NOX2 NADPH OXIDASE PROMOTES PATHOLOGIC CARDIAC REMODELING ASSOCIATED WITH DOXORUBICIN CHEMOTHERAPY

Abstract

Doxorubicin is a highly effective cancer treatment whose use is severely limited by dose-dependent cardiotoxicity. It is well established that doxorubicin increases reactive oxygen species (ROS) production. In this study, we investigated contributions to doxorubicin cardiotoxicity from Nox2 NADPH oxidase, an important ROS source in cardiac cells, which is known to modulate several key processes underlying the myocardial response to injury. Nox2-deficient mice (Nox2^−/−) and wild-type (WT) controls were injected with doxorubicin (12 mg/kg) or vehicle and studied 8 weeks later. Echocardiography indicated that doxorubicin-induced contractile dysfunction was attenuated in Nox2^−/− versus WT mice (fractional shortening: 29.5±1.4 vs 25.7±1.0 %; P<0.05). Similarly, in vivo pressure-volume analysis revealed that systolic and diastolic function was preserved in doxorubicin-treated Nox2^−/− versus WT mice (ejection fraction: 52.6±2.5 vs 28.5±2.3 %, LVdPdt_min: −8379±416 vs −5198±527 mmHg s⁻¹, EDPVR: 0.051±0.009 vs 0.114±0.012; P<0.001). Furthermore, in response to doxorubicin, Nox2^−/− mice exhibited less myocardial atrophy, cardiomyocyte apoptosis and interstitial fibrosis, together with reduced increases in profibrotic gene expression (procollagen IIIαI, TGF-β₃, connective tissue growth factor) and matrix metalloproteinase-9 activity, versus WT controls. These alterations were associated with beneficial changes in NADPH oxidase activity, oxidative/nitrosative stress and inflammatory cell infiltration. We found that adverse effects of doxorubicin were attenuated by acute or chronic treatment with the AT1 receptor antagonist, losartan, which is commonly used to reduce blood pressure. Our findings suggest that ROS specifically-derived from Nox2 NADPH oxidase make a substantial contribution to several key processes underlying development of cardiac contractile dysfunction and remodeling associated with doxorubicin chemotherapy.

INTRODUCTION

Doxorubicin remains one of the most widely-used anti-neoplastic agents in the effective treatment of various cancers (1). Unfortunately, its chronic use is severely limited by dose-dependent cardiotoxicity characterized by progressive cardiac dilatation, contractile dysfunction and ultimately congestive heart failure (2,3). The chronic response to doxorubicin-induced myocardial injury involves complex reorganization of the extracellular matrix (mainly driven by matrix metalloproteinases; MMPs) and substantial alterations in cardiomyocyte biology, such as apoptosis and changes in cell growth, excitation-contraction coupling and cytoskeleton organization (3).

The mechanisms by which doxorubicin causes deleterious structural and functional changes are poorly understood, although activation of several pathways has been proposed, including local release of vasoactive substances, mitochondrial dysfunction, lipid peroxidation, glutathione peroxidase depletion, reduced sarcoplasmic Ca²⁺-ATPase activity and impaired myocardial energetics (1,2). However, recent attention has focused on the potential involvement of myocardial reactive oxygen species (ROS) and oxidative stress, which are increased by chronic doxorubicin treatment and have been proposed to be at least partly responsible for the associated cardiotoxicity (4,5). Indeed, experimental doxorubicin-induced cardiac contractile dysfunction and adverse morphological changes are inhibited by exogenous antioxidant treatment or overexpression of endogenous antioxidant enzyme, and similar protective effects are reported clinically (6-8).

Although, it is well established that oxidative stress modulates several key processes underlying doxorubicin-induced cardiotoxicity, such as extracellular matrix remodeling, cardiomyocyte apoptosis and alteration of cardiac contractile properties (5,9-13), the precise contribution of different ROS sources remains unknown. Identification of wide-ranging cardiovascular actions of ROS has prompted detailed investigation of the roles of potential sources, including the mitochondrial electron transport chain, xanthine oxidase and dysfunctional nitric oxide synthases (NOS) (14). In this regard, recent work has identified a family of NADPH oxidases as important sources of myocardial ROS (10,15,16). The prototypic NADPH oxidase comprises a catalytic core (Nox2/p22^phox), together with several cytosolic subunits (p47^phox, p67^phox, p40^phox, Rac) that associate upon activation (14). Several Nox isoforms have been identified (Nox1-5); the major cardiac isoforms are Nox2 and Nox4, which are expressed in cardiomyocytes, fibroblasts, and endothelial cells (14,15).

Recent studies indicate that NADPH oxidase-derived ROS play a pivotal role in cardiac pathophysiology. NADPH oxidase activity/expression is increased in experimental and clinical heart failure (15-17). Furthermore, NADPH oxidase-dependent ROS production regulates several key components of cardiac remodeling, such as myocyte hypertrophy, contractile dysfunction, apoptosis and fibrosis (10,11,18). Interestingly, our previous studies using Nox2^−/− mice revealed differential effects that were dependent upon the remodeling stimulus. Nox2 NADPH oxidase was critical for cardiac hypertrophy and fibrosis in response to chronic angiotensin II (AngII) infusion (18). Similarly, development of cardiac contractile dysfunction, chamber dilatation, myocyte hypertrophy/apoptosis and fibrosis after myocardial infarction were attenuated in Nox2^−/− compared to wild-type (WT) mice (16). In contrast, after pressure-overload, left ventricular (LV) hypertrophy occurred independently of Nox2, although Nox2 was essential for cardiac fibrosis and contractile dysfunction (10,15). It therefore appears that Nox2 NADPH oxidase may influence individual components of the cardiac remodeling phenotype in a stimulus-specific manner.

The potential role of Nox2 NADPH oxidase in doxorubicin-induced cardiotoxicity has yet to be clearly established. However, doxorubicin increases myocardial NADPH oxidase activity in vivo (19) and stimulates MMP-2 expression/activity and apoptosis in vitro via NADPH oxidase-dependent activation of JNK/ERK and hydrogen peroxide, respectively (5,20). Furthermore, production of several known stimuli of NADPH oxidases which are important in cardiac remodeling (e.g. AngII, aldosterone, endothelin-1) are increased by doxorubicin (21-23). More definitive evidence comes from a clinical study which identified several genetic polymorphisms of NADPH oxidases predisposing patients to increased risk of doxorubicin cardiotoxicity (4). The same group reported that doxorubicin-stimulated cardiac superoxide production was attenuated in Nox2^−/− mice, which were resistant to doxorubicin-induced LV dilatation and contractile dysfunction (4,24), although no further detailed analysis of remodeling was undertaken. The aim of this study was to precisely define the role of Nox2 NADPH oxidase in doxorubicin-induced cardiotoxicity and to investigate how it may modulate individual components of the remodeling phenotype in this setting.

METHODS

Experimental animals

A colony of Nox2^−/− mice is established on a C57BL/6J background (25) in our institution. All experiments were performed in accordance with the Home Office Guidance on the Operation of the Animals (Scientific Procedures) Act 1986 (UK).

Doxorubicin treatment protocol

Male Nox2^−/− and WT littermate controls (8-10 weeks) were injected with a cumulative dose of 12 mg/kg doxorubicin or equivalent volume of vehicle control via 3 weekly injections (4 mg/kg i.p. at 0, 7 and 14 days) and subsequent analyses performed 8 weeks after the first injection. No mortality was associated with this dosing regimen. For ex vivo analyses, animals were sacrificed by sodium pentobarbitone overdose before hearts were excised and either frozen in liquid nitrogen and stored at −80 °C or fixed in 10 % neutral-buffered formalin solution for further studies. A separate cohort of mice were chronically treated with losartan (10 mg/kg/day in drinking water; pre-treatment period of 7 days) (26) and assessed 4 weeks after the first doxorubicin injection. A minimum of 6 animals per group were studied for all protocols.

Echocardiography and invasive assessment of cardiac function

Mice were anaesthetized with 1.5 % isofluorane/oxygen, placed on a warming pad and imaged in the supine position using a Vevo770® ultrasound system with high-frequency 45 MHz RMV707B scanhead (VisualSonics Inc., Canada). M-mode parasternal short-axis scans at papillary muscle level were used to quantify LV end-diastolic and end-systolic diameters (LVEDD, LVESD) from which % fractional shortening was calculated (LVEDD-LVESD)/LVEDD*100).

Isofluorane was then increased to 2 % and the right carotid artery cannulated with a high-fidelity 1.2F pressure-volume catheter (SciSense Inc., Canada), aortic pressure measured and the catheter advanced into the LV for recording of steady-state function. The abdominal inferior vena cava (IVC) was then briefly occluded, allowing construction of variably-loaded pressure-volume loops, from which LV end-systolic and end-diastolic pressure-volume relations (ESPVR, EDPVR) were determined. Absolute volume measurements were corrected for α (derived by simultaneous aortic outflow recording by echocardiography) and parallel conductance (hypertonic saline injection) (27).

Assessment of cardiac remodeling

Following sacrifice, measurements of LV and right ventricular weights were taken and indexed to tibial length. All histological analyses were performed on fixed (10 % neutral-buffered formalin), paraffin-embedded LV sections (5 μm). Cardiomyocyte cross-sectional area was determined by hematoxylin and eosin staining, analyzing cells with centrally-located nuclei. Cardiac interstitial fibrosis was assessed by picrosirius red staining (0.1 % w/v), excluding coronary vessels and perivascular regions (16). Data were quantified by blinded digital image analysis (NIS-Elements, Nikon UK).

Immunohistochemistry for 3-nitrotyrosine and CD45 were performed with rabbit (Millipore, UK) and rat (BD Biosciences, UK) polyclonal antibodies (1:1000), respectively, using diaminobenzidine as the chromogen and nuclear counterstaining with hematoxylin. Data were quantified by blinded digital image analysis (NIS-Elements, Nikon UK).

Cardiomyocyte apoptosis was determined by TUNEL staining (Roche Diagnostics, UK). TUNEL-positive myocyte nuclei were counted, and data expressed as % total nuclei identified by DAPI staining in the same sections.

Real-time RT-PCR

Total RNA was extracted from LV homogenate using TRI reagent (Sigma-Aldrich, UK), and cDNA synthesized by reverse transcription (Applied Biosystems, UK). mRNA expression of GATA-4, procollagen IIIαI, transforming growth factor-β₃ (TGF-β₃), connective tissue growth factor (CTGF), MMP-2, MMP-9, Nox2 and Nox4 were analyzed by real-time RT-PCR using fluorescent SYBR Green (Prism 7300, Applied Biosystems, UK) and β-actin for normalization by the comparative C_t method. Primer sequences are shown in Supplementary Table 1.

Gelatin zymography

Activities of myocardial MMP-2 and MMP-9 were analyzed by gelatin zymography (16). Zymograms were digitized and densitometric values digitally quantified (VisionWorksLS, Ultra-Violet Products, UK) and normalized to the WT control.

NADPH oxidase activity

NADPH-dependent superoxide production was assessed in LV homogenate using lucigenin (5 μmol/L)-enhanced chemiluminescence (NADPH 300 μmol/L; 100 μg protein; 37 °C) (15), performed in triplicate. Potential superoxide sources were assessed by experiments performed in the presence of Tiron (20 mmol/L; cell-permeable superoxide scavenger), diphenyleneiodonium (DPI, 10 μmol/L; inhibits NADPH oxidase and other flavoproteins), N^ω-Nitro-L-arginine methyl ester hydrochloride (L-NAME, 100 μmol/L; inhibits superoxide from dysfunctional NOS), oxypurinol (100 μmol/L; xanthine oxidase inhibitor), or rotenone (20 μmol/L; mitochondrial inhibitor).

Isolated cardiomyocyte studies

Ventricular cardiomyocytes were isolated from male WT mice (aged 10-12 weeks) by collagenase digestion (28), plated at 50 cells/mm² in MEM containing 10 % inactivated FBS and antibiotics, and equilibrated for 1 h. Cells were then incubated with or without doxorubicin (0.5 μmol/L) (5) for 3 h in the presence or absence of the AT1 receptor antagonist, losartan (10 μmol/L, pre-treatment for 30 mins) (29). Immediately after treatment, cardiomyocyte NADPH-dependent superoxide production was quantified using lucigenin-enhanced chemiluminescence and data normalized to cell viability assessed by CellTiter-Fluor™ assay (Promega, UK). The remainder of the cells were frozen in liquid nitrogen and stored at −80 °C for real-time RT-PCR analysis of Nox2, Nox4 and TGF-β₃ mRNA expression.

Statistical analysis

Data were analyzed by one-way ANOVA followed by Bonferroni or Dunnett’s multiple comparison test, or unpaired Student’s t test, as appropriate. P<0.05 was considered to be significant.

RESULTS

Echocardiography

Data are summarized in Table 1. No differences were observed between WT and Nox2^−/− controls. However, Nox2^−/− mice demonstrated better contractile function after doxorubicin treatment compared to WT mice, as indicated by higher % fractional shortening and lower LVESD. Heart rate and LVEDD were similar between groups.

TABLE 1.

LV functional parameters in control and doxorubicin-treated WT and Nox2^−/− mice

	WT control	WT DOX	Nox2−/− control	Nox2−/− DOX
Echocardiography data
n	15	11	16	17
Heart rate (bpm)	511±7	518±16	491±16	503±14
LVEDD (mm)	4.12±0.04	4.12±0.07	4.24±0.06	4.25±0.06
LVESD (mm)	2.73±0.07	3.07±0.08*	2.77±0.09	2.86±0.15
FS (%)	34.4±1.2	25.7±1.0***	34.8±1.4	29.5±1.4*†
Pressure-volume data
n	10	10	8	10
MABP (mmHg)	79.5±3.3	78.2±2.7	79.6±2.3	77.6±3.0
LVESP (mmHg)	105±3.3	98±2.1	100±2.5	100±2.8
LVEDP (mmHg)	6.6±1.0	15.4±2.4**	6.9±1.5	4.9±1.0†††
LVdP/dtmax (mmHg s−1)	9676±386	6114±520***	9481±475	8965±475
LVdP/dtmin (mmHg s−1)	−8218±337	−5198±527***	−8600±472	−8379±416†††
τ (ms)	6.0±0.3	10.3±0.6***	6.3±0.5	5.7±0.5†††
LVESV (μl)	33.5±3.3	63.2±9.0**	32.8±5.4	38.6±3.8†
LVEDV (μl)	70.3±7.1	77.6±8.8	69.0±5.5	69.0±6.0
Stroke volume (μl)	43.2±5.3	24.1±3.2**	43.1±4.3	38.3±3.2†
Ejection fraction (%)	56.5±3.5	28.5 ± 2.3***	58.9±4.9	52.6±2.5†††
SW (mmHg.μl g−1)	37964±5291	16658±3448***	40359±3286	29074±2327
ESPVR	2.43±0.42	1.17±0.38*	2.64±0.29	2.31±0.20
EDPVR	0.040±0.011	0.114±0.012***	0.041±0.008	0.051±0.009†††

DOX, doxorubicin; FS, fractional shortening; LVEDV, LV end-diastolic volume; LVESP, LV end-systolic pressure; LVESV, LV end-systolic volume; MABP, mean arterial blood pressure; SW, stroke work. Mean±SEM.

^*P<0.05,

^**P<0.01,

^***P<0.001, versus control;

^†P<0.05,

^†††P<0.001, Nox2^−/− doxorubicin versus WT doxorubicin.

LV pressure-volume analysis

Steady-state data are presented in Table 1. WT and Nox2^−/− controls exhibited similar function, and consistent with echocardiography, doxorubicin-treated WT mice had impaired systolic function, reflected by decreased LVdP/dt_max, stroke volume and ejection fraction and increased LV end-systolic volume. LVdP/dt_min was also reduced in WT doxorubicin-treated mice, whilst LV end-diastolic pressure (LVEDP) and the isovolumic relaxation time constant τ were increased, indicating diastolic dysfunction. In contrast, none of these steady-state parameters were altered in Nox2^−/− doxorubicin-treated mice. Furthermore, representative steady-state pressure-volume loops (Figures 1A and C) clearly demonstrate that doxorubicin-induced decreases in stroke work (area of loop) observed in WT mice were preserved in Nox2^−/− mice (Table 1).

Figure 1.

Effect of doxorubicin on LV pressure-volume relations in WT and Nox2^−/− mice. Representative (A, C) steady-state pressure-volume loops, (B, D) ESPVR after IVC occlusion.

Figures 1B and D show representative data obtained under variable loading after IVC occlusion. The slope of the LV ESPVR was decreased in WT doxorubicin-treated mice, indicating reduced contractility, but preserved in Nox2^−/− mice. In addition, the slope of the LV EDPVR (measures diastolic stiffness) was increased in WT, but not Nox2^−/− doxorubicin-treated mice (Table 1). Importantly, mean arterial blood pressure remained similar between groups.

Myocardial atrophy

Whilst there was no difference in body weight between WT and Nox2^−/− controls (30.7±0.7 vs 30.3±0.8 g, n=30,33), this was reduced by doxorubicin in both groups (WT: 24.9±0.7 g, n=33; P<0.001 vs control; Nox2^−/−: 27.6±0.4 g, n=37; P<0.05 vs control) with a tendency towards a greater decrease in WT versus Nox2^−/− mice. Morphometric data were therefore normalized to tibial length. Chronic doxorubicin treatment decreased LV/tibial length in WT, but not Nox2^−/− mice (Figure 2A). No differences in right ventricular/tibial length were observed between groups (data not shown). Consistent with this atrophic action of doxorubicin, cardiomyocyte cross-sectional area (Figure 2B) was decreased in WT, but not Nox2^−/− mice. GATA-4 mRNA expression was similar between groups (Figure 2C).

Figure 2.

Effect of doxorubicin on cardiomyocyte remodeling in WT and Nox2^−/− mice. (A) LV/tibial length (TL; n=13-20), (B) cardiomyocyte cross-sectional area (n=7), (C) GATA-4 mRNA expression (n=6-9), (D) TUNEL staining (n=8). Mean±SEM. **P<0.01, versus control; †P<0.05, Nox2^−/− doxorubicin versus WT doxorubicin.

Cardiomyocyte apoptosis

Chronic doxorubicin treatment increased the number of TUNEL-positive cells in LV sections from WT versus Nox2^−/− mice (158±37 vs 23±12 %; P<0.01; Figure 2D) compared to controls. Representative examples of TUNEL-stained LV sections are shown in Supplementary Figure 1.

Cardiac fibrosis

Interstitial fibrosis, assessed in LV sections by picrosirius red staining, was increased in WT versus Nox2^−/− doxorubicin-treated mice (123±20 vs 37±23 %; P<0.01; Figure 3A) compared to controls. Representative examples of picrosirius red-stained LV sections are shown in Supplementary Figure 2. Consistent with this, mRNA expression of the pro-fibrotic genes procollagen IIIαI, TGF-β₃ and CTGF was attenuated in doxorubicin-treated Nox2^−/− compared with WT mice (Figure 3B-D).

Figure 3.

Effect of doxorubicin on extracellular matrix remodeling in WT and Nox2^−/− mice. (A) cardiac interstitial fibrosis (n=7), (B-E) pro-fibrotic gene mRNA expression (n=10-14), (F) MMP-9 activity by gelatin zymography (n=6). Mean±SEM. *P<0.05, **P<0.01, ***P<0.001, versus control; ^†P<0.05, ^†††P<0.001, Nox2^−/− doxorubicin versus WT doxorubicin.

Myocardial MMP activity and expression

Myocardial MMP-9 activity was increased by doxorubicin in WT, but not Nox2^−/− mice (Figure 3F). Similarly, MMP-9 mRNA expression was elevated in WT, but not Nox2^−/− doxorubicin-treated mice (Figure 3E). MMP-2 activity/expression remained similar between groups (data not shown).

NADPH oxidase activity and expression

LV NADPH-dependent superoxide production was increased in WT, but not Nox2^−/− doxorubicin-treated mice (Figure 4A). Superoxide generation was inhibited by Tiron and DPI, but not by L-NAME, oxypurinol and rotenone, in both WT and Nox2^−/− mice (Figure 4B shows WT data), indicating that the observed signal reflected NADPH oxidase activity.

Figure 4.

Effect of doxorubicin on NADPH oxidase activity/expression and oxidative/nitrosative stress in WT and Nox2^−/− mice. (A) NADPH-dependent superoxide production (relative light units, RLU; n=7-9), (B) effect of inhibitors on NADPH oxidase activity in WT mice (oxy=oxypurinol, rot=rotenone; n=6), (C-D) mRNA expression (N.D. = not detected; n=9-13), (E) 3-nitrotyrosine staining (n=6). Mean±SEM. *P<0.05, **P<0.01, ***P<0.001, versus control; ^††P<0.01, Nox2^−/− doxorubicin versus WT doxorubicin.

Myocardial Nox2 mRNA expression was increased in doxorubicin-treated WT mice, but was not detected in Nox2^−/− mice (Figure 4C). Nox4 mRNA expression was also increased in doxorubicin-treated WT, but not Nox2^−/− mice (Figure 4D).

3-Nitrotyrosine staining

LV 3-nitrotyrosine levels, used to assess in situ oxidative/nitrosative stress, were increased in doxorubicin-treated WT mice compared to controls, and this increase was attenuated in Nox2^−/− mice (Figure 4E). Representative examples of 3-nitrotyrosine-stained LV sections are shown in Supplementary Figure 3.

CD45 staining

Immunohistochemistry demonstrated that myocardial infiltration of CD45-positive leukocytes was increased in WT doxorubicin-treated mice (Figure 5). Although there was a tendency towards increased leukocyte infiltration in Nox2^−/− doxorubicin-treated mice, this was not significant and was clearly reduced compared to the WT group.

Figure 5.

Effect of doxorubicin on leukocyte infiltration in WT and Nox2^−/− mice. (A-D) Representative LV sections stained for CD45, (E) positive spleen control, (F) negative secondary-only spleen control, (G) Mean quantification data±SEM, n=9, as CD45-positive cells per high-power field (HPF). *P<0.05, versus control.

Losartan treatment

Chronic treatment with the AT1 receptor antagonist, losartan, resulted in normalization of echocardiographic fractional shortening (Figure 6A) and LVESD (data not shown) in doxorubicin-treated WT mice; heart rate and LVEDD were unaffected (data not shown). Losartan also increased invasively-assessed LVdP/dt_max, slope of the ESPVR (Figure 6B-C), ejection fraction, stroke work and LVdP/dt_min, whilst decreasing τ and slope of the EDPVR (data not shown) in these animals. Furthermore, chronic losartan treatment prevented the development of myocardial atrophy in WT doxorubicin-treated mice, together with increases in NADPH-dependent superoxide production and mRNA expression of Nox2 and MMP2 (Figure 6D-F and H). Losartan also tended to reduce doxorubicin-induced increases in Nox4 mRNA expression, although this failed to reach significance (Figure 6G). Similarly, in vitro studies in isolated WT cardiomyocytes demonstrated that acute losartan treatment attenuated doxorubicin-induced increases in NADPH oxidase activity and mRNA expression of Nox2, Nox4 and TGF-β₃ (Figure 6I-L).

Figure 6.

Effect of losartan (Los) on doxorubicin cardiotoxicity in WT mice. Chronic treatment: (A-C) contractile function (n=5-10), (D) LV/tibial length (TL; n=9-11), (E) NADPH-dependent superoxide production (relative light units, RLU; n=5-8), (F-H) mRNA expression (n=5-8). Acute treatment in isolated cardiomyocytes: (I) NADPH-dependent superoxide production (n=6-7), (J-L) mRNA expression (n=6-7). Mean±SEM. *P<0.05, **P<0.01, ***P<0.001, versus vehicle; ^†P<0.05, ^††P<0.01, Los doxorubicin (DOX) versus doxorubicin; ^‡P<0.05, ^‡‡P<0.01, doxorubicin versus basal.

DISCUSSION

This study clearly demonstrates that Nox2 NADPH oxidase plays an important role in the development of doxorubicin-induced cardiac injury and remodeling. The major evidence for this was that genetic disruption of Nox2 protected against the development of cardiac contractile dysfunction, atrophic cardiomyocyte degeneration, apoptosis, interstitial fibrosis and MMP-9 activation after chronic doxorubicin treatment. These actions on cardiac remodeling were associated with beneficial changes in NADPH oxidase activity, oxidative/nitrosative stress and inflammatory cell infiltration.

Although doxorubicin is extremely effective against many tumors, its well established cardiotoxic side effects severely limit its clinical utilization. The incidence of cardiotoxicity is low at doses <400 mg/m², rising to 7 % at the commonly-employed upper limit (550 mg/m²), steeply increasing thereafter (4). In the present study, a cumulative doxorubicin dose of 12 mg/kg was used and mice studied 8 weeks after first administration. This protocol was adopted in order to reflect common treatment regimens, which usually involve several cycles of chemotherapy, and to approximate the clinical chronology of chronic doxorubicin-induced cardiotoxicity, which typically develops over several months (21). It should be noted that doxorubicin exerts its anti-tumor action by inhibiting DNA topoisomerase II, thereby preventing unwinding of the double helix during replication, and by direct intercalation into the DNA, preventing resealing (20). Importantly, these pathways are not thought to involve ROS; indeed our confirmatory studies in HeLa cells demonstrated that knockdown of Nox2 using a specific siRNA had no significant effect on the anti-tumor efficacy of doxorubicin (see Supplementary Figure 4).

It is well established that doxorubicin increases myocardial ROS production, which has been implicated in the associated cardiotoxicity (4,5). However, the sources of ROS and mechanisms by which they may modulate different components of the cardiotoxic remodeling phenotype are less clear. Potential myocardial ROS sources include the mitochondrial electron transport chain, xanthine oxidase, dysfunctional NOS and NADPH oxidases (14). In this study we focused on NADPH oxidases, which are the only myocardial source whose primary function is ROS generation. NADPH oxidases are activated by several stimuli (e.g. AngII, cyclic stretch, TNF-α) that are pivotal in cardiac remodeling, and importantly, are also capable of modulating other ROS sources (14). Indeed, NADPH oxidases may potentiate mitochondrial ROS generation and convert xanthine dehydrogenase to xanthine oxidase (30,31). Furthermore, NADPH oxidase-derived ROS are known to promote oxidative degradation of tetrahydrobiopterin, leading to NOS uncoupling and superoxide production (32). However, NADPH-dependent superoxide production was unaffected by rotenone, oxypurinol or L-NAME, suggesting that these alternate sources were not major contributors to doxorubicin-induced ROS generation in our study. Furthermore, our data clearly demonstrate that doxorubicin-stimulated NADPH-dependent superoxide production is critically-dependent on Nox2, as increases observed in WT hearts were not seen in Nox2^−/− mice. This pattern was repeated with 3-nitrotyrosine, indicating that in situ oxidative/nitrosative stress was only elevated in WT doxorubicin-treated hearts and was therefore dependent upon Nox2. In addition, Nox2 mRNA expression was increased by doxorubicin in WT hearts and, as expected, was not detectable in Nox2^−/− hearts. Our previous studies investigating Nox2 NADPH oxidase in cardiac remodeling found Nox4 expression to be increased in both WT and Nox2^−/− mice after chronic pressure-overload and myocardial infarction, together with parallel increases in NADPH oxidase activity (15,16). However, in this study, although Nox4 mRNA expression was increased in doxorubicin-treated WT hearts, it remained unaltered in those from Nox2^−/− mice. The reasons for this are unclear although it may suggest potential cross-talk between the two isoforms and dependence of Nox4 on Nox2 in the setting of doxorubicin cardiotoxicity. Nonetheless, it appears that upregulation of both Nox2 and Nox4 may contribute to increased NADPH oxidase activity in doxorubicin-treated WT mice.

Doxorubicin is known to cause cardiac contractile dysfunction (2) and this was clearly evident in our WT mice. These animals developed systolic dysfunction as evidenced by reductions in fractional shortening, both pressure- (LVdP/dt_max) and volume-based indices (LV end-systolic volume, ejection fraction), and the slope of the ESPVR, which is acknowledged to be the gold-standard measure of contractility (27). WT mice also demonstrated impaired LV relaxation (reduced LVdP/dt_min, prolonged τ) and diastolic dysfunction (increased LVEDP and slope of the EDPVR). In contrast, all of these parameters were preserved in doxorubicin-treated Nox2^−/− mice.

The marked attenuation of contractile dysfunction observed in Nox2^−/− mice is likely to have occurred due to beneficial actions on different components of the remodeling phenotype. It appears that reduced cardiomyocyte atrophy and interstitial fibrosis may be the most important contributors, with apoptosis playing a secondary role by reducing LV mass and stimulating reparative extracellular matrix remodeling (3). In addition, it is possible that Nox2-derived ROS could exert adverse effects on contractile function through direct actions on excitation-contraction coupling, myofilament calcium responsiveness and cellular energetics (12,13,33). It is well established that doxorubicin cardiotoxicity is at least partly attributable to cardiomyocyte apoptosis (6). Indeed, in the present study, the incidence of TUNEL-positive apoptotic nuclei was reduced in hearts from doxorubicin-treated Nox2^−/− mice, suggesting that this may contribute to preserved contractile function in these animals. There appears to be considerable debate in relation to the trophic actions of doxorubicin on the cardiomyocyte, with some studies reporting hypertrophy (7,34) whilst others suggest an atrophic action (35,36). Here we clearly demonstrate that chronic doxorubicin treatment results in myocardial atrophy, as reflected by decreased morphometric LV/tibial length ratio and cardiomyocyte cross-sectional area. This was attenuated in Nox2^−/− doxorubicin-treated mice, suggesting that preservation of cardiomyocyte mass may also have contributed to maintained contractile function. It should be noted that reduced cardiomyocyte apoptosis in Nox2^−/− mice in response to doxorubicin is likely to have also influenced preservation of morphological LV mass in these animals (37,38), although the extent of its contribution in this context is unclear. It has been suggested that the atrophic action of doxorubicin may occur secondary to reductions in GATA-4, a key regulator of cardiac development, known to modulate myocardial expression of sarcomeric proteins such as α-myosin heavy chain and troponin I (39,40). However, GATA-4 mRNA expression remained similar between groups, suggesting that this pathway was not involved in our study.

Contractile function is also known to be influenced by alterations in extracellular matrix remodeling and fibrosis. Indeed, chronic doxorubicin treatment was associated with increases in interstitial fibrosis and expression of pro-fibrotic genes (procollagen IIIαI, TGF-β₃, CTGF) in WT, but not Nox2^−/− mice, and these changes are likely to be largely responsible for the differences in diastolic and relaxation properties (e.g. LVEDP, τ, EDPVR) between groups. Furthermore, myocardial expression and activity of MMP-9, but not MMP-2 was increased in WT, but not Nox2^−/− doxorubicin-treated mice. Indeed, it is well established that MMP activation is redox-sensitive (9) and several recent in vitro and in vivo studies have implicated NADPH oxidases, particularly those containing Nox2 (16,41). Doxorubicin may also activate MMPs in a redox-sensitive manner. Myocardial activity and expression of MMP-2 and MMP-9 are increased acutely by doxorubicin in mice in vivo, and this is inhibited by the superoxide dismutase mimetic, MnTMPyP (6). Similarly, MMP-2 and MMP-9 activity/expression are augmented by doxorubicin in H9c2 rat ventricular cardiomyocytes in association with elevated Nox1 expression, and attenuated by the flavoprotein inhibitor, DPI (5). Indeed, in the present study, myocardial activity and expression of MMP-9, but not MMP-2, was increased in WT, but not Nox2^−/− doxorubicin-treated mice. In this regard, it has been suggested that doxorubicin-induced activation of MMP-2 and MMP-9 may involve different pathways. For example, a single bolus doxorubicin dose in miceinduced early activation of myocardial MMP-2, which was followed MMP-9 activation several days later (42). In addition, an in vitro study indicated that p38MAPK may be responsible for doxorubicin-induced MMP-9 activation, whereas JNK/ERK may be involved in MMP-2 regulation (5). Doxorubicin cardiotoxicity is frequently characterized by LV dilatation and MMP activation is thought to play an important role in this process (2,3). However, in the present study, LV end-diastolic structural parameters assessed by both echocardiography and pressure-volume analysis remained similar between groups, despite activation of MMP-9. The reasons for this are unclear although it may reflect a temporal relationship between MMP-9 and LV chamber dilatation or the involvement of other MMPs. It is also conceivable that LV dilatation may have been masked by the atrophic actions of doxorubicin on the cardiomyocyte which were observed in WT, but not Nox2^−/− mice.

Doxorubicin is well known to induce myocardial inflammation (35). In the present study, chronic doxorubicin treatment was associated with myocardial infiltration of CD45-positive cells in WT, but not Nox2^−/− mice, suggesting a role for ROS derived specifically from Nox2 NADPH oxidase. Indeed, Nox2 is known to be expressed in inflammatory cells and their recruitment may influence redox-sensitive processes, such as MMP activation and extracellular matrix remodeling, through generation of cytokines and other factors (14). Interestingly, in mice lacking T cells (RAG1^−/−), AngII-induced hypertension (known to be mediated by Nox2) is blunted and associated with reduced superoxide production and attenuation of endothelial dysfunction and cardiac hypertrophy (43). Not only does this indicate that T cell Nox2 NADPH oxidase is necessary for AngII-induced hypertension, it also suggests that some of the protective effects against doxorubicin-induced cardiotoxicity observed in Nox2^−/− mice in the present study, may occur secondary to activation of Nox2 in inflammatory cells. However, it is interesting to note that although patients with chronic granulomatous disease (caused by defects in genes encoding the Nox2 NADPH oxidase complex) frequently develop recurrent infections, Nox2^−/− mice do not due to compensation by other protective pathways (44,45). Indeed, in our study doxorubicin-treated Nox2^−/− mice exhibited a tendency towards increased myocardial infiltration of CD45-positive cells, suggesting that the observed effects on cardiac remodeling may not be accounted for solely by differences in the inflammation.

It is well established that several of the actions of Nox2 NADPH oxidase in cardiac remodeling are mediated via activation of AngII (11,18), production of which is known to be increased by doxorubicin (21). Indeed, in the present study, chronic treatment with the AT1 receptor antagonist, losartan, prevented development of both contractile dysfunction and myocardial atrophy in response to doxorubicin. Furthermore, increases in NADPH oxidase activity and Nox2/Nox4 mRNA expression induced by doxorubicin both in vivo and in vitro were attenuated by co-treatment with losartan. These data suggest that AngII signalling may play an important role in doxorubicin-induced Nox2 activation, although it is likely that other key upstream mediators are also involved. Interestingly, Rac1-mediated activation of Nox2 NADPH oxidase is implicated in the proximal signaling involved in AngII-induced cardiomyocyte hypertrophy, ahead of downstream Akt activation (46), and doxorubicin cardiotoxicity is attenuated by statins which exert antioxidant effects via Rac1 inhibition (47). Signaling upstream of Rac1 translocation in neonatal cardiomyocytes may also involve PKC, c-Src-family tyrosine kinases and proline-rich tyrosine kinase 2 (48-50). Nonetheless, data on Nox-isoform specific activation in the cardiomyocyte are scarce and further studies are clearly required to establish the precise mechanisms underlying doxorubicin-induced Nox2 activation and these are currently the focus of our research.

In summary, these data suggest that ROS derived specifically from Nox2 NADPH oxidase make a significant contribution to several key processes underlying cardiac remodeling associated with doxorubicin chemotherapy. Selective targeting of NADPH oxidases may provide a novel therapeutic strategy against the cardiotoxic actions of doxorubicin, thereby maximizing its effectiveness as an anti-neoplastic agent.