Interaction of GRASP, a Protein encoded by a Novel Retinoic Acid-induced Gene, with Members of the Cytohesin Family of Guanine Nucleotide Exchange Factors*

A novel, retinoic acid-induced gene, GRP1-associated scaffold protein (GRASP), was isolated from P19 embryonal carcinoma cells using a subtractive screening strat-egy. GRASP was found to be highly expressed in brain and exhibited lower levels of expression in lung, heart, embryo, kidney, and ovary. The predicted amino acid sequence of GRASP is characterized by several putative protein-protein interaction motifs, suggesting that GRASP may be a component of a larger protein complex in the cell. Although GRASP does not harbor a predicted membrane spanning domain(s), the protein was observed to be associated with the plasma membrane of transiently transfected mammalian cells. Yeast two-hy-brid screening revealed that GRASP interacted strongly with the General Receptor for Phosphoinositides 1 (GRP1), a brefeldin A-insensitive guanine nucleotide exchange factor for the ADP-ribosylation factor family of proteins. GRASP z GRP1 interactions were also demonstrated in vitro and in mammalian cells in which GRASP was shown to enhance GRP1 association with the plasma membrane. Furthermore, GRASP colocalized with endogenous ADP-ribosylation factors at the plasma membrane in transfected cells, suggesting that GRASP may modulate

Retinoic acid (RA) 1 exerts important effects in the processes of vertebrate development, cellular growth and differentiation, and homeostasis (reviewed in Ref. 1). Cellular effects of RA are mediated by two families of nuclear receptors, retinoic acid (RAR) and retinoid X receptors, both of which are members of the steroid/thyroid hormone receptor superfamily of ligand-dependent transcription factors (reviewed in Ref. 2). Retinoid signals induce the expression of a wide array of target genes implicated in mediating the pleiotropic effects of RA on cellular function. These genes encode transcription factors (3)(4)(5)(6)(7)(8)(9), metabolic enzymes (10), growth factor receptors (11), extracellular matrix proteins (12,13), and secreted proteins (14 -16) postulated to convey positional information in the developing embryo.
P19 embryonal carcinoma (EC) cells (17) have been widely used as a model of RA-induced cellular differentiation as these cells have the capacity to differentiate into all three primitive germ layers under the appropriate cell culture conditions and inducers (18). When grown in a monolayer and treated with all-trans-retinoic acid (tRA), P19 cells undergo differentiation along endodermal and mesodermal pathways (19). In contrast, tRA has a biphasic effect on P19 cells grown in aggregates. When treated with low (1-10 nM) concentrations of tRA, P19 cell aggregates differentiate primarily via the endodermal pathway, while high concentrations (Ͼ300 nM) of this retinoid induce differentiation into neurons, glial, and fibroblast-like cells (20). Additionally, P19 cell aggregates treated with Me 2 SO (0.5-1.0%) differentiate into skeletal and cardiac muscle cells (21).
In contrast to the retinoid signals, which are generally antiproliferative and promote cellular differentiation, signaling by members of the Ras superfamily of small guanine nucleotidebinding (G) proteins typically promote cellular proliferation, alteration in cellular morphology, or changes in vesicular trafficking pathways (22). ARF1 through ARF6 proteins are members of the Ras superfamily that are implicated in the control of various vesicular trafficking pathways (22) and may also be involved in cytoskeletal reorganization and morphological phenotypy (23,24). Activation of ARF proteins occurs as a result of GDP/GTP exchange, a reaction catalyzed by guanine nucleotide exchange factors (GEFs). ARF GEFs fall into two broad categories, defined primarily by molecular mass and sensitivity to inhibition by brefeldin A (BFA; Ref. 22), a toxic fungal metabolite that inhibits ARF nucleotide exchange (25). Cytohesin-1 (also known as B2-1; Refs. 26 and 27), ARNO (ARF nucleotide-binding site opener; also known as cytohesin-2; Ref. 28), and GRP1 (29) represent a recently identified subfamily of ARF GEFs that are insensitive to the effects of BFA. These * This work was supported in part by Grant CA51993 from the NCI, National Institutes of Health (to M. I. D. and M. L.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM 1 The abbreviations used are: RA, retinoic acid; RAR, retinoic acid receptor; tRA, all-trans-retinoic acid; RT-PCR, reverse transcriptasepolymerase chain reaction; GST, glutathione S-transferase; BFA, brefeldin A; FITC, fluorescein isothiocyanate; TRITC, tetramethylrhodamine isothiocyanate; PH, pleckstrin homology; HEK, human embry-GEF proteins share a common domain organization and exhibit a high degree of overall sequence similarity (80 -90% amino acid identity). The catalytic activity of these GEFs is greatly enhanced by binding of phosphoinositides (28, 30 -32), which is thought to induce GEF translocation to membranes (32,33), thereby facilitating subsequent interactions with ARFs.
Although the cellular functions of the cytohesin subfamily members are not well understood, conflicting data point to two functional consequences of activation of ARFs by these BFAinsensitive GEFs. Because it was initially shown that these GEFs were able to catalyze GDP/GTP exchange on ARF1, which is known to be involved in endoplasmic reticulum-Golgi and intra-Golgi transport (34,35), attempts were made to analyze the cellular consequences of overexpression of these proteins on the function of the Golgi apparatus. Monier and co-workers (36) have shown that overexpression of ARNO (cytohesin-2) leads to a disassembly of the Golgi complex in transfected HeLa cells, a phenomenon that has also been observed upon treatment of cells with BFA (34). Similar effects have been observed in BHK-21 cells upon overexpression of ARNO3, the human homologue of GRP1 (37). However, endogenous ARNO does not cofractionate with Golgi membranes, but rather does so with plasma membrane markers (38). Furthermore, it has been demonstrated that ARNO colocalizes at the cell periphery with ARF6 in transfected cells (38). In a later study, ARNO was demonstrated to function as a GEF for ARF6 in transfected HeLa cells as evidenced by its ability to induce remodeling of the cortical actin cytoskeleton (39), a known function of ARF6 (23). Similar results implicating GRP1 as a GEF for ARF6 were obtained by other laboratories (40,41).
With the aim of identifying potential target genes that mediate the effects of tRA on induction of cellular differentiation, we isolated cDNAs corresponding to differentially expressed mRNAs from P19 cells treated with tRA. We report herein the cloning, expression, and characterization of a novel gene that is induced by tRA in P19 cells. GRASP is strongly up-regulated by tRA in P19 cells and the protein encoded by this gene displays a plasma membrane localization. GRP1 was isolated as a GRASP interaction partner and GRASP was shown to colocalize with both GRP1 and ARFs at the plasma membrane of transfected human embryonic kidney (HEK) 293 cells. Coexpression of GRASP and GRP1 resulted in an increased partitioning of GRP1 into the particulate fraction of cells suggesting that the two proteins are part of a stable complex at the plasma membrane. The presence of putative protein-protein interaction motifs within GRASP is highly suggestive that GRASP may function as an adaptor or scaffolding protein. We propose that GRASP may aid in the localization of GRP1 to discrete signaling networks and facilitate stimulation of ARF-mediated events at the plasma membrane.

MATERIALS AND METHODS
Cell Culture and Transfections-P19 cells were cultured and treated with tRA essentially as described (42). HEK293 cells were maintained and transfected as described previously (43).
Plasmid Constructs-Vectors for the yeast two hybrid system (pBTM116 and pASV; Ref. 44), the yeast reporter strain L40 (44), and the random primed mouse embryo cDNA library inserted into the VP16 activation domain-encoded pASV3 vector (44) were all kind gifts from Drs. R. Losson and P. Chambon (IGBMC, Illkirch, France). The GRP1 construct was kindly provided by Drs. Michael Czech and Jes Klarlund (29). All constructs were prepared using standard techniques. Supplemental details are available upon request.
Subtractive Hybridization and cDNA Cloning of GRASP-Subtractive hybridization was performed as described (45). Full-length GRASP was isolated from a mouse embryo cDNA library (10.5 days post coitum) that was kindly provided by Dr. P. Chambon and J.-M. Garnier.
Isolation of RNA and Northern Blot Analysis-Total RNA was isolated using TRI reagent (Molecular Research Center) according to the manufacturer's protocol. Poly(A) ϩ RNA was isolated by two rounds of purification on oligo(dT)-cellulose (Amersham Pharmacia Biotech). For Northern analyses, 5-10 g of poly(A) ϩ RNA were separated on a 1% agarose-formaldehyde gel, transferred to a ZetaProbe (Bio-Rad) membrane, and probed with [␣-32 P]dCTP-labeled cDNA fragments. Multiple exposures of the blots were used for quantitative densitometric scanning using the MCID Imaging software as described previously (46).
Yeast Two-hybrid Screening and ␤-Galactosidase Assays-Yeast twohybrid screening and ␤-galactosidase assays were performed essentially as described (47) except that the Saccharomyces cerevisiae L40 reporter strain contained the bait plasmid pBTM116/GRASP 180 -257.
GST Pull-down Experiments-GST fusion proteins were produced, crude bacterial lysates were prepared, and GST pull-down experiments were conducted as described previously (47).
Reverse Transcription-Polymerase Chain Reaction (RT-PCR) Assays-RT-PCR and Southern analysis were performed as described (43) using appropriate primers and probes. Detailed information is available upon request.
Production of GRASP Polyclonal Antibodies-Antisera was raised in the goat against purified GST-GRASP by Bethyl Laboratories (Montgomery, TX). The anti-GRASP antisera was affinity-purified by passing serum from the immunized goat multiple times over a GST immunosorbent to absorb anti-GST antibodies. The serum was then passed over a GST-GRASP immunosorbent to capture antibodies specific for GRASP, and this column was eluted at low pH. The extent of anti-GST contamination in the affinity-purified ant-GRASP antisera was found to be less than 5%.
Indirect Immunofluorescence and Confocal Microscopy-Twenty-four hours following transfection, HEK293 cells growing on coverslips were fixed in 4% paraformaldehyde and permeabilized in two changes of phosphate-buffered saline containing 0.02% Triton X-100. Antibody incubations were performed using standard techniques with the anti-Myc monoclonal antibody (Invitrogen), anti-GRASP polyclonal antibody, or the monoclonal antibody ID9 (kindly provided by Dr. Richard Kahn), which has been reported to recognize all human ARF proteins (48). Myc-GRASP immune complexes were detected using fluoroscein isothiocyanate (FITC)-or tetramethylrhodamine isothiocyanate (TRITC)-conjugated goat anti-mouse antibodies (Southern Biotechnology Associates Inc.). FITC-conjugated rabbit anti-goat and TRITC-conjugated rabbit anti-mouse were used as secondary antibodies for the detection of anti-GRASP and ID9, respectively. Images were captured on a Leica inverted confocal microscope model TCS4D and processed using Photoshop 5.0 (Adobe Systems, Inc.) Preparation of HEK293 Extracts and Immunoblotting-Cytoplasmic and membrane extracts of HEK293 cell extracts were prepared essentially as described (49). Twenty micrograms of protein from the various extracts were subjected to immunoblot analysis as described previously (46) using anti-Myc (Invitrogen) or anti-GFP (CLONTECH) monoclonal antibodies. For coimmunoprecipitation analysis, HEK293 cells were collected in solubilization buffer (10 mM Hepes, pH 7.4, 150 mM NaCl, 1 mM EDTA, 1 mM dithiothreitol, 10% (v/v) glycerol, 0.1 mM phenylmethylsulfonyl fluoride, 10 M pepstatin A, 10 M leupeptin, 25 g/ml aprotinin, 0.05% (v/v) Nonidet P-40, and 0.5% (v/v) Triton X-100) and incubated at 4°C with rotation for 30 min. The lysate was cleared by centrifugation at 16,000 ϫ g for 20 min at 4°C. Three hundred micrograms of lysate protein was incubated with the anti-Myc antibody in a volume of 1.5 ml for 30 min at 4°C, after which a rabbit anti-mouse secondary antibody (Southern Biotechnology Associates, Inc) was added and the incubation continued for an additional 30 min. Protein A-Sepharose (100 l, 50 mg/ml in solubilization buffer; Amersham Pharmacia Biotech) was then added, and the incubation was continued at 4°C with rotation for an additional 30 min. The beads were collected by centrifugation (800 ϫ g for 1 min at 4°C) and subsequently washed gently six times with 1.0 ml of solubilization buffer. Immunoprecipitates were subjected to immunoblot analysis as described above. Various exposures of the blots were used for quantitative densitometric scanning as described previously (46).

RESULTS
Cloning of GRASP, a tRA-induced Gene-Subtractive hybridization was used to identify tRA-induced genes in P19 EC cells grown in monolayers. A 2.1-kilobase pair transcript, referred to hereafter as GRASP, was shown to be induced by tRA in a dose-dependent manner by Northern analysis of P19 cells grown in monolayer culture (Fig. 1A, middle panel). As a control, the blot was also probed for RAR␤, a gene known to be strongly induced by RA in P19 cells (4) and for 36B4, which encodes the acidic ribosomal phosphoprotein P0, a highly expressed "housekeeping" gene (50) that is unresponsive to tRA treatment.
Induction of GRASP in the Presence of Cycloheximide-In order to determine if induction of GRASP occurred independently of protein synthesis, mRNA was isolated from P19 cells treated with 1.0 M tRA in the presence and absence of cycloheximide (10 g/ml; Ref. 51), and analyzed by Northern blot-ting. GRASP mRNA levels in P19 cells were increased by cycloheximide treatment (Fig. 1B, middle panel, lane 3) and this was further augmented by treatment with tRA (compare lanes 3 and 4 of Fig. 1B), suggesting that at least part of the induction of GRASP expression by tRA in P19 cells did not require protein synthesis. Induction of RAR␤ by tRA was not attenuated by cycloheximide treatment under these conditions (Fig. 1B, top panel).
Time Course of tRA-induced Expression of GRASP in P19 Cells-Northern analyses were performed to determine the time course of tRA induction of GRASP in P19 cells grown in monolayer and aggregate cultures, as induction of P19 cell differentiation by tRA under these two conditions results in markedly distinct phenotypes (19,20). Induction of GRASP was observed within 2 h of initiation of tRA treatment in both monolayer and aggregate cultures (  (Fig. 1C, lanes 6 -10). Thus, although RAR␤ and GRASP both appear to be direct targets of tRA in P19 EC cells, the pattern of inducibility of these two genes is subtly distinct, suggesting that additional regulatory mechanisms may exist for one or both genes.
GRASP Contains Multiple Sites for Protein-Protein Interactions-The GRASP fragment isolated by subtractive screening was used to isolate a full-length clone from a randomly primed mouse embryo cDNA library. The 1,978-base pair GRASP clone obtained contained an open reading frame encoding a 392amino acid protein with a predicted molecular mass of 42,382 Da. This open reading frame is preceded by a single, in-frame stop codon (nucleotides [35][36][37] and is followed by a polyadenylation-like motif (nucleotides 1944 -1949, Fig. 2A). In addition, the 3Ј-untranslated region of the GRASP transcript contains an AU-rich cluster (nucleotides 1264 -1276) similar to those found in and proposed to account for the very low level expression of many immediate-early genes (52)(53)(54)(55)(56).
A data base search revealed that GRASP is a novel gene that displayed very little overall similarity with known sequences. However, analysis of the predicted amino acid sequence revealed that GRASP contains several known protein-protein interaction motifs (see  (59 -63) and to the regulator of G-protein signaling RGS12 (64).
Distribution of GRASP in Mouse Tissues-RT-PCR analyses revealed that GRASP was very highly expressed in brain relative to the other tissues examined, but expression was also detected  (64). This alignment was performed using ClustalW 1.7 (89). C, expression of GRASP in the embryo and various adult mouse tissues. RT-PCR and Southern analysis were performed as described under "Materials and Methods." Shown in C is a representative experiment that was replicated three times.
in lung and heart, and to a lesser extent in embryo (10 -12.5 days post coitum), kidney, and ovary (Fig. 2C). Expression of GRASP was not observed in thymus, spleen, liver, or testis.
Subcellular Localization of GRASP in HEK293 Cells-To study the subcellular localization of GRASP in mammalian cells and sequence determinants thereof, constructs encoding Myc epitope-tagged full-length and deletion mutants of GRASP were used to transfect HEK293 cells transiently, and the fusion proteins were visualized by indirect immunofluorescence and confocal microscopy. Although GRASP does not harbor predicted membrane-spanning domains, both Myc-GRASP and Myc-GRASP 1-186 appeared to be primarily localized to the plasma membrane (Figs. 3, A and B) but also exhibited a degree of diffuse cytosolic staining. In contrast, Myc-GRASP 189 -392, which lacks the amino-terminal alanine/proline-rich region and the PDZ domain, was not localized to the plasma membrane but rather was associated with large vesicular structures that were distributed diffusely throughout the cytoplasm (Fig.  3C). These results suggest that full-length GRASP is associated with the plasma membrane of transiently transfected HEK293 cells and that the amino-terminal alanine/proline-rich region and/or PDZ domain are responsible for this localization.
Isolation of GRP1 as an Interaction Partner of GRASP-The presence of putative protein-protein interaction motifs and the lack of any known catalytic domains suggested that GRASP may function as an adaptor protein. Toward the goal of elucidating the function of this novel protein, we employed a yeast two-hybrid system (43,44,65) to isolate proteins expressed in the mouse embryo that may interact with GRASP subdomains. From a screen using the leucine-rich region of GRASP (amino acids 180 -257) as a bait, a clone was isolated that encoded amino acids 1-156 of GRP1. The leucine-rich region of GRASP was observed to interact strongly and specifically with GRP1 1-156 in yeast (Fig. 4A).
In order to confirm the interactions of GRASP and GRP1 that we observed in yeast, in vitro GST pull-down experiments were performed. Full-length GRP1 exhibited a strong interaction with both GST-GRASP and GST-GRASP 1-257 (Fig. 4, B  and C, lanes 3 and 4), but not with GST-GRASP 1-186, GST-GRASP 1-82, or GST alone (lanes 5, 6, and 2, respectively). These results demonstrate that the amino acid residues sufficient for interaction with GRP1 are contained within the leucine-rich region of GRASP (residues 180 -257) and that fulllength GRASP is capable of strong interaction with GRP1 in vitro.
Residues in the NH 2 -terminal Region of GRP1 Mediate the Interaction with GRASP-In order to determine the region of GRP1 responsible for mediating interaction with GRASP, in vitro protein-protein interaction assays were carried out using GST-GRASP and truncation mutants of GRP1 (Fig. 5A). Fulllength GRP1 (Fig. 5B, lane 3) and GRP1 1-95 (lane 9) exhibited a strong interaction with GST-GRASP, but neither protein interacted with GST alone (lanes 2 and 8, respectively). GRP1 1-95 contains a coiled-coil domain (residues 18 -59) and a small fragment of the Sec7 domain (residues 72-95; see Fig. 5A). In contrast, GRP1 72-399, which lacks the amino-terminal coiledcoil domain but contains the entirety of both the Sec7 and PH domains, failed to interact with GST-GRASP (Fig. 5B, lane 6), confirming that the latter domains of GRP1 are not required for interaction with GRASP. Because cytohesin-2 is also expressed in P19 cells (see below) and the amino-terminal coiled-coil region of GRP1 and cytohesin-2 share considerable identity (see Fig. 5C), we next determined if GRASP and cytohesin-2 interact in vitro. Cytohesin-2 1-93, which contains the aminoterminal coiled-coil region (see Fig. 5C), like GRP1 1-95, exhibited a strong interaction with GST-GRASP, but not GST alone (Fig. 5B, lanes 12 and 11, respectively). These findings demonstrate that the amino-terminal regions of both GRP1 and cytohesin-2 are sufficient to mediate interaction with GRASP.
BFA-insensitive ARF GEFs Are Differentially Expressed in P19 Cells-Because GRASP and GRP1 interacted strongly in yeast and in vitro, we next sought to determine if GRP1 and the other closely related members of this subfamily, cytohesin-2 and cytohesin-1, are coexpressed with GRASP during RA-induced differentiation of P19 EC cells, which may facilitate a physiologically relevant interaction. RT-PCR and Southern blotting were employed to determine the relative expression of these transcripts in P19 cells treated with vehicle and tRA (1.0 M, 24 h). Expression of cytohesin-1 was not detected in either untreated or tRA-treated P19 EC cells whereas cytohesin-2 was highly expressed in both samples (Fig. 5D, lanes 1 and 2 and  lanes 3 and 4, respectively). GRP1 expression was also detected in P19 cells (Fig. 5D, lanes 5 and 6). Neither cytohesin-2 nor GRP1 expression appeared to be responsive to tRA treatment of P19 cells (Fig. 5D, lanes 3 and 4 and lanes 5 and 6, respectively).
Coimmunoprecipitation of GRASP and GRP1 from HEK293 Cell Extracts-Coimmunoprecipitation analyses were performed using extracts of HEK293 cells transfected with expression vectors encoding both Myc-GRASP and GFP-GRP1 to determine if GRASP and GRP1 were capable of interaction in a cellular context. Myc-GRASP was immunoprecipitated from whole cell extracts using an anti-Myc monoclonal antibody, and immunoprecipitates were then analyzed by immunoblotting using a monoclonal antibody directed against GFP. The anti-Myc antibody coimmunoprecipitated GFP-GRP1 from cells transfected with Myc-GRASP and GFP-GRP1 (Fig. 6, lane 6), and coimmunoprecipitation of GFP-GRP1 was dependent on the presence of Myc-GRASP (compare lanes 5 and 6). However, the anti-Myc antibody did not coimmunoprecipitate GFP in control cells expressing both Myc-GRASP and GFP (Fig. 6, lane 4). These results clearly demonstrate that GRP1 and GRASP interact in extracts of transiently cotransfected mammalian cells.
GRASP Colocalizes with GRP1 at the Plasma Membrane-As we have shown that GRASP is associated with the plasma membrane of transfected cells (Fig. 3A), we employed immunofluorescence and laser-scanning confocal microscopy to determine if GRASP colocalizes at the cell periphery with GFPtagged GRP1. In the absence of GFP-GRP1 expression, Myc-GRASP was detected at the cell periphery and plasma membrane of HEK293 cells (Fig. 7A; see also Fig. 3A). Consistent with these data, Myc-GRASP was found to be stably associated with the membrane fraction of transiently transfected HEK293 cells as detected by immunoblot analysis (data not shown).
In the absence of Myc-GRASP expression, GFP-GRP1 displayed a diffuse cellular staining that included substantial plasma membrane localization as detected by confocal microscopy (Fig. 7B). However, GFP-GRP1 did not appear to be stably associated with the membrane fraction as detected by immunoblot analysis of cell fractions prepared from these cells (Fig.  7F). For example, the ratio of GFP-GRP1 immunoreactivity present in the cytosolic versus the membrane fraction was approximately 10:1 (Fig. 7F, lanes 9 -11). This finding, when considered with the results of confocal microscopy, suggests FIG. 4. Interactions between GRP1 and GRASP. A, interactions between GRP1 and GRASP in yeast. A bait consisting of the LexA DBD fused to GRASP 180 -257 and a prey consisting of the VP16 activation domain fused to GRP1 1-156 were used to cotransform the L40 strain of S. cerevisiae. ␤-Galactosidase assays were performed in liquid yeast culture as described under "Materials and Methods." Each ␤-galactosidase value represents the mean Ϯ standard error of four to six independent determinations. B, schematic of the GST-GRASP fusion constructs used for in vitro protein-protein interaction assays. C, in vitro protein-protein interaction between full-length GRP1 and GRASP truncation mutants. GST, GST-GRASP, or the corresponding deletion mutants of GRASP were bound to glutathione-Sepharose and used as affinity matrices to examine interactions with full-length, [ 35 S]methionine-labeled GRP1. Shown is a gel from a representative experiment that was replicated three times.

FIG. 5. Interactions of GRP1 and cytohesin-2 with GRASP and relative expression of BFA-insensitive ARF GEFs during RAinduced differentiation of P19 EC cells.
A, schematic of the GRP1 truncation mutants. B, in vitro protein-protein interactions between GST-GRASP and GRP1 and cytohesin-2 truncation mutants. C, sequence alignment of the coiled-coil regions of GRP1 and cytohesin-2. Asterisks indicate amino acid identities. D, relative expression of ARF exchange factors in P19 cells treated with tRA. RT-PCR was performed using 1.0 g of total RNA isolated from P19 cells grown in a monolayer and treated with 1.0 M tRA (or 0.1% ethanol) for 24 h. The blots were probed with an end-labeled oligonucleotide that exhibited identity to all three BFA-insensitive ARF GEFs. Shown in B and D are representative experiments that were replicated four and three times, respectively. that GRP1 may be loosely associated with the plasma membrane in the absence of Myc-GRASP expression.
Coexpression of Myc-GRASP and GFP-GRP1 and staining with the corresponding antibodies revealed a precise overlap in the localization of the two proteins at the plasma membrane of HEK293 cells (Fig. 7, C and D, respectively; see also overlay, Fig. 7E). Moreover, coexpression of Myc-GRASP appeared to stabilize association of GFP-GRP1 with plasma membrane loci as indicated by immunoblotting. In contrast to results obtained when GFP-GRP1 was expressed with increasing amounts of an empty expression vector (Fig. 7F, lanes 9 -11; see also above), cotransfection with increasing amounts of a Myc-GRASP altered the subcellular distribution of GFP-GRP1 such that the cytoplasmic:membrane ratio of GFP-GRP1 immunoreactivity approached unity (Fig. 7F, lanes 12 and 13). The distribution of GFP was unaffected by cotransfection with a Myc-GRASP expression vector (Fig. 7F, compare lanes 4 -6 with lanes 7 and 8). Collectively, these results demonstrate an increased association of GRP1 with plasma membrane loci in the presence of Myc-GRASP. These findings also lend further support to the hypothesis that GRASP⅐GRP1 complexes exist in intact, transfected cells.
GRASP Colocalizes with ARFs at the Plasma Membrane-It is hypothesized that recruitment of GRP1 to the plasma membrane leads to its association with and subsequent activation of ARF proteins (32,41). To determine if GRASP localizes to cellular ARF loci, HEK293 cells were transfected with Myc-GRASP and double-labeled with affinity-purified, anti-GRASP antiserum and a monoclonal antibody that recognizes all human ARF subtypes (ID9; kindly provided by Dr. Richard Kahn; Ref. 48). In agreement with previous results, Myc-GRASP was found to associate with the plasma membrane of transfected HEK293 cells and, to a lesser extent, was diffusely distributed in the cytoplasm (Fig. 8, B, E, and H). Endogenous ARFs appeared to be distributed mainly in the cytosol; however, staining was also apparent at the plasma membrane (Fig. 8, A,  D, and G). An overlay of the GRASP and ARF confocal images revealed a colocalization of the two proteins at the plasma membrane (Fig. 8, C, F, and I). Interestingly, colocalization of ARFs and GRASP was also observed in various surface protrusions of the plasma membrane (Fig. 8I), the formation of which has been observed upon activation of ARF6 (23). These results suggest that GRASP may colocalize with endogenous ARF6, lending support to the hypothesis that GRASP may influence ARF signaling. DISCUSSION We report the molecular cloning of GRASP, a novel gene that is up-regulated in P19 EC cells treated with tRA. Induction of GRASP by tRA in P19 cells was not completely inhibited by cycloheximide, suggesting that this gene may be a direct target of tRA. However, induction of GRASP by tRA differed qualitatively and quantitatively in P19 cells grown as monolayers versus aggregates, suggesting that GRASP may be subject to additional regulatory controls. GRASP is selectively expressed in a number of mouse tissues with the highest level of expression observed in brain, heart, and lung. GRASP transcripts were also detected in 10 -12.5-day mouse embryos.
Analysis of the coding sequence revealed that GRASP harbors multiple sites for protein-protein interactions, which raises the possibility that GRASP may function as an adaptor or scaffolding protein. A search of the protein data bases revealed that GRASP displayed partial sequence similarity with clone B3-1, a gene that was originally isolated from a natural killer (NK)/T cell-subtracted library (66). The regions of similarity include the PDZ domain, the leucine-rich region, and residues within the carboxyl-terminal region (data not shown). Interestingly, B2-1 (27), later referred to as cytohesin-1 (26), was cloned from the same subtracted NK/T cell library, which, in light of the results presented herein, raises the possibility that B3-1 and B2-1 may interact and coordinate some NK cell functions. The function of B3-1 remains unknown, although Dixon and colleagues (66) speculated that B3-1 may be a transcription factor because of the presence of a leucine zipper domain and prevalence of negatively charged residues. However, the results of indirect immunofluorescence and confocal microscopy studies reported herein demonstrated that Myc-GRASP was excluded from the nucleus and appeared to be localized at the plasma membrane. Because there are no predicted membrane-spanning domains within its sequence, it seems likely that GRASP is a peripheral membrane protein.
Localization of GRASP at the plasma membrane appeared to be dictated by the amino-terminal region of the protein, which harbors both the alanine/proline-rich region and the PDZ domain. A deletion mutant lacking the alanine/proline-rich region and PDZ domain, Myc-GRASP 189 -392, was observed to be associated with large vesicular structures located diffusely throughout the cell. It is not known whether Myc-GRASP 189 -392 associates with these structures simply as a result of mislocalization, or whether these structures arise as a result of Myc-GRASP 189 -392 expression, which would suggest a dominant inhibitory effect. A more thorough analysis of the nature of these vesicles will be required to clarify this issue. Nevertheless, our data clearly implicate amino-terminal residues, including the alanine/proline-rich region and PDZ domain, as important determinants of GRASP subcellular localization. These results are consistent with reports that PDZ domains play a role in the proper cellular localization of many proteins (67).
It has been proposed that the PDZ domains of several proteins facilitate synaptic neurotransmission by physically linking various components of a signal transduction cascade at specific, subcellular loci (61, 68 -70). For example, InaD contains five PDZ domains, each of which binds a specific protein component in the Drosophila phototransduction signaling pathway facilitating their juxtaposition and enhancing the ef- ficiency and specificity of signal transmission (71). The possibility that GRASP may perform a similar cellular function within the context of ARF signaling pathways is suggested by the present findings, but this remains to be established.
Members of the cytohesin subfamily of ARF GEFs can be recruited to the plasma membrane of cells in response to growth factor stimulation (72)(73)(74). Activation of many growth factor receptors at the cell surface results in stimulation of phosphatidylinositol 3-kinase, and subsequent phosphorylation of the D3 position on phosphatidylinositol and its phosphorylated derivatives (75). Generation of 3Ј-phosphorylated derivatives, specifically phosphatidylinositol (3,4,5)-trisphosphate, is thought to promote the association of cytohesin subfamily members with membranes through interaction of phosphatidylinositol (3,4,5)trisphosphate with the GEF PH domain (73,74,76). Conversely, inhibitors of phosphatidylinositol 3-kinase, such as wortmannin, have been shown to block translocation of these proteins to the plasma membrane, resulting in cytosolic localization (73,74). It is thought that recruitment of these proteins were cotransfected with either 1.0 g of GFP or GFP-GRP1 expression vectors along with the indicated amounts of either a Myc-GRASP expression vector or an empty vector encoding the Myc epitope (pCDNA3ϩ). Cells were harvested for extract preparation 24 h after transfection. Twenty micrograms of protein derived from the indicated cell extracts were analyzed by immunoblotting using the anti-GFP antibody. Expression of Myc-GRASP was confirmed by immunoblotting (data not shown). Cytoplasmic to membrane (C:M) ratios were generated by densitometric scanning as described under "Materials and Methods." Shown is a representative experiment that was replicated three times.  ID9 (A, D, and G), a monoclonal antibody that recognizes all human ARF proteins (48), along with the appropriately labeled secondary antibodies as described under "Materials and Methods." Shown in C, F, and I are overlays of GRASP and ID9 staining from the representative cells. Note that the staining shown in C, F, and I represents only areas of coincident localization. A-C, D-F, and G-I were generated from independent experiments that were replicated four times. The sizing bar corresponds to 5 m in all panels.
to the membrane creates a favorable interface for the interaction of GEF and ARF proteins, leading to stimulation of GDP/ GTP exchange and subsequent activation of ARFs (28,32,38,41). This is supported by in vitro experiments that demonstrated enhanced GEF stimulated GTP␥S binding by various myristoylated ARFs in the presence of lipid vesicles and phosphorylated derivatives of phosphatidylinositol (28,32,38,41). Enhanced GDP/GTP exchange was completely abolished by addition of the inositol headgroup moiety, inositol 1,3,4,5-tetrakisphosphate, to the reaction, indicating that enhanced GEF activity results from its association with membranes and not an allosteric change that promoted its catalytic activity (32). Taken together, these results are consistent with a model that binding of phosphatidylinositol (3,4,5)-trisphosphate to the PH domain of a particular GEF results in the translocation of that GEF to the plasma membrane, at which the protein is able to promote GDP/GTP exchange on ARF proteins.
The results of the present study identify GRASP as a novel interaction partner for GRP1 and suggest that cellular localization of GRP1 may be influenced by the interaction of the GRP1 coiled-coil domain with GRASP. Similar results have been obtained by Neeb and co-workers (77), who demonstrated an interaction of the coiled-coil region of msec7-1, the rat homologue of cytohesin-1, and the presynaptic protein Munc13-1, which serves to target msec7-1 to the active zone at the presynapse. It may be speculated that the presence of a PDZ domain in GRASP could allow the selective subcellular targeting of GRP1 or cytohesin-2 to discrete plasma membrane domains, resulting in subsequent activation of distinct ARFs. This is an attractive possibility that could result in the juxtaposition of GEF(s) and components of specific signaling pathways. Other motifs present in GRASP could also facilitate interactions with possible downstream effectors, implicating GRASP as a putative scaffold protein. A parallel example of this type of subcellular targeting may exist in the Rho family of GTPases. Tiam1, a GEF for Rac1 (78,79), has been implicated in regulating neuronal morphology (80). In addition to a PH and a Dbl homology domain, Tiam1 contains a PDZ domain (78) that may target this GEF to a particular subcellular locus thus facilitating specificity in mediating Rac1-induced events.
Interaction of GRP1 with GRASP or GRASP-like proteins may also be speculated to promote adhesion of the ␣ L ␤ 2 integrin (LFA-1) to its extracellular ligand, ICAM-1. Kolanus and co-workers (26) have shown that cytohesin-1 interacts with carboxyl-terminal residues of the ␤ 2 integrin receptor subunit, increasing the avidity of LFA-1 for ICAM-1. Because of the strong sequence similarity, it seems likely that other members of the cytohesin subfamily may be able to function in this capacity. If GRASP, or a GRASP-like protein, acting through its alanine/proline-rich region and/or PDZ domain, associates with integrin receptors and/or associated proteins, it is conceivable that recruitment of cytohesin family members to this locus may similarly enhance the avidity of LFA-1 for ICAM-1. Such speculation would be consistent with previous studies demonstrating that treatment of various cell types with RA results in modulation of integrin adhesion (81)(82)(83)(84)(85)(86). For example, Katagiri and co-workers (87) have shown that retinoic acid potentiates cellular adhesion and aggregation mediated by LFA-1 and ICAM-1 in HL-60 cells undergoing neutrophilic differentiation. This alteration in cellular adhesiveness does not result from increased levels of expression of either LFA-1 or ICAM-1, but rather is induced by a change in the avidity state of LFA-1 (87). This change in LFA-1 avidity was shown to require de novo protein synthesis and to occur within 24 h of initiation of RA treatment (87), consistent with the time course of maximal induction of GRASP mRNA in P19 cells reported herein.
In summary, RA-induced expression of GRASP may represent a point of convergence between the nuclear receptor signaling pathways and activation of plasma membrane-associated ARF proteins. Induction of GRASP by RA may influence cellular phenotypy via one or more of the multiple pathways described above that may underlie, at least in part, the pleiotropic effects of RA during cellular differentiation and/or in cellular function.