Flumazenil Mimics whereas Midazolam Abolishes Ischemic Preconditioning in a Rabbit Heart Model of Ischemia–Reperfusion

All experiments were conducted following the approval of the Institutional Committee for Animal Care and Laboratory Use (Hadassah-Hebrew University Medical Center, Jerusalem, Israel). The investigation conforms with the Guide for the Care and Use of Laboratory Animals .13 

New Zealand white rabbits weighing between 2.5 and 3.5 kg were initially premedicated with an intramuscular injection of a ketamine (50 mg/ml)–xylazine (10 mg/ml) solution at a volume of 0.6 ml/kg body weight. They were then anesthetized with intravenous pentobarbital (30 mg/kg), administered via  a 20-gauge intravenous cannula into a marginal ear vein. Hetastarch, 5 ml · kg⁻¹· h⁻¹, was infused continuously via  the intravenous cannula. Anesthesia was maintained during the experiment by pentobarbital supplements as needed (according to pedal and palpebral reflexes). Midazolam and flumazenil were used according to the study protocol. Neuromuscular blockers were not administered in order to assess anesthetic depth. The neck was opened with a ventral midline incision, and a tracheostomy was performed. The rabbits’ lungs were mechanically ventilated with positive-pressure ventilation and an inspiratory fraction of oxygen of 1.0. The ventilation rate was 30–35 breaths/min, and tidal volume was approximately 15 ml/kg. The respiratory rate was adjusted to keep the arterial pH between 7.35 and 7.45. End-expiratory carbon dioxide tension was monitored continuously. A 22-gauge catheter filled with heparinized saline was placed in a carotid artery for blood pressure monitoring and blood sampling. Core body temperature was measured via  a rectal temperature probe and maintained at 38.5°± 0.3° (normothermic for rabbits) with radiant heat and a warming blanket. Needle electrodes were inserted subcutaneously in a lead II configuration to enable recording of an electrocardiogram to determine heart rate and help confirm the occurrence of ischemia (ST-segment elevation) and reperfusion of the myocardium distal to the coronary occlusion. A left thoracotomy was performed in the fourth intercostal space and a 4-0 silk suture was passed around a prominent branch of the left coronary artery (approximately halfway between the apex and the base) and threaded through a small vinyl tube to form a snare. Coronary artery occlusion was achieved by tightening the snare around the coronary artery. Regional epicardial cyanosis and ST-segment elevation in the electrocardiogram confirmed myocardial ischemia. Reperfusion was achieved by releasing the snare and was confirmed by visual observation of reactive hyperemia. Ventricular fibrillation, if it occurred, was reversed using direct mechanical stimulation: An index finger was flicked directly against the right ventricle side of the fibrillating heart one to three times to achieve defibrillation. Failure to convert to an organized rhythm after three attempts was defined as intractable fibrillation.

Before a 30-min stabilization period, all animals were given salicylate (100 mg/kg) intravenously. Salicylate is a highly effective hydroxyl free radical scavenger that, upon scavenging ^·OH, forms 2,3- and 2,5- dihydroxybenzoic acid (DHBA) by hydroxylation. Although the 2,3-DHBA is a direct product of the reaction of hydroxyl radicals with salicylate, the formation of 2,5-DHBA from salicylate occurs by two independent routes: direct trapping of ^·OH and biotransformation mediated by the cytochrome P-450 system.14All animals underwent 40 min of regional ischemia followed by 2 h of reperfusion. Preconditioning was elicited by 5 min of coronary occlusion followed by 10 min of reperfusion, beginning 15 min before the period of prolonged coronary occlusion. Rabbits were randomly assigned to the following groups (fig. 1): control group (ischemia and reperfusion without further intervention) (C, n = 10); ischemic preconditioning group (IP, n = 10); flumazenil group (F, n = 10), midazolam group (M, n = 10), or midazolam plus ischemic preconditioning group (M + IP, n = 10). Flumazenil (0.05 mg/kg) or midazolam (0.05 mg/kg) (these doses were chosen because they correspond to therapeutic doses) were administered intravenously 15 min before the 40-min ischemia. In the M + IP group, midazolam (0.05 mg/kg) was given 15 min before the short ischemia (i.e. , 30 min before the 40-min ischemia). An additional group was added to determine whether the benzodiazepine receptor blocking effect of flumazenil is involved in the mechanism of flumazenil-induced preconditioning: flumazenil plus midazolam group (F + M, n = 10). Flumazenil (0.05 mg/kg) was given 15 min before the administration of midazolam (0.05 mg/kg) and the beginning of the 40-min ischemia. To determine where the benzodiazepine receptors influence the preconditioning cascade, 5-hydroxydecanoate (5-HD, 5 mg/kg), a selective mitochondrial K^ATPchannel antagonist, was administered with flumazenil: F + 5-HD group (n = 10). To rule out the possibility that 5-HD itself may influence the generation of ROS during reperfusion, another control group in which 5-HD was administered before ischemia and reperfusion was also included (C + 5-HD, n = 6). In a third control group (n = 6), 5-HD was given to sham-operated rabbits to rule out any possible direct effect of this agent on the baseline levels and production of ROS in nonischemic hearts (data not shown). Arterial pressure, heart rate, and temperature were recorded continuously. Blood samples for hydroxyl free radical measurements were obtained as follows: at the beginning of the experiment (baseline); at 20 and 40 min of occlusion; every 2 min in the first 10 min of reperfusion; and thereafter at 20, 30, 60, 90, and 120 min of reperfusion. In addition, in group M + IP, blood samples for salicylate and hydroxyl radicals were also collected at the end of the first coronary occlusion and after 5 and 10 min during the first reperfusion period.

We quantified the level of ^·OH-mediated conversion of salicylate to its dihydroxybenzoate derivatives by high-performance liquid chromatography coupled with electrochemical detection.15The yield of DHBA derivatives depends on both the flux of the generated hydroxyl radicals and the concentration of salicylate. The values of DHBA derivatives were normalized to account for the difference in salicylate concentration at different time points. Variations in salicylate concentration are attributed to its metabolism and renal clearance. The concentrations of hydroxyl radicals generated in the heart are expressed as the ratio between DHBA (ng) and salicylate (μg).

At the end of the experimental protocol, hearts were excised, mounted on a Langendorff apparatus, and perfused with physiologically buffered saline at 100 cm H²O for 1 min to wash out intravascular blood. The coronary artery was reoccluded, and 0.2% methylene blue was infused into the aortic root to label the normally perfused zone with deep blue color, thereby delineating the risk zone as a nonstained area. The hearts were then removed from the Langendorff apparatus, trimmed of atria and great vessels, weighed, and frozen (in a cold chamber at a temperature of −18°). Hearts were then cut into 2-mm transverse slices. The slices were incubated in 1% 2,3,5-triphenyl tetrazolium chloride in pH 7.4 buffer for 20 min at 37°. The slices were then placed in 10% neutral buffered formalin for 10 min to increase the contrast between stained and nonstained tissue. Because triphenyl tetrazolium chloride stains viable tissue a deep red color, nonstained tissue was presumed to be infarcted. Slices were then photographed, and risk and infarct areas in each slice were measured by computed planimetry. The mass-weighted average of the ratio of infarct area to the area at risk of the ventricle from each slice was determined (percent infarction).

Dihydroxybenzoic acid levels were identified and measured by high-performance liquid chromatography coupled with electrochemical detection using a Varian 5000 liquid chromatograph (Varian Medical Systems, Palo Alto, CA), equipped with a Rheodyne 7125 sample injector (20-μl loop; Rheodyne LLC, Rohnert Park, CA). The column used for separation of salicylate and DHBA was a 25 cm × 4 mm Li Chrospher 100 RP-18, 5 μm (E-Merck, Darmstadt, Germany). The mobile phase contained 0.0 3m citric acid, 0.0 3m acetic acid, 0.2 g/l sodium azide, and 2× methanol. The mobile phase was titrated with solid NaOH to pH 3, followed by titration with CH³COONa to a final pH of 3.6. The flow rate was 1 ml/min. The system was equipped with two detectors in series. Salicylate was identified and measured fluorometrically using a FD-300 model fluorescence detector (Spectrovision, Chelmsford, MA) using excitation and emission wave lengths of 300 and 412 nm, respectively. DHBA derivatives were quantified using an electrochemical amperometric detector (model 4A; Bioanalytical Systems, West Lafayette, IN), with a plastic cell equipped with a glass carbon electrode operated at +0.80 V, using an Ag–AgCl reference electrode. The signals from the detector were acquired on an EZChrome data acquisition and handling system (EZChrome Elite; Scientific Software, San Ramon, CA) and subsequently processed.

Hemodynamic data over time within each group were analyzed using analysis of variance with repeated measures on one factor. Differences in hemodynamics between groups were analyzed using analysis of variance with the Tukey post hoc  test. Incidence of ventricular fibrillation was analyzed with a Kruskal-Wallis test. Intergroup comparisons for infarct sizes were made with one-way analysis of variance, and group differences were detected with Tukey post hoc  test. Differences between the study groups in DHBA levels were assessed by analysis of variance with the Tukey post hoc  testing. Statistical calculations were performed using SPSS 10.0 for Windows software (SPSS Inc., Chicago, IL). Data are expressed as mean ± SEM and significance was assumed for P < 0.05.

A total of 82 rabbits were studied. The results from 71 animals contributed to the final data set: 8 control, 9 ischemic preconditioning, 9 flumazenil, 8 midazolam, 9 midazolam plus ischemic preconditioning, 9 flumazenil plus midazolam, 9 flumazenil plus 5-HD, and 5 control plus 5-HD–treated rabbits and 6 sham-operated animals. The remaining 11 animals were excluded for technical reasons: accidental release of the snare during the ischemia period (n = 3) or intractable ventricular fibrillation (n = 8). The incidence of refractory ventricular fibrillation was not significantly different among groups (2/10 C, 1/10 IP, 0/10 F, 1/10 M, 1/10 M + IP, 1/10 F + M, 1/10 F + 5-HD, 1/6 C + 5-HD, 0/6 sham-operated animals; P = 0.35 with Kruskal-Wallis test).

There was a trend for mean arterial pressure to decrease over time within all groups. Heart rate was similar in all groups, as was the rate–pressure product; therefore, differences in myocardial infarct size between groups could not be attributed to decrease in blood pressure or differences in heart rates. Hemodynamic data are shown in table 1.

The area at risk–to–left ventricular mass ratio did not differ significantly among the groups (51 ± 2% in group C, 51 ± 1% in group IP, 52 ± 1% in group F, 50 ± 3% in group M, 51 ± 2% in group M + IP, 50 ± 3% in group F + M, 52 ± 2% in group F + 5-HD, and 51 ± 1% in group C + 5-HD). These data suggest that changes in the infarct sizes observed in the different experimental groups cannot be related to the percentage of the left ventricular myocardium that was occluded. In group C, the measured infarct size was 57 ± 6% of the area at risk. Ischemic preconditioning and flumazenil resulted in much smaller infarcts, averaging 22 ± 5 and 23 ± 4%, respectively, of the risk zone (P < 0.01). However, the infarct sizes in the M and M + IP groups were similar to that of group C: 53 ± 3 and 49 ± 6%, respectively (P = not significant [NS]), indicating that benzodiazepine receptor activation abolishes ischemic preconditioning. Moreover, when midazolam was added 15 min after flumazenil administration, the cardioprotective effect of flumazenil was abolished (infarct size 51 ± 2%, similar to control). The addition of the K^ATPchannel blocker 5-HD to flumazenil inhibited the beneficial effect of flumazenil (infarct size of 51 ± 4%, similar to control), suggesting that K^ATPchannel opening is downstream to benzodiazepine receptor blockade in the mechanism of preconditioning (fig. 2). The measured infarct size in group C + 5-HD was 54 ± 3% of the area at risk (NS compared with group C), thus indicating that 5-HD itself has no direct influence on infarct size (fig. 2).

In the IP and M + IP groups, the level of 2,3 and 2,5-DHBA did not increase significantly after 5 min of index ischemia during the preconditioning period (12% and 9%, respectively, compared with baseline values in group IP [P = NS], and 9% and 8% in group M + IP, respectively [P = NS]). After 5 min of reperfusion following the preconditioning stimulus, there was an increase of 29% (P < 0.05) above baseline in 2,3-DHBA. A further 36% increase (P < 0.05) above baseline levels was observed after 10 min, at the end of the first reperfusion period. The level of 2,5-DHBA increased by 26% compared with baseline values (P < 0.05) at 5 min of reperfusion and increased further to 32% (P < 0.05) above baseline values after 10 min. Administration of midazolam attenuated the increase in hydroxyl radical levels in group M + IP during the preconditioning period: The measured normalized values of 2,3-DHBA were 13% (P = NS) and 16% (P = NS) above baseline levels after 5 and 10 min of the first reperfusion period, respectively. Furthermore, as was observed with 2,3-DHBA, midazolam attenuated the increase in the concentration of 2,5-DHBA during preconditioning: The measured value of 2,5-DHBA was only 12% (P = NS) above baseline levels at 5 min of reperfusion and 14% (P = NS) above the baseline values at 10 min.

An acute increase of 50% in normalized 2,3-DHBA compared with baseline values (P < 0.05) was already observed in the control group after 2 min of reperfusion. After 10 min of reperfusion, the peak value was measured (81% increase compared with baseline levels; P < 0.01; fig. 3). After 10 min of reperfusion, there was only a 24% and 31% increase in the concentration of 2,3-DHBA in the IP and F groups, respectively (P < 0.01 compared with control group). However, the administration of midazolam (groups M and M + IP) did not produce this effect: The hydroxyl radical production increased to a degree similar to that seen in the control group (increase of 72% and 68%; P = NS compared with group C). The addition of 5-HD to flumazenil resulted in an increase of 68% of the normalized 2,3-DHBA similar to that of the control group. In group C + 5-HD, there was an increase of 87% in the concentration of the 2,3 derivative (P = NS compared with group C), thus ruling out any possible effect of 5-HD itself on the generation of hydroxyl radicals. Maximal production of 2,5-DHBA occurred in the control group after 4 min of reperfusion: a 75% increase compared with baseline values (P < 0.01; fig. 4). Ischemic preconditioning and flumazenil, however, significantly attenuated this increase to only 24% and 33%, respectively, above baseline values (P < 0.05 compared with group C). The increases in 2,5-DHBA in the M (76%), M + IP (69%), and F + 5-HD (81%) groups were comparable with that of the control group (P = NS). As with 2,3-DHBA, 5-HD itself had no effect on the production of 2,5-DHBA. After 4 min of reperfusion, there was an increase of 77% in the level of this derivative compared with baseline levels. This increase was similar to that of the control group (P = NS).

At all time points during reperfusion (up to 30 min), there was a significant difference between the C, M, M + IP, and F + 5-HD groups on one hand, and the IP and F groups on the other hand (figs. 3 and 4).

At 60 min of reperfusion and afterward, there was a return to baseline values in all groups.

In the sham-operated animals, there was no significant difference in the concentration of the DHBA derivatives compared with baseline levels in all time points (data not shown).

In this in vivo  rabbit model of ischemia and reperfusion, flumazenil reduced infarct size and decreased postischemic production of hydroxyl radicals to a degree similar to that produced by ischemic preconditioning. These data thus indicate that flumazenil mimics ischemic preconditioning. Alternately, the benzodiazepine midazolam had the opposite effect; namely, it abolished ischemic preconditioning. Therefore, directly or indirectly, benzodiazepine receptors influence ischemic preconditioning. These findings confirm and extend the previous in vitro  report by Zhang and Yao,6who found that flumazenil reduced cell death in chick cardiomyocytes, despite its discontinuation 10 min before 60-min ischemia. In the current in vivo  study, flumazenil administered before the ischemic period elicited similar cardioprotective effects. The mechanisms by which flumazenil mediates these actions in cardiomyocytes seem to be identical to those seen with classic ischemic preconditioning, where several parallel endogenous signaling pathways seem to be involved, including adenosine, acetylcholine, K^ATPchannels, ROS, protein kinase C, and others.16Yao et al.  17showed in cultured cardiomyocytes that flumazenil induced cardioprotection by generating ROS. An increase in ROS signals activated protein kinase C and subsequently opened mitochondrial K^ATPchannels. It has yet to be demonstrated that the same mechanism is involved in the cardioprotection seen in our in vivo  model.

In contrast to the beneficial effects of ROS and metabolites in triggering ischemic or pharmacologic preconditioning, they have been implicated at being involved in the postischemic reperfusion injury of the heart.10ROS cause lipid peroxidation that disrupts membrane function permitting calcium overload to occur.11ROS can induce structural alterations in both vascular cells and cardiac myocytes, leading to a decrease in heart contractility and a marked increase in coronary vascular resistance. The current study suggests that a simple procedure, such as flumazenil administration, offers another method to protect the heart by attenuating free radical production during reperfusion and also reducing cell death.

Benzodiazepine effects are mediated primarily via  the central benzodiazepine receptors located in the central nervous system, but they also bind to other receptors, called peripheral benzodiazepine receptors, located mainly in peripheral tissues such as the heart.18Leducq et al.  9have shown that a specific peripheral benzodiazepine receptor antagonist, SSR180575, protected the heart against ischemia–reperfusion injury. Although flumazenil is not a selective peripheral benzodiazepine receptor ligand, but rather a potent antagonist of the central benzodiazepine receptors,19we were able to show that it had significant protective cardiac effects. In the current study, however, midazolam, a benzodiazepine receptor agonist, neither limited infarct size nor decreased ROS production during reperfusion after ischemia. When administered before ischemic preconditioning or flumazenil, it inhibited their beneficial effects, suggesting that benzodiazepine receptors may be implicated in their mechanism of cardioprotection. It is of interest that benzodiazepine receptors have been identified in mitochondria.9,20During ischemic preconditioning, minute ROS production in the mitochondria occurs, leading to the opening of mitochondrial K^ATPchannels and subsequent activation of protein kinases. Furthermore, we showed that selective blockade of the mitochondrial K^ATPchannels by 5-HD during flumazenil administration abolished the reductions in infarct size and hydroxyl free radical production seen with flumazenil. Therefore, K^ATPchannel opening is downstream to benzodiazepine receptor inhibition in the flumazenil protective cascade.

This study may have implications in the clinical practice because midazolam is a widely used benzodiazepine in cardiac patients, especially in coronary artery bypass surgery patients. It is of interest that, in the clinical setting, i.e. , coronary surgery, midazolam has been associated with a decreased preservation of early postoperative myocardial function (as demonstrated by poorer hemodynamics, higher concentrations of troponin I, and need for inotropic support) compared with the volatile anesthetics sevoflurane or desflurane.21Alternately, flumazenil, a benzodiazepine antagonist, has been shown to produce minimal coronary and left ventricular hemodynamic responses in patients with coronary artery disease after reversal of flunitrazepam-induced sedation.22Therefore, flumazenil-induced cardioprotection may have a role in clinical situations where the deleterious effects of ischemia–reperfusion are expected, specifically in high-risk patients in whom an ischemic type of preconditioning may further jeopardize diseased myocardium.