Localization of a Portion of Extranuclear ATM to Peroxisomes*

The gene mutated in the human genetic disorder ataxia-telangiectasia codes for a protein, ATM, the known functions of which include response to DNA damage, cell cycle control, and meiotic recombination. Consistent with these functions, ATM is predominantly present in the nucleus of proliferating cells; however, a significant proportion of the protein has also been detected outside the nucleus in cytoplasmic vesicles. To understand the possible role of extra-nuclear ATM, we initially investigated the nature of these vesicles. In this report we demonstrate that a portion of ATM co-localizes with catalase, that ATM is present in purified mouse peroxisomes, and that there are reduced levels of ATM in the post-mitochondrial membrane fraction of cells from a patient with a peroxisome biogenesis disorder. Furthermore the use of the yeast two-hybrid system demonstrated that ATM interacts directly with a protein involved in the import of proteins into the peroxisome matrix. Because peroxisomes are major sites of oxidative metabolism, we investigated catalase activity and lipid hydroperoxide levels in normal and A-T fibroblasts. Significantly decreased catalase activity and increased lipid peroxidation was observed in several A-T cell lines. The localization of ATM to peroxisomes may contribute to the pleiotropic nature of A-T.

in maintaining the integrity of DNA and/or cell cycle control in response to DNA damage. These include mammalian DNA-dependent protein kinase (DNA-PK), Tel1p, and MEC1p from Saccharomyces cerevisiae, Rad3p, from Saccharomyces pombe, and MEI-41 from Drosophila melanogaster (4,5). Like DNA-PK, ATM does not appear to be a lipid kinase but rather functions as a serine/threonine protein kinase (6,7). ATM has been implicated in several signaling pathways including DNA damage recognition (2), radiation signal transduction and cell cycle control (8 -12), and meiotic recombination (13)(14)(15). There is also evidence that ATM plays a more general intracellular signaling role (16 -19). Further support for the importance of ATM has been provided by the complementation of radiosensitivity and cell cycle defects in A-T cells by full-length ATM cDNA (20,21).
Several groups, including our own, have shown that ATM is predominantly a nuclear protein, although a significant amount is present outside the nucleus (13,(22)(23)(24)(25). Many of the features of A-T, such as aberrant meiosis and cell cycle checkpoint control, are consistent with the nuclear localization of ATM. However, the neurological symptoms are difficult to explain in terms of defective DNA damage response. Using immunoelectron microscopy we showed labeling of ATM antibodies in cytoplasmic vesicles of varying sizes (60 -230 nm) in both fibroblasts and lymphoblastoid cells (24). The heterogeneous size of the vesicles labeled with ATM antibodies and the punctate immunofluorescence was suggestive of peroxisomes (24). Peroxisomes are single membrane-bound organelles present in virtually all eukaryotic cells. They contain a range of enzymes involved in a variety of metabolic processes including peroxidebased respiration, oxidation of very long chain fatty acids, and synthesis of plasmalogens and bile acids. The majority of peroxisomal matrix proteins are imported into the peroxisome via a noncleavable peroxisomal targeting signal type 1 (PTS1), which consists of a tripeptide SKL or conserved variant at the extreme C terminus (26). In this report we show that ATM co-localizes with the peroxisomal matrix protein catalase and binds the PTS1 receptor. The presence of ATM in peroxisomes may provide an explanation for some of the pleiotropic symptoms of A-T. enbush strain) were obtained from the Central Animal Breeding House (Pinjarra Hills, Brisbane, Australia). ATM knockout mice were bred from heterozygotes kindly provided by Dr. Phil Leder (Harvard Medical School, Boston, MA). The lipid peroxidation kit was from Cayman Chemicals (Ann Arbor, MI).
Cell Culture-Primary human fibroblasts and neonatal foreskin fibroblasts were obtained from Prof. Kay Ellem of this institute and cultured in RPMI medium containing 10% fetal calf serum. Zellweger fibroblasts from complementation group 4 (GM13267) were obtained from the Human Genetic Mutant Cell Repository, Coriell Laboratories. GM 3487A, GM 3984A, and GM 3395 are nontransformed A-T fibroblasts, and AT5B/VA and AT13LA SV are SV40 transformed A-T fibroblasts that were obtained from Dr. R. Reddel (Westmead Hospital, Sydney, Australia).
Preparation of Antibodies-The ATM4BA antibody has been described previously (24). A mouse monoclonal antibody (CT1) to the extreme C-terminal sequence of ATM (DPKNLSRLFPGWKAWV) was prepared by standard methods. A peptide with this sequence was synthesized by Chiron Mimotopes with an N-terminal cysteine residue to facilitate coupling to keyhole limpet hemocyanin using m-maleimidobenzoyl-N-hydroxysuccinimide ester as described (24). The conjugate was used to immunize BALB/c mice. Spleen cells from these mice were fused with the Sp2/O myeloma cell line. Positive hybridomas were detected by enzyme-linked immunosorbent assay with unconjugated peptide bound to plates (24). The CT-1 monoclonal antibody is of the IgG 1 isotype and recognizes both human and murine ATM.
Fluorescence Immunohistochemistry-For immunofluorescence studies, cells grown on coverslips, were fixed with 4% formaldehyde in PBS for 20 min and then permeabilized with PBS containing 0.3% Triton X-100 for 5 min. Coverslips were incubated with primary antibody in PBS containing 2% fetal calf serum at 4°C overnight, washed three times in PBS containing 0.05% Tween 20, and then incubated with secondary antibody for 3 h at room temperature. The antibodies were used at the following dilutions: ATM-4BA (1:200); CT-1 culture supernatant (undiluted); CT-1 ascites fluid (1:200); and anti-catalase (1:200). Coverslips were mounted using Vectashield fluorescence mounting medium (Vector Labs) and viewed with a Bio-Rad MRC500 confocal microscope using K1/K2 filters.
Yeast Two-hybrid Analysis-A partial catalase clone (309301) was obtained from the IMAGE Consortium cDNA clone library (Genome Systems Inc.). A 1.4-kb nearly full-length EcoRI-NotI catalase fragment was subcloned into the EcoRI/NotI sites of pSL1180. The remaining 5Ј end was subcloned by isolating a 280-bp EcoRI fragment from pCAT 2.2 (a gift from Dr. Nick Hayward of this institute) and inserting it into the EcoRI site of pSL-CAT1.4. Clones were screened for correct orientation of the 280-bp fragment. pSL1180 containing full-length catalase was designated pSL-catalase. Catalase was then subcloned into pGBT9 by polymerase chain reaction. The sense oligonucleotide 5Ј-GGAAGATC-TCCATGGCTGACAGCCGGG-3Ј, which encodes a BglII restriction site, and the antisense oligonucleotide 5Ј-ACTGCATGCATCAGATTT-GCCTTCTCCCTTGC-3Ј, which encodes an NsiI restriction site, were used in a polymerase chain reaction using Pfu DNA polymerase with pSL-catalase as the template. The polymerase chain reaction product was purified and digested with BglII and NsiI and cloned into pGBT9 digested with BamHI and PstI. The sequence integrity of catalase was confirmed by DNA sequencing.
The pGAD10 plasmid containing a cDNA insert for Pex5p was kindly provided by Dr. S. Gould (27). The plasmid ATM 4.01 was obtained by cloning the NsiI/AseI fragment of full-length ATM cDNA into the yeast DNA-binding domain vector pAS2. This fragment contains the C-terminal of ATM corresponding to amino acid residues 1666 -3056 plus some untranslated 3Ј sequence (12). To remove 491 bp from the 3Ј-end of ATM 4.01, the plasmid was digested with NcoI (restriction site at 8964 bp) and NotI (site in vector). The linearized plasmid (3.51 kb) was religated, and purified after transformation. An 8.7-kb ATM fragment (corresponding to amino acid residues 1-2934) was cloned into the NcoI site of pAS2-1 (12).
Subcellular Fractionation and Western Blotting-Two 150-cm 2 flasks each of neonatal foreskin and GM13267 fibroblasts were trypsinized to remove the cells from the culture dish, and the cells were suspended in 2 ml/flask of ice-cold homogenization buffer. The procedures for preparation of nuclear and post-mitochondrial pellet fractions and Western blotting for ATM were as described previously (24).
Purification of Mouse Liver Peroxisomes-Mouse livers were excised, and each was immediately suspended in 7.5 ml of cold sucrose/ethanol solution (0.25 M sucrose and 0.1% ethanol, containing 18 l 0.1 M phenylmethylsulfonyl fluoride, 20 l 10 mg/ml leupeptin, 20 l of 10 mg/ml aprotinin). The livers were minced finely using a McIlwain tissue chopper followed by homogenization on ice with one stroke of a Potter/ Elvehjem homogenizer. The homogenate was centrifuged at 1,000 ϫ g for 10 min at 4°C, and the supernatant was transferred to a fresh tube and centrifuged at 2,800 ϫ g for 15 min. The supernatant from this step was centrifuged at 16,000 ϫ g for 10 min to obtain an organelle pellet. This was suspended in 5 ml of sucrose/ethanol and again centrifuged at 16,000 ϫ g for 10 min at 4°C. The resulting washed organelle pellet was resuspended in 0.5 ml of sucrose/ethanol solution and loaded onto a Nycodenz (Nycomed AS) gradient (15-40%, in STE (0.25 M sucrose containing 10 mM Tris-HCl, pH 7.4, 0.1 mM EDTA)) and centrifuged for 40 min at 80,000 ϫ g using a VTi65.1 rotor (Beckman). Fractions (0.5 ml) were collected and assayed for catalase, succinate dehydrogenase activity, density (refractometry), and protein concentration by standard methods.
For Western blotting, proteins were separated by SDS-polyacrylamide gel electrophoresis on 4% gels in a Protean II apparatus (Bio-Rad), transferred to Hybond C nitrocellulose membrane (Amersham Pharmacia Biotech) at 100V for 1 h, and then blocked for 2 h at room temperature in 5% blotto (5% nonfat milk powder in PBS). The membranes were then incubated with CT-1 antibody (culture supernatant diluted 1:10) overnight at 4°C. This was followed by five washes in PBS containing 0.05% Tween 20 for 5 min and then a 1-h incubation with horseradish peroxidase-conjugated anti-rabbit IgG at room temperature. The membranes were again subjected to repeated washing as before and developed using the Renaissance chemiluminescence system (Dupont).
Lipid Peroxidation Measurements-Determination of lipid peroxides was performed on cell homogenates of normal and A-T fibroblasts using the lipid hydroperoxide kit from Cayman Chemicals according to the manufacturer's instructions.
Catalase Assays-Fibroblasts were grown in 75-cm 2 flasks to 70% confluency. The cells were washed twice in PBS and scraped off the plastic, and the pellet resuspended in 400 l of ice-cold 10 mM Tris-HCl, pH 7.0, containing 0.1 mg/ml digitonin and 0.25 M sucrose. After incu-  2) and atm Ϫ/Ϫ mice (lanes 3 and 4) using CT-1 monoclonal antibody specific for ATM. DNA-PK was used as a loading control. bation on ice for 10 min, Triton X-100 was added to a final concentration of 0.1%. Following a further 5-min incubation on ice, the mixture was vortexed and used for assay of catalase activity as measured by the decrease in absorbance of hydrogen peroxide at 240 nm (28).

RESULTS
The C Terminus of ATM Is Similar to That of Catalase-The heterogeneous size of the vesicles labeled with ATM antibodies, the presence of a defined single membranous structure, and the punctate immunofluorescence, which we observed previously was suggestive of peroxisomes (24). Peroxisomes are singlemembrane-bound organelles present in virtually all eukaryotic cells, and they contain a range of enzymes involved in a variety of metabolic processes including peroxide-based respiration, oxidation of very long chain fatty acids, and synthesis of plasmalogens and bile acids. Furthermore, comparison of the Cterminal sequences of the peroxisomal matrix enzyme catalase, which contains sequence important for its localization to this organelle, with that of ATM revealed some sequence conservation (Fig. 1A). This relatedness is evident in two regions as indicated by bold type in Fig. 1A. It is noteworthy that the C-terminal sequence (last 16 amino acids) shows 100% conservation between mouse and human ATM. A monoclonal antibody (CT-1) was raised against the extreme C-terminal 16 amino acids. The specificity of the CT-1 antibody was determined by immunoblotting with tissue extracts from normal and atm Ϫ/Ϫ mice (Fig. 1B).
Colocalization of ATM and Catalase-To confirm the localization of ATM to peroxisomes, we performed double-labeling experiments using an antibody to human catalase and two different antibodies to ATM, the polyclonal antibody ATM-4BA as described previously (24) and the monoclonal antibody (CT-1). When catalase was stained using a rhodamine-labeled secondary antibody (red) and ATM with a fluorescein isothiocyanate-labeled secondary antibody (green), it is evident that colocalization (yellow staining) occurs ( Fig. 2A), demonstrating that ATM is localized to peroxisomes. However, it is also clear that ATM is present in other cytoplasmic vesicles not stained with anti-catalase. This is consistent with the report of Lim et al. (29), which shows that ATM binds to ␤-adaptin and ␤-neuronal adaptin-like protein, components of clathrin-coated and secretory vesicles. Confocal microscopy was used to confirm the co-localization of catalase and ATM using the CT-1 antibody (Fig. 2B).
ATM Is Present in Purified Peroxisomes-Because peroxisomes are readily purified from mouse liver using Nycodenz gradients, we used this approach to further demonstrate that ATM is present in peroxisomes. The results of such a gradient analysis appear in Fig. 3A, demonstrating that catalase activity peaks in fraction 3. Small amounts of mitochondria are also present elsewhere on the gradient as evidenced by the presence of succinate dehydrogenase peaking in fraction 6. Fractions were run on a SDS gel and checked for the presence of ATM by immunoblotting with CT-1 antibody. It can be seen that ATM is detected in the peak fraction (#3) from the gradient, coinciding with the peak of catalase (Fig. 3B). It is evident that the recovery of ATM in these fractionated peroxisomes is low, but this protein was reproducibly observed to coincide with the catalase peak. This is also not due to nuclear contamination because the much more highly abundant protein DNA-depend- ent protein kinase (DNA-PK) was not detected in any of the gradient fractions (Fig. 3B, lower panel).
ATM Binds to the Peroxisomal PTS1 Receptor-The C-terminal sequence similarity of ATM to catalase suggested that ATM might be imported into peroxisomes in a manner analogous to catalase. The import of proteins into peroxisomes requires a host of peroxisome assembly factors or peroxins that include specific PTS receptors, which recognize the targeting signals (30). In the case of catalase the PTS1 signal was thought to be the SHL sequence (positions 11-9 from the C terminus), which is in the corresponding position to a related sequence in ATM, SRL (Fig. 1A). More recently it has been shown that the extreme C-terminal sequence KANL is sufficient to target catalase to the peroxisome (31,32). The Cterminal sequence of ATM is KAWV; however, it is not yet known whether this constitutes a PTS1 signal. Recognition of the PTS1 signal is mediated by the PTS1 receptor, now known as Pex5p (30). In our study the yeast two-hybrid assay was used to show that Pex5p binds to ATM. A 4.01-kb construct corresponding to the C-terminal half of ATM was cloned into the yeast vector pAS2, which contains the GAL4 DNA-binding domain (12,24). Pex5p cloned into the pGAD10 vector was used to co-transform yeast cells, and interaction between the two proteins was assayed using a liquid culture ␤-galactosidase assay. As a control, the coding sequence of human catalase was also cloned into the pGBT9 vector. The results show that both proteins bind the PTS1 receptor (Table I). Although the observed binding was weak, it nevertheless represents a true interaction because a similar degree of binding was observed with catalase and may reflect the low affinity expected of a transient interaction in vivo. When the 8.7-kb and 3.51-kb ATM cDNA constructs in pAS2, both lacking the extreme C terminus, were used with Pex5p there was no interaction. The 8.7-kb construct expresses almost full-length ATM protein (corresponding to amino acids 1-2934), and we have previously used it to demonstrate an interaction between ATM and p53 (12). The 3.51-kb construct lacks 491 bp at the C terminus. In toto, these data suggest that ATM may be imported into peroxisomes via the PTS1 pathway.
Reduced Extra-nuclear ATM in Peroxisome Biogenesis Disorder Cells-It would be predicted that, where there is a PTS1 import defect, ATM should be absent or reduced in peroxisomes but present in the nucleus. Peroxisome biogenesis disorders result from failure to assemble normal peroxisomes. Peroxisome biogenesis disorder patients can be divided into at least 10 complementation groups, reflecting the complexity of the import process (33). We therefore performed subcellular fractionation, followed by immunoblotting for ATM, on Zellweger complementation group 4 cells (GM13267) compared with normal fibroblasts. Complementation group 4 cells contain a defect in an ATPase (Pex6p), which is required for stability of the receptor, Pex5p (30). It is clear that the relative amount of ATM in the post-mitochondrial pellet fraction versus nuclear ATM in GM13267 cells is reduced compared with that in normal cells (Fig. 4). Immunoblotting with antibody to DNA-PK was used to show that there was no contamination of nuclear material in the post-mitochondrial pellet fraction (data not shown). ATM was also found in the cytoplasm of this peroxisome biogenesis disorder line (data not shown), similar to that observed previously for the ATM protein in the A-T cell line (AT1ABR) expressing mutant ATM protein (24).
Decreased Catalase Activity and Increased Lipid Hydroperoxide Levels in A-T-There have been reports in the literature (although controversial) of defects in catalase activity in some A-T cell lines (34). The significance of these observations required reinvestigation in view of our finding of ATM in the peroxisome. We found significantly reduced catalase activity (50 -70% of normal) in all the A-T fibroblasts tested (Table II). The reduction in activity does not appear to be due to reduced protein levels because no obvious change in the amount of catalase was detected by immunoblotting with anti-catalase antibodies (data not shown). In addition the distribution of catalase in A-T cells as assessed by fluorescence immunohistochemistry appears normal (data not shown). A consequence of reduced catalase activity would be increased hydrogen peroxide levels in the cell, which would in turn be expected to result in increased levels of lipid hydroperoxides. Lipid hydroperoxides were therefore measured in several normal and A-T fibroblasts, and the results are shown in Table III. It is evident that the levels of peroxides are significantly higher than controls in all A-T lines tested. DISCUSSION Our previous studies on the localization of ATM indicated that a proportion of the protein is localized outside the nucleus in vesicular structures (24). Using several different independent techniques and different anti-ATM antibodies, this study has shown that ATM is present in peroxisomes. We further demonstrated binding of the C-terminal half of ATM to Pex5p, the receptor for proteins containing a PTS1 signal. In addition we showed that ATM is markedly reduced in the post-mito-  Fig. 2A) ATM immunoreactivity (indicated by the arrow) is evident in the peak fraction (#3) corresponding to the peak of catalase activity. Lower panel, immunoblotting of peroxisome gradient fractions showing absence of the highly abundant nuclear protein, DNA-PK. Lane 0, nuclear extract from gradient isolation; lanes 1-5, fractions 1-5 from a typical peroxisome isolation, respectively. chondrial pellet of GM13267 cells, which have a defect in the import of proteins into peroxisomes. The nuclear and peroxisomal forms of ATM appear to be identical in as far as the C terminus is concerned, because both are recognized by the monoclonal anti-C-terminal peptide antibody, CT-1. The mech-anisms that determine the distribution of ATM between the nucleus and peroxisomes are presently unknown, and in addition there is no obvious nuclear localization signal in the protein. Determination of the exact peroxisomal targeting signal must also await detailed site-directed mutagenesis studies. In this regard, although catalase has been shown to bind to the PTS1 receptor (Ref. 32 and this study), a recent report indicates that the import of catalase is more complex (35). Although we have shown that ATM also binds the PTS1 receptor, we have not yet demonstrated that ATM is internalized in the peroxisomal matrix. It may be bound to the peroxisome membrane.
As pointed out by Rotman and Shiloh (36), several lines of evidence suggest that the hypersensitivity of A-T cells to ionizing radiation may not result solely from defective cell cycle checkpoints and DNA damage processing. The constitutive activation of several IR-responsive signaling pathways in unirradiated A-T cells, which resembles but does not achieve the same extent as that in irradiated control cells, indicates that these cells are in a state of continuous oxidative stress. In addition, red blood cells from A-T heterozygotes have been reported to contain a lower content of sulfhydryl groups, increased membrane fluidity and increased lipid peroxidation (37). We have now demonstrated decreased levels of catalase and increased levels of lipid hydroperoxides in several A-T cell lines. The major mechanisms for disposal of hydrogen peroxide are catalase and glutathione peroxidase, both of which occur in peroxisomes as well as in the cytoplasm. The continuous state of oxidative stress in A-T could be explained by a deficiency in the detoxification of reactive oxygen species, and this may result from the decreased catalase activity observed here. Intriguingly, catalase activity in the livers of atm Ϫ/Ϫ is not statistically different to normal (data not shown), suggesting fundamental differences between mice and humans or different tissues in response to lack of peroxisomal ATM. This may be significant in terms of the much less severe neurodegeneration seen in atm Ϫ/Ϫ mice (38 -41) and requires further investigation.
It should be pointed out that ATM is also present on nonperoxisomal vesicles co-localizing with ␤-adaptin (29), and ATM binds in vitro to the neuronal homologue of ␤-adaptin, ␤-neuronal adaptin-like protein. Because it has been suggested that ␤-NAP plays an essential role in synaptic vesicle transport in neuronal cells, the absence of ATM from these vesicles could also contribute to neuronal degeneration. Indeed there is evidence that defects in synaptic vesicle transport give rise to ataxia in humans (42). It is intriguing to suggest that the function of the noncatalase vesicles and peroxisomes may be interrelated, in that components of the peroxisome importation machinery (peroxins) have been localized to vesicles distinct from mature peroxisomes (43), and the endoplasmic reticulum plays an essential role in peroxisome biogenesis (44).
Many of the features of A-T, for example defects in cell cycle checkpoints and aberrant meiosis, are consistent with the nu-  No colonies a pVA3 and pTD1 are positive control plasmids encoding a murine p53/GAL4 DNA-binding domain hybrid in pGBT9 and an SV40 large T-antigen/GAL activation domain hybrid in pGAD3F, respectively.
b Only a few colonies were detected.

TABLE II Catalase activity in normal and A-T fibroblasts
Fibroblast extracts were prepared as described under "Experimental Procedures," and catalase activity was measured by the decrease in absorbance of the substrate, hydrogen peroxide, at 240 nm. The experiment was performed three times, and each determination was performed in triplicate. The results represent the means Ϯ S.D. of the combined results. NFF, neonatal foreskin fibroblasts. PGPF and JRRF, normal human dermal fibroblasts. clear localization of ATM. However, an explanation for the basis of the most debilitating and characteristic symptoms of the disease, the neurological dysfunction, has so far been elusive. Some peroxisome biogenesis disorders show similarity with A-T, for example, ataxia that occurs in infantile Refsum's disease (45) and demyelination, which occurs in X-linked adrenoleukodystropy (46). ATM is found predominantly outside the nucleus in Purkinje cells (47). 2 Thus in noncycling cells the primary function of ATM would be in the maintenance of cellular homeostasis. In summary, we have demonstrated that ATM is associated with peroxisomes, which provides a new direction for investigating the role of ATM.