**4. Discussion**

Our study has several key findings: In line with prior reports of cardiovascular side effects of TZDs [45,46], it confirms an increase in cardiac IR injury following the acute administration and

washout of the TZD rosiglitazone in rat isolated hearts. A novel finding in this context however, is that in further dose-response experiments we found a considerable, rosiglitazone-induced, fully reversible increase in mitochondrial oxidation as assessed by decreased NADH and increased FAD autofluorescence. This finding was independent of PPARγ activation as it was not abolished by the PPARγ antagonist GW9662. Moreover, all the observed rosiglitazone e ffects on the mitochondrial redox state were replicated with another member of the TZD family, pioglitazone, which suggests a group- rather than a mere single drug-dependent side e ffect on mitochondrial function.

#### *4.1. PPAR*γ *Activation and Myocardial Protection: Friend or Foe?*

Several options are available to investigate the specific aspects of a particular signaling pathway. For example, the expression, modification and/or activity of a certain protein/enzyme can be measured. An agonist can be used to activate a certain pathway. Or a specific antagonist can be used to attenuate or abolish a certain finding. Using the latter, we have previously shown in a consomic rat model of resistance against myocardial IR that the specific PPARγ antagonist GW9662 prevented endogenous and exogenous cardioprotection [11]. Thus, PPARγ activation is a critical part of cardioprotective pathways that can be initiated by di fferent triggers upstream of PPARγ [12].

Conversely, administration of specific PPARγ agonists should be able to mimic the above phenotype and activate a cardioprotective pathway directly without the need for, e.g., ischemic or anesthetic preconditioning, and their acutely cardio-depressant and other side e ffects [12]. While numerous experiments with the PPARγ antagonist GW9662 have largely shown myocardial protection against IR injury by PPARγ activation [2,12,47–51], experiments with agonists like TZDs have revealed less clear results, particularly under pathological conditions [52]. To the contrary, but in line with our findings, the TZD-triggered aggravation of myocardial outcome was shown in animal studies [13–17] and in humans [18,19]. The apparently contradictory findings among some of these studies may be due to di fferences in the species and experimental models being used, the protocol and duration of TZD administration, co-morbidities, and/or related to the used dosage.

#### *4.2. Specificity of TZDs for PPAR*γ *Activation*

Conclusions from experiments using agonists and/or antagonists generally rely on their respective specificity for any given pathway or receptor. Thus, it is possible that TZDs have PPARγ-independent binding sites [53] that can influence cardiovascular outcome directly or indirectly [54–56]. Reports about the deleterious e ffects of TZDs without their prevention by the use of a specific PPARγ-antagonist [13,15–17] may, therefore, be due to PPARγ-independent side e ffects as previously discussed by Feinstein and colleagues [54]. Moreover, these reports pose the question of whether those side e ffects are specific to individual members of the TZD group or side e ffects of the TZD group as a whole because of their chemical similarities [53]. The use of more than one member of the TZD group within the same study and demonstration of failure to abolish the observed e ffect with GW9662 can add clarification in these regards.

#### *4.3. TZDs A*ff*ect Mitochondrial Function*

Our knowledge about the cardiac mitochondrial side e ffects of TZDs is sketchy [54]. An in-vivo and in-vitro study in mice [57] reported rosiglitazone to cause dysfunction of cardiac mitochondria as evidenced by decreased mitochondrial respiration and substrate oxidation, as well as decreased complex I and IV activities. Rosiglitazone also increased superoxide production from complexes I and III. Neither genetic PPARγ deletion nor the PPARγ antagonist GW9662 prevented these e ffects. Moreover, these findings were associated with decreased ATP synthesis and increased cardiac dysfunction during rosiglitazone administration. The authors emphasize that, similarly to our findings, these were dose-dependent e ffects found at concentrations of 10 μM and higher.

Reactive oxygen species (ROS) are produced from di fferent intracellular sources with mitochondria and di fferent mitochondrial electron transport chain complexes being the major sources in cardiomyocytes [58]. While inhibition of the mitochondrial electron transport chain at specific sites of complex I and/or III [58] can lead to oxidative stress, the latter is not necessarily proof of a blockade but could also be caused by an increase in electron transport, as is the case with uncoupling [59]. Indeed, mitochondrial uncoupling makes oxidative phosphorylation less efficient, increases ROS, and—if not countered by increased delivery of the reducing equivalents, NADH and FADH2, through the Krebs cycle—can lead to a decreased membrane potential and ATP synthesis. Thus, a decrease in membrane potential or ATP synthesis or an increase in ROS production cannot distinguish between the uncoupling and blockade of mitochondrial electron transport, unless the mitochondrial redox state is measured. In this context, our results sugges<sup>t</sup> a dose-dependent and reversible mitochondrial oxidation by two different TZDs independent of PPARγ activation. Excessive mitochondrial oxidation can lead to decreased ATP synthesis and excessive ROS production, all of which can damage the myocardium and/or lead to increased sensitivity to subsequent IR.

#### *4.4. Study Limitations and Summary*

This study needs to be interpreted within its natural constraints. We used one species and an acute rather than chronic experimental IR injury model. Mitochondrial function was assessed by two different redox state measurements in intact hearts, but not by mitochondrial oxygen consumption in isolated mitochondria, and we did not assess ROS production, membrane potential or ATP synthesis, neither of which would have added to differentiating severe blockade of mitochondrial electron transport from uncoupling. The IR injury assessment was limited to one TZD and one dose given twice acutely and washed out before IR. On the other hand, rosiglitazone's effects on the redox state, with or without the PPARγ antagonist GW9662, were closely mimicked by pioglitazone; the NADH results mirrored the FAD measurements; and the GW9662 results were nearly identical to the ones in its absence, serving as internal controls. Although we have not conducted a formal dose-response study with the PPARγ antagonist GW9662 or Western blot analysis, we have chosen a dose (10 μM) commonly used to block PPARγ in the isolated heart [11] and isolated cardiomyocyte studies [41,42]. Demonstration of a change in infarct size remains the gold standard for in- and ex-vivo studies on cardioprotective or -toxic agents and strategies, even in the absence of significant functional changes. Within limits, compensatory mechanisms, such as e.g., mildly increased intracellular calcium in surviving cardiomyocytes, can make up for the loss of function in the infarcted myocardium. The infarct size is among the more sensitive parameters with earlier responses to IR than functional changes [36]. We exposed hearts from non-diabetic animals to an acute dose of rosiglitazone; chronic exposure in diabetic individuals would require the study of chronically diabetic and, thus, hyperglycemic hearts in the control group vs. normoglycemic hearts chronically exposed to a TZD which would complement but not replace the findings in the present study and add additional confounders to the study.

In summary, our study in rat isolated hearts suggests that the off-target effects of the TZDs, rosiglitazone and pioglitazone, include a significant degree of mitochondrial oxidation associated with aggravated myocardial IR injury that can help explain the reported increase in adverse cardiac events. Because of the large number of diabetic patients worldwide who are chronically treated with TZDs, it is important to unravel the mechanisms of these adverse effects and their clinical consequence in future studies, including diabetic IR models, in order to improve overall patient outcome while minimizing unwanted side effects.

**Author Contributions:** Study design by M.L.R., D.W., D.F.S. and A.K.S.C.; experiments conducted by M.L.R. and R.E.; data analysis and presentation by M.L.R. and R.E.; manuscript written by M.L.R. and A.K.S.C.; manuscript revised and approved by M.L.R., R.E., D.W., D.F.S. and A.K.S.C.; agreemen<sup>t</sup> to be accountable for all aspects of the work by M.L.R., R.E., D.W., D.F.S. and A.K.S.C. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was supported in part by the US Department of Veterans Affairs Biomedical Laboratory R&D Service (IK2BX001278 and I01 BX003482), the National Institutes of Health (5R01 HL123227), a Roizen Anesthesia Research Foundation New Investigator Grant from the Society of Cardiovascular Anesthesiologists, and by institutional funds to MLR.

**Acknowledgments:** The authors wish to thank Qunli Cheng, James S. Heisner (Department of Anesthesiology, Medical College of Wisconsin, Milwaukee, WI), Sushrut V. Shidham and Darren S. Nabor (formerly Medical College of Wisconsin Affiliated Hospitals, Milwaukee, WI) and William J. Cleveland (Department of Anesthesiology, Vanderbilt University Medical Center, Nashville, TN) for their valuable contributions to this study.

**Conflicts of Interest:** The authors declare no conflict of interest.
