Optical Control of Cardiac Function with a Photoswitchable Muscarinic Agonist

Light-triggered reversible modulation of physiological functions offers the promise of enabling on-demand spatiotemporally controlled therapeutic interventions. Optogenetics has been successfully implemented in the heart, but significant barriers to its use in the clinic remain, such as the need for genetic transfection. Herein, we present a method to modulate cardiac function with light through a photoswitchable compound and without genetic manipulation. The molecule, named PAI, was designed by introduction of a photoswitch into the molecular structure of an M2 mAChR agonist. In vitro assays revealed that PAI enables light-dependent activation of M2 mAChRs. To validate the method, we show that PAI photoisomers display different cardiac effects in a mammalian animal model, and demonstrate reversible, real-time photocontrol of cardiac function in translucent wildtype tadpoles. PAI can also effectively activate M2 receptors using two-photon excitation with near-infrared light, which overcomes the scattering and low penetration of short-wavelength illumination, and offers new opportunities for intravital imaging and control of cardiac function. cardiac photoregulation, cholinergic, muscarinic, photopharmacology, dualsteric agonist, two-photon


Introduction
Remote spatiotemporal control of physiological processes may provide novel treatment opportunities. Cardiopathies are paradigmatic in this regard because of the rapid time course and complex integration of electrophysiological and molecular events in very specific areas of the heart. For instance, most cardiac rhythm control strategies rely on antiarrhythmic drugs (AADs) targeting ionic currents, whose effects cannot be regulated spatiotemporally. As a result, AADs often give rise to intolerable side effects, including ventricular pro-arrhythmogenicity, and are only partially effective. Overcoming the high failure and complication rates of current therapeutic strategies to treat these diseases will require both patient-personalized determination of the specific physiopathological mechanism and qualitative pharmacological breakthroughs. 1 The application of light and optical techniques in medicine has had a profound impact over the last several decades, in diagnostics, surgery and therapy. 2 In particular, photoexcitation of intrinsic molecules or exogenous light-sensitive agents introduced in the body can affect the tissues and cells within in various ways, via the generation of heat (photothermal), chemical reactions (photochemical), and biological processes (photobiological/photopharmacological or optogenetic). 2 The potential of light as a therapeutic tool with high spatiotemporal resolution has been recently investigated in the cardiovascular field, particularly for arrhythmias, through optogenetics. [3][4][5][6][7] However, the application of such genetic techniques to human subjects with therapeutic purposes is still hampered by safety, regulatory and economic hurdles. Unlike optogenetics, photopharmacology rely on the use of exogenous light-regulated small molecules that can photocontrol native targets and that could be tested and approved using standard drug development procedures. [8][9][10][11][12][13][14][15] These molecules can be used in combination with devices that deliver light to specific locations in the body 7,[16][17][18] in order to remotely control drug dosing and duration of action. Since the activity of drugs is structure-dependent, reversible photoresponsive drugs are obtained by the rational introduction of a molecular photoswitch into the structure of a bioactive compound. 8,14 Cardiac function is controlled by the autonomic sympathetic and parasympathetic nervous systems, which act via adrenoceptors and muscarinic acetylcholine receptors (mAChRs), respectively. 19 In particular, stimulation of β1 and β2 adrenergic receptors increases the heart rate (positive chronotropy) and contractility (positive inotropy), whilst stimulation of M2 mAChRs decreases heart rate and prolongs the atrioventricular conduction time. 20,21 Thus, adrenergic and muscarinic receptors constitute suitable target candidates to control cardiac function with light. Muscarinic acetylcholine receptors (mAChRs) belong to class A G protein-coupled receptors (GPCRs) and are divided in five different subtypes (M1-M5). 22 M2 receptor is extensively expressed in the heart. All five mAChRs are characterized by a high sequence homology in the orthosteric site located in the transmembrane region. This fact limits the development of subtype-selective orthosteric agonists. On the other hand, the allosteric site located in the extracellular loop is less conserved, thus muscarinic allosteric agents are commonly endowed with a more pronounced subtype-selectivity. 23 A chemical strategy commonly applied to overcome such limitation is the incorporation, within the same molecular structure, of two distinct pharmacophore elements belonging to (a) high-affinity orthosteric agonists and (b) highly selective allosteric ligands. 24 Iperoxo-like orthosteric agonist moiety, and (b) the M2-selective allosteric fragments derived from W84 and Naphmethonium (Fig.  1). The incorporation of a photoisomerizable unit into the structure of a dualsteric agonist should enable controlling with light the mutual position of the orthosteric and the allosteric moieties, presumably leading to differences between the two isomers in receptor affinity and efficacy. We chose an azobenzene core as photoresponsive component because of the favourable characteristics that azobenzene-based photoswitches normally display for biological purposes in comparison to other photoswitches, such as design flexibility, large changes in geometry upon isomerization, high photostationary states and fatigue resistance, fast photoisomerization rates, and chemical stability, among others. 33 Moreover, the use of arylazo compounds has been proven safe in humans for some approved drugs and food colorants. 33  PAI (1) and NAI (2) were prepared via two subsequent Menshutkin reactions between the azobenzene linker (4) and the corresponding allo-and orthosteric intermediates (5 and 6, 9) (Scheme 1). Compound 3 was synthesized via the typical Mills reaction and successively brominated photochemically to afford the desired linker 4. Notably, this photochemical reaction exempted us from using a radical initiator 34 and gave an excellent yield (96%), proving for the first time that light-induced benzylic halogenations can be conveniently used also for the preparation of such versatile photoswitchable linkers. Compounds 5, 6 and 9 were prepared as previously reported from commercially available starting materials (Scheme 1 and SI). [24][25][26]31 As a prerequisite for a reversible light-dependent control of their biological activity, PAI and NAI need to effectively behave as reversible photoswitches, which means that the photoisomerization should be relatively fast and quantitatively significant in both directions. UV/Vis spectroscopy experiments showed that PAI and NAI have the typical absorption bands of conventional azobenzenes. PAI can be isomerized to the cis form (about 73% conversion) by applying 365 nm light, while it thermally relaxes back to the trans form in several hours at room temperature. It can be also effectively back-isomerized to the trans form by applying white or blue (460 nm) light (83% trans) ( Fig. 2 and SI, Fig. S2). Surprisingly, NAI resulted refractory to photoisomerization (only 23% cis after 10 min at 365 nm, SI, Fig. S2.1CD), which shows that rational design of azobenzene-containing ligands does not always afford the expected results. We hypothesized that the absorption and emission properties of the naphthalimide moiety 35 could interfere with its photochromism. Given the unsatisfactorily photochromic behaviour of NAI, we selected only PAI for further studies.

PAI allows reversible photo-activation of M2 mAChRs in calcium imaging experiments and molecular docking simulations.
The photopharmacological properties of PAI were first assessed in vitro with real-time calcium imaging assays in transiently transfected HEK cells under 1P-illumination ( Fig. 3 and SI). We tested also the non-photoresponsive muscarinic agonist Iperoxo (IPX) 36 as a control (Fig. 3a). The application of trans-PAI (dark-adapted state) induced cytosolic calcium oscillations indicative of M2 agonism, which were reduced by converting PAI to its cis form upon illumination with UV light (365 nm) (Fig. 3bc). Calcium oscillations could be restored after back-isomerizing PAI to the trans configuration using blue light (460 nm). The time course of calcium responses during activation with trans-PAI displayed a diversity of behaviours in individual cells (Fig. 3c), including oscillatory waves, transient peaks, and step responses as previously observed with PLC-activating GPCRs. 12 Quantification of photoresponses (ΔF/F 0 ) to PAI application and 365 nm illumination shows a reduction in the calcium signal induced by UV light pulses (Fig. 3d). Intriguingly, PAI activated M2 mAChRs in the range of picomolar concentrations, similarly to the super-agonist Iperoxo. 36 Thus, we demonstrated that PAI can effectively activate M2 mAChRs in vitro in its dark-adapted (trans) form and its activity can be reversibly switched off and on with light.In order to account for the observed photoswitchable activity of PAI in M2 mAChR, we looked for putative differences on the receptor level regarding binding efficacy of cis-and trans-PAI using molecular docking simulations (see SI for details). PAI isomers were docked into their theoretical binding site at the human M2 mAChR (PDB 4MQT). Our results suggested that trans-PAI can bind to the M2 mAChR in a typical dualsteric pose compatible with receptor activation (SI, Fig. S5.1a). 37 In contrast, a flipped orientation is favoured in the case of the cis-isomer (SI, Fig. S5.1b). This binding pose is likely incompatible with receptor activation and provides a possible explanation for the light-dependent efficacy of PAI.
Trans-PAI is more effective than cis-PAI at inducing bradycardia and PR lengthening in rats. Once established that PAI allows lightdependent reversible activation of M2 mAChRs, we aimed at testing it as an agent to photocontrol cardiac function in vivo. We initially used Wistar rats for our experiments. The intraperitoneal administration of PAI induced progressive bradycardia and PR lengthening in a dose-dependent manner in both configurations ( Fig. 4 and SI, Fig. S6.1). These effects were accompanied with Fig. 4. In vivo effect of trans-and cis-PAI on the cardiac activity of rats. The activity of dark-relaxed (trans, grey plots) and UV-illuminated PAI (cis, purple plots) administered intraperitoneally in anesthetized rats was tested by means of electrocardiography. The heart rate (left panel) and PR interval (right panel) are plotted as a function of increasing doses of both isomers, which induced progressive bradycardia and PR lengthening in a dose-dependent manner. Significant differences between the dark-relaxed and UV-treated PAI were found in the heart rate and PR interval at the higher doses (*p < 0.05; ***p < 0.001 trans vs cis), in agreement with the higher agonist activity of the trans form observed in vitro and in tadpoles. Three rats in the trans-group died because of extreme bradycardia after the 100 µM/kg dose. The effects of PAI were reversible only upon administration of the muscarinic antagonist Atropine.
variable degrees of systemic parasympathetic effects, such as salivation, urination and defecation. At low doses (≤ 3 µM/kg), both isomers yielded a similar small effect, but remarkably differed at intermediate and high doses. At 10 µM/kg PAI and higher doses, the PR interval was significantly more prolonged in trans; at 30 µM and higher doses, heart rate was also lower in trans. The effects of PAI could not be photoswitched either with blue or with UV light, showing that the ability of light to penetrate murine cardiac tissue at those wavelengths is likely not sufficient to reach M2 mAChR location. Only the administration of atropine (2 mg) completely reverted bradycardia, PR lengthening and systemic parasympathetic effects in both groups (SI, Fig. S6.1). These results demonstrated an enhanced parasympathetic activity for the trans-isomer, and confirmed in mammals the previous findings observed in cells.

PAI enables reversible photocontrol of cardiac activity in
Xenopus tropicalis tadpoles. As an alternative to demonstrate reversible control of cardiac function in vivo, we turned to an animal model in which light scattering is known to be low thus allowing better light penetration. We selected Xenopus tropicalis tadpoles for this purpose since they are translucent and are recognized as an excellent model for studying the human cardiovascular system (Fig. 5). 38,39 Moreover, we had already successfully used video light microscopy to acquire real-time images of the developing beating heart by digitizing the expanding and contracting blood pool in early translucent hearts (Fig. 5a). 38,40 In the absence of PAI, the cardiac rate of tadpoles remained nearly constant at 2.3 ± 0.1 beats·s -1 during control illumination with UV light and at 2.10 ± 0.01 beats·s -1 in the dark. The variability score (V.S.) was 7.97 ± 0.07. Upon administration of 10 µM trans-PAI, heart rate decreased dramatically (0.41 ± 0.02 beats·s -1 in the trace of Fig. 5e) leading in some cases to cardiac arrest (Fig. 5c). Heart beating recovered progressively upon UV illumination (cisisomerisation, 1.28 ± 0.02 beats·s -1 , Fig. 5f), and was not altered in the dark since thermal relaxation is slow. Some animals displayed less stable cardiac rate during UV periods compared to controls  , Fig. S6.2). Adding 10 μM trans-PAI under dim light reduces the heart rate in animals 2, 3 and 4. UV illumination isomerizes PAI to the cis form and the heart rate is partially recovered. Dim red light does not isomerize PAI (SI Appendix, Fig. S6.1) and heart rate is relatively stable. White light converts PAI to the trans isomer, causing cardiac arrest in all 4 animals. UV light restores heartbeat in all animals, some displaying an unstable rate. Several white/UV light cycles were repeated in some animals, showing similar effects. e) Quantification of heart rate during the last minute of each period (beats·s-1, n=4 tadpoles) in control conditions, under white light (trans-PAI) and under UV light (cis-PAI). f) Two-way for repeated measures ANOVA was performed with uncorrected Fisher's LSD test, significance values were established with a p-value = 0.05. Error bars represent standard error of the mean (SEM). The heart rate was significantly higher under UV illumination compared to visible light (p-value < 0.05). Both isomers produced a significant reduction of heart rate in comparison to controls (p-value < 0.001).
(V.S. of 17.4 ± 0.2 and 8.30 ± 0.02, respectively; SI, Fig. S7.3). Subsequent illumination of the animals with visible light (cis-to-trans isomerization) again reduced cardiac rate and eventually interrupted heart beating (V.S. of 374 ± 47.7, SI, Fig. S7.3). Cardiac activity was restored by later exposition to UV light, and further UV/visible light cycles confirmed the reversibility of the pharmacological effects. (see example Supporting Movie S1). Overall, these experiments demonstrated that PAI allows remote and reversible control of heart rate with light in living animals.

Activation of PAI with near infrared (NIR) light.
In order to overcome the scattering and low penetration of violet and visible illumination, we tested whether PAI could be used to activate M2 mAChRs at longer wavelengths. In fact, a critical aspect that must be addressed to unleash the full potential of light-regulated drugs and favour their translation into clinic is their responsiveness to red or NIR radiation, 44,41 which enables higher penetration through tissue, abolishes photodamage and, in the case of 2P excitation, allows three-dimensional subcellular resolution. [42][43][44][45][46] PAI has an excellent thermal stability in both configurations (Fig. S2.3) and is photochemically suited for cis-to-trans photoisomerization with NIR light under 2P-excitation, which encouraged us to test its effects in living cells in real-time calcium imaging assays using a confocal microscope equipped with a pulsed laser. (Fig. 6). PAI was initially applied in its cis (off) state, which as expected did not produce cytosolic calcium oscillations. Subsequent illumination at 840 nm induced robust calcium responses, as previously observed in calcium imaging experiments for cis-to-trans photoisomerization under 1P-excitation (Fig. 6a). These results are quantified in Fig.  6c. The responses (ΔF/F0) obtained for cis-PAI (1P pre-irradiation at 365 nm) are comparable to controls, and 2P-induced isomerization to trans-PAI achieves calcium responses nearly as high as perfusion of iperoxo. It is worth noting that even under NIR excitation PAI maintains an outstanding potency (picomolar) to activate M2 mAChRs, which is rarely observed in photoswitches. 11,44,47,48 Interestingly, 2P microscopy is extensively used for intravital imaging including cardiovascular imaging at subcellular resolution. 49,50 Thus, PAI has a bright future to control cardiac function with light.

Conclusion
The rapid and reversible control of cardiac activity is of particular interest in medicine, including the spatiotemporal manipulation of Error bars are ± SEM. Data were analyzed by using one-way ANOVA with Sidak post hoc test for multiple comparisons for statistical significance (p-value (****) < 0.0001; GraphPad Prism 6).
close anatomic structures bearing different electrophysiological functions in the heart. Light-activated cardiac drugs could be selectively enhanced in certain regions of the heart (e.g., preventing undesired pro-arrhythmogenic ventricular effects when atria are targeted), or at certain times (on-demand, i.e., active only during atrial fibrillation or bradycardia). For that purpose, cardiac patches with integrated electronics and electric stimulation 46 could be further equipped with optoelectronic devices for photostimulation. Drug-based cardiac photoregulation techniques offer potential advantages compared to electric stimulation of cardiac muscle, which produces inhomogeneous areas of de-and hyperpolarization, causes faradaic reactions that alter pH, and produce toxic gases (H 2 , O 2 , Cl 2 ), all of which would be prevented by light-stimulation.
To this end, we have developed the first photoswitchable compound that enables control of cardiac activity with light in wildtype animals without genetic manipulation. To the best of our knowledge, PAI is also the first photoswitchable M2 mAChR agonist to be reported. Despite the changes introduced in the ligand structure in order to photoregulate its activity, PAI retains the high potency of its parent compounds Iperoxo and P-8-Iper. 31,36 PAI activates M2 receptors in its trans configuration and can be reversibly photoswitched with different wavelengths including NIR light under 2P excitation. Future experiments will be addressed to demonstrate that PAI enables precise spatiotemporal control of cardiac function in mammalians in combination with 2P cardiovascular imaging.

Supporting Information
The Supporting Information is available free of charge on the ACS Publications website. Detailed materials and methods, synthetic procedures, chemical analyses, and any additional data and figures as noted in the text (PDF).

Conflicts of interest
There are no conflicts to declare. EG designed and performed in vivo experiments in rats. PG conceived the project and designed experiments. FR, CM, and PG wrote the paper with contributions from all authors. All authors have given approval to the final version of the manuscript. † Fabio Riefolo and Carlo Matera contributed equally.

Authors information
All authors have given approval to the final version of the manuscript.

Funding Sources
This project has received funding from the EU Horizon 2020 Framework Programme for Research and Innovation under the Specific Grant Agreement 2 (Human Brain Project WaveScalES SGA2 No. 785907), AGAUR/Generalitat de Catalunya (CERCA Programme, 2017-SGR-1442 and 2017-SGR-1548), FEDER funds, ERANET SynBio MODULIGHTOR, Fundaluce foundation, Ramón Areces foundation, MINECO (FPI fellowship BES-2014-068169 and project CTQ2016-80066R), and CATCH-ME (grant agreement nº 633196). CM was supported by the Ermenegildo Zegna Founder's Scholarship. LA was supported by the graduate program "Receptor Dynamics -Emerging paradigms for novel drugs" funded by the Elite Network of Bavaria.

ACKNOWLEDGMENTS.
The authors are grateful to Jean-Philippe Pin for providing the chimeric Gi/Gq protein clone and to Núria Camarero for helping during preliminary in vitro experiments. CM and FR are grateful to Prof. Marco De Amici for helpful discussion and continuous support. Molecular graphics and analyses were performed with the UCSF Chimera package. Chimera is developed by the Resource for Biocomputing, Visualization, and Informatics at the University of California, San Francisco (supported by NIGMS P41-GM103311). Mass spectrometry was performed at the IRB Barcelona Mass Spectrometry Core Facility, which actively participates in the BMBS European COST Action BM 1403 and is a member of Proteored, PRB2-ISCIII, supported by grant PRB2 (IPT13/0001 -ISCIII-SGEFI / FEDER).