Structural and Biochemical Basis for the Inhibitory Effect of Liprin-α3 on Mouse Diaphanous 1 (mDia1) Function*

Background: RhoA·GTP activates mDia1 resolving an autoinhibited state leading to formation of unbranched actin filaments in cells. Results: Liprin-α3 uses an α-helical region to bind to mDia1, counteracting mDia1 activation by RhoA. Conclusion: Liprin-α3 competes with RhoA and mDia1 autoinhibition to modulate its activity. Significance: Knowing how mDia1 is inactivated is essential to understand its biology and for therapeutic approaches. Diaphanous-related formins are eukaryotic actin nucleation factors regulated by an autoinhibitory interaction between the N-terminal RhoGTPase-binding domain (mDiaN) and the C-terminal Diaphanous-autoregulatory domain (DAD). Although the activation of formins by Rho proteins is well characterized, its inactivation is only marginally understood. Recently, liprin-α3 was shown to interact with mDia1. Overexpression of liprin-α3 resulted in a reduction of the cellular actin filament content. The molecular mechanisms of how liprin-α3 exerts this effect and counteracts mDia1 activation by RhoA are unknown. Here, we functionally and structurally define a minimal liprin-α3 core region, sufficient to recapitulate the liprin-α3 determined mDia1-respective cellular functions. We show that liprin-α3 alters the interaction kinetics and thermodynamics of mDiaN with RhoA·GTP and DAD. RhoA displaces liprin-α3 allosterically, whereas DAD competes with liprin-α3 for a highly overlapping binding site on mDiaN. Liprin-α3 regulates actin polymerization by lowering the regulatory potency of RhoA and DAD on mDiaN. We present a model of a mechanistically unexplored and new aspect of mDiaN regulation by liprin-α3.

1 m shows that additional mechanisms next to autoinhibition must exist to inactivate formins in a spatiotemporal manner.
Recently, Sakamoto et al. (29) identified liprin-␣3 as a direct mDia1 binding partner. Liprins can be found in vertebrates, flies, and worms (30). They play important roles in regulation of synapse function and assembly but are also involved in regulation of cell motility (31,32). Depletion of liprin-␣ leads to decreased cell migration and cell spreading, whereas its overexpression results in the opposite (33,34).
A region in the liprin-␣ N terminus overlapping the coiledcoil domain has recently been mapped as directly binding the open form of mDia1 contacting the DID of the regulatory N terminus (mDia N ). The expression of this liprin-␣3 domain in HeLa cells displaces mDia1 from the plasma membrane and reduces actin stress fiber formation (29). Binding of active RhoA is not sufficient to fully activate formins, suggesting that additional events, such as post-translational modifications or binding to proteins and phospholipids, are needed (22,43). To switch off formin function, these activating events must be counterbalanced.
Here, we aim at elucidating how liprin-␣3 binds to mDia1 and regulates its function counteracting RhoA activation, resulting in reduction in cellular actin filament formation. Does it use the same binding site as DAD or active RhoA on mDia N , and how is RhoA and DAD binding affected by the presence of liprin-␣3?

Experimental Procedures
Expression and Purification of mDia N , Liprin-␣3, and DAD Fragments-Human RhoA and mouse mDia1 fragments were expressed and purified as described in the final buffer A (50 mM Tris/HCl, pH 7.4, 100 mM NaCl, 5 mM MgCl 2 , 2 mM ␤-mercaptoethanol) (26). Liprin-␣3 fragments and mutants were purified accordingly. Point mutations in mDia1, liprin-␣3, and DAD were created using the QuikChange protocol (Stratagene). All protein concentrations of mDia N, DAD, and liprin-␣3 were determined using the absorption at 280 nm (A 280 ). For DAD(1145-1200)/DAD(1145-1209) and for liprin-␣3 fragments with a low extinction coefficient at A 280 , tryptophan residues were added at the N terminus for concentration determination. The RhoA concentration was determined using the Bradford assay (Expedeon).
Liprin-␣3 Peptide Synthesis-Peptides were synthesized on a Rink amide resin using an automated solid phase peptide synthesizer (MultiSynthec) and following standard Fmoc/tert-butanol strategy. The resin load (Rink amide resin) was 15 mol, and for each coupling step, seven equivalents of Fmoc-amino acid and seven equivalents of activation reagents (ethyl 2-cyano-2-(hydroxyimino)acetate and diisopropylcarbodiimide) were used. N-terminal fluorescein labeling was achieved by using 3 eq of 5,6-carboxyfluorescein (CF) and 3 eq of ethyl 2-cyano-2-(hydroxyimino)acetate and diisopropylcarbodiimide, respectively. All reactions were performed in N,N-dimethylformamide. For final cleavage of the peptides, the resin was treated with TFA/triisopropylsilane/water (95:2.5:2.5 (v/v/v)) for 3 h. Subsequently, peptides were precipitated by the addition of ice-cold diethylether, washed, and lyophilized. All peptides were purified using preparative reverse phase HPLC and analyzed by LC-electrospray ionization-MS (LCQ, Finnigan-MAT, Thermo) with a gradient of 20 -70% acetonitrile (0.1% TFA) in water (0.08% TFA) for  Stopped-flow Kinetics-The experiments were performed at 20°C using a SX20 Applied Photophysics spectrometer (Leatherhead, UK). The mant-GppNHp nucleotide was excited at 350 nm, and the emission was recorded using a 420-nm cut-off filter. For the association kinetics, a constant concentration of 250 nM (final) RhoA⅐mant-GppNHp was titrated with increasing concentrations of mDia N (final 0.6 -40.7 M) without/with the presence of a 2-fold molar excess of liprin-␣3. The obtained data follow a pseudo-first-order binding kinetics. The fluorescence was plotted as a function of time and fitted to a single exponential function to result in the observed rates (k obs ) for each concentration. The slope of the linear fit of the plot of k obs versus protein concentration resulted in the association rate constant (k a ). The dissociation rate constants (k d ) were obtained by mixing a preformed 100 nM (final) mDia N ⅐ RhoA⅐mant-GppNHp complex with a 100-fold excess of unlabeled RhoA Q63L (final 10 M). In the experiments with liprin-␣3, 10 M (final) of liprin-␣3 was added to RhoA Q63L. GraphPad Prism version 6.0 was used for data analysis.
Isothermal Titration Calorimetry (ITC) Measurements-The interactions of liprin-␣3/mDia N , mDia N /DAD, and mDia N / RhoA; the mutational analysis of the liprin-␣3-mDia1 interface; and the competition experiments were performed on an ITC 200 instrument at 20°C (GE Healthcare) (65). For a typical experiment, 2 l of protein in the syringe (200 -500 M) was stepwise injected into the cell solution of protein in concentrations of 20 -50 M, depending on the interaction analyzed. In the competition experiments, 270 M protein in the syringe was titrated to a complex of a 2-fold molar excess of liprin-␣3 (54 M) over mDia1 (27 M) in the cell. The binding isotherms obtained were fitted to a one-site or competitive binding model using Origin version 7.0 and the integrated MicroCal software, in which stoichiometry (N), enthalpy change (⌬H), and the equilibrium association constant (K A ) are directly obtained from the exper-imental data (44). We used the standard EDTA-CaCl 2 sample tests as described by MicroCal to assess the statistical significance of individual observations. These gave values within the manufacturer's tolerances of Ϯ20% for K A values and Ϯ10% in ⌬H.
Fluorescence Polarization Experiments-All experiments were done using a Paradigm Detection Platform (Beckman Coulter). The CF-FITC-labeled F-Lip(567-587) and F-Lip(567-582) peptides were excited at 485 nm, and the emission was detected at 535 nm. The experiments were performed in buffer A at 37°C, and the integration time was 500 ms. F-Lip(567-587) and F-Lip(567-582) (100 nM) were incubated with 10 M mDia N or 10 M mDia N plus 15 M RhoA Q63L. The solutions were preincubated until the equilibrium was reached, and the polarization signal was stable. Measurements without liprin-␣3 peptides served as blank values. Data analysis was done with GraphPad Prism version 6.0.
Crystallization and Structure Determination-Crystals of DID (aa 135-369) and Lip(567-587) grew in mother liquor containing 0.2 M NaCl, 0.1 M Tris/HCl, pH 8.0, 20% PEG 6000 (buffer C) overnight at 20°C using the sitting drop/vapor diffusion method (150 nl of protein at 10 g/liter plus 150 nl of buffer C). The crystals had a final size of ϳ50 ϫ 50 ϫ 50 m and were shock-frozen in liquid nitrogen in buffer C containing 30% (w/v) of D-glucose as cryoprotectant. A native data set was collected at 100 K on beamline X06DA at a wavelength of 1.0 Å using a Dectris Pilatus 2M detector at the Swiss Light Source (SLS), Paul Scherrer Institute (Villigen, Switzerland). The detector distance was 165 mm, the oscillation range was 0.1°, and 1800 frames were collected. The program suite CCP4 was used for structure determination (45). Data were indexed and integrated with MOSFLM version 7.0.9 and scaled with SCALA version 3.3.20 (46 -48). The crystals belonged to the monoclinic space group C2 (C121) and contained two Lip(567-587)⅐DID heterodimers in the asymmetric unit. Initial phases were determined with the program Phaser using residues 135-369 of chain B of the mDia N ⅐RhoC structure (PDB code 1Z2C) to search for two molecules (49). The program Coot version 0.7.1 was used to build the model into the 2F o Ϫ F c and F o Ϫ F c electron density maps in iterative rounds of refinement with REFMAC5 (50,51). In the final model, 100% of the residues are in the allowed regions of the Ramachandran plot, as judged by MolProbity (52,53). All structure figures presented here were made with PyMOL version 1.6.9.0 (54). Data collection and refinement statistics are given in Table 2. As an example, a part of the electron density for the liprin-␣3-peptide (chain D) is shown as an omit map in Fig. 2I. R work is calculated as follows, R work ϭ ⌺͉F o Ϫ F c ͉/⌺F o . F o and F c are the observed and calculated structure factor amplitudes. R free is calculated as R work using the test set reflections only.
Immunofluorescence Analysis of F-actin Content-HeLa cells were grown in DMEM and N2a cells were grown in 1:1 DMEM/ Opti-MEM (plus 10% FCS, 100 units/ml penicillin, and 100 g/ml streptomycin). For transfection, the DNA constructs (0.5 g of pEGFP-N3-RhoA Q63L and mCherry-C1-liprin-␣3) were mixed with 1.5 l of Lipofectamine LTX with 1 l of Plus reagent (Invitrogen) in 100 l of Opti-MEM. Staining of filamentous actin was done with CF647-phalloidin (1:60; Biotium). Immunofluorescence images were taken using an UltraView VoX, PerkinElmer Life Sciences spinning disc confocal microscope. Maximum projection of a confocal Z-stack was used for data processing. The quantification of CF647-phalloidin intensity was done using ImageJ software (National Institutes of Health), determining the mean total fluorescence. Normalization was done to non-transfected cells. The mock-transfected cells were used as a reference to analyze the effect of RhoA G14V or the RhoA G14V/liprin-␣3 overexpression. At least 20 cells were used for analysis per construct. Each construct was transfected and analyzed at least two times.
Analytical size exclusion chromatography runs suggest that Lip(561-587), Lip(567-587), and Lip(457-587) are trimeric in solution (Fig. 1D). Lip(457-587) overlaps in its N terminus with the coiled-coil domain, supporting an oligomerization. However, a low content of secondary structure could also explain the elution profiles observed (Fig. 1D). Lip(217-587) forms a higher oligomer (Fig. 1D). For liprin-␣3 aa 190 -275, a trimeric coiled-coil region is predicted (MultiCoil; data not shown), supporting our analytical gel filtration experiments. The mDia1 DID forms an elongated monomer in solution. Both mDia N and mDia N CC elute as higher oligomers, but from structural data, it is known to form elongated dimers in solution (Fig. 1D). This is supported by coiled-coil predictions ( Fig. 1C) (55). Predictions using MultiCoil showed a low probability for mDia1 to form trimeric coiled-coils (data not shown) (56). We used the DID and a synthesized Lip(567-587) peptide for crystallization trials.

The Liprin-␣3-mDia1 Structure
We determined the atomic structure of the complex of Lip(567-587) and the DID (Lip(567-587)⅐DID) by x-ray crystallography up to a final resolution of 1.65 Å ( Fig. 2 and Table 1).
The crystals belonged to the space group C121 with two copies of Lip(567-587)⅐DID per asymmetric unit. This is the first structure of a part of the liprin-␣3 N-terminal domain in complex with an interaction partner, namely the mDia1 DID. The data showed well defined electron density for the DID as well as for most residues of the Lip(567-587) peptide (Fig. 2I). However, the C-terminal residues 583-587 of liprin-␣3 were not FIGURE 1. Definition of the binding sites in liprin-␣3 and mDia1. A, domain organization of liprin-␣3 and mDia1 and the constructs used in this study. ϩ, interactions of liprin-␣3 fragments with mDia1; Ϫ, fragments not interacting. B, thermodynamic characterization of the shortest liprin-␣3 and mDia1 fragments needed for full interaction using ITC. Lip(567-587) and DID bind with 6.9 M, showing that those fragments contain all residues essential for the interaction. C, coiled-coil prediction of mDia1 and mouse liprin-␣3 using COILS version 2.1. For mDia1, the prediction shows a high score for aa 460 -562 forming a coiled-coil. For liprin-␣3, extended coiled-coil regions were predicted in the N terminus covering residues aa 26 -493 and a second patch in the C terminus from aa 863 to 889. D, analytical size exclusion chromatography of mDia1 and liprin-␣3 fragments as indicated. Shown is the molecular size calculated by calibration (MW calc.) and the molecular weights for the respective monomers (MW mon.). For all proteins, the absorption at 280 nm is shown except for Lip(567-587) (for which A 220 nm is shown). Aprotinin (6.5 kDa), ribonuclease A (13.7 kDa), carbonic anhydrase (29.0 kDa), ovalbumin (43.0 kDa), and conalbumin (75.0 kDa) were used for calibration of the S75 10/300, and for S200 10/300, additionally aldolase (158 kDa), ferritin (440 kDa), and thyroglobulin (669 kDa) were used. visible in the electron density. Using ITC, we found that these residues did not contribute to mDia N binding under the assay conditions because Lip(561-582) binds with 4.5 M affinity ( Table 1).
The structure shows that the liprin-␣3 aa 567-582 (Lip(567-582)) forms an ␣-helix upon binding to the DID Armadillo repeat motifs 3-5 (ARM3 to -5) with Lip(567-582) contacting the third ␣-helices (␣3) of the three-␣-helix-containing ARMs ( Fig. 2A). The third ␣-helices of the ARMs form a concave surface for protein-protein interactions. As stated earlier, the first ARM of the DID is incomplete and contains only helices ␣2 and ␣3 of ARM1 and can therefore be considered as a heat repeat (26,57). The third helix of ARM5 is the extended interdomain helix (ID; ␣17, ␣3 5 ) connecting the DID with the DD (Fig. 2A). Liprin-␣3 makes extensive contacts toward the ID of mDia1, forming salt bridges between Glu-358 D and Glu-362 D (where the superior "D" represents mDia1) in mDia1 and Arg-575 L and Arg-572 L (where the superior "L" represents liprin-␣3) in liprin-␣3 (Fig. 2, B and C). The superposition of the mDia N ⌬G⅐DAD (Protein Data Bank code 2BAP) structure with Lip(567-587)⅐DID shows that the binding sites of liprin and DAD for mDia1 closely match (Fig. 2, D-F). As shown in Fig.  2E, liprin-␣3 overlaps with the DCR (DAD residues 1175-1195) in its C terminus but also with residues following the DCR, including the DAD basic region (DBR; aa 1196 -1209) not visible in the mDia N ⌬G⅐DAD structure but shown biochemically to take the path along the ID (PDB code 2BAP; DAD visible : aa 1196 -1200) (11,26). The presence of the complete DBR in DAD substantially increases the affinity toward mDia1. DAD(1145-1195) binds with a 70-fold reduced affinity compared with DAD(1145-1200), as shown previously (11). Under the conditions used here, the affinity of DAD(1145-1200) toward mDia N is further increased from 37 to 3 nM for DAD(1145-1209) as judged by a competitive ITC-binding assay ( Table 1). Although the major contacts in the DAD-mDia1 interaction were hydrophobic, liprin-␣3 forms hydrophobic interactions as well as specificity-creating salt bridges with DID residues (Fig. 2, B and C). The validation of the geometry of the structure was performed by MolProbity and is shown in Table 2 (52).

Mutational Characterization of the Liprin-␣3 DID Interface
We mutated several hydrophobic residues in mDia1 and found that residues that were reported to completely abolish binding toward DAD, namely A256D D and I259D D (11,14), also strongly interfered with liprin-␣3 binding. Although no binding toward liprin-␣3 could be observed with mDia N I259D D , A256D D reduced the affinity ϳ10-fold from 4 to 42 M (Table 1). Ile-259 D is located directly within the hydrophobic interface with liprin-␣3, whereas Ala-256 D is located at its edge (Fig. 2B). Interestingly, N165D D , which abolishes RhoA⅐GTP binding without affecting DAD binding, also does not affect liprin-␣3 binding (Table 1). For liprin-␣3, the mutations M576E L , L573E L , and L580E L within the hydrophobic core of interaction did completely block binding toward mDia1, as did the mutations R569E L , R572E L , and R575E L , the latter two resolving the salt bridges toward Glu-362 D and Glu-358 D in mDia1. Opposite charge mutations of E358R D and E362R D also completely abolished binding toward liprin-␣3, showing the importance of these electrostatic interactions ( Table 1). The integrity of the mutant mDia1 proteins (N165D D , A256D D , I259D D , R317E D , E362 D , and E358 D ) used was shown previously examining RhoA and DAD affinities by stopped flow and ITC (14,26). We analyzed the DID and liprin-␣3 mutants by analytical gel filtration (data not shown). All proteins showed a running behavior and elution profile similar to those of the wild type proteins (Fig. 1D). To analyze how liprin-␣3 mechanistically impacts the interaction dynamics of mDia N and Rho/ DAD, we performed kinetic studies using stopped flow.

Liprin-␣3 Alters mDia1-RhoA Interaction Dynamics
Association-Based on the superimposed structures of RhoC⅐mDia N (PDB code 1Z2C) and Lip(567-587)⅐DID, it is not directly obvious whether and how liprin-␣3 interferes with Thermodynamic characterization of liprin-␣3 and mDia1 by ITC K D is the equilibrium dissociation constant, ⌬H is the enthalpy, ⌬S is the entropy change, and N is the stoichiometry of the interaction. The reaction partner in the syringe (S) (300 -400 M) was titrated to the protein in the cell (C) (30 -40 M). In the cases of the competition assays, a 270 M concentration of the protein in the cell was titrated to 27 M complex in the cell.

Liprin-␣3 Negatively Regulates Function of Formin mDia1
RhoA binding to mDia1 (Fig. 2F). Sakamoto et al. (29) concluded from their cell biological data that on the one hand, liprin-␣3 does only bind to the open, non-autoinhibited form of mDia1. On the other hand, RhoA, mDia1, and liprin-␣3 were not able to form a ternary complex. We wondered how RhoA displaces liprin-␣3 from mDia1, either using an allosteric mechanism or influencing liprin-␣3 binding sterically or through long range electrostatic effects. Both might involve the liprin-␣3 C-terminal residues not visible in the crystal structure (aa 583-587). Because RhoA shows a more than 1000-fold tighter association with mDia1 compared with liprin-␣3 (4 nM versus ϳ5 M; Table 1), we tested whether the presence of liprin-␣3 at all influences the mDia N -RhoA interaction kinetics. To this end, we performed pre-equilibrium stopped-flow experiments ( Fig. 3 and Table 3).
RhoA loaded with fluorescently labeled mant-GppNHp (a non-hydrolyzable GTP analogue) was titrated with increasing concentrations of mDia N (pseudo-first order conditions) in absence/presence of a 2-fold molar excess of different liprin-␣3 fragments (Lip(217-587), Lip(561-587), Lip(567-587), and Lip(561-582)). The observed association rates (k obs ) were obtained by a single exponential fit of the primary data (not shown). For the reaction in the absence of liprin-␣3, linear fitting of k obs against increasing mDia N concentrations yielded the second-order association rate constant, k a ϭ 1.0 M Ϫ1 s Ϫ1 (Fig. 3A). This linear behavior was also observed for the mDia N -RhoA associations in the presence of all liprin-␣3 fragments tested. Interestingly, all liprin-␣3 fragments similarly reduced the association rates (Lip  Table 3). This suggests that the functionally important region in liprin-␣3 mediating this effect is located within the liprin-␣3 residues 567-582, the residues visible in the crystal structure.
Dissociation-We determined how liprin-␣3 fragments influence the RhoA⅐mDia N dissociation rate constant (k d ) by preforming a complex of mant-GppNHp-labeled RhoA and mDia N and mixing this with a 100-fold excess of unlabeled active RhoA Q63L in the absence/presence of equimolar liprin-␣3 concentrations. We found that the primary dissociation The observed association rate constants (k obs ) were plotted against the mDia N concentration. The association rate constants (k a ) correspond to the slope of the linear fit. B, dissociation rate constant. A preformed mDia N ⅐RhoA⅐mant-GppNHp complex was mixed with a 100-fold excess of active RhoA Q63L, and the increase in fluorescence was followed over time. The data followed a single exponential behavior. Error bars, S.D.

TABLE 2 Data collection and refinement statistics and validation of the geometry of the Lip(567-587)⅐DID structure (molecular replacement)
Values shown in parenthesis for data collection and refinement statistics are for the highest resolution shell. One single crystal was used for data collection. For the definitions of R free and R work , see "Experimental Procedures."

Data collection
Space

TABLE 3 Effect of different liprin-␣3 fragments on the RhoA-mDia1 interaction dynamics as determined by stopped-flow analysis
The k a is the association rate constant, and k d is the dissociation rate constant.

Liprin-␣3 Negatively Regulates Function of Formin mDia1
kinetics data followed a single exponential behavior as reported earlier under slightly different reaction conditions (here, 100 mM NaCl; earlier, 300 mM NaCl; Fig. 3B). The dissociation rate (2.35 ms Ϫ1 ) obtained here for the RhoA⅐mDia N dissociation agrees with the dissociation rate reported earlier (26). We see that the overall dissociation rate, k d , is only slightly increased by the presence of each liprin-␣3 fragment tested (2.82-3.22 ms Ϫ1 ; Fig. 3B and Table 3). In summary, the presence of liprin-␣3 influences the dynamics of the mDia N -RhoA interaction, leading to a much slower RhoA-mDia N association and an only marginally faster dissociation. This holds to a comparable extent for all of the liprin-␣3 fragments tested, again showing that Lip(567-582) does contain all residues mediating the observed effects on the mDia N -RhoA interaction dynamics. To quantify the inhibitory potency of liprin-␣3 on the interactions of mDia1 and RhoA/DAD under equilibrium conditions, we performed competition measurements using ITC.

Displacement of Liprin-␣3 from mDia1 by Active RhoA and DAD(1145-1200)/DAD(1145-1209)
For quantification of the inhibitory potency of liprin-␣3 on the mDia N -RhoA and mDia N -DAD interactions, we performed ITC experiments. We preformed a complex between mDia N and liprin-␣3 using a molar excess of liprin-␣3 over mDia N and titrated this with active RhoA Q63L or DAD.
DAD-Structurally, it is obvious that mDia N cannot simultaneously bind to liprin-␣3 and DAD because they use a highly overlapping binding site on mDia N (Fig. 2, D-F). We performed competition ITC experiments by titrating DID with Lip(561-587) until saturation and injected DAD(1145-1200) into this complex. The presence of liprin-␣3 decreases the DID-DAD(1145-1200) binding affinity from 37 to 155 nM, as judged from fitting the data with a competitive and one-site fit model (Fig. 4B). The affinity obtained by the competitive fit model agrees with a direct measurement of DAD(1145-1200) mDia N interaction under the same reaction conditions (29 nM; Table  1). Notably, this competitive fit model is only an approximation because, although the liprin-␣3 DAD(1145-1200) interaction sites highly overlap, they are not identical. The reaction is mainly entropically driven (T⌬S ϭ ϩ20.2 kcal mol Ϫ1 ) with an unfavorable enthalpic contribution. The enthalpy of the competition reaction (⌬H) corresponds to the enthalpy of direct Lip(561-587) binding to DID but with the opposite sign (from ϩ11.0 to Ϫ10.4 kcal mol Ϫ1 ; Fig. 4B and Table 1). To show that the DAD(1145-1200) displaces liprin-␣3 from mDia N , we ana-

Liprin-␣3 Negatively Regulates Function of Formin mDia1
lyzed the ITC cell content by analytical size exclusion chromatography (S75 10/300; Fig. 4C). Indeed, DAD(1145-1200) eluted with DID in the same peak, whereas liprin-␣3 eluted alone in the second peak. We furthermore showed that the longer Lip(457-587) can be displaced from mDia N ⌬GCC using the DAD(1145-1209) fragment, including the DBR (Table 1). These experiments explain why liprin-␣3 is only capable of binding to the open, active form of mDia1, because the affinity of mDia N toward DAD(1145-1200) and DAD(1145-1209) is more than 2 and 3 orders of magnitude higher compared with liprin-␣3, respectively. As a conclusion, liprin-␣3 is dissociated by DAD using a highly overlapping binding site on mDia N due to its significantly higher binding affinity toward mDia N .
RhoA-Because both RhoA and liprin-␣3 use distinct binding sites on mDia N , we did not apply a competitive model as described for DAD but instead used a one-site model to fit the data. This shows that the presence of liprin-␣3 does reduce the binding affinity (K D ) to active RhoA Q63L from 4 nM to an apparent affinity of 40 -60 nM (Fig. 4A and Table 1). The stoichiometry for the RhoA-mDia N -Lip(567-587) interaction was 1:1, and the reaction shows an exothermic heat profile at 20°C with a total enthalpy change of Ϫ2.0 kcal mol Ϫ1 . The reduction in favorable enthalpy compared with the mDia N -RhoA interaction in the absence of liprin-␣3 (from Ϫ4.0 to Ϫ2.0 kcal mol Ϫ1 ) agrees with the release of liprin-␣3 from mDia N by RhoA binding. The reaction is mainly driven by the favorable, positive entropic contribution (T⌬S ϭ 7.9 kcal mol Ϫ1 ). As a control, we performed these experiments with Lip(561-587) and Lip(561-582) to show that the C and N termini of liprin-␣3 have no additional contribution. C-terminally truncated Lip(561-582) reduces the RhoA-mDia N affinity similarly to 41 nM, and Lip(561-587) reduces it to 58 nM (Table 1). Again, these reactions show exothermic heat profiles, and the reactions are entropically favored. These data show that the presence of liprin-␣3 interferes with the association of RhoA and mDia N , although both use a distinct binding site on mDia N (Fig. 2F). As a control, we excluded a direct binding of liprin-␣3 toward RhoA by ITC (data not shown). Our results support an allosteric displacement mechanism of liprin-␣3 by RhoA rather than competing for the same site on mDia N . However, because the ITC data for RhoA do not exclude a ternary complex formation between mDia N ⅐liprin-␣3 and RhoA, we next performed fluorescence polarization experiments to test whether RhoA displaces liprin-␣3 from mDia N .

Does Liprin-␣3 Form a Ternary Complex with mDia N and RhoA?
In order to examine whether liprin-␣3, mDia N , and RhoA can form a ternary complex, we synthesized N-terminally CFlabeled peptides of Lip(567-587) (F-Lip(567-587)) as well as Lip(567-582) (F-Lip(567-582)), which lacks the C-terminal residues 583-587 not visible in the structure but oriented toward RhoA in a putative ternary complex (Fig. 4D). In fluorescence polarization experiments, we used a 100 nM concentration of the fluorescently labeled liprin-␣3 peptide and added 10 M mDia N . We observed an increase in the fluorescence polarization signal for both liprin-␣3 peptides tested, reflecting a complex formation between both liprin-␣3 peptides and mDia N . To test whether RhoA dissociates the complexes of the liprin-␣3 peptides and mDia N , we added 15 M active RhoA Q63L. This led to a drop of the fluorescence polarization signal to the baseline for Lip(567-587) and Lip(567-582), indicating the dissociation of the liprin-␣3⅐mDia N complexes. These results show that the liprin-␣3 residues 583-587 are dispensable for the displacement of liprin-␣3 from mDia N by active RhoA. As a support, we also performed analytical size exclusion chromatography with mDia N , RhoA Q63L, and Lip(457-587) to exclude the formation of a ternary complex (data not shown).
For Lip(567-587), we placed a longer polypeptide linker between the Lip(567-587) coding region and the N-terminal mCherry (Lip(3GS)(567-587)), preventing the possibility that the mCherry protein sterically occludes the binding of Lip(567-587) to mDia1. Indeed, placing the linker between the mCherry and the Lip(567-587) restored the actin-reducing effect, showing the importance of the N terminus of Lip(567-587) for its function (Fig. 5, A and B). This reflects the mechanistic relevance of the N-terminal region contacting the mDia N ID/DD.
Notably, Sakamoto et al. (29) showed a reduction in cellular F-actin for endogenous full-length liprin-␣3 by siRNA knockdown in HeLa cells. However, we could not observe this overexpressing full-length liprin-␣3 in HeLa cells. To show whether this is cell type-dependent, we overexpressed full-length liprin-␣3 in murine neuroblastoma N2a cells. Indeed, we observed that full-length liprin-␣3 significantly reduces the actin filament content similarly to Lip(561-587) and Lip(3GS)(567-587) (Fig. 5C). As a control, we analyzed the overexpression levels of full-length liprin-␣3 in HeLa and N2a cells by immunoblotting. Both cell types expressed liprin-␣3 substantially and to a comparable extent, with the N2a cells showing an only marginally stronger expression (data not shown). This excludes the possibility that the differences observed in the F-actin content in the two cell types are due to prominent divergent expression levels. Additionally, full-length liprin-␣3 accumulates at the cell periphery in N2a cells in contrast to HeLa cells (Fig. 5, A  and D).

Discussion
The Minimal and Functional Liprin-␣3-Lip(567-582) (liprin-␣3 core region; LCR) affects the mDia N -RhoA interaction dynamics, competes with RhoA and DAD for binding to Liprin-␣3 Negatively Regulates Function of Formin mDia1 mDia N , is not able to form a ternary complex with RhoA/DAD and mDia N , and leads to an actin-reducing effect similar to that of longer liprin-␣3 fragments. We showed that DAD and active RhoA can displace liprin-␣3 from mDia1.
Under physiological conditions, the oligomeric states of mDia1 and liprin might strongly affect its capacity to modulate the interactions of mDia1 and RhoA/DAD. We determined in ITC experiments that the N-terminally longest fragment

Liprin-␣3 Negatively Regulates Function of Formin mDia1
tested, Lip(217-587), binds mDia1 with a 2-3-fold reduced affinity compared with the shorter fragments with a stoichiometry of 0.5 (shorter fragments 1:1). This indicates that one mDia1 molecule binds two liprins, independently of the mDia1 oligomeric state of the fragments tested (Table 1). Additionally, the stopped-flow data showed the lowest effect on the mDia N -RhoA kinetics for Lip(217-587). The formation of a higher oligomeric state in Lip(217-587) might render one mDia-binding site inaccessible. For full-length liprin-␣3, we only found a reduction of the cellular F-actin content in neuronal N2a cells. Furthermore, liprin-␣3 was mostly localized to the cell periphery, in contrast to HeLa cells (Fig. 5D). Therefore, cell type-dependent unidentified mechanisms must exist in cells to bring liprin-␣3 to an mDia1 binding-competent and/or -inactivating state. These may involve liprin oligomerization or a functionality residing in the liprin C terminus, such as recruitment to membrane receptors.
Interplay of Liprin-␣3 and DAD-Deletion and/or mutation of residues ( 1196 RRKR 1199 ) within the DBR (aa 1196 -1209) influences actin filament formation (17). Opposite charge mutations E358R D , E362R D , and D366R D in the mDia N ID decreased binding toward DAD(1145-1200) (26). The Lip(567-587)⅐DID structure presented here shows that Lip(567-587) overlaps significantly with the C-terminal part of the DCR (aa 1175-1195) on mDia N and that it takes a path along the negatively charged ID also used by the DBR (Fig. 2, B  and E). For the DBR, it was postulated that it clusters plasma membrane PIP 2 and inserts into the lipid bilayer partially responsible for an inhibition of mDia1-mediated actin filament elongation (58). The presence of liprin-␣3 might support this by competing with the DBR for binding to the negatively charged ID, reducing the competitive potency of the ID for PIP 2 to bind to the DBR (Fig. 2, B and E). Whether liprin-␣3 has the same inhibitory potency in vivo is not clear, taking into account the intermolecular interaction of mDia N and DAD as studied here, whereas it is intramolecular in vivo and therefore most likely even more potent.
In this context, a comparison of the consensus sequence defined for the DCR, (G/A)(V/A)MDXLLEXL(K/R/Q)X(G/ A)(S/G/A)(A/P), and Lip(567-587) shows that the residues 583-587 not visible in the structure are a perfect match with the C-terminal DCR consensus (Lip(567-587), N-TPRSARLER-MAQALALQAGSP-C) (16). However, DAD and liprin-␣3 are oriented in the opposite direction on mDia1 (Fig. 2E). Determination of whether this has any physiological regulatory role will require further investigation.
Displacement of Liprin-␣3 by RhoA-From the structure presented here, it is not directly obvious how RhoA displaces Lip(567-582) from DID because both use completely independent binding sites (Fig. 2F). Although RhoA binds 3 orders of magnitude tighter to mDia N than liprin-␣3, its presence can weaken the RhoA-mDia N interaction about 10 -15-fold. These structural and equilibrium ITC data support an allosteric displacement mechanism of liprin-␣3 from mDia N by RhoA.
The displacement of liprin-␣3 from mDia N by RhoA is mainly entropically driven. Lip(567-587) forms an ␣-helix in complex with mDia N , but it might be unstructured in the unbound form, which could explain the thermodynamics observed. In initial BioSAXS experiments, we found that unbound liprin-␣3 shows a high degree of flexibility in solution indicative of a low amount of secondary structure (data not shown). The mDia1 DID from the structure presented here superposes well with DID domains from other structures (Fig.  2G). However, there is conformational flexibility within the C-terminal half of the mDia N ID and the DD (Fig. 2G). Liprin-␣3 forms important salt bridges with residues from the ID (Fig. 2, B, C, and E). Small alterations in the conformation of the ID/DD induced by RhoA-binding to mDia N might interfere with liprin-␣3 binding and might support liprin-␣3 dissociation.
Our kinetic data show that the presence of liprin-␣3 reduces the association rates of mDia N and RhoA, whereas the dissociation is nearly unaffected with the concentrations used. It was reported earlier (i.e. for the interaction of Raf-RBD and Ras) that electrostatics play an important role for the association Binding to the DID lowers the DAD affinity moderately, whereas the affinity toward RhoA is more strongly affected. In this way, the DAD is able to compete with RhoA to re-establish the autoinhibited conformation. Moreover, Rho regulators, such as RhoGAPs, could compete with mDia N for RhoA binding, leading to subsequent RhoA and thereby mDia1 inactivation. The mDia1 membrane translocation could be achieved by liprin-␣3 competing for DID/formin interaction partners, such as IQGAP, and by RhoA dissociation and inactivation. The LCR might also interfere with the interplay of the DBR, PIP 2 , and ID, resulting in formin inhibition and reduction in cellular filamentous actin. For simplicity, the model does not include the subcellular localization of mDia1/liprin-␣3 and the dimeric state of mDia1 and possible oligomeric (homo-and/or heterooligomers) states of liprin-␣3. a.u., arbitrary units. rates in protein complex formation (59,60). The opposite charge mutation K133E in the Rho insert helix region was shown to strongly reduce the association rate of RhoA toward mDia N (26). Binding of liprin-␣3 to DID occurs at a highly negatively charged patch along the ID, which is in spatial proximity to the positively charged Rho insert helix in the complexes of mDia N ⅐RhoC (PDB code 2Z2C) and mDia N (TSH)⅐Cdc42 (PDB code 3EG5) (Fig. 2H) (14,26). Liprin-␣3 might interfere with the electrostatic attraction of RhoA by mDia N in a first association step.
The association rates obtained for mDia1 and RhoA are 2-3 orders of magnitude lower than expected for a diffusion-controlled interaction. This suggests that there might be at least a two-step binding with a rate-limiting step possibly involving conformational changes, which we could not observe under the conditions used.
Mechanisms of How Liprin-␣3 Inhibits Formin Function-The shortest still functional liprin-␣3 fragment, Lip(567-582), is presumably not capable of sterically interfering with the catalytic FH1 and FH2 domains. It is unlikely that it simultaneously recruits further factors such as RhoGAPs to down-regulate RhoA activity or factors mediating mDia1 membrane translocation. We postulate that liprin-␣3 down-regulates mDia1 function by several possible mechanisms (Fig. 5E).
First, liprin-␣3 directly interferes with the regulatory machinery to inactivate mDia1 by modulating the RhoA interaction and DAD autoinhibition. The RhoA binding to mDia (K D ϭ 4 nM) is slightly higher compared with the DAD(1145-1209), including the DBR (K D ϭ 3-14 nM). However, the presence of liprin-␣3 reduces the mDia1-RhoA interaction (10 -15fold to 40 -60 nM) more strongly than the mDia-DAD interaction (3-4-fold to 23 nM). Thereby, the capacity of DAD to compete with RhoA for mDia N binding and to re-establish the autoinhibited, closed conformation of the formin is increased. Liprin-␣3 might also directly affect the DBR function and its interplay with membrane-bound PIP 2 and would thereby result in subsequent formin inactivation.
Second, the 10 -15-fold reduction of RhoA-mDia1 affinity might be sufficient in vivo for GAP proteins or other regulators to compete with mDia1 for RhoA binding and its following inactivation. Affinities of many RhoGAPs toward Rho proteins loaded with GTP or its transition state mimic GDP⅐AlF x are reported to be in the low nanomolar range (61).
Third, the presence of liprin-␣3 was shown to translocate an N-terminal mDia1 fragment from the plasma membrane (29). This effect could be due to interference with RhoA binding. Alternatively, it was shown that IQGAP recruits mDia1 to the plasma membrane in phagocytic cups by direct binding to the DID (62, 63) (Fig. 5E). The LCR directly targeting the same binding site as IQGAP on mDia N could thereby interfere with mDia1 membrane translocation. Moreover, neurochondrin, anilin, and the recently identified F-Bar protein CIP-4/Toca-1 have been shown to be involved in formin membrane translocation (64). Determination of whether this is affected by liprins will require further investigation.
Conclusion-Here, we present a detailed mechanistic view down to atomic resolution of how the liprin-␣3 regulates the interplay of mDia1, RhoA, and the mDia1-autoinhibitory switch. However, it is still unclear how liprin-␣3 can bind to mDia N in the first place, considering the much higher affinities of DAD and RhoA for mDia N . Therefore, events that stabilize the formin's open conformation are a prerequisite. Binding to phospholipids and adaptor proteins or post-translational modifications could cooperate in this process. Future investigations will show how liprin-␣3 regulates mDia1 function under physiological conditions. This includes the determination of the effective cellular concentrations of liprin-␣3, mDia1, and RhoA, dependent on the cell and tissue type. Additionally, the oligomeric states of liprins and the formins and its subcellular localization (receptor/membrane-bound versus cytosolic) might also affect its interplay. The results presented here might be of therapeutic interest to inactivate formin function and thereby actin filament formation in processes such as tumor invasion and metastasis.