Therapeutic manipulation of IKBKAP mis-splicing with a small molecule to cure familial dysautonomia

Approximately half of genetic disease-associated mutations cause aberrant splicing. However, a widely applicable therapeutic strategy to splicing diseases is yet to be developed. Here, we analyze the mechanism whereby IKBKAP-familial dysautonomia (FD) exon 20 inclusion is specifically promoted by a small molecule splice modulator, RECTAS, even though IKBKAP-FD exon 20 has a suboptimal 5′ splice site due to the IVS20 + 6 T > C mutation. Knockdown experiments reveal that exon 20 inclusion is suppressed in the absence of serine/arginine-rich splicing factor 6 (SRSF6) binding to an intronic splicing enhancer in intron 20. We show that RECTAS directly interacts with CDC-like kinases (CLKs) and enhances SRSF6 phosphorylation. Consistently, exon 20 splicing is bidirectionally manipulated by targeting cellular CLK activity with RECTAS versus CLK inhibitors. The therapeutic potential of RECTAS is validated in multiple FD disease models. Our study indicates that small synthetic molecules affecting phosphorylation state of SRSFs is available as a new therapeutic modality for mechanism-oriented precision medicine of splicing diseases.

S plicing mutations through the creation or disruption of splicing cis-elements induce constitutive gain or loss or the alternative splicing of target exons, which in turn prevent functional protein production by generating stop codons or frameshifts. Pre-mRNA splicing is catalyzed by five essential small nuclear RNAs (U1, U2, U4, U5, and U6 snRNA) and more than 100 proteins. In addition to these essential factors (e.g., U1 small nuclear ribonucleoprotein (snRNP) to recognize the 5′-splice site and U2 snRNP to recognize the branchpoint sequence), various trans-acting splicing factors are recruited to cis-regulatory elements to orchestrate alternative splicing. Several approaches have been taken to amend mutation-derived splicing defects by targeting either cis-regulatory elements or trans-acting splice factors. The former approach includes the antisense-oligonucleotide drug nusinersen, which rescues deficient SMN2 exon 7 inclusion for spinal muscular atrophy therapy [1][2][3] , and small-molecule compounds, such as SMN-C3 and RG7916, that also promote SMN2 exon 7 inclusion, in part by directly binding to an imperfect RNA helix formed by the exon 7 5′-splice site and U1 snRNA acting for bulge repair [4][5][6][7][8] . Examples of the latter approach include smallmolecule inhibitors for CDC-like kinase (CLK) or serine/arginine-rich protein kinase (SRPK) 9 , which then inactivate serine/ arginine-rich splicing factors (SRSFs) through serine phosphorylation in their carboxy-terminal RS domain. Previously, we synthesized CLK inhibitors 10 and demonstrated their therapeutic potential in disease models of cystic fibrosis (c.3849 + 10 kb C > T mutation of cystic fibrosis transmembrane conductance regulator (CFTR)) 11 , Duchenne muscular dystrophy (c.4303 G > T mutation of dystrophin) 12 , and anhidrotic ectodermal dysplasia with immunodeficiency (EDA-ID; IVS6 + 866 C > T mutation of NF-kappa-B essential modulator (NEMO)) 13 .
In parallel, we developed a dual-color splicing reporter system combining two different fluorescent proteins to identify both ciselements and trans-acting factors involved in targeting alternative splicing events in worms, mice, and cultured cells [14][15][16] . Recently, we applied this system to screen a chemical library and found a small-molecule RECTAS, which corrects abnormal splicing caused by the IVS20 + 6 T > C splicing mutation of the (inhibitor of κ light polypeptide gene enhancer in B-cells, kinase complexassociated protein (IKBKAP) gene 17 . The homozygous mutation is mainly found in the Ashkenazi Jewish population and is responsible for >99.5% of cases of familial dysautonomia (FD) (also known as Riley-Day syndrome or hereditary sensory/autonomic neuropathy type III) [18][19][20] . FD is an autosomal recessive genetic disease, and autonomic neurons and peripheral sensory neurons derived from neural crest cells are impaired in development, function, and survival [18][19][20] . Patients with FD exhibit various symptoms including cardiovascular instability, recurrent pneumonia, vomiting/dysautonomic crisis, gastrointestinal dysfunction, decreased sensitivity to pain and temperature, and defective lacrimation. The IVS20 + 6 T > C mutation causes abnormal IKBKAP exon 20 skipping, due to impaired recognition of the mutant 5′ splice site by U1 snRNP 21,22 , leading to a frameshift and the generation of a premature termination codon in exon 21 of IKBKAP mRNA 20 . This frameshift in turn results in reduced expression of IKAP (or ELP1), compromising tRNA modification and neuronal cell survival 17,23,24 . RECTAS directly affects mRNA splicing, and our transcriptome analyses revealed that RECTAS affects the splicing of only a limited set of genes, suggesting that it has high specificity and a direct role in correcting IKBKAP aberrant splicing 17 . Sequence analysis of the responsive exons suggests that RECTAS promotes exon recognition by either inhibiting splicing repressors, such as heterogeneous nuclear ribonucleoproteins (hnRNPs), or stimulating splicing activators. RBM24 functions as a tissue-dependent splice activator by binding to a cryptic intronic splicing enhancer element (IVS20 + 13 to +29) downstream of the intronic 5′ splice-site mutation of IKBKAP-FD and promotes U1 snRNP recognition of the mutant 5′ splice site 25 . Thus, undetectable RBM24 expression in neuronal tissues explains the greater extent of aberrant splicing of IKBKAP-FD exon 20 in neurons 25 . In this study, we revealed that SRSF6 is also involved in exon 20 recognition and that RECTAS specifically rectifies aberrant splicing by promoting the SRSF6 function.

Results
Binding of SRSF6 to an intronic splice enhancer (ISE) is required for IKBKAP-FD exon 20 inclusion. IKBKAP-FD exon 20 has a suboptimal 5′ splice site, due to the IVS20 + 6 T > C mutation. It lacks complementarity to U1 snRNA at the −1 and +6 positions (Fig. 1a), and such suboptimal splice sites are more prone to be regulated by alternative splicing factors [11][12][13]26,27 , as we reported for promotion of spliceosomal recognition of this 5′ splice site by RBM24 in skeletal muscle tissue 25 . The RNA sequencing (RNA-Seq) evaluation of the transcriptome of FD patient fibroblasts homozygous for the IKBKAP IVS20 + 6 T > C mutation following RECTAS treatment revealed a highly selective action of RECTAS on IKBKAP-FD exon 20, with the secondhighest ΔPSI (percent spliced-in) of 618 altered splicing events with ≥0.1 of | ΔPSI | (Fig. 1b, c and Supplementary Data 1). As we supposed that RECTAS rescued exon 20 skipping as part of the overall context of splicing regulation, we checked whether RBM24 function is involved in RECTAS-induced exon 20 inclusion. Western blot for human tissues as well as HeLa and FD patient fibroblasts, responder cells to RECTAS 17 , indicates RBM24 expression is exclusive in heart and skeletal muscle but absent in RECTAS responder cells, irrespective of RECTAS treatment (Fig. 1d). These observations ruled out the involvement of RBM24 in RECTAS-induced splice modification. Therefore, we hypothesized that RECTAS interacts with an unidentified splicing regulator(s) that promotes the restoration of normal exon 20 inclusion. To test this hypothesis, we examined loss-of-function study by RNAi knockdown for individual SRSFs, a major family of alternative splicing activators, and compared the extent of RECTAS-induced promotion of IKBKAP-FD exon 20 inclusion, using a dual fluorescent splicing reporter system we previously developed ( Fig. 1e) 11,13,17 . We observed a marked reduction (~60%) in exon 20 inclusion when SRSF6 was knocked down (Fig. 1f, g), whereas depletion of nine other SRSFs yielded no or only a modest effect on this splicing event (Fig. 1f, g).
These observations indicate that SRSF6 is a crucial regulator of IKBKAP-FD exon 20 inclusion in the therapeutic splicing environment provided by RECTAS treatment. We further asked whether SRSF6 directly controls IKBKAP-FD exon 20 inclusion. For this purpose, we searched for potential SRSF6-recognition sites based on the systematic evolution of ligands by exponential enrichment (SELEX) motifs for SRSF6 with ESE finder 3.0 28,29 for a region downstream or upstream of the exon 20 5′-splice site (−100 to +100 nucleotides (nt)) of IKBKAP. We identified three candidate motifs 21, 45, and 58 nt downstream of the exon 20 5′ splice site (site-a, site-b, and site-c, respectively, in Fig. 1h). Subsequent validation of SRSF6 binding by RNA pull-down assay with biotin-conjugated RNAs revealed that only site-b has a high affinity to a cellular protein inferred to be SRSF6 (Fig. 1i). The band of expected SRSF6 disappeared in the pull-down products from SRSF6-knockdown cells, confirming its identity as SRSF6 (Fig. 1j). In addition, SRSF6 binding was lost when point mutations (site-b(mt) in Fig. 1h) were introduced in the SRSF6binding motif (Fig. 1k). The functional significance of SRSF6 binding for IKBKAP-FD exon 20 recognition was subsequently examined by introducing a site-b(mt) mutation into the IKBKAP- The IKBKAP exon 20 5′ splice site has nearly optimal pairing to U1 snRNA, whereas the IVS20 + 6 T > C creates an additional mismatch at +6 position, resulting in suboptimal recognition by U1 snRNP. Blue dots indicate base pairing. b Top 100 differential splice events by | ΔPSI (PSI RECTAS -PSI DMSO ) | following RECTAS treatment (2 µM, 8 h) in FD patient fibroblasts. c Sashimi plot with junction read number of IKBKAP-FD exon 19-22 for DMSO-or RECTAS-treated FD patient fibroblasts. The displayed read count range was set to 0-50. d RT-PCR for RBM24 expression in human tissues, and cell lines treated with 10 µM RECTAS or 0.1% DMSO for 24 h. RBM24 and ACTB were detected with primers oAM669 + oAM671 and oAM13 + oAM14, respectively. e Diagram of reporter system for IKBKAP-FD exon 20 splicing 13,17 . GFP is translated when exon 20 is included (E20(+)), whereas RFP is translated when it is skipped (E20(−)). f, g Microscopic images (f) and quantification of IKBKAP-FD exon 20 inclusion (g), following treatment with RECTAS (10 µM) or 0.1% DMSO for 24 h in HeLa cells transfected with the indicated siRNAs for 72 h. h SRSF6binding motifs starting at +21 nt (site-a), +45 nt (site-b), and +58 nt (site-c), predicted by ESE finder 3.0 28,29 are shown. RNA sequences for pull-down assays are indicated. Asterisks in site-b(mt) indicate mutated ribonucleotides. ISE-20 from Sinha et al. 30 is also indicated. i-k Western blot of SRSF6 by 1H4 antibody for RNA pull-down products or no bait RNA control for HeLa cells (i), site-b pull-down products for HeLa cells treated with siRNA for 72 h (j), and site-b and site-b(mt) RNA pull-down products for HeLa cells (k). l Quantification of IKBKAP exon 20 inclusion rate, following RECTAS treatment for 24 h in HeLa cells with IKBKAP-FD or site-b(mt)-mutant IKBKAP (FD site-b(mt)) reporter. Data from four replicates are shown in (f), (g), and (l), and two replicates are shown in (i), (j), and (k). Representative data of two experiments are shown in (d). Columns, mean; bars, SE; n.s., P ≥ 0.05; **P < 0.01; ***P < 0.001 for unpaired two-tailed t test in (g) (vs. the RECTAS 10 µM of si-NS: si-SRSF1, P = 2.3 × 10 −3 ; si-SRSF2, P = 1.7 × 10 −4 ; si-SRSF4, P = 0.94; si-SRSF6, 1.5 × 10 −7 ; si-SRSF7, P = 1.9 × 10 −4 ; si-SRSF8, P = 1.0 × 10 −4 ; si-SRSF9, 3.0 × 10 −3 ; si-SRSF10, P = 0.28; si-SRSF11, P = 6.1 × 10 −4 ; si-SRSF12, P = 1. FD exon 20 splicing reporter, and a reduction in RECTAS-led exon 20 inclusion was observed, comparable to that seen in SRSF6 knockdown cells (~60%) (Fig. 1l). These combined observations indicate that site-b is an ISE, and whose recognition by SRSF6 is crucial to provide a splicing context that restores normal splicing of IKBKAP-FD exon 20 in the presence of RECTAS. We note that the identified SRSF6-dependent ISE overlaps with a 20-nt region with ISE activity (dubbed ISE-20) revealed by our previous antisense-oligonucleotide walk 30 (Fig. 1h). In addition, the SRSF6-binding site (+45 to +50) somewhat close to that of RBM24 (+13 to +29) 25 imply RBM24 binding may interfere with SRSF6 binding, and may impair the effect of RECTAS in tissues with RBM24 expression.
Reciprocal control of IKBKAP-FD exon 20 splicing by REC-TAS and CDC-like kinase (CLK) inhibitors suggests that CLKs are pivotal regulators. As SRSF6 is functionally activated by serine phosphorylation in its carboxyl-terminal RS domain, we next asked if kinases upstream of SRSF6 promote IKBKAP-FD exon 20 inclusion. For this purpose, we examined the effects of the following synthetic inhibitors of potential upstream SRSF kinases on IKBKAP-FD exon 20 recognition with our splicing reporter: TG003 for CLKs; 10,12,13 INDY, CaNDY, and ALGER-NON for dual-specificity tyrosine phosphorylation-regulated kinases (DYRKs) [31][32][33] and CLKs 11 , and SRPIN340 for SRSF protein kinases (SRPKs) 34,35 . We used RECTAS and the plant cytokinin kinetin 17,36,37 as controls with opposite effect to that expected for inhibitors of SRSF6 kinases. Treatment of cells with TG003, INDY, CaNDY, and ALGERNON at 2 and 10 µM resulted in the dose-dependent promotion of IKBKAP-FD exon 20 skipping (Fig. 2a, b), in contrast to the increased inclusion of exon 20 by RECTAS treatment (Fig. 2a, b), whereas the exoninclusion effect of kinetin was considerably weaker, consistent with our prior observation 17 (Fig. 2a, b). SRPIN340 had no effect on IKBKAP-FD exon 20 splicing (Fig. 2a, b). Some inhibitors of CLKs and DYRKs show cross-inhibition due to structural similarity at the ATP-binding pocket between these two kinase families 11,25,[31][32][33]38,39 , and depletion of CLK or DYRK isoforms by RNAi was conducted to distinguish their effects. As a result, we found that the depletion of CLK isoforms inhibited IKBKAP-FD exon 20 recognition similarly to SRSF6-depleted cells (Fig. 2c), whereas depletion of DYRK isoforms had no effect (Fig. 2c). Moreover, overexpression of CLK1 mimicked the effect of RECTAS on the inclusion of IKBKAP-FD exon 20 ( Fig. 2d), whereas the kinase-inactive K191M mutant of CLK1 had no effect (Fig. 2d). Dependency of RECTAS activity on CLK activities was further assessed by evaluating RECTAS response in cells under CLK inhibition. We recently reported CaNDY as a potent pan-CLK inhibitor 11 , and found that there is no exon 20 inclusionpromoting activity by RECTAS when CLKs were functionally inactivated (Fig. 2e). Moreover, we found SRSF6 overexpression also has exon 20 inclusion-promoting activity as seen in RECTAS treatment or CLK1 overexpression (DMSO conditions in Fig. 2f). However, this SRSF6 overexpression effect was also canceled when CLKs are inhibited (CaNDY conditions in Fig. 2f). These observations together support the idea that RECTAS depends on CLK1 activity for exon 20 inclusion-promoting activity. We further validated the effects of compound treatment (RECTAS, TG003, INDY, CaNDY, ALGERNON, and SRPIN340) and knockdown of SRSF6 or CLK in FD patient fibroblasts with the homozygous IKBKAP-FD IVS20 + 6 T > C mutation, and confirmed the increased IKBKAP-FD exon 20 skipping by CLK inhibition (Fig. 2g and Supplementary Fig. 1a) or depletion for SRSF6 or CLK ( Fig. 2h and Supplementary Fig. 2a) for the endogenous transcripts. Primary fibroblasts without IVS20 + 6 T > C mutation of IKBKAP showed no exon 20 skipping following CLK inhibition ( Fig. 2i and Supplementary Fig 1b) or depletion for SRSF6 or CLK ( Fig. 2j and Supplementary Fig. 2b), showing the effect of RECTAS depends on the suboptimal nature of exon 20 donor sites due to IVS20 + 6 T > C (Fig. 1a). In addition, we also compared RECTAS with compounds previously reported for exon 20 inclusion promotion 40 , and we confirmed that RECTAS shows the highest activity (0.72 μM of EC 25 ) over those we tested (Fig. 2k, l). Two compounds, kinetin and (−)-epigallocatechin gallate (EGCG), reached EC 25 of RECTAS (Fig. 2k, m, n), but were 15.5-and 241-fold less potent and there was also cytotoxicity around the effective range for EGCG (Fig. 2n).
Having confirmed that CLK-dependent SRSF6 activation is required for restoration of normal splicing of IKBKAP-FD exon 20 by RECTAS, we then asked whether RECTAS affects the functions of CLKs and SRSF6. For this purpose, we synthesized a biotin-linked RECTAS analog (b-RECTAS) (Fig. 3a), which retained the IKBKAP-FD exon 20 inclusion-promoting activity, whereas biotin alone at 50 μM had no effect (Fig. 3b). We used b-RECTAS in a compound pull-down assay from HeLa cell lysates by neutravidin to evaluate its binding affinity to CLK1 and SRSF6. We found that b-RECTAS binds to CLK1 in a dose-dependent manner but not to SRSF6 (Fig. 3c). Recombinant CLK1 protein also bound to b-RECTAS in a dose-dependent manner, indicating their direct interaction (Fig. 3d). The interaction between b-RECTAS and CLK1 was effectively competed out by incubating with 500 nmol of free RECTAS (lane 6 in Fig. 3d). Western blotting for RECTAS-treated FD patient fibroblasts revealed that the SRSF6 phosphorylation was enhanced following RECTAS treatment (Fig. 3e), and this effect was detected by 6 h and disappeared by 10 h after washout of RECTAS (Fig. 3f). Since it was recently reported that several small molecules stabilize U1 snRNA-donor paring for SMN exon 7 through 5′-splice-site bulge repair 8 , we tested whether RECTAS has such activity on IKBKAP exon 20 donor site. RNA pull-down assay by IKBKAP exon 20 donor sequences with or without IVS20 + 6 T > C reproduced weakened affinity to U1 snRNP (detected by U1-70k and SmB/B′ subunits) by IVS20 + 6 T > C mutation (lanes 4 and 6 in Fig. 3g), whereas RECTAS had no effect on their affinity, denying its role in donor-U1 snRNA paring (lanes 5 and 7 in Fig. 3g).
The therapeutic effect of RECTAS in FD disease models. Having established the mechanism and target selectivity of RECTAS, we then assessed its therapeutic potential for FD, in two FD disease models: iPSC-SNs and IKBKAP-FD transgenic mice. As alternative splicing undergoes tissue-dependent regulation due to a varied distribution of RNA-binding proteins 41 , it is desirable to assess therapeutic potential in the context of sensory or autonomic neurons, which are affected in FD patients [18][19][20] . For efficient preparation of induced pluripotent stem cells-derived sensory neurons (SNs), we employed the embryoid-body method for the initial differentiation step (detailed in "Methods"). This modification resulted in higher viability and rapid derivation of the SNs (BRN3A + /TUBB3 + ) on day 12 (Fig. 4a) (see "Methods" for details). The effect of the compounds was tested by adding kinetin, RECTAS, or solvent only (0.1% dimethyl sulfoxide (DMSO)) to the culture media during the whole differentiation process. We found that 10 µM of RECTAS completely restored IKBKAP-FD exon 20 inclusion (lane 7 in Fig. 4b), while 10 µM of kinetin did not eliminate exon 20 skipping completely in patient iPSC-SNs (lane 6 in Fig. 4b). The effect of RECTAS on exon 20 inclusion was transient, and beginning around 2 h after the incubation and lasting for~4 h after washout ( Supplementary  Fig. 3). We also examined the effect of CLK inhibition for RECTAS responsivity in iPSC-SNs. Consistent with observations ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-24705-5 in IKBKAP-FD reporter-transfected HeLa cells, iPSC-SNs showed impaired RECTAS activity by CLK inhibition with CaNDY treatment (Fig. 4c).
we measured the effect of per os (p.o.) administration of RECTAS, using previously established IKBKAP-FD transgenic mice, in which an IVS20 + 6 T > C-mutant human IKBKAP gene was introduced in a BAC clone, in the presence of the endogenous murine wild-type Ikbkap, as an asymptomatic splicing model 42 . RECTAS was orally administered at 300 or 400 mg/kg body weight (BW) with a booster administration at 4 h intervals. Eight hours after the first administration, total RNA was extracted from dorsal root ganglia (DRG) to assess the splicing profile in sensory neurons. RT-PCR revealed enhanced inclusion (about 70% increase in PSI from~12.5 to~22%) of IKBKAP-FD exon 20 following RECTAS administration at both doses in DRG (Fig. 4g). We subsequently conducted transcriptome analysis for DRG, excised from MOCK and RECTAS-treated (300 mg/kg BW) groups. We found IKBKAP-FD exon 20 inclusion among the highly promoted exons (13th highest alterations) (Supplementary Fig. 5 and Supplementary Data 2), consistent with its selective effect in patient primary fibroblasts (Fig. 1b, c and Supplementary Data 1). Moreover, we confirmed there was no Rbm24 expression detectable in DRG irrespective of RECTAS administration (Fig. 4h), ruling out the involvement of Rbm24 as shown for human tissue cultures (Fig. 1d). Finally, we assessed the preclinical profile of RECTAS, checking for effects on body weight change and food consumption, as well as conducting necropsy and hematology tests, both after single p.o. administration (0, 100, 300, 1000 mg/kg BW) and 14-day repetitive p.o. administration (0, 40, 200, and 1000 mg/kg BW/day) (Supplementary Table 1). There was no genotoxicity for RECTAS in the micronucleus and Ames tests (Supplementary Table 1). These preclinical data together indicate that there was no anticipated adverse effect associated with RECTAS administration, assumingly due to its high target selectivity for IKBKAP-FD exon 20 (Fig. 1b, c and Supplementary Fig. 5).

Discussion
Recent studies revealed that the prevalence of splicing mutations is much higher than previously thought, as~30-50% of diseaseassociated mutations are estimated to cause abnormal splicing 43,44 , and the total number is further increasing due to the identification of deep-intronic splicing mutations through recent whole-genome sequencing and transcriptome analysis 13,[45][46][47][48] . Therefore, an innovative therapeutic strategy to target misregulation of splicing is urgently needed. Small-molecule compounds that target trans-acting splice factors offer an opportunity to systematically target a particular set of mis-splicing events that are regulated by a similar splicing context. In this study, we identified alternative exons that show a dependency on a subset of SRSFs and CLK kinases are rational targets for selective manipulation in both promotive and suppressive ways (Fig. 5). Our data indicate that enhanced SRSF6 activity is a crucial aspect of the splicing environment induced by RECTAS treatment, and the requirement of CLKs in this process suggests that the direct interaction of RECTAS with CLK1 (Fig. 3c, d) sterically activates cellular CLK1. As a precedent for PDK1, the regulatory kinase of AGC kinases, there is an allosteric small-molecule activator that binds to the HM/PIF-pocket to induce a conformational change 49 . We assume that RECTAS binding modifies the intramolecular conformation of CLK1 to promote its catalytic activity or provide accessibility of a signal mediator upstream of CLK1. Understanding the binding mode of RECTAS to CLK1 and the activating pathway for CLKs in the context of suboptimal donor RNA recognition by SRSF6 will provide further insights into this regulation.
We found that RECTAS p.o. administration of 200-300 mg/kg BW for 8 h achieved~22% PSI for exon 20 in DRG of IKBKAP-FD transgenic mice, versus~12.5% baseline PSI in mock or untreated controls (Fig. 4g). In the recent report from Morini et al. 50 , as little as~12.5% of IKBKAP exon 20 PSI (versus~10% baseline) by kinetin administration (200 mg/kg BW/day in the diet for 6 months) was sufficient for the phenotypic rescue of sensory loss. Though these data were obtained with different transgenic mice and cannot be directly compared, we expect that RECTAS should exhibit higher therapeutic efficacy in vivo than kinetin, which already completed a phase 1 clinical trial in FD patients (NCT02274051 in ClinicalTrials.gov). Kinetin has been reported to have high target selectivity in FD patient fibroblasts, and our study revealed a similar target-selective profile for RECTAS. Other RECTAS-responsive exons, such as TATA-box binding protein associated factor, RNA polymerase I subunit A (TAF1A) exon 4 and par-3 family cell polarity regulator (PARD3) exon 21 (Supplementary Data 1) had no apparent effect on cell viability. Presumably, this is because alternative splicing occurs physiologically for most genes, and the safety profile determined in preclinical tests further confirmed that RECTAS is welltolerated in rodents. In addition, we confirmed the higher activity of RECTAS than previously reported IKBKAP splice modifiers in a reporter assay (Fig. 2k), which further highlights the efficient activity of RECTAS for IKBKAP-FD exon 20 inclusion. These observations together suggest that RECTAS may prove to be a potent agent for the treatment of FD, and ongoing medicinal chemistry indicates further improvement in potency.
Searching in existing databases, there are numerous splicing mutations that resemble the context of IKBKAP-FD exon 20. For example, a splicing mutation in the potassium voltage-gated channel subfamily Q member 1 (KCNQ1), which is responsible for congenital long QT syndrome type I, has a similar splice enhancer motif, and the impaired exon of KCNQ1 is also manipulated bidirectionally in a similar manner as IKBKAP IVS20 + 6 T > C (our unpublished observations). The opportunity for bidirectional regulation by CLK inhibitors and agonists offers a strategic approach for individual splicing contexts, as promoting recognition of a target exon, on the one hand, provides a therapeutic effect for splicing mutations like the IVS20 + 6 T > C mutation of IKBKAP, whereas inhibition of a target exon, on the other hand, offers a therapeutic opportunity for splice-site-creating splicing mutations, as we previously demonstrated the therapeutic effects of CLK inhibitors in EDA-ID with the IVS6 + 866 C > T mutation of NEMO 13 and cystic fibrosis with the c.3849 + 10 kb C > T mutation of CFTR 11 . CLK-targeting strategies may also become increasingly relevant, as recent whole-genome and transcriptome analyses highlighted the widespread presence of deep-intronic mutations that lead to the creation of cryptic pseudoexons 13,[45][46][47][48] .
In addition, numerous splicing mutations have also been identified in recent cancer genomics studies 51,52 , although their individual contributions to malignancy remain largely unknown. CLK inhibitors (e.g., T-025 53 and SM08502 54 ) and SRPK inhibitors (e.g., SPHYNX 55,56 , SRPIN340 35 , and SRPKIN-1 57 ) described by other groups and us were recently reported to have anti-cancer activity in several cancer types. Moreover, smallmolecule compounds that target the essential splice factor SF3B, such as spliceostatin A and E7107 58,59 , have shown efficacy in leukemia models harboring mutations in the same or different trans-acting splice factors, namely SF3B1, SRSF2, or U2AF35, which sensitize the cells to the inhibitors 60-63 , though inhibition of SF3B also evokes global intron retention and cytotoxicity. In this regard, targeting CLKs by RECTAS, with a target selectivity sufficient for a favorable safety profile in preclinical tests, may provide an alternative therapeutic strategy to target cancer cells with vulnerable splicing machinery or CLK-dependent causal splicing misregulation.
Overall, our current observations with IKBKAP-FD exon 20 not only highlight RECTAS as a potential therapeutic agent for FD but also suggest the feasibility of chemical control of diseaseassociated missplicing by manipulating the splicing environment composed of SRSFs and CLKs in target cells. Facing the increasing prevalence of splicing mutations associated with human diseases, mechanism-oriented precision medicine for missplicing, as exemplified by this study, may provide a new framework for targeting such splicing diseases.

Methods
Cell lines. HeLa cells were obtained from the Japanese Collection of Research Bioresources (JCRB) Cell Bank of the National Institutes of Biomedical Innovation, Health and Nutrition (NIBIOHN) (JCRB9004) (Osaka, Japan) and cultured in Dulbecco's modified Eagle medium (DMEM) (Bibco of Thermo Fisher Scientific, Waltham, MA, USA) supplemented with 10% fetal bovine serum, 100 U/mL penicillin, and 100 µg/mL streptomycin. Neuro 2A cells were obtained from JCRB Cell Bank of NIBIOHN (IFO50081), and cultured in DMEM (GIBCO of Thermo Fisher Scientific, Waltham, MA, USA) supplemented with 1× GlutaMAXI (GIBCO of Thermo Fisher Scientific), 10% fetal bovine serum, 100 U/mL penicillin, and 100 µg/mL streptomycin. Fibroblasts from 19-and 26-year-old male patients with FD were provided by Coriell Institute (Camden, NJ, USA) (GM02342 (P1) and GM00850 (P2), respectively) and cultured in DMEM supplemented with 15% fetal bovine serum, 100 U/mL penicillin, and 100 µg/mL streptomycin. Primary fibroblasts from healthy donors (36-year-old male (TIG-114, C1) and 40-year-old female (TIG-108, C2)) were obtained from JCRB Cell Bank of NIBIOHN, and cultured in DMEM supplemented with 10% fetal bovine serum, 100 U/mL penicillin, and 100 µg/mL streptomycin. All cells were maintained in a cell incubator at 37°C with 5% CO 2 . The tissue culture study was conducted under ethical approval by the ethics committee of Kyoto University Graduate School and Faculty of Medicine.
IKBKAP (IVS20+6T>C) humanized mice. The IKBKAP (IVS20 + 6 T > C) transgenic mouse strain used in this study was obtained from Drs. Pickel and Slaugenhaupt 42 . Mice were bred in the light/dark cycle of 12 h/12 h, 40-60% humidity and 69-72°F temperature. RECTAS in suspension with 0.5% carboxymethyl cellulose was administered to more than 8-week-old mice p.o. at a dose of 300 or 400 mg/kg BW. Four hours later, an equivalent booster dose was given, and after another 4 h, the mice were euthanized and RNA extraction from dorsal root ganglia was performed. Extracted RNA was further analyzed by RT-PCR to measure the PSI value for exon 20 inclusion of IKBKAP. An animal study using IKBKAP (IVS20 + 6 T > C) mice was conducted under ethical approval by Cold Spring Harbor Laboratories.
Dual-fluorescence reporter system for splicing evaluation. We previously constructed the SPREADD reporter system for IKBKAP IVS20 + 6 T > C 17 , in which the splicing extent of IKBKAP exon 20 was quantified by the ratio of green fluorescent protein (GFP) and red fluorescent protein (RFP). For the reporter assay, the IKBKAP IVS20 + 6 T > C splicing reporter vector was transfected into HeLa cells (0.1 × 10 6 cells/well, seeded in a 12-well plate) with Lipofectamine 2000 transfection reagent (Thermo Fisher Scientific) and changed to fresh media at 4-5 h after transfection to be incubated with or without compounds for 24 h before sample preparation or fixation with 4% paraformaldehyde in phosphate-buffered saline (pH 7.4). Fixed cells were then scanned with a BZ-X710 fluorescence microscope (Keyence, Osaka, Japan), and analyzed for fluorescence intensities with a hybrid-cell count software of BZ-X analyzer (1.4.0.1) (Keyence). The resulting GFP/RFP fluorescence ratio was then plotted to indicate the relative extent of IKBKAP exon 20 inclusion.
Compound pull-down assay. b-RECTAS was synthesized as a RECTAS analog with IKBKAP exon 20-targeting activity with covalent biotin labeling. The resulting b-RECTAS was used for the compound pull-down assay with neutravidin beads. Neutravidin beads were incubated with 0, 100, or 200 nM of biotin-conjugated RECTAS for 2 h at 4°C. Following washing with Tris-buffered saline (pH 7.4), the biotin-RECTAS-bound beads were then incubated with HeLa cell lysate or GST-CLK1 recombinant protein expressed in Sf9 insect cells (ProQinase, Freiburg, Germany) for 16 h at 4°C. For competition assays, 100 nmol RECTAS was incubated during the protein binding step, followed by washing and eluting with Laemmli buffer. Eluted samples were analyzed by western blotting.
Cell viability assay. For cell viability assay, HeLa cells were seeded in a 96-well plate in triplicate and incubated with compounds for 72 h. Cell viability was measured by WST-8 assay using Cell Counting Kit-8 (Dojindo, Kumamoto, Japan). Absorbance at 450 nm was measured with ARVO X5 multimode plate reader (Perkin Elmer, Waltham, MA).
RNA-Seq. Total RNA was extracted from FD patient fibroblasts or DRG from IKBKAP (IVS20 + 6 T > C) transgenic mice, using TRIzol (Thermo Fisher Scientific) with DNase treatment. For FD patient fibroblasts, total RNA (2.5 µg) was poly-A purified with a Dynabeads mRNA DIRECT Micro Kit (Thermo Fisher Scientific), and used for library preparation with an Ion Total RNA-seq Kit v2 (Thermo Fisher Scientific). Single-end sequencing was performed on an Ion Proton System (Thermo Fisher Scientific). For DRG from IKBKAP (IVS20 + 6 T > C) transgenic mice, total RNA (0.6-0.7 µg) was poly-A purified and the library was prepared with TruSeq stranded mRNA library kit (Illumina, Inc. San Diego, CA). RNA-Seq was then conducted with NovaSeq 6000 (Illumina, Inc.), and Real-Time Analysis (RTA) 1.9 software (Illumina, Inc.) was applied for primary data processing. Sequence reads were split into 60-base fragments from the 5′ end and mapped to the human genome (GRCh38) with STAR (ver. 2.4.1d), using the Ensemble annotation GRCh38.94 as reference. 66 Sequence reads with average Phred quality scores lower than 17, and those from rRNA, tRNA, snRNA, snoRNA, or repetitive elements in the Repbase database 67 were eliminated. To calculate the PSI values, we used all internal exons from representative variants for each gene selected based on the nucleotide sequences of variants and coding sequences in UniProt 68 . The PSI value of the exon (n) was calculated as the number of junction reads of exon (n − 1) and (n) divided by the sum of junction reads of exon (n − 1) and (n) and exon (n − 1) and (n + 1). For analysis of FD patient fibroblasts, PSI values obtained from patient fibroblasts P1 and P2 were averaged, and exons with more than 15 reads per million junction read for the sum of their skipping forms and inclusion forms in DMSO and/or RECTAS-treatment conditions were used for the analysis. The Sashimi plots were generated using an Integrated Genomic Viewer (ver. 2.8.0) 69 . The RNA-seq data were deposited at the National Bioscience Database Center database for patient fibroblasts (accession: hum0180), and at the Gene Expression Omnibus of the National Center for Biotechnology Information for transgenic mice (accession: GSE161109).