Binding of Urate and Caffeine to Haemocyanin Analysed by Isothermal Titration Calorimetry

Two major anaerobic metabolites, L-lactate and urate, modulate the oxygen affinity of haemocyanin in some crustacean species (Truchot, 1980; Morris et al., 1985; Bridges and Morris, 1986; Morris and Bridges, 1994). L-Lactate, which accumulates as the main glycolytic end-product in anaerobic Carcinus maenas, Cancer pagurus, Callinectes sapidus and other crustaceans, lowers the P₅₀ of the respiratory pigment independently of the proton concentration and reduces the cooperativity of oxygen binding (Truchot, 1980; Johnson et al., 1984). The lactate-binding site was found to be stereospecific, so a precise arrangement of both the hydrogen atom and the methyl group at the chiral carbon atom was required for full activity of the ligand (Graham et al., 1983; Graham, 1985).

Urate also increases the oxygen affinity of the haemocyanin of some crustaceans, as first described by Morris et al. (1985) in the freshwater crayfish Austropotamobius pallipes. Similar results were obtained in Carcinus maenas, Homarus vulgaris and several other crustaceaceans (Lallier and Truchot, 1989; Bridges, 1990; Nies et al., 1992). Urate accumulates in the haemolymph of these species (Lallier et al., 1987; Zeis, 1994) during hypoxic conditions, when uricase activity is diminished because its second substrate, oxygen, is scarce (Dykens, 1991). Several related purine bases and analogues such as caffeine were examined for their effectiveness in modulating the oxygen affinity of the haemocyanin of A. pallipedes (Morris et al., 1985, 1986; Morris, 1990). The specificity for a purine derivative structure was found to be low, but the purine ring system was essential for modulating haemocyanin oxygen-binding. The simultaneous presence of both in vivo factors, L-lactate and urate, had an additive effect on haemocyanin oxygen-binding in H. vulgaris; however, the increase in the oxygen affinity of the pigment was finite (Zeis et al., 1992).

The haemocyanin of arthropods consist of one, two, four, six or eight basic hexameric assemblies (Markl and Decker, 1992; Van Holde and Miller, 1992; Van Holde and Miller, 1995). Binding studies using ultracentrifugation techniques and equilibrium dialysis revealed two binding sites each for L-lactate and urate per dodecameric haemocyanin molecule of H. vulgaris, which suggest a 1:1 stoichiometry in binding between the effectors and the basic allosteric unit, the hexamer. The dissociation constants were 0.87 mmol l⁻¹ for L-lactate and 0.03 mmol l⁻¹ for urate. No evidence for cooperativity in ligand binding was detected using this method for either effector (Nies et al., 1992). Furthermore, the binding of urate led to a significant decrease in the cooperativity of oxygen binding and in the magnitude of the Bohr effect in H. vulgaris (Zeis et al., 1992; Nies et al., 1992).

A few other studies have examined the stoichiometry and affinity of haemocyanin/effector interactions. In Panulirus interruptus, the California spiny lobster, which possesses a hexameric haemocyanin consisting of three different subunits (a, b and c), one lactate-binding site was found per hexamer. However, an artificial homo-hexamer of subunit b also contained one lactate-binding site per hexamer. The absence of a 1:1 ratio between the number of lactate-binding sites and the number of subunits suggests that the lactate-binding site occurs between subunits within the quartenary structure of the protein (Johnson et al., 1987). These findings support the concept that L-lactate acts as a classic allosteric effector, which means that the effector binds preferentially to oxyhaemocyanin at a site other than the oxygen-binding site and between subunits rather than to a specific site, stabilizing a particular allosteric conformation of the protein. However, approximately three L-lactate-binding sites were found per hexamer in Callinectes sapidus haemocyanin, with a dissociation constant of 3.2 mmol l⁻¹ for each site (Johnson et al., 1984).

By performing urate competition experiments using several urate analogues, it was shown that the purine ring system is essential for interaction with the urate-binding site. The urate analogue caffeine was found to bind with a higher affinity to the urate-binding site than the naturally occurring effector and to be more effective in modulating the cooperativity and affinity of haemocyanin oxygen-binding (Zeis, 1994).

Additional insights into haemocyanin/effector interactions were gained using isothermal titration calorimetry. Binding of oxygen to the haemocyanin of H. vulgaris was known to be linked both to the proton concentration and to the concentration of urate or caffeine. Furthermore, the magnitude of the Bohr effect was found to be dependent on the urate concentration (Zeis et al., 1992; Zeis, 1994). To evaluate whether there was any link between the effectors (urate or caffeine) and protons at the level of the haemocyanin conformation dominant under oxygenated conditions, the binding of effector to the fully oxygenated haemocyanin of H. vulgaris was investigated at different pH values. The same study investigated whether any groups with a pKa in the physiologically relevant pH range (7.55–8.15) were involved in the binding of the effectors urate or caffeine to the haemocyanin of H. vulgaris (Menze et al., 2000).

Caffeine is a strong effector of the oxygen-binding behaviour of the dodecameric haemocyanin of the lobster H. vulgaris and binds with a higher affinity to the pigment than does the naturally occurring effector urate. At sufficiently high concentrations, caffeine is able to inhibit urate binding completely (Zeis et al., 1992; Nies et al., 1992). Furthermore, caffeine is soluble up to concentrations of 100 mmol l⁻¹, whereas urate is soluble only up to 1 mmol l⁻¹, which limits the range of c values obtainable for binding isotherms of the naturally occurring effector.

Fig. 2 shows raw data and integrated signals of a typical experimental titration of haemocyanin with urate and caffeine in 100 mmol l⁻¹ Hepes buffer, pH 8.0, containing 20 mmol l⁻¹ MgCl₂, 20 mmol l⁻¹ CaCl₂ and 150 mmol l⁻¹ NaCl. In Fig. 2A,C the power that has to be applied to keep the reference cell and the sample cell at the same temperature is plotted against the molar ratio of [effector] to [haemocyanin]. Each spike corresponds to a 5 μl injection of a 1 mmol l⁻¹ caffeine or urate solution into a 25 μmol l⁻¹ haemocyanin and a 31.4 μmol l⁻¹ haemocyanin solution, respectively. Titration of the two ligands causes different exothermic reactions. As the ligand accumulates, the heat signal decreases because of the saturation of haemocyanin with the effector. Fig. 2B,C shows the integrated heat signals (in kJ mol ⁻¹ caffeine or kJ mol⁻¹ urate) plotted against the molar ratio of the total caffeine or urate to the haemocyanin concentration. Under the experimental conditions chosen, an almost sigmoidal binding isotherm was obtained for the caffeine titration. Because of the smaller binding constant for urate, a hyperbolic binding curve for this effector was obtained despite the somewhat higher protein concentration.

The binding of urate to the haemocyanin of H. vulgaris was studied by Nies et al. (1992) using equilibrium dialysis. Caffeine binding was studied indirectly using urate competition experiments. No indication of cooperativity in binding for either ligand was found in this study. However, as demonstrated by Menze et al. (2000), ITC binding isotherms for the two ligands could be conveniently explained only if cooperative binding was assumed for caffeine. This conclusion was deduced by analysing caffeine- and urate-binding isotherms using a non-interacting site model. By using this model, different numbers of binding sites n on the dodecameric haemocyanin (H. vulgaris) were obtained for urate (n=2) and caffeine (n=3.2). Since previous non-calorimetric binding studies (Nies et al., 1992) had revealed two binding sites for each of the two structurally similar ligands, a different number of binding sites seemed unlikely. Menze et al. (2000) has shown that a ligand/macromolecule system having two binding sites with different binding constants and slightly different enthalpies of binding can lead to a change in the number of binding sites measured if the binding isotherms are analysed using a non-interacting site model.

A similar pattern was observed by Fisher and Tally (1997, 1998), who studied the binding of ADP and NADPH to bovine liver glutamate dehydrogenase also using ITC. However, in contrast to previous binding studies using different methods, only ITC was successful in detecting substantial interactions between at least three of the individual binding sites of this hexameric protein. In this system, the three sequential binding constants for the formation of the binary and tertiary complexes were nearly identical, while the corresponding ΔH⁰ and ΔS⁰ values changed significantly in both magnitude and sign. These observations led to the new concept of isoergonic cooperativity in ligand binding (Fisher and Tally, 1997, 1998).

An analysis of caffeine and urate binding to the haemocyanin of Homarus vulgaris by ITC revealed two binding sites that displayed cooperative binding for caffeine but not for urate. The binding of urate was characterised by a microscopic binding constant (K) of 8500 l mol⁻¹ and an enthalpy change (ΔH⁰) of −135.1 kJ mol⁻¹ for binding to both sites. Caffeine bound cooperatively to haemocyanin with two microscopic binding constants of 14 100 l mol⁻¹ (K₁) and 40 400 l mol⁻¹ (K₂). The corresponding enthalpy changes for binding were ΔH⁰1=−97.5 kJ mol⁻¹ and ΔH⁰2=−113.4 kJ mol⁻¹ (Menze et al., 2000).

Urate binding exhibits a more unfavourable entropy change (ΔS⁰) of −0.385 kJ mol⁻¹ K⁻¹ than caffeine binding, with ΔS⁰1=−0.252 kJ mol⁻¹ K⁻¹ and ΔS⁰2=−0.298 kJ mol⁻¹ K⁻¹, leading to a higher free energy change for caffeine binding (ΔG⁰1=−23.3 kJ mol⁻¹, ΔG⁰2=−25.9 kJ mol⁻¹) than for urate binding (ΔG⁰=−22.1 kJ mol⁻¹). As with caffeine binding, urate binding is enthalpy-driven at 20 °C. Both binding processes are characterised by a favourable enthalpy change and an unfavourable entropy change with respect to the free energy change of the system. However, for urate, the favourable enthalpy change is compensated by a larger entropy change, resulting in a smaller free energy change upon ligand binding. Therefore, the ratio between unfavourable entropy changes (release of ordered water molecules and increase in vibrational modes) and favourable entropy changes (loss of rotational, translational and conformational freedom) for caffeine binding is smaller than for urate binding. Since caffeine is an uncharged molecule with the same double ring structure as urate, the difference may originate from the amount of water ordered near the purine ring.

The difference between the binding constants for the two successive binding steps for caffeine is too small to be accurately detected by equilibrium dialysis. However, in the ITC experiments, the deviation from non-cooperative binding was amplified by at least two different binding enthalpies, which function as ‘signals’ in the experiments.

How can the fact that caffeine displays cooperative binding and urate does not be explained? Consider two protein conformations (A and B) for oxygenated haemocyanin, one of which is present at a much higher concentration (e.g. [A] ⪢ [B]). If two such conformations exist in equilibrium and differ in affinity for a certain ligand, the addition of this ligand will increase the relative amount of the conformation with the higher affinity. Thus, the dominant conformation A may be characterised by a higher affinity for urate than that of conformation B. With these assumptions, increasing the urate concentration would lead to only a slight change in the distribution of the two conformations since the high-affinity conformation A is already dominant. Under these conditions, cooperativity in urate binding would not be detectable. In contrast, for caffeine, we have to assume a higher affinity for conformation B. If caffeine is titrated, the conformational distribution would change significantly towards B, which would lead to the observed cooperativity in binding. The hypothesis of two haemocyanin conformations under oxygenated conditions is in accordance with an analysis of the pH-dependence of oxygen binding employing the nesting model for haemocyanin of the closely related lobster H. americanus. In this model, two conformations for the dodecameric oxygenated haemocyanin are required (Decker and Sterner, 1990).

The pH-dependence of urate and caffeine binding to haemocyanin from H. vulgaris was also investigated by Menze et al. (2000) using ITC. A possible dependence on pH might be expected since the magnitude of the Bohr effect was found to depend on the urate concentration (Zeis et al., 1992). No pH-dependence for binding for either ligand could be measured in the pH range investigated (7.55–8.15), since the binding isotherms measured at different pH values were found to coincide. Furthermore, no dependence of ΔH⁰app on the ionization enthalpy of three different buffer systems (ΔH⁰ionz between 20.5 and 48.2 kJ mol⁻¹) were found for caffeine binding to H. vulgaris haemocyanin (Menze et al., 2000). Thus, no change in the pKa of any amino acid was involved when caffeine binding occurred. The observed dependence of the magnitude of the Bohr effect on urate concentration could be explained if two haemocyanin conformations occurred under deoxygenated conditions. If these conformations differed in their affinity for protons and urate/caffeine, the observed behaviour might occur. Thus, in toto, four different conformations have to be assumed for the dodecameric haemocyanin of H. vulgaris. Four conformations were also predicted for the haemocyanins of other crustaceans such as H. americanus and Carcinus maenas on the basis of the analysis of the pH-dependence of oxygen binding employing the nesting model (Decker and Sterner, 1990; Dainese et al., 1998). To provide further insights into the allosteric regulation of haemocyanin, the binding of urate and caffeine to haemocyanin should be investigated under deoxygenated conditions.

This work was supported by the Deutsche Forschungsgemeinschaft (Gr 456/20-1) and the State of Nordrhein-Westfalen (M.K.G.).