Kalirin, a cytosolic protein with spectrin-like and GDP/GTP exchange factor-like domains that interacts with peptidylglycine alpha-amidating monooxygenase, an integral membrane peptide-processing enzyme.

Although the integral membrane proteins that catalyze steps in the biosynthesis of neuroendocrine peptides are known to contain routing information in their cytosolic domains, the proteins recognizing this routing information are not known. Using the yeast two-hybrid system, we previously identified P-CIP10 as a protein interacting with the cytosolic routing determinants of peptidylglycine alpha-amidating monooxygenase (PAM). P-CIP10 is a 217-kDa cytosolic protein with nine spectrin-like repeats and adjacent Dbl homology and pleckstrin homology domains typical of GDP/GTP exchange factors. In the adult rat, expression of P-CIP10 is most prevalent in the brain. Corticotrope tumor cells stably expressing P-CIP10 and PAM produce longer and more highly branched neuritic processes than nontransfected cells or cells expressing only PAM. The turnover of newly synthesized PAM is accelerated in cells co-expressing P-CIP10. P-CIP10 binds to selected members of the Rho subfamily of small GTP binding proteins (Rac1, but not RhoA or Cdc42). P-CIP10 (kalirin), a member of the Dbl family of proteins, may serve as part of a signal transduction system linking the catalytic domains of PAM in the lumen of the secretory pathway to cytosolic factors regulating the cytoskeleton and signal transduction pathways.

Cytosolic proteins are involved in the formation of secretory granules (1)(2)(3)(4)(5)(6) and in the trafficking and localization of integral membrane proteins needed for the synthesis of bioactive peptides (7)(8)(9)(10)(11)(12)(13)(14)(15)(16). We have used one of the few integral membrane proteins known to be involved in the biosynthesis of neuropeptides, peptidylglycine ␣-amidating monooxygenase (PAM), 1 to search for cytosolic proteins involved in these proc-esses (17). PAM is a bifunctional enzyme and integral membrane forms contain an NH 2 -terminal signal sequence, the two catalytic domains that catalyze the sequential reactions required for peptide amidation, a single transmembrane domain, and a short cytosolic domain (18).
PAM is involved in the production of all ␣-amidated peptides and functions only after neuroendocrine-specific endoproteases and carboxypeptidases have exposed the COOH-terminal glycine residue that serves as the nitrogen donor for amide formation (19). Immunocytochemical evidence indicates that PAM begins to function in the trans-Golgi network (TGN), but most peptide amidation occurs in immature secretory granules (20). Using immunoelectron microscopy, integral membrane forms of PAM have been localized to the TGN, especially to distal tubuloreticular regions, and to large dense core vesicles (21).
When expressed independently in neuroendocrine cells, each lumenal catalytic domain of PAM is targeted to large dense core vesicles (22). Integral membrane forms of PAM are localized to the TGN region of both neuroendocrine and nonneuroendocrine cells (7,8). A small percentage of membrane PAM is present on the cell surface or in endosomes at steady state. Elimination of the distal 40 amino acids of the 86-amino acid cytosolic domain results in relocation of membrane PAM to the plasma membrane (8,9). When transferred to a plasma membrane protein such as the interleukin 2 receptor ␣-chain (Tac), the cytosolic domain of PAM directs the majority of the protein to the TGN region and confers the ability to undergo internalization from the plasma membrane (9).
Mutation of a tyrosine residue in the COOH-terminal domain of PAM greatly diminishes internalization of PAM from the cell surface without dramatically altering its TGN localization (9). The TGN localization of membrane PAM is greatly compromised upon deletion of an 18-amino acid domain that includes the tyrosine residue essential for internalization. Integral membrane PAM proteins are phosphorylated, and mutagenesis studies indicate that phosphorylation affects routing (23).
Using a rat hippocampal library and the yeast two-hybrid system, we recently identified partial cDNAs encoding two PAM COOH-terminal interactor proteins (P-CIPs) (17). The biological relevance of these interactions is supported by the fact that the interactions are eliminated when the 18-amino acid segment identified as essential for proper routing of PAM is eliminated. P-CIP2 is similar to serine/threonine dual spec-ificity protein kinases, while P-CIP10 contains at least five spectrin-like repeats (17). In this study we identify the fulllength P-CIP10 protein as a member of the Dbl family of GDP/GTP exchange factors (24) and establish the phenotype of stably transfected AtT-20 cell lines expressing PAM-1 and P-CIP10.

MATERIALS AND METHODS
Cloning of Full-length P-CIP10 cDNAs-The 2.0-kb P-CIP10 cDNA fragment identified using the yeast two-hybrid system (17) was used to screen 1 ϫ 10 6 plaque-forming units from a random primed rat hippocampal cDNA library in -ZAPII (Stratagene). Seven positive clones were plaque-purified, and the two largest cDNA inserts recovered were overlapping 4.7-kb (clone 10/28) and 4.1-kb (clone 10/34) fragments. Both strands of these cDNAs were sequenced completely; only the 5Ј ends of clones 10/28 (upstream sequence from nucleotide 36) and 10/34 (upstream sequence from nucleotide 419) differed. Attempts to extend these sequences by 5Ј-rapid amplification of cDNA ends (RACE) were unsuccessful. The five shorter cDNAs were fragments of the larger pieces. DNA manipulations were carried out according to standard protocols.
Since no in-frame stop codon was identified in clones 10/28 and 10/34, 3Ј RACE was used to extend the 3Ј-end of the 10/28 cDNA (25). Briefly, poly (A) ϩ RNA from adult rat parietal cortex (200 ng) or olfactory bulb (140 ng) was reverse transcribed with the Promega reverse transcription system (Madison, WI) using a RACE hybrid primer (RHP) 5Ј-GGAATTCGAGCTCATCGAT 17 -3Ј (0.75 M) and avian myeloblastosis virus reverse transcriptase (15 units). After the initial 35-cycle amplification using 5Ј-CAGGATGCCTTTCAAGTG-3Ј (nt 4375-4392 of final P-CIP10a) and RHP, an aliquot of the PCR product was used in a nested PCR with 5Ј-CCTCTAGAGCACCCCATCCTCAGACAAT-3Ј (nt 4592-4611 of P-CIP10a) and RHP. A 770-base pair fragment (674 nt of new sequence) from independent amplifications of both tissues was purified, subcloned into pBluescript II (SKϪ) (pBS), and sequenced. The additional 3Ј-sequence contained no in-frame stop codon, so 3Ј-RACE was repeated as above using sense primers 5Ј-CTGCTTCTTCCCCCTG-GTGA-3Ј (nt 5097-5116 of P-CIP10a) for the first round of amplification and 5Ј-GGTCTAGAATGGAGGCAAGTCTGAGT-3Ј (nt 5144 -5164 of P-CIP10a) for the second round of PCR reactions with the same RHP. The 650-base pair fragment (444 nt of new 3Ј sequence) obtained in this second 3Ј-RACE reaction had an in-frame stop codon. Reverse transcriptase-PCR was used to verify that the novel sequence contained in the 3Ј-RACE products was contiguous to clone 10/28.
Construction of Expression Vectors-Construction of pGEX-CIP10, an expression vector encoding a GST fusion protein containing P-CIP10a (aa 447-1138), was described (17). To construct pET-HisDH, a bacterial expression vector encoding all of the Dbl homology (DH) domain and most of the pleckstrin homology (PH) domain of P-CIP10a (P-CIP10a (aa 1254 -1537)), the 857-base pair fragment from clone 10/28 was subcloned into pET28a (Novagen) in-frame with the histidine tag. A mammalian expression vector encoding full-length P-CIP10a (pSCEP.P-CIP10a) was constructed by inserting the full-length cDNA piece from pBS.P-CIP10a into pSCEP (26). To construct pBS.Myc.P-CIP10, the c-Myc epitope (underlined) with a Gly 5 linker (MEQKLI-SEEDLNGGGGG) was joined in-frame to Gly 5 of P-CIP10a using standard methods. The full-length cDNA insert was then transferred to pSCEP to generate p.SCEP.Myc.P-CIP10. All PCR-generated cDNA was confirmed by DNA sequencing.
In Vitro Transcription/Translation and Northern Blot Analysis-Truncated forms of P-CIP10 cDNAs were used as templates for in vitro transcription and translation reactions. pBS.10a (nt 1-1142) was generated by digesting pBS.P-CIP10a with BglII (cuts at nt 1142) and BamHI (cuts in 3Ј-MCS) and religating. pBS.10b (nt 1-1525) was generated in the same manner from pBS.P-CIP10b. Radiolabeled proteins ([ 35 S]methionine, 40 Ci/40-l reaction; Amersham Corp.) were synthesized using a rabbit reticulocyte lysate in vitro transcription and trans-lation system (TNT; Promega Corp.) and analyzed by SDS-PAGE with or without prior immunoprecipitation. Total RNA (10 g) prepared from different adult rat tissues was electrophoresed on 1% agarose gels containing formaldehyde (25). Poly(A) ϩ RNA was prepared with the Promega PolyATtract ® mRNA isolation system II. RNA transferred to nitrocellulose membranes was hybridized with a cDNA probe encompassing nt 1363-3398 of P-CIP10a using standard protocols (25). In situ hybridization was carried out as described (17).
Biosynthetic Labeling, Subcellular Fractionation, and Immunoprecipitation-Cells plated on 15-mm culture dishes and grown to 70 -90% confluency were incubated in methionine/cysteine-free complete serumfree medium for 10 min and then labeled with the same medium containing 1 mCi/ml [ 35 S]methionine/cysteine (ProMix; Amersham) for 15 or 30 min followed by a nonradioactive chase in complete serum-free medium. Cells were either directly extracted into SDS buffer (1% SDS, 50 mM Tris-HCl, pH 7.5, 10 mM ␤-mercaptoethanol) by incubation at 95°C for 5 min or were scraped into TMT buffer (10 mM sodium TES, pH 7.5, 20 mM mannitol, 1% Triton X-100) and subjected to three cycles of freezing and thawing and centrifugation to pellet insoluble material. For the subcellular fractionation experiment, cells were removed from the dishes by scraping into an isotonic buffer (50 mM HEPES-KOH, pH 7.5, 250 mM sucrose), disrupted using a ball bearing cell cracker (15-m clearance), and subjected to differential centrifugation (23).
Immunoprecipitation of P-CIP10 utilized Ab JH2000 or Myc mAb 9E10 (23,28). PAM-1 was immunoprecipitated using Ab JH1764 (22). Immunoprecipitated proteins were resolved by SDS-PAGE and visualized by fluorography. Apparent molecular masses were determined using prestained molecular weight standards (Rainbow standards; Amersham). The capacities of Ab JH2000 and mAb 9E10 were determined by adding increasing amounts of AtT-20/P-CIP10a or AtT-20/Myc.P-CIP10 cell extract to a fixed amount of antibody plus radiolabeled in vitro translated Myc.P-CIP10.

Morphological Studies of Stably Transfected Cells-Live
AtT-20 cells were photographed at low magnification using phase contrast optics. Coded images were analyzed using a BioQuant TCW (version 3.00; R & M Biometrics, Nashville, TN) to record the total number of cells, number of cells with processes (longer than 1 cell body length), number of round cells, number of giant cells (cell body bigger than 3 nuclei), length of each process, and number of bifurcations (branch points). Four or five images were analyzed for each cell line (215-460 cells); after decoding, data were analyzed for statistical significance using a t test.
Complex Formation of GST-GTPases with P-CIP10 -E. coli transformed with vectors encoding GST-Rac1, GST-RhoA, and GST-Cdc42Hs were obtained from Dr. Richard C. Cerione (Cornell University, Ithaca, NY). The purified fusion proteins were prepared, dialyzed, bound to glutathione-Sepharose beads, and depleted of bound nucleotide (29,30). AtT-20 cells expressing PAM-1 or PAM-1 with Myc.P-CIP10 were extracted into 20 mM Tris-HCl, pH 7.5, 50 mM NaCl containing 1% Triton X-100. After freezing and thawing three times, extracts were diluted with 3 volumes of the guanine-nucleotide depletion buffer (30) and centrifuged. For each binding reaction, about 50 g of GST-GTPase bound to 50 l of glutathione-Sepharose beads was mixed with an aliquot of cell extract containing 2 mg (0.25 mg/ml) of protein from nonlabeled cells or 5 ϫ 10 7 cpm/ml acid-precipitable protein from radiolabeled cells. After mixing for 3 h at 4°C, the beads were washed with nucleotide depletion buffer. The beads incubated with nonlabeled extracts were eluted with Laemmli buffer, and bound proteins were subjected to SDS-PAGE and Western blot analysis using Ab JH2000. The beads incubated with radiolabeled cell extracts were eluted by boiling in 50 mM Tris-HCl, pH 7.5, 1% SDS, diluted, and subjected to immunoprecipitation (9) with Ab JH2000.

RESULTS
Tissue Distribution of P-CIP10 mRNA-P-CIP10 was identified in a hippocampal/cortical cDNA library prepared from 3-week-old rat pups that had been subjected to a single maximal electroconvulsive stimulus (17). Before trying to clone a full-length P-CIP10 cDNA, we used Northern blot analysis to determine the size of the P-CIP10 transcript and the tissues expressing the highest levels of P-CIP10 (Fig. 1A). A somewhat heterogeneous set of P-CIP10 transcripts was visualized in total RNA prepared from olfactory bulb, parietal cortex, and hippocampus, with lower levels detected in hypothalamus and none detected in cerebellum. P-CIP10 mRNA was detectable in total RNA prepared from kidney and spleen but was not visualized in anterior or neurointermediate pituitary or atrium, tissues that contain high levels of PAM mRNA. Multiple forms of P-CIP10 mRNA were apparent in all of the tissues examined; when poly(A)ϩ mRNA from olfactory bulb and parietal cortex was subjected to Northern blot analysis, distinct bands of 8.0 and 5.7 kb were detected (Fig. 1A, inset).
A similar distribution of P-CIP10 transcripts was observed when in situ hybridization was performed on sections of adult rat brain (Fig. 1, B and D). P-CIP10 transcripts were prevalent in the olfactory bulb, including the internal granular layers, internal plexiform layer, mitral cell layer, and accessory olfac-tory bulb. P-CIP10 transcripts were also prevalent in all layers of the cerebral cortex, piriform cortex, and amygdala and in the dentate gyrus and CA1-3 regions of the hippocampus. P-CIP10 transcripts were present at lower levels in several hypothalamic structures, including the paraventricular, supraoptic, dorsomedial, and arcuate nuclei.
Isolation of Full-length P-CIP10 cDNAs-Our partial cDNA was substantially shorter than the P-CIP10 mRNAs observed in tissues. We used the 2.0-kb P-CIP10 cDNA fragment to screen a rat hippocampal cDNA library. The two largest cDNA fragments isolated (10/28 and 10/34) were identical except at their 5Ј-ends ( Fig. 2A). No in-frame stop codons were found at either end of either cDNA. The GC content of the 5Ј-ends of both clones was high, and attempts to extend the sequences by 5Ј-RACE were unsuccessful (25); sequence and in vitro translation data (see below) indicated that an initiator Met was included in each clone. By sequentially employing 3Ј-RACE, we extended the sequence to include an in-frame stop codon ( Fig.  2A). No sequence diversity was found in the newly amplified 3Ј-fragments. The fragments were assembled to form two fulllength P-CIP10 cDNAs, P-CIP10a (5Ј-end of 10/28) and P-CIP10b (5Ј-end of 10/34) ( Fig. 2A).
A single long open reading frame with a stop codon near the 3Ј-end was found in both P-CIP10 cDNAs. The GC-rich nature of the 5Ј-end of both P-CIP10a and P-CIP10b and the presence of a single Met residue in both unique regions raised the possibility that a transcriptional start site was present in each clone. The nucleotide sequence surrounding each Met agreed with the consensus translational initiation sequence defined for higher eukaryotes (Fig. 2B) (31,32). To determine whether these potential translational initiation sites were functional, we performed coupled in vitro transcription/translation reactions. We truncated P-CIP10a and P-CIP10b at a common site less than 1200 nt from the potential translational initiation sites so that the predicted 20-amino acid difference between the translation products of P-CIP10a and P-CIP10b would be detectable (Fig. 2B). Each P-CIP10 cDNA yielded a protein of the size predicted if translation were initiated at the Met in each unique 5Ј-region (Fig. 2C); for both P-CIP10a and P-CIP10b, the next Met is more than 80 amino acid residues downstream. The P-CIP10a transcription/translation reaction proceeded much more efficiently than the P-CIP10b reaction, and we used the P-CIP10a cDNA for all further studies.
Structure Predicted for the P-CIP10 Protein-P-CIP10a encodes a protein of 1899 amino acids with a calculated molecular mass of 217 kDa and pI of 5.67 (Fig. 2D). P-CIP10 is largely hydrophilic, with the characteristics of a cytosolic protein. The NH 2 terminus of P-CIP10 lacks a hydrophobic signal sequence, and no hydrophobic stretches typical of transmembrane domains are present. By homology search and computer-based structural analysis, P-CIP10 can be divided into five regions: a short NH 2 -terminal region, a region of spectrin-like repeats, a DH domain, a PH domain, and the COOH-terminal region (Fig. 2E).
The NH 2 -terminal 150 amino acids of P-CIP10 are homologous to Trio, a new member of the Dbl family of proteins identified by virtue of its interaction with the cytosolic domain of the leukocyte common antigen-related (LAR) transmembrane protein-tyrosine phosphatase (33). The next 1000 amino acid residues of P-CIP10 are most homologous to Trio (33), spectrin, and fodrin (34,35). Spectrin and fodrin are cytoskeletal proteins involved in the maintenance of plasma membrane structure by cross-linking to actin and to various integral and membrane-associated proteins (36, 37). Secondary structure predictions indicate that this region of P-CIP10 is almost entirely ␣-helical and that the NH 2 -terminal part of P-CIP10 can FIG. 1. Expression of P-CIP10 mRNA is prominent in specific brain regions. A, total RNA (10 g) from the indicated adult rat tissues was subjected to Northern blot analysis. P-CIP10 transcripts were visualized with the P-CIP10 (nt 1363-3398) probe (17). Inset, poly(A) ϩ RNA (8 g) was visualized with the same probe. P-CIP10 transcripts were visualized in a coronal section through the anterior hypothalamic area (B) or the olfactory bulb (D) using an antisense riboprobe; the sense riboprobe control for a corresponding section is shown for comparison (C). AOB, accessory olfactory bulb, IGr; internal granular layer, IPl; internal plexiform layer, Mi; mitral cell layer, EPl; external plexiform layer, Gl; glomerular layer. Bar, 200 m.  2. Cloning strategy, in vitro transcription/translation, and sequence of P-CIP10. A, the partial cDNAs isolated from the yeast two-hybrid PAM CD interactor screen, I-10 and I-19, are indicated. Using I-10 (P-CIP10 (nt 1363-3398)) as probe, cDNAs 10/34 and 10/28 were isolated from a rat brain cDNA library. RACE products 1 and 2 (RACE#1 and RACE#2) were obtained through sequential use of 3Ј-RACE. P-CIP10a and P-CIP10b were constructed from clones 10/28 and 10/34 with RACE products 1 and 2. B, the nucleotide sequences at the putative translational initiation sites in P-CIP10a and P-CIP10b are compared with the optimal translational initiation site (Kozak) (31, 32); identities are in boldface type. C, for in vitro transcription/translation, pBS.10a (nt 1-1142) and pBS.10b (nt 1-1525) were used as templates. The SDS-PAGE analysis and apparent molecular masses of the translation products are shown; the molecular mass predicted for each translation product is shown in parenthesis. D, amino acid sequence obtained by translating the longest open reading frame of the P-CIP10a cDNA (U70373; P-CIP10b is U88156). The protein sequence reported previously (aa 447-1138) (17) is set off by arrows with double bars; the sequence marked above by a cross-hatched bar is P-CIP19 (aa 473-823). The boxed region (aa 992-1013) shows the peptide used to raise antisera. The beginning and end of each spectrin-like repeat is indicated by double-headed arrows. DH and PH domains are overlined with thick and thin lines, respectively. The triplet encoding Lys 1794 was absent in approximately half of the PCR products sequenced, presumably reflecting a splicing variant. Probable human P-CIP10 clones were identified by screening dbEST and dbSTS; the regions of rat P-CIP10 to which the human ESTs correspond are indicated by dots under the sequence, and sites at which the amino acid sequence predicted for human P-CIP10 differs from that of rat P-CIP10 are indicated. a, AA028043; b, AA027938, four frame shifts inserted; c, H09892; d, AA115289, one frame shift, and AA115265. E, major domains of rat P-CIP10 are drawn to scale. be arranged into nine spectrin-like repeats 103-138 amino acids in length (Figs. 2E and 3A).
Separated from the spectrin-like repeats by 50 amino acids is a 200-amino acid region (aa 1258 -1457) with significant homology to the DH domain defined by Dbl, Dbs, and Ost (24, 38 -40) (Fig. 2, D and E, and Fig. 3B). DH domains were identified first in a family of oncogenic proteins and subsequently shown to catalyze the exchange of bound GDP for bound GTP on specific members of the Rho subfamily of small GTP binding proteins (24). The DH domain with the highest homology to the DH domain of P-CIP10 is the first DH domain of human Trio (90% identity) (Fig. 3B). The DH domains of Dbl, Ost, and Dbs share 42-46% identity with that of P-CIP10. Additional proteins that share significant homology with the DH domain of P-CIP10 include Tiam, FGD1, and yeast SCD1 (Fig. 3B) (41)(42)(43).
The region of P-CIP10 immediately following the DH domain (Fig. 2, D and E; aa 1458 -1555) constitutes a PH domain. PH domains are poorly conserved in sequence and are defined by their common three-dimensional structural motifs (44,45). The PH domain of P-CIP10 has greatest similarity to the PH domain of Trio, followed by the PH domains of Dbl, Dbs, and Ost (Fig. 3B). Although not essential for in vitro GEF activity, the PH domain is generally essential to cellular function (46). PH domains are thought to aid in protein localization by proteinprotein or protein-lipid interaction (44,45,47).
Except at its extreme COOH terminus, P-CIP10 exhibits homology to Trio. The COOH-terminal third of the 2861-amino acid Trio protein contains a second DH/PH domain, immunoglobulin-like repeats, and a putative serine/threonine protein kinase domain and these last two domains bear no homology to P-CIP10. Since the human ESTs identified as homologues of P-CIP10 (Fig. 2D) exhibit a greater degree of identity to rat P-CIP10 than does human Trio, we conclude that Trio and P-CIP10 are encoded by separate genes. The sequence homology exhibited by P-CIP10 and Trio suggests that these two proteins define a subfamily of the Dbl proteins.
Expression of P-CIP10 in AtT-20 Cells-To study the properties of P-CIP10 and its effects on cellular function, P-CIP10a and Myc.P-CIP10 cDNAs ( Fig. 2A) were used to doubly transfect AtT-20 cells stably expressing PAM-1 (AtT-20/PAM-1 cells). AtT-20 corticotrope tumor cells do not express P-CIP10 at a level that allows detection by Northern blot analysis of total RNA (data not shown). Although the transfected P-CIP10 mRNA could easily be visualized by analyzing 10 g of total RNA, the P-CIP10 protein proved difficult to detect. At least two clonal cell lines expressing P-CIP10 and two expressing Myc.P-CIP10 were selected based on Northern blot analysis and one of each was studied in detail.
Direct visualization of P-CIP10 proved impossible with available antisera, so we concentrated the Myc.P-CIP10 protein by immunoprecipitation. Extracts of AtT-20/PAM-1 cells expressing Myc.P-CIP10 were adsorbed to Myc monoclonal antibody immobilized on Protein G beads. Bound proteins were subjected to Western blot analysis and visualized with a rabbit polyclonal antiserum to P-CIP10 (aa 447-1138) (17) (Fig. 4A). A cross-reactive protein of 210 kDa was detected in cells expressing Myc.P-CIP10 but not in AtT-20/PAM-1 cells.
Expression of P-CIP10 in AtT-20 cells could also be demonstrated by metabolic labeling and immunoprecipitation. AtT-20/PAM-1 cells expressing P-CIP10, Myc.P-CIP10, or only PAM-1 were incubated in medium containing [ 35 S]Met/Cys for 15 min, extracted, and subjected to immunoprecipitation using antibody to recombinant GST.P-CIP10 (aa 447-1138) (Fig. 4B). A radiolabeled protein of 210 kDa was detected in both P-CIP10 lines and accounted for 0.03 Ϯ 0.01% of the total protein synthesized during a 30-min pulse. The fact that P-CIP10 was more easily identified using metabolic labeling methods suggested that the protein might have a short half-life. Using a pulse/chase paradigm, both P-CIP10a and Myc.P-CIP10 were found to turn over quickly (Fig. 4C). A semilogarithmic plot of the densitized band intensities yielded a half-life estimate of 60 min for both proteins (Fig. 4D).
To localize P-CIP10, AtT-20/PAM-1 cells expressing Myc.P-CIP10 were biosynthetically labeled for 30 min and subjected to differential centrifugation (Fig. 4E). PAM-1 is recovered in fractions enriched in endoplasmic reticulum (P1 and P2) as well as in fractions enriched in TGN (P2 and P3) and secretory granules (P3). Each particulate fraction as well as the cytosolic fraction contained Myc.P-CIP10. The association of P-CIP10 with particulate fractions, many of which contain intact PAM-1 (23), suggests that P-CIP10 may function by interacting with membranous organelles.
Expression of P-CIP10 Changes AtT-20 Cell Morphology-Since many members of the Dbl family of proteins interact with members of the Rho family of GTPases and affect cytoskeletal organization and cell shape (48,49), we examined the morphology of our stably transfected AtT-20 cells (Fig. 5). Many cells expressing P-CIP10 were larger than wild type AtT-20 or AtT-20/PAM-1 cells. In addition, cells expressing P-CIP10 often had very long processes, some of which were branched. Photomicrographs of randomly selected fields of each cell type were analyzed; the number of giant cells, the percentage of cells having neuritic processes, and the lengths and shapes of processes were quantified (Fig. 5E). Approximately 15% of the total population of P-CIP10-expressing cells had greatly enlarged cell bodies, 5-6 times more than for nontransfected as well as for the PAM-1-expressing cells. Twice as many of the P-CIP10expressing cells had processes, although the average number of processes per cell for cells with processes was unaltered. The percentage of branched processes and the percentage of processes longer than 100 m were dramatically increased for the cells expressing P-CIP10.
Expression of P-CIP10 Alters the Metabolism of PAM-1-In AtT-20 cells, the 120-kDa PAM-1 protein is cleaved by neuroendocrine-specific endoproteases after it exits the TGN (50). Cleavage yields soluble 45-kDa PHM and a 70-kDa membrane  (n ϭ 3). E, biosynthetically labeled AtT-20 cells expressing Myc.P-CIP10 were subjected to subcellular fractionation; Myc.P-CIP10 was immunoprecipitated (Ab JH2000) from resuspended particulate fractions (1000 ϫ g for 5 min (P1), 4000 ϫ g for 15 min (P2), 37,000 ϫ g for 15 min (P3), and 435,000 ϫ g for 15 min (P4)) and the final supernatant (S); proteins were visualized by fluorography. All results were replicated three times. protein containing the PAL, transmembrane, and COOH-terminal cytosolic domains. Active, 45-kDa PHM is stored in large dense core vesicles from which its secretion can be regulated by secretagogues. Membrane PAM proteins that are localized to the cell surface can be cleaved at a site near the transmembrane domain, leading to the accumulation of bifunctional 105-kDa PAM in the medium (8).
To determine whether expression of P-CIP10 changed the metabolism of PAM-1, we performed pulse/chase metabolic labeling experiments on AtT-20/PAM-1 cells and AtT-20/PAM-1 cells expressing P-CIP10a or Myc.P-CIP10. Quadruplicate wells of each cell type were incubated in medium containing [ 35 S]Met/Cys and either harvested immediately (P; pulse) or incubated in the presence of unlabeled methionine/cysteine for 1, 2, or 4 h (C; chase) (Fig. 6). Full-length PAM-1 (120 kDa) disappeared more rapidly from cells expressing P-CIP10 than from PAM-1 cells (Fig. 6A). Consistent with this, production of 45-kDa PHM occurred at a slightly earlier time in cells expressing P-CIP10 ( Fig. 6A; 45 kDa). The total amount of PAM protein recovered after 4 h of chase was less in AtT-20/PAM-1 cells expressing P-CIP10. More bifunctional 105-kDa PAM accumulated in the medium of cells expressing PAM-1 than in the medium of cells expressing PAM-1 and P-CIP10 ( Fig. 6B; 105  kDa).
P-CIP10 Interacts with Rac1 GTPase-The presence of a DH domain followed by a PH domain in P-CIP10 strongly suggested that it would interact with members of the Rho subfamily of small GTP-binding proteins (48,49). The GST fusion proteins of Rac1, RhoA, and Cdc42 were depleted of nucleotide, bound to glutathione-Sepharose, and incubated with extracts of AtT-20 cells expressing PAM-1 or PAM-1/Myc.P-CIP10. The bound proteins were analyzed by Western blot or by immunoprecipitation. Extracts of AtT-20 PAM-1/P-CIP10 cells contained a 210-kDa protein that bound to Rac1 but not to RhoA or Cdc42 (Fig. 7). Binding occurred in the absence of bound nucleotide and was not observed when extracts of control AtT-20 PAM-1 cells were used. We conclude that P-CIP10 is capable of interacting specifically with Rac1.

DISCUSSION
Since our previous mutagenesis studies identified the CD of membrane PAM as essential in establishing its steady state localization in neuroendocrine cells, we screened a rat hippocampal cDNA library for PAM CD interactors using the yeast two-hybrid system (17). We previously identified partial cDNAs encoding a putative protein serine/threonine kinase (P-CIP2) and a protein with spectrin-like repeats (P-CIP10). Elucidation of the complete structure of P-CIP10 revealed the presence of elements common to signal transduction pathways, and expression of P-CIP10 in AtT-20 cells altered PAM processing and cell morphology. The itinerary taken by membrane PAM proteins in AtT-20 cells is complex, and it is anticipated that additional proteins capable of interacting with the CD of PAM will be identified.
The expression of P-CIP10 is highest in specific areas of rat brain. PAM is expressed in the same areas, but PAM is also expressed at high levels in many tissues lacking P-CIP10. For example, atrium, anterior, and neurointermediate pituitary express little P-CIP10 but high levels of PAM. P-CIP10 may fulfill a function unique to its sites of expression, or homologous protein(s) may be involved in the routing of PAM in tissues lacking P-CIP10. The existence of P-CIP10 transcripts with different 5Ј-ends and different sizes suggests that alternate splicing generates diverse forms of P-CIP10. Routing signals in the CD of PAM are recognized in both neuroendocrine and nonneuroendocrine cells (7,8), and proteins homologous to P-CIP10, but expressed more widely, may be involved in these interactions.
The co-expression of P-CIP10 and PAM-1 in AtT-20 cells increased the rate and extent of disappearance of newly synthesized PAM-1. After traversing the Golgi stacks, newly syn- The entire analysis was replicated three times; the asterisk indicates nonspecific bands (8,9). thesized membrane PAM exits the TGN and enters immature secretory granules (50). Although some membrane PAM remains in mature secretory granules, most of the membrane PAM leaves the immature granules, perhaps via constitutivelike vesicles. Membrane PAM that reaches the cell surface is efficiently internalized, with mutagenesis studies suggesting that access to lysosomes is affected by phosphorylation (23). The phenotype observed suggests that P-CIP10 affects routing of PAM in the TGN/immature secretory granule region of the cell. P-CIP10 protein, which turns over with a half-life of less than 1 h, was recovered in particulate fractions and in cytosol, and factors affecting its interaction with particulate fractions remain to be identified.
The structure of P-CIP10, a member of the growing Dbl family of proteins, provides a great deal of insight into its possible functions (24,51). The first members of this family, Dbl, Dbs, and Ost, were identified by their oncogenic activity. Family members share DH and PH homology domains, and most have been shown to serve as GDP/GTP exchange factors for specific members of the Rho family of small GTP-binding proteins. Oncogenic forms of Dbl family members often lack NH 2 -terminal regulatory domains, and the effect of expressing P-CIP10 lacking this region remains to be investigated. Trio, which is broadly expressed, is most closely related to P-CIP10 (33). Trio was identified by virtue of the interaction of its COOH-terminal domain with the cytosolic domain of the LAR receptor type protein-tyrosine phosphatase (33); characterized forms of P-CIP10 lack a homologous region. Phogrin, a homologue of the LAR protein-tyrosine phosphatase, is localized to the regulated secretory pathway of neuroendocrine cells (52), and several other receptor type protein-tyrosine phosphatases are neuroendocrine-specific, raising the possibility that they interact with P-CIP10-related molecules (53)(54)(55).
The NH 2 -terminal half of P-CIP10 contains 9 units homologous to the 5-nm-long structural units that make up spectrin (68). Spectrins provide structural support to the plasma membrane and form an extensive cytoskeletal meshwork by binding to specific soluble and integral membrane proteins (36,37,56). The spectrin-based membrane skeleton restricts the mobility of membrane proteins. An isoform of ␤-spectrin associated with Golgi membranes in Madin-Darby canine kidney cells may play a role in sorting proteins into the vesicular transport pathway (57). The fact that P-CIP10 expression facilitates cleavage of newly synthesized PAM by enzymes located in immature secretory granules identifies the TGN/immature secretory gran-ule as a subcellular site at which PAM/P-CIP10 interactions occur.
The DH/PH domain of P-CIP10 is most homologous to the first DH/PH domain of human Trio (33). Consistent with this homology, P-CIP10 binds to Rac1 but not to RhoA or Cdc42. Many members of the Dbl family, through their DH/PH domain, act as GDP/GTP exchange factors for members of the Rho family of small GTP binding proteins and play important roles in regulating cytoskeletal organization (49,58). The ability of P-CIP10 to act as a GDP/GTP exchange factor remains to be tested. The PH domain of known Dbl family members is generally not essential for in vitro GDP/GTP exchange factor activity, but it is essential for cellular function (46,59). PH domains and the structurally related phosphotyrosine binding domains can bind specific lipids, phosphopeptides, or proteins, targeting the neighboring DH domain to the proper subcellular location (45,46,60). The COOH-terminal region of P-CIP10 retains some homology to the corresponding region of Trio but shows no homology to other proteins in the data base.
Expression of P-CIP10 in AtT-20 cells was associated with the presence of more giant cells, more cells with neuritic processes, and more long processes that branched. Although the underlying mechanisms are not clear, the fact that P-CIP10 interacts with Rac1 is consistent with the occurrence of cytoskeletal effects. In Drosophila, Rac1 affects axonal but not dendritic outgrowth (61), and expression of constitutively active Rac1 in the Purkinje cells of transgenic mice alters both axonal and dendritic morphology (62). In fibroblasts, Rac1 regulates formation of lamellipodia and membrane ruffling (49). Rac1 has been shown to interact with a variety of protein and lipid kinases, thus affecting actin cytoskeletal organization, transcriptional activation, cell proliferation, and secretion (49,58,63,64). Studies using perforated cells demonstrated an inhibitory role for Rac and Rho in transferrin receptor-mediated endocytosis through coated pits (65), and kinectin, a vesicle membrane anchoring protein for kinesin binds activated Rac (66). Overexpression of Tiam1 or activated Rac1 in Tlymphoma cells induced invasiveness (67).
Perhaps most intriguing about the identification of a Dbl family member as a PAM CD interactor is the possibility that we have identified a signal transduction pathway linking the lumen of the secretory pathway to the cytosol. P-CIP10, a putative GDP/GTP exchange factor, interacts with integral membrane PAM, whose pH and conformation-sensitive functional domains (27) reside in the lumen of the regulated secretory pathway, and with Rac1, a small GTP-binding protein that is a component of several distinct cytosolic signal transduction pathways. These interactions suggest a relationship between large dense core vesicle biogenesis and the signal transduction machinery. Identification of additional large dense core vesicle and cytosolic interactors with P-CIP10 and studies on the effects of P-CIP10 on signal transduction pathways regulated by Rho-like GTPases should provide a mechanistic understanding of the biogenesis and routing of large dense core vesicles. We propose giving P-CIP10 the name kalirin to signify its ability to interact with multiple proteins (Kali, Hindu goddess with many hands) and its spectrin-like domains.