Characteristics of the ATAD3 genes and proteins.
Atad3 genes are conserved in multicellular organisms, but absent in unicellular organisms, including yeast. In the human genome, two genes encoding ATAD3 proteins, Atad3A and Atad3B, are located on chromosome 1 at locus p36.33. The ExPASy Molecular Biology Server revealed two alternatively spliced human transcripts derived from the Atad3A gene: Q9NVI7-2/ATAD3A (length, 586 amino acids) and variant Q9NVI7/ATAD3A-2 (length, 634 amino acids).
Characterization of ATAD3 transcripts in human U373 and HeLa cells by reverse transcription-PCR (RT-PCR) revealed that only ATAD3A (Q9NVI7-2) mRNA is expressed in these two cell lines (data not shown). The ATAD3A isoform is conserved in all mammals and Drosophila
. The Atad3B
gene is human specific and ATAD3B protein (648 amino acids; 72 kDa) is only expressed at very low level in specific human cell lines (23
). Computer-assisted structural predictions identified two coiled-coil domains (CC1, amino acids 85 to 115; CC2, amino acids 180 to 220) with high oligomerization probability within the N terminus of ATAD3A. The coiled-coil domains are followed by a predicted transmembrane segment (TMS; amino acids 248 to 264) positioned in the central part of the molecule (Fig. ). The C terminus harbors a conserved ATPase domain characteristic of the AAA+
-ATPase subfamily with ATP-binding and ATPase domains (Walker A and Walker B), including an SRH motif, sensor I and sensor II residues, and an Arg finger (Fig. ) (13
FIG. 1. Characteristics of the ATAD3A protein. (a) Schematic representation of human ATAD3A protein. The epitope domains for N-ter and C-ter antibodies, the two coiled-coil domains (CC1 and CC2), the predicted transmembrane sequence (TMS), and the ATP binding (more ...) Mitochondrial ATAD3A topology.
Differential solubility assays showed that ATAD3A behaves as an integral mitochondrial membrane protein. ATAD3A is resistant to carbonate and high salt extraction and is soluble only in buffers containing Triton X-100 (see below). Submitochondrial membrane vesicle fractionation by density gradient centrifugation showed that ATAD3A partitions in heavy-density fractions but is also present in intermediate and light fractions (Fig. ). These results suggest that ATAD3A is a protein anchored to the IM that can also interact with the OM (33
FIG. 2. Characterization of mitochondrial ATAD3A topology. (a) Sucrose-density gradient profile of mitochondrial membrane fragments. Fractions from top to bottom were analyzed by immunoblotting. Markers: Ant-1, an integral protein of the IM; porin, an integral (more ...)
To examine the submitochondrial localization of ATAD3A domains in more details, we analyzed the accessibility of the protein to trypsin digestion under various conditions. In isotonic buffer containing 0.28 M sucrose, ATAD3A was protected from proteolysis (Fig. ). In hypotonic swelling conditions, and in the absence of salt, ATAD3A became highly sensitive to trypsin (Fig. ). At a low trypsin/protein ratio, ATAD3A proteolysis generated two fragments sequentially, ATAD3A* and ΔN-ATAD3A (37 kDa). The ATAD3A* fragment that is recognized by both N-ter and C-ter antibodies was produced with kinetics that paralleled proteolysis of the OM protein TOM20 (Fig. ). This indicates that the first 40 amino acids are positioned close to the mitochondrial surface. At higher trypsin/protein ratios, the ΔN-ATAD3A fragment that is specifically recognized by ATAD3A C-ter antibody remained protected from digestion, whereas OPA1, an inner membrane space (IMS) protein that peripherally associated with the outer surface of the IM (18
), is fully degraded (Fig. ). Under the same conditions, prohibitin, an oligomeric transmembrane protein of the IM with a domain exposed to the IMS, is digested at much lower rate (Fig. ). To investigate whether resistance of the C-terminal ΔN-ATAD3 fragment to trypsin is dependent on its specific submitochondrial localization, we combined trypsin digestion with Triton X-100 treatment in hypotonic buffer. Under these conditions, full proteolysis of the ATAD3A protein was observed (Fig. ). Similarly, if mitochondria were preincubated with a low trypsin concentration in hypotonic swelling conditions to promote formation of the ΔN-ATAD3A fragment, subsequent addition of Triton X-100 induced full proteolysis of this fragment at very low Triton X-100 concentrations (Fig. ). At low trypsin concentrations, Triton X-100 does not increase OPA1 proteolysis as OPA1 is an IMS protein whose accessibility to protease does not require IM permeabilization (Fig. ). These results indicated that the ATAD3A C-terminal domain is localized in a matrix compartment.
When protease treatment of mitochondria was carried out in hypotonic buffer that contained 120 mM NaCl, different proteolysis patterns were observed (Fig. ). In the absence of Triton X-100, ATAD3A, and OPA1 were partially protected from proteolysis. The protective effect of NaCl is assumed to be the result of mitochondrial condensation by the salt which, in turn, promoted OM resealing and shrinkage of the IMS. In the presence of Triton X-100, ATAD3A and OPA1 proteolyses were restored. However, low-molecular-weight trypsin-resistant ATAD3A fragments could be observed that were not present in the absence of salt (Fig. ). Notably, an N-terminal 25-kDa fragment (*) remained visible even at high trypsin concentrations. To investigate whether salt protects ATAD3A from proteolysis via strengthened interactions between mitochondrial membranes, we combined trypsin digestion with digitonin extraction in an isotonic buffer in the absence or presence of 120 mM NaCl. A relatively low concentration of digitonin (0.65 to 0.75%) generate holes in and removes most of the OM, making IMS proteins accessible to cleavage by exogenous proteases (20
). However, contact sites between the OM and the IM can withstand this relatively harsh treatment (20
). It is thus expected that proteins localized at contact sites will be protected from proteolysis in condensed mitochondria. In the absence and in the presence of 120 mM NaCl, solubilization of the OM was detectable by full digestion of OPA1 at low trypsin concentrations (Fig. ). OPA1 was even more efficiently proteolysed in the presence of NaCl. Considering ATAD3A, in the absence of NaCl, the digestion profile was similar to that observed for mitoplasts under hypotonic conditions. However, in the presence of 120 mM NaCl, ATAD3A was protected from proteolysis (Fig. ). Only the first 40 amino acids on ATAD3A were accessible to trypsin, resulting in the production of a single ATAD3A* fragment recognized by both C-ter and N-ter antibodies. As digitonin also punctures the IM, thereby exposing matrix proteins to protease digestion (18
), full ATAD3A proteolysis occurred at high trypsin concentrations with no intermediate ΔN-ATAD3A fragment (Fig. h).
These data led us to conclude that ATAD3A is anchored into the mitochondrial IM and enriched in sites with the potential to form contacts with the OM. The N-terminal domain interacts with the inner surface of the OM and the C-terminal AAA ATPase domain localizes in a specific matrix compartment. Uncertainty remains regarding the exact localization of the 40 first amino acids, but the high sensitivity of this region to protease attack suggests that they may be located close to the mitochondrial surface (see also the discussion section and Fig. ).
FIG. 10. Model for ATAD3A topology deduced from trypsin digestion experiments. (a) Schematic representation of ATAD3A at contact sites between the OM and IM. ATAD3A is shown as a dimer, but higher oligomers are likely through association of the C termini AAA-ATPase (more ...) The C-terminal ATAD3A domain controls mitochondrial targeting and inner membrane insertion.
The complex topology of ATAD3A and its interactions with mitochondrial membranes, in addition to the absence of a predicted mitochondrial targeting signal, raises questions regarding the mitochondrial targeting and anchoring mechanisms. To identify the import signal for ATAD3A, we constructed several Myc-tagged truncated versions of the protein and examined their cellular localization (Fig. ). We first analyzed the 245-586 ATAD3A-Myc protein that carries the predicted TMS and the AAA domain. When expressed in U373 cells, this mutant protein localized exclusively to mitochondria (Fig. ). Solubility experiments demonstrated that the 245-586 ATAD3A-Myc protein behaved, like endogenous wild-type protein, as an integral mitochondrial membrane protein (Fig. ). Trypsin sensitivity experiments in hypotonic conditions demonstrated that the membrane-bound 245-586 ATAD3A-Myc protein was protected from protease attack but was digested in the presence of Triton X-100 (Fig. ). Hence, the C-terminal ATAD3A domain contains all of the information necessary for mitochondrial targeting, membrane insertion, and the positioning of the C-terminal AAA ATPase domain in a matrix compartment.
FIG. 3. Characterization of regions that determine mitochondrial targeting of ATAD3A. (a) Double immunofluorescence analysis of the 245-586 ATAD3A-Myc mutant in transfected U373 cells with anti-Myc (red) and anti-ATAD3-Cter (green) antibodies shows mitochondrial (more ...)
To investigate the contribution of the TMS to mitochondrial targeting, we next analyzed the effect of TMS (residues 246 to 264) on the localization of different N-terminal half mutants. We used truncated Δ50 N-terminal mutants because we found that the first 50 amino acids may participate in interactions with the mitochondrial outer membrane (MOM) (see below). When expressed in U373 cells, the Δ50-250 ATAD3A-Myc protein lacking the TMS localized diffusely in the cytoplasm and accumulated on perinuclear structures with no mitochondrial labeling (Fig. ). The Δ50-280 ATAD3A-Myc protein with the TMS showed the same immunostaining pattern (Fig. ). Further addition of 10 amino acids was sufficient to target the Δ50-290 ATAD3A-Myc protein to mitochondria, where the protein also promoted mitochondrial chain fragmentation (Fig. ). We concluded that both the TMS and an adjacent auxiliary import region cooperate for mitochondrial import.
Interaction of ATAD3A N terminus with the mitochondrial outer membrane requires the first 50 amino acids plus functional features provided by transoligomerization.
To analyze the mechanism of interaction between ATAD3A N terminus and the MOM, we used the 1-250 ATAD3A-Myc mutant lacking the TMS. When expressed in U373 cells, a large fraction of the 1-250 ATAD3A-Myc-tagged protein colocalized and cosedimented with the mitochondrial fraction (Fig. ). Indirect immunofluorescence analysis of transfected cells showed that a population of the 1-250 ATAD3A-Myc protein localizes along the mitochondrial tubules (Fig. ). Solubility experiments revealed that the 1-250 ATAD3A-Myc protein is peripherally associated with the mitochondrial OM. It was resistant to NaCl extraction but was solubilized by carbonate buffer and 2 M urea (Fig. ). Trypsin digestion experiments demonstrated that a 25-kDa fragment harboring the N-ter epitope was selectively resistant to proteolysis (*), whereas the C-terminus Myc tag epitope was digested (Fig. ). The same trypsin-resistant 25-kDa fragment was also observed for ATAD3A digestion in the presence of Triton X-100 and NaCl (Fig. ). This confirms that, upon interaction with mitochondrial OM, the ATAD3A N terminus can be protected from trypsin digestion. To further localize the OM interaction domain, we deleted the first 50 amino acids and found that this was sufficient to abrogate interactions of the Δ50-250 ATAD3A-Myc mutant protein with the mitochondrial OM (Fig. ). The 50 first N-terminal amino acids were nevertheless insufficient for reconstituting a functional mitochondrial OM interaction domain. Indeed, the 1-220 ATAD3A-Myc mutant did not interact or colocalize with mitochondria (Fig. ).
FIG. 4. Characterization of the ATAD3A N-terminal domain. (a) U373 cells transfected with 1-250 ATAD3A-Myc plasmid were double immunostained with anti-Myc antibody (green) and anti-ATAD3A C-ter antibody (red). Cells were observed under confocal microscopy at (more ...)
We could then demonstrate that oligomerization of the N-terminal domain of ATAD3A is probably required for MOM interaction. As expected from the presence of two coiled-coil domains, yeast two-hybrid experiments confirmed that the ATAD3A N terminus (amino acids 1 to 245) forms oligomers (Fig. ). The ATAD3A N terminus can also mediate formation of hetero-oligomers with the full-length ATAD3A cotranslated in rabbit reticulocytes (Fig. ). Dimerization requires the full-length coiled-coil domains, as revealed by the lost of heterodimerization of the 1-220 ATAD3A mutant (Fig. , lanes 1). Finally, cross-linking experiments with the hydrophilic cross-linker dithiobis(sulfosuccinimidyl propionate) (DTSSP) confirmed that the 1-250 ATAD3A-Myc protein does associate as a dimer when bound to the mitochondrial OM (Fig. ). Together, these results indicate that the N terminus of ATAD3A can physically interact with the mitochondrial OM and that this interaction requires the sequence comprising the first 50 N-terminal amino acids plus specific functional features provided by transoligomerization.
ATAD3A is required for normal cell growth and maintenance of mitochondrial morphology.
The high degree of similarity between the Drosophila
ATAD3A (dATAD3A) ortholog and human ATAD3A (70%), suggested that the cellular functions of ATAD3A might be conserved between humans and Drosophila
. As an initial approach to identifying the cellular function of ATAD3A, we disrupted dAtad3A
gene expression in Drosophila melanogaster
. To begin a mutational analysis of dATAD3A
, we searched different Drosophila
databases and identified a piggyBac transposon inserted in the dATAD3A
gene designated PBc05441 (42
). We confirmed by sequencing the flanking region that this insertion was located in the 5′-untranslated region of the dATAD3A
transcription unit, and it will thus be referred to as dATAD3Ac05441
. Homozygous mutants showed growth arrest during larval development, indicating that dATAD3A is required for organism growth (Fig. ). These results complement a recent study showing that Caenorhabditis elegans
ATAD3A is essential for mitochondrial activity and development (22
FIG. 5. ATAD3A is required for cell growth in Drosophila. (a) Size comparison of wild-type and dATAD3Ac05441 homozygous mutant larvae at 120 h of development. After this time point, dATAD3Ac05441 homozygotes fail to enter the pupal stage and die. (b and c) Clonal (more ...)
We then used the “Flp-out” recombination technique (31
) to generate clones of fat body cells expressing a mitochondrial green fluorescent protein (GFP) (32
) alone or in combination with an RNAi construct targeting dATAD3A (Fig. ). Fat body cells are determined during embryogenesis and show an enormous increase in cell size during larval stages without cell division. Recombined fat body cells in which dATAD3A expression was knocked down displayed strong size reduction, demonstrating that dATAD3A is required for cell growth in a cell-autonomous manner. Our genetic analysis is consistent with genomic analysis that identified dATAD3A
as a major gene positively regulated by the TOR signaling pathway involved in cell growth (19
). The mitochondrial GFP also revealed a more fragmented mitochondrial network in cells expressing dATAD3A RNAi (Fig. , lower panels).
ATAD3A regulates steroid biosynthesis at contact sites.
To investigate whether the growth phenotype associated with ATAD3A downregulation is linked with changes at contact sites and altered mitochondrial metabolism, we focused on mitochondrial cholesterol metabolism. In steroidogenic tissues, tropic hormones stimulate steroid biosynthesis from cholesterol through the activation of the cyclic AMP or the calcium messenger systems. The rate-limiting step in all steroidogenic pathways is the transfer of cholesterol from the mitochondrial OM to the IM, where the first step of the steroidogenic cascade, namely, the conversion of cholesterol to pregnenolone by the cytochrome P450 side chain cleavage (P450scc), occurs (7
). Cholesterol is imported into the mitochondrial IM with the assistance of a transduceosome complex that organizes at the contact sites (reviewed in reference 36
). To confirm that ATAD3A functions at contact sites, we investigated its contribution to steroid biosynthesis in the human steroidogenic cell line NCI-H295R, which retains several differentiated functions and is therefore useful for defining the cellular mechanisms regulating steroid production. In NCI-H295R cells only ATAD3A but not ATAD3B is expressed (Fig. ). A marked difference in the expression levels of ATAD3A in NCI-H295R and U373 cells also suggests that ATAD3 proteins are differentially regulated between these two human cell lines (Fig. ). Stimulation with angiotensin II (AII), a Ca2+
-mobilizing steroidogenic hormone, specifically enhanced ATAD3A level in NCI-H295R cells (Fig. ). In contrast, FSK, which directly activates the transduceosome complex, has no such effect (Fig. ). Similar results were obtained with primary bovine adrenocortical cell cultures stimulated with AII (not shown). These results suggest that ATAD3A could participate in the formation of contact sites modulated by hormonal stimulation for cholesterol import into the IM. A direct contribution of ATAD3A to cholesterol import was next confirmed by analyses of the effects of short-term ATAD3A downregulation on cortisol and aldosterone production in NCI-H295R cells stimulated with AII or FSK. Western blot analyses showed marked downregulation of ATAD3A protein in cells transfected with hATAD3A siRNA (Fig. ). Short-term ATAD3A downregulation does not induce major changes in mitochondrial network organization (Fig. ) and did not significantly impact the basal production levels of cortisol and aldosterone in resting cells (Fig. ). In contrast, the stimulatory effects of AII and FSK on cortisol and aldosterone syntheses were markedly inhibited (Fig. ). Taken together, these data suggest that ATAD3A is required for enhanced channeling of cholesterol for hormone-dependent steroidogenesis. They also strongly support the hypothesis that altered cell growth associated with ATAD3A downregulation could be linked with altered mitochondrial metabolism resulting from changes at contact sites.
FIG. 6. ATAD3A assists steroidogenesis in NCI-H295R cells. (a) Time course induction of ATAD3A in NCI-H295R cells stimulated with AII (10 nM). In the left lane is U373 cell extract. (b) NCI-H295R cells were not stimulated (ctl) or were stimulated with FSK (10 (more ...) ATAD3A controls the mitochondrial network in a manner dependent on the ATP-bound state.
Dynamic interactions between the OM and IM also play important roles in regulating mitochondrial fusion and fission processes (5
). To further support a role of ATAD3A at contact sites, we thought to analyze the short-term effect of dominant-negative ATAD3A mutants on mitochondrial network organization in U373 cells. To this end, we considered two remarkable features of the AAA+
ATPase subfamily of proteins, namely, their assembly into functional oligomers and the dependence of the active oligomers on the nucleotide-binding state and the ATPase activity of the AAA domain subunits (1
). It was expected that defective ATP-binding ATAD3A mutants or defective ATP hydrolysis mutants could function as dominant-negative proteins by interfering with normal oligomer functions. To test this possibility, we prepared Myc-tagged wild-type ATAD3A and mutant proteins with well-described AAA mutations in the Walker A and Walker B domains that, respectively, do not bind ATP (358K/E WA ATAD3A) or bind ATP but are defective in ATP hydrolysis (413E/Q WB-ATAD3A) (1
) (Fig. ). Coimmunoprecipitation studies confirmed that Myc-tagged ATAD3A constructs are competent to oligomerize with endogenous wild-type ATAD3A (Fig. ). Overexpression of these proteins in U373 cells revealed that WA ATAD3A produced immediate fragmentation of the mitochondrial network, whereas overexpression of wild-type ATAD3A (Fig. ) or WB mutant (Fig. ) produced no such phenotype. Quantitative indirect immunofluorescence analysis with C-ter ATAD3A antibodies demonstrated that a 4-fold expression of WA ATAD3A-Myc over endogenous wild-type protein was sufficient to fragment mitochondrial chains (Fig. ). These findings were reproduced with mouse NIH 3T3 cells (data not shown).
FIG. 7. ATPase-defective ATAD3A mutant disorganizes mitochondrial tubules in U373 cells. (a) U373 cells transfected with plasmids encoding wild-type ATAD3A-Myc or WA ATAD3A-Myc mutant as indicated were double immunostained with anti-Myc antibody (green) and anti-ATP (more ...) Mitochondrial fragmentation induced by dominant-negative ATAD3A mutants depends on the fission machinery.
To confirm that the rapid mitochondrion fragmentation phenotypes induced by WA ATAD3A mutants were not a secondary consequence of mitochondrial dysfunction, we first verified that in early transfected cells fragmented mitochondria could still be labeled with the membrane potential-dependent dye MitoTracker Red CMXRos (Fig. ). We also ensured that mitochondrial fragmentation did not correlate with leakage of cytochrome c
(not shown) and did not affect mtDNA nucleoid core organization (Fig. ). We finally showed that the fragmentation phenotype associated with WA ATAD3A mutant expression is linked to an activation of the cell fission machinery (Fig. ). In mammalian cells, mitochondrial fission is regulated by Drp1, a cytosolic GTPase (40
). Transfection of U373 cells with Drp1 siRNA induced a profound decrease in Drp1 protein (Fig. ) and inhibited mitochondrial fragmentation induced by WA-ATAD3A overexpression (Fig. ). Counting the number of WA-ATAD3A-expressing cells revealed that 60% had fragmented mitochondria, whereas this dropped to less than 10% in cells transfected with Drp1 siRNA. Together, these results suggest that the loss of function of ATAD3A at contact sites between the OM and IM is sensed by the cell fission machinery.
FIG. 8. Mitochondrial fragmentation induced by WA ATAD3A mutant depends on the mitochondrial fission machinery. (a) U373 cells transfected with plasmids encoding WA ATAD3A-Myc for 20 h were labeled with MitoTracker CMXros (100 pM) for 2 h prior to fixation with (more ...) ATP binding regulates interaction of the ATAD3A N terminus with the mitochondrial OM.
To investigate the molecular mechanisms associated with loss of function of the WA mutant, we first compared the partitioning of WA and WB mutants with submitochondrial membrane vesicles fractionated by density gradient centrifugation (Fig. ). The WB ATAD3A mutant was present in all fractions, whereas the WA mutant concentrated in intermediate and heavy fractions. This suggested that the WA ATAD3A mutant was less prone to interact with the OM, but the protein still localized in the vicinity of contact sites. Because interaction with the OM requires transoligomerization of ATAD3A N terminus (Fig. ), we compared the susceptibility of the wild-type ATAD3A and WA mutant to the hydrophilic cross-linker DTSSP (Fig. ). With control mitochondria, cross-linked ATAD3A oligomers migrated as a single band on SDS-PAGE, with the apparent molecular mass varying between 150 and 300 kDa as a function of the acrylamide concentration. In 6% PAGE cross-linked ATAD3A oligomers migrated in the position of the 150-kDa marker protein (Fig. , left panel). Mass spectrometry analysis of trypsin-digested peptides derived from cross-linked ATAD3A oligomers revealed that they were exclusively composed of ATAD3A peptides, indicating that they are exclusively constituted with ATAD3A molecules (data not shown). Under the same conditions, prohibitin, an oligomeric transmembrane protein of the IM, was not cross-linked with DTSSP (not shown). If purified mitochondria expressing wild-type ATAD3A-Myc are incubated in the presence of DTSSP, the Myc-tagged molecules become cross-linked as homo-oligomers and hetero-oligomers with the endogenous ATAD3A molecules (Fig. , right panel, lanes 1 to 3). Hetero-oligomers (*) could be identified because they migrated below the homo-oligomers and reacted with both anti-Myc and anti-ATAD3A C-ter antibodies (not shown). With mitochondria expressing the WA ATAD3A-myc mutant, hetero-oligomers cross-linked with endogenous ATAD3A were not observed at all (Fig. , right panel, lanes 4 to 6). Moreover, the WA mutant protein was less prone to cross-linking by DTSSP, although it spontaneously formed covalent dimers that can be reduced by DTT (results not shown). These results indicate that loss of function by the WA-ATAD3A mutant could be due to decreased transoligomerization of the N-terminal domains.
FIG. 9. Interaction of ATAD3A N termini with MOM is implicated in the control of the mitochondrial network. (a) Sucrose-density gradient profile of mitochondrial membrane fragments of cells transfected with WB-ATAD3A or WA-ATAD3A mutants. Fractions from top to (more ...)
The interaction of ATAD3A with the MOM requires both transoligomerization of the N terminus and the first 50 amino acids (Fig. ). To confirm that decreased interaction of the WA mutant with MOM could contribute to the mitochondrial chain fragmentation phenotype, we analyzed the effect of ectopic expression of the truncated Δ50 ATAD3A-Myc mutant on mitochondrial tubular network in U373 cells (Fig. ). Coimmunoprecipitation studies revealed that the Δ50 ATAD3A-Myc mutant supports the formation of a heterocomplex with endogenous wild-type protein (Fig. ). Indirect immunofluorescence demonstrated that a 4- to 5-fold expression of Δ50 ATAD3A-Myc over endogenous wild-type protein induced fragmentation of the mitochondrial chains, similar to that observed with the WA mutant (Fig. ). We concluded that the interaction of ATAD3A N termini with MOM is required for correct mitochondria homeostasis and is regulated by ATP.