|Home | About | Journals | Submit | Contact Us | Français|
Regulation of the Ser/Thr phosphatase protein phosphatase 1 (PP1) is controlled by a diverse array of regulatory proteins. However, how these proteins direct the specificity of PP1 is not well understood. More than 1/3 of the nuclear pool of PP1 forms a holoenzyme with the nuclear inhibitor of PP1, NIPP1, to regulate chromatin remodeling, among other essential biological functions. Here, we show that the PP1-binding domain of NIPP1 is an intrinsically disordered protein, which binds PP1 in an unexpected manner. NIPP1 forms an α-helix that engages PP1 at a novel interaction site, using polar rather than hydrophobic contacts. Importantly, the structure also reveals the first shared PP1 interaction site outside of the RVxF motif, the ΦΦ motif. Finally, we show that NIPP1:PP1 substrate selectivity is determined by altered electrostatics and enhanced substrate localization. Together, our results provide the molecular basis by which NIPP1 directs PP1 substrate specificity in the nucleus.
Protein Phosphatase 1 (PP1, ~38.5 kDa) is a major serine/threonine phosphatase (Bollen et al., 2010; Peti et al., 2012). Mammalian genomes contain three different genes that encode four distinct catalytic subunits of PP1: PP1α, PP1β/δ and the splice variants PP1γ1 and PP1γ2, which differ in the amino acid compositions of the N- and C-terminal extremities. Remarkably, free PP1 has very little intrinsic specificity for substrates. Rather, it is tightly regulated by its interaction with >200 known targeting proteins (Bollen et al., 2010), proteins that localize PP1 to distinct regions of the cell and modulate its substrate specificity. The catalytic site of PP1 is at the intersection of three potential substrate binding sites: the hydrophobic, acidic and C-terminal grooves. In addition, most PP1 regulators and also some substrates interact with PP1 via a primary PP1-binding motif, the RVxF motif (Hendrickx et al., 2009; Meiselbach et al., 2006; Wakula et al., 2003). While this interaction is often necessary for regulatory protein binding, it does not influence the enzymatic activity of PP1, as it is 20 Å away from the active site. Additional docking sites, such as the SILK and MyPhoNE motifs, have been identified but it is currently unclear how they influence the substrate specificity of PP1 (Hendrickx et al., 2009; Hurley et al., 2007; Marsh et al., 2010; Ragusa et al., 2010; Terrak et al., 2004). Thus, a detailed understanding and, equally importantly, the ability to predict how the >200 PP1 targeting proteins direct PP1 specificity from sequence alone, is still missing.
Recently, we have shown that the neuronal PP1 targeting protein spinophilin directs PP1 specificity by a novel mechanism: inhibition by steric occlusion of alternative substrate binding sites (Bollen et al., 2010; Dancheck et al., 2008; Kelker et al., 2009; Marsh et al., 2010; Ragusa et al., 2010). However, it is likely that PP1 employs a multitude of mechanisms to generate specific holoenzymes. The nuclear inhibitor of PP1 (NIPP1) is a key PP1 regulator and functions as a signaling-hub in the nucleus. NIPP1 is one of the evolutionary oldest PP1 regulators and, most critically, >1/3 of the nuclear pool of PP1 forms a holoenzyme with NIPP1 (Jagiello et al., 1995). NIPP1 (38.5 kDa) contains three functional domains: (1) an N-terminal Forkhead Associated (FHA) domain (aa 1–143), which specifically binds p-Thr residues followed by a Pro; (2) a central PP1-binding domain (aa 144–225), with 200RVTF203 forming the RVxF motif and (3) a multifunctional C-terminal domain (aa 226–351) that binds RNA, has endoribonuclease activity and inhibits PP1 using a currently unknown mechanism (Beullens et al., 2000; Jagiello et al., 1995; Jagiello et al., 1997; Trinkle-Mulcahy et al., 1999).
Deletion of the NIPP1 gene in mice causes embryonic lethality (Van Eynde et al., 2004), which can be explained by the essential PP1-dependent functions of NIPP1 in transcription, pre-mRNA splicing, cell cycle progression and/or chromatin remodeling (Bollen and Beullens, 2002; Tanuma et al., 2008; Van Dessel et al., 2010). A number of substrates have been reported for the NIPP1:PP1 holoenzyme; however, how NIPP1 turns PP1 into a specific enzyme is currently not understood. Like other PP1 regulatory proteins (Bollen et al., 2010; Peti et al., 2012; Ragusa et al., 2010), NIPP1 has domains that are not involved in PP1 binding, but instead serve other functions. For example, the N-terminal FHA domain (Mahajan et al., 2008) likely functions as a recruitment platform for substrates, including the pre-mRNA splicing factor SAP155, which is a component of the U2 small nuclear ribonuclear protein complex that recruits NIPP1:PP1 to the spliceosome (Boudrez et al., 2002; Wahl et al., 2009) and CDC5L, which is involved in pre-mRNA splicing in a NIPP1:PP1 dependent manner (Boudrez et al., 2000). Some NIPP1:PP1 substrates, like SAP155 and CDC5L, have overlapping cellular functions, increasing the likelihood that NIPP1 facilitates cross-talk between protein complexes that mediate nuclear dependent processes. However, because the central PP1-binding domain of NIPP1 also inhibits the dephosphorylation of a subset of canonical PP1 substrates, including glycogen phosphorylase a, NIPP1 was originally identified as a PP1 inhibitor.
Here we used a combination of NMR spectroscopy, X-ray crystallography and biochemistry to elucidate, at a molecular level, how NIPP1 binds and directs PP1 substrate specificity. We show that the PP1-binding domain of NIPP1 is highly dynamic in its unbound state. The high-resolution crystal structure of the NIPP1:PP1 holoenzyme, together with a significant number of biochemical experiments, shows that NIPP1 interacts with PP1 at both conserved and novel interaction sites and also reveals how the NIPP1:PP1 holoenzyme achieves substrate specificity. Collectively, this work provides fundamental insights into the molecular basis of substrate selection by the PP1: NIPP1 holoenzyme.
Numerous NIPP1 single and multidomain constructs were tested to determine the optimal PP1-binding domain construct for co-crystallization (Figure 1A). 2D [1H,15N] HSQC NMR analysis of NIPP11-225, comprising both the FHA and PP1-binding domains, showed that, as expected, the FHA domain was well-folded. However, ~100 NH cross-peaks showed very little chemical-shift dispersion in the 1HN dimension, indicating that the central PP1-binding domain of NIPP1 is intrinsically disordered. When produced alone, the PP1-binding domain (NIPP1144-225) showed perfect chemical shift overlap with the collapsed chemical-shift dispersion peaks in NIPP11-225, confirming that NIPP1144-225 is intrinsically disordered (Figure 1B). In order to gain further insights, we used carbon chemical shift analysis to test for preferred secondary structures in NIPP1144-225 (Figure 1C). Multiple preferred secondary structure elements were detected in unbound NIPP1144-225 (Marsh et al., 2006): (1) a ~18% populated α-helix formed by residues 146–154; (2) a ~17% populated α-helix formed by residues 163–171; (3) a ~17% populated β-strand formed by residues 192–197 and (4) a ~16% populated α-helix formed by residues 215–221. Interestingly, the ~17% populated β-strand is just N-terminal to the RVxF (200RVTF203) PP1-docking site of NIPP1, which itself adopts a β-strand ~10% of the time. Taken together, these data show that free NIPP1144-225 has few stretches of preferred secondary structure elements.
Despite the highly dynamic nature of the NIPP1 PP1-binding domain in solution, it is functional as it interacts tightly with both bacterially expressed and native PP1. Isothermal titration calorimetry (ITC) measurements using NIPP11-225 and recombinant PP1α7-330 reported a dissociation constant (Kd) of 17.8 nM (Figure 2A). Similarly, ITC measurements with the NIPP1 PP1-binding domain (NIPP1144-225) and PP1α7-330, reported a Kd of 73 nM. While the latter Kd is slightly higher, these measurements confirm that most of the key direct interactions between NIPP1 and PP1 are conserved in the NIPP1144-225:PP1 complex (Figure S1). Thus, our ITC data agrees well with previous experiments that mapped the PP1-binding region to the central domain of NIPP1 (Beullens et al., 1999).
In order to identify which residues within the PP1-binding domain of NIPP1 interact directly with PP1, we generated a NIPP1144-225:PP1α7-330 holoenzyme in which only NIPP1144-225 was 15N-labeled. As free NIPP1144-225 is an intrinsically disordered protein (IDP) that lacks any significant long range interactions, PP1-bound and PP1-unbound residues of NIPP1144-225 will have significantly different NMR relaxation properties. That is, NIPP1 residues that interact directly with PP1 will be invisible in a 2D [1H,15N] HSQC spectrum, while unbound residues will retain their high quality NMR characteristics because they are free to move independently from their PP1-bound neighbors. Consequently, it is readily possible to identify the NIPP1144-225 residues that do not bind PP1 and, in turn, define the optimal NIPP1-binding domain construct for crystallization efforts. Using this procedure, we determined the optimal NIPP1 PP1-binding domain construct to include residues 158–216 (referred to hereafter as NIPP1158-216). ITC measurements using NIPP1158-216 and PP1α7-330 reported a dissociation constant (Kd) of 104 nM (Figure S1), again confirming that most key interactions are conserved in the NIPP1158-216:PP1 holoenzyme.
To determine how NIPP1 binds PP1 at a molecular level and to gain insights into how NIPP1 directs PP1 specificity, we determined the 2.1 Å crystal structure of the NIPP1158-216:PP17-307 complex (hereafter referred to as NIPP1:PP1) by molecular replacement using PP1 (PDBID 3E7A (Kelker et al., 2009)) as a search model. The space group of the crystal was P212121 with four NIPP1:PP1 complexes in the asymmetric unit (Table I). All four copies of the NIPP1:PP1 complexes are essentially identical (Figure S2). Thus, the remainder of the paper is focused on NIPP1:PP1 complex AB.
The structure of the NIPP1:PP1 holoenzyme reveals that the majority, but not all, of the NIPP1 residues become structured when bound to PP1 (Figure 2B). Specifically, the N-terminal NIPP1 residues (aa 160–184) bind to the bottom of PP1 while the C-terminal NIPP1 residues (aa 199–214) bind to the top of PP1. The intervening residues (aa 185–198), which includes a polybasic stretch that functions as a nuclear localization signal (NLS) and has also been reported to play a role in preventing the dephosphorylation of a subset of PP1 substrates, were not visible in the electron density map. The catalytic site of PP1, which contains two essential Mn2+ ions, is ~20 Å away from both the N- and C-terminal NIPP1 structured domains and thus NIPP1 does not interact with the catalytic site of PP1.
A more detailed analysis of the structure of NIPP1:PP1 shows that NIPP1 interacts with PP1 at three sites (Figure 2C). First, as expected, NIPP1 occupies the RVxF binding groove on PP1 (Figure 3). Second, residues C-terminal to the RVxF motif form a β-strand (β1) that extends the central β-sheet of PP1 by binding to PP1 β-strand β14. A nearly identical RVxF-flanking PP1-docking site was also identified in spinophilin (Ragusa et al., 2010), and thus identifies a novel conserved PP1-binding motif that is shared between PP1 interactors (Figure 4). Third, NIPP1 residues 160–175 fold into an α-helix and interact with PP1 at a novel interaction site that was not previously identified in any other PP1 holoenzyme structures (Figure 5). Collectively, the NIPP1:PP1 complex buries ~3186 Å2 of solvent accessible surface area, which occludes ~12% of the surface of PP1.
Three distinct interaction sites are responsible for mediating NIPP1 and PP1-binding. Interaction site 1, the RVxF interaction, includes NIPP1 residues 199SRVTFS204, which dock to the RVxF binding groove of PP1 (Figures 3A, 3B). Residues Val201 and Phe203 bind in an extended conformation, in a manner nearly identical to that seen in the MYPT1:PP1 (Terrak et al., 2004), inhibitor-2:PP1 (Hurley et al., 2007; Marsh et al., 2010), spinophilin:PP1 (Ragusa et al., 2010) and neurabin:PP1 (Ragusa et al., 2010) holoenzymes. Specifically, NIPP1 residues Val201 and Phe203 bind a deep hydrophobic pocket formed by PP1 residues Ile169, Leu243, Phe257, Arg261, Val264, Leu266, Met283, Leu289, Cys291, and Phe293 (Figure 3B). This interaction is further stabilized by backbone hydrogen bonds between NIPP1 residues 202TFS204 and PP1 residues 289LMC291 and a sidechain/mainchain hydrogen bond between Asp242 (PP1) and Val201 (NIPP1).
PP1 interaction site 2 includes NIPP1 residues Ile209 and Ile210, which dock into a second hydrophobic pocket in PP1 centered on the top of PP1 β-strand β14 (Figure 4A). This hydrophobic pocket is formed by PP1 residues Leu75, Tyr78, Met282, Ile295, Leu296 and Lys297 (Figure 4B). While Ile209 is partially exposed to solvent, Ile210, the NIPP1 residue that becomes most buried upon PP1-binding (even more than Phe203, the ‘F’ of the NIPP1 RVxF motif), is completely occluded from solvent. NIPP1 binding at this site is also stabilized by multiple backbone-backbone, backbone-side chain and side chain-side chain hydrogen bonds. Namely, the position of Tyr78 in the NIPP1:PP1 complex is stabilized by a hydrogen bond between the Tyr78 hydroxyl group and the carboxylate of NIPP1 residue Glu208. A similar hydrogen bond is present on the opposite side of the bound NIPP1 chain, mediated by the hydroxyl of PP1 Tyr255 and the carboxylate of NIPP1 Asp207 (Figure 4B). NIPP1 residues 208–210 also form a short β-strand (NIPP1 β1) that hydrogen bonds with PP1 residues 293–297 (PP1 β14), extending the central PP1 β-sheet by one strand (Figure 4C). Collectively, the hydrophobic interactions between NIPP1 residues Ile209 and Ile210 and PP1 as well as the hydrogen bonds between both PP1 (Tyr) and NIPP1 (Asp/Glu) side chains and parallel β-strands (PP1β14-NIPP1β1) stabilize the interactions at this site.
Unexpectedly, spinophilin makes similar, but not identical, interactions with PP1 (Figure 4D). Namely, spinophilin residues Val458 and Phe459 make hydrophobic interactions with PP1 that are similar to those made by Ile209 and Ile210 in the NIPP1:PP1 holoenzyme (Figure 4E). Specifically, Val458, like Ile209 from NIPP1, is partially exposed while Phe459 interacts in the same PP1 pocket as NIPP1 residue Ile210. However, this PP1 hydrophobic pocket is slightly different between the two holoenzymes due to a change in the conformation of PP1 residue Tyr78 between the NIPP1:PP1 and spinophililn:PP1 complexes. In the spinophilin:PP1 holoenzyme, the PP1 Tyr78 sidechain points away from spinophilin Phe459, which allows the two rings to base stack on one another. In contrast, in the NIPP1:PP1 holoenzyme, the sidechain of Tyr78 rotates by nearly 180°, so that it forms the ‘lid’ of the Ile210 binding pocket. If Tyr78 adopted this conformation when bound to spinophilin, it would clash with spinophilin residue Phe459, rationalizing why the sidechain adopts a distinct conformation in the spinophilin:PP1 complex (Figures 4B,4E). Further comparison with all other available PP1 structures reveals that, in contrast to most PP1 residues, which are largely conformationally invariant, the side chain position of PP1 residue Tyr78 is highly variable, adopting a range of conformations depending on the PP1 interaction partner (Figure S3). Thus, the conformational variability of PP1 residue Tyr78 enables PP1 to create distinct pockets to accommodate different PP1 interacting proteins at this PP1 hydrophobic pocket. Because the PP1 regulatory proteins that engage this site do so via two sequential hydrophobic residues, i.e., via a ΦΦ motif, we have named this the PP1 ΦΦ binding pocket.
Interaction 3, the NIPP1helix interaction, includes NIPP1 residues 160–184, with residues 160–175 forming a 4 turn α-helix (Figures 2C,5A,5B) that docks onto the bottom surface of PP1. Interestingly, in their unbound state, NMR chemical shift analysis showed that these NIPP1 residues also exhibited a ~20% α-helical preference. Thus, this interaction between NIPP1 and PP1 is likely mediated by conformational selection, where a preformed conformation is used to guide an efficient binding event (Marsh et al., 2010). Remarkably, unlike interaction sites 1 and 2, which are dominated by hydrophobic interactions, the NIPP1helix is anchored to PP1 mainly by electrostatic interactions, with very few hydrophobic contacts. The interaction is centered on NIPP1 residue Asn174 and PP1 residue Met190 (Figure 5C, 5D). The NIPP1 helix residues that are most buried upon PP1-binding, in descending order, are Asn174, Ile177, Lys175, Thr167, Thr171 and Asn170, while the most buried PP1 residues are Pro192, Met190, and Pro196. Met190, in particular, adopts a rotamer conformation in the NIPP1:PP1 holoenzyme that differs from all other PP1 structures in order to accommodate the imidazole ring of His173. This allows the side chain of Asn174 to make hydrogen bonds with the backbone amide and carbonyl of Met190. As NIPP1 is the first PP1 regulatory protein that has been shown to bind PP1 at this surface, this PP1 interaction demarcates a novel PP1 interaction site.
While the interface between the NIPP1helix and PP1 is largely polar, the surface of the NIPP1helix that faces the solvent is unusually hydrophobic. In particular, Leu166 and Phe169, which are closest to the anchoring residue Asn174, face outwards away from PP1. Leu180 and Ile182 are also solvent accessible. This causes the positively charged surface of PP1 at the NIPP1helix binding site to become more hydrophobic in nature. In addition, the NIPP1helix termini are conformationally variable. The asymmetric unit of the NIPP1:PP1 crystal contains 4 copies of the NIPP1:PP1 holoenzyme (Figure 5E, Figure S2), which differ from one another primarily in the conformation and number of residues in the NIPP1helix. The residues visible in the NIPP1:PP1 electron density map ranged from 156–183 (holoenzyme copy 4; Figure 5E, far right), to 156–178 (holoenzyme copy 2; Figure 5E, second from right); the NIPP1helix residues visible in all four complexes are 160–178. An overlay of all four copies of the holoenzyme (superimposed using PP1) reveals that the termini of the NIPP1helix are variable, adopting a range of distinct conformations that reflect different local environments in the crystal. In contrast, the interactions made and the conformations adopted by residues at the center of the NIPP1helix (residues 162–175) are structurally conserved (Figure 5F). This shows that NIPP1helix residues 162–175 comprise the core of the NIPP1helix-PP1 interaction.
NIPP1 was originally identified as a potent inhibitor of PP1 (Beullens et al., 1992). However, later data revealed that inhibition of PP1 phosphatase activity by the central domain of NIPP1 is substrate dependent (Beullens et al., 2000). Indeed, while NIPP1143-224 potently inhibited PP1 when glycogen phosphorylase a was used a substrate (a specific substrate of the Gm:PP1 holoenzyme), the dephosphorylation of myelin basic protein by PP1 was only weakly affected. Critically, our crystal structure of the NIPP1158-216:PP1α7-330 holoenzyme revealed that NIPP1 does not bind near the catalytic site of PP1 and thus does not inhibit PP1 activity by blocking access to this site. Consistent with these results, NIPP1158-216 was identified as a non-competitive inhibitor for the dephosphorylation of glycogen phosphorylase a (Figure 6A). In-line with our previous work, we showed that the activity of the NIPP1158-216:PP1α7-330 holoenzyme against glycogen phosphorylase a was ~4–5 fold higher following trypsinolysis (Figure 6B), which digests NIPP1 and releases active PP1 (Beullens et al., 2000). This confirms that NIPP1158-216 inhibits the phosphorylase phosphatase activity of PP1. In contrast, the dephosphorylation of SAP155 and CDC5L was not affected by trypsinolysis of NIPP1158-216:PP1. That is, their dephosphorylation was nearly equally efficiently performed by intact NIPP1:PP1 and free PP1.
Interestingly, the measured Michaelis constant, Km, values of the NIPP1:PP1 holoenzyme for these substrates were similar, i.e. 4.2 ± 0.3 μM for glycogen phosphorylase a, 2.9 ± 0.4 μM for SAP155 and 3.0 ± 0.4 μM for CDC5L (for comparison, diffusion controlled dephosphorylation of pNPP, p-Nitrophenyl Phosphate, has a Km of 4.8 ± 0.1 mM). Thus, while the Km values were nearly identical for these substrates, their kcat values were differentially affected by NIPP1. Finally, the dephosphorylation of glycogen phosphorylase a by PP1 purified from rabbit skeletal muscle (mixture of all PP1 isoforms; Figure 6C) or bacterially expressed PP1α7-330 (Figure S4) was inhibited by NIPP1158-216 but did not affect the dephosphorylation of SAP155 and CDC5L, even when added in a molar excess. This shows that the substrate specificity and activity of the NIPP1:PP1 holoenzyme is independent of the source of PP1.
Previously, the inhibitory activity of NIPP1 was correlated with amino acids N-terminal to the NIPP1 RVxF motif and C-terminal to the NIPP1 helix (191RPKRKRKNSR200) (Beullens et al., 1999). These residues were also shown to function as a nuclear localization sequence. Interestingly, no electron density was observed for the majority of these residues in the NIPP1:PP1 holoenzyme crystal structure, indicating that these NIPP1 residues stay flexible even when bound to PP1. Based on these results, the inability of the NIPP1:PP1 holoenzyme to dephosphorylate glycogen phosphorylase a seems to involve highly dynamic electrostatic interactions.
To gain further insight into this mechanism, we analyzed the MYPT1:PP1 holoenzyme structure (Terrak et al., 2004). MYPT1 was also originally defined as a PP1 inhibitor because it too inhibited the dephosphorylation of glycogen phosphorylase a. However, as observed in the NIPP1:PP1 holoenzyme, the active site of PP1 is also fully accessible in the MYPT1:PP1 holoenzyme. Interestingly, like NIPP1, MYPT1 also has a polybasic stretch (30KRKK33) N-terminal to its RVxF motif (35KVKF38). While electron density was visible for these residues in this complex, these residues have the highest B-factors indicating that they also interact more weakly with PP1. In the MYPT1:PP1 holoenzyme, most interactions in this region are mediated via the backbone, likely enabling the side chains to interact with substrates and/or PP1 itself. To confirm that these polybasic amino acids in NIPP11-224 are important for the inhibition of glycogen phosphorylase a, we created a NIPP1 mutant (191RPKRKRKNSR200 was mutated to 191RPAAAAANSR200). As expected, glycogen phosphorylase a activity assays showed that poly-A mutated NIPP11-224 has a reduced inhibitory potency (Figure 7), which correlates well with previous findings (Beullens et al., 2000). Together, these data show that the stretch of dynamic, polybasic amino acids N-terminally preceding the RVxF-motif residues in NIPP1 and MYPT1 are necessary for the inhibition of the phosphorylase phosphatase activity by these two PP1 interacting proteins.
To understand how the FHA-domain of NIPP1 affects the activity of associated PP1 towards its substrates, we compared the dephosphorylation of CDC5L and SAP155 by free PP1 and the NIPP11-225:PP1 holoenzyme. For these experiments, SAP155 or CDC5L were first phosphorylated with recombinant CycA2/CDK2, which resulted in six (CDC5L) or nine (SAP155) phosphorylated residues, as verified by MS. We have previously shown that phosphorylation by CycA2/CDK2 enables the binding of SAP155 and CDC5L to the FHA domain of NIPP1 (Boudrez et al., 2000; Boudrez et al., 2002).
As shown in Figure 8A, NIPP11-225:PP1 dephosphorylated SAP155 and CDC5L several-fold more efficiently than free PP1. In contrast, the dephosphorylation of pNPP, which does not bind to the FHA domain, was not affected by the presence of NIPP11-225. These data suggested that the recruitment of SAP155 and CDC5L by the FHA domain promoted their dephosphorylation by PP1. Consistent with this notion, the PP1 active site is not altered by NIPP1 binding, as demonstrated by the NIPP1:PP1 crystal structure and by the uniform dephosphorylation efficiency of PP1 and NIPP11-225:PP1 for pNPP (Figure 8A). Therefore, changes outside the PP1 active site must be responsible for the increased dephosphorylation efficiency of SAP155 and CDC5L by the NIPP1:PP1 holoenzyme. Specifically, it must result from a modification of the PP1 substrate binding surface, either by a change in the surface electrostatics, by steric blocking, by enlargement of substrate binding sites, by an increase in local concentration of substrates or any combination thereof. Strikingly, the binding of NIPP1 to PP1 has a significant impact on the overall surface charge of PP1. Indeed, the top (RVxF site) of the NIPP1:PP1 holoenzyme becomes nearly entirely negatively charged in the presence of NIPP1, while the bottom (NIPP1helix site) loses a significant positively charged patch and becomes more hydrophobic and acidic (Figure 8B). Thus, it is very likely that this considerable change of surface electrostatics, as already proposed for MYPT1:PP1, together with the recruitment of the substrates by the NIPP1 FHA domain (proximity effect), determines the substrate specificity of NIPP1:PP1. In conclusion, while NIPP1 is a potent inhibitor against a subset of substrates including the most commonly used substrate, glycogen phosphorylase a, NIPP1 is not an inhibitor of the tested physiological substrates.
Our biochemical and NMR data of the NIPP1 PP1-binding domain demonstrates that NIPP1 is an intrinsically disordered protein. NIPP1 undergoes a folding-upon-binding transition upon complex formation with PP1 to form the functional NIPP1:PP1 holoenzyme. As anticipated, NIPP1 binds in the RVxF-binding groove of PP1. However, additional interaction sites, including a hydrophobic interaction with the ΦΦ motif binding grove formed by PP1 Tyr78 and β-strand 14 are necessary for its nanomolar interaction with PP1. Notably, this Tyr78 is conserved in PP1 in nearly all isoforms and species, including all human isoforms and in organisms as diverse as Schistosoma mansoni and Trichoplax adhaerens (the only organism identified without a tyrosine at this position is Lepeophtheirus salmonis, where it is replaced by a cysteine). In contrast, in PP2A and PP3 (PP3, calcineurin), the structurally equivalent residue is replaced by an isoleucine or valine, respectively. Thus, the specificity of PP1 for these regulators is likely mediated in part by the presence of a tyrosine residue at this binding pocket in PP1.
Interestingly, NIPP1 residues 160–175, which only show a ~20% populated α-helix in unbound NIPP1, become fully populated when bound to PP1. In contrast with the RVxF and the ΦΦ hydrophobic pocket interactions, which are mainly stabilized by hydrophobic contacts, the NIPP1helix interaction is stabilized predominately by electrostatic interactions. Importantly, the ΦΦ hydrophobic pocket interaction was also observed in the spinophilin:PP1 complex and is thus the first structurally confirmed interaction site, outside of the RVxF binding groove, that has been detected in multiple PP1 holoenzymes (Peti et al., 2012; Ragusa et al., 2010). Therefore, there is now optimism that the analysis of additional PP1 holoenzymes will lead to a common understanding of how PP1 regulatory proteins bind and direct the activity of PP1.
Direct comparison with the structure of the spinophilin:PP1 holoenzyme shows that while both spinophilin and NIPP1 are IDPs, spinophilin folds completely upon binding to PP1 and becomes rigid, while NIPP1 retains some flexibility as electron density for NIPP1 residues 185–198 was not observed. This shows that different IDP PP1 regulators adopt different levels of structure and rigidity when bound to PP1. We previously showed that spinophilin inhibits the dephosphorylation of the canonical PP1 substrate glycogen phosphorylase a by binding to the C-terminal substrate binding groove and blocking its access to PP1 residue D71, a residue previously shown to be critical for glycogen phosphorylase a binding. Here, we show that a stretch of dynamic polybasic amino acids, N-terminal to the RVxF-motif of NIPP1, is necessary for the inhibition of the dephosphorylation of a subset of PP1 substrates, including glycogen phosphorylase a. These results suggest that the increased, localized positive charge on NIPP1 influences glycogen phosphorylase a substrate binding and/or positioning and thus likely inhibits its PP1-mediated dephosphorylation via a mechanism similar to that proposed for substrate selection by the MYPT1:PP1 holoenzyme, i.e. altered electrostatics (Terrak et al., 2004).
Thus far, two distinct modes of substrate selection by PP1 holoenzymes have been reported. Spinophilin:PP1 selects substrates by sterically occluding specific substrate binding sites (Ragusa et al., 2010). This mechanism for altering the substrate specificity of PP1 is different to that reported for the MYPT1:PP1 holoenzyme, which was proposed to be determined by altered electrostatics, as well as by potentially extending substrate binding grooves (Terrak et al., 2004). The NIPP1:PP1 holoenzyme appears to function in a manner more similar to that of MYPT1:PP1. Specifically, the presence of NIPP1 significantly changes the electrostatic charge distribution of the surface of the NIPP1:PP1 holoenzyme when directly compared with PP1. We have also shown that the NIPP1 FHA domain plays a role in substrate recruitment. Recent peptide:NIPP1 FHA domain NMR interaction studies showed that this interaction is weak, but strong enough to achieve further selectivity for specific phosphorylated substrates (Kumeta et al., 2008).
These observations begin to explain the striking diversity of PP1 holoenzymes, each of which may form a truly unique enzyme with distinctive properties. This is made even more intriguing by the fact that the number of identified PP1 targeting proteins (~200) is still increasing (Bollen et al., 2010; Hendrickx et al., 2009; Peti et al., 2012). If the diversity of interactions observed for PP1 is conserved across ser/thr phosphatases, it would allow the ~40 ser/thr protein phosphatases to form hundreds of unique holoenzymes, ensuring that they are as specific as the 428 known ser/thr protein kinases.
Experimental procedures for cloning and expression, ITC and PP1 inhibition assays are either previously described (Kelker et al., 2009; Peti and Page, 2007) or included as Supplemental Experimental Procedures.
Bacteria were lysed by high-pressure homogenization (Avestin C3 EmulsiFlex) in the presence of EDTA-free protease inhibitor cocktail (Roche). For the expression of NIPP1, CDC5L, SAP155, CDK2 and Cyclin A2, the lysates were clarified by centrifugation at 50,000×g. For PP1, lysate was clarified by ultracentrifugation at 100,000×g. For the purification of NIPP11-225, NIPP1144-225 and NIPP1158-216, clarified lysate was loaded onto a HisTrap HP column (GE Healthcare) and eluted with a 5 – 500 mM imidazole gradient. Peak fractions were pooled and dialyzed overnight at 4°C (20 mM Tris pH 7.5, 50 mM NaCl for NIPP1144-225 and NIPP1158-216, or 200 mM NaCl for NIPP11-225) with TEV protease to simultaneously facilitate his6-tag cleavage. The following day, the cleaved protein was incubated with Ni2+-NTA beads (Qiagen) at 4°C and the flow-through was collected. The intrinsically disordered proteins NIPP1144-225 and NIPP1158-216 were subsequently heat purified at 90°C (600 rpm, 15 min). The supernatant was collected, concentrated to 1 mM, flash frozen in liquid nitrogen and stored at −80°C until further use. Following the Ni2+-NTA subtraction step, NIPP11-225 was further purified by size exclusion chromatography (SEC; Superdex 75 26/60 [GE Healthcare]), concentrated to 1 mM, flash frozen and stored at −80°C. For NMR experiments, 1 mM 13C/15N-labeled NIPP1144-225 was used (20 mM Na-phosphate pH 6.5, 50 mM NaCl, 0.5 mM TCEP, 10% D2O).
SAP155 and CDC5L were purified by HisTrap HP affinity chromatography, dialyzed overnight and purified by gel filtration on Superdex 200 column in 20 mM Tris-HCl pH 8.0, 250 mM NaCl. The proteins were in vitro phosphorylated by incubation with purified CycA2/CDK2 in 50 mM Tris pH 7.5, 2 mM DTT, 2 mM MgCl2, 100 μM ATP, 0.03 mg/mL BSA at 30°C for 90 min. The reaction was stopped by the addition of 10 mM EDTA and phosphorylation was confirmed using Pro-Q Diamond phosphoprotein gel stain (Life Technologies) and MALDI-TOF MS analysis. In addition to Pro-Q Diamond phosphoprotein gel stain analysis for some experiments [γ-32P]ATP was added into the phosphorylation reaction. In these experiments, phosphorylation reactions were stopped by the addition of 20 μM Roscovitine and analyzed by autoradiography. Phosphorylated substrates were flash frozen and stored at −80°C. Glycogen phosphorylase a was prepared as described previously (Beullens et al., 1999).
The CycA2/CDK2 kinase was co-purified by mixing bacterial pellets expressing CycA2 and CDK2 prior to lysis. Clarified lysate containing the kinase complex was immobilized on Ni2+-NTA resin and eluted with 200 mM imidazole prior to overnight dialysis at 4°C in size exclusion chromatography (SEC) buffer (20 mM Tris pH 7.5, 500 mM NaCl, 1 mM DTT). Next, the complex was purified using SEC (Superdex 200 26/60). Peak fractions were again dialyzed overnight at 4°C (20 mM Tris pH 7.5, 150 mM NaCl, 1 mM MgCl2, 0.005% Tween-20, 10% glycerol) flash frozen and stored at −80°C.
To purify the NIPP1158-216:PP1 holoenzyme complex for crystallization trials, a cell pellet expressing PP1α7-307 was lysed in PP1 Lysis Buffer (25 mM Tris pH 8.0, 700 mM NaCl, 5 mM imidazole, 1 mM MnCl2, 0.1% Triton X-100), clarified by ultracentrifugation and immobilized on Ni2+-NTA resin. Bound His6-PP1 was washed with PP1 Buffer A (25 mM Tris pH 8.0, 700 mM NaCl, 5 mM imidazole, 1 mM MnCl2), followed with a stringent wash containing 5% PP1 Buffer B (25 mM Tris pH 8.0, 700 mM NaCl, 0.25 M imidazole, 1 mM MnCl2) and low salt PP1 Buffer A wash (25 mM Tris pH 8.0, 50 mM NaCl, 5 mM imidazole, 1 mM MnCl2) prior to incubation with NIPP1158-216 for 1h at 4°C. The complex was eluted in low salt PP1 Buffer B (25 mM Tris pH 8.0, 50 mM NaCl, 0.25 M imidazole, 1 mM MnCl2). Next, the complex was purified by SEC (Superdex 200 26/60; 20 mM Tris pH 7.5, 500 mM NaCl, 5 mM DTT). Peak fractions were incubated overnight with TEV protease at 4°C. The following day, the NaCl concentration of the sample was reduced to 50 mM prior to purification by ion-exchange chromatography (Mono-Q 5/50 GL [GE Healthcare]) and eluted using a 0.05 – 1 M NaCl gradient. NIPP1158-216:PP1 containing fractions were pooled and SEC purified (Superdex 75 26/60; 20 mM Tris pH 7.5, 500 mM NaCl, 0.5 mM TCEP). Fractions containing the NIPP1158-216:PP1 complex were concentrated to 5.5 mg/mL and immediately used for crystallization trials.
For dephosphorylation assays, purification of PP1α7-330 and NIPP1:PP1 holoenzymes proceeded as described above, using only a single-step Ni2+-NTA affinity purification at 4°C. PP1 was eluted following the stringent wash while NIPP1:PP1 holoenzymes were eluted following complex formation and a low salt PP1 Buffer A wash. The purified proteins were immediately diluted to 0.08 μM and 1 μM for use in protein substrate and pNPP (p-Nitrophenyl Phosphate) dephosphorylation assays, respectively. Native PP1 was purified from rabbit skeletal muscle as described previously (Beullens et al., 1999).
NMR measurements were performed at 298 K on a Bruker Avance 500 MHz spectrometer equipped with a TCI HCN z-gradient cryoprobe. Chemical shift referencing, NMR spectra processing, chemical shift assignments and secondary structure propensity calculations were performed as described (Dancheck et al., 2008).
Crystals (5.5 mg/ml NIPP1158-216:PP1; 20 mM Tris pH 7.5, 500 mM NaCl, 0.5 mM TCEP) were obtained from successive rounds of fine-screening using the sitting drop (200 nl) vapor diffusion method at 4°C. Large, rod-shaped crystals formed in 80 mM Bis-Tris pH 6.9, 320 mM KF, 21% (w/v) PEG1500. Crystals were cryo-protected by a 20 sec soak in mother liquor supplemented with 30% glycerol and immediately flash frozen. A dataset from a single NIPP1158-216:PP17-307 crystal was collected at Beamline X25 at the National Synchrotron Light Source at Brookhaven National Laboratory. The structure was determined using molecular replacement (Phaser in PHENIX (Adams et al., 2010)) using PP1 (PDB: 3EA7, (Kelker et al., 2009)) as the search model in space group P212121 to 2.1 Å resolution. Clear electron density was observed for the portions of the NIPP1158-216 bound to each PP1 in the ASU (4 PP1:NIPP1 complexes). The initial models of the 4 NIPP1158-216 chains were built using Autobuild (PHENIX (Adams et al., 2010)). The structure was completed by successive rounds of manual building using Coot (Emsley et al., 2010) and refinement using PHENIX (Adams et al., 2010). MolProbity was used to evaluate model quality (Chen et al., 2010). 99.9% of all residues were in the allowed region of the Ramachandran diagram, with one outlier (Leu7, chain C). Analyses of protein:protein interactions and buried surface area were carried out using the Protein Interactions Calculator and NACCESS (Tina et al., 2007). Structure figures were created using PyMol (The PyMOL Molecular Graphics System, Version 1.2r3pre, Schrödinger, LLC). Data collection and structure refinement statistics are reported in Table I.
Dephosphorylation reactions were initiated by the addition of 0.02 μM (holo)enzyme to a reaction mixture containing 1 μM phosphorylated CDC5L260-606 or SAP1551-491 to a final enzyme:substrate ratio of 1:50. Reactions were incubated at 30°C for 30 min and quenched by the addition of SDS-PAGE loading buffer and boiling for 10 min at 100°C. Samples were resolved by SDS-PAGE. The phospho-protein and total protein content of substrate bands were detected using Pro-Q Diamond phosphoprotein gel stain and Sypro Ruby protein gel stain, respectively (Life Technologies). The resulting fluorescence signal was measured using a Typhoon 9410 Variable Mode Imager (Ge Healthcare) with excitation and emission wavelengths of 532 nm and 560 nm, respectively. Bands were quantified using ImageQuant TL (GE Healthcare). Pro-Q results were corrected for total protein as previously described (Jacques et al., 2008; Ragusa et al., 2010). The dephosphorylation experiments shown in Figs. 7 and and88 were performed with radioactively phosphorylated SAP155 and CDC5L under the same conditions as described above. The dephosphorylation was quantified by Serenkov counting after SDS-PAGE. The dephosphorylation of glycogen phosphorylase a was measured by quantification of the released Pi, as described previously (Beullens et al., 1999).
The authors thank Dr. M. Allaire (National Synchrotron Light Source, NSLS) for his support at NSLS beamline X25. Use of the NSLS at Brookhaven National Laboratory was supported by the US Department of Energy, Office of Science, Office of Basic Energy Sciences under contract no. DE-AC02-98CH10886. The work was supported by grants from the National Institute of Neurological Disorders and Stroke (R01NS056128) to W.P., the National Institute of General Medicine (R01GM098482) and an American Cancer Society (Research Scholar Grant RSG-08-067-01-LIB) to R.P., and the Fund for Scientific Research-Flanders (Grant G.0478.08) and a Flemish Concerted Research Action (GOA 10/16) to M.B.
All NMR chemical shifts were deposited in the BioMagResBank (BMRB 18198). Atomic coordinates and structure factors have been deposited in the Protein Data Bank (PDB 3V4Y).
Supplemental information includes four figures, Supplemental Experimental Procedures and Supplemental References.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.