Dynamics of Dystrophin’s Actin-Binding Domain

Abstract

We have used pulsed electron paramagnetic resonance, calorimetry, and molecular dynamics simulations to examine the structural mechanism of binding for dystrophin’s N-terminal actin-binding domain (ABD1) and compare it to utrophin’s ABD1. Like other members of the spectrin superfamily, dystrophin’s ABD1 consists of two calponin-homology (CH) domains, CH1 and CH2. Several mutations within dystrophin’s ABD1 are associated with the development of severe degenerative muscle disorders Duchenne and Becker muscular dystrophies, highlighting the importance of understanding its structural biology. To investigate structural changes within dystrophin ABD1 upon binding to actin, we labeled the protein with spin probes and measured changes in inter-CH domain distance using double-electron electron resonance. Previous studies on the homologous protein utrophin showed that actin binding induces a complete structural opening of the CH domains, resulting in a highly ordered ABD1-actin complex. In this study, double-electron electron resonance shows that dystrophin ABD1 also undergoes a conformational opening upon binding F-actin, but this change is less complete and significantly more structurally disordered than observed for utrophin. Using molecular dynamics simulations, we identified a hinge in the linker region between the two CH domains that grants conformational flexibility to ABD1. The conformational dynamics of both dystrophin’s and utrophin’s ABD1 showed that compact conformations driven by hydrophobic interactions are preferred and that extended conformations are energetically accessible through a flat free-energy surface. Considering that the binding free energy of ABD1 to actin is on the order of 6–7 kcal/mole, our data are compatible with a mechanism in which binding to actin is largely dictated by specific interactions with CH1, but fine tuning of the binding affinity is achieved by the overlap between conformational ensembles of ABD1 free and bound to actin.

Introduction

Dystrophin (Dys) is a large muscle cytoskeletal protein of 427 kDa. Structurally, Dys consists of an N-terminal actin-binding domain (ABD1), 24 spectrin-like repeats that form the rod domain and house the second actin-binding domain (ABD2), and a C-terminal region containing a cysteine-rich domain that binds the dystroglycan complex (1). The N- and C-terminal ends of the protein provide a biochemical picture of how Dys can link the costameric F-actin network to both the sarcolemmal membrane and extracellular matrix via the dystroglycan complex. This location in the cytoskeletal network of muscle is crucial for maintaining membrane and costamere integrity, a fact that is best appreciated in the context of Duchenne and Becker muscular dystrophies (DMD and BMD, respectively) (2). In DMD and BMD, a multitude of mutations (missense and nonsense alike) in Dys lead to the development of muscle tissue degeneration. Patients with these forms of muscular dystrophy have elevated cytosolic calcium levels, indicative of an extracellular leakage of the ion through the sarcolemma due to focal membrane tears. Similarly, Dys-deficient model organisms, particularly mdx mice, show the same elevated cytosolic calcium phenotype (3).

The biophysical mechanism by which Dys protects the sarcolemma from mechanical stress is still an active research question. Dys contains several spectrin-like repeat domains (triple helix bundles) between the N-terminal ABD1 and C-terminal ZZ domain that anchor it to the dystroglycan complex. Atomic force microscopy has been used to unfold these types of protein domains from spectrin, showing that they have a low energy barrier to unfolding, which probably serves as a mechanical means to dissipate transduced force propagating through the costamere (4). However, independent of such a mechanism is the impact Dys has on the actin cytoskeleton alone. Previously, it was shown that the large-scale bending and twisting motions of F-actin are greatly restricted when both Dys and its homologous partner utrophin (Utr) bind (5, 6). Intriguingly, the amplitude of these large-scale motions is reduced, but the rate of motion is increased, indicating that both Dys and Utr binding impart resilience to the actin cytoskeleton. Such an effect could be another means by which the two proteins dampen the laterally transduced force of contraction within the costamere.

Although both Dys and Utr were found to restrict the amplitude of motion, the proteins were found to do so differentially, with Utr being the more restrictive of the two proteins (5, 6). Although it is unclear how such functional differences arise, a possibility is that they originate from differences in the structural binding modes of each protein’s actin-binding domains. Previously, the Utr ABD1 was shown to undergo a conformational opening upon association with F-actin; Utr ABD1’s adjacent calponin-homology (CH) domains transitioned from a more closed compact state to an open extended state, forming a well-ordered and stable complex with F-actin (7). We propose here that the Dys ABD1 can also undergo a similar structural transition (Fig. 1) but with distinct nuances that could contribute to a less stable ABD1-F-actin complex and, in turn, a smaller restrictive contribution to overall filament motions. Indeed, spectroscopic evidence in support of Dys ABD1 having a distinct binding mode has been reported (8).

Figure 1.

Proposed structural model for Dys ABD1 upon binding F-actin. In the absence of actin, the two adjacent CH domains are closely packed (closed state, blue). In the presence of actin, CH1 and CH2 can become more separated (open state, red) and sample multiple structural states. To see this figure in color, go online.

Several ABDs from proteins in the spectrin superfamily have been characterized in both open (9, 10) and closed (11, 12, 13) conformations using x-ray crystallography. Solution studies have converged on a similar finding; analytical ultracentrifugation (12) and pyrene excimer fluorescence (8) studies have identified ABDs in a closed conformation, whereas pulsed electron paramagnetic resonance (EPR) experiments show an equilibrium between closed and more open conformations, specifically for utrophin’s ABD. This open/closed dichotomy persists when ABDs are bound to actin. Both mutation studies (14) and cosedimentation assays (15) suggest a predominant role played by CH1 to binding affinity, with CH2 adopting a regulatory role. Excimer fluorescence suggests that Dys ABD1 binds to actin in a closed conformation (8), but double-electron electron resonance (DEER) shows clearly that Utr ABD1 binds to actin in an open conformation (7), and cryo-electron microscopy (cryo-EM) models suggest an open conformation for actin-bound α-actinin and β-III-spectrin ABDs (16, 17). These observations suggest that ABDs are malleable domains whose conformations are susceptible to actin binding and, potentially, even to the small perturbations introduced by different experimental techniques. It is thus appealing to hypothesize that the observed differences in Dys and Utr ABD1 interactions with actin stem from their distinct structural dynamics.

To test this hypothesis in our study, we have used a combination of DEER experiments and molecular dynamics (MD) simulations to examine the conformational dynamics of Dys ABD1. By applying these methods in parallel, we were able to define a mechanism of Dys ABD1 binding to actin that is distinct from that of Utr ABD1, providing insight into the different physical properties of these two proteins that may contribute to physiological function.

Materials and Methods

Protein mutagenesis, expression, purification, and labeling

A plasmid encoding the mouse Dys ABD1 protein fragment (residues 8–246) was subjected to mutagenesis using a Q5 Site-Directed Mutagenesis Kit (New England BioLabs, Ipswich, MA) to remove the protein’s native cysteines and incorporate new cysteines at positions 120 and 239. These positions were chosen to mimic our previous study on Utr ABD1, and also because they were suitable for measuring the proposed distance changes. The mutated plasmid was then transformed into BL21 DE3 Escherichia coli and subsequently grown to an optical density of 0.7, at which point protein expression was induced with 2 mM isopropyl β-D-1-thiogalactopyranoside per liter of growth. Use of mouse Dys ABD1 was chosen so that we could compare spectroscopic measurements to those of Utr ABD1 published previously, which were also from mice (7). Expression was allowed to occur overnight at room temperature. After expression, cells were pelleted at 10,000 rotations per minute (rpm). Cells were subsequently lysed in 50 mM Tris (pH 8), 20% w/v sucrose, 1 mM ethylenediaminetetraacetic acid, and 1 mM dithiothreitol (DTT) by incubation with lysozyme for 1 hr at 4°C, followed by freeze-thaw cycling in a dry-ice isopropanol bath. The resulting lysate was then incubated with DNase I for 1 hr at 4°C. All cell debris was removed from solubilized protein by centrifugation at 18,500 rpm in a SS-34 Sorvall rotor (Thermo Fisher Scientific, Waltham, MA) and was subsequently sterile filtered through a 0.2 μm filter to ensure complete removal of particulates. The lysate was then loaded onto a GE 5 mL SP ion-exchange column (GE Healthcare Bio-sciences, Pittsburgh, PA) and eluted over a linear gradient (0–500 mM) of NaCl buffered by 10 mM Tris (pH 7.5), 1 mM EGTA, and 1 mM DTT. Elution fractions containing the ABD1 fragment were then concentrated and further purified using an S100 sepharose gel-filtration column. Protein fractions pooled after size exclusion were then dialyzed in buffer containing 10 mM Tris, 100 mM NaCl, 2 mM MgCl₂, and 1 mM DTT (pH 8), buffer conditions under which actin bundling does not occur (18). Final protein purity was verified by SDS-PAGE.

For DEER experiments, a fivefold excess (to cysteine residue concentration) of 4-maleimido-TEMPO (4-maleimido-2,2,6,6-tetramethyl-1-piperinyloxy; Sigma-Aldrich, St. Louis, MO) was added to a DTT-free Dys ABD1 sample and allowed to incubate for 3 hr at 25°C. Free label was subsequently removed by four rounds of dialysis in 4 L of 10 mM Tris, 100 mM NaCl, 2 mM MgCl₂, and 1 mM DTT (pH 7.5). The resultant labeled protein was spin counted to determine the label concentration, which, when compared to the protein concentration, resulted in ∼84% labeling efficiency (Fig. S1). Spin-labeled samples were then loaded into quartz capillary tubes (1.1 mm inner diameter, 1.6 mm outer diameter, 15 μL sample volume) containing 7% v/v glycerol as a cryo-protectant, flash frozen in liquid nitrogen, and subsequently stored at −80°C until use.

Double-electron electron paramagnetic resonance

We performed DEER experiments to measure interprobe distances in the range of 2–6 nm. Measurements were made on a Bruker E580 spectrometer (Billerica, MA) operating at Q-band (34 GHz) with an EN5107 resonator (Bruker, Eden Prairie, MN). A four-pulse DEER protocol with a π/2 pulse width of 12 ns and an electron double resonance pulse width of 24 ns was implemented. The electron double-resonance frequency was placed on the pump position, which corresponded to the absolute maximum of the nitroxide absorption spectrum. The observe position was placed 24 Gauss higher than the pump position on the field-swept spectrum. Experiments were performed at 65 K. The resulting DEER waveform was analyzed using the model-independent Tikhonov fit provided in DeerAnaylsis2013.2. The Tikhonov distribution was then fit to multiple Gaussian distance distributions (Eqs. 1 and 2), assuming the existence of discrete structural states as described previously (19):

$$\rho \left(\mathrm{r}\right)=1/\left(\sigma \sqrt{2\pi }\right)\exp \left(-{\left(\mathrm{r}-\mathrm{R}\right)}^2/\left(2{\sigma }^2\right)\right)$$ (1)

and

$$\sigma =\left(\mathrm{FWHM}\right)/\left(2\sqrt{\ln \left(2\right)}\right)\text{.}$$ (2)

Differential scanning calorimetry

Differential scanning calorimetry (DSC) experiments were performed on a NanoDSC (TA Instruments, New Castle, DE) at a scan rate of 1°C/min using micromolar protein concentrations. Buffers consisted of 20 mM 3-morpholinopropane-1-sulfonic acid, 100 mM KCl (pH 7.5). Solutions were extensively degassed under vacuum with gentle stirring before loading into the calorimeter. This helps prevent release of air bubbles from solution during the experiment. From the thermodynamics parameters enthalpy of unfolding (ΔH), melting temperature (T_m), and change in baseline heat capacity (ΔC_p), we calculated a free energy of unfolding (ΔG) using the Gibbs-Helmholtz equation:

$$\mathit{\Delta }\mathrm{G}=\mathit{\Delta }\mathrm{H}\left(1-\mathrm{T}/{\mathrm{T}}_{\mathrm{m}}\right)+\mathit{\Delta }{\mathrm{C}}_{\mathrm{p}}\left(\mathrm{T-}{\mathrm{T}}_{\mathrm{m}}-\mathrm{Tln}\left(\mathrm{T}/{\mathrm{T}}_{\mathrm{m}}\right)\right)\text{.}$$ (3)

Further details of our analysis using multiple approaches to constrain each parameter are discussed in the Supporting Materials and Methods.

MD simulations

MD simulations were performed using GROMACS 5.0.6 (20) starting from chain A of the crystal structure of Dys ABD1 (Protein Data Bank (PDB): 1DXX (10),), corresponding to residues 9–246 of human Dys (Uniprot P11532). Similarly, for Utr, the starting structure was taken from chain A of PDB: 1QAG (9) of human utrophin. For the latter, using the Uniprot P46939 sequence, both the N- and C-termini were elongated with PyMOL by six and five residues, respectively, so that the resulting sequence 25–261 aligns well with the one simulated for dystrophin (Fig. S2). For Dys ABD1, amino acids S10 and S188 were mutated back to cysteine residues using CHARMMGUI (21), and, for both systems, the crystal water surrounding the protein was preserved. All histidine residues were set to neutral charge with hydrogen on the Nε, except for Utr ABD1 His 88 and 190, which were neutral with proton on Nδ. Periodic boundary conditions were implemented through a rhombic dodecahedron box with a box vector length of 11.6 nm. The system was solvated in TIP3P water (22), and KCl was added to achieve electroneutrality and an ionic strength of 150 mM. Covalent hydrogen bonds were constrained with the LINCS (23) algorithm, and the equations of motion were propagated with a time step of 2.0 fs. Electrostatic interactions were treated with the particle-mesh Ewald (24) algorithm with a real cutoff and a grid spacing of approximately of 1.2 and 0.12 nm, respectively. Van der Waals interactions were switched off between 1.2 and 1.0 nm. A constant temperature of 300.0 K was maintained with the V-rescale algorithm (25) applied to the protein and solvent groups independently, and a constant pressure of 1 atm was maintained with the Parrinello-Rahman barostat (26). Simulations were run using two flavors of the CHARMM force field family: CHARMM36 (27) and CHARMM22^∗ (28, 29). For each system (Dys and Utr), five simulations were run with the CHARMM36 force field and five with the CHARMM22^∗ force field for a total of 20 independent simulations. Comparisons of structural parameters derived from each force field are summarized in Fig. S3–S6. For both force fields, the system was initially minimized to remove bad contacts, and then it was equilibrated for 1 ns at constant volume and temperature followed by 3 ns at constant pressure and temperature, during which the harmonic restraints on the protein’s heavy atoms were gradually reduced from 239 kcal mol⁻¹ nm⁻² to zero. At this point, for each force field, five independent simulations were started by randomizing the atoms’ velocities and continued for 500 ns, resulting in a total simulation time of 5 μs for the ten total trajectories for each system. Coordinates of the trajectories were saved every 5 ps, and additional analyses (see Supporting Materials and Methods) indicated that to improve statistics, the trajectories obtained with the two force fields could be pooled together. The trajectories were processed with GROMACS, data analysis and plotting was performed with the software R (30), and structures were visualized with PyMOL (31).

Principal component analysis (32) was performed on the protein’s backbone atoms to identify the main conformational changes that characterize ABD1’s dynamics. To directly compare the conformational states of Dys and Utr, the principal components were calculated by pooling the trajectories of both systems together, and then each system’s trajectories were projected on the common principal components. Residue Q189 for Dys was omitted from the analysis because sequence alignment of Dys to Utr (Fig. S2) shows a gap in Utr in that position. The trajectories were then binned along the first two principal components, and the relative density of the histograms was used to calculate the free-energy difference through the Boltzmann distribution. To identify flexible regions in the protein backbone, the conformations of the Cα of four consecutive residues were binned using the MK32K25 structural alphabet (33), and the corresponding Shannon entropy was calculated using the method of Pandini et al. (34).

As a proxy for the distance between the two spin labels on residues 120 and 239, we monitored the distance between the two Cα carbons of these residues. Additionally, the solvent-accessible surface area (SASA) (35) was measured for the total protein and separated in the hydrophobic and hydrophilic contributions. The differences in SASA between the open conformation of ABD1 and the closed conformations were plotted on the closed conformation’s representative structure using PyMOL.

Results

Dys ABD1’s structure transitions from a closed to open state but with significant structural disorder

To test the hypothesis that the CH domains of Dys ABD1 undergo a conformational opening upon binding of F-actin (Fig. 1), we performed DEER experiments on a spin-labeled ABD1 fragment. Application of DEER allows us to make a direct comparison to existing structural data on the homologous Utr ABD1 fragment (7). Using site-directed mutagenesis, we removed all native cysteine residues in the Dys ABD1 construct and engineered in new cysteines at residue positions 120 and 239, positions similar to those used in Utr ABD1 in (7). Based on the domain-swapped dimer crystal structure previously determined for Dys ABD1 (10), these residue locations would be suitable for detecting a change in interdomain distance upon F-actin association.

After spin labeling the protein, we measured interprobe distances with and without varying concentrations of F-actin. In the absence of F-actin, DEER measurements identified a short interprobe distance, indicating that the CH1 and CH2 domains of Dys ABD1 are in a more compact closed state (Fig. 2, top). This is apparent in the DEER echo amplitude, which shows a rapid decay and mild oscillations in the time domain, both of which indicate a short interprobe distance. In samples containing F-actin, the DEER echo amplitude decays more slowly, indicating longer distances. The oscillations initially present in the absence of F-actin are dampened out in its presence, indicating actin-induced structural disorder (Fig. 2, middle and lower panels).

Figure 2.

DEER data on 80 μM 4-maleimido-2,2,6,6-tetramethyl-1-piperidinylo-labeled Dys ABD1 (blue data sets) in the presence of increasing F-actin from top to bottom. The molar ratio of F-actin (FA) to Dys ABD1 is shown above the time domain data. The Utr ABD1 DEER reported previously in (7) (red data sets) is overlaid for comparison. The left shows the time-domain decays. The right shows the derived distance distributions. Tikhonov distributions for Dys ABD1 (black dotted lines) were fitted to two discrete Gaussian distributions (blue solid lines) and indicate that F-actin shifts interprobe distance toward a more open structural state, but there is considerable structural disorder. Note that this contrasts with Utr ABD1’s Gaussian distributions (shown as red solid lines), which are ordered. Moreover, Utr ABD1’s structural opening is complete. For complete time domains, see Fig. S7. To see this figure in color, go online.

When comparing these structural results to the previously examined Utr ABD1, several features stand out. First, both ABDs share a common mechanism of opening upon binding to F-actin. Second, the Dys ABD1 contains much more structural disorder than seen previously in Utr ABD1. This is evident in the mild or complete lack of oscillations present in the echo amplitude decay of Dys, which, in Utr ABD1, was very well resolved. Lastly, unlike the Utr ABD1, Dys ABD1 does not undergo a complete shift to the open structural state. This is evident by the fact that a shorter interprobe distance is still sampled even in the presence of excess actin. This latter observation is consistent with previous work examining the DysABD1 open-closed structural transition upon binding actin (8); pyrene excimer fluorescence using residues C10 and C188 indicated that Dys ABD1 occupied a closed conformation when bound to actin. This method, however, would not have been able to detect the open conformers identified in this study by DEER. Collectively, the pulsed EPR data identify a common binding mode for Dys and Utr ABD1 but simultaneously highlight distinct structural dynamics in the ABD-actin interaction.

MD simulations characterize a compact and conformationally heterogeneous ABD1

To gain atomistic insight into the factors affecting the compact versus extended conformational equilibrium, we performed MD simulations starting from the Dys and Utr ABD1 crystal structures (9, 10), both of which are in an open conformation. Ten 500 ns simulations were independently run for each ABD, and in 18 out of 20 simulations, the ABD structures similarly collapsed to compact conformations. The results show that not all simulations sampled the same conformation, and a small number of stable compact conformations were identified for both Dys and Utr using principal component analysis (Fig. 3, minima B–F). The free-energy profiles in Fig. 3 show that for both Dys and Utr, at least five to six distinct compact conformations and one extended conformation exist. Moreover, these conformations have similar energies, and the different structures can easily interconvert into each other by overcoming small free-energy barriers (3–4 kcal/mol).

Figure 3.

Free-energy landscape of Dys ABD1 (A) and Utr ABD1 (B) projected on the first two principal components. The two principal components describe a “bending” motion of the two CH domains around a central swivel (PC1) that allows the extended-to-compact transition and the “revolution” motion of one CH domain around the other (PC2). A representative structure for each of the major conformational minima is also plotted. Structural heterogeneity correlates with the measured structural disorder present in DEER distributions and low unfolding free energy. To see this figure in color, go online.

Using the method of Pandini et al. (34), we calculated the backbone Shannon entropy as a proxy of its conformational flexibility and projected the calculated entropy on the structure (Fig. 4 A). The entropy of the linker region spanning from Gly-130 to Asn-135 is comparable to that of the disordered N- and C-termini, indicating that this region is characterized by a great degree of conformational flexibility, as recently suggested by Chakravarty et al. (36). This suggests that the linker region closer to the CH2 domain acts as a swivel, allowing the conformational transitions between compact and extended conformations as well as the revolution of one CH domain around the other. This role is supported by monitoring the change in φ/ψ dihedral angles between the extended (minimum A in Fig. 3 A) and a compact conformation (minimum F in Fig. 3 A), which highlights the very same region as the one in which dihedrals change the most (Fig. S8).

Figure 4.

Structural collapse of dystrophin and utrophin ABD1 derived from MD simulations. (A) The Dys ABD1 Shannon entropy calculated for macrodiehedrals formed by four consecutive Cα carbons is shown. (B) The SASA averaged over the ensemble of structures defining each minimum identified in Fig. 3 is shown, with error bars representing one SD. Representative closed structural states of (C) Dys and (D) Utr ABD1 are shown. Hydrophobic residues that promote closure are highlighted. To see this figure in color, go online.

Next, we aimed at identifying the driving force inducing the transition to compact conformations. Upon further analysis, all compact structures display a smaller SASA when compared to the extended conformation (Fig. 4 B). Although some of the residues involved may vary in each different compact structure, we observed that a core of hydrophobic residues on the same face of the amphipathic linker helix and a cluster of residues on the CH2 domain are predominantly involved in hydrophobic interactions in all compact conformations (Fig. 4, C and D; Fig. S9). Such a pattern could indicate that the closed structural states are stabilized in part by interactions involving hydrophobic side chains. A similar mechanism was recently proposed for the ABD of β-III-spectrin (37), and it is also supported by the fact that the same hydrophobic interactions are present in the extended crystal structure of Dys ABD1 (PDB: 1DXX), although because of the dimeric nature of the crystal, they are domain swapped (Fig. S10).

Experimental validation of MD simulations

To directly test the MD results that the distribution of closed conformers is promoted by hydrophobic contacts, DSC experiments were performed. The increase in a protein’s heat capacity (ΔC_p) during an unfolding transition correlates directly with increased exposure of apolar amino acids, based empirically on the free energy of transfer between organic and aqueous phases (Supporting Materials and Methods). Comparison of this calculated ΔC_p with that experimentally measured by DSC indicates the extent of hydrophobic residue exposure to water in the folded state. We found that the measured value of ΔC_p for thermal unfolding of Dys ABD1 (4.09 ± 0.07 kcal/mol) was substantially less than the calculated value (4.7 kcal/mol) (Table S1). This is consistent with the MD conclusion that the native state of Dys ABD1 has an unusually large fraction of hydrophobic residues exposed to solvent.

Next, to compare the ensemble of conformations generated by MD simulations to the experimental data obtained by DEER, we plotted for Dys ABD1 the probability density of the distance between the Cα of Val-120 and Leu-239, the residue positions at which spin labels were inserted (Fig. S11). The plot shows a broad peak centered at 1.2 nm with a shoulder at 2.0 nm, corresponding to the closed conformations. The features of the peak match the features of the same peak in the DEER experiments in absence of actin (Fig. 2). We characterized the two underlying distributions as both due to compact conformations, differing only by the relative orientation of the two CH domains (Fig. S11). In the DEER data, the peak is centered at ∼1 nm longer distance than in the data from the simulations. This could be explained by the fact that in the simulations, the distance was mapped between the two Cα rather than from the ends of the long TEMPO spin label. In addition, the two mutations introduced to add the spin labels may somewhat interfere with the hydrophobic patch between the linker and the CH2 domain (in particular, perturbing Val-120 and Phe-236, which is only three residues away from Leu-239). Yet it was reassuring to notice that the change in distances metric was preserved. In fact, the DEER data indicate an increase of ∼2.5 nm upon transition from the compact to the extended conformations (Fig. 2, top and bottom panels), and a similar trend is observed in the simulated data when considering the small shoulder at 4.0 nm, which represents the extended conformation.

Discussion

The diverse biophysical methods employed in this study were geared toward identifying structural differences in Dys and Utr ABD1-actin complexes and understanding how they, in turn, may contribute to differences in function. We found that, similar to Utr, the Dys ABD1 can exist in two main structural states: a compact state in which the CH domains are in close apposition and a more extended open state that is stabilized by actin. Unlike our previous study of Utr ABD1 (7), Dys ABD1 did not shift completely to its extended open state; a significant mole fraction of ABD1 continued to sample compact states even with an excess actin concentration (Fig. 2). Moreover, the Dys ABD1-actin complex is much more structurally heterogeneous than Utr ABD1. This indicates that the two protein domains, though homologous, may functionally diverge in part because of distinct structural dynamics.

Support for the distinct structural dynamics comes not only from DEER (Fig. 2), in which the full-width half maximum of Dys ABD1 distance distributions far exceeds that of Utr ABD1, but also MD simulations (Figs. 3 and 4). In the latter case, for instance, an overlay of structural states derived from Fig. 3 indicates that Dys and Utr ABD1 are sampling distinct conformational spaces. If the CH1 domains of each ABD1 are aligned in pairwise fashion and the positions of the linker and CH2 are examined, not a single pair of structural states from the free-energy minima orient exactly the same way (Fig. S12). Although we observed several general similarities, even in the most favorable of cases, the orientation of the two CH domains were still distinct such that no structure for the Dys ABD1 could be mapped onto a structure for Utr ABD1.

When considering how these distinct ABD1 structural dynamics could impact function, two possibilities come to mind: alteration of ABD1 affinity for actin and alteration of actin structural dynamics when bound to filaments. With regard to modulation of affinity, recent work on Dys and Utr ABD1 indicates that actin binding is largely mediated by the CH1 domain (15, 38), as has been suggested in other binding studies of ABDs or through inference from CH2 domain mutations that increase affinity for actin (17, 39). We used this evidence to justify aligning our simulation-derived Dys ABD1 structures with that of the CH1 domain of β-III-spectrin in a recently reported cryo-EM structure and assessed whether any of the stable conformers were capable of docking without steric hindrance (Fig. 5) (17). From this analysis, we found that docking the open extended conformers from the MD simulations is compatible with binding, because the linker and CH2 domain protrude and point away from actin, avoiding any potential steric clash. Steric clash was one of the main reasons originally proposed for ABDs needing to open upon binding (16). We also found, however, that although many of the closed compact conformers result in steric clashes between the CH2 domain and actin, at least some compact conformations for either Dys or Utr are compatible with binding to actin (structure B in Fig. 3 A for Dys ABD1, and structures B, D, and E in Fig. 3 B for Utr ABD1). This is in agreement with the recent work of Shams et al. (40), which suggests that various binding modes between ABD1 and actin could be accessible. Therefore, the conformational ensembles computationally characterized could explain how both compact and extended conformations of ABD1 are sampled when bound to actin.

Figure 5.

Modeling of MD simulation-derived Dys ABD1 conformers on actin filament. When the CH1 domain of Dys ABD1 conformers in free-energy minima of Fig. 3 is aligned with the CH1 domain of β-III-spectrin ABD from a recent 6.9 Ǻ cryo-EM structure (6ANU), some of the structural models are devoid of steric clashes (green, yellow, red). The open conformation (magenta) is similarly free of steric clashes. Although some closed conformations of Dys ABD1 exhibit significant steric clashes with actin (blue), the fact that others do not suggests there are binding-compatible closed states for Dys ABD1, in agreement with previous structural measurements using pyrene excimer fluorescence (8). To see this figure in color, go online.

The notion that Dys and Utr ABD1 closed structural states have high degeneracy and that some are potentially “binding competent” whereas others are not suggests that ABD1’s structural dynamics could directly impact affinity for actin. If, for example, the most stable closed conformers of ABD1 are those that give raise to steric clashes when binding to actin, shifting the equilibrium to a binding competent conformer of higher free energy will result in a reduced binding affinity. Alternatively, if multiple closed conformers are free of steric clashes and thus binding competent, affinity for actin would be increased. In this way, the shifting of conformational equilibria in the structurally dynamic unbound state could play a role in fine-tuning affinity for actin.

It is important to point out that the energetics at play in such a binding model are compatible with existing measurements of ABD1 affinity. In Dys ABD1, for instance, the ΔG of actin binding derived from reported K_d that ranges from 10 to 50 μM (38, 41) is an energetically favorable 6–7 kcal/mole. Our calculations show that the energy difference between the various closed and open ABD1 conformers should be smaller than this range (Fig. 3) and thus be capable of an actin-induced redistribution of the structurally dynamic ensemble. Unfortunately, the computational precision of the free-energy profile constructed from our unbiased simulations does not allow any further extrapolation. Although there appears to be more than one conformer in the ABD1 ensembles that is potentially binding competent (Fig. 5), in the absence of simulation data that include actin filaments, no conclusions can be drawn about which conformation is the most stable when bound. As such, it is not possible to directly explain the differences in the experimentally detected spin label distances in Dys and Utr (Fig. 2) as due to the binding of one specific more stable conformer identified by MD simulations. However, DEER measurements indicate that it is primarily an open conformation for Utr and a heterogeneous mixture for Dys. Future work could potentially probe this structural heterogeneity further using pressure perturbation EPR methods (42).

With regard to how Dys and Utr differentially alter actin dynamics, the structural behavior measured by DEER provides insights into a mechanistic contribution (Fig. 2). When full-length Dys and Utr bind actin, they each restrict the amplitude of bending and twisting motions of the filament, although Utr does so to a much greater degree. Part of the difference is probably explained by binding of each protein’s ABD2 (6), which varies in their proximity to ABD1 (43, 44). However, another contributing factor may be distinct structural dynamics within each protein complex. Because actin itself can adopt many conformations upon twisting and bending (45), binding of ABDs in general may impose on the filament’s conformational exploration. If bound ABDs differ, however, in their ability to conformationally adapt to actin fluctuations, they may consequently have distinct restrictive capacities. If these dynamic differences persist in their full-length structures, they may locally contribute to the overall global differences measured previously (5, 6). Additionally, if bound ABDs differ in the degree to which they can adapt to structural changes of actin, distinct ABD dynamics may also contribute to ABD’s ability to remain bound under mechanical stress experienced by the filament. The fact that the computationally characterized ensembles of Dys and Utr ABD1 have similar general features but differ in the specific orientation of the CH domains supports this as a viable adaptation mechanism.

Overall, the results presented in this study agree well with previous work on Dys ABD1 and other ABDs from the spectrin superfamily and simultaneously provide new insights, to our knowledge. In a prior study examining the question of whether or not the Dys ABD1 opens upon binding, for instance, pyrene excimer fluorescence data indicated that the ABD1 could bind in a closed conformation (8). Although these results could not determine the distribution of conformational states when bound to actin, they are in agreement with our current findings. We showed in Fig. 2 that even in a 4:1 excess of actin, ∼50% of the ABD1 is bound in a compact conformation. However, what our results show in addition is that the remaining ∼50% of the bound population is present in an extended conformation, making our findings complementary to that of Singh et al. (8). We have already pointed out that different experimental techniques can capture ABD domains in both compact and extended conformations. Such differences, when considered in conjunction with a flat free-energy landscape of easily interconvertible conformers, suggest that ABDs are malleable proteins responsive to small external perturbations. A perturbation on the order of just 1 kcal/mole could shift the equilibrium from one conformation to another (Fig. 3).

This proposed mechanism many not only contribute to our understanding of ABDs being found in a variety of conformations but also the impact of disease-causing mutations. If the conformational ensembles of ABDs are easily perturbed, defects such as missense mutations could significantly alter the distribution of structural states. If such mutations redistributed the conformational ensemble in a way that more readily populates binding-competent conformers, for instance, through the mechanism discussed above, it could help explain the modest to significant gains in affinity observed in ABDs associated with disease (37, 46). In addition, mutation-induced redistribution may alter the conformational ensemble in such a way that there is a gain of conformational adaptability that allows binding to persist despite structural fluctuations of the filament.

Application of this generalized ABD model to DMD and BMD mutations in Dys requires some caution. Many DMD and BMD missense mutations that have been studied previously are mutations that have substituted CH-domain core hydrophobic residues for charged or polar residues (10). When such mutations were studied in the isolated Dys ABD1 fragment, the main affect seemed to be misfolding of the domain rather than change in binding affinity (47). Indeed, when many of the same missense mutations were examined in full-length Dys, changes in binding affinity were modest or statistically insignificant (48). This suggests that for DMD and BMD, missense mutations that perturb the core fold of ABD1 CH domains are promoting disease via a misfolding-induced degradation mechanism (49). For some missense mutations—perhaps those that are more superficially located within the ABD1, or those that do not misfold the ABD1 but perturb stability—a functional impact like that discussed above for ABDs in general may become relevant. However, without clear examples of DMD and BMD missense mutations that alter binding affinity while simultaneously preserving the ABD1 fold, this potential mechanism will remain speculative and bear most relevance to other ABDs in which missense-mutation stability-structure-affinity relationships are better established (17).

One of the observations made here points to an interesting feature of Dys ABD1 that will probably require further investigation in future studies. Although DSC measurement of heat capacity helped validate the MD-suggested use of surface hydrophobicity in modulating closed conformer sampling, it also revealed a rather complex endothermic transition (Fig. S13). The absolute heat capacity of a protein is a measure of underlying structural diversity in the folded and unfolded states but can also be used to evaluate the presence of intermediate states. It is not known whether the distribution of conformers identified for Dys ABD1 reflects a distribution of equivalent distinct states or arises as a distribution or sampling of partially folded intermediate states. We find that the diversity of open and closed conformers corresponds energetically to a two-state model comprised of a folded and unfolded state. The unfolded state retains a surprising amount of order, consistent with retaining hydrophobic contacts even under extreme conditions. This suggests that the energetic functional landscape of Dys ABD1 may be capable of absorbing much more energy than anticipated from the distribution of open and closed conformers identified. However, further quantitative assessment of such a thermodynamic prediction will require application of additional biophysical techniques.

Conclusions

We have probed Dys ABD1 dynamics with complementary experimental and computational techniques. We have characterized free ABD1 in solution as adopting a compact conformation. With computer simulations, we have shown that both Dys and Utr ABD1 are conformationally malleable proteins for which low-energy compact and extended conformations are easily accessible through a flat free-energy landscape. We have identified the region of the linker that grants conformational flexibility to the protein and the hydrophobic patches responsible for stabilizing the compact conformations. We have characterized the conformational ensemble of actin-bound Dys ABD1 as originating from an equal contribution of compact and extended structures. Finally, we have linked our findings and proposed a model for how unique structural dynamics of Dys and Utr ABD1 contribute to their distinct functions.

Author Contributions

M.E.F. purified and labeled protein for DEER. M.E.F., A.Y.L., and A.R.T. acquired DEER data. M.E.F., A.Y.L., A.R.T., and D.D.T. analyzed DEER data. B.H. and A.C. performed and analyzed MD simulations. C.C., R.M., J.S., and E.G. purified protein for DSC and performed DSC experiments. C.C., R.M., J.S., E.G., and A.H. analyzed DSC results. M.E.F., A.C., A.H., and D.D.T. wrote the article.

Editor: Laurent Blanchoin.

Supporting Citations

References (50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64) can be found in the Supporting Material.