Competition between bridged dinucleotides and activated mononucleotides determines the error frequency of nonenzymatic RNA primer extension

Abstract

Nonenzymatic copying of RNA templates with activated nucleotides is a useful model for studying the emergence of heredity at the origin of life. Previous experiments with defined-sequence templates have pointed to the poor fidelity of primer extension as a major problem. Here we examine the origin of mismatches during primer extension on random templates in the simultaneous presence of all four 2-aminoimidazole-activated nucleotides. Using a deep sequencing approach that reports on millions of individual template-product pairs, we are able to examine correct and incorrect polymerization as a function of sequence context. We have previously shown that the predominant pathway for primer extension involves reaction with imidazolium-bridged dinucleotides, which form spontaneously by the reaction of two mononucleotides with each other. We now show that the sequences of correctly paired products reveal patterns that are expected from the bridged dinucleotide mechanism, whereas those associated with mismatches are consistent with direct reaction of the primer with activated mononucleotides. Increasing the ratio of bridged dinucleotides to activated mononucleotides, either by using purified components or by using isocyanide-based activation chemistry, reduces the error frequency. Our results point to testable strategies for the accurate nonenzymatic copying of arbitrary RNA sequences.

Graphical Abstract

Correct nucleotide incorporation in nonenzymatic RNA primer extension can be traced to a bridged dinucleotide mechanism, whereas mismatched incorporations arise from the direct reaction of activated mononucleotides.

Figure 1.

Deep sequencing nonenzymatic RNA primer extension. (A) The RNA hairpin construct primes itself and has a six-base random-sequence template (34). The caged bases prevent primer extension from encroaching on downstream regions required for processing. Each ‘caged’ base harbors an NPOM modification that can be removed with long-wavelength UV to generate a native dT (34,61). The 5′ Handle is required for downstream PCR, and the complementary 5′ Handle Block prevents it from interacting with the template. (B, C) The primer is extended in the presence of all four 2-aminoimidazole-activated monoribonucleotides. The nucleotides form a reactive 5′-5′ bridged dinucleotide intermediate required for efficient copying (24). Products have both correct (complementary) and incorrect (mismatched) incorporations. (D) Deep sequencing reports on millions of individual product-template pairs.

Figure 2.

The bridged dinucleotide intermediate determines complementary product sequences. (A) Frequencies of complementary and mismatched nucleotide incorporations (20 mM 2AIrN, 24 h; n = unextended hairpins + total nucleotide incorporation events). (B) Position-dependent base frequencies of complementary products. (C) Inferring bridged dinucleotide identities. The first nucleotide adjacent to the primer becomes incorporated, whereas the downstream second nucleotide functions as a leaving group and diffuses away. (D) Position-dependent frequencies of inferred bridged dinucleotides that participated in generating complementary products.

Figure 3.

Inferred bridged dinucleotide frequencies and complementary product sequence space. (A) A log-linear correlation between predicted equilibrium binding constants and inferred bridged dinucleotide frequencies (20 mM 2AIrN, 24 h incubation, position 1; least squares unconstrained linear fit, R² = 0.76, dashed lines indicate 95% confidence interval on the fit). The Nearest Neighbor Database (NNDB, (43)) was used to calculate predicted equilibrium binding constants for each stretch of priming base, correctly incorporated base, and downstream base of a bridged dinucleotide. These values are shown plotted against the measured frequencies of inferred bridged dinucleotides at position 1 (Figure 2D). (B) The sequence space of primer extension with 20 mM 2AIrN, 24 h incubation. The volume of each sphere is proportional to the frequency of each complementary product at least three bases long, beginning at position 1. Note that the frequencies of some triplets are too low to be visualized (Supplemental Table S3). (C) rG and rC templates are used up by the more reactive bridged dinucleotides, leaving behind higher proportions of rA and rU templates. (D) More reactive inferred bridged dinucleotides participating in primer extension become less frequent with time as less reactive inferred bridged dinucleotides—dominated by combinations of rA and rU—become slightly more frequent.

Figure 4.

Activated mononucleotides are responsible for mismatches. (A) The position-dependent frequency of each mismatch, relative to all mismatches at the indicated position (20 mM 2AIrN, 24 hours; T:P = Template:Product). (B1–3) Mismatches could originate from bridged dinucleotides (B1), or the incorporation of activated mononucleotides (B2). (B3) The templating base distribution downstream of mismatches can be ascertained from the sequencing data. (C) The templating base distribution downstream of position 1 mismatches. (D) The overall mismatch frequency depends on the ratio of activated mononucleotides to bridged dinucleotides (increasing across experiments left to right). At the time of peak bridged dinucleotide concentration, the ratio ≅ 14 for 20 mM 2AIrN (Supplementary Figure S2C), ≅ 50 for 20 mM 2AIrN + 100 mM AI (22), and bridged dinucleotides do not form with 20 mM OAtrN.

Figure 5.

Advantages of a prebiotically plausible bridge-forming activation chemistry. (A) Frequencies of complementary and mismatched nucleotide incorporations (10 mM 2AIrN + MeNC-mediated bridge-forming activation, 24 h; n = unextended hairpins + total nucleotide incorporation events). Product yields with bridge-forming activation but only 10 mM 2AIrN are comparable to that with 20 mM 2AIrN but no activation chemistry (Figure 2A), and the products are less error-prone. Other features, including bridged dinucleotide (B) and mismatch patterns (C) remain the same.