Expanding the Nucleoside Recoding Toolkit: Revealing RNA Population Dynamics with 6-thioguanosine

Abstract

RNA-sequencing (RNA-seq) measures RNA abundance in a biological sample but does not provide temporal information about the sequenced RNAs. Metabolic labeling can be used to distinguish newly made RNAs from pre-existing RNAs. Mutations induced from chemical recoding of the hydrogen bonding pattern of the metabolic label can reveal which RNAs are new in the context of a sequencing experiment. These nucleotide recoding strategies have been developed for a single uridine analogue, 4-thiouridine (s⁴U), limiting the scope of these experiments. Here we report the first use of nucleoside recoding with a guanosine analogue, 6-thioguanosine (s⁶G). Using TimeLapse sequencing (TimeLapse-seq), s⁶G can be recoded under RNA-friendly oxidative nucleophilic-aromatic substitution conditions to produce adenine analogues (substituted 2-aminoadenosines). We demonstrate the first use of s⁶G recoding experiments to reveal transcriptome-wide RNA population dynamics.

Graphical Abstract

The transcriptome is in constant flux between RNA transcription, processing, and decay, but standard RNA-seq experiments only provide a static snapshot of the cellular RNA levels. Metabolic labeling offers a chemical strategy to distinguish RNAs made over different time periods, but traditionally requires biochemical isolation of the metabolically labeled RNA.¹ We and others have developed nucleotide conversion chemistry to study RNA population dynamics transcriptome-wide in an enrichment-free RNA-seq experiment.^2–4 In these experiments, living cells are treated with s⁴U, which is incorporated into newly transcribed RNAs. Total RNA is isolated and either alkylated to produce uridine analogues with a single recoded hydrogen bond (SLAM-seq³) or reacted under oxidative-nucleophilic-aromatic substitution conditions to fully recode s⁴U into cytosine² or cytosine analogues (TimeLapse-seq⁴). After these treatments, newly transcribed RNA can be distinguished from pre-existing RNA by apparent thymidine to cytosine (T-to-C) mutations in sequencing reads, thereby adding a temporal dimension to RNA-seq experiments.

Methods that use nucleotide recoding to study RNA population dynamics are currently limited to a single pyrimidine metabolic label, s⁴U, making them suboptimal or inappropriate for several applications such as studying the turnover of uridine-tailed RNAs⁵, pseudouridylated RNAs⁶, and uridine-poor RNAs. As TimeLapse with s⁴U is similar to convertible nucleoside chemistry developed to post-synthetically modify oligonucleotides^7,8 we were inspired by similar convertible nucleoside approaches that affectively allowed the recoding of purines⁹. Therefore, we sought to expand the scope of recoding chemistry to a purine nucleotide.

Early studies of metabolic labels suggest several nucleotides can be incorporated into the transcriptome. ^10,11 While the most widely used metabolic labels to study RNA population dynamics are uridine analogues, s⁶G has frequently been employed in photocrosslinking experiments (PAR-CLIP¹²) and RNA structural studies¹³. We reasoned that RNA-friendly oxidative-nucleophilic-aromatic substitution (TimeLapse chemistry) that we previously developed for s⁴U could be extended to recode s⁶G to 2-aminoadenosine analogues, thereby expanding the toolkit of recodable RNA metabolic labels. Here we report the development of this chemical approach and demonstrate the first use of a guanosine-based metabolic label to measure RNA population dynamics (Scheme 1).

Scheme 1.

Overview of s⁶G-based TimeLapse-seq to reveal RNA population dynamics.

We hypothesized that the oxidative nucleophilic-aromatic-substitution conditions that we developed for s⁴U TimeLapse (NaIO₄ and 2,2,2-trifluoroethylamine, TFEA)⁴ could also convert thiolated guanine into N6-substituted analogues of adenine. These conditions were previously optimized to minizmize oxidation of guanine and preserve compatibility with RNA-seq. These reagents led to clean conversion of 6-thioguanine to N6-trifluoroethyl substituted 2,6-diaminopurine by ¹H-NMR (see Supporting Information (SI)). We found by a LC-MS time-course that 6-thio-2’-deoxyguanosine (s⁶dG) nucleoside was consumed within 5 min and the corresponding 2-aminoadenosine analogue (hereafter referred to as A*) was produced within 1h (Figure 1A).

Figure 1.

(A) Extracted ion chromatograms corresponding to masses of of 6-thio-2’-deoxyguanosine and 2-amino-6-(2,2,2-trifluoroethyl)-amino-2’-deoxyadenosine. (B) Restriction digest assay of a DNA duplex containing a single 6-thio-2’-deoxyguanosine, treated with 600 mM NH₃ and 10 mM NaIO₄ for 1 h at 45°C and subsequent digestion by SspI restriction enzyme (AATATT) and analysis by Native-PAGE. (C) Quantification of N to A transitions in ACTB mRNA sequencing reads. K562 cells were treated with 500 µM 6-TG for 1 h. Extracted RNA was treated with 600 mM TFEA and 10 mM NaIO₄ for 1 h at 45°C.

From our previous work with s⁴U, we determined that recoding efficiencies of about 50% or greater are sufficient for monitoring RNA population dynamics.⁴ To assess the recoding efficiency of s⁶G in the context of an oligonucleotide, we chose to study a DNA oligonucleotide because of documented challenges incorporating s⁶G into RNA using prokaryotic RNA polymerases for in vitro transcription.¹⁴ We used an in vitro restriction digest assay with a single s⁶dG positioned at a site that creates an endonuclease restriction site upon successful recoding to dA* and PCR amplification. The majority (~59%) of the nucleotide s⁶dG was recoded to dA* in the context of a DNA duplex using ammonia in TimeLapse chemistry (Figure 1B). Similar results were obtained using different amines, including TFEA (see SI). These in vitro results led us to test whether we could use s⁶G to reveal RNA population dynamics of cellular RNAs.

We explored conditions to metabolically label cellular RNAs with s⁶G. While many photo-crosslinking studies use s⁶G nucleoside as a metabolic label¹², we investigated the incorporation of either s⁶G nucleoside or 6-thioguanine (6-TG) nucleobase into cellular RNA and found similar incorporation (see SI). Due to the solubility of 6-TG, subsequent experiments were performed using 6-TG treated cells.

To test if we could use nucleotide conversion chemistry with s⁶G to examine the dynamics of cellular RNAs, we treated human K562 cells with 6-TG. The cells were grown for 1 h to allow time for incorporation of s⁶G into newly synthesized RNA. We did not observe significant toxicity even after 2 h treatment (see SI), consistent with previous reports for s⁶G.¹² Total RNA was then isolated and subjected to TimeLapse chemistry, followed by targeted reverse transcription (ACTB mRNA, see SI) and next generation sequencing. Sequencing reads were mapped to the target transcript and the mutations of each nucleotide to adenosine were counted. We found that s⁶G is incorporated into newly transcribed RNA and consverted into A* as inferred from the increase in G-to-A mutations at all G nucleotides that were analyzed (Figure 1C and see SI). This conversion was 6-TG treatment and TimeLapse chemistry-dependent.

Having established the viability of using 6-TG to label cellular RNAs, we tested whether we could use 6-TG with TimeLapse chemistry to reveal RNA population dynamics transcriptome wide. We labeled K562 cells for 4 h, a labeling time optimized for studying the half-lives of mRNAs.¹ Next we extracted total RNA from human K562 cells, treated the RNA with TimeLapse chemistry and subjected it to sequencing. Once the reads were mapped to the transcriptome we tested whether the 6-TG treatment substantially impacted RNA levels. Expression analysis revealed that the RNA levels from 6-TG treated cells were highly correlated with the levels from untreated and s⁴U-treated cells, indicating that the process of metabolic labeling does not substantially impact the transcriptome in the timeframe of the experiment (see SI, s⁶G vs. s⁴U Pearson’s r ≥ 0.96; s⁶G vs. untreated Pearson’s r ≥ 0.95). Analysis of the sequencing reads revealed a notable increase in G-to-A mutations (Figure 2A). The increase in G-to-A mutations tracked with transcript half lives (as determined using traditional s⁴U TimeLapse-seq), with transcripts that have the shortest half-lives demonstrating the greatest increase in G-to-A mutations (Wilcox test, p < 10⁻¹⁵; Figure 2B). Even transcripts with long half lives had a significant increase in G-to-A mutations upon 6-TG treatment (Wilcox test, p < 10⁻¹⁵). These trends were also apparent when examining individual transcripts (Figure 2C); transcripts with short half lives such as JUN had higher numbers of reads with G-to-A mutations than stable transcripts such as GAPDH (Figure 2C and see SI). As with targeted sequencing, this increase in the G-to-A mutation rate was only found in RNA from cells that had been treated with 6-TG.

Figure 2.

(A) Metabolic labeling of cellular RNAs leads to G-to-A mutations at sites of s⁶G in sequencing reads shown for a region of the EGR1 transcript. (B) The distribution of average G-to-A mutation rate for each transcript separated by half life quantile (calculated from validated s⁴U TimeLapse-seq, 1 = high turnover, 10 = low turnover compared between transcripts from cells treated with 6-TG with identically treated RNA from untreated cells. **** p < 0.0001 based on a two sided Wilcox rank sum test. (C) Genome browser tracks for representative fast (JUN), intermediate (SMG5 and EIF2S1) and slow (GAPDH) turnover transcripts colored by the cumulative number of G-to-A mutations. Gray tracks represent all RNA-seq reads, with blue tracks representing the profile when only reads with the indicated number of G-to-A mutations are considered. (D) Correlation plot comparing transcript half-lives (log10 transformed) calculated using s⁶G TimeLapse-seq and s⁴U TimeLapse-seq. Histograms summarize the distribution of half-lives with the example transcripts indicated. The density of points is indicated by color (yellow, low; blue, high). (E) Correlation plots of RNA-seq profile comparing 6-TG versus untreated cells (top) and TimeLapse chemistry treated and untreated RNA (bottom) with Pearson’s correlation coefficient reported on each plot.

Each read from a TimeLapse-seq experiment reports on the mutational content of a single molecule of RNA that is either new (labeled) or pre-existing (unlabeled). We previously developed a statistical analysis of nucleotide recoding data using a binomial distribution to model the read distribution.⁴ Based on the number of G-to-A mutations in each read, accounting for new reads lost during handling, we calculated the fractions of new RNA that were produced during the treatment for over 4000 transcripts (see SI, Figure 2D). These fractions were reproducible across replicates (Pearson’s r = 0.92, see SI), and correlated well with results using s⁴U TimeLapse chemistry (Pearson’s r = 0.84, Figure 2D). Assuming simple exponential kinetics, we estimate RNA half-lives using these fractions (Figure 2C and 2D).

We found that known fast turnover transcripts such as transcription factors (e.g. JUN) have significantly shorter half lives than SMG5 or slow turnover transcripts such as GAPDH, consistent with previous reports (Figure 2D).⁴ Notably, using TimeLapse-seq with 6-TG allows us to estimate the half-lives of uridine-poor transcripts such as CBX4, whose reads have on average 10 uridine nucleotides, but 60 guanosine nucleotides (SI).

These results demonstrate that s⁶G can be used to monitor transcriptome-wide RNA population dynamics. Specifically, TimeLapse chemistry can be extended beyond s⁴U and can be applied to recode s⁶G to cause specific G-to-A mutations in sequencing experiments. While the lower incorporation rates of s⁶G¹⁵ lead to lower mutation rates induced by s⁶G compared with those induced by s⁴U (s⁶G 1.5%, s⁴U 4.5%), this rate is well above background (0.15% G-to-A mutations in TimeLapse treated or untreated samples without s⁶G) and allows analysis of the fraction of each transcript that is new. Half lives calculated from s⁶G TimeLapse-seq correlate well with those determined using s⁴U TimeLapse-seq (Figure 2D). Similar to s⁴U TimeLapse-seq, the chemical treatment and metabolic labeling with 6-TG preserve the information of traditional RNA-seq (Figure 2E) while providing insight into RNA population dynamics.

This nucleotide recoding chemistry adds a new technique to the larger set of experiments that use mutations to study nucleic acids including the analysis of epigenetic modifications through bisulfite sequencing¹⁶, RNA structure^17–19, RNA-protein interactions^12,20 and posttranscriptional modifications²¹. The impact of chemical approaches to reveal nucleic acids biology through mutational analysis continues to increase as sequencing technologies continue to evolve. TimeLapse-seq with s⁶G expands the power of nucleotide recoding by providing the first determination of RNA half lives using an analogue of guanosine and introduces the potential to analyze multiple timepoints using different metabolic labels in a single RNA-seq experiment.