nominally Ca2+-free solution for 45 to 60 minutes to remove surrounding cells. After injection of RNA or water the oocytes were incubated in 70% L-15 medium (Gibco) containing penicillin (100 µg/ml) and streptomycin (100 U/ml) at room temperature (20° to 25℃). The oocytes were voltage-clamped at 20° to 23℃ by the use of conventional two-micro- electrode techniques.
10. C. Baud, R. T. Kado, K. Marcher, Proc. Natl. Acad. Sci. U.S.A. 79, 3188 (1982).
11. N. Dascal, E. M. Landau, Y. Lass, J. Physiol. (Lon- don) 352, 551 (1984).
12. S. H. Thompson and R. W. Aldrich, in The Cell Surface and Neuronal Function, C. W. Cotman, G. Poste, G. L. Nicolson, Eds. (Elsevier/North-Hol- land, Amsterdam, 1980), pp. 49-85.
13. R. Miledi, Proc. R. Soc. London Ser. B 215, 491 (1982); M. E. Barish, J. Physiol. (London) 342, 309 (1983).
14. K. S. Cole and J. W. Moore, Biophys. J. 1, 1 (1960).
15. N. C. Anderson, J. Gen. Physiol. 54, 145 (1969); C. Y. Kao, in Biology of the Uterus, R. W. Wynn, Ed. (Plenum, New York, ed. 2, 1977), pp. 423-496; J.
Mironneau and J. P. Savineau, J. Physiol. (London)
302, 411 (1980);
C. Mironneau, J. Physiol. (Paris) 77, 851 (1981); G. Vassort, in Smooth Musde, E. Bulbring, A. F. Brading, A. W. Jones, T. Tomita, Eds. (Arnold, London, 1981), pp. 353- 366.
16. C. D. Benham and T. B. Bolton, J. Physiol. (London) 340, 469 (1983).
17. We thank M. Shanabrough for invaluable technical assistance and R. W. Tsien and R. W. Aldrich for their comments on the manuscript.
25 August 1986; accepted 16 December 1986
Coexistence of Guanylate Cyclase and Atrial Natriuretic Factor Receptor in a 180-kD Protein
ARANJANIYIL K. PAUL, RAVI B. MARALA, RAMA KANT JAISWAL, RAMESHWAR K. SHARMA
Atrial natriuretic factor (ANF) is a peptide hormone that is released from atria and regulates a number of physiological processes, including steroidogenesis in adrenal cortex and testes. The parallel stimulation of membrane guanylate cyclase and corticosterone production in isolated fasciculata cells of rat adrenal cortex has supported the hypothesis of a mediatory role for cyclic guanosine monophosphate (cyclic GMP) in signal transduction. A novel particulate guanylate cyclase tightly coupled with ANF receptor was purified approximately 273,000-fold by two-step affinity chromatography. The enzyme had a molecular size of 180 kilodaltons and was acidic in nature with a pI of 4.7. Its specific activity was 1800 nanomoles of cyclic GMP formed per minute per milligram of protein. The purified enzyme bound ANF with a specific binding activity of 4.01 nanomoles per milligram of protein, a value that is close to the theoretical binding activity of 5.55 nanomoles per milligram of protein for 1 mole of the ligand binding 1 mole of the receptor protein. These results indicate that the guanylate cyclase-coupled ANF receptor exists in a 180-kilodalton protein of rat adrenocortical carcinoma and represent a step toward the elucidation of the basic mechanism of cyclic GMP-mediated transmembrane signal transduction in response to a hormone.
S TUDIES WITH ISOLATED FASCICU- lata cells of rat adrenal cortex and rat adrenocortical carcinoma indicated a physiological mediatory role for cyclic gua- nosine monophosphate (cyclic GMP) in ste- roidogenic signal transduction and led to the proposal of a hypothetical working model in which membrane guanylate cyclase was the key enzyme in receptor-mediated cyclic GMP signal pathway [reviewed in (1)]. Until recently a strong bias existed against the presence of a hormone-depen- dent membrane guanylate cyclase in any endocrine or nonendocrine tissue. The belief was that there was only one guanylate cy- clase, a soluble enzyme, which was docu- mented to be hormone-independent and nonspecifically activated by a variety of ni- trite-generating compounds and agents that affect the oxidation-reduction potential of biological reactions (2, 3). These reserva- tions were overcome by the demonstration of two distinct types of guanylate cyclase- membrane and soluble-in rat adrenal and rat adrenocortical carcinoma cells; only the membrane enzyme is hormone-specific (4- 6).
More recently, the above results were corroborated in various rat tissues by dem- onstrating that atrial natriuretic factor (ANF) selectively stimulates particulate gua- nylate cyclase (7); these tissues included the rat adrenal gland (8). In vivo infusion stud- ies with rat adrenal venous blood (9) and in situ studies with isolated fasciculata cells of rat adrenal cortex showed that ANF stimu- lates the production of corticosteroids (10). Similarly, in mouse interstitial (11) and Ley- dig cells (12, 13), testosterone production is increased by ANF. The mechanism of the ANF-dependent stimulation of steroidogen- esis is not known, but the stimulation of membrane guanylate cyclase in parallel with the generation of an ANF-dependent ste- roidogenic signal suggested that this enzyme may have a role in mediating signal trans- duction (10).
Elucidating the biochemical mechanism of the mediatory role of cyclic GMP in receptor-mediated transmembrane signal transduction requires purification of the membrane guanylate cyclase. Only partial purification of any mammalian particulate guanylate cyclase has been achieved to date
(14). We now describe purification of the membrane guanylate cylase and demonstrate that this enzyme is tightly coupled with the ANF receptor.
Membranes isolated from rat adrenocorti- cal carcinoma cells were solubilized as in (15), adjusted to a final concentration of 5 mM MnCl2, and adsorbed onto a guanosine triphosphate (GTP)-agarose affinity resin, which was suspended in and extensively washed with buffer A [25 mM triethanol- amine hydrochloride (pH 7.6), 5 mM MnCl2, and 1 mM 3-[(3-cholamidopro- pyl) - dimethylammonio] - 1 - propanesul- fonate (CHAPS)] until there was no detect- able protein (absorbance at 280 nm). The guanylate cyclase was eluted at room tem- perature with 25 mM triethanolamine (pH 7.6) 1 mM CHAPS, and 2 mM EDTA. The pooled enzymic fractions were adjusted to 5 mM Mn2+ and adsorbed to the cyclic GMP- Sepharose, which had been equilibrated with buffer A. The resin was loaded onto a small column (1.6 by 8 cm); flow-through was cycled back on the column once; and the column was washed extensively with buffer A. The enzyme was eluted with buffer A containing 2 mM EDTA (crossed affinity purification step) (lane 5 in Fig. 1A). The enzyme was thus purified approximately 273,000-fold.
The homogeneity and authenticity of the membrane guanylate cyclase is shown by the following criteria. (i) Sodium dodecyl sul- fate-polyacrylamide gel electrophoresis (SDS-PAGE) of the purified protein shows a single stained band with a molecular mass of 180 kD (Fig. 1A); (ii) isoelectric focusing of the native and iodinated protein indicates a single symmetrical activity peak with a pI of 4.7 ± 0.10 (mean ± SEM) (Fig. 1, B and C); and (iii) Western blot analysis of the GTP affinity-purified enzyme shows a single 180-kD band although the SDS-PAGE of the GTP affinity-purified protein shows multiple Coomassie-stained bands.
The specific activity of the purified partic- ulate guanylate cyclase is 1800 nmol of
Department of Biochemistry, University of Tennessee, Memphis Center for the Health Sciences, Memphis, TN 38163.
Fig. 1. Purity of the particulate guanyl- A 2 KD 1 3 8 4 5 5 B ate cyclase-coupled ANF receptor. (A) SDS-PAGE of membrane guanylate Top 4 7 cyclase at successive purification steps. (Lane 1) Molecular weight markers, 3 D (lane 2) membranes (starting materi- 200 6 +180 2 PH al), (lane 3) solubilized membranes, (lane 4) GTP-affinity step, and (lane 5) 116 5 crossed affinity purification step (0.12 1 ug of protein). The gel was stained 94 Guanylate cyclase activity C 0 4 68 3000 7 43 2000 6 P.H 1000 5 (pmol/min per milliliter) with Coomassie blue. The molecular weight markers were myosin (120,000). B-galactosidase (116,000), phosphorylase b (94,000), bovine se- rum albumin (68,000), and ovalbumin (43,000). (B) Purified guanylate cy- clase (3 ng) was subjected to nondena- turing isoelectric focusing electropho- 125 | Incorporated (count/min) resis. Nondenaturing isoelectric focus- ing was performed with pure nonio- Dye 0 4 dinated enzyme and the 125I-labeled 0 10 20 30 enzyme as previously described (24). Fraction number The gel was sliced into 2-mm slices; the enzyme was extracted with 50 mM tris (pH 7.5) containing 10 mM MgCl2 and 1 mM CHAPS, and assayed for gua- nylate cyclase activity (4). (C) Nondenaturing isoelectric focusing of radioiodinated guanylate cyclase was performed as described above and counted for radioactivity. Data are means (± SD) from two individual experiments each on the native and 125I-labeled enzyme. Each assay was done in duplicate (25).
cyclic GMP formed per minute per milli- gram of protein (Fig. 2). That the enzyme is an authentic guanylate cyclase is demon- strated by immunological studies, which show that antibody to the 180-kD protein blocks up to 90% of the guanylate cyclase activity of the purified enzyme. It is note- worthy that the antibody blocks neither soluble guanylate cyclase nor adenylate cy- clase activities, indicating specificity for the membrane enzyme.
The pure guanylate cyclase binds 1251- labeled ANF. Bound ANF is displaceable by unlabeled ANF. Increases in the specific binding activity of ANF parallel the purifica- tion steps of the guanylate cyclase, reaching a mean ± SEM of 4.01 ± 0.45 nmol per milligram of protein in the pure enzyme (Fig. 2). This value approaches the theoreti- cal binding activity of 5.55 nmol per milli- gram of protein, if it is assumed that 1 mol of the ligand binds 1 mol of the receptor protein. Because of the limited quantity of purified enzyme, we could not determine the detailed ANF binding kinetics, but the Scatchard analysis of the particulate fraction showed a high ANF affinity (KD, 1.5 p.M) and one binding site. Although the purified guanylate cyclase binds ANF, the enzyme is not stimulated by ANF. These results are similar to those for the lung enzyme (16). Since the loss in response to ANF stimula- tion occurred during the detergent solubili- zation of the receptor, it is possible that a lipid component or an accessory protein necessary for original hormonal stimulation is lost in this purification step.
During the course of our investigations Kuno et al. (16) showed that in a highly purified rat lung preparation, ANF receptor and guanylate cyclase are copurified. Al- though the dual presence of the receptor and
Fig. 2. Coexistence of the particulate guanylate cyclase 4500 and ANF receptor at the successive purification steps. Guanylate cyclase activity 2000 was measured as in (4). The 3500 ANF receptor binding as- 1500 2500 says were performed by in- cubating 125I-labeled ANF (0.2 pmol per tube; specific 1000 1500 activity 93 Ci/mmol) with 500 samples at different purifica- 500 10 tion steps at 25℃ for 1 hour 2 in a total volume of 500 pl of incubation buffer [5 mM 1 5 Guanylate cyclase activity (nmol/mg per minute) MgCl2, 50 mM tris-HCI 0.02 0.1 (pH 7.5), 0.2% heat-inacti- 0.015 0.075 vated bovine serum albu- 0.005 0.05 0.025 min, 0.1 mM EDTA, and aprotinin (700 IU/ml)]. For 0.01 determination of nonspecif- Membrane Solubilized GTP-affinity Crossed ic binding, the sample buff- purified affinity purified ers contained, in addition, 1 p.M nonradioactive ANF. The reaction was terminated by the addition of 3 ml of ice-cold 0.9% NaCl (w/v) followed by immediate filtration through Whatman GF/C glass fiber filters for the particulate fraction, and through Whatman GF/B filters treated with 0.32% polyethyleneimine for solubilized and other purified fractions. The filter paper was washed three times with 5 ml of ice-cold buffer, dried, and counted for radioactivity. Results are shown as the means ± SEM (n = 6).
Specific ANF binding (pmol/mg)
enzyme has to be established in a homoge- neous lung protein, there are striking bio- chemical and kinetic differences between the lung and the tumor receptor-coupled en- zymes: the subunit molecular mass of the lung protein is 120 kD; the lung enzyme is stimulated by hemin (17) and is absolutely dependent on Mg2+-GTP (16), whereas the tumor enzyme is able to substitute Mn2+- GTP for Mg2+-GTP and is not stimulated by hemin. In addition, the pI of the lung protein is 6 (14), whereas that of the tumor enzyme is 4.7. In contrast to the near 1:1 stoichiometry of the binding of ANF to the tumor enzyme, the lung enzyme bound only 14.5% of ANF at the noted theoretical value. The structural and kinetic differences between the two enzymes suggest that these
two receptor-guanylate cyclases may be iso- zymes.
Studies done with affinity-labeling tech- niques (18-20) have shown a 120-kD ANF binding protein and those done with ANF cross-linking techniques (19) have shown 60- to 70-kD proteins in various tissues. Preliminary evidence suggests that only cer- tain ANF receptor sites may be coupled to guanylate cyclase (21). It will be of interest to scrutinize the structural and functional features of these receptors.
Coexistence of the ANF receptor and guanylate cyclase activities on a single poly- peptide chain indicates that the mechanism of transmembrane signal transduction in- volving mediation by second messenger, cyclic GMP, is different from the well-estab-
lished adenylate cyclase system. In hormone- dependent adenylate cyclase there is an as- semblage of individual components-recep- tor, GTP-binding protein, and catalytic moiety-for signal transduction (22, 23). In contrast, the presence of dual activities- receptor binding and enzymic-on a single polypeptide chain indicates that this trans- membrane protein contains both the infor- mation for signal recognition and its transla- tion into a second messenger. It is possible that a third signal component (probably a lipid or an accessory protein) is needed to link these two activities functionally.
Note added in proof. Although the anti- body to the 180-kD guanylate cyclase blocks guanylate cyclase activity, it does not inhibit the binding of ANF to the protein. This indicates that either the antibody is solely against the guanylate cyclase epitope of the protein or that there are two tightly coupled 180-kD proteins which are inseparable by the present techniques.
REFERENCES AND NOTES
1. R. K. Sharma, in Hormonally Responsive Tumors, V. P. Hollander, Ed. (Academic Press, New York, 1985), pp. 185-217.
2. N. D. Goldberg and M. K. Haddox, Annu. Rev. Biochem. 46, 823 (1977).
3. F. Murad et al., Advan. Cyclic Nucleotide Res. 11, 175 (1979).
4. P. Nambi and R. K. Sharma, Endocrinology 108, 2025 (1981).
5. Biochem. Biophys. Res. Commun. 100, 508 (1981).
6. P. Nambi, N. V. Aiyar, R. K. Sharma, Arch. Bio- chem. Biophys. 217, 638 (1982).
7. S. A. Waldman, R. M. Rapoport, F. Murad, J. Biol. Chem. 259, 14332 (1984).
8. S. A. Waldman, R. M. Rapoport, R. R. Fiscus, F. Murad, Biochim. Biopbys. Acta 845, 298 (1985).
9. M. Nakamura, K. Odaguchi, T. Shimizu, Y. Naka- mura, M. Okamoto, Eur. J. Pharmacol. 117, 285 (1985).
10. N. Jaiswal, A. K. Paul, R. K. Jaiswal, R. K. Sharma, FEBS Lett. 199, 121 (1986).
11. F. Bex and A. Corbin, Eur. J. Pharmacol. 115, 125 (1985).
12. A. K. Mukhopadhyay, H. G. Bohnet, F. A. Leiden- berger, FEBS Lett. 202, 111 (1986).
13. K. N. Pandey, S. N. Pavlou, W. J. Kovacs, T. Inagami, Biochem. Biophys. Res. Commun. 138, 399 (1986).
14. S. A. Waldman, L. Y. Chang, F. Murad, Prep. Biochem. 15, 103 (1985).
15. P. Nambi, N. V. Aiyar, R. K. Sharma, FEBS Lett. 140, 98 (1982).
16. T. Kuno et al., J. Biol. Chem. 261, 5817 (1986).
17. S. A. Waldman, M. S. Sinacore, J. A. Lewicki, L. Y. Chang, F. Murad, ibid. 259, 4038 (1984).
18. C. C. Yip, L. P. Laing, T. G. Flynn, ibid. 260, 8229 (1985).
19. R. L. Vandlen, K. E. Arcuri, M. A. Napier, ibid., p. 10889.
20. K. S. Misono, R. T. Grammer, J. W. Ribgy, T. Inagami, Biochem. Biophys. Res. Commun. 130, 994 (1985).
21. D. C. Leitman et al., J. Biol. Chem. 261, 11650 (1986).
22. D. C. May, E. M. Ross, A. G. Gilman, M. D. Smigel, ibid. 260, 15829 (1985).
23. M. Rodbell, Cyclic Nucleotide Protein Phosphoryl. 17, 207 (1984).
24. Y. Kuroda, W. C. Merrick, R. K. Sharma, Arch. Biochem. Biopbys. 213, 271 (1982).
25. U. K. Laemmli, Nature (London) 227, 680 (1970).
26. These studies were carried out between 1982 and 1986 and constitute a portion of A.K.P.’s disserta- tion. We thank R. Nutt, Merck Sharp and Dohme Research Laboratories, for the sample of synthetic ANF; W. Y. Cheung, St. Jude Research Hospital, and M. Rodbell, National Institute of Environmen- tal Health Sciences, for the critical review of the manuscript, and J. M. Sharma for editing the manu- script. Supported by NSF grant 8609867 and NIH grant NS23744.
21 July 1986; accepted 28 January 1987
Technical Comments
Computing with Neural Networks
Hopfield and Tank (1) refer to “A new concept for understanding the dynamics of neural circuitry” using the equation (in a slightly different notation)
dt
1 Ri # j=1
(i = 1,… , n) (1)
for the neuron state variables i. The con- cept is that the variables w;(t) approach equilibrium as t → w if the connections Ti are symmetric (Tij = Tji). Hopfield and Tank also state that “a nonsymmetric cir- cuit … has trajectories corresponding to complicated oscillatory behaviors … but as yet we lack the mathematical tools to manip- ulate and understand them at a computa- tional level” (1, p. 629), and that “the symmetry of the networks is natural be- cause, in simple associations, if A is associ- ated with B, B is symmetrically associated with A” (1, p. 629).
Associations are often asymmetric, as in the asymmetric error distributions arising during list learning (2). Neural network models (3) explain these distributions when one uses Eq. 1 supplemented by an associa- tive learning equation for the connections Ty
dTt dt = - ATij + Buif(uj) (2)
Because of the nonlinear term wifi(uj) in Eq. 2, Tij + Tji.
Stability theorem’s (4) have been proved about neural networks which include and generalize Eqs. 1 and 2. Thus symmetry is not necessary to prove associative learning and memory storage by neural networks. Nor is symmetry needed to design stable neural networks for adaptive pattern recog- nition (5). Methods have also been devel- oped (6) for analyzing the oscillatory behav- ior of neural circuits. We believe that the relation between symmetry and stability in neural networks is much more subtle and better understood than Hopfield and Tank (1) suggest.
Nonetheless, symmetry does help to ana- lyze the system represented by Eq. 1. In fact, we (M.A.C. and S.G.) (7) independently discovered an energy function for neural networks “designed to transform and store a large variety of patterns. Our analysis includes systems which possess infinitely many equilibrium points” (7, p. 818), exam- ples of which have been constructed (8). These networks are
da =(4)[b(x)= ad(4)]
# (i=1, … , n) (3)
Given symmetric connections (cij = Cji), the energy function is
*
V =- Σ σπειαπει) αξί + i=1
MIN
# jk=1 (4)
Along system trajectories
# dv = - Za(4)di( M)[ b(Mi) - dt i=1 *
k=1 (5)
If a(w) = 0 and d’(x) = 0, then V = 0, AVSO which is the key property of an energy function. We (M.A.C. and S.G.) have noted that “the simpler additive neural networks . are also included in our analysis” (7, p. 819). The system represented by Eq. 3 reduces to the additive network (Eq. 1) when a;(wi) = Ci1,b;(w;) = - 1/R; i + Ii, Cj = - Ty and d;(u) = f(4). Then
#
#
v = >1 [ {(E)dsg - { If(M) - 0
i=1
0
# jx=1 (6)
which includes the energy functions used in (1). We (M.A.C. and S.G.) (7) also analyzed the more difficult and physiologically im- portant cases where the cells obey mem- brane, or shunting, equations and the signal functions dy(uj) may have output thresholds.