Crystal Structure of the Tumor-promoter Okadaic Acid Bound to Protein Phosphatase-1*

,

The phosphorylation and dephosphorylation of proteins is vital to the regulation of many cellular pathways and processes. Two classes of enzymes in the cell that catalyze cellular dephosphorylation activity are tyrosine phosphatases and serine/threonine phosphatases (1). Classification of serine/threonine phosphatases can be subdivided into four categories: protein phosphatase-1 (PP1), 1 -2A (PP2A), -2B(PP2B) and -2C (PP2C) (2). The first three of these categories comprise what is known as the PPP family of protein phosphatases since they contain extensive sequence similarity in their catalytic domains and little or no sequence homology to PP2C or to tyrosine phosphatases. There are several natural toxin inhibitors of the PPP family of enzymes. These include microcystins, calyculins, tautomycin and okadaic acid (OA) (Fig. 1) (1).
There have been many biochemical and modeling studies on the inhibition of the PPP family of phosphatases by the natural toxins, but the lone crystal structure is of microcystin-LR (MCLR) bound to PP1 (␣ isoform) (8). Here we describe the crystal structure of OA bound to the recombinant catalytic subunit of PP1 (␥ isoform).

EXPERIMENTAL PROCEDURES
Crystallization-The catalytic subunit of protein phosphatase-1 ␥ isoform was purified as described previously (9,10). OA was purified from Prorocentrum lima (9,10). Crystals were obtained by the hanging drop vapor diffusion method at room temperature. The enzyme and inhibitor were mixed in a 1:2 molar ratio with the concentration of protein being ϳ0.4 mM. The PP1⅐OA complex was then mixed with an equivalent volume of mother liquor, which consisted of 2 M lithium sulfate, 100 mM Tris (pH 8.0), 2% polyethylene glycol 400, and 10 mM ␤-mercaptoethanol. The complex crystallized in the space group P4 2 2 1 2 with cell dimensions a ϭ b ϭ 99.18 Å, c ϭ 62.17 Å, with one complex per asymmetric unit.
Data Collection, Structure Determination, and Refinement-A first data set to 2.9 Å was collected at 100 K on a Rigaku RU-H3R rotating anode generator equipped with a Rigaku R-AXIS IVϩϩ image plate detector. The data were processed with the HKL suite of programs (11). This initial structure was solved by molecular replacement with the program AMoRe, using the PP1-MCLR structure (with MCLR removed) as a search model (Protein Data Bank accession code 1FJM) (12). Electron density for both the protein and the inhibitor were clear from the initial map generated from the molecular replacement solution. Another data set to 1.9 Å was then collected on the same instrumentation as above. The first structure was used as a search model for another round of molecular replacement as above. OA was fit to the difference density using the crystal structure of the free inhibitor as the starting model (6). The protein-inhibitor model was subjected to rigidbody refinement in CNS prior to manually fitting the model using the program XtalView (13,14). The model was then subjected to iterative rounds of macromolecular refinement using CNS with a maximum likelihood target. The crystallographic data are listed in Table I 1 The abbreviations used are: PP, protein phosphatase; OA, okadaic validity using WHATCHECK and PROCHECK (15,16). PROCHECK showed that 98% of residues were in allowed Ramachandran plot ranges with an overall G factor of 0.24. The atomic coordinates and structure factors have been deposited in the Protein Data Bank (accession code 1JK7).

RESULTS
Okadaic Acid-PP1 Structure-The catalytic subunit of PP1 is composed of an ␣-helical domain (9 ␣-helices) and a ␤-sheet domain (14 ␤-strands comprising 3 ␤-sheets) (Fig. 1). The active site lies at one of the interfaces between the two domains. Residues within the active site are almost exclusively found on segments of random coil joining regular secondary structural elements, rather than within the regular secondary structural elements themselves. There have been two previously published structures of each of phosphatases PP1 and PP2B (8,(17)(18)(19). The overall structure obtained here with OA-bound is very similar to both the tungstate-inhibited structure of PP1 (C ␣ r.m.s.d. 0.555 Å (residues 6 -288) averaged over the two tungstate⅐PP1 complex molecules in the asymmetric unit) and the structure of PP1 with the inhibitor MCLR-bound (C ␣ r.m.s.d. 0.518 Å (residues 6 -285) averaged over the two PP1⅐MCLR complexes in the asymmetric unit) (Fig. 3).
Inhibitor Binding-The presence of three grooves (hydrophobic, C-terminal, and acidic) on the surface of PP1 has been reported previously ( Fig. 1) (8). The double ring spiroketal moiety of OA is hydrophobic and binds into the hydrophobic groove on the surface of the protein. In the PP1⅐OA complex, Trp-206 and Ile-130 in the hydrophobic groove appear to be the two most important residues in this interaction due to their proximity to the hydrophobic segment of OA (Fig. 2). Other hydrophobic interactions occur between the C-4 to C-16 region of OA (which includes two six-membered rings) and the PP1 residues Phe-276 and Val-250.
The remaining sites of interaction between PP1 and OA involve hydrogen bonding. A conserved acidic motif is present in many PP1 inhibitors, either as a carboxylic acid (C-1 in OA) or as a phosphate (calyculin A) (Fig. 2). In our structure, the acid motif in OA accepts a hydrogen bond from the hydroxyl group of Tyr-272 (20). Other hydrogen bonding interactions occur between Arg-96 and the C-2 hydroxyl of the inhibitor and between Arg-221 and the C-24 hydroxyl group of OA.
A comparison of OA-bound to PP1 and the crystal structure of OA alone reveals that the two structures have a very similar overall conformation (0.397 Å r.m.s.d. over all 56 atoms). In the structure of unbound OA, a hydrogen bond between the C-24 hydroxyl and the C-1 acid keeps the inhibitor in a cyclic structure. This hydrogen bond is present in the PP1⅐OA complex as well. The recent proposal that, based on modeling consider-  ations, OA binds to PP1 in an extended conformation is clearly unfounded (21).

DISCUSSION
Crucial PP1-OA Interactions-The presence of a hydrophobic moiety in many PPP family inhibitors suggests that the hydrophobic groove may play a key role in inhibitor binding. This was seen in the crystal structure of MCLR bound to PP1 where the hydrophobic ␤-(2S, 3S, 8S, 9S)-3-amino-9-methoxy-2, 6, 8-trimethyl-10-phenyldeca-4,6-dienoic acid (Adda) group of MCLR contacts residues in the hydrophobic groove of PP1 and is repeated in our structure where the hydrophobic tail of OA contacts similar residues in the hydrophobic groove. The other hydrophobic interactions that occur are between the C-4 to C-16 region of OA and PP1 residues Phe-276 and Val-250. These interactions may actually be an impediment to OA inhibition since mutation of Phe-276 to a smaller and less hydrophobic cysteine reduces the K i for the acid 40-fold (22). The hydrophobic ring of Phe-276 may inhibit entry of OA into the active site, overshadowing other favorable hydrophobic interactions. A cysteine (Cys-269) is present at the equivalent position in PP2A and may be the reason that PP2A has a higher affinity for OA than does PP1. Interestingly this cysteine in PP2A has recently been implicated as a key residue in the interaction of the PP2A/PP4-specific inhibitor fostriecin (23).
While the hydrophobic interactions between OA and PP1 are predominant, structure-function studies show that the hydrogen bonding interactions are important as well. The importance of the hydrogen bond between Tyr-272 and the acid group of OA is exemplified by the observation that esterification or removal of the acidic moiety in OA results in elimination of its inhibitory activity and by the fact that mutation of Tyr-272 to Phe results in a 50-fold increase in the K i value (20,22). Despite the intimate interaction of the C-2 hydroxyl group of OA with the Arg-96 side chain, removal of the hydroxyl group results in only a 7-fold increase in the K i value (24). In contrast, mutation of Arg-221 to Ser confers resistance to inhibition by OA, underlining the importance of the interaction between the Arg and the C-24 hydroxyl of OA (25).
Comparison to Other PPP Structures-As mentioned, the structure of PP1 obtained here is remarkably similar to the two structures of each of PP1 and PP2B determined previously. Even with only the phosphate-mimic tungstate present, the architecture of the active site of the tungstate-bound PP1 structure is virtually identical to our structure with OA-bound (Fig.  3). This suggests that either tungstate does not accurately mimic the phosphate product or that the binding of OA is enthalpically favorable and does not cause significant changes to the structure of the protein. In contrast, the MCLR-bound structure reveals large changes in the conformation of the active site relative to the tungstate-and OA-bound PP1 structures (Fig. 3). These changes are mainly restricted to the ␤12-␤13 loop, which has been previously implicated in inhibitor specificity (26,27). The loop in the MCLR structure folds back on itself, causing significant shifting of residues 273-278 relative to our structure and to the tungstate-bound structure of PP1. One critical difference between MCLR and OA is the presence of a dehydroalanine residue in MCLR that covalently alkylates the S ␥ of Cys-273 in a time-dependent reaction (1). This covalent linkage is not the primary cause of inhibition of PP1 by MCLR (8). It is unclear whether inhibition by MCLR requires the movement of the ␤12-␤13 loop or whether this movement occurs along with covalent bond formation. Given the strong similarity of the PP1-interacting domains of OA and MCLR, it is likely that the primary mode of inhibition of PP1 by MCLR is similar to that of OA and that the movement of the ␤12-␤13 loop in the MCLR complex is a secondary event accompanying the covalent bonding reaction. An important interaction between PP1 and MCLR that is not present in the   FIG. 2. A, stereo representation of the active site of the PP1⅐OA complex. Pertinent active site residues are labeled and are shown as a ball-and-stick representation with carbon atoms colored gray, oxygen atoms colored red, and nitrogen atoms colored blue. OA is shown as a balland-stick representation with carbon atoms colored yellow and oxygen atoms colored red. The intramolecular hydrogen bond in OA is shown as a dashed line. The active-site manganese atoms are shown as yellow spheres. B, active site of the PP1⅐OA complex. All residues of PP1 within 4 Å of the OA are shown, the closest residue-OA interactions are shown by dashed lines, and the distances of all possible hydrogen bonding interactions are labeled.
PP1⅐OA complex is the hydrogen bond that occurs between Arg-96 (PP1) and the acid of the methyl-aspartate residue (MCLR). This interaction may account for the 100-fold greater inhibition of PP1 by MCLR over OA (28).
The structure of PP1-OA is strikingly similar to the structure of PP2B (Fig. 3) despite the fact that OA does not strongly inhibit PP2B. In the catalytic domain of PP2B, the protein backbone around the active site is virtually identical to that in PP1, including the ␤12-␤13 loop. However, the overall C ␣ r.m.s.d. is 1.04 Å (residues 6 -272), indicating some significant differences in other parts of the protein, namely in the less structured N and C termini regions. Several attempts to model the PP1-OA structure based on the other PPP family structures have been made (29 -32). These models have largely been based on the MCLR-bound structure in which the conformation of the ␤12-␤13 loop differs from PP1 with tungstate-bound and our structure with OAbound. The OA-PP1 structure presented here will greatly facilitate models based on PP1 inhibitors structurally related to OA (e.g. tautomycin).
OA Binding to Other Protein Phosphatases-OA exhibits different inhibitory potential on the structurally similar PPP family members. Okadaic acid inhibits PP2A most strongly (IC 50 Ϸ0.1 nM), PP1 less well (IC 50 Ϸ10 nM), and PP2B weakly (IC 50 Ϸ1-2 M) (28,33). The primary difference in the active site region between PP1 and PP2A is in the ␤12-␤13 loop where PP1 contains the residues GEFD (residues 274 -277) but PP2A has YRCG (residues 267-270) in the equivalent positions (Fig.  2). While mutation of Phe-276 to Cys (in PP2A) clearly facilitates inhibition by OA, there are likely additional reasons for the enhanced inhibition of PP2A by OA. The side chain of Tyr-267 in PP2A should point into solvent, away from the active site, if it is in the same position as the equivalent Gly-274 in PP1, so this change should not be expected to be very significant. However, the allowed torsion angles of a tyrosine would constrain the ␤12-␤13 loop to a larger degree (Gly-274 is in the third position of a type II ␤-turn and is currently in a region of the Ramachandran plot that is disallowed for Tyr residues), and perhaps the loop is brought into a position to allow for a hydrogen bond to occur between the tyrosine and a hydroxyl group on OA (C-7-OH). The tyrosine could also produce a favorable hydrophobic interaction with the C-10 methyl group of OA. Further interactions in PP2A may result from substitution of Glu-275 in PP1 with Arg-268 in PP2A. Hydrogen bonding interactions may occur between this arginine and the C-7-OH group or hydrophobic interactions with the C-10 methyl group. The enhanced inhibitory potency of OA with PP2A is likely a combination of these effects. FIG. 3. A, stereo representation of the backbone carbon alignment of tungstatebound PP1 (gold), PP1⅐OA complex (blue), and PP1⅐MCLR complex (PDB ID 1FJM) (red). OA is shown in ball-and-stick representation with carbon atoms in gray and oxygen atoms in red. The metals are from the PP1-OA structure and are shown as spheres as in Fig. 1. B, stereo representation of the backbone carbon alignment of PP1-OA (blue) and PP2B (PDB ID 1TCO) (gold). C, stereo representation of the backbone carbon alignment between the PP1⅐OA complex (blue) and the PP1⅐MCLR complex (red). OA is shown as ball-and-stick representation and is colored light blue, MCLR is shown as ball-and-stick representation and is colored red.
The decreased inhibition of PP2B with OA is more difficult to interpret. Major active site differences between the two proteins are the replacement of Ile-133 in PP1 with Tyr-159 in PP2B, Tyr-134 in PP1 with Phe-160 in PP2B, Cys-273 in PP1 with Leu-312 in PP2B, and Phe-276 in PP1 with Tyr-315 in PP2B. Of these, changing Phe-276 in PP1 to a tyrosine in PP2B may make the greatest impact on OA inhibition. The portion of the inhibitor that resides in this part of the active site is quite hydrophobic, interacting with Phe-276, the methylene carbon of Cys-273, the side chain of Val-250, and the aromatic ring of Tyr-272. The ring of Phe-276 points almost directly at one of the double ring systems of OA (carbons 4 -12) (Fig. 2), and the introduction of a hydroxyl group in PP2B may be enough to disrupt the interaction between OA and the protein, thus reducing the inhibitory potency of OA for PP2B. The resistance of PP2B to OA may therefore arise from the combination of this residue change and subtle structural changes in key OA contact residues in the potential OA binding site. Determination of the PP1-OA structure presented here will now facilitate rational mutagenesis of the PP2B catalytic subunit to test this hypothesis.