Titin Circular RNAs Create a Back-Splice Motif Essential for SRSF10 Splicing

Supplemental Digital Content is available in the text.

RBM20 and SRSF10 into the pLKO.1-puro plasmid. ShRNA sequences are detailed in Supplemental Table II. To create the pCDH-cTTN1 construct, we amplified two amplicons from genomic DNA of hiPSCs (primers in Supplemental Table I): 1) from the middle of exon 78, including intron 78 and exon 79, till the end of intron 79; 2) from the beginning of exon 145, including intron 145, till the middle of exon 146. In a second PCR step we combined these two amplicons and ligated this product into the BamHI and NotI sites of the pCDH-CMV-MCS-EF1-Puro plasmid.
We created 2 mutations within the SRSF10 motif of pCDH-cTTN1 (from AAAGAACC to AAGGAGCC) by PCR based mutagenesis. Therefore, we amplified 3 PCR products from the pCDH-cTTN1 plasmid with primers introducing these mutations (Supplemental Table I): 1) from the middle of exon 78 to the introduced mutation in exon 79; 2) from the mutation in exon 79 to the introduced mutation in exon 145; 3) from the mutation in exon 145 to the middle of exon 146. In a second PCR step we combined these amplicons and ligated this product into the pCDH-CMV-MCS-EF1-Puro using the BamHI and NotI sites.
We created a full-length SRSF10 overexpression construct (pCDH-flag-SRSF10) with a N-terminal flag-tag by amplifying SRSF10 from cDNA derived from the hiPSC-CMs (primers Supplemental Table I). Furthermore, we created a pCDH-flag-EGFP control construct by amplifying EGFP from the pIRES2-EGFP plasmid. Both PCR products were cloned into the BamHI and NotI restriction sites of pCDH-CMV-MCS-EF1-Puro.
All plasmid sequences were verified by Sanger sequencing and occurrence of mutations excluded.

Virus production
To produce third-generation lentivirus of pLKO.1-puro, pLKO.1-dsRED or pCDH-CMV-MCS-EF1-Puro based constructs we co-transfected 4*10 6 HEK293T cells with 4 μg of the expression plasmid, 2.7 μg pMDLg/pRRE, 1 μg pRSV-Rev, 1.4 μg pVSVG using Genejammer (Agilent; 204130) according to the manufacturer's protocol. The next day medium was replaced to RPMI/B27 for the circRNA targeting viruses and to CDM3 medium for all other viruses. This medium containing the produced lentivirus was collected after 24 hours and either used directly for hiPSC-CM transduction or the amount of transducing units was first determined.
The amount of transducing units is determined by transducing 250.000 HEK293T cells with series of 50/100/200/500/1000 μl of medium with virus of the pLKO.1-dsRED plasmid.
Three days after, the cells were trypsinized and FACS sorted for the dsRED positive population. The condition with 10-20% positive cells was used to calculate the amount of transducing units assuming 1 viral copy per cell. The amount of tranducing units for an experimental virus and its corresponding control were determined in the same FACS experiment.

HiPSC-CM infection
HiPSC-CM were dissociated and replated 2 to 4 days before lentiviral transduction to ensure homogenous cell populations between conditions. Of all viruses containing a puromycin resistance cassette, 2 ml/well of medium with virus was freshly added to the hiPSC-CM in a 6well plate. After 5 days virus was removed and puromycin selection started with 8 μg/ml puromycin for 72 hours after which the cells were fixed or harvested. These experiments where either completely performed in RPMI/B27 or CDM3 medium depending on the medium the virus was generated in.
In the experiments to select the best shRNA to inhibit the circRNA, medium containing the shRNA-encoding lentivirus was freshly added to the hiPSC-CM and the experiment performed as described above without puromycin selection. In the follow-up experiments using cTTN1-shRNAs the amount of transducing units was the same for the cTTN1-shRNA and the negative control shRNA. The amount of transducing units ranged between 250000 and 850000 per well in 6-well plates for RNA, between 20000 and 120000 per well in 24-well plates for immunocytochemistry, and 15000 per well in 96-wellplates for apoptosis experiments, depending on the density of the plated hiPSC-CMs.

Cytoplasmic/Nuclear fractionation for RNA isolation
10 7 dissociated cells were spinned for 5 minutes at 160x g and gently resuspended in 550 μl precooled membrane lysis buffer (50 mM TRIS pH8.0; 140 mM NaCl; 1.5 mM MgCl2; 0.5% NP-40; 10 mM EDTA; 1 mM DTT; 1 μl/ml Superase-In, Invitrogen AM2696). After 5 minutes incubation on ice, lysed cells are centrifuged 5 minutes at 500x g at 4°C. Supernatant is transferred to a new tube and recentrifuged to remove any remaining nuclei. The cleared supernatant is transferred to a new tube and 1 ml TriReagent added to start RNA isolation.
Nuclei pellet of the first centrifugation is washed in 175 μl precooled membrane lysis buffer, by gentle resuspension and centrifugation. The nuclei pellet is again resuspended in 175 μl precooled membrane lysis buffer, 1 ml TriReagent added, samples are vortexed and 5 times passed through a 20-gauge needle to break the nuclear membrane.

RNA isolation
Total RNA was isolated from all samples using 1 ml TriReagent (Sigma Aldrich; T9424).
TriReagent was either added to frozen tissue samples, which were homogenized using a MagNA Lyser (Roche), to a frozen cell pellet after TrypLE Express dissociation, or directly to live cells growing on a dish. Total RNA isolation was performed according to the manufacturer's protocol.

RNA-sequencing & analysis
We performed RNA-seq on 3 biological replicates of hiPSC-CMs transduced with a cTTN1 or negative control shRNA. Therefore, RNA was confirmed to have a RIN score >8 with the Agilent 2100 Bioanalyser. Total RNA samples were sequenced on a Illumina NextSeq 500 platform in paired-end mode with a read length of 150 bp. Sequencing depth was approximately 50 million raw reads per sample. Base-calling was performed using the bcl2fastq 2.0 conversion software from Illumina.
Quality control of fastq files was performed using FASTQC (https://www.bioinformatics. babraham.ac.uk/projects/fastqc/). Trimmomatic version 0.351 38 was used to remove Illumina adapters and low quality bases, using a Phred score cutoff of 30 while discarding reads with a length below 25 bases. The RNA-seq reads were then aligned against the human genome using TopHat2 version 2.0.14 39 with default values and the UCSC hg19 annotation file. Differential gene expression analysis was performed using the R Bioconductor package, DESeq2. 40 GenomicRanges infrastructure was used to count the number of aligned reads overlapping with each gene. To test for exon usage differences we used the R Bioconductor package DEXSeq version 1.20.0 41 , using GenomicRanges infrastructure to count the number of aligned reads overlapping with each exonic region (see Supplemental Excel File IV).
Pathway enrichment analysis was performed using the R package ReactomePA. It implements hypergeometric models to assess whether the number of selected genes associated with a reactome pathway is larger than expected. The gene names of the differentially spliced exons or differentially expressed genes (p-adj cut-off ≤0.05 and absolute log2 fold change ≥1) were used as input for the pathway enrichment analysis.

qRT-PCR for mRNA
To detect mRNA or splice isoform levels, 250 ng to 1 μg RNA was DNAse treated with DNAseI amplification grade (Invitrogen; 18068015) and reverse transcribed using Superscript II reverse transcriptase (Invitrogen; 18064014) with oligo-dT and random hexamer primers according to the manufacturer's protocol. cDNA was diluted 5 times and 2 μl used as input for the qPCR.
qPCR was performed using 1 μM primers (Supplemental Table I) and LightCycler 480 SYBR Green master 1 (Roche; 04887352001) on a LightCycler 480 system II (Roche) using the following cycling program: 5 minutes pre-incubation at 95°C; 40 cycles of 10 seconds denaturation at 95°C, 20 seconds annealing (temperatures in Supplemental Table I), and 20 seconds elongation at 72°C. Data were analyzed using LinRegPCR quantitative PCR analysis software 42 and the starting concentration of transcripts estimated by this software was corrected for the geometric mean of the estimated starting concentration of the three housekeeping genes HPRT, GAPDH and TBP.

(q)RT-PCR for circRNAs
For some circRNA experiments the RNA was first treated with RNase R to remove linear RNA. Therefore we treated 1 μg total RNA with 5 units of RNase R (Epicentre; RNR07250) at 37°C for 10 minutes followed by heat inactivation at 95°C for 3 minutes. Control samples in these experiments were treated similarly without addition of RNase R.
For RT-PCR detection of circRNAs, RNA was DNAse treated and reverse transcribed as described above using a combination of random hexamer primers and oligo dT. For previously RNase R treated samples reaction volumes were doubled. cDNA was diluted 5 times and 1 to 4 μl was used for input in the end-point PCR reactions and 2 μl for input in qPCR.
End-point PCR was performed to detect circRNAs using HOT FIREPol DNA polymerase and 1 μM primers as detailed in Supplemental Table I. Primers are designed in a divergent manner in the two exons that backsplice to create the circRNA of interest. Therefore, the band of interest is the smallest band detected in the agarose gel, which is verified by Sanger sequencing. The program used to detect the circRNAs contained the following cycling conditions: 30 seconds at 95°C denaturation, 30 seconds at variable temperatures (Supplemental Table I) annealing and 10 seconds at 72°C elongation. The amount of cycli ranged between 30 and 45 depending on the type of experiment, amount of starting material and efficiency of the primers/amount of extra amplified amplicons by the primers.
To quantitatively detect cTTN1 we designed a custom made Taqman primer probe set, which contained a non-modified forward primer 5'-AGAGGTGCCCAAGAAGCTC-3' and reverse primer 5'-ATGGATTCTCCCTGCTTTGC-3' and a FAM-labeled MGB probe 5'-AC CCAAAGAACCTCCAAA-3'. Here the probe was designed on the backsplice junction in such a way that the amount of mismatches between cTTN1 and all other circRNAs containing the SRSF10 motif in their backsplice junction was at least one and for most of them 2-4. We performed real-time PCR using 900 nM primers, 250 nM probe and LightCycler 480 probes master (Roche; 04707494001) on a LightCycler 480 system II (Roche) using the following cycling program: 10 minutes pre-incubation at 95°C; 45 cycles of 15 seconds denaturation at 95°C, 1 minute annealing and extension at 60°C. Data were analyzed using LinRegPCR quantitative PCR analysis software. 42

qRT-PCR for miRNAs
To detect miRNA levels, 500 ng to 1 μg RNA was DNAse treated and reverse transcribed using the miScript reverse transcription kit (Qiagen; 218161) with the miScript HiFlex Buffer.
cDNA was diluted 8 times and 2 μl was used as input for the qPCR.
qPCR was performed using High Resolution Melting Master (Roche; 04909631001) in a reaction where an end-concentration of 2.5 mM MgCl2 and 1 μM primers was used. The forward primer was specific for the miRNA (Supplemental Table I) and the reverse primer complementary to the adapter sequence of the RT-primer in the miScript RT kit was general for all miRNAs. Real-time PCR reactions were performed on a LightCycler 480 system II (Roche) using the following cycling program: 10 minutes pre-incubation at 95°C; 40 cycles of 45 seconds denaturation at 95°C, 45 seconds annealing at 55°C, and 45 seconds elongation at 72°C. Data were analyzed using LinRegPCR quantitative PCR analysis software 42 and the starting concentration of miRNAs estimated by this software was corrected for the geometric mean of the estimated starting concentration of the housekeeping genes HPRT and GAPDH.

Immunocytochemistry
All cells for immunocytochemistry were plated on 12 mm glass coverslips. Untransduced hiPSC-CMs were cultured for 1 week in RPMI/B27, and transduced hiPSC-CM for 2 weeks as described above, all on matrigel-coated coverslips. Cells were fixed in 4% paraformaldehyde for 10 minutes at room temperature and washed 3 times in PBS. Cells were permeabilized with 0.1% or 1% Triton X-100 in PBS for 8 minutes and unspecific antibody binding was blocked by 1 hour incubation with 4% goat serum or BSA. Primary antibodies (Supplemental Table III) were diluted in PBS with 4% goat serum or 4% BSA with 1% Triton X-100 and incubated overnight at 4°C or 3 hours at room temperature. Cells were washed 3 times in PBST or PBS and afterwards incubated for 1 hour at room temperature in the dark with 1:250 diluted secondary antibodies (Supplemental Table III) in PBS with 4% goat serum or BSA with 1% Triton X-100. Cells were washed 3 times in PBST or PBS. Nuclei were counterstained with DAPI (1:5000) for 5 min and mounted in Mowiol (Sigma, 81381) Pictures were taken blinded on a Leica TCS SP8 X/DMI6000 inverted confocal microscope with a 63x oil immersion lens. Quantification of myofibrillar structure was performed by separate scoring for striation and sarcomere organization. For sarcomere organization we used the following scores: 0) No myofibrils; 1) randomly scattered, punctuate or fuzzy appearance of myofibrils; 2) not well organized myofibrils structures with no alignment; 3) slightly organized/aligned myofibrils; 4) high degree of myofibril alignment. For striation we used the following scores: 0) No visible striation; 1) Striation visible in some myofibrils; 2) Clearly visible, distinct striation in almost all myofibrils. Quantification of RBM20 localization was performed by dividing cells in 3 groups: 1) only nuclear; 2) nuclear and cytoplasmic; or 3) only cytoplasmic, and the percentage of cells belonging to each group was calculated.

(Immuno)FISH
Cells were dissociated and replated on matrigel coated 12 mm coverslips. While replating 20000 TU of a cTTN1 shRNA or negative control shRNA virus was added to the RPMI/B27 medium with thiazovivin in which the cells were plated to allow immediate transduction. 6 days after transduction, cells were washed in PBS and fixed in 4% formaldehyde (v/v) and 0.05% (v/v) glacial acid in PBS for 17 minutes at room temperature. After 3 PBS washes of 5 minutes each, ice-cold 70% ethanol is added to the cells and they are incubated for 20 minutes at -20°C. Cells are again washed in PBS and incubated 30 minutes at room temperature in 1% Triton X-100 and 1% Saponin in PBS. After another PBS wash, cells are incubated for 15 minutes at 37°C in proteinase K (4 μg/ml in PBS) in the FISH experiments and 5 to 7.5 minutes depending on cell density for the immunoFISH experiments. Cells are washed again in PBS, followed by post-fixation in 4% formaldehyde and 0.05% glacial acid in PBS for 5 minutes.
Cells were then incubated for 30 minutes in 3% peroxide in water and washed in PBS again. This washing step is followed by 3 steps of 5 minutes each of increasing ethanol concentrations from 70 to 100% ethanol after which cells are air-dried. To detect the cTTN1, cells were incubated overnight at 37°C with 10 ng of a FAM-labeled DNA-probe directed towards the cTTN1 backsplice junction ( IBA life sciences). The next day the probe was washed away in 3 washes of 10 minutes at 60°C in 2x SSC buffer and 3 washes in 0.2x SSC buffer, followed by a PBS wash. Afterwards cells were blocked in 10% horse serum in PBS for 30 minutes at room temperature. For immunoFISH experiments cells were incubated for 3 hours with an anti-RBM20 antibody (Supplemental Table III After another PBS wash, cells were incubated for 5 minutes with 1:1000 DAPI in PBS, washed again in PBS and mounted in prolong glass antifade (Invitrogen; P36982). Pictures were taken on a Leica TCS SP8 X/DMI6000 inverted confocal microscope with a 63x oil immersion lens.
Total protein (20 μg) was separated by electrophoresis on pre-cast 4-20% gradient polyacrylamide gels (Bio-Rad; 4561094) and transferred to a methanol activated PVDF membrane (Bio-Rad; 170-4272) using a Trans-Blot Turbo transfer system (Bio-Rad) for 10 minutes at 25V. This membrane was blocked for 1 hour in 5% non-fat dry milk in TBST and incubated with the first antibody (Supplemental Table III) in 5% non-fat dry milk in TBST overnight at 4°C. The next day membranes were washed 3 times 10 minutes in TBST, incubated 1 hour at room temperature with the HRP-linked secondary antibody 1:5000 in 5% non-fat dry milk in TBST, and washed 3 times for 10 minutes in TBST. Bands were detected using the ECL prime western blotting detection reagent (Amersham; RPN2236) and images acquired using the ImageQuant LAS4000 (GE Healthcare).
For immunoprecipitation of SRSF10 in HEK293T cells we co-transfected 4 μg pCDH-cTTN1 or pCDH-cTTN1mutant with 4 μg pCDH-flag-SRSF10 or pCDH-flag-EGFP into 4*10ˆ6 HEK293T cells using genejammer according to the manufacturer's protocol and these cells were collected 48 hours after transfection. Cells were lysed using 1 volume lysisbuffer compared to the volume of the cell pellet according to the protocol of the MagnaRIP RNA binding protein immunoprecipitation kit (Millipore; 17-700). Immunoprecipitation was further performed using 100 μl lysate input per antibody according to the manufacturer's protocol or repeated under more stringent criteria by adding 1 or 3 M Urea to the washing buffer and blocking of the beads with 1% ultrapure BSA (Invitrogen; AM2616) in RIP wash buffer for 1 hour before the lysate was added. After immunoprecipitation and RNA isolation according to the protocol, cDNA is prepared and circRNAs are detected as described above.

Apoptosis measurements
Apoptosis was measured using the Caspase-Glo 3/7 Assay (Promega; G8090). Therefore hiPSC-CMs were plated in 96-wellplates and transduced with 15.000 transducing units of cTTN1 or negative control shRNAs. Apoptosis was measured 4 days after transduction according to the manufacturer's protocol on a Glomax multi detection system (Promega).
For the TUNEL staining were hiPSC-CMs transduced with cTTN1 or negative control shRNA while being plated on 12 mm glass coverslips. Cells were fixed in ice-cold methanol for 10 minutes 4 days after transduction. After fixation cells were air-dried at room temperature
Agarose was allowed to solidify and silicon racks containing 2 posts per well (EHT Technologies; C0001) were positioned in the casting molds. The cell-hydrogel suspension consisting of hiPSC-CMs, fibrinogen, thrombin and matrigel was then poured around the posts and incubated at 37°C in a cell incubator for 1.5-2 hours to enable polymerization. EHTs adhered to the silicon racks were transferred to culture medium (DMEM/F12 low glucose, Sigma Aldrich D5546; 5% heat inactivated horse serum; 1% penicillin/streptomycin; 0.1% (w/v) Aprotinin, Sigma-Aldrich A1153; 0.1% (w/v) Insulin, Sigma Aldrich I9278) and medium was replaced 3 times a week. EHTs demonstrated regular contractions 4-6 days after generation. Video-optical recordings of the EHTs (60 frames per second) were acquired at 7 and 14 days after EHT generation and contraction profiles were assayed using Musclemotion TM algorithm. 25

Calculation of genetic constraint of genomic regions
We compared the observed number of rare variants (  The code used to perform the genetic constraint analysis is available at https://github.com/ImperialCardioGenetics/cTTN. This code was run in R3.6.3 with as required package ggplot2.

Bioinformatic prediction of exons included in cTTN1
For this prediction we aligned the paired-end reads of 3 RNAseR treated RNA samples isolated from human left ventricular tissue 12 against the human genome using TopHat2 version 2.0.14 39 with default values and the UCSC hg19 annotation file. We then used the computational tool circAST 24 to determine which exons within the locus of cTTN1 (exon 79-145) are detected in the RNAseR treated samples to determine which exons are derived from circRNAs in this region and could be included in full-length cTTN1 (Supplemental Excel File III).

Statistics of molecular experiments
Data obtained from hiPSC-CM are a combination of 2 to 5 independent experiments on cells from independent differentiations, with at least n=2 biological replicates per independent experiment. Data of these independent experiments are combined using Factor Correction 20 , where the control condition was used as a reference to calculate the correction factor for which all the datapoints of that experiment were corrected. As a consequence shown data for continuous variables are a mean+/-SEM of n=6-15 biological replicates derived from 2-5 differentiations. For categorical data the percentage of cells in all groups is depicted per condition.
To compare continuous variables between two groups we used the Mann-Whitney U-test and between three groups we used the Kruskal-Wallis test combined with the Dunn post-hoc test, these tests were performed in GraphPad Prism Software version 8. To compare the distribution of cells in different categories (e.g. scoring of phenotype) between two groups we used the chi-square test. To compare the effect of loss of RBM20 over time (day 1-8), we made use of a 2-way ANOVA to test the effect of loss of RBM20, the day effect and the interaction term. To fit the linearity condition we performed a ln transformation for cTTN1, N2BA-G, CACNA1 ex9*-10, and CAMK2D_9. In case the interaction term was significant we determined the differences between the shRNA against RBM20 and the negative control shRNA by pairwise comparison, in which we used Bonferroni correction for multiple testing. Only in RBM20 and CACNA1 ex9-10 the interaction term was not significant. In case of CACNA1 ex9-10 the condition (loss of RBM20) was also not significant in the ANOVA indicating no significant effect of loss of RBM20 and in the case of RBM20 the condition was significant, indicating a significant effect of the shRNA against RBM20 at all timepoints, which was further confirmed by the pairwise comparison. The chi-square tests and the analysis of the RBM20 time-points were performed using IBM SPSS statistics version 26. All performed tests were two-sided and a p-value <0.05 was considered significant.

Data and code availability
Raw sequencing data for circRNA detection in human heart tissue are available via NCBI

Supplemental Figure VII. Engineered heart tissue contraction after loss of cTTN1
Representative contraction profiles from the negative control (SCR) and cTTN1 knockdown EHT groups at 7 and 14 days after generation of the EHTs (a

Supplemental Excel File I: RNA binding protein analysis of RBM20 dependent vs independent circRNAs within TTN
CircRNAprofiler output of the RNA binding protein analysis, containing the coordinates of the circRNAs analysed, the detected motifs in backsplice junction and full circRNA sequences and the comparison between RBM20 dependent and independent circRNAs for the SRSF10 binding motifs.

Supplemental Excel File II: RNA binding protein analysis of cTTN1, all highly expressed TTN circRNAs and linear TTN mRNA transcript
CircRNAprofiler output of the RNA binding protein analysis, containing the coordinates of the TTN-derived circRNAs analysed, the detected motifs in backsplice junctions and full circRNA sequences of these circRNAs, the detected motifs specifically in cTTN1 and the detected motifs in the linear mRNA transcript including all 363 TTN exons. This table also contains the comparison of SRSF10 binding motifs between TTN and non-TTN-derived circRNAs and between the TTN-derived circRNAs and the linear TTN mRNA transcript.

Supplemental Excel File III: CircAST output based on RNAseq of 3 RNAseR treated human myocardial samples
This table shows the included exons in circRNAs within the genomic region spanning exon 79 to 145, which comprises the cTTN1 region and gives an indication which exons could be included in cTTN1. Given are the exons detected in the 3 separate myocardial samples and the exons excluded from circRNAs in all 3 samples.

Supplemental Excel File IV: RNA sequencing analysis after cTTN1 inhibition
This table contains the output of the differential expression and exon usage analysis after shRNA-mediated knockdown of cTTN1 in hiPSC-CM. Furthermore, it contains the results of the pathway analysis based on the differential expressed genes or exons detected in the first analysis (cut-offs padj ≤0.05 and log2FC ≥1).

Supplemental Excel File V: Predicted miRNA binding sites within cTTN1
Results of miRNA binding site prediction by circRNAProfiler for all exons within the cTTN1 genomic region. Shown are the miRNAs included in the analysis after filtering on cardiac expression and the predicted binding sites for these miRNAs, where all the information is included to construct the actual binding sites. The different sheets include miRNA binding sites based on different stringencies: sheet 1 totalMatches=7, maxNonCanonicalMatches=0; sheet 2 totalMatches=7, maxNonCanonical Matches=1; sheet 3 totalMatches=6, maxNon CanonicalMatches=2.

Supplemental Excel File VI: Predicted miRNA binding sites within MYBPHL 3'UTR
Results of miRNA binding site prediction by circRNAProfiler for the 3'UTR of MYBPHL.
Shown are the miRNAs included in the analysis after filtering on cardiac expression and the predicted binding sites for these miRNAs, where all the information is included to construct the actual binding sites. The different sheets include miRNA binding sites based on different stringencies: sheet 1 totalMatches=7, maxNonCanonicalMatches=0; sheet 2 totalMatches=7, maxNonCanonical Matches=1; sheet 3 totalMatches=6, maxNonCanonical Matches=2.