Molecular Insights into the Fungus-Specific Serine/Threonine Protein Phosphatase Z1 in Candida albicans

ABSTRACT The opportunistic pathogen Candida is one of the most common causes of nosocomial bloodstream infections. Because candidemia is associated with high mortality rates and because the incidences of multidrug-resistant Candida are increasing, efforts to identify novel targets for the development of potent antifungals are warranted. Here, we describe the structure and function of the first member of a family of protein phosphatases that is specific to fungi, protein phosphatase Z1 (PPZ1) from Candida albicans. We show that PPZ1 not only is active but also is as susceptible to inhibition by the cyclic peptide inhibitor microcystin-LR as its most similar human homolog, protein phosphatase 1α (PP1α [GLC7 in the yeast Saccharomyces cerevisiae]). Unexpectedly, we also discovered that, despite its 66% sequence identity to PP1α, the catalytic domain of PPZ1 contains novel structural elements that are not present in PP1α. We then used activity and pulldown assays to show that these structural differences block a large subset of PP1/GLC7 regulatory proteins from effectively binding PPZ1, demonstrating that PPZ1 does not compete with GLC7 for its regulatory proteins. Equally important, these unique structural elements provide new pockets suitable for the development of PPZ1-specific inhibitors. Together, these studies not only reveal why PPZ1 does not negatively impact GLC7 activity in vivo but also demonstrate that the family of fungus-specific phosphatases—especially PPZ1 from C. albicans—are highly suitable targets for the development of novel drugs that specifically target C. albicans without cross-reacting with human phosphatases.

PP1␥]). PP1 regulates diverse and essential biological processes by dephosphorylating a variety of protein substrates. Although the intrinsic substrate specificity of PP1 is very low, by interacting with regulatory proteins to form distinct holoenzymes (~200 biochemically confirmed PP1 interactors), PP1 achieves high specificity (6)(7)(8). More than two decades ago, it was discovered that budding yeast (Saccharomyces cerevisiae), fission yeast (Schizosaccharomyces pombe), and the opportunistic fungus Candida albicans also carry a PP1 gene, coding for GLC7, Dis2, and GLC7-like, respectively ( Fig. 1). There is~80% sequence identity between C. albicans GLC7 and human PP1 isoforms ("GLC7" will be used here to refer to the PP1 homolog in fungal species) (9). Like PP1, GLC7 controls a plethora of essential biological processes, and its activity is regulated by its interaction with multiple regulatory proteins, many of which are conserved in humans (8). However, it was also discovered that fungi express a unique family of PP1-like genes, which include those coding for PPZ1 and PPQ1/Sal6 (10,11). Unlike GLC7, these families of fungus-specific PP1-like phosphatases consist of two distinct domains: an N-terminal domain that is enriched in serines and is predicted to be unstructured (i.e., a member of the intrinsically disordered protein [IDP] family) and a C-terminal catalytic domain that has high sequence similarity to GLC7 (Fig. 1). This suggests that the fungus-specific phosphatases may also bind the GLC7-specific regulatory proteins. This was confirmed in S. cerevisiae, in which yeast two-hybrid studies demonstrated that S. cerevisiae PPZ1 (ScPPZ1) binds some, but not all, S. cerevisiae GLC7 (ScGLC7)-specific regulatory proteins (12). However, the molecular determinants that explain why fungus-specific phosphatases bind only a subset of the GLC7specific regulators are still largely unknown.
Furthermore, not only is PPZ1 important for cation homeostasis and cell wall biosynthesis, it is also critical for C. albicans virulence (13,14). Namely, deletion of the Ppz1 gene reduces the ability of C. albicans to infect mice, a function that can be rescued by the reintegration of a single copy of the Ppz1 gene. In addition, PPZ1 has also been shown to play a role in the morphological changes in C. albicans associated with infectivity: that is, the transition from the yeast to the hyphal form. This is because the deletion of Ppz1 reduces the rate of hyphal growth (15). Together, these data suggest that the specific inhibition of PPZ1 will prevent this morphological transition and, as a consequence, block C. albicans infectivity without killing the commensal pathogen. The fungistatic effect of such a treatment would be more beneficial than eliminating C. albicans altogether as the latter might result in uncontrolled bacterial proliferation. Here, we used X-ray crystallography and biochemistry to elucidate the structural and functional characteristics of PPZ1 that are unique to the fungusspecific family of phosphatases. We discovered novel structural elements in PPZ1 that not only explain why PPZ1 binds less effectively to GLC7 regulators but also define new interaction surfaces that may be leveraged for the development of novel, effective antifungal therapeutics.  (14,16). PPZ1 cat and human PP1␣ exhibit 66% sequence identity throughout their catalytic domains, and the 6 residues that coordinate the active site metals are 100% conserved ( Fig. 1; see Fig. S1 in the supplemental material). Accordingly, full-length PPZ1 (PPZ1 FL ) and its individual domains (PPZ1 Nterm and PPZ1 cat ) are readily isolated from Escherichia coli. In addition, PPZ1 FL and PPZ1 cat are active, as they effectively dephosphorylate a small molecule substrate mimetic (p-nitrophenyl phosphate [pNPP]) (see Fig. S2 in the supplemental material). Furthermore, their activities are identical to that of human PP1␣ purified from E. coli (8) or PP1c (a mixture of PP1 isoforms ␣, ␤, and ␥) purified from rabbit muscle (17). However, unlike PP1␣, which expresses solubly and crystallizes readily without its C-terminal disordered residues (residues 301 to 330) (8), a PPZ1 construct truncated at the corresponding residue (PPZ1 cat ⌬466 -484) is largely insoluble compared to PPZ1 cat (Ͼ10-fold reduction in yield). This suggests that the PPZ1 cat C-terminal region is critical for folding and/or stability. Consistent with this conclusion, the secondary structure prediction program PSIPRED (18) predicts that these residues are not disordered as they are in PP1␣ but instead form an ␣-helix (see Fig. S3 in the supplemental material).

PPZ1
The C-terminal residues of PPZ1 cat are structured and form an ␣-helix. In order to understand the molecular consequences of the sequence differences between PPZ1 and PP1␣ and the role of the PPZ1 C-terminal residues in PPZ1 function, we determined the 3-dimensional crystal structure of PPZ1 cat to 2.61 Å resolution ( Table 1). The PPZ1 cat structure includes residues 171 to 478. (The electron density for the last 6 residues, 479 to 484, was not observed, and thus they were not modeled.) As expected, the PPZ1 cat structure adopts the canonical PP1 fold ( Fig. 2A and B), comprising a mixed ␣/␤ protein, whose central loops are positioned to coordinate the active site metals. However, the structures are not identical. The root mean square deviation (RMSD) between PP1␣ (PDB no. 4MOV [19]) and PPZ1 cat is 1.23 Å (backbone). The most significant difference is in the N-terminal region, where the PPZ1 cat helix A= extends for one more turn compared with PP1␣ and the loop connecting helix A= to helix A (L1) adopts a conformation distinct from that in PP1␣ ( Fig. 2B to D). (The RMSD of L1 between PPZ1 cat and PP1␣ is 2.99 Å.) Although the sequences of L1 between PPZ1 cat and PP1␣ are only 33% identical, this change of conformation is primarily due to a single amino acid change in helix C, Tyr144 PP1␣ to Cys309 PPZ1 . The much smaller cysteine side chain creates a hydrophobic pocket into which the side chain of Val193 PPZ1 binds (Fig. 2D). Because this pocket is not present in PP1␣, the corresponding PP1␣ residue, Val28 PP1␣ , interacts directly with helix A=. As a consequence, the C␤ atoms of the two side chains are separated by 5.6 Å between the two structures. The rest of the PPZ1 cat L1 conformation changes to accommodate this new interaction, which in turn widens the pocket defined by helix A=, L1, and helix B (referred to here as the "Z1helix binding pocket").
The newly widened Z1-helix binding pocket results in the second significant structural difference between PPZ1 cat and PP1␣: the C-terminal residues of PPZ1 cat are not disordered like they are in PP1␣, but instead form an ␣-helix that nestles into this newly widened Z1-helix binding pocket ( Fig. 2B to E). This interaction is stabilized by hydrophobic interactions between the C-terminal helix and residues from helices A= and B, with the interaction centered on PPZ1 C-terminal helix residue Met473 PPZ1 (Fig. 2E). This residue is completely buried from solvent via interactions with Leu464 PPZ1 , Leu469 PPZ1 , and Val472 PPZ1 from the C-terminal helix as well as by interactions with Phe185 PPZ1 from helix A= and His235 PPZ1 , Ile238 PPZ1 , and Arg239 PPZ1 from helix B. The majority of these residues are not conserved in PP1␣. In particular, only a single residue (underlined) between the PPZ1 C-terminal helix and the C-terminal disordered tail of PP1␣ is conserved (PPZ1, 466 SAALKQVMKKEKQ 478 ; PP1␣, 301 KNKG-KYGQFSGLN 313 ), with the sequence of PPZ1 consisting of multiple hydrophobic residues and no helix-disrupting glycines (Leu469 PPZ1 , Val472 PPZ1 , and Met473 PPZ1 versus Gly304 PP1␣ , Gly307 PP1␣ , and Gly311 PP1␣ , respectively), rationalizing why the corresponding residues in PP1␣ are unstructured. Finally, the experimental B-factors for residues in the Z1-helix are higher than the rest of PPZ1 cat (Fig. 2F), suggesting that the Z1-helix is more dynamic than the rest of the PPZ1 cat .
The Z1-helix is dynamic. We also determined the structure of PPZ1 cat bound to the marine toxin microcystin-LR (MC [PPZ1 cat -MC]) (Table 1), a potent cyclic peptide inhibitor of PP1 (20). Both the MC-free and MC-bound structures of PPZ1 cat are highly similar, with an RMSD of 0.54 Å (Fig. 3A). The largest difference is in the ␤12-␤13 loop, which contains Cys438 PPZ1 (Fig. 3B). This cysteine forms a covalent bond with the bound MC identical to that observed in the PP1␣-MC complex (21). This causes the 438 CGEFD 442 loop to change conformation and become more dynamic (residues 439 GEF 441 were not modeled in the PPZ1 cat -MC structure due to a lack of density). Unexpectedly, the Z1-helix is also no longer ordered in the PPZ1 cat -MC structure (Fig. 3A). This was not due to MC binding, but instead the Z1helix was displaced by a symmetry-related molecule in the crystal. The observation that the Z1-helix can be displaced supports the observation that the Z1-helix is more dynamic than the rest of the PPZ1 cat . Importantly, the positions of L1 are identical between the PPZ1 cat and PPZ1 cat -MC structures, demonstrating that the conformation of L1 is intrinsic to PPZ1 cat itself and not a consequence of Z1-helix binding.
MC binds the active site of PPZ1 cat essentially identically to that observed in PP1␣ (Fig. 3C). Furthermore, the active sites are also identical (Fig. 3D). The bulk of the MC binds the PPZ1 cat hydrophobic binding pocket (Fig. 3A), while the remainder covers and, as a consequence, blocks the active site. The largest difference between the conformations of MC in PP1␣-MC and PPZ1 cat -MC is a change of the rotomer conformation of the 1Zn residue phenyl group and the Arg guanidinium group (Fig. 3C). The orientation observed in PPZ1 cat -MC is favored because a malonate ion from the crystallization mother liquor is bound in the rotomer position populated in the PP1␣ structure; thus both positions are expected to be equally likely in the absence of malonate. Furthermore, because the residues that mediate MC binding are highly conserved between PPZ1 cat and PP1␣ (79% identical and 90% similar [see Fig. S1 in the supplemental material]), MC is predicted to inhibit their activities with similar 50% inhibitory concentration (IC 50 ) values. We tested this in vitro using both small molecule (pNPP)and peptide (myosin light chain [pMLC])-based substrates, which show that the IC 50 values of MC for all constructs are essentially identical (IC 50 s for pMLC of 0.96 nM for PPZ1 FL , 0.72 nM for PPZ1 cat , and 0.43 nM for PP1c, and IC 50 s for pNPP of 11.8 nM for  The sequence and structural conservation of regulatory protein binding pockets in PPZ1 cat is highly variable. While PP1 exhibits broad specificity, it acts in a highly specific manner by forming stable complexes (holoenzymes) with a host of regulatory proteins that direct PP1 activity toward specific substrates and localize PP1 to specific regions of the cell (8,22). Recent structural studies have revealed that PP1 binds these regulators using small linear interaction motifs (SLiMs) (7,19,23,24). The most well- known is the RVxF SLiM, which is found in~70% of all known regulators (22). Others include the MyPhoNE, SILK, ⌽⌽, and Arg SLiMs (6,7,19,23,25). Mapping of the sequence differences onto the structure of PPZ1 cat shows that the bulk of the changes are distally located from the active site (Fig. 4A). This comparison also reveals that the RVxF interaction residues are 91% conserved between PPZ1 and PP1␣ (the only difference being a conservative Met290 PP1␣ -to-Leu455 PPZ1 substitution [ Fig. 4B; see Fig. S1 in the supplemental material]), suggesting that all PP1 regulators that contain an RVxF sequence should bind PPZ1. Other sites are similarly conserved. For example, inhibitor-2 (I-2) is a specific protein inhibitor of PP1. In addition to an RVxF and SILK motif, it binds PP1 using a long helix, which binds across the PP1 active site (25). Like the RVxF motif, the I-2 helix binding pocket is highly conserved in PPZ1 (86% identical, 93% conserved Fig. 4B; see Fig. S1) suggesting that I-2 can bind productively to PPZ1. Likewise, the conservation of the NIPP1-helix interaction pocket (70% identical and 80% conserved [ Fig. 4B; see Fig. S1]), also suggesting that this PP1 regulatory protein can bind productively to PPZ1 cat .
In contrast, other SLiM binding pockets are much less conserved in PPZ1. For example, the SILK binding sites are only 71% identical and 79% similar between PPZ1 and PP1␣ ( Fig. 4B; see Fig. S1 in the supplemental material). Three of the 4 amino acid differences result in a change from an acidic to uncharged residue (Glu56 PP1␣ /Ala221 PPZ1 , Asp166 PP1␣ /Ala331 PPZ1 , and Glu167 PP1␣ / Gly332 PPZ1 ). As these acidic residues coordinate the basic "K" of the SILK motif, the "K" may not be necessary for PPZ1 to bind SILK motif-containing proteins. The same is true for the My-PhoNE binding pocket (23) (Fig. 4B; see Fig. S1). While the majority of residues that define the pocket are largely similar-57% identical and 75% similar-multiple residues differ between the two proteins (Asp179 PP1␣ /Val344 PPZ1 , Gln198 PP1␣ /Phe363 PPZ1 , Gly215 PP1␣ /Glu380 PPZ1 , and His237 PP1␣ /Ser402 PPZ1 ). The Gly215 PP1␣ /Glu380 PPZ1 substitution results in the largest clash in the superimposed structures, with the Glu380 PPZ1 side chain clashing with that of Trp17 MYPT1 , suggesting that these changes impact MYPT1 binding.
The SLiM binding pocket that is most different between PP1␣ and PPZ1 is the ⌽⌽ interaction site, with an identity of 22% and a similarity of 67% (7, 19) ( Fig. 4B; see Fig. S1 in the supplemental material). The most significant differences are the replacement of Tyr78 PP1␣ , which defines the ⌽⌽ interaction pocket, with Lys243 PPZ1 and the replacement of Ala279 PP1␣ with Trp444 PPZ1 . Both Lys243 PPZ1 and Trp444 PPZ1 are large, bulky residues that hinder access to the pocket. This suggests that PP1 regulators that contain a ⌽⌽ motif will bind PPZ1 with lower affinity than PP1␣. Finally, although the Arg interaction pocket is perfectly conserved between both PP1␣ and PPZ1 (see Fig. S1), the presence of the Z1-helix in PPZ1 is expected to negatively impact the binding of regulators that use this site for binding, especially those that also use the ⌽⌽ binding pocket. Because the Z1-helix binds between these two sites, its presence will block access to the Arg site (Fig. 4C), reducing the ability of PPZ1 to bind regulators that contain these sequential SLiMs.
PPZ1 binds only a subset of PP1 regulatory proteins. To test if PPZ1 binds differentially to known PP1 regulators, we used pulldown assays with three well-characterized regulators-GADD34, PNUTS, and spinophilin ( Fig. 5A and B) [19], and 8.7 nM for spinophilin 417-602 [27]). As predicted, the pulldown assays show that PPZ1 cat binds GADD34 less effectively than PP1␣ (Fig. 5C and D). Because the RVxF site is nearly perfectly conserved between PP1 and PPZ1 cat , this suggests that the weakened affinity is largely due to an inability of GADD34 552-567 to productively bind the ⌽⌽ motif binding pocket. To quantify the reduction in binding, we used isothermal titration calorimetry (ITC). The affinity of GADD34 for PPZ1 cat decreases more than 19-fold, resulting in a K D of only 1,150 Ϯ 90 nM (Fig. 5E). Consistent with this, peptides that contain only RVxF motifs bind PP1 with K D s in the micromolar range (28), similar to those observed for GADD34 552-567 and PPZ1 cat .
While the PP1 binding domain of GADD34 contains only an RVxF motif and ⌽⌽ motif, other regulators, such as PNUTS 394 -433 and spinophilin 417-602 contain additional motifs that facilitate PP1 binding (PNUTS, RVxF-⌽⌽-Arg; spinophilin, RVxF-⌽⌽-Arg-spinophilin 417-602 -helix [ Fig. 5A and B]). We used pulldown assays to determine if the altered ⌽⌽ binding pocket and the presence of the Z1-helix negatively impact the binding of these regulators to PPZ1 (Fig. 5C). The results show that PPZ1, compared to PP1␣, does not effectively pull down either PNUTS or spinophilin (~85% less binding [ Fig. 5C and D]). Furthermore, mutation of either the ⌽⌽ or Arg motif in spinophilin results in a similar reduction in PP1 binding, to levels similar to those observed for PPZ1 and wild-type spinophilin (see Fig. S4 in the supplemental material). Together, these data demonstrate that the altered ⌽⌽-binding pocket coupled with the presence of the Z1-helix (Fig. 5B) negatively impacts the ability of PPZ1 to interact with regulators that require these sites for binding.
The presence of the N-terminal IDP domain reduces the ability of I-2 to inhibit PPZ1. Unlike GADD34, PNUTS, and spinophilin, I-2 does not bind PP1 at the ⌽⌽or PPZ1-specific binding pockets. Instead, it binds PP1 using the SILK-RVxF-I-2-helix interaction pockets (25). With the exception of the SILK motif binding pocket, these interaction sites in PPZ1 are largely conserved with those in PP1␣. Furthermore, the change in the SILK binding pocket suggests that PPZ1 no longer has a strict requirement for the "K" residue (Fig. 4B). This suggests that I-2 will inhibit both PPZ1 and PP1␣ with equal potencies. To test this, we measured the ability of I-2 to bind and inhibit PPZ1 activity and compared it to that determined for PP1c. The data show that both PPZ1 cat and PP1c are inhibited by I-2 with nearly equivalent IC 50 s, measured to be 7 and 12 nM, respectively (Fig. 5F). However, unexpectedly, we also discovered that I-2 is a much less potent inhibitor against PPZ1 FL , which includes the~160-amino-acid N-terminal IDP domain (see Fig. S3 in the supplemental material). With PPZ1 FL , the IC 50 of I-2 increases 39-fold to 278 nM; this increase was observed for both recombinant I-2 and I-2 partially purified from rabbit muscle (not shown). (The latter result is consistent with previous studies [29] that showed that S. cerevisiae PPZ1 FL is also poorly inhibited by I-2.) Furthermore, the specific activities of the two PPZ1 constructs also differ significantly, with 4.6 mU/mg for PPZ1 FL and 1,200 mU/mg for PPZ1 cat . These data suggest that the N-terminal domain likely masks one or more binding sites for I-2 (but not the active site, as the PPZ1 FL protein is fully susceptible to MC inhibition [ Fig. 3E]) and prevents, via an as yet undetermined mechanism, PPZ1 inhibition by this protein inhibitor.

DISCUSSION
Our structural and biochemical studies reveal why, in spite of their high sequence similarity, fungus-specific phosphatases like PPZ1 bind only a subset of the GLC7-specific regulators (Fig. 6A). Specifically, we discovered that there are three distinct mechanisms by which PPZ1 inhibits the binding of many GLC7-specific regulators. First, sequence differences between PPZ1 cat and PP1␣ in canonical SLIM interaction pockets, especially the ⌽⌽ binding pocket, inhibit the subset of regulators that requires these sites for binding (see Fig. S1 in the supplemental material). Second, the presence of the N-terminal IDP domain, which is not found in GLC7, also negatively impacts the binding of at least some regulators (e.g., I-2), likely through an as yet undefined steric mechanism ( Fig. 5 and 6A). Third, structural differences between PPZ1 cat and PP1␣ can also negatively impact regulator binding. In particular, our data show that the presence of the Z1-specific helix hinders access to the C-terminal groove by regulators that bind the Arg site via the ⌽⌽ site (Fig. 5). In addition, the dynamic nature of this helix raises the intriguing possibility that it may also serve a regulatory role in PPZ1 function. Finally, the binding pocket of the Z1-specific helix is also unique, as that of PP5, the only other known serine/threonine protein phosphatase with a C-terminal helix, binds at the front, versus the top, of the core catalytic domain (Fig. 6B) (30). Together, these results show that instead of PPZ1 competing with GLC7 for its regulators, the regulators preferentially bind and control the activity of GLC7 (Fig. 6A). This resolves a long-standing question about why the presence of the fungus-specific phosphatases does not globally disrupt GLC7 function (12).
Our studies also revealed that multiple pockets in PPZ1 cat are unique to this phosphatase and thus can be exploited for drug development. While the PPZ1 cat and PP1␣ active sites are perfectly conserved, the conformation of loop L1 is unique in PPZ1 cat , as is its folded Z1-specific helix. Both of these new structural elements create novel binding surfaces that are not present in human isoforms of PP1. Thus, they are useful for the development of specific, potent antifungals. Because their active sites are per-fectly conserved, one strategy for developing a fungus-specific inhibitor of PPZ1 would be to use "fragment linking" (31): i.e., linking a small molecule that targets one of the unique PPZ1specific binding pockets with a small molecule that, for example, targets the PP1 active site. This would result in an inhibitor that selectively targets PPZ1, as it would bind PPZ1 with higher affinity than human PP1 isoforms. Because PPZ1 is critical for C. albicans hypha formation, a morphological change associated with infectivity, this type of PPZ1-specific inhibitor would stop C. albicans infections without killing the commensal pathogen. This would prevent uncontrolled bacterial proliferation that can accompany C. albicans elimination and thus will result in novel, potent drugs with minimal side effects for the treatment of candidemia.

MATERIALS AND METHODS
Protein expression, purification for pNPP-based assays, and crystallography. The gene encoding PPZ1 cat (aa 171 to 484) was synthesized (Ge-neArt; Invitrogen), subcloned into the RP1B bacterial expression plasmid (32), and expressed largely as previously described (8). Specifically, the plasmid was cotransformed with the pGRO7 plasmid, which encodes the GroEL/GroES chaperone (TaKaRa), into E. coli BL21(DE3) cells (Invitrogen). Cells were grown in LB medium supplemented with 1 mM MnCl 2 at 30°C to an optical density at 600 nm (OD 600 ) of~0.5, at which point arabinose was added (2 g/liter) to induce the expression of the GroEL/ GroES chaperone. At an OD 600 of~1, the temperature was lowered to 10°C, and the expression of PPZ1 was induced using 0.1 mM IPTG (isopropyl-␤-D-thiogalactopyranoside): the protein was allowed to express for~20 h at 10°C. The cells were harvested by centrifugation, suspended in fresh LB medium (again supplemented with 1 mM MnCl 2 and 200 g/ml of chloramphenicol to inhibit the ribosome), and agitated for~5 h at 10°C. Harvested cells were frozen and stored at Ϫ80°C.
All purifications were performed at 4°C. Cells expressing His 6 -tagged tobacco etch virus (TEV) protease sequence-PPZ1 cat were lysed in lysis buffer (25 mM Tris [pH 8.0], 700 mM NaCl, 5 mM imidazole, 1 mM MnCl 2 , 0.1% Triton X-100) using high-pressure homogenization (Avestin C3 EmulsiFlex) in the presence of an EDTA-free protease inhibitor cocktail (Roche). The lysate was clarified by centrifugation at 45,500 ϫ g, filtered through a 0.22-m-pore polyethersulfone (PES) membrane filter (Millipore), and then loaded onto Ni 2ϩ -nitrilotriacetic acid (NTA) resin (GE Healthcare) preequilibrated in buffer A (25 mM Tris [pH 8.0], 700 mM NaCl, 5 mM imidazole, 1 mM MnCl 2 ). Bound His 6 -TEV-PPZ1 cat was washed with 100 ml of buffer A, followed by more stringent wash with 100 ml "stringent wash buffer" consisting of 94% buffer A and 6% buffer B (25 mM Tris [pH 8.0], 700 mM NaCl, 250 mM imidazole, the lack of conservation of key PP1/GLC7 interaction domains (such as the ⌽⌽ binding pocket), and the presence of the PPZ1-specific helix negatively impact the binding of PPZ1 with many GLC7-specific regulators, leaving the regulators to preferentially bind and regulate the activity of GLC7. (B) Overlay of PPZ1 (pink) and PP5 (blue; PDB no. 1S95 [30]), illustrating that the C-terminal helices of the two proteins bind different pockets in the core PSP catalytic domain.
1 mM MnCl 2 ). The bound His 6 -TEV-PPZ1 cat was eluted using 100% buffer B and immediately purified by size exclusion chromatography (SEC) using Superdex 75 26/60 preequilibrated in 20 mM Tris (pH 8.0), 500 mM NaCl, 0.5 mM Tris(2-carboxyethyl)phosphine (TCEP), and 1 mM MnCl 2 . Fractions containing the His 6 -TEV-PPZ1 cat protein were pooled and incubated overnight with TEV protease at 4°C. A second Ni 2ϩ -NTA "subtraction" purification was used to separate cleaved PPZ1 cat from the cleaved His 6 tag and TEV; the flowthrough, which contained the cleaved PPZ1 cat , was pooled and purified in a final step using SEC. PPZ1 cat was pooled, concentrated, and used immediately for crystallization experiments.
Protein expression and/or purification for dephosphorylation assays. The catalytic subunit of rabbit skeletal muscle protein phosphatase 1 (PP1c) was isolated as described (33). The 20-kDa myosin light chain (MLC20) and the myosin light chain kinase (MLCK) were obtained from turkey gizzard, and MLC20 was phosphorylated by MLCK in the presence of [␥-32 P]ATP and Mg 2ϩ as described previously (34). Recombinant 6ϫHis-tagged inhibitor-2 (I-2) was prepared as described previously (28); I-2 was also isolated from rabbit skeletal muscle (35) for some control experiments.
The coding sequences for full-length Candida PPZ1 FL and its conserved C-terminal catalytic domain, termed PPZ1 cat , were generated by PCR from the pET28-CaPPZ1 plasmid (GenBank accession no. GQ357913 [14,36]) and cloned into the E. coli expression vector pGEX6p-1 (Amersham Biosciences); both constructs were sequence verified (UD Genomed, Ltd.). Both constructs were then transformed into E. coli BL21(DE3) RIL cells (Agilent), and expression of N-terminal glutathione S-transferase (GST)-tagged proteins was induced with 0.6 mM IPTG (Sigma) at 18°C. Overnight incubation was used to express GST-PPZ1 cat , while a shorter, 3-h incubation period was required to reduce inclusion body formation and ensure optimal production of soluble GST-PPZ1 FL . In both cases, 0.5 mM MnCl 2 was included in the culture medium, and 1 mM MnCl 2 was present during the subsequent purification steps. The fusion proteins were purified using glutathione-Sepharose 4B (GE Healthcare) resin, and the GST tag was removed using the PreScission protease (GE Healthcare) during elution (following the instructions of the manufacturer). The protein concentration was determined using the Bradford assay, and protein purity was verified by SDS-PAGE.
Protein phosphatase assays. The activities of the recombinant C. albicans phosphatases were measured with 32 P-labeled MLC20 substrate as previously described (34), with the exception that the substrate concentration was 1 M and 2 mM MnCl 2 was included in the assay mixtures. The activity of rabbit PP1 was also determined under identical conditions as a control. The 50% inhibitory concentrations (IC 50 s) of microcystin-LR (obtained from Enzo Life Sciences or purified as previously described [37]) were determined by using either the protein substrate or a small molecule substrate as described previously (38), except that pNPP (Sigma-Aldrich) was used instead of 3-O-methylfluorescein phosphate (OMFP). Reactions were carried out in 96-well plates (Costar). MC concentrations were prepared by serial dilution and added to 150 l of buffered enzyme ( ) in wells containing the reaction mixtures. The high-signal controls were prepared by adding dimethyl sulfoxide (DMSO), matching the amount present in the reaction mixture with the highest concentration of MC. The low-signal controls were the reaction mixtures containing the highest concentration of microcystin-LR. The reactions were initiated by the addition of pNPP substrate and incubated at room temperature for 30 min. The reactions were stopped using 100 l of 300 mM potassium phosphate (pH 10). The absorbance was measured at 405 nm using an Epoch spectrophotometer (BioTek). The percentage of activity of each protein was calculated using the equation [(absorbance Ϫ low-signal control)/(high-signal control Ϫ low signal control)] ϫ 100%. The IC 50 was then calculated using SigmaPlot 12.5.
Crystallization and structure determination. Crystals of apo-PPZ1 cat were obtained using hanging drop vapor diffusion in 1.8 M ammonium citrate tribasic (pH 7.0) in a 2:1 protein/crystallization condition ratio. To generate PPZ1 cat -MC crystals, PPZ1 cat was first incubated with MC at a 1:1 molar ratio for 15 min before crystallization using hanging drop vapor diffusion in 0.06 M citric acid (pH 4.1) plus 16% (wt/vol) polyethylene glycol (PEG) 3350. For data collection, crystals were first cryo-protected with 30% glycerol in mother liquor (PPZ1 cat ) or 3.4 M Na-malonate (pH 4.0) (PPZ1 cat -MC) and then flash-frozen in liquid nitrogen. X-ray data were collected in house at 100 K using a Rigaku FR-Eϩ Superbright rotating copper anode X-ray generator with a Saturn 944ϩ HG charge-coupled device (CCD) detector (Brown University Structural Biology Facility). The data were phased using molecular replacement (Phaser as implemented in PHENIX [39]) using PP1 (PDB no. 4MOV [19]) as the search model. Clear electron density of MC could be observed bound to the active site. The initial models were built using Phenix.Auto-Build (40) followed by iterative rounds of refinement in PHENIX and manual building using Coot (41).
Pulldown assay. GADD34 552-567 was purchased. PNUTS 394 -433 and spinophilin 417-602 were purified according to a method described previously (19,27). Spinophilin 417-602 ⌽⌽¡AA, R469D, and R469E mutants were generated using site-directed mutagenesis. PP1␣ 7-330 and PPZ1 cat were purified by the method described above but retained the His tag. After SEC (20 mM Tris [pH 8.0], 500 mM NaCl, 0.5 mM TCEP, 1 mM MnCl 2 ), 50% (vol/vol) glycerol was added to both PP1␣ 7-330 and PPZ1 cat . The final concentrations of the proteins were adjusted to 5 M. Ten micromoles of each PP1 regulator was added into 500 l of PP1␣ 7-330 or PPZ1 cat . The mixture was incubated under rocking (4°C) for 60 min to allow for complex formation. A 15-l bed volume of Ni-NTA resin (GE Healthcare) was added into tubes, and the mixture was incubated at 4°C for 60 min. The beads were pelleted by centrifugation (2000 ϫ g) and washed three times with 500 l of SEC buffer. Forty microliters of SDS loading buffer was added to the beads, and the samples were boiled at 80°C for 5 min and analyzed on NuPAGE 4 to 12% bis-Tris gels. Gels were stained overnight with SYPRO ruby protein gel stain (Life Technologies) according to the manufacturer's protocols and scanned using a Typhoon 9410 laser scanner (GE Healthcare) with an excitation wavelength of 457 nm and emission filter of 610 nm following destaining. Densitometry was performed using ImageQuant TL 7.0 software for quantification of the band intensity.
Accession number(s). Atomic coordinates and structure factors have been deposited in the Protein Data Bank under accession no. 5JPE for PPZ1 and 5JPF for PPZ1-MC.