Phototransduction in Drosophila is mediated by phospholipase C-dependent hydrolysis of PIP2, and is an important model for phosphoinositide signalling. Although generally assumed to operate by generic machinery conserved from yeast to mammals, some key elements of the phosphoinositide cycle have yet to be identified in Drosophila photoreceptors. Here, we used transgenic flies expressing fluorescently tagged probes (P4M and TbR332H), which allow in vivo quantitative measurements of PI4P and PIP2 dynamics in photoreceptors of intact living flies. Using mutants and RNA interference for candidate genes potentially involved in phosphoinositide turnover, we identified Drosophila PI4KIIIα (CG10260) as the PI4-kinase responsible for PI4P synthesis in the photoreceptor membrane. Our results also indicate that PI4KIIIα activity requires rbo (the Drosophila orthologue of Efr3) and CG8325 (orthologue of YPP1), both of which are implicated as scaffolding proteins necessary for PI4KIIIα activity in yeast and mammals. However, our evidence indicates that the recently reported central role of dPIP5K59B (CG3682) in PIP2 synthesis in the rhabdomeres should be re-evaluated; although PIP2 resynthesis was suppressed by RNAi directed against dPIP5K59B, little or no defect was detected in a reportedly null mutant (dPIP5K18).
Phosphoinositides such as PtdIns(4,5)P2 (PIP2) are ubiquitous and vital regulators of numerous cellular functions (Balla, 2013; Di Paolo and De Camilli, 2006; Hilgemann et al., 2001; Payrastre et al., 2001; Rohacs, 2009; Yin and Janmey, 2003), and their metabolism plays vital roles in most cells. In Drosophila photoreceptors, the light response is mediated by a G-protein-coupled phospholipase C (PLC) cascade (Hardie, 2012; Hardie and Juusola, 2015; Katz and Minke, 2009; Montell, 2012; Yau and Hardie, 2009). Here, hydrolysis of PIP2 by PLC results in activation of two Ca2+-permeable cation channels, transient receptor potential (TRP) and TRP-like (TRPL), which mediate the electrical response to light (Hardie and Minke, 1992; Montell and Rubin, 1989; Phillips et al., 1992). Along with other core components of the transduction cascade, rhodopsin, PLC and the channels are localised in a stack of plasma membrane microvilli forming the light-guiding rhabdomere. Light-activated PLC activity in the photoreceptors is extremely powerful, capable of depleting the entire PIP2 pool in the microvillar membrane within ∼1 s, if not controlled by rapid Ca2+ and PKC-dependent negative feedback (Gu et al., 2005; Hardie et al., 2004, 2015, 2001). Appropriate and timely replenishment of PIP2 is essential not only for maintained visual signalling but also for cell integrity, and photoreceptors undergo degeneration in mutants where PIP2 levels cannot be maintained (Sengupta et al., 2013). Although specialised for vision, the phosphoinositide cycle in the Drosophila eye appears to operate via canonical machinery conserved from yeast to mammals. Because of their experimental accessibility and decades of intensive study, Drosophila photoreceptors have long represented an important genetic model for this ubiquitous pathway (Balakrishnan et al., 2015; Hardie, 2012; Hardie and Juusola, 2015; Katz and Minke, 2009; Montell, 2012; Raghu et al., 2012).
Many of the genes for enzymes and other proteins involved in phosphoinositide turnover in Drosophila photoreceptors have been at least putatively identified. However, certain elements, such as the kinases required for PtdIns(4)P (PI4P) and PIP2 synthesis remain to be identified, and for some of the others there is limited direct evidence demonstrating their involvement in phosphoinositide metabolism in situ. The tools most widely used to monitor phosphoinositide turnover in living cells are fluorescently tagged phosphoinositide binding peptides, which rapidly translocate to and from membranes depending on the concentration of their target lipid (Stauffer et al., 1998; Várnai and Balla, 1998; Várnai et al., 2017). Recently, we adapted this methodology to the fly retina (Hardie et al., 2015). After testing several probes, we selected TbR332H, a mutant variant of the pleckstrin homology (PH) domain from the Tubby protein (Hughes et al., 2007; Quinn et al., 2008) as the most reliable probe for PIP2 in the photoreceptors, whilst the P4M domain of Legionella SidM (Hammond et al., 2014) was chosen for PI4P. Importantly, the probes can be quantitatively imaged in the rhabdomeres of intact living flies, allowing completely non-invasive measurements of this subcellular compartment to be made in vivo over many hours.
In the present study, we used these probes, in combination with electrophysiological approaches, to screen candidate genes that might be involved in phosphoinositide turnover in photoreceptors, using a combination of mutants and RNA interference. We identify PI4KIIIα as the kinase isoform responsible for PI4P synthesis in the microvilli. Our results also indicate that to function, PI4KIIIα requires the additional involvement of orthologues of two genes (Efr3 and YPP1), that have been implicated in PI4KIIIα activity in yeast and mammals (Baird et al., 2008; Chung et al., 2015; Nakatsu et al., 2012). Our results also support the identification of two elements previously implicated by other evidence, namely CDP-diaglycerol synthase and PI-synthase. However, our evidence raises questions about the recently reported central role of dPIP5K59B (CG3682) in PIP2 synthesis in the rhabdomeres (Chakrabarti et al., 2015).
Using GFP-tagged lipid probes to monitor phosphoinositide turnover in vivo
To monitor PIP2 and PI4P depletion and resynthesis in vivo, we used flies expressing specific PIP2 and PI4P fluorescently tagged probes (TbR323H-YFP and P4M-GFP) in the major photoreceptor class (R1-R6) using the promoter for the opsin (Rh1) expressed in these cells (Hardie et al., 2015). As previously described (Hammond et al., 2014; Hardie et al., 2015; Hughes et al., 2007), these probes appear to be specific for PIP2 (as opposed to InsP3) and PI4P (as opposed to PIP2) respectively, unlike other probes sometimes used for monitoring PIP2 (e.g. PH domain of PLCδ) or PI4P (e.g. PH-domains from FAPP1, OSH1 and OSH2). In otherwise wild-type flies, these probes indicate that both PIP2 and PI4P are depleted in the rhabdomeres by the blue excitation light with approximately exponential time courses of ∼7 s for PIP2 (using TbR332H) and ∼12 s for PIP (using P4M). Such stimulation also results in saturating, persistent activation of the transduction cascade, causing a prolonged depolarizing afterpotential (PDA) due to conversion of the majority (70%) of rhodopsin (R) to the active metarhodopsin (M) state (for a review, see Hardie, 1985) and PIP2 remains depleted until M is photoreisomerised to R by long wavelength light (Hardie et al., 2015). Following rapid (∼2 s) photoreisomerisation of M to R, recovery of the respective probes to the microvillar membrane (reflecting resynthesis of PI4P and PIP2) can be monitored in vivo by recovery of fluorescence of the deep pseudopupil (DPP) in completely intact animals as a function of time in the dark (Fig. 1). Such recordings yield half times (t½) of recovery of ∼10 s for PI4P and ∼40 s for PIP2 in otherwise wild-type flies.
Because mutations in many of the genes we investigated are lethal at the organismal or cell level, in many cases we resorted to RNAi knockdown using fly stocks from the Vienna Drosophila Resource Center (VDRC). The UAS-RNAi constructs in these flies must be driven by the Gal4-UAS system (Brand and Perrimon, 1993). In order to achieve strong and selective RNAi knockdown in the eye, we crossed the UAS-RNAi lines to GMRGal4 flies in which Gal4 expression is driven by the GMR (glass multiple repeats) promoter, which is strongly expressed in all cells of the developing eye (Hay et al., 1997). These flies also expressed a UAS-RNAi construct for the white (w) gene (GMRGal4, w-UAS-RNAi, referred to here as GMRw), which results in virtually white-eyed flies (Kalidas and Smith, 2002), hence maximizing reporter fluorescence. Recently, we reported that expression of one copy of GMRGal4 induces a number of subtle phenotypes by itself, including a ∼2- to 3-fold reduction in sensitivity, shorter ommatidia and a ∼30% reduction in area of microvillar membrane (Bollepalli et al., 2017). Since PIP2 turnover had not been investigated in GMRGal4 flies, we first characterised PI4P and PIP2 turnover in control GMRw/+ flies using P4M and TbR332H probes. Although behaviour was broadly similar, the half-times for resynthesis of both PI4P (t½=∼25 s cf. ∼10 s in control) and PIP2 (t½=∼60 s cf. ∼40 s in control) were slightly, but significantly (P<0.0001), longer than values in wild-type backgrounds (Fig. 1). The kinetics of PI4P depletion in particular was also noticeably faster (Fig. 1D).
As additional controls, we also separately tested all RNAi lines with apparent phenotypes in PIP2 and/or PI4P resynthesis by crossing them into otherwise wild-type TbR332H and P4M backgrounds (without GMRGal4), but in this case none showed any significant differences to the wild-type controls (Fig. S1). The effectiveness of UAS-RNAi knockdown driven by GMRGal4 was validated by qRT-PCR on retinal tissue dissected from freeze-dried heads in critical lines showing significant phenotypes (Fig. S2 and see Materials and Methods).
Genetic dissection of PI turnover
Drosophila contains identified and candidate genes for all of the respective elements of the canonical PI cycle (Fig. 1I). Of these, we previously showed that both PI4P and PIP2 synthesis were severely compromised in mutants of PITP (rdgB) and DAG kinase (rdgA), providing in vivo evidence for their essential role in the cycle, whereas, as expected, PIP2 hydrolysis (depletion) is prevented in norpA mutants lacking PLC (Hardie et al., 2015). The primary aim of this study was to test the involvement of the remaining candidate genes using specific PI4P and PIP2 probes and thereby provide a more comprehensive overview of the phosphoinositide cycle machinery in the photoreceptors.
The Drosophila genome contains three putative PI4-kinase genes: PI4KIIα (CG2929), PI4KIIIβ (fwd=CG7004) and PI4KIIIα (CG10260), and limited evidence has implicated all three in development and/or degeneration in Drosophila retina (Forrest et al., 2013; Raghu et al., 2009). Which isoform is responsible for maintaining the PI4P pool required for phototransduction in the microvillar (plasma membrane) rhabdomeres is not known; however, we suspected PI4KIIIα, because in other organisms and tissues it has been implicated in synthesis of PI4P at the plasma membrane whereas PI4KIIα and IIIβ have been implicated on endomembranes (Balla et al., 2008; Burgess et al., 2012; Polevoy et al., 2009; Tan et al., 2014). Of the three kinases, PI4KIIIα is also the isoform enriched most strongly in the eye (7-fold enriched, www.Flyatlas.org).
Because a null mutation (PI4KIIIαΔ123) is lethal, we first tried to generate PI4KIIIα null whole-eye mosaics using the FLP-FRT method (Stowers and Schwarz, 1999); however, this resulted in failure of eyes to develop, indicating the mutation was also cell lethal in the developing eye. We also investigated PI4P and PIP2 resynthesis in PI4KIIIαΔ123 heterozygotes generated by crossing PI4KIIIαΔ123/FM7 flies to flies expressing either P4M or TbR332H probes (Fig. S3). PIP2 resynthesis time courses measured in PI4KIIIα/+ heterozygotes using the TbR332H probe were indistinguishable from controls (whether the parent TbR332H line or FM7/+ siblings). PI4P resynthesis measured using P4M-GFP in PI4KIIIα/+ heterozygotes appeared to be marginally slower than that in sibling controls (t½=11.3 s cf. 9.8 s in controls), but, although significant (P=0.01), because the difference was so small, we could not exclude this being due to other differences in genetic background.
We therefore used RNA interference with UAS-RNAi constructs driven by GMRGal4 (Fig. 2). RNAi knockdown of either PI4KIIIβ (fwd) or PI4KIIα had no detectable influence on the time course of either PI4P or PIP2 resynthesis, with the respective probes recovering with half times of ∼30 s (P4M) and ∼50 s (TbR332H), which were indistinguishable from GMRw controls. However, RNAi knockdown of PI4KIIIα (reducing transcript levels to ∼8%, see Fig. S2), profoundly delayed recovery of both PI4P and PIP2 (t½=∼500 s), suggesting that PI4KIIIα is indeed the main isoform responsible for PI4P synthesis in the rhabdomeres. This also appears to be the PI4P pool required for phototransduction because slow recovery of the probes was mirrored by a similarly slow recovery of the electrical response to light (ERG) after PI4P and PIP2 depletion induced by 30 s intense blue illumination. In fully dark-adapted flies, the response to light was relatively normal, with only a minor reduction in ERG amplitude and sensitivity and reduced synaptic ‘on’ and ‘off’ transients compared with controls (Fig. 3).
We were concerned that PI4KIIIα knockdown might have adversely affected development of the eye; however, there were no obvious structural differences compared with GMRw/+ controls: facet patterns showed only occasional minor irregularities, which can also be seen in GMRw/+ eyes. In terms of photoreceptor structure, dissociated ommatidia prepared for whole-cell recordings were of similar appearance to those from GMRw/+ controls, and whole-cell capacitances (a sensitive measure of microvillar surface area) were indistinguishable from those in GMRw/+ controls (Fig. S4E). Macroscopic light-induced currents had normal kinetics and at most only slightly (but not significantly) reduced amplitudes compared with GMRw/+ controls – although both were reduced, and more variable compared with the wild type (Fig. S4A,B), as previously reported for GMRGal4 flies (Bollepalli et al., 2017). Single-photon responses (quantum bumps) also had normal waveforms, albeit slightly reduced in amplitude (Fig. S4C,D). These results indicate that the photoreceptors and the basic molecular components of the transduction cascade were essentially intact.
Membrane localization and function of PI4KIIIα in yeast and mammals is reported to depend upon association with accessory scaffolding proteins including Efr3, YPP1 and Sfk1 (Baird et al., 2008; Chung et al., 2015; Nakatsu et al., 2012). Interestingly, the Drosophila orthologue of Efr3 is rolling black out (rbo), which was originally suggested to be a DAG lipase by sequence homology of a putative lipase domain (Huang et al., 2004). However, RBO protein has recently been reported to co-immunoprecipitate with PI4KIIIα in Drosophila, and has been suggested to be important in controlling PI4P levels in neurons (Zhang et al., 2017). Null rbo mutations are lethal, but a temperature-sensitive allele (rbots) was reported to have use-dependent defects in the ERG – namely a profound loss of sensitivity to light following bright illumination (Huang et al., 2004). In view of the identification of Efr3 as a scaffolding protein for PI4KIIIα, we asked whether this use-dependent loss of sensitivity might reflect loss of PI4KIIIα function and hence failure to resynthesise PI4P and PIP2.
To address this, we expressed TbR332H and P4M probes in rbots mutants and made measurements at permissive (21-23°C) and restrictive temperatures (∼37°C). We first compared dark-adapted fluorescence intensities of the reporters at 37°C and 22°C in wild-type backgrounds and found that fluorescence at 37°C was slightly decreased (to ∼80% of the level at 22°C) for both probes, consistent with the known temperature dependence of GFP fluorescence (Zhang et al., 2009). For controls, we then investigated the temperature dependence of PIP2 and PI4P turnover in a wild-type background. As might be expected, both PIP2 and PI4P depletion and resynthesis were accelerated at higher temperatures in wild-type flies (Fig. 4). Thus, translocation of TbR332H out of the rhabdomere during blue excitation (reflecting PIP2 depletion), had a time constant of ∼2 s at 37°C, compared with ∼7 s at room temperature (Fig. 4A). Resynthesis was also accelerated with half-times of fluorescence recovery of ∼20 s at 37°C and ∼40 s at room temperature (Fig. 4D). Depletion and resynthesis of PI4P as monitored by P4M were similarly accelerated (Fig. 4C,E). Judging from the Fmax/Fmin ratios, the absolute dark-adapted level of both PIP2 and PI4P at 37°C appeared to be little affected (∼95% of levels at 22°C).
In rbots mutants, depletion and recovery time courses of PI4P and PIP2 at 22°C determined using P4M and TbR332H probes were similar to those measured in a wild-type background, although P4M recovery in particular was slightly but significantly slower (Fig. 4F). However, when flies were exposed to the restrictive temperature, behaviour was profoundly altered (Fig. 4). Fluorescence of the P4M probe declined rapidly in the dark upon raising the temperature to 37°C and within 3 min of activating the heating coil was already reduced to less than 10% of control levels. With shorter heating periods (90 s to 2 min), more residual P4M fluorescence remained, which then decayed to baseline with a similarly accelerated time course to wild-type controls at 37°C, suggesting that PLC activity was normal. However, thereafter, no recovery of fluorescence could be detected in the dark (after the usual M>R reconverting red light) and in most cases, fluorescence decreased even further (Fig. 4C,E). This indicates that resynthesis of PI4P in the microvilli is blocked at the restrictive temperature in rbots, consistent with the notion that RBO (Efr3) is an essential co-factor for PI4KIIIα function.
PIP2 measured using the TbR332H probe behaved in a broadly similar fashion, but with some interesting differences. Firstly, PIP2 (TbR332H fluorescence) remained at >50% of the original room temperature levels in the rhabdomeres of rbots eyes for at least 5 min at the restrictive temperature in the dark. PIP2 was then depleted by blue excitation as in wild-type controls, with a similar acceleration in time course (Fig. 4B), again confirming that PLC activity was intact in rbots at the restrictive temperature. After this initial exposure to blue excitation, there was usually a modest (∼25%) recovery of fluorescent probe in the dark over ∼200 s (Fig. 4B,D). However, in marked contrast to recovery in control conditions, rather than inducing the usual decay of any recovered fluorescence, blue excitation now resulted in a slight increase in fluorescence with each repeated exposure (Fig. 4B), with an exponential time constant similar to but slightly slower than the initial depletion (3.0±0.3 s, n=11). Whilst the slight recovery of fluorescence might be taken to indicate partial resynthesis of PIP2, this would presumably have to come from a source other than PI4P (at least distinct from any pool detected by P4M), whilst the paradoxical increase induced by each light exposure seems hard to explain. An alternative interpretation is that PIP2 in the rhabdomere remained unreplenished like PI4P, but after the initial depletion of PIP2 in the rhabdomere by the first episode of blue excitation, each subsequent episode now results in depletion of PIP2 from plasma membrane outside the microvilli. This could then result in redistribution of some of the probe back to the microvilli, because the probe is expected to report the relative difference in PIP2 levels between competing sinks (i.e. the rhabdomere and the rest of the plasma membrane). On return to room temperature, turnover of both PI4P and PIP2 remained blocked for at least 30 min, but eventually recovered after a couple of hours. We also investigated both PI4P and PIP2 resynthesis with P4M and TbR332H probes at room temperature following rbo RNAi knockdown. Here, we found a very pronounced and significant slowing of both PI4P and PIP2 resynthesis (t½=248 s for P4M and t½=173 s for TbR332H; Fig. 2).
As with PI4KIIIα knockdown, the slow or absent recovery of the probes in rbots and rbo RNAi backgrounds was mirrored by very slow and limited recovery of sensitivity following bright illumination, as measured in the ERG (Fig. 3D). Indeed in rbots, the majority of flies (9/12) showed no detectable recovery of sensitivity for at least 10 min following PIP2 depletion at 37°C.
In yeast and mammals, a second protein (YPP1 in yeast, TTC7 in mammals) is also involved in the scaffolding complex required for PI4KIIIα function (Wu et al., 2014). There appears to be just one orthologue of YPP1 in Drosophila, annotated as CG8325, but with no reported function. According to Fly Atlas (Flyatlas.org), CG8325 is relatively enriched (∼3-fold) in both eye and brain tissue. Mutants of CG8325 [l(2)k14710] are lethal, therefore we sought evidence for its role by RNA interference using GMRGal4 to drive UAS-YPP1-RNAi in flies expressing P4M and TbR332H. As shown in Fig. 2B,D, resynthesis of both PI4P and PIP2 was severely compromised in these flies. Again, this slow recovery was mirrored in a prolonged loss and slow recovery of sensitivity of the light response after bright light adaptation, as measured in the ERG (Fig. 3).
Recently, a third regulator of the PI4KIIIα complex was reported, namely Sfk1 (yeast) and its homologue TMEM150A in mammals, which was reported to be required for association of YPP1 with the Efr3/PI4KIIIα complex (Chung et al., 2015). The most closely homologous genes in Drosophila are CG4025 (∼30% amino acid identity to TMEM150A) and CG7990 (∼20% identity), both of which are enriched in eye and brain tissue (Flyatlas.org). However, using two independent RNAi lines for each gene, we found no obvious effect of either on PI4P or PIP2 resynthesis (Fig. 2B,D).
CDP-diglyceride synthase (cds)
Drosophila cdsA was the first eukaryotic CDP-diglyceride synthase to be cloned and the sole representative in the Drosophila genome. It is highly enriched in the retina and has been implicated in the photoreceptor phosphoinositide cycle on the basis that sensitivity to light in the hypomorphic cdsA1 mutant cannot be maintained during and following prolonged exposure to light (Wu et al., 1995). In addition, more direct evidence for cdsA in microvillar PIP2 resynthesis came from whole-cell recordings from dissociated photoreceptors in cdsA1 mutants using a genetically targeted electrophysiological PIP2 biosensor, Kir2.1 (Hardie et al., 2002). To test the requirement of cdsA for phosphoinositide recycling in vivo we used two independent cdsA RNAi lines crossed to GMRw flies expressing P4M and TbR332H probes. In one line (cdsKK) PI4P recovery was undetectable, whilst recovery of PIP2 was extremely slow (t½=∼800 s, Fig. 2). Sensitivity to light and recovery from light adaptation were also greatly compromised in ERG recordings (Fig. 3). In the second line (cdsGD), PIP2 recovered to ∼50% of original levels relatively quickly (within ∼2 min), but then took a further ∼10 min to recover to pre-depletion levels, but little or no effect was seen on PI4P recovery times. In these flies, sensitivity to light and recovery from light adaptation in ERG recordings were at most only slightly impaired compared with controls, which is to be expected since 50% PIP2 levels are sufficient to mediate near-maximal activation of TRP channels (Hardie et al., 2015).
dPIS is the only recognisable homologue of mammalian PI-synthase in the Drosophila genome, and has been reported to function as a PI-synthase when heterologously expressed in HEK293 cells (Wang and Montell, 2006). It is expressed in photoreceptors and mutants have visual phenotypes, including slow recovery of the light response following light adaptation, and hence it is a prime candidate for the PI-synthase in the photoreceptor PI cycle (Wang and Montell, 2006). Because null dpis mutants are cell lethal, in order to test the requirement of dPIS for PI synthesis in vivo, we again used RNA interference. Resynthesis of both PI4P and PIP2 in dpis-RNAi flies were very substantially slowed (t½=250-400 s; Fig. 2B,D), supporting the essential role for dPIS in the photoreceptor phosphoinositide cycle, as previously proposed (Wang and Montell, 2006).
In the canonical phosphoinositide cycle (Fig. 1I), the final step in the resynthesis of PIP2 is conversion of PI4P to PIP2 via PIP5-kinase. The Drosophila genome contains genes for two putative PIP5K isoforms: skittles (sktl) and dPIP5K59B (CG3682, shortened hereafter to dPIP5K). Both are expressed in the eye, and dPIP5K at least has been reported to immunolocalise to the rhabdomere (Chakrabarti et al., 2015). Severe or null mutations of both genes are organismal lethal or semi-lethal. Null sktl mutations are also cell lethal, but it is possible to generate viable whole-eye mosaics of a reportedly null dPIP5K mutant (dPIP5K18) using the FLP-FRT method (Chakrabarti et al., 2015). Using such mosaics, Chakrabarti et al. (2015) recently reported defects in the electrical light response as well as PIP2 resynthesis and concluded that dPIP5K is the main kinase responsible for the synthesis of PIP2 required for phototransduction in the rhabdomeres. However, technical issues (Hardie et al., 2015 and see Discussion) led us to question some of their findings. We therefore re-examined PIP2 resynthesis in dPIP5K18 eye mosaics using flies provided by the authors of the Chakrabarti et al. (2015) study, confirming their genotype by the effective absence (<1%) of dPIP5K mRNA in dissected retinae using qRT-PCR (Fig. S2C).
In marked contrast to Chakrabarti et al. (2015), when we monitored PIP2 in dPIP5K18 mosaic eyes using TbR332H, we found robust PIP2 resynthesis with only a slight slowing of the time course of recovery, which was significant with respect to that in the wild type (t½=∼49 s and 38 s, respectively; P=0.0002), but not with respect to sibling heterozygote controls (t½=44 s, P=0.06; Fig. 5). Chakrabarti et al. (2015) also reported that overexpression of dPIP5K driven by Rh1Gal4 resulted in a marked acceleration of PIP2 resynthesis; however, when we measured PIP2 resynthesis using TbR332H in the same overexpressing flies, we found the time course slightly slower, although statistically indistinguishable, compared with that in controls (P=0.14, Fig. 5B,C).
Chakrabarti et al. (2015) also reported a ‘profound’ impairment of the photoreceptor response as inferred from ERG recordings in dPIP5K18 eye mosaic mutants. We confirmed a similar reduction in ERG amplitudes (Fig. 6); however, the most conspicuous aspect of the dPIP5K18 phenotype (also noted by Chakrabarti et al., 2015) was the complete lack of the synaptic ‘on’ and ‘off’ transients, which derive from postsynaptic interneurons (large monopolar cells or LMCs). Recently, another mutant defective in synaptic transmission (hdc, which lacks the photoreceptor neurotransmitter histamine) was found to have similarly reduced ERG amplitudes (Fig. 6), suggesting that synaptic feedback from interneurons to photoreceptors normally contributes to the photoreceptor component of the ERG (Dau et al., 2016). Therefore, from these results, it is not clear whether there is in fact any defect at the level of phototransduction. To investigate this directly, we examined dPIP5K18 mutant photoreceptors using whole-cell recordings of light-induced currents from dissociated ommatidia (Fig. 7). Morphologically, the mutant ommatidia appeared normal in appearance, and their capacitances (a sensitive measure of microvillar membrane area) were indistinguishable from wild-type (64±8 pF and 63±5 pF, respectively; n=10-13 cells). Importantly, we failed to detect any phenotype at all in the responses to light: absolute sensitivity, responses to single photons, brief flashes and 1 s steps of increasing intensity were all indistinguishable from the wild type, or dPIP5K18/+ sibling heterozygote controls (Fig. 7). These results indicate that dPIP5K18 mutants have no detectable defects in phototransduction and at most, marginal defects in PIP2 synthesis in the rhabdomere. By contrast, the complete lack of synaptic transients in the ERG suggests that dPIP5K18 mutants may be defective in PIP2 synthesis at synaptic terminals where PIP2 is believed to be critical for normal vesicular exo- and endocytosis, and interacts with key synaptic proteins such as synaptotagmin (Di Paolo et al., 2004; Lauwers et al., 2016; Park et al., 2015).
The dPIP5K18 mutant was generated by a targeted ‘insertion of a selection marker (Pw+) flanked by multiple stop codons within the gene such that the kinase domain of dPIP5K was disrupted’ and lack of protein was confirmed by western blotting (Chakrabarti et al., 2015). Nevertheless, we also proceeded to measure PIP2 resynthesis using UAS-dPIP5K-RNAi (VDRC line 108104KK). Surprisingly, despite the virtual lack of effect of the dPIP5K18 mutation, PIP2 resynthesis measured using TbR332H was in fact markedly slower in UAS-dPIP5K-RNAi flies crossed to GMRw (t½=430 s, Fig. 2A,B). This effect seemed to be specific for PIP2 as PI4P resynthesis measured using P4M was relatively little affected (Fig. 2C,D). This suggests either that the dPIP5K18 mutant was not null as reported (Chakrabarti et al., 2015) or that UAS-dPIP5K-RNAi had influenced PIP2 resynthesis by off-target effects.
The second PIP5K gene in Drosophila is sktl (skittles). Severe alleles, such as sktlΔ1-1 or sktlΔ20 are lethal (Hassan et al., 1998) and attempts to generate whole-eye mosaics with sktlΔ1-1 resulted in failure of eyes to develop, indicating it is cell lethal in the developing eye (see also Chakrabarti et al., 2015). Trans-heterozygotes (sktlΔ1-1/sktlΔ20) are viable, with eyes of normal appearance, but when we expressed TbR332H in sktlΔ1-1/sktlΔ20 trans-heterozygotes, the time course of PIP2 resynthesis was not obviously affected (Fig. S3E). Similarly, we found no obvious slowing of resynthesis in single or double heterozygotes of dPIP5K and sktlΔ1-1 (Fig. S3). Finally, we measured PIP2 resynthesis time courses in UAS-sktl RNAi flies (crossed to GMRw) and found a slight slowing with respect to GMRw controls (t½=88 s and 67 s, respectively), which was significant on a direct t-test, but not on a one-way ANOVA comparing all the RNAi lines (Fig. S3F). However, uniquely amongst the RNAi lines we tested, qRT-PCR of dissected retina revealed no significant knockdown of sktl mRNA in these flies (Fig. S2). In the absence of data from true null alleles or a validated RNAi line, we are reluctant to draw any conclusions from these results, leaving open the possibility that sktl may contribute to PIP2 synthesis in the rhabdomeres.
We have developed transgenic flies and methodology that allow quantitative measurements of phosphoinositide turnover in the microvillar rhabdomeres of completely intact living flies (Hardie et al., 2015). In the present study, we used this approach to screen for candidate genes that might be involved in maintaining phosphoinositide levels in the rhabdomeres. Because little is known about them in the Drosophila eye, much of our study concentrated upon the kinases presumed to be responsible for synthesis of PI4P and PIP2 (PI4-kinase and PIP5-kinase).
The PI4-kinase required for phototransduction had not previously been identified in Drosophila photoreceptors, although PI4KIIα has been implicated in rhabdomere biogenesis (Raghu et al., 2009), whilst downregulation of either PI4KIIIβ (fwd) or PI4KIIIα by RNAi partially rescued retinal degeneration caused by upregulation of PI4P (Forrest et al., 2013). However, PI4KIIIα has consistently been identified as the isoform associated with plasma membrane PI4P pools (Balla et al., 2008; Balla, 2013; Tan and Brill, 2014) and our evidence now strongly supports the identification of PI4KIIIα as the isoform responsible for maintaining the ‘phototransduction pool’ of PI4P in the rhabdomeres. Firstly, using RNAi, we found a marked slowing of both PI4P and PIP2 resynthesis following PI4KIIIα knockdown, whereas RNAi directed against the other two candidates had no effect. This slowing down of PI4P and PIP2 resynthesis was mirrored in a similarly slow recovery of the light response following bright illumination. Secondly, PI4KIIIα activity in both yeast and mammals has recently been found to be critically dependent upon a scaffolding complex including Efr3 and YPP1 (TTC7 in mammals), and mutant and/or RNAi knockdown of the respective Drosophila orthologues of both these genes was found to have very similar effects in blocking or profoundly slowing PI4P and PIP2 resynthesis and recovery of light sensitivity.
The Drosophila orthologue of Efr3 is rolling black out (rbo, also referred to as stmA). Previously, it was found that the temperature-sensitive rbots mutant showed a complete loss of light response following bright illumination (Huang et al., 2004) and an acute blockade of synaptic transmission (Huang et al., 2006), but at that time the only homology noted was to DAG lipase, whilst in vitro biochemistry on whole heads appeared to show an accumulation of PIP2, leading the authors to suggest that RBO lipase activity was somehow required for PLC function. By contrast, our results using both rbots mutants and rbo RNAi indicated that PLC activity (as witnessed by the initial rapid translocation of both TbR332H and P4M probes) was normal at the restrictive temperature, but PI4P (and consequently PIP2) failed to be resynthesised, strongly supporting a role in PI4KIIIα function. A similar block of both PI4P and PIP2 resynthesis was obtained following RNAi knockdown of the Drosophila YPP1 orthologue (CG8325). In yeast and mammals, yet another regulator of the PI4KIIIα complex has been reported (Sfk1, or TMEM150A in mammals). In contrast to rbo and YPP1 the Drosophila genome does not contain an obvious orthologue of this protein, and RNAi knockdown of the two most closely related genes, with 20% (CG7790) or 30% (CG4025) amino acid identity, respectively, had no obvious effect. However, little can be concluded from this negative result and we do not, for example, exclude the possibility that both can contribute in a redundant manner. However, overall, our evidence supports the existence of a PI4KIIIα-Efr3-YPP1 complex in flies and indicates that it is essential for both PI4P and downstream PIP2 synthesis in the microvillar membrane. Because null mutations of all three genes are lethal in the fly, this complex can also be expected to be of more widespread function.
The situation with PIP5-kinase is less clear. Recently, Chakrabarti et al. (2015) concluded that dPIP5K was the key kinase required for synthesis of PIP2 required for phototransduction. They showed that dPIP5K immunolocalised to the rhabdomeres and reported a severe defect in PIP2 resynthesis measured with a fluorescent probe (PLCδ-PH-GFP) in dPIP5K18 mutants. However, even their wild-type recovery times (∼3-4 min) were much slower than should be the case with this probe (∼20 s, see Hardie et al., 2015). This suggests that their measurements were compromised, most likely by failure to rapidly reconvert M to R after blue excitation, so that, in effect, they may have been measuring the rate of M to R photoreisomerisation by continuous red light rather than PIP2 synthesis (see Hardie et al., 2015 for further discussion). Using the TbR332H probe, and ensuring rapid M to R photoreisomerisation, our results reveal only a very minor slowing of PIP2 resynthesis in the same dPIP5K18 mutants. Chakrabarti et al. (2015) also reported ‘profound’ defects in ERG recordings from dPIP5K18 mutants. However, although we confirmed the same ERG phenotype, we attribute this to a defect in synaptic transmission and photoreceptor light responses were quantitatively indistinguishable from controls when studied using more direct whole-cell recording techniques.
Surprisingly, despite the virtual lack of effect of the dPIP5K18 mutation, dPIP5K-RNAi did result in a very pronounced slowing of PIP2 resynthesis. This discrepancy is perplexing and would seem to imply either that dPIP5K-RNAi had unanticipated off-target effects (although this did not appear to include sktl mRNA; Fig. S2B), or that the dPIP5K18 mutation was not null. dPIP5K has multiple transcripts and, although our qRT-PCR primers should have detected all the transcripts reported in the databases (Flybase.org), the existence of an unrecognised transcript that was not eliminated by the mutation cannot be excluded. Another possibility would be compensatory upregulation of alternative gene(s) in the dPIP5K18 mutant, but not in dPIP5K-RNAi flies, although again, this did not appear to include sktl (Fig. S2C).
In conclusion, identification of the PIP-kinase responsible for the synthesis of PIP2 required for phototransduction still requires further investigation. Because of the RNAi effects reported here, dPIP5K (CG3682) still remains a viable candidate; but only, it would seem, on the assumption that dPIP5K18 is not a null mutant. Because we were unable to generate viable sktl null mutants, and because sktl RNAi failed to suppress mRNA (Fig. S2), we cannot exclude a role for sktl, and one possibility is that PIP2 synthesis may be mediated by both dPIP5K and sktl in a redundant manner. In principle, PIP2 could also be synthesised from PI5P via PI5P 4-kinase (dPIP4K). However, synthesis via this route would be difficult to reconcile with the requirement for PI4-kinase (and hence PI4P) for PIP2 synthesis indicated in the present study. Furthermore, null mutants of dPIP4K have essentially normal light responses (Chakrabarti et al., 2015).
We also tested two further genes previously strongly implicated in the PI cycle, namely cdsA and dPIS: in both cases, we confirmed a marked slowing of PI4P and PIP2 resynthesis following RNAi knockdown. Together with previous results (e.g. Hardie et al., 2015), and with the exception of the uncertainty over PIP5-kinase, the forward PI cycle in fly photoreceptors (Fig. 1I) can now be confidently populated with specific genes. However, some reverse steps, in particular PIP2 and PIP phosphatases, which are likely to be important in determining absolute PI4P and PIP2 levels, remain to be identified. Finally, we emphasise that in vivo measurements using TbR332H and P4M are, in principle, simple and routine to perform and should lend themselves to further investigations aimed at identifying and characterising the molecular and cellular machinery underlying this important and ubiquitous pathway.
MATERIALS AND METHODS
Drosophila melanogaster were reared in the dark at 25°C on standard (cornmeal/agar/yeast/glucose) diet. All flies were on white-eyed (w1118) background, but in most cases with one or more Pw+ transfection marker transgenes, resulting in eyes with an orange colour that varied slightly according to the line in question. Stocks used are listed in Table S1. To monitor phosphoinositide levels in the rhabdomeres, we used flies expressing the PIP2-specific probe TbR332H (eYFP tagged), and the PI4P-specific probe P4M (eGFP tagged) in their photoreceptors under control of the rhodopsin (ninaE) promoter (Hardie et al., 2015). For RNA interference, UAS-RNAi constructs in VDRC lines (Table S1) were driven by crossing to flies expressing TbR332H or P4M (on the second chromosome) and GMRGal4 on the third chromosome, together with w-UAS-RNAi to suppress expression of the w gene (combination referred to as GMRw), thus generating flies with very pale orange eye colour. The resulting F1 progeny used for experiments therefore had one copy of GMRw, one copy of the respective UAS-RNAi construct and one copy of TbR332H or P4M. Controls included flies expressing reporters with just one copy of GMRw but no other RNAi construct and also flies expressing one copy of the UAS-RNAi construct (but not GMRw).
The dPIP5K18 mutant and dPIP5K overexpressing line were kindly provided by the authors of the Chakrabarti et al. (2015) study. In the absence of information on the precise location of the disrupting insert, we checked the genotype of dPIP5K18 by qRT-PCR of mosaic retinae, and found that they contained only trace (0.8%) levels of dPIP5K mRNA attributable to contaminating tissue in the dissected retinae (Fig. S2C). In addition, as reported, dPIP5K18 mutants were homozygous lethal/semi-lethal and retained the Pw+ selection marker, whilst mosaic retinae reproduced the ERG phenotype reported by Chakrabarti et al. (2015).
Live imaging of the deep pseudopupil and calibration
Fluorescence from the DPP of intact flies was measured as previously described (Hardie et al., 2015). Briefly, flies were fixed with low melting point wax in truncated pipette tips, mounted on a micromanipulator and observed with a 20×/0.35 NA Fluor objective on a Nikon inverted microscope (Nikon, Kingston-Upon-Thames, UK). The DPP was cropped via a rectangular diaphragm and fluorescence intensity measured using a photomultiplier tube (Cairn Research Ltd, Faversham, UK) collecting fluorescence excited by a blue (470 nm peak) ultrabright LED (Cairn Research) and imaged via 515 nm dichroic and OG515 longpass filters. Fluorescence signals were sampled at ≥100 Hz and analysed using pClamp10 software (Molecular Devices, CA USA). Unless otherwise stated, data were normalised between Fmax (from ‘naïve’ dark-adapted flies or maximum fluorescence after recovery, whichever was greatest) and Fmin (minimum fluorescence after depletion by blue excitation). Photo-reisomerisation of M to R was achieved by long wavelength light delivered by an ultrabright orange/red LED (640 nm; Thorlabs, Ely, UK) via the microscope eyepiece for 2-4 s. In some genotypes, there was sufficient expression of the w gene for there to be significant contribution of the intracellular pupil pigment to the recorded signals. When activated, this rapidly migrates towards the rhabdomeres, causing a rapid (τ=∼1-2 s) decrease in fluorescence. Where this was a potential problem, immediately before each episode of blue excitation, the eye was stimulated for 5-7 s with green (540 nm) light of just sufficient intensity to fully activate the pupil (and hence nullify any differential effect on the fluorescence), but insufficient to cause any significant PIP2 or PI4P depletion. Surprisingly, flies showing a significant pupil effect included those on a GMRw background, despite having only a very pale orange eye colour.
Fly rhodopsin (R) absorbs maximally at 480 nm, and the metarhodopsin state (M) at ∼570 nm. The two states are thermostable, photo-interconvertible and exist in a photoequilibrium determined by their photosensitivity spectra and the spectral content of illumination (Minke and Kirschfeld, 1979). Photoequilibration was achieved within <100 ms for the blue excitation LED (generating ∼70% M, 30% R) and ∼2 s for the orange/red LED (generating ∼1-2% M, 98-99% R).
Electroretinograms (ERGs) were recorded as described previously (e.g. Satoh et al., 2010) from flies immobilised as for optical recordings with low melting point wax in truncated pipette tips. Recordings were made using a DAM60 amplifier (World Precision Instruments, Hitchin, UK) with low resistance (∼10 MΩ) glass microelectrodes filled with fly Ringer (140 mM NaCl, 5 mM KCl, 1.5 mM CaCl2, 4 mM MgCl2, 25 mM proline, 5 mM alanine) inserted into the eye, with a similar electrode inserted into the head capsule near the ocelli as reference. Stimulation was via an ultrabright red LED (640 nm) or white power LED filtered with broadband Schott filters presented at a distance of ∼5 mm from the eye via a liquid-filled light guide. Whole-cell patch clamp recordings from photoreceptors in dissociated ommatidia were made as previously described (Hardie et al., 2001; Katz et al., 2017) using 10-15 MΩ patch pipettes containing 140 mM potassium gluconate, 4 mM Mg-ATP, 1 mM NAD, 0.4 mM Na-GTP and 10 mM TES with the bath solution described above (chemicals from Sigma-Aldrich, Gillingham, UK). Illumination, via a green (520 nm) LED was calibrated in effectively absorbed photons by counting quantum bumps (Henderson et al., 2000).
For ERG and in vivo fluorescence measurements at 37°C, the tip of the plastic pipette containing the mounted fly was inserted into a coil of nichrome resistance wire connected to an Iso-Tech 303DD DC power supply (RS Components, Corby, UK) and constant current applied until the desired temperature was reached. Temperature calibration was performed using a thermistor probe of similar size (1 mm diameter) to a fly, inserted into the same truncated plastic pipette tip used for ERG or DPP imaging. After activating the heating coil with the appropriate current, the desired temperature of 37°C was reached within ∼90-120 s. Unless otherwise stated, fluorescent measurements were started 3 min after activating the coil.
Preparations of nearly pure Drosophila retinal tissue were collected as previously described (Matsumoto et al., 1982; Raghu et al., 2000). Briefly, whole flies were snap frozen in liquid nitrogen and dehydrated in pre-chilled acetone at −20°C for 4 days. The acetone was then drained off and retinae were cleanly separated at the level of the basement membrane using forceps and a flattened insect pin. Total RNA was extracted by RNeasy Plus Micro kit (Qiagen, Manchester, UK) from 20-30 retinae per sample, collected as described above and homogenized by a tissueLyser (Qiagen) with the buffer provided from RNeasy kit and 8 1 mm zirconia beads (Thistle) three times for 50 s. Samples were then transferred to a Qiashredder (Qiagen) to remove the debris, and total RNA extracted according to the manufacturer's instructions. The nucleic acid preparations were quantified by absorbance measurements at 260 nm using a NanoDrop instrument. Quantitative real time qRT-PCR was performed by One-Step SYBR PrimeScript RT-PCR Kit II (Perfect Real Time; Takara, cat. no. RR086A) and ABI 7500 fast instrument (Applied Biosystems, Warrington, UK) using primers: Ef1a48D forward: 5′-TCCTCCGAGCCACCATACAG-3′; Ef1a48D reverse: 5′-GTCTTGCCGTCAGCGTTACC-3′ (used for internal control). For each respective RNAi line, the primers used were as follows: rbo forward: 5′-ATAGATAAGTTGGCGCTGGG-3′; rbo reverse: 5′-GGGTGATCGGTCTGGTTAAG-3′; YPP1 forward: 5′-AGGAAAAGCACTCAGACACC-3′; YPP1 reverse: 5′-TTCACTCAGAGCCTGTTCAAC-3′; dPIP5K forward: 5′-AGATACCCTCCCCGCTTAA-3′; dPIP5K reverse: 5′-TGGTGAATCTTGCCACTGC-3′; sktl forward: 5′-CCTCTAGCAAACTATTCCCTCG-3′; sktl reverse: 5′-TCCAGCGGTTCATTCTCATC-3′; PI4KIIIα forward: 5′-CAGTATGCCGTAAAGACCCTC-3′; PI4KIIIα reverse: 5′-GTGTGCCACTATCTGCGAC-3′.
Statistical tests (two-tailed unpaired t-tests or one-way ANOVAs with post-tests as specified in text and/or figure legends) were performed in GraphPad Prism5. All errors are expressed as s.e.m.
We are grateful to Dr Armin Huber (Hohenheim Uni, Stuttgart, Germany), Dr Padinjat Raghu (NCBS, Bangalore, India) and the Vienna Drosophila Resource Center (VDRC) for providing flies.
Conceptualization: R.C.H.; Methodology: C.-H.L., M.K.B., R.C.H.; Validation: C.-H.L.; Formal analysis: M.K.B., S.A., R.C.H.; Investigation: C.-H.L., M.K.B., S.V.L., S.A., J.T., J.A.B., R.C.H.; Resources: J.T., J.A.B., R.C.H.; Writing - original draft: R.C.H.; Writing - review & editing: C.-H.L., M.K.B., S.A., J.A.B., R.C.H.; Supervision: R.C.H.; Project administration: R.C.H.; Funding acquisition: J.A.B., R.C.H.
This project received funding from the Biotechnology and Biological Sciences Research Council (BBSRC; Grants BB/G0092531/1 and BB/M007006/1 to C.H., M.K.B., C.-H.L.), European Union's Horizon 2020 Research and Innovation Programme (658818-FLYghtCaRe to R.C.H.), Canadian Institutes of Health Research (MOP#81187 to J.A.B.), The Cancer Research Society (#11202 and #16121 to J.A.B.) and a SickKids Restracomp scholarship (J.T.). Deposited in PMC for release after 6 months.
The authors declare no competing or financial interests.