Cross-chiral exponential amplification of an RNA enzyme

Significance

A special form of autocatalysis, rarely seen outside of biology, involves self-replication with exponential growth. Biology relies on building blocks of the same chiral handedness, including nucleic acids that contain only D-sugars, but in the laboratory one is free to explore both sides of the mirror. Here, a system of cross-replicating RNAs was devised based on a ribozyme that joins two RNA fragments of the opposite handedness to form its own enantiomer, which in turn forms a new copy of the original ribozyme. The D- and L-ribozymes grow together exponentially and can be propagated indefinitely, demonstrating a form of autocatalytic self-replication that relies on an RNA enzyme and its mirror image.

Abstract

An RNA ligase ribozyme that catalyzes the joining of RNA molecules of the opposite chiral handedness was optimized for the ability to synthesize its own enantiomer from two component fragments. The mirror-image D- and L-ligases operate in concert to provide a system for cross-chiral replication, whereby they catalyze each other’s synthesis and undergo mutual amplification at constant temperature, with apparent exponential growth and a doubling time of about 1 h. Neither the D- nor the L-RNA components alone can achieve autocatalytic self-replication. Cross-chiral exponential amplification can be continued indefinitely through a serial-transfer process that provides an ongoing supply of the component D- and L-substrates. Unlike the familiar paradigm of semiconservative nucleic acid replication that relies on Watson–Crick pairing between complementary strands, cross-chiral replication relies on tertiary interactions between structured nucleic acids “across the mirror.” There are few examples, outside of biology, of autocatalytic self-replication systems that undergo exponential amplification and there are no prior examples, in either biological or chemical systems, of cross-chiral replication enabling exponential amplification.

In autocatalytic reactions, a product molecule catalyzes its own formation, and the rate of product synthesis increases as more product is formed, until some other aspect of the reaction system becomes rate-limiting. This self-dependent rate acceleration is characteristic of growing populations of living organisms, and, like biological organisms, autocatalytic chemical reactions can lead to the emergence of complex patterns and adaptive behaviors (1–3). An exceptional situation arises when the rate of autocatalysis is directly proportional to the concentration of the catalyst, which causes the amount of catalyst to increase exponentially over time. So long as there is an adequate supply of substrate(s), exponential growth can rapidly enrich a privileged molecule and preserve it against degradation or dilution within its environment (4–7). By analogy to reproduction in biology, autocatalytic chemical systems are sometimes referred to as “self-replicating,” although others reserve this term for systems that undergo exponential amplification or that involve informational macromolecules, or both (8–10).

In extant life, nucleic acids are replicated by proteins, but the synthesis of RNA and DNA can be carried out without the aid of proteins, involving either the polymerization of mononucleotides (11) or the ligation of oligonucleotides (12). It is widely thought that life on Earth originated with such systems, and in the laboratory, Watson–Crick pairing provides a straightforward means for a product nucleic acid to direct the synthesis of additional products through reciprocal templating. Autocatalytic RNA- and DNA-based amplification reactions have been demonstrated in a variety of formats, reflecting an inherent capacity for self-replication, although true exponential growth has proven more difficult to achieve (13–17).

The first reported example of a self-replicating nucleic acid involved a self-complementary hexadeoxynucleotide template that binds two complementary trinucleotides, which are chemically ligated to form another copy of the template (13). In this reaction and many other subsequent examples, however, dissociation of the paired template and product strands is rate-limiting. As a consequence, the reaction rate is proportional to the square root of the concentration of the catalyst, leading to parabolic rather than exponential growth (18–21).

To enable more favorable self-replication kinetics, an RNA ligase ribozyme was developed that joins two component RNA fragments to form another copy of itself (14). The ribozyme engages the two substrates through short regions of base pairing and tertiary contacts, such that product dissociation is not rate-limiting. However, product-dependent growth is overshadowed in this system by the spontaneous formation of new ribozymes in a reaction that involves two pairs of substrates without the need for preformed catalyst. When assembled into a quaternary complex, the substrates catalyze their own ligation. The self-replicating system was subsequently engineered to operate in a cross-replication format, where a pair of ribozymes direct each other’s synthesis from 4 component substrates (22). The spontaneous reaction path is less favorable in the cross-replication format, but still only 2.3-fold slower than the rate of autocatalytic synthesis.

Exponential, autocatalytic growth of RNA was finally achieved after using directed evolution to improve the catalytic efficiency of the cross-replicating ligase ribozymes so that the product-catalyzed reaction dominates over the spontaneous reaction (23). This system exhibits the logistic growth pattern characteristic of population dynamics in biology, with the initial rate of reaction directly proportional to the concentration of catalyst (24). Further improvement by directed evolution achieved an exponential growth rate of 0.14 min^–1, corresponding to a doubling time of 5 min. Exponential growth can be continued indefinitely through a serial-transfer process that provides an endless supply of substrates. A portion of a completed reaction mixture is transferred to a fresh reaction mixture, seeded by whatever catalyst molecules are carried over in the transfer. The most advanced cross-replicating ribozymes have demonstrated an overall amplification of 10¹⁰⁰-fold in 36 h (25).

All extant life is homochiral, relying exclusively on the D-enantiomer of RNA or DNA. Prior examples of autocatalytic systems based on nucleic acids also have relied exclusively on the D-enantiomer. Outside of biology, however, nucleic acids can be prepared in either the D or L form, which have mirror-image structure and identical catalytic function when operating with corresponding mirror-image substrates (26–29). It is not possible to form consecutive Watson–Crick pairs between nucleic acids of the opposite handedness (30, 31), and thus when a D-ligase interacts with an L-RNA substrate it must bind the substrate entirely through tertiary contacts. This mode of cross-chiral interaction avoids the impediment of product inhibition that results from Watson–Crick pairing interactions in a homochiral self-replicating nucleic acid system. Cross-chiral replication also provides a different form of macromolecular information transfer compared to that of extant biology, with information being propagated across the mirror rather than across the double helix.

In a prior study, a cross-chiral ligase ribozyme was obtained that catalyzes the RNA-templated joining of RNA substrates of the opposite handedness. This ribozyme, termed 16.12t, was the result of 16 rounds of directed evolution and subsequent truncation to remove extraneous nucleotides (32). In a subsequent study, further optimization resulted in the 27.3t ligase, which required 11 additional rounds of evolution and removal of extraneous nucleotides (33). When prepared as D-RNA, these ligases can catalyze the joining of multiple L-RNA oligonucleotide substrates at rates up to 1.4 min^–1. They can synthesize functional L-RNA ribozymes, including L-16.12t (32, 33), and can even amplify the L-RNA form of the hammerhead ribozyme through iterative thermocycling (34). The D- and L-ligases can operate simultaneously on their respective L- and D-substrates within a common reaction mixture, but reciprocal synthesis of D- and L-ribozymes has not yet been shown.

If a pair of D- and L-ligases could be made to synthesize each other in a reciprocal manner, it would provide an autocatalytic reaction system that achieves cross-chiral replication. In the present study, a candidate cross-chiral self-replication system was constructed based on the 27.3t ligase, dividing it into 2 fragments that can be ligated by the enantiomeric form of the ribozyme. The D-ligase catalyzes joining of the L-fragments to form the L-ligase, and the L-ligase similarly catalyzes joining of the D-fragments to form the D-ligase (Fig. 1A). The 27.3t ligase required further optimization through directed evolution to improve its catalytic efficiency, leading to the identification of a cross-chiral ribozyme that can undergo exponential amplification. Both the D- and L-ribozymes become amplified through their reciprocal synthesis from 4 component fragments. Amplification is not limited by-product dissociation, and the initial rate of reaction is directly proportional to the concentration of ribozymes. The replication process can be sustained indefinitely through a serial-transfer procedure that provides an ongoing supply of the D- and L-fragments.

Fig. 1.

A cross-chiral, cross-replicating RNA enzyme. (A) The replication cycle entails cross-chiral ligation of 2 D-RNA fragments (D1 and D2), catalyzed by the L-ligase, to form the D-ligase; which in turn catalyzes cross-chiral ligation of 2 L-RNA fragments (L1 and L2) to form the L-ligase. (B) Sequence and secondary structure of the 47.11 ligase, with mutations relative to the starting 27.3t ligase indicated by red circles. Blue circles indicate mutations that were introduced to provide a favorable ligation junction; the curved arrow indicates the site of ligation. Nucleotides in magenta were added prior to directed evolution. Stem elements P1–P6 are labeled.

Results

Reciprocal Cross-Chiral Synthesis.

The 27.3t ligase ribozyme was divided into 2 RNA fragments that can associate noncovalently through base-pairing interactions, retaining the structure of the full-length enzyme (Fig. 1B). This division was made between residues 70 and 71. In addition, an A-to-G mutation was made at position 71, together with a complementary U-to-C mutation at position 46, to provide an especially favorable C•G sequence flanking the ligation junction (33). The D and L forms of the 5′ fragment (termed D1 and L1, respectively) were prepared synthetically, with spectrally distinct fluorescent labels at the 5′ end. The D and L forms of the 3′ fragment (D2 and L2, respectively) also were prepared by chemical synthesis, including chemical 5′-triphosphorylation (35–37).

Both the D- and L-27.3t ligases have previously been shown to be catalytically active in the templated ligation of L- and D-RNA substrates, respectively (33). No ligation is observed when either the 2 D-RNA or the 2 L-RNA fragments of 27.3t are incubated alone, but when all 4 fragments are present, both the D- and L-RNA ligases are produced (Fig. 2A). The D-RNA form of 27.3t catalyzes ligation of the corresponding L1 and L2 fragments, with a first-order rate constant (k_obs) of 0.029 min^–1 in the presence of 10 µM of each fragment and 50 mM MgCl₂ at pH 8.3 and 17 °C (Fig. 2B).

Fig. 2.

Cross-chiral synthesis of the ligase ribozyme. (A) Cross-chiral ligation catalyzed by the 27.3t fragments. The D- or L- fragments alone do not react, whereas all 4 fragments together give rise to both the D- and L-ligases. Reaction conditions: 5 µM of each fragment and 50 mM MgCl₂ at pH 8.3 and 17 °C for 20 h. (B) Cross-chiral ligation under multiple-turnover conditions, employing 0.05 µM of either the D-27.3t (squares) or D-47.11 (circles) ligase, 10 µM each of the corresponding L1 and L2 fragments, and 50 mM MgCl₂ at pH 8.3 and either 17 °C (filled symbols) or 30 °C (open symbols).

Cross-catalytic synthesis of the D- and L-27.3t ribozymes does not exhibit exponential amplification. This result may be due to the modest rate of ligation and/or slow rate of catalytic turnover of the enzyme, either of which could prevent exponential growth (5, 22, 23). Elevated reaction temperature can improve turnover by increasing the rate of dissociation of enzyme and product. However, the 27.3t ligase has an even lower catalytic rate at higher temperatures, with a k_obs of 0.011 min^–1 at 30 °C (Fig. 2B), which corresponds to less than one turnover per hour. The autocatalytic properties of 27.3t might be improved through directed evolution, selecting for ribozymes that operate more rapidly and with multiple turnover at elevated temperature.

Directed Evolution of the Cross-Chiral Ligase.

A population of variants of the D-27.3t ligase was prepared by introducing random mutations throughout the molecule at a frequency of 10% per nucleotide position. The D-RNAs were covalently linked to an L-RNA primer, which was bound to a separate L-RNA template. The D-RNAs were challenged to extend the primer through multiple successive ligation reactions, each involving addition of a 5′-triphosphorylated L-RNA substrate that binds along the L-RNA template. The resulting full-length L-RNA products were isolated, together with the D-ribozyme that had catalyzed their synthesis. These functional ribozymes were reverse transcribed, PCR amplified, and forward transcribed to generate a progeny population of ribozymes.

Eleven rounds of directed evolution were carried out, applying increasingly stringent selection constraints (SI Appendix, Table S1). These 11 rounds began with the 27.3t ligase, and thus to reflect evolutionary continuity are numbered beginning with round 28. The first 3 rounds (28 to 30) required 4 successive ligation reactions, progressively decreasing the reaction time from 7 to 1 h. The next 7 rounds (31 to 37) required 10 successive ligation events and the last round (round 38) required 6 ligation events. The full-length products were selected by various means, including incorporation of a 3′-terminal biotin moiety and subsequent capture on streptavidin-coated beads (rounds 28 to 30), isolation of materials by high-resolution, denaturing polyacrylamide gel electrophoresis (rounds 31 to 33), and functional selection based on the catalytic activity of the synthesized product (rounds 34 to 38).

The last 5 rounds of directed evolution required the D-ribozyme to extend a covalently linked L-RNA primer to synthesize the functional L1 fragment of the original 16.12t ligase (SI Appendix, Fig. S1). Following this reaction, the L-RNA template was removed and the L-RNA extension products were combined with the L2 fragment of the 16.12t ligase. The noncovalent assemblies of the synthesized L1 fragment and the supplied L2 fragment were challenged to ligate a 5′-biotinylated D-RNA substrate to the attached D-ribozyme. The resulting D-RNA ligation products were captured on streptavidin-coated beads, then the corresponding D-ribozymes were reverse transcribed and PCR amplified. One of the PCR primers was designed to bind selectively within the region corresponding to the attached D-RNA substrate.

The ability of the ligase ribozyme to catalyze multiple successive ligation events improved substantially over the final rounds of directed evolution, with the population of RNAs from round 38 yielding sixfold more full-length 16.12t L1 fragment compared to what is obtained when using the 27.3t ligase. The RNAs from round 38 were analyzed by deep sequencing, which revealed high sequence diversity with a broad distribution of mutations, including 12 nucleotide positions with >10% frequency of mutations (SI Appendix, Fig. S2). However, when tested for the ability to ligate the 2 fragments of the L-27.3t ligase, the activity of the round 38 population was similar to that of the D-27.3t ligase (SI Appendix, Table S2). Nonetheless, this population had been selected for the ability to catalyze multiple successive ligation events, which was considered to be conducive to the ultimate goal of multiple-turnover ligation as required for cross-chiral exponential amplification.

Compartmentalized Evolution of the Cross-Chiral Ligase.

A compartmentalized directed evolution strategy was devised to select directly for D-RNA ligases that catalyze multiple-turnover, cross-chiral ligation of the 2 fragments of the L-27.3t ribozyme (Fig. 3A). Microbeads were functionalized with covalently linked DNA primers that are complementary to an extended region that was added to the 3′ end of the ligase. Sense DNA strands derived from the round 38 population of RNAs were hybridized to these primers at a loading ratio of one DNA per bead. The bead-bound DNAs were clonally amplified by emulsion PCR, whereby individual beads were isolated within the aqueous compartments of a water-in-oil emulsion, together with the PCR reagents, then subjected to thermal cycling. The emulsion was broken and the microbeads were re-emulsified together with the reagents for in vitro transcription, which also was carried out in a clonally isolated manner. The transcribed ligase ribozymes were captured by the same bead-linked DNA primers so that each bead displayed many clonal copies of a particular ribozyme and the DNA that encodes it.

Fig. 3.

Compartmentalized directed evolution of the cross-chiral ligase. (A) Microbeads (gray circles) are functionalized with DNA primers (red) that, following emulsion PCR, capture ~10³ clonal copies of double-stranded DNA (green) that encodes a particular ligase ribozyme. 1) The emulsion is broken and the beads are re-emulsified with an in vitro transcription mixture to generate ~10⁴ clonal copies of the D-ligase (orange), which also are captured by the DNA primers. 2) The second emulsion is broken and the beads are functionalized with ~10⁵ copies of the 27.3t L2 fragment (blue), to which the fluorescently labeled (yellow circle) L1 fragment is hybridized. The beads are re-emulsified, together with an RNA-cleaving DNA enzyme that releases the D-ligases into solution, where they can catalyze cross-chiral ligation to form bead-bound copies of the L-27.3t ligase. 3) The third emulsion is broken and the noncovalently bound materials are removed. The beads are sorted by FACS, collecting those with the highest fluorescence. 4) The collected bead-bound DNAs are amplified by PCR, and the PCR products are clonally distributed into a new emulsion to begin the next round of evolution. (B) FACS data from rounds 39, 42, and 47 (gray, tan, and brown, respectively) of compartmentalized evolution. Bead fluorescence was normalized by subtracting the median fluorescence of the corresponding negative control population.

The second emulsion was broken and the L2 fragment of the L-27.3t ribozyme was covalently linked to the beads. The beads were emulsified for a third time, together with the fluorescently labeled L1 fragment of the L-27.3t ribozyme and the reagents for cross-chiral RNA ligation. The L1 and L2 fragments were prehybridized to form an L1•L2 complex, and an RNA-cleaving DNA enzyme was included in the mixture to catalyze release of the ligase ribozymes from the beads into solution within each compartment. The population of D-ligases were then challenged to ligate the L1 and L2 fragments, forming bead-bound L-27.3t ribozymes that carry a fluorescent label. The emulsion was broken and the harvested beads were screened by fluorescence-activated cell sorting (FACS) to isolate beads with the highest level of fluorescence. The selected beads were pooled and the attached DNAs were amplified by PCR to provide material for the next round of compartmentalized evolution.

Over the course of 9 rounds of compartmentalized evolution, the reaction temperature was increased from 17 to 30 °C and both the reaction time and ratio of ribozyme-to-substrate were decreased (SI Appendix, Table S2). The number of ribozymes and substrates per bead was controlled by adjusting the amount of attached DNA primers, the RNA transcription yield, and the amount of attached L2 fragments. Error-prone PCR was performed after each round (38). Approximately 3 × 10⁷ beads were screened by FACS in each round, isolating ~0.5% of the beads with the highest fluorescence. The bead-based fluorescent signal increased over the course of evolution (Fig. 3B), indicating improved cross-chiral ligation activity in a multiple-turnover reaction format. D-RNAs were prepared from each round of materials and assayed for their ability to ligate the 2 fragments of L-27.3t. For the population as a whole, there was a fourfold improvement in catalytic activity, which appeared to plateau by the latter rounds (SI Appendix, Table S2).

Deep sequencing analysis was carried out on materials collected from each round of compartmentalized evolution (rounds 39 to 47) to identify candidate ribozymes for further study. Moderate- and high-abundance sequences were characterized by Levenshtein distance-based principal component analysis. Twelve sequences were identified that collectively cover the densest regions of this distribution and individually had either reached high abundance (>1% of reads) or were significantly enriched (>fourfold) during the last 3 rounds of evolution (SI Appendix, Table S3). All of the mutations that had risen to prominence in the population by round 47 are represented among these 12 sequences, including covarying mutations within the P1 and P2 stems, a pair of mutations at positions 52 and 63 of the L4 loop, and individual mutations at position 24 of the L2 loop and position 106 of the J6/3 region (SI Appendix, Fig. S2). Conversely, sequence diversity had increased within the L5 loop, with 9 distinct variants among the 12 candidate ribozymes. The 5′ and 3′ termini beyond the P1 stem are highly variable in sequence and appear to be unstructured, and so were trimmed from most of the sequences.

The 12 candidate ribozymes were prepared as D-RNAs and tested for their ability to ligate the 2 fragments of L-27.3t (SI Appendix, Table S3). These ribozymes exhibited threefold to eightfold improved activity compared to D-27.3t. Five of the most active ligases were prepared as their respective D- and L-RNA fragments, including the 2 facilitatory mutations adjacent to the ligation junction (C46 and G71). The 4 RNA fragments that comprise the D- and L-ligases were combined at a concentration of 2 µM each, in either the presence or absence of 0.1 µM D-ligase. Autocatalysis with exponential amplification requires that the reaction products enhance the reaction rate compared to the spontaneous rate of reaction of the substrates alone (5). Two of the most active ligases exhibited substantially greater product yield in the presence compared to absence of a starting amount of D-ligase when tested at 30 °C. One of these ligases, designated 47.11, achieved 2.6-fold greater yield of L-47.11 when 0.1 µM D-47.11 was added at the start of the reaction (SI Appendix, Table S3). The 47.11 ligase contains 7 mutations and both 5′- and 3′-terminal insertions relative to the 27.3t ligase (Fig. 1B), and was chosen for subsequent studies of cross-chiral replication. When tested under multiple-turnover conditions, the D-RNA form of 47.11 catalyzes ligation of the L1 and L2 fragments with a k_obs of 0.15 min^–1 at 17 °C and 0.091 min^–1 at 30 °C (Fig. 2B). Each copy of the D-ligase generates 8.8 or 5.8 copies of the L-ligase per hour at either 17 or 30 °C, respectively.

Cross-Chiral RNA Replication with Exponential Growth.

The 47.11 cross-chiral ligase was used to drive a cross-chiral replication cycle in which the enantiomeric D- and L-ribozymes undergo cross-catalytic exponential growth. The D1, L1, and L2 fragments all were prepared by chemical synthesis and D2 was prepared by in vitro transcription. Replication was initiated by combining 10 µM of each of the 4 RNA fragments, and the yield of the D- and L-ligase products was measured over time (Fig. 4A).

Fig. 4.

Cross-chiral exponential amplification of an RNA enzyme. (A) Yield over time of the D- and L-47.11 ligases (orange and blue, respectively), with no starting amount of ligase. The data were fit to the logistic growth equation. (B) Initial rate of synthesis of the D- and L-ligases as a function of the initial concentration of the two ligases. The data were fit to the idealized autocatalytic rate equation. Values are based on three replicates, with error bars representing SE (SI Appendix, Fig. S3). Reaction conditions: 0 to 1 µM of each ligase (D- and L-47.11), 10 µM of each fragment (D1, D2, L1, and L2), and 50 mM MgCl₂ at pH 8.7 and 32.5 °C.

During the initial “lag” phase of the reaction, the rate of product formation is slow because it depends on assembly of and catalysis by a quaternary complex composed of all four fragments. As ligase ribozymes are synthesized, however, they can form a ternary complex in which the ligase engages the two fragments of the opposite handedness. This ternary complex reacts more rapidly compared to the quaternary complex of unligated fragments. The rate of synthesis of both the D- and L-ligases then increases progressively over time as more ligases are generated, reaching a maximum after about 2 h. Beyond this point, the reaction rate decreases as the supply of unreacted fragments becomes depleted.

Cross-chiral replication follows the classic model of resource-constrained population growth. The yields of both D- and L-ligases were fit to the logistic growth equation: [E]_t = a/(1 + be^–ct), where [E]_t is the concentration of either ligase at a given time, a is the maximum yield, b is the degree of sigmoidicity, and c is the rate of exponential growth. The synthesis of both the D- and L-ligases fits well to the logistic growth equation (r² > 0.99), with exponential growth rates of 0.014 ± 0.001 and 0.021 ± 0.001 min^–1, respectively (Fig. 4A). This apparent logistic growth is consistent with resource-constrained autocatalysis, but a sigmoidal product yield profile can result from various reaction idiosyncrasies and is not necessarily indicative of true exponential amplification (39).

A kinetic investigation of cross-chiral replication was conducted to differentiate autocatalysis from other reaction profiles. The initial rate of ligase synthesis was determined as a function of the starting concentration of D- and L-ligases. The data were fit to the equation: (d[E]/dt)₀ = (k_auto [E]₀^ρ) + k_spont, where the initial rate, (d[E]/dt)₀, is proportional to the starting ligase concentration, [E]₀, raised to the reaction order ρ. A plot of (d[E]/dt)₀ vs. [E]₀^ρ has a slope equal to the autocatalytic rate constant, k_auto, and a y-intercept equal to the rate of reaction in the absence of a starting amount of ligase, k_spont. Exponential growth occurs when the reaction order is 1.0 (5).

The 27.3t ligase did not exhibit exponential amplification because, for that system, k_auto [E]₀^ρ is approximately equal to k_spont within the relevant concentration regime, thus blunting the effect of increasing amounts of full-length ligase. In contrast, the behavior of the 47.11 ligase is more strongly determined by the autocatalytic rate constant. Varying and equimolar concentrations of the D- and L-ligases were used to initiate cross-chiral replication and the rate of formation of new copies of each ligase was determined during the initial linear phase of the reaction (Fig. 4B and SI Appendix, Fig. S3). Production of the L-ligase consistently outpaced that of the D-ligase over a range of starting concentrations of ligase, which may be due to the D2 component of the D-ligase being prepared by in vitro transcription, which is expected to have higher chemical purity compared to the components of the L-ligase that were both chemically synthesized. Another contributing factor may be that the D-ligases used to initiate cross-chiral replication were prepared by in vitro transcription, whereas the L-ligases were prepared by RNA-catalyzed ligation of chemically synthesized L1 and L2 fragments.

Fitting these data to the autocatalytic rate equation gives values for the reaction order, ρ, of 1.06 ± 0.13 and 0.94 ± 0.21 for the formation of the D- and L-ligase, respectively (Fig. 4B). Within experimental error, these values are consistent with exponential amplification, and the data fit equally well to the idealized autocatalytic rate equation (where ρ = 1.0). When ρ is fixed at 1.0, the fit provides estimates for k_auto of 0.011 ± 0.001 and 0.031 ± 0.002 min^–1 for formation of the D- and L-ligase, respectively. Although one cannot exclude a reaction order slightly below 1.0, the estimated autocatalytic rates when ρ is fixed at 1.0 are highly consistent with the observed autocatalytic rates as determined using the logistic growth equation (Fig. 4A). The autocatalytic efficiency, k_auto/k_spont, is ~5 × 10⁶ M^–1 for both the D- and L-ligase. Thus, with 10 µM substrates, autocatalysis exceeds the spontaneous reaction by ~50-fold, confirming cross-chiral exponential amplification of RNA.

Sustained Cross-Chiral Replication of RNA.

Cross-chiral replication of the ligase ribozyme can be sustained through a serial-transfer process, whereby amplification is allowed to proceed through the exponential phase, then a portion of the reaction mixture is transferred to a new reaction vessel that contains a fresh supply of the D1, D2, L1, and L2 fragments. The first reaction mixture contained 10 µM each of the 4 fragments and was seeded with 0.2 µM each of the D- and L-ligases. The yield of both the D- and L-ligases reached ~2 µM after 5 h, at which point 10% of the reaction mixture was removed and added to a second reaction vessel that contained 10 µM each of the 4 fragments. This process was continued for 6 cycles of growth and dilution, corresponding to an overall dilution of 10⁶-fold after 30 h (Fig. 5). Exponential amplification of both the D- and L-ligases was consistent over the 6 cycles, with yields of 2.0 ± 0.1 and 2.1 ± 0.1 µM, respectively. There was no decrease in yield in the later cycles and it is expected that this process could be continued indefinitely.

Fig. 5.

Sustained cross-chiral exponential amplification. A serial-transfer process was initiated with 0.2 µM each of the D- and L-47.11 ligases, under reaction conditions as in Fig. 4. After 5 h, 10% of the reaction mixture was transferred to a new vessel containing a fresh supply of the 4 fragments. The concentrations of the D- and L-ligases were determined at the beginning and end of each cycle. For clarity, the data for the D- and L-ligases are shifted by 6 min to the right or left, respectively.

Discussion

Although there are many known examples of autocatalytic reaction systems, few such systems are capable of exponential self-replication (23, 25, 40, 41). Living systems are the most profound example of an exponential replication system, with the ability to propagate heritable information across successive generations. Biological self-replicating systems are invariably homochiral, and all prior examples of chemical self-replicating systems are either indifferent to chirality or enantioselective for components of the same chirality (42, 43). The cross-chiral replicator described here demonstrates that chiral uniformity is not an essential condition for exponential self-replication, which can also be achieved by cooperative catalysis among opposing homochiral enantiomers. The cross-chiral replicator grows exponentially and can sustain growth over many successive generations in the face of dilution and degradation. The replicating molecules could potentially encode heritable information if they could adopt alternative compositions, but in the present study, there was no alternative other than to join the four component substrates to yield the D and L forms of the 47.11 ligase.

Two key hurdles must be overcome to achieve exponential growth in a chemical replication system, where a product molecule catalyzes its own formation from component substrates. The first is that the product-catalyzed reaction, as reflected by the autocatalytic rate constant, must substantially outpace the spontaneous reaction between substrates that can occur in the absence of product. When these rates are similar, the formation of new products will not significantly accelerate the overall reaction rate. This limitation cannot be resolved by improving the intrinsic rate of product formation, unless there is differential improvement favoring the autocatalytic rate over the spontaneous rate. This first criterion was not met for earlier forms of the cross-chiral ligase, such as the 16.12t and 27.3t ribozymes (32, 33).

The second hurdle to achieving exponential growth is that product release (or some other intermediate step of the reaction cycle) must not limit catalytic turnover. Nucleic acid catalysts that depend entirely on Watson–Crick pairing for binding the substrates are inherently limited by-product release (13–21). An exception arises in biochemical systems where mechanisms such as strand displacement or segregation of the product strand enable strand separation to not be rate-limiting.

Rate-limiting product release is a general impediment to self-replication that proceeds via templated self-assembly, and has also been observed in peptide replication based on helical coiled-coil interactions, where the interactions between template and substrates persist in the template–product complex (44). Exponential or near-exponential growth in these systems has only been achieved by either destabilizing the templating interactions (23, 45, 46) or by physical separation of the template and product (47). A different way to achieve exponential self-replication, which does not require strand separation, involves the propagative growth of linear or branched structures, with periodic fragmentation to generate new “seeds” that initiate further propagative growth (40, 48–50).

The cross-chiral replication system provides an alternative solution because it necessarily avoids Watson–Crick pairing between catalyst and substrates due to the inability of nucleic acids of the opposite handedness to form contiguous base pairs (30, 31). Instead, the substrates are bound through cross-chiral tertiary interactions within the active site of the ribozyme. An enzyme active site can preferentially bind the transition state intermediate and thus favor catalytic turnover. Multiple-turnover catalysis has been observed with an in vitro evolved RNA ligase protein that, like the cross-chiral ligase ribozyme, can only bind RNA through tertiary interactions (51). The mechanism of product release has not been determined for the cross-chiral ligase, but the prior 27.3t ribozyme has poor catalytic turnover when synthesizing its enantiomer, generating less than one product molecule per hour (Fig. 2B), which is insufficient to drive autocatalysis.

Two distinct approaches were taken in this study to enhance catalytic turnover of the ribozyme, facilitating the development of a self-replication system that exhibits exponential growth. Unlike most directed evolution procedures, which involve selection based on a single chemical event, the first 11 rounds of directed evolution required the cross-chiral ribozyme to catalyze up to 10 successive ligation reactions (SI Appendix, Fig. S1 and Table S1). Because these reactions took place along a common RNA template, however, the ligase did not need to fully disengage from the product before entering the next catalytic cycle, as would be necessary during multiple-turnover catalysis. Therefore, in the 9 subsequent rounds of directed evolution, the ligase was made to operate in true intermolecular fashion, with clonal isolation of individual variants within the compartments of a water-in-oil emulsion (Fig. 3). This approach has been used successfully to screen for other biochemical traits (52–54), but in this instance was used to select for catalytic turnover in the first half-reaction of the cross-chiral replication cycle, requiring the evolving D-ligases to generate multiple copies of a functional L-ligase.

Compartmentalized selection proved essential to improving catalytic turnover of the ribozyme, although mutations that were prevalent in the final evolved population were already present to a lesser degree following the noncompartmentalized phase of evolution (SI Appendix, Fig. S2). The catalytic activity of the population as a whole improved by only 2.5-fold, but individual ribozymes isolated from the population showed more substantial improvement. After the final round of evolution, population diversity remained high (SI Appendix, Fig. S2 and Table S3), suggesting that further evolution with higher selection stringency might lead to further improvement of catalytic turnover.

The evolved 47.11 ligase is an efficient replicator (Fig. 2B), capable of cross-chiral replication with exponential growth. A common way to assess autocatalytic behavior is to measure the progressive increase in the rate of reaction over time, until the supply of substrates starts to become depleted (Fig. 4A). The characteristic logistic growth profile is indicative, but not necessarily proof, of exponential autocatalysis because biphasic kinetics and other complexities can give rise to similar reaction profiles (39). A more formal way to assess autocatalysis is to measure the initial rate of reaction as a function of the starting concentration of catalyst (5). At the outset of the reaction, the rate is highly dependent on the starting concentration of catalyst. Within the bounds of experimental error, the initial rate appears to be directly proportional to the starting concentration of the 47.11 ribozyme, indicative of true exponential growth (Fig. 4B). The autocatalytic rate determined in this manner closely matches that obtained from the logistic growth equation (Fig. 4A), with values of 0.011 and 0.014 min^–1, respectively, for D-RNA synthesis, and 0.031 and 0.021 min^–1, respectively, for L-RNA synthesis. The close congruence of these values provides further support for exponential growth, with a doubling time of 50 to 63 min.

Exponential amplification confers a type of molecular immortality because, so long as one provides on ongoing supply of substrates, the replicating molecules can be maintained indefinitely against environmental dilution or degradation (3–10). A practical way to achieve ongoing amplification is through a serial-transfer procedure, as was applied to the 47.11 ligase in carrying out six successive rounds of 10-fold amplification, resulting in an overall amplification of 10⁶-fold (Fig. 5). This process could be continued for hundreds of rounds to achieve astronomical amplification factors, although the result would be monotonous when using only a single pair of cross-replicating ligases and their constitutive fragments. If, however, there were multiple replicators competing for common resources, then differences in the autocatalytic rate constants would lead to enrichment or depletion of particular replicators over time, a form of chemical natural selection.

Darwinian evolution in a cross-chiral replication system would further require a form of molecular heritability, where the specific composition of the various replicators is carried forward to products with the same, or nearly the same, composition. Heritable information would need to be transmitted across the mirror through the reciprocal catalytic activities of D- and L-ligases. This mode of information transfer is very different from heredity in biology, which relies on the simplicity of Watson–Crick pairing and semiconservative replication. Instead, information must be transferred across a cross-chiral reaction network via the three-dimensional interface of enzyme–substrate interactions that bring together molecules of the opposite handedness. The D and L forms of a cross-replicating pair need not have the same composition, only the ability to maintain a defined composition of both members of the pair through the replication cycle (55). Another ribozyme that was identified in this study, 47.4, differs from 47.11 by 14 mutations but has similar cross-chiral amplification activity (SI Appendix, Table S3), supporting the potential for more diverse populations of cross-chiral replicators.

Informational complexity could expand combinatorially by increasing the number of D- and L-fragments within a replicator network. This combinatorial expansion could be achieved if the cross-chiral ligases were assembled from more than two fragments, or if the replicator network expanded to include other D- and L-ribozymes that increase the functional diversity of the system. Such increases in combinatorial complexity would require corresponding increases in the catalytic selectivity of the cross-chiral ligases that drive replication, exploiting the broad capacity of D- and L-RNAs to bind to each other via distinct tertiary interactions (56, 57). At a sufficient level of complexity, functional adaptations might arise among populations of D- and L-ribozymes that evolve cooperatively.

Darwinian evolution may have emerged through autocatalytic reaction networks involving nucleic acids, peptides, or other prebiotic molecules, but it seems unlikely that cross-chiral RNA replication had a role in the early history of life on Earth. There would need to be some mechanism for maintaining supplies of both D- and L-substrates that provide the basis for replication, and some explanation for how a cross-chiral system gave rise to the homochiral system of extant biology. Instead, cross-chiral replication should be considered for its potential to integrate functions across genetic molecules of opposite chirality. Cross-chiral genetic systems could exploit a rapidly expanding catalog of functional roles for D- and L-RNA, using exponential growth for target-dependent signal amplification (28, 58, 59) and nonbiological L-RNAs to provide orthogonality to biological processes (60, 61). These systems would also represent an entirely artificial approach to genetic self-organization that exploits chirality in a manner that has not, and perhaps could not, be explored in the natural world. Life as we know it is homochiral, but life as it might be realized in the laboratory need not be considered so narrowly.

Materials and Methods

Materials.

The sequences of all oligonucleotides used in this study are listed in SI Appendix, Table S4. Oligonucleotides were either purchased from IDT, prepared by in vitro transcription (SI Appendix, SI Methods), or synthesized in-house using an Expedite 8909 DNA/RNA synthesizer, with reagents and phosphoramidites from either Glen Research or Chemgenes. All RNAs were purified by denaturing polyacrylamide gel electrophoresis (PAGE) prior to use. D1 and L1 fragments for cross-chiral ligation were prepared synthetically and 5′-labeled with sulfo-cyanine5 and fluorescein, respectively. Unless stated otherwise, D2 and L2 fragments were prepared synthetically and chemically 5′-triphosphorylated, as previously described (35–37). Hot Start Taq and Q5 high fidelity DNA polymerase were from New England Biolabs. Superscript IV reverse transcriptase, MyOne C1 streptavidin magnetic beads, and MyOne carboxylic acid magnetic beads were from Thermo Fisher Scientific. Streptavidin beads were washed according to the manufacturer’s instructions and preblocked with 1 mg/mL tRNA prior to use. Emulsification oil DMF-A-6CS and surfactants KF-6012 and KF-6038 were from Shin-Etsu Chemical. All other chemical reagents were from Sigma-Aldrich.

Cross-Chiral Synthesis of Ligase Ribozymes.

Reactions with the 4 fragments alone used 5 µM each of D1, D2, L1, and L2 in the presence of 50 mM MgCl₂ and 25 mM EPPS (pH 8.3) at various temperatures for 20 h, and were quenched by diluting 17-fold with a solution of 95% formamide and 3.5 mM EDTA (pH 8.0). Measurements of k_obs were conducted during the initial linear phase of the reaction using 0.05 µM D-ligase and 10 µM each of L1 and L2, under conditions as above. The reaction products were separated by PAGE and quantitated using an Amersham Typhoon RGB laser scanner with ImageQuant TL software. The data were fit by linear regression using GraphPad Prism.

Directed Evolution of Sequential Ligation Activity.

A starting pool of DNA templates was prepared based on the 27.3t ligase sequence, introducing random mutations at a frequency of 10% per nucleotide position, flanked by fixed primer binding sites. The DNAs were made double-stranded using SuperScript IV in a reaction supplemented with 1.5 mM MnCl₂ to improve polymerase extension through lesions that can occur during DNA synthesis (62). The double-stranded DNAs were ethanol precipitated, then amplified by PCR using Hot Start Taq polymerase to introduce the T7 RNA polymerase promoter sequence, followed by in vitro transcription using T7 RNA polymerase. The D-RNAs were purified by PAGE, covalently attached to an L-RNA primer via a photocleavable linker (32), and again purified by PAGE.

The first round of evolution reported in this study, designated round 28, is numbered in continuity with 27 rounds of evolution carried out in two prior studies (32, 33). Round 28 was initiated with 4 nmols of D-RNAs, corresponding to ~4 copies each of 6 × 10¹⁴ distinct sequences. Subsequent rounds were initiated with 0.2 to 0.5 nmols of D-RNAs, prepared as above from the DNA output of the preceding round. In each round, the L-RNA primer was annealed to an L-RNA template together with the 5′-triphosphorylated L-RNA substrates, and cross-chiral ligation was carried out in the presence of 200 mM MgCl₂, 250 mM NaCl, and 50 mM Tris-HCl (pH 8.3) at 17 °C. Round 28 used 1 µM D-RNA enzyme with attached L-RNA primer, 2 µM L-RNA template, and 1 µM of each L-RNA substrate. In rounds 29 to 38, no NaCl was added and the MgCl₂ concentration was reduced to 100 mM; and in rounds 31 to 38, the concentration of L-RNA substrates was increased to 10 µM. The reaction times for each round of evolution are listed in SI Appendix, Table S1 and the sequences of the 5′-triphosphorylated L-RNA substrates are listed in SI Appendix, Table S4.

In rounds 28 to 30, D-RNA enzymes were selected based on 3′-terminal incorporation of a biotinylated L-RNA substrate. The full-length reaction products were captured and amplified, as previously described (32). In rounds 31 to 33, D-RNA enzymes were selected based on synthesis of the L1 fragment of the 16.12t ligase from component L-RNA substrates. The L-RNA template was biotinylated, enabling capture of the template-bound products on streptavidin beads, which were washed twice with saline solution [1 M NaCl, 1 mM EDTA, 10 mM Tris-HCl (pH 8.0), and 0.05% Tween-20]. The products then were eluted with alkaline solution (25 mM NaOH, 1 mM EDTA, and 0.05% Tween-20), ethanol precipitated, and the full-length products were isolated by PAGE.

In rounds 34 to 38, D-RNA enzymes were selected based on the catalytic activity of the synthesized L1 fragment of the 16.12t ligase when supplied with the matching L2 fragment. Following round 33, nested PCR was used to introduce a substrate at the 5′ end of the D-RNAs. The D-RNAs were challenged to catalyze synthesis of L1 and the reaction products were recovered by capture and release from the biotinylated L-RNA template as in rounds 31 to 33. 1 µM L-RNA products linked to the D-RNA enzymes responsible for their synthesis were incubated together with 1 μM L2 fragment, 2 µM D-RNA template, 1 µM biotinylated D-RNA substrate, 200 mM MgCl₂, 250 mM NaCl, and 50 mM Tris-HCl (pH 8.3) at 17 °C for 1 h. The ligated D-RNA products were captured on streptavidin beads, which were washed twice with saline solution, once with alkaline solution, and once with a solution of 8 M urea, 1 mM EDTA, and 10 mM Tris-HCl (pH 8.0). Then the RNAs were eluted from the beads in formamide solution [95% formamide, and 10 mM EDTA (pH 8.0)] at 95 °C for 5 min, ethanol precipitated, and amplified by RT-PCR.

Compartmentalized Evolution of Cross-Chiral Ligases.

DNAs from round 38 were modified at both ends using PCR primers, introducing 5′-biotin within the sense strand. The PCR products were captured on streptavidin beads, the nonbiotinylated cDNAs were removed by washing 3 times with alkaline solution, and the biotinylated strands were eluted with formamide solution and purified using the PureLink PCR Purification Kit (Thermo Fisher Scientific). The DNAs were then captured at a loading ratio of one DNA per bead by hybridization to magnetic beads derivatized with primers that are complementary to the 3′ end of the DNA. The bead-bound DNAs were clonally amplified by emulsion PCR, yielding ~3,000 double-stranded DNAs per bead (SI Appendix, SI Methods). The beads were removed from the emulsion, transferred to an in vitro transcription mixture, again clonally emulsified, and incubated at 37 °C for 10 to 60 min to yield clonal copies of RNA that also hybridized to the bead-conjugated primers. The number of RNAs captured per bead was controlled by varying the transcription time and number of primers per bead (SI Appendix, SI Methods). The second emulsion was broken and the L2 fragment of 27.3t RNA was conjugated to the beads, with variable loading determined quantitatively (SI Appendix, SI Methods). The sulfo-cyanine5-labeled L1 fragment of 27.3t was hybridized to the unlabeled, bead-bound L2 fragment at a ratio of 0.9:1 in a solution containing 50 mM MgCl₂, 100 mM NaCl, 50 mM EPPS (pH 8.3), and 0.05% KF-6012, which was agitated at 23 °C for 5 min.

RNA-catalyzed ligation was initiated by adding 1% (by volume) of a 1 mM solution of the 10-23 DNAzyme (63), which cleaves a specific RNA linkage to release the D-ligases from the beads. The beads were immediately isolated by emulsification, the DNAzyme was activated by heating to 37 °C for 10 min, and the released ligases were refolded by incubating at 55 °C for 10 min, then cooling to room temperature. The ligation reaction was carried out at various temperatures and for various times (SI Appendix, Table S2), then the emulsion was broken. The beads were washed twice with alkaline solution to remove unligated L2 fragments, transferred to a fresh tube to remove beads that were bound to the surface of the first tube, washed again with alkaline solution, and resuspended in 137 mM NaCl, 2.7 mM KCl, 10 mM Na₂HPO₄ (pH 7.4), and 0.05% Tween-20. The brightest 0.5% of beads were collected using an Aria Fusion cell sorter, then the bead-bound DNAs were PCR amplified and the PCR products were purified using the PureLink PCR Purification Kit. Random mutations were introduced by error-prone PCR (38), and the resulting double-stranded DNAs were used as input for the next round of evolution.

Deep Sequencing of Evolved Ligases.

DNAs from rounds 38 and 44 to 47 were minimally amplified by PCR using Q5 high fidelity DNA polymerase, purified using the PureLink PCR Purification Kit, and sequenced using the Amplicon-EZ service (Genewiz) for paired-end sequencing. Sequence trimming (64) and mutation frequency calculations were performed according to a published protocol (65). Sequence reads from rounds 44 to 47 were error-corrected, merged, and compiled using DADA2 (66), and sequence space was visualized in R using a custom script. Briefly, sequences with a summed frequency >0.1% were plotted according to the first 2 principal components of the Levenshtein distance matrix (44 and 36% of the variance, respectively). Sequences within regions of high density of the PCA plot were aligned to the 27.3t sequence using Benchling and manually assessed for diversity, abundance, and enrichment.

Evaluation of Individual Ligase Clones.

Twelve candidate ligases were prepared individually by assembly PCR followed by in vitro transcription (SI Appendix, SI Methods). Ligation was carried out in a reaction mixture containing 25 nM D-ligase, 2.5 µM each of the L1 and L2 fragments of 27.3t, 50 mM MgCl₂, 100 mM NaCl, and 50 mM EPPS (pH 8.3) at 30 °C for 4 h, then the products were analyzed by PAGE. For 5 of the most active ligases, the D1, D2, L1, and L2 fragments were prepared and autocatalysis was assessed by conducting reactions with 2 µM each of the 4 fragments in either the presence or absence of 0.1 µM full-length D-RNA ligase, comparing the yields at 1 h.

Cross-Chiral Replication of the 47.11 Ligase.

The RNA components were preincubated at 32.5 °C for 3 min, then mixed with a prewarmed solution to give a reaction mixture containing 10 µM of each RNA fragment, various and equal concentrations of D- and L-47.11, 50 mM MgCl₂, 10 mM EPPS (pH 8.7), and 0.05% Tween-20, which were incubated at 32.5 °C. Aliquots were taken from the reaction at various times, quenched by diluting 17-fold with a solution of 95% formamide and 3.5 mM EDTA (pH 8.0), then analyzed by PAGE. Initial velocity was determined over the first 20 min of the reaction when the yield was <15% of the maximum extent, and the data were fit using linear regression. Serial transfer experiments entailed 10-fold dilution of each completed reaction mixture into a fresh mixture containing 11 µM of each RNA fragment, 56 mM MgCl₂, 11 mM EPPS (pH 8.7), and 0.056% Tween-20.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

Sequencing data, initial velocity data, bioinformatics pipeline, and software script data have been deposited in the Dryad Digital Repository (https://doi.org/10.5061/dryad.0vt4b8h6w) (67).