Protein 4.1N Is Required for Translocation of Inositol 1,4,5-Trisphosphate Receptor Type 1 to the Basolateral Membrane Domain in Polarized Madin-Darby Canine Kidney Cells* 210

Protein 4.1N was identified as a binding molecule for the C-terminal cytoplasmic tail of inositol 1,4,5-trisphosphate receptor type 1 (IP3R1) using a yeast two-hybrid system. 4.1N and IP3R1 associate in both subconfluent and confluent Madin-Darby canine kidney (MDCK) cells, a well studied tight polarized epithelial cell line. In subconfluent MDCK cells, 4.1N is distributed in the cytoplasm and the nucleus; IP3R1 is localized in the cytoplasm. In confluent MDCK cells, both 4.1N and IP3R1 are predominantly translocated to the basolateral membrane domain, whereas 4.1R, the prototypical homologue of 4.1N, is localized at the tight junctions (Mattagajasingh, S. N., Huang, S. C., Hartenstein, J. S., and Benz, E. J., Jr. (2000) J. Biol. Chem. 275, 30573–30585), and other endoplasmic reticulum marker proteins are still present in the cytoplasm. Moreover, the 4.1N-binding region of IP3R1 is necessary and sufficient for the localization of IP3R1 at the basolateral membrane domain. A fragment of the IP3R1-binding region of 4.1N blocks the localization of co-expressed IP3R1 at the basolateral membrane domain. These data indicate that 4.1N is required for IP3R1 translocation to the basolateral membrane domain in polarized MDCK cells.

To study the function of the C-terminal cytoplasmic tail of IP 3 R1, we searched for molecules binding the C-terminal cytoplasmic tail using a yeast two-hybrid system. Protein 4.1N, a homologue of the erythrocyte membrane cytoskeleton protein 4.1, was identified as a binding molecule for the C-terminal cytoplasmic tail of IP 3 R1. Protein 4.1, originally identified in red blood cells and called red blood cell protein 4.1 (4.1R), plays a critical role in the morphology and mechanical stability of the red blood cell plasma membrane (13). Three structural/functional domains have been identified in 4.1R. The N-terminal membrane-binding domain (also called the 30-kDa or FERM domain) (14) possesses binding sites for the cytoplasmic tails of integral membrane proteins such as band 3 (15,16), glycophorin C (17,18), and CD44 (19). An internal domain contains the spectrin-actin binding (also called the 10-kDa domain) activity required for membrane stability (20 -22). The C-terminal domain (CTD, also called the 22-24-kDa domain) has recently been reported to bind to tight junction proteins, ZO-1 and ZO-2 (23), to an immunophilin, FKBP13 (24), and to nuclear mitotic apparatus protein, which appears to mediate the spindle formation in nucleated cells (25). Three homologues of 4.1R have been cloned: the widely expressed homologue 4.1G, the neuronal homologue 4.1N, and the brain homologue 4.1B. They share a high degree of homology with prototypical homologue 4.1R in the three structural/functional domains. Each 4.1 protein is characterized by three unconserved unique domains between the conserved membrane-binding domain, the spectrin-actin binding domain, and the CTD (26).
In this study, we found that 4.1N and IP 3 R1 associated in both subconfluent and confluent Madin-Darby canine kidney (MDCK) cells, a well studied tight polarized epithelial cell line.
Both were predominantly translocated to the basolateral membrane domain when MDCK cells grew from subconfluence to confluence, whereas 4.1R, the prototypical homologue of 4.1N, was localized at the tight junction (23), and other endoplasmic reticulum (ER) marker proteins were still present in the cytoplasm in confluent MDCK cells. The localization of IP 3 R1 at the basolateral membrane domain was determined by its 4.1Nbinding region and could be blocked by a fragment of the IP 3 R1-binding region of 4.1N. These data suggest that 4.1N serves to regulate IP 3 R1 subcellular localization.

MATERIALS AND METHODS
Plasmid Construction-All of the plasmids were propagated in the Escherichia coli strain HB101. All PCR products of the cDNA fragments were generated in frame using Platinum Pfx DNA polymerase (Invitrogen) and were verified by nucleotide sequencing using an ABI PRISM 377 automated sequencer (Applied Biosystems). cDNA encoding the C-terminal cytoplasmic tail of IP 3 R1 (IP 3 R1/CTT, aa 2590 -2749) were generated by PCR from mouse IP 3 R1 cDNA and subcloned into the site of EcoRI and BamHI of pGBT9 (Clontech) to generate pGBT9-IP 3 R1/CTT, into the site of EcoRI and BamHI of pEGFP-C3 (Clontech) to generate GFP-IP 3 R1/CTT, and into the site of BamHI and EcoRI of pGEX-KG (27) to generate GST-IP 3 R1/CTT. Truncated constructs corresponding to different lengths of IP 3 R1/CTT (see Fig. 3A) were subcloned into the site of EcoRI and BamHI of pGBT9. GFP-IP 3 R1-N was generated by subcloning full-length mouse IP 3 R1 fusing with an EGFP cDNA in its N terminus into pcDNA3.1/Zeoϩ (Invitrogen). 2 GFP-IP 3 R1/ ⌬18A10 was generated by replacing the fragment consisting of EcoRI to XhoI site (aa 2216 -2749) in GFP-IP 3 R1-N with a PCR product consisting of aa 2216 -2737 in mouse IP 3 R1. GFP-18A10 was generated by subcloning a synthesized cDNA fragment consisting of aa 2738 -2749 in mouse IP 3 R1 into the site of EcoRI and BamHI in pEGFP-C3 (the terminological base for GFP-IP 3 R1/⌬18A10 and GFP-18A10 is that the last 14 residues of IP 3 R1 contain the specific recognition site by 18A10 antibody) (28). pBact-STneoB-C1 (used to express IP 3 R1) and EGFP-SERCA2a were as described previously (29,30). To construct the ER marker, DsRed2-KDEL, we fused the N-terminal 17 amino acids of calreticulin to the N terminus of DsRed2 (Clontech) and the ER target and retention signal, KDEL, to its C terminus by a two-step PCR as was described previously (31,32). The resulting PCR products were subcloned into pcDNA3.1/Zeoϩ (Invitrogen).
The mouse 4.1N cDNA was obtained by PCR using the primers of 5Ј-ATCGGAATTCATGACAACAGAGACAGGT-3Ј and 5Ј-ATCGTCTA-GATCAGGATTCCTGTGGCTT-3Ј (the underlined letters indicate the EcoRI and XbaI sites for cloning, respectively) corresponding to fulllength of mouse 4.1N sequence (accession number AF061283) using mouse cerebellum cDNA library as template and subcloned into the site of EcoRI and XbaI in pcDNA3.1/Zeoϩ (Invitrogen) (pcDNA3-4.1N). The pcDNA3-HA4.1N/FL was generated by replacing the cDNA fragment of EcoRI to EcoRV consisting of aa 1-298 of 4.1N in pcDNA3-4.1N with a PCR product that contains a corresponding sequence and is amplified using a 5Ј-primer containing the HA tag sequence and EcoRI site and a 3Ј-primer containing an EcoRV site. The pcDNA3-Venus-4.1N/FL was generated by inserting a PCR product amplified using the Venus cDNA (33) as template into the site of BamHI and EcoRI in pCDNA3-4.1N. The pcDNA3-HA4.1N/⌬CTD and the pcDNA3-Venus-4.1N/⌬CTD (aa 1-766) were generated by replacing a 4.1N fragment consisting of aa 663-879 in pcDNA3-HA4.1N/FL and in pcDNA3-Venus-4.1N/FL (ApaI-ApaI), respectively, with a PCR product consisting of aa 663-766 in 4.1N. The pcDNA3-HA4.1N/CTD (aa 767-879) was generated by subcloning a PCR product of 4.1N/CTD amplified using a 5Ј-primer containing the EcoRI site and the HA tag sequence and a 3Ј-primer containing XbaI site and stop codon into the site of EcoRI and XbaI in pcDNA3.1/Zeoϩ. pcDNA3-Venus-4.1N/CTD (aa 767-879) was generated by subcloning two PCR products of Venus (BamHI-EcoRI fragment) and 4.1N/CTD (EcoRI-XbaI fragment) into the sites of BamHI and XbaI in pcDNA3.1/Zeoϩ. GST-4.1N/CTD was generated by subcloning a PCR product of 4.1N/CTD into the site of BamHI and EcoRI in pGEX-KG (27). Truncated constructs corresponding to different lengths of 4.1N (see Fig. 4A) were subcloned into the sites of EcoRI and BamHI in pGAD424 (Clontech).
Yeast Two-hybrid Assay-The GAL4-based MATCHMAKER two-hybrid system II (Clontech) was used for the yeast two-hybrid assays. Plasmid vectors, pGBT9, and pGAD424, encoding the GAL4 DNAbinding domain and the GAL4 activating domain, respectively, were used to express hybrid proteins. To screen for proteins that interact with the C-terminal cytoplasmic tail of IP 3 R1 (IP 3 R1/CTT), a mixture of embryonic and adult human brain cDNA libraries (both from Clontech) in GAL4 activating domain vector pACT2 was screened using the Cterminal cytoplasmic tail of mouse IP 3 R type 1 cloned into GAL4 DNAbinding domain vector pGBT9 (see "Plasmid Construction" for details) as bait in PJ69 -4A yeast. Positive clones were tested further for specificity by co-transformation into yeast either with pGBT9-IP 3 R1/CTT or with pGBT9 alone. DNA from positive clones were isolated, and the GAL4 activating domain plasmids were recovered in E. coli strain HB101 and sequenced. For binding region mapping, yeast was cotransformed with plasmids carrying respective inserts fused to GAL4 DNA-binding domain or GAL4 activating domain and were assayed for nutritional selection of drop-out leucine, tryptophan, adenine, and histidine and for ␤-galactosidase activity on nitrocellulose filters as described in the Clontech manual.
Antibodies-For production of anti-4.1N antibodies, a nucleotide sequence corresponding to amino acid residues 588 -790 of mouse 4.1N, which has no homology to the other members of the 4.1 family, was subcloned into pRSET-C (Invitrogen). E. coli strain BL21 (DE3) was transformed with this plasmid, and His 6 -tagged protein was expressed and then purified over nickel columns. Japanese white rabbits and Wistar rats were immunized with the fusion protein. The rabbit antiserum was affinity-purified against GST-4.1N fragment (aa 588 -790) covalently coupled to CNBr-activated Sepharose 4B (Amersham Biosciences) according to standard protocols. The anti-IP 3 R1 rat monoclonal antibodies 18A10, 4C11, and 10A6 and the anti-IP 3 R1 mouse monoclonal antibody KM1112 were described previously (34 -36). H1L3 is a rabbit polyclonal anti-IP 3 R1 antibody prepared using a purified fusion protein corresponding to the residues 2463-2536 of mouse IP 3 R1 expressed in E. coli. 3 Anti-ZO1 antiserum (T8754) was a generous gift from Dr. S. Tsukita of Kyoto University. Anti-GFP monoclonal antibody and anti-HA polyclonal antibody were purchased from Medical & Biological Laboratory, Ltd., and anti-Na,K-ATPase ␣1 monoclonal antibody was from Upstate Biotechnology.
Cell Culture and Transfection-COS-7 and MDCK cells were maintained in Dulbecco's modified Eagle's medium (Nacalai Tesque) supplemented with 10% heat-inactivated fetal bovine serum. For subconfluent and confluent MDCK cells, MDCK cells were plated at 0.6 -1.0 ϫ 10 5 and 2.25 ϫ 10 5 cells, respectively, on 18 ϫ 18 mm poly-L-lysine-coated coverslips in a 35-mm culture dish and cultured for 1 and 5 days, respectively. The transfections were performed using LipofectAMINE 2000 reagent (Invitrogen) according to the manufacturer's protocol. Transfected COS-7 cells were harvested 1 day after transfection. Transfected MDCK cells on coverslips were fixed 3 days after transfection and then were processed for immunofluorescence staining with antibodies as described below.
Co-immunoprecipitation, Pull-down Binding Assay, and Immunoblotting-Lysates of subconfluent and confluent MDCK cells and of transformed E. coli, transfected COS-7 cells and whole brain of 8-weekold male ICR mice (Japan SLC, Inc. (Shizuoka Ken, Japan)) were prepared as previously described (37) in regard to the preparation the lysate of HEK293 cells and about preparation of P2 fraction from brain tissue, respectively. Co-immunoprecipitation from lysate of MDCK cells and lysate of whole mouse brain was performed as previously described (37). Pull-down binding assay was performed as a modification from the co-immunoprecipitation protocol. Briefly, lysates of E. coli and transfected COS-7 cells were first solubilized in 1% sodium deoxycholate at 36°C for 30 min, followed by adding 0.1 volume of 1% Triton X-100 in 50 mM Tris-Cl, pH 9.0, and the preparations were centrifuged for 30 min at 100,000 ϫ g. The supernatants were then used for in vitro binding assay. For each reaction, 1,000 g of protein of solubilized lysate of E. coli was incubated with 30 l of 1:1 slurry of glutathione-Sepharose 4B (Amersham Biosciences) at 4°C for 2 h and then washed with washing buffer (4 mM Hepes, 150 mM NaCl, 0.5% Triton X-100) three times, and then the spun down complex of glutathione-Sepharose 4B with fusion protein was used. At the same time, 500 g of protein of solubilized lysate of transfected COS-7 cells was incubated with 25 l of 1:1 slurry of glutathione-Sepharose 4B at 4°C for 1 h to clear any nonspecific binding to beads from the lysates. The cleared supernatant of the lysate was then added to the glutathione-Sepharose 4B-protein complex, and the mixture was incubated for 2 h or overnight at 4°C. The complex was then spun down and washed with washing buffer three times. The proteins were eluted by boiling in 1ϫ SDS-PAGE sample buffer for 3 min and were separated by SDS-PAGE. The proteins were transferred to a polyvinylidene difluoride membrane (Millipore, Bedford, MA), and the membranes were probed with anti-GFP Ab, anti-4.1N Ab, 18A10 Ab, 4C11 Ab, or anti-HA Ab.
Fluorescence and Confocal Microscopy-The cells on coverslips were fixed in freshly prepared 1.75% paraformaldehyde in cell culture medium for 15 min at room temperature. Then the cells were washed three times with PBS, permeabilized with 0.2% Triton X-100 in PBS for 5 min at room temperature, blocked with 2% normal goat serum in PBS for 1 h at room temperature, and washed three times with PBS. The cells were then incubated with primary antibodies (rabbit anti-4.1N Ab, H1L3 Ab (for endogenous IP 3 R1 in MDCK cells), KM1112 Ab (for exogenously expressed IP 3 R1), and anti-Na,K-ATPase ␣1 Ab; final concentration, 2 g/ml; anti-ZO-1 antiserum, 1:1; rat anti-4.1N antiserum (only used together with H1L3 Ab for dual immunostaining of endogenous 4.1N and IP 3 R1 in MDCK cells); 1:50) for 1 h at room temperature and washed three times as described above. They were then incubated with suitable Alexa Fluor 488, Alexa Fluor 594 secondary antibodies, or Alexa 594-phalloidin (Molecular Probes) at room temperature for 1 h and washed three times again as described above. Coverslips were mounted using Vectashield mounting medium (Vector Laboratories). Fluorescence images were taken using a confocal scanning microscope (FV-300, Olympus, Tokyo, Japan) attached to an inverted microscope (I ϫ 70; Olympus) with a 60ϫ objective.

Identification of 4.1N as a Binding Protein for the C-terminal
Cytoplasmic Tail of IP 3 R1-To understand the function of the C-terminal cytoplasmic tail of IP 3 R1, a yeast two-hybrid screening was performed to search for binding molecules of the C-terminal cytoplasmic tail of IP 3 R1. Using a fusion protein of the GAL4 DNA-binding domain with the C-terminal cytoplasmic tail of IP 3 R1 (IP 3 R1/CTT; Fig. 1A) as bait, approximately ϳ4.0 ϫ 10 6 yeast transformants were screened with a mixture of embryonic and adult human brain cDNA libraries fused to the GAL4 activation domain. By nutritional selection assay, 19 positive clones were obtained, and their cDNA inserts were sequenced. Among them, seven clones encoded sequences corresponding to various lengths of the C-terminal portion of 4.1N. The sequence of the shortest fragment corresponded to the C-terminal domain (amino acid residues 767-879) of mouse 4.1N ( Fig. 1B; hereafter, this fragment is referred to as 4.1N/ CTD). The interaction between 4.1N and IP 3 R1 was confirmed by a colony lift filter assay for ␤-galactosidase activity.
4.1N and IP 3 R1 Interact in Vitro and in Mouse Brain-To further verify the interaction between 4.1N and IP 3 R1, a pulldown assay was performed using a fusion protein of GST and 4.1N/CTD (GST-4.1N/CTD). GST-4.1N/CTD attached to glutathione-Sepharose beads was incubated with the lysates of COS-7 cells transiently expressing the GFP-tagged C-terminal cytoplasmic tail of IP 3 R1 (GFP-IP 3 R1/CTT) or GFP alone. After extensive washing, proteins bound to GST-4.1N/CTD were separated by SDS-PAGE and probed with anti-GFP antibody. As shown in Fig. 2A, GST-4.1N/CTD bound to GFP-IP 3 R1/CTT but not to GFP. GST alone did not bind to GFP-IP 3 R1/CTT. These results indicate that 4.1N/CTD also binds to the C-terminal cytoplasmic tail of IP 3 R1 in vitro.
To determine whether 4.1N binds to IP 3 R1 in vivo, lysates of whole mouse brain were subjected to co-immunoprecipitation  .1N/CTD, but GFP did not; GFP-IP 3 R1/CTT did not bind to GST. B and C, 4.1N binds to IP 3 R1 in vivo. The lysate of whole mouse brain was solubilized and immunoprecipitated with anti-IP 3 R1 and anti-4.1N antibodies, respectively. The input and immunoprecipitated proteins were separated by SDS-PAGE and probed with anti-4.1N and anti-IP 3 R1 antibodies. B, 4.1N was co-immunoprecipitated by three rat anti-IP 3 R1 antibodies, 4C11, 18A10, and 10A6 antibodies but not by normal rat IgG. C, IP 3 R1 was co-immunoprecipitated by a rabbit anti-4.1N antibody but not by normal rabbit IgG. WB, Western blot. with three distinct rat anti-IP 3 R1 monoclonal antibodies or a rabbit anti-4.1N polyclonal antibody. Co-precipitated proteins were separated by SDS-PAGE and probed with anti-4.1N and anti-IP 3 R1 antibodies. As shown in Fig. 2B, two bands that migrate at 135 and 100 kDa were detected with anti-4.1N antibody in the input sample and the pellets immunoprecipitated with anti-IP 3 R1 antibodies. These signals disappeared when the primary antibody was preincubated with the antigen protein (data not shown). The 135-kDa protein is the prototypical product of gene 4.1N. The 100-kDa protein is an isoform abundant in peripheral tissue, as reported by Walensky et al. (38). Conversely, when the reciprocal immunoprecipitation was performed using a rabbit anti-4.1N antibody, IP 3 R1 was coimmunoprecipitated (Fig. 2C). In the negative control experiments, neither normal rat IgG nor normal rabbit IgG immunoprecipitated 4.1N and IP 3 R1. Taken together, these observations indicate that 4.1N interacts with IP3R1 in vivo.
The Last 14 Amino Acids of IP 3 R1 Are Necessary and Sufficient for Binding to 4.1N-To determine the minimal sequence responsible for IP 3 R1 binding to 4.1N, serial deletions of the C-terminal cytoplasmic tail of IP 3 R1 (⌬1-⌬5) in pGBT9 were constructed, and their associations with 4.1N/CTD were examined using a yeast two-hybrid system. As shown in Fig. 3A, whereas the C-terminal cytoplasmic tail of IP 3 R1 interacted with 4.1N/CTD, none of the deletion mutants interacted with 4.1N/CTD. These results indicate that the last 14 amino acids of IP 3 R1 are necessary for binding to 4.1N.
A pull-down assay was then performed to confirm the yeast two-hybrid results. GST-4.1N/CTD fusion protein was incubated with the lysates of COS-7 cells transiently expressing GFP-IP 3 R1-N (GFP-tagged full-length IP 3 R1), GFP-IP3R1/ ⌬18A10 (GFP-tagged IP 3 R1 lacking the last 14 amino acids), or GFP-18A10 (the GFP-tagged last 14 amino acids of IP3R1). Expression of GFP-IP3R1-N, GFP-IP3R1/⌬18A10, and GFP-18A10 was confirmed by Western blotting assay using both anti-GFP antibody and 18A10 antibody. As shown in Fig. 3B, both GFP-IP 3 R1-N and GFP-18A10 were pulled down by GST-4.1N/CTD, but GFP-IP 3 R1/⌬18A10 was not. Neither GFP-IP 3 R1-N nor GFP-18A10 was pulled down by GST. Taken together, these results clearly indicate that the last 14 amino acids of IP 3 R1 are necessary and sufficient for binding to 4.1N. Additionally, because the last 14 amino acids of IP 3 R1, which contain the specific recognition site by 18A10 antibody of IP 3 R1 (28), share no homology with IP 3 R2 and IP 3 R3, these data support the result of a yeast two-hybrid assay that 4.1N/CTD does not bind to the C-terminal cytoplasmic tails of either IP 3 R2 or IP 3 R3 (data not shown) and suggest that 4.1N specifically binds to IP 3 R1.
The CTD Domain of 4.1N Is Necessary and Sufficient for Binding to IP 3 R1-The results of the yeast two-hybrid screening showed the shortest fragment of 4.1N required for binding to IP 3 R1 to correspond to amino acid sequence 767-879 of mouse 4.1N (Fig. 1B). To test whether other parts of 4.1N also bind to IP 3 R1 and to more precisely determine the IP 3 R1binding site of 4.1N, serial deletions of 4.1N in pGAD424 were constructed, and their associations with IP 3 R1/CTT were examined using a yeast two-hybrid system. As shown in Fig. 4A ization of IP 3 R1 in MDCK cells. Subconfluent and confluent MDCK cells were fixed, permeabilized, and immunostained with anti-IP 3 R1 and anti-4.1N antibodies or phalloidin. In subconfluent MDCK cells, IP 3 R1 existed in the cytoplasm (Fig.  5A). 4.1N was distributed in both the cytoplasm and the nucleus and partially co-localized with IP 3 R1 in the cytoplasm (Fig. 5B). Neither IP 3 R1 nor 4.1N existed at the region of the plasma membrane stained by phalloidin ( Fig. 5A and data not shown). In confluent MDCK cells, on the other hand, apart from slight immunofluorescence of both IP 3 R1 and 4.1N scattering in the cytoplasm, 4.1N and IP 3 R1 co-localized predominantly at the cell-cell junctional region in a punctate pattern (Fig. 5, C and D). These results indicate that both 4.1N and IP 3 R1 are translocated from the cytoplasm (and the nucleus) to the cell-cell junctional region when MDCK cells grow from subconfluence to confluence.
To determine whether 4.1N could also bind to IP 3 R1 in MDCK cells, the lysates of subconfluent and confluent MDCK cells were subjected to immunoprecipitation, respectively, with two rat polyclonal anti-IP 3 R1 antibodies. Co-precipitated proteins were separated by SDS-PAGE and probed with anti-4.1N and anti-IP 3 R1 antibodies. As shown in Fig. 5 (E and F) 1N and IP 3 R1 colocalized at the cell-cell junctional region in confluent MDCK cells (Fig. 5D). However, in contrast, IP 3 R is reportedly localized at the basolateral membrane domain (39), 4.1R, the prototypical homologue of 4.1N, is reportedly localized at the tight junctions in confluent MDCK cells (23). To accurately determine the localization of 4.1N and IP 3 R1, the localization of 4.1N and IP 3 R1 was compared with that of ZO-1, a marker protein at the tight junction in MDCK cells, and Na,K-ATPase, a marker protein at the basolateral membrane domain in MDCK cells. As shown in the three-dimensional illustrations (Fig. 6), neither 4.1N nor IP 3 R1 was localized at the tight junctions immunolabeled by anti-ZO-1 antibody, but both were localized at the basolateral membrane domain immunolabeled by anti-Na,K-ATPase ␣1 antibody. These results indicate that although 4.1R is localized at the tight junction in confluent

MDCK cells (23), 4.1N co-localizes with IP 3 R1 at the basolateral membrane domain in confluent MDCK cells. The 4.1N-binding Region of IP 3 R1 Is Responsible for Localization at the Basolateral Membrane Domain in Confluent
MDCK Cells-In both subconfluent and confluent MDCK cells, 4.1N associated with IP 3 R1 (Fig. 5). To examine the role of the 4.1N-IP 3 R1 interaction in the translocation of IP 3 R1 to the basolateral membrane domain, MDCK cells transiently expressing GFP-IP 3 R1-N, GFP-IP 3 R1/⌬18A10, GFP-18A10, or GFP alone were grown to confluence and observed using a confocal microscope. As shown in Fig. 7, when endogenous 4.1N was recruited to the basolateral membrane domain, GFP-IP 3 R1-N (Fig. 7A) and GFP-18A10 (Fig. 7C) were also recruited to the basolateral membrane domain. However, neither GFP-IP 3 R1/⌬18A10 (Fig. 7B) nor GFP alone (data not shown) was recruited to the basolateral membrane domain. Therefore, these results indicate that the 4.1N-binding region of IP 3 R1 is necessary and sufficient for the localization of IP 3 R1 at the basolateral membrane domain in confluent MDCK cells.  1N/FL (Fig. 8A) and Venus-4.1N/⌬CTD (Fig. 8B) were completely recruited to the basolateral membrane domain, and Venus-4.1N/CTD (Fig. 8C) and Venus alone (data not shown) were distributed in the cytoplasm and the nucleus. These results indicate the N-terminal portion, but not the C-terminal domain, to be responsible for the localization of 4.1N at the basolateral membrane domain.

4.1N/CTD Fragment Blocks IP 3 R1 Localization at the Basolateral Membrane Domain in Confluent MDCK
The CTD domain of 4.1N was necessary and sufficient for binding to IP 3 R1 (Fig. 4), and the 4.1N/CTD fragment was not localized at the basolateral domain in confluent MDCK cells (Fig. 8C). To determine whether the 4.1N/CTD fragment can block IP 3 R1 translocation to the basolateral domain in conflu-ent MDCK cells, MDCK cells were co-transfected with plasmids encoding Venus-4.1N/FL, Venus-4.1N/CTD, or Venus with IP 3 R1. As shown in Fig. 8 (D-F), when IP 3 R1 was coexpressed with Venus-4.1N/FL, both were recruited to the vicinity of the cell-cell junction in confluent MDCK cells. However, when IP 3 R1 was co-expressed with Venus-4.1N/CTD, both remained in the cytoplasm and the nucleus in confluent MDCK cells. This co-expression with Venus did not affect the distribution of IP 3 R1. Taken together, these results indicate that the interaction between IP 3 R1 and 4.1N is responsible for IP 3 R1 translocation and that both the N-terminal portion and the CTD domain of 4.1N are necessary for IP 3 R1 translocation to the basolateral membrane domain in confluent MDCK cells.
Other ER Marker Proteins Are Still Present in the Cytoplasm-IP 3 R1 is known to be predominantly localized on the ER (35,40). Consistently, in subconfluent MDCK cells, IP 3 R1 was found to be localized in the cytoplasm (Fig. 5B), whereas in confluent MDCK cells, IP 3 R1 was found to be localized at the basolateral membrane domain (Figs. 5, C and D, and 6, B2, and BЈ2). To detect the ER localization in confluent MDCK cells, immunostaining for other endogenous ER marker proteins was performed. As shown in Fig. 9A, calnexin was localized in the cytoplasm and was not localized at the basolateral membrane domain. In particular, there is no immunofluorescence of calnexin in the cell-cell junctional region immunolabeled by anti-Na,K-ATPase ␣1 antibody. Additionally, three other ER proteins, calreticulin, calsequenstrin, and sarcoplasmic/endoplasmic reticulum calcium-ATPase (SERCA2a), were also found to be localized in the cytoplasm (data not shown). These results indicate that when IP 3 R1 translocates to the basolateral membrane domain in confluent MDCK cells, other ER marker proteins are still localized in the cytoplasm. To further confirm these observations, MDCK cells were transiently transfected with EGFP-SERCA2a (a fusion protein of SERCA2a with EGFP) and DsRed-KDEL (a fusion protein of the ER target and retention signal peptide of calreticulin with DsRed2) and grown to confluence. Although the endogenous IP 3 R1 was predominantly concentrated at the basolateral membrane domain, the exogenously expressed EGFP-SERCA2a and DsRed2-KDEL were still distributed in the cytoplasm (Fig. 9, B and C). Taken together, these results suggest that although IP 3 R1 is localized at the basolateral membrane domain in confluent MDCK cells, the ER is not.

DISCUSSION
Bush et al. (39) have found IP 3 R to be localized at the basolateral membrane domain as well as in the cytoplasm in confluent MDCK cells by using antibodies against IP 3 R purified from the rat cerebellum. However, the molecular mechanism accounting for this localization has not been elucidated. In this study, we demonstrated an interaction between the C-terminal cytoplasmic tail of IP 3 R1 and the CTD domain of 4.1N using a yeast two-hybrid system, in vitro and in vivo binding assays. By employing specific antibodies against IP 3 R1 and against 4.1N, we found that IP 3 R1 and 4.1N associated in both subconfluent and confluent MDCK cells and were translocated from the cytoplasm (and the nucleus) to the basolateral membrane domain when MDCK cells grew from subconfluence to confluence. The localization of IP 3 R1 at the basolateral membrane domain was determined by its 4.1N-binding region and could be blocked by a fragment of 4.1N/CTD, which was necessary and sufficient for 4.1N binding to IP 3 R1 and could not be recruited to the basolateral membrane domain. Furthermore, we also found that although IP 3 R1 was localized at the basolateral membrane domain, several other endogenous or exogenously expressed ER marker proteins were still present in the cyto-plasm. Our data indicate IP 3 R1 localization at the basolateral membrane domain in confluent MDCK cells to be regulated by an interaction between the C-terminal cytoplasmic tail of IP 3 R1 and the CTD domain of 4.1N.
Organization of proteins into structurally and functionally distinct membrane domains is an essential characteristic of polarized epithelial cells. The details of the mechanism by which 4.1N and IP 3 R1 are restricted to the basolateral membrane domain in confluent MDCK cells have not been clarified. Herein, we identified several features of the translocation process of 4.1N and IP 3 R1 to the basolateral membrane domain. First, although 4.1R and 4.1N share a high degree homology (26), they have different subcellular localizations. Second, the N-terminal portion (aa 1-766) of 4.1N is responsible for the translocation of 4.1N to the basolateral membrane domain. Third, the interaction between 4.1N and IP 3 R1 is necessary for IP 3 R1 translocation to the basolateral membrane domain. Two 4.1R isoforms of 135 and 150 kDa reportedly co-localize with ZO-1, ZO-2, and occludin at tight junction in confluent MDCK cells (23). In contrast, 4.1N is not localized at tight junction (Fig. 6, A1 and AЈ1) but is localized at the basolateral membrane domain in confluent MDCK cells (Fig. 6, A2 and AЈ2). Although the 4.1R/CTD fragment bound to ZO-1 and ZO-2 in vitro, it could not be recruited to the tight junction in confluent MDCK cells (23). The 4.1N/CTD also could not be recruited to the basolateral membrane in confluent MDCK cells (Fig. 8C). The N-terminal portion of 4.1N, which could be completely recruited to the basolateral membrane (Fig. 8B), contains three unique domains and a highly conserved membrane-binding domain (Fig. 1B). The different subcellular localizations of 4.1R and 4.1N might be determined by their unique domains. The membrane-binding domain of 4.1R is known to interact with integral plasma membrane proteins such as band 3 (15,16), glycophorin C (17,18), and CD44 (19). Although the interactions between 4.1N and these proteins have not been verified, the overall high sequence homology between 4.1N and 4.1R leads us to speculate that the N-terminal portion of 4.1N (aa 1-766) may also have the ability to interact with one or more integral plasma membrane proteins. Both of these interactions, which might be induced by signals from MDCK cells at confluence, and the unique domains of 4.1N may allow recruitment of . The cells were grown to confluence, immunostained with anti-IP 3 R1 antibody, and analyzed by a confocal microscope. When IP 3 R1 was co-expressed with Venus-4.1N/FL, both were recruited to the vicinity of the cell-cell junction (D). When IP 3 R1 was co-expressed with Venus-4.1N/CTD, both were distributed in the cytoplasm and the nucleus (E). When IP 3 R1 was co-expressed with Venus, IP 3 R1 was recruited to the cell-cell junction; Venus was still distributed in the cytoplasm and the nucleus (F). Scale bar, 20 m.  IP 3 R1 is known to be predominantly localized on the ER (35,40). In subconfluent MDCK cells, IP 3 R1 was indeed also found in the cytoplasm (Fig. 5B). However, in confluent MDCK cells, IP 3 R1 showed a subcellular localization different from that of other ER marker proteins; although IP 3 R1 was predominantly concentrated at the basolateral membrane domain (Figs. 5, C and D, and 6, B2 and BЈ2), other ER marker proteins remained in the cytoplasm and were clearly absent at the cell-cell junction of the basolateral membrane domain (Fig. 9A and data not  shown). Again, we found that exogenously expressed EGFP-SERCA2a and DsRed2-KDEL were not translocated to the basolateral membrane domain in confluent MDCK cells (Fig. 9,  B and C). Considering that about 10 -20% of total IP 3 R was accessible to externally added biotin, primarily from the basolateral side in nonpermeabilized confluent MDCK cells (39), it is possible that a portion of IP 3 R1 is localized in the vicinity of the basolateral plasma membrane, a basolateral plasma membrane subdomain, or an associated membrane compartment and that a portion of IP 3 R1 exists as integral plasma membrane proteins. It is necessary to investigate the existence status of IP 3 R1 at the basolateral membrane domain to understand the physiological function of IP 3 R1 in confluent MDCK cells.
There is growing evidence suggesting that IP 3 R is localized to or on the plasma membrane. Immunolabeling studies in several cell lines have found a portion of the subcellular IP 3 R pool to be localized to the plasma membrane (41)(42)(43)(44). A number of subcellular fractionation studies have found that IP 3 R often appears in the plasma membrane fraction (45)(46)(47)(48). Tanimura et al. (49) clearly demonstrated all three isoforms of IP 3 R to be externally biotinylated in several cell lines. On the other hand, the CTD domain of 4.1N reportedly binds to the AMPA receptor GluR1 subunit (37) and to the D2 and D3 dopamine receptors (50), and these interactions appear to regulate the cell surface expression of these receptors. 4.1N, a membrane cytoskeletal protein, is expressed not only in neural tissue but also in non-neural tissues (38). The mechanism detected in MDCK cells by which 4.1N serves to regulate IP 3 R1 subcellular localization may have a general significance in other cell lines.
The possible role of IP 3 R1 at the basolateral membrane domain is at present uncertain. Localization of IP 3 R1 near the basolateral plasma membrane may allow IP 3 R1 to efficiently receive the inositol 1,4,5-trisphosphate signal and thereby rapidly induce a local Ca 2ϩ increase, which may modulate nearby actin cytoskeleton through Ca 2ϩ -sensitive actin-binding proteins and play critical roles in cell adhesive and cell polarity maintenance (39). If IP 3 R1 exists as an integral plasma membrane protein at the basolateral membrane domain in confluent MDCK cells, IP 3 R1 could conceivably function as a plasma membrane Ca 2ϩ channel with the IP 3 -binding domain and the 4.1N-binding C-terminal cytoplasmic tail facing the cytoplasm. Additionally, there are data supporting the hypothesis that IP 3 R residing on the plasma membrane can also function as a capacitative Ca 2ϩ entry channel (51)(52)(53).
Wu et al. (54) reported that disruption of the spectrin-protein 4.1 interaction resulted in a decreased thapsigargin-induced global cytosolic Ca 2ϩ response and in selective loss of the endothelial cell I SOC . Other reports have shown the dynamic activity of cytoskeletal actin to mediate the coupling process between Ca 2ϩ store depletion and Ca 2ϩ entry across the plasma membrane (55)(56)(57)(58). In view of the fact that 4.1N is a membrane-binding protein, a spectrin-actin-binding protein, and a component of the cytoskeleton, it would be worthwhile to investigate the functional contribution of the interaction be-tween 4.1N and IP 3 R1 and this interaction-based IP 3 R1 subcellular translocation in the process of Ca 2ϩ entry across the plasma membrane triggered by Ca 2ϩ store depletion.