The PP1 binding code: a molecular-lego strategy that governs specificity

Abstract

Ser/Thr protein phosphatase 1 (PP1) is a single-domain hub protein with nearly 200 validated interactors in vertebrates. PP1-interacting proteins (PIPs) are ubiquitously expressed but show an exceptional diversity in brain, testis and white blood cells. The binding of PIPs is mainly mediated by short motifs that dock to surface grooves of PP1. Although PIPs often contain variants of the same PP1 binding motifs, they differ in the number and combination of docking sites. This molecular-lego strategy for binding to PP1 creates holoenzymes with unique properties. The PP1 binding code can be described as specific, universal, degenerate, nonexclusive and dynamic. PIPs control associated PP1 by interference with substrate recruitment or access to the active site. In addition, some PIPs have a subcellular targeting domain that promotes dephosphorylation by increasing the local concentration of PP1. The diversity of the PP1 interactome and the properties of the PP1 binding code account for the exquisite specificity of PP1 in vivo.

Abbreviations

AnkCap

ankyrin repeat cap

BiSTriP

bipartite docking site of SDS22 that interacts with the α4–6 triangular region of PP1

CBM21

carbohydrate binding module family 21

CPI-17

protein kinase C potentiated inhibitor protein of 17 kDa

david

Database for Annotation, Visualization and Integrated Discovery

iASPP

inhibitory member of the apoptosis stimulating proteins of p53 family

IDoHA

inhibitor-2 docking site for the hydrophobic and acidic grooves

MyPhoNE

myosin phosphatase N-terminal element

Mypt1

myosin phosphatase targeting subunit 1

PIP

PP1-interacting protein

PP1

protein phosphatase 1

PPP

phosphoprotein phosphatase

PPP1R

phosphoprotein phosphatase 1 regulatory subunit

Repo-Man

recruits PP1 onto mitotic chromatin at anaphase

SDS22

suppressor 2 of the dis2-mutant

SH3

SRC homology 3

SpiDoC

spinophilin docking site for the C-terminal groove

Introduction

The superfamily of phosphoprotein phosphatases (PPPs) comprises Ser/Thr protein phosphatases 1–7 (PP1–7) [1,2]. All PPP members have a structurally related catalytic core and identical reaction mechanism but a distinct set of substrates and interacting proteins. Collectively, PP1–7 catalyze over 90% of all eukaryotic protein dephosphorylation reactions [1,3]. The most important protein phosphatase in terms of substrate diversity is PP1, predicted to hydrolyze the majority of Ser/Thr-linked phosphate ester bonds in eukaryotic cells [4]. This implies that PP1 counteracts hundreds of distinct Ser/Thr protein kinases. Consistent with this seemingly promiscuous action, PP1 has a shallow active site and dephosphorylates many phosphoproteins in vitro [4,5]. Numerous enzymes, including protein Ser/Thr kinases and tyrosine phosphatases, have diversified during evolution by gene duplication and speciation, but this clearly does not apply to PP1 [6]. Indeed, eukaryotes express at most a few closely related isoforms of PP1 (α, β, γ1 and γ2 in mammals) and these cannot be distinguished by their enzymatic properties. Also, PP1 has barely changed during more than 1 billion years of eukaryotic evolution, as illustrated by the > 80% sequence identity of PP1 from yeast and man, and the ability of human PP1 to rescue the lethality of a loss of PP1 in yeast [7].

The broad substrate specificity of PP1 in vitro and identical enzymatic properties of its isoforms have contributed to the widespread belief that PP1 serves a rather passive function to end signaling by protein Ser/Thr kinases. However, this view is at variance with numerous in vivo observations showing that PP1 acts in a highly specific and regulated manner [3,4]. The extraordinary specificity of PP1 in vivo can be explained by its ability to form complexes with a large number of structurally unrelated proteins that control its localization, activity and substrate specificity. For vertebrates, close to 200 PP1-interacting proteins (PIPs) have already been identified (Table 1), which create a vast array of PP1 holoenzymes with a distinct set of substrates and mechanisms of regulation. Thus, PP1 has mainly diversified during evolution by expansion of its regulatory toolkit.

Table 1. The PP1 interactome of vertebrates. The gene names (alphabetical order) and protein names of all validated vertebrate PIPs are given. Note that many PIP encoding genes have a PPP1R (phosphoprotein phosphatase 1 regulatory subunit) based nomenclature. If the gene already had another name before its description as a PIP, the PPP1R nomenclature has been adopted as an alias by the HUGO Gene Nomenclature Committee (http://www.genenames.org).

Gene	Protein	Gene	Protein	Gene	Protein
AATK	LMTK1	KIAA0430	LIMKAIN b1	PPP1R21	CCDC128
AHCYL1	IRBIT	KIAA1244	KIAA1244	PPP1R26	DRIM BP
AKAP1	AKAP149	KIF18A	KIF18A	PPP1R27	DYSFIP1
AKAP11	AKAP220	LMTK2	KPI-2	PPP1R32	IIIG9
AKAP5	AKAP79	LMTK3	LMTK3	PPP1R35	LOC221908
AKAP9	AKAP450	MAP1B	MAP1B	PPP1R36	LOC145376
ANKRD28	PITK	MAPT	TAU	PPP1R37	LRRC68
ANKRD42	SARP	MCM7	MCM7	PPP1R42	TLRR
APC	APC	MKI67	MKI67	PREX2	PREX-2
AURKA	Aurora-A	MPHOSPH10	MPHOSPH10	PTK2	FAK
AURKB	Aurora-B	MYO16	MYR 8	RB1	RB
AXIN1	AXIN	MYO1D	Myosin-ID	RB1CC1	RB1CC1
BCL2	BCL2	NCOR1	N-Cor	RBM26	RBM26
BCL2L1	BCL-x	NEFL	Neurofilament L	RIMBP2	RIMBP2
BCL2L2	BCL-w	NEK2	NEK2a	RPGRIP1L	RPGRIP1L
BRCA1	BRCA1	NOC2L	NIR	RPL5	L5
CAMSAP3	NEZHa2	NOM1	NOM1	RRP1B	RRP1B
CASC1	CASC1	NONO	P54nrb	RYR1	Ryanodine receptor 1
CASC5	KNL1	OCLN	Occludin	SACS	SACSIN
CASP2	Caspase 2	OPN3	Opsin-3	SFI1	SFI1
CASP9	Caspase 9	ORC5	ORC5	SFPQ	PSF
CCDC8	CCDC8	PARD3	PAR-3	SH2D4A	SH2D4A
CD2BP2	CD2BP2	PCDH11X	Protocadherin 11X	SH3GLB1	Endophilin B1t
CDC25C	CDC25C	PCDH7	Protocadherin 7	SH3RF2	Hepp1
CDCA2	Repo-Man	PCIF1	PCIF1	SLC12A2	NKCCl
CENPE	CENPE	PFKM	PFK-1	SLC7A14	Solute carrier 7A14
CEP192	CEP192	PHACTR3	Scapinin	SLC9A1	NHE1
CHCHD3	CHCHD3	PHACTR4	PHACTR1-4	SMARCB1	SNF5
CHCHD6	CHCHD6	PHRF1	PHRF1	SPATA2	SPATA2
CLCN7	CLC7	PKMYT1	PKMYT1	SPOCD1	SPOCD1
CNST	Consortin	PLCL1	PRIP-1	SPRED1	SPRED1
CSMD1	CSMD1	POLD3	p68	SPZ1	SPZ1
CSRNP2	CSRNP2	PPP1R1A	Inhibitor-1	SRSF10	SRp38
CSRNP3	CSRNP3	PPP1R1B	DARPP32	STAU1	Staufen
DDX31	DEAD box protein 31	PPP1R1C	IPP5	SYTL2	SYTL2
DLG2	Chapsyn-110	PPP1R2	Inhibitor-2	TMEM132C	TMEM132C
DLG3	SAP102	PPP1R3A	GM	TMEM132D	TMEM132D
DZIP3	DZIP3	PPP1R3B	GL	TMEM225	TMEM225
EIF2AK2	PKR	PPP1R3C	PTG	TNS1	Tensin 1
EIF2S2	eIF2β	PPP1R3D	R6	TP53BP2	p53BP2
ELFN1	ELFN1	PPP1R3E	KIAA1443	TRA2B	TRA2beta
ELFN2	ELFN2	PPP1R3F	HB2E	TRIM28	KAP1
ELL	ELL1	PPP1R7	SDS22	TRIM42	TRIM42
FARP1	FERM	PPP1R8	NIPP1	TRPC4AP	TRPC4AP
FER	FER kinase	PPP1R9A	Neurabin I	TRPC5	TRP5
FKBP15	FK506BP15	PPP1R9B	Spinophilin	TSC2	TSC2
GPATCH2	GPATCH2	PPP1R10	PNUTS	TSKS	TSKS
GPR12	GPR12	PPP1R11	Inhibitor-3	UBN1	Ubinuclein 1
GRM1	mGlu1	PPP1R12A	MYPT1	URI1	URI
GRM5	mGlu5	PPP1R12B	MYPT2	VDR	Vitamin D receptor
GRM7	mGlu7	PPP1R12C	p84	VPS54	VPS54
GRXCR1	Glutaredoxin	PPP1R13B	p53BP2-like	WBP11	SIPP1
HCFC1	HCF1	PPP1R13L	IASPP	WDR81	WDR81
HDAC6	HDAC6	PPP1R14A	CPI-17	WNK1	WNK1
HSPB6	Hsp20	PPP1R14B	PHI-1	WWC1	KIBRA
HYDIN	HYDIN	PPP1R14C	KEPI	YLPM1	ZAP3
IKZF1	Ikaros	PPP1R14D	GBPI-1	YWHAG	14-3-3 gamma
ITGA2B	Integrin αIIB	PPP1R15A	GADD34	ZBTB38	ZBTB38
ITPR1	IP3R1	PPP1R15B	FLJ14744	ZCCHC9	ZCCHC9
ITPR3	IP3R3	PPP1R16A	MYPT3	ZFYVE1	ZFYVE1
KCNA6	KCNA6	PPP1R16B	TIMAP	ZFYVE16	ENDOFIN
KCNK10	KCNK10	PPP1R17	G-substrate	ZFYVE9	SARA
KDM5B	JARID1B	PPP1R18	Phostensin	ZSWIM3	ZSWIM3

Highly connected proteins function as signaling hubs and come in two flavours: party or sociable hubs have multiple protein interaction domains and bind many ligands simultaneously, whereas date or non-sociable hubs only contain a single interaction domain that usually only binds one or two proteins simultaneously [8]. With nearly 200 validated vertebrate PIPs, PP1 ranks among the single-domain hubs with the most diversified protein interactome. In addition, based on the recovery of previously validated PIPs in an unbiased bioinformatics-assisted PIP identification screen, it was estimated that hundreds of vertebrate PIPs remain to be identified [4,9]. Hence, the numbers of distinct PP1 complexes and Ser/Thr protein kinases (encoded by ∼ 420 genes in mammals) are roughly balanced, strongly suggesting that PP1 displays a stringent substrate specificity at the holoenzyme level.

In the following sections, we give an overview of the properties of the regulatory toolkit of PP1. The initial focus is on the interaction mechanism of PIPs, which involves multiple, conserved binding motifs that form the basis of an emerging PP1 binding code. Subsequently, we discuss how PIPs determine the activity and substrate specificity of PP1. Finally, we explore the domain structure, expression pattern and functional diversity of PIPs.

A PP1 binding code based on the combination of docking motifs

How can PP1 form stable complexes with so many structurally unrelated polypeptides? Although some PIPs only interact with one specific PP1 isoform, most are known to bind all PP1 isoforms. Even if one takes into account this limited isoform specificity, the number of PIPs per PP1 isoform remains huge. The structural basis for the high connectivity of PP1 lies in its surface grooves, which represent interaction sites for PIPs. The binding of PIPs is mediated by docking motifs that are often only four to eight residues long and on average occupy ∼ 400 Ų of the surface of PP1 [4]. Even if one assumes that the entire surface of PP1 is available for interaction with PIPs, there is only place for at most 30 non-overlapping docking sites for PIPs, much fewer than the number of validated PIPs. This implies that PIPs must share docking sites on PP1, a conclusion that is supported by ample experimental evidence (see below). Nevertheless, PIPs interact with PP1 in a highly specific manner because they contain a unique combination of PP1 binding motifs (Fig. 1). The combinatorial potential of PP1 docking motifs is enormous. For example, more than 4000 combinations can theoretically be made with three out of 30 predicted PP1 docking motifs (total surface of PP1/surface occupied on average by each motif). The available crystal structures fully support the concept of a PP1 interaction code and show that various PIPs occupy both overlapping and distinct surface areas of PP1, creating holoenzymes with unique properties [10–14]. The use of a large fraction of the PP1 surface as an interface for PIP binding may have hampered its evolution, in accordance with the general finding that proteins with many interaction partners have a slow evolutionary rate [15].

Figure 1

The PP1 binding code. PP1 has multiple surface grooves that bind PIP docking motifs (middle panel). Some of the better characterized binding sites for specific docking motifs (RVxF, SILK, MyPhoNE, SpiDoC and IDoHA) are shown. PIPs combine PP1 docking sites to form unique PP1 holoenzymes. The illustrations are based on known structures of PP1 with inhibitor-2 (A) [11], Mypt1 (B) [10], spinophilin (C) [13] and spinophilin plus inhibitor-2 (D) [14].

A considerable number of PP1 docking motifs have already been functionally characterized (Table 2). The RVxF-type docking motif is present in nearly 90% of the validated PIPs [4]. It binds to a hydrophobic channel that is remote from the catalytic site and functions as an anchor for PP1 recruitment [16,17]. The associated local increase in the concentration of PP1 may promote secondary interactions with docking motifs that by themselves have an affinity that is too low to form a stable interaction with PP1. A similar function has been proposed for the SILK-type docking motif, which occurs in seven PIPs [9]. The myosin phosphatase N-terminal element (MyPhoNE), present in six PIPs, has a poorly defined function in substrate selection [9,10]. The spinophilin docking site for the C-terminal groove (SpiDoC), which is also present in neurabin, prevents substrate recruitment via the C-terminal groove of PP1 [13]. Inhibitor-2 has a conserved docking site for the hydrophobic and acidic grooves (IDoHA), which emanates from the active site [11]. The IDoHA motif binds as an α-helix and blocks phosphatase activity by covering the catalytic site. Phosphorylated fragments of inhibitor-1 [18,19] and CPI-17 [12] have been proposed to inhibit PP1 by acting as pseudosubstrates, although conclusive data are missing. A bipartite docking site of the leucine-rich repeat protein SDS22 interacts with a triangular region that is delineated by helices α4–6 of PP1 (BiSTriP) [20]. Recently, an RNYF sequence in the SRC homology 3 (SH3) domain of the inhibitory member of the apoptosis stimulation proteins of p53 family (iASPP) was identified as a novel PP1 docking site [21]. Finally, a PP1 docking motif has been described that consists of eight ankyrin repeats of the myosin phosphatase targeting subunit Mypt1 and specifically caps the C-terminus of PP1β (AnkCap) [10]. Other PIPs, including aurora kinase A [22], Repo-Man [23] and some testis-specific PIPs [24,25], specifically interact with PP1α, PP1γ1 and PP1γ2, respectively, indicating that they also contain isoform-specific docking motifs. In the case of the neurabins a sequence C-terminal to the RVxF motif has been identified that may preferentially recognize the C-terminus of PP1γ1 [26–28]. Furthermore, the nucleolar targeting of PP1β and PP1γ1 is mediated by a single N-terminal residue [29], hinting at the existence of a nucleolar PIP with a docking site that is specific for the N-terminus of these PP1 isoforms.

Table 2. PP1 docking motifs. For each PP1 docking motif, a representative PIP and the function of the motif is given.

Motif	PIP	Function
RVxF	G-subunit	PP1 anchoring
SILK	Inhibitor-2	PP1 anchoring
MyPhoNE	Mypt1	Substrate selection
SpiDoC	Spinophilin	Substrate selection
IDoHA	Inhibitor-2	Inhibition
RNYF	iASPP	?
BiSTriP	SDS22	?
Pseudosubstrate	Inhibitor-1	Inhibition
AnkCap	Mypt1	Substrate selection

Properties of the PP1 binding code

As is true for any biological code the PP1 binding code adheres to certain general principles. First, the code is specific for interactions with PP1: none of the PP1 docking motifs is known to bind to other PPP members. This contrasts with other PPP members which can share protein interactors. For example, the TAP42 family member α4 can form stable complexes with PP2A, PP4 and PP6, suggesting that these phosphatases have a partially homologous binding interface [30]. Also, most complexes of PP2A, PP4 and PP6 contain a fixed, member-specific scaffolding subunit that mediates binding to a variety of substrate-specifying/binding subunits, which represents a subunit arrangement that is very different from that of PP1 [1,2]. Calcineurin (PP3, PP2B), the closest relative of PP1, also has three established binding sites for conserved docking motifs in substrates and regulatory proteins, but these are not conserved in PP1 [31]. Intriguingly, two of the calcineurin docking motifs bind to surface grooves that are topologically equivalent to binding sites on PP1. Indeed, the ‘PxIxIT’ calcineurin docking motif binds to a groove that corresponds to a channel on PP1 that binds the RVxF motif, and an autoinhibitory domain of calcineurin blocks the catalytic activity in a manner that is similar to the inhibition of PP1 by the IDoHA motif of inhibitor-2.

Second, the PP1 code is universal: the four supertaxa of the eukaryotic crown express both common and lineage-specific PIPs, but the basics of the PP1 interaction code have not changed during eukaryotic evolution. This is well illustrated by the conservation of the RVxF and SILK-type docking motifs, and their binding grooves [6,9]. From an evolutionary viewpoint two types of PIPs can be distinguished: primary PIPs, which originated as PP1 interactors, and secondary PIPs, which only acquired PP1 binding properties later on during evolution [6]. Since PP1 docking motifs are often short and reside in an intrinsically disordered region (see below), this offers a simple rationale for the rapid expansion of the secondary PP1 interactome during eukaryotic evolution. Indeed, intrinsically disordered regions are prone to random mutations, and if such mutations generate a PP1 docking site that offers a functional advantage it will be conserved.

Third, the PP1 binding code is degenerate. For example, the RVxF motif actually conforms to the consensus sequence [K₅₅R₃₄][K₂₈R₂₆][V₉₄I₆]{FIMYDP}[F₈₃W₁₇], in which the subscript numbers are the percentages of residues’ occurrence in validated PIPs and the braced residues are excluded [9,32]. RVxF variants show different affinities for PP1, which explains why the motif is essential for the binding of some, but not all, RVxF-containing PIPs [17,33]. The degeneracy of docking motifs increases the plasticity of the PP1 binding code. In vivo, this plasticity is further enhanced by regulatory pathways that affect the affinity of docking motifs for PP1, e.g. by phosphorylation of flanking residues [4].

A fourth property of the code is that it is nonexclusive: binding of one PIP does not necessarily exclude the recruitment of a second PIP. A number of trimeric PP1 holoenzymes have been identified where PP1 is associated with one targeting or substrate-specifying PIP and one inhibitory PIP. Strikingly, the inhibitory PIPs (inhibitor-1, inhibitor-2 and CPI-17) and their paralogs can each form trimeric complexes with various PIP–PP1 dimers, leading to the enticing hypothesis that most PP1 dimers have the potential to (transiently) recruit an inhibitory PIP. Trimeric complexes may indeed have escaped detection because the interaction with some inhibitory PIPs is phosphorylation dependent and readily lost during cell fractionation [12]. In a trimeric complex the two PIPs can have nonoverlapping binding sites on PP1 but, at least in some complexes (e.g. Mypt1/PP1/CPI-17), these PIPs can additionally have binding sites for each other [12]. Some PIP–PIP interaction sites, e.g. between SDS22 and inhibitor-3, have been lost during eukaryotic evolution [34,35]. The loss of PIP–PIP interaction sites allows the independent recruitment of these PIPs and increases their combinatorial potential, which may have further enhanced the versatility of the PP1 interaction code during evolution. Intriguingly, in some trimeric complexes the two PIPs compete for a common binding site on PP1, e.g. the RVxF binding channel, as has been reported for the GADD34/inhibitor-1 and spinophilin/inhibitor-2 pairs [14,36]. Since each of these PIPs has multiple PP1 docking sites, competition for a common binding site does not affect the composition of the complex.

A fifth property of the PP1 code is that it is dynamic: PIPs compete with each other for binding to PP1. As a consequence, the composition of the cellular complement of PIP–PP1 complexes is determined by the concentration of all PIPs and their relative affinity for PP1 (isoforms), which is subject to regulation, e.g. by phosphorylation. Consistent with this notion, the transient or stable overexpression of a particular PIP leads to an increased targeting of PP1 to the subcellular compartments that harbor the PIP [9,37]. Importantly, the total PIP level exceeds the concentration of PP1 manifold, ensuring that there is no free pool of PP1 that can dephosphorylate substrates indiscriminately and uncontrolled. In medium spiny neurons, where the total PP1 concentration is estimated to be ∼ 20 μm, the concentration of only a single inhibitory PIP (DARPP-32) already exceeds this value by 2.5-fold [38]. A simple biochemical experiment also nicely illustrates this point: when purified PP1 is added in a threefold molar excess to crude cell fractions, its activity towards glycogen phosphorylase is suppressed by ∼ 60% (Fig. 2). This suggests that the added PP1 is captured by endogenous PIPs, 60% of which are known to restrain the phosphorylase phosphatase activity of PP1 [9]. In further agreement with this interpretation, added PP1 remains fully active towards phosphorylase if the PIPs in the cell fraction are first destroyed by trypsinolysis.

Figure 2

PIPs exist in a molar excess to PP1. The upper panel shows that eukaryotic cells contain numerous PIPs (colored shapes) that compete for binding to the limited pool of endogenous PP1 (grey ovals). A cytoplasmic extract also contains multiple PIPs: some of these are bound to endogenous PP1, others bind to exogenously added purified PP1 (middle left panel). PIPs are destroyed by trypsinolysis, leaving the endogenous and added PP1 as active free catalytic subunits (middle right panel). Following the addition of a threefold excess of purified PP1 to the cytoplasmic extract, the added PP1 activity is only partially recovered, hinting at its complexation with PIPs that inhibit the phosphorylase phosphatase activity (lower left panel). This inhibition is not seen after prior destruction of the PIPs by trypsinolysis. Similar results are obtained with nuclear extracts (our unpublished data).

PIPs direct PP1 function

The binding of a PIP does not bring about major conformational changes in PP1 [5,10–13]. Two PIPs, inhibitor-2 and SDS22, render PP1 sensitive to trypsin [34,39]. This was initially taken as evidence for a conformational change of PP1. However, the structure of the PP1–inhibitor-2 complex revealed that this trypsin sensitivity of PP1 may be due to the loss of a metal ion from the catalytic site [11], although this was not confirmed in a subsequent study [14]. The biological relevance of these findings is unclear but an attractive hypothesis is that inhibitor-2 and SDS22 regulate metal recruitment at the active site, either as a regulatory mechanism or as an essential step in the lifecycle of PP1. Similarly, preliminary evidence suggests that the methylesterase PME-1 and the chaperone PTPA regulate PP2A through the removal and reloading of catalytic metal ions [40].

Inhibitory PIPs block substrate binding, either by covering the active site or by acting as pseudosubstrates [11,12,14]. Substrate-specifying PIPs act in a more sophisticated manner: they prevent the dephosphorylation of substrates by occupying specific substrate-binding grooves or they promote dephosphorylation by providing extended docking sites for substrates [10,13,41]. Biochemical studies have revealed that some PIPs use both positive and negative substrate selection mechanisms [42,43], or restrict the action of associated PP1 to a single phospho-site of a substrate that is phosphorylated on multiple residues [44]. However, more structural data are needed to understand the underlying molecular mechanisms. Some PIPs (additionally) contribute to substrate selection by targeting PP1 to a particular cellular compartment that contains a subset of substrates. In that case, it is the enhanced local concentration of PP1 that drives substrate dephosphorylation. A variant of this type of substrate selection is that some PIPs are themselves substrates for associated PP1 [4].

Structural similarities between PIPs

PP1 docking motifs are usually too short and degenerate to be readily used as a feature to identify PIPs. A property that is shared by a majority of PIPs is that their PP1 binding domain is intrinsically disordered in the unbound state but (partially) ordered in the PP1-bound state [4,5]. The absence of a stable fold in the unbound state is due to a lack of sufficient hydrophobic residues needed to form the core of a globular domain. Intrinsic disorder enables PIPs to reach binding sites on the PP1 surface that are remote from one another, an implicit property of the PP1 binding code. Whereas some intrinsically disordered PP1 binding domains never adopt a structure in the PP1-unbound state, others appear to form transient secondary and tertiary structure elements in a subpopulation (20%–80%) of molecules [45,46]. Some of these structure elements have a conformation that corresponds to that seen in the PIP–PP1 complex. The preferential binding of PP1 to these transiently preformed elements (conformational selection) is thought to be the first step in the formation of a stable PIP–PP1 complex and is followed by an additional folding of the PP1 binding domain. Since PP1 recruitment is associated with the folding (induced fit) of PIPs, it can be argued that some PP1 functions may be independent of its enzymatic activity and involve its chaperone function. Similarly, an activity-independent function has already been reported for PP2A [47].

Analysis using the Database for Annotation, Visualization and Integrated Discovery (david) shows that the 189 validated vertebrate PIPs are significantly enriched in specific domains or motifs (Fig. 3). Compared with other proteins, PIPs more often contain the carbohydrate binding module CBM21 (six PIPs), an ankyrin repeat (13 PIPs), a nucleotide binding domain (20 PIPs), a poly-lysine-rich region (11 PIPs), a poly-proline- or proline-rich domain (29 PIPs), a glutamine or glutamic-acid-rich region (13 PIPs) or an SH3 domain (six PIPs). CBM21 domains mediate the targeting of a subset of PIPs, the G-subunits, to glycogen. Other domains/regions that are often found in PIPs may be implicated in PP1 binding. Actually, ankyrin repeats [10] and SH3 domains [21] are already known to be involved in PP1 binding. It is noteworthy that PP1 has hydrophobic and acidic grooves close to the catalytic site [19], which represent potential docking sites for proline-rich and poly-lysine-rich regions, respectively. Of the 20 PIPs with an ATP binding domain, 12 are protein kinases and several of these are known substrates for associated PP1 [48]. With the exception of a proline-rich domain, none of the domains that are prevalent in PIPs can be found in regulatory subunits of other PPP members, in accordance with the first principle (specificity) of the PP1 binding code.

Figure 3

PIPs are enriched in specific domains. Only domains that are enriched in the 189 vertebrate PIPs are shown (P < 0.05), as derived from a david analysis, version 6.7 [52,53]. For comparison, the same analysis was also performed for five sets of 189 randomly chosen human proteins. The numbers of randomly chosen proteins with a given domain are shown as means ± SEM (n = 5).

Expression pattern and functional diversity of PIPs

Approximately a third of all validated PIPs do not show an obvious cell-type or tissue-specific expression (Fig. 4A). However, subsets of PIPs are rather selectively expressed in brain (Fig. 4B), testis (Fig. 4C) or white blood cells (Fig. 4D). Another set of PIPs is highly expressed in most tissues except white blood cells (Fig. 4E). The relatively specific expression of subsets of PIPs in brain, testis and blood is consistent with the very high expression levels of PP1 in these tissues. Indeed, PP1 protein levels have been estimated to be as high as 20 μm in certain parts of the brain [38] and the splice-variant PP1γ2 is abundantly and rather specifically expressed in testis [25,49,50]. There are no data on PP1-protein expression levels in blood, but microarray analyses indicate that this is the tissue that expresses among the highest levels of PP1 transcripts (Fig. 4F).

Figure 4

Tissue-specific expression of PIPs. (A) The validated vertebrate PIPs were clustered by the gene set enrichment analysis (gsea) online tool based on their expression profiles in normal tissues, as compiled in the Human Tissue Compendium (Novartis), and are shown as a heat map with low (blue) to high (red) expression [54]. Striking examples of tissue-specific expression are shown enlarged in (B–E). (F) The tissue-specific expression of the genes encoding PP1α (PPP1CA), PP1β (PPP1CB) and the splice variants of PP1γ (PPP1CC ).

A david analysis also revealed that most PIPs are localized in the nucleus (71 PIPs), the cytosol (29 PIPs) or associated with the plasma membrane (39 PIPs) (Fig. 5). PIPs are significantly enriched in the nucleus and cytosol, but underrepresented in the mitochondria and Golgi apparatus. However, it cannot be ruled out that the underrepresentation of PIPs in some organelles only reflects a lack of research data. Compared with other proteins, PIPs are 1.2-fold more often associated with multiple compartments. This is consistent with the description of PIPs that contain several substrate targeting domains, which explains why they can be associated with several signaling complexes [4].

Figure 5

The subcellular distribution of PIPs. The numbers of PIPs with a specific subcellular localization were obtained by a david analysis [52,53]. The same analysis was performed for five sets of 189 randomly chosen human proteins. The number of randomly chosen proteins with a given localization are shown as means ± SEM (n = 5).

More than 70% of all validated PIPs have an annotated function (Fig. 6). In accordance with the pleiotropic action of PP1 [3,51], PIPs can be linked to a wide variety of cellular processes. However, they function predominantly in signaling processes, including the regulation of metabolism, protein synthesis, cell-cycle progression and stress responses. In accordance with the high diversity of PIPs in brain (Fig. 4), they have numerous neuron-specific functions. david analysis predicts that many functions for which PIPs are enriched are also performed by PP2A regulatory subunits (data not shown). However, a detailed comparison is difficult because of the relatively low number of PP2A regulatory subunits (encoded by 15 genes). It seems likely that task distribution among PPPs occurs at a lower, functional level, i.e. within the same pathway, to prevent situations where a single phosphatase is responsible for both activation and inhibition of a given pathway. Indeed, several examples exist where PP1 and PP2A have an opposite effect within the same pathway [3].

Figure 6

Functional diversity of PIPs. The bars show the functional categories for which vertebrate PIPs are enriched (P < 0.05), as determined by a david analysis [52,53]. The same analysis was performed for five sets of 189 randomly chosen human proteins and are presented as means ± SEM (n = 5).

Conclusions and perspectives

The concept of a PP1 binding code is in accordance with a plethora of biochemical and structural data generated over the last 15 years. It explains how this single-domain phosphatase can form stable complexes with hundreds of structurally unrelated proteins and provides a rationale for its exquisite specificity in vivo. Undoubtedly, many PP1 docking motifs and their target grooves remain to be identified. An obvious bottleneck is that most PP1 interaction domains are intrinsically disordered, even when bound to PP1, which makes obtaining structural data challenging. However, once the PP1 code is more completely elucidated, it should be possible to screen for novel PIPs using combinations of PP1 docking motif consensus sequences and to make predictions about the binding mode and function of poorly characterized PIPs. There may be a future for designer PIPs, i.e. polypeptides consisting of a subcellular targeting domain and a chosen set of appropriately spaced PP1 docking motifs. This lego-block approach can be used to interfere with specific subsets of PP1 functions in vivo. Finally, small-cell-permeable compounds that mimic PP1 docking motifs or interfere with substrate recruitment have considerable therapeutic potential.

Acknowledgements

We thank Dr Wolfgang Peti (Brown University) for sharing his insights into the structure of PP1 holoenzymes and Dr Elke Van Ael for critical comments on the manuscript. The authors’ work is financially supported by the Fund for Scientific Research – Flanders (Grant G.0478.08) and a Flemish Concerted Research Action (GOA 10/16). B.L. holds a postdoctoral fellowship of the Research Foundation – Flanders (FWO).