INAD Signaling Complex of Drosophila Photoreceptors

cross-talk is to physically separate the phototransduction cascade from other signaling pathways. In vertebrate photoreceptors, this is achieved by sequestering the phototransduction proteins into a specialized cellular compartment, the outer segment of rod or cone photoreceptors, that serves only for the detection of light and is connected to the rest of the cell by a small ciliary stalk. In Drosophila photoreceptors, proteins involved in light detection are also confined to a separate cell compartment, the rhabdomere formed by a densely packed stack of microvilli along one side of the cell.

In addition to being confined to the rhabdomere, phototransduction proteins of Drosophila photoreceptors are organized into supramolecular complexes by the scaffolding protein INAD (inactivation no afterpotential D). The identification of the INAD signaling complex in these photoreceptor cells has led to a reevaluation of the principles by which Drosophila phototransduction is activated and regulated. With respect to signaling specificity, it has been suggested that the organization of the phototransduction cascade into signaling complexes avoids cross talk with other signaling cascades [8]. The concept of signaling complexes, in which signaling proteins are permanently linked together, differs fundamentally from a model in which signaling proteins diffuse freely in the photoreceptive membrane and interact with each other by random collisions. The model of freely diffusing signaling components has been established for vertebrate rod photoreceptor cells. In this sensory system, the high concentration of phototransduction proteins in the disk membrane may allow the cascade to operate at sufficient speed and specificity without forming signaling complexes, although some components of this signaling cascade appear to be linked together as well (see “Signalling complexes in vertebrate photoreceptor cells”). The compound eyes of fast-flying insects like Drosophila show a much higher temporal resolution than vertebrate eyes, and their photoreceptors have evolved the fastest G protein-coupled signaling cascades known to date. In Drosophila photoreceptors, the receptor potential can be generated within 20 ms after light stimulation [9]. This is about 10 times faster than in mammalian photoreceptors and about 100 times faster than in toad photoreceptors [3]. Arguably, the assembly of phototransduction proteins into signaling complexes may contribute to the high signaling speed of the cascade by eliminating delays due to the time required for protein diffusion.

In this contribution, we first highlight common and discriminative features of the vertebrate and Drosophila phototransduction cascade. We then focus on the INAD signaling complex of Drosophila photoreceptors and finally discuss the assembly of signaling proteins in vertebrate photoreceptors. For a more comprehensive coverage of these topics, refer to excellent reviews on invertebrate photoreception in general [3, 4, 10–15] or on the INAD signaling complex in particular [8, 16–20].

COMPARISON OF VERTEBRATE AND DROSOPHILA

PHOTOTRANSDUCTION CASCADES

Photoreception in vertebrates and in Drosophila shows common features as well as significant differences. In both visual systems, light is absorbed by the seven- transmembrane-receptor rhodopsin, which is present at high concentration in the disk membranes of rod photoreceptor cells or in the microvillar membranes of rhabdomeral photoreceptor cells. The high density of rhodopsin in the light-absorbing compartments

ensures a high probability of catching a photon and hence a high sensitivity of the photoreceptors. Vertebrate and invertebrate rhodopsins display low but significant sequence identity, for example, 22% amino acid identity between the major Drosophila rhodopsin Rh1 and bovine rhodopsin [21, 22]. It is now generally accepted that all animal rhodopsins evolved from a common ancestor [23]. Conserved amino acids in vertebrate and invertebrate rhodopsins include a lysine in the seventh transmembrane segment that constitutes the chromophore-binding site, serine and threonine residues near the C-terminus that are targets for light-dependent rhodopsin phosphorylation, and consensus sites for N-linked glycosylation near the N-terminus of the protein. Although molecular details in rhodopsin activation and inactivation are markedly different between vertebrates and invertebrates [24, 25], the rhodopsin activation mechanism initiated by conversion of an 11-cis retinal chromophore into its all-trans form follows similar principles as does rhodopsin inactivation by binding of arrestin proteins to the active conformation of the receptor.

To transmit the light signal from rhodopsin to photoreceptor ion channels, vertebrates and invertebrates make use of different kinds of G protein-coupled cascades (Fig. 1). Vertebrate rhodopsins activate the visual G protein transducin, which couples to a phosphodiesterase (Fig. 1A). Activation of the phosphodiesterase results in breakdown of cyclic guanosine monophosphate (cGMP), which is synthesized by a guanylate cyclase and, in the dark, keeps a fraction of cGMP-gated ion channels in an open state. Thus, light activation of most vertebrate photoreceptors leads to the closure of cGMP-gated ion channels, which results in a hyperpolarization of the cells.

In Drosophila, as in most invertebrates, the activated rhodopsin molecule transmits the signal to a heterotrimeric Gq protein consisting of Gαq, Gβe, and Gγe (Fig. 1B). As a result of G protein activation, Gαq couples to the effector enzyme phospholipase Cβ (PLCβ). Activated PLCβ hydrolyzes the membrane phospholipid phosphatidylinositol 4,5-bisphosphate (PIP2) to form the second messengers diacylglycerol (DAG) and inositol 1,4,5-trisphosphate (IP3). The phototransduction cascade terminates in the opening of cation channels composed of ion channel subunits of the TRP protein family, TRP, TRPL, and possibly TRPγ, which results in the depolarization of the photoreceptor cell membrane. Although the gating mechanism of these TRP channels has not yet been entirely clarified, several lines of evidence suggest that the DAG branch, rather than the IP3 branch of the phosphoinositol signaling pathway activates the ion channels [26].

In both vertebrate and invertebrate photoreceptors, the absorption of a single photon results in the generation of a distinct electrophysiological response of the receptor, a so-called quantum bump. Quantum bumps correspond to the coordinated closure or opening of a defined number of ion channels. The macroscopic response to a stronger light stimulus (i.e., the receptor potential) is achieved by summation of the quantum bumps obtained from more or less simultaneous absorption of many photons.

IDENTIFICATION OF THE INAD SIGNALING COMPLEX

In the 1970s, Bill Pak and coworkers isolated a number of Drosophila phototransduction mutants that were identified by their altered electrophysiological characteristics in electroretinogram recordings [27, 28]. This approach eventually led to the identification of major players in Drosophila phototransduction such as rhodopsin Rh1 [21, 22] or the

Fig. 1. Comparison of vertebrate and Drosophila phototransduction cascades. A In vertebrate photoreceptors, the G protein transducin couples to a phosphodiesterase (PDE). The active PDE degrades cyclic guanosine monophosphate (cGMP), which is synthesized by a guanylate cyclase (GC) from guanosine triphosphate (GTP). In the dark, the cGMP-gated channels are kept open by cGMP. Light activation of these photoreceptors leads to the closure of the cGMPgated cation channels and thus to a hyperpolarization of the cell membrane. B In Drosophila photoreceptors, the Gq protein activates a phospholipase Cβ (PLCβ), which then hydrolyzes the membrane phospholipid phosphatidylinositol 4,5-bisphosphate (PIP2) and generates the two second messengers diacylglycerol (DAG) and inositol 1,4,5-trisphosphate (IP3). The phototransduction cascade terminates in the opening of at least two cation channels, TRP (transient receptor potential) and TRPL (TRP-like), which leads to a depolarization of the cell membrane. TRP and TRPL channels may be gated by DAG or by polyunsaturated fatty acids (PUFAs), which could be released from DAG by a DAG lipase. The scaffolding protein INAD (inactivation no afterpotential D) organizes phototransduction proteins into a supramolecular complex. The unconventional myosin NINAC (neither inactivation no afterpotential C) may couple this supramolecular complex to the actin cytoskeleton. An eye-specific protein kinase C (ePKC), which is also part of this complex, appears to be involved in response inactivation and light adaptation. Rh rhodopsin, SMC submicrovillar cisternae.

central effector enzyme PLCβ [29]. Another mutant isolated in this genetic screen was termed inactivation no afterpotential D (inaD). The electroretinogram of this mutant showed inactivation by a strong light stimulus but lacked the afterpotential, a prolonged depolarization after termination of the light stimulus that is observed in wild-type photoreceptors when more than 20% of rhodopsin molecules are activated. The corresponding inaD gene was isolated by a subtractive hybridization screen for eye-enriched complementary DNAs, one of which rescued the inaD phenotype in transgenic Drosophila [30]. The inaD gene encodes a protein of 674 residues with homology to PDZ domain-containing proteins. PDZ domains have been named after the postsynaptic density protein PSD95, Drosophila disks large (Dlg), and the tight-junction protein ZO-1, in which this domain was first identified.

Although initially only two PDZ domains were identified in INAD, Tsunoda et al. [31] realized that INAD is a pure scaffolding protein that is composed almost exclusively of five PDZ domains. As a first interaction partner of INAD, one of the two principle light-activated ion channels, TRP, was identified [32]. The binding of TRP to INAD is abolished in the original inaDP215 mutant by a point mutation (M442K) in the third PDZ domain. The existence of a signaling complex in fly photoreceptors was first proposed when coimmunoprecipitation studies with proteins of purified photoreceptor membranes from larger flies revealed that INAD, TRP, PLCβ, and an eye-specific protein kinase C (ePKC) are tethered together [33, 34]. The assembly of these phototransduction proteins was also shown for Drosophila photoreceptors, and a requirement of INAD for the rhabdomeral localization of TRP was demonstrated [35].

The concept of organizing major components of the fly visual transduction cascade into a signaling complex was further established and extended by Tsunoda et al. [31], who demonstrated that INAD functions as a scaffold to which the phototransduction proteins are bound. In a newly generated INAD-null mutant, in which the photoresponse was almost abolished, the signaling complexes were completely lost, and TRP, PLCβ, and ePKC were mislocalized. Besides the major INAD ligands TRP, PLCβ, and ePKC that are present in the complex at about equimolar ratios [33], additional photoreceptor proteins were shown to interact with INAD [35–37]. These proteins are rhodopsin, the second ion channel TRPL, the unconventional myosin NINAC (neither inactivation no afterpotential C), and calmodulin. They are detected in coimmunoprecipitation studies in substoichiometric concentrations, and their localization to the rhabdomere does not depend on the presence of INAD. While binding of the NINAC myosin III to INAD has been reported to be required for proper termination of the visual response [36], the significance of the proposed interaction of rhodopsin, TRPL, and calmodulin with INAD is not clear.

STRUCTURE OF THE INAD SIGNALING COMPLEX

AND BINDING SPECIFICITY

The three-dimensional structure of PDZ domains was first solved for the third PDZ domain of PSD95 and for the human homolog of the discs-large protein (hDlg) by x-ray diffraction of crystals containing the PDZ domains and short peptides corresponding to the C-terminal region of their targets [38, 39]. The approximately

90 amino acids of these PDZ domains fold into a six-stranded β-sheet (B1–B6) and two α-helices (A1, A2). PDZ domains typically bind their targets by interaction with a short stretch of amino acids at the very C-terminal end of the proteins. The peptides bind in a groove between the B2 strand and the A2 helix and form an antiparallel β-strand that interacts with the B2 strand by hydrogen bonding. The region between A2 helix and B2 strand contains a conserved R/K-X-X-X-G-L-G-F motif that forms a carboxylatebinding loop that interacts with the free COO− group at the C-terminus of the ligand (for reviews, see [40, 41]).

Screening of an oriented peptide library with nine different PDZ domains showed that consensus sequences for PDZ interaction fall into two main categories: (S/T)-X- Φ-COOH and Φ-X-Φ-COOH (X is any amino acid and Φ is a hydrophobic amino acid) [42]. Accordingly, PDZ domains have been grouped into type I PDZ domains that bind (S/T)-X-Φ-COOH and type II PDZ domains that bind Φ-X-Φ-COOH. Additional binding motifs that do not fit into these categories have been identified, for example, the motif D/E-X-Φ-COOH in peptides binding to the PDZ domain of neuronal nitric oxide synthase [43] or H-W-C-COOH at the C-terminus of N-type Ca2+ channel that binds to the first PDZ domain of Mint-1 [44]. Due to this heterogeneity in binding motifs, it is not possible to predict from the amino acid sequence alone whether a given photoreceptor protein is a putative ligand for one of the PDZ domains of INAD.

For the INAD signaling complex, structural details are only available for the binding of PLCβ to PDZ1 of INAD. The interaction of a heptapeptide corresponding to the C-terminal region of PLCβ with PDZ1 has been studied by crystallography [45]. The binding of this peptide to PDZ1 differs from typical PDZ–ligand binding schemes because it involves a covalent interaction. The crystal structure revealed a disulfide bond between the cysteine residue in the F-C-A-COOH motif of PLCβ and a cysteine in the B2 strand of the PDZ domain, which results in a high-affinity interaction between PLCβ and INAD. The functional significance of this covalent interaction is, however, not clear. The PLCβ homolog of the related fly species Calliphora has no cysteine near the C-terminus (A. Huber and M. Bähner, unpublished), and this PLCβ can be removed from INAD with high salt buffer under nonreducing conditions [33]. Studies using GST (glutathione S-transferase) pull-down assays, coimmunoprecipitation of recombinantly expressed PDZ domains with putative binding partners, yeast two-hybrid assays, and analysis of Drosophila mutants with mutations in specific PDZ domains or ligand-bind- ing domains provided additional insights into the binding partners of each of the five PDZ domains of INAD. At least in vitro, it appears that binding of the various INAD ligands is not restricted to one particular PDZ domain for each target, for example, ePKC, which was originally shown to bind to PDZ4 was found to interact also with PDZ2 and PDZ3 [37, 46]. PLCβ was shown to bind to PDZ1 with its C-terminal binding motif F-C-A-COOH and also to PDZ5 with an internal sequence that overlaps with the G protein-binding domain [47]. In vivo, PLCβ binding to PDZ5 is relevant for PLCβ stability because a point mutation of a conserved glycine in PDZ5 or introduction of a stop codon between PDZ3 and PDZ4 that results in the expression of a truncated INAD significantly reduces the level of PLCβ [31, 48]. TRP that appears to be associated with PDZ3 in vivo [31, 32] was found to interact also with PDZ4 [37]. The additional possible binding partners of INAD, rhodopsin, TRPL, and the myosin III NINAC also interact

Fig. 2. Model of proposed supercomplexes consisting of several INAD (inactivation no afterpotential D) signaling complex core units. The core units consist of four molecules INAD (shown as broad lines; PDZ domains are numbered 1–5), four TRP (transient receptor potential) subunits that form a functional cation channel, four molecules PLCβ (phospholipase Cβ), and four molecules ePKC (eye-specific protein kinase C). Possible protein–protein interactions that may contribute in the assembly of supercomplexes as discussed in the text are indicated by dotted lines.

with the PDZ domains of INAD in vitro, while the binding of calmodulin was reported to take place in the region between PDZ1 and PDZ2 [37].

Importantly, Xu et al. (1998) provided evidence that INAD molecules can bind to each other via interaction of PDZ domains 3 and 4. This raised the possibility that the INAD signaling core complexes are organized into larger units composed of several ion channels, phospholipases Cβ, and protein kinases C. These supercomplexes are referred to as signalplex or transducisome [16, 37]. A model for the possible organization of these supercomplexes is shown in Fig. 2. In addition to INAD–INAD interactions, the supercomplexes could be generated by the formation of tetrameric TRP ion channels from TRP momomeres of different INAD core complexes or by the simultaneous interaction of PLCβ with PDZ1 and PDZ5 of different INAD molecules. In addition, it has been shown that PLCβ forms homodimers that could tether together two INAD core complexes [45].

ANCHORING OF THE INAD SIGNALING COMPLEX

TO THE MICROVILLAR MEMBRANE

Early studies on the INAD signaling complex have revealed that INAD is required for proper localization of its major ligands (TRP, ePKC, and PLCβ) to the rhabdomere as these signaling proteins are mislocalized in inaD mutants and become degraded [16, 31, 35]. But, how is INAD itself anchored to the microvillar membrane? The localization of the INAD signaling complex to the rhabdomere could be achieved by interaction with components of the microvillar membrane or with the microvillar cytoskeleton. Insight into this question came from findings showing that TRP is crucial for anchoring INAD to the rhabdomeral membrane [49, 50]. INAD becomes mislocalized in trp null mutants or in mutants lacking the PDZ-binding site of TRP but not in mutants defective in other INAD ligands. Together with INAD, ePKC and PLCβ are also mislocalized in a trp mutant, and they are still attached to INAD in this mutant, as revealed by coimmunoprecipitation. Since TRP is the only transmembrane protein among the major components of the INAD signaling complex, it seems reasonable to assume that TRP is required for membrane attachment of the complex. However, a substantial amount of the signaling complex (ca. 25%) of INAD [50] remains in the rhabdomeres of trp mutants, suggesting the existence of an additional TRP-independent mechanism for INAD anchorage. The signaling complexes remaining in the rhabdomere of trp mutants may account for the fact that the electrophysiological phenotype of a trp null mutant, in which the receptor potential is generated by TRPL channels, is by far not as severe as in the inaD null mutant, which shows almost no light response [31]. Expression of INAD proteins with mutations in either of its five PDZ domains in CHO (Chinese hamster ovary) cells indicated that PDZ1 is involved in attaching INAD to the cell membrane [50]. Besides its interaction with PLCβ, PDZ1 of INAD was shown to bind to the unconventional myosin III NINAC, which in Drosophila photoreceptors could anchor the signaling complex to the microvillar actin cytoskeleton [36]. More recently, an interaction of TRP (and of TRPL) with the actin-binding protein moesin has been reported, which could represent an additional mechanism for linking the signaling complex to the cytoskeleton [51]. This last interaction is light dependent: Moesin is bound to the ion channels in dark-raised flies and released after 1h of light adaptation. The dynamic interaction with TRP and TRPL results in a subcellular redistribution of moesin and concomitant rearrangements of the actin cytoskeleton.

The question whether the TRP-INAD interaction is necessary for targeting the signaling complex to the rhabdomere or whether TRP and INAD reach the rhabdomere independently of each other and then require the interaction for staying there has also been clarified. In late pupae, at a time when maturation of rhabdomeres is completed and signaling proteins are first synthesized, TRP was found in the rhabdomeres independently of INAD, and INAD likewise reached the rhabdomeres independently of TRP. These findings suggest that the TRP–INAD interaction is necessary for anchoring of the signaling complex but not for its targeting [49, 50]. On the other hand, in wild-type flies the assembly of at least three proteins of the signaling complex (INAD, ePKC, PLCβ) seems to occur before its components have reached the rhabdomere. Evidence for this conclusion comes from experiments, in which INAD is expressed in the photoreceptors under the control of a heat shock promoter. While recovery of the electrophysiological