GMF is a cofilin homologue that binds Arp2/3 complex to stimulate filament debranching and inhibit actin nucleation

Summary

Cell locomotion and endocytosis are powered by the rapid polymerization and turnover of branched actin filament networks nucleated by Arp2/3 complex [1]. While a large number of cellular factors have been identified that stimulate Arp2/3 complex-mediated actin nucleation, only a small number of studies so far have addressed which factors promote actin network debranching [2–4]. Here, we investigated the function of a conserved homologue of ADF/cofilin, glia maturation factor (GMF) [5, 6]. We found that S. cerevisiae GMF (also called Aim7) localizes in vivo to cortical actin patches, and displays synthetic genetic interactions with ADF/cofilin. However, GMF lacks detectable actin binding or severing activity, and instead binds tightly to Arp2/3 complex. Using in vitro evanescent wave microscopy, we demonstrated that GMF potently stimulates debranching of actin filaments produced by Arp2/3 complex. Further, GMF inhibits nucleation of new daughter filaments. Together, these data suggest that GMF binds Arp2/3 complex to both ‘prune’ daughter filaments at the branch points and inhibit new actin assembly. These activities and its genetic interaction with ADF/cofilin support a role for GMF in promoting the remodeling and/or disassembly of branched networks. Therefore, ADF/cofilin and GMF, members of the same superfamily, appear to have evolved to interact with actin and actin-related proteins, respectively, and to make mechanistically distinct contributions to the remodeling of cortical actin structures.

Results and Discussion

Glia maturation factor (GMF) is an evolutionarily conserved member of the ADF/cofilin superfamily based on sequence comparisons and structural homology [5, 6]. Sequence alignments and three-dimensional structure comparisons suggest that while many general features of the ADF/cofilin fold are retained in GMF-γ, the actin-binding residues on ADF/cofilin are not well conserved (red bars, Fig. 1A). Therefore, the cellular functions of GMF may have diverged from those of ADF/cofilin [6].

Figure 1. GMF localization and genetic analysis.

(A) Sequence of S. cerevisiae (SC) ADF/cofilin (Cof1) aligned with GMF isoforms (B, beta; G, gamma) from H. sapiens (HS), M. musculus (MM), D. melanogaster (DM), C. elegans (CE) and D. discoideum (DD). The locations of cof1 alleles with actin-binding defects are indicated [10]. (B) Localization of GMF-GFP and Abp1-RFP in the same cells. (C) Comparison of wild type and gmf1Δ cell growth rates in YPD. (D) Strains serially diluted and grown on YPD agar for two days. (E) gmf1Δ, cof1-22, and gmf1Δ cof1-22 cell growth rates in YPD. (F) The GMF-GFP plasmid rescues temperature-sensitive growth defects of gmf1Δ cof1-22 cells.

To study the function of S. cerevisiae GMF (YDR063W; also called Aim7, or altered inheritance rate of mitochondria [7]), we first constructed an internal fusion of GFP to GMF, which was functional (see below). GMF-GFP localized to the cytoplasm and cortical puncta. Localization of GMF-GFP in a strain expressing integrated Abp1-RFP, a marker for cortical actin patches [8], revealed that most GMF-GFP puncta correspond to actin patches (Fig. 1B). However, not all Abp1-RFP patches were labeled with GMF-GFP, suggesting that GMF-GFP may only be present on actin patches for a portion of their lifetime. A similar observation has been made for Cof1-GFP [9]. Thus, GMF, like Cof1, may be a late arriving component of actin patches.

Next, we made a deletion of the GMF1/AIM7 gene. gmf1Δ (aim7Δ) caused no overt defects in cell growth (Fig. 1C) or morphology (not shown), as previously reported [7]. To explore the relationship between GMF and ADF/cofilin, we crossed gmf1Δ to a partial loss of function cofilin allele, cof1-22 [10]. The gmf1Δ cof1-22 double mutant showed more pronounced defects in cell growth compared to cof1-22 after serial dilution and plating (Fig. 1D). The synthetic defects were very evident at 34°C, but also could be observed at 25°C by comparing growth rates of strains in liquid medium (Fig. 1E). The GMF-GFP construct used above rescued the synthetic growth defects of the gmf1Δ cof1-22 strain (Fig. 1F).

We investigated whether GMF, like ADF/cofilin, is able to bind and/or sever filamentous actin (F-actin) in vitro. In contrast to Cof1, purified GMF showed minimal cosedimentation with F-actin, even at the highest concentrations tested (10 and 20 μM), using either rabbit muscle actin or yeast actin (Fig. 2A). In addition, we failed to detect a binding interaction between GMF and NBD-labeled actin monomers (not shown). Further, GMF did not affect the kinetics of actin assembly or disassembly (Figs. 2B-D), whereas Cof1 accelerated both processes, attributable to its severing activity [11]. GMF also did not enhance actin disassembly effects of Cof1 (Fig. 2C), indicating that the two proteins do not synergize in actin severing. Taken together, the genetic, co-localization, and biochemical data suggest that GMF has evolved a function distinct from ADF/cofilin, but one that nevertheless overlaps physiologically with that of ADF/cofilin.

Figure 2. GMF interacts with Arp2/3 complex but not actin.

(A) Co-sedimentation assays using 3 μM GMF (left) or Cof1 (right) and variable concentrations of F-actin. Pellets (P) and supernatants (S) analyzed by SDS-PAGE and Coomassie staining. (B) Polymerization kinetics of 4 μM monomeric actin (5% pyrene-labeled) in the presence of different concentrations of GMF. (C) Effects of Cof1 and/or GMF on rate of disassembly of F-actin (10% pyrene-labeled) induced at time zero by Vitamin D-binding protein. (D) Effects of Cof1 (0.25 μM) and GMF (0.25 μM or 2.5 μM) on rate of assembly of 2 μM monomeric actin (5% pyrene-labeled). (E) Silver stained gel of proteins isolated from gmf1Δ cell lysates on GST-GMF and GST beads. Labeled bands were identified by mass spectrometry. (F) Binding of purified GMF to Arp2/3 complex immobilized on beads or control beads. Pellets (P) and supernatants (S) analyzed by immunoblotting with anti-GMF antibodies (1:1000) and quantified by densitometry.

To better understand GMF cellular function, we isolated from crude cell extracts proteins that bound to GST-GMF beads but not control GST beads (Fig. 2E). Excised gel bands were identified by mass spectrometry as actin, Arp2/3 complex subunits, and the Arp2/3-interacting protein coronin/Crn1 [12]. These results suggested an interaction (either direct or indirect) between GMF and Arp2/3 complex, consistent with previous co-immunoprecipitation results in mammalian cells and two-hybrid studies in yeast [5, 13–15]. Binding assays using purified proteins demonstrated that soluble GMF binds Arp2/3 complex immobilized on beads (Fig. 2F), indicating a direct interaction.

To study the potential effects of GMF on Arp2/3 complex activity, we first compared Arp2/3 complex-mediated actin assembly kinetics at different concentrations of GMF (Fig. 3A). GMF inhibited Arp2/3 complex-induced actin nucleation in a concentration-dependent manner (Fig. 3B), but had no significant effect on actin assembly kinetics in the absence of Arp2/3 complex (Fig. 3A and Fig. 2B). Half-maximal inhibition of 20 nM Arp2/3 complex was observed at approximately 250 nM GMF, suggesting a sub-micromolar dissociation constant. To further investigate these effects, we used time-resolved total internal reflection fluorescence (TIRF) microscopy to image nucleation and growth of branched actin filament networks formed by 20 nM Arp2/3 complex (activated by 300 nM VCA) in the presence and absence of 1 μM GMF (Fig. 3C). Imaging network growth allowed us to unambiguously identify the mother and daughter filaments at each node. Consistent with the bulk kinetic analysis, GMF caused a marked decrease in rate of daughter nucleation events per unit length of filament (Fig. 3D), but no significant change in elongation rate (Fig. 3E). Together, these data suggest that binding of GMF to Arp2/3 complex suppresses nucleation of daughter filament assembly.

Figure 3. Inhibition of Arp2/3 complex-mediated actin nucleation by GMF.

(A) Polymerization of 2 μM monomeric actin (5% pyrene-labeled) with 20 nM Arp2/3 complex, 200 nM GST-VCA, and GMF (0–4 μM). (B) Rates of assembly derived from curves in A. (C) Time-lapse TIRF microscopy on assembly of 1 μM actin, 20 nM Arp2/3 complex, 300 nM VCA, +/− 1 μM GMF. Images were selected from a series captured at 5 sec intervals. (D, E) Effects of 1 μM GMF on rate of daughter filament nucleation per length of total filament (D) and rate of filament elongation (E).

The TIRF experiments were performed under conditions in which actin filaments were not attached to the surface of the microscope slide, but could diffuse in two dimensions within the image plane [16]. In the absence of GMF, branches were stable; detachment was never seen in time-lapse observations of 57 individual branches (lasting on average 10 min from the time each daughter filament began to elongate). However, the addition of low concentrations of GMF (10 nM) induced numerous branch detachments within the first few minutes after daughter nucleation (Fig. 4A). This effect was specific to branch points, as GMF did not induce detectable severing at other locations on filaments. The debranching activity of GMF was potent; it was observed at 10 nM GMF, a concentration below the effective range for inhibition of nucleation and that caused no overt change in rate of branch nucleation or elongation (Fig. 3 and data not shown). 10 nM GMF reduced the characteristic lifetime of branches in 75% of the population to 190 ± 70 sec (Fig. 4B). The remaining 25% of daughter filaments appeared as a stable fraction unaffected by GMF. This fraction most likely did not debranch simply because insufficient GMF was present in the reaction; much of the GMF present may have been sequestered by Arp2/3 complex, which was in molar excess (20 nM Arp2/3 complex vs. 10 nM GMF). At higher GMF concentrations, close to 100% of daughter filaments debranched, and the lifetime was further reduced (120 ± 20 sec at 50 nM GMF; 90 ± 20 sec at 200 nM GMF). In reactions without GMF, the lifetime was estimated to be longer than 1 hour (see Supplemental Methods). The observed debranching kinetics are consistent with a simple mechanism in which GMF binding to Arp2/3 complexes at branch sites induces daughter filament dissociation, and most GMF molecules remain associated with Arp2/3 complex for the duration of the experiment rather than being recycled and made available for subsequent debranching events.

Figure 4. Debranching of actin filaments by GMF.

(A) Time-lapse TIRF microscopy on assembly of 1 μM actin, 20 nM Arp2/3 complex, 300 nM VCA, +/− 10 nM GMF. Red asterisks indicate debranching events. Without GMF, branches survived many minutes (arrow). (B) Fraction of daughter filaments that have debranched over time since their nucleation, with exponential probability distribution fits (red) to the data. Number of branches observed/debranched: 57/0 (control), 76/47 (10 nM GMF), 44/43 (50 nM GMF), 34/33 (200 nM GMF). (C) Disassembly of branched filaments observed by TIRF microscopy before and ~100s after addition of 50 nM GMF and/or 50 nM Cof1. (D) Working model for GMF function. GMF binds Arp2/3 complex and both prunes filament branches and inhibits nucleation. In doing so, GMF may functionally cooperate with ADF/cofilin to promote disassembly of actin networks (coordinated disassembly pathway), or it may independently remodel dendritic networks into unbranched networks (remodeling pathway), which may be later disassembled by ADF/cofilin.

Recently, it was shown that ADF/cofilin not only severs F-actin but also stimulates filament debranching by binding to F-actin and promoting dissociation of Arp2/3 complex [4]. To better understand the respective roles of GMF and ADF/cofilin in remodeling actin networks, we directly compared their effects by TIRF microscopy. In a microscope flow chamber, actin was assembled for 10–15 min with Arp2/3 complex and VCA. Then, buffer containing 50 nM GMF and/or 50 nM Cof1 (without G-actin, Arp2/3 complex, or VCA) was introduced, and filaments were observed for several minutes. Under these conditions, there was no new filament assembly and we could monitor the products of GMF- and Cof1-induced disassembly even though filaments were not tethered to the chamber surface. Exposing branched filaments to GMF produced long, mostly unbranched filaments, whereas addition of Cof1 yielded numerous, very short filaments (Fig. 4C). The fragmentation by Cof1 was extensive (Fig. 4C) and rapid (data not shown), such that we were unable to assess whether or not some of the filaments were derived from debranching events rather than general severing of F-actin. Nonetheless, we observed that some of the short fragments produced by Cof1 remained branched (Fig. 4C). When branched networks were treated with GMF and Cof1 together (50 nM each), the resulting filaments were short and largely unbranched, as expected from the individual effects of both proteins. From these data, it is clear that GMF and Cof1 have strikingly different effects in the remodeling/disassembly of branched filaments, consistent with major differences in their mechanisms (see below). Further analysis is required to learn if and how the separate effects of GMF and Cof1 might be coordinated to remodel actin networks.

The results presented here may help to explain how cellular actin networks are rapidly remodeled during cell motility, endocytosis, cytokinesis, and intracellular transport. Arp2/3 complex binds to sides of preexisting filaments and nucleates formation of daughter filaments at 70-degree angles, producing branched ‘dendritic’ arrays [17–19]. These activities play an important role in driving lamellipodial protrusion, endocytic vesicle internalization, and intracellular transport [20]. Ultrastructural analyses indicate that Arp2/3-nucleated actin networks at these locations consist of branched filaments [21–23], although this remains disputed at the lamellipodial leading edge [24]. It is not yet understood how branched networks nucleated by Arp2/3 are rapidly disassembled in vivo. Branched filaments assembled by Arp2/3 in vitro persist for tens of minutes [19, 25] and this study), but in vivo they are dismantled within a few seconds at the leading edge and endocytic loci [20, 26]. This suggests that cells may require additional factors to disassemble branched filaments. ADF/cofilin is one factor that stimulates debranching, by binding to F-actin and promoting dissociation of Arp2/3 complex, possibly through propagation of conformational changes in F-actin [4].

Here we have identified a new role for GMF as a factor that debranches filaments. Although GMF is a structural cousin of ADF/cofilin, it has a distinct mechanism. GMF binds directly to Arp2/3 complex, not actin, and ‘prunes’ daughter filaments at the branch points without severing elsewhere. This produces long unbranched filaments. GMF also inhibits Arp2/3 nucleation of daughter filaments. Both activities, debranching and inhibition of nucleation, may contribute to remodeling the architecture of dendritic actin networks into linear filaments. In contrast, the dual function of ADF/cofilin to sever and debranch filaments leads to fragmentation of branched networks into very short filaments, which can promote their net disassembly, or in other contexts amplify barbed ends to promote new assembly [27]. These observations highlight the mechanistic distinctions between GMF and ADF/cofilin, and the target specificity of GMF. Interestingly, GMF activities appear to be related to those of coronin. Coronin binds to Arp2/3 complex with high affinity and directly inhibits nucleation [12], and more recently it was reported that coronin promotes debranching [3]. Since we isolated coronin and Arp2/3 complex on GMF affinity columns, this raises the intriguing possibility that GMF and coronin work together in regulating Arp2/3 complex.

Based on the activities, localization, and genetic interactions of GMF, we conclude that it may have two different roles in vivo. In some cellular contexts, GMF and ADF/cofilin may work in concert to promote disassembly/turnover of branched actin structures (Fig. 4D; ‘coordinated disassembly pathway’). In other contexts, GMF may act independently of cofilin to remodel branched networks, by pruning branches and inhibiting nucleation, producing long filaments (Fig. 4D; ‘remodeling pathway’).

Our data also raise many new questions. First, how do low concentrations of GMF dismantle existing branches while only higher concentrations prevent nucleation of new branches? The simplest hypothesis is that both activities share the same underlying biochemical mechanism, in which GMF alters Arp2/3 conformation to antagonize its binding to mother filaments. Saturation of Arp2/3 with GMF may be required to effectively block nucleation, whereas only transient binding of GMF may be needed to induce detachment of a branch. Second, did ADF/cofilin and GMF diverge evolutionarily from a common ancestor to interact with actin and actin-related proteins (Arp2 and/or Arp3), respectively? Third, do the activities of S. cerevisiae GMF extend to mammalian homologues? Mammals have two GMF genes, GMF-β and GMF-γ, which together are widely expressed [5, 28, 29]. In addition, GMF-γ has been localized to the leading edge of mammalian cells, and co-immunoprecipitates with Arp2/3 complex [5]. Finally, what mechanisms control the timing of GMF interactions with Arp2/3 complex? This may involve ATP hydrolysis or phosphate release on Arp2 and/or Arp3 [30, 31] or in actin subunits in the mother or daughter filaments proximal to branch junctions. There is also the possibility that GMF activity, like that of ADF/cofilin, is controlled by phosphorylation, which has been suggested for mammalian GMF-γ [5].

Experimental procedures

For details, see Supplemental Methods.

Actin assembly and disassembly kinetics

Rabbit muscle actin was used in all experiments unless yeast actin is indicated. GMF, Cof1, and VCA fragment of Las17/WASp were purified from E. coli. Arp2/3 complex was purified from S. cerevisiae. Assembly and disassembly assays were performed essentially as described [11] using gel-filtered actin monomers (5% pyrene-labeled) for assembly assays and preassembled F-actin (10% pyrene-labeled) for disassembly assays. In disassembly assays, preassembled F-actin (2 μM) was incubated with GMF and/or Cof1 for 250 sec, then 4 μM Vitamin D-binding protein (human plasma Gc-globulin, Sigma-Aldrich, St. Louis, MO) was added to initiate disassembly.

Cell imaging

A low copy (CEN) plasmid expressing GMF-GFP was transformed into yeast strain BGY3091, which has an integrated copy of Abp1-RFP. Transformed cells were grown in selective media at 25°C to log phase, then imaged on a Zeiss E600 microscope (Thornwood, NY) equipped with a Hammamatsu Orca ER CCD camera (Bridgewater, NJ) running Openlab software (Improvision Inc., Waltham, MA).

TIRF microscopy analysis

RMA was labeled on lysine residues with Alexa-488 NHS (Molecular Probes, Carlsbad, CA), and was mixed at 10% with unlabeled actin (90%), then diluted to 1 μM. Fluorescence emissions were collected on an electron-multiplying charge-coupled device (EMCCD; Andor, South Windsor, CT) for 0.1 sec laser exposures at 5 sec intervals to generate time-lapse recordings. Analysis of actin elongation and branching rates was performed by tracing filaments as described (see Supplemental Methods). Branching rates, k_br, were calculated by counting the number of observed branch nucleation events, N_br, and dividing by the integral of the total filament length, L_tot, observed over the course of the TIRF recording, according to: k_br = N_br/∫L_tot (t)dt. Branch nucleation times were calculated by linear extrapolation of daughter filament elongation back to zero length, and branch lifetimes were then calculated as the time from nucleation to the midpoint between the frame in which the branch was last observed and the frame in which the dissociated branch was first observed.