PP2A holoenzymes, substrate specificity driving cellular functions and deregulation in cancer

Abstract

PP2A is a highly conserved eukaryotic serine/threonine protein phosphatase of the PPP family of phosphatases with fundamental cellular functions. In cells, PP2A targets specific subcellular locations and substrates by forming heterotrimeric holoenzymes, where a core dimer consisting of scaffold (A) and catalytic (C) subunits complexes with one of many B regulatory subunits. PP2A plays a key role in positively and negatively regulating a myriad of cellular processes, as it targets a very sizable fraction of the cellular substrates phosphorylated on Ser/Thr residues. This review focuses on insights made towards the understanding on how the subunit composition and structure of PP2A holoenzymes mediates substrate specificity, the role of substrate modulation in the signaling of cellular division, growth, and differentiation and its deregulation in cancer.

1. Holoenzyme structure assembly and regulation

PP2A is a major serine/threonine protein phosphatase of the PPP family of phosphatases that is highly conserved from yeast to humans. PP2A makes up close to 1% of total cellular protein in certain tissues, making it one of the most abundant enzymes. In cells, PP2A can exist in two basic protein complex types: a heterodimeric “core enzyme,” consisting of a scaffold (A) subunit and a catalytic (C) subunit; and a heterotrimeric holoenzyme, in which the core dimer is in complex with a regulatory (B) subunit (reviewed in Shi, 2009b). PP2A plays a key role in many cellular processes as it targets a large fraction of the cellular substrates phosphorylated on Ser/Thr residues (reviewed in Eichhorn et al., 2009; Kurimchak and Graña, 2012; Kurimchak and Graña, 2015). There are several genes encoding isoforms within each of the three classes of PP2A subunits, allowing for the potential generation of up to 60 distinct PP2A holoenzymes.

The scaffold subunit of PP2A holoenzyme (also known as the A or PR65 subunit) functions as a bridge that connects the regulatory (B) and catalytic (C) subunits (Fig. 1A). There are 2 known isoforms of PP2A/A, PP2A/Aα and PP2A/Aβ, which share approximately 87% sequence identity (Hemmings et al., 1990). In cells, the PP2A/Aα isoform is approximately 10-fold more abundant than the PP2A/Aβ isoform. The tissue-specific expression of the two isoforms of PP2A/A is variable too, as the PP2A/Aα isoform is expressed across many tissue types, while the PP2A/Aβ isoform is mostly expressed in the testis (Zhou et al., 2003). Structural analysis has revealed that the PP2A scaffold subunit contains 15 tandem HEAT (huntingtin-elongation-A subunit-TOR) repeats, each of which is made up of a pair of antiparallel α-helices (Fig. 1A). These HEAT repeats come together to form an elongated, horseshoe-shaped structure (Groves et al., 1999). This structural composition allows for a high degree of flexibility, as the HEAT repeats are able to twist open to varying degrees to accommodate the binding of the regulatory and catalytic subunits (reviewed in Shi, 2009b). It has been shown that N-terminal HEAT repeats of PP2A/A make contact with the various regulatory subunits, while C-terminal HEAT repeats (11–15) mediate contacts with the catalytic subunit (Cho and Xu, 2007; Wlodarchak et al., 2013; Xu et al., 2008; Xu et al., 2006)(Fig. 1B).

Figure 1. Assembly and composition of trimeric PP2A holoenzymes.

(A) The cartoon depicts the linear pathway for the biogenesis of trimeric PP2A holoenzymes, starting from latent, inactive catalytic C subunit complexed with α4. PTPA displaces α4 and loads catalytic metal ions into the active site, allowing the C subunit to now complex with the scaffold A subunit. The reversible carboxymethylation of L309 of the C-terminal tail of the C subunit is then mediated by the PP2A-specific enzymes LCMT1 and PME1, respectively. The core dimer now complexes with one of four families of B regulatory subunits to form an active, trimeric PP2A holoenzyme. The crystal structures of the core dimer and the trimeric holoenzymes containing B (B55α), B’ (B56γ), B” (PR70), and the coiled coil domain of SG2NA (STRN3) have been solved (see text for details). (B) The cartoon depicts the B regulatory and C catalytic subunit contacts with the 15 HEAT repeats of the scaffold A subunit of PP2A. B55α and PR70 contact HEAT repeats 1–7 of PP2A/A, while B56γ makes contact with HEAT repeats 2–8 of PP2A/A. The catalytic C subunit interacts with HEAT repeats 11–15 of the scaffold A subunit. *The coiled coil domain of SG2NA is known to make contacts with HEAT repeats 1–3 of PP2A/A, although it does not exclude possible contacts with additional HEAT repeats.

Like the scaffold subunit, the catalytic subunit of PP2A (also known as the C subunit) exists in two isoforms, PP2A/Cα and PP2A/Cβ, with PP2A/Cα being about 10-fold more abundant in cells than PP2A/Cβ. The two isoforms share approximately 98% sequence identity, with the most variable region being the N-terminus (Stone et al., 1987). The catalytic subunit interacts with the C-terminal HEAT repeats of the scaffold subunit (Fig. 1B); this is different from its contacts with regulatory subunits, which are unique for each B-family (Cho and Xu, 2007; Wlodarchak et al., 2013; Xu et al., 2008; Xu et al., 2006). For example, the contacts between the catalytic subunit and B family substrates are mediated by its highly-conserved C-terminal tail (Xu et al., 2008). In fact, mutation and deletion of a small region in the C-terminal tail (residues 304–309) was found to abrogate binding between B55 and the core enzyme (Ogris et al., 1997). Post-translational modifications to the C-terminal tail can also impact holoenzyme assembly. Methylation of L309 has been shown to increase the binding affinity of the PP2A core enzyme for certain regulatory subunits (such as B55α in cells) (Longin et al., 2007; Yu et al., 2001), but not every subunit family (Ikehara et al., 2007; Ogris et al., 1997; Tolstykh et al., 2000; Wei et al., 2001; Xing et al., 2006; Xu et al., 2008; Xu et al., 2006). It has been also shown that modifications to the C-terminus can impact the catalytic activity of the enzyme, as phosphorylation of T304 and Y307 have been shown to inhibit phosphatase activity (Chen et al., 1992; Longin et al., 2007).

The B regulatory subunit of PP2A is comprised of 4 major families (B, B’, B”, and B”’), with each of these families comprising two to five genes and many containing multiple splice variants (Fig. 1A). This subunit is responsible for subcellular localization and is thought to be the key determinant of substrate specificity for the PP2A complex. While sequence conservation within each regulatory subunit family is high, there is very little sequence similarity across families. In addition, the expression levels of different regulatory subunits vary greatly depending on cell type and tissue (reviewed in Kurimchak and Graña, 2012; Kurimchak and Graña, 2015; Virshup and Shenolikar, 2009).

The B family of regulatory subunits (also referred to as B55, PR55, or PPP2R2) is comprised of 4 major isoforms: B55α, B55β, B55γ, and B55δ. B55α and B55δ are ubiquitously expressed, while B55β and B55γ have been found to be enriched in the brain. Sequence identity among these four B family members is high, with the majority of the structural differences occurring in the N-terminus (reviewed in Eichhorn et al., 2009; Kurimchak and Graña, 2015). The crystal structure of B55α complexed with the PP2A/AC core dimer has been solved (Xu et al., 2008). B55α and the other B-family subunits are seven-bladed β-propellers, with each of the blades being comprised of WD40 repeats. In complex with the core dimer, B-family subunits make several contacts with HEAT repeats 1–7 in the N-terminus of the scaffold but very few contacts with the catalytic subunit (Xing et al., 2006). Further structural analysis of B55α revealed an acidic top adjacent to a deep groove that is thought to mediate binding to its various substrates (Fig. 2). In support of this notion, mutation of residues within this region have been shown in vitro to inhibit dephosphorylation of Tau, a known B55α substrate (Xu et al., 2008). Binding of B55α to one of its other substrates, p107, has also been shown to be abrogated by a single aspartic acid to lysine mutation in the top surface. It is conceivable that this acidic top may use substrate-specific sequences to discriminate between its various binding partners, as Tau and p107 have been shown to share very little overlap between residues necessary for binding (Jayadeva et al., 2010).

Figure 2. Crystal structures of B55α, B56γ, and PR70-containing PP2A holoenzymes.

Pymol surface structures of trimeric PP2A holoenzymes containing B55α, B56γ, and PR70 are depicted in this figure. The scaffold A subunit is colored light green and the catalytic C subunit is colored light orange with microcystin (bright green) in the active site. Each of the B regulatory subunits are depicted with their electrostatic potentials indicated (red indicates acidic, while blue indicates basic). The surfaces of B55α and B56γ facing the active site are more acidic as compared to PR70, supporting the notion that the pocket created by each B subunit immediately adjacent to the active site of PP2A/C is unique in shape and electrostatic nature. The top surface of B55α also features a deep groove surrounding the central hole of the β-propeller in close proximity to the active site. B56γ makes many contacts with the catalytic C subunit as compared to B55α and PR70. The surface of PR70 appears less proximal to the active site of the enzyme than that of B55α and B56γ.

The B’ family of regulatory subunits (also referred to as B56, PR61, or PPP2R5) is comprised of 5 different members: B56α, B56β, B56γ, B56δ, and B56ε. B56γ and B56δ have 3 alternative splicing isoforms, and B56ε has an alternative translation isoform, which account for the 10 total B’ products identified (reviewed in Yang and Phiel, 2010). The crystal structure of B56γ has revealed that B56 regulatory subunits are composed of eight HEAT-like repeats in a structure that is reminiscent of the scaffold subunit. In complex with the core dimer, the B’ regulatory subunit makes many contacts with the scaffold subunit (specifically via HEAT repeats 2–8) and the catalytic subunit (Cho and Xu, 2007; Xu et al., 2006)(Fig. 1A). Structural analysis has also revealed an acidic surface of B’ that is oriented toward the active site of the catalytic subunit (Fig. 2). Based on structural similarities to proteins such as karyopherin and β-catenin (Conti et al., 1998; Graham et al., 2000), it has been proposed that this surface might be responsible for substrate recognition (reviewed in Shi, 2009b). In support of this notion, mutation of a highly conserved glutamic acid in B56β blocked dephosphorylation of one of its known substrates, tyrosine hydroxylase (Saraf et al., 2010).

The B” family of regulatory subunits (also referred to as PR72 or PPP2R3) is comprised of 3 members: PR70/PR48, PR130/PR72, and G5PR. PR70 has been identified as the full-length version of PR48, while PR72/PR130 refer to two subunits differentially transcribed from the same gene (Hendrix et al., 1993; Yan et al., 2000). Unlike the other three families of regulatory subunits, the B” family possess Ca²⁺-binding EF-hand motifs as they require the presence of calcium for binding to the PP2A core enzyme (Janssens et al., 2003). Crystal structures of PR70 and PR72-containing PP2A complexes have been solved (Wlodarchak et al., 2013). Like the B-family of regulatory subunits, B” subunits make multiple contacts with HEAT repeats 1–7 of the scaffold subunit (Fig. 1A). PR70 binding to the scaffold subunit causes it to maximally stretch and makes PR70-containing holoenzymes the most elongated relative to the other B subunit families (Fig 2). The C-terminus of B” subunits mediates binding to the catalytic subunit and is also thought to act as a substrate binding site, as known interactors including CDC6 have been shown to bind here (Davis et al., 2008; Wlodarchak et al., 2013).

The B”’ family of regulatory subunits is comprised of 3 members: Striatin, Striatin-3 (also known as SG2NA), and Striatin-4 (also known as Zinedin). They are localized primarily in dendritic spines of the corpus striatum, which is an essential component of the motor system (Castets et al., 1996; Moqrich et al., 1998; Salin et al., 1998). In addition to interacting with PP2A/A and PP2A/C subunits and carrying out dephosphorylation-related cellular functions, multiple other proteins have been shown to interact with B”’ subunits, including kinases, to form large signaling complexes referred to as Striatin-interacting phosphatase and kinase complexes (or STRIPAKs) (Goudreault et al., 2009). These STRIPAK complexes are involved in several key cellular processes including cell cycle control and cell signaling (reviewed in Hwang and Pallas, 2014). The three Striatin family members share four conserved domains: a coiled coil domain, a caveolin-binding domain, and a calmodulin-binding domain within the N-terminus; and a WD40 repeat domain within the C-terminus (Chen et al., 2014; Moreno et al., 2000). The crystal structure of the coiled coil domain of SG2NA has been solved, which assumes a noncanonical asymmetrical parallel homodimer conformation in which one α-helical chain contains a large bend. The coiled coil domain has been shown to directly interact with PP2A/A subunits to form a 2:2 heterotetrameric core complex. Mutational studies have revealed that SG2NA homodimerization is required for its interaction with PP2A and thus the assembly of STRIPAK complexes (Chen et al., 2014).

Due to the nature of its trimeric holoenzyme configuration, understanding the assembly and disassembly of PP2A complexes is pivotal in order to elucidate its regulation in cells (Fig. 1). Recent structural and biochemical efforts have narrowed down a linear pathway for the biogenesis of PP2A holoenzymes. It has been shown that partially folded PP2A/C is stabilized for latency by the binding of α4, which in part prevents its ubiquitination and subsequent degradation by Midline 1 (MID1) (Jiang et al., 2013; Liu et al., 2001; Short et al., 2002). This provides a source of latent PP2A/C in cells for the biogenesis of heterotrimeric holoenzymes while mitigating the unregulated phosphatase activity of free PP2A/C (Jiang et al., 2013). For the assembly of heterotrimeric complexes to continue, it is necessary for PP2A/C to become activated. This is facilitated by the phosphotyrosyl phosphatase activator (PTPA), which is a specific for PP2A. PTPA plays a critical role in stabilizing an active conformation of the catalytic active site and mediates the loading of catalytic metal ions (Guo et al., 2014). Guo et al. provide evidence that the active site is occupied by a Zn²⁺ ion and ATP is necessary for the loading of a catalytic Mg²⁺ ion for PP2A/C activation. This active form of the catalytic subunit can now bind to the scaffold subunit to form the PP2A core dimer (Fig. 1).

The unstructured C-terminus of PP2A/C is subject to post-translational modifications during the process of trimeric PP2A biogenesis, including phosphorylation of T304 and Y307 and carboxymethylation on L309 (Janssens et al., 2008; Lee and Pallas, 2007; Löw et al., 2014). Carboxymethylation on L309 is reversibly controlled by two PP2A-specific enzymes, leucine carboxy methyltransferase (LCMT-1) and PP2A-specific methylesterase 1 (PME-1) (Stanevich et al., 2011; Xing et al., 2008). Loss of LCMT-1 causes apoptosis in cells, indicating the requirement of PP2A methylation for cell function (Lee and Pallas, 2007). Methylation of PP2A/C is thought to play a role in the recruitment of B regulatory subunit to the core enzyme (Bryant et al., 1999; Longin et al., 2007; Tolstykh et al., 2000; Yu et al., 2001). It is known that methylation of L309 is essential for binding of B family subunits to the core dimer in cells, but not members of the other three regulatory subunit families (Longin et al., 2007; Yu et al., 2001). This is not the case for B subunits in vitro, which could suggest that L309 methylation simply enhances binding affinity between B subunits and the A/C dimer or that it serves as a recruitment signal for other necessary cofactors, among other possibilities (Ikehara et al., 2007; Shi, 2009b). It has been also shown that levels of PP2A/C methylation fluctuate during the cell cycle, indicating a role of PP2A/C methylation and holoenzyme assembly in cell cycle regulation (Janssens et al., 2008; Yu et al., 2001). Active, methylated A/C dimers can complex with a B regulatory subunit to form the complete heterotrimeric holoenzyme (reviewed in Wlodarchak and Xing, 2016).

2. The B subunits determines substrate specificity

Heterotrimeric PP2A holoenzymes are known to coordinate complex spatially and temporally regulated functions within the cell, which requires a high degree of substrate specificity. This has been studied in the context of many binding partners, most prominently the microtubule-binding protein Tau (Drewes et al., 1993; Gong et al., 1994; Xu et al., 2008), PRC1 (Cundell et al., 2016), and p107 (Jayadeva et al., 2010; Kolupaeva et al., 2013; Kurimchak et al., 2013) for PP2A/B55α and SLIM-containing substrates (Hertz et al., 2016) and tyrosine hydroxylase (Saraf et al., 2010) for PP2A/B56, all targeted to the catalytic subunit of PP2A for dephosphorylation by the B regulatory subunits. Crystallization studies of the B, B’, and B” regulatory subunits in the context of their trimeric holoenzymes have revealed divergent structures, which is consistent with their unrelated underlying sequence. In PP2A complexes, the B regulatory subunit acts as the key determinant of substrate specificity and subcellular localization. This is in contrast to PP1 and PP2B protein phosphatases, in which the regulatory and catalytic subunits both participate in substrate determination (reviewed in Shi, 2009a). For these phosphatases, a binding pocket distal from the active site on the catalytic subunit binds specific SLIMs with the consensus sequences RVxF and PxIxIT, respectively (Bollen et al., 2010; Roy and Cyert, 2009). Recently, there have been significant efforts to decipher the mechanisms of PP2A substrate recognition in order to further understand many phosphorylation-mediated eukaryotic signaling processes. The binding properties of B56-containing PP2A complexes are among the most well-characterized.

As indicated above, the B56 family of regulatory subunits is comprised of five human isoforms (α, β, γ, δ, and ε), making it the largest of the four families (reviewed in Shi, 2009b). In PP2A/B56 complexes, the surface of B56 makes extensive interactions with the scaffold subunit and orients toward the active site of the catalytic subunit (Cho and Xu, 2007; Xu et al., 2006)(Fig. 2). Structural similarities to other well-characterized proteins such as karyopherin and β-catenin led to the hypothesis that the surface of B56 subunits acts as a sensor to recognize target substrates for dephosphorylation (Conti et al., 1998; Graham et al., 2000). Deletion mutagenesis studies for two identified substrates, the kinesin KIF4A and the RHOA exchange factor GEF-H1, in comparison to known binding requirements of other substrate proteins, revealed a putative, B56-binding motif termed the LxxIxE motif (Hertz et al., 2016). These motifs are found in intrinsically disordered domains and are highly conserved in essential proteins, classifying them as SLIMs (reviewed in Davey et al., 2015). Further analysis showed that the amino acid composition of the motif modulates the affinity for B56, with some residues conferring stronger binding affinities than others (e.g., L compared to F in position 1) and others being an absolute requirement for binding (e.g., E in position 6). This ultimately determines the phosphorylation status of associated substrates and has critical consequences on their in vivo functions. This was illustrated using the transcription factor FOXO3, in which it was shown that there is a correlation between B56 binding affinity and extent of FOXO3 dephosphorylation, which determines its cellular localization and affects downstream target gene transcription (Hertz et al., 2016). For the B56 family of regulatory subunits, the discovery of the LxxIxE motif in many of its binding partners provided a simple but elegant mechanism of substrate recognition and linked binding affinity with phosphorylation status. Previous studies have identified two consecutive Arg37 and Arg38 residues in Tyrosine Hydroxylase that may interact with B56′β Glu153 and are important for dephosphorylation of Ser31 and Ser40 (Saraf et al., 2010). Ectopic expression of WT B56′β, but not a B56′β Glu153 mutant, results in dephosphorylation of multiple proteins as determined by 2D electrophoresis followed by phosphoprotein staining with Pro Q Diamond, indicating the importance of this acidic residue in substrate dephosphorylation. Of note, while Tyrosine Hydroxylase was not identified in the search for B56 substrates containing SLIMS, analysis of its sequence reveals a potential LxxIxE consensus motif, specifically ^aa431LHCLSE^aa436, which could mediate binding to B56’β conceivably facilitating the interaction of Arg37 and Arg38 with B56′β Glu153 to direct dephosphorylation of Tyrosine Hydroxylase Ser31 and Ser40.

B55α is a ubiquitously expressed regulatory subunit of the B/B55 family and has been reported to target many diverse protein substrates with critical functions in cell division, differentiation, survival, as well as tissue-specific specialized processes (reviewed in Eichhorn et al., 2009; Kurimchak and Graña, 2015). Many of these proteins have been shown to interact with PP2A either by coimmunoprecipitation and/or mass-spectrometry approaches (reviewed in Eichhorn et al., 2009). In spite of an abundance of information on the structure and biochemical properties of PP2A holoenzymes (especially in the context of PP2A/B55α-mediated Tau recognition), the molecular basis for how B55α contacts its substrates and how specific phosphorylated sites are positioned for dephosphorylation by the catalytic subunit of PP2A remain elusive (Shi, 2009a; Xu et al., 2006; Xu et al., 2009). It is also unknown if B55α also binds any SLIMs in substrates as is the case for B56 regulatory subunits. Among the most well-studied substrates of B55α are Tau, PRC1, and p107.

Tau is one of three microtubule-associate proteins (MAPs) in the normal mature neuron (Goedert et al., 1989). In normal cell physiology, Tau is responsible for modulating the assembly and stabilization of the microtubules that support intraneuronal transport (Weingarten et al., 1975). In the pathology of Alzheimer disease (AD) and other related neurodegenerative diseases called tauopathies, Tau protein is abnormally hyperphosphorylated, which results in its aggregation into filament bundles (Grundke-Iqbal et al., 1986). This observation led to the discovery of PP2A as a key mediator of Tau dephosphorylation in the brain and implicated its inactivity in tauopathy disease progression (Drewes et al., 1993; Gong et al., 1994). Using highly purified and recombinant proteins in in vitro dephosphorylation assays as well as structure-guided mutagenesis, the molecular basis for Tau binding to PP2A/B55α has become more clearly understood. Specifically, two “Bα-binding repeats” were identified that had an enrichment of positively charged amino acids, with one of these domains containing 11 lysine residues within a 63 amino acid stretch. Importantly, this observation of basic residues being required for PP2A/B55α-mediated dephosphorylation of Tau agrees well with what was found about the acidic nature of the putative substrate-binding top surface of B55α (Xu et al., 2008)(Fig. 2).

Various PP2A holoenzymes have been shown to interact with the three pocket proteins (p107, p130, and pRB) in the context of extracellular signals or stresses (Cicchillitti et al., 2003; Garriga et al., 2004; Voorhoeve et al., 1999a; Voorhoeve et al., 1999b; Vuocolo et al., 2003). Interestingly, pharmacological inhibition of CDKs, which are responsible for pocket protein phosphorylation and inactivation, results in rapid dephosphorylation of the three pocket proteins. This observation led to the hypothesis that PP2A holoenzymes and CDKs exist in an equilibrium during the cell cycle that ultimately determines the phosphorylation status of the three pocket proteins (Garriga et al., 2004). Cell-based assays using pharmacologic inhibitors of PP2A/PP1 and co-expression of SV40 small t antigen, which is known to displace B subunits from the PP2A core enzyme, implicated PP2A. p107 targeting by PP2A/C was confirmed via immunoprecipitation assays. Finally, a candidate B regulatory subunit approach using purified trimeric PP2A holoenzymes and GST-p107, and an unbiased proteomic approach using Flag-p107 pulldowns from RCS cell lysates, demonstrated that the PP2A interaction with p107 is largely mediated by B55α and to a much lesser extent by B55δ. Reciprocal proteomic analyses in RCS, identified both p107 and pRB in B55α immunoprecipitates (Jayadeva et al., 2010; Kurimchak et al., 2013). These findings were consistent with another catalog approach that independently identified B55α as the B subunit targeting p107 upon FGF1 stimulation in RCS cells (Kolupaeva et al., 2013). The preference of PP2A/B55α for p107 over pRB and p130 could be caused by several factors, including differential innate binding affinities between the pocket proteins and/or subcellular and temporal expression constraints. Further exploration of the precise targeting mechanism and binding properties of p107 and PP2A/B55α are ongoing.

PP2A/B55 complexes play a major role in the temporal coordination of cell division, particularly during mitotic exit (Glover, 2012; Grallert et al., 2015; Mochida and Hunt, 2012; Wurzenberger and Gerlich, 2011). It has been previously shown that dephosphorylation of the anaphase spindle protein PRC1 is controlled by PP2A/B55 holoenzymes (Cundell et al., 2013). During mitotic exit, PP2A/B55 has been also implicated in the reassembly of the Golgi apparatus and nuclear envelope via its interactions with substrates including the Golgi tethering protein GM130 (Lowe et al., 2000; Schmitz et al., 2009). Negative regulation of PP2A/B55 during mitosis is controlled by two key protein substrates, α-Endosulfine (ENSA) and ARPP19 (Cundell et al., 2013). To understand the mechanisms through which its various mitotic substrates are both recognized and differentially dephosphorylated, a combination of phosphoproteomics and kinetic modeling strategies were employed in whole cell extracts. This strategy led to the discovery that the identified B55 substrates have a defined bipartite polybasic motif in which the numbers of basic residues present affects the observed dephosphorylation rate. Thus, the rate of PP2A/B55-mediated dephosphorylation has been proposed to be encoded into substrates via basic residues within this motif. During mitosis, this means that as PP2A/B55 moves through the transition from metaphase to anaphase, it acts on progressively fewer basic substrates by varying its rate of dephosphorylation. It is thought that the different degrees of basicity within the target substrate protein can control the kinetics of dephosphorylation by increasing its residence time on the PP2A/B55 surface (Cundell et al., 2016).

It is known that for effective mitotic exit, PP2A/B55 complexes must dephosphorylate CDK1–Cyclin-B1-phosphorylated serine and threonine sites (SP and TP sites, respectively) in target substrates such as CDC20 (Agostinis et al., 1992; Manchado et al., 2010; Mayer-Jaekel et al., 1993; Mochida et al., 2009; Schmitz et al., 2009). While phosphorylated SP and TP sites are often thought of as interchangeable post-translational modifications, both kinases and phosphatases have been demonstrated to exhibit an intrinsic preference for one over the other. Literature published over three decades ago showed that the turnover rate of PP2A for short peptides containing TP sites is at least 34-fold higher than for otherwise identical SP substrates (Cohen, 1989; Deana et al., 1982; Deana and Pinna, 1988). In order to assess whether this was the case for B55-containing PP2A complexes, biochemical approaches and more recently proteomic analyses were employed to assess PP2A/B55-specific dephosphorylation during mitotic exit. It was found that the preference of PP2A/B55 complexes for phosphothreonine is critically important for the timely activation of the APC/C and the translocation of the chromosomal passenger complex (CPC), which is essential for proper cytokinesis (Hein et al., 2017; van der Horst and Lens, 2014). This result was underscored by the fact that substitutions of TP sites for SP sites (and vice versa) in the APC/C co-activators CDC20 and CDH1 resulted in defects in the temporal regulation of mitosis (Hein et al., 2017). Others performing similar kinetic studies looking at mitotic exit found that there was an absolute requirement for threonine residues for PP2A/B55-mediated dephosphorylation of substrates such as PRC1 and TPX2 (Cundell et al., 2016). It is thought that the preference of PP2A/B55 for phosphothreonine is an inherent property of the PP2A catalytic subunit, as it has also been demonstrated for B56-containing PP2A complexes (Deana et al., 1982; Deana and Pinna, 1988; Pinna et al., 1976). To date, the molecular details that govern this preference remain unclear.

The B” family of PP2A regulatory subunits is made up of 3 known members: PR70/PR48, PR130/PR72, and G5PR. As mentioned above, Ca²⁺-binding EF-hand motifs have been identified in B” family members, which is consistent with their requirement of calcium for binding to the PP2A core enzyme (Janssens et al., 2003). PR70 function is critically important in the cell cycle, especially for progression into S phase, as knockdown in cells causes G1 arrest (Davis et al., 2008). While B and B’ regulatory subunits are thought to use their acidic surfaces facing the catalytic active site of PP2A/C for substrate binding, PR70 subunits have been shown to bind a number of known substrates (including CDC6 and pRB) via their C-terminal regions (Davis et al., 2008; Magenta et al., 2008). PR48 was first shown to bind human CDC6 when it was identified in a yeast two-hybrid screen of a human placental cDNA library using CDC6 as bait (Yan et al., 2000). A recent structural study further characterized the PR70-CDC6 interaction, showing that the dynamic interaction between PR70 and the PP2A catalytic subunit exerts a tight control on CDC6 dephosphorylation by altering its effective entry into the enzymatic active site (Wlodarchak et al., 2013). PR70/PR48 has also been shown to interact with pRB and p130 in in vitro pulldown assays and with pRB in cells (Jayadeva et al., 2010; Magenta et al., 2008). Consistent with its calcium requirement for binding to the PP2A core enzyme, treatment with an intracellular calcium chelator blocked the PP2A/PR70-mediated dephosphorylation of pRB in cells (Magenta et al., 2008). These and other studies give further insight into the mechanisms through which PP2A phosphatase activity is tightly regulated in processes including the cell cycle, and not just a default suppressor of phosphorylation when kinases are inactivated.

The Striatin family of scaffolding proteins (Striatin, SG2NA, and Zinedin) comprises the fourth family of PP2A regulatory subunits (B”’ family). As mentioned above, Striatins not only mediate dephosphorylation-related functions via interactions with the PP2A/AC core dimer, but also serve a unique function in PP2A complexes by acting as molecular scaffolds that organize signaling complexes known as STRIPAKs (Goudreault et al., 2009). Recent studies have implicated STRIPAK complexes as key regulators of signaling pathways including Hippo, MAPK, and non-genomic nuclear receptor signaling (reviewed in Hwang and Pallas, 2014). In the Hippo signaling pathway, Striatin acts as a scaffold for the assembly of a signaling complex consisting of the estrogen receptor (ER), G_αi and eNOS. The assembly of this complex is critical for effective downstream signaling, as disruption of the STRN-ER interaction has been shown to block estradiol-induced activation of MAPK and phosphorylation of eNOS (Bernelot Moens et al., 2012). In the MAPK pathway, a component of the STRIPAK complex known as GCK (or germinal center kinase) can directly activate MAP3Ks and thus acts as a positive regulator of MAPK signaling (Dan et al., 2001). Recent studies have linked loss of function or dysregulation of STRIPAK components to human disease, particularly cancer. For example, elevated expression levels of the STRIPAK kinase MST4 in various prostate cancer cell lines and tumors has been shown to promote cancer cell growth in vitro and in vivo (Lin et al., 2014; Sung et al., 2003; Xiong et al., 2015). Though recent progress has been made in the understanding of STRIPAK complexes, many questions remain, ranging from its specific structural architecture to its mechanisms of substrate regulation.

3. PP2As role in proliferation, differentiation and development

3.1. Cell cycle

a. G2/M, Mitosis and G1 reentry.

PP2A plays an important role throughout the cell cycle. During G2, PP2A activity is high and Cyclin B/CDK1 activity is low. Inhibitory phosphorylation of CDK1 is maintained by WEE1/MYT1 kinases and can be removed by the CDC25 phosphatase at the G2/M transition. In Xenopus egg extracts, it has been demonstrated that PP2A/B55δ is the major phosphatase responsible for dephosphorylation of CDK substrates (Mochida et al., 2009). In fact, PP2A/B55 dephosphorylates and keeps WEE1/MYT1 active and CDC25 inactive in G2, preventing Cyclin B/CDK1 from activation (reviewed in Lorca and Castro, 2013). Since activation of Cyclin B/CDK1 is essential for mitotic entry, PP2A/B55 inhibition is required at the G2/M transition, which is mediated by the Greatwall kinase. Greatwall knockdown has been shown to promote G2 arrest and mitotic defects, which can be resumed by PP2A/C knockdown or pharmacological PP2A inhibition with okadaic acid (Burgess et al., 2010; Castilho et al., 2009). Only with low PP2A/B55 activity can a Cyclin B/CDK1 activation threshold be reached to enter mitosis (Burgess et al., 2010; Lucena et al., 2017; Tuck et al., 2013).

During mitosis, PP2A activity is regulated to prevent mitotic collapse and control mitotic exit. In prophase, each of the duplicated chromosomes condenses into two linked sister chromatids, forming X-shaped structures. While sister chromatids are linked at the centromere, a point of chromosome constriction with highly condensed chromatin, sister chromatids are also held together along the chromatids arms by Cohesins, which are ring protein complexes with multiple subunits (Brooker and Berkowitz, 2014). At this point, PP2A/B55 activity is inhibited by the Greatwall kinase substrates ENSA and ARPP19. This step is critical to prevent mitotic CDK substrates from being prematurely dephosphorylated. Greatwall depletion in HeLa cells has been shown to cause chromosome abnormalities such as incomplete condensation or improper congression to the metaphase plate, which results in mitotic defects including metaphase arrest followed by cell death, metaphase delay followed by aberrant chromosome segregation and cytokinesis, or exiting mitosis absent of metaphase (Burgess et al., 2010). Noticeably, the condensation defects caused by loss of Greatwall can be reversed by co-depletion of B55 isoforms in MEFs, implicating PP2A/B55 as the mediator of mitotic collapse upon Greatwall loss (Álvarez-Fernández et al., 2013).

Of note, though PP2A/B55 activity is inhibited in early mitosis, it has been reported that PP2A/B56 may help prevent spindle assembly checkpoint failure (Jin et al., 2017; Varadkar et al., 2017). In prometaphase, the nuclear envelope breaks down and the kinetochore, a multiprotein complex, starts to form around the centromere. At this point, the kinetochore microtubules, which are polarized long fibers made of α/β-tubulin, extend from two opposing centrosomes toward the kinetochores. Since this step is critical and error-prone, the kinetochore ensures proper attachment via two processes: spindle assembly checkpoint (SAC) and kinetochore-microtubule error-correction, which are both controlled by PP2A/B56 (reviewed in Saurin, 2018). Briefly, SAC senses the unattached kinetochores and error-correction senses and destabilizes the incorrect kinetochore-microtubule attachment. The cell is then arrested until each kinetochore pair is properly attached to microtubules extended from opposing centrosomes to meet the bi-orientation criteria (reviewed in Saurin, 2018). In the simplified error-correction model, incorrect kinetochore-microtubule attachment generates low level of tension, which triggers the phosphorylation of numerous complexes at the outer kinetochore principally by Aurora B, leading to microtubule detachment. The core of the outer kinetochore known as the KMN complex is composed of three sub-complexes: the KNL1 complex, the MIS12 complex and the NDC80 complex (reviewed in Musacchio and Desai, 2017). For a stable kinetochore-microtubule attachment, PP2A/B56 is required to dephosphorylate Aurora B substrates (Foley et al., 2011). PP2A/B56 is recruited to the kinetochores by binding to BubR1 for SAC silencing. Meanwhile, in response to unattached kinetochores, SAC forms the mitotic checkpoint complex (MCC) which directly binds and inhibits the anaphase promoting complex/cyclosome (APC/C), a ubiquitin E3 ligase. Once SAC is satisfied, Cyclin B is degraded following APC^CDC20 activation, CDK1 activity is rapidly reduced, and Cyclin B/CDK1-inhibited PP1 becomes auto-reactivated. In Xenopus egg extracts, it has been demonstrated that PP1 activates PP2A/B55 by inhibiting Greatwall kinase (Ma et al., 2016). In fission yeast, Grallert et al. reported a phosphatase relay mechanism in the regulation of mitotic exit. The inhibitory phosphorylation of B56 at Ser378 is maintained by PLK1. When PLK1 activity decreases after anaphase onset due to PLK1 degradation by APC/C (Schmucker and Sumara, 2014), PP1 is recruited to dephosphorylate B56 (Grallert et al., 2015). As a result, dephosphorylation of specific substrates drives orderly events including chromosome decondensation and nuclear envelope reformation, leading to proper mitotic exit and cytokinesis (Cundell et al., 2016; Mehsen et al., 2018).

b. G0/G1, G1/S.

The commitment of a eukaryote cell to a cell division cycle is made in late G1 when cells depend on extracellular factors and/or nutrients. In metazoans, except for specific embryonic cycles, progression through the restriction point in late G1 requires extracellular mitogenic signals (Pardee, 1989). In absence of mitogen stimulation prior to the restriction point, cells exit the cell cycle and become quiescent, in a stage known as G0. Once re-stimulated by growth factors, cells reenter the cell cycle in G1 and progress through the restriction point. The restriction point is regulated by the phosphorylation status of the three members of the pocket protein family (pRB, p107, and p130). In G0 and early G1, pocket proteins stay active in their hypophosphorylated forms and hold back cell cycle progression by forming a repressor complex with E2F and DP transcription factors at the promoters of E2F-dependent genes (reviewed in Kurimchak and Graña, 2012; Kurimchak and Graña, 2015; Sotillo and Graña, 2010). Within these complexes, pocket proteins recruit chromatin modifying enzymes that help compact chromatin and silence these promoters. In response to mitogenic stimulation, pocket proteins become hyperphosphorylated and inactive, resulting in abrogation of pocket protein/E2F/DP repressor complexes, chromatin decondensation and transcription of E2F-dependent genes, which is further stimulated by the recruitment of activating E2F complexes. This allows for passage through the restriction point with the expression of gene products and enzymes required for DNA and histone synthesis as well as the control of subsequent cell cycle events. This process is fine-tuned by pocket protein phosphorylation via G1 Cyclin/CDKs and dephosphorylation via PP2A. As previously mentioned, PP2A can dephosphorylate pocket proteins in all phases in cell cycle, with pRB as a target of PP2A/PR70 and p107, p130 and likely pRB as targets of PP2A/B55α (Jayadeva et al., 2010; Kurimchak et al., 2013; Magenta et al., 2008). The hyperphosphorylation status of pocket proteins is promoted and maintained by Cyclin D/CDK4/6 in G1, Cyclin E/CDK2 from G1 to S, Cyclin A/CDK2 in S/G2/mitosis and Cyclin B/CDK1 from G2/M until late mitosis, when CDK activity decreases and pocket proteins become abruptly dephosphorylated in preparation for the next cell cycle (reviewed in Kurimchak and Graña, 2015). Of note, PP2A may also regulate progression through interphase via PP2A/B56, which dephosphorylates cyclin E on Ser384 to prevent cyclin E degradation potentially mediating Cyclin E/CDK2 phosphorylation of substrates beyond the G1/S transition (Davis et al., 2017).

c. Mitogen stimulation and other signaling cues.

Two well-studied signaling pathways that promote cell proliferation and growth have been shown to be regulated by PP2A: mitogen-activated protein kinase (MAPK) and PI3K/AKT pathways. In the MAPK pathway, PP2A/B family members B55α and B55δ dephosphorylate RAF1 on Ser259 (Adams et al., 2005; Ory et al., 2003), a residue that when phosphorylated inhibits RAF1 kinase by binding to 14-3-3 proteins (Abraham et al., 2000; Jaumot and Hancock, 2001). The three subunits of the PP2A/B55α holoenzyme were detected in immunopurified RAF1 by mass spectrometry and B55α recruitment to the RAF1-PP2A/AC complex was stimulated by PDGF, which stimulates RAF1-Ser259 dephosphorylation (Ory et al., 2003). Loss and gain of B55α and B55δ function in serum-starved HEK cells re-stimulated with EGF confirmed the implication of PP2A/B55 trimeric holoenzymes in MEK1/2 and ERK1/2 activation via RAF1 activation (Adams et al., 2005). Follow up studies also support a role for PP2A/B55α and PIN1 in the recycling of ERK-inactivated RAF1 and B-RAF (Dougherty et al., 2005; Ritt et al., 2010). In a mechanistic fashion related to RAF1 activation, Kinase Suppressor of Ras 1 (KSR1), the molecular scaffold that assembles the MAPK pathway, is also a dephosphorylation target of PP2A, as mass spectrometry analysis identified the PP2A core dimer and B55α in immunopurified KSR1. In quiescent cells, KSR1 is phosphorylated on Ser297 and Ser392 by C-TAK and sequestered by 14-3-3 in an inactive form. Upon PDGF treatment, PP2A/B55 dephosphorylates KSR1 on Ser392, which releases it from 14-3-3 and allows it to translocate to the cell membrane to facilitate MAPK pathway activation (Ory et al., 2003). More recent studies have found that the guanine nucleotide exchange factor GEF-H1 serves as the bridge linking KSR1 to PP2A, as it is required for KSR1 dephosphorylation and oncogenic RAS signaling via KSR1 (Cullis et al., 2014). However, KSR1 was found to preferentially interact with B56α, β and ε, as compared to B55α, and this interaction was dependent on GEF-H1 and mitogenic stimulation as well as promoted by oncogenic RAS activation (Cullis et al., 2014). Thus, it is possible that in response to different signals or in different cell types, this site might be targeted by distinct PP2A holoenzymes.

In the PI3K/AKT pathway, both PI3K and AKT downstream signaling factors are substrates of PP2A (Kuo et al., 2008; Petritsch et al., 2000). Following stimulation and activation of the PI3K pathway, AKT is phosphorylated on Thr308 at the activation loop by PDK1 and on Ser473 at a hydrophobic motif by TORC2 (target of rapamycin complex 2), DNA-PK (DNA-activated protein kinase), PDK2 and potentially several other kinases, resulting in activation of several downstream pathways implicated in cell survival, protein synthesis and cell growth, metabolism and angiogenesis (reviewed in Fayard et al., 2005). AKT phosphorylation on Thr308 was decreased in serum-starved B55α-overexpressing cells following serum stimulation, while B55α knockdown resulted in increased phosphorylation of Thr308. Consistently, an in vitro dephosphorylation assay showed that PP2A/B55α dephosphorylates AKT on Thr308 preferentially over Ser473, suggesting that B55α selectively and directly regulates AKT phosphorylation on Thr308 (Kuo et al., 2008). More recent work has shown that B56β also targets AKT for dephosphorylation on both Thr308 and Ser473, facilitating the attenuation of insulin signaling (Rodgers et al., 2011). Interestingly, the CLK2 kinase is stimulated by insulin/AKT-mediated phosphorylation of AKT-bound B56β, which promotes the recruitment of the PP2A/AC dimer and subsequent dephosphorylation and attenuation of AKT. Moreover, the PH-domain-leucin-rich protein phosphatase, PHLPP, also dephosphorylates AKT Thr473, suppressing AKT-mediated survival (Gao et al., 2005). While the key phosphatase targeting each of these two residues remains controversial, it is possible that multiple phosphatases target these sites in response to varying attenuation signals. A better understanding of substrate specificity should help resolve these issues.

Downstream substrates of the AKT signaling pathway such as p70S6K and PHD2 have also been reported to be negatively regulated by PP2A. Activation of p70S6K is mediated by sequential phosphorylation on multiple sites by mTOR and PDK1, which is counteracted by PP2A/B56γ (Hahn et al., 2010; Jastrzebski et al., 2007). Although p70S6K Thr389 has also been reported to be dephosphorylated by PHLPP protein phosphatase (Liu et al., 2011), it is plausible as pointed above that the same substrate can be dephosphorylated by multiple phosphatases. PP2A also plays an important in role in autophagy. It has been reported that during amino acid starvation, PP2A is activated due to dissociation of α4 from PP2A/C and increased formation of PP2A/B55α, leading to dephosphorylation of the mTORC1 substrate ULK1 and activation, which promotes the induction of autophagy (Wong et al., 2015). Finally, PP2A/B55α also specifically dephosphorylates PHD2 during hypoxia-induced autophagy. In normoxia, PHD2 promotes hypoxia-inducible factor HIF1α degradation when PHD2 is phosphorylated on Ser125 by p70S6K. Under hypoxia, mTOR and p70S6K are inactivated and PP2A/B55α dephosphorylates PHD2 Ser125, leading to HIF1α stabilization (Di Conza et al., 2017).

3.2. Differentiation and development.

PP2A plays a critical role in regulating development and its dysfunction has been implicated in developmental diseases. It has been shown that Ppp2r1a knockout in mice is embryonic lethal (between E5 and E10.5), suggesting that the scaffold subunit PP2A/Aα is required for embryonic development (Ruediger et al., 2011). Loss of Ppp2r1a in mouse oocytes severely decreases the number of pups by 84% compared to control mice, suggesting an essential role of Ppp2r1a in female fertility (Hu et al., 2014). Similar to the scaffold subunit, constitutive homozygous knockout of Ppp2ca is embryonic lethal (E6.5), whereas heterozygous embryos developed normally (Gu et al., 2012; Götz et al., 1998). Knockout of Ppp2ca in male mouse germ cells leads to infertility (Pan et al., 2015). Although Ppp2ca and Ppp2cb share 97% sequence identity, knockout of Ppp2cb is viable (Gu et al., 2012).The role of B regulatory subunits in development has also been widely studied. PP2A/B55α is important for oocyte maturation and embryonic development. Knockdown of PP2A/B55α in oocytes leads to defective spindle assembly, chromosome misalignment and DNA damage response in MII stage. Moreover, knockdown of PP2A/B55α in mouse embryos shows reduced number of cells developing to 8-cell and blastocyst stages, accomppanied by increased apoptosis (Liang et al., 2017). PP2A/B55γ is critical in in the differentiation of bone marrow mesenchymal stem cells (BM-MSCs) to osteoblasts. Osteoblast differentiation is composed of three processes: proliferation, matrix maturation and matrix mineralization (Rutkovskiy et al., 2016). Knockdown of B55γ in BM-MSCs reduced mineralization with insufficient calcium depoisits and a lack of the characteristic cuoidal cell morphology (Serguienko et al., 2017). It has been reported that mutations of PP2A/B56 family genes correlate with childhood overgrowth and intellectual disability in 5 individuals (Loveday et al., 2015). In addition, B56γ is also critical for heart development. Although not embryonic lethal, B56γ^−/− leads to increased neonatal death in mice. Starting from E16, B56γ^−/− mice exhibit ventricular septal defects in the heart due to increased apoptosis (Varadkar et al., 2014).

PP2A appears to be involved in regulating T cell signaling and the immune response. Full activation of CD4+ and CD8+ T cells requires both T cell receptor (TCR) stimulation and co-stimulation provided by receptors such as CD28, while the inhibitory co-signal CTLA-4 prevents T-cell activation (Gaud et al., 2018; Rudd et al., 2009). NFκB becomes activated upon TCR stimulation, and suppression of TCR-mediated NF-κB activation appears dependent on PP2A/B56γ (Breuer et al., 2014). Knockdown of B56γ in Jurkat T cells stimulated via TCR showed increased phosphorylation of IKK and IκBα, which are required for NF-κB activation. Consistently, the transcription and secretion of the NF-κB target IL-2 are increased upon B56γ knockdown (Breuer et al., 2014). PP2A has also been reported to associate with CD28 and CTLA-4 (Chuang et al., 2000). GST pulldowns from HEK293 cells have shown that the cytoplasmic domains of CTLA-4 and CD28 can interact with PP2A/C. Although how PP2A would affect CTLA-4 is still unclear, inhibition of PP2A by okadaic acid treatment or catalytic inactive mutation H59K in PP2A/C leads to increased IL-2 secretion in anti-CD28 stimulated J32 Jurkat cells (Chuang et al., 2000).

PP2A/B55α helps mediate chondrocyte maturation in response to fibroblast growth factor (FGF1) (Kolupaeva et al., 2013; Kurimchak et al., 2013). Upon FGF1 treatment, p107 is rapidly dephosphorylated in rat chondrosarcoma (RCS) cells, followed by dephosphorylation of p130 and pRB, leading to chondrocyte maturation and cell cycle exit. PP2A/B55α complex formation with p107 and p107 dephosphorylation increases following FGF1 stimulation. Furthermore, PP2A/B55α and simultaneous PP2A/B55α/δ knockdown delay p107 dephosphorylation, confirming that this process is mediated by PP2A/B55α/δ. As a result of p107 dephosphorylation 1.5 hours after FGF treatment, p107 is increasingly recruited to the promoters of E2F dependent genes such as MYC and represses gene expression. Also, upregulation of p21 expression is detected about 10 to 15 hours after FGF stimulation coinciding with p130 and pRB dephosphorylation, implicating that this is a consequence of p21 inhibiting CDK2 and CDK4 rather than a direct regulation by PP2A/B55α. Altogether, FGF-induced dephosphorylation of pocket proteins contributes to chondrocyte maturation and cell cycle exit (Kolupaeva et al., 2013; Kurimchak and Graña, 2015; Kurimchak et al., 2013).

4. PP2A deregulation in cancer

Early support for a role of PP2A as a tumor suppressor came from studies demonstrating that okadaic acid and other potent toxins that inhibit PP2A caused tumors in mice and the observation that the small t antigen of SV40, which disrupts PP2A holoenzymes, was oncogenic (reviewed in Hahn and Weinberg, 2002; Kurimchak and Graña, 2013). Below we outline the diversity of alterations targeting core subunits and PP2A inhibitors that are associated mostly with its tumor suppressor function, but also with a context-dependent oncogenic role.

a. PPP2R1A mutations.

As described above, the scaffold subunit of PP2A has two isoforms, PP2A/Aα and PP2A/Aβ, which are encoded by two 86% identical genes PPP2R1A and PPP2R1B respectively (Zhou et al., 2003). It has been reported that PPP2R1A is frequently mutated in various cancer types, ranging from 18.4% to 43.2% of all cases. Most cancer-associated mutations encode inactive PP2A/Aα or PP2A/Aβ scaffolds, which prevent proper PP2A holoenzyme formation and promote carcinogenesis (Haesen et al., 2016). Anti-HA-tag pulldown assays from cells transfected with HA-PP2A/Aα constructs based on 11 mutations observed in endometrial carcinomas, identified 10 mutations with reduced binding to the PP2A/C subunit. In addition, most PP2A/Aα mutants showed defects in binding to B55α/β subunits, while certain PP2A/Aα mutants also lost the ability to bind to other B subunits such as B56β/γ/δ and PR72. Of note, overexpressing patient-derived PPP2R1A mutations R183G/Q and S256F in HEC-1-A uterine cancer cells, which are wild type for PPP2R1A, promotes both anchorage-independent and tumor xenograft growth (Haesen et al., 2016). Remarkably, PP2A/Aα subunit E64D and E64G mutations found respectively in lung and breast cancer are defective in binding to B56 subunits but remain unaffected in binding to B55 and PR72 (Zhou et al., 2003). Knock-in mouse models harboring E64D or E64G are prone to developing lung cancer (Ruediger et al., 2011; Walter and Ruediger, 2012). In addition, PPP2R1A mutations are associated with poor prognosis of gastrointestinal stromal tumors and dysfunctional PP2A results in enhanced phosphorylation of downstream oncogenic kinases such as AKT and ERK (Akaike et al., 2018; Toda-Ishii et al., 2016).

b. B55 alterations.

Deletions affecting the PPP2R2A gene have been observed in a number of epithelial tumors in prostate, ovarian and breast cancer patients (Cheng et al., 2011; Watt et al., 2017; Youn and Simon, 2013). Of note, shRNA knockdown of PP2A/B55α, B56α and B56γ in breast cancer cells causes a hyper-proliferative-phenotype in 3D culture (Watt et al., 2017). A recent study has identified two rare loss of function mutations in the PPP2R2A gene in three de-novo AML patients (without MDS or not preceded by chemotherapy and/or radiation treatment). Mutations G247T and delG85 result in premature stop codons and are associated with apparent complete loss of B55α expression, despite the presence of a wildtype PPP2R2A allele by an unknown mechanism. Interestingly, absence of B55α expression is associated with increased phosphorylation of AKT on Thr-308 (Shouse et al., 2016), a site previously reported to be regulated by PP2A/B55α (Kuo et al., 2008). Of note, these AML mutant cells are also more sensitive to AKT inhibitor treatment (Shouse et al., 2016).

While there is growing evidence supporting a tumor suppressor role for PPP2R2A, B55α overexpression has been observed in pancreatic ductal adenocarcinoma tissues and correlated with poor patient prognosis (Hein et al., 2016). Consistent with this finding, B55α expression is also elevated in pancreatic cancer cell lines as compared to immortalized normal human pancreatic ductal cells. Moreover, shRNA-mediated depletion of B55α in CD-18/HPAF cells resulted in decreased mitogenic signaling and reduced anchorage-independent growth and tumorigenicity in mice with tumor cells orthotopically implanted into the pancreas (Hein et al., 2016). Given the large number of PP2A/B55α substrates with diverse functions, it is conceivable that this key phosphatase bears context-dependent tumor suppressive or oncogenic functions. In cancers where B55α/PP2A or other PP2A holoenzymes exhibit oncogenic function, selective inhibitors of PP2A activity may have therapeutic potential. With this regard, it has been reported that PP2A inhibition sensitizes cancer cells to kinase inhibitors treatment in HCC and BCR-ABL+ leukemia (Fu et al., 2016; Lai et al., 2018). Combination of the PP2A inhibitor LB-100 (a derivative of norcantharidin) and anti-PD-1 blockage synergistically suppresses tumor growth in mice (Ho et al., 2018), suggesting a potential anti-cancer combination strategy. There are several other examples of LB100 combination with cytotoxic chemotherapy or radiation than have been reviewed recently (Mazhar et al., 2019).

Although B55α is the most abundant isoform among the B55 regulatory family in many cells and tissues (mRNA transcripts per million reported in the Human Protein Atlas), other less abundant B55 gene isoforms have also been found altered in cancer. The B55δ-encoding gene PPP2R2D is downregulated in hepatocellular carcinoma (HCC) tumors and cell lines, by a mechanism that may involve upregulation of miR133b. In the presence of cisplatin, a chemotherapeutic drug used for the treatment of advanced HCC singly or in combination, B55δ expression is increased. The increase in B55δ expression is associated with a G1 cell cycle arrest, inhibition of cell migration, cell colony area reduction, and apoptosis. These effects are diminished with B55δ knockdown, suggesting that B55δ is important for cisplatin-treatment sensitivity. Moreover, B55δ overexpression reduced xenograft tumor volume (Zhuang et al., 2016). PPP2R2B, the gene encoding B55β, has also been found downregulated in 90% of colon cancer samples and cell lines by a mechanism that involves hypermethylation of its promoter (Tan et al., 2010).

c. Other B subunit mutations or alterations.

As described above, the B’ family consists of five isoforms: B56α (PPP2R5A), B56β (PPP2R5B), B56γ(PPP2R5C), B55δ (PPP2R5D), and B56ε (PPP2R5E) (Chen et al., 2004). B’ family members are also dysregulated in cancer and likely associated with tumor suppressive functions. For example, a number of mutations of PPP2R5C have been recently described in lung cancer with a potential tumor suppressor function (Nobumori et al., 2013). One of the first links of a tumor suppressor function of a specific B’ subunit implicated B56γ. This link was established following the observation that the expression of hTERT, in combination with SV40 LT and oncogenic H-Ras^V12 in HEK cells (HEK TERV), is not sufficient for malignant transformation, while HEK cells expressing LT, hTERT, oncogenic H-Ras^V12 and SV40 small t antigen (st) (HEK TERST) are tumorigenic in immunodeficient mice (Sablina et al., 2010). Since st expressed in human cells was known to disrupt PP2A holoenzyme formation by competing or displacing B regulatory subunits, it was hypothesized that disruption of a particular B subunit could mediate st tumor suppressor functions. PPP2R5C knockdown in HEK TERV cells reduced B56γ levels and resulted in increased proliferation, cell transformation and tumorigenicity in immunocompromised mice to an extent similar to that observed in HEK TERST cells (Chen et al., 2004). Consistently, ectopic expression of B56γ in lung, breast and prostate cancer cell lines reduced cell proliferation and colony formation in soft agar, suggesting a potential tumor suppressor role of B56γ in multiple cell types (Chen et al., 2004). Of note, in response to DNA damage, B56γ partially suppresses cell proliferation and anchorage-independent growth by dephosphorylating Thr55 on Ser15-phosphorylated p53, which stabilizes it (Li et al., 2007; Shouse et al., 2008). However, overexpression of B56γ in p53^−/− cells reduces cell proliferation, though to a lesser extent than in wildtype cells, implicating a p53-independent mechanism of growth suppression (Nobumori et al., 2013).

Although PPP2R5D alterations are not frequently seen in human cancers, Ppp2r5d knockout in mice results in frequent spontaneous tumor development (Lambrecht et al., 2018). Specifically, 33% of the B56δ KO mice developed hematologic malignancies early in life (0–6 months). Older B56δ KO mice developed hepatocellular carcinomas (HCC) with a high frequency ranging from 17% (12–18 months) to 57% (18–24 months). Immunoblot analysis of B56δ KO HCC tumors revealed increased cMYC phosphorylation on Ser62 (ERK target) and inactivating phosphorylation of GSK3β on Ser9. Since GSK3β Ser9 is a target site of B56δ, it was hypothesized that the expected suppression of cMYC phosphorylation on Thr58 (GSK3 target) could potentiate cMYC stabilization via independent signaling leading to cMYC phosphorylation of Ser62, although other mechanisms cannot be ruled out (Lambrecht et al., 2018).

PPP2R3B codes for PR70, a regulatory subunit from the B” regulatory family. It is located on the X and Y chromosomes in a homologous region known as the pseudoautosomal region (PAR), which escapes X chromosome inactivation. Although PPP2R3B mutations are rare, the frequency of loss of the inactive X chromosome is high in females with breast, ovarian or melanoma tumors (Pageau et al., 2007; van Kempen et al., 2016). It has been reported that loss of PPP2R3B expression correlates with shorter overall survival in melanomas. Consistently, ectopic expression of PPP2R3B in the MM117 melanoma cell line derived from metastatic melanoma reduces subcutaneous tumorigenicity in mice, while PPP2R3B knockdown in melanoma MM57 cells increased the rate of tumor formation (van Kempen et al., 2016). PR70 overexpression in synchronized melanoma cells is associated with a small delay in progression through the G1/S transition concomitant with an increase in the abundance of the CDC6/CTD1 complex, while PR70 knockdown resulted in a decrease in total and chromatin associated CDC6 and CDT1 (van Kempen et al., 2016). Previous studies had shown that PR70 associates with CDC6, and that calcium enhances the interaction of PR70 with PP2A/C and PP2A/A. However, PR70 knockdown in HeLa cells resulted in an accumulation of CDC6 and a G1/S arrest. Of note, the accumulation of CDC6 was dependent on CDC6 residues that are phosphorylated by CDKs and promote its stability (Davis et al., 2008). While the levels of CDC6 were not examined in MM57 cells (van Kempen et al., 2016), the decrease in CDC6 Ser52 and CDT1 associated with chromatin may be at odds with the findings in HeLa cells (Davis et al., 2008).

d. Inhibitors SET, CIP2A, ENSA/Greatwall.

Aside from the alterations of genes encoding the three types of PP2A subunits, evidence continues to grow supporting that deregulation of endogenous PP2A inhibitors, including SET, CIP2A and Greatwall kinase, is associated with various human cancers.

SET is a physiologic PP2A inhibitor that binds directly to the PP2A/C subunit (Switzer et al., 2011). Consistent with the prevalent view that a fraction of PP2A holoenzymes exhibit tumor suppressor functions, SET appears to act as an oncogene. SET overexpression has been frequently detected and associated with poor prognosis in a large variety of cancers including breast, non-small cell lung (NSCLC) and metastatic colorectal cancer (mCRC) (Cristóbal et al., 2015; Huang et al., 2018; Liu et al., 2015; Pagano et al., 2018). Other studies also suggested that increased mRNA expression of SET is linked to adverse outcomes in chronic lymphocytic leukemia (CLL) patients (Brander et al., 2018). Strategies of restoring PP2A activity by inhibiting SET or the interaction of SET with PP2A/C are being pursued. In human medulloblastoma cells, shRNA knockdown of SET decreases cell viability and proliferation in a p53-dependent manner (Wei et al., 2018). Of note, drugs disrupting the PP2A/SET complex provide a promising therapeutic option in cancer treatment. A peptide designated OP449 that binds SET and activates PP2A has been shown to synergistically kill AML cell lines in combination with selective tyrosine kinase inhibitors (Agarwal et al., 2014). FTY720 (Fingolimod) and derivatives devoid of immunomodulatory function abrogate the SET/PP2A interaction increasing PP2A activity, and are also active in combination with FLT3 inhibitors synergistically inducing cell death in AML cell lines (Smith et al., 2016). In addition, novel derivatives of FTY720 such as alkoxy phenyl-1-propanone derivatives (APPDs) induce cell death in CLL cells through displacement of SET from PP2A, triggering intrinsic apoptosis (Pagano et al., 2018). A more extensive review of combination therapies that target PP2A has been published recently (Mazhar et al., 2019).

Another oncogenic PP2A inhibitor is Cancerous Inhibitor of PP2A (CIP2A). CIP2A was initially identified as a protein immunopurified using PP2A/A as bait by mass spectrometry (Junttila et al., 2007), and was more recently shown to form homodimers that bind B56α and B56γ (Wang et al., 2017). CIP2A also interacts with c-MYC, protecting c-MYC Ser62 from PP2A-mediated dephosphorylation and thus stabilizing it (Junttila et al., 2007). The same study demonstrated that ectopic expression of CIP2A can transform HEK-TERV cells, the human cell transformation model mentioned in the previous section, indicating that CIP2A can substitute for SV40 st in human cell transformation. It has been shown that CIP2A is overexpressed in head and neck squamous cell carcinoma (HNSCC) and colon cancers (Junttila et al., 2007). Overexpression of CIP2A mRNA has also been found in many cancers, including 39% in breast cancer and 87% in gastric cancer (Côme et al., 2009; Li et al., 2008). It has been suggested that CIP2A mRNA overexpression correlates with breast cancer aggressivity (Côme et al., 2009). CIP2A siRNA knockdown has been shown to reduce clonogenic potential in HeLa cells and three gastric cancer cell lines (Li et al., 2008). Consistently, CIP2A depletion in MDA-MB-231 breast cancer cells significantly reduced tumor volume in mouse xenografts (Côme et al., 2009).

As mentioned earlier, the main function of the Greatwall kinase (known as MASTL in mammals) is to inhibit PP2A/B55 by phosphorylating Arpp19 and/or ENSA, which are inhibitors of PP2A/B55 (reviewed in Castro and Lorca, 2018). Growing evidence suggests that upregulation of Greatwall promotes tumorigenesis (Rogers et al., 2018; Sun et al., 2017; Uppada et al., 2018). Increased MASTL mRNA and protein expression in colon cancer as compared to normal adjacent samples correlates with worse overall survival (Uppada et al., 2018). With similar observations made in gastric cancer and breast cancer (Rogers et al., 2018; Sun et al., 2017), Greatwall expression may serve as a cancer biomarker. As expected based on its cell cycle function, the oncogenic role of Greatwall is mediated by PP2A/B55 inhibition, as MASTL depletion causes mitotic cell death through PP2A activation in breast cancer cells (Yoon et al., 2018). On the other hand, it has been proposed that Greatwall can also trigger AKT hyperactivation independent of PP2A activity (Vera et al., 2015).

e. Final thoughts on therapy.

In the past, targeted therapies mainly focused on kinases, while phosphatases that equally contribute to cellular phosphorylation balance have been long neglected (reviewed in Westermarck, 2018). Based on what is discussed above, it is necessary to consider PP2A both as a diagnostic biomarker and a therapeutic target. Several PP2A activating drugs have emerged as novel approaches to cancer treatment, especially in combination therapies (reviewed in Mazhar et al., 2019). In addition to the PP2A activating compounds described above, tricyclic neuroleptics (phenothiazine and chlorpromazine) were found to interact with PP2A and increase its activity after being identified in a screen for drugs synergizing with Notch inhibitors to kill T-cell Acute Lymphoblastic Leukemia (T-ALL) cells (Gutierrez et al., 2014). These compounds were reengineered to eliminate their neuroleptic activities, generating a class of compounds know as small molecular activators of PP2A (or SMAPs) (Kastrinsky et al., 2015). Recently, SMAPs have been shown to have activity towards a number of cancer cells, especially in combination treatment (Kauko et al., 2018; McClinch et al., 2018; Sangodkar et al., 2017). Treatment with SMAP DT-061 in combination with MEK inhibitor AZD6244 effectively suppressed KRAS-driven lung cancer cell line xenografts in mice, even if forced to express cMYC (Kauko et al., 2018). This is consistent with the authors’ finding that PP2A inhibition confers MEK inhibitor and MEK/RAF resistance in these lung cancer cells. Given on one hand the preclinical success of small compounds that activate PP2A but also the many alterations associated with inhibition of subsets of PP2A holoenzymes in cancer, future work aimed at identifying more selective PP2A compounds holds promise and should help prevent toxicities and unwanted secondary effects.