Docking Troponin T onto the Tropomyosin Overlapping Domain of Thin Filaments

Abstract

Complete description of thin filament conformational transitions accompanying muscle regulation requires ready access to atomic structures of actin-bound tropomyosin-troponin. To date, several molecular-docking protocols have been employed to identify troponin interactions on actin-tropomyosin because high-resolution experimentally determined structures of filament-associated troponin are not available. However, previously published all-atom models of the thin filament show chain separation and corruption of components during our molecular dynamics simulations of the models, implying artifactual subunit organization, possibly due to incorporation of unorthodox tropomyosin-TnT crystal structures and complex FRET measurements during model construction. For example, the recent Williams et al. (2016) atomistic model of the thin filament displays a paucity of salt bridges and hydrophobic complementarity between the TnT tail (TnT1) and tropomyosin, which is difficult to reconcile with the high, 20 nM K_d binding of TnT onto tropomyosin. Indeed, our molecular dynamics simulations show the TnT1 component in their model partially dissociates from tropomyosin in under 100 ns, whereas actin-tropomyosin and TnT1 models themselves remain intact. We therefore revisited computational work aiming to improve TnT1-thin filament models by employing unbiased docking methodologies, which test billions of trial rotations and translations of TnT1 over three-dimensional grids covering end-to-end bonded tropomyosin alone or tropomyosin on F-actin. We limited conformational searches to the association of well-characterized TnT1 helical domains and either isolated tropomyosin or actin-tropomyosin yet avoided docking TnT domains that lack known or predicted structure. The docking programs PIPER and ClusPro were used, followed by interaction energy optimization and extensive molecular dynamics. TnT1 docked to either side of isolated tropomyosin but uniquely onto one location of actin-bound tropomyosin. The antiparallel interaction with tropomyosin contained abundant salt bridges and intimately integrated hydrophobic networks joining TnT1 and the tropomyosin N-/C-terminal overlapping domain. The TnT1-tropomyosin linkage yields well-defined molecular crevices. Interaction energy measurements strongly favor this TnT1-tropomyosin design over previously proposed models.

Significance

The troponin T subunit of the troponin complex is a hotspot for mutations that lead to cardiomyopathy and skeletal muscle disease. However, high-resolution structures of troponin on thin filaments have not been solved, and therefore, the assembly of even wild-type troponin onto F-actin-tropomyosin remains uncertain. Here, molecular docking and extensive molecular dynamics provide a comprehensive model of actin-TnT-tropomyosin interaction replete with electrostatic linkages and hydrophobic networking. Our studies provide a baseline for a future examination of regulatory imbalances imposed by disease-rendering point mutations, potentially connecting early stage molecular insults to cardiomyopathy and skeletal muscle pathology.

Introduction

The concerted contributions of the actin-based thin filament regulatory proteins, troponin and tropomyosin, initiate an intricate cooperative, allosteric mechanism to control actin-myosin interaction in skeletal and cardiac muscle (1, 2, 3). Here, troponin responding to changing calcium ion concentration modulates tropomyosin position on actin to unblock myosin-binding sites and begin the activation process. This leads to actin activation of myosin ATPase, cross-bridge cycling on thin filaments, and hence, muscle contraction. For troponin-tropomyosin to effectively communicate with all actin subunits along thin filaments and participate in cooperative/allosteric regulation of actin-myosin interaction, tropomyosin must polymerize head to tail to form continuous cables along each of the filaments’ two long-pitch actin helices. In turn, troponin, a three-subunit complex, is anchored via its structural troponin T (TnT) tail (TnT1) over the tropomyosin head-to-tail overlapping domain (4, 5, 6, 7). TnT thereby acts as a molecular staple to augment and thus assure strong head-to-tail tropomyosin association. In addition, this intermolecular interaction defines the location of the rest of the complex, namely the troponin regulatory head (a.k.a. core domain) containing troponin C (TnC), the Ca²⁺ receptor, and troponin I (TnI), which reversibly influences tropomyosin positioning on actin in response to Ca²⁺ binding to TnC (see Fig. 1). Although it is well known that the TnT structural tail of troponin is a hotspot for disease-linked mutation (8, 9, 10, 11, 12, 13, 14, 15), its detailed residue-residue interactions with tropomyosin have not been well characterized at high-resolution. Hence, the binding of the entire troponin complex and its dynamics on actin-tropomyosin remain uncertain.

Figure 1.

The orientation and component organization of TnT and tropomyosin on F-actin. The molecular model shows pairs of head-to-tail tropomyosin dimers (pink, magenta) associated with the two long-pitch F-actin helices formed by actin subunits (pale blue). The filaments pointed and barbed ends are indicated. Four-helix bundles formed by the head-to-tail junctions between adjacent tropomyosin dimers are marked by brackets, thus joining the N-terminal ends of the lower pink tropomyosin chains and the C-terminal ends of the upper magenta chains. The F-actin-tropomyosin structure is taken from Orzechowski et al. (19). The cardiac TnT1 helix representing residues 89–151 is shown docked onto actin-tropomyosin as determined in the current study (yellow helices on front and back face of F-actin). Explicit interactions between tropomyosin and TnT1 residues 152–202 beyond the docked structure and the beginning of the troponin core domain TnT are uncertain but are acknowledged to extend longitudinally toward the pointed end of the filament to the surrounds of residue 174 on tropomyosin (56). Given the structural uncertainty of this region of TnT1, it is demarcated schematically as a yellow rectangle and shown connected just to the front-facing TnT1 helix for simplicity. Similarly, N-terminal TnT residues (1–88) are also uncharacterized structurally and symbolized by a yellow box on the opposite end of the TnT1 helix. The N-terminal region is not expected to extend longitudinally given previous classic results (7,16). Connectivity between tropomyosin-linked TnT and the core domain TnT2 is not definitive (yellow arrow); its presumed off-actin location is indicated by an asterisk (20). To see this figure in color, go online.

Troponin-tropomyosin organization

The N-terminus of each tropomyosin dimeric coiled coil is directed toward the pointed end of thin filaments and binds to the C-terminus of the next molecule (7,16). The intermolecular junction between adjacent tropomyosin coiled coils creates a short four-helix bundle “overlapping domain” (Fig. 1; (17, 18, 19)). Each tropomyosin dimer spans seven actin subunits and can be subdivided into pseudorepeats that span a single actin subunit (1, 2, 3). In turn, one troponin complex binds each tropomyosin molecule by means of the extended TnT tail (Fig. 1), whose sequence appears to be antiparallel to that of tropomyosin (16). The tropomyosin coiled coil, the bound ternary troponin complex, and the underlying seven actin subunits comprise a single “regulatory unit” (1, 2, 3). Domain mapping suggests that the troponin “core domain,” including TnI, TnC, and C-terminal domains of TnT, localizes over the central pseudorepeats of tropomyosin (Fig. 1; (20)). The elongated N-terminal TnT tail (∼16 nm) points away from the troponin core domain toward the barbed end of the filament (7). The TnT tail (TnT1) (21) thereby is aligned against the C-terminal half of tropomyosin, proceeds over the tropomyosin overlapping domain, and then onto N-terminal pseudorepeat 1 of the adjacent tropomyosin molecule (Fig. 1; (7)). The ∼18.5-nm contour length troponin tracks 4–5 successive actin subunits of the thin filament regulatory unit (20). Thus, in addition to its role in strengthening the polymeric tropomyosin cable, TnT acts as a molecular ruler and defines the orientation and periodicity of the troponin complex on thin filaments. Conformational changes between the troponin components involved in Ca²⁺ and myosin-binding have been studied by using NMR, Förster resonance energy transfer, and chemical cross-linking (reviewed in (22)). However, these methods lack sufficient precision to establish the troponin position along thin filament actin and tropomyosin at an atomic level and thus to definitively follow the regulatory transitions of troponin accompanying those of tropomyosin on thin filaments.

Electron microscopy image reconstruction of isolated thin filaments has yielded the general outlines and orientation of the troponin core domain (20). Rod and mobile domains of TnT and TnI are evident, but the corresponding resolution is insufficient to unambiguously define the key atomic level interactions within molecular envelopes of the reconstructions (20,23). Crystal structures have been obtained for the “core domain” of the troponin complex, resolving TnC, much of TnI, and the C-terminal end of the TnT (24,25). Again, the remaining sections of troponin, namely, the TnT tail and TnI mobile domain, are too flexible to be identified by crystallography (26). Thus, the mechanistic insight about troponin-tropomyosin structure/function is wanting.

The TnT tail

Numerous studies have shown that, although shorter than tropomyosin, the 16-nm-long TnT1 tail aligns lengthwise along actin-based tropomyosin with a 1:1 stoichiometry that specifies the 38.5-nm periodicity attributable to the whole troponin complex on the thin filament. However, the ends of TnT and the termini of tropomyosin do not coincide, and thus, the two elongated molecules are staggered relative to each other (Fig. 1; (6,7,16)). In fact, whereas a small segment of TnT1 binds to the N-terminal end of tropomyosin, the rest traverses the tropomyosin N-/C-terminal overlapping domain and then continues over only ∼60% of the C-terminal part of the adjacent tropomyosin molecule while adopting a sequence arrangement antiparallel to that of tropomyosin. This leaves a troponin-free gap over much of the N-terminal half of tropomyosin before the next TnT binds (Fig. 1, see (7,27)). Models of TnT1 on tropomyosin must be consistent with these structural parameters.

Much of the tail domain of TnT (TnT1) is conserved across evolutionary boundaries. Hence, it is not surprising that tropomyosin isoforms specific to troponin-regulated muscle appear to contain similarly conserved TnT-specific recognition domains on tropomyosin (28), which are absent in troponin-free smooth muscle and cytoskeletal tropomyosin isoforms. N-terminal regions of TnT1 are largely helical, water soluble, and bind to tropomyosin with a high affinity (K_d ∼20 nM) (11,21,27,29, 30, 31). Point mutations or truncation, for example, between residues between 92 and 120 in the human cardiac TnT sequence, perturb such tight binding (6,8, 9, 10, 11, 12, 13, 14, 15). (N-terminal extensions complicate sequence numbering comparison among TnT isoforms; we will use that of human cardiac TNNT2_isoform 6 (P45379-6) here to conform to the numbering in publications of these authors). As mentioned, the TnT tail domain is a well-known hotspot for mutations connected to cardiomyopathies and skeletal muscle disease ((8); reviewed in (6,14)). It follows that residue-specific identification of the amino acids involved in the binding of TnT to tropomyosin will help to characterize deficiencies associated with such mutations (14). Despite these biomedical and biophysical associations, detailed in vitro and in situ, structural characterization of TnT-tropomyosin associations is limited.

Computationally driven docking protocols have now progressed to an extent that outstanding research groups have attempted to generate full atomistic models of the thin filament in silico and thereby address gaps in current experimental approaches (31, 32, 33). In this study, we have built further on these previously proposed macromolecular assemblies. In our work, we have limited docking protocols to predictable single α-helix domain (defined in (34)) helices of TnT, thus restricting localization to the conserved tropomyosin overlapping domain and its surroundings. We have not attempted to dock intrinsically disordered or structurally uncharacterized troponin segments to actin-tropomyosin, which may confound these studies. Furthermore, we did not guide docking to likely but unsubstantiated anchor sites for TnT1 on actin-tropomyosin (see (31, 32, 33)) that might bias outcomes. We therefore broadened our own previously presented preliminary in silico approaches (35) in attempting to determine the regions of the TnT that associate with tropomyosin. Unlike the recently published models (31, 32, 33), the TnT model represented here fits snuggly onto actin-tropomyosin, consistent with tight binding isotherms, and involving well-defined hydrophobic coupling interspersed with numerous salt bridges. We believe that the model generated realistically replicates known energetics of TnT-tropomyosin interaction while predicated on and consistent with reliable moderate resolution structural information.

Although limiting ourselves to docking a single well-characterized segment of TnT1 onto actin-tropomyosin, our work yields structural insights into 1) how the TnT domain enforces the tight binding and polarity of the troponin complex onto actin-tropomyosin; 2) how the domain buttresses the tropomyosin head-to-tail overlap junction to facilitate formation of a mechanically continuous tropomyosin cable on thin filaments; 3) provides a structural blueprint to guide the future assessment of disease-linked TnT mutational hotspots; and 4) describes cardiac-specific crevices at the TnT-tropomyosin interface likely to bind small molecules that could modulate cardiac contractility.

Materials and Methods

Docking studies

Protein-protein docking is a well-established technique for generating atomic level interaction data if structural data is not available or if resolution is insufficient to identify the key residues involved. Successful docking protocols, however, rely on accurate models for the structures of the receptor and the ligand to be docked. The calculations often suggest multiple possibilities with scores that are close to each other; however, high ranking poses generally can be discriminated from each other by the incorporation of biochemical information to eliminate from further consideration poses which are inconsistent with known data. In the case of TnT1 association with tropomyosin, correct docking is expected to feature TnT1 in an antiparallel arrangement to tropomyosin, coupled with interactions with the N-terminus, the overlap domain, and the C-terminus of the tropomyosin molecule (7,16,27). In addition, the docked TnT1 should interact intimately with tropomyosin given their high affinity for each other (27,29,31). In our case, subjective choices were not required because the two top scoring poses fulfilled these requirements, with TnT1 docked to identical, symmetry-related positions on opposite faces of the tropomyosin coiled coil.

The receptor

In the current work, the atomic model of the tropomyosin cable (19), represented by two contiguous tropomyosin dimers joined together head to tail by a centrally located N-/C-terminal overlapping domain, was taken as a “receptor” for docking TnT1 “ligands” in a first level screening protocol. The tropomyosin model (Tpm1.1st (a.b.b.a.), following the nomenclature in (36)) is based on NMR and x-ray crystallography and has been extensively evaluated by molecular dynamics (MD), either in isolation or when linked to F-actin (19,37, 38, 39). Peripheral ends of the construct distal to the overlapping center were not expected to contribute to the TnT1 binding site and, hence, were truncated in the current work so that the model consisted of 340 residues, thereby reducing the computational expense of the workflow; this step did not affect measured outcomes (e.g., when longer tropomyosin sequences were used for validation purposes). Because the TnT1 associates with tropomyosin on actin, docking was also performed with a receptor consisting of the above tropomyosin model and the five associated actin subunits from our previous thin filament model (19) to determine a TnT1 binding site in the context of the thin filament.

The ligand

Different length segments of human cardiac TnT1 between residues 70 and 170 were chosen as ligands because they all have high helical propensity and are known to bind tropomyosin (11,27,29,30,32). This region of the sequence of different TnT isoforms is well conserved phylogenetically (6,40,41), and thus, its study has broad applicability. TnT1 segments were modeled as ideal α-helices using the Discovery Studio Visualizer 2019 software (Dassault Systèmes BIOVIA, http://www.3dsbiovia.com/products/collaborative-science/biovia-discovery-studio/requirements/technical-requirements-2019.html) with amino acid side-chain orientation defined by the most common, low-energy rotamer conformation and the α-helical construct energy minimized.

Note, to be consistent with Palm et al., Tobacman et al., and Williams et al. (11,29,33), we use the sequence numbering scheme for TNNT2_HUMAN-isoform 6 (UniProtKB-P45379) when reporting on docked segments of human cardiac TnT isoform 6. When tested, parallel docking studies of TnT1 segments from rabbit fast skeletal muscle, numbered as in TNNT3-isoform 1 (UniProtKB-P02641-1), yielded virtually the same axial and azimuthal docking (see Results).

Molecular-docking protocol

The programs PIPER and ClusPro (42,43) were used to define likely TnT1-tropomyosin “ligand-receptor” interaction at residue-specific resolution. PIPER and ClusPro were chosen for our analysis (42,43) because the suite of programs has been consistently rated among the best global docking methodologies in the CAPRI challenge (Critical Assessment of Predicted Interaction), particularly when no a priori information on a target complex is known or assumed (43). The programs reliably fitted segments of TnT longitudinally along the tropomyosin coiled coil with antiparallel native orientation. Other programs, such ZDOCK, PatchDock, and GRAMM-X (44, 45, 46), were less successful at these basic tasks, and here, identified TnT poses were less well discriminated from each other.

As an initial screen, a PIPER-initiated global search was first performed to evaluate large numbers (10⁹–10¹⁰) of TnT1 configurations after rotation and translation over an actin-free tropomyosin receptor surface. Configurations were ranked according to attractive and repulsive surface complementarity as well as electrostatic, van der Waals, and hydrophobic interaction potential, either as a balanced metric or separately. After PIPER, ClusPro was applied to ∼1000 of the lowest-energy TnT1-targeted tropomyosin alignments to perform a three-dimensional search for 25–50 groups of closely related “clustered” configurations, reasoning that highly populated structural arrangements most closely identified the global energy minimal of a native binding site. The highest-ranking clusters were then averaged and energy minimized. The ranking was the same when analyzed by electrostatic, van der Waals, or hydrophobic interaction. At this point, the clusters were manually inspected to select the best scoring poses that showed the expected antiparallel alignment and making interactions across the tropomyosin overlap domain. As mentioned above, the highest scoring poses after inspection associated TnT1 longitudinally along tropomyosin with native antiparallel sequence alignment over tropomyosin domains that bind troponin (7,16,27). In fact, the highest scoring poses linked TnT1 to the same residues on tropomyosin for all TnT1 fragments tested whether or not docking, and MD was performed on tropomyosin alone or later on actin-tropomyosin. In contrast, the docking of TnT1 to tropomyosin or to actin-tropomyosin yielded low scoring poses distal to tropomyosin-binding domains or with incorrect polarity; these arrangements were not considered further.

Molecular dynamics

PIPER/ClusPro energy minimization can result in a structure whose final amino acid side-chain orientation may be biased by that in the starting structure. To reduce such bias in the final model and optimize the TnT1-tropomyosin interaction further, tropomyosin-TnT conformers with the top ranked scores were processed additionally by MD for up to 200 ns with NAMD (47), utilizing the CHARMM36m force field (48) as previously described (38). Generally, modest improvement in ligand-receptor side-chain complementarity and local annealing of TnT helices were observed during MD.

Finally, 40 ns or greater MD simulations were carried out on the highest scoring poses of TnT now docked by PIPER/ClusPro to F-actin-tropomyosin. In this case, the tropomyosin cable on actin was formulated using periodic boundary conditions to create an essentially infinite thin filament with end-to-end bonded tropomyosin wrapped continuously around F-actin (hence, mimicking thin filament periodicity in silico). The initial model was derived from the cable model previously published (19). This model was expanded to comprise 28 actin monomers and eight tropomyosin chains. This unit repeats in both directions along the filament axis by simple translation (using actin helical parameters of 27.5 Å translation and 167.1° rotation), therefore fulfilling an unbroken periodic boundary condition. The resulting model was solvated using the cionize, solvate, and ionize plugins in VMD to add water, 0.15 M sodium chloride, and 3 mM magnesium chloride before MD simulations (49). Throughout the simulation, buried actin residues near the filament axis were harmonically constrained to maintain the integrity of the filament structure. Simulations were performed using NAMD (47). Heating to 300 K and constrained equilibrations were performed under constant volume, with a gradual release of constraints on the tropomyosin and surface actin residues. To bring the pressure of the system close to one bar, a short 50,000 step run was performed under constant pressure using the Langevin piston, allowing each axis of the solvent box to change independently. The system was inspected at this point to insure there were no gaps in the solvent or significant changes in the filament axis length. Production runs were then performed under constant volume. This method keeps the repeat of the actin filament close to the starting value so that the periodic boundary does not compress or elongate the filament while bringing the pressure in the system up to a standard value.

In all cases, MD was performed using explicit solvent.

Results

Assessing thin filament models

Relatively recent models by Manning et al. (32) or by Gangadharan et al. (31) describe the sequence alignment and the conformation of TnT interacting with an unusual tropomyosin overlap structure. The models are predicated on a crystal structure from the Murakami group of end-to-end linked N- and C-terminal tropomyosin peptides that were cocrystallized with TnT fragments (50) (Protein Data Bank, PDB: 2Z5H). In the same study, examination of crystal structures of N- and C-terminal tropomyosin fragments (no TnT) (PDB: 2Z5I) uncovered an unexpected N-/C-terminal tropomyosin overlap complex containing an N-terminal dimer that splays into single helices (50). Each single chain of the splayed helices interleaves the dimeric ends of two separate C-terminal peptides to form two three-helix complexes; thus, the asymmetric unit contains an N-terminal dimeric handle of a “two-pronged fork,” in which each prong is represented by a three-helix N-/C-terminal overlap zone, leading into the tail of two C-terminal dimers and an additional six C-terminal peptides that do not make overlap zones. When crystallization trials also included TnT1 fragments (50), TnT bound to the side of the three-helix overlap to form a four-chained bundle, now containing the additional TnT chain (PDB: 2Z5H; Fig. S1 A). Such forked tropomyosin structures are not normally observed experimentally or in situ (compare to (7,16, 17, 18,51)). However, end-to-end binding promiscuity is common among tropomyosin fragments, leading to several examples of non-native tail-tail and head-head tropomyosin associations (28,52) and, in this case, bifurcation.

Although electron density maps used to build the Murakami structures revealed strong C-terminal coiled-coil signal, less convincing, weak helical densities were attributed to the ends of N-terminal tropomyosin and associated TnT peptides; modeling of the N-terminal and TnT densities results in a structure with poor B-factors and poor side-chain and backbone rotamers (50). Thus, the tropomyosin-TnT structure deserved further inquiry to evaluate its merits as a reference in our own work or in the prior computational work performed by others. Hence, we subjected the Murakami TnT-tropomyosin model used by Gangadharan et al. (31) to standard molecular dynamics (MD) simulation in explicit solvent as a test of its conformational stability. Conspicuously, during simulation, the TnT component dissociated from the complex, along with orthogonal contortion of the tip of the N-terminal chain of tropomyosin in relation to the rest of the N-terminal helix and the C-terminal partner (Fig. S1, A–C). In contrast, molecular outcomes of lengthy MD simulations performed on more conventional four-helix bundled N-/C-terminal tropomyosin overlapping structures and on isolated TnT1 helices are stable and unchanged during simulation (Fig. S1, D and E; compare to data in (37,38)). Because models of the Murakami N-/C-terminal tropomyosin overlapping structure and its interaction with TnT are unprecedented (7,16, 17, 18,51) and are unstable in silico, we have discounted them without further experimental evidence as possible crystallization artifacts, resulting from the use of foreshortened engineered tropomyosin constructs attached to non-native GCN4 leucine zipper extensions. Thus, the validity of ever more complex models (e.g., Manning et al. (32); Gangadharan et al. (31)) derived from such crystallography is uncertain. In the latter model, forbidden clashes between component helices were left uncorrected, whereas the arrangement of the overlapping N- and C-terminal coiled-coil chains, stacked one on top of the other (31), is inconsistent with the orthogonal orientation found experimentally (17,18) and which is stable during MD (51). We have additionally observed that overall close apposition of TnT and tropomyosin as well as molecular clashes noted in the Gangadharan model yields unfavorable interaction energy terms (Table 1).

Table 1.

Interaction Energetics between TnT1 and Tropomyosin

Model	Interaction Energy between TnT and Tropomyosin (kcal/mol)			Buried Interface between TnT and Tropomyosin (Å2)
	Electrostatic	van der Waals	Total
Cardiac model	−829.8 ± 90.4	−81.6 ± 6.8	−911.4 ± 92.3	2295.1 ± 145.4
Skeletal model	−1012.6 ± 75.6	−80.7 ± 7.4	−1093.31 ± 76.0	2447.4 ± 127.58
Williams model	−762.8	−1.5	−764.3	809.1
Gangadharan model	−1789.1	+1011	+1011	4616.8
Cardiac model with TnT mutants
A104V	−897.9 ± 63.9	−98.91 ± 7.0	−996.79 ± 65.2	2763.2 ± 202.6
F110I	−636.6 ± 61.2	−59.95 ± 4.9	−696.51 ± 61.8	1863.0 ± 144.9
R130C	−633.4 ± 76.8	−83.10 ± 7.1	−716.46 ± 78.3	2421.0 ± 128.1
R141W	−700.7 ± 89.6	−97.82 ± 6.6	−798.54 ± 91.8	2775.2 ± 135.0

Interaction energetics were computed between TnT1 and tropomyosin for the complex when linked to F-actin. Data were averaged for the last 10 ns of 40-ns molecular dynamics simulations using the NAMD protocol in VMD (49). The buried surface between TnT1 and tropomyosin was calculated by subtracting the exposed surface area for the TnT1 complex on F-actin-tropomyosin from the total surface of the two components in isolation. All calculated values (±SD) are standardized for the same 62-residue stretch of TnT1 (89–151 for cardiac and 79–141 for skeletal muscle) on actin-tropomyosin using our residue numbering convention. Values for the Williams and Gangadharan models were computed directly from their reported pdb coordinates (31,33). Binding energetics calculations for mutant TnT1 on actin-tropomyosin were done in the same way as for control cardiac TnT after 40-ns MD simulations of mutant TnT1 on actin-tropomyosin, in which single amino acids were substituted for wild-type ones in VMD (49), as previously described (37,39).

Recognizing that “the structural understanding” of the tropomyosin overlapping domain “has evolved over time,” Williams et al. (33) built a new full atomistic thin filament model, based in part on newer Förster resonance energy transfer measurements as well as recent structures by Li et al. (38), Orzechowski et al. (19), and Yang et al. (20). The new model diverged from their previous model (32), which relied more strictly on Murakami et al. (50). However, the axial alignment between TnT1 and tropomyosin helices may still reflect possibly spurious residue-residue anchoring derived from Murakami et al. (50). Still, detailed examination of the new Williams model reveals wide TnT1-tropomyosin separation (Fig. S1 F; Table 1) and sparse Coulombic and hydrophobic contact over the interface between the tropomyosin overlapping region and TnT1 (Table 1), in apparent conflict with reports of a 20-nM intermolecular K_d between the two (27,29,31). We therefore also subjected the Williams model to standard MD simulation, during which segments of the TnT1 separated even further from tropomyosin than in the starting model (Fig. S1 G). Clearly, this region of the Williams model, and in particular, that over the tropomyosin N-/C-terminal overlapping domain, required further study.

An unbiased approach to docking TnT onto actin-tropomyosin

The paucity of explicit residue-residue alignment information between TnT and tropomyosin makes anchor-driven MD or steered docking an impractical approach to fitting TnT1 onto thin filaments (as in (32,33)). However, given the power of current computational resources, global docking of TnT1 onto actin-tropomyosin by brute force means seems a plausible alternative, thus allowing a very large sampling of conformational space available to find the lowest-energy structure. Still, the reliability of protein docking protocols, no matter the size of the undertaking, depends on accurately described ligand and target structures. In our case, TnT-free coiled-coiled tropomyosin and its N-/C-terminal overlapping domain structures are now well characterized (17, 18, 19,37,51), hence acceptable docking targets. In contrast, significant regions of the troponin complex structure, including flexible TnT1, have not been solved experimentally by x-ray diffraction, NMR, or cryoelectron microscopy (22,26). Nevertheless, circular dichroism measurements show TnT fragments covering much of human cardiac TnT1, namely between residues 70 and 170, are 90% α-helical (11); comparable results were confirmed here and originally identified as helical by Pearlstone et al. (30). In addition, a segment of cardiac TnT1 isoforms from residue 101 to 163 and a corresponding sequence in much-studied rabbit fast skeletal muscle (residues 91–153 in TnT isoform 1) are well conserved as are human cardiac and rabbit skeletal residues 77–94 and 67–84 (27). In fact, helical propensity programs indicate 100% helicity for a long segment stretching from cardiac TnT residues 89–163 (32), and MD simulations over residues 105–163 performed here of the corresponding peptides are, without interruption, α-helical throughout our 200-ns simulation (Fig. S1) as they were in earlier short MD simulations (32). As mentioned, TnT1 is soluble in aqueous solvent and binds to tropomyosin with a high affinity (21,27,29,31). Therefore, the ligand for docking studies was built by imposing α-helical symmetry on backbone α-carbon chains derived from varying length cardiac TnT1 peptides encompassing residues between 89 and 163 (TNNT2_HUMAN-isoform 6 sequence numbering used, UniProtKB - P45379) because this helical segment of the TnT1 conformer is well behaved experimentally and therefore displays an unique and identifiable structure well suited for global docking routines, despite the lack of direct structural representation for TnT1.

PIPER/ClusPro docking of TnT1 to tropomyosin

We concentrated efforts to dock fragments of cardiac TnT onto tropomyosin to the full span of residues from positions 89 to 181 and to shorter spans ranging from 89 to 163, 104 to 151, 89 to 128, and 89 to 151 using PIPER/ClusPro docking followed by cluster analysis of the docked TnT1 segments to dimeric and overlap regions of tropomyosin to identify the most likely favorable interactions. All docked as α-helices with high scores to the same region on tropomyosin, including the N-/C-terminal overlapping and C-terminal domains of tropomyosin. The highest scoring TnT1 fragment, consisting of residues 104–151, docked over the tropomyosin overlapping domain with a directionally opposed antiparallel sequence alignment to tropomyosin and continued along tropomyosin C-terminal residues up to Asp230. A second fragment (residues 89–128) favored by Jin and Chong to bind tropomyosin (27) also yielded high scores, and the two docked structures superposed seamlessly over tropomyosin residues common to both poses. The TnT1 segment, representing residues 89–151, aligned continuously and intimately alongside tropomyosin, matching residue-residue-specific sequences held in common with the segments comprising 104–151 and 89–128. Virtually identical scores were obtained for poses of TnT docked either to the convex or the concave side of superhelical tropomyosin. C-terminal stretches of TnT1 mentioned above and modeled as helices but ranging beyond residue 151 did not align as closely alongside tropomyosin; hence, we focused on residues 89–151. The longer TnT fragments contain a 15-aa-long series of residues that is highly charged; positions 154–163 have an EERARREEEENRRRK sequence. Analyzing simulations of the docked structure containing these residues showed occasional TnT dissociation from tropomyosin. This suggests that, absent the rest of the troponin structure, residues 154–163 may have been attracted to solvent rather than fixed to tropomyosin (53), under our conditions. We note a second tropomyosin-binding site closer to TnT2 (27) may be needed to hold this transitional region containing nonhelical domains of TnT in place over tropomyosin.

Development of asymmetrical sidedness of TnT1-tropomyosin interaction

Although tropomyosin is slightly bent over relatively short stretches (0.7°/residue), the tropomyosin dimer has a pseudo-twofold symmetry around its superhelical axis; thus, TnT segments dock to the same residues on one or the other side of the coiled-coil dimer with common high scores. This virtual 2:1 TnT/tropomyosin docking propensity matches the stoichiometry sometimes observed experimentally for isolated proteins (7). In marked contrast, high scoring TnT1 segments dock to just one preferred side of actin-bound tropomyosin, when the tropomyosin target is part of the actin-tropomyosin complex. The second side is obscured by tropomyosin location on the actin substrate that competes for TnT1 docking, in particular, where tropomyosin abuts a ridge on actin near to subdomain 1 residues D24 and D25. Hence, here, the single-sided docking agrees with the 1:1 TnT/tropomyosin stoichiometry found for intact thin filaments (54). The residue-residue position of TnT on actin-based tropomyosin is the same as that on actin-free tropomyosin.

Molecular dynamics simulations on TnT1-tropomyosin structures

Although PIPER and ClusPro scoring assesses energy minimized structures, the programs’ accounting of conformational flexibility during docking calculations is limited. To account for the dynamics about the docked structures, we therefore carried out MD on the top ranked poses of cardiac TnT1 residues 101–151, 89–151, and 89–128 linked to isolated tropomyosin and/or to actin-tropomyosin. No shifts were noted in the axial or azimuthal positioning of the TnT fragments docked either onto tropomyosin or onto tropomyosin associated with F-actin. And, in the latter, no shifts in axial or azimuthal positioning of tropomyosin relative to F-actin occurred during MD. However, during MD, modest repair of initial ClusPro minimized structures occurred, and side-chain orientations readjusted accordingly, without other changes.

Comparison to previous models

In contrast to the Williams et al. (33) model of the thin filament that shows a relatively wide gulf between tropomyosin and cardiac TnT1, intimate interaction is observed between our docked TnT1 on tropomyosin and on actin-tropomyosin before or during MD. This is especially apparent when cardiac TnT1-tropomyosin was associated with F-actin; here, the actin substrate evidently sustains tropomyosin curvature and thus facilitates constant one-sided TnT1-tropomyosin interaction during MD (Fig. 2). Extensive interactions between TnT1 and tropomyosin are observed involving hydrophobic packing between TnT residues Leu102, Ala104, Leu105, Ile 106, and Ala108 and the tropomyosin overlapping domain. Nonpolar TnT residues Leu120, Leu123, and Ile127 also flank neighboring C-terminal residues of tropomyosin. This arrangement results in the formation of an elongated and deep hydrophobic pocket and cleft between the two molecules (Fig. 2; in Figs. 2, 3, and 4, results of docking and MD on cardiac TnT residues 89–151 are shown). A series of charged residues on TnT (Arg94, Glu101, Glu116, Glu119, Arg126, Arg130, Arg134, Glu136, Arg144, and Arg148) border the two hydrophobic patches to form closely spaced salt bridges with oppositely charged residues on tropomyosin (Fig. 3). Docked TnT1 fragments do not interact directly with the F-actin substrate along their length, except possibly for the fluctuating C-terminal tip of the longest ones tested.

Figure 2.

Docking followed by molecular dynamics of cardiac TnT1 on actin-tropomyosin. In Fig. 1, a peptide representing residues 89–151 of cardiac TnT1 (yellow) is shown docked onto tropomyosin strands linked to F-actin by our methods (pale blue). (A) A magnified view of Fig. 1 shows the results of docking the mid-piece TnT1 segment containing residues 89–151 onto actin-tropomyosin and its association with the overlap junction of tropomyosin. The N-terminus of the TnT segment interacts with the N-terminal end of tropomyosin (pink), bridges over the C-/N-terminal four-helix overlap domain (marked by brackets), and then extends over the C-terminus of tropomyosin (magenta) in the direction of the pointed end of the thin filament. Acidic and basic residues forming salt bridges between TnT1 and tropomyosin are colored red and blue, respectively; hydrophobic interactions are also shown (white). (B) Shown is a ribbon representation of the TnT1 segment and its interactions with tropomyosin, now with actin hidden and rotated for easier inspection. Brackets mark tropomyosin overlap domain. (C) Shown is a surface rendering of the structure in (B). Arrows mark solvent-exposed pockets between TnT1 and tropomyosin. (D and E) are 180° rotations of (B and C). No attempts were made to dock N-terminal residues 1–89, in the so-called hypervariable region of TnT1, or C-terminal residues leading into the troponin core domain. To see this figure in color, go online.

Figure 3.

Contact maps between tropomyosin and either cardiac or skeletal muscle TnT1. (A) Cardiac TnT1-tropomyosin is shown. (B) Skeletal muscle TnT1-tropomyosin is shown. TnT1-tropomyosin residue-residue alignment is constant after 5 ns of simulation, whereas side-chain contacts typically are dynamic. Contacts shown are taken from energy-minimized averaged MD trajectories of actin-tropomyosin-TnT1 structures recorded over the last 10 ns of 40-ns simulations. Red lines indicate hydrogen bonding as well as salt bridges; violet lines show van der Waal’s contacts (a 5-Å distance cutoff criteria between side chains was applied). The upper tracing in each diagram shows the C-terminal sequence of tropomyosin (Tpm1.1 residues 230–284), and the lower tracing show the N-terminal ones (Tpm1.1 residues 1–20) (i.e., residues in contact with TnT1); note that the extreme N- and C-terminal tropomyosin residues (1–9 and 275–284) interleave to form the tropomyosin overlap. The middle row in each set represents TnT1. The different sequence numbering for cardiac and skeletal TnT relates to the presence of a “hypervariable” N-terminal cardiac TnT extension absent in skeletal muscle TnT (6). Note that residues 89–151 of human cardiac isoform-6 (A) and residues 79–141 of rabbit fast skeletal muscle isoform-1 (B) are almost identical in sequence to each other and in alignment with tropomyosin, with a few conservative substitutions (underlined). Thus, our docking shows both fragments associate with tropomyosin in an antiparallel sequence arrangement and onto the same region of actin-free and actin-associated tropomyosin. However, the diagram indicates locally diminished intimate contact linking cardiac TnT1 to a stretch of tropomyosin from Tyr261 to Leu278, whereas this region is more highly connected in the skeletal muscle TnT1 model (compare with Fig. 4). To see this figure in color, go online.

Figure 4.

Comparison of cardiac and skeletal muscle TnT1 tropomyosin interaction. As above, these structures are taken from averaged, energy-minimized MD trajectories over the last 10 ns of simulation of TnT1 docked onto actin-tropomyosin. (A and B) Shown are paired ribbon diagrams and surface views of the skeletal muscle TnT1 segment (cyan) docked to actin-tropomyosin; shown in a view similar to that of cardiac TnT1 and tropomyosin in Fig. 2, B and C with actin hidden and tropomyosin colored tan and coral. Note that there is more intimate contact between skeletal muscle TnT1 and tropomyosin (C) than for cardiac TnT1 and tropomyosin (D) and that the pocket between TnT1 and tropomyosin (arrows) therefore is diminished in size. To see this figure in color, go online.

Corresponding docking and MD analysis of skeletal muscle TnT1-tropomyosin interaction shows that distances between oppositely charged residues are slightly closer (Fig. 4; Table 1; here, TnT residues 79–141 were docked) and that the hydrophobic pocket in the skeletal muscle example is diminished in its dimensions. Side-chain variance in what appear to be conservative amino acid differences between the two isoforms of TnT (A112N, A122S, E125D, and K129R) may be responsible for differences noted (Fig. 3; Table 1). We also note that the presence of TnT enhances the interaction energetics of tropomyosin for actin. Here, the total interaction energy (electrostatic + van der Waals) between tropomyosin and actin was enhanced by ∼12–13% by either cardiac or skeletal muscle TnT (TnT-free actin-tropomyosin (−4638 ± 196 kcal/mol) versus TnT-associated actin-tropomyosin (−5194 ± 189 and −5263 ± 278 kcal/mol for cardiac and skeletal muscle cases)). Corresponding measurements on the Williams et al. model (33) showed weaker actin-tropomyosin interaction energetics than derived from our model of TnT1-free actin-tropomyosin, and the presence of TnT1 did not appreciably enhance actin-tropomyosin binding (∼2%). These values improved when we carried out extensive MD (up to 200 ns) on the Williams model but never reached those measured for our model.

Interaction energy assessment

Interaction energies were computed between TnT1 residues 89–151 and tropomyosin as well as for the same segments of well-characterized mutant TnTs (F110I, R130C, and R141W) (Table 1). In each case, values for mutant Coulombic and van der Waals interactions showed less favorable energetics. The increased solvent-accessible surface, indicating diminished interfacial complementarity and potential hydrophobic interaction between the mutants and tropomyosin also indicated weaker binding strength. In contrast, A104V showed localized helical unfolding (data not shown) and more intimate contact between N-terminal residues of the TnT fragment assayed and tropomyosin, thus accounting for the apparent enhanced binding energetics.

Discussion

Lacking experimentally determined high-resolution maps of troponin bound to native or reconstituted thin filaments, we and other groups have attempted to model the complex on F-actin-tropomyosin computationally (20,31, 32, 33,55). In the current study, the application of unbiased global fitting protocols, PIPER and ClusPro, was limited to docking a section of TnT1 (cardiac residues 89–151, skeletal muscle residues 79–141) with well-established helicity onto tropomyosin and actin. This conservative approach avoided fitting unstructured domains or ones that interact weakly with actin or tropomyosin. We also avoided reference bias linked to the use of uncertain and/or unsubstantiated TnT1-tropomyosin crystal contacts that would have restricted the conformational space surveyed. Moreover, the conformational exploration afforded here by PIPER and ClusPro is extensive and not limited in direction by steered docking methodology (32,33).

Our work is most easily distinguished from prior attempts to dock TnT alongside tropomyosin by taking note of the varying description of its lengthwise separation from tropomyosin in the reported models. On one extreme, the Gangadharan et al. model (31) places TnT1 in very close apposition to tropomyosin, causing forbidden residue-residue clashes, leading to unfavorable energetics (see computation of van der Waals energies in Table 1). Conversely, models of Manning et al. (32) and Williams et al. (33) place TnT1 at an unrealistically large distance from actin-tropomyosin with little chance for electrostatic and hydrophobic interaction (see computation of respective buried interface and van der Waals interactions in Table 1). Molecular dynamics simulations carried out here suggest that these models have not been optimized sufficiently to document definitive structural interactions and positioning. By contrast, in our own characterization, TnT assumes a radial position on thin filaments in between those previously proposed, in which numerous stable salt bridges and energetically favorable hydrophobic interactions of tropomyosin are possible.

Consistent with previously elucidated experimental work on TnT1-tropomyosin interaction, our docking studies indicate that binding links a midpiece stretch of TnT1 (here, residues 89–151) in human cardiac TnT to the tropomyosin N-/C-terminal overlapping domain as well as to adjacent dimeric coiled-coil regions of tropomyosin. This alignment of TnT to tropomyosin agrees with the recent Williams et al. (33) modeling, showing the same antiparallel arrangement as well as axial residue-residue register of TnT1 with the tropomyosin at the overlap domain. However, the radial separation, axial angle, azimuthal position, and rotation of TnT1 on tropomyosin at or peripheral to this site and therefore along actin-tropomyosin differs significantly, resulting in the divergence of residue-residue register at a distance. In the Williams model, few residues on TnT1 form salt bridges that link TnT1 and tropomyosin. In contrast, our model displays closely packed salt bridges and is replete with hydrophobic interactions. Moreover, the measurement of the buried surface, an indication of surface complementarity, is favorable in our model.

Our molecular docking describes the formation of an interleaving quasi-five-helix bundle that appears to couple the tight binding of TnT on thin filaments to the stabilization of the tropomyosin overlapping domain and, hence, to the maintenance of the polymeric cable. Within a ∼25-Å span, hydrophobic residues between cardiac TnT residues M95 and I106 intermesh with the N-/C-terminal overlapping domain residues M1–L13 and L278 of tropomyosin. Flanking this region, multiple electrostatic interactions link TnT residues R113–R134 to tropomyosin residues D275–E250 over an additional 20–25 Å. The model contains helical segments of TnT1, which run alongside C-terminal tropomyosin pseudorepeats in the direction of the pointed end of the actin filament. In this way, TnT1 can act as a molecular ruler for the entire regulatory complex on tropomyosin with a specificity that defines the polarity of the rest of the troponin complex on the filament.

Breaks in the continuity of the α-helical patterning of TnT1 occur beyond residue 170 on the C-terminal end of TnT1. Here, relatively unstructured connectivity between TnT1 and TnT2 may relieve TnT from strictly following the tropomyosin superhelix. A jointed multipartite organization (as drawn schematically in Fig. 1) or flexible region of TnT1 could thereby relax the peripheral localization of the troponin tail on tropomyosin. Such an arrangement would also leave the interface between actin and tropomyosin unencumbered by TnT1, allowing unobstructed regulatory translation of tropomyosin-TnT over actin. If, in contrast, a continuously helical 16-nm TnT1 strictly followed superhelical tropomyosin, it presumably would come in contact with the actin surface and interfere with actin-tropomyosin interaction while potentially restricting the regulatory movement of tropomyosin over the thin filament surface.

Previously proposed, short helical fragments may define the TnT1-TnT2 junction near tropomyosin residue 175 (33,35,53). It follows that TnT makes secondary contact with tropomyosin near to this junction and, then, via TnT2 connects to the troponin core (20,24,25,27,56). The very tight association between TnT1 and tropomyosin taken together with additional, but still structurally uncharacterized, tropomyosin-TnT2 core domain connectivity limits the diffusion of troponin away from thin filaments, thereby maximizing local interactions of the troponin core domain extensions on the thin filament. Thus, thin filament function relies on a balance between highly specific tight structural interactions that anneal TnT1 to polymeric tropomyosin cables and weaker transient binding processes that allow tropomyosin regulatory movements on actin while coupled to Ca²⁺-dependent TnI-TnC and TnI-thin filament controls.

Weak versus tight regulatory subunit binding on thin filaments

The three-state steric-blocking model provides a general framework for understanding the changing equilibria among thin filament components that control the binding and force the generation of myosin heads on actin (57,58). Here, tropomyosin coiled coils, associated with groups of seven successive actin subunits along thin filaments, translocate azimuthally over thin filaments during muscle activation and relaxation. In relaxed muscle, tropomyosin is “pinned” by troponin in a blocked B-state that interferes with the access of myosin heads to actin on the filaments. After muscle activation, Ca²⁺ binds to TnC of the troponin complex, resulting in the C-terminal mobile domain of TnI dissociating from actin-tropomyosin, whereas its “switch peptide” fits into TnC’s N-lobe (reviewed in (22)). The TnI transition is thought to “unpin” tropomyosin and relieve the steric hindrance. It follows that tropomyosin then moves on actin to a “closed” C-state, still inhibitory but permissive to weak actin-myosin interaction. The myosin-head interaction with actin facilitates cooperative movement of tropomyosin to an “open” M-state now involving the binding of strong (i.e., force-generating) myosin cross-bridges on actin (reviewed in (2,3)). Although these regulatory steps are easily conceptualized as identifiable configurations or as discrete kinetic states, they actually rapidly equilibrate with each other while being biased one way or another by the interplay of troponin, Ca²⁺, and myosin interactions (58,59).

Controlling muscle activation and relaxation is made possible by transient and reversible weak binding interactions between thin filament regulatory proteins and actin. For example, low-affinity binding between single tropomyosin dimers and actin (K_d ∼ 3 mM) (60) allows the tropomyosin cable, once polymerized on thin filaments, to translocate between muscle on- and off-states at low-energy cost. At low Ca²⁺, tropomyosin position on actin is biased by low-affinity association of the C-terminal TnI “mobile domain” (23,61) to actin-tropomyosin, pinning tropomyosin in the myosin-blocking configuration. In turn, at high Ca²⁺, the Ca²⁺-laden N-lobe of TnC and the TnI “switch peptide” associate transiently (K_d is in the 25–150 μM range) and compete with actin for TnI, thereby relieving the steric constraint on myosin-actin association (62,63). Despite this network of low-affinity protein-protein exchange, the close topological organization of the thin filament components ensures the effectiveness of the regulatory transitions by minimizing diffusion of the interacting components. Similarly, transient contacts made by the tip of myosin-binding protein C, otherwise largely tethered to thick filaments, may modulate the behavior of troponin or tropomyosin on closely apposed thin filaments.

Unlike interactions of troponin core and mobile domains with actin, the troponin TnT1 tail binds F-actin-tropomyosin with an exceedingly high affinity (20 nM K_d) (27,29,31), consistent with playing a completely separate, more structural role than exhibited by the rest of the regulatory complex. The proposed atomic models of TnT1 on actin-tropomyosin are consistent with such binding characteristics. Here, targeted interactions of helical domains on cardiac and skeletal muscle TnT1 to specific regions on successive tropomyosin dimers along the thin filaments are ensured by numerous salt bridges and hydrophobic binding. Our characterization of the interaction of TnT1 segment studied here with tropomyosin suggests that it contributes considerably to the corresponding tight binding but cannot realistically ascribe the extent to which other segments of TnT share in the binding, including the as yet uncharacterized, second binding site, at the distal end of TnT (27,64). Whether or not binding of TnT to tropomyosin occurs as a single concerted event or in multiple independent steps is also uncertain.

Conclusions

Four significant outcomes define our work, as follows: 1) We established a likely TnT1 interaction site on actin-tropomyosin and, in particular, its position over the N-/C-terminal tropomyosin overlapping domain. This result sets the register and periodicity of rest of the troponin complex over actin-tropomyosin. 2) We defined the TnT1-tropomyosin interaction at an atomic level. Thus, regulatory imbalances imposed by disease-rendering point mutations can be detailed, as begun here, potentially connecting early stage molecular insults to cardiomyopathy and skeletal muscle pathology. 3) We identified potential pockets that form at the interface of TnT1 and tropomyosin that might be targets for drug binding. 4) We provided a foundation for further energy landscape characterization to describe the structural energetics of tropomyosin cable translation across actin filaments. Our future goal is to refine our models and assess our predictions experimentally.

Author Contributions

W.L. thought of the general approach taken and, along with M.J.R. and E.P., formulated the principal concepts presented. E.P. carried out the docking studies, and E.P. and M.J.R. performed the molecular dynamics simulation and data computation. A.G. performed circular dichroism measurements. W.L., E.P., and M.J.R. analyzed the data. W.L. wrote the manuscript. W.L., E.P., and M.J.R. prepared the figures and tables.

Editor: Steven Rosenfeld.