A novel machine learning-based screening identifies statins as inhibitors of the calcium pump SERCA.

We report a novel small-molecule screening approach that combines data augmentation and machine learning to identify Food and Drug Administration (FDA)-approved drugs interacting with the calcium pump (Sarcoplasmic reticulum Ca2+-ATPase, SERCA) from skeletal (SERCA1a) and cardiac (SERCA2a) muscle. This approach uses information about small-molecule effectors to map and probe the chemical space of pharmacological targets, thus allowing to screen with high precision large databases of small molecules, including approved and investigational drugs. We chose SERCA because it plays a major role in the excitation-contraction-relaxation cycle in muscle and it represents a major target in both skeletal and cardiac muscle. The machine learning model predicted that SERCA1a and SERCA2a are pharmacological targets for seven statins, a group of FDA-approved 3-hydroxy-3-methylglutaryl coenzyme A reductase inhibitors used in the clinic as lipid-lowering medications. We validated the machine learning predictions by using in vitro ATPase assays to show that several FDA-approved statins are partial inhibitors of SERCA1a and SERCA2a. Complementary atomistic simulations predict that these drugs bind to two different allosteric sites of the pump. Our findings suggest that SERCA-mediated Ca2+ transport may be targeted by some statins (e.g., atorvastatin), thus providing a molecular pathway to explain statin-associated toxicity reported in the literature. These studies show the applicability of data augmentation and machine learning-based screening as a general platform for the identification of off-target interactions and the applicability of this approach extends to drug discovery.

A multiscale approach for bridging the gap between potency, efficacy, and safety of small molecules directed at membrane proteins.

Membrane proteins constitute a substantial fraction of the human proteome, thus representing a vast source of therapeutic drug targets. Indeed, newly devised technologies now allow targeting "undruggable" regions of membrane proteins to modulate protein function in the cell. Despite the advances in technology, the rapid translation of basic science discoveries into potential drug candidates targeting transmembrane protein domains remains challenging. We address this issue by harmonizing single molecule-based and ensemble-based atomistic simulations of ligand-membrane interactions with patient-derived induced pluripotent stem cell (iPSC)-based experiments to gain insights into drug delivery, cellular efficacy, and safety of molecules directed at membrane proteins. In this study, we interrogated the pharmacological activation of the cardiac Ca2+ pump (Sarcoplasmic reticulum Ca2+-ATPase, SERCA2a) in human iPSC-derived cardiac cells as a proof-of-concept model. The combined computational-experimental approach serves as a platform to explain the differences in the cell-based activity of candidates with similar functional profiles, thus streamlining the identification of drug-like candidates that directly target SERCA2a activation in human cardiac cells. Systematic cell-based studies further showed that a direct SERCA2a activator does not induce cardiotoxic pro-arrhythmogenic events in human cardiac cells, demonstrating that pharmacological stimulation of SERCA2a activity is a safe therapeutic approach targeting the heart. Overall, this novel multiscale platform encompasses organ-specific drug potency, efficacy, and safety, and opens new avenues to accelerate the bench-to-patient research aimed at designing effective therapies directed at membrane protein domains.

Brugada syndrome trafficking-defective Nav1.5 channels can trap cardiac Kir2.1/2.2 channels.

Cardiac Nav1.5 and Kir2.1-2.3 channels generate Na (INa) and inward rectifier K (IK1) currents, respectively. The functional INa and IK1 interplay is reinforced by the positive and reciprocal modulation between Nav15 and Kir2.1/2.2 channels to strengthen the control of ventricular excitability. Loss-of-function mutations in the SCN5A gene, which encodes Nav1.5 channels, underlie several inherited arrhythmogenic syndromes, including Brugada syndrome (BrS). We investigated whether the presence of BrS-associated mutations alters IK1 density concomitantly with INa density. Results obtained using mouse models of SCN5A haploinsufficiency, and the overexpression of native and mutated Nav1.5 channels in expression systems - rat ventricular cardiomyocytes and human induced pluripotent stem cell-derived cardiomyocytes (hiPSC-CMs) - demonstrated that endoplasmic reticulum (ER) trafficking-defective Nav1.5 channels significantly decreased IK1, since they did not positively modulate Kir2.1/2.2 channels. Moreover, Golgi trafficking-defective Nav1.5 mutants produced a dominant negative effect on Kir2.1/2.2 and thus an additional IK1 reduction. Moreover, ER trafficking-defective Nav1.5 channels can be partially rescued by Kir2.1/2.2 channels through an unconventional secretory route that involves Golgi reassembly stacking proteins (GRASPs). Therefore, cardiac excitability would be greatly affected in subjects harboring Nav1.5 mutations with Golgi trafficking defects, since these mutants can concomitantly trap Kir2.1/2.2 channels, thus unexpectedly decreasing IK1 in addition to INa.

hiPSC-CM Monolayer Maturation State Determines Drug Responsiveness in High Throughput Pro-Arrhythmia Screen.

Human induced pluripotent stem cell derived cardiomyocytes (hiPSC-CMs) offer a novel in vitro platform for pre-clinical cardiotoxicity and pro-arrhythmia screening of drugs in development. To date hiPSC-CMs used for cardiotoxicity testing display an immature, fetal-like cardiomyocyte structural and electrophysiological phenotype which has called into question the applicability of hiPSC-CM findings to the adult heart. The aim of the current work was to determine the effect of cardiomyocyte maturation state on hiPSC-CM drug responsiveness. To this end, here we developed a high content pro-arrhythmia screening platform consisting of either fetal-like or mature hiPSC-CM monolayers. Compounds tested in the screen were selected based on the pro-arrhythmia risk classification (Low risk, Intermediate risk, or High risk) established recently by the FDA and major stakeholders in the Drug Discovery field for the validation of the Comprehensive In vitro Pro-Arrhythmia Assay (CiPA). Here we show that maturation state of hiPSC-CMs determines the absolute pro-arrhythmia risk score calculated for these compounds. Thus, the maturation state of hiPSC-CMs should be considered prior to pro-arrhythmia and cardiotoxicity screening in drug discovery programs.

Dominant frequency increase rate predicts transition from paroxysmal to long-term persistent atrial fibrillation.

BACKGROUND: Little is known about the mechanisms underlying the transition from paroxysmal to persistent atrial fibrillation (AF). In an ovine model of long-standing persistent AF we tested the hypothesis that the rate of electric and structural remodeling, assessed by dominant frequency (DF) changes, determines the time at which AF becomes persistent. METHODS AND RESULTS: Self-sustained AF was induced by atrial tachypacing. Seven sheep were euthanized 11.5±2.3 days after the transition to persistent AF and without reversal to sinus rhythm; 7 sheep were euthanized after 341.3±16.7 days of long-standing persistent AF. Seven sham-operated animals were in sinus rhythm for 1 year. DF was monitored continuously in each group. Real-time polymerase chain reaction, Western blotting, patch clamping, and histological analyses were used to determine the changes in functional ion channel expression and structural remodeling. Atrial dilatation, mitral valve regurgitation, myocyte hypertrophy, and atrial fibrosis occurred progressively and became statistically significant after the transition to persistent AF, with no evidence for left ventricular dysfunction. DF increased progressively during the paroxysmal-to-persistent AF transition and stabilized when AF became persistent. Importantly, the rate of DF increase correlated strongly with the time to persistent AF. Significant action potential duration abbreviation, secondary to functional ion channel protein expression changes (CaV1.2, NaV1.5, and KV4.2 decrease; Kir2.3 increase), was already present at the transition and persisted for 1 year of follow up. CONCLUSIONS: In the sheep model of long-standing persistent AF, the rate of DF increase predicts the time at which AF stabilizes and becomes persistent, reflecting changes in action potential duration and densities of sodium, L-type calcium, and inward rectifier currents.

Inhibition of platelet-derived growth factor-AB signaling prevents electromechanical remodeling of adult atrial myocytes that contact myofibroblasts.

BACKGROUND: Persistent atrial fibrillation (PAF) results in electromechanical and structural remodeling by mechanisms that are poorly understood. Myofibroblast proliferation and fibrosis are major sources of structural remodeling in PAF. Myofibroblasts also interact with atrial myocytes via direct physical contact and release of signaling molecules, which may contribute to remodeling. OBJECTIVE: To determine whether myofibroblasts contribute to atrial myocyte electromechanical remodeling via direct physical contact and platelet-derived growth factor (PDGF) signaling. METHODS: Myofibroblasts and myocytes from adult sheep atria were co-cultured for 24 hours. Alternatively adult sheep atrial myocytes were exposed to 1 ng/mL recombitant PDGF AB peptide for 24 hours. RESULTS: Myocytes making contact with myofibroblasts demonstrated significant reduction (P ≤ .05) in peak L-type calcium current density, shortening of action potential duration (APD), and reduction in calcium transients. These effects were blocked by pretreatment with a PDGF-AB neutralizing anti-body. Heterocellular contact also severely disturbed the localization of the L-type calcium channel. Myocytes exposed to recombinant PDGF-AB peptide for 24 hours demonstrated reduced APD50, APD80 and Peak L-type calcium current. Pretreatment with a PDGF-AB neutralizing antibody prevented these effects. Finally, while control atrial myocytes did not respond in a 1:1 manner to pacing frequencies of 3 Hz or higher, atrial myocytes from hearts that were tachypaced for 2 months and normal myocytes treated with PDGF-AB for 24 hours could be paced up to 10 Hz. CONCLUSIONS: In addition to leading to fibrosis, atrial myofibroblasts contribute to electromechanical remodeling of myocytes via direct physical contact and release of PDGF-AB, which may be a factor in PAF-induced remodeling.

Remodeling of mechanical junctions and of microtubule-associated proteins accompany cardiac connexin43 lateralization.

BACKGROUND: Desmosomes and adherens junctions provide mechanical continuity between cardiac cells, whereas gap junctions allow for cell-cell electrical/metabolic coupling. These structures reside at the cardiac intercalated disc (ID). Also at the ID is the voltage-gated sodium channel (VGSC) complex. Functional interactions between desmosomes, gap junctions, and VGSC have been demonstrated. Separate studies show, under various conditions, reduced presence of gap junctions at the ID and redistribution of connexin43 (Cx43) to plaques oriented parallel to fiber direction (gap junction "lateralization"). OBJECTIVE: To determine the mechanisms of Cx43 lateralization, and the fate of desmosomal and sodium channel molecules in the setting of Cx43 remodeling. METHODS: Adult sheep were subjected to right ventricular pressure overload (pulmonary hypertension). Tissue was analyzed by quantitative confocal microscopy and by transmission electron microscopy. Ionic currents were measured using conventional patch clamp. RESULT: Quantitative confocal microscopy demonstrated lateralization of immunoreactive junctional molecules. Desmosomes and gap junctions in lateral membranes were demonstrable by electron microscopy. Cx43/desmosomal remodeling was accompanied by lateralization of 2 microtubule-associated proteins relevant for Cx43 trafficking: EB1 and kinesin protein Kif5b. In contrast, molecules of the VGSC failed to reorganize in plaques discernable by confocal microscopy. Patch-clamp studies demonstrated change in amplitude and kinetics of sodium current and a small reduction in electrical coupling between cells. CONCLUSIONS: Cx43 lateralization is part of a complex remodeling that includes mechanical and gap junctions but may exclude components of the VGSC. We speculate that lateralization results from redirectionality of microtubule-mediated forward trafficking. Remodeling of junctional complexes may preserve electrical synchrony under conditions that disrupt ID integrity.