• Evidence indicates that imidazoline I₂ binding sites (I₂BSs) are present on monoamine oxidase (MAO) and on soluble (plasma) semicarbazide-sensitive amine oxidase enzymes. The binding site on MAO has been described as a modulatory site, although no effects on activity are thought to have been observed as a result of ligands binding to these sites.
• We examined the effects in vitro of several imidazoline binding site ligands on activities of bovine plasma amine oxidase (BPAO) and porcine kidney diamine oxidase (PKDAO) in a spectrophotometric protocol.
• While both enzymes were inhibited at high concentrations of all ligands, clonidine, cirazoline and oxymetazoline were seen, at lower concentrations, to increase activity of BPAO versus benzylamine, but not of PKDAO versus putrescine. This effect was substrate dependent, with mixed or biphasic inhibition of spermidine, methylamine, p-tyramine and β-phenylethylamine oxidation observed at cirazoline concentrations that increased benzylamine oxidation.
• With benzylamine as substrate, clonidine decreased K_M (EC₅₀ 8.82 μM, E_max 75.1% of control) and increased V_max (EC₅₀ 164.6 μM, E_max 154.1% of control). Cirazoline decreased V_max (EC₅₀ 2.15 μM, E_max 91.4% of control), then decreased K_M (EC₅₀ 5.63 μM, E_max 42.6% of control) and increased V_max (EC₅₀ 49.0 μM, E_max 114.4% of decreased V_max value).
• Data for clonidine fitted a mathematical model for two-site nonessential activation plus linear intersecting noncompetitive inhibition. Data for cirazoline were consistent with involvement of a fourth site.
• These results reveal an ability of imidazoline ligands to modulate BPAO kinetics allosterically. The derived mechanism may have functional significance with respect to modulation of MAO by I₂BS ligands.

Following the discovery that clonidine exerts its hypotensive effect at least in part through an action at novel nonadrenergic receptors in the ventrolateral medulla (Bousquet et al., 1984), interest in these imidazoline receptors, as well as in other binding sites for clonidine and related drugs, has blossomed (see Eglen et al., 1998; Head, 2003 for reviews). At least three subclasses of binding sites have been claimed to exist, with only the imidazoline type-1 receptor (I₁R), which is thought to play a role in central control of blood pressure (see Bousquet & Feldman, 1999), having been shown thus far to be associated with a signal transduction system (Takada et al., 1997; Edwards et al., 2001). Imidazoline-3 binding sites (I₃BSs) encompass a nebulous group of loci associated predominantly with various ion channel subunits, and while ligands may modulate processes such as insulin secretion and catecholamine release (see Molderings, 1997; Morgan & Chan, 2001 for reviews), some of these effects may be due to drugs acting at two or more different binding sites in target tissues (Morgan & Chan, 2001).

While a variety of physiological effects can be ascribed to ligands acting at I₁Rs or I₃BSs, the same may not be said for those binding to imidazoline-2 binding sites (I₂BSs).

The observation that I₂BSs were located on the outer membranes of mitochondria (Tesson et al., 1991) quickly led to the realisation that many such sites were actually located on the monoamine oxidase (MAO) protein (Olmos et al., 1993; Alemany et al., 1995; Carpéné et al., 1995; Tesson et al., 1995). Imidazoline ligands bound with high affinity to these sites and were able to inhibit MAO and other amine oxidases in vitro (Carpéné et al., 1995), while chronic treatment of rats with amine oxidase inhibitors caused either a downregulation (Alemany et al., 1995) or an upregulation (Wiest & Steinberg, 1997) of I₂BSs in the brain or liver. Some functional interplay between the I₂BS on MAO and enzyme activity was thus presumed to occur, and the I₂BS became increasingly referred to as a modulatory or regulatory site on MAO (Tesson et al., 1995).

In support of such a designation, numerous studies have shown clearly that the I₂BS is distinct from the MAO active site (Raddatz et al., 1997; 1999) and that, in rats at least, MAO-B appears to be the predominant enzyme isoform upon which these sites might be found (Raddatz et al., 2000; Remaury et al., 2000). However, while imidazolines certainly bind with high affinity to the I₂BS on MAO, inhibition of enzyme activity requires rather higher ligand concentrations than are necessary to saturate the I₂BS (Ozaita et al., 1997).

A few other enzymes also show affinity for I₂BS ligands, including several amine oxidases of the EC 1.4.3.6 subclass (Carpéné et al., 1995; Holt & Baker, 1995; Holt et al., 2003). This family of copper-containing amine oxidases includes diamine oxidase, plasma amine oxidases and tissue-bound semicarbazide-sensitive amine oxidase (SSAO) (Callingham et al., 1991). It is likely that the active sites of these enzymes are responsible, at least in part, for binding imidazoline ligands, since several endogenous ligands for I₁Rs, I₂BSs and I₃BSs, including agmatine, tryptamine, histamine and polyamines, are substrates for one or more of these enzymes (Holt & Baker, 1995; Lyles, 1996). However, the possibility remains that an I₂BS separate from the active site may exist on semicarbazide-sensitive amine oxidase proteins.

If the high-affinity I₂BS on MAO, and perhaps on semicarbazide-sensitive amine oxidases, corresponds to a modulatory site and not simply to an artefactual binding site, then modulation of substrate turnover would necessarily take place on ligand binding to the enzyme. Some precedent does exist for rapid activation of amine oxidase activities. Human platelet MAO and rat brain MAO can be activated in vitro in a substrate-selective manner by a component of human plasma, as a result of a reduction in substrate K_M values (Yu & Boulton, 1979; Wahlund et al., 1984). Human plasma also activates human lung SSAO, once again in a substrate-selective fashion, by increasing substrate affinity (Dalfó et al., 2003), and captopril, although not possessing an imidazoline structure, activates rat MAO in vitro through increasing V_max (Raasch et al., 2002).

An extensive study by Carpéné et al. (1995) showed concentration-dependent inhibition of MAO by high concentrations of several imidazoline ligands; at low concentrations, MAO activity was similar to control values. However, in the same study, tranylcypromine and cirazoline, an imidazoline I₂BS/I₁R ligand, appeared to induce a small, statistically insignificant potentiation of bovine plasma amine oxidase (BPAO), an EC 1.4.3.6 semicarbazide-sensitive amine oxidase enzyme, at low concentrations of ligand. This effect was not commented upon by the authors. A related amine oxidase from porcine plasma was stimulated in vitro, in a pH-dependent manner, by including ammonia or imidazole in the enzyme reaction (Kelly et al., 1981; Yadav & Knowles, 1981).

Given the clear precedent for rapid modulation of several amine oxidase enzymes and the ability of imidazole, and perhaps cirazoline and tranylcypromine, to stimulate plasma amine oxidases in vitro, it was decided to examine the abilities of several imidazoline binding site (IBS) ligands to influence the activities both of BPAO and of diamine oxidase from pig kidney (PKDAO). Both of these enzymes are semicarbazide-sensitive amine oxidases of the EC 1.4.3.6 subclass, and they share a common quinone cofactor and similar reaction mechanisms (Klinman, 2003).

Assays were performed in duplicate or triplicate wells, at 37°C, with reactions usually initiated by adding amine substrates that had been prewarmed to 37°C. The increase in absorbance at 498 nm was monitored continuously in a Molecular Devices SpectraMax Plus 384 plate reader for an appropriate period of time, and initial rates of enzyme activities were determined by linear regression of data (SOFTmax PRO, version 3.1.2).

Proteins in the solid obtained from Sigma (86% protein, w w⁻¹) were separated by SDS–polyacrylamide gel electrophoresis. Samples were separated under denaturing conditions and lanes were either stained with Coomassie blue or were electroblotted to a nitrocellulose membrane. The membrane was then incubated in 2 M potassium glycinate buffer (pH 10) containing nitroblue tetrazolium (0.24 mM) for 45 min in the dark. This procedure results in deposition of purple formazan on the nitrocellulose paper over protein bands, such as that for BPAO, containing a quinone or quinoid cofactor (Paz et al., 1991; Holt et al., 1998). Gel electrophoresis (not shown) revealed that the commercial enzyme preparation contained several major and numerous minor protein components. Quinone staining of an electroblot identified one major band, at 82 kDa, consistent with the mass of a single BPAO subunit (Mu et al., 1994). Very faint formazan staining was also observed at 101 kDa, similar to a minor band at 102 kDa seen with partially purified porcine aorta SSAO (Holt et al., 1998). Thus, although the commercial BPAO preparation contains only one major quinone amine oxidase activity, it is possible that the 101 kDa component represents a subunit from vascular smooth muscle SSAO. Nevertheless, kinetic evidence in this study did not support participation of a second enzyme activity in benzylamine turnover.

Since benzylamine is also a substrate for MAO-B, the possibility was examined that some platelet MAO-B activity might be present in the bovine plasma extract and thus contribute to reaction velocity. Colorimetric assays were performed as described above, but with benzylamine substrate present at a final concentration of 2 mM, and either in the presence or absence of the MAO inhibitor pargyline (10 μM), added 10 min before substrate. While electrophoresis indicated that one or more proteins were present with masses close to 60 kDa, not inconsistent with the potential presence of MAO-B, the failure of pargyline to have any effect on benzylamine turnover (not shown) indicated that no active MAO-B enzyme was present in the preparation.

In order to assess the purity of drug compounds, aqueous solutions of clonidine and cirazoline were diluted in HPLC-grade methanol and were injected directly onto a Waters Micromass ZQ quadrupole mass spectrometer, with positive electrospray ionisation. Mass spectra (not shown) revealed the presence of parent ions with m/z ratios of 229.93, 231.94 and 233.95 (clonidine, containing ³⁵Cl and ³⁷Cl) and 217.03 (cirazoline), consistent with calculated M+H⁺ values for the drugs. No other charged species could be detected above baseline noise in either sample, suggesting that drugs were devoid of contaminants that might demonstrate efficacy in subsequent assays.

Based on results obtained from kinetic studies, a reaction mechanism was proposed and a novel equation formulated to describe the observed behaviour. Details of the derivation of this mechanism will be published elsewhere.

Figure 1 Preliminary screening results for effects of imidazoline ligands on benzylamine oxidation by BPAO (a) or putrescine oxidation by PKDAO (b). Benzylamine (740 μM) was incubated with BPAO (1 mg ml−1), or putrescine (288 μM) with PKDAO (more ...)

Benzylamine hydrochloride, putrescine dihydrochloride, methylamine hydrochloride, spermidine trihydrochloride, vanillic acid, 4-aminoantipyrine, peroxidase (type II, from horseradish), amiloride hydrochloride, clonidine hydrochloride, efaroxan hydrochloride, guanabenz, idazoxan hydrochloride, moxonidine hydrochloride and oxymetazoline hydrochloride were from Sigma. 2-(2-Benzofuranyl)-2-imidazoline hydrochloride (2-BFI) was purchased from Tocris (Ballwin, MO, U.S.A.). Cirazoline was a generous gift from Sanofi-Synthélabo (Bagneux, France), and guanfacine a gift from Sandoz (Basel, Switzerland; now Novartis Pharma). Sodium phosphate and potassium phosphate buffer salts were of molecular biology grade, and were obtained from Sigma. HPLC-grade methanol was purchased from Fisher Scientific (Nepean, Ontario, Canada). All other chemicals were of reagent grade.

In contrast, PKDAO was inhibited by all compounds tested (Figure 1b), with no indication of activation at lower ligand concentrations. For those compounds that inhibited both enzymes, the rank orders of inhibitory potency were similar (guanfacine=guanabenz>2-BFI), and in the absence of activation, weak inhibition of PKDAO by cirazoline, clonidine and oxymetazoline was evident. While 2-BFI appeared to be able to inhibit BPAO only by around 50%, inhibition of PKDAO by 2-BFI appeared uniphasic and largely complete at 1 mM. Efaroxan, a structural analogue of 2-BFI, which inhibited BPAO to a degree and with a potency similar to that of 2-BFI was virtually without effect on PKDAO.

Activation of BPAO-mediated benzylamine oxidation by cirazoline and clonidine was only observed above pH 6.00 and below pH 8.33 (data not shown). Although activity was highest between pH 7.33 and 7.67, both in control wells and in the presence of cirazoline or clonidine, the greatest degree of potentiation was seen at pH 6.33 for 30 μM cirazoline (152% of control activity at pH 6.33) and at pH 6.67 for 300 μM clonidine (164% of control activity at pH 6.67) in 167 μM sodium phosphate buffer.

The activation of BPAO in the absence of similar effects upon PKDAO suggests that the effects observed with BPAO are not due simply to nonspecific interactions of some imidazoline ligands with all copper-containing quinone cofactor amine oxidases. Furthermore, neither cirazoline nor clonidine interacted directly with components of the chromogenic solution, including peroxidase, or acted as BPAO substrates from which peroxide would be generated (results not shown).

Figure 2 Effects of the I1R ligand moxonidine (MOX), the I2BS ligand idazoxan (IDZ) and the I2ABS ligand amiloride (AM) on the ability of cirazoline (CIR; 60 μM) to potentiate oxidation of benzylamine (300 μM) by BPAO. Moxonidine, idazoxan or amiloride (more ...)

Although inhibition of BPAO by moxonidine, idazoxan and amiloride could only be observed at high concentrations of these drugs, with the consequence that IC₅₀ values could not be determined with any certainty, it was nevertheless evident that inhibitory potencies of these compounds were increased in the presence of cirazoline (Figure 2). The effects of clonidine were less substantial in this regard, with the potency of amiloride in particular remaining entirely unaffected by the presence of clonidine (not shown). These data suggest that binding of cirazoline, and to a lesser extent clonidine, to BPAO increases the affinity of the enzyme for at least some imidazoline ligands.

Figure 3 Effects of cirazoline (0.1 μM–3 mM) on oxidation by BPAO of benzylamine (BZ; 300 μM), p-tyramine (TYR; 135 μM), β-phenylethylamine (PEA; 275 μM), methylamine (MA; 420 μM) or spermidine (SPD; 50 (more ...)

Figure 4a is a representative plot illustrating that cirazoline caused a concentration-dependent reduction in the K_M of BPAO for benzylamine, with an EC₅₀ value in this example of 5.9 μM and an E_max reducing K_M to 44.8% of control. At higher concentrations, the K_M value was seen to increase, due to inhibition of activity by cirazoline. V_max was increased by cirazoline to 110% of control (Figure 4b), with an EC₅₀ of 44.0 μM. However, it was observed that at very low micromolar concentrations, cirazoline appeared already to have reduced V_max by around 6%; two separate experiments suggested that this minor effect was reproducible, and occurred with an EC₅₀ of approximately 2 μM. Thus, the effect of cirazoline in increasing V_max was, in this representative experiment, to activate the enzyme to 117% of baseline activity.

Figure 4 Effects of cirazoline on the kinetic constants KM (a) and Vmax (b) for benzylamine turnover by BPAO. Benzylamine (10 μM–3 mM) was incubated at 37°C with BPAO (0.167 mg ml−1) in the presence of water (controls) or cirazoline (more ...)

Figure 5a illustrates results from a representative experiment, showing that clonidine reduced K_M in a concentration-dependent manner to 78.2% of the control value. The EC₅₀ in this example was 7.0 μM. This stimulatory effect was rapidly obscured by the apparent increase in K_M as a result of enzyme inhibition by clonidine. The major effect of clonidine was to increase V_max (Figure 5b). The EC₅₀ in this example was 133.6 μM, and the E_max of clonidine was apparent as an increase in substrate turnover to 156% of the control V_max value, consistent with the effect on v shown in Figures 1a and 3.

Figure 5 Effects of clonidine on the kinetic constants KM (a) and Vmax (b) for benzylamine turnover by BPAO. Benzylamine (10 μM–3 mM) was incubated at 37°C with BPAO (0.167 mg ml−1) in the presence of water (controls) or clonidine (more ...)

Table 1 lists mean effects of cirazoline and clonidine on kinetic constants for benzylamine turnover by BPAO.

Table 1 Effects of imidazoline ligands on kinetic constants for benzylamine turnover by BPAO

The inhibition by cirazoline of spermidine oxidation (see Figure 3) was examined in greater detail (results not shown). Kinetic plots revealed a mixed pattern of inhibition, with an increase in K_M accompanied by a decrease in V_max. These observations are inconsistent with simple competitive inhibition at the BPAO active site, and instead indicate that cirazoline inhibits activity through binding to at least one site on the enzyme that is distinct from the active site.

Figure 6 Replots of kinetic constants for oxidation of benzylamine by BPAO in the presence of high (inhibitory) concentrations of cirazoline or clonidine. KM and Vmax values for benzylamine oxidation were obtained by nonlinear regression analysis of velocity data, (more ...)

Dixon plots (1/v versus [I]) at a range of substrate concentrations for a representative data set were linear for high (inhibitory) concentrations of both cirazoline and clonidine (not shown), suggesting that the complexes formed between enzyme, benzylamine and either cirazoline or clonidine were unable to yield product (Segel, 1993). However, while common intercepts of Dixon plots confirmed the estimated K_i value for cirazoline of 2.64 mM, they suggested that the K_i for clonidine may be two- to three-fold higher than the value of 8.8 mM, determined from Figure 6c, perhaps a consequence of the outlying point for 50 mM clonidine on that plot. Dixon slope replots (Figure 7) were thus used to confirm estimates for inhibitor constants, and yielded respective values for K_i, γK_i and γ of 2.85 mM, 15.95 mM and 5.6 (cirazoline) and 19.1 mM, 379 mM and 19.8 (clonidine).

Figure 7 Replots of slopes obtained from Dixon plots (1/v versus [I]) at several substrate concentrations and at high (inhibitory) concentrations of cirazoline and clonidine. Mean velocity data from triplicate determinations, representative (more ...)

The present data are consistent with intersecting linear noncompetitive inhibition (Segel, 1993), a simple mixed inhibition system in which the EI complex has a lower affinity for substrate than does free enzyme, and complexes containing both substrate and inhibitor are nonproductive.

A velocity equation (not shown) was devised based on this scheme by an approach that assumed the existence of rapid equilibrium conditions (Segel, 1993). The equation was entered into the user-defined nonlinear regression equations facility of GraphPad Prism, and was then applied to the representative v versus [S] data sets from which Figures 4 and 5 were generated. The variables α, β, K_H and K_L were constrained initially to the estimated values listed in parentheses in Table 2, which were deduced from the results listed in Table 1, while estimates for γ and K_i were obtained from Figure 7. Thereafter, β, K_L, α and K_H were allowed to vary in turn, in order to assess goodness of fit. However, the values for γ and K_i were held constant at their initial values, since the very low affinities of the imidazoline drugs for the inhibitory site meant that at most drug concentrations, large variations in K_i or γ would have a negligible effect on calculated values for v at any given substrate concentration.

Table 2 Variables determined for clonidine and cirazoline acting to induce two-site nonessential activation and intersecting linear noncompetitive (mixed) inhibition of BPAO, with benzylamine as substrate, by applying a novel equation (see text) to a representative (more ...)

Figure 8a shows that velocity curves at all concentrations of clonidine could be described successfully by the equation. Table 2 lists the best-fit variables obtained by nonlinear regression. With most data sets, it was subsequently found that when three of four variables were constrained to their best-fit values, the fourth could be re-obtained by regression even when the initial estimate was quite different from the calculated best-fit value, indicating that the calculated best-fit values were probably not localised minima or maxima.

Figure 8 Representative velocity versus [S] data for clonidine (a) and cirazoline (b) obtained as described in the legend to Figure 4, and fitted to an equation describing enzyme activity in the presence of a drug that is a nonessential activator (more ...)

The reaction scheme devised to describe reactions between BPAO, benzylamine and clonidine provides for one allosteric site, not including the low-affinity inhibitory site, through which modulation of V_max might be effected. However, cirazoline was observed to reduce V_max slightly at concentrations lower than those that subsequently decreased K_M and increased V_max, and a reaction scheme that would account for this further interaction would be rather more complex than that devised for clonidine. As such, the scheme for clonidine may be a little too simplistic with respect to providing an indication of all pathways that might alter reaction kinetics. Nevertheless, since the effect of cirazoline in reducing V_max is rather small, and since it occurs at cirazoline concentrations lower than those required to decrease K_M and increase V_max, it might be expected that the equation would provide an adequate fit for data obtained at higher drug concentrations. Representative data were subsequently analysed as described above for clonidine, and it was found that good fits could be obtained for velocity data in the presence of cirazoline at 100 μM or higher (Figure 8b). At 10 and 30 μM, fits were poor and could be improved only by altering α, K_H, β or K_L to values that were unrealistic. Fits could not be obtained for data obtained in the presence of 1 or 3 μM cirazoline. Table 2 lists the best-fit variables obtained from data sets for 100 μM cirazoline and higher.

The fits obtained support the contention that allosteric modulation of BPAO by cirazoline and clonidine occurs through interactions of these drugs with at least three distinct sites, and thus that alternative explanations, such as binding of two or more ligands within a single site leading to a loss of efficacy, are unlikely to offer a satisfactory explanation for the present observations. Accordingly, these data also provide indirect support for communication between copper amine oxidase subunits.

The present data provide the first conclusive evidence for allosteric modulation of amine oxidase enzymes by imidazoline receptor/binding site ligands. The effects observed with cirazoline and clonidine in this study occurred at micromolar concentrations or higher, and it is doubtful if, for these particular drugs, such effects would be of much consequence in vivo. Rather, these drugs have provided the first indication that imidazoline ligands are able to modulate SSAO enzymes through acting at novel binding sites on these enzymes. Consequently, it is anticipated that other ligands that are rather more potent in this regard will be identified or synthesised.

The present observations are of particular interest in view of the abilities of a number of imidazoline ligands to bind with high affinity to a site on MAO that corresponds, for a subset of human MAO-B enzymes, to a region between Lys 149 and Met 222 (Raddatz et al., 1997; 2000), and it has been proposed that MAO activity may be regulated in a tissue-specific manner by imidazoline ligands interacting with this site (Raddatz et al., 2000). However, with the exception of MAO inhibition through an action claimed to be at the active site (Carpéné et al., 1995; Raasch et al., 1999), effects of these ligands on MAO activity have not been demonstrated at low concentrations around the ligand K_D values for these sites. The enzyme examined in the present study is not a membrane-bound EC 1.4.3.4 flavin enzyme from the MAO family, but rather a soluble copper-containing EC 1.4.3.6 quinoprotein from the SSAO family (Janes et al., 1990). Nevertheless, despite numerous physicochemical differences, these enzyme classes share many substrates and inhibitors, and perhaps some functions (Lyles, 1984; 1996; Callingham et al., 1991). For example, chronic inhibition of both activities in vivo results in a downregulation of rat brain I₁R, presumably as a result of a prolonged increase in the levels of an endogenous I₁R agonist that is also a substrate for both MAO and SSAO (Holt et al., 2003).

Enzyme activation is usually essential in nature, and is most often evident as a requirement for a metal ion (activator) to form a complex with substrate before turnover can occur (Segel, 1993). Nonessential activation, where activity is simply enhanced following binding of the activator to the enzyme protein, is a common phenomenon in cooperative enzyme systems, and activation by low concentrations of a competitive inhibitor at low substrate concentrations is indicative of cooperative binding (MacRae et al., 2000). SSAO enzymes are homodimeric, with one active site per subunit. Crystal structures for EC 1.4.3.6 enzymes from Escherichia coli (Parsons et al., 1995), pea seedling (Kumar et al., 1996) and Arthrobacter globiformis (Wilce et al., 1997) reveal β-ribbon arms that extend from each subunit into the other, with residues from these arms partially lining a channel within the opposite subunit that extends from bulk solvent to the active site and which allows substrates to access the active site (Wilce et al., 1997). Thus, it is conceivable that binding of substrate (or inhibitor) to one active site might influence access of substrate (or inhibitor) to the second through an effect on the second substrate channel mediated through the β-ribbon arm. Since access of substrate to the active site is not thought to be rate limiting in catalysis by these enzymes (Mure et al., 2002), only by limiting access of substrate might a cooperative effect be induced; however, no evidence has been uncovered for negative (or, indeed, positive) cooperativity with respect to substrate turnover in SSAO dimers. This is also true for MAO-B, which is thought to exist as a dimer on the mitochodrial membrane (Binda et al., 2001). Nevertheless, half-site reactivity, an extreme form of negative cooperativity, has been shown to occur with hydrazine and hydrazide inhibitors of SSAO enzymes (Morpurgo et al., 1992; Frébort et al., 1995). Proof of the occurrence of half-site reactivity constitutes proof of crosstalk between subunits. Consequently, the possibility that cooperative substrate binding could occur in BPAO, and thus that low concentrations of an imidazoline ligand binding at one active site might alter substrate binding kinetics at the other active site, is very real.

While the existence of such crosstalk would provide an adequate explanation for the effects of a modulator that acts at a single site, the multisite effects observed with clonidine and cirazoline clearly implicate sites other than the dimer active sites, and the mechanism(s) responsible may thus include, but are not limited to, classical cooperativity. The present data may therefore describe a highly unusual and perhaps unique mechanism.

The model that has been proposed to explain the effects of imidazolines on the BPAO dimer contains a substrate binding site, two sites through which activity is modulated allosterically (either positively or negatively) and a site through which activity is inhibited. The model makes few assumptions, although, for simplicity, it was presumed that binding of an imidazoline drug to any of its sites was without effect on subsequent binding of the drug to the other vacant sites. Experimental data for clonidine fitted well to the derived equation. However, the possibility that cirazoline may bind to one more site than does clonidine indicates that the model may be too simplistic. While clonidine may have bound to an extra site that was not revealed experimentally, the drug failed to influence enzyme activity through binding to that site, and, for clonidine, the model therefore represents an acceptable simplification of the true system.

Cirazoline appears to bind to four sites, causing both a reduction and an increase in V_max, a reduction in K_M, and mixed inhibition at high concentrations. It is possible, but difficult to envisage, that a monomer could contain three or more different allosteric sites with affinities for one drug. Rather, the results can be explained adequately through a homodimeric model with one active site and one modulatory site on each monomer. In order that an explanation based on such a model might fit the present observations with benzylamine and BPAO, the existence of cooperativity between subunits would be necessary. With substrate bound at one active site, cirazoline might then bind to the active site on the second subunit, influencing turnover at the first active site, or at a regulatory site on the first subunit, influencing turnover at the first active site, or at a regulatory site on the second subunit, influencing either turnover of benzylamine at the first active site or binding or efficacy of the cirazoline at the second active site. The latter point is supported by the observation that the inhibitory potencies of cirazoline and clonidine appeared to be increased in the presence of other imidazoline ligands (Figure 2). This suggests that, while it is possible that inhibition is mediated through binding within the active site pocket at which substrate is oxidised, albeit not directly at the substrate binding site itself, it is equally likely that the mixed inhibition instead takes the form of negative cooperativity, with drugs binding to the active site at which substrate is not bound. In deriving an equation to describe this model, no allowance was made for the possibility that a ligand might influence its own affinity or efficacy at one site through binding to an adjacent allosteric site; the hypercubic model including such an interaction would be comprised of 32 enzyme species, 88 equilibrium constants and eight rate constants, and would be both highly complex and of little predictive usefulness.

The IBS on MAO fits the profile for an I₂BS, based partly on an affinity for idazoxan in the nanomolar range (Eglen et al., 1998). While the low micromolar affinity of at least one of the sites on BPAO for clonidine is consistent with binding to an I₂BS (Eglen et al., 1998), the low potencies of idazoxan and amiloride as antagonists of the potentiating effects of cirazoline and clonidine in the present study would seem to suggest that the sites to which cirazoline and clonidine bind are not I₂BSs. Although several of the ligands demonstrated here to have modulatory actions versus BPAO activity are I₁R ligands, binding profiles are also not entirely consistent with an involvement of I₁Rs, which in any case, like I₃BSs, are not associated with amine oxidase enzyme proteins. However, dissociation constants for the interaction of these ligands with BPAO, measured as K_i values versus amine turnover, were determined in the presence of amine substrates, and would not thus be expected to be similar to K_D values determined from radioligand binding assays. Consequently, the possibility that these allosteric sites on the BPAO protein possess characteristics typical of the I₂ sites already identified on MAO might not be precluded at the present time.

The very different effects of cirazoline on five different substrates indicate that this drug is unlikely to influence only the oxidative (second) half-reaction, since this copper- and oxygen-dependent step involves oxidation of an aminoquinol cofactor to regenerate topa quinone cofactor at the enzyme active site (Klinman, 2003); accordingly, this step should be similar with all substrates used. Rather, the reductive (first) half-reaction, which results in the release of aldehyde following binding and oxidation of amine substrate, is more likely to be the process affected by binding of low concentrations of imidazoline drugs, although a concomitant effect on the oxidative half-reaction may not be ruled out at this time. In support of this hypothesis, the rate-limiting step in benzylamine oxidation by BPAO is thought to be C–H bond cleavage during aldehyde production (Mure et al., 2002), an important step in the reductive half-reaction, and thus the rate of C–H bond cleavage must necessarily be increased in any activation process. This would not occur if the mechanism of action of imidazoline ligands was simply to control access of substrates to the hydrophobic channel, which, as has been shown in a bacterial SSAO enzyme, leads from bulk solvent to the active site (Wilce et al., 1997). The unusual temperature dependence of BPAO activity and the large kinetic isotope effect for benzylamine with this enzyme indicate that hydrogen tunnelling, a process by which activation energy requirements are anomalously reduced, occurs during this C–H bond cleavage step (Grant & Klinman, 1989) in a copper-dependent manner. The distance between the active site aspartate base (proton acceptor) and the substrate (proton donor) has a major influence on the degree to which hydrogen tunnelling may occur (Kohen & Klinman, 1999), and very small changes in the distances between donor, acceptor and active site copper following binding of an imidazoline drug to a binding site in the vicinity of the active site could be reflected in very large changes in rates of catalysis. This might explain the differential effects of cirazoline on oxidation of benzylamine, β-phenylethylamine, p-tyramine, methylamine and spermidine, since a minor spatial rearrangement of residues bordering the active site might facilitate hydrogen tunnelling when benzylamine is bound as substrate, but reduce the extent of tunnelling with the other amines.

Activation of SSAO enzymes in vitro is not entirely without precedent. Elegant kinetic work with porcine plasma amine oxidase by Kelly et al. (1981) revealed that this enzyme could be activated and inhibited by ammonia, a product of oxidation of primary amines by SSAO, and by imidazole, a common constituent in biological buffers. Furthermore, the authors concluded that the simplest model that might explain their observations would possess two sites, in addition to the active site, at which modifiers might bind (Kelly et al., 1981). More recently, a component of human plasma was found to activate SSAO from human lung through increasing affinity for benzylamine substrate (Dalfó et al., 2003). The E_max in that instance was to reduce K_M to 36.5% of control, comparable with the E_max of 42.6% determined for cirazoline in this study. While lysophosphatidylcholine was shown also to activate lung SSAO, the efficacy of this plasma phospholipid was rather lower than that of plasma itself, and it was suggested that other components in the plasma may be of greater importance in this regard (Dalfó et al., 2003). Nevertheless, the demonstration that a tissue-bound SSAO enzyme can be activated in vitro through effects of an endogenous species on substrate affinity in a manner analogous to the present observations with soluble SSAO enzymes, and to those of others (Kelly et al., 1981), suggests that allosteric sites similar to those identified in the present study may be present on several different amine oxidase enzymes, and may be physiologically relevant. Furthermore, the findings here and by others of SSAO activation in enzymes from several species and tissue sources, and in both impure and purified enzyme preparations, argue against the likelihood that the imidazoline drugs bind to a second contaminating protein and that this protein–drug complex then interacts with the SSAO dimer to activate or inhibit substrate turnover.

Active site-directed inhibitors of SSAO are being developed as novel anti-inflammatory agents (Koskinen et al., 2004). Such drugs may also reduce atherosclerotic plaque deposition (Yu et al., 2002) and reduce formation of advanced glycation end products and thus vascular damage in diabetics (Yu et al., 2003). In addition, inhibition of adipocyte SSAO results in a reduction in glucose transport into fat cells, and substantial weight loss in obese mice (Yu et al., 2004). However, while selective SSAO inhibition would seem, in some patients, to be beneficial, there is a dearth of inhibitors that show good selectivity for SSAO versus MAO enzymes in vivo, while retaining high potency. The data presented here provide novel information regarding kinetic mechanisms of amine turnover by SSAO, and indicate that novel inhibitors might target sites on the protein separate from the active site, perhaps offering a means by which selectivity could be improved. That such drugs could act in a substrate-dependent manner is an intriguing thought. Furthermore, the observation that benzylamine oxidation by BPAO could be inhibited only by 50% by 2-BFI and efaroxan suggests that these compounds will prove to be useful tools in future examinations of possible cooperative behaviour in SSAO dimers.

In summary, the present study has revealed an ability of some imidazoline ligands to alter reaction kinetics of the EC 1.4.3.6 amine oxidase from bovine plasma in a manner dependent on the substrate used. Similar observations were made in a preliminary experiment with a related enzyme from sheep plasma. Effects, both activatory and inhibitory, appeared to be mediated through drugs binding to three or four sites, and the present results are consistent with some degree of crosstalk occurring between monomers in the SSAO homodimer. A review of the literature indicates that a number of drugs and chemicals can alter reaction kinetics in vitro and potentiate activity of SSAO, and also MAO enzymes, which are known to express a high-affinity I₂BS. An ability of imidazoline ligands to alter MAO reaction kinetics and influence substrate specificity in a tissue-specific manner through an action at the I₂BS on MAO would reveal a novel target for future antidepressant and neuroprotective compounds. Work continues in our laboratory to clarify the nature of the IBSs on SSAO enzymes, and to identify compounds that alter MAO reaction kinetics through binding to the imidazoline site on that enzyme.

We are indebted to Irwin Segel (University of California, Davis) for his extensive advice on appropriate kinetic analytical procedures and to Keith Tipton (Trinity College, Dublin) for assistance with interpretation of data. We also thank Brian Callingham (Cambridge University), David Nutt and Alan Hudson (Bristol University) and Peter Knowles (University of Leeds) for helpful discussions. We are grateful to Monica Palcic (University of Alberta) for assistance with electrophoresis and electroblotting procedures, and to Gail Rauw (University of Alberta) for mass spectrometric analyses. Funding has been provided by the Canadian Institutes of Health Research (CIHR), the Canada Research Chair programme and the George and Dorothy Davey Endowment.