Based on a serendipitous observation that PM targeting of polybasic polarity protein Lgl appeared to be sensitive to hypoxia (Dong et al., 2015), we previously established custom live imaging assays (see below) to demonstrate that all three polybasic polarity proteins, Lgl, aPKC, and Dlg, showed acute and reversible loss of PM targeting under 30–60 min of hypoxia (0.5% O₂) in Drosophila follicle and embryonic epithelial cells in vivo (Dong et al., 2020; Dong et al., 2015; Lu et al., 2021). Since PM PIP2 also appeared to be transiently depleted under hypoxia in such assays (Dong et al., 2015), we decided to systematically investigate how hypoxia impacts the PM PI4P and PIP2 levels in vivo. We used follicular epithelial cells of Drosophila ovaries as the primary system as they are well established for ex vivo live imaging (Prasad and Montell, 2007) and for genetic manipulations such as RNAi knock-down.

We generated transgenic flies that ubiquitously express the PI4P sensor P4M × 2::GFP (Sohn et al., 2018) as well as PIP2 sensors PLCδ-PH::GFP and PLCδ-PH::RFP (hereafter referred as PLC-PH::GFP and PLC-PH::RFP, respectively) (Wills et al., 2018). Consistent with PI4P being mostly enriched at both PM and membranes of intracellular compartments such as endosomes and Golgi, P4M × 2::GFP can be seen in both PM and intracellular puncta in follicle epithelial cells (Figure 1A). To investigate hypoxia-induced turnover of PI4P and PIP2, ovaries dissected from flies expressing both P4M × 2::GFP and PLC-PH::RFP were mounted and imaged in custom micro chambers that can be flushed with either 0.5% O₂/99.5% N₂ gas mixture for hypoxia or normal air for reoxygenation. Within ~60 minutes of hypoxia, both PI4P and PIP2 sensors were gradually lost from PM, with PM PLC-PH::RFP diminished faster than PI4P (Figure 1A, Figure 1—video 1). Once the imaging chamber was reoxygenated by flushing with normal air, both sensors rapidly recovered to the PM within ~10 min. At single-cell level, recovery of PM P4M × 2::GFP clearly and consistently preceded the PLC-PH::RFP (Figure 1A, Figure 1—video 1). Image quantification (see Materials and methods) further confirmed such differences in turnover dynamics between PM P4M × 2::GFP and PLC-PH::RFP (Figure 1A’). The faster depletion under hypoxia and delayed replenishment during reoxygenation of PM PIP2 suggest that PIP2 depletion likely involves its conversion to PI4P and its resynthesis depends on the recovery of PM PI4P.

In addition, under hypoxia the disappearance of P4M × 2::GFP intracellular puncta always preceded the depletion of PM P4M × 2::GFP. PM PI4P showed a transient increase at early phase of hypoxia likely due to depletion of intracellular PI4P leading to increased amounts of free P4M × 2::GFP sensor that binds PM PI4P (Figure 1A and A’). Under reoxygenation, PM P4M × 2::GFP consistently recovered before the appearance of intracellular P4M × 2::GFP puncta, although the latter became brighter after recovery (Figure 1A, Figure 1—video 1). Such early depletion of intracellular PI4P pool under hypoxia and its delayed replenishment under reoxygenation suggest that cells appear to prioritize the maintenance of PM PI4P pool to the intracellular pool of PI4P under energetic stress such as hypoxia.

Finally, we investigated how electrostatic PM targeting of Lgl, a polybasic polarity protein carrying a typical polybasic motif (Dong et al., 2015), correlates with hypoxia-induced turnover of PM PI4P and PIP2. Quantitative live imaging of follicle epithelial cells expressing endogenous Lgl::RFP together with PLC-PH::GFP or P4M × 2::GFP showed that under hypoxia the loss of PM Lgl::RFP preceded PIP2 and PI4P, while under reoxygenation PM recovery of Lgl::RFP lagged behind both (Figure 1B, C, Figure 1—videos 2; 3). Such results are consistent with previous studies that electrostatic PM targeting of Lgl relies on both PIP2 and PI4P, although PIP2 appears to contribute more to the PM targeting of Lgl (Dong et al., 2015). Note that in Figure 1B Lgl::RFP recovery appears lower than in wild type, possibly due to the titration of PIP2 by PLC-PH::GFP expression.

Overall, our quantitatively live imaging data showed for the first time at high subcellular and temporal resolutions that hypoxia triggered a dramatic turnover of PM PI4P and PIP2 in vivo, which directly impacts the electrostatic PM targeting under hypoxia and reoxygenation.

The seven species of PPIns including PI4P and PIP2 are synthesized and interconverted by several dozens of PPIn kinases and phosphatases, many of which are conserved in Drosophila (Balakrishnan et al., 2015). We carried out a targeted RNAi screen to identify which PPIn kinases and phosphatases may be required for regulating the hypoxia-triggered dynamic turnover of PM PI4P and PIP2, using PM Lgl::GFP as a quick readout (Supplementary file 1). By imaging mosaic follicle epithelia containing both wild type and marked RNAi-expressing cells (Figure 2A), we eliminated the variability of each individual hypoxia imaging assay, making it possible to consistently and quantitatively detect even subtle phenotypes in RNAi cells. Among the candidates we identified is PtdIns-4 kinases IIIα (PI4KIIIα), one of the PI4K enzymes that phosphorylate PI to PI4P. Among them, PI4KIIIα is primarily responsible for the biosynthesis of PI4P in the PM (Nakatsu et al., 2012; Tan et al., 2014; Yan et al., 2011), while PI4KIIα and PI4KIIIβ (encoded by four-wheel drive or fwd in Drosophila [Brill et al., 2000]) are responsible for the synthesis of PI4P in endosomes and Golgi (Baba et al., 2019; Burgess et al., 2012; Ketel et al., 2016; Tóth et al., 2006).

Under normal (i.e. normoxia) conditions, PI4KIIIα-RNAi cells showed a moderate reduction of PM PI4P and increased intracellular PI4P puncta (Figure 2A). Under hypoxia, in both RNAi and wild-type cells PI4P intracellular puncta disappeared prior to the loss PM PI4P which showed similar depletion rates in two cell types (Figure 2A, Figure 2—video 1). Under reoxygenation, compared to wild-type cells, the recovery of PM PI4P in RNAi cells was significantly delayed while intracellular puncta showed much faster recovery (Figure 2A and B).

In contrast to P4M × 2::GFP, levels of PM PLC-PH::GFP in PI4KIIIα-RNAi cells were similar to the wild-type cells, suggesting a robust PM PIP2 homeostasis mechanism that compensates well the modest reduction of PI4P under normal conditions (consistent with [Hammond and Burke, 2020; Sohn et al., 2018]). However, once challenged by hypoxia, PI4KIIIα-RNAi cells showed much accelerated loss of PM PLC-PH::GFP (Figure 2A, C, Figure 2—video 2). Strikingly, under reoxygenation PLC-PH::GFP in PI4KIIIα-RNAi cells first formed transient but prominent intracellular puncta which were not seen in wild-type cells, and these puncta rapidly disappeared at the onset of PM PIP2 recovery which was strongly delayed compared to wild-type cells (Figure 2A and C, Figure 2—video 2).

In summary, our data support that PI4KIIIα is required for the efficient replenishment of PM PI4P and PIP2 after their hypoxia-triggered depletion. During reoxygenation, PI4KIIIα knock-down cells showed delayed PM PI4P recovery but enhanced replenishment of intracellular PI4P pool, although the latter could be due to increased amount of free P4M × 2::GFP sensors when PM P4P recovery was delayed. In addition, knocking down PI4KIIIα accelerates the depletion of PM PIP2 but not PI4P under hypoxia, suggesting an increased conversion of PIP2 to PI4P during depletion.

We then investigated how other two PI4K enzymes, PI4KIIα and FWD, contribute to the dynamic turnover of PI4P of PM and intracellular pools. Since neither PI4KIIα nor fwd null mutants are lethal (Brill et al., 2000; Burgess et al., 2012; Polevoy et al., 2009) and single RNAi knock-down against each showed no obvious phenotypes, we used newly published multi-RNAi tools Qiao et al., 2018 to generate fly stocks simultaneously expressing multiple dsRNAs targeting either both PI4KIIα and fwd (‘PI4K-2KD’), or all three PI4Ks (‘PI4K-3KD’). While PI4K-2KD cells showed no discernable phenotypes under either normoxia or hypoxia in our hypoxia assays (Figure 3—figure supplement 1), PI4K-3KD cells showed severely reduced PM P4M × 2::GFP and dramatically increased intracellular P4M × 2::GFP puncta (Figure 3A, B); the latter could be due to more P4M × 2::GFP sensors being bound to the intracellular PI4P pool when PM PI4P is low (Sohn et al., 2018; Wills et al., 2018).

Similar to PI4KIIIα-RNAi cells, PI4K-3KD cells under hypoxia first showed reduction of intracellular P4M × 2::GFP puncta and transient increase of PM P4M × 2::GFP, followed by accelerated depletion of PM P4M × 2::GFP (Figure 3A, Figure 3—video 1). Under reoxygenation, PI4K-3KD cells only showed recovery of P4M × 2::GFP in intracellular puncta but not in PM, suggesting that PI4K-3KD cells are severely deficient in acute resynthesis of PM PI4P after its hypoxia-induced depletion.

Remarkably, PM PLC-PH::GFP levels showed no reduction in PI4K-3KD cells under normal conditions, despite of the severe loss of PM PI4P (Figure 3A, B, Figure 3—video 2). Similar to PI4KIIIα-RNAi cells, in PI4K-3KD cells PM PLC-PH::GFP showed accelerated loss under hypoxia (Figure 3A), and formed transient intracellular puncta during reoxygenation. Despite the apparent absence of PM PI4P recovery in PI4K-3KD cells, PM PLC-PH::GFP still recovered under reoxygenation, although the recovery was much delayed (Figure 3A, Figure 3—video 2).

Our data support that PI4KIIα and/or FWD contribute significantly to the maintenance of PM PI4P under normoxic conditions and to the replenishment of PM PI4P after hypoxia-triggered depletion. The data also suggest that an apparently PM PI4P-independent mechanism maintains the homeostatic level of PM PIP2 under normal conditions and sustains its recovery after hypoxia-triggered depletion. However, rapid replenishment of PM PI4P is clearly required for the efficient recovery of PM PIP2 after hypoxia-triggered depletion.

To investigate how electrostatic PM targeting is affected by the disruptions of PM PI4P and PIP2 turnover, we imaged the PM targeting of Lgl::GFP in PI4KIIIα-RNAi and PI4K-3KD cells undergoing hypoxia. In PI4KIIIα-RNAi cells, we found that Lgl::GFP essentially phenocopied the behavior of PLC-PH::GFP. As shown in Figure 4A, under normoxic conditions PM Lgl::GFP levels in PI4KIIIα-RNAi were similar to wild-type cells, but Lgl::GFP showed accelerated loss from PM under hypoxia and severely delayed recovery to PM during reoxygenation. In addition. Lgl::GFP also formed transient intracellular puncta prior to the onset of its PM recovery (Figure 4A, Figure 4—video 1). Such data are consistent with previous studies that electrostatic PM targeting of Lgl::GFP is more PIP2-dependent (Dong et al., 2015).

In PI4K-3KD cells, PM Lgl::GFP also showed accelerated loss under hypoxia and delayed recovery under reoxygenation (Figure 4A, Figure 4—video 2). Unlike PLC-PH::GFP, Lgl::GFP only formed few very transient puncta prior to the onset of PM recovery under reoxygenation (Figure 4A). However, in half (113/219) of PI4K-3KD cells PM Lgl::GFP was already partially diffused under normal conditions (asterisked in Figure 4A), and in these cells Lgl::GFP also failed to recover to PM during reoxygenation. Even in PI4K-3KD cells with normal PM Lgl::GFP, a third (36/106) failed to recover during reoxygenation. Such data suggest that the electrostatic PM targeting becomes much less resilient in PI4K-3KD cells, especially when cells are challenged by energetic stresses such as hypoxia.

Unlike PI4KIIα and FWD, which localize to intracellular membranes, PI4KIIIα is primarily PM localized (Baskin et al., 2016; Nakatsu et al., 2012). We thus investigated how subcellular localization to PI4KIIIα affects the hypoxia-induced dynamic turnover of PM PI4P. The PM localization of yeast PI4KIIIα (‘Stt4p’) requires EFR3, YPP1and Sfk1 (mammalian TTC7 and TMEM150A, respectively) (Chung et al., 2015; Hammond et al., 2014; Nakatsu et al., 2012). All three proteins are conserved in Drosophila, including EFR3 homologue Rbo (‘Rolling blackout’) (Huang et al., 2004; Vijayakrishnan et al., 2009), dYPP1/dTTC7 (CG8325) and dTMEM150A (“dTMEM”, CG7990 and CG4025), (Liu et al., 2018). Consistent with that EFR3 is a peripheral membrane protein and that its PM targeting appears to be independent of PI4P/PIP2 (Nakatsu et al., 2012), in follicle epithelial cells Rbo::GFP (Huang et al., 2004; Vijayakrishnan et al., 2009) showed exclusive PM localization that was highly resistant to hypoxia (Figure 5—figure supplement 1, Figure 5—video 1). rbo-RNAi cells essentially phenocopied PI4KIIIα-RNAi cells in terms of the dynamic turnover of PM PI4P, PIP2 and Lgl::GFP under hypoxia, with one exception that under reoxygenation Lgl::GFP did not form transient puncta prior to PM recovery, even though PLC-PH::GFP still formed transient and prominent puncta in rbo-RNAi cells prior to the onset of the recovery of PM PLC-PH::GFP (Figure 5A, C and D, Figure 5—videos 2–4). The reason for such difference between Lgl::GFP and PLC-PH::GFP in rbo-RNAi cells is unclear.

YPP1/TTC7 helps to link PI4KIIIα to Rbo/EFR3 and enhances the PM targeting of PI4KIIIα in cultured cells (Nakatsu et al., 2012). Consistently, ttc7-RNAi cells showed similar albeit milder phenotypes in hypoxia-triggered turnover of PM PI4P and PIP2 as well as the PM targeting and retargeting of Lgl (Figure 6A, Figure 6—videos 1–3). Interestingly, ttc7-RNAi cells also showed reduced PM Rbo::GFP (Figure 6B), suggesting that TTC7 enhances the PM targeting of Rbo and that phenotypes in ttc7-RNAi cells could be partially due to the reduction of PM Rbo. Although at present we do not have the tools to directly examine the PI4KIIIα localization in rbo- or ttc7-RNAi cells, our data strongly support a scenario where PM localization of PI4KIIIα is essential for the efficient recovery of PM PI4P and PIP2 after hypoxia-triggered depletion.

How does hypoxia trigger the acute depletion of PM PI4P and PIP2? We showed previously that direct ATP inhibition by antimycin (AM) in follicle cells also induced loss of PM Lgl::GFP in follicle and embryonic epithelial cells (Dong et al., 2015), suggesting that the acute depletion of PM PI4P and PIP2 could be triggered by hypoxia-induced ATP inhibition. We first tested this hypothesis using a genetically encoded ATP sensor AT[NL], which is a FRET-based ratiometric ATP sensor that recently became available and validated in Drosophila (Imamura et al., 2009; Kioka et al., 2014; Tsuyama et al., 2017). As shown in Figure 7—figure supplement 1A, live imaging of follicle cells expressing AT[NL] confirmed an acute and reversible reduction of intracellular ATP levels under hypoxia and reoxygenation. Because AT[NL] is not suitable for imaging together with our current PI4P and PIP2 sensors, we also tested an intensimetric ATP sensor MaLionR (Arai et al., 2018) in HEK293 cells. Under hypoxia, HEK293 cells showed acute and reversible reduction of ATP levels as measured by MaLionR intensity changes (Figure 7—figure supplement 1B, Figure 7—video 1), as well as reversible depletions of PM P4M × 2::GFP and PLC-PH::GFP which correlate with MaLionR intensity changes (Figure 7—figure supplement 1C).

We then subjected HEK293 cells to acute ATP inhibition by the treatment of 2-deoxyglucose (2-DG) and AM, followed by washout to allow ATP recovery. Upon adding 2-DG and AM, MaLionR brightness dropped rapidly within 10–20 min and plateaued afterward (Figure 7A, B, Figure 7—videos 2; 3). In general, P4M × 2::GFP or PLC-PH::GFP began gradually lost from PM at the onset of the ATP drop and became completely cytosolic within ~40–60 min of ATP inhibition (Figure 7A, B). While the reduction of ATP as measured by MaLionR brightness was rather uniform across cells, ATP recovery after the washout of 2-DG and AM was slightly asynchronous across the cells. In general, the recovery of PM PI4P always preceded the detectable increase of MaLionR brightness (Figure 7A, Figure 7—video 2), while PM PIP2 recovery was concurrent with MaLionR brightness increase (Figure 7B, Figure 7—video 3). In some cells PM PI4P recovery appeared up to ten minutes ahead of MaLionR increase (Figure 7A). Given that MaLionR appears to have a dynamic range between ~50 µM and 2 mM of ATP (Arai et al., 2018), such data support that initial replenishment of PM PI4P could start under very low intracellular ATP levels.

Similar to the results from hypoxia assays in Drosophila follicle cells (Figure 1A), in HEK293 cells under ATP inhibition the depletion of PM P4M × 2::GFP consistently lagged behind the loss of PM PLC-PH::GFP, while after drug washout PM P4M × 2::GFP recovery consistently preceded the recovery of PM PLC-PH::GFP (Figure 7C, Figure 7—video 4). We also investigated whether PI4KIIIα is required for the post-ATP inhibition recovery of PM PI4P in HEK293 cells. After ATP inhibition, we washed out drugs with medium containing 20 µM wortmannin (WM) which specifically inhibits PI4KIIIα/β isoforms but not PI4KIIα/β (Balla and Balla, 2006). Washout with WM caused no discernable delay in ATP recovery as measured by MaLionR brightness (Figure 7D), but strongly delayed PM recovery of PI4P, and to much less degree PIP2 (Figure 7D, Figure 7—figure supplement 2). Such results are consistent with the Drosophila RNAi results that PI4KIIIα is required for the efficient PM recovery of PI4P and PIP2 after hypoxia-triggered depletion.

In summary, ATP inhibition in human cultured cells also triggers acute and reversible depletion of PM PI4P and PIP2 similar to the dynamic turnover of PI4P and PIP2 in cells undergoing hypoxia. For reasons unknown, our efforts of making a transgenic MaLionR sensor in Drosophila was unsuccessful, limiting our ATP assays to HEK293 at present. However, our data strongly support that intracellular ATP levels directly dictate the homeostasis and turnover of PM PI4P and PIP2 in both Drosophila and human cultured cells.

We would speculate that the reduction of intracellular ATP levels, through either hypoxia or drug inhibition, triggers acute loss of PM PI4P and PIP2 by two possible mechanisms. PI4P and PIP2 could be maintained at slow turnover rates on the PM, but reduction of ATP activates a specific cellular response to acutely deplete PM PI4P and PIP2. Alternatively, a more parsimonious mechanism would be that PM PI4P and PIP2 are constantly under fast turnover, which requires high activity of PI and PIP kinases. ATP reduction, which directly inhibits the activity of these kinases, pivots the equilibrium to the dephosphorylation process which converts the PIP2 to PI4P and PI4P to PI.

Consistent with the critical role of PI4P in maintaining PM identity and its biological activity, our data revealed that cells undergoing hypoxia/ATP inhibition consistently prioritize the maintenance and recovery of PM PI4P over the intracellular PI4P pool in a PI4KIIIα-dependent manner. However, while PI4KIIIα is well characterized for its essential role in generating the PI4P on the PM (Balla, 2013; Nakatsu et al., 2012), K_mATP values of PI4KIIIα (500–700 µM) and PI4KIIβ(Fwd) (~400 µM) are about one or two orders higher than that of PI4KIIα (10–50 µM) (Balla and Balla, 2006; Carpenter and Cantley, 1990). Such K_mATP differences would suggest that, in contrast to our results, the intracellular PI4P pool should deplete more slowly and recover more quickly than the PM PI4P in cells undergoing hypoxia/ATP inhibition, as PI4KIIIα would be the first PI4K to lose activity under hypoxia and the last to become active under reoxygenation.

One possible reason behind such a discrepancy could be that K_mATP of PI4KIIIα was measured decades ago using purified PI4KIIIα enzymes from tissues such as bovine brains and uterus (Carpenter and Cantley, 1990). Recent data showed that PI4KIIIα forms a highly ordered multi-protein membrane targeting complex essential for its activity (Lees et al., 2017). It is thus possible that PI4KIIIα in the complex may have much lower K_mATP in vivo, and/or has dramatically increased enzymatic activity to produce sufficient PI4P at the PM even when ATP levels are much lower than the measured K_m. Alternatively, the K_mATP of PI4KIIIα is indeed high and our live imaging results actually highlight a prioritized transfer of PI4P from the intracellular pool to maintain or replenish the PM PI4P levels. Phosphatidylinositol (PI) is abundant on intracellular membranes, but not the PM (Pemberton et al., 2020; Zewe et al., 2020). Therefore, during the early phase of reoxygenation when intracellular ATP levels are low, PI4P is first synthesized at the intracellular pool by PI4KIIα but is immediately transferred to the PM. Only after the full replenishment of PM PI4P is the intracellular PI4P pool filled. Supporting this transfer PI4P from intracellular pools to PM pools (Dickson et al., 2014), our data show loss of PM PI4P recovery in PI4K-3KD cells, in which the maintenance of intracellular pool of PI4P is supposedly impaired.

Out data are consistent with the view that in wild type cells under hypoxia/ATP inhibition, PM PIP2 depletes and recovers through direct inter-conversion with PI4P on the PM. Interestingly, in both PI4KIIIα-RNAi and PI4K-3KD cells, PM recovery of PIP2 is preceded with transient intracellular PIP2-positive puncta which were not seen in recovering wild-type cells. It is possible that in the absence or delay of PM PI4P recovery, enzymes such as PIP5K are instead electrostatically recruited to the intracellular PI4P-positive puncta (Dong et al., 2016; Fairn et al., 2009) to convert PI4P to PIP2. It is unclear, however, in PI4K knock down cells whether the delayed PM PIP2 recovery originates from the PIP2 generated in these puncta. Additional sensors are necessary to confirm the co-localization of PI4P and PIP2 on these transient puncta. Notably, MEF cells from PI4KIIIα knock-out mice also showed increased PIP2-positive intracellular vesicles (Dong et al., 2016; Nakatsu et al., 2012).

The existence of intracellular P4M × 2::GFP puncta in PI4K-3KD cells suggest that the knock down of PI4KIIα and fwd is unlikely complete, but the severe reduction of PM PI4P confirms the knock down is strong enough to greatly enhance the defects in PI4KIIIα-RNAi cells. Such partial knock-down by PI4K-3KD is actually necessary for our imaging assays, as completely blocking PI4P synthesis is cell lethal. It is striking that PM PIP2 is well maintained in the near absence of PM PI4P in PI4K-3KD cells. Synthetic biology-based evidence suggested that PIP5K can be sufficient to make PIP2 from PI in E. coli by phosphorylating both its fourth and fifth positions in the absence of PI4Ks (Botero et al., 2019), and it is possible that similar pathway maintains the steady state PM PIP2 levels in PI4K-3KD cells. Nonetheless, our imaging experiments showed that PI4K activity is essential for cells to maintain PM PIP2 levels when cells are subject to hypoxia. In this regard, our study of PI4K-compromised cells repeatedly revealed deficiencies in PI4P/PIP2 turnover and electrostatic PM targeting that can only be observed when cells are subject to energetic stress conditions.

PM targeting of PI4KIIIα strictly depends on its formation of an obligate superassembly with TTC7 (YPP1), FAM126 (Hycin) and EFR3 (Rbo) (Lees et al., 2017; Wu et al., 2014). A recent study also showed that RNAi knock-downs of PI4KIIIα, TTC7 and Rbo yielded similar phenotypes in Drosophila wing discs, such as moderately reduced PM PI4P but no obvious changes of PM PIP2 (Basu et al., 2020). Same RNAi knock-downs in Drosophila photoreceptors also showed similar phenotypes such as reduced PI4P levels and impaired light response, although PIP2 levels also appear to be reduced (Balakrishnan et al., 2018). Moreover, our data showed that knocking down TTC7 also reduced PM localization of Rbo, supporting that components in PI4KIIIα complex may act interdependently for proper PM targeting in vivo.

The hypoxia-resistant PM localization of Rbo/dEFR3 suggests that under hypoxia/ATP inhibition PI4KIIIα maintains its PM localization, which should be essential for its role in recovering the PM PI4P. The core complex of PI4KIIIα/TTC7/FAM126 forms a collective basic surface that electrostatically binds to the acidic inner leaflet of the PM which could be sensitive to the loss of PM PI4P and PIP2. However, TTC7 also interacts with the C-terminus of EFR3 (Chung et al., 2015; Lees et al., 2017). PM targeting of yeast EFR3 requires a basic patch that interacts with general acidic phospholipids but is not disrupted by the loss of PM PI4P and PIP2 (Wu et al., 2014). Mammalian EFR3 contains an additional N-terminal Cys-rich palmitoylation site that is also required for the PM targeting (Nakatsu et al., 2012). Such dual and PI4P/PIP2-independent mechanisms are supported by the hypoxia-resistant PM localization of Rbo as we observed. Future studies will be needed to directly investigate the PM targeting of PI4KIIIα, TTC7 and FAM126 in vivo under hypoxia/ATP inhibition.

While previous studies showed that genetically reducing PM PIP2 levels disrupts the PM localization of several polarity proteins including Lgl, it is difficult to conclude whether such loss of PM targeting is the direct consequence PIP2 reduction (Claret et al., 2014; Gervais et al., 2008). In this study, we are able to quantitatively and qualitatively demonstrate that in cells undergoing hypoxia-reoxygenation the acute and reversible loss of PM targeting of Lgl directly correlates with the turnover of PM PI4P and PIP2. Consistent with the idea that Lgl appears to depend more on PIP2 for its PM targeting, Lgl closely follows the dynamic turnover and relocation of PIP2 during hypoxia and reoxygenation. In particular, ectopic and transient puncta of Lgl::GFP seen in PI4KIIIα-RNAi or PI4K-3KD cells under reoxygenation appear to be strikingly similar to PIP2-positive puncta in these cells, although due to the limited array of biosensors we have not been able to directly confirm the co-localization of Lgl::GFP and PIP2 in these transient puncta. Additional genetically encoded biosensors for PI4P and PIP2 (e.g. P4M × 2::iRFP and PLC-PH::iRFP) are in development for such experiments.

It is notable that in rbo-RNAi cells, Lgl::GFP formed very few transient puncta prior to PM recovery during reoxygenation, even though PLC-PH::GFP showed plenty of prominent puncta. The reason for such difference between Lgl::GFP and PLC-PH::GFP in rbo-RNAi cells is unclear, though likely derives from the requirement of polybasic motif proteins for additional anionic lipids at the plasma membrane, specifically high molar fractions of phosphatidylserine in addition to lower concentrations of polyanionic phosphoinositides (Yeung et al., 2008).

In summary, our study revealed an acute and reversible loss of PI4P and PIP2 from PM under hypoxia/ATP inhibition in both Drosophila and human cultured cells. Such dynamic turnover of PM PI4P and PIP2 explains the dramatic loss of the PM targeting of polybasic polarity proteins such as Lgl under hypoxia. How cells meticulously maintain steady state PM PI4P and PIP2 levels under normal physiological conditions has been extensively studied; our studies highlight the importance of understanding mechanisms controlling this homeostasis and dynamics of phosphoinositides under energetic stresses triggered by hypoxia, ATP inhibition and ischemia, and the critical role of the interplay between polarity proteins and PM phosphoinositides in controlling cell polarity under normal and disease conditions.

Flies of carrying transgenic ubi-P4M::GFP, ubi-P4M × 2::GFP, ubi-PLC-PH::GFP and ubi-PLC-PH::RFP alleles were generated by phiC31-mediated integration protocol (Huang et al., 2009). attP^VK00022 (BL#24868) and attP^VK00040 (BL#35568) stocks were used to integrate the above constructs to the 2nd chromosome and 3rd chromosome, respectively.

PI4K-3KD was generated using pNP plasmid based on the published protocol Qiao et al., 2018. The pNP vector was cut with Nhe I and EcoR I and ligated with annealed oligo-dimmer o to generate pNP-fwd-KD, pNP-PI4KII-KD, and pNP-PI4KIIIα-KD. Primers used for annealing oligo-dimmer were listed in Supplementary file 2:

The pNP-fwd-KD was digested with Spe I and ligated with the shortest fragment cut by Spe I and Xba I from pNP-PI4KIIα-KD to make pNP-PI4K-2KD. To make pNP-PI4K-3KD, pNP-PI4KIIIα-KD was digested with Spe I and ligated with the shortest fragment cut by Spe I and Xba I from pNP-PI4K-2KD. pNP-PI4K-2KD and pNP-PI4K-3KD were used to generate transgenic stocks with the standard protocol.

UAS-AT1.03NL1(DGRC#117011) was used to express the Drosophila-optimized ATeam ATP sensor ‘AT[NL]’ in follicle cells (Tsuyama et al., 2017). cy2-Gal4 (Queenan et al., 1997) was a gift from David Bilder, UC Berkeley. rbo::GFP was a gift from Kendal Broadie, Vanderbilt University (Huang et al., 2004).

w; lgl::mCherry and w; lgl::GFP knock-in stock were previously published (Dong et al., 2015).

Drosophila cultures and genetic crosses are carried out at 25 °C.

Additional stocks used were: UAS-PI4KIIIα-RNAi (BL#35256), UAS-Rbo-RNAi (VDRC#47751), UAS-ttc7-RNAi (VDRC#35881).

Mammalian constructs of P4M::GFP, P4M × 2::GFP, and PLC-PH::GFP were as previously described (Hammond et al., 2014; Hammond et al., 2012; Várnai and Balla, 1998). DNA fragments encoding PLC-PH::GFP, P4M::GFP and P4M × 2::GFP were inserted into pGU vector (Lu et al., 2021) which contains a ubiquitin promoter. MaLionR ATP sensor was obtained from Addgene (#113908) (Arai et al., 2018).

Follicle cells containing over-expressing or RNAi clones were generated by heat-shocking 3 days old (after eclosion) young females of the correct genotype at 37 °C for 15–30 min and ovaries were dissected 3 days later.

Ovaries from adult females of 2 days old were dissected in halocarbon oil (#95) and were imaged according to previously published protocol (Dong et al., 2015; Huang et al., 2011). To ensure sufficient air exchange to samples during the imaging session, dissected ovaries were mounted in halocarbon oil on an air-permeable membrane (YSI Membrane Model #5793, YSI Inc, Yellow Springs, OH) sealed by vacuum grease on a custom-made plastic slide over a 10 × 10 mm² cut-through window. The slide was then mounted in a custom made air-tight micro chamber (~5 cm³) for live imaging under confocal microscope. Oxygen levels inside the chamber were controlled by flow of either air or custom O₂/N₂ gas mixture at the rate of approximately 1–5 cc/s. Images were captured at room temperature (25 °C) on an Olympus FV1000 confocal microscope (60 x Uplan FL N oil objective, NA = 1.3) by Olympus FV10-ASW software, or on a Nikon A1 confocal microscope (Plan Fluo 60 x oil objective, NA = 1.3) by NIS-Elements AR software.

HEK293 cells were cultured in glass bottom dishes (In Vitro Scientific) and were transfected with DNA using X-treme Gene 9 DNA transfection reagent (Sigma Cat# 6365787001). After 24–40 hr of transfection cells were mounted and imaged on a Nikon A1 confocal microscope (Plan Fluo 40 x oil objective, NA = 1.3) by NIS-Elements AR software. For images to be used for quantification, parameters were carefully adjusted to ensure no or minimum overexposure. In addition, when necessary, a fluorescent PM dye (CellMask DeepRed Plasma Membrane Stain, ThermoFisher, Cat#C10046) was added to the cell culture prior to live imaging to help in visualizing the PM for later quantifications.

HEK293 cells one day after transfection were imaged live in temperature control chamber at 37 °C. For hypoxia and ATP inhibition experiments, cells were starved in glucose-free medium six hours prior to live imaging. Hypoxia treatment was carried out using a custom designed culture dish lid which seals the 35-mm glass-bottom culture dish but allows gas to be flushed in and out the sealed dish chamber through two small built-in nozzles. Prior to the sealing, medium inside dish was reduced to ~200 µl and was covered by an air permeable membrane to prevent evaporation. Oxygen levels inside the chamber were controlled by flow of either air or custom O₂/N₂ gas mixture at the rate of approximately 0.1 cc/s. ATP inhibition was initiated by adding equal volume of serum containing 2-DG and antimycin to the final concentrations of 10 mM and 2 µM, respectively. After the end of inhibition, drugs were washout by replacing with normal media. For wortmannin inhibition experiment, wortmannin was added to the washout media to final concentration of 10 µM.

Time-lapse movies were first stabilized by HyperStackReg plug-in in ImageJ. Images or movies containing excessively noisy channels were denoised by PureDenoise plugin in ImageJ prior to quantification. PM localization of GFP or RFP in images or movies was measured in Image J by custom macro scripts. For Drosophila samples, ROIs approximately 20–40 µm² were drawn across selected cell junctions in the first frame of the movie. Custom macros were used to automatically generate PM masks by threshold-segmentation using the mean pixel value of the ROI.

For HEK293 cells, PM masks were generated by an à trous waveleta decomposition method (Hammond et al., 2014; Olivo-Marin, 2002) base on the channel that either contains PM-localized proteins or fluorescent PM dyes. Cytosol masks were generated by segmentation using threshold based on the mean pixel value of the ROI. Cells expressing all transfected fluorescent proteins were selected for measurement. For each cell, one ROI was drawn to cover part of cell that contains PM and cytoplasm.

Due to the use of computer generated PM and cytosol masks, the exact shape of the ROI was not critical, except that PM segments in contact with neighboring expressing cells were avoided. Nuclei and intracellular puncta were also avoided. Custom macros were used to automatically measure PM and cytosolic intensities of each fluorescent protein in each cell marked by ROI in the sample image. Background was auto-detected by the macro based on the minimal pixel value of the whole image.

The PM localization index for each fluorescent protein was auto-calculated by the macro as the ratio of [PM - background]/[cytosol - background]. In live imaging experiments, “Normalized PM Index” was calculated by normalizing (PM Index –1) over the period of recording against the (PM Index –1) at 0 min. Data were further processed in Excel, visualized and analyzed in Graphpad Prism.