Portal de Periódicos da CAPES

Biological significance of agmatine, an endogenous ligand at imidazoline binding sites

2001; Wiley; Volume: 133; Issue: 6 Linguagem: Inglês

10.1038/sj.bjp.0704153

1476-5381

Walter Raasch, Ulrich Schäfer, Julian Chun, Peter Dominiak,

Ion channel regulation and function

Many publications have shown that imidazoline derivatives such as clonidine, moxonidine or rilmenidine reduce sympathetic tone via a central mechanism and that as a result they reduce plasma catecholamines and blood pressure (Reid et al., 1995). This reduction in blood pressure appears not to be regulated via peripheral, presynaptically localized receptors, since neither catecholamine depletion by reserpine, or destruction of the nerve endings with 6-hydroxydopamine produces a notable weakening of the clonidine-induced blood pressure reduction (Haeusler, 1974a,1974b; Kobinger & Pichler, 1976; Finch et al., 1975). In contrast, selective α2-adrenoceptor antagonists such as rauwolscine dose-dependently block hypotension induced by intravertebral application of clonidine. While some classical α2-adrenoceptor antagonists such as SKF86466 (in contrast to the imidazoline derivatives efaroxan and idazoxan) did not inhibit the clonidine-induced hypotension in the CNS (Ernsberger et al., 1988b; 1994; Haxhiu et al., 1994), this was discussed to due to underdosage of the antagonist (Bock et al., 1999). Hence the effects of clonidine are due to a central α2-adrenoceptor-mediated mechanism. The first signs that imidazoline derivatives might also work via non-adrenergic binding sites stem from Ruffolo (1977); the imadazoline derivative tetrahydrozoline was able to antagonize oxymetazoline-induced contraction, but not the contractile response induced by phenylethylamine derivatives such as noradrenaline, methoxamine or phenylephrine. Clear indications for a novel receptor type came from Bousquet et al. (1984), who reported hypotension after microinjection of clonidine into the rostroventrolateral medulla (RVLM). α-methylnoradrenaline showed no blood pressure reducing effect in the same model. The authors therefore assumed that binding sites must be present in the RVLM which preferentially bind imidazolines. Radioligand binding studies on RVLM membranes showed selective binding sites for imidazolines (Ernsberger et al., 1987). These data confirmed the assumptions of Ernsberger et al. (1988a; 1992); Buccafusco et al. (1995) and Bousquet et al. (1984) that the C1 region of the RVLM appeared to be the decisive area involved (Reis et al., 1989). Since then, two different imidazoline binding site subtypes have been identified (Michel & Insel, 1989; Michel & Ernsberger, 1992; Ernsberger et al., 1992). The I1-binding site, which shows a high affinity binding to [3H]-clonidine, is localized in the frontal cortex and the ventrolateral medulla (Bricca et al., 1989; Ernsberger et al., 1990a; 1992; Gomez et al., 1991), an area associated with central blood pressure regulation. Functionally, the I1-binding site seems to be involved in central blood pressure regulation (Ernsberger et al., 1988b; 1994; Haxhiu et al., 1994; Hamilton, 1992a,1992b; Hamilton et al., 1992; Hieble & Ruffolo, 1992), even though its importance remains largely unclarified, since in functional α2A-adrenoceptor 'knock out' mice (D79N; Macmillan et al., 1996) no evidence of I1-imidazoline binding site-mediated effects was revealed (Zhu et al., 1999). Its amino acid sequence and the DNA coding for it also remain to be determined, although an imidazoline binding site antisera cDNA has been isolated and characterized as encoding a 1504 amino acid protein (IRAS-1) showing properties of an I1-binding site (Piletz et al., 2000). I1-binding sites have also been demonstrated in the spinal cord, kidney and pancreas (Regunathan et al., 1993; Ernsberger et al., 1995; Schulz & Hasselblatt, 1989a), but not in the left ventricle (Raasch et al., 2000). Unlike the I1-binding site, the role of the I2-binding sites has been better characterized, whereby I2-binding sites can be further differentiated amongst I2A- and I2B-binding sites depending on their amiloride sensitivity. I2 binding sites have been demonstrated in various tissues such as brain (Brown et al., 1990), liver (Tesson & Parini, 1991) and kidney (Michel et al., 1989), whereby studies by Tesson et al. (1991) concluded that they are associated with mitochondria. Further studies focused their localization to the outer mitochondrial membrane. Functionally, the I2-binding site has been characterized as a regulatory subunit of monoamine oxidase (MAO, Figure 1), and later on it became clear that both MAO A and B share the same I2-binding site as a novel domain on the protein (Limon et al., 1992; Olmos et al., 1993). Moreover, studies on MAO A- and MAO B-deficient mice indicate that (1) the I2 binding sites identified by [3H]-idazoxan reside solely on MAO B, and (2) the binding sites on MAO A and a 28-kDa protein identified in livers of MAO A- and MAO B-deficient mice by photolabelling with 2-[3-azido-4-[(125)l]iodophenoxyl]methylimidazoline ([125I]-AZIPI) may represent additional subtypes of the imidazoline-binding site family (Remaury et al., 2000). In vitro studies have shown that selective ligands of the I2-binding site reduce MAO activity (Carpene et al., 1995; Tesson et al., 1995; Raasch et al., 1996; 1999). Further on, no correlation with a reduced oxygen utilization could be shown in vitro (Raasch et al., 1999). After protein molecular studies showed that I2-binding sites could be found on various MAO isoenzymes with varying molecular weights (Escriba et al., 1996), and that a binding domain (Escriba et al., 1999) for imadazoline derivatives could be identified in the MAO-B (Raddatz & Lanier, 1997; Raddatz et al., 1997; 1999; 2000), the functional role of the I2-binding site, unlike the I1-binding site, could be considered as established. Chronic treatment of rats with specific I2-ligands reduces MAO activity in various organs and catecholamines increase as a consequence (Raasch et al., 1999). In pathophysiological states such as heroin dependency or neurodegenerative disorders such as Alzheimer's disease or Huntington's chorea, the I2-binding site density is reduced (Garcia-Sevilla et al., 1999), which indicates that I2-binding sites might also be of clinical significance. In addition to this correlation to MAO, the I2-binding site has also been suggested to be involved in cell growth (Regunathan et al., 1996a) or analgesic effects (Olmos et al., 1994; Sastre et al., 1996a). However, both are isolated findings and require confirmation with more detailed investigations. In this view, the 28-kDa protein identified in livers of MAO A- and MAO B-deficient mice (Olmos et al., 1994; Remaury et al., 2000) may be of some importance (see Figure 1; for detailed information see corresponding section of this review). Imidazoline binding sites and their suggested functions. Agmatine binds with a moderate affinity (Ki values are taken from Li et al.) to α2-adrenoceptors as well as to I1 and I2 binding sites. Some authors (Chan, 1998; Molderings et al., 1998a; 1999a) attributed functions to binding sites (nonl1/non I2-binding sites) the affinity profile of which is consistent with neither the I1- nor the I2-binding sites. Furthermore, some authors have attributed specific functions such as noradrenaline release (Fuder & Schwarz, 1993; Molderings & Göthert, 1995; Likungu et al., 1996; Molderings et al., 1999b), secretion of gastric acid and pepsin (Molderings et al., 1998a; 1999a) and insulin from β-cells (Chan, 1998) to binding sites for which the affinity profile is not consistent with those of I1- or I2-binding sites (nonI1/non I2-binding sites see Figure 1). Apart from a specific, saturable, high affinity and reversible binding, the corresponding anatomical, histological and subcellular distribution of the putative binding sites as well as the identification of a physiological function, the establishment of the protein sequence, DNA structure, signal transduction and the identification of endogenous ligands are decisive criteria for the establishment of a new receptor system (Ernsberger, 1999). On the basis of the fact that the phenylethyl derivative noradrenaline and the imadazoline derivative clonidine influence blood pressure via central α2-adrenoceptors and/or imidazoline binding site-mediated mechanisms, it had to be asked whether other non-catecholaminergic and until now unidentified substances participate in the regulation of blood pressure and heart rate. Atlas (Atlas & Burstein, 1984a,1984b; Atlas et al., 1987) isolated a substance from rat and calf brain by ion exchange chromatography, electrophoresis and HPLC. This isolate bound specifically to α2-adrenoceptors and displaced clonidine, but not the α1-ligand prazosin or the β-ligand cyanopindolol. Because of this property, the substance was named 'Clonidine Displacing Substance' (CDS). Occasionally, this CDS is referred to as 'classical CDS' (cCDS) to emphasize its detection by radioligand binding studies. Even though Atlas (Atlas & Burstein, 1984a,1984b; Atlas et al., 1987) did not clarify its structure, CDS was characterized as a hydrophobic substance with a molecular weight of 587 Da1 stable to heat, acid hydrolysis and proteolytic enzymes such as trypsin, chymotrypsin, pronase, papain and pyroglutamase. Moreover it was postulated that CDS was not an amino acid and that it possessed no amino groups as shown by a negative ninhydrin and fluorecamine reaction. Due to its electrophoretic properties it was claimed that CDS was positively charged. Wavelengths of 224 and 276 nm represented its absorption maxima, which suggested the presence of aromatic residues in the molecule (see also Table 1; Atlas & Burstein, 1984a,1984b; Meeley et al., 1988a,1988b). Meeley et al. (1988a,1988b) developed a specific antibody directed against the clonidine analogue p-aminoclonidine. Since this antibody revealed a cross-reactivity with CDS, the authors concluded that there were structural similarities between clonidine and CDS and claimed that a phenyl and imidazole ring were mandatory structural characteristics for CDS, which confirmed the findings of Atlas & Burstein (1984a,1984b) regarding the relative hydrophobic character and the positive charge at neutral pH. CDS determined by radioimmunoassays is referred to in some studies as 'immunoreactive CDS' (irCDS) in order to emphasize its mode of determination. Since the distribution of irCDS in various organs of rats is directly correlated with biological activity attributed to cCDS, both cCDS and irCDS have been suggested to be similar (Meeley et al., 1988a). For this reason we have not distinguished between cCDS and irCDS in later sections of this review; only the term CDS is used irrespective of the way in which it was determined. Later on, CDS could be characterized by radioligand binding studies as being 30 fold selective for imidazoline binding sites compared to α2-adrenoceptors (Table 1), which strengthened the hypothesis that CDS might be an endogenous ligand for the imidazoline binding site (Ernsberger et al., 1988a; 1990b). Studies showing that CDS has only a weak affinity towards the inhibitory G-protein also fit in with this idea (Atlas, 1991). Unlike clonidine, CDS was incapable of influencing basal adenylate cyclase activity in human platelets or the noradrenaline-induced inhibition of adenylate cyclase at a concentration able to displace clonidine binding, findings which certainly do not support an α2-adrenoceptor-mediated mechanism of action for CDS. After CDS had initially been isolated from the brains of several species (Atlas & Burstein, 1984a,1984b; Meeley et al., 1986; Ernsberger et al., 1988a; Regunathan et al., 1991a), CDS was also shown in various peripheral tissues by a specific antiserum directed against CDS (see Table 1; Meeley et al., 1988b; 1992; Dontenwill et al., 1988; Hensley et al., 1989). The existence of peripheral CDS was confirmed by a specific bioassay, which is based on the ability of CDS to induce contraction of the vas deferens or gastric fundus (Diamant & Atlas, 1986; Felsen et al., 1987). CDS, which has been isolated from brain, gastric fundus, heart, small intestine, kidney, liver, skeletal muscle and serum, contracted the gastric funds in a manner completely or at least partially antagonizable by verapamil. CDS isolated from the adrenal glands, however, relaxed the gastric fundus. The different extent of antagonism as well as the relaxation by adrenal CDS was explained by the possible co-extraction of other, effect-masking substances. Finally, Synetos et al. (1991) isolated CDS from human plasma and showed its biological effectiveness through the contraction of rat aortal vascular rings. Because of the significantly reduced plasma concentrations of CDS in adrenalectomized rats compared to sham-operated animals, Meeley et al. (1992) suggested the adrenals as the possible source of CDS circulating in the blood. Since CDS and clonidine compete for a common binding site, it was questioned whether CDS would have agonistic or antagonistic activity in a functional test. After central application of CDS, arterial blood pressure increases significantly without any change in heart rate both in the cat and the rat (Bousquet et al., 1986; 1987), which in this way is opposite to the effects seen after central dosing of clonidine (see Table 2; Bousquet et al., 1984). Intracisternal application of CDS produces no change in blood pressure in anaesthetized rabbits (Bousquet et al., 1987). Furthermore, CDS can antagonize clonidine-stimulated hypotension directly, i.e. the blood pressure reduction is reduced (Bousquet et al., 1986) and the dose-response curve for clonidine is clearly shifted to the right by CDS. This finding completely contradicts the results of Meeley et al. (1986), who observed a clear drop in blood pressure and heart rate after injection of CDS into the C1-region of the rat RVLM. Combination experiments with clonidine were not performed in this study. The reasons underlying these discrepant results may lie in the various solvents used for isolating and purifying the CDS (Bousquet et al., 1987). Alternatively, effects from CDS-extract impurities such as aminoacids, CDS-fragments, catecholamines, histamine or potassium might also have lead to these inconsistencies (Reis et al., 1992; Szabo et al., 1995; Singh et al., 1995). Apart from the above described property of CDS to contract various organ preparations, CDS has also been identified as a catecholamine releasing substance (Table 2). CDS binds with a high affinity to membranes of bovine chromaffin cells, whereby the displacement of [3H]-idazoxan by CDS was not impeded by guanosine 5′-(β,γ-imido)triphosphate, indicating that the corresponding imidazoline binding site was not coupled to a GTP binding protein (Regunathan et al., 1991a). It should be noted that this was an I2 like site labelled by [3H]-idazoxan and that I2-binding sites are consistently unaffected by guanine nucleotides. The concentration-dependent adrenaline release from chromaffin cells in response to CDS was comparable to that in response to nicotine, while the CDS-stimulated noradrenaline release was only about a quarter of the noradrenaline release induced by nicotine. Unlike the nicotine response, the release of either catecholamine following CDS can not be blocked by hexamethonium, which suggests a nicotine receptor-independent mechanism (Regunathan et al., 1991a). Since the catecholamine release is also not inhibited by the specific α2-antagonist SKF-86466, which shows no affinity towards imidazoline binding sites (Ernsberger et al., 1990b), but is influenced by cobalt (Regunathan et al., 1991a), a specific imidazoline binding site-dependent and calcium-dependent release mechanism induced by CDS is suggested. For the further functional characterization of CDS, its influence on glucose stimulated insulin release was investigated in isolated Langerhans cells (Table 2 and Figure 2). The existence of imidazoline binding sites was shown in the pancreas (Schulz & Hasselblatt, 1989a,1989b), but the imidazoline binding sites of the β-cells of the pancreas appear to differ from the known I1- and I2-binding sites (Figures 1 and 2; Brown et al., 1993a; Chan et al., 1994; 1995; Morgan et al., 1995), and are also not identical with the binding site for sulphonylurea derivatives (Brown et al., 1993b; Rustenbeck et al., 1997). For this reason Morgan et al. (1999) have speculated about the existence of a pancreas specific I3-binding site. Moreover, it could be shown that imadazoline derivatives such as efaroxan or phentolamine increase the release of insulin by influencing the K-ATP channel (Figure 2; Chan & Morgan, 1990; Dunne et al., 1995; Plant & Henquin, 1990). CDS isolated from rat brain potentiated the glucose (6 mM) induced secretory insulin response concentration-dependently to a similar extent as efaroxan, and reversed the inhibitory effects of diazoxide on glucose-stimulated insulin release just as other similar imadazoline derivatives do. That this CDS effect is possibly mediated via imidazoline binding sites can be concluded from the observation that imadazoline derivatives such as RX801080 and KU14R antagonize the insulin-releasing effect of CDS. In addition, the effects of CDS on insulin secretion were not altered by pretreatment of the CDS extract (protease incubation as well 3000 Da molecular filtration centrifugation; Chan et al., 1997), which confirms the structural properties of CDS postulated by Atlas & Burstein (1984a,1984b), i.e. that CDS is not a peptide, but rather a low molecular weight substance. In closing, efaroxan-pretreated islet cells appear to be desensitized to CDS concerning insulin release, which correlates with observations obtained for efaroxan itself (Chan, 1998). Current working hypothesis depicting mechanisms regulating insulin secretion from pancreatic β-cells and proposed sites for interference by agmatine. ATP generated by glucose metabolism shuts down K+-channels, resulting in depolarization and subsequent influx of Ca2+ through voltage-activated Ca2+-channels. This influx of Ca2+ increases cytosolic Ca2+ concentration, which is accompanied by mobilization of Ca2+ from the endoplasmic reticulum (ER), an event triggering secretory granule translocation and exocytotic release of insulin. The respective binding of imidazolines (such as efaroxan) and sulphonylurea derivatives (such as glibenclamide) to I3-binding sites and the sulphonylurea receptor, also closes the K+-channels. Agmatine does not bind to I3-binding sites. However, agmatine may enhance insulin secretion via its metabolites, after it is taken up by specific transporters. Putrescine, spermidine are necessary for proinsulin biosynthesis, whereas spermine may exert a stimulatory or permissive role in RNA transcription and long-term insulin release. Polyamines are also probably involved in regulation of cytosolic Ca2+-concentration by blocking Ca2+-influx and its release from intracellular stores. Abbreviations: …▪quot; : stimulation; …•: inhibition. Li et al. (1994) succeeded in identifying and characterizing mammalian agmatine by ion and molecular weight exclusion chromatography, high pressure liquid chromatography and mass spectroscopy, as a candidate for CDS. Agmatine, the decarboxylation product of the amino acid arginine, was first identified in 1910 by Kossel in herring sperm and is known as an intermediate in the polyamine metabolism of various bacteria, fungi, parasites and marine fauna (Tabor & Tabor, 1984; Yamamoto et al., 1988; Ramakrishna & Adiga, 1975), where polyamines have been attributed an important function in cellular growth. Agmatine is chemically characterized as follows (Table 1): its molecular mass is 130 Da, the UV absorption maximum of 200 nm suggests an aliphatic structure and a ninhydrin positive reaction confirms the existence of an amino group. Radioligand binding studies on membranes of bovine cerebral cortex, the ventrolateral medulla and on chromaffin cells have revealed Kds towards α2-adrenoceptors, I1- and I2-binding sites of 4, 0.7 and 1 μM, respectively (Figures 1 and 3; Li et al., 1994a). It had low affinity for the α1- and β-adrenoceptors, 5−HT3 serotonin and D2 dopamine binding sites (Li et al., 1994a), or the κ opioid and adenosine A1-receptors (Szabo et al., 1995). An interaction with the sigma3 binding site has been shown on murine neuroblastoma cells (Molderings et al., 1996). As a functional correlate to CDS, agmatine concentration-dependently releases adrenaline and noradrenaline from chromaffin cells (Table 2). Since chromaffin cells express imidazoline binding sites, but not α2-adrenoceptors (Regunathan et al., 1993), this can be considered as an indication for an agonistic function of agmatine at these binding sites. However, since there is a lack of proof that agmatine-induced catecholamine release can be blocked by antagonists of the I1-binding site, it is not certain whether these binding sites mediate this effect. Moreover, data showing an inhibitory potency or no effect of agmatine on noradrenaline release (Häuser & Dominiak, 1995; Häuser et al., 1995; Molderings & Göthert, 1995; Molderings et al., 1997; 2000; Schäfer et al., 1999b) fuels doubts as to whether this catecholamine releasing effect is really mediated via a direct mechanism whereby imidazoline binding sites are involved. On the other hand, it was suggested that agmatine influences noradrenaline release via a dual interaction, namely a competitive antagonism and an allosteric activation of the rat α2D-adrenoceptor, since (1); noradrenaline, moxonidine- or clonidine-induced noradrenaline release in segments of rat vena cava was dose-dependently enhanced or inhibited by agmatine, and (2); binding of clonidine and rauwolscine was inhibited, the rate of association and dissociation of clonidine was altered, and [14C]-agmatine was inhibited from binding to its specific recognition site by agmatine (Molderings et al., 2000). Schematic representation of an agmatinergic synapse: L-arginine enters the nerve ending via a transporter and is decarboxylated by the mitochondrial arginine decarboxylase (ADC) to agmatine (AGM), which is stored in vesicles and metabolized to putrescine (PUT) by agmatinase (AGMase). Agmatine inhibits NO synthase (NOS) as well as monoamine oxidase (MAO) since it was demonstrated that I2-binding site (I2-BS) is a regulative binding site of MAO. After agmatine is released from the neuron it is subject for a specific uptake or it interacts with various pre- and postsynaptic receptors including the I1-binding site (I1-BS), α2 adrenoceptor (α2-R), NMDA, nicotinic cholineric (NIC), 5−HT3 (via the sigma-2 binding site) receptor. Furthermore, agmatine enters postsynaptic neurons via nicotinic and possibly NMDA ion channels. Whether such released agmatine represents a source for serum agmatine has not yet been determined. Peripheral effects of agmatine on blood pressure and cell growth are also a matter of debate. Released agmatine binds to presynaptic imidazoline binding sites and α2 adrenoceptors and in this way is involved in the regulation of catecholamines. Agmatine penetrates glial cells where it also modulates the expression and activity of iNOS. Agmatine arises enzymatically from the activity of arginine decarboxylase (ADC) on arginine (Figures 3 and 4) and is not supplied from nutritional components or bacterial colonization. ADC isolated from rat brain differs from plant or bacteria-derived ADC concerning localization, since ADC is associated with the mitochondria rather than the cytoplasm (as is typical for bacteria). The second difference concerns its substrate specificity. In contrast to bacterial ADC, mammalian ADC uses ornithine in addition to arginine, whereby it is not a typical ornithine decarboxylase, since it is neither cytosolic nor inhibited by diflouromethylornithine, a universal and irreversible inhibitor of all isoforms of ornithine decarboxylase (ODC) (Hunter et al., 1991). Finally, the optimum temperature of mammalian ADC is 30°C. At the bacterial temperature optimum of 37°C the enzyme activity of the mammalian ADC is only one third as active as it is at 30°C. Only the pH optimum (8.25) is similar between mammalian and bacterial ADC (Li et al., 1994a; 1995; Regunathan & Reis, 2000). Inhibition experiments in macrophages with lipopolysaccharides (LPS), transforming growth factor-β (TGF-β) and Interleukin-10 (IL-10) showed that ADC activity is subject to physiological control (Sastre et al., 1998). The co-localization of I2-binding sites and ADC on mitochondria has been discussed as a potential intracellular receptor-controlled regulatory loop for endogenous biosynthesis (Figure 3; Li et al., 1995). However, the organ specific distribution of ADC in rats (Regunathan & Reis, 2000) differs from that of agmatine (Raasch et al., 1995a), revealing some doubt that there is a close correlation between agmatine and its biosynthetic enzyme. Metabolism of L-arginine in the mammalian organism. Using high pressure liquid chromatography, agmatine has been demonstrated in nearly all organs of the rat (Table 1), whereby the highest concentrations are found in the stomach (71 ng g−1 wet weight), followed by the aorta, small and large intestine, and spleen; it is found in lower concentrations (<10 ng g−1 wet weight) in the lungs, vas deferens, adrenals, kidneys, heart, liver, skeletal muscle, brain and testes (Raasch et al., 1995a,1995b). Gas chromatography studies by Stickle et al. (1996) confirmed an organ specific distribution of agmatine. This distribution pattern of agmatine in various organs (Raasch et al., 1995a) differs widely from that of CDS (Table 1; Meeley et al., 1992). As an example, high concentrations of CDS but only low concentrations of agmatine are found in the adrenal gland. Moreover, the low correlation (r=0.2193) between agmatine and CDS tissue levels in both studies indicates clearly that agmatine can not exclusively represent CDS, but that, if at all, it only represents a member of a whole CDS family. The concentration of agmatine in rat plasma is only 0.45 ng ml−1 (Raasch et al., 1995a,1995b), which renders it doubtful that agmatine acts as a circulating hormone since the Kd values for the I1- (0.7 μM) and I2-binding sites (1 μM; (Li et al., 1994a) are approximately 200 – 300 fold higher compared to rat plasma concentrations. In addition, the source for circulating agmatine remains unidentified, since (1) ADC has not been detected in plasma until now, and (2) the adrenals, which were identified as sources for CDS (Meeley et al., 1992), contain only minimal amounts of agmatine (see Table 1; Raasch et al., 1995a). Stimulation experiments on animals designed to investigate this question have not been performed until now. In humans, substantially higher plasma concentrations (47 ng ml−1) were determined when compared to rats (Feng et al., 1997). The reasons underlying this large difference remain to be clarified. An age-dependency for agmatine tissue concentrations could not be established, with the exception of the cerebral cortex, where concentrations declined by nearly 50% with age (Raasch et al., 1995a). Using immunohistochemistry with specific antibodies against agmatine (Wang et al., 1995), agmatine was found to be regionally distributed in the cerebral cortex, the lower brain stem, the midbrain, frontal brain, thalamus and the hypothalamus in rat brain (Otake et al., 1998). In this way the distribution of agmatine-containing neurones correlates with the distribution pattern of α2-adrenoceptors and imidazoline binding sites to the extent that (1) in most agmatine containing regions, α2-adrenoceptors and imidazoline binding sites are also expressed (Kamisaki et al., 1990; de vos et al., 1991; 1994; Bricca et al., 1993; Nicholas et al., 1993; 1996; King et al., 1995; Ruggiero et al., 1995) and (2) agmatinergic neurones are concentrated in brain regions (e.g. cerebral cortex), which project to areas (e.g. striatum and midline thalamus), which contain α2-adrenoceptors and imidazoline binding sites (Berendse & Groenewegen, 1991; Jones & Yang, 1985; Saper et al., 1986). There also appear to be a multitude of interactions between agmatinergic cells and receptor areas in the corticothalamostriatal regulatory loops. At the cellular level, agmatine could be shown in smooth muscle and endothelial cells. Since ADC is expressed only in endothelial cells and not in smooth vascular muscle cells, it was concluded that agmatine either originates in the serum or is taken up from the endothelial cells and stored in the smooth vascular muscle cells, although a corresponding transporter for agmatine has not yet been identified in the vasculature (Regunathan et al., 1996a). There are, however, reports on an agmatine-transport system of bacterial (Kashiwagi et al., 1986; Driessen et al., 1988) and neuronal origin (Sastre et al., 1997). Glial cells not only express imidazoline binding sites, they also synthesize agmatine (Regunathan et al., 1995a), whereby the agmatine concentration and ADC activity in the cultivated cells are substantially higher than the concentrations in brain, which might indicate that glial cells represent the main site for synthesis and storage. It is well known that cells under culture conditions can undergo alterations in distinct features such as receptor population, enzyme activity and this might also alter agmatine content. This could be why there are differences in agmatine content between neuronal and glial cells. In this respect, interferon-gamma (IFN-γ) was able to increase ADC activity without inducing nitric oxide synthase (iNOS) activity significantly in astrocytes, whereas LPS stimulated iNOS but not ADC activity. These data suggest that the ADC activity in neuronal tissue is subject to regulation, and that two different stimuli influence two pathways of arginine metabolism in entirely different ways. At the subcellular level, agmatine was shown by immuncytochemistry to be localized mainly in large dense-core vesicles in the cytoplasm