Search

We have determined the binding affinities of the novel alpha-2 adrenoceptor antagonist radioligand [3H]-MK912 for the cloned human alpha-2A, alpha-2B and alpha-2C adrenoceptors. The KD-values were 1.25 nM, 1.36 nM and 0.086 nM for the alpha-2A, alpha-2B and alpha-2C subtypes, respectively. Thus, the selectivity of [3H]-MK912 for the human alpha-2C adrenoceptor vs. the human alpha-2A and alpha-2B adrenoceptors is 14-fold and 16-fold, respectively. The alpha-2C selectivity, and the very high affinity of [3H]-MK912 for the alpha-2C adrenoceptor subtype (KD = 86 pM) makes this radioligand a promising tool for studying the role of alpha-2C adrenoceptors in the human. A selection of antagonists useful for differentiating between the human alpha-2A, alpha-2B and alpha-2C adrenoceptor subtypes is discussed.

Many different types of receptors couple to the inhibitory guanine nucleotide-binding protein (G protein) Gi. In NG108-15 neuroblastoma-glioma cells, alpha 2b-adrenergic, m4 muscarinic, and delta-opiate receptors all use Gi as a transducer. According to the ternary complex model of receptor-G protein interactions, agonists bind to these receptors with high affinity only in their G protein-associated form. Conversely, G protein affinity for the receptor is increased by agonist binding. We have developed an extended ternary complex model in which multiple receptors couple to a single G protein and we have examined two consequences of the model theoretically and experimentally. First, the simple ternary complex model can account for the observed high and low affinity agonist binding only when G protein is limiting; however, measurements show a significant excess of G protein over receptor. Could this paradox be explained by other receptors competing for the same G protein and limiting the amount of free G protein so that high and low affinity agonist binding would be seen? Our theoretical simulations show that this does not occur unless the receptors and G protein are present in a precise stoichiometric ratio and have an extremely high affinity, such as when agonists for both receptors are present. The second prediction of this model is that binding of an agonist at one receptor should produce competition for G protein used by another receptor. If the G protein pool were limiting and freely mobile, this would result in an unlabeled agonist at one receptor decreasing binding of a radiolabeled agonist to another receptor. Experimentally, the G protein was made limiting by a partial pertussis toxin treatment. Radioligand binding to alpha 2b-adrenergic and m4 muscarinic receptors in these pertussis toxin-treated NG108-15 membranes showed no cross-talk with the delta-opiate or muscarinic receptors, which are present in excess. This could occur because the different receptors interact with structurally different G proteins (e.g., distinct beta or gamma subunits). More likely it is because of limitations of the mobility of G proteins in the membrane due to 1) attachment to structural elements, such as the cytoskeleton, 2) sequestration in lipid pools, or 3) organization into slowly exchanging supramolecular complexes. These results show that we must reexamine the assumptions of the collision coupling and ternary complex models.

Rapid activation of guanine nucleotide-binding protein (G protein)-mediated signal transduction mechanisms occurs in many tissues. The human neutrophil provides a useful model for studying the mechanisms of these fast processes. Fluorescent chemotactic tetrapeptide and pentapeptide exhibit 30-50% quenching of fluorescence upon binding to the neutrophil formyl peptide receptor, and their binding affinity is strongly regulated by the G protein Gi. We used rapid kinetic spectrofluorometric methods to study the assembly and disassembly of the ternary complex of ligand, receptor, and G protein in digitonin-permeabilized human neutrophils. Binding was studied up to 20 nM ligand, where the half-time for association was 1.2 sec. The rate constant of association was near that for diffusion-limited reactions of ligands and proteins, 2 x 10(7) M-1 sec-1. The rate of uncoupling of formyl peptide receptor from G protein in the presence of high concentrations of guanine nucleotide was > or = 5 sec-1 (i.e., t1/2 of 0.14 sec). Thus, disassembly of the formyl peptide receptor-G protein complex occurs in the millisecond time domain and may be faster than the next step in the signal transduction process.

Synthetic peptides are important tools for understanding the sites and mechanisms of receptor/G protein interactions. We examined the structural determinants of receptor-fragment peptides for G protein binding and activation. A dimer of peptides from the carboxyl-terminal (i3c) and amino-terminal (i3n) regions of the alpha 2A-adrenergic receptor is most potent in stimulating guanine-nucleotide exchange of any peptides studied. Stimulation of GTPase by i3n is partially blocked by pertussis toxin treatment, whereas stimulation by i3c is not, which is consistent with action of i3c at the amino terminus of Gi. Both peptides inhibit adenylyl cyclase in Chinese hamster ovary cell membranes, but only the i3c effect is consistent with a pure Gi stimulation. We also examined the mechanism and defined a minimal structural subset of i3c required for G protein activation. Residues 361-365 from the receptor were essential for GTPase stimulation, whereas determinants in the region 368-373 modulated that activity. A specific role for arginines is defined beyond just their positive charge. Complex effects of modifications of Thr373 suggest a regulatory or conformational role of that residue in the previously defined constitutive activation of the alpha 2-adrenergic receptor [J. Biol. Chem. 268:16483-16487 (1993)]. Thus, our data plus recent mutagenesis results support a role for hydrophobicity in the i3n region and a positively charged/arginine-rich region approximately 15-20 residues from the sixth transmembrane span in G protein activation. In contrast, the immediate perimembrane region of i3c seems to have largely conformational effects in producing constitutive activation of the receptors.

The agonist- and antagonist-binding properties of the alpha 2-adrenergic receptor in a purified plasma membrane preparation from human platelets were determined both by direct binding of radiolabeled ligands and by competition with the labeled alpha 2-antagonist, [3H] yohimbine. Binding of [3H]yohimbine was characterized by a single high affinity binding site (Kd = 6.2 +/- 1.4 nM, Bmax = 507 +/- 53 fmol/mg). In direct binding studies, the imidazoline full alpha 2-agonist, [3H]-5-bromo-6-N(2-4,5-dihydroimidazolyl)quinoxaline ([3H] UK 14,304), bound to only one quantifiable high affinity site (Kd = 0.88 +/- 0.17 nM), representing 65 +/- 6% of the number of [3H]yohimbine sites. Binding of the partial agonist [3H]-p-aminoclonidine (PAC) showed nonlinear Scatchard plots. Analysis according to a model of multiple independent binding sites showed the data to be consistent with two sites (Kd1 = 0.62 +/- 0.18 nM and Kd2 = 7.9 +/- 1.4 nM). The high affinity site corresponded to 15 +/- 6% and the low affinity site corresponded to 39 +/- 6% of the number of [3H]yohimbine sites. Competition for binding of the alpha 2-antagonist, [3H]yohimbine, with nonradiolabeled ligands revealed a single affinity for yohimbine. In contrast, competition for [3H]yohimbine binding by the full agonist UK 14,304 and epinephrine is best fit by a model with two independent binding sites. The partial agonist PAC was best characterized by a model with three distinct binding sites. The full agonists UK 14,304 and epinephrine inhibited adenylate cyclase approximately 30%, whereas PAC produced only 12% inhibition. The inhibitory guanine nucleotide-binding protein (Ni) with Mr 40,700 was the sole pertussis toxin substrate in the purified membranes. It was quantitated by pertussis toxin-catalyzed [32P]ADP ribosylation in cholate extracts. There is a 20- to 100-fold excess of Ni over alpha 2-adrenergic receptors. Comparisons made between the experimental data for agonist binding and theoretical predictions of the simple ternary complex model suggest that there is compartmentalization of Ni and/or that the alpha 2 receptors are heterogeneous.

Highly purified cholinergic postsynaptic membranes from Torpedo electric tissue contain, in addition to the acetylcholine receptor (AcChoR), major proteins of Mr 43,000 and Mr approximately 90,000 and minor proteins that can be removed from the membranes by alkaline treatment. We have prepared an antiserum to these alkaline-extractable proteins that reacts with the Mr 43,000 protein but not with any of the other major membrane proteins, including the AcChoR subunits. Immunofluorescent staining of sections of Torpedo electric tissue shows that this antiserum binds to the innervated but not the uninnervated surface of the electrocytes. In rat diaphragm muscle, the antigens recognized by this antiserum are highly concentrated at the synapse. Synaptic staining of muscle is eliminated by prior incubation of the antiserum with the Mr 43,000 protein but not by incubation with affinity-purified AcChoR. This antiserum stains end plates of muscles denervated for 7 days. Antiserum to AcChoR binds to the subsynaptic membranes of electrocytes and muscle but does not react with the Mr 43,000 protein. Purified AcChoR blocks staining of synapses by anti-AcChoR but the Mr 43,000 protein does not. These results indicate that the Mr 43,000 protein is located in the innervated membrane of Torpedo electrocytes and that an immunologically similar component is highly concentrated in the postsynaptic membrane of mammalian muscle.

Regulator of G protein-signaling (RGS) proteins accelerate GTP hydrolysis by Galpha subunits and are thought to be responsible for rapid deactivation of enzymes and ion channels controlled by G proteins. We wanted to identify and characterize Gi-family alpha subunits that were insensitive to RGS action. Based on a glycine to serine mutation in the yeast Galpha subunit Gpa1(sst) that prevents deactivation by Sst2 (DiBello, P. R., Garrison, T. R., Apanovitch, D. M., Hoffman, G., Shuey, D. J., Mason, K., Cockett, M. I., and Dohlman, H. G. (1998) J. Biol. Chem. 273, 5780-5784), site-directed mutagenesis of alphao and alphai1 was done. G184S alphao and G183S alphai1 show kinetics of GDP release and GTP hydrolysis similar to wild type. In contrast, GTP hydrolysis by the G --> S mutant proteins is not stimulated by RGS4 or by a truncated RGS7. Quantitative flow cytometry binding studies show IC50 values of 30 and 96 nM, respectively, for aluminum fluoride-activated wild type alphao and alphai1 to compete with fluorescein isothiocyanate-alphao binding to glutathione S-transferase-RGS4. The G --> S mutant proteins showed a greater than 30-100-fold lower affinity for RGS4. Thus, we have defined the mechanism of a point mutation in alphao and alphai1 that prevents RGS binding and GTPase activating activity. These mutant subunits should be useful in biochemical or expression studies to evaluate the role of endogenous RGS proteins in Gi function.

The functional organization of Gbetagamma is poorly understood. Regions of bovine brain Gbetagamma that interact with a photoaffinity derivative of an alpha2-adrenergic receptor-derived peptide from the third intracellular loop (diazopyruvoyl-modified peptide Q (DAP-Q)) and a hydrophobic membrane probe (3-trifluoromethyl-3-(m-iodophenyl)diazirine (TID)) were examined. We previously showed that DAP-Q cross-links to specific, competable sites on both the alpha and beta subunits of Go/Gi but not on the gamma subunit and that betagamma subunit was required for stimulation of Go/Gi GTPase activity (Taylor, J. M., Jacob Mosier, G. G., Lawton, R. G., Remmers, A. E., and Neubig, R. R. (1994) J. Biol. Chem. 269, 27618-27624). Similarly, we show here that the membrane-associated photoprobe [125I]TID labels alpha and beta but not gamma. We have now mapped the sites of incorporation of DAP-Q and TID into the beta subunit. TID labels both the 14-kDa amino-terminal and the 23-kDa carboxyl-terminal fragments from a partial tryptic digest of beta while DAP-Q labels only the carboxyl-terminal fragment. Further mapping with endopeptidase Lys C reveals substantial labeling of multiple fragments by TID while DAP-Q labels predominantly a approximately 6-kDa fragment within the carboxyl-terminal 60 amino acids of beta1. Thus, regions within the 7th (or possibly 6th) WD-40 repeat of the beta subunit of G protein interact with the receptor-derived peptide while membrane interaction involves multiple sites throughout the beta subunit.

After alkaline extraction, purified subsynaptic fragments isolated from Torpedo electric tissue exhibit on sodium dodecyl sulfate/polyacrylamide gel electrophoresis predominant peptides of apparent Mr 41,000, 50,000, and 65,000 (i.e., the peptides characteristic of the nicotinic receptor purified and isolated in detergent solutions). The peptide of Mr 43,000 that is also found in the isolated postsynaptic membranes is recovered in the supernatant after alkaline extraction. The alkaline-extracted membranes were functionally intact, as demonstrated by the following criteria. The kinetics of binding of [3H]acetylcholine in the presence and absence of 30 micron carbamoylcholine to occupy acetylcholine binding sites, [14C]-meproadifen [2-(diethylmethylaminoethyl)-2,2-diphenylvalerate iodide ] was bound with a dissociation constant, KD, of 0.3 +/- 0.1 micron to 0.3 +/- 0.1 site per [3H]alpha-toxin site. This binding was displaced by perhydrohistrionicotoxin. The carbamoylcholine-stimulated efflux of 22Na+ from the Torpedo vesicles were preserved after alkaline extraction. It is concluded that not only the acetylcholine binding site, but also the local anesthetic binding site, must be associated with the peptides of the cholinergic receptor itself and not that of Mr 43,000. Those peptides remaining after alkaline extraction are also sufficient for permeability control.

alpha 2-Adrenergic receptor (alpha 2-AR) responses are mediated by the pertussis toxin-sensitive guanine nucleotide-binding protein (G protein) Gi. Because all three known Gi subtypes are inactivated by pertussis toxin, it has been difficult to determine which of the subtypes are involved in alpha 2-AR responses. In order to investigate alpha 2-AR/Gi coupling, we performed binding and adenylyl cyclase experiments in membranes from CHO-K1 cells transfected with the human alpha 2A-AR. Antisera directed against the carboxyl-terminal region of the Gi1/Gi2 or the Gi3 proteins were used to determine which subtypes were important for high affinity agonist binding and inhibition of adenylyl cyclase. The CHO-K1 cell membranes exhibited immunoreactivity at an apparent molecular mass of 40-41 kDa for both Gi1/Gi2 and Gi3 antisera. Western blot analysis, using purified bovine brain G proteins for comparison, demonstrated that the transfected CHO-K1 cells possess Gi2 and Gi3. High affinity guanosine 5'-(beta,gamma-imido) triphosphate-sensitive binding of the alpha 2-AR agonists [3H]bromoxidine and p-[125I]iodoclonidine ([125I]PIC) was reduced by 30-50% by either the Gi1/Gi2 or Gi3 antiserum. Bromoxidine (1 microM) and PIC (1 microM) inhibited membrane adenylyl cyclase by 34 and 27%, respectively. Gi3 antiserum reduced the inhibition by 26% and 67% for bromoxidine and PIC, respectively. The Gi1/Gi2 antiserum reduced the inhibition by 56% and 63% for bromoxidine and PIC, respectively. Furthermore, when both antisera were used together, there was a complete reversal of alpha 2-AR-mediated inhibition. These observations provide evidence of alpha 2A-AR coupling to at least two subtypes of Gi proteins and the first evidence of functional involvement of Gi3 in the inhibition of adenylyl cyclase.

The binding properties of p-[125I]iodoclonidine [( 125I]PIC) to human platelet membranes and the functional characteristics of PIC are reported. [125I]PIC bound rapidly and reversibly to platelet membranes, with a first-order association rate constant (kon) at room temperature of 8.0 +/- 2.7 x 10(6) M-1 sec-1 and a dissociation rate constant (koff) of 2.0 +/- 0.8 x 10(-3) sec-1. Scatchard plots of specific [125I]PIC binding (0.1-5 nM) were linear, with a Kd of 1.2 +/- 0.1 nM. [125I]PIC bound to the same number of high affinity sites as the alpha 2-adrenergic receptor (alpha 2-AR) full agonist [3H] bromoxidine (UK14,304), which represented approximately 40% of the sites bound by the antagonist [3H]yohimbine. Guanosine 5'-(beta, gamma-imido)triphosphate greatly reduced the amount of [125I]PIC bound (greater than 80%), without changing the Kd of the residual binding. In competition experiments, the alpha 2-AR-selective ligands yohimbine, bromoxidine, oxymetazoline, clonidine, p-aminoclonidine, (-)-epinephrine, and idazoxan all had Ki values in the low nanomolar range, whereas prazosin, propranolol, and serotonin yielded Ki values in the micromolar range. Epinephrine competition for [125I]PIC binding was stereoselective. Competition for [3H]bromoxidine binding by PIC gave a Ki of 1.0 nM (nH = 1.0), whereas competition for [3H]yohimbine could be resolved into high and low affinity components, with Ki values of 3.7 and 84 nM, respectively. PIC had minimal agonist activity in inhibiting adenylate cyclase in platelet membranes, but it potentiated platelet aggregation induced by ADP with an EC50 of 1.5 microM. PIC also inhibited epinephrine-induced aggregation, with an IC50 of 5.1 microM. Thus, PIC behaves as a partial agonist in a human platelet aggregation assay. [125I]PIC binds to the alpha 2B-AR in NG-10815 cell membranes with a Kd of 0.5 +/- 0.1 nM. [125I]PIC should prove useful in binding assays involving tissues with a low receptor density or in small tissue samples and in studies of cloned and expressed alpha 2-AR.

The functional significance of high affinity agonist binding to receptors that interact with guanine nucleotide regulatory proteins has remained controversial. Preincubation of human platelet membranes with the full alpha 2-agonist UK 14,304 in the absence of GTP increases the potency of the agonist to inhibit adenylate cyclase in a pre-steady state (15-sec) assay. The EC50 after preincubation (6 +/- 1 nM) is within a factor of 2 of the high affinity Kd for [3H]UK 14,304 binding determined under identical conditions (2.7 +/- 0.1 nM). In contrast, in the usual steady state measurements (15 min) or in pre-steady state measurements without agonist preincubation, the EC50 values (74 +/- 1 and 207 +/- 8 nM, respectively) are near the low affinity Kd for [3H]UK 14,304 binding. Reduction of the GTP concentration in steady state adenylate cyclase assays also decreases the EC50 for UK 14,304 from 40 +/- 5 nM at 10 microM GTP to 14 +/- 5 nM with no added GTP. Both sets of experimental observations are accommodated by a complete kinetic model of inhibition in which the high affinity ternary complex of drug, receptor, and G protein leads to the response. Explicit rate parameters are included for agonist binding, receptor-G protein interactions, GTP binding, and hydrolysis. Despite the functional role of the high affinity state of the alpha 2-receptor in this model, the steady state EC50 for agonist-mediated inhibition correlates best with the Kd of low affinity agonist binding in the presence of high levels of GTP. Under conditions in which formation of the high affinity ternary complex is favored, the EC50 for responses approaches the high affinity Kd.

G protein heterocomplex undergoes dissociation and association during its functional cycle. Quantitative measurements of alpha and betagamma subunit binding have been difficult due to a very high affinity. We used fluorescence flow cytometry to quantitate binding of fluorescein-labeled Gi1alpha (F-alpha) to picomolar concentrations of biotinylated G beta gamma. Association in Lubrol solution was rapid (kon = 0.7 x 10(6) M-1 s-1), and equilibrium binding revealed a Kd of 2.9 +/- 0.8 nM. The binding showed a complex dependence on magnesium concentration, but activation of F-alpha with either GDP/aluminum fluoride or guanosine 5'-O-(3-thiotriphosphate) completely prevented formation of the heterocomplex (Kd > 100 nM). The binding was also influenced by the detergent or lipid environment. Unlabeled betagamma reconstituted in biotinylated phospholipid vesicles (pure phosphatidylcholine or mixed brain lipids) bound F-alpha approximately 2-3-fold less tightly (Kd = 6-9 nM) than in Lubrol. In contrast, beta gamma in ionic detergents such as cholate and 3-[(cholamidopropyl)diethylammonio]-1-propanesulfonate exhibited substantially lower affinities for F-alpha. Dissociation of F-alpha from beta gamma reconstituted in lipid vesicles was observed upon addition of aluminum fluoride or excess unlabeled alpha subunit, indicating that myristoylated alpha subunit has only a weak interaction with lipids without the beta gamma subunit. The kinetics of aluminum fluoride-stimulated dissociation were slower than those of the alpha subunit conformational change detected by intrinsic fluorescence. These results quantitatively demonstrate G protein subunit dissociation upon activation and provide a simple but powerful new approach for studying high affinity protein/protein interactions in solution or in a lipid environment.

N-methyl-3'-O-anthranoyl (MANT) guanine nucleotide analogs are useful environmentally sensitive fluorescent probes for studying G protein mechanisms. Previously, we showed that MANT fluorescence intensity when bound to G protein was related to the degree of G protein activation where MANT-guanosine-5'-O-(3-thiotriphosphate) (mGTP gammaS) had the highest fluorescence followed by mGTP and mGDP, respectively (Remmers, A. E., Posner, R., and Neubig, R. R. (1994) J. Biol. Chem. 269, 13771-13778). To directly examine G protein conformations with nucleotide triphosphates bound, we synthesized several nonhydrolyzable MANT-labeled guanine nucleotides. The relative maximal fluorescence levels observed upon binding to recombinant myristoylated Goalpha (myrGoalpha) and myrGialpha1 were: mGTPgammaS > MANT-5'-guanylyl-imidodiphosphate > MANT-guanylyl-(beta,gamma-methylene)-diphosphonate > MANT-guanosine 5'-O-2-(thio)diphosphate. Using protection against tryptic digestion as a measure of the activated conformation, the ability of the MANT guanine nucleotides to maximally activate myrGo alpha correlated with maximal fluorescence. Biphasic dissociation kinetics were observed for all of the MANT guanine nucleotides. The data were consistent with the following model, [formula: see text] where G protein activation (G*-GXP) is determined by a conformational equilibrium between two triphosphate bound states as well as by the balance between binding and hydrolysis of the nucleotide triphosphate. Compared with myrGialpha1, maximal mGTP fluorescence was only 2-fold higher for the myrGialpha1 Q204L mutant, a mutant with greatly reduced GTPase activity, and only 24% that of mGTPgammaS, indicating that partial activation by mGTP was not just due to hydrolysis of mGTP. These results extend our previous conclusion that GTP analogs do not fully activate G protein.

The nonhydrolyzable GTP analogue guanosine 5'-(beta, gamma-imido)triphosphate (GMP-PNP) produced an ATP-dependent but Ca2+-independent stimulation of [3H]norepinephrine release from permeabilized chromaffin cells. This stimulation of secretion was 25-35% of the secretion induced by 10 microM Ca2+. A similar Ca2+-independent stimulation was produced by other non-hydrolyzable GTP analogues. No effect was seen with a variety of other nucleotides, including GTP. The GMP-PNP effect was specifically inhibited by low concentrations of guanine nucleotides. Addition of cAMP did not mimic the Ca2+-independent GMP-PNP effect, but did slightly enhance Ca2+-dependent secretion. Pretreatment with pertussis toxin had no effect on Ca2+-dependent secretion or on the GMP-PNP effect. There was no detectable diglyceride or inositol phosphate produced during GMP-PNP treatment, and addition of diglyceride and inositol trisphosphate did not induce secretion. Guanosine 5'-(beta-thio)diphosphate (GDP-beta-S), in addition to its ability to inhibit the GMP-PNP effect, partially inhibited Ca2+-dependent secretion. At 10 microM free Ca2+, the effects of GMP-PNP and Ca2+ were nonadditive. In fact, secretion in the presence of both GMP-PNP and 10 microM Ca2+ was slightly less than secretion due to Ca2+ alone. These data suggest that a guanine nucleotide-dependent process interacts in some way with one or more components of the normal Ca2+-dependent secretory pathway. However, it may not be an intrinsic part of the mechanism underlying Ca2+-dependent secretion.

Synthetic peptides that activate or inhibit G proteins reveal structural determinants of receptor-G protein interactions and show promise as potential therapeutic agents. A cysteine-containing peptide from the carboxyl-terminal part of the third cytoplasmic loop of the alpha 2-adrenergic receptor (peptide Q) uncouples alpha 2-adrenergic receptors from Gi. Peptide Q readily forms disulfide-linked dimers (Qdimer), as detected by high performance liquid chromatography and mass spectrometry. Qdimer is > 100-fold more potent than monomeric Q peptide in inhibiting p-[125I] iodoclonidine binding to the human alpha 2a-adrenergic receptor in platelet membranes and transfected Chinese hamster ovary cells. In addition, Qdimer is 10-20 times more potent than monomeric Q peptide in inhibiting alpha 2 agonist-stimulated GTPase in cell membranes and in directly stimulating G(o)/Gi GTPase in lipid vesicles. The effect of Qdimer is reversible and not mimicked by cystine. Formylation of both tryptophans greatly reduces the potency of the dimer but a single formyl group is well tolerated, indicating an asymmetric interaction of the dimer with Gi in membranes. A mixed dimer of peptides from the amino- and carboxyl-terminal ends of the third cytoplasmic loop of the alpha 2-adrenergic receptor is most potent in all measures of G protein interactions, suggesting that the dimer of Q peptides mimics multiple intracellular portions of the alpha 2-adrenergic receptor with the G protein. These data confirm the importance of multiple receptor regions in G protein activation and suggest a strategy for examining the role of physically separated regions in protein-protein interactions.