Protein Ser/ Thr phosphatases – the ugly ducklings of cell signalling

Authors

  • David L. Brautigan

    1. Department of Microbiology, Immunology and Cancer Biology, Center for Cell Signaling, University of Virginia, School of Medicine, Charlottesville, VA, USA
    Search for more papers by this author

D. L. Brautigan, Department of Microbiology, Immunology and Cancer Biology, Center for Cell Signaling, University of Virginia, School of Medicine, Charlottesville, VA 22908 USA
Fax: +1 434 924 1236
Tel: +1 434 924 5892
E-mail: db8g@virginia.edu

Abstract

This review traces the historical origins and conceptual developments leading to the current state of knowledge of the three superfamilies of protein Ser/Thr phosphatases. ‘PR enzyme’ was identified as an enzyme that inactivates glycogen phosphorylase, although it took 10 years before this ugly duckling was recognized for its true identity as a protein Ser/Thr phosphatase. Ethanol denaturation for purification in the 1970s yielded a phosphatase that exhibited broad specificity, which was resolved into type-1 and type-2 phosphatases in the 1980s. More recent developments show that regulation and specificity are achieved through assembly of multisubunit holoenzymes, transient phosphorylation and the action of inhibitor proteins. Still not widely appreciated, there are hundreds of discrete protein Ser/ Thr phosphatases available to counteract protein kinases, offering potential therapeutic targets. Signalling networks and modelling schemes need to incorporate the full gamut of protein Ser/ Thr phosphatases and their interconnections.

Abbreviations
G1-P

glucose 1-phosphate

MAPK

mitogen-activated protein kinase

PKA

protein kinase

SAPS

Sit4-associated proteins

Overview of protein kinases and phosphatases

Phosphorylation of proteins occurs predominantly on Tyr, Ser and Thr residues. Figure 1 provides a schematic depiction of protein phosphorylation enzymes. Besides these phospho-monoesters, there is phosphorylation of His, Lys and Arg residues, although, for the present discussion, these are not considered further. Transfer of the γ-phosphoryl group from ATP as a donor onto the protein side chains as acceptors is catalyzed by a superfamily of enzymes called protein kinases. There are approximately 500 protein kinases in the human genome, with approximately 90 being specific for Tyr residues, whereas the remaining 400+ kinases phosphorylate Ser/Thr residues. All the protein kinases, with rare exceptions, share a common 3D structure involving a β-sheet N-terminal lobe that binds the ATP and a predominantly α-helical C-terminal lobe that associates with the peptide substrate. The enzymes undergo conformational changes upon binding substrates, with fast phosphotransfer and product release as the rate-determining step. In this sense, the biochemistry of the kinase superfamily is fairly simple because they all have the same structure and all operate with the same mechanism. By contrast, protein phosphatases evolved a diverse biochemistry, with separate superfamilies of enzymes that have different 3D structures, and different active sites with different mechanisms of hydrolysis. There are approximately 90 phosphatases of the protein Tyr and dual specificity (Tyr and Ser/Thr) phosphatase type that employ a common covalent phosphoenzyme intermediate, involving an active site cysteine residue. This mechanism of hydrolysis is shared by the more distantly-related lipid phosphatases, such as PTEN and myotubularin, and by the specialized CDC25, which dephosphorylates and activates cyclin-dependent kinase CDK1. This entire collection of cysteine-based phosphatases is the subject of a separate review by Tonks [0]. Phosphatases reactive with Ser/Thr residues fall into at least three protein superfamilies, called PPP, PPM and DxDxT phosphatases. The distinctions between these families in terms of protein structure and enzyme mechanisms have been reviewed by Shi [1].

Figure 1.

 Protein kinases and protein phosphatases. Showing separation of protein families, with kinases (top box) separated into Tyr and Ser/Thr specific types, depicted as two-lobed structures with ATP in the N-terminal lobe. Protein phosphatases are shown as four families, with the PTP family of active site Cys phosphatases (lower left) subdivided into different groups (PTP, DUSP and PTEN). The Ser/Thr phosphatases are shown in three families, with active sites depicted as bimetallic iron-zinc for the PPP family related to purple acid phosphatase, Mn-Mg for the PPM1 family dependent on added metal ions, and DxDxT for the family that utilizes an Asp-phospho intermediate in hydrolysis.

Origins: the Cori laboratory and phosphorylase

The story of protein phosphatases, and of protein phosphorylation as a regulatory mechanism, can be traced back to research conducted in the 1930s and 1940s, referred to in those early days as ‘mechanisms of hormone action’, which was later called ‘signal transduction’, eventually becoming known as ‘cell signalling’. The basic issue was to understand how blood glucose is regulated by hormones. Glucose in the blood is maintained within a relatively narrow range of approximately 5 mm through the opposing actions of insulin and glucagon, polypeptide hormones that are synthesized and secreted by the pancreas. Previous discoveries had shown that glucose was stored predominantly in the liver and skeletal muscles as a polymer called glycogen. The synthesis of glycogen is stimulated by insulin. By contrast, the degradation of glycogen (called glycogenolysis) is stimulated by glucagon or by adrenergic agents such as adrenaline (epinephrine). Thus, hormones from the pancreas send signals to the liver to either synthesize or degrade glycogen.

Studies of glycogenolysis were pioneered by Carl and Gerty (Radnitz) Cori, a husband-and-wife team. Carl and Gerty were educated in Prague, where they met as students, and then they worked briefly in Vienna before immigrating to the USA. In an old photograph (Fig. 2), we can see Gerty Cory handling a glass pipette in the process of carrying out an experiment, as Carl watches. Note also the fluted filter paper in the funnel in the foreground; this will become important later. For many years, this was the professional arrangement dictated by external circumstances because, in those days, it was not customary that women were appointed to University faculty positions. Carl (but not Gerty) was a Professor. Instead, Gerty worked in the laboratory as a research associate, for a fraction of Carl’s salary, conducting experiments side-by-side with students and postdoctoral students, six of whom went on to be awarded Nobel prizes, including, Christian de Duve, Arthur Kornberg, Edwin Krebs, Luis Leloir, Severo Ochoa and Earl Sutherland. This is one of the most amazing legacies in all of science. Years later, when Carl was department chairman at Washington University in St Louis, Missouri, Gerty finally was promoted, from a postdoctorate to a full professor in 1 year, which is an unusual career path. It was a good thing that this happened quickly because, in the following year (1947), she was the first American woman and, at the time, only the third woman ever, to be awarded the Nobel Prize [2].

Figure 2.

 Photograph from the Nobel Prize website showing Gerty and Carl Cori in their research laboratory. A fluted filter paper cone similar to those used in the phosphorylase preparations is circled in red.

Before there was phosphorylation: the PR enzyme

Carl and Gerty Coris discovered that glycogen was converted to a glucose phosphoester, glucose 1-phosphate (G1-P), by an enzymatic reaction that used inorganic phosphate instead of water, such that, instead of hydrolysis, this reaction is termed ‘phosphorolysis’, and the enzyme is named ‘phosphorylase’. The Coris did not discover phosphorylase, although they carried out pioneering work on its purification, properties and regulation. They showed that it could be used to both synthesize and degrade glycogen. Many new concepts in biochemistry were found with phosphorylase [3]. It is an enzyme that requires a cofactor derived from a vitamin, pyridoxal phosphate, for catalytic activity; it is activated and inhibited via allosteric mechanisms by metabolites (e.g. ATP, AMP, G6P); and it is a multimeric enzyme, forming dimers and tetramers that correspond to less and more active states, phosphorylase b and a, respectively. The addition of AMP to the inactive phosphorylase b greatly increased activity, giving rise to the hypothesis that AMP was an essential cofactor (pyridoxal phosphatase was not yet known to be the actual active site cofactor). When Gerty purified phosphorylase from fresh rabbit skeletal muscle, which represented an abundant source of the enzyme, it was in the active, a form. In a series of back-to-back papers, Gerty and Carl, together with a student, Arda Green, described the isolation, crystallization and kinetic properties of phosphorylase [4]. In 1945, Gerty and Carl published a paper [5] that described the enzyme conversion of phosphorylase a to b catalyzed by ‘PR enzyme’, named as ‘prosthetic group removing’ with the release of what was assumed to be an organic-phosphate species (Fig. 3). AMP, if added, would re-activate the phosphorylase b. At the time, PR enzyme was not considered to be a phosphatase because intestinal (alkaline) phosphatase did not convert phosphorylase a to b. The PR enzyme indeed was an ugly duckling and was not recognized for its true identity as the first ever protein phosphatase. In retrospect, it is interesting to note that PR enzyme (similar to the phosphatases described later) required cysteine as a reducing agent, was activated by manganese ions (not by magnesium) and inhibited by mannose phosphate, glycerophosphate, G1-P and ammonium sulfate. The true identity of PR enzyme as phosphorylase phosphatase remained a mystery for approximately 10 years, and was only solved after the reverse reaction (i.e. the conversion of phosphorylase b to a) was shown to involve phosphorylation of the protein by the action of another enzyme, phosphorylase b kinase.

Figure 3.

 Interconversion of phosphorylase a and b. The diagram depicts how the highly active tetramer is converted by Gerty Cori’s PR enzyme to a lower activity dimeric state. Re-addition of AMP activates phosphorylase to levels almost equivalent to the a form. The PR enzyme was assumed to remove AMP and/or cleave the enzyme in half. Alternatively, 10 years later, Fischer and Krebs that b to a conversion involved the action of phosphorylase b kinase, transferring 4 mol of phosphate groups from ATP to the tetramer of phosphorylase.

Protein kinase comes out of the ashes

The discovery of the first protein kinase (and of protein phosphorylation as a mechanism to regulate protein function) resulted from failed preparations of phosphorylase. As a postdoctoral fellow, Ed Krebs learned to make crystalline phosphorylase a with Arda Green and Gerty Cory in St Louis. After moving to the University of Washington in Seattle in 1948 to take a position as an assistant professor, Ed initially avoided working on phosphorylase so that he did not compete with his mentors. However, his fellow assistant professor Edmond (Eddy) Fischer, who also was hired by the biochemistry department chairman Hans Neurath, encouraged him to study phosphorylase, although Ed became frustrated with the repeated failure of his preparations. He was only able to get the inactive form of the enzyme, phosphorylase b. Krebs and Fischer took to puzzling about this problem. What was the basis for Gerty’s magic in making active phosphorylase a? What was done differently in Seattle compared to in St Louis? It was the filtration step. Gerty was always in a hurry and emphasized that the filtration step of the muscle extract should be performed quickly. This involved using fluted filter papers to line funnels through which the muscle extract was filtered (Fig. 2). The fluted papers would quickly become clogged and the rate of filtration would slow down, at which time (in St Louis) you would pour the contents of the clogged filter into a new fresh filter paper to speed up the process. By the end of the preparation, the cold room would be littered with many sheets of filter paper. However, Krebs was a new investigator with a limited budget and, aiming to conserve supplies, used only one or two filter papers, and this extended the time of the filtration process. Together, Ed and Eddy hit on the idea that maybe there was something about the filter paper that resulted in the recovery of phosphorylase a. They set out to test this idea by using many sheets of filter paper for fast filtration and, indeed, obtained phosphorylase a. If they pre-washed filter paper, the preparation again yielded inactive phosphorylase b. They extracted the filter paper and added the extract back to muscle extracts, which resulted in the conversion of phosphorylase b to phosphorylase a. In a flair for the dramatic, they even burned sheets of filter paper in the hood, gathered the ash and added filter paper ashes (presumably devoid of any organic compounds) and found that this promoted the conversion of phosphorylase b to phosphorylase a. These experiments appear as table VI in their publication in The Journal of Biological Chemistry in 1955 [6]. What was in the filter paper? It had to be inorganic rather than organic. The answer was metal ions, in particular the calcium and manganese ions that were required for the activation of the protein kinase. Krebs and Fisher used γ-labelled radioactive ATP in the b to a conversion reaction to demonstrate covalently bound phosphate in the product. Active phosphorylase a had 1 mol·mol−1 of phosphate introduced by the enzyme they called ‘phosphorylase b kinase’ [7]. The discovery revealed that the PR enzyme identified by Gerty Cory that catalyzed the reverse reaction was a protein phosphatase that dephosphorylated this site, thereby converting active phosphorylase a to inactive phosphorylase b. Later, the cyclic AMP-dependent protein kinase (PKA) was found to phosphorylate and activate phosphorylase b kinase, exposing the principle of a two-step kinase activation cascade that linked the formation of cyclic AMP to the process of glycogenolysis and the release of the Cory ester (G1-P).

No respect for the housekeepers

Discoveries in the 1960s and 1970s related the second messengers cyclic AMP and Ca2+ to the physiological regulation of a variety of biochemical pathways [8]. Changes in the levels of these second messengers were linked to the activation of protein kinases. Glycogen synthesis and degradation, cholesterol biosynthesis, pyruvate metabolism and fatty acid biosynthesis were all pathways where rate-limiting enzymes were found to be controlled as a result of phosphorylation by these kinases. What emerged was a powerful paradigm that dominated research: hormones activated receptors to produce transient increases in intracellular second messengers that activated protein kinases. These kinases phosphorylated intracellular enzymes to activate physiological responses. Return to the basal state involved some poorly defined and relatively nonspecific protein phosphatases whose activity was rather constant and not regulated. The second messengers and the hormones that stimulated their accumulation did not appear to have any affect on the phosphatase activities. Signalling was all about the kinases and these were the enzymes that attracted the most attention. Phosphatases were referred to as housekeeping enzymes, as a relative pejorative term, implying that they were unresponsive to hormonal signalling and therefore physiologically uninteresting. The most popular question asked at conferences, often during the final session that was consigned to phosphatases, was ‘how do we inhibit their activity?’ to eliminate them as a nuisance. Borrowing a line from the comic actor Rodney Dangerfield, phosphatases ‘couldn’t get no respect’. Students and postdoctoral students (including myself) were given friendly advice to avoid research on phosphatases because this was considered as a dead-end for your career. Phosphatases were the ugly ducklings, shunned by the rest of the signal transduction flock.

Nonetheless, a few groups around the world persisted in the study of protein phosphatases. One considerable experimental challenge was phosphatase enzyme assays. Milligram quantities of the substrate protein were needed, and hence the use of phosphorylase or histones, as well as a purified kinase, for the regular preparation of 32P-phosphoprotein substrate. Phosphatase activity was measured by incubation, followed by precipitation of the substrate with ice-cold trichloroacetic acid and scintillation counting of the inorganic 32P phosphate ions in the supernatant after centrifugation [9]. There were strange phenomena, such as increasing specific activity by the dilution of tissue extracts, and multiple peaks on size exclusion and ion exchange chromatography that were not reproducible week-to-week. Publications described the purification of different phosphatases by column chromatography starting from tissue extracts. Eddy Fisher’s group in Seattle maintained a consistent effort working on the traditional rabbit skeletal muscle PR enzyme of Gerty Cory, using phosphorylase a as a substrate. The activity was reported to be relatively resistant to trypsin digestion and denaturation by urea [10]. In Sendai, Japan, at Tohoku University, Shigeru Tsuiki and coworkers isolated and purified different glycogen synthase and histone phosphatases from the liver [11,12]. Histones were readily available and easily phosphorylated as substrates. Tom Langan had purified a liver phosphatase for histone and protamine before 1970 [13]. However, except for enzyme assays at the time, there was no sequence information and no antibodies, nor any basis for a cross comparison of the various preparations. These years were spent in uncharted territory, with few landmarks, and it was hard to tell in which direction to go, as well as how much headway was being made.

From complexity to simplicity

In the 1970s, there were two key advances that changed the field. One was the discovery by Walter Glinsman and F. Huang at NIH-NICHD of two proteins resistant to boiling water temperature that were inhibitors of phosphorylase phosphatase [14]. One protein, called inhibitor-1, required phosphorylation by PKA for its inhibitory activity, whereas the other protein, called inhibitor-2, was inhibitory as purified, and phosphorylation by PKA had no effect on activity. The second advance was the result of a fortuitous error in Ernie Lee’s laboratory in Miami, which was made when attempting to carry out fractionation by ethanol precipitation, as had been developed by Edwin Cohn for the isolation of plasma proteins (e.g. Fraction V serum albumin). Instead of mixing an ice-cold tissue extract with a low proportion of chilled ethanol, 1 : 5 (for 20% final v/v), by mistake, the addition was carried out with five volumes of EtOH into one volume of extract, giving a final solution of 80% ethanol. Even worse, this was performed at room temperature. This causes all the proteins to precipitate and curdle into lumps, with the appearance of cottage cheese. The precipitate was collected by centrifugation, and extracted by homogenizing with buffer in an attempt to redissolve the protein, which did not appear to work. In a stroke of inspiration, an extract of the precipitate was tested for phosphatase activity. Eureka! There was 10-fold more phosphatase activity than was present in the original muscle extract. Published first in BBRC in 1974 [15] and as a complete purification in 1975 [16], this became the worldwide standard protocol for the isolation of protein phosphatase. When applied to various partially purified preparations of phosphatases, the ethanol precipitation procedure yielded what appeared to be the same 35-kDa active enzyme, which was sensitive to Walter Glinsman’s inhibitor proteins. Protein phosphatase was now simple: a 35-kDa enzyme. The concept was that this phosphatase could associate with inhibitory proteins and/or other regulatory subunits that resulted in various higher molecular weight, ‘latent’ or ‘holoenzyme’ forms of phosphatases. Indeed, in a series of studies, Takeda’s group, in Hiroshima, Japan, demonstrated urea and ethanol dissociation of a 35-kDa catalytic subunit from other subunits that corresponded to changes in phosphatase activity with phosphohistones and phosphorylase [17,18]. Protein Ser/Thr phosphatases proved to be multisubunit enzymes with common catalytic subunits.

The great divide: type-1 and type-2 phosphatases

The 35-kDa phosphatase preparation turned out not to be a single enzyme but rather an admixture of distinctive phosphatase catalytic subunits. Thomas Ingebritsen, working as a postdoctoral fellow in Philip Cohen’s laboratory in Dundee, Scotland, partially separated (by diethylaminoethyl column chromatography) two overlapping peaks of phosphatase activity from the extract of the ethanol precipitation step [19]. These two peaks were distinguished in terms of activity using phosphorylate kinase as a substrate and the heat-stable inhibitor proteins. Phosphorylase kinase is phosphorylated on both its α- and β-subunits by PKA. However, the dephosphorylation of the subunits is catalyzed by different phosphatases: one preferentially reacting with the α-subunit, the other preferentially reacting with the β-subunit. The α-subunit phosphatase was resistant to inhibition by inhibitor 1 or inhibitor 2, whereas the β-subunit phosphatase was potently inhibited by both heat-stable inhibitors. Ingebritsen and Cohen used these properties to classify the phosphatases as type-1 or type-2 [20]. They further subdivided the type-2 phosphatases into three subclasses: 2A, 2B, 2C. The type IIB phosphatase turned out to be the calcium and calmodulin-dependent phosphatase known as calcineurin. The type-2C phosphatase was magnesium-dependent and, in subsequent years, sequence analysis showed that this was an entirely different superfamily of phosphatases dependent on added metal ions, called PPM. Nonetheless, the type-1 and type-2 phosphatase classification scheme proved to be very useful for the next 20+ years for providing a means of distinguishing different Ser/Thr phosphatases. Although still widely used, I suggest that this classification scheme is now too simple and perpetuates some misunderstandings of phosphatases. The description of PP1 and PP2A as specific entities marginalizes other members of the PPP phosphatases (such as PP4, PP5, PP6) and does not give adequate attention to the combinatorial diversity of PPP phosphatases. Cells and tissues each have approximately 100 individual Ser/Thr phosphatases, and not simply two types.

Glycogen and myosin targeting subunits show the combinatorial nature of PP1

During the 1980s, Philip Cohen and his coworkers in Dundee, Scotland, pioneered the concept of ‘targeting’ subunits for PP1 that were responsible for distribution of PP1 into different subcellular fractions and restricting the substrate specificity of the 35-kDa catalytic subunit [21]. These targeting subunits presumably were denatured and removed during the ethanol precipitation step of purification. Phosphorylase phosphatase isolated from skeletal muscle is tethered to glycogen particles and negatively regulated during activation of glycogenolysis [22]. Multiple groups had reported a high molecular weight (approximately 250 kDa) form of ‘phosphorylase phosphatase’; however, the targeting subunit was exceedingly sensitive to endogenous proteases and was elusive, appearing as a ladder of bands on SDS/PAGE. Peter Stralfors, Akira Hiraga and Philip Cohen called this subunit G [23] and Anna De Paoli Roach in Indianapolis cloned the gene, referring to it as RG1 [24], and later produced a knockout mouse that had 90% reduction in muscle glycogen accumulation [25]. The RVSF sequence motif for tethering PP1 was first defined in this glycogen-binding subunit [26]. Later, an entire family of structurally related glycogen-binding PP1 subunits was discovered (GL, PTG, R5, R6) [27–30], although the individual functions of these subunits are still not understood.

A second targeted form of PP1 isolated from muscle myofibrils is known as myosin light chain phosphatase. This phosphatase has a myosin targeting subunit, called MYPT1 or MBS of 130 kDa, which was also recovered as a 60-kDa fragment that bound PP1 and myosin. The myosin light chain phosphatase was purified from smooth muscle tissues such as gizzard by Mary Pato at the NIH [31], David Hartshorne at the University of Arizona [32,33], Philip Cohen in Dundee [34] and Tim Haystead at the University of Virginia [35] who purified it from porcine bladder. The results of these studies on glycogen and myosin-associated phosphatases firmly established a multiple subunit model for PP1, based on tethering of the active catalytic subunit to a RVxF motif common to the regulatory subunits. Furthermore, PP1 catalytic subunit was distributed within cells into different compartments by virtue of binding to these regulatory subunits that enhance PP1 activity against certain substrates at the same time as suppressing activity toward phosphorylase a or other substrates.

In August 1985, the first-ever global gathering of protein phosphatase researchers took place in Belgium, organized by Wilfried Merlevede from the University of Leuven and Joseph DiSalvo from the University of Cincinnati. The featured topics were the glycogen- and myosin-targeted forms of protein phosphatase-1. This international conference brought together biochemists who were studying phosphatases primarily related to glycogen metabolism and physiologists who were studying smooth muscle contraction regulated through reversible phosphorylation of the regulatory light chains by calmodulin-dependent myosin kinase and myosin phosphatase. The scientific sessions in Brussels were held in the elegant Palace of the Academy of Sciences (Fig. 4). The Proceedings were published as the first two volumes of ‘Advances in Protein Phosphatases’, comprising a series that ran for 10 years. After this historic conference, phosphatase researchers have gathered regularly over the past 20+ years; in odd-numbered years, at EuroPhosphatase conferences held at different locales in Europe and, in even-numbered years, at a FASEB Summer Conference called ‘Protein Phosphatases’ in Colorado, USA.

Figure 4.

 Photograph of the participants at the first international conference on protein phosphatases in 1985, gathered in the lobby of the Palace of the Academy of Sciences in Brussels, Belgium. Co-organizers Merlevede and DiSalvo are in the front row (centre).

Extending the family of PPP phosphatases

The years following the first international phosphatase conference witnessed the dawn of a new era, where advances in technologies of protein sequencing by vapour-phase Edman degradation and cDNA cloning provided the first protein sequences and highlighted the existence of more than a dozen other protein Ser/Thr phosphatases. Mark Mumby at UT Southwestern in Dallas, Texas, received the Rosa Lowie award as a result of being the person to first clone and sequence a cDNA for a protein phosphatase catalytic subunit, PP2Ac [36]. This sequence was closely related to the PP1 that was being sequenced by Edman degradation [37] and, using Mumby’s DNA for PP2Ac as a probe to clone PP1, Joaqum Arino instead found a second PP2Ac isoform [38]. After he returned to Spain, Arino cloned other phosphatases from yeast, including PPZ1 and PPG [39,40]. Tricia Cohen and her research team in Dundee, including Neil Brewis, Edgar and Odette da Cruz e Silva, Norbert Berndt and Viktor Dombradi, reported sequences for PP1, PP2A and novel phosphatases, called at the time PPX, PPV and PPY [41]. A type-1 Ser/Thr phosphatase was found to be critical for mitosis in fission yeast Schizosaccharomyces pombe, cloned as Dis2 by M. Yanagida at Kyoto University, Japan [42] and as Bws1 by R. Booher and David Beach at Cold Spring Harbor [43]. Yanagida also cloned related phosphatases Sds23 and Ppe1 from S. pombe [44]. Ron Morris at Rutgers University isolated mutants in the fungus Aspergillus called blocked in mitosis (bim) and found that BimG11 was a Ser/Thr phosphatase [45], which could be replaced (rescued) by rabbit muscle PP1 [46], showing remarkable conservation and functional complementation across species. Kim Arndt at Cold Spring Harbor had cloned yeast Sit4 phosphatase (later called PP6) and showed that it was involved in G1 to S phase progression and had distinctive subunits called Sit4-associated proteins (SAPS) [47–49]. Sequence similarities allowed these proteins to all be grouped together as PPP phosphatases.

Members of the PPP family show a remarkable level of sequence identity across species (approximately 80%), which is among the highest degree of conservation for any enzyme. In a prescient analysis, Bruce Averill and John Vincent noted that three regions of sequence in PPP phosphatases could be aligned with purple acid phosphatase [50]. These regions had common side chains that could serve as metal ligands (Y, H, D, E). They proposed that PP1, PP2A and calcineurin (also known as PP3) resemble purple acid phosphatase, which has an oxo-bridged Fe-Zn bimetallic centre as the active site. Atomic absorption detected iron in calcineurin and PP2A [51,52] supporting the proposal; however, recombinant PPPs do not contain iron, and so it is a matter of debate as to whether purple acid phosphatase is an accurate model for PPPs. A purple acid phosphatase bimetallic active site distinguishes PPP enzymes from other Ser/Thr phosphatases (see below). On the other hand, the PPPs sequences are sufficiently distinctive to separate them into subtypes (i.e. PP1, PP2A, PP4, PP5, PP6) in dendograms [1,41]. Each individual subtype shows remarkable specificity in functional complementation across species (Fig. 5). For example, the yeast sit4 mutant phenotype is rescued by Ppe1, the homologous PP6 from S. pombe, or by PPV from Drosophila or by PP6 from humans [53,54], but not by another subtype of the PPP family, such as PP1. One hypothesis accounting for complementation specificity across species is the existence of regulatory subunits that are specific for each PPP catalytic subunit subtype. The regulatory and catalytic subunits are conserved across species in parallel, and function together to form physiological phosphatases. Proteomic analyses by Anne-Claude Gingras in Toronto, Canada, showed that PP2A, PP4 and PP6 phosphatases have dedicated subunits, as well as some common ones [55]. One apparent over-riding principle is that PP1 binds its subunits directly, using the RVxF motif and other sequences, whereas PPP such as PP2A, PP4 and PP6 use scaffold subunits to indirectly bridge together catalytic and regulatory subunits in heterotrimers (Fig. 6).

Figure 5.

 Scheme showing the homologous relationship of the major PPP phosphatases in the eukaryotic species listed on the left. The separate vertical boxes show conservation of function within each type of phosphatase, separate from related types. The curved arrows show functional complementation, where Drosophila or mammalian versions of the genes are able to restore function in yeast.

Figure 6.

 Interaction of PPP subunits. The phosphatase catalytic C subunits interact with regulatory R subunits either through direct binding, as for PP1, or indirectly by associating with scaffold subunits, as for PP2A, PP4 and PP6.

Toxins as tools for PPP phosphatases

Another indication of the close structural similarity among the PPP family phosphatases was inhibition by a diverse set of natural toxins that have gained great popularity as experimental reagents. Okadaic acid was isolated from marine black sponge Halichondria, and Shoji Shibata in Honolulu, Hawaii, showed that it caused contraction of smooth muscles by increasing phosphorylation of the myosin light chain [56,57]. Akira Takai and Corinna Bialojan, in J. C. Ruegg’s laboratory in Heidelberg, showed that this was a result of inhibition of the myosin light chain phosphatase [58]. Okadaic acid was studied by Hirota Fujiki, along with its derivative dinophysistoxin, as tumour-promoting agents, at the National Cancer Center in Tokyo, led by T. Sugimura [59]. Over the next 20 years, thousands of papers were published using okadaic acid as a tool compound to inhibit PPP phosphatases. Cell permeability is a major advantage for the use of okadaic acid in cell biology; however it is not widely appreciated that the transfer of okadaic acid through the membrane into the cytosol is quite a slow process and, even after many hours, the intracellular levels are well below the concentration applied to cells, as demonstrated in phosphatase assays conducted by Brian Hemming’s group in Basel, Switzerland [60]. A different complex natural product toxin, calyculin A, effectively inhibits PPPs at nanomolar concentrations and is more cell permeable, making it an effective research tool for the inhibition of PPPs in intact cells. Fostriecin made by Streptomyces is perhaps the most highly selective inhibitor for type-2 PPPs compared to PP1, as shown by Richard Honkanen at the University of South Alabama, Mobile [61], although this is an unstable compound, requiring special handling precautions that limit its ready applicability. Cyanobacteria (blue green algae) Microcystis, which grow wild in fresh water ponds, produce several cyclic peptide toxins, including microcystin and nodularin, that were studied over 20 years ago as the cause of liver damage in animals [62–64]. Hirota Fujiki in Tokyo found that, similar to okadaic acid, these toxins were potent inhibitors of PPP phosphatases [65]. In the same year, Carol MacKintosh in Dundee and Richard Honkanen also reported microcystin inhibition of PP1 and PP2A [66,67]. Greg Moorhead, now at the University of Calgary, Canada, and Carol MacKintosh introduced microcystin immobilized on beads for the affinity purification of PPP phosphatases [68,69]. Carol MacKintosh, working with Phil and Tricia Cohen, showed that PP1 covalent adducts are formed by Michael addition of a cysteine residue to the double bond in microcystin [70]. Toxins are potent, wide spectrum inhibitors useful in blocking PPP activities, although they do not inhibit other types of Ser/Thr phosphatases, and cannot fully extinguish phosphatase activity during the processing of cell or tissues. Even scraping tissue culture cells in NaCl/Pi reduces the phosphorylation of proteins, and the most effective recovery of phosphoproteins involves the direct quench of cells or tissues into ice-cold trichloroacetic acid, as described by Hayashi et al. [71], and as adapted for use in immunoprecipitation [72]. This answers the most popular question about phosphatases posed by biomedical researchers over the years.

Highlighting individual Ser/ Thr phosphatases

Protein phosphatase-1

PP1 represents the original ‘PR enzyme’ of Gerty Cori and the historical phosphorylase phosphatase. A milestone accomplishment was the crystal structure of the PP1 catalytic subunit, as carried out by John Kuriyan and Angus Nairn at Rockfeller University in New York [73], and in parallel by David Barford, then at Oxford University in England [74]. These structures revealed a surface Y-shaped channel on the front face of the protein, centered on the bimetallic active centre, which is the site of toxin binding, involving an overhanging β12-β13 loop. On the back side of the protein surface lie hydrophobic sockets for the VxF motif of regulatory subunits [26]. Binding to PP1C not only uses the dominant RVxF motif, found in GM and MYPT1 and the majority of regulatory subunits, but also involves other, more distal regions of these protein partners to achieve stable complexes. Much of PP1 partitions into the particulate fraction of cells, where it has a major role in dephosphorylation of cytoskeletal proteins. Mathieu Bollen of Leuven University in Belgium predicts > 200 regulatory subunits for PP1, based on PP1 binding assays that yielded 78 novel RVxF-containing proteins [75]. Association with regulatory subunits regulates PP1 activity in terms of accelerating the kinetics and restricting substrate specificity and these PP1 holoenzymes offer functional diversity to counteract hundreds of protein Ser/Thr kinases [76].

The illustrative example of multisite contact between PP1 and its subunits was visualized in the co-crystal of MYPT1 N-terminal domain in complex with PP1C, solved by R. Dominguez at the Boston Biomedical Research Institute [77]. In addition to the primary RVxF anchor site, a helix N-terminal to the RVxF motif in MYPT1 makes contact with the lower right front of PP1C and the eight ankyrin helical repeats in MYPT1 are arranged end-on across the top of the PP1C, enveloping the C-terminal tail that is extended. The composite surface made of MYPT1 ankyrin repeats and the face of PP1C is imagined as the myosin-specific phosphatase. More recently Wolfgang Peti and Rebecca Page at Brown University in Providence, Rhode Island, determined the structure of the regulatory subunit neurabin bound to parts of the Y-shaped channel on PP1C, and proposed that this accounts for restricted specificity by occupying substrate sites [78]. Furthermore, they studied a trimeric PP1 complex formed with inhibitor-2, even when neurabin occupies the RVxF site on PP1C [79]. This idea of PP1 inhibitor proteins serving as an ‘extra’ subunit of PP1 complexes is supported by results showing that inhibitor-2 binds to PP1C complexed to Nek2, forming a trimeric complex [80,81]. In addition, inhibitor-2 directly binds to PP1C regulatory subunits LMTK2 [82] and neurabin II [83]. Shirish Shenilokar at Duke University and NUS Singapore has elucidated the trimeric arrangement where GADD34 tethers PP1C using RVxF, and this complex binds inhibitor-1 [84]. These examples suggest that PP1 inhibitor proteins are not simply scavengers in cells for monomeric PP1C, as was previously assumed, but are specific regulators of individual PP1 complexes (Fig. 7).

Figure 7.

 Multiple PP1 inhibitor proteins exhibit selectivity for different holoenzymes. Phosphatase inhibitor-2 (I-2) is phosphorylated during mitosis by CDK:cyclin B1 and associates with a subset of PP1 holoenzymes, including three kinase-PP1 pairs, as well as prolyl isomerase Pin1 (solid arrows; box on the right). By contrast, I-2 does not associate with or inhibit other PP1 holoenzymes (dashed lines; box on the left); however, these forms of PP1 are targets of other inhibitor proteins that are substrates for either PKA (Inh1, DARPP32, Inh5) or PKC (CPI-17, KEPI, PHI-1).

Perhaps the most specific PP1 inhibitor of all is CPI-17, a phosphorylation-dependent inhibitor for myosin phosphatase discovered and studied by Masumi Eto of Thomas Jefferson Medical College in Philadelphia [85]. CPI-17 has no RVxF motif but inhibits PP1C, as well as MYPT1::PP1C, when phosphorylated on Thr38, which induces a 1000-fold increase in potency. The specificity of CPI-17 for free PP1C or PP1C bound to MYPT1 compared to other regulatory subunits, such as GM, is attributed to two effects: the relative rates of dephosphorylation of Thr38 and contacts between CPI-17 and both PP1C and MYPT1[86]. Many PP1 complexes react with CPI-17(pT38) as a substrate; with PP1C and holoenzyme MYPT1::PP1C, the rate of hydrolysis is low, such that CPI-17 acts as an inhibitor. There are approximately a dozen different protein inhibitors of PP1, with possibly the most famous being DARPP-32, as studied intensively by Angus Nairn and Paul Greengard at Rockefeller University [87]; however, most of the other inhibitor proteins remain relatively understudied and poorly understood in terms of specificity, regulation and physiological functions.

PP1 is regulated by phosphorylation of both the catalytic and regulatory subunits. Cyclin-dependent kinases [88], as well as Nek2 [80] and LMTK2 [82], phosphorylate the conserved Pro-(Val/Ile)-Thr-Pro-Pro site in the C-terminal tail of PP1C. Phosphorylation reduces activity with exogenous substrates but auto-dephosphorylation reactivates the PP1, so that inactivation is transient. As exemplary regulatory subunits, both GM and MYPT1 are phosphorylated, at multiple sites by multiple kinases, which alters phosphatase activity. We can expect that many of the regulatory subunits of PP1 will be phosphorylated, coordinating the actions of kinases and phosphatases. Overall, this provides three parallel paths for negative regulation of phosphatases by phosphorylation (Fig. 8). Regulation by phosphorylation poses a challenge in phosphatase research because rapid dephosphorylation frustrates attempts to measure changes in phosphatase activity in extracts or immunoprecipitates prepared without addition of phosphatase inhibitors. Thiophosphorylation using ATP-γS, which retards dephosphorylation, represents one approach for studying the effects of phosphorylation on phosphatases [89–92]. The regulatory subunits, in addition to targeting PP1C to glycogen or myofibrils, also act as scaffolds that bind protein kinases and other enzymes, in effect generating signal transduction particles that recollect the glycogen particles studied in the 1960s for their special properties in response to second messengers [22,93].

Figure 8.

 Negative regulation of PPP phosphatases by phosphorylation. The scheme shows how protein Ser/Thr kinases inactivate protein Ser/Thr phosphatases by: (1) direct phosphorylation of the catalytic subunits, which is reversed by intramolecular auto-dephosphorylation; (2) indirect phosphorylation of inhibitor proteins that are holoenzyme specific by contacting both catalytic and regulatory subunits; and (3) phosphorylation of regulatory subunits to induce conformational inhibition of the bound catalytic subunits. Holoenzymes of PPP phosphatases employ contacts with both regulatory and catalytic subunits to achieve specificity with substrates. The definition of a phosphatase for a particular intracellular substrate or process involves specifying the particular catalytic, regulatory and inhibitory components.

Protein phosphatase-2A

Originally purified as a phosphatase using histone or glycogen synthase as substrates, PP2A is an abundant cytosolic heterotrimer of ABC subunits. The early results described the activation of this phosphatase by various polycations (protamine, spermidine, spermine, Mg2+), which might relate to physiological regulation, even though the phenomenon is still not understood and has more or less slipped out of any conversations about PP2A. (Note that the PCM nomenclature for PP2A subunits was to denote poly-cation modulated.) The A subunit (Tpd3 in yeast) structure determined by David Barford was a banana-shaped protein, composed of prototypical HEAT helical repeats [94]. Subsequent structures of ABC trimers, carried out independently at around the same time by Yigong Shi at Princeton University (now in Beijing) [95] and Wenqing Xu at the University of Washington, Seattle [96], revealed that the A subunit was bent into a horseshoe U-shape. The C subunit bound side-by-side with a B subunit to the inter-helical loops on the same face of the ‘U’, at opposite ends of the A subunit. The structures confirm an early model proposed by Gernot Walter of UCSD, La Jolla, CA, and Marc Mumby [97]. More recently, a fragment of the substrate Shugosin (Sgo1) was co-crystalized with PP2A in contact with both the B and C subunits [98], giving us an example of how substrate specificity can be achieved using multiple points of contact on neighbouring subunits. Although yeast only have two B subunits (Cdc55 and Rds1), animals have at least a dozen genes for B subunits in three families referred to by different names (B, B′, B′ or B55, B56, B72) that compete with one another to associate with the AC core dimer. Although small interfering RNA experiments in Drosophila S2 cells suggest that PP2A is an obligate ABC trimer [99], results obtained by Gernot Walter’s group using different monoclonal antibodies argue that AC (as well as ABC) forms co-exist in mammalian cells [100]. Walter’s work has linked human lung and colon cancer to mutations in the A subunits that affect B subunit binding [101,102]. David Pallas, when working in Tom Robert’s laboratory, discovered PP2A in immunoprecipitates of viral antigens [103], and this led to the finding that polyoma small t antigen binds AC core dimer in place of B subunits, effectively compromising PP2A functions in transformed cells [104,105]. This raises a cautionary message about commonly used cell lines HEK293T and COS that express large T and small t compared to their HEK293 and CV-1 counterparts. Signalling is altered in these cells by the small t complex with PP2A [106,107]. Many investigators have contributed to the isolation, cloning and functional characterization of various B subunits, including Jozef Goris, Anna DePaoli-Roach, David Virshup, Stefan Strack, Brian Wadzinski, David Pallas, Angus Nairn and William Hahn (with apologies to those who are not mentioned). The B subunits dictate the specificity, localization and regulation of PP2A.

The catalytic subunit of PP2A has a characteristic conserved sequence TPDYFL at the C-terminus that undergoes post-translational modifications. Jeffry Stock at Princeton University has studied the reversible methyl esterification of the extreme C terminus Leu309 by a S-adenosyl-l-methionine-dependent methyl transferase, and its reversal by PME-1, protein methyl esterase [108,109]. The effects of methyl esterification are controversial; there are claims that this alters phosphatase activity, although the changes are slight and based on in vitro assays that are hard to relate to activity in cells. Other reports indicate the methyl ester of PP2Ac favours the recruitment of B subunits into trimers in living cells but not in vitro assays. Twenty years ago, Tyr phosphorylation of purified AC dimer of PP2A by src, lck or epidermal growth factor receptor kinases was mapped to Y307 in the trypsin-sensitive C terminus [110]. The Tyr phosphorylation inactivated the phosphatase but was reversed by auto-dephosphorylation that could be blocked by okadaic acid. Fibroblasts stimulated by epidermal growth factor or transformed by v-src boosted PP2A Tyr phosphorylation, corresponding to an increase in mitogen-activated protein kinase (MAPK) activation [92]. The availability of phosphosite specific antibodies to pY307 from several different commercial sources has led to hundreds of reports of the physiological regulation of PP2A by this mechanism. Dephosphorylation of pTyr by PP2A led Jozef Goris to isolate a factor called activator PTPA, which turned out some years later to be a proline isomerase [111–113]. Egon Ogris, the conference organizer in Vienna, studied the yeast genes Rrd1 and Rrd2 in a necessary step for the generation of active PP2A group phosphatases [114,115]. The Rrd are related to PTPA, so that Pro isomerization is probably involved, in addition to the insertion of active site metal ions, although the maturation of phosphatase catalytic subunits is still not understood.

Neglected for some time, but now the focus of considerable attention, are the protein inhibitors of PP2A, first purified and described by Zahi Damuni at Penn State Medical School in Hershey, PA [116]. One inhibitor is also known as SET [117], and Danilo Perrotti at Ohio State [118] and Dale Christensen at a company in North Carolina [119,120] have proposed the activation of PP2A as a possible therapy for cancer by targeting SET. Using MS proteomics with a tagged PP2A B56 subunit, Jukka Westermarck of University of Turku, Finland, discovered CIP2A, a cancer inhibitor of PP2A [121], that has been implicated as a prognostic indicator in a variety of carcinomas. Most recently, the proteins ENSA and Arpp19 were reported to be phosphorylation-dependent inhibitors of PP2A and key for the entry into mitosis [122,123]. The Greatwall kinase (Gwl) was shown to activate these selective inhibitors of AB56δC, leading to the activation of Cdc25 and the generation of MPF. Inhibitor proteins for PP2A (Fig. 9) deserve additional attention, and may offer novel routes for selectively regulating (and even activating) phosphatases to counteract different diseases.

Figure 9.

 Protein inhibitors of protein phosphatase-2A. Regulation of PP2A involves the inhibition of activity by proteins SET and CIP2A. Novel therapeutic approaches involve compounds such as ceramide or FTY-720 that interfere with the inhibitors and therefore activate PP2A. Inhibitors ENSA and Arpp19 were reported as being phosphorylation-dependent mitotic inhibitors of a specific PP2A heterotrimer.

Calcineurin (PPP3)

Claude Klee at NIH described calcineurin years ago as a calmodulin-binding protein in the brain [124]. Philip Cohen’s group in Dundee, Scotland, showed that this was a calmodulin-dependent phosphatase and classified it as type PP2B, and it has been called PPP3 [125]. Calcineurin is regulated by RCN (RCAN) proteins, as studied by Kyle Cunningham at Johns Hopkins University, that are able to either inhibit or activate the phosphatase [126,127]. These RCAN, also called calcipressins, may be involved in the maturation of calcineurin and have been implicated in Alzheimer’s disease, Down’s syndrome and cardiac hypertrophy [128]. Martha Cyert at Stanford University has studied the role of calcineurin in yeast, showing that it dephosphorylates the Crz1 transcription factor, allowing nuclear import and activation of stress response genes [129]. This action in yeast is parallel to the action of calcineurin in T cells, as elucidated by Gerald Crabtree at Stanford [130] and Anjana Rao and Patrick Hogan, formerly at Harvard [131]. Calcineurin dephosphorylates the nuclear factor of activated T cells to allow its nuclear import and activation of transcription. Calcineurin is a bona fide drug target that is blocked by the drugs FK-506 and cyclosporine, which have been used clinically for many years to inhibit the host versus graft response following transplantation [132].

Protein phosphatase-4

Called PPX by Tricia Cohen when it was first cloned approximately 20 years ago [133], PP4 is most closely related to PP2A. PP4 does not bind to the A subunit scaffold of PP2A and, instead, has its own dedicated subunits, purified by Brian Wadzinski [134], and proteomic analyses by Anne-Claude Gingras and genetic analyses from Tricia Cohen relate PP4 (Pph3 in yeast) to cisplatin resistance and the DNA damage response [135,136]. PP4 was localized to centrosomes, and it is implicated in centrosome duplication [137–139]. Tse-Hua Tan at Baylor College of Medicine in Houston reported that PP4 knockout in mice is embryonic lethal and PP4 null thymocytes show decreased proliferation and enhanced apoptosis [140]. PP4 also regulates c-Jun N-terminal kinase signalling and responses to tumour necrosis factor [141]. Other work demonstrated PP4 dephosphorylation of histone H2AX during replication and in response to DNA damage and targeting of RPA2 in homologous recombination [142–144]. Association with centrosomes, effects on apoptosis, repair of DNA damage and cisplatin sensitivity show that PP4 has essential and fundamental roles separate from other PPPs in cell signalling.

Protein phosphatase-5

The canonical PPP catalytic domain is fused to a N-terminal TPR domain in the PP5 member of the family [145]. Tricia Cohen in Dundee, Scotland, and Richard Honkanen and Michael Chinkers, both at the University of South Alabama, have studied PP5 and inhibition by toxins. PP5 is activated from the auto-inhibited state by truncation of the C-terminus, or the addition of unsaturated fatty acids, or binding to Rac1, as described by Sandra Rossie at Purdue University and David Armstrong at NIEHS [146–149]. The TPR domain mediates interactions with the glucocorticoid receptor and Hsp90, and so it appears to fulfill a targeting function [146,150].

Protein phosphatase-6

Discovered in 1989 in budding yeast by Kim Arndt, while a postdoctoral student with Gerald Fink [151], Sit4 was one of four Suppressor of Initiation of Transcription defects that resembled PP2A in sequence. This phosphatase is homologous to Ppe1in S. pombe, as studied by M. Yanagida [152], and to PPV in Drosophila that Tricia Cohen’s group used to rescue sit4 mutant yeast, showing that the N-terminal 50 residues determined specificity [53]. Human PP6 was cloned in 1996 and was shown to be the functional homologue of Sit4p and ppe1p [54]. Kim Arndt, at Cold Spring Harbor, showed that Sit4 was key to the expression of G1 cyclins and progression through the cell cycle. He used epitope tagged Sit4 to isolate four different SAPs and showed that the deletion of the SAPs was equivalent to the deletion of Sit4, indicating that the subunits were required for function [49]. Sequences of KIAA1115, KIAA0685 and C11orf23 produced from the human genome aligned with the core domain of the yeast SAPs, and these were shown by graduate student Bjarki Stefansson in my laboratory to be specific subunits for PP6 [153]. SAPS specific for PP6 are key to the activation of DNA-PK and the repair of double-strand DNA breaks following ionizing radiation [154–156]. The human SAPS bind Sit4 in yeast and restore SAP function, showing that the catalytic and regulatory subunits are conserved in concert with one another [157]. In vertebrates, PP6 has another three subunits of the Ankrd (ankyrin repeat domain) protein family that form heterotrimers with the catalytic plus SAPS subunits [158], although little else is known about these holoenzymes.

PPM phosphatases

Purified 30 years ago as a phosphatase activating liver glycogen synthase [159], the Mg2+-dependent Ser/Thr phosphatase (called PP2C) is the founding member of phosphatase superfamily distinct from the PPP. These phosphatases are now known as PPM1 plus a letter (A, B, C, D, etc.) to denote the different genes. Shinri Tamara at Tohoku University in Sendai, Japan, had purified the phosphatase and cloned the DNA, discovering that the sequence more resembled adenylyl cyclase than other Ser/Thr phosphatases [160]. The complete resistance to inhibition of toxins such as okadaic acid, microcystin and calyculin A and the crystal structure of PPM1A determined by David Barford [161] confirmed that these phosphatases are a separate protein superfamily. Despite some efforts, regulatory subunits or inhibitor proteins for PPM phosphatases were not found, and so the enzymes were considered to function as monomers that could associate with stress-related protein kinases [162]. However, in 2009, Erwin Grill’s group in Munich, Germany, and a team led by Sean Cutler at UC Riverside, California, identified the cytosolic receptor for the plant hormone abscisic acid as a PPM1 heterodimer with RCAR/PYR/PYL, the prototypical member of the START family of proteins [163,164]. Binding of PPM1 to RCAR/PYR/PYL enhances affinity for abscisic acid, and the binding of the hormone ligand (or agonists) to the START co-receptor induces a conformation change that results in an inhibition of phosphatase activity. This breakthrough discovery reveals the hormone inhibition of PPM, a negative regulator of signalling. Negative regulation of a negative regulator gives a positive outcome. In addition, Irute Meskiene at the University of Vienna found that the Arabidopsis phosphatase AP2C1 controls the MAPKs, MPK4 and MPK6, in ethylene and jasmonate-mediated stress signalling [165,166]. Plants have as many as 80 genes for PPMs and use them for signalling responses to various hormones.

Returning to the dephosphorylation of glycogen synthase, which occurs in response to insulin, PPM1A is activated by allosteric binding of a putative insulin second messenger glycan called INS2, which Joe Larner at the University of Virginia has studied since the 1970s. The INS2 activates PPM1A, as well as the related mitochondrial phosphatase for pyruvate dehydrogenase, but only with peptide or protein substrates, and not with p-nitrophenyl phosphate as substrate [167]. This opens the possibility that PPM1 phosphatases can be either inhibited or activated by hormone signalling.

The PPM1D phosphatase also known as Wip1 was described by Ettore Appella at NIH [168] as a gene whose transcription is induced by the action of p53 in response to DNA damage, and forms a negative-feedback loop by the dephosphorylation of p53, MDM2 and stress kinases such as ATM, Chk2 and p38MAPK [169–171]. Because it is induced, and protective, PPM1D provides an inviting target for the development of inhibitors as anti-cancer drugs. Sara Lavi at Tel Aviv University in Israel has demonstrated that the inducible expression of PPM1A increases p53 expression and activation [172].

The human phosphatase PPM1G is a component of the spliceosome and binds to protein YB-1 to affect alternative splicing [173]. Binding with YB-1 was enhanced by phosphorylation of the phosphatase, which is another example of conditional binding of a partner subunit to a PPM phosphatase. Phosphatases with a PH domain for association with phosphoinositides were sought out by Alexandra Newton at UCSD who found two PPM enzymes [174,175] and named them PHLPP1 and 2 (pronounced ‘Flip’). PHLPP1 shows preference for the T308 site in Akt and PHLPPs also promote apoptosis via c-Jun N-terminal kinase signalling as a result of dephosphorylating the T387 inhibitory site in kinase MST1 [176].

DxDxT phosphatases and other surprises

At least one other separate superfamily of Ser/Thr phosphatases coexists with the PPP and PPM, called the DxDxT phosphatases based on the active site motif used in hydrolysis. The Fcp1 phosphatase and related Scp1 are specific for the YSPTSPS sequence repeat in the C-terminus of RNA polymerase II [177]. The catalysis probably involves two steps: a Mg2+-assisted formation of an aspartyl-phospho intermediate and its hydrolysis [178,179]. The structural core of Fcp1/Scp1 is similar to bacterial enzymes including haloacid dehalogenase. Antje Gohla in Gary Bokoch’s laboratory purified a cofilin phosphatase [180] that dephosphorylates a Ser3 site to activate cofilin for the dissolution of cortical actin. This phosphatase has the DxDxT motif that is related to haloacid dehalogenase and was named ‘chronophin.’ These DxDxT phosphatases, similar to the PPMs, are insensitive to the toxins that inhibit PPP phosphatases and involve Mg2+, although probably as a result of interaction with the phosphoryl group of the substrate, rather than as a cofactor bound in the enzyme active site.

Lastly, there is the intriguing case of PGAM5, an isoform of metabolic enzyme phosphoglycerate mutase. Kohsuke Takeda at the University of Tokyo found that PGAM5 coprecipitated with tagged stress-activated kinase ASK1 [181]. The PGAM5 is in the mitochondrial membrane and lacks glyceromutase activity, but activates ASK1. This was related to the Ser/Thr phosphatase activity of PGAM5 that required invariant His105 in the catalytic site. It appears that the protein structure and active site can support either one of two types of catalytic activity involving phosphoryl groups. This raises the possibility that there are even more proteins in cells with Ser/Thr phosphatase activity waiting to be discovered.

Perspectives and speculations

The combinatorial nature and complexity of Ser/Thr phosphatases that has retarded research progress over the years offers tremendous potential for a better understanding of their roles in signalling networks and in the development of new, highly specific drugs. Annotations of signalling networks need to recognize multisubunit PPP holoenzymes as discrete entities, with specific feedback interconnections. Compounds will need to avoid targeting the active sites, which has occurred during evolution to produce cytotoxins that poison the essential actions of the PPP phosphatases. Instead, sites on regulatory subunits that dock substrates, or stabilize intersubunit association, or are critical for binding of endogenous inhibitor proteins, hold more promise as drugable chemical space. As we learn from history, phosphatases have proven difficult to study and, similar to ugly ducklings, have been misunderstood and neglected. Structural biology and proteomics have revealed the complexity of phosphatases and propelled knowledge forward to an extent that fresh thinking can be applied about phosphatases. These ugly duckling enzymes have matured into beautiful swans.

One continuing challenge is to establish phosphatase–substrate relationships, determining which phosphatase dephosphorylates a given phosphosite in a cellular protein. This has to move well beyond the time-worn type-1 versus type-2 scheme from 30 years ago and involve the definition of a phosphatase by molecular identification of multiple components, such as the specific isoforms of catalytic and regulatory subunits, any involved inhibitor protein and other protein scaffolds. Assigning phosphatases to substrates in this way is probably best achieved by proteomics. We can reasonably expect that phosphatases will be associated together with individual substrates in stable complexes. These complexes assemble and orient enzyme–substrate complexes for efficient catalysis. Regulation of the activity of these complexes involves phosphorylation of the catalytic and regulatory subunits, as well as the inhibitor proteins, that modulate phosphatase activity in response to signalling, especially by kinases. Second messengers such as Ca2+, lipids, glycans and polyamines or metabolites are potential modulators of Ser/Thr phosphatases. The development of pharmaceuticals to target phosphatases might reduce activity by blocking substrate docking, or compete with (or mimic) allosteric effectors, or increase activity by interfering with endogenous inhibitor proteins. Each approach holds potential, and will require customized assays. With a well-established and robust world-wide kinase-based pharmaceutical business, it is time to reconsider the inclusion of the phosphatases.

There are still some enduring mysteries to be solved in terms of a fundamental understanding of Ser/Thr phosphatases. Phosphatases exhibit stringent negative-feedback regulation of protein levels by mechanisms that have not been elucidated. Levels of PPP catalytic subunits tend to be relatively constant in cells and resist alteration [182]. The overexpression of ectopic subunits suppresses the levels of the endogenous subunits. The effects are dependent on catalytic activity, based on a more effective transient expression of mutationally inactive phosphatases [183], or the treatment of cells with cell permeable toxin inhibitors, which elicit higher levels of protein. Despite the popularity of using small interfering RNA, there are reasons to be skeptical, or at least very cautious, about drawing conclusions from the knockdown of phosphatases because the system resists knockdown to preserve function. Furthermore, the robust catalytic activity of Ser/Thr phosphatases allows for a fraction of the total protein to exert substantial effects. Autoregulation of phosphatase levels has been attributed to both transcriptional and translational effects. Protein turnover maybe be controlled as well (e.g. by a factor such as α-4). By contrast to specific regulatory subunits, α-4 binds multiple phosphatases, PP2Ac, PP4c and PP6c [55,184], and acts as an adaptor for the E3 ligase Mid1 to promote the ubiquitin-mediated degradation of these PPPs via the proteasome [185]. The turnover of phosphatases has not received much attention to date.

Over the years, the PPP phosphatases have been notoriously difficult to express as soluble, active recombinant proteins. Wolfgang Peti recently developed an improved protocol [78], although the post-translational activation of phosphatases by a process called ‘maturation’ remains poorly understood. The PPP and PPM apoproteins need to have metal ions inserted into the active sites, and this likely involves some chaperone-like action. This is especially an issue for the insertion of iron into the bimetallic active site of all of the PPPs. Furthermore, as noted, the PTPA protein is a prolyl isomerase that activates PP2A, and the related yeast Rrd proteins have actions (possibly Pro isomerization) for PP2A and Sit4 maturation. The actions of inhibitor-2 on PP1 have been described as chaperone-like [186,187], and the RCAN proteins have contradictory effects [127] that might arise from an involvement in the maturation processes for calcineurin. Determining the biochemical mechanisms of maturation and activation, defining networks of regulation and signalling, finding specific roles in physiology and producing new pharmacologic agents all represent avenues for future phosphatase research.

Acknowledgements

This article is based on an introductory lecture presented to the 2011 Europhosphatases EMBO conference held at Baden near Vienna, Austria. The idea of organizer, Egon Ogris, was to offer a historical introduction and an overview of the field for the benefit of those who are new to the area of protein phosphatases, which typically has been approximately half of the attendees at these conferences. This perspective is based on my own experience and the accounts of events provided to me by others, and I implore the forbearance of those whose work has not been discussed in as much detail as that of others. I thank the anonymous reviewer and friends who offered a number of specific comments for revision of the review.

Ancillary