Retinal Targets for Calmodulin Include Proteins Implicated in Synaptic Transmission*

Ca2+ influxes regulate multiple events in photoreceptor cells including phototransduction and synaptic transmission. An important Ca2+ sensor inDrosophila vision appears to be calmodulin since a reduction in levels of retinal calmodulin causes defects in adaptation and termination of the photoresponse. These functions of calmodulin appear to be mediated, at least in part, by four previously identified calmodulin-binding proteins: the TRP and TRPL ion channels, NINAC and INAD. To identify additional calmodulin-binding proteins that may function in phototransduction and/or synaptic transmission, we conducted a screen for retinal calmodulin-binding proteins. We found eight additional calmodulin-binding proteins that were expressed in the Drosophila retina. These included six targets that were related to proteins implicated in synaptic transmission. Among these six were a homolog of the diacylglycerol-binding protein, UNC13, and a protein, CRAG, related to Rab3 GTPase exchange proteins. Two other calmodulin-binding proteins included Pollux, a protein with similarity to a portion of a yeast Rab GTPase activating protein, and Calossin, an enormous protein of unknown function conserved throughout animal phylogeny. Thus, it appears that calmodulin functions as a Ca2+ sensor for a broad diversity of retinal proteins, some of which are implicated in synaptic transmission.

Influx of Ca 2ϩ , in Drosophila photoreceptor cells, occurs in two spatially and temporally distinct phases. The initial Ca 2ϩ influx is activated in response to light stimulation of rhodopsin and is mediated by the TRP and TRPL cation channels situated in the microvillar portion of the photoreceptor cells, the rhabdomeres (reviewed in Ref. 1). TRP-and TRPL-dependent Ca 2ϩ entry contributes both to the depolarization of the photoreceptor cells as well as to Ca 2ϩ -mediated negative-feedback regulation. A second phase of Ca 2ϩ entry in fly photoreceptor cells presumably occurs in the active zones in the presynaptic terminals and is the result of opening of voltage-gated Ca 2ϩ channels. Ca 2ϩ influx via the voltage-gated channels leads to fusion of synaptic vesicles with the plasma membrane and release of neurotransmitter (reviewed in Ref. 2). Thus, fluxes in Ca 2ϩ levels regulate multiple processes that are critical for vision including negative feedback regulation, termination of the light-induced current, and synaptic transmission. Nevertheless, identification of the Ca 2ϩ sensors which regulate these processes is quite incomplete.
Some of the proteins that mediate the various Ca 2ϩ -regulated processes in fly photoreceptor cells are likely to do so through direct interaction with Ca 2ϩ . These include protein kinase C, which functions in adaptation and termination of the photoresponse (3,4), and synaptotagmin, a synaptic vesicle protein that may be one of the Ca 2ϩ sensors that functions in release of neurotransmitter (reviewed in Ref. 5). Other Ca 2ϩbinding proteins implicated in synaptic transmission are the protein phosphatase calcineurin (6), which binds both Ca 2ϩ and Ca 2ϩ /calmodulin, and rabphillin (7,8), a peripheral membrane protein that binds Rab3, a small GTPase that is associated with synaptic vesicles (reviewed in Ref. 2).
Many additional retinal proteins may be controlled indirectly by Ca 2ϩ through association with the Ca 2ϩ -binding regulatory protein calmodulin. A number of studies have indicated that there exists a plethora of retinal calmodulin-binding proteins in vertebrates (9 -11) several of which, such as the cGMPgated ion channel, have been identified (12). Calmodulin is expressed at very high levels in Drosophila photoreceptor cells (13) and mutations and experimental conditions that lower the intracellular calmodulin levels result in defects in negative feedback regulation and termination of the photoresponse (14 -17).
NINAC p174 and p132, proteins that consist of protein kinase and myosin head domains (18), are the major calmodulinbinding proteins expressed in Drosophila photoreceptor cells (13). Disruption of the NINAC/calmodulin interaction causes a decrease in the intracellular concentration of calmodulin, a dramatic alteration in the normal spatial distribution of calmodulin and consequently, defects in adaptation and termination of the photoresponse (13,14). Calmodulin also binds to the light-sensitive ion channels, TRP and TRPL (19,20), and disruption of the calmodulin association with TRPL causes an impairment in inactivation (17). Two other retinal proteins shown to bind calmodulin are RIC, a protein with homology to RAS (21), and INAD (22), a PDZ containing scaffold protein which binds many of the proteins required in phototransduction (reviewed in Ref. 23).
Studies in a variety of organisms indicate that calmodulin functions in synaptic transmission. Calmodulin binds to the synaptic vesicle protein synapsin I, from vertebrates and Drosophila (24 -26), as well as to mammalian Rab3A (27). Furthermore, association of Rab3A with synaptic membranes is disrupted in vitro by Ca 2ϩ /calmodulin (27). Several calmodulindependent protein kinases appear to have roles in synaptic transmission. Ca 2ϩ /calmodulin-dependent protein kinases I and II (CaM kinase I and II) appear to function in synaptic transmission in a variety of neurons and are known to phosphorylate several components associated with synaptic vesicles (reviewed in Ref. 28), such as the synapsin I-like protein expressed in Drosophila photoreceptor cells (29). Additionally, inhibition of myosin light chain kinase (MLCK) 1 disrupts synaptic transmission in rat sympathetic neurons (30). Further evidence that calmodulin may also function in synaptic transmission in the Drosophila visual system is that a reduction in calmodulin levels in photoreceptor cells affects a feature of the electroretinogram that emanates from activity post-synaptic to the photoreceptor cells (13).
Due to the apparent roles of calmodulin in multiple processes in Drosophila vision, we sought to identify potential targets for retinal calmodulin in addition to the five previously described.
In the current report, we identified eight additional calmodulin-binding proteins that are expressed in the Drosophila retina, most of which are enriched in the retina. One group consists of Drosophila proteins related to components previously implicated in synaptic transmission but not known to bind calmodulin. These include proteins with domains similar to the Rab3 GDP/GTP exchange factors (31,32) and a homolog of diacylglycerol-binding proteins conserved from Caenorhabditis elegans to vertebrates, referred to as UNC13 and mUNC13 (33,34). Mutations in the C. elegans UNC13 were isolated in a screen for mutants with uncoordinated movement (35) and subsequent studies suggest that UNC13 functions in synaptic transmission (36). Since we also identified CaM kinase I, CaM kinase II, MLCK, and calcineurin in the Drosophila retina, a minimum of six of the eight calmodulin-binding proteins identified in the current screen are related to components which appear to function in synaptic transmission. A seventh protein, Pollux, has limited homology to two yeast Rab GTPase activating proteins (37) and the last protein, Calossin, is a novel calmodulin-binding protein (Ͼ4100 amino acids) of unknown function conserved throughout animal phylogeny. Therefore, calmodulin may regulate a greater repertoire of proteins that function in synaptic transmission than previously recognized.

EXPERIMENTAL PROCEDURES
Isolation of Clones Encoding Calmodulin-binding Proteins-The retinal expression library used to screen for calmodulin-binding clones was prepared using poly(A) ϩ RNA isolated from adult fly retinas (ZAP library; gift from Charles Zuker, UCSD; cDNAs were non-directionally inserted into the EcoRI site). This was not a subtracted retinal-specific library. The screen for calmodulin-binding proteins was performed by probing 1 ϫ 10 6 plaques with 125 I-calmodulin at 10 7 cpm/ml in 10 mM imidazole (pH 7.4), 150 mM KCl, 0.1 mM CaCl 2 , 0.2% Tween 20, and 5% bovine serum albumin. The positives were counter-screened with a trpl DNA probe to cull out the trpl clones (since TRPL binds very well in overlay assays) and the remaining positive plaques were rescreened twice. The pBluescript SKϪ clones containing the cDNAs were excised by co-infection with 10 7 plaque forming units/ml of R408 helper phage as described (Stratagene protocol) and organized into groups by DNA cross-hybridization and sequencing one end of representative members of each group. After eliminating positives that corresponded to previously described retinal calmodulin-binding proteins (TRPL, NINAC, and RIC), a total of 45 positives remained that fell into eight groups. The complete sequences corresponding to three of the remaining eight, calmodulin-dependent protein kinase II, myosin light chain kinase, calcineurin (at 14B), were previously reported. To obtain addition cDNA sequences encoding the remaining five (CaM kinase I, CRAG, PLX, dUNC13, and CALO), the retinal ZAP library and an adult head ZAP library were rescreened with DNA probes.
DNA Sequencing-DNA sequencing was performed at the DNA Analysis Facility at the Johns Hopkins University School of Medicine using the fluorescent dideoxy terminator method of cycle sequencing (38) in conjunction with a Perkin-Elmer, Applied Biosystems Division (PE/ABd) 373a automated DNA sequencer (39).
Chromosomal in Situ Hybridizations-In situ hybridizations to salivary gland polytene chromosomes from late third instar larvae were performed with digoxigenin-labeled DNA probes. The probes were prepared by boiling 100 -500 ng of DNA (in H 2 O) for 5 min, cooling quickly on ice, and adding 2 l of 10 ϫ DIG Labeling Mixture (Boehringer Mannheim catalog number 1277-065), 2 l of 10 ϫ hexanucleotide mixture (Boehringer Mannheim catalog number 1277-081), and Klenow fragment in a total volume of 20 l. After incubating at room temperature for 3-16 h, 20 g of carrier DNA was added and the DNA was ethanol precipitated and resuspended in 75 l of hybridization buffer (0.6 M NaCl, 50 mM sodium phosphate buffer (pH 6.8), 1 ϫ Denhardt's (0.02% bovine serum albumin, 0.02% Ficoll, 0.02% polyvinylpyrrolidone), and 5 mM MgCl 2 ). After hybridizing to polytene chromosomes, the probes were detected using a 1:500 dilution of anti-digoxigenin-POD (Boehringer Mannheim catalog number 1207-733).
RNA Blots-Tissue from various developmental stages (0.5 g each, w 1118 strain) were collected and total RNAs were isolated with RNAzol (Tel-Test) according to the manufacturer's instructions. Poly(A) ϩ RNA was purified from total RNA using an mRNA isolation kit (Boehringer Mannheim). 50 g of total RNA or 5 g of poly(A) ϩ RNA from each tissue was fractionated on formaldehyde-agarose gels (1.2% agarose), transferred to a nylon membrane, probed with 32 P-labeled cDNA fragments and the signals were detected using a PhosphorImager (Fugi). The cDNA fragments used for probing the blots encoded the following amino acids indicated in parentheses: dUNC13 (642-1174), CaMK1 (80 -405), PLX (164 -1379), CRAG (388 -1209), CALO (1415-4118). The blots were stripped by incubating at 95°C in 0.5% SDS and re-probed with the full-length RP49 cDNA probe.
Generation of GST Fusion Proteins-All the fusion proteins were expressed as glutathione S-transferase (GST) fusions using pGEX vectors (Pharmacia Biotech Inc.). Unless specified otherwise, the fragments used for subcloning were generated by polymerase chain reaction. The dUnc13 fragments were inserted between the BamHI/EcoRI sites of pGEX-5X-3 and the Crag sequences were subcloned into pGEX-5X-2 as XbaI/XhoI fragments. The fusions constructed using Lyncein, the bovine homolog of plx, were generated using a Lyncein cDNA (B61) isolated from a bovine retinal ZAP library (gift of Donald Zack, JHU-SOM). pPollux.9 and pLyncein-L1 through pLyncein-L6 were created by inserting BamHI/EcoRI fragments into pGEX-5X-3. pLyncein-L7 was generated by digesting B61 with XhoI and cloning the fragment nondirectionally into pGEX-5X-3. pLyncein-L8 was obtained by digesting B61 with XhoI and EcoRI and cloning into pGEX-5X-3. pPollux.3 was generated by cloning the original plx isolate, dm265, into the XhoI/EcoRI sites of pGEX-5X-3 and then digesting with BamHI and religating. pPollux.2 was obtained by subcloning between the BamHI/ EcoRI sites of pGEX-5X-3. The amino acid residues encoded by the fragments in the following plasmids are indicated in parentheses: pCrag.1 (388 -664), pCrag. Calmodulin Binding Assays-The fusion proteins were expressed in freshly transformed Escherichia coli BL21(DE3) cells as described (Pharmacia Biotech Inc.). To perform the calmodulin overlay assay, 200 l of bacterial cells, induced with isopropyl-1-thio-␤-D-galactopyranoside, were pelleted and boiled in 20 l of 2 ϫ SDS sample buffer for 5 min. The samples were then fractionated by SDS-PAGE (10% gel) and transferred to polyvinylidene difluoride membranes. Biotinylated calmodulin (Life Technologies) was overlaid onto the filter and detected as described by the manufacturer. To perform overlays in the absence of free Ca 2ϩ , 5 mM EGTA was included in all buffers and no CaCl 2 was added. All procedures were carried out at room temperature.

Summary of Screen for Retinal Calmodulin-binding Proteins-To identify calmodulin-binding proteins expressed in
the Drosophila retina, we probed a retinal expression library with 125 I-labeled calmodulin. The positive clones isolated in the screen were grouped by DNA cross-hybridization and sequencing and represented cDNAs encoding 11 different calmodulinbinding proteins. Three of the positives corresponded to previously reported calmodulin-binding proteins: NINAC (13), TRPL (19), and RIC (Table I; Ref. 21). Two other calmodulinbinding proteins known to be expressed in the retina, TRP (20) and INAD (22), were not identified in the screen. TRP binds only to unmodified or biotinylated calmodulin and not to 125 Ilabeled calmodulin and INAD binds relatively weakly in calmodulin overlay experiments.
Eight calmodulin-binding proteins were identified which were not previously reported to be expressed in the Drosophila retina. The full-length sequences corresponding to three of the calmodulin-binding proteins have been described. These corresponded to two calmodulin-dependent protein kinases, MLCK (40,41) and CaM kinase II (42,43), as well as to one of the two previously described calcineurin proteins (Table I; Ref. 44). We also found that a third calmodulin-dependent protein kinase was expressed in the Drosophila retina, CaM kinase I. No CaM kinase I had previously been reported from any invertebrate raising the possibility that this protein kinase was specific to vertebrates. Nevertheless, the Drosophila CaM kinase I (Table  I) was highly related, ϳ60% identical over 349 amino acids, to vertebrate CaM kinase I (Fig. 1A) (45). The remaining four calmodulin-binding proteins were not known to bind calmodulin prior to the current work (see below). Thus, it appears that a minimum of 13 calmodulin-binding proteins are expressed in the visual system.
Proteins with Domains Similar to a Rab3 GEP and a Rab GAP Bind Calmodulin-At least two of the four novel calmodulin-binding proteins share similarities to components implicated in synaptic transmission. One of these proteins (1441 residues), we refer to as CRAG (calmodulin-binding protein related to a Rab3 GDP/GTP exchange protein; Table I) due to its similarity to a domain in the recently identified rat Rab3 GDP/GTP exchange protein (rRab3 GEP) (31) and the C. elegans homolog, AEX-3 (32), which has been implicated in synaptic vesicle release (Figs. 1B and 2A). The sequences of AEX-3 and the rRab3 GEP were published contemporaneously and were therefore not directly compared. AEX-3 and the rRab3 GEP (1409 and 1602 amino acids, respectively) contained three regions of homology, the first of which (ϳ500 residues) was conserved in CRAG, AEX-3, rRab3 GEP (CAR domain; Fig. 2A) and the human homolog, MADD (death domain MAP kinase activator; Ref. 46). The latter two regions, were conserved in AEX-3 and rRab3 GEP (AR1 and AR2), but not CRAG, and were shorter (ϳ100 and 300 residues, respectively) than the CAR homology. The CAR domain in CRAG was ϳ36 identical over 321 amino acids (residues 95-415) to either rRab3 GEP or AEX-3. In addition, there was weak homology (16%) in the flanking sequences that extended the CAR domain in CRAG to residues 73-490. The C-terminal ϳ800 residues of CRAG did not share significant primary amino acid sequence homology with the rRab3 GEP, AEX-3, or any other protein in the data banks.
POLLUX (PLX) is a protein previously reported to be 732 amino acids in length and required for viability (47). The protein is predicted to have a transmembrane domain and a leucine zipper ( Fig. 2B; Ref. 47). We found that PLX was 1379 amino acids in length and the formerly assigned initiator methionine corresponded to residue 648 (Fig. 1C). A protein related to PLX was TBC1 (48), a mouse protein which had homology to the majority of PLX. The region in PLX that contained the greatest similarity to TBC1 was a 337-amino acid segment (51% identity, residues 676 -1012) that included the putative transmembrane domain (Figs. 1C and 2B). Of particular interest, the region most highly conserved between PLX and TBC1 included a 153-amino acid domain (residues 811-963) that displayed moderate homology to the yeast Rab family GTPase-activating proteins, GYP6 or GYP7 (37). GYP7 was ϳ29% identical to this domain in either PLX or TBC1; however, if two gaps of 18 and 36 amino acids were introduced in PLX and TBC1, the 29% homology extended to over 222 a These proteins were determined to be enriched in the visual system by spatially localizing the proteins on sections of adult heads (18,20,69,70). The column, cDNAs, lists the number of positives that fell into each class. The filters were counter-screened with trpl DNA to cull out TRPL positives.
b Approximately 50 isolates cross-hybridized with trpl and five were subjected to DNA sequence analysis. All five showed 100% identity to trpl. c ND, not determined. The asterisk indicates the methionine (residue 648, indicated in bold) that was assigned as the first residue in PLX in a previous report. A histidine (residue 1286) was identified as a glutamine in the prior sequence. D, dUNC13 (dUNC; Y17921 and Y17922) compared with rat mUNC13-1 (mUNC; U24070). dUNC13 is alternatively spliced and the deduced amino acid sequences encoded by the two RNA isoforms, dUNC13-A (Y17921) and dUNC13-B (Y17922), differ at the N-terminal end. dUNC13-A (dUNC-A), which contains the CBS, and dUNC13-B (dUNC-B) contain Ͼ88 and Ͼ508 amino acids unique to each isoform. Neither initiator codon nor the common stop codon were identified. The lines above the sequences indicate the CBS, the C1 and two C2 domains (C2-1 and C2-2). E, CALO (Y17920) shown with a related protein in C. elegans (cCALO; AF003140) and a variety of mouse ESTs that were artificially fused together to create one sequence (mCALO). Two cysteine-rich domains (CRD1 and CRD2) are indicated. The initiator methionine in CALO was not identified. amino acids (742-963). This ϳ200 amino acid sequence corresponded to the domain previously referred to as a TBC domain due to its similarity to segments in the TRE-2 oncogene and the yeast regulators of mitosis, BUB2 and CDC16. Neither this nor any other domain in PLX shared primary amino acid sequence homology with the recently described Rab3 GTPase activating protein (49).
A Calmodulin-binding Protein Homologous to a Diacylglycerol-binding Protein Implicated in Synaptic Transmission, UNC13-A third protein, not previously known to bind calmodulin, was a Drosophila homolog of UNC13 (dUNC13 ; Table I), a diacylglycerol-binding protein which may be required for release of neurotransmitter from the presynaptic terminal (36,50). dUNC13 was expressed as at least two alternatively spliced forms encoding proteins of Ͼ1304 (dUNC13A) and Ͼ1724 (dUNC13B) amino acids (Fig. 1D). dUNC13A and dUNC13B shared a common C-terminal region of Ͼ1216 amino acids and differed due to unique N-terminal sequences (Ͼ88 and Ͼ508 residues, respectively). dUNC13 contained extensive homology (Ͼ68%) with the C. elegans UNC13 and rat homologs (mUNC13) beginning in the unique region of dUNC13A and extending over the entire region common between both iso- ). UNC13 and mUNC13-1 shared a similar level of homology over the same region and are only weakly related over the N-terminal ϳ500 amino acids. The 508 amino acids specific to dUNC13B were not homologous to the UNC13 proteins or any proteins in the data banks.
Features common between dUNC13 and other members of the UNC13 family include strong homology to two conserved sequence motifs, C1 and C2, originally recognized in various protein kinase C isoforms (51). A large variety of other signaling proteins, such as RAF, diacylglycerol kinase, RAS GTPaseactivating protein, synaptotagmin, and phoshopholipase C, were subsequently found to contain these domains (see Refs. 52 and 53). C1 domains typically bind diacylglycerol and many C2 domains are Ca 2ϩ -binding regulatory domains. Some C2 domains also bind phospholipids and do so in a Ca 2ϩ -dependent manner. Other C2 domains confer Ca 2ϩ dependence to functions, such as protein kinase activity, mediated by domains distinct from C2. Biochemical analyses of UNC13 demonstrates that it is a bona fide Ca 2ϩ -dependent phorbol ester-binding protein (52). The putative C1 domain in dUNC13 (residues 182-232) included six invariant cysteines (Fig. 1D) as well as a seventh cysteine conserved among all UNC13 proteins. Overall, the C1 domain was 92% identical to the corresponding region in mUNC13 (34). The two C2 domains present in each of the three other UNC13 proteins (C2-1 and C2-2) were also found in dUNC13. C2-1 (residues 299 -393) and C2-2 (residues 1170 -1264) were 76 and 67% identical with the same motifs in mUNC13-1.
Calossin, a Novel Calmodulin Protein Conserved throughout Animal Phylogeny-The fourth novel calmodulin-binding protein is referred to as Calossin (CALO) due to its interaction with calmodulin and colossal molecular mass (predicted Ͼ450 kDa). Several overlapping cDNAs were obtained resulting in the identification of a single open reading frame encoding Ͼ4118 amino acids (Fig. 1E). Several hydrophobic regions were predicted according to a computer algorithm (54); however, it is unclear if any is sufficiently long to span a lipid bilayer. CALO was related to a predicted C. elegans protein (cCalossin) of similar size (3864 residues) which was identified as part of the C. elegans Genome Sequencing Consortium. The homology between CALO and cCALO was not uniform but concentrated in several domains. The longest continuous region of identity began at amino acid 2460 and extended ϳ1650 residues to near the C terminus. In addition, there were two shorter stretches of similarity between residues 604 and 1150. The highest levels of identity (each ϳ70%) were in three ϳ50 -100 amino acid regions: 1) residues 604 -649; 2) residues 2587-2638; 3) residues FIG. 1-continued 3276 -3380. The first two of these conserved regions were cysteine-rich domains, CRD1 and CRD2, respectively (Fig. 1E), that resembled different classes of zinc finger domains (55). CRD1 was most similar to the zinc finger family defined by Requiem, a protein required for apoptosis (56), while CRD2 shared features equally well with several families of zinc family proteins and could not be included within a single group.
Several mouse and human expressed sequence tags (ESTs) were also related to CALO. In general, CALO was more highly related to the mammalian ESTs (mCalossin) than to cCALO. In some stretches, the homology to mCALO was twice as high as to cCALO (e.g. residues 2460 -2578; 67 and 32% identity, respectively). Thus, although CALO was not related to any protein of known function, it appeared to be conserved from C. elegans to humans.
Expression of mRNAs Enriched in Adult Eyes-The mRNAs encoding each of the calmodulin-binding proteins described above appeared to be expressed in the retina since the cDNAs were isolated by screening a retinal expression library with calmodulin. To ascertain whether the mRNAs were enriched in the adult visual system, we probed Northern blots containing equal amounts of RNA prepared from heads of wild-type adults and a mutant strain, sine oculis, which fails to develop compound eyes. In addition, RNA isolated from selected developmental stages was included on the blots. We found that the mRNAs encoding each of the four of the novel calmodulinbinding proteins, CRAG, PLX, dUNC13, and CALO, were expressed at the highest levels in the adult (Fig. 3). CaM kinase I mRNAs were also readily detected in the adult, as well as at similar levels during the third instar larval and pupal periods. Although, all the genes were expressed at the highest levels in the adult, after exposing the blots for longer durations, weak signals were detected at earlier developmental stages including embryos (data not shown).
Of particular significance here, the mRNAs encoding all four novel calmodulin-binding proteins were expressed at 3-10-fold higher levels in wild-type heads as compared with sine oculis heads ( Fig. 3 and Table I). Thus, each of these mRNAs appeared to be enriched in the visual system relative to other portions of the adult head. Although CaM kinase I was only slightly enriched in wild-type heads (2-fold), an adult-specific ϳ3.5 kb mRNA was not detected in sine oculis heads. Consequently, this CaM kinase I mRNA appeared to be eye-specific.
Mapping Calmodulin-binding Sites-To confirm that dUNC13, CRAG, and PLX associate with calmodulin and to map the binding sites, we performed gel overlay and pull-down assays. The overlap between the two original dUNC13 calmodulin-binding clones included amino acids 64 -229 (Fig. 4A). To further map the calmodulin-binding site(s), we generated several GST-dUNC13 fusion proteins and performed gel overlay experiments by fractionating the fusions by SDS-PAGE, transferring them to membranes, and probing with biotinylated calmodulin in the presence of Ca 2ϩ or the Ca 2ϩ chelator, EGTA. Calmodulin-binding was detected in the presence of Ca 2ϩ but not EGTA (Fig. 5A). Furthermore, the calmodulinbinding site mapped to residues 64 -88, a portion of dUNC13 specific to the unique N-terminal domain of dUNC13A (Fig.  1D). The absence of a calmodulin binding signal with the fusion missing residues 64 -88 was not due to absence of the protein since the appropriate size GST-dUNC13 fusion was detected upon reprobing the filter with anti-GST antibodies (data not shown). The region of dUNC13 that bound calmodulin included some homology to mUNC13-1, but no sequence similarity to mUNC13-2, mUNC13-3, or UNC13. The gel overlay assay described above tested for interaction between calmodulin and dUNC13 immobilized on a membrane. To address whether dUNC13 was capable of binding to calmodulin in solution, we performed a pull-down assay by incubating a GST-dUNC13 fusion protein with agarose beads linked to calmodulin. The GST-dUNC13 fusion protein, bound calmodulin and did so in a Ca 2ϩ -independent manner (Fig. 6A). GST alone did not bind calmodulin in the presence of Ca 2ϩ or EGTA and neither GST-dUNC13 or any other fusion protein described below bound to agarose beads under any Ca 2ϩ conditions (data not shown; Ref. 21).
The portion of the PLX protein that was isolated in the screen extended from residues 180 -1379. Using a series of overlapping GST fusion proteins and the gel overlay assay, the calmodulin-binding site(s) contained in the original fusion protein was further mapped to residues 657-680 (Figs. 4B and  5B). The sequence of the calmodulin-binding site was not con-served in the mouse homolog, TBC1 (Fig. 1C), but was in several human ESTs (data not shown). A bovine homolog of PLX (Lyncein), which we isolated from a bovine retinal library, was highly conserved in the calmodulin-binding domain (Fig.  1C) despite having no higher overall sequence conservation to PLX than TBC1 (data not shown). Moreover, a fusion protein containing the conserved sequence in Lyncein bound calmodulin (data not shown; Fig. 4B). PLX also bound to calmodulin in a pull-down assay; although this interaction was Ca 2ϩ independent (Fig. 6B).
Both gel overlay and pull-down assays were also used to confirm that CRAG bound calmodulin and to further localize the domain which contained the binding site(s). The initial fusion protein identified in the screen extended from residues 388 to 1209 (Fig. 4C). Using the gel overlay technique, the domain containing the calmodulin-binding site was narrowed to residues 655-822 ( Fig. 4C and 5C), a region that was not conserved in AEX-3 or the rRab3 GEP. As was the case with PLX and dUNC13, the association with calmodulin was Ca 2ϩ dependent in the gel overlay assay, but Ca 2ϩ -independent in the pull-down experiments (Fig. 6C).
Many, but not all, calmodulin-binding sites resemble basic amphipathic helices and/or conform to the IQ consensus motif, a sequence of ϳ25 amino acids containing the core consensus IQXXXRGXXXR (57-59). However, no such sequences were FIG. 3. Expression of RNA analyzed by probing RNA blots. The blots contain RNA prepared from wild-type at the indicated developmental stages as well as from wild-type adults and/or wild-type adult heads, bodies, and from sine oculis heads. A, a blot containing total RNA was probed, stripped of the signal and reprobed as indicated. The blot was also probed with RP49 to determine the relative levels of RNA loaded in each lane. The approximate sizes of the RNAs, relative to the indicated RNA markers, were as follows: CRAG, 2.5 and 4.6 kb; PLX, 5.6 kb. Since CALO RNA was larger than the 9.49-kb marker, the blot was reprobed with the fat gene (data not shown), which encodes an RNA ϳ15-20 kb. CALO co-migrated with FAT and was therefore approximately similar in size. B, a blot containing poly(A) ϩ RNA was used to examine the expression of dUNC13 and CaM kinase I (CaMK1) mRNAs. The major dUNC13 RNAs expressed during pupal development were ϳ2.4, 4, and 12 kb. In addition, a 3.3-kb RNA was detected in adult heads. A low level of RNA was detected in so heads in longer exposures. The major CaM kinase I RNA was ϳ2.5 kb. An additional signal of ϳ3.5 kb was seen in adult heads. After longer exposures, low levels of RNA were detected during earlier developmental stages (data not shown). detected in dUNC13, CRAG, or dUNC13 and there was no obvious similarity among these proteins within the calmodulinbinding domains. Furthermore, although we have mapped calmodulin-binding sites within each of these three proteins, it remains possible that there exist additional sites that were not detected in these analyses.

DISCUSSION
Prior to the current screen, five calmodulin-binding proteins were known to be expressed in the Drosophila retina, four of which, NINAC, TRP, TRPL and INAD, function in phototransduction. However, we found that at least six out of the eight additional calmodulin-binding proteins expressed in the Drosophila retina contained domains related to proteins implicated in synaptic transmission. Two of the proteins were related to components implicated in synaptic transmission which were not previously known to bind calmodulin and four were known calmodulin-binding proteins: calcineurin, MLCK, CaM kinase I, and CaM kinase II. A seventh protein had weak homology to two yeast RAB GTPase-activating proteins and may function in exocytosis. The eighth protein, Calossin, is the largest known calmodulin-binding protein. Although the function of CALO could not be inferred from the sequence, it appeared to be conserved throughout animal phylogeny.
CRAG May Provide a Mechanism for Ca 2ϩ -regulated GDP/ GTP Exchange of Rab3-One of the novel calmodulin-binding proteins identified in the screen, CRAG, contained significant homology to the largest of the three conserved domains in the recently identified Rab3 GEPs (31,32). Moreover, CRAG was similar in size to Rab3 GEPs. While it remains to be determined if CRAG is also a Rab3 GEP, such a finding would have interesting implications regarding the mechanism by which GTP exchange on Rab3 is regulated. Rab3 binds to synaptic vesicles; however, this association only occurs in resting nerve terminals and requires Rab3 in the GTP bound state. Mutation of the C. elegans Rab3 GEP, AEX-3, causes accumulation of Rab3 in neuronal cell bodies and an impairment in release of neurotransmitter (32). Thus, the Rab3 GEP appears to play a critical role in association of Rab3 with synaptic vesicles and in synaptic transmission. The observation that CRAG binds calmodulin implies that the putative GEP activity of this protein could be regulated by changes in Ca 2ϩ levels which are spatially restricted to microdomains near the active zones in presynaptic terminals.
A variety of evidence suggests that the absolute level of Rab3-GTP bound to synaptic vesicles regulates the rate of exocytosis by limiting the number of vesicles that can be fused with the plasma membrane. Thus, formation of Rab3-GTP appears to be a crucial step in synaptic transmission. The mechanisms controlling the GDP-GTP exchange are not known but one possibility is that CRAG is a Rab3 GEP and the associated calmodulin provides a sensor to differentiate between the lower Ca 2ϩ levels in resting nerve terminals and higher levels resulting from Ca 2ϩ influx. While it remains to be determined if CRAG is a Rab3 GEP and whether the exchange activity is regulated through the associated calmodulin, an exchange factor for another small GTPase, RAS, binds to and is regulated by Ca 2ϩ /calmodulin (60).
CRAG bound to calmodulin in solution in either the presence of absence of Ca 2ϩ . However, using the gel overlay assay, calmodulin-binding was strictly Ca 2ϩ dependent. The gel overlay technique inolves fractionating the proteins by SDS-PAGE resulting in extensive protein denaturation. Although the basis for the differences in calcium dependence in the gel overlay and column binding assays is unclear, many other proteins bind calmodulin in a Ca 2ϩ -independent manner under native conditions but show Ca 2ϩ -dependent binding using denaturing conditions. Examples include the vertebrate brain myosin V (formerly referred to as p190; 61), phosphorylase kinase (62), and the small GTP-binding protein RIC (21). Thus, CRAG probably binds calmodulin in a Ca 2ϩ -independent manner since the solution binding assays are more likely to reflect physiological binding conditions. While Ca 2ϩ is typically required for association of calmodulin with its targets, several others also bind in a Ca 2ϩ -independent manner. These include some unconventional myosins, neuromodulin and neurogranin (reviewed in Ref. 63).
UNC13 Proteins May Have Dual Ca 2ϩ Sensors-In addition to aex-3, several other mutations have been identified in C. elegans that appear to disrupt exocytosis of synaptic vesicles and release of neurotransmitter. One such mutation is in the gene encoding the diacylglycerol-binding protein, UNC13. In the current work, we found that one of the calmodulin-binding proteins is a highly conserved Drosophila homolog of the C. elegans and mammalian UNC13 proteins. Although the specific function of UNC13 remains unclear, it may operate in docking and/or fusion of synaptic vesicles since the rat brainspecific mUNC13-1 protein binds directly to two proteins, syntaxin and Doc2␣, which function in Ca 2ϩ -dependent exocytosis (64,65).
The C2 domains present in UNC13 homologs could potentially serve as a Ca 2ϩ sensor which responds to the Ca 2ϩ influx required for exocytosis. Therefore, the question arises as to the function of a potential second type of Ca 2ϩ sensor provided by the binding of calmodulin to dUNC13. One possibility is that each UNC13 protein really has only one Ca 2ϩ sensor and that it is supplied in some isoforms by the C2 domain and in others through Ca 2ϩ /calmodulin. Consistent with this proposal, the calmodulin-binding domain is not conserved in UNC13 sug-gesting that the C2 domain provides the only Ca 2ϩ detector in this protein. The reverse may be the case in mUNC13-1 since this protein does not appear to contain Ca 2ϩ -binding C2 domains (34) but does show sequence similarity to the dUNC13 calmodulin-binding site.
An alternative proposal, which we favor, is that some UNC13 proteins may be regulated by Ca 2ϩ via both C2 domains and calmodulin. Such dual regulation may provide a mechanism for extremely rapid as well as sustained responses to highly transient increases in Ca 2ϩ . The rise in Ca 2ϩ , resulting from opening of the voltage-gated channels in synaptic terminals, occurs in microdomains and collapses within microseconds after closing of the ion channels (reviewed in Ref. 2). C2 domains comprise an unusual Ca 2ϩ binding motif in that Ca 2ϩ appears to regulate this domain through a shift in electrostatic potential rather than a conformational change (reviewed in Ref. 2). As such, C2 domains have the potential to respond very quickly, but transiently, to the rapid Ca 2ϩ flux in the active zones of presynaptic terminal. Although fusion and release of neurotransmitter is extremely rapid (submilliseconds to milliseconds), there is some latency between the opening and closing of the ion channels and these latter events. Ca 2ϩ binding to calmodulin, which induces a conformational change, may induce a more delayed but sustained response to Ca 2ϩ than provided by the C2 domain. Thus, dual binding of Ca 2ϩ to calmodulin and C2 domains may enable UNC13 proteins to sense the Ca 2ϩ rise within a few microseconds and sustain the response for several hundred microseconds to several milliseconds.
Potential Roles of Pollux in Photoreceptor Cells-One of the calmodulin-binding proteins identified in the screen was PLX, a protein suggested to be a novel cell adhesion molecule of 732 residues (47). Based on the deduced amino acid sequence, PLX is predicted to contain at least one transmembrane domain and a leucine zipper. Spatial localization of the protein showed that it is found on the cell surface as well as the lumen of the trachael system and on the plasma membrane of axons in the central nervous system (47). A related mouse homolog, TBC1, described in a contemporaneous report, was 1141 residues and was found to be a nuclear protein (48). Thus, TBC1 and PLX have very disparate spatial distributions.
We found that PLX was 1379 residues rather than 732 amino acids as previously reported. The additional sequence was not due to a chimeric cDNA since multiple plx cDNAs were obtained and TBC1 shared similarity to PLX both N-and Cterminal to the formerly assigned initiating methionine at residue 648. Of particular relevance here, we found that PLX bound calmodulin and did so in a Ca 2ϩ -independent manner. Although the sequence of the calmodulin-binding site was not conserved in TBC1, the region was very similar in Lyncein, a homolog isolated from a bovine retinal library. 2 Furthermore, the Lyncein sequence also bound calmodulin. Thus, it appears that a PLX homolog is expressed in the vertebrate retina.
A possible clue as to the function of PLX in the retina is that it shares some similarity to two yeast Rab GAP proteins (37); although, there was no homology to the Rab3 GAP expressed in the rat brain (49). Nevertheless, the observation that PLX contains a domain related to Rab GAPs combined with the finding that it appears to be localized to the plasma membrane and lumen of the trachael system raises the possibility that PLX may be involved in exocytosis. In the Drosophila visual system, exocytosis is important not only in synaptic transmission but in turn-over of the microvillar membrane of the photoreceptor cells. Shedding of membrane does not occur uni-formly during the diurnal cycle, but occurs maximally soon after dawn (66). Thus, an increase in the exocytotic process is correlated with the light dependent rise in Ca 2ϩ and therefore might be regulated in part by a Ca 2ϩ sensing component in a Rab cycle. Alternative potential functions for PLX in photoreceptor cells include other processes that involve vesicular trafficking such as insertion of new membrane in the microvilli and the budding, targeting, and fusion of rhodopsin carrier vesicles with the plasma membrane. These latter events involve a variety of Rab proteins (67) and also appear to be regulated during the daily light cycle (66).
Concluding Remarks-With the identification of the calmodulin-binding proteins in the current screen, there exists a minimum of 13 targets for calmodulin in the Drosophila retina. Furthermore, each of the calmodulin-binding proteins was enriched in the retina. Drosophila Rab3 is also enriched in the adult heads (68) and may bind calmodulin since vertebrate Rab3 associates with calmodulin (27). Interestingly, most of the same proteins that were identified in the Drosophila retina are also expressed in the vertebrate retina. 3 The great diversity and abundance of retinal calmodulin-binding proteins indicates the particular importance of calmodulin as a Ca 2ϩ sensor in the retina. Currently, there is strong evidence that retinal calmodulin functions in phototransduction. However, Ca 2ϩ /calmodulin may also be utilized to regulate a broad range of vesicular trafficking events in the retina. Such processes may include membrane turnover, protein transport and especially synaptic transmission.