IRX3 depletion enhances hiPSC-derived cardiomyocyte differentiation

Given the established role of Irx3 in mouse ventricular conduction system specification and maturation in vivo [22,24], we investigated whether IRX3 depletion similarly influences the differentiation of hiPSC-CMs. Using CRISPR–Cas9, we generated from the same mutagenesis experiment two IRX3 loss-of-function mutant hiPSC lines, hereafter referred to as IRX3-KO: IRX3^cl1.1-/- and IRX3^cl1.2-/- (S1 Fig). We targeted the end of IRX3 exon 1, upstream of the IRX3 DNA binding domain (Fig 1A). IRX3^cl1.1-/- harbors deletions of 2 and 59 bp in each allele, while IRX3^cl1.2-/- carries deletions of 2 and 227 bp (S1 Fig). All the resulting mutations are in line with previous reports on CRISPR–Cas9 genome editing [30]. Both 2 and 59 bp mutations introduce premature termination codons (PTCs) before the IRX3 DNA-binding domain (S1 and S2 Figs). The 227 bp deletion encompasses the transcription start site (TSS) and should not be transcribed (S1 Fig); however, an alternative start codon 483 bp downstream may give rise to a truncated protein (S2 Fig). The two IRX3-KO hiPSC lines display markedly reduced IRX3 gene expression levels (S1 Fig). Clone identity and homogeneity were confirmed via subcloning and sequencing (S1 Fig), and pluripotency features remained unaffected across all lines, as demonstrated by karyotyping, expression of pluripotency markers, teratoma formation, and SNP array (S1 and S3 Figs; and S1 Table).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. IRX3 depletion increases hiPSC-derived cardiomyocyte differentiation.

(A-B) Schematic of CRISPR–Cas9-mediated generation of loss-of-function mutant hiPSC clones IRX3^cl1.1-/-, IRX3^cl1.2-/-, and IRX3^cl2-/- with respective protocol used for hiPSC-CM differentiation. (C) Representative immunofluorescence of cTnI (red) for IRX3^cl1+/+, IRX3^cl1.1-/-, and IRX3^cl1.2-/- hiPSC-CMs at day 30 of differentiation. (D) Percentage of TnI-positive cells (n = 2,000-4,000 cells analyzed per group of 4 independent replicates). Cell nuclei were counterstained with DAPI (blue). (E-H) Calcium transient analysis of IRX3^cl1+/+, IRX3^cl1.1-/-, and IRX3^cl1.2-/- hiPSC-CMs at day 30 of differentiation. (E) Calcium transient amplitude. (F) Calcium transient maximum slope. (G) Corrected calcium transient duration at 50% (CaD_50c) and 90% (CaD_90c) of decay, n = 19-22 fields of view of 3 independent replicates. (H) Representative curves of calcium transients. Data are presented as box plots with individual replicates. One-way ANOVA. *P < 0.05; **P < 0.01 vs IRX3^cl1+/+. Scale bar = 50 µm.

https://doi.org/10.1371/journal.pone.0351704.g001

To examine the effects of IRX3 depletion on CM development, we differentiated hiPSCs using a Wnt-modulation protocol (Fig 1B). At day 30, IRX3-KO lines showed ~15% higher CM yield, assessed by immunostaining for the sarcomeric marker cTnI (encoded by TNNI3) (Fig 1C-D). Gene expression profiling revealed upregulation of TNNI1 and CX43 (GJA1), a gap junction protein characteristic of working CMs, in IRX3-KO hiPSC-CMs (S4A-B Fig) [31,32]. In contrast, no change was observed in CX40 (GJA5) expression (S4 Fig), while SCN5A, a marker of the ventricular conduction system, was reduced (S4 Fig) [33,34].

As hiPSC-CMs mature, they undergo physiological and metabolic changes that enable a faster and more robust response to stimuli. Calcium handling is central to these changes, governing both contraction and relaxation cycles [35]. In line with a more mature phenotype after IRX3 depletion, we observed that both IRX3^cl1.1-/- and IRX3^cl1.2-/- clones showed a significant increase in peak amplitude (Fig 1E), faster transient upstroke slope (Fig 1F), and shorter calcium transient duration at 50% and 90% decay (CaD_50c and CaD_90c) compared with IRX3^cl1+/+ (Fig 1G), suggesting enhanced calcium handling (Fig 1H).

Together, these findings suggest that IRX3 depletion improves CM differentiation and may bias subtype specification toward working CMs, consistent with in vivo mouse data showing that Irx3 promotes conduction system identity instead of working CM fate [24].

IRX3 depletion improves cardiomyocyte function and structure

To enable higher-resolution structural and functional analyses, we generated an IRX3 KO line using a commercial TROPO-GFP hiPSC reporter line (designated IRX3^cl2-/-), which carries a homozygous 1 bp insertion (S1-S3 Figs). First, we evaluated whether the enhanced calcium handling observed in previous clones was conserved in this reporter line. We confirmed shortening of CaD_50c and CaD_90c (Fig 2A), alongside a 19% increase in peak amplitude (Fig 2B), and an 88% faster transient upstroke velocity (Fig 2C), reflecting improved calcium kinetics in the absence of IRX3 (Fig 2D). These functional improvements were accompanied by the upregulation of SERCA2 and RYR2, the primary regulators of intracellular calcium reuptake and release, respectively [36]. Consistent with this maturation profile, the ventricular isoform MYH7 was significantly upregulated in IRX3^cl2-/- cardiomyocytes. MYH6 expression was also increased, while TNNI3 levels remained comparable to controls (S5 Fig).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. IRX3-KO cardiomyocytes display robust excitation-contraction coupling and mitochondrial organization.

(A) Calcium transient at 50% (CaD_50c) and 90% (CaD_90c) of decay. (B) Calcium transient amplitude. (C) Calcium transient maximum slope, n = 37-40 regions of 2 independent replicates. (D) Representative curves of calcium transient. (E) Maximum contraction speed. (F) Maximum relaxation speed. (G) Total shortening area. (H) Total contraction-relaxation area, n = 5-6 independent regions. (I) Representative curves of contraction and relaxation from IRX3^cl2+/+ (black dash line) and IRX3^cl2-/- (gold dash line) hiPSC-CMs at day 30 of differentiation. (J) Representative immunofluorescence of TOM20 (red) and endogenous TnI-GFP (green). (K) Schematic figure and TOM20 fluorescence distribution analyzed in four regions, ranging from the perinuclear area to membrane. (L) TOM20 fluorescence distribution per region (n = 103-127 individually analyzed cells of 2 independent replicates) from IRX3^cl2+/+ and IRX3^cl2-/- hiPSC-CMs at day 30 of differentiation. (M) Ultrastructure of IRX3^cl2+/+ and IRX3^cl2-/- hiPSC-CMs detected by transmission electron microscopy. Black arrows indicate mitochondria further highlighted by zoom inset. (N) Quantification of mitochondria area (n = 744-1206 individually analyzed mitochondria of 1 independent experiment) from IRX3^cl2+/+ and IRX3^cl2-/- hiPSC-CMs at day 30 of differentiation. Data are presented as box plots (A-C, E-H), mean ± SD (L), and violin plot (N). Student’s t-test was applied to all plots except L, which was analyzed using two-way ANOVA. *P < 0.05; **P < 0.01; ***P < 0.001; and ****P < 0.0001 vs IRX3^cl2+/+. Scale bars = 20 µm.

https://doi.org/10.1371/journal.pone.0351704.g002

Consistent with these calcium handling improvements, action potential durations were significantly reduced in IRX3^cl2-/- hiPSC-CMs by 25% (APD50) and 17% (APD90) compared to controls, with no significant differences observed in action potential amplitude (S5 Fig). Furthermore, contraction and relaxation velocities were increased by 40% and 48%, respectively (Fig 2E-F), while contraction amplitudes were similarly improved (Fig 2G-I). These functional enhancements were accompanied by a more extensive mitochondrial network and increased mitochondrial size (Fig 2J-N), indicative of a shift toward more mature metabolic function.

Immunostaining for the sarcomeric markers cTnI and alpha-actinin-2 at day 30 revealed enhanced myofibrillar alignment in IRX3^cl2-/- hiPSC-CMs (Fig 3A). Quantitative analysis confirmed a 15% increase in organized areas and a 0.03 μm increase in sarcomere length, alongside reduced sarcomeric dispersion (0.67 ± 0.14 vs. 0.62 ± 0.18; p = 0.035), with no changes in cell size (Fig 3B-C). Proliferation analysis indicated an increase at early stages (day 6) in IRX3^cl2-/- cells that attenuated by later stages, suggesting an initial expansion of the progenitor pool (Fig 3D).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 3. IRX3 depletion enhances structural maturation of cardiomyocytes.

(A) Representative immunofluorescence of alpha-Actinin2 (red) and endogenous cTnI-GFP (green). Zoom inset shows the organization of filaments, and the mask of analyzed windows generated with “MorphoScript” software. (B) Order, spacing and dispersion of alpha-Actinin2 filaments, calculated using MorphoScript, presented as frequency distribution and violin plots (n = 103-127 individually analyzed cells of 2 independent replicates). (C) Total area in µm² presented in violin plots (n = 103-127 individually analyzed cells of 2 independent replicates). (D) Proliferation rate from day 0 to day 30 of differentiation (n = 6 independent replicates). (E) Representative immunofluorescence of CX43 (white) and endogenous cTnI-GFP (green), and (F) Quantification of CX43 expression (n = 14 fields of view obtained from 4 independent replicates) from IRX3^cl2+/+ and IRX3^cl2-/- hiPSC-CMs at day 30 of differentiation. Data are presented as violin plots (B-C), mean ± SD (D), and box plots (F). Cell nuclei were counterstained with DAPI (blue). Student’s t-test was applied to all plots, and Mann-Whitney U test for ‘Order’ parameter in Fig 3B. *P < 0.05; **P < 0.01; ***P < 0.001; and ****P < 0.0001 vs IRX3^cl2+/+. Scale bars = 20 µm.

https://doi.org/10.1371/journal.pone.0351704.g003

Additionally, CX43 protein levels were upregulated in IRX3^cl2-/- hiPSC-CMs (Fig 3E-F), corroborating the gene expression trends observed in IRX3^cl1.1-/- (S3 Fig). Functional dye transfer assays confirmed enhanced gap junction communication in IRX3-KO hiPSC-CMs (S5 Fig). Collectively, these data show that IRX3 depletion improves calcium handling, electrophysiological response, contractility, mitochondrial abundance, sarcomere organization, and intercellular communication, which is consistent with a robust enhancement of CM differentiation.

IRX3 depletion activates early cardiogenic gene programs

We next assessed the temporal expression profile of IRX3 in wild-type differentiating hiPSCs. IRX3 mRNA levels peaked at day 4 of differentiation and declined thereafter (Fig 4A), indicating a potential role in early lineage specification.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 4. IRX3 depletion in hiPSCs enhances early cardiac gene expression.

(A) Gene expression profile of IRX3 in IRX3^cl1+/+ clone from days 0 to 15 of differentiation analyzed by microarray. (B) Volcano plot showing downregulated (green) and upregulated (orange) differentially expressed genes between IRX3^cl1+/+ and IRX3^cl1.1-/- at day 4 of differentiation analyzed by microarray (log2fold change, p.adj < 0.05). (C) Gene set enrichment analysis of upregulated genes in IRX3^cl1.1-/- compared with IRX3^cl1+/+. (D) Heatmap showing 12 downregulated and 19 upregulated selected genes when comparing IRX3^cl1+/+ and IRX3^cl1.1-/- at day 4 of differentiation (p.adj < 0.05). (E) Representative immunofluorescence of GATA4 (red), NKX2-5 (green), and KI67 (magenta) in IRX3^cl1+/+ and IRX3^cl1.1-/- cells at day 6 of differentiation. Cell nuclei were counterstained with DAPI (blue). Quantification of GATA4 and NKX2-5 mean fluorescence intensity (MFI) within the nuclei and percentage of KI67-positive cells (p = 0.0545) (n = 3-6 independent experiments, with data averaged from 5 fields of view per sample and 35,000-36,000 cells analyzed per group). Data are presented as box plots. Student’s t-test. **P < 0.01, ****P < 0.0001 vs. IRX3^cl1+/+. Scale bars = 50 µm.

https://doi.org/10.1371/journal.pone.0351704.g004

Microarray analysis of day 4 IRX3^cl1.1-/- cells revealed 937 upregulated and 412 downregulated genes relative to controls (Fig 4B), consistent with the loss of IRX3’s repressive function [37]. Pathway enrichment analysis revealed significant activation of Wnt-related signaling pathways (Fig 4C), which play key roles in early cardiac mesoderm specification [38,39].

Among the differentially expressed genes, we identified 11 cardiac progenitor markers and 9 genes involved in sarcomere maturation (Fig 4D). Upregulated transcription factors included TBX5, HAND1/2, MEF2A/C, MYOCD, KDR, and DAAM1, all associated with cardiogenesis [40–46]. Conversely, 11 pluripotency and germ-layer–associated genes were downregulated, indicating commitment to the cardiac lineage.

To validate these transcriptomic signatures, we assessed the expression of core cardiac transcription factors at Day 4 (S6 Fig). While TBX5 and GATA4 transcripts were consistently upregulated across all IRX3-KO clones, NKX2–5 expression was heterogeneous among clones. However, at the protein level (Day 6), quantitative immunofluorescence revealed a significant increase in nuclear intensity for both GATA4 (+37%) and NKX2–5 (+26%) in IRX3^cl1.1-/- cells relative to controls (Fig 4E). This upregulation of cardiac TFs coincided with a trend toward increased cell proliferation (KI67 + , p = 0.0545), recapitulating phenotypes previously observed in IRX3^cl2-/- progenitors (Fig 3D). Collectively, these data suggest that IRX3 depletion reinforces the expression of early cardiac lineage determinants, potentially priming the population for robust commitment.

In silico analyses suggest IRX3 interacts with cardiac transcription factor targets in CPCs

IRX3 is known to physically interact with master cardiac regulators such as TBX5, GATA4, and NKX2–5 during heart development [24,27]. Consistent with these physical associations, large scale network analyses have mapped the broader regulatory landscape connecting these TFs in human cardiogenesis [27]. To investigate whether IRX3 functionally cooperates with these factors to drive gene expression in progenitors, we performed an integrative analysis using scATAC-seq data (day 5) and ChIP-seq data for TBX5, GATA4, and NKX2–5 (day 6).

Motif scanning identified putative IRX3 binding sequences within open chromatin regions that were also occupied by these TFs (Fig 5A; see S7 Fig for consensus sequence logos). Of the 26,135 genes in the total IRX3 binding universe, we found that 27% (7,156 genes) were co-bound by at least one cardiac factor, while the majority (18,979 genes) constituted ‘IRX3-Solo’ targets (S8 Fig).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 5. IRX3 modulates cardiac differentiation by interacting with the transcriptional network of TBX5, GATA4, and NKX2-5.

(A) Genomic alignment of scATAC-seq open chromatin profiles at the TBX5, GATA4, and NKX2-5 loci. Predicted IRX3 binding motifs are highlighted with enlarged, color-coded markers to indicate their specific genomic context: markers denote motifs located within open chromatin regions (ATAC-seq peaks) that also physically overlap with ChIP-seq binding peaks for the respective transcription factor. (B) Venn diagram intersecting the target genes of TBX5, GATA4, and NKX2-5 (identified by ChIP-seq and containing IRX3 motifs) with the 1,344 differentially expressed genes (DEGs) identified in IRX3^cl1.1-/- at Day 4. (C) Gene Ontology (GO) enrichment analysis for biological processes associated with the 494 overlapping genes. (D) Specific cardiogenic targets modulated by IRX3 depletion that are direct downstream targets of at least one of the three core cardiac TFs.

https://doi.org/10.1371/journal.pone.0351704.g005

Notably, this distinction was not driven by gene biotype, as both the ‘Solo’ and ‘Co-occupied’ populations contained a similar proportion of protein-coding genes (~60%; S8 Fig). However, functional analysis revealed a clear divergence. While the ‘IRX3-Solo’ fraction was associated with non-cardiac lineages (e.g., sensory perception, immune effector process), the co-occupied subset was exclusively enriched for cardiac-specific processes including muscle development, cell-cell junction assembly, and actin-cytoskeletal organization (S8 Fig). Further classification by molecular function suggested a central role in actin binding and structural integrity (S9 Fig).

Within the co-occupied fraction, NKX2–5 showed the highest overlap (5,067 genes), followed by GATA4 (3,227) and TBX5 (1,997) (S8 Fig). Genomic distribution analysis indicated that the majority of these co-occupied peaks were located within promoter regions (S10 Fig). The subset of 655 genes commonly targeted by all three TFs and IRX3 was significantly enriched for terms related to cardiac morphogenesis (S8 Fig), supporting a model where IRX3 modulates cardiac differentiation by regulating the structural machinery, potentially via TF complexes [27].

To validate these genomic predictions, we intersected the 1,344 DEGs identified in our Day 4 IRX3^cl1.1-/- microarray with the TF binding profile. We found that approximately 37% (494 genes) of the IRX3-dependent DEGs correspond to putative IRX3/TF co-targets (Fig 5B), a set that includes central regulators of Wnt signaling and muscle development (Fig 5C). Specifically, among the 22 cardiogenesis-associated genes identified earlier (Fig 4D and S6 Fig), 16 are direct targets of at least one TF, and 8 are targeted by all three (Fig 5D). Notably, this shared regulatory network includes the TFs themselves – TBX5, GATA4 and NKX2–5 (see S2 Table for the complete gene set).

Altogether, these findings support a model in which IRX3 shapes early cardiac commitment not only by regulating the expression of TBX5, GATA4, and NKX2–5, but by directly modulating their downstream transcriptional programs.

Ethics Statement

This study involved two human induced pluripotent stem cells (hiPSCs) lines and animal research, all conducted under approved ethical guidelines.

Donor-derived hiPSC line (SS109 clone): hiPSCs were generated from adult skin fibroblasts after obtaining written informed consent from the donor. The consent process and study were reviewed and approved by the Research Projects Ethics Committee (CAPPesq) of the Clinical Directorate of Hospital das Clínicas, Faculty of Medicine, University of São Paulo, under protocol number 0055/11, approved on March 21, 2012. Project title: Use of iPS (induced pluripotent stem) cells to understand changes in cardiomyocytes of patients with genetically based cardiomyopathies [55].

Commercial hiPSC line (AICS-0037–172): We used the AICS-0037–172 line developed at the Allen Institute for Cell Science (allencell.org/cell-catalog) and available through the Coriell Institute for Medical Research. This line is derived from the WTC parental line (Coriell Cat#GM25256), originally developed by the Conklin Laboratory at the J. David Gladstone Institute. The cell line was generated with IRB approval and written informed consent from the donor, and all identifying information was removed prior to distribution. The use of this line in our study complies with the terms outlined by the Allen Cell Collection (https://catalog.coriell.org/1/AllenCellCollection/MTA-and-Additional-Terms). Use of this line was reviewed and approved by the Committee for Ethics in Animal Use (CEUA) of the University of São Paulo Medical School under protocol number 1656/2021, approved on November 24, 2021. Project title: Cardiomyocytes Derived from Human iPSC (hiPSC-CMs) for Post-infarction Cardiac Repair and Regeneration in Swine.

Animal research: All animal procedures were conducted in accordance with institutional guidelines and the Animal Research: Reporting of In Vivo Experiments (ARRIVE) guidelines. Animal research was approved by the Committee for Ethics in Animal Use (CEUA) of the University of São Paulo Medical School under license number 1089/2018, approved on September 3, 2018. All efforts were made to minimize animal suffering.

Cell lines

Three IRX3-KO cell lines derived from two human iPSC (hiPSC) lines were utilized in this study. The SS109 lineage, developed in our laboratory [55], was first assessed in January 2020 to generate two mutant lines (IRX3^cl1.1-/- and IRX3^cl1.2-/-). The AICS-0037–172 (TROPO-GFP) hiPSC line was obtained from the Coriell Institute (Camden, NJ, USA) in July 2019, first assessed for experimental work at our lab in November 2019 and used to generate the IRX3^cl2-/- line in January 2022. The use of this line follows the citation and distribution requirements outlined by the Allen Cell Collection [56].

All hiPSC lines were maintained in StemFlex medium (Gibco) at 37ºC and 5% CO₂. Cells were passaged using Versene (1x) (Gibco) upon reaching 80−90% confluency and were regularly monitored for mycoplasma contamination. Following CM differentiation, cells were maintained in RPMI 1640 medium supplemented with B-27 with insulin and refreshed every 4 days. When necessary, cells were dissociated with Trypsin-EDTA 0.25%.

hiPSC generation

To exclude protocol and cell line-specific responses, we evaluated the role of IRX3 depletion on CM differentiation using two distinct differentiation protocols and two hiPSCs lines: SS109 and TROPO-GFP (monoallelic mEGFP Tagged TNNI1 WTC. SS109 clone was generated with lentivirus from skin fibroblasts of a healthy donor following the protocol of Somers et.al. and previously published in Muñoz JJAM, 2022 [55,57]. Both cell lines were characterized for the expression of pluripotent markers by immunofluorescence and flow cytometry, in vivo teratoma formation assay and no karyotypic abnormalities were observed (S1and S3 Figs).

Animal research

All animal procedures were conducted in accordance with the Animal Research: Reporting of In Vivo Experiments (ARRIVE) guidelines and approved by the Committee for Ethics in Animal Use (CEUA) of the University of São Paulo Medical School under license number 1089/2018, approved on September 3, 2018. All efforts were made to minimize animal suffering.

Adult male BALB/c mice (2–3 months old) were used for the in vivo teratoma formation assay. Animals were maintained under standard temperature and lighting conditions with ad libitum access to food and water. Mice were anesthetized using 2% isoflurane, and upon adequate sedation, received subcutaneous injections on the back near the hind limb. Each injection contained 5 × 10⁵ hiPSCs in 50 μL, mixed in a 1:1 ratio with Geltrex. Mice were monitored weekly for teratoma formation. Four weeks after injection, animals were euthanized by isoflurane inhalation (4–5%) in a closed chamber until respiratory arrest, followed by intracardiac perfusion to ensure irreversible death. Teratomas were collected, fixed in 10% formalin, and processed for histological evaluation using hematoxylin and eosin staining.

Plasmid transfection (IRX3^cl1-/-)

gRNAs were designed to target exon 1 of IRX3, upstream of the DNA binding domain, using the CRISPOR website (https://crispor.gi.ucsc.edu/). To generate IRX3^cl1.1-/- and IRX3^cl1.2-/- (SS109 clones), 2.5 x 10⁴ hiPSCs were mixed with 4 µg sgRNA (MLM3636, Addgene) and 4 µg plasmid Cas9 (hCas9 Church, Addgene). The cell-plasmid mixtures were electroporated using the Neon transfection system (ThermoFisher Scientific) under the conditions of 1,400 V, 20 ms, and 1 pulse. Subsequently, the electroporated cells were seeded in a 6-well plate containing mTeSR (StemCell Technologies) supplemented with 10 μM ROCK Inhibitor Y-27632 (Sigma-Aldrich) for 24 hours. The next day, the medium was changed to remove Y-27632. On day 2, the hiPSC SS109 clone was dissociated into single cells and replated into individual 96-well plates to generate single-cell clones. After 10–14 days, individual colonies were passaged and screened by PCR and the selected clones had their DNA sequence confirmed by sequencing.

sgRNA design, synthesis and in vitro assembly of Cas9-gRNAs (IRX3^cl2-/-)

The target gene sequences were identified using the CRISPRscan algorithm (http://www.crisprscan.org). gRNAs were designed to target exon 1 of IRX3, upstream of the DNA binding domain, using the CRISPRScan website [58] (scores ranging from 15 to 25, with an average score of 18), with the human genome in the Homo sapiens version as the reference (see the description of gRNAs in the S3 Table). The designed gRNAs obtained an efficiency score predicted by the CRISPRScan website [58]. From CRISPRScan, we also predicted the potential off-target sites for the gRNAs, defined by the CFD score [59]. For the IRX3 gene, zero off-targets and zero mismatches were predicted. For the DNA oligo, a target-specific CRISPR RNA (crRNA) and a trans-activating crRNA (tracrRNA) were designed, combined into a single transcript, with a PAM sequence (NGG) allowing Cas9 to initiate binding. The assembled oligos were synthesized and purified by HPLC by ThermoFisher Scientific company.

The gRNAs were in vitro synthesized using the GeneArt Precision gRNA Synthesis Kit (ThermoFisher Scientific) according to the provided protocol. In brief, the DNA template was first assembled with 25 μL of polymerase chain reaction (PCR) components, including 12.5 μL of Phusion High-Fidelity PCR Master Mix (2x), 1.0 μL of Tracr Fragment + T7 Primer Mix, 1.0 μL of forward and reverse primers (0.3 μM for each primer, S4 Table), and 10.5 μL of nuclease-free water. The PCR was performed at 98 °C for 10 s, followed by 32 cycles of denaturation at 98 °C for 5 s, annealing at 55 °C for 15 s, and a final extension at 72 °C for 1 minute.

In vitro transcription of the sgRNA was conducted by incubating with another 20 μL reaction component, including 8 μL of DNTP mix (100 mM of ATP, GTP, CTP, and UTP), 6 μL of the above-mentioned DNA template, 4 μL of TranscriptAid Reaction Buffer (5×), and 2 μL of TranscriptAid Enzyme Mix for 3 hours at 37 °C. Subsequently, the residual DNA template was digested with 1 μL of DNase I. The sgRNA was purified using the GeneArt gRNA Clean-up Kit, verified by gel electrophoresis and finally stored at −80 °C.

In brief, 8 x 10⁴ cells per well were seeded in 12-well plates. Three hours later, the RNP complexes were formed by combining 0.5 μg equimolar mixture of tracer and gRNAs with 2.5 μg Cas9 nuclease (gRNA targeting IRX3 and Cas9 protein) using Lipofectamine™ CRISPRMAX™ Cas9 Transfection Reagent (Invitrogen, Thermo Fisher Scientific) with TrueCut™ Cas9 v2 (Invitrogen, Thermo Fisher Scientific) protein following the manufacturer’s recommendations of Lipofectamine CRISPRMAX and protocol of Xin Yu et.al. [60]. At 4 hours post-transfection, the media containing the transfection reagent was removed and replaced with fresh mTeSR (STEMCELL Technologies) supplemented with Y-27632 (Sigma-Aldrich) for 1 day. After additional 2 days, the cells were dissociated and plated at single-cell density into 96-well plates. Clones were screened by sequencing of amplicons spanning the target site of IRX3 gene exon 1. The positive hiPSC clones were expanded and characterized (S1-2 and S4 Figs). See S3-4 Tables for a complete list of guides and primers.

Generation of hiPSC subclones by manual colony picking

Clonal sublines were generated by manual colony picking following established protocols for pluripotent stem cell clonal expansion [61,62]. To facilitate clonal isolation, cells were seeded at low density to allow the formation of spatially distinct colonies derived from single cells. Emerging colonies were identified under an inverted microscope based on morphology and size and were manually picked using sterile pipette tips under a stereomicroscope. Each colony was transferred into an individual well of a Geltrex-coated 24-well plate containing pre-warmed hiPSC mTeSR medium (StemCell Technologies). To enhance survival after manipulation, the medium was supplemented with 10 µM ROCK inhibitor (Y-27632) during the initial 24-hour recovery period. Subclones were subsequently expanded stepwise into larger vessels until sufficient cell numbers were obtained for cryopreservation, genomic DNA extraction, and downstream analyses. Only subclones exhibiting typical hiPSC morphology and stable proliferation were selected.

Genomic DNA extraction, PCR amplification, and Sanger sequencing

Genomic DNA was extracted from hiPSC subclones using the DNeasy Blood & Tissue Kit (Qiagen, Hilden, Germany) according to the manufacturer’s instructions. The genomic region encompassing the CRISPR–Cas9 target site at the IRX3 locus was amplified by PCR using primers flanking the sgRNA recognition sequence (S4 Table). Reactions were carried out using a GoTaq® Flexi DNA Polymerase (Promega, USA). PCR products were verified by agarose gel electrophoresis and purified using either the QIAquick Gel Extration kit (Qiagen, Hilden, Germany) or ExoSAP-IT (Affymetrix, USA). Purified amplicons were subjected to bidirectional Sanger sequencing using the same primers employed for amplification. Sequencing reactions were performed using the BigDye® Terminator v3.1 Cycle Sequencing Kit (Applied Biosystems, USA) and analyzed on a 3500xL Genetic Analyzer (Applied Biosystems, USA). Sequencing chromatograms were inspected and analyzed using Geneious Prime (Biomatters, New Zealand) to confirm genotype.

Genome Integrity Assessment by SNP array

To assess genomic stability following CRISPR–Cas9 editing, genome-wide SNP array analysis was performed using the Axiom™ Human Genotyping SARS-CoV-2 Research Array (~870k markers, Thermo Fisher Scientific) on the GeneTitan™ MC Instrument. Genomic DNA was extracted from parental wild-type cells (IRX3^cl1+/+ and IRX3^cl2+/+-) and their respective edited clones (IRX3^cl1.1-/-, IRX3^cl1.2-/-, and IRX3^cl2-/-). DNA (200 ng per sample) was processed using the Axiom™ 2.0 Reagent Kit following the standard workflow, which includes amplification, fragmentation, precipitation, and hybridization to array plates, followed by automated washing, staining, and scanning. Genotype calling and quality control were performed using Axiom™ Analysis Suite v5.1 (Thermo Fisher Scientific). Single Nucleotide Variant (SNV) calling was executed using the standard Genotyping workflow, while Copy Number Variation (CNV) and Loss of Heterozygosity (LOH) were assessed using the CNV Discovery workflow, utilizing GRCh38/hg38 as the reference genome. To verify genomic integrity, SNV concordance analysis was performed by comparing genotype calls between parental and edited cell pairs. Concordance rates >98% between isogenic pairs were used as the threshold for genomic stability, while comparisons between genetically distinct backgrounds served as specificity controls.

Flow cytometry

The hiPSCs were fixed with 4% paraformaldehyde (PFA, Sigma-Aldrich) at room temperature for 20 min, permeabilized with 0.1% Triton X-100 (Sigma-Aldrich) for 15 min at 4 ºC and incubated with blocking solution (3% BSA/PBS, Sigma-Aldrich) for 30 min at 4 ºC. Next, the cells were incubated with the primary antibodies (1:100) diluted in blocking solution for 1 hour at 4 ºC. The fluorescently conjugated antibodies used to characterize undifferentiated cells were Oct4 (Millipore), Sox2 (Millipore), TRA-1–60 (Millipore) and TRA-1–81 (Millipore). Samples were read on BD Accuri™ C6 Personal Flow Cytometry (BD Biosciences) and analyzed with FloJow™ v10.8.1 Software (BD Biosciences).

Differentiation into hiPSC-CMs

The differentiation processes for IRX3^cl1+/+, IRX3^cl1.1-/-, IRX3^cl1.2-/-, IRX3^cl2+/+, IRX3^cl2-/- hiPSCs into CMs were conducted according to Wnt pathway stimulation and inhibition protocols adapted from previous studies. Two well-established CM differentiation protocols were utilized to reinforce the reproducibility of our findings using IRX3^cl1 [63] and IRX3^cl2 [64] clones. For both, a couple of days before the start of the protocol, cells were subcultured in the proportion of 1:3 and seeded into Geltrex (Gibco)-coated 12-well plates. Upon reaching 80–90% confluence, they were treated with CHIR99021 (6–12 μM, Millipore) and cultured in RPMI 1640 medium supplemented with B-27 minus insulin (Gibco). For IRX3^cl1 clones, following a 24-hour 8 µM CHIR99021 treatment, the medium was replaced with RPMI 1640 supplemented with B-27 minus insulin and 10 ng/ml BMP4 (R&D Systems). On day 2, medium was changed to RPMI 1640 supplemented with B-27 minus insulin, 2.5 µM KY2111 (Cayman Chemical) and XAV939 (Cayman Chemical). Starting from day 4, the medium was refreshed every 48 hours with RPMI 1640 supplemented with B-27 with insulin (Gibco). For IRX3^cl2 clones, 48 hours after CHIR00921 treatment, the medium was replaced with RPMI 1640 supplemented with B-27 minus insulin. On day 3, the medium was changed to RPMI 1640 supplemented with B-27 minus insulin and 2 µM Wnt-C59 (Tocris Bioscience) and incubated for 48 hours. On day 5, the medium was replaced again with RPMI 1640 supplemented with B-27 minus insulin. From day 7 onwards, cells were cultured in RPMI 1640 medium supplemented with B-27 with insulin until the end of protocol. Metabolic selection of hiPSC-CMs was conducted upon the observation of beating cells, between days 10 and 12. During this period, cells were cultured with glucose-free RPMI 1640 medium supplemented with B-27 with insulin (Gibco) for 48 hours. When necessary, the medium was renewed for an additional 2 days to ensure optimal purification of the CMs.

RNA isolation, Microarray preparation and analysis

Total RNA was isolated using the RNeasy Micro Kit (Qiagen). The RNA samples for whole transcriptome expression analysis were enzymatically fragmented and biotinylated using GeneChip^TM Whole Transcript Expression Arrays (Applied Biosystems). The samples were then hybridized with the GeneChip^TM WT PLUS Reagent Kit (Applied Biosystems) and scanned using the GeneTitan™ Microarray System (Applied Biosystems). The entire procedure for sample processing, running, and scanning followed the manufacturer’s protocol. Raw expression data for the D4-15 IRX3^cl1+/+sample was subjected to reanalysis from previously published data by our group (GSE188749) [55]. Expression levels of the IRX3^cl1.1-/- were assessed in the same microarray chip (GSE262776). Raw intensity values were background-corrected, normalized, and log2-transformed using the Robust Multi-array Average (RMA) algorithm. Differentially expressed genes (DEGs) were detected using limma R package [65,66]. Genes with logFC >= 0.5 or <= −0.5 and FDR < 0.01 were considered differentially expressed.

RT-qPCR and relative gene expression analysis

Total cellular RNA was extracted with TRIzol^TM Reagent (Invitrogen, Thermo Fisher Scientific) followed by column purification with the RNeasy Micro kit (Qiagen), according to the manufacturer’s instructions. Reverse transcription was performed using 500 ng of total RNA with the SuperScript™ IV First-Strand Synthesis System (Invitrogen, Thermo Fisher Scientific). Quantitative real-time PCR was performed using TaqMan™ Universal PCR Master Mix (Invitrogen, Thermo Fisher Scientific) or QuantiTect SYBR Green PCR Kit (Qiagen). Relative gene expression was determined using the 2^−ΔΔCt method, with GAPDH or ACTB serving as endogenous controls (see S4 Table for primer sequences).

Fluorescence microscopy

hiPSC-CMs were plated at 1 x 10⁴ cells/cm² in 96-well fluorescence plates (Wells Greiner Microplate, Black, Clear Bottom), previously treated with Geltrex. Cells were fixed in 4% PFA, permeabilized with 0.5% Triton X-100/PBS and blocked with 5% BSA for 1 hour. Immunostaining was performed using anti-Troponin I cardiac (1:400, Hytest), anti-KI67 (1:200, Abcam), OCT4 (1:200, Cell Signaling), NANOG (1:200, Cell Signaling) and SOX2 (1:200, Cell Signaling), followed by appropriate incubation with secondary antibodies conjugated with Alexa Fluor 555/647 fluorophores (1:500, ThermoFisher Scientific). DAPI (1 mg/ml, ThermoFisher Scientific) was used to counterstain nuclei. For the high-content screening analysis, 16 up to 64 fields of view/well were collected using EVOS M7000 automated Imaging System (Invitrogen, Thermo Fisher Scientific) or IN Cell Analyzer 2000 (GE Healthcare) at either 20× or 40 × magnification.

After the acquisition, the raw images were analyzed using CellProfiler (4.2.0) [67]. The high-throughput analysis retrieved phenotypic information of more than 2000 single-cells per sample, for cell-based analysis, or 150–500 images for image-based analysis. Nuclear mean fluorescence intensity (MFI) of NKX2.5 and GATA4 were extracted from each individual nucleus within the images after background subtraction correction. For detecting replicating cells, 5 µM 5-ethynyl-20 deoxyuridine (EdU, ThermoFisher Scientific) was added to the cell medium 48 hours. Before incubation with fluorescent secondary antibodies, the cells were processed using the Click-iT™ EdU Alexa Fluor™ 647 HCS Assay (Invitrogen, Thermo Fisher Scientific) according to the manufacturer’s instructions. The EdU proliferative rate was calculated by counting the number of positive-staining cells divided by the total number of DAPI-positive cells multiplied by 100.

Confocal microscopy

On day 27, hiPSC-CMs were plated to confluence at 3 x 10⁴ cells/cm² in CellView Culture Slide (Greiner BioOne), previously treated with Geltrex. Cells were fixed with 4% PFA, permeabilized with 0.5% Triton X-100/PBS and blocked with 5% BSA for 1 hour. Immunostaining was performed using anti-TOM20 (1:200, Santa Cruz Biotechnology), Alpha-Actinin (Sarcomeric) (1:200, Sigma), GATA4 (1:200, Santa Cruz Biotechnology), NKX2–5 (1:200, Cell Signaling), Connexin 43 (1:200, Abcam), followed by incubation with appropriate secondary antibodies conjugated with Alexa Fluor 555/647 fluorophore (1:500, ThermoFisher Scientific). DAPI (1 mg/ml) was used to counterstain nuclei. Images were acquired on an Axio Observer.Z1 (Carl Zeiss) equipped with a Confocal Spinning Disk Unit (CSU-X1) (Yokogawa Life Science) and a Rolera EM-C2 EMCCD camera (Teledyne Photometrics). DAPI, TnI, and TOM20 were, respectively, detected by sequentially illuminating samples with 405 nm, 488 nm, and 561 nm laser lines through a C-Apochromat 40 × /1.20 W Korr objective (Carl Zeiss). Image analysis was performed in CellProfiler after preprocessing in Fiji Software (NIH) [68]. Briefly, “Subtract Background” filter plugin (rolling ball radius = 5) was applied to TOM20 images to reduce background and highlight mitochondria. Preprocessed TOM20 images were then processed in CellProfiler to extract mean TOM20 intensity and distribution in segmented mitochondria. For Connexin-43 quantification, CX43-positive and TnI-positive areas were segmented using the ‘Threshold’ module. The ratio of CX43 area to cardiomyocyte area was then calculated and multiplied by 100 (i.e., CX43 area / TnI-positive area × 100). The resulting values were converted to fold change to facilitate comparison across groups. Representative images were further processed by using the auto function of “Brightness/Contrast” plugin in Fiji, in order to highlight subcellular localization of proteins of interest. We measured the MFI of each protein of interest and, when applicable, cell area. Violin plots represent data from several hundreds of cells.

Scrape loading dye transfer assay (SLDT)

We performed the dye transfer assay to assess intercellular communication through gap junctions. For this, the cells were seeded at 1 x 10⁴ density on a 96-well plate, previously treated with Geltrex. Cells were washed with Ca²⁺/Mg²⁺-PBS (Gibco) and incubated with a solution prepared in Ca²⁺/Mg²⁺-PBS containing Lucifer Yellow dye (1 mg/ml, Sigma-Aldrich), propidium iodide (10 μg/ml, Sigma-Aldrich) and Hoechst 33342 (50 μg/ml, Invitrogen). After addition, a cut was made with a stainless-steel blade of approximately 4 mm specially adapted in-house for this purpose. After 5–40 minutes of incubation, cells were washed again with Ca²⁺/Mg²⁺-PBS to remove background fluorescence, and subsequently fixed with 4% PFA solution in PBS. Then, the images centering the cut were acquired using the EVOS M7000 Imaging System and processed and analyzed using CellProfiler software.

Morphology and contractile structures with MorphoScript

To characterize the hiPSC-CMs morphology as well as the organization of their contractile structures, hiPSC-CMs were dissociated with Trypsin-EDTA 0.25% (Gibco) for 5 min at 37 °C and replated on a 96-well plate with 10 μM ROCK Inhibitor Y-27632 added to the maintenance medium. Low confluence was used to obtain isolated cells (0.7 x 10⁴ cells/well). The cells were then fixed with 4% PFA for 15 min at room temperature, and permeabilized with 0.1% Triton X-100 for 5 min. The blocking was performed with 5% BSA for 60 min at room temperature. Alpha-Actinin (Sarcomeric) (1:200, Sigma-Aldrich) diluted in 2% BSA was used and incubated overnight at 4 °C. After incubation with primary antibody, cells were incubated with appropriate secondary antibody conjugated with Alexa Fluor 555 fluorophore (1:500, ThermoFisher Scientific) and DAPI (1 mg/ml) for 1 hour at room temperature.

The images were acquired using the EVOS M7000 Imaging System at × 40 magnification and analyzed using the MorphoScript Software in MATLAB, as previously described [69]. The isolated cells at the periphery were segmented manually. For the analysis, we set the window size to 15 so each window contains at least five contractile structures labeled with alpha-actinin. For the rest of the parameters, the default values were kept. This analysis allowed us to describe the morphology of hiPSC-CMs through the measurement of the sarcomeric organization (the order), spacing, alignment (the dispersion) and geometric parameters such as the cell surface area.

Contractile properties measurement

For the measurement of contractile properties, hiPSC-CMs were plated in 96-well plates and allowed to beat spontaneously. Recordings were made using the EVOS M7000 Imaging System equipped with a thermostatic chamber (5% CO₂, 37 °C) and 20 × objective. Before any recordings, the cells were allowed to stabilize in the chamber for at least 15 minutes. Spontaneously beating hiPSC-CMs were recorded for 15–20 seconds per field of view, with the imaging system set to a capture rate of 30 frames per second. Contractility analysis was conducted using ContractionWave Software, an open-source tool designed for large-scale analysis of CM contraction properties [70]. The recorded videos were imported into the Software, and parameters related to time, speed, area and frequency of contraction were extracted as previously described [70].

Calcium transient measurement

We employed two different calcium-sensitive fluorescent dyes to analyze calcium transients. For the single-wavelength calcium indicator method, IRX3^cl2+/+ and IRX3^cl2-/- hiPSC-CMs were seeded at a density of 3 x 10⁴ cells on a 96-well plate. Five days after plating, cells were loaded with the Fluo-4 NW Calcium Assay Kit (Molecular Probes) for 45 min at 37 °C. Following incubation, cells were washed and underwent a 5-minute de-esterification process with Tyrode’s solution (NaCl [135 mM], KCl [5.4 mM], MgCl [1.0 mM], HEPES [10 mM], CaCl₂ [1.8 mM], glucose [5.0 mM] [pH 7.4]). Fluorescent signals were recorded from the center of the plates using the EVOS M7000 Imaging System. Excitation was performed at 488 nm and emission fluorescence signals were collected at 516 nm for a duration of 20 s.

The recorded frames were imported into Fiji Software for analysis. After applying filtering, calcium transients from fluorescence signals were obtained. Calcium transients for each sample were then blinded and imported into pClamp software (version 11.2, Molecular Devices) for further analysis. Following baseline correction, the period containing each individual calcium transient was identified in pClamp software. The software’s built-in algorithms were used to calculate the parameters. The amplitude of the calcium transient was measured as the difference between the peak fluorescence intensity and the baseline fluorescence intensity. The time of decay was calculated by fitting an exponential function to the decay phase of the signal for 50% and 90% decay. To account for variations in spontaneous beating rates, calcium transient durations (CaD₅₀ and CaD₉₀) were normalized using Fridericia’s formula (CaDc = CaD / RR^1/3, where RR is the transient interval). The maximum rise slope was determined as the highest value of the first derivative during the rise phase. The average value of each parameter for each sample was then calculated, and the results were plotted and compared between the two groups.

For the dual-wavelength measurement, IRX3^cl1+/+ and IRX3^cl1-/- hiPSC-CMs were plated on laminin-coated glass coverslips. Seven days after plating, the cells were loaded with Fura-2, AM (ThermoFisher Scientific) diluted in Tyrode’s free Ca²⁺ solution (final concentration of 2 μM) and incubated for 10 min at room temperature. Next, cells were incubated for an additional 5 min at 37 °C and washed once with Tyrode free Ca²⁺ to exclude the contribution of trans-sarcolemma Ca²⁺ influx via voltage gated Ca²⁺ channels. Prior to measurements, cells were incubated for 5 min to enable complete de-esterification of intracellular Fura-2. Intracellular Ca²⁺ events were recorded using a 40 × objective on an Olympus IX70 microscope fitted with an IonOptix system (Olympus; IonOptix) at 37 °C. Cells were alternately excited at 340 nm and 380 nm, and the emitted fluorescence was collected at 510 nm. The 340/380 nm fluorescence ratio was used to quantify intracellular calcium levels. The recorded fluorescence ratios were analyzed using IonWizard software to determine calcium transient parameters, including amplitude, calcium decay and max slope for each sample.

Membrane potential

Electrical activity was recorded using the FluoVolt™ Membrane Potential Kit (Molecular Probes). Fluovolt dye was diluted in Tyrode’s solution pre-warmed to 37 °C containing PowerLoad (a solubilizing agent provided in the kit) and Blebbistatin (100 μM, Sigma-Aldrich) to prevent contraction artifacts. Cells were incubated with the solution for 10 min at room temperature, followed by an additional 5 min at 37 °C. After incubation, the cells were washed once with Tyrode’s solution to remove excess dye. Fluorescence intensities were measured using a multi-mode microplate reader (SpectraMax iD3, Molecular Devices) with excitation and emission wavelengths set at 480 nm and 540 nm, respectively. To capture electrical signals with maximal time resolution, a minimal sampling interval of 10 ms was used. All experiments were conducted at a controlled temperature of 37 °C to maintain physiological conditions. The acquired fluorescence signals were analyzed using pClamp software (version 11.2, Molecular Devices) to determine the average action potential duration for each sample.

Single-cell ATAC-seq and ChIP-seq (in silico) analysis

scATAC-seq data (GEO ID: GSE181346) [71] and ChIP-seq data (GEO ID: GSE159411) [72] were obtained from the Gene Expression Omnibus (GEO) database. The scATAC-seq samples consisted of CMs derived from induced pluripotent stem cells (iPSCs) after 5 days of differentiation. ChIP-seq samples were collected after 6 days of induced differentiation from iPSCs to CMs, targeting GATA4, NKX2–5, and TBX5 transcription factors. scATAC-seq profiles were aggregated across cells and analyzed as pseudo-bulk data: using the preprocessed BED files to identify open chromatin genomic regions, which were further annotated using the ChIPseeker package in R [73]. Promoter regions were defined as ±1 kb around the transcription start site (TSS) of the nearest gene. This pseudo-bulk strategy for scATAC-seq peak calling and annotation is similar to approaches used previously to derive consensus peak sets from aggregated scATAC-seq datasets [74]. Next, IRX3 motifs within scATAC-seq data were identified using motifs obtained from the “encode_pwms” object in the chromVARmotifs package [75]. Specifically, we utilized motifs derived from [76] (core sequence: ACATGT) and [77] (core sequence: ACATGA) to capture both canonical and variant binding specificities (visualized in S7 Fig). Overlapping genomic regions between scATAC-seq peaks and predicted IRX3 motifs were calculated using the matchMotifs function from the motifmatchr package [75]. The output was processed with the “reduce” function from the IRanges package [78] to merge overlapping or adjacent genomic regions.

After identifying IRX3 motifs, ChIP-seq data analysis was performed. For each ChIP target (GATA4, NKX2–5, TBX5), overlapping peaks from downloaded BED or BB files were merged using the ChIPseeker package [73]. The data, initially in human genome version hg19, was lifted over to hg38 using the liftOver function from the rtracklayer package [73,78]. The resulting peaks were annotated. To identify genomic regions with open chromatin and IRX3 motifs for each ChIP target, we used the function findOverlapOfPeaks from the package ChIPpeakAnno [79]. The inputs included the ChiP targets positive regions and the scATAC-seq + IRX3 Motif genomic intervals. This analysis takes multiple sets of genomic intervals (peaks) as input and computes their overlaps, returning both pairwise and higher-order intersections. Using the default overlap criteria (minimum 1 bp overlap within the same genome build), findOverlapsOfPeaks extracted the subset of genomic regions where all four peak sets intersected, which we defined as co-localized regions of open chromatin, IRX3 motif presence, and TF binding. These co-localized regions were then carried forward for gene assignment and downstream ontology enrichment analyses.

We repeated this process for NKX2–5 and TBX5 ChIP data. The findOverlapOfPeaks function was then applied again, with the overlapping regions from the previous analysis as input. The resulting overlapping regions were then extracted and annotated. For data visualization, Venn diagrams were generated using the gplot package to illustrate the overlap between different sets of genomic regions. Lastly, Gene Ontology enrichment analysis was performed using the clusterProfiler package [80]. Genes annotated for each transcription factor, the intersection between them, and all targets were selected for Biological Process and Molecular Function enrichment analyses. The OrgDb parameter was set ‌‌for both analyses to “org.Hs.e.g.,db.”

Transmission electron microscopy

Cells were grown to 90% confluence before being washed, trypsinized, pelleted, and fixed in 2.5% glutaraldehyde in 0.1 M sodium cacodylate buffer (pH 7.2) overnight. Samples were then post fixed in 1% osmium tetroxide for 2 h, stained in bloc with 0.5% uranyl acetate, rinsed, and dehydrated in graded ethanol. After immersion in propylene oxide, samples were embedded in epoxy resin (Spurr, Electron Microscopy Sciences, EMS, Hatfield PA, USA) and polymerized for 42 hours at 75 °C. Ultrathin sections were stained with lead citrate and uranyl acetate and TEM images were collected with a JEM-1400 Plus microscope (JEOL, Japan) applying tension of 120 kV and using camera OneView 4K x 4K (Gatan, USA). Mitochondria segmentation was manually performed in Fiji software by an observer who was blind to the experimental conditions.

Statistical analysis

Data were derived from 3–6 independent replicates and are presented as means ± standard deviation (SD), box plot showing the median, interquartile range (25^th–75^th percentiles), or violin plot ranging from the 1st through the 3rd quantile. One-way ANOVA followed by Tukey’s multiple comparisons test was used to analyze data from Fig 1D-E and Fig 4A-B; two-way repeated measures ANOVA with Sidak’s correction for multiple comparisons was used for Figs 2L and 3D; and the Mann-Whitney test was used for Fig 2B (“Order” plot). Student’s t-test was applied for the remaining figures (Figs 1-5 and S1-S5 Figs), unless otherwise stated. All t-tests were two-tailed and paired. Statistical significance was considered achieved when p < 0.05. Further statistical details are provided in the figure legends.

Resource availability

Requests for further information and resources should be directed to and will be fulfilled by the lead contact, Jose Eduardo Krieger (j.krieger@hc.fm.usp.br).

Materials availability

This study did not generate new unique reagents.