Purple acid phosphatases has been identified as an attractive target for the development of chemotherapeutics to treat osteoporosis (16,21,22). The purpose of this study was to identify new inhibitor leads by screening a fragment library of compounds – the philosophy behind this approach is to discover new leads (i.e., small molecules) that can then be elaborated to generate potent and specific PAP inhibitors. Although the desired target for such inhibitors is human PAP, the enzymes from pig or red kidney bean (which have similar active sites) were used here for initial testing because of their ready availability.
A number of inhibitors of PAPs have been discovered but most of these do not possess drug-like properties and so are not suitable for further modifications that may convert them into chemotherapeutic agents. We therefore embarked upon on a FBS approach using a library of small molecules that possess drug-like properties to discover novel compounds as inhibitor leads of PAP. The initial screening was based on a plate assay where all 500 compounds were tested for inhibitory activity at a concentration of 1 mm against pig PAP. Three compounds, CC27209, CC24201 and MO07123 (Table 1; Figure S1), inhibit pig PAP by >70% at that concentration. Subsequently, the modes of inhibition and magnitudes of inhibition constants were determined (Table 2). The three compounds are all competitive inhibitors of pig PAP with Ki values in the 30–60 μm range. The three compounds also inhibit rkbPAP in a similar manner (Table 2; Figure S1), an observation that suggests that despite some structural variations between the two enzymes (discussed in detail later), the overall interactions between inhibitor and enzyme are largely conserved. This interpretation is also in agreement with previous studies using phosphonate derivatives as inhibitors for pig and rkbPAP – in each case, the observed modes and magnitudes of inhibition were similar in the two enzymes (32,33).
Table 2. Kinetic data for Maybridge compounds against reduced pig purple acid phosphatases (PAP) and red kidney bean PAP at pH 4.9
| ||Pig PAP||Red kidney bean PAP|
| K ic (μm)|| K iuc (μm)||Ligand efficiencya (competitive)|| K ic (μm)|| K iuc (μm)||Ligand efficiencya (competitive)|
|CC24201||53 ± 19||223 ± 181||0.39||37 ± 16||NSE||0.40|
|CC27209||59 ± 21||NSE||0.44||43 ± 14||359 ± 336||0.46|
|MO07123||42 ± 14||NSE||0.40||33 ± 10||NSE||0.41|
Structure and sequence comparisons of plant and animal PAPs
An obvious caveat in the preceding section is the use of pig and rkbPAP instead of the human enzyme. As pointed out earlier, the rationale for this approach is that the sufficient amounts of pure human PAP required for screening a large number of compounds cannot be obtained easily (48). Furthermore, as it is an aim of this study (as a necessary step towards visualizing enzyme–inhibitor interactions) to obtain crystal structures of a PAP with a bound fragment, it is not practical to use human PAP. In contrast, pig and rkbPAP can be obtained easily in large quantities, and in particular, plant PAPs have been shown to crystallize readily, thus facilitating the screening and visualizing process (38,49). Although pig PAP is available in large quantities and crystallization using previously identified conditions (9) has been successful, it has not been possible to obtain crystals of this enzyme where access to the active site is not blocked by a phosphate anion, and attempts to soak the fragments into pre-grown pig PAP crystals to replace the phosphate anion have been unsuccessful. In an attempt to further substantiate the use of both pig and rkbPAP as valid models for human PAP, we carried out a detailed sequence and structure comparison, focusing on the vicinity of the active site of these three enzymes. A partial structure-based sequence alignment of plant, human and pig PAP shows that the metal-binding ligands, as expected from previous studies (9,15,50,51), are completely conserved across the three species and that the amino acids that make up the surface of the active site are identical for the human and pig enzymes, with only three differences in rkbPAP (Figure 1). The two differences closest to the metal centre occur at positions R170 and Y365 (numbering according to the rkbPAP sequence), both of which are occupied by phenylalanine side chains in the animal enzymes (i.e., F244 and F56 in pig PAP). The most significant difference between the animal and plant enzymes lies in the deletion of a short loop in the sequence of rkbPAP relative to the animal PAPs (Figure 1). In animal PAPs, this region of the polypeptide chain forms a surface-exposed loop that consists of 21 amino acids (i.e., N144-X145–163-T164 in pig PAP). For the human enzyme, this loop has been shown to be susceptible to proteolytic cleavage, resulting in an increase in activity of up to tenfold (including increased efficiency towards ATP as a substrate depending on the protease used for cleavage) (15,47,52). In rkbPAP, the repression loop is replaced by a single amino acid R258, which comes from the adjoining subunit of the dimer.
Figure 1. Structure-based sequence alignment for residues in the active site region of plant and animal purple acid phosphatases (PAPs). Residues that have a red arrow coordinate directly to the metal ions. Residues that are located on the surface of the active site are highlighted in purple. The residues in the green box form an extended loop in human and pig PAPs. When the peptide bond residues A161 and R162 are proteolytically cleaved (green arrow) in human PAP, the enzyme is ten times more active. In plant PAP, this loop is missing, but R258 from the other subunit of the dimer reaches across to occupy this space.
Download figure to PowerPoint
The residues that contribute to the exposed surface of the active site and are located within the repression loop in the animal enzymes are N142, S143, D144 and Q149 (numbering according to the human PAP sequence). Thus, the active sites of the two animal enzymes are virtually indistinguishable. When comparing the plant and animal enzymes, the positions of the metals and the residues that are close to the metal centres are conserved, but some differences do exist at the periphery. It is therefore postulated that because of the small size of the fragments and similarity of the active site environment surrounding the metal centre, the binding modes of the fragments would be conserved between rkbPAP and animal PAPs. The structural similarities, the ready availability of rkbPAP and the fact that rkbPAP crystallizes easily make this enzyme suitable for crystallographic studies of inhibitor binding.
Crystallographic investigations into the fragment binding modes
The structures of the three fragments in complex with rkbPAP were determined by X-ray crystallography to 2.3 Å for CC24201 and MO07123 and to 2.7 Å resolution for CC27209 (Table 3). The asymmetric unit in all of the structures consists of two dimers of rkbPAP. The alignment of the dimeric structure reported here onto the first reported crystal structure of rkbPAP (PDB ID: 4KBP) gave an RMSD value of 0.38 Å after superimposition of all Cα atoms, demonstrating that the overall fold is conserved across all of the structures.
Table 3. Data collection and refinement statistics for the red kidney bean purple acid phosphatases – fragment complexes
| Temperature (K)||100||100||100|
| Resolution range (Å)||19.86–2.30||19.90–2.70||19.89–2.30|
| Total number of observations||677 862||420 125||618 281|
| Total number unique||126 115||77 045||115 253|
| Completeness (%)||99.8 (99.8)||99.7 (100)||94.8 (96.3)|
| Rmergea||0.11 (0.58)||0.11 (0.58)||0.15 (0.55)|
| Rmeas (all I+ & I−)||0.13 (0.65)||0.12 (0.64)||0.17 (0.63)|
| Rpim (all I+ & I−)||0.05 (0.28)||0.05 (0.27)||0.07 (0.28)|
| Mean I/σI||14.0 (3.5)||13.6 (2.9)||8.4 (2.6)|
| Multiplicity||5.4 (5.1)||5.5 (5.5)||5.4 (4.5)|
| Unit cell lengths (Å)|| a = b = 127.71 c = 299.76|| a = b = 126.76 c = 298.95|| a = b = 126.10 c = 297.98|
| Space group|| P31 21|| P31 21|| P31 21|
| Total number of atoms||16 288||15 253||16 359|
| Number of water molecules||1424||432||1609|
| Wilson B-factor (Å2)||27.46||45.43||25.46|
| RMS bonds (Å)||0.015||0.017||0.014|
| RMS angles (°)||1.1||1.4||1.1|
|Ramachandran statistics (%)|
For the CC24201 complex, the inhibitor was modelled into the difference electron density (σ > 4.5) in one of the subunits (A). However, it was not possible to unequivocally fit it or any other ligand to the electron density in subunits B, C and D (the correlation of the model to 2Fo-Fc electron density was poor). For the MO07123 complex, the inhibitor was modelled in three active sites and covered by electron density (σ > 3.7) in subunits A, B and C, but for subunit D, there was only electron density for the thiazole ring and carboxylate group of the inhibitor. Variation in the binding mode of this inhibitor in the different active sites was observed, with one binding mode being more similar to that of CC24201 (i.e., the phenyl ring forms π-cation interactions with the side-chain imidazole ring of H295) and the other binding mode being more consistent with docking predictions (this is illustrated in Figure 2A). For CC27209, the inhibitor was modelled in all four active sites (σ > 5.0), but also with variation between the individual sites, differing particularly in its magnitude of interaction with R258 and in the spatial arrangement of the functional groups of the inhibitor (this variability is illustrated in Figure 2B). The binding of CC24201 in active site A, MO07123 in active site B and CC27209 in active site B is described later.
Figure 2. Superimposition of subunit B (cyan) and subunit A (yellow) of the complex formed between rkbPAP and (A) MO07123 (B) CC27209. For CC27209, the major difference in inhibitor binding is observed as a stronger interaction with R258 (yellow) and a change of the conformation of the inhibitor. For MO07123, the major difference in inhibitor binding is driven by the interaction of the phenyl ring with the side chains that line the active site. One fragment (yellow) forms π-cation interactions through the phenyl ring with a nitrogen atom in the H295 side-chain imidazole ring, whereas another (cyan) forms π-cation interactions through the phenyl ring with the R258 side-chain guanidino group.
Download figure to PowerPoint
CC24201 coordinates to the metal centre through its carboxylate group with one oxygen bridging between the metals (2.7 Å from Fe and 2.9 Å from Zn) and the other oxygen forming a hydrogen bond with the N201 side-chain amine (2.3 Å) and an electrostatic interaction with a nitrogen (3.1 Å) in the side-chain imidazole of H202 (Figure 3). This oxygen interacts only weakly with Zn (3.3 Å). The pyridine nitrogen forms a hydrogen bond (2.4 Å) with the side-chain hydroxyl of Y365 and with a water molecule (3.1 Å), and the phenyl ring forms a π-cation interaction with a nitrogen in the side-chain imidazole group of H295.
Figure 3. (A) Fo-Fc electron density map (σ = 4.0) and stick representation showing the complex formed between the active site of rkbPAP (green) and CC24201 (cyan); a water molecule (red sphere) is observed bridging the inhibitor to rkbPAP. (B) Surface and stick representation showing the fit of CC24201 (cyan) in the active site of rkbPAP (yellow). (C) Surface and stick representation of the predicted binding mode of CC24201 (green) to red kidney bean purple acid phosphatases (PAP) (yellow) based on MVD docking simulations. (D) Surface and stick representation of the predicted binding mode of CC24201 (green) to human PAP (cyan) based on MVD docking simulations.
Download figure to PowerPoint
CC27209 is bound to the metal centre with the aromatic ring being approximately co-planar to the metals and the closest heavy atom to the metal centre being a carbon from the phenyl ring (2.6 Å from Fe and 2.7 Å from Zn) (Figure 4). The hydroxyl oxygen of the inhibitor forms a hydrogen bond with the hydroxyl oxygen in the Y365 side-chain (2.5 Å), and the oxygen in the 5-membered ring is hydrogen bonded to a water molecule (2.8 Å) that bridges the inhibitor to a nitrogen atom (2.8 Å) in the side-chain imidazole of H295. The unexpected observation here is that the aromatic carbons form close contacts with the metal centre, whereas the hydroxyl group on the inhibitor faces away from the metal centre and interacts with the side-chain hydroxyl of Y365.
Figure 4. (A) Fo-Fc electron density map (σ = 5.0) and stick representation showing the unusual complex formed between the active site of rkbPAP (green) and CC27209 (cyan); two water molecules (red spheres) are observed bridging the inhibitor to rkbPAP. (B) Surface and stick representation showing the fit of CC27209 (cyan) in the active site of rkbPAP (yellow). (C) Surface and stick representation of the predicted binding mode of CC27209 (green) to rkbPAP (yellow) based on MVD docking simulations. (D) Surface and stick representation of the predicted binding mode of CC27209 (green) to human purple acid phosphatases (cyan) based on MVD docking simulations.
Download figure to PowerPoint
For MO07123, the carboxylate group binds to the metal centre in a bidentate manner, with one oxygen coordinating to Fe (1.9 Å) and to Zn (3.1 Å) (Figure 5). The second oxygen of the carboxylate group binds to Zn (2.0 Å), while forming a hydrogen bond (2.9 Å) with the N201 side-chain amide. The thiazole ring of MO07123 forms π-cation interactions with a nitrogen atom in the H202 side-chain imidazole and the phenyl ring forms close contacts with the side-chain guanidino group of R258.
Figure 5. (A) Fo-Fc electron density map (σ = 3.7) and stick representation showing the complex formed between the active site of rkbPAP (green) and MO07123 (cyan); a water molecule (red sphere) is observed bridging the inhibitor to rkbPAP. (B) Surface and stick representation showing the fit of MO07123 (cyan) in the active site of rkbPAP (yellow). (C) Surface and stick representation of the predicted binding mode of MO07123 (green) to rkbPAP (yellow) based on MVD docking simulations. (D) Surface and stick representation of the predicted binding mode of MO07123 (green) to human purple acid phosphatases (cyan) based on MVD docking simulations.
Download figure to PowerPoint
Docking studies using MVD (47) were undertaken to model the binding of the three Maybridge inhibitors to rkbPAP. It was rationalized that if the program could successfully predict the binding modes of the inhibitors to rkbPAP, then it should be useful for predictions of the binding of the inhibitors to human PAP. For the CC24201-rkbPAP complex, MVD successfully predicted the key elements of molecular recognition (i.e., the interaction of the carboxylate group with the metal centre, the interaction of the pyridine nitrogen with Y365 and the interaction of the phenyl ring with R258) but the predicted orientation of the inhibitor was slightly different to that observed in the crystal structure (RMSD = 1.56 Å for all atoms; Figure 3). For the CC27209-rkbPAP complex, MVD predicted a binding orientation similar to that of CC27209 observed in subunit A (RMSD = 0.91 Å for all atoms). This result is illustrated in Figure 4. For the rkbPAP-MO07123 complex, MVD predicted a binding orientation similar to that of MO07123 observed in subunit B (RMSD = 0.64 Å for all atoms; Figure 5).
Subsequent docking studies were undertaken to predict the binding of the three inhibitors to human PAP (Figures 3–5). Analysis shows that they can all bind to the binuclear centre in a similar fashion to that observed in the rkbPAP but that there is an ∼180° rotation when their orientations are compared to how they bind to rkbPAP. For MO07123 and CC24201, the carboxylate group is predicted to coordinate directly to the metal centre and interact with the conserved N89 and H90 (N201 and H202 in rkbPAP), whilst the thiazole sulphur or pyridine nitrogen are oriented towards a pocket in the interface between the conserved metal ligating residues N89 and H221 (N201 and H325 in rkbPAP) and the mammalian repression loop (N142, S143 and D144, spatially equivalent to R258 in rkbPAP), whereby the sulphur or nitrogen are held in place by electrostatic interactions with the side chains of these residues. The phenyl ring of both inhibitors is observed forming hydrophobic interactions with F54 (spatially equivalent to R170 in rkbPAP). For CC27209, the phenyl group is held close to the metal centre and forms π-cation interactions with H90, whilst the 5-membered ring and the dimethyl tail form hydrophobic interactions with H90. The hydroxyl group is oriented towards the same pocket discussed earlier, where it is held in place by hydrogen bonds and electrostatic interactions.