Abstract

   Mitochondrial trifunctional protein deficiency, due to mutations in
   hydratase subunit A (HADHA), results in sudden infant death syndrome
   with no cure. To reveal the disease etiology, we generated stem
   cell-derived cardiomyocytes from HADHA-deficient hiPSCs and accelerated
   their maturation via an engineered microRNA maturation cocktail that
   upregulated the epigenetic regulator, HOPX.  Here we report, matured
   HADHA mutant cardiomyocytes treated with an endogenous mixture of fatty
   acids manifest the disease phenotype: defective calcium dynamics and
   repolarization kinetics which results in a pro-arrhythmic state. Single
   cell RNA-seq reveals a cardiomyocyte developmental intermediate, based
   on metabolic gene expression. This intermediate gives rise to
   mature-like cardiomyocytes in control cells but, mutant cells
   transition to a pathological state with reduced fatty acid
   beta-oxidation, reduced mitochondrial proton gradient, disrupted
   cristae structure and defective cardiolipin remodeling. This study
   reveals that HADHA (tri-functional protein alpha), a
   monolysocardiolipin acyltransferase-like enzyme, is required for fatty
   acid beta-oxidation and cardiolipin remodeling, essential for
   functional mitochondria in human cardiomyocytes.

   Subject terms: Mechanisms of disease, Cardiovascular diseases
     __________________________________________________________________

   Mutations in the gene HADHA result in mitochondrial tri-functional
   protein (MTP) deficiency and can result in sudden infant death syndrome
   for which there is no treatment. Here the authors show that the MTP
   deficient pathology in human cardiomyocytes leads to an abnormal
   cardiolipin pattern and suggests that cardiolipin affecting compounds
   may serve as a potential therapy.

Introduction

   Mitochondrial trifunctional protein (MTP/TFP) deficiency is thought to
   be a result of impaired fatty acid oxidation (FAO) due to mutations in
   hydroxyacyl-CoA dehydrogenase/3-ketoacyl-CoA thiolase/enoyl-CoA
   hydratase subunit A (HADHA HADHA/LCHAD) or subunit B (HADHB)^[65]1. A
   major phenotype of MTP-deficient newborns is sudden infant death
   syndrome (SIDS), which manifests after birth once the child begins
   nursing on lipid-rich breast-milk. Defects in FAO have a role in
   promoting a pro-arrhythmic cardiac environment; however, the exact
   mechanism of action is not understood, and there are no current
   therapies^[66]2,[67]3.

   Pluripotent stem cell derived cardiomyocytes (hPSC-CM) provide a means
   to study human disease in vitro but have limitations due to their
   immaturity as they are representative of fetal cardiomyocytes (FCM)
   instead of adult cardiomyocytes (ACM)^[68]4,[69]5. Due to the lack of
   knowledge in how committed cardiomyocytes transition from an immature
   FCM to a mature ACM, many cardiac diseases with postnatal onset are
   poorly characterized^[70]6–[71]11 During cardiogenesis, FCMs go through
   developmental states and once past cardiomyocyte commitment exhibit:
   exit of cell cycle, utilization of lactate, cessation of spontaneous
   beating, and then at the postnatal stage utilization of fatty acids and
   cardiolipin maturation^[72]12–[73]17. Since immature hPSC-CMs are
   unable to utilize fatty acids through FAO as an energy source, they are
   limited in their use to model FAO disorders.

   Current approaches to mature hPSC-CMs toward ACM focus on prolonged
   culture^[74]18, physically stimulating the cells with either
   electrical^[75]19 or mechanical stimulation^[76]20 or by 2D surface
   pattern cues to direct cell orientation^[77]21. An emerging area of
   hPSC-CM maturation is in manipulating microRNAs (miRs)^[78]22–[79]24.
   Overexpressing just one miR, Let-7, can accelerate human embryonic stem
   cell derived cardiomyocyte (hESC-CM) maturation towards an ACM-like
   state^[80]24. However, no maturation regimen has been able to mature
   hPSC-CMs to an adult state.

   In this study, we analyze mitochondrial trifunctional protein
   deficiency by generating stem cell derived cardiomyocytes from
   HADHA-deficient human induced pluripotent stem cells (hiPSCs) and
   accelerate their maturation by our engineered microRNA maturation
   cocktail. The data reveal the essential dual role of HADHA in fatty
   acid beta-oxidation and as an acyltransferase in cardiolipin remodeling
   for cardiac homeostasis.

Results

Generation of MTP deficient CMs

   To recapitulate the cardiac pathology of mitochondrial trifunctional
   protein deficiency on the cellular level in vitro (Fig. [81]1a), we
   used the CRISPR/Cas9 system to generate mutations in the gene HADHA of
   human iPSCs. From our wild-type (WT) hiPSC line, that serves as our
   isogenic control, we generated HADHA mutant hiPSC lines using two
   different guides targeting exon 1 of HADHA (Supplemental Fig. [82]1A,
   B). To identify the phenotypes specifically caused by mutations in
   HADHA and to control for potential background effects, we chose to
   study a knockout (KO) HADHA (HADHA^KO) and compound heterozygote
   (HADHA^Mut) hiPSC lines that were generated using gRNA1. We also
   utilized the hiPSC line HEL87.1, which was derived from a patient
   carrying the founder point mutation most common in mitochondrial
   trifunctional protein disorder, HADHA c.1528G>C (Supplemental
   Fig. [83]1C, D)^[84]25.

Fig. 1.

   [85]Fig. 1
   [86]Open in a new tab

   Generation of HADHA Mutant and Knockout stem cell derived
   cardiomyocytes. a Schematic of fatty acid beta-oxidation detailing the
   four enzymatic steps. b Schematic of HADHA KO DNA and protein sequence
   from WTC iPSC line showing a 22 bp deletion, which resulted in an early
   stop codon. c Schematic of HADHA Mut DNA and protein sequence from WTC
   iPSC line showing a 2 bp deletion and 9 bp insertion on the first
   allele and a 2 bp deletion on the second allele. RNA-Sequencing read
   counts show that the HADHA Mut expresses exons 4–20 resulting in a
   truncated protein. d Western analysis of HADHA expression and
   housekeeping protein β-Actin in WTC iPSCs. e Confocal microscopy of WT,
   HADHA Mut and HADHA KO hiPSC-CMs for the cardiac marker αActinin
   (green) and HADHA (red). f Seahorse analysis trace of fatty acid
   oxidation capacity of WT, HADHA Mut and HADHA KO hiPSC-CMs. n = 6–7
   biological replicates. Source data are provided as a Source Data file

   Examining the DNA sequence of the HADHA^KO line showed a homozygous
   22 bp deletion, which resulted in an early stop codon in exon 1
   (Fig. [87]1b and Supplemental Fig. [88]1E). The HADHA^Mut line had a
   2 bp deletion and 9 bp insertion on the first allele and a 2 bp
   insertion on the second allele (Fig. [89]1c and Supplemental
   Fig. [90]1F). Both lines showed no off-target mutations on the top
   three predicted sites (Supplemental Fig. [91]1G). The mutations found
   in the HADHA^Mut line resulted in a predicted early stop codon on both
   alleles (Supplemental Fig. [92]1H). However, when we examined the
   protein in each line we found that HADHA was expressed in the WT hiPSC
   line, not expressed in the HADHA^KO line and was still expressed, to a
   lower degree, in the HADHA^Mut line (Fig. [93]1d). We then examined the
   transcript of HADHA expressed in WT and HADHA^Mut lines. We found the
   WT line expressed the full length HADHA transcript from exon 1–20 while
   the HADHA^Mut line skipped exons 1–3 and expressed HADHA exons 4–20
   (Fig. [94]1c). It is possible that the mutations generated at the
   intron-exon junction induced an alternative splicing event and a new
   transcript since there is no known transcript of HADHA from exon 4–20
   (Supplemental Fig. [95]1I). The observed reduction in the HADHA mutant
   molecular weight (Fig. [96]1d) supports this hypothesis. The expressed
   HADHA^Mut protein skips the expression of exons 1–3, 60 amino acids,
   generating a truncated ClpP/crotonase domain, which likely compromises
   the mitochondrial localization and protein folding of the enzyme pocket
   resulting in the inability to stabilize enolate anion intermediates
   during FAO (Supplemental Fig. [97]1J).

   Using a monolayer directed differentiation protocols^[98]26,[99]27 we
   generated human induced pluripotent stem cell derived cardiomyocytes
   (hiPSC-CMs) from the WT and hiPSC lines with HADHA mutations. We found
   that the reduction or loss of HADHA did not hinder the ability to
   generate cardiomyocytes (Fig. [100]1e). However, we found that all CMs,
   even the control CMs, were unable to utilize long-chain FAs
   (Fig. [101]1f) and needed to be matured^[102]14,[103]24,[104]28.

Screening microRNAs for hPSC-CM maturation

   MicroRNAs have recently been shown to regulate the key, opposing
   processes of cardiomyocyte regeneration, maturation and
   dedifferentiation^[105]24,[106]29,[107]30. We cross-referenced in vivo
   miR-sequencing data of human fetal ventricular to adult ventricular
   myocardium^[108]24,[109]31,[110]32 and combined multiple miRs together
   with Let-7 to rapidly mature hPSC-CMs by promoting a more complete
   adult like transcriptome. Analyzing each miR’s predicted targets
   affecting glucose and/or fatty acid metabolism, cell growth and
   hypertrophy and cell cycle, six miRs were chosen to assess for their CM
   maturation potency: three upregulated miRs (miR-452, −208b^[111]33 and
   −378e^[112]34) and three downregulated miRs (miR-122, −200a, and −205)
   (Fig. [113]2a).

Fig. 2.

   [114]Fig. 2
   [115]Open in a new tab

   Cardiomyocyte maturation microRNA screen. a Schematic of the workflow
   performed to determine candidate microRNAs to screen for cardiomyocyte
   maturation. b Schematic of the workflow performed to generate microRNA
   transduced stem cell derived cardiomyocytes. c Cell area analysis of
   microRNA treated hiPSC-CMs. MicroRNA-208b OE lead to a significant
   increase in cell area while miR-205 KO led to a significant decrease.
   EV: 2891 μm^2, 208b: 5802 μm^2. *p < 0.05, one-way ANOVA on Ranks
   performed. n = 16–51 cells measured. d Micro-electrode array analysis
   of microRNA treated hiPSC-CMs corrected field potential duration
   (cFPD). MiR-452 OE led to a longer cFPD. EV: 296 ms, 452: 380 ms.
   n = 3–6 biological replicates. e Single cell twitch force analysis
   using a micropost assay. MiR-200a KO led to a significant increase in
   twitch force of hiPSC-CMs. EV: 30.8nN, miR-200a: 51.7nN. *p < 0.05,
   one-way ANOVA on Ranks performed. n = 12–41 cells measured. f Seahorse
   analysis of the maximum change in oxygen consumption rate (OCR) due to
   FCCP after oligomycin treatment of microRNA treated hiPSC-CMs. MiR-122
   KO led to a significant increase in maximum OCR while miR-208b OE,
   −378e OE and −200a KO led to significant decreases in maximum OCR.
   miR-122 KO: 1.35-fold change compared to EV. **p < 0.01, ***p < 0.001,
   one-way ANOVA performed. n = 3–17 biological replicates. Box plot
   middle line represents the median, x represents mean, bottom line of
   the box represents the median of the bottom half (1st quartile) and the
   top line of the box represents the median of the top half (3rd
   quartile). The whiskers extend from the ends of the box to the
   non-outlier minimum and maximum value. Source data are provided as a
   Source Data file

Functional analysis of candidate microRNAs

   These six miRs were assessed using four functional tests to determine
   hPSC-CM maturation: cell area, force of contraction, metabolic
   capacity, and electrophysiology. WT D15 hiPSC-CMs were transduced with
   a lentivirus to either OE a miR or KO a miR using CRISPR/Cas9. Cells
   were then lactate selected to enrich for the cardiomyocyte population
   and puromycin selected to enrich for the population containing the
   viral vector. Functional assessment was performed after two weeks of
   miR perturbation on D30 (Fig. [116]2b).

   An important feature of cardiomyocyte maturation is an increase in cell
   size. We found only miR-208b OE brought a significant increase in cell
   area (Fig. [117]2c and Supplemental Fig. [118]2A). Immature hPSC-CMs
   spontaneously beat at a high rate and have a short field potential
   duration when studied by extracellular micro-electrodes. Using
   micro-electrode arrays, we found only miR-452 OE increased the
   corrected field potential duration (cFPD) to a more adult like duration
   (Fig. [119]2d). One of the hallmarks of cardiomyocyte maturation is the
   increase in contractile force generated by the cell. We performed
   single cell force of contraction analysis using a micropost
   platform^[120]24,[121]35,[122]36 and found only the KO of miR-200a
   brought about a significant increase in force of contraction
   (Fig. [123]2e and Supplemental Fig. [124]2B). Finally, we assessed the
   metabolic capacity and found only the KO of miR-122 brought about a
   significant increase in maximum oxygen consumption rate (OCR)
   indicating more active mitochondria (Fig. [125]2f and Supplemental
   Fig. [126]2C).

Bioinformatic analysis of candidate microRNAs

   RNA-Sequencing was performed after alterations of some of the miRs
   (miR-378e OE, −208b OE, −452 OE, −122 KO, or −205 KO) to assess their
   global transcriptional impact in hPSC-CMs. In each sample, ~11,000
   protein-coding genes were expressed with an aggregated expression of at
   least three FPKM (fragments per kilobase of transcript per million
   mapped reads) across all samples were used for principle component
   analysis (PCA). PCA showed that each miR was able to bring a
   significant change from their respective controls (Supplemental
   Fig. [127]2D). Furthermore, since none of the miRs clustered with one
   another, each miR was capable of inducing a unique expression
   signature.

   Each miR’s function was then analyzed by specifically examining
   pathways that are essential for cardiac maturation. A pathway
   enrichment heatmap was generated showing how each miR influenced seven
   different pathways chosen as hallmarks of cardiomyocyte maturation
   (Supplemental Fig. [128]2E).

   From these data, we generated a MicroRNA Maturation Cocktail we termed
   MiMaC, consisting of: Let7i OE, miR-452 OE, miR-122 KO, and miR-200a
   KO. Let7i was chosen due to our previous study showing the potency of
   this miR to bring about cardiomyocyte maturation^[129]24. From each of
   the functional assays, we chose a miR that brought a significant
   increase in maturation to generate a cocktail that consisted of the
   smallest number of miRs.

Functional assessment of MiMaC

   To assess MiMaC treated hPSC-CM maturation we performed force of
   contraction, cell area, and metabolic assays (Fig. [130]3a). MiMaC
   treated hiPSC-CMs had a statistically significant increase in twitch
   force and power as compared to control cells (Fig. [131]3b–d). MiMaC
   treated hiPSC-CMs and hESC-CMs had a statistically significant increase
   in cell area (Fig. [132]3e, f and Supplemental Fig. [133]3A).
   Furthermore, both MiMaC-treated hESC-CMs and miMaC-treated hiPSC-CMs
   were able to utilize palmitate significantly greater than control CMs
   (Fig. [134]3g and Supplemental Fig. [135]3B).

Fig. 3.

   [136]Fig. 3
   [137]Open in a new tab

   MiMaC accelerates hiPSC-CM maturation. a Schematic of the four
   microRNAs combined to generate MiMaC. b Single cell force of
   contraction assay on micro-posts showed that MiMaC treated hiPSC-CMs
   led to a significant increase in twitch force. EV: 24nN, MiMaC: 36nN.
   **p < 0.01, t-test followed by a Mann-Whitney rank sum test. n = 40–54
   cells measured. c Representative trace of an EV (control) and a MiMaC
   treated hiPSC-CM. d Single cell force of contraction assay on
   micro-posts showed that MiMaC treated hiPSC-CMs led to a significant
   increase in power. EV: 22fW, MiMaC: 38fW. *p < 0.05, t-test followed by
   a Mann-Whitney rank sum test. n = 40–54 cells measured. e Cell size
   analysis showed that MiMaC treated hiPSC-CMs led to a significant
   increase in area. EV: 2389 μm^2, MiMaC: 3022 μm^2. ***p < 0.001, t-test
   followed by a Mann-Whitney rank sum test. n = 220–298 cells measured. f
   Representative confocal microscopy images of EV and MiMaC treated
   hiPSC-CMs. αActinin (green), phalloidin (red) and DAPI are shown. g
   Seahorse analysis of fatty acid oxidation capacity showed that MiMaC
   treated hiPSC-CMs matured to a point where they could oxidize palmitate
   for ATP generation while controls cells were not able to utilize
   palmitate. MiMaC hiPSC-CMs had a significant increase in OCR due to
   palmitate addition. *p < 0.05, t-test was performed. n = 8–9 biological
   replicates. Error bars are standard error. h Venn diagram of KO
   microRNA predicted targets and the identification of HOPX as a common
   predicted targeted between all KO miRs screened for cardiomyocyte
   maturation. i Plot of HOPX expression from RNA-Sequence data during
   cardiomyocyte maturation. HOPX expression is significantly higher in
   D30 and 1-year hESC-CMs and 1-year hESC-CMs have significantly higher
   HOPX as compared to D30 hESC-CMs. * denotes significance vs D20. #
   denotes significance vs D30. **p < 0.01 and *p < 0.05 are vs D20,
   #p < 0.05 is vs D30, one-way ANOVA was performed. n = 2–4 biological
   replicates. Error bars are standard error. j HOPX expression in adult
   human ventricle tissue is significantly higher than fetal human
   ventricular tissue. Plotted using RNA-sequencing data. *p < 0.05, a
   negative binomial test was used, n = 6 for fetal samples, n = 35 for
   adult samples. k RT-qPCR of HOPX expression showed that MiMaC treated
   hiPSC-CMs at D30 had a statistically significant higher level of HOPX
   as compared to EV control D30 hiPSC-CMs. **p < 0.01, t-test followed by
   a Mann-Whitney rank sum test. n = 5–6 biological replicates. l Single
   cell RNA-Seq tSNE plot of unbiased clustering of microRNA treated
   hPSC-CMs. m Cluster plot detailing which treatment groups are enriched
   in each cluster. n Heatmap of maturation categories based on MiMaC
   cluster. o Heatmap of in vivo human maturation markers that are
   upregulated with maturation (yellow). Box plot middle line represents
   the median, x represents mean, bottom line of the box represents the
   median of the bottom half (1st quartile) and the top line of the box
   represents the median of the top half (3rd quartile). The whiskers
   extend from the ends of the box to the non-outlier minimum and maximum
   value. Bar graphs show mean with standard error. Source data are
   provided as a Source Data file

Transcriptional assessment of MiMaC

   To gain a better understanding of how MiMaC was affecting the
   transcriptome of hiPSC-CMs we performed RNA-Sequencing comparing D30 EV
   control CMs to D30 MiMaC treated CMs. Pathway enrichment analysis using
   a hallmark gene set (Supplemental Table [138]S1) showed that many cell
   maturation and muscle processes were upregulated such as: myogenesis
   and epithelial mesenchymal transition^[139]37. The top downregulated
   pathways were associated with cell cycle, a key feature of
   cardiomyocyte maturation. Using STRING Analysis, we determined the
   network of significantly upregulated and interconnected genes
   associated with two pathways: myogenesis and epithelial mesenchymal
   transition (Supplemental Fig. [140]3C). STRING analysis was also used
   to show that the significantly downregulated and interconnected genes
   were associated with repression of cell cycle, specifically, the
   mitotic spindle and G2M checkpoint (Supplemental Fig. [141]3D). These
   findings show that the MiMaC tool promotes a more mature transcriptome
   in hiPSC-CMs.

HOPX is a regulator of CM maturation

   To better understand the molecular mechanisms that are critical for
   cardiac maturation, the overlapping predicted targets of the screened
   six miRs were determined. We had previously studied the predicted
   targets of Let7, insulin receptor-pathway and polycomb repressive
   complex 2 function, and their role in cardiomyocyte maturation^[142]24.
   We now found that all four downregulated miRs during maturation had
   five common predicted targets. One of these predicted targets, HOPX
   (Fig. [143]3h), is important for cardiomyoblast specification^[144]38.
   Furthermore, we have recently shown that HOPX is involved in
   cardiomyocyte maturation^[145]39. However, the regulation of HOPX
   expression and mechanism of HOPX action during maturation are not
   understood. We found HOPX expression was upregulated in vitro, in vivo
   and in MiMaC treated hiPSC-CMs (Fig. [146]3i-k). We also found HOPX was
   upregulated 6.8-fold in D30 miR-122 KO hiPSC-CMs while Let7i OE matured
   hiPSC-CMs had no effect on HOPX expression (Supplemental Fig. [147]3E).
   These data suggest that Let7i OE maturation does not govern HOPX
   cardiac maturation pathways. This highlights the necessity of combining
   multiple miRs together to generate a robust maturation effect in
   hPSC-CMs.

   We previously assessed the role of HOPX OE in cardiomyocyte maturation
   and found that HOPX OE led to an increase in CM size^[148]39. Using
   STRING analysis, we found the differentially expressed genes associated
   with cell division in the HOPX OE group generated a
   highly-interconnected network with key cell cycle genes highly
   downregulated (Supplemental Fig. [149]4A). This recapitulated the cell
   cycle repression we found during the in vitro CM maturation process
   (MiMaC treated hiPSC-CMs). We then generated four clusters using Kmeans
   clustering: regulation of mitotic cell cycle, cell division, inhibition
   of cilia and ubiquitin protein. Representative cell cycle genes, BUB1,
   MKI67, and CENPE, were downregulated while the inhibitor of many G1
   cyclin/cdk complexes, CDKN1C, was significantly upregulated in the HOPX
   OE condition (Supplemental Fig. [150]4B). These data suggest that HOPX
   OE mechanistically increases cell size by driving the exit from cell
   cycle and inducing cardiomyocyte hypertrophy.

HOPX regulates cell cycle via SRF genes

   HOPX is a homeodomain protein that does not bind DNA but rather is
   recruited to locations in the genome by serum response factor
   (SRF)^[151]40. HOPX in turn recruits histone deacetylase (HDAC) and
   removes acetylation marks resulting in the silencing of genes
   (Supplemental Fig. [152]4C). HOPX OE led to a significant
   down-regulation of 294 SRF targets (hypergeometric test p-value is
   1.31x10^−5) (Supplemental Fig. [153]4D). We validated using qPCR a
   known SRF target gene that should be repressed during cardiomyocyte
   maturation, natriuretic peptide precursor A (NPPA). After 2 weeks of
   HOPX OE, NPPA was significantly repressed, while cardiac troponin C, a
   non-SRF cardiac gene was unaffected by HOPX OE. The ventricular isoform
   of myosin light chain, MYL2, which is upregulated as ventricular
   cardiomyocytes mature, increased 1.87-fold in expression after two
   weeks of HOPX OE (Supplemental Fig. [154]4E).

   We determined the SRF target genes in common between HOPX OE vs. the
   negative control (NC) hiPSC-CMs and the human adult vs. fetal
   myocardium (ventricular myocardium) transitions. 76 SRF targets were
   common between the two groups and formed a significant group of genes
   (hypergeometric test p-value is 5.44x10^^−24) (Supplemental
   Fig. [155]4F) with a strong association for repression of cell cycle
   (Supplemental Fig. [156]4G). Using STRING analysis, we determined the
   network of connected genes out of the 76 genes in common and ran Kmeans
   clustering to generate four clusters (Supplemental Fig. [157]4H); two
   cell cycle clusters, DNA repair and muscle development gene clusters.
   The network of SRF regulated cell cycle genes that was in common
   between the HOPX OE line and adult cardiomyocytes (Supplemental
   Fig. [158]4I) showed genes associated with cell cycle with 7 of the 10
   genes associated with the spindle machinery. These data indicate that
   MiMaC acts through HOPX to repress SRF cell cycle targets.

scRNA-sequencing analysis of miR treated CM maturation

   Using single cell RNA-sequencing (scRNA-Seq), we utilized the MiMaC
   tool to provide further insight into the underlying mechanisms of
   cardiomyocyte maturation. We performed scRNA-Seq and unbiased
   clustering on five groups of miR treated CMs: EV, Let7i & miR-452 OE,
   miR-122 & −200a KO, MiMaC and MiMaC + FA. The enrichment of the miR
   perturbation was analyzed in the five identified clusters
   (Fig. [159]3l, m) using a Chi-square test. The EV group was enriched in
   clusters 0 and 3, Let7i and miR-452 OE group was enriched in clusters 0
   and 1, miR-122 and −200a KO group was enriched in clusters 0 and 3 and
   MiMaC and MiMaC + FA were enriched in clusters 1 and 2. Cluster 4
   mainly consisted of cells with poor read counts and was not analyzed
   further. Characterizing the cell fate in each subgroup showed the
   majority of cells were cardiomyocytes with a very small subset of cells
   in cluster 1 displaying fibroblast (ENC1, DCN, and THY1) and epicardial
   markers (WT1, TBX18) (Supplemental Fig. [160]5A).

   To rank which clusters had a higher degree of cardiomyocyte maturation
   we assessed the genes highly up- and downregulated in the MiMaC
   enriched cluster, cluster 2, compared to cardiac markers, oxidative
   phosphorylation genes (Fig. [161]3n and Supplemental
   Table [162]S2)^[163]41, and in vivo human cardiac maturation markers
   (Fig. [164]3o and Supplemental Table [165]S3). We found that cluster 2
   had high upregulation of genes associated with myofibril structural
   proteins and in vivo maturation markers (Fig. [166]3n, o; P < 2x10^−16,
   using linear mixed effects model). We also found the MiMaC-treated
   cells were the most mature (Supplemental Fig. [167]5B-E and
   Supplemental Table [168]S4). Based on these findings, we ranked each
   cluster from least mature to most mature as cluster: 0 < 1 < 3 < 2.
   Cluster 2, the most mature CM cluster enriched for the MiMaC treated
   CMs, showed the highest expression of HOPX, a gene that is upregulated
   in maturation and is the predicted target of the downregulated miRs in
   MiMaC (Supplemental Fig. [169]5F and Supplemental Table [170]S3).
   Importantly, these data indicate that the observed transcriptional
   maturation mirrors normal in vivo cardiomyocyte maturation
   (Fig. [171]3o).

   Finally, we assessed the addition of fatty acids with MiMaC to increase
   cardiomyocyte maturation. Three long-chain fatty acids, palmitate,
   linoleic and oleic acid were added to the basal cardiac media
   used^[172]42. We found the MiMaC + FA cells were enriched in cluster 2.
   Importantly, while some studies have shown lipotoxicity with some FAs,
   the analysis of our carefully optimized FA-treatment procedure showed
   no increase in transcripts indicative of apoptosis, revealing minimal
   lipotoxicity in this assay (Supplemental Table [173]S5)^[174]43. These
   data indicate that MiMaC was essential for a robust transcriptional
   maturation of hiPSC-CMs.

scRNA-Seq reveals an intermediate CM maturation stage

   After unbiased analysis of the miR treated CMs it was clear each miR
   combination resulted in enrichment of different states of CM
   maturation. Interestingly, cardiomyocyte cluster 1, enriched for Let7i
   and miR-452 OE, showed a robust upregulation of OXPHOS and Myc target
   genes but was not yet significantly increased in most cardiomyocyte
   maturation markers (Fig. [175]3n, o and Supplemental Fig. [176]5G,H;
   Supplemental Table [177]S6). Hence, treatment of Let7i and miR-452 OE
   created an intermediate maturity CM in which metabolic maturation was
   the leading force. These data suggest a possible intermediate stage is
   a necessary transition stage between a fetal like CM to a more mature
   CM, which requires transient up-regulation of OXPHOS genes.

HADHA-Deficient CMs display reduced mitochondrial function

   The generation of the MiMaC tool allowed us to study HADHA CM disease
   etiology. First, we assessed the maximum OCR of WT, HADHA Mut, and KO
   CMs. MiMaC treated WT CMs had a statistically significant increase in
   maximum OCR as compared to control cells (Fig. [178]4a, b).
   Interestingly, control and MiMaC-treated HADHA Mut CMs had maximum OCR
   similar to control WT-CMs while the HADHA KO CMs had depressed maximum
   OCR. These data suggest defective mitochondrial activity of HADHA Mut
   and KO CMs.

Fig. 4.

   [179]Fig. 4
   [180]Open in a new tab

   Fatty acid challenged HADHA Mut CMs displayed elevated cytosolic
   calcium levels leading to increased beat rate irregularities. a
   Seahorse mitostress assay to analyze maximum oxygen consumption rate
   after oligomycin and FCCP addition. MiMaC treated CMs showed a
   significant increase (2.2-fold change) in maximum OCR compared to
   control EV CMs. *p < 0.05, one-way ANOVA on ranks vs WT-CM EV was
   performed. n = 1–14 biological replicates. b Representative trace of
   the mitostress assay. c Seahorse analysis of fatty acid oxidation
   capacity showed that MiMaC treated hiPSC-CMs matured to a point where
   they could oxidize palmitate for ATP generation while controls cells
   were not able to utilize palmitate. MiMaC hiPSC-CMs had a significant
   increase in OCR due to palmitate addition. Both MiMaC treated Mut and
   KO hiPSC-CMs were unable to oxidize palmitate. *p < 0.05, one-way ANOVA
   on ranks vs WT-CM EV was performed. n = 1–14 biological replicates. d
   Representative trace of the change in fluorescence during calcium
   transient analysis. e Quantification of the maximum change in
   fluorescence during calcium transients. Mut CMs as compared to WT CMs
   after 12D of Glc + FA media treatment had a statistically significantly
   lower change in calcium. WT CM: 2.03, Mut CM: 1.55. ***p < 0.001,
   t-test was performed. n = 28–30 cells measured. f Quantification of the
   tau-decay constant. Mut CMs as compared to WT CMs after 12D of Glc + FA
   media treatment had a higher tau-decay constant. WT CM: 0.63 s, Mut CM:
   0.76 s. n = 28–30 cells measured. g Representative trace of the change
   in cell membrane potential during whole-cell patch clamp analysis. h
   Quantification of the APD90 in WT and HADHA Mut CMs after 12D of
   Glc + FA media. WT 541 ms, Mut 1068 ms. *p < 0.05, t-test was
   performed. n = 9–10 cells measured. i Quantification of the tau-decay
   constant in WT and HADHA Mut CMs after 12D of Glc + FA media.
   p = 0.066, t-test followed by a Mann-Whitney rank sum test. n = 9–10
   cells measured. j Time to wave duration 50% is significantly longer in
   Mut CMs as compared to WT CMs after 12D of Glc + FA media treatment.
   ***p < 0.001, t-test was performed. n = 18–36 cells measured. k
   Representative beat rate trace of Mut CM in Glc or Glc + FA media. l
   Quantification of the change in beat interval (ΔBI). Mut CMs in
   Glc + FA media as compared to Mut CMs in Glc media had a statistically
   significant higher ΔBI. *p < 0.05, t-test followed by a Mann-Whitney
   rank sum test. n = 13–16 cells measured. m Poincaré plot showing
   ellipses with a 95% confidence interval for each group. The more
   rounded ellipse of the Mut Glc + FA condition shows that these cells
   had a greater beat to beat instability as compared to Mut Glc CMs. Box
   plot middle line represents the median, x represents mean, bottom line
   of the box represents the median of the bottom half (1st quartile) and
   the top line of the box represents the median of the top half (3rd
   quartile). The whiskers extend from the ends of the box to the
   non-outlier minimum and maximum value. Source data are provided as a
   Source Data file

   We showed that only MiMaC treated WT CMs showed a statistically
   significant increase in oxygen consumption due to palmitate addition
   (Fig. [181]4c). WT control CMs along with control and MiMaC treated
   HADHA Mut and KO CMs were unable to utilize FAs. These data show that
   MiMaC treated CMs have the capacity to utilize long-chain FAs; however,
   MiMaC-treated HADHA Mut and KO CMs are unable to do so.

Abnormal calcium handling of HADHA Mut CMs

   MTP-deficient infants can present with sudden, initially unexplained
   death after birth^[182]44. It is possible that the stress of lipids,
   the main substrate for ATP production found in a mother’s breast-milk,
   is what precipitates the early infant death due to MTP deficiency. We
   chose to utilize a combination of three long-chain fatty acids
   supplemented to our base cardiac media which contains glucose (Glc + FA
   media): palmitate, oleic, and linoleic acid, since these FAs are the
   most abundant in the serum of breastfed human infants^[183]42,[184]45.
   Palmitate, as a fatty acid substrate, is one of the most abundant fatty
   acids circulating during the neonatal period, representing 36% of all
   long-chain free fatty acids^[185]46. While challenging CMs with FAs can
   lead to lipotoxicity, we have carefully developed a concentration and
   combination of three fatty acids that do not result in lipotoxicity
   (Fig. [186]3l and Supplemental Fig. [187]6A)^[188]42,[189]47,[190]48.
   Moreover, other groups in the field of hPSC-CM maturation have also
   found that carefully chosen and fully conjugated FAs stimulate aspects
   of CM maturation^[191]49.

   To better understand the way in which MTP-deficient CMs may be
   precipitating an arrhythmic state leading to SIDS, we measured calcium
   transients in our WT and HADHA Mut CMs (Fig. [192]4d and Supplemental
   Fig. [193]6B). We found, the fold change in calcium being cycled was
   significantly higher in WT CMs as compared to HADHA Mut CMs
   (Fig. [194]4e). This suggested calcium was being cycled from the
   cytosol and stored in an aberrant manner in HADHA Mut CMs. When
   examining the tau-decay constant, we found HADHA Mut CMs had a higher
   average value (Fig. [195]4f). This suggested the rate at which calcium
   was being pumped back into the sarco/endoplasmic reticulum was slower
   in the HADHA Mut CMs.

Beat rate abnormalities in HADHA Mut CMs

   Since HADHA Mut CMs cultured in Glc + FA media exhibited abnormal
   calcium cycling, we assessed whether or not these CMs also exhibited
   abnormal electrophysiology. Using single cell whole-cell patch clamp,
   we found there was a significant increase in action potential duration
   in the HADHA Mut CMs as compared to WT CMs (Fig. [196]4g, h). This
   elongation period is seen during the plateau phase of the action
   potential where calcium ions are opposing the change in voltage due to
   potassium ions. This extent of elongation is indicative of a
   pathological state and suggests calcium handling as a potential source
   of this abnormal action potential (Fig. [197]4d–f). Furthermore, as in
   the calcium handling data, the tau-decay constant was higher in HADHA
   Mut CMs as compared to WT CMs (Fig. [198]4i). There was no difference
   in resting membrane potential (Supplemental Fig. [199]6C). Since,
   elongations of action potentials have been shown to result in
   arrhythmic heart conditions, phase 2 re-entry^[200]50, our whole-cell
   patch clamp data suggest the HADHA Mut CMs may be in a pro-arrhythmic
   state suggesting that indeed, HADHA Mut CMs when challenged with FAs
   could result in SIDS due to abnormal calcium handling.

   To assess the electrophysiology of a monolayer of cardiomyocytes, as a
   syncytium of cardiomyocytes is more representative of “tissue level”
   myocardium, we determined membrane potential changes using a
   voltage-sensitive fluorescent dye, Fluovolt. We found that while HADHA
   Mut CMs had no change in the maximum change in voltage amplitude or the
   rate of depolarization (Fig. [201]4j and Supplemental Fig. [202]6D–H),
   significant differences were observed when examining repolarization
   rates. We found that the time to wave duration (WD) 50% (WD50) and 90%
   (WD90) were significantly longer in the HADHA Mut CMs as compared to WT
   CMs (Fig. [203]4j and Supplemental Fig. [204]6H). These data suggest
   that the HADHA Mut CMs, at a monolayer level, had impaired
   repolarization resulting in slightly longer action potentials. Both the
   single cell and monolayer electrophysiology data showed abnormal action
   potential durations, both of which could be caused by abnormal calcium
   handling found in HADHA CMs.

   We tracked the spontaneous beating of HADHA Mut CMs in the presence of
   FAs and found that the HADHA Mut CMs had a significantly higher beat
   interval (Fig. [205]4k; Supplemental Fig. [206]6I) and a significantly
   higher change in beat-to-beat interval (ΔBI) than controls
   (Fig. [207]4l). These data suggest that HADHA Mut CMs beat on average
   slower and the time between beats was more variable. Furthermore, we
   quantified the percentage of ΔBI that were greater than 250 ms and on
   average the HADHA Mut CMs had a higher percentage of potentially
   arrhythmic ΔBIs (Supplemental Fig. [208]6J)^[209]51. Finally, we
   generated a Poincaré plot with fitted ellipses (95% confidence
   interval) around each group’s beat interval data (Fig. [210]4m). A
   narrow and elongated ellipse suggested uniform beat intervals while a
   more rounded ellipse suggested beat rate abnormalities. Taking the
   ratio of the major to minor axis of each ellipse we found that the
   HADHA Mut Glc condition had a ratio of 4.36 while the HADHA Mut
   Glc + FA condition had a ratio of 3.12 indicating that the HADHA Mut
   Glc + FA condition had a more rounded ellipse, indicating a higher
   beat-to-beat variability. These data combined suggest that the FA
   treated HADHA Mut CMs enter a pro-arrhythmic state potentially due to
   abnormal calcium handling, which results in elongated action potentials
   and abnormal repolarization.

scRNA-sequencing identifies HADHA Mut CM subpopulations

   Single-cell RNA-Sequencing and unbiased clustering were performed to
   better understand how the HADHA Mut CM population was behaving when
   challenged with FAs. A tSNE plot detailing each of the sequenced cell
   groups showed a clear distinction between WT and HADHA Mut CMs, with a
   small but significant overlap (Fig. [211]5a). When performing unbiased
   clustering, six clusters were found: 0 HADHA Mut CMs non-replicating, 1
   an intermediate maturation population of WT and Mut CMs, 2 HADHA Mut
   CMs replicating, 3 healthy CMs, 4 fibroblast like population and, 5
   epicardial like population (Fig. [212]5b, c and Supplemental
   Fig. [213]7A).

Fig. 5.

   [214]Fig. 5
   [215]Open in a new tab

   scRNA-Seq reveals multiple disease states of fatty acid challenged
   HADHA Mut CMs. a Single cell RNA-sequencing tSNE plot of WT compared to
   HADHA Mut CMs shows a clear distinction between these two groups. Four
   conditions of D30 CMs: 6 days of FA treated MiMaC WT CM, 6 days of FA
   and SS-31 MiMaC WT CMs, 6 days of FA treated MiMaC HADHA Mut CMs and 6
   days of FA and SS-31 treated MiMaC HADHA Mut CMs. b Unbiased clustering
   revealed six unique groups. c Heatmap detailing the enrichment of
   conditions in each cluster. d Heatmap of maturation categories based on
   MiMaC cluster. e Heatmap of in vivo mouse maturation markers that are
   upregulated with maturation. f Confocal microscopy showing that HADHA
   Mut CMs have more nuclei than WT CMs. Blue—DAPI, green—ATP synthase
   beta subunit, and pink—Titin. Inset is of the nuclei shown in grey
   scale. g Histogram of the frequency of cells with either 1, 2, 3-, or 4
   or more nuclei. HADHA mutant CMs have a significant number of cells
   with three or more nuclei. ***p < 0.001, Chi-square test with three
   degrees of freedom performed. n = 150–225 cells measured. h
   Downregulated metabolic pathways in cluster 0 (non-replicating HADHA
   CMs) as compared to cluster 3 (WT CMs). i Downregulated metabolic
   pathways in cluster 2 (endoreplicating HADHA CMs) as compared to
   cluster 3 (WT CMs). j Upregulated metabolic pathways in cluster 2
   (endoreplicating HADHA CMs) as compared to cluster 3 (WT CMs).
   Metabolic bubble plot circle size is proportional to the statistical
   significance. The smaller the p-value, the larger circle. Adjusted
   p-value 0.01 used as cutoff. k Branching trajectory kinetics plot based
   on CellTrails clustering. l Heatmap detailing the enrichment of
   conditions in each state. Source data are provided as a Source Data
   file

   To assess the degree of maturation and disease state we categorized
   each cluster based on the key categories described above
   (Fig. [216]3n). Upregulated genes in cluster three were associated with
   myofibril assembly and striated muscle cell development. Interestingly,
   a subset of both WT and HADHA Mut CMs were identified in an
   intermediate CM maturation cluster, cluster 1, as described above
   (Figs. [217]3l, [218]5d). This cardiac population had a high
   up-regulation of OXPHOS and Myc target genes (Supplemental
   Fig. [219]7B). WT cells that further developed from this intermediate
   state were identified in the more mature CM state, cluster 3. HADHA Mut
   cells, however, entered two sequential pathological states of disease.
   We postulate that first, HADHA Mut cells lose many highly expressed and
   repressed cardiac markers along with cell cycle inhibitor CDKN1A, as
   seen in cluster 0 (Supplemental Fig. [220]7C). Finally, very diseased
   HADHA Mut CMs in cluster 2 up-regulated genes that should be highly
   repressed in mature CMs, and activated cell cycle genes (Fig. [221]5d,
   Supplemental Fig. [222]7C, D and Supplemental Table [223]S7). We
   benchmarked these stages of maturation and disease progression against
   in vivo mouse and human maturation markers and found a similar trend
   for maturation, disease progression and loss of cardiac identity
   (Fig. [224]5e and Supplemental Fig. [225]7E–G).

   Examining significantly changed hallmark pathways between HADHA Mut CM
   clusters and the WT CM cluster we found OXHPOS, cardiac processes and
   myogenesis, being depressed in the mutant cells (Supplemental
   Table [226]S8, [227]S9). Furthermore, while WT CMs showed strong
   expression of cell cycle repressor CDKN1A, both HADHA Mut CM
   populations lost this expression. Cluster 2, the replicating HADHA Mut
   CMs, had an upregulation of DNA replication, G2M checkpoint and mitotic
   spindle genes (Supplemental Table [228]S9). Moreover, genes that are
   expressed in replicating and/or endocycling cells such as MKI67 and
   RRM2 were expressed only in cluster 2 HADHA Mut CMs (Supplemental
   Fig. [229]7C). To address potential pathological outcomes of the
   abnormal cell cycle marker increase, we analyzed the number of nuclei
   per cell in HADHA Mut CMs. Importantly, we observed a significant
   increase of the nuclei per cell in HADHA Mut CMs as compared to WT CMs
   (Chi-square test p < 0.001) (Fig. [230]5f, g). The majority of WT CMs
   were mono- or bi-nucleated, which is the healthy state found in vivo
   for nuclei number in CMs^[231]52. However, the number of mono-nucleated
   HADHA Mut CMs was significantly reduced while the number of bi- and
   multi-nucleated HADHA Mut CMs were increased suggesting a pathological
   state in the HADHA Mut CMs^[232]53. These data support the surprisingly
   high cell cycle transcript expression we found in a subpopulation of
   HADHA Mut CMs (cluster 2), and suggest multiple stages of disease state
   in the HADHA Mut CMs.

   To ensure cell cycle was not the underlying difference between all
   clusters, we examined cell cycle genes in each cluster (Supplemental
   Fig. [233]7H). Unlike previous studies which found that the bias
   imposed on cluster differences was dictated by which state of the cell
   cycle the cells were in^[234]54, we found that only cluster 2
   (Fig. [235]5b and Supplemental Fig. [236]7H) showed upregulated cell
   cycle genes. We also re-processed the clustering data with the removal
   of cell cycle genes and all clusters remained, except for original
   cluster 2, high cell cycle HADHA Mut CMs (Supplemental Fig. [237]7I).
   These findings suggest that cell cycle is the underlying reason for
   cluster 2 phenotype, but not for the rest of the cell populations
   (Fig. [238]5a, b).

   Finally, we found two genes (BAX and MFN2) that were highly expressed
   in cluster 3, MiMaC cluster, but downregulated in cells defective for
   HADHA, cluster 0 (Supplemental Fig. [239]7J). These findings support a
   recent study showing that HADHA mutants have defects in mitochondrial
   fission and fusion machinery, specifically, they also found the gene
   MFN2 to be mis-regulated leading to punctate malfunctioning
   mitochondria^[240]55.We postulate three different states of pathology
   in HADHA Mut CMs challenged with FAs: intermediate
   state(cluster1)::non-replicating CM state(cluster 0)::replicating CM
   state(cluster 2^[241]56). Cluster 1 showed an intermediate state of CM
   maturity, characterized by elevated OXPHOS and Myc target genes
   (Supplemental Table [242]S10). Importantly, both WT and HADHA CMs are
   found in cluster 1, suggesting that the HADHA CMs only manifest
   pathological phenotypes that separate them from wild-type cells later
   in development, during the maturation process, similar to that seen in
   human development.

   We performed unbiased metabolic pathway analysis, screening 68
   metabolic pathways and found HADHA Mut CM clusters, 0 and 2, displayed
   reduced metabolic pathway gene expression in comparison to WT CM,
   cluster 3 (Fig. [243]5h, i). Specifically, OXPHOS was one of the most
   downregulated pathways followed by cholesterol metabolism and fatty
   acid oxidation. Interestingly, in cluster 2, there were two highly
   upregulated metabolic pathways: nucleotide interconversion and folate
   metabolism, two key metabolic processes involved in DNA synthesis
   (Fig. [244]5j). Since HADHA Mut CMs displayed a down-regulation of many
   metabolic pathways including fatty acid and OXPHOS genes, we next
   wanted to examine the mitochondria and myofibrils of these cells.

Predicted cell trajectory from healthy to diseased state

   Since we had identified two different disease states,
   replicating/endoreplicating or intermediate state, we wondered if we
   could find a trajectory upon which a healthy CM follows to a disease
   state in an unbiased manner. To do this, we used the CellTrails^[245]57
   method as it allows for branching. Based on CellTrails clustering, we
   found six different states with a clear distinction between mutant and
   WT CMs (Supplemental Figure [246]8A). Clusters 1 and 4 were identified
   as WT cells, while clusters 2, 3, 5, and 6 were identified as HADHA Mut
   CMs. To identify the intermediate states, we utilized two hallmark
   maturation genes, one that goes up with ventricular maturation (MYH7)
   and one that goes down with ventricular maturation (MYL7)
   (Fig. [247]5k, l; Supplemental Figure [248]8B). S4, enriched for WT
   CMs, and S3, enriched for HADHA Mut CMs, both had low expression of the
   maturation gene MYH7 and a high expression of the immature gene MYL7.
   S5, also enriched for HADHA Mut CMs, had a scattered expression of both
   genes, MYL7 and MYH7 suggesting further progression to a pathological
   state. These data suggest that S4 and S3 are less mature CMs,
   representing the previously identified intermediate CM that can
   transition into a normal mature CM (S1) or fall into a diseased state
   (S5).

   Finally, our kinetics model identified two branches as end states of
   the pathological HADHA Mut CMs, S6 and S2 (Supplemental Fig. [249]8C).
   These two branches identified the two different HADHA Mut states from
   our scRNA-Seq analysis, non-replicating (S2) and
   replicating/endocycling (S6) HADHA Mut CMs. Together, these kinetics
   data in an unbiased manner, identified WT CMs as one potential starting
   point that led, through an intermediate CM state that had not been able
   to mature to a matured WT CM, to the pathological non-replicating and
   replicating/endocycling HADHA Mut CMs.

Structural and mitochondrial deficiencies of HADHA Mut and KO CMs

   When HADHA Mut and KO CMs were cultured in glucose-media alone, no
   obvious defects were observed in HADHA Mut and KO compared to the WT
   CMs (Supplemental Figure [250]9A,B). However, when cultured 6–12 days
   in FA media, sarcomere and mitochondrial defects manifested in the
   HADHA Mut and KO CMs, while the WT CMs appeared normal (Fig. [251]6a,
   Supplemental Fig. [252]9C). After 12D of Glc + FA media treatment, WT
   CMs had healthy myofibrils while the HADHA Mut CMs showed sarcomere
   dissolution, as α-actinin staining became punctate and actin filaments
   were difficult to detect (Fig. [253]6a). We assessed mitochondrial
   health since the HADHA Mut and KO CMs were unable to process long-chain
   FAs by first staining for mitochondrial ATP synthase beta subunit to
   examine the presence of a mitochondrial network. We found that both the
   WT and HADHA Mut CMs had many connected mitochondria while the KO CMs,
   at 6D FA, had lost their mitochondrial network to small, more circular
   mitochondria. To assess the functionality of these mitochondria, we
   analyzed the mitochondrial proton gradient via mitotracker orange
   staining. After 12-days of Glc + FA rich media, HADHA Mut CMs had
   highly depressed mitochondrial membrane proton gradient (Fig. [254]6a,
   b). Using a mitochondrial calcium sequestering dye, Rhod2, we examined
   the relative fluorescence intensity in the mitochondria of WT and HADHA
   Mut CMs and found a significant decrease in colocalization and
   intensity of the calcium indicator dye in the HADHA Mut CMs treated for
   12-days of glucose and fatty acid medium as compared to the WT CMs
   (Fig. [255]6c and Supplemental Fig. [256]9D).

Fig. 6.

   [257]Fig. 6
   [258]Open in a new tab

   Fatty acid challenged HADHA Mut CMs displayed swollen mitochondria with
   severe mitochondrial dysfunction. a Representative confocal images of
   WT and Mut CMs in 12D of Glc + FA media. b Quantification of
   mitotracker and ATP synthase β colocalization and intensity.
   ***p < 0.001, t-test followed by a Mann-Whitney rank sum test. n = 40
   cells per group measured. c Quantification of Rhod-2 and ATP synthase β
   colocalization and intensity. ***p < 0.001, t-test followed by a
   Mann-Whitney rank sum test. n = 53–66 cells measured. d Transmission
   electron microscopy images of WT and Mut CMs after 12D of Glc + FA
   media showing sarcomere and mitochondria structure. e Histogram of
   mitochondria circularity index for WT and HADHA Mut CMs after 12 days
   of Glc + FA media showed HADHA Mut CMs mitochondria are rounder.
   n = 81–251 mitochondria measured. f Histogram of mitochondria area for
   WT and HADHA Mut CMs after 12 days of Glc + FA media showed HADHA Mut
   CMs mitochondria are smaller. g Quantification of maximum OCR from
   mitostress assay. Mut and KO CMs as compared to WT CMs after 12D of
   Glc + FA media had a significantly lower max OCR. WT CM: 359
   pmoles/min/cell, Mut CM: 190 pmoles/min/cell, KO CM: 125
   pmoles/min/cell. ***p < 0.001, one-way ANOVA was performed vs WT 12D
   Glc + FA. n = 6–19 biological replicates. h Quantification of ATP
   production from mitostress assay, calculated as the difference between
   baseline OCR and OCR after oligomycin. Mut and KO CMs as compared to WT
   CMs after 12D of Glc + FA media had significantly lower ATP production.
   WT CM: 93 pmoles/min/cell, Mut CM: 51 pmoles/min/cell, KO CM: 43
   pmoles/min/cell. ***p < 0.001, one-way ANOVA was performed vs WT 12D
   Glc + FA. n = 6–18 biological replicates. i Quantification of proton
   leak from mitostress assay, calculated as the difference between OCR
   after oligomycin and OCR after antimycin & rotenone. Mut and KO CMs as
   compared to WT CMs after 12D of Glc + FA media had significantly higher
   proton leak. SS-31 treated Mut CMs after 12D of Glc + FA had a
   significantly lower proton leak and non-treated Mut CMs. WT CM: 3.64
   pmoles/min/cell, Mut CM: 7.66 pmoles/min/cell, KO CM: 10.52
   pmoles/min/cell. **p < 0.01, ***p < 0.001, one-way ANOVA was performed
   vs WT 12D Glc + FA. n = 4–19 biological replicates. j Schematic of
   patient harboring the point mutation, c.1528G > C, in the gene HADHA
   and the process of obtaining patient skin cells, reprograming them into
   iPSCs and then differentiating them into iPSC-CMs. k Representative
   confocal images of HADHA c.1528G > C CMs treated for 12-days of Glc or
   Glc + FA media. l Quantification of mitotracker and ATP synthase β
   colocalization and intensity. ***p < 0.001, t-test followed by a
   Mann-Whitney rank sum test. n = 31–35 cells measured. Box plot middle
   line represents the median, x represents mean, bottom line of the box
   represent the median of the bottom half (1st quartile) and the top line
   of the box represents the median of the top half (3rd quartile). The
   whiskers extend from the ends of the box to the non-outlier minimum and
   maximum value. Source data are provided as a Source Data file

   To better assess the sarcomere and mitochondrial disease phenotype we
   performed transmission electron microscopy (TEM) on WT and HADHA Mut
   CMs after 12D of Glc + FA exposure (Fig. [259]6d). WT CMs showed
   abundant myofibrils, clear Z bands but indistinct A-bands and I-bands,
   and no M-lines, indicating an intermediate, normal stage of CM
   myofibrillogenesis. Furthermore, WT CMs showed healthy mitochondria
   with good cristae formation. In contrast, HADHA Mut CMs showed poor
   myofibrils with a disruption of Z-disk structure replaced by punctate
   Z-bodies^[260]58 and disassembled myofilaments in the cytoplasm.
   Interestingly, HADHA Mut CM mitochondria were small and swollen with
   very rudimentary cristae morphology (Fig. [261]6d)^[262]59. Quantifying
   the WT and HADHA Mut CM mitochondria revealed HADHA Mut mitochondria
   were smaller in area and more rounded as compared to WT mitochondria
   (Fig. [263]6e, f and Supplemental Fig. [264]9E, F). Finally, examining
   complex I–V proteins showed that HADHA Mut CMs had depressed complex
   I–IV protein expression in Glc + FA conditions (Supplemental
   Figure [265]9G). These data show HADHA, but not control CMs lose
   sarcomere structure, mitochondrial membrane potential and calcium
   homeostasis and morphology when exposed to FAs.

SS-31 rescues aberrant proton leak in HADHA Mut CMs

   To better understand the pathological state of HADHA Mut and KO CMs
   exposed to FAs we functionally assessed their mitochondria. We found
   that the maximum OCR of Glc + FA treated HADHA Mut and KO CMs were
   significantly depressed as compared to WT cells (Fig. [266]6g).
   Furthermore, HADHA Mut CMs displayed reduced oxygen-dependent ATP
   production (Fig. [267]6h) and reduced glycolytic capacity (Supplemental
   Fig. [268]6H). Since exposure to FAs led to a reduction in
   mitochondrial membrane potential and reduced ATP production, we
   postulated that this may be due in part to an increased proton
   leak^[269]60,[270]61. We found HADHA Mut and KO CMs had a significantly
   higher proton leak than WT CMs (Fig. [271]6i). Previous studies have
   revealed that elamipretide (SS-31), a mitochondrial-targeted peptide
   can prevent mitochondrial depolarization, the proton leak^[272]62.
   Interestingly, a 1 nM treatment of HADHA Mut cardiomyocytes with
   elamipretide (SS-31) rescued the increased proton leak in Glc + FA
   challenged Mut CMs (Fig. [273]6i). These data suggest that HADHA Mut
   and KO CMs exposed to FAs resulted in reduced mitochondrial capacity
   due in part to increased proton leak.

HADHA c.1528G > C CMs display structural and mitochondrial abnormalities

   To test if observed HADHA Mut and KO CM disease pathology resembled the
   human clinical disease and to ensure the phenotype was not due to
   off-target gRNA mutations, we analyzed a hiPSC line that was derived
   from a patient that has the founder point mutation most common in
   mitochondrial trifunctional protein disorder, HADHA c.1528G > C
   (Fig. [274]6j and Supplemental Fig. [275]1C, D). We differentiated the
   HADHA c.1528G > C hiPSCs to cardiomyocytes to assess their sarcomere
   structure and mitochondria in the presence and absence of fatty acids.
   HADHA c.1528G > C CMs cultured in glucose have well-formed myofibrils
   and sarcomeres as seen by phalloidin and α-actinin staining
   (Fig. [276]6k). When HADHA c.1528G > C CMs were cultured in Glc + FAs
   for 12-days, we found the same sarcomere dissolution and loss of
   mitochondrial potential gradient, as shown via mitotracker, as we
   previously had shown with our HADHA Mut CM line (Fig. [277]6k, l and
   Fig. [278]6a, b). These data show a cardiac disease phenotype in human
   patient cells with the founder mutation, HADHA c.1528G > C, in a cell
   culture model. Furthermore, this third, independent HADHA mutant line,
   with a different background from our CRISPR/Cas9 modified hiPSC lines
   with only a single point mutation in the gene HADHA at c.1528G > C,
   recapitulated the disease phenotype found in our HADHA Mut and KO CMs.
   Consequently, the data show CRISPR/Cas9 disease phenotype phenocopies
   that of the human clinical phenotype.

Loss of HADHA leads to long-chain fatty acid accumulation

   To assess the disruption of long-chain fatty acid oxidation in HADHA
   Mut and KO CMs, we performed untargeted lipidomic analysis to
   characterize global lipidomic changes. We found an increase in
   long-chain acyl-carnitines in HADHA Mut and KO CMs as compared to WT
   CMs with no significant change in medium-chain acyl-carnitine levels
   (Fig. [279]7a, b and Supplemental Figure [280]10A). These data suggest
   that a mutation in HADHA led to an accumulation of long-chain FAs in
   the mitochondria. During the first step of long-chain FAO, saturated
   FAs will be processed into FAs with a single double bond, for instance:
   14:0→14:1, 16:0→16:1 and 18:0→18:1, while unsaturated FAs, on the
   carboxyl end, will go through the first step of FAO and gain another
   double bond, for instance: 18:1→18:2 and 18:2→18:3. Accordingly, we
   found minimal variation in the levels of the saturated FAs: 14:0, 16:0
   and 18:0 (Supplemental Fig. [281]10B–D) but, large increases in the
   abundance of 14:1, 16:1, 18:1 in the HADHA Mut and KO CMs along with
   slight increases in 18:2 and 18:3 in the HADHA KO CMs (Fig. [282]7c–e
   and Supplemental Fig. [283]7E,F). We also examined the total abundance
   of triglycerides, neutral lipids, in the HADHA Mut and KO CMs and found
   those to be significantly higher than WT CMs suggesting FAs are not
   being catabolized and are instead being packaged for storage due to
   their accumulation (Fig. [284]7f). These data recapitulate the in vivo
   findings from short lived HADHA KO mice serum levels of elevated FA
   species^[285]44.

Fig. 7.

   [286]Fig. 7
   [287]Open in a new tab

   Fatty acid challenged HADHA KO and Mut CMs have elevated fatty acids
   and abnormal cardiolipin profiles. a Model of long-chain FA
   intermediate accumulation after the first step of long-chain FAO due to
   the loss of HADHA. b The sum of all long-chain acyl-carnitines in WT,
   Mut and KO FA treated hPSC-CMs. *p < 0.05, one-way ANOVA was performed
   vs WT 12D Glc + FA. n = 2–6 biological replicates. c Amount of
   physeteric acid in the free fatty acid state in WT, Mut and KO FA
   treated hPSC-CMs. ***p < 0.001, one-way ANOVA was performed vs WT 12D
   Glc + FA. n = 2–6 biological replicates. d Amount of palmitoleic acid
   in the free fatty acid state in WT, Mut and KO FA treated hPSC-CMs.
   ***p < 0.001, one-way ANOVA was performed vs WT 12D Glc + FA. n = 2–6
   biological replicates. e Amount of oleic acid in the free fatty acid
   state in WT, Mut and KO Glc + FA treated hPSC-CMs. **p < 0.01,
   ***p < 0.001, one-way ANOVA was performed vs WT 12D Glc + FA. n = 2–6
   biological replicates. f The sum of all triglycerides in WT, Mut and KO
   FA treated hPSC-CMs. ***p < 0.001, one-way ANOVA was performed vs WT
   12D Glc + FA. n = 2–6 biological replicates. g The sum of all
   hydroxylated long-chain acyl-carnitine species found in WT, Mut and KO
   Glc + FA treated hPSC-CMs. ***p < 0.001, one-way ANOVA was performed vs
   WT 12D Glc + FA. n = 2–6 biological replicates. h Relative amount of
   tetra[18:2]-CL in WT and HADHA KO CMs treated with either Glc or
   Glc + FA. ***p < 0.001, one-way ANOVA was performed vs WT 12D Glc + FA.
   n = 3 biological replicates. i Cardiolipin profile generated from
   targeted lipidomics for WT, HADHA KO and HADHA c.1528G > C CMs treated
   with either Glc or Glc + FA. *p < 0.05, **p < 0.01, ***p < 0.001,
   one-way ANOVA was performed vs WT 12D Glc. n = 3 biological replicates.
   j Cardiolipin profile generated from global lipidomics for WT CMs 12D
   Glc + FA, HADHA Muts CM 6D and 12D Glc + FA and HADHA KO CMs 12D
   Glc + FA. *p < 0.05, **p < 0.01, ***p < 0.001, one-way ANOVA was
   performed vs WT 12D Glc + FA. n = 2–6 biological replicates. k The sum
   of all CLs that have myristic acid (14:0) in their side chain in WT,
   HADHA Mut and HADHA KO CM FA treated hPSC-CMs. **p < 0.01,
   ***p < 0.001, one-way ANOVA was performed vs WT 12D Glc + FA. n = 2–6
   biological replicates. l The sum of all CLs that have palmitic acid
   (16:0) in their side chain in WT, HADHA Mut and HADHA KO CM FA treated
   hPSC-CMs. ***p < 0.001, one-way ANOVA was performed vs WT 12D Glc + FA.
   n = 2–6 biological replicates. m Schematic diagram of how HADHA works
   in series with TAZ to remodel CL. Box plot middle line represents the
   median, x represents mean, bottom line of the box represents the median
   of the bottom half (1st quartile) and the top line of the box
   represents the median of the top half (3rd quartile). The whiskers
   extend from the ends of the box to the non-outlier minimum and maximum
   value. Bar graphs show mean with standard error. Source data are
   provided as a Source Data file

   One of the hallmarks for clinically testing whether a patient has MTP
   deficiency is to examine if there are increased levels of hydroxylated
   long-chain FAs in patient blood serum or cultured
   fibroblasts^[288]3,[289]63–[290]65. We found a significant increase in
   the sum of all hydroxylated long-chain FAs in the HADHA KO CMs as
   compared to the WT CMs (Fig. [291]7g) while little increase in
   medium-chain hydroxylated FAs (Supplemental Fig. [292]10G). We next
   identified all hydroxylated fatty acid species found within our WT,
   HADHA Mut, and HADHA KO CM samples. We identified 14 hydroxylated fatty
   acid compounds, many more than what has been previously published for
   either human or mouse studies. We found the fold increase of many
   long-chain hydroxylated fatty acids to be significantly elevated as
   compared to the WT CMs (Table [293]1). These data re-exemplify the
   specificity of mutations in HADHA resulting in long-chain fatty acid
   disruption. Furthermore, these data match well with in vivo mouse data
   from short lived HADHA KO mice along with data from MTP-deficient
   and/or LCHAD deficient patient serum and cultured
   fibroblasts^[294]3,[295]63–[296]65.

Table 1.

   Hydroxylated FA Species Fold-Change Abundance Compared to WT 12D
   Glc + FA
   Hydroxylated species WT 12D Glc + FA Mut 6D Glc + FA Mut 12D Glc + FA
   KO 6D Glc + FA
   C6:0 1 1.79 5.82* 2.85
   C8:0 1 3.61** 9.28** –
   C10:0 1 2.36 4.49 1.8
   C12:0 1 2.70** 5.17*** 6.08***
   C14:1 1 4.84* 16.52*** 10.52***
   C14:0 1 1.98 6.82 24.15***
   C16:1 1 3.58 14.07 102.47***
   C16:0 1 3.05 9.97 189.69***
   C18:3 1 2.33 11.08* 55.36***
   C18:2 1 3.72 10.14 446.42***
   C18:1 1 5.37 14.32 449.16***
   C18:0 1 3.01 7.63 209.22***
   C20:1 1 1.62 3.98 96.82***
   C20:0 1 1.37 6.35 149.29***
   [297]Open in a new tab

   One-way ANOVA, n = 2–6 biological samples. *p < 0.05, **p < 0.01 and
   ***p < 0.001

   These data show that disruption and KO of HADHA leads to a specific
   long-chain FA intermediate accumulation. Yet, one of the striking
   phenotypes we observed were rounded and collapsed mitochondria and not
   bursting mitochondria due to potential FA overload^[298]66. We
   therefore decided to examine another phospholipid category that
   regulates mitochondrial structure, cardiolipins.

HADHA and TAZ act in series to mature cardiolipin

   Cardiolipin (CL) is a phospholipid essential for optimal mitochondrial
   function and homeostasis as it maintains electron transport chain
   function along with other mitochondrial functions^[299]67,[300]68. CL
   is the major phospholipid of the mitochondrial inner membrane that is
   synthesized in the mitochondria and is dynamically remodeled during
   postnatal development and disease^[301]16,[302]17. The most abundant
   species of CL in the human heart is tetralinoleoyl-CL
   (tetra[18:2]-CL)^[303]69. In cardiac diseases such as diabetes,
   ischemia/reperfusion and heart failure, or due to a specific mutation
   in a cardiolipin remodeling enzyme tafazzin (TAZ), which leads to Barth
   syndrome, tetra[18:2]-CL levels are abnormal^[304]70–[305]73. Using
   targeted lipidomics, we analyzed WT CMs supplemented with and without
   FAs. We found that FA treated WT CMs resulted in a significant increase
   in tetra[18:2]-CL (Fig. [306]7h), similar to previously observed
   findings during in vivo cardiomyocyte postnatal
   maturation^[307]16,[308]17. These data show that CL maturation in
   cardiomyocytes can be induced in vitro. However, the HADHA KO CMs,
   after FA treatment, were unable to increase the amount of
   tetra[18:2]-CL as compared to WT FA treated CMs. Furthermore, as shown
   in postnatal in vivo development, WT CMs shift their CL profile to a
   more mature CL profile showing a significant decrease in CLs with
   [16:1] and increased CLs with carbons greater than 18, including the
   intermediate [18:1][18:2][18:2][20:2]^[309]16. However, HADHA KO CMs
   and patient derived HADHA c.1528G > C CMs were unable to remodel their
   CL profiles as efficiently as WT CMs (Fig. [310]7i). These data show
   that, surprisingly, HADHA, in addition to its role in long-chain FAO,
   is also required for the cardiomyocyte CL remodeling process.

   Since HADHA KO CMs showed a CL remodeling defect, we next analyzed the
   cardiolipin species in more detail in WT, HADHA Mut and KO CMs using
   full lipidomics. Reinforcing our targeted lipidomics results, we found
   that HADHA Mut and KO CMs challenged with FAs showed an increased
   abundance of lighter chain CLs and a depletion of heavier chain CLs
   (Fig. [311]7j and Supplemental Figure [312]10H). Three CL species,
   tetra[18:1], [18:1][18:1][18:1][18:2] and [18:1][18:1][18:2][18:2] were
   significantly enriched in the HADHA Mut and KO CMs (Fig. [313]7j).
   Interestingly, [18:1][18:1][18:2][18:2] CL is specifically depleted in
   Barth syndrome patients who have a mutation in TAZ^[314]74,[315]75.

   It has been previously shown that the HADHA protein has a similar
   enzymatic function to monolysocardiolipin acyltransferase (MLCL
   AT)^[316]76,[317]77. MLCL AT transfers mainly unsaturated fatty
   acyl-chains to lyso-CL. It therefore seems plausible that HADHA has a
   direct role in remodeling cardiolipin to produce mature tetra[18:2]-CL
   species in cardiomyocytes. If TAZ and HADHA are acting in parallel to
   produce remodeled CL, they should both be equally depleting the MLCL
   pool. When TAZ is KO’d, there is a dramatic increase in MLCL, showing
   the direct usage of MLCL by TAZ to generate mature CL^[318]78,[319]79.
   However, when HADHA is KO’d, there is no change in the MLCL pool
   (Supplemental Fig. [320]10I). This would suggest that HADHA does not
   remodel MLCL but rather CL. If TAZ and HADHA are acting in parallel,
   the KO of each should not result in the inverse accumulation
   relationship to specific CL intermediates. For instance, TAZ KO results
   in the decrease of [18:1][18:1][18:2][18:2] CL^[321]74,[322]75. Yet in
   our HADHA KO we see an accumulation of the same species. We propose
   that TAZ first remodels MLCL to an intermediate of CL such as
   [18:1][18:1][18:2][18:2] and then HADHA continues to remodel the CL
   species to tetra[18:2]-CL.

Loss of HADHA function does not augment ALCAT1 function

   To garner a better understanding of how the cardiolipin profile was
   changing due to the lack of HADHA, we examined which new CL species
   became enriched in the HADHA Mut and KO CMs. CL species that had fatty
   acid acyl-chains of saturated fatty acids, such as 14:0 and 16:0, were
   enriched in the HADHA Mut and KO CMs (Fig. [323]7k, l). We did not
   identify any CL acyl-chains that had 18:0. Typically, nascent CL with
   multiple saturated fatty acid acyl-chains (CL[Sat]), have been
   synthesized from cardiolipin synthase (CLS) (Fig. [324]7m)^[325]80.
   During the remodeling process of CL[Sat], the saturated fatty acid
   acyl-chains are replaced by unsaturated fatty acid acyl-chains. Our
   data suggest a nascent CL[Sat] accumulation in HADHA mutants.

   We next examined ALCAT1 (acyl-CoA:lysocardiolipin acyltransferase-1) as
   a means for the HADHA Mut and KO CMs to utilize for CL remodeling.
   Since ALCAT1 has no preference for fatty-acyl substrate, it should
   utilize whichever fatty-acyl-CoA substrate is present^[326]81,[327]82.
   Hallmarks of ALCAT1 activity are an increase in polyunsaturated fatty
   acid acyl-chains being incorporated to CL^[328]83. However, when we
   examined the CL species that had acyl-chains with fatty acids with a
   carbon length 20 or greater, the majority of the HADHA Mut and KO CMs
   actually had less species as compared to WT CMs (Fig. [329]7j).
   Furthermore, there was no increase in CL species that had multiple
   acyl-chains with fatty acids with a carbon length 20 or greater in any
   of the groups. Consequently, these data suggest that ALCAT1 is not
   being engaged in the HADHA Mut and KO CMs to compensate for the loss of
   HADHA.

Discussion

   We have generated the first human MTP-deficient cardiac model in vitro
   utilizing MiMaC matured hiPSC-CMs and discovered that a TFPα/HADHA
   defect in long-chain FAO and CL remodeling results in disease like
   erratic beating suggesting a pro-arrhythmic state. We further showed a
   mechanism of action; mutations in HADHA resulted in abnormal
   composition of the prominent phospholipid, CL due to HADHA’s acyl-CoA
   transferase activity^[330]76.

   CL is important for mitochondrial architecture, it has been shown to
   function in organizing the cristae and electron transport chain (ETC)
   higher order structure, important for ETC activity, and acts as a
   proton trap on the outer leaflet of the inner mitochondrial
   membrane^[331]84. CMs with defective HADHA are unable to generate large
   amounts of tetra[18:2]-CL, due to their inability to remodel nascent CL
   during CM maturation. Hence the mitochondrial morphology becomes
   rounded in HADHA Mut, KO and c.1528G > C CMs. These data suggest that
   FAO phenotypes alone might not explain the defects observed in HADHA
   Mut, KO and HADHA c.1528G > C CMs and that CL remodeling is
   particularly important during the CM maturation process. Hence, we
   propose that a mutation in the HADHA enzyme during the CM maturation
   process results in an over accumulation of immature CL-saturated
   species, that may be causal for the observed mitochondrial defects and
   pathology (Fig. [332]7m).

   Here we have shown that the MTP-deficient pathology in CMs leads to an
   abnormal cardiolipin pattern that results in severe mitochondrial
   defects and calcium abnormalities that pre-dispose CMs to erratic
   beating in HADHA Mut CMs. We identified SS-31 as a therapy to rescue
   the proton leak phenotype of FA challenged HADHA Mut CMs. This suggests
   that SS-31, or other cardiolipin affecting compounds, may serve as a
   potential treatment to help mitigate aspects of mitochondrial
   dysfunction in MTP deficiency.

Methods

hESC and hiPSC and cardiac differentiation

   The hESC line RUES2 (NIHhESC-09–0013, WiCell, RUESe002-A) and hiPSC
   line WTC #11 (Coriell Institute, GM25256), previously derived in the
   Conklin laboratory^[333]85, were cultured on Matrigel growth
   factor-reduced basement membrane matrix (Corning) in mTeSR media
   (StemCell Technologies). A monolayer-based directed differentiation
   protocol was followed to generate hESC-CMs and hiPSC-CMs, as done
   previously^[334]26. hiPSC-CM cardiolipin assay was done with a small
   molecule monolayer-based directed differentiation protocol, as done
   previously^[335]27. Fifteen days after differentiation hPSC-CMs were
   enriched for the cardiomyocyte population using a lactate selection
   process^[336]86. We generated cardiomyocyte populations ranging from
   40–60% that were then enriched to 75–80% cardiomyocytes after 4 days of
   lactate enrichment.

HADHA line creation

   Using LentiCrisprV2 plasmid^[337]87 (lentiCRISPRV2 was a gift from Feng
   Zhang (Addgene plasmid # 52961) two different gRNAs targeted to Exon 1
   of HADHA were designed using CRISPRScan^[338]88. Sequences for the
   gRNAs can be found in Supplemental Table [339]S12. The gRNA and Cas9
   expressing plasmids were transiently transfected into the WTC line
   using GeneJuice (EMD Millipore). Twenty-four hour after transfection,
   WTCs were puromycin selected for 2 days and then clonally expanded. DNA
   of the clones was isolated, the region around the targeting guides was
   PCR amplified (see guides in Supplemental Table [340]S12) and sequenced
   to determine the insertion and deletion errors generated by CRISPR-Cas9
   system in exon 1 of HADHA. Western analysis was performed to determine
   the levels of HADHA protein in HADHA mutants. Thirty-one clones were
   sent for sequencing from gRNA1 experiment, six clones (19%) had no
   mutations while 25 clones (81%) were found to have mutations.
   Twenty-four clones were sent for sequencing from gRNA2, one clone had
   no mutations (4%) while 23 clones (96%) were found to have mutations.
   Two of the mutant lines were analyzed further in this study.

CRISPR off-target

   The potential off-targets of the HADHA gRNA were identified using
   Crispr-RGEN’s Cas-OFFinder tool^[341]89. The top predicted off-targets
   were then amplified by GoTaq PCR and sequenced. Off-target primers can
   be found in Supplemental Table [342]S13. Sequencing for primer pair #1
   can be found in Supplemental Fig. [343]1G.

HADHA c.1528G > C patient and the cell line

   The patient manifested the disease during the first months after birth
   with hypoketotic hypoglycemia and failure to thrive, with metabolic
   acidosis, cardiomyopathy, and hepatomegaly. The skin sampling to obtain
   fibroblast cultures was performed with informed consent of the parents,
   as approved by the ethical review board of Helsinki University
   Hospital, and according to Helsinki Declaration. The HEL87.1 LCHAD
   patient cell line was isolated from skin fibroblasts and reprogrammed,
   by Sendai vector-based CytoTune-TM method as previously
   described^[344]25. Their stemness characteristics were confirmed by
   immunohistochemistry (expression of pluripotent markers OCT4, SSEA4,
   Tra-1–60). Quantitative PCR was used to assess the expression levels of
   endogenous stem cell markers and to confirm removal of transgene
   vectors. The karyotype was confirmed to be diploid, 46XY. The HEL87.1
   iPSCs were cultured and differentiated as described above for hiPSCs
   and hESCs. Point mutation was confirmed via Sanger sequencing by
   amplifying a region around the c.1528G > C mutation using primers found
   in Supplemental Table [345]16.

RNA extraction and qPCR analysis

   RNA was extracted from cells using Trizol and analyzed with SYBR green
   qPCR using the 7300 real-time PCR system (Applied Biosystems). Primers
   used are listed in Supplemental Table [346]S14. Linear expression
   values for all qPCR experiments were calculated using the delta-delta
   Ct method.

Protein extraction and western blot analysis

   Cells were lysed directly on the plate with a lysis buffer containing
   20 mM Tris-HCl pH 7.5, 150 mM NaCl, 15% Glycerol, 1% Triton X-100, 1 M
   β-Glycerolphosphate, 0.5 M NaF, 0.1 M Sodium Pyrophosphate,
   Orthovanadate, PMSF and 2% SDS^[347]90. 25U of Benzonase Nuclease (EMD
   Chemicals, Gibbstown, NJ) was added to the lysis buffer right before
   use. Proteins were quantified by Bradford assay (Bio-rad), using BSA
   (Bovine Serum Albumin) as Standard using the EnWallac Vision. The
   protein samples were combined with the 4x Laemmli sample buffer, heated
   (95 °C, 5 min), and run on SDS-PAGE (protean TGX pre-casted 4%-20%
   gradient gel, Bio-rad) and transferred to the Nitro-Cellulose membrane
   (Bio-Rad) by semi-dry transfer (Bio-Rad). Membranes were blocked for
   1 h with 5% milk and incubated in the primary antibodies overnight at
   4 °C. The membranes were then incubated with secondary antibodies
   (1:10,000, goat anti-rabbit or goat anti-mouse IgG HRP conjugate
   (Bio-Rad) for 1 h and the detection was performed using the
   immobilon-luminol reagent assay (EMD Millipore). Full blots can be
   found in the source data file. Primary antibodies are as follows: Alpha
   tubulin antibody Cell Signalling Technologies (2144) 1:2000, Beta
   tubulin Promega (G7121) anti-mouse 1:4000, Beta Actin Cell Signalling
   Technologies (4970) 1:4000, HADHA Abcam (ab54477 anti-rabbit 1:1000,
   UCP3 Abcam (ab3477) anti-rabbit 1:200, SLC25A4 (ANT1) Sigma
   (SAB2105530) anti-rabbit 1:1000, OXPHOS MitoSciences (MS604/G2830)
   anti-mouse 1:1000, anti-GFP Invitrogen (A-11122) anti-rabbit 1:1000.

microRNA overexpression and Knockout

   We used LentiCrisprV2 plasmid (Addgene 52961) to KO microRNAs-141,
   −200a, −205, and −122. gRNAs for each miR that had either the
   protospacer adjacent motif (PAM) NGG cut site adjacent or in the seed
   region of the mature microRNA were chosen to test. gRNAs can be found
   in Supplement Table [348]S12. The global reduction of each miR was
   assessed via TaqMan RT-qPCR with probes specific against the mature
   form of each respective miR.

   We used the pLKO.1 TRC vector (pLKO.1—TRC cloning vector was a gift
   from David Root (Addgene plasmid # 10878) to OE a microRNA^[349]91. The
   genomic sequence 200 bp up- and downstream of the mature microRNA was
   amplified and purified. Primers for each microRNA can be found in
   Supplemental Table [350]S15. The amplicons were cloned between AgeI and
   EcoRI sites of pLKO.1 TRC vector under the human U6 promoter.

Viral production

   HEK 293FT cells were plated one day before transfection. On the day of
   transfection, the OE or KO plasmid of choice was combined with
   packaging vectors psPAX2 (psPAX2 was a gift from Didier Trono Addgene
   plasmid # 12260) and pMD2.G (pMD2.G was a gift from Didier Trono
   Addgene plasmid # 12259) in the presence of 1 μg/μL of polyethylenimine
   (PEI) per 1 μg of DNA. Medium was changed 24 h later and the
   lentiviruses were harvested 48 and 72 h after transfection. Viral
   particles were concentrated using PEG-it (System Biosciences, Inc).

hiPSC-CM transduction and selection

   hiPSC-CMs were transduced on day 14 post-induction in the presence of
   hexadimethrine bromide (Polybrene, 6 μg/ml). Lentivirus was applied for
   17–24 h and then removed. Cells were cultured for an additional two
   weeks. Lactate selection was employed to obtain an enriched population
   of cardiomyocytes^[351]86. Puromycin selection was used to select for
   cells that have positively incorporated the vector. After 2 weeks of
   culture, cells were harvested for end point analysis. For the MiMaC
   group, hiPSC-CMs were transduced with a lower dose of the four
   different lentiviruses concurrently while controls were transduced with
   both control vectors: pLKO.1 and the LentiCRISPRv2 empty vector.

Immunocytochemistry and morphological analysis

   Cells were fixed in 4%(vol/vol) paraformaldehyde, blocked for an hour
   with 5% (vol/vol) normal goat serum (NGS), and incubated overnight with
   primary antibody in 1% NGS, followed by secondary antibody staining in
   NGS. Measurements of CM area were performed using Image J software.
   Quantification of mitotracker intensity were performed using Image J
   software and following previously published methods on colocalization
   quantification^[352]92. Analysis was done on a Leica TCS-SPE Confocal
   microscope using a ×40 or ×63 objective and Leica Software. Primary
   antibodies used were: αActinin 1:250 Sigma A7811 anti-mouse, HADHA
   1:250 abcam ab54477 anti-rabbit, ATP Synthase β 1:250 abcam ab14730
   anti-mouse, Titin 1:300 Myomedix TTN-9 (cTerm) anti-rabbit, GFP 1:300
   Invitrogen A-11122 anti-rabbit. Secondary antibodies and other reagents
   used were: DAPI at a concentration of 0.02 μg/mL, phalloidin alexa
   fluor 568 1:250, alexa fluor 488 or 647-conjugated goat anti-mouse and
   anti-rabbit secondary antibodies 1:500 (Molecular Probes).
   MitotrackerCMTMRos Life technologies (M7510) used at a final
   concentration of 300 nM in RPMI with B27 plus insulin supplement,
   incubated with cells for 45 min prior to fixation.

Mitochondrial calcium

   HADHA Mut and WT CMs were plated following lactate enrichment at 20,000
   cells per Matrigel-coated well in a 24 well, glass bottom plate
   (Cellvis) and treated with Glc + FA medium for 12 days. Cells were
   stained using 4.5 mM Rhod-2 (Thermofisher R1244) in DMSO and 2 nM
   Mitotracker green (Thermofisher M7514) in DMSO for 30 min^[353]93.
   Cells were rinsed with PBS and returned to culture medium for imaging
   on the heated, 5%CO2 stage of an inverted Nikon eclipse Ti equipped a
   Yokogawa W1 spinning disk. Colocalization analysis was performed as
   previously described^[354]92.

Micro-electrode array

   Electrophysiological recording of spontaneously beating cardiomyocytes
   was collected for 2 min using the AxIS software (Axion Biosystems).
   After raw data collection, the signal was filtered using a Butterworth
   band-pass filter and a 90μV spike detection threshold. Field potential
   duration was automatically determined using a polynomial fit T-wave
   detected algorithm.

Microposts (force of contraction and beat rate)

   Arrays of polydimethylsiloxane (PDMS) microposts were fabricated as
   previously described^[355]36. The tips of the microposts were coated
   with mouse laminin (Life Technologies), and cells were seeded onto the
   microposts in Attoflour® viewing chambers (Life Technologies) at a
   density of approximately 75,000 per cm^2 in RPMI medium with B27
   supplement and 10% fetal bovine serum. The following day, the media was
   removed and replaced with serum-free RPMI medium, which was exchanged
   every other day. Once the cells resumed beating (typically 3–5 days
   after seeding), contractions of individual cells were imaged (at a
   minimum of 70 FPS) using a Hamamatsu ORCA-Flash2.8 Scientific CMOS
   camera fitted on a Nikon Eclipse Ti upright microscope using a ×60
   water-immersion objective. Prior to imaging, the cell culture media was
   replaced with a Tyrode buffer containing 1.8 mM Ca2+, and a live cell
   chamber was used to maintain the cells at 37 °C throughout the imaging
   process. A custom-written matlab code was used to track the deflection,
   [MATH: <msub><mrow><mi
   mathvariant="normal">Δ</mi></mrow><mrow><mi>i</mi></mrow></msub> :MATH]
   , of each post i underneath an individual cell, and to calculate the
   total twitch force,
   [MATH: <msub><mrow><mi>F</mi></mrow><mrow><mi
   mathvariant="normal">twitch</mi></mrow></msub><mo>=</mo><munderover
   accent="false" accentunder="false"><mrow><mo>
   ∑</mo></mrow><mrow><mi>i</mi><mo>=</mo><mn>1</mn></mrow><mrow><mi>#</mi
   ><mspace width="0.25em"></mspace><mi
   mathvariant="normal">posts</mi></mrow></munderover><msub><mrow><mi>k</m
   i></mrow><mrow><mi
   mathvariant="normal">post</mi></mrow></msub><mo>×</mo><msub><mrow><mi
   mathvariant="normal">Δ</mi></mrow><mrow><mi>i</mi></mrow></msub> :MATH]
   ^[356]36, where
   [MATH: <msub><mrow><mi>k</mi></mrow><mrow><mi
   mathvariant="normal">post</mi></mrow></msub><mo>=</mo><mn>56.5</mn><mi>
   n</mi><mi>N</mi><mo>∕</mo><mi>μ</mi><mi>m</mi> :MATH]
   and the spacing between posts was 6 μm.

Seahorse assay

   The Seahorse XF96 extracellular flux analyzer was used to assess
   mitochondrial function as previously described^[357]24. The plates were
   pre-treated with 1:60 diluted Matrigel reduced growth factor (Corning).
   At around 28 days after differentiation, cardiomyocytes were seeded
   onto the plates with a density of 50,000 cells per XF96 well. The
   seahorse assays were carried out 3 days after the seeding onto the XF96
   well plate. One hour before the assay, culture media was exchanged for
   base media (unbuffered DMEM (Seahorse XF Assay Media) supplemented with
   sodium pyruvate (Gibco/Invitrogen, 1 mM) and with 25 mM glucose (for
   MitoStress assay), 25 mM glucose with 0.5 mM Carnitine for Palmitate
   assay. Injection of substrates and inhibitors were applied during the
   measurements to achieve final concentrations of 4-(trifluoromethoxy)
   phenylhydrazone at 1 μM (FCCP; Seahorse Biosciences), oligomycin
   (2.5 μM), antimycin (2.5 μM) and rotenone (2.5 μM) for MitoStress
   assay; 200 mM palmitate or 33 μM BSA, and 50 μM Etomoxir (ETO) for
   palmitate assay. The OCR values were further normalized to the number
   of cells present in each well, quantified by the Hoechst staining
   (Hoechst 33342; Sigma–Aldrich) as measured using fluorescence at 355 nm
   excitation and 460 nm emission. Maximal OCR is defined as the change in
   OCR in response to FCCP compared to OCR after the addition of
   oligomycin. ATP production was calculated as the difference between the
   basal respiration and respiration after oligomycin. Proton leak was
   calculated as the difference between respiration after oligomycin and
   after antimycin and rotenone. Cellular capacity to utilize palmitate as
   an energy source was calculated as the difference between the average
   OCR after second palmitate addition and the final respiration value
   before the second addition of palmitate. The reagents were from Sigma,
   unless otherwise indicated.

Whole-cell patch clamp analysis

   WT and Mut CMs were plated as single cells onto Matrigel-coated glass
   coverslips and treated with Glc + FA medium for 12 days. Whole-cell
   patch clamp recordings were performed on an inverted DIC microscope
   (Nikon) connected to an EPC10 patch clamp amplifier and computer
   running Patchmaster software (HEKA). Coverslips were loaded onto the
   stage and bathed in a Tyrode’s solution containing 140 mM NaCl, 5.4 mM
   KCl, 1.8 mM CaCl2, 1 mM MgCl2, 10 mM glucose, and 10 mM HEPES. The
   intracellular recording solution (120 mM L-aspartic acid, 20 mM KCl,
   5 mM NaCl, 1 mM MgCl2, 3 mM Mg2 + -ATP, 5 mM EGTA, and 10 mM HEPES) was
   loaded into borosilicate glass patch pipettes (World Precision
   Instruments). Patch pipettes with a resistance in the range of 2–6 MΩ
   were used for all recordings. Offset potentials were nulled before
   formation of a gigaΩ seal and fast and slow capacitance was compensated
   for in all recordings. Membrane potentials were corrected by
   subtraction of a 20 mV tip potential, calculated using the HEKA
   software. Voltage-clamp and current-clamp experiments were then
   performed. To generate a single action potential, a 5 ms depolarizing
   current pulse of sufficient intensity was applied in current-clamp
   mode. Inward and outward currents were evoked by a series of 500 ms
   depolarizing steps from −120 to +70 mV with +10 mV increments in
   voltage-clamp mode. Gap-free recordings of spontaneous activity of
   patched cardiomyocytes were performed for 30 s with 0 pA current
   injection to provide a measure of the maximum diastolic potential
   (resting membrane potential) held by the cell without current input.
   All recordings and analyses were performed using the HEKA Patchmaster
   software suite.

RNA-sequencing

   Day-30 hiPSC-CMs were harvested for RNA preparation and genome wide
   RNA-seq (>20 million reads). RNA-seq samples were aligned to hg19 using
   Tophat, version 2.0.13^[358]94. Gene-level read counts were quantified
   using htseq-count^[359]95 using Ensembl GRCh37 gene annotations. Genes
   with total expression above 1 normalized read count across RNA-seq
   samples in each binary comparison were kept for differential analysis
   using DESeq^[360]96. Princomp function from R was used for Principal
   Component Analysis. TopGO R package^[361]97 was used for Gene Ontology
   enrichment analysis. To assess the effects of miR perturbation on
   cardiac maturation pathways, each condition was compared against their
   empty vector (EV), and upregulated genes (>1.5-fold change) and
   downregulated genes were identified (<−1.5-fold change). A
   hypergeometric test was performed on up- and downregulated genes
   separately for enrichment against a curated set of pathways that are
   beneficial for cardiac maturation, resulting in a m by n matrix, where
   m is the number of pathways (m = 7) and n is the number of conditions
   (n = 6, including EV). The negative log[10] of the ratio between
   enrichment p-value for up- and downregulated genes were calculated to
   represent the overall net “benefit” of a treatment: large positive
   value (>0) means the treatment results in more upregulation of genes in
   cardiac maturation pathways than downregulation of these genes, and
   more negative values means the treatment results in more downregulation
   of genes in cardiac maturation pathways. Human fetal and adult
   ventricular data was obtained from the Roadmap Epigenomics
   Consortium^[362]98.

Single-cell RNA-sequencing

   Raw single-cell RNA-seq data are processed through the CellRanger
   pipeline from 10X Genomics. Output of the CellRanger pipeline is
   further analyzed using Seurat R package^[363]99. Cells with more than
   40% of reads mapped to mitochondrial genes, less than 200 detected
   genes or less than 2000 Unique Molecular Identifiers (UMIs) are
   removed. Remaining cells are scaled by number of UMIs and % mapped to
   mitochondrial genes. Parameters for tSNE analysis of maturation single
   cell RNA-seq data were 2905 top variable genes, top 10 principal
   components, and resolution 0.5. Parameters for tSNE analysis of HADHA
   mutant single cell RNA-seq data were 3375 top variable genes, top 10
   principal components, and resolution 0.4. Cell cycle genes from
   Kowalczyk et al^[364]54. and the CellCycleScoring function in the
   Seurat package were used to assess the effects of cell cycle on
   clustering. Genes detected in at least 25% of cells in either cluster
   and have false discovery rate <0.1 are defined as differentially
   expressed. Expression values are normalized for each gene across all
   cells plotted in the heat maps (i.e., Z-scores). Human in vivo
   maturation markers are based on genes upregulated in adult heart
   compared to fetal heart in the Roadmap Epigenomics Project^[365]98.
   Mouse in vivo maturation markers are based on genes upregulated in the
   in vivo cardiomyocyte single cell RNA-seq data from Delaughter et
   al.^[366]41. We also used PCA projection to assess the maturity of our
   MiMac single cell RNA-seq data to single cell RNA-seq of human fetal
   heart development^[367]100. We selected genes significantly higher in
   adult heart compared to fetal using DESeq (2 fold higher in adult,
   FDR < 0.05). We then intersected these genes with the top 30 most
   highly expressed genes in each scRNA-seq cluster to get the final gene
   list for heatmap in Fig. [368]3o. Gene Ontology enrichment is performed
   using the TopGO package^[369]97. We used the CellTrails method^[370]57
   to further dissect the dynamics of HADHA perturbation. CellTrails is a
   new algorithm that uses lower-dimensional manifold learning for de novo
   chronological ordering. The parameter for gene selection were keeping
   genes with fano factor >1.2; parameters for clustering were
   min_size = 0.02, min_feat = 10, max_pval = 1e-2, min_fc = 1.5 (each
   cluster should include at least 2% of all cells, and contain at least
   10 genes that are expressed 1.5 fold higher in that cluster compared to
   others). The yEd graph editor
   ([371]https://www.yworks.com/products/yed) was used to generate
   visualization of single cell dynamic trails, as recommended by the
   CellTrails algorithm.

Calcium transient analysis method

   Cardiomyocytes were plated on Matrigel-coated round glass coverslips.
   The cardiomyocytes were incubated for 25 min at 37 °C with 1 mM Fluo-4
   AM (Life Technologies, [372]F14201) in Tyrode’s buffer (1.8 mM CaCl[2],
   1 mM MgCl[2], 5.4 mM KCl, 140 mM NaCl, 0.33 mM NaH[2]PO[4], 10 mM
   HEPES, 5 mM glucose, pH to 7.4). The substrate was then transferred to
   a 60 mm Petri dish fresh with pre-warmed Tyrode’s buffer for imaging.
   Samples were imaged using a Hamamatsu ORCA-Flash2.8 Scientific CMOS
   camera fitted on a Nikon Eclipse Ti upright microscope. Videos were
   taken with a ×40 water-immersion objective at a framerate of at least
   20 frames per second. The fluorescence power was adjusted to ensure
   adequate capture of fluorescence change during depolarization without
   bleaching, and the same fluorescence power was used for all
   experiments. The cardiomyocytes were biphasically stimulated at 5 V/cm
   with carbon electrodes (Ladd Research, 30250) at either 0.5 Hz or 1 Hz,
   and at least five beats were captured during each video for analysis.

   Videos were analyzed with a custom MATLAB code; calcium transients were
   obtained finding the cell boundary and averaging the fluorescence
   within the boundary for each video frame. The background fluorescence
   was determined automatically for each video frame and subtracted from
   the calcium transients. The calcium transients were then analyzed to
   find the peak fluorescence (F), baseline fluorescence (F[0]), time to
   peak (T[peak]), and time to 50 and 90% relaxation (T[50R], T[90R]). The
   rates to peak, 50%, and 90% relaxation (R[peak], R[50R], R[90R]) were
   calculated by dividing the respective fluorescence change by the
   respective time. An exponential decay function (
   [MATH:
   <msup><mrow><mi>e</mi></mrow><mrow><mo>−</mo><mi>t</mi><mo>∕</mo><mi>τ<
   /mi></mrow></msup> :MATH]
   ) was fit to the relaxation between 10 and 90% relaxation to determine
   the relaxation coefficient, τ. All of these measurements were obtained
   for at least four beats in each video and averaged for comparison.

TEM

   Cells were fixed in 4% glutaraldehyde in sodium cacodylate buffer, post
   fixed in osmium tetroxide, en bloc stained in 1% uranyl acetate,
   dehydrated through a series of ethanol, and embedded in Epon Araldite.
   70 nm sections were cut on a Leica EM UC7 ulta microtome and viewed on
   a JEOL 1230 TEM.

Glucose and fatty acid media

   The base media, which we are calling Glucose Media, is RPMI
   supplemented with B27 with insulin. The fatty acid media is the glucose
   media with oleic acid conjugated to BSA (Sigma O3008): 12.7 μg/mL,
   linoleic acid conjugated to BSA (Sigma L9530): 7.05 μg/mL, sodium
   palmitate (Sigma P9767) conjugated to BSA (Sigma A8806): 52.5 μM and
   L-carnitine: 125 μM. Fatty acid (FA) experiments used this
   B27 + insulin + the three FAs (oleic acid, linoleic acid and palmitic
   acid), in RPMI media. This media was changed every other day during the
   6-days or 12-days of treatment.

Elamipretide (SS-31)

   SS-31 came from Stealth BioTherapeutics and was dissolved in PBS. A
   final concentration of 1 nM was used in experiments.

Box plots

   The ‘x’ in each box plot denotes the average value while the horizontal
   bar denotes the median value, no outlier values are shown. * denotes
   P < 0.05.

Bar graphs

   Bar graphs show the mean with error bars showing standard error.

STRING analysis

   Protein association maps were generated using STRING version 10.5. In
   each diagram, genes connected to one another have an association with
   one another. There are three action effects: arrow – > positive,–| −
   negative and line with a circle on the end—unspecified. There are also
   eight different action types that are denoted by line color:
   green—activation, blue—binding, cyan—phenotype, black—reaction,
   red—inhibition, purple—catalysis, pink—post-translational modification,
   and yellow—transcriptional regulation. Kmeans clustering was used to
   identify the significantly changed genes due to MiMaC for: muscle
   structure development and extracellular matrix organization. Markov
   Clustering Algorithm (MCL) was used to identify genes MiMaC had
   downregulated to control cell division.

Statistical ānalysis

   P-values were calculated using student t-test or one-way ANOVA. For
   one-way ANOVA analysis that failed the normality test, ANOVA a
   Kruskal-Wallis one-way ANOVA of Variance on Ranks was performed. All
   statistical tests used an α = 0.05.

Targeted cardiolipin analysis Using LC-MS/MS

   We used WT hiPSC-CMs treated for 12D Glc + FA media and HADHA Mut
   hiPSC-CMs treated for 6D and 12D Glc + FA media. Immediately before
   extraction, each cell pellet was dissolved in 40 μL DMSO and the
   membranes were disrupted by sonication. Cells were subjected to
   sonication using 3 cycles consisting of 20 s on, 10 s off. Care was
   taken to keep the cells on ice during sonication. After shaking, the
   suspension was transferred into a 2 mL glass LC vial.

   For cardiolipin extraction, an extraction mixture consisting of 20 mL
   chloroform/methanol mix (2:1 v/v) and 30 µL internal standard solution
   (5 mg PC(18:0/18:1(9Z)) (Avanti Polar Lipids, Inc., Alabaster, AL) was
   prepared. Next, 600 μL of the extraction mixture was added to the
   samples, followed by vortexing and incubation at −20 °C for 20 min. The
   samples were then sonicated in an ice bath for 15 min. Purified water
   (100 µL) was added, and the samples were shaken for 30 min at room
   temperature. After centrifugation at 12,000 × g for 10 min at 4 °C. The
   bottom phase was transferred to a new glass LC vial and dried under
   vacuum. The residue was then reconstituted by adding 150 µL
   acetonitrile/isopropanol/H[2]O (65:30:5, v/v/v), and centrifuged at
   20,000 × g for 10 min at 4 °C. The supernatant was transferred to
   individual glass vials for MS analysis.

   For targeted cardiolipin measurements, 2 µL of each prepared sample was
   injected into a 6410 Agilent Triple Quad LC-MS/MS system for analysis
   using an electrospray ionization source and negative ionization mode.
   Chromatographic separation was achieved on an Agilent 300 SB-C8 RRHD
   column (1.8 μm, 2.1 × 50 mm). The mobile phase A was 10 mM ammonium
   acetate in acetonitrile/H[2]O (6:4, v/v), and mobile phase B was 10 mM
   ammonium acetate in isopropyl alcohol/acetonitrile/H[2]O (90:10:1,
   v/v/v). The mobile phase composition changed from 60% A to 1% A over
   the 12 min separation, followed by a rapid increase to 60% A and
   equilibration to prepare for the next injection. The total experimental
   time for each injection was 20 min. The flow rate was 0.26 mL/min, the
   auto-sampler temperature was 4 °C, and the column compartment
   temperature was set to 55 °C. Targeted MS/MS data were acquired using
   multiple-reaction-monitoring (MRM) mode. MassHunter Workstation
   Software Quantitative Analysis for QQQ B.07.00 (Agilent) was used to
   integrate extracted MRM peaks.

Untargeted lipidomic analysis

   One million cells were extracted with 225 µl of methanol at −20 °C
   containing an internal standard mixture of PE(17:0/17:0),
   PG(17:0/17:0), PC(17:0/0:0), C17 sphingosine, ceramide (d18:1/17:0), SM
   (d18:0/17:0), palmitic acid-d[3], PC (12:0/13:0), cholesterol-d[7], TG
   (17:0/17:1/17:0)-d[5], DG (12:0/12:0/0:0), DG (18:1/2:0/0:0), MG
   (17:0/0:0/0:0), PE (17:1/0:0), LPC (17:0), LPE (17:1), and 750 µL of
   MTBE (methyl tertiary butyl ether) (Sigma–Aldrich) at −20 °C containing
   the internal standard cholesteryl ester 22:1. Cells were vortexed for
   20 sec, sonicated for 5 min and shaken for 6 min at 4 °C with an
   Orbital Mixing Chilling/Heating Plate (Torrey Pines Scientific
   Instruments). Then 188 µl of LC-MS grade water (Fisher) was added.
   Samples were vortexed, centrifuged at 14,000 rcf (Eppendorf 5415D). The
   upper (non-polar, organic) phase was collected in two 350 µL aliquots
   and evaporated to dryness. One organic phase aliquot was re-suspended
   in 100 µL of methanol:toluene (9:1, v/v) mixture containing 50 ng/mL
   CUDA ((12-[[(cyclohexylamino)carbonyl]amino]-dodecanoic acid) (Cayman
   Chemical). Samples were then vortexed, sonicated for 5 min and
   centrifuged at 16,000 rcf and prepared for lipidomic analysis. Method
   blanks and pooled human plasma (BioreclamationIVT) were included as
   quality control samples. WT FA CM and HADHA Mut 12D FA were N = 2 with
   each N being a pool of 1–3 samples. Mut 12DFA #1, #3 and #4 were
   averaged as 1 sample and WT 12DFA #1, #2 and #4 were averaged as 1
   sample.

   HADHA KO CM were N = 3 and HADHA Mut 6D FA was n = 6 where N = 2 with 3
   technical replicates per N.

Chromatographic and mass spectrometric conditions for lipidomic RPLC-QTOF
analysis

   Re-suspended samples were injected at 3 µL and 5 µL for ESI positive
   and negative modes respectively, onto a Waters Acquity UPLC CSH C18
   (100 mm length × 2.1 mm id; 1.7 µm particle size) with an additional
   Waters Acquity VanGuard CSH C18 pre-column (5 × 2.1 mm id; 1.7 µm
   particle size) maintained at 65 °C was coupled to a Vanquish UHPLC
   System. To improve lipid coverage, different mobile phase modifiers
   were used for positive and negative mode analysis^[373]101. For
   positive mode 10 mM ammonium formate and 0.1% formic acid were used and
   10 mM ammonium acetate (Sigma–Aldrich) was used for negative mode. Both
   positive and negative modes used the same mobile phase composition of
   (A) 60:40 v/v acetonitrile:water (LC-MS grade) and (B) 90:10 v/v
   isopropanol:acetonitrile. The gradient started at 0 min with 15% (B),
   0–2 min 30% (B), 2–2.5 min 48% (B), 2.5–11 min 82% (B), 11–11.5 min 99%
   (B), 11.5–12 min 99% (B), 12–12.1 min 15% (B), and 12.1–15 min 15% (B).
   A flow rate of 0.6 mL/min was used. For data acquisition a Q-Exactive
   HF Hybrid Quadrupole-Orbitrap Mass Spectrometer was used with the
   following parameters: mass range, m/z 100–1200; MS^1 resolution 60,000:
   data-dependent MS^2 resolution 15,000; NCE 20, 30, 40; 4 targets/MS^1
   scan; gas temperature 369 °C, sheath gas flow (nitrogen), 60 units, aux
   gas flow 25 units, sweep gas flow 2 units; spray voltage 3.59 kV.

LC-MS data processing using MS-DIAL and statistics

   Untargeted lipidomic data processing was performed using
   MS-DIAL^[374]102 for deconvolution, peak picking, alignment, and
   identification. In house m/z and retention time libraries were used in
   addition to MS/MS spectra databases in msp format^[375]103. Features
   were reported when present in at least 50% of samples in each group.
   Statistical analysis was done by first normalizing data using the sum
   of the knowns, or mTIC normalization, to scale each sample. Normalized
   peak heights were then submitted to R for statistical analysis. ANOVA
   analysis was performed with FDR correction and post-hoc testing.

Reporting summary

   Further information on research design is available in the [376]Nature
   Research Reporting Summary linked to this article.

Supplementary information

   [377]Supplementary Information^ (3.3MB, pdf)
   [378]Peer Review File^ (1.2MB, pdf)
   [379]Reporting Summary^ (112.1KB, pdf)
   [380]Source Data^ (7.9MB, xlsx)

Acknowledgements