Counteracting chromatin effects of a splicing-correcting antisense oligonucleotide improves its therapeutic efficacy in spinal muscular atrophy

Summary

Spinal Muscular Atrophy (SMA) is a motor-neuron disease caused by mutations of the SMN1 gene. The human paralog SMN2, whose exon 7 (E7) is predominantly skipped, cannot compensate for the lack of SMN1. Nusinersen is an antisense oligonucleotide (ASO) that upregulates E7 inclusion and SMN protein levels by displacing the splicing repressors hnRNPA1/A2 from their target site in intron 7. We show that by promoting transcriptional elongation, the histone deacetylase inhibitor VPA cooperates with a nusinersen-like ASO to promote E7 inclusion. Surprisingly, the ASO promotes the deployment of the silencing histone mark H3K9me2 on the SMN2 gene, creating a roadblock to PolII elongation that inhibits E7 inclusion. By removing the roadblock, VPA counteracts the chromatin effects of the ASO, resulting in higher E7 inclusion, without large pleiotropic effects. Combined administration of the nusinersen-like ASO and VPA in SMA mice strongly synergized in SMN expression, growth, survival, and neuromuscular function.

Introduction

Spinal Muscular Atrophy (SMA) is a motor-neuron disease caused by loss-of-function mutations of the SMN1 gene. Humans have a paralog, SMN2, whose exon 7 (E7) is predominantly skipped, and so it cannot fully compensate for the lack of SMN1. Both SMN2 and SMN1 are about 30 kbp-long each but they differ in only 11 bp, which indicates a recent duplication in evolution. The only critical difference between the two SMN genes is the C>T base change 6 bp inside exon 7 that causes E7 skipping (Monani et al., 2000), giving rise to a truncated and inactive version of the SMN protein. Both genes encode a splicing silencer sequence in intron 7 that is the target site for the negative splicing factors hnRNPA1 and A2 that contribute to E7 skipping in SMN2. This basic mechanistic knowledge led to the development of an antisense oligonucleotide (ASO) therapeutic strategy for SMA: Nusinersen (Spinraza) is a splicing-correcting ASO drug approved for clinical use. Nusinersen targets the splicing silencer located in SMN2 intron 7 pre-mRNA and, by blocking the binding of hnRNPA1 and A2, it promotes higher E7 inclusion, increasing SMN protein levels. Nusinersen was the first drug approved for SMA therapy and also the first-splicing-corrective drug. Although nusinersen is administered to patients by lumbar puncture, to readily reach motor neurons, systemic administration robustly rescues severe symptoms in an SMA mouse model (Hua et al., 2008, 2011, 2015) which suggests that restoring adequate SMN levels in peripheral tissues might be therapeutically important.

Alternative splicing is not only regulated by the binding of activators and repressors to splicing enhancers and silencers, but also by the rate of RNA polymerase II (RNAPII) transcriptional elongation and chromatin structure. According to the kinetic coupling mechanism, slow elongation can promote either exon inclusion or skipping, depending on the type of exon (de la Mata et al., 2003; Dujardin et al., 2014, Fong et al., 2014). In class I exons, slow elongation promotes inclusion by improving the recruitment of constitutive or regulatory splicing factors to the nascent mRNA, whereas in class II exons slow elongation enhances the binding of inclusion inhibitory factors to their target sites. Genome-wide studies showed that approximately 20% of a cell’s alternative splicing events are sensitive to elongation and, depending on the cell type, class I exons represent between 50 and 80%, while class II exons are between 20 to 50% of the elongation-sensitive alternative exons (Fong et al, 2014; Ip et al., 2011, Muñoz et al., 2009, 2017; Maslon et al., 2019). On the other hand, intragenic deployment of specific histone marks can create more compact or more relaxed chromatin regions that inhibit or promote RNAPII elongation, respectively and consequently may affect alternative splicing. For example, acetylation of histone H3 lysine 9 (H3K9Ac) along gene bodies promotes skipping of class I exons (Schor et al., 2009), whereas dimethylation of the same residue (H3K9me2) promotes their inclusion (Alló et al., 2009; Schor et al., 2013). Conversely, intragenic H3K9 acetylation promotes inclusion of class II exons, which is consistent with the fact that treatments that inhibit elongation promote their skipping (Dujardin et al., 2014).

We show here that SMN2 E7 is a class II alternative exon, which opened the way to explore the regulation of its inclusion using kinetic coupling tools. We found that by promoting transcriptional elongation, histone deacetylase (HDAC) inhibitors cooperate with a nusinersen-like ASO to upregulate E7 inclusion. Surprisingly, the ASO also elicits the deployment of the silencing histone mark H3K9me2 on the SMN2 gene, creating a roadblock to RNA polymerase II elongation that acts negatively on E7 inclusion. By removing the roadblock, HDAC inhibition counteracts the undesired chromatin effects of the ASO, resulting in higher E7 inclusion. Combined systemic administration of the nusinersen-like ASO and HDAC inhibitors in neonate SMA mice had strong synergistic effects on SMN expression, growth, survival, and neuromuscular function. Thus, we suggest that HDAC inhibitors have the potential to increase the clinical efficacy of nusinersen, and perhaps other splicing-modulatory ASO drugs, without large pleiotropic effects, as assessed by genome-wide analyses.

Results

Assessment of SMN2 E7 alternative splicing

From a technical point of view it should be noted that the SMN1 and SMN2 genes are present within a 500 kb duplicated region on chromosome 5. Due to the extremely high sequence identity between them, reads of deep sequence analyses and primers used to amplify segments in, for example, ChIP-qPCR protocols (see below) cannot be assigned unequivocally to either SMN gene and so an assessment in parallel is imposed. The exceptions are the end-point RT-PCRs of: (1) the transfected SMN2 minigene (Lorson et al., 1999); (2) the human SMN2 transgene in mice (see below); and (3) the human endogenous SMN2 gene. In the latter, exon 8 of SMN2 contains a DdeI restriction site that is absent in SMN1. Hence, RT-PCR followed by restriction digestion with DdeI distinguishes SMN2 and SMN1 transcripts (Parsons et al., 1996). We will refer to merged SMN1/2 genes when a parallel assessment is performed and as SMN2 alone when a gene-specific method is applied.

Slow elongation inhibits SMN2 E7 inclusion

To evaluate the impact of elongation rate on SMN2 E7 inclusion, we assessed the effects of a slow mutant of RNAPII. We co-transfected HEK293T cells with an SMN2 E7 reporter minigene (Lorson et al., 1999) and a plasmid encoding an α-amanitin-resistant large subunit of RNAPII, with or without the point mutation R749H (de la Mata et al., 2003), previously shown to slow elongation in vivo by 2- to 3-fold (Boireau et al., 2007; Maslon et al., 2019). When transcription was carried out by the slow polymerase, SMN2 E7 skipping was greater than upon transcription by the α -amanitin-resistant wild-type polymerase (Fig. 1A). The use of a minigene splicing reporter is necessary to assess the effects of the slow polymerase by transient co-transfections. However, and more importantly, the effect is also observed on E7 of the endogenous merged SMN1/2 genes as shown by re-analysis of published genome-wide sequencing data (Fong et al., 2014) that confirmed that not only the slow polymerase causes E7 skipping, but that a fast mutant (E1126G) has the expected opposite effect, promoting E7 inclusion (Fig. 1B). Consistently, treatment of cells with the DNA topoisomerase I inhibitor camptothecin (CPT), which indirectly inhibits elongation (Listerman et al., 2006; Dujardin et al., 2014), also promoted E7 skipping from the endogenous SMN2 gene (Fig. 1C). These results indicate that SMN2 E7 is a class II exon.

Figure 1. SMN2 E7 inclusion into the mature mRNA is controlled by RNAPII elongation behaving as a class II exon.

(A) Transcription by a slow RNAPII mutant inhibits SMN2 E7 inclusion. HEK293T cells were co-transfected with the SMN2 reporter minigene and expression vectors for WT^res and hC4 RNAPII (slow mutant), followed by addition of α-amanitin. Alternative splicing was analyzed by radioactive RT-PCR followed by native polyacrylamide gel electrophoresis and autoradiography. Bars display means ± SD of percentage of the radioactivity in the full-length (FL) band, containing E7, over the sum of radioactivity in the FL and ΔE7 (lacking E7) bands of at least three independent transfection experiments. A two-tailed Student’s t test was used to determine the significance between events. (B) Re-analysis of the RNA-seq data published by the Bentley laboratory (Fong et al., 2014). Due to the almost identical DNA sequences of the human SMN1 and SMN2 genes (only 11 bp differences in ~30 kbp), sequencing reads were assigned to a merge of the two genes. For this reason, splicing junction data had to be corrected considering 100% inclusion in the case of the SMN1 gene. Top: “Sashimi plots” indicating the number of reads for the E6–8 (skipping), E7-E8 (inclusion) and E6-E7 (inclusion) junctions of the merged SMN1/2 genes in cells stably transfected with the slow (R749H, in red), fast (E1126G, in green) and WT (in gray) RNAPII large subunit constructs, all of which had a second mutation that confers resistance to α-amanitin, a drug that was used to treat the cells before RNA extraction. Bottom: Raw and corrected quantification of the levels of E7 inclusion as assessed by analysis of the E6-E7 junction reads (top) and E7-E8 junction reads (bottom). Inclusion levels are expressed as both percentage of E7 inclusion and E7⁺/ E7⁻ (i.e. FL/Δ7) ratios. (C) Effect of camptothecin (CPT) on alternative splicing of endogenous SMN2 E7. HEK293T cells were incubated for at least 6 hours with 3 µM CPT. Alternative splicing was assessed as in (A).

Cooperative effects of ASO1 and HDAC inhibitors

In view of the above results, we reasoned that chromatin opening by histone acetylation should foster E7 inclusion by promoting RNAPII elongation. Therefore, we measured the effects of HDAC inhibitors, such as trichostatin A (TSA), on E7 inclusion in HEK293T cells, either alone or in combination with transfection of a nusinersen-like 2’-O-(2-methoxyethyl) (MOE) phosphorothioate-modified antisense oligonucleotide (named ASO1 in this paper) at suboptimal concentrations. Nusinersen (5’-TCACTTTCATAATGCTGG-3′) was originally dubbed ASO10–27 (Hua et al., 2008), and the equally effective ASO1 variant (5′-ATTCACTTTCATAATGCTGG-3′) has a 2-nt extension and can be described as ASO10–29. TSA alone was less potent than ASO1 in upregulating E7 inclusion, but combining both drugs greatly enhanced the effect (Fig. 2A). The cooperative effect of TSA was dose-dependent (Suppl. Fig. 1A) and reproducible in other cell types, such as HeLa (Suppl. Fig. 1B) and SMA patient fibroblasts (Suppl. Fig. 1C). We obtained similar results by combining ASO1 and other HDAC inhibitors such as SAHA (N-hydroxy-N′-phenyl-octanediamide) (not shown) and valproic acid (VPA, Fig. 2B). Also, similar effects were obtained with an ASO with the same sequence as ASO1 but with a different chemical backbone (2’-O-methyl phosphorothionate) (not shown).

Figure 2. HDAC and histone methylation inhibitors potentiate ASO1’s upregulation of E7 inclusion.

Combined effects on endogenous SMN2 E7 alternative splicing of transfection with 25 nM ASO1 and treatment with 3 µM TSA for 24 hr. (A), 10 mM VPA for 24 hr. (B), or 25 µM 5-AZA for 48 hr. (C). Combined effects on HEK293T endogenous SMN2 E7 alternative splicing of transfection with 25 nM each of si-hnRNPA1 and si-hnRNPA2, and treatment with 3 µM TSA for 24 hr. (D) or 10 mM VPA for 24 hr. (E). An oligonucleotide with a ASO1 scrambled sequence was used as control (scrambled oligo). Alternative splicing was assessed as in Fig. 1A.

H3K9me2 is a transcriptionally repressive mark that competes with the transcriptionally permissive H3K9Ac (Ghare et al., 2014). Accordingly, with the caveat of potential indirect effects, inhibition of methylation by 5-aza cytidine (5-AZA), a drug that inhibits both DNA and histone methylation (Wozniak et al., 2007) enhanced the effects of ASO1 as effectively as the HDAC inhibitors (Fig. 2C). Finally, in agreement with the fact that ASO1 displaces the splicing repressors hnRNPA1/A2 from their pre-mRNA target site in intron 7, we found that upon treatment with either TSA (Fig. 2D) or VPA (Fig. 2E), the effects of hnRNPA1/A2 depletion in promoting E7 inclusion were stronger.

VPA was previously assessed in the context of SMA, but not in combination with nusinersen (Swoboda et al., 2010). ChIP analysis (Fig. 3A) shows that VPA promoted intragenic H3K9 acetylation along the merged SMN1/2 genes, with a conspicuous peak around the E7 region, which may explain why E7 inclusion is upregulated, according to the kinetic-coupling mechanism of alternative splicing (Kornblihtt et al., 2013). We ruled out the possibility that the HDAC inhibitors may be acting through modulation of SMN protein levels, which in turn might affect SMN2 E7 splicing (Jodelka et al., 2010), because overexpressing SMN did not alter the effect of ASO1 (Suppl. Fig. 1D)

Figure 3. ASO1 promotes H3K9 dimethylation along the merged SMN1/2 genes, creating a roadblock to RNAPII.

(A) H3K9Ac distribution along the merged SMN1/2 genes, assessed by ChIP-qPCR in HEK293T cells treated (red) or untreated (CTRL, gray) with 10 mM VPA for 12 hr. H3K9me2 (B), total RNAPII (C), P-Ser5 RNAPII (D), and P-Ser2 RNAPII (E) distribution along the merged SMN1/2 genes, assessed by ChIP-qPCR in HEK293T cells transfected with the scrambled oligo (CTRL, gray) or ASO1 (blue and purple) and treated with 10 mM VPA (red and purple). Vertical red arrows indicate the target site of ASO1. Four independent immunoprecipitation replicates were conducted per experiment. Data are represented as mean ± S.D. (n = 4, *p < 0.05, two-tailed Student’s t test).

Chromatin effects of ASO1

Due to the fact that ASO1 is a single-stranded oligonucleotide that hybridizes with an intronic sequence at the pre-mRNA level, we wondered if it could have chromatin effects similar to those of siRNAs directed to intronic regions, as previously described (Alló et al., 2009, Schor et al., 2013). These reports showed that siRNAs whose guide strands are complementary to pre-mRNA intronic sequences (antisense) located downstream of an alternative exon regulate the splicing of that exon by promoting silencing marks, such as H3K9me2, which in turn act as roadblocks to RNAPII elongation. This effect was shown to be counterbalanced by factors that favor chromatin opening or transcriptional elongation. Remarkably, here we observed that transfection of HEK293T cells with ASO1 promoted extensive H3K9 dimethylation along the merged SMN1/2 genes, with peaks that reached an 8-fold increase at the promoter and the alternative E7 areas (Fig. 3B, cf. blue and gray lines). Other histone marks, like H3K27me3 or H3K9Ac, were not affected by ASO1 transfection (Suppl. Fig. 1E). Most importantly, VPA not only reduced the H3K9me2 marks relative to the control (Fig. 3B, cf. red and gray lines) but also completely abolished the effect of ASO1 (Fig. 3B, cf. blue and purple lines), confirming antagonistic roles of histone methylation and acetylation.

We next tested whether RNAPII densities were altered. Transfection with ASO1 greatly increased total, P-Ser5 and P-Ser2 (Figs. 3C–E) RNAPII densities at the promoter and at a distinct peak near the ASO1 target site on the transcript. In all three cases, the RNAPII peak was abolished when cells were additionally treated with VPA.

RNase H, an enzyme that degrades RNA in RNA-DNA hybrids, did not alter the effect of ASO1 on SMN2 E7 inclusion (Suppl. Fig. 1F). Moreover, the effect of ASO1 was unchanged by depletion of AGO1 (Suppl. Fig. 1G), known to be required for the chromatin effects of siRNA (Alló et al., 2009).

Uncoupling the two opposite roles of ASO1

The above results suggest a model whereby in the absence of HDAC inhibitors, ASO1 has two opposing effects on E7 inclusion (Fig. 4A, left panel). It promotes inclusion by blocking hnRNPA1 and A2 binding to the pre-mRNA, but concomitantly results in a compact chromatin structure and a roadblock to elongation that in turn promotes E7 skipping. Although both effects co-exist, at high ASO1 concentrations, the chromatin effect is evidently surpassed by the hnRNPA1/A2 effect. On the other hand, in the presence of HDAC inhibitors (right panel), the chromatin-silencing effect is abrogated, and the levels of E7 inclusion increase further. To test the validity of this model, we investigated the effects of a second ASO, ASO2, whose target site is also located in intron 7, but downstream of the ASO1 target site (Suppl. Fig. 2A). ASO2 bears no sequence identity with ASO1, and does not overlap hnRNPA1/A2 binding sites. Even so, if ASO2 elicits the chromatin effect, it should inhibit E7 inclusion. Fig. 4B shows that, whereas ASO1 increases, ASO2 inhibits E7 inclusion. Furthermore, similarly to ASO1, ASO2 promoted H3K9 methylation (Fig. 4C) and higher total, P-Ser5, and P-Ser2 (Figs. 4D–F) RNAPII densities, consistent with roadblocks to elongation that in all cases were abolished by treatment with VPA. As specificity controls, neither ASO1 nor ASO2 promoted H3K9 dimethylation in a gene located in the same topologically-associated domain (TAD) as SMN1 and SMN2 (Lefebvre et al., 1995) (Suppl. Fig. 2B), or in genes located outside the SMN1 and SMN2 TAD with either high (Suppl. Fig. 2C) or low (Suppl. Fig. 2D) basal levels of H3K9 methylation.

Figure 4. Uncoupling the two opposing effects of ASO1: ASO2 also promotes H3K9 dimethylation and imposes an RNAPII roadblock, but inhibits SMN2 E7 inclusion.

(A) Model of two opposing roles of ASO1. In the absence of HDAC inhibition, ASO1 promotes H3K9 dimethylation and subsequent chromatin compaction (left); in the presence of HDAC inhibitors, H3K9 acetylation increases and chromatin becomes more relaxed, counteracting the compaction promoted by ASO1 (right). (B) Effects on endogenous SMN2 E7 alternative splicing of transfecting HEK293T cells with 25 nM ASO2 using the scrambled oligo and ASO1 as negative and positive controls respectively. Alternative splicing was assessed like in Fig. 1A. H3K9me2 (C), total RNAPII (D), P-Ser5 RNAPII (E), and P-Ser2 RNAPII (F) distribution along the merged SMN1/2 genes, assessed by ChIP-qPCR in HEK293T cells transfected with the scramble oligo (CTRL, gray) or ASO2 (orange and purple) and treated with 10 mM VPA (red and purple). Vertical red arrows indicate the target site of ASO2. Four independent immunoprecipitation replicates were conducted per experiment. Data are represented as mean ± S.D. (n = 4, *p < 0.05, two-tailed Student’s t test).

To confirm the mechanism proposed in Fig. 4A, we next performed CRISPR ablation of the gene encoding the histone H3K9 dimethyl transferase G9a (Fiszbein et al., 2016). The effectiveness of this gene knock out was confirmed by western blot (Fig. 5A). Notably this promoted the upregulation of SMN2 E7 inclusion by ASO1 (Fig. 5B).

Figure 5. Characterization of the chromatin effects of ASO1.

(A-B) Ablation of the H3K9 dimethyl transferase G9a potentiates upregulation of SMN2 E7 inclusion by ASO1. (A) Western blot showing the absence of G9a protein in the cells in which the G9a gene was knocked out. (B) Alternative splicing was analyzed, as in Fig. 1A, in WT and CRISPR G9a-ablated HEK293T cells using the scrambled oligo as control. (C-D) Chromatin effects of ASOs are sequence-specific and not restricted to the SMN genes or to introns. (C) H3K9 dimethylation on three regions of the ELP1 gene (amplicons a, b and c) assessed by ChIP-qPCR in HEK293T cells transfected with the scrambled oligo (CTRL, gray), ASO1 (blue and purple) and ASOs targeting intron 19 (i19, left panel) or exon 20 (e20, right panel). (D) H3K9 dimethylation on two regions of the FOXM1 gene (amplicons a and b) assessed by ChIP-qPCR in HEK293T cells transfected with the scrambled oligo (CTRL, gray), ASO1 (blue and purple) and an ASO targeting the intron 8/exon 9 junction (5’SS-1).

These results with ASO1 and ASO2 still leave open the question as to whether the chromatin effects of these ASOs reflect a general phenomenon or a particular one restricted to the SMN gene, to its intron 7, or to introns in general. To this end we assessed the effects of ASOs, directed at two other genes, that were previously used in splicing experiments: the ELP ASOs were designed by the Krainer lab as controls to modulate E20 inclusion of the IKBKAP gene carrying a point mutation that causes familial dysautonomia (Sinha et al., 2018); the FOXM1 ASOs were shown by the Valcárcel lab to promote skipping of E9 in the FOXM1 gene (Martín et al., 2021). Notably, these ASOs, both unrelated in function, sequence or chromosome location to the SMN1 or SMN2 genes, also enhance deployment of the H3K9me2 mark around their respective target sites. As relevant controls, ASO1 has no effect on either ELP1 or FOXM1 (Figs. 5C and 5D).

Genome-wide analyses

We performed genome-wide characterization of the effects of ASO1 and VPA in HEK293T cells, including nascent-mRNA profiles, mRNA expression, and alternative splicing, using the recently developed POINT-seq technology (Sousa-Luís et al., 2021) as well as more standard RNA-seq. Metagene analysis of POINT-seq on 2741 expressed protein-coding genes confirmed that VPA dramatically changed the pre-mRNA read-count ratio between the promoter region (P) and the gene body (GB) (Fig. 6A), whereas ASO1 did not (Fig. 6B). Similar results were obtained when assessing the number of genes with different P/GB values for control versus VPA- or ASO1-treated cells (Fig. 6C).

Figure 6. Global effects of VPA and ASO1 on transcription elongation and gene expression.

Metagene analysis of POINT-seq signal in normalized transcription units from the transcription start site (TSS) to the cleavage and polyadenylation site (pA) of non-overlapping protein-coding genes of HEK293T cells treated or not with VPA (A) or ASO1 (B). (C) Histograms representing the log2 ratio distribution of read density between the promoter-proximal region (P) and gene body (GB) for POINT-seq samples. The VPA experiment is shown on the left, and ASO1-treated cells on the right. Invalid log2 values were avoided by adding 1 unit to each ratio. (D-F) RNA-seq volcano plots showing differentially expressed genes in HEK293T cells treated with VPA (D), ASO1 (E) and VPA + ASO1 (F) with respect to control untreated cells. Only red points represent significantly differentially expressed genes, by showing a p-value lower than 1 × 10⁻⁵ and an absolute log2FoldChange value higher than 1. Green points represent genes for which a fluctuation between conditions was observed, but it was inconsistent between replicates, either due to technical variability or a low number of reads. Grey dots are genes with log2FoldChange smaller than 1. Three replicates per condition were employed. (H-G) Genome browser views of POINT-seq for the PLCB4 and PAK5 genes (G), and the merged SMN1/2 genes (H) of HEK293T cells treated with or without with VPA.

RNA-seq analysis revealed that only modest changes in gene expression and alternative splicing were detected. Thus only 69 genes out of 29,000 transcription units analyzed were significantly altered by VPA treatment (Fig. 6D). Notably none of the 29,000 transcription units significantly changed expression in response to ASO1 (Fig. 6E), and the combined effects of VPA + ASO1 were also minimal (104 genes over 29,000 transcription units, Fig. 6F).

Similarly, RNA-seq revealed that VPA only affected the patterns of 48 of the 79,549 alternative splicing events analyzed (0.06%, Suppl. Figs. 3A and B), when a relatively high threshold (ΔPSI>30%) is applied, and confirmed the cooperative effects of ASO1 and VPA on merged SMN1/2 genes E7 inclusion (Suppl. Figs. 3C and D). In conclusion, these data are consistent with a global increase in transcription elongation caused by VPA, as previously shown for individual genes, where elongation speed increased from ~2 to ~4 kb/min the with HDAC inhibitor TSA (Dujardin et al., 2014), but with no conspicuous global changes in gene expression and alternative splicing. In parallel, transfection with ASO1 had no global effects on elongation or gene expression.

The recently designed POINT-seq method assesses the distribution of nascent mRNA reads associated to elongating RNAPII along transcribed genes. As mentioned above, it becomes impossible to assign POINT-seq reads individually to each of the two genes. For these reasons, POINT-seq analysis must necessarily refer to the merged SMN1/2 genes. In spite of these limitations, POINT-seq data showed that, unlike ASO1, VPA reduced the ratio of read counts at the Promoter over the Gene Body (P/GB ratio) from 4.17 to 1.73 (Suppl. Fig. 3E), consistent with the global effect of VPA on elongation. However, when focusing on the region around ASO1’s target site (Suppl. Fig. 3F), we clearly observe that ASO1 increased the ratio of reads mapping upstream versus downstream of the target site (U/D ratio), which is highly consistent with the roadblock to elongation evidenced by RNAPII ChIP (Fig. 3).

To assess if the increase in elongation by VPA is accompanied by an increase in overall transcription of the merged SMN1/2 genes, we selected three genes from the RNA-seq data of Fig. 6D: one whose expression is augmented by VPA (PAK5); another whose expression unaffected by VPA (PLCB4); and the merged SMN1/2 genes and compared their POINT-seq transcription profiles with and without VPA. Similar to PLCB4 and unlike PAK5, POINT-seq profiles of the merged SMN1/2 genes indicate that its transcription is unaltered by VPA (Figs. 6 G and H).

The increase in H3K9 dimethylation of the merged SMN1/2 genes caused by ASO1, shown by ChIP-qPCR (Fig. 4C), was confirmed by ChIP-seq analysis (Suppl. Fig. 4). ASO1 caused a nearly 10-fold increase in the H3K9me2 signal around its target site, compared to a 2-fold increase in the rest of the gene (regions B and A, respectively in Suppl. Figs. 4A and 4B). Consistent with above results (Fig. 4C), ASO1’s effect was abolished by VPA. As a negative control, neither ASO1 nor VPA affected the deployment of the H3K9me2 mark in the β-actin gene (Suppl. Fig. 4C).

Combined treatment in SMA mice

Nusinersen, administered subcutaneously in newborn SMA mice, distributes broadly and strongly rescues survival and motor function (Hua et al., 2011). We therefore sought to test the combined effect of ASO1 and VPA treatment using a mouse model of severe SMA. Mice have only one Smn gene, and Smn^−/− mutants are embryonic lethal (Hsie-Li et al., 2000; Monani et al., 2000). We used Smn-null transgenic mice containing two copies of the human SMN2 transgene, which develop a severe SMA-like phenotype with a mean survival of ~7–10 days (Hua et al., 2011). We administered two consecutive low-dose subcutaneous injections of ASO1 (18 µg/g) at P0 and P1 and/or one subcutaneous injection of VPA (10 µg/g) at P1. The rationale for the choice of a low or suboptimal dose of ASO1 is to elicit a partial phenotypic effect that allows us to evaluate improvements by co-administration of HDAC inhibitors. Since the generation of mice with the two mouse SMN alleles disrupted is performed by crossing of Smn^+/− heterozygous and the Smn^−/− (SMA mice) and Smn^−/+ phenotypes are not distinguished at birth, injections were blind with respect to the genotypes and genotyping was performed after injections. Fig. 7A shows that SMA mice injected with VPA alone died before P10, similarly to vehicle-treated controls. The suboptimal dose of ASO1 alone extended median survival to ~20 days, while co-injection of VPA with ASO1 increased the median survival to ~70 days. All ASO1-alone-injected mice were dead at P62, whereas ~60% of those treated with ASO1 and VPA were still alive at P62. The difference between the two treatments is statistically significant (p = 0.00081).

Figure 7. ASO1 and VPA have synergistic effects in SMA mice.

Kaplan-Meier survival plot (A) and growth curves (B) of SMA mice, following subcutaneous administration at P0 and P1 of 16.8 μg ASO1 (n=18) or saline (n=14), one subcutaneous dose at P1 of 10 µg per g of body weight VPA (n=12), or both treatments together (n=20). ASO1-treated Smn^+/− heterozygote littermates (n=15) served as controls. Righting reflex (C) and grip strength (D) tests of P7 SMA animals, treated as indicated in (A) and (B). In the righting-reflex test, we measured the time it takes a mouse to right itself when placed on its back on a flat surface. In the grip-strength test, we measured the angle at which the pups fall from a tablet with a rough surface to which they hold on by their forelimbs. Scramble (Ctrl) (n=12, 40 trials), VPA (n=13, 50 trials), ASO1 (n=11, 50 trials), ASO1 + VPA (n=17, 68 trials), and untreated heterozygotes (n=12, 36 trials). Statistical significance was analyzed by two-way repeated measures ANOVA. P<0.05 was considered statistically significant; data are represented as mean + SD.

Body weight gain in the surviving mice was also greatly improved by VPA plus ASO1 injections, compared to ASO1 alone (Fig. 7B). We obtained similar results with a combined treatment of ASO1 and TSA (Suppl. Figs. 5A and B). Western blot analysis shows that the combined treatment greatly increased the levels of human SMN protein in liver, kidney, skeletal muscle, spinal cord, brain, and heart (Suppl. Fig. 5C–H). The largest effect on SMN expression was in the liver (Suppl. Fig. 5C), underscoring the proposed role of this organ in SMA pathogenesis (Hua et al., 2011).

Suppl. Fig. 6 illustrates conspicuous differences in size, posture, glassy eyes, and absence of grooming between the ASO1 alone and the ASO1+TSA-treated mice at 11, 14, 17 and 42 days after birth. Both the untreated and TSA-only-treated mice died around P7-P8.

We estimated the hazard ratio (HR) of the combined treatment versus a single drug treatment or no treatment at all. An HR greater than 1 suggests an increased risk, and an HR below 1 suggests a smaller risk. We obtained an HR = 0.19 for the combined treatment with VPA + ASO1 versus ASO1 alone, and an HR = 0.12 for the combined treatment with TSA + ASO1 versus ASO1 alone. Thus, far from being risky, these data demonstrate that the combined treatments are highly beneficial.

In addition, we performed two noninvasive neuromuscular function tests, appropriate for mouse pups. In the righting reflex test (Feather-Schussler et al., 2016), whereas wild-type mice righted immediately, untreated mutant mice took ~45 seconds. This delay was not significantly changed by VPA treatment, was greatly reduced by ASO1 treatment (~10 seconds), and was virtually eliminated by the combined treatment (Fig. 7C, Supplemental video 1). In the grip-strength test, animals treated with both ASO1 and VPA could hold on at much higher inclination angles (~ 70°), compared with the untreated mutants (~35°) or mice treated with VPA alone (~30°) or limiting ASO1 alone (~45°) (Fig. 7D, Supplemental video 2).

As in cultured cells (Fig. 3), ASO1 injection into SMA mice also promoted higher H3K9 dimethylation and RNAPII accumulation on the SMN2 transgene in liver and brain (Suppl. Fig. 7). This result strongly suggests that the mechanism we elucidated in cultured cells also underlies the survival and behavioral effects observed in mice.

Discussion

This study shows that SMN2 E7 is a class II alternative exon, whose inclusion into the mature mRNA is inhibited by a slow mutant of RNAPII and increased by a fast mutant. Consistently, by promoting transcriptional elongation, histone deacetylase inhibitors cooperate with the nusinersen-like antisense oligonucleotide ASO1 to enhance SMN2 E7 inclusion. These effects were not only observed in human cells in culture, including SMA-patient fibroblasts, but also in a transgenic mouse model of SMA in which the combined administration of ASO1 and HDAC inhibitors improved growth, survival and neuromuscular function, compared to either drug alone. Among HDAC inhibitors, unlike TSA and SAHA, VPA is approved for clinical use (Wirth et al., 2006) and has been previously used, without any ASO, in clinical trials that failed to demonstrate efficacy (Swoboda et al., 2010; Kissel et al., 2011). The rationale for VPA’s clinical testing was that chromatin opening at the SMN2 promoter would increase transcription and therefore SMN levels (Kernochan et al., 2011). In our experiments, VPA increases elongation (Fig. 6A) but not overall transcription (Fig. 6G and 6H) of the SMN gene. It has been previously shown (Schor et al., 2009) that an increase in intragenic elongation does not necessarily imply an increase in gene transcription and therefore total mRNA levels. This is because such treatment may not further increase the recruitment of RNAPII to the promoter. Thus, changes in elongation rates may affect the quality (splicing isoform ratio) but not quantity of the encoded mRNA. Results in cells coincide with a lack of corrective effect of VPA alone in our SMA mouse experiments (Fig. 7 and Suppl. Fig. 5). In particular, administration of VPA alone did not stimulate the production of SMN protein tested by Western blot in six different organs of the treated mice, but strongly stimulated ASO1’s effect (Suppl. Fig. 5. C–H). It should be mentioned that our mouse experiments involved a single dose of HDAC inhibitors alone or together with ASO1 at postnatal days 1 or 2, whereas HDAC inhibitors alone have shown slight beneficial effects in mice when dosed chronically (Avila et al., 2007, Chang et al., 2001).

Recently, an additive effect on E7 inclusion of the HDAC inhibitor LBH589 and a nusinersen-like phosphorodiamidate morpholino oligomer (PMO) was reported for SMA patient-derived fibroblasts and transgenic-mouse neural stem cells (Pagliarini et al., 2020). Similarly, the combination of VPA with a splice-switching PMO conjugated to a cell-penetrating peptide was shown to promote more E7 inclusion in patient fibroblasts than either agent alone (Farrelly-Rosch et al., 2017). However, neither study explored the underlying mechanism nor tested animal models.

Our mechanistic analysis in this paper reveals that the nusinersen-like ASO promotes the deployment of the silencing histone mark H3K9me2 on the SMN2 gene, especially around its target site, similarly to reported nuclear effects of siRNAs that target pre-mRNA sequences creating a roadblock to RNAPII elongation that affects alternative splicing (Alló et al., 2009; Schor et al., 2013). We show that both the increase in H3K9 dimethylation and the accumulation of RNAPII at ASO1’s target region are abrogated by HDAC inhibitors. ASO1 triggers accumulation of total RNAPII and its two main phosphoisoforms, suggesting that it acts as a general steric impediment, rather than regulating RNAPII phosphorylation. The chromatin effect of ASO1 does not appear to involve R-loop formation (Tang-Wong et al., 2019) because the effect of ASO1 is not altered by overexpression of RNase H, an enzyme that degrades RNA in RNA-DNA hybrids (Suppl. Fig. 1F). This suggests that ASO1 does not act through hybridization to DNA, as opposed to RNA, consistent with the previous demonstration that an oligonucleotide with complementary sequence to nusinersen (sense oligo) has no effect on SMN2 E7 inclusion (Rigo et al., 2014).

These findings reveal unforeseen alterations to the deployment of histone marks induced by ASO treatment at the targeted gene. Most importantly, we show that the chromatin effects of the ASOs are sequence-specific, not restricted to intronic regions or to the merged SMN1/2 genes (Figs. 5C and D) and therefore, potentially able to affect, either positively or negatively, other ASO-based therapies. In particular, the uncoupling experiments with SMN2 ASO2 (Figs. 4B–F) confirm the model of Fig. 4A, in which the nusinersen ASO1 has two opposite effects on SMN2 E7 inclusion: a positive one through displacement of the negative splicing factors hnRNP A1/A2; and a negative one through triggering the deployment of the H3K9me2 mark. In this context, promotion of histone acetylation mitigates the negative effect at suboptimal ASO1 concentrations.

We support the idea that co-transcriptional base-pairing of ASOs with their target sites on pre-mRNAs promote local recruitment of histone methylating enzymes. However, further investigation will be necessary to decipher in depth the precise mechanism by which ASOs trigger H3K9 dimethylation.

Genome-wide analyses reinforce the evidence that the combined treatment with ASO1 and HDAC inhibitors enhances SMN2 E7 inclusion, accompanied by only marginal alterations—both in the number of genes affected and in the magnitude of the change—of global gene expression and alternative splicing (Fig. 6 and Suppl. Fig. 3).

Our results pave the way towards clinical assessments in patients. Although we expected VPA to have large pleiotropic effects, our RNA-seq data of cells in culture indicated, in our conditions, that global effects on gene expression and splicing were minimal, which is consistent with our results in mice and with the fact that VPA is currently used to treat several neurological disorders (Chen et al., 2014). In any case, the drug concentrations and short times of VPA treatment we employed were sufficient to enhance the effect of ASO1 on splicing, without major alterations to gene expression. In fact, comparable RNA-seq results were obtained in other studies, under similar VPA-treatment conditions. For example, Balasubramanian et al. (2019) reported that treatment of RN46A cells with 0.5 mM VPA for 72 h caused regulation of only 88 of the ~16,500 genes analyzed. However, treatment of rats with VPA doses 60 times higher than the ones we used in mice here (see below) change expression of about 3,000 genes (Zhang et al., 2018). This implies a concentration-dependent effect on the gene-regulation.

Compared to other HDAC inhibitors, VPA may be a good choice, because it was reported that at molar concentrations 1000-fold lower than VPA, other HDAC inhibitors like SAHA elicit expression changes in 10 times more genes than VPA, strongly affecting many signaling pathways (Lunke et al., 2021). In addition, although VPA has limited efficacy at best in SMA patients, it is safe and well tolerated (Elshafay et al., 2019). The lack of an effect per se in SMA mice (Figs. 7A and B) contrasting with the strong effect in improving ASO1’s therapeutic effects, suggests that an optimal VPA dose can be established that minimizes potential toxicities in patients. We speculate that systemic VPA administration will improve the efficacy of nusinersen, particularly in peripheral tissues. SMN is a ubiquitously expressed protein playing fundamental roles in splicing of all tissues. However, currently nusinersen is only administered via intrathecal injection to mainly reach the central nervous system. We foresee that, at low concentrations of nusinersen present in peripheral tissues, due to CSF clearance after intrathecal injection (Chiriboga et al., 2016), the combined systemic treatment with VPA or other HDAC inhibitors may increase SMN levels in the periphery, as shown in Suppl. Figs. 5C–H, and improve nusinersen performance in peripheral tissues.

Finally we would like to underscore that our results support the relevance of the kinetic coupling between pre-mRNA processing and transcriptional elongation in a whole organism. Except for the alternative splicing/elongation response to light and dark in whole Arabidopsis plants (Godoy Herz et al., 2019) and non-physiological intervention with the slow polymerase mutant in mice (Maslon et al., 2019), most research supporting kinetic coupling was previously performed in cultured cells.

Limitations of this study

In our view, the main limitation of this proof-of-principle study is that we tested only one dosing protocol for the combined administration of the nusinersen-like ASO1 and HDAC inhibitors in SMA mice. This protocol involved a single subcutaneous injection of the drugs at P0/P1, alone or in combination, which proved to be effective in demonstrating the advantages of combined treatment on growth, survival and motor function. The blood-brain barrier is immature at this stage, so ASOs do reach the CNS and very effectively rescue the motor-neuron defects, with a long duration of action (Hua et al., 2011, 2015). However, since nusinersen is administered directly to CSF (intrathecally) in patients, additional studies will be necessary to adjust the effects of intrathecal administration of nusinersen combined with systemic administration of VPA, including the demonstration that nusinersen not only increases H3K9me 2 levels in whole tissues (Suppl. Fig. 7) but also that VPA reduces these levels.

STAR Methods

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Dr. Alberto R. Kornblihtt (ark@fbmc.fcen.uba.ar)

Data and code availability

The raw and processed data derived from POINT-seq, RNA-seq and ChIP-seq analyses as generated in this study are deposited in NCBI GEO (GSE167762). All code supporting POINT analyses are available on request. The reanalyzed published data used in this study can be found at GEO as indicated in the Key Resources Table.

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Mouse anti-hSMN	BD Biosciences	Cat#: 610646
Rabbit Polyclonal - Isotype Control (ChIP Grade)	Abcam	Cat#: ab171870
Mouse monoclonal anti-α-Tubulin	Sigma-Aldrich	Cat#: T5168
IRDye 800CW Goat anti-Mouse IgG	Licor	Cat#: 26-32210
IRDye 800CW Goat anti-Rabbit IgG	Licor	Cat#: 926-32211
Mouse monoclonal anti Histone H3K9me2 (di methyl K9)	Abcam	Cat#: ab1220
Rabbit polyclonal anti Histone H3K9ac (acetyl K9)	Abcam	Cat#: ab4441
Rabbit monoclonal anti Histone H3	Abcam	Cat#: ab12079
Mouse Monoclonal anti c-Myc	Milipore	Cat#: MABE282
Rabbit polyclonal anti RNase H1	Proteintech	Cat#:15606-1-AP
Rabbit polyclonal anti actin	Sigma-Aldrich	Cat#: A2066
Rabbit monoclonal anti G9A	Cell Signaling Technology	Cat#: 3306S
Rabbit monoclonal anti Ago1	Cell Signaling Technology	Cat#: 5053S
Rabbit monoclonal anti Rpb1 NTD (D8L4Y)	Cell Signaling Technology	Cat#: 14958S
Rabbit monoclonal Phospho-Ser2 CTD (E1Z3G)	Cell Signaling Technology	Cat#: 13499S
Rabbit monoclonal Phospho-Ser5 CTD (D9N5I)	Cell Signaling Technology	Cat#: 13523S
Mouse monoclonal anti RNAPII (8WG16)	Santa Cruz	Cat#: sc-56767
Rabbit polyclonal anti H3K9me2 (di methyl K9)	Active Motif	Cat#: 39753
Bacterial and virus strains
DH5α chemically competent E. coli cells	Thermo Fisher Scientific	Cat#: EC0111
Chemicals, peptides, and recombinant proteins
Dulbecco’s modified Eagle’s medium (DMEM)	Thermo Fisher Scientific	Cat#: 12800017
Lipofectamine 2000	Thermo Fisher Scientific	Cat#: 11668500
Opti-MEM Reduced Serum Medium	Thermo Fisher Scientific	Cat#: 11058021
PBS	Thermo Fisher Scientific	Cat#: 10010023
Trypsin/EDTA Solution	Thermo Fisher Scientific	Cat#: R001100
Trichostatin A (TSA)	Sigma-Aldrich	Cat#: T8552
Valproic Acid (VPA)	Sigma-Aldrich	Cat#: P4543
5-Azacytidine (5-AZA)	Sigma-Aldrich	Cat#: A2385
α-Amanitin	Sigma-Aldrich	Cat#: A2263
Dimethyl Sulfoxide - Calbiochem (DMSO)	Merck	Cat#: 317275-500ML
TRIzol Reagent	Thermo Fisher Scientific	Cat#: 15596018
M-MLV Reverse Transcriptase	Invitrogen	Cat#: 28025021
SuperScript III Reverse Transcriptase	Invitrogen	Cat#: 18080044
GoTaq DNA Polymerase	Promega	Cat#: M7848
Dynabeads Protein G	Thermo Fisher Scientific	Cat#: 10009D
Dynabeads M-280 Sheep Anti-Mouse	Thermo Fisher Scientific	Cat#: 11202D
Complete EDTA-free protease inhibitor cocktail	Sigma-Aldrich	Cat#: 05056489001
Critical commercial assays
NEBNext Ultra II DNA Library Prep Kit for Illumina (NEB)	New England Biolabs	Cat#: E7645L
NovaSeq 6000 platform (Illumina) by.	Novogene UK
Deposited data
Genome-wide data are available under GEO number GSE167762	This paper	GEO: GSE167762
Human reference genome Ensambl GRCh38 Homo sapiens	N/A	N/A
FlyBase dm6 genome assemblies Drosophila melanogaster	N/A	N/A
Experimental models: Cell lines
HEK293T	ATCC	Cat#: CRL-3216
HeLa	ATCC	Cat#: CRM-CCL-2
SMA type I homozygous fibroblasts (3813)	Coriell Cell Repositories	Cat#: GM03813
SMA type I carrier fibroblasts (3814)	Coriell Cell Repositories	Cat#: GM03814
HEK293T G9a KO	Fiszbein et al., 2016	N/A
Drosophila Schneider 2 (S2) Cells	Gibco	Cat#: R69007
Experimental models: Organisms/strains
Mouse: US-SMA model, FVB.Cg-Tg(SMN2)89Ahmb Smn1tm1Msd/J	The Jackson Laboratory	JAX: 005024
Mouse: Taiwanese-SMA mouse model, FVB.Cg-Tg(SMN2)2Hung Smn1tm1Hung/J	The Jackson Laboratory	JAX: 005058
Mouse: wild-type FVB/N	The Jackson Laboratory	JAX: 001800
Oligonucleotides
ASO CTRL (5′-AATCATTTGCTTCATACAGG-3′)	Hua et al., 2008	N/A
ASO1 (MOE) (5′-ATTCACTTTCATAATGCTGG-3′)	Hua et al., 2008	N/A
ASO2 (5′-AAAGTATGTTTCTTCCACAC-3′)	N/A	N/A
ASO1 (OMe) (5′-AUUCACUUUCAUAAUGCUGG-3′)	Hua et al., 2008	N/A
I19 ELP1 (5′-TTAATTAAGTAGAAAACATT-3′)	Sinha et al., 2018	N/A
E20 ELP1 (5′-GCTCGATGATGAACAACTTC-3′)	Sinha et al., 2018	N/A
5′-SS-1 FOXM1	Martin et al., 2021	N/A
Human non targeting siRNA	Dharmacon	Cat#: NC1567415
ON-TARGETplus Human HNRNPA1 siRNA SMARTpool	Dharmacon	Cat#: L-008221-00-0010
ON-TARGETplus Human HNRNPA2 siRNA SMARTpool	Dharmacon	Cat#: L-011690-01-0010
siLuc Sense: GAUUAUGUCCGGUUAUGUAUU siLuc Antisense: UACAUAACCGGACAUAAUCUU	IDT	N/A
Primers for PCR, see Table S1	N/A	N/A
Recombinant DNA
RNAPII (Rpb1) wild-type (WTres; pAT7Rpb1αAmr)	Nguyen et al. 1996	N/A
RNAPII hC4 mutant (pAT7Rpb1αAmrR749H)	Nguyen et al. 1996	N/A
RNAPII α-amanitin-sensitive vector (WTs) GH111110	Nguyen et al. 1996	N/A
pCI-SMN2	Lorson et al., 1999	Addgene cat.# #72287
pcDNA3.1+SMN1mycHIS	Dreyfuss Lab	Addgene, cat.# 71687
pEGFP-RNASEH1	Bubeck et al., 2011	Addgene, cat.# 108699
Software and algorithms
ImageJ	Schneider et al., 2012	https://imagej.nih.gov/ij/
FastQC (v0.11.5)	https://www.bioinformatics.babraham.ac.uk/projects/fastqc/	N/A
BWA mem (v0.7.15-r1140)	https://doi.org/10.1093/bioinformatics/btp324	Li et al, 2009
Samtools (v1.11)	http://samtools.sourceforge.net/	Li et al., 2009
TrimGalore (v0.4.4)	https://www.bioinformatics.babraham.ac.uk/projects/trim_galore/	N/A
Picard MarkDuplicates (v2.24.0)	https://broadinstitute.github.io/picard/	N/A
BedTools2 coverage (v2.29.2)	https://bedtools.readthedocs.io/en/latest/content/installation.html	Quinlan and Hall, 2010
ggplot2 (v3.3.3)	https://ggplot2.tidyverse.org	Wickham et al, 2016
STAR (v2.7.0)	https://github.com/alexdobin/STAR	Dobin et al., 2013
Kallisto (v0.46.0)	https://github.com/pachterlab/kallisto	Bray et al., 2016
ggsashimi (v1.0.0)	https://github.com/guigolab/ggsashimi	Garrido-Martín et al, 2018
pysam (v0.15.4)	https://github.com/pysam-developers/pysam	Li et al,2009
DESeq2 (v1.28.1)	https://bioconductor.org/packages/release/bioc/html/DESeq2.html	Love et al, 2014
EnhancedVolcano (v1.8.0)	https://github.com/kevinblighe/EnhancedVolcano	N/A
rMATS (v4.1.1)	http://rnaseq-mats.sourceforge.net	Shen et al, 2014
Other
High Sensitivity DNA ScreenTape Analysis	Agilent	Cat#: 5067-5584
Concentrator columns	Zymoresearch	Cat#: D5201
33-gauge custom removable needle	Hamilton	Cat#:7803-05
600 Series Microliter Syringes	Hamilton	Cat#:87942

Materials availability

All unique/stable reagents generated in this study are available from the lead contact upon request.

Experimental model and subject details

Animals

All mouse protocols were in accordance with Cold Spring Harbor Laboratory’s Institutional Animal Care and Use Committee guidelines. The severe SMA model (Smn^−/−; SMN2^2TG/0) was generated as previously described (Riessland et al., 2010). Animals were housed in groups and fed standard chow diets. Mice of the same age and of both sexes were used for the studies. Weights were monitored on a daily basis, and animals were sacrificed at day 7 and tissues were harvested.

Genotyping

For each animal, the genotype was verified by PCR reactions using tail-tip DNA. Primer sequences and PCR conditions were previously described (Rigo et al., 2014).

Administration of Oligonucleotides to hSMN2 Transgenic Mice

Oligonucleotide solutions in saline were injected subcutaneously into the upper back at P0 and P1 with a 5-mL syringe and 33-gauge custom removable needle (Hamilton) as described (Hua et al., 2011). All drugs were injected subcutaneously before P3, contralateral to the oligonucleotide-injection site, with TSA (10 mg/kg) or VPA (10 mg/kg) or vehicle. Mouse tissues and organs, including liver, thigh muscles, kidney, and spinal cord, were snap-frozen in liquid N₂ and stored at −70 °C.

Power calculations were done according to standard requirements for animal-protocol approval by CSHL IACUC. Based on anticipated effect sizes, we aimed to have 12–15 pups per group for the survival analysis. Because the crosses to obtain SMA mice yield ~50% of non-SMA heterozygote mice, we needed a total of 120 pups to obtain ~60 SMA mice (15 pups x 4 treatments) for each HDAC inhibitor treatment with its own controls. To obtain this number of mice, a total of 60 breeding cages were used.

The number of 12–15 pups per group was decided following the conditions of a Hua et al., 2011. This number was chosen in order to test the effect of VPA/TSA on ASO1 therapy with inferential statistics (i.e., using p-values) and at the same time to avoid sacrificing an excessive number of mice, according to https://research.usu.edu//irb/wp-content/uploads/sites/12/2015/08/A_Researchers_Guide_to_Power_Analysis_USU.pdf and to the NIH Guide for the Care and Use of Laboratory Animals (2011).

Method details

Antisense oligonucleotide synthesis

ASO1 (5′-ATTCACTTTCATAATGCTGG-3′), a seven-mismatch control (5′-AATCATTTGCTTCATACAGG-3′) and ASO2 (5′-AAAGTATGTTTCTTCCACAC-3′) 2′-O-methoxyethyl-modified oligonucleotides with phosphorothioate backbone and all 5-methyl cytosines were purchased from IDT, and a 2′-OMe-modified phosphorothioate oligonucleotide (5′-AUUCACUUUCAUAAUGCUGG-3′) (Hua et al, 2008) was purchased from TriLink. The oligonucleotides were dissolved in 0.9% w/v saline.

RNAPII expression vectors and alternative splicing reporter minigene

The expression vectors for α-amanitin-resistant variants of the large subunit of human RNAPII (Rpb1) wild-type (WTres; pAT7Rpb1αAmr vector), and the hC4 mutant (pAT7Rpb1αAmrR749H) were previously described⁴. An α-amanitin-sensitive Rpb1 expression vector (WTs) was used as a control. The pCI-SMN2 (Addgene, Plasmid #72287) minigene vector for SMN2 was described previously (Lorson et al., 1999).

Cell culture and treatments

HEK293T, HeLa, and SMA type I homozygous and carrier fibroblasts 3813 and 3814 (Coriell Cell Repositories, Camden, New Jersey, United States) were grown in Dulbecco’s modified Eagle’s medium (DMEM) containing 4.5 g of glucose and 10% fetal bovine serum (Gibco) at 37 °C. Cells were plated at a density of 2×10⁵ cells per well in 12-well plates 24 hr before transfection. siRNA (25 nM), plasmid (500 ng), or ASO (25 nM) transfections were performed 24 hr after cells were plated, using 3 μl of Lipofectamine 2000 (Thermo Fisher Scientific) per well in 12-well plates. 24–48 h later, cells were treated with Trichostatin A (Sigma, T8552), Valproic Acid (Sigma, P4543), 5-Azacytidine (Sigma, A2385) or vehicle for the indicated time, and harvested for downstream procedures.

RNA extraction and RT-PCR

Twenty milligrams of mouse tissue was pulverized in liquid N₂ with mortar and pestle, and homogenized with 1 mL of Trizol (Invitrogen). Total RNA was isolated according to the manufacturer’s directions. Two methods were used for RT-PCR reactions in different experiments: 1. One microgram of RNA was reverse-transcribed with M-MLV reverse transcriptase (Invitrogen) and oligo-dT primer, and the cDNA was amplified. Amplification and analysis of SMN2 transcripts was performed as described (Lorson et al., 1999). 2. One microgram of total RNA was reversed transcribed using Superscript III (Thermo Fisher Scientific) reverse transcriptase and random primers. After PCR amplification using Gotaq (Promega) and SMN2 primers surrounding E7, products were either loaded in a 6% acrylamide (37.5:1) TBE gel and stained with Ethidium Bromide for visualization or analysed by Tapestation 4150 using high sensitivity DNA D1000 screentapes (Agilent) for quantification. Primer sequences are listed in the Supplementary Table 1. Data were analyzed using Excel and plotted in GraphPad Prism. A minimum of three biological replicates were always analyzed, each in technical triplicate. Samples form each RT-PCR experiment were run together on the same gel. Non-consecutive lanes were denoted by a vertical white line of separation.

Western blot

Mouse organs:

Twenty milligrams of tissue was pulverized in liquid N₂ and homogenized in 0.4 mL (liver, kidney, muscle, spinal cord, brain, and heart) of 1× protein sample buffer containing 2% (w/v) SDS, 10% (v/v) glycerol, 50 mM Tris-HCl (pH 6.8), and 0.1 M DTT. Protein samples were separated by 12% SDS-PAGE and electroblotted onto nitrocellulose membranes. The blots were probed with mAb anti-hSMN (BD Biosciences, 610646), or pAb anti-β-tubulin (Sigma), followed by secondary IRDye 800CW-conjugated goat anti-mouse or anti-rabbit antibody. Protein signals were detected with an Odyssey instrument (LI-COR Biosciences).

Cells in culture:

Cells were lysed in 1X protein sample buffer (50mM Tris-HCl pH6.8, 2% SDS, 10% glycerol, 5% β-mercaptoethanol, 0.0025% Bromophenol blue). Protein samples were separated by 4–12% SDS-PAGE (NuPage, Life Technologies) and electroblotted onto PVDF membranes. The blots were probed with anti-H3K9me2 (Abcam, ab1220), anti-H3K9ac (Abcam, ab4441), anti-H3 (Abcam, ab12079), anti-Myc (Millipore, MABE282), anti-tubulin (Sigma-Aldrich, T5168), anti-RNase H1 (Proteintech, 15606-1-AP), anti-actin (Sigma-Aldrich, A2066), anti-G9A (CST, 3306S) or anti-Ago1 (CST, 5053S) antibodies.

RNAi knockdown

Downregulation of hnRNP A1 and A2 was performed using ON-TARGET plus SMARTpool siRNA oligonucleotides (Dharmacon). siRNA oligos were delivered to cells following the manufacturer’s instructions, and allowed to act for 72 hr. Accell siRNA anti-human non targeting siRNA (Dharmacon, NC1567415) was used as a control.

siRNA oligos were delivered to cells following the manufacturer’s instructions, and allowed to act for 72 hr. siLuc (Sense: GAUUAUGUCCGGUUAUGUAUU Antisense: UACAUAACCGGACAUAAUCUU) was used as a control.

Chromatin immunoprecipitation (xChIP) followed by q-PCR

Approximately 2 × 10⁶ HEK293T cells per sample were treated for 10 min in 1% (v/v) formaldehyde at room temperature to crosslink protein-DNA complexes. Crosslinking was stopped with glycine at a final concentration of 125 mM. Cells were washed twice with cold PBS and swelled on ice for 10 min in 25 mM HEPES pH 8, 1.5 mM MgCl₂, 10 mM KCl, 0.1% NP-40, 1 mM DTT and 1× protease inhibitor cocktail set III (Calbiochem). Following Dounce homogenization, the nuclei were collected and resuspended in 1 ml sonication buffer (50 mM HEPES pH 8, 140 mM NaCl, 1 mM EDTA, 0.1% sodium deoxycholate, 0.1% SDS and 1× protease inhibitor cocktail). DNA was sonicated in an ultrasonic bath (Bioruptor Diagenode) to an average length of 200–500 bp. After addition of 1% (v/v) Triton X-100, samples were centrifuged at 15,000 ×g. Supernatants were immunoprecipitated O/N with 40 µl of pre-coated anti-IgG magnetic beads (Dynabeads Protein G, Invitrogen) previously incubated with the antibody of interest for 6 hr at 4 °C. The antibodies used were: rabbit anti-H3 (2 μg, Abcam ab1791), mouse anti-H3K9me2 (4 μg, Abcam ab1220), rabbit anti H3K9me3 (4 μg, Abcam ab8898), rabbit anti H3K9Ac (2 μg, Abcam ab4441), rabbit anti Rpb1 NTD (2 μg, Cell Signaling, D8L4Y), rabbit Phospho-Rpb1 CTD Ser2 (2 μg, Cell Signaling, E1Z3G), rabbit Phospho-Rpb1 CTD Ser5 (2 μg, Cell Signaling, D9N5I). Control immunoprecipitations were performed with rabbit IgG (1 μg, Abcam ab171870). Beads were washed sequentially for 5 min each in Low-salt (20 mM Tris-HCl pH 8, 150 mM NaCl, 2 mM EDTA, 1% Triton X-100, 0.1% SDS), High-Salt (20 mM Tris-HCl pH 8, 500 mM NaCl, 2 mM EDTA, 1% Triton X-100, 0.1% SDS) and LiCl buffer (10 mM Tris pH 8.0, 1 mM EDTA, 250 mM LiCl, 1% NP-40, 1% Na-deoxycholate) for 5 min at 4 °C and then twice in TE 1× for 2 min at room temperature. Beads were eluted in 1% SDS and 100 mM NaHCO₃ buffer for 15 min at 65 °C and crosslinking was reversed for 6 hr after addition of NaCl to a final concentration of 200 mM. Chromatin was precipitated with ethanol overnight, treated with 20 μg proteinase K, and purified by phenol-chloroform extraction. Immunoprecipitated DNA (1.5 μl) and serial dilutions of the 10% input DNA (1:4, 1:20, 1:100, and 1:500) were analyzed by SYBR-Green real-time qPCR. The oligonucleotide sequences used are listed in Supplementary Table 1.

Chromatin immunoprecipitation followed by deep sequencing (ChIP-seq)

HEK293T or S2 cells were treated for 10 min in 1% (v/v) formaldehyde at room temperature to crosslink protein-DNA complexes, and the crosslinking was then quenched with 125 mM glycine for 5 min. Cells were washed twice with cold PBS and resuspended in 600 µL of Cell Lysis Buffer (5 mM PIPES pH 8.0, 85 mM KCl, 0.5% NP-40 and 1x complete) and incubated on ice for 10 min. They were then centrifuged at 2,400 rpm for 5 min and nuclear pellets were resuspended in 400 μL of Nuclear Lysis buffer (0.5% SDS, 5 mM EDTA, 25 mM Tris-HCl pH 8.0 and 1x complete) and incubated on ice for 10 min. Chromatin were sonicated for 10 min (high power, 30 s on-off repeats) to obtain an average size of 200–500 bp (Bioruptor, Diagenode). Soluble chromatin was obtained after sonicated nuclei were centrifuged at 13,000 rpm for 10 min. Thirty micrograms of HEK293T cells chromatin were mixed with 2 µg of S2 cells chromatin and diluted 10 times with IP dilution buffer (10 mM Tris-HCl pH8.0, 5 mM EDTA, 0.5% Triton X-100 and 0.15 M NaCl). Nucleosomes were isolated from supernatant by IP with 3 μg of H3K9me2 antibody overnight at 4C. Thirty microliter of washed Dynabeads™ M-280 Sheep Anti-Mouse IgG (11202D, Thermo Fisher) was added and samples were rotated for 2–3 hrs at 4C. IPed DNA/beads were washed with 1 mL of buffer A (20 mM Tris-HCl pH 8.0, 2 mM EDTA, 0.05% SDS, 1% Triton X-100 and 0.165 M NaCl) once, 1 mL of buffer B (20 mM Tris-HCl pH8.0, 2 mM EDTA, 0.05% SDS, 1% Triton X-100 and 0.5 M NaCl) once, 1 mL of buffer C (10 mM Tris-HCl pH8.0, 1 mM EDTA, 1% NP-40, 1% Sodium Deoxycholate and 0.25 M LiCl) and then 1 mL of buffer D (10 mM Tris-HCl pH8.0 and 1 mM EDTA) twice. Next IPed beads were incubated with 0.01 mg/mL RNase A (Ambion) in 300 μL of buffer E (1% SDS, 0.1 M NaHCO3 and 0.5 M NaCl) at 65°C for at least 4 hr. After RNase treatment, 30 μL of 10x Proteinase K mixture (200 mM Tris-HCl pH 6.5, 150 mM EDTA and Proteinase K 0.3 mg/mL) were added and then incubated at 45°C for 2 hrs. DNA fragments were isolated using ChIP DNA clean and concentrator columns (D5201, Zymoresearch) and quantified using Qubit Fluorometric Quantification. DNA libraries were made according to NEBNext Ultra II DNA Library Prep Kit for Illumina (NEB) manual, using 5 ng of DNA, and 10∼12 cycles of PCR were used to amplify the libraries. Deep sequencing was conducted on a NovaSeq 6000 platform (Illumina) by Novogene UK.

Chip-seq analysis

Raw data quality were checked with fastQC (v0.11.5) and preprocessed with TrimGalore (v0.4.4) (Krueger et al., 2015) to remove possible sequencing adapters and to filter sequences by Phred quality scores (“-q 20 --length 10 --stringency 1). Reads were further aligned with bwa mem (v0.7.15-r1140) (Li et al., 2009) against the ENSEMBLE GRCh38 and FlyBase dm6 genome assemblies, which correspond to Homo sapiens and Drosophila melanogaster, respectively. Reads fully aligning in multiple regions were randomly assigned to one of those top score matches, using the aligner default parameters. Also, bitwise flags read paired (0×1) and read mapped in proper pair (0×2) were required using samtools (v1.11) (Heng et al., 2009). Duplicated reads were identified and removed with Picard MarkDuplicates (v2.24.0) (Broad Institute, 2019). BedTools2 coverage (v2.29.2) (Heng et al., 2010) was employed to reveal the number of reads overlapping each analysed feature. The total number of mapped spike-in reads (SpkIn) per sample were counted and used as a normaliser as follows:

$$\begin{array}{c}Sf=\frac{Hs}{SpkIn}\end{array}*0.01$$

For the downstream analyses, all raw read counts were multiplied by the scaling factor (Sf), giving the normalised counts.

Illumina data pre-processing

Quality control for raw Illumina short-reads was performed using the FastQC tool (https://www.bioinformatics.babraham.ac.uk/projects/fastqc/). Read adaptors were trimmed using TrimGalore (https://www.bioinformatics.babraham.ac.uk/projects/trim_galore/) in paired-end mode, removing reads with less than 10 nucleotides (nt) and/or low-quality ends (20 Phred score cut-off). The resultant reads were aligned against the reference human genome (GRCh38) using STAR software (Dobin et al., 2013), requiring a minimum alignment score (–outFilterScoreMin) of 10.

Identification of expressed genes

Generated strand-specific pA+ RNA-seq data was used to identify expressed genes in HEK116 cell-line. Thus, Kallisto (Bray et. al 2016 – Nature biotechnology) was used to map the reads against the human transcriptome (Ensembl v90), and TPM measurement for each transcription unit (TU) from the output was acquired. The transcript with highest TPM was selected per gene. Genes having no transcript with TPM higher than 4 were discarded. Moreover, filtered TUs must have protein-coding tag as a biotype, which was extracted from Ensembl GTF file version 90. To better detect signal levels from POINT technology, overlapping TUs were excluded. For that, an extra window of 10 Kb upstream and downstream of each TU was added. A final number of 2741 non-overlapping expressed genes was identified.

Sashimi plots

For sashimi plots generation, ggsashimi v1.0.0 (https://doi.org/10.1371/journal.pcbi.1006360) was used on RNA-seq samples, with the following parameters: -M 100 -s MATE2_SENSE -S plus. The 3 replicates per condition were given as input. The number of read counts supporting each event was internally aggregated using the replicates mean. The following table shows the number of read counts per replicate, separated by a semicolon:

	Control	ASO1	VPA	ASO1+VPA
E6-E7	384; 930; 720	1002; 977; 994	1058; 1851; 1928	854; 2030; 1570
E7-E8	452; 1018; 854	1327; 1053; 1040	1216; 1921; 2075	1008; 2313; 1818
E6-E8	338; 1196; 636	370; 617; 361	1187; 970; 1172	339; 604; 441

Meta-profiles

Previous identified expressed genes (N=2741 genes) were divided into 100 bins. For each bin, the overlapping number of fragment counts was measured with pysam (Li et al., 2009), which was then divided by the number of million reads mapped against the human genome, obtaining the Fragment counts per Million Mapped Reads.

Importantly, the 1% top and bottom extreme values per bin were remove, and the mean was measured excluding these values.

POINT-seq windows ratios

Promoter-proximal region was defined as a window from TSS to 1500bp downstream. From this point to TES it was considered gene body window. The gene set employed was the same as used in the meta-profiles (N=2741 genes). The number of fragment counts per window was obtained using pysam. Then, fragment counts were divided by window length, to accommodate for differences in gene size. Thus, the ratio between promoter-proximal and gene body regions was obtained by the division between the number of fragments in these respective windows, normalised to the window’s size. For the cases where no reads were found in the gene body, a value of zero was attributed. For the histogram, when there were not reads in the promoter-proximal region a value of zero was obtained by adding 1 unit to the ratio and by measuring the log2.

The ratio between the upstream and downstream regions of ASO1 target was obtained as described for promoter-proximal / gene body regions. Here, the upstream window considered was chr5:70951601-70951993 and the downstream chr5:70951993-70952250.

Identification of differentially expressed genes

Kallisto mapped RNA-seq reads against the human transcriptome to produce estimated gene expression values, which were then gathered in a non-normalized count matrix.

Using it as input, significant differentially expressed genes were detected with the DESeq2 package (10.1186/s13059-014-0550-8). A cut-off of 1×10⁻⁵ for p-value and 1 for the absolute value of log2FoldChange was applied over DESeq2’s own two-sided statistical test results. Volcano plots were generated with the EnhancedVolcano v1.8.0 package (10.18129/B9.bioc.EnhancedVolcano).

Differential Alternative spliced events

Genome-wide splicing variation induced by VPA treatment were investigated with rMATS v4.1.1 (doi: 10.1073/pnas.1419161111). Ensembl human v90 was given as annotations. The following paraments were also given to rMATS: -t paired --readLength 150 --libType fr-firststrand. The output was then stringently filtered by maser package (https://doi.org/doi:10.18129/B9.bioc.maser), thus an event was considered significant when the average number of reads per condition supporting the predominant class (inclusion or exclusion) was > 20, dPSI > 0.3 and FDR < 0.01. Scatter plot was obtained with the dotplot function of the maser package and represents the average PSI among the 3 replicates, per condition.

Quantification and statistical analysis

GraphPad Prism 9 was used for statistical analysis. One-way or two-way ANOVA tests were followed by Bonferroni’s corrected multiple comparisons between pairs of conditions, or TTEST comparisons were used with just two pair of conditions. Unless otherwise indicated in the figure legends, we analyzed three biological replicates for each data point in all graphs, and the level of significance was as follows: p < 0.05.

Supplementary Material

Sup 2. Supplemental Figure 2. Antisense oligonucleotides and their effects on chromatin (related to Fig. 3).

(A) Exert of the SMN2 gene sequence comprising the E7 alternative exon. Blue uppercase, exon 7 nucleotide sequence. Lowercase, intronic sequences. Red uppercase, exon 8 nucleotide sequence. Green highlight, sequence corresponding to the binding site for ASO1 and hnRNPA1/A2 i the encoded pre-mRNA. Red highlight, idem for the binding site for ASO2. (B-D) Levels of H3K9me2 deposition along (B) Small EDRK-rich factor 1A (SERF1A, upstream of SMN2), (C) Myoblast Determination Protein (MYOD), and (D) Glyceraldehyde 3-phosphate dehydrogenase (GAPDH), assessed by ChIP-qPCR in HEK293T cells. Data represented as mean S.D. (n = 4, *p < 0.05, two-tailed Student’s t test).

Sup 1. Supplemental Figure 1. Histone acetylation potentiates the ASO effect on E7 inclusion (related to Fig. 2).

Combined effects on endogenous SMN2 E7 alternative splicing in transfected (A) HEK293T cells treated with 25 nM ASO1 and increasing doses of TSA for 24 hr., (B) HeLa cells treated with 10 nM ASO1 and 3 μM TSA for 24 hr., and (C) SMA patient fibroblasts (3813) treated with 10 nM ASO1 and 3 μM TSA for 24 hr. Bars display means SD of percentage of the radioactivity in the FL band over the sum of radioactivity in the FL and DE7 bands of at least three independent transfection experiments. Control experiments for the opposing roles of ASO1 on chromatin and splicing (related to Fig. 2). (D) Overexpression of the SMN protein does not alter the ASO1 effect on E7 splicing. Top: control Western blot for efficient expression of the c-Myc-SMN fusion protein encoded by plasmid pcDNA3.1+SMN1mycHIS (Addgene, cat.# 71687) using anti-tubulin (tub.) as loading control. Bottom: SMN2 E7 RT-PCR of cells co-transfected with ASO1 and plasmid. (E) Transfection of HEK293T cells with ASO1 does not increase H3K27 trimethylation or H3K9 acetylation over the SMN2 gene. Distribution of these histone marks was assessed by ChIP-qPCR, with amplicons mapping near the numbered exons (E) and introns (i). Three independent immunoprecipitation replicates were conducted per experiment. Data are represented as mean S.D. (n = 5; in all cases p was much higher than 0.05, two-tailed Student’s t test). (F) Overexpression of the RNase H enzyme does not affect the ASO1 effect on E7 splicing. Top: control Western blot for efficient expression of the GFP-RNase H fusion protein encoded by plasmid pEGFP RNASEH1 (Addgene, cat.# 108699), using anti-actin (act.) as loading control. Bottom: SMN2 E7 RT-PCR of cells co-transfected with ASO1 and the RNase H plasmid. (G) siRNA-mediated AGO1 knockdown does not alter the ASO1 effect on E7 splicing. Top: control Western blot for efficient AGO1 knockdown using anti-tubulin (tub.) as loading control. Bottom: SMN2 E7 RT-PCR of cells co-transfected with ASO1 and the AGO1 siRNA. RT-PCR conditions for panels a, c and d were as in main Figure 1A.

Sup 4. Supplemental Figure 4. H3K9me ChIP-seq analysis (related to Fig. 6).

(A) Top, genome browser views of H3K9me2/input ChIP-seq signal for the merged SMN1/2 genes of HEK293T cells transfected with control or ASO1 and treated or not with VPA. Bottom, quantification of the H3K9me2 signal in regions A (blue, bulk of the gene) and B (orange, ASO1 target region). (B) Zoom-in of region B and part of region A. (C) Genome browser views of H3K9me2/input ChIP-seq signal for the B-actin gene of HEK293T cells transfected with control or ASO1 and treated or not with VPA.

Sup 5. Supplemental Figure 5 (related to Fig. 7). ASO1 and TSA combined treatment in mice.

Kaplan-Meier survival plot (A) and growth curves (B) of SMA mice after subcutaneous administration at P0 and P1 of 16.8 μg ASO1 (n=20) or vehicle (n=12), one subcutaneous dose of 10 μg per g of body weight TSA (n=15) at P2, or both treatments together (n=24). ASO1-treated heterozygotes (n=18) served as controls. Statistical significance was analyzed by two-way repeated measures ANOVA. P < 0.05 was considered statistically significant; data are represented as mean + SD. (C-H) Western-blot analysis of P7 tissues from liver (C), kidney (D), muscle (E), spinal cord (F), brain (G) and heart (H) from SMA mice, using anti-hSMN (BD Biosciences, upper panel) and anti-β-tubulin (Sigma, lower panel) antibodies.

Sup 6. Supplemental Figure 6. Phenotypes of mice treated with subotimal doses of ASO1 alone or in combination with TSA (related to Fig. 7).

(A) Comparison of body sizes at P11 of mice heterozygous (Het) and homozigous (SMA) for the mutated mouse SMN gene inhected with ASO1 (left) or ASO1 + TSA (right). (B-D) Phenotypes of SMA mice at P14 (B), P17 (C) and P42 (D) treated with ASO1 or ASO1 + TSA.

Sup 7. Supplemental Figure 7. SMA mice injected with ASO1 show increased H3K9 dimethylation and RNAPII roadblocks along the SMN2 gene in brain and liver (related to Fig. 7).

H3K9me2 (left) and total RNAPII (right) distribution along the human SMN2 transgene, assessed by ChIP-qPCR, in brain (A) and liver (B) of P7 SMA mice injected with ASO1 as indicated in Methods. Red vertical arrows indicate the target site for ASO1 on the pre-mRNA. Three independent immunoprecipitation replicates were conducted per experiment. Data are represented as mean S.D. (n = 3, *p < 0.05, two-tailed Student’s t test).

Sup 3. Supplemental Figure 3 (related to Fig. 6). Global effects of VPA on alternative splicing revealed by RNA-seq.

(A) Scatter-plot showing PSI (Percentage Spliced-In) of exon skipping events detected by rMATS, for RNA-seq samples. Red dots are differentially spliced events for which the inclusion is higher in the control cells, and blue dots represent events for which the inclusion is higher in VPA-treated cells. Grey dots represent not significant cases. Cases were only deemed significant if the number of supportive reads > 20, dPSI > 0.3 and FDR < 0.01. (B) Table showing the number of alternative splicing events (ASEs) of different modes, whose patterns are significantly affected by VPA. Total ASE are all alternative splicing events with more than 20 reads. Differential ASE are the ones for which supportive reads were > 20, dPSI > 0.3 and FDR < 0.01. ss: splice site. (C) Sashimi plot showing the average number of reads supporting each splice junction of the merged SMN1/2 genes, based on the aligned RNA-seq data, upon treatments of HEK293T cells with ASO1, VPA or both together. (D) Raw and corrected quantification of the levels of E7 inclusion as assessed by analysis of the E6-E7 junction reads. Inclusion levels are expressed as E7 PSI considering the E6-E7 reads for inclusion. POINT-seq analysis on the merged SMN1/2 genes (related to Fig. 6). (E) POINT-seq signal normalized to the library size on the merged SMN1/2 genes of HEK293T cells treated with VPA. Ratios between the accumulated reads at the promoter region (P) and gene body (GB) are shown on the right. (F) POINT-seq signal at a zoom-in of the target region of ASO1, displaying the impact of ASO1-binding on Pol II progression. The top track is the control sample, and the bottom corresponds to ASO1-treated cells. Read density ratios of the upstream (U) over the downstream (D) regions are shown on the right side of the corresponding track. The control profile is superimposed as a dotted line on the ASO1 profile.