A single PDZ domain protein interacts with the Menkes copper ATPase, ATP7A. A new protein implicated in copper homeostasis.

The homeostatic regulation of essential elements such as copper requires many proteins whose activities are often mediated and tightly coordinated through protein-protein interactions. This regulation ensures that cells receive enough copper without intracellular concentrations reaching toxic levels. To date, only a small number of proteins implicated in copper homeostasis have been identified, and little is known of the protein-protein interactions required for this process. To identify other proteins important for copper homeostasis, while also elucidating the protein-protein interactions that are integral to the process, we have utilized a known copper protein, the copper ATPase ATP7A, as a bait in a yeast two-hybrid screen of a human cDNA library to search for interacting partners. One of the ATP7A-interacting proteins identified is a novel protein with a single PDZ domain. This protein was recently identified to interact with the plasma membrane calcium ATPase b-splice variants. We propose a change in name for this protein from PISP (plasma membrane calcium ATPase-interacting single-PDZ protein) to AIPP1 (ATPase-interacting PDZ protein) and suggest that it represents the protein that interacts with the class I PDZ binding motif identified at the ATP7A C terminus. The interaction in mammalian cells was confirmed and an additional splice variant of AIPP1 was identified. This study represents an essential step forward in identifying the proteins and elucidating the network of protein-protein interactions involved in maintaining copper homeostasis and validates the use of the yeast two-hybrid approach for this purpose.

Copper homeostasis is essential for the survival and proper functioning of all organisms. The growing number of diseases in which the homeostatic regulation of copper balance is disrupted include Menkes and Wilson diseases as well as neurological diseases such as amyotrophic lateral schlerosis, Alzheimer, Parkinson, and prion diseases (1,2). Over the last decade, significant progress has been made toward unraveling the biological processes that maintain a fine balance of copper in the body. Several mammalian proteins have been identified and shown to be critical for copper handling. They include metallothioneins, small proteins likely to be involved in the detoxification of excess copper (3); CTR1, a plasma membrane copper uptake protein (4,5); ATP7A and ATP7B, ATPases required for copper transport to cuproenzymes and excretion of excess copper (6 -8); Atox1, a chaperone for copper delivery to the ATPases (9, 10); hCCS, required for copper delivery to Cu/Znsuperoxide dismutase (10); hCox17, Cox11, hSco1, and hSco2, required for copper incorporation into cytochrome c oxidase in the mitochondrion (10,11); and most recently, MURR1 (COMMD1), a protein implicated in vesicular copper sequestration (12). These proteins probably represent only a small fraction of those that are part of the network of protein interactions that mediate copper homeostasis.
Within the cell, the biological pathways controlling fundamental processes, such as cell growth, cell cycle progression, metabolic pathways, and signal transduction, are initiated and maintained through a series of tightly regulated and coordinated protein-protein interactions (13,14). Many of these interactions are transient and facilitated by distinct protein interaction domains that recognize specific exposed motifs on partner proteins. For example, Src2 and Src3 homology domains recognize specific motifs that are phosphorylated or proline-rich (14), respectively, whereas PDZ (PSD-95/Discs-large/ZO-1) domains recognize C-terminal motifs (15)(16)(17). Among the copper proteins, specific interactions have been demonstrated between each of the ATPases and Atox1 (18,19) and between ATP7B and MURR1 (20), but none of these interactions appear to involve any of the known interaction modules.
To identify new proteins implicated in copper homeostasis, one approach is to identify interacting partners of proteins known to be essential for copper transport. We have used ATP7A to initiate a search for such interacting proteins. ATP7A is a large transmembrane, coppertransporting P-type ATPase that has two roles within a cell: to transport copper to copper-requiring enzymes at the trans-Golgi network (TGN) 3 and in the secretory pathway, and at the plasma membrane to efflux excess copper from the cell (7,8). To fulfill its dual role in the cell, ATP7A maintains a steady state localization at the TGN while constitutively cycling between the TGN and plasma membrane. With elevated intracellular copper, its steady state location shifts to the plasma membrane until copper levels return to normal and the majority of the protein cycles back to the TGN (21)(22)(23). Note that in polarized cells, ATP7A traffics to the basolateral surface (24). ATP7A comprises an N-terminal region with six metal binding motifs (GMXCXXC) that each can bind one copper ion, eight transmembrane domains, and several conserved residues or motifs that are characteristic of heavy metaltransporting ATPases (19,25). ATP7A mediates copper translocation across cellular membranes through ATP-driven cycles of phosphoryla-tion and dephosphorylation, and its copper-induced trafficking is linked to its catalytic activity (7,8). A diverse array of protein-protein interactions are likely to be integral to the localization and activity of ATP7A.
Ion transport proteins are endowed with multiple sorting motifs. These motifs interact with components of the cell sorting machinery to specify and regulate protein trafficking to and maintenance at specific cell surfaces (e.g. apical versus basolateral), as well as trafficking between intracellular storage compartments and the plasma membrane. Often these motifs are located within cytoplasmic C-terminal tails (26). Whereas many transport proteins employ classical targeting motifs such as tyrosine-and dileucine-based signals, a variety of unique signals are utilized by many proteins for correct targeting (26). Potentially, cells can control the transport activity of ion transporters by regulating the trafficking of these proteins through modulation of protein-protein interactions.
For ATP7A, a C-terminal dileucine motif is required for TGN retrieval (27,28), whereas a 38-amino acid sequence containing transmembrane domain 3 mediates TGN retention (29). Within the C terminus, there is also a stretch of acidic residues whose role, if any, is unknown, as a well as a putative class I PDZ binding motif (1497DTAL1500). Recently, this motif was shown to be required for the targeting and/or retention of ATP7A at the basolateral surface in polarized Madin-Darby canine kidney cells (24). A putative targeting signal also exists within the N-terminal metal binding site 6 (30).
To identify cellular constituents that are likely to play a role in regulating and/or facilitating the copper transport and trafficking activities of ATP7A and, by inference, implicated in copper homeostasis, we have embarked on a study that utilizes the cytoplasmic regions of ATP7A as bait proteins for yeast two-hybrid screening. This approach has been utilized to identify interacting partners of a vast array of proteins including the Na,K-ATPase and gastric H,K-ATPase (31). With the multiplicity of signals often found within the C termini of proteins, screening of a human brain cDNA library commenced with the ATP7A C terminus as the bait. This report focuses on the identification and initial characterization of one of several proteins that were found to interact with the ATP7A C terminus. This protein, designated AIPP1 (ATPase-interacting PDZ protein) is a small protein with a single PDZ domain. Coimmunoprecipitation experiments in mammalian cells confirmed the interaction between ATP7A and AIPP1, while mutagenesis experiments showed that AIPP1 interacted within the last 15 amino acids of ATP7A, with no involvement of the dileucine motif. Alternatively spliced transcripts of AIPP1 were identified and potentially encode two isoforms of this protein, one with a putative signal peptide. AIPP1 represents a new protein in copper homeostasis, and its identification establishes the validity of the yeast two-hybrid approach in identifying other proteins with key roles in this important process.
Cloning and Mutagenesis-To generate a bait construct encoding the C terminus of ATP7A, a 312-bp fragment encoding 97 amino acids of the ATP7A C terminus, from the end of transmembrane domain 8 to the stop codon (amino acids 1404 -1500), was amplified by PCR using oligonucleotides Y2H7 (5Ј-ccccatatgTCTCTCTTCCTTAAACTT-TAC-3Ј) and Y2H8 (5Ј-cccgtcgacTTATAATGCAGTGTCATCATC-3Ј) (NdeI and SalI restriction sites in lowercase, boldface type) and pCMB19 containing the ATP7A cDNA (35) as a template. This fragment was ligated into the NdeI/SalI sites of the vector pAS2-1 (Clontech) in frame with the GAL4 DNA binding domain (DNA-BD) to generate plasmid pSLB29.
A fragment encoding the ATP7A C terminus in which the 1487LL1488 was converted to 1487AA1488 was generated by PCR using pCMB300, in which the mutation was previously incorporated (27), as template DNA. PCR using the primers hMNK25 (5Ј-cccgggct-gcagTTATAATGCAGTGTCATCATC-3Ј), which introduced a PstI restriction enzyme site (lowercase, boldface type) at nucleotide position 4504 (immediately after the stop codon), and Y2H7, which introduced an NdeI restriction enzyme site upstream of nucleotide position 4210 (amino acid position 1404), led to amplification of a 315-bp product. This PCR product was cloned into the pGEM-T-Easy vector (Promega) to generate plasmid pCMB545. The insert from this plasmid was isolated with NdeI/PstI and ligated into the corresponding sites of pAS2-1 to create an in-frame fusion with the GAL4 DNA-BD in plasmid pSLB63.
A bait construct was generated in which the last 15 amino acids were deleted from the ATP7A C terminus (pAS2-1/ATP7A⌬1485-1500). PCR amplification was used to generate ATP7A (⌬1485-1500) using the wild-type ATP7A cDNA in pCMB19 (35) as a template and oligonucleotides Y2H7 and hMNK23 (5Ј-ccctgcagTCTCCCACCAG-GAGTCAGTGCTTGTCAGG-3Ј) that introduced a stop codon at nucleotide position 4456 (amino acid position 1486) and a PstI site at nucleotide position 4473. The 263-bp PCR product was ligated with the pGEM-T-Easy vector and was designated pCMB540. The insert was isolated with NdeI/PstI and ligated into the NdeI/PstI site of pAS2-1 to generate pSLB65.
For yeast two-hybrid interaction assays with the ATP7B C terminus (ATP7B-C), a 294-bp cDNA fragment that encoded the last 97 amino acids of ATP7B (amino acids 1369 -1465) was amplified by PCR using pCMB278, containing the wild-type ATP7B cDNA as a template (36), and primers hWND17 (5Ј-ccccatatgTCATCCCTGCAGCTCAAGT-GC-3Ј) and hWND18 (5Ј-cccgtcgacTCAGATGTACTGCTCCTC-3Ј), which contained NdeI and SalI restriction sites (lowercase, boldface type), respectively. The PCR product was initially cloned into the Nde-I/SalI sites of pAS2-1 to generate pSLB13. However, since the ATP7B-C/GAL4 DNA-BD fusion in this vector caused autoactivation of the reporter genes, the cDNA encoding ATP7B-C was cloned into the pG-ADGH activation domain vector (Clontech). The insert from pSLB13 was isolated as an EcoRI fragment and ligated into the EcoRI site of pGADGH to create plasmid pSLB20 (pGADGH/ATP7B-C). To carry out interaction assays with ATP7B-C expressed from pGADGH, the AIPP1 cDNA was cloned into pAS2-1. The AIPP1 cDNA was isolated as a BamHI/BglII fragment from the original pACT2-based library plasmid (pSLB56) and ligated into the BamHI site of pAS2-1 in frame with the GAL4 DNA-BD to generate plasmid pSLB75. As an unrelated control for the interaction assays, the plasmid pCMB152 was used, which encoded the full-length human Atox1 in-frame with the GAL4 activation domain (AD) in pGADGH (19).
Expression constructs to enable the detection of AIPP1 in mammalian cells were generated and encoded AIPP1 with an N-terminal c-Myc epitope tag (EQKLISEEDL). The original plasmid pSLB56 lacked the first 33 bp of the AIPP1 open reading frame (ORF) and was used as the initial template to first generate a full-length ORF. Primers were designed to amplify the entire AIPP1 ORF and included the 33 bp that were missing from the pSLB56 insert, based on the data base sequence (accession number NM_016484) (LOC51248#2F, 5Ј-cggaattc 142 CCC-GAGATGGACAGCCGGATTCCTTATGATGACTACCCGGTGG-TTTTCTTGCCTGCCTATGAGAATCC 189 -3Ј; LOC51248#3R, 5Ј-cccgcctcgagaacttta 551 CTAGTGCACAGTCCTCTCTTTTTGGC-G 524 -3Ј) (EcoRI and XhoI sites for cloning in lowercase, boldface type). The PCR product was ligated into the pGEM-T-Easy vector to generate plasmid pSLB81. This plasmid was used as a template with oligonucleotides LOC51248#4F (5Ј-gcggaatt 142 CCCGAGATGGAACAAAAAT-TGATCAGCGAAGAGGATCTTGACAGCCGGATTCCTTATGA-TGACTACCCG 161 -3Ј) and LOC51248#3R to incorporate an N-terminal c-Myc tag (boldface type and underlined) immediately after the AIPP1 start codon. Primers also incorporated EcoRI/XhoI sites (lowercase, boldface type) for cloning. PCR products were ligated into the EcoRI/XhoI site of the mammalian expression vector pcDNA3 (Invitrogen). The final plasmid construct was designated pSLB84.
Yeast Two-hybrid Screening-A human brain cDNA library constructed in the GAL4 AD vector pACT2 (Clontech) was screened by co-transformation of the library DNA and the bait construct pSLB29 into S. cerevisiae YGH1 as previously described (37). Transformants were screened for HIS3 and lacZ gene expression using protocols described in Ref. 70. Positive clones were verified by curing the bait plasmid from the HISϩ/LacZϩ positive yeast clones, isolating the library plasmid(s) from Leuϩ/TrpϪ/(LacZϩ) auxotrophs, and retransforming library plasmids back into S. cerevisiae YGH1 either with the pAS2-1 vector alone or with the original bait plasmid pSLB29. Transformants were then reassessed for ␤-galactosidase activity. For transformants that remained positive for lacZ expression with the bait but negative with the vector alone, plasmid DNA was isolated, transformed into E. coli DH10B, and subjected to further detailed analyses.
DNA Sequence Analysis-Sequencing of plasmid DNA was carried out either using DYEnamic ET Terminator Chemistry (Amersham Biosciences) with reactions analyzed at the Australian Genome Research Facility (Melbourne, Australia) or using ABI PRISM BigDye Terminator chemistry (Applied Biosystems) with reactions analyzed on an Applied Biosystems 3730S capillary sequencer apparatus at Micromon DNA Sequencing Facility, Monash University (Melbourne, Australia). Sequence analysis and basic sequence manipulation were performed using Sequencher (GeneCodes). Electronic data base searches were conducted using the BLASTN and BLASTP algorithms on the NCBI Web site (www.ncbi.nlm.nih.gov/BLAST/). Domain structure of proteins was investigated using the Simple Modular Architecture Research Tool Web site (smart.embl-heidelberg.de/) or the Conserved Domain Database (www.ncbi.nlm.nih.gov/Structure/ccd/wrpsb.cg). To determine the signal peptide cleavage sites in amino acid sequences, the SignalP 3.0 Server (available on the World Wide Web at www.cbs. dtu.dk/services/SignalP/) and PSORT II server (38) were used. Amino acid sequence alignments were carried out using the ClustalX algorithm (39).
␤-Galactosidase Filter and Liquid Assays-␤-Galactosidase filter and liquid assays on S. cerevisiae YGH1 transformants were carried out as described in Ref. 70. For the liquid assays, the protocol using chlorophenol red-␤-D-galactopyranoside (Roche Applied Science) as the substrate was followed.
Immunoblot Analysis of S. cerevisiae YGH1 Transformants-Total protein extracts from S. cerevisiae YGH1 transformants were prepared using a trichloroacetic acid extraction method as described in Ref. 70. Approximately 50 g of protein was subjected to SDS-PAGE (12.5% (v/v) acrylamide). Proteins were transferred to Hybond-C nitrocellulose membrane (Amersham Biosciences) using a Trans-Blot semi-dry electrophoretic transfer apparatus (Bio-Rad) and detected with rabbit-derived antibodies directed against the ATP7A C terminus (diluted 1:500), the GAL4 DNA-BD (2 g/ml) (Sigma), or the GAL4 AD (1 g/ml) (Sigma). Secondary antibody used for detection was a goat anti-rabbit IgG peroxidase conjugate (Chemicon). Protein bands were detected after incubation in Lumi-Light Chemiluminescence blotting substrate (Roche Applied Science) and visualized using the Luminescent Image Analyzer LAS-3000 (Raytest Isotopenmessgeraete).
Mammalian Cell Transfection and Co-immunoprecipitation-Transient transfection of plasmid DNA into mammalian fibroblast cell lines was carried out using Lipofectamine (Invitrogen) and recommended protocols. Co-immunoprecipitation was carried out essentially as previously described (40). Briefly, cell lysates were prepared from confluent monolayers of transfected or untransfected fibroblasts cultured in a 75-cm 2 flask as described (40). For immunoprecipitation, 5 g of either a rabbit polyclonal anti-c-Myc antibody (Sigma) or rabbit preimmune serum was added to an aliquot of each lysate, and ϳ50 l of Immuno-Pure immobilized Protein G Plus (50% slurry; Pierce) was added to precipitate immune complexes. Immune complexes were resuspended in Laemmli sample buffer, boiled for 5 min, and fractionated by SDS-PAGE (12.5% and 7.5% (v/v) acrylamide). Proteins were transferred to Hybond-C nitrocellulose membrane (Amersham Biosciences) using a Trans-Blot semi-dry electrophoretic transfer apparatus (Bio-Rad) and detected with either an affinity-purified preparation of a sheep anti-ATP7A antibody (diluted 1:1000) 4 or a polyclonal rabbit anti-c-Myc antibody (Sigma). Secondary antibodies used for detection were a rabbit anti-goat IgG peroxidase conjugate (Chemicon) and a goat anti-rabbit IgG peroxidase conjugate (Chemicon), respectively. Protein bands were detected after incubation in Lumi-Light Chemiluminescence blotting substrate (Roche Applied Science) and visualized using the Luminescent Image Analyzer LAS-3000 (Raytest Isotopenmessgeraete).
RT-PCR and Northern Blot-Total RNA was isolated from cell pellets using a combination of TRIZOL reagent (Invitrogen) and RNeasy Mini Kit (Qiagen) following the manufacturer's recommendations for the RNeasy Mini Kit. For Northern blot analysis, fractionation of ϳ10 g of RNA on a 1% (w/v) agarose gel containing 1ϫ MOPS buffer (pH 7) (20 mM MOPS, 2 mM NaOAc, 1 mM EDTA) and 0.66 M formaldehyde, transfer to a Hybond-Nϩ membrane (Amersham Biosciences), and hybridization of labeled probe DNA were carried out as previously described (41). Membranes were exposed to an imaging plate (Fujifilm) overnight at 4°C, and bands were visualized using Bio-imaging Analyzer Systems BAS-1800 (Raytest Isotopenmessgeraete). Synthesis of radiolabeled probes was carried out using [␣-32 P]dCTP Redivue tip (3000 Ci/mmol) (Amersham Biosciences) and a Random Primed DNA Labeling Kit (Roche Applied Science).
For cDNA synthesis, ϳ5 g of total RNA was mixed with 25 ng of random hexanucleotide primers (Roche Applied Science), 0.3 mM dNTPs, 1ϫ avian myeloblastosis virus reverse transcriptase buffer (Roche Applied Science), 2.5 units avian myeloblastosis virus reverse transcriptase (Roche Applied Science) and incubated at 42°C for 60 min. Approximately 2 l of this reaction was used for amplification of the cDNA using a MasterTaq Kit (Eppendorf) according to the manufacturer's recommendations and the appropriate oligonucleotide primers. The reactions were placed in a thermocycler, and the temperature was cycled as follows: 95°C for 2 min; 95°C for 20 s, 55°C for 30 s, 72°C for 1 min (for a total of 10 cycles); 95°C for 20 s, 55°C for 30 s, 72°C for 2 min (for a total of 20 cycles); 72°C for 5 min, 5°C for 10 s.

RESULTS
Identification of an ATP7A-interacting Partner-DNA encoding the entire cytoplasmic C terminus of ATP7A (ATP7A-C), from amino acid 1404 to the stop codon, was cloned into the bait vector pAS2-1 (TRP) to create an in-frame fusion with the GAL4 DNA-BD (pSLB29). To screen for ATP7A-interacting partners, this plasmid was co-transformed into S. cerevisiae YGH1 with a human brain cDNA library constructed in the pACT2 (LEU) vector, which encoded the GAL4 AD. Approximately 3.3 ϫ 10 6 clones were screened on medium lacking histidine, of which 2466 were Hisϩ and 77 were consistently Hisϩ/LacZϩ. After curing the bait plasmid from the first 20 of these Hisϩ/LacZϩ positive clones, analysis of plasmid DNA from the TrpϪ/Leuϩ auxotrophs revealed that two were true positives. DNA sequence analysis of the inserts from these two plasmids followed by a search of the nucleotide data base revealed that they were identical to sequences previously deposited in GenBank TM and encoded uncharacterized proteins. This report focuses on one of these, a hypothetical protein that corresponded to several independent data base entries (TABLE ONE) and in this study is encoded by plasmid pSLB56. The ϳ900-bp insert in this plasmid contained 390 bp of the 423-bp ORF, lacking the first 33 bp, and 486 bp of 3Ј-untranslated sequence to the poly(A) tail.
In a similar study that utilized the C terminus of the plasma membrane calcium ATPase (PMCA) splice variant 2b as a bait to screen a human brain cDNA library, the same protein was identified as a PMCAinteracting partner and was designated PISP (PMCA-interacting single-PDZ protein) (42). We propose a change in name for this protein to AIPP1 (ATPase-interacting PDZ protein) to reflect its role in interacting with both of these ATPases.
␤-Galactosidase filter and liquid assays were carried out on S. cerevisiae YGH1 co-transformants that expressed AIPP1 and wild-type ATP7A-C (pSLB56 ϩ pSLB29), AIPP1 and mutated/truncated ATP7A-C (pSLB56 ϩ pSLB63 or pSLB65), AIPP1 and the ATP7B C terminus (pSLB75 ϩ pSLB20), and AIPP1 and Atox1 as an unrelated protein (pSLB56 ϩ pCMB152). As a control, transformants that contained pAS2-1/ATP7A-C(pSLB29) ϩ pACT2 vector were included. These experiments demonstrated a significant and specific interaction between AIPP1 and ATP7A-C (Fig. 1). The control showed that the positive ␤-galactosidase result with ATP7A-C and AIPP1 did not arise from nonspecific interaction of ATP7A-C with pACT2 vector-derived proteins. There was no interaction between AIPP1 and the unrelated copper protein Atox1. Significantly, there was no interaction between AIPP1 and the copper ATPase, ATP7B, which is closely related in structure and function to ATP7A. When either the last 52 (not shown) or 15 amino acids were deleted from the ATP7A C terminus, the interaction with AIPP1 was abolished, hence localizing the region of interaction to within the last 15 amino acids of ATP7A (Fig. 1). All of the co-transformants expressed the GAL4 DNA-BD and AD fusion proteins of the expected size, so that the negative ␤-galactosidase results did not arise through lack of expression of the appropriate proteins (Fig. 1A).
Co-immunoprecipitation experiments were carried out to confirm the interaction between ATP7A and AIPP1 in mammalian cells. An expression construct was generated that encoded the full-length AIPP1 ORF with an N-terminal c-Myc epitope tag. When this construct was transiently transfected into either CHO-K1 cells or human fibroblasts and analyzed for expression of the Myc-tagged protein (AIPP1-Myc) by Western blot analysis with an anti-c-Myc antibody, a protein with an apparent molecular mass of ϳ17 kDa was detected, consistent with the predicted molecular mass based on its amino acid sequence (data not shown). Therefore, this construct was employed for transfection and co-immunoprecipitation experiments. A clonally pure fibroblast cell line (A12-H9) was originally derived from a Menkes patient and was previously immortalized and transfected to stably express the ATP7A protein (34). For co-immunoprecipitation experiments, this cell line was transfected to transiently express AIPP1, after which a total cell lysate was prepared and incubated with either a rabbit anti-c-Myc antibody or rabbit preimmune serum as a control. SDS-PAGE analysis of the immune complexes precipitated with Protein G beads showed that the anti-c-Myc antibody precipitated both AIPP1-Myc and ATP7A Database accession numbers of all known AIPP1 sequences All GenBank TM entries corresponding to AIPP1 splice variants are shown together with the NCBI protein accession numbers, the length of each sequence entry in bp up to, but not including, the poly(A) tail, the length of the 5Ј-and 3Ј-untranslated region (UTR) available for each entry, the length in amino acids (aa) of the predicted protein, and the reference and/or source from which the sequence was derived if known. NA, not available. from the lysate, whereas the preimmune serum did not precipitate either protein (Fig. 2). We concluded that the interaction between ATP7A and AIPP1 represented a true interaction in a mammalian cell environment.
The ATP7A-C-AIPP1 Interaction Is Not Dependent on Copper Levels in S. cerevisiae-Within AIPP1 and the C terminus of ATP7A, there were no identifiable metal binding motifs or ligands likely to bind copper, so we did not expect the interaction to be copper-dependent. Surprisingly, when yeast transformants that expressed ATP7A-C and AIPP1 were cultured under conditions of copper depletion, by the addition of the copper chelator BCS to the yeast growth medium, a significant increase in the extent of the interaction was observed (Fig. 3). The addition of 50 M copper (as CuSO 4 ) to copper-depleted cells decreased the interaction to basal levels (Fig. 3). Higher concentrations of copper to 500 M were tested and yielded similar results (data not shown). However, Western analysis of BCS and copper-treated cultures showed that there was a detectable increase in the levels of expression of the GAL4 DNA-BD/ATP7A-C and GAL4 AD/AIPP1 fusion proteins as copper was depleted from the medium and a significant decrease when copper was added back (Fig. 3). This observation suggested that the increase in ␤-galactosidase activity with copper depletion could be attributed to increased protein expression levels and that the interaction between ATP7A-C and AIPP1 was not copper-dependent. A possible explanation for the increased protein levels is that copper may compete with zinc for binding to transcription factors, so that depletion of cellular copper enhances transcription factor activity, leading to increased protein expression levels and consequently increased ␤-galactosidase activity. Our observations that the addition of zinc to BCS-treated cultures caused consistently high levels of ␤-galactosidase activity (at least 2-fold that observed in BCS-treated cultures) (data not shown) is consistent with this possibility.
AIPP1 Is a Single PDZ Domain Protein-Protein sequence analysis of AIPP1 failed to identify any predicted transmembrane helices, but identified a putative PDZ domain (Fig. 4). The PDZ domain comprises ϳ80 of the 140 amino acids of AIPP1 and was the region that showed greatest similarity in data base searches with other PDZ domain-containing proteins. Other single PDZ domain proteins with a high level of similarity to AIPP1 included the MALS/VELIS family of proteins, which are the mammalian homologues of Caenorhabditis elegans Lin7, the Golgi-associated PDZ and coiled-coil motif-containing protein (GOPC), and Figure 1. Yeast two-hybrid ␤-galactosidase assay of the interaction between AIPP1 and the ATP7A C terminus. A, ␤-galactosidase activity of S. cerevisiae YGH1 co-transformants. S. cerevisiae YGH1 was co-transformed with plasmids that encoded AIPP1 and either the wild-type ATP7A C terminus (ATP7A-C), mutated (ATP7A-C[LL-AA]) and truncated (ATP7A-C[⌬1485-1500]) variants of ATP7A-C, the wild-type ATP7B C terminus (ATP7B-C), or Atox1, as indicated. A control was included that was transformed with pAS2-1/ATP7A-C ϩ pACT2 AD vector. ␤-Galactosidase activity of overnight cultures was measured using chlorophenol red-␤-D-galactopyranoside as the substrate. Within a single experiment, the ␤-galactosidase activity of triplicate samples was measured. The values shown represent ␤-galactosidase activity calculated in Miller units and using the average value obtained from at least three independent experiments Ϯ S.D. The panels above the graph show the GAL4 DNA-BD and AD fusion proteins (indicated above each lane) that are expressed in each co-transformant as detected by Western blot with affinity-purified GAL4 DNA-BD and AD antibodies (Sigma), respectively. The proteins in each lane were derived from the same co-transformants that were tested for ␤-galactosidase activity shown directly below. The position of protein molecular weight markers (Fermentas) are shown on the left in kDa. In the bottom panel, the expected protein bands are indicated by an asterisk. B, amino acid sequence of the ATP7A and ATP7B C termini. The sequence of the last 97 amino acids of the wild-type ATP7A C terminus is shown together with the sequence of the LL-AA mutated and truncated forms used for interaction assays with AIPP1. The numbers correspond to the amino acid positions within the ATP7A sequence. The dileucine and dialanine motifs are shown in boldface type, and the region of interaction with AIPP1 is shaded gray. The PDZ binding motif (DTAL) is underlined. The last 96 amino acids of wild-type ATP7B is also shown. ␤ 2 -syntrophin, with amino acid identities within the PDZ domain ranging between 36 and 47%. PDZ domains contain a highly conserved GLGF motif. Of the two glycine residues, only the second is absolutely conserved. In AIPP1, this motif is represented by the sequence QLGF (Fig. 4).
Gene Structure and Transcript Analysis of AIPP1-The gene encoding AIPP1 is located on the X chromosome at Xq13.1 and comprises seven exons and six introns that span ϳ4 kb of DNA. There are eight data base entries derived from independent DNA profiling studies that correspond to the AIPP1 cDNA sequence (TABLE ONE). Of these, seven encode a 140-amino acid protein, whereas one (AY358829) contains a longer ORF that encodes a polypeptide with an additional 32 N-terminal amino acids. When the nucleotide sequences corresponding to AY358829 and NM_016484 were used to search the human genome data base using the BLASTN algorithm, the different start codons of both predicted proteins and the additional sequence of AY358829 were found to reside within exon 2. This observation suggested that the transcript that corresponded to AY358829 may be derived by alternative splicing of exon 2, which introduces an extra 252 nucleotides into the transcript, of which 156 represent 5Ј-untranslated sequence and 96 encode the additional N-terminal 32 amino acids. The sequence upstream of the AY358829 exon 2 sequence corresponded to exon 1 sequence and is common among the AIPP1 entries. Therefore, AY358829 appears to encode an additional isoform of AIPP1. We propose that the shorter isoform be designated AIPP1a and the longer isoform AIPP1b (Fig. 5).
To determine the size and relative abundance of the AIPP1 transcripts, Northern blot analysis was carried out. Total RNA isolated from normal human fibroblasts was probed with an AIPP1 cDNA fragment common to both transcripts. Two transcripts were identified, a minor one at ϳ1.4 kb and a major form at ϳ1.1 kb (Fig. 5A). Based on the length of the cDNA sequences deposited in GenBank TM , the AY358829 cDNA may be derived from the minor transcript, whereas AF151061, AK024746, and BC012996 are likely to be derived from the major transcript. The cDNAs corresponding to NM_016484 and BX537725, although encoding the shorter AIPP1 isoform, are significantly longer (Ͼ1.7 kb) due to an extended 3Ј-untranslated region arising from an alternative polyadenylation start site. A transcript corresponding to these cDNAs was not detected, either due to a much lower abundance or due to tissue-specific expression. In the same experiment, the effect of intracellular copper levels on AIPP1 transcript abundance was investigated. Total RNA was isolated from a Menkes patient fibroblast line (Me32a-T22/2L), and the same cell line transfected to stably express ATP7A (A12-H9), which were previously shown to have high and low intracellular copper levels, respectively (34,43). Hybridization of these RNAs with the AIPP1 cDNA probe fragment revealed that AIPP1 transcript levels were not affected by variation in intracellular copper concentrations (Fig. 5A).
To further verify the existence of the alternatively spliced products in vivo, RT-PCR was carried out on total RNA using oligonucleotide primers complementary to common regions of the AIPP1 cDNA sequences (Fig. 5B). A shorter major product and two less abundant products were amplified. Sequence analysis revealed that the shorter, most abundant product (602 bp) represented AIPP1a, whereas the largest and least abundant product (841 bp) represented AIPP1b. A third product (674 bp) also was amplified and encoded AIPP1a but contained additional 5Ј-untranslated sequence. This splice product was designated AIPP1aЈ. The existence of AIPP1a and AIPP1b were confirmed by RT-PCR using several primer combinations that included those specific for AIPP1b (data not shown). Fig. 5C shows a schematic representation of the intron-exon arrangement from which AIPP1a and AIPP1b are derived.

DISCUSSION
The molecular dissection of ATP7A to identify amino acid residues and signals important for its copper translocation and copper-induced trafficking activities is an ongoing process that has, over the last decade, yielded significant insight into the mechanism of action of this transport protein. N-and C-terminal as well as transmembrane sequences have been implicated in regulating its subcellular localization (27)(28)(29)(30)44), whereas the N-terminal metal binding motifs as well as those residues that are highly conserved and characteristic among the ATPases are required for its copper transport activity (45)(46)(47)(48). Protein kinase-dependent phosphorylation also may be implicated in regulating the localization and/or activity of ATP7A (49). However, the proteins in the cell that act upon and interpret these various signals to regulate and/or facilitate the activities of ATP7A have not been identified.
This study represents part of a larger project whose objective is 2-fold: 1) to identify proteins that interact with ATP7A to provide insight into the mechanisms and interactions that regulate its activities and 2) through the identification of ATP7A-interacting partners, to identify other proteins important in the maintenance of copper homeostasis. With a number of C-terminal signals already identified within ATP7A, the search for interacting partners commenced with a yeast two-hybrid approach in which the entire C terminus was used as a bait to screen a human brain cDNA library.
We identified a new protein, AIPP1, implicated in copper homeostasis through its interaction with ATP7A. AIPP1 is a small, 140-amino acid protein composed mainly of a single PDZ domain. PDZ domains were first recognized in three proteins (the postsynaptic density protein, PSD-95, the Drosophila septate junction protein, Discs-large, and the Figure 2. Co-immunoprecipitation of ATP7A and c-Myc-tagged AIPP1 from transfected human fibroblasts. The fibroblast cell line derived from a Menkes patient and previously transfected to stably express ATP7A (A12-H9) was transfected with pSLB84 to transiently express AIPP1-Myc. Following recovery for 48 h, lysates were prepared from transfected and untransfected cells. ATP7A and AIPP1 were co-immunoprecipitated from the transfected lysate using an affinity-purified rabbit anti-Myc antibody (0.1 g/l) (Sigma). As a control, rabbit preimmune serum (0.3 g/l) was incubated with a separate aliquot of the same lysate. Immune complexes were precipitated with Sepharose-Protein G beads (Pierce) and eluted by boiling in SDS-PAGE/Laemmli loading buffer, prior to loading on SDS-polyacrylamide gels. One-half of each sample was separated on a 7.5% SDS-polyacrylamide gel (top panel) and the other half on a 12.5% SDS-polyacrylamide gel (bottom panel). As controls for protein size and expression, aliquots (100 l) of lysates derived from transfected and untransfected cells were prepared by adding 25 l of 10% (w/v) SDS and 25 l of loading buffer, and 30 l of each of these samples were loaded on each gel. Following electrophoresis, proteins were transferred to Hybond-C nitrocellulose membranes and detected with an affinity-purified antibody directed against the ATP7A N terminus (␣-ATP7A; 1:1000 dilution) or an affinity-purified rabbit polyclonal anti-c-Myc antibody (␣-Myc; 1:250 dilution) (Sigma). The sizes of prestained protein molecular weight markers (Fermentas) are shown on the right in kDa. The ATP7A and AIPP1 proteins are indicated by the arrows on the left. Human proteins with similarity to AIPP1 were aligned using the ClustalX algorithm (39) and BioEdit software (67). Only the AIPP1b sequence was used in the alignment, since it is identical to AIPP1a except for the additional 32 N-terminal amino acids. The predicted signal peptide cleavage site for AIPP1b is indicated by an arrow. Residues that are identical or similar in the majority (60%) of sequences are shaded in black and gray, respectively. The PDZ domain is boxed, and the conserved "GLGF" motif is indicated by asterisks. GenBank TM protein accession numbers are as follows: AIPP1b, AAQ89188; GOPC, NP_065132;  S. cerevisiae YGH1 co-transformants expressing AIPP1 and the wild-type ATP7A C terminus (ATP7A-C) were passaged for two nights in normal growth medium, in medium supplemented with the copper chelator BCS (Sigma), or in medium supplemented with BCS and CuSO 4 at the concentrations indicated. ␤-Galactosidase activity was measured using chlorophenol red-␤-D-galactopyranoside as the substrate. Within a single experiment, the ␤-galactosidase activity of triplicate samples was measured. The values shown represent ␤-galactosidase activity calculated in Miller units and using the average value obtained from at least three independent experiments Ϯ S.D. The data were analyzed by Student's t test, and significant differences (p Ͻ 0.05) were found for the following pairs. *, 0 M BCS plus 0 M copper versus 100 M BCS plus 0 M copper; ***, 100 M BCS plus 0 M copper versus 100 M BCS plus 50 M copper. The panels below the graph show Western blots of the GAL4 DNA-BD/ATP7A-C and GAL4 AD/AIPP1 fusion proteins that are expressed in each culture represented in the graph. Protein extracts from the untreated and treated cultures were subjected to SDS-PAGE and Western blotting with an ammonium sulfate-precipitated preparation of antibodies directed against a peptide within the ATP7A C terminus (diluted 1:500) or with affinity-purified GAL4 AD antibodies (1 g/ml) (Sigma). An image of the Ponceau Red-stained membrane is also shown to demonstrate approximately equivalent protein loading in each lane. epithelial tight junction protein, ZO-1) and are among the most common protein interaction domains in organisms (15,17). The vast majority of proteins that contain PDZ domains are associated with the plasma membrane (50). PDZ proteins have an essential role in transporting and targeting transmembrane, membrane-associated, and cytosolic protein complexes to distinct subcellular locations, as well as organizing and maintaining the formation of functional protein complexes in these locations (reviewed in Refs. 15-17 and 50). PDZ domains interact with specific C-terminal motifs that typically comprise the last four amino acid residues of partner proteins and are classified based on these PDZ binding motifs (16,17,51). A putative class I PDZ binding motif (X(S/ T)X), characterized by a serine or threonine at the Ϫ2-position, was identified at the C terminus of ATP7A (1497DTAL1500).
In yeast, the interaction between ATP7A-C and AIPP1 was not dependent on copper levels. Mutagenesis experiments demonstrated that the dileucine motif was not implicated in the AIPP1/ATP7A-C interaction but that AIPP1 interacted within the last 15 amino acids of ATP7A, consistent with its predicted interaction with the DTAL motif. In a recent study, deletion of this sequence from ATP7A had no effect on its Golgi localization in polarized Madin-Darby canine kidney cells under basal copper conditions but caused the protein to mislocalize to the apical surface and intracellularly under conditions of elevated copper. When basal copper conditions were restored, ATP7A cycled back to the Golgi area (24). These data suggested that the PDZ binding motif is important for the sorting and/or retention of ATP7A at the basolateral membrane under elevated copper conditions. Taken together, data from this study and that of Greenough et al. (24) suggest that AIPP1 is likely to represent the PDZ protein that interacts with DTAL of ATP7A to target and/or stabilize ATP7A at the basolateral membrane under elevated copper conditions.
Recently, the C terminus of PMCA2b was used as a bait to screen a human brain cDNA library for interacting proteins (42). A protein iden- Figure 5. Identification of alternatively spliced variants of AIPP1. A, Northern blot analysis of AIPP1 mRNA transcript levels in human fibroblast cells with normal (GM2069), high (Me32a-T22/2L), and low (A12-H9) intracellular copper environments. Approximately 10 g of total RNA was isolated from each cell line and separated on a formaldehyde-agarose gel (1%). Following transfer to Hybond-N membrane and prehybridization, hybridization was carried out with a 920-bp AIPP1 cDNA fragment random prime-labeled with [␣-32 P]dCTP Redivue tip (3000 Ci/mmol) (Amersham Biosciences). The membrane was washed and subjected to audioradiography using the Bio-imaging Analyzer Systems BAS-1800. Top panel, autoradiograph showing AIPP1 mRNA transcripts as indicated on the right. Bottom panel, formaldehyde-agarose gel showing the 28 S and 18 S rRNA bands as a control for loading. B, RT-PCR amplification of AIPP1 splice variants. i, schematic diagram showing the location of the primers used to obtain PCR fragments. The long black line represents the AIPP1 cDNA sequence. The short black line above represents the unique region of the AIPP1b splice product that is additional to the common AIPP1 sequence at the point where the two dotted lines meet. The start and stop codons are indicated. The primers used for PCR are indicated by arrows. ii, agarose gel (1.5%) depicting the fragments generated by RT-PCR. Total RNA was isolated from a normal human fibroblast cell line (950026), and cDNA was synthesized as described. Primers (LOC51248#7F and LOC51248#8R), corresponding to regions common to both AIPP1a and AIPP1b sequences, were employed for PCR using the cDNA as a template. The sizes of all DNA fragments are shown in bp. The size of each PCR product is indicated on the right. The sizes of the DNA Molecular Weight Marker VII fragments (Roche Applied Science) are indicated on the left. C, schematic representation of the genomic structure of AIPP1 splice variants. The black lines represent intron sequence. The blocks represent exon sequence. Coding sequence is represented by gray blocks, and noncoding sequence is represented by hatched blocks. The genomic structure was determined with the aid of the BLAST search engine against the human genome using the cDNA sequences obtained from GenBank TM (AIPP1a, NM_016484; AIPP1b, AY358829). The sequences were then aligned with the corresponding region of the X chromosome using Sequencher (Genecodes). tified in this screen, PISP, was a single PDZ domain protein that interacted with all PMCA b-splice variants, most likely through recognition of the putative class I PDZ binding motif ETS(L/V) at the C terminus of these proteins. The data base accession numbers that corresponded to PISP were the same as those that corresponded to AIPP1. Therefore, AIPP1 and PISP are the same protein that interacts with two different cation-transporting ATPases. We propose a change in name for this protein to AIPP1 to reflect its interaction with both ATPases. Based on its localization in polarized Madin-Darby canine kidney cells in a punctate pattern within the cytosol and close to the basolateral membrane, the latter known to be enriched in PMCAs, Goellner et al. (42) proposed a role for AIPP1 in sorting of the PMCA b-splice variants to or from the plasma membrane. A similar role for AIPP1 could be envisaged in the context of ATP7A trafficking. Since AIPP1 appears to act in the copperregulated trafficking pathway of ATP7A, this suggests that this pathway is the same or at least utilizes some of the same cellular components as those utilized by the PMCAs to reach the basolateral membrane.
The study by Goellner et al. (42) demonstrated that AIPP1 is expressed ubiquitously in tissues consistent with the ubiquitous expression of both ATP7A and PMCAs. Their study also reported the presence of one AIPP1 transcript and two protein bands as detected by Northern and Western blot, respectively. In contrast, we identified two AIPP1 mRNA transcripts, one of which was a minor form that encoded an additional potential isoform of AIPP1 with a putative signal peptide. This form could feasibly encode the larger protein product detected by Goellner et al. (42). Based on its interaction with two different ATPases and the presence of potentially more than one AIPP1 isoform, it is likely that AIPP1 has multiple roles in the cell, quite possibly interacting with a range of other proteins. As with all of the mammalian copper proteins studied to date, AIPP1 mRNA transcript levels do not appear to be affected by changes in intracellular copper levels.
AIPP1 showed a high degree of similarity with the PDZ domains of other single PDZ proteins, such as GOPC (52)(53)(54), ␤ 2 -syntrophin (55), and in particular the MALS family of proteins, of which there are three members (MALS-1, -2, and -3), which are the mammalian homologues of the C. elegans Lin7 protein (56). All of these proteins play a role in the intracellular transport or targeting of their partner proteins (16,53,(57)(58)(59)(60)(61). The MALS family may be of particular relevance in providing insight into the role of AIPP1. The MALS proteins are expressed throughout the nervous system in the axonal and dendritic compartments and also in epithelial cells. In the kidney and in polarized Madin-Darby canine kidney cells, it has been suggested that MALS plays an active role in targeting proteins to the basolateral membrane (61,62). In neural cells, the PDZ domain of MALS interacts with the NMDA receptor subunit NR2B to form part of a complex that uses microtubules to transport the NMDA receptor, which is contained within vesicles, to the dendrite (16). In addition, a recent study demonstrated that NMDA receptor activation in hippocampal neurons resulted in the trafficking of ATP7A independent of copper concentration and in ATP7A-dependent copper efflux (63). MALS-2, which is specifically expressed in the brain, interacts directly with NR2B and was shown to have a binding preference for C-terminal peptides with the sequence, E(T/S)(R/X)(V/ L/F) (56). This sequence conforms to a class I PDZ binding motif and matches the C terminus of NR2B (ESDV) and is also similar to the ATP7A PDZ binding motif (DTAL). Conceivably, upon NMDA receptor activation, MALS-2 may bind to this motif in ATP7A in addition to NR2B and potentially other ion channels and thus may explain the mechanism of ATP7A trafficking under these conditions. This possibility is supported by the finding that AIPP1 interacts with two different cation transporters with similar but not identical class I PDZ binding motifs, but further experiments will be required to test this hyopothesis.
If AIPP1 functions similarly to MALS and other PDZ proteins, then it is likely to interact with other proteins in the cell in mediating the transport, targeting, or retention of its interacting partners at the basolateral surface. The C terminus of the closely related copper ATPase ATP7B did not interact with AIPP1, and a putative PDZ binding motif, according to current classification systems, could not be identified at its C terminus. This was not an unexpected result, since ATP7B localizes differently from ATP7A, toward the apical membrane of hepatocytes under elevated copper conditions (64 -66).
Disruption of PDZ domain interactions is known to result in disease phenotypes (17). For example, a deletion mutation that removes the last three amino acids of CFTR accounts for the symptoms of cystic fibrosis (CF) in a minority of CF patients (26). It is feasible that disruption of the ATP7A/AIPP1 interaction could cause a Menkes-like, copper deficiency phenotype. In this context, it is interesting to note that the AIPP1 gene is located on the X chromosome (Xq13.1), in the vicinity of the ATP7A locus at Xq13.2-Xq13.3, so that mutations in this gene, as with Menkes disease, would affect predominantly males. Hence, AIPP1 offers itself as an additional candidate in cases where patients might present with such symptoms but in whom a mutation in ATP7A cannot be identified.
The identification of the novel PDZ protein, AIPP1, as an interacting partner of ATP7A represents an important first step toward discovering new proteins in copper homeostasis and offers an exciting development in the study of ATP7A trafficking in the cell. The added benefit of discovering new proteins that modulate copper homeostasis is that these molecules represent potential new targets for therapy in disorders where copper balance is disrupted, and they could potentially serve as candidates for study in disorders of copper transport for which the genetic basis has not yet been uncovered.