Actin Monomers Activate Inverted Formin 2 by Competing with Its Autoinhibitory Interaction*

Background: INF2 is not regulated by typical formin autoinhibition. Results: INF2 is autoinhibited in cells and is constitutively active in biochemical actin polymerization assays containing only actin monomers but is inhibited by proteins that bind actin monomers. Conclusion: INF2 can be activated by actin monomers. Significance: A component of INF2 regulation might be the ability to sense free actin monomer levels. INF2 is an unusual formin protein in that it accelerates both actin polymerization and depolymerization, the latter through an actin filament-severing activity. Similar to other formins, INF2 possesses a dimeric formin homology 2 (FH2) domain that binds filament barbed ends and is critical for polymerization and depolymerization activities. In addition, INF2 binds actin monomers through its diaphanous autoregulatory domain (DAD) that resembles a Wiskott-Aldrich syndrome protein homology 2 (WH2) sequence C-terminal to the FH2 that participates in both polymerization and depolymerization. INF2-DAD is also predicted to participate in an autoinhibitory interaction with the N-terminal diaphanous inhibitory domain (DID). In this work, we show that actin monomer binding to the DAD of INF2 competes with the DID/DAD interaction, thereby activating actin polymerization. INF2 is autoinhibited in cells because mutation of a key DID residue results in constitutive INF2 activity. In contrast, purified full-length INF2 is constitutively active in biochemical actin polymerization assays containing only INF2 and actin monomers. Addition of proteins that compete with INF2-DAD for actin binding (profilin or the WH2 from Wiskott-Aldrich syndrome protein) decrease full-length INF2 activity while not significantly decreasing activity of an INF2 construct lacking the DID sequence. Profilin-mediated INF2 inhibition is relieved by an anti-N-terminal antibody for INF2 that blocks the DID/DAD interaction. These results suggest that free actin monomers can serve as INF2 activators by competing with the DID/DAD interaction. We also find that, in contrast to past results, the DID-containing N terminus of INF2 does not directly bind the Rho GTPase Cdc42.

whereas the INF2-non-CAAX variant is cytoplasmic and plays a role in Golgi organization (18). INF2 may play additional roles in vesicular trafficking, microtubule stabilization, and centrosome orientation (19 -21). Mutations in the DID region of INF2 result in two human diseases, focal segmental glomerulosclerosis (22) and Charcot-Marie-Tooth disease (23).
Despite the similar C-terminal requirements for full activity, both the exact nature of the formin C-terminal effect and the C-terminal sequence involved are protein-specific for the three best characterized formins (INF2, mDia1, and FMNL3). The DAD of INF2 also binds an actin monomer with submicromolar affinity through an interaction similar to that of a WASp homology 2 (WH2) motif ( Fig. 1 (15)). In contrast, amino acids C-terminal to the DAD of mDia1 play a role in its stimulatory effect on the FH2 domain, but core DAD residues do not appear important to this effect (13). Furthermore, the C terminus of mDia1 binds actin monomers weakly, with an estimated dissociation constant of Ͼ 50 M (13,14). FMNL3 represents a third variation, with a WH2-like sequence N-terminal to, but distinct from, its DAD. This WH2-like sequence binds actin monomers with low micromolar affinity, intermediate between INF2 and mDia1 C termini (14).
One unresolved question concerns the relationship between actin interaction and regulation in the C-terminal region. This question is particularly acute for INF2 because actin-and DIDbinding sequences overlap in the DAD (Fig. 1). Indeed, mutation of three leucine residues in the DAD disrupts both actin and DID binding (15,24). These results suggest that actin monomers and DID may compete for DAD binding, with the result that actin monomers can relieve autoinhibition. In this work, we present evidence supporting this hypothesis.
Cellular Experiments-U2OS human osteosarcoma cells (a gift from Duane Compton, Geisel School of Medicine) were maintained in Dulbecco's modified Eagle's medium with 4.5 g/liter glucose, 584.0 mg/liter L-glutamine, 110.0 mg/liter sodium pyruvate, and 10% calf serum (Atlanta Biologicals) at 37°C and 5% CO 2 . Lipofectamine 2000 (Invitrogen) was used for all plasmid transfections according to the protocol of the manufacturer. 100 ng of each plasmid DNA was used for all transfections, and the cells were analyzed 16 -18 h post-transfection. Cells were fixed with 4% formaldehyde in PBS (pH 7.4) for 15 min at room temperature. After washing with PBS, cells were permeabilized on ice with 0.1% Triton X-100 in PBS for 15 min. Cells were then washed with PBS prior to blocking with 2.5% calf serum in PBS for 1 h at room temperature. Actin was stained using 100 nM TRITC-phalloidin (Sigma). Images were captured using one of the following systems: Wave FX spinning disc confocal system (Quorum Technologies, on a Nikon  Eclipse microscope) using the 491-nm laser and 525/20 filter  for GFP, the 403-nm laser and 460/20 filter for DAPI, and the  561-nm laser and 593/40 filter for Texas red and the laserscanning Nikon A1RSi confocal workstation with a PMT DU4 and Galvano scanner and 405-, 488-, 561-, and 639.5-nm lasers. Images were acquired using Metamorph and were processed using Nikon Elements and Adobe Photoshop CS.
Full-length INF2 was expressed in Sf9 cells using the Bac to Bac expression system (Invitrogen) following the instructions of the manufacturer. Cells were harvested for protein purification 2.5 days post-infection, washed three times in cold PBS, resuspended in EB (50 mM Hepes 7.6 (at 4 ºC), 300 mM NaCl, 1 mM MgCl 2 , 1 mM DTT, 2 g/ml leupeptin, 10 g/ml aprotinin, 1 g/ml calpeptin, 1 g/ml calpain inhibitor 1, 1 g/ml pepstatin A, and 1 mM benzamidine) and lysed by Dounce. The highspeed supernatant was passed over Q-Sepharose followed by Source Q15, eluting from the latter using an NaCl gradient between 150 and 300 mM NaCl. Actin concentration was quantified by its extinction coefficient (25,974 M Ϫ1 cm Ϫ1 at 290 nm). Other protein concentrations were quantified by Bradford assay (Bio-Rad) and confirmed by Coomassie-stained SDS-PAGE using known concentrations of actin as standards. Antibody Preparation-Rabbit polyclonal antibodies against the INF2 N terminus or mDia1 N terminus were produced through Covance Inc. Antibodies were affinity-purified using the GST fusion proteins covalently attached to Sulfo-link resin (Pierce) and eluted in 200 mM glycine-HCl (pH 2.5). Antibodies were dialyzed into PBS.
Actin Polymerization Assays-Unlabeled and pyrene-labeled actin were mixed in G buffer to produce a 5% pyrene-actin stock. This stock was converted to Mg 2ϩ salt by 2 min of incubation at 23°C in 1 mM EGTA/0.1 mM MgCl 2 immediately prior to polymerization. Polymerization was induced by addition of 10ϫ KMEI (500 mM KCl, 10 mM MgCl 2 , 10 mM EGTA, and 100 mM imidazole (pH 7.0)) to a concentration of 1ϫ, with the remaining volume made up by G-Mg (G buffer containing 0.1 mM MgCl 2 instead of CaCl 2 ). Pyrene fluorescence (excitation 365 nm, emission 410 nm) was monitored in a 96-well fluorescence plate reader (Tecan Infinite M1000, Mannedorf, Switzerland). The time between mixing of final components and start of fluorimeter data collection was measured for each assay and ranged between 15 and 20 s.
GST Pull-down Assays-Purified GST fusions were mixed with glutathione-Sepharose 4B beads in assay buffer (1ϫ KMEI ϩ 0.5% thesit ϩ 1 mM DTT) overnight at 4°C, washed in assay buffer, and then the amount of bound GST fusion was quantified by Bradford assay. Cdc42 was charged with 5Ј-guanylyl imidodiphosphate (GMP-PNP) or GDP by mixing 10 mM Cdc42 with 1 mM nucleotide in 1ϫ KMEI. To this solution, EDTA was added to 2 mM from a 500 mM stock, and the mixture was incubated for 10 min at 30°C. An additional 2 mM MgCl 2 was added from a 1 M stock, and the mixture was placed on ice. Binding reactions contained 9 mM GST fusion and 3 mM of the putative binding partner (Cdc42 or INF2-FFC) in assay buffer and were mixed for 30 min at 4°C. After centrifugation (1800 ϫ g for 2 min in a swinging bucket rotor), supernatants were removed, and pellets were washed briefly in assay buffer. Supernatants and pellets were analyzed by Coomassie-stained SDS-PAGE.

RESULTS
To test whether the DID/DAD interaction contributes to INF2 regulation in cells, we mutated a key DID residue for DAD binding in a plasmid expressing full-length GFP-INF2-CAAX and transfected this construct into U2OS cells. This mutation, alanine 149 to aspartate (A149D) is predicted to strongly compromise INF2 DID/DAD affinity on the basis of the effect of the analogous mutation (A256D) in the DID of mDia1 (7,9), which adds both charge and bulk into the hydrophobic binding pocket. Our modeling studies suggest similar structures for INF2 and mDia1 DIDs (Fig. 1). As in our past findings (18,24), wild-type INF2-CAAX associates with the ER but causes only a low level of actin polymerization on the ER surface, as judged by TRITC-phalloidin staining ( Fig. 2A). In contrast, the A149D mutant causes a significant proportion of the phalloidin staining to shift from stress fibers to the ER surface (Fig. 2B). To confirm that the GFP-INF2-CAAX localization represents the ER, we used CFP-Sec61p as an ER marker and found that its staining overlapped with GFP (Fig. 2, A and B). Accumulation of actin on the ER occurs in over 90% of A149D cells examined while occurring in less than 10% of wild-type cells (Ͼ 100 cells examined).
We extended these results by conducting a similar experiment on the INF2-non-CAAX isoform, which localizes to the cytosol with some enrichment in the perinuclear region around the Golgi (18). Wild-type INF2-non-CAAX causes no apparent change in the phalloidin staining pattern compared with nontransfected cells (Fig. 2C), whereas INF2-non-CAAX-A149D causes a marked increase in cytosolic phalloidin staining (Fig.  2D). Quantification of perinuclear phalloidin staining versus cellular GFP intensity shows that the A149D mutant, but not the wild type, causes increased filament accumulation versus the wild type at a range of transfection levels (not shown). These results suggest that the DID and DAD of INF2 participate in an autoinhibitory interaction, similar to other formins, such as mDia1.
However, the results contradict our previous biochemical results showing that the DID-containing N terminus (INF2- To test this hypothesis, we expressed and purified full-length INF2 (non-CAAX) from an insect cell system (Fig. 3A) and tested its actin polymerization activity using a pyrene-actin assay. Full-length INF2 retains equivalent actin polymerization activity to INF2-FFC (Fig. 3B). Unlike INF2-FFC, however, full-length INF2 does not accelerate actin depolymerization (Fig. 3B). This result is also similar to the trans-assay, in which INF2-NT inhibits actin depolymerization by INF2-FFC (24). Thus, the cellular data suggesting DID/DAD-dependent INF2 autoinhibition of actin polymerization are in conflict with biochemical results showing no autoinhibition of polymerization but inhibition of depolymerization.
One possible explanation is that actin monomer binding to the DAD of INF2 competes with DID binding, relieving autoinhibition and activating polymerization. Indeed, INF2-DAD binds actin monomers and resembles an actin monomer-binding WH2 motif (15). To test this hypothesis, we conducted pyrene-actin polymerization assays using varying concentrations of profilin, which binds at an overlapping site on actin to WH2 motifs (30,31). Increasing concentrations of profilin should progressively inhibit the polymerization activity of full-length INF2 if free actin monomers activate INF2 through DAD binding. As a control, we tested the effect of profilin on INF2-FFC polymerization activity and found a minimal inhibitory effect (Fig. 3C). In contrast, full-length INF2 polymerization activity is strongly inhibited by increasing profilin concentration (Fig.  3D). We quantified the polymerization activity of INF2 by measuring the half-time (t1 ⁄ 2 ) to full polymerization, with a larger t1 ⁄ 2 denoting lower polymerization activity. We plotted t1 ⁄ 2 versus free actin monomer concentration (meaning not bound to profilin) in addition to plotting against profilin concentration (Fig. 3E). These results show a concentration-dependent increase in INF2 activity with increasing free actin monomer concentration. To test this effect further, we used the WH2 motif from WASp to compete with the DAD of INF2 for actin monomer binding. Because WASp-WH2 binds actin monomers more weakly than profilin, a higher concentration is needed but produces a similar inhibitory effect on full-length INF2 (but not INF2-FFC) activity (Fig. 3F).
We have shown previously that the DAD-containing C terminus of mouse INF2 binds to actin monomers (15). To verify this interaction for the human protein, we labeled human INF2-Cterm with TMR and then conducted fluorescence anisotropy experiments in the presence of varying concentrations of latrunculin B (LatB)-stabilized actin monomers, similar to past studies. These assays show that TMR-INF2-C binds the actin monomer with a K d app of 0.12 M (Fig. 4A). To test whether the TMR label contributes to the binding affinity, we conducted competition assays in which fixed concentrations of TMR-INF2-C and LatB-stabilized actin monomers are mixed with unlabeled INF2-C. In these experiments, increasing unlabeled INF2-C reduces the anisotropy of TMR-INF2-C, suggesting competitive binding with a K d app of 0.18 M (Fig. 4A, inset).
These experiments show that, similar to mouse INF2, the C terminus of human INF2 binds actin monomers.
Our previous results showed that mouse INF2-Nterm was unable to inhibit mouse INF2-FFC when added in trans (15). We confirmed this result with the human proteins, testing 50 M INF2-Nterm and finding no inhibition on 20 nM INF2-FFC (Fig. 4B). We next conducted these assays in the presence of 12 M profilin, which allowed inhibition of the full-length protein.
Under these conditions, INF2-Nterm inhibits INF2-FFC, whereas the A149D point mutation does not (Fig. 4C). We were unable to reach a saturating concentration of wild-type INF2-Nterm in these assays, testing up to 50 M (Fig. 4D). These results suggest that the DID/DAD interaction of INF2 is lowaffinity compared with that of mDia1 (4,8). Indeed, fluorescence anisotropy assays using TMR-INF2-Cterm failed to show saturating binding at the highest testable concentrations of INF2-Nterm (not shown).
If actin monomers activate INF2 by competing with DID for DAD binding, the reverse scenario should also be true. Factors that compete with DAD for DID binding should activate INF2. This model is similar to the activation mechanism for mDia1 whereby GTP-bound RhoA binds to a region overlapping the DAD binding site (7)(8)(9)12). Another Rho family GTPase, Cdc42, has been proposed to bind and activate INF2, on the basis of assays using cell extracts as the source of Cdc42 (20, 21). We tested the ability of GTP-charged Cdc42 to directly bind the DID-containing N terminus of INF2 using GST pull-down assays similar to those used in the previous work but in which both Cdc42 and INF2 N terminus were present as purified proteins. These assays show no evidence for direct binding between INF2 N terminus and Cdc42, whereas the positive control reactions (Cdc42 binding to the WASp CRIB (Cdc42/ Rac interactive binding) domain and INF2 N terminus binding to INF2-FFC) do display binding (Fig. 5). The Cdc42 used in this experiment was produced in insect cells and purified from the membrane fraction. Thus, it is likely to contain the C-terminal prenyl modification. Use of bacterially expressed, non-prenylated Cdc42 produces similar results (not shown).
We also tested the effect of Cdc42 on full-length INF2 in actin polymerization assays containing profilin. A 50-fold molar excess of GMP-PNP-charged Cdc42 over INF2 fails to change the polymerization time course, suggesting that Cdc42 does not disrupt the autoinhibitory DID/DAD interaction (Fig.  6A). Similarly, use of GDP-charged Cdc42 has no effect on INF2 activity (not shown). In the absence of a physiological INF2 activator, we used a polyclonal antibody against the N terminus if INF2 (anti-INF2-NT) as a means of disrupting the DID/DAD interaction. Full-length INF2-non-CAAX is premixed with the antibody, followed by its addition into pyrene-actin polymerization assays containing profilin. The anti-INF2-NT antibody activates INF2 activity (Fig. 6A), whereas a nonspecific antibody does not (not shown). These results suggest that antibody binding to DID disrupts the DID/DAD interaction, activating actin polymerization.

DISCUSSION
We show here that INF2 is autoinhibited but that free actin monomers can compete with the DID/DAD interaction, activating actin polymerization (Fig. 6B). It is unclear whether this mechanism is generally applicable, although actin monomer binding to regions surrounding the DAD occurs for several formins (13)(14)(15)). mDia1 appears to be autoinhibited even in the presence of free actin monomers (4,32), which might be due to the extremely low affinity of its DAD region for actin or to the fact that the actin binding site does not overlap the DID binding site (13,14). The C-terminal region of FMNL3 binds actin monomers with micromolar affinity, but the region responsible for this binding is distinct from the canonical DAD region. Thus, it might not be predicted to play a significant role in regulation (14,33). Structural information on the autoinhibitory interaction of FMNL formins would be extremely interesting in this respect.
Could actin monomer binding represent a physiological activation mechanism for INF2? On the basis of the cytoplasmic concentrations of actin, profilin, and thymosin ␤4 (a sequestering protein whose actin monomer binding site overlaps that of INF2-DAD), the concentration of free actin monomer is estimated to be in the micromolar range in mammalian cells (34,35). Fluctuations in cellular actin polymerization, increasing or decreasing the free monomer pool, could conceivably increase or decrease INF2 activity, similar (but in an opposite manner) to models of serum response factor-negative regulation by actin monomer levels (36). Because INF2 plays a role in mitochondrial dynamics (17) and because mitochondria play central roles in cellular homeostasis (37,38), such a mechanism could  be a means for monitoring cellular state. Interestingly, the activity of mDia1 is also increased by increases in actin monomer concentration (35,39), but the mechanism of this effect appears different from that of INF2 because this effect does not require DID or DAD regions.
Our findings suggesting no high affinity direct interaction between the N terminus of INF2 and Cdc42 are at odds with previous results (20,21). The differences in the source of Cdc42 may, however, explain the conflicting results. In the previous studies, Cdc42 was from cell extracts, whereas we used purified Cdc42 (both prenylated and non-prenylated). The source of INF2 was very similar in both studies, being a bacterially expressed DID-containing construct fused to the C terminus of glutathione S-transferase (INF2 amino acids 1-340 in Refs. 20, 21 and 1-420 in this study). Therefore, it is possible that Cdc42 and INF2 interact through additional proteins present in cellular extracts and that these additional proteins might serve as stabilizing factors for the DID/DAD interaction.