A dPIP5K Dependent Pool of Phosphatidylinositol 4,5 Bisphosphate (PIP2) Is Required for G-Protein Coupled Signal Transduction in Drosophila Photoreceptors

Multiple PIP2 dependent molecular processes including receptor activated phospholipase C activity occur at the neuronal plasma membranes, yet levels of this lipid at the plasma membrane are remarkably stable. Although the existence of unique pools of PIP2 supporting these events has been proposed, the mechanism by which they are generated is unclear. In Drosophila photoreceptors, the hydrolysis of PIP2 by G-protein coupled phospholipase C activity is essential for sensory transduction of photons. We identify dPIP5K as an enzyme essential for PIP2 re-synthesis in photoreceptors. Loss of dPIP5K causes profound defects in the electrical response to light and light-induced PIP2 dynamics at the photoreceptor membrane. Overexpression of dPIP5K was able to accelerate the rate of PIP2 synthesis following light induced PIP2 depletion. Other PIP2 dependent processes such as endocytosis and cytoskeletal function were unaffected in photoreceptors lacking dPIP5K function. These results provide evidence for the existence of a unique dPIP5K dependent pool of PIP2 required for normal Drosophila phototransduction. Our results define the existence of multiple pools of PIP2 in photoreceptors generated by distinct lipid kinases and supporting specific molecular processes at neuronal membranes.


Introduction
The detection and conversion of external stimuli into physiological outputs is a fundamental property of neurons and depends on intracellular signal transduction pathways. Phosphoinositides, the seven phosphorylated derivatives of phosphatidylinositol are key signalling molecules and of these the most abundant PIP 2 has multiple roles in neurons. Several neuronal receptors (such as the metabotropic glutamate, growth factor and sensory receptors) transduce stimuli into cellular information using the hydrolysis of PIP 2 by phospholipase C enzymes. Additionally, within the context of neuronal cell biology PIP 2 has several roles including cytoskeletal function [1] [2] and several ion channels and transporters (eg: Kir, TRP and Na + /Ca 2+ exchanger ) require PIP 2 for their activity [3]. At the pre-synaptic terminal, a regulated cycle of PIP 2 turnover is essential to regulate synaptic vesicle cycling. Thus PIP 2 plays multiple roles at the plasma membrane of neurons; hence not surprisingly, changes in phosphoinositide metabolism have been linked to several inherited diseases of the human nervous system [reviewed in [4]]. Finally, one of the molecular targets of lithium, used in the treatment of bipolar disorders, is inositol monophosphatase a key regulator of PIP 2 turnover in neurons [5].
Given the multiple functions of PIP 2 at the plasma membrane, it is unclear if a common pool of PIP 2 supports all these functions. Alternatively, if there are distinct pools, it is unclear how these are generated and sequestered on the nanoscale structure of the membrane. In principle, PIP 2 can be generated by the activity of two classes of phosphatidylinositol phosphate kinase (PIPK) enzymes, designated PIP5K and PIP4K; PIP5K phosphorylates PI4-P at position 5 of the inositol ring, whereas PIP4K phosphorylates PI5-P at position 4 [[6]]. Although PIP4K and PIP5K synthesize the same end product, they are not functionally redundant [7] and studies of the mammalian enzymes has defined the molecular basis of substrate specificity [8]. Genes encoding PIP5K are present in all sequenced eukaryotes; however PIP4K appears to be a feature of metazoans; mammalian genomes contain three distinct genes for each of these two activities. However, the functional importance of these two classes of enzymes in generating plasma membrane PIP 2 has remained unclear.
Drosophila photoreceptors are a well-established model for analyzing phosophoinositide signaling in-vivo [9]. In these cells, the absorption of photons is transduced into neuronal activity by G-protein coupled, phospholipase Cβ (PLCβ) mediated PIP 2 hydrolysis [10]. Thus, during phototransduction, PIP 2 needs to be resynthesized to match consumption by ongoing PLCβ activity. PIP 2 turnover is tightly regulated in photoreceptors; mutants in molecules that regulate PIP 2 turnover show defects in phototransduction [11]. However the role of PIPK enzymes in regulating PIP 2 synthesis during phototransduction is unknown. In this study we have analyzed each of the three PIPK encoded in the Drosophila genome that could generate PIP 2 in the context of phototransduction. Our analysis defines three pools of PIP 2 supporting distinct molecular processes in photoreceptors.

Multiple PIP kinases are expressed in the Drosophila eye
In silico analysis of the Drosophila genome sequence revealed that there are four distinct genes that encode open reading frames that include the Interpro domain IPR00002498 which is the "PIP kinase catalytic domain". These include CG6355, CG3682, CG9985 and CG17471. Of these CG6355 encodes a FYVE domain containing protein that is the single ortholog of yeast Fab1, a protein with 1-phosphatidylinositol 3-phosphate 5 kinase activity [12] [13]. CG17471 (dPIP4K) has recently been shown to encode a PIP4K activity that can generate PIP 2 by 4 kinase activity using PI5P as a substrate [14]. The remaining two genes namely CG9985 (sktl) and CG3682 could encode putative PIP5K activity. sktl has been proposed to encode a Drosophila PIP5K [15] [16]. CG3682 is an independent gene that also encodes a putative PIP5K activity. Previous studies have shown that the activation loop region of PIPKs contains specific residues that are conserved among PIP5K and are distinct from PIP4K enzymes [8]. A multiple alignment of PIP5K and PIP4K proteins from mammals and Drosophila reveals that sktl and CG3682 have activation loop residues that are highly diagnostic of those seen in mammalian PIP5K enzymes (Fig. 1A). Both SKTL and dPIP5K show high level of sequence similarity with all the isoforms of mammalian PIP5K. In catalytic domain the identity is more than 80%, whereas the overall sequence homology is from 55-65% with different mammalian isoforms. SKTL is ubiquitously expressed in all organs (Fig. 1B) suggesting its function in many/all cell types. In mammals this kind of expression pattern is evident for α and β isoforms of the PIP5K [17]. By contrast, the γ isoform of PIP5K is mostly expressed in neuronal tissues [18,19] an expression pattern recapitulated by dPIP5K. In addition dPIP5K has multiple splice variants with a conserved catalytic domain and variable C-terminal extensions [19]. The splicing pattern and protein isoforms of dPIP5K so generated as well as its expression pattern (enriched in the adult head Fig. 1B) recapitulates that seen for mammalian PIP5Kγ. The functional significance of the splice variants of dPIP5K remains to be established. In summary it is very likely that the Drosophila genome contains two genes that encode PIP5K activity namely sktl and CG3682. We have named CG3682 as dPIP5K. Thus, collectively there are three phosphoinositide kinases (PIPK), sktl, dPIP5K and dPIP4K all of which could generate PIP 2 .
In order to identify the PIPK that would generate PIP 2 in adult Drosophila photoreceptors, we studied the expression of all three genes. We found that while sktl and dPIP4K were ubiquitously expressed in adult Drosophila, dPIP5K expression was mainly restricted to the head (Fig. 1B). All three genes were expressed in the Drosophila retina; while sktl RNA was present at very low levels and showed no eye-enrichment, dPIP5K and dPIP4K, showed some degree of enrichment in the eye (Fig. 1C).

Loss-of-function mutants in dPIP5K
In order to reveal the function of dPIP5K in vivo, a loss-of-function mutant was generated using ends-out homologous recombination [20]. This results in the insertion of a dominant selection marker (Pw + ) flanked by multiple stop codons within the gene such that the kinase domain of dPIP5K was disrupted and the mutant allele should produce no protein. A total of eight independent knock-out alleles were isolated by following phenotypic markers, genetic mapping and molecular screening. Two of these namely dPIP5K 18 and dPIP5K 30 were studied in detail and are described in this study. All eight alleles were semi-lethal; very few homozygous mutant flies emerged as viable adults. Using a polyclonal antibody generated against a relatively Figure 1. PIP kinase genes in Drosophila genome. A) Multiple alignment of the protein sequences of PIP4K and PIP5K genes. The sequence around the activation loop is presented with the region indicated by a grey line. Amino acid residues common between PIP5K and PIP4K proteins are marked in blue; residues unique to PIP5K are marked in yellow and unique residues for PIP4K in grey. The red arrow indicates the single residue described as responsible for the unique substrate specificity of PIP4K and PIP5K. Notations used for gene names are; hs-Homo sapiens; a-alpha; b-beta; g-gamma; CG17471-Drosophila PIP4K. B) RNA expression pattern of PIP kinases in various fly tissues: Qualitative RT (reverse transcription) PCR analysis with RNA extracted from various fly tissues. The tissue sources are labeled above the lanes. '+' denotes +RT and '−' denotes −RT. The corresponding gene names are indicated on the left side of the agarose gel picture. C) Comparative real time PCR analysis showing eye enrichment of dPIP5K and dPIP4K; the X-axis indicates gene names and the Y-axis represents transcript level expression in arbitrary units (A.U). White bars represent expression levels from cDNA samples of wild type fly heads and black bars represent samples from heads of so D (mutants that lack eyes). Values shown are the means ± S.D of three independent samples. p values between wild type and so D samples were determined using an unpaired t-test. The stars represent level of significance (***p< 0.001; **p< 0.01; *p< 0.05) doi:10.1371/journal.pgen.1004948.g001 unique C-terminal region of dPIP5K, we found that dPIP5K 18 and dPIP5K 30 were protein null alleles of dPIP5K ( Fig. 2A).
In this study we also used a protein null allele of dPIP4K (dPIP4K 29 ) that has already been described [14]. Homozygous deletions in sktl (eg: sktl Δ20 ) are larval lethal [15] and analysis using mitotic clones revealed that loss of sktl is also cell lethal in the developing eye. Thus for the analysis of sktl we have used an allelic combination sktl Δ20 / + and over expression of a kinase dead version of SKTL. These are not protein null alleles but represent the most severe alleles of sktl that give viable eyes.
Abnormal light response in dPIP5K 18/30 photoreceptors Since phototransduction in Drosophila involves rapid G-protein coupled PIP 2 hydrolysis, it might be predicted that loss-of-function mutants in a PIPK that generates the PIP 2 (which is the substrate for photoreceptor PLCβ) might also show a defective electrical response to light. We studied the response to light of mutants in all three genes encoding PIP kinases that can in principle generate PIP 2 , namely dPIP5K, sktl and dPIP4K. A widely accepted way to study the electrical response to light is the electroretinogram (ERG), an extracellular recording of lightinduced electrical changes in the eye. Using ERG, we examined the electrophysiological responses of dPIP5K 18 photoreceptors to light. Since very few homozygous mutant adults were obtained, FLP/FRT mediated mitotic recombination was used to obtain mosaic animals in whom the whole eye was homozygous mutant for dPIP5K 18 [21,22]. ERGs were performed on day 0 (< 24hrs post-eclosion) flies. Characteristically, wild type photoreceptors respond with a large depolarization associated with on and off-transients. By contrast, photoreceptors from dPIP5K 18 produced a much smaller receptor potential in response to a stimulus of equivalent intensity associated with a very slow activation kinetics and response termination. Sample traces of voltage changes in response to a 2s stimulus of green light from wild type and mutants are shown (Fig. 2B). This phenotype was seen in all eight knock-out alleles of dPIP5K that we isolated. Responses in dPIP5K 18 displayed abnormal kinetics: the time of rise to the peak of the response was substantially prolonged (Fig. 2F) and the time for decay back to baseline following the end of the light stimulus was also increased (Fig. 2E). dPIP5K 18 photoreceptors did not display any on or off transients that represent synaptic activity at the first synapse between photoreceptors and the brain. An intensity response function analysis using flies of matched eye colour showed that dPIP5K 18 photoreceptors have a reduction in response sensitivity when measured over several log units of light intensity (Fig. 2C,D). Introduction of a genomic rescue transgene in dPIP5K 18 flies was able to largely correct the peak amplitude, response termination as well as restore both "on" and "off" transients completely (Fig. 2B, E, F). These results demonstrate that dPIP5K is required to support a normal electrical response to light in Drosophila photoreceptors.
By contrast light responses were unaffected in dPIP4K 29 [protein null mutant of dPIP4K [14]] photoreceptors (Fig. 2G,H). Although the amplitude of ERG responses from dPIP4K 29 look smaller than controls, this is because these flies are smaller than controls due to a growth defect during larval development [14]. However the kinetics of the light response were only marginally different in dPIP4K 29 . Finally we studied the most severe allele of sktl that gives rise to adult photoreceptors, namely sktl Δ20 /+ (S1A Fig.); these flies gave normal response to light. Further overexpression of a kinase dead version of sktl had no effect on the electrical response to light as measured by ERGs (S1B Fig.). Together these results suggest that dPIP4K and sktl are most likely dispensable for a normal electrical response to light in Drosophila photoreceptors. dPIP5K 18 photoreceptors show no major defects in ultrastructure or levels of transduction proteins A number of mechanisms could account for the abnormal light response in dPIP5K 18 photoreceptors. PIP 2 is known to be an allosteric regulator of a number of proteins involved in both vesicular transport as well as the cytoskeleton. Thus, loss of dPIP5K function might impact photoreceptor structure through defects in these processes and the abnormal light response may be a consequence of abnormal ultrastructure as seen in the case of mutants such as rdgA and rdgB [23].
To test this hypothesis, we studied photoreceptor ultrastructure using transmission electron microscopy (TEM). This revealed that photoreceptors R1-R7 from 0 day old flies were normal in dPIP5K 18 (Fig. 3A). Microvilli were completely intact and showed no vesiculation or blebbing and only minimal defects were seen at the base of the microvilli; however these changes did not increase with age or illumination and rhabdomere structure remained completely intact (Fig. 3B).
A reduced light response could also result from reduction in the levels of key proteins required to support the phototransduction cascade. Thus the abnormal light response in dPIP5K 18 photoreceptors could be due to altered levels of any one of the key proteins required for phototransduction such as Gq and Rh1. Gαq 1 , a severe hypomorph of dGq expressing less than 1% of the wild type protein levels shows more than 1000-fold reduction in sensitivity to light [24]and ninaE mutants characterized by large decrease in the level of Rh1 also show reduced sensitivity to light. Alternatively, the abnormal ERG could also be the consequence of reductions in the level of proteins like NORPA, TRP, INAD which act downstream of photon absorption to generate a normal electrical response to light. This led us to check the protein level of all the key components of the phototransduction cascade in dPIP5K 18 . Western blot analysis (Fig. 3C) showed that neither the levels of Rh1, Gαq, Gβ and Gγ, nor the levels of NORPA, TRP, TRPL, INAD and INAC were appreciably reduced compared to controls. This result suggests that the abnormal light response observed in dPIP5K 18 is not due to reduced levels of any of the key protein required for phototransduction. Further the subcellular localization of Rh1 and TRP were also studied and found to be not different between control and dPIP5K 18 (Fig. 3D). Together, these findings suggest that ultrastructural defects or changes in the levels and localization of the major transduction proteins cannot explain the abnormal electrical response in dPIP5K 18 photoreceptors. The genotypes of the flies are labeled above each lane. The arrows indicate three bands corresponding to three different forms of dPIP5K detected with this antibody in wild type flies. All three bands are missing in both the knockout lines labeled as dPIP5K 18 and dPIP5K 30 . A protein loading control is shown labeled with a black arrow. (B) Representative ERG traces depicting the response of control and dPIP5K 18 photoreceptors to single 2s flash of green light of intensity 3 (cf. X-axis of Fig. 2D). The genotypes corresponding to each trace are indicated on the right side. Scale bar at the bottom shows the axis; X-axis represents time in seconds and Y-axis represents amplitude of the response in mV. The duration of the light pulse is indicated. Genotypes: Control-wild type; Mutant-dPIP5K 18 ; Bac[dPIP5K] represents BAC clone containing the dPIP5K gene. (C) Graphical representation comparing the light response between control and dPIP5K 18 . The X-axis represents increasing light intensity in log units. The Y-axis represents peak amplitude of each response in mV. Error bars: Mean +/− S.D. (D) Intensity response function of the light response in control and dPIP5K 18 flies. Wild type and dPIP5K 18 flies with matched eye color are shown. The X-axis represents increasing light intensity in log units and Y-axis the peak response amplitude at each intensity normalized to the response at the maximum intensity. p values were determined using an unpaired t-test. The stars represent level of significance (***p< 0.001; **p< 0.01; *p< 0.05). Quantification of the decay time (time taken for the amplitude of the ERG response to reach 50% of its peak amplitude) (E) and the rise time of the ERG response (F) in control, dPIP5K 18  dPIP5K regulates light induced PIP 2 dynamics in photoreceptors dPIP5K is a PIP kinase that is predicted to convert PI4P into PIP 2 . To test its requirement in regulating the dynamics of light induced PIP 2 turnover at the rhabdomeral membrane we used a live fly preparation in which a fluorescent biosensor consisting of the PH domain of PLCδ fused to GFP (hereafter called PIP 2 biosensor) is expressed in photoreceptors. When eyes are illuminated with bright light (λ max 488 nm) high rates of PLC activation result in the hydrolysis of PIP 2 and as a consequence the fluorescent biosensor is detached from the membrane and diffuses out of the microvillar cytoplasm resulting in the loss of the fluorescent pseudopupil signal. Under red light illumination that converts metarhodopsin to rhodopsin thus terminating PLCβ activity, PIP 2 levels recover as a consequence of ongoing PIP 2 resynthesis and the PIP 2 biosensor signal in the pseudopupil recovers ( Fig. 4A,B). The kinetics of the fluorescent pseudopupil are blocked in norpA P24 which lacks appreciable PLCβ activity (Fig. 4B). Using this approach we studied the kinetics of light induced PIP 2 turnover in dPIP5K 18 and compared it to wild type controls. This analysis revealed a clear delay in the kinetics of PIP 2 resynthesis in dPIP5K 18 compared to wild type controls and suggests that dPIP5K activity is required to support normal PIP 2 resynthesis following phototransduction (Fig. 4B).
We also tested the effects of overexpressing dPIP5K in adult photoreceptors; the level of protein overexpression was established by Western blots of retinal extracts using an antibody to dPIP5K (Fig. 4D). Overexpression of dPIP5K in photoreceptors resulted in a marked acceleration in the recovery of the PIP 2 biosensor following stimulation with light (Fig. 4E, F). This result implies that dPIP5K is able to regulate the rate of PIP 2 resynthesis following illumination in photoreceptors.
dPIP5K does not support cytoskeletal function at the microvillar membrane PIP5K has been shown to have a role in actin remodeling in both yeast and mammalian systems [25]. In dPIP5K 18 , photoreceptor ultrastructure was largely normal by TEM analysis and growing flies under conditions of bright light illumination did not result in disruption of microvillar ultrastructure. This suggests that the actin cytoskeleton is unaffected by the absence of dPIP5K activity. Further, phalloidin staining suggested that the actin cytoskeleton was largely unaffected in dPIP5K 18 (Fig. 5A). The Ezrin/Radixin/Moesin (ERM) family of proteins are regulated by PIP 2 and act as cross-linkers between cortical actin and plasma membrane thus playing a key role in maintaining membrane projections, such as microvilli and filopodia [26] [27]. ERM proteins are regulated by PIP 2 and in the presence of PIP 2 the active, phosphorylated form of the protein is attached to the membrane; on PIP 2 hydrolysis the proteins are dephosphorylated and the inactive form is released to cytosol [28]. Drosophila has only a single ERM protein, dMoesin, which is required for morphogenesis and maintenance of microvillar structure in photoreceptors. Moesin localizes to the base of the rhabdomere in wild-type flies in the dark [29], but is dephosphorylated and translocates to the cytosol under bright illumination. Using immunolabelling we studied the distribution of p-Moesin in photoreceptors and found this to be no different between controls and dPIP5K 18 (Fig. 5B). Collectively these findings  dPIP5K is not required to support rhodopsin turnover at the microvillar plasma membrane In Drosophila photoreceptors during illumination rhodopsin, the G-protein coupled receptor for light is endocytosed into a vesicular compartment called rhodopsin loaded vesicles (RLV) [30], a key compartment in the turnover of rhodopsin. PIP 2 is an important regulator of multiple steps in the endocytic cycle [31]. To test if dPIP5K supports the synthesis of the PIP 2 pool regulating endocytosis, we observed the number of RLV in photoreceptors from dPIP5K 18 photoreceptors; these were very similar to those in wild type (Fig. 5C). We also tested the effect of dPIP5K 18 on the retinal degeneration phenotype of norpA this has previously been shown to depend on endocytosis. norpA mutants undergo light dependent retinal degeneration due to the accumulation of excessive amounts of metarhodopsin-arrestin2 (Arr2-Rh) complexes stabilised at the microvillar membrane [32,33]. This process has been shown to depend on clathrin-mediated endocytosis, a process requiring a number of PIP 2 dependent steps. To test if this was affected in dPIP5K 18 , we generated double mutants of norpA P24 ;dPIP5K 18 and studied the time-course of light dependent retinal degeneration in comparison with norpA P24 alone. If the PIP 2 pool produced by dPIP5K was required to mediate endocytosis then one might expect reduced endocytosis of Arr2-Rh complexes in dPIP5K 18 photoreceptors thus suppressing the retinal degeneration phenotype of norpA P24 . Such an effect is seen when dynamin function is removed using the shibire mutant that results in suppression of degeneration in norpA [32]. However we found that light dependent retinal degeneration was not suppressed in norpA P24 ; dPIP5K 18 , although the time course of degeneration was marginally slower (Fig. 5D, E). Additionally the number of RLVs in norpA P24 and norpA P24 ; dPIP5K 18 were not significantly different (Fig. 5F). These results suggest that dPIP5K function is not required to support the endocytic turnover of rhodopsin.

Subcellular localization of dPIP5K
Given our prediction that dPIP5K generates PIP 2 that is used as a substrate for light induced PLCβ activity, it is likely that the enzyme is localized to the microvillar plasma membrane where PLCβ is localized. Initial fractionation experiments showed that almost all of the dPIP5K is localized to a membrane fraction (Fig. 6A) where it co-fractionates with key phototransduction proteins such as INAD. We attempted to establish the localization of dPIP5K expressed at endogenous levels using immunocytochemistry; under these conditions the dPIP5K antibody was able to detect the protein localized at the microvillar membrane (Fig. 6B); this kept during the experiment. Black color indicates total dark and red indicates in red light illumination. The time points are labeled above the bar in minute. The detailed experimental procedure is discussed in material and methods section. (B) Fluorescent deep pseudopupil (dpp) imaging to study PIP 2 dynamics using flies expressing the PIP 2 biosensor (see main text for details). The time scale of the imaging is indicated on the top of each panel. Arrows indicate the timing of a 90 ms flash of blue light used for imaging the dpp. Images were acquired from control, dPIP5K 18 and norpA P24 . The genotypes used for the image acquisition are labeled at the left of the image panel. norpA P24, which is a protein null mutant of PLCβ, is used to show the dependence of dpp dynamics on PLCβ activity. (C) Quantitative representation of PIP 2 dynamics. X-axis represents time in minutes between the depleting flash of blue light and the next image acquired. During this period eyes were illuminated in red light. Y-axis represents the level of fluorescence represented as a % of the value in the initial image. Error bars represents mean +/− S.D from five flies. p values were calculated using an unpaired t-test. The stars represent level of significance (***p< 0.001; **p< 0.01; *p< 0.05) (D) Western blot from head extracts depicting the level of dPIP5K protein expression in wild type flies and those overexpressing dPIP5K. The blot was probed with antibody to dPIP5K. Tubulin was used as loading control. (E) Representative images of dpp imaging in control flies and those overexpressing dPIP5K. (F) Quantification of PIP 2 dynamics in flies overexpressing dPIP5K compared to controls. X-axis represents time in minutes and Y-axis represents the level of fluorescence represented as a % of the value in the initial image. Error bars represents mean +/− S.D from five flies. p values were determined using an unpaired t-test. The stars represent level of significance (***p< 0.001; **p< 0.01; *p< 0.05).  was abolished in retinae from dPIP5K 18 photoreceptors that are protein null alleles for this gene. Double labeling experiments showed that dPIP5K co-localizes with Rh1 at the microvillar membrane (Fig. 6C). We exploited a genetic tool [34] that allowed us to elevate the expression level of untagged endogenous dPIP5K in photoreceptors, expressed from the endogenous gene locus. Under these conditions too, we found dPIP5K clearly localized to the microvillar membrane (Fig. 6D). By contrast a dPIP4K::GFP transgene was excluded from the microvillar plasma membrane (Fig. 6E).

Interaction of dPIP5K with rdgB
Photoreceptors of the Drosophila rdgB mutant show defects in the electrical response to light as well as light dependent degeneration. rdgB encodes a large multi domain protein including an N-terminal phosphatidylinositol transfer protein (PITP) domain [reviewed in [35]]. In vitro the PITP domain can bind and transfer phosphatidylinositol (PI) between two membrane bound compartments and it is presumed, though not demonstrated that the PI delivered to the acceptor compartment is the substrate for phosphorylation by PIPKs that generate phosphorylated versions of PI. In the case of PIP 2 this would include the sequential phosphorylation of PI and PI4P by PI4K and PIP5K respectively. Although the precise molecular function of RDGB in photoreceptors in unknown, it has previously been shown that rdgB mutant photoreceptors have a defect in restoring the level of microvillar PIP 2 following transduction triggered by a bright flash of light [36]. Thus rdgB mutants represent an opportunity to test the importance of a potential PIP5K that might generate microvillar PIP 2 required for phototransduction. To test the relevance of dPIP5K in generating PIP 2 required for G-protein coupled PLCβ activity, we generated photoreceptors that are double mutant rdgB 9 ; dPIP5K 18 ; importantly we used the rdgB 9 allele that is a strong hypomorph and expresses a small amount of this protein and therefore has a residual response to light. We compared the light response of rdgB 9 photoreceptors with those of rdgB 9 ; dPIP5K 18 ( Fig. 7A). Under similar conditions, while rdgB 9 photoreceptors have peak ERG amplitudes of ca. 1.5 mV (Fig. 7A), rdgB 9 ; dPIP5K 18 photoreceptors respond with a amplitude of only 0.4 mV (Fig. 7B). This observation suggests that dPIP5K function is required to support the residual light response in rdgB 9 photoreceptors. We also studied a second phenotype of rdgB 9 namely light dependent retinal degeneration and found that, rdgB 9 ; dPIP5K 18 photoreceptors degenerated faster than rdgB 9 alone (Fig. 7C,D). By contrast loss of dPIP4K or sktl did not exacerbate the electrical response to light or the retinal degeneration phenotype of rdgB 9 .

Discussion
The hydrolysis of PIP 2 by PLC in response to receptor activation is a widespread mechanism of signalling at the plasma membrane. In some cells such as neurons, activation of cell surface receptors by neurotransmitter ligands (e.g glutamate, Ach) or sensory stimuli triggers high rates of PLC activation and rapid consumption of PIP 2 . Under these conditions, it is essential that levels of PIP 2 , the substrate for PLC are maintained as failure to do so would likely result in desensitization. In mammalian cells, multiple classes of PIPK, the enzymes that resynthesize PIP 2 Loaded Vesicle (RLV) involved in Rh1 endocytosis and recycling. (D) Rate of photoreceptor degeneration in norpA P24 and norpA P24 ; dPIP5K 18 monitored using optical neutralization. The flies were reared in continuous light at 2300 lux. The X-axis represents age of the flies and the Y-axis represents number of rhabdomere visualized in each ommatidium. Error bars represent mean +/− S.D from 50 ommatidia taken from at least five flies. (E) Representative transmission electron micrographs of retinae from norpA P24 and norpA P24 ; dPIP5K 18 flies showing the degree of preservation of ultrastructure. Images shown are from flies that are three days old gown under the same illumination conditions as for panel D. (F) Quantitative representation of Rhodopsin Loaded Vesicles (RLVs) in control, norpA p24 and norpA p24 ;dPIP5K 18 . The genotype of the fly is shown in the X-axis and Y-axis represents the count of RLVs.
doi:10.1371/journal.pgen.1004948.g005 have been described; yet the contribution of these enzymes to PIP 2 resynthesis following PLC activation during cell signalling in vivo remains unclear. Broadly two classes of PIPK can synthesize PIP 2 have been described; PIP5K that phosphorylates PI4P at position 5 [37] or PIP4K that can phosphorylate PI5P at position 4 [38]. In this study we have analyzed the consequence of loss of each of these two classes of PIPK to resynthesis following PLC mediated PIP 2 depletion during Drosophila phototransduction. Loss of dPIP5K function results in profound defects in the light activated electrical response as well as slower recovery of plasma membrane PIP 2 levels. Conversely overexpression of dPIP5K was able to substantially accelerate the recovery of PIP 2 levels following stimulation with a bright flash of light. We found that dPIP5K is localized to the microvillar plasma membrane, the site at which PIP 2 needs to be produced to support ongoing light induced PLC activity. Finally, we found that loss of dPIP5K enhances the ERG defect in a hypomorphic allele of rdgB, a gene with a well-established defect in the response to light. Collectively these observations strongly suggest that dPIP5K activity underlies the conversion of PI4P to PIP 2 at the microvillar membrane where it is then available as a substrate for light induced PLCβ activity (Fig. 7E). By contrast loss of the only PIP4K enzyme in the Drosophila genome has minimal effects on phototransduction and this enzyme is not targeted to the microvillar plasma membrane. Our findings also imply that dPIP4K activity (and hence the conversion of PI5P into PIP 2 ) is dispensable for maintaining PIP 2 levels during Drosophila phototransduction. This is consistent with a previous study which found no reduction in the levels of PIP 2 in flies lacking dPIP4K function [14]. Our observations validate the conclusion from biochemical studies in mammalian cells that the levels of PI5P are substantially lower than those of PIP 2 and hence it is unlikely to be the source of the majority of PIP 2 in cells [38]. The identity of the PI4K isoform that generates the substrate, PI4P used by dPIP5K remains unknown although a recent study in mammalian systems suggests that PI4KIIIα is likely to be the relevant isoform [39].
Although the ERG response is severely affected in dPIP5K 18 , it is not abolished as seen in null mutants of PLCβ (norpA) that are not able to hydrolyse PIP 2 . Additionally, the resting levels of PIP 2 as detected by the PIP 2 biosensor are comparable to wild type and following a bright flash of light that depletes PIP 2 , its levels do recover albeit at a slower rate than in wild type photoreceptors. Given that dPIP5K 18 is a protein null allele, these observations imply that there must be a second pool of PIP 2 in dPIP5K 18 cells that is able to support phototransduction and microvillar PIP 2 re-synthesis albeit with lower efficiency (Fig. 7F). This second pool of PIP 2 is likely available with low efficiency for PLC activity in the absence of the dPIP5K dependent pool thus accounting for the residual light response and observed PIP 2 dynamics in dPIP5K 18 photoreceptors.
We found that the ultrastructure of dPIP5K 18 photoreceptors was essentially normal. This was particularly surprising given that in addition to phototransduction, PIP 2 at the microvillar  membrane is also expected to regulate multiple processes required to maintain normal microvillar structure including dynamin dependent endocytosis [40][32] as well as cytoskeletal function [41]. However, using multiple readouts we found that molecular readouts of endocytosis and cytoskeletal function were unaffected in dPIP5K 18 photoreceptors (Fig. 5). These observations imply that the PIP 2 required for these processes is not dependent on dPIP5K activity; rather PIP 2 generated by a separate PIPK supports these processes. Thus far, dPIP4K has not been detected on the microvillar plasma membrane, dPIP4K 29 photoreceptors show normal ultrastructure on eclosion and do not undergo light dependent microvillar degeneration; thus dPIP4K is unlikely to be the critical enzyme that generates the PIP 2 required to support dynamin dependent endocytosis, p-Moesin localization or phototransduction. The Drosophila genome encodes an additional PIP5K activity, sktl that is expressed at low levels in the adult retina but is localized to both the microvillar and basolateral membrane and hence could synthesize PIP 2 at both these locations. Complete loss of sktl function is cell lethal and overexpression of sktl in developing photoreceptors results in a severe block in rhabdomere biogenesis [42] whereas overexpression of sktl results in light dependent retinal degeneration in post-development photoreceptors. These findings presumably reflect an essential and non-redundant role for SKTL in supporting fundamental PIP 2 dependent cellular processes such as endocytosis and cytoskeletal function that are not dependent on PIP 2 hydrolysis by PLC. This model is consistent with the cell-lethal phenotype of photoreceptors that are null for sktl and previous studies showing a role for sktl in supporting cytoskeletal function and endocytosis in other Drosophila tissues and processes such as spermiogenesis [43] and oogenesis [44].
Collectively, our observations imply that there are at least two pools of PIP 2 in photoreceptors; one generated by dPIP5K that is required to support a normal electrical response to light but is dispensable for non-PLC dependent functions of PIP 2 in photoreceptors and another that is generated by enzymes other than dPIP5K (most likely SKTL) that is also capable of supporting PIP 2 synthesis during the light response albeit with reduced efficiency. In summary the PIP 2 pool synthesized by dPIP5K is unique in that it is required for a normal light response and apparently dispensable for other PIP 2 dependent functions/processes. It also reflects the existence of distinct/segregated pools of PIP 2 on the same microvillar plasma membrane that are maintained by distinct kinases.
A number of previous studies have shown that in multiple eukaryotic cell types, plasma membrane PIP 2 levels are remarkably stable, undergoing transient fluctuations despite ongoing PLC mediated PIP 2 hydrolysis [36,[45][46][47]. However the reasons for this remarkable finding have remained unclear although pharmacological studies have suggested the importance of PIP 2 resynthesis in this process [47,48]. One potential explanation for this idea is the existence at the plasma membrane of two pools of PIP 2 , a larger but less dynamic pool of that is not normally accessed by PLC and supporting non-PLC dependent functions of this lipid and a second, quantitatively smaller but more dynamic pool that is the substrate for PLC activity. What control, dPIP5K 18 , rdgB 9 and rdgB 9 ; dPIP5K 18 . Y-axis represents mean +/− S.D of the peak amplitude from five flies. (C) Representative optical neutralization images from control, rdgB 9 , dPIP5K 18 and rdgB 9 ; dPIP5K 18 showing the exacerbation of degeneration in rdgB 9 ; dPIP5K 18 compared to rdgB 9 , dPIP5K 18 alone does not show any degeneration. The representative images shown are collected from one-day-old flies maintained in L/D cycle (900 lux). (D) Quantification of the effect of dPIP5K 18 loss of function on photoreceptor structure in rdgB 9 . (E) Schematic diagram showing the light induced PIP 2 cycle in Drosophila photoreceptors. Genes encoding a given enzyme activity where identified are marked in italics. Notations used: PI(4,5)P 2 -phosphatidylinositol 4,5 bisphosphate, norpA-no receptor potential A, PLCβ-phospholipase C beta, DAG-diacylglycerol, DGK-diacylglycerol kinase, rdgA-retinal degeneration A, laza-lipid phosphate phosphohydrolase (PA phosphatase), PA-phosphatidic acid, CDP-DAG-Cytidine diphosphate diacylglycerol; CDS CDP-DAG synthase, PI-phosphatidylinositol, rdgB-retinal degeneration B, PITP-phosphatidylinositol transfer protein, PI(4)P-phosphatidylinositol 4-phosphate (F) Schematic representation of the pools of PIP 2 in Drosophila photoreceptor membranes. Representations are only semi-quantitative. The total pool of PIP 2 in the plasma membrane is shown bounded by the solid black line. The PLC sensitive PIP 2 pool sensitive to light induced PLC activity is shown in the rectangle bounded by the broken/dashed line indicating that PLC will likely also use a non-dPIP5K dependent pool. The basal PIP 2 pool is indicated in green. The major function of each pool is indicated. Enzymes responsible for the synthesis of each pool are marked.
doi:10.1371/journal.pgen.1004948.g007 underpins such pools of PIP 2 ? The existence of separate enzymes that generate unique pools of PIP 2 has been previously suggested but there have been limited experimental studies to support this model. In murine platelets where thrombin induced PIP 2 hydrolysis appears to be dependent on PIP5K1β but not PIP5Kγ [49]; since both these enzymes are expressed in platelets this implies the existence of two pools of PIP 2 in these cells of which the PIP5K1β dependent pool is available for thrombin dependent PIP 2 turnover. This finding together with our study in Drosophila photoreceptors implies that the plasma membrane in general may contain a specific pool of PIP 2 dedicated for the use of receptor dependent PLC signalling and synthesized by a specific PIPK. It is possible that given the high rates of PLC activated PIP 2 turnover at the plasma membrane (such as the microvillar membrane in photoreceptors) eukaryotic cells have evolved a mechanism to generate distinct PIP 2 pool for this purpose so that other PIP 2 dependent functions at the plasma membrane remain unaffected by ongoing receptor activated PIP 2 hydrolysis. It is likely that dPIP5K and mammalian PIP5K1β represent PIP5K enzymatic activities required to support such a pool of PIP 2 at the plasma membrane. It is presently unclear what properties might make dPIP5K more suitable for generating PIP 2 in the context of receptor triggered PLC activity. One possibility is that the kinetic properties of the enzyme encoded by dPIP5K is distinct from that encoded by sktl allowing it to function in the context of high rates of PIP 2 turnover. Alternatively (or additionally) within the nanoscale organization of the microvillar plasma membrane, it is possible that dPIP5K is segregated such that PIP 2 generated by this enzyme is available within molecular distances of the phototransduction machinery. Interestingly, Drosophila photoreceptors contain within their microvillar membrane a macromolecular signalling complex organized by the PDZ domain protein INAD. It is presently not known if dPIP5K is part of a similar complex but the existence of such mechanisms has been previously shown for mammalian PIP5K1γ in the context of focal adhesion function [50,51]. Interestingly, it has been reported that the INAD protein complex that includes PLCβ is recruited to detergent resistant membranes during light stimulation [52] which themselves have been previously implicated in the formation of PIP 2 microdomains and receptor activated PIP 2 turnover [53] [54]. It is possible that the two PIPKs, SKTL and dPIP5K show differential localization to such domains thus generating and segregating such pools of PIP 2 and further studies in this direction are likely to provide insight into this issue. Nevertheless our study provided evidence for the concept of distinct PIPK enzymes as the basis for functionally distinct pools of PIP 2 at the plasma membrane. Further analysis in this system is likely to reveal the molecular basis for the organization of PIP 2 pools at cellular membranes.

Fly culture and genetics
Flies (Drosophila melanogaster) were reared on medium containing corn flour, sugar, yeast powder, and agar along with antibacterial and antifungal agents. Flies were maintained at 25°C and 50% relative humidity. There was no internal illumination within the incubator and the flies were subjected to light pulses of short duration only when the incubator door was opened. When required, flies were grown in an incubator with timed illumination from a white light source (Intensity mentioned in the figure legends of each experiment).

Generation of a dPIP5K antibody
In order to generate an antibody to dPIP5K, the antigenic fragment (an * 250 amino acid unique sequence at the C-terminus of dPIP5K) was expressed as a recombinant protein and purified by affinity chromatography. Polyclonal antibodies were generated in rats using standard immunization protocols.

Generation of dPIP5K knockout by homologous recombination
A knockout of dPIP5K was generated using 'ends-out' homologous recombination [20]. A 5.4 kb sequence of dPIP5K genomic sequence was used to generate the donor construct. It consisted of two pieces of genomic dPIP5K cloned as insert 1 (3.17 kb) and insert 2 (2.3 kb) separated by a marker gene white (Pw + ) which was flanked by stop codons. These targeting sequences were cloned into the vector pW25 [55]. Transgenic flies carrying this construct were generated and used to perform homologous recombination as previously described [20]. Potential recombinants, which were mapped onto chromosome II, were subjected to molecular analyses using a PCR-based method. Finally, eight individual mutant alleles of dPIP5K (termed as PC4, PC5, PC8, PC18, PC30, PC33, PC60, and PC62) were confirmed and one of these dPIP5K 18 was characterized in detail and used in all experiments described in this study. Since homozygous mutants in dPIP5K are semi-lethal during pupal development we recombined dPIP5K 18 onto a chromosome with an FRT site at 42B. This allele was used to generate mosaic animals in which only adult retinae were homozygous mutant [21,22]. dPIP5K genomic rescue fly A BAC clone encompassing dPIP5K,CH321-03B05 (in attB-Pacman-Cm R vector) was obtained from p[acman] resource [56]. This BAC clone was 57.178 Kb long and included the dPIP5K gene with extended 5' and 3' regions having the promoter and most of the regulatory regions of the gene. The presence of dPIP5K in the clone was verified by PCR using specific primers. This clone was microinjected into embryos and inserted via FC31 integration into the VK22attP docking site in the fly genome to generate the wild-type dPIP5K genomic transgene {Bac [dPIP5K]}. Classical genetic crosses were used to move Bac[dPIP5K] into the dPIP5K 18 mutant background. Protein expression from Bac[dPIP5K] was verified using Western blotting using a dPIP5K specific antibody.

Optical neutralization and scoring retinal degeneration
Flies were immobilized on ice, decapitated using a blade and fixed on a glass slide using a drop of colorless nail varnish. It was imaged using 40× oil objective of Olympus BX43 microscope. Quantitation of degeneration was done as previously described [57] Isolation of pure retinal tissue Pure preparations of retinal tissue were collected using previously described methods [58]. Briefly, 0-to 12-hr-old flies were snap-frozen in liquid nitrogen and dehydrated in acetone at −20°C for 48 hr. The acetone was then drained off and the retinae dried at room temperature. They were cleanly separated from the head at the level of the basement membrane using a scalpel blade.

RNA extraction and QPCR
RNA was extracted from Drosophila head using TRIzol reagent (Invitrogen). Purified RNA was treated with amplification grade DNase I (Invitrogen). cDNA conversion was done using SuperScript II RNase H-Reverse Transcriptase (Invitrogen) and random hexamers (Applied Biosystems). Quantitative PCR (QPCR) was performed with the Applied Biosystem 7500 Fast Real Time PCR instrument. Primers were designed at the exon-exon junction following the parameters recommended for QPCR. Transcript levels of the ribosomal protein 49 (RP49) were used for normalization across samples. Three separate samples were collected from each genotype, and duplicate measures of each sample were conducted to ensure the consistency of the data. The primers used for QPCR were as follows:

Subcellular fractionation
Fractionation was performed as described in [52]; in short frozen fly heads were homogenized in lysis buffer (20 mM

Immunohistochemistry
For immunofluorescence, retinae from flies (0-12 hour post eclosion) were dissected under low red light in phosphate buffered saline (PBS) and then fixed in 4% paraformaldehyde in PBS with 1mg/ml saponin in fixing solution for 30 min at room temperature. Fixed eyes were washed 3 times in PBST (PBS with 0.3% TritonX-100) for 10 minutes. The tissues were then incubated in blocking solution [5% fetal bovine serum (FBS) in PBST] for 2 hours at room temperature, after which the tissues were incubated with primary antibodies diluted in blocking solution [anti-p-Moesin-1:200 [29]], anti-Rh1-1:50 (4C5, Developmental Studies Hybridoma Bank), anti-TRP-1:250,anti-dPIP5K-1:100, anti-GFP-1: 200 (Abcam)] overnight at 4°C. Appropriate secondary antibodies conjugated with a fluorophore were used at 1:300 dilutions [Alexa Fluor 633/568 IgG, (Molecular Probes)] and incubated for 2 hours at room temperature. Wherever required, during the incubation with secondary antibody, Alexafluor 568-phalloidin (Invitrogen) was also added to the tissues to stain the F-actin. After three washes in PBST, the tissues were washed in PBS for 10 min, mounted in mounting medium (70% glycerol in PBS). The whole-mounted preparations were viewed under Olympus FV1000 laser scanning confocal microscope.

Electroretinogram recordings
Flies were anesthetized and immobilized at the end of a disposable pipette tip using a drop of low melt wax. Recordings were done using glass microelectrodes filled with 0.8% w/v NaCl solution. Voltage changes were recorded between the surface of the eye and an electrode placed on the thorax. Following fixing and positioning, flies were dark adapted for 6 min. ERG was recorded with 2 second flashes of green light stimulus. Stimulating light was delivered from a LED light source to within 5 mm of the fly's eye through a fiber optic guide. Calibrated neutral density filters were used to vary the intensity of the light source. Voltage changes were amplified using a DAM50 amplifier (WPI) and recorded using pCLAMP 10.2. Analysis of traces was performed using Clampfit (Axon Laboratories).

PIP 2 dynamics
To monitor PIP 2 dynamics in live flies, transgenic flies expressing PH-PLCδ::GFP (PIP 2 biosensor) were anesthetized and immobilized at the end of a pipette tip using a drop of low melt wax and fixed by clay on the stage of an Olympus IX71 microscope. The fluorescent deep pseudopupil (dpp, a virtual image that sums rhabdomere fluorescence from *20-40 adjacent ommatidia) was focused and imaged using a 10× objective. Time-lapse images were taken by exciting GFP using a 90ms flash of blue light and collecting emitted fluorescence. The program used for this purpose was created in Micromanager. Following preparation, flies were dark adapted for seven minutes after which the eye was stimulated with a 90 ms flash of blue light. The blue light used to excite GFP was also the stimulus to rapidly convert the majority of rhodopsin (R) to metarhodopsin (M) thus activating the phototransduction cascade and triggering depletion of rhabdomeric PIP 2 . Between the blue light stimulations, photoreceptors were exposed to long wavelength (red) light that reconverts M to R. The resurgence in dpp fluorescence with time indicates translocation of the probe from cytoplasm to rhabdomere membrane upon PIP 2 re-synthesis. The dpp intensity was measured using ImageJ from NIH (Bethesda, MD, USA). Cross sectional areas of rhabdomeres of R1-R6 photoreceptors were measured and the mean intensity values per unit area were calculated.

Electron microscopy
Samples for TEM were prepared as mentioned in Ref. 44 Retinae were labeled with an antibody to Rh1. Co-localization of both proteins to the microvilli is shown. (D) Quantification of PIP 2 dynamics in flies overexpressing sktl compared to controls. X-axis represents time in minutes and Y-axis represents the level of fluorescence represented as a % of the value in the initial image. Error bars represents mean +/− S.D from five flies. (TIF)