PITX2 deficiency leads to atrial mitochondrial dysfunction Authors: Jasmeet S Reyat, Laura C Sommerfeld, Molly O’Reilly, Victor Roth Cardoso, Ellen Thiemann, Abdullah O Khan, Christopher O’Shea, Sönke Harder, Christian Müller, Jonathan Barlow, Rachel J Stapley, Winnie Chua, S Nashitha Kabir, Olivia Grech, Oliver Hummel, Norbert Hübner, Stefan Kääb, Lluis Mont, Stéphane N Hatem, Joris Winters, Stef Zeemering, Neil V Morgan, Julie Rayes, Katja Gehmlich, Monika Stoll, Theresa Brand, Michaela Schweizer, Angelika Piasecki, Ulrich Schotten, Georgios V Gkoutos, Kristina Lorenz, Friederike Cuello, Paulus Kirchhof, Larissa Fabritz Published: 2024 ABSTRACT AbstractAimsReduced left atrial PITX2 is associated with atrial cardiomyopathy and atrial fibrillation (AF). PITX2 is restricted to left atrial cardiomyocytes (aCMs) in the adult heart. The links between PITX2 deficiency, atrial cardiomyopathy, and AF are not fully understood.Methods and resultsTo identify mechanisms linking PITX2 deficiency to AF, we generated and characterized PITX2-deficient human aCMs derived from human induced pluripotent stem cells (hiPSC) and their controls. PITX2-deficient hiPSC-derived atrial cardiomyocytes showed shorter and disorganized sarcomeres and increased mononucleation. Electron microscopy found an increased number of smaller mitochondria compared with isogenic controls. Mitochondrial protein expression was altered in PITX2-deficient hiPSC-derived atrial cardiomyocytes. Single-nuclear RNA-sequencing found differences in cellular respiration pathways and differentially expressed mitochondrial and ion channel genes in PITX2-deficient hiPSC-derived atrial cardiomyocytes. PITX2 repression in hiPSC-derived atrial cardiomyocytes replicated dysregulation of cellular respiration. Mitochondrial respiration was shifted to increased glycolysis in PITX2-deficient hiPSC-derived atrial cardiomyocytes. PITX2-deficient human hiPSC-derived atrial cardiomyocytes showed higher spontaneous beating rates. Action potential duration was more variable with an overall prolongation of early repolarization, consistent with metabolic defects. Gene expression analyses confirmed changes in mitochondrial genes in left atria from 42 patients with AF compared with 43 patients with sinus rhythm. Dysregulation of left atrial mitochondrial (COX7C) and metabolic (FOXO1) genes was associated with PITX2 expression in human left atria.Conclusion PITX2 deficiency causes atrial mitochondrial dysfunction and a metabolic shift to glycolysis in human aCMs. PITX2-dependent metabolic changes can contribute to the structural and functional defects found in PITX2-deficient atria.Keywords: Atrial fibrillation, Mitochondrial dysfunction, Human induced pluripotent stem cells, Metabolic shift, PITX2, Human heart tissue Time of primary review: 30 days Introduction Atrial fibrillation (AF) is common and the associated cardiovascular mortality and morbidity profoundly affect patients, their families, and society.1,2 Better concepts to prevent and treat AF are needed to improve this situation.3,4 Genome-wide association studies found over 100 different common gene variants that are associated with AF.5–8 The most prominent signals are clustered in a genomic region on chromosome 4q25, close to the PITX2 gene.5–8 The gene variant-carrying locus regulates the PITX2 gene,9–11 and deletion of AF risk alleles reduces left atrial PITX2 concentrations.9,10PITX2 messenger RNA (mRNA) is confined to left atrial cardiomyocytes (aCMs) in the adult human heart12,13 and in mice.12,14 Recurrent AF after thoracoscopic AF ablation is related to reduced left aCM PITX2.13Pitx2 mRNA regulates transcription in the adult heart15 with multiple effects on cardiac function and structure: Partial deletion of Pitx2 modulates atrial electrical function,12,16–19 while complete Pitx2 deficiency alters atrial structure, calcium handling, and ion channel composition.20–22 To identify PITX2-dependent cellular processes and pathways contributing to these atrial changes in human cells and in patients with AF, we generated and characterized human induced pluripotent stem cell (hiPSC)-derived PITX2-deficient aCMs and wild-type (WT) controls. Results were validated in human left atrial tissue from patients with and without AF and compared with published findings. Methods Cell culture Based on the known effects of PITX2 deletion in murine and zebrafish models, we chose to delete the intron-exon region of exon 6 of the PITX2 gene for this study to enable observation of a clear PITX2-dependent phenotype.19–22 The human control iPSC line (F1; MPIMBMi011-A) and the otherwise isogenic, genome-edited PITX2-deficient line were donated by the group of Boris Greber and have previously been described.23 HiPSCs were maintained in Gibco StemFlex Medium (Thermo Fisher Scientific, Rugby, A3349401) on Geltrex (Thermo Fisher Scientific, A1569601)-coated plates. The differentiation of hiPSCs into atrial cardiomyocytes and ventricular cardiomyocytes (vCMs) was optimized based on a published protocol.24 Briefly, on day 0, the medium was replaced with differentiation medium [RPMI-1640 with GlutaMAXTM and HEPES (Thermo Fisher Scientific, 72400047) containing 0.5 mg/mL human recombinant albumin (Sigma-Aldrich, Gillingham, A9731), 0.2 mg/mL L-ascorbic acid 2-phosphate (Sigma-Aldrich, 49752)] supplemented with 4 µM CHIR99021 (Sigma-Aldrich, SML1046) to promote mesoderm induction. On day 2, the medium was replaced with a differentiation medium containing 5 µM IWP-2 (Sigma-Aldrich, I0536) to promote cardiac progenitor cell differentiation. After day 4, cells were maintained in the cardiomyocyte differentiation medium. To induce aCM specification, 1 µM retinoic acid (Sigma-Aldrich, R2625) was supplemented to the medium between days 3 and 6 of differentiation. On day 6, the medium was changed to cardiomyocyte maintenance medium (cardiomyocyte differentiation medium supplemented with 2% B-27TM, Thermo Fisher Scientific, 17504044), and the medium was refreshed every 48 h. Beating cardiomyocytes were observed from as early as day 8 of differentiation. At day 12, aCMs and vCMs were re-plated at a lower density by dissociating cells using StemPro Accutase Cell Dissociation Reagent (Thermo Fisher Scientific, A1110501) and cultured in cardiomyocyte plating medium [cardiomyocyte maintenance medium with the addition of 10% KnockOutTM Serum (Gibco, 10828028) and 1 µM Thiazovivin (Sigma-Aldrich, SML1045)] for 24 h before the medium was changed to cardiomyocyte selection medium [RPMI 1640 no glucose (Gibco, 11879020) supplemented with 0.5 mg/mL human recombinant albumin, 0.2 mg/mL L-ascorbic 2-phosphate and 4 mM lactate (Sigma-Aldrich, 1614308)] for an additional five days. Afterwards, aCMs and vCMs were maintained in a cardiomyocyte maintenance medium until day 30, a time-point in which hiPSC-derived aCMs and vCMs express key cardiac markers.24,25 Immunofluorescence staining HiPSC-derived atrial cardiomyocytes were fixed with 4% paraformaldehyde, blocked with 4% goat serum, and incubated with primary antibodies (see Supplementary material online, Table S1) overnight at 4°C on a rocker. Cells were subsequently washed and stained with the corresponding Alexa Fluor secondary antibody conjugates (Thermo Fisher Scientific) for 1 h at room temperature and then counterstained with DAPI (1:10,000) for 5 min and mounted using Prolong Gold Anti-fade reagent (Thermo Fisher Scientific) ready for imaging using a Zeiss LSM 880 Airyscan confocal microscope (Carl Zeiss NTS Ltd., Oberkochen). Images were analysed using Fiji software. Sarcomere structure analysis was carried out using a previously published MATLAB (MathWorks) script.26 Analysis of nuclei parameters was carried out using a previously described pipeline in Cell Profiler 4.2.1.27 Electron microscopy HiPSC-derived atrial cardiomyocytes were cultured in 3.5-cm plastic dishes for 3 days, fixed in a mixture of 4% paraformaldehyde and 1% glutaraldehyde (Science Services, ) in 0.1 M phosphate buffer overnight at 4°C. Samples were rinsed three times in 0.1 M sodium cacodylate buffer (pH 7.2–7.4), scraped off the cell culture dish, and osmicated using 1% osmium tetroxide in cacodylate buffer. Following osmication, the samples were dehydrated using ascending ethanol concentrations, followed by two rinses in propylene oxide. Infiltration of the embedding medium was performed by immersion in a 1:1 mixture of propylene oxide and Epon (Science Services, Germany), followed by neat Epon and hardening at 60°C for 48 h. For electron microscopy, ultra-thin sections (60 nm) were cut and mounted on copper grids and stained using uranyl acetate and lead citrate. The sections were analysed with a JEM- 2100Plus Transmission Electron Microscope at 200 kV (Jeol, Germany). Images were acquired with the XAROSA CMOS camera (Emsis, Germany). Flow cytometry HiPSC-derived atrial cardiomyocytes were processed using the FoxP3/Transcription Factor Staining Buffer kit (eBiosciencesTM, 00–5523–00) according to manufacturer’s instructions before being incubated with primary antibodies (see Supplementary material online, Table S1) overnight at 4°C on a rocker. Subsequently, samples were induced with corresponding Alexa Flour secondary antibody conjugates (Thermo Fisher Scientific) for 30 min at 4°C. For experiments looking at cell proliferation, hiPSC-derived atrial cardiomyocytes were incubated with 5-ethynyl-2′-deoxyuridine (EDU) using the Click-iTTM EDU Alexa Fluor 488 Flow Cytometry Assay Kit (Thermo Fisher Scientific, C10420) according to the manufacturer’s instructions. Samples were processed using a BD LSR FortessaTM (BD Biosciences), and data were analysed using FlowJo software. RNA isolation and quantitative real-time PCR Total RNA was isolated from aCMs and vCMs using the RNeasy Mini Kit (QIAGEN, 74104) and reverse transcribed into cDNA using the High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems, 4368814) using a total of 1 µg of RNA. RNA was quantified using the QubitTM RNA high-sensitivity kit (Invitrogen, Q32852) using a Qubit Fluorometer. cDNA was diluted to a working concentration of 5 ng/µL using RNA-free water (QIAGEN, 129112). Quantitative real-time PCR (RT-qPCR) was performed using 10 ng of template cDNA and PowerUpTM SYBRTM Green Master Mix (Applied Biosystems, A25742). Samples were run on the 7500 Fast Real-Time PCR system (Thermo Fisher Scientific). Gene of interest Ct values were normalized to housekeeping gene Ct values using the ΔΔCt method.28 Sequences of primers used for RT-qPCR are provided in Supplementary material online, Table S2. Proteomics Protein quantification, quality assessment, imputation, differential expression analysis, and enrichment analyses were conducted by the UKE Bioinformatics Core, Hamburg, Germany. HiPSC-derived atrial cardiomyocytes from six independent differentiation runs were pelleted, washed with sterile PBS, and shock-frozen in liquid nitrogen. Samples were prepared using established proteomic techniques (for details see Supplementary materials and methods). Extracellular flux analysis Mitochondrial oxidative phosphorylation and glycolytic flux were measured with a Seahorse XF-96 Analyser (Agilent). HiPSC-derived atrial cardiomyocytes were plated into XF-96 well (Agilent, 103794-100) Geltrex-coated plates at a cell density of 50 000 cells per well. Measurements were made in XF RPMI Medium pH 7.4 supplemented with 10 mM glucose, 1 mM HEPES, 2 mM L-Glutamine, and 1 mM sodium pyruvate. Mitochondrial oxidative phosphorylation and glycolytic proton efflux were assessed using the following parameters: oligomycin (2 µg/mL), BAM15 (3 µM), rotenone and antimycin A (2 µM), and desoxyglucose (2-DG; 50 mM). ATP supply fluxes and corrections of mitochondrial acidification were calculated as previously described.29–31 Analysis of mitochondrial membrane potential The mitochondrial membrane potential was analysed using the mitochondrial-selective dye tetramethylrhodamine methyl ester (TMRM; 2.5 nM). To normalize mitochondrial content, hiPSC-derived atrial cardiomyocytes were stained with MitoTrackerGreen (200 nM, 1 h). HiPSC-derived atrial cardiomyocytes were plated on gelatine-coated glass coverslips and cultured for 6–7 days at 5% CO2 and 37°C. Measurements were performed on a Leica TCS SP5 confocal microscope at basal conditions or in response to oligomycin A treatment (2 µM). TMRM was excited at 561 nm and emission assessed between 580 and 700 nm. MitoTrackerGreen was excited at 488 nm and emission assessed between 500 and 530 nm. The images were processed using LAS × software (version 3.5.6.21594). Mean intensity values of TMRM fluorescence (corrected for background) were normalized to the mean intensity value of MitoTrackerGreen fluorescence (corrected for background) per image to correct for mitochondrial content. n = 3 independent hiPSC-derived atrial cardiomyocyte differentiation runs and 20 images per condition per hiPSC-derived atrial cardiomyocyte batch were analysed. Data were normalized to mean values of WT hiPSC-derived atrial cardiomyocytes at basal conditions. Western blotting Protein isolation and western blotting were carried out as previously described.13 Briefly, proteins were isolated from hiPSC-derived atrial cardiomyocytes using 1% Triton X-100 (Sigma-Aldrich, T8787) and protease and phosphatase inhibitors (Thermo Fisher, 78440) and subsequently quantified using the DC Protein Assay kit (Bio-Rad, 500-01112). SDS-polyacrylamide gel electrophoresis and western blot analysis were performed using NovexTM WedgeWellTM 4–20% Tris-Glycine gels (Thermo Fisher, XP04205). Membranes were blocked in Intercept® (Tris-buffered Saline) blocking buffer (LI-COR, 927-60001) and incubated at 4°C overnight on an orbital shaker. On the next day, membranes were incubated overnight at 4°C with primary antibodies (see Supplementary material online, Table S1). Membranes were then washed and incubated with mouse and rabbit fluorescently conjugated secondary antibodies (LI-COR) for 2 h at room temperature before visualization on the LI-COR Fc Dual-Mode Imaging System. Quantification of western blots was carried out using Image Studio Lite software (LI-COR) with quantification normalized to GAPDH expression. Single-nuclei RNA-sequencing of WT and PITX2−/− hiPSC-derived atrial cardiomyocytes and analysis In order to assess changes in gene expression resulting from suppression of the PITX2 gene at the single-cell level, we applied single-nuclei RNA-sequencing (snRNAseq) to hiPSC-derived atrial cardiomyocytes. Nuclei from hiPSC-derived atrial cardiomyocytes were isolated and processed for snRNAseq as described.32 We compared two replicates of the PITX2−/− hiPSC-derived atrial cardiomyocyte cell line with three replicates of the WT hiPSC-derived atrial cardiomyocytes as controls. Data were mapped to the human genome (GRCh38) using 10X cellranger version 6.1.2 (www.10xgenomics.com), processed to remove doublets and identify nuclei that met high-quality standards, and harmonized to remove batch effects.33 Manifolds were constructed using Uniform Manifold Approximation and Projections (UMAPs) for all individual nuclei of knock-out and control samples (https://arxiv.org/pdf/1802.03426.pdf). Populations were defined by assignment of nuclei to individual clusters based on Leiden-annotation with a resolution of 0.5.34 To perform differential gene expression analyses (DGE) between PITX2−/− hiPSC-derived atrial cardiomyocytes and WT groups we created aggregated pseudo-bulk samples from our single-nuclei dataset (one pseudo-bulk sample per each cluster/Leiden annotated35). To be considered, one sample should have at least five nuclei per cluster. To test for DGE we used edgeR implemented in R. Before fitting our quasi-likelihood negative binomial generalized log-linear model, we filtered for genes that have sufficient counts (at least 10) and that were expressed in at least 50% of the samples (min.prop = 0.5) to be considered in statistical analysis. We used the empirical Bayes quasi-likelihood test (glmQLFtest) to perform gene-wise tests across contrasts. Pathway enrichment analysis of RNASeq data from aCM was performed using Bioconductor packages in R (Version 4.3.3) and RStudio (Version 2023.12.1). To compare gene expression changes in response to PITX2 repression, a recently published data set of hiPSC-derived ventricular-like cardiomyocytes exposed to PITX2-repressing siRNA or to scrambled control RNA was accessed.36 Pathway analysis using gene ontology and expression of metabolic differentially expressed genes of interest was performed. Kyoto Encyclopedia of Genes and Genomes pathways and GO terms of differentially expressed genes were determined by Benjamini–Hochberg tests with a P-value threshold of 0.05. Whole-cell patch-clamp electrophysiology HiPSC-derived atrial cardiomyocytes were plated at a density of 25 000–35 000 cells on Geltrex-coated 15 mm round glass coverslips to obtain single-cell distribution. The cells were maintained in culture for a minimum of seven days until experiments were carried out. Action potential (AP) recordings were made using the whole-cell patch-clamp technique on an Axopatch 200B amplifier (Molecular Devices), recorded in the current-clamp configuration. Briefly, cells were superfused at 3 mL/min, 36–37°C, with a solution containing in mM: 145 NaCl, 5.4 KCl, 5 HEPES, 1.8 CaCl2, 1.2 MgCl2, 0.33 NaH2PO4, 0.83 MgSO4.7H20, and 11 glucose, pH 7.4 with NaOH. The internal pipette solution contained in mM: 130K-glutamate, 10 KCl, 10 NaCl, 0.5 MgCl2, 10 HEPES, and 5 MgATP, pH 7.2 with KOH (all reagents from Sigma-Aldrich). Pipette resistance ranged between 1.5 and 3 MΩ. Spontaneously occurring APs were recorded for 60 s before APs were triggered by 1 ms current injections (1 nA). AP trains were stimulated at 1, 2, or 3 Hz for 60 s to allow rate adaptation and digitized at 50 kHz using CED micro1401 driven by Signal v6 (Cambridge Electronic Design). Only spontaneously beating cells were used for experiments. APs were analysed for diastolic membrane potential and AP duration (APD 30, APD50, APD70, and APD90) using modified algorithms from ElectroMap software.37 Information on additional parameters measured can be found in the Supplementary materials online. Bulk RNA-sequencing of human left atrial appendages and analysis Bulk RNA-sequencing was performed on human left atrial appendages (see study approval) collected from patients undergoing open-heart surgery with excision of left and right atria at six centres as published.13 Sequencing was performed at the University of Münster, Germany (M Stoll). Good-quality samples were aligned to the human genome (GRCh38p12) using the HISAT2 alignment tool.38 The aligned files were sorted and indexed using samtools.39 Feature counts (transcript level) were computed using the htseq tool.38 Htseq readcounts were normalized using DESeq2. Data was transformed using regularized log transformation with DEseq2 prior to visualization. DGE analysis was performed by modelling the Benjamini–Hochberg false discovery rate (FDR). Differentially expressed genes were defined as FDR <0.05 and Log2(fold change) >0.1. For heatmaps, data were visualized as log-normalized counts from DESeq2. GO Pathway Enrichment analysis was carried out using Gene Ontology. Statistical analysis Data were analysed using PRISM (GraphPad Software Inc., version 6), and results are presented as mean ± SD unless otherwise stated. All experiments were repeated a minimum of three times using different batches of differentiated cells accounting for biological replicates, which are specified in the figure legends. The number of samples (n) and the statistical test used for each analysis are indicated in the figure legends. Where possible, experimenters were blinded to the genotypes of tissue samples/cells. P-values are stated in the figures. Study approval Biopsies from left atrial appendages were sampled during open-heart surgery from six separate cohort studies run at the universities of Barcelona, Birmingham, Maastricht, Muenster, Munich, and Paris (all part of the CATCH ME consortium), and immediately frozen in liquid nitrogen to prevent RNA degradation. All study participants provided written informed consent. The investigation complied with the principles that govern the use of human tissues outlined in the Declaration of Helsinki. The Medical Ethics Committee of each participating centre approved the study and its protocols. Overall governance was provided by Maastricht University. Details of the clinical characteristics of the patients have been published.40 This analysis compared patients who were in sinus rhythm (SR) at the time of surgery and who did not have a history of AF prior to surgery (SR) with patients who had established, permanent AF. Results Generation and differentiation of PITX2-deficient hiPSC-derived atrial cardiomyocytes PITX2-deficient hiPSC line and the respective control (WT) cells showed a normal karyotype and pluripotency status (see Supplementary material online, FigureS1A–C). PITX2-deficient hiPSCs and WT hiPSCs were successfully differentiated into aCMs (Figure 1A) with a high yield (see Supplementary material online, FigureS1D). Time course analysis revealed robust induction of PITX2 expression, peaking between days 8 and 12 of differentiation in WT cells (Figure 1B). This early peak reflects the known role of PITX2 in right-left patterning during early mesodermal development.41–43BMP10 expression was detected from 30 days of differentiation, reflecting differentiation of the cells into hiPSC-derived atrial cardiomyocytes (Figure 1C).24,25 As intended, PITX2 was reduced in PITX2−/− hiPSC-derived atrial cardiomyocytes (Figure 1D). As expected, WT hiPSC-derived vCMs showed no PITX2 expression (Figure 1D). Subsequent analysis of cardiomyocyte developmental transcriptional factors revealed a reduction in MYCOD and an increase in TBX5 expression in the PITX2−/− hiPSC-derived atrial cardiomyocytes (see Supplementary material online, FigureS2A). aCM markers BMP10, KCNJ3, NR2F1, and NR2F2 expression was reduced as expected (see Supplementary material online, FigureS2B). Ventricular-specific genes were largely undetectable in WT and PITX2−/− hiPSC-derived atrial cardiomyocytes (see Supplementary material online, FigureS2C). Characterization of WT and PITX2−/− hiPSC-derived atrial cardiomyocytes (aCMs). (A) Schematic overview of differentiation protocol used to generate hiPSC-derived aCMs. (B) Gene expression analysis of PITX2 and (C) BMP10 over the time course of atrial cardiomyocyte differentiation using WT hiPSC-derived aCMs (WT aCMs) as assessed by RT-qPCR (n = 3). Dashed line represents the basal expression of PITX2 or BMP10 in WT hiPSCs. (D) Gene expression of PITX2 in day 30 aCMs from WT and PITX2−/ (PITX2−/− aCMs) lines as assessed by RT-qPCR. Day 30 hiPSC-derived ventricular cardiomyocytes from the WT line (WT vCMs) were used as a control (n = 6). (E) Confocal microscopy of immunofluorescently labelled α-actinin in WT and PITX2−/− aCMs. Scale bar = 10 µm. (F) Sarcomere length measurements in WT and PITX2−/− aCMs (WT aCMs = 63 images; PITX2−/− aCMs = 62 images). (G) Gene expression of MYH6, ACTN2, TNNT2, TNNI1 and TNNI3 in WT and PITX2−/− aCMs as assessed by RT-qPCR (n = 6). Data are expressed as the mean relative expression and presented as box and whisker plots (min to max). Mann–Whitney U tests were used to compare gene concentrations between groups. (H) Bi-nucleated and mono-nucleated cell analysis in WT and PITX2−/− aCMs. Altered cardiomyocyte structure and nuclear morphology in PITX2-deficient hiPSC-derived atrial cardiomyocytes PITX2−/− hiPSC-derived atrial cardiomyocytes exhibited sarcomere disarray (Figure 1E) and shortened sarcomeres (Figure 1F) compared with WT controls. mRNA concentrations of the sarcomeric transcripts MYH6, TNNT2, and TNNI1 mRNA were increased in PITX2−/− hiPSC-derived atrial cardiomyocytes compared with WT controls (Figure 1G). PITX2−/− hiPSC-derived atrial cardiomyocytes displayed a greater ratio of mono-nucleated cardiomyocytes compared with multi-nucleated cardiomyocytes (Figure 1H). These nuclei were increased in number, larger, and displayed an altered shape (see Supplementary material online, Table S4). Given that mononucleation is associated with an increased proliferative capacity in cardiomyocytes,44 we next investigated the proliferation status of PITX2−/− hiPSC-derived atrial cardiomyocytes. PITX2−/− hiPSC-derived atrial cardiomyocytes displayed increased 5-ethylnyl-2′-deoxyuridine (EdU) incorporation compared with WT hiPSC-derived atrial cardiomyocytes (see Supplementary material online, FigureS2D) and showed a proliferative gene signature with increased expression of CCNA1 and CCNB1 (see Supplementary material online, FigureS2E) and a reduction in the cellular quiescence genes TP53, CDKN1a, CDKN2a and HES1 (see Supplementary material online, FigureS2F), confirming increased proliferation. Proteomic analysis identifies altered mitochondrial and metabolic pathways in PITX2-deficient hiPSC-derived atrial cardiomyocytes Principal component analysis of the proteomic data revealed close clustering of the PITX2−/− hiPSC-derived atrial cardiomyocytes (Figure 2A). In total, 150 out of 3128 proteins were differentially expressed between genotypes (Figure 2B). Gene-Set Enrichment Analysis identified differentially expressed mitochondrial proteins (Figure 2C) and up-regulated Normalized Enrichment Scores in PITX2−/− hiPSC-derived atrial cardiomyocytes for processes affecting mitochondria, the generation of metabolites, energy allocation and mitochondrial translation and organization, specifically of the cristae and enhanced collagen biosynthesis (Figure 2D). Endoplasmic reticulum and ribosome organization, translation, and extracellular matrix organization were down-regulated. These data identify a link between PITX2 deficiency and expression of proteins relevant to mitochondrial and metabolic function in hiPSC-derived atrial cardiomyocytes. Targeted comparisons of key proteins relevant for mitochondrial fission and fusion (Figure 2E) and of mitophagy and biogenesis (Figure 2F) were differentially expressed in PITX2−/− hiPSC-derived atrial cardiomyocytes. Proteomic analysis of PITX2−/−­ hiPSC-derived atrial cardiomyocytes (aCMs). (A) Principal component analysis (PCA) of samples used in proteomic analysis. (B) Volcano plot showing protein enriched in WT aCMs vs. PITX2−/− aCMs. Significantly enriched proteins (log2FC > 1) are shown in black. (C) Differentially expressed mitochondrial proteins in WT aCMs and PITX2−/− aCMs presented as a heatmap. (D) Gene-set enrichment analysis of enriched and down-regulated pathways in WT aCMs and PITX2−/− aCMs. Proteins with an FDR < 0.05 and an absolute log2-fold-change > 1 were considered significantly changed. Further information on data analysis can be found in the Supplementary materials. (E, F) Expression of proteins linked to mitochondrial fission and fusion (E) and related to mitochondrial biogenesis and mitophagy (F) in WT and PITX2−/− aCMs (n = 6). PITX2-dependent changes in gene expression based on single-nuclear RNA-sequencing pseudo-bulk analysis of single-nuclei RNA-sequencing data from PITX2−/− and WT hiPSC-derived atrial cardiomyocytes showed differential expression of a large number of transcripts (Volcano plot in Figure 3A). Gene ontology analysis identified respiration as one of the main affected processes (Figure 3B). Based on Leiden-annotated UMAP clustering six distinct cell populations were found, with 60.8% of all cells belonging to cluster C1 (Figure 3C). Cell clusters C1, 2, 4, and 6 consist of both PITX2−/− and WT aCM nuclei. Clusters 3 and 5, containing ∼10% of cells, consist predominantly of PITX2−/− aCM nuclei (Figure 3D). These results identify an increased heterogeneity of PITX2−/− hiPSC-derived atrial cardiomyocyte nuclei. Among the possible comparisons of cell lines and cells belonging to the different clusters, two were considered important: differences between genotypes in the largest cluster (C1) and differentially expressed genes in the PITX2−/−-enriched cluster (C3) compared with the main WT cluster (C1). Among the top differentially expressed genes between WT and PITX2−/− hiPSC-derived atrial cardiomyocytes in cell cluster C1 were the mitochondrial genes COX6a and ABCA1 and the sodium channel SCN9A. Among the top differentially expressed genes between PITX2−/− of C1 and PITX2−/− of C3 were cell-cell contact and structural proteins and transcription factors (Figure 3E). Analysis of published36 gene expression data in hiPSC-derived ventricular-like cardiomyocytes exposed to PITX2-repressing RNA or scrambled control RNA identified similar pathways regulated in response to PITX2 repression using gene ontology (Figure 3F). Transcriptional changes in PITX2-deficient hiPSC-derived atrial cardiomyocytes (aCM). (A) Volcano plot of differentially expressed genes in ‘pseudo-bulk’ mRNA sequencing analysis of nuclei from PITX2−/− and WT control hiPSC-derived atrial cardiomyocytes. (B) Gene ontology analysis of the bulk RNA-sequencing data. (C) Leiden plot of single-nuclei RNA-sequencing identifies six clusters of cells, including one cluster containing mainly PITX2−/− cells. (D) Differential gene expression patterns in the single-nuclear RNA-sequencing data sets of WT (PITX2+/+) and PITX2−/− aCM depicted by cell cluster (Leiden plot). (E) List of 56 most differentially expressed genes in PITX2−/− hiPSC-derived atrial cardiomyocytes (PK) vs. WT based on the single-nuclear RNA-sequencing analysis. (F) Gene expression differences in a published data set36 of hiPSC-derived cardiomyocytes exposed to PITX2-small interfering RNA (siRNA) or scrambled control siRNA. Left panel: Volcano plot. Right panel: Gene ontology analysis of differentially expressed genes. Changes in metabolism and mitochondrial function in PITX2-deficient hiPSC-derived atrial cardiomyocytes Electron microscopy revealed no overt morphological defects between PITX2−/− hiPSC-derived atrial cardiomyocytes and WT controls. However, mitochondria in PITX2−/− hiPSC-derived atrial cardiomyocytes were smaller and less structured: some mitochondria showed a fractured outer membrane. Mitochondria in WT cells appeared elongated with visible cristae (Figure 4A). Expression of FOXO1, PPARGC1a, and PYGM was increased in PITX2−/− hiPSC-derived atrial cardiomyocytes compared with WT controls (Figure 4B), suggesting increased glycolytic activity. Seahorse experiments confirmed increased glycolysis in PITX2−/− hiPSC-derived atrial cardiomyocytes (Figure 4C and D). PITX2−/− hiPSC-derived atrial cardiomyocytes showed decreased SLC27A6 expression (Figure 4E). Glycolytic metabolism in PITX2−/− hiPSC-derived atrial cardiomyocytes (aCMs). (A) Electron microscopy revealed no overt morphological differences between genotypes. Mitochondria appeared elongated and structured in WT aCMs, while they were smaller with in part fractured outer membranes in PITX2−/− aCMs. G: golgi; L: lipid droplet; M: mitochondria; scale bar 500 nm. (B) Gene expression of FOXO1, PFKM, PPARAGC1a, PYGM and SCL2A1 in WT and PITX2−/− aCMs (n = 6) as assessed by qRT-PCR. Data are expressed as the mean relative expression and presented as box and whisker plots (min to max). (C) Measurement of glycolysis (glycoPER) as assessed by Seahorse measurements (n = 6). Traces shown are PER corrected after subtracting non-glycolytic acidification from the rates post 2-DE addition and mitochondrial acidification contributions.29,30 For representation purposes, oligomycin A and BAM addition have been removed from the trace as these are not relevant to the glycolytic measurements reported. (D) Quantification of basal glycolysis (glycoPER) and maximal glycolysis (Max glycoPER). (E) Gene expression of ADIPOR2, CD36, LPIN1, PPARA, SLC27A1 and SLC27A6 in WT and PITX2−/− aCMs (n = 6) as assessed by qRT-PCR. Data are expressed as the mean relative values and presented as box and whisker plots (min to max). Statistical analyses were carried out using Mann–Whitney U tests to compare between two groups. The mitochondrial/nuclear DNA ratio showed no difference between WT and PITX2−/− hiPSC-derived atrial cardiomyocytes (Figure 5A). PITX2−/− hiPSC-derived atrial cardiomyocytes showed more mitochondrial membrane content by TOMM20 flow cytometry (Figure 5B). RT-qPCR of common mitochondrial genes revealed increased COX7C and reduced MCU expression in PITX2−/− hiPSC-derived atrial cardiomyocytes (Figure 5C). Functional analysis of mitochondrial respiration revealed lower basal and maximal mitochondrial respiration in PITX2−/− hiPSC-derived atrial cardiomyocytes without changes in proton leak and oligomycin-sensitive ATP generation (Figure 5D and E). These experiments also found a higher glycolytic index in PITX2−/− hiPSC-derived atrial cardiomyocytes (Figure 5F). Basal mitochondrial membrane potential was higher compared with WT control cells, suggesting that PITX2−/− hiPSC-derived atrial cardiomyocytes already exhibit a more glycolytic metabolic state under normal culture conditions. Mitochondrial membrane potential (Figure 5G) was more sensitive to Oligomycin A in WT than in PITX2−/− hiPSC-derived atrial cardiomyocytes. Representative fluorescent microscopy images for TMRM and MitoTrackerGreen for both genotypes are shown (Figure 5H). Together, these results suggest that PITX2 deficiency causes a metabolic shift to glycolysis in hiPSC-derived atrial cardiomyocytes. PITX2−/− hiPSC-derived atrial cardiomyocytes increase their number of mitochondria, likely to compensate for the less efficient energy generation. Mitochondrial respiration in PITX2−/− hiPSC-derived atrial cardiomyocytes (aCMs). (A) Mitochondrial (ND1) to nuclear DNA (B2M) ratio as assessed by RT-qPCR (n = 6). (B) Flow cytometry analysis of mitochondrial membrane content in WT and PITX2−/− aCMs using TOMM20 staining (n = 6). (C) Gene expression of COX7C, MCU, NRF1, PRDX5, SOD1 and SOD2 in WT and PITX2−/− aCMs (n = 6) as assessed by RT-qPCR. (D) Traces showing oxygen consumption rates (OCR) in WT and PITX2−/− aCMs (n = 6). (E) Quantification of OCRs shown in (D). (F) Quantification of JATP from either oxidative phosphorylation or glycolytic sources. Data are expressed as glycolytic indexes (GI) showing absolute values of ATP supply. (G) aCMs were loaded with the mitochondrial membrane-sensitive dye tetramethylrhodamine methyl ester (TMRM) and MitoTrackerGreen as a mitochondrial-selective loading control. Subsequently, aCMs were exposed to oligomycin A (2 µM for 10 min). Alterations in mitochondrial membrane potential of PITX2−/− aCMs (blue) or the isogenic control cells (wild-type WT; black) at baseline (−) or in response to oligomycin A (+) were expressed as the ratio of TMRM/MitoTrackerGreen fluorescence as fold change of the WT. The graph represents the data summarized from three independent experiments of at least 20 images per experiment from three independent aCM differentiation runs. One-way ANOVA with Sidak post-test for multiple comparisons was performed. (H) Exemplary fluorescence images used to generate the mitochondrial potential data shown in Figure 5G. Scale bar indicates 25 µm. Faster beating rates and more heterogeneous and prolonged atrial APs in PITX2-deficient hiPSC-derived atrial cardiomyocytes As expected from Pitx2-dependent suppression of pacemaker activity in the murine left atrium,45 spontaneously beating PITX2−/− hiPSC-derived atrial cardiomyocytes showed an increased beating frequency compared with WT hiPSC-derived atrial cardiomyocytes (Figure 6A). Concentrations of the sino-atrial node gene SHOX2 and the myocardial gene NKX2-5 were increased in PITX2−/− hiPSC-derived atrial cardiomyocytes (Figure 6B). To compare AP morphologies, we applied unbiased clustering to all AP waveforms recorded in WT and PITX2−/− hiPSC-derived atrial cardiomyocytes (Figure 6C). Atrial AP clustered into three distinct morphologies (clusters 1–3). PITX2−/− hiPSC-derived atrial cardiomyocytes consistently showed more AP waveforms belonging to ‘cluster 3’ APs (with prolonged APs) compared with WT hiPSC-derived atrial cardiomyocytes during pacing and spontaneous beating (Figure 6C and Supplementary material online, Table S6). The additional AP morphology is one of the reasons why, on average, PITX2−/− hiPSC-derived atrial cardiomyocytes showed prolonged AP durations (APD, Figure 6D and Supplementary material online, FigureS3A and B). AP amplitude (Figure 6E) and peak upstroke velocity (dV/dtmax) hiPSC-derived atrial cardiomyocytes (Figure 6G) were reduced in PITX2−/− hiPSC-derived atrial cardiomyocytes compared with WT. The diastolic membrane potential was variable, but not different between genotypes (Figure 6F and G). These electrophysiological changes were less pronounced at high pacing rates (2 and 3 Hz, Supplementary material online, FigureS3C–E). Exclusion of more depolarized, less normal-appearing APs prior to clustering led to almost identical results (data on file). PITX2−/− hiPSC-derived atrial cardiomyocytes showed reduced KCNA5 expression and increased KCNA4 and KCNH2 gene expression (Figure 6H). Protein concentrations of KCNA5 and hERG were reduced in PITX2−/− hiPSC-derived atrial cardiomyocytes, and Kv1.4 concentrations were increased (Figure 6I). Electrophysiological characterization of PITX2−/− hiPSC-derived atrial cardiomyocytes (aCMs). (A) Spontaneous beating rate in WT and PITX2−/− aCMs (WT n = 43, PITX2−/− n = 87). (B) Gene expression of NKX2-5, NPPA, SHOX2, and TBX3 in WT and PITX2−/− aCMs (n = 6) as assessed by RT-qPCR. (C) Combined APs from 1, 2, and 3 Hz WT and PITX2−/− aCMs following unsupervised clustering categorized into three distinct clusters. Computationally modelled APs are shown (top) with the percentage of APs representative of those traces in WT and PITX2−/− aCMs quantified (below). (D) Representative action potential (AP) traces of spontaneously beating or 1 Hz paced WT aCMs and PITX2−/− aCMs using whole-cell patch-clamp (top). Quantification of action potential duration (APD) at APD30, 50, 70 and 90 in spontaneously beating or 1 Hz paced WT and PITX2−/− aCMs (spontaneously beating WT n = 43, PITX2−/−n = 87; 1 Hz WT n = 82, PITX2−/−n = 112 over five batches of independently differentiated cells: below). (E) Action potential amplitude (APA) in 1 Hz paced WT or PITX2−/− aCMs (1 Hz—WT n = 82, PITX2−/− n = 112). Diastolic potential and peak upstroke velocity (dV/dtmax) in spontaneously beating (F) and 1 Hz paced (G) WT or PITX2−/− aCMs (spontaneously beating—WT n = 43, PITX2−/− n = 87; 1 Hz—WT n = 82, PITX2−/− n = 112). Note that only some cells showed spontaneous beating, resulting in different diastolic potential values than in paced cells. (H) Gene expression of KCNA5, KCNA4, KCNJ12 and KCNH2 in WT and PITX2−/− aCMs (n = 6) as assessed by RT-qPCR. (I) Western blot analysis of KCNA5, Kv1.4, Kir2.2 and hERG in WT and PITX2−/− aCMs (n = 4). Western blots are shown on top with quantification below. GAPDH was used as a loading control. Data are expressed as the mean relative expression and presented as box and whisker plots (min to max). For electrophysiological analysis, statistics were carried out using a repeated measures ANOVA to compare differences in electrophysiological parameters. For gene and protein analysis, Mann–Whitney U tests were used to compare between two groups. Differential expression of metabolic genes in left atrial tissue from patients with AF RNA-sequencing data in left atrial appendage tissue collected from 85 patients during open-heart surgery were compared between patients in SR without a diagnosis of AF (‘SR’) and patients with AF diagnosed prior to surgery and in AF during tissue collection (Figure 7A, clinical details in Supplementary material online, FigureS4A–B). Gene enrichment analysis identified 1150 up-regulated genes in left atrial appendage tissue from patients with AF compared with patients in SR (Figure 7A). Biological processes linked to mitochondrial organization, ion transport, and muscle contraction were up-regulated in AF patients (see Supplementary material online, FigureS4C and Tables S7 and S8). COX7A1 gene expression was up-regulated and SLC25A4 gene expression was down-regulated in atrial tissue from patients with AF compared with patients in SR (Figure 7C). Bulk RNA-sequencing of left atrial appendage tissue from AF and SR patients. Left atrial tissue was collected during open-heart surgery from patients without known AF and in sinus rhythm (SR) during the operation (‘SR’) and from patients with permanent AF including during surgery. (A) Volcano plot showing genes in patients with AF (permanent AF) vs. those in SR at the time of tissue harvest. Significantly enriched genes [false discovery rate (FDR) < 0.05] in AF patients (blue) and significantly enriched genes in SR patients (grey) are shown. (B) Differentially expressed genes in individual samples of patients in SR and AF. Selected genes represent the top 10 enriched genes in either SR patients (top) or AF patients (bottom). (C) Expression of COX7A1 and SLC25A4 in sinus rhythm (SR) and permanent AF patients’ atrial tissue (Sinus rhythm n = 42; AF n = 43). Correlation analysis of PITX2-regulated genes in patients with chronic (permanent) AF implicated in (D) metabolism (SLC27A6, FOXO1, and PYGM), (E) mitochondrial function (COX7C, MCU, and NRF1) and (F) cardiomyocyte structure (MYH6, TNNT2, and TNNI1). Data represent n = 43 with Spearman r values and corrected P-values shown on graph. A detailed analysis of genes that surround the chromosome 4q25 locus topological associating domain identified only reduced PITX2 in left atria from patients in AF when compared with patients in SR (see Supplementary material online, FigureS4D). Five up-regulated genes and 14 down-regulated genes were also found to be regulated in both the human left atrial RNAseq and the PITX2−/− hiPSC-derived atrial cardiomyocyte proteomic data sets (see Supplementary material online, FigureS5A). Integrated analysis using our proteomics data set and two published data sets of PITX2-deficient heart tissue from zebrafish and mice revealed nine common genes up-regulated and eight common genes down-regulated in PITX2−/− hiPSC-derived atrial cardiomyocytes and Pitx2−/− heart tissue (see Supplementary material online, FiguresS5B and S3C). Association of PITX2 with metabolic and ion channel genes in human left atria with AF Three genes implicated in glycolytic metabolism were associated with PITX2 expression in both AF patients and PITX2−/− hiPSC-derived atrial cardiomyocytes (SLC27A6, forkhead box protein O1 (FOXO1) and glycogen phosphorylase (PYGM) Figure 7D). Consistent with findings in PITX2−/− hiPSC-derived atrial cardiomyocytes, COX7C expression was positively associated with PITX2 expression (Figure 7E). The PITX2 correlation of MYH6 and TNNT2 was also replicated in human atrial tissue (Figure 7F). Genes implicated in cell cycling and quiescence (CCNA1, CCNB1, and HES1) showed no correlation with PITX2 in AF patients (see Supplementary material online, FigureS6A). The ion channel genes KCNA5 and KCNH2 were associated with PITX2, consistent with findings in PITX2−/− hiPSC-derived atrial cardiomyocytes (see Supplementary material online, FigureS6B). Exploratory analyses of the human LA appendage RNAseq data set and proteomic data from PITX2−/− hiPSC-derived atrial cardiomyocytes replicated differences in the expression of genes required for mitochondrial oxidative processes and ATP generation (see Supplementary material online, FigureS7). Overall, these associations support a role for PITX2-dependent regulation of oxidative phosphorylation, mitochondrial structure and function, and cardiac ion channels in patients with AF. Discussion Main findings PITX2 deficiency reduces mitochondrial respiration and induces a metabolic shift towards enhanced glycolysis in hiPSC-derived atrial cardiomyocytes. Similar results can be replicated in human left atria with AF. In addition, PITX2 deficiency affects metabolic and respiratory pathways in hiPSC-derived atrial cardiomyocytes and increases heterogeneity of nuclear RNA expression. These PITX2-dependent effects can interact and contribute to the structural and functional changes found in PITX2-deficient atria and lead to AF. Our results suggest a potential effect of metabolic interventions to prevent and treat PITX2-dependent atrial defects and AF. PITX2-dependent mitochondrial and metabolic dysfunction PITX2 deficiency led to altered protein and gene expression (Figures 2–4) that include reduced mitochondrial respiration and a metabolic shift towards increased glycolysis in atrial cardiomyocytes (Figures 4 and 5). Such defects and the resulting metabolic dysfunction can lead to fatty deposits,46,47 promote fibrosis,48 and underlie sarcomeric dysfunction (Figure 1, similar findings in21) in experimental AF,49 thereby contributing to three key features of AF. A similar FOXO-dependent metabolic switch has been described in PITX2-deficient skeletal muscle.50 Differential expression of metabolic genes was confirmed in human left atrial tissue (Figure 7). Single-cell nuclear RNA-sequencing identified an additional cell cluster in PITX2−/− CMs (Figure 3C and D) that can further add to electrical heterogeneity (Figure 6). Our findings are consistent with the role of PITX2 in the maintenance of mitochondrial structure and function and in the regulation of mitochondrial genes in the murine heart suggested by others.51,52 Mitochondrial capacity in the heart declines during ageing,53 leading to increased mitochondrial oxidative stress in cardiomyocytes.54 Subtle PITX2-dependent mitochondrial defects could aggravate ageing-induced mitochondrial dysfunction and oxidative stress53,54 and thereby promote AF. Further studies testing metabolic challenges in PITX2-deficient atrial models are warranted to unmask subtle metabolic defects and to study whether PITX2 is involved in atrial protection against hypoxia and oxidative stress.51 Our findings support the concept that metabolic support of the atria conveys at least a part of the AF-preventing effects of SGLT2 inhibitors55,56 and PARP inhibition.57 PITX2-dependent regulation of cellular function and metabolic predisposition to AF Cardiomyocyte function including ion homeostasis requires sustained and high energy production. The increased heterogeneity of atrial APs (Figure6), shorter sarcomeres (Figure 1), and contractile dysfunction49 seen in PITX2−/−hiPSC-derived atrial cardiomyocytes and in other models of PITX2 deficiency in mice20,22 can be caused by mitochondrial dysfunction altering atrial calcium handling58–61 and repolarization,62 in addition to direct, PITX2-dependent regulation of ion channel expression (see Supplementary material online, FigureS6). The pathway analyses in PITX2−/− hiPSC-derived atrial cardiomyocytes and of published data in cardiomyocytes with post-differentiation repression of PITX2 (Figure 3) show dysregulation of metabolic and mitochondrial respiration, suggesting that metabolic dysfunction is one of the main changes associated with PITX2 deficiency in cardiomyocytes. The increased functional heterogeneity in PITX2−/− hiPSC-derived atrial cardiomyocytes may also reflect the effects of an additional cell cluster found by single-nuclear sequencing (Figure 3). Future interventional studies aiming at restoring mitochondrial function can determine the role of metabolic dysfunction in these PITX2-dependent changes. Structural defects have been described in conditionally PITX2-deficient hearts before.20 This study finds structural defects in PITX2-deficient hiPSC-derived atrial cardiomyocytes kept in culture, showing a direct effect of PITX2 deficiency on structural alterations in the heart. Combined PITX2−/− hiPSC-derived atrial cardiomyocytes proteomics, single-nuclear RNA-sequencing and analysis of human atrial RNA-sequencing identified PITX2-regulated ion channel and mitochondrial genes. Changes in mitochondrial genes are consistent with recent RNAseq data sets in animal models of AF,63 and in patients with AF (ETFB gene).64 Correlation of PITX2 gene expression and metabolic gene expression in atria from patients with AF (Figure 7), dysregulation in PITX2-deficient hiPSC-derived atrial cardiomyocytes (Figure 3) and changes in cardiomyocytes exposed to PITX2 siRNA (Figure 3) support metabolic gene regulation by PITX2. Compared with the shortening of atrial APs in murine models of Pitx2 deficiency,12,18 the AP prolongation observed in this study (Figure 6) was unexpected. The increased heterogeneity of early repolarisation found here is consistent with PITX2-deficiency-dependent electrophysiological changes in another, independently generated PITX2-deficient hiPSC-derived atrial cardiomyocyte model.19 Metabolic and other PITX2-dependent effects and inter-species variability may contribute to these differences. The more subtle electrical phenotype in heterozygous Pitx2-deficient (Pitx2c+/−) mice12,16–18 is consistent with a less profound, dose-dependent defect. Key next steps to better understand the interactions between mitochondrial and metabolic state, gene expression, cardiomyocyte structure, ion channel dysregulation, and altered atrial electrophysiology are metabolic challenges and interventions aiming to restore mitochondrial function to assess the resulting phenotypic changes and a role of PITX2. Our findings suggest that therapies improving cardiomyocyte metabolism could help to prevent AF linked to PITX2. The prevention of AF by SGLT2 inhibitors55,65 is a clear clinical sign that metabolic interventions have potential for AF treatment. Strengths and limitations Strengths of the study are a human aCM model enabling the observation of structural and functional PITX2-dependent changes in atrial cardiomyocytes in the absence of arrhythmias and other cardiovascular stressors, the hypothesis-generating characterization of the hiPSC-derived atrial cardiomyocytes and the confirmation of key findings in human atria with AF. Independent validation in hiPSC-derived cardiomyocytes and in other experimental and clinical models is desirable, including in organoid models and animals with left and right atria. Our single-nuclei RNA-sequencing analysis confirms metabolic changes and finds an increased cellular heterogeneity affecting ∼10% of cells. This illustrates the multifaceted effects of suppression of PITX2 in cardiomyocytes. Future research is needed to define potential dose-dependent, milder metabolic phenotypes in other PITX2-deficient cells and animal models.12,16–18 Further research is also needed to identify the mechanisms of mitochondrial dysfunction and to identify potential therapeutic targets. Putative crosstalk between cardiomyocytes and other atrial cells requires further studies in multicellular hiPSC-derived, animal, and human models. Another limitation is the relatively high variability of electrical function in the hiPSC-derived atrial cardiomyocytes19 which reflects different cell clusters and variable maturation.66 This variability may have obviated subtle differences, e.g. in diastolic potential, between genotypes. Improved hiPSC-atrial cardiomyocyte maturation using engineered heart tissue19 and three-dimensional growth techniques67 may generate more mature cells and organoids suitable to address these questions. Finally, although RNA-sequencing of left atrial appendages enabled us to evaluate PITX2’s function in patients, these analyses were limited to bulk sequencing of atrial tissue obtained during open-heart surgery. Limited access to cardiac tissue outside of surgical procedures renders this limitation difficult to overcome. The single-nuclei sequencing removed mitochondria prior to sequencing. Genes encoded by mitochondrial DNA (13 genes) were not included in the single-nuclear sequencing analyses. In view of the large number of mitochondrial genes encoded by nuclear DNA, this is a minor limitation in our view. Translational perspective The strongest genetic predisposition for AF is located on chromosome 4q25, close to the PITX2 gene. This study in human iPS-derived atrial cardiomyocytes shows that deletion of PITX2 leads to genetic and proteomic changes resulting in metabolic and mitochondrial dysfunction in atrial cardiomyocytes. Similar PITX2-dependent changes are found in human left atria. Our results identify metabolic and mitochondrial dysfunction as a novel contributor to AF in patients with a genetic predisposition. They support the evaluation of metabolic therapies to prevent and reverse functional and structural defects related to AF and its genetic basis. Supplementary Material Acknowledgements We would like to thank Boris Greber for providing the hiPSC lines used in this study and Daniela Moralli (University of Oxford Karyoypting Core Facility) who assisted with karyotyping analysis. The Seahorse Extracellular Flux analysis was supported by the Cellular Health and Metabolism Facility in the College of Life and Environmental Sciences at the University of Birmingham. In addition, we would like to thank the Translational Research on Heart Failure and Arrhythmias Cluster, ICVS Birmingham for useful insights and thoughtful discussions on the manuscript. Contributor Information Jasmeet S Reyat, Institute of Cardiovascular Sciences, College of Medical and Dental Sciences, University of Birmingham, Wolfson Drive, B15 2TT Birmingham, UK; Division of Cardiovascular Medicine, Radcliffe Department of Medicine, University of Oxford, John Radcliffe Hospital, OX3 9DU Oxford, UK. Laura C Sommerfeld, Institute of Cardiovascular Sciences, College of Medical and Dental Sciences, University of Birmingham, Wolfson Drive, B15 2TT Birmingham, UK; Department of Cardiology, University Heart and Vascular Center Hamburg, University Medical Center Hamburg-Eppendorf, Martinistraße 52, 20246 Hamburg, Germany; DZHK (German Center for Cardiovascular Research), Partner Site Hamburg/Kiel/Lübeck, University Medical Center Hamburg-Eppendorf, Martinistraße 52, 20246 Hamburg, Germany; University Center of Cardiovascular Sciences, University Medical Center Hamburg-Eppendorf, Martinistraße 52, 20246 Hamburg, Germany. Molly O’Reilly, Institute of Cardiovascular Sciences, College of Medical and Dental Sciences, University of Birmingham, Wolfson Drive, B15 2TT Birmingham, UK. Victor Roth Cardoso, Institute of Cardiovascular Sciences, College of Medical and Dental Sciences, University of Birmingham, Wolfson Drive, B15 2TT Birmingham, UK; Institute of Cancer Genomics, College of Medical and Dental Sciences, University of Birmingham, B15 2TT Birmingham, UK. Ellen Thiemann, Department of Cardiology, University Heart and Vascular Center Hamburg, University Medical Center Hamburg-Eppendorf, Martinistraße 52, 20246 Hamburg, Germany; DZHK (German Center for Cardiovascular Research), Partner Site Hamburg/Kiel/Lübeck, University Medical Center Hamburg-Eppendorf, Martinistraße 52, 20246 Hamburg, Germany; Institute of Experimental Pharmacology and Toxicology, Cardiovascular Research Center, University Medical Center Hamburg-Eppendorf, Martinistrasse 52, 20246 Hamburg, Germany. Abdullah O Khan, Institute of Cardiovascular Sciences, College of Medical and Dental Sciences, University of Birmingham, Wolfson Drive, B15 2TT Birmingham, UK. Christopher O’Shea, Institute of Cardiovascular Sciences, College of Medical and Dental Sciences, University of Birmingham, Wolfson Drive, B15 2TT Birmingham, UK. Sönke Harder, Institut für Klinische Chemie und Laboratoriumsmedizin, Massenspektrometrische Proteomanalytik, University Medical Center Hamburg-Eppendorf, Martinistrasse 52, 20246 Hamburg, Germany. Christian Müller, UKE Bioinformatics Core, University Medical Center Hamburg-Eppendorf, Martinistrasse 52, 20246 Hamburg, Germany. Jonathan Barlow, Cellular Health and Metabolism Facility, College of Life and Environmental Sciences, University of Birmingham, B15 2TT Birmingham, UK. Rachel J Stapley, Institute of Cardiovascular Sciences, College of Medical and Dental Sciences, University of Birmingham, Wolfson Drive, B15 2TT Birmingham, UK. Winnie Chua, Institute of Cardiovascular Sciences, College of Medical and Dental Sciences, University of Birmingham, Wolfson Drive, B15 2TT Birmingham, UK. S Nashitha Kabir, Institute of Cardiovascular Sciences, College of Medical and Dental Sciences, University of Birmingham, Wolfson Drive, B15 2TT Birmingham, UK. Olivia Grech, Institute of Cardiovascular Sciences, College of Medical and Dental Sciences, University of Birmingham, Wolfson Drive, B15 2TT Birmingham, UK. Oliver Hummel, Max Delbrück Centrum for Molecular Medicine, Robert-Rössle-Straße 10, 13125 Berlin, Germany. Norbert Hübner, Max Delbrück Centrum for Molecular Medicine, Robert-Rössle-Straße 10, 13125 Berlin, Germany; Charite—Universitätsmedizin Berlin, Charitéplatz 1, 10117 Berlin, Germany; DZHK (German Center for Cardiovascular Research), Partner Site Berlin, Germany. Stefan Kääb, Department of Medicine I, University Hospital Munich, Ludwig Maximilian University of Munich (LMU), Marchioninistraße 15, 81377 Munich, Germany; DZHK (German Centre for Cardiovascular Research), Partner Site Munich Heart Alliance, Munich, Germany. Lluis Mont, Hospital Clínic, Universitat de Barcelona, Villarroel, 170, 08036, Barcelona, Catalonia, Spain; Institut de Recerca Biomèdica, August Pi- i Sunyer, Roselló, 149-153, 08036 Barcelona, Catalonia, Spain; Centro Investigación Biomedica en Red Cardiovascular, Av. Monforte de Lemos, 3-5. Pabellón 11. Planta 0, 28029 Madrid, Spain. Stéphane N Hatem, INSERM UMRS1166, ICAN—Institute of Cardiometabolism and Nutrition, Sorbonne University, Institute of Cardiology, Pitié-Salpêtrière Hospital, 91 Boulevard de l’Hôpital, 75013 Paris, France. Joris Winters, Department of Physiology, Cardiovascular Research Institute Maastricht, Maastricht University, Minderbroedersberg 4-66211 LK Maastricht, The Netherlands. Stef Zeemering, Department of Physiology, Cardiovascular Research Institute Maastricht, Maastricht University, Minderbroedersberg 4-66211 LK Maastricht, The Netherlands. Neil V Morgan, Institute of Cardiovascular Sciences, College of Medical and Dental Sciences, University of Birmingham, Wolfson Drive, B15 2TT Birmingham, UK. Julie Rayes, Institute of Cardiovascular Sciences, College of Medical and Dental Sciences, University of Birmingham, Wolfson Drive, B15 2TT Birmingham, UK. Katja Gehmlich, Institute of Cardiovascular Sciences, College of Medical and Dental Sciences, University of Birmingham, Wolfson Drive, B15 2TT Birmingham, UK. Monika Stoll, Institute of Human Genetics, Genetic Epidemiology, WWU Münster, Albert-Schweitzer-Campus 1, D3, Domagkstraße 3, 48149 Münster, Germany; Cardiovascular Research Institute Maastricht, Genetic Epidemiology and Statistical Genetics, Maastricht University, Universiteitssingel 50, 6229 ER Maastricht, The Netherlands. Theresa Brand, Institute of Pharmacology and Toxicology, University of Würzburg, Versbacher Straße 9, 97078 Würzburg, Germany. Michaela Schweizer, Department of Morphology and Electron Microscopy, Center for Molecular Neurobiology, University Medical Center Hamburg-Eppendorf, Martinistraße 52, 20246 Hamburg, Germany. Angelika Piasecki, Institute of Experimental Pharmacology and Toxicology, Cardiovascular Research Center, University Medical Center Hamburg-Eppendorf, Martinistrasse 52, 20246 Hamburg, Germany. Ulrich Schotten, Department of Physiology, Cardiovascular Research Institute Maastricht, Maastricht University, Minderbroedersberg 4-66211 LK Maastricht, The Netherlands. Georgios V Gkoutos, Institute of Cancer Genomics, College of Medical and Dental Sciences, University of Birmingham, B15 2TT Birmingham, UK. Kristina Lorenz, Institute of Pharmacology and Toxicology, University of Würzburg, Versbacher Straße 9, 97078 Würzburg, Germany; Leibniz-Institut für Analytische Wissenschaften—ISAS—e.V., ISAS City, Bunsen-Kirchhoff-Straße 11, 44139 Dortmund, Germany. Friederike Cuello, DZHK (German Center for Cardiovascular Research), Partner Site Hamburg/Kiel/Lübeck, University Medical Center Hamburg-Eppendorf, Martinistraße 52, 20246 Hamburg, Germany; Institute of Experimental Pharmacology and Toxicology, Cardiovascular Research Center, University Medical Center Hamburg-Eppendorf, Martinistrasse 52, 20246 Hamburg, Germany. Paulus Kirchhof, Institute of Cardiovascular Sciences, College of Medical and Dental Sciences, University of Birmingham, Wolfson Drive, B15 2TT Birmingham, UK; Department of Cardiology, University Heart and Vascular Center Hamburg, University Medical Center Hamburg-Eppendorf, Martinistraße 52, 20246 Hamburg, Germany; DZHK (German Center for Cardiovascular Research), Partner Site Hamburg/Kiel/Lübeck, University Medical Center Hamburg-Eppendorf, Martinistraße 52, 20246 Hamburg, Germany. Larissa Fabritz, Institute of Cardiovascular Sciences, College of Medical and Dental Sciences, University of Birmingham, Wolfson Drive, B15 2TT Birmingham, UK; Department of Cardiology, University Heart and Vascular Center Hamburg, University Medical Center Hamburg-Eppendorf, Martinistraße 52, 20246 Hamburg, Germany; DZHK (German Center for Cardiovascular Research), Partner Site Hamburg/Kiel/Lübeck, University Medical Center Hamburg-Eppendorf, Martinistraße 52, 20246 Hamburg, Germany; University Center of Cardiovascular Sciences, University Medical Center Hamburg-Eppendorf, Martinistraße 52, 20246 Hamburg, Germany. Supplementary material Supplementary material is available at Cardiovascular Research online. Author contributions Conception and design of the research: J.S.R., P.K., L.F.; acquisition of data: J.S.R., L.C.S., M.o.R., V.R.C., A.O.K., S.H., C.M., J.B., O.H., N.H., R.J.S., S.N.K., O.G., M.S., T.B., F.C., A.P.; analysis and interpretation of the data: J.S.R., L.C.S., M.o.R., V.R.C., E.T., A.O.K., C.o.S., S.H., N.H., O.H., C.M., M.S., A.P., J.B., T.B., K.L., F.C.; statistical analysis: J.S.R., V.R.C., J.B., W.C., J.W., T.B., C.L., F.C., S.Z.; supervising the experiments: J.S.R., F.C., P.K., L.F.; drafting the manuscript: J.S.R., P.K., L.F.; critical revision of the manuscript for important intellectual content: L.C.S., W.C., S.K., L.M., S.N.H., N.V.M., J.R., K.G., K.L., F.C., M.St., U.S., G.V.G. All authors approve the current version of the manuscript. Funding This work was supported by the European Commission (grant agreements no. 633196 [CATCH ME]) to P.K., L.F., M.S. and U.S. no. 116074 [BigData@Heart EU MI] to P.K., no. 985286 [MAESTRIA] to L.F., U.S., British Heart Foundation (FS/13/43/30324 and AA/18/2/34218 to P.K. and L.F.), German Centre for Cardiovascular Research supported by the German Ministry of Education and Research (DZHK, to P.K. and LF), Foundation Leducq (140VD01) to P.K, Deutsche Forschungsgemeinschaft (DFG, 509167694; KI73/4-1) to P.K and (CU53/5-1; CU53/10-1; CU53/12-1) to FC, and SFB1525/453989101 and SFB/TR 296/424957847 to K.L. J.S.R. and K.G. acknowledge support from a British Heart Foundation Accelerator Award (AA/18/2/34218). K.G. is supported by the Medical Research Council (MR/V009540/1) and the National Centre for the Replacement, Refinement and Reduction of Animals in Research (NC/T001747/1). A.O.K. and C.o.S. are Sir Henry Wellcome Fellows (218649/Z/19/Z; 221650/Z/20/Z). J.R. is a British Heart Foundation Intermediate Fellow (FS/IBSRF/20/25039). U.S. is supported by grants from the Netherlands Heart Foundation (CVON2014-09 [RACE-V] and grant number 01-002-2022-0118, EmbRACE) and the European Commission (no. 860974 [ITN Network Personalise AF] and no. 952166 [REPAIR]). NH is supported by the DFG SFB-1470-B03, the Chan Zuckerberg Foundation, and an ERC Advanced Grant under the European Union Horizon 2020 Research and Innovation Program (AdG788970). V.R.C. and G.V.G. acknowledge support from the MRC Health Data Research UK (HDRUK/CFC/01), an initiative funded by the UK Research and Innovation, Department of Health and Social Care (England). M.O. is the recipient of a Dutch Heart Foundation Dekker grant (03-004-2022-0036) and ZonMW Off-Road research grant (04510012110049). A CC BY or equivalent license is applied to AAM arising from this submission, in accordance with the grant’s open access conditions. Data availability All data associated with this study are present in the paper or in the Supplementary Materials. Sequencing datasets used in this study can be requested from the corresponding author. The mass spectrometry proteomics dataset has been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD037189. The snRNA sequencing data set has been deposited in the European Nucleotide Archive (ENA) at EMBL-EBI under accession number PRJEB75160 (https://www.ebi.ac.uk/ena/browser/view/PRJEB75160). References Associated Data Supplementary Materials Data Availability Statement All data associated with this study are present in the paper or in the Supplementary Materials. Sequencing datasets used in this study can be requested from the corresponding author. The mass spectrometry proteomics dataset has been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD037189. The snRNA sequencing data set has been deposited in the European Nucleotide Archive (ENA) at EMBL-EBI under accession number PRJEB75160 (https://www.ebi.ac.uk/ena/browser/view/PRJEB75160).