A PtdIns(4)P-driven electrostatic field controls cell membrane identity and signalling in plants

Abstract

Many signalling proteins permanently or transiently localize to specific organelles. It is well established that certain lipids act as biochemical landmarks to specify compartment identity. However, they also influence membrane biophysical properties, which emerge as important features in specifying cellular territories. Such parameters include the membrane inner surface potential, which varies according to the lipid composition of each organelle. Here, we found that the plant plasma membrane (PM) and the cell plate of dividing cells have a unique electrostatic signature controlled by phosphatidylinositol-4-phosphate (PtdIns(4)P). Our results further reveal that, contrarily to other eukaryotes, PtdIns(4)P massively accumulates at the PM, establishing it as a critical hallmark of this membrane in plants. Membrane surface charges control the PM localization and function of the polar auxin transport regulator PINOID as well as proteins from the BRI1 KINASE INHIBITOR1 (BKI1)/MEMBRANE ASSOCIATED KINASE REGULATOR (MAKR) family, which are involved in brassinosteroid and receptor-like kinase signalling. We anticipate that this PtdIns(4)P-driven physical membrane property will control the localization and function of many proteins involved in development, reproduction, immunity and nutrition.

Extended Data Figure 1. Construction of membrane surface charge (MSC) probes and additional MSC reporters confirming the high electronegativity of the plasma membrane in plants

a) Sequence alignment between the different MSC probes showing the polybasic stretch in each construct (or the associated mutations) and their respective net positive charges. Cationic residues (K, R) are in red, acidic residues (E, D) in light blue, hydrophobic and aromatic residues in the amphipatic helixes Rit-tail and KRphy are in dark blue (F, W, L, V), the C-terminal farnesylation sequence CVIM (CxxM box) is in green, the C-terminal geranylgeranylation sequence CAIL (CxxL box) is in purple and, the N-terminal myristoylation sequence MGSSK is in pink. b) Schematic representation of the lipid modifications used in our MSC probes: myristoyl (pink, N-terminal modification, covalently linked to the second glycine), farnesyl (green, C-terminal modification, covalently linked to the cysteine in the CxxM motif) and geranylgeranyl (purple, C-terminal modification, covalently linked to the cysteine in the CxxL motif). c–f) Representative confocal images of root epidermal cells expressing the indicated MSC probe. All the constructs are expressed by the pUBQ10 promoter and are fused with cYFP at their N-terminus, except the myristoylated constructs, which are fused with cYFP at their C-terminus. bars, 5μm. c) Localization of the full collection of farnesylated probes from 8Q-Farn (0+) to 8K-Farn (8+). The farnesylated MSC probes are based on the C-terminal tail of the human small GTPase K-Ras4B. K-Ras is targeted to the PM via a C-terminal farnesyl anchor in conjunction with an adjacent unstructured polybasic peptides made of 8 lysines. Our bioprobes consist of a fusion between a tandem repeat mCITRINE fluorescent protein (cYFP) and the K-Ras C-terminal tail, in which we modified the net positive charges via site directed mutagenesis of the lysine stretch. The least cationic probe (0+), in which 8 lysines have been substituted by glutamine (8Q-Farn), is localized in numerous endomembrane compartments. This suggests that farnesylation of the 8Q-probe is sufficient to provide membrane anchoring in the absence of its adjacent lysines and that this probe, which is targeted mainly by hydrophobic interactions, confers targeting to intracellular membranes. The gradual addition of cationic charges should increase electrostatic interactions with anionic lipids and thereby shift the probes localization toward more negatively charged membranes. Indeed, we observed that the more cationic the probe is, the more it is targeted to the PM at the expense of endomembrane localization. d) The cysteine in the CxxM motif of 8K-Farn was substituted by an alanine thereby prohibiting C-terminal addition of a farnesyl lipid anchor (8K-noFarn). This non-farnesylated probe failed to associate with any membrane and was fully soluble, despite being strongly cationic (8+). This suggests that electrostatic interactions by themselves are not sufficient for membrane targeting and that stable membrane association requires some type of hydrophobic interactions. e) Localization of the Myr-8Q (0+), Myr-4K4Q (4+) and Myr-8K (8+) probes. Note that, by contrast with the 8Q-Farn (Extended Data Fig. 1c, top left pannel) and 8Q-Gege probes (see main Fig. 1i), the Myr-8Q probe is already partly associated with the PM in the absence of electrostatic interactions. This showed that these different lipid anchors have different intrinsic targeting properties but that they each failed to provide PM specificity on their own. Nonetheless, like for the farnesylated reporters, the gradual addition of net positive charges next to the myristoyl modification gradually increases PM association: Myr-4K4Q (4+) has an intermediate PM/endomembrane localization and Myr-8K is specifically localized at the PM. Together, our results support the notion that strong electrostatic interactions provide PM specificity regardless of the lipid anchor type. f) Localization of the KA1^Kcc4p reporter at the PM and in the nucleus. Similar to KA1^MARK1, KA1^Kcc4p is specifically localized at the PM and not in endomembrane compartments, confirming that this specific localization at the cell surface is a property of the KA1 domain in general rather than a specific feature of the MARK1 protein. However, unlike KA^MARK1, KA1^Kcc4p was also partly localized in the cytosol and the nucleus, which makes this domain less convenient as a MSC readout. For this reason, from now on, we decided to use the KA1^MARK1 domain in subsequent experiments. g) Sensitivity of KA1^MARK1 (left and middle panel) and 8K-Farn (8+ probe, right panel) to 90 min of BFA treatment at the indicated concentration. To show that BFA was active during our treatment we used the endocytic tracer FM4-64 and found that it was accumulated in BFA bodies at both 25 μM and 100 μM of BFA. FM4-64 was used at 1 μM and added 10 min prior confocal observations in the continuous presence of BFA.

Extended Data Figure 2. The high electronegativity of the PM is a common feature of many cell types and at least two plant species

a) Localization of KA1^MARK1, 8K-Farn (8+), 8Q-Farn (0+) in transiently transformed N. benthamiana leaves. Blue arrowheads show PM localization and yellow arrows show endomembrane localization. b–c) Confocal picture of b) the shoot and c) the root of transgenic Arabidopsis lines stably expressing cYFP-KA1^MARK1. Scale bars, 10 μm.

Extended Data Figure 3. Polarity indices in Arabidopsis root epidermis of various fluorescent PM proteins

Charts showing the polarity index for each fusion protein indicated at the bottom. Different italicized-letters indicate significant differences among means (P<0.0001, Tukey’s test). Note that only PIN2-GFP (red) is significantly different from all the other genotypes. The polarity indices of phosphoinositide (green) and MSC (pink) sensors fluctuate between 1.2 and 1.4, which is close to the numbers reported for PI4P and PI(4,5)P₂ reporters. However, we found that expected non-polar controls (blue), including the lipid dye FM4-64 and plasma membrane proteins Lti6b and PIP2a aquaporin (aqPIP2a) have similar polarity indices. Therefore, we could not detect significant statistical differences between our phosphoinositides or MSC sensors and expected non-polar controls. Although we cannot exclude that these non-polar control are in fact polar, we favor the hypothesis that confocal images of root cells might be biased for apical/basal signal over lateral signal because of the topology of these cells. First, the apical pole of one cell is tighlty juxtaposed to the basal pole of its neighbouring cell, which tend to enhance the apparent apical/basal signal over the lateral one. In addition, pinhole-based microscopes have high thickness of the z-sections. As a result, the z resolution is much lower than x and y resolution, so the volume collected by the microscope is not an isodiametric cube but cuboid; therefore a straight membrane in z will appear stronger than a curved one - which is the case of the apical and basal root membranes compared to the lateral membranes. Therefore we concluded that phosphoinositides and PM MSC are likely not polar in Arabidopsis root epidermis. Method for quantification of polarity index. 7 days old transgenic lines were analyzed to determine the “Polarity index” in root tip epidermis. “Polarity index” is the ratio between the fluorescence intensity (Mean Grey Value function of Fiji software) measured at the PM apical/basal side and PM lateral sides (Line width=3). We selected only cells for which the PM at each pole (apical, basal and laterals) were easily viewable and we selected cells that were entering elongation (at least as long as wide, but no more than twice as long as wide). Quantification was conducted in 100 cells over more than 15 independent plants. This Polarity index reveals the degree of polarity of the fluorescent reporters between the apical/basal side and lateral sides of the PM.

Extended Data Figure 4. Sensitivity of phosphoinositides and MSC sensors to PI3K and PI4K inhibitors

a) Schematic representation of the action of the drugs used to perturb phosphoinositides production and lipid sensors used as read-out. b–e) Confocal pictures of Arabidopsis root epidermis from the genotype indicated on the left, treated with the drug concentration indicated at the top for 90 min (mock, LY294002 and WM) or 30 min (PAO). b) cYFP-2xFYVE^HRS. As reported previously, 90 min of PI3K inhibition leads to swelling of late endosomes labelled by the PI3P sensors 2xFYVE^HRS, rather than a release of the probe into the cytoplasm. Yellow arrowheads show enlarged endosomes. Endosome swelling suggested that WM and LY294002 are active, although we noticed that WM had a more drastic effect at 30 μM than 1 μM. On the other hand, PAO treatment had no effect on 2xFYVE^HRS-labelled endosomes. c) cYFP-1xPH^FAPP1. PI3K inhibition by LY294002 and 1 μM WM had no effect on 1xPH^FAPP1 localization. However PI3K and PI4K inhibition by 30 μM WM partially released 1xPH^FAPP1 into the cytosol and PI4K inhibition by PAO fully solubilized this PI4P sensor. In the 60 μM 30 min PAO treatment (right) both the PM and endosomal pools of 1xPH^FAPP1 were solubilized. This result is surprising given that the endosomal pool of 1xPH^FAPP1 can rely only on ARF1-binding for endomembrane localization (See Fig. 3 of main text). The PH domain of FAPP1 interacts specifically with GTP-loaded ARF1 and it is possible that PI4K inhibition inhibits ARF1 activation. For example, the ARF GTPase Activating Protein (ARF-GAP) VAN3, which binds ARF1 in Arabidopsis, has a PI4P-binding PH domain and its GAP activity is enhanced by PI4P. d) cYFP-2xPH^PLC. Only PAO 60 μM 30 min (right) had a slight effect on the PM localization of the PI(4,5)P₂ bionsensor 2xPH^PLC, which becomes significant after prolonged treatment (45 to 60 min of 60 μM PAO, see g). e) cYFP-C2^Lact. Inhibition of PI3K and/or PI4K had no effect on the PM localization of the PS bionsensor C2^Lact. However, we noticed that 60 μM PAO for 30 min (right) decreased the number of endomembrane compartments labeled by this probe, suggesting some impact of PI4K activity on the intracellular localization of PS. f) Confocal picture of Arabidopsis root epidermis from the genotype indicated at the top, treated with the drug concentration indicated on the left for 30 min. g) Dissociation index (mean ±SEM) for the genotype and drug concentration indicated at the bottom. All treatments were performed during 30 min, except when indicated otherwise. Different italicized-letters indicate significant differences among means (P<0.0001, Kruskal Wallis test); only different treatments with the same genotype were compared (separated by grey-dashed lines). Scale bars in b to f, 5 μm.

Extended Data Figure 5. 2x and 3xPH^FAPP1 have longer residency time at the PM than 1xPH^FAPP1

High affinity lipid binding domains (LBDs) are expected to localize more specifically to the membrane compartment that accumulates the most its cognate lipid, while lower affinity LBDs are more likely to have a broader localization domain (a–c). Low affinity sensors (here 1xPH^FAPP1) are less efficient in discriminating between two membranes with two different concentrations of their targeted lipid species (here PI4P) and as a result they might be targeted to both of these membranes (a). By contrast, high affinity sensors (2xPH^FAPP1 and 3xPH^FAPP1) will have increased dwell time at the membrane that is the most enriched in the targeted lipid and they will accumulate preferentially in this compartment (b and c). In other words, high affinity sensors work like a “Velcro fastener”: they will grab more strongly to a surface with more spikes (in this case the spikes being PI4P). In order to confirm that our PH^FAPP1-based sensors behaves according to the scenario explained above, we performed a FRAP experiment (See main Fig. 3a to c). This analysis showed that the recovery was much faster in the case of 1xPH^FAPP1 and kymographic analysis showed that the recovery of fluorescence has an oval shape, indicating recovery from both the side (i.e. the PM) and the cytosol (Fig. 3a–c of the main text). This result is compatible with the idea that 1xPH^FAPP1 has a fast exchange rate between the PM and the cytosol. On the contrary, in the case of 2xPH^FAPP1 and 3xPH^FAPP1 the recovery was slower and kymographic analysis (Fig. 3b of the main text) showed that the recovery of fluorescence is centripetal (triangle shape). 1xPH^FAPP1 localizes at the PM and in endomembranes, while 2x and 3xPH^FAPP1 are not (or less) present in intracellular compartments. Therefore, it is conceivable that the fast recovery of the 1xPH^FAPP1 reporter might come from fast endocytic recycling that is not happening in the case of the 2x and 3xPH^FAPP1 proteins. To exclude this possibility, we tested whether pharmacological inhibition of protein recycling by BFA had any impact on the recovery time of the 1xPH^FAPP1 construct and (d–e). We found that cYFP-1xPH^FAPP1 had similar recovery time in the presence or absence of BFA (100 μM 60 min). These results are consistent with the notion that the 2xPH^FAPP1 and 3xPH^FAPP1 sensors have a longer residency time at the PM than 1xPH^FAPP1 and repopulate the bleached area by lateral diffusion with their cognate lipids.

Extended Data Figure 6. P4M^SidM is specifically localized to the PM in various cell types in Arabidopsis

Confocal pictures of a) the root and b) the shoot of transgenic Arabidopsis lines stably expressing cYFP-P4M^SidM. Scale bars, 10 μm.

Extended Data Figure 7. Full scan of lipid-protein overlay experiments and associated western blots

Full scan of lipid overlay assays presented in the main Figure 5a, and their associated western blots. Top left is shown the position of the different lipids spotted on each membrane: Lysophosphatidic acid (LPA), Lysophosphocholine (LPC), Phosphatidylinositol (PI), Phosphatidylinositol-3-phosphate (PI3P), Phosphatidylinositol-4-phosphate (P4P), Phosphatidylinositol-5-phosphate (PI5P), Phosphatidylethanolamine (PE), Phosphatidylcholine (PC), Sphingosine 1-Phosphate (S1P), Phosphatidylinositol-3,4-bisphosphate (PI(3,4)P₂), Phosphatidylinositol-3,5-bisphosphate (PI(3,5)P₂), Phosphatidylinositol-4,5-bisphosphate (PI(4,5)P₂), Phosphatidylinositol-3,4,5-trisphosphate (PI(3,4,5)P₃), Phosphatidic acid (PA), Phosphatidylserine (PS) and mock. Anionic phospholipids are indicated in red. Note that with this in vitro interaction technique, we systematically found a stronger signal with PS and to a lesser extent with PA. This was also the case for HA-KA1^MARK1 which is known to bind PS, PA and PI(4,5)P₂ with similar binding affinities in Surface Plasmon Resonance experiments. It is important to point out that these fat blots experiments are qualitative rather than quantitative. The main point of these experiments is to show that PID, BKI1 and MAKR1 to MAKR4 are indeed able to bind anionic phospholipids in vitro, and that this binding relies on their respective membrane hook (for PID and BKI1). We do not think that these experiments faithfully pin point particular lipid preferences. In fact, we expect PID, BKI1 and MAKR1 to MAKR4 to rely on membrane surface charges (non specific electrostatic interactions) in vivo and we therefore concentrated our experiments using in vivo assays (see yeast and in planta experiments).

Extended Data Figure 8. Using the cho1Δ yeast mutant to test the requirement of PM MSC for protein localization

a) In yeast, the PM is highly electronegative, mainly due to the presence of phosphatidylserine (PS), which massively accumulates in this membrane. Endomembrane compartments are of intermediate electronegativity, likely because of the marginal presence anionic lipids in these compartments (e.g. PI4P in the Golgi, PI3P in endosomes). b) The yeast cho1Δ mutant is impaired in PS biosynthesis and therefore lacks a strong PM electrostatic field^,. As a result of this loss of PM MSC, endomembranes become more electronegative than the PM in cho1Δ and cationic proteins relocalize to endomembranes at the expense of the PM. This situation is exemplified by the localization of GFP-C2^Lact (c and d), a PS biosensor, and GFP-KA1^MARK1 (e and f), a MSC reporter. c) GFP-C2^Lact is specifically localized at the PM in yeast confirming that the main pool of this lipid is in this membrane. d) On the contrary, GFP-C2^Lact is soluble in the cho1Δ. This soluble GFP-C2^Lact is a localization by default in the absence of PS to target this sensor to membranes. e) GFP-KA1^MARK1 is localized at the PM in yeast, confirming that the PM is highly electronegative in this system. f) GFP-KA1^MARK1 is sensitive to PS depletion and relocalizes to endomembranes in the cho1Δ, which become more electronegative than the PM in this mutant. Therefore, the cho1Δ mutant assay allows discriminating between proteins that are targeted to the PM by specific interactions with PS (e.g. C2^Lact) or by reading-out the PS-dependent PM MSC (e.g. KA1^MARK1). Proteins that specifically interact with PS are solubilized in cho1Δ, while MSC effector proteins are depleted from the PM and relocalize to endomembranes. g) Three representative images showing the localization of the indicated constructs in WT and cho1Δ yeast. h) Quantification of localization of the indicated construct in yeast. Cytoplasm = cytosol and/or endomembrane (n=300 cells). The localization quantification was performed using three categories according on the fluorescence expression pattern, “Plasma membrane”, “Cytoplasm” and “Plasma membrane and Cytoplasm”. Here, we took cytoplasm in a broad sense, including both soluble proteins (see for example localization of BKI1^8A-GFP in WT yeast or localization of the PS sensor GFP-C2^Lact in cho1Δ) but also proteins associated with endomembranes (see for example localization of GFP-KA1^MARK1 in cho1Δ). For each GFP-tagged proteins, three independent experiments were performed and the localization was recorded in 100 yeasts in each experiment (300 cells total). Note that PID-GFP has a dual localization in yeast at the PM and in endomembrane compartments. PID^MH-GFP is more specifically localized at the PM than full length PID-GFP, while PID^9Q-GFP is localized in endomembrane compartments but not at the PM. These results suggest that in yeast PID likely has two localization sequences, one PM targeting sequence that corresponds to PID^MH and a second, so far unknown sequence, that targets PID to endomembranes. The situation is likely similar in planta, since PID-cYFP has a dual PM and endosomal localization, while PID^9Q-cYFP localizes only to endosomes but not at the PM (see Fig. 5c and 6b of the main text).

Extended Data Figure 9. Localization of MAKR1 to MAKR4 and their respective C-terminal deletion in Arabidopsis root

a) Representative images of full length MAKR1 to MAKR4 localization in roots of stably transformed transgenic Arabidopsis lines. b) Representative images of MAKR1 to MAKR4 N-terminus localization in root of stably transformed transgenic Arabidopsis lines. Note that, similar to BKI1, all the MAKR proteins are localized to the PM and cytoplasm. Furthermore, in some cases they are also present in the nucleus (see for example MAKR3 or MAKR4). Nuclear localization has also been reported for GFP-MAKR4 but the functional significance of this localization is currently unknown. Scale bars, 10 μm.

Fig. 1. The plant PM and the cell plate are highly electronegative, a property that correlates with PI4P localization

a–o) Confocal pictures of MSC probes in Arabidopsis root epidermis. Probes are indicated at the bottom and net charges at the top. p) Tukey boxplot showing the distribution of intracellular compartments (spots) per cell for each MSC reporter. Different italicized-letters indicate significant differences among means (P<0.0001, Kruskal-Wallis test). q) Schematic representation of MSC organization in plants. r–s) Dual-color imaging during cytokinesis in Arabidopsis root epidermis. Plants co-expressing: r) 2xCyPet-1xPH^FAPP1 (top) and cYFP-2xPH^PLC (bottom) or s) 2xCHERRY-1xPH^FAPP1 (top) and cYFP-KA^MARK1 (bottom). Confocal images are color-coded with respect to pixel intensity based on the scale shown in the top right corner. Scale bars, 5 μm.

Fig. 2. PI4K activity is required to maintain the PM electrostatic signature

a) Schematic representation of the drugs used to perturb phosphoinositides production and lipid sensors used as read-out. b–i) Dual-color imaging of plants treated with the indicated time and drug concentration. PS, PI(4,5)P₂ and MSC reporters are pseudo-colored in green (left), 1xPH^FAPP1 are pseudo-colored in purple (middle). Colocalizations are showed in white in the merged channel (right). b–c) Plants co-expressing 2xCHERRY-C2^Lact and cYFP-1xPH^FAPP1. d–e and h–i) Plants co-expressing cYFP-2xPH^PLC and 2xCyPet-1xPH^FAPP1. f–g) Plants co-expressing cYFP-KA1^MARK1 and 2xCHERRY-1xPH^FAPP1. j–s) Confocal pictures of cYFP-KA1^MARK1 MSC reporter treated with the indicated time and drug concentration and t) corresponding dissociation index (mean ±SEM). Different italicized-letters indicate significant differences among means (P<0.0001, Kruskal-Wallis test).

Fig. 3. PI4P is a hallmark of the plant PM

a–c) FRAP analyses of 1x, 2x and 3xPH^FAPP1 sensors. a) Representative confocal pictures, b) kymograms of protein diffusion within the PM and c) trace of fluorescence intensity during FRAP analyses. d–h) Confocal pictures of PI4P probes in Arabidopsis root epidermis. Probes are indicated at the top. Scale bars, 5 μm. i) Tukey boxplot showing the distribution of intracellular compartments (spots) per cell for each PI4P reporter. Different italicized-letters indicate significant differences among means (P<0.0001, Kruskal-Wallis test). j–o) Confocal pictures of PI4P probes in Nicotiana benthamiana leaf epidermis. Probes are indicated at the top and mutations in PH^FAPP1 at the bottom. Bottom panels show schematic representations of PH^FAPP1 membrane recruitment mechanism according to the different mutations used. Orange arrowheads indicate endosomal localization of 1xPH^FAPP1. Scale bars, 20 μm.

Fig. 4. PM PI4P drives the electrostatic field of the cell membrane

a) Schematic representation of the genetic system used to specifically deplete PM PI4P. b–c) mTURQUOISE2 imaging of MAP-mTU2-SAC1^DEAD (b) and MAP-mTU2-SAC1 (c) in Nicotiana benthamiana leaf epidermis. d–m) cYFP imaging of the lipid or MSC reporter indicated at the top in Nicotiana benthamiana leaf epidermis, co-expressed with MAP-mTU2-SAC1^DEAD (d–h) or MAP-mTU2-SAC1 (i–m). n) Quantification of localization observed in d–m. o) Schematic representation of PI4P and MSC organization in non-perturbed cells (left) or cells with reduced PM PI4P (right). Scale bars, 20 μm.

Fig. 5. PINOID and BKI1/MAKRs are effectors of the PM electrostatic field

a) Lipid overlay assays with recombinant PID-Flag, PID^9Q-FLAG, BKI1-Flag, BKI1^8A-Flag, BKI1^Nter-Flag, MAKR1-Flag, HA-MAKR2, HA-MAKR3, HA-MAKR4 and the HA-KA1^MARK1 control. Anionic lipids are indicated in blue. b) Three representative confocal pictures showing the localization of the indicated GFP-fused protein in WT and cho1Δ yeast. c) Representative images in mock or PAO treated plants. Numbers at the bottom indicates the proportion of cells with signal at the PM or not. Scale bars, 5 μm.

Fig. 6. PM targeting by PID cationic membrane hook is required for function

a) Schematic representation of the PID protein. b) Confocal picture of pUBQ10::PID-cYFP and pUBQ10::PID^9Q–cYFP in Arabidopsis root meristem epidermis. Scale bars, 5μm. c–g) Representative picture of root hair phenotypes (left) and localization of the indicated construct (right). Each picture was taken with identical setting indicating that each transgenic line expressed comparable level of PID protein. Blue arrowheads indicate elongated root hairs and yellow arrows indicate root hairs with inhibited growth. Scale bars, 100 μm. h) Tukey boxplot showing the quantification of root hair length in the following lines: pEXP7::PID-cYFP (P-Y, orange); pEXP7::PID^9Q-cYFP (P^9Q-Y); pEXP7::PID^9Q-cYFP^5K3Q-Farn (P^9Q-Y^5K-F); pEXP7::2x-cYFP^8K-Farn (Y^8K-F) and WT. Different italicized-letters indicate significant differences among means (P<0.0001, Kruskal-Wallis test).