External Push and Internal Pull Forces Recruit Curvature-Sensing N-BAR Domain Proteins to the Plasma Membrane

LETTERS

External push and internal pull forces recruit curvature-sensing N-BAR domain proteins to the plasma membrane

Milos Galic1,7, Sangmoo Jeong2, Feng-Chiao Tsai1, Lydia-Marie Joubert3, Yi I. Wu4,6, Klaus M. Hahn4, Yi Cui5 and Tobias Meyer1,7

Many of the more than 20 mammalian proteins with N-BAR of cortical actin structures3 and clathrin-mediated endocytosis4,5. domains1,2 control cell architecture3 and endocytosis4,5 by Although the cellular functions of many N-BAR domain proteins have associating with curved sections of the plasma membrane6. It been extensively studied, the fundamental chicken and egg causality is not well understood whether N-BAR proteins are recruited dilemma of whether they primarily sense or form curvature is not directly by processes that mechanically curve the plasma resolved. This question applies to all curvature binding domains but is membrane or indirectly by plasma-membrane-associated particularly relevant for N-BAR domain proteins where in vitro assays adaptor proteins that recruit proteins with N-BAR domains that showed that they selectively bind to6 as well as form7 membranes of then induce membrane curvature. Here, we show that externally high positive curvature. During endocytosis, where N-BAR domains induced inward deformation of the plasma membrane by have been most extensively studied, the N-BAR domain proteins cone-shaped nanostructures (nanocones) and internally amphiphysin 1, endophilin 2 and BIN2 are recruited only after the induced inward deformation by contracting actin cables both assembly of clathrin-coated pits in a step that preceded vesicle scission8. trigger recruitment of isolated N-BAR domains to the curved This raised the question of whether N-BAR domain proteins are plasma membrane. Markedly, live-cell imaging in adherent recruited directly by curved membranes created by clathrin-mediated cells showed selective recruitment of full-length N-BAR formation of a tubular neck or indirectly by first binding to adaptor proteins and isolated N-BAR domains to plasma membrane proteins that then recruit N-BAR domains to curve the membrane. Sim- sub-regions above nanocone stripes. Electron microscopy ilarly, it is not clear how N-BAR domain proteins that bind9,10 N-WASP conﬁrmed that N-BAR domains are recruited to local and other actin-binding proteins11,12 are recruited to actin-rich regions membrane sites curved by nanocones. We further showed that of the plasma membrane to then in turn regulate the actin architecture. N-BAR domains are periodically recruited to curved plasma Our study aimed to distinguish between sensing and forming curved membrane sites during local lamellipodia retraction in the front membranes by investigating recruitment of N-BAR domain protein to of migrating cells. Recruitment required myosin-II-generated the plasma membrane in living cells. We focused on two examples of force applied to plasma-membrane-connected actin cables. N-BAR domains, the one from nadrin 2 (refs 3,13), a regulator of actin Together, our results show that N-BAR domains can be directly polymerization, and the one from amphiphysin 1 (ref. 4), a regulator recruited to the plasma membrane by external push or internal of endocytosis. We developed a method to directly deform the plasma pull forces that locally curve the plasma membrane. membrane in live cells using cone-shaped tin oxide nanostructures14 (nanocones). The top panels in Fig. 1a show top and side electron N-BAR domains are versatile membrane-binding regulatory elements microscopy (SEM) views of the formed nanocones and the bottom left that function in a wide range of cellular processes including regulation panel shows a cell growing on top of the nanocone surface.

1Department of Chemical and Systems Biology, Stanford University, 318 Campus Drive, Clark Building W200, Stanford, California 94305, USA. 2Department of Electrical Engineering, Stanford University, 476 Lomita Mall, McCullough Building 217, Stanford, California 94305, USA. 3Cell Sciences Imaging Facility, Stanford University School of Medicine, Stanford, California 94305, USA. 4Department of Pharmacology, Lineberger Comprehensive Cancer Center, University of North Carolina, 120 Mason Farm Road, Genetic Medicine Building, Chapel Hill, North Carolina 27599, USA. 5Department of Materials Science and Engineering, Stanford University, 476 Lomita Mall, McCullough Building 343, Stanford, California 94305, USA. 6Present address: Center for Cell Analysis and Modeling, University of Conneticut Health Center, 400 Farmington Avenue, Cell and Genome Sciences Building, Farmington, Connecticut 06032, USA. 7Correspondence should be addressed to M.G. or T.M. (e-mail: [email protected] or [email protected])

Received 1 September 2011; accepted 30 May 2012; published online 1 July 2012; DOI: 10.1038/ncb2533

LETTERS

acNanocones Nanocones Unroofed 3T3 cells on nanocones (SEM, top) (SEM, top) (SEM, side)

200 nm 200 nm

200 nm

3T3 cells on nanocones 10 µm (SEM, top) 100

25 Transparency 1 µm

(normalized to glass) 0

Glass 200 nm600 nm 1,000 nm 200 nm 200 nm b 3T3 cells (nanocones > 600 nm) 9%

d 3T3 cells on nanocones (TEM, side) 91% 5 µm

Glass Nanocones (600 nm)

3T3 cells (nanocones 200 nm) 200 nm

48% 52%

20 µm Glass Nanocones (200 nm)

Figure 1 Nanocones induce inward curved plasma membrane deformations on glass with blue. After 48 h, the cell density on 600 nm cones (brown) is at the basal plasma membrane of adherent cells. (a) SEM micrographs of signiﬁcantly lower when compared with adjacent glass surface, whereas no nanocones on a glass substrate shown from the top (top, left) and the side changes are visible for 200 nm nanocones (yellow). Analyses of cell density (top, right). SEM micrograph of 3T3 cells (red) grown on 200 nm nanocones for nanocones of 600 nm (n = 1,484 cells, top) and 200 nm (n = 780 cells, (bottom, left). Transparency of nanocones decreases with increasing cone bottom) height are shown to the right. (c) SEM micrographs of the inner side height (bottom, right). For each height, the transparency was measured at of the plasma membrane of cells grown on 200 nm nanocones. Cells were eight different regions of the sample. Error bars represent s.e.m. of the mean sonicated to remove everything except the plasma membrane of cells grown value. (b) 3T3 cells cultured on 20-µm-wide stripes of 600 nm nanocones on nanocones (red). Note the nanocone-induced deformation of the plasma (top, left) and 200 nm nanocones (bottom, left). The sawtooth symbols at membrane. (d) TEM micrographs of 3T3 cells grown on 200 nm nanocones. the bottom of the panels and the white lines depict the location of the Cells are coloured in red. Note that nanocones do not penetrate the plasma nanocone stripes. Cells on top of nanocones are marked with red, and cells membrane.

We tested an array of cone sizes that can be formed by this process electron microscopy (TEM) images of cross-sections of the basal plasma and found an optimal height of 200 nm with a diameter of 50 nm at membrane (Fig. 1d). the base. The 200 nm nanocones had a 90% transparency compared To compare the effect of nanocone-induced plasma membrane with glass without nanocones (Fig. 1a, bottom right). When cultured curvatures on N-BAR domain recruitment within the same cell, on 200 nm but not on 600 nm nanocones, cells showed no significant we designed a pattern of 3-µm-wide stripes, alternating between preference between nanocone and flat surfaces (Fig. 1b) and freely smooth and nanocone-coated surfaces, plated 3T3 cells expressing migrated across 20-µm-wide bands of nanocones (data not shown). fluorescently tagged N-BAR proteins on these patterned nanostructures When culturing cells on these nanocones, we found small-diameter and imaged them with a confocal microscope. Markedly, we found positively curved sections (inward bend) in the basal plasma membrane puncta selectively over nanocone stripes when we expressed the YFP- of unroofed cells in SEM micrographs (Fig. 1c), and in transmission conjugated N-BAR protein nadrin 2 (Fig. 2a). Control experiments

LETTERS

a Nadrin 2 (FL) / membrane d Nadrin 2 N-BAR (TEM, immunogold) 3 μm 200 nm 5 4

N-BAR density 1 ∗∗

(goldparticles per 100 nm) 0 Nanocones No nanocones e Nadrin 2 (N-BAR)

15 N-BAR puncta

b Nadrin 2 (N-BAR) 10 ∗∗ μ 3 m ∗∗ Cytosol 900 5 600 10 μm 0 Nadrin 2 intensity

Nadrin 2 300 increase)

(normalized grey values) (normalized grey Background 0 –5 0153045 60 (percentage of puncta (percentage Control (glass) Nadrin 2 (FL) 03060 Time (s) Nadrin 2 (N-BAR) Time (s)

f c Nadrin 2 (N-BAR) Amphiphysin 1 (N-BAR) –20" –15" –10" –5" 100 t = 13.8 ± 0.3 s 3 μm 1/2 ∗∗ 5" 10" 15" Bleach ∗∗ 300 20" 25" 30" 35" 200 50

100 40" 45" 50" 55" Nadrin 2 intensity increase) Amphiphysin 1

0 in percentage) (recovery Control (glass) 60"65" 70" 75" (percentage of puncta (percentage 0 Amphiphysin 1 (FL) 0306090 120 Amphiphysin 1 (N-BAR) 1 μm Time (s)

Figure 2 Nanocone-induced membrane deformation triggers N-BAR gold-conjugated secondary antibodies accumulate over nanocone-induced domain recruitment to the plasma membrane. (a) Nadrin 2 forms puncta inward membrane deformations. The immunogold density was measured over nanocones. 3T3 cells were grown on 3-µm-wide stripes of nanocones over nanocones (n = 25) where deformation of the plasma membrane was and transfected with nadrin 2 (red) together with membrane marker observed and compared with adjacent regions within the same image. CAAX (green). Nadrin 2 puncta formation was selectively observed Error bars represent s.e.m. of the mean value. (e) Nanocone-induced over nanocone stripes. A magnification of a selected region (white N-BAR domain puncta of nadrin 2 are stationary and long-lived. 3T3 rectangle) is shown to the right. (b,c) The N-BAR domain is sufficient cells plated on nanocones were transfected with the fluorescently tagged for puncta formation. Quantitative analysis of the increase in puncta N-BAR domain of nadrin 2 and imaged every 500 ms for 60 s. Kymograph density induced by nanocones for full-length (FL, light red, n = 14 cells) of two representative puncta (orange) is shown below. To the right, a and the isolated N-BAR domain (N-BAR, dark red, n = 12 cells) of graph depicting traces of individual N-BAR domain puncta (red), cytosol nadrin 2 (b) and full-length (light red, n = 10 cells) and the isolated (green) and background (black) is shown. (f) FRAP analysis indicates N-BAR domain (dark red, n = 12 cells) of amphiphysin 1 (c), respectively. a stable, long-lasting and dynamic fraction of comparable sizes over Error bars represent s.e.m. of the mean value. (d) TEM micrograph individual nanocones. Magnified area showing a time-series following a showing that the N-BAR domain of nadrin 2 accumulates on membrane single nanocone-induced N-BAR domain accumulation through a FRAP deformations induced by nanocones. 3T3 cells were transfected with cycle (left) and analysis of multiple experiments (right) are shown. About the fluorescently tagged N-BAR domain of nadrin 2, fixed and incubated half of the BAR domains present in the puncta are exchanged with a with primary antibodies directed against the fluorescent tag. Note that half-life (t1/2) of ∼14 s. showed that the average local increase in nadrin 2 intensity in puncta We then showed that the N-BAR domain of nadrin 2 is sufficient was 113 ± 9% over a uniform cytosolic background whereas a plasma to mediate the punctuate recruitment to the regions with nanocone membrane marker increased at the same sites only by 45 ± 3%, stripes (Fig. 2b left). In a quantitative single-cell analysis, full-length arguing that nadrin 2 is recruited to local sites within the plasma protein and the isolated N-BAR domain generated significantly more membrane. A correlation analysis resulted in the same conclusion puncta in the plasma membrane over nanocone stripes when compared (Supplementary Fig. S1a,b). with the plasma membrane over the flat surfaces (Fig. 2b, right panel).

LETTERS

This recruitment by nanocone stripes was not restricted to the N-BAR by variability in cell thickness or membrane accumulation (Fig. 3a,b,e). domain of nadrin 2, but also applied to the isolated N-BAR domain of Furthermore, the N-BAR domain of nadrin 2 did not co-localize amphiphysin 1 (Fig. 2c). with a marker for clathrin-mediated endocytosis at the leading Using electron micrographs, we then validated the precise edge (Supplementary Fig. S2h). However, we observed transient localization of the N-BAR domain over nanocones. Staining cells co-localization with clathrin puncta behind the leading edge, suggestive transfected with the YFP-tagged N-BAR domain of nadrin 2 of temporary recruitment of the isolated N-BAR domain of nadrin 2 to with an antibody directed against the fluorescent tag showed a these internal sites during endocytosis8,18 (Supplementary Fig. S2i). significant enrichment of gold particles over nanocones (Fig. 2d and Together, these results argued that N-BAR domain proteins are Supplementary Fig. S1c). Next, we measured the persistence of N-BAR recruited to the actin-rich leading-edge membrane in a process domain puncta over nanocones in live cells. We found the puncta that does not involve endocytosis and can be mediated by their themselves to be stable for periods longer than 1 min (Fig. 2e), whereas N-BAR domains alone. about half of the BAR domains within the puncta exchanged with To more directly investigate the effect of peripheral actin a time constant of ∼14 s, as shown in fluorescence recovery after polymerization on the localization of N-BAR domains in the photobleaching (FRAP) experiments (Fig. 2f). front, we transiently changed actin dynamics in live cells. First, As many N-BAR domain proteins have roles in endocytosis4,5, we monitored N-BAR domain puncta at the leading edge of 3T3 we determined, as a control, whether the observed puncta reflect cells on depolymerization of actin filaments. N-BAR domain puncta endocytotic sites that might be induced by nanocones. We did not disappeared within less than 5 min after the addition of 4 µM observe a nanocone-dependent increase in the average endocytosis latrunculin A (Fig. 3f) or 5 µM cytochalasin D (Fig. 3g) and reappeared rate (Supplementary Fig. S2a). Although, clathrin light chain and once the drug was washed out. In a complementary experiment, we dynamin both enriched over stripes of nanocones (Supplementary rapidly triggered actin polymerization taking advantage of an assay Fig. S2b,c), co-expression experiments with nadrin 2 and clathrin based on inducible translocation of TIAM1, an exchange factor for light chain showed only minimal overlap (Supplementary Fig. S2d), RAC (ref. 19). 3T3 cells were co-transfected with a fluorescently tagged arguing that N-BAR domain recruitment by nanocones can be induced N-BAR domain together with a construct containing FKBP fused to independently of endocytosis. This low correlation was confirmed in TIAM1 and a membrane-anchored FRB domain. The addition of a spatial cross-correlation analysis where the co-localization of the rapamycin triggered dimerization of FRB and FKBP, rapidly recruiting near identical CFP-tagged nadrin 2 and YFP-tagged nadrin 2 was TIAM1 to the plasma membrane, and causing the activation of the used as a reference (Supplementary Fig. S2e,f). Tracking individual small GTPase RAC1. This led to increased levels of RAC–GTP at CFP–clathrin light chain puncta over nanocones over time provided the plasma membrane causing extensive actin polymerizations and, no evidence for sequential co-localization, ruling out that N-BAR with a delay, the translocation of the isolated N-BAR domain to the domain recruitment follows clathrin accumulation with a delay plasma membrane (Supplementary Fig. S3). A comparable result (Supplementary Fig. S2g). Together, these results argue that nanocones was observed when a photoactivatable version of the small GTPase provide an external local push force that induces stable membrane RAC1 (LOV–RAC1; ref. 20) was used to trigger actin polymerization deformations and that an inward curved plasma membrane is sufficient (Supplementary Fig. S4). Together, these experiments argue that to recruit N-BAR domains. increases in dynamic cortical actin polymerization result, with a delay, We then considered that internal forces that pull on local membrane in the recruitment of curvature-sensing N-BAR domain proteins. sites during actin reorganization or endocytosis might be equally Cortical actin polymerization often occurs in cycles where a capable of inducing the high inward curvatures necessary for N-BAR polymerization phase that extends the lamellipodia outward is followed domain recruitment. We tested this hypothesis by focusing on by a retraction phase whereby the actin motor myosin II pulls migrating cells that experience significant membrane deformation at lamellipodia partially back before the next extension21–23. Markedly, the leading edge15. As shown in Fig. 3a, the plasma membrane-localized time-lapse imaging of cyclic lamellipodia in the front of polarized N-BAR domain proteins nadrin 2 and amphiphysin 1 were enriched in 3T3 cells showed that N-BAR recruitment occurs primarily during the actin-rich front of polarized migrating 3T3 cells. This localization the retraction phase rather than the extension/polymerization phase only required the N-BAR domain, because the same polarized (Fig. 4a). A time-course analysis over multiple expansion–retraction localization was observed for full-length and isolated N-BAR domains cycles showed that the concentration of the N-BAR domain of nadrin 2 and amphiphysin 1 (Fig. 3b). Deletion of the amphipatic (normalized to the total amount of actin filaments) remained low helix, a sequence motif required for binding and stabilization of as the lamellipodia moved outward and only increased as it was curved membranes7,16,17, prevented enrichment of the N-BAR domain retracted (Fig. 4b and Supplementary Movie S1). This observation was protein, arguing that localization critically depends on this structural confirmed using an autocorrelation analysis comparing normalized feature required for membrane binding and present in all N-BAR N-BAR domain concentration and lamellipodia position (Fig. 4c). This domain proteins (Fig. 3c). cyclic nature of N-BAR domain recruitment during extension and We observed an overlap between membrane regions with high retraction is schematically shown in Fig. 4d. cortical actin and either nadrin 2 or amphiphysin 1 N-BAR domains, We reasoned that myosin-II-mediated lamellipodia retraction suggesting that curved membrane sites are created above cortical involves pulling forces applied to the plasma membrane through actin networks (Fig. 3d). Control experiments with markers for membrane-anchored actin cables that cause local inward membrane the cytoplasm and the plasma membrane showed that this partial curvature while the distributed surrounding actin meshwork resists co-localization of N-BAR domains with cortical actin was not caused the retraction. Consistent with such a role of myosin II, we observed

LETTERS

a Nadrin 2 (FL) Amphiphysin 1 (FL) CFP (cytosol) CFP (cytosol) f PreLatA Wash

Merge Merge 10 μm

Nadrin 2 Amph1 (FL) (FL)

Cytosol Cytosol Nadrin 2 10 μm 10 μm (N-BAR) b Nadrin 2 (N-BAR) Amphiphysin 1 (N-BAR) f-tractin CFP (cytosol) CFP (cytosol)

Merge Merge 100 Nadrin 2 Amph1 (N-BAR) (N-BAR) 50 Pre Cytosol Cytosol ∗∗ 10 μm 10 μm LatA (5 min)

Nadrin 2 (N-BAR) Wash (60 min) c Nadrin 2 (ΔN-BAR) Amphiphysin 1 (ΔN-BAR) 0 CFP (cytosol) CFP (cytosol) (puncta intensity in percentage)

Merge Merge g Nadrin 2 Amph1 Pre CytoD Wash Δ (ΔN-BAR) ( N-BAR) 10 μm

Cytosol Cytosol 10 μm 10 μm Nadrin 2 f-tractin Nadrin 2 (N-BAR) Amphiphysin 1 (N-BAR) d Nadrin 2 f-tractin (actin) f-tractin (actin) (N-BAR)

Merge Merge f-tractin

Nadrin 2 Amph1 (N-BAR) (N-BAR)

f-tractin f-tractin μ μ 10 m 10 m 100

Nadrin 2 (N-BAR) Amphiphysin 1 (N-BAR) e CAAX (membrane) CAAX (membrane) 50 Pre ∗∗ Merge Merge CytoD (5 min)

Nadrin 2 (N-BAR) Wash (60 min) Nadrin 2 Amph1 0 (N-BAR) (N-BAR) (puncta intensity in percentage)

CAAX CAAX 10 μm 10 μm

Figure 3 N-BAR domains are dynamically recruited to local membrane cells were transfected with a marker for filamentous actin (green) and sites at the leading edge of migrating 3T3 cells. (a) The proteins nadrin 2 the N-BAR domain of nadrin 2 (left, red) or amphiphysin 1 (right, red), and amphiphysin 1 are enriched in the leading edge of migrating 3T3 respectively. (e) N-BAR domain and a plasma membrane marker only cells. (b) Enrichment in the leading edge is dependent on the N-BAR partially overlap. 3T3 cells transfected with a membrane marker (CAAX, domain. Cells expressing a cytosolic marker (CFP, green) and the isolated green) and the N-BAR domain of nadrin 2 (left, red) and amphiphysin 1 N-BAR domain of nadrin 2 (left, red) and amphiphysin 1 (right, red). (right, red) are shown. (f) The addition of the actin polymerization inhibitor Note the polarized localization of the isolated N-BAR domain to the latrunculin A (LatA) reversibly inhibits N-BAR domain puncta formation. leading edge. (c) Enrichment in the leading edge is dependent on the A total of 1,303 individual puncta from 12 cells were analysed for the amphipatic helix. Cells expressing a cytosolic marker (CFP, green) and drug washout experiments. Pre, pre-treatment. (g) Addition of the actin the isolated N-BAR domain of nadrin 2 (left, red) and amphiphysin 1 polymerization inhibitor cytochalasin D (CytoD) reversibly inhibits N-BAR (right, red) that is lacking the amino-terminal amphipatic helix show no domain puncta formation. A total of 428 puncta from eight cells were polarized localization to the leading edge. (d) N-BAR domain patches analysed for the drug washout experiments. Error bars represent s.e.m. of show significant overlap with marker for filamentous actin (f-tractin). 3T3 the mean value. ∗∗P < 0.01. during lamellipodia retraction a parallel increase of the N-BAR domain transitions into the lamella (Fig. 4e and Supplementary Movie S2). To of nadrin 2 at the front and of the heavy chain of non-muscle myosin further validate the hypothesis that myosin II may pull on actin cables II-A (NMHC-II-A) in a region closely behind where the lamellipodia to promote retraction, we used ML-7, an inhibitor of myosin light

LETTERS

a 3T3 cell (lamellipodia protrusion–retraction cycle) 5 μm f Nadrin 2 (N-BAR) 0 Merge 2.5 Nadrin 2 (N-BAR) 5.0 7.5 f-tractin 10.0 0 1020 30 40 50 60 70 80 90 100 110 120 130 140150 160 170 180 10 μm 12.5 Time (s) Pre ML-7 Time (s) 100

b (position in pixels)

2.0 6 Leading edge 50 4 ∗∗ Percentage of puncta intensity

f-tractin 2 1.0 0 O 2 Nadrin 2 (N-BAR) 0 Pre H 0.6 ML-7 0 100200 300 400 500600 700 800 900 1,000 1,100 1,200 1,300 1,400 Time (s)

c d g 0.05 Low Amph1 (N-BAR) 0 0 (N-BAR) 5 Collapse Collapse (f-tractin) –0.10 High 10

Growth Growth 15 versus position) (N-BAR/f-tractin Cross-correlation –0.20 –100 0 100 20 Leading edge 10 μm Time difference (s) 25 (movement in pixels) Time (s) Pre ML-7 Time (s) e 3T3 cell (lamellipodia protrusion–retraction cycle) 5 μm 100

Merge ∗∗ 50 Nadrin 2

(N-BAR) Percentage of puncta intensity 0 O NMHC-II-A Pre H 2 ML-7 0 1020 30 40 50 60 70 80 90 100 110 120 130 140150 160 170 180 Time (s)

Figure 4 Myosin-II-dependent contraction of actin cables induces N-BAR Graph depicting the average cross-correlation of the normalized N-BAR domain recruitment to the lamellipodia plasma membrane. (a) N-BAR concentration at the leading edge with the position of the leading edge over domain of nadrin 2 recruits to contracting lamellipodia. Montage depicting time. (d) Model of the normalized N-BAR domain of nadrin 2 concentration an expansion–collapse cycle in 3T3 cell lamellipodia. Merged images of as a function of the leading-edge position. (e) NMHC-II-A and the N-BAR the N-BAR domain of nadrin 2 (red) and f-tractin (green) at an interval domain of nadrin 2 both enrich during lamellipodia retraction, N-BAR close of 10 s are shown in the top row. For clarification, separate channels of to the front and myosin II a bit further back. The montage of a 3T3 cell the fluorescently tagged N-BAR domain of nadrin 2 (middle row) and expressing nadrin 2 (red) and NMHC-II-A (green) shows accumulation of f-tractin (bottom row) are shown below. Note the delayed accumulation both proteins during lamella retraction. (f,g) Inhibition of myosin II causes of nadrin 2 during the retractive phase of lamellipodia. (b) Comparison rapid disassembly of N-BAR domain puncta at the leading edge of 3T3 cells. of normalized N-BAR domain intensity during an expansion–collapse The addition of the MLCK inhibitor ML-7 triggers rapid disassembly of the cycle of the lamellipodia. Correlation of the N-BAR domain concentration N-BAR domain of nadrin 2 (f, red, n = 8 cells) and amphiphysin 1 (g, red, normalized to f-tractin (red) and lamellipodia position (blue) are shown for n = 12 cells) puncta, respectively. In contrast, no effect was visible on the an individual position on the leading edge monitored over an interval of addition of water (white, n = 6 cells). For better visualization, a montage of 1,400 s. (c) The fluorescence level of the N-BAR domain (after normalization a magnified section is depicted next to the pictures. Error bars represent to f-tractin intensity) anti-correlates with the expansion of the leading edge. s.e.m. of the mean value, ∗∗P < 0.01. chain kinase (MLCK) that controls myosin II activity of non-muscle of migrating cells and observed significant inward curved sections cells. Consistent with a key role for MLCK and myosin II in the process along the leading edge (Fig. 5a). Actin cables oriented perpendicular to of N-BAR domain recruitment, the patches of the respective N-BAR the plasma membrane often pointed to the inward curved section domains of nadrin 2 and amphiphysin 1 rapidly disappeared on the of the plasma membrane (Fig. 5b and Supplementary Fig. S5a), addition of the inhibitor (Fig. 4f,g). suggesting that force applied to these actin cables was responsible This suggested that myosin-II-mediated force on actin cables induces for the deformation. Furthermore, immunogold localization showed local inward deformations of the plasma membrane at sites where the that the isolated N-BAR domain of nadrin 2 locally enriched at these cables contact the leading-edge plasma membrane. We used electron inward curved membrane sites (Fig. 5c and Supplementary Fig. S5b). microscopy24 to analyse the topography of the plasma membrane Together, these findings introduce a recruitment mechanism whereby

LETTERS

a 3T3 cell (SEM) b 3T3 cell (TEM) c Immunogold (N-BAR)

200 nm 200 nm

8 7 6 5 4 3 2 N-BAR density 1 ∗∗ 1

(particles per 100 nm) 0 Inward curved membranes Flat membranes

500 nm d Nanocone-induced Myosin-II-dependent Clathrin-mediated plasma membrane plasma membrane plasma membrane deformation deformation deformation

Membrane N-BAR domain protein Nanocones Cortical actin meshwork Myosin II Plasma-membrane-anchored 10 μm actin filaments Clathrin-coated vesicle

Figure 5 External push and internal pull forces applied to local plasma of nadrin 2, fixed and incubated with primary antibodies directed against membrane sites recruit N-BAR domains to inward curved membrane. (a) SEM the fluorescent tag. Enrichment of immunogold particles to curved sites micrographs depicting membrane deformation at the leading edge. For better was measured (n = 15 cells) and compared to adjacent regions within the visualization, a magnification of a selected region (white rectangle, bottom) same image. Note that gold-conjugated secondary antibody is significantly depicting a lamellipodium is shown (top). (b) TEM micrograph of a 3T3 cell enriched on inward curved membrane sites (graph, bottom). (d) Different parallel to the glass plane indicating actin filaments pulling at the plasma types of force-dependent membrane deformation recruit N-BAR domain membrane as a source of plasma deformation. For better visualization the proteins. N-BAR domain proteins (red) sense positive (inward) membrane ends of individual actin cables are highlighted in red. (c) The N-BAR domain curvature induced by external force such as the nanocone substrate (left), of nadrin 2 accumulates on membrane deformations at the leading edge. during myosin-II-triggered actin contraction (middle), and during endocytosis 3T3 cells were transfected with the fluorescently tagged N-BAR domain (right). Error bars represent s.e.m. of the mean value; ∗∗P < 0.01.

MLCK and myosin-II-mediated actin contraction pulls on local plasma N-BAR domain recruitment (Fig. 5d, middle), a finding that has broad membrane–actin cable contact sites to create local inward curved implications in actin-dependent processes such as cell polarization and membrane sites that then recruit N-BAR domain proteins. Notably, migration. Third, our results also probably apply to endocytosis where N-BAR puncta disappeared when plasma membrane tension was clathrin coats may provide the force that recruits N-BAR domains increased by lowering the external osmotic strength (Supplementary by creating a neck with a curved membrane tube that connects the Fig. S5c). Correspondingly, lowering the membrane tension with forming clathrin vesicle to the rest of the plasma membrane (ref. 8; high osmotic pressure or the surface relaxant deoxycholate25 triggered Fig. 5d, right). Although these different recruitment mechanisms argue puncta formation (Supplementary Fig. S5d,e), suggesting that N-BAR that local force application and induction of membrane curvature are recruitment can further be regulated by changes in the global plasma needed to trigger the initial N-BAR domain recruitment in live cells, it membrane tension parameter. is likely that N-BAR domains, once recruited, have a complementary Our results argue that different modes of mechanical deformation role in stabilizing the curved membrane section. of the plasma membrane exist that trigger the recruitment of N-BAR- Our study shows that cytosolic N-BAR domains continuously domain-containing regulatory proteins. First, the ability of external scan the plasma membrane for highly inward curved sections and nanocones to directly trigger N-BAR domain recruitment suggests that accumulate at sites where the membrane is sufficiently deformed a receptor-independent mechanical signalling mechanism may exist by local forces. Our study further suggests that any type of push whereby extracellular matrix components of sufficient stiffness can force applied to local plasma membrane sites from the outside or trigger local deformations of the plasma membrane to directly recruit pull force applied to local sites from the inside will induce N-BAR and activate intracellular signalling proteins with N-BAR domains domain recruitment as long as the local force is sufficiently strong (Fig. 5d, left). Second, our results with actin and myosin II argued and is counteracted by distributed forces on the membrane in the that internal force applied to membrane connected actin cables can opposing direction. Finally, different force-generating mechanisms exert sufficient pull force to curve the plasma membrane and cause involving clathrin, microtubules, actin cables or extracellular matrix

LETTERS may create in different contexts either inward (positive) or outward 5. Ringstad, N. et al. Endophilin/SH3p4 is required for the transition from early (negative) curved plasma membrane that probably provides only two to late stages in clathrin-mediated synaptic vesicle endocytosis. Neuron 24, 143–154 (1999). types of distinct recruitment signal. To gain specificity, additional co- 6. Bhatia, V. K. et al. Amphipathic motifs in BAR domains are essential for membrane regulatory mechanisms are probably needed to link specific mechanical curvature sensing. EMBO J. 28, 3303–3314 (2009). 7. Peter, B. J. et al. BAR domains as sensors of membrane curvature: the amphiphysin deformations to distinct signalling pathways and cell functions. BAR structure. Science 303, 495–499 (2004). 8. Taylor, M. J., Perrais, D. & Merrifield, C. J. A high precision survey of the METHODS molecular dynamics of mammalian clathrin-mediated endocytosis. PLoS Biol. 9, e1000604 (2011). Methods and any associated references are available in the online 9. Yamada, H. et al. Dynamic interaction of amphiphysin with N-WASP regulates actin version of the paper at www.nature.com/naturecellbiology assembly. J. Biol. Chem. 284, 34244–34256 (2009). 10. Yarar, D., Waterman-Storer, C. M. & Schmid, S. L. SNX9 couples actin assembly Note: Supplementary Information is available on the Nature Cell Biology website to phosphoinositide signals and is required for membrane remodeling during endocytosis. Dev. Cell 13, 43–56 (2007). ACKNOWLEDGEMENTS 11. Rocca, D. L., Martin, S., Jenkins, E. L. & Hanley, J. G. Inhibition of Arp2/3-mediated The authors thank the members of the Meyer laboratory for comments and actin polymerization by PICK1 regulates neuronal morphology and AMPA receptor discussion. M.G. was supported by the Swiss National Science Foundation endocytosis. Nat. Cell Biol. 10, 259–271 (2008). (No. PBBSP3-123159), and a Novartis Jubilaeumsstiftung and Stanford Deans 12. Salazar, M. A. et al. Tuba, a novel protein containing bin/amphiphysin/Rvs and Dbl Postdoctoral Fellowship. S.J. was supported by a Korea Foundation for Advanced homology domains, links dynamin to regulation of the actin cytoskeleton. J. Biol. Studies graduate fellowship. Y.C. acknowledges the partial support from a DOE- Chem. 278, 49031–49043 (2003). 13. Richnau, N. & Aspenstrom, P. Rich, a ρ GTPase-activating protein domain- EFRC at Stanford: Center on Nanostructuring for Efficient Energy Conversion (No. containing protein involved in signaling by Cdc42 and RAC1. J. Biol. Chem. 276, DE-SC0001060). T.M. acknowledges financial support from the National Institutes 35060–35070 (2001). of Health, MH064801, MH095087 and GM063702. 14. Jeong, S., McDowell, M. T. & Cui, Y. Low-temperature self-catalytic growth of tin oxide nanocones over large areas. ACS Nano 5, 5800–5807 (2011). AUTHOR CONTRIBUTIONS 15. Giannone, G. et al. Periodic lamellipodial contractions correlate with rearward actin M.G. performed all experiments and analysed the data. F-C.T. developed the waves. Cell 116, 431–443 (2004). temporal cross-correlation analysis. S.J. and Y.C. designed the nanocones. L-M.J. 16. Gallop, J. L. et al. Mechanism of endophilin N-BAR domain-mediated membrane helped with the SEM. Y.I.W. and K.M.H. developed the photoactivatable RAC curvature. EMBO J. 25, 2898–2910 (2006). construct. M.G. and T.M. designed the experiments, interpreted the results and 17. Masuda, M. et al. Endophilin BAR domain drives membrane curvature by two newly wrote the manuscript. All authors discussed the results and the manuscript. T.M. identified structure-based mechanisms. EMBO J. 25, 2889–2897 (2006). supervised the study. 18. Ferguson, S. M. et al. Coordinated actions of actin and BAR proteins upstream of dynamin at endocytic clathrin-coated pits. Dev. Cell 17, 811–822 (2009). COMPETING FINANCIAL INTERESTS 19. Inoue, T. & Meyer, T. Synthetic activation of endogenous PI3K and Rac The authors declare no competing financial interests. identifies an AND-gate switch for cell polarization and migration. PLoS One 3, e3068 (2008). Published online at www.nature.com/naturecellbiology 20. Wu, Y. I. et al. A genetically encoded photoactivatable Rac controls the motility of Reprints and permissions information is available online at www.nature.com/reprints living cells. Nature 461, 104–108 (2009). 21. Burnette, D. T. et al. A role for actin arcs in the leading-edge advance of migrating 1. Habermann, B. The BAR-domain family of proteins: a case of bending and binding? cells. Nat. Cell Biol. 13, 371–381 (2011). EMBO Rep. 5, 250–255 (2004). 22. Machacek, M. et al. Coordination of ρ GTPase activities during cell protrusion. 2. Suetsugu, S., Toyooka, K. & Senju, Y. Subcellular membrane curvature mediated by Nature 461, 99–103 (2009). the BAR domain superfamily proteins. Semin. Cell Dev. Biol. 21, 340–349 (2010). 23. Giannone, G. et al. Lamellipodial actin mechanically links myosin activity with 3. Rollason, R., Korolchuk, V., Hamilton, C., Jepson, M. & Banting, G. A CD317/ adhesion-site formation. Cell 128, 561–575 (2007). tetherin-RICH2 complex plays a critical role in the organization of the subapical 24. Abercrombie, M., Heaysman, J. E. & Pegrum, S. M. The locomotion of fibroblasts actin cytoskeleton in polarized epithelial cells. J. Cell Biol. 184, 721–736 (2009). in culture. IV. Electron microscopy of the leading lamella. Exp. Cell Res. 67, 4. David, C., McPherson, P. S., Mundigl, O. & de Camilli, P. A role of amphiphysin in 359–367 (1971). synaptic vesicle endocytosis suggested by its binding to dynamin in nerve terminals. 25. Raucher, D. & Sheetz, M. P. Cell spreading and lamellipodial extension rate is Proc. Natl Acad. Sci. USA 93, 331–335 (1996). regulated by membrane tension. J. Cell Biol. 148, 127–136 (2000).

DOI: 10.1038/ncb2533 METHODS

METHODS was included. The primers were 50-CACCATGGCCGACATCAAGACGG-30 and Nanocone formation. A thin film of tin (35 nm thickness) was deposited by 50-GCGGAGAGGACCCGAATCAC-30. The 1N-BAR construct, lacking the initial electron-beam evaporation on a glass coverslip at room temperature. The glass amphipatic helix, was encoded following previous publications7,16,17 by amino with the deposited tin was then exposed to a nitrogen gas environment with a low acids 28–256 using the following primers: 50-AGCCTTCCTCGAGATGCTGGGGA- concentration of oxygen (about 1 part per million) and heated to 350 ◦C for 90 min. AAGCTGATGAG-30 and 50-AGCCTTCGAATTCGCGGAGAGGACCCGAATC-30. To make the nanocone structures transparent, the glass coverslip with nanocones Full-length nadrin 1 (Entrez Gene ID: 55114) was obtained from the ORFeome was further heated to 400 ◦C for 3 h in air. The annealing to the glass and the library (internal ID: 14668). The N-BAR domain spanning amino acids 1–260 formation of the replicate nanocone shapes occurred during this step. was generated using the following primers: 50-AGCCTTCGAATTCATGAAG- AAGCAGTTCAACCGC-30 and 50-AGCCTTCGGTACCCCTCTTCAGGTGTT- Electron micrographs. 3T3 cell were cultured on nanocone-coated glass coverslips CTTC-30. The 1N-BAR construct, lacking the initial amphipatic helix, was and fixed using 2% glutaraldehyde (8% stock EM grade) and 4% paraformaldehyde encoded by amino acids 18–260 using the following primers: 50-AGCCTTCGAA- in 0.1 M HEPES, Na cacodylate buffer at pH ∼7.4 for 10 min. TTCATGGCTGAGAAAACAGAAGTC-30 and 50-AGCCTTCGGTACCCCTCTTC- For the TEM micrographs in Figs 1d, 5b and Supplementary Fig. S5a, the fixation AGGTGTTCTTC-30. medium was then replaced with 1% OsO4 in double-distilled H2O (ddH2O) for 1 h, Full-length GMIP (Entrez Gene ID: 51291) was obtained from the ORFeome and the cells were washed three times in ddH2O and incubated with 1% uranyl library (internal ID: 54593). acetate in ddH2O overnight. Samples were then dehydrated as follows: 1× 50% Full-length TRIP10 (Entrez Gene ID: 9322) was obtained from the OR- ethanol for 5 min; 1× 70% ethanol for 5 min; 1× 95% ethanol for 10 min; 2× 100% Feome library (internal ID: 6046). The crystal structure of TRIP10 has ethanol for 15 min; and finally moved to propylene oxide for 15 min. Epon (20 ml previously been solved26. To generate the isolated F-BAR domain con- EMbed 812, 16 ml DDSA, 8 ml NMA, 0.77 ml DMP-90) was prepared and sample struct, we used the following primers decoding amino acids 1–298: 50-AGC- infiltration performed in a series of steps: 1:1 (Epon/propylene oxide) for 1 h; 2:1 CTTCAAGCTTATGGATTGGGGCACTGAG-30 and 50-AGCCTTCCGGTACCGC- (Epon/propylene oxide) overnight; 100% Epon for 3 h; fresh 100% Epon for 2 h. TGTCGGAGGGTGCACG-30. After the final step, the Epon was polymerized at 65 ◦C overnight. The glass coverslips Full-length FCHO2 (Entrez Gene ID: 115548) was obtained from the and nanocones were then dissolved in 49% hydrofluoric acid (10 min), and samples ORFeome library (internal ID: 7370). The crystal structure of FCHO2 has 27 were washed three times with ddH2O. Samples were then embedded a second time previously been solved . To generate the isolated F-BAR domain con- in Epon and polymerized overnight. Finally, samples were cut orthogonally to the struct, we used the following primers decoding amino acids 1–274: 50- plane of the glass slide and imaged on a JEM-1230 electron microscope using AGCCTTCGAATTCATGATGATGGCGTATTTCGTCGAG-30 and 50-AGCCTT- ×5 to ×30,000 magnification. CGGTACCAGCAGTGTCACATTCTTC-30. For the SEM micrographs in Figs 1a,c and 5a, cells were rinsed with 0.1 M Na Full-length BAIAP2 (Entrez Gene ID: 10458) was obtained from the ORFeome cacodylate buffer (pH 7.2) after primary fixation and were postfixed for 1 h with 1% library (internal ID: 10073). The crystal structure of BAIAP2 has previously been 28 OsO4 in water, washed briefly with water and dehydrated in an ascending ethanol solved . We generated the isolated I-BAR domain construct spanning amino acids series (50, 70, 90 and 100% (twice), 20 min each) before critical-point drying with 1–252: 29 liquid CO2 in a Tousimis 815B. Filters were mounted on colloidal graphite on 15 mm Other constructs, LOV–RAC1 (ref. 20), f-tractin , Lyn–FRB (ref. 30), aluminium stubs (Ted Pella), and the upper portion of the well was cut away to retain CFP–FKBP–TIAM1 (ref. 30) and NMHC-II-A (ref. 31) (Addgene no. 11347) were only the filter area with attached cells for SEM visualization. Mounted specimens previously described. were sputter-coated with 100 Å Au/Pd using a Denton Desk 11 TSC sputter coater. Visualization of samples was performed with a Zeiss Sihma field-emission SEM Cell transfection. Proteins were transfected with Lipofectamine 2000 (Invitrogen, operated at 5 kV, working distance 6 mm and an inlens SE detector under high no. 11668) according to the manufacturer’s protocol. Excessive overexpression of vacuum conditions. Images were captured in TIFF format. full-length and the isolated N-BAR domain of amphiphysin 1 led to aggregation For the SEM micrograph in Fig. 1c, 3T3 cells were unroofed, by brief sonication that induced deformations of the plasma membrane. Amphiphysin 1 promoted (Sonics Vibracell, VCX130). In detail, cells were fixed using 2% glutaraldehyde (8% invaginations that originated at the plasma membrane and occasionally polymerized stock EM grade) and 4% paraformaldehyde in 0.1 M HEPES, Na cacodylate buffer forming dynamic, tubular structures that protruded into the cytosol. Supporting our at pH ∼7.4 for 10 min. The fixation medium was then replaced with 1× PBS and the notion that N-BAR domain proteins do not deform the plasma membrane at the samples were placed in a 15 ml glass beaker filled with 10 ml of 1× PBS. A 1/8 inch native concentration, membrane tubulation did not occur at low expression levels. microprobe was placed 1 cm above the sample in the beaker and two 5 s pulses of To account for this potential overexpression artefact, all experiments on nanocones 40% intensity were applied. Cells were then put back into the fixation solution and were performed 12–24 h post-transfection, when expression was still low, and cells processed as described above. that exhibited visible large protein aggregates were excluded. For the TEM micrographs in Figs 2d, 5c and Supplementary Figs S1c and S5b, 3T3 cells were transiently transfected with the fluorescently tagged isolated N-BAR Live-cell microscopy. Live-cell dual-colour measurements were performed on a domain of nadrin 2. At 24 h post-transfection, the cells were FACS sorted and spinning-disc confocal microscope. CFP and YFP were excited using the 442 nm the YFP-positive cells were plated and fixed. The N-BAR domain of nadrin 2 was line of a helium cadmium laser (100 mW; Kimmon Electrics) and the 514 nm detected using polyclonal antibodies directed against the fluorescent tag (Invitrogen, line of an argon laser (300 mW; Melles Griot), respectively. Images were captured A11122, 1:500) and visualized using gold-conjugated secondary antibodies (Ted using an EMCCD (electron-multiplying charge-coupled device) camera (QuantEM Pella, 15726-1, 1:50). 512SC), driven by Micromanager which was mounted on the side port of an inverted microscope (model IX-71; Olympus). Co-localization studies were carried out on a Cloning. Double-stranded oligonucleotides encoding full-length and BAR do- custom-made prism-based total internal reflection fluorescence microscope using mains of the respective proteins were cloned into the pECFP–N3 or pEYFP–N3 an EMCCD camera (QuantEM 512SC) driven by Micromanager. expression vectors (CLONTECH Laboratories). All constructs were validated by sequencing before use. The following primers (with selective restriction sites) were Background subtraction algorithm. For better visualization, images Fig. 2b,c used. were processed in Matlab using a high-pass filter (http://www.csse.uwa.edu.au/∼pk/ For full-length nadrin 2 (Entrez Gene ID: 9912), 50-AGCCTTCGAATTCATGA- Research/MatlabFns/FrequencyFilt/highpassfilter.m). AGAAGCAGTTCAATCGCATGC-30 and 50-AGCCTTCGGTACCTCCTGCGCCG- AGGGCGGTGCTCTCAGACTCCTC-30were used. The nadrin 2 N-BAR domain Spatial cross-correlation analysis. Spatial cross-correlation of individual images is predicted to be encoded by amino acids 1–244 (ref. 7). Our construct con- in Supplementary Figs S1b and S2d–f was assessed using the Matlab function xcorr. tained an additional linker of 10 amino acids. The sequences 50-AGCCTT- CGAATTCATGAAGAAGCAGTTCAATCGCATGC-30 and 50-AGCCTTCGGTAC- Drugs. The following inhibitors were used: latrunculin A (4 µM; Cal Biochem no. CTCCTGCGCCCCCGAAGGAAGGCTTCTCTAC-30 were used as primers. The 428026); cytochalasin D (5 µM; Invitrogen no. PHZ1063), ML-7 (20 µM; Santa Cruz 1N-BAR construct, lacking the initial amphipatic helix, was encoded fol- Biotechnology no. sc-200553A). lowing previous publications7,16,17 by amino acids 18–254 using the following primers: 50-AGCCTTCGAATTCATGGCTGAAAAGACAGAAGTT-30 and 50- FKBP–TIAM1 assay. 3T3 cells were triple-transfected with Lyn–FRB, AGCCTTCGGTACCTCCTGCGCCCCCGAAGGAAGGCTTC-30. CFP–FKBP–TIAM1 and the YFP-tagged N-BAR domain of amphiphysin 1. At Full-length amphiphysin 1 (Entrez Gene ID: 273) was obtained from the 12–24 h post-transfection, individual cells were imaged before and at 10 s intervals ORFeome library (internal ID: 8429). The N-BAR domain of amphiphysin 1 after the addition of 100 nM rapamycin (Sigma-Aldrich, no. B0560) with a ×63 spans amino acids 1–236 (ref. 7). A 20-amino-acid additional linker sequence objective and an ×1.5 Optovar module.

METHODS DOI: 10.1038/ncb2533

LOV–RAC1 assay. HeLa cells were typically co-transfected with CFP–LOV–RAC1 of N-BAR intensity to f-tractin intensity with respect to the localization of the cell and a YFP-tagged gene of interest, using Lipofectamine 2000 and imaged boundary for each pixel was analysed. 12–36 h after transfection. All experiments were performed with constitutively active CFP–LOV–RAC1. Cells with relatively high LOV–RAC1 expression levels Statistical analysis. Statistical significance was tested using the two-tailed Student were identified (on the basis of CFP fluorescence) and selected for the light- t-test. The results from individual experiments are presented as mean values; error induced translocation studies. As LOV–RAC1 activity is triggered with blue light20, bars depict the s.e.m. acquisition of CFP–LOV–RAC1 using a 442 nm laser automatically activated the construct. Identified cells were therefore kept in the dark for 5 min to allow the 26. Shimada, A. et al. Curved EFC/F-BAR-domain dimers are joined end to end into a LOV domain to refold (half-life of LOV–RAC1 for refolding is 43 s; ref. 20) before ﬁlament for membrane invagination in endocytosis. Cell 129, 761–772 (2007). imaging. Subsequently, YFP followed by a CFP image were acquired and compared 27. Henne, W. M. et al. Structure and analysis of FCHo2 F-BAR domain: a dimerizing with a second image pair obtained 15 s later. For each cell, membrane-localized and membrane recruitment module that effects membrane curvature. Structure 15, puncta of N-BAR domain proteins were identified using the find Maxima function 839–852 (2007). in ImageJ. Individual thresholds were manually adjusted for the pre-activation 28. Millard, T. H. et al. Structural basis of ﬁlopodia formation induced by the image, and then applied the same way to the post-activated image. The negative IRSp53/MIM homology domain of human IRSp53. EMBO J. 24, 240–250 (2005). ratio of pre/post puncta stimulation in the experiments without LOV–RAC reflects 29. Johnson, H. W. & Schell, M. J. Neuronal IP3 3-kinase is an F-actin-bundling protein: a fraction of the puncta (<5%) falling below the detection level probably owing to a role in dendritic targeting and regulation of spine morphology. Mol. Biol. Cell 20, slight photobleaching. 5166–5180 (2009). 30. Inoue, T., Heo, W. D., Grimley, J. S., Wandless, T. J. & Meyer, T. An inducible translocation strategy to rapidly activate and inhibit small GTPase signaling pathways. Temporal cross-correlation analysis. Temporal cross-correlation has been Nat. Methods 2, 415–418 (2005). 32 previously described . In brief, a mask of the migrating cell was generated for 31. Wei, Q. & Adelstein, R. S. Conditional expression of a truncated fragment of individual frames and each pixel at the boundary was identified, containing nonmuscle myosin II-A alters cell shape but not cytokinesis in HeLa cells. Mol. Biol. information of N-BAR and f-tractin intensity. Individual pixels (at the cell Cell 11, 3617–3627 (2000). boundary) were tracked over multiple frames, allowing analysis on position and 32. Tsai, F. C. & Meyer, T. Ca(2+) pulses control local cycles of lamellipodia retraction velocity of the lamellipodia. Finally, temporal cross-correlation between the ratio and adhesion along the front of migrating cells. Curr. Biol. 22, 837–842 (2012).

SUPPLEMENTARY INFORMATION

DOI: 10.1038/ncb2533

a Nadrin2 (FL) / CAAX b Crosscorrelation Nadrin2 (FL) 0.8

0.4 Y-axis 0 Cross-correlation 1µm (Nadrin2-FL/CAAX)

Nadrin2 CAAX -0.4 -3.5 0 3.5 translocation X-axis (∆ µm)

c Nadrin2, N-BAR (TEM, Immunogold)

Figure S1 N-BAR domain recruitment to Nanocones is curvature-dependent. 2 vs. CAAX upon lateral shift. (c) Transmitted electron micrographs (a) 3T3 cells cultured on Nanocones were transfected with the full length show accumulation of immunogold particles directed against the isolated version of nadrin 2 (red) and the membrane marker CAAX (green). For N-BAR domain of nadrin 2 to inward curved plasma membranes formed by better visualization, a magnified section (white box) of the individual Nanocones (red arrows). 3T3 cells were transfected with a fluorescently channels is shown to the right. Note non-uniform fluorescent intensity for tagged N-BAR domain of nadrin 2 and plated on Nanocones. Fluorescently the membrane marker, likely reflecting individual or groups of Nanocones tagged N-BAR domain of nadrin 2 is detected with polyclonal anti-GFP N-BAR domain recruitment to Nanocones is Nadrin2 vs. CAAX upon lateral shift. Transmitted that deform theFigure plasma S1membrane, causing membrane accumulation in the antibody and visualized with gold-conjugated secondary(c) antibodies. Note z-direction. Nadrincurvature-dependent. 2 puncta occurred on (a) top 3T3 of individualcells cultured puncta. on However,Nanocones the increasedelectron density micrographs of gold particlesshow accumulation on inward curved of immunogold membranes on N-BAR domain werepuncta transfected appeared not with on the all membranefull length indentations.version of Nadrin2 (b) (red) top of particlesNanocones directed (red arrows) against compared the isolated to the N-BARcytosolic domain background of (blue Correlation analysisand thefor full membrane length nadrin marker 2 (red) CAAX and (green). the membrane For better marker arrows),Nadrin2 indicative to inward of a curvature-sensing curved plasma recruitment membranes of theformed N-BAR by domain CAAX (green). Note,visualization, no significant a magnified drop in correlation section (white is evident box) for of nadrinthe to suchNanocones. sites. Scale 3T3bars; cells (a), 1were µm; (c),transfected 200nm. with a fluorescently individual channels is shown to the right. Note non-uniform tagged N-BAR domain of Nadrin2 and plated on fluorescent intensity for the membrane marker, likely Nanocones. Fluorescently tagged N-BAR domain of reflecting individual or groups of Nanocones that deform the Nadrin2 is detected with polyclonal anti-GFP antibody and plasma membrane, causing membrane accumulation in the visualized with gold-conjugated secondary antibodies. Note z-direction. Nadrin2 puncta occurred on top of individual the increased density of gold particles on inward curved puncta. However, N-BAR domain puncta appeared not on membranes on top of Nanocones (red arrows) compared to

WWW.NATURE.COM/NATURECELLBIOLOGYall membrane indentations. (b) Correlation analysis for full the cytosolic background (blue arrows) is indicative of a 1 length Nadrin2 (red) and the membrane© marker 2012 Macmillan CAAX Publishers Limited.curvature-sensing All rights reserved. recruitment of the N-BAR domain to such (green). Note, no significant drop in correlation is evident for sites. Scale bars; (a), 1 µm; (c), 200nm.

SUPPLEMENTARY INFORMATION

Figure S2 N-BAR domain proteins recruit to Nanocone-triggered inward two channels before and after a lateral shift by 10 pixel (=1.6µm). Error curved sections of the PM independently of clathrin-dependent endocytotic bars represent s.e.m. for the 125 individual rows (y-axis) of the heatmap events. (a) Nanocones do not increase the average endocytosis rate. shown in panels (d) and (e). Note the significant drop in correlation for FM1-43 uptakeFigure assay S2. shows N-BAR that thedomain average proteins fluorescent recruit intensity to Nanocone- within nadrinchannels 2 vs. before nadrin and2 (red) after but a notlateral for nadrinshift by 2 10vs. pixelmembrane (white) and the same celltriggered (left graph) inward was reducedcurved sections28+/-8% of by the the PM Nanocones. independently The of nadrin(=1.6µm). 2 vs. clathrinError bars light represent chain (gray). s.e.m. (g) for N-BAR the 125 domain individual and clathrin light average fluorescentclathrin-dependent intensity of cells endocytotic grown on events. Nanocones (a) (right Nanocones graph) do chainrows puncta (y-axis) on of Nanocones the heatmap remain shown separate in panels over time.(d) and 3T3 (e). cell plated on was 27+/-5%not less increase than for the cells average grown on endocytosis glass. Nanocone-dependent rate. FM1-43 uptake NanoconesNote the significant were transfected drop in with correlation fluorescently for nadrin tagged 2 N-BARvs. domain of change in intensityassay showswithin athat single the cell average was measured fluorescent for 20intensity cells (left within nadrinnadrin 2 2(red) (red) and but clathrin not for lightnadrin chain 2 vs. (green) membrane and imaged (white) every and 5 second graph). Averagethe sameintensities cell (leftwere graph) measured was for reduced 44 cells 28+/-8% on glass byand the 60 fornadrin 3 minutes. 2 vs. clathrin Kymograph light depictingchain (gray). traces (g) of individualN-BAR domain puncta are shown cells on NanoconesNanocones. (right Thegraph). average Error bars fluorescent represent intensity s.e.m. of of the cells mean below.and clathrin (h) In lightthe absence chain puncta of Nanocones, on Nanocones the N-BAR remain domain of nadrin 2 value. (b, c)grown Endocytotic on Nanocones proteins are (right enriched graph) over was Nanocones. 27+/-5% less 3T3 than cells doesseparate not colocalize over time. with 3T3 markers cell plated for clathrin-dependent on Nanocones wereendocytosis at the over-expressingfor cellsfluorescently grown on tagged glass. clathrin Nanocone-dependent light chain (b) or dynaminchange in leadingtransfected edge ofwith migrating fluorescently 3T3 cells. tagged Cells N-BAR were co-transfected domain of with the (c) were grownintensity on 3µm within wide stripesa single of cell Nanocones. was measured Both proteins for 20 cellswere (left N-BARnadrin domain 2 (red) ofand nadrin clathrin 2 (red) light and chain clathrin (green) light and chain imaged (green), a marker for enriched overgraph). stripes ofAverage Nanocones. intensities For better were visualization, measured for a magnified44 cells on clathrin-dependentevery 5 second for endocytosis 3 minutes. and Kymograph imaged 24h depicting later on atraces TIRF microscope. section is shownglass to and the 60right cells (white on box).Nanocones Note, that (right dynamin-aggregates graph). Error bars Theof individualN-BAR domain puncta did are not shown colocalize below. with (h)endocytic In the sitesabsence at the leading on Nanoconesrepresent formed tubular s.e.m. structuresof the mean (red value. box). (b,(d) c) Correlation Endocytotic analysis edge.of Nanocones, (i) In the absencethe N-BAR of Nanocones, domain of thenadrin N-BAR 2 does domain not of nadrin 2 is for full lengthproteins nadrin 2are (red) enriched with clathrin over Nanocones. light chain (Clc, 3T3 green) cells shows low transientlycolocalize recruited with markers to endocytic for clathrin-dependent sites at the center endocytosis of the cell. Cells were colocalization.over-expressing Single channels fluorescently for the inset tagged(white) areclathrin shown light below. chain (e) (b) imagedat the leadingon a TIRF edge for 3of minutes. migrating Kymograph 3T3 cells. shows Cells pulsatile were appearance Control correlationor dynamin analysis (c) for were CFP-tagged grown on full 3µm length wide nadrin stripes 2 (red)of with ofco-transfected the isolated N-BAR with thedomain N-BAR of nadrin domain 2 (red)of nadrin to these 2 (red) sites, and indicating YFP-tagged fullNanocones. length nadrin Both 2 (green).proteins Magnificationwere enriched of over single stripes channels of transientclathrin formationlight chain of (green), curved plasmaa marker membrane for clathrin-dependent sites by clathrin light chain are shown below.Nanocones. (f) Panel For depicting better visualization,difference in correlation-score a magnified section for the is (green).endocytosis Scale andbars imaged(a, g, h, 24hi), 10 later µm; on (b, a c), TIRF 3 µm; microscope. (d, e), 1 µm. shown to the right (white box). Note, that dynamin- The N-BAR domain did not colocalize with endocytic sites aggregates on Nanocones formed tubular structures (red at the leading edge. (i) In the absence of Nanocones, the box). (d) Correlation analysis for full length nadrin 2 (red) N-BAR domain of nadrin 2 is transiently recruited to 2 with clathrin light chain (Clc, green) shows low colocaliza- endocytic sites at the centerWWW.NATURE.COM/NATURECELLBIOLOGY of the cell. Cells were imaged tion. Single channels for the inset (white) are© 201 shown2 Macmillan below. Publishers on Limited. a TIRF All for rights 3 minutes. reserved. Kymograph shows pulsatile (e) Control correlation analysis for CFP-tagged full length appearance of the isolated N-BAR domain of nadrin 2 (red) nadrin 2 (red) with YFP-tagged full length nadrin 2 (green). to these sites, indicating transient formation of curved Magnification of single channels are shown below. (f) plasma membrane sites by clathrin light chain (green). Panel depicting difference in correlation-score for the two Scale bars (a, g, h, i), 10 µm; (b, c), 3 µm; (d, e), 1 µm.

SUPPLEMENTARY INFORMATION

Figure S3. Rapamycin-induced actin polymerization of rapamycin-induced translocation of the isolated N-BAR Figure triggeredS3 Rapamycin-induced BAR domain actin puncta polymerization formation triggered in 3T3 cells.BAR domain (a) Additiondomain of of rapamycin-induced nadrin1. (f) No translocation of isthe observed isolated N-BAR when domain puncta3T3 formation cells werein 3T3 co-transfected cells. (a) 3T3 cells with were a membrane-anchored co-transfected with a ofthe nadrin1. amphipatic (f) No helix, translocation encoded is observedby the first when 18 the amino amphipatic acids helix,of membrane-anchoredFRB, as well FRB,as with as well fluorescently as with fluorescently and FKBP-tagged and FKBP-tagged encodednadrin1 by is the deleted. first 18 (g)amino Addition acids of of nadrin1 rapamycin is deleted. does (g) not Addition of Rho-GEFRho-GEF TIAM1 and TIAM1 the N-BAR and the domain N-BAR of nadrin domain 2. Additionof nadrin of 2.rapamycin rapamycininduce translocation does not induce of translocationthe BAR domain of the proteinBAR domain GMIP. protein (h) GMIP. triggeredAddition hetero-dimerization of rapamycin of FRBtriggered and FKBP hetero-dimerization triggering rapid translocation of FRB (h)Addition Addition of ofrapamycin-induced rapamycin-induced translocation translocation of theof the F-BAR F-BAR domain of FKBP-TIAM1and FKBP to thetriggering PM (blue, rapid left translocationpanels). With a of delay, FKBP-TIAM1 this triggered to proteindomain TRIP10. protein (i) TRIP10. No translocation (i) No translocationis observed when is observedthe isolated F-BAR formationthe of PM RAC-GTP, (blue, leftcausing panels). excessive With lamellipodia a delay, this formation triggered and domainwhen theof TRIP10, isolated encoded F-BAR by domain the first of 298 TRIP10, amino acids encoded is expressed by translocationformation of the of N-BAR RAC-GTP, domain causing of nadrin excessive 2 (orange, rightlamellipodia graph). (b) separately.the first 298 (j) aminoAs for TRIP10, acids is no expressed translocation separately. is observed (j)for theAs isolatedfor No translocationformation is and observed translocation when the amphipaticof the N-BAR helix, domain encoded of by nadrin the F-BARTRIP10, domain no translocationof FCHO2 (AA1-274). is observed (k) Addition for the ofisolated rapamycin-induced F-BAR first 18 amino acids of the N-BAR domain of nadrin 2 is deleted. (c) translocation of full length BAIAP2. (l) No translocation is observed when 2 (orange, right graph). (b) No translocation is observed domain of FCHO2 (AA1-274). (k) Addition of rapamycin- Addition of rapamycin-induced translocation of the isolated N-BAR domain the isolated I-BAR domain of BAIAP2, encoded by the amino acids 1-252 when the amphipatic helix, encoded by the first 18 amino induced translocation of full length BAIAP2. (l) No of amphiphysin 1. (d) No translocation is observed when the amphipatic is expressed. Scale bars, 10µm. Error bars represent s.e.m. of the mean helix ofacids amphiphysin of the N-BAR 1, encoded domain by the of first nadrin 28 amino 2 is deleted.acids is deleted. (c) (e) value.translocation **P < 0.01. is observed when the isolated I-BAR domain Addition of rapamycin-induced translocation of the isolated of BAIAP2, encoded by the amino acids 1-252 is N-BAR domain of amphiphysin 1. (d) No translocation is expressed. Scale bars, 10µm. Error bars represent s.e.m. observed when the amphipatic helix of amphiphysin 1, of the mean value. **P < 0.01. WWW.NATURE.COM/NATURECELLBIOLOGYencoded by the first 28 amino acids is deleted. (e) Addition 3 © 2012 Macmillan Publishers Limited. All rights reserved.

SUPPLEMENTARY INFORMATION

Figure S4 Induction of actin-dynamics caused recruitment of N-BAR domain RAC1 (bottom) are shown next to the images. Note that LOV-RAC1 strongly proteins to the plasma membrane. (a) Scheme of the LOV-RAC1 activation enriched at the nuclear membrane while no N-BAR domain puncta located mechanism. Exposure of blue light induces a conformational change of the to this structures. At the plasma membrane, LOV-RAC1 was uniformly LOV domain which exposes the constitutive active form of RAC1. Activity of distributed while the N-BAR domain of nadrin 2 appeared in puncta facing RAC1 ends when the LOV domain returns to its native state. (b) The LOV- the cytosol. Together, these data suggest that translocation of the N-BAR RAC1 (green) was activated with a 100ms pulse of 442nm light. Within domain is not due to direct interaction with the RAC1 upon light-activation. 15 seconds the isolated N-BAR domain of nadrin 2 (red) was recruited (c) Analysis of the observed increase in puncta density is shown to the in punctate to the plasma membrane. For better illustration, the images right for the isolated N-BAR domain of nadrin 2 (top) and amphiphysin 1 were background-subtracted. As a reference, original images depicting the (bottom). Error bars represent s.e.m. of the mean value. Scale bar, 10µm. N-BAR domain or nadrin 2 (top), a cytosolic reference (middle) and LOV- ** P < 0.01. Figure S4. Induction of actin-dynamics caused recruitment LOV-RAC1 (bottom) are shown next to the images. Note of N-BAR domain proteins to the plasma membrane. (a) that LOV-RAC1 strongly enriched at the nuclear membrane Scheme of the LOV-RAC1 activation mechanism. Expo- while no N-BAR domain puncta located to this structures. At sure of blue light induces a conformational change of the the plasma membrane, LOV-RAC1 was uniformly distrib- LOV domain which exposes the constitutive active form of uted while the N-BAR domain of nadrin 2 appeared in RAC1. Activity of RAC1 ends when the LOV domain returns puncta facing the cytosol. Together, these data suggest that to its native state. (b) The LOV-RAC1 (green) was translocation of the N-BAR domain is not due to direct activated with a 100ms pulse of 442nm light. Within 15 interaction with the RAC1 upon light-activation. (c) seconds the isolated N-BAR domain of nadrin 2 (red) was Analysis of the observed increase in puncta density is 4 recruited in punctate to the plasma membrane. For better shown to the right for the isolatedWWW.NATURE.COM/NATURECELLBIOLOGY N-BAR domain of nadrin illustration, the images were background-subtracted. As a 2 (top) and amphiphysin 1 (bottom). Error bars represent reference, original images depicting the N-BAR© 201 domain2 Macmillan or Publisherss.e.m. Limited. of theAll rights mean reserved. value. Scale bar, 10µm. ** P < 0.01. nadrin 2 (top), a cytosolic reference (middle) and

SUPPLEMENTARY INFORMATION

Figure S5. Actin cables pulling at the plasma membrane recruit curvature- domain puncta in COS cells. Cells were transfected with YFP tagged N-BAR sensing N-BARFigure domain S5. proteins. Actin cables (a) Transmitted pulling at electronthe plasma micrographs membrane of 3T3 domainpuncta of in nadrin COS 2cells. (top) Cellsand amphiphysin were transfected 1 (bottom). with YFPAfter replacement of cells cut parallelrecruit to the curvature-sensing surface and contrast N-BAR enhanced domain with proteins. osmium tetroxide. (a) regulartagged medium N-BAR (DMEM+10%PBS; domain of nadrin 310+/-10 2 (top) and mOsm/kg) amphiphysin with pre-warmed 1 One sample withTransmitted no actin filamentselectron visiblemicrographs (top, left) of and3T3 seven cells samplescut parallel with to hypo-osmotic(bottom). After medium replacement (1:6 dilution of regularof medium:ddH2O; medium 50 mOsm/kg) the actin filamentsthe extending surface andto the contrast PM are shown.enhanced Note with that osmium actin filaments tetroxide. are N-BAR(DMEM+10%PBS; domain puncta 310+/-10 disappeared mOsm/kg) within 200 with seconds. pre-warmed (d) Hyper-osmotic oriented perpendicularOne sample to the with plasma no actin membrane filaments ending visible close (top, to deformed left) and shockhypo-osmotic increased mediumN-BAR domain (1:6 dilution puncta ofdensity. medium:ddH2O; Addition of hyper-osmotic 50 section of theseven plasma samples membrane. with (b)actin Fluorescently filaments extending tagged N-BAR to the domain PM are mediummOsm/kg) (regular the mediumN-BAR domainsupplemented puncta with disappeared 1% 10x PBS; within 340+/-10 mOsm/ of nadrin 2 isshown. detected Note with thatpolyclonal actin filamentsanti-GFP antibody are oriented and visualized perpendicular with kg)200 triggered seconds. an increase (d) Hyper-osmotic in N-BAR domain shock puncta increased number N-BAR for nadrin 2 (top) gold-conjugatedto the secondary plasma antibodies. membrane Note ending the high close density to deformed of gold particles section anddomain amphiphysin puncta 1density. (bottom) Addition within 100 of hyper-osmoticseconds. (e) . Membrane medium relaxation Figure S4. Induction of actin-dynamics caused recruitment LOV-RAC1 (bottom) are shown next to the images. Note associate withof curved the plasma regions membrane. (red arrows) of(b) the Fluorescently plasma membrane tagged compared N-BAR caused(regular rapid medium increase supplemented in N-BAR domain with puncta1% 10x number. PBS; 340+/-10 N-BAR domain of N-BAR domain proteins to the plasma membrane. (a) that LOV-RAC1 strongly enriched at the nuclear membrane to the low cytosolicdomain background of nadrin (blue2 is detected arrows). with(c) Membrane polyclonal tension anti-GFP controls punctamOsm/kg) number triggered of nadrin an 2 increaseand amphiphysin in N-BAR 1 increased domain punctawithin 60 seconds Scheme of the LOV-RAC1 activation mechanism. Expo- while no N-BAR domain puncta located to this structures. At N-BAR domainantibody puncta andformation. visualized Hypo-osmotic with gold-conjugated shock dissolved secondary N-BAR afternumber addition for nadrinof 400 2µM (top) Deoxycholate. and amphiphysin Scale bars, 1 (bottom) (a, b), 200nm; within (c-e), 10µm. sure of blue light induces a conformational change of the the plasma membrane, LOV-RAC1 was uniformly distrib- antibodies. Note the high density of gold particles associate 100 seconds. (e) . Membrane relaxation caused rapid LOV domain which exposes the constitutive active form of uted while the N-BAR domain of nadrin 2 appeared in with curved regions (red arrows) of the plasma membrane increase in N-BAR domain puncta number. N-BAR domain RAC1. Activity of RAC1 ends when the LOV domain returns puncta facing the cytosol. Together, these data suggest that compared to the low cytosolic background (blue arrows). puncta number of nadrin 2 and amphiphysin 1 increased to its native state. (b) The LOV-RAC1 (green) was translocation of the N-BAR domain is not due to direct (c) Membrane tension controls N-BAR domain puncta within 60 seconds after addition of 400 µM Deoxycholate. activated with a 100ms pulse of 442nm light. Within 15 interaction with the RAC1 upon light-activation. (c) formation. Hypo-osmotic shock dissolved N-BAR domain Scale bars, (a, b), 200nm; (c-e), 10µm. seconds the isolated N-BAR domain of nadrin 2 (red) was Analysis of the observed increase in puncta density is recruited in punctate to the plasma membrane. For better shown to the right for the isolated N-BAR domain of nadrin WWW.NATURE.COM/NATURECELLBIOLOGY 5 illustration, the images were background-subtracted. As a 2 (top) and amphiphysin 1 (bottom). Error bars represent reference, original images depicting the N-BAR domain or s.e.m. of the mean value. Scale bar, 10µm. ** P < 0.01. © 2012 Macmillan Publishers Limited. All rights reserved. nadrin 2 (top), a cytosolic reference (middle) and