Molecular characterization of linker and loop-mediated structural modulation and hinge motion in the C4-C5 domains of cMyBPC

Abstract

Introduction:

The central C4 and C5 domains (C4C5) of cardiac myosin binding protein C (cMyBPC) contain a flexible interdomain linker and a cardiac-isoform specific loop. However, their importance in the functional regulation of cMyBPC has not been extensively studied.

Methods and results:

We expressed recombinant C4C5 proteins with deleted linker and loop regions and performed biophysical experiments to determine each of their structural and dynamic roles. We show that the linker and C5 loop regions modulate the secondary structure and thermal stability of C4C5. Furthermore, we provide evidence through extended molecular dynamics simulations and principle component analyses that C4C5 can adopt a completely bent or latched conformation. The simulation trajectory and interaction network analyses reveal that the completely bent conformation of C4C5 exhibits a specific pattern of residue-level interactions. Therefore, we propose a “hinge-and-latch” mechanism where the linker allows a great degree of flexibility and bending, while the loop aids in achieving a completely bent and latched conformation. Although this may be one of many bent positions that C4C5 can adopt, we illustrate for the first time in molecular detail that this type of large scale conformational change can occur in the central domains of cMyBPC.

Conclusions:

Our hinge-and-latch mechanism demonstrates that the linker and loop regions participate in dynamic modulation of cMyBPC’s motion and global conformation. These structural and dynamic features may contribute to muscle isoform-specific regulation of actomyosin activity, and have potential implications regarding its ability to propagate or retract cMyBPC’s regulatory N-terminal domains.

1. Introduction

Myosin binding protein C (MyBPC) is an important regulatory protein in striated muscle. It consists of three paralogs: slow skeletal (sMYBPC), fast skeletal (fMyBPC) and cardiac (cMyBPC) isoforms encoded by the MYBPC1, MYBPC2, and MYBPC3 genes, respectively. All MyBPC paralogs localize to the C zone of the sarcomere (Heling et al., 2020) and share common structural features which include seven immunoglobulin I-set (IgI) domains (C1-C5, C8, C10) and three fibronectin type III domains (C6, C7, C9); however, there are some key isoform-specific differences. fMyBPC consists of features that are shared with the cardiac isoform like the N terminal Pro/Ala (PA)-rich motif, domains C1-C10, and a 10 amino acid linker between C4 and C5 (C4C5 linker). On the other hand, sMyBPC is the product of four alternative splice variants with insertions or exclusions in the PA region, C7, and the C terminal (Ackermann and Kontrogianni-Konstantopoulos, 2011). The cardiac isoform differs from the fast skeletal isoform by the addition of a C0 domain, the M-motif between C1 and C2, the C4C5 linker, and a 28 amino acid loop structure in C5 (Oakley et al., 2004). Although the M-motif and the C4C5 linker are both part of fMyBPC and cMyBPC, the cardiac isoform’s M-motif has a 9-residue insertion with important cardiac specific phosphorylation sites (Gautel et al., 1995; Jia et al., 2010). Additionally, the cardiac isoform’s C4C5 linker shares only 50% sequence identity compared to the fast skeletal isoform (Fig. 1). The C0 domain and the 28 amino acid C5 loop are strictly cardiac specific (Flashman et al., 2004).

Fig. 1.

Sequence alignment of the C4C5 domains of slow, fast, and cardiac isoforms of myosin binding protein C (MyBPC) from Homo sapiens, Mus musculus, Rattus norvegicus, Felis catus, Canis lupus familiaris, Sus scrofa, and Bos tarus. The C4C5 domains contained in the box represent their physical length in comparison to the full length cMyBPC molecular model. The RefSeq and UniProtKB Accession numbers for each species-specific MyBPC protein isoform are given in Fig. S1. The linker region is highlighted in green, and the loop region is highlighted in orange.

The cardiac isoform of MyBPC has been given special attention in the research community due to its role in hypertrophic cardiomyopathy (HCM). This progressive disease has a prevalence of 1:200–500 (Baudhuin et al., 2015) and can cause disability or death due to heart failure or stroke (Maron, 2002). In fact, more than 250 HCM-causing mutations in cMyBPC have been reported (Carrier et al., 2015; Harris et al., 2011). The prevalence of such mutations throughout the entire length of cMyBPC, including C4 and C5 domains (Stenson et al., 2017), suggests that cMyBPC function or dysfunction is dependent on the structure of all 11 domains. It is known that cMyBPC is a critical modulator of the speed and magnitude of force generation (Coulton and Stelzer, 2012; Flashman et al., 2004; Heling et al., 2020; Mamidi et al., 2014, 2016). Notably, the N-terminus of cMyBPC (C0-C2) contains physiologically important phosphorylation sites that have been shown to regulate its binding with myosin S1/S2 and actin (Belknap et al., 2014; Gautel et al., 1995; Li et al., 2020; Moss et al., 2015; Mun et al., 2014), whereas the C-terminal domains (C8-C10) mediate the localization of cMyBPC to the A band and anchor the molecule to light meromyosin (Alyonycheva et al., 1997; Gilbert et al., 1999; Moos et al., 1975). In contrast, very little is known about the precise structure, function and regulation of the central domains of cMyBPC (C3-C7).

It is of interest to elucidate the roles of the central domains of cMyBPC because they contain unique structural features that may have potential functional relevance. The linker between C4 and C5 is a 10-amino acid extension beyond the predicted N-terminal of the C5 domain’s IgI motif, which may allow hinge motion between C4 and C5 (Oakley et al., 2004). Rotary shadowing electron microscopy studies revealed that cMyBPC indeed forms three major particle shapes: V-shaped (58%), elongated (20%), or amorphous (22%) (Hartzell and Sale, 1985). A more recent study showed that cMyBPC formed bent rods with one or two hinge points (50%), straight and potentially folded molecules (31%) and amorphous structures (19%) (Previs et al., 2016). Of the bent conformations, about 60% had one hinge forming a V shape while 40% had two hinges (Previs et al., 2016). Based on measured lengths of the hinge arms, the two hinge points most likely reside between domains C1-C2 and C4-C5 (Previs et al., 2016).

In addition to the 10-amino acid linker between C4 and C5, the C5 domain contains a cardiac specific 28-amino acid loop inserted between its C and D β-strands (CD-loop). An NMR study found that the C5 loop is unstructured, highly dynamic and extended, and destabilizes the domain (Idowu et al., 2003). However, it is stabilized by contacts with other domains of the protein, which compensate for the absence of a folded CD-loop (Idowu et al., 2003). A computational study of the isolated C5 domain confirmed its unstable nature through kinetic simulations of the unfolding transition (Guardiani et al., 2008). Furthermore, due to a relatively high number of proline and alanine residues in the C5 loop, it has been speculated that the loop may function as a stable scaffold for cardiac-specific ligand interactions of signal transduction molecules, specifically SH3 domains (Gautel et al., 1995; Musacchio et al., 1994). However, the precise role of this cardiac-specific feature has not yet been identified.

The ability of native cMyBPC to adopt a wide-range of conformations such as a V-shape or completely bent elongated shape requires a high degree of inter-domain flexibility that can be attributed to the central domains of the molecule and suggests that the C4C5 linker and/or the C5 loop may be important for dynamic transition and regulation of cMyBPC structure. To date, there have been over 30 disease causing missense mutations identified in relative proximity to or within the C4C5 linker and the C5 loop (Stenson et al., 2017), which could affect its folding, conformation, or function. In addition, various studies have found post translational modifications in the C4C5 domains, although the precise mode of regulation has not been elucidated (Hornbeck et al., 2012; Huttlin et al., 2010; Kooij et al., 2013; Lundby et al., 2013; Schumacher et al., 2007; Wagner et al., 2012). Furthermore, it has been observed that the binding affinity of myosin S1 to full-length cMyBPC is significantly greater than to C0-C2 alone (Nag et al., 2017), and that the strongest binding sites between cMyBPC and myosin S1 are localized to its central domains (Ponnam and Kampourakis, 2021). Since no study thus far has systematically explored the functional effects and unique features of the C4C5 domains in one combined system, the aim of this study was to determine the structural characteristics of the linker and loop regions and the dynamics of the C4C5 domains of cMyBPC.

2. Methods

2.1. Sequence alignment and analysis

Multiple sequence alignment of the slow, fast, and cardiac isoforms of myosin binding protein C from seven species was built using the NCBI Protein Database and Cluster Omega (Madeira et al., 2019). The reference sequence accession numbers for each species are as follows: Homo sapiens, slow (NP_002456.2), Mus musculus, slow (NP_001239301.1), Rattus norvegicus, slow (NP_001094228.2), Felis catus, slow (XP_023113138.1), Canis lupus familiaris, slow (XP_022259120.1), Sus scrofa, slow (XP_005664259.1), Bos taurus, slow (NP_001104243.1), Homo sapiens, fast (NP_004524.3), Mus musculus, fast (NP_666301.2), Rattus norvegicus, fast (NP_001099727.1), Felis catus, fast (XP_006941016.1), Canis lupus familiaris, fast (XP_533608.2), Sus scrofa, fast (XP_003127419.3), Bos taurus, fast (XP_024834509.1), Homo sapiens, cardiac (NP_000247.2), Mus musculus, cardiac (UniProtKB O70468), Rattus norvegicus, cardiac (NP_001099960.1), Felis catus, cardiac (XP_019667955.1), Canis lupus familiaris, cardiac (NP_001041571.1), Sus scrofa, cardiac (XP_020940874.1), and Bos taurus, cardiac (NP_001070004.1). The FASTA sequences used in the sequence alignment are included in Fig. S1.

2.2. Gene synthesis and protein expression

pET-30a(+) plasmid vector carrying DNA fragments of mouse cMyBPC C4C5 domains tagged with N-terminal His₆-tags were obtained from GenScript (Piscataway, NJ). The HGVS protein nomenclatures and abbreviated names in parentheses for the protein constructs are listed as follows (one-letter abbreviations were used for each amino acid): O70468:p.M1_K539del; E629_Q1270del (C4), O70468:p.M1_Q638del; V767_Q1270del (C5), O70468:p.M1_K539del; V767_Q1270del (C4C5), O70468:p.[M1_K539 del; E629_Q638del; V767_Q1270del] (Linker^del), O70468:p.[M1_K539del; V683_D710del; V767_Q1270del] (Loop^del), and O70468:p.[M1_K539del; E629_Q638del; V683_D710del; V767_Q1270del] (Double^del). The C4C5, Linker^del, Loop^del, and Double^del proteins correspond to the wild-type C4C5 fragment, mutant with the linker deletion, mutant with the loop deletion, and mutant with both linker and loop deletions, respectively. E. coli BL21 Star (DE3) cells were transfected with the above plasmid vectors using heat shock at 47 °C for 45 s and grown on agar plates. E. coli pilot culture was grown in lysogeny broth (LB) - Miller formulation (10 g/L tryptone, 5 g/L yeast extract, 10 g/L NaCl) containing Kanamycin (34 μg/ml working concentration) at 37 °C / 225 rpm on MaxQ 4000 Benchtop Orbital Shaker for 16 h. 15% glycerol stocks of each of the constructs were made, flash frozen in liquid N₂, and stored at −80 °C for subsequent steps. The pilot 50 ml culture was diluted into 1 L of the same growth medium and continued to incubate for 3–5 h at 37 °C shaken at 225 rpm until OD_600nm was approximately 1.0. At that time, the culture was cooled down on ice for 30 min and 0.5 mM IPTG was added. The expression of the recombinant C4C5 constructs were induced overnight at 16 °C with shaking at 225 rpm. The cells were harvested by centrifuging at 6,000 rcf at 4 °C for 10 min, and the pellet was resuspended in 25 ml of cold lysis buffer (20 mM HEPES pH 7.5, 500 mM KCl, 10 mM Imidazole, 1 mM AEBSF), flash frozen in liquid N₂, and stored at −80 °C until purification.

2.3. Recombinant protein purification

All molecular biology techniques were conducted as described previously (Li et al., 2020). The frozen cells were thawed in RT (~21 °C) and resuspended in about 50 ml of cold lysis buffer plus a protease tablet (cOmplete ULTRA Tablets, Mini, EDTA free, EASYpack Protease Inhibitor Cocktail, Roche). The cells were lysed with the Avestin Emulsiflex C3 homogenizer (ATA Scientific) at a pressure of 10,000–15,000 psi at 4 °C. The cell homogenates were clarified at 50,000 rpm at 4 °C using the Beckman Coulter Optima L-100 XP Ultracentrifuge with Type 70 Ti Rotor for 1 h. The supernatant was incubated with HisPur Ni-NTA Resin (Thermo Scientific) that was pre-washed in wash buffer (20 mM HEPES pH 7.5, 500 mM KCl, and 50 mM Imidazole) on a rocker for 2–3 h at 4 °C. Proteins were placed in a 15-ml gravity flow column (Econo-Pac Chromatography Column, Bio-Rad) and washed with 10–15 ml of wash buffer. The His₆-tagged recombinant proteins were eluted from the column using a total of 4 ml of elution buffer (20 mM HEPES pH 7.5, 500 mM KCl, and 250 mM imidazole). Due to a relatively high purity at this stage, all fractions were collected and concentrated using a centrifugal concentrator (Amicon Ultra Centrifugal Filter Unit, 10 kDa molecular weight cut off) according to manufacturer’s instructions. The concentration was determined by the UV absorbance at 280 nm (A₂₈₀) with the Beckman Coulter DU 800 Spectrophotometer, corrected for background light scattering using the two point Rayleigh approximation method to account for RNA/DNA contaminants and turbidity (Porterfield and Zlotnick, 2010). The full UV spectra from 800 to 200 nm was measured for each construct. The theoretical extinction coefficients were calculated from ExPASy ProtParam tool (Gasteiger et al., 2005) using the predicted Edelhoch method equation: ε(280 nm) = 5500*(# Trp) + 1490*(#Tyr) + 125*(#Cystine). Although the best method to measure protein concentration is the direct Edelhoch method, it has been shown that the predicted extinction coefficient is quite reliable for proteins that contain Trp residues (Pace et al., 1995). Since our recombinant proteins each had 2–3 Trp residues, calculating their concentrations from the predicted Edelhoch equation was sufficient. Appropriate amounts of protein samples were injected in the NGC Medium-Pressure Liquid Chromatography System with an ENrich High-Resolution Size Exclusion Column 70 (10 × 300 mm, BioRad) equilibrated in HEPES buffered saline (20 mM HEPES pH 7.5, 150 mM NaCl). The last 6 fractions for C4 and last 7 fractions of C5 were collected, while all fractions containing C4C5 constructs were collected and pooled. A final stock protein concentration was determined by UV absorbance at 280 nm as described above. Purified proteins were at least 90–95% pure by Coomassie staining. All experiments were done with fresh protein samples within 1 week of purification or immediately after thawing.

2.4. SDS-PAGE, Coomassie staining and Western blots

In preparation for sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE), protein samples were diluted, prepared in 1x Laemmli sample buffer and heated at 95 °C for 10 min. Samples were loaded and electrophoretically separated using 4–12% TruPAGE Precast Gels (Sigma-Aldrich) at 180 V for 45 min. For total protein level determination, the gel was stained with GelCode Blue Stain Reagent (Thermo Scientific) or “Blue silver” colloidal Coomassie formulation (10% phosphoric acid, 10% ammonium sulfate, 0.12% G-250 dye, and 20% methanol) (Candiano et al., 2004). Confirmation of C4C5 constructs was performed by Western blot analysis. Briefly, proteins were transferred to PVDF membranes and probed with anti-His-tag antibody (Santa Cruz Biotechnology Inc. AD1.1.10, catalog # sc-53073, 1:10,000). Membranes and gels were imaged using the Azure Biosystems c600 Imaging system. Densitometric quantifications were performed using the ImageJ software available from the U.S. National Institutes of Health, Bethesda, MD, USA (Schneider et al., 2012).

2.5. Size exclusion chromatography-multiple angle light scattering (SEC-MALS)

SEC-MALS analyses of the solution oligomerization state(s) of all constructs were performed at room temperature using an 18-angle light scattering (LS) detector (DAWN HELEOS II) and a differential refractive index (dRI) detector (The Optilab T-rEX) from Wyatt Technology, connected in tandem to an upstream ÄKTA Explorer FPLC system (Cytiva, formerly GE Healthcare Lifesciences). C4, C5, C4C5, Linker^del, Loop^del, and Double^del constructs were injected at concentrations of 150 μM for single domain and 50 μM for double domain constructs, respectively, in a 500 μl volume through the Superdex 200 increase 10/300 GL column (Sigma-Aldrich) using HBS buffer with 1 mM DTT, for in-line LS and dRI data collection. Data were analyzed using the ASTRA 7.1.4 software package (Wyatt Technology).

2.6. Far-UV circular dichroism (CD) spectroscopy

CD spectra of C4, C5, and C4C5 constructs were recorded on the Jasco J-815 CD Spectropolarimeter or the AVIV Model 215 CD Spectrometer at 25 °C in boric acid- sodium fluoride buffer, pH 7.4 (10 mM boric acid, 150 mM NaF) using a quartz cuvette with a 0.1 cm path length. The cuvette was cleaned with 2 M nitric acid, rinsed with water, and dried with nitrogen gas after each day of use. The CD signal was recorded from 260 −180 nm with a bandwidth of 1 nm, step size of 1 nm, and averaging time of 5 s. Each experiment consisted of three trials with 10 accumulations each. All accumulations were averaged, then background corrections were done by subtracting three independent averaged spectra of the boric acid- NaF buffer from the averaged experimental spectra. Protein concentrations used in calculating mean molar residue ellipticity ([θ]_MRE) were determined by UV absorbance at 280 nm, with Rayleigh correction as discussed above. The average concentrations were 0.247 ± 0.007 mg/ml for C4, 0.223 ± 0.008 mg/ml for C5, 0.112 ± 0.011 mg/ml for C4C5, 0.096 ± 0.003 mg/ml for Linker^del, 0.137 ± 0.003 mg/ml for Loop^del and 0.132 ± 0.004 mg/ml for Double^del. [θ]_MRE (deg * cm² * dmol⁻¹) was calculated using the following formula: [θ]_MRE = (θ_obs * MRW) / (10 * l * c), where θ_obs is the raw CD signal in milli-degrees, l is the path length in cm, and c is the concentration in mg/ml. MRW or mean residue weight is M/(N-1), where M is the molecular weight in g/mol and N is the number of amino acids. The spectra were not double normalized to any wavelength. The raw CD spectra in MRE of all groups were smoothed using a 2nd order Savitzky-Golay filter with a point window of 10. The smoothed data were analyzed by the DichroWeb online server using three standard algorithms (SELCON3, CONTIN-LL, CDSSTR) with reference sets 6 and 7 (Whitmore and Wallace, 2004). We also used the BeStSel webserver (Micsonai et al., 2018) since it addresses the spectral diversity of β-structured proteins and provided more detailed analyses of β-sheet secondary structures. All results were combined in the consensus data.

2.7. Differential scanning fluorimetry

The thermal stability of the C4C5 domains of cMyBPC and its truncation mutants were explored using the Protein Thermal Shift assay (Applied Biosystems by Life Technologies). Each experiment was conducted according to manufacturer’s instructions. Briefly, each reaction component consisting of 1 μg/μl of C4C5 constructs in Protein Thermal Shift Buffer and 1x SYPRO Orange dye was added and mixed in the 96-well format Applied Biosystems MicroAmp Optical Reaction Plate at 4 °C. The plate was sealed with the MicroAmp Optical Adhesive Film and centrifuged at 200g for 1 min at 4 °C. A StepOnePlus real time PCR instrument was used to run “Melt Curve” experiments with the following settings: standard ramp speed, ROX reporter dye, no quencher, no passive reference, 20 μl reaction volume, continuous ramp mode, temperature range from 25 to 95 °C, and 1% ramp rate (0.025 °C/s for 20 μl reaction volume, according to the Protein Thermal Shift Studies user guide). The data were analyzed with the Protein Thermal Shift Software, v1.4 package (Applied Biosystems by Life Technologies). After data normalization from the “no-protein control” reference, the florescence melt curves and the first derivative method were used to determine the multi-peak melting temperatures (T_D1, T_D2). No noise reduction was performed because the replicate averages reduced high frequency noise. The Boltzmann method was also used to determine the overall melting temperature (T_B). T_B is calculated by automated selection of a region of analysis (ROA) and fitting the curve to the Boltzmann equation: F(T) = F (pre) + [(F(post)-F(pre))/(1 + ê((T_B-T)/C))], where F(T) is the fluorescence at a temperature, F(pre) is the fluorescence at the start of the ROA, F(post) is the fluorescence at the end of the ROA, T_B is the melting temperature, and C is the enthalpy of the reaction.

2.8. CD thermal melting and cooling experiments

Similar to CD spectroscopy experiments to obtain secondary structure, the CD thermal melt/cooling experiments utilized the Jasco J-1500 CD Spectrophotometer with a quartz cuvette with a 0.1 cm path length (Greenfield, 2006a; Miles et al., 2021). All of the proteins were dialyzed in boric acid- sodium fluoride buffer, pH 7.4 (10 mM boric acid, 150 mM NaF). The protein concentrations were determined by UV absorbance at 280 nm with baseline correction at 340 nm via NanoDrop 2000 spectrophotometer (Thermo Scientific). The protein samples were filled to the brim in a sealed cuvette to prevent evaporation and resulting fluctuations in concentration. The “Temperature Interval Measurement” software was used to ramp up the temperature from 25 °C to 101 °C with a temperature gradient of 1 °C/min, data pitch of 2.0 °C, and delay time of 10 s while obtaining a full spectrum of 175–260 nm with a data pitch of 1 nm and D.I.T. of 2 s. One accumulation was obtained at each temperature pitch with a scanning speed of 100 nm/min. The “reverse” setting was used to collect the reverse melting data. Three trials were done for each protein construct. After the buffer spectrum at each corresponding temperature was subtracted from each of the experimental spectrum, the resulting data were converted to MRE as described above. The MRE data were smoothed using a 2nd order Savitzky-Golay filter with a frame length of 11. Smoothed data were used to calculate secondary structure using the “multiple spectra analysis” option on BeStSel (Micsonai et al., 2018). The smoothed data were averaged together for representative images. All processing was done using MATLAB R2021b. The melting curves at 230 nm and temperature-dependent changes in secondary structure basis components were fitted to a Boltzmann sigmoid curve on GraphPad Prism 6 using the equation Y = (Min+(Max-Min)) / (1 + exp((Tm-X)/Slope)), where Tm is the inflection point of the sigmoidal curve, and Slope is the curvature coefficient (Heusser et al., 2021). One wavelength of 230 nm was chosen based on the goodness of fit across all constructs for consistency and previously published reports on Ig domain melting (Pricolo et al., 2020; Rao et al., 2010).

2.9. Computational modeling

Computational modeling was done as described previously (Doh et al., 2019). The full length mouse cMyBPC (UniProtKB Accession No. O70468) model was made by RaptorX server (Källberg et al., 2012). In order to create a C4-C5 model, a literature search of the Protein Data Bank (PDB) was done, and available experimental structures were identified: 2DLT (mouse, fast isoform) and 2YUZ (human, slow isoform) for the C4 domain and 1GXE (human, cardiac isoform) for the C5 domain. The most representative NMR models of 2DLT and 1GXE (model 2 for both) were selected for further modeling using an analysis of NMR structure ensembles (OLDERADO) (Kelley and Sutcliffe, 1997). Template-based homology modeling of the mouse cMyBPC C4 and C5 domains were done with the I-TASSER server (Yang et al., 2014) using the mouse cMyBPC sequence (UniProtKB Accession No. O70468). The specific amino acids used in the modeling process were identical to the recombinant protein constructs without the N-terminal methionine and His₆-tags. The C4 and C5 domains were separated and iteratively refined using locPREFMD (Feig, 2016), Galaxy WEB (Heo et al., 2013), and MolProbity (Williams et al., 2018) to reduce steric clashes and improve protein geometry and global/local conformations. The linker region between C4 and C5 was also modeled using I-TASSER and refined as described above. In order to connect the C4 and C5 domains through the linker region, the overlapping regions of C4 domain/linker and linker/C5 domain were aligned with each other. Then, the linker peptide was truncated and each side of the truncated linker peptide was manually connected using the “bond” command in The PyMOL Molecular Graphics System (Version 2.4.1. Schrödinger, LLC). In order to determine the effect of initial structure on the end result, two other C4C5 models were made with a kink in the C4-C5 linker region so that each model had widely different initial starting conformations. The final C4C5 models were refined and validated using the procedure described above, and protonated using the PlayMolecule ProteinPrepare tool (Martínez-Rosell et al., 2017).

2.10. Molecular dynamic simulation (MDS)

The truncation mutations in C4C5 involving the 10 amino acids in the linker region (EVKID FVPRQ), the 28 amino acids in the loop region in C5 domain (TVTQG KKASA GPHPD APEDA GADEE WVF), or both were introduced in silico using the CHARMM-GUI server (Jo et al., 2008; Lee et al., 2016). C4C5 and the mutant models were solvated with 0.15 M potassium chloride ions and explicit TIP3P water molecules using the Monte Carlo method in cubic boxes (204 ų). Periodic boundary conditions were set using the Particle-mesh Ewald (PME) method. Although the C4C5 model is part of a multi-domain protein with potential N- and C-terminal constraints, no additional constraints for the simulations were used because we wanted to explore the high flexibility and dynamics that C4 and C5 domains may have. The simulation time step was set to 2.0 femtoseconds (fs) and coordinates were saved every 10 picoseconds (ps). After a 100 ps minimization under the NVT ensemble, the system was equilibrated and simulated under the NPT ensemble at 310.15 K and 1.0 atm with Langevin dynamics/piston method using NAMD 2.12 (Phillips et al., 2020) and CHARMM36m force field (Huang et al., 2017). For the three C4C5 domain replicates, they were run for 700 ns, along with the Linker^del, Loop^del, and Double^del models. The 700 ns end states for one of the C4C5 models along with the three mutant models were used to simulate three different 200 ns runs for each model. The 500–700 ns segment for the three C4C5 domain replicates were used to study the effects of initial position. In total, we have simulated 700 ns triplicates of C4C5 (2100 ns), 700 ns of the 3 mutants (2100 ns), and 200 ns triplicates of one C4C5 model and 3 mutant models (2400 ns). This resulted in a combined total simulation time of 6600 ns or 6.6 μs.

2.11. Trajectory analyses of MDS

All analyses of MDS trajectories were performed using the Anaconda distribution of Python version 3.7 with the MDAnalysis package (Michaud-Agrawal et al., 2011). Frame-by-frame root mean square deviation (RMSD) calculations were computed with reference to the first frame following energy minimization steps on trajectories aligned to this reference frame (0.1–700.1 ns). The RMSD results showed that large structural changes dominated the initial portion of the trajectories for some constructs. As described further in the results, these large changes represent the movements of the two immunoglobulin domains as they find their most stable conformation relative to each other. Because we were only interested in the dynamics of the constructs about their mean steady-state conformation, we only considered the last 200 ns segments of the C4C5 model triplicates and the triplicate extension runs of the mutant models for the remainder of the MDS analyses described unless otherwise noted. For example, the root mean square fluctuation (RMSF) for each α carbon was computed from trajectories that were initially aligned to the first frame, and then aligned again to the mean structure of all frames across the three 200 ns replicates.

The gross geometric properties of the constructs were assessed at each time step of a trajectory via three quantities: radius of gyration (R_g), end to end distance (EED), and hinge angle. R_g values were calculated using the built in method offered by the MDAnalysis package. The EED was a calculated using a geometric distance between the nitrogen atom of K540 and the carboxylic acid carbon atom of D766. The hinge angle was defined as the angle connecting the center of geometry (CG) of each immunoglobulin domain and the CG of the linker. Residues K540 – M628 were used to calculate CG of C4. Residues E639-D766 were used to calculate the CG of C5; the loop region (residues V683-D710) is always excluded to avoid complications due to its presence or absence. The residues between the C4 and C5 domains constituted the linker (E629-Q638). In the cases when the linker was deleted, the CG of the C-terminal residue of C4 and the N-terminal residue of C5 was used to calculate the CG (residues M628 and E639). R_g, EED, and hinge angle data were visualized via histograms and characterized using a Gaussian approximation.

Principal component analysis (PCA) of α carbons was used to characterize the essential dynamics of the constructs. To make comparisons of the PCA results between constructs possible, only α carbons shared between all constructs were used in the analysis (C4 domain and C5 domain without the loop: residues K540- M628, E639-T681, and K711-D766). PCA was performed on trajectories that had been aligned to an initial frame and then to the replicates’ mean structure. This ensures that the analysis captures only the internal dynamics of the construct instead of global translations and rotations. The PCs are calculated by diagonalizing the mean-subtracted covariance matrix for a given trajectory, resulting in eigenvectors (i.e. PCs) and associated eigenvalues. The PCs represent orthogonal displacements of α-carbons about their mean positions throughout the trajectory. The associated eigenvalues indicate how much of a structure’s positional variance during its trajectory can be accounted for by a given PC. The percent accumulated variance was plotted to assess how well the first few PCs captured the total dynamics. Then, each frame of the mean-subtracted trajectory was projected onto a PCs to generate a series of weights. The magnitude of a weight for a given frame indicates how significant the associated PC is in that frame. Pairwise scatter plots and 2D distributions were generated for the weights of the three most significant PCs. 1D distributions were also generated for the three PCs. The distributions were generated using a Gaussian kernel density estimator whose bandwidth was determined using Scott’s rule.

Principle component movements were visualized with VMD plugin NMWiz (Bakan et al., 2011; Humphrey et al., 1996) with the following settings: RMSD = 2.0, draw cutoff: 1.0, arrow cylinder radius: 0.1, arrow cone radius: 0.3, arrow cone height: 1.0, graphics material: AOChalky, graphics resolution: 50. Image rendering and ray tracing were done with the VMD Tachyon ray tracer.

2.12. Interaction network analysis and trajectory visualization

The last 900 ns frame of the three extended replicate C4C5 simulations were extracted and submitted to the server RING 2.0 for the calculation of residue interaction network (Piovesan et al., 2016). The results were manually sorted and organized into interactions that were directly on the interface of the C4 and C5 domains, or those that involved the linker or loop residues. Multiple layers of interaction networks exist in the system, but only those with immediate connection to the latched conformation were considered for this analysis. For each trial, the average bond free energy was estimated with the combined hydrogen bond (Hbond), ionic, van der Waals (vdW) energies were used per software guidelines (Piovesan et al., 2016). In order to determine the surface area of the interacting interface between C4 and C5, the solvent accessible surface area (SASA) of the 0 ns frame and three 900 ns frames of the C4C5 model were individually calculated with the VMD Timeline tool, probe radius of 1.4 Å (Humphrey et al., 1996). The buried surface area of the interacting interface was calculated as (SASA_0ns-SASA_900ns)/2. The detailed residue-level interaction networks were made in PyMOL by selecting all interaction residues and changing the appearance to the cartoon and surface representations with 60% and 80% transparency, respectively. The interaction bonds were manually drawn using the “dist” command, then “hide labels” command. Display settings were optimized for quality and the lighting conditions were changed as follows: ambient (0.66), reflect (0.02), and shininess (2). The desired images were rendered using a simple “draw” setting or ray-traced. For trajectory visualization, representative 900 ns simulations of C4C5, Linker^del, Loop^del, and Double^del models were made into videos using the VMD Movie Plugin and the VideoMach software. After the trajectories were loaded and graphically adjusted in VMD, the time stamps were overlaid in the frame view using a VMD script. The videos were made using the Movie Maker plugin and VideoMach software, with internal Tachyon rendering.

2.13. Data and statistical analysis

Data were reported as mean ± standard deviation (SD) unless mentioned otherwise. Statistical comparisons of the individual C4 and C5 domain measurements were performed using a 2-tailed Student’s t test. Comparisons of the double domain constructs (C4C5, Linker^del, Loop^del, and Double^del) were made using a one-way ANOVA followed by the Tukey Kramer post hoc test. The criterion for statistical significance was set at P < 0.05. All statistical analysis were performed using GraphPad Prism version 6.01 for Windows (GraphPad Software, https://www.graphpad.com) or R (R Core Team (2018) Vienna, Austria. https://www.R-project.org/). Data figures were made by GraphPad Prism or Anaconda distribution of Python 3.7 (Matplotlib and Seaborn plotting libraries). Molecular models and related images were rendered using PyMOL. The trajectory videos were created using VMD (University of Illinois at Urbana-Champaign, Champaign, IL, USA).

2.14. Accession numbers

The accession numbers reported in this paper includes the solution NMR structures of mouse fast MyBPC C4 domain (PDB ID: 2DLT) (to be published) and human cardiac MyBPC C5 domain (PDB ID: 1GXE) (Idowu et al., 2003). Reference sequences used in sequence analysis of slow, fast, and cardiac MyBPC are Homo sapiens (NP_002456.2, NP_004524.3, NP_000247.2), Mus musculus (NP_001239301.1, NP_666301.2, UniProtKB O70468), Rattus norvegicus (NP_001094228.2, NP_001099727.1, NP_001099960.1), Felis catus (XP_023113138.1, XP_006941016.1, XP_019667955.1), Canis lupus familiaris (XP_022259120.1, XP_533608.2, NP_001041571.1), Sus scrofa (XP_005664259.1, XP_003127419.3, XP_020940874.1), and Bos taurus (NP_001104243.1, XP_024834509.1, NP_001070004.1).

3. Results

3.1. Sequence alignment and analysis

The sequence alignment of the three MyBPC isoforms from seven species of animals showed that there is a linker region between domains C4 and C5 that is absent in the slow isoforms, and a loop region in domain C5 only present in the cardiac isoform (Fig. 1). Due to variations in sequence alignment algorithms, studies have defined the positions of the linker and loop regions differently (Idowu et al., 2003; Oakley et al., 2004). We chose to define the linker as EVKIDFVPRQE and the loop as VTQGKKASAGPHPDAPEDAGADEEWVFD (O70468), according to the Idowu et al., 2003 study. The linker region is 100% conserved between the seven different species in both types of striated muscle, and the sequence for the fast skeletal muscle type (EIKVEYVPKQ) and cardiac muscle type (EVKIDFVPRQ) have neutral pIs of 6.14 and 6.04 respectively. However, only 29% (8/28) of the cardiac specific loop region residues are conserved, with a consensus sequence of **T***K****P***A********EWVF, where * indicates difference. Human and mouse loop regions are both negatively charged (−4) with pIs of 4.29 and 3.96, respectively. All sequences used in the alignment are included in Fig. S1. The evolutionarily conserved linker sequence may indicate structural or functional importance, and the loop may have a role in isoform-specific muscle function or regulation.

3.2. Recombinant protein expression, purification and characterization

In order to study the C4 and C5 domains separately, as well as together with truncation mutations in the linker and the loop regions, single domain constructs (C4 and C5) and double domain constructs were expressed and purified. The C4C5, Linker^del, Loop^del, and Double^del proteins correspond to wild-type C4C5 fragment, mutant with linker deletion, mutant with loop deletion, and mutant with both linker and loop deletion, respectively. A summary of the C4C5 constructs used in this study is shown in Fig. 2.

Fig. 2.

Schematic representation of the six cMyBPC C4C5 constructs used in this study. The Met-His₆ represents the N-terminal His₆-tags. The linker is represented by the curved line connecting the C4 and C5 domains. The loop is represented by the elongated curve on the C5 domain.

The specific murine C4, C5, C4C5, Linker^del, Loop^del, and Double^del sequences used in gene synthesis and protein expression in this study are shown in Fig. S2 with each of their protein parameters calculated from the ProtParam tool (Table S1). Although all of the C4C5 protein constructs were highly pure after purification with Ni-NTA resin, size exclusion chromatography (SEC) was used to exchange buffer to HBS (Fig. 3).

Fig. 3.

Averaged size exclusion chromatography elution profile for the A. single domain constructs (C4 and C5) and B. the double domain C4C5 constructs, both with the Bio-Rad size exclusion standards. The dotted lines represent the retention volumes of lyophilized mixtures of thyroglobulin (670 kDa)/bovine γ-globulin (158 kDa) merged together, chicken ovalbumin (44 kDa), and equine myoglobin (17 kDa).

Each fraction from the size exclusion chromatography was collected and run on SDS-PAGE, which showed that most of the peaks contained mainly C4C5 constructs (Fig. S3). Therefore, the multiple peaks seen in the FPLC runs are likely to be oligomerized C4C5 proteins. When the multimer peak was gathered and ran on the FPLC again, it showed the same pattern of multiple peaks- suggesting that the C4C5 constructs were in dynamic equilibrium among many oligomeric forms (data not shown). Another likely reason for the wide FPLC peak is due to heavy loading, since the purification process yielded greater than 1–2 mg of protein. The molecular weights of the constructs were similar to predicted values from the respective amino acid sequences, at around 10–15 kDa for the single domains and 25 kDa for the double domains (Table S1, Fig. S3). We confirmed that each of the final purification products for C4, C5, C4C5, Linker^del, Loop^del, and Double^del were greater than 90–95% pure by SDS-PAGE (Fig. 4). Finally, the combined final purification products were probed for the His₆-tag using a Western blot to confirm that the proteins we isolated were indeed the C4C5 constructs (Fig. S4).

Fig. 4.

Representative products of the protein purification process. The single domain (C4 and C5 on the left) and double domain (C4C5, Linker^del, Loop^del, and Double^del on the right) recombinant proteins are shown with the molecular marker (M), the lysate pellet after homogenization (L), the combined eluents from Ni-NTA resin (E), and the final concentrated SEC FPLC proteins (F).

Due to the very wide FPLC peaks that each C4C5 domain construct displayed, size exclusion chromatography-multiple angle light scattering (SEC-MALS) was done to determine the oligomerization state of the C4C5 domain constructs and the size of the proteins (Fig. 5). The results showed that the C4 domain was relatively monomeric with a monodisperse peak. However, the C5 domain showed polydispersity up to ~25 kDa, showing potential dimer formation and higher molecular weight impurities. All double domain constructs showed oligomerization as well, ranging from dimers and trimers to higher oligomers. Based on these observations, it seems that the C5 domain confers a tendency to form oligomers. The molecular mass determined by SEC-MALS experiments is summarized in Table 1.

Fig. 5.

Representative SEC-MALS for each study construct showing their normalized differential refractive index (dRI) and molar mass (kDa) vs. retention volume (mL).

Table 1.

SEC-MALS based molecular weight analysis.

Construct	Loading concentration (μM)	Molar mass of monomeric peak (kDa)	Predicted molar mass (kDa)
C4	150	12.0 ± 3.3%	10.8
C5	150	14.9 ± 1.6%	14.7
C4C5	50	22.3 ± 3.7%	25.9
Linkerdel	50	32.0 ± 3.9%	24.7
Loopdel	50	33.6 ± 6.2%	23.0
Doubledel	50	26.5 ± 6.8%	21.8

3.3. C4C5 domain secondary structure determination

Overall, far UV circular dichroism (CD) experiments showed proper folding of all of the recombinant proteins and appropriate dynode voltages across the relevant wavelength ranges (Fig. 6). Since different fitting algorithms produce different secondary structure results, a consensus average % of secondary structure using BeStSel, CDSSTR, SELCON3, and CONTIL-LL was calculated for easier visualization (Table 2). The raw secondary structure quantification results from BeStSel, CDSSTR with set 6 and 7, SELCON3 with set 6 and 7, and CONTIN-LL with set 7 are reported in Table S2 (Whitmore and Wallace, 2004).

Fig. 6.

Averaged far UV circular dichroism spectra of the single and double domain constructs normalized to trial-matched control buffer spectra with corresponding dynode high tension voltage. Values are expressed as mean residue ellipticity ± SEM from three independent trials.

Table 2.

Consensus estimates of secondary structures using BeStSel and sets 6 and 7 of CDSSTR, SELCON3, and CONTIL-LL combined.

Groups	Helix	Strand	Turn	Others/Unordered
C4	5.3 ± 0.7	42.3 ± 1.6	22.0 ± 2.3	30.9 ± 1.3
C5	7.1 ± 0.6	29.6 ± 0.9*	19.0 ± 1.2	44.6 ± 0.8*
C4C5	9.8 ± 1.1	34.9 ± 1.3	19.9 ± 1.8	35.0 ± 1.6
Linkerdel	7.0 ± 0.8	30.7 ± 1.3	19.5 ± 1.4	42.7 ± 0.9
Loopdel	10.5 ± 1.2	32.8 ± 0.9	20.1 ± 1.6	36.2 ± 1.5
Doubledel	10.2 ± 1.1	30.3 ± 0.5	19.5 ± 1.4	40.4 ± 1.6

Values are expressed as mean ± SEM from three independent replicates of seven separate algorithms. Consensus secondary structures were calculated by combining sub-classifications of helices and strands into single entries for easy comparisons. Statistical comparisons of the individual C4 and C5 domain measurements were performed using a 2-tailed Student’s t test.

^*Significant difference (P < 0.05) from C4. Comparisons of the double domain constructs (C4C5, Linker^del, Loop^del, and Double^del) were performed using a one-way ANOVA followed by the Tukey Kramer post hoc test. The pairwise comparison results are provided in Table S3.

The CD spectra of all constructs showed similar trends: they had mostly β-strands (positive band at 195 nm and negative band at 218 nm) and some characteristics of α-helices (positive band at 193 nm and negative bands at 222 nm and 208 nm) (Fig. 6) (Greenfield, 2006b). C4 showed the most β-strand content, as well as the narrowest negative band at 218 nm (Fig. 6, Table 2). On the other hand, C5 showed lower ellipticity above 210 nm and negative bands near 195 nm which are characteristics of irregular secondary structures (Fig. 6). This finding is consistent with other studies that C4 alone consists of mostly β-strands, and the loop region in the C5 domain may alter the overall secondary structure to be more disordered (Guardiani et al., 2008; Idowu et al., 2003; Suay-Corredera et al., 2021). C4 and C5 had statistically significant differences in the contents of β-strands (42.3 ± 1.6% in C4 vs. 29.6 ± 0.9% in C5, p < 0.01) and disordered structures (30.9 ± 1.3% in C4 vs. 44.6 ± 0.8% in C5, p < 0.01) (Table 2). Interestingly, the weighted average of the C4 and C5 spectra is not identical to the combined C4C5 domains. This may be due to differences in the conformation of the combined domain construct from the single domain counterparts stemming from interdomain interactions, which we will explore further (Klimtchuk et al., 2010).

The CD spectra of C4C5, Loop^del, and Double^del were all qualitatively very similar, but the Linker^del construct was somewhat different than the other 3 proteins. Loop^del and Double^del constructs predictably contained more β-strands and less disordered structure, such as seen in the isolated C4 construct. However, C4C5 with the loop region would be predicted to make the combined C4 and C5 system more disordered. Instead, we saw that C4C5 contained β-strands on par with the constructs without the loop, and contained less disordered secondary structures compared to the Linker^del construct (Fig. 6). The double domain constructs were compared using a one-way ANOVA, and this analysis showed that C4C5 had significantly less disordered structures (35.0 ± 1.6% in C4C5 vs. 42.7 ± 0.9% in Linker^del, p = 0.03) (Table S3). This suggests that the linker region (the only difference between C4C5 and Linker^del) may play a role in reducing the overall disordered secondary structures. Despite qualitative differences, none of the pairwise comparisons of the Linker^del and Double^del showed significant differences. This may indicate that the loop may have a localized single domain effect, but does not cause a measureable global double domain change in secondary structure (Table S3).

3.4. Evaluation of thermal stability through denaturation experiments

In order to understand the role that the linker and the loop may have on the thermal stability of C4C5, two sets of experiments were performed: differential scanning fluorimetry (DSF) and circular dichroism (CD) thermal denaturation/renaturation. CD thermal denaturation involves monitoring the change in ellipticity or secondary structure as the protein sample is heated, while DSF utilizes the fact that more hydrophobic residues become exposed as the protein unfolds and measures the amount of hydrophobic dye that binds and fluoresces. Experimental DSF melt curves were normalized to control solvent melt curves to eliminate the potentially confounding contribution of excess dye binding to the solvent. These orthogonal techniques were used to gain an understanding of the thermal characteristics of the C4C5 and mutant proteins.

The DSF melting experiments gave insight into the potential unfolding mechanism of the C4C5 domains. For our primary analysis, the Boltzmann fit method was used to obtain the overall thermal melting temperature (T_B) (Fig. 7 A–E, Table 3, Table S4). The individual single domain constructs seem to have monophasic melting processes (Fig. 7A) with significantly different melting temperatures (51.4 ± 1.2 for C4 and 40.1 ± 3.2 for C5, p < 0.01) (Fig. 7E, Table 3). The C4C5 and Loop^del constructs, which contain intact linker regions, were found to have high initial signal, indicating that the linker may provide many hydrophobic pockets for the dye to interact with, given that 50% of the residues in the region are hydrophobic (Fig. 7C). Additionally, constructs without the linker (Linker^del and Double^del) showed a more cooperative melting transition. This may indicate that the absence of the linker causes both domains to unfold together.

Fig. 7.

Differential scanning fluorimetry. The normalized thermal melt fluorescence curve (A.U.) of the single domain (A) and double domain (C) constructs and the numerical derivative traces of the respective melting curves (B, D) are shown. The scatter and box-whisker plots of T_B (E) and the biphasic T_D1 and T_D2 data (F) are presented as mean ± SD.

Table 3.

Summary of all types of melting temperatures measured by differential scanning fluorimetry.

Groups	TB (°C)	TD1 (°C)	TD2 (°C)
C4	51.4 ± 1.2(34)	52.2 ± 1.3(34)	NA
C5	40.0 ± 3.2(31)*	40.6 ± 3.5(31)*	NA
C4C5	46.5 ± 0.8(37)	44.3 ± 0.9(36)	52.1 ± 1.6(9)
Linkerdel	47.9 ± 0.5(37)	49.9 ± 1.0(37)	NA
Loopdel	45.6 ± 0.6(37)	44.8 ± 0.3(37)	55.6 ± 2.2(6)
Doubledel	47.6 ± 0.6(37)	49.5 ± 0.2(37)	NA

Values are expressed as mean ± SD (N). Statistical comparisons of the individual C4 and C5 domain measurements were performed using a 2-tailed Student’s t test.

^*Significant difference (P < 0.05) from C4. Comparisons of the double domain constructs (C4C5, Linker^del, Loop^del, and Double^del) were performed using a one-way ANOVA followed by the Tukey Kramer post hoc test. The pairwise comparison results are provided in Table S4.

The constructs with the linker region showed biphasic melting behavior (Fig. 7 C, D). However, due to the small magnitude of the normalized fluorescence slopes, a precise interpretation of these results is challenging (Fig. 7F, Table 3). In brief, multi peak melting temperature (T_D) analysis using the derivative method showed that the biphasic mean melting temperatures, T_D1 and T_D2 were 44.3 ± 0.9 °C (n = 36) and 52.1 ± 1.6 °C (n = 9) for C4C5 and 44.8 ± 0.3 °C (n = 37) and 55.6 ± 2.2 °C (n = 6) for Loop^del (Fig. 7F, Table 3, Table S4). One data point from the C4C5 group was excluded due an ambiguous three phase melting temperature measurement. As implied by the smaller sample size of the second peaks, not all traces showed calculable biphasic melting. Based on our individual C4 and C5 domain melting curves, the first and second peaks potentially represent C5 and C4 denaturation, respectively. This finding may be attributable to the primary structural separation of the C4 and C5 domains by the linker region. For constructs without the linker region, the two domains seem to melt together in a single monophasic relationship.

In order to study the thermal stability of the C4C5 domain constructs using an orthogonal technique, CD denaturation experiments were done to measure the changes in MRE with temperature in the range of 25 °C to 101 °C. A ramp speed of 1 °C/min and data pitch of 2 °C were used to ensure that the melting speed was slow enough for protein denaturation processes, and the temperature was ramped up to 101 °C to ensure complete unfolding (Fig. 8). An estimated overall melting temperature for each construct was obtained at 230 nm (Fig. 8, Table 4). The melting temperature trends seen between the C4 and C5 constructs are consistent with the DSF data, as well as previous reports (Suay-Corredera et al., 2021). There were no statistical differences in melting temperatures among the double domain constructs. The melting temperatures of the double constructs were generally higher than the data obtained from DSF experiments which may be due to the differences in buffer conditions (Table 4).

Fig. 8.

Average melting (in red) and cooling (in blue) curves monitored at 230 nm were plotted with a corresponding Boltzmann sigmoid fit. An overlay of the complete circular dichroism spectra from the full temperature range of the melting and cooling experiments are also shown for each construct (25, 35, 39, 43, 47, 51, 55, 59, 63, 67, 75, 91, and 101 °C). The arrows show the direction of spectral changes during the melting process. The Boltzmann sigmoidal fit of the cooling curves are not shown. Values are expressed as mean ± SEM from three independent replicates.

Table 4.

Fitted melting parameters from circular dichroism experiments monitored at 230 nm.

Groups	Tm	Adjusted R2
C4	54.0 ± 0.1	0.996
C5	45.0 ± 1.0*	0.993
C4C5	54.4 ± 0.1	0.987
Linkerdel	53.6 ± 0.7	0.994
Loopdel	54.8 ± 0.3	0.989
Doubledel	52.8 ± 0.4	0.979

Values are expressed as mean ± SEM. Statistical comparisons of the individual C4 and C5 domain measurements were performed using a 2-tailed Student’s t test.

^*Significant difference (P < 0.05) from C4. Comparisons of the double domain constructs (C4C5, Linker^del, Loop^del, and Double^del) were performed using a one-way ANOVA followed by the Tukey Kramer post hoc test, however no significant differences were found in any of the multiple comparisons.

Although the selection and monitoring of one wavelength as a proxy for protein unfolding has been done previously by many, we acknowledge that changes in one wavelength may not capture the complex structural changes that occurs during thermal melting and cooling. Thus, we calculated secondary structure basis components that utilize information from multiple different wavelengths for the entire melting and cooling process, so that the temperature-dependent changes in secondary structures could be explored (Fig. S5). The results show that each secondary structure basis component has its own unique melting temperatures, and some undergo changes without an identifiable sigmoidal relationship that can be fitted well to a Boltzmann sigmoidal relationship (Fig. S5). However, those with a fairly robust fit produced trends consistent with melting temperatures obtained at 230 nm.

Additionally, an important consideration in characterizing protein melting temperature is whether it is reversible or irreversible. The renaturation curves (blue in Fig. 8) for all constructs showed greater than 50% refolding, but much of the cooperativity of melting/cooling was lost and none of the constructs returned completely back to the starting MRE at 230 nm. Furthermore, a qualitative inspection of all of the constructs showed that the initial CD spectra at 25 °C was different from the final one at 25 °C after melting/cooling. From the detailed secondary structure basis component analysis, it was also evident that some secondary structures returned to their original % structural content; however, not all did (Fig. S5). There may also be cooling-induced structure formation that makes the interpretation of renaturation difficult. With the experimental conditions and techniques that we have utilized, the melting and cooling process of the C4C5 constructs is complex due to varied type and degree of heat or cold-induced destruction/formation of select secondary structures. Overall, the six C4C5 constructs seem to undergo at least a partially irreversible melting process, which may indicate that there is aggregation or rearrangement/formation of alternate side chain conformations. Although it is uncertain whether this occurs during the melting process or at the maximum temperature, the possibility of aggregation can skew the estimation of melting temperatures, and the interpretation of the melting temperature data warrants a more careful consideration.

3.5. Comparison of the C4C5 and truncation mutation models with molecular modeling and dynamics simulation

Molecular dynamics simulations (MDS) were run for each of the four double domain C4C5 models. These models were started from a straight conformation (similar to the schematic view in Fig. 2) and given time to adequately equilibrate. There were no signs of denaturation, evidenced by a stable root mean squared deviation (RMSD) time series (Fig. 9A). The relatively long time to reach steady state is attributable to the high number of residues simulated and the local and global flexibility provided by the loop and the linker regions, respectively. The linker and loop structures do not seem to significantly contribute to the domain integrity, as all truncation models were stable. Steady state was reached around 500 ns, but the simulations were extended to 700 ns to guarantee that the protein structures were fully stable. After this, a triplicate of 200 ns trials was done (3 × 200 ns for each model) and used for subsequent geometric analyses in order to capture structural changes in the fully equilibrated models. The root mean squared fluctuations (RMSF) per residue for each of the constructs showed a generalized increase in fluctuations in the Linker^del and Double^del models where the linker was deleted. The fluctuations in the loop region were decreased in the C4C5 model compared to the Linker^del model, indicating the stabilization of the loop (Fig. 9B).

Fig. 9.

A. Root mean squared deviation (RMSD) over the 700 ns simulation of four C4C5 models. The high deviations in the C4C5 model are due to the large bending movement from an initially straight molecule of the C4 and C5 domains. B. The average per-residue root mean squared fluctuations (RMSF) from triplicates of 200 ns simulations were calculated for each model. The residues in the linker and loop regions are highlighted in gray.

With manual inspection of the trajectories, we noticed a striking finding. The C4C5 model bent completely onto itself with the linker acting as a hinge, and the loop latching onto the bent conformation (Supplemental Video 1). In order to quantitatively study this phenomena, we looked at geometric measurements as well as principle component analyses. First, protein compactness or mean radius of gyration (R_g) ± SD was calculated from the 200 ns molecular dynamics simulation (MDS) replicates. Based on the R_g data, the linker region appears responsible for the C4C5 domains’ ability to form a more compact conformation, as demonstrated by lower overall Rg values for constructs that contain the linker. However, the loop helps in increasing the compactness of C4C5 in the presence of the linker evidenced by a lower R_g in C4C5 than Loop^del (19.5 ± 0.5 Å in C4C5 vs. 22.5 ± 2.4 Å in Loop^del, p = 0.01) (Fig. 10A, Table 5, Table S5). Additionally, the loop alone without the linker does not seem to make the two domains more compact (25.2 ± 1.3 Å in Linker^del vs. 24.9 ± 1.3 Å in Double^del, ns) (Fig. 10A, Table 5, Table S5). The loop region without the presence of the linker is likely free to become flexible and move dynamically.

Fig. 10.

The pooled data from three separate 200 ns simulations were used to calculate a combined histogram of (A) the radius of gyration, (B) end to end distance as defined in the methods section, and (C) the angle between the C4 and C5 domains.

Table 5.

Summary of the geometric measurements from the C4C5 domain simulations.

	Rg (Å)	EED (Å)	Angle (deg, °)
C4C5	19.5 ± 0.5	29.3 ± 4.0	49.8 ± 3.2
Linkerdel	25.2 ± 1.3	78.1 ± 6.6	143.3 ± 16.8
Loopdel	22.5 ± 2.4	63.9 ± 9.7	84.2 ± 18.5
Doubledel	24.9 ± 1.3	79.2 ± 7.1	145.5 ± 15.8

The mean ± SD of geometric analysis parameter measurements were obtained from combining three 200 ns replicate simulations. Statistical comparisons of the double domain constructs (C4C5, Linker^del, Loop^del, and Double^del) were performed using a one-way ANOVA followed by the Tukey Kramer post hoc test. The pairwise comparison results are provided in Table S5.

In order to further study the dynamics of the C4C5 domains and the effect of regional mutations, the mean end to end distance (EED) ± SD of the C4C5 domains (defined as the distance between the backbone nitrogen atom of the first residue to the carbon atom of the last residue) was calculated over time. For the fully equilibrated structures, the EED is smallest in the C4C5 model (29.3 ± 4.0 Å), intermediate in the Loop^del model (63.9 ± 9.7 Å), and largest in Linker^del (78.1 ± 6.6 Å) and Double^del (79.2 ± 7.1 Å) models (Fig. 10B). All pairwise comparisons of the EED between groups were statistically significant except for the Linker^del and Double^del pair (Table 5, Table S5). These data again suggest that the linker is essential in initial hinge formation, and that the loop facilitates C5 to latch onto C4 and maintain a completely bent conformation. In addition, the angle between the C4 and C5 domains (measured as the angle formed from the centers of geometries of C4, linker and C5 without the loop region) showed that C4C5 achieved the smallest mean ± SD angle of 49.8 ± 3.2°, while the Loop^del (84.2 ± 18.5°), Linker^del (143.3 ± 16.8°) and Double^del (145.5 ± 15.8°) models showed larger equilibrium angles (Fig. 10C, Table 5, Table S5). All pairwise comparisons of angles between groups were statistically significant except for the Linker^del and Double^del pair (Table S5). The intact C4C5 model forms a tightly bent conformation based on the angle measurements as well as the Rg and EED measurements. In contrast, Loop^del is able to close about 77%, while both Linker^del and Double^del only close an average of 21% normalized to the mean angle achieved in the C4C5 model.

Next, principal component analysis (PCA) was performed on the motions of the conserved α-carbons between the four constructs from three pooled trajectory replicates to identify and examine their essential dynamics. The first several PCs of each construct, accounting for 60–85% of the total variances, are visualized in Fig. 11 and Fig. S6. As shown in Fig. 11C, the first two PCs of the Linker^del and Double^del represent twisting motions of the C4 and C5 domains about their central joint. The diffuse scatter and contour plots generated by projecting the Linker^del and Double^del trajectories onto these PCs show that the constructs simply fluctuate about their mean structures without adopting a distinct conformation (Fig. 11A). By contrast, projecting the trajectories of the Loop^del construct onto its first two PCs show a primary clustering about a bent conformation and a less prevalent unbent conformation further along the first PC (Fig. 11A). The presence of these distinct bent and unbent conformations compared to the Linker^del and Double^del’s ill-defined conformation suggests that the linker indeed acts as a hinge for the C4 and C5 domains. In addition to the hinge motion, the Loop^del’s domains also twist with respect to one another as demonstrated by movement along the second PC (Fig. 11C). Finally, projecting the C4C5 trajectories onto its first two PCs revealed remarkably constrained dynamics about a single tightly bent structure. In fact, C4C5 had the lowest dynamic range among all four constructs and its PCs merely reflect relatively small and insignificant fluctuations. This is further reflected by the fact that each of C4C5’s PCs account for less of the total variance compared to the other constructs (Fig. 11B). Importantly, these results show evidence that the linker enables the bent conformation of C4C5 and the loop strengthens and stabilizes the conformation.

Fig. 11.

A. The two main principle components PC1 and PC2 are visualized as scatter plots (top) and contour plots (bottom). The PC plots were obtained from three 200 ns simulation replicates. The C4C5 model shows tight clustering indicating very limited motion due to a stable bent conformation, whereas the Linker^del or Double^del models shows a very wide clustering of the principle components. The Loop^del model shows fairly narrow clustering representing a less stable bent conformation, with an island of clustering where the linker region allowed for a more straight conformation. B. The cumulative variance curve for the principle component analysis of the four double domain C4C5 constructs. The analysis used the pooled data from three 200 ns replicate simulations. C. Representative movements of the first and second principal components of the four C4C5 models, superimposed on the mean structures. The Loop^del model shows the most collective bending motion, while the C4C5 model showed a fairly small and rigid movement in various directions. The PCA analysis shows that the linker enables a primarily bending motion seen in the Loop^del model.

3.6. Exploration of the bent conformation of the C4C5 model

Since the bent conformations of the C4C5 models were stable for more than 200 ns, the residue-level interaction landscape was evaluated for interface hydrogen bonds (Hbonds), van der Waals (vdW) interactions, and salt bridges. The analysis of the final conformation of the three 900 ns C4C5 simulation replicates showed that the folding interface contained an average of 4.3 ± 1.5 hydrogen bonds, 1.3 ± 1.5 ionic bonds, and 14.7 ± 3.2 vdW interactions. No π- π stack or π-cation interactions were found. Each of the above interactions contributed to an overall average bond free energy of 188.3 ± 60.3 kJ/mol to the interface (Piovesan et al., 2016). Although the bond free energy is quite high, this is an estimation according to the average bond strength of each interaction. The buried surface area of the bent conformation compared to the starting open conformation is 88.0 ± 85.1 Ų, which is not quite high enough to form a stable complex, which requires >~500 Ų (Day et al., 2012). Additionally, the specific interaction partner analysis shows that most interactions are simulation-dependent, and there are only a few identical interaction partners that form in multiple replicates. Despite the fact that most interaction partners are not conserved amongst the end conformation replicates, it is still insightful to observe the locations and different types of interactions that can form the interface between C4 and C5 (Table S6).

Intriguingly, the majority of the interactions occurring at the hinge (near the linker residues) were van der Waals and hydrophobic interactions. In contrast, the interactions with higher average bond energies, like Hbonds and ionic interactions, predominated in the “latching” interface, where the C4 domain and the loop interfaced (Fig. 12). Furthermore, the C4C5 model’s simulation trajectory showed that the hydrophobic, linker-associated interactions formed first, then the two domains came closer together with the loop latching onto the C4 domain subsequently (Supplemental Video 1). Although more in vitro experiments need to be done to determine the precise dynamic mechanisms of the C4C5 domains, the sequence of events described above may be one of the possible bending pathways.

Fig. 12.

A-C. Detailed atomic-level description of the interface interaction residues of the final 900 ns frame conformation of the replicate C4C5 simulation trials 1 (A), 2 (B), and 3 (C). A list of all of the interactions found in each trial of the C4C5 simulations shown here is found in Table S6. D. A schematic diagram of the interaction location and type of bonds formed by the C4C5’s bent conformation interface. The simulation results indicated three main interfaces of the C4C5’s bent conformation. The C4-Loop interface consisted of mixed Hbond, ionic, and vdW interactions (14 combined across three end states). The C4 and C5 domain interface consisted of the most number of interacting residues (34), more vdW pairs than the C4-Loop interface, and a substantial amount of HBonds. The linker interface was mainly vdW hydrophobic interactions with itself or the neighboring C4 and C5 residues (12 combined interaction pairs).

In order to study whether the starting conformation impacted the ability of the C4C5 domains to fold into the above described bent conformation, we conducted two more replicate experiments, in addition to the initial first trial, with manually adjusted models that had varying start position angles compared to the initial “straight” model we have used (represented in Fig. 2). After 500 ns simulations and a 200 ns extension to ensure full equilibration (a total of 700 ns), basic geometric and PCA analyses were done on the last 200 ns segments. We observed that constructs that started in different conformations returned to the initial first model’s more “straight” conformation after some time (Supplemental Videos 2–4). At the end of the 700 ns simulation, the geometric results showed that the second trial was close to reaching the fully bent conformation (Fig. 13). On the other hand, the third model bent in the opposite direction – almost 360° away from original bent conformation studied in Fig. 12. Although the replicate simulations did not reach the fully bent conformations, the various completely bent conformations may simply be inaccessible with the limitations of the simulation time. However, as shown previously in the C4C5 model’s trajectory video (Supplemental Video 1) and the large range of RMSD (Fig. 9A), the 10 residue flexible linker confers the C4C5 domain exceptional flexibility, and provides a great degree of movement. Therefore, based on the replicate runs of the C4C5 models, we believe that our initial bent C4C5 conformation with the specific interaction networks likely represents only one of many potential bent conformations that are possible from the diverse hinge orientations (Figs. 11, 13).

Fig. 13.

Simulation-based geometric and principle component analyses for the replicate C4C5 model simulations with differing starting conformations. A. RMSD plots of the three replicate trails of C4C5 with differing starting conformations. B. Histograms of Rg, EED, and hinge angles calculated from the last 200 ns of the C4C5 replicate simulations. C. PC1 and PC2 scatter and contour plots of the three trials of C4C5. D. The cumulative variance calculations of the C4C5 replicate simulations.

4. Discussion

4.1. The linker and loop regions modulate the secondary structure and thermal stability of the C4C5 domains

A prior NMR study performed on just the C5 domain found that the loop region was unstructured, highly dynamic and extended, and destabilizes the domain (Idowu et al., 2003). Additional evidence from a computational study of the isolated C5 domain supported the unstable nature of the C5 domain through kinetic simulations of the unfolding transition (Guardiani et al., 2008). In our secondary structure determination using CD, we also observed that the isolated C5 contained more disordered structures than well-defined β-strands (Fig. 6). However, the destabilizing effect of the C5 loop seems to be a localized-single domain effect, as the double domain constructs without the loop failed to show any significant differences in the distribution of secondary structures (Table 2). Furthermore, it was proposed that the loop is stabilized by contacts with other parts of the protein in order to compensate for the absence of a more conventional CD-loop (Idowu et al., 2003). In fact, we found that the C4C5 construct contained significantly less disordered structures than the Linker^del construct (Table 2), even though they both contain the loop region. Based on our simulation results (Figs. 11 and 12), we surmise that this stabilization and increased ordered structures in C4C5 may come from the bent conformation of the C4C5 domains. When the linker is present, it acts as a hinge in order to facilitate the C4C5 domains to adopt a bent conformation (Fig. 12). The loop region likely stabilizes the C4C5 domains further by latching onto the C4 domain, increasing the proportion of ordered structures that can be formed by the loop (Figs. 6, 12, Table 2).

The impact of the linker and loop regions on the thermal stability of the C4C5 domains seem to be equally explained by our hinge-latch model of the C4C5 domains. Although the multiple pairwise comparisons of the melting temperatures obtained from CD denaturation experiments did not yield any significant differences between groups, the DSF results showed that the Linker^del construct denatured at a significantly higher temperature than C4C5 (Fig. 7, Table 3, Table S4). This may be contradictory to the Idowu et al. study, where they claimed that the peak dispersion typical of folded proteins is absent in the “N-short” C5 construct with the linker deletion. However, the N-short construct used in their study removed the residues RQEPPKI, which are actually the last 2 residues of the linker (RQ) and the first 5 residues of the C5 domain itself based on our domain definitions. Therefore, we believe that their observation of a significant pattern of unfolding via NMR was due to a substantial deletion in the integral part of the C5 domain itself. In our study, the “linker” definition (EVKIDFVPRQ) is different and shifted more to the N-terminal end compared to the linker definition of Idowu et al. (RQEPPKI). In addition, the DSF experiments showed that the effect of the loop on domain stability may differ based on the presence of the two domain system in our study (C4C5). One potential explanation for the lower melting temperature of C4C5 is that as the temperature is increased, the bent conformation which is formed by only a few interaction partners and shallow interface surface area (Fig. 12) will be broken. Thus, the excess motion that is possible with the flexible C4C5 linker provides more opportunities for secondary structure unfolding of C4C5 at lower temperatures. In contrast, the more rigid Linker^del construct may be more resistant to unfolding through excessive linker movements. In fact, deletion of the linker seems to be associated with a monophasic melting curve, without two separate domain unfolding pathways, and a generally higher melting temperature (Fig. 7, Table 3, Table S4).

4.2. The linker and loop regions participate in a hinge and latch mechanism to mediate and stabilize cMyBPC’s bent conformation

A key observation from rotary shadowing experiments is that both the skeletal and cardiac isoforms of MyBPC can adopt sharply bent V-shaped conformations (Hartzell and Sale, 1985; Previs et al., 2016; Swan and Fischman, 1986). A hypothesis implied by these studies is that the flexible ten amino acid linker region between the C4 and C5 domains present in both the above isoforms acts as a hinge to enable the tight bending required for the V-shaped conformations. Indeed, our MD simulations of various C4C5 constructs support the aforementioned notion that the linker functions as a hinge; constructs without the linker (Linker^del and Double^del) were unable to form as tight of angles between the C4 and C5 domains compared to constructs with the linker present (C4C5 and Loop^del) (Figs. 10, 11).

By contrast, the cardiac-specific loop region’s role and how it alters cMyBPC function from skeletal isoforms remains poorly understood. Some have speculated that the loop may be a regulatory site of cMyBPC function or a docking location for cardiac-specific ligands, but little data exists to support any specific mechanism (Gautel et al., 1995; Musacchio et al., 1994). Based purely on sequence, the Loop^del construct is structurally similar to the fast skeletal form of C4C5 while the Double^del construct is structurally similar to the slow skeletal C4C5. By studying MD simulation of C4C5 constructs with and without the loop region, we provide the first structurally based hypothesis of its functional role in different isoforms. In essence, our simulations indicate that the loop functions as a latch to stabilize C4C5’s bent conformation by interacting with the C4 domain. In doing so, the C4C5 construct achieved a dramatically smaller R_g, EED, and interdomain hinge angle than the Loop^del construct that lacked the loop region (Fig. 10). Although only two replicates of the C4C5 models reached or started to reach the same bent conformation, it provides evidence that this extreme bending is physically possible and can occur (Supplemental Videos 1–4, Fig. 12). As further evidence of the linker region’s stabilizing role, the C4C5 construct had an extremely constrained dynamic range compared to the Loop^del, which still exhibited hinge and twisting motions between its domains (Fig. 11). However, it is important to note that our simulations also suggest that the loop region requires the linker region to perform the proposed latch function. With the loop present but the linker deleted (i.e., the Linker^del construct), the RMSF data (Fig. 9B) shows that the loop is much floppier than in the C4C5 construct because the loop cannot reach the C4 domain and be stabilized. Furthermore, the PC analysis shows that the Linker^del model’s dynamics look the same as if both the loop and linker were deleted, as in the Double^del construct (Fig. 11, Fig. S6).

Our hypothesized hinge and latch mechanism can explain previously unexplored differences between skeletal and cardiac cMyBPC isoforms seen in rotary shadowing studies (Hartzell and Sale, 1985; Swan and Fischman, 1986). For instance, while both cardiac and skeletal MyBPC molecules were observed in V-shaped conformations, a more significant population of skeletal MyBPC molecules were seen in the less sharply bent, U-shaped conformations. Interestingly, the U-shaped confirmation was not observed in cardiac MyBPC molecules. Based on our simulation results, we would predict that this is because the skeletal form, similar in structure to our Loop^del construct, lacks the cardiac-specific loop that can latch on to the C4 domain, and strengthen the closure of the hinge. The loop’s presence may also explain why only 32% of skeletal MyBPC molecules were observed in the V-shaped conformation (Swan and Fischman, 1986) compared to 58% of cardiac MyBPC molecules (Hartzell and Sale, 1985). The rotary shadowing studies also found isoform-dependent differences in the V-shaped conformations’ bending angles: 76 ± 23° in skeletal MyBPC (Swan and Fischman, 1986) versus the tighter 59.1 ± 19° in cardiac MyBPC (Hartzell and Sale, 1985). Remarkably, the experimentally determined skeletal and cardiac MyBPC angles agree within error with our interdomain angle calculations from the Loop^del and C4C5 construct, respectively (Table 5).

Taken together, our findings strongly suggest that differences in hinge flexibility as mediated by the presence of the cardiac-specific loop region may be a critical factor in defining the differential functions of the skeletal and cardiac MyBPC isoforms.

4.3. The proposed hinge and latch mechanism of cMyBPC’s central domains may be important in dynamic regulation of cardiac function

cMyBPC is known to interact with and to modulate various important contractile proteins such as actin and myosin (Heling et al., 2020), and the proposed hinge and latch mechanism of the C4C5 domains may fine-tune the reach and movement of cMyBPC within the cardiac sarcomere lattice. Previous studies have estimated that the surface-to-surface spacing of adjacent thick and thin filaments is 9–16 nm depending on sarcomere lengths of 3.6–1.9 μm (Julian et al., 1978). Electron tomography studies of intact muscle sarcomeres showed that cMyBPC is able to span the myosin-actin interfilament space (Luther et al., 2011). Given that the 10 amino acids that form the linker is thought to be a novel extension to a conventional IgI domain structure (Oakley et al., 2004), the large dynamic translational and hinge movements of cMyBPC modulated by the linker region (and indirectly by the latch mechanism of the cardiac specific loop region) appear plausible. In addition, as seen in the I-band of titin, multiple Ig domains may extend by straightening out their interdomain linkers, without the need to completely unfold the Ig domains themselves (Granzier and Labeit, 2004; Linke et al., 1999). Because the striated muscle-specific linker seems to be able to modify the reach and movement of fMyBPC and cMyBPC, it may contribute to a similar intrinsic molecular elasticity and stretch-sensing, as observed in the multiple tandem Ig domains in titin (von Castelmur et al., 2008). This type of physical extension of the MyBPC molecule, or a stretch-sensing mechanism could modulate changes in sarcomere length that occur in vivo during systole and diastole.

In contrast, the linker region could also function indirectly by providing extra extension or retraction back to aid other regulatory regions of cMyBPC, such as the N-terminal domains. It is thought that the orientation of the N-terminal of cMyBPC is facilitated by the presence of the internal hinges in cMyBPC that are present in the C1-C2 and C4-C5 domains (Colson et al., 2016; Previs et al., 2016). Since the N-terminal domains of cMyBPC influence actin rotational dynamics, the large movements of cMyBPC enabled by the hinge domains may directly affect the formation of myosin-actin cross bridges, thus contributing to the dynamic regulation of cardiac function (Colson et al., 2016; Nadvi et al., 2016). Additionally, a hypothetical model of the interaction of cMyBPC with myosin from the Spudich lab showed that the C4 and C5 domains may have interactions with the mesa of the blocked myosin head or interacting heads motif (IHM) and the proximal myosin S2 region (Nag et al., 2017). In fact, the internal domains such as the C4-C6 domains may regulate the SRX state of myosin by stabilizing the IHM (Rahmanseresht et al., 2021). In an in vitro experiment, the population of human β-myosin in the SRX state was shown to be increased in the presence of the C0C7 cMyBPC fragment (Sarkar et al., 2020). A recent microscale thermophoresis study also demonstrated that there is high affinity binding between the central domains of cMyBPC and myosin S1, so the central region of cMyBPC likely acts as an active element in regulating the N-terminal domain dynamics (Ponnam and Kampourakis, 2021). Additionally, various regulatory events affecting the extension or compaction of the linker may transduce those signals all the way to the N-terminal of cMyBPC.

Moreover, C4C5 may also be a site of contractile regulation through ligand binding and other post-translational modifications. For example, there is a confirmed phosphorylation site at S690 in the C5 loop region of mouse cMyBPC, although this serine residue is not present in humans (Huttlin et al., 2010). Due to a relatively high number of proline and alanine residues in the C5 loop, it has been speculated that the loop may also function as a stable scaffold for cardiac-specific ligand interactions of signal transduction molecules, specifically SH3 domains (Gautel et al., 1995; Musacchio et al., 1994). Additionally, it has been shown that in vitro S-glutathionylation of the C4C5 domains of cMyBPC at C627 and C655 increases myofilament Ca²⁺ sensitivity (Patel et al., 2013).

Lastly, the potential functional importance of the C4C5 domains is highlighted by the fact that over 30 HCM- and DCM-causing mutations have been mapped to this region according to the HGMD database (Stenson et al., 2017). Because many of these mutations are expected to form proteins that incorporate into the sarcomere rather than causing haploinsufficiency (Harris et al., 2011), it is likely that mechanisms contributing to cardiac dysfunction involve a disruption in the biological structure and function of the linker and loop regions of these domains. Some of these human disease causing mutations in the C4 domain (M555T, C623X), linker region (Q642X), and C5 domain (P679S, Q689H, N755K, V757M) were found to be some of the interface residues from our C4C5 interaction network analysis (human sequences). Therefore, the mechanism of disease caused by these mutations may potentially be explained by a disruption in the sbility to form the bent conformation of C4C5. In addition to the many mutations present in the C4C5 domains, the potential importance of the loop region in proper cMyBPC function is highlighted by the presence of four HCM-causing mutations in humans (Q689H, A693S, A701T, E710K in human sequence) (Bashyam et al., 2012; Brion et al., 2010; Millat et al., 2010; Olivotto et al., 2008). Of these, Q689H was previously mentioned to participate in the interaction network found in our study, and the A701T and E710K mutations are located at sites that are 100% conserved sites between the seven species studied. While the pathophysiology underlying these mutations has yet to be studied, the change from a negatively to positively charged residue in the E710K mutation, or the possible phosphorylation of the A693S and A701T mutant residues may disrupt native electrostatic interactions required for normal cMyBPC behavior. One possibility is that the change in charge may prevent the loop from stabilizing C4C5’s bent conformation. Alternatively, these mutations could alter the binding affinity between the C4C5 domains and myosin as predicted by the previously mentioned model by the Spudich lab and recent experimental data suggesting that the central cMyBPC domains may be in part responsible for maintaining myosin’s SRX state (Nag et al., 2017; Rahmanseresht et al., 2021).

5. Limitations

Due to the experimental techniques and instruments utilized, there are some potential limitations of our study. The protein concentration calculated from the predicted extinction coefficient and absorbance at 280 nm may be less accurate than using amino acid analysis. As noted above, the overall melting and cooling behavior of the C4C5 constructs is highly complex and it may be impossible to completely eliminate the potential contributions of oligomerization, aggregation, or structural rearrangements of the recombinant proteins in the measurements of the thermal denaturation assays (DSF and CD). In addition, calculations using linear combinations of basis spectra can lead to errors in inaccuracies in CD secondary structure determination. Since the sequence definitions of the linker and loop regions have not been established, partly due to the lack of a crystal structure of both the C4 and C5 domains together, it has been arbitrarily derived from our sequence and computational model analysis. Due to current limitations in computer simulation technologies, it is also possible that the flexible and large protein models did not have the chance to explore all possible conformations in space over the time course used in our simulations. As a result, it is possible that our steady state model is at a local rather than an absolute minimum energy level. Our current study focused on only two domains of cMyBPC in isolation, and since the full-length protein has 9 additional domains linked in series, long-range structural or interactional effects could not be measured directly. An interesting observation from the C4C5 simulation showed that the loop of the C5 domain reaches very close to the N-terminal of the C4 domain, potentially allowing it to contact both the C3 and C6 domains. If there are interactions among the central domains, our simulations were not able to show those possible conformations. Future studies will have to be performed to address these specific possibilities and their potential impact on cMyBPC structure and function, as well as cardiac physiology and pathophysiology.