Ouabain–digoxin antagonism in rat arteries and neurones


  • Hong Song,

    1. Department of Physiology, University of Maryland School of Medicine, Baltimore, MD, USA
    2. Center for Heart, Hypertension and Kidney Disease, University of Maryland School of Medicine, Baltimore, MD, USA
    Search for more papers by this author
  • Eiji Karashima,

    1. Department of Physiology, University of Maryland School of Medicine, Baltimore, MD, USA
    2. Center for Heart, Hypertension and Kidney Disease, University of Maryland School of Medicine, Baltimore, MD, USA
    Search for more papers by this author
  • John M. Hamlyn,

    Corresponding author
    1. Department of Physiology, University of Maryland School of Medicine, Baltimore, MD, USA
    2. Center for Heart, Hypertension and Kidney Disease, University of Maryland School of Medicine, Baltimore, MD, USA
    Search for more papers by this author
  • Mordecai P. Blaustein

    Corresponding author
    1. Department of Physiology, University of Maryland School of Medicine, Baltimore, MD, USA
    2. Department of Medicine, University of Maryland School of Medicine, Baltimore, MD, USA
    3. Center for Heart, Hypertension and Kidney Disease, University of Maryland School of Medicine, Baltimore, MD, USA
    Search for more papers by this author


This article is corrected by:

  1. Errata: Corrigendum Volume 592, Issue 21, 4803, Article first published online: 31 October 2014

  • H. Song and E. Karashima were equal first authors and J. M. Hamlyn and M. P. Blaustein were equal second authors in this study.


Key points

  • ‘Classic’ cardiotonic steroids (CTSs) all inhibit Na+,K+-ATPase (Na+ pumps) and exert cardiotonic and vasotonic effects. Nevertheless, prolonged ouabain, but not digoxin, administration induces hypertension; moreover, digoxin antagonizes the hypertensinogenic effect of ouabain.
  • To examine acute ouabain–digoxin interactions, we tested these and related CTSs on myogenic tone (MT) in pressurized rat mesenteric small arteries and glutamate-evoked Ca2+ transients in primary cultured rat hippocampal neurones.
  • The CTSs (0.3–10 nm) all augmented MT at 70 mmHg and Ca2+ signals, but separated into two functional groups according to whether they were ouabain- or digoxin-like. CTSs within each group were synergistic, but between groups, were antagonistic to one another in both assays.
  • Na+ pump αβ protomers may function as tetraprotomers ((αβ)4) with quarter-site reactivity; simultaneous ouabain- and digoxin-like molecule binding promotes tetraprotomer disaggregation, enabling partial protomer reactivation.
  • These results may reveal why some patients respond poorly to digoxin therapy, and why Na+ pumps may be a novel target for therapeutic development.


‘Classic’ cardiotonic steroids (CTSs) such as digoxin and ouabain selectively inhibit Na+,K+-ATPase (the Na+ pump) and, via Na+/Ca2+ exchange (NCX), exert cardiotonic and vasotonic effects. CTS action is more complex than previously thought: prolonged subcutaneous administration of ouabain, but not digoxin, induces hypertension, and digoxin antagonizes ouabain's hypertensinogenic effect. We studied the acute interactions between CTSs in two indirect assays of Na+ pump function: myogenic tone (MT) in isolated, pressurized rat mesenteric small arteries, and Ca2+ signalling in primary cultured rat hippocampal neurones. The ‘classic’ CTSs (0.3–10 nm) behaved as ‘agonists’: all increased MT70 (MT at 70 mmHg) and augmented glutamate-evoked Ca2+ (Fura-2) signals. We then tested one CTS in the presence of another. Most CTSs could be divided into ouabain-like (ouabagenin, dihydroouabain (DHO), strophanthidin) or digoxin-like CTS (digoxigenin, digitoxin, bufalin). Within each group, the CTSs were synergistic, but ouabain-like and digoxin-like CTSs antagonized one another in both assays: For example, the ouabain-evoked (3 nm) increases in MT70 and neuronal Ca2+ signals were both greatly attenuated by the addition of 10 nm digoxin or 10 nm bufalin, and vice versa. Rostafuroxin (PST2238), a digoxigenin derivative that displaces 3H-ouabain from Na+,K+-ATPase, and attenuates some forms of hypertension, antagonized the effects of ouabain, but not digoxin. SEA0400, a Na+/Ca2+ exchanger (NCX) blocker, antagonized the effects of both ouabain and digoxin. CTSs bind to the α subunit of pump αβ protomers. Analysis of potential models suggests that, in vivo, Na+ pumps function as tetraprotomers ((αβ)4) in which the binding of a single CTS to one protomer blocks all pumping activity. The paradoxical ability of digoxin-like CTSs to reactivate the ouabain-inhibited complex can be explained by de-oligomerization of the tetrameric state. The interactions between these common CTSs may be of considerable therapeutic relevance.


blood pressure


cardiotonic steroids




effective concentration for a half-maximal effect


endogenous ouabain


endoplasmic reticulum






mass spectroscopy


myogenic tone in arteries pressurized to 70 mmHg


Na+/Ca2+ exchange(r)


passive diameter at 70 mmHg intralumenal pressure


plasma membrane




response units


surface plasmon resonance


sarco-/endoplasmic reticulum


The ‘classic’ cardiotonic steroids (CTSs), exemplified by ouabain, digoxin and bufalin, all have a steroid nucleus with a 5- or 6-member lactone ring at C17 (Fieser & Fieser, 1959; Hoch, 1961). They are all, as the class name implies, cardiotonic, i.e. they increase the force of cardiac contraction. They are also all Na+ pump (i.e. ATP-dependent Na+ and K+ transport or Na+,K+-ATPase) inhibitors (Blanco & Mercer, 1998), as discovered 60 years ago (Schatzmann, 1953). The minimal functional Na+ pump unit consists of paired α and β subunits, i.e. an αβ protomer (Vilsen et al. 1987; Martin & Sachs, 2000), in which β is a chaperone and the catalytic (α) subunit is the only well-documented CTS ‘receptor’ (O'Brien et al. 1994; Lingrel, 2010; Yatime et al. 2011).

The enigma of how Na+ pump inhibitors exert a cardiotonic effect was solved by the recognition that the Na+/Ca2+ exchanger (NCX) is the missing link (Baker et al. 1969). Na+ pump inhibition raises the intracellular Na+ concentration ([Na+]i). This, in turn, promotes NCX-mediated net gain of Ca2+, thereby enhancing contractile activation (Baker et al. 1969; Wier & Hess, 1984; Blaustein et al. 1998).

The fact that nearly all vertebrate cells express Na+ pumps with CTS receptors (dog red blood cells are a rare exception; Parker, 1973) fostered speculation that there must be an endogenous ligand for this receptor (Szent-Gyorgi, 1953). Indeed, skin glands in certain toads synthesize and secrete bufadienolide CTS (Bagrov et al. 2009). A CTS was purified from human plasma and identified, by mass spectroscopy (MS), as ‘ouabain’ (Hamlyn et al. 1991), a steroidal rhamnoside originally purified from the plants Strophanthus gratus and Akocanthera schimperi (Fieser & Fieser, 1959; Hoch, 1961). Endogenous ouabain (EO) was also purified and identified analytically by MS and nuclear magnetic resonance in bovine adrenals (Tamura et al. 1994; Schneider et al. 1998) and hypothalamus (Kawamura et al. 1999), and in rat plasma (Jacobs et al. 2012). EO is an adrenocortical hormone whose secretion is stimulated by ACTH, angiotensin II and catecholamines (Hamlyn et al. 1991; Laredo et al. 1994, 1997; Sophocleous et al. 2003).

Ouabain is widely used in biomedical research because it is a water-soluble, selective Na+ pump inhibitor. The hypothesis that an EO-like compound plays a role in the pathogenesis of salt-sensitive hypertension (Blaustein, 1977) spurred the discovery of EO (Hamlyn et al. 1982, 1991). The link to hypertension was fostered by the observation that prolonged subcutaneous administration of ouabain to normal rats induces hypertension (Yuan et al. 1993). This link was confirmed by: (i) the demonstration that ouabain- and ACTH-induced forms of hypertension are prevented by mutation of the α2 Na+ pump high ouabain affinity binding site (Dostanic-Larson et al. 2005; Dostanic et al. 2005; Lorenz et al. 2008); and (ii) evidence that plasma EO is significantly elevated in nearly 50% of patients with essential hypertension and a majority of patients with hypertension due to aldosterone-producing adenomas, and is correlated with blood pressure (BP; Rossi et al. 1995; Pierdomenico et al. 2001; Manunta et al. 2011). The inverse relationship between α2 Na+ pump expression and BP (Zhang et al. 2005; Pritchard et al. 2007) is also consistent with the clinical findings: both elevated ouabain/EO and reduced α2 expression should reduce the activity of the high ouabain affinity α2 Na+ pumps in arterial smooth muscle (Dostanic et al. 2005; Zhang et al. 2005; Linde et al. 2012). This should elevate the sub-plasma membrane Na+ concentration (sub-PM [Na+]) and promote NCX-mediated Ca2+ gain by the myocytes, and thereby enhance myogenic tone and raise BP (Zhang et al. 2005). Conversely, overexpression of α2 Na+ pumps might be expected to maintain a very low sub-PM [Na+] and, thereby, low [Ca2+] and reduced BP (Pritchard et al. 2007).

More surprising was the finding that digoxin and digitoxin, classic CTSs extracted from plants of the genus Digitalis, did not induce hypertension but, in fact, antagonized the hypertensinogenic effect of ouabain (Manunta et al. 1993, 2000). Conversely, other CTSs derived from plants of the genus Strophanthus, dihydro-ouabain (DHO) and ouabagenin, mimicked the effect of ouabain on BP (Manunta et al. 2001). The anti-hypertensinogenic effect of digoxin was confirmed in a salt-sensitive model of hypertension (Huang et al. 1999). Nevertheless, all of the aforementioned ‘ouabain-like’ and ‘digoxin-like’ agents are ‘classic CTSs’: i.e. they are Na+ pump inhibitors with roughly similar affinities for the CTS binding site (Yoda, 1973, 1974; Yoda et al. 1973), and they are cardiotonic (Movitt, 1949; Smith, 1986a).

A few publications describe antagonisms between CTSs. Ouabain–DHO antagonism was reported in the heart (Godfraind et al. 1982). Recently, ouabain was found to antagonize the digoxin- and bufalin-induced inhibition of membrane endocytosis in NT2 human neuronal precursor cells (Feldmann et al. 2007), and to attenuate digoxin- and bufalin-induced cardiotoxicity in vivo (Nesher et al. 2010).

The bufadienolide, marinobufagenin (MBG), has also been implicated in the pathogenesis of hypertension (Fedorova et al. 2002; Puschett et al. 2010). There is, however, no evidence for ouabain–MBG or digoxin–MBG antagonism.

Here we describe the antagonism between ‘ouabain-like’ and ‘digoxin-like’ compounds on myogenic constriction in rat small arteries, and on glutamate-induced Ca2+ signalling in primary cultured rat hippocampal neurones. We suggest that this antagonism occurs directly on the α2 and α3 Na+ pumps in arterial myocytes and neurones, respectively. This antagonism may help to explain why digoxin and digitoxin antagonize the vasopressor effect of ouabain, mentioned above, and why digoxin may be efficacious as an antihypertensive in some patients with essential hypertension (Abarquez, 1967). It may also be the reason why some congestive heart failure patients are more likely than others to benefit from digoxin therapy (Rathore et al. 2002).



Mesenteric arteries were obtained from Sprague–Dawley rats (150–200 g males; Charles River, Wilmington, MA, USA). The rats were killed by CO2 overdose followed by decapitation; the mesenteric arcade was rapidly dissected and placed in ‘artery dissection solution’ (5°C). Pregnant (18 days) Sprague–Dawley rat dams (Charles River) were killed by CO2 asphyxiation followed by thoracotomy. The E18 fetuses were then removed and killed by decapitation, and neurones were cultured from the hippocampi. All rat protocols were approved by the Institutional Animal Care and Use Committee of the University of Maryland School of Medicine.

Myogenic tone in rat mesenteric small arteries

Artery preparation

Segments of 4th order mesenteric arteries were dissected from the mesenteric artery arcade on the external surface of the small intestine. The arteries were cannulated at both ends, pressurized to 70 mmHg (with no internal flow), and superfused with physiological salt solution (PSS) at 35–37°C to permit myogenic tone (MT70) to develop. External diameter was continuously monitored with a custom on-line edge detection system based on a Nikon TMS microscope (Nikon, Melville, NY, USA). Images were analysed with customized Interactive Data Language software (IDL; Research Systems, Boulder, CO, USA). Passive diameter at 70 mmHg intralumenal pressure (PD70) was determined by superfusing the arteries with Ca2+-free PSS for 10 min at the end of each experiment. Thus, MT70 = [(diameter decrease from PD70 evoked by being pressurized to 70 mmHg in the absence of a CTS)/(PD70)] × 100. Arteries that developed significant leaks were discarded. Full details are published in Zhang et al. (2002, 2005).

Endothelium removal

To eliminate the complicating effects of CTSs on the endothelium (Raina et al. 2010) so that we could focus on the arterial myocytes, the endothelium was destroyed at the start of some experiments. To remove the endothelial cells, an air bubble was introduced into the cannula line by opening the distal stopcock and allowing air to enter the cannula attached to the intralumenal perfusion pump. The air bubble was then slowly pushed into the artery lumen, and the distal stopcock was closed to stop the solution flow. The air bubble was then pushed through the lumen by briefly re-opening the distal stopcock. Measurements were made after a 30 min waiting period. Arteries normally dilated to PD70 in response to 10 μm acetylcholine (ACh); thus, the adequacy of endothelium removal was judged by the lack of arterial dilatation to ACh (relaxation of no more than 10% of PD70). Perfusion with an air bubble was repeated as necessary to achieve this (Raina et al. 2010).

Effects of CTSs on myogenic constriction

CTS-evoked percentage increases in MT70 were calculated as the percentage increases in constriction compared to the control MT70, i.e. [(external diameter decrease from PD70 in the presence of a CTS)/(diameter decrease from PD evoked by pressurization to 70 mmHg in the absence of a CTS)] × 100. The influence of the addition of a second CTS, as a percentage, was then calculated as: [(the increase or decrease in diameter evoked by the second CTS)/(the decrease in diameter evoked by the first CTS)] × 100. Decreases in diameter are reported as ‘positive’ increases in MT; increases in diameter (i.e. vasodilatation) are reported as ‘negative’ increases in MT.

Ca2+ signalling in primary cultured rat hippocampal neurones

Neurones were generated in co-culture with astrocytes from rat E18 embryos and Ca2+ signals (with Fura-2 and digital imaging) were measured in cells identified as neurones, as described in detail in Song et al. (2013).

Use of mass spectrometry (MS) and surface plasmon resonance (SPR) to test for ouabain–digoxin aggregation

Cholesterol and several related steroids self-associate under some conditions (Strauch et al. 1969; Haberland & Reynolds, 1973; Chang & Cardinal, 1978). To test for the possibility that ouabain–digoxin antagonism might be due to the association of ouabain with digoxin (thereby reducing the effective concentrations of both agents), we measured the binding of ouabain to digoxin with MS and SPR technologies.

Mass spectrometry

Ouabain and digoxin were co-infused into a Bruker HCT Ultra ion trap mass spectrometer (Bruker Corp., Fremont, CA, USA) via the electrospray interface. The steroids were co-infused at 25–250 nm in a solution of 50% CH3CN when protonated molecular ions were desired. Alternatively, when lithiated molecular ions were monitored, the steroids were co-infused in the same solution containing 250 μm LiCO3. For further information, see the online Supporting information.

Surface plasmon resonance analysis

The binding of ouabain to digoxin, and vice versa, was also measured with SPR technology using a Biacore T200 instrument (Biacore Life Sciences, GE Healthcare, Piscataway, NJ, USA). Initially, the carbohydrate moieties of ouabain and digoxin were both oxidized as follows: ouabain (1 mmol) was dissolved in 200 ml of water. While stirring, 20 ml of 0.5 m NaIO4 in water was added drop-wise. The reaction was dried by vacuum centrifugation. The dried materials were reconstituted in 100 ml water, filtered (0.45 μm) and separated by C18 preparative scale HPLC using a CH3CN gradient. The elution of oxidized ouabain was monitored at 220 nm and was collected and dried. Digoxin (40 μmol) was dissolved in 5 ml 70% CH3CN in H2O. Then 2 ml of 0.5 m NaIO4 in water was added drop-wise while vortexing. The reaction was dried by vacuum centrifugation and the materials were reconstituted in 100 ml 4% CH3CN in water. The reconstituted material was applied to a preconditioned C18 column (20 g Mega-BE, Varian, Agilent Technologies, Santa Clara, CA, USA). The column was washed with 40 ml of water and 120 ml 2.5% CH3CN. Bound materials were eluted with 80 ml 70% CH3CN and dried by vacuum centrifugation. Both oxidized compounds were dissolved in DMSO for coupling to the Biacore chips.

Oxy-ouabain and oxy-digoxin were coupled to flow cells 2 and 3, respectively, of a Biacore CM5 sensor chip, to a total of ∼200 response units (RU). Coupling used the EHS kit (Biacore) to derivatize the surface with ethylamine, followed by incubation with either oxy-ouabain or oxy-digoxin and reduction of the resulting Schiff bases with NaBH4, as recommended by the manufacturer. Flow cell 1 of the chip was treated identically but was not conjugated to an oxidized CTS, for use as a control. Binding of ouabain and digoxin to the coupled surfaces was performed with solutions of 20 nm and 200 μm in PSS with 2% DMSO. Binding of ouabagenin was performed with a solution of 100 nm in PSS (no DMSO). Following injection into flow cells 1, 2 and 3, the association of the soluble CTS to the oxidized CTS on the chip was recorded by SPR with a Biacore T200 (Biacore, Inc.,). Between reactions, the chip surface was washed with PSS for 3 min. The signal from flow cell 1 was subtracted from the signal from flow cell 2 or 3 to correct for changes in refractive index, changes of the solution, injection noise and non-specific binding to the blank surface. A blank injection with buffer alone was subtracted from the resulting curves. As a positive control for the presence of active CTS on the chip surface, we used a monoclonal anti-ouabain antibody (Enzo Life Sciences, Uniondale, NY, USA), at 6 nm and 60 nm. Sensorgrams generated by the instrument were analysed using the Biacore T200 Evaluation software.

Solutions and reagents

Artery dissection solution contained (in mm): 145 NaCl, 4.7 KCl, 1.2 MgSO4.7H2O, 2.0 Mops, 0.02 EDTA, 1.2 NaH2PO4, 2.0 CaCl2.2H2O, 5.0 glucose and 2.0 pyruvate; 1% albumin was also added (pH 7.4 at 5°C). The PSS perfusion solution contained (in mm): 112 NaCl, 25.7 NaHCO3, 4.9 KCl, 2.5 CaCl2, 1.2 MgSO4.7H2O, 1.2 KH2PO4, 11.5 glucose and 10 Hepes (pH to 7.3–7.4). Ca2+-free PSS was made by omitting Ca2+ and adding 0.5 mm EGTA. Solutions were gassed with 5% O2–5% CO2–90% N2 at 35–37°C; the measured O2 level in the open artery chamber was ∼12%.

All reagents used for solutions were reagent grade. Ouabain, dihydro-ouabain, ouabagenin, digoxin, digitoxin, digoxigenin, bufalin, strophanthidin, proscillaridin, prazosin and 4-amino-5-(4-chlorophenyl)-7-(dimethylethyl)pyrazolo[3,4-d]pyrimidine (PP2) were purchased from Sigma-Aldrich (St Louis, MO, USA) or as indicated. MBG (purity and structure verified by MS) was a gift from Dr Alexei Y. Bagrov (National Institute on Aging, NIH, Baltimore, MD, USA). SEA0400 was a gift from Dr Takahiro Iwamoto (University of Fukuoka, Fukuoka, Japan). Rostafuroxin (PST2238; Quadri et al. 1997) was a gift from Dr Patrizia Ferrari (Prassis-Sigma Tau, Milan, Italy).

Statistical analyses

The data are expressed as means ± SEM; n denotes the number of arteries or number of neurones studied. Comparisons of data were made using ANOVA or Student's paired or unpaired t test, as appropriate. Differences were considered significant at P < 0.05.


Part 1. Studies on rat mesenteric small arteries

Nanomolar doses of ‘classic’ CTSs augment myogenic constriction

Previous studies revealed that nanomolar ouabain rapidly (within 1–3 min) and reversibly increases myogenic tone (MT70) and myogenic reactivity in rat and mouse mesenteric small arteries. The ouabain concentration that elicits a half-maximal effect (EC50) is ∼0.5–1.0 nm (Zhang et al. 2005; Raina et al. 2010). The augmented vasoconstriction (vasotonic effect) is associated with an increase in arterial myocyte cytoplasmic Ca2+ concentration ([Ca2+]CYT; Zhang et al. 2005). The enhancement of the Ca2+ signal and the vasotonic response are both antagonized by the NCX inhibitor SEA0400 (Iwamoto et al. 2004; Zhang et al. 2005), and by smooth muscle-specific knockout of NCX type 1 (NCX-1; Zhang et al. 2010). Moreover, ouabain and digoxin both augment vasoconstrictor-evoked Ca2+ signals in primary cultured rat mesenteric artery myocytes (Arnon et al. 2000).

Figure 1A–C shows that 10 nm doses of two cardenolides, the Digitalis steroid digoxin and the Strophanthus (ouabain-like) steroid strophanthidin, and the bufadienolide proscillaridin A, a Scilla toxin, reversibly increase MT70 in rat mesenteric small arteries pressurized to 70 mmHg. Similar results were obtained with a number of other Strophanthus, Digitalis and bufadienolide CTSs (Table 1). Previously, we also tested the synthetic steroid, rostafuroxin (Zhang et al. 2005), a derivative of digoxigenin (Quadri et al. 1997). In contrast to natural CTSs, rostafuroxin (5 μm), a ouabain antagonist that does not itself inhibit Na+ pumps (Ferrari et al. 1998; Ferrari, 2010), also does not augment MT70 (Zhang et al. 2005). Table 1 summarizes the results from a number of experiments comparable to the ones in Fig. 1A–C: it shows the increase in MT70 induced by a 10 min treatment with a 3 nm dose of each of the agents.

Figure 1.

Figure 1. Classic cardiotonic steroids (CTSs), at nanomolar concentrations, constrict pressurized (70 mmHg) rat small arteries

A–C, strophanthidin (Stroph, A), digoxin (Dig, B) and proscillaridin A (ProA, C), all at 10 nm concentrations, reversibly constrict rat mesenteric small arteries with myogenic tone (MT). Each panel shows a representative experiment. PD, passive external diameter measured in Ca2+-free PSS at the end of the experiment. D, in some experiments similar to those of panels A–C, cumulative CTS dose–constriction curves were measured. Summary CTS dose–response curves for five CTSs (Ouab, ouabain; Dig, digoxin; DTX, digitoxin; Stroph, strophanthidin; ProA, proscillaridin A); ordinate is the percentage increase in MT70 compared to the control MT70. Each curve shows the averaged data from 4–10 arteries; bars indicate ± SEM.

Table 1. Effects of several cardiotonic steroids on myogenic tone (MT70) in rat mesenteric small arteries
Cardiotonic steroid (CTS)Control MT70a (% of PD)MT70 + 3 nm CTSb (% of PD)CTS-induced ΔMT70c (% of control)
  1. aThe data in this table are from experiments similar to those illustrated in Fig. 1AC. Mesenteric small arteries were pressurized to 70 mmHg and myogenic tone (control MT70) was measured before adding a CTS. In the large majority of cases, each artery was from a different rat although, in a few instances, two arteries from a single rat were used. Mean passive external diameter (PD) was 241.7 ± 4.4 μm; mean control MT70 was 19.6 ± 0.9%; total n = 53 arteries. bMT70 in the presence of 3 nm CTS; number of arteries given in parentheses. **P < 0.002 and ***P < 0.001 vs. control MT70 (Student's paired t test). In all cases, except for MBG, control MT70 was the average of the ‘before CTS’ and ‘after CTS’ control, and experiments were accepted only if the arteries recovered after CTS washout. In the case of MBG, however, washout was very slow and incomplete; thus, only the ‘before CTS’ control was used. cPercentage increase in MT70 due to addition of 3 nm CTS.

Ouabain18.4 ± 0.825.0 ± 0.8 (7)***33.9 ± 8.4
Dihydro-ouabain (DHO)21.7 ± 1.030.0 ± 1.2 (5)***42.7 ± 8.9
Strophanthidin16.1 ± 0.921.9 ± 1.1 (5)***37.0 ± 6.2
Ouabagenin20.7 ± 0.925.0 ± 1.1 (5)***33.0 ± 16.1
Digoxin19.8 ± 0.826.1 ± 0.9 (6)***32.1 ± 3.8
Digitoxin20.5 ± 0.924.5 ± 1.0 (5)***34.1 ± 20.1
Digoxigenin21.4 ± 1.028.2 ± 1.2 (5)**37.2 ± 16.7
Bufalin20.6 ± 1.026.9 ± 1.2 (5)**40.4 ± 17.5
Proscillaridin19.0 ± 0.926.5 ± 1.1 (5)***41.4 ± 5.0
Marinobufagenin (MBG)19.3 ± 1.024.8 ± 1.2 (5)***29.3 ± 7.5

In some similar types of experiments, cumulative CTS dose–arterial diameter data were obtained. Summarized data from 4–10 such experiments for each of five different CTSs are presented in Fig. 1D. These results, and data for several other CTSs, indicate that 3–10 nm doses of these CTSs are near-maximally effective, and that the apparent EC50 values for all of these CTSs are in the range of 0.1–1.0 nm (Fig. 1D). In other words, all of these chemically distinct ‘classic CTSs’ are vasotonic and have fairly similar EC50 values in the small artery MT70 assay.

Ouabain–digoxin antagonism in small arteries

The antagonism of ouabain's chronic hypertensinogenic action by digoxin (Manunta et al. 1993, 2000, 2001; Huang et al. 1999) raised the possibility that ouabain and digoxin might interact at the CTS binding site on arterial smooth muscle α2 Na+ pumps. To test this possibility, we explored the effect of digoxin on the ouabain-induced augmentation of MT70 in rat mesenteric small arteries. As illustrated in Fig. 2Aa, when 3 nm ouabain increased MT70 in a pressurized artery, addition of 10 nm digoxin to the ouabain-containing superfusion medium reversibly reduced the ouabain-induced constriction. On average, 10 nm digoxin inhibited 3 nm ouabain-induced constriction by 65% (Fig. 2B and C), even though digoxin, by itself, constricts pressurized arteries about as much as does ouabain, and the two CTSs have similar EC50 values (Fig. 1, Table 1 and Raina et al. 2010). The antagonism of ouabain's effect by digoxin was dose dependent and appeared to be competitive; when the digoxin concentration was increased above 10 nm, the constriction again increased (Fig. 2C), presumably because the constriction by digoxin began to predominate. Moreover, when the sequence of additions was reversed, and the increase in MT70 was induced by 3 nm digoxin, 10 nm ouabain reversibly antagonized that constriction (Fig. 2Ab, and B).

Figure 2.

Figure 2. Antagonism of ouabain-induced vasoconstriction by digoxin, and of digoxin-induced vasoconstriction by ouabain

Aa, digoxin (Dig, 10 nm) reversibly reduces (arrow ‘b’) the constriction induced by 3 nm ouabain (Ouab; arrow ‘a’) in a rat mesenteric small artery with myogenic tone (MT). PD, passive external diameter measured in Ca2+-free PSS. Ab, the converse experiment to that shown in panel Aa: 3 nm ouabain reversibly antagonizes the constriction induced by 10 nm digoxin in an artery with myogenic tone (MT70). B, summarized data from 6 experiments, each with protocols similar to the ones in panels Aa and Ab. The ordinate values are the constrictions induced by ouabain or digoxin alone (‘a’), or the constrictions in the presence of both CTSs (a–b); *P < 0.05; **P < 0.01 for the indicated pairs. Mean passive diameter (PD) was 224 ± 12 μm; mean control MT70 was 18.0 ± 1.3% of PD (n = 12 arteries). C, digoxin log dose–response curve shows the effect of digoxin concentration on the constriction induced by 3 nm ouabain. The data are from panel A and similar experiments plotted as: (b/a) × 100% vs. digoxin concentration. The data points are mean values from 2–9 arteries. The bar above the graph indicates the approximate ranges of the therapeutic, intermediate and toxic concentrations of digoxin in humans (white, grey and black regions, respectively; Smith, 1986b; Rathore et al. 2002).

Both ouabain and digoxin, individually, increased MT70 in de-endothelialized arteries (data not shown; see Zhang et al. 2005 and Raina et al. 2010 for the ouabain data). In fact, de-endothelialization shifted the CTS dose–arterial constriction curve toward lower CTS doses (i.e. it increased the apparent affinity for ouabain), presumably because it abolished vasoconstriction-induced, endothelium-dependent vasodilatation (Raina et al. 2010). The antagonism between ouabain and digoxin was still apparent after removal of the endothelium, but it occurred with much lower concentrations of both ouabain and digoxin, e.g. 0.1 and 0.3 nm, respectively (Fig. 3A and D). The ouabain–digoxin antagonism was also observed in the presence of 100 nm prazosin (Fig. 3B and D), an α1-adrenoceptor antagonist, which was used to block the effect of released noradrenaline (norepinephrine; NA) on the myocytes (Raina et al. 2010). This implies that at least part of the ouabain–digoxin antagonism resides in the arterial myocytes, and is not related to endothelial effects on the smooth muscle, or to NA release from the sympathetic nerve endings or NA's activation of smooth muscle contraction.

Figure 3.

Figure 3. Ouabain–digoxin antagonism in de-endothelialized arteries and arteries treated with prazosin or PP2

A, digoxin (0.3 nm) antagonizes the increase in myogenic constriction (MT70) induced by 0.1 nm ouabain in a de-endothelialized mesenteric small artery. B, digoxin (10 nm) antagonizes the increase in myogenic constriction (MT70) induced by 3 nm ouabain in a mesenteric small artery treated with α1-adrenoceptor blocker prazosin (100 nm). C, digoxin (10 nm) antagonizes the increase in myogenic constriction (MT70) induced by 3 nm ouabain in a mesenteric small artery treated with the c-Src-kinase inhibitor PP2 (1 μm). D, summarized data on the antagonism of ouabain-induced increases in MT70 in de-endothelialized arteries, and in arteries treated with 100 nm prazosin or 1 μm PP2; n = 5 arteries in each case. Endothelium removal or treatment with prazosin or PP2 did not significantly affect the antagonism of the ouabain-induced increase in MT70 by digoxin (P > 0.05 in all cases).

In contrast to these acute effects of CTSs, chronic administration of ouabain, both in vivo and in vitro, greatly increases expression of several Ca2+ transport proteins, including NCX type 1 and C-type transient receptor potential protein-6 (TRPC6), in arterial smooth muscle (Zulian et al. 2010). The protein up-regulation is manifested, functionally, by an increase in arterial myocyte Ca2+ signalling. Those chronic effects appear to be mediated by a ouabain-activated, α2-mediated serine/threonine protein kinase signalling cascade, probably involving c-Src. Ouabain, in vivo, phosphorylates and activates c-Src in arterial smooth muscle, and the effects of chronic in vitro ouabain on protein expression are blocked by 1 μm PP2, a c-Src-kinase inhibitor (Zulian et al. 2012). The same concentration of PP2 did not affect the acute 3 nm ouabain-induced increase in MT70 or the antagonism of this effect by 10 nm digoxin (Fig. 3C and D).

Other ‘digoxin-like’ CTSs also antagonize the vasotonic effect of ouabain

Several other CTSs were also tested for their ability to antagonize the effect of ouabain, using a protocol similar to that illustrated in Figs 2A and 4. Interestingly, none of the ‘ouabain-like’ CTSs (i.e. Strophanthus steroids) that we tested (ouabagenin, DHO and strophanthidin; Table 1), at 10 nm concentrations, antagonized the vasotonic effect of 3 nm ouabain (Fig. 4A and B show examples; Fig. 5, upper green bars, are summarized data). Rather, the addition of these compounds marginally increased MT70. This implies that the 3 nm dose of ouabain was near-maximally effective, and the additional effect of these other CTSs, like raising the ouabain concentration to 10 nm, had only a small additive effect. In contrast, all of the tested ‘digoxin-like’ (Digitalis) CTSs, digoxin (Fig. 2A), digitoxin, and digoxigenin, at 10 nm concentrations, reversibly reduced the ouabain-induced increase in MT by ∼70% (Fig. 5, upper red bars). Previously, we demonstrated that the ouabain-antagonist rostafuroxin (5 μm) has no vasotonic effect of its own, but it, too, antagonizes the vasotonic effect of ouabain (Zhang et al. 2005).

Figure 4.

Figure 4. Examples of the effects of three cardiotonic steroids on the ouabain-induced increase in MT70 (A–C), and of digoxin on the increase in MT70 induced by three cardiotonic steroids (D–F)

Bufalin (Buf, C), but not dihydro-ouabain (DHO, A) or strophanthidin (Stroph, B), all at 10 nm concentrations, antagonizes the 3 nm ouabain-induced increase in MT70. Digoxin (10 nm) antagonizes the increase in MT70 induced by 3 nm DHO (D) and 1 nm proscillaridin A (ProA, E), but not that induced by 3 nm MBG (F). In all panels, ‘MT’ is the control MT70 in mesenteric small arteries pressurized to 70 mmHg.

Figure 5.

Figure 5. Summarized data illustrate the relative abilities of various cardiotonic steroids (CTSs) to antagonize the 3 nm ouabain-induced increase in MT70 (upper row of bars), and the ability of 10 nm digoxin to antagonize the increase in MT70 induced by various CTSs (lower row of bars)

Data (5–8 arteries per bar) are from experiments similar to those illustrated in Fig. 4, and are expressed as: b/a × 100%. Positive values indicate vasodilatation (or antagonism); negative values indicate vasoconstriction (or synergism). The ability of the various CTSs to induce hypertension when infused subcutaneously (Manunta et al. 2000, 2001) is indicated at the bottom: ‘+’ agents induce hypertension; ‘–’ agents do not induce hypertension and are anti-hypertensinogenic (i.e. they antagonize the hypertensinogenic effect of ouabain); ND, not determined.

The three bufadienolide CTSs that we tested on the ouabain-induced constriction, bufalin, MBG and proscillaridin A, exhibited different behaviours. Bufalin (10 nm) was ‘digoxin-like’: it antagonized the vasotonic effect of ouabain (Figs 4C and 5), whereas neither MBG nor proscillaridin A exhibited antagonism to ouabain in this protocol (Fig. 5, upper blue bars).

An alternative protocol for testing the antagonism between CTSs was to increase the MT70 constriction by incubating pressurized arteries with 3 nm of the various Strophanthus, Digitalis or bufadienolide CTSs, and then add 10 nm digoxin. A sample protocol, in which MT70 was increased with 3 nm DHO, is illustrated in Fig. 4D. The results were complementary to those obtained when 3 nm ouabain was the initial vasotonic agent. Digoxin (10 nm) antagonized the vasotonic effects of the four ‘ouabain-like’ (Strophanthus) steroids and proscillaridin A (Fig. 4E), but had no effect on the vasotonic effect induced by digitoxin, digoxigenin or bufalin, i.e. the ‘digoxin-like’ CTSs (Fig. 5, lower bars). Also, digoxin, too, did not antagonize the effect of MBG (Figs 4F and 5, lower bars).

If the ouabain-augmented arterial constriction is associated with increased [Ca2+]CYT (Zhang et al. 2005; Raina et al. 2010), a key question is whether the antagonism by digoxin or other digoxin-like CTSs is linked to a decrease in [Ca2+]CYT. It is very difficult to address this issue in pressurized small arteries because the additional time required to load arteries with a Ca2+-sensitive dye before beginning a relatively long experiment means that only a small fraction of the experiments could be completed successfully. Therefore we needed a more convenient preparation in which to study Ca2+ signalling during CTS antagonism. This also enabled us to address another critical question, namely whether CTS antagonism could be observed in preparations other than arteries. The answers to these questions are provided in the next section.

Part 2. Studies on primary cultured rat hippocampal neurones

NCX mediates ouabain- and digoxin-induced augmentation of glutamate-evoked Ca2+ signals in neurones

Low glutamate (Glu) concentrations (3–4 μm) induce rapid, large, metabotropic glutamate receptor-mediated Ca2+ signals in neurones, and only small, delayed signals in astrocytes in primary cultured rat hippocampal neurone–glia co-cultures (Song et al. 2013). Moreover, low nanomolar concentrations of either ouabain or digoxin augment the Glu-evoked Ca2+ signals in the neurones (Song et al. 2013). As in arteries (Arnon et al. 2000; Zhang et al. 2005), signal augmentation by ouabain appears to be mediated by NCX, and involves increased Ca2+ storage in the endoplasmic reticulum (ER; Song et al. 2013). The effects of ouabain depend upon Na+, and are blocked by the NCX inhibitor SEA0400 (Arnon et al. 2000; Zhang et al. 2005; Song et al. 2013).

SEA0400 (300 nm) was used to test for the role of NCX in augmenting neuronal 3 μm Glu-evoked Ca2+ signals by digoxin (Fig. 6). This dose of SEA0400 had no effect on the control Glu-evoked signals, but blocked virtually all of the 3 nm ouabain-induced augmentation (Song et al. 2013). Figure 6A shows data from a representative neurone and illustrates the protocol. Figure 6B shows summarized data from 20 cells: 300 nm SEA0400 markedly attenuated the 3 nm digoxin-induced augmentation of Glu-evoked Ca2+ signals in neurones. These data indicate that most of the CTS-induced signal augmentation is mediated by NCX. The implication is that, rather than a direct action of digoxin on ER Ca2+ release (McGarry & Williams, 1993; Nishio et al. 2004), digoxin's effect, like that of ouabain, is largely the result of Na+ pump inhibition and elevation of sub-PM [Na+].

Figure 6.

Figure 6. SEA0400 antagonizes digoxin's augmentation of Glu-evoked Ca2+ signals in rat neurones

Effects of 3 nm digoxin ± 300 nm SEA0400 (NCX inhibitor) on 3 μm Glu-evoked Ca2+ signals in primary-cultured rat hippocampal neurones. A, time course of Fura-2 F360/F380 ratio signals from a representative neurone that illustrates the protocol. The diluent for the digoxin and SEA0400 was DMSO (final concentration = 0.33%, which alone has no effect on the signal; see Song et al. 2013). The cells were pre-incubated with digoxin ± SEA0400 for 5 min before the stimulation with 3 μm Glu. There was a 13–14 min washout period between each of the records shown. B, summary data for 20 cells like the one in A. SEA0400 reduced the digoxin-induced augmentation by 69%. ***P < 0.001 for the indicated pairs (ANOVA).

To test for a possible role of c-Src in the rapid augmentation of the Glu-evoked Ca2+ signals by ouabain, we employed PP2. Control Glu-evoked signals were unaffected by 1 μm PP2 (not shown). Pre-treatment with 3 nm ouabain for 5 min augmented the amplitude of the 3 μm Glu-evoked increase (Δ) in the Fura-2 fluorescence ratio, F360/F380, from 0.36 ± 0.06 (arbitrary units) to 1.02 ± 0.08 (P < 0.08; n = 19 cells). Following preincubation with both 3 nm ouabain and 1 μm PP2, the ΔF360/F380 was 1.04 ± 0.13, not significantly different from the ΔF360/F380 with only ouabain. The augmentation by ouabain, with or without PP2, was completely reversible. Thus, these signals, like the ouabain-induced increase in MT70 (Fig. 3D), do not appear to be mediated by c-Src kinase.

Ca2+ signalling as a surrogate for direct measurements of Na+ pump activity

It would be useful to study nanomolar CTS-induced changes in cytosolic [Na+] or Na+ pump activity directly. Unfortunately such studies are not feasible for several reasons. The high ouabain affinity α2 (artery) and α3 (neurone) Na+ pumps are present in low abundance in most cell types; for example, they constitute only ∼20% of the total Na+ pumps in astrocytes and arterial myocytes (Golovina et al. 2003; Zhang et al. 2005). Thus, complete inhibition or knockout of α2/α3 Na+ pumps causes only a very small (1–2 mm) rise in ‘bulk’ cytosolic [Na+] (Golovina et al. 2003). Moreover, α2 and α3 Na+ pumps are confined to plasma membrane (PM) microdomains at PM-sarco-/endoplasmic reticulum (S/ER) junctions, where they co-localize with NCX and help regulate cell Ca2+ and Ca2+ signalling (Juhaszova & Blaustein, 1997; Linde et al. 2012). Therefore, while patch clamp or Na+,K+-ATPase studies on these Na+ pumps may be prohibitively difficult, the robust CTS-induced, Na+-dependent, NCX-mediated Ca2+ signals are a convenient surrogate for direct α2 or α3 Na+ pump activity measurements.

Ouabain–digoxin antagonism in rat neurones

Rat hippocampal neurones also exhibit ouabain–digoxin antagonism (Fig. 7): ouabain (3 nm) amplified the 3 μm Glu-evoked Ca2+ signal, and the addition of 10 nm digoxin markedly and reversibly reduced the signal amplification (Fig. 7A–C). The same result was obtained when the sequence was reversed: 10 nm digoxin, alone, augmented the Ca2+ signal, and the augmentation was greatly reduced by the addition of 3 nm ouabain (Fig. 7D–F). When the ouabain concentration was raised to 10 or 30 nm, however, 10 nm digoxin was unable to antagonize the effect of ouabain (Fig. 7G and H). The implication is that digoxin and ouabain appear to be unusual competitive antagonists in both the arterial and neuronal assays. The antagonism never resulted in complete block of the response: the maximum reduction of the inhibitory effect of either CTS was ∼75% (e.g. Fig. 7B and C; and compare Fig. 2C and D). Moreover, the reduction of the response could be overcome by raising the concentration of either ouabain (Fig. 7G and H) or digoxin (Fig. 2C).

Figure 7.

Figure 7. Ouabain–digoxin antagonism on Glu-evoked Ca2+ signals in rat hippocampal neurones

Effects of 3 nm ouabain (Ouab, A–C) and/or 10 nm digoxin (Dig, D–F) on the F360/F380 ratio signal evoked by 3 μm Glu. A, B and D, E are data from representative neurones. The panels in A and D show F360 excitation images (Aa and Da) and peak response background-subtracted F360/F380 ratio cell images (b–g) which were captured at the times indicated in the time-course graphs in panels B and E, respectively. The cells were pre-incubated with ouabain and/or digoxin for 5 min before stimulating with 3 μm Glu. There was a 13–14 min washout period (break in the data) between each recording. C and F are, respectively, the summarized data from 17 cells similar to the one in A/B, and 12 cells similar to the one in D/E. The addition of digoxin reduced the ouabain-induced augmentation by 77.5% (C); the addition of ouabain reduced the digoxin-induced augmentation by 66% (F). G and H, experiments were performed with the same protocol as shown in ‘A and B’, but the ouabain concentration was raised to 10 nm (G, n = 11) or 30 nm (H, n = 9), respectively. *< 0.05, **< 0.01, ***P < 0.001; NS, not significantly different for the indicated pairs (ANOVA).

The ouabain aglycone ouabagenin was also tested. Like the other CTSs, ouabagenin augmented the Glu-evoked Ca2+ signals in neurones (Fig. 8). Digoxin (10 nm) attenuated the effect of 3 nm ouabagenin by about 69% (Fig. 8B) whereas 10 nm ouabain did not (Fig. 8A). When 3 nm digoxin was tested with 3 nm ouabagenin, there was less attenuation – about 30% (not shown). Thus, digoxin, but not ouabain, is also an antagonist of the ouabagenin effect in both arteries (Fig. 5) and neurones.

Figure 8.

Figure 8. Ouabagenin–digoxin antagonism of Glu-evoked Ca2+ signals in rat neurones

Summary data showing the effects 3 nm ouabagenin (Ouabgn) ± 10 nm ouabain (Ouab, A; n = 34 cells) or digoxin (Dig, B; n = 39 cells) on 3 μm Glu-evoked Ca2+ (Fura-2 F360/F380 ratio) signals evoked by 3 μm Glu in primary-cultured rat hippocampal neurones. The protocol was comparable to the ones used for the experiments in Figs 6A, and 7B and E. ***P < 0.001; NS, not significantly different for the indicated pairs (ANOVA).

The digitoxigenin derivative rostafuroxin behaves as a ‘pure’ ouabain antagonist

Rostafuroxin (Rosta, 6 μm), which does not inhibit Na+,K+-ATPase at this concentration (Ferrari et al. 1998), had no effect on Glu-induced Ca2+ signals or on the 3 nm digoxin-induced amplification of the signals in hippocampal neurones (Fig. 9A and B). Rostafuroxin did, however, block the ouabain-induced augmentation of the Glu-evoked signal (Fig. 9C). These results are analogous to the effects observed in pressurized arteries where rostafuroxin had no effect on control MT70, but blocked the ouabain-induced increase in MT70 (Zhang et al. 2005). The data suggest that rostafuroxin may be a ‘pure’ ouabain antagonist with negligible ‘agonist’ (Na+ pump inhibitory) effect in both of these preparations. These results are also consistent with the fact that rostafuroxin is a digitoxigenin derivative, and is structurally, as well as functionally, a member of the ‘digoxin-like’ compounds in terms of its antagonism of ouabain's, but not digoxin's effect.

Figure 9.

Figure 9. Effect of rostafuroxin on the augmentation of Glu-evoked neuronal Ca2+ signals by ouabain or digoxin

A, lack of effect of 6 μm rostafuroxin (Rosta) on the Ca2+ transients evoked by 3 μm Glu. Rostafuroxin was dissolved in DMSO (final concentration 0.367%), which alone had no effect on the Ca2+ transients. Data from a representative neurone is shown in a and summarized data from 11 neurones in b. B, effect of 3 nm digoxin ± 6 μm rostafuroxin on 3 μm Glu-evoked Ca2+ transients. Fura-2 F360/F380 ratio data are from a representative neurone (a) and summary data are from 16 neurones (b). C, effect of 3 nm ouabain ± 6 μm rostafuroxin on 3 μm Glu-evoked Ca2+ transients. Fura-2 F360/F380 ratio data are from a representative neurone (a) and summary data are from 42 neurones (b). Rostafuroxin reduced the ouabain-induced augmentation by 71.4% (Cb), but had no effect on the response to digoxin (Bb). ***P < 0.001; NS, not significantly different for the indicated pairs.

Figure 10 shows the rostafuroxin dose–response curve for antagonism of the ouabain-induced increase in the Glu-evoked Ca2+ signals. The apparent EC50, 1.4 μm, is very close to the EC50 (1.7 μm) for the displacement of 3H-ouabain from dog kidney Na+,K+-ATPase by rostafuroxin (Ferrari et al. 1998).

Figure 10.

Figure 10. Rostafuroxin dose–response curve: percentage inhibition of the ouabain-induced augmentation of the Glu-evoked Ca2+ signals in neurones

The data are from a total of 12 experiments, including those of Fig. 9C, and comparable experiments at the other rostafuroxin (Rosta) concentrations. The data points are the means ± SEM of 2 (at 0.26 μm) to 4 (at 0.88 μm rostafuroxin) experiments. The numbers of cells/experiment ranged from 4 to 17, with an average of 9/experiment. The fitted curve, which has been extrapolated to 0% at 0.01 μm, and 100% at 100 μm rostafuroxin, has an EC50 of 1.4 μm and a Hill coefficient of 1.5.

Ouabain–bufadienolide interactions

The bufadienolides, bufalin and proscillaridin A (both 10 nm), and MBG (3 nm), all augmented Glu-evoked Ca2+ signals in hippocampal neurones (Fig. 11). Bufalin also blocked the ouabain-induced, but not the digoxin-induced, augmentation of the Glu-evoked Ca2+ transient (Fig. 12A and B) and is, thus, ‘digoxin’-like’ in its actions (Fig. 5 shows comparable effects in the MT70 assay). On the other hand, neither proscillaridin A nor MBG antagonized either the ouabain- or the digoxin-induced amplification of the Glu-evoked signal (Fig. 12C–F). Thus, proscillaridin A appears to be MBG-like in this assay.

Figure 11.

Figure 11. Augmentation of Glu-evoked Ca2+ signals in neurones by three bufadienolides

Summarized data showing the effects of bufalin (Buf, 10 nm; A, n = 13 cells), proscillaridin A (ProA, 10 nm; B, = 21) and marinobufagenin (MBG, 3 nm; C, = 28) on the F360/F380 ratio signal evoked by 3 μm Glu. ***P < 0.001.

Figure 12.

Figure 12. Summary data showing the effects of three bufadienolides on the augmentation of Glu-evoked Ca2+ signals by ouabain or digoxin in neurones

Bufalin (Buf, 10 nm) reduced the 3 nm ouabain-induced augmentation of the 3 μm Glu-evoked Ca2+ transients by 72% (Ouab, = 28 cells, A), but had negligible effect on augmentation induced by 3 nm digoxin (Dig, = 22, B). Neither proscillaridin A (ProA, = 25 and 19 cells, respectively, C and D), nor marinobufagenin (MBG, = 36 and 26 cells, respectively, E and F), antagonized the augmentation induced by 3 nm ouabain (C and E) or digoxin (D and F). ***P < 0.001; NS, not significant (ANOVA).

Can cardiotonic steroids self-aggregate?

We asked whether the apparent antagonism between, for example, ouabain and digoxin might be the result of the formation of stable aggregates. This would lower the ‘free’ concentrations of both CTSs and thereby reduce ongoing inhibition of the Na+ pump. A number of reports indicate that some steroids can self-associate into stable complexes. Although some such aggregations occur in micelles (Haberland & Reynolds, 1973; Gilbert et al. 1975), others involve pairs (or higher order complexes) of molecules such as bile acids and α-tocopherol (Strauch et al. 1969; Chang & Cardinal, 1978; Rossi et al. 1995). There is precedent for self-association of molecules with a steroid nucleus such as bile salts because they may have ‘hydrophilic’ and ‘hydrophobic’ sides (Garzelli et al. 1992; Rossi et al. 1995; Hamlyn et al. 1996). Moreover, when equimolar ouabain and digoxin solutions ranging from 25 to 250 nm were examined by electrospray mass spectrometry, protonated and lithiated molecular ions at m/z 1366.2 and 1372.2, respectively, were observed (Supporting information Fig. S1), indicating the formation of ouabain–digoxin complexes. However, the concentration of the complexes was ≤1% of the total CTSs. Thus, physical complexation of these two CTSs exists in the gas phase but is quantitatively minor under the conditions tested.

To explore further the possibility of more extensive ouabain–digoxin aggregation, especially in aqueous media, the same high ionic strength physiological solutions used for the contraction and Ca2+ imaging experiments were employed together with surface plasmon resonance. Oxidized ouabain and digoxin were coupled to Biacore CM5 sensor chips. The coupling was confirmed by the binding of anti-ouabain antibodies, which reacted with both ligands, as expected. Binding to the uncoupled chip surface was tested in parallel, to correct for non-specific ligand binding. The oxidation of these CTSs does not significantly alter binding because oxidized ouabain and oxidized and biotinylated digoxin all bind to the Na+,K+-ATPase (Hegyvary, 1975; Nutikka et al. 1991), and oxidized ouabain is even a little more effective than ouabain as a cardiotonic in vivo (Arad et al. 1993).

With 60 nm antibody, oxy-ouabain coupled to the chip surface specifically bound ∼600 RU (response units) of antibody, and oxy-digoxin bound ∼1000 RU; at 6 nm antibody, the binding was ∼260 RU and ∼310 RU, respectively (Supporting information Figs S2 and S3). Based on these values, if stable ouabain–digoxin complexes were formed on the chips, we would expect signals of ∼5–20 RU when ouabain was tested with the oxidized digoxin chip and digoxin was tested with the oxidized ouabain chip. At CTS concentrations of 20 nm or even 200 μm, neither ouabain nor digoxin produced a detectable background-corrected signal (Supporting information Fig. S4 shows the digoxin data; the detection limit for the Biacore T200 is 1 RU). In addition, we tested 100 nm ouabagenin to circumvent the possibility that the interaction between ouabain and oxidized digoxin involved the rhamnose moiety on the ouabain, which we had oxidized to enable coupling to the chip surface. In this case, too, the signal was undetectable (Supporting information Figs S2 and S3). These data indicate that the observed ouabain–digoxin antagonism that occurs in the presence of arterial and neuronal preparations in physiological media is not due to formation of stable ouabain–digoxin aggregates.


Classic notions of CTS action; some observations do not fit the dogma

All the ‘classic CTSs’ tested in this study are Na+ pump inhibitors that augment myogenic tone at low nanomolar concentrations (Fig. 1 and Table 1). This effect apparently results from a rise in sub-PM [Na+] (Blaustein, 1993; Poburko et al. 2007) and a secondary increase in [Ca2+]CYT mediated by NCX (Arnon et al. 2000; Iwamoto et al. 2004; Zhang et al. 2005, 2010). Several CTSs were also tested on glutamate-evoked cytosolic Ca2+ signals in neurones, and all did, indeed, augment those signals (Figs 6 and 9, and Song et al. 2013). In neurones too, the effects of the CTSs appear to be mediated by NCX because they are inhibited by the NCX blocker SEA0400 (Fig. 6 and Song et al. 2013).

Three astonishing findings raised doubts about this ‘central dogma’ that all CTSs have virtually identical modes of action. First, chronic administration of digoxin or digitoxin to normal rats does not raise BP, whereas ouabain does (Huang et al. 1999; Manunta et al. 2000). Second, digoxin and digitoxin lower BP in rats made hypertensive with ouabain (Manunta et al. 2000). Third, whether in vivo (Pulina et al. 2010; Zulian et al. 2012), or in cultured arterial myocytes in vitro (Pulina et al. 2010; Linde et al. 2012; Zulian et al. 2012), prolonged treatment with low-dose ouabain, but not digoxin, induces increased expression of several myocyte Ca2+ transport proteins. These transporters include: the NCX, the S/ER Ca2+ pump SERCA2 and the receptor-operated channel component TRPC6. Moreover, chronic in vitro digoxin antagonizes the effect of prolonged ouabain treatment on Ca2+ transporter expression (Zulian et al. 2012). The present study was therefore undertaken to examine acute CTS interactions in two different read-out systems, rat mesenteric small arteries with myogenic tone and primary cultured rat hippocampal neurones.

Some CTSs antagonize the effects of others on both Ca2+ signalling in neurones and myogenic tone in small arteries

Antagonism between the effects of ouabain and digoxin was observed in both the neuronal Ca2+ signalling and arterial MT70 assays, and was overcome by increasing the concentration of either digoxin, tested in the MT70 assay (Fig. 2C), or ouabain, tested in the Ca2+ signalling assay (Fig. 7). Also, the antagonism in the arteries did not depend upon the endothelium and was observed even after the α1-adrenoceptors were blocked by prazosin (Fig. 3B and D). In both assay systems, the antagonist as well as agonist effects of the CTSs appeared to be a direct result of high affinity CTS binding to, and inhibition of, Na+ pumps (α2 in arteries, α3 in neurones) and the consequent NCX-mediated gain of Ca2+ by the respective cell types (see next section). This conclusion is also consistent with the fact that the classic CTSs act directly on the Na+ pump and do not interfere with cell metabolism (Schatzmann, 1953).

The influence of CTS structure on CTS antagonism was examined in arteries by comparing the ability of various CTSs to antagonize the vasotonic effect of ouabain, and by the ability of digoxin to antagonize the vasotonic effect of several CTSs. These experiments enabled us to classify a number of CTSs as either ‘digoxin-like’ or ‘ouabain-like’ (Figs 4 and 5). The vasotonic effect of the ouabain-like agents (ouabain, DHO, ouabagenin and strophanthidin) were antagonized by digoxin, but not by ouabain, with which they were synergistic (Fig. 5). Conversely, the digoxin-like agents (digoxin, digitoxin, digoxigenin and bufalin) antagonized the vasotonic action of ouabain, but not the vasotonic effects of the digoxin-like agents, with which they were synergistic (Fig. 5). This ouabain–digoxin and ouabain–bufalin antagonism is consistent with in vivo studies on the heart (Nesher et al. 2010), and in vitro data from NT2 cells (Feldmann et al. 2007).

In the neuronal Ca2+ signalling assay too, bufalin was ‘digoxin-like’: it antagonized the effects of ouabain, but not those of digoxin (Fig. 12A and B). Similarly, the synthetic derivative of digoxigenin, rostafuroxin, antagonized the effect of ouabain, but not of digoxin, in this assay (Fig. 9). Thus, both rostafuroxin, which is also a ouabain antagonist in the MT70 assay (Zhang et al. 2005), and bufalin are ‘digoxin-like’ in terms of their antagonist actions. Rostafuroxin is different, however; it blocked the augmentation by ouabain in both assays, but had no ‘agonist’ effect: it did not augment Ca2+ signals (Fig. 9) or vasoconstriction (Zhang et al. 2005). Thus, rostafuroxin appears to be a ‘pure’ antagonist of ouabain at the doses tested. The implication is that it can bind to the Na+ pump with relatively low affinity (EC50 ∼1.5 × 10−6 m), without blocking the cation transport channel and inhibiting Na+ extrusion – effects that are required for augmentation of Ca2+ signals and vasoconstriction (Wray et al. 1985; Blaustein, 1993; Blaustein et al. 1998). This is consistent with the evidence that rostafuroxin displaces 3H-ouabain from dog kidney (α1) Na+,K+-ATPase with an EC50 of 1.5 × 10−6 m, but inhibits ATP hydrolysis (and, presumably, Na+ transport) with an EC50 of 4 × 10−5 m, whereas ouabain has identical EC50 values (∼2 × 10−8 m) in the two assays (Ferrari et al. 1998).

It is noteworthy that, in both the myogenic tone assay (Fig. 2D; Zhang et al. 2005) and the Ca2+ signalling assay (Figs 7B and C, and 9C), micromolar doses of rostafuroxin but only nanomolar concentrations of digoxin were needed to antagonize the effects of ouabain. These data indicate that rostafuroxin is a much (∼100- to 1000-fold) weaker ouabain antagonist than digoxin. This may explain the absence of any antihypertensive effect of rostafuroxin in clinical trials where very low doses were used in unselected hypertensive patients (Staessen et al. 2011).

Two bufadienolide CTSs, MBG and proscillaridin A, did not fit this simple classification, although both were vasotonic and both also augmented neuronal Ca2+ signalling. There was no evidence of ouabain–MBG or digoxin–MBG antagonism in either assay (Figs 4F, 5 and 12E and F). Proscillaridin A was anomalous. It appeared to be ‘ouabain-like’ in the MT70 assay in that it did not antagonize the vasotonic effect of ouabain, while digoxin antagonized the vasotonic effect of proscillaridin A (Figs 4D and 5). In neurones, however, there was no evidence of either ouabain– or digoxin–proscillaridin A antagonism; thus, in this preparation, proscillaridin A behaved as an ‘MBG-like’ agonist (Fig. 12C–F).

What is the ‘common anti-ouabain pharmacophore’?

A critical question is what structural features distinguish the ouabain-like from the digoxin-like molecules. The comparative functional responses enable us to generalize about the significance of structural differences between the ouabain-like and digoxin-like CTSs. As the sugars are not relevant, they are excluded. The steroid nucleus (5β androstane), with numbered carbons, is illustrated in Fig. 13; it includes a C14β-OH, which is common to both classes, and a C3β-oxygen, because both classes include aglycones as well as C3β-glycosides. The ouabain-like CTSs all have 1β, 5β and 11α OH groups, a 12β-H, a C19-CH2OH and a C17 5-member lactone ring (reduced in the case of dihydro-ouabain); strophanthidin is the exception, with 1β- and 11α-H, and a C19-CHO. In contrast, digoxin-like CTSs are all 1β-, 5β- and 11α-H, and C19-CH3; they may be either 12β-H or 12β-OH, and may have either a 5- or 6-member lactone ring at C17. The lactone ring is common to both ouabain and digoxin groups and, thus, is not a key determinant of the antagonism.

Figure 13.

Figure 13. Prototypical cardiotonic steroid skeleton core

The primary features are a steroid skeleton in which the rings are fused in a cistranscis arrangement. All ouabain-like steroids in this study have a 14β-OH, an unsaturated lactone ring attached via C17 in the β configuration, and a methyl group at C18. The sugars, if present, are attached via the 3β-OH. See text for details. Reproduced with permission from Manunta et al. (2000).

Rostafuroxin, which is a digoxin-like ouabain antagonist, is structurally similar to the other digoxin-like compounds, but has a 17α-OH, and a 17β 3-furyl group in place of the lactone ring. Proscillaridin A, which appeared to be ouabain-like in the arterial assay, and MBG-like in the neuronal assay, is structurally more similar to the digoxin-like molecules, but differs in that it has a C4–C5 double bond that affects the configuration of the A/B rings. Also, unlike the other glycosylated digoxin-like molecules (digoxin and digitoxin both have three digitoxoses), proscillaridin A has a single rhamnose at C3.

Marinobufagenin, which is functionally neither ouabain-like nor digoxin-like in our assays, is structurally more similar to the digoxin-like compounds (including bufalin). However, MBG has a 5β-OH (similar to all the ouabain-like, and none of the digoxin-like, CTSs), and an oxygen epoxide between C14 and C15, rather than the 14β-OH that is present in both the ouabain- and digoxin-like CTSs.

The present work therefore indicates that the general structural features necessary for activity as a ouabain antagonist are restricted to the steroid nucleus, and are influenced by the configuration of, and substitutions within, the A/B ring area. As the number and spatial complexity of the major groups appears to be relatively small, novel ouabain antagonists that lack a steroidal nucleus seem feasible.

Anomalous antagonism between ‘ouabain-like’ and ‘digoxin-like’ CTSs

The antagonism between ‘ouabain-like’ and ‘digoxin-like’ CTSs is striking because these agents are recognized classically as fully-active ‘agonists’ (i.e. Na+ pump inhibitors): they all bind with high affinity and thereby inhibit the α2 and α3 Na+ pumps. By adding an ‘antagonist’ to an ‘agonist’, progressive, dose-dependent reversal of inhibition is expected, with complete reversal at very high antagonist concentrations (Segel, 1975). Conversely, if both substances are pure inhibitory agonists that bind to the same sites, one would normally expect the second substance simply to bind to, and inhibit, more pumps until all the sites are saturated. Clearly, this is not the case here, and another explanation is needed. One possibility is that the CTS interaction (antagonism) occurs at two binding sites remote from one another, i.e. one on the Na+ pump and another in a downstream signalling pathway (e.g. one involving a protein kinase). The CTSs do not, however, directly affect Ca2+ signalling, or the contraction mechanisms. This is exemplified by the normal, or even amplified, Ca2+ transients and vasoconstrictor responses to neurotransmitters in the presence of CTSs (Zhang et al. 2005; Song et al. 2013). Rather, the effects appear to be related to occupancy of the high affinity CTS binding sites on the Na+ pumps and inhibition of Na+ extrusion. Two other alternatives, however, require further consideration. One is that heterologous CTSs may aggregate in solution. The second is that Na+ pumps may function as oligoprotomers.

Regarding the first alternative, the extensive, but never complete, antagonism raised the possibility that heterologous (i.e. ouabain-like and digoxin-like) CTSs might form relatively stable complexes in solution. Our Biacore data (Supporting information Figs S2–S4) and electrospray mass spectroscopy data (Supporting information Fig. S1), however, are inconsistent with this possibility.

Na+ pump α subunits have a single ouabain binding pocket

An alternative explanation for the ouabain–digoxin antagonism is related to the fundamental stoichiometry of α subunits and CTS binding sites. X-ray crystallography of purified pig kidney (α1) Na+,K+-ATPase reveals that the Na+ pump catalytic (α) subunit contains a single high affinity binding site for ouabain (Yatime et al. 2011). The α1, α2 and α3 subunit isoforms are highly homologous, with ∼87% sequence identity. Most of the differences are located in the cytoplasmic N-terminal segment and the large cytoplasmic loop that are probably involved in isoform-specific sorting and regulation (Blanco & Mercer, 1998; Song et al. 2006). The ouabain binding pocket is conserved in all the α isoforms despite ouabain affinity differences, which are attributable primarily to specific amino acids in the extracellular loop (H1–H2) adjacent to the first two transmembrane helices (Dostanic et al. 2003; Dostanic-Larson et al. 2006; Lingrel, 2010). Those amino acids are located near the mouth of the cation channel/ouabain binding pocket (Yatime et al. 2011). Ouabain inserts into the channel from the extracellular side, lactone ring first, and plugs the channel, thereby blocking Na+ extrusion (Ogawa et al. 2009; Yatime et al. 2011); the mouth of the channel then closes around the ouabain (Yatime et al. 2011). Reduction of Na+ extrusion, with a local sub-PM rise in [Na+], secondarily promotes NCX-mediated net Ca2+ gain (Arnon et al. 2000; Golovina et al. 2003); this enhances Ca2+ signalling, and vasoconstriction and cardiac contraction (Wray et al. 1985; Blaustein, 1993).

Our experimental evidence is inconsistent with a single class of CTS binding sites on the functional pump. For example, rostafuroxin displaces 3H-ouabain from the pump without significantly inhibiting the Na+,K+-ATPase activity (Ferrari et al. 1998) or depressing cation transport (inferred from Fig. 9). This is difficult to explain unless rostafuroxin binds to pump molecules to which 3H-ouabain is already bound, but at a second site that doesn't plug the ion channel. Moreover, the effects are CTS structure-selective: rostafuroxin disinhibits the ouabain- but not digoxin-inactivated pumps (Fig. 9B and C). In other words, Na+ pumps are sensitive to certain discrete differences in the structure of the second CTS ligand that binds; this requires a minimum of two interacting ligand binding sites.

Do Na+ pumps function as mono- or oligo-αβ protomers?

The observations reported here, including the evidence that at least two separate CTS binding sites are involved, raise a number of questions concerning the relationship between αβ protomers. X-ray crystallography and other studies reveal that each α subunit has only one CTS binding site, and that mono-αβ protomers are fully functional (Cornelius, 1995; Martin & Sachs, 2000; Yatime et al. 2011). Thus, our results require that Na+ pumps operate as αβ oligomers.

Numerous investigators have suggested that Na+ pumps, including native pumps, in situ (Hah et al. 1985), function as di- or tetraprotomers (Askari, 1987; Skriver et al. 1989; Donnet et al. 2001; Laughery et al. 2004; Barbosa et al. 2010) with full, half-, or quarter-of-the-sites reactivity (Vilsen et al. 1987; Buxbaum & Schoner, 1991; Taniguchi et al. 2001). Nevertheless, this view is still controversial (Clarke & Fan, 2011). Importantly, nearly all studies of Na+,K+-ATPase structure and activity (other than measurements of ion transport) have been performed on detergent-treated, ‘purified’ native and/or cloned enzyme with an α1 subunit. Detergents radically alter the lipid composition and membrane environment of the native Na+ pumps in ways that are likely to affect the formation and dissociation of αβ oligoprotomers: high detergent concentrations tend to favour oligoprotomer dissociation (Ogan et al. 2007; Barbosa et al. 2010).

Diprotomeric, or (αβ)2, Na+ pump complexes might help to explain some of the acute ouabain–digoxin antagonism, but cannot account, quantitatively, for the data in Figs 2, 5, 7A–F and 12A. For example, binding of ouabain to one α subunit in a diprotomer might induce a conformational change that prevents the second protomer in the complex from binding digoxin with high affinity. At the same time, the presence of digoxin (perhaps at a low affinity site) might prevent ouabain from binding to the paired protomer. This model is not feasible for two reasons, however. First, it could only offset 50% of the inhibitory effect of ouabain, a value that is considerably less than the maximum of ∼70–80% activity recovery that we routinely observed (Figs 2B and C, 5, 7A–F and 12A). Second, it requires that digoxin interact at a low affinity site, but digoxin is equivalent to ouabain as a high affinity ligand (Fig. 1D) and thus would inhibit the second αβ protomer, leading to no net pump activity. Further, it seems very unlikely that the stimulatory effect of digoxin reflects an interaction at a low affinity site because both inhibitory and stimulatory effects of digoxin are readily detected even at 0.1 nm (Figs 1D, 2C and 3B). Thus, a more complex model is needed. In the following section, we explore models in which the Na+ pumps operate as oligotetramers with a variable fraction (half or quarter) of reacting protomers, and that may disaggregate and re-aggregate.

Ouabain–digoxin antagonism requires a Na+ pump oligotetramer with quarter-site reactivity

Figure 14 presents a series of simple models starting with the ‘classic’ interaction of a single α2 (or α3) subunit with ouabain and then digoxin (Model 1). In all the models, we assume that each α represents an αβ protomer, but we ignore the β subunit. In each model, four α2 subunits are shown for consistency; the percentage of active α2 subunits (below each model) is given relative to the control condition for that model (at the left). In Model 1, addition of ouabain at its EC50, by definition, leads to loss of activity of two of the four α2 subunits. Addition of digoxin at a higher concentration (>EC50) leads to additional activity loss due to increased occupancy of CTS receptors; there is no ouabain–digoxin antagonism. In Models 2 and 3 (the latter includes a dimer with half-of-the-sites reactivity), addition of digoxin also promotes further loss of activity without evidence of antagonism.

Figure 14.

Figure 14. Models for the interaction of ouabain and digoxin with the Na+ pump

In all models, four α2 Na+ pump subunits are shown for consistency. White and red ovals denote active and inactive α2 subunits, respectively. For simplicity, in Models 1–4, each reaction initially involves addition of one CTS (usually ouabain, black triangle) at its EC50, so that, at steady state, exactly one half of the subunits are inactivated (residual pump activity = 50%). Addition of a second, structurally different, CTS (e.g. digoxin, grey triangle), at a concentration approximately twice the EC50 for ouabain, is then shown. Model 1 uses monoprotomers, while Models 2–4 invoke diprotomers, either without (Model 2), or with (Models 3 and 4), half-of-the-sites reactivity. Models 5 and 6 invoke tetraprotomers each with quarter-site reactivity. In the latter models, binding of one CTS molecule to a single tetramer is sufficient to stop all ion pumping. Models 4 and 6 invoke disaggregration of complexed subunits upon addition of a second CTS. In each case, the relative ion pumping activity after addition of the first and second CTS (as a percentage of baseline) is shown below each model. See text for further discussion.

In Model 4, an additional feature is invoked: digoxin-induced disaggregation of the α2 protomers (Hah et al. 1985). In this model, the liberated α2 subunits offset the additional inhibitory effect expected when digoxin is added (total [CTSs] sufficient to give ∼75% of maximal binding), so that the apparent net pump activity is similar before and after addition of digoxin. This surprising implication arises from the dimer model; it suggests that even more complex models, in which Na+ pumps function in clusters of four α2 subunits (i.e. tetraprotomers), and in which only one α is active at a time (quarter-of-the-sites reactivity; Taniguchi et al. 2001), may be needed to explain the data. Accordingly, Model 5 is based on the assumption that the binding of one ouabain molecule to one of the four complexed α2 subunits causes all pumping in the tetraprotomer to cease (activity = 0%). Clearly, this is insufficient to enable ouabain–digoxin antagonism unless disaggregation of the complex also occurs, as in Model 6 (and see Model 4).

In Model 6A, addition of digoxin is proposed to promote a concentration-dependent (see Fig. 2C) disaggregation of the ouabain-inactivated complex. At the EC50 for ouabain, the disaggregation alone would, theoretically, allow two of the previously inactive α subunits in the complex that do not have CTSs bound to become fully active (active pumps = 200%). With digoxin and ouabain both present, however, the total CTS concentration is substantially above the EC50 for either CTS alone. The final net activity of the four α2 subunits reflects a dynamic balance between disaggregation of the complex and consequent activation of individual α2 subunits versus inhibition by the two CTSs. In the example with 3 nm ouabain, the net activating effect of digoxin is first apparent at 0.1 nm and peaks at 10 nm at which there is residual net inhibition of only ∼25–30% in the system (i.e. pump activity = 70–75% of baseline; see Figs 2C and 7C). Relative to the starting conditions, this paradoxical partial recovery of Na+ pump activity is sufficient to drive down cell Ca2+ and reverse most of the ouabain-evoked increase in MT (Figs 2 and 6). As the digoxin concentration is increased progressively above the optimal 10 nm, inhibition of all Na+ pumps will occur (i.e. active pumps → 0% activity, and augmentation by the CTS is restored; see Fig. 2C). Model 6B is identical to 6A but, in this case, the digoxin is added first, and ouabain acts as the antagonist (see Fig. 7D–F). Model 6 therefore offers a framework for understanding the symmetry in the results shown in Figs 2A and B, and 7.

The data in Figs 2-4, 7 and 12A show that the antagonistic effects of digoxin-like and ouabain-like CTSs are readily and fully reversible upon removal of the second CTS. Our model suggests that rapid re-aggregation of the tetrameric complex is responsible, with a return of the complex to an inactive state with one molecule of ouabain-like (or digoxin-like) CTS bound. This inference is supported by the observation (Figs 2-4) that both the onset and offset of the digoxin antagonism are nearly equivalent and relatively rapid (typically on a time scale of tens of seconds, with a new steady state achieved in 2–5 min). The rapidity of the events (including the slower secondary changes in cell Na+ and Ca2+ and MT) argues against trafficking of Na+ pumps in and out of intracellular organelles; it is certainly too fast to be explained by new pump synthesis and/or altered pump degradation.

To our knowledge, only one previous study has addressed the oligomeric state of the Na+ pump in intact cells. In radiation inactivation analysis in human red cells, the apparent molecular weight of the functional Na+ pump was ∼620 kDa, consistent with (αβ)4, whereas in washed membranes the target size fell to 320 kDa, consistent with (αβ)2 (Hah et al. 1985). Thus, in living cells, and in contrast to many biochemical preparations of the Na+,K+-ATPase, the Na+ pumps appear to be tetraprotomers. Moreover, in red cells treated with 1 μm digitoxigenin, the target size fell to 330 kDa, a value similar to the size found in washed membranes; this is consistent with disaggregation of αβ tetramers to dimers. We, too, conclude that CTSs induce functional tetraprotomer disaggregation, but we suggest further that the co-presence of digoxin- and ouabain-like CTSs induces disaggregation to monoprotomers via a short-lived diprotomer state. Additionally, our model suggests that a large free energy change is associated with the disaggregation of diprotomers, implying that energy is required to maintain the tetrameric state in intact cells. This energy appears to be missing in some broken membranes and detergent-treated preparations of the Na+,K+-ATPase, where αβ diprotomers or monoprotomers are typically observed (Vilsen et al. 1987; Martin & Sachs, 2000; Clarke & Fan, 2011).

Models 5 and 6 in Fig. 14 invoke α2 (or α3) subunits in the tetraprotomeric state. Stoichiometric mixtures of α1 and α2 subunits in tetraprotomers may, a priori, be possible. In this case, however, an especially interesting feature of a tetramer whose pump activity is inhibited by a single ouabain–α2 subunit interaction would be the liberation of two α1 subunits upon digoxin-induced disaggregation (assuming that the tetramers contain two α1 and two α2 protomers). Our data argue against this particular functional mixture of α subunit isoforms simply because the addition of digoxin at 10 nm would liberate two rodent CTS-resistant (Blanco & Mercer, 1998) α1 subunit isoforms. The combined pump activity of these α subunit isoforms would then far exceed (active pumps = 200%) both the baseline pump activity (active pumps = 100%), and that observed when nanomolar digoxin is present (Fig. 2C, active pumps = 70–75%). Thus, functional mixtures of α1 and α2 isoforms are not quantitatively feasible. This conclusion is consistent with studies which indicate that these isoforms exhibit little, if any, co-localization (Lee et al. 2006; Song et al. 2006).

Rostafuroxin–Na+ pump interactions fit the tetraprotomer model

Figure 15 uses Model 6 (Fig. 14) to describe the interactions between rostafuroxin (Rosta) and either digoxin or ouabain with the tetrameric Na+ pump. Panel A shows that while Rosta (inverted green triangle) binds, it neither inhibits the pump (indicated by the white, Rosta-bound protomer) nor induces disaggregation. Then, when digoxin, the much higher affinity ligand, is added, it displaces Rosta and inhibits those pumps (indicated by the red, digoxin-bound protomer), thereby augmenting Ca2+ signals (Fig. 9B). The binding of digoxin does not induce disaggregation because digoxin and Rosta are both ‘digoxin-like’ molecules, and the disaggregation of tetramers requires the simultaneous binding of a ‘digoxin-like’ and an ‘ouabain-like’ ligand.

Figure 15.

Figure 15. Models for the interaction of rostafuroxin, an ouabain antagonist, with the Na+ pump

As in Fig. 14, models 6A and B, these models are based on tetrameric Na+ pumps with quarter-site reactivity in which the binding of one molecule of digoxin (grey triangle) or ouabain (black triangle) stops ion pumping. The white and red ovals denote active and inactive pump protomers, respectively. The relative ion pumping activity after each addition, as a percentage of baseline, is shown below each model. A shows that rostafuroxin, a digitoxigenin derivative (Rosta, green triangle), has no effect on pumping when all tetramers are active (i.e. in the absence of digoxin or a ouabain-like CTS). Upon addition of the much higher affinity ligand, digoxin, Rosta is displaced and ion pumping is inhibited without disaggregation because digoxin and Rosta are structurally similar. In B, with ouabain already bound, addition of Rosta (green triangle) evokes disaggregration of the ouabain-complexed tetramer. B and C differ only in the order of ouabain and Rosta addition. Using the assumption that Rosta binds and can maintain disaggregation (but see text) without being inhibitory, a large (3-fold) increase in activity is predicted. This predicted increase would be a true stimulation of activity above the control condition. See text for further discussion of the quantitative effects of Rosta, and why such stimulation is not observed experimentally.

As illustrated in Fig. 15B and C, this simple model, does not appear to fit the experimental data in Fig. 9C. The model predicts that the combination of ouabain and Rosta should lead to a large net increase in Na+ pump activity relative to baseline (i.e. a true Na+ pump stimulation above the control, and not simply the relief from ouabain inhibition). This would be consistent with tetramer disaggregation and the inability of Rosta to inhibit Na+ pumps. The experiments (Figs 9C and 10) show, however, that some residual inhibition by ouabain was still present; this demonstrates that Rosta, at the doses tested (Fig. 10), was unable to reverse, completely, the inhibitory action of ouabain.

Some important considerations were ignored in the preceding analysis. First, higher concentrations of Rosta than those used in Figs 9 and 10 should, in principle, drive the reaction further to the right, assuming that there are no other rate limiting events. This should enhance disaggregation and increase Na+ pump activity >100% (i.e. above the control – a true stimulation). This was not testable with Rosta because the highest dose used, 6 μm, was near the solubility limit, and additional solvent vehicle was toxic to the preparations.

Second, and most important, none of the previously mentioned models takes reaction kinetics into account. This was unnecessary for ouabain and digoxin because they probably have relatively similar kinetics; i.e. relative to Rosta, they remain bound. Rosta antagonizes ouabain-enhanced arterial MT (Zhang et al. 2005) and neuronal Ca2+ signalling (Figs 9C and 10). This may be due to displacement of bound ouabain from the Na+ pumps by Rosta (Ferrandi et al. 1998) and/or to disaggregation of Na+ pump tetramers. Rosta is unlikely to remain bound, however, because the EC50 for Rosta is ∼1000-fold higher than the EC50 for digoxin or ouabain (compare Figs 2 and 10; Ferrari et al. 1998; Song et al. 2013). Thus, Rosta may be unable to maintain the disaggregation, and active Na+ pump monomers should quickly drift back to the tetrameric state even if ouabain remains bound to some of those tetramers. The net effect of Rosta on disaggregation is therefore expected to be limited, and certainly far less than the maximal (300%) activation illustrated in Fig. 15B and C. Indeed, the net effect of Rosta should reflect the influence of Rosta on ouabain binding as well as the balance between tetrameric and active monomeric states which is a function of the concentrations of Rosta and ouabain, and their kinetic properties. Thus, the net stimulation implied in Fig. 15B and C may not be possible. In order for tetramer dissociation and effective antagonism to occur (see Figs 2-4, 7 and 8, and Model 6 in Fig. 14), two structurally dissimilar CTSs must be bound simultaneously and must remain bound. The behaviour of Rosta in Figs 9 and 10 is, therefore, consistent with Fig. 14 Model 6.

Acute versus chronic ouabain–digoxin antagonism

Recent work reveals that the α2 Na+ pump also behaves as a ‘biased receptor’ for CTSs during prolonged exposure to these agents (Zulian et al. 2012), but this effect appears to be independent of the cation transport function of the α2 Na+ pump. Biased receptors generate different ligand-dependent signals and have been described in other critical receptor–ligand systems (DeWire & Violin, 2011; Kenakin, 2012). For example, the prolonged high affinity binding of ouabain to α2 Na+ pumps led to activation of c-Src, a PP2-sensitive protein kinase, while digoxin was unable to activate c-Src. Instead, digoxin behaved as an antagonist that completely blocked the stimulating effect of ouabain on Ca2+ transporter up-regulation. In sharp contrast to those results, the present acute studies show that both digoxin and ouabain behave as full agonists when administered alone (Figs 1, 2 and 7, and Song et al. 2013). Further, the short-term actions of these CTSs, described here, are symmetrical: the acute signalling effects evoked by addition of ouabain followed by digoxin are indistinguishable from those when digoxin is added prior to ouabain. Thus, there is no evidence that the biased receptor behaviour of α2 Na+ pumps is necessary to explain the acute effects of ouabain or digoxin. This is a second major difference between the short- and long-term actions of these CTSs.

Antagonism of CTS inhibition versus stimulation of the Na+ pump in vivo

In vivo, circulating endogenous CTSs that resemble ouabain, digoxin and MBG have been reported (Schoner & Scheiner-Bobis, 2007; Bagrov et al. 2009; Blaustein & Hamlyn, 2010). Their existence implies that some high CTS affinity Na+ pump tetramers must be inhibited by one or more of these CTSs. We omit further discussion of MBG here because this steroid does not antagonize the effect of ouabain or digoxin (Figs 5 and 12E and F). In Fig. 16, we focus on the predicted effects of adding ouabain (A) or digoxin (B) into the in vivo environment where endogenous digoxin- or ouabain-like steroids, respectively, are already bound to tetrameric Na+ pumps (Model 6 in Fig. 14). In both cases (Fig. 16A and B), a biphasic concentration–activity relationship is predicted: there is an apparent stimulation of activity induced by addition of low concentrations of the second (dissimilar) CTS, followed by inhibition at high concentrations. It is critical to realize that the portion of CTS-inhibited tetramers typically will be unknown in vivo and, more likely, not considered by the experimenter. The increasing Na+ pump activity evoked by low doses of CTS is, therefore, likely to be interpreted as stimulation. Panels A and B in Fig. 16 make it clear, however, that this phenomenon is not a true stimulation but, rather, a relief from pre-existing inhibition.

Figure 16.

Figure 16. Models that predict the in vivo situation where endogenous CTSs are present

Starting with the tetrameric model of Fig. 14 (Models 6A and B), A and B show the initial in vivo condition in which 33% of the available tetramers are blocked by digoxin (grey triangle, A) or endogenous ouabain (black triangle, B). The two panels then show the effect when the experimenter adds a low (nm) dose, and then a high (μm) dose of ouabain (A) or digoxin (B) into the environment. The same assumptions as in Models 6A and B (Fig. 14) are used here. The relative ion pumping activity after each addition, as a percentage of the baseline, is shown below each model. The in vivo ‘starting condition’ is the assumption that only 67% of the available tetramers are active; this reflects the interaction of endogenous CTSs with the tetramers. This baseline value is also shown normalized to 100% to emphasize subsequent changes in activity as seen by the experimenter. The addition of low dose CTSs stimulate activity above the baseline condition while higher doses cause inhibition. The models make it clear that the stimulatory effect is actually the relief from pre-existing inhibition. See text for further details, including evidence for the in vivo starting condition.

In accord with these considerations, a number of investigators have reported similar biphasic ouabain dose–response curves on various Na+,K+-ATPase and intact cell preparations that presumably include tetramers (e.g. Lee & Yu, 1963; Repke, 1963; Hougen et al. 1981; Cook et al. 1983; Hamlyn et al. 1985). In the Hamlyn et al. study, on crude dog kidney Na+,K+-ATPase, the basal specific activity was increased and the stimulatory effect of ouabain was absent when the enzyme was prewashed in a high K+, but not high Na+, solution. Thus, in the case where stimulation of the enzyme by low dose ouabain occurred, it was, in actuality, ‘due to the relief from a pre-existing state of partial inhibition’ that was removed by the K+ pre-treatment (Hamlyn et al. 1985). Canine renal Na+,K+-ATPase consists, almost exclusively, of α1 Na+ pumps (which, in the dog, have high CTS affinity), whereas the present results are from rat α2 and α3 Na+ pumps in arteries and neurones, respectively. This implies that all three of the main α subunit isoforms can form tetramers and should, in principle, produce biphasic dose–response curves under the appropriate conditions.

Conclusion and significance

We have shown, using different assays in rat arteries and neurones, that ouabain-like and digoxin-like CTSs can antagonize one another on short time scales (seconds to minutes). When present alone, members of these two groups of ‘classic CTSs’ have similar ‘classical’ effects on Na+,K+-ATPase activity, Na+ transport, Ca2+ signalling, and cardiac and vascular contraction. Yet, when used in combination, their effects are remarkably distinct and incompatible with the simple additive classic model for their action. The results reported here lead to a new concept of CTS–Na+ pump interactions that helps to reconcile a number of divergent, seemingly inconsistent observations and ideas about Na+ pump function.

The existence of Na+ pumps as tetraprotomeric complexes, in situ, raises a question as to the normal functional significance of these complexes. Clearly, any mechanism that can promote the dissociation of tetramers would trigger large and rapid increases in Na+ pump activity. One such mechanism might be the rapid increase in surface Na+ pump activity initiated by adenosine monophosphate (AMP) in skeletal muscle (Benziane et al. 2012). In the case of EO, the blood levels of which are regulated, the introduction of an exogenous digoxin-like CTS, at a therapeutically relevant dose, would evoke the rapid disaggregation of tetramers and thereby tend to broaden the concentration–response curve for Na+ pump inhibition. In this context, the digoxin-evoked disaggregation may be beneficial in minimizing toxicity: it would help keep the number of active pumps virtually constant while simultaneously preventing ouabain-complexed pumps from triggering long term signalling events (Zulian et al. 2012), an action that apparently requires tetramers. In addition, since ouabain is endogenous to mammals (Blaustein & Hamlyn, 2010; Blaustein et al. 2012), disaggregation of tetramers would help EO to modulate α2-mediated Na+ transport and Ca2+ signalling with negligible influence on α1 function in the same cells even when, as in humans, the affinity of ouabain for α1 is only ∼10-fold lower than for α2 (Linde et al. 2012).

Importantly, plasma EO is elevated in several forms of hypertension (Rossi et al. 1995; Pierdomenico et al. 2001), and correlates positively with BP in essential hypertension (Tripodi et al. 2009; Manunta et al. 2011); EO levels of 0.15–0.4 nm are usually observed in hypertension, but may range as high as 1–2 nm. Plasma EO is also elevated in heart failure (Gottlieb et al. 1992). Thus, our results are relevant to the use of CTSs in high EO states. For example, digitalis glycosides were a mainstay of therapy for congestive heart failure for two centuries (Wray et al. 1985), and they are still widely used for the therapy of atrial fibrillation, especially in the elderly (Gheorghiade et al. 2013). Intriguingly, the use of digitalis has been controversial; it is associated with increased morbidity and mortality in some patients with heart failure, but not others (Digitalis Investigative Group, 1997; Rathore et al. 2002), for reasons that are not clear. Because elevated levels of EO circulate in many patients with congestive heart failure (Gottlieb et al. 1992), treatment with digoxin would, based upon the dynamic ouabain–digoxin antagonism we describe, seem likely to promote vascular dilatation in patients – especially in those given lower doses of digoxin (Kirch, 2001; Rathore et al. 2002). Further, low doses of digoxin, which are associated with better therapeutic outcomes (Rathore et al. 2002), acted as a vasodilator when administered to arteries pre-contracted by ouabain, whereas higher concentrations of digoxin, that would be toxic in vivo, promoted vasoconstriction (Fig. 2C). We suggest that the functional antagonism between ouabain and digoxin is relevant to the use of digitalis glycosides as a vasodilator in heart failure (Kirch, 2001).

Finally, the ouabain antagonism exemplified by rostafuroxin, as well as the previously mentioned antihypertensive effects of digoxin, digitoxin and rostafuroxin (Quadri et al. 1997; Ferrari et al. 1998; Huang et al. 1999; Manunta et al. 2000), are noteworthy. In vivo, both the short-term and long-term (‘biased’) effects of the CTS will be linked, and are therefore of special relevance during therapy. In the high ouabain/EO models mentioned above, more EO-bound Na+ pump tetramers and up-regulated arterial c-Src and NCX expression would be expected. Under these circumstances, the administration of digoxin, digitoxin, or rostafuroxin should dissociate EO-bound tetramers and increase arterial Na+ pump activity. These acute actions would be followed by the slow dampening of ‘biased signalling’ with a progressive decline in c-Src, NCX expression (Zulian et al. 2012) and blood pressure over several days (e.g. Manunta et al. 2000).

Overall, our data provide new and strong support for the idea that the Na+ pump CTS binding site is an important, under-appreciated and intriguing therapeutic target. Our observations may therefore have far-reaching implications.

Additional information

Competing interests

The authors have no conflicts of interest to disclose.

Author contributions

M.P.B., E.K., H.S. and J.M.H. designed the experiments. E.K. and H.S. performed the experiments. E.K., H.S., M.P.B. and J.M.H. analysed the data, and M.P.B., J.M.H., H.S. and E.K. interpreted the results. J.M.H. and M.P.B. generated the models. M.P.B. and J.M.H. drafted the manuscript. All four authors edited and approved the final manuscript.


This work was supported in part by NIH grants HL78870, HL45215, HL10755 and NS16106.


We thank the University of Maryland School of Medicine Biosensor Core Facility (Dr Y. Zhang, operator, and Dr R. J. Bloch, director) for performance of the SPR assays. We thank Dr A. Y. Bagrov (National Institute on Aging, NIH, Baltimore, MD, USA) for a gift of marinobufagenin, Dr P. Ferrari (Prassis-Sigma Tau, Milan, Italy) for a gift of rostafuroxin and Dr. T. Iwamoto, University of Fukuoka, Fukuoka, Japan, for a gift of SEA0400. We thank an anonymous reviewer for suggesting the analysis shown in Fig. 15.

Author's present address

E. Karashima: Department of Internal Medicine, Division of Cardiology, Shimonoseki City Hospital, 1-13-1 kouyou-chou Shimonoseki, Yamaguchi 750-8520, Japan.