Rhotekin, a new putative target for Rho bearing homology to a serine/threonine kinase, PKN, and rhophilin in the rho-binding domain.

Using a mouse embryo cDNA library, we conducted a two-hybrid screening to identify new partners for the small GTPase Rho. One clone obtained by this procedure contained a novel cDNA of 291 base pairs and interacted strongly with RhoA and RhoC, weakly with RhoB, and not at all with Rac1 and Cdc42Hs. Full-length cDNAs were then isolated from a mouse brain library. While multiple splicing variants were common, we identified three cDNAs with an identical open reading frame encoding a 61-kDa protein that we named rhotekin (from the Japanese “teki,” meaning target). The N-terminal part of rhotekin, encoded by the initial cDNA and produced in bacteria as a glutathione S-transferase fusion protein, exhibited in vitro binding to 35S-labeled guanosine 5′-3-O-(thio)triphosphate-bound Rho, but not to Rac1 or Cdc42Hs in ligand overlay assays. In addition, this peptide inhibited both endogenous and GTPase-activating protein-stimulated Rho GTPase activity. The amino acid sequence of this region shares ~30% identity with the Rho-binding domains of rhophilin and a serine/threonine kinase, PKN, two other Rho target proteins that we recently identified (Watanabe, G., Saito, Y., Madaule, P., Ishizaki, T., Fujisawa, K., Morii, N., Mukai, H., Ono, Y., Kakizuka, A., and Narumiya, S. (1996) Science 271, 645-648). Thus, not only is rhotekin a novel partner for Rho, but it also belongs to a wide family of proteins that bear a consensus Rho-binding sequence at the N terminus. To our knowledge, this is the first conserved sequence for Rho effectors, and we have termed this region Rho effector motif class 1.

The Ras superfamily of small GTPases encompasses a group of ubiquitous regulatory proteins related by both structure and function. The products of such genes are involved in a plethora of intracellular signaling processes (1). These proteins are gen-erally regarded as being activated in the GTP-bound form. The intrinsic hydrolytic activity of these proteins is responsible for the reversion to the resting GDP-bound form. Proteins of the Rho subfamily play a pivotal role in the regulation of cytoskeletal organization and the determination of cell polarity. Strongly linked to the formation of stress fibers and focal adhesions (2), regulation of cell motility (3), aggregation (4,5), cell cycle progression (6), and contractile ring formation and cytokinesis (7,8), the Rho proteins occupy key positions in many fundamental cellular processes.
A large number of regulatory proteins for Rho have been characterized, including nucleotide exchange proteins (9 -11), GTPase-activating proteins (GAPs) 1 (12,13), and guanine nucleotide dissociation inhibitors (14,15). In contrast, there has been surprisingly little information available on the nature of the molecules that are directly regulated by Rho. Recently, Rho has been proposed to regulate phosphatidylinositol-4-phosphate 5-kinase and to regulate actin polymerization through increases in phosphatidylinositol 4,5-bisphosphate levels (16). However, a direct interaction with a regulatory element that may give rise to this effect has yet to be demonstrated.
The two-hybrid system was used successfully to demonstrate in vivo interactions between Ras and its downstream effectors, Byr-1 (17), Raf-1, and CYR1 (18). More recently, this system has been used to dissect precisely which features of Ras are involved in interactions with multiple effectors and how this contributes to oncogenesis (19). To isolate and examine downstream Rho targets, we conducted a library screening using the yeast two-hybrid system and complemented our data with in vitro confirmation of the interaction. By these procedures, we have identified a Ser/Thr protein kinase (PKN), a PKN-related protein (rhophilin), and a 180-kDa coiled coil-containing protein (citron) as potential Rho target molecules (20,21). We have also isolated a novel Ser/Thr protein kinase, p160 ROCK , as a potential Rho effector (22) that displays a structural similarity to citron. The present report describes the isolation of a new putative target protein that binds to the GTP-bound form of Rho and inhibits its GTPase activity. The Rho-binding region of this protein appears to be related to those of PKN and rhophilin. London) were expressed in Escherichia coli as GST fusion proteins and were prepared as described (23). Plasmids pVP16 and pBTM116 for use in the two-hybrid system were gifts of Drs. Stan Hollenberg, Rolf Sternglanz, Stan Fields, and Paul Bartel.
Yeast Two-hybrid System Screening-Two-hybrid system screening was conducted essentially as described previously (18), except that strain AMR70 was used in conjunction with L40 in the mating strategy. The initial screening was conducted with a RhoC mutant with a deletion at residue 181, lacking the CAAX box and the polybasic region. Deletion was carried out through PCR amplification using the original rhoC cDNA (27) as a template. This PCR, using a 5Ј-end primer of AGCGGATCCATGGCTGCAATCCGAAAGAAG and a 3Ј-end primer of CCAGAATTCAGACCTGGAGGCCAGCCCGAG, introduced a BamHI site at the 5Ј-end before codon 1 and a EcoRl site at the 3Ј-end after codon 181. This cDNA was then subcloned into the multiple cloning site of a modified pBTM116 plasmid (pBTM116M) (21). This vector was called pBTM116M-rhoC⌬C and drove the expression of a LexA-RhoC⌬C fusion protein. Similar deletions were made also by PCR for rhoA using a 5Ј-end primer of AGCGGATCCATGGCTGCCATCCGGAAGAAA and a 3Ј-end primer of CCAGAATTCAAGCTTGCAGAGCTCTCG and for rhoB with a 5Ј-end primer of AGCGGATCCATGGCGGCCATCCGCAA-GAAG and a 3Ј-end primer of CCAGAATTCACTTCTGCAGCGCG-GCGCGCG with rhoA cDNA (23) and rhoB cDNA (27) as templates, respectively; the resulting cDNAs were inserted similarly to pBTM116M. Full-length rac1 and CDC42Hs cDNAs were excised from the respective pGEX plasmid DNA with BamHI and inserted into pBTM116M to produce LexA-Rac1 and LexA-Cdc42, respectively. A murine day 10.5 embryonic library in pVP16 (18) was screened with a bait plasmid featuring LexA fused to a RhoC deletion mutant (pBTM116-rhoC⌬C). These clones were then used directly for the analysis of LacZ expression. From 1.2 ϫ 10 7 initial transformants, we identified 256 LacZ ϩ histidine prototrophs, 79 of which were cured of pBTM116-rhoC⌬C. Interactions with other proteins were evaluated after mating with yeast strain AMR70 harboring various test baits. Of the 79 cured clones, 22 were LacZ ϩ with the initial screening bait and LacZ Ϫ with the lamin fusion construct. Of these, 12 clones appeared to carry pVP16 containing the same cDNA insert. A 291-base pair insert was excised from the plasmid of clone 21 and designated as C21.
cDNA Cloning of the Full-length Rhotekin-A murine brain oligo(dT)-primed cDNA library in ZAP II (Stratagene) was used to isolate a full-length rhotekin cDNA. A total of 1.1 ϫ 10 6 independent clones were screened on nylon filter membranes (DuPont NEN PlaqueScreen) by hybridization with a 32 P-labeled C21 cDNA. Hybridization of the probe and subsequent washing of filters were carried out as described (28). Positives were rescreened once, and plasmid DNA was rescued using XL-1 Blue E. coli and helper phage VCS M13 (Stratagene) according to the manufacturer's instructions. Nucleotide sequencing was carried out on both strands by the use of the dideoxy chain termination method. To examine the interaction of the full-length rhotekin in the two-hybrid system, the full coding sequence of rhotekin cDNA from the FspI site to the 3Ј-XhoI site in the multiple cloning site of pBluescript SK was inserted in the NotI site of pVP16 to create plasmid pVP16rhotekin (full length).
Northern Blotting-Total RNA was isolated from dissected murine tissues as described (28), and poly(A) ϩ RNA was purified using oligo(dT)-latex beads (Pharmacia Biotech Inc.). Two g poly(A) ϩ RNA was separated by electrophoresis on a 1.2% formaldehyde-agarose gel, transferred to a nylon membrane, and immobilized by UV cross-linking. The RNA was then hybridized with a 32 P-labeled XhoI-XhoI fragment of the full-length rhotekin cDNA in 50% formamide, 5 ϫ SSC, 50 mM Tris⅐HCl, pH 7.5, 5 ϫ Denhardt's solution, 0.1% SDS, and 200 g/ml yeast RNA at 42°C for 16 h. The filter was washed finally with 0.5 ϫ SSC and 0.1% SDS at 65°C and analyzed using a Fuji BAS2000 Bioimage analyzer.
Ligand Overlay Assays-Ligand overlay assays were employed as an in vitro confirmation of positives. The insert from pVP16-C21 was transferred to pGEX-3X to give pGEX-C21. pGEX-rhotekin (amino acids 7-113) was created by introducing the rhotekin coding sequence between FspI and BamHI sites into pGEX-3X. The plasmids were expressed as GST fusion proteins in bacteria, and 5 g of protein of total bacterial lysate was subjected to SDS-polyacrylamide gel electrophoresis and electrophoretically transferred to nitrocellulose membranes (Schleicher & Schuell). The adherent proteins were renatured on the filter and then incubated with radiolabeled small GTPase as described (21,29). Each of the small GTPases was preloaded either with [ 35 S]GTP␥S or with [ 35 S]GDP␤S (both at 1000 Ci/mmol). The bound radioactivity was determined by filter assay, and a 5 nM concentration of the radiolabeled protein was added to the incubation. After incuba-tion, the filter was washed briefly and rapidly dried. Interactions were imaged by autoradiography.
GAP Protection Assay-GAP protection was carried out essentially as described previously (29,30). GST-RhoA (80 -100 pmol) was first loaded with [␥-32 P]GTP (30 Ci/mmol) in buffer A (20 mM Hepes/NaOH, pH 7.5, 50 mM NaCl, 1 mM MgCl 2 , 2 mM EDTA, 1 mg/ml bovine serum albumin, and 10 mM dithiothreitol). The intrinsic rate of GTP hydrolysis was examined by incubating 20 pmol of [ 32 P]GTP-RhoA in 50 l of buffer B (20 mM Hepes/NaOH, pH 7.5, 1 mM MgCl 2 , and 1 mg/ml bovine serum albumin) at 30°C. Duplicate aliquots of 5 l were removed after 0, 2.5, 5, 10, and 15 min and applied to a BA85 membrane. The amount of RhoA-bound [ 32 P]GTP remaining unhydrolyzed after each incubation was determined by the amount of radioactivity adhering to the filter. The GAP-catalyzed rate of GTP hydrolysis was examined in the presence of 1 g of purified GST-rhoGAP. The effect of GST-C21 on the rate of intrinsic and GAP-stimulated GTP hydrolysis was determined by preincubating [ 32 P]GTP-RhoA with 5 g of GST-C21 on ice for 5 min prior to transfer to a bath at 30°C. The dose dependence of the inhibitory effect of GST-C21 on the intrinsic and GAP-catalyzed GTPase activity of RhoA was determined by preincubating 0, 1, 2.5, 5, 7.5, or 10 g of GST-C21 with 8 pmol of GTP-RhoA in 50 l of buffer B for 5 min on ice, followed by transfer to a bath at 30°C for 5 min, with or without the addition of 0.5 g of GST-rhoGAP.

RESULTS AND DISCUSSION
We conducted a library screening with the yeast two-hybrid system in order to identify new partners for Rho. The bait vector pBTM116 drives the expression of the LexA transcription factor fused to a bait protein. The complementary plasmid, pVP16, drives the expression of a nuclear localization sequence and the VP16 transcription activation domain (VAD) fused to a random-primed day 10.5 murine embryonic cDNA library. Positive interactions between the bait and proteins expressed from the library plasmid led to the assembly of a transcriptionally active complex driving the expression of yeast HIS3 and bacterial lacZ genes. An initial screening bait was constructed by deleting the C-terminal polybasic region and the CAAX box of human RhoC (LexA-RhoC⌬C). This strategy was followed on the premise that this region strongly directs Ras and related proteins to the plasma membrane (31,32) and would interfere with the nuclear localization required for transcriptional activation of the reporter genes in this assay.
We identified several clones displaying the same interaction profile. They appeared to bear the same cDNA in pVP16. Sequencing this insert revealed a novel 291-base pair cDNA, which we called C21. In the two-hybrid system, VAD-C21 was strongly positive with RhoC⌬C and RhoA⌬C and weaker with RhoB⌬C (Fig. 1A). Truncation of the C-terminal domains of Rho proteins gave rise to a far stronger interaction than did the full-length forms, possibly either because of a tendency for the CAAX box to favor localization to the plasma membrane or through an increased preference of the full-length baits to interact with endogenous yeast proteins. We presume that this reduces the amount of bait protein available for the formation of transcriptionally active complexes. There appeared to be no interaction with either LexA-Rac1 or LexA-Cdc42.
We then expressed the C21 peptide as a bacterial GST fusion protein and examined its interaction with various small GTPases in vitro by the ligand overlay assay (Fig. 1B). A specific interaction was seen with GTP-RhoA and GTP-RhoB, but not with GDP-RhoA, GTP-Rac1, or GTP-Cdc42. Our attempts to express RhoC as a GST fusion protein were unsuccessful. Taken together, these results are in agreement with the interaction profile observed in the two-hybrid system and clearly demonstrate a specific interaction with GTP-bound Rho proteins. However, due to the method of construction of the library, which included a PCR step, C21 contained one missense PCR mutation and 24 base pairs of the 5Ј-noncoding region as deduced from the full-length cDNA for rhotekin (see below). To ensure that these did not contribute to or interfere with the Rho binding properties of this peptide, we expressed a fragment of the full-length rhotekin containing the N-terminal coding sequence (amino acids 7-113). This peptide was also found to be positive in the overlay assay with GTP-RhoA (Fig.  1B). Moreover, when the full-length rhotekin coding region was introduced into pVP16, this, too, displayed an identical interaction profile in the two-hybrid system as did the original cDNA clone (Fig. 1A). This indicated that in vivo Rho binding activity is a property of the full-length protein as well as the restricted N-terminal fragment.
GST-C21 could be purified from E. coli as a soluble protein, allowing us to investigate its effect upon the endogenous and GAP-stimulated GTPase activity of RhoA in vitro. We found that this peptide inhibited both endogenous and GAP-stimulated GTP hydrolysis ( Fig. 2A), and this inhibition occurred in a dose-dependent manner (Fig. 2B), indicating that not only does this protein inhibit the interaction of a GAP with Rho, but that it can also modify the inherent hydrolytic activity of the cognate GTPase. Similar interactions between a small GTPase, its effector, and GAP have been reported on Rac1/Cdc42 and p65 PAK or Rac1 and p120 ACK and their GAP, Bcr (29,33).
Screening a mouse brain cDNA library using C21 cDNA as a hybridization probe yielded 15 positives from 1.1 ϫ 10 6 independent clones. Three 2.7-kilobase cDNAs were found to be identical and presumed to be full-length (type 1 cDNA) (Fig.  3A). Northern blot analysis of rhotekin mRNA expression using this cDNA as a probe revealed the presence of a transcript of ϳ3 kilobases in brain and kidney tissues (Fig. 4). Weaker expression was also seen in lung, testis, skeletal muscle, heart, and thymus. The size of the transcript appeared to be different in some tissues, and there appeared to be multiple mRNA species in kidney. Consistent with this finding, multiple splicing arrangements were detected also in the brain library, and these inserts appeared also to be full-length (Fig. 3A). Type 2 FIG. 2. GAP protection studies. A, time course. Twenty pmol of [␥-32 P]GTP-RhoA was incubated alone (E), with 1 g of GAP (Ç), with 5 g of GST-C21 (Ⅺ), or with both 1 g of GAP and 5 g of GST-C21 (छ), and GTP hydrolysis was determined as described under "Experimental Procedures." B, effect of increasing concentrations of GST-C21 on the intrinsic (Ⅺ) and GAP-stimulated (E) GTPase activity of RhoA. Eight pmol of [␥-32 P]GTP-RhoA was incubated for 5 min at 0°C in the presence of varying amounts of GST-C21. At time 0, the reaction was transferred to 30°C, and 0.5 g of GAP was added (E). After 5 min, the remaining [␥-32 P]GTP was determined by filter binding assay. The results displayed represent typical results; replicated data varied by ϳ7%.

FIG. 1. Interaction of rhotekin with members of the Rho protein family.
A, in vivo two-hybrid system. Shown is the interaction of various LexA fusions with VAD alone (i), with VAD-C21 (ii), and with VAD-rhotekin (full-length) (iii). The same interaction profile was seen with VAD-C21 and VAD-rhotekin (full length). L40 strains harboring the pVP16 vector, pVP16-C21, or pVP16-rhotekin were mated with strain AMR70 expressing various bait constructs in pBTM116. Diploids were cultured as patches on selective medium for both plasmids and transferred to filter papers (Whatman No. 1), and ␤-galactosidase activity was determined as described (18). B, ligand overlay assays. cDNA contains two exon changes at nucleotide 376 (sequence GAG/GC) and at nucleotide 1910 (sequence ATG/GC). The former was localized in the 5Ј-noncoding region, and the latter caused a 185-base pair insert in the 3Ј-region of type 1 cDNA. The third splicing variant showed an exon change at nucleotide 662 (sequence GAG/GA) of type 1 cDNA and had a different 5Ј-end (type 3 cDNA; data not shown). As only one cDNA clone was obtained for each of types 2 and 3, they were not fully characterized. Type 1 cDNA featured a single open reading frame starting at the ATG codon at base 591 and encoding a protein of 551 amino acid residues with a calculated molecular mass of 61 kDa, which we named rhotekin (Fig. 3B). Two proline-rich motifs were found toward the C terminus of rhotekin (amino acids 421-427 (PAPRKPP) and amino acids 525-533 (PLPPQRSPK)). Such regions have recently been described as general cognate ligands for numerous SH3 groups (34). C21 cDNA covers nucleotides 567-858, which encodes the rhotekin N-terminal peptide (amino acids 1-89). This, together with the finding obtained with the rhotekin fragment (amino acids 7-113), could locate the Rho-binding domain between amino acids 7 and 89. This region showed significant homology to the Rho-binding domains of PKN (35,36) and rhophilin (Fig. 3C) (20). Another common feature is that they are localized in the N terminus of each molecule. However, besides the Rho-binding sites, these proteins are unrelated. Furthermore, data base searches failed to identify any other rhotekin-related proteins. This strongly suggests that this Rho-binding motif is a modular entity that may feature in the regulation of a range of effectors with a spectrum of unrelated activities. We have tentatively termed this domain Rho effector motif class 1 (REM-1). It should be noted that REM-1 has no similarity to the binding motifs of p65 PAK and p120 ACK , which are the Rac/Cdc42 effectors (29,33), nor is it present in the coiled coil-bearing Rho effectors such as p160 ROCK and citron. Thus, REM-1 may define a particular class of Rho effectors. PKN and rhophilin have been proposed to function as Rho effectors because the kinase activity of PKN is stimulated by Rho binding (20). In addition, as expected for effector molecules for the small GTPases, the REM-1-bearing proteins may inhibit the GTPase activity of Rho, as shown for rhotekin in this study.
Each of the putative Rho target molecules we identified by the two-hybrid system showed some difference in their interaction with Rho proteins in this assay. While the strength of a signal in the two-hybrid system is not an absolute indicator of affinity for interacting molecules, two-hybrid data have been shown to broadly reflect relative affinities for related molecules (37). Rhotekin interacted with RhoC and RhoA equally well (this study), whereas rhophilin interacted exclusively with RhoA (20), and citron acted more preferentially on RhoC (21). These findings may indicate a degree of subtlety and complexity of Rho signaling that different Rho proteins may communicate downstream through different patterns of activation of various effectors. To date, no specific actions have been assigned for each member of the Rho protein family. However, differences in expression and cellular localization have been reported for these Rho proteins (38 -40). The above finding also raises the possibility that Rho-effector interaction does not occur through the so-called switch regions alone because these regions are identical among three members of the Rho protein family. Indeed, Diekmann et al. (41) showed that a region other than switch regions of Rho was also important in elicitation of Rho-mediated stress fiber formation.
In conclusion, we have identified a new putative effector for Rho, with a region homologous to other Rho effector molecules.
This region specifically binds GTP-Rho and may constitute the first consensus effector sequence for Rho small GTPases.