Interactions with PDZ domain proteins PIST/GOPC and PDZK1 regulate intracellular sorting of the somatostatin receptor subtype 5.

By yeast two-hybrid screening we have identified interaction partners for the intracellular C-terminal tail of the human and rodent somatostatin receptor subtype 5 (SSTR5). Interactions with the PDZ domain-containing proteins PIST and PDZK1 are mediated by the PDZ ligand motif at the C terminus of the receptor; in case of the human and mouse (but not the rat) receptors, a slight sequence variation of this motif also allows for binding of the peroxisomal receptor PEX5. PIST is Golgi-associated and retains SSTR5 in the Golgi apparatus when coexpressed with the receptor; PDZK1 on the other hand associates with the SSTR5 at the plasma membrane. Endogenous SSTR5 in the neuroendocrine AtT-20 tumor cell line is colocalized with PIST in the Golgi apparatus. On a functional level, removal of the PDZ ligand motif of the receptor does not interfere with agonist-dependent internalization of the receptor or its targeting to a Golgi-associated compartment; however, recycling of the receptor to the plasma membrane after washout of the agonist is inhibited, suggesting that the PDZ-mediated interaction of SSTR5 is required for postendocytic sorting.

In recent years it has been appreciated that the signaling and the subcellular distribution of G-protein-coupled receptors may be influenced by proteins that interact with the intracellular regions of the receptors, most notably the C termini (1,2). Several interactions have been mapped to the membrane proximal domain of the C terminus, including the so-called helix 8, which encompasses the palmitoylation motifs found in most receptors of the rhodopsin-related type I family (3)(4)(5). These contacts especially involve proteins that act in the postendocytic, intracellular sorting of multiple receptors, i.e. the sorting nexin SNX1, N-ethylmaleimide sensitive factor and the G-protein-coupled receptor-associated sorting protein GASP. A second type of interactions involves the distal C termini of GPCRs, 4 which in many cases contain motifs for the interaction with PDZ domains. PDZ domain proteins frequently act as scaffold molecules because of their multiple protein interaction motifs (6). Thus this type of interaction has the potential to target a receptor to specific subcellular domains and into specific signaling complexes. Interaction of the ␤ 1 -adrenergic receptor with the third PDZ domain of PSD-95, for example, is believed to anchor this receptor at the postsynaptic density of excitatory synapses (7). The interaction of the parathyroid hormone receptor with Na ϩ /H ϩ exchanger regulatory factor/EBP-50 on the other hand physically links the receptor to phospholipase C␤, thereby shifting the second messenger response from adenylate cyclase activation to the hydrolysis of phospholipids (8).
We have recently initiated a search for intracellular interaction partners for the various members of the somatostatin receptor (SSTR) family (9,10). There are five SSTR subtypes in mammalian species, which form a rather homogeneous subfamily within the larger family of GPCRs, as they all preferentially couple to inhibitory, pertussis toxinsensitive G-proteins and exhibit common intracellular coupling patterns such as inhibition of adenylate cyclase (11) and activation of GIRK potassium channels (12). Thus, SSTR1 was detected on the axon terminals of hypothalamic neurons extending to the median eminence (13), SSTR4 was found on dendrites in hippocampal neurons (14), and SSTR3 is localized to neuronal cilia in the rodent brain (15). The SSTR5 is of particular pharmacological relevance as this receptor subtype regulates growth hormone secretion from the pituitary as well as insulin secretion from pancreatic ␤-cells (16,17). In the AtT20 cell line, a widely used model for the regulation of hormone release by somatostatin, SSTR2 is present at the cell surface whereas SSTR5 is largely confined to the trans-Golgi-network (TGN, Ref. 18).
Here we identify the PDZ domain proteins PIST and PDZK1 as intracellular binding partners of the SSTR5. Our data indicate that these proteins determine the intracellular sorting of the receptor; in particular, the PDZ-mediated interactions of SSTR5 appear to regulate targeting of the receptor from the Golgi/TGN to the plasma membrane.

MATERIALS AND METHODS
Yeast Two-hybrid Screen-A cDNA fragment coding for the C-terminal, intracellular domain of the human SSTR5 (amino acids 324 -364) was amplified by PCR and subcloned into the pAS2 vector (BD Biosciences). This bait plasmid was cotransformed together with a pretransformed human brain Matchmaker cDNA library in pACT-2 (BD Biosciences) into the CG1945 yeast strain. Clones growing in the absence of Leu, Trp, and His were picked, and their phenotypes were confirmed by ␤-galactosidase assay and retransformation. Plasmids from positive candidates were isolated and sequenced.
Fusion Proteins and Antibodies-Expression of GST-and His 6tagged fusion proteins and purification using glutathione-Sepharose (AP Biotech, Freiburg, Germany) or nickel-nitrilotriacetic acid-agarose (Qiagen, Hilden, Germany) were performed as described by the manufacturers. Guinea pig anti-PIST antiserum was generated using a His 6fusion protein of PIST (amino acids 120 -455; custom antiserum generation by Biogenes GmbH, Berlin, Germany). The rabbit anti-serum against the C terminus of the rat SSTR5 has been described before (19). Anti-FLAG and anti-myc were from Sigma, and anti-T7 was from Novagen.
Expression Constructs-The mouse SSTR5 cDNA was kindly provided by Wolfgang Meyerhof (Potsdam, Germany). Expression constructs for SSTR5 were generated by cloning PCR fragments coding for the cDNAs of human, rat, and mouse SSTR5 into a modified pcDNA3 vector (Invitrogen), which encodes an N-terminal T7-epitope tag (20). The full-length coding regions of PIST and PEX5 (Peroxin 5), as well as residues 15-478 of human PDZK1 were cloned into pEGFP-C1. Fulllength rat PDZK1 was obtained by reverse transcription-PCR from rat kidney and expressed as fusion with the myc-epitope tag in pcDNA3.1 (Invitrogen).
Pull-down Assays-GST fusions were prepared and left on glutathione beads (10 g of protein for individual experiment). These beads were incubated with 10 g of His 6 -tag fusion protein in 500 l of TBS-T (Tris-buffered saline containing 0.05% Tween) for 2 h at 4°C. Samples were precipitated by centrifugation, and supernatants were removed. After extensive washing of the beads with TBS-T, input, supernatant, and precipitates were analyzed by Western blotting using the anti-His 6antibody.
AtT20 cells grown on poly-L-lysine-coated coverslips were fixed with 4% paraformaldehyde at room temperature and rinsed thoroughly (2 ϫ phosphate-buffered saline, 2 ϫ TBS, pH 7.4). Subsequent to preincubation for 20 min in blocking buffer (3% normal goat serum, 0.1% Triton X-100 in TBS) cells were incubated overnight at 4°C with a mixture of the primary antibodies diluted in TBS containing 0.5% normal goat serum and 0.1% Triton X-100 (rabbit-anti-SSTR5 1:500, guinea pig anti-PIST 1:1000). Following three rinses in TBS, the cells were incubated for 1 h at room temperature with goat-anti-rabbit IgG coupled to Alexa 488 and goat-anti-guinea pig IgG conjugated to Alexa 594 (both from Molecular Probes) diluted 1:500 in TBS. Controls included omission of either one of the primary antibodies and incubation with both secondary antibodies to control for possible cross-reactivities. To assess the effect of stimulation on the interaction of PIST and SSTR5 in the AtT20 cell line, cells were equilibrated for 10 min at 37°C in Earle's buffer (130 mM NaCl, 5 mM KCl, 1.8 mM CaCl 2, 0.8 mM MgCl 2 , HEPES 20 mM, pH 7.4) supplemented with 1.0% bovine serum albumin and 0.1% glucose before stimulation for various periods of time (0 -30 min) with the same buffer containing 20 nM D-[Trp 8 ]SRIF prior to fixation and immunostaining as described. After rinsing, the coverslips were mounted using Aquapolymount (Polysciences) and imaged on a Zeiss LSM 510 confocal microscope. Single optical sections through the region of the nucleus were acquired. Image processing was carried out using the Zeiss LSM Image Browser software and Adobe Photoshop release 6.0.

RESULTS
By yeast two-hybrid screening in various cDNA libraries derived from human or rat tissues, we identified three putative interaction partners of the intracellular C-terminal tails of human and rodent SSTR5 (TABLE  ONE). PDZ domain protein interacting specifically with Tc10 (PIST, also known as GOPC or CAL; Refs. 21-23), a PDZ domain protein that is widely expressed in the brain and periphery, was found to be a potential interaction partner of the human as well as the rat SSTR5. Besides the PDZ domain, PIST consists of two coiled-coil motifs, the second of which is interrupted by a leucine zipper. A deletion analysis demonstrated that the interaction between PIST and SSTR5 is mediated by the PDZ domain (TABLE ONE). We also identified the PDZ domain protein from kidney (PDZK1), a protein with four PDZ domains, which is expressed in kidney, liver, gastrointestinal tract, and pancreas (24). PDZK1 also interacted with the C termini of both rat and human receptors. The third protein we identified, PEX5, also known as the peroxisomal receptor, does not contain a PDZ domain but contains a set of seven tetratricopeptide repeats. Via these repeats, PEX5 has been shown to identify C-terminal target sequences of proteins destined for sorting into peroxisomes. The consensus sequence for binding to PEX5 is S/C/ A-K/R/H-L/M-stop (25), which is similar to the consensus sequence for binding of C-terminal sequences to type I PDZ proteins (X-S/T-X-⌽- Identification of interaction partners for the SSTR5 C-terminal tail A, baits used for yeast two-hybrid screening. Numbers indicate sequence fragments used in the pAS2 bait vector. The C-terminal 12 residues from human, rat, and mouse are shown for comparison. B, summary of interaction partners of human and rat SSTR5. Targets obtained from the yeast two-hybrid system were tested by retransformation into yeast strain CG1945 using either the empt bait vector (pAS2) or bait vector containing the C-terminal sequences of rat or human SSTR5. In the case of PIST, deletion constructs were also generated, to map the interacting domain, as indicated. ϩ, ϩϩϩ, Ϫ indicate the strength of the interaction, as determined by the ability of transformed yeast cells to grow on selective media. (-Trp, -His, -Leu). nd, not determined. PDZ, PSD-95/discs large/ZO-1; L, leucine zipper; CC, coiled-coil; T, tetratricopeptide repeat. stop, where ⌽ is a large and hydrophobic residue; see Ref. 6 for review). The sequence comparison in TABLE ONE shows that only the human but not the rat SSTR5 fits to the consensus sequence for PEX5; also the mouse receptor C-terminal should bind PEX5, as the C-terminal residue is leucine in mouse, compared with isoleucine in rat SSTR5. This was confirmed by a pull-down assay with a synthetic peptide corresponding to the mouse SSTR5 C terminus (data not shown). Thus the mouse and human receptors, which conform to the consensus sequence for PEX5 binding, do bind PEX5, whereas the rat receptor does not correspond to the consensus and consequently lacks binding to PEX5. On the other hand, both rat and human receptors bind PIST and PDZK1 in a typical PDZ domain-mediated manner.
Specific interactions were in all cases confirmed in vitro by pull-down assays using purified fusion proteins. GST fusions of the C terminus of the rat and human SSTR5, as well as GST alone, were prepared and left on the glutathione matrix. Potential interacting partners were prepared as His 6 -tagged fusions and incubated with the GST matrices. After washing and elution by SDS sample buffer, all three proteins, PIST, PDZK1, and PEX5 (as well as an apparent degradation product of PIST) were detected in the eluate from the human SSTR5-GST matrix but not in the eluate from the control GST matrix (Fig. 1A). In addition, PDZK1 and PIST also bound to the rat SSTR5-GST matrix, whereas the PEX5 did not, further confirming the species-specific interaction of SSTR5 with this protein seen in the yeast two-hybrid system. In the case of PDZK1, we used this pull-down assay to further map the region of PDZK1 that interacts with the SSTR5. For this purpose, constructs coding for parts of PDZK1 were used to express the corresponding His 6tagged fusions and used in the pull-down assay with the human SSTR5-GST fusion. Only those fusion proteins that contained the third PDZ domain were bound by SSTR5-GST in this assay, strongly indicating that interaction between SSTR5 and PDZK1 is mediated by this domain (Fig. 1B).
We generated stable HEK-derived cell lines expressing the epitopetagged mouse SSTR5; these cells, or normal HEK cells, were subsequently transfected with PEX5-EGFP, PDZK1-, or PIST-EGFP (or EGFP as a control) expression vectors. Cells were lysed and, after immunoprecipitation of the receptor from cell lysates using the T7 antibody, precipitates were analyzed for the presence of the receptor and interacting proteins by Western blotting. In all cases, receptor and interacting proteins were detected strongly in precipitates, and precipitation of PEX5, PIST, and PDZK1 was dependent on the presence of the mSSTR5 in transfected cells (Fig. 2, A-C). In the case of PIST, we also analyzed the requirement for the PDZ ligand motif at the C terminus of the receptor by deleting the last eight residues in the expression construct (Fig. 2D). When PIST was coexpressed with this receptor construct in HEK cells, it could no longer be precipitated with the receptor, further confirming that the interaction between SSTR5 and PIST is mediated by a PDZ-type interaction involving the receptor C terminus.
Further analysis was devoted to the PDZ domain-containing proteins PDZK1 and PIST only, because their interactions appeared stronger than that of PEX5 ( Figs. 1 and 2) and were not limited to the human system. The subcellular distribution of expressed proteins was analyzed by immunocytochemistry. As expected, the receptor was found at the cell surface when expressed alone (Fig. 3A). PIST exhibited a Golgi/ TGN-like staining pattern, as has been described before (e.g. Ref. 26). PDZK1 was found at the plasma membrane (Fig. 3A); this is presumably because of the interaction of PDZK1 with membrane-associated protein of 17 kDa (MAP17; Ref. 27). When PIST was transfected into mSSTR5expressing cells, it was again localized to a Golgi-like structure; however, the distribution of the mSSTR5 changed dramatically, as most of the protein was now found in an intracellular compartment overlapping with the distribution of PIST (Fig. 3B). In contrast, the mSSTR5s receptor, which lacks the PDZ ligand motif, was not affected by PIST coexpression and was found at the cell surface (Fig. 3C). Transfection of a PDZK1 expression plasmid into the mSSTR5-expressing cell line changed neither the distribution of the receptor nor that of the PDZ protein, as both remained colocalized at the plasma membrane (Fig.  3D). Thus it appears that the presence of PDZ domain proteins influences the subcellular targeting of the receptor.
To determine whether these findings were relevant to cells that express endogenous SSTR5, we investigated the localization of SSTR5 and PIST in mouse pituitary AtT20 cells, which express SSTR5 and have been frequently used to study the effects of somatostatin on hormone secretion. Immunolocalization of endogenous SSTR5 using a polyclonal antibody raised against the receptor C terminus showed that SSTR5 is localized in an intracellular compartment, previously identified as the TGN by virtue of its immunostaining for syntaxin-6 (18). Costaining with the PIST antibody raised in guinea pig showed that a large proportion of the SSTR5 signal overlaps with the fluorescence for PIST, suggesting that both proteins reside and interact in the Golgi/TGN in AtT20 cells (Fig. 4, A-C). We also treated these cells with SST for var- After washing, fractions of the input, the supernatant (S) and the precipitated (P) material were analyzed by Western blotting using anti-His 6 antibody. B, mapping of the interacting domain in PDZK1. Pull-down assay was performed as in A using His 6 -tagged fusions of fragments of PDZK1, as indicated on the right. Note that the P12 fusion could be prepared only in minute amounts; however, all the material remained in the supernatant indicating that PDZ domain 3 is required for binding. ious periods of time, to see whether SST signaling would affect the distribution of the receptor. However, the localization of SSTR5 immunoreactivity did not change dramatically under these conditions, although some SSTR5-immunoreactive material was visible throughout the cytoplasm after 30 min of stimulation (Fig. 4, D and E). Expression of the truncated mutant mSSTR5s in these cells indicated that the intracellular localization of the endogenous receptor is due to the C-terminal PDZ ligand motif, as the truncated receptor was localized to the plasma membrane, with little or no intracellular staining (Fig. 4F).
Finally, to investigate the functionality of the interaction of the receptor with PDZ domain proteins, we examined the effect of PDZ domain deletion on SSTR5 receptor trafficking. It has been widely documented that SSTR5 and many other GPCRs are internalized after prolonged agonist treatment (28). When HEK cells expressing the mSSTR5 were treated with an agonist for 30 min, almost the entire immunoreactivity for the receptor was redistributed from the plasma membrane into an intracellular location, which could be identified as the Golgi compartment, as it costained with the Golgi-specific marker Golgin-97 (Fig. 5A). Washout of the agonist (including an acid wash procedure to disrupt agonist/receptor interaction) and subsequent incubation in normal medium led to recycling of the receptor to the plasma membrane. When this procedure was performed on the mSSTR5s mutant receptor, which lacks the PDZ binding motif, the receptor was similarly internalized into a Golgi-like structure, indicating that the PDZ interaction is not required for internalization. However, the truncated receptor completely lacked the ability to be recycled to the plasma membrane (Fig. 5B).
Stimulation of cells coexpressing PIST led to a similar picture for the wild type mSSTR5, as it was found largely in the Golgi (Fig. 5C, as seen before in the absence of agonist, Fig. 3). In the presence of PDZK1, on the other hand, some internalization of the receptor was observed, but this remained rather incomplete. In cells that showed internalization, PDZK1 did comigrate with the receptor into an intracellular compartment (Fig. 5c). We also analyzed the influence of PIST coexpression on receptor recycling; here it became evident that the overexpressed PIST prevented a complete recycling of the mSSTR5. Thus PIST retains the receptor in the Golgi apparatus independently of its agonist-treated status.

DISCUSSION
All mammalian somatostatin receptor subtypes, with the exception of a splice variant of SSTR2 (29), contain a potential C-terminal PDZ ligand motif; even the closest known homologs in invertebrates, the Drostar receptors for allatostatin C from Drosophila melanogaster (30), do contain such a motif. So far our studies have indicated that all SSTRs differentially interact with C-terminal binding partners, and these interactions may link the receptors to specific subcellular locations and signaling pathways (9,10). The PDZ domain motif of SSTR5 from mouse and man is unique as it also conforms to the consensus binding sequence for the peroxisomal receptor PEX5, and this protein has consequently been pulled out in our yeast two-hybrid screen. In fact, analysis of SSTR5 sequences on a peroxisomal targeting data base (see Ref. 31) suggests that the human and mouse receptors should be targeted to peroxysomes because of this motif. Although we did detect a physical interaction between SSTR5 and PEX5 in transfected HEK cells, we have not obtained evidence for peroxisomal targeting of SSTR5 in these cells but found the receptor at the plasma membrane instead (data not shown). This suggests that additional factors besides the C-terminal motif are required to target a membrane protein to peroxisomes. On the other hand it may mean that PEX5 participates in so far unknown signaling events. Nevertheless, we have focused our work mainly on the PDZ-type interactions of the SSTR5, which could not only be investigated for human and mouse SSTR5 but also for the rat subtype.
Our data demonstrate a quite robust interaction between SSTR5 and the PDZ domains of PIST (which occurs at the level of the Golgi apparatus) and of PDZK1 (which occurs at the plasma membrane). The SSTR5 is in this respect most similar to two regulators of chloride conductance, chloride channel 3b and the cystic fibrosis transmembrane regulator (CFTR), both of which also interact with PIST and PDZK1 through their C-terminal PDZ ligand motifs (26). However, CFTR does also interact with the PDZ domain of Na ϩ /H ϩ exchanger regulatory factor (32, 33), whereas the human SSTR5 does not (5). Chloride channel 3b is found in the Golgi similar to PIST, whereas the CFTR resides at the apical membrane of epithelial cells (e.g. Ref. 26). Unfortunately, the possible role of the PDZ ligand motif is unclear in the case of the CFTR due to conflicting evidence in the literature. The C-terminal interaction may be of pathological importance, as two studies suggested that the lack of the PDZ ligand motif leads to intracellular accumulation and mistargeting of the CFTR, similar to that observed in cystic fibrosis patients (34,35). This was however contradicted by other studies that suggested a role of the PDZ interaction in the regulation of chloride transport activity instead (36,37). Cheng et al. (38) demonstrated that overexpressed PIST targets the CFTR for lysosomal degradation.   8 ]SST for the times indicated before fixation. In F, the T7-mSSTR5s construct was expressed, and the receptor was stained using the T7 antibody (red fluorescence), whereas PIST was again stained as above. Receptor distribution was analyzed by staining using monoclonal anti-T7, directed against the N-terminal T7 epitope, followed by Cy3-anti-mouse. c, cells coexpressing the wild type receptor and PIST (upper, lower panel) or PDZK1 (middle panel) were treated with SST28 as above. In the lower panel, the agonist was washed out, followed by a 30-min recycling period. The localization of receptor and interacting proteins was in each case determined by staining with the appropriate antibodies, or the EGFP autofluorescence in the case of PIST, as described in the legend to Fig. 3.
Here we show that the PDZ ligand motif in the C terminus of the SSTR5 determines its intracellular location; the C-terminally truncated receptor is localized at the plasma membrane of AtT-20 cells, unlike its wild type counterpart, which is found mostly in the Golgi apparatus. Golgi localization is specific to the SSTR5, whereas the SSTR2 which is expressed in the same cells (and which does not bind PIST), is found at the plasma membrane under resting conditions (18).
The role of the C-terminal motif in postendocytic sorting is similar to that of the PIST/PDZK1 binding motif in the CFTR, as lack of this motif leads to a defect in receptor recycling (35). Although for CFTR the additional interaction of the C terminus with the Na ϩ /H ϩ exchanger regulatory factor PDZ domain may contribute to recycling to the plasma membrane (as has been observed also for the ␤ 2 -adrenergic receptor, Ref. 39), we have no evidence for additional proteins (besides PDZK1 and PIST) binding to the SSTR5 C terminus, suggesting that the relative contributions of PIST and PDZK1 may determine the postendocytotic fate of the receptor. Interestingly, Gage et al. (40) have recently provided evidence that type I PDZ ligands are sufficient to promote rapid recycling of the ␤ 2 -adrenergic receptor independent of their ability to bind to Na ϩ /H ϩ exchanger regulatory factor. As we could not detect endogenous PDZK1 in HEK cells (data not shown), we currently favor a model where the receptor is sorted to the Golgi compartment after internalization independently of its PDZ ligand motif. PIST recognizes the endocytosed receptor in the Golgi/TGN and directs its sorting back to the plasma membrane. Via its coiled-coil domain, PIST associates with syntaxin-6, a SNARE protein that is involved in trafficking between endosomes and the TGN and might therefore assist in the proper sorting of membrane proteins such as the SSTR5 (22,41,42). In this respect, PIST might be a general sorting molecule for a wide variety of transmembrane proteins ranging from GPCRs (SSTR5 and the ␤ 1 -adrenergic receptor; Ref. 43) to cell adhesion molecules (44) to ion channels (chloride channel 3b) or their associated proteins such as stargazin (45). PDZK1 also recognizes multiple types of membrane proteins. This is in contrast to proteins such as GASP or the ␤-arrestins, which so far appear to be specifically devoted to the postendocytic sorting of GPCRs only.
Rapid recycling of the SSTR5 and targeting to the membrane from intracellular stores has been described previously (46). In addition, it has been shown that stimulation of cell surface receptors promotes targeting of reserve receptors from the Golgi/TGN to the plasma membrane (46). It is tempting to speculate that intracellular sorting events governed by the interactions with PIST and/or PDZK1 may contribute to delivery of receptors to the cell surface "on demand." One such signal might be transduced by the small GTPase Tc10, which, in its activated form, leads to translocation of PIST to the plasma membrane (47).
We tried to assess whether somatostatin itself might be the signal to bring the SSTR5 to the cell surface in AtT-20 cells, either by activation of SSTR2 or the residual plasma membrane associated SSTR5. Although we did observe only little translocation to the cell surface, further work will be needed to address this, as the presence of SST will immediately lead also to further internalization. Nevertheless we expect that controlled delivery of receptor to the cell surface might have a strong influence on the ability of somatostatin to regulate growth hormone secretion from the pituitary or insulin secretion from the pancreas.