The patient groups from whom explanted heart tissues were collected included explanted hearts from healthy donors (nonfailing (NF)) and end-stage heart failure patients, who were reported to have either ischemic heart failure (I-HF) or non-ischemic heart failure (NI-HF). The demographics for the patients included information on age, gender, and race. Patients' ages ranged from 41 to 69 years with each condition represented (NF, NI-HF, and I-HF) (Supplementary file 1) and were similarly distributed among the three groups. In an attempt to provide gender and race balance, one female was included in each group and at least one African American (Figure 1). We utilized mass spectrometry to investigate whether PTMs could be found on β-MHC isolated from these hearts and utilized MD simulations to better understand the functional impact these PTMs may have on human cardiac disease presentation. For workflow, see the schematic in Figure 1—figure supplement 1.

Unique high-confidence peptides bearing PTMs were identified using bottom-up mass spectrometry data from NF, ischemic, and nonischemic failing human hearts (Figures 2 and 3). The samples were purified using SDS-PAGE and were digested using in-gel trypsinization. We report eight PTMs: six acetylated lysines and two phosphorylated residues, one serine and one threonine: K34-Ac, K58-Ac, S210-P, K213-Ac, T215-P, K429-Ac, K951-Ac, and K1195-Ac (Tables 1 and 2). Six of these PTMs are distributed throughout the myosin motor domain (Figure 4A) and two within the tail (Figure 4B). The myosin motor is comprised of four domains: the N-terminal domain, the upper and lower 50 kDa domains, and the converter arranged around a central β-sheet (Figure 4A). The structure of myosin gives rise to three functional regions: the actin-binding cleft, which is formed between the upper and lower 50 kDa domains and interacts with the thin filament; the nucleotide-binding pocket, which is comprised of several regulatory loops that coordinate the nucleotide and transmit structural information throughout the structure; and the converter domain, which converts small-amplitude changes in the motor domain into large-amplitude lever arm motions.

A schematic of the β-MHC protein is included in Figure 4—figure supplement 1 and indicates key regions of the protein along with the newly identified PTMs. In Figure 4A, the three-dimensional structure of the human β-MHC sequence complexed with Mn-AMPPNP (PDB: 4DB1) was used to model the PTM sites at K58-Ac and doubly modified peptide K213-Ac/T215-P. This structure represents a post-rigor ATP-bound state of myosin and was used to assess proximity of the modified residues to structurally important regions in the myosin motor domain. An S2 fragment structure (PDB: 2FXM) was used to model K951-Ac.

Residue K34 is located between the SH3-like domain and the converter domain (Figure 4A). K58 is located within the SH3-like domain, which typically serves as a module for protein–protein interactions. Therefore, it is anticipated that acetylation of either K34 or K58 may interfere with normal interactions of this domain. Residues S210, K213, and T215 are sequentially close in the primary structure, and it is interesting to note that they lie together on one face of the myosin motor domain near the ATP-binding pocket in a region called loop 1. This loop is notable for its influence on ATP/ADP cycling (Murphy and Spudich, 1998). It is plausible, therefore, to expect that these modifications may influence ATP binding and/or Pi and ADP-release dynamics (Figure 4A). S210-P was found to exist as a single modification only. The location and dynamics between modifications of residues K213-Ac and T215-P are interesting as they appear to be co-modifications. Although K213-Ac was only found coincident with phosphorylation of T215, T215-P was also found on its own, without K213-Ac. The K213-Ac/T215-P-modified sequence (amino acids 207–234, m/z 766.36) was detectable as an isotope dot product (idotp) with a value below 0.5, and therefore was unquantifiable (Figure 4—figure supplement 2). The acetylated residue K429 is located within the myosin head-like domain at the actin-binding interface. The remainder of the modified residues K951 and K1195 are located within the coiled-coil of the S2 region, although K1195 is missing from the model in Figure 4A. The panels within Figure 4B show closeup views of secondary structural elements and side-chain interactions that neighbor the K58-Ac, K213-Ac, T215-P, and K951-Ac modifications.

To investigate the potential significance of these newly identified PTMs for cardiac function, we assessed their abundance in the human heart samples we analyzed. The relative abundance of the modifications was determined by calculating peak areas of modified peptide and normalizing to IRP peak area (amino acids 1504–1521, m/z 974.49). The MS/MS spectrum of the trypsin-digested common IRP is shown in Figure 5—figure supplement 1, and b-ions indicate N-terminal fragment ions and y-ions indicate C-terminal fragment ions.

Overall, the ratios of modified PTM sites were variable, ranging between 1 and 14 in the NF donor hearts, with a tendency for decreased abundance in failing hearts. In Figure 5, the ratio of modified/IRP is shown for peptides 1 (K34-Ac), 2 (K58-Ac), 3 (K429-Ac), 4 (K951-Ac), and 5 (K1195-Ac). Beneath each histogram are the respective tryptic peptides where the reported PTMs were found. The residues shown in brackets are the trypsin digestion sites. Of interest is PTM K951-Ac as acetylation at this site is significantly decreased with ratios of approximately 14 in control hearts and 5 in failing hearts. Following the same trend (but not statistically significant) is K1195 also in the tail region. Both K34-Ac and K429-Ac are found in the myosin motor domain where they may be protected from HDAC activity. K429-Ac is a low-abundance PTM with ratios of modified peptides/IRP ratios of approximately 1 and 1.4. This site is buried within the actin-binding cleft and may have been modified prior to protein folding. K34-Ac, on the other hand, is more abundant with approximate modified peptide ratios of 5–10. Perhaps the most interesting observation can be made for the co-modified peptide containing K58-Ac as it may be more susceptible to removal by HDACs or lower HAT activity under I-HF conditions.

In Figure 6, the modified/IRP ratios are shown for the phosphorylated peptides. The average ratios of T215-P ranged from 3 to 2, while in the control nondiseased hearts the modification was nearly undetectable in two samples. The modifications on the other high-confidence peptides may have functional significance, although the ratio of significantly modified/IRC was not altered in diseased (I-HF and NI-HF) compared to NF hearts (Figures 5 and 6). For this reason, we expect they may have a limited role in driving pathogenesis, but may still have functional relevance. Another phosphorylated residue was the phospho-serine at S210 on peptide 6, with ratios of approximately 1–2 modified peptide/IRP in all groups. T215 on peptide 7 had ratios of approximately 1–2 of the modified peptides/IRP in all the groups (Figure 6). When examining the 3D structure of the β-MHC motor domain, we observed three modifications – at the lateral face of the myosin motor domain – with potential importance due to their proximity to the ATP-binding pocket: (1) S210-P (single), (2) T215-P (single), and (3) K213-Ac plus T215-P (double). However, the question remains whether PTMs with low occupancy on a protein such as β-MHC may in fact have subtle yet important roles in fine-tuning of its function. Additional factors to consider are whether specific PTM sites reported here are more labile and prone to action by phosphatases and deacetylases, which may have increased activity in the failing heart.

Model building and MD were used to study putative relationships between PTMs identified on β-MHC and cardiovascular disease. Here, we performed MD simulations of unmodified and modified β-MHC (K58-Ac, K213-Ac/T215-P, K951-Ac) of PTMs that existed in functionally significant regions of the protein seen in Figure 4. For details regarding simulation runs, refer to Tables 3 and 4. From Figure 4, it can be seen that phosphorylation and acetylation sites were distributed throughout the structure of the motor domain and S2 fragment.

Several studies have provided insight on how phosphorylation of cardiac sarcomeric proteins alters myofilament performance (Kawai et al., 2017; Sadayappan et al., 2006; Wijnker et al., 2014; Kuster et al., 2012; Solaro and Kobayashi, 2011; Yamasaki et al., 2002). Kawai et al. utilized LC-MS and found that the rod region of α-MHC (predominant myosin isoform in the adult mouse heart) was hypophosphorylated in the HCM-linked cTnC-A8V mouse model, leading to perturbed cross-bridge kinetics (Kawai et al., 2017). Additionally, the Frank-Starling mechanism may in turn be modified to increase cardiac output under conditions of increased venous return (Monasky et al., 2013). Incidentally, phosphorylation of myofilament proteins by PKCβII and PKA influences length-dependent prolongation of heart muscle relaxation (Monasky et al., 2010; Monasky et al., 2008). Moreover, sarcomeric protein phosphorylation is regulated by reactive oxygen species (ROS), whereby oxidative stress tends to increase protein phosphorylation due to inhibition of protein phosphatase activation and stimulation of protein kinases (Sumandea and Steinberg, 2011). Crosstalk between PTMs may in fact provide a higher order of regulation, for example, same-site competition, structural changes in secondary sites that make it more accessible for modification by other PTMs, or direct modification of the modification of a secondary PTM (Liddy et al., 2013).

Our study identified novel PTMs on β-MHC in nonfailing and failing human hearts. Overall, there tended to be less PTM abundance in the heart failure conditions examined here, which may be due to increased phosphatase or HDAC activity. Our modeling data suggest that some of these PTMS have potential to alter dynamics within β-MHC and thereby fine-tune myofilament function. It remains to be seen whether loss of these PTMs in failing hearts represents the removal of a beneficial, albeit subtle regulation of myofilament function.

Explanted donor human heart tissues were obtained from the Ohio State University Tissue program. The de-identified samples were obtained from patients that were 41–69 years of age with I-HF or NI-HF, and healthy donors after informed consent. Clinical information about patient hearts that were utilized in this study is included in Supplementary file 1.

Human heart tissue was homogenized in Laemmli buffer and separated by SDS-PAGE (12 %), stained with Coomassie blue, and bands that corresponded to β-MHC (~223 kD) were excised as shown in Figure 1—figure supplement 1. The samples were homogenized in 1× Laemmli sample buffer with protease inhibitor cocktail, phosphatase inhibitor cocktail, 1 µM trichostatin A, 1 µM quisinostat, 5 mM nicotinamide, and 1 mM sodium vanadate.

Sample peptides were processed using an externally calibrated high-resolution electrospray tandem mass spectrometer (Thermo Q Exactive HF; Thermo Fisher Scientific, Waltham, MA) in conjunction with a Dionex UltiMate 3000 RSLCnano System (Thermo Fisher Scientific). 5 µl of sample peptide was aspirated into a 50 µl loop and then loaded onto the trap column (Acclaim PepMap100 C18, 5 μm, 100 Å, 300 μm i.d. × 5 mm, Cat# 160454, Thermo Fisher Scientific). Separation on an analytical column (Acclaim PepMap RSLC 75 μm and 15 cm nanoViper; Thermo Fisher Scientific) was conducted with a flow rate of 300 nl/min. A 60 min linear gradient from 3% to 45% B (0.1% formic acid in acetonitrile) was performed. The LC eluent was directly nanosprayed into Q Exactive HF mass spectrometer (Thermo Scientific). During the chromatographic separation, the Q Exactive HF was operated in a data-dependent mode and under direct control of the Thermo Excalibur 3.1.66 (Thermo Scientific). The MS data were acquired using the following parameters: 20 data-dependent collisional-induced dissociation (CID) MS/MS scans per full scan (350–1700 m/z) at 60,000 resolution. MS2 were acquired in centroid mode at 15,000 resolution. Ions with single charge or charges more than 7 as well as unassigned charge were excluded. Raw data were searched with Proteome Discoverer 2.2 (Thermo Fisher Scientific) using Sequest HT and Mascot search engines and percolator as the PSM validator with species-specific FASTA database. Phosphorylation and acetylation were used as a dynamic modification in SequestHT and Mascot and PTMs were scored by ptmRS node in Proteome Discoverer 2.2. MS1-based quantification of peptides was performed using Skyline 4.0. The mass spectrometry data generated in this study has been deposited at Dryad and can be found at doi:10.5061/dryad.s4mw6m97g.

The area under the curve (AUC) for each peptide sequence was measured in Skyline 20.1, an open-source Windows client application software for targeted proteomics data analysis and bioinformatics (MacLean et al., 2010). To determine the modification levels of each PTM site, the AUC of each modified peptide was measured and divided by the AUC of the unmodified common internal reference peptide (IRP) (Figure 5—figure supplement 1). The ratio of the modified peptide to IRP was compared among human heart samples from nondiseased controls and patients with I-HF and NI-HF. The IRP was chosen based upon recommendations from Sherrod et al. utilizing a known unmodified peptide between 7 and 20 amino acids (no methionine and cysteine residues), eluted across the chromatogram, and demonstrating consistent signal stability (Sherrod et al., 2012).

Starting coordinates for the wildtype (WT) β-MHC motor domain simulations were obtained from an X-ray crystallography structure of the post-rigor, ATP state of β-MHC in the Protein Data Bank (PDB, https://www.rcsb.org; PDB ID: 4DB1, 2.6 Å, residues 1–777; Berman et al., 2000). Starting coordinates for the WT S2 fragment simulations were obtained from an X-ray crystallography structure of S2D (PDB ID: 2FXM, 2.7 Å, residues 838–963) (Blankenfeldt et al., 2006). Missing heavy atoms were built using Modeller (Webb and Sali, 2016), and conformer ‘A’ was chosen among residues with multiple conformations in the PDB entries. Hg atoms were removed from the 2FXM structure and ANP•Mn was replaced with ATP•Mg in 4DB1. Starting coordinates for the post-translationally modified variants were obtained via in silico modification of the WT structures using the leap module of AMBER. There were two modified variants of 4DB1: 4DB1-K58-Ac (corresponding to acetylation of Lys 58) and 4DB1-K213-Ac/T215-P (corresponding to simultaneous acetylation of Lys 213 and phosphorylation of Thr 215). There was one modified variant of 2FXM: 2FXM-K951-Ac (corresponding to acetylation of Lys 951). For 2FXM, K951 residues in both chains A and B were modified. Because both K951 residues were modified, our simulations sample the ‘most aggressive’ effects of acetylation at this site: diminished effects may be found for singly modified systems.

All 4DB1 simulations were performed with the AMBER20 package (Geeves and Holmes, 1999; Tang et al., 2016) and the ff14SB force field (Maier et al., 2015). Water molecules were treated with the TIP3P force field (Sumandea and Steinberg, 2011). Metal ions were modeled using the Li and Merz parameter set (Li and Merz, 2014; Li et al., 2015a; Li et al., 2015b). ATP molecules were treated with parameters from Meagher et al., 2003. Parameters for phosphothreonine (called ‘TPO’) and acetyllysine (called ‘ALY’) were obtained from Raguette et al. and Belfon et al., respectively. The SHAKE algorithm was used to constrain the motion of hydrogen-containing bonds (Hammonds and Heyes, 2020). Long-range electrostatic interactions were calculated using the particle mesh Ewald (PME) method (Essmann et al., 1995).

Hydrogen atoms were modeled onto the initial structure using the leap module of AMBER, and each protein was solvated with explicit water molecules in a periodic, truncated octahedral box that extended 10 Å beyond any protein atom. Na⁺ and Cl^- counterions were added to neutralize the systems and then 120 mM Na⁺ and Cl^- ions were added. Each system was minimized in three stages. First, hydrogen atoms were minimized for 1000 steps in the presence of 100 kcal/mol restraints on all heavy atoms. Second, all solvent atoms were minimized for 1000 steps in the presence of 25 kcal/mol restraints on all protein atoms. Third, all atoms were minimized for 8000 steps in the presence of 25 kcal/mol restraints on all backbone heavy atoms (N, O, C, and C atoms). After minimization, systems were heated to 310 K during three successive stages. In each stage, the system temperature is increased by ~100 K over 100 ps (50,000 steps) using the NVT (constant number of particles, volume, and temperature) ensemble. During all heating stages, 25 kcal/mol restraints were present on the backbone heavy atoms (N, O, C, and C atoms). After the system temperatures reached 310 K, the systems were equilibrated over five successive stages using the NPT (constant number of particles, pressure, and temperature) ensemble. During each stage, the systems were equilibrated for 5.4 ns in the presence of restraints on backbone atoms. The strength of the restraints was decreased from 25 kcal/mol during the first stage to 1 kcal/mol during the fourth stage. During the final equilibration stage, the systems were equilibrated in the absence of restraints.

Production dynamics for conventional MD simulations were then performed using the canonical NVT ensemble with an 8 Å nonbonded cutoff and 2 fs time step. Coordinates were saved every picosecond. Simulations were run in triplicate for 500 ns each. Unless specified otherwise, simulations were analyzed separately, and the results of replicate simulations were averaged together. To account for potential equilibration effects, the first 100 ns were excluded from subsequent analyses.

2FXM contains a linear fragment of S2 spanning ~130 residues. Due to the length of this fragment, we performed implicit solvent simulations of 2FMX using the Generalized Born model. This approach was previously used to study S2 fragments, and our methods were chosen to best match earlier simulations (Korkmaz et al., 2016). Simulations were performed using the GB model described by Mongan et al., 2007.

The C_α RMSD, C_α RMSF, SASA, interatomic distances, and interatomic contacts were calculated with cpptraj (Roe and Cheatham, 2013). The C_α RMSD was calculated after alignment of all The C_α atoms to the minimized structure. The C_α RMSF was calculated about average MD structures for each simulation. Two residues were considered in contact with one another if at least one pair of heavy atoms were within 5 Å of one another. All protein images were prepared using UCSF Chimera (Pettersen et al., 2004; Sanner et al., 1996). Electrostatic potentials of molecular surfaces were generated using default parameters for the Coulombic Surface coloring method in UCSF Chimera.

Mass spectrometry and MD simulations data analyses were performed using one-way ANOVA, followed by Bonferroni’s post hoc test and Student’s t-test, respectively.