A trapped human PPM1A–phosphopeptide complex reveals structural features critical for regulation of PPM protein phosphatase activity

Metal-dependent protein phosphatases (PPM) are evolutionarily unrelated to other serine/threonine protein phosphatases and are characterized by their requirement for supplementation with millimolar concentrations of Mg2+ or Mn2+ ions for activity in vitro. The crystal structure of human PPM1A (also known as PP2Cα), the first PPM structure determined, displays two tightly bound Mn2+ ions in the active site and a small subdomain, termed the Flap, located adjacent to the active site. Some recent crystal structures of bacterial or plant PPM phosphatases have disclosed two tightly bound metal ions and an additional third metal ion in the active site. Here, the crystal structure of the catalytic domain of human PPM1A, PPM1Acat, complexed with a cyclic phosphopeptide, c(MpSIpYVA), a cyclized variant of the activation loop of p38 MAPK (a physiological substrate of PPM1A), revealed three metal ions in the active site. The PPM1Acat D146E–c(MpSIpYVA) complex confirmed the presence of the anticipated third metal ion in the active site of metazoan PPM phosphatases. Biophysical and computational methods suggested that complex formation results in a slightly more compact solution conformation through reduced conformational flexibility of the Flap subdomain. We also observed that the position of the substrate in the active site allows solvent access to the labile third metal-binding site. Enzyme kinetics of PPM1Acat toward a phosphopeptide substrate supported a random-order, bi-substrate mechanism, with substantial interaction between the bound substrate and the labile metal ion. This work illuminates the structural and thermodynamic basis of an innate mechanism regulating the activity of PPM phosphatases.

Reversible protein phosphorylation signaling pathways are shaped by opposing actions of protein kinases and phosphatases. These pathways regulate the response of cells and organisms to changing environmental and physiological circumstances; dysregulation of these pathways underlies many human diseases. Although phosphorylated protein residues are dominated by phosphoserine (pSer) and phosphothreonine (pThr), with a minimal proportion of phosphotyrosine (pTyr), the numbers of phosphatases involved in reversing the pSer/ pThr and pTyr modifications are more evenly divided (1)(2)(3). Analysis of global dynamics of pSer/pThr and pTyr-based signaling suggests distinct patterns of regulation (4). The phosphoprotein phosphatases (PPP), 3 with 13 members, and the metal-dependent phosphatases (PPM), with 18 members, provide most of the serine/threonine protein phosphatase activity in human cells (5,6). The activity, substrate specificity, and subcellular localization of PPP are regulated primarily by binding of regulatory or inhibitor subunits, of which over 100 have been identified in human cells (7,8). PPM phosphatases generally are monomeric, but regulation of their activity remains incompletely understood (9 -11). The evolutionarily distinct PPP and PPM phosphatases both feature tightly-bound bimetal clusters in their active sites, but only the PPM phosphatases require supplementation with millimolar concentrations of Mg 2ϩ or Mn 2ϩ salts to exhibit detectable activity in vitro, suggesting that a weakly-bound ion is necessary for activity (5,6).
Existing structures of full-length human PPM1A were determined from crystals formed at pH 5.0 to 5.5 and exhibit two closely-spaced Mn 2ϩ ions in the active site (12,21). The PPM1A catalytic domain forms an ␣␤␤␣ sandwich, with a peculiar structure termed the Flap positioned at one end of the active site; a C-terminal extension comprising helices ␣7 to ␣9 is specific to PPM1A/PPM1B homologues (12,22). The Flap structure has been proposed to increase specificity through interactions with substrates (23). The bi-metal center features an indirectly coordinated phosphate ion and a bridging water/ hydroxide ion that is proposed to function as the nucleophile for the S n 2 hydrolase reaction (12,21). As these Mn 2ϩ ions have sub-micromolar affinities, their presence does not explain the requirement for supplementation with millimolar concentrations of Mg 2ϩ or Mn 2ϩ ions (5,6). An additional divalent ion, M3, which is displayed as a third ion present in the active sites of some crystal structures of bacterial or plant PPM phosphatases, has been proposed as the weakly-bound, catalytically essential metal ion (24 -28). As most M3-coordinating residues are highly conserved, the involvement of M3 in metazoan PPM catalysis was anticipated (24, 26 -28). Recently, we used isothermal titration calorimetry (ITC) to detect millimolar-affinity binding of Mg 2ϩ to PPM1A (29). Site-directed mutagenesis of Asp-146 or Asp-243 was used to distinguish two mutually exclusive subsites for the third metal ion: the Asp-146/Asp-239 subsite, which supports catalysis, and the Asp-239/Asp-243 subsite, which increases metal and substrate affinities but is not essential for catalysis (29). Asp-239 coordinates M1 in addition to its involvement in both M3 subsites; Asp-239 mutants are inactive. We used hydrogen-deuterium exchange-MS (HDX-MS) to investigate the conformational mobility of PPM1A in the presence of low or moderate magnesium ion concentrations (30). We found that the binding of Mg 2ϩ to the catalytically essential Asp-146/Asp-239 subsite restricted conformational mobility of the active site and specific regions of the Flap subdomain, suggesting that the third metal ion stabilizes the native structure of the enzyme (30). These studies support the idea that conformational stabilization of the Flap, which is critical for specific substrate recognition, may be coupled to catalytic activity through the binding of the third metal ion.

Substrate binding by inactive mutants of PPM1A
Substrates of PPM1A exhibit few commonalities in flanking sequences (Fig. 1A) (14,15,18,31). Apart from the prevalence of hydrophobic residues in the ϩ1 position, the targeted pSer/ pThr residues are flanked by an assortment of charged, polar, and hydrophobic residues without a unitary pattern. To gain insight into structural aspects of substrate recognition and the involvement of the third metal ion in human PPM1A catalytic activity, we sought conditions for co-crystallization of an inactive mutant of PPM1A in complex with a phosphopeptide substrate under physiological pH and in the presence of Mg 2ϩ ions. Trapping of substrate phosphoproteins by inactive phosphatases has been used to isolate phosphatase-substrates complexes (29,32,33). PPM1A D146A is catalytically inactive (29,34), a defect that may result from reduced substrate binding and/or disruption of catalysis. Hence, we investigated the binding of fluorescein-labeled p38(175-185, 180pT) to PPM1A D146A by fluorescence anisotropy (Fig. 1B). The inactive mutant binds the labeled phosphopeptide with K d ϭ 48 Ϯ 16 M, thus demonstrating that the D146A mutation does not abolish substrate binding. Michaelis constants, determined from the substrate concentration dependence of the initial rates, provide an estimate of the strength of substrate binding. The K d value for the labeled phosphopeptide binding to PPM1A D146A is 1.8-fold weaker than the reported K m value of PPM1A for unlabeled phosphopeptide (27 Ϯ 4 M) and is nearly equal to the corresponding K m value of PPM1A D243A (53 Ϯ 9 M) (29). Note that although PPM1A D243A is catalytically active, it exhibits reduced affinity for the catalytically-essential metal ion and a higher K m for the substrate (29).
We further investigated the binding of the phosphopeptide to PPM1A D146A by NMR spectrometry chemical shift perturbations (Fig. 1C). TOCSY spectra of p38(175-185, pThr-180) alone (blue contours) or in the presence of full-length PPM1A D146A (10:1 molar ratio, red contours) showed significant chemical shift changes of the pThr-180 and Gly-181 peaks, consistent with conformational changes of the modified and adjacent residues. A screen for crystallization conditions for the PPM1A D146A-p38(175-185, 180pT) complex failed to produce co-crystals.
The recombinant PPM1A protein used for these experiments retained four linker residues (SGGT) and the position 1 methionine following removal of the N-terminal His 6 tag (29). As the initiating methionine is absent in mature, endogenous PPM1A and existing crystal structures begin with Gly-2, the additional amino acids may interfere with crystallization. Moreover, HDX-MS experiments suggested increased conformational mobility of PPM1A D146A compared with the WT enzyme (30). To address these concerns, we reverted the N terminus to the endogenous form, introduced the more conservative Asp to Glu mutation at position 146, and removed helices ␣7, ␣8, and ␣9 to produce the WT PPM1A cat and PPM1A cat D146E variants of the catalytic domain (aa 2-297). As expected, PPM1A cat exhibited nearly the same activity as had been reported for His 6 tag-free, full-length PPM1A, both for a phosphopeptide substrate, p38(175-185, 180pT), Trapped PPM1A-phosphopeptide complex and a nonphysiological substrate, para-nitrophenyl phosphate (pNPP) ( Table 1) (29,34). PPM1A cat D146E did not exhibit detectable activity for either substrate. The Motif 5 aspartate residue that aligns to PPM1A Asp-146 is highly conserved and intolerant of substitution, supporting its proposed role in coordinating a metal ion essential for catalysis (29, 34 -37). We sought a tighter-binding substrate to increase the stability of potential protein-phosphopeptide complexes. Based on the p38␣ activation loop sequence, the cyclic peptide c(MpSIpYVA) (Fig. 1D) had been developed as a substratecompetitive inhibitor of the human PPM paralogue PPM1D; interestingly, it is a substrate of PPM1A, with a K m of 11 M (38). The shorter enzyme, PPM1A cat , also was active toward the cyclic peptide with a K m of 5.9 Ϯ 1.4 M, whereas PPM1A cat D146E did not exhibit measurable activity (Table 1). We used ITC to analyze the binding of c(MpSIpYVA) to PPM1A cat D146E (Fig. 1E). After subtracting heats of dilution, integrated injection heats were fitted to a single-site binding model, yield- substrates. Target phosphoserine/phosphothreonine residues are indicated by red lowercase letters. Classification of flanking residues as hydrophobic (green), polar (lavender), positively-charged (blue), or negatively-charged (pink) is indicated by colored boxes. Lowercase letters indicate proximal sites of phosphorylation. B, fluorescence anisotropy titration of fluorescein-labeled p38(175-185, 180pT) with PPM1A D146A at 25°C. Data represent the mean Ϯ S.E. of four titrations and were fit to a single-site binding model with K d ϭ 48 Ϯ 16 M. C, NMR chemical shift perturbation study of the binding of p38(175-185, 180pT) to PPM1A D146A at 30°C. TOSCY spectra of p38(175-185, 180pT) peptide alone (blue contours) or in the presence of PPM1A D146A (red contours, 10:1 molar ratio). D, primary structure of the cyclic thioether isopeptide, c(MpSIpYVA). E, isothermal titration calorimetry analysis of the binding of c(MpSIpYVA) to PPM1A cat D146E at 25°C. Left panel, representative thermal trace for injection of c(MpSIpYVA) into PPM1A cat D146E solution is shown. Right panel, integrated heats of injection for the binding of c(MpSIpYVA) to PPM1A cat D146E were fit to a single-site binding model after subtraction of peptide heats of dilution. Data represent the mean Ϯ S.E. of two titrations.

Crystal structure of PPM1A cat D146E in complex with the cyclic peptide c(MpSIpYVA)
High quality monoclinic crystals of the PPM1A cat D146Ec(MpSIpYVA) complex were obtained from a pH 7.3 solution containing Ca 2ϩ ions ( Table 2). Refinement of the structure to 2.2 Å resolution revealed an asymmetric unit comprising three similar protein-peptide complexes related by two pseudo-rotation axes. Consistent with structures of full-length PPM1A (12,21), the truncated protein forms the characteristic PPM fold with the metal ion-containing active site embedded in the top edge of the opposed ␤-sheets flanked by ␣-helices ( Fig. 2A). The Flap subdomain is positioned alongside the active site. The mutated residue, D146E, is located at the apex of the ␤6-␤7 loop. The three complexes in the asymmetric unit exhibit moderate conformational variability, especially in flexible regions of the protein that exhibit high crystallographic B values (Fig. S1, A and B).
Importantly, the active site contains three Ca 2ϩ ions and the bound cyclic peptide substrate. Two of the metal ions occupy positions corresponding to those of the two Mn 2ϩ ions found in structures of phosphate-bound, full-length PPM1A obtained at pH 5.0 -5.5 (12,21). Specifically, M1 is coordinated by Asp-60, Asp-239, and Asp-282, with additional waters of coordination, and M2 is coordinated by Asp-60, Gly-61, and one of the phosphate oxygens of the cyclic peptide phosphoserine residue, with additional waters of coordination (Fig. 2B). Importantly, the bridging water/hydroxide, which has been proposed to serve as the nucleophile for the S n 2 reaction, is jointly coordinated by M1 and M2, as in the structures of full-length PPM1A (12,21). The third metal ion is bound between the side chains of Asp-239 and Asp-243, one of the two proposed subsites for M3 (29). Based on the effects of specific mutations, the M3 Asp-239/ Asp-243 subsite contributes to substrate affinity, whereas the M3 Asp-146/Asp-239 subsite is essential for catalytic activity (29). The cyclic peptide sits in the active site with the phosphoserine positioned asymmetrically over the M1-M2 cluster and with the phosphotyrosine side chain proximal to the PPM1A N-terminal residue, Gly-2 (Fig. 2B). The three phosphoserine phosphate oxygens each can form a hydrogen bond with an M1-or M2-bound water molecule, supporting the importance of the M1/M2 cluster for charge stabilization of the transition state (12). As an S n 2 reaction, the phosphomonoester hydrolase reaction proceeds with inversion at the phosphorus atom (6,39). Importantly, the position of the incipient leaving group, pSer-2 O␥, opposite the M1/M2 bridging water molecule/hydroxide ion provides structural support for the role of the latter as the S n 2 nucleophile. Additional hydrogen bonds between the Arg-186 side chain and the cyclic peptide Met-1 and pSer-2 backbone carbonyl groups stabilize the complex. Unlike the compact solution conformation of unbound c(MpSIpYVA) (38), the bound peptide exhibits an open conformation stabilized by intra-backbone hydrogen bonds (Fig. S1C). Except for the C␦ and O⑀ atoms of the mutated D146E residue, positions of the metal ions with coordinating waters, the cyclic peptide, and active-site residues are well-defined by localized electron density (Fig. 2C). The loss of defined conformation for the  Trapped PPM1A-phosphopeptide complex mutated side chain is consistent with the importance of the Asp-146 residue for activity. Although mutation of Asp to Glu is regarded as a conservative substitution, in the context of the PPM-active site, the additional methylene group displaces the O⑀ atoms from the proper distance to allow coordination of a divalent metal ion jointly with an Asp-239 carboxylate oxygen.
Representation of Coulombic potential in the active site shows the three divalent ions embedded in the highly negative field produced by active-site aspartic acid ␦O atoms and surmounted by the positively-charged side chains of Lys-165, Arg-186, and Arg-281 (Fig. 2D). Notable hydrophobic contacts between the protein and cyclic peptide are found between Flap Hydrogen bonds are indicated in magenta. The bridging water molecule/hydroxide ion (asterisk) is jointly coordinated by M1 and M2. M1 is also coordinated by Asp-60, Asp-239, Asp-282, and two water molecules. M2 is also coordinated by Asp-60, Gly-61, one of the phosphate oxygens of the cyclic peptide phosphoserine residue, and two water molecules. M3 is coordinated by Asp-239, Asp-243, and four water molecules. C, simulated, annealed composite omit electron density map contoured at 1 (gray mesh) overlaid on stick representation with coloring as in A. Localized electron density for the C␦ and O⑀ atoms of the D146E mutated residue is absent. D, surface representation of electrostatic potential (coulombic) of PPM1A cat (semi-transparent surface, 10 kcal/mol⅐e, blue; Ϫ10 kcal/mol⅐e, red) in complex with the cyclic peptide (stick representation, sandy brown). E, surface representation of residue mean hydrophobicity according to the Kyte-Doolittle hydrophobicity scale (4.5, hydrophobic, orange; Ϫ4.5, polar, blue).

Trapped PPM1A-phosphopeptide complex
residues Ile-184 and Leu-191 with cyclic peptide residues Val-5 and Ile-3, respectively, and between PPM1A residues Ala-3 and Phe-4 with cyclic peptide residues pTyr-4 and Met-1, respectively (Fig. 2E). Calcium ions inhibit PPM1A activity, although the mechanism of inhibition has not been established (29,40). Coordination geometries of the metal ions are altered in the calciumcontaining PPM1A cat D146E-c(MpSIpYVA) complex (Fig.  S1D). In existing structures of full-length PPM1A, both Mn 2ϩ ions exhibit octahedral coordination (12,21). For all three complexes in the asymmetric unit, M2 exhibits octahedral geometry, but the sixth coordination position is occupied by a phosphate oxygen of the cyclic peptide phosphoserine. The best-fitting coordination geometry for M1 varies among the three complexes in the asymmetric unit. M1 is coordinated by side-chain carbonyl oxygens of Asp-60, Asp-239, and Asp-282, with three or four water molecules occupying the remaining positions for trigonal prism, square face monocapped (chain A, n ϭ 7, r.m.s.d. ϭ 0.312), pentagonal bipyramid (chain B, n ϭ 7, r.m.s.d. ϭ 0.383) or trigonal prism, and square face bicapped (chain C, n ϭ 8, r.m.s.d. ϭ 0.635) geometries. Importantly, the bridging water/hydroxide ion, which has been proposed to serve as the nucleophile for the S n 2 reaction, is positioned similarly in the three chains. The weakly-bound M3 is coordinated by Asp-239, one or both side-chain oxygens of Asp-243, and water molecules for the remaining positions of octahedral or face capped octahedral coordination geometry.

Comparison of phosphate-bound and cyclic peptide-bound forms of the PPM1A catalytic domain
The structures of the PPM1A cat D146E-c(MpSIpYVA) complex and the catalytic domain portion (aa 2-297) of phosphate-bound, full-length (aa 2-382) PPM1A are similar, with substantial conformational differences evident for the N termini, the ␤3-␤4 loops, the ␣2-␤5 bridge structures, and the protruding points of the Flap (Fig. 3A). Within the active site, smaller conformational changes reflect specific interactions with the substrate (Fig. 3B). A shift in the position of the N terminus permits hydrogen bonding between the N-terminal hydrogens and the phosphotyrosine O oxygen. In the fulllength protein, the side chain of Arg-33 protrudes into the active site, allowing hydrogen bonding or electrostatic interactions between both N⑀ hydrogens and two of the phosphate oxygens (12). In the complex, the Arg-33 side chain is retained at the edge of the active site by a hydrogen bond between N⑀(H) and the protein Ala-3 backbone carbonyl oxygen. The side chains of Flap residues Ile-184 and Gln-85 are shifted ϳ3 Å toward the bound substrate, compared with their positions in the phosphate-bound, full-length protein. In the complex, the N⑀1 and N⑀2 of Arg-186 form hydrogen bonds with the cyclic peptide Met-1 and pSer-2 backbone carbonyl oxygens.
Comparative studies of protein sequences among homologues may allow discrimination between regions of the protein that are functionally constrained from those that are conformationally permissive. We compared the locations of conformational differences between the structures of the free protein and the cyclic peptide-protein complex with locations of sequence variability or insertions and deletions among 150 PPM1A homologues (Fig. 3C). Interestingly, the N-terminal amino acids and the Flap domain, which showed substantial differences in conformation between the free and bound protein, exhibit low sequence variability, suggesting the conformational differences may have functional significance. The ␣2-␤5 bridge, also located on the top of the protein, exhibited substantial differences in conformation but exhibited a moderate level of sequence variability. The ␤3-␤4 and ␣1a-␣2 loops, located on the bottom of the protein, exhibited substantial differences in conformation and large variations in sequence, suggesting tolerance for structural variation. Insertions or deletions were tolerated only in the ␣2-␤5 bridge and the ␤3-␤4 and ␣1a-␣2 loops, regions with high to moderate sequence variation. The ␤5-␤6 and ␤7-␤8 loops, also located on the bottom of the protein, exhibited substantial sequence variation but did not show conformational changes between the phosphopeptide-bound and phosphate-bound forms. The structural comparison of the free and substrate-bound forms of the PPM1A catalytic domain and the analysis of sequence variation among metazoan homologues support the importance of the first few N-terminal residues, the ␣2-␤5 bridge, and the Flap subdomain in interactions between the enzyme and substrate. Recently, crystal structures of the Arabidopsis thaliana PP2C phosphatase TAP38/PPH1 in the free state and as the D180E mutant in complex with a phosphothreonine-peptide substrate were described (37). Similar to conformational changes of PPM1A residues Arg-33 and Arg-186 described above, the conformations of the TAP38/PPH1 residues Arg-69 and Arg-225 are altered upon binding of the substrate (37).

Small-angle X-ray scattering and molecular dynamics
Recently, we used HDX-MS to characterize the conformational mobility of full-length PPM1A and PPM1A D146A under conditions of low (0.1 mM) and moderate (2 mM) Mg 2ϩ concentrations (30). The Flap subdomain was identified as a conformationally mobile region in which amide hydrogen-deuterium exchange was affected both by Mg 2ϩ concentrations and the identity of the residue at position 146. The Flap subdomain of bacterial PP2C phosphatases had been characterized as flexible based on its variable conformations in crystals where its conformation may be affected by interactions with neighboring molecules (26 -28). To gain additional understanding of the effects of the cyclic peptide binding on solution conformations of PPM1A, we used small angle X-ray scattering (SAXS) and molecular dynamics (MD) simulations. SAXS scattering profiles for PPM1A cat and PPM1A cat D146E are indistinguishable (Fig. S2A), indicating that the D146E mutation does not induce detectable changes in the solution conformation of the phosphatase under these conditions. Analysis of scattering profiles of the free and cyclic peptide-bound forms of PPM1A cat D146E (Fig. 4A, cyan and pink dots, respectively, and Fig. S2B) indicates that the radius of gyration of the complex (R g , ϭ 21.07 Ϯ 0.06 Å) is slightly reduced compared with that of the free protein (R g , ϭ 21.51 Ϯ 0.06 Å). Scattering profiles calculated from the crystal structures of phosphate-bound (residues 2-297 of PDB code 1A6Q) PPM1A and cyclic peptide-bound (present work, chains A and D) PPM1A cat D146E produced accurate fits to the scattering data ( ϭ 1.18 for the free protein and ϭ 1.53 Trapped PPM1A-phosphopeptide complex for the complex) (Fig. S2C). We used constrained MD simulations to generate conformational ensembles for free and cyclic peptide-bound PPM1A cat D146E proteins. Scattering profiles predicted from the constrained MD conformational ensembles provided slight improvements of the fits to the scattering data for the free protein and the complex (Fig. 4A, blue ( ϭ 1.12) and red ( ϭ 1.47) curves, respectively). The MD simulations also suggest that the Flap subdomain is more flexible in the free protein than in the protein-cyclic peptide complex. Analysis of conformational variation across the ensembles for the free protein (blue curve) and the protein-cyclic peptide complex (red curve) identifies regions of differential variability (Fig. 4B). Specifically, the MD simulations indicate reduced conformational variability in the ␣2-␤5 bridge and the Flap subdomain in the complex, compared with the free protein. The reduction in Flap subdomain conformational variability upon complex formation is consistent with the observed small reduction in R g upon complex formation.

Trapped PPM1A-phosphopeptide complex
To gain additional insight into the effects of cyclic peptide binding and the D146E mutation on conformational variability, we also performed MD simulations on the WT protein, without and with bound cyclic peptide, and we compared these to the corresponding states of the D146E protein. Overlays of snapshots spanning the simulation interval for the WT protein, in free and cyclic peptide-bound forms (Fig. S3, A and B, respectively), and the D146E mutant protein, in free and cyclic peptide-bound forms (Fig. S3, C and D, respectively), illustrate conformational differences. In these simulations, M3 is bound to the Asp-146/Asp-239 subsite for the WT enzyme and to the Asp-239 /Asp-243 subsite for the D146E mutant (Fig. 4C). In all four states, the weakly bound M3 ion exhibits greater positional variation than the tightly bound M1 and M2 ions (Fig. 4C). Among the four states, the positional variation of M3 is the largest for the WT enzyme in the unbound state (Fig. 4C, upper  left). Conversely, the Arg-33, His-62, and Arg-186 side chains exhibit the largest conformational variation for the D146E protein in the unbound state (Fig. 4C, lower left). Interestingly, for both the WT enzyme and the D146E protein, the Arg-33, His- Trapped PPM1A-phosphopeptide complex 62, and Arg-186 side chains exhibit reduced conformational variability in the cyclic peptide-bound state compared with the free state (Fig. 4C, upper right and lower right panels, respectively). To provide a quantitative measure of the conformational variability of the Flap, we analyzed the distributions for the M1 to Arg-186 C␣ distance and the M2 to Arg-186 C␣ distance for the four MD conditions (Fig. 4D, upper and lower panels, respectively). As noted above, the positions of the tightly bound M1 and M2 ions show little variation across the simulation and provide relatively fixed reference points in the active site, whereas the C␣ of Arg-186 serves as an indicator of the position of the Flap. For both the WT and D146E proteins, the free protein exhibited considerably greater conformational variation compared with the respective bound states. The distributions for the free WT protein (Fig. 4D, blue curves) contain two peaks and are broader than those of the D146E protein (green curves). The distance distribution for the PPM1A cat -cyclic peptide complex (Fig. 4D, orange curve) is similar to that of the PPM1A cat D146E-cyclic peptide complex (red curve) but additionally contains a small peak with substantially larger distances. In the minor population, the Arg-186cyclic peptide hydrogen bond is disrupted, and the Flap is positioned farther from the active site. Mutation of PPM1A Asp-146 to either alanine or glutamic acid abolishes catalytic activity, implying that divalent ion occupancy of the M3 Asp-146/Asp-239 subsite is essential for catalytic activity. The simulations suggest that the binding of M3 to the Asp-146/Asp-239 subsite supports a conformationally restricted substrate-binding conformation but also supports a minor population featuring a more distant flap conformation.

Catalytically-essential metal ion and the substrate bind in either order
An important aspect of the PPM1A cat D146E-c(MpSIpYVA) complex structure is the combined presence of a substrate and a third metal ion in the active site. This structure provides a context for investigating the interactions between binding of the substrate and binding of the third metal ion. A surface representation of the complex suggests that M3 bound in the Asp-239/Asp-243 subsite is accessible to the solvent, implying that the weakly bound ion can dissociate and re-associate in the presence of bound substrate (Fig. 5A). Moreover, the structure suggests that the catalytically essential Asp-146/Asp-239 subsite also is accessible to mobile ions. These observations suggest that the random-order, bi-substrate mechanism, in which the substrate (S) and the essential metal ion (M) may bind in either order, is the appropriate model (Fig. 5B). Note that for the purposes of kinetic mechanism, the essential metal ion refers to the weakly-bound ion provided by supplementation; the M1 and M2 ions, which also are necessary for catalysis but remain fully bound under normal conditions, are formally considered as intrinsic components of the free enzyme. The ordered sequential mechanism, an alternative mechanism, was proposed on the basis of product inhibition studies (40). In the ordered sequential mechanism, the obligatory initial binding of the metal ion is followed by binding of the substrate to form the catalytically competent ternary complex; after the catalytic step, the alcohol, phosphate ion, and metal ion are released sequentially (40). Evidence against the sequential mechanism is provided by the structure of a complex of an A. thaliana PP2C phosphatase with a trapped phosphopeptide substrate that contains two metal ions in the active site (37) and by an extensive mutagenesis study of a bacterial PP2C phosphatase that demonstrated the requirement for an intact M1-M2 bimetal cluster for efficient substrate binding (33). To investigate interactions between binding of the substrate and binding of the third metal ion, we determined the substrate and magnesium ion concentration dependences of the initial rates for WT PPM1A cat and for PPM1A cat D243A, which retains only the Asp-146/Asp-239 subsite for M3. The substrate and magnesium ion concentration dependence of the initial rates each follow apparent Michaelis-Menten kinetics ( Fig. 5C and Fig.  S4). Analysis of the functional dependence of the apparent Michaelis constants on Mg 2ϩ or substrate concentration, as appropriate, allows estimation of the kinetic constants for the random-order bi-substrate mechanism ( Fig. 5D and Table 3). The values of k cat estimated by sequential fitting to the paired Equation 3 and Equation 4 or Equation 6 and Equation 7 agreed to better than 1%. The D243A mutation had little effect on the catalytic rate constant, in agreement with previous work (29). For both the WT and the D243A mutant enzyme, this analysis suggests substantial interaction between the binding of the weakly-bound metal ion and the binding of the phosphopeptide substrate, as indicated by the small value of ␣ (0.07-0.08). This analysis assumes that binding events involving the substrate and metal ion are in rapid equilibrium in the presence of a rate-determining catalytic step. The rapid equilibrium assumption implies that the product of equilibrium constants for the formation of the ternary enzyme⅐metal⅐substrate (EMS) complex via initial binding of the metal ion (K 1 ⅐K 2 ) must equal the product of equilibrium constants for its formation via initial binding of the substrate ((K 2 /␣)⅐(␣K 1 )) (Fig. 5B, upper and lower paths, respectively). Hence, the affinities of the phosphopeptide for the free enzymes (K 2 /␣) are weaker than the affinities for enzymes with M3 bound (K 2 ) by about a factor of 12. Similarly, the affinities of the essential divalent ion for the free enzymes (K 1 ) are weaker than the affinities for the enzymephosphopeptide substrate complexes (␣K 1 ) by about a factor of 12. Moreover, the values of ␣ for the WT and D243A mutant enzymes are similar. Interestingly, published data describing the Mn 2ϩ and pNPP concentration dependences of PPM1A phosphatase activity (40), when subjected to the same analysis, support the random-order, bi-substrate mechanism with ␣ ϭ 0.4 (Fig. S5). Hence, the coupling between the binding of Mg 2ϩ and that of a phosphopeptide substrate is five times stronger than the coupling between the binding of Mn 2ϩ and that of the artificial substrate pNPP.

Discussion
In this work, we have presented the structure of a trapped cyclic phosphopeptide substrate complexed with the catalytic domain of PPM1A. The enzyme is inactive due to the presence of Ca 2ϩ ions and the D146E mutation, either of which abolishes activity and enables trapping of the substrate (29,40). The cyclic phosphopeptide substrate, c(MpSIpYVA), is a tighterbinding variant of the p38 MAPK activation loop, a physiolog-Trapped PPM1A-phosphopeptide complex ical substrate of PPM1A (14, 38). Although the binding of millimolar-affinity Mg 2ϩ to PPM1A was detected by ITC (29) and the presence of the third metal ion in the active site was anticipated based on homology (24, 26 -28), the PPM1A cat D146Ec(MpSIpYVA) complex provides the first structural confirmation of a third metal ion in the active site of a metazoan PPM phosphatase. In the complex, the cyclic peptide serine phosphoester moiety is positioned asymmetrically over the M1-M2 cluster in the active site, with altered conformations of key PPM1A residues Arg-33 and Arg-186. The binding of c(Mp-SIpYVA) to PPM1A cat D146E is dominated by the favorable enthalpy of complex formation, consistent with small conformational adjustments in the complex compared with the free enzyme. Analysis by a combination of SAXS and MD simulations of the free and cyclic peptide-bound forms of PPM1A cat D146E revealed a slightly more compact solution Hydrolases frequently feature metal ions in their active sites (41). For PPM phosphatases, the closely-spaced M1-M2 cluster provides a highly specific binding environment for phosphate ions, pSer-and pThr-peptides, and phosphoprotein substrates (6,12,21). The requirements for specific binding of phosphoproteins are distinct from requirements for catalytic activity, as specific binding of substrates was observed by catalytically inactive mutants that preserve the M1-M2 cluster or in the presence of Ca 2ϩ ions, as reported previously (29,33,37) and in this work. The M1 and M2 ions jointly coordinate a bridging water molecule/hydroxide ion that has been proposed to function as the nucleophile in the S n 2 phosphate monoester hydrolase reaction (6,12,21). In the original crystal structure of PPM1A, the position of the bound phosphate failed to indicate unambiguously which phosphate oxygen should model the serine/threonine leaving group, and thus the identification of the S n 2 nucleophile remained tentative (12). In the complex, the position of the phosphoserine O␥, as the insipient leaving group, opposite the bridging water molecule/hydroxide ion provides structural support for its identification as the S n 2 nucleophile. The M1-M2 cluster also may contribute to catalytic activity by reducing charge accumulation in the transition state (39), a function supported by the direct coordination of serine/threonine phospho-monoester oxygen atoms by M2, as seen in the A. thaliana PHH1-trapped substrate complex (37) and in this work. Thus, the M1-M2 cluster is necessary for substrate binding, activates the water molecule/hydroxide ion to serve as the S n 2 nucleophile, and stabilizes the transition state through coordinated water molecules.
The role of the weakly-bound M3 in the PPM catalytic mechanism remains to be demonstrated. The absence of detectable enzymatic activity without supplementation with millimolar concentrations of divalent Mg 2ϩ or Mn 2ϩ is one of the defining characteristics of the PPM phosphatases (7,35). In eukaryotic and eukaryotic-like bacterial PPM phosphatases, detection of a third metal ion in the active site is correlated with the presence of the sequence GDS in motif 5, corresponding to aa 145-147 in PPM1A (6,35,36); mutation of the motif 5 aspartate residue in these phosphatases results in loss of activity (24, 26 -29, 37). The location of the Asp-146/Asp-239 subsite suggests that M3 is needed to facilitate protonation of the serine/threonine leav-ing group by positioning an activated, metal-bound water molecule near the fissile phosphate monoester bond (6,37). This role is supported by the location of the cyclic peptide phosphoserine O␤-O␥ bond in the structure of the complex near the presumed Asp-146/Asp-243 M3 subsite. In addition, this role is supported by the correlated reductions in catalytic activity and apparent metal-binding affinity (K metal ) with decreasing pH values (29).
Divalent calcium ions were found to inhibit the activity of PPM1A toward pNPP as a competitive inhibitor of Mn 2ϩ with a K i ϭ 4.45 mM (40), suggesting that catalytic inhibition by Ca 2ϩ resulted from its occupancy of the weak M3-binding site. As the Lewis acidities of Mg 2ϩ , Mn 2ϩ , and Ca 2ϩ are similar, the failure of calcium to support catalysis may result from its increased size or its propensity to adopt alternative coordination geometries (42,43). Note that in the intracellular environment, competitive inhibition of PPM phosphatase activity by Ca 2ϩ ions is considered inconsequential because intracellular calcium ion concentrations are orders of magnitude lower than the K i values.
The dependence of the initial rates on phosphopeptide substrate and Mg 2ϩ concentrations supports a random-order, bisubstrate mechanism, with substantial interaction between the binding of the substrate and the labile metal ion. This mechanism is supported by the structure of the complex, which indicates that both the Asp-146/Asp-239 and Asp-239 /Asp-243 subsites are accessible to ions in solution in the presence of bound substrate, and by solution studies suggesting that the binding of either Mg 2ϩ or substrate results in reduced conformational mobility (30). Although the substrate can bind in the absence of the weakly-bound metal ion, the binding affinity is considerably higher for the metal-bound state. Similarly, although Mg 2ϩ can bind to the substrate-free enzyme, the binding affinity is considerably higher for the substrate-bound state. As only the ternary complex is catalytically active, this mechanism suggests that the activities of PPM phosphatases can be regulated by the concentrations of both substrate and metal ions to effect negative regulation of signaling pathways. In most eukaryotic cells, the concentration of free Mg 2ϩ is less than 1 mM, with extensive competition among proteins, nucleic acids, and lipids for binding (44). In the absence of stress signaling, low concentrations of phosphoprotein substrate result in limited formation of the ternary EMS complex. Following activation of a stress-signaling pathway, the increased abundance of phosphorylated signaling protein drives formation of the active EMS ternary complex due, in part, to the increased affinity for Mg 2ϩ . In addition, stress signaling may affect intracellular metal ion concentrations through altered ion channel activity, with consequent activation of PPM phosphatase activity (45)(46)(47). The resulting high activity of the phosphatase efficiently dephosphorylates the signaling protein, resulting in decommissioning of activated stress signaling.

Expression and purification of His 6 tag-free PPM1A D146A
Expression and purification of His-TEV-PP2C␣ D146A was performed as described previously, including treatment with Table 3 Kinetic constants for random-order, bi-substrate mechanism Experiments were performed at 30°C with 0 -30 mM MgCl 2 and 0 -300 M p38␣ (175-185, 180pT). Initial rates were fitted to an apparent Michaelis-Menten equation with explicit dependence on Mg 2ϩ or substrate concentration. Trapped PPM1A-phosphopeptide complex TEV protease to remove the N-terminal His 6 tag (29). The protein was further purified by anion-exchange and gel-filtration chromatography. For anion-exchange chromatography, the protein was dialyzed against Buffer B (50 mM HEPES (pH 7.5), 2 mM ␤-mercaptoethanol, 2 mM MgCl 2 , 0.1 mM EGTA, and 10% glycerol) containing 100 mM NaCl. The protein was then loaded onto a Q-Sepharose column (GE Healthcare) that was preequilibrated with Buffer B containing 100 mM NaCl. The column was subsequently washed with the same buffer, and the protein was eluted by using a 100 -400 mM NaCl gradient in the same buffer. The pooled fractions from Q-Sepharose were concentrated. Gel filtration was performed by FPLC using Superdex-75 16/60 (GE Healthcare) in the presence of Buffer B containing 100 mM NaCl, resulting in purified, full-length PPM1A D146A. Note that this protein retained five additional amino acids (SGGTM) at the N terminus, compared with endogenous PPM1A. The protein was Ͼ95% pure as determined by SDS-PAGE followed by Coomassie Blue staining. Protein concentrations were determined by optical spectrometry (Nanodrop) using ⑀ 280 ϭ 35,410 M Ϫ1 cm Ϫ1 .

Peptide synthesis and purification
All chemicals and amino acids used were of analytical grade. The peptide p38(175-185, 180pT) (NH 2 -TDDEMpTGYVAT-COOH) and cyclic peptide c(MpSIpYVA) were synthesized by solid-phase peptide synthesis using Fmoc/tert-butyl chemistry on Wang resin and Rink amide AM resin, respectively. Phosphorylated threonine, serine, and tyrosine were incorporated as a Fmoc-Thr(PO(OBzl)OH)-OH, Fmoc-Ser(PO(OBzl)OH)-OH, and Fmoc-Tyr(PO(OBzl)OH)-OH, respectively. For final thioether cyclization of c(MpSIpYVA), the N terminus of the peptide was chloroacetylated using chloroacetic anhydride and N,N-diisopropylethylamine (DIEA), following established methods with minor modifications (38,48). For consistency in reporting structural information concerning macrocyclic peptides, the peptide-like unit containing the thioether bond is designated as the defined chemical component 48V (49,50). Cleavage of the peptides from the resin was achieved using a mixture of Reagent K (TFA/thioanisole/water/phenol/1,2-ethanedithiol, 82.5:5:5:5:2.5 v/v) and 1% triisopropylsilane (TIPS) for 3 h at room temperature. After removal of the resin by filtration, the filtrate was concentrated by flushing with nitrogen gas. Crude peptides were precipitated with diethyl ether. For synthesis of c(MpSIpYVA), after cleavage from the resin, crude peptide was cyclized by dissolving in a 1% triethylaminecontaining distilled, deionized water (Millipore) (ϳ3 mM) and stirring at room temperature overnight. The cyclization reaction was quenched by acidification with acetic acid. Crude peptides were purified using reversed-phase HPLC (RP-HPLC) on a preparative C4 column (BioAdvantage Pro 300, Thomson Liquid Chromatography) with 0.05% TFA/water/acetonitrile as the solvent. Purified peptides were characterized by matrixassociated laser desorption ionization TOF MS (MALDI MicroMX, Waters). The purities of the peptides were found to be Ͼ98% using analytical RP-HPLC.

Peptide labeling
The peptide p38(175-185, 180pT) was labeled with 5(6)-carboxyfluorescein (Novabiochem) on the solid phase. Dry resinbound N-terminal deprotected peptide was reacted with an excess of 5(6)-carboxyfluorescein, 1-hydroxybenzotriazole/ hexafluorophosphate benzotriazole tetamethyl uronium, and DIEA in N-methyl-2-pyrrolidone (NMP) in the dark for 24 h. The resin was washed thoroughly with 20% piperidine in NMP until the wash solution became colorless. Resin was then washed consecutively with NMP and diethyl ether five times and finally dried under an N 2 atmosphere. After labeling, peptides were cleaved as described above and purified by RP-HPLC, and masses were checked by MALDI MicroMX.

Trapped PPM1A-phosphopeptide complex Fluorescence anisotropy binding assay
Fluorescence anisotropy measurements were performed at 25°C using an LS55 luminescence spectrometer (PerkinElmer Life Sciences). The protein PPM1A D146A was dialyzed overnight against 20 mM Tris (pH 7.5), containing 100 mM NaCl, 2 mM ␤-mercaptoethanol, 30 mM MgCl 2 , 0.1 mM EGTA, and 10% glycerol. Titrations were carried out in the same buffer using an initial concentration of 5(6)-carboxyfluorescein-labeled p38(175-185, 180pT) peptide of 1 M. Fluorescence anisotropy was measured with excitation at 480 nm and emission at 520 nm using 5-nm bandwidths. After mixing with PPM1A D146A, solutions were incubated for 2 min before fluorescence anisotropy measurements were taken. Each titration point was measured twice, and titrations were performed four times. The binding constant, K d , was estimated from nonlinear leastsquares fitting (GraphPad Prism 7.01) of background-corrected, averaged anisotropy values, A, to Equation 1, where A f and A b are the anisotropy of the free and bound peptide, respectively, and [P t ] is the dilution-corrected total protein concentration.

NMR spectroscopy
NMR experiments were conducted at 30°C using a 600 MHz Bruker Biospin NMR spectrometer with a TCI cryoprobe. The protein PPM1A D146A and the peptide p38(175-185, 180pT) were co-dialyzed overnight against phosphate buffer (20 mM sodium phosphate (pH 7.5), 100 mM NaCl, 5 mM MgCl 2 ), and experiments were carried out in the same buffer. Total correlation spectroscopy (TOCSY; mixing time, 60 ms) and NOESY (mixing time, 400 ms) spectra were recorded in water-suppression mode.

Steady-state kinetics assays
Phosphatase activity assays were performed as described previously (29). Under standard conditions, initial rates of phosphatase activity were determined by incubating proteins (15 ng, 9.2 nM) with various concentrations of phosphopeptide in phosphatase assay buffer (50 mM Tris⅐HCl (pH 7.5), 40 mM NaCl, 0.1 mM EGTA, 0.02% ␤-mercaptoethanol) supplemented with 30 mM MgCl 2 for 9 min at 30°C. The dependence of initial rates on divalent metal ion concentration was determined as described above, except that the MgCl 2 concentrations were varied from 0.5 to 30 mM. Phosphate was determined using the Biomol green assay (Enzo Life Sciences), following the manufacturer's protocol. Phosphatase activity toward pNPP was measured using 120 ng (73.2 nM) of protein in phosphatase assay buffer supplemented with 10 mM MnCl 2 for 12 min at 30°C. The assay was terminated by adding 2 volumes of a 1:1 molar ratio of 2 N NaOH and 0.5 M EDTA (pH 8.0). The amount of p-nitrophenol released was determined spectrophotometrically (⑀ ϭ 18,000 M Ϫ1 cm Ϫ1 at 410 nm). To determine K m and k cat values, the initial rates were fitted to the Michaelis-Menten equation using GraphPad Prism 7.01. Values of k cat were calculated from the concentration of the enzyme, [E] 0 and the equa-

Fitting initial rates to the rapid equilibrium, random-order, bi-substrate mechanism
In the rapid equilibrium approximation, the initial rate of an enzymatic reaction that requires the binding of both substrate (S) and an essential divalent metal ion cofactor (M), in either order, is given by Equation 2,

ITC
ITC measurements were performed at 25°C using a VP-ITC MicroCalorimeter (MicroCal). PPM1A cat D146E and the cyclic peptide c(MpSIpYVA) were individually co-dialyzed against Trapped PPM1A-phosphopeptide complex Buffer C. ITC experiments were carried out in the same buffer with a protein concentration of 16 M. Integrated heats of injection were corrected for the heats of dilution and were fitted to a one-site binding model using Origin 7.0 software.

Crystallization and data collection and data processing
PPM1A cat D146E was purified as described above except that the buffer used for the Superdex 75 16/60 gel filtration had the following composition: 10 mM MES (pH 7.0), 150 mM NaCl, 2 mM TCEP, and 2 mM MgCl 2 . After concentration of the protein to 20 mg/ml, the protein and cyclic peptide were co-dialyzed against the above buffer. The complex was formed by combining the protein and c(MpSIpYVA) in a 1:2 molar ratio and incubating at 25°C for 1 h. In the crystallization trial, 3 l of the complex was mixed with 3 l of the well solution consisting of 0.1 M HEPES (pH 7.3), 0.1 M calcium acetate, and 40% PEG400 in a hanging drop vapor diffusion experiment. Crystals grew over 3 days to an overall size of 200 ϫ 200 m at 20°C and were cryoprotected by transfer into a solution prepared by diluting the complex buffer into 0.1 M HEPES (pH 7.3), 0.1 M calcium acetate, 40% PEG400, and 26% glycerol in 1:1 ratio for 25 min. Crystals were flash-frozen directly in a nitrogen cryostream. Data were collected at 95 K, using a Rigaku 007HF rotating anode X-ray source producing CuK ␣ radiation (1.5418 Å) and equipped with multilayer focusing mirrors on a Saturn A200 mosaic CCD detector in 0.25°oscillation images. The data were integrated and scaled internally using XDS and XSCALE (51).

Structure solution and refinement
The crystal structure was determined by molecular replacement using AMoRe (52) with a model of PPM1A (PDB code 1A6Q) used in the search. Three molecules were found in the asymmetric unit. All subsequent calculations were carried out using CNS1.3 (53). The model was refined at 2.2 Å resolution with Cartesian-simulated annealing, energy minimization, and atomic displacement parameter optimization. Electron density was displayed, and the model was manually modified where needed using O (54). Difference electron density for cyclic peptide c(MpSIpYVA) was apparent for all three molecules in the asymmetric unit. Chemical parameters restraining the cyclic peptide in crystallographic refinement were constructed so that they were consistent with the Engh and Huber (55) protein restraint parameters used throughout the process. After further refinement, 10 Ca 2ϩ ions with their associated coordinating waters and 606 additional ordered solvent water molecules were added. The final model was verified with a compositesimulated annealed omit electron density map calculated in CNS 1.3. Data collection and refinement statistics are shown in Table 2. The Ramachandran plot of the final model indicated that 89.9% of all residues were in the most favored regions of the plot, whereas three residues (one in each monomer) were in the disallowed region.

Small-angle X-ray scattering
Before the experiments, the proteins and the c(MpSIpYVA) cyclic peptide were co-dialyzed against 10 mM MES, pH 7.0, containing 150 mM NaCl, 2 mM TCEP, and 2 mM MgCl 2 , and experiments were carried out in the same buffer. Solution X-ray scattering data were collected for samples containing between 3.0 and 0.5 mg/ml protein to investigate the effect of sample concentration on the scattering data. SAXS data collections were performed at beam line 12IDB at the Advanced Photon Source synchrotron (Argonne National Laboratory, Argonne, IL) or using a lab-based instrument (SAXSLAB, Institute for Bioscience and Biotechnology Research, Rockville MD), all carried out at 25°C. Synchrotron data were collected using 14-keV incident radiation and Pilatus 2M detector covering the q-range between 0.008 and 0.96 Å Ϫ1 . Lab-based SAXS data were collected using 8-keV incident radiation from a Rigaku 007HF rotating anode source and a Pilatus 300K detector with programmable positioning covering the q-range from 0.008 to 0.8 Å Ϫ1 . Multiple data sets, including concentration series, were collected on both instruments yielding scattering profiles that were indistinguishable apart from data uncertainty. Scattering measurements reported here comprise 30 independent data frames acquired sequentially at Advanced Photon Source with means calculated over the 30 frames and uncertainty in the mean calculated as the root-mean-square deviations over the 30 frames divided by the square root of 30. No data frames were removed. Two-dimensional scattering data were processed using instrument-specific routines correcting for the sample transmission, detector pixel sensitivity, and solid angle per pixel while using dynamic masking to remove the data impacted by stray cosmic radiation. Buffer scattering was subtracted from sample curves using protein concentration-based solvent volume fractions. The radii of gyration were determined using Guinier fits within the q-ranges corresponding to q max ⅐R g Ͻ 1.1-1.2 using Primus software (56) while ensuring the absence of the systematic biases in the fit residuals. All fits of the scattering data to the atomic coordinates of the proteins were performed using AXES software (57), averaging over multiple ensemble members where appropriate.

Molecular dynamics simulations
All simulations were done with substitution of Mg 2ϩ ions for the Ca 2ϩ ions in the crystal structure. Four different states of the PPM1A cat protein were investigated as follows: 1) the WT protein with a third Mg 2ϩ ion occupying the Asp-146/Asp-239 subsite; 2) as in state 1 but with bound cyclic peptide; 3) the D146E mutant protein with a third Mg 2ϩ ion occupying the Asp-239/Asp-243 subsite; 4) as in state 3 but with bound cyclic peptide. Trajectories were calculated with the NAMD 2.9 software (58) using the CHARMM36 all-hydrogen topologies and parameters (59). Positional harmonic constraints were applied to the ␣-carbons of the core residues 13-16, 24 -27, 39 -42, 56 -60, 129 -132, 140 -143, 148 -151, 159 -162, 220 -223, 234 -237, 284 -288, and 293-295. Proteins were centered in a cubic box with initial dimensions of 75.0 Å. The solvent was represented by the explicit TIPS3P model, and periodic boundary conditions were applied. A CUTOFF of 12.0 Å was used for the energy functions, with a SWITCHDIST of 10.0 Å and PAIR-LISTDIST of 14.0 Å. The Langevin and Langevin-Piston algorithms were used to propagate the dynamics and maintain constant temperature and pressure of 300.0 K and 1 atm, respectively. Rigid bonds were used to allow for an integration time step of 2.0 fs. Trajectories for each state were propagated Trapped PPM1A-phosphopeptide complex