Biophysical Mechanism of Allosteric Regulation of Actin Capping Protein

Abstract

Actin capping protein (CP) can be regulated by steric and allosteric mechanisms. The molecular mechanism of the allosteric regulation at a biophysical level includes linkage between the binding sites for three ligands: F-actin, Capping-Protein-Interacting (CPI) motifs, and V-1/myotrophin, based on biochemical functional studies and solvent accessibility experiments. Here, we investigated the mechanism of allosteric regulation at the atomic level using single-molecule Förster resonance energy transfer (FRET) and molecular dynamics (MD) to assess the conformational and structural dynamics of CP in response to linked-binding site ligands. In the absence of ligand, both single-molecule FRET and MD revealed two distinct conformations of CP in solution; previous crystallographic studies revealed only one. Interaction with CPI-motif peptides induced conformations within CP that bring the cap and stalk closer, while interaction with V-1 moves them away from one another. Comparing CPI-motif peptides from different proteins, we identified variations in CP conformations and dynamics that are specific to each CPI motif. MD simulations for CP alone and in complex with a CPI motif and V-1 reveal atomistic details of the conformational changes. Analysis of the interaction of CP with wild-type (wt) and chimeric CPI-motif peptides using single-molecule FRET, isothermal calorimetry (ITC) and MD simulation indicated that conformational and affinity differences are intrinsic to the C-terminal portion of the CPI motif. We conclude that allosteric regulation of CP involves changes in conformation that disseminate across the protein to link distinct binding-site functions. Our results provide novel insights into the biophysical mechanism of the allosteric regulation of CP.

Introduction

Actin assembly and disassembly are important for directing the shape and movement of cells and tissues during normal development and physiological function as well as for aiding in the motility of some pathogens upon infection ^{1 2}. Actin filaments grow and shrink by gaining and losing monomeric subunits from their two ends – barbed and pointed. Actin filaments near membranes can push on membranes by growth from free barbed ends, following nucleation by Arp2/3 complex. Actin and Arp2/3-based assembly and motility can require contributions from actin capping protein (CP), ADF/cofilin, profilin and thymosin ^{2, 3}. Regulation of filament polymerization and depolymerization via actin capping protein (CP) and its binding partners is essential to provide the force and energy for movements of many cellular membranes, including vesicles and the plasma membrane.

CP is expressed in a wide range of organisms, from budding yeast to humans ⁴. In vertebrates, CP is as an α / β heterodimer, and each subunit occurs as three distinct isoforms. The α1, α2, and α3 isoforms are expressed from three different genes ⁵. The β isoforms are all expressed from one gene and differ by alternative splicing ⁶. α3 and β3 are expressed exclusively in male germ cells ^{7, 8}. α1and α2 are expressed in a wide variety of cells and tissues, at different ratios ⁹. The β1 isoform is expressed in striated muscle and is located at Z-lines, where it binds the barbed end of the sarcomeric thin filament ⁶. CP purified from skeletal muscle contains the β1 isoform, which is often called “CapZ” because of its Z-line localization ¹⁰. The β2 isoform is predominantly expressed in non-muscle cells and tissues; β2 is also expressed in striated muscle cells, and it does not localize to Z lines ^{6, 11}. The molecular basis for their differing localization in striated muscle is not known.

The CP heterodimer adopts a mushroom-like shape ⁴. The mushroom cap is comprised of interlaced β sheets and α helices. The mushroom stalk protrudes perpendicularly from the bottom of the cap surface and is composed of α helices (Figure 1). The top surface of the mushroom cap binds to filament barbed ends, preventing association and dissociation of actin subunits ⁴.

Figure 1. Illustration of dye positions, using previously published structures of CP-ligand complexes (PDB 3LK3 and 3AAA).

FRET construct positions are shown as teal circles (α9β161) and orange circles (α44β84). CP α subunit shown as dark green ribbon, and CP β subunit as light green ribbon. CARMIL1 CPI-motif peptide shown as blue space-filling (modified from PDB 3LK3), and V-1 shown as red space-filling (modified from PDB 3AAA).

A number of biomolecules bind directly to CP and regulate the interaction of CP with barbed ends. Those molecules comprise polyphosphoinositides, the protein V-1 / myotrophin, and a diverse set of proteins with CP-interacting (CPI) motifs, including the CARMIL, CKIP and WASHCAP (FAM21) protein families ^{4, 15, 16}. In cytoplasm, CP and V-1 are both present at high concentration in micromolar quantities, they both diffuse freely, and they bind tightly to each other with nanomolar affinity ¹⁷. In contrast, CPI-motif proteins are present in far smaller amounts, and they are generally targeted to specific membrane locations ^{15, 16}. The effect of CPI-motif binding to CP includes weakening the binding affinity of CP for V-1 ^{18, 19}. These observations raised the possibility, proposed by Hammer and colleagues ¹⁸, that CPI-motif proteins activate CP locally at a membrane by promoting dissociation of the CP inhibitor V-1 The binding of CP to V-1, F-actin, and CPI motifs, are linked molecular processes ¹⁹. V-1 binds directly to the cap region of CP and sterically blocks the site required for binding to actin-filament barbed ends ^{17, 18, 20-22}. Structures of co-complexes show that the binding sites for F-actin and V-1 overlap extensively, but not completely ¹⁹. In contrast, the binding sites for CPI-motif proteins are instead located in the stalk region of CP and are spatially different from those for F-actin and V-1; they bind to the stalk region of CP (Figure 1) ^{15, 16, 21, 22}. Therefore, CPI-motif proteins inhibit the binding of CP to actin filament barbed ends and to V-1 by an allosteric mechanism ^{15, 16, 21, 22}.

The physical basis of the allosteric regulation and linkage between these two distinct sites was investigated with hydrogen-deuterium exchange mass spectrometry (HDX-MS) ²³. The HDX-MS study revealed that the solvent accessibility of CP was altered at both binding sites when either a CPI-motif peptide or V-1 was added to CP in solution. The findings indicate that interaction with either ligand can induce conformational changes to sites within CP that are distant from the ligand binding surface. Available crystal structures of co-complexes show only slight differences in the conformations of CP when complexed with a CPI-motif peptide or V-1 ^{21, 24, 25}; however, rather large biochemical effects are observed when CP is bound to either ligand ^{18, 19, 21}. Together, these findings are consistent with linked changes in the conformation and/or the structural dynamics of the two sites. The atomistic details of the linkage are not clear because the spatial resolution of HDX-MS is limited to proteolytic peptides, and crystal structures do not capture potential dynamics within different states.

Here, to advance our understanding of the allosteric linkage mechanism that regulates CP function, we sought evidence at the atomic level for changes in the conformation or dynamics of CP in solution, by comparing free CP (Apo-CP) with CP complexed with either a CPI-motif peptide or V-1. We chose to use CP composed of the α1 and β2 isoforms because this is commonly found in non-muscle cells. We performed confocal single-molecule Förster resonance energy transfer (FRET) of molecules in solution, which allows one to quantify conformational changes in the protein alone and in complex with regulators. We complemented the experimental results with molecular dynamics (MD) analyses of the conformation of CP in those settings, to provide an atomistic description of the structural changes identified by single-molecule FRET.

Materials and Methods

Protein purification and labeling.

CP Regulators.

Human V-1 (UniProt P58546) was purified as described ²³. CPI-motif peptides from human CARMIL1, WASHCAP, and CKIP-1 were obtained and used as described ¹⁹.

Wild-type CP.

Wild-type mouse CPα1β2 heterodimer (UniProt Q5RKN9 and Q923G3) (pBJ 2041), assembled from the two co-expressed subunits, was purified as described ²³.

Cys-null CP.

To enable the introduction of specific labeling sites in CP, all nine intrinsic Cys residues were changed to Ser using the mutagenesis service of GeneWiz (South Plainfield, NJ). A bacterial expression plasmid for mouse CPα1β2 (pBJ 2041) ²³ was modified to produce a “Cys-null” CP expression plasmid (pBJ 2454). This plasmid simultaneously expressed two CP subunits: His-tagged mouse CP α1 (C124S, C141S, C157S) and non-tagged mouse CP β2 (QC8S, C36S, C62S, C147S, C206S, C272S). The CP heterodimer, assembled from the two co-expressed subunits, was purified as described ²³. The heterodimer was stable during purification.

We assayed Cys-null CP for two biochemical activities: binding to a CARMIL1 CPI-motif peptide with ITC, and capping of F-actin barbed ends with pyrene-actin polymerization assays, both performed as previously described ^{7, 14}. The assays for capping F-actin barbed ends showed similar activities for wild-type CP and Cys-null CP (Supplementary Figure 1). Wild-type CP and Cys-null CP showed equivalent binding to a CARMIL1 CPI-motif peptide (Supplementary Table 1).

CP FRET Constructs.

For each single-molecule FRET experimental construct, two Cys residues were introduced on the surface to permit labeling with donor and acceptor fluorophores. Residues were chosen for labeling based on considerations described in the Supplementary Text. The Cys residues corresponding to the chosen labeling positions were introduced in the Cys-null CP expression plasmid (pBJ2454) by GeneWiz. The first construct used α1 S9C and β2 S161C mutations (pBJ 2478), and the second construct used α1 N44C and β2 E84C mutations (pBJ 2488).

CP subunits were co-expressed from one plasmid in E. coli NiCo21(DE3) (New England BioLabs, Ipswich, MA) as described ²³. The fusion protein was isolated on Ni Sepharose^® 6 Fast Flow (Cytiva, Marlborough, MA). PreScission protease (GenScript, Piscataway, NJ) was added to cleave the His-tags. Eluted CPs were bound to ceramic hydroxyapatite type I, 40 mm (Bio-Rad, Hercules, CA) in 20 mM Tris-HCl, 20 mM NaCl, 2 mM NaH₂PO₄, 125 mM CaCl₂, 1 mM NaN₃, 5 mM DTT (pH 7.5) and eluted with a linear gradient to 200 mM NaH₂PO₄, 20 mM NaCl, 125 mM CaCl₂, 1 mM NaN₃, 5 mM DTT (pH 7.6). Fractions containing purified CPs were dialyzed into 20 mM Bis-Tris-propane, 1 mM Tris(2-carboxyethyl)phosphine (TCEP), 0.1 M NaCl, 1 mM NaN₃ (pH 7.0).

Dual-labeled variants were obtained by sequential labeling, with labeling efficiency confirmed by mass spectrometry. CPs were diafiltered into 20 mM Bis-Tris-propane, 0.1 M NaCl, 1 mM NaN₃ (pH 7.0), and immediately labeled with 0.7 mol Alexa Fluor 488 C5 maleimide (Invitrogen, Carlsbad, CA) per mol CP. Alexa 488-labeled CPs were isolated on Mono-Q HR 5/5 (Cytiva) in 20 mM Bis-Tris-propane, 1 mM TCEP, 1 mM NaN₃ (pH 7.0), and eluted with a NaCl gradient. Alexa 488-labeled CPs were diafiltered into 20 mM Bis-Tris-propane, 0.1 M NaCl, 1 mM NaN₃ (pH 7.0), and immediately labeled with 5 mol Alexa Fluor 594 C5 maleimide (Invitrogen) per mol CP. Alexa 488, Alexa 594-labeled CPs were isolated on Mono-Q HR 5/5 (Cytiva) in 20 mM Bis-Tris-propane, 1 mM TCEP, 1 mM NaN₃ (pH 7.0), and eluted with a NaCl gradient. Column fractions were frozen individually at −70°C.

For each construct, mass spectrometry analyses revealed quasi-site-specific labeling of the CP heterodimer, with donor fluorophore predominantly (or exclusively) attached to the β subunit and acceptor fluorophore predominantly or exclusively attached on the α subunit for both constructs (see Methods, Figure 1 and Supplementary Figure 2).

Labeling of CP Constructs for ITC.

We used ITC to test whether addition of the dyes altered binding of CPI-motif peptides to the CP FRET constructs. The sequential labeling strategy did not produce quantities of dual-labeled donor-acceptor material sufficient for multiple ITC experiments. As an alternative, we labeled the CP FRET constructs with acceptor dye in quantities sufficient for coupling at both Cys residues, and we tested those double-labeled CP preparations by ITC.

Cys-null CP with α1 N44C β2 E84C (pBJ 2488) and Cys-null CP with α1 S9C β2 S161C (pBJ 2478) were expressed and purified using the same methods described above. The CPs were diafiltered into 20 mM Bis-Tris-propane, 0.1 M NaCl, 1 mM NaN3 (pH 7.0), and immediately labeled with 6 mol Alexa Fluor 594 C₅ maleimide (Invitrogen) per mol CP. Alexa 594-labeled CPs were isolated on Mono-Q HR 5/5 (Cytiva) in 20 mM Bis-Tris-propane, 1 mM TCEP, 1 mM NaN3 (pH 7.0), and eluted with a NaCl gradient. Fractions containing Alexa 594-labeled CPs were dialyzed into 20 mM 3-(N-morpholino)propanesulfonic acid (MOPS), 100 mM KCl, 1 mM TCEP, 1 mM NaN3 (pH 7.2) and stored at −70 °C.

Experimental and Computational Methods

Mass Spectrometry Analysis of Labeled CP.

Labeled CP samples were prepared by exchanging into volatile salt solutions, with a final exchange into 2.5 mM ammonium acetate. Intact protein samples were analyzed on a Thermo Scientific Orbitrap Eclipse using Protein Mode. The solution was diluted into 50 / 50 acetonitrile / water with 0.1% formic acid for direct infusion via syringe pump into the mass spectrometer. Spectra were collected at 240K with 5 microscans and averaged over 100 scans prior to deconvolution using Thermo Scientific Xtract.

ITC Binding Assays.

ITC experiments with fluorescent labeled CP and with unlabeled Cys-null CP were performed as described ¹⁹. The concentrations of Alexa 594 labeled CPs were determined using UV-visible absorbance with Alexa 594 extinction coefficient ε₅₉₄ = 96,000 M⁻¹ cm⁻¹. The concentration of unlabeled Cys-null CP was determined using UV-visible absorbance with ε₂₈₀ = 77810 M⁻¹ cm^-1. Concentrations of V-1 and CPI-motif peptides were determined as described ¹⁹. ITC data were fit to a single-site binding model, constraining the stoichiometry to 1.0.

Single-molecule FRET experiments.

Single-molecule FRET measurements were performed on a modified Picoquant MT200 instrument (Picoquant, Berlin, Germany) as described ²⁶. Experiments were conducted at a protein concentration of 100-250 pM (estimated from dilutions of samples with known concentration based on absorbance measurements), 20 mM 3-(N-morpholino)propanesulfonic acid (MOPS), 100 mM KCl, 1 mM TCEP, 1 mM NaN3 (pH 7.2), 200 mM β-mercaptoethanol (for photoprotection), 0.001% Tween 20 (for surface passivation), at a room temperature of 295 ± 0.5 K. Pulsed interleaved excitation (PIE) enabled selection of bursts with 1:1 donor:acceptor stoichiometry.

Single-molecule FRET data were analyzed using the “Fretica” package developed by Daniel Nettels and Ben Schuler (University of Zurich, Zurich, CH) and available online at https://schuler.bioc.uzh.ch/wpcontent/uploads/2020/09/Fretica20200915.zip. Fluorescence lifetimes were obtained via a convolution with the Instrument Response Function (IRF), measured as described ²⁶. Time resolved anisotropies were computed as described ²⁶. Fits for titration curves and 95% confidence intervals were calculated using Mathematica (Princeton, NJ).

MD simulations and analysis.

For molecular dynamics simulations, we used crystal structures for free CP (PDB 1IZN) ²⁷ and the V-1-CP complex (PDB 3AAA) ²¹ as starting points, rebuilding any missing residues using Modeller ²⁸. There is no co-complex structure for WASHCAP and CP, but CD2AP has good sequence alignment with WASHCAP. We used the CD2AP complex PDB 3LK4 ²⁵ to build the N-terminal portion of WASHCAP (starting at V992) and PDB 3AA6 ²¹ to build the C-terminal part (up to A1024). The two missing residues at either end of the WASHCAP CPI were added using Modeller ²⁸. The C-terminal tentacle of CPβ is dynamic and has a wide range of motion. In control simulations with and without the β-tentacle, we found no significant difference in the dynamics of the rest of CP and therefore truncated CPβ at R530 so that the simulation box would be much smaller, and simulations would run faster.

Each system (free CP, V-1-CP and WASHCAP-CP) was solvated using TIP3P water with 10 Å padding. Na⁺ and Cl⁻ were added to neutralize the system and give an ionic strength of 50 mM. NAMD 3.0 ²⁹ was used to perform the simulations using the CHARMM36 forcefield ³⁰. Following minimization, the systems were heated and equilibrated at a temperature of 300 °K and 1 atm of pressure (NpT conditions). Simulations were performed with 2 fs timesteps by fixing hydrogens. We employed Particle Mesh Ewald methodology ³¹ for long-range electrostatics, and we used a 10 Å cut-off and 8.5 Å switch distance for van der Waals interactions. Following 0.5 μs equilibration of each system, 3-5 independent replicates of each system were run for at least 1.5 μs each, resulting in more that 10 μs of total simulation data. Analysis was performed using Bio3D ^{32, 33} in R.

Calculation of Transfer Efficiencies from MD Simulations.

Using the full MD trajectories, we adopted the geometric accessible volume (AV) method to simulate and predict FRET efficiencies utilizing the Python module FRETraj ³⁴. For every conformation of the trajectory, AV projects a spheroid onto the donor and acceptor locations that represents all sterically allowed dye positions ³⁵. The spheroid is calculated using five dye parameters: three radii of the dye as well as the width and length of the linker.

Results

To monitor in-solution conformational changes via single-molecule FRET, full-length CP constructs were designed to interrogate structural changes upon ligand binding (see Supplementary Text, Design of Labeling Positions on CP). Two sets of labeling positions were identified; they sample different regions of the protein and provide complementary information. CP α1N44C β2E84C (henceforth “α44β84”) and CP α1S9C β2S161C (henceforth “α9β161”) (Figure 1) probe the conformations between the stalk and the lower cap region and between the stalk and the higher cap region, respectively. Cysteine residues were inserted in a Cysteine-less CP variant that maintains the biochemical activity of wild-type CP (see Supplementary Figure 1 and Supplementary Table 1).

The effect of inserting Cysteine residues at specific probe positions in CP, coupled with the addition of fluorescent dyes, was measured based on the binding of CPI-motif peptides and V-1 (see Methods). Dissociation binding constants (K_D) obtained from ITC measurements with double-acceptor labeled CP are listed in Table 1. The K_D values for CARMIL1 peptide were similar to ones obtained from ITC measurements with wild-type CP and Cys-null CP (¹⁹ and Supplementary Table 1), and the K_D values for V-1 were similar to values for wild-type CP (¹⁹ and Supplementary Table 1).

Table 1. K_D values for ligand binding to dye-labeled CP constructs measured by single-molecule FRET and ITC.

For single-molecule FRET, CP α9β161 and CP α44β84 were labeled with donor and acceptor dyes. K_D values are calculated from the change in transfer efficiency as a function of the specified ligand. For ITC, CP α9β161 and CP α44β84 were labeled with acceptor dye at both donor and acceptor positions. K_D values are from ITC titrations with specified ligand.

Ligand	CP Constructs	Single-molecule FRET		ITC
		KD (nM) Individual titrations	KD (nM) (mean ± s.d.)	KD (nM) Individual titrations
	α9β161	13 ± 4		14 ± 1
		21 ± 7	17 ± 4	24 ± 4
CARMIL1		17 ± 6
	α44β84			65 ± 9
		No change	n/a	66 ± 10
	α9β161	7.4 ± 1.4		21 ± 3
		4 ± 2	7 ± 3	26 ± 3
CKIP		9 ± 3
	α44β84	10 ± 2		29 ± 3
		17 ± 4	13 ± 4	38 ± 4
		11 ± 4
	α9β161	0.37 ± 0.04		5 ± 1
		0.7 ± 0.2	1.4 ± 1.2	6 ± 1
		1.5 ± 0.3
WASHCAP		3.1 ± 0.6
	α44β84	19 ± 3		13 ± 3
		20 ± 4	22 ± 5	20 ± 3
		28 ± 3
		8 ± 3		52 ± 9
	α9β161	8 ± 3	15 ± 12	53 ± 8
V-1		28 ± 9		67 ± 13
		11 ± 4		18 ± 4
	α44β84	22 ± 10	14 ± 7	20 ± 5
		9 ± 4

Conformational Changes Between Stalk and Lower Cap Region

The α44β84 construct probes the configuration between the stalk and the underside of the cap region of CP (Figure 1). In absence of ligand (Apo-CP), the histogram of FRET efficiencies reveals one population with mean transfer efficiency (Ē) of 0.458 ± 0.007 (Figure 2 A). The plot of donor lifetime vs Ē suggests that the transfer efficiency population reflects a dynamic ensemble of the protein and not simply a single rigid distance (Supplementary Figure 3). After accounting for the contribution of the dye linker, we estimated the distance distribution sampled by the protein (Supplementary Figure 4), which can be compared to the value calculated from the Apo crystal structure (PDB: 1IZN). The single-molecule measurements report on a dynamic ensemble with a root mean square distance of approximately 55 Å (Supplementary Figure 3-5). The crystal structure C_α – C_α distance is 45 Å, which is within the distribution sampled by the observed conformational ensemble.

Figure 2. Single-molecule FRET analysis of CP α44β84 binding to WASHCAP and V-1. A.

Transfer efficiency histograms for free CP (0.07 nM Apo-CP, 0 nM ligand, green) and CP titrated with WASHCAP CPI-motif peptide (blue) or V-1 (red). The concentration of CP is low to ensure measurement of single molecules in the confocal volume. WASHCAP induces a shift to a higher Ē value, and V-1 causes a shift to a lower Ē value. Measurements were performed in triplicate; representative histograms are shown. B. Titration curves for WASHCAP (blue) or V-1 (red) plot the mean normalized change in Ē (Ē₀ - Ē^[WASH/V-1]) vs increasing concentrations of WASHCAP or V-1. Plotted data points are mean values, and error bars are standard deviation. Gray shading represents the 95% confidence interval of the fit as determined by Mathematica.

CPI motifs bind directly to the stalk of CP (Figure 1). To understand whether CPI-motif binding affects the conformations of the stalk region, single-molecule FRET measurements were performed in the presence of the CPI-motif peptide from WASHCAP, which is a potent regulator of CP ¹⁹. With increasing concentration of ligand, the distribution of transfer efficiencies shifted to higher values and reached saturation at a value above 1 μM (Figure 2 B). The change in Ē was sufficiently small (ΔE^[WASH] = +0.071 ± 0.001, Supplementary Table 2) that one cannot resolve two distinct peaks for the Apo-CP and ligand-bound states. The titration curve of Ē vs ligand concentration is well fit assuming a 1:1 binding model, which results in a K_D of 22 ± 5 nM (Table 1 and Figure 2 B), similar to the values from corresponding ITC measurements for CP labeled at both Cys residues with the acceptor dye (Table 1 and Supplementary Figure 6).

A structural interpretation of the shift in transfer efficiency as a function of distance requires one to rule out contributions from quenching or rotational hindrance of the fluorophores. To this end, the fluorescence lifetimes were measured for the donor-only and corresponding anisotropy decays in Apo-CP with and without ligand (Supplementary Tables 3-4). No significant changes were apparent, excluding quenching and rotational effects Therefore, the transfer efficiency and the fluorescence lifetime of the donor in presence of acceptor are most consistent with a small decrease in the mean distance between the positions probed by the dyes (~ 3 Å) as well as a reduction in the width of the corresponding distance distribution (Supplementary Figure 3-5). This distribution encompasses distances calculated from crystal structures of other CPI-motif peptides in co-complex with CP, suggesting that our observed conformational changes are compatible with previously determined structures.

In biochemical assays, CPI-motif peptides antagonize the binding of V-1 to CP, and V-1 antagonizes CPI-motif binding to CP ^{18, 19}. To investigate this mechanism, we measured how binding of the V-1 ligand on the top surface of the CP cap affects the conformation of the stalk and the underside of the cap, where CPI-motif peptides bind (Figure 1). With increasing concentrations of V-1 (Figure 2 A and B), the mean transfer efficiency Ē monotonically shifted to lower values. Most significantly, the change in Ē for V-1 moved in the opposite direction (more expanded with respect to Apo-CP) compared to the change in Ē produced by the WASHCAP CPI-motif peptide (more collapsed with respect to Apo-CP) (Supplementary Table 2, Figure 2). The magnitude of the change in Ē from the Apo state to saturation was small, ΔE^[V-1] = −0.017 ± 0.004. The fact that V-1 induced a decrease in Ē is consistent with the difference in distance between residues α44 and β84 in corresponding crystal structures ^{21, 25, 27}. No significant alterations are introduced in the width of the distance distribution (Supplementary Figures 3-5), when compared to the Apo configuration. Fit of the change in transfer efficiency yields a K_D of 14 ± 7 nM, assuming a 1:1 stoichiometry, in agreement with K_D values obtained by ITC (Table 1 and Figure 2 B) and with previously published values for binding affinity and stoichiometry ^{19, 20}.

Conformational Changes Between Stalk and Upper Cap Region

The CP α9β161 construct probes the configuration between the stalk and the upper cap region of CP (Figure 1), which is the region where CP interacts with F-actin barbed-ends and V-1. In contrast to α44β84, the distribution of transfer efficiencies of Apo-CP α9β161 contained two populations (Ē₁^Apo = 0.18 ± 0.02 and Ē₂^Apo = 0.37 ± 0.03, Figure 3 A, top histograms, and Supplementary Table 2), suggesting two conformational states of the protein. This result was not predicted by previous structural studies of CP, including crystal structures. Analysis of fluorescence anisotropies (Supplementary Table 4) indicated that the two populations do not arise because of slow rotation of the fluorophores; instead, they reflect two different slow-exchange (>1 ms) configurations of CP.

Figure 3. Conformational changes observed for CP α9β161 upon binding of WASHCAP CPI-motif peptide or V-1. A.

Two conformations are observed for Apo-CP, indicated as E₁^Apo and E₂^Apo (0.25 nM Apo-CP, 0 nM ligand, green). For WASHCAP (blue), the titration histograms show a complete shift to a higher mean transfer efficiency for both populations upon saturation. Titration histograms for V-1 (red) show a shift to a lower mean transfer efficiency for both E₁ and E₂ populations upon binding; however, a portion of the low-efficiency population remains, in contrast to the results for WASHCAP. Measurements were performed in triplicate. Representative histograms are shown. B. Titration curves for WASHCAP (blue) or V-1 (red) plot the mean normalized change in Ē₁ (Ē₁^[WASH]- Ē₁^Apo and Ē₁^[V-1]- Ē₁^Apo_, open circles) or Ē₂ (Ē₂^[WASH]- Ē₂^Apo and Ē₂^[V-1]- Ē₂^Apo, closed circles) vs increasing concentrations of WASHCAP or V-1. Each data point is the mean of three values. Error bars indicate standard deviation. The gray shading represents the 95% confidence interval of the fit. Data for Ē₁^[WASH]- Ē₁^Apo do not display a significant difference with addition of WASHCAP; therefore, a fit was not performed.

Upon titration with WASHCAP peptide, the Ē₁^Apo and Ē₂^Apo populations decreased in favor of a new third population with change in mean transfer efficiency Ē₂^[WASH] - Ē₂^Apo = +0.112 ± 0.006 (Figure 3, panel A and B, Supplementary Table 2, and Supplementary Figure 7). At saturation with ligand, only the Ē₂^[WASH] population was observed. The transfer efficiency change of the main population can be fitted to a 1:1 binding model, which reports a K_D of 1.4 ± 1.2 nM (Table 1 and Figure 3 B). To better account for the multiple states identified, the relative fraction of each population (associated with the mean transfer efficiencies Ē₁^Apo, Ē₂^Apo, and Ē^[WASH]) was estimated by computing the corresponding areas under the curve at each ligand concentration (Supplementary Figure 8 A and B, WASHCAP). A global fit of the three populations was performed, assuming a 3-state model (Apo-CP 1, Apo-CP 2, Bound) (Supplementary Text and Supplementary Figure 9-10). From the model, an effective K_D* was computed, representing the dissociation constant between the Apo and bound configurations, of 2 ± 2 nM; this value is similar to values obtained by ITC for labeled molecules (Supplementary Table 5). However, a close inspection of histograms reveals a change in the equilibrium between the two Apo configurations upon addition of ligands, indicating that a 4-state model (Apo-CP 1, Apo-CP 2, Bound 1, Bound 2) may be more appropriate (Supplementary Figure 9-10). It is important to note that connectivity between different states cannot be ruled out in equilibrium experiments; therefore, the most generic case was considered, in which both Apo states can bind the ligand and the bound states can interconvert one into another. Corresponding analysis revealed two dissociation constants K_D1 and K_D2 (between the corresponding Apo and Bound states, Supplementary Table 6) equal to 20 ± 20 and 1 ± 1 nM, respectively. The one order of magnitude difference in the K_D1 value explains why the data were adequately described with a 3-state model, since the contribution from the Bound 1 state is negligible.

Next, CP α9β161 was titrated with increasing concentrations of V-1. The E₁^Apo and E₂^Apo peaks shifted toward lower mean transfer efficiency values, with ΔĒ₁^[V-1] = −0.056 ± 0.012 and ΔĒ₂^[V-1] = −0.033 ± 0.003 (Figure 3, panel A and B, and Supplementary Table 2, and Supplementary Figure 11). These changes were in the same direction as the changes observed for the CP construct α44β84 (Supplementary Figure 5). Fitting the change in transfer efficiency as a function of the V-1 concentration resulted in a K_D1 of 50 ± 20 nM for the Apo-CP 1 state (representing the transition from unbound to Bound 1) and a K_D2 of 12 ± 3 nM for the Apo-CP 2 state (representing the transition from unbound to Bound 2) (Figure 3). To better assess the binding affinity in the model, the distribution of transfer efficiencies was analyzed globally assuming a 4-state model (Apo-CP 1, Apo-CP 2, Bound 1, Bound 2). The global fit of the titration provided a K_D1 of 25 ± 2 nM for V-1 binding to Apo-CP 1 and K_D2 of 19 ± 2 nM for V-1 binding to Apo-CP 2 (Supplementary Figure 12 and Supplementary Table 6). Reducing the model to represent only three states did not reproduce the experimentally determined distribution of transfer efficiencies at high (μM) ligand concentration (Supplementary Figure 12-13).

Therefore, the experimental results were consistent with the coexistence of four states across the titration. The similarity between the two determined dissociation constants, K_D1 and K_D2, is consistent with the coexistence of the two bound populations at saturating concentration, and the overall values are similar to the corresponding ITC results (Supplementary Table 5).

Comparison of lifetime vs transfer efficiency for Apo-CP and V-1 bound states suggests that the binding of V-1 does not significantly alter the distribution of conformations, rather impacts the relative abundance of the two configurations (Supplementary Figures 3-5).

CPI sequence-specific effects on CP conformations

The amino-acid sequences of CPI-motif regions from different CPI-motif proteins are conserved across evolution, and CPI-motif peptides from those proteins have distinct biochemical effects on CP ¹⁹, raising the question of whether these biochemical differences stem from disparate conformational changes upon binding. To this end, the effect of interaction of WASHCAP on CP conformations was compared with the effect of two different CPI-motif peptides: CKIP and CARMIL1 (Figure 4 A).

Figure 4. Conformational changes observed for CP α44β84 upon CPI-motif peptide binding.

A. Sequences of the CPI-motif peptides and the consensus sequence for the CPI region. Asterisks denote key residues necessary for the biochemical activity of the CPI-motif peptides on CP. B. Transfer efficiency distributions and titration curves for WASHCAP (blue), CKIP (cyan) and CARMIL1 (orange). Measurements were performed in triplicate, and representative histograms are shown. Free CP transfer efficiency distributions are shown in green (0.07 nM Apo-CP, 0 nM ligand). C. Titration curves plot the change in mean transfer efficiency (ΔĒ) with increasing concentrations of WASHCAP (blue), CKIP (cyan) and CARMIL1 (orange). Data points are mean and standard deviation from the replicates. Color shading indicates the 95% confidence interval of the fit. Data of CARMIL1 are not fitted because a binding curve is not apparent; in this case, the line denotes the average of the values, and the shading is standard deviation.

As discussed above, for CP α44β84, binding of WASHCAP produced a positive increment of ΔĒ^[WASH] = 0.071 ± 0.001 (Figure 4 B and C and Supplementary Table 2). The interaction with CKIP resulted in a smaller increase, +0.031 ± 0.002, whereas the binding of CARMIL1 produced no significant shift in Ē (Figure 4 B and C and Supplementary Table 2). To exclude the possibility that the absence of changes for CARMIL1 is due to a lack of interaction of the peptide with the labeled CP, the binding was verified by ITC (Table 1). The labeled CP gave K_D values similar to published results with wild-type CP ²³ (Supplementary Table 1). To confirm that these results reflect conformational variations, fluorescence lifetimes and time-resolved anisotropy were measured for donor-only and acceptor-only populations of the Apo-CP and the CPI-motif-saturated CP samples. No significant differences in lifetimes were observed, indicating the absence of dye quenching and a lack of deviations from the assumption of rapid rotation of the fluorophores (Supplementary Table 3). Analysis of donor dye lifetime in presence of the acceptor (at saturating conditions of ligand) suggests that binding of CARMIL1 does restrict the distance distribution explored by CP while maintaining an average distance similar to that seen for Apo-CP (Supplementary Figures 3-5). In contrast, the changes in Ē with WASHCAP and CKIP represent structural variations characterized by larger mean squared inter-dye distances and equal or broader distributions of distances compared to the Apo-CP (Supplementary Figures 4-5).

Next, the effects of the three CPI-motif peptides on CP α9β161 were measured and compared. Similar to the case for CP α44β84, CKIP produced a shift in transfer efficiency, ΔĒ₂^[CKIP] = +0.068 ± 0.008 (Supplementary Table 2) smaller than that of WASHCAP, ΔĒ₂^[WASH] = +0.112 ± 0.006 (Figure 5 A and B and Supplementary Table 2). Different from the case of CP α44β84, a measurable change in the mean transfer efficiency was observed for CARMIL1, ΔĒ₂^[CARMIL1] = +0.042 ± 0.002 (Supplementary Table 2). Corresponding values for K_D are reported in Table 1, and data are shown in Supplementary Figures 7 and 8. Analysis of the lifetime versus transfer efficiency for CP α9β161 revealed that interaction with both CARMIL1 and WASHCAP decreased the distribution of distances sampled by CP, whereas interaction with CKIP results in a similar distribution to Apo-CP (Supplementary Figures 3-5).

Figure 5. Conformational changes observed for CP α9β161 upon CPI binding.

A. Transfer efficiency histograms for WASHCAP (blue), CKIP (cyan) and CARMIL1 (orange) CPI-motif peptides. Measurements were performed in triplicate, and representative histograms are shown. Free CP transfer efficiency distributions are shown in green (0.25 nM Apo-CP, 0 nM ligand). B. Titration curves plot the mean normalized change in transfer efficiency Ē₁ (Ē₁^[WASH]- Ē₁^Apo, Ē₁^[CKIP]- Ē₁^Apo and Ē₁^[CARMIL1]- Ē₁^Apo_, open circles) or Ē₂ (Ē₂^[WASH]- Ē₂^Apo, Ē₂^[CKIP]- Ē₂^Apo and Ē₂^[CARMIL1]- Ē₂^Apo, closed circles) vs increasing concentrations of WASHCAP (blue), CKIP (cyan) and CARMIL1 (orange). Titrations were performed in triplicate in separate experiments and plotted as the mean and standard deviation. Shading represents the 95% confidence interval of the fit.

One important difference among the three CPI-motif peptides is that they produce quantitatively different changes in mean transfer efficiency and lifetime for both CP constructs, α44β84 and α9β161, indicating that the extent to which the CP conformations are modulated upon interaction with the CPI-peptide depends on the nature of the CPI-peptide ligand itself. Comparing the distance distributions determined from lifetime analysis of the CARMIL1 – CP interactions for both probe sets indicates that the CPI-motif of CARMIL1 has the strongest effect on restricting CP conformations, whereas CKIP has the least effect, allowing for a broader dynamic ensemble of CP (Supplementary Figures 3-5).

Atomistic models of conformational ensembles from MD

The single-molecule FRET results described herein reveal two distinct conformational states of Apo-CP, which have not been observed previously. The results also reveal small but significant structural changes in response to the binding of different regulatory ligands. Single-molecule FRET experiments provide a simple coarse-grained perspective (limited to the measured distances of labeled pair residues) of the conformational changes in CP upon binding. Therefore, to investigate and understand the possible nature of conformational changes throughout the entirety of the CP molecule, a molecular dynamics approach was employed to construct atomistic models of the conformational states in the Apo and ligand-bound conditions. Simulations were performed for each starting configuration based on the protein crystal structures in the absence of ligand, with WASHCAP CPI, and with V-1 (see Methods). Multiple independent simulations were performed in each case, with over 10 μs of aggregate simulation time.

For Apo-CP, MD simulations showed two stable conformations. This result supports the experimental observation of two states for CP α9β161 in the single-molecule FRET measurements. Movies showing interpolations between the two equilibrium states revealed the stalk region of CP undergoing twisting and tilting (Figure 6 A, Movie 1 and Supplementary PDB files). One state, termed Apo-CP 2, was occupied more often than the other, termed Apo-CP 1 (illustrated in Figure 6 A and Supplementary Figure 14). This difference is qualitatively consistent with the relative abundance of the two populations observed in the single-molecule FRET experiments.

Figure 6. Molecular dynamics simulation results: Equilibrium conformations for the two states of unbound CP (Apo-CP 1 and Apo-CP 2) and for the CP-V-1 and CP-WASHCAP complexes.

A. The Apo-CP 1 and Apo-CP 2 states primarily differ in the tilt and twist of their stalk regions, relative to the cap. The V-1 and WASHCAP structures have very similar stalk structures, in which the α helices are more aligned with each other than they are in the Apo-CP structures. The locations of the FRET probes are shown in cyan (α9 and β161) and orange (α44 and β84). B. MD simulations of C_α – C_α distance distributions for the attachment points of the fluorophores in the two constructs, comparing Apo-CP (green) with the bound states of V-1 (red) and WASHCAP CPI peptide (blue). Abscissa values are for α44β84, and ordinate values are for α9β161. The distance for the V-1-bound state is longer than the primary Apo-CP 2 state for both FRET pairs while the WASHCAP state is shorter for both pairs.

The binding of ligand, for both V-1 and WASHCAP CPI peptide, resulted in the α helices of the stalk region of CP becoming more aligned with each other, relative to the two Apo states. The effect on the mushroom cap was also similar for both ligands, resulting in a flatter cap as compared to the Apo states (illustrated in Figure 6 A, Movie 2, Movie 3 and Supplementary PDB files). For the dye positions in the cap region, there were relatively small changes between the Apo, V-1 and WASHCAP states (Figure 6 A). Conversely, the helices in the stalk region showed changes in both alignment and rotation of the helix bundle, and this led to larger changes in the dye positions, and hence, the distances between FRET pairs.

These results are consistent with the conclusions of the single-molecule FRET experiments, in which CPI-motif binding pulls the dyes closer to one another and V-1 binding pushes the dyes farther apart. In the simulations, only a single population was reported for both WASHCAP-bound and V-1 bound CP conformations (Figure 6 B and Supplementary Figure 14). This outcome is consistent with WASHCAP single-molecule measurements but is at variance for V-1, where two bound populations were experimentally observed at saturation of ligand (Figure 3). We interpret this discordance between experiments and simulations as an indication of an incomplete exploration of the protein energy landscape in the bound state (where low frequency states may be rarely explored) or a small variation in energetics (temperature).

To provide a more detailed comparison of simulations and experiments, the molecular dynamics simulation trajectories were used to calculate transfer efficiencies that would be expected for a confocal single-molecule FRET experiment. Using the conformations from the simulations, a dye cloud was generated for each fluorophore to illustrate the possible dye conformations (Supplementary Figure 15), using the Accessible Volume method described by Sindbert and colleagues ³⁵. These dye clouds, coupled with photon detection statistics (shot-noise), then allowed for calculation of a distribution of transfer efficiencies for each CP construct in the Apo conformation and when bound to V-1 and WASHCAP.

For CP α44β84, the two different conformational states of the protein resulted in two populations with mean transfer efficiencies of 0.407 and 0.567, respectively (Supplementary Figure 14). This result differs from the single-molecule FRET measurements of a single peak with mean transfer efficiency of 0.458 in the experiments (Figure 2). With WASHCAP CPI peptide bound, the simulated transfer efficiency shifted to a higher value (Supplementary Figure 14), which agrees qualitatively with the experimental measurements for single-molecule FRET (Figure 2). In addition, when V-1 is bound, the simulated transfer efficiency shifted to a lower value (Supplementary Figure 14), again consistent with the results of single-molecule FRET (Figure 2).

For the construct CP α9β161, the transfer efficiency distribution from the two Apo-CP states reflected the corresponding fractions observed in the simulations and partially overlapped (Supplementary Figure 14). This result agrees qualitatively with what was observed by single-molecule FRET (Figure 3), although predicted mean transfer efficiencies were lower than the detected ones. We speculate that the shift in measured transfer efficiency reflects changes in the Förster radius due to alterations in the κ² parameter (see Supplementary Text and Supplementary Table 7). One cannot, however, exclude that the two Apo-CP states observed in simulation would be detected as a single averaged mean transfer efficiency in the single-molecule FRET experiment if their exchange occurs on a timescale faster than a millisecond (approximately the diffusion time of the protein). A weighted average of the simulated mean transfer efficiencies of 0.14 and 0.22 was indeed consistent with the experimentally measured mean value of Ē₁^Apo = 0.18 ± 0.02, suggesting that the observation of Ē₂^Apo = ~ 0.37 in the simulations would require a further exploration of the protein energy landscape. In the following, the interpretation of simulations and their comparison with experiments was restricted to the first scenario, where discrepancies are attributed to alterations in the κ² parameter.

With WASHCAP CPI-motif peptide bound, the simulated transfer efficiency shifted to a single population at a higher transfer efficiency (Figure 6 B and Supplementary Figure 14). This result corresponds to what was measured by single-molecule FRET upon WASHCAP binding (Figure 3). With V-1 bound, the simulated transfer efficiency shifted to a single population at a higher efficiency (Supplementary Figure 14, α9β161, lower panel). This result was not predicted from the C_α – C_α distance measurements provided by the molecular dynamics simulation (Figure 6 B). In addition, this result differed from the result for single-molecule FRET. The single-molecule FRET results showed a shift to a lower mean transfer efficiency for both Apo states, and both states were present after V-1 saturation (Figure 3).

Overall, the MD simulation results are in general agreement with the single-molecule FRET results and conclusions, within known limitations in the conversion of distances from simulated data to mean transfer efficiencies. The MD simulations provide a consistent atomistic description of the conformational changes that occur in CP following regulator ligand binding. This information provides new insight into possible details of the allosteric mechanism of CP regulators.

Dissecting the CPI-motif

The three CPI-motif peptides produced consistently different results for both CP constructs – α9β161 and α44β84, posing the question of which portions of the CPI-motif peptides are responsible for the differences. Examining the amino-acid sequences and performing whole-molecule alignment of crystal structures of the CP / CARMIL1 (PDB 3LK2) and CP / CKIP (PDB 3AA1) peptide complexes, one observes that the α-carbon backbones in the N-terminal halves of the CARMIL1 and CKIP peptides are in very good alignment, while the α-carbon backbones differ substantially in the C-terminal halves (Supplementary Figure 16 and Movie 4). This point suggests that the single-molecule FRET differences between the CARMIL1 and CKIP CPI-motif peptides might result from the differences in their C-terminal halves. To test this hypothesis, chimeras of CPI-motif peptides were created, combining the N-terminal and C-terminal halves of WASHCAP (which had the largest changes in mean transfer efficiencies) with those of CARMIL1 (which had the smallest changes in mean transfer efficiencies). The dividing line between the N-terminal half of one peptide and the C-terminal half of the other peptide was chosen as C-terminal to the central core of the most highly conserved amino-acid residues essential for all CPI-motif peptide binding to CP, namely KXRXK (Figure 7 A) ^{36, 37} . These two chimeras are referred to as WASH_N-C1_C (N-term WASHCAP with C-term CARMIL1) and C1_N-WASH_C (N-term CARMIL1 with C-term WASHCAP).

Figure 7. Dissection of CPI-motif peptide effects with chimeras.

A. Sequences of chimeric peptides, WASH_N-C1_C (Chimera 1) and C1_N-WASH_C (Chimera 2). CARMIL1 residues are orange, and WASHCAP residues are blue. B. Transfer efficiency histograms for CP α9β161 with WASHCAP (blue), Chimera 1: WASH_N -C1_C chimera (fuchsia), CARMIL1 (orange), and Chimera 2: C1_N - WASH_C chimera (purple). Experiments were performed in triplicate and representative histograms are shown for ligand saturation at 1 μM. Free CP transfer efficiency distributions are shown in green (0.25 nM Apo-CP, 0 nM ligand). C. Titration curves for all concentrations of peptide are shown with the average change in mean transfer efficiency relative to Apo-CP, ΔĒ, plotted versus concentration of peptide. The value for Ē₁^Apo did not change with chimeric peptide concentration, similar to results for WASHCAP and CARMIL1 peptides.

The chimeric peptides were titrated into CP α9β161 to saturation, and the corresponding transfer efficiency distributions were compared to each other and to those of the wild-type (non-chimeric) peptides. The Ē difference upon binding of WASH_N-C1_C was ΔĒ₂^[WN-C1C] = +0.047 ± 0.004 (Figure 7 C and Supplementary Table 2), which is equivalent to the change observed for CARMIL1 of +0.042 ± 0.002 (Supplementary Table 2). The change in Ē upon binding of C1_N-WASH_C was ΔĒ₂^[C1N-WC] = +0.114 ± 0.006 (Supplementary Table 2), which is within errors of the +0.112 ± 0.006 change observed for WASHCAP (Supplementary Table 2). For both CP constructs, the chimera results show that the effects on transfer efficiency correspond to the C-terminal portion of the peptides. K_D values for the chimeric peptides were determined from single-molecule FRET and ITC titrations with similar results (Supplementary Table 8 and Supplementary Figure 17). The WASH_N-C1_C affinity trended with CARMIL1 affinity, while C1_N-WASH_C affinity trended with that of WASHCAP, suggesting that the C-terminal portion of the peptides largely defines the interaction with CP.

Analogous experiments were performed with CP α44β84. The chimera WASH_N-C1_C produced very little change in Ē on binding (Supplementary Figure 18), similar to the result for non-chimeric CARMIL1. In contrast, the chimera C1_N-WASH_C produced a larger change in transfer efficiency, with ΔĒ^[C1N-WC] = +0.039 ± 0.001 (Supplementary Table 2). This change was significant, although not as large as that observed for the WASHCAP peptide. These results further support the hypothesis that the C-terminal portions of CPI-motif peptides are responsible for differences in Ē.

CP dynamics related to interaction with the chimeric peptides were analyzed further by comparing the lifetime versus transfer efficiency. In contrast to the effect of CARMIL1, which appeared to restrict conformations of the protein, both chimeric peptides had a minor effect on the range of distance distributions for both CP α44β84 and CP α9β161, relative to Apo-CP. Taken together, the results indicate that the C-terminal region of the chimeras determines the overall register of the distance distribution (mean value), but the whole CPI-motif is required to control both CP dynamics (width of the distribution) and affinities.

To extend the analysis, the experimental FRET results were compared with MD simulation results. MD simulations of the WASHCAP CPI-motif peptide bound to CP showed differences in the dynamics of residues along the length of the peptide. The N-terminal section was much more dynamic, with few specific and/or stable interactions with CP. The C-terminal section was more constrained and maintained more stable interactions with CP (Supplementary Figure 19). The MD results agree with the single-molecule FRET results, supporting the conclusion that the C-terminal portion of the peptide is responsible for the differences in conformational changes and affinities (Figure 7).

Discussion

We investigated the molecular mechanisms for allosteric regulation of CP, using the complementary approaches of single-molecule FRET experiments and MD simulations. Our study identified two distinct CP conformations in solution, which are further modulated by two classes of regulators: CPI-motif proteins and V-1. One important discovery is that the conformational changes induced by the two classes of regulators are in opposite directions. This result accounts for the antagonism of the two classes of regulators with respect to their binding to CP. Together, the results provide new evidence and insights into the antagonistic mechanism utilized by CP regulators to control its function.

CP Conformations in Solution

Time-dependent solvent exposure of Apo-CP was revealed by previous HDX-MS experiments, suggesting that CP may exist in a more dynamic configuration than what has been captured by crystallographic studies ²³. Here, we provide the first physical evidence for the existence of multiple CP conformations in solution. Both single-molecule FRET measurements and MD simulations agree on the presence of two distinct states (Apo-CP 1 and Apo-CP 2, Figure 8) characterized by a difference in the configuration of the stalk and the cap of the CP heterodimer with respect to each other. From the analysis of the distribution of transfer efficiencies and lifetimes, we conclude that the two states are in slow exchange (longer than milliseconds) with each state characterized by an ensemble of configurations that occur on the nanosecond to microsecond timescale.

Figure 8. Equilibrium scheme of CP states in Apo and bound states.

CP interconverts between two Apo unbound configurations, Apo-CP 1 (CP₁) and Apo-CP 2 (CP₂), which can both bind to either V-1 or CPI motif peptides. Equilibrium constants are shown between each state.

The slow timescale of exchange between the two major Apo configurations implies an energy barrier separating the two states, possibly due to restructuring of the protein conformations and/or disruption of interactions between two interfaces of the protein. From the MD simulations, we gain atomistic details for the corresponding structural features of the two states, representing a torsional change of the stalk propagating to the cap. This new conformational ensemble varies from the structures previously identified for the CP ligand-bound states, suggesting that indeed the Apo-CP conformations identified here are unique to CP in its ligand-free state. We further investigated key residue pairs within the heterodimer that are in close contact (< 0.6 nm) in one of the Apo configurations but undergo large conformational changes (> 1 nm) from one configuration to another. Excluding contributions from the disordered N- and C-terminal ends of the subunits, residues β41–42 (β-side of the stalk) and β63–64 (β-region of the cap bottom) emerged from the analysis as part of a possible interface of interest. It is possible that slow interchange dynamics reflects long standing interactions between these residues (and surrounding regions) of the heterodimer. Note that cysteine residues are present in nearby proximity of these regions, but are far apart in previously obtained crystal structures, suggesting that reported results are not specific to Cys-null CP but also apply to wild-type CP.

Effects of Regulators on CP Conformation

CPI-motif proteins and V-1 bind directly to CP at different sites ^{21, 22, 25, 38} . Their binding displays negative linkage ^{18, 19}; that is, the binding of one ligand decreases the affinity of the other ligand. This negative linkage is particularly important because the current model for CP function in cells, as proposed by Hammer and colleagues ¹⁸, is based on antagonism between the effects of CPI-motif proteins, which are found in small amounts and are restricted to distinct membrane locations ¹⁵, versus the effects of V-1, which is found in high concentration and diffuses freely in cytoplasm ¹⁷. A molecular mechanism of linkage is suggested by the observation of changes in solvent accessibility for CP at CPI-motif regulator binding site in HDX-MS analysis when bound by V-1 and vice versa ²³. Solvent accessibility reflects changes in conformation and/or structural dynamics but lacks specific information about the molecular nature of the differences. Here, regulators were found to introduce changes in CP conformations and dynamics in solution; most important, the conformational changes induced by interaction with the two classes of regulators occur in opposing directions (Supplementary Figure 5). In particular, the results demonstrate that CP, in solution, adopts distinct conformations when bound to either CPI-motifs (the stalk and β-subunit cap regions move closer together in distance) or when bound to V-1 (the stalk and the β-subunit cap regions move farther apart). Recent work analyzing available crystal structures for CPI-motif-bound CP and V-1-bound CP found similar results ³⁷.

The conformational states induced by the binding of V-1 and CPI-motif proteins do not correspond exactly to either of the two states observed for Apo-CP. The experimental results indicate that the binding of V-1 does not restrict CP from exploring the two states; however, the binding of CPI-motif peptides favors a single bound conformation (Figure 5). These findings support the hypothesis that the conformational changes in CP underlie linkage, and they account for antagonism in binding. In this model, the steric inhibition of actin capping produced by V-1 binding is antagonized by CPI-motif binding, and the net effect is activation of CP. Thus, diffuse inhibition leads to spatially limited activation. This model can account for regulation of actin polymerization at barbed ends, either by the promotion of Arp2/3-based polymerization or by the inhibition of formin-controlled polymerization.

Detailed observations from MD simulations

The stalk conformation for V-1 and WASHCAP structures are very similar (Figure 6) while the conformation of the binding site on CP is very different for these two states (Supplementary Figure 20). Since WASHCAP binds to the stalk, this suggests WASHCAP could bind to CP-V-1 and then release V-1 from the complex. Conversely, V-1 would not bind as strongly to the CP-WASHCAP complex since the V-1 binding site is in a very different conformation. There is also a significant difference in the V-1-binding site between the Apo-CP 1 and Apo-CP 2 structures (Supplementary Figure 20). The V-1 site on Apo-CP 1 diverges more from the V-1-CP structure, suggesting that V-1 would interact more weakly with Apo-CP 1 than with Apo-CP 2. These findings are consistent with experiments that showed that binding of CARMIL to CP alters the stability of the V-1 binding site as well as changing the accessibility to the binding site by a key tryptophan residue in V-1 ²⁴.

The MD simulations revealed substantial alignment of the α-helices of the stalk region along with flattening of the cap region when comparing Apo-CP with both bound states: CPI-motif-bound CP and V-1-bound CP (Figure 6, Movies 2 and 3). Flattening of the cap region has previously been observed in structures of CP bound to V-1 ²¹, to twinfilin-actin ³⁹, and to the barbed end of F-actin ^{12, 13}. Comparing the two ligand-bound structures from the MD simulations, we observed that the V-1-bound CP structure showed greater twisting of the stalk region of CP, compared to the WASHCAP CPI-motif-bound CP structure, along with slightly more flattening of the cap (Figure 6, Movie 5).

Overall, the biggest conformation differences associated with ligand binding are between the stalk region and the β-subunit cap region. When WASHCAP CPI-motif is bound, the stalk and cap regions move closer together; however, when V-1 is bound, the stalk and the cap regions move farther apart. These findings further demonstrate the linked conformational changes from the stalk through the cap regions, which account for the effects on regulator binding and actin filament capping.

Differences among CPI-motif peptides: contrasts with biochemical effects

A comparison of the biochemical effects with conformational changes imposed by CPI-motif peptides on CP shows interesting differences. One might expect that strong biochemical effects correlate with large conformational changes; however, CARMIL1 produced the smallest conformational change but has substantial biochemical effects in terms of binding affinity for CP and inhibition of V-1 binding ¹⁹. This apparent discrepancy seems to suggest a disconnect between the functional activity of CP and conformational changes induced upon interaction with CPI-motif proteins. However, the combination of lifetime information and transfer efficiencies indicates that CP adopts a dynamic conformational ensemble where regulators alter both the mean and the width of the distribution of distances. Based on the placement of the dye pairs, with one in the cap and one in the stalk, the mean of the distribution reports on the average distance between the two, while the width captures the associated dynamics. With respect to dynamics, the binding of CARMIL1 and WASHCAP restricts the range of CP conformations (CARMIL1>WASHCAP), while interaction with CKIP allows for a broader range compared to Apo-CP (Supplementary Figure 5). The fine-tuned biochemical effects imparted on CP by CPI-motif peptides may be described by the change in both the conformation and dynamics. As suggested by the simulations, altering the mean conformation of CP may result in a different binding surface along the cap, which has a lower affinity for V-1 or actin. At the same time, changes in dynamics may modulate affinity by enabling the exploration of more (or less) binding-competent CP conformations. Recent computational work reported similar findings by exploring distinct modulation of the flexibility of CP whether bound to CARMIL or twinfilin, both of which contain CPI motif ⁴⁰. One might speculate that changes in CP conformation and dynamics allow for fine-tuning of function as encoded by the specific sequence of the CPI motif peptides not only in terms of changes in affinity, but for adapting surfaces of interactions to different ligands.

This study compared different CPI motifs. CARMIL family proteins, but not other CPI-motif proteins, have a second CP binding domain on the C-terminal side of CPI motif termed the CSI (CARMIL specific interaction) domain ²⁵. CARMILs also have a third domain C-terminal to the CSI motif, termed “membrane-binding” (MB) ⁴¹ The CSI and MB domains play additional roles in modulating the affinity and function of CP ^{25, 38, 41, 42}; therefore, they may further restrict the dynamics and the conformation of CP.

CPI-motif sequence encodes for conformations, dynamics, and affinities

To better understand how the sequence of CPI-motifs encodes for the register, dynamics, and affinities of interaction, two chimera peptides were studied that encompass the N- and C-terminal halves of WASHCAP and CARMIL1. The C-terminal half was found to largely control the register (mean conformation) and affinity, while the N-terminal half further modulates dynamics (range of conformational changes) and affinity. In particular, the binding affinities for WASH_N-C1_C were higher than the affinity observed for C1_N-WASH_C. These observations, derived from the single molecule experiments, are consistent with MD simulations of the WASHCAP peptide where the N-terminal half of the sequence was largely flexible compared to the C-terminal half (Supplementary Figure 19). The simulation results provide a different perspective on CPI conformations in solution when compared to previous crystal structures of CP with CPI-motif peptide bound ²¹, where the N-terminal half of different peptides can be aligned very precisely (Supplementary Figure 16). We speculate this discrepancy may arise from the specific experimental conditions used to determine the protein structure.

Potential implications for other CP isoforms.

One possible limitation of these results is the inclusion of only a single non-muscle isoform for each CP subunit, α1 and β2. While the relevance of these findings to CP composed of other isoforms remains to be established, some expectations regarding the β1 (CapZ) and β2 isoforms can be extrapolated from the comparison of current experiments and simulations.

Indeed, the β1 and β2 polypeptides differ only in their extreme C-terminal region ⁶, as they are produced by alternative splicing from one gene. In both cases, the C-terminal regions form amphipathic helices, and the hydrophobic side of that helix lies in a hydrophobic groove of an actin subunit at the barbed end of the filament ^{12, 13}. The actin-binding abilities of β1 and β2 are similar ¹⁴

Since the extreme C-terminal region of β2 (β-tentacle) has been omitted in the simulations but not in the experiments, the reasonable agreement between experiments and simulations suggests that the existence of two distinct configurations in Apo-CP and the directional effect of ligands should be identical for both β1 and β2. We cannot exclude that differences in the β-tentacle and mutations distinguishing α-subunit isoforms will further modulate conformations and dynamics of CP as well as affinities for the regulators.

Conclusions

The single-molecule experiments and MD simulations here revealed the coexistence of two distinct conformational states for CP in absence of ligands, characterized by a torsion of the protein stalk and a change in the configuration of the cap. Interaction with ligands alters both the equilibrium between the two configurations of the protein as well as its dynamics. The binding of CPI-motif peptides and V-1 were found to act in “opposite” directions, with V-1 holding CP in an intermediate configuration between the two Apo states and CPI-motif peptides altering the cap curvature and positioning. These observations together provide a plausible explanation of the different biochemical effects of these ligands on CP function. In addition, the C-terminal region of CPI-motif peptides was found to dictate the register of binding to the stalk of CP, while the N-terminal region further refines dynamics and binding affinity. Altogether, these observations provide a quantitative description of CP dynamics in absence and presence of ligands, and they enable bridging and reconciling observations from previous structural studies of the protein.

Abbreviations

CP

actin capping protein

Apo-CP

free unbound CP

F-actin

filamentous actin

CPI

capping protein interacting

HDX-MS

hydrogen-deuterium exchange mass spectrometry

FRET

Förster resonance energy transfer

MD

molecular dynamics

ITC

isothermal calorimetry

CARMIL

capping protein, Arp2/3, and myosin I linker