A portable poison exon for small molecule control of mammalian gene expression

Aptamer-regulated poison exon as a strategy for traceless gene regulation

To achieve traceless drug-induced gene regulation, we took inspiration from poison exons, which are exons with premature stop codons (Fig. 1a). Typically, poison exons undergo alternative splicing in response to the binding of splicing factors^12,13. When the poison exon is spliced into the transcript, it introduces a premature stop codon, resulting in a truncated and inactive protein. In contrast, skipping of the poison exon enables the production of the full-length protein due to the complete removal of the poison exon from the mRNA sequence.

We aimed to generate a poison exon whose skipping is controlled by a small molecule (Fig. 1b). Small molecules that regulate splicing, such as risdiplam and branaplam^7,14, bind both the 5’ splice site and the U1 snRNP (U1 small nuclear ribonucleoprotein) to promote exon inclusion rather than skipping. However, these small molecules can affect splicing throughout the transcriptome⁷. This may account for the toxicities seen with branaplam⁸.

We therefore chose to use small molecules that bind RNA aptamers rather than splice sites. Some aptamers undergo conformational changes upon binding their ligands, which we and others have used to regulate the folding of various RNAs, including fluorogenic aptamers¹⁵ and ribozymes^16,17. The small molecule-binding aptamer typically contains a helical stem that is directly connected to a functionally important helical stem in the target RNA. When the small molecule binds, it induces a conformational change that stabilizes both helical stems. We therefore asked if any existing exons contain a stem-loop structure that is functionally linked to exon skipping.

One notable example of an exon containing a functionally relevant stem-loop structure is exon 7 of the SMN genes (SMN1 and SMN2)¹⁸. This stem-loop, termed terminal stem-loop 2 (TSL2) (Fig. 1c,d), is located at the end of exon 7. The base of the stem includes two nucleotides from the 5’ splice site in the adjacent intron. These nucleotides are part of the binding site of the U1 snRNP. Because TSL2 partially occludes the U1-binding sequence, exon 7 is skipped in approximately 6% of SMN1 transcripts¹⁸. Mutant versions of TSL2 with additional base pairing with the U1-binding sequence lead to more skipping of exon 7¹⁸. Conversely, decreasing the number of base pairs to increase accessibility to the U1-binding sequence reduces skipping of exon 7¹⁸. Thus, the functional importance of TSL2 suggests that it could be a useful entry point for small-molecule aptamers to enable drug-regulated exon skipping.

Selection of acyclovir for the control of exon skipping

To control splicing, we wanted to use an aptamer that binds a molecule with minimal or no effects on cellular physiology. Tetracycline has been used for making ligand-controlled RNA devices in mammalian cells^16,19–22, but antibiotic use promotes antibiotic resistance¹¹, and tetracycline induces mitochondrial dysfunction even at low doses^9,10. We therefore chose an aptamer that binds acyclovir (Fig. 1e,f)²³, an antiviral^24–28 that can be administered to cells at concentrations up to 10 mM with no apparent cytotoxicity^29–31. We confirmed that acyclovir has minimal effect on gene expression or cell viability, in contrast to known splicing regulators (Extended Data Fig. 1). Therapeutic concentrations of acyclovir typically range between 18.4 μM and 36.0 μM²⁴, and up to ~114 μM in some cases³². These concentrations are higher than the measured ~1.3 μM K_D of acyclovir binding to its aptamer (Extended Data Fig. 2). The nontoxic nature of acyclovir, even at high concentrations, makes it desirable for controlling synthetic RNA devices in cells.

Conversion of SMN1 exon 7 into a portable poison exon

Since exon 7 splicing relies on sequences in the introns preceding and following the exon, our initial cassette comprised SMN1 intron 6-exon 7-intron 7 (Fig. 1g). Exon 7 contains the SMN1 stop codon but would function as a premature stop codon when inserted within a target gene.

A problem with the intron 6-exon 7-intron 7 cassette is the length of intron 6 (5764 nt). Intron 6 can be truncated to the first 62 nt and the last 138 nt of the original intron 6 (Extended Data Fig. 3a and Supplementary Table 1), and still be efficiently spliced³³. Intron 7 was not truncated, as previous studies showed that truncation impairs its splicing³³. We confirmed that both introns, i.e., truncated intron 6 (termed intron 6_trunc) and the entire 444-nt intron 7, are efficiently spliced out when inserted into a Renilla luciferase gene (Extended Data Fig. 3b).

We next asked if the contiguous intron 6_trunc-exon 7-intron 7 cassette would splice exon 7 into the mRNA when inserted into a target gene (Fig. 1g). To identify an appropriate site for inserting the cassette, we examined the Renilla luciferase sequence for AG/G or UG/G motifs^34,35. Introns are typically located between (A/U)G and G nucleotides. We inserted the intron 6_trunc-exon 7-intron 7 cassette between UG521–522 and G523 (Extended Data Fig. 3c). This location maintains the reading frame with the stop codon in exon 7. Thus, successful splicing of exon 7 into the luciferase transcript is expected to block luciferase activity.

After insertion of the intron 6_trunc-exon 7-intron 7 cassette, luciferase expression decreased by ~92% (Fig. 1g). This indicates that poison exon 7 was successfully spliced into the luciferase transcript. RT-PCR showed that the majority (~83%) of the luciferase transcript contained an additional ~50 nt sequence, which reflects the size of exon 7 (Fig. 1g). Thus, intron 6_trunc-exon 7-intron 7 is a portable cassette that can cause exon 7 to be spliced into a heterologous transcript as a poison exon.

Conversion of TSL2 into hairpins suitable for aptamer control

Our strategy involves replacing the upper part of the TSL2 stem-loop with the acyclovir aptamer (Fig. 1b). The acyclovir aptamer would be relatively unfolded in the absence of acyclovir. This unfolded aptamer would perturb the stability of the helix below, which contains the U1-binding sequence. Because the U1-binding sequence would be accessible, poison exon 7 would be included, blocking expression of the protein of interest. However, when acyclovir is added, the acyclovir aptamer and its basal stem would fold, promoting the folding of the helix below. This would sequester the U1-binding sequence, causing the exon to be skipped, leading to production of the native transcript and subsequent protein expression.

We first needed to determine if varying the folding of TSL2 provides a range of splicing from 0 to 100%. To test this, we removed base pairs in TSL2 (Fig. 1h), which would in principle make the U1-binding sequence fully accessible. This led to no skipping of exon 7 and essentially complete loss of luciferase activity (Fig. 1h). In contrast, adding nucleotides to obtain complete base pairing with the U1-binding sequence caused nearly complete exon skipping and full luciferase activity (Fig. 1i). This indicates that controlling TSL2 hybridization with the U1-binding sequence could enable a broad range of splicing control.

To improve our chances of finding an optimal entry point for insertion of the acyclovir aptamer, we wanted to develop multiple hairpins that sequester the U1-binding sequence and efficiently induce exon 7 skipping. The first potential hairpin for insertion was the TSL2 mutant containing a helical stem that is fully hybridized to the U1-binding sequence (Fig. 1i). We designated this TSL2 mutant as “Hairpin 1.”

For “Hairpin 2,” we considered a hairpin at an alternative location within the intron 6_trunc-exon 7-intron 7 cassette. TSL2 traverses the exon-intron junction. We therefore considered an “intron-only” hairpin with the U1-binding sequence on the 5’ side of the helix, followed by a stem-loop and then a newly inserted 6-nt-long U1-complementary sequence (Fig. 1i). We mutated U1-complementary nucleotides in TSL2 to prevent disruption of Hairpin 2 folding downstream, as both structures hybridize to the same U1-binding sequence (Fig. 1d,h,i). The intron 6_trunc-exon 7-intron 7 cassette containing Hairpin 2 maintained highly efficient luciferase expression, indicating efficient skipping of exon 7 (Fig. 1i). Thus, the acyclovir aptamer can be inserted into either Hairpin 1 or Hairpin 2 to control exon skipping.

Interestingly, in the constructs containing either Hairpin 1 or Hairpin 2, the level of protein expression was higher than a control luciferase transcript lacking the cassette. This likely reflects the stimulatory effect of splicing on mRNA nuclear export³⁶, capping and polyadenylation of mRNAs^36–38, which collectively lead to enhanced protein expression.

Control of SMN1 exon 7 skipping using an acyclovir aptamer

To prepare Hairpins 1 and 2 for acyclovir regulation, we removed extraneous base pairs above the U1-binding sequence in each hairpin (Fig. 2a). Simultaneously, we truncated the extended stem below the putative acyclovir-binding site in the acyclovir aptamer (Fig. 2a)²³. The modified acyclovir aptamer was then fused to the hairpins, creating the acyclovir-regulated splicing cassettes, Ac-Hairpin 1 and Ac-Hairpin 2 (Fig. 2a).

To test the effectiveness of these cassettes, we inserted Ac-Hairpin 1 and Ac-Hairpin 2 into Renilla luciferase and measured acyclovir-induced luciferase expression. Ac-Hairpin 1 did not show any acyclovir-induced increase in luminescence: baseline luciferase expression was 5.3% of the control and did not increase upon acyclovir addition (500 μM, Fig. 2b). In contrast, Ac-Hairpin 2 resulted in a 29-fold increase in luminescence (0.4% to 12.7%) upon acyclovir treatment (500 μM, Fig. 2b).

To enhance acyclovir-induced exon skipping, we increased sequestration of the U1-binding sequence by adding one, two, or three additional base pairs between the acyclovir aptamer and Hairpin 2 (Fig. 2c). Transducer 1, which contained an additional U•A base pair, showed the greatest improvement in gene activation (35.4%) while maintaining low background expression (1.0%) (Fig. 2d).

We next tested transducer sequence variants (Transducers 7–12) (Fig. 2c,d). In the case of Transducer 7 (Ac-H2Tr7), the signal ranged from 0.3% luciferase expression (no acyclovir) to 90.5% expression (500 μM acyclovir), which reflected a 295-fold gene induction by acyclovir (Fig. 2d,e and Supplementary Table 3).

Design of a poison exon that functions in multiple cell lines

We expressed Renilla luciferase containing Ac-H2Tr7 in three additional cell lines of diverse origins: HeLa, Huh-7 (human hepatocellular carcinoma cell line), and HT-1080 (human fibrosarcoma). In Huh-7 and HT-1080 cells, we observed elevated background expression, indicating that the poison exon was skipped, to a small extent, even without acyclovir (Fig. 3a). To address this, we sought to modify the flanking introns by using more efficient intron junctional sequences to enhance poison exon splicing.

We tested several well-studied introns that splice with high efficiencies (Supplementary Table 2): MAP3K7 intron 4³⁹, ORAI2 intron 1³⁹, TMEM14C intron 1³⁹, HBB intron 2⁴⁰, a β-globin/IgG chimeric intron^41,42, the MINX intron⁴³, and TPI1 introns 4 and 5⁴⁴. We first truncated the long introns to 250 nt by removing internal sequences while preserving splice junctions and branch points (Extended Data Fig. 4a). Then, each intron was inserted into Renilla luciferase to test its ability to undergo splicing (Extended Data Fig. 4b). Among these introns, HBB intron 2 and the β-globin/IgG chimeric intron showed better performance than the SMN1 introns (Extended Data Fig. 3b and 4c). Besides restoring luciferase expression by being spliced out of the coding sequence, HBB intron 2 and the chimeric intron enhanced overall protein expression by 2-fold and 3-fold, respectively.

We next wanted to confer the improved splicing properties to the Ac-H2Tr7-containing cassette by partially replacing the SMN1 introns with the stronger introns identified above. We preserved the intron 6 and intron 7 sequences immediately flanking exon 7. The remaining parts of introns 6 and 7 were replaced, alone or in combination, with stronger introns to create six chimeric introns (Extended Data Fig. 4d). We then tested their performance in Huh-7 and HT-1080 cells (Extended Data Fig. 4e). Among these, we found a modified version of intron 6 that exhibited almost no background expression and high levels of acyclovir-induced activation in both cell lines. This construct comprised the first 100 nt of the β-globin/IgG chimeric intron and the terminal 138 nt of intron 6, and maintained the SMN1 intron 7. We termed this construct Ac-H2Tr7In4 (Fig. 3b).

We tested Ac-H2Tr7In4 in five additional cell lines, HeLa, U-2 OS, HEK293T, HCT 116, and Hepa 1–6, and found low background and high acyclovir-induced luciferase expression in all cell types (Fig. 3c). This demonstrates that Ac-H2Tr7In4 functions consistently across diverse cell lines. We designated the Ac-H2Tr7In4-containing cassette Cyclone (acyclovir-controlled poison exon).

Tunable and reversible protein expression by acyclovir-dependent skipping of the poison exon

To determine if Cyclone allows a graded response based on acyclovir concentration, we expressed Renilla luciferase containing Cyclone in HEK293T cells treated with increasing concentrations of acyclovir (0 μM to 2 mM) for 48 h. In the absence of acyclovir, RT-PCR detected a single product (250 nt) corresponding to poison exon inclusion (Fig. 3d). As the acyclovir concentration increased, a 196-nt product appeared, reflecting poison exon skipping (Fig. 3d). Protein expression increased with acyclovir concentration, reaching levels equivalent to the control lacking the poison exon at 250 μM acyclovir (Fig. 3e). Acyclovir-dependent regulation of luciferase expression was also seen using promoters of different strengths (Fig. 3f). Similar acyclovir dose-dependent protein expression was observed with Cyclone-regulated EGFP and mCherry constructs (Fig. 3g,h). Together, these data demonstrate that acyclovir dose controls poison exon skipping and protein expression levels.

Notably, Cyclone enhanced luciferase expression by ~2–3-fold compared to Ac-H2Tr7 at every acyclovir dose tested (Fig. 3e). This suggests that the chimeric introns facilitate mRNA processing and nuclear export. At acyclovir concentrations above 250 μM, the Cyclone construct yielded protein expression levels exceeding 100% of the no-Cyclone control. The maximal acyclovir-induced activation of Ac-H2Tr7 was 101.4% relative to the control (2 mM acyclovir), while Cyclone reached 349.5% at this concentration. Thus, Cyclone enables fine-tuning of gene expression across a wide range, from subphysiological to supraphysiological levels, depending on acyclovir dose.

To investigate the time course of acyclovir-induced poison exon skipping, we treated HEK293T cells with 500 μM acyclovir. Poison exon skipping was detectable ~4 h after treatment, with maximal skipping observed at 6 h (Fig. 3i). To assess reversibility, we pre-treated cells with 500 μM acyclovir for 24 h before replacing the medium with fresh, acyclovir-free medium. By 6 h after acyclovir removal, nearly all transcripts had reverted to poison exon inclusion (Fig. 3j).

The background expression and induction level of Cyclone were superior to a highly optimized transfection-based Tet-On system (Fig. 3k and Extended Data Fig. 6a)⁴⁵. When using cell lines optimized for expression of the reverse tetracycline-controlled transactivator, rtTA^45–48, we observed similar background and induction levels compared to Cyclone (Extended Data Fig. 6b). These results demonstrate that Cyclone provides robust gene control with a simplified experimental setup.

Cyclone enables traceless acyclovir-dependent expression of diverse transgenes

We next tested whether Cyclone could be used for traceless and acyclovir-dependent expression across diverse transgenes. To enable traceless gene regulation, Cyclone must be inserted directly into the gene of interest. Based on the most common splice site consensus motifs^34,35, the Cyclone cassette can be inserted between any NAG and GNN codon. In this way, the stop codon of Cyclone will be in-frame, and exon 7 will function as a poison exon. Since many codons can undergo synonymous mutations that maintain the identity of the encoded amino acid, Cyclone can be inserted in-frame after any glutamine, glutamate, or lysine that is followed by a valine, alanine, aspartate, glutamate, or glycine. This versatility ensures that multiple potential insertion sites will likely exist for most genes of interest.

To test traceless control of gene expression, we inserted Cyclone into the mouse thrombopoietin (Thpo) gene. Thrombopoietin requires traceless gene expression control because sequence changes in the protein are generally not tolerated⁴⁹. We inserted Cyclone at two positions within Thpo (Fig. 4a): between Q82 and D83, or between Q117 and V118 (Supplementary Table 4). We expressed each Cyclone-Thpo construct in HEK293T cells, and thrombopoietin levels were measured in the culture media by ELISA (Fig. 4b). At both insertion sites, thrombopoietin was expressed in an acyclovir dose-dependent manner, with minimal background and similar degrees of maximal activation.

We also generated acyclovir-regulated Cre recombinase. We inserted the poison exon into the coding sequence of Cre recombinase (Supplementary Table 4) and transfected it into a loxP-GFP Cre reporter N2a cell line⁵⁰ to measure Cre activity (Fig. 4c). In the absence of acyclovir, no Cre expression was detected by western blot (Fig. 4d), and minimal GFP expression was observed (Fig. 4e,f). Cre expression increased with higher doses of acyclovir (Fig. 4d), with approximately 40% of cells exhibiting green fluorescence (Fig. 4e,f).

We next explored a strategy to generate and select cells with acyclovir-inducible expression of a model transgene. Using EGFP as an example, we generated a puromycin N-acetyltransferase (PAC)-P2A-EGFP fusion protein (Fig. 4g). The P2A sequence is a self-cleaving peptide⁵¹ that causes PAC and EGFP to be produced as separate proteins. We inserted Cyclone into the pac gene (Supplementary Table 4) and expressed the construct in HEK293T cells cultured in the presence of acyclovir to induce the expression of the fusion protein. Puromycin-selected HEK293T cells exhibited acyclovir dose-dependent PAC (Fig. 4h) and EGFP expression (Fig. 4i). These results suggest that this PAC-based selection strategy may be adaptable for other genes of interest by substituting EGFP.

Pac-Cyclone enables acyclovir control of endogenous genes

In contrast to transgenes, it remains challenging to control the expression of endogenous genes with small molecules. In contrast to the promoters used by transgenes, endogenous promoter elements are controlled by diverse aspects of chromatin organization, which ensure tissue-specific, developmental, and regulated patterns of gene expression. Therefore, it is desirable to develop small-molecule systems that can precisely control the magnitude and duration of endogenous gene expression.

Although Cyclone can be knocked into any gene, we used a Pac-Cyclone cassette (Fig. 5a and Supplementary Table 5) for the rapid selection of clones with genomic integration. The Pac-Cyclone cassette is designed to be inserted after the start codon of an endogenous gene via CRISPR knock-in through homology-directed repair (HDR). The start codon of the endogenous gene initiates the translation of PAC, followed by the P2A sequence, and finally the protein encoded by the endogenous gene. The intervening P2A site ensures that these are produced as separate proteins. P2A cleavage leaves a proline which is followed by a 3X-FLAG epitope tag that was added to the beginning of the endogenous protein to facilitate detection.

We knocked Pac-Cyclone into two endogenous genes, WTAP (WT1 associated protein) (Fig. 5b) and PCIF1 (phosphorylated CTD interacting factor 1) (Fig. 5c). These genes were chosen because they have different exon-intron structures and produce proteins of different sizes. We selected WTAP because it is an essential gene and therefore stable knockout cell lines cannot be generated⁵². In both knock-in experiments, we used HDR donor templates that included Pac-Cyclone flanked by homology arms from each gene (Supplementary Table 6,7). After knocking Pac-Cyclone into HCT 116 cells (Fig. 5d), the cells were cultured with acyclovir (250 μM) to maintain expression of the endogenous gene. Puromycin was subsequently added to select for stable integration. In the case of WTAP, clonal cell lines with homozygous insertion were selected.

Prior to the selection of clonal cell lines, the cell population contained a mixture of homozygous and heterozygous insertions. This was readily detected using anti-WTAP western blotting which revealed the endogenous WTAP, and a slightly larger band corresponding to 3xFLAG-WTAP (Fig. 5e). The expression of the 3xFLAG-WTAP band was controlled by the acyclovir dose. Similar results were obtained with PCIF1 (Fig. 5f). For these experiments, cells were first cultured in 250 μM acyclovir and then switched to media containing 0, 62.5, or 250 μM acyclovir for 2 days.

To obtain HCT 116 cells with homozygous knock-in, we performed clonal selection (Extended Data Fig. 7). In these clonal cell lines, we observed only a single WTAP band, reflecting the 3xFLAG-WTAP (Fig. 5g). We cultured the cells in varying concentrations of acyclovir for 5 days and found dose-dependent WTAP expression with no detectable WTAP at 0 μM acyclovir. Notably, doses of acyclovir above 62.5 μM resulted in WTAP levels that exceeded wild-type levels. Acyclovir-induced poison exon skipping was confirmed using RT-PCR with primers annealed adjacent to the targeted WTAP genomic locus (Fig. 5g). This demonstrates that Cyclone can enable acyclovir-dependent control of endogenous gene expression, allowing both subphysiological and supraphysiological expression levels.

M2 DMS MaP reveals the structural basis of ligand regulation

Cyclone is notably different from many other aptamer-based gene regulatory switches in terms of its large fold increase in gene expression in response to acyclovir. Previous systems typically exhibit less than a 20-fold change in gene expression^{16,17,19,21,23}. Therefore, we wanted to understand the structural basis for the robust acyclovir response seen with Cyclone. To test this, we used a multidimensional chemical-mapping method called “mutate-and-map” dimethyl sulfate mutational profiling (M2 DMS MaP). M2 DMS MaP is a powerful method for identifying pairs of interacting nucleotides^53,54. M2 DMS MaP relies on mutagenesis to introduce sparse nucleotide mutations in an RNA of interest. These mutations induce structural perturbations in partner nucleotides, which are then detected through their altered chemical reactivity as measured by DMS probing. Heatmaps generated by M2 DMS MaP reveal helix signatures as diagonal lines⁵³.

We performed M2 DMS MaP on an in vitro transcribed 800-nt long Cyclone RNA sequence, including the adjacent intronic sequences. Mutations were introduced using error-prone PCR. We performed DMS probing in the absence or presence of acyclovir (250 μM). Targeted RT-PCR and amplicon sequencing were then performed on the region surrounding the acyclovir aptamer and the stem containing the U1-binding sequence (P1) (Fig. 6a).

To identify structural changes induced by acyclovir, we plotted the population-average DMS reactivity onto the secondary structure (Fig. 6a,b). Acyclovir lowers the overall population-average DMS reactivity in the aptamer domain and the P1 stem, suggesting ligand-induced folding (Fig. 6a). The residues at the putative acyclovir-binding pocket undergo the most significant decrease in DMS reactivity.

To understand the structural changes induced by acyclovir, we used sequencing data to generate M2 heatmaps for the acyclovir-free and acyclovir-bound states (Fig. 6c,d). The M2 heatmap of the acyclovir-bound state displays two large helical domains, as well as intervening box-like patterns (Fig. 6d), which are notably absent in the ligand-free heatmap (Fig. 6c). Box-like patterns in M2 heatmaps often indicate the presence of complex tertiary conformations⁵⁴ and likely represent the folding of the ligand-binding pocket upon the addition of acyclovir. Thus, the M2 heatmaps collectively demonstrate that acyclovir induces a large conformational change in the acyclovir aptamer, reflecting the assembly of an acyclovir-binding pocket.

We next examined the effect of acyclovir on the P1 stem, which contains the U1-binding sequence. In the M2 heatmaps, a diagonal pattern reflecting the helical organization of P1 was detected both with and without acyclovir (Fig. 6c,d), suggesting that P1 likely exists in a folded:unfolded equilibrium. The strengthening of P1 by acyclovir (Fig. 6c,d) suggests that acyclovir shifts the equilibrium a sufficient degree to induce exon skipping.

Notably, the acyclovir-induced box-like network of interactions includes the immediately adjacent base pairs in the P1 stem, suggesting that these residues are directly regulated by acyclovir binding. This base pairing likely stabilizes the remainder of the P1 stem, thereby facilitating the overall sequestration of the U1-binding sequence. The remarkably close arrangement of the acyclovir-binding pocket and the P1 stem likely facilitates the ability of acyclovir to influence the U1-binding function of the P1 stem.

Aptamer “plug-and-play” for the change of Cyclone specificity

To test whether the small molecule regulator of Cyclone could be swapped, we replaced the acyclovir-binding aptamer with a tetracycline aptamer at the site identified by M2 DMS MaP as the acyclovir-binding pocket. The tetracycline-binding pocket was identified through in vitro selection⁵⁵ and later confirmed by crystallography⁵⁶. We therefore inserted this domain into the P1 stem, using zero, one, or two base pairs of varying composition between the tetracycline aptamer and the P1 stem (Extended Data Fig. 8a).

When expressed in HEK293T cells, each construct exhibited minimal luciferase expression in the absence of tetracycline, indicating efficient inclusion of the poison exon (Extended Data Fig. 8b). However, three constructs showed robust tetracycline-induced luciferase expression. Notably, Transducer 8 exhibited nearly 80% luciferase expression at 125 μM tetracycline relative to a non-splicing luciferase control and achieved ~40-fold gene activation between 0 and 125 μM tetracycline (Extended Data Fig. 8b,c).

These findings suggest that Cyclone’s control ligand can be readily switched to other molecules.

Plasmid cloning, characterization and purification

Introns and poison exon cassettes were ordered as gBlocks from IDT (Coralville, IA). Alternatively, they were ordered as double-stranded DNA gene fragments from Twist Bioscience (South San Francisco, CA). Each gene fragment contains 30–40 bp of vector-specific homologous regions on the tails of the fragment to accommodate Gibson Assembly with a linearized vector backbone.

Plasmid backbones were linearized by double digestion using restriction enzymes purchased from New England Biolabs (NEB) (Ipswich, MA). Appropriate experimental conditions and protocols for double digestion were generated using the NEBcloner web tool. The digestion mixture was separated on a 1% agarose gel to isolate the linearized fragment. Gel purification was carried out using the QIAquick Gel Extraction Kit (catalog number 28706). The psiCHECK-2 vector was purchased from Promega (catalog number C8021). pINDUCER20 (Plasmid #44012), pcDNA3-EGFP (Plasmid #13031), pcDNA3-mCherry (Plasmid #30125), pCMV-Cre (Plasmid #123133), and LentiGuide Puro-P2A-EGFP (Plasmid #137729) were purchased from Addgene (Watertown, MA). The Thpo Mus musculus cDNA ORF clone in a pcDNA3.1+/C-(K)-DYK backbone was ordered from GenScript (Piscataway, NJ).

Gene fragments were assembled with linearized backbones using the Gibson Assembly Master Mix (NEB E2611L) following the manufacturer’s protocol. The assembly mixture was transformed into NEB^® 5-alpha Competent E. coli (High Efficiency) (NEB C2987I) following standard cloning procedures. All assembled plasmids were verified via Sanger sequencing or whole plasmid sequencing by GENEWIZ (Azenta Life Sciences).

Drug preparation

Acyclovir (catalog number A1915–5G) was purchased from TCI AMERICA. It was dissolved in DMSO to create a 200 mM concentrated stock solution and kept frozen at −20°C for long-term storage. Further dilutions of the stock solution into aqueous buffers or cell culture media should be made prior to performing biological experiments. No drug precipitation was observed as long as the final concentration did not exceed 2 mM.

Doxycycline (Takara Bio 631311) was diluted to 1 mg/ml in double-distilled H₂O, filtered sterilize, aliquoted, and stored at −20°C in the dark.

Cell culture conditions and transfection reagents

All cell lines were obtained directly from the American Type Culture Collection (ATCC) except for a loxP-GFP Cre reporter N2a cell line, which was a gift from the Zhang Lab at Broad Institute. The base media was supplemented with 10% heat-inactivated Fetal Bovine Serum (FBS), 100 units/mL penicillin, and 100 μg/mL streptomycin (Gibco 15140122). HEK293T and Huh-7 were cultured in Dulbecco’s Modified Eagle’s Medium (DMEM) (catalog number 11995065 or ATCC 30–2002), while HeLa, HT-1080, and N2a were cultured in Eagle’s Minimum Essential Medium (EMEM) (ATCC 30–2003). U-2 OS and HCT 116 were cultured in McCoy’s 5a Medium Modified (ATCC 30–2007). HEK293T cells used for Tet-On experiments were cultured in Tet System Approved FBS (Takara Bio 631105) instead of regular FBS to minimize tetracycline contamination. All cell lines were cultured at 37°C in a 5% CO2-humidified incubator under standard tissue culture conditions.

Transient transfections of HEK293T, Huh-7, HeLa, and U-2 OS were carried out using FuGENE HD transfection reagent (Promega E2311) following the manufacturer’s protocol. Transient transfections of HT-1080 and HCT 116 were done using Lipofectamine LTX reagent with PLUS reagent (Invitrogen A12621), and transient transfection of N2a was done using FuGENE 4K transfection reagent (Promega E5911). Transfection for CRISPR/Cas9 knock-ins in HCT 116 cells was carried out using Lipofectamine CRISPRMAX Cas9 transfection reagent (catalog number CMAX00008).

Luciferase assay

One day before transfection, cells were seeded into a clear-bottom, white, TC-treated 96-well microplate (Corning 3610). Each well holds 15,000 cells in 100 μl of growth medium. The edge wells of the plate were not seeded with cells; instead, they were filled with 1x PBS (phosphate-buffered saline) to minimize evaporation.

One hour before performing the transfection, the 200 mM acyclovir stock solution was diluted in the growth medium to achieve a concentration that was three times (3x) the final desired concentration. Then, 50 μl of this diluted acyclovir solution was added to the 100 μl of growth medium already present in each well without disturbing the cells. This process resulted in the final desired concentration of acyclovir (1x) in the wells.

Transfection was performed using Opti-MEM (Gibco 31985062) and transfection reagents appropriate for the cell line used in the experiment. Each plate contained blank wells for background subtraction and wells transfected with psiCHECK-2 control vectors for expression normalization, facilitating comparison among plates. Cells in each well were transfected with 100 ng of DNA plasmids. Transfection was performed according to the manufacturer’s protocol. After transfection, the cells were returned to the incubator.

Two days after transfection, the acyclovir-containing medium was replaced with 75 μl of fresh medium per well. Luminescence was measured using the Dual-Glo Luciferase Assay System (Promega E2920). To measure Firefly luminescence, 75 μl of Dual-Glo Luciferase reaction mix was added to each well. After a 10-minute incubation at room temperature, the luminescence signal was measured using the SpectraMax iD3 Multi-Mode Microplate Reader. Renilla luminescence was measured by adding 75 μl of Dual-Glo Stop & Glo reaction mix to each well and following the same steps as for the Firefly luminescence measurements.

Luminescence data analysis was performed using Excel (Microsoft) and Prism (GraphPad). The ratio between Renilla and Firefly luciferase activity was calculated for each well to normalize for transfection efficiency and cell number. This ratio was then normalized to the ratio generated by the psiCHECK-2 control vector, expressed as a percentage of expression.

RNA extraction and RT–PCR

Total RNA was extracted using TRIzol Reagent (Invitrogen 15596018) according to the manufacturer’s protocol. RNA concentration was determined using a NanoDrop spectrophotometer (Thermo Scientific), and 1 μg of total RNA was treated with DNase I (NEB M0303S). cDNA was generated from the RNA using gene-specific primers and SuperScript IV Reverse Transcriptase (Invitrogen 18090010) according to the manufacturer’s protocol. The cDNA was amplified by PCR using Phusion High-Fidelity PCR Master Mix (NEB M0531S) and then separated on a 2% agarose gel for imaging with a Bio-Rad Gel Doc XR imager.

Western blot

Cells expressing transgenes were treated with acyclovir for 48 hours after transfection before being harvested for western blot analysis. Cells expressing acyclovir-controlled WTAP or PCIF1 were treated with acyclovir for 5 days before being harvested for western blot analysis. Total protein was extracted from the cells using RIPA buffer with protease inhibitor (Thermo Scientific #78440) and quantified using the Quick Start Bradford Protein Assay (Bio-Rad 5000201). Equal amounts of total protein for each condition were separated on 4–12% NuPAGE Bis-Tris Mini Protein Gels (Invitrogen NP0321BOX) and transferred to Immobilon-P PVDF Membrane (Millipore IPVH00010). The membranes were blocked with 5% milk in PBS-T and incubated with the corresponding primary antibodies (anti-PCIF1 antibody (abcam ab300487); anti-WTAP antibody (Cell Signaling Technology 56501s); anti-puromycin N-acetyltransferase antibody (Invitrogen #702389); anti-FLAG antibody (Sigma-Aldrich F1804); anti-β-tubulin antibody (proteintech 10068–1-AP); anti-GAPDH antibody (GeneTex GT239)) at the recommended dilution factors at 4°C overnight, followed by incubation with secondary antibodies at room temperature for 1 hour. The membranes were treated with Pierce ECL Western Blotting Substrate (Thermo Scientific 32106) and imaged using a Bio-Rad Gel Doc XR imager.

Fluorescence imaging

Cells were plated in 24-well glass bottom tissue culture plates (MatTek) treated with Poly-D-Lysine (Cultrex 3439–100-01) and mouse Laminin (Cultrex 3400–010-02) and cultured for 24 hours before transfection. After transient transfection and acyclovir treatment, the cells were incubated for another 48 hours before being subjected to imaging experiments. To reduce background fluorescence, the culture media were replaced with complete FluoroBrite DMEM (Gibco A1896701) 1 hour before imaging. Hoechst 33342 was added to a final concentration of 0.5 μg/ml. The 24-well plate was transferred to a Tokai Hit stage top incubator (37°C and 5% CO₂). Live cell fluorescence images were obtained with a CoolSnap HQ2 CCD camera (Photometrics) through a ×20 air objective mounted on a Nikon Eclipse Ti-E microscope and analyzed with NIS-Elements software to detect EGFP (470 ± 20 nm excitation, 495 nm dichroic, 525 ± 25 nm emission) and Hoechst (350 ± 25 nm excitation, 400 nm dichroic, 460 ± 25 nm emission). Image analysis was conducted in ImageJ software.

Flow Cytometry Analysis

HEK293T cells were seeded in 6-well plates and treated with DMSO, acyclovir (500 μM), or doxycycline (250 ng/mL or 10 μg/mL) for two days. Cells were then harvested using TrypLE^™ Express Enzyme (Gibco 12604013), resuspended in 300 μl buffer (PBS, 2% FBS, 2 mM EDTA), and kept on ice. DAPI staining (Sigma D9542) was added before flow cytometry analysis (BD FACSymphony A5 Cell Analyzer, BD FACSDiva 9.9). In the EGFP experiment, a total of 250,000 events were recorded using the same voltage settings for all conditions. Similarly, a total of 100,000 events were recorded in the mCherry experiment with consistent voltage settings. Data were analyzed using FlowJo (FlowJo 10.10.0) with the gating strategy shown in Extended Data Fig. 5.

Isothermal titration calorimetry (ITC)

ITC binding experiments were performed using an Affinity ITC (TA Instruments). Titrations were carried out with the acyclovir aptamer RNA in the sample cell and acyclovir ligand as the titrant in the syringe. Acyclovir aptamer RNA (70 μM) was folded by heating for 10 minutes at 75°C in ITC running buffer (100 mM HEPES pH 7.4, 500 mM NaCl, 0.5 mM MgCl₂, and 0.25% (v/v) DMSO) and then cooled to room temperature over 1 hour on the benchtop. The acyclovir stock solution (200 mM) was diluted to a final concentration of 495.6 μM in ITC running buffer to serve as the titrant solution. The sample cell was loaded with acyclovir aptamer RNA, and the injection syringe was loaded with the acyclovir titrant solution. At the start of the ITC run, the instrument was equilibrated at 25°C, and the acyclovir titrant solution was injected into the sample cell with an initial injection of 1 μl followed by 16 injections of 3 μl. Data analysis was performed using NanoAnalyze (TA Instruments). After baseline correction, the area under each peak was integrated and fitted to a one-site independent binding model. The heat of dilution was subtracted from the integrated peak area. The first data point was removed from all data sets due to the different injection volume.

ELISA

HEK293T cells were plated in a 96-well plate as described in the luciferase assay section. After transient transfection and acyclovir treatment, cells were incubated for 24 hours before removing 2.5 μl of media to measure Thpo protein concentration using the Thrombopoietin (TPO/THPO) Mouse ELISA Kit (Invitrogen #EMTHPO) according to the manufacturer’s protocol. ELISA samples were measured with a SpectraMax iD3 Multi-Mode Microplate Reader.

M2 DMS MaP reagents and equipment

Dimethyl sulfate (CAS 77–78-1) was purchased from Sigma Aldrich (St. Louis, MO). DNA primers were ordered from IDT. DNA Clean & Concentrator-5 kit, RNA Clean & Concentrator-5 kit, and RNA Clean & Concentrator-25 kit were purchased from Zymo Research. AMPure XP Bead-Based Reagents were purchased from Beckman Coulter. Phusion High-Fidelity PCR Master Mix with HF Buffer, Induro Reverse Transcriptase, 5X Induro RT Reaction Buffer, and Barcoded sequencing oligos were purchased from NEB. AmpliScribe T7 High Yield Transcription Kit was purchased from Biosearch Technologies. RNaseOUT Recombinant Ribonuclease Inhibitor was purchased from Invitrogen.

Error-prone PCR and in vitro transcription for M2 DMS MaP

The DNA template for mutate-and-map experiments was prepared as previously described^53,61, using error-prone PCR with Taq polymerase. The DNA amplicons were evaluated on an agarose gel and purified with the DNA Clean & Concentrator-5 kit. We used 200 ng of mutated DNA as a template to perform in vitro transcription with the AmpliScribe T7 High Yield Transcription Kit. The reaction was purified with the RNA Clean & Concentrator-25 kit. RNA quality was assessed by visualization on 10% Novex TBE-Urea Gels (Invitrogen EC6875BOX) stained with SYBR Gold.

DMS modification of in vitro-transcribed RNA

In vitro-transcribed RNA was probed as previously described, with some minor changes to the protocol62. Briefly, 1 μg of non-mutagenized or mutagenized RNA was refolded by heating to 95°C for 3 minutes, then snap-cooled on ice for 1 minute in a buffer that promotes the modification of m¹G and m³U (200 mM Bicine, 100 mM KCl, pH 8.37 at room temperature). MgCl₂ (5 mM) and acyclovir (250 μM) or DMSO were then added, followed by incubation of the RNA at 37°C for 10 minutes. DMS probing was performed by adding neat DMS (1.7 M in anhydrous ethanol) to the folded RNA and incubating at 37°C for 6 minutes. The reaction was quenched by adding 4 volumes of 20% β-mercaptoethanol in water while incubating on ice. The RNA was purified from the quenched DMS reaction using the RNA Clean & Concentrator-5 kit and eluted in 10 μl water. The RNA was either processed directly or stored at −80°C.

Reverse transcription and sequencing library preparation

Reverse transcription was performed using Induro Reverse Transcriptase (NEB M0681S) according to the manufacturer’s protocol, with the following changes. RNA (500 ng) was mixed with 10 mM dNTP (2 μl), 2 μM RT primer (1 μl), and water to a final volume of 13 μl during the annealing step. A custom RT master mix was prepared by combining 5X Induro RT Reaction Buffer (4 μl), 100 mM DTT (1 μl), RNaseOUT Recombinant Ribonuclease Inhibitor (0.5 μl), Induro Reverse Transcriptase (200 U/μl, 1 μl), and water (0.5 μl). Reverse transcription reactions were incubated at 25°C for 10 minutes, followed by 57°C for 4 hours. To hydrolyze the RNA, 1 N NaOH (5 μl) and 500 mM EDTA (5 μl) were added to the reaction, which was then incubated at 95°C for 3 minutes. The cDNA was purified using the DNA Clean & Concentrator-5 kit and eluted in 15 μl water.

Sequencing libraries were generated by a 2-step PCR protocol using Phusion High-Fidelity PCR Master Mix with HF Buffer (NEB M0531S). Step 1 PCR amplifies cDNA to install the Illumina oligo binding site. cDNA (2 μl) was amplified for 20 cycles using the primers below. The reaction mix was purified using the DNA Clean & Concentrator-5 kit and eluted in 15 μl water.

Step 1 PCR forward primer (5’ 3’)

ACCACGACGCTCTTCCGATCTNNNNNTTTTCCTTACAGGGTTTCAGACAAAATCAAAAAGA

Step 1 PCR reverse primer (5’ 3’)

GACGTGTGCTCTTCCGATCTNNNNNACATTTGTTTTCCACAAACCATAAAGTTTTACAAA

Step 2 PCR was performed with NEBNext Multiplex Oligos for Illumina. The Step 2 PCR reaction mix contained purified DNA from Step 1 (2–10 ng in 12 μl water), i5 index primer (4 μl), i7 index primer (4 μl), and Phusion High-Fidelity PCR Master Mix with HF Buffer (20 μl). The PCR cycle number was optimized for each experiment. The reaction mix was size-selected with 0.9X AMPure XP Beads, eluted in 21 μl water, and quantified using Qubit 1X dsDNA High Sensitivity (HS) and Broad Range (BR) Assay Kits (Invitrogen). Libraries were pooled, cleaned up again with 0.9X AMPure XP Beads, and eluted in 28 μl water. The final library was quantified using Qubit 1X dsDNA High Sensitivity (HS) and Broad Range (BR) Assay Kits (Invitrogen) and sequenced on the NovaSeq X Plus PE 2 × 100 (~10 million reads per sample in the library).

Data Processing

DMS MaP sequencing data were processed using the ShapeMapper pipeline63^,64. Sequencing data were aligned to a custom FASTA sequence for the Cyclone amplicon. Sequencing data alignment was performed with ShapeMapper v2.2 using the following settings: --amplicon --output-parsed --output-aligned-reads --output-counted-mutations --min-mutation-separation 2⁵⁴.

For M2 DMS MaP, sequencing data were aligned to a single FASTA sequence with ShapeMapper v2.2 using the following settings: --amplicon --dms --output-parsed --min-mutation-separation 2. M2 heatmaps were generated as previously described⁵⁴. Briefly, the output parsed mutation file was processed with the script mut_to_simple.py, which retains only sequencing reads that cover the entire length of the RNA. The resulting .simple file was then used to calculate a unique mutational profile for each point mutant with the script simple_to_M2_map.py. Z-score normalization was performed using the script M2_Map_to_Zscores.py. The normalized matrix was then visualized in Prism GraphPad as a heatmap.

Design of sgRNA and HDR donor constructs

The overall workflow of our knock-in experiment is built upon the widely used endogenous tagging protocol based on the 2A peptide homology-directed repair (HDR) donor cassette65^,66. Our Pac-Cyclone HDR donor is inserted after the start codon and therefore tags the endogenous gene at the N-terminus.

Before making any designs, we recommend researching the gene of interest using the NCBI (National Center for Biotechnology Information) Gene database (https://www.ncbi.nlm.nih.gov/gene/) or the Ensembl genome database (https://useast.ensembl.org/index.html). Besides identifying the location of the start codon, it is also helpful to know which chromosome the gene is located on to determine the copy number in the cell line to be used. Learning the orientation of the gene is informative during the design step of the HDR donor. Additionally, it is important to ensure that the knock-in design can target all transcripts and isoforms of the gene of interest. Next, it is recommended to check the expression level of this gene in the cell line to be used for the knock-in experiment using The Human Protein Atlas database (https://www.proteinatlas.org/) to confirm that the gene will be expressed.

The sgRNAs and HDR donors targeting the WTAP and PCIF1 genes were generated using a web-based design tool from GenScript (Homology-Directed Repair (HDR) Knock-in Templates). This tool provides a graphic display and interactive interface of the target genomic locus for convenient design of the sgRNAs, HDR donors, and custom primers for assessing editing efficiency. To design the gene-specific HDR donor containing the poison exon, search for the gene ID and transcript ID, or provide the genome coordinates of the start codon. After placing the cursor behind the start codon, input the sequence of Pac-Cyclone (Supplementary Table 5). The design tool will by default add 300 bp long homology arms on both ends to ensure that the donor overlaps with the locus for efficient homology-directed repair. Choose the “add silent mutations” option to disrupt the PAM sequence and avoid sgRNA binding and re-editing events. Design one HDR donor with mutations for testing multiple sgRNAs. The HDR donor was ordered directly from GenScript in single-stranded DNA format (3 μg). The web tool recommends the top five sgRNAs in this region of the genome. Choose 2–3 sgRNAs with cut sites within 30 bp of the insertion site for better editing efficiency. It is also advised to consider the “on-target score” to avoid off-target editing events. Finally, custom primers for assessing editing efficiency were generated automatically by the design tool.

Delivery of CRISPR-Cas9 editing components and cell culture

Before delivering the editing components into cells, the efficiency of the sgRNAs was tested using the GenCrispr sgRNA Screening Kit (GenScript L00689). HCT 116 cells were chosen due to their predominantly diploid nature, fast growth rate, and susceptibility to transfection. The knock-in experiment was started at the third passage of the HCT 116 cells, which were obtained directly from ATCC.

Two days before transfection, cells were plated into a 6-well TC-treated plate (0.4 million cells per well in 2 ml medium). Seed additional wells as transfection control and puromycin selection control. One day before transfection, cells were treated with nocodazole (5 ng/ml) (catalog number SML1665–1ML) for 15 hours. Nocodazole-containing medium was replaced with acyclovir-containing medium (250 μM acyclovir, 3 ml per well) 30 minutes before transfection. GenCRISPR^™ Ultra NLS-Cas9 protein (6.25 μg) (GenScript Z03621), sgRNA (1.2 μg), and single-stranded HDR donor (2.5 μg) were delivered by chemical transfection using Lipofectamine^™ CRISPRMAX^™ Cas9 Transfection Reagent (Invitrogen CMAX00008). Replace the medium with fresh medium containing 250 μM acyclovir 24 hours after transfection. Incubate for another 48 hours before starting puromycin selection.

Three days after transfection, cells were passaged from the 6-well plate into a 10-cm dish using medium containing acyclovir and puromycin (250 μM acyclovir, 0.4 μg/ml puromycin) (Sigma-Aldrich P4512). The puromycin concentration was determined by establishing an antibiotic kill curve to find the minimum concentration required to kill unedited cells. Wash cells daily to remove dead cells and debris. Passage cells to a 15-cm dish if needed to avoid contact inhibition. Puromycin selection was maintained for 5 days as the control cells all died. Cells were then recovered in acyclovir-only medium for two days before starting clonal expansion.

Limited dilution and clonal expansion

After puromycin selection, limiting dilution was performed to isolate monoclonal cell populations for genotyping. During this process, cells were cultured in complete medium containing 250 μM acyclovir. Cells were washed with 1x PBS (Gibco 10010023) and then trypsinized using TrypLE^™ Express Enzyme (Gibco 12604013) to obtain a homogenized cell solution. Cell concentration was quantified using the Countess 3 Automated Cell Counter (Invitrogen AMQAX2000). Prepare 10 ml of cell solution with a concentration of 5 cells/ml for each 96-well plate to be used. Transfer 100 μL of this cell solution into each well of a 96-well plate to obtain an average density of 0.5 cells/well. Seed one well at the corner with ~1000 cells to help find the focal plane. Cells were left undisturbed in an incubator for 7–14 days. After ~7 days, scan the plate for cell growth. Wells containing monoclonal populations were expanded into 24-well plates. One monoclonal population was split into two wells: one for genotyping and one for further expansion. After genotyping, populations with undesired genotypes were discarded. Edited clones with both alleles tagged were expanded into larger culture dishes, screened for acyclovir-controlled expression, and then frozen into aliquots and stored in liquid nitrogen.

Genotyping by genomic PCR

Monoclonal cells for genotyping were grown in 24-well plates as described in the previous section. Extract genomic DNA from wild-type and tagged cells by adding 200 μl of QuickExtract^™ DNA Extraction Solution (catalog number QE09050) to cells trypsinized and collected from each well. Genomic DNA (ranging from 0.5 μl to 5 μl, depending on cell density) was used for PCR amplification with Phusion high-fidelity polymerase (NEB M0531L) using primers that anneal outside the donor sequence to check for both successful editing and homozygosity. Amplicons were analyzed on 1% agarose gels. Wild-type cells show a single band corresponding to two unedited alleles. Heterozygous editing shows the appearance of an upper band corresponding to integration of the donor sequence. A single band of higher molecular weight indicates homozygous editing of both alleles. Amplicons from homozygous clones were sent for Sanger sequencing using the amplicon sequencing service provided by GENEWIZ.

RNA-seq

HEK293T cells were plated in 10-cm tissue culture dishes (Corning 430167) and cultured under standard conditions (37°C, 5% CO₂). Six experimental conditions were established in biological triplicate. The first four were processed together: (1) DMSO control for acyclovir treatment, (2) DMSO control for branaplam treatment, (3) acyclovir treatment (250 μM), and (4) branaplam treatment (100 nM). Two additional conditions—a DMSO control and a 500 μM acyclovir treatment—were processed in a subsequent batch following the same protocol. After three days of treatment, total RNA was extracted using TRIzol LS Reagent (Invitrogen 10296010) following the manufacturer’s protocol, treated with DNase I (NEB M0303S), and purified using the RNA Clean & Concentrator-25 kit (Zymo R1018).

RNA-seq libraries were prepared with the Illumina TruSeq Stranded Total RNA Library Preparation Kit (with rRNA depletion) at the Weill Cornell Medicine Genomics Resources Core Facility. Sequencing was performed on an Illumina NovaSeq X Plus platform, generating ≥70 million 150 bp paired-end reads per sample. The 500 μM acyclovir group generated ≥ 32 million 150 bp paired-end reads per sample. Raw FASTQ files were processed using nf-core/rnasplice (v1.0.4), a standardized and reproducible Nextflow-based pipeline (v24.10.0)67. Reads were aligned to the GRCh38 human genome reference, and all pipeline parameters—including trimming, alignment, quantification, and splicing analysis—were retained at their default settings.

For alternative splicing analysis, rMATS results from nf-core/rnasplice were used to compare branaplam-treated samples to branaplam controls and acyclovir-treated samples to acyclovir controls. Splicing events were quantified from rMATS output files reporting inclusion junction counts (IJC) and skipping junction counts (SJC) for each event. Events were retained if either condition had all replicates with IJC and SJC counts ≥ 10, ensuring robust splicing metric estimation. Significant alternative splicing events were defined as those with false discovery rate (FDR) < 0.05 and |Δ percent spliced in (PSI)| > 0.1. FDR was calculated using the Benjamini-Hochberg method. Individual splicing patterns were visualized using ggSashimi⁷³ by merging triplicates and displaying mean coverage based on STAR-aligned BAM files and genomic coordinates extracted from the rMATS results.

For differential gene expression analysis, raw read count data were obtained from featureCounts output files generated by nf-core/splice. Raw counts were processed to retain only Ensembl gene identifiers. Counts for each gene were aggregated across samples and merged into a single count matrix for analysis with DESeq2. Genes with low expression (row sum ≤10) were removed prior to analysis. Differentially expressed genes were defined as those with an adjusted p-value < 0.05 and |log₂ fold change| > 1.

All software versions, parameters, and computational workflows are fully documented in the nf-core/rnasplice pipeline67. Complete DGE and rMATS datasets for branaplam- and acyclovir-treated samples are provided as additional files.

Statistics and Reproducibility

All experiments, including micrographs and immunoblots, were independently repeated three times with similar results.