Novel proteins that interact with the COOH-terminal cytosolic routing determinants of an integral membrane peptide-processing enzyme.

The steady state distribution of membrane forms of peptidylglycine α-amidating monooxygenase (PAM) in the secretory pathway of neurons and endocrine cells depends on signals in its cytosolic COOH-terminal domain (CD). Mutagenesis studies yielded catalytically active PAM proteins that are not properly localized or internalized. Employing the yeast two-hybrid system, we isolated two distinct cDNAs whose protein products showed a strong interaction with the CD of PAM. The interaction of these novel PAM COOH-terminal interactor proteins (P-CIPs) did not occur with a misrouted CD mutant as bait in the yeast system. Both proteins, P-CIP2 and P-CIP10, were expressed as fusion proteins that interacted in vitro with solubilized integral membrane PAM. P-CIP2 was homologous to several serine/threonine and dual specificity protein kinases, while P-CIP10 contained spectrin-like repeats. Endogenous P-CIP2 was localized to the Golgi region of AtT-20 corticotrope tumor cells, and expression of integral membrane PAM disrupted the distribution of endogenous P-CIP2. Both P-CIP2 and P-CIP10 mRNAs were found to be expressed in rat brain neurons also expressing PAM proteins. P-CIP2 and P-CIP10 may be members of a family of cytosolic proteins involved in the routing of membrane proteins that function in the regulated secretory pathway.

In neurons and endocrine cells, biologically active peptides are stored in large dense core vesicles (LDCVs), 1 which undergo regulated release (1)(2)(3). After exocytosis, the membrane proteins of LDCVs are retrieved from the plasma membrane and targeted to the trans Golgi network (TGN) or lysosomes (1)(2)(3). Little is known about the mechanisms underlying the sorting of membrane proteins that function in LDCVs. Peptidylglycine ␣-amidating monooxygenase (PAM; EC 1.14.17.3), which catalyzes the two reactions involved in the COOH-terminal ␣-amidation of bioactive peptides (4 -6), occurs in soluble and integral membrane forms (Fig. 1) and serves as a tool to compare the routing of soluble and membrane proteins to LDCVs (4 -6). Although the luminal domains of PAM catalyze the amidation reaction, the facts that the COOH-terminal do-mains (CD) of mammalian and Xenopus laevis PAM are highly conserved and that the nervous system expresses almost exclusively the integral membrane forms of PAM indicate that tethering of PAM to the membrane is functionally important (4 -7).
In AtT-20 corticotrope tumor cells, integral membrane forms of PAM have access to immature secretory granules and accumulate in tubuloreticular structures in the distal part of the TGN (8 -11). Soluble PAM proteins accumulate in LDCVs in the peripheral processes of AtT-20 cells. At steady state, only a small percentage of the membrane PAM is on the surface or in endosomes and can be detected by surface enzyme assays and binding and internalization of ectodomain antibodies (8 -11) (Fig. 1).
The PAM CD contains multiple routing signals recognized at different subcellular locations. Truncation of half of the CD (Fig. 1) leads to diminished storage of PAM in LDCVs, plasma membrane localization of active enzyme, and failure of internalization (9 -12). Transfer of the cytosolic and transmembrane domains of PAM to the luminal domain of the interleukin 2 receptor ␣ chain relocated this protein from the plasma membrane to the TGN and supported internalization of bound antibody (10). Further mutagenesis studies identified residues 928 -945 in the PAM CD as critically important and eliminated a role for amino acids distal to residue 957 (10). Mutation of Tyr 936 to Ala (Y936A) disrupted internalization of membrane PAM without greatly affecting cleavage in LDCVs (10). The CD of membrane PAM is phosphorylated on Ser/Thr residues in cells; mutation of a protein kinase C site at Ser 937 to Ala (S937A) resulted in mistargeting of the internalized PAM⅐PAM antibody complex (11). The importance of the cytosolic domains of many other membrane proteins in their retention and/or targeting to the appropriate subcellular compartment is well established (13)(14)(15)(16)(17)(18)(19)(20)(21).

MATERIALS AND METHODS
DNA Techniques-The methods of Sambrook et al. (25) were used; sequence analysis employed the Sequenase kit (U.S. Biochemical Corp.) or was performed by the Johns Hopkins Genetics Core. The polymerase chain reaction was used to introduce restriction sites for SalI and EcoRI at the 5Ј and 3Ј ends of amplified cDNA fragments for various versions of PAM CD, and the inserts generated by SalI/EcoRI digestion were cloned into pPC97 (22,23). This produced plasmids pPC97-CDT (CD truncated), which encodes PAM-1-(891-961), and pPC97-dCDT (encoding PAM-1-(891-961) with residues 928 -945 deleted) ( Fig. 2A). For construction of the expression vectors pGEX-CIP2 and pGEX-CIP10, P-CIP2 and P-CIP10 cDNAs were excised from the isolated pPC86 library vectors by SalI/NotI digestion and ligated to pGEX5X.2 (Pharmacia Biotech Inc.) digested with the same restriction enzymes.
In Vitro Binding Assay-E. coli strain BL21 transformed with pGEX5X.2 (glutathione S-transferase; GST) (Pharmacia), pGEX-CIP2, and pGEX-CIP10 were grown in 2 ϫ YT medium and induced with isopropyl ␤-thiogalactopyranoside. GST and GST⅐P-CIP2 fusion proteins were immobilized on glutathione-Sepharose (Pharmacia) beads. Since the GST⅐P-CIP10 fusion protein was insoluble, the pellet was collected after cell lysis and solubilized in phosphate-buffered saline containing 6 M urea. The 30,000 ϫ g supernatant fraction was diluted to 1 M urea and applied to glutathione-Sepharose beads. All beads were washed with 500 mM NaCl and equilibrated with binding buffer (10 mM Tris-HCl, 150 mM potassium acetate, 1 mM MgCl 2 , and 0.5 mM CaCl 2 , pH 7.5). AtT-20 cells stably expressing PAM-2 were extracted with 20 mM sodium TES, pH 7.5, 10 mM mannitol, 1% Triton X-100 containing a protease inhibitor mixture (8). The extract was then diluted 10 times with binding buffer containing the protease inhibitor mixture. A 50-g aliquot of the diluted AtT-20 cell extract was mixed with 5 g of GST, GST⅐P-CIP2, and GST⅐P-CIP10 immobilized to glutathione-agarose and incubated at 4°C for 2 h.
Cell Culture and Immunostaining-AtT-20 cells (8) were fixed with ice cold methanol, rinsed with saline, blocked, and incubated with primary antibodies for P-CIP2 and PAM overnight at 4°C. Immunostaining was visualized using fluorescein isothiocyanate-conjugated goat anti-mouse immunoglobulin and Cy 3 -conjugated donkey-anti-rabbit immunoglobulin (Jackson ImmunoResearch) (8). Monoclonal antibodies 6E6 to the PAM CD and 18E5 to PHM were raised as described (10). Polyclonal antibody JH1764 was specific for PHM (9). Rabbit antibodies were raised to a synthetic polypeptide from the COOH terminus of the predicted P-CIP2 protein (PKENPGRGQVFVEYANADG) linked to keyhole limpet hemocyanin with glutaraldehyde (9) (Hazleton HRP, Inc., Denver, PA).
In Situ Hybridization and Immunocytochemistry-Male Sprague-Dawley rats were anesthetized with 50 mg/kg pentobarbital sodium and perfused with ice-cold 4% paraformaldehyde in phosphate buffer. Brains were stored (4°C) for 1 day in fixative and in 25% sucrose for at least 7 days (26). A 551-base pair NotI/BamHI fragment of P-CIP2 cDNA and a 603-base pair BamHI fragment of P-CIP10 cDNA were subcloned in pBluescript and used to synthesize sense and antisense RNA probes. These probes visualized discrete bands on Northern blots. For immunocytochemistry, sections subjected to in situ hybridization were then washed, blocked, rinsed with 3% H 2 O 2 , and incubated with antiserum to PAM (Ab 571) using an avidin-biotin kit (Vector, Burlingame, CA) (26).

Identification of Interactor Proteins in the Yeast Two-hybrid
System-To identify cytosolic factor(s) involved in the routing of membrane PAM, we used the two-hybrid system, a yeastbased genetic assay that uses the modular nature of transcriptional activators to detect protein-protein interactions (22)(23)(24). The PAM CD (CDT) was used as bait for the two-hybrid screen of a rat hippocampal cDNA library ( Fig. 2A). Truncation at amino acid residue 961 eliminated the highly charged PEST segment located near the COOH terminus of PAM-1, which might cause nonspecific ionic interactions or rapid degradation (10,11,27). A screen of 1.2 ϫ 10 6 double transformants on histidine-deficient medium yielded 20 primary clones; four displayed strong ␤-galactosidase activity. Plasmids were rescued from each of these yeast colonies and named P-CIP1, -2, -10, and -19. None of the four interactor clones supported growth of yeast in deficient medium in the absence of the CDT plasmid or in the presence of the nonfunctional, internally deleted dCDT plasmid (Fig. 2B). Thus, the pattern of growth of the double transformants suggests that the interaction between the protein products of the isolated cDNAs and the PAM CD is similar to the physiological interactions that determine PAM routing (Fig. 2B).
Predicted Structures of Interactor Proteins P-CIP2 and P-CIP10 -The sequences of the proteins encoded by P-CIP2 and P-CIP10 are shown in Fig. 3, A and B, respectively. The sequence of P-CIP19 was contained within P-CIP10 (Fig. 3B). When P-CIP1, P-CIP2, and P-CIP10 were searched for homology to the Non-redundant Protein Data base maintained at the National Center for Biotechnology Information using the BLASTP algorithm, they were found to encode novel proteins. The 333-amino acid sequence predicted for P-CIP1 did not exhibit significant homology to any sequences in the data bases, and studies on it will be reported separately.
The predicted P-CIP2 protein that follows the GAL4 transactivator domain contains 392 amino acids (M r ϭ 43,567) followed by an in-frame stop codon and a 264-nucleotide 3Ј-untranslated region. According to BLASTP analysis, the P-CIP2 protein (Fig. 3A) shows significant homology to members of the serine/threonine-protein kinase family (e.g. the Leishmania mexicana Cdc2-related protein kinase Crk1; accession number 585007) and dual specificity Ser/Thr/Tyr-protein kinases (e.g. human phosphotyrosine picked threonine kinase, PYT; accession number 312816), suggesting that P-CIP2 is a protein kinase. The catalytic core domains of protein kinases typically begin with a nucleotide binding loop (GXGXXGXV) followed by an invariant Lys, also involved in nucleotide binding (28). The third Gly in the nucleotide loop is sometimes substituted with Ser or Ala; P-CIP2 contains Ala at this position (Ala 8 ). P-CIP2 also contains the highly conserved protein kinase catalytic domain triplet DFG (residues 150 -152), which is involved in Mg 2ϩ binding (29), in addition to the consensus triplet APE (residues 172-174) thought to be involved in catalysis. The sequence DLXXXN is often an indicator of protein kinase substrate specificity (28). Residues 131-137 (DLKPRN) of P-CIP2 suggest that it is a serine/threonine-specific protein kinase. P-CIP2 shares sequence identity with KIS, a protein fragment encoded by a partial cDNA identified by virtue of interaction of its product protein with stathmin in a yeast two-hybrid screen (30).
The protein encoded by the P-CIP10 cDNA is 677 amino acids long (M r ϭ 78,435) and lacks an in-frame stop codon, suggesting that full-length P-CIP10 is extended in both the COOH-terminal and NH 2 -terminal directions (Fig. 3B). Secondary structure predictions for P-CIP10 indicate that it is almost entirely ␣-helical, and BLAST analysis revealed that the P-CIP10 protein has the greatest sequence homology with human ␣-fodrin, chick ␣-spectrin, and rat ␣-II spectrin. These proteins all contain multiple repeats of a 100 -120-amino acid domain thought to fold into a three-helix bundle (31)(32)(33)(34)(35). Spectrin and fodrin are involved in cross-linking actin filaments to form the meshwork underlying the plasma membrane (3, 31-33). P-CIP10 includes five full spectrin-like repeats varying in length from 105 to 131 amino acids with partial spectrin-like domains at the NH 2 -and COOH-terminal ends. P-CIP19 includes two complete spectrin-like domains (Fig. 3B); secondary structure predictions suggest that the B and C helices of the second complete spectrin-like repeat in P-CIP19 are interrupted by a longer than average nonhelical loop.
Interactor Proteins P-CIP2 and P-CIP10 Bind to the CD of Integral Membrane PAM-The interaction between membrane PAM proteins and P-CIP2 and P-CIP10 proteins was confirmed in vitro using GST fusion proteins. From the SDS-polyacrylamide gel electrophoresis patterns of the expressed GST fusion proteins (not shown), the P-CIP2 and P-CIP10 portions of the fusion proteins have apparent molecular masses of 44 and 84 kDa, respectively, in agreement with the sizes predicted from the amino acid sequence data in Fig. 3. Glutathione-Sepharose beads with immobilized GST, GST⅐P-CIP2, and GST⅐P-CIP10 were prepared, and integral membrane PAM-2 was applied to each resin. As demonstrated in Fig. 4, the 105-kDa PAM-2 protein bound to the GST⅐P-CIP2 and GST⅐P-CIP10 resins but not to the GST resin.
Integral Membrane PAM Interacts with P-CIP2 in Cultured Cells-In order to determine whether integral membrane PAM and P-CIP2 interact in a cellular environment, we generated an antiserum to a peptide from the COOH-terminal region of P-CIP2 (Fig. 3A). This antiserum visualized a single major endogenous cytosolic protein of 47 kDa when used for Western blot analysis of AtT-20 corticotrope tumor cell extracts (data not shown). Nontransfected AtT-20 cells were fixed and immunostained simultaneously with polyclonal antisera specific for P-CIP2 and monoclonal antibodies directed against PAM. Staining for endogenous P-CIP2 was heaviest in the perinuclear region near the Golgi/TGN, with less intense staining apparent throughout the cytosol (Fig. 5B). Nontransfected AtT-20 cells have a low level of endogenous PAM and exhibit no significant immunostaining for PAM (Fig. 5A) (8).
We then examined AtT-20 cell lines stably expressing PAM-1 to determine whether expression of exogenous integral membrane PAM altered the localization of endogenous P-CIP2. As expected, staining for PAM in AtT-20 cells expressing integral membrane PAM-1 was concentrated primarily in the perinuclear TGN region with staining at the tips of the cellular processes apparent in more heavily stained cells (Fig. 5C) (8,9). Expression of integral membrane PAM-1 in AtT-20 cells resulted in the redistribution of the endogenous P-CIP2 (Fig. 5D). Immunostaining for P-CIP2 in AtT-20 cells expressing PAM-1 was no longer concentrated in the perinuclear TGN region but was redistributed throughout the cytosol. Thus, expression of integral membrane PAM-1 in AtT-20 cells disrupted the steady state distribution of P-CIP2.
Interactor Proteins P-CIP2 and P-CIP10 Are Found in Neurons That Express PAM-A meaningful interaction between PAM and either P-CIP2 or P-CIP10 can occur only if the proteins are expressed in the same cells, so the localizations of PAM protein and transcripts encoding P-CIP2 and P-CIP10 in rat brain were compared using dual in situ hybridization and immunocytochemistry. Sections of brain were first analyzed by in situ hybridization using radiolabeled sense and antisense RNA probes synthesized from fragments of the P-CIP2 and P-CIP10 cDNAs. The sections were then probed with polyclonal PAM antibodies. As shown in Fig. 6, neurons that showed high levels of hybridization with P-CIP2 and P-CIP10 riboprobes could also be immunostained for PAM protein; the control sense riboprobes for P-CIP2 and P-CIP10 yielded no signal (not shown). Many nearby neurons that did not contain PAM protein also lacked the transcripts encoding P-CIP2 and P-CIP10. PAM protein is found in neurons expressing P-CIP2 and P-CIP10 transcripts, making the postulated interactions possible.

DISCUSSION
The yeast two-hybrid system was used to identify proteins that interact with the cytosolic domain of PAM and thus might be involved in the trafficking of membrane PAM in neurons and endocrine cells. Four clones exhibited the proper selection characteristics, suggesting that they were capable of physiologically relevant interactions with the cytosolic CD of membrane PAM.
P-CIP2 is a putative serine/threonine or dual specificity protein kinase containing all of the conserved residues characteristic of the catalytic core of a serine/threonine-protein kinase (Fig. 3A) (28,29). The mass of the AtT-20 protein identified on Western blots by antisera to P-CIP2 (ϳ47 kDa) is only a few kilodaltons greater than the mass predicted based on the partial cDNA identified in the yeast two-hybrid screen (ϳ44 kDa). A short NH 2 -terminal extension in full-length P-CIP2 presumably accounts for the observed size difference. The COOHterminal region of P-CIP2 is highly homologous to the third ribonucleoprotein consensus sequence domain of pre-mRNA splicing factor U2AF 65 (36). Ribonucleoprotein consensus domains are involved in RNA recognition and binding; their significance to the function of P-CIP2 is unknown. P-CIP2 is the rat homologue of a mouse protein designated KIS, differing at only 2 of the 201 amino acid residues predicted for KIS (30). KIS was identified in a yeast two-hybrid screen for proteins that interact with stathmin, a ubiquitously expressed 19-kDa cytosolic protein that is phosphorylated in response to a variety of secretagogues (30). Although KIS was recognized as a putative serine/threonine-protein kinase, the partial sequence reported lacks the NH 2 -terminal part of the protein kinase catalytic core, and no direct interaction of KIS with stathmin was demonstrated (28 -30).
The interaction of a protein kinase with the CD of membrane PAM could play an important role in the routing and steady state distribution of the protein. In cultured cells, the CD of membrane PAM is phosphorylated on Ser and Thr residues by an unknown protein kinase (11). Recombinant PAM CD is phosphorylated on Ser 932 and Ser 937 by protein kinase C (11). Expression of mutant PAM-1/S937A in AtT-20 cells resulted in its misrouting following internalization from the plasma mem- brane; PAM-1/S937A appeared to be routed to lysosomes instead of returning to the TGN (11). Thus, phosphorylation of the CD plays a role in the trafficking of membrane PAM, and P-CIP2 may be directly or indirectly involved in this process. P-CIP10 is most homologous to human ␣-fodrin and chick ␣-spectrin, and the amino acid sequence of P-CIP10 can all be arranged into spectrin-like repeats (Fig. 3A) (32,34,35). More distantly related members of this family include ␤-spectrin, Dbs (a protein with a guanine nucleotide exchange factor domain and a pleckstrin homology domain), and dystrophin. PAM is clearly expressed in neurons lacking P-CIP10, and P-CIP10 may represent only one of a family of related proteins that interact with the PAM CD. In erythrocytes and nonerythroid cells, spectrin and fodrin provide structural support to the plasma membrane, and spectrin binds to specific soluble and integral membrane proteins to form an extensive cytoskeletal meshwork that is tightly associated with the plasma membrane (32,33,35). An isoform of ␤-spectrin is associated with Golgi membranes in Madin-Darby canine kidney cells and dissociates from the Golgi membranes at the time when the Golgi complex is extensively fragmented and dispersed (33), suggesting a role for ␤-spectrin in membrane protein retention. Spectrin homologues like P-CIP10 may play an important role in the routing of membrane PAM proteins in neuroendocrine cells.
The redistribution of endogenous P-CIP2 in AtT-20 cells stably expressing integral membrane PAM was striking (Fig.  5). Expression of exogenous integral membrane PAM disrupted the perinuclear localization of endogenous P-CIP2, causing the protein to adopt a diffuse, cytosolic localization. P-CIP2 is not an integral membrane protein, and expression of membrane PAM must perturb the normal steady state equilibrium interactions involved in localizing P-CIP2. P-CIP2 might phosphorylate the CD of PAM or might bind to the CD of PAM and phosphorylate another protein involved in routing. Localization of endogenous P-CIP2 to the Golgi region of AtT-20 cells suggests that it plays a role in the cycling of membrane PAM into or out of this compartment rather than in internalization from the plasma membrane. P-CIP2 and P-CIP10 interact with integral membrane PAM in test tube binding studies and with the PAM CD in the yeast two-hybrid system. The co-localization of P-CIP2 and P-CIP10 transcripts in the same nerve cell bodies expressing PAM proteins provides support for the physiologic relevance of these interactions (Fig. 6). These observations lead to the suggestion that full-length P-CIP2 and P-CIP10 are involved in the routing of membrane PAM proteins. These proteins would link PAM, a transmembrane protein whose luminal domain is critical to the production of bioactive peptides, to the cytoskeleton. Like dystrophin and the dystroglycan complex, these proteins could mediate communication between these two subcellular compartments. Isolation of the full-length cDNAs encoding P-CIP2 and P-CIP10 and the systematic biochemical and physiological analysis of these molecules will begin to provide an understanding of the mechanisms involved in routing membrane PAM proteins and, thus, will shed light on the biogenesis of LDCVs in neurons and endocrine cells.