ARF1-mediated actin polymerization producesmovement of artificial vesiclesJulien Heuvingh*†‡§¶, Michel Franco�, Philippe Chavrier*§**, and Cecile Sykes*†‡**

*Centre de Recherche, Institut Curie, F-75248 Paris, France; Unites Mixte de Recherche †168 and §144, Centre National de la Recherche Scientifique, F-75248Paris, France; ‡Universite Paris VI, F-75248 Paris, France; and �Institut de Pharmacologie Moleculaire et Cellulaire, Unite Mixte de Recherche 6097, CentreNational de la Recherche Scientifique, F-06560 Valbonne, France

Edited by Thomas D. Pollard, Yale University, New Haven, CT, and approved September 6, 2007 (received for review May 21, 2007)

Vesicular trafficking and actin dynamics on Golgi membranes areboth regulated by ADP-ribosylation factor 1 (ARF1) through therecruitment of various effectors, including vesicular coats. Actinassembly on Golgi membranes contributes to the architecture ofthe Golgi complex, vesicle formation, and trafficking and is medi-ated by ARF1 through a cascade that leads to Arp2/3 complexactivation. Here we addressed the role of Golgi actin downstreamof ARF1 by using a biomimetic assay consisting of liposomes ofdefined lipid composition, carrying an activated form of ARF1incubated in cytosolic cell extracts. We observed actin polymeriza-tion around the liposomes resulting in thick actin shells and actincomet tails that pushed the ARF1 liposomes forward. The assaywas used to characterize the ARF1-dependent pathway, leading toactin polymerization, and confirmed a dependency on CDC42 andits downstream effector N-WASP. Overall, this study demonstratesthat actin polymerization driven by the complex multicomponentsignaling cascade of the Golgi apparatus can be reproduced witha biomimetic system. Moreover, our results are consistent with theview that actin-based force generation at the site of vesicleformation contributes to the mechanism of fission. In addition toits well established function in coat recruitment, the ARF1 machin-ery also might produce movement- and fission-promoting forcesthrough actin polymerization.

transport vesicle � Golgi � membrane scission � motility � CDC42

Intracellular traffic is mediated by transport vesicles that budfrom a donor membrane through the assembly of specific coat

proteins and are then transported and fused with the acceptororganelle (1, 2). One important, yet not fully understood, step ofvesicle formation is the fission and separation of the transportcarrier from the donor compartment. Extensive studies mostlyfocusing on the formation of clathrin-coated vesicles duringendocytosis from the plasma membrane revealed the facilitatingrole of actin polymerization for the pinching off of vesicles andtheir movement away from the donor membrane (3). In thesecretory pathway, it is not yet clear whether actin fulfills thesame role. However, fundamental similarities between the reg-ulation of actin assembly at the plasma membrane and on Golgisubcompartments have been uncovered (4, 5). Besides its im-plication in the maintenance of Golgi architecture, the actincytoskeleton might play a role in the biogenesis of Golgi-derivedtransport vesicles (6–8), just as actin dynamics is coupled to theformation of clathrin-coated vesicles and cargo exit from thetrans-Golgi network (TGN) (9, 10) and subsequent trafficking inthe early secretory pathway (11–13).

Roles for Golgi coat proteins, namely coat protein I (COPI)and clathrin coats, in the control of actin cytoskeleton dynamicshave been documented, connecting the mechanism of vesicleformation to the actin cytoskeleton (9, 11, 12). Recruitment ofCOPI and clathrin coat–protein complexes on Golgi subcom-partments is regulated by the small GTP-binding protein ADP-ribosylation factor 1 (ARF1), and ARF1 has been shown toregulate actin dynamics on Golgi membranes (14). ARF1-mediated recruitment of COPI on cis-medial Golgi compart-

ments triggers Arp2/3 complex-dependent actin polymerizationin a cascade that involves the Rho GTPase CDC42 and itsdownstream effector, N-WASP (11, 12, 15, 16). Although notcompletely understood, the mechanism of CDC42 association toGolgi membranes may involve the binding of CDC42 to theCOPI �-subunit (15). In addition, it was recently found thatGTP-ARF1 on the Golgi complex recruits a CDC42 GTPase-activating protein (GAP), ARHGAP21, which further regulatesCDC42 activity and Arp2/3 complex dynamics on Golgi mem-branes (17, 18). Whether and how ARF1-mediated assembly ofclathrin-coated vesicles is coupled to actin dynamics on the TGNis less clear. It was recently reported that activated ARF1 triggersthe recruitment of a cortactin/dynamin-2 complex to Golgimembranes with consequences for post-Golgi transport (10).This pathway downstream of ARF1 would couple Arp2/3 com-plex activation by cortactin to the stimulation of vesicle fission bydynamin-2 (10). In addition, ARF1 is known to promote theproduction of phosphoinositides, PtdIns 4-phosphate (PI4P),and PtdIns 4,5-biphosphate (PI4,5P2) at the Golgi complex and,hence, to affect both actin dynamics and membrane traffic (19).These observations suggest that ARF1-regulated actin dynamicscontribute to the trafficking of Golgi-derived vesicles by facili-tating the formation and possibly the dissociation of transportcarriers from Golgi subcompartments.

Here we designed an in vitro reconstituted system analyzingthe role of ARF1-dependent actin dynamics on liposomes. Thissystem is composed of lipid vesicles of defined composition, towhich ARF1 is bound. GDP-ARF1 is soluble, but it binds tomembrane in its GTP-bound form by its myristoylated amphi-pathic helix, allowing liposomes to be readily functionalized byGTP-ARF1 in vitro. In practice, GTP-�-S, a nonhydrolyzableanalog of GTP, was used to stabilize the association of ARF1 toliposomes. These liposomes were incubated in HeLa cell ex-tracts, as opposed to purified protein mixes, because the requiredcomponents of ARF1-dependent actin dynamics are not yet fullyidentified. Under these conditions, liposomes promoted actinassembly at their surface in an ARF1-dependent manner andmoved by the polymerization of an actin comet tail. This assayallowed for the characterization of the biochemical pathway

Author contributions: P.C. and C.S. contributed equally to this work; P.C. and C.S. designedresearch; J.H. performed research; M.F. contributed new reagents/analytic tools; J.H.analyzed data; and J.H., P.C., and C.S. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Abbreviations: GUV, giant unilamellar vesicle; LUV, large unilamellar vesicle; TGN, trans-Golgi network.

¶Present address: Universite Paris Diderot and Physique et Mecanique des Milieux Hetero-genes, Ecole Superieure de Physique et de Chimie Industrielles, F-75231 Paris, France.

**To whom correspondence may be addressed. E-mail: philippe.chavrier@curie.fr or cecile.sykes@curie.fr.

This article contains supporting information online at www.pnas.org/cgi/content/full/0704749104/DC1.

© 2007 by The National Academy of Sciences of the USA

16928–16933 � PNAS � October 23, 2007 � vol. 104 � no. 43 www.pnas.org�cgi�doi�10.1073�pnas.0704749104

Dow

nloa

ded

by g

uest

on

July

20,

202

1

downstream of ARF1, leading to actin polymerization andmovement.

ResultsARF1 Liposomes. Liposomes were prepared from a mixture ofsynthetic lipids for a simple and well controlled system. Unlessotherwise indicated, they consisted of 1,2-dioleoyl-sn-glycero-3-phosphatidylcholine (DOPC) with 1% 1,2-dioleoyl-sn-glycero-3-phosphatidylinositol 4,5-biphosphate (PIP2) because PIP2 isknown to regulate the architecture of the Golgi complex down-stream of ARF1 (20, 21). Liposomes of different sizes wereobtained by using two different techniques. Giant unilamellarvesicles (GUVs), 1 to 20 �m in diameter, formed by the gentleswelling method (22) were used for microscopy observations andquantification of actin polymerization by epifluorescence. Largeunilamellar vesicles (LUVs), 0.8 �m in diameter, formed by theextrusion method (23) were generally used for biochemicalcharacterization. To bind ARF1 to liposomes, we made use of itsnatural property to insert into membranes upon GTP loading(24). For this purpose, recombinant-purified, myristoylatedGDP-ARF1 was loaded with the nonhydrolyzable analog ofGTP, GTP-�-S, at a low-Mg2� concentration in the presence ofliposomes, and insertion was verified by sedimentation analysis[see supporting information (SI) Fig. 4].

GTP-ARF1 Induces Actin Shell and Comet Formation. GUVs loadedwith GTP-�-S-ARF1 (ARF1-GUVs) were incubated in HeLacell extract supplemented with G-actin (including 5% fluores-cently labeled actin) and ATP. An actin shell, visible as a darkring by phase-contrast microscopy or a fluorescent ring byepif luorescence microscopy, was detected around the liposomesas early as 5 min after adding liposomes to the extract. Thethickness of the actin gel increased progressively around theGUVs as a function of time (Fig. 1 A–C). In some cases, after2 to 3 h, the actin gel growth led to the disappearance of thevesicle, producing a homogeneous ball of actin gel (Fig. 1D).Because �10 �M residual-free GTP-�-S was present in theliposome preparation, a control was performed by using GUVsthat were preincubated in GTP-�-S without ARF1. These GUVsnever developed an actin gel (Fig. 1 E and F) (the exposure ofthe fluorescent images is enhanced 10 times compared with Fig.1 A–D), although faint patches of fluorescent actin could be seenat the periphery of some of these GUVs (Fig. 1F). To quantifythe extent of actin polymerization around ARF1-GUVs, themean actin fluorescence intensity in the fluorescent ring aroundrandomly selected GUVs was measured 30–45 min after mixingwith the extract, as described in Material and Methods (Fig. 2).Then �95% of ARF1-GUVs displayed a pronounced fluores-cent signal (Fig. 2 A1), contrasting with control GUVs preincu-bated with GTP-�-S without ARF1 (Fig. 2 A2). We conclude thatthe residual GTP-�-S left in the assay is not sufficient to triggeractin polymerization in the absence of ARF1.

Actin comets visible by phase-contrast microscopy were ob-served on 18% of ARF1-GUVs (66 comets for 376 GUVs) andcorrelated with movement of these GUVs (Fig. 1G). Of note,comets were never observed for GUVs of �2 �m in diameter, anobservation that is consistent with experiments on beads (25). Theaverage speed of rocketing liposomes was 0.52 � 0.30 �m/min (n �20), comparable to Arp2/3-mediated movement in other in vitrosystems (26). Actin comet-based movement was similarly observedwith LUVs, on which ARF1 was bound (Fig. 1H).

To confirm that the formation of actin shells and comets wasthe result of de novo barbed-end actin polymerization, and notbecause of recruitment of preexisting actin filaments from theextract, the assay was carried out in the presence of 0.1 to 2 �Mcytochalasin D, which is an inhibitor of actin polymerization atbarbed ends in this concentration range (27). The amount ofactin present around ARF1-GUVs was significantly reduced in

the presence of the drug at �0.5 �M (P � 0.0001) (Fig. 2B). Thisinhibition, together with the observed propulsion of small lipo-somes by actin tails, showed that barbed-end actin polymeriza-tion, and not recruitment of preexisting filaments, was occurringaround ARF1 liposomes, and that actin dynamics were able togenerate a propulsive force.

To address whether the presence of PIP2 was important forthe production of actin at the membrane, we used ARF1-GUVscomposed of 96% DOPC and 4% phosphatidylinositol (PI).Thus, these vesicles lacked PIP2, but bore the same electrostaticcharge as the standard 1% PIP2-containing ARF1-GUVs. WhenDOPC-PI liposomes were mixed with HeLa cell extracts, weobserved a shell of actin, although it was markedly reducedcompared with standard 1% PIP2-containing ARF1-GUVs(compare Fig. 2 A3 and A1), demonstrating that PIP2 is requiredfor optimal ARF1-dependent polymerization of actin aroundliposomes. When standard (1% PIP2) ARF1-GUVs were incu-bated in buffer containing G-actin and ATP without extract, no

Fig. 1. Actin polymerizing around ARF1 liposomes produces actin shells andcomets. (A–F) For all image pairs, the left image is a phase-contrast image, andthe right is fluorescence microscopy showing labeled actin. The expositiontime for all fluorescence images is the same, except for E and F, whereexposition is 10 times longer. (Scale bar: 5 �m.) (A and B) GUV loaded withGTP-�-S-bound ARF1 incubated for 30 min in cytosolic extract. A shell of actinis visible around the GUVs. (C and D) ARF1-GUV incubated for 2 h in extract.A wide actin shell is observable around the GUV in C, whereas a ball of actinwith no enclosed volume is shown in D. (E and F) GUV without ARF1 incubatedfor 30 min in supplemented extracts. Small faint patches of actin are visible inF. (G and H) The elapsed time between each image is 30 sec. (Scale bars: 2 �m.)(G) ARF1-GUV propelled by an actin comet. All images are phase-contrastmicroscopy, except for the first image, which is fluorescent microscopy ofactin. (H) ARF1-LUV propelled by an actin comet. All images are phase-contrastmicroscopy, except for the last image, which is fluorescent microscopy of actin(red) and lipids (green).

Heuvingh et al. PNAS � October 23, 2007 � vol. 104 � no. 43 � 16929

CELL

BIO

LOG

Y

Dow

nloa

ded

by g

uest

on

July

20,

202

1

actin polymerization was observed around the liposomes (datanot shown).

Thus far, these results indicate that the presence of activatedARF1 on synthetic liposome membranes triggered actin poly-merization and movement. To provide additional evidence forthe dependency of ARF1 in our system, we made use of ourrecent finding that ARHGAP21, a CDC42 GAP regulating actindynamics on Golgi, associates with Golgi membranes throughthe interaction of its central ARF-binding domain (ARFBD)with GTP-bound ARF1 (17). Addition of 0.1 or 1 �M purifiedrecombinant GST-tagged ARFBD to the assay resulted in adose-dependent inhibition of actin polymerization on ARF1-GUVs, compared with 1 �M GST alone (Fig. 2C). Our inter-pretation of these data are that ARHGAP21-ARFBD, whichbinds GTP-bound ARF1 with a high affinity (Kd � 50 nM) (18),prevents further activation of effector pathways involved in actinpolymerization downstream of ARF1.

Overall, our data demonstrate the absolute requirement ofGTP-ARF1 for actin polymerization around liposomes and thenecessity of cytosolic factors, most probably proteins, to supportactin polymerization downstream of ARF1.

ARF1-Dependent Actin Polymerization Requires CDC42 and N-WASP.Although the machinery regulating actin dynamics on Golgimembranes is far from understood, it is now clear that ARF1plays a critical role in Golgi actin assembly through the controlof a cascade leading to CDC42/N-WASP/Arp2/3 complex acti-vation (11, 12, 28). We checked that this activation was at workin our assay by immunoblotting for CDC42 on liposomes sedi-

mented by centrifugation. This experiment revealed that, com-pared with naked LUVs, ARF1-LUVs recruited significantlyhigher amounts of CDC42 from the cytosolic extract, and thisfinding correlated with increased association of F-actin withLUVs (SI Fig. 5).

Because CDC42 has been shown to play a role in actinpolymerization downstream of ARF1 on isolated Golgi mem-branes (11, 12), we characterized the role of CDC42 in ourARF1-liposome assay as follows. Specific inhibitors of theCDC42 cascade were added to the polymerization assay, and theactin shell around ARF1-GUVs was quantified as before. Se-cramine, a recently described specific inhibitor of CDC42 shownto block traffic of secreted proteins out of the Golgi complex(29), was first added to the assay at concentrations ranging from1 to 20 �M. The addition of 1 �M secramine was sufficient tosignificantly reduce ARF1-dependent actin polymerization (P �0.006) (Fig. 3A). In addition, the CDC42/Rac interactive binding(CRIB) region of PAK-1, which binds GTP-bound CDC42 (andRac1) (30), was purified as a GST-fusion protein and added tothe assay at concentrations ranging from 0.25 to 2 �M. Signif-icant inhibition of actin polymerization around ARF1-GUVswas already observed at the lowest concentration of PAK-CRIBdomain used (0.25 �M) (Fig. 3B). These results revealed thatGTP-bound ARF1 on liposomes was able to trigger a cascadeleading to CDC42 activation and, hence, to actin polymerization.

In its GTP-bound state, CDC42 binds directly to N-WASP,relieving N-WASP autoinhibition and allowing N-WASP toactivate the actin polymerization nucleating capacity of theArp2/3 complex (31). Wiskostatin was recently characterized as

23 46

821

652

215

4201

8402

6904

2918

fluorescence intensity (a.u.)

Cytochalasin D

10

40

20

30

0

sem

oso

pil fo

%

Controls

50

10

80

20

30

0

10

40

20

30

0

sem

oso

pil fo

%se

mo s

opil f

o %

sem

oso

pil fo

%s e

mos

opi l f

o %

sem

oso

pil fo

%

0µMn=27

0.5µMn=34

p<0.0001

2µMn=29

p<0.0001

n=67

no pip2n=66

p<0.0001

no ARF1n=47

p<0.0001

90

% o

f lip

oso

mes

% o

f lip

oso

mes

ARFBD

% o

f lip

oso

mes

0µMn=40

10

40

20

30

0

50

60

0.1µMn=85

p<0.0001

1µMn=39

p<0.0001

10

40

20

30

0

50

10

40

20

30

0

10

40

20

30

0

10

40

20

30

0

50

10

40

20

30

0

50

23 46

821

652

215

4201

8402

6904

2918

fluorescence intensity (a.u.)

23 46

821

652

215

4201

8402

6904

2918

fluorescence intensity (a.u.)

23 46 821

652

215

4201

8402

6904

2918fluorescence intensity (a.u.)

23 46 821

652

215

4201

8402

6904

2918

fluorescence intensity (a.u.)

23 46 821

652

215

4201

8402

6904

2918

fluorescence intensity (a.u.)

A B C

Fig. 2. Quantification of actin polymerization around ARF1-GUV. Quantification of actin polymerization around GUV after 30–45 min of incubation insupplemented extracts. The histogram shows the percentage of GUVs as a function of circumferential mean of fluorescence arranged by exponential bin size(i.e., the fluorescence value in a column is twice the value of the column on its left and half the value of the column on its right). (A Top) ARF1-GUV (1% PIP2,99% DOPC). (A Middle) GUV (1% PIP2, 99% DOPC) with no ARF1. (A Bottom) ARF1-GUV (4% PI, 96% DOPC); statistical comparison with A Top is provided. (Band C) ARF1-GUV (1% PIP2, 99% DOPC) incubated in supplemented extracts with increasing concentrations of drugs or proteins. Statistical comparison with thepreceding concentration is provided. (B) Incubation with increasing concentrations of cytochalasin D. Concentration of drug vehicle (DMSO) is kept constant(1‰). (C) Incubation with increasing concentrations of the ARFBD fragment of the ARHGAP21 protein.

16930 � www.pnas.org�cgi�doi�10.1073�pnas.0704749104 Heuvingh et al.

Dow

nloa

ded

by g

uest

on

July

20,

202

1

a chemical inhibitor of N-WASP, which stabilizes N-WASP in itsautoinhibited conformation insensitive to activation by CDC42and PIP2 (32). Addition of 10 �M Wiskostatin significantlyinhibited actin polymerization at the surface of ARF1-GUVs(P � 0.0001) (Fig. 3C), indicating that ARF1 acts upstream ofa well established CDC42/N-WASP/Arp2/3 cascade (33), leadingto actin polymerization.

These findings demonstrate that membrane-bound ARF1 iscompetent for triggering CDC42 recruitment and activity, leadingto N-WASP/Arp2/3 complex-dependent actin polymerization.

Role of Actin Dynamics in Membrane Remodeling in Vitro. Becauseour ARF1-reconstitution assay reproduced the signaling cascadedissected in cellular systems, it provided a good tool for studyingmembrane deformations induced by the presence of ARF1. Twotypes of membrane reorganization were observed: liposomescarrying a comet tail or balls of actin with no enclosed volume(see Fig. 1 D–F). Liposomes with comets were observed 30 to 45min after the start of actin polymerization and underwentmotion by actin comet growth during the 30 min. Balls of actinappeared upon longer incubation times (�2 h) and were pre-ceded by actin shell growth around the liposomes. Note that ballsof actin were in the range of several micrometers in diameter andwere thus bigger than the liposomes supporting comet growth(�2 �m in diameter).

Liposome separation (i.e., one part of the vesicle moving awayfrom the other part by the growth of an actin comet) wasobserved on ARF1-GUVs on rare occasions (see SI Movie 1).These events occurred on vesicles with an unusual oblate shapeand an asymmetric actin shell.

DiscussionOur data demonstrate that the ARF1 cascade leads to actinpolymerization by CDC42 activation and further triggering ofN-WASP and likely the Arp2/3 complex at the cytosol/membrane interface. In addition, GTP-ARF1 appears to beindispensable for actin polymerization on DOPC/PIP2 GUVs,somewhat contrasting with previous findings that liposomescontaining phosphoinositides could produce actin comets in cellextracts (34). However, it should be noted that these experimentsused liposomes with different lipid composition (equal amountsof PI and PC and 4–33% PIP2) incubated in different extracts(Xenopus eggs instead of HeLa cells) (34). Another importantfinding from our study is that the effect of several CDC42inhibitors, including the Golgi-associated, CDC42-specific GAP,ARHGAP21 (17), clearly demonstrates that activation ofCDC42 (GTP-loading) downstream of ARF1 is indispensablefor actin polymerization, and that this step can be reconstitutedat the surface of liposomes incubated in cytosolic extracts in vitro.This finding is interesting given that the mechanisms leading toCDC42 activation on Golgi subcompartments downstream ofARF1 are not understood. The COPI could provide a linkbetween ARF1 and CDC42 because it is recruited on the Golgiby binding to ARF1 and it is known to interact with GTP-boundCDC42 (11, 15). However, COPI alone is probably not directlyresponsible for CDC42 activation because replacement of GDPfor GTP on Rho proteins requires specific guanine nucleotideexchange factors (GEFs). The larger family of Rho GEFs is theDbl-related GEFs that comprise 69 distinct members in humans(35). Among these, only a few have been localized to the Golgi

Secramine CRIB Wiskostatin

0µMn=52

1µMn=38

p<0.006

20µMn=40

p<0.0001

0µMn=47

0.25µMn=43

p<0.0005

2µMn=44

p<0.0001

0µMn=31

10µMn=44

p<0.0001

20µMn=49

p<0.0001

% o

f lip

oso

mes

% o

f lip

oso

mes

% o

f lip

oso

mes

10

40

20

30

0

50

60

10

40

20

30

0

50

10

40

20

30

0

50%

of l

ipo

som

es%

of l

ipo

som

es%

of l

ipo

som

es

10

40

20

30

0

50

60

10

40

20

30

0

50

10

40

20

30

0

50

% o

f lip

oso

mes

% o

f lip

oso

mes

% o

f lip

oso

mes

10

40

20

30

0

50

60

10

40

20

30

0

50

10

40

20

30

0

50

23 46

821

652

215

4201

8402

6904

2918

fluorescence intensity (a.u.)

23 46

821

652

215

4201

8402

6904

2918

fluorescence intensity (a.u.)

23 46

821

652

215

4201

8402

6904

2918

fluorescence intensity (a.u.)

23 46 821

652

215

4201

8402

6904

2918

fluorescence intensity (a.u.)

23 46 821

652

215

4201

8402

6904

2918

fluorescence intensity (a.u.)

23 46 821

652

215

4201

8402

6904

2918

fluorescence intensity (a.u.)

A B C

Fig. 3. Inhibition of proteins of the ARF1 actin polymerization cascade. Quantification of actin polymerization around GUV after 30- to 45-min of incubationin supplemented extracts. Histograms of the radial mean of fluorescence for different GUVs arranged by bins of exponential size. ARF1-GUV (1% PIP2, 99% DOPC)incubated in supplemented extracts with different concentrations of drugs or proteins. Statistical comparison with the preceding concentration is provided. (A)Incubation with increasing concentrations of the secramine drug targeting CDC42. Concentration of drug vehicle (DMSO) was kept constant (1‰). (B) Incubationwith increasing concentrations of the PAK-CRIB domain, which binds to and competes for activated CDC42. The 2 �M concentration of GST was kept constant.(C) Incubation with increasing concentrations of Wiskostatin that targets N-WASP. Concentration of drug vehicle (DMSO) was kept constant (1‰).

Heuvingh et al. PNAS � October 23, 2007 � vol. 104 � no. 43 � 16931

CELL

BIO

LOG

Y

Dow

nloa

ded

by g

uest

on

July

20,

202

1

complex so far, including possibly Fgd1 and Dbs (36, 37).Whether these or other Rho GEFs are involved in CDC42activation and regulation of actin dynamics on Golgi subcom-partments is presently unknown. Moreover, the inhibition ofARF1-dependent actin polymerization by secramine (Fig. 3A) isa further indication that this in vitro system is able to reconstitutethe full activation cascade of Cdc42. Indeed, in vitro secraminewas shown to inhibit binding of Cdc42 to membranes in aRhoGDI-dependent manner (29). Therefore, this drug isthought to inhibit RhoGDI-dependent shuttling of GDP-boundCdc42 between the cytosol and the target membrane, whereGEF-assisted GDP/GTP exchange takes place. Finally, theinhibition of actin polymerization on GUVs we observed in thepresence of the ARF1-binding domain of ARHGAP21(ARFBD) is likely due to binding to GTP-ARF1 and compet-itive displacement of cytosolic ARF1 effectors required for actinassembly. We previously reported that expression of ARFBD incells displaces ARF1 effectors such as AP-1 and COPI from theGolgi complex (17). This assumption also is supported by thenanomolar-range binding affinity of ARFBD for GTP-boundARF1 (Kd � 50 nM) (18). In addition, we observed a furthersignificant inhibition of actin polymerization in the presence ofARFBD fused with the RhoGAP domain of ARHGAP21 (seeSI Fig. 6). This effect might be due to RhoGAP-stimulated GTPhydrolysis on CDC42 associated with ARF1-GUVs and henceinhibition of CDC42-dependent actin polymerization.

Our findings that an unidentified CDC42-GEF is involveddownstream of ARF1 in activating CDC42 on Golgi membranes,together with our previous observations that ARF1 recruits theCDC42-GAP ARHGAP21 protein on the Golgi complex (17),indicate that ARF1 is able to control the full GDP/GTP cycle ofCDC42, and hence regulates actin dynamics on Golgi mem-branes. The availability of an in vitro system supporting acomplex multicomponent signaling cascade leading to CDC42activation downstream of ARF1 should help to clarify theimplication of upstream regulators in the activation of CDC42 onGolgi membranes.

It is important to note that the positive influence of COPI onthe recruitment of CDC42 could only account for actin poly-merization on cis-medial Golgi cisternae, and not on the trans-side of the Golgi complex, the TGN, where ARF1 is known tocontrol vesicle budding through clathrin and adaptor proteins,APs and GGAs (14). However, roles for CDC42 for exit ofcargos out of the TGN also have been described (38), andadditional mechanisms must exist to ensure CDC42-dependentactin polymerization on the TGN. In the endocytic pathway,CDC42 activity is regulated by the intersectin families of CDC42GEFs, which also bind to N-WASP and are part of a corecomplex composed of dynamin, cortactin, ABP1, and syndapininvolved in clathrin-dependent endocytosis (3, 39, 40). A homo-logue complex consisting of syndapin II and dynamin II pro-motes vesicle formation at the TGN (41). Whether this complexis linked to the ARF1/cortactin/dynamin pathway regulatingpost-Golgi transport (5, 10) and to CDC42 activity is presentlyunknown.

Analyzing the stresses and forces developed during actinpolymerization allows us to understand the formation of eitheractin comet tails or balls of actin in our assay (see SI Discussionfor a more complete analysis). We know from previous studieson polystyrene beads that the growth of an actin gel throughArp2/3 nucleation generates stresses inside the gel that lead togel rupture and bead rocketing (42). On liposomes, which arewater-permeable, forces developed during actin polymerizationcreate an osmotic pressure difference that can lead to membranerupture and solute leakage. Two scenarios can be imagined:Either the stress in the actin gel produces a comet-propelledliposome before pressure is too high to create solute leakage orthe liposome leaks and shrinks slowly while actin continues

polymerizing, thus producing balls of actin with no detectableliposome. The osmotic pressure at which the vesicle leaksdepends inversely on the radius (see SI Discussion). Therefore,larger liposomes are more likely to leak, whereas smaller lipo-somes are more stable and allow comet formation to occur,which is what we see in our experiments.

Although balls of actin are probably not relevant in the Golgi,smaller liposomes propulsed by comets are reminiscent of whatis observed when vesicles bud from the Golgi. We show in ourassay that movement of liposomes carrying ARF1 can be trig-gered by actin polymerization. The force generated during thisprocess has been well characterized in Arp2/3-based assays(43–46), and this force can lead to membrane separation, asobserved in other Arp2/3-based liposome reconstitution assayswhere actin-associated vesicles were observed budding fromlarger membrane masses (43, 44). Actin polymerization mightnot be the sole factor responsible for vesicle separation. How-ever, we propose that actin polymerization mediated by ARF1might help vesicles to transiently pinch off and move away fromdonor compartments, reminiscent of models proposed for en-docytosis (47, 48). The reconstitution of ARF1-dependent actinpolymerization in a model system paves the way for a betterunderstanding of the mechanism of vesicle dynamics on Golgimembranes.

Materials and MethodsMaterials. Rabbit muscle actin was from Cytoskeleton (Denver,CO), cytochalasin D and GTP-�-S were purchased from Sigma–Aldrich (St. Louis, MO), and Wiskostatin was from Calbiochem(San Diego, CA). Lipids were obtained from Avanti Polar Lipids(Alabaster, AL), Alexa Fluor 594 rabbit muscle actin was fromMolecular Probes (Eugene, OR), and anti-PIP2 antibody fluo-rescein conjugate was from Echelon (Salt Lake City, UT).Secramine was a gift from Eric Macia (Institut de PharmacologieMoleculaire et Cellulaire, Valbonne, France).

HeLa cell extracts were prepared as described in SI Materialsand Methods. Total protein concentration of the extract, asmeasured by Bradford assay with a BSA standard, was 14 mg/ml.

Myristoylated ARF1 was prepared from Escherichia coli-coexpressing bovine ARF1 and yeast N-myristoyltransferase bya procedure described previously (24). Purification of GST-fusion protein of the PAK-CRIB and indicated domains ofARHGAP21 has been described previously (17).

Liposome Preparation. One- to 20-�m-diameter GUVs were pre-pared by the gentle hydration method (22, 49). Three hundredmicrograms of lipids solubilized in organic solvent was depositedon a Teflon disk and dried under vacuum for 2 h to ensurecomplete evaporation of the solvent. The Teflon disk was placedin a beaker, and 10 ml of a 0.28 M sucrose solution was carefullyadded. The liposomes grew spontaneously during an overnightincubation. The �0.80-�m-diameter LUVs were prepared byextrusion (23). When noted, 2% NBD-PC was added to the lipidsas a fluorescent probe.

The difference in optical index between the sucrose solutioninside the liposomes and the outside buffer solutions permitted theobservation of the liposomes with phase-contrast microscopy. Theosmolarity of these solutions was checked to prevent the bursting ofthe liposomes. Because PIP2 is notoriously difficult to mix withother lipids, we verified its presence by incubating the GUV withfluorescein-labeled anti-PIP2 antibody.

ARF1 Binding on Membranes. To bind myr-ARF1 on GUVs, weused a modification of the procedure described in ref. 24. Thirtymicroliters of the liposome stock solution were mixed with 30 �lof Glu buffer [50 mM Hepes (pH 7.5), 2 mM MgCl2, 180 mMD-glucose] supplemented with 1.5 mM final DTT, 350 �M finalGTP-�-S, 1.6 �M final of purified myr-ARF1, and 2.6 mM

16932 � www.pnas.org�cgi�doi�10.1073�pnas.0704749104 Heuvingh et al.

Dow

nloa

ded

by g

uest

on

July

20,

202

1

EDTA. GDP/GTP exchange and myr-ARF1 insertion on GUVswere allowed for 45–60 min at room temperature, and then 2.6mM MgCl2 was added. We washed the GUV with Glu buffer by200 � g centrifugation for 20 min to reduce the concentration offree GTP-�-S that could interact with other G proteins (residualGTP-�-S concentration �10 �M). When theARHGAP21 effectwas tested, a second centrifugation step was added (residualGTP-�-S final concentration �1 �M). We prepared ARF1-LUVs in the same way, except that the concentration of lipo-somes was 0.25 mg/ml and centrifugation was omitted (residualGTP-�-S concentration �10 �M).

Motility and Polymerization Assay. For motility and polymerizationassays, 1-�l extracts were mixed with 1 �l of G-actin (partly labeled)in G buffer, 1 �l of a solution containing ATP, creatin-phosphateand 1,4-diazabicyclo[2.2.2]octane (DABCO), and 4 �l of ARF1-GUVs washed in Glu buffer. When drugs or peptide were addedto the assay, it was diluted in 1 �l of Glu buffer, and 1 �l ofARF1-GUVs was omitted instead. Final concentrations were 2mg/ml extracts, 3.8 �M rabbit muscle G-actin, 290 nM Alexa Fluor594 G-actin, 1 mM ATP, 20 mM creatin-phosphate, and 135 �M1,4-diazabicyclo[2.2.2]octane (DABCO). Final salinity was equiv-alent to 80 mM KCl. A higher salinity (150 mM KCl) did notsignificantly affect the assay (see SI Fig. 7). GUVs were observedwith an IX70 Olympus inverted microscope and an Olympus�100/N.A. 1.35 phase-contrast objective (Olympus, Tokyo, Japan).Fluorescent-labeled molecules were excited by a 200-W mercurylamp (OSRAM, Munich, Germany). Images were recorded with aCCD camera (Roper Scientific, Trenton, NJ) driven by Meta-

Morph software (Universal Imaging Corporation, Downingtown,PA).

Quantification of Actin Polymerization Around GUVs. Fluorescentimages were analyzed by using ImageJ software (http://rsb.info.nih.gov/ij) to quantify the amount of fluorescently la-beled actin present around the membrane of the GUVs. Usingthe plug-in ‘‘Oval Profile Plot,’’ the maximum intensity valuealong a radius starting from the center of the GUV was taken.This measure was repeated 360 times by rotating the radius,producing a circumferential profile. The mean value of thisprofile was taken as a measure of the quantity of actin arounda GUV. For each experimental condition, 40–70 GUVs wereanalyzed, and results were gathered in a histogram. Because thedistribution of actin quantity was similar to a log-normal law, weused a logarithmic scale for the fluorescence intensity (i.e., wedistributed the data into bins of exponentially increasing size).Different experimental conditions were compared by using aWilcoxon rank-sum test. Because of some variability in the assay,statistical analysis was performed only on data sets obtained onthe same day with the same batch of ARF1-GUVs and extracts.

We thank Tomas Kirchhausen and Eric Macia for providing secramineand helpful comments; Thierry Dubois and Nadia Elkhatib for prepa-ration of GST-fusion proteins PAK-CRIB and PAK-ARHGAP21; andAlexis Gautreau, Julie Plastino, and Bruno Antonny for fruitful discus-sions. This work was supported by Ligue Nationale Contre le Cancer‘‘Equipe Labellisee’’ and Fondation BNP-Paribas grants (to P.C.), aHuman Frontier Science Program grant (to C.S.), a Program Incitatifand Cooperatif from Institut Curie and ACI Nanoscience grant (to C.S.and P.C.), and an ACI Nanoscience fellowship (to J.H.).

1. Rothman JE, Wieland FT (1996) Science 272:227–234.2. Bonifacino JS, Glick BS (2004) Cell 116:153–166.3. Kaksonen M, Toret CP, Drubin DG (2006) Nat Rev Mol Cell Biol 7:404–414.4. Stamnes M (2002) Curr Opin Cell Biol 14:428–433.5. McNiven MA, Thompson HM (2006) Science 313:1591–1594.6. di Campli A, Valderrama F, Babia T, De Matteis MA, Luini A, Egea G (1999)

Cell Motil Cytoskeleton 43:334–348.7. Valderrama F, Babia T, Ayala I, Kok JW, Renau-Piqueras J, Egea G (1998)

Eur J Cell Biol 76:9–17.8. Egea G, Lazaro-Dieguez F, Vilella M (2006) Curr Opin Cell Biol 18:168–178.9. Carreno S, Engqvist-Goldstein AE, Zhang CX, McDonald KL, Drubin DG

(2004) J Cell Biol 165:781–788.10. Cao H, Weller S, Orth JD, Chen J, Huang B, Chen JL, Stamnes M, McNiven

MA (2005) Nat Cell Biol 7:483–492.11. Chen JL, Lacomis L, Erdjument-Bromage H, Tempst P, Stamnes M (2004)

FEBS Lett 566:281–286.12. Fucini RV, Chen JL, Sharma C, Kessels MM, Stamnes M (2002) Mol Biol Cell

13:621–631.13. Valderrama F, Luna A, Babia T, Martinez-Menarguez JA, Ballesta J, Barth H,

Chaponnier C, Renau-Piqueras J, Egea G (2000) Proc Natl Acad Sci USA97:1560–1565.

14. D’Souza-Schorey C, Chavrier P (2006) Nat Rev Mol Cell Biol 7:347–358.15. Wu WJ, Erickson JW, Lin R, Cerione RA (2000) Nature 405:800–804.16. Matas OB, Martinez-Menarguez JA, Egea G (2004) Traffic 5:838–846.17. Dubois T, Paleotti O, Mironov AA, Fraisier V, Stradal TE, De Matteis MA,

Franco M, Chavrier P (2005) Nat Cell Biol 7:353–364.18. Menetrey J, Perderiset M, Cicolari J, Dubois T, Elkhatib N, El Khadali F,

Franco M, Chavrier P, Houdusse P (2007) EMBO J 26:1953–1962.19. De Matteis MA, Di Campli A, Godi A (2005) Biochim Biophys Acta 1744:396–

405.20. Godi A, Pertile P, Meyers R, Marra P, Di Tullio G, Iurisci C, Luini A, Corda

D, De Matteis MA (1999) Nat Cell Biol 1:280–287.21. De Matteis MA, Godi A (2004) Nat Cell Biol 6:487–492.22. Needham D, Evans E (1988) Biochemistry 27:8261–8269.23. Macdonald RC, Macdonald RI, Menco BPM, Takeshita K, Subbarao NK, Hu

LR (1991) Biochim Biophys Acta 1061:297–303.24. Franco M, Chardin P, Chabre M, Paris S (1995) J Biol Chem 270:1337–1341.25. Cameron LA, Footer MJ, van Oudenaarden A, Theriot JA (1999) Proc Natl

Acad Sci USA 96:4908–4913.26. Plastino J, Sykes C (2005) Curr Opin Cell Biol 17:62–66.27. Cooper JA (1987) J Cell Biol 105:1473–1478.

28. Rohatgi R, Ho HY, Kirschner MW (2000) J Cell Biol 150:1299–1310.29. Pelish HE, Peterson JR, Salvarezza SB, Rodriguez-Boulan E, Chen JL,

Stamnes M, Macia E, Feng Y, Shair MD, Kirchhausen T (2006) Nat Chem Biol2:39–46.

30. Burbelo PD, Drechsel D, Hall A (1995) J Biol Chem 270:29071–29074.31. Kim AS, Kakalis LT, Abdul-Manan M, Liu GA, Rosen MK (2000) Nature

404:151–158.32. Peterson JR, Bickford LC, Morgan D, Kim AS, Ouerfelli O, Kirschner MW,

Rosen MK (2004) Nat Struct Mol Biol 11:747–755.33. Rohatgi R, Ma L, Miki H, Lopez M, Kirchhausen T, Takenawa T, Kirschner

MW (1999) Cell 97:221–231.34. Ma L, Cantley LC, Janmey PA, Kirschner MW (1998) J Cell Biol 140:1125–

1136.35. Rossman KL, Der CJ, Sondek J (2005) Nat Rev Mol Cell Biol 6:167–180.36. Estrada L, Caron E, Gorski JL (2001) Hum Mol Genet 10:485–495.37. Kostenko EV, Mahon GM, Cheng L, Whitehead IP (2005) J Biol Chem

280:2807–2817.38. Musch A, Cohen D, Kreitzer G, Rodriguez-Boulan E (2001) EMBO J

20:2171–2179.39. Hussain NK, Jenna S, Glogauer M, Quinn CC, Wasiak S, Guipponi M,

Antonarakis SE, Kay BK, Stossel TP, Lamarche-Vane N, et al. (2001) Nat CellBiol 3:927–932.

40. Kessels MM, Engqvist-Goldstein AEY, Drubin DG, Qualmann B (2001) J CellBiol 153:351–366.

41. Kessels MM, Dong JX, Leibig W, Westermann P, Qualmann B (2006) J CellSci 119:1504–1516.

42. van der Gucht J, Paluch E, Plastino J, Sykes C (2005) Proc Natl Acad Sci USA102:7847–7852.

43. Giardini PA, Fletcher DA, Theriot JA (2003) Proc Natl Acad Sci USA100:6493–6498.

44. Upadhyaya A, Chabot JR, Andreeva A, Samadani A, van Oudenaarden A(2003) Proc Natl Acad Sci USA 100:4521–4526.

45. Boukellal H, Campas O, Joanny JF, Prost J, Sykes C (2004) Phys Rev E StatNonlin Soft Matter Phys 69:061906.

46. Marcy Y, Prost J, Carlier MF, Sykes C (2004) Proc Natl Acad Sci USA101:5992–5997.

47. Liu J, Kaksonen M, Drubin DG, Oster G (2006) Proc Natl Acad Sci USA103:10277–10282.

48. Merrifield CJ (2004) Trends Cell Biol 14:352–358.49. Reeves JP, Dowben RM (1969) J Cell Physiol 73:49–60.

Heuvingh et al. PNAS � October 23, 2007 � vol. 104 � no. 43 � 16933

CELL

BIO

LOG

Y

Dow

nloa

ded

by g

uest

on

July

20,

202

1