Spatiotemporal control of phosphatidic acid signaling with optogenetic, engineered phospholipase Ds

Phosphatidic acid (PA) is a pleiotropic lipid second messenger with several physiological and pathological functions (Liu et al., 2013; Wang et al., 2006). PA can modify membrane charge and curvature and also engage and activate cytosolic effector proteins (Kooijman and Burger, 2009; Jang et al., 2012; Putta et al., 2016). These effects can lead to cellular changes, including in cytoskeletal organization, membrane trafficking, gene expression, growth, and migration. As such, dysregulation of PA homeostasis occurs in many diseases, including cancer, neurodegeneration, and infection (Gomez-Cambronero, 2014; Nelson and Frohman, 2015; Bruntz et al., 2014; Oliveira and Di Paolo, 2010).

PA is produced by three pathways: acylation of lysophosphatidic acid (LPA) by lysophosphatidic acid acyltransferases (LPAATs), phosphorylation of DAG by DAG kinases (DGKs), and hydrolysis of phosphatidylcholine (PC) by phospholipase Ds (PLDs; Bradley and Duncan, 2018; Selvy et al., 2011; Shulga et al., 2011). Pools of PA produced via these different routes are suggested to have distinct cellular functions. For example, PA produced by LPAAT in the ER functions as an intermediate in de novo phospholipid and triglyceride biosynthesis (Vance and Vance, 2004). PA produced by DGKs and PLDs on other organelle membranes can affect diverse and distinct processes, including actin polymerization, macropinocytosis, secretory vesicle formation, mTOR signaling, and, recently, the Hippo pathway (Selvy et al., 2011; Eichmann and Lass, 2015; Nelson and Frohman, 2015; Totaro and Piccolo, 2019; Fang et al., 2001; Foster, 2013; Yoon et al., 2015). In particular, Hippo signaling, which controls cell size and proliferation, was recently shown to be downregulated by PLD-derived PA (Han et al., 2018a).

Cells employ multiple routes to produce PA for several reasons. LPAATs, DGKs, and PLDs have different subcellular localizations, enabling production of local pools of PA on different organelle membranes (Bradley and Duncan, 2018; Eichmann and Lass, 2015; Selvy et al., 2011; Du et al., 2003, 2004; Shulga et al., 2011). Because of different substrate acyl tail compositions (LPA, DAG, and PC, respectively) and intrinsic acyl tail preferences, the enzymes produce different collections of PA species, some of which can differentially impact signaling pathways. Further, the availability of many isozymes from three different classes gives cells ample opportunities to exert control over PA production from diverse upstream stimuli. Given the central position that PA occupies in phospholipid metabolism (Vance and Vance, 2004; Vance, 2015), redundancy and many levels of regulation are a key feature of PA metabolism; however, these properties make it challenging to decipher specific biological functions of spatially segregated pools of PA.

PA levels can be manipulated using loss- and gain-of function approaches. Typical loss-of-function approaches involve LPAAT, DGK, or PLD inhibition, RNAi, or gene knockouts. The ability of PA-biosynthesizing enzymes to partially compensate for one another can, however, make it challenging to ascribe specific biological functions to subcellular pools of PA. For example, knockout or inhibition of PLD1 and PLD2, the two PLD isoforms responsible for PA production by hydrolysis of PC, results in modest to minimal changes, which can be cell-type and stimulus dependent, in the total PA levels (Su et al., 2009; Antonescu et al., 2010; Bohdanowicz et al., 2013; see Fig. S1, A and B). Loss-of-function studies can also be complicated by noncatalytic roles of some enzymes (Selvy et al., 2011).

Complementary gain-of-function methods to manipulate PA levels present different challenges. One approach, adding PA-containing liposomes to the media, affords control of acyl chain composition, but not subcellular location. By contrast, overexpression of endogenous PA-producing enzymes gives some spatial control over PA localization and acyl tail composition. However, this approach offers poor temporal resolution, can produce ectopic pools of PA due to nonphysiological enzyme localization upon overexpression (Du et al., 2003), and can be complicated by noncatalytic functions of the enzymes. Finally, some PA-synthesizing enzymes have intrinsically low catalytic activity, requiring activation by additional endogenous factors (Powner and Wakelam, 2002; Du et al., 2000).

Critically, existing tools are not suitable for determining whether acute, local production of PA on specific organelle membranes is sufficient for driving individual physiological processes. This technological gap has hampered efforts to unravel the pleiotropy of PA, which would establish biological functions for spatiotemporally distinct pools of this lipid. Here, we develop a toolset for producing physiologically relevant pools of PA with spatial and temporal precision and apply it to define roles for organelle-specific PA in directing a critical signal transduction pathway. We first report the design, optimization, and validation of a suite of light-controlled, optogenetic PLDs (optoPLDs) capable of producing varying amounts of PA on specific organelle membranes. Using these optoPLDs, we then establish the sufficiency for a pool of PA at the plasma membrane (PM), but not on several intracellular organelles, in attenuating mammalian Hippo signaling. Because mammalian PLDs exhibit different localizations, durations, and strengths of activity when stimulated by several upstream factors, the tunable activity of the optoPLD system, in space, time, and amplitude, enables effective mimicry of the different activation states of endogenous, PLD-dependent PA signaling.

OptoPLD comprises two fusion proteins encoded on a single plasmid and connected by a P2A self-cleaving peptide: (1) CRY2 fused to an mCherry-tagged PLD and (2) CIBN fused to a variable organelle targeting tag (Fig. 1 A). Among several light-mediated heterodimerization systems, CRY2–CIBN was chosen because its dimerization is rapidly induced by nontoxic, 488-nm light, but its dissociation rate in the dark is relatively slow, making it ideal for sustained heterodimerization (and concomitant PA production) using intermittent and thus relatively minimal sample illumination (Kennedy et al., 2010; Idevall-Hagren et al., 2012; Benedetti et al., 2018; Niu et al., 2016).

For the PA-producing enzyme, we selected a PLD from Streptomyces sp. PMF (PLD_PMF) for several reasons. We chose a PLD because its substrate, PC, is much more abundant than DGK and LPAAT substrates (DAG and LPA); consequently, activation of PA production would minimally impact other lipid levels. We selected a heterologous PLD to avoid effects from protein–protein interactions with mammalian proteins and because mammalian PLDs are subject to complex regulation, requiring activation by various factors. Among the many microbial PLDs, PLD_PMF has only ∼20% sequence identity to human PLDs. This well-characterized enzyme exhibits high activity under physiological conditions (37°C, pH 7.4), and the availability of its crystal structure could enable protein-engineering efforts (Leiros et al., 2004).

To target CRY2-mCherry-PLD_PMF to desired organelles, we fused CIBN to localization tags for the PM (CAAX motif), endosomes (2xFYVE domain of Hrs1), the TGN (sialyltransferase transmembrane domain), or the ER (Sac1 transmembrane domain). By encoding CRY2-mCherry-PLD_PMF and CIBN tag on a bicistronic vector (Liu et al., 2017), we ensured their equimolar expression and eliminated the need to fluorescently tag CIBN (Fig. 1 A).

In the absence of blue light, CRY2-mCherry-PLD_PMF was localized in the cytosol, but upon illumination with 488-nm light, it was recruited to the appropriate, CIBN-tagged membrane (Fig. 1 B). Colocalization analysis revealed that 25–45% of the total optoPLD relocalized from the cytosol to the desired membrane (Fig. S2 A). To quantify the turn-on/off kinetics of optoPLD photodimerization, we generated an infrared fluorescent protein (iRFP)–tagged CIBN-PM and examined colocalization with CRY2-mCherry-PLD_PMF. The light-mediated PLD_PMF recruitment to the PM was very rapid (t_1/2 ∼5 s), and its dissociation in the dark was much slower (t_1/2 ∼10–15 min; Fig. 1 C). Similar kinetics was observed for the association and dissociation of other organelle-targeted optoPLDs (Fig. S2 B). Thus, for sustained PLD_PMF recruitment to target membranes, blue light pulses every 2–5 min was sufficient.

We next ascertained whether light-induced optoPLD recruitment to target membranes would result in PLD-catalyzed PA synthesis. Here, we used a genetically encoded biosensor derived from the Spo20 PA-binding domain, termed PASS (Lu et al., 2016; Zhang et al., 2014). We coexpressed GFP-PASS with each different organelle-targeted optoPLD in HEK 293T cells and monitored both PASS and optoPLD fluorescence by confocal microscopy after 30 min of intermittent blue light illumination (5-s pulses every 2 min). On each organelle membrane tested, we observed the colocalization of GFP-PASS with optoPLD (Fig. 2). Under identical conditions, expression of catalytic-dead optoPLD bearing the H170A mutation in PLD_PMF (termed optodPLD) resulted in a mostly cytosolic localization of GFP-PASS, confirming that recruitment of PASS to the target membrane requires PLD activity (Fig. 2). We note that Spo20 has affinity for other anionic PM lipids, including phosphatidylinositol (4,5)-bisphosphate, which might explain a minor partial and optoPLD-independent localization of this probe at the PM (Horchani et al., 2014). A time-course experiment revealed that PASS accumulates on the target organelle membrane as little as 10 min after optoPLD activation and returns to the cytosol 1–3 h after optoPLD is inactivated by keeping the cells in the dark (Fig. S3). These results demonstrate the ability of optoPLD to produce spatiotemporally controlled pools of PA.

We determined that the PA generated by optoPLD comprised molecular species similar to PA generated by endogenous, mammalian PLDs. This issue is important because PA species with different acyl tail compositions can differentially affect signaling (Adachi et al., 2016; Yoon et al., 2015; Tanguy et al., 2018), and PLD_PMF and mammalian PLDs have very different sequences, though they both share the characteristic HKD active site motifs (Selvy et al., 2011).

To characterize PA made by PLDs as opposed to total cellular PA, we took advantage of IMPACT (Imaging PLD Activity with Clickable alcohols via Transphosphatidylation), our recently reported chemoenzymatic method that generates tagged lipids that report on levels of PLD activity (Bumpus and Baskin, 2016, 2017; Bumpus et al., 2018). IMPACT relies on the ability of PLDs to catalyze transphosphatidylation reactions of PC with short primary alcohols to produce phosphatidyl alcohol lipid reporters of PLD activity (Brown et al., 2007; Morris et al., 1997).

In an IMPACT experiment, a functionalized primary alcohol such as 3-azido-1-propanol (AzProp) is used to produce an azido phosphatidyl alcohol (Fig. 3 A). A subsequent click chemistry tagging step with an alkyne-quaternary ammonium tag (Alk-QA) enables highly sensitive identification and quantification of the phosphatidyl alcohol species (varying by acyl tail composition) by using liquid chromatography–mass spectrometry (LC–MS) analysis. Importantly, the identities of the phosphatidyl alcohols match those of the natural PA species produced by the PLDs (Bumpus and Baskin, 2017; Brown et al., 2007). Alternatively, click chemistry tagging using a bicyclononyne-fluorophore conjugate (e.g., bicyclononane BODIPY fluorophore [BCN-BODIPY]) generates fluorescent phosphatidyl alcohols, which can be visualized by fluorescence microscopy in live cells or quantified by HPLC from lipid extracts (Fig. 3 B; Bumpus and Baskin, 2017).

To eliminate background IMPACT signal from endogenous PLDs in our optoPLD validation experiments, we generated double knockout (DKO) HEK 293T cell lines of the two mammalian PLD isoforms, PLD1 and PLD2, using CRISPR/Cas9-mediated mutagenesis. Knockout was confirmed by Western blot and IMPACT using click chemistry tagging with BCN-BODIPY and HPLC and flow cytometry analyses (Fig. S1, C–E).

We expressed various optoPLDs in PLD DKO cells and performed IMPACT with click chemistry tagging using Alk-QA. LC–MS analysis revealed that optoPLDs targeted to the PM, endosomes, or ER produced transphosphatidylation products with acyl chain compositions similar to corresponding products of IMPACT labeling in WT HEK cells (i.e., coming from endogenous mammalian PLDs; Fig. 3 B). These data indicate that optoPLD has substrate preferences very similar to mammalian PLDs, which mirror the natural PA species produced by mammalian PLDs (Bumpus and Baskin, 2017). These LC–MS studies validate the potential of optoPLD to generate functionally relevant pools of PA in mammalian cells.

To test whether optoPLD is equally active when targeted to different organelle membranes, we performed single-cell analysis of optoPLD activity using IMPACT. To eliminate undesired IMPACT signal from endogenous PLDs, we again expressed the optoPLDs in PLD DKO cells, which were treated with AzProp with or without blue light (5-s pulses every 2 min), followed by rinsing and BCN-BODIPY click chemistry tagging in live cells. To quantify the extent of optoPLD activity at each organelle, we performed two-color flow cytometry after IMPACT labeling, enabling us to normalize cellular IMPACT-derived fluorescence (BODIPY, green) to optoPLD expression level (mCherry, red; Fig. S2 C).

Gratifyingly, optoPLDs recruited to all of the tested organelle membranes exhibited high and roughly equivalent activity when stimulated with blue light (Fig. 3 C). The ER-targeted optoPLD showed a small but statistically significant increase, in line with the slightly higher abundance of the PLD substrate PC in this organelle membrane (van Meer and de Kroon, 2011). IMPACT labeling from optoPLD in the dark was still slightly higher than that from the catalytic-dead optodPLD, suggesting that cytosol-localized CRY2-mCherry PLD_PMF had a small level of undesired background activity (Fig. 3 C).

To ensure that optoPLD had the highest possible signal to background, we optimized the interaction of PLD_PMF with the membrane by switching the positions of CRY2, mCherry, and PLD_PMF within the optoPLD constructs, creating mCherry-CRY2-PLD_PMF, CRY2-mCherry-PLD_PMF, PLD_PMF-mCherry-CRY2, and mCherry-PLD_PMF-CRY2. Each of these was expressed in PLD DKO cells with CIBN-PM, and two-color flow cytometry revealed that performed CRY2-mCherry-PLD_PMF exhibited the highest signal to background (i.e., IMPACT signal under blue light illumination vs. in the dark; Fig. S2 D). We thus continued to use this version of optoPLD for all future studies.

Mammalian PLD1 and PLD2 exhibit a wide range of physiologically relevant activities (i.e., between basal and fully activated states) as a result of their ability to be stimulated by several protein and lipid interactors (Du et al., 2000; Powner and Wakelam, 2002; Selvy et al., 2011; Petersen et al., 2016; Peng and Frohman, 2012; Bruntz et al., 2014). Because PLDs are regulated by changes in both subcellular localization and catalytic activity, we sought to develop a panel of optoPLDs with a range of activities to mimic the ability of cells to adjust endogenous PLD activity in different physiological contexts.

To accomplish this goal, we performed directed evolution of PLD_PMF. We developed a modified version of yeast display (Liu, 2015), termed yeast membrane display, for high-throughput screening of PLD_PMF mutants (termed PLD^#). Because PLD is an enzyme that acts on membrane lipids, we reasoned that, by tethering PLD^# members to an appropriate PC-rich organelle membrane within Saccharomyces cerevisiae, we could take advantage of IMPACT and FACS to enrich new PLD^# enzymes with desired catalytic activities.

Our yeast membrane display platform involved three steps: (1) PLD^# library construction, (2) cloning and expression in yeast, and (3) IMPACT labeling and FACS-based enrichment of yeast based on the expressed PLD^# activity (Fig. 4 A). First, we built a construct to express PM-binding PLD_PMF (PM-PLD), containing PLD_PMF, mCherry and the tandem pleckstrin homology domain of phospholipase C (2xPLCδ^PH), which binds to phosphatidylinositol (4,5)-bisphosphate in the PM, in a β-estradiol-inducible expression system (Várnai and Balla, 1998; Stauffer et al., 1998; McIsaac et al., 2011). Using this system, maximal PM-PLD expression was observed within 3 h of β-estradiol treatment, with no observed toxicity (Fig. S4 A). To determine PM-PLD activity in yeast, we adapted IMPACT to work in this organism. Fortunately, S. cerevisiae treated with AzProp exhibited only very low endogenous transphosphatidylation activity compared with yeast expressing PM-PLD (Fig. S4, B and C). Thus, AzProp labeling of yeast, followed by paraformaldehyde fixation and tagging via click chemistry with BCN-BODIPY, enabled the selective fluorescent labeling of PM-PLD–expressing cells (Fig. 4 B).

To construct a diverse PLD^# library, we performed random mutagenesis on PLD_PMF using error-prone PCR and performed IMPACT-FACS–based enrichment of yeast expressing PM-PLD^#. Yeast expressing the PM-PLD^# library exhibited a broad range of IMPACT labeling, indicating that changes in PLD^# activity caused by mutations are reflected in the IMPACT labeling efficiency (Fig. 4 C). After three enrichment rounds (Fig. S4 D) and subsequent DNA isolation and amplification, we determined the identities of the mutated residues in the enriched PLD^# population by next-generation sequencing.

Among the many potential hits, we selected six highly reoccurring PLD_PMF mutants for testing in mammalian cells: A40T, T49I, S168G, K294M, R375S, and G429D. We generated PM-targeted optoPLD^# constructs with each of these six mutant PLDs, expressed them in PLD DKO HEK 293T cells, and evaluated their PLD activities using IMPACT followed by flow cytometry. Excitingly, we discovered that these six optoPLD^#s exhibited a diversity of catalytic activities, ranging from as low as the basal activity of endogenous human PLDs in this cell line to as high as PLD activity elicited by very strong stimulation with PMA (Fig. 4 D). LC–MS analysis of the IMPACT-derived lipids produced by two representative optoPLD^#s (G429D and K294M) established that these enzymes have very similar acyl tail substrate preferences as WT PLD_PMF (Fig. S4 E).

These studies demonstrate that our optoPLD^# collection spans the physiological range of endogenous PLD activity. Notably, the strongest optoPLD^# (G429D) exhibited higher activity than WT PLD_PMF, which is the first example to our knowledge of an evolved PLD variant with higher catalytic ability for PA generation than its WT counterpart. A key benefit of this collection of optoPLD^#s is that PA formation can be tuned by simply swapping out the library member rather than by changing expression levels, which is harder to precisely control, particularly within a two-component, optogenetic heterodimerization system.

The availability of the G429D variant of PLD_PMF allowed us to directly visualize the subcellular localizations of optoPLD activity using real-time (RT) IMPACT, a recently reported, highly time-resolved IMPACT variant (Liang et al., 2019). In RT-IMPACT, AzProp is replaced with an oxo-trans-cyclooctene alcohol (oxoTCO) for the transphosphatidylation step, and its corresponding click chemistry partner is a fluorogenic tetrazine-BODIPY (Tz-BODIPY) dye, enabling tagging via the tetrazine ligation click chemistry reaction (Liang et al., 2019; Lambert et al., 2017; Carlson et al., 2013; Devaraj et al., 2008; Blackman et al., 2008; Wu and Devaraj, 2016; Oliveira et al., 2017; Fig. 5 A). Due to the rapid kinetics and fluorescence turn-on of the tetrazine ligation, the fluorescent lipid products of IMPACT can be visualized within tens of seconds after the beginning of the click chemistry reaction. During the development of RT-IMPACT, we discovered that IMPACT-derived fluorescent lipids produced by PM-localized endogenous PLDs trafficked rapidly, over tens of seconds, from the PM to the ER; however, early time points of the click chemistry time-lapse movie could reveal accurate localizations of IMPACT lipid products before the bulk of this rapid diffusion and trafficking to other organelle membranes (Liang et al., 2019).

Having characterized the rapid, unidirectional PM-to-ER trafficking of the fluorescent lipid reporters of PLD activity and now having optoPLDs in hand, we were eager to use RT-IMPACT to establish that PM- and ER-targeted optoPLDs had activity on the expected organelle membrane. Because oxoTCO is a poorer PLD substrate compared with AzProp (Liang et al., 2019) and shorter transphosphatidylation durations are optimal in RT-IMPACT, to minimize trafficking of lipid reporters during this step, we used the highly active G429D optoPLD^# to afford the highest possible IMPACT labeling.

Indeed, PLD DKO HEK 293T cells expressing PM-targeted optoPLD exhibited a substantial fraction of RT-IMPACT fluorescence at the PM, as well as some intracellular fluorescence likely at the ER (Liang et al., 2019), in an early time point of the time-lapse movie (Fig. 5 B). Interestingly, the internalization of these fluorescent lipids appears to be more rapid in HEK 293T cells than in HeLa or NIH 3T3 cells, which we used in the development of RT-IMPACT (Liang et al., 2019). Importantly, cells expressing ER-targeted optoPLD and labeled in an analogous manner exhibited strong ER signal and no fluorescence at the PM (Fig. 5, B and C). Thus, these studies, in conjunction with the PASS experiments (Fig. 2), support the conclusion that PM- and ER-targeted optoPLDs exhibit enzymatic activities at the expected and desired membrane locations.

Having established optoPLDs as tools capable of generating organelle-specific pools of PA, we sought to demonstrate the ability of optoPLD-derived PA to affect a cellular signaling pathway. Among the many potential pathways influenced by PA signaling, we focused our efforts on the Hippo pathway, a mechanism by which cells can control growth and proliferation (Yu et al., 2015). Hippo signaling reduces growth and proliferation by activating the large tumor suppressor 1/2 (LATS1/2) kinases, which phosphorylate a key progrowth transcription factor, Yes-associated protein (YAP), and prevent its translocation to the nucleus.

A recent study demonstrated that PLD-derived PA attenuates Hippo signaling, leading to the decreased phosphorylation and correspondingly increased nuclear translocation of YAP (Han et al., 2018a). Different PA effectors were proposed as relevant PA effectors, including the LATS1/2 kinases themselves as well as an upstream factor, neurofibromin 2 (NF2)/merlin. The effects of PA on Hippo signaling were determined both by loss of function (inhibition and knockout of PLDs) and by gain of function (addition of PA-containing liposomes; Han et al., 2018a). Given the dynamic localizations of endogenous PLDs and the multiple proposed PA effectors in this pathway, we reasoned that Hippo signaling would represent an excellent test case for our optoPLDs.

We first recapitulated a key finding of this study, demonstrating that addition of PA-containing liposomes to serum-starved HEK 293T cells caused a translocation of YAP from the cytosol to the nucleus, evaluated by immunofluorescence (IF; Fig. S5 A). We then used optoPLD to generate PA at the PM, the most likely location of action given the known site of action of NF2 in Hippo signaling (Yin et al., 2013). Excitingly, PA production by PM-targeted optoPLD induced a robust nuclear translocation of YAP, which we did not observe with either PM-optoPLD in the dark or using the catalytic-dead PM-optodPLD (Fig. 6 A). Intriguingly, when we replaced PM-optoPLD with versions targeted to other organelles (endosomes, TGN, or ER), we found that these optoPLDs were significantly less effective in downregulating the Hippo pathway (Fig. 6 B), despite their similar levels of PA-synthesizing activity (Fig. 3 C). These results both confirm the role of PA in allowing cells to attenuate Hippo signaling and, interestingly, suggest that local production of PA in the PM can control these signaling events.

Finally, using the sensitive and quantitative IF-based assay for YAP localization, we examined whether our collection of optoPLD^#s of varying activities could differentially influence Hippo signaling. Using PM-targeted versions of each of our six optoPLD^#s, we found that those with lower PLD catalytic activity, determined by IMPACT, caused less YAP nuclear translocation and those with higher PLD activity caused more translocation, meaning that the highly active optoPLD^# members were most effective in downregulating the Hippo pathway (Fig. 6 C). Remarkably, when we plotted the mean fraction of nuclear YAP against the mean intensity of IMPACT labeling (correlating to optoPLD catalytic activity), we observed a strong linear correlation (Fig. 6 D), indicating that by titrating more PA into the PM, we could precisely tune the strength of Hippo signaling. Altogether, these results demonstrate that the PA produced by optoPLD can downregulate the Hippo pathway in both a location- and dose-dependent manner.

In this study, we set out to develop an addition to the toolkit for studying the cell biology of PA. There are excellent pharmacological and genetic knockout tools for decreasing the activity of PA-synthesizing enzymes from the PLD, DGK, and LPAAT families (Su et al., 2009; Scott et al., 2009; Lewis et al., 2009; O’Reilly et al., 2014; Purow, 2015; McCloud et al., 2018; Sakane et al., 2016; Hideshima et al., 2003). However, using these tools to controllably tune local pools of PA can be challenging due to dynamic and still controversial localizations of these enzymes, as well as evidence of crosstalk and compensation between PA-synthesizing pathways. To address these challenges, we sought to develop a complementary gain-of-function tool, optoPLD, to study PA signaling with spatiotemporal precision.

Several factors were critical to the design of optoPLD. First was the selection of PLD_PMF as the desired exogenous PA-biosynthesizing enzyme, due to several factors, including its functionality in and orthogonality to eukaryotic cells. Second was the use of the CRY2–CIBN light-mediated dimerization system, which affords extremely fast turn-on kinetics and, equally desirable for sustained PA generation on the minute-to-hour timescale, relatively slow turn-off kinetics, necessitating only occasional and periodic illumination with low-toxicity, visible blue light (Kennedy et al., 2010; Tucker et al., 2014; Benedetti et al., 2018). We found dramatic differences in catalytic efficiencies of optoPLD constructs by varying the relative positions of CRY2, PLD_PMF, and mCherry, highlighting the importance of the order of proteins in chimeric protein design in this context. We suggest that, of the four constructs we investigated, CRY2-mCherry-PLD_PMF and PLD_PMF-mCherry-CRY2 exhibited higher activity because positioning mCherry in between CRY2 (the membrane anchor, through its interaction with CIBN) and PLD_PMF (the lipid-modifying enzyme) enhances the accessibility of PLD_PMF to the membrane due to the structural features of mCherry, whose N- and C-termini extend from the same end of the β-barrel (Shu et al., 2006).

Another factor key to the development of optoPLD was IMPACT, a chemical biology tool that we had originally designed to study the activity and localization of endogenous mammalian PLDs (Bumpus and Baskin, 2016, 2017, 2018). Here, we confirmed that IMPACT could be applied to quantify the activity of an exogenous, bacterial PLD within mammalian cells. In addition, the application of IMPACT in a new organism (S. cerevisiae) was key to our efforts to use directed evolution to generate a panel of optoPLDs with varying catalytic activities. For these protein engineering studies, we developed a high-throughput selection strategy, yeast membrane display, wherein libraries of membrane-tethered PLDs expressed in yeast could be enriched according to PLD activity using IMPACT labeling followed by FACS. Through these directed evolution studies, we produced a panel of PLD_PMF mutants (and, subsequently, optoPLD variants) with activities that were both lower and higher than WT PLD_PMF. We envision that our yeast membrane display platform could be widely useful for the further evolution of PLD_PMF mutants with varying and interesting catalytic properties.

As an application to validate the functionality of the PA produced by the optoPLDs, we revisited the recent discoveries concerning the roles of PLD and PA signaling in modulating the Hippo pathway, which controls cell growth and proliferation (Meng et al., 2018; Han et al., 2018a). In a detailed mechanistic study, Han et al. (2018a) proposed that PLD-derived PA downregulates Hippo signaling by binding to both NF2/merlin and LATS1/2, inhibiting the phosphorylation of YAP and ultimately leading to increases in YAP-dependent gene expression. Using similar experimental conditions, we recapitulated this finding by bulk addition of PA-containing liposomes to serum-starved HEK 293T cells. Excitingly, we found that optoPLD could, with organelle-level precision, cause a similar downregulation of Hippo signaling, but only when the optoPLD-generated PA was produced in the PM and not, at similar levels, on intracellular organelle locations (endosomes, TGN, or ER). This work implicates a pool of PA produced at the PM in allowing cells to downregulate, or bypass, the Hippo pathway.

Interestingly, concurrent work by Meng et al. examined the role of mechanotransduction on the Hippo pathway and concluded that PLD-derived PA had the opposite effect on Hippo signaling, namely that it promoted the pathway (Meng et al., 2018). Mechanistically, they proposed the Rab-like GTPase RAP2 as the relevant PA effector, but importantly, these studies were performed on cells grown in a low-stiffness environment (i.e., on collagen-coated hydrogels). Under those conditions, the Hippo pathway is controlled by the stiffness of the extracellular environment. In fact, the Hippo pathway is regulated by several upstream stimuli, including serum/growth factor availability, cell–cell contacts, and ECM stiffness, and the determination of the exact constellation of upstream factors that regulate Hippo signaling in different contexts is an interesting and active area of research (Meng et al., 2016).

Relevant to these studies, both serum starvation and low ECM stiffness activate the Hippo pathway and result in YAP exclusion from the nucleus. Although the Hippo activation caused by serum starvation was counteracted by optoPLD-derived PA, we found that the activation caused by low ECM stiffness overrode the effect of optoPLD (Fig. S5 B), consistent with the study by Meng et al. (2018). Thus our work using optoPLD lends support to the hypothesis emerging from these two studies that PA indeed has different effects on the Hippo pathway in different contexts and experimental conditions (Stampouloglou and Varelas, 2018; Han et al., 2018b). Future work using optoPLD in combination with other toolsets could reveal the causes and mechanisms underlying the differential roles of PA in modulating Hippo signaling.

In summary, optoPLD is a tool for precise, spatiotemporally controlled production of local pools of PA on individual organelle membranes that should complement existing toolsets for modulating PA using inhibition or knockout/knockdown of biosynthetic enzymes. OptoPLD interfaces with a variety of imaging-based tools for monitoring the dynamics of PA metabolism, including genetically encoded biosensors, FRET-based imaging of PLD–substrate binding events, and our recently developed chemoenzymatic method, IMPACT (Zhang et al., 2014; Kassas et al., 2012, 2017; Bumpus and Baskin, 2018; Kim et al., 2015; Lu et al., 2016; Hughes et al., 2002). We validated the physiological relevance of optoPLD-derived PA pools to confirm and elucidate differential roles of PA signaling in the modulation of Hippo signaling. We envision that our collection of optoPLDs, which in principle could be expanded to target additional organelles that are sites of PA-dependent signaling (e.g., lysosomes, mitochondria; Yoon et al., 2011, 2015; Yang and Frohman, 2012; Kameoka et al., 2018; Adachi et al., 2016), as well as potential implementation of PLD_PMF in chemical-induced dimerization systems (Stanton et al., 2018; DeRose et al., 2013), could prove to be widely useful for understanding roles of spatially and temporally defined pools of PA in controlling the many signaling pathways modulated by this multifunctional lipid signaling agent.

The following reagents were obtained from commercial suppliers: Lipofectamine 2000, Invitrogen (catalog number 11668019); poly-L-lysine, Sigma-Aldrich (catalog number P2636), 3-aminopropyltriethoxysilane, TCI (catalog number A0439); tetrakis(acetonitrile)copper(I) tetrafluoroborate (CuBF₄), Sigma-Aldrich (catalog number; glutaraldehyde solution, 50% in water, Beantown Chemical (catalog number 139070); dichlorodimethylsilane, VWR (catalog number 102572-962); acrylamide, 40% in aqueous solution, VWR (catalog number 97064-522); bis-acrylamide, 2% in aqueous solution, Alfa Aesar (catalog number J63265-K2); sulfo-SANPAH, VWR (catalog number 10189-584); tetramethylethylenediamine, VWR (catalog number 0761); ammonium persulfate, VWR (catalog number 0486); 5-gluoro-2-indolyl des-chlorohalopemide, Cayman Chemical (catalog number 13563); DGK inhibitor I (R59022), Enzo Life Sciences (catalog number ALX-430-027); PMA, ChemCruz (catalog number sc-3576); DTT, VWR (catalog number 97061-338); 8-oxo-dGTP, TriLink BioTechnologies (catalog numberN-2034); 2′-deoxy-P-nucleoside-5′-triphosphate, TriLink BioTechnologies (catalog number N-2037); β-estradiol: Sigma-Aldrich (catalog number E8875); 1,2-dioleoyl-sn-glycero-3-PC (DOPC), Cayman Chemical (catalog number 15098); 1,2-dioleoyl-sn-glycero-3-phosphate, Avanti Polar Lipids (catalog number 840875); mouse anti-PLD1, Santa Cruz Biotechnology (catalog number sc-28314, RRID:AB_677324); rabbit anti-PLD2, Cell Signaling Technology (catalog number 13904, RRID:AB_2798341); mouse anti-YAP, Santa Cruz Biotechnology (catalog number sc-101199, RRID:AB_1131430); goat anti-mouse HRP, Bio-Rad (catalog number 1706516, RRID:AB_11125547); donkey anti-Mouse Alexa Fluor 488, Invitrogen (catalog number A-21202, RRID:AB_141607); and goat anti-rabbit-HRP, Bio-Rad (catalog number 1706515, RRID:AB_11125142). The following compounds were prepared as described previously: BCN-BODIPY (Alamudi et al., 2016), AzProp (Yuan et al., 2012), oxoTCO ((R)-oxoTCO–C1; Lambert et al., 2017), and Tz-BODIPY (Carlson et al., 2013).

mCherry-CRY2-PLD_PMF was made by swapping INPP5E in mCherry-CRY2-INPP5E (a gift from the De Camilli laboratory, Yale University, New Haven, CT) for PLD_PMF (a coding sequence of PLD Streptomyces PMF, UniProt P84147_STRSM, primers: 5′-GCG​ATC​GGG​CGG​ATC​TGC​AGA​TTC​AGC​TAC​ACC​ACA-3′ and 5′-GTG​GTA​CCT​TAA​GCG​TTG​CAG​ATT​CCT​CTT​G-3′) using PvuI/KpnI. The CRY2 sequence corresponds to UniProt Q96524, amino acids 1–498; the CIBN sequence corresponds to UniProt A0A178V495, amino acids 1–170. CRY2-mCherry-PLD_PMF was cloned by first removing CRY2 from mCherry-CRY2-PLD_PMF using double XhoI sites to clone mCherry-PLD_PMF, followed by insertion of CRY2 using the megaprimer PCR method (primers: 5′-CGG​TGG​GAG​GTC​TAT​ATA​AGC​AGA​GCT​GGT​TTA​GCA​TGA​AGA​TGG​ACA​AAA​AGA​CTA​TAG-3′ and 5′-GGC​GCT​AGC​GGA​TCT​GAC​GGT​TCC​TGA​CCC​TCC​CGA​GCC​ACC​TGC​TGC​TCC​GAT​CAT​G-3′). For CIBN-Sac1, CIBN (amplified from CIBN-GFP-CAAX, a gift from the De Camilli laboratory, primers: 5′-TTT​AAA​CTT​AAG​CTT​GGT​ACC​ATG​AAT​GGA​GCT​ATA​GGA​G-3′ and 5′-CGC​CAC​CGG​TCA​CAT​GAA​TAT​AAT​CCG​TTT​TC-3′) and Sac1 (UniProt Q9NTJ5; amino acids 521–587, amplified from CFP-FRB-Sac1, a gift from the De Camilli laboratory, primers: 5′-TAT​TCA​TGT​GAC​CGG​TGG​CGC​CTT​CCT​G-3′ and 5′-AGT​GGA​TCC​GAG​CTC​GGT​ACT​CAG​TCT​ATC​TTT​TCT​TTC​TGG​ACC-3′) were cloned into a KpnI-digested pCDNA5-FRT-TO vector (Thermo Fisher) using Gibson assembly. CIBN-CAAX (the CAAX amino acid sequence is LYKGKKKKKKSKTKCVIM), CIBN-TGN (the TGN sequence is UniProt Q64685, amino acids 2–37), and CIBN-2xFYVE (the FYVE sequence is UniProt O14964, amino acids 147–220) were cloned by swapping out Sac1 in CIBN-Sac1 for CAAX (amplified from CIBN-GFP-CAAX, primers: 5′-GTA​CCG​GTG​GAT​CAG​GAT​CAC​TGT​ATA​AGG-3′ and 5′-TGG​GCC​CCT​ACA​TAA​TTA​CAC​ACT​TTG​TCT​TTG-3′), TGN (constructed by overlap-extension PCR, primers: 5′-GTA​CCG​GTA​TAC​ATA​CTA​ATC​TGA​AAA​AGA​AGT​TCA​GCC​TCT​TCA​TCC​TGG​TCT​TCC​TGC​TCT​TCG​CTG​TGA​TC-3′ and 5′-TGG​GCC​CTT​ACA​GTG​TAA​GGG​CCT​CGT​AAT​CAG​AGC​CCT​TCT​TCC​AAA​CGC​AGA​TCA​CAG​CGA​AGA​GCA​GG-3′), or 2xFYVE (amplified from RFP-2xFYVE, a gift from the De Camilli laboratory, primers: 5′-GTA​CCG​GTG​GCG​CCT​TAG​AAT​TCG​AAA​G-3′ and 5′-CAT​GAT​CTA​GAG​GGC​CCT​ATT​C-3′) using AgeI/ApaI. CIBN-iRFP-CAAX was cloned by replacing GFP in CIBN-GFP-CAAX with iRFP (Addgene; 31857) using the megaprimer PCR method (primers: 5′-GGA​ATT​GGA​GAA​AAC​GGA​TTA​TAT​TCA​TGT​GAC​CGG​TGG​ATC​AAT​GGC​TGA​AGG​ATC​CG-3′ and 5′-ACA​CTT​TGT​CTT​TGA​CTT​CTT​TTT​CTT​CTT​TTT​ACC​ACC​GGT​CTC​TTC​CAT​CAC-3′). CRY2-mCherry-PLD_PMF or mCherry-CRY2-PLD_PMF were amplified by PCR with insertion of a P2A sequence (primers: 5′-TTT​AAA​CTT​AAG​CTT​GGT​ACG​CCA​CCA​TGA​AGA​TGG​AC-3′ or 5′-TTT​AAA​CTT​AAG​CTT​GGT​ACG​CCA​CCA​TGG​TGA​GCA​AG-3′ and 5′-TAT​AGC​TCC​ATT​CAT​GGT​ACC​GGG​ACC​TGG​GTT​TTC​CTC​GAC​ATC​GCC​AGC​CTG​CTT​GAG​CAG​ACT​AAA​GTT​TGT​AGC​GCC​GCT​TCC​AGC​GTT​GCA​GAT​TCC​TCT​TG-3′), and the resultant product was cloned into KpnI-digested CIBN-X (X = Sac1, CAAX, TGN, or 2xFYVE) plasmids using Gibson assembly to clone CRY2-mCherry-PLD_PMF-P2A-CIBN-X or mCherry-CRY2-PLD_PMF-P2A-CIBN-X plasmids. For PLD_PMF-mCherry-CRY2-P2A-CIBN-X plasmids, PLD_PMF (primers: 5′-GCG​TTT​AAA​CTT​AAG​CTT​GCT​AAG​CCA​CCA​TGG​CAG​ATT​C-3′ and 5′-AGA​TCC​GCC​TCC​GGA​AGA​TCC​ACC​AGC​GTT​GCA​GAT​TCC​TCT​TG-3′), mCherry-CRY2 (amplified from PLD_PMF-mCherry-CRY2, primers: 5′-GGT​GGA​TCT​TCC​GGA​GGC​GGA​TCT​GTG​AGC​AAG​GGC​GAG​GAG-3′ and 5′-TAG​CGC​CGC​TTC​CTG​CTG​CTC​CGA​TCA​TGA​TC-3′), and P2A-CIBN-X (amplified from CRY2-mCherry-PLD_PMF-P2A-CIBN-X, primers: 5′-CGG​AGC​AGC​AGG​AAG​CGG​CGC​TAC​AAA​C-3′ and 5′-CTA​GAA​GGC​ACA​GTC​GAG​GCT​GAT​CAG​CGG​GTT​TAA​ACG-3′) were cloned into KpnI/ApaI-digested pCDNA5-FRT-TO vector using Gibson assembly. For mCherry-PLD_PMF-CRY2-P2A-CIBN-X plasmids, PLD_PMF (primers: 5′-CGC​CGA​GGG​CCG​CCA​CTC-3′ and 5′-TTC​GAG​ATC​TTC​CGG​AAG​ATC​CAC​CAG​C-3′) and CRY2-P2A (amplified from PLD_PMF-mCherry-CRY2-P2A-CIBN-X, primers: 5′-ATC​TTC​CGG​AAG​ATC​TCG​AAG​CGC​GGC​C-3′ and 5′-TAT​AGC​TCC​ATT​CAT​GGT​ACC​GGG​AC-3′) were cloned into BsrGI/KpnI-digested mCherry-CRY2-PLD_PMF-P2A-CIBN-X using Gibson assembly. A catalytically dead mutant of PLD_PMF was made by mutating histidine 170 to alanine using site-directed mutagenesis (primers: 5′-AAA​AAC​AGC​TTT​CTC​CTG​GAA​CGC​TAG​CAA​AAT​CTT​GGT​TGT​AGA​C-3′ and 5′-GTC​TAC​AAC​CAA​GAT​TTT​GCT​AGC​GTT​CCA​GGA​GAA​AGC​TGT​TTT​T-3′). iRFP-PASS was cloned by swapping out GFP in GFP-PASS (a gift from the Du Laboratory, University of Texas Health Science Center at Houston, Houston, TX) for iRFP (amplified from iRFP-C1 [Addgene; 54786], primers: 5′-GTA​GCT​AGC​GCC​ACC​ATG​GCT​GAA​GG-3′ and 5′-GTC​TGT​ACA​TCT​CTT​CCA​TCA​CGC​CGA​TC-3′) using NheI and BsrGI. For Gal1-PLD_PMF-mCherry-2xPLCδ^PH, Gal1 (amplified from pFA6a-TRP1-PGAL1-3HA [a gift from the Emr laboratory, Cornell University, Ithaca, NY], primers: 5′-GGA​TCC​ACT​AGT​TCT​AGA​GCG​GCC​GCT​CCT​TGA​CGT​TAA​AGT​ATA​GAG​GTA​TAT​TAA​CAA​TTT​TTT​G-3′ and 5′-CCA​TGA​CGT​CCG​GAT​TAG​AAG​CCG​CCG​AG-3′), PLD_PMF-mCherry (amplified from PLD_PMF-mCherry-CRY2, primers: 5′-TTC​TAA​TCC​GGA​CGT​CAT​GGC​AGA​TTC​AGC​TAC​AC-3′ and 5′-GGA​GCC​CGT​GGG​ATC​TGA​GTC​CGG​ACT​TG-3′), 2xPLCδ^PH (amplified from pGPD-mCherry-2xPLCδ^PH [a gift from the Emr laboratory], primers: 5′-ACT​CAG​ATC​CCA​CGG​GCT​CCA​GGA​TGA​C-3′ and 5′-TAT​ATG​ATC​ACT​ACT​TCT​GCC​GCT​GGT​CCA​TG-3′), and Yps1_term (amplified from S. cerevisiae genomic DNA, primers: 5′-GCA​GAA​GTA​GTG​ATC​ATA​TAG​TTC​CAT​CTT​TAC-3′ and 5′-TGG​AGC​TCC​ACC​GCG​GTG​GCA​GGC​GTT​CTG​AAA​AGA​AG-3′) were cloned into NotI-digested pRS416 vector using Gibson assembly.

HEK 293T cells were grown in DMEM (Corning) supplemented with 10% FBS (Corning) and 1% penicillin/streptomycin (Corning) at 37°C in a 5% CO₂ atmosphere. For serum starvation, HEK 293T cells were cultured in DMEM supplemented with 1% penicillin/streptomycin (serum-free medium). Cells were transfected 20–24 h after seeding using Lipofectamine 2000 (Thermo Fisher) by following the manufacturer’s protocol. Briefly, cells were incubated in regular growth media or serum-free media containing plasmids premixed with Lipofectamine 2000 (0.33 µg DNA and 0.75 µl Lipofectamine 2000 per well for a 24-well plate). Experiments were performed 20–22 h after transfection.

A lentiviral plasmid for PLD1 knockout was made by cloning the sgRNA (5′-GTG​AGC​CCA​CAA​ATA​GAC​GG-3′) into the LentiCRISPRv2-Puro vector (Addgene; 98290), and a lentiviral plasmid for PLD2 knockout was made by cloning the sgRNA (5′-GGC​ACC​GAA​AGA​TAT​ACC​AG-3′) into the LentiCRISPRv2-Blast vector (Addgene; 98293). 7.5 µg of each LentiCRISPR plasmid was cotransfected with packaging plasmids (5.6 µg of Gag and 1.8 µg of Env) into HEK 293TN cells seeded on a 10-cm dish using Lipofectamine 2000. The virus-containing medium was collected 48 h and 72 h after transfection, and the cell debris was removed by passing through a 0.45-µm filter. For lentiviral transduction, HEK 293T cells seeded on a 6-well plate were treated with transduction medium (freshly prepared by mixing 1.5 ml virus-containing medium, 0.5 ml DMEM media, and polybrene to a final concentration of 8 µg/ml) for 48 h. During the transduction, the transduction medium was replaced every 12 h. After transduction, the cells were placed in regular DMEM for 24 h and then reseeded on a 10-cm dish in DMEM containing 1 µg/ml puromycin (for PLD1 knockout), 5 µg/ml blasticidin (for PLD2 knockout), or 1 µg/ml puromycin + 5 µg/ml blasticidin (for PLD1/2 knockout, i.e., DKO). The cells were kept in the selective medium for 3–5 d until all cells on a nontransduced control plate had died. The selected cells were lysed in RIPA lysis buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1% Triton-X, 0.5% sodium deoxycholate, 0.1% SDS, and 1 mM EDTA) for subsequent Western blot analysis, with detection by chemiluminescence using the Clarity Western ECL Substrate (Bio-Rad) and acquisition on a Bio-Rad ChemiDoc MP System.

A homemade light box was built by attaching four strips of dimmable, 12 V blue-LED tape light (1000Bulbs.com; 2835–60-IP65-B1203) on the inside of a styrofoam box. For optogenetics experiments, the light box was placed in the CO₂ incubator using an AC Outlet Power Bank (Omars; 24,000 mAh, 80 W) as a power supply. To switch the light on and off without disturbing the controlled atmosphere (i.e., temperature and CO₂ levels) in the incubator, a Wi-Fi controlled outlet (ECOPlugs) was used to enable remote control via a smartphone. For optogenetic stimulation of cells, light illumination was applied for 5-s intervals every 2 min for the duration of the experiment unless otherwise noted.

1 d (20–24 h) before transfection, HEK 293T cells were seeded on 35-mm glass-bottom imaging dishes (MatTek Corporation) for live-cell imaging or 24-well plates (Corning) for flow cytometry, both of which were pretreated with 50 µg/ml poly-L-lysine for coating. For IMPACT labeling using AzProp and BCN-BODIPY (Bumpus and Baskin, 2017), cells were treated with 2 mM AzProp in regular growth media for 40 min at 37°C with or without light illumination as above and rinsed three times with PBS (pH 7.4). Then the cells were treated with 1 µM BCN-BODIPY for 10 min at 37°C, rinsed another three times with PBS, and then incubated in regular growth media for 10 min at 37°C to allow for remaining excess, unreacted BCN-BODIPY to be washed out. For imaging, media was replaced with imaging buffer (135 mM NaCl, 5 mM KCl, 20 mM Hepes, 1 mM CaCl₂, 100 µM MgCl₂, 1 mg/ml BSA, and 1 mg/ml glucose). For flow cytometry, cells were lifted with trypsin-EDTA (Corning; 0.25% trypsin and 0.1% EDTA). For RT-IMPACT labeling using oxoTCO and Tz-BODIPY (Liang et al., 2019), serum-starved cells were treated with 3 mM oxoTCO in serum-free media for 3 min at 37°C with light illumination as above and rinsed three times with PBS. After adding 100 µl of PBS onto the center of the imaging dish to keep the cells wet, cells were subjected to imaging with a time-lapse movie. Between the second and third frames of the movie, 100 µl of 0.66 µM Tz-BODIPY was added to initiate the click chemistry. Images were acquired at room temperature using Zeiss Zen Blue 2.3 on a Zeiss LSM 800 confocal laser scanning microscope equipped with Plan Apochromat objectives (40× 1.4 NA) and two gallium arsenide phosphide detectors. Solid-state lasers (405, 488, 561, and 640 nm) were used to excite DAPI, BODIPY, mCherry, and iRFP, respectively. Acquired images were processed using FIJI. For flow cytometry, cells were analyzed using a BD FACSAria Fusion Fluorescence Activated Cell Sorter. Acquired data were analyzed using FlowCytometryTools via Python.

Fluorescence-based measurement of mammalian PLD activity was performed as described previously (Bumpus and Baskin, 2017). Cells were treated with 5 mM AzProp in regular growth media for 40 min at 37°C, rinsed once with PBS, and subjected to lipid extraction using a modified Bligh-Dyer extraction. Cells were scraped in 250 µl methanol, 125 µl acetic acid (20 mM in water), and 100 µl PBS using cell scrapers (Corning), and the suspension was transferred into 1.5-ml microcentrifuge tubes. After addition of 500 µl chloroform to the suspension, the mixture was vortexed vigorously for 5 min and centrifuged at 17,000 ×g for 1 min. The organic layer (bottom) was collected in a new tube and dried under a stream of N₂ gas to form a lipid film. The film was dissolved in 40 µl chloroform/methanol/water (73:23:3) with 5 µM BCN-BODIPY and incubated overnight at 42°C. The solution was then diluted with 70 µl of a chloroform/methanol/water mixture (73:23:3), filtered through a syringe-driven filter unit (Millex; 0.45 µm), and subjected to analysis using a Shimadzu Prominence HPLC equipped with an in-line fluorescence detector. Separation was achieved using a Luna 3 µm Silica LC Column (Phenomenex; 150 × 4.6 mm) with a binary gradient elution system where solvent A was chloroform/methanol/ammonium hydroxide (95:7:0.5) and solvent B was chloroform/methanol/water/ammonium hydroxide (60:34:5:0.5).

Measurement of PLD transphosphatidylation activity or levels of cellular PA by LC–MS was performed as described previously (Bumpus and Baskin, 2017). Cells were treated with either 5 mM AzProp in regular growth media for 40 min at 37°C (for measuring transphosphatidylation activity) or with the indicated inhibitor (for measuring levels of PA), rinsed once with PBS, and subjected to Bligh-Dyer lipid extraction as described above. The lipid film was dissolved in 7 µl of chloroform, followed by addition of 30 µl of a click chemistry reaction master mix (prepared freshly by mixing 78 µl of 10 mM Alk-QA [Bumpus and Baskin, 2017] in degassed methanol, 2.95 µl of 25 mM CuBF₄ in DMSO, and 312 µl of degassed ethanol). The tube was then flushed with argon gas and incubated overnight at 42°C. The incubated solution was diluted with 200 µl of a chloroform/methanol/water mixture (73:23:3), filtered, and subjected to high-resolution LC–MS analysis on an Agilent 6230 electrospray ionization–time-of-flight MS coupled to an Agilent 1260 HPLC equipped with a Luna 3 µm Silica LC Column (Phenomenex; 50 × 2 mm) using a binary gradient elution system where solvent A was chloroform/methanol/ammonium hydroxide (85:15:0.5) and solvent B was chloroform/methanol/water/ammonium hydroxide (60:34:5:0.5).

S. cerevisiae (strain SEY6210) stably expressing Gal4dbd.ER.VP16 (GEV) was obtained from the Emr laboratory. Yeast were grown in yeast extract–peptone-dextrose (YPD) medium before transformation, and yeast nitrogen base (YNB; supplemented with necessitate amino acids) medium was used after transformation. Plasmids were transformed into yeast cells using electroporation by using slight modifications to a standard protocol (Richman et al., 2009). Briefly, an overnight yeast culture was diluted to approximately OD 0.1 in 100 ml of YPD and shaken for 5–6 h at 30°C. The cells were pelleted by centrifugation in a 50-ml conical tube (4,000 ×g for 3 min), resuspended in 10 ml LiOAc-TE buffer (0.1 M lithium acetate, 10 mM Tris-HCl, pH 8.0, and 1 mM EDTA), and gently shaken for 45 min at 30°C. The culture was added with 250 µl DTT (1 M in water), gently shaken for 15 min at 30°C, transferred into a 50-ml conical tube, diluted with 40 ml of water, and centrifuged at 4,000 ×g for 3 min. The cells were then rinsed by centrifugation and resuspension three times at 4°C with 25 ml of water, 3 ml of 1 M sorbitol, and finally 50 µl of 1 M sorbitol. To 40 µl of the obtained cell slurry, 50 ng of linearized vector and 50 ng of insert were added (total volume of DNA was <5 µl). The mixture was incubated on ice for 20 min, transferred into a prechilled 0.2-cm-gap electroporation cuvette, and pulsed at 1.5 kV, 25 µF, 200 Ω using a Bio-Rad Gene Pulser. Immediately after the pulse, the mixture was diluted with 1 ml of cold 1 M sorbitol and incubated, with shaking, in 10 ml of YPD + 1 M sorbitol for 3 h at 30°C to increase the survival rate of transformed yeast cells. Cells were then resuspended in 50 ml of YNB (supplemented with appropriate amino acids and lacking uracil) and shaken for 48 h at 30°C. The culture was then diluted by a factor of 1:100 in 50 ml YNB and shaken for another 24 h to reduce untransformed cells.

Transformed yeast cells were subcultured to OD ∼0.2 in 10 ml of YNB and shaken for 3–4 h at 30°C. When cell density was OD 0.8, 20 µl β-estradiol (100 µM in water, final concentration of 200 nM) was added to the culture to induce PLD expression. For IMPACT labeling, cells after the induction were harvested for OD 3 equivalent, resuspended in 1 ml YNB supplemented with 10 mM AzProp and 200 nM β-estradiol and shaken for 1 h at 30°C. The cells were then washed three times by centrifugation at 10,000 ×g for 1 min and resuspension in 0.1 M potassium phosphate buffer (prepared by mixing 0.1 M K₂HPO₄ and 0.1 M KH₂PO₄ to pH 6.5). Cells were fixed by resuspension of the pellet in 3.7% formaldehyde in 680 µl of potassium phosphate buffer and gentle shaking for 45 min at room temperature. The fixed cells were washed three times with 0.1 M potassium phosphate buffer, resuspended in 300 µl of potassium phosphate buffer supplemented with 1 µM BCN-BODIPY, and shaken for 30 min at 30°C for fluorescent tagging via click chemistry. The labeled cells were washed three times with 0.1 M potassium phosphate buffer and then imaged using a Zeiss LSM 800 confocal laser scanning microscope or sorted using a BD FACSAria Fusion Fluorescence Activated Cell Sorter. In FACS, 1–2 × 10⁵ cells were collected for subsequent plasmid extraction.

Prior to plasmid extraction, yeast cells enriched by FACS were supplemented with a 1,000-fold excess of nontransformed yeast cells (e.g., 10⁸ nontransformed cells for 10⁵ of collected cells) to increase total DNA concentration in the solution and aid in efficient DNA recovery during the plasmid extraction. After centrifugation (10,000 ×g for 1 min), 200 µl of glass beads (Scientific Industries; SI-BG05) and 200 µl of miniprep buffer 1 (50 mM Tris-HCl, 10 mM EDTA, and 100 µg/ml RNaseA, pH 8.0) were added to the cell pellet, followed by vigorous vortexing for 10 min. The mixture was mixed with 200 µl of miniprep buffer 2 (0.2 M NaOH and 1% SDS), incubated for 10 min at room temperature, mixed with 280 µl of miniprep buffer 3 (4.2 M guanidine-HCl and 0.9 M potassium acetate, pH 4.8), and centrifuged for 10 min at 4°C. The supernatant was transferred into a miniprep column (BioBasic; EZ-10). After washing the column once with 750 µl of miniprep buffer 4 (10 mM Tris-HCl and 80% ethanol, pH 8.0), the DNA was eluted into 30 µl of water. The PLD fragment was amplified by a two-step, nested PCR sequence, using 1 µl of eluted DNA for a 20-µl reaction mixture for the first PCR reaction. The PLD fragment amplified by nested PCR was then mutagenized by error-prone PCR using dNTP analogues (8-oxo-dGTP and 2′-deoxy-P-nucleoside-5′-triphosphate) according to established procedures (Angelini et al., 2015).

Cells were treated with 50 mM AzProp in YNB for 1 h at 30°C and rinsed three times with 0.1 M potassium phosphate by centrifugation at 10,000 ×g for 1 min. To the cells were added 200 µl of methanol and 200 µl of glass beads, and the mixture was vortexed vigorously for 10 min. To the mixture was then added 100 µl of acetic acid (20 mM in water), 80 µl of 0.1 M potassium phosphate buffer, and 400 µl of chloroform, followed by further vortexing for 5 min. Afterward, the lipid extract was tagged with BCN-BODIPY by click chemistry and analyzed by HPLC as described previously for mammalian cells.

2D hydrogels of different stiffnesses were prepared following established procedures (Tse and Engler, 2010). Briefly, 125 µl of 0.1 M NaOH was evaporated on a 12-mm cover glass (Fisherbrand, Thermo Fisher) to coat the glass with a uniform NaOH film. The glass was then treated with 50 µl of 3-aminopropyltriethoxysilane for 5 min, rinsed extensively with water, incubated in 2 ml of 0.5% glutaraldehyde in PBS for 30 min, and dried under air after aspiration of the glutaraldehyde solution. In parallel, a microscope slide (Fisherbrand, Thermo Fisher) was treated with 200 µl dichlorodimethylsilane for 5 min, rinsed for 1 min with water, and wiped with a Kimwipe to dry. For preparing stiff hydrogels (∼40 kPa), 2 ml of 40% acrylamide and 2.4 ml of 2% bisacrylamide were mixed with 5.6 ml water. For preparing soft hydrogels (∼1 kPa), 1.25 ml of 40% acrylamide and 0.15 ml of 2% bisacrylamide were mixed with 8.6 ml of water. To the solution, 1/100th volume of 10% ammonium persulfate and 1/1,000th volume of tetramethylethylenediamine were added. A small amount (6.25 µl) of the gel solution was quickly applied onto the chloro-silanated microscope slide, and then the amino-silanated cover glass was placed on the top. The gel solution was allowed to polymerize for 15 min, and then the microscope slide was carefully removed to obtain a hydrogel that was adhered to the cover glass. The hydrogel was rinsed three times with water for 5 min each, treated with 125 µl of 0.4 mM sulfo-SANPAH in 50 mM Hepes under 365-nm UV illumination (Thermo Fisher; catalog number UVP95004207) for 15 min, rinsed three times with 50 mM Hepes, and incubated in 50 µg/ml collagen type I (Corning; catalog number 354236) in 50 mM Hepes at 37°C for 8–16 h. After rinsing two times with PBS, the collagen-coated hydrogel was submerged in PBS, placed in a tissue culture hood for 30 min under UV irradiation for sterilization, and used for cell culture.

1 d (20–24 h) before transfection, HEK 293T cells were seeded on either 12-mm cover glass (Fisherbrand) pretreated with poly-L-lysine or homemade 2D hydrogels coated with collagen type I (described above). Approximately 20 h after transfection in serum-free media, optoPLD was activated by 488-nm light illumination (5 s of illumination at 2-min intervals) for 30 min at 37°C. The cells were then fixed in 4% formaldehyde for 10 min at room temperature, rinsed three times with PBS, permeabilized with 0.5% Triton X-100 in PBS for 5 min at room temperature, and blocked with 1% BSA and 0.1% Tween-20 in PBS (blocking buffer) for 30 min. The cells were then treated with a 1:100 dilution of anti-YAP antibody in blocking buffer for 1 h at room temperature, rinsed three times with 0.1% Tween-20 in PBS (PBS-T), treated with a 1:1,000 dilution of anti-mouse–Alexa Fluor 488 antibody conjugate in blocking buffer for 1 h at room temperature, and rinsed three times with PBS-T. After the final rinse, the cells were incubated in PBS-T for 5 min, mounted on microscope slides (Fisherbrand, Thermo Fisher) using ProLong Diamond Antifade Mountant with DAPI (Thermo Fisher), and incubated overnight at room temperature in the dark before imaging. Image acquisition by laser-scanning confocal microscopy was performed as described above by using solid-state lasers (405, 488, 561, and 640 nm) to excite DAPI, Alexa Fluor 488, mCherry, and Alexa Fluor 647, respectively. Acquired images were analyzed using FIJI and Python.

For experiments involving quantification of comparisons between more than two independent groups, statistical significance was calculated using the Kruskal–Wallis test, followed by the Games–Howell post hoc test, operated using the StatsModels package in Python (for Kruskal–Wallis) and userfriendlyscience package in R (for Games–Howell). For confocal microscopy, colocalization was calculated in FIJI using Coloc 2 plugin to determine Pearson correlation coefficients. All experiments were performed in at least three biological replicates on different days. Exact numbers of replicate experiments and sample sizes are provided in each figure legend.

Fig. S1 includes MS data, HPLC traces, and Western blot data to characterize PA levels, PLD levels, and PLD activity in PLD-deficient HEK 293T cells. Fig. S2 includes confocal microscopy colocalization analysis to characterize optoPLD localization and flow cytometry plots of experiments to optimize optoPLD activity. Fig. S3 includes confocal micrographs of PA biosensor and optoPLD localization during a time course of optoPLD activation and inactivation. Fig. S4 includes confocal micrographs, HPLC traces, and MS data to characterize the PM-targeted PLD in yeast used for directed evolution. Fig. S5 includes confocal micrographs of IF staining to evaluate changes in YAP localization upon optoPLD activation in cells grown on hydrogels of different stiffnesses.

We thank the Cornell Biotechnology Resource Center Imaging Facility (National Science Foundation grant DBI-1428922). We thank Alex Batrouni, Jeff Jorgensen, Sho Suzuki, and Lu Zhu for helpful discussions; the Emr, Fromme, and Yu laboratories (Cornell University, Ithaca, NY) for equipment and reagents; and Jonathan Lee, Paris Ghazi, and Adam Wojno for assistance.

J.M. Baskin acknowledges support from a Beckman Young Investigator award, a Sloan Research Fellowship, and the National Science Foundation (CAREER CHE-1749919). R. Tei was supported by Funai Overseas and Cornell Fellowships.

The authors declare no competing financial interests.

Author contributions: R. Tei and J. Baskin conceived of the project, designed experiments, interpreted the results, and wrote the manuscript. R. Tei carried out all experiments and data analysis.