Allosteric modulation of cardiac myosin mechanics and kinetics by the conjugated omega-7,9 trans-fat rumenic acid

Abstract

Key points

• Direct binding of rumenic acid to the cardiac myosin-2 motor domain increases the release rate for orthophosphate and increases the Ca²⁺ responsiveness of cardiac muscle at low load.
• Physiological cellular concentrations of rumenic acid affect the ATP turnover rates of the super-relaxed and disordered relaxed states of β-cardiac myosin, leading to a net increase in myocardial metabolic load.
• In Ca²⁺-activated trabeculae, rumenic acid exerts a direct inhibitory effect on the force-generating mechanism without affecting the number of force-generating motors.
• In the presence of saturating actin concentrations rumenic acid binds to the β-cardiac myosin-2 motor domain with an EC₅₀ of 200 nM. Molecular docking studies provide information about the binding site, the mode of binding, and associated allosteric communication pathways.
• Free rumenic acid may exceed thresholds in cardiomyocytes above which contractile efficiency is reduced and interference with small molecule therapeutics, targeting cardiac myosin, occurs.

Based on experiments using purified myosin motor domains, reconstituted actomyosin complexes and rat heart ventricular trabeculae, we demonstrate direct binding of rumenic acid, the cis-delta-9-trans-delta-11 isomer of conjugated linoleic acid, to an allosteric site located in motor domain of mammalian cardiac myosin-2 isoforms. In the case of porcine β-cardiac myosin, the EC₅₀ for rumenic acid varies from 10.5 μM in the absence of actin to 200 nM in the presence of saturating concentrations of actin. Saturating concentrations of rumenic acid increase the maximum turnover of basal and actin-activated ATPase activity of β-cardiac myosin approximately 2-fold but decrease the force output per motor by 23% during isometric contraction. The increase in ATP turnover is linked to an acceleration of the release of the hydrolysis product orthophosphate. In the presence of 5 μM rumenic acid, the difference in the rate of ATP turnover by the super-relaxed and disordered relaxed states of cardiac myosin increases from 4-fold to 20-fold. The equilibrium between the two functional myosin states is not affected by rumenic acid. Calcium responsiveness is increased under zero-load conditions but unchanged under load. Molecular docking studies provide information about the rumenic acid binding site, the mode of binding, and associated allosteric communication pathways. They show how the isoform-specific replacement of residues in the binding cleft induces a different mode of rumenic acid binding in the case of non-muscle myosin-2C and blocks binding to skeletal muscle and smooth muscle myosin-2 isoforms.

Introduction

Trans fatty acids (TFAs) are produced industrially during partial hydrogenation of vegetable fats (iTFAs) and by the bacterial metabolism of polyunsaturated fatty acids in the rumen of ruminants (rTFAs). The amount of trans fatty acids in partially hydrogenated vegetable oils can be as high as 60%, with different isoforms of trans-octadecenoic acid (trans 18:1) accounting for 80−90% of the total trans fatty acid content. The main component of iTFAs is elaidic acid (18:1 trans-Δ9). Additional products of industrial hydrogenation have their trans double bonds distributed in a Gaussian-like fashion around position 9 (Stender et al. 2008). In contrast, during biohydrogenation trans fatty acids are formed with a preference for the double bond in position 11 with vaccenic acid (18:1 trans- Δ11) being the major product. The maximum content of rTFA in ruminant fat and milk reaches about 6% (Mangwe et al. 2020). Unlike elaidic acid, vaccenic acid is efficiently converted by desaturases from a wide range of organisms to rumenic acid, the cis-Δ9-trans-Δ11 isomer of conjugated linoleic acid. The exact levels of rTFA in the fat of ruminants also depend on the feed given to the animals and accordingly can show typical seasonal variations (Jakobsen et al. 2006).

The dietary intake of large amounts of iTFAs has been consistently linked to increased low-density lipoprotein and cholesterol levels and decreased high-density lipoprotein levels (Judd et al. 1994; Park & Koehler, 2019). Because of this, the intake of large amounts of iTFAs has been associated with adverse effects on health, including a higher risk of hypertension, coronary heart disease, stroke and diabetes (Willett et al. 1993; Marklund et al. 2020; Wang et al. 2020). While the number of observational studies that have specifically investigated the effects of rTFAs on human health is comparatively small, on the whole they do not establish a strong association between increased intake of rTFAs and a higher risk of coronary heart disease (Jakobsen et al. 2008; Brouwer et al. 2013). To the contrary, higher plasma levels of rTFAs have been associated with a lower risk of heart failure (Tokede et al. 2013; Da Silva et al. 2015; Chandra et al. 2020). In particular, increased concentrations of rumenic acid in fatty tissue seem to correlate with a lower risk of myocardial infarction (Smit et al. 2010).

Small molecule effectors that specifically interact with β-cardiac myosin were shown to have the potential to reverse the effects of cardiomyopathy-causing mutations and to restore cardiac output by altering chemomechanical coupling in the motor domain (Green et al. 2016; Metra & Teerlink, 2017; Maack et al. 2019; Trivedi et al. 2020). An allosteric mode of action has also been shown for arachidonic acid (C20:4 all-cis-Δ5,8,11,14), which stimulates ATP turnover as well as force generation of cardiac and smooth muscle myosin-2 (Damron & Summers, 1997; Katayama et al. 2010). Initial tests performed in our laboratories to identify further fatty acids that interact directly with the myosin motor domain, led to the identification of several C18 and C20 polyunsaturated fatty acids as modulators of cardiac myosin-2 activity. Saturated C16, C18, and C20 fatty acids as well as C16 polyunsaturated and monounsaturated fatty acids had no detectable activity. Of the fatty acids tested, rumenic acid showed particularly strong activation of basal and actin-activated ATP turnover by the subfragment-1 portion of β-cardiac myosin. Rumenic acid, like most dietary conjugated linoleic acids comes from ruminant meat and dairy products. Typical daily intake of these products has been estimated to be approximately 210 mg for men and 150 mg for women (Silveira et al. 2007). The dietary intake of rumenic acid is in part direct (Precht & Molkentin, 1999; Ritzenthaler et al. 2001) and in part mediated by a mammalian delta-9 desaturase, which effectively converts vaccenic acid into rumenic acid (Kuhnt et al. 2006). Vaccenic acid is the main monounsaturated rTFA found in the fat of ruminants and dairy products, where it is 2 to 3 times more abundant than rumenic acid (Hervas et al. 2020).

Here, we describe the effect of rumenic acid on the in situ performance of rat cardiac myosin-2 and the in vitro performance of porcine β-cardiac myosin-2 (Liu et al. 2015). For comparison, we also characterized the influence of rumenic acid on the enzymatic activities of rabbit skeletal muscle myosin (Gyimesi et al. 2020), human non-muscle myosin-2 isoform C⁰ (NM2C⁰) (Chinthalapudi et al. 2017), and chicken smooth muscle myosin-2 (Dominguez et al. 1998). Our results show that rumenic acid binding to the motor domain affects the enzymatic activity of both cardiac myosin-2 and non-muscle myosin-2 isoforms. In contrast, skeletal muscle and smooth muscle myosin-2 activity is not affected by the presence of physiological concentrations of rumenic acid. In the case of cardiac myosin-2 isoforms, the affinity of rumenic acid for cardiac myosin is increased 50-fold in the presence of saturating actin concentrations. Under low-load conditions, the presence of rumenic acid enhances Ca²⁺ responsiveness but the maximum rate of ATP turnover and the maximal velocity of shortening appear not to be affected. In Ca²⁺-activated demembranated (skinned) trabeculae from rat heart, mainly containing α-cardiac myosin, rumenic acid exerts a direct inhibitory effect on the force-generating mechanism without affecting the number of force-generating motors and their dependence on Ca²⁺ ion concentration. Together, the in vitro and in situ results show that direct binding of rumenic acid to the cardiac myosin-2 motor domain reduces the force output per motor but increases the Ca²⁺ responsiveness of cardiac muscle at low load. At higher loads, Ca²⁺ responsiveness remains unchanged. Low micromolar concentrations of rumenic acid differentially affect the ATP turnover rates of non-actin-bound β-cardiac myosin motors in the super-relaxed and disordered relaxed states. The associated uncoupling by a net increase in off-actin ATP turnover is predicted to increase thermogenesis and reduce the energy efficiency of myocardial contraction. Based on the results of molecular docking, we describe models of the protein-ligand interactions for rumenic acid and the motor domains of cardiac and non-muscle myosin-2 isoforms in the pre-powerstroke state.

Methods

Animals and ethical approval

Rats used for the mechanical experiments in Florence were treated in accordance with both the Italian regulation on animal experimentation (authorization no. 956/2015 PR) in compliance with Decreto Legislativo 26/2014 and the EU regulation (directive 2010/63). Male rats (Rattus norvegicus, strain Wistar Han, weight 230−280 g, aged 2–3 months) were provided by Charles River Laboratories and were housed at the Centro di Stabulazione Animali da Laboratorio, University of Florence, Italy, under controlled conditions of temperature (20 ± 1°C), humidity (55 ± 10%), and illumination (lights on for 12 h daily, from 07.00 to 19.00 h). Animals were kept with free access to food and water prior to use. Rats were anaesthetized with isoflurane (5%, v/v); as soon as the animal attained a state of deep anaesthesia, the heart was rapidly excised and perfused with a modified Krebs-Henseleit solution as previously described (Pinzauti et al. 2018).

Protein production and purification

Porcine cardiac myosin-2 was purified as described previously (Jacques et al. 2008; Pant et al. 2009). The β-cardiac isoform is the dominant myosin heavy chain (MHC) isoform in the porcine and human heart, while the α-cardiac isoform is the dominant isoform in rat hearts. Human, porcine and rat β-cardiac myosin motor domains share >95% sequence identity with each other and >90% sequence identity with the rat α-cardiac myosin motor domain.

The HMM fragment of porcine β-cardiac myosin was produced by proteolytic digestion of the purified full-length protein using TLCK-treated α-chymotrypsin. The proteolytic products were purified by gel filtration as described previously (Radke et al. 2014). The motor domain of human non-muscle myosin-2C0 (NM2C⁰) was produced and purified as described previously (Heissler & Manstein, 2011). Smooth muscle myosin protein was purified from chicken gizzards and treated with papain to liberate subfragment-1 (S1) (Margossian & Lowey, 1973; Suzuki et al. 1978).

G-actin from chicken pectoralis major muscle was purified according to the method of Lehrer and Kerwar with minor modifications(Lehrer & Kerwar, 1972; Radke et al. 2014). Expression constructs for human cardiac TnC (UniProt identifier: P63316), TnI (UniProt: P19429) and TnT (UniProt: P453796) were cloned into pET 11c vector and over-expressed in E. coli strain BL21(DE3)pLysS competent cells (Merck). Troponin subunits were purified as described previously (al-Hillawi et al. 1994; Krüger et al. 2003). The troponin complex was reconstituted by mixing troponin subunits at a molar ratio of 1 TnT : 1 TnI : 3 TnC (Burton et al. 2002). The human Tpm α-1-cardiac gene with a 5′ modification encoding an N-terminal Met- Ala-Ser extension was cloned into expression vector pJC20 (Heald & Hitchcock-DeGregori, 1988; Monteiro et al. 1994). The N-terminal Met residue is removed by post-translational processing. Recombinant ASα-Tpm was over-produced in E. coli strain BL21(DE3)pLysS. Tpm purification from bacterial cell pellets was performed as described previously (Coulton et al. 2006).

Steady-state ATPase assay

Steady-state ATPase activities were measured using an NADH-coupled assay (Furch et al. 1998) performed at 37 ± 1°C in standard ATPase buffer (25 mM HEPES, pH 7.3, 5 mM MgCl₂, 25 mM KCl and 0.5 mM dithiothreitol (DTT)) supplemented with varying concentrations of CaCl₂. ATP turnover is reported as molecules ATP hydrolysed per second and per myosin head. Actin-activated steady-state ATPase activities of the different myosin constructs were determined at actin concentrations ranging from 0 to 40 μM and fitting of the data to the Michaelis-Menten equation. K_app,actin is the actin concentration at half-maximum activation of ATP turnover, k_cat corresponds to the maximum value of ATP turnover in the presence of saturating actin concentrations, and k_basal to ATP turnover in the absence of actin. Free Ca²⁺ concentrations, expressed as pCa values, were calculated using Maxchelator software (Bers et al. 2010). The change in absorbance at 340 nm due to NADH oxidation was monitored using a Multiskan FC Microplate Photometer (Thermo Fisher Scientific, Waltham, MA, USA). β-Cardiac HMM concentration in the assay was 0.3 μM. Basal ATPase was measured in the absence of F-actin.

Regulated thin filaments were reconstituted in ATPase buffer by mixing F-actin, human α-cardiac Tpm and α-cardiac troponin holo-complex to a final concentration in the assay of 18 μM, 9 μM and 6 μM, respectively. Reconstitution was performed 30 min before the experiment. Ca²⁺ titrations were performed over the range from pCa 9 to 4 and the ATPase-pCa was fitted by the Hill equation:

(1)

where y represents ATP turnover; the steepness of the transition is given by n and thus estimates the cooperativity of Ca²⁺ activation; pCa_50, the pCa at which 50% of the maximum value is attained, estimates the Ca²⁺ sensitivity of the parameter under investigation.

Rumenic acid (90140 from Cayman Chemicals, Ann Arbor, MI, USA) was added to the reaction mixtures shortly before the reaction was started. Stock solutions of rumenic acid were stored under nitrogen gas at −80°C and aliquots were thawed and diluted immediately before setting up the experiment. Measurements were started by the addition of 2 mM ATP. Control measurements were carried out in the absence of rumenic acid. Each reaction mixture including controls contained 5% (v/v) ethanol, which was used as solvent for rumenic acid.

Phosphate release, ADP release and single turnover measurements in the presence of F-actin

Phosphate release was measured essentially as described previously (Brune et al. 1994; Behrens et al. 2017) using MDCCPBP (N-[2-(1-maleimidyl)ethyl]-7-(diethylamino)coumarin-3-carboxamide) labelled phosphate binding protein (obtained from Life Technologies, Carlsbad, CA, USA). Measurements were performed at 20°C using a Hi-tech Scientific SF-61 double mixing stopped-flow system (TgK Scientific Limited, Bradford on Avon, UK) in assay buffer (20 mM MOPS [3-(N-morpholino)-propanesulfonic acid], pH 7.0, 50 mM KCl, 5 mM MgCl₂, 1 mM DTT, 5% ethanol). MDCCPBP fluorescence was excited at 436 nm and detected using a 455 nm cut-off filter. Final concentrations after double mixing were 0.5 μM β-cardiac HMM, 0.25 μM ATP and 10 μM F-actin in the absence or presence of 20 μM rumenic acid in all solutions. To obtain the rate constant of actin-activated phosphate release, transients were fitted with single exponential equations.

ADP release from actomyosin was determined indirectly by recording the decrease in the light scattering signal that indicates the dissociation of the actin-myosin-ADP complex after the addition of excess ATP. Final concentrations after mixing were 0.5 μM β-cardiac HMM, 25 μM ADP, 0.75 μM F-actin and 500 μM ATP in the absence or presence of 20 μM rumenic acid in all solutions.

The consequences of rumenic acid binding on the turnover of ATP by actin-activated β-cardiac myosin were additionally followed in single-turnover experiments (Tsiavaliaris et al. 2002). Using the extrinsic fluorescence probe 2′-/3′-O-(N′-methylanthraniloyl)-ATP (mantATP) instead of ATP as substrate, the observed signal changes can be attributed to three phases. The initial fast rise of fluorescence signal intensity reflects the binding of the ATP analogue to the myosin active site, the following plateau phase monitors the duration of the hydrolysis reaction, and the third phase, corresponding to a decrease in fluorescence signal intensity, monitors the rate of product release.

Single mantATP turnover kinetics with reconstituted cardiac myosin thick filaments

Single ATP turnover kinetic experiments using fluorescent 2′-/3′-O-(N′-methylanthraniloyl)-ATP (mantATP) were performed with reconstituted cardiac myosin thick filaments following a two-step mixing protocol as described previously (Stewart et al. 2010; Anderson et al. 2018; Gollapudi et al. 2020). First, synthetic thick filaments (STFs) were assembled by diluting porcine β-cardiac myosin (5 μM) in buffer (20 mM Tris-HCl pH 7.4, 150 mM KCl, 1 mM EGTA, 3 mM MgCl₂, 1 mM DTT) with subsequent incubation for 1–2 h on ice (Gollapudi et al. 2020). All following steps were performed at 25°C. A 100 μl aliquot of 0.8 μM STFs was mixed with 50 μl of 3.2 μM mantATP in a 96-well plate and the reaction mixture was aged for 60 s to allow binding and hydrolysis of mantATP. Subsequently, 50 μl 16 mM ATP was added (final concentrations: STFs: 0.4 μM, mantATP: 0.8 μM, ATP: 4 mM). Subsequent changes in fluorescence intensity were recorded using a plate reader system (CLARIOstar Plus, BMG Labtech, Ortenberg, Germany) with excitation at 365 nm and fluorescence detection with a band-pass filter (415–440 nm). The recorded biphasic fluorescence transients were fitted with double exponential decay functions to determine amplitudes and rates of fast and slow phases representing the DRX and SRX states, respectively. The experimental dead time of our plate-based assay and the increase in ATP turnover rate of myosin heads mediated by rumenic acid in the DRX state allow reliable tracking of the changes in the observed rate constant for the fast phase and the ratio of fluorescence amplitudes associated with the slow and fast phases up to a maximum concentration of 20 μM rumenic acid.

Homology modelling and docking analysis

To identify the binding sites and estimate the associated binding scores for rumenic acid in the motor domains of bovine β-cardiac myosin, human β-cardiac myosin, human non-muscle myosin-2C, rabbit skeletal muscle myosin, and chicken smooth muscle myosin, we performed computational docking experiments with Schrödinger Maestro (Schrödinger, LLC, New York, NY, USA; 2018) (Jacobson et al. 2002, 2004). Pre-powerstroke state structures of the motor domains of bovine β-cardiac myosin (PDB-code: 5N69), human non-muscle myosin-2C (PDB-code: 5I4E), rabbit skeletal muscle myosin (PDB-code: 6YSY), and chicken smooth muscle myosin (PDB-code: 1BR2) were used as target structures. A homology model of the human β-cardiac myosin motor domain was built using the Schrödinger Prime module with the bovine β-cardiac myosin motor domain (PDB-code: 5N69) as a template. The quality of the resulting model was assessed with stereochemical analysis tools in COOT (Emsley & Cowtan, 2004).

Pre-docking minimization and optimization of the ligand structure were performed with the LigPrep module, using an OPLS_2005 force-field. Protein PDB structures were prepared using the Protein Preparation Wizard module of Maestro. All solvent atoms were removed from the receptor structures before the docking procedure. Several grid maps were generated to cover the complete motor domain volume. The Glide extra-precision (XP) docking procedure was used for docking the ligand to the receptor structures (Nagpal et al. 2012). The XP Glide Scoring Function (XP GlideScore) includes a linear combination of coulomb, van-der-Waals, binding, and penalty energy terms (Friesner et al. 2004). The resulting ensembles of top rumenic acid binding poses were selected according to the low-binding energy cut-off, equal to one hydrogen bond (−3.0 kcal mol⁻¹). These ensembles were exported to PyMOL (The PyMOL Molecular Graphics System, Version 2.0 Schrödinger, LLC), COOT (Emsley et al. 2010; Bond et al. 2020) for analysis, and UCSF Chimera 1.14 for visualization (Pettersen et al. 2004).

in vitro motility experiments

in vitro motility experiments were performed using the Kron and Spudich assay geometry with previously described modifications (Kron & Spudich, 1986; Radke et al. 2014). All experiments were carried out at 37 ± 1°C and cytochrome C was used as blocking reagent. For temperature control, a water-purged glass cuvette was attached to the flow cell with grease and connected to an external water bath. In order to remove any remaining ATP-insensitive ‘rigor’ heads, β-cardiac HMM was mixed in an equimolar ratio with F-actin, the ATP concentration was adjusted to 2 mM and the solution was immediately centrifuged at 100,000 × g for 20 min at 4°C.

Regulated thin filaments were prepared 30 min before use by incubating 1 μM tetramethyl rhodamine (TMR)-phalloidin-labelled F-actin with 0.5 μM human α-cardiac Tpm and 0.5 μM human cardiac troponin holo-complex. Immediately before insertion into the flow cell, regulated thin filaments were diluted in assay buffer to a concentration of 40 nM (based on the actin concentration), and 100 nM α-cardiac Tpm and 100 nM α-cardiac troponin were added to saturate the filaments with the regulatory components (Homsher et al. 1996). Each reaction mixture including controls contained 5% (v/v) ethanol, which was used as solvent for rumenic acid. Ca²⁺ titrations of the thin filament sliding velocity (V_f) were performed over the range from pCa 9 to 4 and the V_f–pCa relationships were fitted by the Hill equation (eqn. (1)) with y representing V_f. The cooperativity of Ca²⁺ activation is given by n and Ca²⁺ sensitivity by pCa₅₀. Tracking of TMR-phalloidin-labelled actin filaments was performed with the program DiaTrack 3.04 (Vallotton et al. 2017).

Mechanical experiments

The rat heart ventricular trabeculae used in these experiments are pillar-like multicellular preparations with a width of 200−400 μm and length of 2−3 mm. Their size and shape allow the application of sarcomere-level mechanics formerly developed and refined on intact skeletal muscle fibres (Kentish et al. 1986; Lombardi & Piazzesi, 1990; Pinzauti et al. 2018). MHC isoform composition in the rat ventricle is 84% α-MHC and 16% β-MHC (Pinzauti et al. 2018). The choice of trabeculae reflects a necessary compromise for the in situ study of the effect of rumenic acid on cardiac myosin mechano-kinetics at the sarcomere level in view of the difference in isoform composition between rat and pig heart 16% β-MHC (Locher et al. 2011; Pinzauti et al. 2018). Thin, unbranched and uniform trabeculae were dissected from the right ventricle and transferred into a thermo-regulated trough perfused with a modified Krebs-Henseleit solution equilibrated with carbogen (95% O₂, 5% CO₂) (Pinzauti et al. 2018). The trabecula was skinned by 30 min perfusion at room temperature with 1% (v/v) Triton X-100 in relaxing solution (5.44 mM ATP, 7.7 mM MgCl₂, 25 mM EGTA, 100 mM TES, pH 7.1, 19.11 mM creatine phosphate, 10 mM glutathione). 20 mM 2,3-butanedione monoxime (BDM) was added to the skinning solution to prevent the sample from contraction and damage during dissection (Mulieri et al. 1989). The trabecula was mounted in a drop of relaxation solution between the lever arms of a loudspeaker motor and a capacitance force transducer (Lombardi & Piazzesi, 1990). A multi-drop exchange solution system, driven by a stepper motor, allowed rapid change of the solution bathing the trabecula (Linari et al. 2007). A striation follower (Huxley et al. 1981) continuously recorded the changes in sarcomere length in a selected trabecula region (0.6–1.0 mm long). To reduce the compliance of damaged sarcomeres at the clamped ends, we stabilized the ends of the trabecula with a rigor solution containing glutaraldehyde (5% v/v) and then glued them to the T-clips with shellac dissolved in ethanol. The sarcomere length (sl) was set at 2.2 μm and the corresponding trabecula length (L₀) and cross-sectional area (CSA) were measured. The osmotic agent dextran T-500 (5% w/v) was added to all solutions to restore pre-skinning inter-filament distance (Pinzauti et al. 2018). We used a temperature jump protocol to activate trabeculae (Linari et al. 2007). To prevent disruption of the sarcomere structure during the time required for Ca²⁺ to equilibrate within the preparation, trabeculae were first transferred to the activation solution at a temperature of 0–1°C. Force development only occurred following the transfer of the trabeculae into the activation solution at the selected experimental temperature (15°C). To prevent sarcomere shortening against the compliant end regions during force development, we used as feedforward signal the sarcomere length shortening recorded during force development in fixed-end conditions (FE) in the first activation cycle (length clamp conditions, LC; see also Caremani et al. 2016). The time course of force redevelopment T(t) following unloaded shortening can be fitted by an equation with both an exponential and a linear component (Linari et al. 2007):

(2)

where A is the amplitude of the exponential component, k_tr is the rate constant, a is the slope of the linear component and t is time.

The composition of the solutions for the mechanical experiments is detailed in Table 1. The activation solution at a given pCa was obtained by mixing relaxation and activation solutions. Control measurements were carried out in the absence of rumenic acid. Each reaction mixture including controls contained 5% (v/v) ethanol, which was used as solvent for rumenic acid. Preliminary tests were performed that showed that the addition of 5% ethanol to the bathing solution does not affect the mechanical performance of the preparation. Ca²⁺ titrations of the maximum isometric force (T₀) were performed in the pCa range 6.8−4.5 and the relation between T₀ normalized for the corresponding T₀ at pCa 4.5 (T_0,4.5) and pCa was fitted using the equation:

(3)

where n and pCa₅₀ estimate the cooperativity and Ca²⁺ sensitivity respectively, as defined before.

Table 1. Composition of solutions used for sarcomere level mechanical experiments

Solution	[ATP]	[MgCl2]	[EGTA]	[HDTA]	[CaEGTA]	[TES]	[CP]	[GSH]
Relaxation	5.4	7.7	25	—	—	100	19.1	10
Preactivation	5.5	6.9	0.1	24.9	—	100	19.5	10
Activation	5.5	6.8	—	—	25	100	19.5	10

• The concentrations listed in the table are given in mM. All solutions contained 1 mg ml⁻¹ creatine phosphokinase, 10 mM trans-epoxysuccinyl-l-leucylamido-(4-guanidino)butane (E-64) and 20 mg ml⁻¹ leupeptin. The ionic strength in the solutions used was in the range of 188–195 mM and the concentration of free Mg²⁺ ions was 1.2 mM. Chemicals were obtained from Merck KGaA (Darmstadt, Germany). Abbreviations: ATP, adenosine 5′-triphosphate; EGTA, ethylene glycol-bis-(β-aminoethyl ether)-N,N,N′,N′-tetraacetic acid; HDTA, 1,6 diaminohexane-N,N,N′,N′-tetraacetic acid; TES, N-tris[hydroxymethyl]methyl-2-aminoethanesulphonic acid; CP, N-[(phosphonoamino) methyl]-N-methylglycine; GSH, glutathione.

Half-sarcomere stiffness measurements

Half-sarcomere (hs) stiffness of Ca²⁺ activated skinned trabeculae was measured by applying small length changes (ranging from −3 to +3 nm per hs, stretch positive), complete in 110 μs, on an isometrically contracting trabecula (Linari et al. 2007). Hs-stiffness (k) was estimated by the slope of the relation between the tension attained at the end of the step and the change in sarcomere length (T₁ relation). A train of different-sized steps at 200 ms intervals was applied during each activation in order to enhance the precision of stiffness measurement. Sarcomere length and isometric tension at the start of each test step were kept constant by applying a step of the same size but in the opposite direction 50 ms after each test step (Linari et al. 2007; Percario et al. 2018). Hs-stiffness was measured over the range pCa 6.8–4.5, in control conditions and in the presence of 20 μM rumenic acid.

The intercept on the abscissa of the T₁ relation corresponds to the hs-strain (Y₀) just before the step. Since T₀ is modulated by [Ca²⁺] via a proportional change in the number of attached motors, without any change in motor strain and force (Linari et al. 2007), the relation between hs-strain and Ca²⁺-modulated force (Y₀–T₀ relation) is expected to be linear (Linari et al. 2007), in terms of a simple mechanical model that describes the dependence of hs-strain on the force with the first order equation:

(4)

where C_f (the slope) is the compliance of the filaments and s₀ (the ordinate intercept) is the average strain of the attached motors.

Dividing any term in eqn (4) by T₀ we can obtain:

(5)

where C_hs, the hs-compliance, is the reciprocal of the hs-stiffness (k) and s₀/T₀, the compliance of the array of motors, is the reciprocal of β·e, where e is the stiffness of the motor array when all motors are attached and β the fraction of attached motors, which is 1 in rigor (Cooke & Franks, 1980).

Statistical analysis

Origin 2020b (OriginLab Corp., Northampton, MA, USA) and dedicated computer software written in LabVIEW (National Instruments, Austin, TX, USA) were used for the analysis. All data are expressed as the mean ± SD unless otherwise stated. Statistical comparisons were done using the unpaired Student's t test.

Results

Effect of rumenic acid on the basal and actin-activated ATPase activity of β-cardiac HMM

Basal and actin-activated ATPase activities of porcine β-cardiac HMM were measured in the presence of increasing concentrations of rumenic acid (range 1−50 μM; Fig. 1A). The rate of basal ATP turnover by porcine β-cardiac HMM corresponds to 0.075 ± 0.003 s⁻¹ in control measurements and shows a ∼1.9-fold increase to 0.141 ± 0.009 s⁻¹ in the presence of 50 μM rumenic acid (P = 0.00012, n = 4; Fig. 1A). The step-wise addition of actin leads to the expected activation of ATP turnover, with further increases in ATP turnover observed upon addition of rumenic acid. Fitting of the observed dependences of ATP turnover on the concentration of rumenic acid to hyperbolic equations shows that the half-maximal effective concentration (EC₅₀) for rumenic acid decreases with increasing actin concentrations (Fig. 1A). A secondary plot of the observed EC₅₀ values against the concentration of actin is best fitted by a hyperbola and predicts an approximately 50-fold decrease in the EC₅₀ of β-cardiac HMM for rumenic acid, from 10.5 ± 1.0 μM in the absence of actin to 200 ± 80 nM in the presence of saturating concentrations of actin (Fig. 1B). While the presence of increasing concentrations of actin leads to in an increased affinity of β-cardiac HMM for rumenic acid, the affinity of β-cardiac HMM for actin is not affected by the presence of rumenic acid (Fig. 1C).

Figure 1. Rumenic acid effect on basal and actin-activated ATPase activity of β-cardiac HMM

A, dependence of the Mg²⁺-ATPase activity of β-cardiac HMM on rumenic acid concentration in the presence of 0, 5, 10, 20 and 40 μM actin. The continuous lines describe hyperbolic fits to the data measured in the presence of 0, 5, 10, 20 and 40 μM actin. Error bars indicate SD (n = 4). B, graph showing the observed changes in EC₅₀ values for the RA-mediated increase in ATPase activity plotted against the actin concentration. The continuous line corresponds to a sigmoid dose-response relationship. C, dependence of β-cardiac HMM ATPase activity on actin concentration measured in the absence (▼) and presence (▲) of rumenic acid (50 μM). According to the hyperbolic fit curves (continuous lines), the maximum ATP turnover rate k_cat is increased approximately 2-fold, whereas the apparent actin affinity K_app,actin is not affected by the presence of 50 μM rumenic acid. Temperature = 37 ± 1°C.

In single turnover measurements performed in the presence of actin, we observed that the initial two phases reflecting the binding and hydrolysis of mantATP remained unchanged in the presence of 20 μM rumenic acid while the rate constant for the third phase, corresponding to a decrease in fluorescence signal associated with the release of the hydrolysis products phosphate and mantADP, increased 1.6-fold from 0.173 ± 0.005 s⁻¹ to 0.279 ± 0.008 s⁻¹ (Fig. 2A). Phosphate release constitutes the rate-limiting step in the actin-activated ATPase cycle of β-cardiac myosin (Radke et al. 2014). Therefore, we measured the rate of actin-activated phosphate release from β-cardiac HMM in double-mixing stopped-flow experiments in the absence and presence of rumenic acid. The rate of phosphate release is again approximately 1.6-fold accelerated in the presence of 20 μM rumenic acid from 0.09 ± 0.01 s⁻¹ to 0.14 ± 0.01 s⁻¹ (P < 0.0001, n = 4, Fig. 2B). The rate of displacement of ADP from acto-HMM by excess ATP increased from 169.5 ± 8.3 s⁻¹ to 187.6 ± 6.2 s⁻¹ in the presence of 20 μM rumenic acid (P = 0.0352, n = 5, Fig. 2C). Consistent with the activation of ATP turnover in the presence of rumenic acid, we observed upon addition of 20 μM rumenic acid to the in vitro motility assay buffer an 18% increase in the velocity of actin sliding from 237 ± 22 nm s⁻¹ to 280 ± 18 nm s⁻¹ (P = 0.023, n = 4, Table 2). Changes in kinetic and functional parameters upon rumenic acid binding to β-cardiac HMM and EC₅₀ values for rumenic acid in the absence and presence of actin are summarized in Table 2.

Figure 2. Rumenic acid effect on individual steps of the β-cardiac HMM ATPase cycle

A, actin-activated single-turnover of mantATP by β-cardiac HMM. The dotted lines describe double exponential fits to the data that define rate constants for the binding of mantATP (fluorescence increase) and the release of the hydrolysis products orthophosphate and mantADP (fluorescence decay). The observed rate of release of the hydrolysis products is increased 1.6-fold in the presence of 20 μM rumenic acid. B, actin-activated phosphate release is measured by the fluorescence increase of MDCC-PBP upon binding of phosphate. Single exponential fits to the data (dotted lines) show a 1.6-fold accelerated phosphate release in the presence of 20 μM rumenic acid. C, ADP-dissociation from acto-HMM. Displacement of ADP (25 μM) from 0.5 μM acto-HMM with excess ATP (0.5 mM) is monitored by the change in the light scattering signal. Single exponential fits to the data (dotted lines) define an 11% faster rate in the presence of 20 μM rumenic acid. Temperature = 20 ± 1°C.

Table 2. Changes in functional parameters upon rumenic acid binding to β-cardiac HMM

	β-Cardiac HMM			NM2C^0-2R
Parameter	+ Rumenic acid	Control	P value	+ Rumenic acid	Control	P value
kbasal (s^−1)	0.141 ± 0.009^§	0.075 ± 0.003	0.00012	0.132 ± 0.005^§	0.067 ± 0.003	0.0002
EC50 − basal ATPase (μM − rumenic acid)	10.5 ± 1.3	n.a.	n.a.	14.85 ± 1.94	n.a.	n.a.
kcat (s^−1)	0.51 ± 0.05^§	0.27 ± 0.02	0.0016	0.206 ± 0.01^§	0.082 ± 0.007	0.0006
EC50 − actin-activated ATPase (μM − rumenic acid)	0.20 ± 0.08^Ω	n.a.	n.a.	6.03 ± 1.59^$	n.a.	n.a.
Kapp, actin (μM − actin)	24.8 ± 9.7^§	25.9± 4.2	0.5	>140	>140	0.9
kcat/Kapp,actin (μM^−1 s^−1)^Π	0.0199 ± 0.0001	0.0098 ± 0.0001	0.01	—	0.002 ± 0.0002	n.a.
Phosphate release from acto-myosin kobs (s^−1)	0.14 ± 0.01^λ	0.09 ± 0.01	0.0001	n.d.	n.d.	n.d.
ADP release from acto-myosin^λ k-AD (s^−1)	187.6 ± 6.2^§	169.5 ± 8.3	0.0352	n.d.	n.d.	n.d.
Filament velocity (nm s^−1)	280 ± 18^#	237 ± 22	0.023	n.d.	∼90	n.d.

• ^§ In the presence of saturating concentrations of rumenic acid;
• ^Π Values for k_cat/K_app.actin were determined in the absence and presence of 50 μM rumenic acid from the initial slope of the data fit to the Michaelis-Menten equation at [actin] << K_app.actin (49);
• ^# In the presence of 20 μM rumenic acid;
• ^λ In the presence of 10 μM actin;
• ^Ω In the presence of 40 μM actin;
• ^$ In the presence of 20 μM actin.

Control measurements probing the effect of rumenic acid on other myosin-2 isoforms, which serve a different physiological function, were performed with rabbit skeletal muscle myosin-2, human NM2C⁰, and chicken gizzard smooth muscle myosin-2. In the case of NM2C⁰, we observed similar changes in ATP turnover following the addition of rumenic acid compared to those observed with β-cardiac myosin. The activation of both basal and actin-activated ATPase activity of NM2C⁰ is best described by hyperbolic fit curves (P = 0.0002, n = 4 and P = 0.0006, n = 4; Fig. 3A). The reduction in EC₅₀ values for rumenic acid observed between basal ATP turnover and actin-activated turnover in the presence of 20 μM actin is ∼2.5-fold for NM2C⁰ (14.9 μM/6.0 μM) (Fig. 1B). Single turnover experiments performed with NM2C⁰ in the absence and presence of 20 μM rumenic acid show that ATP binding and ATP hydrolysis are not affected, while the observed rate constant associated with the release of the hydrolysis products increases approximately 1.3-fold from 0.107 s⁻¹ to 0.136 s⁻¹ (Fig. 3B). The addition of up to 50 μM rumenic acid to our assay solutions did not induce any significant changes in basal and actin-activated turnover of ATP by skeletal muscle myosin-2 and smooth muscle myosin-2 (P = 0.186, n = 4 and P = 0.894, n = 4; Fig. 3C).

Figure 3. Rumenic acid effect on basal and actin-activated ATP turnover of NM2C⁰ and smooth muscle myosin-2

A, basal (△) and actin-activated (▲) ATPase activities of NM2C⁰ are both accelerated in the presence of increasing concentrations of rumenic acid. The continuous lines describe hyperbolic fits to the data and define EC₅₀ values of 14.85 ± 1.94 μM and 6.03 ±1.59 μM, respectively. Error bars indicate SD (n = 4). Temperature = 37 ± 1°C. B, actin-activated single-turnover of mantATP by NM2C⁰. The dotted lines describe double exponential fits to the data and define rate constants for mantATP binding, ATP hydrolysis and product release. The rate of product release is increased 1.27-fold in the presence of 20 μM rumenic acid. Temperature = 20 ± 1°C. C, basal (△) and actin-activated (▲) ATPase activities of smooth muscle myosin-2 are both unaffected by the addition of up to 50 μM rumenic acid. Data are best fitted by horizontal lines. Error bars indicate SD (n = 4). Temperature = 37 ± 1°C.

Computational analysis of rumenic acid binding sites on myosin motor domains

To predict the binding site for rumenic acid in the myosin motor domain, we performed computational docking studies with β-cardiac myosin, NM2C isoforms, skeletal muscle myosin-2, and smooth muscle myosin-2. To obtain comparable results, we used in each case a pre-powerstroke state structure, which was solved to 2.45 Å resolution for the bovine β-cardiac myosin motor domain (Planelles-Herrero et al. 2017), to 2.9 Å resolution for the chicken smooth muscle myosin motor domain (Dominguez et al. 1998), to 2.25 Å resolution for the human NM2C motor domain (Chinthalapudi et al. 2017), and to 3.25 Å for rabbit skeletal myosin (Gyimesi et al. 2020). Since an experimental structure of the human β-cardiac myosin motor domain in the pre-powerstroke state is not available, we constructed a homology model based on the structure of the bovine protein (Planelles-Herrero et al. 2017). The motor domains of human and bovine β-cardiac myosins share >96% sequence identity and the resulting homology model has a close resemblance to the template structure with an r.m.s.d. value of 0.26 Å. To ensure unbiased results, we initially performed global docking of rumenic acid using grid maps covering the complete motor domain followed by targeted docking to areas with binding scores better than -2.5 kcal mol⁻¹. The binding scores for the best poses and experimentally determined EC₅₀ values, determined by measuring the dependence of the increase of basal myosin ATPase activity on the concentration of rumenic acid, are summarized in Table 3.

Table 3. Docking scores and EC50 values of rumenic acid bound to myosin motor domains

Myosin	Docking score for thebest pose (kcal mol^−1)	Experimental EC50actin-activated ATPase
β-Cardiac muscle myosin-2^1	−4.5	2.1 ± 0.9 μM
Non-muscle myosin-2C^2	−3.6	6.0 ± 1.3 μM
Skeletal muscle myosin-2^3	−2.3	No effect
Smooth muscle myosin-2^4	>0	No effect

• ¹ Homology model of human β-cardiac myosin-2 based on the pre-powerstroke structure of β-cardiac myosin from B. Taurus (PDB-code: 5N69);
• ² Human NM2C⁰ in the pre-powerstroke state (PDB-code: 5I4E);
• ³ Rabbit skeletal muscle myosin in the pre-powerstroke state (PDB-code: 6YSY);
• ⁴ Chicken gizzard smooth muscle myosin-2 in the pre-powerstroke state (PDB-code: 1BR2);
• ⁵ The Extra Precision Glide Scoring Function (XP GlideScore) corresponds to a linear combination of Coulomb, van-der-Waals, binding, and penalty energy terms (Friesner et al. 2006);
• ⁶ In the presence of 20 μM actin

In the case of human and bovine β-cardiac myosin, the docking results predict rumenic acid to bind to an allosteric site located in the interface area between the upper 50 kDa, the lower 50 kDa, and the converter domains (Fig. 4A). Figure 4B shows a sequence comparison of myosin-2 family members in the region predicted to mediate the binding of rumenic acid with the position of secondary structure elements shown above the sequences. Human, rat and pig β-cardiac myosin motor domain sequences share greater than 95% identity with each other, but less than 51% identity with human NM2C⁰ and chicken smooth muscle myosin-2. The overall identity between the motor domain sequences of human, rat and porcine β-cardiac myosin and that of rat α-cardiac myosin is greater 90% and greater 97% for the predicted rumenic acid binding region. The sequences of the rat cardiac myosin isoforms are identical in the rumenic acid binding region.

Figure 4. Post-docking interactions of rumenic acid with residues in the motor domain of pre-powerstroke state human β-cardiac myosin

A, ribbon model of human β-cardiac myosin with an ensemble of rumenic acid docking poses having binding scores between −4.5 (best) and −3.0 kcal mol⁻¹. The semi-transparent grey surface shows the volume occupied by the docking poses of the rumenic acid ensemble. The binding site is part of a large cleft between the N-terminal domain (blue), the transducer (cyan), the relay helix (green), and the converter (red). ATP is shown in orange semi-transparent surface representation. B, sequence comparison of myosin-2 family members in the region predicted to mediate binding of rumenic acid to β-cardiac myosin. Positions of associated secondary structure elements are displayed above the aligned sequences. Human β-cardiac myosin heavy chain residues that are predicted to be directly involved in rumenic acid binding regions are shown in bold. Proteins and abbreviations used in the alignment with UniProt identifiers given in parentheses: R.n. αCM: rat α-cardiac myosin heavy chain (P02563); H.s. βCM: human β-cardiac myosin heavy chain (P12883); R.n. βCM: rat  β-cardiac myosin heavy chain (G3V8B0); B.t. βCM: bovine β-cardiac myosin heavy chain (Q9BE39); S.s. βCM: porcine  β-cardiac myosin heavy chain (P79293); H.s. sklM: human skeletal muscle myosin heavy chain, adult 2 (Q9UKX2). In the region of interest, the following substitutions occur in the myosin-2 isoform of rabbit skeletal muscle (Q28641): K136, P137A, A753G, I757V; G.g. SMM: chicken smooth muscle myosin heavy chain (P10587); H.s. NM2C: human non-muscle myosin-2C heavy chain (Q7Z406). The alignment was generated using Clustal Omega 1.2.4 (Sievers & Higgins, 2018). [Color figure can be viewed at wileyonlinelibrary.com]

The binding site in the β-cardiac myosin motor domain is formed by upper 50 kDa domain helices HD, HE, and the loop connecting them, and the converter domain helices HZ and HB′ together with strand βa connecting helices HY, which is frequently referred to as SH1 helix, and HZ. The ensemble of rumenic acid docking poses occupies a compact, horseshoe-shaped volume (Figs 4A and 5A), with binding scores ranging from −4.5 kcal mol⁻¹ to −3.2 kcal mol⁻¹. The ensemble conformers make hydrophobic and hydrogen bond contacts to a subset of residues that are conserved in all myosin-2 isoforms (outlined in red in Fig. 5B) and residues that are only conserved between the cardiac myosin isoforms shown with dotted red outlines. The hydrogen bonds with conserved N711 play an important role in rumenic acid binding to the allosteric site. The interaction with rumenic acid extends the pre-existing hydrogen bond network between N711, F765 and Y164. Among the residues conserved only in α- and β-cardiac myosin but not in other closely related myosin isoforms, Y164 and N160 contribute significantly to rumenic acid binding. Y164 makes hydrophobic contacts with four carbons of the aliphatic chain. These interactions are responsible for the bent conformation of rumenic acid and facilitate the formation of a hydrogen bond between the carboxyl group and N711. N160 plays a similar role, contacting two carbons of the aliphatic chain and facilitating the interaction with N711. The contact area formed by residues N160, T163, Y164, T167 and D168 (helix HE) is linked to the P-loop and the ATP binding site via strand β4. Changes mediated via this link are likely to contribute to the functional consequences observed upon binding of rumenic acid to β-cardiac myosin.

Figure 5. Comparison of the rumenic acid binding sites in the β-cardiac myosin and human NM2C motor domains and the omecamtiv mecarbil binding site in the β-cardiac myosin motor domain

A, close-up view of the predicted rumenic acid binding site in human β-cardiac myosin. B, 2D diagram of ligand-protein interactions mediated by hydrogen bonds and hydrophobic contacts in β-cardiac myosin. Residues with an interaction distance of less than 3.9 Å are shown. Residues are coloured in accordance with the colour code used in Fig. 4A. Residues that are similar or identical in sarcomeric, smooth, and cytoskeletal myosin-2 isoforms are shown with continuous red outlines. Residues that are highly conserved between cardiac myosins are shown with dotted red outlines. C, superposition of the omecamtiv mecarbil binding pose (Planelles-Herrero et al. 2017) with the in β-cardiac myosin. Omecamtiv mecarbil is shown as yellow ball-and-stick model and the surface of the rumenic acid pharmacophore is coloured blue. The carboxymethyl-piperazine and the fluoro-benzyl rings of omecamtiv mecarbil fit well into the volume occupied by the rumenic acid ensemble, while the methyl-pyridinyl ring extends beyond the pharmacophore boundaries towards the relay helix. D, 2D diagram of omecamtiv mecarbil–protein interactions mediated by hydrogen bonds in β-cardiac myosin. The carboxymethyl-piperazine, fluoro-benzyl ring, carbamoyl-amino linker, and methyl-pyridinyl ring of omecamtiv mecarbil are labelled with numbers 1, 2, 3 and 4, respectively. Depiction and colouring of residues are identical to those used in panel B. E, close-up view of the predicted rumenic acid binding site in human NM2C. F, 2D diagram of ligand-protein interactions mediated by hydrogen bonds and hydrophobic contacts in NM2C. Depiction and colouring of residues are identical to those used in panels B and D. The 2D diagrams are based on interaction plots created with LigPlot+ (Wallace et al. 1995) and Maestro Release 2020–4: (Schrödinger, LLC, New York, NY, USA). [Color figure can be viewed at wileyonlinelibrary.com]

In the case of NM2C, rumenic acid is predicted to bind to the same region, but with the binding site shifted towards the relay helix and away from the converter. Interacting residues belong to strand β3, which is part of the seven-stranded β-sheet that is generally referred to as the transducer, the neighbouring relay and HW helices, and strand βa of the convertor domain. Instead of the compact, horseshoe-shaped volume observed with cardiac myosin-2, the top docking poses for NM2C occupy a more elongated S-shaped volume. The alternative binding mode appears to be driven by the replacement of β-cardiac residues N160 and Y164 with NM2C residues G180 and S184, thereby removing two key sidechains for the establishment of an interaction networks in the allosteric site. Stabilizing interactions are contributed by hydrogen bonds between the carboxyl group of rumenic acid and the guanidinium groups of R520 and R735, and by hydrophobic and van-der-Waals contacts between the aliphatic side chain of rumenic acid and residues L505, H508, L513, M682 and S686.

The small-molecule compound omecamtiv mecarbil was shown to bind to the same region in an extended mode that involves interactions with residues from the converter domain, the third β-strand of the transducer and the relay helix (Planelles-Herrero et al. 2017). The volume occupied by omecamtiv mecarbil shows a large overlap with the volume occupied by rumenic acid in cardiac myosin and partially overlaps with the volume occupied by rumenic acid in NM2C. Omecamtiv mecarbil binding to this allosteric site was shown to be functionally relevant for stabilizing the lever arm in a primed position, resulting in accumulation of β-cardiac myosin in the primed state (Planelles-Herrero et al. 2017).

In line with the results of our biochemical studies, the docking of rumenic acid to smooth muscle myosin results in binding scores > 0 kcal mol⁻¹. The replacements N160T and Y164S in smooth muscle myosin-2 is predicted to necessitate the same shift as observed for NM2C towards the relay helix and away from the region between helix HE and the converter. However, the presence of R675 in smooth muscle myosin-2 instead of S686 in the binding cleft of NM2C creates a steric hindrance that prevents binding to this alternative location. In the case of skeletal muscle myosin-2, the replacement Y164F disrupts the strong hydrogen bonding to N711 (Fig. 5B). As a result, the sidechains of N711 and F164 change orientation, which leads to further conformational changes in the affected region. Together, these changes result in an unfavourable geometry for rumenic acid binding, which is reflected in a poor docking score below the binding energy cut-off (Table 3).

Effect of rumenic acid on the Ca²⁺ sensitivity of the regulated cardiac actomyosin system under zero load

We examined the effect of rumenic acid on Ca²⁺ activation of the regulated β-cardiac myosin-thin filament system by performing Ca²⁺ titrations over the range from pCa 9 to 4 of the steady-state ATPase activity (Fig. 6A) and unloaded thin filament sliding velocity (V_f) (Fig. 6B). The ATPase–pCa and V_f–pCa relationships were fitted by the Hill equation (eqn (1)). The Ca²⁺ ion dependences of ATP turnover and V_f are similar (pCa₅₀ ∼6.3) in the absence of rumenic acid. In the presence of 20 μM rumenic acid; both show an increase in Ca²⁺ responsiveness as indicated by left shifts of 0.37 and 0.29 pCa units, respectively. The cooperativity of the Ca²⁺ dependences of ATP turnover and V_f are not significantly affected by rumenic acid (Table 4).

Figure 6. Effect of rumenic acid on the Ca²⁺ activation of the regulated β-cardiac myosin-thin filament system

A, ATP turnover-pCa relation of β-cardiac HMM activated by reconstituted thin filaments in the absence (○) and in the presence of 20 μM rumenic acid (●). B, V_f-pCa relation of reconstituted thin filaments moving on β-cardiac HMM in the absence (○) and in the presence of 20 μM rumenic acid (●). Both assays were performed at 36°C; values are means ± SD (n = 4); the lines represent fits of the data to the Hill equation (dashed line: control; continuous line: 20 μM rumenic acid).

Table 4. Effect of rumenic acid on the Ca²⁺ activation of the regulated β-cardiac myosin-thin filament system under zero load conditions^*

	ATP turnover			Filament velocity
	Control	± Rumenic acid^*	P value	Control	± Rumenic acid^*	P value
n	1.16 ± 0.18	1.23 ± 0.15	0.78	2.22 ± 0.48	2.39 ± 0.31	0.78
pCa50	6.31 ± 0.06	6.68 ± 0.05	0.003	6.36 ± 0.04	6.65 ± 0.03	0.001

• * 20 μM rumenic acid was added to the assay buffer.

Effect of rumenic acid on sarcomeric off-actin states of β-cardiac myosin

The balance of myosins in the SRX and DRX conformations affects cardiomyocyte contraction, relaxation and metabolism. Impairment of this balance has been associated with functional, energetic and cellular abnormalities that confer an increased risk of heart failure and atrial fibrillation, especially in susceptible patients such as those with hypertrophic cardiomyopathy (Toepfer et al. 2020; Nag & Trivedi, 2021). We used a plate-based single-nucleotide turnover assay to define the effect of rumenic acid on the ATP turnover of β-cardiac myosin DRX and SRX states and the equilibrium between the states (Anderson et al. 2018). In the absence and presence of up to 20 μM of rumenic acid, the decay rate of mant-nucleotide fluorescence (Fig. 7A) was best fitted by two exponential rate constants, the faster rate constant representing ATP turnover by DRX-state myosin heads and the slower representing turnover by SRX myosin heads (Stewart et al. 2010). In the absence of rumenic acid, the observed rate constant and amplitude for the fast phase (DRX state) are 22 × 10⁻³ ± 1 × 10⁻³ s⁻¹ and 72 ± 3%. The corresponding values for the slow phase (SRX state) are 3.6 × 10⁻³ ± 0.5 × 10⁻³ s⁻¹and 28 ± 3%. In the range from 0.5 to 20 μM rumenic acid, the fraction of fluorescence amplitude associated with the slow phase increased linearly from 28 ± 3% to 36 ± 3%, indicating a matching change in the relative occupancies of SRX and DRX states. DRX occupancy, calculated as 1 − SRX, goes upon addition of 20 μM rumenic acid from 72 ± 3% to 64 ± 3%, a decrease of 11%. The rate of ATP turnover of myosin heads in the DRX state increased linearly with increasing rumenic acid concentrations, reaching a value of 120 × 10⁻³ ± 10 × 10⁻³ s⁻¹ at 20 μM (P = 0.001, n = 3; Fig. 7B). The decrease in the activity of the SRX state in the presence of increasing concentrations of rumenic acid to a plateau value of 0.32 × 10⁻³ ± 0.06 × 10⁻³ s⁻¹ was best fitted by a sigmoidal dose-response curve with an apparent EC₅₀ value of 1.2 ± 0.05 μM (P = 0.0018, n = 3; Fig. 7C).

Figure 7. Effect of rumenic acid on the ATP turnover of β-cardiac myosin DRX and SRX states

A, representative fluorescence decay transients of mant-nucleotide release from β-cardiac myosin STFs. Fitting the traces with double exponential functions yields the observed rate constants and amplitudes of the DRX (fast phase) and SRX (slow phase) states. B, within the measuring range of 0 to 20 μM, the ATP turnover rate of the fast phase (DRX state) increases linearly with increasing concentrations of rumenic acid. C, the decrease in the activity of the SRX state in the presence of increasing concentrations of rumenic acid is best fitted by a dose-response curve with an apparent EC₅₀ value of 1.2 ± 0.05 μM. Temperature = 25 ± 1°C.

Effect of rumenic acid on isometric force

The maximum Ca²⁺-activated isometric force (T₀) developed by rat trabeculae is in the presence of rumenic acid reduced in a dose-dependent manner. The decrease saturates at rumenic acid concentrations ≥ 40 μM at a level that corresponds to ∼76% of the control value (T_0,c) and the EC₅₀ corresponds to 6.2 ± 0.4 μM (Fig. 8A). Force redevelopment following a period of unloaded shortening at saturating [Ca²⁺] (pCa 4.5) in the absence or presence of 20 μM rumenic acid is shown in panels B and C, respectively in Fig. 8. The time course of force development was fitted with eqn (2) comprising an exponential and a linear component. Both in the absence and the presence of rumenic acid, more than 85% of force redevelopment within the first second of contraction can be accounted for by the exponential component alone. k_tr, estimated by the exponential component of the fit, is 9.4 ± 0.9 s⁻¹ in the absence and 8.6 ± 1.2 s⁻¹ in the presence of 20 μM rumenic acid. The observed difference is not significant (P = 0.29, n = 5), as is evident by superimposing the time courses normalized to the steady force reached before the release (Fig. 8D).

Figure 8. Effect of rumenic acid on isometric force development

A, titration of the effect of rumenic acid concentration on T₀ at pCa 4.5. Data are expressed relative to T₀ in control (T_0,c). The continuous line corresponds to a fit of the data to the Hill equation. Error bars indicate SD (n = 3). B and C, force redevelopment following unloaded shortening at pCa 4.5. Changes in force (lower trace) and half-sarcomere length (upper trace) in the absence (B) and in the presence of 20 μM rumenic acid (C). D, superimposed force redevelopment curves from B and C following normalization to the force before the release step.

Effect of rumenic acid on the force-pCa dependence

To assess the effect of rumenic acid on the force-pCa dependence, we measured T₀ in the range of pCa 6.8–4.5 (Fig. 9A). To isolate the effect of rumenic acid on the Ca²⁺ sensitivity of the force T₀ at any given Ca²⁺ ion concentration, Fig. 9B shows the results normalized for the corresponding T₀ at pCa 4.5 (T_0,4.5). Values for n and pCa₅₀ obtained from fitting the curves with eqn (3) are reported in Table 5. The results show that rumenic acid causes slight reductions in cooperativity and Ca²⁺ responsiveness of force development. However, the significance of the effect is scarce and the effect on n is even less significant (P = 0.07) than that on pCa₅₀plusmn (P = 0.007).

Figure 9. Force-pCa relations in the absence (open circles) and in the presence of 20 μM rumenic acid (filled circles)

A, T₀ as a function of pCa is shown in absolute units. B, normalized graph showing the ratio of T₀ at selected Ca²⁺ concentrations to T₀ at saturating Ca²⁺ (T_0,4.5). Data points are means ± SE from four trabeculae. Dashed (control) and continuous (20 μM rumenic acid) lines correspond to a fit of the data to the Hill equation.

Table 5. Effect of rumenic acid on the Ca²⁺ dependence of force generation

	Control	+ Rumenic acid^*	P value
T0/T0,4.5c	1	0.68 ± 0.10	< 0.0007
n	3.00 ± 0.23	2.34 ± 0.20	0.07
pCa50	5.96 ± 0.01	5.87 ± 0.02	0.007

• * 20 μM rumenic acid was added to the assay buffer.

Effect of rumenic acid on number and force of attached motors at T₀

To elucidate the mechanism of depression of T₀ by rumenic acid, we used a previously developed experimental approach (Linari et al. 2007), in which the stiffness of the array of actin-attached motors in each half-sarcomere (an estimate of the fraction of attached motors) is obtained by determining the relation between hs-stiffness and Ca²⁺-modulated force. Stiffness is measured by superimposing on the steady isometric force a series of four steps of different size (Fig. 10A). Each step is followed after 50 ms by the same step in the opposite direction to keep the sarcomere length constant before the next step. The force peak attained at the end of the step (T₁, arrow in Fig. 10B) depends on the stiffness of the half-sarcomere (Huxley & Simmons, 1971). In Fig. 10C, the relations between T₁ and the step size in the absence (open circles) and presence of 20 μM rumenic acid (filled circles) for measurements performed with trabeculae at pCa 4.5 are compared. The slope of the linear fit to the data corresponds to the hs-stiffness (k) and its intercept on the abscissa corresponds to the hs-strain (Y₀) just before the step. k is not significantly altered by the presence of rumenic acid (12.54 ± 0.46 kPa nm⁻¹ (control) versus 11.94 ± 0.51 kPa nm⁻¹, P = 0.4, n = 4), suggesting that the force depression by rumenic acid is mainly caused by a reduction in hs-strain. T₁ relations were determined at various pCa (range 6.8–4.5) in either condition. The relations between hs-strain and Ca²⁺-modulated force (Y₀–T₀ relation) either in control (Fig. 10D open circles) or in the presence of rumenic acid (filled circles) can be fitted by eqn (4). The slope of the linear fit (C_f) specifies the compliance of the filaments, while the ordinate intercept (s₀) gives the average strain of the attached motors. Actually, in each relation there is a downward deviation of the experimental point at the lowest force with respect to the linear fit to the three points at the higher force (dashed line control, continuous line rumenic acid). Deviations from a linear response at low forces have been previously observed in both skeletal and cardiac intact myocytes and are generally explained by the presence of an additional elastic element with constant small stiffness in parallel with that of attached myosin motor. The contribution of this element emerges when the number of attached motors and thus the stiffness of the motor array drops to a value comparable to the parallel element stiffness (Fusi et al. 2014). At higher forces, the data follow the predicted linear relationship both in the presence and absence of rumenic acid. The values for slope and ordinate intercept are summarized in Table 6. C_f is almost identical for both relations (∼16 nm MPa⁻¹) as expected, while s₀ is reduced from 4.48 to 3.47 nm (77% the control value) in the presence of 20 μM rumenic acid, which fully accounts for the reduction of T₀ by rumenic acid (76%; see Fig. 8A).

Figure 10. Effect of rumenic acid on half-sarcomere stiffness at T₀

A, protocol for measuring hs-stiffness during isometric contraction at saturating [Ca²⁺]: force response (lower trace) to a series of stepwise hs-length changes (upper trace). Each step (marked by a bar) is followed after 50 ms by the same step in the opposite direction to keep the sarcomere length constant before the next step. B, superimposed hs-length changes (upper traces) and force response (lower traces) for four steps of different sizes and directions imposed at a force = T_0,4.5. T₁, the force attained at the end of the length step, is marked by the arrow. C, T₁ relationships determined in the absence (open circles and dashed line) and in the presence of 20 μM rumenic acid (filled circles and continuous line). Lines are linear regression equations fitted to the experimental data. The intercept of the lines with the x-axis measures hs-strain before the length step (Y₀). D, dependence of Y₀ on T₀ in the absence (open circles) and in the presence of 20 μM rumenic acid (filled circles) (n = 4). Data are grouped in classes of 15 kPa. Dashed (control) and continuous (20 μM rumenic acid) lines are the linear fits to data for forces > 20 kPa.

Table 6. Rumenic acid induced changes in the performance of rat trabeculae

Parameters of the T1 relationships^§	Control	+ Rumenic acid	P value
Cf (nm MPa^−1)	15.5 ± 3.0	16.5 ± 3.3	0.8
k (kPa nm^−1)	12.54 ± 0.46	11.94 ± 0.51	0.4
s0 (nm)	4.48 ± 0.17	3.47 ± 0.12	0.003
F0 (pN)	4.79 ± 0.18	3.71 ± 0.13	0.003

• ^§ Estimates of the relevant mechanical parameters of the half-sarcomere (filament compliance, C_f, hs-stiffness, k, and myosin motor strain, s₀), according to Model 1 reported in Fusi et al. (2014) in the absence (control) and in the presence of 20 μM rumenic acid. The force generated per myosin motor F₀, was calculated by assuming a constant motor stiffness of 1.07 pN nm⁻¹ (Pinzauti et al. 2018).

These results strongly suggest that rumenic acid reduces the isometric force by reducing the force per motor without significant effect on the number of motors attached. At saturating [Ca²⁺], C_hs ( = 1/k) is 79.8 ± 2.9 nm MPa⁻¹ in the absence and 83.8 ± 3.1 nm MPa⁻¹ in the presence of rumenic acid and C_f equals 15.5 ± 3.0 and 16.5 ± 3.3, respectively (Table 6). Thus, s₀/T₀, the compliance of the array of motors, is only slightly changed by the addition of rumenic acid, from 64.3 ± 5.9 nm MPa⁻¹ to 67.3 ± 6.3 nm MPa⁻¹. In the absence of rumenic acid, the corresponding average value for the stiffness of the attached motors (T₀/s₀ = βe = 15.6 ± 1.4 kPa nm⁻¹) divided by the stiffness of the motor array in rigor, when all motors are attached (e = 133 ± 12 kPa nm⁻¹; Pinzauti et al. 2018) gives an estimate of 0.12 ± 0.02 for the fraction β of attached motors at T₀ with saturating Ca²⁺ ion concentrations. When the same calculation is made using the compliance of the array of motors estimated at T₀ in the presence of 20 μM rumenic acid (67.3 ± 6.3 nm MPa⁻¹), with the assumption that the stiffness of the motor is not affected by rumenic acid, one obtains a fraction β of 0.11 ± 0.02 attached motors at T₀ with saturating Ca²⁺-ion concentrations. Thus, the presence of rumenic acid appears to have no significant impact on the fraction of attached motors under these conditions.

Discussion

Our results show that rumenic acid binds directly to the myosin motor domains of cardiac and non-muscle myosin-2 isoforms, where it affects ATP turnover and the coupling between the nucleotide and actin binding sites via an allosteric mechanism. The dissociation of orthophosphate (P_i) from the pre-power stroke myosin-ADP-Pi complex, the rate limiting step of basal turnover of ATP by cardiac myosin, is accelerated 1.6-fold in the presence of 20 μM rumenic acid. The acceleration of the phosphate release step can lead to a partial uncoupling of ATP turnover from motor activity with a corresponding drop in energy efficiency. In addition, a rumenic acid-induced faster transition from the low actin-affinity myosin-ADP-Pi state to high actin-affinity myosin-ADP state increases the duty ratio. This is offset under low-load conditions, if only slightly, by a faster ADP release from acto-myosin.

The addition of rumenic acid leads to different outcomes in regard to Ca²⁺ responsiveness and peak activity of cardiac myosin when the motor works against near zero loads or against high loads on regulated actin filaments. The translation and interpretation of these divergent results between isolated actomyosin systems and fully assembled sarcomeres needs to take into account that load-dependent rates, possibly influenced by rumenic acid, are more pronounced in fibre experiments than in isolated actomyosin complexes, and that the behaviour of motor ensembles cannot be linearly extrapolated from results obtained from in vitro experiments with the isolated proteins (Liu et al. 2018; Mansson et al. 2018). Moreover, different cardiac isoforms were shown to have important kinetic and mechanical differences and the isoform composition of the ventricular myocardium varies considerable between the species used in our study (Palmiter et al. 1999; Alpert et al. 2002; Milani-Nejad & Janssen, 2014). While these limitations are certainly pertinent, our approach does have advantages arising from the fact that the allosteric pocket residues in α- and β-cardiac myosin isoforms are 100% conserved and that we have identified myosin isoforms with an altered and blocked binding mode.

Under low-load conditions, the presence of rumenic acid enhances Ca²⁺ responsiveness of cardiac myosin-2, as indicated by the left-shift observed for the pCa-ATP turnover and pCa–V_f relationships, while the maximum rate of Ca²⁺-activated ATP turnover and the maximal velocity of actin sliding in an in vitro motility assay are not affected. Under high-load conditions, the maximum isometric force T₀ is reduced in the presence of rumenic acid to 76% of the control value, while the lack of a rumenic acid-induced shift in the pCa–force relationship shows that Ca²⁺ responsiveness remains unchanged. Stiffness measurements analysed in terms of a simple mechanical model of the half-sarcomere indicate that the reduction in isometric force is accounted for by a 23% reduction in the force developed per motor (from 4.79 pN in control to 3.71 pN in the presence of rumenic acid), which in turn appears a direct consequence of the drop in energy efficiency related to the acceleration of the phosphate release step.

The interaction between the myosin motor domain and actin is not only important for the process of power generation, but contributes as well to the degree of cooperation between thick and thin myofilaments and to the regulation of Ca²⁺-responsiveness of cardiac sarcomeres (Moore et al. 2016). It is well known that in the presence of Tpm-free F-actin the activation of ATP turnover by myosin subfragment-1 (S1) is linear as a function of added S1, yet becomes sigmoidal when either Tpm or troponin-Tpm is present. Hence, Tpm inhibits the actomyosin ATPase in the presence and absence of troponin and Ca²⁺ at low S1:F-actin ratios and activates the ATPase at moderate to high S1:F-actin ratios more strongly than pure F-actin filaments, relative to control F-actin values. Inhibition and activation by Tpm are greatest when troponin is present (Lehrer & Morris, 1982). Ca²⁺ responsiveness can be increased by extrinsic factors such as low-ATP and/or high-ADP concentrations or factors intrinsic to the myosin motor domain such as a high duty-ratio. Alternatively, factors that shorten the duration of strongly attached states such as high concentrations of ATP, orthophosphate and a low duty-ratio tend to reduce Ca²⁺ responsiveness.

The increase in Ca²⁺ responsiveness observed under low load conditions can be explained by the fact that [Ca²⁺] controls the number of attached motors (Fig. 11 and see Linari et al. 2007; Pinzauti et al. 2018), which in isometric condition, a condition in which the duty ratio is constant, can be taken as an indication of the number of motors available for actin interaction. According to this hypothesis, ATP turnover and in vitro motility will reach a maximum value when the number of interacting motors has reached a critical threshold. This interpretation is supported by a study investigating the regulation of force and unloaded sliding speed by means of individual reconstituted thin filaments (Homsher et al. 2000). Here, it was observed that the speed of actin sliding becomes slower when the fraction of thin filament activation falls below 0.15. The hypothesis that the presence of rumenic acid affects the threshold number of interacting motors necessary for the maximum velocity in the in vitro motility assay has been tested in Fig. 11 by plotting V_f (normalized values from Fig. 8B) against the fraction of available motors, as deduced from the isometric force–pCa relation in Fig. 9B. The curves obtained in the absence and presence of rumenic acid can be best fitted by hyperbolae with half-maximum values corresponding to 0.080 and 0.018, respectively. The greatest changes in the fraction of available myosin motors occur in the physiological pCa range, attained during cardiac systole. Under these conditions, we can conclude that rumenic acid facilitates the sliding velocity of thin filament at zero to low loads by reducing the level of activation of thick filaments required for the development of maximum V_f. At saturating Ca²⁺ concentrations, neither k_tr, the rate constant of force redevelopment, nor β, the fraction of motors attached at T_0, are affected by rumenic acid. Thus, the reduction in T₀ is accounted for by a proportional reduction of F₀, the force generated per myosin motor.

Figure 11. Calculated relationships between V_f and the fraction of available motors

A and B, V_f/V_fmax - fraction of available motors relationship in the absence (A) and presence (B) of 20 μM rumenic acid. The hyperbolic curves give estimates for K_app values of 0.080 ± 0.001 in the absence and 0.018 ± 0.001 in the presence of 20 μM rumenic acid. Numbers next to dots on the curve indicate the pCa value at which measurements were performed. The ordinate is normalized for the maximum value at pCa 4.

In summary, our results provide mechanistic insights about the allosteric mechanism and the extent to which rumenic acid modulates the function of mammalian cardiac and non-muscle myosin-2 isoforms. Our results show that binding of rumenic acid to the motor domain of cardiac myosin-2 reduces the force output per motor by 23% during isometric contraction. High postprandial plasma free rumenic acid concentrations can be reached following a meal that includes ruminant meat or dairy products. The half-maximal effective concentrations for rumenic acid-induced changes in ATP-turnover and isometric force production lie in the region between 4 and 7 μM. The fact that the K_M values of mammalian long-chain acyl-CoA ligase isoforms that regulate the uptake of free fatty acids and control the intracellular equilibrium with CoA-activated fatty acids are in the same range, 3 to 7 μM for saturated and unsaturated C18 fatty acids (Klett et al. 2017), makes it appear likely that the concentration of free rumenic acid can exceed threshold levels in cardiomyocytes where the compound starts to effectively inhibit myosin motor activity. Similar to investigational drug candidates such as omecamtiv mecarbil, CK-274, mavacamten and MYK-491, rumenic acid acts as an allosteric effector of cardiac myosin motor domains (Malik et al. 2011; Kawas et al. 2017; Planelles-Herrero et al. 2017; Anderson et al. 2018). While the exact therapeutic mechanism of action of these drug candidates is still under investigation, our results suggest that rumenic acid binds to the same region of the myosin motor domain. Therapeutic upregulation of the myosin SRX state by mavacamten has been discussed as a particularly promising approach for the treatment of genetic cardiomyopathies and for the reduction of metabolic energy demand during ischaemia in stunned myocardium (Anderson et al. 2018; Repetti et al. 2019). Although compared to the effect of 1 μM mavacamten, 20-fold higher, non-physiological concentrations of rumenic acid are required to achieve a comparable increase in SRX state occupancy in rat heart ventricular trabeculae, EC₅₀ values of 0.20 ± 0.08 μM and 1.2 ± 0.05 μM, measured for the rumenic acid-mediated activation of the actin-activated ATPase of porcine β-cardiac HMM and the inhibition of SRX turnover rates in reconstituted porcine cardiac myosin thick filaments, show the potential of the compound to interfere with mavacamten. In the presence of physiologically relevant concentrations of rumenic acid (≤ 5 μM), activation of the DRX state and an associated increase in metabolic energy demand are the main effects. These outweigh by far the concomitant inhibition of the SRX-ATPase and a slight population shift in favour of the SRX state. The outcome of clinical trials testing the potential of mavacamten and related small molecule therapeutics aimed at cardiac myosin may thus be affected by the dietary uptake of rumenic trans fats by participants in these trials. Dietary intake of rumenic trans fats should therefore be closely controlled during such trials and possibly restricted once they are approved as therapeutics.

Biographies

• 

  Irene Pertici received her PhD in Molecular Medicine in 2018 under the supervision of Vincenzo Lombardi at the PhysioLab, University of Florence. The present study, defining the mechanism of rumenic acid-mediated allosteric modulation of cardiac myosin, was partly realized during her stay at the Institute for Biophysical Chemistry at the Hannover Medical School. She is currently a postdoctoral researcher at the University of Florence.

• 

  Manuel H. Taft is a group leader at the Institute for Biophysical Chemistry, Hannover Medical School. He obtained his PhD in biochemistry for work on the molecular mechanisms of cytoskeletal myosin motor function. As a postdoctoral researcher, he conducted research on small molecule effectors of myosin function. His current work addresses disease mechanism associated with actinopathies and myosinopathies, with a special focus on the role of class 18 myosins.

Open Research

Data availability statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Additional information

Competing interests

The authors declare that they have no competing interests.

Author contributions

D.J.M. conceived the study. I.P., M.H.T., R.F., D.J.M., and M.C. designed experiments and implemented procedures. I.P., J.N.G. and M.H.T. purified proteins, reconstituted complexes and optimized assay conditions. I.P., J.N.G. and M.H.T. performed ATPase and in vitro motility measurements. R.F. and M.H.T. performed homology modelling and docking analysis. I.P. and M.C. performed mechanical experiments. M.H.T. performed fast transient kinetics and single mantATP turnover kinetics. All authors participated in data analysis and contributed sections of the manuscript text. I.P., M.H.T., R.F., M.C. and D.J.M. generated figures. I.P., M.H.T., M.C. and D.J.M. contributed to drafting and editing of the manuscript. All authors have approved the final version of the manuscript and agree to be accountable for all aspects of the work. All persons designated as authors qualify for authorship, and all those who qualify for authorship are listed.

Funding

I.P. was supported by an Erasmus+ Higher Education Scholarship and the PhD Program in Molecular Medicine, University of Siena. MHT received support from the Volkswagen Foundation, Joint Lower Saxony–Israeli Research Projects, grant number VWZN3012. M.C. was supported by the University of Florence (competitive project rictd1819). D.J.M. was supported by Deutsche Forschungsgemeinschaft (DFG) grant MA1081/23-1. R.F. was supported by DFG grant FE 1510/2-1. I.P., J.N.G. and D.J.M. are members of the European Joint Project on Rare Diseases Consortium ‘PredACTINg’ with support from the Italian Ministry of Education, Universities and Research under Decreto Direttoriale 1638 del 19/10/2020 (IP) and the German Federal Ministry of Education and Research under Grant Agreement 01GM1922B. D.J.M. and R.F. are members of the Cluster of Excellence RESIST (EXC 2155) with support from the DFG (Project ID 39087428-B11).

Acknowledgements

We thank Professor Vincenzo Lombardi for his criticism and insightful comments on the manuscript and Professor Marco Linari for discussion and suggestions during the course of this work.

Open access funding enabled and organized by Projekt DEAL.