ZIP3, a new splice variant of the PKC-zeta-interacting protein family, binds to GABAC receptors, PKC-zeta, and Kv beta 2.

The correct targeting of modifying enzymes to ion channels and neurotransmitter receptors represents an important biological mechanism to control neuronal excitability. The recent cloning of protein kinase C-zeta interactingproteins (ZIP1, ZIP2) identified new scaffolds linking the atypical protein kinase PKC-ζ to target proteins. GABAC receptors are composed of three ρ subunits (ρ1–3) that are highly expressed in the retina, where they are clustered at synaptic terminals of bipolar cells. A yeast two-hybrid screen for the GABAC receptor ρ3 subunit identified ZIP3, a new C-terminal splice variant of the ZIP protein family. ZIP3 was ubiquitously expressed in non-neuronal and neuronal tissues, including the retina. The ρ3-binding region of ZIP3 contained a ZZ-zinc finger domain, which interacted with 10 amino acids conserved in ρ1–3 but not in GABAA receptors. Consistently, only ρ1–3 subunits bound to ZIP3. ZIP3 formed dimers with ZIP1–3 and interacted with PKC-ζ and the shaker-type potassium channel subunit Kvβ2. Different domains of ZIP3 interacted with PKC-ζ and the ρ3 subunit, and simultaneous assembly of ZIP3, PKC-ζ and ρ3 was demonstratedin vitro. Subcellular co-expression of ZIP3 binding partners in the retina supported the proposed protein interactions. Our results indicate the formation of a ternary postsynaptic complex containing PKC-ζ, ZIP3, and GABAC receptors.

Recently, two proteins interacting with the TM3-TM4 loops of GABA C receptor 1 and 2 subunits were identified. A new C-terminal splice variant of the glycine transporter GLYT-1 bound to 1 (32), while the microtubule-associated protein 1B (MAP-1B) interacted with 1 and 2 and was co-localized with GABA C receptors in the retina (33)(34)(35). Although MAP-1B binds microtubuli, the protein was not essential to anchor GABA C receptors at the cytoskeleton of bipolar cell synapses since GABA C receptor expression in MAP-1B-deficient mice was indistinguishable from wild-type animals (36). Importantly, neither GLYT-1 nor MAP-1B interacted with the GABA C receptor 3 subunit or GABA A receptors.
Screening a rat brain cDNA-library for proteins binding to the GABA C receptor 3 subunit, we identified ZIP3, a new C-terminal splice variant of a protein family interacting with the -isoform of protein kinase C (PKC-; Refs. 37,38). ZIP proteins link PKC-to shaker-type potassium channel ␤ subunits (39) and play an important scaffold role in the activation of the transcription factor NF-B (40,41). This study identified a new interaction between ZIP3 and GABA C receptors in vitro and postulates the formation of a PKC-/ZIP3/GABA C receptor macromolecular complex.

EXPERIMENTAL PROCEDURES
All experiments were performed in compliance with the guidelines for the welfare of experimental animals issued by the Federal Government of Germany, the National Institutes of Health, and the Max Planck Society.
Yeast Two-hybrid Techniques-Unless specified otherwise, all reagents were parts of the MATCHMAKER GAL4 Two-hybrid System 3 (Clontech, Palo Alto, CA). The large intracellular loop between TM3 and TM4 of the rat GABA C receptor 3 subunit (TM3-TM4, amino acids 344 -445) was PCR-amplified and subcloned in-frame into the EcoRI-BamHI site of the bait vector pGBKT7-BD for expression as a GAL4 fusion protein. Yeast AH109 cells were sequentially transformed with the 3 bait vector and 1 mg of a rat brain MATCHMAKER cDNA-library cloned in pGADT7-AD using the lithium acetate method (42) and plated on media selecting for reporter gene activations containing 10 mM 3-amino-1,2,4-triazole (3-AT, Sigma) to suppress background growth. Yeast colonies were incubated for 4 days at 30°C and transferred to plates containing 5-bromo-4-chloro-3-indoxyl-␤-D-galac-topyranoside (Sigma) to test ␤-galactosidase reporter gene activation. Prey plasmids were isolated from positive yeast colonies using CHROMA SPIN-1000 columns (Clontech), shuttled into Escherichia coli DH5␣ (Invitrogen), and retransformed into yeast cells together with pGBKT7-Lam encoding for human lamin C to test for transactivation. Library inserts of positive, retested interactors were sequenced (43) using an automatic DNA sequencer (AbiPrism 377, Applied Biosystems, Foster City, CA) and analyzed with protein and nucleotide databases of the National Center for Biotechnology Information (NCBI, Bethesda, MD) using the Basic Local Alignment Search Tool (BLAST, 44).
To map interacting protein domains, PCR-amplified DNA fragments of ZIP3, GABA C , and GABA A receptor subunits (see Fig. 2 for details) were subcloned in bait or prey yeast vectors using the EcoRI-XhoI sites. After generating individual yeast strains expressing the constructs, protein-protein interactions were analyzed by transforming bait strains with pGADT7-constructs and prey strains with pGBKT7-constructs. Activation of the ADE2, HIS3, and ␤-galactosidase reporter genes was analyzed on selection plates as described above containing 3 mM or 10 mM 3-AT. Transactivation of all constructs was tested using one of the two positive control plasmids pGBKT7-53 or pGADT7-T of the MATCHMAKER GAL4 Two-hybrid System 3 (Clontech) encoding for the murine tumor suppressor protein p53 and the SV40 large T-antigen, respectively. Semi-quantitative intensities of protein-protein interactions were calculated according to the "Yeast Protocols Handbook" from Clontech (Palo Alto, CA) using o-nitrophenyl-␤-D-galactopyranoside (Sigma) as a substrate. Values are expressed as arbitrary ␤-galactosidase units and represent the mean of the reporter gene activity of three yeast colonies. Error bars are Ϯ S.E.
Pull-down Techniques-Unless otherwise stated, all reagents were purchased from Novagen (Madison, WI). The TM3-TM4 loops of the rat GABA C receptor 1-3 subunits, the complete coding sequences of the rat ZIP1-3, and of the rat shaker-type potassium channel Kv␤2 subunit were ligated in-frame to the coding sequence of glutathione-S-transferase (GST) in pET-41 or fused to the His tag of pET-30. The coding sequences of ZIP3 and PKC-were tagged with a T7-epitope by cloning in pET-21. Plasmids were transformed in Escherichia coli BL21(DE3)pLysS, and protein expression was induced by adding 1 mM isopropyl-beta-D-thiogalactoside (Sigma). Fusion proteins were purified under native conditions from frozen bacteria pellets by incubating for 30 min in ice-cold lyses buffer (50 mM NaH 2 PO 4 , 300 mM NaCl, pH 8.0) containing 25 units/ml Benzonase (Novagen), 1 mg/ml Lysozyme (Sigma) and a mixture of protease inhibitors (Roche Molecular Biochemicals), and subsequent sonication (6 bursts for 10 s at 300 W). Alternatively, the "BugBuster GST-Bind-Purification Kit" from Novagen was used. GST-and His-tagged fusion proteins were immobilized to glutathione-Sepharose or Ni-NTA beads that were pre-incubated with 0.1% Triton-X100 and 0.1% bovine serum albumin in the binding buffers for GST beads (in mM: 4.3 Na 2 HPO 4 , 1.47 KH 2 PO 4 , 137 NaCl, 2.7 KCl, pH 7.3) or Ni-NTA agarose (in mM: 500 NaCl, 20 Tris-HCl, 5 imidazole, pH 7.9). The protein concentration of the coated beads was estimated from a Coomassie Brilliant Blue R-250 (Serva, Heidelberg, Germany)stained SDS-PAGE. Similar concentrations of immobilized fusion proteins were incubated with the cytosolic fraction of E. coli expressing T7-tagged ZIP3 or PKC-for 2 h at 4°C under slow agitation, followed by four washes (GST: 4.3 Na 2 HPO 4 , 1.47 KH 2 PO 4 , 137 NaCl, 2.7 KCl, 0.1% Triton X-100, pH 7.3; Ni-NTA: 500 NaCl, 20 Tris-HCl, 60 imidazole, 0.1% Triton X-100, pH 7.9; all concentrations in mM). To obtain comparable conditions in competition experiments, E. coli protein extracts of similar protein concentrations, as measured at 280 nm, were used. The total volume of these samples was adjusted to 500 l (defined as 100%) by adding protein extract of non-transfected E. coli BL21(DE3)pLysS. Bound proteins were eluted by boiling in SDS sample buffer, separated by SDS-PAGE, and analyzed by Western blotting using a monoclonal anti-T7 immunserum and the enhanced chemiluminescence system (ECL; Amersham Biosciences).
The complete coding sequence of the rat 3 subunit was ligated in pCR3.1 (Invitrogen) by PCR cloning techniques. The protein was synthesized using the T7 promoter of the plasmid according to the manuals provided with the RiboMAX RNA Production-Kit (Promega) and the Flexi Rabbit Reticulocyte Lysate System (Promega) in the presence of 2.4% canine pancreatic microsomal membranes vesicles (Promega) and 0.8 mCi/ml (specific activity Ͼ1 mCi/mmol) [ 35 S]methionine (Amersham Biosciences). The reaction mixture was incubated with ZIP3 immobilized on glutathione-Sepharose in the presence of 0.1% Triton X-100. Bound proteins were washed as described above and separated by SDS-PAGE. Gels were dried and exposed to Hyper film (Amersham Biosciences) for 16 h to 5 days.
Preparation of Protein Extracts from HEK-293 Cells and Rat Brain-Human embryonic kidney cells (HEK-293, ATCC CRL1573) were transfected (45) with cDNAs encoding for T7-tagged ZIP3, the 3 TM3-TM4 loop fused to GST, or GST alone. Proteins were expressed under control of the Rous sarcoma virus promoter. Cells were lyzed in RIPA buffer (1% Triton X-100, 0.5% sodium deoxycholate, 0.1% SDS in phosphatebuffered saline, pH 7.4) containing 1 mM dithiothreitol (Sigma), 25 units/ml Benzonase, and a protease inhibitor mixture (Roche Molecular Biochemicals) and sedimented for 30 min at 13,200 ϫ g at 4°C. The supernatant was diluted 1:4 in phosphate-buffered saline supplemented with 1 mM dithiothreitol and mixed with glutathione-Sepharose beads (pre-incubated as above) for 5 h at 4°C. Subsequent washing, elution, and Western blot detection of interacting proteins was performed as described above.
Adult rat brains were homogenized on ice in 10 ml/mg tissue of homogenization buffer I (0.32 M saccharose, 20 mM Tris-HCl, pH 7.4) containing 10 mg/ml DNaseI and protease inhibitors (Roche Molecular Biochemicals) using a glass/Teflon homogenizer and centrifuged at 30,000 ϫ g for 30 min, and the resulting crude membrane pellet was homogenized again in 5 ml/mg tissue of hypotonic homogenization buffer II (20 mM Tris-HCl, pH 7.4, plus protease inhibitors) to release proteins inside cytosolic vesicles. After centrifugation at 30,000 ϫ g for 30 min, the supernatant was separated completely from the pellet, which was solubilized in 5 ml/mg tissue of buffer III (20 mM Tris-HCl, 200 mM NaCl, 1% Triton X-100, pH 7.4, plus protease inhibitors) for 2 h. Ultracentrifugation was carried out at 100,000 ϫ g for 1 h, and the supernatant was saved (P2). For binding assays, ϳ5 mg of S1 and P2 protein fractions were incubated with beads coated with GST-ZIP3, the His-tagged 3 TM3-TM4 loop, or GST and His tags as negative controls for 5 h under slow agitation. Bound proteins were washed, eluted and visualized as described above. All protein preparation steps were carried out on ice or at 4°C.
Immunocytochemistry-Adult Wistar rats were anesthetized with halothane and decapitated. The posterior eye cups with the retinas attached were immersion-fixed for 15-30 min in 4% (w/v) paraformaldehyde in phosphate buffer (PB; 0.1 M, pH 7.4). The retinas were dissected and cryoprotected in 10% (w/v), 20% (w/v) sucrose in PB for 1 h each, and in 30% (w/v) sucrose in PB overnight at 4°C. Pieces of retinas were mounted in freezing medium (Reichert-Jung, Bensheim, Germany), sectioned vertically at a thickness of 12 m with a cryostat and processed for immunocytochemistry.
The following antibodies were used: a rabbit polyclonal antiserum (1:100) recognizing the 1, 2, and 3 subunits of the rat GABA C receptor (19), a mouse monoclonal antibody against PKC-␣ (1:200; Dunn Labortechnik, Asbach, Germany)-labeling rod bipolar cells, and a rabbit polyclonal antiserum against PKC-(1:10.000; Sigma). In double-labeling immunocytochemical experiments, co-expression of PKC-␣ and PKC-and of PKC-␣ and GABA C receptors was examined. The binding sites of the primary antibodies were revealed by the secondary antibodies Alexa TM 594 (red fluorescence) and Alexa TM 488 (green fluorescence) goat anti-mouse or goat anti-rabbit IgG (H ϩ L) conjugates, respectively (1:500; Molecular Probes, Eugene, Oregon). Double-labeled sections were examined and analyzed with a confocal laser-scanning microscope (LSM 5 Pascal, Zeiss, Oberkochen, Germany), and resulting images were adjusted in brightness and contrast using Adobe Photo-Shop 5.5 (Adobe Systems Inc., San Jose, CA).
Isolation of RNA and Reverse Transcription-PCR-Total RNA was extracted from brain, lung, liver, kidney, and spleen using the TRIZOL Reagent according to the manufacturer's protocol (Invitrogen) and from cerebellum, cortex, hippocampus, olfactory bulb, retina, thalamus, and spinal cord of an adult rat following a method described by Chomczynski and Sacchi (46). For cDNA-synthesis, 3 g of total RNA were incubated in 20 l of cDNA-synthesis buffer (in mM: 50 Tris-HCl, 3 MgCl 2 , 75 KCl, 10 dithiothreitol, pH 8.3), 0.5 mM of each dNTP (Amersham Biosciences), 250 ng of p(dN) 6 (Boehringer, Mannheim, Germany), 20 Units RNasin (Roche Molecular Biochemicals) and 400 units of SuperscriptII RNaseH Ϫ reverse transcriptase (Invitrogen). Incubation times were 15 min at room temperature followed by 2 h at 42°C. PCR amplification was performed with 250 ng of reverse transcribed RNA in 50 l of PCR-buffer (in mM: 20 Tris-HCl, 50 KCl, 1.5 MgCl 2 , 0.2 dNTPs, and 0.5 M each ZIP3 primer (sense 5Ј-CAGCAAGCTCATCTTTCCcaac-3Ј, nt 480-501; antisense 5Ј-CTACTTATGACACTTAAAgcca-3Ј, nt 684 -705), 5 units Taq-polymerase (Invitrogen), pH 8.0) in a thermocy-cler (Applied Biosystems) using the following parameters: 94°C for 2 min followed by 30 cycles at 94°C for 45 s, 60°C for 60 s, 72°C for 30 s and a final incubation at 72°C for 10 min. To compare amounts of RNA isolated from each tissue as well as the efficiency of reverse transcription between RNA samples, a PCR with oligonucleotides recognizing ␤-actin (sense: 5Ј-tgagaccttcaacaccccag-3Ј, nt 372-391; antisense: 5Јgtagacgaccttccacctgt-3Ј, nt1065-1046) was performed for 25 cycles with the parameters described above. Five microliters of each PCR product were separated on a 1.5% agarose gel and stained with ethidium bromide. Controls were treated as above without adding template and/or reverse transcriptase and showed no PCR products. To identify the PCR products, a Southern blot was performed using standard techniques (47). Oligonucleotides for ZIP3 (5Ј-catgggcactttggctggc-3Ј, nt 559-577), and ␤-actin (5Ј-cggtcaggtcatcactatc-3Ј, nt 732-750) were tailed with DIG-ddUTP, hybridized and detected following the protocol of the "DIGoligonucleotide Tailing Kit" and the "DIG Luminescent Detection Kit" (Roche Molecular Biochemicals) as described in the user manuals.

ZIP3 Represents a New C-terminal Splice Variant of the PKC--interacting Protein
Family-To identify proteins that bind to the GABA C receptor 3 subunit, a rat brain cDNA library was screened against the intracellular TM3-TM4 loop of this subunit. Of ϳ1.9 ϫ 10 7 transformed yeast cells, 100 yeast colonies were isolated on selection plates and analyzed further. Upon sequencing, clones 56 and 72 were of specific interest since they represented two independent clones coding for an unknown C-terminal splice variant of the PKC--interacting proteins ZIP1 and ZIP2 (38). Therefore, the protein was termed ZIP3. The N-terminal part of ZIP3 was identical to ZIP1 and ZIP2 and contained a recently characterized acidic putative protein-binding motif described as cdc-homology domain (37,39), a ZZ-zinc finger domain that also has been associated with protein-protein interactions (48), and two consensus sequences for phosphorylation by PKC (Fig. 1A). However, due to the new C terminus, ZIP3 did not contain the PEST and ubiquitinassociated domains present in ZIP1 and ZIP2 (39). Interestingly, the ZIP3-specific C terminus started at the same splice site that is used to generate the ZIP1-specific cassette missing in ZIP2 ( Fig. 1A; Refs. 39,49).
To analyze the distribution of ZIP3, total RNA from different tissues was reverse-transcribed and subjected to PCR amplification. ZIP3 was present at similar levels in all organs analyzed, including the brain (Fig. 1B). Within the central nervous system, ZIP3 was abundantly expressed in spinal cord, thalamus, cortex, and the retina (Fig. 1B, right panels). PCR products of similar intensity for ␤-actin indicated that approximately equal amounts of cDNA from each tissue were used.
Mapping of Interacting Domains between ZIP3 and GABA C Receptor Subunits-The highest sequence diversity between GABA C receptor subunits is found in their TM3-TM4 loops. Thus we analyzed whether besides the 3 subunit additional subunits of the GABA C or the structurally related GABA A receptor were able to interact with ZIP3. PCR products representing TM3-TM4 loops were cloned in the bait vector of the yeast two-hybrid system and expressed fusion proteins were analyzed for their ability to interact with ZIP3. Protein-protein interaction was monitored by the ability of transformed yeast cells to grow on selective media, containing 3 mM or 3-AT. To our surprise, TM3-TM4 loops of all three subunits bound ZIP3, with the 3 loop showing the highest binding affinity ( Fig. 2A, left panel). In contrast to GABA C receptor subunits, no interaction was observed between ZIP3 and subunits of the GABA A receptor ( Fig. 2A, middle panel).
The specific interaction of ZIP3 with the subunits of the GABA C receptor allowed the identification of amino acids important for the binding. An alignment of the TM3-TM4 loops of the rat 1-3 subunits identified two regions of high similarity, located at the very N-and C-terminal ends of the loops (Fig. 2B, upper half). To determine the involvement of these regions in ZIP3 binding, subsequent N-and C-terminal deletions were generated in the 3 TM3-TM4 loop, and the resulting protein fragments were tested directly for their ability to interact with ZIP3 in yeast cells (Fig. 2B, lower half) as described above. While amino acid regions located at the C-terminal part of the 3 loop showed no binding to ZIP3, 10 amino acids at the very N-terminal part of the loop were sufficient for the interaction (gray background in Fig. 2B). To analyze if the deletions changed the 3 binding affinity for ZIP3, the relative binding strength was estimated using a semi-quantitative ␤-galactosidase assay. Compared with the binding strength of the complete 3 TM3-TM4 loop, the relative ZIP3 binding affinities of the 3 constructs decreased slightly with subsequent deletions. Thus, construct 3-D3 contained a minimal ZIP3 binding region; however, additional amino acids present in constructs 3-D1 and 3-D2, which are absent in the TM3-TM4 loops of 1 and 2, support the interaction. This result is consistent with our finding that all three subunits did interact with ZIP3 and that 3 showed a higher binding strength than 1 and 2 ( Fig. 2A, left panel).
To map the location of amino acids that would act in combination with the 10 amino acids of the 3-D3 construct, we generated three additional deletions of the 3 TM3-TM4 loop. Upon deletion of the first 10 amino acids (construct 3-D6), the binding affinity for ZIP3 was reduced by about 50%, indicating that this domain is an important but not the only mediator for the interaction between 3 and ZIP3. When the first 10 amino acids were used in combination with more C-terminal domains of the 3 TM3-TM4 loop (constructs 3-D7 and 3-D8), the binding strength increased to values comparable with the wildtype sequence. Next, regions of ZIP3 were analyzed for their capability to bind the 3 TM3-TM4 loop, again using a strategy of subsequent deletions. The ZIP3/3 interaction was mediated by the ZZ-zinc finger and a C-terminal adjacent protein region (amino acids 119 -221, gray background in Fig. 2C), while the ZIP3specific C terminus and the cdc-homology region were not involved in the binding. Dividing the interacting protein region in two parts resulted in a reduction in binding affinities (constructs ZIP3-D2 and ZIP3-D7). However, interaction with 3 was still present, indicating that amino acids in both protein regions contribute to the binding site in a synergistic manner. Indeed, the ZZ-zinc finger domain alone (construct ZIP3-D9) was able to bind the 3 TM3-TM4 loop at a binding intensity similar to those of constructs ZIP3-D2 and ZIP3-D7.
ZIP3 Is Able to Dimerize and to Interact with GABA C Receptor Subunits, Kv␤2, and PKC--Several proteins have been described in the literature to physically interact with ZIP1 or ZIP2, including PKC-, shaker-type potassium channel ␤ subunits, or the ZIP proteins themselves (37,39). Therefore, we analyzed the binding characteristics of the newly identified member of the ZIP protein family, ZIP3. Immobilized GST fusion proteins were incubated with E. coli protein extracts and bound proteins were analyzed in Western blots. Unspecific interactions were excluded using immobilized GST. ZIP3 bound specifically to ZIP1, ZIP2, and ZIP3 and to the TM3-TM4 loops of the 1, 2, and 3 subunits (Fig. 3A), consistent with the results obtained from the yeast two-hybrid experiments. Furthermore, ZIP3 interacted with Kv␤2 (Fig. 3A) and with PKC- (Fig. 3B). Since the N-terminal regions of ZIP1, ZIP2, and ZIP3 are identical, our data indicate that the ZIP dimerization domain and the binding regions for PKC-and Kv␤2 are located in the N termini. Indeed, the region important for dimerization and binding of ZIP1 to PKC-was mapped to the cdc-homology region (see Fig. 1; Refs. 37,39,41,50).
In a next step, we analyzed the capability of native proteins to interact with ZIP3. Immobilized GST fusion proteins were incubated with cytosolic (S1) or membrane (P2) protein fractions prepared from adult rat brains. Consistent with our pre-

FIG. 2. Domain mapping in ZIP3
and GABA C receptor subunits. A, the TM3-TM4 loops of GABA A/C receptor subunits were tested in binary yeast twohybrid experiments for their ability to bind ZIP3. Encoded amino acids are represented by numbers in parentheses. Activation of HIS3, ADE2, and ␤-galactosidase reporter genes was monitored by the ability of transformants to grow on selective media with the addition of 3 mM (ϩ) or 10 mM 3-AT (ϩϩϩ), indicating low or high binding strength, respectively. (Ϫ) indicates no growth of yeast colonies. B, upper half: alignment of the TM3-TM4 loops of the rat 1-3 subunits. (/) replaces a region not shown because of low homology. Lower half: the complete 3 TM3-TM4 loop and deletion constructs were tested for binding to ZIP3 as described in A. The identified minimal interacting region identical between the subunits is highlighted in gray. Relative binding intensities of interacting constructs were quantified to 1.88 Ϯ 0.41 (3) vious results, ZIP3 interacted specifically with PKC-and Kv␤2 (Fig. 3C). Nonspecific interactions were excluded using antibodies against protein phosphatase 1C (PP1C; 51) and against a protein interacting with C-kinase (PICK1; Ref. 52) as negative controls for cytosolic (S1) and membrane associated (P2) protein preparations, respectively. So far, no report described a physical interaction between PP1C and PICK1 with ZIP proteins, and indeed, no binding was detected under our experimental conditions, demonstrating the specificity of the assay (Fig. 3C).
Specific immunsera to detect native ZIP3 or the GABA C receptor 3 subunit on Western blots do not exist, which prevented us from analyzing the interaction between these two proteins in native tissue. To circumvent this fact, we analyzed whether the interaction between ZIP3 and the GABA C receptor 3 subunit occurred when the binding partners were synthesized in mammalian cells. For this purpose, HEK-293 cells were co-transfected with T7-tagged ZIP3, and the 3 subunit TM3-TM4 loop fused to GST or T7-ZIP3 and GST as a control. Proteins interacting with the 3 loop were precipitated using glutathione-Sepharose beads and analyzed on a Western blot using an anti-T7 immunserum. In agreement with our previous data, ZIP3 was able to bind specifically to the 3 subunit, while no interaction could be observed with GST (Fig. 3D) or in cells co-transfected with the green fluorescent protein instead of ZIP3 (data not shown).
In addition, we tested whether the full-length 3 subunit would be able to bind to ZIP3. The 3 subunit was synthesized in vitro using a rabbit reticulocyte lysate in the presence of canine pancreatic microsomal membranes and subsequently incubated with immobilized GST-ZIP3 fusion proteins. Consistent with our data, we observed an interaction between the full-length 3 subunit and ZIP3, but not between 3 and GST (Fig. 3E).
Indication for a PKC-/ZIP3/GABA C Receptor-containing Macromolecular Complex-Our findings demonstrated that ZIP3 was able to dimerize with ZIP family members and to interact with GABA C receptor subunits, PKC-and Kv␤2, in vitro. We mapped the ZIP3 binding site for subunits to a region different from the cdc-homology domain, that has been shown to mediate the interaction with PKC-(see Fig. 2C; Ref. 37). Therefore, ZIP3 could bind to subunits and PKC-at the same time, acting as a scaffold, physically linking PKC-to GABA C receptors. To analyze whether the 3 subunit and PKC-were indeed able to bind simultaneously to ZIP3, competition experiments were performed. Glutathione-Sepharose beads coated with the TM3-TM4 loop of the 3 subunit were first incubated with 500 l (equivalent to 100% of the total volume) of ZIP3 containing E. coli protein extract, washed extensively, and subsequently mixed with increasing amounts of PKC-containing protein extract. PKC-was not able to displace ZIP3 from binding to the 3 subunit (Fig. 4A), indicating that PKC-and the 3 subunit interacted with different regions of ZIP3, consistent with our previous data (Fig. 2C). Furthermore, this experiment directly demonstrated that ZIP3, PKC-, and the 3 subunit were able to form a ternary complex in vitro, suggesting the possibility that ZIP3 could serve as a scaffold protein at synapses expressing GABA C receptors.

FIG. 3. Recombinant and native proteins binding to ZIP3.
To analyze protein interactions of ZIP3 in vitro, GST and GST fusion proteins were immobilized on glutathione-Sepharose beads and incubated with T7-tagged ZIP3 (A) or PKC-(B) purified from E. coli as indicated. Bound proteins were detected on Western blots (upper panels) using a monoclonal anti-T7 immunserum. The concentration of GST and GST fusion proteins bound to Sepharose beads is shown on Coomassie-stained SDS-PAGE (arrowheads in lower panels). C, to analyze the binding of native proteins to ZIP3, cytosolic (S1) or membrane (P2) protein preparations of rat brain were incubated with GST or GST-ZIP3 immobilized on glutathione-Sepharose. Bound proteins were detected on Western blots using specific immunsera as indicated on the right. Antibodies against PP1C and PICK1 served as negative controls to ensure specificity of the assay. The concentration of GST and GST fusion proteins bound to Sepharose beads is shown on Coomassie-stained SDS-PAGE (arrowheads in the lowest two panels). D, to verify the interaction between ZIP3 and the GABA C receptor 3 subunit in mammalian cells, HEK-293 cells were co-transfected with ZIP3 and GST or ZIP3 and GST fused to the TM3-TM4 loop of the 3 subunit. Cell lysates were incubated with glutathione-Sepharose beads, and bound proteins were analyzed using a monoclonal anti-T7 antibody. E, to verify the interaction between ZIP3 and the full-length 3 subunit, 3 was synthesized in vitro using radioactive methionine in the presence of microsomal membranes. The lysate was incubated with immobilized GST-ZIP3, and the bound 3 subunit was detected radiographically. In all panels, the Benchmark prestained protein ladder (Invitrogen) or calculated sizes of interacting proteins are indicated in kDa.
ZIP3 might serve as a linker, bringing PKC-in close vicinity of 3-containing GABA C receptors. The 3 TM3-TM4 loop contains a consensus sequence for phosphorylation by PKC; however, a direct binding of PKC-to the 3 subunit would be needed for its phosphorylation. Indeed, we could show a direct interaction between PKC-and the 3 TM3-TM4 loop in native protein preparations (Fig. 4B). Nonspecific interactions between 3 and the protein extract were excluded using antibodies against PP1C (51), that has not been reported to bind GABA C receptors. In addition to the native proteins, PKC-and the 3 loop did also interact when recombinant proteins synthesized in E. coli were used (data not shown).
PKC-and GABA C Receptor Subunits Are Co-expressed in the Same Cellular Compartments in the Mammalian Retina-A prerequisite for any protein-protein interaction is the co-expression of the binding partners in the same cellular compartments. As said before, ZIP3-specific immunsera needed to detect the protein in native tissues were not available. Therefore, we analyzed whether proteins that physically interact with ZIP3, namely PKC-and GABA C receptor subunits, were co-expressed in the rat retina.
In the rat central nervous system, the highest concentration of GABA C receptor subunits was detected in the retina, where they are clustered at synaptic terminals of bipolar cells (22)(23)(24)(25). Vertical cryostat sections of adult rat retinas were doublelabeled with antibodies recognizing GABA C receptor subunits/PKC␣ and PKC␣/PKC-. Stained sections were analyzed using confocal laser-scanning microscopy. Co-expression of PKC-and the subunits of the GABA C receptor had to be shown indirectly as the antisera are generated in the same species (rabbit). Fig. 5A shows that at the rod bipolar cell terminals that stratify deep in the inner plexiform layer and are stained with the antibody against PKC␣, GABA C receptor subunits are clustered. Importantly, in the terminals of the PKC␣-labeled rod bipolar cells, PKC-is co-expressed (Fig. 5B). Co-expression can be clearly seen in the merge of the stainings, showing the terminals of the rod bipolar cells in a higher power view (broken lines mark the region of the inner plexiform layer shown). The results of the staining experiments suggest, although indirectly, the co-expression of GABA C receptor subunits and PKC-in rod bipolar cell terminals.

DISCUSSION
Increasing evidence underlines the importance of macromolecular signaling complexes containing ion channels, neurotransmitters receptors, kinases, and phosphatases for the specific regulation of neuronal excitability. The modulation of neurotransmitter receptors by kinases and phosphatases represents an important mechanism to regulate neuronal activity (54). However, factors controlling the specific targeting of these enzymes are largely unknown. The recently discovered ZIP proteins (38) form a new protein family and physically link the atypical protein kinase C isoform PKC-to target proteins, such as potassium channels (39) or the death-domain kinase RIP, a protein involved in the activation of the transcription factor NF-B (40,41). In this study, we identified ZIP3 as a new member of the ZIP protein family, which is abundantly expressed in the brain and retina. ZIP3 was able to form homoand heterodimers and bound to PKC-and Kv␤2. Furthermore, ZIP3 interacted with GABA C receptor subunits using a different binding site, allowing simultaneous binding of subunits and PKC-in vitro. These results suggest a possible formation of a PKC-/ZIP3/GABA C receptor containing macromolecular complex.
Mapping of protein regions mediating the ZIP3/GABA C receptor binding identified 10 amino acids that are conserved in the TM3-TM4 loops of 1-3 subunits. Compared with the complete 3 TM3-TM4 loop, subsequent C-terminal deletions of the 3 loop resulted in a slight decrease in ZIP3 binding affinity, indicating that the 10 amino acids represent a minimal binding motif, while additional 3 sequences support the interaction. Indeed, when these 10 amino acids were deleted, the binding affinity for ZIP3 was reduced by nearly 50%, while deletion of internal regions of the loop did not significantly alter the binding strength compared with the wild-type. Therefore, we propose at least two binding domains in the 3 TM3-TM4 loop contacting ZIP3, one present in the first 10 amino acids that acts in combination with a more C-terminally located motif. TM3-TM4 loops of GABA A receptor subunits were not able to bind ZIP3. This is consistent with the fact that the identified 10 amino acids of the subunits are not conserved in GABA A receptor subunits.
In this study, we identified a region of ZIP3, including the ZZ-zinc finger domain that has been associated with proteinprotein interactions (48), to be important for the binding to the 3 TM3-TM4 loop. This protein region is identical between ZIP1, ZIP2, and ZIP3, and indeed all ZIP proteins bound to 3 in yeast cells (data not shown). ZIP1 and ZIP2 had been described to bind PKC-and to form homo-as well as het- FIG. 4. ZIP3, PKC-, and the GABA C receptor 3 subunit form a ternary complex in vitro. A, to analyze if PKC-and the 3 subunit would interact simultaneously with ZIP3, competition experiments were performed. Equal amounts of GST/3TM3-TM4 immobilized on Sepharose beads (arrowhead in lowest panel) were first incubated with constant quantities of ZIP3-containing E. coli protein extracts, washed, and subsequently mixed with increasing concentrations of PKCextracts (12-100% of the total volume). Different volumes of PKC-protein extracts were adjusted to 500 l (defined as 100% of the total volume) by adding protein extracts of non-transfected E. coli cells. Bound proteins were detected on Western blots as described in Fig. 3A. B, to test for a direct interaction between PKCand the 3 subunit, cytosolic protein preparations of rat brain were incubated with the His-tagged TM3-TM4 loop of the 3 subunit immobilized on Ni-NTA beads. Bound proteins were detected using a PKC-specific immunserum. In all panels, the Benchmark prestained protein ladder (Invitrogen) or calculated sizes of interacting proteins are indicated in kDa. erodimers using the cdc-homology domain (37,39,41,50). This domain is identical in ZIP3, and accordingly, ZIP3 was able to dimerize with all ZIP proteins and to bind to PKC-. The cdchomology region is located N-terminal of the domain mediating the ZIP3/ subunit interaction, indicating that the binding of ZIP3 to GABA C receptors might be independent from its binding to PKC-or other ZIP proteins. Therefore, ZIP3 could physically link PKC-to the TM3-TM4 loops of GABA C receptor subunits.
To test this hypothesis directly, we competed the binding of the 3 subunit to ZIP3 with increasing amounts of PKC-and could demonstrate the in vitro formation of a ternary complex composed of ZIP3, PKC-, and the GABA C receptor 3 subunit. The binding sites for PKC-and for the 3 subunit were mapped to N-terminal domains of ZIP proteins, which are identical between ZIP1, ZIP2, and ZIP3. Therefore, in principle all ZIP proteins might be able to associate with PKC-and GABA C receptor subunits into a ternary protein complex. However, ZIP3 lacks the PEST and ubiquitin domains present in ZIP1 and ZIP2. Since these domains are associated with protein degradation, ZIP3-containing macromolecular complexes might be more stable within cells compared with ZIP1and ZIP2-containing complexes.
A prerequisite for the formation of the proposed ternary complex composed of ZIP3, PKC-, and GABA C receptors is the expression of the proteins in the same cellular compartments. Since ZIP3 specific immunsera are not available, we compared the expression patterns of the proteins that bind to ZIP3, namely PKC-and GABA C receptors. Within the central nervous system, the highest concentration of GABA C receptor subunits is observed in the retina (22,24,25), where also ZIP3 is abundantly expressed (this study). Therefore, we analyzed the distribution of GABA C receptors and of PKC-in rat retina with immunocytochemistry and confocal microscopy. The results of the experiments suggest the co-expression of the GABA C receptor subunits and of PKC-in the terminals of rod bipolar cells.
Earlier studies that analyzed the retinal distribution of PKC-showed contradictory results. While one study found that PKC-co-localized with PKC-␣ in bipolar cells (55), another study showed PKC-expression exclusively in the inner segments of photoreceptors (56). Here we confirm the co-localization of PKC-and PKC-␣ in rod bipolar cells. Cerebellar Purkinje cells also express GABA C receptors (23) and, similar to the results reported in this study, co-express ZIP proteins and PKC- (39).
PKC consensus sequences are present in the TM3-TM4 loops of rat 1-3 subunits, and GABA C receptor currents are modulated by PKC (27)(28)(29). This suggests a functional link between GABA C receptors and PKC-. Indeed, we could show a direct interaction between PKC-and the 3 TM3-TM4 loop in recombinant and native protein preparations, which would be needed for a protein phosphorylation. However, mutation of PKC consensus sequences of the 1 subunit did not prevent its modulation by PKC (30,31,57), indicating that residues other than those of the consensus sequences are phosphorylated (58). On the other hand, proteins other than the 1 subunit might be the target of the PKC activity. So far, no similar studies have been performed for the 3 subunit, and therefore it remains elusive whether the PKC consensus sequences of this protein are used. Alternatively, two consensus sequences for PKC present in ZIP3 might represent new players in the modulation of GABA C receptors. Furthermore, the synaptic clustering of GABA C receptors might be influenced by its interaction with ZIP3 and PKC-since PKC-interacts with tubulin and the actin cytoskeleton (60 -61).
In summary, we present ZIP3 as a new member of the PKC--interacting protein family that it highly expressed in the mammalian retina and demonstrated in vitro its interaction with PKC-, GABA C receptor subunits and Kv␤2. Therefore we suggest the formation of a postsynaptic macromolecular protein complex at GABA C receptor-containing synapses using ZIP3 as a scaffold.