Crystal Structures of Protein Phosphatase-1 Bound to Motuporin and Dihydromicrocystin-LA: Elucidation of the Mechanism of Enzyme Inhibition by Cyanobacterial Toxins

doi:10.1016/j.jmb.2005.11.019 J. Mol. Biol. (2006) 356, 111–120

Jason T. Maynes1, Hue A. Luu1,MaiaM.Cherney1, Raymond J. Andersen2 David Williams2, Charles F. B. Holmes1* and Michael N. G. James1

1Canadian Institutes of Health The microcystins and nodularins are tumour promoting hepatotoxins that Research, Group in Protein are responsible for global adverse human health effects and wildlife Structure and Function fatalities in countries where drinking water supplies contain cyanobacteria. Department of Biochemistry The toxins function by inhibiting broad speciﬁcity Ser/Thr protein Faculty of Medicine, University phosphatases in the host cells, thereby disrupting signal transduction of Alberta, Edmonton, Alta pathways. A previous crystal structure of a microcystin bound to the Canada T6G 2H7 catalytic subunit of protein phosphatase-1 (PP-1c) showed distinct changes in the active site region when compared with protein phosphatase-1 2Department of Chemistry structures bound to other toxins. We have elucidated the crystal structures University of British Columbia of the cyanotoxins, motuporin (nodularin-V) and dihydromicrocystin-LA Vancouver, BC, Canada bound to human protein phosphatase-1c (g isoform). The atomic structures of these complexes reveal the structural basis for inhibition of protein phosphatases by these toxins. Comparisons of the structures of the cyanobacterial toxin:phosphatase complexes explain the biochemical mechanism by which microcystins but not nodularins permanently modify their protein phosphatase targets by covalent addition to an active site cysteine residue. q 2005 Elsevier Ltd. All rights reserved. *Corresponding author Keywords: crystal structure; cyanobacteria; microcystin; nodularin

Introduction with hepatocyte deformation due to reorganization of microﬁlaments.7 Liver tumour promotion may be Microcystin and nodularin (cyclic peptide linked to the ability of these toxins to promote hepatotoxins) are cyanobacterial metabolites cytokeratin hyperphosphorylation associated with found worldwide both in freshwater and in marine morphological changes in rat hepatocytes.8 environments.1 These toxins are responsible for Most heptapeptide microcystins differ in the wildlife fatalities and adverse human health effects nature of two variable L-amino acids and the in countries where drinking water supplies contain absence of methyl groups on D-erythro-b-methyl toxic cyanobacteria. The nodularins and microcystins aspartic acid (Masp) and/or N-methyldehydro- are powerful liver tumour promoters and potent alanine (Nmda) residues (Figure 1(a)).9,10 Nodularins inhibitors of the protein phosphatase-1 and -2A are structurally related cyclic pentapeptides that catalytic subunits (PP-1c/PP-2Ac).2–6 Inhibition of inhibit PP-1c/PP-2Ac with similar potency to that of these enzymes in the liver has been associated the microcystins but only contain one variable amino acid site.11 The hydrophobic pentapeptide motuporin (MOT, a member of the nodularin class of inhibitors Abbreviations used: PP-1c, protein phosphatase-1; and also termed nodularin-V due to the valine in its Masp, D-erythro-b-methyl aspartic acid; MOT, motuporin; variable position, Figure 1(b)), was identiﬁed from the Adda, 3-amino-9-methoxy-2,6,8-trimethyl-10-phenyl- 12 deca-4,6-dienoic acid; OA, okadaic acid; Nmda, N- marine sponge Theonella swinhoei. methyldehydroalanine; MC-LR, microcystin-LR. Previously the solution structures of microcystin- E-mail address of the corresponding author: LR (MCLR, LZleucine and RZarginine in two [email protected] variable positions) and of MOT using 1H nuclear

molecules through the a-carboxyl group of its g-linked D-glutamic acid moiety and the adjacent carbonyl oxygen. This accounts for the observation that the glutamic acid a-carboxyl group is absol- utely necessary for toxicity of the cyclic peptide.6 In addition, the carboxyl group of the Masp residue of MCLR interacts with Arg96 and Tyr134 of PP-1c. Thus, MCLR completely blocks access to the active centre of the enzyme.16 The long hydrophobic tail, composed of the Adda residue of MCLR, is placed in the hydrophobic groove region of PP-1c, adjacent to the active site. Interactions of cyclic peptide toxins also occur at the toxin-sensitive b12-b13 loop (residues 268–281) of PP-1c. A covalent linkage between the Nmda side-chain of the toxin and the sulphur of C273 of PP-1c was observed in the crystal structure, verifying earlier biochemical analyses.17–21 This covalent reaction most likely proceeds via a Michael addition type-reaction to the double bond of Nmda. Comparison of the crystal structures of PP-1c with and without MCLR bound22,23 suggested that the b12-b13 loop of PP-1c shifts to avoid steric conflict between Tyr272 and the Nmda residue of the toxin. This observation first suggested that flexibility of the b12-b13 loop might be important for toxin binding and inhibition of the enzyme. However, the covalent linkage that occurs is secondary to the inhibitor activity of the toxin and likely occurs as a delayed Figure 1. Chemical structure of the cyanobacterial toxins. (a) Structure of dihydromicrocystin-LA (MCLA- reaction in solution. The biological relevance of this 2H). #The two sites of modification for the microcystins, in reaction is unknown. MCLA-2H these sites are Leu and Ala and in MCLR these Whilst sharing similar biological properties, sites are Leu and Arg. *Thesiteofhydrogenation important functional differences exist between the that removes the N-methyldehydroalanine residue microcystins and nodularins with respect to their and creates an N-methylalanine residue with no interaction with PP-1c/PP-2Ac. Although both Michael addition properties. The 3-amino-9-methoxy- toxins initially bind non-covalently and inhibit 2,6,8-trimethyl-10-phenyl-deca-4,6-dienoic acid (Adda) these enzymes, the nodularins, including MOT, residue includes all of the hydrophobic tail region of the do not bind covalently to PP-1c/PP-2Ac even as a inhibitor. (b) Structure of motuporin (nodularin-V). The delayed reaction.17,18 This is despite possessing an Adda residue includes all of the hydrophobic tail region of the inhibitor. (c) Structure of nodularin-R. N-methyldehydrobutyrine residue that could undergo a Michael addition reaction with C273 similar to the reaction of the Nmda residue in MCLR. magnetic resonance spectroscopy were deter- Recently, the crystal structures of PP-1cg bound mined.13 Subsequently, the solution structure of to two other natural toxins (okadaic acid (OA) and the related nodularin-R (variable amino acid is calyculin) have been determined.24,25 The overall arginine, Figure 1(c)) was also elucidated.14 The structure of PP-1c in all of these complexes is similar peptide ring of MCLR has a saddle-shaped motif to PP-1ca in the PP-1c:MCLR complex, except for with carboxyl residues as stirrups at the sides. Both prominent differences in the orientation of the the 3-amino-9-methoxy-2,6,8-trimethyl-10-phenyl- b12-b13 loop region. These findings suggested that deca-4,6-dienoic acid (Adda) and arginine side- the different orientation of the b12-b13 loop in the chains are highly flexible, whereas the leucine PP-1c:MCLR complex15 may have been solely due side-chain of the toxin is well defined. The cyclic to the unusual covalent binding between C273 of peptide ring of MOT forms a smaller, well-defined PP-1c and MCLR. In order to verify this hypothesis saddle shape having less pucker than MCLR. we have determined the structure of PP-1cg bound Determination of the crystal structure of MCLR to MOT, which, besides being a novel inhibitor bound to PP-1c (a isoform) showed that the toxin of PP-1c, possesses a dehydrobutyrine residue that binds to PP-1c through interactions with three could potentially undergo a Michael addition regions of the phosphatase15 including the active reaction with C273, similar to the microcystins. We site with the two catalytic metals, the hydrophobic have also determined the structure of a microcystin groove and the b12-b13 loop region. Microcystin-LR congener (dihydromicrocystin-LA(MCLA-2H); coordinates with the two catalytic metal atoms of Figure 1(a)) in which the Nmda residue has the phosphatase indirectly by binding two water been hydrogenated, thereby removing the ability Cyanobacterial Toxin Recognition 113 of the microcystin to form a covalent bond with C273 and the hydrogenated Nmda residue is 8.2 A˚ . C273 of PP-1c. In addition, there are hydrogen bonding contacts between MCLA-2H and PP-1c including interactions with Y134, R221 and Y272 (Figure 4(a)). Of Results note is the lack of a hydrogen bond between Arg96 and the g-linked glutamic acid, which is present Structure of motuporin (nodularin-V) bound in the PP-1c:MOT complex (Figure 2). to PP-1c Comparison of PP-1c structures bound The toxins of PP-1c contain three common to motuporin, dihydromicrocystin-LA features: a carboxyl group, a hydrophobic tail and and microcystin-LR a large macrocyclic domain (the macrocycle is created via an intramolecular hydrogen bond in Comparison of PP-1c structures with different OA). Motuporin binds to PP-1c in a manner similar toxins bound reveals overall similarity in the to other toxins in that its macrocyclic ring occupies enzyme structures. The structure of PP-1c bound most of the active site region and its hydrophobic to MCLA-2H has a Ca RMSD value of 0.3 A˚ (over tail occupies the hydrophobic groove region of 293 atoms, including the b12-b13 loop) when PP-1c that is adjacent to the active site (Figure 2). compared with the PP-1c:OA complex and 0.5 A˚ The a-carboxyl moiety from the g-linked D-glutamic (over the same 293 atoms) when compared with the acid of MOT lies in proximity to similar organic PP-1c:MCLR complex. The equivalent values for acids from MCLR and OA and to the phosphate the PP-1c:MOT complex are 0.3 A˚ and 0.5 A˚ for the from calyculin A. These negatively charged groups OA and MCLR complexes, respectively. However, serve to bind indirectly to the active site metals of comparison of PP-1c structures bound to MOT, the enzyme via bridging water molecules. MCLA-2H and MCLR illustrate a significant change The hydrophobic tail of MOT, composed of the in the conformation of the toxin-sensitive b12-b13 long Adda side-chain, primarily interacts with loop due to covalent modification of C273 in PP-1c residues W206, V223 and I130 of PP-1c (Figure 2). (Figure 5). The Ca RMSD value over residues There are a significant number of hydrogen 270–276 of the loop region is 2.0 A˚ for comparison bonding contacts between MOT and PP-1c, includ- of the MCLR complex to both the MCLA-2H and ing strong interactions with Arg96 (three potential MOT complexes. This contrasts with a Ca RMSD bonds, all involving guanidinium nitrogen atoms), value of 0.1 A˚ when comparing the equivalent Y134, R221 and Y272 (Figure 2(a)). These inter- regions of the MCLA-2H and MOT complexes. The actions include the conserved carboxylate motif conformation of this loop is identical in the structures present in the g-linked glutamic acid and the of PP-1c bound to MOTand MCLA-2H; however, it is carboxylate of the Masp residue. Motuporin inter- substantially shifted in the MCLR complex. acts with the active site metals indirectly through a hydrogen bond to a bridging water. In comparison with the previous MCLR structure where the Discussion dehydroalanine residue reacted with C273, the dehydrobutyrine residue of MOT is 4.7 A˚ from The PP-1c:MOT structure presented here is the the cysteine and forms no covalent adduct. fourth novel PP-1c-toxin structure to be eluci- The structure of MOT bound to PP-1c is quite dated.15,24,25 Each determined structure is suffi- different from the minimized average solution ciently unique to contribute to the hypothesis that structure determined by NMR (RMSD of 1.7 A˚ toxin/substrate binding to PP-1c can occur by a over 55 atoms) (Figure 3(a)).13 This is in contrast to multitude of modes. The hydrophobic tail of MOT the similarity of PP-1c-bound and free structures does not interact with PP-1c as intimately as the observed for OA (RMSD of 0.7 A˚ over 70 atoms of corresponding hydrophobic portion of OA, as the toxin in a comparison of the bound and illustrated by the higher thermal coefficients and unbound crystal structures)25 but similar to MCLR poorer electron density of the region. The average (RMSD of 1.8 A˚ over 57 atoms in a comparison thermal factor for the atoms of the hydrophobic of PP-1-bound crystal and the unbound NMR tail of the Adda residue of MOT is 5.8 A˚ 2 greater minimal-averaged structure).15 than the average thermal factor for the entire inhibitor. The equivalent hydrophobic tail atoms Structure of dihydromicrocystin-LA bound of OA have an average thermal factor 1.3 A˚ 2 above to PP-1c the average thermal factor for the entire inhibitor. The hydrophobic tail of OA has five potential van Dihydromicrocystin-LA binds to PP-1c in a der Waals interactions with the hydrophobic groove similar manner to its congener MCLR (Figures 4 that are within 4 A˚ , whereas MOT only has two and 5) with the notable exception being the lack of a potential interactions. covalent interaction with C273 in the b12-b13 loop of The macrocylic ring portion of OA that occupies the phosphatase. Significantly, this loop adopts a the active site region of PP-1c has very few very different conformation in the absence of a hydrogen bonding interactions with the enzyme covalent bond with the toxin. The distance between and therefore hydrophobic binding may provide 114 Cyanobacterial Toxin Recognition

Figure 2. Crystal structure of motuporin bound to protein phosphatase-1c. (a) Structural formula representation of MOT showing active site contacts with PP-1c within 4 A˚ . All potential hydrogen bonds have their distances shown. (b) Stereo representation of the active site contacts in the crystal structure of MOT bound to PP-1c. The enzyme is shown as a green backbone ribbon trace with grey carbon atoms on the side-chains of pertinent residues. Motuporin is displayed as sticks with orange carbon atoms. (c) Stereo representation of the electron density seen for MOT K superimposed on the surface of PP-1c (colored grey). All density shown is a sA-weighted 2Fo Fc map contoured at 1s. the larger proportion of the binding energy (two ring, specifically the two acidic moieties of the ring, potential hydrogen bonds exist between the inhibitor and therefore could rely less on hydrophobic and the enzyme, both of which are relatively binding energy (five potential hydrogen bonds distant, being close to or greater than 3 A˚ in length, exist between the inhibitor and the enzyme, three with putative donor-hydrogen acceptor hydrogen of which are relatively strong, being !2.9 A˚ in bond angles of 1348 and 1478 (defined as length, hydrogen bond angles between 1028 and the putative donor–hydrogen-acceptor angle). 1538). The comparison of enzyme-bound and free Motuporin has considerably more hydrogen inhibitor structures also shows evidence of this. bonds between the enzyme and the macrocylic It was noted previously that the bound and Cyanobacterial Toxin Recognition 115

Figure 3. Superimposition of the determined nodularin structures. (a) Overlay of the MOTstructure bound to PP-1c, as determined by X-ray crystallography (green) and solution MOTstructure as determined by NMR (orange).13 Orientation of the inhibitors is the same as in Figure 2. (b) Overlay of the MOT (nodularin-V) structure bound to PP-1c, as determined by X-ray crystallography (green) and a solution structure of nodularin-R.14 Orientation of the inhibitors is the same as in Figure 2. free structures of OA are virtually identical and In the complexes of PP-1c with MCLA-2H and therefore the entropic cost of binding is minimized.25 MOT, Y272 functions as a hydrogen bonding A similar observation was made for the enzyme- partner with an acid moiety of the inhibitors. In bound and NMR solution structure of MCLR.13 this capacity, the function of the tyrosine is When comparing the solution NMR structure of equivalent for both the MCLA-2H and MOT. MOT with the enzyme-bound structure, differences Previous kinetic data showed that mutation of in the ring region and flexibility in the hydrophobic Y272 affected the two inhibitors to significantly tail region are noted (Figure 3(a)).13 A second different degrees.26 Abolition of the hydrogen comparison to the NMR solution structure of a bonding capacity but maintenance of the large related nodularin (nodularin-R) shows similar flexi- size of tyrosine by mutation to phenylalanine or bility in the tail region (Figure 3(b)).14 As evidence tryptophan had virtually no effect on the affinity of that OA relies more on hydrophobic binding and PP-1c for either inhibitor, suggesting that entropy, the buried protein surface area that OA the hydrogen bond from Y272 to the inhibitor occupies is 738 A˚ 2, where as MOT only occupies carboxylate may not be overly important (the same 500 A˚ 2. Taken together this evidence shows that MOT carboxyl group of the inhibitor also functions to most likely relies less on hydrophobic binding energy bind the catalytic metals of the phosphatase and minimizing the entropic cost of binding than indirectly via bridging water molecules). However, does OA and MCLR. Instead, MOT has numerous mutation of this tyrosine to other amino acids of hydrogen bonding interactions with the enzyme that varying chemical properties had quite different compensate for the lack of the former interactions. results. Replacement of the tyrosine with alanine, The increased hydrogen bonding interaction between serine/threonine, cysteine, glutamate or lysine typi- MOTand PP-1c may explain the increased potency of cally increased the IC50 value of MCLR by 10 to 50- this toxin towards PP-1c when compared with OA fold whereas the same substitutions for nodularin-R Z (IC50 0.06 nM versus 20 nM, respectively). increased the IC50 values by 150–250-fold. This 116 Cyanobacterial Toxin Recognition

Figure 4. Crystal structure of MCLA-2H bound to protein phosphatase-1c. (a) Linear representation of MCLA-2H showing active site contacts within 4 A˚ . All potential hydrogen bonds have their distances shown. Toxin is labelled as for Figure 1. (b) Stereo representation of the active site contacts in the crystal structure of MCLA-2H bound to PP-1c. The enzyme is shown as a blue backbone ribbon trace with grey carbon atoms for the side-chains of pertinent residues. Motuporin is displayed as sticks with orange carbon atoms. tyrosine occupies a noteworthy position in the that could participate in Michael addition reactions enzyme active site, near the catalytic metals and at (N-methyldehydroalanine in the microcystins and the base of the hinge of the b12-b13 loop region. In the N-methyldehydrobutyrine in MOT). Determination apoenzyme form, Y272 most likely does not have a of the crystal structure of MOT bound to PP-1c hydrogen-bonding partner, other than metal-bound provides a molecular explanation for the lack of water molecules. It most likely is more important in covalent binding to the phosphatase. The reactive creating the hydrophobic surface of the b12-b13 loop residue in PP-1c, C273, is not orientated sufficiently region and maintaining its position. Removal of the close to N-methyldehydrobutyrine to enable tyrosine might allow for greater movement of the covalent binding to occur. The lack of proximity of loop. This is consequential for the microcystins, since the sulphur nucleophile of C273 may disrupt the they possess a significant number of potential Burgi–Dunitz angle of approach necessary to obtain contacts in the region of the b12-b13 loop (Figure 4). a covalent Michael addition with this residue in Motuporin does not have this capacity, since its MOT.27 While the lack of covalent modification macrocyclic ring does not lie in proximity to the rest of affects the final complex that exists between MOT the b12-b13 loop. This may explain why mutations of and the cognate enzyme, it is unlikely to affect the Y272 affect MOT to a larger degree than do the inhibitory potential of MOT, since the covalent microcystins. reaction observed with MCLR was shown to differ Significant differences exist in the mechanism of temporally from initial inhibition of PP-1c.17 PP-1c binding between the motuporin/nodularin Comparison of PP-1c structures bound to MOT, class and the microcystin class of cyanobacterial MCLA-2H and MCLR illustrates a significant toxins in that the nodularin cyanotoxins do not bind change in the conformation of the toxin-sensitive covalently to their enzyme target. This occurs b12-b13 loop due to the covalent modification of despite the presence of residues in both classes C273 in PP-1c. The conformation of this loop is Cyanobacterial Toxin Recognition 117

Figure 5. Comparison of PP-1c structures bound to toxins. (a) Stereo representation of the overlay of PP-1c bound to 15 MCLA-2H (blue), MOT (green) and MCLR (orange). Motuporin is shown as sticks and the b12-b13 loop is labeled. (b) Stereo representation of the active site regions of the PP-1c-bound toxin structures. The b-methylaspartic acid (Masp), g-linked D-glutamic acid and N-methyldehydroalanine (Nmda) residues are labeled. In MCLA-2H, the N-methylalanine residue and in MOT, the N-methyldehydrobutyrine residues are both in the equivalent position to Nmda. The disulphide bond between the Nmda residue of MCLR and C273 is shown. Colouring is the same as in (a). identical in the presence of either MOT or non- MacKintosh et al.18 determined that mutation of covalently bound MCLA-2H (Figure 5), but changes C273 to Ser or Leu increased the IC50 for toxin substantially with the formation of the Michael inhibition of PP-1c by ten and 20-fold, respectively, addition complex between Nmda of MCLR and as well as abolishing covalent binding. Mutagenesis C273. Determination of the PP-1c structure bound of residues 273–277 of the b12-b13 loop of PP-1c to to MCLA-2H shows conclusively that the change in the corresponding residues (L273, D274, V275, Y276 the b12-b13 loop conformation in the PP-1c:MCLR and N277) from the related phosphatase calcineurin complex is due to the formation of a covalent by Maynes et al. resulted in a chimeric mutant that bond between the phosphatase and bound toxin. showed a similarly increased 20-fold resistance to Presumably, it is possible for the Michael addition microcystin inhibition.29 Determination of the reaction to occur without significant loop move- crystal structure of this PP-1c-calcineurin hybrid ment and, therefore, the loop rearrangement most and systematic mutation of each of the five residues likely occurs after the Michael addition reaction has in the b12-b13 loop, confirmed that a single amino taken place. acid change (C273L) and not a change in the overall Previously, site-directed mutagenesis has shown b12-b13 loop conformation, was the most influential that the initial non-covalent interactions between factor in the increased toxin resistance exhibited by C273 and the microcystins contribute to the initial mutation of C273.29 and rapid inhibition (in min) of PP-1c prior to a Several studies have suggested that the much slower covalent modification step. Runnegar N-methyldehydroalanine residue in microcystins et al.19 and MacKintosh et al.18 independently does not play a role in the initial enzyme-inhibitor showed that PP-1c binds covalently to MCLR, interactions. Microcystins immobilized to Sepharose confirming modification of the N-methyldehydro- beads via an aminoethanethiol adduct retain PP-1c alanine moiety of microcystin with PP-1c at C273.28 inhibitory activity.30 Conversion of MCLR to a Covalent binding of microcystins to PP-1c is not microcystin-glutathione conjugate via Michael observed when the phosphatase is denatured,18 addition to the N-methyldehydroalanine residue indicating that the interaction depends on an intact diminished its in vivo toxicity in animal models, but 28 Z active site. did not abolish it (LD50 630 mgperkgversus 38 mg 118 Cyanobacterial Toxin Recognition per kg for microcystin-LR, respectively).31 Finally, hypothesis has recently been obtained by Hastie Craig et al. confirmed a two step mechanism for et al.,35 who have reported that the dehydrobutyrine irreversible protein phosphatase inhibition by (Dhb)-containing microcystin variant [3-Asp, microcystins.17,28 Kinetically, covalent modification 5-ADMAAdda, 7-Dhb]microcystin-HtyR from of C273 (and the analogous residue C266 in the Nostoc sp. does not interact covalently with PP-2A, related protein phosphatase PP-2A) occurs slowly yet is as potent a PP-1c inhibitor as MCLR and 17,28 Z over several hours. This contrasts with the almost as toxic (LD50 100 mg per kg). initial rapid non-covalent binding and inhibition of the enzyme by multiple toxin interactions described here. Taken together, our crystal struc- Experimental Procedures tures show that these interactions must first occur with the b12-b13 loop in the same position as the Purification of natural product toxins previously determined PP-1c-OA and calyculin 24,25 complexes. This is entirely consistent with our Microcystin-LA was purified from natural previous conclusion that stoichiometric formation blooms of Microcystis aeruginosa using procedures of a covalent complex adventitiously follows after described.17 Motuporin was purified and charac- 16,17 prolonged incubation of toxin and phosphatase. terised as described by Desilva et al.12 Site-specific Evidence has been obtained for the persistent reduction of MCLA by NaBH4 was carried out as presence of covalent complexes between PP-1c and described;17 the final product was fully characterized MCLR in vivo in microcystin-poisoned salmon by high resolution mass spectrometry (FABMS) and afflicted by netpen liver disease,32 in salt water 1H-NMR analysis(data not shown). mussels33 and in microcystin-poisoned trout.34 Therefore, although not required for protein phos- Crystallization phatase inhibition, covalent modification may play a role in the bioavailability of the toxin and The catalytic subunit of human protein phospha- may influence temporally distinct signalling path- tase-1g was purified as described.17,25 Crystals were 32,34 ways involving Ser/Thr protein phosphatases. obtained by co-crystallization using the hanging- The data presented here suggest that non- drop vapor-diffusion method at room temperature. covalently bound complexes of phosphatase and For both MOT and MCLA-2H, the enzyme and toxin (MOT) also have the potential to be toxic, inhibitor were mixed in a 1:3 molar ratio with the since the nodularins possess similar toxicity to the final concentration of protein being 12 mg/ml. The 1,6,10 microcystins. Additional support for this PP-1c inhibitor complexes were then mixed with an

Table 1. Data collection and reﬁnement statistics

PP-1c–MOT complex PP-1c–MCLA–2H complex Data collection Unit cell aZbZ101.0 A˚ , cZ63.49 A˚ , aZbZgZ908 aZbZ100.0 A˚ , cZ62.9 A˚ , aZbZgZ908 Space group P42212 P42212 Wavelength (A˚ ) 1.00 1.00 Resolution (A˚ )a 40–2.1 (2.20–2.10) 40–2.3 (2.42–2.30) Total number of reflections 186,190 113,187 Number of unique reflections 19,726 14,724 Completeness (%)a 99.9 (99.9) 99.9 (99.9) Redundancya 3.7 (3.8) 7.7 (7.6) hI/s(I)ia 13.5 (4.6) 20.4 (6.7) a,b Rsym (%) 6.5 (29.3) 7.2 (28.7) Mosaicity (deg.) 0.37 0.52 Refinement Resolution (A˚ ) 40–2.1 40–2.3 Protein atoms 2350 2338 Waters 73 35 r.m.s. deviations Bond length (A˚ ) 0.02 0.013 Bond angles (deg.) 1.84 1.57 Average B-factors Protein atoms (A˚ 2) 26.3 32.7 Inhibitor atoms (A˚ 2) 32.3 43.7 Solvent (A˚ 2) 29 30.1 c Rcryst (%) 0.22 0.21 d Rfree (%) 0.26 0.25 a Data in parentheses correspond to highest resolution shell. b Z K = Rsym SjI hIij ShIi, where I is the observed intensity and hIi is the average intensity obtained from multiple observations of symmetry-related reflections. c Rfree was calculated as for Rcryst with the 5% of the data omitted from structural refinement. d Z K = Rcryst SjFoj jFcj SjFoj, where jFoj and jFcj are the observed and calculated structure factor amplitudes, respectively. Cyanobacterial Toxin Recognition 119 equivalent volume of mother liquor, that consisted References of 2.3 M lithium sulfate, 100 mM Tris–HCl (pH 8.0), 2% (w/v) polyethylene glycol 400 and 10 mM 1. Carmichael, W. W. (1994). Toxins of cyanobacteria. Sci. b-mercaptoethanol. Both complexes crystallized in Am. 270, 78–86. 2. Yoshizawa, S., Matsushima, R., Watanabe, M. F., the space group P42212 with one complex per asymmetric unit. The crystal data for both Harada, K., Ichihara, A., Carmichael, W. W. & complexes are given in Table 1. Fujiki, H. (1990). Inhibition of protein phosphatases by microcystis and nodularin associated with hepato- toxicity. J. Cancer Res. Clin. Oncol. 116, 609–614. 3. Honkanen, R. E., Zwiller, J., Moore, R. E., Daily, S. L., Data collection, structure determination Khatra, B. S., Dukelow, M. & Boynton, A. L. (1990). and refinement Characterization of microcystin-LR, a potent inhibitor of type-1 and type-2a protein phosphatases. J. Biol. The X-ray diffraction data for both complexes were Chem. 265, 19401–19404. collected at Beamline 8.3.1 at the Advanced Light 4. MacKintosh, C., Beattie, K. A., Klumpp, S., Cohen, P. Source in Berkeley, California on an ADSC Q210 & Codd, G. A. (1990). Cyanobacterial microcystin-LR ˚ is a potent and specific inhibitor of protein phospha- detector. All data were collected at 1 A wavelength. tases 1 and 2A from both mammals and higher plants. Thedatawereprocessedusingtheprograms 36 37 FEBS Letters, 264, 187–192. MOSFLM and SCALA. Both structures were 5. Nishiwaki-Matsushima, R., Nishiwaki, S., Ohta, T., solved by molecular replacement with the program Yoshizawa, S., Suganuma, M., Harada, K. et al. (1991). AMoRe38 using the structure of PP-1c as determined Structure–function-relationships of microcystins, in the complex bound to OA25 as the starting model. liver-tumor promoters, in interaction with protein The electron densities for the protein and bound phosphatase. Jpn. J. Cancer Res. 82, 993–996. inhibitors were clear in both instances (Figure 2(c)). 6. Holmes, C. F. B. & Boland, M. P. (1993). Inhibitors of The protein-inhibitor model was fitted manually protein phosphatase-1 and phosphatase-2a. Two of using the program XtalView.39 The starting model the major serine threonine protein phosphatases involved in cellular regulation. Curr. Opin. Struct. for MOT came from the NMR structure of the 13 Biol. 3, 934–943. inhibitor and the starting model for MCLA-2H 7. Eriksson, J. E., Toivola, D., Meriluoto, J. A. O., came from the previous crystal structure of MCLR 15 Karaki, H., Han, Y. G. & Hartshorne, D. (1990). bound to PP-1c. The structural models were Hepatocyte deformation induced by cyanobacterial subjected to iterative rounds of macromolecular toxins reflects inhibition of protein phosphatases. refinement using the program REFMAC40 with a Biochem. Biophys. Res. Commun. 173, 1347–1353. maximum likelihood target. The crystallographic 8. Ohta, T., Nishiwaki, R., Yatsunami, J., Komori, A., data are listed in Table 1. The final structures Suganuma, M. & Fujiki, H. (1992). Hyperphosphor- consisted of PP-1c residues 6 through 298 and ylation of cytokeratins 8 and 18 by microcystin-LR, a complete inhibitor models for both complexes. The new liver-tumor promoter, in primary cultured rat Carcinogenesis 13 models were checked for validity using the programs hepatocytes. , , 2443–2447. 41 42 9. Craig, M., McCready, T. L., Luu, H. A., Smillie, M. A., WHATCHECK and PROCHECK. The final Dubord, P. & Holmes, C. F. B. (1993). Identification PP1:MOT structure had 97.3% of residues in allowed and characterization of hydrophobic microcystins in Ramachandran plot regions with a G-factor of K0.2. Canadian fresh-water cyanobacteria. Toxicon, 31, The final PP1:MCLA-2H structure had 97.7% of 1541–1549. residues in allowed Ramachandran plot regions 10. Rinehart, K. L., Harada, K., Namikoshi, M., Chen, C., with a G-factor of K0.1. Harvis, C. A. et al. (1988). Nodularin, microcystin, and the configuration of adda. J. Am. Chem. Soc. 110, 8557–8558. Protein Data Bank accession codes 11. Ohta, T., Nishiwaki, R., Suganuma, M., Yatsunami, J., Komori, A., Okabe, S. et al. (1993). Significance of the cyanobacterial cyclic peptide toxins, the microcystins The atomic coordinates and structure factors and nodularin, in liver-cancer. Mutat. Res. 292, have been deposited in the RCSB Protein Data 286–287. Bank with accession codes 2BCD (PP-1:MOT) and 12. Desilva, E. D., Williams, D. E., Andersen, R. J., 2BDX (PP-1:MCLA-2H). Klix, H., Holmes, C. F. B. & Allen, T. M. (1992). Motuporin, a potent protein phosphatase inhibitor isolated from the Papua-New-Guinea sponge Theonella swinhoei Gray. Tetrahedron Letters, 33, 1561–1564. 13. Bagu, J. R., Sonnichsen, F. D., Williams, D., Acknowledgements Andersen, R. J., Sykes, B. D. & Holmes, C. F. B. (1995). Comparison of the solution structures of J.T.M. is a Pfizer MD/PhD Student and M.N.G.J. microcystin-LR and motuporin. Nature Struct. Biol. 2, 114–116. holds a Canada Research Chair in Protein Structure 14. Annila, A., Lehtimaki, J., Mattila, K., Eriksson, J. E., and Function. Supported by Canadian Institutes Sivonen, K., Rantala, T. T. & Drakenberg, T. (1996). of Health Research operating grants to C.F.B.H. Solution structure of nodularin. An inhibitor of and M.N.G.J. C.F.B.H. is an Alberta Heritage serine/threonine-specific protein phosphatases. Foundation for Medical Research Scientist. J. Biol. Chem. 271, 16695–16702. 120 Cyanobacterial Toxin Recognition

15. Goldberg, J., Huang, H. B., Kwon, Y. G., Greengard, P., 29. Maynes, J. T., Perreault, K. R., Cherney, M. M., Nairn, A. C. & Kuriyan, J. (1995). Three-dimensional Luu, H. A., James, M. N. G. & Holmes, C. F. B. structure of the catalytic subunit of protein serine/ (2004). Crystal structure and mutagenesis of a protein threonine phosphatase-1. Nature, 376, 745–753. phosphatase-1: calcineurin hybrid elucidate the role 16. Holmes, C. F. B., Maynes, J. T., Perreault, K. R., of the beta 12-beta 13 loop in inhibitor binding. J. Biol. Dawson, J. F. & James, M. N. G. (2002). Molecular Chem. 279, 43198–43206. enzymology underlying regulation of protein phos- 30. Moorhead, G., MacKintosh, R. W., Morrice, N., phatase-1 by natural toxins. Curr. Med. Chem. 9, Gallagher, T. & MacKintosh, C. (1994). Purification 1981–1989. of type 1 protein (serine/threonine) phosphatases by 17. Craig, M., Luu, H. A., McCready, T. L., Williams, D., microcystin-Sepharose affinity chromatography. FEBS Andersen, R. J. & Holmes, C. F. B. (1996). Molecular Letters, 356, 46–50. mechanisms underlying he interaction of motuporin 31. Kondo, F., Ikai, Y., Oka, H., Okumura, M., Ishikawa, N., and microcystins with type-1 and type-2A protein Harada, K. et al. (1992). Formation, characterization, phosphatases. Biochem. Cell Biol. 74, 569–578. and toxicity of the glutathione and cysteine conjugates 18. MacKintosh, R. W., Dalby, K. N., Campbell, D. G., of toxic heptapeptide microcystins. Chem. Res. Toxicol. 5, Cohen, P. T., Cohen, P. & MacKintosh, C. (1995). The 591–596. cyanobacterial toxin microcystin binds covalently to 32. Williams, D. E., Craig, M., Dawe, S. C., Kent, M. L., cysteine-273 on protein phosphatase 1. FEBS Letters, Holmes, C. F. B. & Andersen, R. J. (1997). Evidence for 371, 236–240. a covalently bound form of microcystin-LR in salmon 19. Runnegar, M., Berndt, N., Kong, S. M., Lee, E. Y. C. & liver and dungeness crab larvae. Chem. Res. Toxicol. 10, Zhang, L. F. (1995). In vivo and in vitro binding of 1293. microcystin to protein phosphatase-1 and phospha- 33. Williams, D. E., Dawe, S. C., Kent, M. L., Andersen, R. J., tase-2a. Biochem. Biophys. Res. Commun. 216, 162–169. Craig, M. & Holmes, C. F. (1997). Bioaccumulation and 20. Dawson, J. F. & Holmes, C. F. B. (1999). Molecular clearance of microcystins from salt water mussels, mechanisms underlying inhibition of protein phos- Mytilus edulis,andin vivo evidence for covalently bound 35 phatases by marine toxins. Front. Biosci. 4, D646–D658. microcystins in mussel tissues. Toxicon, , 1617–1625. 34. Fischer, W. J., Hitzfeld, B. C., Tencalla, F., Eriksson, J. E., 21. McCluskey, A., Sim, A. T. R. & Sakoff, J. A. (2002). Mikhailov, A. & Dietrich, D. R. (2000). Microcystin-LR Serine-threonine protein phosphatase inhibitors: toxicodynamics, induced pathology, and immunohis- development of potential therapeutic strategies. tochemical localization in livers of blue-green algae J. Med. Chem. 45, 1151–1175. exposed rainbow trout (oncorhynchus mykiss). Toxicol. 22. Egloff, M. P., Cohen, P. T., Reinemer, P. & Barford, D. Sci. 54, 365–373. (1995). Crystal structure of the catalytic subunit of 35. Hastie, C. J., Borthwick, E. B., Morrison, L. F., human protein phosphatase 1 and its complex with Codd, G. A. & Cohen, P. T. (2005). Inhibition of tungstate. J. Mol. Biol. 254, 942–959. several protein phosphatases by a non-covalently 23. Barford, D. (1996). Molecular mechanisms of the interacting microcystin and a novel cyanobacterial protein serine threonine phosphatases. Trends Bio- peptide, nostocyclin. Biochim. Biophys. Acta, 18, 18. chem. Sci. 21, 407–412. 36. Powell, H. R. (1999). The Rossmann Fourier auto- 24. Kita, A., Matsunaga, S., Takai, A., Kataiwa, H., indexing algorithm in MOSFLM. Acta Crystallog. sect. Wakimoto, T., Fusetani, N. et al. (2002). Crystal D, 55, 1690–1695. structure of the complex between calyculin A and 37. Bailey, S. (1994). The CCP4 suite—programs for the catalytic subunit of protein phosphatase 1. protein crystallography. Acta Crystallog. sect. D, 50, Structure, 10, 715–724. 760–763. 25. Maynes, J. T., Bateman, K. S., Cherney, M. M., 38. Navaza, J. & Saludjian, P. (1997). AMoRe: an Das, A. K., Luu, H. A., Holmes, C. F. B. & James, automated molecular replacement program package M. N. G. (2001). Crystal structure of the tumor- Macromolecular Crystallography, Part A, vol. 276 pp. promoter okadaic acid bound to protein phosphatase- 581–594. 1. J. Biol. Chem. 276, 44078–44082. 39. McRee, D. E. (1999). XtalView Xfit—a versatile 26. Zhang, L., Zhang, Z., Long, F. & Lee, E. Y. (1996). program for manipulating atomic coordinates and Tyrosine-272 is involved in the inhibition of protein electron density. J. Struct. Biol. 125, 156–165. phosphatase-1 by multiple toxins. Biochemistry, 35, 40. Murshudov, G. N., Vagin, A. A. & Dodson, E. J. (1997). 1606–1611. Refinement of macromolecular structures by the 27. Burgi, H. B., Dunitz, J. D. & Shefter, E. (1973). maximum-likelihood method. Acta Crystallog. sect. Geometrical reaction coordinates. 2. Nucleophilic D, 53, 240–255. addition to a carbonyl group. J. Am. Chem. Soc. 95, 41. Laskowski, R. A., Moss, D. S. & Thornton, J. M. (1993). 5065–5067. Main-chain bond lengths and bond angles in protein 28. Drahl, C., Cravatt, B. F. & Sorensen, E. J. (2005). structures. J. Mol. Biol. 231, 1049–1067. Protein-reactive natural products. Angew Chem. Int. 42. Hooft, R. W., Vriend, G., Sander, C. & Abola, E. E. Ed. Engl. 44, 5788–5809. (1996). Errors in protein structures. Nature, 381, 272.

Edited by M. Guss

(Received 1 September 2005; received in revised form 6 November 2005; accepted 7 November 2005) Available online 22 November 2005