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

Summary

PIP2 has been implicated in multiple functions at the plasma membrane. Some of these require its hydrolysis by receptor-activated phospholipase C, whereas others, such as membrane transport and cytoskeletal function, involve the interaction of the intact lipid with cellular proteins. The mechanistic basis underlying the segregation of these two classes of PIP2 dependent functions is unknown; it has been postulated that this might involve unique pools of PIP2 generated by distinct phosphoinsoitide kinases. We have studied this question in Drosophila photoreceptors, a model system where sensory transduction requires robust phospholipase C mediated PIP2 hydrolysis. We find that the activity of phosphatidylinositol-4-phosphate 5 kinase encoded by dPIP5K is required to support normal sensory transduction and PIP2 dynamics in photoreceptors. Remarkably, non-PLC dependent functions of PIP2, such as vesicular transport and the actin cytoskeleton, were unaffected in dPIP5K mutants. Thus, dPIP5K supports a pool of PIP2 that is readily available to PLC, but has no role in sustaining other non-PLC mediated PIP2 dependent processes. These findings support the existence of at least two non-overlapping pools of PIP2 at the plasma membrane, and provide a platform for future studies of PIP2 regulation at the plasma membrane.

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 PIP₂ 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 PIP₂ by phospholipase C enzymes. Additionally, within the context of neuronal cell biology PIP₂ has several roles including cytoskeletal function [1] [2] and several ion channels and transporters (eg: Kir, TRP and Na⁺/Ca²⁺ exchanger ) require PIP₂ for their activity [3]. At the pre-synaptic terminal, a regulated cycle of PIP₂ turnover is essential to regulate synaptic vesicle cycling. Thus PIP₂ 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 PIP₂ turnover in neurons [5].

Given the multiple functions of PIP₂ at the plasma membrane, it is unclear if a common pool of PIP₂ 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, PIP₂ 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 PIP₂ 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 PIP₂ hydrolysis [10]. Thus, during phototransduction, PIP₂ needs to be resynthesized to match consumption by ongoing PLCβ activity. PIP₂ turnover is tightly regulated in photoreceptors; mutants in molecules that regulate PIP₂ turnover show defects in phototransduction [11]. However the role of PIPK enzymes in regulating PIP₂ synthesis during phototransduction is unknown. In this study we have analyzed each of the three PIPK encoded in the Drosophila genome that could generate PIP₂ in the context of phototransduction. Our analysis defines three pools of PIP₂ supporting distinct molecular processes in photoreceptors.

Results

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 PIP₂ 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 PIP₂.

In order to identify the PIPK that would generate PIP₂ 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 dPIP5K¹⁸ and dPIP5K³⁰ 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 unique C-terminal region of dPIP5K, we found that dPIP5K¹⁸ and dPIP5K³⁰ were protein null alleles of dPIP5K (Fig. 2A).

In this study we also used a protein null allele of dPIP4K (dPIP4K²⁹) 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 PIP₂ hydrolysis, it might be predicted that loss-of-function mutants in a PIPK that generates the PIP₂ (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 PIP₂, namely dPIP5K, sktl and dPIP4K. A widely accepted way to study the electrical response to light is the electroretinogram (ERG), an extracellular recording of light-induced electrical changes in the eye. Using ERG, we examined the electrophysiological responses of dPIP5K¹⁸ 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 dPIP5K¹⁸[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 dPIP5K¹⁸ 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 dPIP5K¹⁸ 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). dPIP5K¹⁸ 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 dPIP5K¹⁸ 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 dPIP5K¹⁸ 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 dPIP4K²⁹ [protein null mutant of dPIP4K [14]] photoreceptors (Fig. 2G,H). Although the amplitude of ERG responses from dPIP4K²⁹ 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 dPIP4K²⁹. 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.

dPIP5K¹⁸ photoreceptors show no major defects in ultrastructure or levels of transduction proteins

A number of mechanisms could account for the abnormal light response in dPIP5K¹⁸ photoreceptors. PIP₂ 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 dPIP5K¹⁸(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 dPIP5K¹⁸ photoreceptors could be due to altered levels of any one of the key proteins required for phototransduction such as Gq and Rh1. Gαq¹, 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 dPIP5K¹⁸. 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 dPIP5K¹⁸ 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 dPIP5K¹⁸ (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 dPIP5K¹⁸ photoreceptors.

dPIP5K regulates light induced PIP₂ dynamics in photoreceptors

dPIP5K is a PIP kinase that is predicted to convert PI4P into PIP₂. To test its requirement in regulating the dynamics of light induced PIP₂ 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 PIP₂ 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 PIP₂ 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, PIP₂ levels recover as a consequence of ongoing PIP₂ resynthesis and the PIP₂ 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 PIP₂ turnover in dPIP5K¹⁸ and compared it to wild type controls. This analysis revealed a clear delay in the kinetics of PIP₂ resynthesis in dPIP5K¹⁸ compared to wild type controls and suggests that dPIP5K activity is required to support normal PIP₂ 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 PIP₂ biosensor following stimulation with light (Fig. 4E, F). This result implies that dPIP5K is able to regulate the rate of PIP₂ 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 dPIP5K¹⁸, 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 dPIP5K¹⁸ (Fig. 5A). The Ezrin/Radixin/Moesin (ERM) family of proteins are regulated by PIP₂ 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 PIP₂ and in the presence of PIP₂ the active, phosphorylated form of the protein is attached to the membrane; on PIP₂ 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 dPIP5K¹⁸(Fig. 5B). Collectively these findings suggest that dPIP5K activity is not required to support cytoskeletal function in adult Drosophila photoreceptors.

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. PIP₂ is an important regulator of multiple steps in the endocytic cycle [31]. To test if dPIP5K supports the synthesis of the PIP₂ pool regulating endocytosis, we observed the number of RLV in photoreceptors from dPIP5K¹⁸ photoreceptors; these were very similar to those in wild type (Fig. 5C). We also tested the effect of dPIP5K¹⁸ 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 PIP₂ dependent steps. To test if this was affected in dPIP5K¹⁸, we generated double mutants of norpA^P24;dPIP5K¹⁸ and studied the time-course of light dependent retinal degeneration in comparison with norpA^P24 alone. If the PIP₂ pool produced by dPIP5K was required to mediate endocytosis then one might expect reduced endocytosis of Arr2-Rh complexes in dPIP5K¹⁸ 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; dPIP5K¹⁸, although the time course of degeneration was marginally slower (Fig. 5D, E). Additionally the number of RLVs in norpA^P24 and norpA^P24; dPIP5K¹⁸ 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 PIP₂ 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 was abolished in retinae from dPIP5K¹⁸ 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 PIP₂ 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 PIP₂ 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 PIP₂ required for phototransduction. To test the relevance of dPIP5K in generating PIP₂ required for G-protein coupled PLCβ activity, we generated photoreceptors that are double mutant rdgB⁹; dPIP5K¹⁸; importantly we used the rdgB⁹ 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 rdgB⁹ photoreceptors with those of rdgB⁹; dPIP5K¹⁸ (Fig. 7A). Under similar conditions, while rdgB⁹ photoreceptors have peak ERG amplitudes of ca. 1.5 mV (Fig. 7A), rdgB⁹; dPIP5K¹⁸ 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 rdgB⁹ photoreceptors. We also studied a second phenotype of rdgB⁹ namely light dependent retinal degeneration and found that, rdgB⁹; dPIP5K¹⁸ photoreceptors degenerated faster than rdgB⁹ 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 rdgB⁹.

Discussion

The hydrolysis of PIP₂ 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 PIP₂. Under these conditions, it is essential that levels of PIP₂, 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 PIP₂ have been described; yet the contribution of these enzymes to PIP₂ resynthesis following PLC activation during cell signalling in vivo remains unclear. Broadly two classes of PIPK can synthesize PIP₂ 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 PIP₂ 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 PIP₂ levels. Conversely overexpression of dPIP5K was able to substantially accelerate the recovery of PIP₂ levels following stimulation with a bright flash of light. We found that dPIP5K is localized to the microvillar plasma membrane, the site at which PIP₂ 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 PIP₂ 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 PIP₂) is dispensable for maintaining PIP₂ levels during Drosophila phototransduction. This is consistent with a previous study which found no reduction in the levels of PIP₂ 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 PIP₂ and hence it is unlikely to be the source of the majority of PIP₂ 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 dPIP5K¹⁸, it is not abolished as seen in null mutants of PLCβ (norpA) that are not able to hydrolyse PIP₂. Additionally, the resting levels of PIP₂ as detected by the PIP₂ biosensor are comparable to wild type and following a bright flash of light that depletes PIP₂, its levels do recover albeit at a slower rate than in wild type photoreceptors. Given that dPIP5K¹⁸ is a protein null allele, these observations imply that there must be a second pool of PIP₂ in dPIP5K¹⁸ cells that is able to support phototransduction and microvillar PIP₂ re-synthesis albeit with lower efficiency (Fig. 7F). This second pool of PIP₂ 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 PIP₂ dynamics in dPIP5K¹⁸ photoreceptors.

We found that the ultrastructure of dPIP5K¹⁸ photoreceptors was essentially normal. This was particularly surprising given that in addition to phototransduction, PIP₂ 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 dPIP5K¹⁸ photoreceptors (Fig. 5). These observations imply that the PIP₂ required for these processes is not dependent on dPIP5K activity; rather PIP₂ generated by a separate PIPK supports these processes. Thus far, dPIP4K has not been detected on the microvillar plasma membrane, dPIP4K²⁹ 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 PIP₂ 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 PIP₂ 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 PIP₂ dependent cellular processes such as endocytosis and cytoskeletal function that are not dependent on PIP₂ 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 PIP₂ 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 PIP₂ in photoreceptors and another that is generated by enzymes other than dPIP5K (most likely SKTL) that is also capable of supporting PIP₂ synthesis during the light response albeit with reduced efficiency. In summary the PIP₂ pool synthesized by dPIP5K is unique in that it is required for a normal light response and apparently dispensable for other PIP₂ dependent functions/processes. It also reflects the existence of distinct/segregated pools of PIP₂ 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 PIP₂ levels are remarkably stable, undergoing transient fluctuations despite ongoing PLC mediated PIP₂ hydrolysis [36,45–47]. However the reasons for this remarkable finding have remained unclear although pharmacological studies have suggested the importance of PIP₂ resynthesis in this process [47,48]. One potential explanation for this idea is the existence at the plasma membrane of two pools of PIP₂, 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 underpins such pools of PIP₂? The existence of separate enzymes that generate unique pools of PIP₂ has been previously suggested but there have been limited experimental studies to support this model. In murine platelets where thrombin induced PIP₂ 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 PIP₂ in these cells of which the PIP5K1β dependent pool is available for thrombin dependent PIP₂ turnover. This finding together with our study in Drosophila photoreceptors implies that the plasma membrane in general may contain a specific pool of PIP₂ 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 PIP₂ turnover at the plasma membrane (such as the microvillar membrane in photoreceptors) eukaryotic cells have evolved a mechanism to generate distinct PIP₂ pool for this purpose so that other PIP₂ dependent functions at the plasma membrane remain unaffected by ongoing receptor activated PIP₂ hydrolysis. It is likely that dPIP5K and mammalian PIP5K1β represent PIP5K enzymatic activities required to support such a pool of PIP₂ at the plasma membrane. It is presently unclear what properties might make dPIP5K more suitable for generating PIP₂ 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 PIP₂ turnover. Alternatively (or additionally) within the nanoscale organization of the microvillar plasma membrane, it is possible that dPIP5K is segregated such that PIP₂ 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 PIP₂ microdomains and receptor activated PIP₂ 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 PIP₂ 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 PIP₂ at the plasma membrane. Further analysis in this system is likely to reveal the molecular basis for the organization of PIP₂ pools at cellular membranes.

Materials and Methods

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).

The wild-type strain was Red Oregon-R. The following fly alleles and insertions were obtained for the experiments described here: so^D, norpA^p24 (Bloomington Stock Center), sktl^Δ20 (Hugo Bellen), rdgB⁹ (R. C. Hardie, Cambridge University), dPIP5K overexpression line- GS200386 (DGRC-Kyoto).

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 dPIP5K¹⁸ 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 dPIP5K¹⁸ 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 ΦC31 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 dPIP5K¹⁸ mutant background. Protein expression from Bac[dPIP5K] was verified using Western blotting using a dPIP5K specific antibody.

Western immunoblotting

Heads from 1 day old flies were homogenized in 2× SDS-PAGE sample buffer followed by boiling at 95°C for 5 min. Samples were separated using SDS-PAGE and electro blotted onto nitrocellulose membrane (Hybond-C Extra; GE Healthcare) using semidry transfer assembly (Bio-Rad). Following blocking with 5% Blotto (Santa Cruz Biotechnology, CA), blots were incubated overnight at 4°C in appropriate dilutions of primary antibodies [anti-α-tubulin (1:5000 dilution; E7 DSHB), anti-Gαq (1:1000 dilution), anti-TRP (1:5000 dilution), anti-Rh1 (1:200,4C5 DSHB), anti-INAD (1:2000) and anti-NORPA (1:1000)]. Protein immunoreacted with the primary antibody was visualized after incubation in 1:10,000 dilution of appropriate secondary antibody coupled to horseradish peroxidase (Jackson Immuno Research Laboratories) for 2 h at room temperature. Blots were developed with ECL (GE Healthcare) and imaged using a LAS 4000 instrument (GE Healthcare).

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:

• RP49 fwd: CGGATCGATATGCTAAGCTGT;RP49 rev: GCGCTTGTTCGATCCGTA;

• dPIP4K fwd: CATCCGTACGTTGTGGAGAG; dPIP4K rev: AGATCCACATCGTTGCTCAG;

• sktl fwd: CTCATGTCCATGTGTGCGTC; sktl rev: TTAATGGTGCTCATCAGTG;

• dPIP5K fwd: AGCAGAGAAAACCGCTTAGG; dPIP5K rev: GGCGATTCACTGACTTATTCC

Subcellular fractionation

Fractionation was performed as described in [52]; in short frozen fly heads were homogenized in lysis buffer (20 mM Hepes, 30 mM NaCl, 5 mM EDTA) with protease inhibitors (Roche) at 4°C. Homogenate was centrifuged at 600×g for 3 min to remove chitonous material. The supernatant was spun at 55,000 rpm (ca. 100K g) for 30 minutes at 4°C (Beckman Ultracentriguge, Optima LE-80K ultracentrifuge, using a SW 50.1 rotor) to separate membrane fraction from cytosol. Equal volume of each fraction were subjected to SDS PAGE and analyzed by Western immunoblotting.

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).

PIP₂ dynamics

To monitor PIP₂ dynamics in live flies, transgenic flies expressing PH-PLCδ::GFP (PIP₂ 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 PIP₂. 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 PIP₂ 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. Briefly eyes were bisected in ice-cold fixative (2.5% glutaraldehyde in 0.1 M PIPES buffer [pH 7.4]). After 10hrs of fixation at 4°C, eyes were washed with 0.1M PIPES, post-fixed in 1% OsO₄ (30min), and stained en bloc in 2% uranyl acetate (1 hr). Eyes were dehydrated in ethanol series and embedded in epoxy. Ultrathin sections (50 nm) were cut and viewed on a Tecnai G2 Spirit Bio-TWIN electron microscope.

Supporting Information

Attachment 1