates yielding concatemeric RNA products, which can be processed (i.e., resected) and re-circularized by an encoded ribozyme in a scheme reminiscent of viroid replication. Thus, one-half of a viroid replication cycle ((−)-strand replication leading to a self-circularizing (+)-strand) can be carried out by just two ribozymes. Completing a full viroid replication cycle would then require the reverse (+)-strand replication leading to a self-circularizing (−)-strand product. This may require a second ribozyme (e.g., a second µHHz) encoded in the (−)-strand akin to the mechanism used by natural viroids (Flores et al., 2000).
MD simulations indicate that the RCS process may be aided by accumulating strain in the nascent dsRNA segment leading to increased displacement (peeling off) of dsRNA 5′- and 3′-ends (i.e., strand displacement). In turn, this peeling off creates a more dynamic environment potentially aiding 5′-end invasion by extending the 3′-end. This topological strain-induced strand displacement may be general and independent of the precise RCS mechanism on scRNA templates and thus should also apply to non-enzymatic polymerization of RNA. Our observation that RCS can be enhanced by the use of branched extension, FT thermocycling, and pre-freezing dilution may also be related to this. Indeed, while the precise mechanistic basis for these enhancements is currently unknown, it seems plausible that all of these enhance 5′-end displacement on the product strand by accelerating conformational equilibration through RNA un- and refolding as observed previously (Mutschler et al., 2015).
In biology, both viroids and Hepatitis D virus (HDV) replication proceeds through RCS on circular RNA genomes mediated by proteinaceous RNA polymerases, but RCS has also been reported for circular DNA templates and proteinaceous DNA polymerases in nature (Wawrzyniak et al., 2017) and in biotechnology (Daubendiek et al., 1995; Givskov et al., 2016; Kristoffersen, 2017; Mohsen and Kool, 2016). dsDNA persistence length is somewhat shorter than dsRNA (dsDNA: 45–50 nm [140–50 nt] vs. dsRNA 60 nm [200 nt]) and stacking interactions in dsDNA are weaker than in dsRNA (Kebbekus et al., 1995; Svozil et al., 2010). Therefore, dsDNA may more readily adopt a circular shape or allow internal kinks to alleviate build-up of strain or to adopt strong bends (Wolters and Wittig, 1989). Nevertheless, we would expect a similar strand displacement effect would play a part in RCS on small circular DNA templates. Indeed, in both cases, RCS proceeds efficiently for circular genomes ranging from a few hundred nt to over 1.5 kb (Mohsen and Kool, 2016). In contrast, RNA-catalyzed polymerization (current maximum approx. 200 nt products; Attwater et al., 2013) is currently limited to RCS on scRNAs. A more efficient RNA-catalyzed RCS-based replication strategy will likely require improvements to the ribozyme polymerase catalytic activity, speed, and processivity as well as the design of the template. Improvements to ribozyme polymerase processivity, which is known to be poor (Johnston et al., 2001; Lawrence and Bartel, 2003), might have the greatest impact and might be realized either through tethering or other topological linkages to the circular template (Cojocaru and Unrau, 2021). A more processive polymerase ribozyme should also result in less non-templated triplet terminal transferase activity, which appears to be a consequence of slow RCS extension and is likely aggravated by dissociation of the 3′-end. Thus, more efficient RCS may also require the stabilization of the 3′-end triplet junction in the ribozyme active site in the same way as primer/nascent strand termini are stabilized within the active sites of proteinaceous RNA- and DNA polymerase (Chim et al., 2018; Houlihan et al., 2020). Finally, introduction of secondary structure motifs in the RNA nascent strand might drive increased 5′-end dissociation (e.g., through formation of stable hairpin structures) relieving strain at the 3′-end.
Larger circular RNA templates might provide advantages for the RCS as they are less strained. They would also provide increased access to the internal face of the circle and might be able to encode the whole ribozyme itself. On the other hand, reduced torsional strain on the dsRNA would be expected to reduce strand invasion and ‘peeling off’ of the product strand. All of these factors merit detailed investigation.
Our motivation for investigating RNA-catalyzed RCS was as a potential solution toward the so-called ‘strand separation problem,’ the inhibition of RNA replication by exceedingly stable RNA duplex products. This form of product inhibition has both a thermodynamic component, that is, the high amount of energy required to dissociate RNA duplexes, and a kinetic component as RNA replication must outpace duplex reannealing, which is rapid unless duplex concentrations are low or reactions take place in a highly viscous medium (He et al., 2017; Tupper and Higgs, 2021). In this context, we reasoned that RCS might provide favourable properties: synthesis and strand displacement on a circular template can in principle proceed essentially iso-energetically as base-pairing (H-bonding/stacking) interactions broken at the 5′-end during strand displacement are continuously compensated for by new base-pairing interactions formed at the nascent strand 3′-end. In turn, this enables an open-ended formation of template-coupled stochiometric excess of single-stranded RNA product strand to encode functions to further aid replication as we show here with resection and recircularization by an encoded ribozyme.
In the course of this work, we discovered another property of RNA synthesis on circular RNA templates that might contribute toward overcoming the strand separation problem. MD simulations indicate that—at least—on scRNA templates—the build-up of strain in the nascent dsRNA segment could aid strand displacement (Figure 3). However, the MD simulations also suggest that strain is non-directional destabilizing both nascent strand 5′- as well as 3′-ends equally. Thus, the potential advantages of scRNA RCS seem to be tempered by opposing effects such as strain as well as reduced template accessibility due to circular RNA ring geometry (Figure 1). Nevertheless, we find that a viroid-like replication strategy can be accomplished by RNA catalysis alone, with one ribozyme performing RCS on circular RNA templates yielding concatemeric RNA products, which can be processed (i.e., resected) and recircularized by a second ribozyme. Future improvements in polymerase ribozyme activity and processivity may allow all necessary components of such a replication cycle to be encoded on a circular RNA ‘genome’ and propagated by self-replication and -processing reactions.
Materials and methods
Oligonucleotides
Base sequences of all oligonucleotides used throughout this work can be found in Supplementary file 1.
In vitro transcription dsDNA templates (containing T7 promotor sequence at the 5′-end upstream of the region to transcribe) for in vitro transcription was generated by ‘fill-in’ using three cycles of mutual extension using GoTaq HotStart (Promega, Madison, WI) between the relevant oligonucleotide and primers: 5T7 or HDVrt (the latter for defined 3′ terminus formation Schürer et al., 2002). The T7 transcription protocol used is based on Megascript. Briefly explained, transcriptions of RNA requiring a triphosphate at the 5′-end (termed GTP Transcription) reaction were carried out under the following conditions: 40 mM Tris-Cl pH 8, 10 mM DTT, 2 mM spermidine, 20 mM MgCl2, 7.5 mM each NTP (Thermo Fisher Scientific), dsDNA templates (varying amount, preferably >5 μM), 0.01 units/μl of inorganic pyrophosphatase (Thermo Fisher Scientific, Waltham, MA), and ~50 μg/μl of T7 RNA polymerase (homemade by Isaac Gallego). Reactions were left overnight (~16 hr) at 37°C. Then 0.5×volume EDTA (0.5 M) was added together with (at least) 2.5×volume of loading buffer (final conditions >50% formamide or >8 M Urea and 5 mM EDTA). For transcription of RNA with a monophosphate 5′-end (termed GMP Transcription), the same procedure is followed as for NTP Transcription, however, 10 mM GMP and 2.5 mM of each NTP instead of the higher amount of NTP used for GTP transcription.
Gel electroporation for analysis or purification
Request a detailed protocolThe sample in the appropriate loading buffer was separated on 10%–20% 8 M Urea denaturing PAGE gel using an EV400 DNA Sequencing Unit (Cambridge Electrophoresis). The product band was visualized by UV shadowing (for non-labeled RNA) or fluorescence scanning (Typhoon scanner, Amersham Typhoon) (for labeled RNA). When needed the identified product based on relative migration was excised. The excised gel fragment was then thoroughly crushed using a pipette tip and suspended in 10 mM Tris-Cl pH 7.4 to form a slurry. For freeze and squeeze extraction, the slurry was frozen in dry ice, then heated to 50°C (~5 min), and finally left rotating at room temperature (from 2 hr to overnight) to elute the product from the gel material. The eluate was then filtered using a Spin-X column (0.22 μm pore size, Costar), ethanol precipitated, (100 mM Acidic acid and 80% ethanol (10 μg glycogen carrier was present when noted)). UV absorbance was measured with a Nanodrop ND-1000 spectrophotometer (Thermo Fisher Scientific) to determine yield of redissolved purified RNA.
Calculation of extension efficiency
Request a detailed protocolGel images from the Typhoon scanner were analyzed in ImageQuant software (Cytiva Life Science) for quantifying band intensity. Quantified band intensities were exported to Excel (Microsoft, Redmond, WA) for further analysis. Extension efficiency (E) for a given band (b) was calculated as the sum of the intensities (I) of all the bands from b to n (n being the highest detectable band), divided by the sum of I of all bands from b−1 to n:
Thus, E represents the efficiency of the given ligation junction (Lb) to allow the production of the extension product in band b, that is, the extension efficiency.
Triplet transcription
Request a detailed protocolTriplets were prepared via run-off in vitro transcription with T7 RNA polymerase. More details on the method can be found in Attwater et al., 2018. Reaction conditions were as follows: 100 pmols of DNA template for each triplet was mixed with equimolar DNA oligo 5T7 to form the template for transcription. For triplets starting with purines, the NTP transcription protocol was used as described above with a total NTP concentration of 30 mM but only adding the nucleotides necessary for the triplet (e.g., AUA was transcribed with only ATP and UTP). For triplets starting in pyrimidines, a lower total NTP concentration was used (4.32 mM) as this yielded better defined bands for purification. About 50 μl transcription reactions were stopped with 2 μl EDTA (0.5 M) and 5 μl of 100% glycerol was added to facilitate gel loading. The samples were separated by 30% 3 M Urea denaturing PAGE as described above. Correct sequence composition was confirmed by A260/280 absorbance ratio, measured with the Nanodrop.
Circularization of RNA
Request a detailed protocolLinear 5’-end monophosphate labelled RNA to be used for circularization was either prepared by in vitro transcription (300 μl reaction volume) or ordered directly as chemically synthesized RNA (Integrated DNA Technologies [IDT], IA). Linear RNA was gel purified as described above. When needed purified RNA was treated with T4 polynucleotide kinase (PNK) (New England Biolabs (NEB), Ipswich, MA) to remove 3′-end cyclic phosphate then RNA was phenol/chloroform extracted, ethanol precipitated, and redissolved in ddH2O. For splinted ligation, 3 pmol purified RNA was mixed with equimolar splint RNA in 262.5 μl ddH2O. The sample was heated to 80°C (2 min) followed by cooling to 17°C (10 min) and finally incubated on ice (5–30 min). Then reaction conditions were adjusted to 50 mM Tris-HCl pH 7.5, 2 mM MgCl2, 1 mM DTT, and 400 μM ATP (1× T4 RNA ligase 2 reaction buffer [NEB]) including 0.25 units/μl T4 RNA ligase 2 (Neb) (final volume 300 μl) and samples left overnight (~16 hr) at 4°C. For non-splinted ligation, 0 pmol gel purified RNA was mixed in 237 μl ddH2O followed by heating to 95°C and then quickly moved to ice. Then reaction conditions were adjusted to 50 mM Tris-HCl, pH 7.5, 10 mM MgCl2, 1 mM DTT (1× T4 RNA ligase reaction buffer [Neb]), 100 μM ATP including one unit/μl T4 RNA ligase 1 (NEB) (final volume 300 μl) and samples left overnight (~16 hr) at 16°C. Circularized RNA was electrophorated by 10% 8 M Urea denaturing PAGE for analysis and purification as described above.
Templated RNA-catalyzed RNA synthesis (the primer extension assay)
Request a detailed protocolRibozyme activity assay was performed essentially as described in Attwater et al., 2018. In a standard reaction (modified where specified), ribozyme heterodimer (5 TU/t1), template, primer (5 pmol of each), and triplets (50 pmol of each) were annealed in 7.5 μl water (80°C 2 min, 17°C 10 min). Then 2.5 μl 4× reaction buffer was added (final volume 10 μl) and samples were left on ice for ~5 min to ensure folding. Final pre-frozen conditions were (unless otherwise noted) either (Tris buffer system) 50 mM Tris (pH 8.3 at 25°C), 100 mM MgCl2, and 0.01% Tween 20, or (CHES buffer system) 50 mM CHES (pH 9 at 25°C), 150 mM KCl, 10 mM MgCl2, and 0.01% Tween 20. At this point some samples (noted in the text) were diluted by adding ddH2O (e.g., 50× dilution corresponds to adding 490 μl ddH2O to the 10 μl samples). Finally, samples were frozen on dry ice and incubated at –7°C in an R4 series TC120 refrigerated cooling bath (Grant, Shepreth, UK) to allow eutectic phase formation and reaction, respectively. To end the incubations, samples that had been diluted were thawed, moved to 2 ml tubes, ethanol precipitated (with glycogen carrier) and redissolved in 10 μl ddH2O. This step was avoided for undiluted samples that were already 10 μl. Finally, 0.5 μl EDTA (0.5 M) was added to all samples to a final volume of 15 μl. (In experiments where the effect of dilution was investigated, e.g., as experiment presented in Supporting Figure 5, ddH2O was added to all the thawed samples to reach the same volume before precipitation). To prepare for separation of extension products, 3 μl of the reacted samples after the addition of EDTA (corresponding to 1 pmol template RNA) was diluted to reach the final loading conditions: 166 mM EDTA, 6 M Urea (+Bromophenol blue), and 10–20 pmol competing RNA (to prevent long product/template reannealing) (final volume 10 μl). Finally, samples were denatured (95°C for 5 min) and RNA separated by 8 M Urea denaturing PAGE.
Sequencing of extension products
In the primer extension reactions used for sequencing, the primer extension was performed as described above except for the following changes: 5 pmol ribozyme heterodimer/template, 20 pmol primer (with a 5′ adapter sequence), and 100 pmol of each triplet was used. In the cases where multiple templates were mixed in the same reaction (one-pot), the final template concentration remained 5 pmol in total. All reactions were done in the CHES buffer and were diluted 50× as standard.
Adapter ligation and RT-PCR
View detailed protocolAfter Urea PAGE separation of the extension products, the noted region of the gel was dissected out, and carefully recovered as described above. The RNA was ethanol precipitated (80% ethanol with 10 µg glycogen carrier) resulting in a dry RNA pellet. To append an adaptor sequence to the 3′-end of the purified RNA products the dry RNA was redissolved in conditions allowing adenylated adapter ligation by T4 RNA ligase 2 truncated K227Q (Neb) following manufacturers’ descriptions. Final adapter ligation conditions were: 50 mM Tris-HCl, pH 7.5, 10 mM MgCl2, and 1 mM DTT (1× T4 RNA ligase reaction buffer [NEB]), 15% PEG8000, 0.04% Tween 20, 5 pmol adenylated DNA primer (Adap1, for base sequences see Supplementary file 1), and 20 U/µl T4 RNA ligase 2 truncated K227Q (Neb) (final volume 10 µl). The samples were then ligated at 16°C for 2 hr. Pre-adenylation of Adap1 using Mth RNA Ligase (Neb) was performed following manufacturers’ descriptions. After adapter ligation, samples were diluted tenfold to achieve conditions for performing RT-PCR (25 cycles) using 0.5 µM forward (PCRp3) and reverse primer (RTp1) and the SuperScript III One-Step RT-PCR System with Platinum Taq DNA polymerase (Thermo Fisher Scientific). Finally, RT-PCR products were gel purified in 3% agarose gel and cleaned up using QIAGEN gel extraction kit (QIAGEN, Hilden, Germany).
Sanger sequencing
Request a detailed protocolPurified RT-PCR products were cloned into pGEM vector using pGEM-T Easy Vector Systems (Promega) as described by the manufacturer and transformed into heat-competent 10-Beta cells (NEB). Inserts from single colonies were PCR amplified (using primers pGEM_T7_Fo and pGEM_SP6_Ba) and send in for Sanger sequencing (Source Bioscience) (using pGEM_T7_Fo as sequencing primer).
Illumina sequencing
Request a detailed protocolIllumina adaptors were added to purified RT-PCR products by PCR (15 cycles) using 0.5 µM forward (Illx_Fo, x denotes different barcodes 1–15, see oligo sequences in Supplementary file 1) and reverse primers (Ill_Ba) and Q5 Hot-Start High-Fidelity 2X Master Mix (Neb). PCR products were gel purified in 3% agarose gel and qPCRed (using NEBNext Library Quant Kit for Illumina) to quantify concentration. Finally, the DNA (consisting of Illumina adapters, barcodes, and RT-PCRed sequence from the RNA extension) were prepared following manufactures protocol for MiSeq Illumina sequencing (Illumina, San Diego, CA) (see, e.g., MiSeq System Guide).
Sequencing data analysis
Illumina Sequencing data were acquired and processed as FASTQ files using Terminal (and available software packages such as FASTX-toolkit). Prior to analysis, the whole output file from Illumina sequencing runs (containing also unrelated sequences) was split based on barcodes identifying the individual samples and trimmed starting with the original (P91) primer sequence (GAAGAACTG). After the P91 sequence, the triplets at positions 1, 2, 3, and so on, would be identifiable representing extension products made by the ribozyme. The presence of the 3′ adapter sequence (GTCGAATAT…) in the aligned sequences marked the end of the original RNA extension product. Sequencing data can be found as described below under section data availability: File 1 includes sequence data for circular and linear one-pot analysis (C1 and L1, respectively), File 2 includes sequence data for branched RCS analysis (B3).
Analysis of the one-pot experiments
Request a detailed protocolBy counting the number of times a given triplet was present at a given position, we were able to calculate the fidelity for each triplet at this position. Identifying and counting the sequencing reads (n) for each position was done using grep (in Terminal) with a list of all relevant sequences (positions 3–18) and the sequencing files. The triplet at position 3, the first barcode position, was used to classify the sequences into coming from templates A to D and thus has 100% fidelity for the correct triplet (Figure 4B).
For example, for analyzing the fidelity (F) of position 4, the following list was used: GAAGAACTG(primer)GAA(pos1)GAA(pos2)YYY(pos3)XXX(pos4). Here, YYY was either of the first barcode triplets for templates A–D, (ATA(template A), AAA(template B), TTA(template C), or ATC(template D)) and XXX was either of the 14 possible triplets (CTG, ATA, CCA, CCC, AAA, CAC, GGG, TTA, TCC, GGC, ATC, GAT, CGC, and GAA). F at position 4 was then calculated for templates A–D as the number of occurrences of a triplet in position 4 (e.g., CCA) divided by the sum of occurrences of all the triplets multiplied by 100%. A generalized term for calculating the F at all positions (3–18) and for all templates (A–D) is:
Here, F is the fidelity, a is the position of the triplet, Y is the templates A–D. n is the number of sequencing reads for a given triplet (xxx) for position a on template Y or for all the 14 triplets (XXX), for a on Y. Eventually, the fidelity for positions 3–15 in the context of templates A–D for all triplets was plotted in Figure 4B. Accumulated chance for a product of reaching position X (shown in plot in Figure 4C) was calculated by multiplying all fidelities for moving from position three to position X with correct triplets (fidelities found in Figure 4C). Data for this analysis can be found as described in the Data availability section below (File 1). Numerical data and calculation are supplied in Figure 4—source data 1.
Analysis of the branched RCS
Request a detailed protocolBy counting the number (n) of correct sequences with a specific length ending in the 3′ adapter sequence, we identified long RCS products (Figure 5D). This was done using grep (in Terminal) with a list of all relevant sequences (positions 9–30, both products I and II), and the sequencing file. Data for this analysis can be found as described in the Data availability section below (File 2). Numerical data and calculation are supplied in Figure 5—source data 1.
Self-circularizing micro hammerhead ribozyme assay
Request a detailed protocolRNA-catalyzed synthesis of fluorophore labeled self-circularizing µHHz was prepared in 2× large (500 pmol) reactions set up and incubated as described above. Specifically, 500 pmol ribozyme heterodimer (5 TU/t1) and circular template (scHHz_temp), 2000 pmol primer (HHrzP12) and 50 µmol of each of the triplets were annealed followed by adding buffer to 50 mM CHES, pH 9, 150 mM KCl, 10 mM MgCl2, and 0.05% Tween 20 (1 ml). Then the sample was diluted 50 times to a final volume of 50 ml. After 4 weeks incubation at –7°C, EDTA was added (5 mM final concentration), reactions were thawed and concentrated to a final volume of ~300 µl using a centrifugation filter (Amicon Ultra, 3 kDa cutoff) retaining long RNA products. µHHz RNA (marked in Figure 6C) was purified by gel electrophoresis and excised product was dissolved to 10 µM in H2O with 0.5 mM EDTA.
Chemically synthesized fluorophore labeled self-circularizing µHHz RNA (IDT) was gel purified as described above and excised product was dissolved to 10 µM in H2O with 0.5 mM EDTA.
Micro-hammerhead ribozyme cleavage/circularization assay
Request a detailed protocolSelf-circularization assays of chemically synthesized fluorophore labeled µHHz comprise 10 pmol µHHz annealed (80°C 2 min, 17°C 10 min) in 4 µl water with 1 µl 5× reaction buffer, final reaction conditions: 50 mM CHES, pH 9, 150 mM KCl, and 10 mM MgCl2 (same as for the templated RNA-catalyzed RNA synthesis). Then incubated in ice for 5 min to ensure folding. This was then frozen on dry ice and either moved to –7°C for eutectic phase formation (reaction) or –80°C (control). After incubation, 10 µl loading buffer (95% formamide, 25 mM EDTA, and bromophenol blue) was added directly to the cold samples to stop the reaction and mixed while thawing. Finally, reactions were analyzed by 20% denaturing PAGE like described above. 5′-phosphorylation of µHHz RNA with polynucleotide kinase (NEB) was carried out following the manufacturer’s directions. RNA was then phenol/chloroform extracted, precipitated, and dissolved in ddH2O with 0.5 mM EDTA to 10 µM (determined by Nanodrop).
Molecular dynamics simulations
Request a detailed protocolAll simulations were set up with the AMBER 18 suite of programs and performed using the CUDA implementation of AMBER’s pmemd program (Case, 2018). A linear ssRNA of 36 nt with the sequence (UUC)12 was built using the NAB utility, which was then circularised using an in-house program (Pyne et al., 2021). From there, the complementary strand containing GAA triplets was progressively grown representing the different stages of the rolling circle replication, containing 9, 18, 21, 24, 27 till 30 nt of dsRNA keeping the rest single-stranded. For each stage, a representative structure was used as a scaffold to grow the dsRNA part and thus build the structure to model next stage. A linear dsRNA fragment containing four GAA triplets with a nick between the first and second was run as a control. This molecule had a total length of 16 bp as it was capped by a CG dimer on each end.
The AMBER99 forcefield (Cheatham et al., 1999) with different corrections for backbone dihedral angles including the parmBSC0 for α and γ (Pérez et al., 2007) and the parmOL3 for χ (glycosidic bond) (Zgarbová et al., 2011) were used to describe the RNA. All initial structures were explicitly solvated using a truncated octahedral TIP3P box with a 14-Å buffer. They were neutralized by two different types of salt, KCl, and MgCl2, described by the ‘scaled charged’ Empirical Continuum Correction (ECC) set of ion parameters (Duboué-Dijon et al., 2018), which substantially reduce the overestimation of ion-ion interactions with respect to water-mediated interactions typical of empirical forcefield calculations (Fingerhut et al., 2021; Kirby and Jungwirth, 2019) (see Figure 3—figure supplement 6). The necessary ion pairs (Machado and Pantano, 2020) were added for matching 0.2 M in the case of KCl, and 0.1 and 0.5 M in the case of MgCl2. Simulations were performed at constant T and P (300K and 1 atm) following standard protocols (Noy and Golestanian, 2010) for 400 ns.
The last 100 ns sampled every 10 ps were used for the subsequent analysis. AMBER program CPPTRAJ (Roe and Cheatham, 2013) was used to determine base-pair step parameters, radial distribution functions of ions around RNA and distances between atoms, including groove width and hydrogen bonds. The latter were defined with a distance cutoff of 3.5 Å and an angle cutoff of 120°. Counterion-density maps were obtained using Canion (Lavery et al., 2014) and were subsequently visualized with Chimera (Pettersen et al., 2004). SerraNA software was used to calculate curvatures at different sub-fragment lengths (Velasco-Berrelleza et al., 2020).
Data availability
All data generated or analyzed in this manuscript is supplied within the manuscript or supporting file; Source Data files containing original unedited gels images as well as numeric data have been provided for Figures 1,2,4 and 5, as well as figure supplements when relevant. Modelling data and sequencing data are provided as described in the data availability section in the manuscript.
-
Dryad Digital RepositoryDeep Sequencing data for document titled: Rolling Circle RNA Synthesis Catalysed by RNA.https://doi.org/10.5061/dryad.tht76hf10
References
-
Single-molecule measurements of the persistence length of double-stranded RNABiophysical Journal 88:2737–2744.https://doi.org/10.1529/biophysj.104.052811
-
Ice as a protocellular medium for RNA replicationNature Communications 1:76.https://doi.org/10.1038/ncomms1076
-
In-ice evolution of RNA polymerase ribozyme activityNature Chemistry 5:1011–1018.https://doi.org/10.1038/nchem.1781
-
Unified prebiotically plausible synthesis of pyrimidine and purine RNA ribonucleotidesScience (New York, N.Y.) 366:76–82.https://doi.org/10.1126/science.aax2747
-
Direct labelling of BAC-DNA by rolling-circle amplificationThe Plant Journal 45:857–862.https://doi.org/10.1111/j.1365-313X.2005.02637.x
-
Highly Efficient DNA Synthesis by the Phage ϕ 29 DNA PolymeraseJournal of Biological Chemistry 264:8935–8940.https://doi.org/10.1016/S0021-9258(18)81883-X
-
A modified version of the Cornell et al. force field with improved sugar pucker phases and helical repeatJournal of Biomolecular Structure & Dynamics 16:845–862.https://doi.org/10.1080/07391102.1999.10508297
-
Processive RNA polymerization and promoter recognition in an RNA WorldScience (New York, N.Y.) 371:1225–1232.https://doi.org/10.1126/science.abd9191
-
Rolling-Circle RNA Synthesis: Circular Oligonucleotides as Efficient Substrates for T7 RNA PolymeraseJournal of the American Chemical Society 117:7818–7819.https://doi.org/10.1021/ja00134a032
-
Discovering viroids--a personal perspectiveNature Reviews. Microbiology 1:75–80.https://doi.org/10.1038/nrmicro736
-
Avsunviroidae family: viroids containing hammerhead ribozymesAdvances in Virus Research 55:271–323.https://doi.org/10.1016/s0065-3527(00)55006-4
-
Viroids: survivors from the RNA world?Annual Review of Microbiology 68:395–414.https://doi.org/10.1146/annurev-micro-091313-103416
-
Cofactors are Remnants of Life’s Origin and Early EvolutionJournal of Molecular Evolution 89:127–133.https://doi.org/10.1007/s00239-020-09988-4
-
Engineering of hairpin ribozyme variants for RNA recombination and splicingAnnals of the New York Academy of Sciences 1447:135–143.https://doi.org/10.1111/nyas.14052
-
RNA-catalyzed RNA polymerization: accurate and general RNA-templated primer extensionScience (New York, N.Y.) 292:1319–1325.https://doi.org/10.1126/science.1060786
-
A Model for the Emergence of RNA from a Prebiotically Plausible Mixture of Ribonucleotides, Arabinonucleotides, and 2’-DeoxynucleotidesJournal of the American Chemical Society 142:2317–2326.https://doi.org/10.1021/jacs.9b11239
-
Charge Scaling Manifesto: A Way of Reconciling the Inherently Macroscopic and Microscopic Natures of Molecular SimulationsThe Journal of Physical Chemistry Letters 10:7531–7536.https://doi.org/10.1021/acs.jpclett.9b02652
-
The contribution of metal ions to the structural stability of the large ribosomal subunitRNA (New York, N.Y.) 10:1366–1379.https://doi.org/10.1261/rna.7390804
-
The biogenesis, biology and characterization of circular RNAsNature Reviews. Genetics 20:675–691.https://doi.org/10.1038/s41576-019-0158-7
-
Rolling-circle amplification under topological constraintsNucleic Acids Research 30:574–580.https://doi.org/10.1093/nar/30.2.574
-
Circular RNAs: diversity of form and functionRNA (New York, N.Y.) 20:1829–1842.https://doi.org/10.1261/rna.047126.114
-
Analyzing ion distributions around DNANucleic Acids Research 42:8138–8149.https://doi.org/10.1093/nar/gku504
-
Processivity of ribozyme-catalyzed RNA polymerizationBiochemistry 42:8748–8755.https://doi.org/10.1021/bi034228l
-
The difficult case of an RNA-only origin of lifeEmerging Topics in Life Sciences 3:469–475.https://doi.org/10.1042/ETLS20190024
-
Split the Charge Difference in Two! A Rule of Thumb for Adding Proper Amounts of Ions in MD SimulationsJournal of Chemical Theory and Computation 16:1367–1372.https://doi.org/10.1021/acs.jctc.9b00953
-
The Discovery of Rolling Circle Amplification and Rolling Circle TranscriptionAccounts of Chemical Research 49:2540–2550.https://doi.org/10.1021/acs.accounts.6b00417
-
Complete, closed bacterial genomes from microbiomes using nanopore sequencingNature Biotechnology 38:701–707.https://doi.org/10.1038/s41587-020-0422-6
-
Freeze-thaw cycles as drivers of complex ribozyme assemblyNature Chemistry 7:502–508.https://doi.org/10.1038/nchem.2251
-
The structural basis of ribosome activity in peptide bond synthesisScience (New York, N.Y.) 289:920–930.https://doi.org/10.1126/science.289.5481.920
-
The chirality of DNA: elasticity cross-terms at base-pair level including A-tracts and the influence of ionic strengthThe Journal of Physical Chemistry. B 114:8022–8031.https://doi.org/10.1021/jp104133j
-
RNA circularization strategies in vivo and in vitroNucleic Acids Research 43:2454–2465.https://doi.org/10.1093/nar/gkv045
-
UCSF Chimera--a visualization system for exploratory research and analysisJournal of Computational Chemistry 25:1605–1612.https://doi.org/10.1002/jcc.20084
-
Lipid-assisted synthesis of RNA-like polymers from mononucleotidesOrigins of Life and Evolution of the Biosphere 38:57–74.https://doi.org/10.1007/s11084-007-9113-2
-
PTRAJ and CPPTRAJ: Software for Processing and Analysis of Molecular Dynamics Trajectory DataJournal of Chemical Theory and Computation 9:3084–3095.https://doi.org/10.1021/ct400341p
-
Thermal Habitat for RNA Amplification and AccumulationPhysical Review Letters 125:048104.https://doi.org/10.1103/PhysRevLett.125.048104
-
A universal method to produce in vitro transcripts with homogeneous 3’ endsNucleic Acids Research 30:e56.https://doi.org/10.1093/nar/gnf055
-
Crystal structure of the catalytic core of an RNA-polymerase ribozymeScience (New York, N.Y.) 326:1271–1275.https://doi.org/10.1126/science.1174676
-
Viruses with Circular Single-Stranded DNA Genomes Are Everywhere!Annual Review of Virology 4:159–180.https://doi.org/10.1146/annurev-virology-101416-041953
-
Enzyme-free ligation of dimers and trimers to RNA primersNucleic Acids Research 47:3836–3845.https://doi.org/10.1093/nar/gkz160
-
The eightfold path to non-enzymatic RNA replicationJournal of Systems Chemistry 3:2208.https://doi.org/10.1186/1759-2208-3-2
-
Rolling-circle and strand-displacement mechanisms for non-enzymatic RNA replication at the time of the origin of lifeJournal of Theoretical Biology 527:110822.https://doi.org/10.1016/j.jtbi.2021.110822
-
SerraNA: a program to determine nucleic acids elasticity from simulation dataPhysical Chemistry Chemical Physics 22:19254–19266.https://doi.org/10.1039/d0cp02713h
-
RNA Splicing by the SpliceosomeAnnual Review of Biochemistry 89:359–388.https://doi.org/10.1146/annurev-biochem-091719-064225
-
Construction of a 42 base pair double stranded DNA microcircleNucleic Acids Research 17:5163–5172.https://doi.org/10.1093/nar/17.13.5163
-
Refinement of the Cornell et al. Nucleic Acids Force Field Based on Reference Quantum Chemical Calculations of Glycosidic Torsion ProfilesJournal of Chemical Theory and Computation 7:2886–2902.https://doi.org/10.1021/ct200162x
-
Potentially Prebiotic Activation Chemistry Compatible with Nonenzymatic RNA CopyingJournal of the American Chemical Society 142:14810–14813.https://doi.org/10.1021/jacs.0c05300
-
Template-Directed Copying of RNA by Non-enzymatic LigationAngewandte Chemie (International Ed. in English) 59:15682–15687.https://doi.org/10.1002/anie.202004934
Decision letter
-
Timothy W NilsenReviewing Editor; Case Western Reserve University, United States
-
James L ManleySenior Editor; Columbia University, United States
-
Jiri SponerReviewer
In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.
Decision letter after peer review:
Thank you for submitting your article "Rolling Circle RNA Synthesis Catalysed by RNA" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, and the evaluation has been overseen by Timothy Nilsen as Reviewing Editor and James Manley as the Senior Editor. The following individual involved in review of your submission has agreed to reveal their identity: Jiri Sponer (Reviewer #3).
The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission.
Essential revisions:
Reviewer #1:
In the origin of life scenario where RNA is assumed to be the first replicator, a key problem is how RNA can replicate itself. Or how can RNA polymerase copy itself, since copying requires an open flexible structure that can be read. While the polymerase needs to be a topological rigid structure in order to catalyse the RNA polymer.
The manuscript describes how a special kind of synthesis of the RNA, rolling circle synthesis on small pieces of circular RNA, could template and build RNA strands. The method apparently could help in avoiding the strand inhibition problem where stable RNA duplex in long strands hinders replication.
It is a bit unclear how, but I suspect this i down to the size of the ring, since the problem increase with the length of the polymer. The authors also touch on this themselves in their MD simulations, since the entropically unfavourable confinement of a string onto a circle can alleviate part of the problem. Nevertheless, the authors also show by MD that the rings become small tight structures that actually hinders replication.
The study depends on trinucleotide triphosphates (triplets) as substrates. They demonstrate a viroid like replication that show how a template (-) is able to make a mirror copy (+) of circular RNA by a rolling circle synthesis with out the need for enzymes or other catalyst apart from RNA itself.
While the conclusion of the work becomes a bit muddled, but honest, the work is a very important piece that demonstrates the huge potential role of circular RNA in the very early stages of life.
I must confess I am very far off my field of competences. I have tried my best to understand the paper, and the methods involved, but obviously I cannot give much feedback on the methods used. So, I cannot suggest improvements on that part.
For the most part, the paper is well written, and the structure is sound. I would strongly recommend publication.
It was unclear to me how it was certain that it was actually rings. There is the MD simulations, but apart from the principle drawing they are obviously not circular in shape.
There are some parts of the manuscript that too me is very difficult to understand. Some statements seem to reflect a community understanding that might be obvious for those working with similar system on a day to day basis, but for the general reader (where I put myself) this becomes gibberish. for instance:
RCS has potentially unique properties with regards to the strand inhibition problem where RNA duplex melting in principle can be effected by continuous toehold strand displacement driven by nucleotide hybridization and the ratchet of nascent strand extension by triphosphate hydrolysis. In an idealized RCS mechanism, such strand invasion and displacement processes are both isoenergetic and coordinated to nascent strand extension (Blanco et al., 1989; Daubendiek et al., 1995), with rotation of the single-stranded RNA (ssRNA) preventing the build-up of topological tension (Kuhn et al., 2002). Thus, RCS is a potentially open-ended process leading to the synthesis of single-stranded multiple repeat products (concatemers) with an internally energized strand displacement circumventing the "strand inhibition problem" (Tupper and Higgs, 2021).
It might not be possible to rephrase this so that everyone can understand. But I think the clarity of manuscript could be improved.
But also because there is references to previous work where long statements does not make it more clear what is actually meant. for instance:
Similar to what was described previously, RNA synthesis by the TPR best in the eutectic phase of water ice, due to beneficial reaction conditions for ribozyme catalysis such as reduced RNA hydrolysis and high ionic and RNA substrate concentrations (Attwater et al., 2010). This was also the case on scRNA templates.
That said, I think the overall message is clear and the paper is very interesting, I expect to reference it when it comes out.
Reviewer #2:
The RNA World theory is one of the most widely-believed explanations for the origin of life. This relies on the idea that there were self-replicating RNA systems in the early stages of life. Usually ,it is supposed that there were polymerase ribozymes that were able to use another RNA strand as a template for synthesis of the complementary strand. As there are no naturally-occurring polymerase ribozymes, there has been a sustained effort over several decades to develop polymerase ribozmes in the lab by in vitro selection. This paper contributes to this by presenting a polymerase ribozyme that can copy a circular template. Circular templates are thought to be important because replication of a circular template can occur via the rolling circle mechanism, in which a polymerase continues multiple times around the same circle, and the far end of the growing strand is displaced from the template at the same time as new bases are added to the growing end. This avoids the problem of strand inhibition (i.e the difficulty of separation of stable double strands that are expected to form when copying linear templates).
This paper considers rolling circle replication on very short circles of around 36 nucleotides. It is shown that replication proceeds by addition of triplets beyond the full length of the circle. As the circle is short, and the double-stranded part is stiff, it is not possible for the whole of the circular template to be double-stranded at the same time. It is shown that roughly half of the circle is double-stranded, and that the separation of the two strands occurs at a point which is on the opposite side of the circle from the point of primer extension.
The rolling circle mechanism involves cleavage of the growing strand by a self-cleaving hammerhead ribozyme that is encoded in its sequence. The mechanism also requires the reconnection of the ends of the new strand in order to form a new circular template. Both the cleavage and re-circularization steps are demonstrated in this paper.
This experiment still falls short of a fully self-replicating ribozyme system, because in order for continued replication to occur, both plus and minus strands of the circle would have to encode a hammerhead ribozyme, and in order for the system to be self-sustaining, the circles would also have to encode the polymerase ribozyme itself (which is supplied separately in this paper and is not replicated). Nevertheless, this paper makes an important step, and continues to bring us closer to developing self-replicating RNA systems.
Lines 122-126 – It is implied that triplets are better than monomers for rolling circle replication because triplets help to open up other double stranded regions. However, it is not obvious that this should be the case. To put a new triplet down you have to displace three bases from the displaced strand, whereas to put a monomer down you only have to displace one base. It is not easy to predict which of these is faster without measuring it. Furthermore, in the actual mechanism occurring here, there is no prior strand to be displaced at the point of attachment, because the displacement is occurring at the other side of the circle and it does not directly interfere with the attachment. So it is not clear whether this argument applies. Has replication of a circular strand actually been attempted with a monomer ribozyme? Is it known whether a triplet ribozyme is better than a monomer ribozyme on circular templates? If not, it would be better to avoid implying this.
Figure 1 – the periodic effect seen in 1E is claimed to be due to the difference of accessibility of template bases on the inside and outside of the circle. However, the results are measured by averaging over many different circular templates. I would expect that different copies of a circular template would have different configurations and would not always have the same bases on the inside. So the inside-outside difference should average out. Could the variability of 1E be explained by variation in rates of addition according to the sequence of the template rather than an inside-outside effect? The sequence effect would be the same in multiple copies of the same template sequence. Is there a similar variability seen when copying linear templates?
Figure 1C shows a TPR dimer. Is the polymerase actually in two parts? Is this important?
Line 213 – It is not clear why the 9 bp primer goes straight to 18 bp. What happened to the lengths in between?
Line 220 – The word "extended" is used to mean that the unhybridized portion is stretched. There is a possible confusion with extending a strand by ligation of a triplet. Maybe the use of a word like stretched is better?
Line 226 – The simulations show that the double-stranded part of the circle is stiff and only covers roughly half of the circle. The point at which the primer extension occurs is therefore far away from the point at which the two strands separate. This is important for very short circles. For longer circles, the stiffness should be less relevant, and there will come a point where the whole circle becomes double stranded. There will then need to be a true strand displacement occurring very close to the point of primer extension. How long would we need the circle to be before it switches to a double-stranded circle? Does the stiffness effect seen here with the short circles make the primer extension reaction easier or more difficult than a true strand displacement reaction on a double stranded circle?
Lines 228-33 – This paragraph is not very clear. The meaning is not coming through.
Line 288 – "orbit" is an odd word. Is there a better one?
Figure 4 – Overall Figure 4 is not clear.
– I have not understood the notation n: 3E5, 2E5 etc.
– For sequence A Pos 4, GAA is the darkest shade, so I am presuming the template is CUU (in the reverse direction). But the second darkest shade is GGC. Why should GGC bind to CUU more strongly than others (for example GGA)? I am not sure whether I have understood this diagram correctly.
– Part C shows fold difference. It would be easier if rates where shown for linear and circular strands separately. Why is sequence D a worse template? Or maybe it is not worse – it's just that the ratio of circular to linear templates is lower? It is not easy to understand this.
– Part D Figure 2 seems to show a double-stranded triplet being added. Why not just a single-stranded triplet.
Figure 6D – It is unclear what is happening at each step. Particularly the backwards and forwards diagrams in step 4. Also, shouldn't the red strand be still attached to the blue circle before the cleavage occurs? The chemical structure in the middle is a bit distracting. I think the structure drawn in A is the same as D step 6. Maybe put parts A and D together and make B and C a separate figure?
Line 436 – The reference to the virtual circular genome is misleading at this point. In the proposal of Zhou et al., there are no real circles, there are simply linear fragments that can be aligned to form a virtual circle. This does not fit with the rest of this paragraph. Either the reference to Zhou et al. should be omitted or it should be explained properly what the virtual circle proposal is.
Reviewer #3:
Technically the experiments are sound, really comprehensive, convincing and the paper is well written. The documentation (the composed Figures etc.) is very nice. The MD simulations nicely complement the experiments. Strong point is that the simulations address a qualitative question and are clearly directed to solve it. It is a preferable application of the MD technique. The basic methodology is correct, the standard AMBER OL3 force field appears appropriate, as the first choice multipurpose RNA version. It is known to lead to over-compacted unstructured ssRNA ensembles, as all biomolecular force fields that are good for folded biopolymers. For the double strand, the circle and their flexibility it should be an optimal choice. The simulations are quite short by contemporary standards, though I do not think their prolongation would change the essence of the findings.
As noted above, I consider the experiments as very convincing. Strong point is that the accompanying simulations address a qualitative question and are clearly directed to solve it. It is a preferable application of the MD technique. So, I really like it, though I have some ideas for potential minor improvements, may be explanatory comments, all for supporting information.
There are some occasional typos, e.g. l. 180 stand displacement, l. 182 this suggest. I think on l. 56. where progress in non-enzymatic synthesis is overviewed, the reference could be more balanced. Appears to me that some groups are represented by duplicate citations while some research is omitted, for example a recent progress in template-free non-enzymatic RNA polymerization of 3',5' cyclic nucleotides , https://chemistry-europe.onlinelibrary.wiley.com/doi/10.1002/syst.202100017.
The short "jumping" supplementary movies are difficult to follow, I assume it is because of the size of the movies. Would it be possible to create a few SI Figures showing details of the most interesting parts of the structure, to focus on the key details, to accompany the movie?
In the simulations, lot of emphasis is paid on the Mg2+ simulations up to 500 mM MgCl. First point, is this condition relevant to the origin of life? Second, inclusion of divalents into MD is always risky, as they sample poorly (which is further exacerbated by the lack of bulk background due to the small periodic box, which may lead to glassy-like ion behavior around the solute). 400 ns is not sufficient to converge Mg2+. In addition, divalents, especially the high charge density Mg2+, are beyond the pair-additive MM approximation. It is impossible to simultaneously balance ion hydration and inner vs. outer shell binding to different coordination sites with the simple MM models. Could the authors briefly comment on initial placement of the ions after equilibration and during MD? Was it always hexacoordinated outer-shell binding to the RNA? Could the authors in SI briefly comment on it, and also comment if they had some specific reasons to choose the Duboue-Dijon parameters over parameters that have been more commonly used in biomolecular simulations? As I am not sure these specific parameters were tested/calibrated for RNA interactions (but I am not fully familiar with the work). Again, as the results are qualitative, I do not expect any effect on basic outcome of the work, so I am not suggesting any new computations.
https://doi.org/10.7554/eLife.75186.sa1Author response
Essential revisions:
Reviewer #1:
In the origin of life scenario where RNA is assumed to be the first replicator, a key problem is how RNA can replicate itself. Or how can RNA polymerase copy itself, since copying requires an open flexible structure that can be read. While the polymerase needs to be a topological rigid structure in order to catalyse the RNA polymer.
The manuscript describes how a special kind of synthesis of the RNA, rolling circle synthesis on small pieces of circular RNA, could template and build RNA strands. The method apparently could help in avoiding the strand inhibition problem where stable RNA duplex in long strands hinders replication.
It is a bit unclear how, but I suspect this i down to the size of the ring, since the problem increase with the length of the polymer. The authors also touch on this themselves in their MD simulations, since the entropically unfavourable confinement of a string onto a circle can alleviate part of the problem. Nevertheless, the authors also show by MD that the rings become small tight structures that actually hinders replication.
We welcome these comments. In the revised manuscript we have rewritten the relevant sections in order to clarify our arguments with respect to aspects of rolling circle synthesis that might aid RNA-catalyzed RNA replication.
In brief, our argument relies on the conjecture (supported by our MD simulations) that in small RNA rings (small circular RNAs, scRNAs), increasing strain upon RNA synthesis can contribute to the dissociation of double-stranded (ds) RNA into single-stranded (ss) RNA at both the 5’- or the 3’-RNA ends. Only the 3’- (but not the 5’-) RNA end is extended by the triplet polymerase ribozyme (TPR) and this process is irreversible. Therefore, over time there will be an overall shift of the dsRNA segment around the circle in the 3’- direction resulting in a 5’-ssRNA “tail” of increasing length. The efficiency of this process relies critically on the speed and processivity of the triplet polymerase ribozyme (TPR) (which is currently poor) and its ability stabilize the RNA 3’-end bound to the circRNA template and extend it before it can dissociate again.
Another potential issue is the fact that (as referee1 correctly observes), our MD simulations suggest RNA synthesis on scRNA templates stretches out the scRNA into an extended, rigid structure, which may be a poor template for replication. However, our experimental data suggest that rolling circle RNA synthesis is not only possible but can proceed over multiple full-length scRNA circles. We hypothesize that this reflects the fact that these rigid structures are likely in equilibrium with more relaxed structures, with more significant dissociation of the dsRNA segment from the circRNA template and / or kinks which relieve ring strain (even though these are not observed in our MD simulations) and it is those that can serve as templates for 3’-extension. Finally, although not supported by one-pot RCS experiments presented in Figure 4, it cannot be ruled out that the nascent strand comprises extension on two or even multiple scRNA templates. The precise mechanistic details do merit further study and this work is ongoing in our lab, but would lead too far for the current publication.
The study depends on trinucleotide triphosphates (triplets) as substrates. They demonstrate a viroid like replication that show how a template (-) is able to make a mirror copy (+) of circular RNA by a rolling circle synthesis with out the need for enzymes or other catalyst apart from RNA itself.
While the conclusion of the work becomes a bit muddled, but honest, the work is a very important piece that demonstrates the huge potential role of circular RNA in the very early stages of life.
We thank the reviewer for his comments and have endeavored to make our conclusions as clear as possible. Furthermore, we have rewritten part of the Discussion section of the manuscript to increase clarity.
I must confess I am very far off my field of competences. I have tried my best to understand the paper, and the methods involved, but obviously I cannot give much feedback on the methods used. So, I cannot suggest improvements on that part.
For the most part, the paper is well written, and the structure is sound. I would strongly recommend publication.
It was unclear to me how it was certain that it was actually rings. There is the MD simulations, but apart from the principle drawing they are obviously not circular in shape.
Referee 1 raises an important point. Our study depends on the circularity of the RNA template molecules, i.e. small circular RNAs (scRNAs). We have therefore gone to some lengths to develop robust methods to efficiently generate such scRNAs, as well as to confirm their circularity experimentally. This is shown in Figure 1B and Supplementary Figure 1 (and described in Supplementary information Materials and Methods). scRNAs are generated by ligation from linear RNAs and circularity is confirmed by exonuclease degradation (exoT). Only scRNA can resist degradation by exoT, as only circular RNAs do not present a free RNA 3’-end as a substrate (Figure 1B). Circularity is further confirmed by electrophoretic mobility shift (Figure 1 —figure supplement 1 and Figure 4 —figure supplement 1).
To illustrate these points more clearly we have expanded the Figure legends for Figure 1B and Figure 1 —figure supplement 1, Figure 4 —figure supplement 1 and have also expanded the Materials and methods section describing experimental procedures to generate and purify scRNAs.
There are some parts of the manuscript that too me is very difficult to understand. Some statements seem to reflect a community understanding that might be obvious for those working with similar system on a day to day basis, but for the general reader (where I put myself) this becomes gibberish. for instance:
RCS has potentially unique properties with regards to the strand inhibition problem where RNA duplex melting in principle can be effected by continuous toehold strand displacement driven by nucleotide hybridization and the ratchet of nascent strand extension by triphosphate hydrolysis. In an idealized RCS mechanism, such strand invasion and displacement processes are both isoenergetic and coordinated to nascent strand extension (Blanco et al., 1989; Daubendiek et al., 1995), with rotation of the single-stranded RNA (ssRNA) preventing the build-up of topological tension (Kuhn et al., 2002). Thus, RCS is a potentially open-ended process leading to the synthesis of single-stranded multiple repeat products (concatemers) with an internally energized strand displacement circumventing the "strand inhibition problem" (Tupper and Higgs, 2021).
It might not be possible to rephrase this so that everyone can understand. But I think the clarity of manuscript could be improved.
We would like to apologize for failing to make our arguments sufficiently clear and free of jargon. We have now rephrased this passage in effort to make it both clearer as well as more accessible to the general reader. Please find below the rewritten section as found in the revised manuscript:
“Specifically in the context of triplet-based RNA replication on a circular template, duplex dissociation and strand separation may in principle be driven by trinucleotide (triplet) hybridization and ligation, leading to extension of the nascent strand 3’-end and an equal displacement of the 5’-end in triplet increments (Figure 1A). Triplet binding to the template strand and dissociation of an equal trinucleotide stretch from the 5’-end are both equilibrium processes and nearly isoenergetic. However, extension (i.e. ligation of the bound triplet to the growing 3’-end) is an irreversible step. Thus, in this scenario RCS would be expected to proceed in ratchet-like fashion with strand displacement driven by triphosphate hydrolysis and triplet ligation.”
But also because there is references to previous work where long statements does not make it more clear what is actually meant. for instance:
Similar to what was described previously, RNA synthesis by the TPR best in the eutectic phase of water ice, due to beneficial reaction conditions for ribozyme catalysis such as reduced RNA hydrolysis and high ionic and RNA substrate concentrations (Attwater et al., 2010). This was also the case on scRNA templates.
Again, we would like to apologize for failing to make our arguments sufficiently clear to be easily understood. We have now rephrased this passage in an effort to make it both clearer as well as more accessible to the general reader. Please find below the rewritten section as found in the revised manuscript:
“As described previously, RNA synthesis by the TPR is most efficient in the eutectic phase of water ice, due its beneficial reaction conditions for ribozyme catalysis (Attwater et al., 2018). Specifically, eutectic ice phases aid TPR activity by the reduced degree of RNA hydrolysis under low temperature conditions, reduced water activity, and the high concentrations of reactants (ribozyme, scRNA template, triplet substrates and Mg2+ ions) present in the eutectic brine phase that arise by excluding solutes from growing ice crystals and remains liquid at subzero temperatures (Attwater et al., 2010). Thus, all RCS experiments were carried out under eutectic conditions."
That said, I think the overall message is clear and the paper is very interesting, I expect to reference it when it comes out.
Reviewer #2:
[…]
Lines 122-126 – It is implied that triplets are better than monomers for rolling circle replication because triplets help to open up other double stranded regions. However, it is not obvious that this should be the case. To put a new triplet down you have to displace three bases from the displaced strand, whereas to put a monomer down you only have to displace one base. It is not easy to predict which of these is faster without measuring it. Furthermore, in the actual mechanism occurring here, there is no prior strand to be displaced at the point of attachment, because the displacement is occurring at the other side of the circle and it does not directly interfere with the attachment. So it is not clear whether this argument applies.
Referee 2 raises an important point and we welcome the opportunity to clarify our arguments.
The implication that triplets are able to “open up” double-stranded RNA regions is based on data in our previous publication (Attwater et al., 2018), which describes the properties of the triplet polymerase ribozyme (TPR) and the general TPR performance on linear RNA templates, which form stable secondary structures (such as hairpins). The key observation is that the TPR can replicate through such RNA structures as a function of triplet concentration, while the standard mononucleotide RNA polymerase ribozyme (RPR) cannot.
Referee 2 is of course correct in that triplets require the displacement of a 3 nt stretch from the primer / nascent strand 5’-end, whereas a mononucleotide RPR would require only a single base-pairing interaction to be disrupted. However, if this were the main energetic bottleneck of strand extension, surely the same argument would apply to template secondary structures, where again only one base-pairing interaction at the hairpin base would need to be disrupted compared to three for triplet invasion. However, our data show that only triplets endow the polymerase ribozyme with a general ability to invade and replicate through template secondary structures.
While we do not know the precise mechanistic basis for this, the fact that secondary structure invasion is a cooperative effect that occurs as a function of triplet concentration suggests, that it may be based on the ability of triplets to bind to intermittently accessible template structures. Triplet binding would stabilize them in their open form enabling the ribozyme to covalently link them to the nascent strand. Indeed, extension is an irreversible process (a ratchet), while 5’-end displacement is an equilibrium process suggesting that given sufficient time even an inefficient extension “ratchet” will “win”.
We of course accept that here may be other mechanisms for processive RNA synthesis on circular templates, such as e.g. tethering or topological linkage mechanism (as explored e.g. by (Cojocaru and Unrau, 2021)). However, we note that Cojocaru and Unrau do not investigate beyond full length circle synthesis and it is therefore currently unclear if their ribozyme is merely more processive on an “open“ template or if its superior activity extends to strand displacement.
We have now sought to explain our arguments more clearly in the revised manuscript (see also comments above).
Has replication of a circular strand actually been attempted with a monomer ribozyme? Is it known whether a triplet ribozyme is better than a monomer ribozyme on circular templates? If not, it would be better to avoid implying this.
We have not tested the mononucleotide RPR on circular RNA templates as – at the time this work was performed – all RPR variants required a hybridization tether to the template for processive RNA synthesis. On a circular RNA template this would likely cause both topological issues as well as – for full length synthesis – require displacement of the tether from the template for further extension. In contrast, the TPR does not require a tether for processive RNA synthesis by virtue of its non-catalytic t1 RNA subunit (Attwater et al., 2018) and this – together with above (2.1) described properties of triplets with regards to template secondary structure invasion – were the reasons we sought to explore triplet-based RCS. Recently, a more advanced RPR version has been described that appears to use a σ factor-like initiation mechanism that enables processive synthesis specifically on much larger, circular RNA templates (ca. 200 nt), where strain may be a lesser concern (Cojocaru and Unrau, 2021). However, as discussed above (2.1), strand displacement was not investigated in this report.
We have rewritten the relevant sections of the manuscript to clarify our arguments
Figure 1 – the periodic effect seen in 1E is claimed to be due to the difference of accessibility of template bases on the inside and outside of the circle. However, the results are measured by averaging over many different circular templates. I would expect that different copies of a circular template would have different configurations and would not always have the same bases on the inside. So the inside-outside difference should average out. Could the variability of 1E be explained by variation in rates of addition according to the sequence of the template rather than an inside-outside effect?
This is a perceptive observation and we have considered the exact same problem. Our explanation, even though speculative, is that the primer binds first and remains bound stably and therefore orients the nascent strand in a particular configuration with decreasing flexibility as the double-stranded RNA segment grows (as indicated by the MD simulations). Indeed, with primer bound and stably oriented – as illustrated in Figure 1F, we would expect to see the observed banding pattern based on accessibility of the triplet junctions. Again, the positioning of the primer is supported by the MD simulations (Figure 3A). As the template in this case is a UUC repeat, sequence variations (or variations in the rate of triplet addition according to the sequence of the template) are unlikely to account for the pattern observed. Indeed, we did not observe such a pattern in linear or non-repetitive circular template sequences. Furthermore, the same (or a very similar) periodic effect was observed on other repetitive circular templates, but not on their linear counterparts of identical sequence (see Figure 2 —figure supplement 1B).
The sequence effect would be the same in multiple copies of the same template sequence. Is there a similar variability seen when copying linear templates?
When copying linear templates (of similar sequence as the circularized RNA template) the periodic effect is not seen.
Figure 1C shows a TPR dimer. Is the polymerase actually in two parts? Is this important?
Yes, the triplet polymerase ribozyme holoenzyme is a heterodimer comprising an active (catalytic) RNA subunit (5TU) and an inactive subunit (t1). While the catalytic subunit by itself is a polymerase ribozyme, the non-catalytic subunit aids processive RNA synthesis (see Attwater et al., 2018). Therefore, the dimeric form of the TPR is the holoenzyme. This is important as the noncatalytic subunit is responsible for its ability to synthesize RNA in the absence of a template hybridization tether, which is specifically beneficial on circular RNA templates (as already outlined in more detail in 2.2). An in-depth description of the TPR and is properties can be found in our previous paper (Attwater et al. 2018). This has been specified in line 129-136 in the revised manuscript.
Line 213 – It is not clear why the 9 bp primer goes straight to 18 bp. What happened to the lengths in between?
In order to limit the number of MD runs, we concentrated on scRNAs with dsRNA segments reflecting later triplet additions (+3 triplets, 18 / + 4 triplets, 21 etc.) as we anticipated that ring strain would only begin to manifest itself in these molecules. Indeed, on scRNAs simulations with 9 bp and 18 bp of dsRNA, we observed an equally unbent double helical part, because the structure of the ssRNA part is longer and more relaxed and so it doesn’t exert any pulling force on the rest of the circle. We therefore did not model intermediate lengths 12 ( +1 triplet) and 15 (+ 2 triplets).
Line 220 – The word "extended" is used to mean that the unhybridized portion is stretched. There is a possible confusion with extending a strand by ligation of a triplet. Maybe the use of a word like stretched is better?
We thank the reviewer for this suggestion. “Extended” is indeed a questionable choice of word and liable to be confused with triplet extension. We have now replaced it with “stretched.”
Line 226 – The simulations show that the double-stranded part of the circle is stiff and only covers roughly half of the circle. The point at which the primer extension occurs is therefore far away from the point at which the two strands separate. This is important for very short circles. For longer circles, the stiffness should be less relevant, and there will come a point where the whole circle becomes double stranded. There will then need to be a true strand displacement occurring very close to the point of primer extension. How long would we need the circle to be before it switches to a double-stranded circle?
This is an important point. In the manuscript we suggest that this limit should be around the persistence length of RNA, which is about 200 bp. This mechanical property indicates the minimum length from which a polymer starts to behave flexibly, i.e. can significantly bend.
Does the stiffness effect seen here with the short circles make the primer extension reaction easier or more difficult than a true strand displacement reaction on a double stranded circle?
As discussed to some extent in (1.1) there are two conflicting mechanisms at work. On one hand, ring strain should aid strand displacement and therefore RNA synthesis beyond full length on a circular RNA template by destabilizing 5’-end hybridization and thereby aiding binding of triplets and extension. On the other hand, ring strain would be equally likely to destabilize 3’-end hybridization and thereby hinder extension and these two processes should therefore cancel each other out. Furthermore, there are other confounding factors such as restricted accessibility of some triplet junctions due to ring geometry (see Figure 1), which may further reduce extension efficiency compared to a linear RNA template. But 3’-end extension by covalent triplet ligation is a unidirectional, irreversible process and therefore overall we observe primer extension, rolling circle synthesis and strand displacement.
Lines 228-33 – This paragraph is not very clear. The meaning is not coming through.
We would like to apologize for not making the parsing clear and thank Reviewer 2 for making us aware of this. We have now reprised the paragraph to the following in effort to clarify:
“In the experimental data, we also observed an inhibitory effect for insertion of the final triplets (+8, +9, and +10 (beyond full length) / extension to 33, 36 and 39 nt of RNA in Figure 2D) into the corresponding scRNA template. This may indeed reflect the onset of the 3’- and 5’-end destabilization observed in the MD simulations (Figure 3), which would likely attenuate primer extension by the ribozyme. Note however that the extension efficiency recovered beyond full length (+11 / extension to 41 nt, Figure 2F), although at lower speed (Figure 2E).”
Line 288 – "orbit" is an odd word. Is there a better one?
We have now replaced “orbit” throughout the manuscript with “full length circle synthesis”. We agree with referee 2 that orbit may be liable to misinterpretation.
Figure 4 – Overall Figure 4 is not clear.
– I have not understood the notation n: 3E5, 2E5 etc.
– For sequence A Pos 4, GAA is the darkest shade, so I am presuming the template is CUU (in the reverse direction). But the second darkest shade is GGC. Why should GGC bind to CUU more strongly than others (for example GGA)? I am not sure whether I have understood this diagram correctly.
The diagram represents the deep sequencing result of a one-pot extension reaction where all four templates were mixed, in the presence of the TRP and triplets (CUG, AUA, CCA, CCC, AAA, CAC, GGG, UUA, UCC, GGC, AUC, GAU, CGC and GAA). The reads from the four templates were assorted relative to position 3 (the first barcode triplet). For the rest of the positions, the diagram presents (in %) which triplets were identified. The notation “n:” stands for number of reads used for the analysis at each position. 3E5, 2E5 represent the number (3*10^5, 2*10^5 etc.). This has now been specified in Figure 4B line 356-369 in revised manuscript.
The darkness of the shading represents the amount (%) of triplets found at the noted position. As correctly stated in the comments, the (reverse order) template for pos 4 is CUU and indeed GAA is the expected correct triplet. The expected triplet for each position is marked with a box in the diagram. In pos 4, the expected GAA is darkest shade. However, as he sequencing data indicated, GGC clearly is also misincorporated at this position to some degree and this is represented in the figure. This has been specified in the text line 328-341 in revised manuscript.
The overall fidelity of the TPR (per base) for linear templates is 97% (Atwater et al. 2018), but positional fidelity can vary and GC-rich triplets can be particularly prone to misincorporation both in a templated context (due to e.g. G-U mispairing) as well as in a non-templated context, as triplet terminal transferase extension of free unpaired 3’- ends.
– Part C shows fold difference. It would be easier if rates where shown for linear and circular strands separately. Why is sequence D a worse template? Or maybe it is not worse – it's just that the ratio of circular to linear templates is lower? It is not easy to understand this.
The fold difference in fidelity that is shown in Part C, is calculated based on the fidelity (%) presented in B, thus the ratio of templates should not play a role. However, to ensure that the concentration and integrity (circular or linear) of the templates were validated (by gel purification and absorbance measurement) and an equal amount of each were added in the experiment.
Calculation of plot in C is now specified in line 365-367 (in revised manuscript).
– Part D Figure 2 seems to show a double-stranded triplet being added. Why not just a single-stranded triplet.
A double stranded segment (consisting of e.g. two complementary triplets (triplet and anti-triplet, e.g. GGC and CCG), or a template and a number of triplets) will stack and may form a substrate for non-templated terminal transferase addition by the TPR. We have not discussed this in the manuscript as more data would be needed for a conclusive mechanistic description. As this remains therefore a somewhat speculative model (although consistent with the data obtained), we have now changed the figure to show a single stranded triplet as this is, as suggested, more intuitive. We have also added an explanation to the Figure 4 legend to the effect that the precise molecular nature of non-templated addition is currently unknown.
Figure 6D – It is unclear what is happening at each step. Particularly the backwards and forwards diagrams in step 4. Also, shouldn't the red strand be still attached to the blue circle before the cleavage occurs? The chemical structure in the middle is a bit distracting. I think the structure drawn in A is the same as D step 6. Maybe put parts A and D together and make B and C a separate figure?
We have changed the order in the figure so that the diagram with the reaction steps is presented before the data (see new Figure 6). In lines 424-430 (see revised manuscript) we have specified what the steps in the diagram denotes. Also, the Figure annotations for Figure 6A-D has been updated to fit the new Figure.
Line 436 – The reference to the virtual circular genome is misleading at this point. In the proposal of Zhou et al., there are no real circles, there are simply linear fragments that can be aligned to form a virtual circle. This does not fit with the rest of this paragraph. Either the reference to Zhou et al. should be omitted or it should be explained properly what the virtual circle proposal is.
We welcome this perceptive point and accept that our reference may be misleading in the context. However, while it is true that in the virtual circular genome proposal of Zhou et al., there are no real circles during the replication phase, Zhou et al. still propose circular RNAs arising from abiotic circularization reactions as an initial source of a circular genome. Nevertheless, we have now rewritten this section to clarify our arguments and removed the reference to Zhou et al.
Reviewer #3:
Technically the experiments are sound, really comprehensive, convincing and the paper is well written. The documentation (the composed Figures etc.) is very nice. The MD simulations nicely complement the experiments. Strong point is that the simulations address a qualitative question and are clearly directed to solve it. It is a preferable application of the MD technique. The basic methodology is correct, the standard AMBER OL3 force field appears appropriate, as the first choice multipurpose RNA version. It is known to lead to over-compacted unstructured ssRNA ensembles, as all biomolecular force fields that are good for folded biopolymers. For the double strand, the circle and their flexibility it should be an optimal choice. The simulations are quite short by contemporary standards, though I do not think their prolongation would change the essence of the findings.
As noted above, I consider the experiments as very convincing. Strong point is that the accompanying simulations address a qualitative question and are clearly directed to solve it. It is a preferable application of the MD technique. So, I really like it, though I have some ideas for potential minor improvements, may be explanatory comments, all for supporting information.
There are some occasional typos, e.g. l. 180 stand displacement.
We have corrected this in the revised manuscript.
l. 182 this suggest. I think on l. 56. where progress in non-enzymatic synthesis is overviewed, the reference could be more balanced. Appears to me that some groups are represented by duplicate citations while some research is omitted, for example a recent progress in template-free non-enzymatic RNA polymerization of 3',5' cyclic nucleotides , https://chemistry-europe.onlinelibrary.wiley.com/doi/10.1002/syst.202100017.
We have focused our introduction and citations mainly on enzymatic (RNA-catalyzed) polymerisation of RNA and believe we provide a fair and balanced account and citation of the main advances in that specific field. Where we have briefly alluded to prebiotic chemistry including references to non-enzymatic RNA replication, we have concentrated on chemical approaches that are closely analogous to the enzymatic chemistry e.g. polymerization via an (activated) 5’phosphate. But reviewer 3 is of course correct that we should not discount alternative prebiotic mechanisms as these may have similarly given rise to early (circular) RNA templates for enzymatic replication. We have now therefore expanded and modified the citations in the introduction to include e.g. the above-mentioned publication as well as others.
The short "jumping" supplementary movies are difficult to follow, I assume it is because of the size of the movies. Would it be possible to create a few SI Figures showing details of the most interesting parts of the structure, to focus on the key details, to accompany the movie?
We thank the reviewer for his suggestion. We have created Figure 3 —figure supplement 1 and 2 with a series of snapshots along the two trajectories showing the main transition events.
In the simulations, lot of emphasis is paid on the Mg2+ simulations up to 500 mM MgCl. First point, is this condition relevant to the origin of life?
As now outlined in more detail in the manuscript, the RNA polymerization activity of our triplet polymerase ribozyme is enhanced in the eutectic phase of water ice at subzero temperatures. Such eutectic phases from when water containing solutes freeze and solutes such as Mg2+ ions, RNA etc. are excluded from the growing water-ice crystals into a liquid brine (eutectic) phase that surrounds the ice crystals and remains liquid at subzero temperatures. This process is accompanied by a profound concentration and dehydration effect, which enhances RNA synthesis and reduces RNA degradation, but also increases counter-ion concentrations to the above levels (0.5M).
The formation and presence of water ice on the early earth can of course not be proven, but is not implausible under currently favored geochemical scenarios for example at the poles or as part of seasonal or diurnal temperature variations. We have outlined our thinking on the beneficial properties of such ice phases in detail in one of the cited refs (Attwater et al., 2010) and now also discuss it in more detail within the main body of the revised manuscript. Furthermore, we now include an explicit reference to the rationale for simulations at high Mg2+ conditions with regards to the importance of eutectic ice phases in the experimental section.
Second, inclusion of divalents into MD is always risky, as they sample poorly (which is further exacerbated by the lack of bulk background due to the small periodic box, which may lead to glassy-like ion behavior around the solute). 400 ns is not sufficient to converge Mg2+. In addition, divalents, especially the high charge density Mg2+, are beyond the pair-additive MM approximation. It is impossible to simultaneously balance ion hydration and inner vs. outer shell binding to different coordination sites with the simple MM models. Could the authors briefly comment on initial placement of the ions after equilibration and during MD? Was it always hexacoordinated outer-shell binding to the RNA? Could the authors in SI briefly comment on it, and also comment if they had some specific reasons to choose the Duboue-Dijon parameters over parameters that have been more commonly used in biomolecular simulations? As I am not sure these specific parameters were tested/calibrated for RNA interactions (but I am not fully familiar with the work). Again, as the results are qualitative, I do not expect any effect on basic outcome of the work, so I am not suggesting any new computations.
We thank the reviewer for raising all these important points. We agree that Mg2+ represents a challenge for empirical force-field calculations due to strong polarization and charge-transfer effects. Indeed, nonpolarizable Mg2+ parameters tend to overestimate the formation of contact ion pairs by a factor of 2 to 3 (Fingerhut et al., 2021). We chose the parameters developed by Jungwirth and coworkers (Duboue-Dijon et al., 2018) because they drastically reduce ion−ion interactions with respect to ion−water interactions (see new Figure 3—figure supplement 6; Fingerhut et al; Kirby and Jungwirth, 2019). They account for the quantum effects by assuming a mean field approach using the so-called Electronic Continuum Correction (ECC), which is numerically implemented by rescaling the charge of multivalent ions, in the case of Mg2+ to 1.5.
We agree with the reviewer that 400 ns is short for an accurate description of the ion environment. However, we have added Figure 3 —figure supplement 6 showing reasonable convergence of ion location around RNA within the first and second hydration shell for the time course of our simulations.
We have added this extra information into the revised manuscript (Method section and Figure 3 – Supplement Figure 6).
https://doi.org/10.7554/eLife.75186.sa2Article and author information
Author details
Funding
Carlsbergfondet (CF17-0809 & CF19-0019)
- Emil Laust Kristoffersen
Medical Research Council (MC_U105178804)
- Philipp Holliger
Engineering and Physical Sciences Research Council (EP/N027639/1)
- Agnes Noy
Engineering and Physical Sciences Research Council (EP/R513386/1)
- Matthew Burman
Engineering and Physical Sciences Research Council (EP/T022205/1)
- Agnes Noy
The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.
Acknowledgements
This work was supported by the Carlsberg Foundation (CF17-0809) (ELK), by the Medical Research Council (MRC) program Grant program no. MC_U105178804 (PH), by the Engineering and Physical Sciences Research Council (EPSRC) Grant EP/N027639/1 (AN) and by the EPSRC (EP/R513386/1) (MB). Simulations were performed on JADE (EP/T022205/1). The authors thank the HecBiosim consortium (EP/R029407/1), Cambridge Tier-2 (EP/P020259/1), and the local York facilities. Correspondence and requests for materials should be addressed to PH.
Senior Editor
- James L Manley, Columbia University, United States
Reviewing Editor
- Timothy W Nilsen, Case Western Reserve University, United States
Reviewer
- Jiri Sponer
Version history
- Received: November 1, 2021
- Preprint posted: November 30, 2021 (view preprint)
- Accepted: February 1, 2022
- Accepted Manuscript published: February 2, 2022 (version 1)
- Version of Record published: March 21, 2022 (version 2)
- Version of Record updated: March 22, 2022 (version 3)
Copyright
© 2022, Kristoffersen et al.
This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.
Metrics
-
- 4,749
- Page views
-
- 687
- Downloads
-
- 15
- Citations
Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.
Download links
Downloads (link to download the article as PDF)
Open citations (links to open the citations from this article in various online reference manager services)
Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)
Further reading
-
- Biochemistry and Chemical Biology
- Cell Biology
Pancreatic a-cells secrete glucagon, an insulin counter-regulatory peptide hormone critical for the maintenance of glucose homeostasis. Investigation of the function of human a-cells remains a challenge due to the lack of cost-effective purification methods to isolate high-quality a-cells from islets. Here, we use the reaction-based probe diacetylated Zinpyr1 (DA-ZP1) to introduce a novel and simple method for enriching live a-cells from dissociated human islet cells with ~ 95% purity. The a-cells, confirmed by sorting and immunostaining for glucagon, were cultured up to 10 days to form a-pseudoislets. The a-pseudoislets could be maintained in culture without significant loss of viability, and responded to glucose challenge by secreting appropriate levels of glucagon. RNA-sequencing analyses (RNA-seq) revealed that expression levels of key a-cell identity genes were sustained in culture while some of the genes such as DLK1, GSN, SMIM24 were altered in a-pseudoislets in a time-dependent manner. In conclusion, we report a method to sort human primary a-cells with high purity that can be used for downstream analyses such as functional and transcriptional studies.
-
- Biochemistry and Chemical Biology
- Cell Biology
Eukaryotic cells control inorganic phosphate to balance its role as essential macronutrient with its negative bioenergetic impact on reactions liberating phosphate. Phosphate homeostasis depends on the conserved INPHORS signaling pathway that utilizes inositol pyrophosphates and SPX receptor domains. Since cells synthesize various inositol pyrophosphates and SPX domains bind them promiscuously, it is unclear whether a specific inositol pyrophosphate regulates SPX domains in vivo, or whether multiple inositol pyrophosphates act as a pool. In contrast to previous models, which postulated that phosphate starvation is signaled by increased production of the inositol pyrophosphate 1-IP7, we now show that the levels of all detectable inositol pyrophosphates of yeast, 1-IP7, 5-IP7, and 1,5-IP8, strongly decline upon phosphate starvation. Among these, specifically the decline of 1,5-IP8 triggers the transcriptional phosphate starvation response, the PHO pathway. 1,5-IP8 inactivates the cyclin-dependent kinase inhibitor Pho81 through its SPX domain. This stimulates the cyclin-dependent kinase Pho85-Pho80 to phosphorylate the transcription factor Pho4 and repress the PHO pathway. Combining our results with observations from other systems, we propose a unified model where 1,5-IP8 signals cytosolic phosphate abundance to SPX proteins in fungi, plants, and mammals. Its absence triggers starvation responses.