ATPase family AAA domain-containing protein 3 (ATAD3) is a mitochondrial membrane protein from the family of ATPases associated with diverse cellular activities and conserved in metazoans. ATAD3 gene absence results in embryonic lethality in worms (Hoffmann et al., 2009), flies (Gilquin et al., 2010; Harel et al., 2016), and mice (Goller et al., 2013; Peralta et al., 2018), indicating that it plays a role in early developmental stages and may be essential for proper mitochondrial function. In hominids, ATAD3 has been duplicated twice to form an array of three paralog genes organized in tandem close to the telomere in chromosome 1p: ATAD3A, ATAD3B, and ATAD3C, whereas other species, such as fruit flies and mice, harbor only one gene (Li et al., 2014). ATAD3B differs from the ancestral paralog ATAD3A by having a C-terminal extension of 62 amino acids, which is caused by a mutation in the original stop codon, while ATAD3C seems to be a truncated gene, missing the first 70 amino acids of the protein (Li and Rousseau, 2012; Merle et al., 2012).

ATAD3A has been described within mitochondria as spanning both mitochondrial membranes with its C terminus facing the matrix and the N-terminal region in the outer membrane (Gilquin et al., 2010; Baudier, 2018). The N-terminal domain comprises two transmembrane domains (TM1 and TM2), two coiled-coil domains (Cc1 and Cc2) important for protein–protein interactions and ATAD3A oligomerization, and a proline-rich domain (PR) of unknown function (Hubstenberger et al., 2010). The C-terminal region of ATAD3A contains an ATPase domain in the mitochondrial matrix with two conserved Walker A and Walker B motifs for ATP binding and ATPase activity (Figure 1). As a member of the AAA+ ATPase family, ATAD3 is predicted to form hexameric ring structures (Frickey and Lupas, 2004).

Functionally, ATAD3A has been associated with several roles within mitochondria including regulation of the inner membrane structure, protein assembly (Peralta et al., 2018; Arguello et al., 2021), mitochondrial DNA (mtDNA) nucleoid organization (He et al., 2007; Gerhold et al., 2015; Desai et al., 2017), cholesterol trafficking, and lipid metabolism (Issop et al., 2015) among others (van den Ecker et al., 2015; Fang et al., 2010; Lang et al., 2020; Jin et al., 2018). However, the primary role of the protein remains unknown.

ATAD3A is considered one of the five most common nuclear genes associated with mitochondrial diseases in childhood (Frazier et al., 2021). Due to the extensive sequence homology among the paralogs ATAD3A, ATAD3B, and ATAD3C, the 1p36.33 region is prone to non-allelic homologous recombination (NAHR), resulting in copy number variants (CNVs). Disease-causing mutations in ATAD3A include duplications and deletions among the ATAD3 paralogs (Harel et al., 2016; Desai et al., 2017; Peeters-Scholte et al., 2017; Gunning et al., 2020; Frazier et al., 2021; Ebihara et al., 2022; Tawfik et al., 2023) and single nucleotide variants (SNVs) in the highly expressed ATAD3A gene (Cooper et al., 2017; Peralta et al., 2019; Dorison et al., 2020; Hanes et al., 2020; Muñoz-Oreja et al., 2024; Al Madhoun et al., 2019; Chen et al., 2023). The allelic spectrum of ATAD3A-associated diseases includes null, hypomorphic, and antimorphic alleles and the variants can be inherited and de novo. The SNVs reported to date in ATAD3A are depicted in Figure 1.

In 2016, pathogenic variants in ATAD3A were first associated with Harel-Yoon syndrome (MIM:617183), characterized by global developmental delay and muscular hypotonia, along with other features such as hypertrophic cardiomyopathy, optic atrophy, congenital cataracts, and peripheral neuropathy (Harel et al., 2016). Since then, ATAD3A variants have been associated with different neurodevelopmental disorders. The dominantly inherited heterozygous variant c.1064G > A (p.Gly355Asp) in the Walker A domain of ATAD3A was associated with hereditary spastic paraplegia and axonal neuropathy (Cooper et al., 2017). The heterozygous c.1396.C > T (p.Arg466Cys) missense variant, involved in the ATP hydrolysis, produces a form of neurological syndrome associated with optic atrophy (Muñoz-Oreja et al., 2024). Bi-allelic deletions of ATAD3 via NAHR and compound heterozygous variants (some combinations of deletions/truncating mutations with missense mutations) have been associated with neonatal lethal pontocerebellar hypoplasia, hypotonia, and respiratory insufficiency syndrome (PHRINL, MIM618810; Desai et al., 2017; Peeters-Scholte et al., 2017; Skopkova et al., 2023). Moreover, a dominant 68 kilobase (kb) de novo duplication in the ATAD3 locus was reported in 22 patients from 21 families, associated with a severe multisystemic disorder characterized by neonatal respiratory insufficiency, hypotonia, and cardiomyopathy, resulting in death in the first weeks of life (MIM:618815; Gunning et al., 2020; Frazier et al., 2021). These duplications generate an extra copy of ATAD3B and an in-frame ATAD3A/C fusion gene that forms a stable chimeric ATAD3A/C protein disrupting regular ATAD3 oligomerization. The extra copy of ATAD3B is not thought to play a role in the phenotype as healthy individuals and patients with distinct phenotypes were found to have benign duplications affecting only ATAD3B.

To functionally evaluate the potential pathogenicity of the ATAD3A variants several transgenic flies have been generated using the UAS-Gal4 system for tissue-specific expression (Table 1). The Drosophila melanogaster ortholog to the human ATAD3A is called Belphegor (bor), hereafter referred to as “Drosophila ATAD3” or “dATAD3.”

In 2016, Harel et al. studied the recurrent heterozygous de novo variant c.1582C > T (p.Arg528Trp), located in the ATPase domain of the protein (equivalent in Drosophila, p.R534W; Figure 1; Harel et al., 2016). This variant was found in five unrelated families associated with global developmental delay, axonal neuropathy, and hypertrophic cardiomyopathy. They demonstrated that ubiquitous (tub-Gal4 and Ubi-Gal4), pan-neuronal (n-syb-Gal4), and motoneuronal (D42-Gal4) expression of UAS-dAtad3^R534W resulted in complete embryonic lethality with no viable adult flies. Muscle-specific expression (C57-Gal4) led to approximately 90% lethality. In contrast, overexpression of the wild-type allele UAS-dAtad3^WT with the same Gal4 drivers consistently produced viable flies with no phenotype. On a cellular level, the p.R534W mutation induced a significant reduction of mitochondria in the ventral nerve cord, axons, and synaptic boutons, suggesting increased mitophagy. Transmission electron microscopy (TEM) showed that muscle tissue contained very few and small mitochondria with highly aberrant cristae and a substantial increase in autophagic intermediates. Overexpression of dAtad3^WT, however, resulted in the opposite phenotype, with large, elongated mitochondria. The authors proposed that ATAD3 may promote mitochondrial fusion or inhibit fission, while the mutation likely inhibits fusion and/or promotes fission.

In 2021, Yap et al. investigated five ATAD3A missense variants inherited in trans to loss-of-function (LOF) alleles and associated with distinct neurological phenotypes (Yap et al., 2021). The variants included c.150C > G (p.Phe50Leu), c.229C > G (p.Leu77Val), in the N-terminal of ATAD3A, c.508C > T (p.Arg170Trp) located between the coiled-coil domains; c.707G > T (p.Gly236Val) in the transmembrane domain, and c.980G > C (p.Arg327Pro) in the ATPase domain (Figure 1). The equivalent variants in Drosophila are dAtad3^F56L, dAtad3^L83K, dAtad3^R176W, dAtad3^G242V and dAtad3^R333P (Table 1). Using CRISPR/Cas9-mediated genome editing, the researchers integrated a gene cassette into the first intron of dAtad3 generating a LOF allele. All flies harboring this LOF allele and a null allele (PBac{PB}dAtad3a^c05496), resulted in a functional knockout. The lethality was rescued by UAS-dAtad3^WT expression using a pan-neuronal Gal4 driver (elav^C155-Gal4), confirming the severe LOF nature of the introduced cassette. The expression of dAtad3^L83V and dAtad3^R176W variants also rescued the developmental lethality caused by ATAD3A loss, suggesting that they are partial LOF alleles. The other 3 variants, dAtad3^F56L, dAtad3^G242V, and dAtad3^R333P failed to rescue lethality indicating that they are severe LOF alleles.

Further characterization of the dAtad3^L83V and dAtad3^R176W variants showed a decreased lifespan, and locomotion and flight defects. TEM analysis showed both mutations caused small mitochondria with cristae abnormalities and increased autophagic intermediates. Confocal images showed increased expression of the autophagic marker p62 in adult thorax muscles expressing dAtad3^L83V and dAtad3^R176W. The authors concluded that ATAD3A function is required for the homeostasis of mitochondrial dynamics and mitophagy. One possible mechanism proposed was through increased interaction with Drp1, as the coiled-coil domain was shown to interact with Drp1 promoting mitochondrial fission (Zhao et al., 2019). Moreover, this study revealed the functional importance of the N-terminal, coiled-coil, and transmembrane domains of ATAD3A.

Recently, Muñoz-Oreja et al. investigated the heterozygous p.Arg466Cys variant (p.R472C in Drosophila), affecting a conserved arginine finger crucial for ATP hydrolysis (Figure 1). Transgenic flies with UAS-dAtad3^R472C were created and crossed with various tissue-specific Gal4 drivers (Muñoz-Oreja et al., 2024). Ubiquitous expression resulted in complete lethality, likely through a dominant-negative mechanism. Expression in nervous and muscle tissue (Atad3-Gal4) or neurons (elav^C155-Gal4) was lethal as well. However, expression driven by eyeless-Gal4 (ey-Gal4), which is limited to the eyes and part of the brain, resulted only in partial lethality (65%), with the surviving flies exhibiting abnormal or missing eyes. Moreover, using the neuroblast-specific inscuteable-Gal4 (insc-Gal4) driver or the late-onset eye and neuronal driver glass multiple reporter-Gal4 (GMR-Gal4), viable flies expressing the mutant variant dAtad3^R472C were produced similarly to controls. Therefore, the Arg466Cys variant is highly deleterious unless expression is highly restricted. Expression of the p.Arg466Cys variant in neuroblasts led to the formation of membrane-bound cholesterol aggregates and increased lysosomal content. The cholesterol aggregates, detected by the reporter mKate-D4, co-localized with the lysosomal marker LAMP-GFP, suggesting that this excess of cholesterol is targeted to the lysosomal pathway. In agreement with the results obtained in Drosophila, patient-derived fibroblasts also exhibited membrane-embedded cholesterol aggregation in the form of membrane whorls and increased lysosomal content. Interestingly, flies expressing the p.Arg466Cys variant under the ey-Gal4 driver showed higher dependency on dietary cholesterol. A diet with reduced cholesterol significantly decreased the number of viable adults, and, by contrast, cholesterol supplementation on the diet enhanced viability. Based on these results, the authors propose a model where defective ATAD3 results in a mitochondrial cholesterol deficit that is attenuated by increasing the cytosolic cholesterol levels. This increased cholesterol would be a cellular compensatory response that leads to an aberrant aggregation in membranes that can cascade to lysosomal insufficiency contributing to the pathomechanism of the disease.

In mice, the ubiquitous disruption of ATAD3 was embryonic lethal (Peralta et al., 2018; Goller et al., 2013). To understand the in vivo function of ATAD3 in mammals we generated two different LOF mouse models: the ATAD3 skeletal muscle deficient mice (Peralta et al., 2018) and ATAD3-neuron deficient mice (Arguello et al., 2021). The main features observed in the animal models are summarized in Table 1.

Animal models for ATAD3/Atad3 dysfunction have become indispensable for studying ATAD3 functions in vivo and effectively recapitulate many features observed in patients’ cells with ATAD3 variants. In the last years, several advanced techniques have been used to study these features.

High-resolution approaches like TEM have provided critical insights into mitochondrial cristae structure. Both animal models and patient cells show disrupted mitochondrial cristae morphology (Peralta et al., 2019; Dorison et al., 2020) and mitochondrial fragmentation (Cooper et al., 2017). This aligns with ATAD3A’s interaction with other proteins like PROHIBITIN and LETM1, that are essential for maintaining cristae morphology (Arguello et al., 2021; Antonicka et al., 2020). Also, in human embryonic kidney (HEK) cells, ATAD3A showed a remarkably regular distribution across the mitochondrial membrane, a typical characteristic of scaffolding proteins (Arguello et al., 2021).

Furthermore, fluorescence microscopy plays a crucial role in the analysis of cholesterol metabolism. In patient cells harboring ATAD3 gene cluster deletions, or duplication resulting in the formation of an ATAD3A/C fusion protein, elevated levels of unesterified cholesterol have been identified by filipin staining (Desai et al., 2017; Gunning et al., 2020). The animal models for ATAD3 dysfunction also reflect patient findings in terms of altered lipid metabolism. In 2024, a study using the Drosophila model introduced the novel application of the cholesterol reporter mKate-D4, which enabled the detection of membrane-bound cholesterol in vivo (Muñoz-Oreja et al., 2024). This novel approach facilitates to detect this difficult to study area of the cell metabolism, the cholesterol metabolism, which has been previously implicated in the disease. The researchers demonstrated that Atad3 dysfunction leads to a compensatory increase in cellular cholesterol, resulting in its abnormal aggregation in membranes that can cascade into lysosomal insufficiency, which may contribute to the pathomechanism of the disease.

Aberrant organization of mtDNA was observed in ATAD3 patients’ fibroblasts by immunocytochemistry staining (Desai et al., 2017; Gunning et al., 2020; Muñoz-Oreja et al., 2024), but has not been studied in animal models. However, in mouse models, mtDNA depletion occurs after cristae disorganization, suggesting that ATAD3 affects mtDNA maintenance in vivo by stabilizing the mitochondrial cristae morphology (Peralta et al., 2018; Arguello et al., 2021).

As previously described, ATAD3 LOF in skeletal muscle tissue of mice resulted in a milder phenotype than the neuron model. This result highlights the existence of different compensatory mechanisms to counteract ATAD3 dysfunction in different tissues. Indeed, Frazier et al. also indicated this tissue specificity, where complex I activity was more profoundly reduced in cardiac tissue than in skeletal muscle or fibroblasts from ATAD3-deficient patients (Frazier et al., 2021).

However, species-specific variations observed in the ATAD3A gene structure present a potential limitation in the direct application of findings from animal models to human disease. To illustrate, consider the case of the NAHR-mediated duplication syndrome, which is characterized by pontocerebellar hypoplasia, seizures, and respiratory insufficiency (Gunning et al., 2020; Frazier et al., 2021). On a molecular level, the duplications typically result in a stable chimeric ATAD3A/C protein harboring 29 missense changes, including the previously referenced p.Arg466Cys variant, which is postulated to be a significant contributor to pathogenicity. However, the p.Arg466Cys variant itself is associated with a milder phenotype including syndromic dominant optic atrophy with neurological involvement (Muñoz-Oreja et al., 2024). This suggests that the variant is only partially responsible for the phenotype observed in the duplication syndrome. Nevertheless, ubiquitous expression of the Arg466Cys variant in Drosophila proved to be lethal. Thus, the animal model, which lacks the ATAD3B and ATAD3C genes, is unable to fully elucidate the underlying pathomechanisms in this case.

Overall, the results obtained from the Drosophila and mouse models, along with other in vitro studies, have yielded valuable insights into the function of ATAD3A. In summary, it has been demonstrated that ATAD3A LOF or CNVs in the ATAD3 locus result in mitochondrial dysfunction, due to the protein’s role in the structural organization of mitochondrial membranes and its impact on essential processes such as mtDNA maintenance and cholesterol metabolism. Nevertheless, the precise function of ATAD3A and the cellular mechanisms underlying ATAD3-associated disorders remain unclear and require further research. The integration of genomic, transcriptomic, proteomic, lipidomic, and metabolomic data, coupled with the application of novel advanced technologies, is a promising avenue for advancing our understanding of pathomechanisms and may also facilitate the identification of potential therapeutic targets.