Cardiac glycosides cause cytotoxicity in human macrophages and ameliorate white adipose tissue homeostasis

Cardiac glycosides inhibit Na+/K+‐ATPase and are used to treat heart failure and arrhythmias. They can induce inflammasome activation and pyroptosis in macrophages, suggesting cytotoxicity, which remains to be elucidated in human tissues.

Cardiac glycosides cytotoxicity causes different forms of cell death (apoptosis or necrosis) in cells other than cardiomyocytes (Xiao et al., 2002) and in particular in metastatic cells (McConkey et al., 2000). Thus, cardiac glycosides have been proposed as an alternative therapy in cancer through immunogenic cell death (Menger et al., 2012) and apoptosis of senescent cells (Guerrero et al., 2019;Triana-Martinez et al., 2019).
One particular cell type that is particularly sensitive to intracellular potassium levels is the macrophage. Low concentration of potassium activates the NLRP3 inflammasome and causes the maturation and secretion of IL-1β (Petrilli et al., 2007). The decrease in intracellular K + can be induced by ATP and other damage-associated molecular patterns (Franchi et al., 2007;Munoz-Planillo et al., 2013;Petrilli et al., 2007), and K + efflux is mediated through membrane channels such as P2X7 receptors (Chen & Nunez, 2010) or TWIK2 (Di et al., 2018). Activation of P2X7 receptor also induces TNF release in macrophages (Barbera-Cremades et al., 2017), suggesting that consequences of a decrease in intracellular K + can be broader than NLRP3-mediated IL-1β production.
Here, we show that human monocyte-derived macrophages undergo cell death following incubation with nanomolar concentrations of cardiac glycosides and in particular with ouabain (50 nM < IC 50 < 100 nM). We show that non-adherent peripheral blood mononuclear cells do not show the same sensitivity to ouabain-induced cytotoxicity and the sugar moiety on the molecule is essential in driving cardiac glycoside-mediated cell death. We also report that ouabaininduced cell death is through the canonical activity of Na + /K + -ATPase. In a proof-of-principle study, we then used ouabain-mediated selective macrophage killing in a setting where these cells cause tissue damage and persistent inflammation and fibrosis. White adipose tissue explants from morbidly obese patients cultured with nanomolar concentrations of ouabain caused almost complete depletion of macrophages, decreased type VI collagen levels, while ameliorating insulin sensitivity ex vivo.
These results suggest that the usage of nanomolar concentration of cardiac glycosides can be an attractive therapeutic avenue in metabolic syndrome characterized by pathogenic infiltration and activation of macrophages in the omental white adipose tissue. and CD3 + positive; $75%) and non-haematopoietic (CD45 − ; $25%).
They were then treated with 50 nM, 100 nM, 500 nM and 5 μM of ouabain for 24 h before performing the cell viability assay.
Adherent peripheral blood mononuclear cell monolayer was washed twice with HBSS and monocytes were differentiated into human monocyte-derived macrophages for 5 days in RPMI 1640 containing 100 ngÁml −1 macrophage colony-stimulating factor (M-CSF, PeproTech) and 10% fetal calf serum. The macrophage purity was confirmed as previously described (Papathanassiu et al., 2017).  For white adipose tissue-derived mature adipocytes, viability was measured by adding alamarBlue (Thermo Fisher) directly into the culture (10%) for 4 h. Light absorbance was measured using Multiskan Ascent (Thermo Fisher). The percentage of reduced alamarBlue and cell viability were quantified following the manufacturer's instructions.

| Measurement of intracellular ions
Human monocyte-derived macrophages were incubated with Asante Potassium Green-2 AM (Abcam, 2 μM), CoroNa Green, AM (Invitrogen, 5 μM) and Fluo-4 AM (Invitrogen, 2 μM) probes for 30 min before fixation with 1% PFA for 4 h. Pictures were taken using an epifluorescent Olympus BX40 microscope equipped with a digital camera Retiga 2000R. A 5 × 5 square grid was overlayed to each picture and two cells per square were selected as region of interest.
Intracellular K + , Na + and Ca 2+ were measured from a total of 150 cells per condition from three independent experiments by quantifying the mean grey values using ImageJ software (ImageJ, RRID:SCR_003070).
Cells from all donors were assigned evenly across the experimental conditions. Mean grey values were normalized as relative to a control value, which was designated as "1".

| Western blotting
White adipose tissue explants were lysed in RIPA buffer (Thermo Fisher) supplemented with protease and phosphatase inhibitor cocktails (Thermo Fisher). Total proteins were quantified by BCA protein assay (Thermo

| Adipocyte mean areas
White adipose tissue explant adipocyte mean areas were measured on Sirius red-stained sections, using ImageJ Adiposoft plug-in (Galarraga et al., 2012). The scale was set using the scale bar and the minimum and maximum areas were set to 20 and 10,000 μm 2 , respectively. Pictures taken from each condition (4× magnification) were uploaded in an automatic directory processing mode and the adipocyte mean areas were calculated from a total of 8954 adipocyte areas across all the conditions.

| ELISA
Detection of human IL-1β and TNF-α (Invitrogen) in human monocyte-derived macrophages culture supernatants was performed by sandwich ELISA, using technical duplicates, following the manufacturer's instructions. Light absorbance was measured using Multiskan Ascent (Thermo Fisher) plate reader.

| Data and analysis
All the experiments that were conducted to draw the main conclusions of the paper use n = 6 or n = 5 healthy human donor or patients undergoing bariatric surgery. The occasional usage of n < 5 patients reflects a confirmation of the main message, using independent assays. Data that use n < 5 patients were labelled as "preliminary" in the figure legends.
Results are expressed as the mean ± SEM and were analysed using

| Cardiac glycosides induce cell death in human monocyte-derived macrophages
We first assessed cardiac glycoside cytotoxicity using annexin V/PI staining followed by flow cytometry analysis in human monocyte-derived macrophages. These cells were treated with either 0.01% DMSO (control) or cardiac glycosides (ouabain, digoxin, digitoxin and bufalin) at 50 nM, 100 nM, 500 nM and 5 μM for 24 h. Cardiac glycoside-treated human monocyte-derived macrophages showed reduced cell viability in a dose-dependent manner at nanomolar concentrations (Figures 1a and   S1A). Among the cardiac glycosides tested, ouabain was the most cytotoxic one (50 nM < IC 50 < 100 nM) in human monocyte-derived macrophages ( Figure 1a). In order to test for the cell specificity of this cytotoxic effect, we treated the non-adherent peripheral blood mononuclear cells with the same range of concentrations of ouabain for 24 h and found that ouabain did not exert cytotoxic effects in these cells (Figure 1b).
These results indicate that human monocyte-derived macrophages are particularly susceptible to the cytotoxic effects of nanomolar concentrations of cardiac glycosides when compared with non-adherent peripheral blood mononuclear cells.
It has been previously reported that ouabain, which contains a L-rhamnose unit ( Figure S1B), has higher affinity to the Na + /K + -ATPase than its aglycone ouabagenin, which lacks the Lrhamnose unit (Cornelius et al., 2013). Accordingly, ouabagenin was not cytotoxic when incubated for 24 h at 50, 100 and 500 nM in human monocyte-derived macrophages ( Figure S1B), suggesting that the cytotoxic effects of ouabain were caused by inhibition of Na + /K + -ATPase. To further test this hypothesis, we treated human monocytederived macrophages with either ouabain (50 nM In order to evaluate the potential cytotoxicity of ouabain in nonimmune resident cells such as adipocytes, human white adipose tissue-derived stromal cells were differentiated into pre-adipoytes and mature adipocytes and treated with ouabain (50, 100, 500 nM and 5 μM) for 24 h (Figure 2e,f). Ouabain cytotoxicity showed no differences among all the groups (Figure 2e,f), indicating that ouabain had no impact on pre-adipocyte and mature adipocyte viability.

| Ouabain improves insulin sensitivity and reduces type VI collagen levels in the white adipose tissue ex vivo
Macrophage infiltration is a hallmark of inflammation and insulin resistance in visceral white adipose tissue depots in mice, rats and humans (Koppaka et al., 2013;Olona et al., 2018;Wentworth et al., 2010).
Insulin sensitivity was assessed by measuring the levels of AKTser473 phosphorylation, a major effector of insulin signalling, in control and ouabain-treated white adipose tissue explants (Figure 3a). Ouabain treatment increased AKTser473 phosphorylation by $50% compared with control white adipose tissue explants (Figure 3b), indicating that ouabain promotes insulin sensitivity in the ex vivo cultured white adipose tissue. We then confirmed this finding by measuring ADIPOQ mRNA expression and CD36 protein levels by qRT-PCR and western blot, respectively. ADIPOQ encodes for adiponectin, an adipokine that promotes insulin sensitivity (Shetty et al., 2009) and was significantly up-regulated in ouabain-treated white adipose tissue explants compared with control ones (Figure 3c). Similarly, protein levels of CD36, which have an important role in fatty acid uptake upon insulin stimulation, were also increased in ouabain-treated white adipose tissue explants (Figure 3d). We next investigated whether increased CD36 and fatty acid uptake could affect adipocyte size, as previously reported (Vroegrijk et al., 2013). Adipocyte mean areas were $20% bigger in ouabain-treated compared with control white adipose tissue explants (Figure 3e). These results show that ouabain improves tissue function, promoting insulin sensitivity in white adipose tissue from obese patients.
Adipose tissue macrophages have been shown to regulate collagen deposition and fibrosis in visceral white adipose tissue during obesity, inducing tissue remodelling and insulin resistance (Keophiphath et al., 2009). In line with this, type VI collagen knockout mice upon high-fat diet showed unrestrained adipocyte expansion and improved white adipose tissue insulin sensitivity (Khan et al., 2009). Type VI collagen immunostaining was performed and mean grey intensities showed a down-regulation of type VI collagen in ouabain-treated compared with control white adipose tissue explants (Figure 3f). This coincided with COL6A1 mRNA expression down-regulation (Figure 3f), confirming a decrease in type VI collagen synthesis and deposition upon ouabain treatment in the white adipose tissue.

| DISCUSSION
Here, we show that nanomolar concentration of cardiac glycosides compromises human macrophage viability in vitro. In the white adipose tissue explants isolated from obese patients, ouabain treatment causes macrophage depletion. We also observe that cardiac glycoside-dependent macrophage cytotoxicity results in TNF secretion in vitro, which we could not detect in the ex vivo white adipose tissue (data not shown). TNF has been shown to induce apoptosis by activating directly caspase 3 or through the Bcl-2 family protein BH3 interacting-domain death agonist (known as BID) in macrophages (Liu et al., 2004). Collectively, these results show that cardiac glycosidemediated decrease in intracellular K + and the net increase in Ca 2+ levels trigger both pyroptotic (caspase 1-dependent IL-1β secretion) and apoptotic (cell death through caspase 3) cell death pathways in human macrophages.
Pyroptosis is caspase-1 dependent and causes the assembly of inflammasome complex, which initiates the efficient release of IL-1β (Gross et al., 2011). While pyroptosis is an inflammatory form of cell death triggered by caspase 1, apoptotic cell death is mediated by effector apoptotic caspases such as caspases 3, 6 and 7 (Man & Kanneganti, 2016). Recent evidence shows a significant crosstalk and compensation between apoptotic and pyroptotic cell death mechanisms and caspases such as caspase 6 having master regulatory role in both events (Zheng et al., 2020). Hence, it is likely that cardiac glycosides trigger broader caspase activation pathways in human macrophages, which causes IL-1β and TNF release, in line with reports showing TNF secretion following co-treatment with LPS and ATP (Barbera-Cremades et al., 2017;Di et al., 2018). Interestingly, TNF secretion has been previously shown to be caspase-1 dependent in macrophages (Miggin et al., 2007), which supports the broader effects of inflammasome activation on inducing pro-inflammatory cytokine secretion.
We report that the cell death caused by cardiac glycosides in human macrophages is dependent on K + flux and requires the presence of the sugar moiety attached to the steroid part of the compound. Indeed, the addition of KCl in the media showed a complete rescue of cell viability and intracellular levels of Na + and Ca 2+ . This also confirms that, similar to cardiomyocytes, macrophage survival is tightly dependent on potassium efflux and the sensitivity to cardiac glycosides is more prominent in human macrophage-like cells when compared with murine ones (LaRock et al., 2019). In macrophages, potassium efflux is mediated by membrane receptors such as P2X7 receptors and TWIK2 (Di et al., 2018), and the inhibition of Na + /K + -ATPase likely deprives the cell from potassium through the inhibition of its influx, while its export remains intact. Furthermore, our data suggest that, among other peripheral blood mononuclear cells, cardiac glycosides seem to have a particularly potent effect on macrophages.
This leads to the question of why human macrophages among other cells undergo apoptosis and pyroptosis as a result of decreased intracellular potassium? Interestingly, cardiac glycosides such as digoxin and digitoxin selectively induce apoptosis of senescent cells and digitoxin has a senolytic (i.e. selective killing of senescent cells) activity at a nanomolar range concentration that is closed to the one observed in cardiac patients treated with this drug (Guerrero et al., 2019;Lopez-Lazaro, 2007). This raises the possibility of phenotypic similarities between senescent cells and macrophages (Behmoaras & Gil, 2021), and a recent report shows that chemotherapy-induced senescent breast cancer cells are highly enriched for macrophage genes and can perform phagocytosis (Tonnessen-Murray et al., 2019).
It is therefore tempting to speculate that the macrophages and senescent cells show a shared sensitivity to the cardiac glycoside-mediated cell death, although the underlying mechanisms may differ.
When used at nanomolar concentrations, cardiac glycosides showed effective depletion of macrophages from ex vivo cultured white adipose tissue explants isolated from obese patients undergoing bariatric surgery. We observed an almost complete down-regulation of CD206 (mannose receptor). Indeed, adipose tissue-infiltrating macrophages are responsible for persistent lowgrade inflammation that underlies the systemic insulin resistance observed in obesity and CD206 is a specific marker for adipose tissue macrophages, which controls adipogenesis (Nawaz et al., 2017).
Despite being described as an M2 macrophage marker, CD206 has been previously found to be expressed by pro-inflammatory white adipose tissue macrophages in humans (Wentworth et al., 2010). In fact, the partial depletion of CD206 + macrophages enhances insulin sensitivity (Igarashi et al., 2018) and we show that this is indeed the case with the usage of ouabain (50-100 nM) in the ex vivo cultured white adipose tissue explants. This white adipose tissue macrophage depletion could be partly through caspase 1-mediated pyroptosis, though we cannot exclude apoptotic cell death. The mechanistic link between different caspases and the macrophage death remains to be identified and is a limitation of our study. Furthermore, although we could detect an increasing IL-1β upon ouabain treatment in the white adipose tissue, the overall detection range was low (<10 pgÁml −1 , data not shown) possibly due to relatively small number of macrophages in the explants when compared with the predominant adipose fraction.
We report an improved white adipose tissue function upon treatment with ouabain. Inflammation and fibrosis are two major pathways that dysregulate white adipose tissue homeostasis in obesity (Olona et al., 2018) and current therapeutic strategies aim to target these pathways (Kusminski et al., 2016). Our findings show that ouabainmediated macrophage cytotoxicity increases adipocyte hypertrophy and induces metabolic activity through up-regulation of CD36, which facilitates the uptake of long-chain fatty acids (Christiaens et al., 2012). Moreover, ouabain treatment reduces type VI collagen, the main collagen in the white adipose tissue, which accumulates following metabolically challenging conditions (Khan et al., 2009). Interestingly, macrophages have been shown to produce type VI collagen in the lung (Ucero et al., 2019), which suggests that ouabain-mediated macrophage depletion is the direct cause of the decreased levels of this collagen type in the white adipose tissue.
We show that stromal vascular fraction-derived cultured adipocytes and pre-adipocytes are not sensitive to ouabain-induced cell death and our data obtained in the peripheral blood suggest that macrophages are likely to be the only resident immune cells in the white adipose tissue that are cleared from the tissue through mechanisms that remain to be identified. Furthermore, the improvement of white adipose tissue homeostasis following ex vivo treatment with ouabain can be through soluble factors secreted by dying macrophages and/or direct effects of the drug on adipocytes. For instance, adipocytes treated with ouabain up-regulate Glut4, indicating a possible non-canonical role of this cardiac glycoside (Brewer et al., 2019).
The identification of all soluble factors accompanying ouabainmediated macrophage death and elucidating the pathways triggered by cardiac glycosides in adipocytes will be important in fully dissecting the mechanisms involved in the beneficial role of these compounds in inflamed tissues.
Here, we show the beneficial role of the usage of ouabain in a tissue where the increase in macrophage infiltrates is associated with tissue damage and inflammation. Notably, the white adipose tissue does not seem to act as a cardiac glycoside reservoir as pharmacokinetics of digoxin is identical between obese and non-obese patients (Abernethy et al., 1981). This suggests that local administration of cardiac glycosides could be the preferred therapeutic route for treating white adipose tissue inflammation in metabolic disease. In an infectious disease context, the use of cardiac glycosides can be detrimental because of the bactericidal role of mononuclear phagocytes (Esposito, 1985). Hence, the therapeutic advantage of nanomolar range usage of cardiac glycosides can be studied in sterile inflammation where tissue macrophage presence is generally associated with poor clinical outcome (e.g. inflammatory brain disorders and autoimmune disease). In keeping with this, the local administration of cardiac glycosides can have therapeutic effects through selective cytotoxicity towards tissue-resident macrophages, bypassing the unwanted side effects related to the reported toxicity of these compounds. Considering that some cardiac glycosides such as digoxin are frequently prescribed medicines in elderly population, these findings may also re-evaluate the usage of these in the context of tissue macrophage integrity in health and disease.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from the corresponding author upon reasonable request. Some data may not be made available because of privacy or ethical restrictions.