Identification of NaK-ATPase Inhibitors in Human Plasma as Nonesterified Fatty Acids and Lysophospholipids*

Elevated plasma levels of factors with cardiac gly- coside-like activity have been implicated in the re- sponse to volume expansion in animals and in the path-ogenesis of certain human diseases. We recently described four fractions (IR1, EI1, E12, El3) from normal human plasma that inhibit NaK-ATPase, displace ouabain from the enzyme, and exhibit digoxin-like im- munoreactivity (Kelly, R. A., O’Hara, D. S., Canessa, M. L., Mitch, W. E., and Smith, T. W. (1985) J. Biol. Chem. 260, 11396-11405). In this report, we identify the active component of these plasma fractions as long-chain nonesterified fatty acids (NEFA) and lysophos- pholipids. These lipids were present in fractions EI1, E12, and E13 in quantities sufficient to account for all of the NaK-ATPase inhibitory activity. The digoxin-like immunoreactivity in fraction IR1 could be attrib- uted to hydrocortisone and other endogenous steroids. To explore the nature of the lipid-NaK-ATPase inter- actions, we examined the effects of various ATP or sodium concentrations on the NaK-ATPase activity measured in the presence of NEFA. Varying sodium did not affect the inhibition of NaK-ATPase by linoleic acid. At c0.15 mM ATP, linoleic acid stimulated NaK-ATPase, concentrated extract of plasma were injected onto a Waters HPLC system with a 3.9 mm X 30 cm phenylpropylsilane column (pBonda-pak phenyl) and eluted with a nonlinear 0-100% gradient elution system at 0.5 ml/min (31). One-ml fractions were collected, lyophilized, reconstituted in water, and assayed for cardiac glycoside-like activity with one of the assay techniques described below. The three fractions with NaK-ATPase inhibitory activity and digoxin-like immunoreactivity (EIl, EI,, E13) were then subjected to additional chromatographic steps. Each fraction was reconstituted in 75% acetonitri1e:water and injected onto a normal phase NH2 radial compression column (Waters) and eluted with a linear 75:0% aceto- nitri1e:water gradient at l ml/min. The middle fractions of each peak containing NaK-ATPase inhibitory activity were lyophilized, recon- stituted in 20% acetonitrile:water, injected onto a C18 3.9 mm X 30 cm reversephase column (C18 pBondapak, Waters) and eluted with a linear 20-80% acetonitri1e:water gradient at 1 ml/min. Again, the middle fractions of the peak with biologic activity were lyophilized and reconstituted in acetonitri1e:water. The fourth fractionation was accomplished using a NovaPak Cls radial compression column (Waters) and a shallow linear gradient elution system. The middle fractions of the peak with biological activity were analyzed by fast atom bombardment mass spectroscopy using a Finnigan MAT 112 mass spectrometer.

extracellular volume and peripheral vascular resistance (1,10,15,16,23,24). Although factors that might regulate the release of such a hormone and its tissue of origin have been the subject of considerable debate, there has been little disagreement regarding the identity of the cell-surface receptor for this hormone (25-31). It has been assumed that the hypothetical hormone inhibits active sodium transport by interacting with the digitalis binding site on the a-subunit of NaK-ATPase. Indeed, the fact that the digitalis binding site has been so highly conserved throughout phylogeny has led to a spirited search for native ligands for these sites.
The existence of high affinity ligands for these sites, i e . the cardiac glycosides, has naturally led to the postulate that any physiologically important sodium pump inhibitor would structurally resemble these plant-derived compounds. Consequently, techniques for assaying cardiac glycosides in plasma, including radioimmunoassays and erythrocyte sodium pump inhibition assays, have been used to search for endogenous NaK-ATPase inhibitors in biologic fluids. Recently, we reported that desalted deproteinized plasma from normal humans contains at least four fractions resolved by HPLC' (IR1, EI1, E12, and E13) that resemble the cardiac glycosides in that they are all low molecular weight heat-and protease-resistant compounds that appear to cross-react with digoxin-specific antibodies. Three of these fractions also inhibited erythrocyte sodium pump function and NaK-ATPase activity in uitro, and they displaced [3H]ouabain from its binding site on the enzyme (31). However, there were important differences between these plasma fractions and the cardiac glycosides, particularly in the nature of their interaction with NaK-ATPase (31). In this manuscript, we report the identification of biologically active components within fractions EI1, E12, and E13 in human plasma as nonesterified fatty acids (NEFA) and lysophospholipids.

MATERIALS AND METHODS
Extracts of plasma from normal humans were prepared as described previously (31). Briefly, plasma from normal humans was deproteinized by heating to 90 "C for 20 min and then centrifuged for 60 min at 35,000 X g. The supernatant was desalted by preparative CIS resin chromatography (Waters). The concentrations of sodium and potassium in the extract following this step were always 4 0 and C0.5 mM, respectively; vanadate was undetectable by atomic absorption spectroscopy. These steps concentrated the NaK-ATPase inhibitory activity as much as 200-fold over the activity measured in untreated plasma.

HPLC
The desalted deproteinized plasma extract was separated into four discrete fractions using the following HPLC method. Two ml of the The abbreviations used are: HPLC, high pressure liquid chromatography; NEFA, nonesterified fatty acids; Br-Mac, 4-bromomethyl-7-acetoxycoumarin. concentrated extract of plasma were injected onto a Waters HPLC system with a 3.9 mm X 30 cm phenylpropylsilane column (pBondapak phenyl) and eluted with a nonlinear 0-100% gradient elution system at 0.5 ml/min (31). One-ml fractions were collected, lyophilized, reconstituted in water, and assayed for cardiac glycoside-like activity with one of the assay techniques described below.
The three fractions with NaK-ATPase inhibitory activity and digoxin-like immunoreactivity (EIl, EI,, E13) were then subjected to additional chromatographic steps. Each fraction was reconstituted in 75% acetonitri1e:water and injected onto a normal phase NH2 radial compression column (Waters) and eluted with a linear 75:0% aceto-nitri1e:water gradient at l ml/min. The middle fractions of each peak containing NaK-ATPase inhibitory activity were lyophilized, reconstituted in 20% acetonitrile:water, injected onto a C18 3.9 mm X 30 cm reversephase column (C18 pBondapak, Waters) and eluted with a linear 20-80% acetonitri1e:water gradient at 1 ml/min. Again, the middle fractions of the peak with biologic activity were lyophilized and reconstituted in acetonitri1e:water. The fourth fractionation was accomplished using a NovaPak Cls radial compression column (Waters) and a shallow linear gradient elution system. The middle fractions of the peak with biological activity were analyzed by fast atom bombardment mass spectroscopy using a Finnigan MAT 112 mass spectrometer.

Assay Techniques
The assay techniques for measuring cardiac glycoside-like activity have been given in detail previously and will be described only briefly here (31). All reagents were obtained from Sigma unless otherwise specified.
NaK-ATPase Inhibition-The hydrolysis of [Y-~'P]ATP by NaK-ATPase, derived from canine kidney cortex or from the supraorbital salt glands of ducks (32), was used to determine NaK-ATPase activity. The specific activity of the canine kidney cortex enzyme averaged 90pmol of Pi/mg of protein/h and that of the avian enzyme -500 Fmol of P,/mg of protein/h. For both enzymes, the concentration of enzyme in the reaction mixture was 1 mg/ml, and the reaction was carried out in incubation buffer consisting of 100 mM NaC1, 50 mM Tris-C1 (pH 7.4), 0.25 m M Na,EDTA, 5 mM MgC12, and 5 mM ATP. Unless stated otherwise, enzymes, buffer, and the inhibitory fractions from plasma were preincubated for 2 h before the addition of KC1 (final concentration, 20 mM) and 0.3 Ci of [T-~*P]ATP. This mixture was incubated for 30 min at 37 "C; then the reaction was terminated by the addition of ice-cold activated charcoal suspension (Norit A, Fisher Scientific).
Radioreceptor Assay-The ability of the NaK-ATPase inhibitory fractions from plasma extracts to inhibit [3H]ouabain binding to NaK-ATPase was tested by preincubation with the enzyme in the incubation buffer for 2 h at 37 "C. 0.4 pCi of (3H]ouabain (New England Nuclear, 18 Ci/mmol) was added, and the incubation was continued for 30 min before the reaction was terminated by adding ice-cold buffer and filtering on 6 GF/C (Whatman) glass-fiber filters. The [3H]ouabain-enzyme complex on the filter was quantified by liquid scintillation spectrometry.
Radioimmunoassay-We used a variety of polyclonal and monoclonal anti-digoxin antibodies to test for the presence of apparent digoxin-like immunoreactivity (31). A dilution of each antibody was chosen to yield 50% binding of 10 nCi of a 1251-labeled histamine derivative of digoxin (New England Nuclear). Separation of bound from free labeled ligand was achieved using either dextran-coated charocoal or a second antibody precipitation technique.

Lipid Analysis
Preparatioe Chromatography-Preparative low-pressure aminopropyl bonded-phase columns were used to separate lipids contained in the three NaK-ATPase inhibitory fractions using the technique of Kaluzny et al. (33). The NaK-ATPase activity in each HPLC fraction was quantified by assaying serial dilutions of the fractions in the enzyme inhibition assay. Then, an aliquot of each HPLC fraction (EIl, E12, EIs) sufficient to cause 50% inhibition of NaK-ATPase was lyophilized and reconstituted in 0.5 ml of chloroform. This aliquot corresponds to -5 ml of plasma. Aminopropyl columns (Bond-Elut, Analytichem International, Harbor City, CA) were washed twice with 2 ml of hexane; then each of the HPLC fractions in chloroform was applied to a column. The column was then eluted with 4 ml of a 2:l mixture of chloroform:2-propranol to obtain eluate I containing neutral lipids. It was then sequentially eluted with 4 ml of 2% acetic acid in diethyl ether (to obtain nonesterified fatty acids) and 4 ml of methanol (to obtain the phospholipids). Eluate I was dried, reconstituted in hexane, and applied to a second aminopropyl column, which was eluted with hexane to obtain cholesteryl esters. A third aminopropyl column was then positioned in series below this second column, and 6 ml of hexane containing 1% diethyl ether and 10% methylene chloride were applied to elute the triglycerides. These tandem columns were eluted with hexane containing 5% ethyl acetate to obtain the cholesterol fraction. The columns were then separated, and the upper column was sequentially eluted with 4 ml of 15% ethyl acetate in hexane to obtain diglycerides and then 4 ml of a 2:l mixture of ch1oroform:methanol to obtain monoglycerides. This technique for separating complex lipid mixtures yielded recoveries of radiolabeled lipids that were >95% if 510 mg of total lipid were initially applied to each column, in agreement with the report of Kaluzny et al. (33). Following each solvent elution, the eluate was dried under nitrogen, reconstituted in water, and assayed for NaK-ATPase inhibitory activity.
Analysis of Nonesterified Fatty Acids-Each of the HPLC fractions (EI1, E12, EI3) with biologic activity was also analyzed for nonesterified fatty acid (NEFA) content using the technique of Tsuchiya et al. (34) with the modifications described below. This technique is capable of quantifying very small amounts of these compounds by derivatization of free carboxylic acids with a fluorescent derivitizing reagent, 4-bromomethyl-7-acetoxycoumarin (Br-Mac). Br-Mac was synthesized as described by Tsuchyiya et al. (35). Fatty acid standards or the HPLC fractions (EL, EI,, E&) were derivatized by incubating with 2-3 mg of potassium bicarbonate and sodium sulfate (Ll, v/v), 50 liters of dibenzo-18-crown-6 in acetone (40 nmol/50 pl), and 50 liters of Br-Mac in acetone (100 nmol/50 liters) in the dark for 30 min at 50 "C in a sealed glass ampoule. Heptadecanoic acid served as an internal standard. The derivatized NEFA were injected onto a Cle radial compression column (Nova-Pak, Waters) at room temperature and eluted at 2 ml/min using an increasing ratio of solvent B (90% methanol in water) to solvent A (methanol:acetonitrile:water, 35:35:30). A third low-pressure pump combined the eluate from the radial compression column with a solvent containing 0.2 M NaOH in 80% methanol at 0.4 ml/min. Alkaline hydrolysis of the Br-Mac-NEFA complex was accomplished in a postcolumn mixing coil at 50 "C to increase the fluorescence of the complex. Fluorescence was monitored at 365 nm excitation and 460 nm emission.

RESULTS
The NaK-ATPase inhibitory activity that we previously identified in fractions of desalted deproteinized human plasma was characterized as relatively low molecular mass (t2000 daltons), heat and protease resistant, and extractable into polar organic solvents (methanol or chloroform); the activity did not extract into hexane. As shown in Fig. 1, A and B, additional normal phase and reverse phase HPLC separations of fraction EI, yielded a compound that could be identified using fast atom bombardment mass spectrometry. The compound had a molecular weight of 496 corresponding to that of lysophosphatidylcholine with palmitic acid as the acyl group. As shown in Fig. 2, this same lipid inhibited NaK-ATPase activity in vitro; it also inhibited [3H]ouabain binding to the enzyme in the radioreceptor assay. In both assays, the steep concentration-effect curves for lysophosphatidylcholine were similar to that of fraction EI, (31). Interestingly, phosphatidylcholine itself was a much less active inhibitor (Fig.  2).
The biologically active factors within the other two HPLC peaks (EL and E13) were subjected to similar normal and reverse phase HPLC separations. In both cases, FAB-mass spectrometric analysis yielded a number of molecular ions in the 220-350-dalton range. There was no predominant molecular ion with a mass greater than 500 daltons that could not be attributed to a dimer or trimer of smaller molecular ions.
We tested the activity of a number of saturated and unsaturated fatty acids in both the NaK-ATPase inhibition assay and the 13H]ouabain radioreceptor assay; these assays gave equivalent results (31). Long chain mono-and polyunsaturated NEFA inhibited [T-~'P]ATP hydrolysis and [3H]oua- bain binding about equally, with an IC5o of approximately 20 FM (Fig. 3A). Saturated NEFA also inhibited [3H]ouabain binding, but only at concentrations approaching the limits of their solubility in aqueous media (Fig. 3A). Finally, NEFA inhibited the binding of labeled digoxin to antibody, but only at concentrations that were about 5-10 times higher than those that inhibited NaK-ATPase activity (Fig. 3B).
Since at least one of the factors responsible for NaK-ATPase inhibitory activity in plasma was a lipid and because others have reported that certain NEFA and phospholipids can interact with cell membrane proteins, we determined whether fractions EI1, E12, and EIs contained lipids. Each fraction was separated chromatographically into the major lipid classes and each lipid class was assayed for biological activity. All the NaK-ATPase inhibitory activity in peaks EII and E13 eluted in the NEFA fraction. The biological activity in fraction E12 was resolved into approximately equal fractions by solvents that elute NEFA and phospholipids. Importantly, with this chromatographic technique, none of the NaK-ATPase inhibitory activity present in any fraction could be detected in the void volume of these columns, nor was any biologic activity found when the column was eluted with solvents other than those eluting nonesterified fatty acids and lysophospholipids.
To identify and quantitate free fatty acids within peaks EI1, EL, and E13, an aliquot of each peak that yielded 50% inhibition of NaK-ATPase activity was derivatized with Br-Mac and separated using reverse phase HPLC. The concentrations of NEFA in each peak are shown in Fig. 4. By comparison with Fig. 3, it can be seen that peaks EI1 and E13 contained a sufficient quantity (approximately 20-25 PM) NEFA to account for the [3H]~~abain-displacing activity measured in the original aliquot. For peak E12, the total NEFA concentration was 13 PM, and this could account for approximately half of the ouabain-displacing activity present. The remainder of the activity in peak E12 can be attributed to lysophospholipids contained in this fraction.
Certain synthetic detergents at concentrations below their critical micellar concentration have been reported to affect substrate-NaK-ATPase interactions (36). To determine whether a similar property could be identified for physiologic concentrations of NEFA, we varied ATP and sodium concentrations and measured the capacity of NEFA to inhibit NaK-ATPase. There was a small but significant (p < 0.05) increase in NaK-ATPase activity induced by 25 PM linoleic acid at low ATP concentrations (<0.15 mM) (Fig. 5, A and B). At concentrations of ATP > 0.15 mM, linoleic acid inhibited enzyme activity. In addition, the concentration-effect curve for the inhibition of NaK-ATPase by linoleic acid was shifted to the right as the ATP concentration was reduced. Between 0.1 and 5 mM ATP, the ICso for linoleic acid more than doubled (Fig.  6). Changes in sodium concentration did not affect the ICso for linoleic acid when the concentration of ATP was held constant. The effect of 25 p~ linoleic acid on the sodium activation curve is illustrated in Fig. 7. Similar results (data not shown) were obtained with other mono-and polyunsaturated fatty acids.
We previously reported that a fourth fraction of normal human plasma, termed IRl, appeared to cross-react with antidigoxin antibodies, but did not inhibit NaK-ATPase or prevent [3H]ouabain binding to the enzyme. This fraction was further resolved using sequential normal and reverse phase  inhibition of NaK-ATPase activity in an enzyme inhibition assay were derivatized with 4-brornomethyl-7-acetoxycoumarin and separated by reverse phase HPLC using a radial compression C,, column; heptadecanoic acid was used as an internal standard. The amount of nonesterified fatty acids detected was less than the calculated enzyme inhibitory activity present only in fraction EI,. This fraction also contained palmitoyl lysophosphatidylcholine (Fig. 1B) which accounted for the remainder of the NaK-ATPase inhibitory activity in this fraction. HPLC and analyzed by mass spectroscopy. Although the factor or factors responsible for the immunoreactivity were not purified to homogeneity, no molecular ion with a mass >600 daltons could be detected. The peak did contain several NEFA, predominantly palmitoleic acid, but not in concentrations sufficient to have caused appreciable inhibition of Iabeled digoxin binding to antibody. The total NEFA content was only 5 PM at concentrations of IRI that yielded 50% inhibition of 1251-digo~in binding.
Because of reports that endogenous steroids could be responsible for at least some of the apparent digoxin-like immunoreactivity in plasma (37, 38), we examined the crossreactivity of a number of endogenous steroids with each of the antibodies used in the digoxin radioimmunoassay. Hydrocortisone coeluted with fraction IR, (Fig. 8), and the amount of cross-reactivity of hydrocortisone with each anti-digoxin antibody population paralleled the apparent immunoreactivity of peak IR, by each of the same antibody populations (Fig.   9). Using a specific hydrocortisone radioimmunoassay, the concentration of hydrocortisone present in a concentration of fraction IR, that yielded 50% displacement of 1Z51-digo~in binding to the rabbit polyclonal antibody was 960 nM, an amount that would account for about 30% of the apparent digoxin-like immunoreactivity detected in this peak.

DISCUSSION
The present data suggest that most, if not all, of the cardiac glycoside-like activity we identified (31) in plasma is due to lysophospholipids and nonesterified fatty acids. These lipids are present in the plasma fractions that inhibited NaK-ATPase activity and [3H]ouabain binding to NaK-ATPase in isolated membrane preparations (Figs. 2 and 3A). They also appeared to cross-react with polyclonal and monoclonal antidigoxin antibodies (Fig. 3B).
The present results shed light on the basis for the steepness of the concentration-effect curves for NaK-ATPase inhibition and [3H]ouabain displacement by each of the HPLC peaks; similarly steep curves were obtained using lysophosphatidylcholine and several NEFA. This phenomenon is likely due to the fact that an increasing lysophospholipid or unsaturated fatty acid content in isolated membrane fragments containing NaK-ATPase eventually will reach a threshold concentration that results in a phase transition of membrane lipids, thereby altering the conformation of all enzyme units (39). This conformational change could prevent hydrolysis of ATP and binding of ouabain to the enzyme. It is noteworthy that the effects of lysophospholipids and NEFA occurred a t levels below their critical micellar concentrations (40). Interestingly, in the digoxin radioimmunoassays, similarly steep concentration-effect curves were not obtained using any of the HPLC peaks. In this case, hapten-antibody binding was presumably inhibited by a discrete interaction between lipids and individual antibody binding sites. Since this apparent digoxin-like immunoreactivity always required concentrations of the HPLC fractions that were 5-10 times greater than the concentration inhibiting NaK-ATPase, it may have represented a nonspecific detergent effect of these lipids.
The fact that some NEFA and phospholipids can alter homeostasis (1,15,16,23). In most of these reports, the methods used to isolate and quantitate the endogenous inhibitors would not have excluded NEFA or lysophospholipids. Indeed, Tamura  NaK-ATPase from mammalian kidney appears to be a cytochrome P-450 derivative of arachidonic acid (45). Micromolar quantities of similar compounds have been reported by Jacobson and co-workers to inhibit active sodium transport by rabbit cortical collecting tubules (46, 47).
Both direct and indirect mechanisms have been suggested by which NEFA and certain phospholipids could modulate active cation transport across membranes. These lipids might interact directly with hydrophobic regions of enzymes to alter the enzyme's affinity for substrates or for modulating ions (36,48,49). It has been proposed as well that NEFA could cause changes in the boundary lipid environment of the enzyme, thereby altering conformational changes occurring during transport of ions (50,51). Changes in membrane fluidity, due to an increase in membrane NEFA content or alterations in the fatty acyl composition of membrane phospholipids, could also directly alter active transport systems (50,52,53). These modulating effects of lipids need not be limited to intrinsic membrane proteins, since changes in NEFA have been shown to alter the activity of a cytosolic proteinase from rat skeletal muscle (54). Finally, certain lipids might act as high affinity ligands for the cardiac glycoside binding site on NaK-ATPase, as has been implied for some cytochrome P-450 derivatives of arachidonic acid, but this remains to be conclusively demonstrated (45).
Indirect means by which NEFA could influence active cation transport have also been examined. NEFA may bind cations within cells or cell organelles, thereby reducing the activity of these ions (55). In addition, changes in membrane fluidity or in specific acyl groups that are covalently attached to proteins forming ionic channels could affect the rate of transport of ions through the membrane (48,56), thereby indirectly affecting their active transport rate.
A number of mechanisms exist within the intact cell to limit these effects of NEFA on membranes. Cells rapidly reesterify NEFA into existing membrane, cytosolic, or organellar lipids. If intracellular ATP levels were low, as could occur with ischemia, reesterification of NEFA would decrease, and NEFA would accumulate. This could explain the observation that NEFA and some esterified lipid intermediates accumulate during myocardial ischemia in association with reduced sodium pump function (57,58). To determine if the inhibitory activity of NEFA is affected by the ATP level, we measured NaK-ATPase activity in the presence of NEFA and various concentrations of ATP. As demonstrated in Fig. 5, there was a small but significant stimulation of NaK-ATPase activity a t low ((0.2 mM) ATP concentrations by 25 WM linoleic acid. As ATP concentrations were increased above 0.2 mM, an inhibitory effect of linoleic acid appeared and became progressively more marked. Similar results were obtained with other mono-and polyunsaturated fatty acids. These data may reflect an interaction of lipids with NaK-ATPase at a lowaffinity regulatory site for ATP (36,59). Other enzyme modulators were also examined for their effects on NaK-ATPase activity in the presence and absence of NEFA. Potassium is known to diminish the inhibition of NaK-ATPase activity caused by NEFA (43, 60); this effect was confirmed in our previous report (31), at least for peak E13. However, we did not note any effect of varying the sodium concentration on the inhibitory effect of linoleic acid on enzyme activity. It is possible that accumulation of NEFA or lysophospholipids might impair transmembrane cation flux in ischemia or in diseases characterized by abnormal or reduced lipid binding by plasma proteins in diabetes (61), chronic renal failure (2-4, 12, 62, 63), or Reye's syndrome (64), but it is not obvious how these lipids could influence ion transport in response to extracellular volume expansion with saline (18). Most of the NEFA and lysophospholipid in plasma are bound to plasma proteins and, therefore, are unavailable for interaction with NaK-ATPase and other membrane-bound proteins under normal physiologic conditions. However, there are other possible explanations for the observation that plasma levels of cardiac glycoside-like activity, which we believe to be fully accounted for by NEFA and lysophospholipid, rise in response to volume expansion. Since NEFA oxidation and sodium reabsorption are linked (65) and since active transport of sodium by the kidney falls with volume expansion, it is possible that plasma NEFA levels would rise in response to decreased renal extraction of NEFA for energy utilization. Alternatively, volume expansion might raise plasma levels of arachidonate precursors which could be used to form vasodilatory or natriuretic prostaglandins (66). In conclusion, we have shown that what has been termed cardiac glycoside-like activity in plasma is attributable to lysophospholipids and a number of NEFA, not just the arachidonate precursors. Nevertheless, the possibility that changes in plasma lipids may affect, either directly or indirectly, ion transport across cell membranes in certain disease states deserves further study.