Drosophila Photoreceptors and Signaling Mechanisms

Fly eyes have been a useful biological system in which fundamental principles of sensory signaling have been elucidated. The physiological optics of the fly compound eye, which was discovered in the Musca, Calliphora and Drosophila flies, has been widely exploited in pioneering genetic and developmental studies. The detailed photochemical cycle of bistable photopigments has been elucidated in Drosophila using the genetic approach. Studies of Drosophila phototransduction using the genetic approach have led to the discovery of novel proteins crucial to many biological processes. A notable example is the discovery of the inactivation no afterpotential D scaffold protein, which binds the light-activated channel, its activator the phospholipase C and it regulator protein kinase C. An additional protein discovered in the Drosophila eye is the light-activated channel transient receptor potential (TRP), the founding member of the diverse and widely spread TRP channel superfamily. The fly eye has thus played a major role in the molecular identification of processes and proteins with prime importance.

translocation play a major role. (iv) Light-induced translocation of G q α and the excess of G q β over G q α, which prevents spontaneous activation of the G q -protein in the dark. (v) The dual role of lightactivated phospholipase Cβ (PLCβ) in vivo as a G-protein-mediated activator and negative regulator of phototransduction via its action as a GTPase activating protein (GAP). (vi) Unitary signaling events (e.g. single photon responses, quantum bumps). (vii) The lightactivated channels, TRP and TRPL, the founding members of the TRP superfamily channel proteins. (viii) Light-induced translocation of the TRPL channel. (ix) The inactivation no afterpotential D (INAD) multimolecular signaling complex, which binds the TRP channel, its activator, the PLC and its regulator, eye-specifi c protein kinase C (ePKC).

STRUCTURAL AND OPTICAL PROPERTIES OF THE DIPTERA COMPOUND EYE GENERAL ANATOMY
Two distinct types of eyes have evolved through evolution; the lens eye (or camera eye) typically encountered in vertebrates, and the compound eye typically encountered in arthropods. Many insects encompass both types of eyes. While, the compound eye is the primary image forming organ, the ocelli lens eye is small and primitive Franceschini, 1968, 1969). The compound eyes are composed of many repeat and well-organized units termed ommatidia (Figure 1B) embedded in a sphere ( Figure 1A). The number of ommatidia in insects vary from just a handful in the primitive Archaeognatha (jumping bristletails) and Thysanura (silverfi sh or bristletails) to several hundred up to thousands in Diptera (which includes the house fl y Musca and the fruit fl y Drosophila).
In Drosophila, the ommatidium consists of about 20 cells, in which 8 (6-21 in different insect species) are the photoreceptor cells (RZ, Figure 1A). Each ommatidium contains a dioptric apparatus composed of transparent chitinous cuticle, which forms the cornea (C, Figure 1A) and an extracellular fl uid-fi lled cavity, called the pseudocone (PC, Figure 1A). The fl oor of the cavity is formed by four

INTRODUCTION
Vision of invertebrate species has been one of the fi rst senses to be thoroughly studied, and many fundamental principles relevant to all senses have been fi rst discovered in invertebrate eyes. A notable example is the discovery of lateral inhibition in the compound eye of the Limulus by the Nobel Prize Laurie, Haldan Keffer Hartline (Ratliff, 1990). Surprisingly, invertebrate phototransduction, the process by which light quanta are translated into electrical signal is still not entirely understood in terms of its underlying molecular mechanism. The pioneering experiments, which exploited the size of giant photoreceptor cells in some invertebrate species like the Limulus (reviewed in Dorlochter and Stieve, 1997), were followed by studies on Drosophila melanogaster, exploiting its great molecular genetics power (reviewed in Minke and Hardie, 2000;Montell, 1989;Pak, 1995;Ranganathan et al., 1995). In the present review, we focus on processes and molecules that have been discovered in invertebrate eyes in general and in the Drosophila eye in particular, which shed light on crucial functions of other cells and tissues. These landmark discoveries include: (i) Structural and optical properties of Diptera compound eyes. (ii) Bistable photopigments in which both the rhodopsin (R) and metarhodopsin (M) states of the photopigment are dark stable and photoconvertible. (iii) The photochemical cycle in which phosphorylated arrestin (ARR) and ARR Semper cells (SZ, Figure 1A) and the walls by primary pigment cells (PZ, Figure 1A, red), which together circle the pseudocone, shielding the photoreceptor from stray light coming from adjacent ommatidia. The photoreceptor cells are highly polarized epithelial cells, with a specialized compartment known as the rhabdomere (Rh, Figure 1A), consisting of a stack of ∼30,000-50,000 microvilli each ∼2 µm long and ∼60 nm in diameter. The transduction machinery is located in these tightly dense structures, while the nucleus and cellular organelles (N, Figure 1C), such as submicrovillar cisternae (SMC, Figure 1C) reside in the cell body. Pioneering studies conducted by Franceschini and Kirschfeld in Diptera (mainly in Musca) have elucidated the remarkable optics of the compound eye. In their studies, they showed that the highly ordered rhabdomeres form light guides (Kirschfeld and Snyder, 1976) that have been widely exploited experimentally (Figure 3). For example, the screening for retinal degeneration mutants of Drosophila has used the optical phenomenon designated deep pseudopupil (dpp), by Franceschini and Kirschfeld (1971), that is associated with their light guide property (see Figure 3). The dpp, which disappears in retinal degeneration mutant fl ies such as in R defective mutants, has been used as an effi cient tool for a fast screen of large populations of putative mutant fl ies. Other examples are spectral measurement of the compound eye such as the eye shine, resulting from tapetal refl ection, transmittance spectra of photopigments and fl uorescent measurements of M.

OPEN AND CLOSED RHABDOM
Two kinds of rhabdomere architecture exist: closed rhabdom, in which all rhabdomeres are fused at the center of the ommatidium (Figures 2A,C) and open rhabdom, in which the rhabdomeres are separated (Figures 2A,B), forming a polygon pattern depend-ing on the number of photoreceptors (hexagonal in Drosophila). Each ommatidium is connected by axons to the ganglionic layers providing a single or several image elements of space, depending on the rhabdomere architecture ( Figure 1A). In open rhabdomere, the repeated elements are arranged in a specifi c geometrical patterning and spacing, ensuring visual connectivity between adjacent ommatidia. Accordingly, the angles between the individual rhabdomeres in one ommatidium are identical to those between adjacent ommatidia. As a result, each of the seven rhabdomeres in one ommatidium portrays the same fi eld of view as a rhabdomere in a neighboring ommatidium (Figures 2D,E; Kirschfeld, 1967). In addition, all six rhabdomeres that share a common fi eld of view send their axons to the same place in the fi rst ganglionic layer -the lamina (La, Figures 1A and 2E). The central rhabdomeres send their axons to the second ganglionic layer -the medulla (Me, Figure 1A). In Drosophila, the seven rhabdomeres of each ommatidium are separated from each other and function as independent light guides ( Figure 1D) forming open rhabdomere architecture (Figures 2A,B). In contrast, bees, beetles and various mosquitoes have a closed rhabdom architecture, in which rhabdomeres within each ommatidium are fused to each other, thus sharing the same visual axis (Figures 2A,C). Recently, the power of Drosophila genetics was exploited to elucidate the molecular factors participating in the transition between open and closed rhabdom architecture by screening, isolating and characterizing Drosophila genes involved in this process. The study identifi ed two genes, spacemaker (spam) and prominin (prom) which when mutated cause the collapse of the intra-rhabdomere space (IRS; Figure 2B) resulting in the conversion of an open rhabdom system into a closed rhabdom architecture. Further analysis showed that SPAM is a secreted protein expressed in the IRS, which acts together with PROM, which is an evolutionary conserved transmembrane (TM) protein often associated with microvilli. Secretion of SPAM into the IRS forces the separation of the stalk membrane, pushing the rhabdomere apart, and the recruitment of SPAM to the microvilli surface by the binding to PROM prevents inter-rhabdomere adhesion. Furthermore, targeted expression of spam to photoreceptors of a closed system markedly reorganizes the architecture of the compound eyes to resemble an open system (Zelhof et al., 2006).
The unusual stiffness of SPAM has been exploited in mechanoreceptors of Drosophila. Accordingly, a recent study has demonstrated the involvement of SPAM in maintaining cell shape and tone, crucial for integrity of the mechanosensory neurons. The authors argued that for poikilothermic organisms, like insects, changes in temperature may impact the function of mechanoreceptor neurons. SPAM role was found as protective of mechanosensory organ from massive cellular deformation caused by heat-induced osmotic imbalance, by forming an extracellular shield that guards mechanosensory neurons from environmental insult (Cook et al., 2008).

FUNCTIONAL RETINAL ORGANIZATION
Drosophila ommatidia consist of eight photoreceptors that can be divided into two functional groups according to their position, functional involvement, spectral specifi city and axonal projection. The R1-R6 cells (marked 1-6 in Figure 1B) represent the major class of photoreceptors in the retina and are involved in image formation and motion detection. These cells have peripherally located rhabdomeres extending from the basal to the apical side of the retina. They express a single opsin called Rh1, which when (C) EM cross-section of Ephestia ommatidium in the region of distal tracheole ends with closed rhabdom architecture (from Fischer and Horstmann, 1971).

(D)
Diagram of seven facets of the compound eye. The encircle rhabdomeres receive light from one and the same point in space (modifi ed from Kirschfeld, 1967). (E) Diagram of axonal connections between ommatidia. Axons of photoreceptor cells one to six receiving light from the same point in space are drawn converging on one and the same cartridge of the lamina (modifi ed from Kirschfeld, 1967).

FIGURE 3 | Deep pseudopupil (dpp) observed in the eye of a living
Drosophila under white orthodromic illumination. The dpp is the superposition of virtual images of adjacent ommatidia observed when a low power microscope is focused at the center of the curvature of the compound eye of Diptera. (A) Dark adapted. (B) After 60 s of medium intense illumination. Note that the central image corresponding to R7/R8 still appears red while the six peripheral images corresponding to R1-R6 refl ect green light. (C) After 60 s of intense illumination (twofold higher). Note that all images refl ect green light. The disparity between (B) and (C) arises from the difference in the absorption spectra between rhodopsin expressed in R1-R6 compared to R7 (upper panels). Schematic representation of the positioning of the pigment granules at each of the above states (lower panels) (modifi ed from Franceschini and Kirschfeld, 1971). combined with 11-cis 3-hydroxy retinal, forms a blue-absorbing R and orange-absorbing M. The R1-R6 cells ( Figure 1B) project their axons to the fi rst optic lobe, the lamina (La, Figure 1A green). The second group consists of two cells in the center of each ommatidium termed, R7 (marked 7 in Figure 1B) and R8 (located below R7) each spanning only half of the retina in length. The central cells R7 and R8 are involved in color vision and detection of polarized light and project their axons to the second optic lobe, the medulla (Me, Figure 1A, pink; Wernet et al., 2006).
Color vision requires comparison between the electrical signals of photoreceptors that are sensitive to different ranges of wavelengths of light. In Drosophila, this is achieved by the inner photoreceptors (R7 and R8) that contain different Rs. The R7 rhabdomere is located distally in the retina and expresses one of two opsins, Rh3 or Rh4, characterized by a UV-absorbing R and blue-absorbing M. The R8 rhabdomere is located proximally in the retina, beneath the R7 rhabdomere (not shown) and expresses one of three opsins, Rh3, Rh5 or Rh6, characterized by a UV-, blue-or green-absorbing R, respectively. On the basis of opsin expression in the R7 and R8 cells, three ommatidia subtypes can be distinguished. The R7 and R8 cells in ommatidia, residing in the dorsal rim area of the eye, which functions as a polarized light detector, both express Rh3 opsin. The "pale" ommatidia subtype express Rh3 in R7 cells and Rh5 in R8 cells and constitute ∼30% of the total ommatidia, while the "yellow" ommatidia subtype express Rh4 in R7 cells and Rh6 in R8 cells and constitute ∼70% of the total ommatidia. Two types of comparisons, required for color vision, can thus occur in the fl y: between the R7 (UV sensitive) and R8 (blue or green sensitive) photoreceptor cells within one ommatidium or between different ommatidia that contain spectrally distinct inner photoreceptors (Wernet et al., 2006).
The intriguing repeated structure of fl y compound eye has been a major scientifi c preparation for research of various aspects of cell differentiation and development. For example, in Drosophila, the hexagonal chiral orientation of the six rhabdomeres in the ommatidia is identical at the upper hemisphere of the compound eye and is reverted by 180° at the equator ( Figure 1D).This phenomenon is generally referred to as the planar cell polarity (PCP) of a tissue, a unique polarization within the plane of epithelium. Genetic screens in Drosophila pioneered the discovery of core PCP factors, which subsequently were found to be evolutionarily conserved. In vertebrate, the PCP factors participate in several developmental processes such as convergence extension, neural tube closure, eyelid closure, hair bundle orientation in inner ear sensory cells, and hair follicle orientation in the skin (for review see Wang and Nathans, 2007).

PUPIL MECHANISM
The pupil-like mechanism of the compound eye was fi rst discovered and studied by Franceschini and Kirschfeld. Upon bright light illumination, tiny pigment granules about 0.2 µm in diameter migrate from dispersed areas of the cell body to the cytoplasmic face of the rhabdomere (Figure 3; Kirschfeld and Franceschini, 1969). The accumulation of pigment granules attenuates propagation of light along the rhabdomere by reducing the refractive index of the interface between the rhabdomere and the adjacent cell body region, thereby changing the waveguide property of the rhabdomere. As a consequence, the amount of light traveling through the rhabdomere is attenuated, activating less photopigment, much like a pupil. This mechanism can attenuate the light fl ux in the rhabdomere by up to one order of magnitude and operates in a time scale of seconds, making it an elegant adaptation mechanism. The pigment granule migration is Ca 2+ -dependent, as evidenced by injecting Ca 2+ chelators into fl ies eye, resulting in the inhibition of pupil closure (Kirschfeld and Vogt, 1980). The pupil mechanism was later found to occur transiently in the trp mutant fl y (Lo and Pak, 1981;Zuidervaart et al., 1979) and was used by Minke as supporting evidence for his hypothesis that TRP is a major route for Ca 2+ entry into the photoreceptor cell (see below, Minke and Selinger, 1991). Recently, the molecular mechanism of pigment granules migration has been elaborated. It was shown that pigment migration is myosin V (MyoV), lightoid, calmodulin (CaM) and cytoplasmic myosin light chain dependent. A model of pigment migration has been put forward by Ready, in which MyoV pulls the pigment granules to the base of the rhabdomere upon Ca 2+ elevation. Accordingly, lightoid, a Rab-related protein, links MyoV to pigment granules while both CaM and myosin light chain bind the long, multi-IQ domain of MyoV lever arms. Together, this Ca 2+ -dependent protein complex migrates to the plus ends (+) of the actin microfi laments, designated the rhabdomere terminal web, at the base of the rhabdomere upon Ca 2+ infl ux induced by illumination .

THE PHOSPHOINOSITIDE CASCADE OF VISION
The signaling proteins of the phototransduction cascade are tightly assembled in the microvillar structure and linked to the actin cytoskeleton (F-actin) via two proteins: Dmoesin, which binds the TRP and TRPL channels, at the base of the microvilli to F-actin (Chorna-Ornan et al., 2005), and no inactivation no afterpotential C (NINAC), that associates INAD to F-actin . The only protein that diffuses during the phototransduction cascade is G q α (Figure 4).
Upon absorption of a photon, R is converted into the active state of the photopigment, M (Figure 4). This leads to the activation of heterotrimeric G-protein (DG q ) by promoting the guanosine diphosphate (GDP) to guanosine triphosphate (GTP) exchange. In turn, this leads to activation of PLCβ, which hydrolyzes the minor phospholipid, phosphatidylinositol 4,5-bisphosphate (PIP 2 ) into the soluble inositol 1,4,5-trisphosphate (InsP 3 ) and the membrane-bound diacylglycerol (DAG). Subsequently, two classes of light-sensitive channels, TRP that is highly permeable to Ca 2+ and TRPL that is a non-selective cation channel, open by a still unknown mechanism. PLC also promotes hydrolysis of the bound GTP, resulting in G q α bound to GDP and this ensures the termination of G q α activity. The TRP and TRPL channel openings lead to elevation of calcium ions extruded by the Na + /Ca 2+ exchanger CALX. Elevation of DAG and Ca 2+ promote ePKC activity, which regulates channel activity. PLC, ePKC and the TRP ion channel form a supramolecular complex with the scaffolding protein INAD (for reviews on the phototransduction cascade see Minke and Cook, 2002;Montell, 1989).

UNITARY EVENTS
Dim light stimulation induces discrete voltage (or current) fl uctuations in most invertebrate species, which are called quantum bumps (Yeandle and Spiegler, 1973; see Figure 5A). Each bump is assumed to be evoked by the absorption of a single photon. The discrete nature of the unitary events of the photoreceptor cells is not due to the quantized nature of light. This has been demonstrated by the application of a non-quantized stimulus such as GTPγS, which elicit quantum bump-like events (Fein and Corson, 1981). The bumps vary in latency, time course and amplitude for identical stimulation and are the consequence of synchronized activation of many light-sensitive channels. The number of channels, which are activated to produce a bump vary greatly in different species: few tens in Drosophila and up to several thousands in Limulus ventral photoreceptors (Nasi et al., 2000). Bump generation is a stochastic process described by Poisson statistics where each effective absorbed photon elicits only one bump (Yeandle and Spiegler, 1973). However  . This photoconversion leads to the activation of heterotrimeric G-protein (G q α) by promoting the GDP to GTP exchange. In turn, this leads to activation of phospholipase Cβ (PLCβ), which hydrolyzes PIP 2 into the soluble InsP 3 and the membrane-bound DAG. Subsequently, two classes of light-sensitive channels, the TRP and TRPL open by a still unknown mechanism. PLC also promotes hydrolysis of the bound GTP, resulting in G q α bound to GDP and this ensures the termination of G q α activity. The TRP and TRPL channel openings lead to elevation of cellular Ca 2+ . Elevation of DAG and Ca 2+ promote eyespecifi c protein kinas C activity, which regulates channel activity. PLC, PKC and the TRP ion channel form a supramolecular complex with the scaffolding protein INAD.
in at least two Drosophila mutants (ninaC and arr, see Figure 5), absorption of a single photon elicits a train of bumps which do not overlap but are separated by intervals (Figures 5B,C). This train of bumps is thought to be caused by a failed R inactivation process and a refractory period of the microvilli (Scott et al., 1997).
A detailed study in Limulus photoreceptors has indicated that the latency of the bump is not correlated with the bump waveform, thus strongly suggesting that the triggering mechanism of the bump arises from different molecular processes than those determining the bump waveform (Dorlochter and Stieve, 1997). These fi ndings are partly explained by models in which the amplifi cation process is preceded by a series of non-amplifying latency producing steps. To produce realistic bumps by such a model means that no step in the transduction cascade could have a life time greater than the duration of a bump generating mechanism which includes the latency, bump duration and bump refractory period. The single photonsingle bump relationship requires that each step in the cascade must have not only an effi cient "turn-on" mechanism, but also an equally effective "turn-off " mechanism (see below). The functional advantage of such a transduction mechanism is obvious; it produces a sensitive photon counter, very well suited for both the sensitivity and the temporal resolution required by the visual system.
A recent study has presented a quantitative model explaining how bumps emerge from stochastic non-linear dynamics of the signaling cascade. Three essential "modules" govern the production of bumps in this model: (i) an "activation module" downstream of PLC but upstream of the channels, (ii) a "bump-generation module" including channels and Ca 2+ -mediated positive feedback and (iii) a Ca 2+ -dependent "negative-feedback module". The model shows that the cascade acts as an "integrate and fi re" device conjectured formerly by Henderson et al. (2000) much like the generation of spikes. The model explains both the reliability of bump formation and low background noise in the dark and is able to capture mutant bump behavior and explains the dependence on external calcium, which controls feedback regulation (Pumir et al., 2008).

THE PHOTOCHEMICAL CYCLE: THE "TURN-ON" AND "TURN-OFF" OF THE PHOTOPIGMENT BISTABLE PIGMENTS
The G-protein-coupled receptor (GPCR), R, is composed of a 7-TM protein, opsin and the chromophore, 11-cis 3 hydroxy retinal (in Diptera; Vogt and Kirschfeld, 1984). Isomerization of the chromophore by photon absorption induces conformational change in the opsin, which is photoconverted into the dark stable physiologically active photoproduct, M. The action spectrum of this reaction depends on the R type (see above) and spans a wavelength range between UV and green lights. To ensure high sensitivity, high temporal resolution and low dark noise of the photoresponse, the active M has to be quickly inactivated and recycled (Figure 6). The latter requirement is achieved, in invertebrates, by two means: the FIGURE 6 | The photochemical cycle: the "turn-on" and "turn-off" of the photopigment. Upon photoconversion of rhodopsin (R) to metarhodopsin (M), by illuminating with blue light (wavy blue arrow), M is phosphorylated at multiple sites by rhodopsin kinase and the fl y ARR2 binds to phosphorylated M. ARR2 is then phosphorylated by Ca 2+ calmodulin-dependent kinase (CaMKII). Photoconversion of phosphorylated M (Mpp) back to phosphorylated R (Rpp) is achieved by illuminating with orange light (wavy red arrow). Upon photoregeneration of Mpp to Rpp, phosphorylated ARR2 is released and the phosphorylated rhodopsin (Rpp) is exposed to phosphatase activity by rhodopsin phosphatase (encoded by the rdgC gene). Unphosphorylated ARR2 also binds to myosin III (NINAC) in a Ca 2+ calmodulin (Ca-CaM)-dependent manner (modifi ed from Liu et al., 2008;Selinger et al., 1993).
absorption of an additional photon by the dark stable M, which photoconverts M back to R (Hillman et al., 1972(Hillman et al., , 1983, or by a multistep photochemical cycle (Figure 6). The action spectrum of M to R conversion in the R1-R6 cells of Drosophila is in the orange range. The red screening pigment of the Drosophila eye prevents massive conversion of R to M, by formation of a red fi lter, which is preferential for M to R conversion. Genetic removal of the red screening pigment and application of blue light (which is preferentially absorbed by the R state) enables a large net photoconversion of R to its dark stable photoproduct M with a minimal conversion of M back to R (Figure 6). A large net photoconversion of R to M, prevents phototransduction termination at the photopigment level when light is turned off (Minke et al., 1975a). This is because the net photoconversion of R to M exceeds the amount of ARR (see below) and thereby its ability to inactivate M, resulting in a large amount of dark stable M, which does not undergo inactivation and thus remains physiologically active in the dark (Byk et al., 1993;Dolph et al., 1993). This brings the capacity of the phototransduction process to its upper limit and results in a phenomenon called prolonged depolarizing afterpotential (PDA; Hillman et al., 1972Hillman et al., , 1983. Illumination with red light photoconverts M back to R and terminates the PDA after the light is turned off. The PDA protocol has been used effi ciently to screen for phototransduction defective Drosophila mutants (Pak, 1995) and has been widely exploited in studies of Drosophila phototransduction.

THE ROLE OF ARRESTIN IN PHOTOINACTIVATION
The ARR family of proteins plays a key role in regulating the activity of GPCRs (Violin and Lefkowitz, 2007). In Drosophila, two homologues of vertebrate ARR exist, which participate through binding, in M inactivation. Both ARRs undergo light-dependent phosphorylation by Ca 2+ calmodulin-dependent kinase II (CaMKII) originally discovered by Matsumoto (Kahn and Matsumoto, 1997). This phosphorylation is unique to the invertebrate visual ARRs and crucial for ARR dissociation from M (Alloway and Dolph, 1999;Kiselev et al., 2000;Yamada et al., 1990).
The study, which clarifi ed the regulatory role of ARR2, used in vitro assays of ARR2 and M, in Drosophila and Musca eyes. Upon photoconversion of R to M, by illumination with blue light (wavy blue arrow, Figure 6), the fl y ARR2 is found predominantly in the membrane fraction, while photoconversion of phosphorylated M (Mpp) back to phosphorylated R (Rpp), by illumination with orange light (wavy red arrow), result in the detection of ARR2 in the supernatant fraction (cytosol). ARR1 on the other hand, always remains membrane bound. The in vitro studies indicated that the functional role of ARR2 binding to M is to terminate its activity (Byk et al., 1993). The isolation of Drosophila mutant fl y arrestin2 (arr2), enabled demonstrating the physiological effect, in vivo, of ARR2 on the light response (Dolph et al., 1993). Accordingly, these fl ies showed a slow response termination at the macroscopic level ( Figure 5B). Further investigations have shown that single photon absorption in these fl ies results in a train of quantum bumps while in wild-type fl ies it elicits a single bump ( Figure 5B).The train of bumps is a manifestation of the M's incapability to inactivate, and explains the slow response termination seen at the macroscopic level (Scott et al., 1997). Moreover, under the assumption that each bump is produced in a single microvillus, the train of bumps separated by intervals suggests a possible inactivation process of the microvilli .
The binding of ARR2 also protects the Mpp from phosphatase activity (Figure 6). Only upon photoregeneration of Mpp to Rpp, is ARR2 released and the Rpp is exposed to phosphatase activity by rhodopsin phosphatase, encoded by the rdgC gene (Steele et al., 1992). These combined actions are crucial for preventing reinitiating of phototransduction in the dark, as the dissociation of ARR2 is coupled to conversion of Mpp to Rpp, thereby directing the protein phosphatase only towards the inactive Rpp (Byk et al., 1993). Subsequent studies have revealed that both CaMKII-dependent phosphorylation of ARR2 at Ser366 and photoconversion of Mpp are required to release phosphorylated ARR2. They furthermore showed that the phosphorylation of ARR2 is required for its dissociation from Mpp upon photoconversion and that ARR2 phosphorylation prevents endocytotic internalization of the ARR2-Mpp complex by a clathrin-mediated mechanism (Alloway and Dolph, 1999;Alloway et al., 2000;Kiselev et al., 2000).
Upon illumination, ARR2 translocates from the cell body to the rhabdomere, thereby elevating its concentration in the signaling compartment (Byk et al., 1993). This process enables the ARR2dependent inactivation of M, operating in massive photoconversion of R to M in bright daylight, thus preventing response saturation and ensures suffi cient time resolution of the light response. A further study has shown that ARR2 translocation requires a phosphoinositide-mediated interaction with myosin III (NINAC; Lee and Montell, 2004). Interestingly, the electrophysiological phenotype of the ninaC mutant is similar to that of arr2 mutant (Figures 5B,C) and may be the consequence of reduced ARR2 concentration in the rhabdomere caused by the ninaC mutation. A recent study suggests that under low Ca 2+ conditions, ARR2 binding to M is slowed down by its sequestration to NINAC. Accordingly, in physiological conditions, light-induced Ca 2+ infl ux acting via CaM (Ca-CaM), rapidly releases ARR2 from NINAC and allows its binding to M and consequently, M inactivation (Liu et al., 2008; Figure 6).

LIGHT-ACTIVATED G q -PROTEIN: THE ROLES OF G q α AND G q β
It has been well established in photoreceptors of several invertebrate species that photoexcited R activates a heterotrimeric G-protein (Fein, 1986). The fi rst experiments, conducted on fl y photoreceptors, showed that when pharmacological agents, known to activate G-proteins, are applied to Musca photoreceptors in the dark, they mimic the light-dependent activation of the photoreceptor cells (Minke and Stephenson, 1985). Later studies using genetic screens isolated two genes encoding visual specifi c G-protein subunits. These genes, dgq (Lee et al., 1990) and gβe (Dolph et al., 1994), encode a G q α and G q β subunit, respectively. The isolated eye-specifi c DG q α, shows ∼75% identity to mouse G q α, which is known to activate PLC (Lee et al., 1990). The most direct demonstration that DG q α participates in the phototransduction cascade came from studies of mutants defective in G q α which showed highly reduced sensitivity to light. In the isolated Gαq 1 mutant, DG q α protein levels are reduced to ∼1%, while G q β, PLC and R protein levels are virtually normal. The Gαq 1 mutant exhibits a ∼1000-fold reduced sensitivity to light and slow response termination (Scott et al., 1995), strongly suggesting that there is no parallel pathway mediated by the G-protein, as proposed for the Limulus eye (Dorlochter and Stieve, 1997). Manipulations of the DG q α protein levels by the inducible heat-shock promoter made it possible to show a strong correlation between the sensitivity to light and DG q α protein levels, further establishing its major role in Drosophila phototransduction (Scott et al., 1995).
The Drosophila fl y has an eye-specifi c G q β (G q β e ) which shares 50% amino acid identity with other Gβ homologue proteins. Two defective G q β e (Gβ e 1 and Gβ e 2 ) mutants with highly reduced G q β levels were isolated and showed a greatly (∼100-fold) decreased sensitivity to light and slow response termination (Dolph et al., 1994). Studies conducted on these mutants revealed that G q α is dependent on G q βγ for both membrane attachment and targeting to the rhabdomere, suggesting that the decreased light sensitivity of these mutants may result from the mislocalization of the G q α subunit (Elia et al., 2005). Attachment of G q α to G q βγ prevents spontaneous GDP-GTP exchange and anchors G q α to the plasma membrane. Therefore, in Gβ e mutants G q α concentration is highly reduced in the rhabdomere (Figures 7A,B). Analysis of the stoichiometry between the G q α and G q β subunits revealed a twofold excess of G q β over G q α. Genetic elimination of the G q β excess leads to spontaneous activation of the visual cascade in the dark, demonstrating that G q β excess is essential for the suppression of dark electrical activity produced by spontaneous GDP-GTP exchange of G q α. Reestablishing the excess of G q β over G q α, by a double hetrozygote mutant fl y, suppresses the dark electrical activity (Elia et al., 2005; Figures 7C bottom trace and D). These studies show a dual role for G q β: retention of G q α in the signaling membrane and prevention of spontaneous activation of G q α in the dark.
Heterotrimeric G-proteins relay signals between membranebound receptors and downstream effectors. Little is known, however, about the regulation of Gα subunit localization within the natural endogenous environment of a specialized signaling cell. Studies using Drosophila fl ies showed that prolonged lights cause massive and reversible translocation of G q α to the cytosol (Kosloff et al., 2003), in similar manner to light-induced translocation of the vertebrate G t α transducin (Arshavsky, 2003;Trojan et al., 2008). A long exposure to light followed by minutes of darkness resulted in reduction in the effi ciency with which each absorbed photon elicited single photon responses, while the size and shape of each single photon response did not change. To dissect the physiological signifi cance of G q α translocation by light, a series of Drosophila mutants were used. Genetic dissection showed a pivotal role for light-induced translocation of G q α from the signaling membrane and the cytosol. Biochemical studies revealed that the sensitivity to light depends on the membrane G q α concentration, which can be modulated either by light or by mutations that impair its membrane targeting. Thus, long-term adaptation is mediated by the movement of G q α from the signaling membrane to the cytosol, thereby reducing the probability of each photon to elicit a bump (Frechter et al., 2007).

DUAL ROLE FOR LIGHT-ACTIVATED PLC PLC ROLE IN LIGHT EXCITATION
Evidence for a light-dependent G q α-mediated PLC activity in fl y photoreceptors came from combined biochemical and electrophysiological experiments. These experiments, conducted in membrane preparations and intact Musca and Drosophila eyes, showed illumination and G q α-dependent accumulation of InsP 3 and InsP 2 , derived from PIP 2 hydrolysis by PLC (Devary et al., 1987; Figure 4).
The key evidence for the participation of PLC in visual excitation of the fl y was achieved by the isolation and analysis of Drosophila PLC gene, designated no receptor potential A (norpA). The norpA gene encodes a β-class PLC, predominately expressed in the rhabdomeres. Mutant fl ies in the norpA gene show a drastically reduced receptor potential. Transgenic Drosophila, carrying the norpA gene on a null norpA background, rescued the transformant fl ies from all the physiological, biochemical and morphological defects, which are associated with the norpA mutants (Bloomquist et al., 1988). The norpA mutant thus provides essential evidence for the critical role of inositol-lipid signaling in phototransduction, by showing that no excitation takes place in the absence of functional PLC (Bloomquist et al., 1988;Minke and Selinger, 1992). However, the events required for light excitation downstream of PLC activation remain unresolved.

PLC ROLE IN RESPONSE TERMINATION
In general, the cytoplasmic GTP concentration in cells is much higher than GDP, making the inactivation process of Gα by hydrolysis of Gα-GTP to Gα-GDP unfavorable. In order to accelerate the GTPase reaction and terminate Gα activity, a specifi c GAP exists (Mukhopadhyay and Ross, 1999). In vitro studies of mammalian PLC-β1 reconstituted into phospholipid vesicles with recombinant M1 muscarinic receptor and G q/11 (Berstein et al., 1992) have shown that, upon receptor stimulation, the addition of PLC-β1 increases the rate at which G q hydrolyses GTP by three orders of magnitude, suggesting its action as GAP. A reduction in the levels of PLC in mutant fl ies affects the amplitude and activation kinetics of the light response (Pearn et al., 1996), but also mysteriously slows response termination (compare Figure 8A to Figure 8B, lower panels). Biochemical and physiological studies conducted in Drosophila have revealed the requirement for PLC in the induction of GAP activity in vivo. Using several Drosophila norpA mutant fl ies, a high correlation between PLC protein level, GAP activity and response termination was observed (Cook et al., 2000). The virtually complete dependence of GAP activity on PLC provides an effi cient mechanism for ensuring the one photon, one bump relationship (Yeandle and Spiegler, 1973), which is critical for the fi delity of phototransduction in dim light. The apparent inability to hydrolyze GTP without PLC ensures that every activated G-protein eventually encounters a PLC molecule and thereby produces a response by the downstream mechanisms. The instantaneous inactivation of the G-protein by its target, the PLC, guarantees that every Gprotein produces no more than one bump (Cook et al., 2000). This apparently complete dependence of GTPase activity on its activator PLC, in fl ies, differs from the partial dependence of GTPase activity on additional GAP factors in vertebrate phototransduction (Chen et al., 2000). Vertebrate phototransduction depends on specifi c GAPs (Arshavsky and Pugh-EN, 1998;Makino et al., 1999). Accordingly, genetic elimination of regulators of G-protein signaling (RGS) proteins reduces and slows down GAP activity and leads to slow response termination to light (Chen et al., 2000).
The dual action of PLC as an activator and a negative regulator nicely accounts for all features of the PLC-defi cient mutants. A June 2009 | Volume 3 | Article 2 | 9

Katz and Minke
Drosophila photoreceptors and signaling mechanisms striking demonstration of the poor temporal resolution of mutants with reduced PLC levels relative to wild-type fl ies is shown in Figure 8C, which compares the ability of wild type and the PLCdefi cient mutant norpA P76 to discriminate between intense lights of different durations (fl ash, red arrow, pulse, blue line). In contrast to the wild-type fl y, where there is a pronounced difference between the responses to a fl ash compared with a long stimulus, no such difference is observed in the norpA mutants, where the two responses overlap (Figure 8C). This result indicates that the PLC-defi cient mutants cannot discriminate between long and short light stimuli. When PLC levels in the signaling membranes are low relative to the amount of the active G-protein, light induces production of G q α-GTP at a higher rate than it is inactivated by PLC. The G q α-GTP that has accumulated during illumination continues to produce bumps in the dark until all active G q α-GTP molecules are hydrolyzed via GAP activity of the scarce PLC ( Figure 8B, lower panel, inset). Hence, the fl ies' temporal resolution is reduced and become virtually blind at low levels of PLC (Cook et al., 2000).

THE PHOSPHOINOSITIDE (PI) CYCLE
In the phototransduction cascade of Drosophila, light triggers the activation of PLCβ. This catalyzes hydrolysis of the membrane phospholipid PIP 2 into water soluble InsP 3 and membrane-bound DAG (Berridge, 1993). The continuous functionality of the pho-toreceptors during illumination is maintained by rapid regeneration of PIP 2 in a cyclic enzymatic pathway (the PI pathway, Figure 9). Moreover, the PI pathway has emerged to be most important for activation of the TRP and TRPL channels (Hardie, 2003;Raghu and Hardie, 2009). This slow response termination can be resolved into continuous production of bumps in the dark at a later time (inset, at higher magnifi cation).
(C) Electroretinogram (ERG) responses showing superimposed traces recorded from wild-type and norpA P76 (a weak norpA allele) to a brief fl ash (red arrow) and continuous light. The graph plots the relative steady state amplitude of the ERG to prolonged lights as a function of relative light intensity. The ERG responses of norpA P76 to a brief fl ash and to continuous light are indistinguishable.

FIGURE 9 | The phosphoinositide cycle.
In the phototransduction cascade, light triggers the activation of phospholipase Cβ (PLCβ). This catalyzes hydrolysis of the membrane phospholipid PIP 2 into InsP 3 and DAG. DAG is transported by endocytosis to the endoplasmic reticulum and inactivated by phosphorylation converting it into phosphatidic acid (PA) via DAG kinase (DGK) and to CDP-DAG via CDP-DAG syntase. Subsequently, CDP-DAG is converted into phosphatidylinositol (PI), which is transferred back to the microvillar membrane, by the PI transfer protein. PIP and PIP 2 are produced at the microvillar membrane by PI kinase and PIP kinase, respectively. There are probably two PIP kinases (PIPK I, PIPK II, which are unifi ed in the scheme). PA can also be converted back to DAG by lipid phosphate phosphohydrolase. PA is also produced from phosphatidylcholine (PC) by phospholipase D (PLD). DAG is also hydrolyzes by DAG lipase into poly unsaturated fatty acids (PUFA).
The phospholipid branch of the PI cycle, following PLC activation, begins by DAG transport through endocytosis to the endoplasmic reticulum (SMC) and subsequently, inactivation by phosphorylation and conversion into phosphatidic acid (PA), via DAG kinase (DGK), encoded by the retinal degeneration A (rdgA) gene (Masai et al., 1993(Masai et al., , 1997. Then, CDP-DAG syntase encoded by the cds gene (Wu et al., 1995) produces DAG-CDP from PA. Both RDGA and CDS are located in the SMC (Figure 1C). Subsequently, DAG-CDP is converted into phosphatidylinositol (PI), which is transferred back to the microvillar membrane, by the PI transfer protein (PITP), encoded by the rdgB gene (Vihtelic et al., 1991) located in the SMC. PIP and PIP 2 are produced at the microvillar membrane by PI kinase and PIP kinase, respectively. PA can be reconverted back to DAG by lipid phosphate phosphohydrolase, LPP, also designated phosphatidic acid phosphatase, PAP, encoded by the laza gene (Garcia-Murillas et al., 2006;Kwon and Montell, 2006) or produced from phosphatidylcholine (PC) by phospholipase D, PLD, encoded by the Pld gene (LaLonde et al., 2005). DAG is also hydrolyzed by DAG lipase encoded by the inaE gene (Leung et al., 2008) predominantly localized outside the rhabdomeres, into polyunsaturated fatty acid (PUFA, Figure 9).
Mutations in most proteins of the PI pathway result in retinal degeneration. For example, rdgA mutant fl ies show light-independent retinal degeneration, thought to occur due to a sustained Ca 2+ infl ux through the light-activated TRP and TRPL channels, making the PI pathway crucial for understanding phototransduction and TRP channels activation. Although it is possible to partially rescue the degeneration phenotypes by reducing the level of TRP (Raghu et al., 2000b), it is still unclear whether this mutation promotes channel opening directly or through an indirect change in the photoreceptor, leading to channel opening.

THE LIGHT-ACTIVATED CHANNELS, TRP AND TRPL, THE FOUNDING MEMBERS OF THE TRP SUPERFAMILY THE trp MUTANT AND THE DISCOVERY OF THE TRP CHANNEL
A spontaneously occurring Drosophila mutant, showing a decline in the receptor potential to baseline during prolonged illumination (Cosens and Manning, 1969), was designated transient receptor potential (trp) by Minke et al. (1975b) (Figure 10B, right). Minke and Selinger suggested in a review article, that the trp gene encodes a Ca 2+ channel/transporter, mainly because application of the Ca 2+ channel blocker La 3+ to wild-type photoreceptors mimicked the trp phenotype (Minke and Selinger, 1991). The cloning of the trp locus by Montell and Rubin (1989) revealed a novel membrane protein. The available sequence of the trp gene led, several years later, to the discovery of mammalian TRPs and the TRP superfamily (Wes et al., 1995;Zhu et al., 1995). However, the signifi cance of the trp sequence, as a gene encoding a putative channel protein, was only fi rst appreciated after a trp homolog, the trp-like (trpl) gene was cloned. This was done by a screen for calmodulin-binding proteins which identifi ed a TM protein. A comparison of its TM domain to that of voltage gated Ca 2+ channels and the TRP protein led to the conclusion that this protein is a putative channel protein with high identity to TRP (Phillips et al., 1992). The fi rst direct physiological evidence for the notion that TRP is the major light-activated channel came from a comparative patch clamp study of isolated ommatidia of wild type and the trp mutant (Hardie and Minke, 1992). The use of Ca 2+ indicator dyes and Ca 2+ -selective microelectrodes, directly demonstrated that the TRP channel is the major route for Ca 2+ entry into the photoreceptor cell (Peretz et al., 1994a,b). The fi nal evidence showing that TRP and TRPL are the light-activated channels came from the isolation of a null mutant of the trpl gene and the construction of the double mutant, trpl;trp, which is blind (Niemeyer et al., 1996). A third TRP homolog channel designated TRPγ has been cloned and sequenced (Xu et al., 2000). Heterologous expression in HEK293 cells has revealed a functional channel (Jors et al., 2006;Xu et al., 2000). However, in Drosophila photoreceptors this channel cannot generate any light-activated conductance in isolation as revealed in the trpl;trp double null mutant and therefore its role in phototransduction, if any, is not clear.

BIOPHYSICAL PROPERTIES OF THE TRP AND TRPL CHANNELS
The Drosophila light-sensitive channels, TRP and TRPL, can be studied separately by utilizing the trpl 302 and trp P343 null mutants, respectively (Scott et al., 1997; Figure 10). The channels are permeable to a variety of monovalent and divalent ions including Na + , K + , Ca 2+ and Mg 2+ and even to large organic cations such as TRIS and TEA (Ranganathan et al., 1991). The reversal potential of the light-induced current (LIC) shows a marked dependence on extracellular Ca 2+ indicating a high permeability for this ion. Permeability ratio measurement for a variety of divalent and monovalent ions, determined under bi-ionic conditions, confi rmed a high Ca 2+ permeability of ∼57:1 = Ca 2+ :Cs + in the trpl mutant and ∼4.3:1 = Ca 2+ :Cs + for the trp mutant (Reuss et al., 1997). The large Ca 2+ permeability of TRP is refl ected in its positive reversal potential (E rev ; Figures 10C,D).
The TRP and TRPL channels show voltage-dependent conductance during illumination. An early study revealed that the light response can be blocked by physiological concentrations of Mg 2+ ions (Hardie and Mojet, 1995). The block mainly infl uenced the TRP channel and affected its voltage dependence. Later, detailed analyses described the voltage dependence of heterologously expressed TRPL channels in S2 cells and of the native TRPL channels, using the Drosophila trp null mutant. These studies indicated that the voltage dependence of the TRPL channel is not an intrinsic property, as is thought for some other members of the TRP family, but arises from divalent cations open channel block that can be removed by depolarization. The open channel block by divalent cations is thought to play a role in improving the signal to noise ratio of the response to intense light and may function in light adaptation and response termination (Parnas et al., 2007).
A comparison with voltage-gated K + channels and cyclic nucleotide gated (CNG) channels, postulates that both TRP and TRPL are assembled as tetrameric channels, thus raising the question whether they assemble as homomultimers or as heteromultimers. Since null trp and trpl mutants both respond to light, each can clearly function without the other. However, heterologous co-expression studies and co-immunoprecipitation, led to the suggestion that the TRP and TRPL channels can assemble into heteromultimers (Xu et al., 1997). Detailed measurements of biophysical properties, questioned this conclusion since they found that the wild-type conductance could be quantitatively accounted for by the sum of the conductances determined in the trp and trpl mutants (Reuss et al., 1997). In addition, a study demonstrated that the TRPL, but not the TRP channel reversibly translocates from the rhabdomere to the cell body upon illumination (Bahner et al., 2002) further imply that TRPL assemble as homomers.

LIGHT-REGULATED SUBCELLULAR TRANSLOCATION OF DROSOPHILA TRPL CHANNELS
In neurons the expression pattern of ion channels determines the physiological properties of the cell. Besides regulation at the level of gene expression that determines which channels are present in a given neuron, traffi cking of ion channels into and out of the plasma membrane is an important mechanism for manipulating the number of channels at a specifi c cellular site (for reviews see Lai and Jan, 2006;Sheng and Lee, 2001). In Drosophila photoreceptors activation of the phototransduction cascade and the infl ux of Ca 2+ through the TRP channels initiate the translocation of the TRPL but not the TRP channels from the signaling compartment, the rhabdomere, to the cell body (Bahner et al., 2002;Meyer et al., 2006). The TRPL translocation process occurs in two stages, a fast translocation (5 min) to the neighboring stalk membrane and a slow translocation (over 6 h) to the basolateral membrane (Cronin et al., 2006). Thus, the TRPL translocation timescale conforms to day night cycle and act in light adaptation (Bahner et al., 2002). While, Ca 2+ infl ux has been shown to be necessary for TRPL translocation the molecular mechanism and structural determinants of the TRPL involved in translocation, are still unknown. Signal dependent translocation of mammalian TRP channels was found to be a widespread phenomenon (Bezzerides et al., 2004;Kanzaki et al., 1999;Zhang et al., 2005). Nevertheless, many of these researches are conducted on TRP channels expressed in tissue culture cells. This makes the Drosophila photoreceptors a unique system in which TRPL channels translocation can be studied in vivo.

ACTIVATION MECHANISMS OF TRP AND TRPL CHANNELS
It has been well established that hydrolysis of PIP 2 by PLC, encoded by the norpA gene, activates the light-sensitive channels TRP and TRPL in Drosophila photoreceptors. However, the mechanism by which PLC activity results in channels opening is still under debate. Several hypotheses have been presented through the years. (i) The InsP 3 hypothesis, suggested that the elevation of InsP 3 , following PIP 2 hydrolysis, activates the InsP 3 R (InsP 3 receptor) resulting in Ca 2+ store depletion and activation of the channels in a store-operated mechanism . This mechanism of activation has also been suggested for a number of mammalian TRPC channels (Putney, 2007;Yuan et al., 2007). In addition, direct activation of the channels as in the Limulus ventral photoreceptors (Payne et al., 1986) using caged Ca 2+ or InsP 3 to elevate Ca 2+ did not activate the channels (Hardie, 1995;Hardie and Raghu, 1998). Rather, direct application of Ca 2+ in excised inside-out patches inhibits expressed TRPL channels in S2 cells by an open channel block mechanism (Parnas et al., 2007), suggesting an inhibition rather than activation effect of Ca 2+ . Furthermore, genetic elimination of the only InsP 3 R in Drosophila had no effect on the light response (Acharya et al., 1997;Raghu et al., 2000a). Therefore, the InsP 3 hypothesis was abundant. It therefore became evident that the alternative branch of PLC, DAG production should be investigated. The most familiar action of DAG is to activate the classical protein kinase C (PKC) synergistically with Ca 2+ . However, mutations in the ePKC, encoded by the inaC gene lead to defects in response termination with no apparent effects on activation compare Figure 11A to Figure 11B). (ii) The PUFA or DAG hypothesis argues that the elevation of DAG and consequently of PUFA acting as second messengers results in channel opening. This hypothesis emerged from a detailed pharmacological study which tested the effect of various fatty acids (including PUFAs) on TRP and TRPL channels activation in vivo and TRPL expressed in Drosophila S2 cells (Chyb et al., 1999). In addition, a detailed analysis of the rdgA mutant encoding DAG kinase has established the importance of the DAG branch in channel activation. This mutant shows light-independent retinal degeneration and constitutive activity of the light-activated channels, while a partial rescue of the degeneration is achieved by eliminating the TRP channel in the double mutant rdgA;trp P343 (Raghu et al., 2000b). Furthermore, it has been shown that the double mutant norpA P24 , rdgA partially rescues the light response in the almost null norpA P24 mutant. This fi nding further supports the hypothesis that DAG or its surrogate PUFA are involved in channel activation . Several lines of evidence challenge this hypothesis: fi rst, application of DAG to intact ommatidia does not activate the channels (unpublished data), while application of DAG analogs 1-oleoyl-2-acetyl-sn-glycerol (OAG) at low concentration (2 µM) in inside-out patches excised from the microvilli of dissociated ommatidia result in activation of the TRP and TRPL channels in kinetics slower by three orders of magnitude (∼60 s after application) compared to the light stimuli (Delgado and Bacigalupo, 2009). Second, the localization of RDGA in the SMC, a relatively distant cellular compartment from the transduction machinery (Masai et al., 1997) makes it unlikely that DAG could act as a second messenger without considerably slowing response termination kinetics, which does not fi t to the fast termination of the response to light. Further establishment of this hypothesis requires identifi cation of a functional-binding domain for DAG or PUFA on the TRP and TRPL channel and further elucidating the complex enzymatic machinery of PUFA production by DAG lipases. Recently, the inaE gene was identifi ed as encoding a homologue of mammalian sn-1 type DAG lipase and was shown to be expressed predominantly in the cell body of Drosophila photoreceptors (Figure 9). Mutant fl ies, expressing low levels of the inaE gene product, have an abnormal light response, while the activation of the light-sensitive channels was not prevented (Leung et al., 2008). The discovery of the inaE gene is a fi rst step in an endeavor to elucidate lipids regulation of the channels (see review, Raghu and Hardie, 2009). Thus, the participation of DAG or PUFAs in TRP and TRPL activation in vivo needs further exploration. (iii) The PIP 2 depletion and DAG accumulation hypothesis argues that PIP 2 acts as a negative modulator, while DAG or its surrogates acts as positive modulators of the TRPL channel. Schilling and colleagues demonstrated in Sf9 cells expressing TRPL that application of DAG or PUFA activates the channels, while application of PIP 2 in inside-out patches inhibit their activity (Estacion et al., 2001). However,  showed that the trp P343 mutant phenotype (in which the light response decays to baseline) is a result of PIP 2 depletion which is not compatible with a PIP 2 inhibitory action. In addition, a PIP 2 -binding domain has not been functionally identifi ed in the TRP or TRPL channels. Together, the above arguments put the PIP 2 depletion and DAG accumulation hypothesis in question. (iv) A recently new hypothesis was formulated, suggesting that plasma membrane lipid-channel interaction controls channel gating. Accordingly, disruption of this interaction by membrane lipid modifi cation through PLC activation causes the opening of the channels (Parnas et al., 2009). It is important to realize that PLC activation, which converts PIP 2 , a charged molecule, containing a large hydrophilic head-group, into DAG, devoid of the hydrophilic head-group, is known to cause major changes in lipid packing and lipid-channel interactions (Janmey and Kinnunen, 2006). It is therefore possible that neither PIP 2 hydrolysis nor DAG production affect the TRP and TRPL channel as second messengers, but rather act as modifi ers of membrane lipid-channel interactions. This may in turn act as a possible mechanism of channel activation (Parnas et al., 2009). This hypothesis evades the two main problems of the DAG or PUFA hypothesis: the need of RDGA at closed proximity to the channels and a channel-binding domain for DAG and/or PUFA. This hypothesis suffers from insuffi cient direct demonstration both in cell expression systems and in vivo.

ORGANIZATION IN A SUPRAMOLECULAR SIGNALING COMPLEX VIA THE SCAFFOLD PROTEIN INAD
An important step towards understanding Drosophila phototransduction has been achieved by the fi nding that some of the key elements of the phototransduction cascade are incorporated into supramolecular signaling complexes via a scaffold protein, INAD (Figure 4). The INAD protein was discovered using a PDA screen which isolated a defective Drosophila mutant (inaD). The fi rst discovered inaD mutant, the inaD P215 , was isolated by Pak (1995) and was subsequently cloned and sequenced by Shieh and Niemeyer (1995). Later studies in Calliphora have shown that INAD binds not only TRP but also PLC (NORPA) and ePKC (INAC) (Huber et al., 1996). The interaction of INAD with TRP, NORPA and INAC was later confi rmed in Drosophila (Tsunoda et al., 1997). It was further found that inaD is a scaffold protein, which consists (A-C) Lower panels: Whole-cell voltage clamp recordings of normalized responses to a 500-ms light pulse of the above mutants. A slow termination of macroscopic response is observed in inaC P209 null mutant and in the inaD P215 mutant in which the binding of INAD to TRP is disrupted (Chevesich et al., 1997;Shieh and Zhu, 1996). of fi ve ∼90 amino acid (aa) protein interaction motifs called PDZ (PSD95, DLG, ZO1) domains. These domains are recognized as protein modules which bind to a diversity of signaling, cell adhesion and cytoskeletal proteins (Dimitratos et al., 1999;Schillace and Scott, 1999) by specifi c binding to target sequences typically, though not always, in the fi nal three residues of the C-terminal. The PDZ domains of INAD bind to the signaling molecules as follows: PDZ1 and PDZ5 bind PLC (Shieh et al., 1997;van Huizen et al., 1998), PDZ2 or PDZ4 bind ePKC (Adamski et al., 1998) and PDZ3 binds TRP (Chevesich et al., 1997;Shieh and Zhu, 1996). This binding pattern is still under debate due to several contradictory reports. Contrary to TRP, TRPL appears not to be a member of the complex, since unlike INAC, NORPA and TRP it remains strictly localized to the microvilli in the inaD 1 null mutant (Tsunoda et al., 1997). Several studies have suggested that, in addition to PLC, PKC and TRP, other signaling molecules such as CaM, R, TRPL and NINAC bind to the INAD signaling complex. Such binding, however, must be dynamic. Biochemical studies conducted in Calliphora have revealed that both INAD and TRP are targets for phosphorylation by the nearby ePKC (Huber et al., 1998). Accordingly, the association of TRP into transduction complexes may be related to increasing speed and effi ciency of transduction events as refl ected by the immediate vicinity of TRP to its upstream activator, PLC, and its possible regulator, ePKC (Huber et al., 1998). Indeed, genetic elimination of INAC affected the shape of the quantum bump of the inaC null mutant, by inducing slow termination of the bump, composed of dumped oscillating current noise of an unclear underlying mechanism Henderson et al., 2000; Figure 11B). Interestingly, a similar phenotype was observed in the inaD P215 mutant, whereby the INAD complex and TRP channel are dissociated (Henderson et al., 2000; Figure 11C), also with a still unclear underlying mechanism.
TRP plays a major role in localizing the entire INAD multimolecular complex. Association between TRP and INAD is essential for correct localization of the complex in the rhabdomeres, as found in other signaling systems (Arnold and Clapham, 1999). This conclusion was derived from the use of Drosophila mutants in which the signaling proteins, which constitute the INAD complex, were removed genetically, and also by deletions of the specifi c binding domains, which bind TRP to INAD. These experiments showed that INAD is correctly localized to the rhabdomeres in inaC mutants (where ePKC is missing) and in norpA mutants (where PLC is missing), but severely mislocalized in null trp mutants (Li and Montell, 2000;Tsunoda et al., 2001), thus indicating that TRP but not PLC or PKC is essential for localization of the signaling complex to the rhabdomere. To demonstrate that a specifi c interaction of INAD with TRP is required for rhabdomeric localization of the complex, the binding site at the C-terminal of TRP was removed or three conserved residues in PDZ3, which are expected to disrupt the interaction between PDZ domains and their targets were modifi ed. As predicted, both TRP and INAD were mislocalized in these mutants. The study of the above mutants was also used to show that TRP and INAD do not depend on each other for targeting to the rhabdomeres. Thus, INAD-TRP interaction is not required for targeting but for anchoring of the signaling complex (Li and Montell, 2000;Tsunoda et al., 2001). Additional experiments on TRP and INAD further showed that INAD has other functions in addition to anchoring the signaling complex. One important function is to preassemble the proteins of the signaling complex. Another important function, at least in the case of PLC, is to prevent degradation of the unbound signaling protein.
A recent study by Ranganathan and colleagues has suggested that the binding of signaling proteins to INAD may be a dynamic process that allows an additional level of phototransduction regulation. Their study showed two crystal structural states of isolated INAD PDZ5 domain, differing mainly by the formation of a disulfi de bond. This conformational change has light-dependent dynamics that was demonstrated by the use of transgenic Drosophila fl ies expressing INAD with a point mutation disrupting the formation of the disulfi de bond. They proposed a model in which, ePKC phosphorylation at a still unknown site promotes the light-dependent conformational change of PDZ5, distorting its ligand-binding groove to PLC and thus regulating phototransduction (Mishra et al., 2007).

CONCLUDING REMARKS
The study of fl y photoreceptors has opened new avenues in biological research, mainly through the exploitation of the power of Drosophila molecular genetics. Processes and proteins that were discovered in Drosophila have been found to be highly conserved through evolution and thus paved the way for the discovery of important proteins and mechanisms in development and cell signaling in mammals. A striking example is the discovery of the TRP channel protein in the Drosophila photoreceptors, which led to the discovery of the widespread TRP superfamily, which plays crucial roles in sensory signaling of insects and mammals. The activation and regulation of Drosophila TRPs by the inositol-lipid signaling pathway and the major role of PLC in the activation of these channels has wide implications for understanding the activation and regulations of mammalian TRPs. Even today Drosophila photoreceptors are one of the few systems in which TRP channels are studied in vivo. Another novel molecule that was discovered in Drosophila photoreceptors is the INAD scaffold protein which forms a supramolecular signaling complex. This protein has introduced new concepts in cell signaling dynamics which are still under investigations. An additional advantage of using the fl y for research on cellular signaling is that frequently the fl y system is less evolutionary evolved relative to mammals, making it simpler to study, while maintaining its core function. It is therefore anticipated that research using the Drosophila sensory and motor systems will continue to identify new proteins and mechanisms of high biological importance.