The RIM/NIM family of neuronal C2 domain proteins. Interactions with Rab3 and a new class of Src homology 3 domain proteins.

RIM1 is a putative effector protein for Rab3s, synaptic GTP-binding proteins. RIM1 is localized close to the active zone at the synapse, where it interacts in a GTP-dependent manner with Rab3 located on synaptic vesicles. We now describe a second RIM protein, called RIM2, that is highly homologous to RIM1 and also expressed primarily in brain. Like RIM1, RIM2 contains an N-terminal zinc finger domain that binds to Rab3 as a function of GTP, a central PDZ domain, and two C-terminal C(2) domains that are separated by long alternatively spliced sequences. Unexpectedly, the 3'-end of the RIM2 gene produces an independent mRNA that encodes a smaller protein referred as NIM2. NIM2 is composed of a unique N-terminal sequence followed by the C-terminal part of RIM2. Data bank searches identified a third RIM/NIM-related gene, which encodes a NIM isoform referred to as NIM3; no RIM transcript from this gene was detected. To test if NIMs, like RIMs, may function in secretion, we investigated the effect of NIM3 on calcium-triggered exocytosis in PC12 cells. NIM3 induced a dramatic increase in calcium-evoked exocytosis (50%), with no significant effect on base-line release, suggesting that NIMs, like RIMs, regulate exocytosis The combination of conserved and variable sequences in RIMs and NIMs indicates that the individual domains of these proteins provide binding sites for interacting molecules during exocytosis, as shown for the zinc finger domain of RIM, which binds to GTP-bound Rab3s. To search for additional interacting proteins for RIMs, we employed yeast two-hybrid screens with the C-terminal half of RIM1. Two members of a new family of homologous brain proteins, referred to as RIM-binding proteins (RIM-BPs), were identified. RIM-BPs bind to RIM in yeast two-hybrid and GST pull-down assays, suggesting a specific interaction. In RIMs, the binding site for RIM-BPs consists of a conserved proline-rich sequence between the two C(2) domains, N-terminal to the beginning of NIMs. RIM-BPs are composed of multiple domains, including three fibronectin type III-domains and three Src homology 3 domains, of which the second Src homology 3 domain binds to RIMs. With the RIM-BPs, we have identified a partner for RIMs that may bind to RIMs at the synapse in addition to Rab3.

Rab proteins are GTP-binding proteins that are generally believed to be essential components of the membrane trafficking machinery of eukaryotic cells (reviewed in Refs. [1][2][3][4]. In brain, a family of Rab proteins collectively referred to as Rab3s is particularly abundant. Four Rab3 isoforms are known (Rab3A, -3B, 3C, and -3D). Of these, Rab3A and Rab3C are concentrated on synaptic vesicles (5,6), while the localization of Rab3B and Rab3D in brain is less clear (7)(8)(9). Rab3A is the best characterized and most abundant Rab protein in the brain, accounting for approximately 25% of total GTP binding by Rabs in brain. Rab3A and Rab3C are attached to synaptic vesicles via a C-terminal lipid modification. Both coordinately dissociate from synaptic vesicles during or after exocytosis, and reassociate after endocytosis (6). In chromaffin cells, exogenous Rab3A inhibits exocytosis, suggesting a regulatory role in exocytosis (10,11). Knockout experiments in mice showed that Rab3A, although the most abundant Rab3 isoform, is not essential for exocytosis or for brain function (12). However, Rab3A performs an important role in regulating the extent of neurotransmitter release in response to Ca 2ϩ (13). In addition, Rab3A is essential for some forms of long term potentiation and long term depression (14,15). Other Rab3 isoforms probably perform similar regulatory roles in synaptic membrane fusion, but their localizations and functions have not been determined.
As GTP-binding proteins, it is likely that Rab3s perform their functions by binding to an effector in a GTP-dependent manner. Two putative effectors for Rab3s have been identified: rabphilin and RIM (16 -18). Rabphilin is a synaptic vesicle protein that is recruited to the vesicles by Rab3 and dissociates from the vesicles together with Rab3 (19). The N-terminal half of rabphilin contains a zinc finger domain that binds Rab3 (17); the crystal structure of this zinc finger domain complexed with Rab3A revealed an unusual fold in which there are multiple contacts between this domain and Rab3 (20). The C-terminal half of rabphilin is composed of two C 2 domains that are similar to the C 2 domains of synaptotagmin and probably also bind Ca 2ϩ (21,22). These findings suggested that rabphilin may couple a Ca 2ϩ -dependent aspect of synaptic vesicle traffic to Rab3 activation by GTP. The second putative Rab3 effector, RIM, is similar to rabphilin in that it also contains an Nterminal zinc finger domain and two C-terminal C 2 domains (18). The zinc finger domain of RIM is homologous to that of rabphilin; both specifically only bind to Rab3s in the GTPcomplexed form. The C 2 domains, however, are quite different in that the RIM C 2 domains do not contain the consensus calcium binding sites that were defined in the synaptotagmin C 2 domains and that are present in the rabphilin C 2 domains (21). Other properties of RIM are even more strikingly distinct from those of rabphilin. RIM is much larger than rabphilin, is subject to extensive alternatively splicing, and contains multiple additional domains that are absent from rabphilin, most notably a central PDZ domain (18). Furthermore, RIM is not localized to synaptic vesicles but instead associated with the synaptic active zone (18). Biochemically, native RIM in brain is a component of the matrix of the active zone; it is virtually insoluble in all detergents except under denaturing conditions. Thus, the two Rab3 effectors interact with Rab3 via homologous N-terminal domains but differ in their subcellular localizations, biochemical properties, and domain structures.
The discovery of two distinct putative effectors for Rab3 at the synapse was somewhat surprising. It raised the question of whether the two effectors mediate different Rab3 functions, whether they cooperate in the same Rab3 function, or whether one of them modifies the function of the other by competing with it. The distinct localizations of rabphilin and RIM suggest different functions; rabphilin would be in an ideal position to mediate a Ca 2ϩ -dependent effect that requires GTP-Rab3 but occurs on free vesicles in the backfield of the synapse, while RIM can only interact with Rab3 when synaptic vesicles are attached to the active zone. In fact, the localization of RIM to the active zone suggests that it could function to recruit vesicles to the active zone in a "tethering" reaction, a function that is similar to what has been porposed for other Rab proteins (3).
The precise respective functions of RIM and rabphilin are currently unclear. A large body of work on rabphilin suggests a variety of possible functions (reviewed in Ref. 23) that involve processes ranging from endocytosis and recycling over assembly of the cytoskeleton to exocytosis. The best controlled studies on rabphilin are probably experiments in chromaffin cells, which demonstrated that rabphilin directly participates in exocytosis, although the mechanism of action remained unclear (24,25). However, rabphilin is not an essential component of the exocytic machinery in brain, since mice lacking rabphilin exhibit no major phenotype (26). Strikingly, mice without rabphilin are viable and fertile without apparent morbidity and exhibit no measurable changes in neurotransmitter release or in the regulation of synaptic transmission (26). This result suggests that RIM may be the more important effector for Rab3, in agreement with its localization to the active zone of the synapse. As a potential effector for Rab3 in regulating neurotransmitter release, RIM in turn presumably binds to downstream target proteins. Although a large number of such proteins have been described for rabphilin, no interacting protein for RIM is known. The large size of RIM and its molecular architecture with multiple separate domains suggest that it binds to several targets in performing its function and that its function is not restricted to an interaction with Rab3. However, it is unclear at this point what these interactions and functions are.
In the current study, we have identified a novel gene for a protein related to RIM, and now refer to the protein products of the two genes as RIM1 and RIM2. Both RIMs are unusually polymorphic due to extensive alternative splicing. Furthermore, we uncovered a second gene product of the RIM2 gene, called NIM2, that is transcribed from the 3Ј-end of the RIM2 gene and composed of only the C-terminal parts of RIM2, including the C 2 B domain. We describe a third related gene, which encodes an analogous protein, called NIM3, and demonstrate that NIM3 also functions in exocytosis as suggested by its homology to RIMs. Finally, we use yeast two-hybrid screens to identify a new class of proteins interacting with RIMs. These proteins are modular SH3 1 domain proteins that we refer to as RIM-BPs. Although it is likely that many additional interacting partners for RIMs remain to be discovered, this is the first step toward elucidating a mechanism of action of RIMs. Together, our data describe a large family of related proteins, the RIMs and NIMs, which function in regulating exocytosis presumably by distinct binding interactions that are partially shared between them and partially specific for the each class of proteins.

Materials and Vectors
All chemicals and enzymes used in the current study were obtained commercially and were of the highest purity available. The following vectors, partly derived from the cDNA clones whose isolation is described below, were constructed using standard procedures (Refs. 27-30; see Figs. 1 and 7).
cDNA Screening, Sequencing, and Sequence Analyses RIM1 and RIM2-GenBank TM searches identified several EST sequences encoding a novel RIM-related protein. One of the corresponding EST clones (yf51e04.r1 Soares infant brain 1NIB Homo sapiens cDNA clone IMAGE:25349) was purchased from the IMAGE consortium and used to screen a rat brain cDNA library in ZAPII by standard methods (27). Thirty-four distinct overlapping clones covering the entire coding region with 50 base pairs of 5Ј-untranslated and 500 base pairs of 3Ј-untranslated region were isolated, and a full-length sequence was assembled from their combined nucleotide sequences and translated into amino acid sequences (Fig. 1A). Among the cDNA clones isolated, eight clones covered the region of alternative splicing (splice site 2) at residue 543 ( Fig. 1B) with six clones with the "A" insert and two clones with the "B" insert. Of eight cDNA clones that contained the alternatively spliced region (splice site 3) at residue 732, two included and six lacked the insert. At the alternatively spliced region at residue 1012 (splice site 4), three cDNA clones contained an insert composed of residues 1013-1034, four cDNA clones contained an insert composed of only residues 1035-1094, and one clone had the entire insert of residues 1013-1094. Five cDNA clones covered the alternatively spliced region at residues 1132-1318 (splice site 5). Of these, three clones lacked the entire insert, one clone lacked residues 1132-1145, and one clone contained the entire insert. Some of these events of alternative splicing were confirmed by GenBank TM searches in EST sequences. For RIM1, sequencing of cDNA clones in addition to those reported earlier (18) and GenBank TM EST sequence analyses identified additional splice variants in splice sites 1 and 5 (see Fig. 1).
NIM3-NIM3 was first identified as a human KIAA0237 sequence in GenBank TM . The rat homolog was cloned by PCR from total rat brain cDNA. To ensure that the 5Ј-end of the NIM3 cDNA was correct and contains a stop codon, extensive cDNA screening of a rat brain cDNA library in ZAPII was performed with a probe from the coding region of NIM3. Ten distinct overlapping clones were isolated that together cover the entire coding region with 0.450 kb of 5Ј-untranslated and 1.5 kb of 3Ј-untranslated region. Three of the cDNA clones contained the initiator ATG.
RIM-BPs-Using the yeast two-hybrid RIM-BP bait clone as a probe (see yeast two-hybrid screens), we screened a rat brain cDNA library extensively to isolate 15 independent cDNA clones covering almost the entire coding region of RIM-BP except for the N terminus. Sequence analysis revealed only one region of alternative splicing where multiple clones either contained or lacked a sequence (residues 1021-1082; Fig.  8). All sequences reported here were submitted to GenBank TM (accession nos. AF199322-199338).

Yeast Two-hybrid Screens and Interactions
Two yeast two-hybrid screens of a rat brain cDNA library in pVP16 -3 (from postnatal day 8) were performed and evaluated as described (18,28,29) using bait vectors encoding the RIM1 PDZ domain (pLexN-RIM1PDZ; residues 492-772) or C 2 domains (pLexN-RIM1C2; residues 713-1588). Of more than 100 positive clones obtained with the PDZ domain bait, 40 were sequenced and tested for interactions with a RIM2 PDZ domain bait. Of the original isolates, only nine clones corresponding to a pioneer sequence without homologies to data bank entries, a single isolate of a homolog of KIAA0378 (also a pioneer sequence without homologies), and a single isolate of an MHC class I antigen were positive with both RIM1 and RIM2 upon retransformation. The screen with the RIM1 C 2 domain bait also resulted in more than 100 positive clones, of which 55 were sequenced and in addition tested for interactions with the RIM2 C 2 domain bait. Of the clones analyzed, a single isolate of a mouse VAP33 homolog (31) and five isolates of RIM-BP1 (see below) were still positive and studied in detail. For mapping of the region of RIM that interacts with RIM-BP and for the relative strength of yeast two-hybrid interactions, yeast strain L40 was co-transformed with the various RIM2 bait vectors listed above and the RIM-BP prey vector isolated in the initial yeast two-hybrid screen. Interactions were first evaluated by ␤-galactosidase staining on a filter; quantitations were then performed for selected bait/prey pairs with a liquid ␤-galactosidase assay that was normalized for protein content (18).

Exocytosis Assays Using Transfected PC12 Cells
To study the effect of NIM3 on exocytosis, PC12 cells were co-transfected with a vector encoding hGH and either an empty second expression vector (control) or a second expression vector encoding NIM3. Transfected cells were then stimulated by KCl depolarization 3 days after transfection, and the amount of hGH secreted as a function of stimulation was determined exactly as described before (18,32). All experiments were performed in triplicate, repeated multiple times, and analyzed statistically with a two-tailed t test.

GST Fusion Protein Affinity Chromatography
We homogenized frozen rat brains (obtained from Pelfreeze) in 0.5% Triton X-100, 1 mM EDTA, 0.1 M NaCl, protease inhibitors, and 50 mM Hepes-NaOH, pH 7.4, and obtained a total brain homogenate after pelleting insoluble material by centrifugation (36). To examine the binding of Rab3 to RIM2, glutathione-agarose columns containing either GST alone or N-terminal RIM2-GST fusion proteins with the zinc finger domain (residues 1-466) or without the zinc finger domain (residues 159 -319) were incubated with the brain homogenate at 4°C overnight in the presence of either 0.5 mM GDP␤S or GTP␥S. Samples were washed three times in the same buffer without nucleotides before analysis by SDS-polyacrylamide gel electrophoresis and immunoblotting with Cl42.1 antibody against Rab3 and antibody C7.2 against synaptophysin. To test the binding of RIM-BP to RIMs, glutathioneagarose columns containing GST alone or GST-RIM-BP fusion proteins with the N-terminal (residues 650 -1270) or C-terminal region of RIM-BP (residues 1271-1625) were used for pull-downs. Since RIMs are almost completely insoluble in brain homogenates with nondenaturing detergents (18), RIMs first had to be solubilized with 1% SDS from the homogenates. The SDS-containing brain extract was then diluted with binding buffer containing Triton X-100 to quench the SDS, resulting in final concentrations of 0.16% SDS and 1% Triton X-100. This brain extract was used in the pull-down experiments with GST-RIM-BP fusion proteins as described above for Rab3 binding, except that no nucleotides were added.

RNA Blotting Experiments
RNA blotting experiments were performed using multiple tissue blots purchased from CLONTECH. Northern blots were hybridized at high stringency with a DNA fragment encoding residues 159 -319 of RIM2, the entire coding region of NIM3, residues 1271-1625 of RIM-BP, or a ubiquitously expressed control protein (VAP33).

Identification and Molecular Cloning of RIM2: The Conserved Molecular Architecture of a Family of Rab3-interacting
Molecules-Data bank searches uncovered several EST sequences that were homologous to, but distinct from, the RIM protein that we had originally isolated (data not shown). To determine if these sequences originated from a RIM-related protein, we used EST clones as probes to isolate overlapping cDNA clones from a rat brain library. Assembly of the sequences of these clones revealed that they encode a large protein of 1555 residues that is closely related to RIM (Fig. 1A). The new protein is similar to RIM over its entire length, suggesting that it represents a novel RIM homolog. This prompted us to refer to the old and new proteins as RIM1 and RIM2, respectively. Partial human sequences for RIM1 and RIM2 were deposited in GenBank TM as random cDNA sequences (KIAA0340 and KIAA751, respectively). Their translated sequences are highly homologous to the rat RIM1 and RIM2 sequences reported here, suggesting that both RIMs are evolutionarily conserved in vertebrates (data not shown).
The two RIMs exhibit the same overall domain architecture consisting of an N-terminal zinc-finger module, a single central PDZ domain, and two C-terminal C 2 domains (Fig. 2). These modules are highly conserved between RIMs and are connected by sequences of variable length and conservation. Some of the connecting sequences are very homologous between RIMs, while others diverge (Fig. 1A). For example, the coupling of a PDZ domain with a C 2 domain is characteristic of RIMs (Fig. 2), and the short sequence that links these two domains is almost identical between the RIMs, indicating that the connection between the two domains is functionally important. Furthermore, the sequence preceding the PDZ domain is also highly conserved among RIMs. In contrast, the more N-terminal region between the zinc finger and PDZ domains contains islands of similarity separated by stretches of variable sequences (Fig.  1A). Some of the conserved sequences outside of the identified domains may represent novel domains that have not yet been defined; future studies will have to determine if these sequences represent independently folding modules. The fact that only some of the connecting sequences are conserved while others are variable suggests that the conserved sequences are functionally meaningful.
Alternative Splicing of RIMs-Analysis of multiple independent RIM2 cDNA clones revealed heterogeneity between clones at several positions, suggesting that RIM2 is extensively alternatively spliced. Similar alternative splicing was previously observed for RIM1 (18). We analyzed the alternative splicing of RIM2 by sequencing multiple independent cDNA clones, and reanalyzed RIM1 alternative splicing by a further characterization of cDNA clones for this protein. In these analyses, we accepted as bona fide alternative splicing only events that could be reproduced in multiple independent cDNA clones or independently observed in sequences reported in Gen-Bank TM or detected in both RIM1 and RIM2. Of the five sites of alternative splicing that were identified in this manner, one site was only found in RIM1 and two only in RIM2 (all three in the N-terminal half of RIMs), while the remaining two sites of alternative splicing were present in both RIM1 and RIM2 (in the C-terminal half of RIMs) ( Figs. 1 and 2). To facilitate discussion of the sites of alternative splicing, we number the sites consecutively from the N to the C terminus for both RIMs, although the N-terminal sites of alternative splicing were not shown to be present in both RIMs (Fig. 1A). As described below, the presence of multiple types and combinations of inserts in these sites generates a large diversity of RIM proteins with potentially more than 100 isoforms, suggesting that RIMs are highly polymorphic neuronal proteins. Fig. 1. Primary structures and alternative splicing of RIM1, RIM2, N1M2, and NIM3. A, the rat sequences of RIM2, RIM1, N1M2, and NIM3 as deduced by cDNA cloning are aligned for maximal homology in single-letter amino acid code. Sequences are identified on the left and numbered on the right. The numbering corresponds to the sequences shown, since the actual number of amino acids differs among the various splice forms. Shared residues are highlighted in a color code corresponding to protein domains as follows: green, the zinc finger domain that binds to Rab3s; blue, the PDZ domain; red, the two C 2 domains; yellow, identical residues outside of defined domains. Cysteine residues in the Rab3-interacting domain that are coordinating the two zinc atoms are shown on a black background. The SH3 domain-binding site between the C 2 domains is underlined, and the prolines involved in binding are also shown on a black background. Dashes indicate gaps. Alternatively, spliced sequences are shown in italic cyan-blue type (see "Experimental Procedures"). Sites of alternative splicing are numbered 1-5 above the sequences, with the same numbering for both RIMs, although not all of the sites have been demonstrated to be alternatively spliced in both genes. B, sequence variants at splice site 2 N-terminal to the PDZ domain of RIM2. Two variants were observed in rat cDNAs, referred to as variants a and b. Variant a is similar to rat and human RIM1 (KIAA0340; GenBank TM accession no. AB002338), and variant b is similar to the human RIM2 sequence encoded by KIAA0751 (hRIM2b; GenBank TM accession no. AB018294), suggesting that the two variants are not cloning artifacts. Splice site 1 is in the N-terminal Rab3-interacting domain. Here RIM1 displays a variable sequence that disrupts the homology with RIM2 and with the Rab3-binding domain of rabphilin (shown in green in Fig. 1A). Previous cDNA cloning identified variants lacking residues 83-105 and 83-106 at this site (18). In addition, we observed a human EST sequence (GenBank TM accession no. AA774730) that lacks residues 57-105. Thus, there are at least four variants of RIM1 in this site, although none have yet been identified for RIM2. In RIM2, two different inserts were observed at splice site 2 immediately before the PDZ domain (residues 544 -564 in Fig. 1A; aligned in Fig. 1B). The shorter variant A (20 residues) is homologous to the corresponding sequence observed in RIM1. The longer variant B (68 residues) exhibits no homology to RIM1. The human RIM2 sequence in the random cDNA clone KIAA0751 also contains this insert, suggesting that it does not represent a cloning artifact. In splice site 3 at the beginning of the first C 2 domain, residues 733-748 were either present or absent in multiple RIM2 cDNA clones. No alternative splicing in sites 2 and 3 was observed for RIM1 at this point.
Probably the most interesting alternative splicing of RIMs occurs in the region between the two C 2 domains at the C terminus at splice sites 4 and 5. This region contains two sites of alternative splicing that are separated by only 38 and 37 residues in RIM1 and RIM2, respectively. These sites of alternative splicing are composed of the largest number variants, contain the longest inserts (up to 187 amino acid residues), and are conserved between RIM1 and RIM2.
In RIM2, three variants were observed in splice site 4, the first of the two C-terminal splice sites: a variant containing residues 1013-1034 but lacking residues 1035-1094, a variant containing only the second sequence (residues 1037-1097), and a variant containing both sequences (residues 1015-1097). It seems likely that a variant lacking all inserts also exists. Analysis of new cDNA clones revealed that RIM1 is expressed in at least four variants at this splice site: a variant lacking all inserts that joins residue 981 to residue 1118 and variants containing the full insert (residues 1032-1167) or different partial inserts (residues 1083-1167 and 1107-1167). A picture of a modular design of alternatively spliced sequences emerges from these observations, with as many as four blocks of sequences that can be variably inserted into, or omitted from, the RIM sequences at this position. These four blocks of sequences correspond to residues 1032-1053, 1054 -1082, 1083-1106, and 1007-1167 in RIM1. Of these blocks, the splice sites following residues 1031, 1106, and 1176 were observed in both RIMs. Splice site 5 of RIM2 was also subject to complex alternative splicing in that either the whole region or only residues 1132-1145 or 1145-1318 were spliced out. Here, RIM1 exhibits only two variants (18). Again the alternatively spliced inserts are highly homologous between the two RIMs. In the analysis of the limited number of RIM1 and RIM2 cDNA clones performed in our studies, we observed no evidence that alternative splicing of RIMs at the different sites is interdependent. If the different splice sites are used independently, combinatorial mixing of different inserts in the various sites would result in more than 100 distinct RIM proteins that differ by as much as 30 kDa.
GTP-dependent Interaction of RIM2 with Rab3s-RIM1 was initially identified because of the GTP-dependent binding of its N-terminal zinc finger domain to Rab3A and Rab3C. The sequence homology between RIM1 and RIM2 in the N-terminal zinc finger domain suggests that RIM2 may have a similar activity. To test this, we performed GST-pull-down experiments with the N-terminal region of RIM2 (Fig. 3). The Nterminal zinc finger domain of RIM2 specifically captured both Rab3A and Rab3C, but not synaptophysin used as a negative control, from rat brain homogenates. Binding was GTP-dependent, since only GTP␥S and not GDP␤S allowed binding. The full-length zinc-finger domain (residues 1-466) was required for binding, while GST alone or a fragment of the zinc finger domain (residues 159 -319) exhibited no background binding of Rab3s (Fig. 3). Thus, RIM2 specifically binds Rab3s, similar to RIM1, in a GTP-dependent interaction.
Cloning and Characterization of NIMs-In addition to the various splice variants, sequencing of RIM2 cDNA clones uncovered two overlapping cDNA clones from the C terminus of RIM2 that differed from the other cDNA clones. These two clones contained all of the sequences 3Ј of splice site 5 of RIM2 but diverged N-terminally from the RIM2 sequence at the 3Ј junction of splice site 5. The 5Ј-end of the inserts of the two clones encoded a unique sequence not found in the other RIM cDNA clones; this sequence included a consensus initiator methionine and was preceded by a stop codon in the clone with the largest insert. This result suggested that in addition to fulllength RIM2, a separate shorter mRNA is transcribed from the RIM2 gene, probably by an independent promoter. This shorter mRNA encodes a protein composed of a unique N-terminal sequence followed by the entire sequence of RIM2 from splice site 5 onwards, including the C-terminal C 2 B domain (Fig. 1A). Because this protein lacks a Rab3-interacting domain and thus cannot bind to Rab3 (see Fig. 3) but is nevertheless derived from the RIM2 gene, we called it NIM2.
Interestingly, further data bank searches identified EST sequences and a random human cDNA clone (KIAA0237; Gen-Bank TM accession no. D87074) that were highly homologous to, but distinct from, the C-terminal parts of RIM1 or RIM2. Similar to NIM2, these human sequences N-terminally diverged from the RIM structures at splice site 5 and also contained in-frame stop codons at the 5Ј-end, suggesting that these entries correspond to a novel NIM3 protein. This finding raised the question of whether a corresponding RIM3 protein exists for NIM3, since there is a RIM2 protein for NIM2. To test this possibility, we performed extensive cDNA screening experi-FIG. 3. GTP-dependent binding of rat brain Rab3A and Rab3C to recombinant RIM2 using GST pull-downs. Glutathione-agarose beads containing GST alone or two different GST fusion proteins of RIM2, a larger protein that includes the entire N-terminal zinc finger domain (GST-RIM2-(1-466)) or a shorter protein with part of this domain (GST-RIM2-(159 -319)), were used in pull-down experiments with total rat brain homogenates solubilized with Triton X-100. Pulldowns were carried out in the presence of 0.5 mM GDP␤S or GTP␥S as indicated. Proteins bound to the beads and the starting homogenate (H) were analyzed by SDS-polyacrylamide gel electrophoresis and immunoblotting with antibodies to Rab3A and Rab3C as a test and to synaptophysin as a negative control. Positions of the immunoreactive bands on the gels are identified on the right. ments with a rat NIM3 probe generated by polymerase chain reaction. We isolated a total of 10 cDNA clones, three of which contained the N terminus of NIM3 as defined by the KIAA0237 sequence (Fig. 1A). In these clones, the putative initiator methionine was also preceded by in-frame stop codons, supporting the notion that NIM2 and NIM3 consist only of a C-terminal C 2 B domain domain preceded by a relatively short sequence. No cDNA clone was isolated that shared homology with RIMs N-terminally to splice site 5. No sequence variants were observed in 10 independent cDNA clones of NIM3, indicating that NIM3, like NIM2 but in contrast to RIM1 and -2, is not subject to alternative splicing. Together these results provide evidence that although NIM3 is a shorter version of RIMs analogous to NIM2, the NIM3 gene probably gives rise to only a NIM and not a RIM transcript (Fig. 2).
Sequence alignment of NIM3 with NIM2, RIM1, and RIM2 reveals that NIM3 is highly homologous to the C-terminal sequences of these proteins. As in NIM2, the homology covers not only the C 2 B domain but also the sequences preceding and following it (Fig. 1A). These regions in fact are the most homologous sequences in the RIMs, indicating that these regions are functionally particularly important. The human NIM3 gene is present on a fully sequenced genomic clone from chromosome 1q33-34.2 (clone 739H11; GenBank TM accession no. AL031289). The gene has six coding exons distributed over 15.5 kb. The exon-intron junction after the first coding exon corresponds to the boundary of alternative splicing at splice site 5 in RIMs, supporting the notion that alternative splicing of RIMs involves alternative use of exons. No RIM-like sequences were detected in the NIM3 gene upstream of the N terminus of NIM3, providing further evidence that the NIM3 gene probably does not encode an analogous RIM3. In addition, the 3Ј-end of the human RIM1 gene has also been elucidated in the course of the human genome project and localized to chromosome 6q12-13 (clone RP5-1046G13; GenBank TM accession no. AL035633). Analysis of this sequence failed to uncover a NIMlike exon, suggesting that the RIM1 gene, in contrast to the RIM2 gene, may not encode a "NIM" protein (data not shown).
Tissue Distribution of RIM2, NIM2, and NIM3 Expression-We performed RNA blotting analyses of rat tissues with RIM2-and NIM3-specific probes to elucidate where they are expressed. Both proteins were found to be encoded by large mRNAs (approximately 7 kb) that were synthesized at high levels exclusively in brain of the tissues analyzed (Fig. 4). RNA blots hybridized with a 5Ј probe from RIM2 detected only a single size mRNA. In contrast, hybridizations with a 3Ј probe revealed two classes of mRNAs consistent with separate mRNAs for RIM2 and NIM2 (data not shown). Both the RIM2 and NIM2 mRNAs were primarily expressed in brain, with low level expression in testis but no other tissue investigated; NIM3 was completely brain-specific. These results mirror the previously described expression pattern of RIM1 (18) and suggest that all four proteins of this family are highly enriched in brain.
Regional Expression of RIMs and NIM3 in Rat Brain-To examine the protein products of the RIM2 and NIM3 mRNAs, we raised antibodies to synthetic peptides and GST fusion proteins containing parts of their sequences. The specificity of the RIM2 and NIM3 antibodies was probed with COS cells transfected with expression constructs of the various proteins and brain homogenates (Fig. 5 and data not shown). In brain homogenates, the RIM2 antibodies reacted with multiple closely spaced bands of the same size as RIM1 (data not shown; also see below). The multiplicity of these bands is consistent with the extensive alternative splicing observed in the RIM2 cDNAs (Fig. 1). NIM3 antibodies, by contrast, recognized a single major protein in brain that had the same size as NIM3 expressed by transfection in COS cells, demonstrating that the cloned NIM3 contains a full-length coding sequence (Fig. 5). In  addition to the major band, several minor bands were observed. We think that these minor bands are probably due to crossreactivity with unrelated proteins and not to alternative NIM3 transcripts, because our extensive cDNA cloning and the RNA blotting experiments failed to reveal any evidence for alternative transcripts.
We then analyzed the immunoblotting signals for RIMs and NIM3 in different brain regions (Fig. 6). In all brain regions, RIM1 and RIM2 were detected as multiple bands consistent with extensive alternative splicing. RIMs were differentially distributed in the various brain regions in a pattern distinct from that of synaptophysin and Rab3A, two synaptic vesicle proteins. Whereas synaptophysin was similarly present in all brain regions, RIMs were more abundant in evolutionarily new rostral brain regions (cortex, cerebellum, olfactory bulb) than in evolutionarily old caudal brain regions (midbrain, hind brain, spinal cord). This difference in expression levels was even more pronounced for NIM3 (Fig. 6). Significant levels of NIM3 could only be detected in the rostral brain regions; we observed no signal in spinal cord, hind brain, or midbrain.
NIM3 Enhances Ca 2ϩ -dependent Secretion in Transfected PC12 Cells-The homology of NIMs to RIMs and their derivation from the same or similar genes (Figs. 1A and 2) suggest that NIMs, like RIMs, may be involved in exocytosis. Such a function would have to be mediated, however, via a different mechanism, since NIMs, in contrast to RIMs, do not bind to Rab3s. To test a possible involvement of NIMs in exocytosis, we measured the effect of overexpression of NIM3 on Ca 2ϩ -dependent secretion in transfected PC12 cells (Fig. 7). For this purpose, we used a system that we and others have extensively employed previously (e.g. see Refs. 18,24,25,32,and 37), namely co-transfection of hGH as a reporter gene with the protein that is being investigated. We co-transfected hGH into PC12 cells with either an empty expression vector or the NIM3 expression vector, stimulated the cells by KCl depolarization, and determined the amount of hGH release under control or stimulation conditions. In previous studies we had shown that under the conditions used, KCl depolarizaiton triggers Ca 2ϩdependent exocytosis that is inhibited by tetanus toxin (32). Measurement of hGH secretion as a function of NIM3 expression revealed that NIM3 had no significant effect on base-line hGH release from unstimulated cells but dramatically en-hanced Ca 2ϩ -dependent exocytosis stimulated with KCl (Fig.  7). This was a surprising result, because in previous studies, we identified several proteins that inhibit release significantly (e.g. truncated syntaxin 1, synaptogyrins, and synaptophysins (32,37)), whereas proteins that enhance secretion have been much less frequently observed. These results suggest that NIM3 functions in exocytosis.
Identification of RIM-BPs Using Yeast Two-hybrid Screens-The modular domain structure of RIMs suggests a potential role as a scaffolding molecule that functionally connects Rab3 to other synaptic proteins. To identify proteins that interact with RIMs, we performed yeast two-hybrid screens. In these experiments, we chose as a bait the C-terminal half of RIM1 that includes the C 2 domains, because this part of RIMs is the most highly conserved (Fig. 1). We screened a rat brain library with this bait and tested all positive clones by retransformation and sequencing. Among 55 positive clones isolated, we observed two homologous SH3 domain proteins, which we named RIM-BP1 and RIM-BP2. RIM-BP1, the protein encoded by the prey vector with the larger insert, was chosen for further study because it interacted with RIM1 and with RIM2, was obtained in multiple independent isolates in the screens, and was expressed at high levels only in brain (see below). Since the initial RIM-BP1 prey clone contained only part of the coding sequence (residues 602-1686), we isolated additional overlapping cDNA clones from a rat brain library that together cover almost the complete coding region (Fig. 8). During the course of this study, a sequence related to RIM-BPs was published as that of PRAX1, a protein interacting with the mitochondrial peripheral benzodiazepine receptor (38). In addition, data banks contain a second protein sequence related to RIM-BPs, KIAA0318 (GenBank TM accession no. AB002316), which is from a random human cDNA clone. PRAX1 is more homologous to RIM-BP1 FIG. 6. Distribution of RIM1, RIM2, and NIM3 in different brain areas. Homogenates from the indicated rat brain areas were analyzed by immunoblotting with antibodies to RIM1, RIM2, NIM3, synaptophysin, and Rab3A as shown. Protein amounts used were approximately equal and were normalized for the synaptophysin signal.

FIG. 7. Effect of NIM3 on Ca 2؉ -triggered exocytosis in transfected PC12 cells.
An hGH expression vector was co-transfected into PC12 cells with an empty control vector or a NIM3 expression vector. Transfected cells were then exposed to control buffer or to depolarizing KCl buffer, and the amount of secreted hGH was determined as a percentage of total expressed hGH for each dish. Results from three independent experiments performed in triplicates were normalized to the amount of stimulated secretion in control cells (100%) and are depicted here as means Ϯ S.E. Only the increase in NIM3-transfected cells of Ca 2ϩ -triggered exocytosis is statistically significant (asterisks; p Ͻ 0.0002) and not the slight change in basal secretion. than to RIM-BP2, and KIAA0318 is more homologous to RIM-BP2 than RIM-BP1, suggesting that they may be the human orthologs of RIM-BP1 and RIM-BP2, respectively.
Sequence analyses revealed that RIM-BPs, PRAX1, and KIAA0318 contain three dispersed SH3 domains and three contiguous fibronectin type III repeats (shown on red and blue backgrounds, respectively, in Fig. 8). These domains are the most closely related sequences in RIM-BPs, PRAX1, and KIAA0318 (Fig. 8) and are flanked by highly charged sequences that are less well conserved. The sequences of the SH3 and fibronectin type III domains are rather atypical, which is probably the reason why they were overlooked in the initial analysis of the PRAX1 sequence (38). The sequences outside of the SH3 domains and fibronectin type III repeats are rich in arginine, lysine, and glutamic acid; these sequences contain strings of charged residues with up to 13 consecutive glutamic acid residues.
Although RIM-BP1 and PRAX1 are very similar, there are stretches of significant sequence divergence between the two proteins (e.g. residues 1166 -1241 in PRAX1; see Fig. 8). The sequence differences between RIM-BP1 and PRAX1 exceed those we have typically observed between the rat and human homologs of many brain proteins (e.g. see synaptophysin (39)), raising the question of whether RIM-BP and PRAX are true orthologs or just homologs. Analysis of multiple rat cDNA clones encoding RIM-BP1 identified a 61-residue sequence that is alternatively present or absent in three independent cDNA clones each, suggesting that it is differentially spliced (Fig. 8). The human gene for PRAX1 was sequenced during the genome project (GenBank TM accession no. AC004687) and shown to be a relatively small gene (approximately 25 kilobases) on chromosome 17. Curiously, the gene sequence reveals that the 63-residue sequence that is alternatively spliced in RIM-BP1 is in the middle of a large exon. This raises the question of how this sequence can be alternatively spliced if it is in the middle of an exon. In contrast, a random human cDNA sequence that also reports the PRAX1 sequence (KIAA0612; GenBank TM accession no. AB014512), lacks residues 191-250. Here the variable sequence is precisely encoded in a single exon, indicating alternative splicing. Together with the limited sequence homology between RIM-BP1 and PRAX1, these findings raise the possibility that RIM-BP1 and PRAX1 may be closely related but do not actually represent orthologs. Furthermore, RIM-BP1 is highly expressed only in the brain (see below), while the peripheral benzodiazepine receptor is a ubiquitously functioning protein outside of the brain (40). Because of these uncertainties, we will continue to refer to the RIM-binding proteins as RIM-BP1 for the purpose of the current discussion.
RIM-BP1 Is Specifically Expressed in the Brain-We performed RNA blotting experiments to determine which tissues express RIM-BP1 (Fig. 9). A single large mRNA was found that was expressed at high levels only in brain. In addition, a smaller weaker signal was found in testis, which we interpret as an artifact because we have observed a similar signal with a number of unrelated probes (data not shown).
The Second SH3 Domains of RIM-BPs Bind to a Proline-rich Sequence in RIM-We identified RIM-BPs as RIM-binding proteins by yeast two-hybrid screens. To confirm this interaction, we used GST pull-down experiments (Fig. 10). Rat brain homogenates were dissolved in SDS and then quenched in Triton X-100 in order to solubilize and renature RIMs because RIMs are not solubilized from the active zone in the absence of denaturing detergents. The homogenate was then incubated with GST-RIM-BP1 fusion proteins containing either the highly charged sequence between the third fibronectin III domain and the second SH3 domain or the second SH3 domain (Fig. 8). Immunoblotting of the bound proteins with antibodies to RIMs showed that only the SH3 domain was capable of capturing RIMs from brain homogenates, thereby confirming the interaction observed in the yeast two-hybrid system (Fig.  10).
We then turned to yeast two-hybrid assays to identify which sequence in RIMs binds to RIM-BP1. Comparison of the insert sequences of the RIM-BP1 and RIM-BP2 prey clones showed that they overlap only in the highly conserved second SH3 domain, suggesting that the interaction of this domain with RIMs led to the isolation of RIM-BPs in the yeast two-hybrid screens (Fig. 8). To confirm this, we first analyzed the interaction of the RIM-BP1 prey vector with a series of prey constructs encoding the C terminus of RIM2. Assays using the nonquantitative ␤-galactosidase filter assay identified a short sequence in RIM2 (residues 1097-1149) that was involved in the interaction (Table I; Fig. 1A). Quantitative liquid ␤-galactosidase assays confirmed that this sequence is sufficient for binding RIM-BP1 (Table II). Inspection of the RIM sequences in this area reveals that it contains a single conserved classical SH3binding motif (sequence RQLPQ(L/V)P), suggesting that RIM-BP1 binds to RIMs at this position. It is interesting that this sequence is present in the short region between the two alternatively spliced sequences in the C terminus of the RIMs and is the only conserved sequence in this short region (Fig. 1A). DISCUSSION Neurotransmitter release is one of the most tightly regulated processes in biology, because the precise control of synaptic signaling is fundamentally important for information processing in the brain. Rab3 is a GTP-binding protein of synaptic vesicles that regulates neurotransmitter release at the synapse (reviewed in Ref. 41). Mechanistically, Rab3 is thought to function by interacting with effector proteins in a GTP-dependent manner. Two such effectors are known, RIM1 and rabphilin (16 -18). Current data suggest that RIM1 is the more important of these two effectors because it appears to be essential for synaptic function, while rabphilin is not (26), 2 and because the strategic localization of RIM1 at the active zone would allow RIM to mediate a GTP-dependent tethering of synaptic vesicles at the active zone (18). We now demonstrate that RIM1 belongs to a larger family of proteins composed of RIM1 and RIM2 and the related NIM2 and NIM3, and we identify a family of RIMinteracting proteins called RIM-BPs. Furthermore, we show that RIM2, similar to RIM1, specifically binds to Rab3 in a GTP-dependent manner, while NIMs do not; nevertheless, NIMs are probable regulators of exocytosis, since NIM3 overexpression greatly facilitates exocytosis. Together, these results expand our view of the components and complexity of the Rab3-dependent regulation of neurotransmitter release.
The present study describes the identification and molecular characterization of a family of novel proteins related to RIM, RIM2, NIM2, and NIM3, and of a separate class of brain proteins that bind to RIMs, named RIM-BPs. These data show that the RIM/NIM family contains at least three closely related genes: RIM1, RIM2, and NIM3, which produce transcripts encoding at least four proteins (RIM1, RIM2, NIM2, and NIM3). Analysis of human genomic sequences containing the RIM1 and NIM3 genes failed to identify coding sequences for NIM1 and RIM3 proteins, respectively, which would suggest that only the RIM2/NIM2 gene encodes both variants. Rabphilin and associated proteins (DOC2, NOC) are more distantly related members of this gene family because they also have zinc finger and/or C 2 domains; however, these domains are less homologous to those of RIMs and NIMs than they are to each other (schematically shown in Fig. 2). Thus, RIMs, NIMs, rabphilin, DOCs, and NOCs form a superfamily of proteins composed of combinations of Rab3-interacting zinc finger domains and C 2 domains. Overall, the RIM/rabphilin protein family is surprising in the way in which C 2 domains are coupled in some members of this family to Rab3-interacting zinc fingers (i.e. in RIMs and rabphilin), while in other family members (i.e. in NIMs and DOC2s) they are present without any connection to Rab3s.
An interesting feature of RIMs is their conserved extensive alternative splicing with a large number of variants (as opposed to a single splice variant in rabphilin; see Refs. 17 and 42). This suggests that RIMs are not only diversified in the expression of two genes but that the exact shape of these isoforms can differ. Although we have not investigated whether alternative splicing of RIMs is regulated in brain and whether it results in functionally different proteins, it is striking that the alternatively spliced sequences are often strategically located next to identified domains (e.g. splice site 2 immediately N-terminal of the PDZ domain, or splice sites 4 and 5, which flank the binding site for RIM-BPs). More importantly, the position and inserts of splice sites 4 and 5 are conserved between RIM1 and RIM2, with inserts that can be very large (almost 200 residues). As a result, this alternative splicing is expected to change the proteins produced considerably.
A further interesting feature of RIMs is their interaction with RIM-BPs. The modular domains structure of RIMs and NIMs indicates a possible interaction with multiple other proteins in addition to Rab3s in the case of RIMs. With the RIM-BPs, our study identifies the first of such potential interacting partners. Two related but distinct RIM-BPs were independently isolated in yeast two-hybrid screens, and the interaction of RIM-BP1 with RIMs was verified in GST pull-down experiments and quantitative yeast two-hybrid assays. The fact that RIM-BPs use an SH3 domain to bind to RIMs and that the binding site in RIMs is located in the small island of constant sequence located between large alternatively spliced sequences (sites 4 and 5; Fig. 1A) suggests that the interaction may be biologically important. RIM-BPs have an unusual domain organization with fibronectin type III repeats that are rarely found in intracellular proteins, indicating a role as a scaffolding protein that would agree well with a localization to the active zone. A rat multitissue RNA blot was hybridized at high stringency with a cDNA probe from the C terminus of RIM-BP1 (top) and a ubiquitously expressed control probe (bottom). The RIM-BP1 blot is shown as an overexposure to illustrate that there are no cross-hybridizing bands outside of the brain except for testis, where a smaller mRNA is detected (asterisk).
FIG. 10. Analysis of RIM binding to RIM-BP1 using GST pulldowns. Glutathione-agarose beads containing GST fusion proteins of the central region of RIM-BP1 (residues 652-1270) or of the C-terminal region of RIM-BP1 (residues 1271-1626) were incubated in the presence and absence of rat brain homogenates that had been first solubilized in SDS and then quenched in Triton X-100 because RIMs cannot be solubilized from the active zones in nondenaturing detergents but, once solubilized, remain in solution. Bound proteins were then analyzed with an antibody to RIMs.  Fig. 1A) were co-transformed with prey clones encoding residues 1271-1625 of RIM-BP into L40 yeast cells. Cells harboring both plasmids were analyzed for activation of ␤-galactosidase transcription using a filter assay as described (28,29 Table I for the RIM2 sequences encoded by these clones) and with the following prey clones: pVP16-RIM-BP1 and -RIM-BP2 encoding residues 602-1625 and 1271-1625 of RIM-BP, respectively; pVP16-EHSH encoding the C-terminal two SH3 domains of EHSH/intersectin (42); and the pVP16 prey vector. Single yeast colonies were grown in liquid culture for 40 h and harvested, and the ␤-galactosidase activity and protein content of each culture were determined as described (18,28,29). Data shown are means Ϯ S.E. from triplicate determinations in nmol/min and mg of protein. A function of RIMs in exocytosis is likely because these proteins interact with Rab3 in a regulated manner. However, no such supposition can be made for NIMs. Here the only clue to a possible function in neurotransmitter release is their brain-specific expression, similarity to RIMs outside of the Rab3-binding domains, and content of a C 2 domain that is often found in exocytotic proteins. To address the question of whether or not NIMs are functionally related to exocytosis, we examined the effect of NIM3 overexpression on Ca 2ϩ -triggered secretion in transfected PC12 cells (Fig. 7). We found that NIM3 induces a dramatic enhancement of secretion, confirming a function for this protein as well in exocytosis. Thus, it appears likely that all members of the RIM/NIM superfamily are involved in the control of exocytosis.
Our results also raise a number of questions. On an immediate, more technical level, one wonders if the RIM1 gene also produces a NIM1 product, if the NIMs are generated by independent promoters within the RIM genes, and whether there is a spatially differentiated transcription of RIMs and NIMs in brain; i.e. are their promoters differentially regulated? Another more technical question is whether the interaction of RIM-BP with RIM is specific, since many SH3 domain interactions are known to be quite promiscuous in vitro. Puzzling here, giving cause to caution, is the prior description of an ortholog or close homolog of RIM-BP1 as a protein called PRAX-1 that binds to the peripheral benzodiazepine receptor, a ubiquitous mitochondrial protein. However, three findings of our experiments support the notion that RIM-BPs are physiological interactors of RIMs. First, we independently isolated two different isoforms of RIM-BPs in a yeast two-hybrid screen with RIM, although RIM-BPs are not abundant proteins. We have performed more than 20 yeast two-hybrid screens with the same library and isolated other SH3 domain proteins with other baits, but never RIM-BPs; thus, these proteins are not simply "sticky." This suggests a high degree of specificity and argues against a promiscuous interaction. Second, the distribution of RIM-BP1, which is primarily expressed in the brain, resembles that of RIMs much more than that of the peripheral benzodiazepine receptor, which, as indicated by its name, is not brain-specific. Furthermore, the modular structure of RIM-BPs with SH3 and fibronectin type III repeat domains would fit well into a role as an organizer of the active zone. Third, we confirmed an interaction of RIM-BP1 with RIM biochemically and localized the binding site in RIMs by quantitative yeast two-hybrid assays. The fact that an SH3 domain of RIM-BPs binds to a conserved PXXP sequence in RIMs gives credence to the interaction, because it conforms well to the known general properties of SH3-mediated interactions.
On a conceptual level, our results raise more important questions. At this point, we do not know the precise functions of the proteins described here beyond the fact that RIMs interact with Rab3, an essential regulator of neurotransmitter release, and that NIM3 enhances calcium-triggered exocytosis in PC12 cells. However, preliminary results with RIM1 knockout mice indicate that these mice are severely abnormal, 2 providing further evidence that Rab3 coupling to RIMs is functionally important in regulating neurotransmitter release. Nevertheless, these results provide no definitive insight into the how and what. In this respect, the identification of RIM-BPs is the first step in the direction of a mechanistic understanding. The composition of RIM-BPs with multiple fibronectin type III repeats and SH3 domains is consistent with a role in the presynaptic cytomatrix next to the active zone as a scaffolding mole-cule and as a cytoskeletal organizer. However, RIMs are likely to have additional binding partners, thereby creating a molecular net of protein-protein interactions. Future studies will have to investigate the nature of these proteins and the architecture of their interactions.