Three of the nine (33.3%) patients in cohort 1 (Table 1) had neonatal onset (onset within 1 month of age), and two (66.7%) of these had infantile death (death between 1 month and 1 year of age). One patient with neonatal onset and infantile death carried a frameshift variant (c.23_33del11) with a loss of function. In contrast, five other patients (5/9, 55.6%) had late onset (at age 4–8 years) and were alive at 14–28 years, suggesting a relatively mild manifestation of the disease. The most frequent presentations in this cohort were less life-threating and included central nervous system (CNS) symptoms, such as seizures (8/9, 88.9%), ataxia (6/9, 66.7%), and cognitive defects (5/9, 55.6%). In contrast, lactic acidosis (2/9, 22.2%), cardiomyopathy (2/9, 22.2%), respiratory distress or failure (3/9, 33.3%), and hypotonia (3/9, 33.3%) were the presentations of the patients who had early disease onset and death. Imaging indicated that cerebral or cerebellar abnormalities were common (8/9, 88.9%) in this cohort (Table 1). There was no sex bias in this cohort, as the female-to-male ratio was 5 to 4 (Table 1).

In contrast to cohort 1, 11 of the 14 (78.6%) patients in cohort 2 (Table 2) had neonatal onset, and 9/11 (81.8%) died in the first year of life, suggesting greater disease severity. The most frequent presentations in this cohort were multi-systemic, including severe lactic acidosis (10/14, 71.4%), respiratory failure or distress (11/14, 78.6%), cardiomyopathy (9/14, 64.3%), and hypotonia (7/14, 50.0%). Similar to the findings in cohort 1, these presentations were associated with early onset and death, suggesting that they are more life-threatening. CNS symptoms, such as seizures (5/14, 35.7%), ataxia (2/14, 14.3%), and cognitive defects (2/14, 14.3%) were less frequent, probably because these presentations are relatively late-onset and thus undetectable due to early death. The main imaging findings were cerebral and cerebellar abnormalities (7/9, 77.8%). In rare cases, abnormalities in the basal ganglia (1 case) and brainstem (1 case) were noted (Table 2). The female-to-male ratio in this cohort was 9 to 5, indicating a sex bias (Table 2), which suggests that male patients have greater disease severity and may have a higher prenatal lethality rate. However, seven of the nine patients with infantile death were females, suggesting that females were more sensitive to the disease (Table 2). Therefore, whether sex contributes to disease severity remains to be further investigated.

Nine patients in cohort 3 (Table 3) were homozygous for c.370G > A, and six patients were compound heterozygotes. Four of the nine (44.4%) homozygous patients had neonatal onset, but only 1 (25%) had infantile death. Among the five patients who were alive at the last follow-up, the oldest patient was ∼8 years old. The six compound heterozygous patients carried the c.370G > A variant with c.371G > T, c.402+1G > C, and c.533G > A variants. The c.371G > T variant resulted in a different amino acid change (p.Gly124Val) at the same codon as the c.370G > A variant (p.Gly124Ser). The patient with this genotype had neonatal onset and was alive at 3 years of age, which was the last follow-up. Similar to the patient with a loss of function frameshift variant (c.23_33del11) in cohort 1, three of four (75%) patients with the splicing variant c.402+1G > C had neonatal onset, and two (66.7%) of them had infantile death, suggesting that this splicing variant could be a loss of function variant, which is associated with greater disease severity. As was observed in cohort 2, the patient with the compound heterozygous c.370G > A/c.533G > A variants had neonatal onset and infantile death. The most prevalent presentations among the c.370G > A homozygotes and heterozygotes were similar and included multi-systemic symptoms, such as lactic acidosis (14/15, 93.3%), developmental delay (13/15, 86.9%), seizures (12/15, 80.0%), respiratory distress (10/15, 66.7%), cardiomyopathy (9/15, 60.0%), hypotonia (9/15, 60.0%), and dystonia (5/15, 33.3%). Similar to the variation in clinical presentations, imaging showed more diverse abnormalities in this cohort than in cohorts 1 and 2 (Table 3). The frequently affected tissues included the cerebrum (7/12, 58.3%), cerebellum (6/12, 50.0%), and corpus callosum (5/12, 41.7%). Noticeably, abnormalities in the basal ganglia and midbrain were also noted in five (5/12, 41.7%) patients, including two c.370G > A homozygotes and three c.370G > A heterozygotes, suggesting that a specific Leigh syndrome-like phenotype is correlated with this genotype (Table 3). Therefore, it can be concluded that the East Asian-specific c.370G > A variant conferred intermediate disease severity that was between the severity of the patients with variants in exons 1–4 and 5–7. Lastly, similar to cohort 1, the female-to-male ratios for c.370G > A homozygous and heterozygous patients were 5 to 4 and 3 to 3, respectively, indicating no sex bias.

In this cohort (Figure 1A), five patients initiated CoQ10 therapy at different ages and at different doses (1,000–∼5,000 mg/day), and four (80%) showed positive effects, including improvements in neurological condition and walking ability (Table 1). Compared to the responsive patients, the unresponsive patient carried variants in exon 4 and had a relatively early onset. As was observed in patients with the c.370G > A variant, the disease manifestations associated with changes in this exon may be more severe than those associated with variants in the upstream exons. Moreover, CoQ10 therapy was initiated relatively late (at 19 years of age). These could be the reasons for this patient’s unresponsiveness to CoQ10 therapy. Nevertheless, CoQ10 therapy was effective in most patients in this cohort.

In contrast to cohort 1, four patients in cohort 2 (Figure 1B) received CoQ10 therapy, and three (75%) were unresponsive, which is in accordance with the life-threatening presentations of patients in this cohort (Table 2). The responsive male patient had both a relatively late onset and early initiation of CoQ10 therapy, suggesting that disease severity and initiation of therapy are the essential determinants of CoQ10 therapy effectiveness.

Six homozygous patients in this cohort were treated with CoQ10 supplementation, and five (83.3%) (three females and two males) were responsive to the therapy (Figure 1C), suggesting that CoQ10 therapy effectively improved the health of these homozygous patients.

For the compound heterozygous patients with the c.370G > A variant, responsiveness to CoQ10 therapy depended on the other variant they carried. For example, the patient with the c.371G > T variant responded to CoQ10 therapy, which is in accordance with the similar clinical presentations of this patient and those of c.370G > A homozygotes. In contrast, none of the four patients carrying the c.402+1G > C variant benefited from CoQ10 therapy, which is also in accordance with the severe manifestations caused by this loss-of-function variant. These results further support the hypothesis that the effectiveness of CoQ10 therapy is associated with disease severity and the timing of therapy initiation.

Generally, the major clinical presentations caused by pathogenic COQ4 variants in exons 1–4 (i.e., amino acid changes in the N-terminus of COQ4) are CNS symptoms which are less life-threating, and are associated with late onset, responsiveness to CoQ10 therapy, and a relatively long lifespan. In contrast, the clinical presentations caused by pathogenic COQ4 variants in exons 5–7 (i.e., amino acid changes in the C-terminus of COQ4) included multi-systemic symptoms which are more fatal and are associated with early onset, unresponsiveness to CoQ10 therapy, and early death. Patients with the East Asian-specific c.370G > A variant displayed intermediate disease severity with multi-systemic dysfunction, which was between that of the patients with variants in exons 1–4 and 5–7. Specifically, the c.370G > A homozygous patients had early disease onset, were responsive to CoQ10 therapy, and a relatively long lifespan. For the c.370G > A heterozygotes, case with other allelic variant located in exon 4 had late onset, was responsive to CoQ10 therapy, and relatively long lifespan, whereas case with allelic variant located in exon 6 had early onset, was unresponsive to CoQ10 therapy, and early death. Moreover, loss-of-function variants, such as frameshift and splicing variants, were associated with early onset, unresponsiveness to CoQ10 therapy, and poor outcomes. Lastly, based on the current data, sex is unlikely to be associated with disease severity.

The mechanism underlying the exon-dependent genotype-phenotype correlation may be associated with the structure and function of COQ4. Yeast Coq4, which is a functional ortholog of human COQ4, is located at the inner mitochondrial membrane on the matrix side and interacts with other Coq proteins, such as Coq3, Coq5, Coq6, Coq7, and Coq9 (Belogrudov, et al., 2001; Marbois, et al., 2005; Awad, et al., 2018). Two functional motifs have been identified in the C-terminus of Coq4; one is a geranylgeranyl monophosphate lipid-bound long hydrophobic α-helix (PDB: 3KB4, Northeastern structural genomics program), and the other is a highly conserved putative zinc ligand motif [HDxxH-(x)11-E] (Marbois et al., 2009). Coq4 is thought to function as a scaffold or organizer that simultaneously anchors the inner mitochondrial membrane, CoQ synthetic complex, and long polyisoprenyl tail of CoQ intermediates and/or CoQ, to organize the ring modifications in the CoQ biosynthetic pathways, and the C-terminus of Coq4 is essential for maintaining these functions (Marbois et al., 2009; Rea et al., 2010; He et al., 2014). In the predicted structure of human COQ4 with Phyre2 (Kelley, et al., 2015), there are transmembrane helices in the C-terminus (Figure 2). Therefore, it is reasonable to speculate that the C-terminus of human COQ4 is likely also important for protein function; thus, COQ4 variants in exons 5–7, which are more likely to disrupt the function of the C-terminus, could cause more serious manifestations than variants in exons 1–4. Similar associations between pathogenic variants in specific domains and certain phenotypes have been also reported in other COQ genes. For example, missense variants within the COQ8 protein-specific KxGQ domain were shown to be associated with cortical and pyramidal tract dysfunction (Traschutz, et al., 2020). Nonetheless, more data, including a high-resolution protein structure and in-depth mechanistic investigations, are required to confirm this speculation.

Notably, a dominant mode of inheritance was suggested for two patients: a 3-year-old boy who carried a de novo heterozygous 3.9-Mb deletion affecting at least 80 genes, including COQ4, who displayed intellectual disability, encephalomyopathy, and dysmorphic features (Salviati, et al., 2012), and a 4-year-old girl with a heterozygous COQ4 c.483 G > C variant who displayed severe metabolic/mitochondrial deficits, extensive muscle damage, and lethal rhabdomyolysis (Romero-Moya, et al., 2017b). However, some speculative reasons may explain these exceptions. For the first patient, the presence of a second allelic variant, for example, a variant in a regulatory region, would have been missed by only sequencing the exon regions. Moreover, the clinical presentations could be attributed to, either fully or in part, other causative genes other than COQ4, since at least 80 genes were affected by the deletion. Similarly, the presence of a second allelic variant may have been missed in the second patient, which could explain why genetic rescue of mitochondrial and skeletal muscle impairment was observed with patient-origin induced pluripotent stem cells (iPSCs) when the c.483 G > C variant was corrected. However, as mentioned above, the phenotypic heterogeneity is high among COQ10D7 patients, even for siblings who share similar genetic backgrounds, suggesting that other genetic or environmental contributors are involved in the pathogenesis of COQ10D7. Moreover, CoQ10 is involved in several biological processes, but it remains unclear whether any of the clinical presentations are attributed to the impairment of other biological processes instead of MRC deficiency in COQ10D7. Therefore, it is possible that, in rare cases, clinical manifestations could be associated with hemizygous or heterozygous COQ4 variants.

In accordance with the role of CoQ10 in the MRC, reduced levels of CoQ10 and complex II + III activity in muscle or fibroblast samples are hallmarks of COQ10D7. In vivo assessments of the CoQ biosynthetic rate using cultured fibroblasts were available, which enabled us to discriminate primary deficiency from secondary deficiency (Rodriguez-Aguilera, et al., 2017). Although the detection of CoQ10 biosynthesis and MRC activity have some diagnostic value, their clinical utility for COQ10D7 diagnosis is limited because of the invasive sampling, relatively long turnaround time for cell culture, limited data regarding diagnostic accuracy, and the lack of a consensus or comprehensive evaluation of detection methodologies. Therefore, genetic testing, particularly high-throughput sequencing, either whole exome and genome-based, should be used as the first-line tool to diagnose COQ10D7 in a rapid and definitive manner, especially when point-of-care high-throughput sequencing is available (Sun, et al., 2020; Sanford, et al., 2019). In fact, all reported COQ4 variants were diagnosed using WES. However, it should be noted that, for new potentially pathogenic variants identified by genetic testing, evaluations of CoQ10 biosynthesis and MRC activity are necessary to confirm their pathogenicity (Brea-Calvo et al., 2021). Phenotypic rescue by CoQ10 treatment would also provide convincing additional evidence for the pathogenicity of a COQ gene variant (Romero-Moya et al., 2017a).

In addition to muscle and skin biopsies, measurements of biomarkers in plasma, serum, white blood cells, or urine show promise for clinical applications due to their less invasive sampling. Although an elevated lactate level is not specific for the diagnosis of COQ10D7, it is valuable for evaluating disease severity and prognosis. Another biomarker, serum CoQ10, was detected at relatively low levels in three COQ10D7 patients (Yu, et al., 2019; Caglayan, et al., 2019), suggesting that it could be used for the preliminary diagnosis of COQ10D7 when accompanied with clinical presentations. Moreover, assessment of plasma or serum CoQ10 is simple and rapid, and could provide the first clue and promote timely CoQ10 supplementation. However, more plasma and serum CoQ10 data are needed to establish reference intervals and to validate this concept. Although decreased plasma or serum CoQ10 is not COQ10D7 specific and is associated with all types of primary CoQ deficiency, patients with all types of primary CoQ deficiency could likely benefit from CoQ10 supplementation, and no detrimental side effects have been associated with this biomolecule (Hargreaves 2014). Nevertheless, if skin fibroblasts could be obtained less invasively, they would be an ideal option due to the analytical accuracy of their biochemical analysis and the ability of functional studies (Romero-Moya, et al., 2017b; Alcazar-Fabra, et al., 2021).

Clear evidence supporting the use of any specific drug for the treatment of mitochondrial diseases (MD) is lacking. However, based on numerous cases with positive responses and the lack of noticeable toxicity, exogenous CoQ10 is widely used for the treatment of MD, particularly primary CoQ10 deficiency (Hargreaves 2014; Orsucci, et al., 2019). Recently, ubiquinol was approved as an orphan drug for the treatment of primary CoQ10 deficiency (Hernandez-Camacho, et al., 2018). Although the delivery and utilization of exogenous CoQ10 is inefficient because of its high molecular weight and low aqueous solubility, when CoQ10 is administered at high doses or in specific formulations, CoQ10 levels are increased in all tissues, including the heart and brain (Zaki 2016). However, approximately 43% (9/21) of patients with COQ10D7 did not benefit from CoQ10 therapy. Due to inconsistent treatment regimens, including variations in formulation and dose, and the lack of data regarding in vivo CoQ10 levels, it is currently impossible to determine the conclusive reasons for unresponsiveness in these patients. However, according to the limited data available, the responsiveness to CoQ10 therapy was highly associated with the location of the pathogenic COQ4 variants and the severity of clinical presentation. A rational speculation was that CoQ10 is more likely to prevent further harm rather than reverse the present impairments. Therefore, unresponsiveness to CoQ10 therapy was more likely to be observed in patients with early onset and life-threatening manifestations. In general, CoQ10 therapy, when administered in time, would be effective, to various extents, for those patients with COQ4 variants in exons 1–4 in both alleles, including the c.370G > A homozygotes, thus highlighting the importance of early genetic diagnosis and intervention.

For responsive patients, the symptoms of multisystem dysfunction, including the CNS, were improved, although it is difficult for CoQ10 to penetrate the blood-brain barrier (Musumeci, et al., 2001; Lamperti, et al., 2003; Artuch, et al., 2006; Mancuso, et al., 2012; Mortensen, et al., 2014). It has been shown that ubiquinol has greater therapeutic effectiveness than ubiquinone due to the improved mitochondrial bioenergetics in the brain and reduced oxidative stress and astrogliosis (Garcia-Corzo, et al., 2014). However, the differences in the therapeutic efficacy of ubiquinol and ubiquinone for COQ10D7 treatment are unclear due to limited data. Since CoQ10 is a lipophilic molecule, its gastrointestinal uptake and assimilation may be inefficient (Masotta, et al., 2019). Therefore, CoQ10 derivatives of a more absorbable nature could be promising alternatives to enhance CoQ10 biosynthesis or bypass deficient steps in the CoQ10 biosynthetic pathway (Awad, et al., 2018). Recently, β-resorcylic acid, a structural analog of the CoQ precursor 4-hydroxybenzoic acid (4-HB), was shown to be a powerful therapy for patients with COQ9 or COQ7 variants, thus shedding light on the application of 4-HB analogs for other forms of primary CoQ10 deficiency, including COQ10D7 (Hidalgo-Gutierrez, et al., 2019). In addition to CoQ therapy, symptomatic treatments were also helpful to some extent for improving the health condition of patients with MD. In general, there is no substantial difference in symptomatic treatments for the same conditions between patients with MD and non-MD patients (Orsucci, et al., 2019). However, in terms of COQ10D7 treatment, the effectiveness of drugs, such as creatine, thiamine, riboflavin, and l-carnitine, requires further evaluation, although some of these drugs may be associated with positive outcomes in rare cases. Moreover, other drugs, such as valproate, linezolid, and aminoglycosides, should be used carefully because of their potential for mitochondrial toxicity (Mancuso, et al., 2012; Orsucci, et al., 2019). Notably, two c.370G > A homozygous patients died due to infections and associated complications (Ling, et al., 2019; Yu, et al., 2019), suggesting that infection management is crucial for COQ10D7 patients since they could lead to metabolic crises in patients with MD (Orsucci, et al., 2019).

Based on findings regarding genotype-phenotype correlations, in-depth functional investigations of COQ4 domains would be useful for the development of targeted drugs for COQ10D7 patients with specific variants. Similarly, targeted drugs could be developed based on an understanding of the presentations associated with impairment of other biological processes besides MRC. Revolutionary advances in multifaceted omics technologies, including genomics, metabolomics, proteomics, epigenomics, transcriptomics, and interactomics, will continue to expand our knowledge of mitochondrial pathology (Rahman and Rahman 2018) and provide us with the possibility of developing targeted drugs based on newly identified molecular phenotypes.

In addition to conventional therapies, targeted gene therapy is a novel and attractive prospect for COQ10D7. With the development of induced pluripotent stem cells (iPSCs) and gene editing technologies, in vitro and in vivo disease models can be established more readily to advance the development of gene therapy (Karpe, et al., 2021). For example, phenotypic rescue was achieved in patient-derived iPSCs and mouse models via rectification of pathogenic variants in amyotrophic lateral sclerosis and disruption of splicing regulatory elements in spinal muscular atrophy (SMA) using gene editing (Wang, et al., 2017; Son, et al., 2019; Lin, et al., 2020). A similar investigation was initiated for COQ10D7 in which genetic rescue of mitochondrial and skeletal muscle impairment was attained in iPSCs with a COQ4 c.483G > C variant (Romero-Moya et al., 2017a). Currently, three targeted drugs (nusinersen, onasemnogene abeparvovec-xioi, and risdiplam) based on gene replacement or splicing modification have been approved by the US Food and Drug Administration for use in the clinical treatment of SMA (Menduti, et al., 2020), confirming the promising clinical utility of targeted gene therapy. However, more comprehensive in vivo evaluations of the efficacy and safety of gene therapy for COQ10D7 are required because the genetic abnormalities in COQ10D7 are more heterozygous than those in SMA.

Whether conventional or targeted gene therapies are administered, a unified, sensitive, and valid design for measuring outcomes is a prerequisite for accurate evaluation of therapeutic efficacy (Mancuso, et al., 2017; Steele, et al., 2017; Koene, et al., 2018). Such comprehensive evaluations were lacking in previous COQ10D7 studies, partially due to the rapid deterioration of patients who have the disease. However, this should be emphasized in future studies, if possible. Randomized, double-blind, placebo-controlled, crossover trials are also essential for accurate evaluation of therapeutic efficacy in a subset of COQ10D7 patients, for example, patients who have COQ4 variants in exons 1–4 in both alleles and present with mild symptoms.