ReviewBiochemistry

Cracking the Phosphatase Code: Docking Interactions Determine Substrate Specificity

See allHide authors and affiliations

Science Signaling  08 Dec 2009:
Vol. 2, Issue 100, pp. re9
DOI: 10.1126/scisignal.2100re9

Abstract

Phosphoserine- and phosphothreonine-directed phosphatases display remarkable substrate specificity, yet the sites that they dephosphorylate show little similarity in amino acid sequence. Studies reveal that docking interactions are key for the recognition of substrates and regulators by two conserved phosphatases, protein phosphatase 1 (PP1) and the Ca2+-calmodulin–dependent phosphatase calcineurin. In each case, a small degenerate sequence motif in the interacting protein directs low-affinity binding to a docking surface on the phosphatase that is distinct from the active site; several such interactions combine to confer overall binding specificity. Some docking surfaces are conserved, such as a hydrophobic groove on a face opposite the active site that serves as a major recognition surface for the “RVxF” motif of proteins that interact with PP1 and the “PxIxIT” motif of substrates of calcineurin. Secondary motifs combine with this primary targeting sequence to specify phosphatase binding. A comprehensive interactome for mammalian PP1 was described, analysis of which defines several PP1-binding motifs. Studies of “LxVP,” a secondary calcineurin-binding sequence, establish that this motif is a conserved feature of calcineurin substrates and that the immunosuppressants FK506 and cyclosporin A inhibit the phosphatase by interfering with LxVP-mediated docking.

Introduction

More than 30% of proteins are phosphorylated (1), and identification of the kinases and phosphatases responsible for this versatile modification has provided fundamental insights into cellular regulation. Most protein kinases phosphorylate a characteristic cluster of amino acid residues that can be identified in candidate substrates by readily available bioinformatic tools. In stark contrast, protein phosphatases dephosphorylate phosphosites with little similarity in amino acid sequence, and the entire complement of protein phosphatases is encoded by far fewer genes—about one-tenth of the number that encode kinases (2); however, these enzymes recognize their substrates with remarkable specificity. What are the rules that enable phosphatases to recognize such a wide range of phosphorylated substrates? Studies demonstrate that substrates and regulators bind to the related phosphatases PP1 and calcineurin through multiple low-affinity interactions at conserved docking surfaces that are distinct from the active site. Phosphatase binding partners, therefore, contain combinations of small, degenerate docking motifs distributed throughout their sequences. These docking interactions determine substrate specificity and the overall affinity of the phosphatase for its interacting protein. Furthermore, blocking any single interaction can substantially disrupt phosphatase function. Genetic and structural studies have identified several of these docking motifs and suggest that we are close to cracking the phosphatase code.

A Conserved Family of Phosphatases

The phosphoprotein phosphatase (PPP) family is one of three distinct gene families that encode phosphoserine (pSer)– and phosphothreonine (pThr)–specific phosphatases. Family members include PP1 and PP2A, the two most abundant protein phosphatases, and the Ca2+-calmodulin–regulated phosphatase calcineurin (also called PP2B or PP3) (3). Although each of these phosphatases contains a related core catalytic domain, the compositions of their holoenzymes differ substantially. The catalytic subunit of PP1 combines with a host of structurally unrelated binding partners and functions as a heterodimer, or in some cases, a heterotrimer in vivo (4). These interacting proteins target the enzyme to specific intracellular locations, regulate its activity, modify its substrate specificity, and in some cases are dephosphorylated by PP1. In contrast, active calcineurin is a well-defined heterotrimer that is activated solely through increases in the concentration of cytosolic Ca2+. The A subunit of calcineurin (CNA) contains the globular catalytic domain, binding sites for its regulators the calcineurin B subunit (CNB) and calmodulin, and an autoinhibitory region (5, 6). The CNA-CNB heterodimer is maintained in an inactive state due to the binding of the C-terminal autoinhibitory domain of CNA to the catalytic cleft. Under conditions of elevated concentrations of cytosolic Ca2+, the binding of Ca2+ to CNB together with the interaction of Ca2+-calmodulin with the calmodulin-binding domain on CNA displaces the autoinhibitory domain, resulting in activation of the enzyme (6, 7). Thus, although they are closely related, the interactions of PP1 and calcineurin with distinct sets of proteins confer on each phosphatase distinct physiological and biochemical properties.

Interactions at the Active Site

PP1 and calcineurin each coordinate two metal ions at their active sites that bind directly to phosphate in the substrate and activate a water molecule to promote single-step dephosphorylation (3, 8, 9). Both enzymes have a shallow catalytic cleft and a relatively open molecular surface that accommodates a wide range of phosphorylated substrates (8, 9). Neither PP1 nor calcineurin efficiently dephosphorylates short peptides, and the activity of each phosphatase is only modestly influenced by amino acid residues adjacent to the phosphosite (10). Instead, the enzymes rely on the recognition of broader structural features in their substrates and regulators to achieve selectivity (1012). This contrasts with serine-threonine kinases whose active sites often contain a peptide-binding cleft that directs sequence-specific interactions with substrates (13, 14).

Primary Docking Sites Target Substrates and Regulators to PPPs

The structural conservation between PP1 and calcineurin extends beyond the active site to regions throughout their catalytic domains. Notably, both PP1 and calcineurin contain a hydrophobic channel on a surface opposite the catalytic site, defined in part by β14 (3) (Fig. 1). This channel forms a critical docking surface on both enzymes, serving as a primary site of regulatory-subunit interaction in PP1 and as a substrate-binding region in calcineurin (15, 16).

Fig. 1

Structural similarities between the docking of PxIxIT to calcineurin and the docking of RVxF to PP1. (A) X-ray structure of PVIVIT bound to calcineurin [based on Protein Data Bank (PDB) accession code 2P6B (38)]. Calcineurin B (CNB) is shown in purple, and its bound Ca2+ ions are shown in yellow. Calcineurin A (CNA) is colored according to its secondary structure: α helices are in cyan, β sheets in blue, and connecting loops in gray; associated Zn2+ and Fe2+ ions at the active site are shown in red. The bound PVIVIT peptide, which forms interactions along the edge of β-14, is shown in orange. (B) X-ray structure of RVxF motif from Gm bound to PP1 [based on (15)]. PP1 is colored according to its secondary structure, as for (A). Mn2+ ions at the active site are shown in red. The RVxF peptide, in yellow, associates with the hydrophobic groove formed by β-14 and β-11.

The RVxF motif specifies binding to PP1

The structure of PP1 cocrystallized with a short peptide derived from the G subunit (GM), which targets PP1 to glycogen particles in muscle (15), revealed that six amino acids of GM, RRVSFA, bind 20 Å away from the catalytic site in the β-14 hydrophobic channel (Fig. 1B). Most PP1-interacting proteins contain a version of this short binding sequence, commonly termed the “RVxF” motif, mutation of which abrogates binding to PP1 (Table 1). This motif provides the primary source of specificity for most proteins that interact with PP1.

Table 1

Summary of docking motifs in PP1- and calcineurin-interacting proteins. Abbreviations for the amino acid residues are as follows: A, Ala; C, Cys; D, Asp; E, Glu; F, Phe; G, Gly; H, His; I, Ile; K, Lys; L, Leu; M, Met; N, Asn; P, Pro; Q, Gln; R, Arg; S, Ser; T, Thr; V, Val; W, Trp; and Y, Tyr. x, any amino acid. Any of the amino acid residues within [ ] can appear at that position. {P} denotes that any residues except P can appear at that position.

View this table:

Can this RVxF motif be used to identify the entire complement of PP1-interacting proteins? The short, degenerate nature of the motif makes this task extremely challenging. Sequence analysis of known interacting proteins established a consensus sequence, [R,K]-X(0,1)-[VI]-{P}-[FW], which fits the PP1-binding motif of 90% of all known PP1-binding proteins (17) (Table 1). However, this motif occurs randomly in ~25% of all proteins, which reduces its usefulness for the de novo identification of PP1-interacting proteins. Targeted structural analyses defined a more restricted consensus sequence that identifies only 40% of known PP1-interacting proteins, but gives a lower number of false-positives (18) (Table 1). The first comprehensive identification of PP1-interacting proteins in mammals was achieved (19) by identifying proteins with an RVxF motif that conforms to both consensus sequences (17, 18); is conserved in human, rat, and mouse homologs; and is surface-accessible in structural predictions. Seventy-eight of the 115 candidates thus identified were validated by examining their binding to PP1 in vitro and in vivo (19). This more than doubles the number of proteins in the mammalian PP1 interactome and establishes the power of combining bioinformatic and experimental approaches to identify binding partners with limited sequence similarity. However, even this relatively comprehensive approach failed to identify a class of PP1-binding proteins that contain an alternative RVxF type motif (Table 1). Inhibitor-2, a conserved regulator of PP1, binds to the RVxF-docking groove in the phosphatase through a nonconsensus sequence, KSQKW (amino acid residues 44 to 48) (20). Gln and Trp in this docking site make contacts similar to those of Val and Phe in RVxF, with additional side-chain interactions stabilizing the binding (20). Thus, alternative interactions with the same docking groove further broaden the scope of possible sequence variation in PP1-interacting proteins and suggest that additional partners for this phosphatase await discovery.

The PxIxIT motif specifies binding to calcineurin

Elucidating the structural motifs that calcineurin detects in its substrates has also been challenging. In mammals, nuclear localization of the members of the nuclear factor of activated T cells (NFAT) family of transcription factors is triggered by calcineurin-dependent dephosphorylation and promotes the activation of T cells, the development of nervous and vascular tissues, and pathophysiological responses such as cardiac hypertrophy (21). Extensive analyses of these physiologically prominent substrates have been critical for identifying conserved structural features recognized by calcineurin.

The primary site of NFAT binding to calcineurin is a short sequence, distinct from calcineurin-regulated phosphosites, whose consensus among NFAT isoforms (NFATc1 to NFATc4) is PxIxIT (2224). This motif was subsequently identified in calcineurin substrates, regulators, and anchoring proteins (Table 1) (2529), and structural studies determined its docking to a surface on calcineurin analogous to the RvXF-binding surface of PP1 (see below). PxIxIT sequences have also been identified in several substrates in Saccharomyces cerevisiae in which calcineurin mediates survival in response to environmental stress (2934). Although most of these sites contain a proline at position 1, hydrophobic residues at positions 3 and 5, and a hydrophilic residue at position 6, none are a perfect match to the PxIxIT consensus sequence (Table 1), illustrating that, as for the RVxF motif, bioinformatic criteria alone are not sufficient for the identification of PxIxIT motifs. For example, experimental determination of the PxIxIT sequence in regulator of calcineurin 1 (RCAN1) identified a region distinct from the region that was previously assumed, on the basis of sequence similarity, to serve as the docking site (28, 29).

Different PxIxIT sequences confer distinct regulatory properties to substrates. The PxIxIT site of NFATc1, SPRIEIT, has a low affinity for calcineurin [dissociation constant (Kd) ~20 mM], which reflects the transient nature of this enzyme-substrate interaction, and mutations that increase or decrease this affinity perturb NFAT signaling (35, 36). Furthermore, native PxIxIT sequences vary in their strength of binding to calcineurin, with sites from five different yeast substrates displaying a more than 200-fold range in affinity [inhibition constant (Ki) values of 15 to ~3 mM] (29, 34). All of these affinities are well below that of the PVIVIT peptide (Kd ~ 1 mM), a PxIxIT sequence that was selected in vitro for its increased binding to calcineurin (Table 1) (35). PxIxIT is a key determinant of overall substrate affinity for calcineurin, as demonstrated by modification of Crz1, a yeast transcription factor that is evolutionarily distinct from NFAT but regulated by calcineurin in an analogous manner. PIISIQ, the native PxIxIT site in Crz1 (Ki ~15 mM), was mutated to PVIAVN, a low-affinity PxIxIT site from the yeast protein Hph1 (Ki ~ 250 mM). Crz1PVIAVN has decreased affinity for calcineurin and decreases Ca2+-dependent gene expression in vivo. In contrast, Crz1PVIVIT, which contains the high-affinity PxIxIT site, displays increased affinity for calcineurin and gives rise to increased gene expression in vivo (34). Thus, calcineurin-substrate affinity can be fine-tuned through specific changes in the PxIxIT sequence to modify the Ca2+-concentration dependence of signaling. This may be an important factor for signaling specificity in vivo, because Crz1PVIVIT causes growth defects under certain environmental conditions (34), and NFATPVIVIT causes severe developmental abnormalities in mice (37).

The region of calcineurin that interacts with PxIxIT was identified by crosslinking, and the structure of calcineurin bound to PVIVIT was determined by both solution nuclear magnetic resonance (NMR) and x-ray crystallography studies (16, 38, 39). PVIVIT forms contacts along the edge of two β sheets in CNA, β11 and β14, similar to the interaction of RVxF with PP1 (Fig. 1A). This hydrophobic docking groove is remote from the catalytic cleft, which explains the ability of PxIxIT to associate with CNA in the absence of CNB and to interact with both the active and inactive forms of the enzyme (40, 41). PxIxIT is critical for substrate tethering, and whereas the high-affinity peptide PVIVIT antagonizes calcineurin-mediated dephosphorylation of NFAT and other protein substrates, it fails to inhibit the activity of calcineurin toward a small-molecule substrate, p-nitro-phenyl phosphate (pNpp), or a phosphopeptide derived from the RII subunit of protein kinase A (11) (35). Mutation of the PxIxIT-binding surface on calcineurin also severely reduces the function of calcineurin in vivo by compromising the interaction between calcineurin and its substrate rather than by reducing its enzymatic activity (16, 34). Nonetheless, the existence of alternative docking interactions on calcineurin is indicated by the resistance of some substrates to inhibition by PxIxIT-related peptides (35, 36).

Substrates and Regulators of PPPs Make Critical Secondary Contacts

The impact of secondary sites on phosphatase-binding specificity is dramatically illustrated by crystallographic analysis of PP1 in complex with the N-terminal domain of myosin phosphatase targeting subunit 1 (MYPT1), a protein that targets PP1 to, and increases its specificity for, myosin light chain. In addition to an RVxF-type motif (K35VKF38), two other contacts occur between PP1 and MYPT1 (42) (Fig. 2). The N terminus of MYPT1 wraps around PP1 and interacts with the base of its catalytic cleft, and ankyrin repeats in MYPT1 interact with and stabilize the extended C terminus of PP1. These secondary interactions alter the shape and charge distribution of the catalytic cleft of PP1, and it is this extended surface that adopts myosin-specific characteristics. Thus, the RVxF motif is necessary for anchoring MYPT1 to PP1, but secondary interactions are responsible for conferring the specific enzymatic characteristics of the complex. This paradigm occurs in other regulators of PP1, which similarly contain an RVxF motif that acts in concert with weaker secondary interactions. Bioinfomatic analyses of 143 PP1-interacting proteins elucidated two such motifs (19). A consensus was established for myosin phosphatase N-terminal element (MyPhoNE), which directs binding of the N terminus of MYPT1 to PP1 (42) and occurs at an increased frequency in PP1-interacting proteins (6 out of 143, or 4.2%) compared to all human proteins in the ENSEMBL database (47 out of 30,086, or 0.16%) (19) (Table 1 and Fig. 2) (43, 44). Similarly, the SILK motif, a PP1-binding sequence previously described in inhibitor-2, was identified in 7 of 143 PP1-interacting proteins (Table 1) (19, 20). These analyses establish that PP1 partners use distinct combinations of binding sites to achieve specific interaction with the phosphatase. Furthermore, the relative contribution of each docking motif to overall affinity varies, because mutation of the same motif in different interacting proteins results in binding defects of differing severity (19).

Fig. 2

Structure of Mypt1 bound to PP1 [based on PDB accession code 11S70 (42)]. PP1 is shown in green, with Mn2+ ions at its active site shown in red. Mypt 1 is depicted in yellow, with two regions of interaction with PP1 shown in orange (the MyPhonE motif at the N terminus of Mypt1 on the left-hand side) and purple (the RVxF motif, in the right hand view). A third contact consists of multiple regions of ankyrin repeats interacting with the extended C terminus of PP1 (at the top of the diagram). Two views differing by 180° rotation are shown.

The LxVP motif directs binding to active calcineurin

An important secondary contact site in calcineurin-interacting proteins is a 15–amino acid region termed the LxVP sequence, which was first identified in NFAT (4548). LxVP is a conserved feature of substrates of calcineurin that is critical for their dephosphorylation (41). LxVP, like PxIxIT, contributes to the overall affinity of the substrate for calcineurin, and variation in the sequence modifies the binding of the substrate to the phosphatase. For example, substitution of the low-affinity LxVP region of NFATc2 with the higher-affinity sequence from NFATc1 results in the increased interaction of NFATc2 with calcineurin (47). The properties of LxVP, however, are distinct from those of PxIxIT. Peptides that contain either motif effectively inhibit the dephosphorylation of NFAT (41, 45, 47), but a peptide containing the LxVP region from NFATc1 specifically inhibits the interaction between NFAT and the activated form of calcineurin (47). Furthermore, in contrast to PVIVIT, the LxVP region of NFATc1 requires both CNA and CNB subunits for binding, inhibits calcineurin-dependent dephosphorylation of the RII phosphopeptide, and stimulates the phosphatase activity of calcineurin against pNpp (41). In silico modeling of the binding of LxVP to calcineurin identified a hydrophobic cleft that is formed at the interface of CNA and CNB, and thus not conserved in PP1 (Fig. 3). Mutations in this region disrupt the binding of LxVP to calcineurin and, when introduced into yeast calcineurin, broadly disrupt its function and interactions with substrates. These substrates include Rcn1, the yeast member of the conserved RCAN family of calcineurin regulators, which contains a consensus LxVP site that is required for binding to calcineurin (41). LxVP sites have been identified in additional mammalian substrates, such as RCAN and kinase suppressor of Ras 2 (KSR2), and the N-terminal amino acids of the RII phosphopeptide, which are required for efficient dephosphorylation by calcineurin, match the consensus sequence DLDVP (11). LxVP sites could not be readily identified, however, by sequence analysis of other yeast substrates. Thus, LxVP sites may be present in a subset of substrates or, more likely, are more variable in sequence than is currently appreciated. Experimental identification of LxVP-type sites in additional substrates is required to establish a more accurate consensus for this motif, and structural studies are also needed to more precisely determine its interaction with active calcineurin.

Fig. 3

Predicted sites of binding on calcineurin for LxVP in calcineurin overlap with contact regions for cyclosporin A–cyclophilin. The x-ray structure of calcineurin in complex with cyclosporin A–cyclophilin is shown [based on PDB accession code 1MF8 (61)]. CNB is in green, with bound Ca2+ ions in yellow. Cyclosporin is shown in magenta, cyclophilin in blue, and CNA in cyan, with Trp352 and Phe356, two predicted sites of contact for LxVP, shown in orange (41).

Differing roles of LxVP- and PxIxIT-docking sites in calcineurin substrates

Although LxVP and PxIxIT are both conserved features in substrates of calcineurin, these motifs seem to play distinct roles in substrate dephosphorylation. PxIxIT binds to calcineurin on a surface behind the catalytic cleft independently of its activation state and likely promotes dephosphorylation by increasing the local concentration of substrate, a mechanism similarly proposed for docking sites on protein kinases (16, 34). In contrast, the LxVP site interacts with its target region only in activated calcineurin, suggesting that this second interaction allows proper orientation of substrate phosphosites relative to the catalytic cleft. This may explain the ability of calcineurin to dephosphorylate a large number of phosphosites (13) in NFAT in a concerted manner (49). It is unclear, however, how binding of LxVP and PxIxIT is spatially or temporally coordinated, especially because the relative spacing and orientation of these sites differ greatly in NFAT and RCAN. As in PPI-interacting proteins, the relative contributions of the two motifs may vary in different substrates of calcineurin. For example, those substrates that either lack PxIxIT sites or contain those with exceptionally low affinity, such as Rcn1 and Hph1 (29, 34), may compensate with stronger interactions at secondary docking sites. Similar to PP1, calcineurin may recognize additional sequences in its interaction partners; thus, obtaining a full complement of calcineurin substrates and regulators will be critical for identifying such motifs. In particular, the surfaces of PP1 that interact with the MyPhoNE and SILK motifs are structurally conserved in calcineurin and might similarly serve as docking surfaces in that phosphatase.

Modulators of PPP Activity Target Active Site and Docking Interactions

PP1 and calcineurin are important cell regulators, and modulation of their activities has substantial implications for human health. In vivo, mechanisms exist for both enzymes to inhibit their active sites directly. Cocrystals of PP1 and inhibitor-2, a potent natural protein inhibitor of PP1, show that in addition to containing an RVxF-related motif and a SILK motif, a third region of inhibitor-2 lies across the active site and makes specific contacts with the catalytic metal ions (20). The structure of this region is exactly analogous to that in calcineurin, in which the C-terminal auto-inhibitory domain interacts with its active site, which suggests a common mode of inhibition that directly blocks catalysis of all substrates (20, 50). The function of calcineurin is also regulated in vivo by the RCAN family of proteins. RCANs use both PxIxIT- and LxVP-docking sites to interact with calcineurin, indicating that they compete with the recognition of other substrates through these sites (28, 29, 41). Furthermore, a conserved C-terminal sequence in mammalian RCAN directly inhibits the activity of calcineurin in vitro (28). However, RCANs have both stimulatory and inhibitory affects on calcineurin activity in vivo (5154), which suggests that the mechanisms by which these proteins modulate calcineurin are likely to be complex. Other endogenous regulators of calcineurin include A-kinase anchoring protein 79 (AKAP-79) (27), calcineurin-binding protein–1 (Cabin-1/Cain) (55, 56), and the A238L protein encoded by Swine fever virus (25, 57). All of these proteins have PxIxIT-mediated docking interactions with calcineurin; however, the mechanisms by which they alter calcineurin activity are currently unknown.

A more sophisticated understanding of the key interaction sites for PP1 and calcineurin may enable the design of more selective inhibitors that disrupt the interaction of the phosphatase with a particular substrate or regulator. These inhibitors would be expected to inhibit competitively, just as peptides that contain interaction motifs (for example, RVxF, PVIVIT, and LxVP) compete with phosphatase-binding proteins in vivo and in vitro. Cell-permeable peptides have been used with some success to inhibit both PP1 and calcineurin (58, 59), and small organic inhibitors of calcineurin appear to inhibit NFAT or PVIVIT binding (60). Furthermore, the immunosuppressants FK506 and cyclosporin A, natural inhibitors of calcineurin that are in wide clinical use to suppress rejection of organs after transplantation, also seem to target the enzyme’s interaction with substrates. FK506 and cyclosporin A combine with the proteins FK506-binding protein (FKBP) and cyclophilin, respectively, and the resulting immunosuppressant-immunophilin complexes (IS-IPs) bind to and inhibit calcineurin. Studies of crystal structures, however, revealed no contact between either IS-IP complex and the active site of calcineurin (9, 50, 61, 62) (Fig. 3), and suggested that these drugs act by sterically blocking the access of large phosphoprotein substrates to the active site.

Studies of the binding of LxVP to calcineurin suggest an alternative explanation (41). Key contact residues for IS-IP complexes overlap with those predicted to bind to the LxVP motif (Fig. 3), and there are striking parallels between the inhibition of calcineurin by LxVP and that by IS-IP complexes. In each case, the activated form of the enzyme is required for binding, dephosphorylation of RII phosphopeptide is inhibited, and phosphatase activity against pNPP is activated (41, 63). Furthermore, IS-IP and LxVP directly compete for binding to calcineurin in vitro, which suggests that IS-IPs inhibit calcineurin by preventing the binding of LxVP sites in substrates to the active phosphatase. In support of this model, enzymatic analyses indicate that cyclosporin A and FK506 are competitive inhibitors of calcineurin, although there is some disagreement about this (64, 65). FK506 and cyclosporin A are thought to universally inhibit calcineurin’s functions; thus, this model of inhibition also predicts the presence of LxVP sites in all of the substrates of calcineurin, a hypothesis that can only be addressed once the composition of LxVP sites and their interaction with calcineurin are more fully characterized.

Recent progress has substantially increased our understanding of the structural motifs that guide the interactions between phosphatases and cellular proteins, and established that specific recognition of substrates and regulators occurs through multiple docking interactions at phosphatase surfaces that are distinct from the active site. Future challenges include determining the mechanisms by which these interacting proteins modulate phosphatase activity and developing high-fidelity tools for in silico identification of binding partners for phosphatases. Meanwhile, the lengthening list of substrates and regulators of PP1 and calcineurin is a reminder of the wide-ranging roles that these versatile proteins play in diverse signaling pathways.

Acknowledgments

We acknowledge J. M. Redondo, A. Rodriguez, D. Barford, and P. Hogan for helpful discussion and comments on the manuscript. M.S.C. and J.R. are supported by NIH grant GM-48728.

References and Notes

View Abstract

Navigate This Article