|Home | About | Journals | Submit | Contact Us | Français|
Vardenafil has higher affinity to phosphodiesterase-5 (PDE5) than sildenafil and lower administered dosage for the treatment of erectile dysfunction. However, the molecular basis for these differences is puzzling because two drugs have similar chemical structures. Reported here is a crystal structure of the fully active and nonmutated PDE5A1 catalytic domain in complex with vardenafil. The structure shows that the conformation of the H-loop in the PDE5A1-vardenafil complex is different from those of any known structures of the unliganded PDE5 and its complexes with the inhibitors. In addition, the molecular configuration of vardenafil differs from that of sildenafil when bound to PDE5. It is noteworthy that the binding of vardenafil causes loss of the divalent metal ions that have been observed in all the previously published PDE structures. The conformational variation of both PDE5 and the inhibitors provides structural insight into the different potencies of the drugs.
Cyclic nucleotide phosphodiesterases (PDEs) are key enzymes controlling cellular concentrations of the second messengers cAMP and cGMP (Mehats et al., 2002; Houslay and Adams, 2003; Goraya and Cooper, 2005; Bender and Beavo, 2006; Lugnier, 2006; Conti and Beavo, 2007; Omori and Kotera, 2007). The human genome encodes 21 PDE genes that are categorized into 11 families. Alternative mRNA splicing of the PDE genes produces approximately 100 iso-forms of PDE proteins that distribute in various cellular compartments and control myriad physiological processes. PDE molecules contain a variable regulatory domain and a conserved catalytic domain but show distinct substrate specificity and inhibitor selectivity. Family-selective PDE inhibitors have been widely studied as therapeutic agents for treatment of various human diseases, including cardiotonics, vasodilators, smooth muscle relaxants, antidepressants, antithrombotics, antiasthmatics, and agents for improving learning and memory (Truss et al., 2001; Rotella, 2002; Schrör, 2002; Castro et al., 2005; Houslay et al., 2005; Lipworth, 2005; Blokland et al., 2006; Menniti et al., 2006).
The most successful examples of this class of drugs are the PDE5 inhibitors (Fig. 1) sildenafil (Viagra), vardenafil (Levitra), and tadalafil (Cialis), which have been used for treatment of male erectile dysfunction (Rotella et al., 2002). Sildenafil (Revatio) has also been approved for treatment of pulmonary hypertension (Galié et al., 2005). Korean authorities have recently approved udenafil (Fig. 1) for treatment of male erectile dysfunction (Salem et al., 2006) . Although these four PDE5 inhibitors have been successfully approved as the drugs for treatment of human diseases, the enthusiasm for development of novel PDE5 inhibitors continues. PDE5 inhibitors have been shown to have potential for other medical applications, including improvement of memory and treatment of cancer and heart disease (Blokland et al., 2006; Salem et al., 2006; Stehlik and Movsesian, 2006; Supuran et al., 2006; Padma-Nathan et al., 2007; Palmer et al., 2007; Zhu and Strada, 2007; Sandner et al., 2008). Much attention has been focused on the recent development of the second generation of PDE5 inhibitors that have the same or different scaffolds from the current drugs but different pharmaco-kinetic profiles (Palmer et al., 2007).
Sildenafil, vardenafil, and udenafil have similar chemical formulae (Fig. 1) and possess similar key pharmacophores that provide for their function. These inhibitors also have the same target and interact with many of the same residues at the active site of PDE5, as shown by the crystal structures of the isolated PDE5 catalytic domain in complex with sildenafil and vardenafil (Sung et al., 2003; Card et al., 2004; Huai et al., 2004; Zhang et al., 2004; Wang et al., 2006). Although the head-to-head comparison is still lacking, the pharmaco-kinetic and pharmacodynamic analyses showed that these PDE5 inhibitors have similar efficacy and tolerance but exhibit some functional differences both in vitro and in vivo (Briganti et al., 2005; Shabsigh et al., 2006; Supuran et al., 2006; Wright, 2006; Doggrell, 2007; Mehrotra et al., 2007). For example, vardenafil shows 10- to 40-fold tighter binding with PDE5 than sildenafil and has an area under the curve of 74.5 μg · h/liter at a 20-mg dosage compared with 1965 for sildenafil at a 100-mg dosage (Shabsigh et al., 2006; Mehrotra et al., 2007).
However, the structural basis for the different potencies of these inhibitors is still puzzling. The early studies on the crystal structures of PDE5 in complex with sildenafil and vardenafil by two groups showed inconsistent results. Vardenafil and sildenafil have the same extended configuration in the crystal structures reported by Sung et al. (2003), in contrast to the folded configuration of both inhibitors in the report by Zhang et al. (2004). Because the PDE5A catalytic domain used by Sung et al. (2003) is basically inactive and the structure reported by Zhang et al. (2004) contains a chimeric replacement of the PDE5A H-loop with the PDE4D H-loop, the biologically relevant conformation of these drugs has remained a question.
To address this question, the crystal structure of the fully active and nonmutated catalytic domain of PDE5A1 in complex with vardenafil has been determined and compared with the previously published cocrystal structures of the enzyme with sildenafil and vardenafil. The structural comparison shows dramatic differences between the vardenafil and sildenafil complexes in both PDE5 protein conformation and the inhibitor configuration. These differences are likely to contribute to the different properties of these drugs.
The cDNA of the catalytic domain of human PDE5A1 was generated by site-directed mutagenesis of the gene of bovine PDE5A, as described previously (Wang et al., 2006). The coding region for amino acids 535 to 860 of PDE5A1 was amplified by PCR and subcloned into the expression vector pET15b. The resultant plasmid pET-PDE5A1 was transferred into Escherichia coli strain BL21-CodonPlus (Stratagene, La Jolla, CA) for overexpression. The E. coli cell carrying pET-PDE5A1 was grown in Luria-Bertani medium at 37°C to absorption A600 = 0.7 and then 0.1 mM isopropyl β-d-thiogalactopyranoside was added for further growth at 15°C overnight. Recombinant PDE5A1 was passed through the nickelnitrilotriacetic acid affinity column (QIAGEN, Valencia, CA), subjected to thrombin cleavage to remove the His tag, and further purified by Q-Sepharose and Sephacryl S300 column chromatography (GE Healthcare, Chalfont St. Giles, Buckinghamshire, UK). A typical purification yielded over 10 mg of PDE5A1 with a purity >95% from a 2-liter cell culture.
The cocrystal of PDE5A1 (535–860) with vardenafil was grown by vapor diffusion. The complex of PDE5A1-vardenafil was prepared by mixing 1 mM vardenafil with 15 mg/ml PDE5A1 at 4°C overnight. The protein drop was set up by mixing 2 μl of protein solution with 2 μl of well buffer and crystallized against a well buffer of 12% polyethylene glycol 3350, 15% glycerol, and 0.1 M sodium acetate pH 4.6 at 25°C. The PDE5A1-vardenafil crystals have the space group P212121 with cell dimensions of a = 68.9, b = 87.8, and c = 138.5 Å (Table 1). Diffraction data were collected on beamline X29 at Brookhaven National Laboratory and processed by program HKL (Otwinowski and Minor, 1997).
The structure of the PDE5A1-vardenafil cocrystal was solved by molecular replacement program AMoRe (Navaza and Saludjian, 1997), using the PDE5A1-IBMX structure without the H-loop as the initial model. The atomic model was built with the use of the program O (Jones et al., 1991) against the electron density map that was improved by the density modification package of CCP4. The structure was refined by CNS (Crystallography and NMR System) (Table 1; Brünger et al., 1998). The atomic coordinates and structural factors have been deposited into the Protein Data Bank with accession code 3B2R.
The enzyme of the PDE5A1 catalytic domain used in these studies was fully active and exhibited kinetic properties (kcat, Km) similar to those for the full-length PDE5A1 (Wang et al., 2006). The structure of the PDE5A1 catalytic domain (residues 535–860) in complex with vardenafil consists of 15 α-helices (Fig. 2). Most of the residues in the PDE5A1-vardenafil cocrystal had solid electron density and were traced without ambiguity. Residues 660 to 672 and 792 to 806, which are parts of the H- and M-loops, lacked electron density and were disordered. The superimposition of PDE5A1-vardenafil over other previously determined PDE5A1 structures (Huai et al., 2004; Wang et al., 2006) yielded root-mean-square deviations of 0.49, 0.54, 0.51, and 0.47 Å, respectively for Cα atoms of 270 comparable residues (536–657, 686–787, and 813–859) of the unliganded PDE5A1 and its complexes with IBMX, icarisid II, and sildenafil, indicating the overall similarity among the PDE5 structures. However, the PDE5A1-vardenafil structure shows dramatic conformation differences in the H-and M-loops from the known PDE5 structures.
The H-loop of PDE5 was previously shown to have four different conformations depending on the liganded state of the protein: 1) a coil conformation in the unliganded state, 2) two short α-helices (H8 and H9) at residues 664 to 667 and 672 to 676 in the IBMX complex, 3) a 310 helix in the sildenafil complex, and 4) two short β-strands in the icarisid II complex (Wang et al., 2006). In addition, these conformation changes of the H-loop upon the inhibitor binding are coupled with the dramatic positional movements, up to 7, 24, and 35 Å in these three complexes, respectively. The position and conformation of the H-loop in the PDE5A1-vardenafil structure also differs significantly from those of the known PDE5 structures. First, helix H9 in the PDE5A1-vardenafil cocrystal contains residues Ser675 to Ile680, compared with the sequence of 671 to 675 in the PDE5A1-IBMX complex; the latter composition of the H-loop resembles the residues of helix H9 in other PDE families (Ke and Wang, 2007). Second, the H-loop in the PDE5A1-vardenafil complex shows a positional shift of as much as 20 Å from that in the unliganded form. Third, the N-terminal residues 680 to 685 of helix H10 that contains residues 680 to 693 in all the early structures of PDE5 and other PDE families are in a coil conformation in the PDE5A1-vardenafil structure. Finally, residues 792 to 808 of the M-loop in the PDE5A1-vardenafil structure are disordered. This disorder is similar to features of this region found in the unliganded and IBMX-bound PDE5A1 structures (Wang et al., 2006) but is in contrast to the ordered conformation of the M-loop in the cocrystal structures of PDE5-sildenafil and PDE5-icarisid II. Because the M-loop contains a coiled fragment around Leu804 that contacts the inhibitors and most likely the cGMP substrate (Fig. 2), its disorder in the PDE5A1-vardenafil structure is likely to have some implication for both inhibitor and substrate binding. However, this possibility will require further study.
The most surprising feature of the PDE5A1-vardenafil structure is the absence of divalent metals at the active site (Fig. 2). This is in contrast to the absolute conservation of the binding of two divalent metal ions at the active sites of all early reported structures of PDE5 and other PDE families (Wang et al., 2006; Ke and Wang, 2007), even in the presence of the metal-chelating agent EDTA during the protein purification of the PDE4B2B catalytic domain (Xu et al., 2000). Because the divalent metals and their binding residues are not involved in crystallographic lattice contacts, the loss of the divalent metals in the PDE5-vardenafil complex is unlikely to be an artifact of the crystal packing. Rather, the loss of the metal ions in the PDE5-vardenafil structure is apparently due to the influence of the conformational changes in the H-loop. A careful examination shows that two of the zinc-binding residues (His617 and Asp764) are well superimposed with those in the other PDE5 structures (Fig. 2D), as shown by small positional differences of 0.15 and 0.19 Å for their Cα atoms between the structures of PDE5A1-vardenafil and PDE5A1-sildenafil. However, two other important metal-binding residues (His653 and Asp654) in the PDE5A1-vardenafil complex show shifts of 0.76 and 0.80 Å from those in the PDE5A1-sildenafil complex although their conformations are retained; the magnitudes of these shifts are almost twice the overall root-mean-square deviation of 0.47 Å for all the atoms in the structures. In addition, the positioning of His684 in the PDE5-vardenafil complex is completely different from that in the PDE5-sildenafil structure, and its imidazole ring is now located at the site normally occupied by the second metal ion or magnesium (Fig. 2D). Thus, the positional and conformational changes of Asp654 and His684 apparently act to eliminate binding of both divalent cations. Because Asp654 and His684 are located, respectively, at the N and C termini of the H-loop, the conformational change of the H-loop upon vardenafil binding seems to be the driving force that causes loss of the metals. This suggests that loss of catalytic activity in the presence of vardenafil is due to two factors: 1) direct competition between vardenafil and cGMP for access to the catalytic site and 2) vardenafil-induced loss of divalent cations from the catalytic site.
To study whether the inactive PDE5-vardenafil complex can regain the catalytic activity, the PDE5A1 catalytic domain (residues 535–860) was mixed with 1.5 mM vardenafil for 4 h and then passed through a Sephacryl S300 gel filtration column in a plain running buffer of 20 mM Tris-HCl, pH 7.5, 50 mM NaCl, and 1 mM 2-mercaptoethanol without inhibitor vardenafil. The specific activities of the native PDE5, the PDE5-vardenafil complex, and the fraction eluted from the S300 column were measured at five repeats by using a method previously described (Wang et al., 2006) and the assay buffer of 20 mM Tris-HCl, pH 7.5, 1.5 mM dithiothreitol, 10 mM MgCl2, and 0.2 μM cGMP. They were 39.8 ± 0.6 nmol/min/mg for the native PDE5 catalytic domain, 0.09 ± 0.02 for the protein in complex with 1.5 mM vardenafil, and 2.0 ± 0.4 for the protein after passed the S300 column. Addition of 10 and 100 nM zinc to the assay buffer increased the specific activity for the protein eluted from the S300 column by approximately 2-fold (3.9 ± 1.1 and 4.4 ± 1.0 nmol/min/mg). The second time passing through the S300 column did not further increase the activity. These experiments suggest that the limited catalytic activity can be regained from the inactive PDE5-vardenafil complex by passing the gel filtration column. The small percentage (10%) of activity recovery implies the improper elution conditions or the trap of vardenafil in the closed active site.
Vardenafil directly competes with cGMP for access to the catalytic pocket of PDE5A1, as does sildenafil (Figs. 2 and and3).3). The binding involves three hydrogen bonds formed between O6 and N1 of imidazotriazinone of vardenafil and Nε2 and Oε1 of Gln817 of PDE5A1 and between a sulfonamide oxygen of vardenafil and the backbone nitrogen of Cys677 of PDE5A1. In addition, two water molecules bind to O6 and N8 of the imidazotriazi-none. For hydrophobic interactions, the propyl-imidazotriazinone of vardenafil stacks against Phe820 of PDE5A1 and also contacts residues Tyr612, His613, Ile680, Leu765, Ala767, Ile768, Leu782, Phe786, and Gln817. The ethoxy group orients to a hydrophobic pocket and interacts via van der Waals forces with Ala779, Val782, Ile813, and Gln817. The phenyl group forms hydrophobic interactions with Tyr676, Met816, Gln817, and Phe820. The ether oxygen of the ethoxyphenyl group has a distance of 3.18 Å to the amide oxygen of the side chain of Gln817. This distance is an unfavorable interaction because both oxygen atoms have no proton for formation of a hydrogen bond and also because Gln817 is unlikely to switch its side-chain orientation because of its pre-existing hydrogen bond with Gln775. The sulfonamide group (-SO2N) of vardenafil contacts Tyr676, Cys677, and Ile680 of the H-loop, in addition to the stack against Phe820. The ethylpiperazine group orients to the surface of the binding pocket and interacts with Met816, Gly819, and Phe820.
Vardenafil binding to PDE5 shares a number of similarities with the binding characteristics of sildenafil, including similar location of their ethoxyphenyl and imidazotriazinone/ pyrazolopyrimidinone groups, the same stacking against Phe820, and the hydrogen bonds with Glu817 (Fig. 3). In addition, both inhibitors are buried in the catalytic pocket. The solvent-accessible area of the bound vardenafil is only 8%, which compares well with 9.4% of the bound sildenafil (Wang et al., 2006). However, the bound vardenafil shows molecular configuration and interactions different from those of the bound sildenafil (Fig. 3). The key difference is the orientation of the piperazine portion of two drugs. The ethylpiperazine of vardenafil orients to the surface of the binding pocket and is extended to interact with residues of Tyr676 to Ile680. In comparison, the methylpiperazine of sildenafil folds back to its molecular entity and interacts with residues Asn662 to Ile665 of the H-loop, but not Tyr676-Ile680, via van der Waals’ interactions.
Vardenafil possesses a chemical structure very similar to sildenafil, but is 10- to 40-fold more potent than sildenafil for PDE5 inhibition and a smaller clinically administered dosage for treatment of erectile dysfunction (Saenz de Tejada et al., 2001; Corbin et al., 2004; Setter et al., 2005; Supuran et al., 2006; Mehrotra et al., 2007). Thus, the structural basis for their different biochemical and physiological properties has been a puzzle. The results of the early studies on the crystal structures of PDE5A isolated catalytic domain in complex with sildenafil and vardenafil showed the similar binding mode of these inhibitors but did not entirely agree (Sung et al., 2003; Zhang et al., 2004). Vardenafil and sildenafil had the same extended configuration in the structures by Sung et al. (2003), in contrast to the folded configuration of both inhibitors in the report by Zhang et al. (2004) (Fig. 3). The different orientations of the piperazine tails of the inhibitors seem to result from the rotation of the single C-S bond. Although the energy barrier for the single bond rotation is minimal in theory, it is rare that the same inhibitors adopt different conformations when bound to their receptors. Because the PDE5A catalytic domain used by Sung et al. (2003) was basically inactive and the structure reported by Zhang et al. (2004) contains a chimeric replacement of the PDE5A H-loop with the PDE4D H-loop, the biologically relevant configurations of these inhibitors is in question. The configuration of vardenafil in our structure is similar to that in the structure reported by Sung et al. (2003), whereas our sildenafil configuration is similar to that reported by Zhang et al. (2004). Because our PDE5 protein is fully active and contains all native PDE5 residues, the configuration difference between vardenafil and sildenafil is likely to be relevant to the biological and physiological properties of these inhibitors.
Our previous studies have already shown that the H-loop of PDE5A is highly flexible and can adopt different conformations and positional locations upon binding of the inhibitors (Wang et al., 2006). This study adds significant new structural information pertaining to the interactions between PDE5 and vardenafil. Although it is not clear whether the individual conformation of the H-loop can be exploited for design of new PDE5 inhibitors, the flexibility of the H-loop is likely to be an important allosteric mechanism that affects substrate and inhibitor binding. The isolated PDE5A1 catalytic domain showed similar binding affinity with vardenafil and sildenafil, and the vardenafil affinity is significantly boosted by the involvement of GAF-B of PDE5A1 regulatory domain (Blount et al., 2006). This implies that the unique features associated with the interaction of each of these inhibitors with the PDE5 catalytic domain are likely to play a major role in determining the influence of the regulatory domain on inhibitor affinity. A full understanding of the molecular effects of these inhibitors will require further structural study within the context of the PDE5 holoenzyme.
It is interesting to note that the PDE4-selective inhibitor rolipram caused relocalization of the full-length PDE4A4 and was suggested to trigger a conformational change on a loop interacting with Mg2+ (Terry et al., 2003). Besides, the binding of cAMP to the GAF domain of trypanosome PDEB1 (Laxman et al., 2005) and cGMP to the GAF domain of human PDE5 (Zoraghi et al., 2005) induced the allosteric conformational changes of the enzymes. These observations, together with the structural study done here, suggest that conformational changes promoted by the binding of inhibitors and substrates in the certain PDE families are essential for the regulations such as phosphorylation of PDE4 and cGMP/cAMP binding to GAF domains of PDEs. The H-loop of PDE5 likely serves as a key mediator of the allosteric regulation of enzymatic activity.
We thank beamline X29 at NSLS for collection of diffraction data.
This work was supported by National Institutes of Health grants GM59791 (to H.K.) and DK58277 (to S.F.).