Clinical, biochemical, and genetic features
Overall, the SA MADD cohort displayed heterogeneous clinical presentations (Table 1 and Supplementary Material 1), together with the characteristic diagnostic urine metabolites associated with MADD (Table 2 and Supplementary Material 2). The variants identified, their location in ETFQO, and their pathogenicity according to the American College of Medical Genetics and Genomics (ACMG) are described in Table 3. Three distinct groups were observed, based on disease severity and response to treatment, as discussed below.
Severe, Rb-unresponsive MADD
The first group included three patients (P6, P8 and P12), who presented with the hallmark features of neonatal-onset MADD. Clinically, their symptoms were severe and progressive, and patients were metabolically and phenotypically unresponsive to treatment with Rb. The first case was a male neonate (P6), born prematurely (at 38 weeks) to parents of White SA ancestry, who presented within the first week of life with acute metabolic decompensation. Key clinical features included congenital cardiac abnormalities (pulmonary stenosis, a patent foramen ovale with a right to left shunt, and a large atrial septum aneurysm) and convulsions with early neonatal death on day 9 of life. Metabolic profiling indicated the typical MADD biochemical fingerprint [16], including dicarboxylic aciduria, increased 2-hydroxyglutaric acid, elevated glycine conjugates (short-, short-branched-, and medium-chain-related), increased disease-associated acylcarnitine conjugates, as well the characteristic increase in sarcosine on the amino acid profile. Both mitochondrial fatty acid oxidation and branched-chain amino acid catabolism were affected, correlating with previous studies on severe MADD cases [11]. The patient displayed a MADD-DS3 score of 30, further supporting a diagnosis of type I MADD. A then novel homozygous variant – c.[1067G > A] (p.[Gly356Glu]) – with in silico and structural evidence of pathogenicity (ACMG classification: likely pathogenic) was identified in this neonate [15]. The second patient was a female infant (P8), born at term to parents of mixed ancestry, who presented at birth with metabolic acidosis, hypotonia and feeding difficulties. Disease-specific metabolic markers were similar to those of P6, and the MADD-DS3 score of 21 was high. The patient succumbed at the age of three months. The homozygous c.[1067G > A] variant was subsequently identified; however, based on the lack of any reported congenital features (no post-mortem examination), a diagnosis of type I/II MADD was given. The third case was a female neonate (P12), born at term to parents of mixed ancestry. A urinary organic acid profile typical of MADD was confirmed by the hospital that made the diagnosis (no residual urine collected before or after treatment was available to re-analyse for the purpose of this study). Clinical features included metabolic acidosis, hypoglycaemia, hyperammonaemia, new-onset pancytopaenia, acute kidney injury, hyponatremia, and hypocalcaemia, and the patient succumbed at 14 days of age. Based on the absence of any reported congenital features (no post-mortem examination) and the high MADD-DS3 score of 24, a diagnosis of type I/II MADD was given. Whole-exome sequencing (WES) and segregation analysis revealed that the patient was compound heterozygous for c.[1067G > A];c.[976G > C] (p.[Gly356Glu];p.[Gly326Arg]). Variant c.[976G > C] is classified as a likely pathogenic variant of unknown significance according to the ACMG criteria and, to our knowledge, has been reported to occur in only one late-onset case of Chinese ancestry by Xi et al. as compound heterozygous with the common variant, c.[250G > A] (p.[Ala84Thr]; ACMG classification: pathogenic) [17].
Considering the treatment-unresponsive metabolic profile (P6, P12) and rapid clinical deterioration of patients P6, P8 and P12, it may be suggestive that c.[1067G > A] is a highly pathogenic variant. Its presence on both alleles, or its bi-allelic combination with another variant affecting the same protein domain (ubiquinone-binding domain) of ETFQO, appears to lead to insufficient enzymatic compensation for adequate ETFQO activity. While it is evident that L-carnitine supplementation facilitated the excretion of toxic organic acids as acylcarnitine conjugates in these patients, we hypothesise that the c.[1067G > A] and c.[1067G > A];c.[976G > C] genotypes result in an ETFQO protein of which the folding cannot be sufficiently rescued/stabilised by Rb treatment.
Moderate, variably Rb-responsive MADD
The second group included eight patients who presented with moderate, heterogeneous phenotypes, and showed a varying response to treatment. The onset of symptoms ranged between the neonatal period (P1), infancy (P2, P7, P9, P10), and childhood (P3, P11, P13) and all patients displayed the characteristic clinical features of MADD. These included metabolic decompensation (n = 5), muscle weakness (n = 4), muscle pain (n = 3), hypotonia (n = 5), neck flexor weakness (n = 5), susceptibility to fatigue (n = 2), restrictive ventilatory defect (n = 1), gastrointestinal involvement (n = 6), elevated serum creatine kinase (CK) (n = 2), recurrent infections (n = 2), lethargy (n = 2), cognitive disability (n = 2), delayed gross motor development (n = 2), migraine/paroxysmal headache (n = 3), seizures (n = 2), coma (n = 3), skeletal involvement (n = 2), and liver dysfunction (n = 4). Uncommon symptoms included ketosis at the time of metabolic crisis (n = 5) and Beevor’s sign (n = 1). Owing to the availability of data, the baseline urine organic acids of only five of the eight patients are reported, and the data of P7 represent the urinary organic acids present upon considerable decompensation near the time of demise3. At first presentation, patients P1, P7 and P9–P11 displayed an increase (to a variable extent) in the diagnostic urine organic acid markers associated with MADD, albeit less pronounced than that of P6, P8 and P12. Most of these patients had increased concentrations of urinary glutaric acid, ethylmalonic acid, dicarboxylic acids, and 2-hydroxyglutaric acid. The excretion of N-hexanoylglycine and, to a variable extent, branched-chain-related glycine conjugates were mostly observed. Moreover, all eight patients displayed increased disease-associated urinary acylcarnitines and elevated sarcosine. The biomarker assessment correlated with previous observations in moderate cases where FAO seems to be initially/mostly affected and branched-chain amino acid catabolism is influenced to a lesser extent [11]. It is important to note that the metabolic profiling was greatly dependent on the time of sample collection and that P3 and P9 had received L-carnitine treatment from a very early stage in their lives due to the prior diagnosis of a sibling with MADD. Apart from P7, the clinical symptoms, together with most of the urine metabolites, improved upon dietary adjustment in combination with treatment with L-carnitine (P9, P10), L-carnitine and Rb (P2, P3, P11, P13), or L-carnitine, Rb, and coenzyme Q10 (P1). Carnitine conjugation indicated that accumulating acyl-CoAs were being detoxified via the carnitine transportation system, which likely explains the less prominent MADD organic acid signature observed. By contrast, the metabolic response to treatment of P7 – who succumbed to a stroke at the age of 23 years – was more comparable to that of severe MADD, a finding that correlated with the severity of the clinical presentation as summarised in Table 1 and Supplemental Method 1.
WES and segregation analysis by Sanger sequencing revealed four compound heterozygous variants in this group of White SA-ancestry patients. These included: (i) c.[740G > T];c.[1448C > T] (p.[Gly247Val];p.[Pro483Leu]) in P13, (ii) c.[287dup*];c.[1448C > T] (p.[Asp97Glyfs*24];p.[Pro483Leu]) in siblings P2 and P3, and (iii) c.[1067G > A];c.[1448C > T] (p.[Gly356Glu];p.[Pro483Leu]) in P1, P7, P9, P10, and P11. The novel c.[287dup*] variant affects the third exon of ETFDH, leading to a premature stop codon, and is classified as likely pathogenic according to the ACMG criteria. The c.[740G > T] variant shares the same classification and has been reported only as a compound heterozygous variant along with the likely pathogenic c.[389A > T] (p.[Gly247Val]) variant in one late-onset MADD case of Chinese ancestry [18]. All variants identified in this group encode for highly conserved amino acids.
Once again, the MADD-DS3 scores confirmed that the disease burden increases when the c.[1067G > A] variant is present. However, in four of the five patients affected by this variant (P9, P10, P11 and P1), MADD was found to be very amenable to treatment. It is therefore reasonable to conclude that when c.[1067G > A] is encountered as a compound heterozygous variant along with a variant which affects a different protein domain of ETFQO (e.g., c.[1448C > T]), sufficient enzymatic compensation occurs to allow for adequate ETFQO activity. Based on the onset of disease, a diagnosis of either type II (P1) or type III MADD (P2, P3, P7, P9, P10, P11 and P14) was given to the patients in this group.
Mild, Rb-responsive MADD
The final group included three patients (P4, P5 and P14) of White SA ancestry, who presented later in life with mild and non-progressive (treatment-related) phenotypes. Clinically, their symptoms were heterogenous, but to some extent characteristic of MADD. The disease presentation included metabolic decompensation (n = 2), muscle weakness (n = 3), neck flexor weakness (n = 2), gastrointestinal involvement as the disease progressed (n = 1), elevated serum CK (n = 2), lethargy (n = 1), encephalopathy (n = 1), cerebral white matter abnormalities (n = 1), and liver dysfunction (n = 1), with two patients exhibiting ketosis. Initially, statin-induced myositis was suspected in P5 until hepatic features, including lipid deposits on the liver biopsy and raised transaminases prompted a metabolic work-up. P5 and P14 displayed urine metabolites associated with MADD, including dicarboxylic aciduria, raised 2-hydroxyglutaric acid (P5), acylglycine conjugates (with less prominent branched-chain-related conjugates), short- and medium-chain acylcarnitines as well as the presence of sarcosine on the amino acid profile. As indicated earlier in the moderate MADD group, FAO tends to be more affected compared to the branched-chain amino acid catabolism in the less severe cases [11], which correlated with our findings. Literature shows that plasma acylcarnitine profiling is typically most informative when diagnosing a late-onset MADD case, as organic acid profiling may only be remarkable at the time of a metabolic crisis or catabolic status induced by fasting. However, acylcarnitine assessments have been inconclusive in some cases, particularly if the patient has insufficient free carnitine available to promote conjugation [10, 16]. The latter has also led to false negative NBS results in mild MADD cases, as reported by Lin et al. [19]. In our study, P14 showed a free carnitine concentration below the limit of detection and normal butyryl/isobutyryl-, isovaleryl- and glutarylcarnitine levels in plasma. Interestingly, this patient presented with severe ketosis, as well as a prominent increase in 2-hydroxyglutaric acid and cis-4-decenedioic acid with unremarkable increases in glutaric acid and ethylmalonic acid.
The clinical and biochemical aberrations essentially normalised following dietary adjustment and treatment with L-carnitine, Rb and coenzyme Q10 (P5 and P14). P4, however, presented with only increased ethylmalonic acid. This patient, the mother of P2 and P3, underwent metabolic testing following the children’s diagnosis and was on L-carnitine treatment at the time of sample collection. Clinical-biochemical improvements, specifically after therapeutic intervention, have been observed in several cases of late-onset MADD, and our clinical and metabolic results (although only three cases were included) strongly correlate with previous investigations [10, 20].
Genetic analyses revealed that all patients harboured the known pathogenic, homozygous c.[1448C > T] (p.[Pro483Leu]) variant. This genotype has been reported in numerous other cases that, similarly to this group, presented with adult-onset, Rb-responsive MADD [8, 20]. The Rb-responsive nature of this variant is further corroborated in a study by Cornelius et al. [21], in which it was shown that ETFQO activity in c.[1448C > T] modified HEK293 cells could be restored from ~ 45% to ~ 85% (corresponding to an increase of ~ 50% to ~ 80% steady-state ETFQO) when moderately Rb-deficient cultures were treated with a saturating concentration of Rb. Therefore, considering the literature, together with the time of onset, severity and treatment response of this group, a diagnosis of type III MADD was given to P4, P5 and P14, despite moderate MADD-DS3 scores.
Allelic frequency spectrum and haplotypes
To determine the allele frequency of the variants identified, PCR–RFLP analysis was used to screen the two most frequently occurring variants, c.[1067G > A] and c.[1448C > T], in the four largest population groups in SA. The study yielded no homozygous or heterozygous genotypes for any of the variants in any of the population groups assessed so that the allele frequencies of the population groups investigated were calculated as < 0.00067% (African, White SA, and Indian ancestry) or < 0.00084% (mixed ancestry) (Table 3). All five variants identified in the cohort were subsequently compared to gnomAD (v2.1.1) [22] and the Human Heredity and Health in Africa (H3Africa) project [23]. Of these variants, only two were recorded on gnomAD, displaying exome allele frequencies of < 0.0001% (c.[1448C > T] and c.[740G > T]); hitherto, none of the variants has been identified in the H3Africa data. The absence of the variants in the above-mentioned population databases, together with the frequency at which they were identified in the SA cohort, indicates with high probability the pathogenicity of these five variants and supports their causative role in MADD.
Haplotyping was performed on those patients for whom sufficient DNA could be obtained. Upon recruitment, patients and their families were invited to self-report their ancestry and region of birth. Consequently, haplotyping was performed on 10 patients of White SA ancestry (P1–P6, P9–P11 and P13) and one patient of mixed ancestry (P8) born in various geographical areas across SA (including the central, north-eastern, north-western, south-eastern, and south-western provinces).
Apart from the siblings, P2 and P3, and their mother, P4, all patients were found to be unrelated down to the second degree. DNA samples from those patients harbouring the variant c.[1067G > A] (P1, P6, P8, P9–P11), displayed variable lengths of a shared haplotype on one allele, with a minimal overlapping region of 7.2 Mb (Fig. 1). Of these samples, the two with homozygous c.[1067G > A] genotypes (P6 and P8) exhibited homozygosity for the shared haplotype in the region of ETFDH. Similar results were obtained for the DNA samples of those patients harbouring the c.[1448C > T] variant [P1, P4 (including P2 and P3 due to their first-degree relation), P5, P9–P11, and P13], with a minimal overlapping region of 4 Mb (Fig. 2). Again, the two samples homozygous for c.[1448C > T] (P4 and P5), displayed homozygosity for the shared haplotype in the ETFDH region. These findings suggest the presence of two separate founder haplotypes on which the c.[1067G > A] and c.[1448C > T] variants arose in the White SA population. However, without access to additional control data from the same population(s), it is currently not possible to estimate the haplotype frequency with confidence.