Differential Effects of 5-Hydroxytryptamine₄ Receptor Agonists at Gastric versus Cardiac Receptors: An Operational Framework to Explain and Quantify Organ-Specific Behavior

Gastrointestinal (GI) prokinetic drugs activating 5-HT₄ receptors are commonly used to facilitate GI transit. They have proven their value in the treatment of gastroesophageal reflux disease and are still used to treat symptoms in patients suffering from the constipation predominant form of irritable bowel syndrome or from gastroparesis; both conditions can cause considerable impairment of quality of life (Bouras et al., 1999; Galligan and Vanner, 2005). However, the presence of functional 5-HT₄ receptors in the human heart might be a concern when using these drugs to treat functional GI disorders (Tonini et al., 1999).

5-HT₄ receptors are seven-transmembrane domain receptors, positively linked to adenylyl cyclase, resulting in the production of cAMP, which is the first step in a cell type-specific cascade of events leading to effect (Martin and Humphrey, 1994). In the human GI tract, different effects have been related to 5-HT₄ receptor activation. Relaxant 5-HT₄ receptors are present on human colonic circular smooth muscle cells (McLean and Coupar, 1996; Prins et al., 2000b). Mucosal 5-HT₄ receptors in human small intestine stimulate secretory processes (Borman and Burleigh, 1993). Moreover, 5-HT₄ receptors have been shown to be involved in the initiation of the peristaltic reflex in human jejunum (Foxx-Orenstein et al., 1996). In addition, in the human stomach and colon, activation of 5-HT₄ receptors located on myenteric excitatory cholinergic neurones can lead to an increased release of acetylcholine from these neurones (Prins et al., 2000a; Leclere and Lefebvre, 2002; Leclere et al., 2005). Therefore, 5-HT₄ receptors are a compelling target for promotility agents since the facilitation of fast synaptic transmission in myenteric ganglia will increase motor reflexes, whereas the facilitation of excitatory input to the muscles will enhance motility directly (Galligan and Vanner, 2005).

The cardiac location of 5-HT₄ receptors acquired a lot of attention because of the cardiac side effects observed with cisapride. Although the observed cardiotoxicity of cisapride is not 5-HT₄ receptor mediated (Mohammad et al., 1997), 5-HT₄ receptor agonists do have inotropic and chronotropic effects in human atria and inotropic effects in human ventricles (Kaumann et al., 1991; Brattelid et al., 2004). Despite the clear tissue-dependent effects of benzamidic 5-HT₄ receptor agonists, a detailed comparison of GI versus cardiac effects of 5-HT₄ receptor agonists within the same species has not yet been performed. Because the interesting potential of 5-HT₄ receptor agonists in the treatment of GI disorders cannot be seen separately from the possible cardiac effects of these prokinetics, we thought it timely to study cardiac and GI effects of 5-HT₄ receptor agonists in an integrated approach.

The pig has been demonstrated to be a relevant species to study human atrial 5-HT₄ receptor interactions (Kaumann, 1990). It is also a good species for studying human digestive function given the similarities between the morphology and physiology of their GI tracts (Miller and Ullrey, 1987), and it has been shown that porcine proximal stomach can be used to investigate presynaptic modulation of myenteric acetylcholine release (Leclere and Lefebvre, 2001).

We now developed a functional assay in porcine proximal stomach to study gastric neuronal 5-HT₄ receptors. By means of a “coupled” in vitro assay, we compared the cardiac versus gastric behavior of the 5-HT₄ receptor agonists prucalopride, tegaserod, R149402, R199715, and 5-HT. Furthermore, we implemented the framework provided by the operational model of agonism (OMOA) to analyze and classify the agonists, using a mixed-effects modeling approach. This model allows for simultaneous agonist- as well as antagonist-related parameter estimation and provides a tool to predict the expression of partial agonism of new ligands in the heart and, hence, their potential side effects.

Materials and Methods

Tissue Preparation

Female pigs (10–11 weeks, 22–27 kg), obtained from local farms, were anesthetized with an i.v. (50 mg/kg) sodium pentobarbital (Kela N.V., Hoogstraten, Belgium) injection. After exsanguination, the heart and the entire stomach were dissected and placed in Krebs-Henseleit solution (11.1 mM glucose, 2.51 mM CaCl₂, 25 mM NaHCO₃, 1.18 mM MgSO₄, 1.18 mM KH₂PO₄, 4.69 mM KCl, 0.033 mM EDTA, and 118 mM NaCl). Right and left atrium preparations were obtained as described previously (De Maeyer et al., 2006). In brief, for chronotropic effect studies, the right atrium was removed and mounted in toto. Left atrial pectinate muscles (9–12 per left atrium) with a thickness <1 mm and a length varying between 3 and 7 mm were dissected away from the endothelial surface and mounted between platinum wire electrodes, aligned perpendicularly to the tissue.

The stomach was opened along the lesser curvature, and the contents were rinsed out. After removal of the mucosa, the tissue was placed at 37°C, and muscle strips of approximately 1.5 cm in length were prepared from the ventral side of the proximal stomach in the direction of the longitudinal muscle layer. Most of the circular muscle layer was removed, taking care not to damage the myenteric plexus. The longitudinal muscle strips were vertically attached onto tissue holders, equipped with two coaxially aligned platinum wire electrodes. The tissue was kept at 37°C and continuously gassed with 95% O₂ and 5% CO₂ in an organ bath set-up containing 20 ml of Krebs-Henseleit solution.

Changes in isometric force of the tissues were recorded via Statham UC2 force transducers (Gould, Cleveland, OH) and DBA 18 digital bridge amplifiers (Anerma, Westerlo, Belgium) on a Powerlab data acquisition system and recorded using Chart version 5.1.1 software. Electrical field stimulation (EFS) was performed with a constant voltage stimulator (Janssen Pharmaceutica N.V., Beerse, Belgium). All tissues were used on the day of preparation. The study was approved by the ethical committee from Johnson and Johnson Pharmaceutical Research and Development, a division of Janssen Pharmaceutica N.V.

Experimental Protocols

Left Atrial Pectinate Muscles. The protocol used for the pectinate muscles was essentially as described previously (De Maeyer et al., 2006). In brief, pectinate muscles were electrically stimulated (0.5 Hz, 5 ms, just above threshold voltage, resting length being half of L_max, the length at which maximal active tension was developed upon EFS). For this study, a cumulative concentration-response (C-R) curve to the agonists with semilog unit concentration increments was established in the paced pectinate muscles in the presence of propranolol (0.2 μM) to avoid indirect β-adrenoreceptor interactions due to the release of noradrenaline, cocaine (6 μM), to reduce tissue capture of 5-HT, and 3-isobutyl-1-methylxanthine (IBMX; 20 μM), and to block the action of phosphodiesterase enzymes. Experiments were terminated by the administration of a saturating concentration of isoprenaline (0.1 mM).

Right Atrium. Spontaneously beating right atria were studied as described previously (De Maeyer et al., 2006). In brief, a cumulative C-R curve was established for all of the agonists under study (one curve per preparation) in the presence of 0.2 μM propranolol and 6 μM cocaine. In the end, isoprenaline (0.1 mM) was administered to the solution.

Gastric Longitudinal Muscle Strips. Because preliminary experiments showed an inhibitory effect of endogenous nitric oxide on EFS-induced contractions, experiments with gastric muscle strips were carried out in the presence of 0.1 mM nitric-oxide synthase inhibitor N^G-nitro-l-arginine-methylester (l-NAME). Indomethacin (1 μM) was added to avoid spontaneous contractions due to the synthesis of prostaglandins. The tissues were allowed to equilibrate for 90 min with rinsing every 15 min under a resting tension of 20 mN. After this stabilization period, the strips were contracted with carbachol (3 μM) to ensure their viability and responsiveness. After washout, this step was repeated at 30-min intervals until two similar successive responses to carbachol were obtained (usually three times). Resting tone was reset at 20 mN between each step. After the last washing step, organ baths with muscle strips that were to receive 5-HT as agonist were supplied with methysergide (1 μM) to block interactions of 5-HT at 5-HT₁, 5-HT₂, 5-ht₅, 5-HT₆, and 5-HT₇ receptors, and granisetron (0.3 μM) to exclude 5-HT₃ receptor activation. Cocaine (6 μM) was also added to these organ baths, because preliminary experiments had shown that the presence of cocaine produced a leftward shift (one-half log unit; P < 0.01) of the C-R curve to 5-HT (cumulative as well as noncumulative). Thirty minutes after the last washing step, the gastric muscle strips were electrically stimulated (every 3 min, a 10-s pulse train at 4 Hz, 0.5 ms, and 20 V, corresponding current 0.4–1 A). Preliminary experiments showed that, under these conditions, a stimulation voltage of 20 V resulted in maximal contractions. Once successive electrically evoked contractions became reproducible, the applied voltage was adjusted to reduce the contraction force to 50% of the force developed at 20V (corresponding current 0.11–0.5 A), after which the electrical stimulation was set on hold for 20 min. EFS was then reinstated at this adjusted voltage, and after at least 15 min, the agonist was added in a single concentration (seven increasing concentrations in seven parallel strips) or in a semilog unit increment cumulative manner (one strip) from 0.01 to 1 nM onwards, depending on the agonist; when the agonist was added cumulatively, the time interval between successive concentration increments was agonist-dependent, ranging from 10 to 15 min (corresponding to three to five EFS trains) for 5-HT to 30 min (10 EFS trains) for tegaserod. From 3 μM prucalopride onwards, a supramaximal sharp increase in electrically induced gastric contraction occurred, probably mediated through inhibition of cholinesterase enzymes (J. A. J. Schuurkes, unpublished data). The involvement of this non-5-HT₄ receptor-related interaction in the generation of the response at lower concentrations of prucalopride is unlikely, considering the binding profile of prucalopride. Therefore, the effects of prucalopride up to this concentration were used for monophasic fitting procedures. The strips that received one concentration of agonist were followed for 70 (5-HT, prucalopride, and R149402) or 120 (tegaserod and R199715) min.

Assessment of the Antagonizing Effect of GR113808 versus Prucalopride and 5-HT. Both in pectinate muscles as well as in gastric muscle strips, the effect of the 5-HT₄ receptor antagonist GR113808 on the cumulative C-R curve of prucalopride was evaluated. In left atrial pectinate muscle preparations, GR113808 (1, 3, 10, 30, 100, or 300 nM) was administered to parallel tissues 30 min before constructing a cumulative C-R curve. In the gastric preparations, GR113808 (1, 3, 10, 30, 300, or 1000 nM) was added 60 min before starting the cumulative administration of prucalopride. The effect of GR113808 (stomach, 1 and 10 μM; left atrium, 0.3 μM) was also evaluated on the responses evoked by cumulatively administered 5-HT in the stomach and the left atrium.

Data Analysis

The average contraction to three EFS trains (stomach) or during 2 min (left atrium) or the average beating rate during 2 min (right atrium) before the addition of agonist was taken as the initial value. All responses were expressed relative to this initial value per se for gastric experiments or relative to the increase above the initial value caused by 0.1 mM isoprenaline for atrial experiments. Increases in contraction force were quantified using the maximal response. When fitting C-R curves, the concentrations where the response declined more than 10% under the response obtained at the previously administered agonist concentration were excluded.

Hill Equation Curve Parameters. To obtain curve parameter estimates for midpoint location [EC₅₀, estimated as –log(EC₅₀)], upper asymptote of the observed maximal effect (α), and Hill slope (n_H), C-R curves to the agonists were fitted to the Hill equation by nonlinear regression using a mixed-effects approach (i.e., interindividual variation in the parameters was accounted for; PROC NL-MIXED; SAS version 9.1; SAS Institute, Cary, NC).

OMOA. To quantify the differences in the expression of agonism of the 5-HT₄ receptor agonists across the different tissues, the OMOA was applied (Black and Leff, 1983). where E_max is the maximal pharmacological effect achievable in the system, K_A is the agonist-receptor complex dissociation constant (the reciprocal of which defines agonist affinity), n_t is the slope index for the transducer relation (the occupancy-effect function) and therefore is a measure of the sensitivity with which the system transduces AR into E, and τ is the efficacy parameter, which is defined by the ratio of total functional receptor concentration (R₀), and K_E, which is the value for half-saturation of the transducer function. Therefore, the reciprocal of τ represents the fraction of receptors that needs to be occupied to achieve half-maximal tissue response.

Modeling. The experimental data were fitted to the operational model of agonism (eq. 2) using a nonlinear mixed-effects model (nonlinear mixed-effects package as implemented in S-PLUS version 6.2; Insightful Corporation, Seattle, WA; the S-PLUS code can be obtained with the authors). The efficacy parameter τ is estimated for each compound and for each tissue. Because within the same tissue the tissue-dependent aspects of K_E are canceled and only the drug-specific aspects of efficacy for the agonists are relevant (relative values of the activated receptor, G-protein complex dissociation constant, K_AR), the ratio of the τ values between the different compounds is kept constant across the different tissues (or vice versa, the ratio of τ values between the different tissues is compound-independent). For all models, interindividual variability on the τ parameter for all compounds was modeled by an exponential equation. In general, where P_i is the individual value for the parameter (τ_{i, compound}), P_pop is the population value for the parameter (τ for that compound), i.e., the value for a typical individual, and η_i is the random deviation of P_i from P_pop. The values of η_i are assumed to be independently normally distributed with mean zero and variance ω². For each tissue, a different E_max and n_t is estimated because these parameters are assumed to be tissue-dependent (Leff et al., 1990). In the stomach, responses are expressed as percentage increase, without normalization. Therefore, since Van Der Graaf and Danhof (1997) have shown that ignoring interindividual variation in E_max may have resulted in erroneous estimates of affinity and efficacy, interindividual variability of E_max in the stomach was accounted for and modeled by an exponential equation (eq. 3). In the left and right atria, on the other hand, interindividual variability of E_max was assumed to be insignificant because, in these tissues, expression of the response relative to that evoked by isoprenaline reduces the variation in the maximal achievable contractile force between the individual animals. It was not possible to account for interindividual variability of n_t, because this resulted in overparameterization of the model. However, the effect of interindividual variation of the slope parameter on affinity and efficacy parameter estimates has been shown to be relatively small (Van der Graaf and Danhof, 1997). For each compound, K_A was assumed to be common for all tissues since receptor affinity is generally considered to be constant across different tissues of the same individual. Moreover, since K_A is assumed to be constant across individuals of the same strain, interindividual variability of pK_a was assumed to be insignificant. K_A and τ were estimated as –log(K_A) and log τ, respectively, because these parameters are assumed to be log-normally distributed (Leff et al., 1990; Van der Graaf and Danhof, 1997). Heteroscedasticity in the intraindividual variance (residual errors) was modeled by a power variance function: where yp_ij is the j^th response for the i^th animal predicted by the model, ym_ij is the measurement, and ϵ accounts for the residual deviation of the model predicted value from the observed response. This results in an additive or proportional residual error model when the power factor is 0 or 1, respectively (and c = 1). For each tissue, a different value of c is estimated. The values for ϵ are assumed to be independently normally distributed with mean zero and variance σ². The residuals were analyzed to verify this assumption. For all compounds in all tissues, the mean of the residuals was near zero. Therefore, the residuals distributed evenly around the fitted function, and the model does not deviate in a systematic manner from the data. The deviation of the measured data from the fitted response can therefore be attributed to random deviation. To compare models with a different fixed effects structure, a maximum likelihood estimation method was used during the optimization and model building process. Criteria for selecting the final model included improvement in the residual plots, increased precision in the parameter estimates, reduced residual variances for the parameter estimates, and reduced the objective function value (–2 × log likelihood). Population values for the final model parameters and variances ω and σ are estimated using a restricted maximum likelihood function.

The model was also applied to the near full data set, including the data with prucalopride in the presence of increasing concentrations of GR113808 (see Results). The same modeling procedure was followed as described above. K_B (see Results) was estimated as –log(K_B), because it has a log-normal distribution. To account for the presence of antagonist, residual error was modeled by eq. 4 in which the value of the constant c is allowed to vary with the presence or absence of antagonist.

Drugs

The following drugs were used: IBMX and l-NAME (Fluka, Buchs, Switzerland); propranolol HCl (Sigma-Aldrich, St. Louis, MO); methysergide maleate (Sigma/RBI, Natick, MA); indomethacin (Merck, Darmstadt, Germany); GR113808 (Tocris Cookson Inc., Bristol, UK); atropine sulfate and 5-hydroxytryptamine creatinine sulfate (5-HT; Acros Chimica, Geel, Belgium); tetrodotoxin (TTX; Serva, Heidelberg, Germany); and cocaine HCl, granisetron HCl, isoprenaline HCl, tegaserod, prucalopride HCl, R149402 HCl, R199715 HCl, and carbachol (Johnson and Johnson Research and Development, Beerse, Belgium). All compounds were dissolved and diluted in distilled water, with the exception of GR113808, tegaserod, and indomethacin. GR113808 was freshly dissolved in dimethyl sulfoxide to obtain a stock solution of 10 mM; dilutions were made with distilled water. A stock solution of tegaserod (1 mM) was made in distilled water containing 20% cyclodextrin and diluted with distilled water; indomethacin was dissolved in 9.1 ml of distilled water supplemented with 0.9 ml of 2% Na₂CO₃. These solutions were stored at –20°C.

Results

We previously reported an in-depth analysis of the 5-HT₄ receptor-mediated effects in the porcine left and right atria (De Maeyer et al., 2006). The curve location parameters of the cumulative C-R curves for the inotropic effect in left atrium and the chronotropic effect in right atrium used for comparison with the gastric responses in the current study are given in Table 1. No previous description of functional gastric 5-HT₄ receptors in the pig was made. Therefore, the gastric results will be presented in detail before we describe the comparison between atrial and gastric tissues.

Effects of 5-HT₄ Receptor Agonists on Porcine Gastric Muscle Strips

Under isometric conditions, the muscle strips displayed very little to no spontaneous contractility throughout the experiment. In preliminary experiments, it was shown that the EFS-induced contractions were blocked by both tetrodotoxin (0.3 μM) and atropine (0.3 μM), indicating a neuronal mechanism involving acetylcholine.

Effects of 5-HT on the Proximal Stomach. 5-HT induced a concentration-dependent increase of the resting tension (Fig. 1B) starting from 0.03 μM 5-HT onwards; the highest concentration tested (10 μM) resulted in an increase in tone of 22.4 ± 7.3%, expressed as percentage of the response to carbachol; the maximal effect was not yet reached. When constructing a cumulative C-R curve to 5-HT (up to 100 μM 5-HT) in the presence of GR113808 (1 and 10 μM), the increase in basal tone by 5-HT was not prevented, and basal tone continued to increase up to the concentration of 100 μM 5-HT. The amplitude of the EFS-induced contractions was always measured from the newly prevailing basal tone.

5-HT nontransiently enhanced EFS-induced contractions, both in the cumulative (Fig. 1B) and the noncumulative administration procedure. The increase in EFS-induced contraction force reached a maximum after 9 to 12 min (three to four electrical pulse trains). In approximately 50% of the animals, the response to the higher single concentrations spontaneously decayed. No difference in the mean location parameters between the C-R curves obtained with the two administration methods was found. The location parameters for the cumulative C-R curve of 5-HT in gastric muscle are given in Table 1.

Effects of Prucalopride, Tegaserod, R149402, and R199715 on the Proximal Stomach. Prucalopride, tegaserod, R149402, and R199715 had no influence on resting tension, but all nontransiently enhanced EFS-induced contractions (as shown for a single concentration of prucalopride in Fig. 1A). The kinetics of these increases in EFS-induced contraction force were agonist-specific, reaching a maximum after 15 to 50 min (the higher the concentration of agonist, the quicker a maximum was reached) for prucalopride and after 40 to more than 100 min for tegaserod. No difference between cumulative and noncumulative administration was found for any of the compounds; the location parameters for the cumulative C-R curves are given in Table 1.

Because methysergide (1 μM) was present in experiments with 5-HT and methysergide itself enhanced the EFS-induced contractions, it was tested whether the presence of this antagonist influenced the response to the selective 5-HT₄-receptor agonist prucalopride to justify the comparison with 5-HT. In these experiments, basal EFS-induced contractions were 30.8 ± 0.9%, whereas in the absence of methysergide, a significantly lower basal response of 24.2 ± 1.1% was measured (both values, n = 48 tissues from six animals; as percentage of the response to carbachol; P < 0.001, unpaired Student's t test between basal values in experiments with prucalopride in the presence versus those in the absence of methysergide). We have no explanation for the effect of methysergide per se on EFS-induced contractions. However, the maximal effect and pEC₅₀ of the cumulative C-R curve of prucalopride in the presence of methysergide were not different from those in the absence of methysergide. Likewise, no difference in response to prucalopride in the absence or presence of methysergide was observed for any concentration tested in the noncumulative setup.

Antagonistic Effect of GR113808 versus Prucalopride and 5-HT in Stomach and Left Atrium

The effect of preincubation with increasing concentrations of GR113808 on the cumulative concentration-effect curves to prucalopride in the gastric muscle preparations and in the left atrial trabeculae is shown in Fig. 2, A and D, respectively. The effect of GR113808 on the 5-HT-induced contractions in stomach and atrium is shown in Fig. 3, A and B, respectively. The maximal effect of the concentration-effect curves shows a downward tendency with increasing GR113808 concentrations (Figs. 2, A and D, and 3A).

Development of a Framework to Compare Gastric and Atrial Responses

To compare the behavior of the agonists at atrial and gastric 5-HT₄ receptors and to explain the differences observed between the location parameters of the cumulative C-R curves in the three tissues (Table 1), we applied the OMOA to the data. Table 2 shows the resulting population estimates for the different tissues together with the interindividual variability (only allowed for log τ and for the E_max in the stomach) expressed as a coefficient of variation. To fit the data simultaneously, including those with prucalopride in the presence of GR113808, the OMOA was adapted to account for the presence of a competitive reversible antagonist. In this equation, B represents the antagonist concentration, and K_B is the antagonist dissociation constant. The variability of the Hill slope in the right atrium seems to be greater than in the other tissues (Table 1). Therefore, interindividual variability of the slope factor in the right atrium was accounted for (which was not possible in the basic model because this resulted in overparameterization of the model). In the stomach, in the presence of the higher concentrations of GR113808 (above 30 nM), the contractile force measured just before the first concentration of prucalopride that evoked a clear response was often increased relative to the basal response (Fig. 2A). Therefore, we introduced the parameter E_basal to account for the increased baseline for these data. This largely improved the model fit (135-point change in the objective function value). The resulting model fit is shown in Fig. 2, B and E. This model results in a pK_B estimate of 9.43 ± 0.04 and 9.63 ± 0.05 in atrium and stomach, respectively. In the atrium, the model-predicted response in the presence of increasing concentrations of GR113808 is systematically overestimating the observations at high-agonist concentrations (Fig. 2E). This is less clear in the stomach, because the maximal effect is often not yet reached (Fig. 2B). However, the maximal effect in the presence of 3 and 10 nM GR113808 seems to overestimate the observed responses. Furthermore, also for 5-HT, an influence of 1 and 10 μM GR113808 on the upper asymptote was observed in the stomach (Fig. 3A).

Because eq. 5 cannot account for a change in the upper asymptote, the model was further adapted by allowing for possible partially insurmountable antagonism of GR113808. The parameter q was introduced, with q being the reduction of τ in the presence of GR113808 (i.e., allowing for a change in the operational receptor number or in the signal transduction efficiency by GR113808). If q is interpreted as the fraction of receptors available for interaction with an agonist after “pseudoirreversible” blockade with an antagonist (Zernig et al., 1996), it corresponds to the q value of Furchgott (Black et al., 1985; Zernig et al., 1996). q depends on the concentration of antagonist used and is modeled by an hyperbolic equation to allow for the saturation of the relation between log [B] and α. In this equation, q_max is the maximal reduction of τ by GR113808, whereas B₅₀ denotes the antagonist concentration at which the reduction is 50% of the value of q_max. The complete data set was therefore fitted using the following model: Because no assumptions are made about the underlying mechanism(s) of the insurmountable antagonism, q_max can well possess (partially) tissue-related properties. Therefore, interindividual variability of this parameter was considered. K_B and B₅₀ are antagonist-related properties, and interindividual variability was assumed to be insignificant.

The model, adapted to accommodate for partially insurmountable antagonism by GR113808 in the atrium, predicts the observations very well and causes a change in the objective function value of 45 points. An additional change by 30 points was obtained when gastric data were also modeled using eq. 7. This model resulted in a pK_B and pB₅₀ estimate in the atrium of 9.02 ± 0.07 and 9.16 ± 0.07, respectively, whereas in the stomach, a pK_B and B₅₀ of 7.91 ± 0.14 and 9.48 ± 0.13 was found (Table 3). This means that, in the atrium (Fig. 2F) and the stomach (Fig. 2C), the data can be fitted if we assume a concurrent insurmountable and competitive action of GR113808. In the stomach, these two actions of GR113808 appear with a different potency. Table 3 summarizes the population predictions obtained with this final model. The parameters are well defined, with the interindividual variability in the atrium slightly higher than that in the stomach. Population-predicted curves for all agonists in each tissue or for each agonist in the three tissues from this final OMOA fit are given in Fig. 4, A to C, and Fig. 5, A to E, respectively. The apparent difference in relative efficacy between R149402 and R199715 in left and right atria (Fig. 4), R149402 being more efficient than R199715 in the left atrium and vice versa in the right atrium, is caused by the rather large interindividual variability for R149402 and especially R199715 in the atria.

Discussion

Differences in benzamide-induced 5-HT₄ receptor-mediated responses between different tissues or species mainly originate from the either partial or full agonistic behavior of these compounds in these different test systems (Langlois and Fischmeister, 2003; Bockaert et al., 2004). In fact, because of the difference with mouse colliculi neurones (Dumuis et al., 1989), 5-HT₄ receptors in the heart were first designated as 5-HT₄-like (Kaumann et al., 1991). Most of these studies use the empirical Hill equation to characterize the effects of benzamides at 5-HT₄ receptors. However, the parameters of the empirical Hill equation contain mixed information on drug-specific properties and characteristics of the biological system, which makes it difficult to appreciate the functional potency of a given 5-HT₄ receptor ligand. This complicates the use of this equation for extrapolation and prediction and renders it inappropriate as a model to predict the expression of agonism and thus possible undesirable side effects of a given agonist (Zuideveld et al., 2004). The OMOA, on the other hand, has already been demonstrated to be very useful to quantify and describe tissue-related efficacy differences (Leff et al., 1990; Janssen et al., 2004). Here, we report the establishment of a framework based on the OMOA that allows for quantitative assessment of the tissue-dependent differences in expressed efficacy of 5-HT₄ receptor agonists in the pig as a model for the human situation.

For the pig to be a representative animal to study tissue-specific 5-HT₄ receptor agonism in man, a relevant GI model needed to be developed in this animal because no proper bioassay was available to analyze the GI motility promoting potency of 5-HT₄ receptor agonists. The data presented in this study now clearly demonstrate that 5-HT, as well as the selective 5-HT₄ receptor agonists prucalopride, tegaserod, R149402, and R199715, concentration-dependently enhanced EFS-induced contractions in the porcine proximal stomach. The involvement of other than 5-HT₄ receptors in the 5-HT-induced effects was ruled out by the presence of methysergide and granisetron (Prins et al., 2000a). Because both TTX and the muscarinergic receptor antagonist atropine blocked the contractions, the 5-HT₄ receptors are probably located on cholinergic neurones. We postulate that the 5-HT₄ receptors are operating presynaptically, facilitating the release of the neurotransmitter. Indeed, the same 5-HT₄-mediated effect on neurotransmission has been described in the human GI tract on neurones projecting to both muscle layers in the colon, as well as to the circular muscle layer in the stomach, whereas no contractile postsynaptically operating 5-HT₄ receptor has been described in the GI tract (Prins et al., 2000a; Leclere and Lefebvre, 2002; Leclere et al., 2005).

Despite the blockade of every 5-HT receptor, with the exception of the 5-HT₄ receptor, 5-HT, but not the other agonists, induced a TTX-insensitive (data not shown) tonic contraction that could not be inhibited by GR113808. This is in line with a previous study of Janssen et al. (2002) who showed, in the same preparation as used in this study, the involvement of an uncharacterized high-affinity receptor population and of the 5-HT_2A receptor in the basal contractile response to 5-HT; in their study, the basal response to 5-HT in the presence of 1 μM methysergide was the same as in our study (Janssen et al., 2002).

We now applied the OMOA to quantify the differences in the expression of 5-HT₄ receptor-mediated agonism in the newly described gastric bioassay and in the porcine left and right atria, which we previously described in great detail (De Maeyer et al., 2006). We used a mixed effects approach, which allows one to fit all data simultaneously and which provides, by incorporating interindividual variability, the necessary flexibility for the model to be practicable with decreased need for a priori assumptions. As a noteworthy example, there is no need for the assumption that the maximal response to a full agonist equals E_max, which is undermined by the inherent difficulty in defining what a full agonist is (Black and Shankley, 1990).

Our data show that the application of the OMOA allows the description and prediction of the tissue-dependent efficacy of 5-HT₄ receptor agonism. The results show that tegaserod has the same τ value as 5-HT, whereas prucalopride, R149402, and R199715 are less efficacious than 5-HT. Operationally, this implies that in a low-efficacy system like the atrium, the latter compounds show less effect than 5-HT and hence that the intrinsic power of prucalopride to produce cardiac side effects is lower than this of tegaserod.

The estimated affinity values for 5-HT, prucalopride, R149402, and R199715 (the latter two being prucalopride-derived experimental 5-HT₄ receptor agonists, developed by Johnson and Johnson Pharmaceutical Research and Development) are in good agreement with results found in receptor binding experiments [pIC₅₀ against GR113808 in HEK cells transfected with the human 5-HT_4(b) receptor, 6.6, 7.0, 8.0, and 8.8, respectively; J. A. J. Schuurkes, unpublished data]. The estimated affinity for tegaserod on the other hand (pK_a, 5.83) is lower than found in receptor binding experiments (pIC₅₀ of 7.4). The estimated slope of the transducer relation in all tissues is smaller than one, implicating a decrease of the Hill slope with increasing values of τ (Black et al., 1985). However, in the case of tegaserod, the Hill slope was smaller than can be expected from its τ value in the different tissues. Difficult penetration of tegaserod into the tissue or possible partial precipitation because of its low solubility could be explanations for the underestimated K_A and the low Hill slope. This idea is supported by the very slow kinetics of the response following the administration of tegaserod compared with the other agonists.

In the left and right atria, the maximal tissue response is estimated to be approximately 100%, i.e., identical to the response to a supramaximal isoprenaline concentration. This implies that 5-HT₄ receptors and β-receptors, which are both positively coupled to cAMP, allow for the same maximal achievable tissue response. The left atrial experiments were performed in the presence of IBMX to exclude the actions of phosphodiesterase enzymes, a condition that greatly enhanced the 5-HT₄ receptor-mediated responses (De Maeyer et al., 2006). It was not possible to include the data in the absence of IBMX in our model since, at equilibrium conditions, none of the agonists showed a positive inotropic effect. 5-HT actually behaved as a partial agonist in the non-IBMX-treated left atrium, and the predicted efficacy for the left atrium in the present study may be overrating the significance for the in vivo situation. In the right atrium, where chronotropic responses to the agonists can be obtained in the absence of IBMX, the E_max estimate greatly exceeded the maximum of the C-R curve for 5-HT, implying that also in this tissue, 5-HT behaved as a partial agonist. A similar finding was reported, when applying the OMOA, for the contractile effect of 5-HT in rings of rabbit aorta (Leff et al., 1990). The inclusion of cardiovascular in vivo data with 5-HT4 receptor agonists in the model would extend its predictive properties and would allow for the prediction of tissue-dependent efficacy in vivo (Van der Graaf et al., 1999).

The flexibility of our framework is clear from the fact that it can accommodate antagonist data. Liu et al. (1992) adapted the OMOA in a similar way to explain their observations with angiotensin II receptor antagonists. From a modeling point of view, our results suggest that GR113808 behaves as a partial insurmountable antagonist in the atrium and in the stomach (GR113808-dependent change of τ and hence of the operational receptor number or the coupling constant). These observations do not stand alone since deviations from competitive antagonism, including insurmountable antagonism and deviation from linear Schild slopes, have been described for 5-HT₄ receptor antagonists, e.g., GR113808, in different tissues, including the piglet right atrium (Medhurst and Kaumann, 1993; Gale et al., 1994; Tam et al., 1995). Furthermore, we have shown previously that in the left atrium, in the presence of IBMX, GR113808 reverted the increased contraction by a single concentration of prucalopride or tegaserod to a level below basal (De Maeyer et al., 2006). This might fit into the recent theories on insurmountable antagonism in which the antagonist bound receptor may adopt multiple conformations and/or states (Vauquelin et al., 2002). Claeysen et al. (2003) have shown that a single mutation in the 5-HT₄ receptor generates a receptor that cannot be bound and activated by 5-HT or tegaserod. Numerous synthetic ligands, on the other hand, as well as GR113808, which behaves as a partial agonist on this receptor, can still stimulate this receptor. This suggests an agonist (or antagonist)-dependent requirement of certain binding sites or conformations. Furthermore, this could explain the different efficacy associated with 5-HT and tegaserod compared with the other agonists (see above). GR1138808 has been shown previously to behave as a partial agonist on the 5-HT_4(h) splice variant, a variant that has been described in porcine tissue (Ullmer et al., 1995; Bender et al., 2000). Thus, it could well be that GR113808 acts as a very weak partial agonist, resulting in the loss of some receptors for restimulation with an agonist. However, finding an explanation for the complex behavior of GR113808 was beyond the scope of this study.

In conclusion, we established a framework that allows quantitative assessment of the tissue dependent differences in expressed efficacy associated with agonists at 5-HT₄ receptors. Our results clearly demonstrate that this provides a practical test for quantifying the expression of partial agonism in the heart by new ligands. This can provide a useful tool for selecting drugs based on their property to exert a balanced GI effect without relevant cardiac effects. Furthermore, our model allows incorporation of antagonist data. The use of mixed-effects modeling allows one to fit all data simultaneously and provides the necessary flexibility to be used without restricting assumptions and opens perspectives to extend the model to in vivo data.