Inhibition of Sodium-Calcium Exchange by Ceramide and Sphingosine*

Na 1 /Ca 2 1 exchange activity in Chinese hamster ovary cells expressing the bovine cardiac Na 1 /Ca 2 1 exchanger was inhibited by the short chain ceramide analogs N acetylsphingosine and N -hexanoylsphingosine (5–15 m M ). The sphingolipids reduced exchange-mediated Ba 2 1 influx by 50–70% and also inhibited the Ca 2 1 efflux mode of exchange activity. The biologically inactive ceramide analog N -acetylsphinganine had only modest effects on exchange activity. Cells expressing the D (241– 680) and D (680–685) deletion mutants of the Na 1 /Ca 2 1 exchanger were not inhibited by ceramide; these mutants show defects in both Na 1 -dependent and Ca 2 1 -de-pendent regulatory behavior. Another mutant, which was defective only in Na 1 -dependent regulation, was as sensitive to ceramide inhibition as the wild-type exchanger. Inhibition of exchange activity by ceramide was time-dependent and was accelerated by depletion of internal Ca 2 1 stores. Sphingosine (2.5 m M ) also inhibited the Ca 2 1 influx and efflux modes of exchange activity in cells expressing the wild-type exchanger; sphingosine did not affect Ba 2 1 influx in the D (241–680) mutant. The effects of the exogenous sphingolipids were reproduced by blocking cellular ceramide utilization pathways, suggesting that exchange activity is inhibited by increased levels of endogenous ceramide and/or sphingosine. We propose that sphingolipids

The Na ϩ /Ca 2ϩ exchanger is the principal Ca 2ϩ efflux mechanism in cardiac myocytes and plays a critical role in regulating the force of cardiac muscle contraction (1). Its stoichiometry is thought to be 3 Na ϩ /Ca 2ϩ (2), although a recent report suggests that it may have a higher, or variable, stoichiometry (3). Exchange activity is regulated by Ca 2ϩ -dependent and Na ϩ -dependent processes; cytosolic Ca 2ϩ activates exchange activity by binding to high affinity regulatory sites located within a large (546 residues) hydrophilic domain situated between the fifth and sixth transmembrane segments of the exchanger (4,5). Cytosolic Na ϩ is thought to induce the time-dependent formation of an inactive state (Na ϩ -dependent inactivation) after it binds to the Na ϩ translocation sites on the exchanger (6). Na ϩ -dependent inactivation can be antagonized by ATP-dependent synthesis of phosphatidylinositol 4,5bisphosphate, by elevated concentrations of cytosolic Ca 2ϩ and by certain mutations within the "regulatory" hydrophilic domain of the exchanger (6 -8).
The physiological significance of these regulatory mechanisms is uncertain (9,10). In intact cells, the K d for Ca 2ϩ -dependent activation of the exchanger is ϳ50 nM (11-13), suggesting that the exchanger may be nearly fully activated under "resting" conditions. The low cytosolic Na ϩ concentration and the high levels of ATP and phosphatidyl 4,5-bisphosphate in healthy cells preclude a major role for Na ϩ -dependent inactivation in regulating exchange activity under physiological conditions. Indeed, there is no direct evidence that Na ϩ /Ca 2ϩ exchange activity is in fact regulated in functioning cardiac myocytes. The present report addresses the possibility that sphingolipids such as ceramide and sphingosine serve as physiological regulators of Na ϩ /Ca 2ϩ exchange activity.
Ceramide is a central component of the sphingomyelin cycle, a stress-activated signaling pathway that participates in the induction of apoptosis and growth arrest. The activating or inhibiting effects of ceramide on a host of intracellular signaling pathways have been described in several recent reviews (14 -16). Ceramide can also be converted to sphingosine and sphingosine phosphate, two other signaling lipids with important regulatory effects of their own.
Here we show that short chain ceramide analogs and sphingosine inhibit Na ϩ /Ca 2ϩ exchange activity in transfected Chinese hamster ovary (CHO) 1 cells expressing the bovine or canine cardiac Na ϩ /Ca 2ϩ exchangers. Similar effects were noted when the cells were treated for 60 min with an inhibitor of endogenous ceramide metabolism. The differential effects of ceramide on exchanger mutants defective in Na ϩ -or Ca 2ϩ -dependent regulation suggest that ceramide blocks the conformational transitions associated with Ca 2ϩ -dependent activation of exchange activity. We propose that in functioning cardiac myocytes, sphingolipids and diastolic Ca 2ϩ levels interact to control the distribution of exchangers between the Ca 2ϩ -activated and nonactivated forms.
cardiac Na ϩ /Ca 2ϩ exchanger (17). The ⌬(241-680) deletion mutant was prepared and expressed in CHO cells (CK138 cells) as described (18). For XIPA cells, the basic residues in the XIP region of the exchanger ( 219 RRLLFYKYVYKRYRAGKQR) were replaced by alanines ( 219 AALL-FYAYVYAAYAAGAQA). For this purpose, RsrII and BclI restriction sites were introduced into the cDNA for the bovine exchanger (p17; Ref. 19) by site-directed mutagenesis at positions 1016 and 1079, respectively; an oligonucleotide cassette with the desired sequence was ligated into the cDNA after cleavage with RsrII and BclI to create the XIPA mutant. cDNAs for the ⌬(680 -685) mutant of the canine exchanger, as well as the wild-type canine exchanger, was kindly provided by Drs. Kenneth D. Philipson and Debora A. Nicoll, UCLA School of Medicine; the full-length cDNAs were ligated into the pcDNA3 expression vector (Invitrogen) and used for transfection of CHO cells following the procedures described previously (17). The cells were grown in Iscove's modified Dulbecco's medium containing 10% fetal calf serum and antibiotics as described (17).
Materials and Solutions-Na-PSS contained 140 mM NaCl, 5 mM KCl, 1 mM MgCl 2 , 10 mM glucose, and 20 mM Mops adjusted to pH 7.4 (37°C) with Tris. K-PSS had the same composition as Na-PSS except that NaCl was replaced with KCl (total concentration, 140 mM). Na-PSS was diluted 7-fold with K-PSS to yield 20/120 Na/K-PSS. Fura-2/acetoxymethylester was purchased from Molecular Probes, Inc. (Eugene, OR). Ceramide analogs and sphingosine were obtained from Biomol Research Laboratories (Plymouth Meeting, PA). All other biochemicals were purchased from either Sigma or Calbiochem (La Jolla, CA).
Fura-2-based Assays of Ba 2ϩ Influx-The Ba 2ϩ influx assay for Na ϩ / Ca 2ϩ exchange activity is fully described in Ref. 20. Cells were grown to confluence in 75-cm 2 culture flasks, washed 3 times with Na-PSS and incubated for 1 min at 37°C with Na-PSS containing 5 mM EDTA to detach cells from the flask. The suspended cells were centrifuged at 700 ϫ g for 1 min, resuspended in Na-PSS ϩ 1 mM CaCl 2 , centrifuged again, and resuspended in 4 -5 ml of Na-PSS ϩ 1 mM CaCl 2 containing 1% bovine serum albumin. The cells were divided into 300-l aliquots and incubated for 30 min with 3 M fura-2/acetoxymethylester and 0.25 mM sulfinpyrazone (to retard transport of fura-2 out of the cells). The fura-2 and sulfinpyrazone were added as 1000-fold concentrated stock solutions in dimethyl sulfoxide.
After the 30-min loading period, the cells were centrifuged for several seconds in an Eppendorf Mini-centrifuge, washed, and preincubated for 5 min in 100 l of Na-PSS ϩ 1 mM CaCl 2 or with other additions and conditions as indicated in individual experiments. These cells were then added directly to fluorescence cuvettes containing 3 ml of 20/120 Na/K-PSS ϩ 0.3 mM EGTA. Gramicidin (1 g/ml) was added to the cuvettes to equilibrate transmembrane Na ϩ and K ϩ gradients. Sphingolipids were then added to the cuvette as 1000-fold concentrated stock solutions in dimethyl sulfoxide. After monitoring fura-2 fluorescence for 60 s, 1 mM BaCl 2 (30 l of 0.1 M stock solution) was added to the cuvette, and fura-2 fluorescence was followed for an additional 240 s.
Fura-2 fluorescence was measured at 510 nm with alternate excitation at 350 and 390 nm using a Photon Technology International RF-M 2001 fluorometer (South Brunswick, NJ); data points were obtained at 1.8-s intervals. All fluorescence values were corrected for autofluorescence using cells that had not been loaded with fura-2. Data are presented as the ratio of fluorescence for excitation at 350/390 nm and represent the mean values (Ϯ S.E.; error bars shown in figures) for the number of experiments (n) indicated in the figure legends. Statistical testing utilized Student's t test for unpaired samples, unless indicated otherwise.
Assay for Ca 2ϩ Efflux-To measure Ca 2ϩ efflux activity, the cells were placed in cuvettes containing either Na-or K-PSS ϩ 0.3 mM EGTA. Ca 2ϩ release from internal stores was initiated by adding thapsigargin (3 l; 1 mM in dimethyl sulfoxide), and fura-2 fluorescence was monitored at the excitation wavelengths appropriate for Ca 2ϩ (340 and 380 nm). Dimethyl sulfoxide alone at the concentrations used had no effect on Ca 2ϩ release. The difference in the amplitude and duration of the ionomycin-induced [Ca 2ϩ ] i transient in K-PSS versus Na-PSS was taken as an indication of exchange-mediated Ca 2ϩ efflux activity (21). Treatment with PPMP-Cells were distributed in 300-l aliquots as described above for the fura-2-based assay of Ba 2ϩ influx, except that 0.1% bovine serum albumin was present in the medium instead of 1% bovine serum albumin. PPMP (30 M) was added from a 1000-fold concentrated stock solution in dimethyl sulfoxide; for the control cells, an equal volume of dimethyl sulfoxide alone was added. After 30 min of incubation, 3 M fura-2/acetoxymethylester was added, and after an additional 30 min of incubation, the cells were centrifuged, washed, and preincubated for 2 min in Na-PSS with the additions specified in individual experiments. PPMP was not present in the wash or the 2-min preincubation solutions.

Ceramide Analogs Inhibit Exchange-mediated Ba 2ϩ Influx-
The data in Fig. 1 show that 15 M N-acetylsphingosine (C2ceramide), a short chain fatty acid analog of ceramide, inhibited Na ϩ /Ca 2ϩ exchange activity in transfected CHO cells expressing the bovine cardiac Na ϩ /Ca 2ϩ exchanger (CK1.4 cells). Exchange activity was assayed as the rate of Ba 2ϩ influx in cells treated with gramicidin in a medium containing 20 mM NaCl and 120 mM KCl; the gramicidin was added to equalize the Na ϩ concentrations across the plasma membrane (20). As shown in Fig. 1, the initial rate of Ba 2ϩ influx was inhibited by 38% (p Ͻ 0.02) when 15 M C2-ceramide was included in the assay medium. The degree of inhibition increased with time, as shown by the downward curvature of the trace in Fig. 1. By 200 -300 s, the rate of Ba 2ϩ influx in the presence of C2ceramide was 72% less (p Ͻ 0.001) than that of control cells over the same interval (Fig. 1B).
The inhibition of Ba 2ϩ influx by C2-ceramide was mimicked by C6-ceramide, another short chain ceramide analog ( Fig. 2A, trace c). In this experiment, the cells were preincubated with thapsigargin prior to assaying exchange activity. Thapsigargin inhibits the Ca 2ϩ -ATPase in the endoplasmic reticulum (22), leading to the release of Ca 2ϩ from internal stores. This treatment accelerates the development of inhibition by ceramide and yields a greater initial inhibition of Ba 2ϩ influx (cf. "Discussion"). Under these conditions, C2-and C6-ceramide inhibited Ba 2ϩ influx by 55 and 51%, respectively (p Ͻ 0.005). As shown in trace b in Fig. 2A, C2-dihydroceramide, a biologically inactive analog (14), had only a small effect on exchange activity. The initial rate of Ba 2ϩ influx was reduced 25% by C2dihydroceramide (p Ͻ 0.05), but at later times, the rate of Ba 2ϩ influx in the presence of C2-dihydroceramide was nearly identical to that of the control cells.
Ceramide and Ca 2ϩ Efflux-The traces in Fig. 3 depict the FIG. 1. Inhibition of Na ؉ /Ca 2؉ exchange by C2-ceramide. A, CK1.4 cells were loaded with fura-2 and preincubated for 5 min in Na-PSS ϩ 1 mM CaCl 2 . The cells were assayed for Na ϩ /Ca 2ϩ exchange activity in 20/120 Na/K-PSS ϩ 0.3 mM EGTA containing 1 g/ml gramicidin, with or without 15 M C2-ceramide, as indicated. BaCl 2 (1 mM) was added (arrow) to initiate exchange mediated Ba 2ϩ influx. B, rates of Ba 2ϩ influx, given by the slopes of the traces in A between 65-95 and 200 -300 s, as indicated (n ϭ 5). RU/s, ratio units/second. effects of C2-ceramide on Na o ϩ -dependent Ca 2ϩ efflux. In this experiment, cells were placed in either Na-or K-PSS, and thapsigargin (1 M) was added 60 s later. Following the addition of thapsigargin, an increase in [Ca 2ϩ ] i was observed because of the release of Ca 2ϩ from internal stores. The traces in panel A for control cells show that the rise in [Ca 2ϩ ] i was reduced in Na-PSS compared with K-PSS; as discussed in detail elsewhere (21), this difference is due to Ca 2ϩ efflux by the Na ϩ /Ca 2ϩ exchanger. When 15 M C2-ceramide was included in the assay medium, the difference between Na-and K-PSS was substantially reduced, indicating that exchangemediated Ca 2ϩ efflux was inhibited by the presence of ceramide. The peak rise in [Ca 2ϩ ] i in K-PSS was reduced in the ceramide-treated cells compared with untreated cells, and a small, gradual rise in [Ca 2ϩ ] i was observed during the 30 s prior to the addition of thapsigargin (Fig. 3B). The rise in [Ca 2ϩ ] i evoked by ceramide was clearly evident during a 3-min incubation period (data not shown), suggesting that ceramide itself induced a slow release of Ca 2ϩ from the stores.
Ceramide and Mutant Exchangers-The exchange protein consists of multiple transmembrane segments with a large hydrophilic domain of 546 amino acids on the cytosolic membrane surface between the fifth and sixth transmembrane segments (4,5). The hydrophilic domain is thought to be essential for the regulation of exchange activity because its elimination by proteolysis or mutagenesis eliminates normal regulatory behavior (23). The data in Figs. 4 and 5 show the effects of ceramide in cells expressing several different mutations involving the hydrophilic domain. The traces in Fig. 4A show that C2-ceramide had only a modest effect on exchange activity in cells expressing a mutant, ⌬(241-680), in which 440 amino acids were deleted from the exchanger's hydrophilic domain. The initial rate of Ba 2ϩ influx was reduced by 28% in the presence of C2-ceramide (p ϭ 0.05, paired t test), a value similar to that seen with the wild-type exchanger when treated with the biologically inactive C2-dihydroceramide (Fig. 2). As with C2-dihydroceramide, little or no inhibition was observed during the later portions of the time course of Ba 2ϩ uptake.
A 20-residue region at the N-terminal end of the hydrophilic domain plays an important role in exchanger regulation (24 -26). This region contains positively charged and hydrophobic residues in a pattern that suggests an amphipathic helix. A peptide corresponding to this region inhibits Na ϩ /Ca 2ϩ exchange activity when applied to the cytosolic membrane surface, and this has therefore been designated the XIP region for exchange inhibitory peptide (25). Site-directed mutagenesis studies have shown that substitution of key positively charged residues in this region with uncharged amino acids abrogates sodium-dependent inactivation, suggesting an important contribution of this region to regulatory behavior. In Fig. 4, the traces in Panel B show the effect of ceramide on cells expressing a mutant exchanger, designated XIPA, in which all the positively charged residues in the XIP region have been substituted with alanines. The data show that C2-ceramide strongly inhibits exchange activity in this mutant. Traces a and b in Fig. 4B show Ba 2ϩ influx in a Na ϩ -free assay medium; under these conditions, exchange activity is absent, and Ba 2ϩ enters cells through passive leak pathways. After correcting for passive Ba 2ϩ entry, the initial rate of Ba 2ϩ influx was found to be inhibited by 68% (p Ͻ 0.001) in the presence of C2-ceramide. The biologically inactive C2-dihydroceramide (15 M) had no effect on exchange activity in this mutant (data not shown).
The behavior of cells expressing a small deletion, ⌬(680 -685), within the hydrophilic domain of the canine cardiac exchanger is presented in Fig. 5A. The activity of this mutant is not regulated by cytosolic Ca 2ϩ and shows little or no Na ϩ -dependent inactivation (27). The extent of Ba 2ϩ influx in these cells was reduced compared with wild-type cells (see below), and so 5 mM BaCl 2 was used in these experiments to increase the signal obtained. As shown in Fig rate of passive Ba 2ϩ influx in the absence of Na ϩ and demonstrate that exchange activity can readily be distinguished from passive Ba 2ϩ entry. The data in Fig. 5B show the results of experiments carried out under identical conditions with cells expressing the wild-type canine cardiac exchanger. In these cells, the rate of Ba 2ϩ influx, after subtracting the rates of passive Ba 2ϩ entry (traces a and b), was inhibited by 64% (p Ͻ Ͻ 0.001) in the presence of 15 M C2-ceramide. The time course of Ba 2ϩ entry for the canine exchanger under these conditions was similar to that for cells expressing the bovine exchanger (Fig. 2). However, for cells expressing the ⌬(680 -685) deletion mutant, the rate of Ba 2ϩ influx declined markedly within 40 s following the addition of Ba 2ϩ . Thereafter, the traces were nearly parallel to the traces for passive Ba 2ϩ influx (traces a and b in Fig. 5A). The explanation for this curious behavior is not known.
Ceramide and [Ca 2ϩ ] i -Cytosolic Ca 2ϩ activates Na ϩ /Ca 2ϩ exchange activity by binding to regulatory sites located in the exchanger's central hydrophilic domain (reviewed in Refs. 1,9). To determine whether the effects of ceramide could be overcome by increasing cytosolic Ca 2ϩ , we preincubated cells with ionomycin in medium containing either 0.3 mM Ca 2ϩ or 0.3 mM EGTA, so that the initial values of [Ca 2ϩ ] i prior to adding Ba 2ϩ were 110 Ϯ 11 and 44 Ϯ 5 nM, respectively. As shown in Fig. 6, the initial rate of Ba 2ϩ influx (in the absence of ceramide) was 3.7-fold greater at the higher value of [Ca 2ϩ ] i (compare traces in panels A and B and rates in panel C). C2-ceramide (15 M) inhibited Ba 2ϩ influx in both cases but seemed to be less effective for the cells preincubated in EGTA (35% inhibition) compared with cells preincubated with 0.3 mM CaCl 2 (59% inhibition) (Fig. 6C). However, this difference disappeared when the data were corrected for the rate of passive Ba 2ϩ leakage into the cell in the absence of Na ϩ /Ca 2ϩ exchange. Traces a and b in Fig. 6A show passive Ba 2ϩ influx when the cells were preincubated and assayed in medium lacking Na ϩ . The Ba 2ϩ leak was very low under control conditions (trace b) but increased 2.3-fold (p Ͻ Ͻ 0.01) following the addition of C2-ceramide (trace a), suggesting that ceramide had increased the passive permeability of the cell membrane to Ba 2ϩ . This increase in passive Ba 2ϩ entry was not observed in the experiments described earlier (Figs. 4 and 5) and may reflect a leakiness induced by the combination of ceramide and ionomycin. Even in the presence of C2-ceramide, however, passive Ba 2ϩ influx was less than the rate of Ba 2ϩ influx because of Na ϩ /Ca 2ϩ exchange. When the passive rates of Ba 2ϩ influx were subtracted from the data obtained under the standard assay conditions for Na ϩ /Ca 2ϩ exchange, the inhibition by C2ceramide was essentially the same for cells preincubated in either 0.3 mM CaCl 2 or 0.3 mM EGTA (67% versus 62% inhibition, respectively). We conclude that within this range of [Ca 2ϩ ] i values, increased activation of exchange activity by cytosolic Ca 2ϩ did not antagonize the inhibitory effects of ceramide.
Two other features of the data in Fig. 6 deserve mention. First, the slope of the control trace in Fig. 6A increases with time following Ba 2ϩ addition; the rate of Ba 2ϩ influx at 150 -200 s is 84% greater than at 65-95 s (p Ͻ Ͻ 0.001). This pattern was also observed for the XIPA cells (Control trace, Fig. 4B). We have previously suggested that this behavior reflects the autoactivation of exchange activity by Ba 2ϩ , acting at the Ca 2ϩ regulatory sites to accelerate exchange activity as it accumulates within the cytosol. In the ceramide-treated cells, the traces remain linear throughout the entire time course, indicating that ceramide blocks this effect. Second, for the cells that were preincubated with 0.3 mM Ca 2ϩ , the slope of the trace for Ba 2ϩ influx showed a time-dependent decline in the presence of C2-ceramide (trace labeled C2 in Fig. 6B). By 200 -300 s, the rate of Ba 2ϩ influx became identical (104%) to that of cells FIG. 4. Effect of C2-ceramide in cells expressing ⌬(241-680) and XIPA mutations of the exchanger. A, cells expressing the ⌬(241-680) exchanger mutant were pretreated with thapsigargin and assayed for Ba 2ϩ influx with or without 15 M C2-ceramide as described in Fig. 2 (n ϭ 6). B, as in A, but with XIPA cells (see text; n ϭ 9 -10). For traces a and b, cells were pretreated with 1 M thapsigargin in K-PSS/EGTA and assayed for Ba 2ϩ influx in Na-free assay medium (K-PSS/EGTA ϩ 1 g/ml gramicidin) (n ϭ 3).

FIG. 5. Effect of C2-ceramide in cells expressing ⌬(680 -685) mutation of the exchanger (A) and the canine wild-type exchanger (B).
A, cells expressing the canine ⌬(680 -685) exchanger mutant were pretreated with thapsigargin and assayed for Ba 2ϩ influx with (n ϭ 10) or without (n ϭ 18) 15 M C2-ceramide; conditions were as described in Fig. 2. except that the Ba 2ϩ concentration was 5 mM. B, As in A., but with cells expressing the canine wild-type exchanger (n ϭ 10 -11). For traces a and b in both panels, the cells were pretreated with 1 M thapsigargin in K-PSS and assayed for Ba 2ϩ influx (5 mM Ba 2ϩ ) in sodium-free assay medium (K-PSS ϩ 1 g/ml gramicidin; n ϭ 9 -13). C, initial rates of Ba 2ϩ influx, given as the slopes of the traces in panels A and B between 64 and 82 s. preincubated in EGTA (p Ͼ 0.3). This was not simply due to saturation of the fura-2 or to Ba 2ϩ uptake achieving a steady state, because other traces with C2-ceramide (e.g. in Fig. 2) did not display a downward curvature over a similar range of fura-2 ratios. We suggest that [Ca 2ϩ ] i probably continued to decline following the addition of Ba 2ϩ in these experiments, leading to gradual decline in the Ca 2ϩ -dependent activation of exchange activity in the ceramide-treated cells.
Sphingosine Inhibits Exchange Activity-The ceramide analogs used in this study are short chain acyl derivatives of sphingosine. The data in Fig. 7 demonstrate that sphingosine itself (2.5 M) inhibited exchange-mediated Ba 2ϩ influx in cells expressing the wild-type exchanger but not in cells expressing the ⌬(241-680) mutant. At a concentration of 2.5 M, sphingosine did not make the cells leaky to Ba 2ϩ , as shown in Fig. 7A (trace a). At higher concentrations, however, a substantial Ba 2ϩ leak was induced, and so we were restricted to sphingosine concentrations Յ2.5 M. At this concentration, sphingosine inhibited the initial rate of exchange-mediated Ba 2ϩ influx by 68%, after correcting for the passive rates of Ba 2ϩ influx. The data in Fig. 7D show that sphingosine partially inhibited the Ca 2ϩ efflux mode of exchange activity. The difference between the traces for Tg-induced Ca 2ϩ release in K-PSS versus Na-PSS was reduced compared with the control cells (Fig. 7C), indicating a reduction in Na ϩ -dependent Ca 2ϩ efflux. Note that sphingosine also induced a substantial leak of Ca 2ϩ from internal stores, as shown by the increase in the fura-2 signal prior to the addition of Tg; this probably accounts for the more rapid rise and shorter duration of the Tg-evoked [Ca 2ϩ ] i transient in the presence of sphingosine compared with controls.
Endogenous Sphingolipids and Exchange Activity-DL-Threo-1-phenyl-2-palmitoylamino-3-morpholino-1-propanol (PPMP) and its shorter chain analog PDMP inhibit glucosylceramide synthase and sphingomyelin synthase in CHO cells (28) and other cell types (reviewed in Ref. 29). Because they block important pathways of ceramide utilization, they have been used to elevate endogenous ceramide levels in various cell types (30 -33). To determine whether an elevation of endogenous sphingolipids would mimic the effects of the exogenous short chain ceramide analogs, we incubated CK1.4 cells with 30 M PPMP for 60 min prior to assaying Na ϩ /Ca 2ϩ exchange activity. The results are shown in Fig. 8 and Table I.
The initial rate of Ba 2ϩ influx was inhibited by 49% in CK1.4 cells following the PPMP treatment ( Fig. 8A and Table I). In this experiment, the cells were treated with thapsigargin following the preincubation with PPMP and then assayed for Ba 2ϩ influx. When the experiment was repeated with cells expressing the ⌬(241-680) mutant (Fig. 8B), only 21% inhibition was observed, a value that was not statistically significant (p ϭ 0.09). With the CK1.4 cells (Fig. 8A) but, curiously, not the ⌬(241-680) cells, the initial value of [Ca 2ϩ ] i (prior to adding Ba 2ϩ ) was lower following PPMP treatment than for the control cells (Table I). The lower value for [Ca 2ϩ ] i probably reflects the reduced Ca 2ϩ entry and the greatly decreased size of internal Ca 2ϩ stores observed in the PPMP-treated cells (see below). Because exchange activity is highly dependent upon [Ca 2ϩ ] i in this concentration range (11), it was important to determine whether or not the inhibition induced by PPMP was merely a secondary consequence of the reduced [Ca 2ϩ ] i .
To examine this issue, we pre-incubated the PPMP-treated and control cells for 2 min with ionomycin in Na-PSS containing either 0.3 mM EGTA or 0.3 mM CaCl 2 , exactly as described above for Fig. 6. As shown in Fig. 8C (traces a) and in Table I, the initial values for [Ca 2ϩ ] i were identical for the PPMPtreated and control cells after exposure to ionomycin ϩ 0.3 mM CaCl 2 . Importantly, the rate of Ba 2ϩ influx under these conditions was inhibited by 42% in the PPMP-treated cells, indicating that the PPMP treatment inhibited exchange activity by a mechanism that was independent of changes in [Ca 2ϩ ] i . This experiment also shows that the initial rates of Ba 2ϩ influx for the control and PPMP-treated cells were stimulated 3.6-and 4.1-fold, respectively, in the cells incubated with 0.3 mM CaCl 2 compared with those incubated with EGTA ( Fig. 8D and Table I). Thus, PPMP treatment did not block the activation of the exchanger by Ca 2ϩ , despite the lower overall exchange activity. As in the experiments with C2-ceramide (Fig.  6), the rate of Ba 2ϩ influx for the PPMP-treated cells incubated with 0.3 mM CaCl 2 gradually declined following the addition of Ba 2ϩ and by 200 -300 s became identical to that for the EGTAtreated cells (110%, p ϭ 0.17). (Note that with this batch of cells, the initial values of [Ca 2ϩ ] i following treatment with ionomycin ϩ 0.3 mM CaCl 2 were lower than for the experiment in Fig. 6; the reasons for this variability are not known.) In two of the experiments shown in Fig. 8 (A and trace b. Cont in C), the slopes of the traces for Ba 2ϩ influx in control cells increased following the addition of Ba 2ϩ . In both cases, this upward curvature was not observed in the PPMP-treated cells. This behavior is essentially identical to that observed with C2-ceramide, as discussed in connection with Fig. 6.
Some additional observations with PPMP-treated cells are listed below, although detailed data will not be presented in most instances: (a) The inhibitory effects of PPMP required time to develop. For example, after 30 min of incubation with 30 M PPMP, inhibition of exchange activity was poor compared with cells incubated for 60 min. This behavior is consistent with a time-dependent accumulation of endogenous ceramide and is not compatible with a direct pharmacological effect of PPMP on exchange activity. Indeed, 30 M PPMP added directly to the cuvette, without preincubation, did not inhibit exchange activity. (b) The data shown in Fig. 8 (A and B) were obtained with cells that had been treated with thapsigargin before assaying for Na ϩ /Ca 2ϩ exchange activity. Without thap- sigargin, Ba 2ϩ influx in the PPMP-treated cells was inhibited by 64% relative to controls (Table I). Thus, thapsigargin treatment did not enhance or accelerate inhibition of exchange activity, in contrast to its effects with C2-ceramide ( Figs. 1 and  2). For the cells expressing the ⌬(241-680) mutant, a 35% inhibition (p Ͻ 0.02) of exchange activity by PPMP was observed when the thapsigargin treatment was omitted (Table I). Thus, the exchange activity of this mutant can be inhibited by PPMP, but less effectively than the wild type, under some experimental conditions. (c) PDMP produced similar effects to PPMP. However, higher concentrations and/or longer incubation times were required for PDMP. This finding is consistent with reports that PPMP is considerably more effective than PDMP in inhibiting glucosylceramide synthesis in intact cells (34). (d) Elevation of cellular ceramide levels has been shown to inhibit store-dependent Ca 2ϩ entry in T lymphocytes (35), and we have found that exogenous C2-ceramide inhibits store-dependent Ca 2ϩ entry in CHO cells. As a positive control for the studies conducted above, we observed that the influx of Mn 2ϩ , a Ca 2ϩ surrogate often used to provide an index of store-dependent Ca 2ϩ entry, was strongly inhibited in the PPMPtreated CK1.4 cells. Internal Ca 2ϩ stores (measured as ionomycin-or thapsigargin-releasable Ca 2ϩ ) were markedly reduced in the PPMP-treated cells, presumably because of the reduced Ca 2ϩ entry. It should be noted that store-operated channels do not contribute to Ba 2ϩ influx during the assay for Na ϩ /Ca 2ϩ exchange; the depolarized membrane potential resulting from the gramicidin treatment (see "Experimental Procedures") blocks any measurable Ba 2ϩ entry by this route (20). (e) Finally, we note that the PPMP treatment did not induce a passive leak for Ba 2ϩ , as measured in gramicidin-treated cells in the absence of Na ϩ .

DISCUSSION
The results presented here demonstrate that short chain ceramide analogs inhibited both the Ca 2ϩ influx and Ca 2ϩ efflux modes of Na ϩ /Ca 2ϩ exchange activity (Figs. 1-3). A biologically inactive ceramide analog, in which the essential double bond in the sphingosine moiety is hydrogenated, had only minor effects on exchange activity (Fig. 2). Moreover, certain mutant exchangers that are defective in their regulatory behavior were not inhibited by ceramide (Figs. 4 and 5). These data suggest that the effects of ceramide are not simply due to a generalized membrane perturbation but that ceramide acts in a biologically relevant manner and specifically targets the mechanisms that regulate exchange activity. Recent results FIG. 7. Sphingosine inhibits Na ؉ / Ca 2؉ exchange activity. A, fura-2 loaded CK1.4 cells were preincubated for 2 min in Na-PSS ϩ 0.3 mM EGTA with 1 M thapsigargin and then assayed for exchange activity as described in the legend to Fig. 1, with (SPH) or without 2.5 M sphingosine present (n ϭ 11-12). For traces a and b, the cells were preincubated in K-PSS with thapsigargin and assayed for Ba 2ϩ influx in K-PSS ϩ gramicidin with (a) or without (b) sphingosine. (n ϭ 5-6). B, as in A, except that cells expressing the ⌬(241-680) deletion mutant were used (n ϭ 4 -6). C, cells were preincubated for 5 min. in Na-PSS ϩ 1 mM CaCl 2 and placed in cuvettes containing either Na-PSS or K-PSS as indicated. Thapsigargin (1 M) was added after 60 s (n ϭ 6 -7). Vertical arrow, Tg. D, conditions were identical to C, except that 2.5 M sphingosine was included in the medium in the cuvettes (n ϭ 8).
FIG. 8. Endogenous sphingolipids inhibit exchange activity. A, CK1.4 cells were preincubated for 60 min with 30 M PPMP and loaded with fura-2 as described under "Experimental Procedures." The cells were then preincubated for an additional 2 min in Na-PSS ϩ 0.3 mM EGTA with 1 M thapsigargin and assayed for exchange activity as described in the legend to Fig. 1 (n ϭ 7). B, as in A, except that cells expressing the ⌬(241-680) mutant were used (n ϭ 9 -10). C, CK1.4 cells were preincubated with 30 M PPMP and loaded with fura-2 as described under "Experimental Procedures." The cells were then incubated for 2 min in Na-PSS ϩ 10 M ionomycin containing either 0.3 mM CaCl 2 (traces a) or 0.3 mM EGTA (traces b) and assayed for exchange activity as described in the legend to Fig. 1 (n ϭ 7). D, initial rates of Ba 2ϩ influx (34 -64 s) for the data in A-C, as indicated.
indicate that 15 M C2-ceramide also inhibits Na ϩ /Ca 2ϩ exchange currents in cardiac myocytes but does not inhibit either Na ϩ or Ca 2ϩ channel activity. 2 Ceramide is a central component of the sphingomyelin signaling pathway and plays a critical role in the activation of the caspase cascade that leads to apoptosis (14 -16). In cardiac myocytes, ceramide and sphingosine levels increase following ischemia/reperfusion (36,37) and after exposure of the cells to tumor necrosis factor ␣ (38). Ceramide induces a multitude of cellular responses, including induction of the stress-activated protein kinase pathway, activation or inhibition of various individual protein kinases, activation of protein phosphatase 2A, inhibition of mitochondrial respiration, activation of cytochrome c release from mitochondria, and inhibition of phospholipase D (reviewed in Refs. 14 -16). Moreover, ceramide can be rapidly converted to ceramide-1-phosphate, sphingosine, and sphingosine-1-phosphate, signaling lipids that have multiple effects of their own. The rapid time course of sphingolipidinduced inhibition of exchange activity in the present study suggests that these agents exert their effects directly on the exchanger rather than through one of the above signal transduction pathways. This conclusion is strongly supported by the recent finding that both sphingosine and C2-ceramide inhibit exchange activity when applied to excised patches from cardiac myocytes. 3 Short chain ceramide analogs sometimes induce effects that are not mimicked by increases in endogenous ceramide (reviewed in Ref. 39). Moreover, the short chain analogs may themselves bring about alterations in endogenous ceramide and/or sphingosine levels (35,40,41). To address these issues, we preincubated the cells for 60 min with PPMP, an agent that blocks conversion of ceramide to glucosylceramide and has been used to elevate endogenous ceramide levels in several different cell types (30 -33). As shown in Fig. 8, treatment with PPMP mimicked the effects of the short chain ceramide analogs on Na ϩ /Ca 2ϩ exchange activity. We conclude that exchange activity is inhibited by increased endogenous ceramide and/or sphingosine as well as by the exogenous analogs.
Ceramide and sphingosine did not inhibit exchange activity in certain regulatory-deficient mutants, and we therefore con-clude that the sphingolipids target one of the mechanisms that regulate exchange activity. The best characterized regulatory mechanisms involve two time-dependent processes that promote inactive states of the exchanger (6, 7). The first process is called "Na ϩ -dependent inactivation" and is observed in excised patches as an exponential decay of current to a steady-state value following application of cytosolic Na ϩ . Na ϩ -dependent inactivation is counteracted by the presence of phosphatidylinositol 4,5-bisphosphate (42), by high concentrations of cytosolic Ca 2ϩ (6), and by mutations involving key basic residues or tyrosines in the XIP region (24,43). The second regulatory process involves the interaction of Ca 2ϩ with high affinity regulatory sites within the central hydrophilic domain of the exchanger (7,44); the binding of Ca 2ϩ to these regulatory sites appears to be required for all modes of exchanger operation (1,9). Na ϩ -dependent inactivation does not appear to be involved in the effects of sphingolipids, because the XIPA mutant was as sensitive to inhibition by ceramide as the wild type (Fig. 4B); this mutant does not display Na ϩ -dependent inactivation because of the alteration of critical basic residues in the XIP region. Moreover, ceramide inhibited wild-type exchange activity equally well at high (95 mM) and low (9 mM) Na ϩ concentrations (data not shown). Because Na ϩ -dependent inactivation requires high concentrations of cytosolic Na ϩ (6), these results provide another indication that ceramide does not promote this inactivation process.
The findings with the various exchanger mutants and the effects of alterations in Ca 2ϩ homeostasis suggest that sphingolipids interfere with the regulatory activation of the exchanger by Ca 2ϩ . C2-ceramide did not inhibit the activities of the deletion mutants ⌬(680 -685) and ⌬(241-680) (Figs. 4 and 5), which are defective in both Ca 2ϩ -dependent activation and Na ϩ -dependent inactivation. As mentioned above, C2-ceramide did inhibit the activity of the XIPA mutant (Fig. 4B), in which regulatory Ca 2ϩ activation is intact, but Na ϩ -dependent inactivation does not occur (24,45). Thus, ceramide inhibited the activity of exchangers that display Ca 2ϩ -dependent activation but not in exchanger mutants in which this regulatory mechanism was defective.
Experimental conditions that altered Ca 2ϩ homeostatic processes also affected the response of exchange activity to ceramide. For example, the effects of ceramide developed slowly in TABLE I Effect of PPMP treatment on Na ϩ /Ca 2ϩ exchange activity Cells were incubated for 60 min with or without 30 M PPMP and loaded with fura-2 as described under "Experimental Procedures." They were then centrifuged, washed, and incubated (in the absence of PPMP) for an additional 2 min in 100 l of Na-PSS plus the additions shown under "Post-PMPP treatment." The concentrations of thapsigargin and ionomycin were 1 and 10 M, respectively. The cells were then transferred to cuvettes to assay for Ba 2ϩ influx. [Ca 2ϩ [ i values were determined during the 10-s interval prior to Ba 2ϩ addition. The significance level for differences in the rates of Ba 2ϩ influx (p) were determined by Student's t test (2-tailed (Fig. 1A). When internal Ca 2ϩ stores were depleted by prior treatment of the cells with thapsigargin or ionomycin, the development of the inhibition of ceramide was accelerated, and the traces for Ba 2ϩ influx no longer displayed a downward curvature (Figs. 2 and 6A). The results suggest that there is a link between the filling state of intracellular Ca 2ϩ stores and the susceptibility of the exchanger to ceramide inhibition. The basis for this observation is not known; perhaps filled Ca 2ϩ stores generate local gradients of elevated [Ca 2ϩ ] i in the vicinity of the exchanger, and this antagonizes ceramide inhibition. For PPMP-treated cells, inhibition of exchange activity in the absence of thapsigargin was immediate (Table I) and did not increase with time following Ba 2ϩ addition (data not shown). This observation is entirely consistent with the observations described above, because internal Ca 2ϩ stores in the PPMPtreated cells were already depleted, presumably because of an inhibition of Ca 2ϩ influx (see discussion under "Results").
In several experiments, control cells showed a time-dependent increase in the rate of Ba 2ϩ influx, as shown by the upward curvature of the traces (Figs. 4B, 6A, 7A, and 8, A and C, trace b. Cont). This behavior was not seen with the ⌬(241-680) mutant (Figs. 4B, 7B, and 8B), but it was observed with the XIPA mutant (Fig. 4B). We have previously suggested that the acceleration in Ba 2ϩ influx is due to the auto-activation of the exchanger by cytosolic Ba 2ϩ through its interaction with the Ca 2ϩ regulatory sites (11). In the presence of ceramide or sphingosine or in PPMP-treated cells, the corresponding traces for Ba 2ϩ influx remained linear throughout the entire time course, indicating that the sphingolipids blocked this process.
These considerations suggest that ceramide/sphingosine impairs the regulatory activation of exchange activity by Ca 2ϩ . Its precise mechanism of action is not known, however. It does not seem likely that these hydrophobic lipids would interact directly with high affinity Ca 2ϩ regulatory sites, because these are located in the hydrophilic domain of the exchanger. A more plausible possibility is that they interact with the transmembrane segments of the exchanger and stabilize the inactive conformation that is attained upon dissociation of Ca 2ϩ from its regulatory binding sites (Fig. 9). In this way, the sphingolipids might increase the K d for Ca 2ϩ activation, slow the conformational transitions involved, or block Ca 2ϩ activation of the exchanger altogether in a subpopulation of exchangers. In any event, it is clear that ceramide does not completely block Ca 2ϩ activation, because exchange activity was stimulated by increasing [Ca 2ϩ ] i in the ionomycin experiments, both in the presence of C2-ceramide (Fig. 6C) and in the PPMP-treated cells ( Fig. 8D and Table I).
In the experiments with ionomycin-treated cells (Figs. 6 and 8C), we sought to determine whether an increase in [Ca 2ϩ ] i would antagonize the inhibitory effects of ceramide or PPMP treatment, as expected if a competitive effect were involved. In both cases, however, the degree of inhibition was similar at high and low [Ca 2ϩ ] i . The highest value of [Ca 2ϩ ] i attained in these experiments was only ϳ110 nM, however, and it will be important to re-examine this issue in an experimental system (e.g. excised patches) that permits a broader range of [Ca 2ϩ ] i values to be studied. These experiments also showed that for the ceramide-or PPMP-treated cells at the elevated [Ca 2ϩ ] i , the rate of Ba 2ϩ influx declined gradually following Ba 2ϩ addition and eventually became equal to that seen at the lower [Ca 2ϩ ] i (Figs. 6B and 8C). This behavior probably reflects the gradual reduction in regulatory activation of exchange activity because the initially high level of [Ca 2ϩ ] i declined following Ba 2ϩ addition. In the corresponding experiments with control cells, the rate of Ba 2ϩ influx did not decline in this manner, consistent with our hypothesis that the ability of Ca 2ϩ and/or Ba 2ϩ to activate exchange activity is impaired by the sphingolipids.
What are the physiological implications of our results? Ceramide and sphingosine are elevated in cardiac myocytes during stress. The resulting inhibition of exchange activity could be viewed as a protective measure to preserve Ca 2ϩ stores and maintain contractile strength under stressful conditions. Alternatively, when cytosolic [Na ϩ ] is elevated, as in ischemia, inhibition of exchange activity could be a means of protecting the cell against Ca 2ϩ overload by reducing exchange-mediated Ca 2ϩ influx. A more interesting possibility, however, is that under nonpathological conditions endogenous sphingolipids and the exchanger work coordinately as a Ca 2ϩ -dependent feedback mechanism to control the distribution of exchangers between active and inactive states, in the manner described below.
The high affinity of the exchanger for regulatory Ca 2ϩ (K d ϳ50 nM) (11-13) would seem to provide no opportunity for meaningful regulation of exchange activity within a physiological range of [Ca 2ϩ ] i values. In a functioning cardiac myocyte, the small fraction of exchangers that become inactive because of dissociation of regulatory Ca 2ϩ during diastole ([Ca 2ϩ ] i ϳ100 nM) would be rapidly reactivated by the ensuing rise in [Ca 2ϩ ] i during the next systole. However, if endogenous ceramide/ sphingosine were to interfere with Ca 2ϩ -dependent activation of the exchanger by any of the mechanisms suggested above (Fig. 9), a fraction of the inactive exchangers would be retained in the inactive state (I 2 in Fig. 9) despite the rise in [Ca 2ϩ ] i . Over multiple contraction/relaxation cycles, the distribution of exchangers between active and inactive states would be determined by two principal factors: diastolic [Ca 2ϩ ] i and endogenous levels of ceramide/sphingosine. A fall in diastolic [Ca 2ϩ ] i would increase the population of inactive exchangers, thereby reducing exchange-mediated Ca 2ϩ efflux and eventually restoring diastolic [Ca 2ϩ ] i to its normal level. An increase in endogenous ceramide or sphingosine, e.g. during stress, would have the same effect and establish a new steady-state relation between diastolic [Ca 2ϩ ] i and exchange activity.
This hypothesis, although speculative, provides a welcome framework for understanding the physiological role of Ca 2ϩ -dependent activation of exchange activity in light of the high FIG. 9. Proposed mechanism of ceramide inhibition. When [Ca 2ϩ ] i dissociates from regulatory binding sites in the exchanger's hydrophilic domain, a conformational transition from an active state to an inactive state (I 2 ) occurs. The designation I 2 is to distinguish this inactive state from the one produced by Na ϩ -dependent inactivation (I 1 ; Refs. 6 and 7). Elevation of [Ca 2ϩ ] i (left arrow) leads to refilling of the regulatory binding sites and a rapid reactivation of exchange activity because of re-establishment of the exchanger's active state. We propose that ceramide stabilizes the I 2 state through an interaction with the transmembrane regions of the exchanger and thereby interferes with the I 2 3 A transition. In beating cardiac myocytes, dissociation of Ca 2ϩ from regulatory sites during diastole would act in conjunction with endogenous ceramide to govern the distribution of exchangers between the A and I 2 states. affinity of the exchanger for regulatory Ca 2ϩ (11-13). In considering the physiological role of sphingolipid-exchanger interactions, it will be essential to define more precisely its effects on Ca 2ϩ -dependent activation and the influence of internal Ca 2ϩ stores on this process. Work toward this end is currently in progress.