ITPase Deficiency Causes Martsolf Syndrome With a Lethal Infantile Dilated Cardiomyopathy

Martsolf syndrome is characterized by congenital cataracts, postnatal microcephaly, developmental delay, hypotonia, short stature and biallelic hypomorphic mutations in either RAB3GAP1 or RAB3GAP2. Through genetic analysis of 85 unrelated “mutation negative” probands referred with Martsolf syndrome we identified two individuals with different homozygous null mutations in ITPA, the gene encoding inosine triphosphate pyrophosphatase (ITPase). The probands reported here each presented with a lethal and highly distinctive disorder; Martsolf syndrome with infantile-onset dilated cardiomyopathy. Severe ITPase-deficiency has been previously reported with infantile epileptic encephalopathy (MIM 616647). ITPase acts to prevent incorporation of inosine bases (rl/dl) into RNA and DNA. In Itpa-null cells, dI was undetectable in genomic DNA. dI could be identified at a low level in mtDNA but this was not associated with detectable mitochondrial genome instability, mtDNA depletion or biochemical dysfunction of the mitochondria. rI accumulation was detectable in lymphoblastoid RNA from an affected individual. In Itpa-null mouse embryos rI was detectable in the brain and kidney with the highest level seen in the embryonic heart (rI at 1 in 385 bases). Transcriptome and proteome analysis in mutant cells revealed no major differences with controls. The rate of transcription and the total amount of cellular RNA also appeared normal. rI accumulation in RNA – and by implication rI production - correlates with the severity of organ dysfunction in ITPase deficiency but the basis of the cellulopathy remains cryptic. While we cannot exclude cumulative minor effects, there are no major anomalies in the production, processing, stability and/or translation of mRNA. Author Summary Nucleotide triphosphate bases containing inosine, ITP and dITP, are continually produced within the cell as a consequence of various essential biosynthetic reactions. The enzyme inosine triphosphate pyrophosphatase (ITPase) scavenges ITP and dITP to prevent their incorporation into RNA and DNA. Here we describe two unrelated families with complete loss of ITPase function as a consequence of disruptive mutations affecting both alleles of ITPA, the gene that encodes this protein. Both of the families have a very distinctive and severe combination of clinical problems, most notably a failure of heart muscle that was lethal in infancy or early childhood. They also have features of another rare genetic disorder affecting the brain and the eyes called Martsolf syndrome. We could not detect any evidence of dITP accumulation in genomic DNA from the affected individuals. A low but detectable level of inosine was present in the mitochondrial DNA but this did not have any obvious detrimental effect. The inosine accumulation in RNA was detectable in the patient cells. We made both cellular and animal models that were completely deficient in ITPase. Using these reagents we could show that the highest level of inosine accumulation into RNA was seen in the embryonic mouse heart. In this tissue more than 1 in 400 bases in all RNA in the cell was inosine. In normal tissues inosine is almost undetectable using very sensitive assays. The inosine accumulation did not seem to be having a global effect on the balance of RNA molecules or proteins.


Author Summary
Nucleotide triphosphate bases containing inosine, ITP and dITP, are continually produced within the cell as a consequence of various essential biosynthetic reactions. The enzyme inosine triphosphate pyrophosphatase (ITPase) scavenges ITP and dITP to prevent their incorporation into RNA and DNA. Here we describe two unrelated families with complete loss of ITPase function as a consequence of disruptive mutations affecting both alleles of ITPA, the gene that encodes this protein. Both of the families have a very distinctive and severe combination of clinical problems, most notably a failure of heart muscle that was lethal in infancy or early childhood. They also have features of another rare genetic disorder affecting the brain and the eyes called Martsolf syndrome. We could not detect any evidence of dITP accumulation in genomic DNA from the affected individuals. A low but detectable level of inosine was present in the mitochondrial DNA but this did not have any obvious detrimental effect. The inosine accumulation in RNA was detectable in the patient cells. We made both cellular and animal models that were completely deficient in ITPase.
Using these reagents we could show that the highest level of inosine accumulation into RNA was seen in the embryonic mouse heart. In this tissue more than 1 in 400 bases in all RNA in the cell was inosine. In normal tissues inosine is almost undetectable using very sensitive assays. The inosine accumulation did not seem to be having a global effect on the balance of RNA molecules or proteins.

Introduction
It is 40 years since two brothers were reported with severely delayed neurocognitive development, spasticity, postnatal microcephaly, short stature, congenital cataracts and primary hypogonadism [1], characterising a disorder that is now termed Martsolf syndrome (MIM 212720). Warburg Micro syndrome (MIM 600118, 614225, 615222, 615663) is an overlapping condition that was described in 1993, which also has microphthalmia/microcornea, retinal dystrophy, optic nerve atrophy and intracranial malformations as clinical features [2]. 60% of cases referred to us with a diagnosis of Warburg Micro syndrome have loss-of-function mutations in either RAB3GAP1, RAB3GAP2, RAB18 or TBC1D20 [3][4][5][6]. 44% of Martsolf syndrome cases have mutations in RAB3GAP1 or RAB3GAP2, which perturb but do not completely abolish the expression or function of the encoded protein [7,8]. The relatively high proportion of unexplained cases in both syndromes indicates that there are likely to be more disease loci and/or causative genetic mechanisms to be discovered.
Infantile-onset dilated cardiomyopathy (iDCM) is a rare, aetiologically heterogeneous disorder that may present as acute, commonly lethal, and with cardiogenic shock [9].
Isolated iDCM may be caused by genetically determined primary abnormalities of heart muscle (sarcomere, Z-disc, desmosomes etc) while iDCM as a component of a multisystem disorder is most commonly secondary to an inborn error of metabolism ( Table 1) [10][11][12] with the prognosis being dependent on the underlying cause. Early genetic testing is recommended in iDCM as it may help direct the clinical management [13].
Here we report two families with a very distinctive clinical presentation of lethal iDCM and Martsolf syndrome associated with homozygous null mutations in ITPA which encodes inosine triphosphate pyrophosphatase (ITPase). ITPase is an enzyme that functions to prevent incorporation of inosine bases (rI/dI) into RNA and DNA by scavenging ITP/dITP in the cell. An autosomal recessive partial deficiency of inosine triphosphate pyrophosphatase (ITPase) has been recognised since the late 1960's via accumulation of inosine triphosphate (ITP) in erythrocytes [14]. This is a relatively common trait that is clinically asymptomatic although it may influence sensitivity to certain drugs [15]. The trait is caused by hypomorphic mutations in ITPA (the gene encoding ITPase) which affect splicing and/or protein stability [16]. Biallelic loss-of-function mutations in ITPA have recently been reported as the cause of an early infantile encephalopathy (EIEE35, MIM #616647) [17]. We present data testing and refuting various hypotheses regarding the molecular consequences of ITPase deficiency on the genome, transcriptome and proteome.

Clinical Information
In Family 4911 ( Figure 1A) a maternal uncle (4911 V:5) and aunt (4911 V:7) of the proband (VI:3) had been described in a clinical paper as Martsolf syndrome with a previously unreported association with an early-onset cardiomyopathy [18]. The proband in the present study, their nephew, died at the age of 2 years. No post mortem examination was carried out and the exact cause of his death could not be confirmed. Prior to his demise he had been noted to have postnatal onset microcephaly with severe delay in all aspects of his development. He had bilateral cataracts diagnosed at the age of 13 months. Generalised seizures began at the age of 14 months. He was noted to have small genitalia. A clinical diagnosis of Martsolf syndrome was made and he had a negative screen for RAB3GAP1, RAB3GAP2, RAB18 and TBC1D20. His elder brother, two maternal uncles and a maternal aunt all had a very similar pattern of problems ( Table 2) and all had died in early childhood with evidence of cardiac failure [18]. Serial echocardiograms in 4911 VI:3 had shown persistent but mild dilation of his left ventricle and he is assumed to have died as a result of the progression of his cardiac disease.
In family 5196 ( Figure 1B) the affected proband (5196 III:3) was a girl who died at the age of 4 years. She had previously been diagnosed as having Martsolf syndrome on the basis of profound developmental delay, failure to thrive, microcephaly, seizures and congenital cataracts. Screening of the known Martsolf syndrome and Warburg Micro genes was negative. She presented in severe cardiac failure and died shortly after this. She had not previously been suspected of having any cardiac disease. In addition to the known anomalies, a post mortem examination revealed marked dilation of the left ventricle with increased trabeculation and mild fibroelastosis (Figure 2A-C). Fatty infiltration was noted of the right ventricle ( Figure 2B). Neuropathology showed cerebellar atrophy (Figure 2F), microgliosis of dentate and olivary nuclei, vacuolation of white matter ( Figure 2E) with scattered axonal spheroids ( Figure 2H) and gliosis of the hippocampus. 5196 III:3 had two maternal uncles who died in infancy (5196 II:3 and 5196 II:4; Figure 1B) who were suspected of having the same disorder although no clinical details were available from either individual.

Whole Exome Sequencing and Cohort Resequencing
In an effort to identify additional causative genes for Martsolf (Figure 1). The alignment, read depth and estimated heterozygous SNP detection sensitivity of each individual is given in Table S1 [19].
Sequence variants were filtered using minor allele frequency (MAF) of < 0.001, plausibly deleterious consequence and bi-allelic inheritance.
In 4911 VI:3, ten rare homozygous variants ( Table 3) were reviewed by manual assessment of read quality using IGV2.3 software [20] and by Sanger sequencing of selected variants for their segregation in the family. Of these only a nonsense mutation, c.452G>A, p.Trp151* (rs200086262) in ITPA (NM_033453, MIM 147520) segregated within the family in a manner consistent for an autosomal recessive disease-causing mutation ( Figure 1A). This variant has been previously identified as disease associated [17] and is present in gnomAD (genome Aggregation Database) with a minor allele frequency of 0.0058%. In the affected individual from Family 5196 (III:3), homozygosity for an apparently unique 40bp deletion spanning the splice donor site of ITPA exon 7 was detected on WES. Subsequent Sanger sequencing confirmed this to be chr20 hg19 g.3202531-3202570del; c.456_488+7del. This deletion is likely to have been microhomology-mediated as a nine base pair perfect repeat is present at the 5' end of the deleted region and the genomic region immediately 3' to the breakpoint ( Figure 1D). Both parents (5196 II:1 and 5196 II:2) were heterozygous for this mutation. Western blotting of lysates from lymphoblastoid cell lines (LCLs) from 5196 III:3 and her mother showed that ITPA protein was completely absent in the cells derived from the affected girl ( Figure 1E). Sanger sequencing of ITPA in the remaining members of the cohort of 85 "mutation negative" families [8] revealed no further plausibly disease-associated mutations. The primers used for this analysis are given in Table S2.

Inosine ribobases (rI) are incorporated in RNA from an affected individual
ITPA encodes inosine triphosphate pyrophosphatase (ITPase) which hydrolyzes both inosine triphosphate (rI) and deoxyinosine triphosphate (dI) [21,22]. Its major function is thought to be to ensure the exclusion of these "non-canonical" purines from RNA and DNA in order to avoid transcript and genome instability. Complete deficiency of ITPase in all tissues would thus be predicted to result in an increase in the incorporation of rI and dI into RNA and DNA respectively. To test this we first purified cellular RNA from a lymphoblastoid cell line (LCLs) from 5196 III:3. This RNA was digested to single nucleotides and analysed using a combination of HPLC and mass spectrometry (HPLC/MS). Using this approach we found that rI was present in RNA at a level of 725±158 SEM nucleotides of rI per 10 6 nucleotides of AMP in 5196 III:3, a significantly higher level than in RNA from LCLs derived from either her father (17±11 SEM rI:rA x 10 6 ) or her mother (71±60 SEM rI:rA x 10 6 ) ( Figure   3A). This equates to approximately one rI base in every 5500 bases of RNA from the null LCL.

Inosine ribobases (rI) are incorporated in RNA from Itpa-null mouse ES cells and embryonic tissues
We generated Itpa-null mouse embryonic stem (ES) cells using CRISPR/Cas9 genome editing [23,24] (Figure 4A, primers encoding the guide RNAs are detailed in Table   S2). In these cells, rI was detectable in RNA at 1889 nucleotides rI per 10 6 nucleotides AMP ± 295 SEM ( Figure 3B).
To determine if there was a correlation between the level of rI in different tissues in vivo and the organs affected in the human disease, we generated mice heterozygous for Itpa null alleles using direct cytoplasmic injection of Cas9 mRNA and guide RNAs into zygotes. Heterozygous animals were crossed to generate Itpa-null embryos and wild-type littermate embryos. Both genotyping and Western blot analysis ( Figure 4A) were used to confirm the null status of each embryo ( Figure 4A). As in previously reported targeted inactivation of Itpa, we found reduced body size (with a proportionate reduction in heart size) in Itpa-null embryos [25] and no other obvious morphological differences compared to wildtype controls ( Figure 4B). The level of rI in RNA from Itpa-null hearts (10382 nucleotides IMP per 10 6 nucleotides AMP ± 2008 SD) was significantly higher than in either Itpa-null brain or kidney (p<0.05 and p<0.01 respectively, student's t-test) and equated to approximately one rI for every 385 bases of RNA ( Figure 3C). rI was present at very low levels in RNA derived from control tissues.

Inosine base incorporation is detectable in mtDNA but not genomic DNA
The bacterial endonuclease V (nfi/EndoV) cleaves DNA at dI bases creating nicks in the dsDNA [26]. Digestion of genomic DNA from control and Itpa-null ES cell lines using EndoV (New England Biolabs) followed by alkaline-gel electrophoresis revealed no measurable difference in migration between the samples ( Figure 3D). However, a small but reproducible increase in the EndoV sensitivity of mtDNA from Itpa-null ES cells as compared to that in controls was seen ( Figure 3E).
To assess whether this increased inosine incorporation was associated with increased instability of the mitochondrial genome (mtDNA), we used quantitative PCR (qPCR) to compare levels of mtDNA to levels of nuclear DNA. We also used long-range PCR (LR-PCR) of mtDNA to look for any increase in the frequency of mtDNA rearrangements. Neither assay showed any differences between Itpa-null cells and controls ( Figure 3F,G), or between Itpa-null tissues and controls ( Figure S1A-B). We used Ion Torrent sequencing to detect base substitutions and MinION sequencing to detect largescale mtDNA rearrangements amplified from control and Itpa-null kidneys and hearts. No differences between wild type and Itpa-null kidney or heart were detected (Figure S1C-D).
Using HPLC/MS analyses we were unable to detect dI in the digested DNA samples from 5196 III:3, 5196 II:2 or the Itpa-null mouse ES cells (data not shown). To assess secondary effects of low-level dI incorporation on genome stability, a commercial comet assay kit (Trevigen) was used. No increase in DNA strand breaks could be detected in 5196 III:3 compared to 5196 II:2 LCLs, or in Itpa-null compared to wild-type ES cells (Figure 3H,I).

Biochemical Assessment of Mitochondrial Function
To assess whether ITPase deficiency had any effect on mitochondrial function, we carried out metabolic tracer analysis on the ES cells using 13 C 5 -glutamine, and conducted functional histopathology on tissue samples. In the tracer experiments, 13 C-incorporation into fumarate and citrate was analysed ( Figure S1E). The (m+4) isotopologues of both metabolites indicated that ITPA-loss did not affect normal oxidative TCA cycle function and a low level of reductive carboxylation of oxoglutarate, as measured by citrate (m+5), was again minimally altered upon loss of ITPA. Functional histopathology on the tissue samples was adapted from analyses used in a clinical diagnostic setting. Samples were reacted for cytochrome c oxidase (COX) and succinate dehydrogenase (SDH) activities, with sequential COX-SDH histocytochemical analyses carried out in order to identify low level, focal COXdeficiency. No differences were seen between control and Itpa-null tissues and no COXdeficient cells were identified in Itpa-null heart ( Figure 4C).

Transcriptome and Proteome Analysis of Itpa null cells/tissues
We compared the transcriptome of control and Itpa-null mouse hearts using the Affymetrix MTA1.0 expression microarray. There was very strong concordance between transcript levels in control and Itpa-null samples when all loci were examined together ( Figure 5A,B) or when a subset of loci that have been implicated in dilated cardiomyopathy in mice or humans were examined separately ( Figure 5E). When specific cardiac disease genes were examined using ddPCR, modest reductions could be observed in Itpa-null heart tissue ( Figure 5G). However it was not possible to determine if reductions of this magnitude would significantly alter cardiomyocyte function and, more importantly, we could not distinguish whether these changes in transcript levels were primary effects or secondary to an early disease process in heart.
A generalized effect on transcription produced by increased inosine incorporation into RNA in Itpa-null cells might not be identified on transcriptome analysis if its effects on individual transcripts are proportionate. Therefore, in order to assess any changes in transcription rate or transcript stability, we labelled RNA transcribed over the course of 30 minutes with the ribonucleotide analogue 4-thiouridine (4sU). 4sU incorporation was assayed immediately after labelling, providing a measure of transcription rate, and then at subsequent time points, at which any changes in the rate of RNA turnover would be revealed ( Figure S2A). 4sU incorporation was assayed by biotinylation of its thiol group and then quantification of biotin using a fluorescence-based kit. When RNA was harvested immediately following treatment, 4sU incorporation appeared lower in Itpa-null cells than in control cells. However, this effect was not confirmed when 4sU in digested RNA was measured directly by HPLC (data not shown). Similarly, reduced incorporation was not seen when an alternative assay utilizing 5-Ethynyl Uridine was used ( Figure S2C). No differences in the 4sU content of control and Itpa-null cells were observed at the later time points, suggesting that inosine incorporation does not affect RNA stability globally (Figure S2A-B).
Label-free quantitative mass spectrometry was performed to examine the whole proteome in control and Itpa-null mouse heart tissue. Apart from the absence of ITPase, no significant differences were detectable on inspection of either the whole dataset or the dilated cardiomyopathy-associated proteins specifically (Figure 5C,E).

Discussion
Bi-allelic loss-of-function mutations were recently reported in ITPA in seven affected individuals from four families with early-infantile encephalopathy, a distinctive pattern of white matter disease evident on brain MR imaging, microcephaly and progressive neurological disease [17]. While no measurement of rI/dI incorporation into RNA or DNA was presented from these cases, the clinical and genetic evidence for causation was compelling in this group of children. Here we have shown that Martsolf syndrome with iDCM, is an allelic disorder. There is also evidence of phenotypic overlap between the disorders as one of the seven affected individuals reported by Kevelam et al. [17] had iDCM and three had early onset cataracts. Taken together with the existing mouse genetic data [25,27], these data strongly support an essential role for ITPase activity in development and maintenance of brain, eye and heart function in mammals. There were obvious candidate mechanisms for a cellulopathy associated with ITPase null state. First, instability of the nuclear genome induced by dITP incorporation into DNA (as seen in E coli [26]); second, instability of the mitochondrial genome via the same mechanism; third, inhibition of RNA polymerase II by rI (previously demonstrated in vitro [28]); fourth, instability of mature transcripts through EndoV-mediated degradation of rI-enriched mRNA; finally, induction of energy deficiency state due to biochemical perturbation of mitochondrial function. In this paper we have attempted to address each of these and failed to observe any single major effect. The differential incorporation of inosine bases between DNA and RNA is interesting. This may reflect the evolution of efficient DNA surveillance and repair mechanisms to deal with deamination of adenosine bases to form dI with the steady state for inosine in DNA being ~1 per 10 6 nucleotides [29]. This would suggest that even moderate increases in the incorporation of dITP into DNA in ITPase null cells are likely to be below the limit of detection for the assays used here.
In this regard it is significant that we could detect low-level dI incorporation into the mitochondrial genome. Importantly we could detect no effect on either the quantity or structural integrity of mtDNA from hearts from Itpa-null embryos as compared to controls ( Figure S1). The lack of evidence for a DNA-based mechanism taken together with the correlation of rI incorporation in RNA with organ severity suggested to us that there may be a transcriptomic mechanism of disease.
Enzymatic A-to-I editing in RNA is used to "recode" specific transcripts in a highly regulated manner [30] and this may explain why there is no rI-induced repair system for RNA. However, human EndoV [29] is capable of cleaving RNA at inosine bases [31,32] and over-incorporation of rI could lead to a generalized instability of the transcriptome. We found evidence of a reduction in the transcript abundance of some longer mRNA extracted from Itpa-null mouse embryonic heart. This effect is difficult to interpret given that it is a relatively minor change and is plausibly a secondary effect of the disease process rather than due to rI-induced RNA instability. One way to address this problem would be to create animals who are null for both Itpa and Endov and thus determine if loss of the ribonuclease activity would rescue any or all of the Itpa phenotypes. Although we have not reported the details here, our preliminary work using Itpa/EndoV double KO mouse embryos suggested no reduction in rI incorporation into RNA.
It seems probable that the major disease mechanism in severe ITPase deficiency related to either inosine base production or rI incorporation. It is not clear why the heart, brain and developing eye are more sensitive to the perturbation. A clear understanding of the disease mechanism is important as it may lead to therapies that will ameliorate the progressive cardiac and neurological effects of this rare but important disease.

Clinical samples and information
Our cohort consists of DNA samples from 85 families, submitted by referring clinicians for research screening for Warburg Micro syndrome and Martsolf syndrome [8]. Affected individuals in this cohort are negative for causative variants in the coding sequences of RAB3GAP1, RAB3GAP2, RAB18 and TBC1D20, the genes previously associated with these disorders. Informed consent has been obtained from all participating families, and studies were approved by the Scottish Multicentre Research Ethics Committee (04:MRE00/19).

Whole exome sequencing and analysis
DNAs from Families 4911 and 5196 (nuclear trios) were enriched for exonic sequence using kits indicated in Table S1 and sequenced using Illumina HiSeq technology as described previously [33]. Sequence reads were aligned to the GRCh37 human genome reference assembly with BWA mem 0.7.10 [34]. Duplicate reads were marked with Picard MarkDuplicates 1.126. Reads were re-aligned around indels and base quality scores recalibrated with GATK 3.3 [35]. Single nucleotide variants and small indels were called with GATK 3.3 HaplotypeCaller on each sample and GenotypeGVCFs to produce a raw variant call set. Variants were annotated using the Ensembl Variant Effect Predictor [36]. Statistics for alignment, read depth and estimated heterozygous SNP detection sensitivity for each individual are listed in Table S1. Sequence variants were filtered using minor allele frequency (MAF) of < 0.001, plausibly deleterious consequence and bi-allelic inheritance.

Sanger sequencing
PCR amplification of the coding exons of ITPA and intron-exon boundaries was carried out using flanking primers with M13 tags to facilitate later sequencing (see Supplementary   Table S2). Primers were designed using ExonPrimer software on the basis of the reference sequence NM_033453. Sequencing reactions were carried out with BigDye Terminator 3.1 reagents (Applied Biosystems), according to manufacturer's instructions. Sequencing data was analysed with Mutation Surveyor software (SoftGenetics).

Antibodies and Western blotting
Rabbit polyclonal antibody raised to full-length ITPA was obtained from Millipore.
Rabbit polyclonal antibody raised to an N-terminal portion of ITPA was obtained from LSBio.
Goat polyclonal antibody raised to β-tubulin was obtained from Abcam.
Cells were lysed in a buffer containing 0.5% (v/v) Nonidet P-40 in a solution of 150mM NaCl, 10mM EDTA and 50mM Tris-HCl (pH=7.5) to which a protease inhibitor cocktail (Roche) was added. Tissue samples were lysed directly in 1x NuPage LDS Sample Buffer (Thermo Fisher) containing 5% β-mercaptoethanol. SDS-PAGE and Western blotting were carried out according to standard methods. ECL 2 Western blotting substrate (Pierce) was used to produce chemiluminescent signal, HyperFilm ECL (General Electric) was developed using a Konica Minolta SRX-101A.

Detection of Inosine bases in hydrolysed RNA and DNA
Cellular RNA and DNA were purified using RNAeasy (Qiagen) and BACC2 (GE Healthcare) kits respectively. mtDNA isolation was performed using a mitochondrial DNA isolation kit (Abcam).
For analysis of nucleic acid composition by mass spectrometry, digestion to single nucleotides was carried out. A combination of either 50µg/ml RNAseA (RNA) or 20U/ml DNAseI (DNA)(Roche Diagnostics) respectively and 80U/ml NucleaseP 1 (Sigma) was used as previously described [37]. Both purification and digestion was carried out in the presence of a 20µM concentration of the adenosine deaminase (ADA) inhibitor deoxycoformycin (DCF)(Sigma). Digestions were carried out in a buffer containing 1.8mM ZnCl 2 and 16mM NaOAc, pH=6.8 at 37°C overnight. Nucleases were then removed with 10,000 MW cut-off spin columns (Amicon). Samples were loaded onto a ZIC-pHILIC column using a Dionex RSLCnano HPLC and the eluate was applied to a Q Exactive mass spectrometer in negative mode. The instrument was operated in tSIM mode and data were quantified using XCalibur 2.0 software.

For analysis of genomic and mitochondrial DNA composition by Endov-digestion and
alkaline-gel electrophoresis, DNA samples were treated with 10 U of Endonuclease V (NEB) with the supplied buffer for 2 hours at 37°C. DNA strands were separated by incubation at 55°C in loading buffer containing 3% Ficoll (type 400) and 300mM NaOH. Samples were separated on agarose gels (50mM NaOH, 1mM EDTA) with a solution of 50mM NaOH, 1mM EDTA used as running buffer. After electrophoresis, gels were neutralized and stained with SYBR Gold (Invitrogen).

Comet assays
Alkaline comet assays were carried out using the Trevigen CometAssay qPCR analysis was carried out on a LightCycler 480 (Roche). Amplification from mtDNA and gDNA was carried out using pairs of primers designed to amplify from the mtCO1 locus of the mitochondrial genome and the Gapdh locus of the autosomal genome.
Amplification from cDNA was carried out using primers designed to amplify from TTN, RYR2, TNNT2 and GAPDH cDNAs. Amplification from NKX2-5 was carried out using commercial TaqMan probes (Mm01309813_s1_Nkx2-5). PCR amplification with unlabelled primers was quantified through binding of specific mono color hydrolysis probes (Roche).
Data were analyzed using LightCycler 480 software version 1.5.0 (SP4) (Roche). Primers were designed using the Universal ProbeLibrary Assay Design Center and are listed in Supplementary Table S2.
Droplet Digital PCR (ddPCR) reactions were carried out according to manufacturer's instructions (Biorad). In each reaction, cDNAs were combined with a VIC-labeled TaqMan control probe, mouse GAPD or eukaryotic 18S (Life Technologies). Primers specific for target genes are as above. Droplets were generated using a Biorad QX200 or QX200AutoDG droplet generator, PCRs were carried out using a C1000 Touch Thermal Cycler, and droplets were analyzed on a QX100 Droplet Reader. The data were analyzed using Quantasoft software (QuantaLife).

Mitochondrial enzyme histochemistry
Tissue samples from Itpa-null embryos and littermate controls (e16.5-e18.5) were frozen in liquid nitrogen-cooled isopentane prior to sectioning. Cryostat sections were stained for individual activities of COX and SDH and also for sequential COX/SDH activity.

Label-free quantitative proteomics
Protein was extracted from embryonic mouse hearts in a buffer containing 8M Urea, 75mM NaCl and 50mM Tris, pH=8.4 by sonication at 0-4°C in a Bioruptor device (Diagenode) together with silica beads. Protein concentrations were quantified using a BCA assay (Pierce) and then 100µg of each sample was subjected to in-solution tryptic digest.
Samples were loaded onto a C18 column using a Dionex RSLC Nano HPLC and the eluate was applied to a Q Exactive mass spectrometer. The data were quantified using XCalibur 2.0 software.
Recombinant plasmids were verified by direct sequencing. For each targeted locus, the E14 ES cells were transduced with 1µg of each vector using the Neon system (Life Technologies) according to manufacturer's instructions. Cells were allowed to recover for 24 hours, then treated for 24h with 1 µg ml −1 puromycin in order to select for cells containing the px462 construct. To select single cells also containing the px461 construct, fluorescence activated cell sorting into 96-well plates was carried out using a FACSJazz instrument (BD Biosciences). Clonal cell lines were analysed by direct sequencing of targeted alleles and by Western blotting. Sequencing primers are shown in Supplementary Table S2. To facilitate sequence analysis, PCR products were cloned into pENTR/D-TOPO vectors prior to sequencing.
Cytoplasmic zygotic injection of wild-type Cas9 mRNA together with in vitro transcribed gRNAs was used to generate Itpa-null mouse embryos. This approach was also used to produce heterozygous-null animals used to establish transgenic mouse lines. The plasmid vectors described above were used as a template for PCR amplification together with forward primers incorporating T7 promoter sequences and a universal reverse primer (see Supplementary Table S2). RNA was synthesised using a HiScribe T7 High Yield RNA Synthesis Kit (New England Biolabs) according to manufacturer's instructions. DNA preparations were made from tissue samples using Viagen reagent (Viagen Biotech).
Genotyping was carried out by direct sequencing of targeted alleles and by Western blotting as above. Following initial genotyping of Itpa-null animals produced by crossing heterozygous-null founders, subsequent genotyping of transgenic lines was conducted through PCR analysis.

Optical projection tomography and morphometry
E16.5 mouse embryos were mounted in 1% agarose, dehydrated in methanol and then cleared overnight in a solution containing 1 part Benzyl Alcohol and 2 parts Benzyl Benzoate. Imaging was conducted with a Bioptonics OPT Scanner 3001 (Bioptonics, UK) using brightfield analysis to detect tissue autofluorescence for capture of anatomical and signal data (wavelengths: excitation at 425 nm, emission: 475 nm). The resulting data were reconstructed using Bioptonics proprietary software (Bioptonics, MRC Technology, Edinburgh, UK), automatically thresholded to remove background signal, then merged into a single 3D image output using Bioptonics Viewer software. Measurements of internal chest cavity diameter, maximum heart diameter, cardiac wall cross-sectional area and total heart cross-sectional area were taken for five embryos per genotype.

Supplemental Data
Supplementary data consists of two tables and three figures.      Supplementary Table S2.

Supporting Information Legend
There is a single supporting information file associated with this manuscript which contains two tables, two figures and the associated methods and references. Table S1 documents the technical metrics relating to the trio whole exome sequencing on families 4911 and 5196. Table S2 lists the oligonucleotide primers used for; i) sequencing the candidate human genes (including ITPA); ii) creation of genome editing reagents; iii) sequencing mouse Itpa; iv) mitochondrial genome analysis; v) quantitative RTPCR. Figure   S1 is titled "Sequencing of mtDNA in control and Itpa-null tissues and metabolic analysis of mitochondrial function in control and Itpa-null ES cells".