Polymerization of actin filaments directed by the actin-related protein (Arp)2/3 complex supports many types of cellular movements1. However, questions remain regarding the relative contributions of Arp2/3 complex versus other mechanisms of actin filament nucleation to processes such as path finding by neuronal growth cones; this is because of the lack of simple methods to inhibit Arp2/3 complex reversibly in living cells. Here we describe two classes of small molecules that bind to different sites on the Arp2/3 complex and inhibit its ability to nucleate actin filaments. CK-0944636 binds between Arp2 and Arp3, where it appears to block movement of Arp2 and Arp3 into their active conformation. CK-0993548 inserts into the hydrophobic core of Arp3 and alters its conformation. Both classes of compounds inhibit formation of actin filament comet tails by Listeria and podosomes by monocytes. Two inhibitors with different mechanisms of action provide a powerful approach for studying the Arp2/3 complex in living cells.
We used fluorescence assays to screen a library of >400,000 small molecules for inhibitors of actin polymerization stimulated by human (Homo sapiens, Hs) Arp2/3 complex and either full-length human Wiskott–Aldrich syndrome protein (HsWASP) with acrylodan-actin or HsWASP residues 105–502 with pyrenyl-actin. These screens each identified two inhibitors of human and bovine (Bos taurus, Bt) Arp2/3 complex, namely CK-0944636 (abbreviated CK-636) and CK-0993548 (abbreviated CK-548) (Fig. 1a). These compounds inhibited BtArp2/3 complex with IC50 (half-maximal inhibitory concentration) values of 32 µM for CK-636 and 11 μM for CK-548 (Fig. 1c). CK-636 inhibited actin polymerization stimulated by fission yeast (Schizosaccharomyces pombe, Sp) Arp2/3 complex (IC50 = 24 μM), but 100 µM CK-548 did not (Fig. 1c, Supplementary Table 1). Fluorescence microscopy of the products of these reactions stained with Alexa 488 phalloidin showed branched actin filaments in controls (Fig. 1e, left panel). Samples with 100 μM CK-636 contained fewer branched filaments (Fig. 1e, centre panel), while samples with 100 μM CK-548 contained only unbranched filaments (Fig. 1e, right panel). We tested a number of compounds structurally related to CK-548 or CK-636 that had no effect on actin polymerization at concentrations up to 200 μM and are useful as controls for experiments with cells (Fig. 2g). Supplementary Table 2 lists one inactive compound from each class.
The following control experiments showed that both compounds interact with Arp2/3 complex rather than actin or WASP. At 20 μM, neither compound inhibited polymerization of actin alone or polymerization stimulated by the FH2 domain of formin Cdc12p (Cdc12(FH2); Fig. 1b), so neither compound interacts directly with actin. As actin can polymerize on its own, neither compound eliminated actin polymerization in mixtures with Arp2/3 complex and WASP or a carboxy-terminal construct of Neural Wiskott–Aldrich syndrome protein (N-WASP-VCA) (Fig. 1c, d). Both compounds inhibited polymerization stimulated by HsArp2/3 complex and ActA, an Arp2/3 complex activating factor from Listeria monocytogenes2, suggesting that they do not act on WASP (Supplementary Fig. 1). A fluorescence anisotropy assay showed that neither compound had a large effect on Arp2/3 complex binding N-WASP-VCA (Supplementary Fig. 2). The affinity of rhodamine-N-WASP-VCA for BtArp2/3 complex was the same with 50 µM CK-636 (Kd = 470 ± 50 nM) as controls (Kd = 510 ± 30 nM), but 50 μM CK-548 decreased the affinity approximately twofold (Kd = 1.0 ± 0.2 μM).
We used a Listeria motility assay to determine whether CK-636 and CK-548 can inhibit actin polymerization mediated by Arp2/3 complex in live cells. Both compounds reduced the formation of actin filament comet tails by Listeria in infected SKOV3 cells (Fig. 2a–c). The concentration dependence of this inhibition gave IC50 values of 22 µM for CK-636 and 31 µM for CK-548. We used the actin polymerization and Listeria comet tail assays to search for more potent inhibitors related to CK-636 and CK-548.
Compound CK-0944666 (abbreviated CK-666) has a fluorobenzene rather than the thiophene ring of CK-636 (Fig. 1a). CK-666 was a better inhibitor of actin polymerization with either BtArp2/3 complex (IC50 = 32 µM for CK-636 verses 17 μM for CK-666) or SpArp2/3 complex (IC50 = 24 µM for CK-636 versus 5 μM for CK-666) (Fig. 1d, Supplementary Table 1). Both CK-636 and CK-666 had IC50 values of 4 μM for inhibiting HsArp2/3 complex. CK-666 reduced actin polymerization around intracellular Listeria to background levels (Fig. 2d, e and g) at lower concentrations than CK-636 (Fig. 2c). Actin filament haloes and comet tails reformed when the compounds were removed after one hour of treatment (Fig. 2f, g). Inactive control compound CK-689 had no effect on comet tail formation (Fig. 2g).
CK-0157869 (abbreviated CK-869) has methoxy groups in the ortho and para positions replacing the ortho-hydroxy and meta-chlorine substituents of CK-548 (Fig. 1a). CK-869 inhibited actin polymerization with BtArp2/3 complex similarly to CK-548 (IC50 = 11 µM for both compounds), but CK869 inhibited comet tail formation by Listeria more effectively (IC50 = 7 μM for CK-869 versus 31 μM for CK-548). Like CK-548, CK-869 did not inhibit either budding or fission yeast Arp2/3 complexes (Fig. 1d).
Both CK-548 and CK-636 also inhibited the formation of podosomes by the THP-1 monocyte cell line (Fig. 2h–k). Podosomes are adhesive structures that depend on Arp2/3 complex, WASP and actin polymerization3. Podosomes also depend on microtubules4, but neither CK-548 or CK-636 at 100 µM disrupted the microtubule network in THP-1 cells (Supplementary Fig. 3). Neither class of compound shows irreversible or off-target morphological effects on cells, such as apoptosis, at concentrations up to 100 μM over 24 h. Our most potent compound, CK-666, had no effect on the mitotic index of SKOV cells at concentrations up to 80 μM, while inhibiting actin assembly around Listeria completely at 10 μM. Fifty micromolar CK-666 or CK-869 each altered the morphology and slightly slowed but did not stop the gliding motility of fish keratocytes (data not shown), as expected, as ∼10% of Arp2/3 complex (∼500 nM) would still be active. Thus Listeria comet tails and monocyte podosomes are more sensitive than lamellar motility to inhibition of Arp2/3 complex.
We determined X-ray crystal structures of BtArp2/3 complex with either CK-636 or CK-548 bound, using molecular replacement with the apo-form BtArp2/3 complex as the search model (PDB accession 1K8K)5. Structures of Arp2/3 complex co-crystallized with the compounds and with compounds soaked into the crystals were nearly identical, but the resolution of soaked crystals was higher (2.70 Å for CK-636 and 2.85 Å for CK-548, Supplementary Table 3). As in previous crystal structures of BtArp2/3 complex5,6,7, none of these structures had appreciable electron density for subdomains 1 and 2 of Arp2, which appears only after chemical crosslinking7.
After one round of rigid body refinement, strong Fo - Fc electron density was present at the interface between Arp2 and Arp3 in maps calculated with data from crystals soaked in 1 mM CK-636 (Fig. 3). This density improved on further refinement (Supplementary Fig. 4a), and CK-636 was modelled into the density when the Rf was 29.2%. CK-636 binds in a pocket between subdomain 4 of Arp2 and subdomain 1 of Arp3. The contact surface is mainly hydrophobic and buries 161 Å2 of the accessible surface of Arp2 and 79 Å2 of Arp3. CK-636 forms hydrogen bonds with two residues in the pocket. The side chain of Asp 248 forms a hydrogen bond with the nitrogen in the indole ring of the inhibitor, and the backbone amide of Ala 203 forms a hydrogen bond with the carbonyl oxygen on the linker between the two ring systems. Both trans and cis conformations of the amide functional group of CK-636 fit into the electron density with the hydrogen bonds described above preserved in both conformations. Because trans-amides are generally 1–5 kcal mol-1 more stable than cis-amides in solution8, and the cis conformation is rarely observed in small molecules, we modelled CK-636 in the trans conformation.
A difference electron density map shows that CK-636 causes minor conformational changes in the complex (Fig. 3b). Arp2 Arg 250 from β-strand 13 rotates out of the pocket and adopts an extended conformation. Arp2 Ala 203 from the loop (residues 199–208) connecting α-helices αE and αF moves towards the pocket, causing the side chain of Leu 198 to adopt a new rotamer. These changes and small changes in the backbone of residues surrounding the pocket result in an overall r.m.s. deviation of 0.63 Å for an overlay of subdomains 3 and 4 (191 Cα atoms) of Arp2 with and without inhibitor bound.
We modelled CK-666 into the CK-636 binding pocket with the aromatic fluorine pointed towards a concave surface in the back wall of the pocket formed by Tyr 202 in Arp2 and residues 118 to 120 in Arp3 (Supplementary Fig. 6). One hundred steps of conjugate gradient minimization in CNS9 showed that CK-666 is stable in this conformation and does not clash with residues in the pocket. The fluorine atom provides additional van der Waals interactions with the back wall of the pocket, and the benzene ring completely fills the hydrophobic pocket created by Ile 252, Tyr 202 and the aliphatic portion of Thr 119. These interactions may allow CK-666 to bind more tightly than CK-636.
During formation of a branch Arp2 moves 31 Å relative to Arp3 from its position in inactive Arp2/3 complex10, so the location of the CK-636 binding pocket between Arp2 and Arp3 suggests that CK-636 and CK-666 lock Arp2/3 complex in an inactive conformation. CK-636 at a concentration of 100 µM did not interfere with N-WASP-VCA binding to Arp2/3 complex (Supplementary Fig. 2), but it prevented WASP-VCA binding from increasing the fluorescence of Arp2/3 complex loaded with etheno-ATP (Supplementary Fig. 5). Thus CK-636 inhibits a conformational change caused by activator binding. This supports the hypothesis that CK-636 locks the complex in an inactive conformation.
The residues contributed by Arp2 and Arp3 to form the CK-636 binding pocket are conserved across a broad range of species (Supplementary Table 4), so we expect CK-636 and CK-666 will be useful for inhibiting Arp2/3 complex from diverse species. These residues are also conserved in actin, but CK-636 does not inhibit actin polymerization at concentrations up to 200 μM. Only half of the binding site is available on an actin monomer, and the two half-sites are not juxtaposed in actin filaments11.
CK-548 has a single stereocentre and was used as a racemic mixture in our experiments. The 2S enantiomer binds in a hydrophobic cavity in the core of the Arp3 subunit (Fig. 4a, b and Supplementary Fig. 4b). The isolated 2R enantiomer, which does not fit in the binding pocket, should provide a useful control for in vivo experiments. Binding of CK-548 requires a substantial conformational change: the loop (residues 76–85) connecting β6 (73–75) and αB (86–98) in subdomain 1 of Arp3 flips upward 7.2 Å, exposing the binding site in the hydrophobic core of Arp3 (Fig. 4b, c). CK-548 contacts 239 Å2 of the inner surface of this cavity, which includes residues Cys 12, Ile 78, Trp 86, Met 89, Met 93 and Leu 112. A hydrogen bond between the side chain amide of Asn 118 and the carbonyl oxygen of CK-548 anchors the inhibitor to the loop between β7 (110–114) and αC (120–132), which forms the bottom lip of the cavity. The ability of CK-548 to bind crystallized Arp2/3 complex suggests that the β6/αB loop can open the hydrophobic binding pocket both in crystals and in solution, even in the absence of inhibitor. The average B-factor for all residues in the loop is 63 Å2 in the closed conformation and 82 Å2 in the open conformation, suggesting that the loop is more flexible when open.
When modelled into the CK-548 binding site, the para-methoxy group of CK-869 projects towards the upper lip of the pocket and is sandwiched by the side chains of Ile 78, Glu 84 and Met 89 (Supplementary Fig. 7). The ortho-methoxy group points towards a crease formed at the ends of strands β1 and β7, which could be exploited to design higher affinity inhibitors.
Both CK-548 and CK-869 inhibited human and bovine Arp2/3 complexes, but neither inhibited budding or fission yeast Arp2/3 complexes (Fig. 1c, d). The residues that contact CK-548 are conserved in Arp2/3 complexes of most mammals, but tryptophan replaces Met 93 at the back of the CK-548 binding pocket in many other species, including budding and fission yeast (Supplementary Table 5). The bent side chain of Met 93 allows CK-548 to bind, but tryptophan at this position blocks binding to non-mammalian Arp3 and actin.
The mechanism of action of CK-548/869 is less apparent than that of CK-636/666. A fluorescence anisotropy assay showed that CK-548 decreased the affinity of rhodamine-labelled N-WASP-VCA for Arp2/3 complex twofold, but this should be inconsequential under the conditions of our assay (Supplementary Fig. 2). The conformation induced by CK-548/869 must interfere with one or more of the reactions along the pathway to branch formation, such as binding of the pointed end of Arp3 to a mother filament10, nucleotide binding to Arp3, or conformational changes that activate branch formation.
Compounds were purchased from Chemdiv: CK-0944636 (catalogue number 8012-5103), CK-0993548 (K205-1650), CK-0944666 (8012-5153) and CK-0157869 (K205-0942). We purified native Arp2/3 complex from human platelets12, bovine thymus6, Schizosaccharomyces pombe13 and Saccharomyces cerevisiae (Supplementary Methods), actin from chicken skeletal muscle14, recombinant full length HsWASP from Freestyle 293-FS cells (Invitrogen), HsWASP (residues 105–502), HsWASP-VCA (residues 428–502), Cdc4212, N-WASP-VCA (residues 428–505) (Supplementary Methods), GST-ActA (residues 36–170) (Supplementary Methods) and S. pombe Cdc12p(FH2) (residues 973–1390)15 from Escherichia coli. We used standard assays to measure polymerization of pyrenyl-actin16 and to visualize actin filaments by fluorescence microscopy17. Binding of etheno-ATP to Arp2/3 complex was performed as described previously with slight modifications18. We crystallized BtArp2/3 complex7 with either 0.5 mM CK-548 or 1 mM CK-636 in DMSO or soaked these compounds into crystals for 24 h before freezing in liquid nitrogen. Diffraction data were collected at beamline X29A at Brookhaven National Laboratories. SKOV3 cells were infected with L. monocytogenes and fixed with 2% formaldehyde, permeabilized with 0.1% Triton-X in PBS, stained with Listeria antibody (US Biologics) and Alexa Fluor 568 phalloidin (Molecular Probes), and imaged by fluorescence microscopy. We used an Isodata threshold on background-subtracted images of Listeria to isolate individual bacteria and measure the ratio of colocalized actin to Listeria fluorescence. Monocyte THP-1 cells were differentiated in 50 nM phorbol myristate acetate (Sigma-Aldrich Fluka) to form podosomes before treatment with compounds. Black molly keratocytes19 were observed by time-lapse phase contrast microscopy.
Human, bovine and S. pombe Arp2/3 complexes were purified as previously described11,12,20. S. cerevisiae Arp2/3 complex was purified from strain BN020 (MAT a or α,ura3-52, his3-200, leu2-3, lys2-801, trp1-1, Δarp2::TRP1, Δarp3::HIS, [pRS315-Arp3], [pRS317-ARP2]). Cultures were grown overnight at 30 °C to A600 = 1 and pelleted by centrifuging at 5,000g for 8 min. Cells were washed in lysis buffer (20 mM Tris pH 8.0, 50 mM NaCl, 1 mM EDTA) and pelleted. The pellet was frozen in liquid nitrogen and stored at -80 °C. Cells were thawed and 1 ml lysis buffer was added per g wet cell pellet. Protease inhibitor tablets (Sigma) were added (one tablet per 100 ml cell suspension) and cells were lysed by five passes at 30,000 p.s.i. on a model 110EH microfluidizer (Microfluidics). The lysate was spun at 30,000g for 20 min and the supernatant was collected and centrifuged for 1 h 15 min at 125,000g. The supernatant was filtered through cheesecloth and proteins were precipitated with 50% saturated aqueous (NH4)2SO4. Pelleted proteins were resuspended in 50 ml of PKME (25 mM PIPES pH 7.0, 50 mM KCl, 1 mM EGTA, 3 mM MgCl2, 1 mM EGTA, 1 mM DTT and 0.1 mM ATP) and dialysed against the same buffer overnight. The dialysate was purified on a GST-N-WASP-VCA affinity column, and Arp2/3 complex-containing fractions were pooled and concentrated to 1 ml before loading onto a Superdex200 gel filtration column (GE Healthcare) equilibrated in 20 mM Tris pH 8.0 and 100 mM NaCl. Peak fractions were pooled and concentrated to approximately 10–20 μM, flash-frozen in liquid nitrogen, and stored at -80 °C. Bovine N-WASP-VCA was constructed by amplifying residues 428–505 from pGEX-2T-NWASP-VCA (a gift of H. Higgs) and cloning into pGV67 using BamHI and EcoRI restriction sites. This vector was used to transform E. coli strain BL21(DE3). N-WASP-VCA was expressed and purified as previously described11. Chicken skeletal muscle was purified and labelled with pyrene iodoacetamide21. Recombinant full length human WASP was purified from 293 cells and recombinant human WASP (residues 105–502) was purified from E. coli for the initial screens with HsArp2/3 complex. Recombinant Listeria ActA residues 36–170 were expressed as an N-terminal GST fusion and purified from E. coli. S. pombe formin Cdc12(FH2) domain (residues 973–1390, abbreviated Cdc12(FH2) in text) was purified from E. coli.
Actin polymerization assay
We measured the nucleation activity of Arp2/3 complex from the time course of actin polymerization. Polymerization reactions of 100 μl were assembled as previously described12 and fluorescence measurements were made at 8 s intervals in a 96-well plate using a Gemini XPS spectrofluorimeter (Molecular Devices) with an excitation wavelength of 365 nm and an emission wavelength of 407 nM. The rate of polymerization at each time point was determined by calculating the slope in five-point intervals and multiplying by [total polymer]/ΔRFU, where [total polymer] is the total actin minus the critical concentration (0.1 μM) and ΔRFU is the difference in relative fluorescence units of the reaction upon completion (RFUmax) and the initial reaction mix (RFUmin). We plotted the maximum elongation rate (R) verses inhibitor concentration ([I]) and fitted the curve using the equation:Here Rmax is the maximum rate of elongation with Arp2/3 complex and no inhibitors and Rmin is the maximum rate of elongation of actin alone. The products of polymerization reactions were diluted into Alexa 488-phalloidin and observed by fluorescence microscopy16.
The endogenous cysteine (Cys 431) near the end of the amino terminus in purified N-WASP-VCA was labelled with rhodamine maleimide12. Fixed concentrations of Rho-N-WASP-VCA were titrated with BtArp2/3 complex and the fluorescence anisotropy measured12. Binding constants were determined by fitting the anisotropy curves to the following equation:where rf is the signal of the free receptor (R), rb is the signal of the bound receptor, and [L] is the total concentration of the ligand (species titrated). rb and Kd were fitted using Kaleidagraph (Synergy Software).
Etheno-ATP fluorescent assay
Binding of etheno-ATP to Arp2/3 complex was performed as described previously with slight modifications17,18. Human Arp2/3 complex (0.7 μM) was incubated with 2.5 μM etheno-ATP, 0.2 M acylamide (used as an etheno-ATP fluorescence quencher) in 2 mM Tris-HCl pH 8.0, 50 mM KCl, 0.05 mM EGTA 0.8 mM MgCl2, 0 and 0.5 mM DTT. Ethano-ATP emission spectra were acquired upon excitation at 340 nm using spectrofluorometer FluoroMax (Horiba JY). Inhibition of Arp2/3 complex by 100 μM CK-636 was performed in the presence or absence of 2.8 μM WASP-VCA.
Listeria comet, motility and podosome assays
SKOV3, THP-1 and L. monocytogenes (ATCC 984) cells were from ATCC. SKOV3 cells were cultured in RPMI with 5% FBS without antibiotic at 37 °C and 5% CO2. Listeria were grown in shaking culture in brain heart infusion (Difco Inc., DF0037-15) at 37 °C. SKOV3 cells were infected with L. monocytogenes, incubated for 90 min and then treated for 60 min with either compound dissolved in DMSO or a DMSO control before fixing with 2% formaldehyde. Fixed cells were permeabilized with 0.1% Triton-X in PBS at room temperature for 15 min and reacted with a 1:200 dilution of Listeria antibody (US Biologics) in PBS for 60 min at room temperature. Antibody was removed and cells were stained with a 1:100 dilution of Alexa Fluor 568 phalloidin and a 1:400 dilution Alexa Fluor 488 goat anti-rabbit secondary antibody (both from Molecular Probes) for 60 min at room temperature. After washing, cells were imaged at 20× with an Axon Instruments automated fluorescence microscope system (Molecular Devices). The THP-1 monocyte cell line was grown in RPMI with 10% FBS/2 mM l-glutamine, differentiated in 50 nM phorbol myristate acetate (Sigma-Aldrich Fluka) for 48 h at 37 °C, trypsinized and plated at 5 × 105 cells ml-1 in black Nunc glass-bottom 96-well plates. After incubation at 37 °C for 2 h, cells were treated with compounds for 15 min before fixing and staining with Alexa Fluor 568 phalloidin and DAPI (Sigma-Aldrich Fluka) as above. Supplementary Fig. 2 details the metrics used for podosome quantification. Black molly keratocytes were isolated19, treated briefly with trypsin and observed by time lapse phase contrast microscopy on plastic Petri dishes at room temperature.
Crystals of BtArp2/3 complex were grown at 4 °C using the hanging drop vapour diffusion method as previously described7. Crystals were soaked for 24 h in 18% polyethylene glycol 8000, 50 mM HEPES pH 7.5, 100 mM potassium thiocyanate, 20% glycerol and either 0.5 mM CK-548 or 1 mM CK-636 in DMSO before freezing in liquid nitrogen. Data were collected at beamline X29A at Brookhaven National laboratories and processed with HKL200022. To generate initial models, protein atoms from 1K8K were refined against the data using 30 cycles of minimization in which each subunit was allowed to move as a separate rigid body. CK-548 structure was refined in Refmac23 with topology files generated in ccp4i sketcher. The CK-636 structure was refined in CNS8 using topology files generated with the prodrg server24.
This work was supported by Cytokinetics, Inc., NIH research grant GM-066311 (to T.D.P.), an NSF graduate research fellowship (C.D.M.) and a Ruth Kirschstein postdoctoral fellowship (GM074374-02 to B.J.N.). We thank L. Belmont, Z. Khurshid, O. Ezizika, J. Lee, S. Leuenroth, Z. Cournia and H. Chen for help with the project.
Author Contributions B.J.N., N.T., A.R., D.W.P., Z.J. and J.H. designed and carried out experiments; C.D.M. analysed data; R.S. and T.D.P. supervised research; and B.J.N., N.T., A.R., J.H. and T.D.P. wrote the paper.
This file contains Supplementary Figures 1-7 with Legends and Supplementary Tables S1-S5.