Dynamic charge interactions create surprising rigidity in the ER/K α-helical protein motif

ER/K α-Helix Is Recurrent in Nature.

The number of both known and hypothetical or predicted protein sequences in the database with the ER/K motif is revealed by a BLAST search in the PubMed database of nonredundant protein sequences including all species. To ensure a nonbiased evaluation of search hits, (E₄K₄)₂ and (E₄R₄)₂ were used as search templates, and the list of sequences that scored better than any species of myosin X or myosin VI was examined to assess the relative number of proteins that contain the ER/K motif. Because tail fragments in myosin VI and myosin X are known to be isolated α-helices in solution, we propose that our search criteria provide a conservative estimate of similar α-helices in other protein structures. We find the ER/K motif is present in at least 123 known distinct proteins in 137 organisms ranging from archaea to humans, with at least 20 distinct proteins in mammals [e.g., supporting information (SI) Table S1]. As mentioned above, the ER/K motifs in the three proteins that are known to form stable α-helices in solution are sparsely interspersed with residues A, L, I, V, M, Q, and D. The relative content of residues E, R, and K in these helices is 70% in caldesmon, 78% in myosin X, and 79% in myosin VI. When we analyzed the results of the BLAST search for continuous ER/K regions matching the search criteria, with the boundaries of this region determined by the presence of amino acids not listed above, we found that the mean content of E, R, and K in helices obtained from the search is 80%. By using this approach, the ER/K motif is >30 residues in at least 40 distinct proteins with the longest continuous ER/K region being 346 residues for a Kelch motif family protein in Trichonomis vaginalis G3.

MD Simulations Reveal Dynamic Charge Interactions in ER/K α-Helices.

To understand the effects of E and R/K side-chain interactions on the stability and function of ER/K α-helices, we performed REMD simulations on the sample sequence (E₄K₄)₂ in explicit solvent. We selected (E₄K₄)₂ as a suitable candidate because it allows the side chains to sample multiple interactions (i to i ± n where n = 1–7) and seek out the most energetically favorable ones. Also, (E₄K₄)₂ is known have high α-helical content in solution with known temperature, pH, and salt-concentration dependence, as measured by circular dichroism (CD) (16). These experimental measurements provide a quantitative benchmark of our REMD simulations. Last, the (E₄K₄)₂ sequence is well represented in the protein mannosyltransferase (Saccharomyces cerevisiae), which contains four continuous perfect repeats of (E₄K₄)₂. In the absence of known crystal structures of ER/K α-helices, we performed REMD simulations starting with the (E₄K₄)₂ peptides in both random (no secondary structure) and fully α-helical conformations (see Methods). Convergence of simulation was assessed by insignificant differences in fraction helix (see SI Methods) at all simulated temperatures (Fig. 1A and Fig. S1). The convergence of two different initial conditions to a unique solution is essential to obtain equilibrium conformations for both the peptide backbone and side-chain interactions, especially because the objective of these MD simulations is to study side-chain interactions.

The thermal melt curves obtained from REMD (Fig. 1A) are very similar to experimental data obtained for (E₄K₄)₂ by using CD (16). CD data for the MT segment of myosin VI are shown in Fig. S2 with the fraction helix displayed in Fig. 1A. The fraction helix for the myosin VI MT with ≈70 residues is higher than that for (E₄K₄)₂ with 16 residues, as expected due to the increased length of the peptide chain (11).

We used REMD simulations to equilibrium sample a large range of protein conformations by using finite computing power (17). However, due to periodic exchange between different replicas, REMD does not provide trajectories of side-chain interactions in real time. To obtain this information, we randomly selected (E₄K₄)₂ replicas that were completely folded from random initial configuration and equilibrated them in explicit solvent for 25 ns at 274 K. These simulations were continued for an additional 25 ns with trajectories of all protein atoms monitored over this time period. The N_ζ atom of residue K7 (i) oscillates between the center of mass of the two O_ε atoms of E4 (i − 3) and E11 (i + 4) (Fig. 1B). The center of mass of the two O_ε atoms of E10 (i), alternately, oscillates between the N_ζ atom of residue K6 (i − 4) and K13 (i + 3) (data not shown). This preference for (i − 3)/(i + 4) interactions at K(i) residues and (i − 4)/(i + 3) interactions at E (i) residues is seen for all of the center eight residues (K5-E12) of the (E₄K₄)₂ helix. The preference for these interactions is reflected in the histograms of the frequency of all distances for i − 3/4 and i + 3/4 interactions, at each of the eight residues K5–E12. These histograms reveal clear 3-Å peaks for the specific interactions listed above (Fig. S3). Residues in the first turn of the helix (E1–E4) do not have i − 3 and i − 4 interactions, whereas residues in the last turn of the helix (K13–K16) do not have i + 3 and i + 4 interactions.

Closer examination of the instantaneous structures of the folded (E₄K₄)₂ α-helices show that the particular pattern of interactions for E and K residues arises from a combination of the following: permissible dihedral angles between the peptide backbone and the C_α–C_β side chain of the E and K residues; and the right-handed structure of the α-helix causing residues to rise left to right from the N to the C terminus of the peptide backbone of the α-helix (Fig. 1 C and D). For example, K27, E31, K34, E38, K41, E45, K48, and E52 form a network of interactions that spans nearly 4 nm along one face of the α-helix (Fig. 1C). This alternating pattern of + 4 then + 3 closely approximates 3.6 residues per turn of an α-helix. Therefore, the ER/K motif allows for continuous series of interactions along the same face of the α-helical backbone. We use this alternating pattern of + 4 then + 3 to draw a map of preferred interactions along the α-helical backbone, both for the extended (E₄K₄)_n helix and for ER/K α-helices in general (Fig. 1C and Fig. S4). A striking feature in such maps is several continuous series of interactions in all ER/K α-helices (Fig. 1 C and D; boxes in yellow). In the (E₄K₄)_n α-helix, eight residues form a chain of oscillating interactions, whereas in upf-2 regulator of nonsense transcripts (Table S1) we see a maximum of 12 residues (Fig. 1D, arrows).

The oscillation of E and K side chains between their i − 3/4 and i + 3/4 neighbors implies that at every time step of the MD simulation, the minimum distance between each N_ζ of K and the centers of mass of O_ε of all its E neighbors (and vice versa) is the dominant contributor to the interaction between E and K side chains. Fig. 2A shows the histogram of these minimum distances collected >25 ns for residue K7, and Fig. 2B shows the combined histogram of these minimum distances for residues K7 through K12. The preference for (i − 3)/(i + 4) interactions at K(i) residues and (i − 4)/(i + 3) interactions at E(i) residues, described above, holds true in this combined histogram, in that for 90% of the simulation time the closest pairing residue for K or E follows this preferred pattern. The histogram can be best characterized by a maximum-likelihood fit to the sum of three Gaussian distributions (SI Methods) with centers located at 3, 5, and 8 Å. The relative times spent in the three modes of interaction, in turn, are given by the areas under the individual Gaussian distributions to be 45%, 37%, and 18%, respectively. Detailed examination of interatomic distances reveals that the interaction centered at 3 Å corresponds to hydrogen bonding between the H_ζ of K and O_ε of E (Fig. 2C and Fig. S5). For the interaction centered at 5 Å, the E and K side chains are located approximately equidistant from their i − 3/4 and i + 3/4 neighbors (Fig. 2C). For these conformations, we find that the O_ε atom of E is located 3–6 Å from H atoms in the K side chain, suggesting a combination of direct interaction through Coulomb forces between opposite charges and indirect interactions through solvent-separated salt-bridge contacts (Fig. S5). The third peak centered at 8 Å has contributions from the following: the continuous series of alternating E and K interactions located seven residues apart implies that residues K8 and E9 do not have i − 3 and i + 3 neighbors, respectively (Fig. S3); and due to oscillations of side chains between their i − 3/4 and i + 3/4 pairs, each side chain spends a small fraction of its time (<10%) without an interacting partner.

Side-Chain Interactions Significantly Influence The Stability and Rigidity of ER/K α-Helices.

The stability of isolated α-helices in the presence of polar solvents such as water is limited by disruption of hydrogen bonds in the α-helical backbone (4, 5). Previous comparisons of MD simulations of alanine-based α-helices with SAXS measurements suggest that these helices undergo continuous transient breaks in the α-helical backbone that reduce the end-to-end distance of the α-helix (11). Our MD simulations show that there are multiple E and R/K side-chains interactions for each turn of the ER/K α-helix, which prevent transient breaks in backbone hydrogen bonds from unraveling the α-helix. Additionally, the ER/K α-helix partially excludes water molecules from the backbone (Fig. 3A), approximately to the same extent as the alanine-based F_s peptide (simulation data not shown). This partial water exclusion is potentially a combination of space occupied by the long E and R/K side chains and the hydrophobic methylene groups in them. Therefore, outer charges in the E and R/K side chains that interact favorably, along with inner hydrophobic groups that exclude water molecules, offer a unique combination to stabilize the backbone conformation of ER/K α-helices in the aqueous cellular environment.

There are two potential sources of bending rigidity of isolated ER/K α-helices, the α-helical backbone and side-chain interactions. Previous computational analyses of alanine α-helices that ignore transient breaks in the α-helical backbone report a bending stiffness of ≈1 pN/nm (scaled for a 10-nm long α-helix) (18, 19). However, in the presence of polar solvent, with partial water exclusion, the contribution of the α-helical backbone to the bending stiffness is unclear. With regard to side-chain contribution to bending stiffness, the backbone carbonyl oxygen, amide hydrogen, and C_α atoms are of 2.2 Å from the helix axis, whereas N_ζ and O_ε of the K and E side chains, respectively, are located on average at 6 Å from the axis of the ER/K α-helix. Thus, bending forces on the ER/K α-helix will first be “sensed” and, therefore, resisted by side-chain interactions before they propagate to the backbone (Fig. 3B). In this regard, the E and R/K side-chain interactions resemble a continuous narrow tube of radius 6 Å that reinforces the bending rigidity of ER/K α-helices. To examine the contribution of E and R/K side-chain interactions to the bending rigidity, we use a simplified model for the bending of the ER/K α-helix as a cantilever beam with side-chain interactions representing the elastic elements that resist bending (Fig. S6). Our model is valid only for external mechanical forces applied over time scales that are significantly greater (10 μs) than the dynamic time scale for side-chain interactions shown here (100 ns). Fig. 3B shows an exaggerated schematic of the bending of the α-helix with the resulting deformations along the α-helical backbone.

To estimate the stiffness of individual side-chain interactions, we interpret the two Gaussian distributions centered at 3 and 5 Å as representing two favored potential wells. The width of these three Gaussian distributions can be used to estimate the harmonic spring constants of 3,200 and 440 pN/nm, respectively, for these potential wells. Whereas the hydrogen bonding interaction at 3 Å is approximately 10-fold stronger than the 5-Å one, the softer 5-Å interaction has an approximately 3-fold greater reach due to larger standard deviation (0.3 vs. 0.9 Å). For side-chain-interaction distances corresponding to the 5-Å potential well, each E (i) interacts simultaneously with K (i − 4) and K (i + 3). However, for the 3-Å potential well, each E (i) interacts with only one of K (i − 4) and K (i + 3). Thus, the 5-Å potential well, in addition to its longer reach, provides a continuous sheet of interactions without breaks along the α-helical backbone. Therefore, we model the side-chain interactions as a continuous narrow tube with stiffness corresponding to the 5-Å potential well. We thus estimate a stiffness of 0.5 pN/nm entirely from side-chain interactions (Fig. S6 and SI Methods). Our model suggests that the ER/K α-helix can act as a relatively stiff structural element in mechanoenzymes such as myosin VI and X, based on side-chain interactions alone.

Single-Molecule Analyses and SAXS Are Consistent with High Rigidity of ER/K α-Helices.

We studied the contribution of the MT of myosin VI (MT ER/K α-helix) to its effective lever arm stroke under load applied by a dual beam optical trap (see Methods) (20). The myosin VI lever arm strokes through ≈180° parallel to the F-actin filament (21), resulting in a lateral bending force transmitted through the MT ER/K α-helix (Fig. 4A). A measure of the stroke size of two myosin VI constructs, one that extends to the beginning and another to the end of the MT ER/K α-helix, enabled us to measure the deflection of the MT ER/K α-helix under lateral bending load.

To characterize the stroke of myosin VI, it is essential to understand the structural elements involved in the myosin VI lever arm. We have previously described the tail domain of myosin VI as being composed of three segments, a globular proximal tail (PT), a central region containing the MT ER/K α-helix, and a terminal globular segment containing the distal tail (DT) and cargo binding domain (8). We had previously shown by using SAXS analysis that the PT alone is an ≈3-nm globular structure, whereas the SAXS envelope for the MT-DT (120 residues) is suggestive of a single α-helix (MT) coupled with a folded domain (DT).

Here, we use SAXS analysis to get insight into the contribution of the PT to the myosin VI lever arm, along with a SAXS envelope for the MT ER/K α-helix alone. To understand the contribution of the PT to the myosin VI lever arm, we obtained a SAXS envelope structure for a segment of the myosin VI molecule starting with the second IQ motif (with bound calmodulin) to the end of the PT (Fig. 4B Upper). Our structure for this complex comes from a fusion of the IQ-calmodulin structure-based on previously modeled myosin VI structures (22) and Rosetta prediction of the PT (see Methods). The IQ and PT were fused at Pro-835 and the backbone dihedral angles at the IQ–PT junction were adjusted manually to fit the structure optimally into the SAXS envelope. The modeled structure was used to predict a 15-nm stroke size of the myosin VI lever arm to the beginning of the MT ER/K α-helix (Fig. 4A). This prediction is verified by a 15 ± 3-nm stroke size measured in an optical trap (Fig. 4C Upper).

The SAXS envelope of the MT ER/K α-helix reveals an ≈10-nm long envelope that matches the predicted length of a single isolated α-helix containing ≈70 residues comprising the MT (Fig. 4B Lower). Interestingly, the SAXS envelope has two distinct bends whose locations along the α-helix correspond to two highly conserved hydrophobic patches along the length of the MT (Fig. 4B Lower). These SAXS results would be most consistent with these hydrophobic patches being relatively rigid kinks rather than flexible nodes. To understand the role of hydrophobic patches on stability of the α-helix, we performed MD simulations of different segments of the MT folded into complete α-helices and then simulated at 400 K (Fig. S7). We found that all segments of the MT remain complete α-helices, whereas the peptide (E₂K₂)₄ unfolds to a random coil (Fig. S7), suggesting that these hydrophobic patches in the MT do not unravel the ER/K α-helix. However, the simulations do show that regions that contain these hydrophobic residues undergo transient unfolding and refolding more frequently than for completely charged ER/K repeats. The functional significance of this observation remains to be elucidated.

The maximal load applied in the optical trap is the stiffness of the optical trap multiplied by the stroke size (displacement of the bead). Our theoretical model predicts that side-chain interactions contribute significant stiffness to ER/K α-helices. We performed optical trapping measurements at a trap stiffness of ≈0.005 pN/nm, which should result in small deflections (<1 nm) of ER/K α-helices with a stiffness of ≈0.5 pN/nm estimated from our model (Fig. S6). We found that the stroke size of a myosin VI construct that extends to the end of the MT ER/K α-helix is 33 ± 4 nm, which is the same as the predicted stroke size of 33 nm measured from the modeled structure of the same myosin VI construct. It should be emphasized that the angle between the PT and MT ER/K α-helix is unknown. However, the orientation shown in the schematic illustration is consistent with the SAXS structures of PT-MT and PT-MT-DT published previously (8) and aligns the MT ER/K α-helix parallel to the F-actin filament to allow dimerization of the cargo binding domains and the ≈36-nm processive stepping of myosin VI (23).

In summary, we show here that the ER/K α-helix forms a relatively rigid and stable isolated secondary structural element in proteins. The ER/K α-helix is much more prevalent in protein sequences than previously reported. We find a unique oscillating interaction of charged side chains distributed along the α-helical backbone. This dynamic interaction of side chains can tolerate significant change in interaction distances that result from lateral bending of the α-helix, compared with locked interactions resulting from salt-bridge formation. We propose a model of the ER/K α-helix as a cantilever beam, which despite its many simplifications is sufficient to account for the high stiffness of the ER/K α-helix observed in both SAXS and single-molecule optical-trapping experiments. The ER/K α-helix in myosin VI, myosin X, and caldesmon acts as a bridge between protein subdomains. Specifically, in myosin VI a load applied at the end of the cargo binding domain must be transmitted through the ER/K α-helix to the head, where it is known to alter the kinetics of nucleotide binding and release (24). The rigidity of the MT ER/K α-helix is likely to impact the force transduced to the myosin VI catalytic head and, therefore, mechanosignaling of myosin VI. Last but not least, the ability of the ER/K α-helix to efficiently span large distances as an isolated protein structural element makes it a promising tool for use in designing synthetic proteins.

MD Simulations.

REMD of the E₄K₄ peptide (CH₃CO-EEEEKKKKEEEEKKKK-NHCH₃) were performed with GROMACS (25, 26) by using the AMBER2003 force field (27). Two simulations, one starting with the peptide in complete α-helix form and the other starting with the peptide as a random coil, were performed for 80 and 42 ns, respectively, for each replica. Trajectories of side-chain interactions were determined from 50 ns simulations of individual replicas at 274 K. Details of simulation parameters and data analysis are described in SI Methods.

Biophysical Characterization of Segments of Myosin VI.

IQ-PT (814–918 Homo sapiens) bound to human calmodulin in an equimolar ratio along with myosin VI MT (916–981 H. sapiens) were purified by using standard protein purification techniques (SI Methods). SAXS measurements, data analysis, and reconstructions were carried out as described in Spink et al. (8) (see SI Methods and Fig. S8 for specific details). CD, multiangle laser light scattering (MALLS), and dynamic light scattering (DLS) measurements were carried out as described in Spink et al. (8) (see SI Methods for details).

Single-Molecule Optical-Trap Assays.

Optical trapping was performed by using myosin VI molecules extending to the beginning (1–918 Sus scrofa) and end (1–981 S. scrofa) of the MT (see SI Methods for details). Optical trapping was performed as described in Bryant et al. (21).