We identified three early-onset paroxysmal AF probands from separate families harboring the same rare TTNmv (Chr 2q31.2 (GRCh38): g. 178539797G>A) without additional loss-of-function TTN variants. The clinical characteristics of the three probands are summarized in Table 2. This missense variant localizes in the A-band of TTN and causes the substitution of a conserved residue threonine by an isoleucine (p.Thr32756Ile, NM_001267550.2 (TTN), c.98267C>T Figure 1C–D, Figure 1—figure supplement 1A). In the gnomAD reference population, the TTN-T32756I variant has a frequency of 0.000245 overall, 0.0046 within the subpopulation of African ancestry, and is impartially distributed across sexes and ages (Figure 1E, Figure 1—figure supplement 1B–E). The TTNmv has also been described in Clinvar (NM_001267550.2(TTN):c.98267C>T (p.Thr32756Ile), Supplementary file 7Supplementary file 7) and gnomAD v4.0.0 (SNV:2–178539798 G-A(GRCh38)) in patients with DCM and other diseases.

To functionally characterize the TTN-T32756I variant, we exploited wild-type (WT) human iPSC lines and introduced the point mutation c.98267C>T in the TTN allele (Figure 2A, Figure 2—figure supplement 1A–B). The iPSC lines expressing the pluripotent markers Sox2 and Oct4 and showing normal karyotype were first differentiated into cardiomyocytes (Figure 2—figure supplement 1C–D), and then a retinoic acid-based and comprehensive maturation protocol was applied to generate the iPSC-aCMs as we described previously (Ly et al., 2021; Ly et al., 2022). Given that cardiac contractility is a key component of heart function and is linked to cardiac disorders associated with TTN (Schick et al., 2018), we first assessed the contractility of the WT and mutant iPSC-aCMs and observed both reduced contraction duration and abnormal relaxation in TTN-T32756I-iPSC-aCMs (Figure 2B–E, Figure 2—figure supplement 2A–B). Compared to WT, the beating frequency of the TTN-T32756I-iPSC-aCMs was significantly increased (52±7.8 vs 98±7.5 beats per min, p=0.001; Figure 2C) coupled with the reduction of the contraction duration (456.5±61.45 vs 262.9±48.16 msec, p=0.032; Figure 2D), the peak-to-peak time (1529±195.5 vs 636.6±135.8 msec, p=0.004; Figure 2—figure supplement 2B), and the relaxation duration (281.5±42.95 vs 79.40±21.14 msec, p=0.003; Figure 2—figure supplement 2A). In contrast, the contraction amplitude of the mutant was increased (8.1±0.8 vs 10.6±7 au, p=0.040) without any significant changes in time-to-peak (Figure 2E, Figure 2—figure supplement 2C), suggesting an increased contractile force by the TTN-T32756I-iPSC-aCMs. As sarcomere disorganization is often observed in TTN-related cardiomyopathies and underlies contractile dysfunction (Hinson et al., 2015), we explored whether TTNmvs perturbed the organization of the sarcomere. Surprisingly, we did not find any disarray in the sarcomeres of either WT or TTN-32756I-iPSC-aCMs. Using transmission electron microscopy (TEM) and immunofluorescence (IF), we observed symmetrical arrays of recurring sarcomeres in the WT and the mutant atrial cardiomyocytes with no changes in the sarcomere length (Figure 2F–G, Figure 2—figure supplement 2D–F). Although no sarcomeric perturbation by the T32756I variant was observed, the aberrant contractility displayed by the TTN-T32756I-iPSC-aCMs suggests that the TTNmv may be pathogenic.

As studies have shown that abnormal atrial electrophysiology and calcium handling underlie atrial arrhythmogenesis and contractile dysfunction (Nattel et al., 2020; McCauley et al., 2020), we hypothesized that the TTN-T32756I variant modulates atrial APs, ion channels, and intracellular calcium. We first tested the spontaneous and paced AP characteristics of the isolated WT and the TTN-T32756I iPSC-aCMs, which revealed a significant shortening of the AP duration (APD) at 10%, 50%, and 90% repolarization; however, there was no change in the peak amplitude (Figure 3A–C, Figure 3—figure supplement 1A–B, Figure 3—figure supplement 2A–D). Since increased potassium current (I_k), especially the augmented delayed rectifier potassium current (I_ks), causes AF by reducing the APD (Ly et al., 2022; Menon et al., 2019), we assessed I_ks by using whole-cell voltage clamping. We observed an increase in both I_k and I_ks in TTN-T32756I-iPSC-aCMs compared with the WT (Figure 3D–F, Figure 3—figure supplement 1C–E); thus, confirming that the APD shortening is partly due to the increased I_k. To test the effect of T32756I on atrial calcium handling, we measured the intracellular calcium transients at the excitation-contraction coupling moment of the iPSC-aCMs using fluorescent calcium imaging. Compared to WT, spontaneously beating TTN-T32756I-iPSC-aCMs exhibited increased frequency along with a significant reduction in the time to 50% and 90% decline of calcium transients (Figure 3G–I, Figure 3—figure supplement 1F). We also saw a similar decile of calcium transients in the paced iPSC-aCMs (Figure 3—figure supplement 2E–H). However, TTN-T32756I-iPSC-CMs exhibited similar calcium transient amplitudes as the WT, indicating that there was no change in the availability of the intracellular calcium for each contraction (Figure 3—figure supplement 1G, Figure 3—figure supplement 2F). Our findings suggest that the reduced APD resulting from increased potassium current and the altered timing of the calcium transient may create a reentrant substrate for AF.

To elucidate the underlying molecular mechanism by which TTN-T32756I is associated with contractile defects and ion channel remodeling, we performed transcriptomic sequencing of mutant iPSC-aCMs and WT. Comparison of total RNA levels in TTN-T32756-iPSC-aCMs with WT-iPSC-aCMs and differential expression analysis showed genes related to cardiac muscle contraction and calcium handling were predominantly affected (Figure 4A, Figure 4—figure supplement 1A–C). Gene ontology (GO) pathway enrichment further showed that the top significantly downregulated cardiac-related GO Biological Processes (-BP) in TTN-T32756I iPSC-aCMs included key processes such as cardiac myofibril assembly, skeletal muscle myosin thick filament assembly, extracellular matrix organization, and regulation of potassium ion transmembrane transporter assembly among others (Figure 4B). As for GO-Molecular function pathways (GO-MF), outward potassium channel activity, voltage-gated potassium channel activity, calcium channel activity, and gap junction channel activity were enriched (Figure 4C). Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis identified downregulation in critical pathways including Adrenergic signaling in cardiomyocytes, the Phosphoinositide 3-kinase (PI3K) signaling pathway, Dilated cardiomyopathy pathway, Hypertrophic cardiomyopathy pathway, and Calcium signaling pathway (Figure 4D), which are crucial for normal cardiac electrophysiology and contractility. Upregulated pathways in both GO and KEGG analyses were mostly restricted to neuronal pathways, including axon and neuron development, the Hippo signaling pathway, and the Notch signaling pathway. In the transcriptomic analysis of TTN-T32756I-iPSC-aCMs, several transcription factors (TFs) showed significant alterations, highlighting their potential roles in the mutation’s impact (Figure 4E, Figure 4—figure supplement 1D). Top-cardiac-related upregulated TFs include MYC, linked to cell growth and apoptosis, suggesting an increase in cellular proliferation. Stress-related TFs such as NFE2L2 and CEBPB are also elevated, indicating a heightened stress response. Additionally, TFs like ESRRA and FOXM1, which regulate metabolic processes and cell cycle, respectively, and FOXO1, which regulates metabolism and stress resistance. Key TFs in muscle function, such as MYOD1, MYOCD, KLF5, and NKX2.5, were enhanced, alongside SMAD2, CREBBP, PITX2, NFKB1, and GATA4, which are vital for diverse roles ranging from signal transduction to immune regulation (Figure 4E). Conversely, downregulated transcription factors include SIRT1, which plays a role in cellular stress resistance and mitochondrial function, pointing to potential vulnerabilities in cellular defenses. Cardiovascular-related TFs such as KLF2 and KLF3, along with SMAD7, involved in TGF-beta signaling, were reduced, possibly impacting structural and signaling integrity. Additionally, MXD1 and IKZF1, crucial for cellular differentiation and immune development, were also diminished (Figure 4E). These transcriptional changes underscore a complex network of regulatory adjustments that could contribute to the altered EP and structural properties in TTN-T32756I-iPSC-aCMs.

Using the KEGG and GO pathways, we performed Ingenuity Pathway Analysis (IPA) to identify novel titin-interacting proteins that could account for the increased I_ks activity in TTN-T32756I-iPSC-aCMs. The IPA suggests FHL2, which is known to transduce mechanical signaling through its interactions with titin and is enriched in the mutant iPSC-aCMs (Figure 4F). As FHL2 has also been shown to interact with the I_ks-binding partner mink (KCNE1) directly and increases its current density (Kupershmidt et al., 2002), we postulate that elevated FHL2 levels in TTN-T32756I-iPSC-aCMs may enhance binding to I_ks and thus increase its activity.

To determine if TTN-T32756I increases I_ks by modulating the interaction between KCNQ1-KCNE1 and FHL2, we performed co-immunoprecipitation studies and confocal microscopy in both WT and TTN-T32756I-iPSC-aCMs. The co-localization between KCNE1 and FHL2 increased ~threefold in TTN-T32756I-iPSC-aCMs, suggesting an increased interaction between them (Figure 5A, Figure 5—figure supplement 1). Since FHL2 enhances I_ks activity, we investigated whether inhibiting FHL2 could reverse the increased I_ks observed in the TTN-T32756I-iPSC-aCMs. To inhibit FHL2, we employed small interfering (si) RNA to suppress FHL2 expression and measured I_ks activity in both WT and TTN-T32756I-iPSC-aCMs. Both the WT and mutant iPSC-aCMs transfected with FHL2-specific siRNA showed a substantial decrease in FHL2 expression (Figure 5B). As shown in Figure 5C–D, voltage-clamp recordings demonstrate that the I_ks in FHL2-suppressed TTN-T32756I-iPSC-aCMs are significantly reduced compared to the corresponding I_ks in TTN-T32756I-iPSC-aCMs but comparable to the WT (Figure 5D). Therefore, inhibition of FHL2 by the siRNA rescues the increased I_ks of the TTN-T32756I-iPSC-aCMs, which is similar to that of WT I_ks. Overall, our data suggest that the TTNmv creates an EP substrate for AF by modulating I_Ks activity, in part by increasing the interaction between the KCNQ1-KCNE1 complex and FHL2 (Figure 5E).

The UIC Multi-Ethnic AF Biorepository was established in 2015 to explore the genetic basis of AF across racial and ethnic groups. This study was approved by the UIC Institutional Review Board (#2015–0681). Subjects are prospectively enrolled from outpatient and inpatient sites within the University of Illinois Health (UIH) healthcare system. Subjects must have a documented history of AF by electrocardiogram, Holter/event monitor, or implantable cardiac device, with confirmation by an attending cardiologist. At the time of enrollment, blood is drawn for DNA extraction and genetic analysis. Baseline demographics are obtained from provider notes in the electronic health record (EHR) and, as necessary, confirmed with the patient at the time of enrollment. Longitudinal clinical outcomes are followed through serial review of the EHR. Written informed consent was obtained from all participants. Adults 18 years of age or older at the time of AF diagnosis were prospectively enrolled between August 25, 2015, and May 19, 2019. Samples from 161 subjects who identified as non-Hispanic Black (NHB) or Hispanic/Latinx (HL) and had an echocardiogram performed within 3 months of enrollment date underwent whole exome sequencing. A total of 27 subjects with congenital or rheumatic disease, severe mitral stenosis, and end-stage renal disease on dialysis were excluded. Three subjects who carried a predicted loss-of-function frameshift or stop-gain TTN variant were additionally excluded. The primary exposure was the presence of one or more missense TTN variants. The primary outcome was time from initial AF diagnosis to hospitalizations with a primary diagnosis of either AF or acute decompensated HF recorded in the clinical discharge summary, captured as repeating events. Clinical data was obtained through manual review of EHR and comorbidities were defined if present in clinical notes on or prior to AF diagnosis date. Subjects were censored on the date of death, last known clinical encounter (up to August 26, 2023), or at 10 years after AF diagnosis. Echocardiographic criteria for left ventricular dilatation were defined as left ventricular end diastolic diameter (LVEDD) >2 standard deviations (SD) above sex-specific mean according to standard guidelines (Lang et al., 2015). Subjects meeting echocardiographic criteria for left ventricular dilatation fulfilled clinical criteria for nonischemic dilated cardiomyopathy if left ventricular ejection fraction (LVEF) <50% and ischemic cardiomyopathy was ruled out by coronary angiogram.

Samples were sequenced with support of the National Human Genome Research Institute (NHGRI) Centers for Common Disease Genetics (CCDG) program. Sequencing was performed at the Broad Institute following methods established by the National Heart, Lung, and Blood Institute (NHLBI) Trans-Omics for Precision Medicine (TOPMed) Atrial Fibrillation Study (available under TOPMed Whole Genome Sequencing Methods: Freeze 9 here). Samples underwent WES using an Illumina HiSeq X system with 150 bp paired-end read length, at 20 x depth for at least 85% of targets. Alignment to human genome reference GRCh38/hg38 was performed using BWA-MEM (Burrows-Wheeler Aligner, v0.7.15.r1140). TTN variant calls were filtered to select those with read depth of ≥20 X, genotype quality ≥20, minor allele frequency (MAF) ≤1% across gnomAD subpopulations, and excluding intronic, regulatory, or UTR 3’/5’ variants. Synonymous variants, homozygous reference calls, non-canonical variants, and any variants with percent spliced in (PSI) index <90 were excluded. Missense variants predicted to be deleterious were identified using REVEL score of ≥0.7, a previously proposed cutoff to predict deleterious effect in dilated cardiomyopathy (Ioannidis et al., 2016; Morales et al., 2020). Because of the uncertain role of missense TTN variants across AF and other cardiac diseases, we examined all TTN missense variants regardless of REVEL score. Variant annotation data were obtained from Ensembl Variant Effect Predictor (McLaren et al., 2016), downloaded January 11, 2024, and CardioDB (https://cardiodb.org/titin) for TTN region and percent spliced in (PSI) (Roberts et al., 2015), downloaded November 16, 2023. The TTN-T32756I variant (REVEL Score: 0.58758; Supplementary file 1) was prioritized because it occurred in multiple unrelated individuals within our clinical AF cohort, despite no reported family history of AF in affected individuals. While no parental inheritance was observed, the possibility of de novo origin cannot be excluded. Furthermore, this variant is located within a region overlapping a deletion mutation recently shown to cause AF in a zebrafish model, supporting its potential pathogenicity (Jiang et al., 2024). Notably, the affected individuals did not carry additional loss-of-function TTN variants.

Human iPSC-CMs were derived from reprogrammed peripheral blood mononuclear cells (PBMCs) as previously described (Hong et al., 2021). Mycoplasma testing revealed no contamination. 80–90% confluent iPSC-CMs were differentiated using the Cardiomyocyte Differentiation Kit (Gibco) and guided toward the atrial subtype using all-trans RA (Argenziano et al., 2018). The cellular population was purified through glucose starvation and lactate replacement, resulting in contracting monolayers of iPSC-aCMs. Our protocol typically yields ~80 to 90% pure iPSC-aCMs and <6% fibroblasts based on immunostaining analysis as we have previously described (Hong et al., 2021; Argenziano et al., 2018). Human iPSC-aCMs were then matured following dissociation and replating on fibronectin-coated plates and maintained in Cardiomyocyte Maintenance Media supplemented with T3, insulin-like growth factor-1, and dexamethasone as previously described (Ly et al., 2022).

Due to the patients' unavailability of peripheral blood mononuclear cells (PBMCs), we utilized a healthy iPSC line and introduced the TTN-T32756I variant using CRISPR/Cas9 genome editing. This approach ensures an isogenic background, thereby minimizing genetic variability and providing a controlled system to study the direct effects of the mutation. We edited the genome of WT iPSCs to introduce the T32756I using the CRISPR-Cas9 technique (Hong et al., 2021). The TTN-T32756I CRISPR-Cas9 was designed from WT allele sgRNA and two ssODN that target the TTN genomic locus: 31111–31119 and genomic DNA location: chr2:178539785–178539811 (GRCh38.p13). An amplification of the exon 352 cDNA donor RNP was constructed and electroporated into TTN gene WT hiPSCs. Gene-editing efficiency was confirmed with next-generation sequencing (NGS). Karyotype analysis was carried out by the Cytogenetics laboratory at WiCell Research Institute Inc Cells were collected and chromosomes were evaluated using the Giemsa trypsin Wright (GTW) banding method. Metaphase cells were analyzed, all of which were concluded to have a normal karyotype (46, XY).

Contractility of hiPSC-aCMs was performed with a versatile open-source software, MUSCLEMOTION (Sala et al., 2018). MUSCLEMOTION is an ImageJ plugin that utilizes a video-based system to assess contractile functions. Optical voltage mapping recordings were performed on the IonOptix system myopacer system using the fluovolt membrane potential kit (Thermo Fisher). HiPSC-aCMs were cultured in confocal dishes and were incubated with Tyrode’s solution (140 mM NaCl, 4.56 mM KCl, 0.73 mM MgCl₂, 10 mM HEPES, 5.0 mM dextrose, 1.25 mM CaCl₂) plus 1 x Fluovolt (Sigma/Aldrich) for 15–20 min. Cells were then rinsed with normal Tyrode’s solution and recorded APDs on the IonOptix system. For the pacing experiment, iPSC-aCMs were electrically stimulated at a frequency of 0.5 Hz with platinum wires separated 2–4 mm, and stimulation voltage was 25% above threshold (10 V, 4 ms). These experiments were performed at room temperature (25 °C).

Whole-cell patch clamping on hiPSC-aCMs I_K recordings was performed according to previously published protocols (Ly et al., 2022; Hong et al., 2021). Briefly, voltage clamps were achieved by using an Axopatch 200B amplifier controlled by pClamp10 software through an Axon Digidata 1440 A. External solution for I_K recordings contained: 140 mM NaCl, 4 mM KCl, 1.8 mM CaCl2, 1.2 mM MgCl2, 10 mM glucose, 10 mM HEPES, and 0.01 mM nifedipine, adjusted to pH 7.4 with NaOH. I_Ks recordings were isolated as 1 μM HMR-1556–sensitive current. The intracellular solution contained 100 mM potassium aspartate, 2 mM MgCl2, 20 mM KCl, 5 mM Mg-ATP, 5 mM EGTA, and 10 mM HEPES adjusted to 7.2 with KOH. I_Ks currents were elicited by using 3 s voltage-clamp steps to test potentials of −60 to +60 mV from holding potential of –40 mV and with 20 mV increments.

Calcium transients were measured using Fluo-4-AM (Invitrogen) dye dissolved in 2.5% Pluronic F-127 (MilliporeSigma). To get a working concentration of 5 μM, the dye solution was added to Tyrode’s solution (1 mM Ca²⁺). The cells were treated with Fluo-4-AM in 1 mM Ca²⁺ Tyrode’s solution, and then allowed to sit at room temperature for 20 min in the dark. Tyrode’s solution without indicators and containing 2 mM calcium was used to wash the cells. Using a 40× objective and the Zeiss LSM 710 confocal equipped with a BiG module, line scans were acquired and examined using ImageJ. The corrected minimum and maximum fluorescence values were determined by normalizing the fluorescence using a baseline background region that was unique to each cell. Fluorescence signals were normalized to basal cell fluorescence after fluo-4 loading (F0). For the pacing experiment, iPSC-aCMs were incubated at room temperature with 1 μM Fura 2-AM for 10 min and then washed out with normal tyrode 1.8 mM Ca for 20 min. The cells were then electrically stimulated at a frequency of 0.5 Hz with platinum wires separated 2–4 mm; stimulation voltage was 25% above threshold (10 V, 4 ms). Estimations of intracellular Ca²⁺ are reported as changes in ΔF/F0, where ΔF=F-F0. Calcium transients’ amplitude 340/380 nm ratio, and kinetics were analyzed using IonOptix software.

iPSC-aCMs were fixed with 2.5% glutaraldehyde in 0.1 M Sorenson’s Buffer for 60 min at room temperature. The cells are then carefully collected into a microcentrifuge tube pre-filled with the same fixation buffer. The samples were then centrifuged at 2500 x g for 10 min at room temperature. The pellet was then removed and inverted with a hypodermic needle to ensure fixative solution thoroughly permeates the sample. The pellet is then allowed to incubate at room temperature to fix for an additional 60 min at room temperature. The fixation solution was then removed and substituted with 1% glutaraldehyde +4% paraformaldehyde in 0.1 M Sorenson’s buffer and stored in 4 °C. Fixed samples were embedded in resin and microtome sections were imaged on a JEOL JEM-1400 Flash TEM.

Cells were cultured on Matek(R) glass bottom dishes that were coated with vitronectin. The cells were then allowed to grow for a period of 2 days. The cells were then fixed with 4% paraformaldehyde (PFA) for 10 min and permeabilized with phosphate-buffered saline (PBS) containing 0.1% Triton X-100 for 10 min. After that, the cells were blocked with 3% bovine serum albumin (BSA) in PBS for 1 hr at room temperature. The staining procedure was carried out overnight at 4 °C using primary antibodies that were diluted in 3% BSA in PBS. The cells were rinsed three times with PBS and then exposed to secondary antibodies (anti-mouse-FITC or anti-rabbit-PE, Santacruz, 1:1000) for 1 hr at room temperature. Subsequently, the cells were rinsed 3 x with PBS and co-stained with 1 µg/ml of DAPI. The antibodies Anti-SOX2 (Abcam, 1:200) and anti-OCT4 (Santacruz, 1:200) were used to assess the generation of iPSCs, while anti-cTnT (Invitrogen, 1:300) and anti-Kv1.5 (Alomone labs, 1:200) were utilized to confirm the differentiation of atrial cardiomyocytes. The samples were visualized using an LSM710 Meta Confocal Microscope manufactured by Zeiss. For the co-localization experiment, anti-FHL2 antibody (Abcam, #ab202584) and anti-KCNE1 antibody (Alomone Labs, #APC-163) were used and the subsequent captured images were analyzed using ImageJ software with the Coloc2 plugin. Pearson’s correlation coefficient was calculated to quantify the degree of co-localization between FHL2 and KCNE1 signals.

We performed Western blots as previously described (Ly et al., 2022; Hong et al., 2021). For western blots, cells on six-well plates were washed with ice-cold DPBS without Ca²⁺ and Mg²⁺, after which 250 µL of 1 X RIPA with protease and phosphatase inhibitors was added per well. Lysate concentrations were measured using BCA Assay and diluted with 4 X Laemmli buffer with 10% 2-Mercaptoethanol. Per sample, 25 µg of protein was then run on an SDS-polyacrylamide gel to separate, and resolved gels were electro-transferred on 0.2 µm PVDF membranes. Membranes were blocked with 5% BSA for 1 hr and then probed with corresponding antibodies of target proteins (anti-FHL2 antibody, Abcam#ab202584). The blots were developed using either anti-rabbit HRP or anti-mouse HRP and scanned on C280 imaging systems (Azure Biosystems). Protein signal densities were determined using ImageJ and normalized to corresponding β-actin signal densities.

For co-immunoprecipitation experiments, cells are rinsed with ice-cold DPBS without Ca²⁺ and Mg²⁺, after which 500 µL of 1 X RIPA with protease and phosphatase inhibitors was added to each well of a six-well plate. Lysates are sonicated on ice 3x5 s, then centrifuged for 10 min at 14,000 x g at 4 °C. 100 µL of Protein A/G agarose bead slurry was added to the lysate and incubated at 4 °C on a rotator for 60 min. The cell lysate with bead slurry was centrifuged for 10 min at 1000 x g at 4 °C and the supernatant was transferred to a fresh Eppendorf tube. 2 µg of primary antibody (Anti-KCNE1 [Alomone# APC-163]: 2.5 µL) was added to 500 µg of precleared cell lysate, and the mixture was incubated with gentle rotation overnight at 4 °C for a day. 40 µL of Protein A/G bead slurry was washed 3 x with RIPA, and the precleared cell lysate with the primary antibody was added to the bead slurry, and then followed by incubation for 3 hr at 4 °C. The mixture is centrifuged for 10 min at 1000 x g at 4 °C, and the flow through was saved to verify the immunoprecipitation. The bead pellet was rinsed with 5x500 µL RIPA and resuspended in 30 µL 4 X Laemmli sample buffer, vortexed, then centrifuged for 30 s. The sample was then boiled for 5 min, then centrifuged at 14,000 x g for 5 min, representing the IP sample. 15 µL of supernatant and immunoprecipitate samples were loaded onto a 10% SDS-PAGE gel and analyzed by western blotting.

The RNA was isolated using the Maxwell RSC simplyRNA Cells Kit (Promega AS1390) according to the instructions provided by the manufacturer. Library preparation was carried out using the Universal Plus mRNA-Seq kit (NuGen 0520-A01). Briefly, RNA underwent poly-A selection, enzymatic fragmentation, and generation of double-stranded cDNA using a mixture of Oligo (dT) and random priming. The cDNA was subjected to end repair, ligation of dual-index adaptors, strand selection, and 15 cycles of PCR amplification. The concentrations of the purified library were determined using the Qubit 1 X dsDNA HS Assay Kit (Invitrogen Q33231). The libraries were then combined in equal amounts, considering the Qubit concentration and the average size determined by TapeStation. The pooled libraries were subsequently run on the MiniSeq instrument for index balancing. The ultimate, refined pool was measured using quantitative polymerase chain reaction (qPCR). The raw reads were aligned to the reference genome hg38 using the STAR alignment tool. The quantification of ENSEMBL gene expression was performed using FeatureCounts. The edgeR package was used to calculate normalized and differential expression statistics. To account for multiple testing, p-values were adjusted using the false discovery rate (FDR) correction method. We detected differentially expressed genes (DEGs) with a q value of less than 0.05 and a log 2 fold change of at least 2 between the Control and AF iPSC-aCMs. We conducted unsupervised hierarchical clustering of all genes that showed differential expression using the Euclidean distance and complete linkage method. Additionally, we created volcano plots using the R programming language. The up- and down-regulated genes were analyzed individually using the DAVID functional annotation tool against the Gene Ontology Biological Process (GO BP) database.

Total RNA was isolated from hiPSC-aCMs using TRIzol reagent (Invitrogen), following the manufacturer’s instructions to ensure the extraction of high-quality RNA. The concentration and purity of the isolated RNA were carefully assessed using a NanoDrop 2000 spectrophotometer (Thermo Fisher Scientific), with 1 µg of total RNA utilized for each reverse transcription reaction. Reverse transcription to synthesize cDNA was conducted using SuperScript III Reverse Transcriptase (Thermo Fisher Scientific). For the analysis, specific assays and primers were selected for target gene FHL2 (Forward sequence: GTGGTGTGCTTTGAGACCCTGT, Reverse sequence: GAGCAGTGGAAACAGGCTTCATG) with glyceraldehyde 3-phosphate dehydrogenase (GAPDH) serving as the normalization reference gene. qPCR reactions were performed on an ABI QuantStudio 5 system (Applied Biosystems), using SYBR Green PCR Master Mix to accurately detect and quantify PCR amplification products. Relative expression levels of the target genes were calculated employing the ΔΔCt method, by the quantification of gene expression changes in the experimental samples relative to control. For the ΔCT Calculation, the cycling time (CT) value of the target gene was subtracted from the CT value of GAPDH in the same sample using ΔCT=CT_{target gene} − CT_{reference gene}. The ΔΔCT value was then calculated using ΔΔCT = ΔCT_Experimental – ΔCT_Control. The relative expression for the gene was in turn calculated using Relative gene expression = 2^−ΔΔCT.

FHL2-specific siRNAs (#SR301594) or scrambled siRNAs were applied to mature hiPSC-aCMs using Lipofectamine RNAiMAX (Invitrogen, Carlsbad, CA). Opti-MEM medium was used to dilute stock solutions of lipofectamine and siRNA, each prepared at a concentration of 10 µM. The siRNA-lipid complex was then made by mixing these solutions in a 1:1 ratio and incubated for five minutes. After adding this complex to the cells dropwise, the media was changed, and the cells were allowed to incubate for a further two days.

For clinical data, categorical variables are represented as count and percentage (%) and tested by Fisher’s exact test. Continuous variables are reported as mean (standard deviation [SD]) or median and interquartile range (IQR) where specified and tested with Mann-Whitney U and Kruskal-Wallis tests. Ordinal variables were tested with the Kruskal-Wallis test. Univariable and multivariable Cox proportional hazards models were used to evaluate risk of the primary outcome as recurring events, related to presence of TTN missense variants. A partially adjusted model was first tested using covariates of age and sex, followed by a fully adjusted model which additionally incorporated covariates of ethnicity and baseline LVEF <50%. Interaction between TTN missense variants and LVEF was additionally evaluated. To identify whether associations were limited to variants with predicted deleterious effect, analysis was repeated stratifying subjects according to presence and predicted impact of TTNmv: presence of one or more predicted deleterious (REVEL ≥0.7) variants, presence of predicted benign (REVEL <0.7) variants only, or no TTNmv present. Additional sensitivity analyses were performed excluding subjects with multiple TTNmvs and with nonischemic dilated cardiomyopathy. Analysis was performed in R (version 4.2.1, packages: arsenal, ggplot2, ggsurvfit, gtsummary, reda, survival). Experiments were performed at least two times (biological replicates) to ensure reproducibility. Unless otherwise noted, experimental data on hiPSCs are shown as mean ± SD. *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001 indicate significance, while p>0.05 is regarded as non-significant. Statistical analyses include unpaired nonparametric and two-tailed Mann-Whitney U test for data with normal distribution, and either one-way or two-way ANOVA with post hoc Bonferroni’s corrections for several groups. The first and third quartiles of a median are used to express skewed data. Fisher’s Exact test is used to compare categorical data, and unpaired Student’s t-test or ANOVA is applied to assess continuous variables.