Golgi-localized GAP for Cdc42 functions downstream of ARF1 to ...

17 downloads 0 Views 2MB Size Report
Mar 27, 2005 - Thierry Dubois1, Olivia Paléotti2, Alexander A. Mironov Jr3, Vincent Fraisier4, Theresia E. B. Stradal5, Maria. Antonietta De Matteis3, Michel ...
A RT I C L E S

Golgi-localized GAP for Cdc42 functions downstream of ARF1 to control Arp2/3 complex and F-actin dynamics Thierry Dubois1, Olivia Paléotti2, Alexander A. Mironov Jr3, Vincent Fraisier4, Theresia E. B. Stradal5, Maria Antonietta De Matteis3, Michel Franco2 and Philippe Chavrier1,6 The small GTP-binding ADP-ribosylation factor 1 (ARF1) acts as a master regulator of Golgi structure and function through the recruitment and activation of various downstream effectors. It has been proposed that members of the Rho family of small GTPases also control Golgi function in coordination with ARF1, possibly through the regulation of Arp2/3 complex and actin polymerization on Golgi membranes. Here, we identify ARHGAP10 — a novel Rho GTPase-activating protein (Rho-GAP) that is recruited to Golgi membranes through binding to GTP-ARF1. We show that ARHGAP10 functions preferentially as a GAP for Cdc42 and regulates the Arp2/3 complex and F-actin dynamics at the Golgi through the control of Cdc42 activity. Our results establish a role for ARHGAP10 in Golgi structure and function at the crossroads between ARF1 and Cdc42 signalling pathways. Eukaryotic cells utilize the GDP/GTP switch of ARF1 to control budding of distinct populations of vesicles from the Golgi complex, and to regulate Golgi architecture and function through the recruitment of a variety of effector proteins including vesicle coat components, phosphoinositide modifying enzymes, and elements of the Golgi actin filament binding protein network1–4. Regulation of Golgi function also involves RhoA and Cdc42, small GTP-binding proteins of the Rho subfamily, possibly through the control of local actin filament assembly5–8. A cascade linking RhoA to its downstream effectors ROCK-II and Citron-N, as well as Profilin-IIa, was recently shown to affect Golgi structure. Cdc42, however, which is recruited to Golgi membranes by binding to COPI, has been proposed to regulate actin assembly at the Golgi apparatus through the activation of Wiskott–Aldrich-syndrome protein (WASP) and the Arp2/3 actin nucleating complex5,6,8,9. Because GTP-bound ARF1 controls COPI binding to the Golgi, these findings would position ARF1 upstream of a cascade mediating Cdc42-dependent assembly of Golgi-associated actin cytoskeleton10,11. However, how the activities of ARF1 and Rho GTPases could be coordinated in the Golgi complex is not clear. Here, we identify ARHGAP10 and show that it functions as a GAP for Cdc42 and regulates the Arp2/3 complex and F-actin dynamics at the Golgi. RESULTS ARHGAP10 interacts with GTP-bound ARF We identified ARHGAP10 (also known as ARHGAP21), an as yet uncharacterized human Rho-GAP12, as a potential partner of GTPbound ARF in a yeast two-hybrid screen. Figure 1b illustrates that a

region corresponding to residues 885–1096 of ARHGAP10 (called ARF-BD hereafter), containing a pleckstrin homology (PH) domain (see Fig. 1a for a schematic representation of ARHGAP10), interacted with constitutively activated mutants of ARF6 (ARF6Q67L) and ARF1 (ARF1Q71L), but not with wild-type or nucleotide-binding defective ARF6 (ARF6T27N). Different approaches were used to confirm the two-hybrid interactions. First, glutathione S-transferase (GST) pull-down experiments showed that GST–ARF-BD, but not GST, associated specifically with ARF6Q67L and ARF1Q71L (Fig. 1c). In contrast, no interaction was seen with ARF6T27N or ARF1T31N (Fig. 1c). To assess whether binding of ARHGAP10 to GTP-bound ARFs was direct, GST–ARF-BD was incubated with purified recombinant ARFs loaded with GTP or GDP. Under these conditions, ARF-BD associated specifically with GTP-bound ARF6 (Fig. 1d) as well as GTP-bound recombinant ARF1 (data not shown). The region of ARHGAP10 that is necessary for interaction with GTP-bound ARF was narrowed down on the basis of a series of seven GST constructs (see Fig. 1a and 1e, upper panel), which were incubated with a lysate of cells expressing haemagglutinin (HA)-tagged ARF6Q67L. Constructs comprising the PH domain only (amino acids 929–1052, construct 2) or the PH domain in association with its amino-terminal flanking region (885–1052, construct 3) did not bind to active ARF6 (Fig. 1e, bottom panel), whereas binding was observed when the PH domain was flanked by its carboxy-terminal region (929–1096, construct 4). However, the C-terminal region was not sufficient for binding because no interaction could be detected with construct 7 (1042–1346), indicating that the interaction of ARF-BD with GTP-bound ARF required the PH domain.

1 Membrane and Cytoskeleton Dynamics Group, Institut Curie, CNRS-UMR144, 75248 Paris, France. 2Institut de Pharmacologie Moléculaire et Cellulaire, CNRS-UMR 6097, F-06560 Valbonne, France. 3Department of Cell Biology and Oncology, Consorzio Mario Negri Sud, 66030 Santa Maria Imbaro (CH), Italy. 4Digital Imaging Platform, Institut Curie, CNRS-UMR144, 75248 Paris, France. 5German Research Centre for Biotechnology (GBF), Department of Cell Biology, Mascheroder Weg 1, D-38124 Braunschweig, Germany. 6 Correspondence should be addressed to P.C. (e-mail: [email protected])

Published online: 27 March 2005; DOI: 10.1038/ncb1244

NATURE CELL BIOLOGY VOLUME 7 | NUMBER 4 | APRIL 2005

353

©2005 Nature Publishing Group

print ncb1244.indd 353

16/3/05 1:29:17 pm

ARH-GAP10

1957 aa 885

D AD

F-B

ARF6Q67L ARF6T27N

1052 1052

ARF6

1096 1346 1346 1346

1064

ARF1Q71L + His

GS AR T − F-B D

T GS

Mr(K) 150 100 75 -

P

− His

GST constructs

e

GT

GD

P

P G ST P

ARF6

GT

GST−ARF-BD

GD

F1 Q 7

1L

1042

d

GST

D

pLex

1346

1096

AR

AR

F1 T 31

N

7L

1 (ARF-BD) 885 2 (PH) 929 3 885 4 929 5 (ARF-BD/Rho-GAP) 929 6 (Rho-GAP) 7

AR F6 T 27 N AR F6 Q 6

Rho-GAP

PH

pG

Rho-GAP

AR

AR

PH

PDZ

pG A

b

a

c

F-B

D

A RT I C L E S

1

2

3

4 5 6 7

50 -

Bound

37 Lysate

Input 25 -

Anti-ARF6

Coomassie blue

T GS

Ly sa

tes

Anti-HA

1

2

3

4

5

6 7 Anti-HA

Figure 1 ARHGAP10 interacts with GTP-bound ARF1 and ARF6. (a) ARHGAP10 and the different GST- and HA-tagged constructs used in this study are schematically represented. ARHGAP10 is a Rho-GAP protein that, in addition to a conserved Rho-GAP domain, contains an N-terminal PDZ domain and a PH domain12. The vertical bar indicates the position of the peptide (residues 1878–1895) used to raise polyclonal antibodies specific to human ARHGAP10. (b) Two million clones of a human placenta cDNA library were screened with the constitutively active ARF6 mutant (ARF6Q67L) fused to the LexA DNA-binding domain as a bait. Forty-three clones were isolated that associated specifically with ARF6Q67L, and among these, 12 clones corresponded to overlapping fragments of ARHGAP10. Growth on medium lacking histidine (−His) indicates a positive two-hybrid interaction of the shortest isolated ARHGAP10 clone, called ARF-BD (corresponding to residues 885–1096, expressed as a fusion with the Gal4 transcriptional-activation domain), with the activated form of ARF6 and ARF1 (ARF1Q71L). (c) HeLa cells were transfected with C-terminally

HA-tagged ARF1 or ARF6 variants and cell extracts were incubated with GST–ARF-BD or GST, as a control. Complexes were immobilized on glutathione-Sepharose beads, washed and analysed by SDS–PAGE together with aliquots of the starting cell lysates. Detection of bound proteins was performed by immunoblotting with anti-HA tag monoclonal antibody. (d) GST or GST–ARF-BD, bound to glutathione-Sepharose beads, was incubated with purified recombinant ARF6 loaded with GDP or GTP. ARF6 bound to the beads was analysed by SDS–PAGE together with an aliquot of input ARF6 loaded as a standard. ARF6 was detected by western blot using anti-ARF6 polyclonal antibodies. (e) GST and the different GST– ARHGAP10 constructs (constructs 1–7, see Fig. 1a) were purified from E. coli and analysed by SDS–PAGE and Coomassie blue staining (4 µg each, upper panel). Equimolar amounts of GST fusion proteins were incubated with lysates of cells expressing HA-tagged ARF6Q67L. Pull-downs were performed as described for c, and bound ARF6Q67L was detected with anti-HA tag monoclonal antibody (bottom panel).

Finally, the Rho-GAP domain (1158–1346, construct 6) did not seem to contribute to binding (Fig. 1e). These findings demonstrate a specific interaction of ARHGAP10 with GTP-bound ARF1 and ARF6, mediated by a region of ARHGAP10 that encompasses the PH domain and a Cterminal flanking region.

the Golgi complex, including GM130 (Golgi matrix protein, Figs 2c–e and 3k, l), and two Golgi coat/adaptor complex proteins depending on GTP-ARF1 for their Golgi localization, β-COP (Fig. 2f–h) and AP1 (Figs 2i–k and 3i, j), as well as ARF1 (data not shown). We also noticed a staining of vesicular cytoplasmic structures that we could not assess further because polyclonal anti-ARHGAP10 antibodies did not work for electron microscopic analysis. In contrast, there was no significant overlap of ARHGAP10 with early endosomes that were stained with EEA1 (early endosome antigen 1, not shown). Overall, our results are consistent with a broad association of ARHGAP10 to the Golgi complex that is reminiscent of ARF1 distribution1. Next, we investigated the effect of mutant forms of ARF1 and ARF6 on ARHGAP10 localization. As previously reported13,14, expression of constitutively active ARF1Q71L induced a change in Golgi morphology with Golgi membranes appearing around the nucleus as large collapsed structures that were positive for both ARF1Q71L and ARHGAP10 (Fig. 3a, b, arrows). In addition, inhibition of ARF1 activity by expression of the dominant inhibitory ARF1T31N mutant, which prevents GTP/GDP

Association of ARHGAP10 with the Golgi complex is regulated by GTP-ARF1 Affinity-purified antibodies directed against a peptide derived from the C-terminal end of human ARHGAP10 (Fig. 1a) detected a single band of apparent relative molecular mass ≥250,000 (Mr 250K) by immunoblotting analysis of HeLa cell extract, slightly above the predicted relative molecular mass of ARHGAP10 (Mr 217.3K). This band was observed almost equally between high-speed membrane and cytosolic fractions (Fig. 2a, b). Immunofluorescence analysis of several human cell lines, including breast adenocarcinoma MCF-7 cells and HeLa cells (Figs 2 and 3, respectively), with anti-ARHGAP10 antibodies revealed that the main staining was juxtanuclear and overlapped with markers of 354

NATURE CELL BIOLOGY VOLUME 7 | NUMBER 4 | APRIL 2005 ©2005 Nature Publishing Group

print ncb1244.indd 354

16/3/05 1:29:21 pm

A RT I C L E S a

c

Peptide Mr(K)



d

e

GM130

GAP10

Merge

f

g

h

+

250 150 100 75 -

- GAP10

50 37 25 20 15 -

PN

S

b S

P

- GAP10

- RhoA

β-COP

GAP10

Merge

i

j

k

AP1

GAP10

Merge

- TfR

Figure 2 Distribution of endogenous ARHGAP10 by cell fractionation and immunofluorescence analyses. (a) Immunoblotting analysis of HeLa cell extract with anti-ARHGAP10 polyclonal antibodies. Rabbit polyclonal antibodies were raised against a synthetic peptide corresponding to residues 1878–1895 of human ARHGAP10 and affinity purified on the immunogenic peptide. These antibodies detected a protein with Mr ≥ 250K protein (GAP10) by immunoblotting analysis of HeLa cell lysate (− peptide). Pre-incubation of the antibodies with an excess of antigenic peptide abolished detection of ARHGAP10 (+ peptide). Relative molecular masses are indicated (Mr(K)). (b) Distribution of ARHGAP10 in subcellular fractions. HeLa cells were homogenized and a postnuclear supernatant was prepared, and centrifuged at 100,000g to give a soluble (S) and a pellet (P) fraction. In each lane,

volumes of fractions representing the same numbers of cells were loaded. The presence of endogenous ARHGAP10, RhoA GTPase (mostly cytosolic) and transferrin receptor (TfR, integral membrane protein) in the fractions was revealed by immunoblotting analysis. (c–k) Immunofluorescence localization of endogenous ARHGAP10 in human MCF-7 breast adenocarcinoma cells, chosen because of their flat extended Golgi complex morphology. The localization of endogenous ARHGAP10 (d, g, j) was compared with that of the Golgi matrix protein GM130 (c), the Golgi coat component β-COP (f) and the adaptor protein AP1 (i) on single optical sections by confocal immunofluorescence microscopy. Merged images showing ARHGAP10 in red and the Golgi markers in green (e, h, k) depict extensive colocalization of ARHGAP10 with the markers. Scale bars, 10 µm.

exchange on endogenous ARF1, caused the disassembly of the Golgi complex and a redistribution of ARHGAP10 to a diffuse cytosolic staining (Fig. 3c, d, arrows). Similarly, expression of ARFGAP1, which promotes GTP hydrolysis on ARF1, triggered a redistribution of ARHGAP10 in the cytoplasm (data not shown). In sharp contrast, constitutively active and dominant-negative mutants of ARF6 had no effect on the juxtanuclear distribution of ARHGAP10 (Fig. 3e–h). To further substantiate the relationship between GTP-ARF1 and ARHGAP10, HeLa cells were treated with brefeldin A (BFA), which forms a non-competitive complex with ARF1’s guanine nucleotide-exchange factors (GEFs), but not with ARF6’s GEFs, thereby causing GDP-bound ARF1 and its downstream effectors to be released in the cytosol within minutes after treatment, and resulting in Golgi breakdown15,16. When treated with BFA, ARHGAP10 disappeared from its juxtanuclear Golgi localization with fast kinetics similar to that of adaptor complex AP1 (see ref. 17; Fig. 3, compare panels i, j with m, n), on timescales in which Golgi complex was unaffected as judged by staining with antibodies against GM130 (Fig. 3, compare

panels k, l with o, p). Expression of ARF1Q71L, previously shown to protect against the dissociation of COPI from the Golgi complex upon BFA treatment14,18, similarly prevented the redistribution of ARHGAP10 that was induced by BFA even after prolonged exposure to the drug (10 min, Fig. 3q, r, arrows). Furthermore, we observed that overexpressed green fluorescent protein (GFP)-tagged ARF-BD (construct 1, see Fig. 1a), which localized mostly to the Golgi complex in HeLa cells (Fig. 3s, arrows), triggered the disappearance of Golgi-localized AP1 staining, most likely by competing for GTP-bound ARF1 (Fig. 3s, t). Altogether, these data demonstrate that the localization of ARHGAP10 to the Golgi complex is dependent on GTP-bound ARF1, and identify ARHGAP10 as a novel GTP-ARF1-interacting protein in the Golgi complex. ARHGAP10 PH domain does not bind to lipids The ARF-binding region of ARHGAP10 encompasses a PH domain (see Fig. 1a) that could bind to specific lipids, thereby further contributing to the association of ARHGAP10 with Golgi membranes. The ability

NATURE CELL BIOLOGY VOLUME 7 | NUMBER 4 | APRIL 2005

355

©2005 Nature Publishing Group

print ncb1244.indd 355

16/3/05 1:29:22 pm

A RT I C L E S ARF1Q71L

ARF1T31N

a

b

c

ARF1Q71L

GAP10

ARF1T31N

d

GAP10 ARF6T27N

ARF6Q67L

e

f

g

h

ARF6Q67L

GAP10

ARF6T27N

GAP10

i

j

k

l

AP1

GAP10

GM130

GAP10

m

n

o

p

AP1

GAP10

GM130

GAP10

No BFA

2 min BFA

ARF1Q71L / 10 min BFA

GFP−ARF-BD

q

r

s

t

ARF1Q71L

GAP10

GFP−ARF-BD

AP1

Figure 3 ARHGAP10 localizes to the Golgi complex in a GTP-ARF1-dependent manner. (a–h) Distribution of endogenous ARHGAP10 in cells expressing mutants of ARF proteins. HeLa cells were transiently transfected with constructs encoding for HA-tagged activate ARF1Q71L (a, b), dominant inhibitory ARF1T31N (c, d), active ARF6Q67L (e, f), or inhibitory ARF6T27N (g, h). After 24 h, cells were fixed and stained with anti-ARHGAP10 antibodies (GAP10; b, d, f, h) and anti-HA tag monoclonal antibody (a, c, e, g). Transfected cells are indicated by arrows. Double-headed arrows indicate non-transfected cells with ARHGAP10 localization to the Golgi complex. (i–p) BFA redistributes ARHGAP10 to the cytosol. HeLa cells were fixed before (i–l) or after a 2-min treatment with BFA added to a final concentration of 5 µg ml−1 (m–p). Cells were fixed and then stained with anti-ARHGAP10 (GAP10; j, l, n, p), and anti-AP1 (i, m) or anti-GM130 (k, o) antibodies. Confocal sections shown in i–p were recorded the same

356

day using identical settings of the confocal microscope. Pixel intensities are directly comparable and are indicative of the amount of proteins in the perinuclear Golgi region. (q, r) Expression of ARF1Q71L protects against BFA-mediated effects. HeLa cells were transfected with a construct coding for HA-tagged ARF1Q71L for 16 h. BFA (5 µg ml−1) was then added to the cells, and after 10 min cells were fixed and stained with anti-ARHGAP10 antibodies (r) and anti-HA monoclonal antibody (q). (s, t) Displacement of AP1 by ARF-BD interacting with GTP-ARF1 on the Golgi complex. HeLa cells were transiently transfected with a construct coding for GFP-tagged ARF-BD (construct 1, see Fig. 1a). Cells were fixed after 16 h and stained with anti-AP1 antibodies (t). Arrows point to Golgi complexes positive for ARF-BD with decreased AP1 levels. Double-headed arrows indicate nontransfected cells. All the experiments were repeated at least three times with identical results. Scale bars, 10 µm.

NATURE CELL BIOLOGY VOLUME 7 | NUMBER 4 | APRIL 2005 ©2005 Nature Publishing Group

print ncb1244.indd 356

16/3/05 1:29:24 pm

A RT I C L E S of ARF-BD to interact with liposomes of defined lipid composition was analysed in order to assess this possibility. Liposomes made up of a mixture of phosphatidylcholine/phosphatidylethanolamine (PC/PE) supplemented with phosphatidylserine and various phosphoinositides were incubated with ARF-BD and sedimented by centrifugation. Quantification of the sedimentation data showed that ARF-BD did not associate with PC/PE liposomes, even when they were supplemented with phosphatidylserine and phosphatidylinositol (PtdIns), PtdIns(3)P, PtdIns(4,5)P2 or PtdIns(3,4,5)P3 (Fig. 4a). Parallel studies based on lipid overlay assays further confirmed that the PH domain of ARHGAP10 did not bind to PtdIns, PtdIns(3)P, PtdIns(4)P, PtdIns(5)P, PtdIns(3,4)P2, PtdIns(3,5)P2, PtdIns(4,5)P2 or to PtdIns(3,4,5)P3 (D. Alessi, unpublished observations). Next, we tested whether ARF-BD could be recruited to liposomes through binding to recombinant myristoylated GTP-ARF1 (Myr-ARF1), which sedimented with PC/PE liposomes as previously reported19 (Fig. 4c). Incubation of ARF-BD with Myr-ARF1-GTP triggered a significant binding of ARF-BD together with Myr-ARF1 to the liposome pellets, irrespective of their lipid composition (Fig. 4b). In contrast, ∆17-ARF1-GTP — a mutant with the 17 N-terminal residues deleted — which was not myristoylated and did not interact with liposomes, did not allow ARF-BD association with liposomes, although it did interact with ARF-BD (data not shown). In addition, in contrast to overexpressed GFP-ARF-BD that localized to the Golgi complex in transfected HeLa cells (Fig. 3s), the isolated ARHGAP10 PH domain (construct 2, Fig. 1a) was exclusively cytosolic (data not shown). In conclusion, although ARF-BD comprises a domain that is likely to adopt a PH domain fold based on sequence homologies, our data indicate that Golgi association of ARHGAP10 is mediated through a direct interaction of ARF-BD with GTP-ARF1 without detectable association of the PH domain to membrane lipids. ARHGAP10 is a Cdc42-GAP The above findings suggest that ARHGAP10 is a novel Rho-GAP protein that localizes to the Golgi apparatus through binding to GTP-bound ARF1. To address the substrate specificity of the ARHGAP10 Rho-GAP domain in vitro (construct 6, Fig. 1a), we performed Rho-GAP assays in the presence of recombinant 32P-γ-GTP-preloaded GST–RhoA, Rac1, or Cdc42 fusion proteins. Figure 5 shows that ARHGAP10 was a potent GAP for Cdc42, although it is much less effective on RhoA (4.5-fold less active compared with Cdc42; see Fig. 5d) or Rac1 (9.9-fold less). We could not assess the effect of full-length ARHGAP10 on cells because we failed to overexpress this protein, probably because of its high molecular weight. To characterize the GAP activity of ARHGAP10 in vivo, we expressed the Rho-GAP domain together with the flanking ARF-BD in fusion with GFP in HeLa cells (ARF-BD/Rho-GAP, construct 5; see Fig. 1a). ARF-BD/Rho-GAP, like ARF-BD, showed a juxtanuclear localization corresponding to the Golgi complex and induced morphological changes of transfected cells, giving a characteristic snowflake shape (see Supplementary Information, Fig. S1a, b). F-actin staining revealed a complete disorganization of the cytoskeleton with formation of actin aggregates and loss of stress fibres (see Supplementary Information, Fig. S1a, b). Golgi targeting of the Rho-GAP domain harbouring an arginine-toalanine mutation at a residue critical for GAP activity and conserved in all Rho-GAP domains (Rho-GAPR1183A) abolished both cell shape changes and F-actin reorganization (see Supplementary Information, Fig. S1c, d), indicating that Rho-GAP activity is required for these changes to occur.

a PC/PE + PS ARF-BD + PS/PtdInsP + PS/PtdInsP2 + PS/PtdInsP3

b PC/PE + PS + PS/PtdInsP

ARF-BD + Myr-ARF1 + PS/PtdInsP2 + PS/PtdInsP3

c Myr-ARF1

PC/PE + PS + PS/PtdInsP2 0

20

40

60

80

100

Bound to liposome (percentage of total)

Figure 4 ARF-BD is recruited to lipids through direct interaction with GTPbound myristoylated ARF1. ARF-BD (2 µM) was incubated in the presence of 0.8 g l−1 phospholipid vesicles of defined composition containing 50% phosphatidylcholine (PC) and 50% phosphatidylethanolamine (PE), or 35% PC/35% PE and 30% phosphatidylserine (+ PS), or further supplemented with 2% PtdIns(4)P (+ PS/PtdInsP), 2% PtdIns(4,5)P2 (+ PS/PtdInsP2) or 2% PtdIns(3,4,5)P3 (+PS/PtdInsP3), in the absence (a) or in the presence of 3 µM GTP-bound myristoylated ARF1 (Myr-ARF1) (b). The lipid vesicles with the protein(s) bound to them were sedimented by centrifugation, and bound proteins were analysed by SDS–PAGE and Sypro Orange staining. Bars represent the fraction of protein sedimented in the lipid vesicle pellet as determined by densitometric analysis and expressed as a percentage of the total protein in the assay. (a) Fraction of ARF-BD bound to vesicles of indicated phospholipid composition. (b) Fraction of ARF-BD bound to phospholipids vesicles in the presence of myristoylated GTP-ARF1. The means ± s.e.m. of three separate experiments are plotted (a, b). (c) Fraction of myristoylated GTP-ARF1 bound to phospholipid vesicles.

Finally, overexpression of the isolated Rho-GAP domain led to an accumulation of the protein in the cytosol of transfected cells (corresponding to construct 6, Fig. 1a) and resulted in mild morphological alterations (see Supplementary Information, Fig. S1e, f). Altogether, these results demonstrate that ARHGAP10 is a bona fide Rho-GAP protein acting preferentially on Cdc42 over RhoA and Rac1 in vitro, and which affects the overall morphology of the cells and underlying actin cytoskeleton upon targeting to the Golgi complex. ARHGAP10 is required for Golgi complex organization Next, we used two independent short interfering RNAs (siRNAs) named siGAP10#1 and siGAP10#2 (see Methods) to lower ARHGAP10 expression in HeLa cells, in order to shed light on the function of this ARF1-interacting protein in the Golgi complex. Western blot analysis showed that expression of ARHGAP10 was strongly diminished in cells that were treated for 72 h with GAP10 siRNAs (see Fig. 6a, b). Depletion of ARHGAP10 induced cell spreading and accumulation of F-actin stress fibres (Fig. 6; compare panels e and g with c). Notably, overexpression of a constitutively active GTPase-defective mutant form of Cdc42 (Cdc42Q61L) in HeLa cells similarly induced cell spreading and stress fibre formation, although less prominently than the thick actin cables induced upon expression of RhoAL63Q (see Supplementary Information, Fig. S2). Therefore the possibility exists that, as already reported20, induction of stress fibre formation upon depletion of ARHGAP10 may be an indirect consequence of Cdc42

NATURE CELL BIOLOGY VOLUME 7 | NUMBER 4 | APRIL 2005

357

©2005 Nature Publishing Group

print ncb1244.indd 357

16/3/05 1:29:25 pm

A RT I C L E S

a

b

Cdc42

bound (percentage of time zero)

80

100

60

40

32P-γ-GTP

bound (percentage of time zero)

100

32P-γ-GTP

RhoA

GAP10 10 nM GAP10 35 nM GAP10 100 nM GAP10 300 nM Exp 1 GAP10 300 nM Exp 2 GAP10 900 nM Exp 1 GAP10 900 nM Exp 2 p50-Rho-GAP 100 nM GST 100 nM

20

0

1

2

3

4

5

6

GAP10 35 nM GAP10 100 nM GAP10 300 nM Exp 1 GAP10 300 nM Exp 2 GAP10 900 nM Exp 1 GAP10 900 nM Exp 2 GAP10 900 nM Exp 3 p50-Rho-GAP 100 nM GST 100nM

80

60

40

20

0

7

1

2

Time (min)

c

5

6

7

7 Rac1

Cdc42

6

GAP10 300 nM Exp 1 GAP10 300 nM Exp 2 GAP10 900 nM Exp 1 GAP10 900 nM Exp 2 p50-Rho-GAP 100 nM GST 100 nM

80

5 Kapp (min−1)

bound (percentage of time zero)

4

d 100

32P-γ-GTP

3

Time (min)

60

40

Cdc42 RhoA Rac1

4 3 2

GAP activity (min−1 nM−1) (6.33 ± 0.27) × 10−3 (1.41 ± 0.23) × 10−3 (0.64 ± 0.13) × 10−3

RhoA

20 1

0

1

2

3

4

5

6

7

0

Time (min)

Rac1

200

400

600

800

1,000

GAP10 (nM)

Figure 5 The Rho-GAP domain of ARHGAP10 acts preferentially on Cdc42 in vitro. (a–c) 1 µM Cdc42 (a), RhoA (b) and Rac1 (c) preloaded with 32P-γGTP were incubated with increasing concentrations of the Rho-GAP domain of ARHGAP10 (construct 6, see Fig. 1a) at 37 °C for the indicated period of time. 32P-γ-GTP bound on the small GTPases was determined at each time point by a nitrocellulose filtration assay and expressed as a percentage of 32 P-γ-GTP bound at time zero. GST (100 nM) or GST–Rho-GAP domain of p50Rho-GAP (100 nM) were used as controls. Kinetics of GTP hydrolysis

were fitted as exponentials (y = a(−Kapp×t) + b) where Kapp represents the apparent rate constant of GTP hydrolysis (min−1). Kinetics of GTP hydrolysis in the presence of 300 or 900 nM of ARHGAP10 Rho-GAP domain were repeated two to three times as indicated. (d) GAP activity of ARHGAP10 Rho-GAP domain on Cdc42, Rac1 and RhoA. Plotting Kapp as a function of the concentration of the Rho-GAP domain showed a linear relationship from which we determined the specific activity (in min−1 nM−1). Data represent means ± s.d. of two to three experiments.

activation leading to RhoA-mediated pathways. Furthermore, depletion of ARHGAP10 had marked consequences on the morphology of the Golgi complex, which lost its compact ribbon organization and broke into dispersed elements (Fig. 6; compare panels f and h with d, and see Fig. 7). This effect was observed in most of the cells treated with ARHGAP10 siRNAs. Electron microscopy was used to analyse the ultrastructure of the Golgi fragments accumulating in ARHGAP10-depleted cells. Although smooth cisternae that organized into stacks could be observed, the mean length (855 ± 46 nm, n = 106) of the cisternae was significantly shorter (P < 0.005) as compared with control cells (1,152 ± 69 nm, n = 122) (Fig. 6i, j). These data indicate that knocking down ARHGAP10 led to the constitutive activation of a Rho GTPase, very likely Cdc42, resulting

in actin cytoskeleton remodelling and suggesting a role for ARHGAP10 in controlling Golgi complex organization.

358

Regulation of Arp2/3 complex dynamics by ARHGAP10 downstream of ARF1 ARF1 was recently shown to control a pathway leading to Arp2/3 complex activation and actin polymerization on purified Golgi membranes9,10. In addition, binding of Cdc42 to COPI-coated membranes is thought to control actin assembly through a pathway involving neural WASP (N-WASP) and the Arp2/3 complex6,9–11,21. Because ARHGAP10 is recruited to the Golgi complex by GTP-bound ARF1 and stimulates the GTPase activity of Cdc42, we wondered whether ARHGAP10 might NATURE CELL BIOLOGY VOLUME 7 | NUMBER 4 | APRIL 2005

©2005 Nature Publishing Group

print ncb1244.indd 358

16/3/05 1:29:27 pm

250 ARHGAP10

10 #2

10 #1 siG

siG

AP

AP

FP siG

Mr(K)

Mo ck

AP

b

siG

AP siG

Mr(K)

siG

Mo ck

a

FP

10 #

10 #2

1

A RT I C L E S

150 Clathrin heavy chain

c

e

g

F-actin siGAP10#2 siGFP

d

siGAP10#2

f

h

CTR433

i

Mock G

M

E

j

siGAP10#1 E

G E

G

M

G

M M M

M

M

M M G

M

N

G

M M

Figure 6 RNAi-mediated depletion of ARHGAP10 induces cell spreading, stress fibre formation and shortening of Golgi stacks. (a, b) HeLa cells were treated for 72 h with two independent siRNAs derived from human ARHGAP10 cDNA (siGAP10#1 or siGAP10#2), or siRNA specific for GFP (siGFP) or mock-treated. Cell lysates were prepared and equal amounts of total protein were separated by SDS–PAGE and analysed by immunoblotting with anti-ARHGAP10 antibodies (a) or anti-clathrin heavy chain as a control (b). Analysis of the immunoblots revealed 8 and 4% of residual ARHGAP10 protein in cells treated with siGAP10#1 and siGAP10#2, respectively (see Methods). (c–h), F-actin and CTR433 distribution in ARHGAP10-depleted cells. HeLa cells treated with control siRNA (siGFP) for 72 h (c, d) or cells

treated with ARHGAP10-specific siRNAs (siGAP10#1 (e–f) or siGAP10#2 (g–h)) were fixed and double-labelled using a monoclonal antibody against the medial-Golgi marker CTR433 (d, f, h) and fluorescently labelled phalloidin to detect actin filaments (c, e, g). Analysis of three independent experiments revealed that 90 ± 3.6% and 84.3 ± 8% of siGAP10#1- and siGAP10#2-treated cells showed F-actin stress fibres, respectively, as compared with 10.7 ± 7% of siGFP-treated control cells. Scale bars, 10 µm. (i, j) Electron microscopic analysis of ARHGAP10-depleted cells. Mocktreated HeLa cells (i) or cells treated with ARHGAP10 siRNA (siGAP10#1) (j) for 72 h were fixed and processed for electron microscopy. G, Golgi stack; M, mitochondrion; E, endosome; N, nucleus. Scale bars, 500 nm.

act downstream of ARF1 to control Cdc42/N-WASP/Arp2/3-dependent actin cytoskeleton dynamics. We did not observe an accumulation of F-actin in the Golgi region in ARHGAP10-depleted cells nor in control cells (data not shown). This was unsurprising because Golgi actin filaments are notoriously difficult to label with standard reagents such as phalloidin22. We thus labelled HeLa cells treated with siRNAs against ARHGAP10 with two different monoclonal antibodies directed against the p16Arc (Fig. 7, upper panels) and p34Arc (Fig. 7, lower panels)

subunits of the Arp2/3 complex23. In control (siGFP-treated) cells, p16Arc staining was mostly diffuse with some barely visible accumulations in the vicinity of the Golgi apparatus (Fig. 7c). Arp2/3 complex-positive perinuclear structures were more visible with anti-p34Arc monoclonal antibody (Fig. 7d–f). Interestingly, labelling of ARHGAP10-depleted cells with both antibodies revealed a marked increase of Arp2/3 complex staining in the perinuclear region (compare panels a–f with g–r). Arp2/3 complex accumulated in punctate structures that formed an extensive

NATURE CELL BIOLOGY VOLUME 7 | NUMBER 4 | APRIL 2005

359

©2005 Nature Publishing Group

print ncb1244.indd 359

16/3/05 1:29:28 pm

A RT I C L E S siGFP

siGAP10#1

siGAP10#2

a

g

m

b

h

n

Giantin

p16Arc

16.7 ± 8.3%, n = 372

92.7 ± 7.6%, n = 361

94.0 ± 2.6%, n = 376

c

i

o

d

j

p

e

k

q

Merge

Giantin

p34Arc

32.3 ± 0.6%, n = 462

f

69.0 ± 2.6%, n = 421

l

82.3 ± 3.0%, n = 432

r

Merge

Figure 7 Accumulation of juxtanuclear Arp2/3 complex structures upon loss of ARHGAP10 expression. Hela cells treated for 72 h with control siRNA (siGFP, a–f) or ARHGAP10 siRNA (siGAP10#1, g–l, and siGAP10#2, m–r) were fixed and immunostained for giantin (a, g, m, d, j, p) together with the p16Arc (b, h, n) or the p34Arc (e, k, q) subunit of the Arp2/3 complex. Images are from a single confocal plane. Confocal sections were recorded the same day using identical settings of the confocal microscope. Pixel intensities are directly comparable and indicative of the accumulation of the proteins in the perinuclear region. Percentages represent the percentage of cells showing a perinuclear accumulation of p16Arc (b, h, n) or p34Arc (e, k, q) (n represents the total number of cells examined from two independent experiments). Merged images show p16Arc (c, i, o) and p34Arc (f, l, r) (in red) accumulating in the vicinity of Golgi elements labelled with giantin (in green). Arrows in c and f point to some Arp2/3-postive structures that are visible in the Golgi region of untreated HeLa cells. These results are representative of five independent experiments. Scale bars, 10 µm.

network in a region corresponding to the position of the Golgi complex and even extending in the cytoplasm beyond the Golgi region. We observed that the Arp2/3 complex network filled the spaces between fragmented Golgi ministacks (labelled with giantin) in ARHGAP10depleted cells (Fig. 7, merged images). These data suggest that loss of 360

ARHGAP10 results in a massive activation/recruitment of the Arp2/3 complex in the juxtanuclear region, probably as a consequence of the lack of GTP hydrolysis on Golgi-localized Cdc42. Finally, we investigated whether ARF1 may affect the Arp2/3 complex and F-actin dynamics through its action on the ARHGAP10/Cdc42 pathway. Staining of ARF1Q71L-expressing HeLa cells for p16Arc, p34Arc and F-actin revealed the formation of cytoplasmic structures in 55 ± 4% of transfected cells (Fig. 8m), and the structures were positive for the three markers and were heterogeneous both in shape and size (see Fig. 8a–h). ARF1Q71L-induced structures, typically one to five per cell, had a comet-like shape (as shown in Fig. 8a–h), or appeared as rings or dots, which were strongly positive for Arp2/3 complex and F-actin. Some of these structures were observed in the vicinity of the Golgi apparatus (see Fig. 8, merge images in d and h), whereas others were present in the cytoplasm at some distance from the Golgi complex. We did not detect giantin on these structures. That ARF1Q71L-induced structures required Cdc42 activity for their formation was demonstrated by two converging observations. First, co-expression of ARF1Q71L and wild-type Cdc42 increased significantly (P < 0.0001) both the frequency of cells showing these structures (reaching a maximum value of 79 ± 2%; Fig. 8m), and the number of structures per cell (see Fig. 8i–l). Second, co-expression of dominant-inhibitory Cdc42T17N significantly impaired (P < 0.0001) formation of Arp2/3 complex positive structures in ARF1Q71L-expressing cells (18 ± 6%; Fig. 8m). Expression of Cdc42 or Cdc42T17N alone did not allow formation of such structures. The most significant inhibition was obtained on co-expression of ARF1Q71L and a construct comprising both the ARF-BD and Rho-GAP domains of ARHGAP10 (ARF-BD/ Rho-GAP; construct 5, Fig. 1a), which localized to the Golgi complex and almost completely abolished ARF1Q71L structure formation (6 ± 1% of doubly transfected cells had Arp2/3-positive structures; Fig. 8m). Interestingly, expression of Golgi-localized ARF-BD, alone or in fusion with a dead Rho-GAP domain (Rho-GAPR1183A), led to an intermediary inhibition, as did expression of the isolated wild-type cytosolic Rho-GAP domain of ARHGAP10 (see Fig. 8m). Our interpretation of these results is that full inhibition of structure formation by the ARHGAP10 ARFBD/Rho-GAP construct results from a dual action of ARF-BD and of the Rho-GAP domain. First, ARF-BD, by competing with ARF1 effectors — including ARHGAP10, AP1 (see Fig. 3s, t) and some other effector(s) required for the formation of Arp2/3-mediated structures (such as COPI that recruits Cdc42 on the Golgi) — for binding to GTP-ARF1 on Golgi membranes, may interfere with the formation of the Arp2/3 complex structures downstream of ARF1Q71L. Second, the Rho-GAP domain that very probably promotes GTP hydrolysis on endogenous Cdc42, may negatively regulate Cdc42 downstream pathway(s). DISCUSSION This study reveals that ARF1 has the ability to control the Arp2/3 complex and F-actin dynamics on Golgi membranes in vivo, confirming and extending previous observations in cell-free systems9,10. This is clearly illustrated by the formation of comet-like as well as more globularshaped Arp2/3 and F-actin-rich structures in the cytoplasm of cells expressing active ARF1. These structures, which probably represent an extreme situation because of the constitutively activated state of the ARF1 mutant, are reminiscent of Arp2/3 complex and F-actin accumulations detected in the perinuclear Golgi/trans-Golgi network (TGN) region of normal cells (see Fig. 7 and refs 24, 25). Another important NATURE CELL BIOLOGY VOLUME 7 | NUMBER 4 | APRIL 2005

©2005 Nature Publishing Group

print ncb1244.indd 360

16/3/05 1:29:30 pm

A RT I C L E S a

ARF1Q71L

b

F-actin c

p16Arc

e

ARF1Q71L

f

F-actin g

p34Arc

d

h

i

j

ARF1Q71L

Cdc42 k

m

p34Arc

l

(n = 232) (n = 194) (n = 191) (n = 183) (n = 191) (n = 379) (n = 434)

Rho-GAPR1183A Rho-GAP ARF-BD ARF-BD/Rho-GAPR1183A ARF-BD/Rho-GAP Cdc42T17N Cdc42

ARF1Q71L + Rho-GAPR1183A ARF1Q71L + Rho-GAP

* (n = 441) ** (n = 457)

ARF1Q71L + ARF-BD ARF1Q71L + ARF-BD/Rho-GAPR1183A ARF1Q71L + ARF-BD/Rho-GAP

*** (n = 453) ** (n = 458) *** (n = 394) *** (n = 415)

ARF1Q71L + Cdc42T17N ARF1Q71L + Cdc42 ARF1Q71L

*** (n = 415) (n = 543) 0

10

20

30

40

50

60

70

80

90

100

Cells with Arp2/3 structures (%)

Figure 8 Expression of ARF1Q71L promotes the assembly of F-actin and Arp2/3 complex structures that depend on Cdc42 and ARHGAP10 activity. (a–l) HeLa cells were transfected with a construct encoding ARF1Q71L, fixed after 24 h and then stained with anti-HA antibodies to visualize ARF1Q71L (a, e, i) in combination with fluorescently labelled phalloidin (b, f), and antibodies against the p16Arc (c) or p34Arc (g) subunit of the Arp2/3 complex. A series of z-planes (0.2-µm increment steps) corresponding to the entire cell thickness were collected and deconvoluted as described in the Methods section. Images show deconvoluted stacked z-planes. Panels d and h show merged images of the boxed areas in a–c and e–g, respectively, with F-actin in green, Arp2/3 complex subunits in red and ARF1Q71L-HA in blue. (i–l) HeLa cells were cotransfected with constructs encoding for HA-tagged ARF1Q71L and GFP-tagged Cdc42, fixed after 24 h and stained with anti-HA to detect ARF1Q71L (i). (j) GFP–Cdc42 signal.

(k) Staining with antibody against p34Arc. (l) Merge signals for ARF1Q71L and p34Arc corresponding to the boxed area in i and k. Scale bars, 10 µm. (m) HeLa cells were transfected with a construct encoding for HA-tagged ARF1Q71L, alone or in combination with the indicated constructs. Cells were fixed 24 h post-transfection and double-stained with anti-HA and antip34Arc antibodies. The percentage of ARF1Q71L-expressing cells showing one or more p34Arc-labelled structure(s) was determined for each plasmid combination on duplicate samples in two independent experiments. The upper part of the graph shows the percentage of cells expressing the indicated construct with one or more p34Arc-labelled structure(s). Numbers in parentheses correspond to the total number of ARF1Q71L cells examined in each condition. Data represent the means ± s.d. of three independent experiments. Statistical analysis was derived from the Student’s t-test (***P < 0.0001, **P < 0.001, *P < 0.01).

aspect of this study is that ARF1 exerts its control on the Arp2/3 complex and F-actin dynamics through the regulation of the GDP/GTP cycle of Cdc42. Our results identify ARHGAP10 as a GAP acting preferentially on Cdc42 and suggest that GTP-ARF1, by interacting with ARHGAP10, is able to control GTP hydrolysis on Cdc42 at Golgi

membranes. Our observation that induction of the Arp2/3 complex and actin recruitment by active ARF1 can be enhanced upon overexpression of wild-type Cdc42, while it is blocked by dominant inhibitory Cdc42T17N, suggests that ARF1 may regulate a pathway leading to GTPloading on Cdc42 through an as yet uncharacterized Golgi-localized

NATURE CELL BIOLOGY VOLUME 7 | NUMBER 4 | APRIL 2005

361

©2005 Nature Publishing Group

print ncb1244.indd 361

16/3/05 1:29:34 pm

A RT I C L E S Cdc42 GEF(s). These observations, together with the fact that Cdc42 associates with Golgi membranes in an ARF1-dependent manner through interaction with γ-COP5,6, reveals a complex interplay between both GTPases in the control of Golgi function. Interestingly, overexpression of a constitutively active mutant form of ARF6 was shown to induce F-actin structures that required Arp2/3 complex activity, similar to the ones we describe here26. However, in contrast to our findings, formation of ARF6-induced structures was independent of Cdc42 activity26, indicating that the pathways involved downstream of the two related ARF proteins leading to Arp2/3 complex activation are distinct. Rho proteins have been shown to regulate vesicle fusion and fission events through their regulatory role on the actin cytoskeleton3. Binding of Cdc42 to COPI-coated membranes is thought to control local actin assembly through a pathway involving its downstream effector N-WASP and the Arp2/3 complex, and is involved in retrograde transport of the KDEL receptor6,9–11,21. One study also links Arp2/3 complex and F-actin recruitment to the transport of clathrin-coated vesicles (CCVs) from the TGN24. Budding of CCVs from the TGN is controlled by GTP-ARF1 interacting with adaptor protein complexes and the GGAs1,27. In addition, in epithelial cells, exit from the TGN is regulated by Cdc42 (ref. 28). Whether ARF1 and Cdc42 cooperate to regulate exit from the TGN is not yet known. However, based on the broad Golgi distribution of ARHGAP10, crosstalk between ARF1 and Cdc42 signalling at the level of the TGN is possible. The extensive Arp2/3 complex network formed in the peri-Golgi space of ARHGAP10-depleted cells is likely to be the consequence of a lack of GTP hydrolysis on Golgi-associated Cdc42, triggering an ‘on’ signal for Arp2/3 complex activation. Accumulation of Arp2/3 complex structures around Golgi membranes may affect Golgi stack organization and dynamics, in keeping with in vitro studies that show that actin can affect the release of COPI-coated vesicles10. Although its precise role remains to be fully clarified, Arp2/3 complex-dependent actin polymerization at the Golgi complex is thought to facilitate budding of vesicles, a function that has clearly been attributed to ARF1. The present study, which identifies ARHGAP10 as a downstream partner of ARF1, reveals a new aspect of the regulation of Golgi structure and function by ARF1 through the control of Cdc42 activity and actin cytoskeleton dynamics within the Golgi complex. METHODS Reagents, drugs and cell culture. Egg phosphatidylcholine, liver phosphatidylethanolamine, brain phosphatidylserine and brefeldin A were purchased from Sigma (St Louis, MO). Brain PtdIns(4,5)P2 and PtdIns(3,4,5)P3 were from Avanti Polar Lipids (Birmingham, AL). Synthetic dipalmitoyl PtdIns(3)P was from Echelon (Biosciences Inc., Salt Lake City, UT). Nucleotides were from Sigma. Sypro Orange gel protein staining was from BioRad (Hercules, CA). Texas-Red, Alexa-488 and Alexa-546 phalloidin were purchased from Molecular Probes (Eugene, OR). MCF-7 human breast adenocarcinoma cells (from the European Collection of Cell Cultures, Salisbury, UK) and HeLa cells (a gift from A. Dautry, Institut Pasteur, Paris, France) were grown in DMEM supplemented with 10% fetal calf serum, 2 mM glutamine and 150 µg ml−1 penicillin/streptomycin at 37 °C in 7% CO2. When indicated, HeLa cells were treated with brefeldin A (10 µg ml−1, Sigma) for the indicated time period at 37 °C, then fixed and processed for immunofluorescence analysis. Antibodies. Polyclonal antibodies against ARHGAP10 were generated against a synthetic peptide (STREIATTDTPLSLHCNT) corresponding to residues 1878– 1895 of the human protein. Anti-ARHGAP10 antibodies from rabbit sera were purified by affinity chromatography onto a column of ARHGAP10 peptide immobilized on HiTrap NHS-activated (Amersham Pharmacia biotech, Piscataway, NJ).

362

Mouse monoclonal antibody against the p16Arc (clone 323H3) subunits of the Arp2/3 complex was described earlier23. The monoclonal antibody (subtype IgG1, clone 334D4) recognizing the p34 subunit of Arp2/3 complex (ArpC2) was raised using purified bovine Arp2/3 complex (provided by T. Pollard, Yale) as described elsewhere29 (323H3 and 334D4 monoclonal antibodies are available from Synaptic Systems, Göttingen, Germany). Rat monoclonal antibody against HA tag (clone 3F10) was purchased from Roche Diagnostics GmbH (Mannheim, Germany). AntiGM130 (clone 35), -AP1, -EEA1 and -clathrin heavy chain monoclonal antibodies were obtained from BD Transduction Laboratories (San Jose, CA). Mouse monoclonal antibodies against RhoA (clone 26C4) and CTR433 were gifts of J. Bertoglio (Inserm U461, Chatenay-Malabry, France) and M. Bornens (Institut Curie, Paris, France), respectively. Polyclonal anti-giantin antibodies were raised against the N-terminal domain of mouse giantin (52–701) and affinity purified. Polyclonal antibodies against c-Myc tag were obtained from Santa Cruz Biotechnology (Santa Cruz, CA). Polyclonal antibodies against ARF6 were a gift from V. W. Hsu (Brigham and Women’s Hospital, Boston, USA). Monoclonal anti-β-COP (clone maD) was purchased from Sigma. Cy2-, Cy3- and Cy5-conjugated secondary antibodies were purchased from Jackson ImmunoResearch (West Grove, PA). Alexa-conjugated secondary antibodies were purchased from Molecular Probes. Subcellular fractionation. HeLa cells were scrapped from the dish with a rubber policeman and resuspended in 1 ml hypotonic buffer (20 mM Tris pH 7.4, 5 mM MgCl2, 1 mM EDTA, 250 mM sucrose, 1 mM dithiothreitol (DTT) with a cocktail of protease inhibitors (Roche)) by repeated passages through a 26G needle. The homogenate was subjected to a 1,000g spin to yield a postnuclear supernatant. The postnuclear supernatant was subjected to a 100,000g spin to yield soluble and pellet fractions. Pellets were resuspended in sample buffer to volumes equal to that of the corresponding supernatants and equal volumes of each fraction were loaded per lane (corresponding to 25 µg of total protein from the soluble fraction). Proteins were separated on 7.5% polyacrylamide gels and transferred to PVDF membranes. Detection of endogenous ARHGAP10, RhoA and transferrin receptor was performed with the enhanced chemiluminescence (ECL) procedure (Amersham Pharmacia biotech). Yeast two-hybrid screening. Two-hybrid screening was performed using a mating protocol with L40 and Y187 yeast strains using a fusion of LexA with ARF6Q67L as bait, and a human placenta random-primed cDNA library30. DNA constructs and transfection. The plasmids pGEX-RhoA, pGEX-Rac1 and pGEX-Cdc42 were provided by R. Cerione (Cornell University, NY). Constructs encoding Myc-Cdc42Q61L, Myc-RhoAQ63L and GST–p50Rho-GAP (198–439) were obtained from A. Hall (University College London, London, UK). DNA sequences encoding residues 885–1096 (ARF-BD), 929–1052 (PH), 885–1052, 929–1096, 1158–1346, 929–1346 (ARF-BD/Rho-GAP) and 1064–1346 (RhoGAP) of human ARHGAP10 were obtained by PCR using the KIAA1424 cDNA as a template31, and the multi-cloning sites of pGEX4T1 (Amersham Pharmacia Biotech) and pEGFP-C2 (Clontech) were inserted to give in-frame fusion with the N-terminal GST or EGFP tag. All constructs were verified by double-stranded DNA sequencing. Constructs encoding HA-tagged ARF1 and ARF6 variants have been described30. Constructs encoding GFP-tagged Cdc42 and Cdc42T17N were obtained from P. Fort (CNRS, Montpellier, France). HeLa cells plated onto coverslips were transfected using the calcium phosphate procedure. Cells were processed for immunofluorescence studies 16–24 h after transfection. RNA interference. Two independent ARHGAP10 siRNAs were used named siGAP10#1 and siGAP10#2. siGAP10#1 consisted of a mixture of four siRNA duplexes designed from the human ARHGAP10 cDNA sequence (Smartpool M-004775-00-50, Dharmacon, Lafayette, CO) with the following sequence: duplex 1, 5′-TAAAGAAGCTGTCATCCTA-3′ (position 2627–2645 of ARHGAP10 cDNA, accession number NM_020824); duplex 2, 5′-GGAGACAGCTCTTCAGTTC-3′ (position 5497–5515); duplex 3, 5′-AAAGGAAACTTCTCAGTAA-3′ (position 2748–2766); duplex 4, 5′-GTAAATCACTTGCATCAGA-3′ (position 2388–2406). siGAP10#2 corresponded to a single duplex with sequence: 5′-GGATCTGTGTCGCAGTTTA3′ (position 2166–2184). An siRNA duplex designed to target GFP (siGFP, 5′-GAACGGCATCAAGGTGAAC-3′) was used as a control (siGAP10#2 and siGFP were purchased from Proligo (Paris, France).

NATURE CELL BIOLOGY VOLUME 7 | NUMBER 4 | APRIL 2005 ©2005 Nature Publishing Group

print ncb1244.indd 362

16/3/05 1:29:36 pm

A RT I C L E S HeLa cells were plated at 20% confluence and transfected with indicated siRNAs (320 nM) using Oligofectamine reagent according to the manufacturer’s instruction (Invitrogen, Carlsbad, CA). To measure ARHGAP10 depletion after 72 h siRNA treatment, HeLa cells were lysed in RIPA buffer (50 mM Tris pH 7.5, 150 mM NaCl, 1% Triton X-100, 0.5% deoxycholate acid, 0.1% SDS, 2 mM EDTA, 1 mM DTT) containing protease inhibitors. Equal amounts of proteins from mock or siRNA-treated cells were loaded on a 7.5% polyacrylamide gel and transferred to PVDF membrane, and equal loading was verified by immunoblotting with anti-clathrin heavy chain antibody. ARHGAP10 depletion was measured by immunoblotting with anti-ARHGAP10 polyclonal antibodies. After detection of bound antibodies with the ECL procedure, films were scanned and the intensity of ARHGAP10 signals estimated using the NIH Image software. The residual level of ARHGAP10 in siGAP10#1- and siGAP10#2-treated cells was 8.6 ± 3.8% (six independent experiments) and 3.9 ± 1.8% (four independent experiments) of the level in siGFP-treated cells, respectively. ARHGAP10 level in siGFP-treated cells was not significantly different as compared with mock-treated cells (data not shown). Indirect immunofluorescence, confocal and deconvolution microscopy. HeLa cells plated onto coverslips were fixed in 4% paraformaldehyde and processed for immunofluorescence microscopy as described previously30. Confocal optical sections were taken with a Leica TCS SP2 confocal microscope equipped with the AOTF module to prevent cross-contamination between fluorochromes. We took z-series of images (6 to 12 planes) at 0.3-µm increments. For deconvolution microscopy, cells were examined under a motorized upright wide-field microscope (Leica DMRA2) equipped for image deconvolution. Acquisition was performed using an oil immersion objective (×100 PL APO HCX, 1.4 NA) and a high-sensitive cooled interlined CCD (charge-coupled device) camera (Roper CoolSnap HQ). Z positioning was accomplished using a piezo-electric motor (LVDT, Physik Instrument, Auburn, MA) mounted underneath the objective lens. The system was steered by Metamorph 5.0.7 Software (Universal Imaging Corporation, Downingtown, PA). z-series of images (12 to 15 planes) were taken at 0.2-µm increments. Deconvolution was performed by the 3D deconvolution Metamorph module, using the fast Iterative Constrained PSF-based algorithm32. Electron microscopy. Cells were fixed with 2% glutaraldehyde in 0.2 M HEPES (pH 7.2) for 1 h, then cells were postfixed with reduced osmium tetroxide for 1 h, dehydrated in alcohols and embedded in Epon (Fluka, Buchs, Switzerland). Sections of 70 nm were cut on Leica ultramicrotome (Leica, Vienna, Austria) and contrasted with lead citrate. Sections were analysed with a Tecnai 12 electron microscope (FEI, Eindhoven, The Netherlands) and pictures were taken with a CCD camera. Measuring of cisternae length was performed using analySIS software (Soft Imaging Systems, Munster, Germany). Protein purification. pGEX plasmids were transformed in Escherichia coli BL21(DE3) strain and GST-fusion proteins were produced and purified with standard procedures. Recombinant ARF6 loaded with GDP or GTP were purified as previously described30. Myristoylated ARF1 was produced in E. coli by co-expression with yeast N-myristoyltransferase and purified as described33. ∆17-ARF1 was expressed in E. coli and purified on an ACA44 (Ultrogel; IBF Biotechnics France, Villeneuve la Garenne, France) gel filtration column33. Binding assays. HeLa cells were transfected with plasmids encoding HA-tagged ARF1 or ARF6 variants using the calcium-phosphate procedure. After 16–20 h, cells were lysed in 50 mM Tris pH 7.4, 137 mM NaCl, 10 mM MgCl2, 10% glycerol, 1% Triton-X100 and a cocktail of protease inhibitors (Roche) and centrifuged at 15,000g for 5 min at 4 °C. Supernatants were incubated with 2 µM GST or GST–ARF-BD for 1 h at 4 °C in the presence of 0.1% BSA. Then, glutathioneSepharose beads were added for a further 1 h. Beads were washed and bound proteins were eluted using SDS sample buffer and separated by SDS–PAGE. The presence of ARF in the bound material was detected by western blot analysis using anti-HA monoclonal antibody. For mapping the region that interacts with active ARF, different regions of ARHGAP10 fused with GST (2 µM) were incubated with lysates from cells over-expressing HA-tagged ARF6Q67L. For direct binding assay with purified recombinant proteins, GDP- or GTPbound ARF6 (4 µM) was incubated for 1 h at 4 °C in the presence of 1 µM GST or GST–ARF-BD immobilized onto glutathione-Sepharose beads in binding buffer (50 mM Tris-HCl pH 7.5, 100 mM NaCl, 2 mM MgCl2, 1 mM DTT, 0.5% Triton-X100) containing 100 µM GDP or GTP, and 0.5% BSA (w/v). Beads were

washed and proteins were separated by SDS–PAGE. Bound ARF6 was detected by western blot analysis using anti-ARF6 antibodies. An aliquot of input ARF6 (1%) was loaded on the gel as a control. Liposome binding assays. Vesicles of defined phospholipid composition were prepared by extrusion through a 0.1-µm pore polycarbonate filter as previously described33. The concentration of each phospholipid is indicated in the legend of Fig. 4. Purified ARF1 or ∆17-ARF1 (3 µM) were first incubated at 30 °C for 40 min in the presence of 1 mM GTP and 1 µM free Mg2+ in the vesicle suspension (0.8 g l−1). Then, the free magnesium concentration was raised to 1 mM, and ARFBD (2 µM, cleaved from the GST moiety with thrombin) was added for 30 min at 25 °C. Samples were centrifuged for 25 min at 400,000g at 25 °C. Supernatants were removed, and pellets resuspended in 20 mM HEPES pH 7.5 and 100 mM NaCl. Equal volumes of pellet and supernatant fractions were analysed on 15% SDS–PAGE and densitometry after Sypro Orange staining. Rho-GAP assay. Purified GST–Rho proteins (1 µM) were loaded with 15 µM 32 P-γ-GTP in 50 mM HEPES (pH 7.5), 100 mM KCl, 2 mM MgCl2, 5 mM Pi and 1 mM DTT for 15 min (30 min for RhoA) at 37 °C. At the initiation of GTP hydrolysis, the 32P-γ-GTP-loaded protein represented 8–30% of the total protein. Then, GST (100 nM), or GST fused to the Rho-GAP domains of ARHGAP10 or p50Rho-GAP were added at the indicated concentrations. Aliquots (25 µl) were removed at the indicated incubation time and diluted into 2 ml of ice-cold ‘stop buffer’ (50 mM HEPES pH 7.5, 100 mM NaCl, 10 mM MgCl2) and immediately filtered on BA85 nitrocellulose membranes (Schleicher and Schuell, Perkin Elmer France, Courtaboeuf, France). Filters were washed twice with 2 ml of the same buffer, dried, and bound 32P-γ-GTP was quantified by scintillation counting. The kinetics of GTP hydrolysis were fitted as exponentials (y = a(−Kapp × t) + b) where Kapp represents the apparent rate constant of GTP hydrolysis (min−1). Plotting Kapp as a function of the concentration of the Rho-GAP domain showed a linear relationship from which the specific activity was determined (in min−1 nM−1). BIND identifiers. Five BIND identifiers (www.bind.ca) are associated with this manuscript: 217712, 217713, 217785, 217786 and 217787. Note: Supplementary Information is available on the Nature Cell Biology website. ACKNOWLEDGEMENTS We are indebted to the Kazusa DNA Research Institute for the gift of the cDNA clone KIAA1424 and to D. Alessi for performing lipid overlay assay. We are grateful to E.G. Berger, J. Bertoglio, M. Bornens, V. W. Hsu, J. Wehland and I. Majoul for providing antibodies, and to D. Cassel, R. Cerione, A. Hall and P. Fort for plasmids. We would like to thank F. Perez and R. Pepperkok for helpful discussions throughout this study, and J. Plastino for critical reading of the manuscript. We also thank A. Letellier and C. Boulard for technical assistance. T.D. is the recipient of a post-doctoral fellowship from ARC (Association pour la Recherche sur le Cancer). This work was supported by an EC Training Network Grant HRPN-CT-2000-00081 (to P.C. and M.A.D.M.). Work in M.A.D.M.’s laboratory is supported by the Italian Association for Cancer Research (AIRC), Telethon Italia and the Italian Ministry for Universities and Research (MIUR). Work in P.C.’s laboratory is supported by grants from the CNRS, Institut Curie, La Fondation BNP-Paribas and La Ligue Nationale contre le Cancer (‘équipe labellisée’). COMPETING FINANCIAL INTERESTS The authors declare that they have no competing financial interests. Received 22 December 2004; accepted 25 February 2005 Published online at http://www.nature.com/naturecellbiology. 1. Chavrier, P. & Goud, B. The role of ARF and rab GTPases in membrane transport. Curr. Opin. Cell Biol. 11, 466–475 (1999). 2. De Matteis, M. A. & Morrow, J. S. Spectrin tethers and mesh in the biosynthetic pathway. J. Cell Sci. 113 (Pt 13), 2331–2343 (2000). 3. Stamnes, M. Regulating the actin cytoskeleton during vesicular transport. Curr. Opin. Cell Biol. 14, 428–433 (2002). 4. Nie, Z., Hirsch, D. S. & Randazzo, P. A. Arf and its many interactors. Curr. Opin. Cell Biol. 15, 396–404 (2003). 5. Wu, W. J., Erickson, J. W., Lin, R. & Cerione, R. A. The gamma-subunit of the coatomer complex binds Cdc42 to mediate transformation. Nature 405, 800–804 (2000). 6. Luna, A. et al. Regulation of protein transport from the Golgi complex to the endoplasmic reticulum by CDC42 and N-WASP. Mol. Biol. Cell 13, 866–879 (2002). 7. Miura, K. et al. ARAP1: a point of convergence for Arf and Rho signaling. Mol. Cell 9, 109–119 (2002). 8. Camera, P. et al. Citron-N is a neuronal Rho-associated protein involved in Golgi

NATURE CELL BIOLOGY VOLUME 7 | NUMBER 4 | APRIL 2005

363

©2005 Nature Publishing Group

print ncb1244.indd 363

16/3/05 1:29:39 pm

A RT I C L E S organization through actin cytoskeleton regulation. Nature Cell Biol. 5, 1071–1078 (2003). 9. Chen, J. L., Lacomis, L., Erdjument-Bromage, H., Tempst, P. & Stamnes, M. Cytosolderived proteins are sufficient for Arp2/3 recruitment and ARF/coatomer-dependent actin polymerization on Golgi membranes. FEBS Lett. 566, 281–286 (2004). 10. Fucini, R. V. et al. Activated ADP-ribosylation factor assembles distinct pools of actin on golgi membranes. J. Biol. Chem. 275, 18824–18829 (2000). 11. Fucini, R. V., Chen, J. L., Sharma, C., Kessels, M. M. & Stamnes, M. Golgi vesicle proteins are linked to the assembly of an actin complex defined by mAbp1. Mol. Biol. Cell 13, 621–631 (2002). 12. Basseres, D. S., Tizzei, E. V., Duarte, A. A., Costa, F. F. & Saad, S. T. ARHGAP10, a novel human gene coding for a potentially cytoskeletal Rho-GTPase activating protein. Biochem. Biophys. Res. Commun. 294, 579–585 (2002). 13. Dascher, C. & Balch, W. E. Dominant inhibitory mutants of ARF1 block endoplasmic reticulum to Golgi transport and trigger disassembly of the Golgi apparatus. J. Biol. Chem. 269, 1437–1448 (1994). 14. Zhang, C. J. et al. Expression of a dominant allele of human ARF1 inhibits membrane traffic in vivo. J. Cell Biol. 124, 289–300 (1994). 15. Donaldson, J. G. & Jackson, C. L. Regulators and effectors of the ARF GTPases. Curr. Opin. Cell Biol. 12, 475–482 (2000). 16. Peyroche, A. et al. Brefeldin A acts to stabilize an abortive ARF-GDP-Sec7 domain protein complex: involvement of specific residues of the Sec7 domain. Mol. Cell 3, 275–285 (1999). 17. Robinson, M. S. & Kreis, T. E. Recruitment of coat proteins onto Golgi membranes in intact and permeabilized cells: effects of brefeldin A and G protein activators. Cell 69, 129–138 (1992). 18. Teal, S., Hsu, V., Peters, P., Klausner, R. & Donaldson, J. An activating mutation in ARF1 stabilizes coatomer binding to Golgi membranes. J. Biol. Chem. 269, 3135– 3138 (1994). 19. Antonny, B., Beraud-Dufour, S., Chardin, P. & Chabre, M. N-terminal hydrophobic residues of the G-protein ADP-ribosylation factor-1 insert into membrane phospholipids upon GDP to GTP exchange. Biochemistry 36, 4675–4684 (1997).

364

20. Nobes, C. D. & Hall, A. Rho, rac, and cdc42 GTPases regulate the assembly of multimolecular focal complexes associated with actin stress fibers, lamellipodia, and filopodia. Cell 81, 53–62 (1995). 21. Valderrama, F. et al. Actin microfilaments facilitate the retrograde transport from the Golgi complex to the endoplasmic reticulum in mammalian cells. Traffic 2, 717–726 (2001). 22. Percival, J. M. et al. Targeting of a tropomyosin isoform to short microfilaments associated with the Golgi complex. Mol. Biol. Cell 15, 268–280 (2004). 23. Olazabal, I. M. et al. Rho-kinase and myosin-II control phagocytic cup formation during CR, but not FcγR, phagocytosis. Curr. Biol. 12, 1413–1418 (2002). 24. Carreno, S., Engqvist-Goldstein, A. E., Zhang, C. X., McDonald, K. L. & Drubin, D. G. Actin dynamics coupled to clathrin-coated vesicle formation at the trans-Golgi network. J. Cell Biol. 165, 781–788 (2004). 25. Matas, O. B., Martinez-Menarguez, J. A. & Egea, G. Association of Cdc42/N-WASP/ Arp2/3 signaling pathway with Golgi membranes. Traffic 5, 838–846 (2004). 26. Schafer, D. A., D’Souza-Schorey, C. & Cooper, J. A. Actin assembly at membranes controlled by ARF6. Traffic 1, 896–907 (2000). 27. Bonifacino, J. S. The GGA proteins: adaptors on the move. Nature Rev. Mol. Cell Biol. 5, 23–32 (2004). 28. Musch, A., Cohen, D., Kreitzer, G. & Rodriguez-Boulan, E. cdc42 regulates the exit of apical and basolateral proteins from the trans-Golgi network. EMBO J. 20, 2171–2179 (2001). 29. Niebuhr, K., Lingnau, A., Frank, R. & Wehland, J. in Cell Biology: A Laboratory Handbook (ed. Celis, J. E.) 398–403 (Academic, San Diego, 1998). 30. Prigent, M. et al. ARF6 controls post-endocytic recycling through its downstream exocyst complex effector. J. Cell Biol. 163, 1111–1121 (2003). 31. Kikuno, R. et al. HUGE: a database for human KIAA proteins, a 2004 update integrating HUGEppi and ROUGE. Nucleic Acids Res. 32, D502–D504 (2004). 32. Sibarita, J.-B., Magnin, H. & De Mey, J. Proc. 2002 IEEE International Symposium on Biomedical Imaging 769–772 (2002). 33. Macia, E., Chabre, M. & Franco, M. Specificities for the small G proteins ARF1 and ARF6 of the guanine nucleotide exchange factors ARNO and EFA6. J. Biol. Chem. 276, 24925–24930 (2001).

NATURE CELL BIOLOGY VOLUME 7 | NUMBER 4 | APRIL 2005 ©2005 Nature Publishing Group

print ncb1244.indd 364

16/3/05 1:29:43 pm

S U P P L E M E N TA R Y I N F O R M AT I O N

Figure S1 Recruitment of the RhoGAP domain of ARHGAP10 to the Golgi complex induces morphological changes. HeLa cells were transiently transfected with constructs encoding GFP-tagged ARFBD/RhoGAP (construct #5, see Fig. 1a) (a-b), ARFBD/RhoGAP(R1183A) harbouring an arginineto-alanine mutation in a conserved catalytic residue in the RhoGAP domain (c-d), or the isolated RhoGAP domain (construct #6 in Fig. 1a, panels

e-f). After 30 h, cells were fixed and labelled with fluorescently-labelled phalloidin (right panels). GFP signals are shown on the left panels. Arrows in panel (b) point to ARFBD/RhoGAP-expressing cells with a disorganized actin cytoskeleton. The results presented in this figure are representative of three independent experiments. Bars, 10 µm.

WWW.NATURE.COM/NATURECELLBIOLOGY

1

© 2005 Nature Publishing Group

S U P P L E M E N TA R Y I N F O R M AT I O N

Figure S2 Expression of constitutively active Cdc42 induces cell spreading and stress fibre formation in HeLa cells. HeLa cells were transiently transfected with a construct encoding for myc-tagged Cdc42-Q61L (a and

b) or RhoA-Q63L (c and d). After 24 h, cells were fixed and stained with anti-myc antibodies (a and c) and fluorescently-labelled phalloidin (b and d). Transfected cells are indicated by arrows. Bars, 10 µm.

2

WWW.NATURE.COM/NATURECELLBIOLOGY

© 2005 Nature Publishing Group