Drug discovery with an RBM20 dependent titin splice reporter identifies cardenolides as lead structures to improve cardiac filling

Figures

• 

  Fig 1. Dual luciferase splicing reporter assay (DLR assay) to identify splicing modulators. (a) Minigene containing titin PEVK exons 4–13. Boxes indicate exons, blue lines mature transcripts generated with and without RBM20. (b) PCR products of alternative transcripts produced from the PEVK minigene by RBM20. RBM20 increases the amount of transcripts lacking all alternative exons or only retaining alternative exon 5. (c) Quantitative PCR to relate inclusion of PEVK exon 8 to the constitutive exon 13 presented as percent spliced in (PSI). RBM20 reduces inclusion of exon 8 by ~70% (N = 3). (d) Dual luciferase splicing reporter with firefly luciferase (FLuc) integrated into exon 8 and renilla luciferase (RLuc) downstream of exon 13. The FLuc/RLuc ratio reflects the inclusion ratio of alternative exon 8. (e) RBM20 shifts alternative splicing of the dual luciferase reporter (DLR) construct to exclude all alternative exons. (f, g) Quantitative PCR (N = 3) and FLuc/RLuc activity (N = 8) produce similar readouts with increased sensitivity of the luciferase based assay. ***P<0.001 versus CTRL (Dunnett’s post-test). Data are presented as mean ±SD.

• 

  Fig 2. Identification and validation of splice active compounds by a semi-automated high-throughput screen. (a) Optimization of the splicing reporter to RBM20 ratio by co-transfection of HEK293 cells. The assay is saturated at a 5-fold excess of RBM20 (N = 8). (b) Assay kinetics with maximum effect after 60 hours of incubation (N = 8). Polypyrimidine Tract Binding Protein 1 (PTBP1) served as negative control not leading to exon exclusion of the splice reporter. (c) Assay suitability for a high-throughput approach—z’ values >0.5 are adequate. (d) Screening strategy to identify splice active compounds. The DLR assay was adapted to the 384-well format followed by the pilot screen with >34,000 compounds at 10 μM. Potentially active compounds were re-evaluated in 9 serial dilutions. Resulting candidates were validated manually in the 96-well format leading to the identification of 7 inhibitors that belong to the group of cardenolides. (e) Number of compounds (cpds) passing the different steps of the screening procedure. *P<0.05, **P<0.01, ***P<0.001 versus CTRL (Dunnett’s post-test). Data are presented as mean ±SD.

• 

  Fig 3. Inhibitors of titin splicing identified by HTS. For each compound concentration dependent activity in the dual luciferase reporter assay and cell viability are plotted. Dashed lines indicate the concentrations used for validation on RNA level. The tables provide kinetic information. Validation by RT-PCR (agarose gel) is quantified by calculating the percent spliced in values (PSI). (a-c) Inhibition of alternative splicing was validated manually in the 96-well format using the DLR assay and conventional as well as quantitative RT-qPCR using an independent genomic minigene derived from TTN exons 241-43 (N = 4). Cardenolides efficiently reduce titin splicing by RBM20 with different potency (splicing IC50). The effect on splicing translates to reduced viability of HEK293 cells (IC50 values splicing vs. viability). (d) The steroid hydrocortisone does not interfere with splicing activity (N = 4). Compared to the cardenolides it lacks the lactone ring at C17 (chemical structures provided on the right). *P<0.05, **P<0.01, ***P<0.001 versus CTRL (Dunnett’s post-test). Data are presented as mean ±SD.

• 

  Fig 4. Cardenolides are potent inhibitors of RBM20 dependent titin splicing. Figure layout and panels as described in Fig 3. (a, b) Additional cardenolides not present in the original library were selected to evaluate their potency to inhibit RBM20 dependent titin splicing. The structural analogs digoxin and digitoxin both efficiently inhibit RBM20 mediated splicing. (DLR N = 4; RNA N = 3). ***P<0.001 versus CTRL (Dunnett’s post-test). Data are presented as mean ±SD.

• 

  Fig 5. Specificity of the inhibitory effect of cardenolides on RBM20-mediated titin splicing. Figure layout and panels as described in Figs 3 and 4. (a) An independent minigene containing FMNL3 exons 25-26 was co-transfected with RBFOX1 and does not respond to digoxin until cell viability is below 30% (N = 3). (b) Digitonin does not affect the splice reporter readout until cell viability is below 20%. There is no effect on the mRNA level (N = 3). (c) Isoginkgetin treatment reduces alternative splicing and mRNA maturation (reduced PCR product at 507 bp in the control at 30 μM). (d, e) Digoxin treatment of co-transfected HEK293 cells leads to decreased expression of RBM20, which is not affected by addition of the proteasome inhibitor MG132. (f) Digoxin dependent splicing inhibition does not depend on the reporter to RBM20 ratio. High concentrations of RBM20 co-transfected with the reporter do not compensate the inhibitory effect. 2-WAY-ANOVA: interaction, column and row factor <0.0001, (N = 4). *P<0.05, ***P<0.001 versus CTRL (Dunnett’s post-test). Data are presented as mean ±SD.

• 

  Fig 6. Digitoxin affects mRNA processing and various signaling pathways related to cell division, signal transduction and cancer. Gene ontology according to WikiPathways analysis of transcripts differentially regulated by digitoxin in HEK293 suggest specific effects that include pathways related to growth and cell cycle as well as differential expression of > 60 genes involved in mRNA processing (adjusted P ≤ 0.05 [Benjamin-Hochberg], two-sided hypergeometric test).

• 

  Fig 7. Digitoxin alters transcription of splicing related genes. (a) Stepwise assembly of major spliceosomal small nuclear ribonucleo proteins (snRNPs) on pre-mRNA. The removal of non-coding sequence from pre-mature mRNA transcripts requires five RNP complexes U1, U2, U4, U5 and U6. Boxes display the genes related to RNA unwinding and complex remodelling/recycling (DExD/H box proteins) as well as factors necessary for the transesterification. 9 of 12 genes are significantly downregulated by at least 20%. Boxes with thick outlines represent proteins binding RBM20 [22]. (b) Target specific factors from human spliceosomal complexes A, B, Bact and C that are dynamically integrated into active spliceosomes. Almost half of the genes in RBM20 dependent splicing complexes are significantly regulated by digitoxin.