The animal protocol was approved by the Committee on the Ethics of Animal Experiments of the Canton Ticino, Switzerland (TI32/18). The study was carried out in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the Directive 2010/63/EU. The study protocol is depicted schematically in Figure 1A. Sprague Dawley female rats (10–12 weeks old; from Charles River Laboratories) were subdivided into four groups. In the Dox group, rats were injected i.p. with 6 doses of Dox hydrochloride (Sigma-Aldrich), one dose each other day (from d1 to d11), for a cumulative dosage of 20 mg/kg, followed by six doses of phosphate-buffered saline (PBS; pH 7.4), one dose each other day (from d19 to d29), as described (Milano et al., 2014). In the Trz group, rats received 6 doses of PBS (from d1 to d11) followed by 6 doses of Trz (Roche), one dose each other day (from d19 to d29), for a cumulative dosage of 20 mg/kg. In the combined Dox/Trz group, rats received 6 doses of Dox hydrochloride (from d1 to d11) followed by 6 doses of Trz (from d19 to d29). Control (Ctrl) rats received 12 doses of PBS at the time points corresponding to drug administration in treated groups (from d1 to d11, and from d19 to d29). Notably, Trz monotherapy was started at d19 of the study protocol to match the time point of Trz administration in the combined therapy group. Trz-mediated changes in LV function at later points were measured in a separate series of experiments (Supplementary Figure 1).

Heart function was monitored by echocardiography using a VEVO 2100 high-resolution imaging system (VisualSonics) at d0, d12, d19, d30, and d37, as described. Anesthesia was induced using 2% isoflurane mixed with 100% oxygen in an induction chamber. Rats were then placed on a heat pad in the supine position and kept at 37°C to minimize fluctuations of body temperature. Data acquisition was performed in rats lightly anesthetized with 0.5–1% isoflurane in order to maintain HR ≥ 350 bpm. Two-dimensional short-axis M-mode echocardiography was performed and LV end-systolic (LVESV) and end-diastolic (LVEDV) volumes, ejection fraction (LVEF) and fractional shortening (LVFS) were determined, as previously described (Barile et al., 2018).

Isolated CMs from in vivo treated rats were analyzed at d19 (early Dox time point), and at d37 (late Dox time point, early Trz time point, combined Dox/Trz treatment). For CMs isolation, rats were anesthetized using a cocktail of ketamine (100 mg/kg) and xylazine (75 mg/kg) and humanly euthanized by cervical dislocation. Hearts were harvested and perfused ex vivo in a Langendorff mode, as previously described (Rocchetti et al., 2014). CMs were isolated separately from LV and right ventricular (RV) free walls. Rod-shaped, Ca²⁺ tolerant CMs were used for patch-clamp measurements and confocal microscopy less than 12 h after tissue dissociation.

To investigate the impact of each treatment on the organization of TT, sarcolemmal membranes were marked by incubating CMs with 20 μM 3-di-ANEPPDHQ (Life Technologies, Carlsbad, CA, United States) and TT oriented transversely along z-lines were visualized (Rocchetti et al., 2014). 3-di-ANEPPDHQ was dissolved in the following solution (in mM): 40 KCl, 3 MgCl₂, 70 KOH, 20 KH₂PO₄, 0.5 EGTA, 50 L-Glutamic acid, 20 Taurine, 10 HEPES, 10 D-glucose (pH 7.4) for 10 min at RT (Sacconi et al., 2012). Cell contraction was prevented by adding blebbistatin (17 μM; Sigma). CMs were washed with the same solution before confocal microscopy analysis. Images of loaded CMs were acquired by laser-scanning microscopy (images: 1,024 × 1,024 pxls, 78 μm × 78 μm) using a confocal microscope (Nikon C2 plus) with a 40× oil-immersion objective. Eight-bit gray-scaled images were analyzed by spatial Fast Fourier Transform analysis to quantify periodic components of pixel variance. To compensate for staining differences among cells, a raw power spectrum was generated with ImageJ (v.1.4) and normalized to its central peak. TT density was quantified by normalizing the area under the harmonic relative to the spatial frequency of 0.5 μm^–1 (between 0.3 and 0.7 μm^–1) to the area of the entire spectrum (Rocchetti et al., 2014).

Action potentials (AP) were recorded by pacing CMs at 1 Hz in current-clamp conditions. AP duration measured at 90% and 50% of the repolarization phase (APD₉₀ and APD₅₀, respectively), diastolic potential (E_diast) and maximal AP phase 0 depolarization velocity (dV/dt_max) were determined. Single cells were superfused with standard Tyrode’s solution containing (in mM): 154 NaCl, 4 KCl, 2 CaCl₂, 1 MgCl₂, 5.5 D-glucose, and 5 HEPES-NaOH (pH 7.35). Experiments were carried out in whole-cell configuration; the pipette solution contained (in mM): 23 KCl, 110 KAsp, 0.4 CaCl₂, 3 MgCl₂, 5 HEPES-KOH, 1 EGTA-KOH, 0.4 NaGTP, 5 Na₂ATP, 5 Na₂PC (pH 7.3). Delayed afterdepolarizations (DADs) were defined as diastolic depolarizations with amplitude ≥1 mV. The percentage of cells exhibiting DADs was quantified. Beat-to-beat variability of repolarization (BVR) was expressed as the short-term variability (STV) of APD₉₀ (i.e., the mean orthogonal deviation from the identity line (Heijman et al., 2013; Altomare et al., 2015), calculated as follows:

for 30 consecutive APs (n_beats) at steady-state level. STV data are shown using APD_n versus APD_{n + 1} (Poincaré) plots.

Cardiac myocytes were incubated in Tyrode’s solution for 45 min with the membrane-permeant form of the dye, Fluo4-AM (10 μmol/L), and then washed for 30 min to allow for the de-esterification process. Fluo4 emission was collected through a 535 nm band pass filter, converted to voltage, low-pass filtered (100 Hz) and digitized at 2 kHz after further low-pass digital filtering (FFT, 50 Hz). Intact CMs were field-stimulated at 1, 2 and 4 Hz at 37°C during superfusion with standard Tyrode’s solution. Ca²⁺ transient (CaT) amplitude at steady-state and the sarcoplasmic reticulum (SR) Ca²⁺ content (CaSR) estimated by an electronically timed 10 mmol/L caffeine pulse were evaluated at each cycle length. The diastolic fluorescence was used as reference (F₀) for signal normalization (F/F₀) after subtraction of background luminescence. For intergroup comparisons, Ca²⁺ handling parameters measured at each cycle length in a treated group were normalized to values measured in Ctrl. Rate-dependency of CaT decay kinetic was expressed as half-time decay (T_0.5). Na⁺/Ca²⁺ exchanger (NCX) function was estimated by mono-exponential fit of caffeine-induced CaT. Frequency of resting Ca²⁺ waves was assessed under 1 min resting conditions before pacing. A Ca²⁺ wave was defined as a Ca²⁺ oscillation occurring at rest with an amplitude >3 SD over resting fluorescence levels (F_rest). Comparable results were obtained using amplitude cutoffs up to 5 F_rest SD. Frequency of spontaneous CaT occurring at rest (resting CaT) was assessed as an additional parameter of SR instability and Ca²⁺ overload.

Spontaneous unitary Ca²⁺ release events (Ca²⁺ sparks) were recorded at RT in Fluo4-AM (10 μM)-loaded CMs under resting conditions. Tyrode bath solution contained 2 mM CaCl₂. Images were acquired at 60× magnification in line-scan mode (xt) at 0.5 kHz by confocal Nikon A1R microscope. Each cell was scanned along a longitudinal line and #10 xt frames (512 × 512 pxls) were acquired. Background fluorescence was measured. Confocal setting parameters were kept constant among experimental groups. Images were analyzed by SparkMaster plugin (Fiji) software (Picht et al., 2007). Automatic spark detection threshold was 3.8. Only in-focus Ca²⁺ sparks (amplitude > 0.3) were included in analyses. The following spark parameters were measured: frequency (event number/s/100 μm), amplitude (ΔF/F0), full width at half-maximal amplitude (FWHM; μm), full duration at half-maximal amplitude (FDHM; ms), full width (FW; μm) and full duration (FD; ms), time-to-peak (TtP, ms), and decay time constant (τ; ms). Spark mass (ΔF/F₀^∗μm³), an index of Ca²⁺ spark volume (Hollingworth et al., 2001), was calculated as spark amplitude^∗1.206^∗ FWHM³. Spark-mediated SR Ca²⁺ leak was calculated as the product of spark mass and frequency.

Proteins from CMs extracts were separated by SDS-polyacrylamide gel electrophoresis (4–12% Bis-Tris Criterion BIO-RAD gels), blotted for 1 h, incubated with polyclonal anti-SERCA2 primary antibody (N-19; Santa Cruz Biotechnology) at 4°C overnight, followed by incubation with a specific secondary antibody labeled with fluorescent markers (Alexa Fluor or IRDye) for 1 h. Signal intensity was quantified by Odyssey Infrared Imaging System (LI-COR). SERCA2 protein levels were normalized to actin levels, as measured using polyclonal anti-actin Ab (Sigma). Data are shown as percent changes vs. Ctrl.

Results are shown as mean ± SE. Unpaired Student’s t-test was used to test for significant differences in two-group analyses. One-way and two-way ANOVA were used to test for significance among multiple groups, with post hoc comparison analyses using Bonferroni’s multiple comparison test. Chi²-test was used for comparison of categorical variables. The statistical test used in each analysis is mentioned in the respective figure legends. Statistical significance was defined as p < 0.05.

Echocardiographic results are shown in Figure 1B. Compared to Ctrl, Dox-treated animals showed increases in LVESV at d19, d30, and d37, and in LVEDV at d19 and d37, along with decreases in LVEF and LVFS at all time points. In the absence of Dox pre-treatment, Trz-treated animals showed an increase in LVESV at d30 (i.e., one day after administration of the last Trz dose), and decreases in LVEF and LVFS. Of note, in this study Trz was administered from d19 to d29 to mimic clinical protocols that involve the sequential administration of the two agents to attenuate toxicity. In a separate study, we evaluated the effects of Trz on LV function for up to 37 days (i.e., the same time frame used for Dox). The results are shown in Supplementary Figure 1. Trz induced a transient decrease in LVFS at d12 and a transient increase in LVESV at d19. Animals receiving Dox/Trz combined therapy showed increases in LVESV and LVEDV, along with decreases in LVEF and LVFS at d30 and d37. Significant differences between Trz monotherapy and the combined Dox/Trz treatment were observed for LVESV and LVEF at d37, consistent with additive toxic effects by the two agents. Body weight, tibia length, and heart weight did not significantly differ among groups (Supplementary Table 1).

Representative confocal region of interest (ROI) of 3-di-ANEPPDHQ–treated CMs in the different groups are shown in Figure 2A. Disruption of TT architecture in the Dox and Dox/Trz groups, but not in the Trz group, can be appreciated visually. Representative spatial Fast Fourier Transform analyses of TT-power in LV and RV CMs are shown in Figure 2B. Quantitative analysis confirmed that Dox, but not Trz, treatment induces TT disarray (Figure 2C).

LV and RV CMs were analyzed separately because electrical measurements are influenced by ventricular loading conditions. AP recordings in isolated CMs from either ventricle showed increases in both APD₅₀ and APD₉₀ in the Dox group at d19, but not at d37 (Figures 3A,B), indicating negligible effects at later time points. In the absence of Dox pre-treatment, Trz did not affect APD, whereas it significantly prolonged it in Dox pre-treated rats (Supplementary Table 2). Depolarizing events during diastole and systole were recorded as DADs and early afterdepolarizations (EADs), respectively. The percentage of cells exhibiting DADs at 1 Hz-stimulation was increased in Dox-treated animals at d19, but not at d37. In analogy to its effect on APD, Trz increased the frequency of DADs selectively in rats pre-treated with Dox (Figures 3C,D). Similar changes were found for beat-to-beat variability of repolarization (BVR), which reflects APD₉₀ time-variability (i.e., electrical instability) and represents a pro-arrhythmic parameter (Johnson et al., 2010). The dispersion of APD₉₀ values around the identity line in Poincaré plots was increased in the Dox group at d19 and in the Dox/Trz group, but not in the Dox group at d37 neither in Trz monotherapy (Figure 4A). Quantitative STV data are shown in Figure 4B. The slope of the linear correlation between STV and APD₉₀ was comparable in all groups (Supplementary Figure 2), indicating that increases in BVR were strictly dependent on APD prolongation. These results indicate pro-arrhythmic conditions in both the Dox and Dox/Trz groups.

In principle, TT disarray and changes in APD, DADs, and BVR can impact intracellular Ca²⁺ handling. We analyzed evocated Ca²⁺ transients (CaT) at 1, 2 and 4 Hz in field-stimulation, as well as caffeine-induced CaT which reflects SR Ca²⁺ content (CaSR; Supplementary Figure 3). Trz treatment resulted in significant decreases in CaT amplitude and CaSR in LV CMs, and to a slightly lesser extent in RV CMs. Unaltered or even increased CaT amplitudes were observed in Dox and Dox/Trz CMs (Figure 5A). Studies of the CaT decay kinetics showed a smaller decay half-time (T_0.5) in Ctrl LV CMs compared to Ctrl RV CMs, in line with faster SR Ca²⁺ sequestration in LV CMs compared to RV CMs (Sathish et al., 2006). CaT decay in Dox d19 and to a lesser extent in Dox/Trz LV CMs was slower than in Ctrl (Figure 5B), suggesting Dox-induced changes in Ca²⁺ removal kinetics; a similar trend was observed in RV CMs. Accordingly, SERCA protein levels were significantly decreased in LV and RV CMs in both Dox and Dox/Trz CMs (Figure 5C). Trz alone did not significantly affect CaT decay kinetic and SERCA protein levels. To assess removal of intracellular Ca²⁺ through NCX, caffeine-induced CaT decay kinetic was evaluated. As shown in Supplementary Figure 4, rate-dependent NCX activity was observed in LV but not in RV Ctrl CMs, accordingly to previous data on chamber-specific NCX expression (Correia Pinto et al., 2006). However, rate-dependent NCX activity was absent in CMs from treated animals, particularly in the Trz group, suggesting faster NCX-dependent Ca²⁺ removal under these conditions, especially at slow pacing rates. The percentage of LV CMs exhibiting Ca²⁺ waves at resting was increased in all treated groups, with a similar increase in RV CMs from Dox d37 rats. The frequency of spontaneous CaT at resting in LV CMs was increased in both Dox and Dox/Trz groups (Figure 5D). These findings are in good agreement with our results on DAD frequency (Figures 3C,D).

To sum up, in spite of SERCA downregulation, global Ca²⁺ handling was preserved in field-stimulated CMs in Dox and Dox/Trz groups, while SR instability was observed mainly at resting. This suggests that compensative mechanisms (i.e., increased Ca²⁺ influx during prolonged APs) may take place in Dox and Dox/Trz groups. Trz alone did not affect electrical activity while directly affect intracellular Ca²⁺ handling.

The effects of Dox on Ca²⁺ waves and resting CaT led us to investigate SR stability by quantifying spontaneous SR Ca²⁺ release events visualized as Ca²⁺ sparks. Representative images of Ca²⁺ sparks in the different LV groups are shown in Figure 6A. A similar pattern was observed in RV CMs. Increased frequencyof Ca²⁺ sparks was readily apparent in the Dox and Dox/Trz group, whereas Trz alone did not significantly impact this parameter. Spark mass was increased in LV CMs from Trz-treated animals, with similar trends in Dox-treated ones (Llach et al., 2019), but not in the Dox/Trz group. Dox, and to a lesser extent Trz, induced an increased spark-mediated SR Ca²⁺ leak quantified by the spark mass^∗spark frequency index (Figure 6B). Dox, and to a lesser extent Trz, increased the number of so-called Ca²⁺ “embers,” defined as Ca²⁺ sparks with FDHM > 20 ms, the peak value of the FDHM distribution (Louch et al., 2013; Figure 6C and Supplementary Figure 5). Moreover, FDHM and the decay time constant (tau_decay), two markers of altered RyR openings, were increased in LV myocytes of Dox-treated animals (Supplementary Table 3), supporting Dox-induced SR instability.