Phospholipase Cβ1 induces membrane tubulation and is involved in caveolae formation

Takehiko Inabaa,1, Takuma Kishimotoa,b,1, Motohide Muratea,1, Takuya Tajimaa,c,1, Shota Sakaia, Mitsuhiro Abea, Asami Makinoa, Nario Tomishigea, Reiko Ishitsukaa, Yasuo Ikedac, Shinji Takeokac, and Toshihide Kobayashia,d,2

aLipid Biology Laboratory, RIKEN, Saitama 351-0198, Japan; bDepartment of Biochemistry, Kyorin University School of Medicine, Mitaka, Tokyo 181-8611, Japan; cResearch Group of Biomolecular-Assembly, Department of Science and Medical Bioscience, Graduate School of Advanced Science and Engineering, Waseda University, Shinjuku-ku, Tokyo 162-8480, Japan; and dUMR 7213 CNRS, University of Strasbourg, 67401 Illkirch, France

Edited by Paul A. Janmey, University of Pennsylvania, Philadelphia, PA, and accepted by Editorial Board Member Edward D. Korn May 13, 2016 (received for review March 8, 2016) membrane curvature plays important roles in various physio- extract as a that induces the tubulation of phosphatidy- logical phenomena. Curvature-regulated dynamic membrane remod- linositol-4,5-bisphosphate (PIP2)-containing liposomes (9). eling is achieved by the interaction between and . So Using mouse brain extract, the present study identified phos- far, several membrane sensing/sculpting proteins, such as Bin/ pholipase Cβ1(PLCβ1), which induces tubulation of the phospha- amphiphysin/Rvs (BAR) proteins, are reported, but there remains tidylethanolamine (PE)- and phosphatidylserine (PS)-containing the possibility of the existence of unidentified membrane-deforming membranes. The results indicate that the characteristic C-ter- proteins that have not been uncovered by sequence . To minal sequence, but not the conserved inositol phospholipid- identify new lipid membrane deformation proteins, we applied binding pleckstrin homology (PH) domain or catalytic domain of liposome-based microscopic screening, using unbiased-darkfield mi- PLCβ1, is involved in the tubulation of liposomes. An in vitro croscopy. Using this method, we identified phospholipase Cβ1(PLCβ1) study suggests that sensing and/or modulation of the membrane as a new candidate. PLCβ1 is well characterized as an cata- curvature by the C-terminal domains is involved in the activation lyzing the hydrolysis of phosphatidylinositol-4,5-bisphosphate (PIP2). of PLCβ1. Knockdown of PLCβ1 in Swiss 3T3 cells resulted in a In addition to activity, our results indicate that PLCβ1 possessed deficiency of caveolae, indicating the importance of this protein the ability of membrane tubulation. Lipase domains and inositol in caveolae formation. BIOLOGY phospholipids binding the pleckstrin homology (PH) domain of PLCβ1 were not involved, but the C-terminal sequence was responsible for Results this tubulation activity. Computational modeling revealed that the C Phospholipase Cβ1 Induces Tubulation of - terminus displays the structural homology to the BAR domains, which Containing Membranes. We screened for a protein that induce is well known as a membrane sensing/sculpting domain. Overexpres- membrane deformation by incubating various tissue extracts with sion of PLCβ1 caused plasma membrane tubulation, whereas knock- giant unilamellar vesicles (GUVs) of defined lipid composition down of the protein reduced the number of caveolae and induced the under darkfield microscopy. We homogenized the brain, heart, and evagination of caveolin-rich membrane domains. Taken together, our liver of C57BL/6 mice, and then the supernatant (sup) fraction was β results suggest a new function of PLC 1: plasma membrane remodel- prepared. The sup was added to GUVs composed of phosphati- ing, and in particular, caveolae formation. dylcholine (PC)/PS (8:2) or PE/PC/PS (6:2:2). After 1 h incubation

phospholipase Cβ1 | membrane tubulation | microscopy screening | Significance caveolae | BAR-like domain

Lipid membrane curvature plays important roles in various he alteration of membrane curvature is crucial in various physiological phenomena. Using darkfield microscopy, we per- Tcellular events, such as cell division, membrane traffic, and formed nonbiased screening of a protein that induces deforma- migration. Membrane curvature is generated by the preferential tions of nonlabeled liposomes. We identified phospholipase Cβ1 binding of specific proteins to a curved membrane. The Bin/ (PLCβ1), which induces tubulation of the phosphatidylethanol- amphiphysin/Rvs (BAR) domain superfamily is a group of well- amine and phosphatidylserine-containing membranes. The char- – studied cytosolic proteins that causes membrane deformation (1 6). acteristic C-terminal sequence of PLCβ1, but not the conserved The BAR domains are crescent-shaped modules with different ra- inositol phospholipid-binding pleckstrin homology (PH) domain or dii. The binding of the BAR domain proteins forms and stabilizes catalytic domains of PLCβ1, is involved in the tubulation of li- membrane tubules of different diameters, depending on the cur- posomes. The C-terminal sequence is predicted to have the Bin/ vature of the domain. Recent structural and analyses amphiphysin/Rvs (BAR)-like conformation by computational have led to the identification of a number of proteins that belong to modeling. Our results indicate that sensing and modulation of different groups of the BAR family. However, there remains the the curvature by the C-terminal BAR-like domains is involved in possibility of the existence of unidentified membrane-deforming the activation of PLCβ1. The present results also reveal the role proteins that have not been uncovered by . of PLCβ1 in caveolae formation. In vitro, BAR proteins induce the tubulation of liposomes. In this study, we screened a protein that induces the tubulation of Author contributions: T.I., T. Kishimoto, M.M., T.T., R.I., and T. Kobayashi designed research; liposomes of defined lipid composition. The tubulation process T.I., T. Kishimoto, M.M., T.T., S.S., M.A., A.M., and N.T. performed research; T.I., M.M., M.A., and N.T. contributed new reagents/analytic tools; T.I., T. Kishimoto, M.M., T.T., S.S., A.M., and was often followed, using electron microscopy or fluorescent N.T. analyzed data; and T.I., T. Kishimoto, M.M., Y.I., S.T., and T. Kobayashi wrote the paper. – microscopy (1 3). However, the low throughput preparation of The authors declare no conflict of interest. samples in electron microscopy does not fit the screening pur- This article is a PNAS Direct Submission. P.A.J. is a guest editor invited by the Editorial pose, and the addition of even a trace amount of a fluorophore Board. can change the physical properties of liposomes. Darkfield mi- 1T.I., T. Kishimoto, M.M., and T.T. contributed equally to this work. croscopy allows real-time, in situ observation of low-contrast 2To whom correspondence should be addressed. Email: [email protected]. samples such as liposomes without labeling them (7, 8). This This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. method was previously used to identify septin from porcine brain 1073/pnas.1603513113/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1603513113 PNAS Early Edition | 1of6 Downloaded by guest on September 30, 2021 at room temperature, the specimens were observed under darkfield PC/PS (8:2) GUV containing 1 mol% PIP2. Similar to the results microscopy (Fig. 1A). Incubation of PE/PC/PS GUVs with mouse with GUVs in the absence of PIP2, tubulation was observed in a brain extract resulted in the tubulation of 25 ± 4% (mean ± SD; n = PE-dependent manner [24 ± 3% of PE/PC/PS/PIP2 (6:2:2:0.1); 3) of the liposomes (Fig. 1A). In contrast, only 1 ± 1% (n = 3; P < t test between PE/PC/PS with and without PIP2, P > 0.05]. 0.001) of the PC/PS liposomes exhibited tubulation, even in the PLCβ1b bound to PC/PS/PIP2 (8:2:0.1)MLVs(Fig.2C)and presence of brain extract. Extracts prepared from mouse heart and not to PC/PIP2 (10:0.1) MLVs, indicating that the presence of liver did not induce any membrane tubulation of either PC/PS or PIP2 is not sufficient for binding of PLCβ1b. We also examined PE/PC/PS GUVs. Mouse brain extract did not induce tubulation of the binding of PLCβ1b to egg PG or PS MLVs (Fig. 2C). The PC, phosphatidylglycerol (PG)/PC (1:1), phosphatidylinositol/PC protein bound both MLVs. However, PLCβ1b did not induce any (1:1), and cardiolipin/PC (1:1) GUVs (0 ± 0%). These results tubulation of PS and PG GUVs. We also showed that PLCβ1b indicate that mouse brain extract contains a factor or factors that bound PE/PS (1:1) MLVs, but not PE/PC (1:1) MLVs. In Fig. S1 are able to induce the tubulation of PE-containing membranes. we compared the binding of PLCβ1b to various lipids by ELISA. We postulated that the factor or factors bind the PE/PC/PS, but ELISA enables us to examine binding of the protein to nonbilayer not the PC/PS, membrane. To identify the protein factor or factors lipids such as diunsaturated dioleoyl-sn-glycero-3-phosphoethanol- responsible for the tubulation of the PE/PC/PS membrane, we in- amine (DOPE). Fig. S1 indicates that PLCβ1b strongly binds PE, cubated multilamellar vesicles (MLVs) composed of PC/PS (8:2) or PS, and PIP2, and PC works as an inhibitor in the binding. Our PE/PC/PS (6:2:2) with mouse brain extract, followed by pre- results indicate that the binding of the protein does not correlate to cipitation of liposomes. Proteins cosedimented with MLVs were tubulation and highlight the importance of PE in tubulation. analyzed by SDS/PAGE (Fig. 1B). Coomassie brilliant blue staining The effects of the fatty acid composition of PE on mem- showed two bands of proteins cosedimented with the PC/PS lipo- brane binding and the tubulation induced by PLCβ1b were then somes. These proteins were identified, using matrix assisted laser examined. In Fig. 1D, we used 1-palmitoyl-2-oleoyl-sn-glycero-3- desorption/ionization-time of flight (MALDI-TOF) mass spec- phosphoethanolamine (POPE), which has one saturated palmitic trometry (MS), as the heavy chain (220 kDa) and acid and one unsaturated oleic acid conjugated to a glycer- (100 kDa), respectively. In addition to these two bands, 150 kDa ophosphoethanolamine backbone. In Fig. 2 D and E,weexamined protein was precipitated when the brain extract was incubated with the binding of PLCβ1b and the protein-induced tubulation of PE/PC/PS. This protein was identified as PLCβ1. DOPE/PC/PS (6:2:2) and disaturated, dipalmiroyl-sn-glycero-3- PLCβ1 is highly expressed in the cerebral cortex, hippocampus phosphoethanolamine (DPPE)/PC/PS (6:2:2) liposomes, in addition (10), and cardiovascular system (11, 12). PLC is an enzyme that to POPE/PC/PS. GUVs were used to measure tubulation, whereas hydrolyzes PIP2 to inositol-1,4,5-triphosphate (IP3) and diac- binding was measured using MLVs. PLCβ1b bound all of the 2+ ylglycerol (DAG). IP3 releases Ca from endoplasmic reticulum MLVs, although DOPE increased and DPPE slightly decreased the through the specific receptor, whereas DAG activates protein binding. However, tubulation was not observed in DPPE/PC/PS C (13, 14). PLC isozymes are classified into six subfamilies GUVs, further indicating that protein binding is not enough for the (PLC-β, PLC-δ, PLC-γ, PLC-e, PLC-ζ, and PLC-η). Mammalian tubulation. The difference of PE acyl chain might affect the mem- PLCβ1 exists in two isoforms expressed by alternative splicing, brane fluidity. The gel-to-liquid crystalline phase transition tem- PLCβ1a and PLCβ1b (15, 16), differing in their C-terminal se- perature Tc is DPPE, 64 °C; DOPE, −8 °C; and POPE, 25 °C. In quences (the C-terminal regions after 1,024, hereafter Fig. 2D, we also measured the tubulation of dilauroyl phosphati- Ctail). Fig. 1C indicates that the recombinant PLCβ1a and dylethanolamine (DLPE) (Tc = 31 °C)-containing GUVs. DLPE/ PLCβ1b bound PE/PC/PS (6:2:2), but not PC/PS (8:2) MLVs. PC/PS GUVs did not tubulate, despite similar Tc of DLPE and Recombinant PLCβ1a and PLCβ1b induced tubulation of 14 ± POPE. These results indicate that membrane fluidity is not essential 2% and 28.0 ± 2.0% of PE/PC/PS GUVs, respectively (Fig. 1D for PLCβ1b-induced membrane tubulation. The reported bending − and Movie S1). These results indicate that PLCβ1 induces rigidity of DLPE is 1.7 × 10 19 J (17), whereas that of DOPE is 0.94 × − tubulation of PE-containing membranes. Because PLCβ1b in- 10 19 J (18), consistent with the ability of DOPE to tubulate duced the tubulation more efficiently than PLCβ1a, we used membrane. However, differences in lipid headgroup do not appear mainly PLCβ1b for further analysis. to have a large influence on the bending modulus (19). Thus, Fig. 2A shows a negative staining electron micrograph of PE/PC/ bending rigidity does not fully explain the different effect of PE PS (6:2:2) liposomes incubated with PLCβ1b. The width of tubules from that of other lipids on membrane tubulation. + was 10–40 nm. Immunoelectron microscopy revealed that the pro- The membrane binding of Ca2 -binding C2 domains of PLCδ tein bound on the tubules and buds formed on the liposome isoforms has been reported (20). Fig. 2 F and G examined the effect membrane (Fig. 2B, arrowheads). of the addition of 1 mM CaCl2 on PLCβ1b binding and the tubu- We then examined the effect of the addition of the lation of PE/PC/PS (6:2:2) liposomes. CaCl2 reduced the binding of of PLCβ1, PIP2.PLCβ1b was added to PE/PC/PS (6:2:2) or PLCβ1b and tubulation (16 ± 3%; t test compared between with

Fig. 1. PLCβ1 induces tubulation of PE-containing membranes. (A) Darkfield image of PE/PC/PS (6:2:2) and PC/PS (8:2) GUVs in the presence and absence of mouse brain extract. The arrowheads indicate tu- bules. (Scale bar, 5 μm.) (B) SDS/PAGE analysis of brain extract proteins cosedimented with PE/PC/PS (6:2:2) and PC/PS (8:2) MLVs. Precipitated proteins were identified. *Myosin heavy chain. **Phospholi- pase Cβ1. ***Dynamin. (C) Binding of PLCβ1a and PLCβ1b to PE/PC/PS (6:2:2) and PC/PS (8:2) MLVs by floatation assay. (D) Darkfield image of PE/PC/PS (6:2:2) GUVs incubated with PLCβ1a and PLCβ1b. The arrowheads indicate tubules. (Scale bar, 5 μm.)

2of6 | www.pnas.org/cgi/doi/10.1073/pnas.1603513113 Inaba et al. Downloaded by guest on September 30, 2021 of PE/PC/PS (6:2:2) GUVs were tubulated when 1,840 nM PLCβ1b was applied. Less efficient but significantly positive (14 ± 4%) tubulation of PE/PC/PS GUVs was observed at 1,840 nM PLCβ1b in Tris-buffered saline (TBS) at pH 7.5 as a physiolog- ical ionic strength buffer (Fig. 2H). What is the role of PE on the tubulation of liposome? PE has a small headgroup and tends to form the negative curvature membrane. This negative curvature is enhanced by the unsatura- tion of the acyl chain. During the tubulation of liposomes, the negative curvature emerges at the outer membrane leaflet of bud necks and the inner membrane leaflet of tubes. The manipulation of the liposome by optical tweezers has shown that the maximum force is required at the budding of tubes from the flat membrane and lower, and the constant force is required for subsequent tube elongation (24). There is an energy barrier to bend the flat mem- brane to the curved shape, and the redistribution of lipids reduces the energy by its spontaneous curvature. In fact, PE is associated on the edge of caveolae (25), and the addition of PE to GUV increased tubulation efficiency (26).

The C-Terminal Domain of PLCβ1 Is Responsible for Tubulation of Liposomes. Similar to other PLCs, PLCβ contains a catalytic core composed of an N-terminal PH domain, EF-hand motif, and the active sites X and Y, followed by a C2 domain (27). In addition to the catalytic X and Y domains, PLCβ is characterized by the presence of an extended C terminus of ∼400 amino acids (Fig. 3A). This C-terminal extension contains a highly conserved

Gαq (28) and an elongated ∼300-amino acid coiled–coil CELL BIOLOGY domain (27). Whereas the PLCδ PH domain binds PIP2 with high specificity and affinity, the binding of the PLCβ PH domain to the lipid is weak (29, 30). In contrast, the C-terminal extension is believed to be the primary membrane binding determinant in PLCβ (31–33). In Fig. 3 B and C, we show the effects of various C-terminal fragments of PLCβ1b on the binding to PE/PC/PS (6:2:2) MLVs and the tubulation of PE/PC/PS GUVs. The C2-Gαq- Ctail and Gαq-Ctail fragments selectively bound to PE/PC/PS α Fig. 2. Effects of lipid composition and calcium ion on binding and tubulation MLVs, whereas C2-G q bound both PE/PS/PC and PS/PC. In of liposomes induced by recombinant PLCβ1b. (A) Negative staining electron micrograph of PE/PC/PS (6:2:2) liposome incubated with PLCβ1b. (Scale bar, 500 nm.) (B) Immunoelectron micrograph of PE/PC/PS (6:2:2) liposome incubated with PLCβ1b. (Scale bar, 100 nm). The arrowheads indicate gold particles. (In- set) Enlarged image (dashed rectangle). (C) Binding of recombinant PLCβ1b

to PC/PS/PIP2 (8:2:0.1), PC/PS (8:2), PC/PIP2 (10:0.1), PE/PC (1:1), PE/PS (1:1), PG, and PS liposomes by floatation assay. (D) Effects of different PE on tubulation efficiency of PE/PC/PS (6:2:2) GUVs by recombinant PLCβ1b (mean ± SD). (E) Effects of different PE on binding of recombinant PLCβ1b to PE/PC/PS (6:2:2) MLVs by floatation assay. (F) Tubulation efficiency of PE/PC/PS (6:2:2) + liposomes by PLCβ1b in the absence (control) and presence (Ca2 )of1mM

CaCl2, followed by the addition of 10 mM EDTA (mean ± SD). (G) Binding of + PLCβ1b to PE/PC/PS (6:2:2) in the absence (control) and presence (Ca2 )of

1 mM CaCl2 followed by the addition of 10 mM EGTA by floatation assay. (H) The effect of buffer on tubulation efficiency. PLCβ1b concentrations are 200 nM in buffer A at pH 8.5 and 1,840 nM in buffer A and TBS at pH 7.5.

and without CaCl2, P < 0.01). The decrease of the tubulation may be a result of the decrease of binding of the protein. The binding of PLCβ1 and tubulation (22 ± 4%; t test compared between with EDTA and without CaCl2, P > 0.05) were recovered by the addition + of Ca2 chelating reagents. The surface charge of lipid head groups was affected by the basic pH (21, 22). PIP2 ionization and domain formation were reported in the presence of lipids with hydrogen bond donor capabilities (21, 23). We also analyzed the effect of neutral pH on PLCβ1b-induced tubulation. When buffer A was adjusted to Fig. 3. C-terminal fragments of PLCβ1b induces tubulation of PE-containing β pH 7.5, PE/PC/PS GUVs stuck to the glass surface. The glass liposomes. (A) Domain structures of PLC 1b. (B) Binding of fragments of PLCβ1b to PE/PC/PS (6:2:2) and PC/PS (8:2) MLVs by floatation assay. surface was siliconized to reduce nonspecific binding of GUVs. (C) Tubulation efficiency of liposomes induced by the fragments of PLCβ1b β Under these conditions, 200 nM PLC 1b (our standard condi- (mean ± SD). (D) Distribution of mKate-C2-Gαq-Ctail of PLCβ1b on PE/PC/PS tion) was not enough to induce tubulation at pH 7.5, and 33 ± 10% (6:2:2) GUV. (Scale bar, 5 μm.) The arrowhead indicates a tubule.

Inaba et al. PNAS Early Edition | 3of6 Downloaded by guest on September 30, 2021 contrast, C2 alone did not bind MLVs, whereas Gαq weakly bound PE/PC/PS MLVs. Although both C2-Gαq-Ctail and Gαq-Ctail spe- cifically bound PE/PC/PS MLVs, only C2-Gαq-Ctail efficiently in- duced tubulation of PE/PC/PS GUVs (Fig. 3C). Three other mutant proteins containing Gαq slightly tubulated PE/PC/PS GUVs, whereas C2 domain alone did not exhibit tubulation. Far-red fluo- rescent protein mKate-conjugated C2-Gαq-Ctail demonstrated the concentration of the protein to the tubules (Fig. 3D and Fig. S2). Fig. S2 shows that even in the spherical liposomes, mKate fluores- cence was accumulated to one or several locations on the mem- brane surface, consistent with the observation in Fig. 2B. The mean maximum intensity of mKate-C2-Gαq-Ctail was 4,577 ± 1,640 on thetube(n = 6) and 3,119 ± 2,874 on the spherical part of tubulated liposomes (n = 6), and 2,209 ± 1,339 on the smooth membrane of spherical liposomes (n = 12). This suggests that the C2-Gαq-Ctail fragment accumulates and stabilizes the tubules shape. Lyon et al. reported that C-terminal domains of PLCβ3 have the BAR domain-like structure (27). From the homology model- ing (SWISS-MODEL) based on their PLCβ3structure,C2-Gαq- Ctail domains of PLCβ1 are revealed to form coiled coils and similar topology to BAR structure with positively charged clusters (Fig. S3). Homology modeling suggests that C2-Gαq-Ctail of PLCβ1 is the BAR-like domain and generates/stabilizes the tubulation by a BAR modulating mechanism.

Overexpression of PLCβ1b and the C-Terminal Fragment Induce Fig. 4. Overexpression of PLCβ1b and C-terminal fragments induces the plasma Plasma Membrane Tubulation in Swiss 3T3 Cells. Overexpression of membrane tubulations in Swiss 3T3 cells. (A) Fluorescence images of Swiss 3T3 cells BAR proteins (1) and fps/fes related-Cdc42-interacting protein 4 overexpressing mKate-tagged full-length PLCβ1b, H331A/H378A mutant, and C2- α Δ α (FER-CIP4) homology-BAR (F-BAR) proteins (2) in cultured G q-Ctail-deficient mutant [ (C2-G q-Ctail)] under the Tet-inducible system are β cells has been reported to induce plasma membrane tubulation. shown. (B) Quantification of the localization of mKate-tagged PLC 1b or those of We used tetracycline-regulated expression system to highly express its mutants under the Tet-inducible expression in Swiss 3T3 cells. (Lower)Repre- β sentative patterns of the plasma membrane (red), cellular internal compartment mKate-PLC 1b and its derivatives in Swiss 3T3 cells. When cells (yellow), cytosol (green), and nucleus (blue). Results are the means ± SD of four were treated with doxycycline (Dox) to trigger tetracycline (Tet)- independent experiments in which more than 100 cells were counted. (Scale bars, inducible expression systems, mKate-PLCβ1b induced plasma 20 μm.) (C) The effect of Tet-inducible overexpression of mKate-tagged PLCβ1b membrane tubulation (Fig. 4A, Left). In the absence of Dox, and its mutants on membrane tubulation in Swiss 3T3 cells. Results were scored as mKate-PLCβ1b was observed as the leaky expression, and located a percentage of the number of the transfected cells. Means ± SD of four in- to the plasma membrane but did not induce tubulation (Fig. S4A). dependent experiments in which more than 100 cells were counted. Likewise, mKate-PLCβ1b expressed by the plasmid containing the constitutive CMV promoter also did not induce tubulation (Fig. β composed of PE/PC/PS/PIP2 (6:2:2:1) (Fig. 5). The sizes and S4B). Similar to full-length PLC 1b, the mKate conjugate of the – phospholipase-dead mutant H331A/H378A (34) induced plasma lamellarity of liposomes were examined using freeze fracture elec- tron microscopy (Fig. S6). The mean diameter of freeze-thawed membrane tubulation, indicating that phospholipase activity is not ± = A liposome was 484 200 nm (n 23), and that of 100 nm pore- required for the induction of membrane tubulation (Fig. 4 , ± = Center). In contrast, the mKate conjugate of a PLCβ1b mutant extruded liposomes was 102 21 nm (n 100). No multilamellar liposomes were observed in 100-nm liposomes, and fracture images defective in C2-Gαq-Ctail did not result in localization to the plasma membrane or membrane tubulation (Fig. 4A, Right). in 4% of freeze-thawed liposomes showed membrane fragments Fig. 4B summarizes the localization of the mKate conjugate of derived from other lamellae, indicating multilamellar vesicles. To full-length PLCβ1b and its derivatives. Both full-length PLCβ1b and compare the lipase activity on these liposomes, the DAG pro- phospholipase-dead H331A/H378A mutants were exclusively lo- duction was quantified by MS. Both proteins showed higher activity on 100-nm liposomes than freeze-thawed liposomes, suggesting cated at the plasma membrane, whereas C2-Gαq-Ctail-defective mutants were distributed in cellular internal compartments and the high curvature efficiently activates phospholipase activity of PLCβ1. cytosol. Protein fragments that exhibited binding to the PE/PC/PS β β membrane (C2-Gα -Ctail, C2-Gα ,Gα -Ctail, and Gα )(Fig.3B) Knockdown of PLC 1 Alters the Morphology of Caveolae. PLC 1b q q q q α β were localized to the plasma membrane. In contrast, the C2 and and subunit G q, which directly activates PLC , are Ctail domains were either in cellular internal compartments or the reported to localize to the caveolar fraction (37). Previously, we cytosol. Fig. 4C indicates the percentage of cells that exhibited showed that PE and PIP2 were accumulated at the cytoplasmic plasma membrane tubulation. Consistent with the model mem- leaflet of the edge of caveolar membranes (25). To address the role β brane study, C2-Gα -Ctail efficiently induced plasma membrane of PLC 1 in caveolae formation, we compared the ultrastructures of q β tubulation (Fig. S5A). The other fragments did not significantly theplasmamembraneincontrolandPLC 1 knockdown Swiss 3T3 induce tubulation. cells. We used two sequences of small-interfering RNA (siRNA) to knockdown the expression of PLCβ1. Both siRNAs reduced the Liposome Size Affects the Activity of PLCβ1. It has been reported amount of PLCβ1to∼20% of control (Fig. S7). The number of that membrane curvature affects the binding and insertion of var- caveolae was dramatically decreased in the PLCβ1knockdowncells ious proteins (35). Ahyayauch et al. (36) reported that PLC from (Fig. 6A, black arrowheads). In addition, small electron-lucent Bacillus cereus has a higher level of activity in smaller liposomes. evaginations were often protruded from the plasma mem- To gain insight into the physiological role in the alteration of brane of PLCβ1 knockdown cells (Fig. 6A, red arrowhead). membrane curvature by PLCβ1, we compared the activity of To demonstrate the cellular distribution of caveolae in PLCβ1 PLCβ1a and PLCβ1b on freeze-thawed and extruded liposomes knockdown cell plasma membrane in relation to the localization of

4of6 | www.pnas.org/cgi/doi/10.1073/pnas.1603513113 Inaba et al. Downloaded by guest on September 30, 2021 the high curvature of the membrane further recruits the protein. Membrane has to be provided for elongation of tubules; however, if tubulation starts in many positions of the membrane, it is difficult to provide the membrane for the development of tubules. This might occur when protein concentration is very high. It is possible that distribution of lipids among GUVs is inhomogeneous, and tubu- lated liposomes and sphere liposomes have altered lipid composi- tion. Moreover, tubule part and spherical part of the liposomes may contain different lipid composition. Recent development in probes that bind specific lipids (38) will be used to further examine detailed lipid distribution during tubulation. PLCβ1b is enriched in the nuclear fraction of erythroleukemia cells (39). However, visible fluorescent protein conjugates of PLCβ1a and PLCβ1b expressed in HeLa cells were localized at the plasma membrane (40). In our hands, mKate-PLCβ1b, as well as Fig. 5. Liposome size affects the activity of PLCβ1. The amount of produced the mKate-conjugates of the C-terminal fragments of PLCβ1b DAG was analyzed by MS with PLCβ1 and no treatment (non). Total DAG is expressed in Swiss 3T3 cells, were located at the plasma membrane the sum of each molecular species. Data were expressed as the absolute (Fig. 4). Interestingly, PLCβ1b is enriched in the caveolae of car- amount of DAG (mean ± SD). Significant differences (P < 0.05) were ob- diomyocytes (41). A recent freeze–fracture immunoelectron mi- served between a and b, a and c, and b and c. croscope study reported an association of both PIP2 andPEtothe edge of the caveolae at the plasma membrane (25). It is well α caveolin, SDS-freeze fracture replica labeling was performed. The established that the G q-subunit activates PLCβ-lipase (42, 43). Recent results indicate that membranes are densities of caveolae in caveolae-rich membrane regions were de- β α creased in knockdown cells [13.0 ± 3.3/μm2 in control cells (n = 16) essential for the activation of PLC isozymes by G q-subunit (44). vs 4.8 ± 3.9/μm2 in knockdown cells (n = 18); P < 0.01]. The surface Previously, Escribá et al. showed that PE enhances the binding of α areas of caveolae in the plasma membranes were also decreased in G -subunit to model membranes (45). FRET experiments using 2 X-rhodamine isothiocyanate (XRITC)-Gα -subunit, XRITC-Gβγ- knockdown cells [7,510 ± 1,100 nm in control cells (n = 338) vs. q CELL BIOLOGY α 5,060 ± 580 nm2 in knockdown cells (n = 233; P < 0.01)]. Immu- subunit, and caveolin-1 (Cav1)-EGFP showed that G q-subunit has βγ nogold-labeling of caveolin in the replicated plasma membrane of a stronger affinity for Cav1-EGFP than G -subunit has (46). Our PLCβ1 knockdown cells demonstrated the concentrated localiza- results, together with results of others, suggest that PE provides a β tions of caveolin in the cytoplasmic leaflets not only in the neck of platform for PLC 1 activation by sensing curved membranes, such caveolar invaginations (black arrowheads) but also around evagi- as caveolae, via the C-terminal domains. In addition to caveolin nations (Fig. 6B, red arrowheads). These results indicate the im- (47), cavin (48) and pacsin2 (49, 50) are reported to be involved in portance of PLCβ1 in caveolae invagination. the formation of caveolae. In particular, pacsin2 localizes at and stabilizes the caveolae neck, and the removal of pacsin2 is coupled Discussion In the present study, we show that the screening using darkfield video microscopy, which previously identified the membrane deformation activity of septin (9), has the potential capacity to identify a novel protein that has no sequence similarity to pre- viously reported proteins that induce membrane deformation. However, only the relatively abundant proteins in the fraction can be identified in this method. Using this procedure, we identified PLCβ1 from mouse brain extract as a protein that tubulates PE-containing liposomes. The PLCβ family shares extra C-terminal domains that are structurally similar to the BAR domain (27). BAR domain proteins are reported to induce membrane tubulation (1–6). Our results indicate that, indeed, the C2-Gαq-Ctail fragment of PLCβ1b is predicted to have BAR-like conformation and alters membrane curvature in a PE-dependent manner (Fig. 3). This observation and the following findings indicate that membrane tubulation is not the result of phospholipid hydrolysis: PE, PS, and PC, used to prepare lipo- somes, are not the substrates of PLCβ, and a phospholipase-null mutant also induces tubulation of PE/PC/PS liposomes. In our GUV study, we used 100 μM lipids and 100 nM PLCβ1b. Esti- mating the diameter of GUV as 5 μm and considering the surface area of phospholipid as 50 A2, one can calculate that roughly 3 × 105 proteins interact with a liposome. Under this condition, PLCβ1b Fig. 6. Knockdown of PLCβ1 alters caveolae. (A) Transmission electron mi- induced tubulation of 28.0 ± 2.0% of PE/PC/PS GUV. In the initial croscopic images of ultrastructural observation of negative control (NC) and experiments, we increased the protein concentration. However, siRNA-treated knockdown (KD) Swiss 3T3 cells. KD cells have smaller numbers of caveolae (black arrowheads) than NC cells and electron-lucent evagination tubulation efficiency was not increased. Excess proteins were rather – inhibitory. Fig. 3D and Fig. S2 indicate that PLCβ1b preferentially (red arrowhead) from the plasma membrane. (Scale bar, 500 nm.) (B) Freeze fracture electron microscopic images of caveolar region in NC and KD cell binds tubules compared with the spherical part of liposomes. It is plasma membranes. Gold particles showing the distribution of caveolin were also noteworthy that in most liposomes, the numbers of tubules observed not only at the edge of caveolae (black arrowheads) but also in were one or, at most, a few. We speculate that the tubulation en- evaginations (red arrowheads). (Inset) Enlarged image of caveolae and an hances the cooperative binding of PLCβ1. Once a tubule is formed, evagination (dashed rectangle). (Scale bars, 500 nm; Inset, 100 nm.)

Inaba et al. PNAS Early Edition | 5of6 Downloaded by guest on September 30, 2021 to the flattening of plasma membrane (51). Our results highlight and Mami Kishimoto, respectively, for encouragement during this work. the importance of PLCβ1 in caveolae assembly. T.I. and T. Kishimoto were supported by the Special Postdoctoral Fellows program of RIKEN. Mouse Swiss Albino embryo fibroblast Swiss 3T3 cells were Materials and Methods provided by the RIKEN BRC through the National Bio-Resource Project of the Ministry of Education, Culture, Sports, Science and Technology, Japan. This work Detailed materials and methods are provided in SI Materials and Methods. ’ The experiments using mice followed the RIKEN Regulations for the Animal was supported by Integrated Lipidology Program of RIKEN, RIKEN President s “ ” Experiments and approved by the institutional review board of RIKEN. Fund 4D Measurements for Multilayered Cellular Dynamics, Grant-in Aid for Scientific Research 23790115 (to A.M.), 23590251 and 15K08167 (to M.M.), ACKNOWLEDGMENTS. We thank Prof. Michael Kozlov of Tel Aviv University 24770135 (to T. Kishimoto), and 22390018 and 25293015 (to T. Kobayashi) and Prof. Hiroshi Takahashi of Gunma University for their valuable from the Ministry of Education, Culture, Sports, Science and Technology of Japan comments. T.I. and T. Kishimoto are especially grateful to Ayumi Inaba and Naito Foundation.

1. Peter BJ, et al. (2004) BAR domains as sensors of membrane curvature: The amphi- lipid bilayers and the beta gamma subunits of heterotrimeric G proteins. Biochemistry physin BAR structure. Science 303(5657):495–499. 38(5):1517–1524. 2. Itoh T, et al. (2005) Dynamin and the actin cytoskeleton cooperatively regulate plasma 31. Kim CG, Park D, Rhee SG (1996) The role of carboxyl-terminal basic amino acids in membrane invagination by BAR and F-BAR proteins. Dev Cell 9(6):791–804. Gqalpha-dependent activation, particulate association, and nuclear localization of 3. Tsujita K, et al. (2006) Coordination between the actin cytoskeleton and membrane -beta1. J Biol Chem 271(35):21187–21192. deformation by a novel membrane tubulation domain of PCH proteins is involved in 32. Jenco JM, Becker KP, Morris AJ (1997) Membrane-binding properties of phospholi- endocytosis. J Cell Biol 172(2):269–279. pase C-beta1 and phospholipaseC-beta2: Role of the C-terminus and effects of pol- 4. Dawson JC, Legg JA, Machesky LM (2006) Bar domain proteins: A role in tubulation, scission yphosphoinositides, G-proteins and Ca2+. Biochem J 327(Pt 2):431–437. and actin assembly in clathrin-mediated endocytosis. Trends Cell Biol 16(10):493–498. 33. Lyon AM, Dutta S, Boguth CA, Skiniotis G, Tesmer JJ (2013) Full-length Gα(q)-phos- 5. Qualmann B, Koch D, Kessels MM (2011) Let’s go bananas: Revisiting the endocytic pholipase C-β3 structure reveals interfaces of the C-terminal coiled-coil domain. Nat BAR code. EMBO J 30(17):3501–3515. Struct Mol Biol 20(3):355–362. 6. Rao Y, Haucke V (2011) Membrane shaping by the Bin/amphiphysin/Rvs (BAR) domain 34. Ramazzotti G, et al. (2008) Catalytic activity of nuclear PLC-beta(1) is required for its . Cell Mol Life Sci 68(24):3983–3993. signalling function during C2C12 differentiation. Cell Signal 20(11):2013–2021. 7. Horio T, Hotani H (1986) Visualization of the dynamic instability of individual mi- 35. Epand RM, D’Souza K, Berno B, Schlame M (2015) Membrane curvature modulation – crotubules by dark-field microscopy. Nature 321(6070):605 607. of protein activity determined by NMR. Biochim Biophys Acta 1848(1 Pt B):220–228. 8. Takiguchi K, et al. (2002) Liposomes possess drastic capabilities for topological 36. Ahyayauch H, Villar AV, Alonso A, Goñi FM (2005) Modulation of PI-specific phos- – transformation. ChemPhysChem 3(7):571 574. pholipase C by membrane curvature and molecular order. Biochemistry 44(34): 9. Tanaka-Takiguchi Y, Kinoshita M, Takiguchi K (2009) Septin-mediated uniform 11592–11600. – bracing of phospholipid membranes. Curr Biol 19(2):140 145. 37. Lanzafame AA, et al. (2006) Inositol phospholipids localized to caveolae in rat heart 10. Homma Y, Takenawa T, Emori Y, Sorimachi H, Suzuki K (1989) Tissue- and cell type- are regulated by alpha1-adrenergic receptors and by ischemia-reperfusion. Am J specific expression of mRNAs for four types of inositol phospholipid-specific phos- Physiol Heart Circ Physiol 290(5):H2059–H2065. – pholipase C. Biochem Biophys Res Commun 164(1):406 412. 38. Kishimoto T, Ishitsuka R, Kobayashi T (2016) Detectors for evaluating the cellular landscape 11. Ushio-Fukai M, Griendling KK, Akers M, Lyons PR, Alexander RW (1998) Temporal of sphingomyelin- and -rich membrane domains. Biochim Biophys Acta, 10.1016/ dispersion of activation of phospholipase C-beta1 and -gamma isoforms by angio- j.bbalip.2016.03.013. tensin II in vascular smooth muscle cells. Role of alphaq/11, alpha12, and beta gamma 39. Fiume R, et al. (2005) Nuclear phospholipase C beta1 (PLCbeta1) affects CD24 ex- G protein subunits. J Biol Chem 273(31):19772–19777. pression in murine erythroleukemia cells. J Biol Chem 280(25):24221–24226. 12. Arthur JF, Matkovich SJ, Mitchell CJ, Biden TJ, Woodcock EA (2001) Evidence for se- 40. Adjobo-Hermans MJ, et al. (2013) PLCβ isoforms differ in their subcellular location and lective coupling of alpha 1-adrenergic receptors to phospholipase C-beta 1 in rat their CT-domain dependent interaction with Gαq. Cell Signal 25(1):255–263. neonatal cardiomyocytes. J Biol Chem 276(40):37341–37346. 41. Grubb DR, Vasilevski O, Huynh H, Woodcock EA (2008) The extreme C-terminal region 13. Berridge MJ (1987) Inositol trisphosphate and diacylglycerol: Two interacting second of phospholipase Cbeta1 determines subcellular localization and function; the “b” messengers. Annu Rev Biochem 56:159–193. splice variant mediates alpha1-adrenergic receptor responses in cardiomyocytes. 14. Nishizuka Y (1992) Intracellular signaling by hydrolysis of phospholipids and activa- FASEB J 22(8):2768–2774. tion of C. Science 258(5082):607–614. 42. Gresset A, Sondek J, Harden TK (2012) The phospholipase C isozymes and their reg- 15. Bahk YY, et al. (1994) Two forms of phospholipase C-beta 1 generated by alternative ulation. Subcell Biochem 58:61–94. splicing. J Biol Chem 269(11):8240–8245. 43. Lyon AM, Taylor VG, Tesmer JJ (2014) Strike a pose: Gαq complexes at the membrane. 16. Caricasole A, Sala C, Roncarati R, Formenti E, Terstappen GC (2000) Cloning and Trends Pharmacol Sci 35(1):23–30. characterization of the human phosphoinositide-specific phospholipase C-beta 1 (PLC 44. Charpentier TH, et al. (2014) Membrane-induced allosteric control of phospholipase beta 1). Biochim Biophys Acta 1517(1):63–72. C-β isozymes. J Biol Chem 289(43):29545–29557. 17. Mutz M, Helfrich W (1990) Bending rigidities of some biological model membranes as 45. Escribá PV, et al. (1997) Role of lipid polymorphism in G protein-membrane interac- obtained from the Fourier analysis of contour sections. J Phys France 51:991–1002. tions: Nonlamellar-prone phospholipids and peripheral protein binding to mem- 18. Chen Z, Rand RP (1997) The influence of cholesterol on phospholipid membrane branes. Proc Natl Acad Sci USA 94(21):11375–11380. curvature and bending elasticity. Biophys J 73(1):267–276. 46. Sengupta P, Philip F, Scarlata S (2008) Caveolin-1 alters Ca(2+) signal duration 19. Marsh D (2006) Elastic curvature constants of lipid monolayers and bilayers. Chem Phys Lipids 144(2):146–159. through specific interaction with the G alpha q family of G proteins. J Cell Sci – 20. Ananthanarayanan B, Das S, Rhee SG, Murray D, Cho W (2002) Membrane targeting 121(Pt 9):1363 1372. of C2 domains of phospholipase C-delta isoforms. J Biol Chem 277(5):3568–3575. 47. Parton RG, Simons K (2007) The multiple faces of caveolae. Nat Rev Mol Cell Biol 8(3): – 21. Graber ZT, Jiang Z, Gericke A, Kooijman EE (2012) Phosphatidylinositol-4,5-bisphosphate 185 194. ionization and domain formation in the presence of lipids with hydrogen bond donor 48. Kovtun O, Tillu VA, Ariotti N, Parton RG, Collins BM (2015) Cavin family proteins and – capabilities. Chem Phys Lipids 165(6):696–704. the assembly of caveolae. J Cell Sci 128(7):1269 1278. 22. Tsui FC, Ojcius DM, Hubbell WL (1986) The intrinsic pKa values for phosphatidylserine and 49. Hansen CG, Howard G, Nichols BJ (2011) Pacsin 2 is recruited to caveolae and func- – phosphatidylethanolamine in phosphatidylcholine host bilayers. Biophys J 49(2):459–468. tions in caveolar biogenesis. J Cell Sci 124(Pt 16):2777 2785. 23. Kooijman EE, King KE, Gangoda M, Gericke A (2009) Ionization properties of phospha- 50. Senju Y, Itoh Y, Takano K, Hamada S, Suetsugu S (2011) Essential role of PACSIN2/ – tidylinositol polyphosphates in mixed model membranes. Biochemistry 48(40):9360–9371. syndapin-II in caveolae membrane sculpting. J Cell Sci 124(Pt 12):2032 2040. 24. Inaba T, et al. (2005) Formation and maintenance of tubular membrane projections 51. Senju Y, Suetsugu S (2015) Possible regulation of caveolar endocytosis and flattening require mechanical force, but their elongation and shortening do not require addi- by phosphorylation of F-BAR domain protein PACSIN2/Syndapin II. BioArchitecture tional force. J Mol Biol 348(2):325–333. 5(5-6):70–77. 25. Murate M, et al. (2015) Transbilayer distribution of lipids at nano scale. J Cell Sci 52. Bhat HB, et al. (2013) Binding of a pleurotolysin ortholog from Pleurotus eryngii to 128(8):1627–1638. sphingomyelin and cholesterol-rich membrane domains. J Lipid Res 54(10):2933–2943. 26. Stachowiak JC, Hayden CC, Sasaki DY (2010) Steric confinement of proteins on lipid 53. Subach OM, Cranfill PJ, Davidson MW, Verkhusha VV (2011) An enhanced monomeric membranes can drive curvature and tubulation. Proc Natl Acad Sci USA 107(17): blue fluorescent protein with the high chemical stability of the chromophore. PLoS 7781–7786. One 6(12):e28674. 27. Lyon AM, Tesmer JJ (2013) Structural insights into phospholipase C-β function. Mol 54. Yamaji-Hasegawa A, et al. (2003) Oligomerization and pore formation of a sphin- Pharmacol 84(4):488–500. gomyelin-specific toxin, lysenin. J Biol Chem 278(25):22762–22770. 28. Waldo GL, et al. (2010) Kinetic scaffolding mediated by a phospholipase C-beta and 55. Iwamoto K, et al. (2007) Curvature-dependent recognition of ethanolamine phos- Gq signaling complex. Science 330(6006):974–980. pholipids by duramycin and cinnamycin. Biophys J 93(5):1608–1619. 29. Tall E, et al. (1997) Phosphoinositide binding specificity among phospholipase C iso- 56. Makino A, et al. (2006) D-threo-1-phenyl-2-decanoylamino-3-morpholino-1-propanol zymes as determined by photo-cross-linking to novel substrate and product analogs. alters cellular cholesterol homeostasis by modulating the endosome lipid domains. Biochemistry 36(23):7239–7248. Biochemistry 45(14):4530–4541. 30. Wang T, Pentyala S, Rebecchi MJ, Scarlata S (1999) Differential association of the 57. Schwede T, Kopp J, Guex N, Peitsch MC (2003) SWISS-MODEL: An automated protein pleckstrin homology domains of phospholipases C-beta 1, C-beta 2, and C-delta 1 with homology-modeling server. Nucleic Acids Res 31(13):3381–3385.

6of6 | www.pnas.org/cgi/doi/10.1073/pnas.1603513113 Inaba et al. Downloaded by guest on September 30, 2021