The Hepatic Angiotensin I1 Receptor I. CHARACTERIZATION OF THE MEMBRANE-BINDING SITE AND CORRELATION WITH PHYSIOLOGICAL RESPONSE IN HEPATOCYTES*

The angiotensin I1 binding sites on hepatic plasma membranes were characterized using the lZ6I-radiola-beled hormone and a partially purified preparation of rat liver membranes. Equilibrium binding studies carried out under optimal conditions yielded a curvilinear Scatchard plot. The nontransformed data were evaluated by computer and optimally fit to a two-site model with a high affinity site Kdl = 0.21 f 0.06 nM and NI = 229 f 72 fmol/mg of protein and a low affinity site K a = 2.9 f 0.5 nM and Nz = 1820 f 185 fmol/mg of protein. The angiotensin antagonist, '261-sarcosine',a1anine8-an- giotensin I1 (saralasin) bound to only one class of sites, corresponding to the low affinity high capacity site for angiotensin I1 with a & = 3.3 k 0.8 nM and N = 1780 f 171 fmol/mg of protein. The divalent cations Ca2+, M&+, and Mn2+ stimulated binding of angiotensin II greater than %fold, with half-maximal stimulation of binding occurring at 0.6-0.9 mM for the three ions. In contrast, saralasin binding was relatively unaffected by divalent cation concentrations between 0-2.0 mM and was inhibited at divalent ion concentrations above 2.0 mM. The high affinity binding site for angiotensin 11 was undetectable in the presence of 2 mM EDTA and progressively increased to 398 f 213 fmol/mg of protein as the level of free M&+ value which covers this 958 confidence range in both directions. All other data obtained from pooled experiments are expressed as means -C S.E. Preparation of Hepatocytes and Measurement of Enzyme Actiui- ties-Isolated liver cells were prepared from 200- to 300-g fasted male Wistar rats by published methods (27). The hepatocytes were resus- pended in Krebs-Ringer bicarbonate buffer to a concentration of about 20 mg of protein/ml and gassed with 95% 0 2 , 5%' Con. One to two milliliters of cell suspension were preincubated for 30 min with 16 mM L-lactate and 4 mM pyruvate under 95% 02, 5% COS at 37 "c. Following the preincubation, cells were stimulated with the various angiotensin analogues for 2 min prior to sampling for the phospho- rylase assay and for 5 min prior to sampling for the glycogen synthase assay. These sample times have been determined to give optimal responses (4). Cell extracts for the assay of phosphorylase and glycogen synthase were prepared by published methods (27). Phospho- rylase activity was assayed according to the method of Stalmans and Hers (28) adapted to the filter paper technique, and glycogen synthase was assayed as described by Thomas et al. (29).

The angiotensin I1 binding sites on hepatic plasma membranes were characterized using the lZ6I-radiolabeled hormone and a partially purified preparation of rat liver membranes. Equilibrium binding studies carried out under optimal conditions yielded a curvilinear Scatchard plot. The nontransformed data were evaluated by computer and optimally fit to a two-site model with a high affinity site Kdl = 0. A positive correlation was established between the ability of angiotensin analogues to stimulate phosphorylase activity in intact hepatocytes and the analogues' potency in competing for angiotensin I1 binding sites in liver membranes. Together these data fulfill the criteria for positive identification of a hormone receptor including saturability, reversibility, specificity, and correlation with biological activity. These results provide strong evidence that this membrane-binding site is the hormone receptor which mediates the metabolic effects of angiotensin I1 in the liver. To whom correspondence should be addressed.
Although the liver is not usually considered a target tissue for angiotensin 11, this organ exhibits a number of responses to AII' including stimulation of glycogenolysis (1, Z), stimulation of gluconeogenesis (3,4), inhibition of fatty acid synthesis (5, 6), and stimulation of renin substrate production (7). In addition, preliminary studies have reported a hepatic binding site for angiotensin I1 (8, 9).
Physiologically, angiotensin I1 mimics the action of glucagon on carbohydrate metabolism in the liver. Glucagon acts in liver via the well understood mechanisms involving adenylate cyclase (lo), cyclic AMP ( l l ) , and the cyclic AMPdependent protein kinase (4,12,13). In contrast, relatively little is known about the mechanism of action of angiotensin 11, although it appears to act through a Ca"-requiring cyclic AMP-independent pathway. The available data suggest that a possible sequence of angiotensin I1 action might be: binding of the hormone to a membrane receptor (8,9); stimulation of phosphatidylinositol breakdown (14, 15); elevation of cytoplasmic Ca2+ levels (16, 17); activation of one or more Caz+dependent protein kinases which phosphorylate the regulatory enzymes of carbohydrate metabolism (4).
Prior to investigating the biochemical events involved in the hepatic actions of angiotensin 11, a binding assay for rat liver angiotensin I1 receptors was developed. The binding sites demonstrate high affinity, specificity, saturability, reversibility, appropriate kinetics, and correlation of binding with a biological effect. These characteristics suggest that the binding sites possess the properties expected of a hormone receptor and that these receptors mediate the metabolic effects of angiotensin I1 in the liver.
Ci/pg) were from New England Nuclear; anti-saralasin antibody was a gift from Dr. W. B. Campbell, Southwestern Medical School, Dallas, TX. The following reagents were obtained from Sigma Chemical Co.: fraction V bovine serum albumin, Tris, EDTA, chloramine-T (Nchloro-p-toluenesulfonamide sodium), /I-mercaptoethanol-type I, dithiothreitol, and bacitracin. No. 30 glass fiber filters (25 mm diameter) were purchased from Schleicher and Schuell.

Methods
Preparation of Rat Liver Plasma Membranes-Rat liver membranes were prepared from 200-to 300-g male Wistar rats according to the protocol (through step 10) of Pohl (18) which is based on the original procedure of Neville (19). However, 10 mM Tris (pH 7.4) was substituted for 1 mM NaHC03, and dithiothreitol (1 or 5 mM) or /3mercaptoethanol (0.1%) was added to preserve adenylate cyclase activity. For experiments testing the effects of ions on binding of angiotensin to its receptor, membranes were prepared in a buffer consisting of 5 mM EDTA in 10 mM Tris, pH 7.4. Membranes were stored in liquid NP for up to 4 months at a concentration of 5-6 mg of protein/ml without loss of binding activity.
Electron micrographs of the liver plasma membranes revealed that the preparation was composed of long continuous sheets of membrane folded back on itself giving an overall vesicular appearance (not shown). In membranes prepared in EDTA, the basal activity of the marker enzyme adenylate cyclase was 9 pmol of cAMP/mg of protein/ min which increased to 214 pmol of cAMP/mg of protein/min in the presence of lo-" M glucagon and M GTP. These adenylate cyclase activities are in the range of those previously reported for this membrane preparation (20). Adenylate cyclase was assayed as described in the accompanying article (21).
Iodination of Peptides a n d Determination of Specific Activity a n d Purity-Angiotensin 11 and saralasin were iodinated using the chlor-Freedlender and Goodfriend (23). The products of the iodination amine-?' procedure of Hunter and Greenwood (22) as modified by reaction were separated using a DEAE-Sephadex A-25 column, and the monoiodinated peptide fraction was collected and stored in 200pl aliquots in 0.8 M glycine buffer, pH 7.5, at -75 "C. Aliquots of the '"I-labeled peptide were thawed immediately before use in radioligand-binding studies and never reused. The purity of the iodinated peptides was evaluated by thin layer chromatography on Eastman plastic-backed cellulose plates developed in a-butyl alcoho1:acetic acid:pyridine:water (15:3:10:12). About 95-100% of both the I2'II-AII prepared in this laboratory and that purchased from New England Nuclear co-migrated with the angiotension standard in the chromatography system. The purity of the I2'I-saralasin was judged to be 85-90% by the same criteria. The specific activity of the "'1-saralasin was 926 Ci/mmol as determined by the self-displacement method of Berson et al. (24) using a rabbit antibody to saralasin. The "'I-AI1 used for saturation binding isotherms was obtained from New England Nuclear and used according to its stated specific activity. While the specific activity of the angiotensin I1 iodinated in this laboratory was not measured, based on its binding activity it appeared to have a specific activity similar to that obtained from New England Nuclear. experiments (20 mM HEPES, 150 mM NaCI, 0.2% crystalline BSA, Binding Experiments-The buffer originally chosen for binding pH 7.4) was similar to the conditions employed by Glossmann et al. (25). However, experiments designed to optimize binding conditions resulted in conversion to another buffer: 20 mM Tris, 100 mM NaCI, 10 mM MgCI2, 0.2% crystalline BSA, pH 7.5 (see "Results"). Though there was no difference in binding whether Tris or HEPES was used as the buffer, Tris was necessary for experiments in which the absence of sodium was desired and was generally adopted. The concentration of l2'I-A1I was 0.1-0.3 nM in all binding experiments. For binding isotherms, unlabeled hormone was added to the ''2''I-AII to achieve higher concentrations. Time course studies were done to define equilibrium conditions for the range of hormone concentrations used in saturation binding isotherms. The binding reaction was initiated by the addition of liver membranes (0.5 mg of protein/ml, final concentration) and incubated in a total volume of 100-200 p1 for 1 h at 12 "C. Five ml of ice-cold 20 mM Tris was added to the assay tube, and bound and free hormone were separated by filtration using glass fiber fiiters which had been presoaked at room temperature in a 0.2% solution of fraction V BSA. This was followed by a 5-ml rinse of the incubation tube. Pilot experiments were performed to optimize rinse volume and composition. Controls were included in all experiments to monitor changes in binding activity between the first and last samples. In time course experiments a 50-pl aliquot was removed from the incubation tube a t varying times, filtered, and rinsed twice with 2 ml of Tris buffer. The filters were counted in a Beckman Biogamma Counter with an efficiency of 50-55%. Experimental results are expressed as saturable binding, which was defined as that portion of total binding able to be displaced by an excess concentration ( 1 0~~5 M ) of unlabeled ligand. Saturable binding constituted 90-95% of total binding a t '"I-angiotensin I1 concentrations of 0.1-0.3 nM.
Hormone Metabolism Studies-Metabolism of angiotensin I1 by rat liver plasma membranes was examined by incubating 0.1-0.3 nM "'I-labeled ligand with membranes at a concentration of 0.5 mg of protein/ml for 60 min a t 0, 12, 22, and 37 "C. Bound and free ligand were separated by centrifugation. The bound hormone was solubilized by adding 100 pl of 50% glacial acetic acid to the pellet followed by incubation in a boiling water bath for 5 min. The free and bound hormone, along with radiolabeled angiotensin standards, were spotted on an Eastman plastic-backed cellulose thin layer chromatography plate and developed in n-butyl a1cohol:acetic acid:pyridine:water (15:3:10:12). Plates were scanned with a Berthold thin layer radio scanner as well as cut up into strips and counted in a Beckman Biogamma Counter. High voltage paper electrophoresis was also used to monitor metabolism of bound and free "'I-angiotensin I1 using a solvent system of water:acetic acid:pyridine (272:l) at pH 4.2.
Analysis ofBinding Data-The dissociation constants and number of binding sites for all saturation binding isotherms were determined using a nonlinear least squares curve fitting program (26) to fit the data to a one-or two-site model based on the equation where B = ligand bound, L = free ligand concentration, N = concentration of binding sites, and K,, = the dissociation constant of the binding site. For a one-site model the second term on the right side of the equation is dropped. The program was executed on a Tektronix 4006 graphics terminal linked to a Control Data Corporation Cyber 720 computer. Goodness of fit was judged by the variance obtained for a given curve fit. A graphic display of the line drawn through the experimental points allowed a visual confirmation of the computercalculated fit. The computer program determined the parameters for the one-or two-site equation directly from the saturation binding data. However, the linearity or deviation from linearity of a Scatchard analysis more dramatically depicts the presence of one or more binding sites. For this reason, some of the data presented under "Results" have been transformed to a Scatchard plot.
Computer analysis of individual experiments yielded the binding site numbers and affinities with their 95% confidence limits. In the text and legends associated with Figs. 1 and 2 these binding parameters are expressed k the minimum value which covers this 9 5 8 confidence range in both directions. All other data obtained from pooled experiments are expressed as means -C S.E.
Preparation of Hepatocytes a n d Measurement of Enzyme Actiuities-Isolated liver cells were prepared from 200-to 300-g fasted male Wistar rats by published methods (27). The hepatocytes were resuspended in Krebs-Ringer bicarbonate buffer to a concentration of about 20 mg of protein/ml and gassed with 95% 0 2 , 5%' Con. One to two milliliters of cell suspension were preincubated for 30 min with 16 mM L-lactate and 4 mM pyruvate under 95% 02, 5% COS a t 37 " c . Following the preincubation, cells were stimulated with the various angiotensin analogues for 2 min prior to sampling for the phosphorylase assay and for 5 min prior to sampling for the glycogen synthase assay. These sample times have been determined to give optimal responses (4). Cell extracts for the assay of phosphorylase and glycogen synthase were prepared by published methods (27). Phosphorylase activity was assayed according to the method of Stalmans and Hers (28) adapted to the filter paper technique, and glycogen synthase was assayed as described by Thomas et al. (29).

Establishment of Optimal
Conditions for Binding-Preliminary experiments were carried out to provide information on the kinetics, ion, and temperature dependence and stability of binding. These experiments employed a buffer containing 100 mM NaC1,0.2% crystalline BSA, and 20 mM Tris, following the conditions of Glossmann et al. (25). The rate of association of angiotensin I1 with rat liver membranes was directly proportional to the temperature of incubation. Maximal binding occurred in 25 min at 22 "C compared to 90 min at 4 "C. The stability of binding was inversely proportional to temperature. At 22 "C binding began to decrease immediately from maximal levels, while a t 4 "C maximal binding was maintained for more than 30 min. Binding kinetics at 12 "C offered a compromise, with maximal levels being reached within 50-60 min and maintained for an additional 20 to 30 min.
Degradation of angiotensin I1 by the hepatic membranes could explain the short-lived steady state binding observed at 22 "C. However, a number of agents, including dithiothreitol, /I-aspartyl napthylamide, bacitracin, and the oxidized /I-chain of insulin, which have been shown to inhibit the degradation of angiotensin and other peptides (25,(30)(31)(32), failed to prevent the rapid loss of membrane-bound angiotensin I1 at 22 "C. Direct measurement of '"I-angiotensin I1 degradation by both ascending chromatography and high voltage electrophoresis indicated that there was negligible metabolism of bound "' 1-AI1 after a 1-h incubation with membranes a t 4, 12, 22, or 37 "C. Furthermore, there was no detectable metabolism of free '"I-AII following a 1-h incubation with membranes at 4, 12, or 22 "C. Even a t 37 "C, 90% of the free radioactivity comigrated with radiolabeled hormone standard. These data demonstrated that there was minimal degradation of the ligand and suggested that the angiotensin binding site itself might be labile. In support of this hypothesis, preincubation of membranes at 4, 12, 22, or 37 "C demonstrated a temperature-dependent decrease in 1251-AII binding. The greatest decrease in binding (70-10076) occurred after preincubation of the membranes for 75 min a t 37 "C.
Attempts to stabilize the AI1 binding activity of rat liver membranes revealed that 10 mM Mg'+ prevented the timedependent loss of binding when this ion was present during both the preincubation of membranes and the equilibration with the ligand. At all preincubation temperatures except 37 "C, Mg'+ prevented the loss of binding capacity. After preincubation at 37 "C, binding in the presence of Mg2+ was 50% of control values. Furthermore, the presence of Mg2+ in the incubation mixture was able to restore the binding activity lost after preincubating membranes in a Mg'"free buffer at 4, 12, 22, or 37 "C. Consequently, in most studies incubations were performed in 100 mM NaCI, 10 mM MgC12, 20 mM Tris, and 0.2% crystalline BSA, at 12 "C. However, in experiments designed to study ion dependence the buffer contained only Tris and BSA.
Determination of Binding Parameters for Angiotensin 11 a n d Saralasin- Fig. 1 shows the dose-dependent binding of I-angiotensin I1 to liver membranes. The relatively flat binding isotherm spans more than three orders of magnitude (Fig. lA), with a threshold concentration of 5 X M AI1 and saturation of sites occurring a t a concentration of about 2 X 10" M AII. As suggested by the shallow dose response curve, when the data were transformed according to Scatchard Since other investigators have reported marked differences between the total number of binding sites detected by aadrenergic agonist and antagonist ligands in rat liver plasma membranes (34), it was of interest to see if such differences existed for angiotensin I1 binding in hepatic membranes. To in a system such as the hepatic angiotensin receptor which potentially has two or more affinity states that are dependent on Mg'f ion (Table  I) and guanine nucleotide (21). However, assuming that the association rates are similar for both the high and low affinity binding sites one can use the method reported by Fields et al. (35) to obtain an association rate for AI1 binding of 2.14 X 10' M" rnin". The dissociation rate for angiotensin I1 can be determined by adding an excess concentration of unlabeled ligand to the system after steady state binding conditions have been achieved. When this is done, a nonlinear least squares analysis yields a best-fit to a biphasic dissociation curve with rate constants of 0.064 min" and 0.0045 min". Assuming that the faster off-rate corresponds to the low affinity site and the slower rate represents the high affinity site, using the experimentally deter- IO this end, '251-saralasin, an antagonist radioligand, was prepared. Fig. 2 represents a typical saturation isotherm for Iz5Isaralasin binding to rat liver membranes. The data, expressed as a Scatchard plot, yielded an apparent straight line indicative of one class of binding sites. This finding was confirmed by computer analysis of the data which yielded a best fit to a single class of sites with K d = 3. Effects of Cations on Agonist and Antagonist Binding-In a number of other systems ions have been shown to enhance or depress hormone binding as well as have differential effects on agonists and antagonists (36)(37)(38)(39). Based on these findings it was of interest to investigate the ion dependence of IZbIangiotensin I1 and '"1-saralasin binding to liver membranes.
These experiments were performed with membranes prepared and washed several times in Tris buffer containing 5 mM EDTA to minimize their divalent cation concentration. As shown in the top half of Fig. 3, the divalent cations, Mg2+, Ca2+, and Mn2+, stimulated AI1 binding up to %fold over control levels. The ED, for this divalent cation potentiation of agonist binding ranged from 0.6 mM for Mg'+ to 0.9 mM for Mn". At divalent cation concentrations above 20 mM, binding gradually decreased but still remained well above control (0---0). After 60 min at 12 "C saturable binding was determined as described under "Methods." Data are expressed as "% contro1/100" which is equivalent to "-fold increase over control," control being 20 m~ Tris, 0.2% crystalline BSA, pH 7.5, with no added ions. Data points represent pooled results from one, two, or three experiments, each one performed in triplicate. Standard errors of the mean were omitted for clarity's sake but were 58% in all cases. The inset in the bottom is a replot of the saralasin data with an expanded ordinate.  I-AI1 and rat liver membranes (0.5 mg of protein/ml) were incubated for 60-70 min a t 12 "C, and bound and free ligand was separated by filtration as described under "Methods." Binding parameters were also determined in the presence of Krebs-Ringer buffer with the bicarbonate being replaced by Tris (KR Tris). Each point was determined in triplicate with a control for nonsaturable binding. The data were computer fit to a one-or two-site model as described under "Methods." Results in this table are presented as the mean of [3][4][5] experiments f S.E. using three separate membrane preparations.
Statistical significance was evaluated using the unpaired Student's t test.
levels. The monovalent cations, sodium and potassium, had less effect on AI1 binding. Sodium-dependent binding exhibited a bell-shaped curve with a maximal %fold stimulation of binding at 100 mM NaCl. Potassium, from 2.0 to 30 mM, did not affect binding. The bottom half of Fig. 3 shows that in contrast to agonist binding, divalent cations between a concentration of 0.5 and 5.0 mM only slightly increased saralasin binding. However, at concentrations above 5.0 mM there was a marked decrease in saralasin binding with nearly complete inhibition at 100 mM divalent ion. The monovalent cations, sodium and potassium, did not display any potentiation of saralasin binding but at the higher concentrations also caused significant inhibition.
Since ions were found to have such a pronounced effect on angiotensin I1 binding, it was of interest to examine the effect of Mg2+ on the concentration or affinity of the two AI1 binding sites. Using EDTA-prepared membranes, the addition of 2 mM EDTA to the incubation mixture lowered AI1 binding by 25-75% depending on the particular membrane preparation. The EDTA in the assay also slightly lowered the half-maximal concentration of free Mg2+ needed for stimulation of AI1 binding from 0.6 to 0.5 mM. Angiotensin I1 saturation binding curves were generated in the presence of threshold (0.01 mM), half-maximal (0.5 mM), or maximal (10.0 mM) concentrations of free Mg2' as determined in pilot experiments. The data were fit by computer to a one-or two-site model as described under "Methods." In the presence of 2 mM EDTA without added Mg2+ no binding of '"I-AI1 to the high affinity low capacity site was observed (Table I). Binding data were best fit to a single class of low affinity sites with a Kd = 13.4 nM represent the mean of pooled data from two separate experiments, of unlabeled angiotensin analogues or other hormones. After incubaeach performed in triplicate. Standards errors of the mean were tion at 12 "C for 60 min, bound and free ligands were separated by omitted for clarity's sake but were 56% in all cases. Sur, sarcosine; filtration as described under "Methods." Data are corrected for non-Iso, isoproterenol. Kc/ values were obtained from competition curves in which angiotensin 11 (0.3 nM) was incubated with liver membranes (0.5 mg of protein/ml) in the presence of varying concentrations of the unlabeled analogues listed in this table. Saturable binding was determined as described under "Methods." ID5lI values were obtained graphically from the curve described by the data points for each analogue. The competiton curve for the AI1 data was also computer generated with a nonlinear curve fitting program and agreed well with the graphically determined curve. Kc/ values were obtained from the equation K, = IDw (1 -f ) where f equals fractional occupancy of the binding site by "'I-AIL Kact values were determined graphically from the data of Fig. 5. is defined as the concentration of agonist needed to provide 50% of maximal phosphorylase activation. These observations demonstrate that binding to the high affinity low capacity site is Mg"-dependent and that potentiation of angiotensin IT binding by Mg" is the result of changes in both the affinity and number of the two classes of binding sites.
Binding of angiotensin 11 to rat liver membranes was determined in a Krebs-Ringer buffer in which the bicarbonate was replaced by Tris to facilitate comparison of the binding results obtained in membranes with the biological effects of angiotensin I1 measured in intact hepatocytes. As seen at the bottom of Table I the affinity and concentration of both classes of binding sites obtained with Krebs-Ringer Tris buffer were not significantly different from those observed in the presence of 10 mM Mg'. Fig. 4 demonstrates the specificity requirements of the angiotensin molecule for interaction with its binding site in rat liver membranes. Substitution of sarcosine for aspartic acid in position 1 and several aliphatic amino acids for phenylalanine in position 8 converts the native hormone into a competitive antagonist in most angiotensin target tissuesi3 (42, 43). In liver membranes this class of analogues displayed an affinity for the AI1 binding site which was equipotent with the native hormone, 1le'"AII. Sar'-S-Me-Cys'-AII, a noncompetitive angiotensin antagonist in smooth muscle (44), was 65% as potent as AI1 in displacing '251-AII from liver membranes (data not shown).

Specificity for Hormone Binding and Correlation of Binding and Biological Potency-
Des-Asp',IleH AII, the .'In gut (40) and kidney cortex (41) sar', aliphatic amino acidnangiotensin 11s are full agonists. angiotensin I11 antagonist, had a Kd seven times greater than that of Ile5-AIL Asn',a-aminobutyrate'-AII, a potent agonist for stimulation of water transport in rat jejunum, had an IDso greater than 10 ~L M for inhibition of 12?-AII binding to liver membranes. Glucagon, isoproterenol, and vasopressin, other hormones which regulate carbohydrate metabolism in liver, did not compete for angiotensin binding sites in liver membranes.
The potency of a series of angiotensin agonists in displacing '2511-AII bound to liver membranes was compared with their ability to stimulate glycogen phosphorylase activity in intact rat hepatocytes (Fig. 5). Phosphorylase activation was chosen for this comparison because the response to angiotensin I1 is large, rapid, and closely correlated with receptor occupancy (45, 46). Table I1 compares Ang) agonists with their ability to displace the binding of lZ6I-angiotensin 11. Top, isolated hepatocytes were preincubated for 30 min, stimulated with the indicated angiotensin analogues for 2 min, and then sampled and assayed for phosphorylase activity. The data were normalized to % of control. Basal phosphorylase activity was 0.25 & 0.02 pmol of glucose-1-P/mg of protein/l5 min ( N = 12). All angiotensin analogues except des-Asp'-angiotensin I caused a maximal stimulation of phosphorylase equal to 0.70 5 0.06 pmol of glucose-1-P/mg of protein/l5 min (+285%). Each point represents the mean from 2-3 experiments, with each assay performed in duplicate. Bottom, Iz5I-angiotensin I1 (-0.3 nM) was incubated with liver membranes (0.5 mg/ml) in the presence of varying concentrations of unlabeled AII, AIII, AI, or des-Asp'-AI. After incubation at not identical for a given analogue, the relative potencies among the angiotensins for binding and biological activity were very similar. The K,, values for binding of angiotensin analogues were 3-to 5-fold greater than the respective KaCt values for glycogen phosphorylase activation. However, the reverse was true for des-Asp'-AI, the least potent agonist tested. Stimulation of phosphorylase activity by angiotensin I was not affected by captopril (data not shown), which indicated that the activity of the decapeptide was not dependent upon its metabolism to AI1 by angiotensin-converting enzyme. In addition to the close correspondence with phosphorylase activation there was also a good correlation between the K,, of AI1 binding (4.1 nM) and its ability to inhibit glycogen synthase activity (Kat, = 1.7 nM) in intact hepatocytes (data not shown).

DISCUSSION
This study characterizes the rat hepatic angiotensin I1 binding site and provides evidence in icating this site to be the receptor which mediates the meta t olic effect of angiotensin on carbohydrate metabolism in the liver. The following criteria, necessary for positive identification of a hormone receptor, have been satisfied: affinity commensurate with physiologic concentrations of the hormone (Fig. l), saturability (Fig. l), reversibility (21), specificity (Figs. 4 and 5), and most importantly, correlation of binding and biological effect of hormone analogues (Fig. 5). The binding affinities described here for liver are very similar to those seen by others for angiotensin I1 binding to rat adrenal cortex homogenate in which high and low affinities of 0.2 nM and 6.3 nM, respectively, were reported (25). Our data, which show two classes of angiotensin I1 binding sites, are similar to those reported in bovine (25), dog (47), and rat (48) adrenal cortical homogenate. In contrast, only one class of binding sites for angiotensin I1 was observed in rat mesenteric artery (37) and dog uterine homogenate (47). However, since the affinity and concentration of the two sites in liver membranes appear to be critically dependent on ionic and other experimental conditions (text, Fig. 3, and Table I), it is difficult to compare hepatic angiotensin binding data with that obtained in other tissues under different experimental conditions. The specificity for both hepatic binding and activation of glycogen phosphorylase by angiotensin analogues displayed the same rank order of potency if not absolute affinity as that observed by others in adrenal cortex (25,47,48) and vascular smooth muscle (37). In all of these systems the following order of potency was observed: AI1 2 AI11 > AI >> des-Asp1-AI. In terms of the binding potency of various angiotensin analogues relative to that of AII, our results agree most closely with data reported for vascular smooth muscle (37).
The possibility that the binding site described here is a degradative enzyme for angiotensin I1 is unlikely. The affinities of AI1 for the binding sites (Kds = 0.21 nM, 2.9 nM) are three to four orders of magnitude greater than the corresponding affinity for degradative enzymes of AI1 (K, = 10 PM) (49).
In addition, the specificity for the plasma membrane binding sites is much stricter than the substrate specificity for enzymes which metabolize angiotensin I1 (50). Finally, angiotensin I1 was not metabolized by this liver plasma membrane preparation.
In contrast to our studies, LaFontaine et al. (8) reported degradation of '"I-angiotensin I1 in rat liver plasma membranes. LaFontaine et al. also detected only one class of binding sites (Kc/ = 0.1 nM) while two classes of sites (Kc,, = 0.21 nM, Kc,, = 2.9 nM) were observed in this study. This lack of correspondence could be due to one of the following possibilities. 1) The procedure for the preparation of rat liver plasma membranes used by LaFontaine et al. was slightly different from that used in the study presented here.
2) LaFontaine et al. used '"I-labeled Asnl,Val"-AII (Hypertensin, Ciba-Geigy Chemicals) in their binding studies whereas '2511-Asp',Ile5-AII was used in the present study. Since these two peptides are degraded by human plasma aminopeptidases a t different rates (51, 52), it is also possible that they are degraded by rat liver plasma membranes to different degrees.
Angiotensin I1 affects the movement of ions in several tissues (41,53), and conversely, ions affect the interaction of angiotensin I1 with its receptors (36,37). Divalent cations (Mg"+, Mn", and Ca2+) caused a marked potentiation of angiotensin I1 binding to hepatic membranes indicating that divalent cations are important not only in the mechanism of action of angiotensin I1 in the liver cytosol but also in regulating the initial hormone-receptor interaction. Sodium caused a comparatively small (%fold) potentiation of AI1 binding while potassium had no effect. This profile of ion dependence of AI1 binding is very similar to that reported by Gunther et al. (37) for rat mesenteric artery homogenate, though the magnitude of binding potentiation in the mesenteric artery was much less than what we observed in rat liver plasma membranes.
There is a striking resemblance between the effects of cations on angiotensin binding presented here and the data reported by Williams et al. (38) for /3-adrenergic agonist and prostaglandin binding to frog erythrocyte membranes. These authors demonstrated that the divalent cations Mg'+, Mn2+, and Ca2+ stimulated /3-agonist binding with an EDSO of 0.4-0.7 mM. These results agree well with our observation that divalent cations stimulate angiotensin I1 binding to liver membranes with an ED5(, range of 0.5-0.9 mM. Furthermore, consistent with our results the stimulation of /3-adrenergic agonist binding was decreased at higher divalent cation concentration. More importantly, the potentiating effects of divalent cations on both /3-adrenergic and angiotensin binding were agonistspecific. For example, magnesium did not increase the affinity of propranolol binding to erythrocyte membranes (38) nor that of saralasin binding to liver membranes. Similarly, Bird and Maguire (39) reported that the potentiating effect of Mg'+ on the binding of P-adrenergic ligands to membranes from S49 lymphoma cells was agonist-specific. One distinguishing feature of the S49 membranes was that the effects of ions were limited to Mg".
Williams et al. (38) postulated that Mg'+ ion did not directly affect the interaction of ligand with its receptor but rather acted on adenylate cyclase or the guanine nucleotide regulatory protein to promote the formation of a high affinity receptor state. Perhaps more revealing from a mechanistic point of view are the data of Bird and Maguire directly implicating the involvement of the guanine nucleotide regulatory protein in the action of Mg2+ (39). Their work demonstrated a lack of Mg2+ stimulation of /3-adrenergic agonist binding in genetic variants of S49 lymphoma cells in which the guanine nucleotide regulatory protein is absent or functionally inactive. The parallels between the data obtained with /3-adrenergic agents and that presented in the study reported here suggest a similarity between angiotensin I1 and adenylate cyclase-linked hormones. This hypothesis is investigated further in the accompanying article.
Our data differ from that of Bird and Maguire (39) and Williams et al. (38) in that while we observed a Mg"-dependent increase in both receptor affinity and number, these investigators demonstrated an effect on receptor affinity only. It is not clear how Mg'+ is able to cause a change in the apparent number of hepatic angiotensin receptors. However, the differences between our results and those obtained with