Actin-induced dimerization of palladin promotes actin-bundling Ravi Vattepu, Rahul Yadav, and Moriah R. Beck* Chemistry Department, Wichita State University, Wichita, Kansas 67260 Received 6 August 2014; Accepted 9 October 2014 DOI: 10.1002/pro.2588 Published online 11 October 2014 proteinscience.org Abstract: A subset of actin binding proteins is able to form crosslinks between two or more actin filaments, thus producing structures of parallel or networked bundles. These actin crosslinking proteins interact with actin through either bivalent binding or dimerization. We recently identified two binding sites within the actin binding domain of palladin, an actin crosslinking protein that plays an important role in normal cell adhesion and motility during wound healing and embryonic development. In this study, we show that actin induces dimerization of palladin. Furthermore, the extent of dimerization reflects earlier comparisons of actin binding and bundling between different domains of palladin. On the basis of these results we hypothesized that actin binding may promote a conformational change that results in dimerization of palladin, which in turn may drive the crosslinking of actin filaments. The proximal distance between two actin binding sites on crosslinking proteins determines the ultrastructural properties of the filament network, therefore we also explored interdomain interactions using a combination of chemical crosslinking experiments and actin cosedimentation assays. Limited proteolysis data reveals that palladin is less susceptible to enzyme digestion after actin binding. Our results suggest that domain movements in palladin are necessary for interactions with actin and are induced by interactions with actin filaments. Accordingly, we put forth a model linking the structural changes to functional dynamics. Keywords: actin; crosslinking; dimerization; palladin; bundling Introduction Abbreviations: ABP, actin binding protein; BS3, bis(sulfosuccinimidyl)suberate; CH, calponin homology; DFDNB, 1,5-difluoro2,4-dinitrobenzene; DSP, dithiobis(succinimidylpropionate); DTT, dithiothreitol; ECM, extracellular matrix; F-actin, filamentous actin; FH2, formin homology domain; ICL, intramolecular crosslinked; Ig, immunoglobulin; IPTG, isopropyl b-D-1-thiogalactopyranoside; Kd, dissociation constant; OD, optical density; SDS-PAGE, sodium dodecyl sulfate polyacrylamide gel electrophoresis; TEV, tobacco etch virus; Vt, vinculin tail domain. Additional Supporting Information may be found in the online version of this article. Grant sponsor: Start-up funds from Wichita State University and a Center of Biomedical Research Excellence (COBRE) grant from the National Center for Research Resources; Grant number: 5P20RR017708; Grant sponsor: The National Institute of General Medical Sciences; Grant number: 8P20GM103420; Grant sponsor: Institutional Development Award (IDeA) from the National Institute of General Medical Sciences from the National Institutes of Health; Grant number: P20GM103418. *Correspondence to: Moriah R. Beck, 1845 Fairmount St., Wichita State University, Wichita, KS. E-mail: moriah.beck@wichita.edu 70 PROTEIN SCIENCE 2015 VOL 24:70—80 Actin regulates diverse cellular processes including cell division, adhesion, motility, and vesicle trafficking through dynamic turnover between monomeric (G-actin) and filamentous (F-actin) forms. Actin binding proteins (ABPs) are responsible for constructing a wide array of filament assemblies that range from meshworks of individual crosslinked actin filaments to closely spaced bundles of F-actin.1 The architecture of these assemblies is largely dictated by the type of actin crosslinking protein and its spatial and temporal regulation.2 Assemblies of loosely spaced and orthogonally connected threedimensional crosslinked actin filaments are present in the cellular cortex to support the cell membrane, lamellipodium, and membrane ruffles. These structures provide mechanical strength and shape as well as fluidity to irreversible deformation of the cell.3,4 Likewise actin bundles can be tightly or loosely packed into parallel actin filaments that provide mechanical support to microvilli, stereocilia, and C 2014 The Protein Society Published by Wiley-Blackwell. V filopodia for their specialized functions.5,6 The architecture of filament meshworks is also determined by the binding affinity and stoichiometry of the crosslinking protein involved.7 The distinction between actin crosslinking and bundling activities of ABPs is complicated in some cases because the ability to transform actin filaments from a weakly crosslinked complex into a tightly bundled network is partially determined by the their local concentration, which is only evident in in vitro assays. Espin, fascin, and aactinin, for example, all form crisscrossed arrays of actin filaments at low crosslinker concentrations that evolve into tightly bundled filaments at higher crosslinker concentrations.8–10 Nonetheless, other classes of crosslinking proteins only produce isotropically crosslinked networks at any crosslinker ratio. The actin binding protein palladin is ubiquitous in developing and adult vertebrate tissues and localizes in actin rich structures such as stress fibers, focal adhesions, podosomes, and Z-discs.11 Many recent studies suggest that deregulation of palladin expression may play a key role in the invasive cell motility that characterizes metastatic cancer cells as well as in the development of cardiovascular diseases.12–15 Localization of palladin in invadopodia and filopodia alongside other actin organizing proteins in these structures suggests that palladin is an important player in both normal and invasive motility. Despite the growing correlative evidence linking palladin to increased invasiveness, anti-migratory functions for actin bundling by palladin have also been reported.16 Because of this lack of clarity regarding the precise role of palladin in cancer progression, our goal was to determine the specific actin crosslinking mechanism that palladin employs to enable a better understanding of the dynamic coordination between the cytoskeleton and tumor cell invasion. Palladin belongs to a novel subfamily that includes two structurally related proteins, myotilin and myopalladin, that are present in skeletal and heart muscle, respectively, and all three are characterized by multiple immunoglobulin (Ig) domains [Fig. 1(A)]. Similar Ig domains are also found in other cytoskeletal organizing proteins including myomesin, titin, myBP-C, and myBP-H, where they maintain the structural integrity of the sarcomere.17,18 Crosslinking of F-actin was initially only associated with the tandem Ig3–4 domain of palladin. However, while characterizing the actin-binding functions of the Ig3 domain, Beck et al. observed that the Ig3 domain could also crosslink F-actin at higher concentrations than employed in initial experiments.19 Beck et al. also identified mutations to two basic patches on the surface of the Ig3 domain that are involved in binding actin. It was inferred that crosslinking of actin filaments could Vattepu et al. Figure 1. No evidence of homodimerization from chemical crosslinking of palladin domains. A: Schematic representation of palladin’s largest (200 kDa) isoform, which includes immunoglobulin (Ig) domains 1 through 5 and two polyproline regions represented by triangles. The brackets demarcate the most abundant isoform (90 kDa) of palladin that incudes only the second polyproline region to the Cterminus. The Ig3–4 domains of palladin are highlighted with the dashed square. B– E: Various palladin domains (7.5 mM) were incubated with varying concentrations of the chemical crosslinker (15–30 mM) for the indicated times, ranging from 10 to 40 min. Reaction products were subsequently analyzed for crosslinking on SDSPAGE. Crosslinking of Ig3 (B), Ig4 (C), Ig3–4 (D) with DFDNB does not show any dimer or other oligomeric species. E: Ig3–4 chemical crosslinking with BS3 reveals the formation of an intramolecular crosslinked monomer (denoted by star) that migrates faster than the monomeric species in SDS gels. occur at these two sites, however this does not rule out the possibility that palladin self-associates to promote actin bundling. Homodimerization is a common mechanism for proteins with small actin-binding domains to bind two filaments simultaneously. In fact, the palladinrelated protein myotilin forms dimers via two Ig domains that are homologous to palladin’s Ig PROTEIN SCIENCE VOL 24:70—80 71 domains 4 and 5.20 Involvement of regions outside the actin-binding domain cannot be ruled out, as the tandem Ig3–4 domain of palladin shows significantly higher binding and bundling activity than the isolated Ig3 domain.19 Moreover, Chin and Toker have identified an Akt1-mediated phosphorylation of a specific serine residue (S507) in the linker region between the Ig 3 and 4 domains of palladin that seems to affect not only migration of breast cancer cells but also the in vivo bundling activity of palladin.16 This highlights the importance for determining the mechanism by which the actin-binding domain of palladin is able to modulate actin filament architecture. In contrast to villin and a-actinin that can both self-associate even in the absence of actin, our group has shown that Ig3 domain remains monomeric in the absence of actin.19 Therefore we wanted to determine whether palladin uses a crosslinking mechanism similar to vinculin, VASP, and scruin whereby actin binding results in homodimerization.21 We have employed chemical crosslinking techniques to determine the role of the different palladin Ig domains in actin crosslinking as well as the role actin itself plays in controlling the activity of palladin. These results shed light on the mechanism whereby actin-binding induces dimerization of palladin, which in turn affects the phase transition of actin from single filaments to bundles. Results Palladin immunoglobulin domains exist as monomers in the absence of actin Our previous study revealed that the Ig3 domain of palladin has two F-actin binding sites, suggesting that bundling of F-actin occurs through binding at these two sites.19 However, actin bundling by the tandem Ig3–4 domain is much more robust than Ig3 alone, resulting in more bundles at lower concentrations of palladin.19 One explanation is that the Ig4 domain of palladin may be involved in dimerization to enhance actin bundling. We hypothesized that the F-actin bundling activity of Ig3, as well as the enhanced bundling by Ig3–4 domains of palladin, may be due to the selfassociation of palladin. Therefore, we wanted to determine whether the palladin domains (Ig3, Ig4, and/or Ig3–4) have any propensity to form homodimers in solution in the absence of actin. Previous sedimentation equilibrium measurements indicated that these domains have similar apparent and predicted molecular weights, which suggests that these domains exist as monomers in solution;19 however, these experiments were carried out at micromolar concentrations of protein and therefore may not be able to detect weak oligomeric species. We have used the chemical crosslinkers DFDNB and BS3 to detect oligomerization of the Ig3, Ig4, 72 PROTEINSCIENCE.ORG and Ig3–4 domains in the absence of actin. First, we used DFDNB with reactive fluorine groups that form stable aryl-amine bonds with nearby amine groups. Our results in Figure 1(B–D) show that neither of the C-terminal Ig domains of palladin form any high molecular weight oligomeric species in the absence of actin under moderately stringent crosslinking conditions (7.5 mM protein, 30 mM crosslinker, and 40 min reaction time). Increased crosslinker concentrations and reaction times can induce minor dimerization in the absence of actin, which likely do not represent a biologically relevant dimeric state. Interestingly, when Ig3–4 was allowed to react with the BS3 crosslinker in the absence of actin, a significant fraction of protein was found in an intramolecular crosslinked monomer [Fig. 1(E)], which migrates faster than native Ig3–4 in a SDSPAGE gel. We do not see this intramolecular crosslinked species of Ig3–4 in the presence of DFDNB [Fig. 1(D)], which is likely due to the shorter spacer arm length of DFDNB (3.0 Å) as compared with BS3 (11.4 Å). This suggests that both of the Ig domains in Ig3–4 maintain a close distance in the absence of actin. Actin induces dimerization in the actin-binding domain of palladin Previous studies with the vinculin tail (Vt) domain indicate that actin can play an active role in modulating its own filament organization by inducing oligomerization.22,23 Therefore, we next employed chemical crosslinking studies to ascertain whether dimerization of the Ig3, Ig4, and Ig3–4 domains of palladin occurs in the presence of actin. DFDNB crosslinking of the Ig3 domain in presence of actin reveals two different oligomeric crosslinked species, corresponding to a 24 kDa Ig3 homodimer and a 54 kDa Ig3:actin heterodimer [Fig. 2(A)]. The composition of these homo- and hetero-dimeric species, as well as other high molecular weight oligomers of Ig3 and actin, were confirmed by probing the Western blots with both palladin and actin antibodies [Fig. 3(A)]. Chemical crosslinking of Ig3–4 with DFDNB in the presence of actin likewise reveals Ig3–4 homodimers of 53 kDa as well as 69 kDa heterodimers [Fig. 2(D)], also confirmed by immunoblotting [Fig. 3(B)]. Homodimeriztion of Ig3 requires greater concentrations of protein as compared to the Ig3–4 actin induced dimerization. These results indicate that the Ig3 domain is sufficient for actin-induced dimerization, just as it is sufficient for both binding and bundling. However, the tandem Ig3–4 domain has a lower threshold for dimerization, which recapitulates the increased bundling activity of Ig3–4 that was previously observed.24 Similar crosslinking experiments with the isolated Ig4 domain do not show any dimer in the Actin-Induced Dimerization of Paladin used the DFDNB crosslinker in a mixture of equal concentrations of Ig3 and Ig4 domains in the presence of F-actin, with reaction conditions identical to the tandem Ig3–4 domains. Our results for the mixture of Ig3 and Ig4 reveal actin-induced dimers of Ig3 at 24 kDa and 54 kDa [Fig. 2(C)] that are similar in quantity to previous actin-induced Ig3 oligomerization in absence of Ig4 [Fig. 2(A)]. Moreover actin cosedimentation assays with a mixture of Ig3 and Ig4 revealed that actin binding is not higher than Ig3 alone (data not shown); suggesting that the free Ig4 domain in solution does not affect Ig3 binding to F-actin. Another possible explanation is that the linker region between Ig3 and Ig4 orients the Ig4 domain favorably during actin bundling and thereby exposes a second dimerization site on Ig4. This conformation could manifest itself only in the presence of the tandem Ig3–4 domains. Chemically crosslinked products were also sedimented at high speed to separate the actin bound and free fractions of palladin. Palladin Ig3–4 homodimers cosediment exclusively with the actin filaments [Fig. 4(A), pellet lanes]. This finding is consistent with the idea that actin binding induces palladin dimerization and suggests that the homodimeric form of palladin is functionally active [Fig. 4(A)]. To determine how the concentration of actin effects palladin dimerization, we performed chemical crosslinking reactions where the concentration of Ig3–4 was held constant while varying the actin concentration (1.25–20 lm) in the presence of a chemical crosslinker. Our results show that even very low concentrations of F-actin can induce palladin dimerization [Fig. 4(B)]. This substantiates the proposition that actin has a critical role in dimerization of palladin’s actin binding domain. Interdomain flexibility of Ig3–4 is important for F-actin binding and bundling Figure 2. Palladin homodimerizes in presence of actin. Palladin domains (5–10 mM) were incubated with F-actin (10 mM) and varying concentrations of the chemical crosslinker DFDNB (30–100 mM) for 20/40 min. Reaction products were analyzed for crosslinking by SDS-PAGE. A: Ig3, B: Ig4, C: Ig3 1 Ig4, D: Ig3–4 chemical crosslinking with DFDNB. Palladin homodimers (arrowhead) and palladin:actin heterodimers (star) are observed in Ig3, Ig3 1 Ig4, and Ig3–4 as indicated. The size of various dimers correspond to: Ig3 homodimer, 24 kDa; Ig3:actin, 54 kDa; Ig3–4 homodimer, 53 kDa; and Ig3– 4:actin heterodimer, 69 kDa. presence of actin, which is most likely due to inability of Ig4 to interact directly with actin [Fig. 2(B)]. However, the presence of the Ig4 domain in tandem with the Ig3 domain has shown much greater bundling of F-actin compared to Ig3 alone, suggesting a yet unknown role of Ig4 in enhancing actin crosslinking activity. To examine the role of Ig4, we have Vattepu et al. We further explored the possibility of changes in the interdomain interactions and the importance of the flexible linker region between the Ig3 and Ig4 domains using a series of chemical crosslinking experiments. As we have shown in Figure 1(E), crosslinking of the tandem Ig3–4 domain by BS3 induces a significant fraction of Ig3–4 to form an intramolecular crosslinked species in the absence of actin. To determine the conformational flexibility and/or role of the tandem domains in the presence of actin, we have used BS3 crosslinking of Ig3–4 in presence of actin under three different conditions. First, Ig3–4 and BS3 crosslinker were both added to polymerized actin at the same time to establish whether intramolecular crosslinking of monomeric Ig3–4 is favorable in the presence of F-actin [Fig. 5(A), Lane 1]. In the second condition, Ig3–4 was first crosslinked with BS3 in the absence of F-actin. This crosslinking reaction was quenched to prevent PROTEIN SCIENCE VOL 24:70—80 73 Figure 3. Western blots confirm the presence of both homo- and heterodimeric species upon chemical crosslinking of palladin in the presence of F-actin. Samples from the previous experiment [Fig. 2(A,D)] were probed against palladin and actin antibodies to identity the species formed during chemical crosslinking reactions. DFDNB crosslinking reactions containing (A) 10 mM Ig3 1 10 mM actin and (B) 5 mM Ig3–4 1 10 mM actin with either the palladin or actin antibody alongside the molecular weight ladder. any further crosslinking and then followed by addition of F-actin, thereby giving ample time for Ig3–4 to form intramolecular crosslinked species before actin binding [Fig. 5(A), Lane 2]. In the final condition, Ig3–4 was incubated first with F-actin for 60 min and then BS3 crosslinker was added, giving the least opportunity for Ig3–4 to form intramolecular crosslinked species [Fig. 5(A), Lane 3]. Our results Figure 4. Homodimers of palladin present in actin bound fraction and at very low concentrations of actin. A: Purified palladin domains (10 mM) were incubated with F-actin (10 mM) and DFDNB before being subjected to a high-speed centrifugation (100,000g). Pellets (P) and supernatants (S) and were analyzed on SDS-PAGE. The Ig3–4 dimer (star) and actin:Ig3–4 heterodimer (arrowhead) were observed in the pellet fraction only. B: The Ig3–4 domain of palladin was incubated with DFDNB and varying actin concentrations (1–20 mM) for 1 h. Reaction products were then analyzed on SDS-PAGE. Both the Ig3–4 homodimer (star) and actin:Ig34 heterodimer (arrowhead) were observed at all concentrations of F-actin. 74 PROTEINSCIENCE.ORG show that Ig3–4 homodimers and Ig3–4:actin heterodimers are still present when F-actin is available for interaction before crosslinker was added [Fig. 5(A)]. Moreover, the quantity of intramolecular crosslinked Ig3–4 species varied significantly under different conditions, suggesting different positions of the Ig4 domain with respect to Ig3 in the presence versus absence of actin [Fig. 5(B)]. In conditions where Ig3– 4 was crosslinked before actin binding, the absence of any Ig3–4:actin heterodimer combined with the increased amount of intramolecular crosslinked species also suggest that domain movement within Ig3– 4 is necessary or is induced when palladin interacts with F-actin. To further explore the importance of flexibility between the Ig3–4 domains we have used the crosslinker DSP, which is similar to BS3 in its spacer arm length but has the advantage of a disulfide bond in the spacer arm that can be cleaved upon reduction. We first carried out an actin cosedimentation assay with Ig3–4 in the presence of DSP without any reducing agent, which resulted in formation of the intramolecular crosslinked species. Furthermore this intramolecular crosslinked species of Ig3–4 caused a reduction in the amount of actin binding and bundling [Fig. 5(C,D)]. This reduction could be due to the intramolecular crosslinked structural confinement of Ig3–4 or may result from the chemical modification of lysine residues, for lysine residues in palladin have been shown to be involved in actin binding and bundling.19 Actin binding and bundling activity were both partially restored by cleaving the spacer arm of DSP, suggesting that the loss of actin bundling and binding is the result of intramolecular crosslinking and does not result from chemical modifications of lysine residues. Our results also indicate that actin bundling is abolished completely in the absence of reducing agent, Actin-Induced Dimerization of Paladin Figure 5. Intramolecular crosslinking of palladin Ig3–4 domains displays cause and effect relationship with actin binding and bundling. A: Lane 1 contains Ig3–4 (10 mM), BS3 (30 mM), and F-actin (10 mM) that were incubated together for 1 h. Lane 2 is Ig3–4 treated with BS3 for 1 h at RT in the absence of F-actin to allow the formation of intramolecular crosslinked monomers (star), followed by a 1 h incubation with F-actin. Lane 3 is Ig3–4 incubated with F-actin for 1 h, before addition of BS3. B: Quantification of the intramolecular crosslinked (ICL) Ig3–4 species from lanes in A. C and D: Ig3–4 was treated with the thiol-cleavable BS3 analog DSP for 1 h in the absence of F-actin to yield intramolecularly crosslinked monomers (ICL). Following a 30 min incubation in either the absence or the presence of 50 mM DTT (C), the binding of DSP treated Ig3–4 to Factin was assessed by a co-sedimentation assay followed by SDS-PAGE and quantification of Ig3–4 band by densitometry. D: Bundling of F-actin by DSP treated Ig3–4 was also assessed by a differential sedimentation assay and the actin bands were quantified in similar manner. The percentage of actin present in each fraction is represented as supernatant (light gray), pellet (dark gray), and bundle (black). whereas actin binding is only diminished [Fig. 5(C,D)]. This dichotomy suggests that interdomain flexibility is more important for bundling and suggests that actin-binding induces a conformational change in Ig3–4 that leads to the exposure of an additional dimerization site on Ig3 or Ig4. This additional binding site could then enhance homodimeriztion of Ig3–4 to a greater degree than for Ig3 and provides an explanation for the previous results that revealed that Ig3–4 is a better actin bundling protein than Ig3. Actin induced structural changes in Ig3–4 are revealed by limited proteolysis We carried out limited proteolysis experiments with Ig3–4 in the presence and absence of F-actin to Vattepu et al. determine whether Ig3–4 undergoes a conformational change upon binding F-actin. First, we performed proteolysis of Ig3–4 with trypsin and chymotrypsin at varying incubation times and protease concentrations, and our results show that the rate of degradation is greater for trypsin as compared to chymotrypsin (Supporting Information, Fig. S1). Furthermore we found that in the absence of Factin, Ig3–4 is more susceptible to enzyme digestion than in the actin bound form (Fig. 6). Trypsin completely digested Ig3–4 in the supernatant within 20 min, whereas actin-bound Ig3–4 was not completely digested until 60 min. The amount of intact Ig3–4 that remained after proteolysis was also quantified and reveals that 60% of actin-bound Ig3–4 remains after 20 min [Fig. 6(C,D)]. On the other hand, Ig3–4 that is not bound to actin was nearly completely digested after only 10 min [Fig. 6(A,B)]. These results suggest that actin binding by Ig3–4 causes a change in conformation or orientation of the two domains that causes protection from proteolysis. We also performed proteolysis of Ig3 under similar conditions, however, we did not observe any change in extent of proteolysis between free Ig3 and actinbound Ig3 (Supporting Information Fig. S2). Ig3 remained quite stable even when we employed higher concentrations of both the proteases, and furthermore binding of Ig3 to F-actin does not appear to involve in any detectable change in conformation. We therefore suggest that this actin-induced conformational change most likely occurs in the tandem Ig3–4 domain and results in a new orientation of the two domains relative to each other. Discussion The general understanding from previous study with a variety of crosslinking proteins is that the crosslinking activity results from more than one actin binding site. This can be achieved either through two structurally related but different binding sites on the same polypeptide or one binding site on each polypeptide that are then brought together through oligomerization. The proximity of these binding sites appears to correlate with the architecture of the resulting actin filament network. Two proximal actin-binding sites, in fimbrin and in the tandem calponin homology (CH) domains of fascin, both result in tightly packed parallel bundles of actin filaments.25,26 Non-covalent dimerization is another a common actin crosslinking mechanism. For example, villin has an N-terminal dimerization site that regulates self-association, with or without actin, but is also critical for bundling actin filaments.27,28 In a similar fashion, the formin family of proteins homodimerize and bundle actin filaments via their FH2 domain.29 Homodimers of filamin, where C-terminal self-association and N-terminal actin binding sites are PROTEIN SCIENCE VOL 24:70—80 75 Figure 6. Actin binding protects palladin from proteolysis. The Ig3–4 domain of palladin (50 mg) was incubated with F-actin (50 mg) and then centrifuged at 100,000g to separate the actin bound (pellet) and free (supernatant) fractions of Ig3–4. Both fractions were then subjected to limited proteolysis by using trypsin (1:1000) before separation and detection of fragments by SDSPAGE (A and B). The percentage of intact Ig3–4 (arrowhead) found in the supernatant (A and C) or actin bound pellet (B and D) was quantitated by densitometry measurements of corresponding gels. distantly located, provide flexibility to crosslink distant actin filaments and assemble into a loose and orthogonal network.30,31 Similarly, crosslinking of actin filaments by a-actinin is facilitated by the central rod domain of spectrin repeats that form an antiparallel dimer, while N-terminal CH domains provide a pair of actin binding sites.32 To understand the network mechanics and organization of the actin cytoskeleton, it is imperative to determine the binding interactions and structural role of the growing list of ABPs using well-defined in vitro biochemical techniques. In this study, we have demonstrated the dimerization of palladin’s actin binding domain can be induced upon binding to actin, which translates into bundling of actin filaments. Palladin domains (Ig3, Ig4, and Ig3–4) do not self-associate in the absence of actin, as revealed by our chemical crosslinking experiments with DFDNB crosslinker and also confirmed in previous studies of palladin oligomerization.19,22,27 Intramolecular crosslinked species were observed with Ig3–4 in the presence of crosslinker (BS3, 11.4 Å) with a longer spacer arm length than DFDNB (3.0 Å); suggesting that domains in Ig3–4 may exist in an anti-parallel, “U” shaped orientation connected through the 41 residue linker (discussed more thoroughly in later section). 76 PROTEINSCIENCE.ORG The homodimers of Ig3 and Ig3–4 that we observed only in the presence of actin (Fig. 2) suggest that palladin utilizes a mechanism whereby two Ig3 or tandem Ig3–4 domains self-associate only upon binding to actin in order to crosslink two actin filaments. In contrast to a-actinin, utropin, and filamin, where actin binding and dimerization activities reside in two separate domains;31–33 the Ig3 domain of palladin seems to possess both an actin binding site and at least one dimerization site that represents the minimum requisite for actin bundling activity. Having different actin binding and dimerization domains separated by a flexible linker, as observed in various other ABPs, alters the structural and viscoelastic properties of the crosslinked actin filaments. Filamin, for example, contains more linker domains than a-actinin and this results in actin filament structures that are characterized by actin bundles that have more viscoelasticity than those formed by a-actinin.34 The viscoelastic properties of actin filament structures formed by the isolated Ig3 and Ig3–4 domains have not been studied yet, however, the rheological properties of full length palladin-actin networks have been found to be similar to those formed by a-actinin.35 Conventional wisdom would suggest that since actin binding and dimerization sites reside in the same domain (Ig3), Actin-Induced Dimerization of Paladin Figure 7. Model for actin bundling and palladin Ig3–4 self-association. 1: Dimerization of palladin requires a conformational change to take place, shown here as a loop displacement. Dimers of palladin may sporadically form but they are unstable and the equilibrium is shifted toward the monomeric form. 2: Palladin binds to F-actin through the Ig3 domain, which results in conformational change that favors dimerization. 2 and 3: Upon dimerization, palladin promotes interaction between actin filaments and their subsequent zippering into tight bundles. 4: Palladin-induced bundles could then elongate by polymerization of crosslinked actin filaments (4) and, although not determined yet, it is possible that palladin assembles both unipolar and mixed polarity bundles. then palladin should form a stiffer actin network than those formed by a-actinin. On the contrary, our results suggest that actin bundling by palladin is not a straightforward single domain mechanism as the Ig3–4 domain displays far greater bundling and actin-induced dimerization capabilities. Although this study has focused on these two domains, it has implications for full-length palladin as all but one of the seven isoforms contain the Ig3 and Ig4 domains. It is important to note that other regions of palladin are involved either directly or indirectly in actin binding because the binding affinity of full-length palladin is significantly greater (Kd  2 mM) than the Ig3 domain (Kd  60 mM).24 This difference in actin binding affinity is also evident from the difference in intensity of homodimer protein bands present in the case of Ig3 versus the more prevalent Ig3–4, when similar protein and actin concentrations are used (Fig. 2). This reduction in homodimer formation for the single domain could likewise result from the lower actin binding affinity of Ig3 (Kd  60 mM) as compared to Ig3–4 (Kd  9 mM), which would require an increase in the Ig3:actin molar ratio to saturate binding and cause more substantial actin bundling. Distinct homodimers and higher actin binding affinity were observed with Ig3–4 and this can also be correlated with significantly higher bundling of actin filaments, as shown even at lower Ig3–4/actin molar ratios (Fig. 5). Despite the fact that the isolated Ig4 domain of palladin has no direct interaction with actin, this domain enhances the binding affinity of tandem Ig3–4 by an unknown mechanism. On the basis of our results, we put forth a mechanism to explain this increased actin binding and crosslinking Vattepu et al. whereby a conformational change occurs in Ig3–4 that is induced by actin binding. In this model, the conformational change leads to (1) the exposure of either second binding or dimerization site on the Ig4 domain or (2) exposure of an otherwise cryptic actin binding site on the Ig3 domain or linker between Ig3 and Ig4 domains that enhances the affinity of Ig3 for actin (Fig. 7). Intramolecular crosslinked species that migrate faster in SDS-PAGE were only observed in samples of Ig3–4 containing the crosslinkers with a longer spacer-arm length (BS3 and DSP), which suggests that the orientation of the actin-bound form of Ig3–4 domain assumes a more extended form than the free form. We found that the presence of intramolecular crosslinked species inversely correlates with Ig3–4 bundling activity. This indicates that interactions between actin and Ig3–4 inhibit intramolecular crosslinking, which results in a more open conformation of the Ig3–4 domain [Fig. 5(A,B)]. Probing this phenomenon further with the reversible DSP crosslinker revealed that the decreased actin binding and bundling caused by intramolecular crosslinking could be partially restored by cleaving the spacer arm of DSP [Fig. 5(C,D)]. Similar intramolecular crosslinked species were also observed in the case of vinculin tail fragment (V884-1066), which likewise decreased actin-bundling activity.22 This outcome suggests that interdomain flexibility allows for an open conformation is critical for actin-bundling activity and is more important for bundling than for binding to actin. Our results suggest that actin binding induces a conformational change in Ig3–4 that favors both dimerization and bundling; therefore, actin plays an active role alongside palladin in PROTEIN SCIENCE VOL 24:70—80 77 transforming cytoskeletal organization. This active role for actin in controlling bundling is a mechanism common to structurally different actin bundling proteins, as previously identified in WLIM1 from Nicotiana tabacum, where actin promotes and stabilizes the monomer to dimer transition. This dimerization of WLIM1 then promotes actin crosslinking.36 Vinculin uses a similar mechanism, where actin inhibits native dimerization of the tail fragment but exposes a cryptic dimerization site that is essential for actin bundling through conformational change in tail domain.22 We further analyzed the actin-induced conformational change in Ig3–4 by limited proteolysis; and, as seen with vinculin tail domain (Vt), conformational changes within the domain brought about after actin binding resulted in an increase in protease susceptibility. Contrary to what was observed for Vt, actin binding by Ig3–4 results in diminished protease reactivity. This suggests that an open conformation of Ig3–4 is induced or favored by actin binding, which results in stronger Ig3–4 homodimerization and subsequently decreases the susceptibility. We demonstrate that the actin-binding and bundling domains of palladin remain monomeric in the absence of actin, but oligomerize following incubation with actin. Our evidence that actin binding decreases the susceptibility of palladin to proteases supports the idea of actin-induced restrictions in conformational flexibility of palladin. Interestingly, we also observed that restricted domain movements brought about by intramolecular crosslinking prevented actin binding and bundling by palladin. Together our results provide a causal link between actin-palladin interactions and oligomer formation. Similar actin-induced protein oligomerization mechanisms have been reported for vinculin, vilin, and WLIM1.22,36,37 Our results suggest that palladin joins these structurally unrelated ABPs in a common mechanism, whereby control of actin-bundling is facilitated by actin itself. On the basis of these results we put forth a model for palladin dimerization and bundling activity, where actin induces a conformational change in palladin that gives rise to a palladin structure that is more amenable to dimerization (Fig. 7). This conformational change in turn promotes binding to multiple actin filaments and subsequent crosslinking of these actin filaments into tight bundles. Materials and Methods Materials Muscle acetone powder for actin purification was purchased from Pel-Freez Biologicals (Rogers, AR). Crosslinkers DFDNB (1,5-difluoro-2,4-dinitrobenzene), DSP or Lomant’s reagent [dithiobis(succinimidyl 78 PROTEINSCIENCE.ORG propionate)], and BS3 [bis(sulfosuccinimidyl)suberate], and trypsin were purchased from Thermo Fisher Scientific (Waltham, MA). Chymotrypsin was obtained from Worthington Biochemical Corporation (Lakewood, NJ). Polyclonal IgG antibodies against palladin and alpha actin were purchased from Proteintech (Chicago, IL). Biotinylated anti-IgG secondary antibody and the immune detection system R Elite ABC kits) were purchased (VECTASTAINV from Vector Laboratories (Burlingame, CA). Expression and purification of recombinant palladin protein domains The coding sequences of Ig3, Ig4, and Ig3–4 domains of palladin, previously cloned into pMAL-c2x,24 were subcloned into the pTBSG vector containing an Nterminal hexa-histidine tag and TEV protease cleavage site for purification of recombinant proteins.38 pTBSG Ig3–4 was further modified by inserting the MBP coding sequence between the histidine tag and TEV protease site to avoid aggregation during recombinant protein expression. All constructs were verified by sequencing and were transformed into BL21(DE3) derived T7 Express lysY competent Escherichia coli cells (New England Biolabs; Ipswich, MA) for protein expression. Cell cultures were grown at 37 C and induced by 0.5 mM IPTG at OD600 of 0.7 and further grown overnight at 18 C. Cell cultures were pelleted down and resuspended in lysis buffer (50 mM Tris pH 8.0, 300 mM NaCl, 10 mM imidazole) followed by sonication and clearing of cell lysate by centrifugation. Supernatant was loaded onto a pre-equilibrated Ni-NTA column and protein was purified according to the manufacturer’s guidelines (Thermo Fisher Scientific, Waltham, MA). The N-terminal histidine tag was cleaved by TEV protease and further purified as previously reported.24 All three palladin domains were finally dialyzed in HBS buffer (20 mM HEPES, pH 7.4, 100 mM NaCl). Chemical crosslinking Chemical crosslinking was done on purified Ig3, Ig4, and Ig3–4 proteins in the presence and absence of F-actin. We have used three crosslinkers (DFDNB, DSP, and BS3) that react with primary amines and differ in their spacer arm length with 3.0 Å, 12.0 Å, and 11.4 Å, respectively. The DSP crosslinker was used for its DTT or TCEP cleavable disulfide bond in the spacer arm. Stock solutions of 2 mM DFDNB, DSP, and BS3 were prepared just before each experiment by dissolving the crosslinker in DMSO and/or water, respectively. All crosslinkers were used at final concentrations ranging from 15 to 100 mM and crosslinking reaction times ranging from 10 to 40 min, then reaction was quenched by addition of Laemmli buffer. G-actin for crosslinking experiments was prepared from the purified actin stock made by Actin-Induced Dimerization of Paladin dialyzing G-actin in to buffer containing 2 mM NaHCO3, pH 7.6, 0.2 mM CaCl2, 0.2 mM ATP, and 0.02% NaN3. for 0 to 80 min. The reaction was quenched by addition of 1 mM PMSF. Samples were run on 12–15% polyacrylamide gel. Immunoblotting Acknowledgments Homo- and hetero-dimers of palladin and actinpalladin fragments were verified by immunoblotting. Samples containing palladin and actin with and without crosslinkers were run on SDS-PAGE in duplicate and transferred on to PVDF membranes (Millipore). Membranes were blocked at 4 C for 6 h with 5% skim milk in TBS-T (0.05% Tween-20). Blots were incubated overnight at 4 C with primary antibody against palladin (1:1000 dilutions) or alpha actin (1:1000 dilutions). Blots were washed with TBS-T buffer and further incubated for 1 h at 25 C with biotinylated anti-rabbit secondary antibody (1:500 dilution) followed by incubation with Vectastain ABC solution for 30 min. For detection, 1 mg/ mL DAB was dissolved in 100 mM imidazole, pH 7.0, 2.5 mM CoCl2 and immediately before use 0.002% H2O2 was added. The reaction was stopped by washing with water twice. The authors thank Owen Nadeau at KUMC for his assistance with pilot chemical crosslinking studies of palladin, Philip Gao, and Anne Cooper in the COBRE-PSF Protein Production Group at KU for cloning of pTBSG vectors, and Isabel Hendry at WSU for help with Western Blots. Actin co-sedimentation assay in presence of crosslinkers Actin co-sedimentation assays in the presence of crosslinker were carried out by incubating both the crosslinker and protein at various concentrations, and followed by addition of F-actin in buffer containing 10 mM Hepes, pH 7.5, 100 mM KCl, 2 mM MgCl2. Samples were incubated for 60 min at room temperature then reaction was quenched by addition of 1M Tris, pH 7.5 and centrifuged at 100,000g for 30 min. Supernatants were removed carefully followed by pellet resuspension in 2X SDS buffer. Samples were run on a 12 or 15% polyacrylamide gel according to the requirements of the experiment. Gels were stained in PageBlue protein staining solution (Thermo-Fisher) and analyzed by using ImageJ.39 Bundling of F-actin by Ig3–4 was carried out in the presence of crosslinker as before except with an additional step involving centrifugation of samples at low speed (5,000g) to pellet high molecular weight bundled actin before high speed centrifugation. Limited proteolysis Ig3–4 was incubated in presence of F-actin in 20 mM Hepes, 100 mM NaCl buffer at room temperature for 60 min, and then centrifuged at 150,000g for 30 min. The supernatant containing unbound Ig3–4 was separated and the pellet containing actin bound Ig3–4 was then re suspended in 20 mM Hepes, 100 mM NaCl. Both the unbound and bound Ig3–4 were then digested by the addition of either chymotrypsin (1:500) or trypsin (1:1000) to each sample, followed by incubation at room temperature Vattepu et al. References 1. Matsudaira P (1994) Actin crosslinking proteins at the leading edge. Semin Cell Biol 5:165–174. 2. Tseng Y, Kole TP, Lee JS, Fedorov E, Almo SC, Schafer BW, Wirtz D (2005) How actin crosslinking and bundling proteins cooperate to generate an enhanced cell mechanical response. Biochem Biophys Res Commun 334:183–192. 3. Salbreux G, Charras G, Paluch E (2012) Actin cortex mechanics and cellular morphogenesis. Trends Cell Biol 22:536–545. 4. Small JV, Stradal T, Vignal E, Rottner K (2002) The lamellipodium: where motility begins. Trends Cell Biol 12:112–120. 5. Revenu C, Athman R, Robine S, Louvard D (2004) The co-workers of actin filaments: from cell structures to signals. Nat Rev Mol Cell Biol 5:635–646. 6. Bartles JR (2000) Parallel actin bundles and their multiple actin-bundling proteins. Curr Opin Cell Biol 12: 72–78. 7. Lieleg O, Claessens MMAE, Bausch AR (2010) Structure and dynamics of cross-linked actin networks. Soft Matter 6:218–225. 8. Purdy KR, Bartles JR, Wong GC (2007) Structural polymorphism of the actin-espin system: a prototypical system of filaments and linkers in stereocilia. Phys Rev Lett 98:058105. 9. Tseng Y, Fedorov E, McCaffery JM, Almo SC, Wirtz D (2001) Micromechanics and ultrastructure of actin filament networks crosslinked by human fascin: a comparison with alpha-actinin. J Mol Biol 310:351–366. 10. Falzone TT, Lenz M, Kovar DR, Gardel ML (2012) Assembly kinetics determine the architecture of alphaactinin crosslinked F-actin networks. Nat Commun 3: 861. 11. Parast MM, Otey CA (2000) Characterization of palladin, a novel protein localized to stress fibers and cell adhesions. J Cell Biol 150:643–656. 12. Goicoechea SM, Bednarski B, Garcia-Mata R, PrenticeDunn H, Kim HJ, Otey CA (2009) Palladin contributes to invasive motility in human breast cancer cells. Oncogene 28:587–598. 13. Ryu B, Jones J, Hollingsworth MA, Hruban RH, Kern SE (2001) Invasion-specific genes in malignancy: serial analysis of gene expression comparisons of primary and passaged cancers. Cancer Res 61:1833–1838. 14. Wang W, Goswami S, Lapidus K, Wells AL, Wyckoff JB, Sahai E, Singer RH, Segall JE, Condeelis JS (2004) Identification and testing of a gene expression signature of invasive carcinoma cells within primary mammary tumors. Cancer Res 64:8585–8594. 15. Jin L (2011) The actin associated protein palladin in smooth muscle and in the development of diseases of PROTEIN SCIENCE VOL 24:70—80 79 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 80 the cardiovasculature and in cancer. J Muscle Res Cell Motil 32:7–17. Chin YR, Toker A (2010) The actin-bundling protein palladin is an Akt1-specific substrate that regulates breast cancer cell migration. Mol Cell 38:333–344. Otey CA, Dixon R, Stack C, Goicoechea SM (2009) Cytoplasmic Ig-domain proteins: cytoskeletal regulators with a role in human disease. Cell Motil Cytoskeleton 66:618–634. Otey CA, Rachlin A, Moza M, Arneman D, Carpen O (2005) The palladin/myotilin/myopalladin family of actin-associated scaffolds. Int Rev Cytol 246:31–58. Beck MR, Dixon RD, Goicoechea SM, Murphy GS, Brungardt JG, Beam MT, Srinath P, Patel J, Mohiuddin J, Otey CA, Campbell SL (2013) Structure and function of palladin’s actin binding domain. J Mol Biol 425:3325–3337. Salmikangas P, Mykkanen OM, Gronholm M, Heiska L, Kere J, Carpen O (1999) Myotilin, a novel sarcomeric protein with two Ig-like domains, is encoded by a candidate gene for limb-girdle muscular dystrophy. Hum Mol Genet 8:1329–1336. Huttelmaier S, Harbeck B, Steffens O, Messerschmidt T, Illenberger S, Jockusch BM (1999) Characterization of the actin binding properties of the vasodilatorstimulated phosphoprotein VASP. FEBS Lett 451:68– 74. Johnson RP, Craig SW (2000) Actin activates a cryptic dimerization potential of the vinculin tail domain. J Biol Chem 275:95–105. Janssen ME, Kim E, Liu H, Fujimoto LM, Bobkov A, Volkmann N, Hanein D (2006) Three-dimensional structure of vinculin bound to actin filaments. Mol Cell 21:271–281. Dixon RD, Arneman DK, Rachlin AS, Sundaresan NR, Costello MJ, Campbell SL, Otey CA (2008) Palladin is an actin cross-linking protein that uses immunoglobulin-like domains to bind filamentous actin. J Biol Chem 283:6222–6231. Jansen S, Collins A, Yang C, Rebowski G, Svitkina T, Dominguez R (2011) Mechanism of actin filament bundling by fascin. J Biol Chem 286:30087–30096. Klein MG, Shi W, Ramagopal U, Tseng Y, Wirtz D, Kovar DR, Staiger CJ, Almo SC (2004) Structure of the actin crosslinking core of fimbrin. Structure 12:999– 1013. PROTEINSCIENCE.ORG 27. George SP, Wang Y, Mathew S, Srinivasan K, Khurana S (2007) Dimerization and actin-bundling properties of villin and its role in the assembly of epithelial cell brush borders. J Biol Chem 282:26528–26541. 28. Khurana S, George SP (2008) Regulation of cell structure and function by actin-binding proteins: villin’s perspective. FEBS Lett 582:2128–2139. 29. Harris ES, Rouiller I, Hanein D, Higgs HN (2006) Mechanistic differences in actin bundling activity of two mammalian formins, FRL1 and mDia2. J Biol Chem 281:14383–14392. 30. Nakamura F, Osborn TM, Hartemink CA, Hartwig JH, Stossel TP (2007) Structural basis of filamin A functions. J Cell Biol 179:1011–1025. 31. Pudas R, Kiema TR, Butler PJ, Stewart M, Ylanne J (2005) Structural basis for vertebrate filamin dimerization. Structure 13:111–119. 32. Sjoblom B, Salmazo A, Djinovic-Carugo K (2008) Alpha-actinin structure and regulation. Cell Mol Life Sci 65:2688–2701. 33. Keep NH, Winder SJ, Moores CA, Walke S, Norwood FL, Kendrick-Jones J (1999) Crystal structure of the actin-binding region of utrophin reveals a head-to-tail dimer. Structure 7:1539–1546. 34. Esue O, Tseng Y, Wirtz D (2009) Alpha-actinin and filamin cooperatively enhance the stiffness of actin filament networks. PLoS One 4:e4411. 35. Grooman B, Fujiwara I, Otey C, Upadhyaya A (2012) Morphology and viscoelasticity of actin networks formed with the mutually interacting crosslinkers: palladin and alpha-actinin. PLoS One 7:e42773. 36. Hoffmann C, Moes D, Dieterle M, Neumann K, Moreau F, Tavares Furtado A, Dumas D, Steinmetz A, Thomas C (2014) Live cell imaging reveals actin-cytoskeletoninduced self-association of the actin-bundling protein WLIM1. J Cell Sci 127:583–598. 37. George SP, Chen H, Conrad JC, Khurana S (2013) Regulation of directional cell migration by membraneinduced actin bundling. J Cell Sci 126:312–326. 38. Qin H, Hu J, Hua Y, Challa SV, Cross TA, Gao FP (2008) Construction of a series of vectors for high throughput cloning and expression screening of membrane proteins from Mycobacterium tuberculosis. BMC Biotechnol 8:51. 39. Abramoff MD, Magalhaes PJ, Ram SJ (2004) Image processing with ImageJ. Biophotonics Intl 11:36–42. Actin-Induced Dimerization of Paladin