Attenuation of doxorubicin-induced cardiotoxicity in a human in vitro cardiac model by the induction of the NRF2 pathway

Dose-dependent cardiotoxicity is the leading adverse reaction seen in cancer patients treated with doxorubicin. Currently, dexrazoxane is the only approved drug that can partially protect against this toxicity in patients, however, its administration is restricted to those patients receiving a high cumulative dose of anthracyclines. Investigations into the mechanisms of cardiotoxicity and efforts to improve cardioprotective strategies have been hindered by the limited availability of a phenotypically relevant in vitro adult human cardiac model system. Here, we adapted a readily reproducible, functional 3D human multi-cell type cardiac system to emulate patient responses seen with doxorubicin and dexrazoxane. We show that administration of two NRF2 gene inducers namely the semi-synthetic triterpenoid Bardoxolone methyl, and the isothiocyanate sulfurophane, result in cardioprotection against doxorubicin toxicity comparable to dexrazoxane as evidenced by an increase in cell viability and a decrease in the production of reactive oxygen species. We further show a synergistic attenuation of cardiotoxicity when the NRF2 inducers and dexrazoxane are used in tandem. Taken together, our data indicate that the 3D spheroid is a suitable model to investigate drug induced cardiotoxicity and we reveal an essential role of the NRF2 pathway in cardioprotection providing a novel pharmacological mechanism and intervention route towards the alleviation of doxorubicin-induced toxicity.


Introduction
The anthracycline doxorubicin, although irst introduced in the 1960s, continues to be a commonly used and highly efective chemotherapeutic drug used to treat both adults and children for a wide range of cancers including Hodgkin's and non-Hodgkin's lymphoma, acute myeloid leukemia, breast and ovarian cancers. However, the dose-dependent cardiotoxic efects of doxorubicin can limit patient exposure and therapeutic eicacy. Even at relative low cumulative doses (200-250 mg/m 2 ) an increase of 8% of adverse cardiac events has been observed [1], further increasing to 26% with cumulative doses of 550-600 mg/m [2]. In fact, after secondary malignancies the leading cause of mortality in cancer patient survivors is cardiovascular disease (CVD), with an increased eight-fold risk when compared to the general population [3,4]. Side efects of doxorubicin include left ventricular wall stress leading to reduction in the left ventricular ejection fraction (LVEF), ventricular and atrial arrhythmia, chronic cardiomyopathy and congestive heart failure.
Multiple mechanisms have been implicated in both the antineoplastic and cardiotoxic action of doxorubicin, although both remain incompletely understood [1,5]. The antineoplastic mechanisms include the stabilisation and intercalation of doxorubicin with the topoisomerase II-α (TOP2 A)-DNA cleavage complex, which in turn leads to the prevention of DNA break repair, as well as the inhibition of DNA replication in rapidly proliferating cancer cells leading to cancer cell death [6]. Further, doxorubicin has been implicated with the generation of free radicals that cause damage to DNA, cellular proteins and mitochondria. Presently, mechanisms of cardiotoxicity include myocardium injury through the generation of reactive oxygen species (ROS) leading to subsequent membrane damage, impaired mitochondrial function, upregulation of proteins that mediate apoptosis and disruption of calcium (Ca 2+ ) homeostasis causing Ca 2+  T sarcomeric disarray and myoibril deterioration [1,6,7]. This disruption of functioning sarcomeres highlights a potential mechanism for the loss of contractile force that is symptomatic of patients with doxorubicinassociated cardiotoxicity [8][9][10]. Experimental animal model systems have been and will continue to be instrumental in providing important insights into the molecular basis of cardiotoxicity and CVD. However, it is widely recognised that the physiology of human cardiac cells difers from those of rodents; the most commonly used in vivo experimental model [11,12]. Therefore, research into mechanisms of CVD would be strengthened signiicantly by the availability of in vitro models based on human cells. Advances in the ability to generate cardiomyocytes from human embryonic stem cells (hESC-CM) as well as the reprogramming from induced pluripotent stem cells (iPS-CM) [1] have started to alleviate this hurdle and provide researchers with tools to study CVD and test new drugs. However, to date, the stem cells used in CVD research remain in an immature state and are therefore not relective of the phenotype of adult cardiomyocytes [1]. Indeed, this cellular immaturity was recently exempliied in an in vitro study in which dexrazoxane, the only drug approved by the FDA for protection against doxorubicin induced cardiotoxicity, was found to exacerbate the condition in iPS-CM [1,13].
Oxidative stress resulting from an increase in the generation of ROS is one of the most widely cited mechanisms regarding the molecular pathway underlying anthracycline induced cardiotoxicity. However, the administration of generic anti-oxidants has been shown to result in no alleviation of the condition [14,15]. The activation of the transcription factor NF-E2-related factor 2 (NRF2) has been shown to antagonise oxidative stress via the upregulation of multiple cytoprotective genes [16,17]. This has stimulated the development of several pharmacological activators of NRF2, including semi-synthetic triterpenoids which have been shown to ofer protection against ROS both in vivo [18][19][20] and in vitro [21,22], and have subsequently entered clinical trials.
Here, we show in vivo like responses in a readily constructible 3D microtissue spheroid in which human cardiomyocytes are grown in combination with human cardiac ibroblasts and endothelial cells. Consistent with this, the spheroids show a concentration-dependent toxicity response to doxorubicin that is reduced in the presence of dexrazoxane. Making use of this mechanistically-relevant model, we show that Bardoxolone methyl (BARD, also known as 2-cyano-3,12dioxooleana-1,9(11)-dien-28-oic acid methyl ester (CDDO-me)) as well as Sulforaphane, an isothiocyanate from the cruciferous vegetable family, both known NRF2 activators [16,23] can protect against doxorubicin induced cardiotoxicity to a similar extent as dexrazoxane but, have a pronounced synergistic efect when co-administered in the presence of dexrazoxane.

Formation of 3D cardiac spheroids
We formed monoculture 3D cardiac microtissue spheroids comprised of human cardiomyocytes alone and a tri-culture spheroid comprised of human cardiomyocytes, ibroblasts and endothelial cells. Two week-old monoculture cardiomyocyte formed loosely associated spheroids that showed no sign of striation upon immunoluorescentstaining using the cardiomyocyte speciic marker α-actinin (Fig. 1A). In contrast, the tri-culture formed tightly associated spheroids as early as day 5 that displayed limited striations in addition to the presence of endothelial cells visualised using CD31 (endothelial cell marker) and extensive ibroblast-generated type I collagen extracellular matrix (Fig. 1B). All markers were retained at 2 weeks in culture (Fig. 1C). Both monoculture and tri-culture spheroids initiated beating after two weeks of culture which coincided with the co-localization of immunopositive staining for HCN4 and α-actinin, showing localization of pacemaker cells at this time point compared to a more dispersed staining at 5 days (Supplemental Fig. 1). The beating rate of monoculture spheroids was signiicantly lower (46.8 ± 0.86 bpm) when compared to the tri-culture spheroids (74.8 ± 2.3 bpm; p = 0.00002, student's ttest) (Fig. 1D), the latter being more emulative of the beating rate range of a resting human adult heart (60-100 bpm) [24,25].

Oxygen profiles of spheroids
To ensure that the dimensions of spheroid system permitted an adequate oxygen supply throughout the system we carried out in silico predictions of oxygen (O 2 ) proiles. Spheroids of radii 88 μm (the average radius of spheroids in this study, blue curve/dot in Fig. 2A and B) have only a minimal decrease in O 2 with 21% O 2 at the outer surface decreasing to 16.3% O 2 in the core. The core O 2 % concentration then decreases almost linearly with a further 1% O 2 reduction at the core for each 7.4 μm increase in the spheroid radius. Moreover, hypoxia (O 2 less than 1%) was only predicted in spheroids of radii greater than 190 μm (red curve/dot in Fig. 2A and B and Cii) ensuring that the size of the spheroids used in this study (Fig. 2Ci) have a well-maintained oxygen supply.

Cellular morphology of 3D microtissues
Since the tri-culture system showed striations and beating rates similar to adult human cardiomyocytes, this model was examined further to conirm cellular morphology. Electron microscopy (EM) images of the two-week spheroids, when the spheroids irst initiate beating, showed no evidence of organized cardiac tissue (Supplemental Fig. 2). However, spheroids maintained in culture for ive weeks showed extensive α-actininpositive striations by immunoluorescence (Fig. 3A) and electron microscopy (EM) (Fig. 3B). EM images display evidence of ultrastructure with the alignment of Z-disks and formation of I-, H-, A-and M bands, as well as presence of myoibrils (Mf) (Fig. 3B) and numerous mitochondria (Supplemental Fig. 3). To determine spheroid functionality we examined the activation of cardiomyocyte-expressed ryanodine receptor (RyR2) intracellular calcium (Ca 2+ ) channels in response to cafeine. Quantiication of cafeine-induced Ca 2+ transients showed a signiicantly (p = 0.02, students t-test) increased response in the 5 week spheroids when compared to those seen at the earliest stage of contraction in 2 week-old spheroids ( Fig. 3C-E). Blockade of the L-type Ca 2+ channels with verapamil resulted in termination of Ca 2+ transients and further stimulation with cafeine following verapamil blockage caused Ca 2+ release from the sarcoplasmic reticulum by both 2 and 5 week-old spheroids (Fig. 3F, G). In addition, inhibition of Ca 2+ -ATPase (SERCA) ion-transport activity by the addition of thapsigargin led to the cessation of cafeine-induced Ca 2+ transients due to depletion of Ca 2+ from the sarcoplasmic reticulum ( Fig. 3H, I). Taken together these results indicate the presence of a functional sarcoplasmic reticulum in our 3D spheroid system.

Mass spectrometry analysis of cardiac cell development
We next, applied global mass spectrometry, database searching and statistical iltering at the peptide and protein level (< 1% FDR), followed by relative quantiication based on spectral counts on both 5 day and 5 week tri-culture spheroids to conirm the presence of known proteins and to identify previously unannotated proteins diferentially regulated upon cellular development. Analysis of three biological repeats resulted in 2497 human protein identiications with a inal set of 1766 passing selection criteria (< 1% FDR and ≥ 2 peptides per protein identiied) (Supplemental Table 2). Comparative proteomics using Progenesis identiied 1126 diferentially expressed protein clusters (p ≤ 0.05, ANOVA and/or 1.5 ≥ fold change; Fig. 4A). The hallmarks of cardiac development is an increase in structural proteins involved in the adult myocardium, as well as proteins involved in calcium handling and cardiac contraction. We observed a signiicant increase (p < 0.05) in expression of the structural proteins; sarcomeric α-actinin, β-myosin, myosin binding protein C (MYBPC3), as well as an increase in myosin light chain that is predominantly seen in adult ventricular tissue (MLC- Fig. 1. Verification of 3D cardiac spheroids structure A) Immunoluorescence of 2 week mono-culture spheroids show no striations after staining with α-actinin B) immunoluorescence of tri-culture spheroids at 5 day and C) 2 weeks show visible striations which are retained as seen with α-actinin antibody, ibroblast matrix stained with collagen I and endothelial cells with CD31 staining. All immunoluorescence images were counterstained with Hoechst (nuclei), D) Beating rate of spheroids shows signiicant diference between tri-culture and monoculture spheroids (data shown as mean of beating rate ± SE, n = 10 (3 independent runs); p < 0.05, student's t-test). 2v, MLC1SB) [26,27] and adult atrial tissue (MLC-2a) [27,28] in 5 week compared to 5 day spheroids (Fig. 4B). This trend was also seen in calcium handling proteins involved in cardiac contractions; namely Ca 2+− ATPase (SERCA), phospholamban (PLN), Ryanodine receptor (RyR2) and the sodium potassium ATPase (Na/K ATPase) (Fig. 4B). In addition, we observed a signiicant (p = 0.0027, ANOVA) increase in expression of Tmem65, a recently identiied maturation marker of cardiomyocytes [29]. Gene ontology (GO) analysis calculated using DAVID (http://david.abcc.ncifcrf.gov) of the signiicantly up regulated proteins in the 5 week spheroids showed enrichment in processes associated with cardiac development such as ion transport (n = 35, p ≤ 0.05), muscle contraction (n = 29, p ≤ 0.05) and heart development (n = 18, p ≤ 0.05; Fig. 4C). Interestingly, in agreement with the terminally diferentiated nature of mature cardiac cells, the GO term analysis of the signiicantly down-regulated proteins included a decrease in cell motion (n = 31, p ≤ 0.05), cell cycle process (n = 24, p ≤ 0.05) and cell morphogenesis (n = 17, p ≤ 0.05; Fig. 4C). Ingenuity pathway analysis (IPA) identiied cardiovascular system development and function as the top biological functional pathways (n = 72, p ≤ 0.005) with the key nodes and their corresponding proteins shown in Fig. 4D. IPA analysis highlighted pathways involved in the development of cardiac tissue between pre-and post-beating cardiac spheroids. It further highlighted a complex of proteins that were upregulated in cardiac contraction, which included those identiied in Fig. 4D, but extended this data to include other known regulators such as histidine-rich calcium binding protein (HRC), a recently identiied SERCA binding protein [30,31], as well as PRKACA and CAMK2D. Upstream analysis identiied known cardiac transcription factors; MYOCD, GATA4, HAND2, and TBX5 to be activated, inducing the up and down-regulation of proteins involved in cardiac development and contraction [32] (Fig. 4E).
To gain a better understanding of proteins that were signiicantly increased in cardiac development we compared all 416 proteins (≥ 1.5 fold and p ≤ 0.05 in the 5 week spheroid) to publically available human adult heart microarray data [33] and found that 251 proteins were identiied with 3-fold intensity above global median. A comparison to our previously determined chamber speciic proteins identiied 11 atria speciic markers and 30 ventricle speciic markers enriched within this dataset [34]. Finally, in order to determine which proteins have previously been implicated in disease, we searched our developmentally regulated proteins against previous links to cardiac function on the basis of phenotype ontology (http://www.informatics.jax.org). This identiied 75 previously implicated proteins thereby identifying 341 previously phenotypically uncharacterized human cardiac associated proteins (Fig. 4F).

Doxorubicin induced cardiotoxicity
Since 5 week old spheroids showed greater in vivo like morphology, we used them to investigate doxorubicin induced cardiotoxicity. Lightsheet-based immunoluorescence microscopy (Zeiss LightSheet Z.1) showed that doxorubicin penetrated cells throughout the spheroid microtissue and was not restricted to the exposed surface area resulting in a homogeneous reporter system (Fig. 5A). Spheroids displayed a concentration-dependent loss of cell viability in response to doxorubicin, resulting in an IC 50 of 0.35 μM (Fig. 5B). Moreover, pre-treatment of cells with the cardioprotectant dexrazoxane diminished the toxic efect of doxorubicin leading to an increased IC 50 of 6.6 μM (Fig. 5B). We next sought to assess the protective efects of activating NRF2, by pre-treating cells with bardoxolone methyl (from here on known as CDDO-me) prior to treatment with 0.001 mM doxorubicin. Under these conditions there was a signiicant 2-fold (p ≤ 0.05) increase in cell viability (Fig. 5C). Interestingly, when cells were pretreated with both dexrazoxane and CDDO-me a signiicant 5-fold synergetic cardioprotective response (p ≤ 0.005) was seen when compared to the doxorubicin treated cells and also a signiicant (p ≤ 0.05) increase in viability compared to dexrazoxane administered alone (Fig. 5C). In order to conirm a reduction in the oxidative stress inferred by doxorubicin we measured levels of reactive oxygen species (ROS) which revealed a signiicant decrease (p ≤ 0.05) in ROS levels in the presence of dexrazoxane and CDDO-me independently (Fig. 5D) but to a greater extent (p ≤ 0.01) when administered in tandem (Fig. 5D). To further conirm the protective efect of NRF2 induction we preincubated spheroids with an alternative pharmacological activator, sulforaphane, prior to doxorubicin treatment. This consistently showed a signiicant (p ≤ 0.05) increase in cell viability and a signiicant synergetic cardioprotective response (p ≤ 0.005) when incubated together with dexrazoxane (Fig. 5E). No signiicant change of basal levels of ATP (Supplemental Fig. 4A) or ROS (Supplemental Fig. 4B) were seen in the presence of NRF2 stimulants in the absence of a cellular challenge by doxorubicin indicating that the NRF2 inducers modulated ATP only in the presence of cellular distress by doxorubicin. In keeping with previously observed results [1] spheroids in the presence of doxorubicin showed a loss of integrity and increased disarray within the sarcomeric structure of the cardiomyocytes (Fig. 5F). However, in the presence of dexrazoxane and CDDO-me, morphological analysis showed the retention of sarcomeric structures (Fig. 5F), consistent with the observed cardioprotective efects of the drug combination.
In order to directly ascertain that the pre-treatment with both the Fig. 2. Oxygen profile of spheroids A) Graphs show the oxygen proile of spheroids with difering radii with the outer core exposure set at atmospheric 21%. Blue curve represents average radii of spheroids used in this study and red indicates size that would lead to a hypoxic core B) Graph of percentage oxygen distribution at the core of the spheroids blue dot is the percentage in the spheroids used in this study and red in a hypoxic core C) Heat proile of oxygen distribution of size of spheroids i) used in this study ii) predicted to have a hypoxic core (For interpretation of the references to colour in this igure legend, the reader is referred to the web version of this article). drugs did not afect the antineoplastic eicacy of doxorubicin, we administered the corresponding concentrations to AU565 breast cancer cells and showed no signiicant diference in cell viability when compared to the administration of doxorubicin alone (Fig. 5G).

Role of NRF2 in cardioprotection
To conirm that CDDO-me efectively activated NRF2 signalling in cardiac spheroids we determined the expression levels of NRF2 target Fig. 3. Morphological and functional assessment of spheroids A) Immunoluorescence staining at 5 weeks show extensive striations (αactinin, red), B) Electron microscopy images of spheroid ultrastructure showed sarcomeres; Z disks, black arrow; H zones, M lines and I-bands; mitochondria (m) and the presence of myoibril bundles (Mf). Representative traces of Ca 2+ release in response to 10 mM cafeine in C) 2 week spheroids and D) 5 week spheroid. E) Percentage cafeine-induced change of peak luorescence intensity at 2 and 5 weeks (graph indicates mean ± s.e after normalizing the peak luorescence intensity before cafeine administration) * denotes signiicant diference (p = 0.02, n = 3, students t-test). Representative traces of Ca 2+ release in response to 1 mM verapamil before and after administration of cafeine at F) 2 weeks and G) 5 weeks. Representative traces of Ca 2+ release in response to 2 μM Thapsigargin before and after administration of 10 mM cafeine at H) 2 weeks and I) 5 weeks (n = 3). F, luorescence intensity; F0, luorescence intensity at baseline; F/F0, F normalized by F0. (For interpretation of the references to colour in this igure legend, the reader is referred to the web version of this article). genes by quantitative PCR (qPCR). This analysis showed a signiicant synergistic induction of established NRF2 targets NQO1, SRXN1, AKR1C1 and TXNRD1 in the presence of dexrazoxane and CDDO-me (Fig. 6A). To conirm the involvement of the NRF2 pathway in the cardioprotective mechanism we next used ML385, a small molecule inhibitor of NRF2 [35]. Western blot analysis showed a signiicant dose dependent decrease in NRF2 levels at the protein level ( Fig. 6B and  (Fig. 6B). Treatment of cells in the absence of doxorubicin showed no signiicant change in viability of cells upon treated with ML385 alone (Fig. 6C) or in the presence of CDDO-me or Sulforaphane (Supplemental Fig. 6). In the presence of doxorubicin, ML385 incubation of cells pre-treated with either CDDO-me or sulforaphane led to no signiicant protection against doxorubicin induced cardiotoxicity (Fig. 6D).

Discussion
Chemotherapy-induced cardiotoxicity is a serious concern for established as well as newer anticancer agents [36]. The relevance of in vitro human cardiac models for research into cardiac disease and toxicity clearly depends on an ability to emulate the phenotype of in vivo cardiac tissue as closely as possible. Eforts to promote greater physiological relevance in multiple cells types have suggested that 3D systems have improved functionality and cellular morphology when compared to their 2D counterparts [37][38][39][40]. With this in mind, we used a multi-cell type 3D cardiac spheroid model to investigate doxorubicin induced decrease in viability and structural integrity. In keeping with patient observations dexrazoxane treatment showed cardioprotection in this 3D microtissue system leading to an increase in cell viability and the maintenance of structural integrity in the presence of doxorubicin. This clinically-relevant observation, has not previously been reported for monolayer iPSC derived cardiomyocytes with suggestions that this resulted from cellular immaturity [1]. We show NRF2 stimulated cardioprotection to similar levels as dexrazoxane and further showed synergistical cardioprotection against doxorubicin induced toxicity with the co-administration of dexrazoxane and two NRF2 inducers namely CDDO-me and sulforaphane. Used together these drugs showed a greater signiicant increase in cell viability, greater decrease in the amount of ROS generated and restoration of sarcomeric structural integrity when compared to that seen with dexrazoxane alone. This coincided with a signiicant increase in downstream targets of NRF2, which is regarded as a master regulator of defence against the efects of oxidative stress [41]. The cessation of a signiicant cardioprotective efect in the presence of the small molecule NRF2 inhibitor ML385, further emphasized the involvement of NRF2 in doxorubicin-induced cardiotoxicity.
The multi-cell type 3D cardiac system described here is an easily reproducible model of cardiac cell development. In many of the current systems, cardiomyocytes are used in isolation and therefore not subjected to the physiological cross talk and signaling between the different cell types within the human heart. Indeed, it has previously been noted that greater in vitro cardiomyocyte development occurs in the presence of non-cardiomyocyte cells [42,43]. Although the 3D system initiated cellular beating activity and showed the development of striations by 2 weeks, EM analysis showed that ultrastructure development occurred after 5 weeks. The responsive behavior of Ca 2+ handling proteins through stimulation and inhibition by cafeine, thapsiargin and verapamil conirmed that this system is functionally active. Taken together these data indicate the presence of a functional sarcoplasmic reticulum within the multi-cell type spheroid microtissue system. However, it must be noted that although this system exhibits key features of cardiac tissue, it remains somewhat electrically immature as is evident from its spontaneous beating, since fully mature cardiac cells are quiescent [44] and therefore future research will be guided into a more in-depth understanding of this parameter. The small size (∼88 μm in diameter) of the spheroid ensures suicient oxygen delivery as previously noted for the engineered biowire structure, which have a radius of approximately 300 μm [45] and therefore these cardiac spheroids can be maintained in culture without perfusion.
We determined that the majority of proteins found in the MS analysis of the multi-cell type spheroid microtissue system were previously found in publically available adult heart microarray data [33]. This included an increase in adult isoforms of an extensive range of sarcomeric proteins including β-myosin, Troponin T-I and C-proteins, tropomyosin, actin, myomesin, obscurin, nebulette, MLC2v that are essential for mechanical force generation [26,44]. In addition, we observed sarcomeric proteins previously not seen in 3D cultures such as alpha-sarcomeric protein and myosin-binding protein C [44]. We also found an increase in key proteins involved in contraction including RyR2, SERCA2 and PLN, as well as a signiicant increase in expression of Tmem65, an essential protein in cardiac contraction [29].
To the best of our knowledge this is the irst study in which dexrazoxane has been administered in tandem with the NRF2 activators CDDOme and sulforaphane to monitor their cardioprotective potential. Dexrazoxane, the only FDA approved cardioprotectant has been shown in clinical trials to be efective in multiple cancer modalities in limiting the cardiotoxic efect seen by anti-cancer drugs [46][47][48][49]. However, despite this, it is currently only administered to women with metastatic breast cancer who have received a cumulative dose of anthracycline of 300 mg/ m 2 or higher [3]. Initially, evidence from short-term clinical trials suggested that administration of dexrazoxane may increase the risk of secondary malignancies or decrease the antineoplastic eicacy of the anticancer drugs. These claims however were challenged in long-term follow ups [3] with some studies now supportive of the concomitant administration of dexrazoxane and anthracyclines to prevent cardiovascular complications [49,50]. However, despite this the European Medicines Agency declined widening the indication of dexrazoxane to all cancer patients and dexrazoxane remains only available to those patients who are expected to receive high cumulative doses of anthracyclines [50] even though low dose exposure has been associated with cardiac damage [1,51,52]. With this in mind, our study shows that for those patients on low dose anthracyclines to whom dexrazoxane remains unavailable, an alternative means of cardioprotection may be through the stimulation of the NRF2 pathway. We show here that CDDO-me (Fig. 6A) or sulforaphane (Fig. 6C) used alone increased cell viability and decreased ROS production (Fig. 6B) to the same eiciency as dexrazoxane alone.
Several mechanisms of dexrazoxane cardioprotection have been proposed including its ability to be a strong metal-chelating agent which chelates free iron thereby preventing the formation of the ironanthracycline complexes that are responsible for the generation of superoxide free radicals via redox reactions [53]. CDDO-me and sulforaphane on the other hand are potent inducers of the NRF2 pathway, a major cellular defence mechanism against oxidative stress. Used at low concentrations (1-100 nM) CDDO-me is cytoprotective but can induce apoptosis when administered at high concentration [54]. Here, we show that a low concentration of CDDO-me is cardioprotective but does not alter the chemotherapeutic eicacy of doxorubicin towards breast cancers cells (Fig. 5G).   Fig. 4. Mass spectrometry analysis of spheroid development A) Volcano plot of the total proteins identiied at 5 day and 5 weeks plotted with p-value against the log fold change from 3 independent experiments. Diferentially expressed proteins (p-value ≤ 0.05) are shown in red. Horizontal line represents p-value = 0.05 and vertical lines represent fold change of ± 1.5 fold, B) Changes in levels of key proteins involved in cardiac development relative to GAPDH C) Gene Ontology (GO) term analysis of diferentially expressed proteins (p-value ≤ 0.05), red represents upregulated and green down regulated proteins, D) IPA analysis of cardiovascular system development and function further indicated enrichment of proteins involved in development and contraction of the heart, red represents upregulated and green down regulated proteins, E) IPA analysis of upstream regulators identiied key transcription factors involved in cardiac development, orange represents signiicantly up regulated proteins and pathways, blue represents down regulated pathways and green represents down regulated proteins. F) Heatmap of 416 signiicantly up regulated proteins (≥ 1.5 fold and p ≤ 0.05 in the 5 week spheroid) by 3-fold intensity above global median in the adult and tissue microarray, had a previously known cardiac phenotype and were enriched in our previous atria [34] and ventricle proteome [29,34], red represents presence and black absence of proteins. (For interpretation of the references to colour in this igure legend, the reader is referred to the web version of this article).
This study clearly shows that dexrazoxane treatment does not inluence NRF2 activation (Fig. 6A) but does signiicantly decrease ROS production (Fig. 5D) and that CDDO-me when added alone also signiicantly decreases the ROS production to an equivalent degree (Fig. 5D) potentially through activation of downstream antioxidant genes including SRXN1, TXNRD1 [55] and NQO1 (Fig. 6A) [56]. The addition of CDDO-me + dexrazoxane in tandem further show a greater signiicant improvement in viability and decrease in ROS than either reagent alone. However, the addition of CDDO-me + dexrazoxane does not signiicantly decrease the levels of ROS beyond that of dexrazoxane or CDDO-me alone suggesting that addition of CDDO-me in tandem with dexrazoxane may promote additional protection via the NRF2 (caption on next page) L. Tomlinson, et al. B io m e d ic in e & P h a r m a c o t h e r a p y 1 1 2 ( 2 0 1 9 ) 1 0 8 6 3 7 pathway through a secondary mechanism of perhaps by increasing the glutathione levels as previously suggested [57]. Several clinical trials have focused on the preventive and therapeutic activities of CDDO-me against a range of cancers [21] with initial analysis showing growth suppression of malignant cells at higher doses of CDDO-me with no reported severe adverse reaction [21,54]. Amongst several completed and ongoing clinical trials of CDDO-me, only the BEACON trial in stage 4 chronic kidney disease patients has been terminated due to an increase in the frequency of cardiac events (to 8.8%) when compared to placebo [58]. However, post-hoc analysis on this patient population determined those who sufered cardiac events had previously been hospitalised for heart failure and exhibited elevated brain natriuretic peptide (BNP) levels. Those with no previous history of heart failure showed rates of adverse cardiac events similar to those seen in the placebo group (2.0%) [59]. Therefore, it is unlikely that the combined use of dexrazoxane and CDDO-me would cause cardiac events in a suitably stratiied patient population. Sulforaphane, has previously been implicated as cardioprotective in cultured rat cardiomyocytes [60] and phase II clinical trials in men with prostate cancer have deemed it a safe treatment although with limited eicacy in this population [61].
To mitigate the complex cardiotoxic action of doxorubicin, a multimechanistic cardioprotective approach is required. Since the administration of dexrazoxane to patients receiving a low dose of anthracyclines is restricted, alternative cardioprotective methods must be developed as cardiovascular damage in these patients is well documented.
In this study we have demonstrated the potential of exploiting cardioprotective efects of NRF2 induction to minimise doxorubicin induced oxidative stress. We further demonstrate a synergistic protective efect when both dexrazoxane and CDDO-me are administered in tandem. Dexrazoxane, CDDO-me and Sulforaphane have been independently assessed in clinical trials, but this study highlights the need for further investigations into the value of co-administration as a novel strategy to overcome anthracycline induced cardiotoxicity.  5. Synergistic cardioprotection against doxorubicin toxicity by dexrazoxane and CDDO-Me A) Immunoluorescence and phase contrast Lightsheet microscopy images of doxorubicin (red) using 0.001 mM doxorubicin after a 30 min incubation. Three planes (x, y and z) of images show penetration of drug throughout spheroid B) Assessment of cell viability at increasing doses of doxorubicin in 3D human spheroids in the presence and absence of 100 μM Dexrazoxane (Dex). Data normalised to vehicle only control and vehicle + 100 μM Dex respectively, graph represent mean ± SE, n = 3 C) Fold change of cell viability seen in the presence of 0.001 mM doxorubicin. Signiicant changes in viability seen upon pre-treatment with 100 μM Dex and 100 nM CDDO-me administered independently and in tandem. * p ≤ 0.05; ** p ≤ 0.005 students t-test, n = 3. D) Fold change in ROS revealed a signiicant decrease (*p ≤ 0.05) in ROS levels in the presence of dexrazoxane and CDDO-me independently but to a greater extent (**p ≤ 0.01) when administered in tandem. E) Signiicant changes in viability seen upon pre-treatment with 100 μM Dex and 10 μM sulforaphane administered independently and in tandem. * p ≤ 0.05; ** p ≤ 0.005 students t-test, n = 3. F) Structural integrity was monitored using α-actinin (red) showed a dose-dependent decrease in structure that was partially rescued in the presence of Dex and CDDO-me. G) Viability of AUG565 breast cancer cells showed a signiicant decrease in viability in the presence of 0.001 mM and 0.0001 mM Dox. Pre-treatment with 100 μM Dex and 100 nM CDDO-me in tandem showed no signiicant change to those treated with Dox alone (graphs represent mean ± SE, n = 3 for all experiments) (For interpretation of the references to colour in this igure legend, the reader is referred to the web version of this article).

Spheroid formation
Tri-culture spheroids were formed in ultra-low adhesion plates (ULA; Corning) by mixing human embryonic stem cell-derived cardiomyocytes (GE healthcare), human cardiac ibroblasts (Promocell) and human cardiac endothelial (Promocell) cells at a ratio of 4:2:1. A maximum of nine passages were permitted for the ibroblasts and endothelial cells, whereas the cardiomyocytes were used directly from the purchased vial. A total of 500 cells were pipetted into each well of a 96-well plate and cells were cultured in a 1:1 v/v ratio of RPMI (Sigma) media supplemented with B27 (Gibco) and endothelial cell media (MV2, Promocell). Spheroids generated from stem cell-derived cardiomyocytes alone were formed using a total of 1000 cells per well in 96-well ULA plates with RPMI media supplemented with B27 [62]. Spheroids were then allowed to mature until age required for experimentation.

Mass spectrometry
Approximately 100 spheroids at either 5 days or 5 weeks were suspended in 160 μl of 25 mM ammonium bicarbonate; 10 μl of 0.05% Rapigest (Waters, Manchester) was added and spheroids shaken at 550 rpm for 10 min at 80°C. The samples were reduced by the addition of 10 μl of 60 mM Dithiothreitol (DTT) with incubation at 60°C for 10 min and then alkylated by the addition of 10 μl of 180 mM iodoacetamide at room temperature for 30 min in the dark. Trypsin (Promega UK Ltd., Southampton, proteomics grade) 2 μg was added to the samples followed by overnight incubation on a rotating mixer at 37°C. The digestion was terminated and Rapigest removed by acidiication using 1 μl of triluoroacetic acid (TFA) with incubation at 37°C for 45 min followed by centrifugation at 15,000 x g for 15 min. For LC-MS/MS analysis, a 10 μl injection was analysed using an Ultimate 3000 RSLC nano system (Thermo Scientiic, Hemel Hempstead) coupled to a QExactiveHF mass spectrometer (Thermo Scientiic). The sample was loaded onto the trapping column (Thermo Scientiic, PepMap100, C18, 300 μm × 5 mm), using partial loop injection, for seven minutes at a low rate of 4 μl/min with 0.1% (v/v) formic acid (FA). The sample was resolved on the analytical column (Easy-Spray C18 75 μm × 500 mm 2 μm column) using a gradient of 97% A (0.1% formic acid) 3% B (99.9% ACN, 0.1% formic acid) to 70% A 30% B over 120 min at a low rate of 300 nl/min. The data-dependent program used for data acquisition consisted of a fullscan 60,000 resolution MS scan (AGC set to 3e6 ions with a maximum ill time of 100 ms) the 18 most abundant peaks were selected for MS/MS using a 30,000 resolution scan (AGC set to 1e5 ions with a maximum ill time of 45 ms) with an ion selection window of 1.2 m/z and a normalised collision energy of 28. To avoid repeated selection of peptides for MS/MS the program used a 30 s dynamic exclusion window.

Proteomic analysis
The MS data was processed using Progenesis Q.I. for Proteomics, version 2 (Nonlinear Dynamics, Newcastle upon Tyne, UK). Samples were aligned according to retention time using a combination of manual and automatic alignment. Default peak picking parameters were applied and features with charges from 1 + to 4 + containing three or more isotope peaks were retained. Database searching was performed using Mascot (Matrix Science, London, UK) against UniHuman Reviewed and UniMouse databases. A Mascot Generic File, created by Progenesis, was searched against UniHuman Reviewed database. A ixed carbamidomethyl modiication for cysteine and variable oxidation modiication for methionine were speciied. A precursor mass tolerance of 10 ppm and a fragment ion mass tolerance of 0.01 Da were applied. The results were then iltered to obtain a peptide false discovery rate of 1%.

Ca 2+ imaging
Ca 2+ transients were measured utilizing the Ca 2+ indicator Fluo-4 (Thermo Fisher). Briely, spheroids were incubated with 5 μM Fluo-4 AM for 30 min at 37°C after which they were washed twice with dye-free medium and placed back into the incubator for 30 min. A Zeiss Axio Observer microscope was used to measure the luorescence intensity of Fluo-4 AM treated with or without diferent pharmacological agents.

Cell viability
3D microtissue spheroids were incubated with varying concentrations of doxorubicin (Selleck Chemicals) for 48 h and cell viability measured using a CellTiter-Glo Luminescent Cell Viability Assay (Promega) as described by the manufacturer's instructions. 100 μM Dexrazoxane was added 3 h and 100 nM CDDO-me, 10 μM sulforaphane and 10 μM ML385, 24 h prior to doxorubicin incubation.

In silico estimates of spheroid O 2
A reaction-difusion equation was used to estimate the oxygen proiles within the spheroids (full details in supplementary material). A Michaelis-Menten oxygen uptake term was assumed with maximal oxygen consumption rates (Vmax) for the three cell types taken from a study by Sekine et al. [63] which estimated oxygen consumption rates of 1.97 × 10 −16 , 5.28 × 10 -17 , 6.11 × 10 -17 mol/s/cell, respectively, for human iPSC-CM, cardiac ibroblasts (FB) and human cardiac microvascular endothelial cells (EC). The Michaelis-Menten half maximal parameter (Km) and the rate of oxygen difusion within the spheroid were taken to be 6.24 × 10 -3 mol/m 3 [64] and 6.33 × 10 -10 m 2 /sec [65], respectively. Cell densities were calculated using the average spheroid radius of 88 μm which comprises of a total of 498 cells in a mixed ratio of 4:2:1 for CM:FB:EC (285:142:71 cells) -giving 9.89 × 10 13 , 4.97 × 10 13 , 2.487 × 10 13 cells/m 3 for CM, FB and EC cells, respectively. These same cell densities were then assumed for all spheroid radii considered. The oxygen concentration at the spheroid boundary was ixed at 21% for each spheroid radii. Radial spherical symmetry was assumed and the steady state oxygen proiles for the spheroid interiors were calculated for spheroid radii ranging from 50 to 200 μm.

Quantitative PCR
Total RNA was isolated from vehicle and drug treated spheroids using the RNeasy mini kit (Qiagen), as per the manufacturer's instructions. cDNA was synthesized using a GoScript cDNA synthesis kit (Promega), as per the instructions provided for a 96 well plate format. Quantitative PCR (qPCR) analysis of NRF2, NAD(P)H dehydrogenase quinone 1 (NQO1), Suliredoxin-1 (SRXN1), Aldo-keto reductase family 1 member C1 (AKR1C1), Thioredoxin reductase 1 (TXNRD1) and glyceraldehyde-3-phosphate dehydrogenase (GAPDH) genes was performed on a Viia 7 real-time PCR instrument (Applied Biosystems) using Power SYBR green PCR Master Mix (ThermoFisher). GAPDH was used as a normalization control. Primer sequences are detailed in Supplementary Table 1.

Reactive oxygen species (ROS) measurements
3D microtissue spheroids were incubated with varying concentrations of doxorubicin (Selleck Chemicals) for 48 h and ROS measured using the ROS-Glo H 2 O 2 Assay (Promega) as described by the manufacturer's instructions. 100 μM Dexrazoxane was added 3 h and 100 nM CDDO-me or 10 μM sulforaphane, 24 h prior to doxorubicin incubation.

Author contribution statement
PS designed the study and LT, RB and PS wrote the manuscript text. ZQL carried out bioinformatics analysis and prepared Fig. 4. PS, HEC, CM, IC, MJC guided experiments. SDW carried out mathematical modelling and prepared Fig. 2. All authors reviewed and edited the manuscript.