Structure and Dynamics of the Flexible Cardiac Troponin T Linker Domain in a Fully Reconstituted Thin Filament

Abstract

The structural analysis of large protein complexes has been greatly enhanced through the application of electron microscopy techniques. One such multiprotein complex, the cardiac thin filament (cTF), has cyclic interactions with thick filament proteins to drive contraction of the heart that has recently been the subject of such studies. As important as these studies are, they provide limited or no information on highly flexible regions that in isolation would be characterized as inherently disordered. One such region is the extended cardiac troponin T (cTnT) linker between the regions of cTnT which have been labeled TNT1 and TNT2. It comprises a hinge region (residues 158–166) and a highly flexible region (residues 167–203). Critically, this region modulates the troponin/tropomyosin complex’s position across the actin filament. Thus, the cTnT linker structure and dynamics are central to the regulation of the function of cardiac muscles, but up to now, it was ill-understood. To establish the cTnT linker structure, we coupled an atomistic computational cTF model with time-resolved fluorescence resonance energy transfer measurements in both ±Ca²⁺ conditions utilizing fully reconstituted cTFs. We mapped the cTnT linker’s positioning across the actin filament, and by coupling the experimental results to computation, we found mean structures and ranges of motion of this part of the complex. With this new insight, we can now address cTnT linker structural dynamics in both myofilament activation and disease.

Graphical Abstract

INTRODUCTION

Within structural biology, a central dogma is that function depends on the structure. In recent years, however, it has also been understood that portions of proteins that do not adopt a fixed secondary structure are critical to function. These so-called intrinsically disordered proteins are often central to function. P53 is a tumor suppressor that is at least partially disordered until it interacts with its target, and mutations can result in significantly greater susceptibility to cancers and poorer overall outcomes. The cardiac thin filament (cTF) is a complex molecular machine that contains both structured and unstructured regions. It is a multiprotein complex composed of filamentous actin (F-actin), tropomyosin (Tm), and the troponin (cTn) complex in a 7:1:1 molar ratio. The cTn complex anchors Tm to F-actin and is a trimeric complex comprising the Ca²⁺ regulatory binding protein, troponin C (cTnC), the inhibitory subunit, troponin I (cTnI), and the Tm-binding domain, troponin T (cTnT) in a 1:1:1 molar ratio.¹ The cTF’s basic function is to transduce chemical signals throughout the myofilament complex and directly regulate the actomyosin cross-bridge cycle.^2,3 Specifically, upon cTnC binding Ca²⁺, allosteric changes occur, releasing the cTnI inhibitory domain^4,5 and azimuthally shifting Tm to the inner domain of actin, which exposes myosin binding sites and allows the cross-bridge cycle to ensue.⁶ cTnT is at least partially the effector of the signal transduction from Ca²⁺ binding, and in the center of this protein, there is a region which connects two structured regions and is itself highly mobile and thus unresolvable in previous structural interrogations but central to the regulatory function of the thin filament. Because the cTF is a highly cooperative and dynamic machine where primary function depends on allosterically mediated changes, to understand the allosteric mechanisms of the cTF, an intact high-resolution structure of the full complex is required.

The current picture of the cTF began with Takeda et al.’s 2003 landmark study⁷ of the cTn core in saturating calcium conditions. However, this model only included fragments of cTnC, cTnI, and cTnT found within the cTn core, resulting in a partial cTF structure. This initial cTF core structure was then confirmed and expanded upon by subsequent studies^8-22 to include actin, Tm, and additional fragments of cTnC, cTnI, and cTnT. These comprehensive studies provided crucial insight into the molecular mechanisms of this multi-subunit machine; however, despite these multifaceted approaches, a complete cTF structure remained elusive. Specifically, there are cTF regions where the structure and precise nature of the protein–protein interactions remain unknown.

One such region is the extended cTnT linker (~50 amino acids) between two structured regions of cTnT referred to as TNT1 and TNT2. It is known to be essential because mutations in this area result in highly penetrant, severe forms of the disease. However, a picture of this domain remains structurally obscure due to inherent high flexibility in the region.^7,23,24 The cTnT linker is composed of a highly charged hinge region (residues 158–166) and a flexible region (residues 167–203).²⁵ It is located immediately distal to the extended Tm–cTnT binding domain⁸ and modulates the cTn–Tm complex’s position on actin, affecting the efficiency of cross-bridge cycling and thus the primary function of the cTF.²⁶ Alterations in the linker’s structure, position, and flexibility certainly affect both inter- and intraprotein interactions, significantly altering myofilament activation. As evidence for this, highly penetrant mutational hotspots in this evolutionarily conserved region have been linked to severe forms of both hypertrophic and dilated cardiomyopathies.^27-30 The current defined structure of the domain is summarized in Figure 1. Until the cTnT linker’s structure and dynamic range are obtained, our ability to understand the role of this domain in cTF activation and the molecular pathogenesis of cardiomyopathy causing mutations in this region remains significantly limited. The interactions of the linker region of cTnT with tropomyosin may be viewed as a paradigm of the effect of both normal and mutated highly mobile protein segments in a complex protein machine (unstructured is a misnomer applied due to limitations of extant experimental techniques).

Figure 1.

Structure of the cTF highlighting the conformations of the cTnT linker: Computational model of the human cTF: troponin complex (cTnT: teal; cTnI: pink; cTnC: golden brown); Tm dimer (green and orange); and actin monomer (transparent). (A) Original computational model incorporating Yamada et al.’s refined structural components. (B,C) The computational model was constrained to the measured FRET distances (Table 1) and then allowed to evolve unconstrained to obtain the cTnT short (B) and long (C) linker representations. The black spheres show the locations of cTnT FRET labels. All models are for the +Ca²⁺ condition.

It was not until 2014 with improvements in cTF staining and electron microscopy (EM) imaging processing by the Lehman group that an electronic density could be attributed to the TNT1 tail (residues 1–156) running adjacent to Tm.^14,18 With the TNT1 and TNT2 regions structurally mapped, we could assume that the linker would lie between these two regions within structural models. Most recently, the cryo-EM study by Yamada et al. resulted in a structure of the cTF in both ±Ca²⁺ conditions.¹⁰ For the first time, this model suggested a short and long cTnT linker structural configuration; however, the linker’s three-dimensional structure remained unresolved.⁸ Since Yamada’s structure was published, the results have been challenged, revised, or confirmed.^8,9,11,12 Importantly, Risi et al. also proposed short and long cTnT linker conformations to explain the observed difference of 48 Å (residues 151–198) versus 82 Å (residues 150–199) in two different cTn complexes; however, here, again, there was a lack of structural data to fully explain this observation.^8,9 Tobacman recently suggested that for thin filament activation to occur, the cTn core must attach to the same tropomyosin region, pseudorepeat 4, resulting in a short and long cTnT linker configuration.⁸ The short and long cTnT linker conformations proposed, however, were based solely on the location of the cTn core and not on experimentally obtained cTnT linker structural data. While these studies have refined the cTnT linker proximal location, even with these advances, the overall cTnT linker structure remains unrefined.

To develop a more complete understanding of thin filament structure and dynamics, our group has previously coupled an atomistic computational cTF model with in vitro experiments.^25,31-35 Herein, to establish a cTnT linker structure, we report on studies which couple time-resolved fluorescence resonance energy transfer (TR-FRET) performed on a fully reconstituted thin filament with an all-atom model. (Note that experimental and computational details of the methods are provided in the Materials and Methods section to provide a reasonable précis of the work for readers not fully immersed in either or both of the two techniques.) The goal of the current study is to develop a method that allows extension of structural models to regions that are too flexible to be resolved by other techniques and in particular apply the method to the cTnT flexible linker, a region central to function and a known locus linked to mutations that cause human cardiomyopathies. To address these goals by TR-FRET, we first determine the cTnT linker orientations and confirm the proposed short and long conformations in a fully reconstituted cTF system; we then proceed to map the cTnT linker’s position relative to actin. We emphasize that this is the first time such approaches have been used on such a large protein complex system in its fully functional state. Previous work applied a related idea to a small, nonfunctioning portion of the complex.³⁶ For TR-FRET, cTnT linker residues 168, 177, 192, and 198 were sequentially cysteine-substituted and labeled with the energy donor 5-([([2-iodoacetyl]amino)ethyl]amino)naphthalene-1-sulfonic acid (IAEDANS). The energy acceptor, 4-dimethylaminophenylazophenyl-4′-maleimide (DABMI) was attached to cTnT A168C or S198C to confirm the short and long cTnT linker configurations. For measurements of actin, the energy acceptor, 5-iodoacetamidofluorescein (5-IAF), was attached to cysteine 374 of actin. TR-FRET measurements were obtained for wild-type thin filaments under different biochemical conditions (±calcium) to probe for distinct conformational changes. The computational cTF model was then constrained to reflect the cTnT linker–actin pair distances measured by TR-FRET, and the most probable actin monomer candidates were chosen. These constrained structures were then relaxed to an equilibrated state and then further evolved without constraints to create refined structures. This study used an iterative process to, for the first time, structurally define the cTnT linker. This new computational cTF model can now provide information regarding the structural dynamics of the cTnT linker in both myofilament activation and disease. We note that a recent publication used a related set of tools to find structures of unconstrained, highly mobile C-terminal domains of cardiac troponin I.⁴⁹

MATERIALS AND METHODS

Protein Expression and Purification.

cDNA sequences encoding human cTnT (hcTnT), human cardiac troponin I (hcTnI), human cardiac troponin C (hcTnC), and Ala–Ser α-Tm were inserted into pET3D vectors and provided by J.D. Potter (University of Miami). The single cysteine-substituted variants (hcTnT A168C, A177C, A192C, and S198C) and double cysteine-substituted variant (hcTnT A168C S198C) for TR-FRET experiments were introduced via the Quick-Change II XL site-directed mutagenesis kit (Agilent Technologies). Each clone was sequenced by the University of Arizona Genetics Core through direct DNA sequencing and verified using the SnapGene Viewer (GSL Biotech LLC).

The multiple hcTnT clones were transformed into Rosetta (DE3) Competent Cells–Novagen (EMD, Millipore), while the pET3D vectors containing WT hcTnI, WT hcTnC, and Ala–Ser α-Tm cDNA were transformed into BL21 (DE3) cells (Agilent Technologies). All transformed cells were streaked onto Luria broth (LB)–ampicillin agar plates and incubated at 37 °C overnight. A single colony from each plate was inoculated into a 5 mL LB medium starter culture and grown at 37 °C for 6–8 h while shaking at 250 rpm. 1–2 mL of each starter culture was inoculated into 2 L of autoclaved Overnight Express TB medium with ampicillin for hcTnC or 2 L of autoclaved ZYP medium (1% tryptone, 0.5% yeast, 0.5% (w/v) glycerol, 0.05% glucose, and 0.2% lactose) with 5% 20X P-buffer (1 M Na₂HPO₄, 1 M KH₂PO₄, and 0.5 M (NH₄)₂SO₄), 1 mM MgSO₄, and ampicillin for hcTnT, hcTnI, and Tm. The large cultures were grown overnight at 37 °C while shaking at 250 rpm. Bacterial cells were centrifuged at 2204g for 30 min at 4 °C.

For hcTnT, the bacterial pellet was resuspended in 50 mL of Sp-Sepharose buffer (6 M urea, 50 mM Tris base, 2 mM ethylenediaminetetraacetic acid (EDTA), 1 mM dithiothreitol (DTT), pH 7.0). The pellet was then frozen at –80 °C and thawed at 4 °C. The suspended pellet was then sonicated on ice for 30 s × 12 with 30 s pauses using an ultrasonic cell disruptor, Microson XL-2000 (Misonix). All remaining steps were done at 4 °C unless otherwise noted. The resuspended pellet was then centrifuged at 39 800g for 30 min, and the collected supernatant was loaded at 1.3 mL/min onto the Sp-Sepharose column (Sigma; packed in a Bio-Rad Econo-Column with a 100 mL bed volume) pre-equilibrated with 5–7× column volume of Sp-Sepharose buffer. The column was then washed with 5–7× column volumes of Sp-Sepharose buffer at 1.3 mL/min. Protein was eluted with a linear gradient from 0–0.6 M KCl in Sp-Sepharose at 1.3 mL/min. Fractions containing hcTnT were determined through Coomassie staining of sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) gel, which were then pooled and dialyzed overnight against 2 L of Q-Sepharose buffer (6 M urea, 20 mM Tris base, 1 mM EDTA, 0.3 mM DTT, pH 7.8). Dialyzed protein was loaded on a Q-Sepharose column (Sigma; packed in a Bio-Rad Econo-Column with a 100 mL bed volume), pre-equilibrated with 5–7× column volume of Q-Sepharose buffer, then washed, and eluted with 0–0.6 M KCl in Q-Sepharose buffer. Finally, fractions containing pure hcTnT were determined through Coomassie staining of the SDS-PAGE gel, then pooled, and stored at –80 °C.

hcTnI was purified on an Sp-Sepharose column and eluted with a linear gradient of 0–0.6 M KCl as described for hcTnT. Pure hcTnI fractions were determined through Coomassie staining of the SDS-PAGE gel, then pooled, and dialyzed against 2 × 2 L of TnC affinity buffer (50 mM Tris base, 2 mM CaCl2, 0.5 M NaCl, 1 mM DTT, pH 7.5). Dialyzed hcTnI was then loaded onto a TnC affinity column prepared as per the manufacturer’s protocol for conjugating proteins to a cyanogen bromide-activated Sepharose 4B gel (Sigma) and packed in an Econo-Column (Bio-Rad), pre-equilibrated with 5–7× column volumes of TnC affinity buffer. The column was washed with 5–7× column volumes of TnC affinity buffer, and the protein was eluted using both urea and EDTA gradients (0–6 M and 0–3 mM, respectively) in TnC affinity buffer. Pure fractions were determined through Coomassie staining of the SDS-PAGE gel, then pooled, and stored at –80 °C.

hcTnC was purified on a Q-Sepharose column as previously described for hcTnI and eluted with a linear gradient of 0 to 0.6 M KCl. Pure hcTnC fractions were determined through Coomassie staining of the SDS-PAGE gel, then pooled, and dialyzed against 4 × 2 L of phenyl Sepharose buffer A (50 mM Tris–HCl, 1 mM CaCl₂, 1 mM MgCl₂, 50 mM NaCl, 1 mM DTT, pH 7.5). Dialyzed protein was then brought to room temperature (RT), and solid ammonium sulfate [(NH₄)₂SO₄] was slowly added to a final 0.5 M concentration; dialyzed protein was loaded at RT onto a phenyl Sepharose column (Sigma; packed in a Bio-Rad Econo-Column with a 100 mL bed volume) pre-equilibrated with 5–7× column volumes of phenyl Sepharose buffer A + 0.5 M ammonium sulfate. After washing the column with 5–7× column volumes of the same buffer, hcTnC was eluted with phenyl Sepharose buffer C (50 mM Tris–HCl, 1 mM EDTA, 1 mM DTT, 0.5 M ammonium sulfate, pH 7.5). Pure fractions were determined through Coomassie staining of the SDS-PAGE gel, then pooled, and stored at –80 °C.

The Tm bacterial pellet was suspended in double-distilled water (ddH₂O) and digested with 0.1 mg/mL of lysozyme at RT. The mixture was stirred on ice for 1 h before freezing at –80 °C and thawing at 4 °C. Solid NaCl was added to a 1 M final concentration and then was sonicated for 3 min × 3, with 3 min pauses between each pulse. The sample was then centrifuged at 39,800g for 20 min at 4 °C, and the supernatant was boiled for 5 min in a water bath and then centrifuged at 39 800g for 20 min at 4 °C after reaching RT. To precipitate Tm, 1 M HCl was added dropwise, and the pH was slowly adjusted to 4.4–4.6. Next, the sample was centrifuged at 39 800g for 10 min at 4 °C, and the resultant pellet was dissolved in 1 M KCl by adjusting the pH to 7–8. The pellet was then clarified at 39 800g for 10 min at 4 °C. The Tm precipitation and suspension procedure was repeated 3–4 times or until the protein was clear and free of DNA contamination and then was stored at –80 °C.

hcTnT Double Cysteine-Substituted Labeling.

The double cysteine-substituted hcTnT proteins to be labeled (hcTnT A168C, S198C) were dialyzed for at least 6 h against TR-FRET labeling buffer (6 M urea, 50 mM Tris base, 0.15 M KCl, 1 mM EDTA) with 1 mM DTT. This was followed by two dialysis processes in TR-FRET labeling buffer with 30 μM DTT for 4–6 h. Prior to labeling, protein concentrations were acquired spectroscopically [DU730 Life Science UV/vis spectrophotometer (Beckman Coulter)] with ε280 = 16960 M⁻¹ cm⁻¹. IAEDANS (Molecular Probes, LN 2018257 and 1893559) dissolved in dimethylformamide (DMF) (Sigma-Aldrich) was added to a 0.5- to 1.5-fold molar concentration of the hcTnT protein. Labeling proceeded overnight while being stirred at RT. Labeling was terminated by adding 10 mM DTT. Excess label was removed by centrifugation and dialyzed out of the sample. Labeling ratios were verified spectroscopically by utilizing the IAEDANS lot data, LN 2018257 and 1893559: ε337 = 5400 M⁻¹ cm⁻¹. The desired labeling concentration was 30%. Upon labeling verification, half of the sample was saved for a donor-only sample. The other half was dialyzed for at least 6 h against TR-FRET labeling buffer with 1 mM DTT. This was followed by two dialysis processes in TR-FRET labeling buffer with 30 μM DTT for 4–6 h. DABMI (Setareh Biotech) dissolved in DMF was added to a 2- to 3-fold molar concentration of the hcTnT protein measured prior to the first labeling. Labeling proceeded overnight while being stirred at RT. Labeling was terminated by adding 10 mM DTT. Excess label was removed by centrifugation and dialyzed out of the sample. Labeling ratios were verified spectroscopically utilizing ε460 = 24 800 M⁻¹ cm⁻¹ for DABMI. The desired labeling concentration was 30%.

To confirm accurate labeling of the hcTnT A168C S198C, the labeling order was swapped so that hcTnT A168C S198C was labeled first with DABMI followed by IAEDANS in the same excess-fold molar concentration indicated above. Both of these double-labeled hcTnT A168C S198C were used to determine the R₀ of the system and in the intra-FRET experiments.

hcTnT Single Cysteine-Substituted Labeling.

The hcTnT proteins to be labeled (single cysteine-substituted hcTnT) were dialyzed for at least 6 h against TR-FRET labeling buffer with 1 mM DTT. This was followed by two dialysis processes in the TR-FRET labeling buffer with 30 μM DTT for 4–6 h. Prior to labeling, protein concentrations were acquired spectroscopically with ε280 = 16960 M⁻¹ cm⁻¹. IAEDANS dissolved in DMF was added to a 10-fold molar concentration of the hcTnT protein. Labeling proceeded overnight while being stirred at RT. Labeling was terminated by adding 10 mM DTT. Excess label was removed by centrifugation and dialyzed out of the sample. Labeling ratios were verified spectroscopically by utilizing the IAEDANS lot data, LN 2018257 and 1893559: ε337 = 5400 M⁻¹ cm⁻¹, and were >0.8 labeling in all cases.

Actin Purification.

All experiments were performed in accordance with the NIH guidelines and approved by the Institutional Animal Care and Use Committee of the University of Arizona. All steps were performed at 4 °C unless otherwise noted. To purify actin, minced rabbit skeletal muscle was used to prepare acetone powder. The proteins were extracted with 1 L of Guba-Straub solution (0.3 M KCl, 0.1 M KH₂PO₄, 0.05 M K₂HPO₄; pH 6.5) for 20 min with gentle stirring. The mixture was centrifuged at 5820g for 10 min and then resuspended in 1665 mL of 0.4% NaHCO₃, 0.1 mM CaCl₂ solution for 30 min prior to vacuum filtration. The resulting residue was resuspended in 500 mL of 10 mM NaHCO₃, 10 mM Na₂CO₃, and 0.1 mM CaCl₂ solution with gentle stirring for 10 min and then diluted with 10 volumes of RT H₂O and filtered immediately. The collected residue was resuspended in 666 mL of cold acetone and stirred for 30 min at RT. The filtered residue underwent this acetone treatment four times and was dried overnight at RT prior to storing at −80 °C.

To purify F-actin, actin acetone powder was extracted for 30 min ×2 using cold ddH₂O. The extract was then filtered and clarified at 39 800g at 4 °C for 30 min. The supernatant was immediately polymerized with 30 mM KCl for 1 h at RT. Next, solid KCl was gradually added at a ratio of 4.2 g solid KCl per 100 mL of solution over 2 h at 4 °C. The polymerized actin was then centrifuged overnight at 39 800g at 4 °C. The collected pellet was completely homogenized and dialyzed in G-buffer (10 mM Tris base, pH 7.5, 0.2 mM ATP-sodium, 0.2 mM CaCl₂) to depolymerize into G-actin. G-actin was then clarified by centrifugation at 39 800g at 4 °C for 1 h. Finally, G-actin was polymerized with 0.05 M KCl and 3 mM MgCl₂ for 1 h at RT.

Actin Labeling.

Prior to polymerization, the concentration was determined spectroscopically using ε290 = 26 600 M⁻¹ cm⁻¹. Unlabeled G-actin was then polymerized with 0.05 M KCl and 3 mM MgCl₂ at RT for 1–2 h. Labeling of F-actin Cys-374 with 5-IAF was labeled with a 10-fold molar excess of 5-IAF (Molecular Probes, LN 1792368) in DMF overnight at 4 °C. The reaction was terminated by adding 10 mM DTT. The labeled F-actin was centrifuged at 356 000g for 30 min. The pelleted labeled F-actin was washed and resuspended in G-buffer (10 mM Tris base pH 7.5, 0.2 mM ATP-sodium, 0.2 mM CaCl₂). Labeled G-actin was clarified by spinning at 288 000g for 10 min and then immediately polymerized with 0.05 M KCl and 3 mM MgCl₂ at RT for 1 h. Labeled F-actin was centrifuged at 356 000g for 30 min and resuspended in TR-FRET working buffer [0.4 M KCl, 30 mM 3-(N-morpholino)-propanesulfonic acid (MOPS), 1.25 mM MgCl₂, 1.25 mM CaCl₂, pH = 7.0, 1.5 mM DTT]. The dye concentration of labeled F-actin was determined spectroscopically using ε493 = 74000 M⁻¹ cm⁻¹ for 5-IAF. The protein concentration was determined using the bicinchoninic acid assay with unlabeled G-actin as the standard.

Recombinant cTn Complex and Full Thin Filament Reconstitution for TR-FRET.

Labeled hcTnT (IAEDANS or DABMI), hcTnC, and hcTnI were individually dialyzed against a 6 M urea solution (6 M urea, 0.5 M KCl, 30 mM MOPS, 1.25 mM MgCl₂, 1.25 mM CaCl₂, 1.5 mM DTT; pH 7) for at least 8 h. Protein concentrations were measured spectroscopically. The following extinction coefficients were used: ε280 = 9970 M⁻¹ cm⁻¹ for hcTnI, ε280 = 4470 M⁻¹ cm⁻¹ for hcTnC, ε337 = 5400 M⁻¹ cm⁻¹ for IAEDANS-labeled-hcTnT (IAEDANS: Molecular Probes, LN 2018257 and 1893559), and ε460 = 24800 M⁻¹ cm⁻¹ for DABMI-labeled-hcTnT (DABMI: Setareh Biotech). The cTn complex was reconstituted at a molar ratio of 1:1.2:1.2 hcTnT:hcTnC:hcTnI and underwent a stepwise dialysis to properly form the cTn complex and remove excess urea as follows: sequentially dialyzed for at least 4 h against solutions containing 30 mM MOPS, 1.25 mM MgCl₂, 1.25 mM CaCl₂, 1.5 mM DTT; pH 7 and (1) 6 M urea + 0.5 M KCl, (2) 4 M urea + 0.5 M KCl, (3) 2 M urea + 0.5 M KCl, and (4) 0 M urea + 0.5 M KCl.

cTn complex, Tm, and F-actin (unlabeled and 5-IAF-labeled) were then individually dialyzed in 0 M urea and 0.4 M KCl solution for at least 8 h. The following extinction coefficients were used to measure protein concentration: ε460 = 24800 M⁻¹ cm⁻¹ for cTn[DABMI], ε337 = 5400 M⁻¹ cm⁻¹ for cTn[IAEDANS], ε280 = 17 880 M⁻¹ cm⁻¹ for Tm dimer, ε493 = 74000 M⁻¹ cm⁻¹ for 5-IAF-labeled actin, and ε290 = 26 600 M⁻¹ cm⁻¹ for unlabeled actin. The thin filament was reconstituted at a 0.8:1.0:7.5 cTn:Tm:actin molar ratio, close to the 1:1:7 canonical ratio, similar to other in vitro experiments, to minimize free cTn.^8,37 The reconstituted thin filaments were dialyzed first against 0 M urea, 0.4 M KCl solution, and then against TR-FRET working buffer (0.15 M KCl, 50 mM MOPS, 1 mM ethylene glycol tetraacetic acid, 5 mM MgCl₂, and 1 mM DTT; pH 7.0) for at least 1 h each. Samples were diluted in TR-FRET working buffer to a 1 μM final concentration of the donor label, IAEDANS, and experiments were performed without calcium and with 1 mM calcium. cTF reconstitutions were derived from at least three independent experiments. Each experiment was carried out using at least two independent protein preparations (all components re-expressed) to ensure rigor and reproducibility of distance measurements.

Determining Förster Distance, R₀.

The Förster distance (R₀) characterizes the 50% efficiency point of the energy- transfer between the donor–acceptor (DA) pair. The DA pair’s R₀ is used to determine the distance between the TR-FRET DA (R) as shown in eq 1

$$R=R_0{\left(\frac{1}{E}-1\right)}^{1∕6}$$ (1)

where E is the TR-FRET transfer efficiency. However, a DA pair’s R₀ can change due to the probes’ environment, making it vital to determine each system’s R₀. R₀ can be calculated in Angstrom if the wavelength is in nanometers from the spectral properties of the donor and the acceptor and the donor quantum yield (eq 2)

$$R_0=0.211{\left(k^2{n}^{-4}{Q}_{\mathrm{D}}{J}_{\text{da}}\right)}^{1∕6}$$ (2)

where κ² is the orientation factor (set to 2/3, corresponding to random orientation), n is the refractive index of protein in aqueous solution (1.39), Q_D is the donor fluorescence quantum yield in the absence of the acceptor, and J_da is the overlap integral between normalized donor emission and acceptor absorption spectra. To obtain these measurements, a Beckman–Colter DU-730 UV–vis spectrophotometer was used to measure the absorption spectrum, and a PTI Quanta-Master 40 was used to capture the emission spectrum on a trace using FelixGX software. The a∣e – UV–vis–IR Spectral Software 1.2 (FluorTools, www.fluortools.com) was used to calculate the Q_D and J_da.

The Q_D is determined by comparison to a known standard as seen in eq 3

$$Q_{\mathrm{D}}=Q_{\mathrm{R}}\frac{n^2}{n_{\mathrm{R}}^2}\frac{I}{I_{\mathrm{R}}}\frac{A_{\mathrm{R}}}{A}$$ (3)

where Q_R is the reference quantum yield, n is the refractive index of the solvent, I is the integrated fluorescence intensity, and A is the absorbance at the excitation wavelength. The reference sample used was quinine sulfate which has a Q_D of 0.577 at an excitation of 350 nm,^38,39 and its emission and absorption spectrum was obtained from the Fluorescence SpectraViewer (Thermo Fisher).

Equation 4 was used to determine the J_da

$$J_{\text{da}}={\int }_0^{\infty }{\epsilon }_{\mathrm{A}}(\lambda ){\lambda }^4{F}_{\mathrm{D}}(\lambda )\mathrm{d}\lambda$$ (4)

where ε_A is the acceptor extinction coefficient spectrum (M⁻¹ cm⁻¹), λ is the wavelength (nm), and F_D is the wavelength-dependent donor emission spectrum normalized to an area of 1. Once the Q_D and J_da have been calculated, eq 2 determines the R₀ for each DA pair.

For all DA pairs, an IAEDANS-labeled-donor site on cTnT A168C, A177C, A192C, or S198C was utilized. However, the acceptor depended on the TR-FRET experiment: (1) cTnT A168C to S198C–DABMI-labeled cTnT A168C or S198C; (2) cTnT linker to actin–5-IAF-acceptor-labeled site on Actin374C. The samples were in a fully reconstituted cTF with either the donor- or acceptor-labeled site, and the samples were diluted in the TR-FRET working buffer to a 1 μM final label concentration. The resulting R₀ for experiment 1, cTnT A168C to S198C, was 43.71 Å, which is within the literature range.^40,41 The resulting R₀ for experiment 2, cTnT linker to actin, are as follows: cTF cTnT 168-actin = 51.78 Å, cTF cTnT 177-actin = 50.71 Å, cTF cTnT 192-actin = 47.54 Å, and cTF cTnT 198-actin = 47.65 Å (Table S1). All R₀s determined for cTnT linker to actin are within the literature range.^41,42

Time-Resolved FRET Measurements and Analysis.

50 μL of 1 μM sample aliquots was loaded manually into a 384-well black polypropylene microplate (#781209; Greiner Bio-One). Fluorescence lifetime measurements were acquired using a high-precision fluorescence lifetime plate reader (FLTPR; Fluorescence Innovations⁴³). The sample was excited using a 355 nm microchip laser (Teem Photonics), and the emission was filtered using 409 nm long-pass and 470/20 nm band-pass filters. The FLTPR allows for high-throughput fluorescence lifetime detection at high precision by using direct waveform-recording technology.⁴⁴

All data were analyzed using FargoFit, a global analysis software.⁴⁵ The donor-only decay was fit to a sum of exponentials as shown in eq 5, which was convolved by the instrument response function (IRF)

$$F_{\mathrm{D}}(t)=\sum _{i=1}^n{A}_i\text{exp}(-t∕{\tau }_{\text{Di}})$$ (5)

where n is the number of decay constants (three decay constants had the best χ² fit while not increasing the complexity of the system), A_i values are the amplitude terms, t is the time, and τ_Di values are the decay constant of each exponential. Each exponential decay rate increases in response to a nearby acceptor, with R₀ being the distance at 50% energy transfer and R being the distance between the TR-FRET DA pair (eq 6)

$$F_{\text{DA}}=\sum _{i=1}^n{A}_i{\text{exp}}^{(-t∕{\tau }_{\text{Di}})[1+(R_0∕R{)}^6]}$$ (6)

The R₀ for each DA pair was determined as described above since R₀ can be affected by the probes’ environment. The model used to describe this system was derived from our earlier work where TR-FRET occurred between a donor on cTnT and two acceptors on Tm.^33,46 However, this model considers three distinct population interactions in the fluorescent decay: (1) a population where the donor is not undergoing TR-FRET, (2) a second population undergoing TR-FRET with one acceptor, and (3) a third population undergoing TR-FRET with a different acceptor. These populations are shown in eq 7

$$\begin{gathered} F_{\mathrm{D}+\mathrm{A}}=X_{\mathrm{D}}{F}_{\mathrm{D}}+(1-X_{\mathrm{D}})[X_{\text{DA}1}{F}_{\text{DA}1}+X_2{F}_{\text{DA}2}] \\ {X}_{\mathrm{D}}=X_{\mathrm{f}1}\cdot X_{\text{DA1}}=X_{\mathrm{f}2}∕(1-X_{\mathrm{f}1})X_{\text{DA}2}=X_{\mathrm{f}3}∕(1-X_{\mathrm{f}1}) \end{gathered}$$ (7)

where F_D, F_DA1, and F_DA2 represent the lifetime decays of each population and X_D, X_DA1, and X_DA2 represent the proportions of each population. X_f1, X_f2, and X_f3 are the original fractions determined within FargoFit. The distances are derived from the populations interacting with each acceptor, F_DA1 and F_DA2.

All values are reported as mean ± SD. For the cTnT A168C to S198C measurements, it is necessary to determine the number of populations measured. To test whether these distances represent a single, normally distributed, mean, a Shapiro–Wilk normality test was performed. If these distances fail to pass this hypothesis test, distinct populations of means were split based on model predictions, and again subjected to the Shapiro–Wilk normality test in order to determine if the non-normal distribution was attributable to a bimodal distribution, representing two populations of normally distributed distances around separate means, and thus considered representative of multiple linker conformations. To assess differences in distances of cTnT linker–actin interactions, two-way ANOVA with Sidak correction for multiple comparisons (MCs) was used. Specifically, within the two-way ANOVA, MCs were used to determine whether the measured distances, along the length of the linker, were significantly closer to or further from the actin backbone. Lastly, to analyze the effect of the biochemical condition (±Ca²⁺) on the linker’s position, within sites, a two-way ANOVA with Sidak correction for MC was used. All analyses were performed in GraphPad Prism version 8.0.0 for Windows (GraphPad Software, San Diego, California USA). A level of p < 0.05 was considered statistically significant.

Molecular Dynamics Simulation.

Generation of computational cTF models for calcium-depleted (−Ca²⁺) and calcium-saturated (+Ca²⁺) thin filaments follows a similar procedure utilized previously.⁴⁷ Our previous computational cTF model⁵¹ was refined according to recently published cryo-EM images of the cTF that demonstrated the orientation of the cTn core within the cTF thin filament and the organization of cTnT within the overlap region.¹⁰ This previous model differs from the recent cryo-EM structures in the placement of the TnT helix within the overlap region of the thin filament and the orientation of the Tn core. Hence, these coordinates and the placement of Tm on the actin surface had to be updated according to the two calcium concentrations. Although there remains no atomic model for the unstructured protein chains within the thin filament, these portions have always been present in our full atomic model, with initial structures predicted for the unstructured segments of the chains with a secondary structure prediction tool.²⁵ Atomic coordinates were constrained to those presented explicitly in PDBs 6KN7 and 6KN8 appropriately, while the remaining degrees of freedom remained unconstrained and refined according to the CHARMM force field.⁴⁸ Another cryo-EM study also examined the structure of natural thin filaments, as opposed to the reconstituted thin filaments utilized by Yamada et al.⁹ We note that the two studies are largely in agreement in terms of structural determination, only disagreeing on what is an appropriate Ca²⁺ concentration to define the saturated state.

Once computational cTF models were established for the calcium-depleted state and calcium-abundant state, based on PDBs 6KN7 and 6KN8, respectively, the atomic structures were solvated in a water box with TIP3 waters⁵⁰ and ionized with potassium and chloride ions to neutralize the system to a set concentration of 0.15 M. These prepared structures were independently subject to molecular dynamics (MD) simulations with the CHARMM36 force field in the NanoScale Molecular Dynamics package version 2.13.⁴⁸ Initially, structure minimization was performed for 5000 steps with the conjugate gradient method, heated to a final temperature of 300 K at a rate of 1 K/ps, and then equilibrated at 300 K for 2 ns at a constant pressure of 1 atm. After equilibration, each system was subject to three 10 ns production runs, with each run being randomly assigned velocities initially according to the proper Boltzmann distribution.

From these production runs, hereby referred to as the unrefined cTnT linker runs, distances between the TR-FRET pairs defined earlier were monitored with the various actin monomers present. From these unrefined cTnT linker simulations, we observed that for each cTnT TR-FRET site, at most two actin monomers were in close contact such that changes in the TR-FRET signal could be observed with the previously mentioned TR-FRET labels. Once distance pairs were established, final conformations from the unrefined cTnT linker simulations were subject to constraints such that cTnT linker–actin pair distances reflected those measured by TR-FRET. These constrained structures were subject to minimization, heating, and equilibration in a similar fashion as described earlier except at a temperature of 70 K. Once equilibrated, the newly produced, refined structures were again subject to minimization, heating, and equilibration at a temperature of 300 K but with all constraints removed. Afterward, another production ensemble was produced as described previously. From this ensemble, average TR-FRET distances were measured and observables such as the structure and flexibility of the cTnT linker were monitored.

RESULTS

Intra-TR-FRET and Computational Measurements: WT cTnT Linker Residues 168–198.

Using fully reconstituted cTFs with IAEDANS-donor-labeled cysteine-substituted residues (cTnT A168C or S198C) and DABMI-acceptor-labeled residues (cTnT A168C or S198C) as TR-FRET probes, we were able to confirm that two cTnT linker orientations exist, short and long, in the +Ca²⁺ condition (Figure 2). For the −Ca²⁺ condition, the data had a three-exponential decay D-only data optimal fit and a 1-Gaussian distance distribution donor–acceptor (DA) fit. while the +Ca²⁺ condition had a three-exponential decay D-only optimal fit and a 2-discrete distance DA fit (Table 2). The −Ca²⁺ distance between cTnT 168 and 198 was 36.54 ± 5.66 Å (n = 12); yet, the computational cTF model predicted two confirmations for the −Ca²⁺ cTnT linker: short = 46.19 Å and long = 44.41 Å (Table 2; Figure 2). However, since the computational cTF model predicted distances less than 2 Å apart, the TR-FRET distances are indistinguishable as two populations. With the addition of calcium (+Ca²+), a significant shift occurred in the data distribution from a single, normally distributed mean to a non-normal, bimodal distribution centered around two means (confirmed by Shapiro–Wilks normality test, p < 0.01): a short distance (11.60 ± 6.78 Å, n = 9), and long distance (42.05 ± 1.44 Å, n = 7) (Figure 2E). We wish to note that Figure 2A-E are schematics intended to show the intra-FRET scheme. Actual data is shown in Figure 2E. The TR-FRET agreed with the computational cTF model (which was in turn based on recent cryo-EM experiments) which showed two conformations of the cTnT linker between the short and long side (Figure 2). To confirm this in the TR-FRET data, we separated these distances into their probable bins, and a Shapiro–Wilk test confirmed that the two populations of distances, when categorized into a short and long conformation, were normally distributed (data not shown). This indicated that for the +Ca²⁺ condition, these TR-FRET measurements were in alignment with the computational cTF model predictions and experimentally verified the observation of a short and long cTnT linker seen in recent cTF structural studies.^8,10 We observed both short and long cTnT linker conformations, dependent on the Ca²⁺ condition.

Figure 2.

A168C to S198C Intra-FRET confirms short and long cTnT linker conformations: the FRET donor and acceptor probe locations (black) on the cTF as shown by the computational model (A,C) and schematic representations (B,D). (A,B) represent the short cTnT linker orientation, and (C,D) represent the long cTnT linker orientation. All models are for reference and are in the +Ca²⁺ condition. (E) Intra-FRET measurements in the ±Ca²⁺ conditions were checked for normality via the Shapiro–Wilk test, which demonstrates the shift in data distribution in the +Ca²⁺ condition. **p < 0.01 Shapiro–Wilk test, n = 12 for −Ca²⁺; n = 16 for +Ca²⁺. The subsequent normality test upon mapping FRET distances, based on computational model predictions, showed two normally distributed means in the +Ca²⁺ condition (data not shown).

Table 2.

Computational and Intra-FRET Distances Measured between cTnT 168 and 198: Intra-FRET Distances Are Reported as Mean ± SD^a

	computational model distances		intra-FRET distances
	−Ca2+	+Ca2+	−Ca2+	+Ca2+
short cTnT linker	46.19	9.22	36.54 ± 5.66	11.60 ± 6.78
long cTnT linker	44.41	24.75		42.05 ± 1.44

^an = 12 for −Ca²⁺; n = 9 for +Ca²⁺ short cTnT linker; n = 7 for +Ca²⁺ long cTnT linker. All distances are given in Angstroms.

TR-FRET and Computational Measurements: WT cTnT Linker to Actin.

TR-FRET alone does not give us information on precisely which 374C actin monomer each labeled cTnT site is interacting with. Thus, we constrained the computational cTF model to reflect the cTnT linker–actin pair distances measured by TR-FRET in both cTnT linker confirmations, and the most plausible cTnT linker–actin monomer interactions were identified for both the short and long cTnT linker (Figure 3). Using fully reconstituted cTFs with IAEDANS-donor-labeled cysteine-substituted residues within the cTnT linker region (A168C, A177C, A192C, and S198C) and 5-IAF-acceptor-labeled actin 374C as TR-FRET probes, we were able to determine the cTnT linker’s position in relationship with actin in both ±Ca²⁺. The data had an optimal fit with a three-exponential decay for the D-only data and a two-discrete distance distribution for the DA. For each DA pair, we obtained two discrete distances. The distances of the WT cTnT linker site, −Ca²⁺, to the two closest actin probes were cTnT 168 = 38.76 ± 5.40 and 56.06 ± 2.97 Å; cTnT 177 = 39.84 ± 1.14 and 54.55 ± 1.33 Å; cTnT 192 = 37.54 ± 1.61 and 51.36 ± 1.10 Å; cTnT 198 = 38.26 ± 5.75 and 53.30 ± 4.96 Å (Table 1). With the addition of calcium, the distances were cTnT 168 = 37.24 ± 6.99 and 57.09 ± 2.05 Å; cTnT 177 = 42.00 ± 4.33 and 57.10 ± 4.12 Å; cTnT 192 = 37.78 ± 2.16 and 51.30 ± 2.02 Å; and cTnT 198 = 37.18 ± 5.11 and 52.26 ± 4.66 Å (Table 1). The values obtained from the TR-FRET measurements were in strong alignment with the computational cTF model predictions and recent cTF structural studies.^8,10

Figure 3.

cTnT linker and actin FRET donor and acceptor sites mapped to the computationally derived short and long linker conformations: FRET donor (black) and acceptor (red) probe locations on the computational model and schematic of the cTF. (A,B) represents the short cTnT linker orientation and (C,D) represents the long cTnT linker orientation. All models are for the +Ca²⁺ condition.

Table 1.

cTnT-Linker-to-Actin FRET Distances: WT cTF FRET Distances Measured in Angstroms, for Each cTnT Linker–Actin Interaction in the ±Ca Condition^a

	−Ca2+		+Ca2+
	distance 1	distance 2	distance 1	distance 2
cTF 168	38.76 ± 5.40	56.06 ± 2.97	37.24 ± 6.99	57.09 ± 2.05
cTF 177	39.84 ± 1.14	54.55 ± 1.33	42.00 ± 4.33	57.10 ± 4.12
cTF 192	37.54 ± 1.61	51.36 ± 1.10	37.78 ± 2.16	51.30 ± 2.02
cTF 198	38.26 ± 5.75	53.30 ± 4.96	37.18 ± 5.11	52.26 ± 4.66

^aDistances reported as mean ± SD. n = 6–7 for all measurements.

For the short cTnT linker conformation, computational modeling determined that all 4 cTnT sites interacted with AC19 and AC21, with the longer of the two measured distances interacting with AC19 and the shorter with AC21. In the order of increasingly distal residues (A168C, A177C, A192C, and S198C), the −Ca²⁺ condition distances from AC19 were 56.06 ± 2.97, 54.55 ± 1.33, 51.36 ± 1.10, and 53.30 ± 4.96 Å (Figure 4 and Table 3). A two-way ANOVA identified a main effect of the cTnT linker site (p = 0.0003) for distances from AC19; thus, MCs were used to determine which sites accounted for this variation (Table 4). For cTnT linker distances to AC19, site A192C was found to be trending closer to AC19 than site A168C (p = 0.0590) (Figure 4, Tables 3, and S2). No other differences in position relative to neighboring sites were observed. The other interaction of the short cTnT linker was with AC21, as indicated by the shorter distances measured via TR-FRET. In the order of increasingly distal residues (A168C, A177C, A192C, and S198C), the −Ca²⁺ condition distances from AC21 were 38.76 ± 5.40, 39.84 ± 1.14, 37.54 ± 1.61, and 38.26 ± 5.75 Å (Figure 4 and Table 3). Unlike the interaction with AC19, no differences were observed in the distances measured between sites for the cTnT linker and AC21. After enforcing TR-FRET distances within the computational cTF model as constraints for an initial refinement of the cTnT linker and subsequently performing MD runs with the constraints relaxed to zero, the measured distances were determined (Table 5). For the short linker in the −Ca²⁺ condition, the computational cTF model also observed cTnT 192 lying closer to AC19 than the other pairs (Table 5). However, the computational cTF model predicts that the distal linker residues cTnT 192 and 198 are also closer to AC21, where the TR-FRET shows no difference in the cTnT–actin distances along the linker (Tables 3, 5, and S2). Taken together, these data suggest that the position of the short cTnT linker to the actin backbone in the −Ca²⁺ condition is stable, likely with residue A192C lying closer to AC19 than the surrounding residues.

Figure 4.

cTnT linker-to-actin FRET distance measurements as mapped to the computational model in each biochemical condition: FRET distances mapped to their probable actin monomer are shown in the −Ca²⁺ and +Ca²⁺ condition. Two-way ANOVA was used to analyze the change in each linker site position relative to their neighboring sites and to assess the effect of changing biochemical condition (±Ca²⁺) for each linker site. MCs (Sidak) were used to assess specific differences when a main effect was observed. * vs cTnT 168, # vs cTnT 177, *p < 0.05. Exact p-values are reported in Tables S2 and S3.

Table 3.

FRET Distances Mapped to the Computational Model: The Computational Model Was Used to Determine Which Actin Monomer Each cTnT Linker Donor Site Was Interacting with; AC19, AC21, and AC23 Refer to Specific Actin Monomers Identified by the Computational Model^a

	short cTnT linker				long cTnT linker
	−Ca2+		+Ca2+		−Ca2+			+Ca2+
cTnT site	AC19	AC21	AC19	AC21	AC19	AC21	AC23	AC19	AC21	AC23
168	56.06 ± 2.97	38.76 ± 5.40	57.09 ± 2.05	37.24 ± 6.99	56.06 ± 2.97	38.76 ± 5.40	n/a	57.09 ± 2.05	37.24 ± 6.99	n/a
177	54.55 ± 1.33	39.84 ± 1.14	57.10 ± 4.12	42.00 ± 4.33	54.55 ± 1.33	39.84 ± 1.14	n/a	57.10 ± 4.12	42.00 ± 4.33	n/a
192	51.36 ± 1.10	37.54 ± 1.61	51.30 ± 2.02* #	37.78 ± 2.16	n/a	37.54 ± 1.61	51.36 ± 1.10	n/a	37.78 ± 2.16	51.30 ± 2.02
198	53.30 ± 4.96	38.26 ± 5.75	52.26 ± 4.66#	37.18 ± 5.11	n/a	38.26 ± 5.75	53.30 ± 4.96	n/a	37.18 ± 5.11	52.26 ± 4.66

^aThe short cTnT linker interacts with AC19 and AC21, while the long cTnT linker interacts with AC19, AC21, and AC23. The distances are reported as mean ± SD (n = 6–7) for each cTnT linker–actin distance measurement ±Ca²⁺ condition. Two-way ANOVA was used to analyze the change in each linker site position relative to their neighboring sites and to assess the effect of changing biochemical conditions (±Ca²⁺) for each linker site. MCs (Sidak) were used to assess specific differences when a main effect was observed.

^*vs cTnT 168

^#vs. cTnT 177

^*p < 0.05. Exact p-value reported in Tables S2 and S3.

Table 4.

Result of Two-Way ANOVA’s Main and Interaction Effects: Distances Measured via FRET for Each cTnT Linker–Actin Interaction^a

		main effect		interaction effect
cTnT linker orientation	interacting actin monomer 374C	±Ca2+	cTnT linker position	±Ca2+ × cTnT linker position
short	AC19	0.4907 ∣	0.0003 ***	0.5365 ∣
	AC21	0.9673 ∣	0.2138 ∣	0.7290 ∣
long	AC19	0.1307 ∣	0.5185 ∣	0.5137 ∣
	AC21	0.9673 ∣	0.2138 ∣	0.7290 ∣
	AC23	0.6907 ∣	0.2945 ∣	0.7226 ∣

^aResults of two-way ANOVA are reported by assessing the main and interaction Effects. ∣ = ns

^***p < 0.001

Table 5.

Distances Obtained from Constraining the cTF Computational Model to Measured FRET Distances: The Computational Model Was Constrained to the Measured FRET Distances and Then Allowed to Run Unconstrained to Obtain the Reported Distances (Mean ± SD)^a

	short cTnT linker				long cTnT linker
	−Ca2+		+Ca2+		−Ca2+			+Ca2+
cTnT Site	AC19	AC21	AC19	AC21	AC19	AC21	AC23	AC19	AC21	AC23
168	51.17 ± 6.51	48.75 ± 3.51	53.80 ± 4.77	48.46 ± 7.48	47.98 ± 9.97	39.96 ± 8.41	n/a	51.50 ± 5.44	36.75 ± 5.31	n/a
177	51.10 ± 5.88	42.68 ± 5.71	58.29 ± 5.17	50.96 ± 14.36	49.21 ± 8.53	31.03 ± 8.22	n/a	51.24 ± 3.98	31.41 ± 10.00	n/a
192	47.26 ± 3.59	27.98 ± 5.60	52.76 ±4.18	49.46 ± 5.97	n/a	33.53 ± 2.77	63.96 ± 5.25	n/a	34.40 ± 4.21	74.26 ± 11.79
198	50.86 ± 3.23	29.18 ± 5.31	47.99 ± 5.68	43.45 ± 4.75	n/a	29.90 ± 2.98	57.25 ± 3.49	n/a	29.37 ± 4.42	70.35 ± 13.04

^aAC19, AC21, and AC23 refer to the actin monomers identified by the computational model to interact with each cTnT linker site. This table presents averages of distances from three independent 10 ns runs for each condition. The total number of frames in the analysis is 3000 (1000 per independent run).

With the addition of calcium (+Ca²⁺), the distances for the short linker (ordered as above) from AC19 were 57.09 ± 2.05, 57.10 ± 4.12, 51.30 ± 2.02, and 52.26 ± 4.66 Å (Figure 4 and Table 3). Again, a main effect of the linker site was identified (p = 0.0003) via two-way ANOVA, indicating that differences exist along the linker in its relative position to AC19 (Table 4). MCs were used within the two-way ANOVA to determine what specific changes were occurring. Significant differences in distance from AC19 were found between A168C and A192C (p = 0.0158), A177C and A192C (p = 0.0108), and A177C and S198C (p = 0.0487) (Figure 4, Tables 3, and S2). The +Ca²⁺ distances to AC21 were 37.24 ± 6.99, 42.00 ± 4.33, 37.78 ± 2.16, and 37.18 ± 5.11 Å (Figure 4 and Table 3). As denoted above, no significant differences in distance between the sites and AC21 were observed (Table S2). Similar observations were made in the computational cTF model, with 192 and 198 lying closer to the actin surface in the +Ca²⁺ state (Table 5). The computational cTF model also suggested that the 198-AC21 distance is smaller than the remaining pairs in the +Ca²⁺ state, which was not supported by the TR-FRET data. These data indicate that in the +Ca²⁺ condition, the cTnT linker lies closer to AC19 more distal along the linker, and the computational cTF model confirmed these changes. Notably, more distal residues (both A192C and S198C) were found to be closer to AC19 for the +Ca²⁺ condition than the −Ca²⁺condition. Thus, to determine whether these changes were calcium-dependent, two-way ANOVA comparing the biochemical conditions within each site was performed (Table 4). While the ANOVA demonstrated a main effect of the cTnT linker site (p = 0.0003), no main effect of biochemical condition was observed, indicating that these changes are not calcium-dependent. Biologically, this is understandable because we would not expect a significant change in the cTnT linker’s position due to being anchored by structured domains on either side and also due to the fact that this could alter the cTF’s ability to transduce the allosteric changes upon Ca²⁺ activation and potentially “break” the system.

Computational modeling determined that unlike the cTnT short linker conformation, in the long linker conformation, AC19 interacts only with A168C and A177C, AC21 with all four sites, and an additional actin (AC23) interacts with A192C and S198C. These actin monomers are different but equivalent to those on the short side based on the rotation of the filament; for consistency, the naming convention remains the same. The distances in the −Ca²⁺ condition from AC19 to A168C and A177C were 56.06 ± 2.97 and 54.55 ± 1.33 Å, respectively (Figure 4 and Table 3). No difference in these distances was found via two-way ANOVA (Figure 4, Tables 3, 4, and S3). Similar to the short cTnT linker, all four sites (A168C, A177C, A192C, and S198C) of the long cTnT linker conformation interact with AC21. In the order of increasingly distal residues (A168C, A177C, A192C, and S198C), the −Ca²⁺ condition distances from AC21 were 38.76 ± 5.40, 39.84 ± 1.14, 37.54 ± 1.61, and 38.26 ± 5.75 Å, respectively (Figure 4 and Table 3). No significant change in the cTnT linker distance was identified via two-way ANOVA (Figure 4, Tables 3, 4, and S3). For AC23, the −Ca²⁺ condition distances were 51.36 ± 1.10 and 53.30 ± 4.96 Å for cTnT 192 and 198, respectively (Figure 4 and Table 3). Again, no significant change in distance was observed in two-way ANOVA for cTnT linker distances to AC23 (Figure 4, Tables 3, 4, and S3). Additionally, little difference was observed in the computational cTF model for AC19 distances to 168 and 177 (Table 5). For the AC21 average distances, the model indicated that all cTnT linker sites lie approximately the same distance away from actin. The 168-AC21 distance is a notable exception, with an average distance of 39.96 Å, placing it further from AC19 than the neighboring residues (Table 5). However, the distance distributions of both the 168 and 177 pairs are much larger, approximately 8–9 Å, suggesting that the linker is sufficiently flexible to adopt a closer conformation to the actin surface. Additionally, 192-AC23 is slightly farther from actin than 198-AC23, but there again is an overlap in the distance distributions that suggests little to no difference. This demonstrated that compared to the short linker, the long linker exhibits a more precise acceptable confirmation, which is less tolerant to change.

With the addition of calcium, the distances from AC19 were 57.09 ± 2.05 and 57.10 ± 4.12 Å for A168C and A177C, respectively (Figure 4 and Table 3). The distances to AC21, in the order of increasingly distal residues, were 37.24 ± 6.99, 42.00 ± 4.33, 37.78 ± 2.16, and 37.18 ± 5.11 Å (Figure 4 and Table 3). Finally, the distances from AC23 were 51.30 ± 2.02 and 52.26 ± 4.66 Å for A192C and S198C, respectively (Figure 4 and Table 3). Similar to the −Ca²⁺ condition, no significant difference in distance relative to the actin backbone was observed for any site in the +Ca²⁺ condition (Figure 4, Tables 3, and S3). Furthermore, there was no main effect of adding calcium for any site, as analyzed via two-way ANOVA (Table 4). No difference in the linker’s relative position to actin was further confirmed by the computational cTF model (Table 5). This demonstrated that compared to the short linker, the long linker exhibits a more precise acceptable conformation, which is less tolerant to change, regardless of biochemical condition. Together, these data further confirmed the precise conformation of the long linker and demonstrated that it is not tolerant to change even under different biochemical conditions.

DISCUSSION

Given the importance of mobile protein segments in general and of the cTnT linker in the cTF, it is critical to understand its average structure and dynamics in order to understand myofilament activation. However, the cTnT linker’s structure has remained unresolved in the calcium-activated crystal structure of human cTnT and subsequent EM studies due to its inherent flexibility.^7,14,18 Most recently, a short and long cTnT linker configuration has been suggested to explain the existence of two cTn core populations seen in the recent cryo-EM studies;^8-10 yet, the associated cTnT linker structural insight from these studies was acquired via a relatively low contour level in electron density maps.¹⁰ The central importance of this region in both biological function and disease, and the approach we have developed as a paradigm, motivates this study of both the average location of the linker in all states and the range of motion expected due to inherent flexibility. A new set of techniques was required to find this structure that plays an important role in the biological function of the cTF. It is becoming clear that understanding the average structure and dynamics of previously unresolvable regions of proteins and protein complexes is necessary in order to fully understand the complex function of biological systems.

In agreement with recent cryo-EM observations,^9,10 our TR-FRET results experimentally verified the existence of two cTnT linker populations, short and long, in the +Ca²⁺ condition (Figure 2). Interestingly in the −Ca²⁺ condition, only one cTnT linker population was measured. Our computational model predicted, however, that these populations differed by less than 2 Å. When the computational cTF model was refined by the cTnT linker–actin TR-FRET distances (Table 1), the resulting short and long cTnT linker conformations were confirmed (Figure 2). In the short cTnT linker, the distance between the alpha-helical portion of cTnT within the cTF overlap and cTnT helix 1 within the core is such that the linker takes a condensed orientation. On the opposing side, this alpha-helical cTnT–core distance is larger due to the core being relatively shifted by approximately one actin monomer to maintain cTn core–Tm interaction sites. Therefore, the long cTnT linker may adopt a more elongated conformation. Furthermore, the shifting of ~25 and ~5 Å for the short and long cTnT linkers, respectively, between the −Ca²⁺ and the +Ca²⁺ conditions structurally demonstrates the cTnT linker’s ability to undergo calcium-induced conformational changes.

Previous work from our group explored the relationship between the cTnT linker residues A168C, A177C, A192C, and S198C and Tm C190 in multiple biophysical conditions (± calcium and ± myosin S1).⁴⁶ This study supported the hypothesis that the cTnT linker was highly tolerant to these changing biochemical conditions such that the cTn core’s position could be maintained throughout the cardiac cross-bridge cycle. However, the existence of the short and long cTnT linker conformations was not observed likely due to the mobility of both the cTnT donor and Tm acceptor sites, which only enabled mapping of the cTnT linker’s relationship with Tm. It is important to note that the cTnT linker’s position is influenced by the cTn core’s positioning.^9,10 The cTn core is a rigid stable structure that can be thought of as stationary with respect to actin and the cTn IT arm and moves as a single unit.²⁵ Thus, to accurately map the full cTnT linker’s positioning, the backbone of the thin filament, actin, was utilized as the reference “anchor”. Not only does actin work as an “anchor” for the cTnT linker, but it will also act as a control for the cTn core’s position in both the short and long cTnT linker conformations. This added complexity to the TR-FRET measurements as the number of DA pairs that contributed to the determined TR-FRET intensities could not be known a priori. A carefully chosen DA could limit the number of DA contributions, though this requires knowledge of every possible DA pair within the DA’s acceptable R₀ distance range. The cTnT linker amino acid sequence has always been structurally present within the computational cTF model, but any conformational insight was strictly limited to information inferred from the TNT1 and TNT2 structures, as well as those explored on the timescale of the MD simulations. This provided us with an initial assumption for the general placement and motion of the cTnT linker and aided in the overall choice of the TR-FRET DA label pair, IAEDANS-5-IAF. Additionally, it was observed that the placement of the donor-IAEDANS on cTnT 168/177/192/198 and acceptor-5-IAF on actin 374C would likely allow for two different DA pairs contributing to the overall TR-FRET signal, resulting in the two distance measurements. The measured TR-FRET distances could then influence the computational cTF model by shifting the linker into a likely conformation that could then be further sampled with MD simulations. Enhanced sampling methods can be applied along with MD simulations to further increase the sampling of configuration space for a given system, especially disordered proteins; This, however, comes at the cost of computational efficiency and can be inhibitory for large, million-atom systems.

When the computational cTF model was constrained by the measured TR-FRET distances, it was observed that the actin reference points were different, depending on whether the linker was “short” or “long”, and therefore, the distances were heterogeneous according to the two positions of the cTn core with respect to the overlap when comparing each side. Overall, by incorporating the TR-FRET measurements into the computational cTF model, we were able to generate likely positions for the respective amino acids that comprise the cTnT linker and provide some insight into the motility of the linker. Initially, possible TnT–actin monomer pairs were monitored during the unrestrained linker to observe which specific actin monomers came within a distance range capable of contributing significantly to the FRET decays. This allowed for the ruling out of any possible contributions, for a given linker, to the opposing actin monomers on the opposite side of the thin filament. In observations of the Yamada structure,¹⁰ the opposing actin monomers come within 50 Å of residue 198/199. We again monitored TnT 198–actin distances to the opposing actin monomers during our refined linker ensemble (Figure S1). Although one production run does show that the linker was capable of producing distances as observed in the Yamada structure, however, the remaining runs observed the linker adopting conformations much farther away and overall averages to a distance within the range of 70–80 Å. Within these distances, limited FRET contribution would be expected. Additionally, relatively short timescales of MD simulations are necessitated by the large number of atoms in the full model–the linear shape of the cTF results in a huge number of water molecules in the simulation “box”, and hence, the configuration space searched in our three separate simulations may not be fully exhaustive of all possible linker positions, but, of course, this may be said of any MD simulation.

For the short and long cTnT linker conformations, there was no main effect of adding calcium for any site (Table 4), which indicated that any differences found were cTnT linker site-dependent. In the −Ca²⁺ condition, short cTnT linker residue A192C was likely lying closer to AC19 than the surrounding residues (Figure 4, Tables 3, and S2). However, with the addition of calcium, +Ca²⁺, both A192C and S198C were found to be closer to AC19 (Figure 4, Tables 3, and S2). It is interesting to note that cTnT A192C is found to be lying closest to AC19 in both ±Ca²⁺, as it was previously shown to be the closest of the four cTnT linker residues to Tm in the WT cTF structure.⁴⁶ It is possible that at cTnT A192C, the cTnT linker-Tm and actin interactions are more strongly influenced by the presence and absence of Ca²⁺ than the rest of the cTnT linker, indicating that this site may well be central to the function of the cross-bridge cycle. We have previously suggested that a myosin binding site could exist near the distal cTnT linker and that the addition of calcium initiates the necessary structural rearrangements for “priming” the linker for myosin binding,⁴⁶ which is further indicated here. For the long cTnT linker, no significant difference in distance relative to the actin backbone was observed for any site in both ±Ca²⁺ conditions (Figure 4, Tables 3, and S3). This could imply that because the long cTnT linker must cross over the body of actin in order to maintain proper cTn core–Tm binding, it must traverse about one actin monomer longer distance than the short cTnT linker (Figure 1). This stretching could force the long cTnT linker to have a more precise confirmation which is less tolerant to change. This finding is in agreement with recent cryo-EM observations.^9,10

In summary, we coupled in vitro FRET experiments with atomistic computational modeling to, for the first time, provide a structure (both average and range) of the extended cTnT linker domain. We experimentally confirmed (in a fully reconstituted system) the existence of the short and long cTnT linker confirmations that were observed in recent cryo-EM studies.^8-10 We presented evidence that the cTnT linker is relatively stable regardless of the biochemical condition. These findings provide a well-defined starting point for further investigation of the primary functional role of the cTnT linker in the regulatory thin filament and to determine how cTnT linker mutations linked to dilated and hypertrophic cardio-myopathies potentially alter this region to disrupt function and cause disease. The work presented in this article also provides a paradigm for the investigation of flexible regions in any protein system for which FRET and computation are accessible.