Timed use of digoxin prevents heart ischemia-reperfusion injury through a REV-ERBα-UPS signaling pathway

Abstract

Myocardial ischemia-reperfusion injury (MIRI) induces life-threatening damages to the cardiac tissue and pharmacological means to achieve cardioprotection are sorely needed. MIRI severity varies along the day-night cycle and is molecularly linked to components of the cellular clock including the nuclear receptor REV-ERBα, a transcriptional repressor. Here we show that digoxin administration in mice is cardioprotective when timed to trigger REV-ERBα protein degradation. In cardiomyocytes, digoxin increases REV-ERBα ubiquitinylation and proteasomal degradation, which depend on REV-ERBα ability to bind its natural ligand, heme. Inhibition of the membrane-bound Src tyrosine-kinase partially alleviated digoxin-induced REV-ERBα degradation. In untreated cardiomyocytes, REV-ERBα proteolysis is controlled by known (HUWE1, FBXW7, SIAH2) or novel (CBL, UBE4B) E3 ubiquitin ligases and the proteasome subunit PSMB5. Only SIAH2 and PSMB5 contributed to digoxin-induced degradation of REV-ERBα. Thus, controlling REV-ERBα proteostasis through the ubiquitin-proteasome system is an appealing cardioprotective strategy. Our data support the timed use of clinically-approved cardiotonic steroids in prophylactic cardioprotection.

Introduction

The mammalian circadian clock is a critical cell-autonomous mechanism regulating many, if not all, aspects of cellular, tissular and organismal homeostasis ¹. Heart physiology makes no exception to this rule and electrophysiological, mechanical and metabolic functions of the cardiac tissue are submitted to cyclic daily variations ^2,3, which are dependent on the integrity of the molecular clock ⁴. Quite expectedly, genetic or environmental perturbations of the circadian clock in mouse models lead to cardiovascular pathologies including impaired cardiac contractility, fibrosis and cardiomyopathy. Human observational studies also largely concur to demonstrate that circadian rhythm misalignments are associated with cardiometabolic diseases ^2,5. The prevalence of adverse cardiovascular events such as stroke and myocardial infarction at the sleep-to-wake transition underlines the complex interplay between (mechanisms of) timed cyclic blood pressure control, heart function and vascular tone regulation. In addition to altering heart responsiveness to pathological events, circadian (dys)regulations also affect the heart and cardiomyocyte’s ability to recover from acute injury.

Susceptibility of the heart to myocardial ischemia-reperfusion injury (MIRI) relies on transient adaptative mechanisms to hypoxia ⁶ and rapid restoration of oxidative metabolism and contractile efficiency ^7–9. During ischemia, the myocardial tissue undergoes ATP depletion, acidosis by accumulation of lactic acid and activation of a series of immediate- and late- response intracellular signaling pathways called “RISK” and “SAFE” pathways ¹⁰. These signaling pathways, which are also stimulated upon preconditioning (PC) ¹¹, protect mitochondrial function. Reperfusion causes massive cellular production of reactive oxygen species ¹⁰, alkalinization and calcium overload. This induces hypercontractility of the myofibers and alters mitochondrial function, thereby opening the mitochondrial permeability transition pore (MPTP) and triggering cell death by apoptosis ¹². Sensitivity to MIRI is time-of-the-day (ToD)-dependent and is exacerbated during the sleep-to-wake transition in human and mouse ¹⁴. Genetic alteration of circadian rhythmicity in mice ^14,15 or circadian rhythm (dys)synchrony in humans correlate with increased acute cardiac injury ^16–18. We recently demonstrated that timed pharmacological antagonism of the cyclically-expressed nuclear receptor REV-ERBα, a heme-binding transcriptional repressor within the molecular clock machinery ¹⁹, preventively protects mouse heart from deleterious effects of MIRI in a p21-dependent manner ²⁰. This observation mechanistically linked the ToD dependency of MIRI with the core clock gene machinery, identifying REV-ERBα as a novel target in pharmacological PC.

The search for pharmacologically protective compounds in MIRI has proven difficult ²¹. Despite their continued use in atrial fibrillation to ameliorate heart contractile properties ²², cardiotonic steroids (CTS) such as digoxin are not recommended in MIRI cardioprotection at therapeutic concentrations ²³. Renewed interest in CTS has however sparkled from the identification of dissociated inotropic and non-ion pumping signaling properties of Na⁺-K⁺ ATPase (NKA), a target of CTS. Isoforms of the 3 NKA subunits [α, β and γ/FXYD] associate to make up the heteromeric membrane NKA complex

Inotropic effects of CTS are essentially mediated by their ability to bind to and inhibit NKA ATPase activity ²⁵, thereby triggering intracellular accumulation of Na⁺ and decreasing Ca⁺⁺ efflux through the Na⁺/Ca⁺⁺ exchanger NCX1 (encoded by the Slc8a1 gene) ²⁶. CTS-induced signaling responses can be induced at ATPase sub-inhibitory concentrations and stems at least in part from NKA’s ability to act as a CTS plasma membrane receptor ^{27 28,29}. CTS binding to NKA α1 subunit allows the activation of multiple downstream intracellular signaling pathways, including part of RISK or SAFE pathways, hence suggesting a potential usefulness of CTS in preventive MIRI protection ³⁰.

Thus, given the temporal gating of signaling pathways, the involvement of both the NKA signalosome and REV-ERBα in MIRI protection, and the known susceptibility of REV-ERBα activity to intracellular signaling pathways, we investigated whether CTS-induced signaling pathways could alter REV-ERBα functions and sensitivity to MIRI. Here we show that digoxin and other CTS activate signaling pathways in cardiomyocytes and trigger REV-ERBα protein degradation through the ubiquitin-proteasome system (UPS) in an SRC kinase-dependent manner. Timed in vivo treatment with digoxin prevented REV-ERBα protein accumulation and conferred protection against MIRI.

Results

The cardiac protective effect of digoxin is time of the day-dependent

We previously demonstrated, using the Langendorff heart perfusion model, that isolated mouse hearts display a ToD-dependent sensitivity to MIRI. Mouse hearts are intrinsically more resistant to MIRI at the beginning of the rest phase [zeitgeber time 0 (ZT0), beginning of the light phase] than at the end of this period (ZT12, average infarct size 27% vs 47%)(²⁰ and Fig. 1A). We asked whether ex vivo PC efficiency follows a similar pattern. Submitting ZT0 isolated hearts to a short iterative ischemia-reperfusion sequence (PC) prior to a prolonged ischemic period neither protected nor aggravated infarct size. In contrast, MIRI-prone ZT12 hearts were protected by this local PC protocol (average infarct size 31% vs 47%) (Fig. 1A). To shed light on these differential sensitivities, the molecular phenotype of ZT0 and ZT12 hearts was investigated by comparing their transcriptomic profiles (Fig. 1B and Supp. Table 1). Only a few protein-encoding genes were found to be differentially expressed (n=145, Fig. 1B), with 10-12% being core circadian genes including REV-ERBα, which is lowest expressed at ZT0. In addition, ≈75% of differentially expressed genes are cyclically expressed with a period of 20-24 hours (Supp. Table 1, Fig. 1B, right panel and Fig. 1C). None of them belonged to RISK or SAFE cardioprotective pathways. Thus, these data suggest that ZT12 hearts are intrinsically more sensitive to MIRI, are responsive to PC and exhibit specific molecular phenotypic features related to molecular clock-driven processes.

Fig. 1. Time of the day- and REV-ERBα-dependent effects of digoxin in mouse heart.

(A) Time of the day dependency of mouse heart sensitivity to MIRI and PC. Upper panel: experimental outline. Lower panel: infarct sizes at ZT0 and ZT12, with or without prior exposure to an iterative short ischemia/reperfusion sequence. Results are shown as mean+/-SEM (n=8-9) which were compared using Brown-Forsythe & Welch ANOVA test, followed by Dunnett’s multiple comparison test. Infarct sizes were calculated based on necrosis area/area at risk. (B) Left panel: overrepresented Gene Ontology Biological Processes (GO BP) terms in differentially expressed genes (ZT0 vs ZT12) mouse hearts; right panel: up- or downregulated and/or rhythmically-expressed (period=22-24h) genes. (C) Time-dependent expression of mouse heart genes. (D) MIRI at ZT0 or ZT9 in wild type mouse hearts after injection of vehicle or digoxin (IP: 1mg/kg). Top: experimental outline and representative transversal sections of TTC-stained mouse hearts post IR (red stain: healthy tissue, unstained: necrotic tissue); Bottom: infarct size measurements. Results are shown as mean+/-SEM (n=7-13) which were compared using Brown-Forsythe & Welch ANOVA test, followed by Dunnett’s multiple comparison test. (E) REV-ERBα protein level in mouse heart whole extracts. Upper panel: representative Wes analysis of REV-ERBα protein levels in mouse whole hearts. Bottom panel: quantification of REV-ERBα protein levels as a function of time. Results are shown as mean+/-SEM (n=3-4, see also Supp. Fig. 2A) which were and compared using Brown-Forsythe & Welch ANOVA test, followed by Dunnett’s multiple comparison test. (F) In vivo digoxin ED₅₀. REV-ERBα levels were assessed at ZT9 in whole heart extracts. Results are shown as mean+/-SEM (n=9-16, see also Supp. Fig. 2B) which were and compared using Brown-Forsythe & Welch ANOVA test, followed by Dunnett’s multiple comparison test. (G) MIRI at ZT9. Myocardial IR tolerance was evaluated at ZT9 in wild type vs Nr1d1^-/- or Cdkn1a^-/- mouse hearts treated either with vehicle or digoxin (ZT5, IP: 1mg/kg). Results are shown as mean+/-SEM (n=6-7) which were compared using Brown-Forsythe & Welch ANOVA test, followed by Dunnett’s multiple comparison test. All measurements were from distinct samples. *: P<0.05, **: P<0.01, ***: P<0.005 and ****: P<0.0001

We then assessed whether the previously reported ex vivo PC properties of digoxin can translate in vivo and whether they are ToD-dependent. We injected a single bolus of digoxin 4 hours prior to organ sampling at ZT0 or ZT9 at a pharmacologically well-tolerated dose (1 mpk IP ³¹). This timing is based on pharmacokinetics studies of digoxin in rodents showing plasma concentration peaking within 1 hour after injection, followed by a 4- to 6-hour tissue distribution phase ³² yielding an estimated extracellular fluid concentration of about 5-10μM ³³. The infarct size in untreated mice was as in Fig. 1A, larger by 20% at ZT9 when compared to ZT0 hearts (Fig. 1D), in agreement with ²⁰. Whereas digoxin pretreatment showed a clear, albeit non-significant (p=0.08) trend to increase MIRI at ZT0, it significantly reduced infarct size by 20% at ZT9 (Fig. 1D). Pre-ischemic monitoring of heart performances with a constant flow perfusion did not reveal significant changes in contraction force, perfusion pression and myocardial contractility (Supp. Fig. 1A-D), thus ruling out an effect of digoxin treatment on coronary vascular tone, endothelial function as well as on systolic and diastolic dynamics. In the MIRI model, a significant proportion of hearts displayed post-ischemic ventricular arrhythmias, making the analysis of cardiac contractile performance upon recovery infeasible.

The cardioprotective effect of digoxin is REV-ERBα-dependent

Since the selected time points correspond to the nadir (ZT0) and zenith (ZT9) of REV-ERBα protein level, respectively (Fig. 1E and Supp. Fig. 2A), we assessed whether digoxin treatment affects REV-ERBα mRNA and protein levels. Increasing doses of digoxin administered at ZT5 led to a dose-dependent decrease REV-ERBα protein level at ZT9 to reach 60% at 1 mpk (Fig. 1F and Supp. Fig. 2B). In contrast, the expression of the REV-ERBα-encoding Nr1d1 gene, whose expression precedes REV-ERBα protein expression by approximately 2 hours (Fig. 1C and Supp. Fig. 2C), was not affected by digoxin treatment. The expression of REV-ERBα target genes (Arntl/Bmal1 and Cdkn1a/p21, Supp. Fig. 2C), which mirror the REV-ERBα protein level pattern in homeostatic conditions (Fig. 1C and Supp. Fig. 2C), were increased by digoxin treatment (Supp. Fig. 2D) as well as P21 protein levels (Supp. Fig. 2E). Digoxin did not affect REV-ERBα protein level at ZT0, which was barely detectable at this time point (Supp. Fig. 2F). This ToD sensitivity to digoxin was not due to changing levels during the day/night cycle of NKA isoforms ATP1A1, or ATP1A2 and ATP1A3 mRNAs or of corresponding protein levels at the time of treatment (Supp. Fig. 3A,B). Their expression was neither affected by REV-ERBα KO (Supp. Table 2) nor by digoxin treatment (Supp. Fig. 3A). Thus, these data suggest that digoxin interferes with REV-ERBα protein levels through post-transcriptional regulation.

Digoxin administered at ZT4 did not protect mouse hearts from MIRI at ZT9 in Nr1d1^-/- mice but rather increased infarct size (+11%) (Fig. 1G), phenotypically mimicking ZT0 hearts (Fig. 1D). REV-ERBα via p21 expression contributes to increased resistance to MIRI ²⁰. A similar experiment in p21-depleted mice confirmed the higher susceptibility of these mice to MIRI but digoxin treatment was still protective in these mice (-18%, average infarct size 55%)(Fig. 1G). Thus, REV-ERBα expression is required to confer cardioprotective properties to a prophylactic digoxin treatment without involving the anti-apoptotic protein P21.

Injected digoxin preferentially accumulates in skeletal muscles, heart and liver ³⁴. Similar to the cardiac tissue, digoxin-exposed livers displayed reduced REV-ERBα protein level and increased target gene expression (E4bp4/Nfil3, Bmal1/Arntl, cdkn1a/p21), while Nr1d1 mRNA levels were left unchanged (Supp. Fig. 4A-E). REV-ERBα expression was also sensitive to digoxin in the human hepatoma cell line HepG2 (Supp. Fig.4F). Taken together, these observations raised the question of how digoxin downregulates REV-ERBα protein expression.

Digoxin affects REV-ERBα protein stability in a cell-autonomous manner

Analysis of bulk cardiac tissue as above may overlook outcomes stemming from multiple inter- or intracellular crosstalks. We therefore asked whether REV-ERBα protein degradation occurs in isolated digoxin-treated cardiomyocytes. We first performed experiments using the human cardiomyocyte cell line AC16, derived from adult human ventricular heart tissue ³⁵. Since digoxin and other CTS have significant pro-apoptotic properties ^36,37, we investigated whether digoxin induced AC16 cell death cells. AC16 cells were synchronized at T_-2 by a dexamethasone pulse (2 hours) which transiently induces Per1 gene expression ³⁸ and subsequently incubated with digoxin at T₀. Cell viability was only mildly affected by digoxin at tested concentrations (Supp. Fig. 5A). Probing apoptosis-related protein levels in AC16 whole cell extracts did not reveal any significant change in 43 players of the apoptotic cascade (Supp. Fig. 5B). In these conditions, REV-ERBα reached a maximal expression 24 hours after synchronization (T₂₄). Digoxin blunted REV-ERBα expression at all time-points (Fig. 2A and Supp. Fig. 5C). REV-ERBα protein levels were normalized to HSP90α levels, whose half-life is 36-40 hours, vs 1 hour for REV-ERBα. NKA subunits α1 and α2 mRNAs and protein levels were steadily detectable in these conditions (Supp. Fig. 3C,D). Digoxin’s effect on REV-ERBα protein level was fully prevented by an anti-digoxin antibody (Supp. Fig. 5D), excluding any artifactual modulation of cellular pathways by contaminants. The digoxin effect was dose-dependent with an observed EC₅₀= 96nM (Supp. Fig. 5E), which fits with the reported binding constant of digoxin to the human NKA α1 subunit (K_D=87nM), the major CTS signalosome relay ^39,40. It is also consistent with the observed efficiency of the non-glycosylated CTS bufalin [EC₅₀<0.1μM (Supp. Fig. 5F), K_D=42nM ³⁹]. The structurally-related ouabain also decreased REV-ERBα stability (Supp. Fig. 5G).

Fig. 2. Digoxin affects REV-ERBα protein stability in a cell-autonomous manner.

(A) Cyclic REV-ERBα protein levels in AC16 cells. Top panel: experimental outline. REV-ERBα levels were determined in synchronized human AC16 cells treated with vehicle or digoxin (0.5μM) at T0 and harvested at T0, T6, T12, T18 and T24. Bottom panel: Representative image from WES analysis. HSP90α was used as a protein loading control. Numbers indicate mean fold-changes [relative to maximal REV-ERBα expression (24h, untreated cells)]. Results are shown as mean+/-SEM (n=3) which were compared using a one-way ANOVA test, followed by a Tukey’s multiple comparison test (see Supp. Fig. 5A). (B) NR1D1 and BMAL1 mRNA expression in synchronized AC16 cells. Cells were treated with vehicle or digoxin (0.5μM) at T0 and harvested at T24 (n=3-5). Results were analyzed as in (A). (C) Real-time monitoring of the BMAL1 promoter activity. Top panel: experimental outline. Bottom panel: bioluminescence signal measured using a KRONOS (Atto) luminometer in AC16 cells transfected with the Bmal1-Luc(iferase) plasmid and treated with vehicle or digoxin (0.5 μM) (n=3). (D) CDKN1A/P21 protein level in AC16 cells. P21 level was assessed using the ProteinSimple Wes system in AC16 cells treated with vehicle or digoxin (0.5 μM). HSP90α was used as a protein loading control. All measurements were from distinct samples. *: P<0.05, **: P<0.01, ***: P<0.005 and ****: P<0.0001

Under basal conditions, expression levels of NR1D1 and of the REV-ERBα target gene ANRTL/BMAL1 showed, as expected, an antiphasic transcript expression in untreated AC16 cells (Fig. 2B). Digoxin increased at T₂₄ both NR1D1 and ANRTL/BMAL1 mRNAs, both known to be directly and cyclically repressed by REV-ERBα (Fig. 2B) and ANRTL/BMAL1 promoter activity became acyclic (Fig. 2C). P21 protein level increased upon digoxin treatment (Fig. 2D), alike after in vivo digoxin injection (Supp. Fig. 2E).

Digoxin’s effect was assessed using the human osteosarcoma U2OS cell line, a commonly used and robust circadian model (Supp. Fig. 6A-C). REV-ERBα expression peaked 16-20 hours after synchronization, and digoxin blunted REV-ERBα protein level at all time-points (Supp. Fig. 6A). The expression of RORα, a transactivator counteracting REV-ERBα activity, remained unchanged (Supp. Fig. 6A) whereasNR1D1 and BMAL1/ARNTL transcripts were higher expressed (Supp. Fig. 6B), with ANRTL/BMAL1 promoter activity becoming acyclic (Supp. Fig. 6C). Similar results were obtained using human primary cardiomyocytes (Supp. Fig. 6D,E). Thus CTS strongly decrease REV-ERBα protein level in several representative cardiomyocyte models and unrelated cell lines (U2OS, HepG2), suggesting a highly conserved mechanism involving NKA.

Digoxin induces multiple intracellular signaling pathways

Digoxin’s effect on REV-ERBα levels can be detected after a short treatment (210 minutes, Supp. Fig. 5H). Therefore, digoxin (at 5x EC₅₀=0.5μM) and other modulators were, in further experiments, added at T₁₈, thus 6 hours prior to the REV-ERBα protein peak in order to minimize secondary responses to treatment(s) (Fig. 3A, top panel). In these conditions, we first characterized the potential involvement of NKA as a bifunctional protein regulating ion fluxes and exerting signaling functions, as the calculated in vivo concentration of digoxin (5-10μM, Fig. 1D) has been reported to activate the signalosome-related PKCε in mouse heart ⁴¹. In vitro, 3,4,5,6 tetrahydroxyxanthone (THX) inhibits NKA ATPase activity with an efficiency similar as ouabain (EC₅₀=1.6μM), while being unable to activate Src ⁴². We therefore compared the activity of THX to that of digoxin on REV-ERBα stability. In contrast to digoxin, THX treatment did not alter REV-ERBα levels in AC16 (Fig. 3A and Supp. Fig. 7A) and U2OS cells (Supp. Fig. 7B), suggesting that the ion pumping activity of NKA might be dispensable for digoxin-induced REV-ERBα decrease. Indeed, pharmacological perturbation of calcium fluxes through inhibition of voltage-dependent Ca⁺⁺ channels by diltiazem and verapamil (Supp. Fig. 7C), through SERCA inhibition by thapsigargin (Supp. Fig. 7D) or by blocking the Na⁺-Ca⁺⁺ exchange system NCX by KB-R7943 (Supp. Fig. 7E) did not regulate REV-ERBα cellular protein levels. Taken together, this data rules out a contribution of intracellular calcium, hence of the inotropic arm of NKA.

Fig. 3. Intracellular pathways activation by digoxin.

(A) NKA ion pumping vs signalosome activities. Top panel: experimental outline. Synchronized AC16 cells were treated as indicated for 6 hours at T18 and harvested at T24. Bottom panel: REV-ERBα protein levels in control and treated cells. HSP90α was used as a protein loading control. Right panel: normalized REV-ERBα protein level. Results are shown as mean+/-SEM (n=10, see also Supp. Fig. 7) which were compared using a one-way ANOVA test, followed by a Tukey’s multiple comparison test. (B,C) Kinase activation trees in AC16 cells generated by KinMap using kinase profiling data at low (B) or high (C) digoxin treatment concentration. Circle diameters are proportional to the level of activity of kinases. (D) Heat map quantifying dose-dependent changes in kinase activities (n=3; 1-way ANOVA, *: P<0.05). MFS: median final score. (E) Transcriptomic analysis in vehicle- or digoxin-treated (0.5μM, 5μM) AC16 cells. The heatmap shows the top-ranking 25 genes and indicates dose-dependent changes in mRNA expression. Stars indicate genes whose expression is also altered in mouse heart after digoxin treatment at ZT9. (F) Gene set enrichment analysis of upregulated genes in digoxin-treated AC16 cells.

The ability of digoxin to trigger intracellular pathways at low and moderate concentrations (0.5μM and 5.0μM) was thus evaluated using serine/threonine (S/T) kinase arrays (Fig. 3B-D). Digoxin activated multiple S/T kinases belonging mostly to the human CMGC (Cyclin-dependent kinases, Mitogen-activated protein kinases, Glycogen synthase kinases, CDC-like kinases), CAMK (Calcium and Calmodulin-regulated Kinases) and the related AGC (PKA, PKG, PKCs) groups (Fig. 3B-D). The IκB kinase complex was also detected as being significantly activated. Activation of representative kinases of the CMGC (ERK1&2) and the AGC (AKT1&2) groups was validated by an immunoblotting approach (Supp. Fig. 8A &B). To disentangle downstream effects of digoxin-mediated signalosome activation, a genome-wide transcriptomic analysis of AC16 cells incubated with digoxin was performed. As reported in other systems ⁴³, digoxin increased the expression of immediate-early genes (IEGs) such as EGR-1, c-FOS and c-JUN both at low and high concentrations (Fig. 3E and Supp. Table 3). Gene set enrichment analysis of upregulated genes identified the NFκB pathway as most robustly induced in these conditions (Fig. 3F), in line with the observed phosphorylation of IkB kinase subunits (Fig. 3B, 3C). Monitoring gene expression in digoxin-exposed vs control mouse ZT9 hearts (Supp. Table 4) showed an induction of Egr-1, c-Fos and c-jun and other immediate early genes, with a 31% (12/38) overlap (Fig. 3E, see *) between in vitro, digoxin-upregulated genes in AC16 cells and in vivo digoxin-upregulated genes in mouse heart. Of note, biological term enrichment analysis identified gene-associated molecular functions such as “response to growth factor” and “regulation of ERK1&2 cascade”, in line with the observed in vitro activation of the NKA signalosome.

Digoxin has been ascribed multiple effects owing to its ability to modulate the activity and/or expression of several proteins: it modulates transcriptional activity of the nuclear receptor RORγ ^44,45, and interacts with PKM2 ⁴⁶, which is a target gene of HIF1α whose translation is inhibited by digoxin ⁴⁷. Modulation of these pathways in AC16 cells by either a RORγ inverse agonist (ML209, Supp. Fig. 9A), knocking down PKM2 (Supp. Fig. 9B) or through HIF1α overexpression (Supp. Fig. 9C) or activation (Supp. Fig. 9D) did not blunt or exacerbate digoxin-induced REV-ERBα degradation. Thus, these digoxin targets can be ruled out as effectors of digoxin action on REV-ERBα.

Taken together, our observations show that digoxin induces multiple signaling pathways, in murine cardiac tissue and cell lines, without engaging previously identified CTS targets. The causal link between digoxin-mediated induction of intracellular signaling pathways and altered REV-ERBα protein levels was thus investigated.

The protein kinase SRC contributes to REV-ERBα stability

As a first step, we pharmacologically modulated the known digoxin-sensitive SRC tyrosine kinase and the top-ranking digoxin-activated serine/threonine kinases identified above. Accordingly, we employed, in conjunction or not with digoxin, widely-used chemical modulators for SRC (PP2), PI₃K (ZSTK474), ERK1&2 (SCH772984), AKTs (MK2206), CAMK1&2 (KN93), MEK1&2 (PD98059), PKG1&2 (RP8-pCPT-cGMPS), PKD (CRT0066101), p70S6K1 (PF-4708671), RSK1&2&3 (BRD7389) and for components of the NFκB pathway such as LUBAC (Bay11-7082), HOIP (HOIPIN-8), IKKα/β (BMS345541), IKKβ (TCPA-1), IKKε and TBK1 (amlexanox) (Fig. 4 and Supp. Fig. 10). This pharmacological approach proved highly efficient at inhibiting kinase activity (see Supp. Fig. 8C as an example). The results were subsequently validated for a number of targets using either overexpressed constitutively active kinase mutants, other inhibitors or siRNAs (Supp. Fig. 11,12).

Fig. 4. Target screening with kinase specific and pathway specific inhibitors.

REV-ERBα protein levels were quantified after a 6 hour-treatment (at T18) with or without digoxin and with or without enzyme inhibitors: (A) PP2 (20μM, Src kinase inhibitor), (B) ZSTK474 (10μM, PI3 kinase inhibitor), (C) SCH772984 (20μM, ERK 1&2 inhibitor), (D) MK2206 (1μM, AKT1/2/3 kinase inhibitor), (E) KN93 (1μM, CAMK 2&4 inhibitor), (F) PD98059 (20μM, MEK 1&2 inhibitor), (G) RP-8-pCPT-cGMPS (20μM, PKG1&2), (H) CRT0066101 (5μM, PKD inhibitor), (I) PF-4708671 (10μM, P70S6K1 inhibitor), (J) BAY11-7082 (10μM, LUBAC and UPS inhibitor), (K) HOIPIN8 (10μM, LUBAC-HOIP specific inhibitor), (L) BMS-345541 (10μM, IKK α/β inhibitor), (M) TPCA-1 (20nM, IKKβ (N) Amlexanox (20μM, IKKε & TBK1 inhibitor). HSP90α was used as a protein loading control. Results are shown as mean+/-SEM (n=3-6) which were compared using a one-way ANOVA test, followed by a Tukey’s multiple comparison test. All measurements were from distinct samples. *: P<0.05, **: P<0.01, ***: P<0.001, ****: P<0.0001.

Testing the initial panel of inhibitors revealed stabilizing effects on basal REV-ERBα levels when inhibiting PI3K, ERKs or MEKs (Fig. 4B,C). However, this was not confirmed using distinct inhibitors or siRNAs with the exception of PI3K (Supp. Fig. 11 B,C). The latter effect was however not detected in U2OS cells (Supp. Fig. 13A). Taken together, these data show that functional interference with this set of kinases did not significantly affect REV-ERBα stability in basal conditions.

Blunting ERKs, PI3K, AKTs, CAMKs, MEKs, PKGs, PKD1, P70S6K1, RSKs, IKKs or TBK activity did not impair digoxin’s ability to trigger REV-ERBα degradation, a conclusion validated through the use of alternative approaches (Supp. Fig. 11, 12). In contrast, inhibition of SRC (with PP2, Fig. 4A and with saracatinib, Supp. Fig. 11A, 12A) blunted digoxin-induced REV-ERBα protein decrease, an effect which, albeit less pronounced, was also observed in U2OS cells (Supp. Fig. 13A). This data was nevertheless suggestive of the involvement of SRC in the chain of events leading to decreased REV-ERBα levels.

Bay 11-7082, an inhibitor of NFκB activation process, efficiently stabilized REV-ERBα in non-challenged cells and significantly protected REV-ERBα from digoxin-induced decreased expression in AC16 (Fig. 4J and Supp. Fig. 10J) and U2OS cells (Supp. Fig. 13B). Initially described as a linear ubiquitin chain assembly complex (LUBAC) inhibitor, this compound is also able to significantly inhibit several E2 ligases and the proteasome itself ⁴⁸. The use of a more specific inhibitor of the LUBAC complex, HOIPIN-8 ⁴⁹(Fig. 4K and Supp. Fig. 10K), of a HOIL-targeting siRNA (Supp. Fig. 11J) and the overexpression of the OUT domain-containing deubiquitinase with linear linkage specificity (OTULIN, Supp. Fig. 11K) did not show significant effects in our assay, thereby excluding the LUBAC complex as contributing to the REV-ERBα degradation process.

Since the NFκB pathway was significantly activated upon digoxin treatment, we addressed more directly the potential involvement of this transcription factor in several ways. We first assessed by mRNA analysis of cardiac tissues from Nr1d1^-/- whether a transcriptomic blueprint of the activated NFκB pathway could be observed. A comparative analysis of Nr1d1^-/- vs Nr1d1^+/+ heart transcriptome at ZT12 evidenced, much like the ZT0 vs ZT12 comparison (Fig. 1B, C), strong dysregulation of clock-related processes without any significant enrichment in NFκB-driven biological processes (Supp. Fig. 14A). Based on aggregated ChIP-seq studies, up- and downregulated genes were searched against the EnrichR ChEA database to identify transcription factors known to bind in the vicinity of their promoters ⁵⁰. Such an analysis did not identify NFκB as a potential regulator of dysregulated genes (Supp. Fig. 14B). A functional approach using either overexpressed inhibitory wild type or super repressor (SR) IκBα ⁵¹ did not protect or enhance digoxin’s effect on REV-ERBα stability (Supp. Fig. 14C). Finally, the reported pharmacological NFκB activator betulinic acid ⁵² did not prevent the effect of digoxin on REV-ERBα protein levels, definitively ruling out this pathway as a potential player in the observed REV-ERBα degradation (Supp. Fig. 14D).

Taken together, our results thus show that SRC kinase, described as part of the digoxin-controlled “NKA signalosome”, contributes to the observed decrease in REV-ERBα protein levels in AC16 and U2OS cells. Bay 11-7082, which inhibits IκBα, E2 ligases Ubc13 and UbcH7, the E3 ligase LUBAC and the proteasome ⁶¹ exerts a potent activity in our system, the effect of which being unrelated to the inhibition of linear ubiquitinylation by the LUBAC complex. These results rather point to the involvement of UPS in REV-ERBα degradation.

Digoxin induces the proteasomal degradation of heme-bound REV-ERBα

The UPS and autophagy are 2 main branches of the protein quality control pathway important for cellular homeostasis. Preliminary investigations ruled out other branches relying on post-translational modifications such as neddylation (using the NEDD8 inhibitor MLN4924/pevonedistat) and sumoylation (using the SUMO inhibitor 2-D08) as controlling REV-ERBα stability (data not shown). Protein arrays assaying the expression of components of the autophagy response did not show a significant modulation of this pathway by digoxin (Supp. Fig. 15A), in contrast to several reports on digoxin’s effect on cancerous cells ^36,53. Accordingly, pharmacological modulation of autophagy by either chloroquine, bafilomycin or rapamycin did not highlight a significant contribution of this process to the observed digoxin-induced effects (Supp. Fig. 15B-D). Knocking down P62/SQSTM1 expression, which was the sole component of the autophagy pathway upregulated by digoxin (Supp. Fig. 15A), also had no effect (data not shown).

We thus explored if the UPS is functionally implicated in the digoxin-induced REV-ERBα degradation. The proteasome catalytic subunit beta 5 (PSBM5) inhibitors bortezomib (BTZ) and clasto-lactacystin β-lactone, albeit to a lesser extent, protected REV-ERBα from digoxin-induced degradation (Fig. 5A, B and Supp. Fig. 16A, B). Protection by BTZ was also observed in U2OS cells (Fig. 5C and Supp. Fig. 16C), suggesting that UPS is a conserved component of the REV-ERBα degradation pathway. As exogenously-expressed wild type REV-ERBα is equally sensitive to digoxin treatment (Fig. 5D and Supp. Fig. 16D), we assessed whether REV-ERBα undergoes ubiquitinylation and is modified upon digoxin treatment using overexpressed Flag-REV-ERBα in AC16 cells. Mono-, multi- and poly-ubiquitinylated proteins were isolated from whole cell extracts and REV-ERBα content was analyzed by western blot analysis. The REV-ERBα protein was eluted along with other mono- and poly-ubiquitinated proteins, and its concentration decreased in digoxin-exposed cellular extracts, and increased in BTZ and digoxin-treated cells (Fig. 5E), indicative of a digoxin-controlled REV-ERBα ubiquitinylation. The H602F REV-ERBα mutant, which is unable to bind its natural ligand heme ⁵⁴, showed higher expression in basal conditions (as previously reported ⁵⁵) and was digoxin-resistant (Fig. 5D and Supp. Fig. 16D), suggesting that heme binding to REV-ERBα promotes a conformation required for proteasomal degradation.

Fig. 5. Digoxin triggers REV-ERBα protein degradation through UPS.

REV-ERBα protein levels were quantified after a 6 hour-treatment with or without digoxin and with or without the following proteasome inhibitors in synchronized AC16 cells (A) clasto-lactacystin β-lactone (10μM, proteasomal inhibitor), (B) bortezomib (BTZ, 200nM, PSMB5 inhibitor) and in synchronized U2OS cells (C) treated with either vehicle, digoxin (0.5μM) and/or bortezomib (200nM, PSMB5 inhibitor. Results are shown as mean+/-SEM (n=3-6) which were compared using a one-way ANOVA test, followed by a Tukey’s multiple comparison test. (D) Normalized REV-ERBα protein level was assessed in synchronized AC16 cells transfected with Flag-Rev-ERBα or Flag-Rev-ERBα-H602F-encoding plasmids and treated with vehicle or digoxin (0.5μM). HSP90α was used as a protein loading control Results are shown as mean+/-SEM (n=5) which were compared using a one-way ANOVA test, followed by a Tukey’s multiple comparison test. (E) Western blot analysis of total ubiquitinylated proteins and of ubiquitinated REV-ERBα protein from cell lysates of U20S cells transfected with Flag-Rev-ERBα plasmid and treated with either vehicle, digoxin (0.5μM) and/or bortezomib (BTZ, 200nM, PSMB5 inhibitor) for 2 hours (T18 to T20). Mono- and poly-ubiquitinylated proteins in the lysate were enriched by passing it through a High binding affinity UBI-QAPTURE-Q® matrix. All measurements were from distinct samples. *: P<0.05, **: P<0.01, ***: P<0.001, ****: P<0.0001.

To determine which component(s) of the UPS pathway could be involved in REV-ERBα degradation, we isolated REV-ERBα interacting partners by a RIME (rapid immunoprecipitation followed by mass spectrometry identification of endogenous proteins) assay ⁵⁶ using the digoxin-sensitive U2OS and HepG2 cells. We then filtered the generated list of interactors against the Gene Ontology “cellular component” or “molecular functions” databases using “proteasome” or “ubiquitin” as biological terms and interacting significantly with REV-ERBα (Supp. Table 6). We identified 16 UPS-related proteins interacting with REV-ERBα, including several E2 (UBE2L3) and E3 (PRPF19, TRIM21) ubiquitin ligases, the NCoR complex component TBL1XR1, and also 4 subunits of the proteasome complex (PSMA4, PSMA5, PSME3, PSMD2) (Supp. Fig. 17A). To this list of proteins potentially funneling REV-ERBα towards proteasomal degradation, we added a number of E2 and E3/E4 ubiquitin ligases previously reported to regulate REV-ERBα protein stability (SIAH2 ⁵⁷, FBWX7 ⁵⁸, HUWE1 ⁵⁹) or predicted to act on REV-ERBα using the proteome-wide E3 ligases-substrate interaction network (UbiBrowser ⁶⁰) such as BRCA1, CBL, MDM2, STUB1, TRIM33 and the ubiquitination factors UBE4A and UBE4B. The final list totaled 28 proteins (Supp. Fig. 17B), out of which 19 were tested for their involvement in REV-ERBα degradation. AC16 cells were thus transfected with siRNAs or treated with pharmacological inhibitors when available. Results are summarized in Supp. Fig. 17B from the quantification (Supp. Fig. 18) of raw data (Supp. Fig. 19). NEDD8, PSME3, the FBXW7-dependent destabilizing kinase CDK1 ⁵⁸, the REV-ERBα stabilizing kinase GSK3β ⁶¹, BRCA1, UBE4A, MDM2, STUB1 and TRIM33 were inactive in our assay (Supp. Fig. 18). Six entities significantly contributed to increased REV-ERBα stability in basal conditions including the REV-ERBα-interacting, NCoR complex component TBL1XR1 and the E3 ligases SIAH2, HUWE1, FBXW7, CBL and UBE4B. Quite unexpectedly, the impaired expression of the E2 ligase-encoding gene UBE2L3/UBCH7 decreased basal REV-ERBα expression (Supp. Fig. 18D). However, none of these E3 ligases involved in basal REV-ERBα proteostasis modulated the digoxin-mediated REV-ERBα degradation process, with the exception of SIAH2, whose knockdown partially alleviated REV-ERBα degradation (Supp. Fig. 18A). Taken together, these data suggest that the digoxin-mediated activation of the UPS triggers REV-ERBα degradation through a complex chain of events yet to be identified, but including the E3 ligase SIAH2 and the proteasome subunit PSMB5 (Fig. 6).

Figure 6. Hypothetical scheme for digoxin action in cardiomyocytes.

Discussion

In the current era of precision medicine wherein therapies are formulated and administered to attain maximal benefit through increased efficiency and reduced side effects, consideration of chronotherapy is an important step further towards this goal. As an edifying example, inclusion of chrono-chemotherapy has been proven beneficial in combating cancer by using a circadian window of higher drug tolerability and efficiency ^62,63. This concept has been recently extended to hypertension treatment at bedtime, which proved more efficient to diminish CVD events ⁶⁴.

The unevenly distributed rate of occurrence of cardiac events suggest a differential sensitivity of the heart to injury along the 24h day-night period ¹⁴. ToD dependent functional variations have also been reported in heart physiology, which are partly reflected at the transcriptional level by the significantly oscillating expression of approximately 13% of cardiac genes, which are involved in multiple cellular functions ranging from signaling and growth to transcription and cardiac remodeling ^65,66. Similarly, ToD-dependent proteomic analysis established striking differences in night versus day protein level patterns in the heart ⁶⁷, suggesting that timed cardiac proteostasis is crucial to proper heart function. Along these lines, we reported a strong correlation between the time of surgery and perioperative myocardial tolerance to injury in patients undergoing aortic valve replacement. This ToD-dependent differential myocardial sensitivity to IRI was attributed to the rhythmic expression of the nuclear receptor REV-ERBα and of its anti-apoptotic target gene P21 ²⁰. We also suggested the therapeutic potential of pharmacological inhibition of REV-ERBα for prophylactic cardioprotection, providing an alternative to physical ischemic preconditioning (PC) in which multiple distal, brief ischemic episodes induce cardioprotective signaling pathways that bestow protection in a future event of prolonged ischemia. Identified as the reperfusion injury salvage kinase (RISK) and the survivor factor enhancement (SAFE) pathways, these signaling pathways encompass multiple components, including apoptotic signals, which are potential targets for pharmacologically-induced PC ^68,69.

With the realization that NKA, the primary target of digoxin and other CTS, controls the activation of multiple signaling pathways ^70,71 including part of the RISK and SAFE pathways such as AKT, PKC and tyrosine kinases (Fig. 3), the use of cardiac glycosides as PC agents has been proposed ³⁰. Cardioprotection is observed in rats dosed with a transient sub-ionotropic dose of ouabain followed by washout similarly to IPC ^40,72. However, this strategy has not yet gained momentum, as digoxin use in heart failure and atrial fibrillation has stirred a 200-year long history of controversies and is still a matter of intense debate, leading to clinical hesitancy to use digoxin and other CTS in PC and other pathologies ^73–76.

Given the molecular connection of hypoxia-induced cellular damages with the molecular clock ^77–79, the potential usefulness of cardiac glycosides as myocardial protective agents and their debated usefulness, we investigated whether digoxin exert a ToD-dependent action on myocardial functional integrity. We observed a decreased susceptibility to IRI in mice dosed with digoxin when REV-ERBα expression reaches its maximum (ZT9), which could be potentially attributed to 2 principal effects of digoxin: (1) circadian rhythm alteration and/or (2) pharmacological cardiac preconditioning through the rapid and potent activation of many components of the RISK and SAFE pathway including PI3K, AKT, ERK1/2, GSK3β, and NFκB (Fig. 3). Genetic ablation of REV-ERBα showed that its expression is mandatory to observe digoxin protective effects. Digoxin treatment in mice relieved REV-ERBα repression on several genes including the cardioprotective Bmal1 and cdkn1a (Supp. Fig. 2 and Supp. Table 4). We showed that the anti-apoptotic P21 is not required for the cardioprotective effect of digoxin. The cardioprotective effect of BMAL1 has been elegantly documented through the use of genetic models (CCM and CBK mice ⁸⁰). The ZT12-digoxin phenotype is reminiscent of that of CCM mice, whose clock is locked at the wake-to-sleep transition (ZT0), display improved tolerance to MIRI at ZT12 and in which Bmal1 is similarly upregulated with dampened cyclicity ¹⁴. Identified BMAL1 target genes include Nampt or Coq10b which replenish cellular stores in cardioprotective NAD⁺ and ubiquinone respectively ⁸¹. However, digoxin-exposed hearts displayed no altered expression of these and other BMAL1 target genes (Supp. Table 4), an outcome likely due to the short time frame (4 hours) of the acute digoxin treatment. Thus, BMAL1 is unlikely to play a role in the observed phenomenon.

In contrast to these cardioprotective effects at ZT9, digoxin treatment of REV-ERBα-depleted hearts (ZT0) resulted in, albeit non-significant, trend to increased susceptibility to MIRI. As our transcriptomic studies did not point to obvious mechanisms, the molecular basis for this differential sensitivity to digoxin at ZT0 is unclear at this stage. Multiple parameters not assessible by gene expression studies may indeed control heart sensitivity to MIRI. Circadian variations of heart metabolism and temporal gating of signaling pathways may contribute to this time-specific response ⁸⁰, while NKA α1 expression is constant throughout the day-night cycle (Supp. Fig. 3 and ⁸²). Alternatively, the subcellular localization of NKA subunits, which is exquisitely regulated and dependent on protein-protein interactions, may vary and affect it signaling properties.

An unexpected, yet one of the most important findings in our study is the impact of digoxin on REV-ERBα protein levels in vivo and in human primary and immortalized cardiomyocytes. Exogenously overexpressed REV-ERBα was found to be equally sensitive to digoxin and ubiquitination assays using cells overexpressing REV-ERBα protein confirmed its ubiquitination. This differs from the reported mechanism of action of digoxin on the stability of the pro-angiogenic HIF1α, which was ascribed to a translational block occurring in the 5’ UTR of the Hif1a gene ⁴⁷. In contrast, the exogenously expressed ligand binding-crippled H602F REV-ERBα mutant was resistant to digoxin-induced degradation. Thus, heme-binding to REV-ERBα, which may induce specific conformational changes in the aporeceptor structure, seems to be an important prerequisite for physiological (^83,84 and Fig. 5D) and digoxin-triggered degradation through the proteasomal pathway. Interestingly, heme regulatory motif (HRM)- or PAS (PER-ARNT-Sim)-containing transcription factors such as BACH1, NPAS2, CLOCK, PER1 and PER2 are also subjected to heme-dependent protein degradation ^85–88, intimately linking heme synthesis and binding to protein stability and transcriptional regulation.

Our data thus clearly point to a beneficial effect of acute REV-ERBα antagonism in cardioprotection. This observation may seem at odds with investigations reporting improvement of failing heart 1-day post-MI following treatment with SR9009, a REV-ERB agonist ⁸⁹, which induces cardiac transcriptomic alterations ^90,91, and SR9009-induced blockade of impending cardiac remodeling through an anti-inflammatory mechanism ⁹². It should be kept in mind however that prolonged agonist treatment decreases, through a negative auto-feedback loop, REV-ERBα protein level ⁸⁹. Alternatively, it may also trigger systemic immune cell mobilization ⁹³, a process unlikely to be at play in our acute system.

CTS-controlled protein degradation is not unprecedented. Ouabain and digoxin triggers the degradation of the nuclear estrogen receptor ERα in primary and metastatic breast cancer cells, providing an additional mechanism for the reported CTS cytotoxicity against cancer cells ⁹⁴. Bufalin, which also induces REV-ERBα degradation (Supp. Fig. 5F), strongly impacts on p160 steroid receptor coactivator SRC1 and SRC3 protein stability ⁹⁵. Knockdown experiments however ruled out a contribution of SRC1 and 3 in our system (data not shown). The pleotropic effects of CTS thus call for a careful examination of their side-effects. However, their use in cardiac PC as a timed, single, low dose administration is an appealing therapeutic strategy to improve patient outcome, as unwanted effects of these FDA-approved drugs will be predictably limited. On the contrary, prolonged exposure to CTS in either CVD or cancer therapy might impinge on the molecular clock in several organs including the liver (Supp. Fig. 4), inducing metabolic alterations with unpredictable outcomes. Alternative approaches may be considered to target REV-ERBα for elimination. Indeed, our search to identify digoxin-controlled E3 ligase/ligases has proven to be difficult, but nevertheless confirmed or identified several E3 ligases controlling REV-ERBα protein stability in naïve conditions. Combining REV-ERBα ligands to an E3 recognition motif to generate a PROteolysis TArgeting Chimera (PROTAC) ⁹⁶ may emerge as an innovative strategy to lessen heart susceptibility to MIRI.

Materials And Methods

Reagents

All reagents, kits and chemicals references are described in the accompanying Appendix (see the Supplemental data section).

Animal Experimentation

All experiments were carried out following the guidelines set by the Comité d’Ethique en Expérimentation Animale du Nord-Pas de Calais CEEA75. To eliminate sex as a confounder, male mice were used throughout this study. C57BL6/J wild-type male mice (8-10 weeks) were purchased from Charles River Laboratories and housed in a temperature-controlled environment (23-25°C) with a 12h/12h light-dark cycle, ZT0 being lights-on. Mice had free access to water and were fed ad libitum a standard chow diet (Safe Diet A04). Cdkn1a/p21 ^-/-mice and their wild-type littermates were obtained from Charles River (France, B6.129S6(Cg)-Cdkn1atm1Led/J - SN 16565), Nr1d1^-/- mice were obtained from B. Vennström ⁹⁷ and were bred, housed and fed as above ⁹⁸.

To study the impact of digoxin in vivo, animals were injected intraperitoneally with digoxin or vehicle (10% ethanol, 40% propylene glycol, 0.08% citric acid, 0.3% sodium phosphate) at ZT5 or ZT20 at indicated concentrations (in most cases 1 mpk). The investigator was blinded to the treatment. Mice were sacrificed by cervical dislocation at ZT9 or ZT0. Organs were harvested and snap-frozen for protein and RNA analysis. Heart sensitivity to ischemia-reperfusion injury was tested using the ex vivo Langendorff perfused mouse heart model ⁹⁹. Isolated hearts were retrogradely perfused with a 5% CO₂-95% O₂-saturated modified Krebs–Henseleit (118.5 mM NaCl, 25 mM NaHCO₃, 4.7 mM KCl, 1.2 mM MgSO₄, 1.2 mM KH₂PO₄) buffer (1.4 mM Ca⁺⁺; 11 mM glucose) at a constant flow rate of 2.5 mL/min (Masterflex L/S peristaltic pump) at 37°C with an imposed pacing (9 Hz, 540 bpm). Hearts were stabilized for 20 min in these conditions then subjected to a normothermic, 30 min global ischemia by switching off the coronary perfusion. Hearts were then re-perfused for 30 min. Heart linear contractions were measured by passing a hook through the heart apex with a pre-load of 1.5 g and were recorded, and perfusion pressure with a Labtrax 4/24T (World Precision Instruments) device and integrated using the LabScribe data acquisition and analysis software (v 2.0). Ex vivo preconditioning was achieved by submitting hearts to an iterative (4 times) sequential ischemia-reperfusion sequence (2x 5 min)(see Fig. 1A).

Infarct sizes were measured using the 2,3,5-triphenyltetrazolium chloride (TTC) staining method. TTC (1%) was injected through the aortic canulae and incubated for 10 min at 37°C. Infarcted (white) and viable (red) areas were measured by computerized planimetry (Adobe Photoshop). Exclusion criteria for perfused hearts were a time to perfusion >3 min, a coronary perfusion pressure > 140 mm Hg or < 40 mm Hg or raising transiently above 180 mm Hg.

Cell lines and human primary ventricular cardiomyocytes

The AC16 human cardiomyocyte cell line ³⁵ was from Millipore (SCC109). Cells were maintained in DMEM/F12 Ham (Sigma-Aldrich) supplemented with 2mM L-glutamine (Thermo Fisher Scientific), 12.5% of fetal bovine serum (FBS) and 1% penicillin-streptomycin (Thermo Fisher Scientific).

The U2OS cell line was from ATCC (HTB-96) and maintained in McCoy 5A medium (Thermo Fisher Scientific) supplemented with 10% FBS and 1% penicillin–streptomycin.

Human primary ventricular cardiomyocytes were from PromoCell (C-12811). These cells were isolated from the ventricles of an explanted heart obtained from a male donor who underwent heart transplantation. Cells were tested and validated for cell specific markers, morphology, viability and adherence rate. Cells were maintained in Myocyte Growth Medium (C-22070, PromoCell) supplemented with Myocyte growth medium supplement mix (C-39270, PromoCell).

For all experiments, confluent cells were synchronized using 100nM dexamethasone (Sigma) for 2 hours, then washed in 1x PBS (Thermo Fisher Scientific). This defined T₀, after which cells were maintained in their respective growth media for further experimentation. Cell sampling and treatment were performed at indicated times. Cell viability was determined at indicated concentrations (see Fig. legends) using CellTiter-Glo® Luminescent Cell Viability Assay (Promega, G7570). DMSO was used at concentrations below 0.02% to avoid unwanted effects on REV-ERBα protein stability. Luciferase real-time recording was as described ³³, with the T₀-T₁₂ period being considered as a signal stabilization period.

All cell lines were routinely (once a month) monitored for mycoplasma contamination in our laboratory.

RNA extraction and RT-qPCR analysis

RNA extraction from cultured cells was carried out using the NucleoSpin® RNA Midi kit according to the manufacturer’s instruction (Macherey-Nagel). Total RNA from mouse heart was extracted from samples previously pulverized in liquid nitrogen using TRIzol™ Reagent according to the manufacturer’s instructions (Thermo Fisher Scientific). Prior to cDNA preparation, all total RNA extracts were subjected to DNase I treatment. The High Capacity cDNA Reverse Transcription kit (Thermo Fisher Scientific) was used to prepare cDNA. cDNAs were quantified using TaqMan Gene Expression Assays. Gene expression was normalized to 18S rRNA expression and relative gene expressions were calculated using the 2^-ΔΔCt method ¹⁰⁰. TaqMan probes used in this study are listed in the Appendix.

Protein extraction, Western blotting (WB) and Simple Western immunoassays

Total protein was extracted from cells using M-PER cell lysis buffer (Pierce) supplemented with Halt Protein Phosphatase and Protease cocktail (1:100 dilution, Pierce). Following a brief vortexing and centrifugation (30 min., 16,000 G), the supernatant was transferred to fresh tubes and protein concentration was determined using the BCA protein assay kit (Pierce).

Protein level was studied either by western blotting or by using the WES Simple Western protein detection system (Simple Western™) ¹⁰¹.

• Western blotting: 20μg of protein were resolved by 12% SDS-PAGE under reducing conditions and blotted onto a nitrocellulose membrane. The membrane was blocked in 5% BSA in 1x TRIS-buffered Saline-Tween (TBS-T) and then incubated overnight with the indicated antibody (see Fig. legends and Appendix for references). Membranes were washed 3 times in 1x TBS-T and then incubated with an anti-mouse HRP-labelled secondary antibody (1:10,000) for 1 hour. SuperSignal™ West Dura Extended Duration substrate (Thermo Fisher Scientific) was used to detect antigen-antibody complexes.

• WES analysis: 3.8 μg proteins were resolved by size by capillary electrophoresis. After immobilization onto the capillary by photoactivated capture chemistry, target proteins were identified using specific primary antibodies (see Appendix for references) and immunoprobed using an HRP-conjugated secondary antibody and a chemiluminescent substrate (ProteinSimple). Chemiluminescent signals were detected and quantified using the Compass for SW software (ProteinSimple).

Cloning and transfection

Transfections were performed 24 hours (T_-24) prior to dexamethasone-induced synchronization. The commercially available pEZ-M11-hREV-ERBα (Genecopoeia) plasmid (referred to as Flag-REV-ERBα) was used to overexpress flagged human REV-ERBα or its derivatives. Human flagged REV-ERBα mutant (hREV-ERBα-H602F) that is deficient in binding heme was generated in house from using the Flag- REV-ERBα plasmid by site-directed mutagenesis by substituting H602 with F ¹⁰².

The human Bmal1-luciferase construct (Bmal-Luc) suitable for real-time bioluminescence studies was obtained by cloning the Bmal1 promoter region from -350 to + 100 relative to the gene transcriptional start site into the SV40-driven, red-emitting luciferase pCBR reporter vector (Promega). Overexpression of wild type and constitutively active HIF1α was achieved by using HA-HIF1α-pcDNA3 (WT- HIF1α) and HA-HIF1α P402A/P564A-pcDNA3 (CA-HIF1α), respectively ¹⁰³. These plasmids were provided by W. Kaelin (Addgene plasmid # 18949, RRID: Addgene #18955). All plasmids used in this study are listed in the Appendix.

The Fugene HD® (Promega) transfection reagent was used to transfect AC16 cells. Transfection of U2OS was achieved using JetPEI ® (Polyplus Transfection). For immunoprecipitation experiments, 10μg pEZ-M11-REV-ERBα were transfected into 1.1x10⁷ AC16 or U2OS cells. Transfection efficiency was controlled by RT-qPCR, western blotting or WES analysis.

Small interfering RNA

Transfections were performed 24 hours (T_-24) prior to dexamethasone-induced synchronization. AC16 cells were transfected for 24 hours with 10nM siRNAs (Control: On-target plus non-targeting pool or On-target plus Smart pool siRNAs; see details in the Appendix) using Lipofectamine RNAiMAX as per the manufacturer’s instructions (Thermo Fisher Scientific). Silencing experiments in U2OS cells were carried out by transfecting 10nM siRNAs using INTERFERin® according to the manufacturer’s recommendation (Polyplus Transfection). Silencing efficiency was monitored by RT-qPCR or western blotting.

Immunoprecipitation and ubiquitination assay

Using the Crosslink Magnetic IP/Co-IP kit (Pierce), immunoprecipitation was carried out as described earlier ³³. For ubiquitination assays, 1.7x10⁷ cells were transfected with 10μg pEZ-M11-REV-ERBα plasmid. On the following day, cells were treated with 0.5μM digoxin (Sigma-Aldrich) with or without 1 μM clasto-lactacystin β-lactone (Sigma-Aldrich) or 10μM bortezomib (Sigma-Aldrich) for 24 hours to inhibit proteasomal activity. To enrich for ubiquitinylated proteins, the UBI-CAPTURE-Q® kit was used as per the manufacturer’s instruction (ENZO Life Sciences). Enriched ubiquitinylated proteins were resolved by 10% SDS-PAGE and were probed with anti-REV-ERBα antibody (Cell Signaling Tech.) or the anti-mono/polyubiquitinylated conjugates monoclonal antibody FK2 (ENZO Life Sciences).

PamGene phospho-protein array (PamGene Kinase Array)

• Sample preparation: Kinase activity profiles were determined in AC16 cells treated or not with digoxin. Briefly, AC16 cells were plated in 12-well fibronectin-gelatin coated plates (Sigma). Cells were synchronized using 100nM dexamethasone as described above. After a 20-hour recovery, cells were treated with 0.5 or 5μM digoxin. After 4 hours of incubation, cells were washed twice with 250 μL 1x PBS and lysed in 250 μL M-PER cell lysis buffer (Pierce) supplemented with Halt Protein Phosphatase and protease inhibitors (Pierce). Lysates were vortexed for 30 seconds and centrifugated (30 min., 16,000 G, 4°C). Protein concentration was determined with the BCA protein assay kit (Pierce) as recommended by the manufacturer.

• PamStation Kinomic analysis: Kinomic profiling of lysates was performed using a PamStation®12 (PamGene Intl) and Serine/Threonine Kinase (STK) PamChip® Arrays. Each STK PamChip® Array contains 144 peptides: 4 positive control phosphorylated peptides and 140 consensus phospho-peptide sequences representing the Ser/Thr kinome. These embedded peptides are 15 amino acids long of which 13 are derived from identified phosphorylation sites in human proteins. Briefly, PamChip arrays were blocked with 2% BSA (1-30 cycles). After the initial blocking step, 2 μg proteins from cell or tissue lysates premixed with 4μl of 10x PK buffer (PamGene Intl), 0.4 μl of 4mM ATP, 0.4 μl of 100x BSA, and 0.5 μl anti-STK primary antibody adjusted to a total volume of 40 μl in distilled water were added to individual arrays from a PamChip. The mix was then pumped through the PamChip® to facilitate interaction between active kinases in the sample and the embedded peptides on the array (30 to 90 cycles). After 30 cycles of pumping, arrays were washed in 1x PBS/0.01% Tween buffer following by 30 μl of PamChip detection mix containing 0.4 μl of FITC-labelled anti-STK antibody (1 μg/μl in TRIS buffer, PamgGene), 3μl of 10x Ab buffer (PamGene) and 26.6 μl H₂O was pumped through each array (2 cycles). Real-time fluorescence signal of phosphorylation was quantified by EVOLVE (PamGene Intl), a kinetic image capture program that captures ‘Pre-wash’ images starting from the 92nd cycle to the 122nd cycle and ‘after-wash’ image (124th cycle).

• BIONAVIGATOR- Image and data analysis of kinomic profiles: The BIONAVIGATOR software Suite (PamGene) was used to perform image analysis and log₂ transformation of signal intensities. The kinase upstream analysis algorithm within the BIONAVIGATOR Software Suite was employed to identify putatively active kinases. Based on substrate phosphorylation signals, the algorithm further estimates specificity and dependency power levels of each corresponding kinase. A final score is calculated by the software based on the scores obtained for specificity and significance. Kinases are ranked based on the mean final score (MFS). A heat map was plotted using MFS scores for corresponding kinases for low (0.5μM) and high (5μM) digoxin treatment groups. To visualize kinase activity changes in digoxin-treated versus untreated cells/tissue, kinase trees were generated using the KinHub platform (http://www.kinhub.org/kinmap/).

Microarray analysis, term enrichment analysis and network analysis

Gene expression from mouse hearts or AC16 cells was analyzed with Affymetrix GeneChip Hu (or Mo)Gene 2.0 ST arrays after RNA amplification, sscDNA labeling and purification. Briefly, RNA was amplified using the GeneChip™ WT PLUS Reagent Kit (Thermo Fisher Scientific), retrotranscribed to cDNA and labeled using GeneChip™ WT Terminal Labeling Kit (Thermo Fisher Scientific), followed by hybridization on the GeneChip Mouse or Human Gene 2.0 ST Array (Affymetrix) according to the manufacturer’s instructions. Raw data were processed further using GIANT (Galaxy-based interactive tools for Analysis of Transcriptomic data; GitHub (https://github.com/juliechevalier/GIANT). This user-friendly tool suite was developed in-house for microarray and RNA-seq differential data analysis ¹⁰⁴. It consists of modules allowing to perform quality control (QC), Robust Multi-Average method normalization, LIMMA differential analysis, volcano plot and heatmaps. Using the GSEA module, normalized data were converted toward a GSEA compatible format for further analysis.

Real-time measurement of bioluminescence

Real-time monitoring of bioluminescence was performed using a dish-type luminometer (Kronos AB-2550; ATTO). To achieve equal distribution of the Bmal-Luc signal in all dishes, 1.1x10⁷ AC16 or U2OS cells were first plated on P150 cell culture plates. When reaching 80% confluency, cells were transfected with the Bmal-Luc plasmid (10μg) using either Fugene HD® (Promega) or JetPEI® (Polyplus Transfection) for AC16 or U2OS cells, respectively. Twenty-four hours after transfection, cells were re-plated onto single 35mm cell culture dishes. Cells were synchronized with 100nM dexamethasone for 2 hours and the medium was replaced with AC16 or U2OS phenol red-free growth media. Cells were treated with vehicle or digoxin at indicated concentrations (see Fig. legends). Prior to starting light emission monitoring, 200 μM beetle luciferin (Promega) was added to cell culture dishes. Bioluminescence was measured for 1 min. at intervals of 10 min. under 5% CO₂ at 37°C for 4 days. The signal obtained was quantified and analyzed/detrended using the ATTO KRONOS software.

Rapid immunoprecipitation and mass spectrometry of endogenous proteins (RIME)

RIME assays were performed using U2OS overexpressing Flagged-hREV-ERBα or HepG2 cells. To overexpress REV-ERBα, cells were transfected with a Flag-REV-ERBα expression vector (GeneCopoeia) using JetPEI (Polyplus transfection). Twenty-four hours after transfection, the medium was replaced by fresh medium and cells were further incubated for 24h. Cells were synchronized for 2h with 100 nM dexamethasone, washed with 1x PBS and incubated for 20h with fresh complete media. For both experiments, cells were cross-linked for 10 min with 1% formaldehyde at 37°C in 1x DMEM. After quenching with 1.25 M glycine, cells were washed twice with ice-cold 1x PBS. Fixed cells were then collected into ice-cold 1x PBS supplemented with protease and phosphatase inhibitors (PIC, Roche). Cells were collected by centrifugation (1,000g, 10 min, 4°C) and lysed in 50 mM HEPES-KOH pH 7.5, 140 mM NaCl, 1 mM EDTA, 10% glycerol, 0.5% NP40, 0.25% Triton X-100 buffer supplemented with PIC. Cytosolic fractions were collected by centrifugation (1,000g, 10 min, 4°C) and pellets were washed with 10 mM TRIS-HCl pH8.0, 200 mM NaCl, 1 mM EDTA and 0.5 mM EGTA buffer containing PIC. After centrifugation, pellets were incubated for 10 min into 10 mM TRIS-HCl pH8.0, 200 mM NaCl, 1 mM EDTA and 0.5 mM EGTA, 0.1% sodium deoxycholate and 0.5% N-laurylsarcosine buffer (with PIC). After centrifugation, nuclei were re-suspended into 50 mM HEPES-KOH pH 7.5, 1 mM EDTA, 0.5 M LiCl₂, 0.7% sodium deoxycholate, 1% NP40 and 1x PIC. After a 10 min sonication, Triton X-100 (1% v:v final) was added and samples were centrifuged at 20,000g for 10 min. In parallel, SureBeads™ Protein A or G magnetic beads (Bio-Rad) were saturated with 5 mg/mL BSA. Four μg of rabbit monoclonal anti-REV-ERBα antibody (#13418, Cell Signalling Technology), 10 μg of mouse monoclonal anti-Flag antibody (#F1804, Sigma-Aldrich), 10 μg of control rabbit IgG (sc-2027, Santa Cruz Biotechnology) or mouse IgG (sc-2025, Santa Cruz Biotechnology) were immobilized on SureBeads™ Protein A (Rabbit) or G (Mouse) magnetic beads. Cytosolic or nuclear extracts were incubated overnight with immobilized antibodies at 4°C. Beads were washed with 1x RIPA buffer [50 mM HEPES-KOH, pH 7.6, 1 mM EDTA, 0.7% (v:v) sodium deoxycholate, 1% (v:v) NP-40 and 0.5 M LiCl] and then with 0.1 M ammonium hydrogen carbonate. Finally, beads were dried and flash-frozen in liquid nitrogen, and LC-MS/MS analysis was performed to identify protein partners as previously described ¹⁰².

Apoptosis and autophagy antibody arrays

To evaluate global changes in the apoptosis and autophagic pathways in digoxin-treated cells, the Human Apoptosis Array C1 (AAH-APO-1-2, Raybiotech) and the Autophagy Array C1 (AAH-ATG-1-2, Raybiotech) were employed. These arrays can simultaneously detect 43 and 20 human proteins regulating apoptosis and autophagy, respectively. Briefly, AC16 cells were plated in P150 cell culture plates. At confluence, cells were synchronized with 100nM dexamethasone as described above. Eighteen hours post synchronization, cells were treated either with 0.5μM digoxin or vehicle and lysed 6h later (T₂₄) in lysis buffer (Raybiotech). Antibody arrays were blocked using the blocking buffer (2 mL, Raybiotech) for 30 minutes and 250μg of fresh total protein lysate (1 mL) was added to each antibody array (1 sample/array). Arrays were incubated overnight at 4°C, then washed thoroughly in washing buffer (1 mL/wash, Raybiotech) and were further incubated overnight at 4°C with biotinylated antibody cocktail (1 mL, Raybiotech). Arrays were washed twice (1 mL/wash) then incubated in 1x HRP-Streptavidin Concentrate for 2h at room temperature. All blocking, washing and incubation steps were performed under gentle rotation (0.5-1 cycle/sec). After a 2h incubation, arrays were washed once followed by chemiluminescence detection using an Invitrogen™ iBright™ FL1500 Imaging System (Thermo Fisher Scientific). Spot intensities on each array were quantified using Image Studio™ Lite (LI-COR Biosciences). The RayBiotech Microsoft® Excel-based analysis software tool was used to average, normalize and subtract the background signal. Heatmaps were generated using GraphPad Prism (v. 9).

Statistics and reproducibility

Statistical analysis was performed using GraphPad Prism (v. 9). Microarray data analysis was carried out using GIANT (Galaxy-based interactive tools for Analysis of Transcriptomic data, available on GitHub (https://github.com/juliechevalier/GIANT). PamGene Kinase Array analysis was carried out the proprietary BIONAVIGATOR software suite (Pamgene https://pamgene.com/contact/).

Data are plotted as the mean ± SEM. At least 3 independent experimental replicates were obtained. In vitro data were determined to have equal variances using the F test. For 2-group comparisons, an unpaired 2-tailed t-test with Welch correction was used. For multiple comparisons with one variable, a 1-way ANOVA followed by the Tukey multiple comparison test (each group compared to every other group) was used. Multiple comparisons with more than one variable were carried out using a 2-way ANOVA followed by a Tukey’s multiple comparison test. Cyclical patterns of gene expression or of protein levels were determined using JTK_Cycle ¹⁰⁵. Individual expression/level data were plotted in Prism as a time series and non-linearly fitted using the “sine wave with nonzero baseline” function, using a wavelength constraint previously determined by JTK_cycle (in the 20-24 hours range). For infarct size measurements, based on an observed standard deviation of 8%, with α set at 0.05, β at 0.20 and a size effect of 10%, the group size was set at n=7. In vivo datasets were considered to have unequal variances and groups were compared using either a one-sided Welch’s t-test, or a Welch ANOVA followed by the Dunnett multiple comparison test unless mentioned otherwise. In all instances, P values < 0.05 were considered statistically significant.

Differential gene expression was analyzed after normalization of signals to the median of all samples, log₂ transformation and exclusion of the 10^th lowest percentile considered as technically unreliable. The Limma package, based on an Empirical Bayes method, is embedded into the Giant suite³⁷ and was used to identify genes exhibiting a fold-change > 1.2 with an FDR < 0.05. Biological term enrichment analysis was performed using the DAVID NCBI portal (v6.8) ¹⁰⁶.

Data availability

Microarray data that support the finding of this study has been deposited to GEO and is available using the accession number GSE183991. Other datasets have been described in ²⁰ and deposited using the accession number GSE62459. Whole mouse heart circadian gene expression was analyzed using the GSE180108 dataset ⁶⁵. All other data that support the finding of this study are available from the corresponding author on a reasonable request.