Distinct Rab Binding Specificity of Rim1, Rim2, Rabphilin, and Noc2 IDENTIFICATION OF A CRITICAL DETERMINANT OF Rab3A/Rab27A RECOGNITION BY Rim2* □ S

of three independent Miscellaneous Procedures— The reverse transcriptase-PCR analysis also performed as described previously (27, Multiple sequence alignment and depiction of the phylogenetic tree were also performed by the ClustalW program (available on the World Wide Web at hypernig. set at the default parameters scribed

Rabphilin was originally identified as a Rab3A-binding protein on synaptic vesicles (1), and it binds the GTP-bound activated form of Rab3A through its N-terminal domain, consisting of two ␣-helical regions (␣ 1 and ␣ 2 ) and zinc finger motifs (2,3). Since a similar Rab3A binding domain is also found in the N-terminal domain of Rim and Noc2 (4 -7), rabphilin, Rim, and Noc2 have previously been proposed to be specific effectors for the Rab3 isoforms (i.e. Rab3A, Rab3B, Rab3C, and Rab3D) and to regulate secretory vesicle exocytosis in neurons and in some endocrine cells (6, 8 -11) (reviewed in Ref. 12). However, the results of a recent genetic analysis of rabphilin knock-out animals strongly refute this notion, because there are no obvious genetic interactions between Rab3 and rabphilin in neurotransmitter release (13,14). Consistent with this, Rab3A has been shown to function in exocytosis of endocrine cells independent of rabphilin (15,16). In addition, an analysis of a Caenorhabditis elegans unc-10 mutant (homologue of vertebrate Rim) has revealed that there is little genetic interaction between Rim and Rab3, indicating that Rim functions as more than just an effector of Rab3 (17). All of these observations strongly suggest that Rab3A is not a major in vivo binding partner of Rim, rabphilin, and Noc2 and that another unidentified binding protein(s) (possibly other Rab subfamilies) must exist in the body. However, no attempt to identify a genuine ligand of these molecules has ever been made, despite the fact that such information is critical to understanding the different phenotypes in Rim and rabphilin mutant animals at the molecular level.
In this study, I examined the Rab binding activities of Rim1, Rim2, Noc2, and rabphilin by cotransfection assay in COS-7 cells using 42 different Rab proteins (including 37 subfamilies reported in the mouse data bases (18,19)) and found that these molecules exhibit distinct Rab binding specificity. Rim1, Rim2, Noc2, and rabphilin interact differently with several Rab proteins that belong to the Rab functional group III (Rab3, Rab26, Rab27, and Rab37) and/or VIII (Rab8 and Rab10) (18) (see also Fig. 3), whereas the synaptotagmin-like protein (Slp) 1 homology domain (SHD) of Slp homologue lacking C2 domains-a (Slac2-a)/melanophilin specifically recognizes Rab27A/B. I also identified the acidic cluster in the ␣ 1 region as a critical determinant of Rab3A recognition by Rim2. These results suggest that Rim, Noc2, and rabphilin are not specific Rab3 effectors and may activate various Rabs differently than previously thought.

EXPERIMENTAL PROCEDURES
Molecular Cloning of the Mouse Rim1, Rim2, Noc2, and the Rab Family and Construction of Expression Vectors-cDNA encoding an N-terminal Rab binding domain (RBD) of mouse Rim1, Rim2, and Noc2 was amplified from the Marathon-Ready adult brain cDNA (Clontech) by reverse transcriptase-PCR as described previously (20) using the following pairs of oligonucleotides with restriction enzyme sites (underlined) or stop codons (boldface type) designed on the basis of the mouse * This work was supported in part by grants from the Science and Technology Agency of Japan (to M. F.) and Grant 13780624 from the Ministry of Education, Science, and Culture of Japan (to M. F.). 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. sequence reported in the databases: 5Ј-CGGATCCATGTCCTCGGCC-GTGGGGCC-3Ј (Rim1-Met primer; sense) and 5Ј-TCACACCTCAGAT-CCAGCACCTG-3Ј (Rim1-RBD-3Ј primer; antisense; GenBank TM accession number AJ310531); 5Ј-CGGATCCATGTCGGCTCCGCTCGGG-CC-3Ј (Rim2-Met primer; sense) and 5Ј-TCAGGCTTCCTCATTTCGA-AGCC-3Ј (Rim2-RBD-3Ј primer; antisense; accession number NM_053271); and 5Ј-CGGATCCATGGCTGACACCATCTTCAG-3Ј (Noc2-Met primer; sense) and 5Ј-TCACAGAGGTCGGAAGTGGGGA-T-3Ј (Noc2-RBD-3Ј primer; antisense; accession number XM_110922). Three different Rim1 cDNAs (no deletion type (original reported Rim1; simply referred to as Rim1 below), ⌬83-105 (deletion of amino acid residues 83-105 in Rim1), and ⌬56 -105 (deletion of amino acid residues 56 -105 in Rim1)) were produced by alternatively splicing at the RBD1(␣ 1 ) (4, 6) (see Fig. 4A). The longest form of Rim1 was dominant in mouse brain (see Fig. 4B, left lane), whereas the shortest form of Rim2 (corresponding to the Rim1⌬56 -105 form) was dominant (original reported Rim2; simply referred to as Rim2 below) (see Fig. 4B, middle lane).
Miscellaneous Procedures-The reverse transcriptase-PCR analysis was also performed as described previously (27,28). Multiple sequence alignment and depiction of the phylogenetic tree were also performed by the ClustalW program (available on the World Wide Web at hypernig. nig.ac.jp/homology/clustalw.shtml) set at the default parameters as described previously (29).
To determine whether rabphilin interacts with Rabs other than the Rab3 isoforms, I first prepared 42 different recombi- A, schematic representation of three alternative splicing isoforms of Rim1, Rim2, Noc2, and rabphilin. The Rab binding domain of Rim1, Rim2, Noc2, and rabphilin consists of two potential ␣-helical regions, RBD1 (same as ␣ 1 ) and RBD2 (same as ␣ 2 ) (black boxes and shaded boxes, respectively) that are separated by two Cysbased Zn 2ϩ finger motifs. Note that at least three forms of Rim1 are produced by alternative splicing at the C terminus of RBD1 (named Rim1, Rim1⌬83-105, and Rim1⌬56 -105) and that the longest form of Rim1 is dominant in the mouse brain (arrow in B). Similar alternative splicing events may occur in Rim2, according to the results of data base searching (data not shown), but the single form of Rim2 corresponding to Rim1⌬56 -105 is found in mouse brain (see B). B, reverse transcriptase-PCR analysis of the Rab binding domain of Rim1, Rim2, and Noc2 in mouse brain. Note that alternative splicing events in Rim1 and Rim2 were differently regulated in mouse brain. The size of the molecular weight markers (100-bp marker) is shown at the right. The gels shown here are representative of two independent experiments.
nant Rab proteins (including 37 subfamilies reported thus far in the mouse data bases (18,19)) and tested their interaction in COS-7 cells by cotransfection assay with T7-tagged rabphilin and 37 different FLAG-tagged Rab proteins classified into different subfamilies (24,25). As expected, rabphilin strongly interacted with three different Rab species (Rab3A, Rab8A, and Rab27A) and marginally with Rab15, but not with any of the other Rabs tested (Fig. 2B, top panel). By contrast, Slac2-a exclusively interacted with Rab27A and not with any other Rabs, including Rab3A, the closest homologue of Rab27A (Fig.  2C, top panel), consistent with the results of my previous studies (24,25,34). It should be noted that Rab27 (which belongs to Rab functional group III) and Rab8 (which belongs to Rab functional group VIII) have been suggested to regulate the transport of secretory vesicles (18,31,32,44,45), although group III and VIII Rabs clearly form different branches of the phylogenetic tree (Fig. 3).
The distinct Rab binding specificities of Rim1 and Rim2 were surprising, because the N-terminal Rab binding domains are highly conserved in Rim1 and Rim2 (Fig. 1A). During the FIG. 5. Distinct Rab binding specificity of Rim1, Rim2, Noc2, and rabphilin. pEF-T7-Rim1-RBD (or pEF-T7-Rim2-RBD, pEF-T7-Noc2-RBD, or pEF-T7-rabphilin-RBD) and pEF-FLAG-Rabs (Rab3A/B/C/D, Rab8A/B, Rab13, Rab26, Rab27A/B, or Rab37) were cotransfected into COS-7 cells. The proteins expressed were immunoprecipitated with anti-T7 tag antibody-conjugated agarose as described previously (20,25). A, total expressed FLAG-Rabs (one-eightieth volume of the reaction mixture) used for immunoprecipitation. B-G, co-immunoprecipitated (IP) FLAG-Rabs were first detected with HRP-conjugated anti-FLAG tag antibody (1:10,000 dilution) (top panels; Blot, anti-FLAG; IP, anti-T7). The same blots were then stripped and reprobed with HRP-conjugated anti-T7 tag antibody (1:10,000 dilution) to ensure that the same amounts of T7-tagged proteins had been loaded (bottom panels; Blot, anti-T7; IP, anti-T7). Note that rabphilin and Noc2 interacted with Rab27A/B and Rab8A in addition to Rab3 isoforms ( lanes 5, 10, and 11), whereas Rim1 and Rim2 did not interact with Rab27A/B. The positions of the molecular weight markers (ϫ 10 Ϫ3 ) are shown on the left. course of cDNA cloning of mouse brain Rim1, I found that at least three alternative splicing isoforms of Rim1 are expressed in mouse brain (named Rim1, Rim1⌬83-105, and Rim1⌬56 -105; see Fig. 4, A and B) (6), although the longest form of Rim1 was dominant in mouse brain. It should be noted that alternative splicing events occurred in RBD1(␣ 1 ) (Fig. 1A, arrowhead), because RBD1 has been suggested to be important for Rab3A binding of Rim1 (7), and the corresponding region of Slac2-a (i.e. SHD1) is critical for Rab27A binding (34). Interestingly, a single isoform of Rim2 corresponding to Rim1⌬56 -105 was detected in the mouse brain (Fig. 4B). I therefore hypothesized that broader Rab binding specificity of Rim1 and distinct Rab binding specificities of Rim1 and Rim2 may be attributable to the alternative splicing events at the RBD1 (i.e. deletion in the RBD1 of Rim1 may alter the Rab binding specificity of Rim1). To verify this hypothesis, I tested the Rab binding specificity of Rim1⌬83-105 and Rim1⌬56 -105. As expected, the binding activity toward Rab10 and Rab37 was decreased by deletion of amino acids 56 -105, whereas its binding activity toward Rab26 was unchanged (Fig. 5, B-D, lanes 7, 9, and 12 in the top  panels). Although all three Rim1 splicing isoforms bound Rab3A/B/C/D, Rab3D binding activity seemed to be increased by deletion of amino acids 56 -105 (Fig. 5, B-D, lane 4). However, since Rim1⌬56 -105, which corresponds to Rim2, still bound Rab26 but not Rab8A, I concluded that the distinct Rab binding specificities of Rim1 and Rim2 cannot be explained by alternative splicing events in RBD1.
Identification of Critical Determinants on the Rab3A⅐Rim2 Complex in Rab3A-Since Rim2 (or Rim1) recognized Rab3 isoforms but did not recognize Rab27A/B, I next attempted to determine the structural determinant(s) of the Rim2⅐Rab3A complex in Rab3A by chimeric analysis of Rab3A and Rab27A. To do so, I initially focused on the switch I and switch II regions of Rab3A, because structural analysis has shown these regions to directly interact with rabphilin (3), and the specific sequence in the switch II region of Rab27A is critical for Rab27A recognition by the SHD of Slac2-a (34). As noted previously, the switch I region is highly conserved in Rab3 and Rab27, but several amino acids (e.g. Tyr-84, Thr-86, Ile-87, Tyr-91, Tyr-92, and Gly-94) are not conserved in Rab27A/B (Fig. 6A, asterisks in the switch II region) (34). Thus, it is highly possible that Rim2-RBD recognizes a specific sequence in the switch II region of Rab3 isoforms. As shown in Fig. 6B, Rim2 only weakly recognized the mutant Rab3A(Y84F/T86S/I87L/Y91F/Y92F/ G94D) (named Rab3A(27-switch II)) (lane 3), which mimics the Rab27A switch II region, whereas rabphilin bound this mutant, the same as it bound the wild-type protein (lane 6). By contrast, mutations in the switch I in Rab3A (IDF/A3; asterisks in the switch I region) completely abrogated the recognition of Rab3A by both Rim2 and rabphilin (Fig. 6B, lanes 2 and 5), and similar results were obtained for Rab27A. The Rab27A(IDF/A3) switch I mutant did not interact with either Slac2-a or rabphilin (data not shown), and the Rab27A(L84I/F88Y/D91G) switch II mutant interacted with rabphilin but not with Slac2-a (34). These results indicate that both the switch I and switch II regions of Rab3A and Rab27A are essential for recognition by the Rim2-RBD and the Slac2-a-SHD, respectively. However, since Rim seemed to bind Rab3 isoforms with somewhat different affinity (Fig. 5, B-E), additional region(s) of Rab3A may partly contribute to its specific recognition by Rim, because the switch I and switch II regions of Rab3A/B/C/D are almost identical (Fig. 6A).
Identification of Critical Determinants on the Rab3A⅐Rim2 Complex in Rim2-In the final set of experiments, I further attempted to determine the critical determinant of Rab3A recognition by the RBD of Rim2. Since the minimal Rab3A binding domain of Rim1 and rabphilin and the minimal Rab27A binding domain of Slac2-a has been mapped to the first ␣-helical region, critical determinant(s) of Rab3A or Rab27A recognition should be present in this region (7,34). In a previous study, I identified several amino acids that are essential for Rab27A recognition by the Slac2-a SHD (e.g. Glu-14, Val-18, Val-21, and Glu-32 of Slac2-a) (34). However, since these residues are also conserved in Rim1 and Rim2, the corresponding residues of Rim are unlikely to be involved in specific Rab3A recognition. To identify the critical determinant of Rab3A recognition, I carefully compared the amino acid sequences of Rim1, Rim2, rabphilin, Noc2, Slac2s, and Slps again and found an acidic cluster (EEE) in the middle region of the ␣ 1 of Rim1 and Rim2, which is not conserved in the Slp and Slac2 families (Fig. 7A,  number signs). In many members of the Slp and Slac2 families, this position is occupied by L(R/K)(R/K) (i.e. basic residues). To test whether the acidic cluster is involved in Rab3A recognition by Rim2, I prepared mutant Rim2 that carries EEE-to-LRR (mimics Slac2-a) or EEE-to-MEA (mimics rabphilin) substitutions. To my surprise, the Rim2(LRR) mutant did not interact with Rab3A but did interact with Rab27A (Fig. 7B, compare  lanes 2 and 5), although the binding activity of Rim2(LRR) toward Rab27A was weaker than that of Slac2-a (data not shown). Interestingly, the Rim2(MEA) mutant still interacted with Rab3A and weakly with Rab27A, consistent with the fact that rabphilin is capable of interacting with both Rab3A and Rab27A (Fig. 2B). DISCUSSION Although Rim, rabphilin, and Noc2 are widely believed to be Rab3 isoform-specific effectors, the results of the present study demonstrate that Rim1, Rim1⌬83-105, Rim1⌬56 -105, Rim2, rabphilin, and Noc2 exhibit broad Rab binding specificities and bind differently to several Rab proteins that belong to the Rab functional groups III (Rab3A/B/C/D, Rab26, Rab27A/B, and Rab37) and/or VIII (Rab8A and Rab10) ( Fig. 5; summarized in Fig. 1C). This finding is in considerable contrast to the SHDs of the Slp and Slac2 families, which specifically recognize Rab27A and Rab27B but not Rab3, a closely related homologue of Rab27 (24,25). The broad and different Rab binding specificities of rabphilin and Rim may explain the different phenotypes of Rim (Unc-10), rabphilin, and Rab3 mutants in C. elegans and mice (13,14,17). Rabphilin and Rim may function as a Rab8, Rab27, and/or Rab37 effector as well as a Rab3 effector in vivo. Indeed, mammalian homologues of Rab8, Rab27, and Rab37 are also present in C. elegans, although their function remains to be clarified (18). In addition, Noc2 preferentially binds Rab27A/B rather than Rab3A and Rab8A (5) (Fig. 5, F and G), suggesting that Noc2 mainly functions as a Rab27 effector in vivo. Consistent with this notion, both Noc2 and Rab27A are predominantly expressed in pancreatic ␤-cells (5,31), and Noc2, rabphilin, and Rab27A are also coexpressed in PC12 cells (5,8,32).
Sequence comparison between Rim and Slac2s (or Slps) and site-directed mutagenesis have clearly shown that the acidic cluster located in the middle region of the ␣ 1 of Rim2 (Glu-50, Glu-51, and Glu-52) is a critical determinant of Rab3A recognition (Fig. 7). It should be noted that such an acidic cluster was absent in the Slp and Slac2 family and that L(R/K)(R/K) (hydrophobic residue and two basic residues) occupies this position. In addition, structural analysis has shown that the Glu-65 of rabphilin, which corresponds to the Glu-50 of Rim2, directly interacts with Rab3A (3). Interestingly, rabphilin and Noc2, both of which are capable of binding Rab3 and Rab27, contain only one acidic residue and no basic residues (Fig. 7A,  open box). Most importantly, the replacement of EEE by LRR in Rim2 altered Rab3A/Rab27A recognition (Fig. 7B), suggesting that a balance between acidic and basic residues in the middle FIG. 7. Acidic cluster in RBD1 (␣ 1 ) of Rim2 is a critical determinant of Rab3A recognition. A, sequence alignment of the SHD1 of the Slp and Slac2 families and RBD1 of Rim1⌬56 -105, Rim2, Noc2, and rabphilin (7,22,24,28,29). Residues that are conserved and similar in half of the sequences are shown against black and gray backgrounds, respectively. The number signs indicate the acidic cluster of Rim (open boxed), which is not conserved in the SHD1 of the Slp and Slac2 families. The arrowheads indicate the amino acid substitutions in Rim2. Amino acid numbers are indicated on the right. B, pEF-T7-Rim2-RBD mutant and pEF-FLAG-Rab3A or (pEF-FLAG-Rab27A) were cotransfected into COS-7 cells. Co-immunoprecipitated (IP) FLAG-Rab mutants and immunoprecipitated T7-tagged proteins are shown in the middle (Blot, anti-FLAG; IP, anti-T7) and bottom panels (Blot, anti-T7; IP, anti-T7), respectively. The top panel indicates total expressed FLAG-Rab proteins (one-eightieth volume of the reaction mixture) used for immunoprecipitation. Note that the LRR-to-EEE conversion altered the Rab binding specificity of Rim2. The positions of the molecular weight markers (ϫ 10 Ϫ3 ) are shown on the left. region of the first ␣-helical region may be the primary determinant of the Rab3/Rab27 recognition of Rim, Noc2, rabphilin, Slp, and Slac2. Interestingly, the same acidic cluster was also found in the C. elegans Rim (Unc-10) (17), suggesting that Rim functions as a Rab3 (and possibly Rab37) effector but not a Rab27 effector, across phylogeny. By contrast, C. elegans rabphilin contains an SKS sequence (i.e. basic residue) instead of an acidic cluster (14), suggesting that C. elegans rabphilin may function as a Rab27 effector but not a Rab3 effector. Consistent with this, there were no genetic interactions between rabphilin and Rab3 mutants (14). Interaction of C. elegans rabphilin with C. elegans Rab3 or Rab27 is now under investigation in my laboratory. I also noted that analysis of C. elegans rabphilin mutants has revealed genetic interaction between rabphilin and SNARE proteins (syntaxin, SNAP-25, and synaptobrevin/ VAMP) (14), because mammalian Rab27 effectors (Slp4-a and Slp3-a) interact with Munc18 -1 (31, 46), 2 one of the well known syntaxin I-binding proteins that is essential for secretory vesicle exocytosis (47)(48)(49). It is therefore tempting to speculate that both vertebrate and invertebrate Rab27 and their effectors (i.e. Slp and rabphilin) control secretory vesicle exocytosis through interaction with SNARE-related proteins. Further work is necessary to elucidate this possibility.
In summary, I have demonstrated that Rim, rabphilin, and Noc2, which were previously believed to be Rab3-specific effectors, interact differently with functional groups III and VIII Rabs. I have also identified that the EEE/LRR sequence in the first ␣-helical region of the Rab binding domain of Rim is the critical determinant of specific Rab3/Rab27 recognition. Based on these results, I suggest that the previous functional block studies using the Rab binding domain of Rim, rabphilin, and Noc2 should be reevaluated, because these domains are unlikely to function as a specific Rab3 trapper.