Deep sequencing of non-enzymatic RNA primer extension

Abstract

Life emerging in an RNA world is expected to propagate RNA as hereditary information, requiring some form of primitive replication without enzymes. Non-enzymatic template-directed RNA primer extension is a model of the copying step in this posited form of replication. The sequence space accessed by primer extension dictates potential pathways to self-replication and, eventually, ribozymes. Which sequences can be accessed? What is the fidelity of the reaction? Does the recently illuminated mechanism of primer extension affect the distribution of sequences that can be copied? How do sequence features respond to experimental conditions and prebiotically relevant contexts? To help answer these and related questions, we here introduce a deep-sequencing methodology for studying RNA primer extension. We have designed and vetted special RNA constructs for this purpose, honed a protocol for sample preparation and developed custom software that analyzes sequencing data. We apply this new methodology to proof-of-concept controls, and demonstrate that it works as expected and reports on key features of the sequences accessed by primer extension.

Figure 1.

Protocol for preparing RNA hairpin constructs for sequencing. (A) NERPE-Seq RNA hairpin constructs contain a hairpin loop that connects the template to the primer so that the product of non-enzymatic primer extension and the corresponding template are on one continuous RNA strand. Two caged bases (magenta Xs) prevent primer extension from encroaching on the downstream 5′ Handle (Supplementary Figure S3). The 5′ Handle Block is complementary to the 5′ Handle and prevents it from interfering with primer extension (Supplementary Figure S8). (B) The primer extension reaction is quenched with a desalting size-exclusion spin column, the caged bases are uncaged and the target RNA is further gel purified. (C andD) The pre-adenylated DNA RT Handle (blocked on its 3′ end to prevent self-ligation) is ligated to the 3′ end of the RNA hairpin (the site of primer extension) (Supplementary Figure S4). (E) The ligase is removed by a Proteinase K digestion, the target RNA–DNA is phenol–chloroform extracted, and the RT primer is annealed to the RT Handle. (F andG) RT generates the cDNA (Supplementary Figure S5); the RNA is degraded, and the cDNA is isolated with a spin column. The ROI harbors the template, hairpin and any product sequences. (H) PCR is used to barcode the DNA and add flanking sequences (Supplementary Figure S6). Each barcode identifies DNA from a specific experiment and enables the sequencing of samples from multiple experiments at the same time. (I) The target PCR products are purified, and validated by automated electrophoresis and quantitative PCR prior to sequencing (Supplementary Figure S7).

Figure 2.

Sequencing a Construct that Mimics Full-length Non-enzymatic Primer Extension. (A) The hairpin construct CTEx was designed to mimic an efficient non-enzymatic primer extension reaction. The RNA was converted into cDNA using the optimized protocol. The ROI includes the template and ‘product’ sequences. After exposure to mock primer extension conditions and desalting, CTEx ran as a well-defined band at the expected position in PAGE. After uncaging and additional purification, the RT Handle was ligated to the 3′ end with very high efficiency. The product was purified and reverse transcribed, yielding a distinct DNA band, which was then isolated. (B) The NERPE-Seq protocol accurately measures the CTEx template and ‘product’ sequences without the template or product identities being provided during analysis. The heat maps show the frequency of each base at the indicated position.

Figure 3.

Data Analysis. (A) Cartoon of a hairpin construct after RT Handle ligation. The labels and color coding indicate the various sequence regions used during data processing. The hairpin and handles are defined sequences (‘fixed’), the Prefix is a four-base motif with the two caged bases, the Template is of defined length but the analysis does not specify a defined sequence, and the Product is of indeterminate length and sequence. (B) The double-stranded DNA generated by barcoding PCR (the barcode is to the 3′ of Fix 2 and not shown). The final location of each region from the construct is labeled and color-coded. Paired-end sequencing provides both the forward (R1) and reverse (R2) sequences, which are compared against each other for quality control. A series of checks identifies the fixed sequences, filters out low-quality reads, and extracts the Template and Product. (C) The end result of Pre-processing is a set of Template-Product pairs. Unextended bases are indicated by an asterisk, and a placeholding A is included as an internal marker. (D) Template-Product pairs are queried in the Characterize stage to assay the sequence properties of templates and complementary products, and indicate the positions and contexts of mismatches.

Figure 4.

NERPE-Seq Can Accurately Measure Primer Extension on a Defined Template. (A) CTB Mimic, used in PAGE analysis, is a construct that mimics the CTB hairpin (B). (C) The extent of primer extension with 20 mM each of 2AIrG and 2AIrC after 24 h using CTB Mimic, as measured by PAGE analysis. (D) The same reaction as in (C), but measured by NERPE-Seq on CTB. (Number of read pairs prior to filtering = 1.8 × 10⁶.) (E) The extent of primer extension with 500 μM 2AI-CGG after 24 h using CTB Mimic, as measured by PAGE analysis. (F) The same reaction as in (E), but measured by NERPE-Seq on CTB. (Number of read pairs prior to filtering = 1.6 × 10⁶.) (G) NERPE-Seq reveals the expected pattern of products from the polymerization reaction (same experiment as in D) and (H) from the non-enzymatic ligation reaction (same experiment as in F). The heat maps show the frequency of each base at each position, including nulls. Note in (G) the mismatched over-extension of G and C products across the templating A in position 4. In (H) the relative intensities at each position are equivalent because all products result from trimer ligation (i.e.: all ligation events contribute equivalently to the +1, +2 and +3 positions). Finally, the overall intensities in (G) are higher than in (H) because primer extension by polymerization is in this case more efficient than ligation.

Figure 5.

NERPE-Seq can measure mismatch frequencies. (A) The extent of primer extension on a 3′-GCC-5′ template incubated with 20 mM 2AIrU for 24 h as measured by PAGE analysis. (B) The same experiment as in (A), but measured by NERPE-Seq. (Number of read pairs prior to filtering = 1.6 × 10⁶.) (C) NERPE-Seq reveals the expected pattern of products. The heat map shows the proportion of each base at each position, including nulls. Note the intensity scale bar maximum value is set to 0.1 because there are relatively few products. (D) The position-dependent distribution of all possible mismatches. The heat map shows the proportion of each mismatch at each position relative to all mismatches in that position. As expected, the Template:Product (T:P) mismatch of G:U in the first position dominates the data, and C:U is also evident in the second position. The data become noisy even by position 2 because there are very few extension events beyond +1 (>98% of the data is in position 1).