Small-molecule activation of SERCA2a SUMOylation for the treatment of heart failure

Decreased activity and expression of the cardiac sarcoplasmic reticulum calcium ATPase (SERCA2a), a critical pump regulating calcium cycling in cardiomyocyte, are hallmarks of heart failure. We have previously described a role for the small ubiquitin-like modifier type 1 (SUMO-1) as a regulator of SERCA2a and have shown that gene transfer of SUMO-1 in rodents and large animal models of heart failure restores cardiac function. Here, we identify and characterize a small molecule, N106, which increases SUMOylation of SERCA2a. This compound directly activates the SUMO-activating enzyme, E1 ligase, and triggers intrinsic SUMOylation of SERCA2a. We identify a pocket on SUMO E1 likely to be responsible for N106's effect. N106 treatment increases contractile properties of cultured rat cardiomyocytes and significantly improves ventricular function in mice with heart failure. This first-in-class small-molecule activator targeting SERCA2a SUMOylation may serve as a potential therapeutic strategy for treatment of heart failure.

T here are 26 million people who suffer from heart failure (HF) in the world (5.8 million patients in the United States) 1 . Of all the cardiovascular diseases, HF is the only diagnosis increasing in both incidence and prevalence 2 . For this reason, there is a critical need for novel targets and treatment strategies. The cardiac sarcoplasmic reticulum Ca 2 þ -ATPase (SERCA2a) is a key pump responsible for intracellular calcium handling and contractility in cardiac cells. Impaired calcium reuptake resulting from decreased expression and activity of SERCA2a is a hallmark of HF. The work in our laboratory has led to the successful completion of Phase 1 and Phase 2 clinical trials of adeno-associated vector type 1 (AAV1)-mediated gene transfer of SERCA2a in patients with severe HF, showing clinical benefits in patients receiving AAV1.SERCA2a [3][4][5][6] . An international Phase 2b/3 trial of 250 patients has recently completed enrolment. Beyond its effects on enhancing contractility, SERCA2a gene transfer has been shown to restore cardiac energetics 7,8 , decrease ventricular arrhythmias 9,10 , block smooth muscle cell proliferation and enhance coronary flow through the activation of nitric oxide synthase in endothelial cells 11 . We recently found that in HF, there are changes of posttranslational modifications (PTMs) of SERCA2a that render it dysfunctional and we showed that restoration of SERCA2a by gene transfer does not abrogate the PTMs of the transporter. Therefore, SERCA2a's enzymatic dysfunction, in addition to its decreased expression, needs to be addressed in the failing heart, to normalize calcium cycling.
SUMOylation, a ubiquitin-like reversible PTM where SUMO (small ubiquitin-like modifier) covalently attaches to a target protein through specific enzyme cascades reaction, has been shown to be implicated in controlling many cellular processes, including regulation of protein function, stability and localization 12 . Several studies have demonstrated that protein SUMOylation is associated with critical cellular pathways, many of them ultimately affecting cardiac function and development, suggesting SUMOylation as a valuable target for the treatment of cardiovascular diseases 13,14 . We have recently found that the levels and activity of SERCA2a in cardiomyocytes to be modulated in parallel with the levels of small ubiquitin-like modifier type 1 (SUMO-1). SUMO-1 gene transfer led to restoration of SERCA2a levels, improved haemodynamic performance and reduced mortality in a murine model of pressure overload-induced HF 15 . We further demonstrated that gene transfer of SUMO-1 in combination with SERCA2a led to reversal of HF in a porcine model of ischaemic HF 16 .
In this report, we have conducted a screening study to discover small molecules hat increase SUMOylation. We identify a small molecule, N106 (N-(4-methoxybenzo[d]thiazol-2-yl)-5-(4-methoxyphenyl)-1,3,4-oxadiazol-2-amine (molecular weight 354 g per mol)), that enhances SERCA2a SUMOylation, resulting in enhanced contractility both in vitro and in vivo. In addition, we characterize the putative site on E1 ligase where N106 docks and confers its activity. Thus, our studies have established a compelling rationale for targeting SUMOylation of SERCA2a as a new therapeutic approach for the treatment of HF.

Results
Identification of potent SERCA2a SUMOylation activators. To identify small-molecule activators of the SUMOylation pathway, we used the data from in-vitro SUMOylation of Ran GTPase-activating protein 1 (RanGAP1, a well-known substrate of SUMO-1) 17 . To find novel small molecules capable of activating SUMOylation, an AlphaScreen (PerkinElmer) assay was developed to detect SUMOylation of RanGAP1 with SUMO-1. AlphaScreen is a bead-based fluorescence resonance energy transfer (FRET) technology that results in emission of light when donor and acceptor beads are brought into proximity. A His-tagged RanGAP1 was used as the substrate for the addition of a glutathione S-transferase (GST)-tagged form of SUMO-1 in the presence of E1, E2 (Ubc9) and ATP. Following the SUMOylation reaction, glutathione-coated donor beads and Ni 2 þ -coated acceptor beads are added to interact with GST and His, respectively, effectively capturing the SUMOylated substrate. When the beads are brought into proximity by the SUMOylated protein, irradiation of the donor beads at 680 nm will result in emission of light at 520-620 nm. The K d determined by FRET between SUMO-1 and RanGAP1 is compatible with those determined with other traditional approaches, such as isothermal titration calorimetry and surface plasmon resonance. We have used this technology to screen over 100,000 compounds from large NCI libraries and a few compounds have been shown to activate SUMOylation at a concentration of 10 mM. Among molecules that increased the RanGAP1 SUMOylation, we performed subsequent functional assays. A compound was considered a primary hit if it increased the FRET signal greater than threefold the s.d. above the mean of the vehicle controls when tested at 10 mM. Hits were subsequently filtered to remove all compounds containing substructural features common to promiscuous compounds 18 . Compounds that also appeared as frequent hits in other FRET-based assays were also excluded. We began by screening 50 small molecules for their ability to SUMOylate SERCA2a in HEK-293 cells at 10 mM concentration, 14 of which were non-toxic. Of the 14 compounds tested, we selected eight compounds that increased SERCA2a SUMOylation in HEK-293 cells overexpressing SERCA2a ( Supplementary  Fig. 1a). We then assessed the mechanical properties of the eight compounds using a video-based edge-detection system in isolated cardiomyocytes ( Supplementary Fig. 1b,c). Endogenous SERCA2a SUMOylation status was also determined in the same set of cells ( Supplementary Fig. 1d). In this study, we focused on the small molecules that activate SERCA2a SUMOylation and that improve cardiac function. Of them, only two compounds (N6 and N106) showed both positive inotropic effects and enhancement of endogenous SERCA2a SUMOylation without upregulation of Ubc9 ( Supplementary Fig. 1e,f).
Mechanism of action by N106 through SUMO E1 enzyme. To understand the mechanisms of action by which N106 enhances SERCA2a SUMOylation, we tested the effects of N106 on SUMO E1 activity. We decided to focus on the SUMO E1 activity, as only SUMO E1 and SUMO E2 enzymes were used in the initial screen. Furthermore, the SUMOylation pathway uses a single SUMO E1 enzyme and SUMO E1 is the first control point of SUMOylation cascade. The SUMO E1 activity was determined by three independent experiments. N106 stimulated ATP-dependent activation of SUMO E1 (EC 50 ¼ 2.45 ± 0.13 mM, Fig. 3a), thioester formation between the SUMO-1 and SUMO E1 (Fig. 3b), and subsequent transfer of the activated SUMO-1 from SUMO E1 to the unique SUMO E2 enzyme, Ubc9, in a dose-dependent manner (Fig. 3c). The SUMO conjugation was not observed in the presence of reducing agent, b-mercaptoethanol, indicating a thioester bond ( Supplementary Fig. 2a,b). We further asked whether there was a dose-limiting effect. The maximal tolerated doses of N106 for in-vitro thioester formation between the SUMO E1 and SUMO-1 was B20 mM range. The SUMO-1 conjugation   to SUMO E1 enzyme and SUMO E1-dependent thioester bond formation between SUMO E2 enzyme and SUMO-1 were less pronounced at 50 mM concentration and that was inhibited at over 100 mM concentration ( Supplementary Fig. 3a,b). In addition, the positive effect of N106 on cardiomyocytes contractility was no longer observed at over 50 mM concentration of N106 ( Supplementary Fig. 3c). Taken together, all three experiments suggest that SUMO E1 is a direct target of N106. To better understand the structural basis for the action of N106 on SUMO E1, we performed computational modelling studies. As SUMO E1 undergoes massive conformational changes as it transitions between the open (the resting state) and closed (the activating state) states during the SUMOylation process 19 , we used structures that represent both states as modelling templates. Interestingly, we found that a promising binding pocket for N106 is located at the interface between subunit 1 (SAE1) and subunit 2 (SAE2) that only exists in the closed state of SUMO E1 and not in the open-state structure (Fig. 3d). There are two relevant residues in this pocket, valine 315 (V315) and glutamine 312 (Q312) on SAE1. According to the docking model, V 315 (in purple) points into the N106 binding site; hence, we decided to mutate it to a larger residue such as phenylalanine (V315F). The docking solution also suggests that Q312 (in green) establishes hydrogen bonds with the N106 compound. In this case, we mutated Q312 to alanine (Q312A), to remove the capacity for hydrogen-bond formation. To validate the putative binding site of SAE1 on the pocket that exist in the closed state of SUMO E1, we tested SUMO E1 activity using these mutants of SAE1 with N106. The Q312A and V315F mutants of putative N106 binding site showed significantly lower ATP hydrolysis activity as compared with the wild-type (WT) protein (Fig. 3e). Moreover, on ATP-dependent SUMO E1 activation in the presence of N106, SUMO E1-SUMO and SUMO E2-SUMO conjugations were weaker in the mutant SAE1 as compared with the WT protein (Fig. 3f,g). Finally, Q312A mutant showed the lowest ATP hydrolysis, SUMO conjugation and thioester transfer to the E2, which may suggest the importance of Q312 residue in SUMO E1 for N106-mediated activation. Moreover, N106-induced increase in SUMOylation of SERCA2a was weaker in mutant SAE1-overexpressing cells than in overexpressed WT ( Supplementary Fig. 4). Dosedependent increases in the profiles of SERCA2a SUMOylation induced by N106 was not observed in cells overexpressing mutants ( Supplementary Fig. 4). These data suggest that both the Q312 and V315 residues on SAE1 could be the potential N106-binding sites.
Importantly, N106 did not significantly stimulate any in-vitro activity of the ubiquitin-activating enzyme (Ub E1) at concentrations up to 100 mM, including ubiquitin activation, conjugation of ubiquitin to Ub E1 enzyme and Ub E1-dependent conjugation of ubiquitin to the E2 enzyme cdc32 ( Supplementary Fig. 5). It is important to note that the N106-binding pockets are not present in the Ub E1 enzyme, further indicating that the N106 compound is specific for SUMO E1 enzyme. It is also well known that each ubiquitin-like modifier (for example, ubiquitin, SUMO, NEDD8 and so on) has its own specific activating E1 enzyme that serves as the entry point for an enzyme cascade 20 . Thus, our screen identified a novel activator of the SUMO E1 enzyme.
Haemodynamic effects of N106 in mouse model of HF. On the basis of our biochemical and computational studies, and given the findings obtained in cardiomyocytes, N106 should increase cardiac function via SERCA2a SUMOylation activation in an animal model of HF. To test our hypothesis, we measured haemodynamic parameters to evaluate the acute effects of N106 infusion on the cardiac function of the mouse model of HF. In the first set of experiments, mice were exposed to pressure overload by transverse aortic constriction (TAC) or sham operation and received either N106 or vehicle. At 2 months, post TAC animals developed HF. The left ventricle (LV) was severely dilated and the cardiac function as determined by fractional shortening and ejection fraction, were significantly decreased ( Supplementary  Fig. 6a-d). The SERCA2a protein levels were significantly decreased by B40% in TAC mice as compared with the sham-operated mice (Supplementary Fig. 6e). In addition, the positive chronotropic effects of dobutamine on myocardium were significantly abrogated in TAC mice ( Supplementary Fig. 6f). Following cannulation of the carotid artery and external jugular vein, infusion of N106 or vehicle was performed as described in the Methods section. Pressure and volume measurements were made simultaneously and pressure volume curves were constructed as the vehicle or N106 were infused.
We found a beneficial effect on haemodynamic parameters following the infusion of N106. The end systolic pressure-volume relationship (ESPVR) in the LV was significantly steeper when N106 was infused compared with baseline (Fig. 4a). More importantly, infusion of N106 induced an increase in the index of contractility and ESPVR slope in a dose-dependent manner (Fig. 4b). Maximum dP/dt, an index of LV contractility, increased with increasing concentration of N106 (Fig. 4c), whereas t, a parameter of relaxation, decreased, indicating enhanced relaxation (Fig. 4d). Changes in haemodynamic parameters are shown in Table 1. The vehicle infusion induced no change in the pressure-volume relationship and did not affect the index of contractility, ESPVR and þ dP/dt max in both sham and TAC mice ( Supplementary Fig. 7). We showed enhanced contractility in normal cardiomyocytes after N106 treatment. However, in sham-operated groups there was an increasing trend; however, it was not statistically significant. Significant dose-dependent effects were not observed in those mice ( Supplementary Fig. 8). This may be due to differences between the model systems (cell-based system versus in-vivo system) and bioavailability of the N106 compound in vitro versus in vivo.
Haemodynamic improvement occurred independent of the protein levels of SERCA2a and its regulatory protein. Western blot analyses revealed that SERCA2a protein levels were not significantly changed by N106 infusion (Fig. 4e). Neither protein levels nor phosphorylation states of phospholamban were changed by N106 treatment (Supplementary Fig. 9a,b). There were not significant changes in protein levels of cardiac ryanodine receptor and sodium-calcium exchanger ion channels in N106treated mice ( Supplementary Fig. 9c,d). However, in TAC hearts, SUMOylation of SERCA2a was significantly increased with the N106 treatment as compared with the vehicle control (Fig. 4f). There was an increasing trend in sham-operated group (P ¼ 0.07, Student's t-test). SERCA2a's ATPase activity was somewhat higher in N106-treated TAC hearts than vehicle-treated hearts (P ¼ 0.06, Student's t-test; Supplementary Fig. 9e). These results indicate that N106 activates SUMOylation-dependent activation of SERCA2a proteins.
We further evaluated whether infusion of N106 would affect other SUMO-1 substrates. In TAC heart, N106 treatment significantly induced SUMOylated specificity protein 1 (Sp1) form. SUMOylation of Sp1 is known to decrease its protein stability 21 ; however, protein levels of Sp1 were not changed following N106 treatment (Supplementary Fig. 10a). N106 had no significant effect on GATA4 SUMOylation and its protein levels ( Supplementary Fig. 10b). The whole SUMO-1 conjugates were not significantly changed by N106 infusion ( Supplementary  Fig. 9f).
Haemodynamic effects of N106 in the SERCA2a knockout mouse. In a second set of experiments, we tested whether N106 exerts its effects through SERCA2a. We used adult mice with an inducible cardiomyocyte-specific excision of the Atp2a2 (Serca2) gene (Serca2 KO) 22 . Four weeks after induction of Serca2 gene excision, the mice displayed a substantial reduction in systolic and diastolic function ( Supplementary Fig. 11a). Ejection fraction was significantly lower in the Serca2 knockout (KO) mice compared with the WT control   Supplementary Fig. 11b). In addition, the LV fractional shortening was significantly decreased (Serca2 KO, 43.80±0.79%; WT, 65.04 ± 0.29%, Po0.001, Student's t-test; Supplementary  Fig. 11c) and the LV dimensions, as measured by LV end-diastolic internal diameters and LV end-systolic internal diameters, were significantly increased (Po0.001; Supplementary Fig. 11d,e). The decay constant, t-value, was nearly 1.6-fold (Serca2 KO, 11.78 ± 0.94 ms; WT, 7.0 ± 0.25 ms, Po0.001, Student's t-test) larger in the Serca2 KO mice compared with the WT control ( Supplementary Fig. 11f). As expected, the expression of SERCA2a was completely abolished in the Serca2 KO mice ( Supplementary  Fig. 11g). As shown in Fig. 5 and Table 2, N106 did not induce any improvements in haemodynamic parameters in the Serca2 KO. ESPVR slope (P ¼ 0.17, versus baseline, Student's t-test), þ dP/dt max (P ¼ 0.28, versus baseline, Student's t-test), P max (P ¼ 0.66 versus baseline, Student's t-test), t (P ¼ 0.10, versus baseline, Student's t-test) and heart rate (HR; P ¼ 0.06, versus baseline, Student's t-test) were not different in the presence of N106 (Fig. 5a-f). Furthermore, the dose-dependent enhancement of cardiac contractility by N106 disappeared in the Serca2 KO mice (Fig. 5g,h). Taken together, the animal data strongly suggest that N106 improves cardiac function through SERCA2a protein.

Discussion
In this study, we demonstrated that small-molecule-mediated activation of SERCA2a SUMOylation provided direct beneficial effects on contractility in failing hearts. Based on its ability to SUMOylate SERCA2a, a specific compound (N106) was chosen. It improved contractility in a dose-dependent manner and calcium-transient property in isolated cardiomyocytes. In the pressure-overload-induced HF model, we found that N106 induced sustained reversal of contractility and was associated with an increased SERCA2a SUMOylation profile.
We tested the specificity of this novel compound N106 for SERCA2a, using molecular, cellular and animal model   Fig. 9f). The strongest evidence that the beneficial effects of N106 are through SERCA2a comes from our results in the Serca2 KO mice, where N106 had no effects. However, SUMOylation of other myocardial targets aside from SERCA2a may also contribute to the reversal of cardiac dysfunction.
One concern about an agent that enhances SUMOylation is off-target effects including tumour growth. Several studies reported direct involvement of SUMO E1 in the setting of cancer. In patients with hepatocellular carcinoma, UBA2 gene expression was higher and directly correlated with lower survival rate 25 . SUMO E1 was also reported to be an important factor for myc-dependent tumour growth in mice and there is a correlation between low SUMO E1 gene expression and longer metastasisfree survival rate of patients with myc-high breast cancers 26 . Thus, we sought to determine whether N106 would have an effect on cancer cell growth. We performed the NCI-60 human tumour cell screen used by the National Cancer Institute's Developmental Therapeutics Program (http://dtp.cancer.gov/docs/compare/ compare.html). The NCI-60 set includes leukemia, lymphomas and carcinomas of ovarian, renal, breast, prostate, colon, lung and central nervous system origin. The results of the screen validated that the N106 did not induce any tumour cell growth and proliferation including myc-dependent human breast cancer cell line MDA-MB-231 ( Supplementary Fig. 12). Even though the NCI-60 screen revealed that N106 does not increase tumour cell line growth, further in-vivo toxicology studies are warranted to diminish further perceived concerns for the proliferative potential for this pathway.
N106 exhibited a pharmacokinetic profile suitable for in-vivo evaluation in mice. In a murine model, the half-life of N106 was determined to be B65.4 min with a C max of B2.24 mM when the mice received 10 mg kg À 1 of N106 by intravenous injection. The oral bioavailability (F%) was 56% and 50%, and terminal elimination half-life (t 1/2 ) was 19 min (Supplementary Fig. 13). This pharmacokinetic profile of N106 allows us to consider using N106 intravenously and orally as well. Further experiments will determine the most effective way of delivering this small molecule. Although the half-life of N106 in vivo is quite rapid, N106 increased cell contractility, calcium-transient SERCA2a's ATPase activity and SUMOylation within 10 min of exposure, and these effects were sustained at 24 h in cardiomyocytes (Fig. 2). Thus, chronic studies need to be completed to evaluate long-term efficiency and toxicology of N106.
We also demonstrated the efficacy of N106 in a mouse model of HF. A limitation of using mice to study HF is related to significant differences in cardiac characteristics between mouse and human such as HR, small size, oxygen consumption and adrenergic receptor ratio 27,28 . Therefore, experiments in larger animal models and other models of HF will be essential to develop our discoveries from murine models into clinical therapies.
In summary, a number of experimental studies and now clinical trials have validated SERCA2a as an important target for the treatment of HF. In this study, we identified a small molecule that activates SERCA2a through SUMOylation and induces beneficial effects in vitro and in vivo. This first-in-class smallmolecule activator may provide a novel therapeutic strategy for the treatment of HF.

Methods
Animals. All mice and rats were housed and treated in accordance with NIH and Institutional Animal Care and Use Committee guidelines and the used protocols were approved by the Icahn School of Medicine at Mount Sinai, Animal Care and Use Committees. Studies were conducted in male B6C3F1 mice aged 8-10 weeks (weight, 25B30 g) purchased from Jackson Laboratories (Bar Harbor, ME). Male rats aged 3 months and weighting 180B250 g were obtained from Jackson Laboratories.
Cardiomyocyte isolation and cardiac physiology. Ventricular heart muscle cells were isolated from 180B250 g male Sprague-Dawley rats as described 29 . The hearts were perfused sequentially with low-calcium buffer, enzyme solutions containing collagenases and proteases, and high-potassium solution. Next, the hearts were minced, filtered and resuspended in high-potassium solution. The cardiomyocytes were plated on laminin-coated plates and cultured in medium M199 (Life Technologies, CT, USA) containing proper supplements. The basic culture medium consisted of medium M199 with Earle's salts, 5 mmol l À 1 creatine (Sigma-Aldrich, Saint Louis, MO, USA), 2 mmol l À 1 l-carnitine (Sigma-Aldrich), 5 mmol l À 1 taurine (Sigma-Aldrich), 100 IU ml À 1 penicillin (Life Technologies) and 100 mg ml À 1 streptomycin (Life Technologies). At 4 h after plating, the culture medium was changed and small molecules were added. As indicators of cardiac functions, measurement of intracellular calcium and cell shortening were performed. In brief, the cardiomyocytes were loaded with the calcium indicator Fura-2 and stimulated at different frequencies (0.1B3.0 Hz) using an external stimulator (Grass Technologies, Warwick, RI, USA), whereas a dual-excitation spectrofluorometer (IonOptix Inc., Milton, MA, USA) was used to record fluorescence emissions used to calculate intracellular calcium concentration. Cell shortening was measured using video microscopy motion-detector system.
Transverse aortic constriction. B6C3F1 mice were acquired from Jackson Laboratories. Mice underwent TAC using a supraclavicular construction model as described 15,30 . TAC or sham surgery was performed in 8B10-week-old male mice (25B30 g). Mice were anaesthetized with intraperitoneal ketamine (95 mg kg À 1 ) and xylaxine (5 mg kg À 1 ), and ventilated with a tidal volume of 0.1 ml and a respiratory rate of 110 breaths per minute (Harvard Apparatus, MA, USA). A 2-to 3-mm longitudinal cut was made in the proximal portion of the sternum, allowing visualization of the aortic arch. The transverse aortic arch was ligated between the innominate and left common carotid arteries with an overlaying 27-gauge needle. The needle was immediately removed, leaving a discrete region of constriction. Animal sham group underwent a similar procedure without ligation.
Only animals with fractional shortening o 50% were determined as being in HF and suitable for further study.
Echocardiography. Mice were anaesthetized with intraperitoneal ketamine (100 mg g À 1 ) for echocardiographic analysis. Transthoracic echocardiography was performed using a Vivid 7,000 equipped (GE Healthcare, Salt Lake City, UT, USA) with a H13L transducer (14 MHz). Two-dimensional and M-mode images were obtained in the short-axis view. The HR, LV end-diastolic internal diameter and LV end-systolic internal diameter were measured over the course of at least three repeated cardiac cycles. The fractional shortening and ejection fraction were then calculated.
Invasive P-V loop measurement of cardiac function. One day after echocardiography, we performed in-vivo haemodynamics. The mice were anaesthetized with an intraperitoneal injection mixture of urethane (1 mg g À 1 ), etomidate (10 mg g À 1 ) and morphine (1 mg g À 1 ), and were then intubated via a tracheotomy and mechanically ventilated at 7 ml g À 1 tidal volume and 120 respirations per minute. A small incision was made in the apex and a French Millar cathether connected a high-fidelity pressure transducer (Scisense, London, ON, Canada) inserted into the LV. Haemodynamic recordings were performed after 5B10 min of stable HR. The rate of pressure rise ( þ dP/dt) and maximal pressure, the time constant of isovolumic relaxation and HR were recorded and analysed. To determine the relative effects of the admittance and conductance techniques to contractility change induced by N106 compound infusion, whole haemodynamic function was recorded at both baseline and after administration of N106 compound through the external jugular vein. N106 compound was solubilized in 10% DMSO (Sigma-Aldrich), 10% Tween-80 (Sigma-Aldrich) and saline for in-vivo experiments. N106 or vehicle solutions were continuously infused through the external jugular vein at a rate of 50 ml min À 1 for 2 min in each dose. Whole haemodynamic functions of increasing dose of N106 of each mouse were recorded for B4 h. Each dose of N106 solution was infused after back to baseline of heart function.