Atomic Structure of PDE4: Insights into Phosphodiesterase Mechanism and Specificity

See allHide authors and affiliations

Science  09 Jun 2000:
Vol. 288, Issue 5472, pp. 1822-1825
DOI: 10.1126/science.288.5472.1822


Cyclic nucleotides are second messengers that are essential in vision, muscle contraction, neurotransmission, exocytosis, cell growth, and differentiation. These molecules are degraded by a family of enzymes known as phosphodiesterases, which serve a critical function by regulating the intracellular concentration of cyclic nucleotides. We have determined the three-dimensional structure of the catalytic domain of phosphodiesterase 4B2B to 1.77 angstrom resolution. The active site has been identified and contains a cluster of two metal atoms. The structure suggests the mechanism of action and basis for specificity and will provide a framework for structure-assisted drug design for members of the phosphodiesterase family.

Cyclic nucleotides are intracellular second messengers that play a key role in many physiological processes. The levels of these nucleotides are tightly regulated at the point of synthesis by receptor-linked enzymes (such as adenylyl or guanylyl cyclase) and at the point of degradation by a family of enzymes known as phosphodiesterases (PDEs). The PDEs constitute a large, divergent family whose members catalyze the hydrolysis of adenosine 3′,5′-monophosphate (cAMP) and guanosine 3′,5′-monophosphate (cGMP), resulting in the formation of the 5′-nucleotide metabolites (AMP and GMP). Ten PDE gene families have been identified, and more than 50 mRNA splicing isoforms have been detected in various human tissues [reviewed in (1–5)] (6). These isoforms are differentially expressed and regulated in different cell types, and they appear to modulate the location, substrate selectivity, kinetics, and response to activators or inhibitors of the enzyme (7–9). PDEs are clinical targets for a range of biological disorders such as retinal degeneration, congestive heart failure, depression, asthma, erectile dysfunction, and inflammation (10,11). Knowledge of the structure of a PDE will allow for the rational design of therapeutics to treat many of these disorders.

The structure of the catalytic domain (residues 152 to 528) of human phosphodiesterase 4B2B (PDE4B2B) was determined by multiwavelength anomalous dispersion in combination with multiple isomorphic replacement (Tables 1 and2) (12). The catalytic domain of PDE folds into a novel compact structure composed of 17 α helices, which form three subdomains (Fig. 1). The first subdomain forms a bundle of four α helices (H3, H5, H6, and H7), with two short interconnecting α helices (H2 and H4). The residues of H0 and H1 are poorly conserved across the PDE family and may not be an essential part of the catalytic domain (13). The second subdomain is formed by two short α helices (H8 and H9) that sit perpendicular to a pair of long antiparallel α helices (H10 and H11). The third subdomain is composed of five α helices (H12 to H16) and an extended loop that forms a β hairpin between H12 and H13. Helix 15 (H15a/H15b) is kinked by partially conserved Pro453 and is probably a continuous helix in the absence of proline. In one of the two molecules in the asymmetric unit, residues 496 to 508 form an additional helix (H17), which packs against a neighboring molecule in the crystal. We believe that the position of this helix is an artifact of crystallization and is not a relevant dimer contact.

Figure 1

Ribbon diagram of the secondary structure of the catalytic domain (residues 152 to 489) of PDE4B2B. ME1 is shown as a silver sphere (ME2 is behind H13 in this orientation). The NH2-terminal subdomain of the molecule (residues 152 to 274) is colored cyan, the middle subdomain green (residues 275 to 347), and the COOH-terminal subdomain is colored yellow (residues 348 to 489).

Table 1

X-ray data and phasing statistics. Data sets for SmCl3 and GdCl3 were collected on beamline 17ID at the Advanced Photon Source in Argonne National Laboratory (APS); all others were collected on beamline X12C at the National Sychrotron Light Source in Brookhaven National Laboratory (NSLS). All data were processed with the HKL package (36).

View this table:
Table 2

Refinement statistics.

View this table:

A deep pocket in the center of the COOH-terminal subdomain extends to the junction of the three subdomains. Of the 21 residues absolutely conserved across all 10 PDE families, 20 lie near this pocket and 12 lie within it (Fig. 2) (13). The pocket has a volume of 440 Å3 and is the only pocket large enough to accommodate a molecule of the cAMP substrate (232 Å3) or AMP product (256 Å3). The pocket is lined with conserved hydrophobic and negatively charged residues. At the bottom of the pocket are several metal atoms, which bind to residues that are completely conserved in all PDE family members.

Figure 2

Surface representation of the PDE4 molecule. The orientation of this view is an 80° counterclockwise rotation around the vertical axis of that shown in Fig. 1. The surface colors denote the level of sequence conservation across the 10 PDE families. Completely conserved residues are shown in red, partially conserved residues in white, and highly variable residues in blue. ME1 and ME2 are represented as yellow spheres. A view of the opposite side of the molecule is shown in the inset at the lower left.

One of these metal atoms (ME1) coordinates residues in H6 in subdomain 1, the loop between H7 and H8 (linking subdomains 1 and 2), and H13 in subdomain 3. ME1 may stabilize the structure by acting as a linchpin that locks the three subdomains of the protein together. It may also play a catalytic role because of its proximity to the active site. ME1 has a high occupancy (as indicated by the intensity of the peaks in electron density maps) and is located deepest in the pocket (Fig. 3). It was present despite the presence of chelators and the absence of additional exogenous divalent metal ions during the purification or crystallization of the protein. ME1 interacts with the Nɛ atoms of His238 (2.0 Å) and His274 (2.1 Å) and the Oδ atoms of Asp275 (2.3 Å) and Asp392 (2.2 Å) in a trigonal bipyramid configuration where the Asp residues occupy the axial positions (Fig. 3). The remaining equatorial position is occupied by a solvent atom (1.9 Å), which also acts as a bridging ligand to a second metal ion. Although the identity of the metal ions cannot be absolutely determined from our structure, the observed geometry of the metal coordinating ligands, the anomalous x-ray diffraction behavior, the existing biochemical evidence (14–17), and the known high affinity of PDE4 for zinc (18) all suggest that ME1 is a Zn2+ ion.

Figure 3

Overview of the active site. A 2F obsF calc electron density map of the metal binding site calculated from 18 to 1.77 Å native data is shown in brown at a level of 2.5σ. An anomalous difference map calculated from 20 to 3.3 Å using refined phases and data from an Er-soaked crystal collected at λ = 1.48 Å is shown in gray at a level of 5σ. ME1 and ME2 are shown in magenta, water molecules are red, and the protein carbon atoms are colored as in Fig. 1. Metal ion interactions are shown in white. Several observations indicate that a third metal ion visible in the electron density is a crystallization artifact arising from an opportunistic arsenate ion (As*) present in the crystallization buffer. This ion shows variable occupancy and makes only three water-mediated contacts with the protein. Anomalous difference density at the site of the third metal ion (shown in cyan) was observed in data collected at λ = 1.0 Å but not in data collected at λ = 1.4 Å, which is consistent with an arsenate ion. The position of the arsenate ion, 2.6 Å above the water that bridges the two metal atoms, suggests a possible binding site for phosphate, or a phosphate-containing species in the active site.

The second metal ion (ME2) varied in occupancy between crystals when no additional metal ions were added to the solution. However, it was always highly occupied when exogenous metals (Mn, Mg, or Zn) were added to the protein. ME2 makes only one direct interaction with the enzyme to the second Oδ of Asp275 (2.4 Å). It further interacts with the water molecule (2.4 Å) that binds to the first metal ion. Four additional water molecules complete the distorted octahedral geometry of the second metal site. These water molecules in turn form hydrogen bonds with residues His278, Glu304, His307, and Ser345 of the enzyme. The preferred in vivo ion cannot be determined from this study. Others have shown that many ions are capable of supporting enzymatic activity (19, 20); however, it is likely that either Mn2+ or Mg2+ is the relevant physiological ion (21). ME1 and ME2 coordinate several of the residues found in the two highly conserved PDE sequence motifs His-Asn-X-X-His (residues 234 to 238) and His-Asp-X-X-His (residues 274 to 278), each of which was previously considered to form a separate metal binding site (22).

The two metal ions are ∼3.9 Å apart and constitute a binuclear motif similar to that found in other enzymes that catalyze the hydrolysis or formation of amide or phosphate bonds. The motif in PDE is particularly similar to that found in the active site of 3′,5′-exonuclease of DNA polymerase I Klenow fragment, which contains Zn2+ and Mg2+ ions with similar coordination spheres (23). Little else about the active site of PDE4 resembles these enzymes.

A model of cAMP docked in the PDE4B2B active site (Fig. 4) shows that a water molecule coordinated to one or both metal ions could act as the nucleophile in the hydrolysis reaction. The metal ions would also serve to stabilize the transition state. The strictly conserved His234 is positioned to accomplish the protonation of the O3′ leaving group. The position of His234 is maintained by a hydrogen bond between its Nδ and a carboxylate oxygen of Glu413, which is itself stabilized by hydrogen bonds between the other side-chain oxygen and two backbone amides (Tyr233 and His234). The ring of His234 is stacked against the ring of His278 and with the edge of the aromatic ring of Tyr233. These stacking and hydrogen bond interactions to three other conserved residues further stabilize the position of His234 and probably alter the pK a of the Nɛ so it more closely matches that of the O3′ phosphate ester. The enzyme appears to have committed four amino acids to donate a proton, which speaks to the importance of this protonation step in the evolution of this enzyme to catalyze cAMP hydrolysis.

Figure 4

Model of cAMP bound to PDE4. A molecular docking procedure was used to fit cAMP into the proposed active site (34). The preferred model is shown in which cAMP adopts the anti conformation with the adenine base inserted into a lipophilic pocket formed by Leu393, Pro396, Ile410, Phe414, and Phe446. The cyclic phosphate group binds to ME1 and ME2, replacing the observed arsenate ion shown in Fig. 3. The 1-N and 6-NH2 groups form hydrogen bonds with the side chain of Gln443, while the 7-N position forms a more distorted hydrogen bond with Asn395. The ribose ring binds loosely against Met347 and Leu393, with a hydrogen bond between His234 and the O3′ oxygen, but with no obvious interaction to the O2′, O4′, and O5′ atoms. Consistent with our model, an experimental synthetic cAMP analog study found that PDE4 makes important interactions with the 1-N, 6-NH2, and 7-N positions, but not with the 2′-OH (35).

The proposed binding mode of cAMP (Fig. 4) also offers insight into cAMP and cGMP specificity within the PDE superfamily. Compared with adenine, guanine has a 1-NH, a 2-NH2, and a 6-CO group on the purine base. The hydrogen-bonding character at the 1- and 6- positions is thus essentially reversed. Gln443, which “reads” the 1- and 6- position groups in the anti conformation model, is conserved in all 10 PDE gene families. However, this glutamine could accommodate the 1- and 6- position groups in cGMP by rotating its amide group by 180°. The reversible hydrogen-bonding character of glutamine should thus permit hydrolysis of both cAMP and cGMP, as occurs in dual-specificity PDEs. In PDE4, Tyr403forms a hydrogen bond with Gln443 and provides additional steric constraints that may be responsible for the cAMP specificity. Residues proximal to Gln443 in the other PDEs may likewise influence its orientation and optimize the shape of the pocket to favor binding of specific nucleotides.

The PDE4 structure further clarifies the functions of many other residues that were characterized in the PDE family. For example, mutations of residues in rat PDE3 corresponding to PDE4 His307 and Thr345 completely diminished the enzyme activity (24). His307 and Thr345 are in the cAMP binding pocket and serve a critical function by forming water-mediated interactions to the second metal ion, positioning it for efficient catalysis. Other studies have implicated several residues on helix 14 in cAMP-cGMP specificity (25). One of these residues (Thr407) makes direct contacts with the nucleotide base in our docked substrate model, whereas the other residue (Tyr403) appears to play a key role in orienting Gln443.

Previous studies have shown that rolipram, an efficient inhibitor of PDE4, is a competitive inhibitor (26) and binds with a stoichiometry of 1:1 within the catalytic domain (27). Well-characterized mutations that affect rolipram binding (28) are primarily located in the binding pocket, and many of these residues are predicted to be in contact with the cAMP substrate in our model. It is likely that many of the known PDE inhibitors bind within the active-site pocket. The PDE4B2B crystal structure, as a representative of this therapeutically important class of enzymes, will assist the drug discovery process by providing binding and specificity information based on the crystal structure.

  • * Present address: Molecular Statistics and Bioinformatics Section, National Cancer Institute, Bethesda, MD 20892, USA.


View Abstract

Navigate This Article