ATPase family AAA domain-containing 3A is a novel anti-apoptotic factor in lung adenocarcinoma cells

Major features of lung adenocarcinoma (LADC) are rapid growth and high metastatic potential as well as resistance to irradiation and chemotherapy (Rosell et al., 2006). Cell proliferation and metastasis are regulated by the balance between growth factors and inhibitory molecules (Schaefer et al., 2007). Using suppression subtractive hybridization (SSH), microarray and hierarchical clustering to investigate gene expression patterns in patients with lung cancer, we found that hepatocyte growth factor (HGF) and HGF receptor (HGFR, or product of proto-oncogene met, MET) were highly expressed in advanced LADC patients who smoked. Cigarette smoking was further shown to be a key factor of disease progression and treatment failure (Chen et al., 2006). However, a portion of patients who did not respond well to therapies in Taiwan were women and nonsmokers (Sun et al., 2007).

We used the same strategy to identify genes that were highly expressed in LADC. We then subtracted this LADC-specific gene pool from smoking-related genes. The resulting genes were subcategorized for ATPase and GTPase. The genes, of which the enzyme activity was activated by receptors, such as hepatocyte growth factor receptor (HGFR), epidermal growth factor receptor (EGFR) and HER2/neu, were excluded. Using this procedure, we identified three genes, encoding for dynamin-related protein 1 (DRP1) (Chiang et al., 2009), mitofusin 2 (Mfn-2) (de Brito and Scorrano, 2008a) and the ATPase family, AAA domain-containing protein 3 (ATAD3) (Hubstenberger et al., 2008), which were upregulated in LADC. DRP1 and Mfn-2 are GTPases, and ATAD3 is an ATPase.

Three types of ATAD3 have been documented in the NCBI database (http://www.ncbi.nlm.gov): a 66-kDa ATAD3A (BC033109), a 72.6-kDa ATAD3B (NM_031921) and a 46-kDa ATAD3C (NM_001039211; the differences in protein sequences among ATAD3A, 3B and 3C are summarized in supplementary material Fig. S1A). Although protein sequence alignment indicates that ATAD3A and 3C are truncated isoforms of ATAD3B, they are encoded by different genes located in non-overlapping regions on chromosome 1 (supplementary material Fig. S1B,C). Moreover, ATAD3A contains 16 exons, 3B 15 exons and 3C 12 exons, indicating that ATAD3A and 3C are not alternatively spliced variants of ATAD3B.

Using autoantibody-mediated identification of antigens (AMIDA), Gires et al. detected overexpression of KIAA1273/TOB3 (ATAD3B) in patients with head and neck cancer (Gires et al., 2004). Applying phage display to probe tumor-associated antigens, Geuijen et al. identified ATAD3A in acute myeloid leukemic (AML) blasts (Geuijen et al., 2005). Inhibition of ATAD3B expression by siRNA increased apoptosis (Schaffrik et al., 2006), but whether ATAD3B or ATAD3A were directly involved in programmed cell death, was not clear. In this study, we determined the expression level of ATAD3A in LADC cells and pathological specimens. The correlation between ATAD3A expression and patient survival was evaluated statistically. The effect of ATAD3A on cell growth and apoptosis was characterized in vitro.

Expression of ATAD3A was examined by RT-PCR in one HeLa and eight lung cancer cell lines. ATAD3A was detected in all cell lines (Fig. 1A). In eight pairs of lung cancer biopsy samples, overexpression of ATAD3A was detected in six LADC samples (Fig. 1B). Following sequence analysis, which was performed using fluorescently labeled dideoxy nucleotides (Mission Biotech, www.missionbio.com.tw, Taipei, Taiwan), and a DNA sequencing ladder read using an ABI PRISM 3700 DNA Analyzer (CD Genomics, Shirley, NY), nucleotide sequence homology of cDNA fragments from the nine cell lines and four LADC specimens was searched using a web program (http://blast.ncbi.nlm.nih.gov/Blast.cgi). They matched that of ATAD3A: NM_033109, Homo sapiens ATPase family, AAA domain containing 3A (ATAD3A). No mutation was detected (GenBank, BankIt1285471, GU189416). ATAD3B and ATAD3C were not detected by RT-PCR or DNA sequencing in LADC cells (data not shown).

Following determination of specificity and sensitivity, monoclonal antibodies were used to detect ATAD3A expression in LADC cells. A 66-kDa protein band corresponding to the anticipated molecular mass of ATAD3A was recognized in all the cell lines (Fig. 2A). A 70-kDa protein was only highly expressed in H23 and H2087 cells, but it was not detected in A549, HeLa or mouse embryonic fibroblasts. To determine the identity of the two proteins, cell lysate from H23 was immunoprecipitated and the respective protein band was subjected to analysis using matrix-assisted laser desorption/ionization-time-of-flight mass spectrometry (MALDI-TOF). The peptide mass fingerprints of both the 66-kDa and 70-kDa proteins matched (MS-Fit search; http://prospector.ucsf.edu/): those of the full-length ATAD3A: GenBank CAI22955.1, Homo sapiens, ATPase family AAA domain-containing 3A. The matched peptides covered 36% (211/586 amino acids) of the protein (supplementary material Fig. S2B).

Both the 66- and 70-kDa proteins were characterized by MALDI-TOF. The peptide mass fingerprints of both the 66-kDa and 70-kDa proteins matched that of AATD3A: MS-Fit search (http://prospector.ucsf.edu/): CAI22955, ATPase family, AAA domain containing 3A [Homo sapiens] (supplementary material Fig. S2B); however, they covered only 25.0% (167/648 AAs) of the ATAD3B. Moreover, three MALDI-TOF resultant fragments did not match with ATAD3B (the mismatched sequences are shown in supplementary material Fig. S2C,D). These data indicated that both 66-kDa and 70-kDa proteins were ATAD3A (CAI22955), and that the 70-kDa protein could be a post-translationally modified ATAD3A in LADC cells. The 85-kDa protein, which was determined by MALDI-TOF, was identified as DRP1, suggesting that ATAD3A interacted with DRP1 (supplementary material Fig. S2B1 and Fig. S2B3).

Immunocytochemical staining showed that ATAD3A was abundantly present in cytoplasm. The granular appearance of subcellular structures suggested that ATAD3A could be present in mitochondria (Fig. 2B). A MitoTracker Red CMXRos (Molecular Probes, Eugene, OR) uptake assay and confocal fluorescence immunocytochemistry (Fig. 2C) as well as immunoblotting of sucrose gradient-separated organelle fractions (Fig. 2D) confirmed that ATAD3A was localized in light membrane and mitochondria-associated membrane fractions of endoplasmic reticulum (ER) as well as in mitochondrial fraction. In addition to ATAD3A, apoptosis-inducing factor (AIF) and glucose response protein (GRP) 78 were detected in the same fractions. The results showed that as suggested by the web prediction program (http://www.ch.embnet.org/software/TMPRED_form.html), ATAD3A (BC033109) carried a transmembrane domain with a coiled-coil domain exposed to the cytoplasm which could interact with DRP1 and/or Mfn-2 (supplementary material Fig. S2E and Fig. S3A-C).

Using immunoblotting, we identified that the major type of ATAD3 expressed in LADC specimens was the 66-kDa ATAD3A (Fig. 3A). Immunohistochemistry detected ATAD3A in 93 (86.9%) of the pathological samples from patients with LADC. The signal was predominantly localized in cancer cells (Fig. 3B1), but not in non-tumor lung tissue (NTLT; Fig. 3B2). ATAD3A expression was also detected in 89.3% (50/56) of metastatic lymph nodes (data not shown). Statistical analysis showed that overexpression of ATAD3A in tumors correlated with tumor stage and lymphovascular involvement (Table 1), suggesting that ATAD3A expression could be associated with the metastatic potential of LADC. Interestingly, among the 93 patients who had high levels of ATAD3A, 39 (41.9%) patients had tumor recurrence during follow-up examination. Among the 14 patients who had low levels of ATAD3A, three had tumor recurrence (21.4%). All 42 patients who had recurrence developed tumors within 24 months of the operation. The risk of recurrence for patients with high levels of ATAD3A was 3.01-fold higher than that for patients with low levels of ATAD3A (P=0.045). Survival of patients with low ATAD3A levels was significantly better than that of patients with high ATAD3A levels (Fig. 3C). The hazard ratio between these two groups was 2.415, and the difference in cumulative survival was significant (P=0.0027). Multivariate analysis, however, revealed that the difference in ATAD3A expression between the two groups was marginal (P=0.052).

As shown previously, ATAD3A appeared as a 66-kDa form in A549 and HeLa cells as well as in mouse embryonic fibroblasts. The 70-kDa form, however, was only highly expressed in H23 and H2087 cells. In H1437, H226, H838 and H2009 cells, expression level of ATAD3A varied. Moreover, the amino acid sequence of the 70-kDa protein, which was frequently detected in lung cancer cell lines, but not in pathological specimens (Fig. 4A, also refer to Fig. 2A and Fig. 3A), matched that of ATAD3A (supplementary material Fig. S2B-D), suggesting that the 70-kDa protein might not be an ATAD3B, but a phosphorylated ATAD3A. To resolve the issue, we ran the protein sequence of ATAD3A through a web NetPhos program to predict for phosphorylation sites (http://www.cbs.dtu.dk/services/NetPhos/), and a NetPhosK program to predict specific kinase (http://www.cbs.dtu.dk/services/NetPhosK/) in eukaryotic proteins. The results showed that for ATAD3A (BC033109), the most probable kinase was protein kinase C (PKC) at the possible phosphorylation sites Thr335, Thr338, Thr359, or PKA at Thr118 (supplementary material Fig. S4A). For ATAD3A_L (NM-018188) and ATAD3B, the most probable kinase was PKA (supplementary material Fig. S4B,C). The most probable kinase for ATAD3C was PKC at position Thr184 (supplementary material Fig. S4D). However, since this segment is located inside the membrane of the prospective vesicle, the prediction might not be applicable. When the cell lysate of H23 cells was treated with calf intestinal alkaline phosphatase (CIP) before immunoblotting, signals of the 70-kDa protein band reduced markedly (Fig. 4B1). Treatment with CIP also reduced the 66-kDa protein to 63 kDa, suggesting that both the 66-kDa and 70-kDa proteins were phosphorylated. When ATAD3A antibody-precipitated proteins were probed with antibodies specific to phosphoserine/threonine (S-P/T-P) or phosphotyrosine (Upstate, Millipore Corporate, Billerica, MA), both 66-kDa and 70-kDa protein bands were positive for S-P/T-P (Fig. 4B2), but not phosphotyrosine (data not shown), indicating that the phosphorylated residues in both the 66-kDa and 70-kDa proteins were at serine or threonine.

To search for the kinase that is responsible for ATAD3A phosphorylation, we treated H23 cells with a panel of kinase inhibitors. As shown in Fig. 4C1, only addition of calphostin C, a protein kinase C (PKC) inhibitor, reduced the intensities of the 70-kDa and 66-kDa protein bands, indicating that PKC is the major kinase for ATAD3A phosphorylation. The results excluded a possibility of PKA in ATAD3A phosphorylation. Since calphostin C treatment reduced 70-kDa and 66-kDa proteins in a dose-dependent fashion (Fig. 4C2), the data also suggested that phosphorylation was essential for maintaining ATAD3A stability. To identify the PKC isozyme that phosphorylated ATAD3A, A549 cells, which only expressed 66-kDa ATAD3A, and H23 cells were transfected with plasmids carrying various isozyme of PKC genes. Interestingly, level of the 70-kDa proteins increased in cells that ectopically expressed PKCγ, PKCι and PKCζ (Fig. 4D1,2). These results confirmed that PKC was responsible for ATAD3A phosphorylation.

Because PKC activity is associated with growth factor receptor-related cellular events and cell cycle progression (Hirai et al., 1989; Black, 2000), the fact that PKC is involved in ATAD3A phosphorylation suggests that ATAD3A expression may be regulated by growth factor receptor- or cell cycle progression-related pathways. We therefore analyzed the expression pattern of ATAD3A during cell cycle progression by double-thymidine block (DTB) and serum starvation-reactivation methods. Results of DTB and release showed that levels of ATAD3A increased in S phase (4 hours after release from DTB) and decreased in the G1 phase (16 hours after release from DTB; Fig. 5A), indicating that ATAD3A biosynthesis was at S phase of cell cycle progression. Surprisingly, serum starvation increased levels of ATAD3A as well. The increase was dose (Fig. 5B1) and time dependent (Fig. 5B2). Moreover, increase of ATAD3A during serum starvation was associated with increase of cisplatin resistance (Fig. 5C). However, serum starvation did not increase levels of ATAD3A mRNA, but it did increase expression of a panel of metastasis- and angiogenesis-related genes, including those for HGF, vascular endothelial growth factor-B and matrix metalloproteinases (Table 2). Increase of AATD3A protein level could be a translational activation. The results showed that serum starvation induced growth arrest of cells at G1 phase and down-regulated expression of replication-related genes, which reflected reduced DNA damage and less drug toxicity (Chow and Ross, 1987; Chen et al., 2008). The loss of function effect of ATAD3A on drug sensitivity is yet to be resolved.

As anticipated, silencing of ATAD3A expression by siRNA (Fig. 6A1) increased cisplatin sensitivity (Fig. 6A2). Moreover, knockdown of ATAD3A (ATAD3A^kd) expression increased mitochondrial fragmentation (Fig. 6B1), and decreased colocalization of the ER [ER was visualized by ER retention signal KDEL-conjugated green fluorescence protein (GFP); Fig. 6B2, upper row] and mitochondria (Fig. 6B2, upper row, red fluorescence). Silencing of ATAD3A expression reduced the total amount of mitochondria as well, but it increased the amount of enlarged ER (Fig. 6B2, center row). These results suggested that ATAD3A could be detected on both ER and mitochondria, and like mitofusin-2 (Mfn-2) and DRP1, ATAD3A could be involved in mitochondrial shaping, i.e. fission and fusion, as well as communication between ER and mitochondria (Fig. 6B2, upper row). Interestingly, knockdown of DRP1 (DRP1^kd) also increased the amount of prominently enlarged ER (Fig. 6B2, bottom row). The presence of prominently enlarged ER in DRP1^kd cells was confirmed by electron microscopy (Fig. 6B3A,B). In ATAD3A^kd cells, we identified two notable features, small vesicles around the dilated ER that appeared to be budding off from the ER (Fig. 6B3C) and mitochondria encased in vacuoles (Fig. 6B3D). Although the corresponding features of engulfed mitochondria in double-labeled confocal fluorescence micrographs were not detected, it is possible that the enzyme activity in vacuole-encased mitochondria was obscured. As noted above, ATAD3A, Mfn-2 and DRP1 were all involved in mitochondrial shaping. Results of a PSORT II prediction program (http://psort.ims.u-tokyo.ac.jp/) further showed that all three molecules contained notable stretch of coiled-coil (supplementary material Fig. S4A-C), suggesting that these proteins might interact with each other, and act together in communication between the ER and mitochondria. Using immunoprecipitation and immunoblotting, Mfn-2 and ATAD3A were co-precipitated by the respective antibodies (Fig. 6C1). Moreover, using the same method to react with mitochondria-associated membrane and light membrane fractions of sucrose gradient ultracentrifugation before immunoblotting, mitochondrial proteins, such as optic atrophy protein 1 (OPA1), AIF and Mfn-2, were co-precipitated by antibodies specific to ATAD3A (Fig. 6C2), suggesting that DRP1, Mfn-2 and ATAD3A were involved in a, yet to be determined, protein transport between the ER and mitochondria. In this case, silencing expression of Mfn-2, an important molecule for membrane fusion, would increase darkening of the ER, mitochondrial fragmentation, the number of small vesicles (Fig. 6D1,2), and cisplatin sensitivity (data not shown). Increased darkening of the ER would suggest protein accumulation in the organelle. Taken together, the data suggested that these proteins played a pivotal role in maintaining normal morphology of the ER and mitochondria. A defect in these proteins would increase drug sensitivity and cell apoptosis.

Our results indicate that the ATAD3A is the major type of ATAD3 family detected in LADC. Overexpression of ATAD3A in patients with LADC correlated with significantly higher incidence of early tumor recurrence and increased drug resistance, which ultimately reflected poor survival.

By demonstrating that the 70-kDa protein, which was frequently detected in LADC cell lines, was sensitive to CIP, our data suggested that the 70-kDa protein was a phosphorylated ATAD3A. Interestingly, the 66-kDa protein was sensitive to CIP as well, indicating that 66-kDa ATAD3A was also phosphorylated, and that the phosphorylation sites were at serine/threonine residues. Treatment with calphostin C, a pan-PKC inhibitor (Kobayashi et al., 1989), for 2 hours reduces phosphorylation and protein level of both 66- and 70-kDa ATAD3A, confirming that PKC is the kinase that catalyzes ATAD3A phosphorylation, and that phosphorylation is essential for ATAD3A stability. These findings corresponded well with immunoblotting results of patients' biopsy specimens, and suggested that ATAD3A detected in non-tumor lung tissue was a 63-kDa protein, which might not be readily phosphorylated and was labile. In addition, ectopic expression of PKC showed that the PKC isozymes responsible for ATAD3A phosphorylation were PKCγ, PKCι and PKCζ (Hirai and Chida, 2003). It is therefore worth noting that ATAD3A expression was upregulated in S phase of cell cycle. However, serum starvation also increased ATAD3A. It is unclear how serum starvation induces ATAD3A expression, but the findings correspond well with pathological observations that expression of ATAD3A increases with disease progression of LADC. Our previous results showed that hypoxia increased expression of hepatocyte growth factor (HGF) and interleukin-8 as well as synthesis of prostaglandin F_2α (Chiang, 2009). Since rapid growth of cancer cells during disease progression often results in inadequate supply of oxygen and nutrients in the tumor nest, our findings provide further explanation of how short-term serum deprivation-induced genes work in concert on survival and possibly metastasis of cancer cells (Tavaluc et al., 2007).

As noted previously, using AMIDA, Gires et al. detected overexpression of ATAD3B in patients with head and neck cancer (HNC) (Gires et al., 2004). In addition, Schaffrik et al. showed that two forms of ATAD3B, ATAD3B_l (large) and ATAD3B_s (small), were detected in HNC (Schaffrik et al., 2006). Unlike ATAD3B_l and ATAD3B_s, which differ at their N-termini, the discrepancy between ATAD3A and 3B is at their C-termini (supplementary material Fig. S1A-D). Using MALDI-TOF to determine specific antibody-precipitated proteins, we found that peptide mass fingerprints of both the 66-kDa and 70-kDa proteins match those of ATAD3A, including the N-terminus (supplementary material Fig. S2A-D). However, three MALDI-TOF fragments did not match ATAD3B or ATAD3A_L (NM_018188), indicating that the ATAD3A overexpressed in LADC cells is ATAD3A (BC033109). Applying phage display, Geuijen et al. identified ATAD3A as a significant tumor-associated antigen in AML blasts (Geuijen et al., 2005). Our observations support their data, indicating that AML and LADC might overexpress ATAD3A to facilitate cell growth and possibly metastasis. Inhibition of ATAD3A expression, by contrast, increases apoptosis and drug sensitivity, suggesting that ATAD3A is an anti-apoptotic factor.

It is interesting to note that silencing of ATAD3A expression increased mitochondrial fragmentation and mitochondria-containing autophagic vacuoles, phenomena that are frequently detected in Mfn-2 knockout-associated and DRP1-related apoptotic cells (Honda et al., 2005; de Brito and Scorrano, 2008a; Gomez-Lazaro et al., 2008; Knott et al., 2008; Chiang et al., 2009), implying that ATAD3A could be involved in mitochondrial fusion. Recently, de Brito and Scorrano showed that Mfn-2, an essential mitochondrial fusion protein, was detected on the ER, in particular in the mitochondria-associated membrane and the mitochondrial outer membrane. They suggested that the coiled-coil structure on the cytoplasmic side of Mfn-2 on both organelles tethered the two organelles together to coordinate Ca²⁺ flow (de Brito and Scorrano, 2008b; Merkwirth and Langer, 2008). However, results of a PSORT II prediction program (http://psort.ims.u-tokyo.ac.jp/) showed that although a cleavage site for mitochondrial presequence was detected between amino acid residues 20 and 21 (KRH|MA), no evident mitochondrial targeting sequence or ER membrane retention signal was identified in Mfn-2. On the contrary, two peroxisomal targeting signals, KINGIFEQL starting at amino acid residue 38 and RLKFIDKQL at residue 400 (supplementary material Fig. S4A), were found in the protein. Since peroxisomes are derived from ER, their findings suggested that Mfn-2 could be targeted to both organelles by a, yet to be determined, factor or the protein may take an alternative route (Pfanner and Geissler, 2001) via the ER before transport to mitochondria.

Like Mfn-2, ATAD3A is a transmembrane protein containing an extensive coiled-coil in the cytoplasmic domain of the N-terminus (supplementary material Fig. S4B). Moreover, like Mfn-2 and ATAD3A, DRP1, which is essential for mitochondrial fission, also contains a stretch of coiled-coil (supplementary material Fig. S4C). Interestingly, knockdown of DRP1 increased bulging of the ER, which was detected by ectopic expression of a KDEL-conjugated GFP. Electron micrographs confirmed that the ballooning structure observed by fluorescence microscopy was the ER, suggesting that DRP1 was involved in configuration changes of the ER. When DRP1 activity was kept undisturbed, knockdown of ATAD3A increased the number of transport vesicles, which appeared to be budding off of a dilated region of ER. Knockdown of any gene concomitantly changed the morphology of the ER and mitochondria, suggesting a functional connection between the two organelles involving Mfn-2, DRP1 and ATAD3A.

It is worth noting that ATAD3A and DRP1 are concurrently upregulated in the early S phase of cell cycle progression in LADC cells (Chiang et al., 2009). Elegant studies by Shiao et al. (Shiao et al., 1995) and Jackowski (Jackowski, 1996) showed that the level of phospholipids that are synthesized in the ER and transported to mitochondria via mitochondria-associated membranes also increases in the early S-phase of cell cycle progression. Moreover, using sucrose gradient ultracentrifugation, ATAD3A, glucose response protein 78 (GRP78) (Sun et al., 2006) and apoptosis AIF (Chen et al., 2008) were detected in fractions of light membrane, mitochondria-associated membrane and mitochondria. These data, considered together with studies using confocal immunofluorescence microscopy (CIM) and electron microscopy (EM), suggest that ATAD3A, an ATPase, could be required for an alternative transport of proteins, such as GRP78, AIF and membrane-anchored Mfn-2, from the ER to the mitochondria. We are examining such a possibility in an ongoing study.

In conclusion, immunoblotting and immunohistochemistry revealed abundant expression of ATAD3A in lung adenocarcinoma cells. Pathological results suggest that ATAD3A expression is associated with lymphovascular invasion, which reflects the increased metastatic potential of LADC and poor prognosis of patients. In vitro, serum starvation increased expression of ATAD3A and the level of cisplatin resistance in lung adenocarcinoma cells. Our finding that ATAD3A was present in the mitochondria-associated membrane of the ER and mitochondria, and that silencing of ATAD3A increased transport of vesicle-like figures and mitochondrial fragmentation as well as cisplatin sensitivity suggest that in addition to material transport between the ER and the mitochondria ATAD3A might play a role in drug resistance of lung cancer cells, in particular during deprivation of nutrients. At that stage, the cancer cells might stop proliferating and prepare for moving out of the gradually deteriorating microenvironment.

The patients in this study were from the same cohort used in the previous study. The protocols of both studies were approved by the Medical Ethics Committee. All clinical data were identical to those in the previous study (Chen et al., 2006).

Eight NSCLC cell lines (H23, H226, H838, H1437, H2009, H2087, A549 and SK-MES-1) were used for the in vitro evaluation of gene expression. H23, H838, H1437, H2009, H2087 and A549 are LADC cells, and H226 and SK-MES-1 are epithelial type cells. HeLa is a uterine cervical epithelial cell line. Cells were grown at 37°C in a monolayer in RPMI 1640 supplemented with 10% fetal calf serum (FCS), 100 IU/ml penicillin and 100 μg/ml streptomycin.

Following total RNA extraction and synthesis of the first-strand cDNA, an aliquot of cDNA was subjected to 35 cycles of PCR to determine the integrity of β-actin mRNA (Chen et al., 2006). The cDNA used in the following RT-PCR was adjusted according to the quality and quantity of β-actin mRNA. The primer sequences were selected using Primer3 (http://frodo.wi.mit.edu/cgi-bin/primer3). For ATAD3A, the primers were: ATAD3As: 5′-GGTCTACTCAGCCAAGAAT-3′ (sense primer, nts 889-907, BC033109) and ATAD3Aa: 5′-CACTTCCTCCCGTAGTCAAA-3′ (antisense primer, nts 1633-1653). The primers for ATAD3B were: ATAD3Bs: 5′-GGTCTACTCAGCCAAGAAT-3′ (sense primer, nts 881-899, NM_031921) and ATAD3Ba: 5′-GCGCATCTTCTGTCGGTACT-3′ (nts 1792-1811, NM_031921) and those for ATAD3C were: ATAD3Cs: 5′-GTGACAGACCGGGACAAAGT-3′ (sense primer, nts 1188-1207, NM_001039211) and ATAD3Ca: 5′-CACTTCCTCCCGTAGTCAAA-3′ (antisense primer, nts 2014-1995). ATAD3A and 3B shared the same sense primer, but had the different antisense primers. ATAD3A and 3C had the different sense primers, but shared the same antisense primer. The anticipated cDNA fragments were 765 base-pairs (bp) for ATAD3A, 931 bp for ATAD3B and 827 bp for ATAD3C. For N-terminal ATAD3A, the primers were: ATAD3A Ns: 5′-TGCGAGCATGTCGTGGCTCTTCGG-3′ (sense primer, nts 84-107, BC033109) and ATAD3A Na: 5′-GGGACGTCTCCCTCACTAGG-3′ (antisense primer, nts 958-939).

Total cell lysate was prepared by mixing 5×10⁷ cells/100 μl phosphate-buffered saline with equal volume of 2×NP-40 lysis buffer [40 mM Tris-HCl, pH 7.6, 2 mM EDTA, 300 mM NaCl, 2% NP-40 and 2 mM phenylmethylsulfonylfluoride (PMSF)]. Protein-G-Sepharose™ (Amersham Biosciences AB, Uppsala, Sweden) was pre-washed before mixing with 500 μg of total cell lysate. The reaction mixture was incubated at 4°C for 60 minutes, and then centrifuged at 800 g for 1 minute. The supernatant was reacted with 5 μg of purified monoclonal antibodies and 20 μl of fresh protein-G-Sepharose at 4°C for 18 hours. The reaction mixture was centrifuged at 800 g for 1 minute. Following removal of the supernatant, the precipitate was washed with 1× PBS, and dissolved in loading buffer (50 mM Tris, pH 6.8, 150 mM NaCl, 1 mM disodium EDTA, 1 mM PMSF, 10% glycerol, 5% β-mercaptoethanol, 0.01% Bromophenol Blue and 1% SDS). Electrophoresis was carried out in two 10% polyacrylamide gels with 4.5% stacking. One gel was processed for immunoblotting (Chen et al., 2006), and the other gel was stained with Coomassie Blue. Protein bands on the Coomassie-stained gel, which corresponded to the immunoblotting-positive bands, were extracted from the gel for identification by MALDI-TOF on a Voyager-DE™ pro biospectrometry workstation (Applied Biosystems, Milpitas, CA, USA). Fragments of peptide fingerprints were matched with those on the SwissProt database by MS-fit (ProteinProspector 4.0.5., The Regents of the University of California). After electrophoresis, proteins on the first gel were transferred to a nitrocellulose membrane for immunoblotting. The membrane was probed with specific antibodies. The signal was amplified by biotin-labeled goat anti-mouse IgG, and peroxidase-conjugated streptavidin. The protein was visualized by exposing the membrane to an X-Omat film (Eastman Kodak, Rochester, NY) with enhanced chemiluminescent reagent (NEN, Boston, MA).

Immunoblotting and immunohistochemistry were performed as described previously (Chen et al., 2006). Antibodies for β-actin were obtained from Chemicon International (Temecula, CA). Antibodies to ATAD3A were raised in the laboratory (supplementary material Fig. S1). For immunocytochemistry, the cells were grown overnight on slides, and then fixed with cold methanol-acetone at 4°C for 10 minutes before staining. Immunological staining was performed by an immunoperoxidase method. Antibodies to ATAD3A were not added for the negative control group.

DNA sequence corresponding to C-terminal amino acids 281-534 of ATAD3A was amplified by primer sequences containing BamHI (sense) and HindIII (antisense) restriction sites. The primer sequences were ATAD3A-forward: 5′-ATCGGATCCATGGTCTACTCAGCCAAGAAT-3′ (BamHI site is underlined) and ATAD3A-reverse: 5′-ATCAAGCAACTATCACTTCCTCCCGTAGTCAAA-3′ (HindIII site is underlined).

The restriction fragment of ATAD3A was cloned into an expression vector pET-32b⁺ (pET32⁺-AIF; Promega KK, Tokyo, Japan). Bacterial colonies containing pET32⁺-ATAD3A was selected, and induced by isopropyl-β-D-thiogalactopyranoside (IPTG) to mass-produce recombinant protein fragments of ATAD3A. The recombinant protein was purified using a nickel-affinity column. Affinity-purified ATAD3A was used to immunize BALB/c mice, and sensitivity of antiserum (OD₄₀₅>0.3 at 1:6000 dilutions) was measured by enzyme-linked immunosorbent assay (ELISA). Specificity of antibodies was determined by the appearance of a 66-kDa band in immunoblots of lung cancer cell extract (Geuijen et al., 2005). Monoclonal antibodies were produced by a hybridoma technique using mouse myeloma cells NS1, and ATAD3A-specific antibodies were screened by the above-mentioned methods. In some cells, two protein bands, a 66-kDa and a 70-kDa one, were detected by the antibodies in the immunoblotting. Sensitivity of the antibodies, which was measured by a serially diluted mouse ascites, reached 1:51,200 dilutions (supplementary material Fig. S2A1). In mouse tissues, the antibodies recognized a 66-kDa protein in tissue extracts from heart, lung, muscle and spleen. The 70-kDa protein was only detected in kidney and liver (supplementary material Fig. S2A2). In order to determine the identity of the 66-kDa and 70-kDa human proteins, the respective bands were excised from a Coomassie-stained gel and subjected to an analysis by matrix-assisted laser desorption/ionization and time-of-flight mass spectrometry (MALDI-TOF). The results showed that both immunoprecipitated 66-kDa and 70-kDa proteins matched ATAD3A (CAI22955; MS-Fit data shown in supplementary material Fig. S2B). The peptides matched 36.0% (211/586AAs) of ATAD3A. However, they matched only 25.0% (167/648AAs) of ATAD3B, and three MALDI-TOF fragments did not match ATAD3B (the mismatched sequences are noted in supplementary material Fig. S2C and D). The results excluded the possibility that LADC cells expressed ATAD3B.

It is worth noting that two different ATAD3As, ATAD3A (BC033109) and ATAD3A (NM_018188), are listed in GenBank (http://www.ncbi.nlm.nih.gov/entrez). The difference between ATAD3A (BC033109) and ATAD3A (NM_018188) is an insert of a peptide fragment containing 48 extra amino acid residues between Lys94 and Glu143 in ATAD3A (NM_018188), which has not been identified in ATAD3A (BC033109), 3B (NM_031921) or 3C (NM_001039211) (as shown in supplementary material Fig. S1D). Our results of MALDI-TOF analysis of immunoprecipitated proteins show that the ATAD3A, present in LADC, is ATAD3A (BC033109). To avoid confusion, we renamed ATAD3A (NM_018188) as ATAD3A_L in this manuscript.

In each pathological section, non-tumor lung tissue served as an internal negative control. Slides were evaluated by two independent pathologists with no knowledge of the clinicopathology of the specimens. The ImmunoReactive Scoring system was adapted for this study (Remmele and Schicketanz, 1993). Briefly, a specimen was considered to have strong signals when more than 50% of cancer cells were positively stained; intermediate, if 25-50% of the cells stained positive; weak, if less than 25% or more than 10% of the cells were positively stained; and negative, if less than 10% of the cancer cells were stained. Cases with strong and intermediate ATAD3A signals were classified as ATAD3A⁺, and those with weak or negative ATAD3A signals were classified as ATAD3A⁻ (Chen et al., 2006; Chiang et al., 2009).

Correlation of ATAD3A level with clinicopathological factors was analyzed by either the χ²-test or the Fisher's exact test. Survival curves were plotted using the Kaplan-Meier estimator (Kaplan and Meier, 1958). Statistical difference in survival between different groups was compared by the log rank test (Mantel, 1966). Statistical analysis was performed using GraphPad Prism5 statistics software (San Diego, CA). Statistical significance was set at P<0.05.

Electron microscopy was carried out using a routine protocol. Briefly, cells were fixed with 2.5% glutaraldehyde (EM grade, Sigma) in 100 mM phosphate buffer (PB; pH 7.2), incubated at 4°C overnight. The cells were washed with PB three times before post-fixation with 1% osmium tetroxide in PB for 2 hours. After removal of the fixative with distilled water, the cells were suspended in 2% molten agar. The agar blocks were trimmed and dehydrated in a serial dilution of ethanol for 15 minutes each. The blocks were further dehydrated with 100% ethanol three times, for 15 minutes each, and infiltrated with 100% ethanol-LR white (1:1) mixture overnight. The blocks were changed to the pure LR white (Agar Scientific Ltd., Essex, England) and infiltration was continued at 4°C for 24 hours, before transferring to capsules filled with LR white, which were polymerized and solidified at 60°C for 48 hours. The resin blocks were trimmed and cut with an ultramicrotome (Leica Ultracut R, Leica Mikrosysteme GmbH, Vienna, Austria). The thin sections were transferred to 200 mesh copper grids, and stained with 2% uranyl acetate for 30 minutes, and 2.66% lead citrate (pH 12) for 10 minutes, before observation with an electron microscope (JEM1400, JEOL USA, Inc., Peabody, MA) at 100-120 kV. For gene silencing experiments, cells were harvested 48 hours following siRNA treatment.

We thank Dr Jae-Won Soh (Department of Chemistry, Inha University, Incheon, Korea) for generously providing pHACE-PKCγ-CAT, pHACE-PKCδ-CAT, pHACE-PKCε-CAT, pHACE-PKCη-CAT, pHACE-PKCι-CAT and pHACE-PKCζ-CAT, and Dr Jiuping Ding (Key Laboratory of Molecular Biophysics, Huazhong University of Science and Technology, Wuhan, Hubei, China) for KDEL-conjugated GFP. RNAi for silencing DRP1 or ATAD3A gene expression was obtained from the National RNAi Core Facility in the Institute of Molecular Biology/Genomic Research Center, Academia Sinica, Taipei, Taiwan, supported by the National Research Program for Genomic Medicine Grants of NSC (NSC 97-3112-B-001-016). This study was supported, in part, by the Comprehensive Academic Promotion Projects (NCHU 975014 to K.C.C.).