The human histamine H4 receptor (hH4R), coexpressed with Gαi2 and Gβ1γ2 in Sf9 insect cells, is highly constitutively active, and thioperamide [THIO; N-cyclohexyl-4-(imidazol-4-yl)-1-piperidinecarbothioamide] is one of the most efficacious hH4R inverse agonists. High constitutive hH4R activity may have pathophysiological implications in which case inverse agonists may behave differently than neutral antagonists. To learn more about the structural requirements for hH4R inverse agonism, we investigated 25 compounds (indole, benzimidazole, and thienopyrrole derivatives) structurally related to the standard antagonist JNJ-7777120 [1-[(5-chloro-1H-indol-2-yl)carbonyl]-4-methyl-piperazine]. We characterized the compounds in radioligand binding assays by using [3H]histamine ([3H]HA) and in steady-state GTPase assays in the presence (antagonist mode) and absence (inverse agonist mode) of the agonist HA, yielding the following results: 1) Twenty-two compounds were inverse agonists (efficacy: 15–62% of the THIO effect), and only three compounds (12%) showed neutral antagonism. Thus, inverse agonism is far more common than neutral antagonism. 2) The inverse agonistic efficacy of the R5-monosubstituted indole-derived compounds increased with the volume of R5. R5 may interact with Trp6.48 of the rotamer toggle switch and stabilize the inactive receptor conformation. 3) A subset of compounds showed large differences between the Ki value from [3H]HA competition binding and the EC50 value from steady-state GTPase assays, whereas the Kb values were closer to the Ki values. Thus, the two-state model should be extended to a model comprising a constitutively active hH4R state, which can be discriminated by inverse agonists from a structurally distinct HA-stabilized active state.
The concept of inverse agonism is derived from the two-state model of receptor activation (Seifert and Wenzel-Seifert, 2002). According to this model, GPCRs exist in an equilibrium of an active G protein-coupling conformation (R*) and an uncoupled inactive state (R). R* promotes GDP/GTP exchange at the Gα subunit and shows a higher affinity for agonists than R. Thus, agonists activate the receptor by stabilizing the R* state. Neutral antagonists bind to R and R* states with the same affinity without altering the equilibrium. Some receptor molecules spontaneously adopt the R* state and promote G protein signaling in the absence of agonists, which is referred to as constitutive activity. Inverse agonists are ligands that bind preferentially to the R state and reduce the percentage of spontaneously active receptors and, thereby, reduce constitutive activity. Likewise, inverse agonism can be explained by the extended ternary complex model, which takes into account that both free and ligand-occupied receptors can interact with G proteins (Samama et al., 1994).
In the last two decades, constitutive receptor activity and inverse agonism of ligands have been reported for a large number of wild-type GPCRs, e.g., for the δ-opioid receptor (Costa and Herz, 1989), the human formyl peptide receptor FPR-26 (Wenzel-Seifert et al., 1998), the cannabinoid CB1 receptor (Bouaboula et al., 1997), and the histamine receptors H1R (Leurs et al., 2002), H2R (Alewijnse et al., 1998), H3R (Wieland et al., 2001), and H4R (Liu et al., 2001; Schneider et al., 2009). Constitutive activity is not only a common property of wild-type GPCRs, but also is the cause of several diseases (Seifert and Wenzel-Seifert, 2002). Even cancerous cell transformation may be caused by a long-term elevation of second messengers, resulting from constitutive activity of overexpressed receptors (Seifert and Wenzel-Seifert, 2002; Kenakin, 2004). Thus, it is obvious that the targeted development of inverse agonistic drugs will gain more and more importance. However, for most receptors only very little is known about the structural requirements governing inverse agonism and neutral antagonism. Although efforts have been made to establish structure–activity relationships for inverse agonists at GPCRs (Soudijn et al., 2005), most drug development programs are still not designed to systematically investigate this issue.
As we recently reported (Schneider et al., 2009), the hH4R is highly constitutively active when coexpressed with Gαi2 and Gβ1γ2 in Sf9 insect cells using the baculovirus expression system. In steady-state GTPase and [35S]GTPγS binding assays thioperamide (THIO) was found to be one of the most efficacious inverse agonists at the hH4R. The benzimidazole derivative JNJ-7777120 [1-[(5-chloro-1H-indol-2-yl)carbonyl]-4-methyl-piperazine], which behaves as a neutral hH4R antagonist in reporter gene assays (Lim et al., 2005), showed a partial inverse agonistic effect in our system, reaching ∼30% of the THIO efficacy (Schneider and Seifert, 2009; Schneider et al., 2009).
In contrast to the hH3R, which plays an important role in neurological and psychiatric diseases, the hH4R is an interesting drug target for the therapy of inflammation, allergy, and autoimmune disorders (Tiligada et al., 2009). Specifically, the hH4R is involved in the pathogenesis of pruritus and asthma (Thurmond et al., 2008; Neumann et al., 2010). In past years, a large number of structurally diverse compounds targeting the hH4R were developed with an increasing focus on the aminopyrimidine scaffold (Sander et al., 2009; Smits et al., 2009). If hH4R-related diseases are caused or promoted by extraordinarily high constitutive hH4R activity, inverse agonists at the hH4R would be an interesting therapeutic option. However, if inverse agonists caused receptor up-regulation, resulting in rebound effects after discontinuation of the therapy, neutral hH4R antagonists would be preferred.
Therefore, we started a project to investigate structure–activity relationships for inverse agonism and neutral antagonism at the hH4R. Here, we report on the investigation of a series of 25 indole, benzimidazole, and thienopyrrole derivatives that were described previously as hH4R ligands (Venable et al., 2005) and are structurally related to the reference compound JNJ-7777120. We characterized the compounds by using Sf9 cell membranes coexpressing the hH4R with Gαi2 and Gβ1γ2. We performed steady-state GTPase assays and high-affinity radioligand binding assays with [3H]HA. Structure–activity relationships were established by molecular modeling and correlating the efficacy values with different molecular descriptors such as logP and molar volume.
Materials and Methods
The baculoviruses encoding the N-terminally FLAG and C-terminally hexahistidine-tagged hH4R were prepared as described recently (Schneider et al., 2009). Recombinant baculovirus for the unmodified versions of the Gβ1γ2 subunits was a kind gift of Dr. P. Gierschik (Department of Pharmacology and Toxicology, University of Ulm, Ulm, Germany). Baculovirus encoding Gαi2 was donated by Dr. A. G. Gilman (Department of Pharmacology, University of Southwestern Medical Center, Dallas, TX). The antibody selective for Gαi1/2 was purchased from Calbiochem (San Diego, CA), and the M1 anti-FLAG antibody was obtained from Sigma-Aldrich (St. Louis, MO). The series of H4R antagonists investigated in this article were synthesized as described previously (Venable et al., 2005) and provided by Johnson and Johnson Pharmaceutical Research and Development (San Diego, CA). Stock solutions (10 mM) were prepared in Me2SO. The dilution series was prepared in 10% (v/v) dimethyl sulfoxide (DMSO). Because of low solubility, for compounds 9 and 10 50% (v/v) DMSO was used for the preparation of the 1 mM solution. THIO was obtained from Tocris Bioscience (Bristol, UK), and the dilution series was prepared in 10% DMSO. The final concentration of DMSO in the sample was 5% for compounds 9 and 10 and 1% for the other compounds. [3H]HA (specific activity 14.2–18.1 Ci/mmol) was obtained from PerkinElmer Life and Analytical Sciences (Waltham, MA). We prepared [γ-32P]GTP by using GDP and [32P] (orthophosphoric acid, 150 mCi/ml; obtained from PerkinElmer Life and Analytical Sciences) according to a previously described enzymatic labeling procedure. All other reagents were of the highest purity available from standard suppliers. Radioactive samples were counted in a PerkinElmer Life and Analytical Sciences Tricarb 2800TR liquid scintillation analyzer.
Generation of Recombinant Baculoviruses, Cell Culture, and Membrane Preparation.
Sf9 cells were cultured in 250- or 500-ml disposable Erlenmeyer flasks at 28°C under rotation at 150 rpm in SF 900 II medium (Invitrogen, Carlsbad, CA) supplemented with 5% (v/v) fetal calf serum (Biochrom, Berlin, Germany) and 0.1 mg/ml gentamicin (Lonza Walkersville, Inc., Walkersville, MD). Cells were maintained at a density of 0.5 to 6.0 × 106 cells/ml. Recombinant baculoviruses were generated in Sf9 cells by using the BaculoGOLD transfection kit (BD Pharmingen, San Diego, CA) according to the manufacturer's instructions. After initial transfection, high-titer virus stocks were generated by two sequential virus amplifications. The supernatant fluid from the second amplification was stored under light protection at 4°C and used as routine virus stock for membrane preparations.
Infection of the cells with baculoviruses encoding hH4R, Gαi2, and Gβ1γ2 was performed as described previously (Preuss et al., 2007). Sf9 membranes were prepared as described previously (Gether et al., 1995), using 1 mM EDTA, 0.2 mM phenylmethylsulfonyl fluoride, 10 μg/ml benzamidine, and 10 μg/ml leupeptin as protease inhibitors. Membranes were suspended in binding buffer (12.5 mM MgCl2, 1 mM EDTA, and 75 mM Tris/HCl, pH 7.4). All membrane preparations were stored at −80°C until use.
[3H]HA Binding Experiments.
Before the experiments, membranes were sedimented by a 10-min centrifugation at 4°C and 13,000 rpm and resuspended in binding buffer (12.5 mM MgCl2, 1 mM EDTA, and 75 mM Tris/HCl, pH 7.4). For competition binding experiments 10 nM [3H]HA and four to six appropriate concentrations between 1 nM and 100 μM of the test compound were used. Nonspecific binding was determined in the presence of THIO (10 μM). Incubations were performed for 60 min at 25°C with shaking at 250 rpm. Bound radioligand was separated from free radioligand by filtration through GF/C filters pretreated with 0.3% (m/v) polyethyleneimine and washed three times with 2 ml of ice-cold binding buffer (4°C). Filter-bound radioactivity was determined by liquid scintillation counting.
Steady-State GTPase Assay.
Steady-state GTPase assays were essentially performed as described previously (Schneider et al., 2009). The samples did not contain NaCl, and the reaction temperature was 25°C. Inverse agonist pEC50 values were determined in the absence of HA, whereas pKb values were calculated from pIC50 values determined in the presence of agonist (HA, 100 nM). For all assays four to six appropriate concentrations between 1 nM and 100 μM of the test compound were used. To quantify efficacies, THIO (10 μM) was used as reference for the maximum inverse agonistic effect.
Based on the crystal structure of the human β2 adrenergic receptor a homology model of the inactive hH4R was generated as described previously (Deml et al., 2009). Glu5.46 was modeled in its protonated state, as proposed (Jongejan et al., 2008). The unsubstituted compound (compound 2) was docked manually into the binding pocket of hH4R. Subsequently, molecular dynamic simulations were performed as described previously (Strasser et al., 2008), including the natural surrounding of the receptor, such as lipid bilayer and water. For all molecular dynamic simulations, the software Gromacs (http://www.gromacs.org) was used.
Determination of Descriptors for Quantitative Structure–Activity Relationship.
To search for dependence of efficacy, several ligand-specific descriptors, such as logP, polar surface area, molar refractivity, molar volume, refraction index, and polarizability, were calculated (ACD/PhysChem, release 11.0; ACD Labs, Toronto, Canada).
Protein concentrations were determined with the Bio-Rad DC protein assay kit (Bio-Rad Laboratories, Hercules, CA). Membrane proteins were separated on SDS polyacrylamide gels containing 10 or 12% (m/v) acrylamide. Proteins were transferred onto Trans-Blot nitrocellulose membranes (Bio-Rad Laboratories) and reacted with M1 anti-FLAG (1:1000) or anti-Gα1/2 (1:1000). Protein bands were visualized by enhanced chemiluminescence (Pierce Chemical, Rockford, IL) using goat anti-mouse IgG (Sigma-Aldrich; 1:2000) or donkey anti-rabbit IgG (GE Healthcare, Little Chalfont, Buckinghamshire, UK; 1:10,000), both coupled to horseradish peroxidase.
Radioligand competition experiments and steady-state GTPase assays were analyzed by nonlinear regression with Prism 5.01 software (GraphPad Software Inc., San Diego, CA). For every compound the curves of all assays were merged in one file, and a global fit over all data was performed. All values are given as means ± S.E.M. Significance was defined as p < 0.05 (confidence interval 95%). In the case of multiple t tests (comparisons of pEC50 with pKi or pKb) the Bonferroni correction was applied (44 t tests), resulting in a new significance level of p < 0.001.
The compounds were characterized in radioligand competition binding assays using [3H]HA (10 nM) and membranes from Sf9 cells coexpressing the hH4R with Gαi2 and Gβ1γ2. As shown in Fig. 1A, there is a good correlation (p < 0.0001; r2 = 0.88; slope 1.12 ± 0.09) between the pKi values determined in the Sf9 cell system and previously reported results (Venable et al., 2005), indicating that the rank order of affinity within the compound series was retained in our experimental system. However, the pKi values determined in Sf9 cell membranes were on average by 0.68 log units lower than the literature data.
Our results show (Table 1) that in the series of piperazine-methylated and R5-monosubstituted indole derivatives the pKi values are influenced by the R5 residues in the following rank order (compound numbers in parentheses): Cl (4) > Br (5) ∼ F (3) > CH3 (6) ∼ NH2 (8) ∼ H (2) ≫ OCH3 (7) ∼ CF3 (9). Thus, halogen substitution in position R5 increases binding affinity, whereas an OCH3 or a CF3 residue in this position considerably reduces the pKi value. Introduction of an additional F or Cl in position R7 of compound 3 or 4 yields the structures 10 and 11 with slightly (–0.30 log units) reduced affinity (Table 1). By contrast, R5/R7 disubstitution with CH3 (12) did not significantly alter the binding affinity, compared with the corresponding R5-monosubstituted compound 6. This indicates that substituents in R7 are well tolerated without major reduction of affinity (Table 1). In fact, R7 monosubstitution with NH2 (15) and CH3 (16) is even beneficial and increases the binding affinity by ∼0.5 log units, compared with the totally unsubstituted compound 2 (Table 1).
We also investigated different positions of a Br residue in the ring system and found that the pKi value correlates with the position of Br in the rank order R5 (5) ≫ R6 (13) ∼ R7 (14), with ∼1 log unit affinity difference between Br in R5 and R6 (Table 1). This shows again that halogen substitution in position R5 is favored and leads to compounds with high binding affinity. Compound 4, the experimentally most widely used H4R antagonist JNJ-7777120 (Jablonowski et al., 2003; Thurmond et al., 2004), shows the highest affinity in the investigated series and is substituted with Cl in position R5.
When the indole scaffold is replaced by a benzimidazol structure (Table 2), in most cases the affinity is reduced by at least 0.3 log units (compounds 17, 18, 20, and 21 in Table 2 versus compounds 2, 3, 6, and 16 in Table 1). Only when a CF3 substituent is located in position R5, a change from the indole (9) to the benzimidazole (19) structure leads to an increase of the pKi by ∼0.3 log units. Replacement of the phenyl moiety of the indole scaffold by the bioisosteric thiophene ring yielded the thienopyrrol series shown in Table 2. When completely unsubstituted, this change of scaffold led to a reduction of the pKi value by ∼0.5 log units (indole derivative 2 in Table 1 versus thienopyrrole derivative 23 in Table 2). Compound 25, which is substituted with Cl in position R4 and CH3 in R5 shows an affinity comparable with the (R5-) Cl-substituted compound 4 (JNJ-7777120).
Steady-State GTPase Assays: Efficacies.
We also investigated the functional properties of the compounds in steady-state GTPase assays. As shown in Tables 1 and 2, the majority of compounds exhibited inverse agonism with efficacies between 15 and 62% of the effect of the hH4R inverse agonist THIO. It is noteworthy that we also found three neutral antagonists (compounds 9, 14, and 19). Figure 2 shows examples for compounds with effects between neutral antagonism and 60 to 70% of the THIO effect. As suggested by the structures of 9 (indole derivative) and 19 (benzimidazole derivative), a CF3 residue in position R5 favors neutral antagonism. Compound 14, which has a Br substituent located in position R7, is also a neutral antagonist. Replacement of Br in R7 by either H (2), NH2 (15), or CH3 (16) results in inverse agonists with efficacies of −0.40 (2), −0.37 (15), or −0.29 (16). This indicates that neutral antagonism requires an electronegative, electron-withdrawing substituent in position R7.
Relatively high inverse agonistic efficacies (range between −0.45 and −0.65) were found for compounds 1, 3, 10, 11, 17, 21, and 22. It is noteworthy that compound 1 is not methylated at the piperazine ring. Although the piperazine methyl substitutent increases affinity in radioligand binding (compound 1 versus 2 in Table 1), it is obviously not necessary for a large inverse agonistic effect. Within the series of R5-monosubstituted indole derivatives (compounds 2–9, Table 1), the efficacy depends on the R5 substituent in the following rank order: CF3 (9) < OCH3 (7) < Cl (4) < CH3 (6) < Br (5) < H (2) < NH2 (8) < F (3). When in the structures of 3 (R5 = F), 4 (R5 = Cl), or 6 (R5 = CH3) a second F (10), Cl (11), or CH3 (12) substituent is introduced in position R7 this does only slightly change inverse agonistic efficacy by 0.10 to 0.20 units. The Br substituent can be moved from R5 (5) to R6 (13) without a significant change in efficacy (Table 1). However, as already mentioned, Br in position R7 leads to a complete loss of inverse agonistic effect, yielding a neutral antagonist.
In the benzimidazole series, for compounds 18, 19, and 20 (Table 2), the efficacies were not significantly different from those of the corresponding indole compounds 3, 6, and 9 (Table 1). Only in case of the completely unsubstituted and the (R7-) CH3-substituted compounds did a change from the indole structure (2 and 16 in Table 1) to the benzimidazole scaffold (17 and 21 in Table 2) increase inverse agonistic efficacy by 0.15 to 0.20. The compounds of the thienopyrrole series show only moderate to low inverse agonism. The thienopyrroles 24 and 25 and the indole derivative 12 represent the three inverse agonists with the lowest efficacy in the series, thus representing almost neutral antagonists. It is noteworthy that all of these compounds have a CH3 substituent located in similar regions of the molecule (R7 in the indole series or R5 of the thienopyrrole scaffold).
Steady-State GTPase Assays: pEC50 and pKb Values.
The pEC50 values of the inverse agonists were determined in the absence of HA (“inverse agonist mode”). The pKb values were calculated from pIC50 values (Cheng-Prusoff equation), determined in the presence of 100 nM of HA (“antagonist mode”). As shown in Fig. 1B, we obtained a clear correlation (p < 0.001; r2 = 0.62) between pEC50 values from steady-state GTPase assays and radioligand binding pKi values. It is noteworthy that Tables 1 and 2 indicate that most of the compounds show tendency toward increased pKi and pKb values compared with the corresponding pEC50 values. This is also demonstrated in Fig. 1, B and D by the reduced slope of the regression lines and by the fact that the majority of data points are located left of the diagonal line. This trend reached significance (p < 0.001, t test with Bonferroni correction for the total number of t tests) for compounds 3 and 11 (pKi ≫ pEC50; filled circles in Fig. 1B) and compounds 1, 3, 7, 10, 11, and 17 (pKb ≫ pEC50; filled circles in Fig. 1D). The pKb or pKi values that are significantly higher than the corresponding pEC50 values are also in bold in Tables 1 and 2.
The correlation between pEC50 values and pKb values was significant, but widely scattered (Fig. 1D; p < 0.001; r2 = 0.36). The best correlation was found for the pKb values from antagonist mode GTPase assays compared with the pKi values from radioligand competition binding (Fig. 1C; p < 0.0001; r2 = 0.66).
Binding Mode of Compound 2.
Molecular dynamics revealed a stable interaction between compound 2 and the receptor (Fig. 3A). The positively charged terminal amino moiety of the piperazine establishes an electrostatic interaction with the highly conserved Asp3.32. A bivalent interaction is established between the side chain of the noncharged Glu5.46 and the carbonyl moiety and the NH of the indole moiety, respectively. The indole moieties of the ligand and Trp6.48 establish a hydrophobic, aromatic interaction. This interaction stabilizes Trp6.48 in its vertical conformation, which is proposed to be an essential part of the so-called rotamer toggle switch (Shi et al., 2002) and is considered typical for the inactive state of a biogenic amine receptor. Thus, the binding mode of 2 described here differs significantly from that of the related compound 4, which has been described previously (Jójárt et al., 2008). These significant differences in binding mode with regard to the structurally highly related compounds 2 and 4 may be explained by different hH4R models. The model of Jójárt et al. (2008) is based on the crystal structure of bovine rhodopsin, whereas the hH4R model used within this study is based on the crystal structure of hβ2R. Another reason for the different findings may be the differences in simulation and docking protocol.
A comparison of the backbone of TMVII of the compound 2–hH4R complex with the backbone of TMVII of the ligand-free hH4R revealed a slight shift of TMVII away from TMVI (Fig. 3B). The indole moiety of the ligand switches into the pocket neighbored to Trp6.48 and thus induces a slight movement of Trp6.48 toward TMVII. As a consequence, TMVII is slightly shifted away from TMVI. Furthermore, the simulations revealed two small pockets in the R5 and R7 positions of the indole derivatives (Fig. 3C). These pockets can be filled with additional substituents in the corresponding positions. Docking of 11 into the binding pocket reveals that both chlorine atoms optimally fit into these additional pockets (Fig. 3D). Thus, the interactions between ligand and receptor are not disturbed. Further docking studies with compound 7 (data not shown) revealed that the space-filling methoxy moiety does not fit into the additional pocket. Because the benzimidazole derivatives can exist in two tautomeric states, both tautomers of derivative 20 were docked into the binding pocket of hH4R. The first tautomer (methyl moiety in R5 position) fit well into the binding pocket with its additional methyl moiety, and no loss in interaction was observed (Fig. 3E). In contrast, the second tautomer (methyl moiety in R6 position) did not fit into the binding pocket, because in the position of the methyl moiety, there was no additional space left (Fig. 3F). Further simulations showed (data not shown) that the second tautomer (methyl moiety in R6 position) shifted away from TMVI. As a consequence, the interaction with Glu5.46 was lost. It is possible that the indole moiety rotates approximately 180°. In this conformation, there is sufficient space for the additional methyl group, but again, because of the rotation of the indole, the interaction between the NH of the indole and Glu5.46 was lost.
Dependence of the Efficacy of Compounds 2–9 on Molar Volume.
To describe the efficacy of the indole derivatives with differences in substitution pattern in R5, we calculated the descriptors logP, polar surface area, molar refractivity, molar volume (van der Waals volume of 1 mol), refraction index, and polarizability. The calculations showed that the efficacy significantly depends on the molar volume of the ligand, but not on the other properties mentioned above. A correlation of the molar volume with the efficacy for compounds 2–9 led to a linear relationship (Fig. 4) described by the following equation: with the statistical parameters: n = 8, r2 = 0.782, F = 21.56, standard error of estimate = 0.085, and p < 0.005. Thus, the efficacy of compounds 2–9, with variations in the substitution pattern in R5 position, strongly depends on the volume of the substituent. All other properties did not show any influence on efficacy. However, the efficacy of compounds 10–16 with different substitution patterns in R7 does not correlate with the molar volume or any of the other properties mentioned above.
The extraordinarily high and Na+-insensitive constitutive activity of the hH4R (Schneider et al., 2009) suggests that inverse agonists and neutral antagonists may behave differently in clinical conditions. If high constitutive receptor activity leads to pathological condition, then inverse hH4R agonists may be more effective than neutral antagonists. On the other hand, neutral antagonists could be advantageous, because they may induce less up-regulation of binding sites compared with inverse agonists.
Therefore, it is very important for the development of hH4R antagonistic drugs that the structural requirements for neutral antagonism and inverse agonism are elucidated, enabling the synthesis of tailored compounds with the intended quality of action. To learn more about structure–activity relationships for hH4R neutral antagonists and inverse agonists we investigated a series of 25 hH4R antagonists described previously as H4R ligands (Venable et al., 2005). The compounds were characterized in the recently reported Sf9 insect cell-based test system, coexpressing hH4R with Gαi2 and Gβ1γ2 (Schneider et al., 2009). The expression level of Gαi2 in Sf9 cells is more than 100-fold higher than the hH4R expression level (Schneider et al., 2009). Thus, signal transduction of hH4R is not hampered by a limited Gαi2 availability. It has to be emphasized that we determined inverse agonistic activity at the point of GDP/GTP exchange, which is very proximal in the signal transduction cascade. Functional assays at more distal points, such as adenylyl cyclase or reporter gene assays, often suffer from nonlinear signal transduction, e.g., because of limited availability of distal signaling components. Such effects can bias determination of correct inverse agonistic activities.
Radioligand competition assays with [3H]HA revealed a good correlation of our data with previously reported results (Venable et al., 2005). It is noteworthy that among the 25 compounds investigated in our study three compounds (12%) were neutral antagonists, suggesting that neutral antagonism is rather exceptional. This is confirmed by a previously reported survey of 105 articles describing 380 antagonists acting on 73 GPCRs. Only 15% of these compounds acted as neutral antagonists (Kenakin, 2004). The similar percentage of inverse agonists potentially indicates a general mechanism of regulation of constitutive activity that most GPCRs may have in common. It is noteworthy that the neutral antagonists identified in our studies all show very low affinities with pKi values approximately 6.50 (compound 14) or lower (compounds 9 and 19). We do not have an explanation for these data. There is no indication in the literature that neutral antagonism goes along with low affinity. In fact, some low-affinity H4R ligands are actually highly efficacious inverse agonists (Deml et al., 2009).
All inverse agonists investigated in our study exhibited lower efficacies than the reference compound THIO, confirming that THIO is one of the most effective inverse hH4R agonists. However, as recently reported, THIO does not exert the maximum possible inverse agonistic effect (Schneider et al., 2009), and a quinazoline derivative with higher inverse agonistic efficacy than THIO has been described previously (Smits et al., 2008). This compound is closely related to the structures investigated in our study, suggesting that small structural changes can largely affect inverse agonistic efficacy. This was also observed within our compound series. For instance, introduction of Br in position R7 of the unsubstituted indole derivative 2 (efficacy = −0.40) eliminates inverse agonism and produces the neutral antagonist 14. A similar sensitivity of inverse agonistic efficacy for subtle structural changes was described previously, e.g., for α1AR antagonists (Rossier et al., 1999) or 5-HT1BR antagonists (Gaster et al., 1998).
Unexpectedly, we found a subset of compounds (filled circles in Fig. 1, B and D) with pEC50 values significantly lower than the pKi and/or pKb values. Such discrepancies are not unusual for inverse agonists (Table 3). The pKb values in our compound series were much closer to the pKi values than the pEC50 values. Because pKb and pKi values were determined in the presence of HA, but pEC50 values were determined in its absence, it is possible that some of the inverse agonists discriminate between the constitutively active receptor state and the active state stabilized by histamine. However, this phenomenon seems to be complex, because the pKi/pEC50 difference was observed only in some hH4R inverse agonists. Also the cimetidine/famotidine pair in Table 3 (Alewijnse et al., 1998) confirms that even in the same system the EC50/Ki ratio may show large variability. The literature examples shown in Table 3 for the cannabinoid CB2R (Ross et al., 1999) and the dopamine D2R (Roberts et al., 2004) show that systems with good correlation between EC50 and Ki values also exist. It is noteworthy that the difference in EC50/Ki ratios obtained with the neutral D2R antagonist [3H]spiperone and the D2R agonist [3H]-N-propylnorapomorphine suggest that the quality of action of the radioligand may result in different Ki values. Our results and the literature examples in Table 3 show that the two-state model of receptor activation, which describes only one active and one inactive receptor state, is not sufficient to describe the pharmacological properties of inverse agonists. Hence, the model must be extended to include different active states, dependent on whether the receptor is agonist-activated or constitutively active. Other data confirm that the two-state model of receptor activation has to be expanded into a more complex model. For example, the replacement of GTP by xanthosine 5′-triphosphate switches the effect of ICI 118551 [erythro-dl-1-(7-methylindan-4-yloxy)-3-isopropylaminobutan-2-ol] at the β2AR-Gsαs fusion protein expressed in Sf9 cells in the cAMP assay from inverse agonism to partial agonism (Seifert et al., 1999). Furthermore, certain β2AR agonists act as partial agonists or inverse agonists depending on whether the functional assay is performed in intact cells or cell membranes (Chidiac et al., 1994).
Molecular dynamic simulations revealed a stable binding mode for 2, different from the already published binding mode of the related compound 4 (Jójárt et al., 2008). However, the binding mode described in this work is similar to the recently published binding mode of clozapine at hH4R (Jongejan et al., 2008). The indole or the benzimidazole moiety of the compounds described here are embedded in a small pocket, parallel to the indole moiety of Trp6.48 in its inactive, more vertical conformation. A conformational change, the rotamer toggle switch of the highly conserved amino acids Trp6.48 and Phe6.52 is necessary for GPCR activation (Crocker et al., 2006; Strasser et al., 2008). Molecular dynamic simulations revealed that the hydrophobic interaction between the benzimidazole or indole moiety of the compounds stabilizes the inactive conformation of the Trp6.48. Because of this interaction between the ligand and Trp6.48, Trp6.48 cannot switch into its active, vertical conformation. Thus, the inverse agonism or antagonism of the analyzed compounds can be explained. The exchange of the hydrogen in R5 and/or R7 position in the indole derivatives into more space-filling substituents, such as chlorine, as given in compounds 4 and 11, leads to an increase in affinity, caused by an increase in contact area between ligand and receptor. If the substituent is too large, such as in compound 7, a decrease in affinity is observed. However, because no significant dependence between affinity and molar volume of compounds 2–9 could be observed further, and as yet unknown descriptors may be relevant. By contrast, the efficacy of compounds 2–9 depends significantly on the molar volume. With decreasing molar volume, corresponding to a decreasing volume of substituent in R5 position, a significant decrease in efficacy is observed (Fig. 4). Thus, small substituents in R5 position shift the equilibrium between inactive and active conformation toward the inactive state.
It should be noted that our study comprised a relatively small number of compounds. Important substitution patterns (e.g., benzimidazoles derivatives substituted in analogy to the indole derivatives) were, most unfortunately and despite substantial efforts, not available to us. Therefore, our study does not allow more detailed conclusions about the interactions at the binding site. Binding modes of structurally completely different H4R inverse agonists, e.g., of thioperamide or some 2,4-diaminopyrimidines (Sander et al., 2009), are not predictable from our results. In addition, establishing analogies between the binding mode of the JNJ compounds and other compounds, would lead to an extensive modeling study, which is beyond the scope of this article.
Nevertheless, to the best of our knowledge, our study is one of the first that systematically investigated structure–activity relationships of inverse GPCR agonists, using structurally closely related compounds. Our experiments showed that explaining the impact of structural alterations on the pharmacological properties of inverse agonists is much more difficult than expected, because structure–activity relationships could be established only for substituents in position R5 of the indole-derived compounds. Seemingly, the sole determination of pKb and pKi values is not sufficient to establish structure–activity relationships for inverse agonists. Similarly to previously reported results for 2,4-diaminopyrimidine-derived hH4R ligands (Sander et al., 2009), our results show that small changes in molecular structure lead to extensive efficacy and potency differences, caused by yet unpredictable major changes in binding mode and receptor conformation. This could be elucidated by cocrystallization of hH4R with structurally diverse inverse agonists. The recently published crystal structure of the β2 adrenoceptor, cocrystallized with the inverse agonist carazolol (Cherezov et al., 2007; Rasmussen et al., 2007), shows that this approach is possible.
We thank Mrs. Gertraud Wilberg and Mrs. Astrid Seefeld for excellent technical assistance and the reviewers for their helpful critique.
This work was supported by the Research Training Program (Graduiertenkolleg) “Medicinal Chemistry: Molecular Recognition–Ligand-Receptor Interactions” of the German Research Foundation (Deutsche Forschungsgemeinschaft) [Grant GRK 760]; and the European Cooperation in the Field of Scientific and Technical Research Action BM0806 (“Recent Advances in Histamine Receptor H4>R Research”), funded by the European Commission (Seventh European Framework Program).
Article, publication date, and citation information can be found at http://jpet.aspetjournals.org.
- human histamine H4 receptor
- G protein-coupled receptor
- thioperamide (N-cyclohexyl-4-(imidazol-4-yl)-1-piperidinecarbothioamide)
- ICI 118551
- transmembrane domain
- dimethyl sulfoxide
- [35S]guanosine 5′-O-(3-thio)triphosphate.
- Received January 16, 2010.
- Accepted May 17, 2010.
- Copyright © 2010 by The American Society for Pharmacology and Experimental Therapeutics