The biogenesis of active protein phosphatase 2A holoenzymes: a tightly regulated process creating phosphatase specificity


V. Janssens, Laboratory of Protein Phosphorylation & Proteomics, Department of Cellular and Molecular Medicine, University of Leuven, Herestraat 49, PO Box 901, B-3000 Leuven, Belgium
Fax: +32 16 330 735
Tel: +32 16 330 684


Protein phosphatase type 2A (PP2A) enzymes constitute a large family of Ser/Thr phosphatases with multiple functions in cellular signaling and physiology. The composition of heterotrimeric PP2A holoenzymes, resulting from the combinatorial assembly of a catalytic C subunit, a structural A subunit, and regulatory B-type subunit, provides the essential determinants for substrate specificity, subcellular targeting, and fine-tuning of phosphatase activity, largely explaining why PP2A is functionally involved in so many diverse physiological processes, sometimes in seemingly opposing ways. In this review, we highlight how PP2A holoenzyme biogenesis and enzymatic activity are controlled by a sophisticatedly coordinated network of five PP2A modulators, consisting of α4, phosphatase 2A phosphatase activator (PTPA), leucine carboxyl methyl transferase 1 (LCMT1), PP2A methyl esterase 1 (PME-1) and, potentially, target of rapamycin signaling pathway regulator-like 1 (TIPRL1), which serve to prevent promiscuous phosphatase activity until the holoenzyme is completely assembled. Likewise, these modulators may come into play when PP2A holoenzymes are disassembled following particular cellular stresses. Malfunctioning of these cellular control mechanisms contributes to human disease. The potential therapeutic benefits or pitfalls of interfering with these regulatory mechanisms will be briefly discussed.


chaperonin-containing T-complex polypeptide complex


E3 ligase identified by differential display

HEK-TER cells

human embryonic kidney cells expressing SV40 large T, telomerase catalytic subunit and oncogenic Ras


immunoglobulin-α-binding protein 1


leucine carboxyl methyl transferase 1


mitogen-activated protein kinase kinase 3


midline 1




polyA-binding protein


protein phosphatase type 2A methyl esterase 1


protein phosphatase type 2A


protein phosphatase type 4


protein phosphatase type 6


phosphatase 2A phosphatase activator


suppressor of His4 transcription 4-associated protein


suppressor of His4 transcription 4


phosphatase 2A-associated protein of 42 kDa


T-complex polypeptide


type 2A-interacting protein


phosphatase 2A-associated protein of 42 kDa-interacting protein of 41 kDa


target of rapamycin signaling pathway regulator-like


target of rapamycin


T-complex polypeptide-1 ring complex

Introduction – how to create protein phosphatase type 2A (PP2A) specificity?

The vast majority (> 98%) of regulatory protein phosphorylations in a normal cell occur on Ser/Thr residues [1], and this process is coordinated by the opposing actions of Ser/Thr protein kinases and Ser/Thr protein phosphatases. Paradoxically, the estimated number of human genes encoding Ser/Thr phosphatases (40) is one order of magnitude smaller than the number of genes encoding Ser/Thr kinases (428) [2], resulting in the misconception that Ser/Thr phosphatases must be much less specific enzymes [3]. The highly conserved PP2A family represents a good example of how this paradox can be explained, as these phosphatases have acquired specificity by assembling a large number of different multisubunit holoenzymes from a limited number of catalytic subunits. In human cells, PP2A enzymes constitute an extended family of at least 96 holoenzymes, assembled from just two catalytic subunits [4]. Although approximately one-third of PP2A is estimated to occur as a dimeric complex composed of one catalytic C subunit and one structural A subunit (PP2AD) [5], the majority of these PP2A complexes are heterotrimers, consisting of one catalytic C subunit, one structural A subunit, and one variable regulatory B-type subunit (Fig. 1). The C subunit and A subunit are each encoded by two different genes, giving rise to almost identical α and β isoforms. Despite their near sequence identities, the isoforms of the C subunit and the A subunit, respectively, are functionally nonredundant [6–9]. For the B-type subunits, 15 human genes have been described, belonging to four distinct families: PPP2R2 (A to D) (encoding PR55/B or B55 subunits), PPP2R5 (A to E) (encoding PR61/B′ or B56 subunits), PPP2R3 (A to C) (encoding PR72/B′′ subunits) and the striatin genes STRN, STRN3 and STRN4 (encoding PR93/PR110/B′′′) (Fig. 1). Some of these genes generate multiple isoforms arising from alternative splicing or translation [4,10–12], eventually giving rise to at least 23 different B-type subunit isoforms. The final composition of the holoenzyme, resulting from the combinatorial assembly of just one C subunit, one A subunit, and one B-type subunit, eventually provides the essential determinants for subcellular targeting [13], substrate specificity and fine-tuning of phosphatase activity (reviewed in [3,14–16]), and these holoenzyme properties are mainly determined by the regulatory B-type subunits. A number of structural studies have provided insights into the mechanisms by which the regulatory B-type subunits act to regulate PP2A substrate specificity and activity [17–20]. In addition, the spatial and temporal expression patterns of the regulatory subunits within tissues, cells or subcellular locations can be highly divergent [11,13,21–25]. Together, the broad diversity in PP2A holoenzyme composition creates specificity and forms the basis for PP2A’s multiple physiological functions (reviewed in [16]). For instance, PP2A holoenzymes encompassing a PR55/B-type subunit are important regulators in mitosis (reviewed in [26,27]), apoptosis (reviewed in [28]), extracellular signal-related kinase signaling (reviewed in [29]), transforming growth factor-β signaling [30] and tau dephosphorylation in brain (reviewed in [31]), whereas many PP2A holoenzymes encompassing a PR61/B′-type subunit have other targets in mitosis/meiosis (reviewed in [26,27]), apoptosis (reviewed in [28]), development (reviewed in [32]) and brain (reviewed in [33]), and are considered to be the holoenzymes responsible for the major tumor suppressive properties of PP2A through their roles in Wnt/c-myc and Akt signaling (reviewed in [16,29,32,34–37]). PR72/B′′-containing trimers have reported roles in noncanonical Wnt signaling (reviewed in [29]), epidermal growth factor signaling [38], pocket protein regulation (reviewed in [39]), DNA synthesis [16,40] and neuronal signaling [33], whereas B′′′ trimers are functionally important in Hippo signaling [41]. However, most of these functional data originate from cellular models; the true biological functions of the B-type subunits remain relatively underexplored, owing to a lack of appropriate in vivo knockout models, particularly in higher animals. Only very recently have the functional consequences of knocking out a single B-type subunit in mice been reported, revealing a specific role for PR61/B′δ in the central nervous system, and providing the first direct evidence in vivo for nonredundant roles of individual B-type subunits in a complex mammalian system [42].

Figure 1.

 PP2A holoenzyme structure and PP2A C tail modifications. The majority of PP2A complexes within a cell are heterotrimers, consisting of a catalytic C subunit, a structural A subunit, and one regulatory B-type subunit. In humans, the C and A subunits are each encoded by two different genes (indicated in italic), giving rise to an α and β isoform of the C subunit (PPP2CA and PPP2CB), and A subunit (PPP2R1A and PPP2R1B) respectively. For the B-type subunits, 15 human genes (indicated in italic) have been described, giving rise to > 23 isoforms belonging to four different families, PR55/B (or B55, encoded by PPP2R2A to PPP2R2D), PR61/B′ (or B56, encoded by PPP2R5A to PPP2R5E), PR72/B′′ (encoded by PPP2R3A to PPP2R3C), and PR93/PR110/B′′′ (or the striatins, encoded by STRN, STRN3 and STRN4). This diversity in composition creates specificity and forms the basis for PP2A’s multiple physiological functions. Several post-translational modifications of the highly conserved PP2A C tail constitute important determinants of holoenzyme assembly, in particular reversible C-terminal methylation, which is catalyzed by the methyltransferase LCMT1 (encoded by LCMT1) and reversed by the methylesterase PME-1 (encoded by PPME1). In addition, Thr304 and Tyr307 phosphorylation, presumably catalyzed by a cyclin-dependent kinase (CDK) and a src-like or receptor tyrosine kinase (RTK), may also affect the recruitment of particular B-type subunits in the complex. In these cases, dephosphorylation is thought to occur by an autodephosphorylation mechanism.

Although the molecular structure of PP2A ensures the generation of many functionally divergent phosphatases, during the biogenesis of these PP2A holoenzymes the promiscuous phosphatase activity of the free catalytic subunit poses an inherent risk for the cell as long as its activity is not restricted by the interaction with other subunits [43]. In this review, we will describe how cells deal with this issue by synthesizing the free C subunit as an inactive form [44], necessitating a subsequent holoenzyme assembly-coupled activation mechanism [43]. These important insights led to a first model of PP2A biogenesis in yeast [43]. However, during this process, PP2A C stability needs to be ensured as well, as free PP2A C is supposed to be unstable and quickly degraded [45–47]. Here, we will describe how PP2A stability, enzymatic activity and holoenzyme assembly are tightly controlled by a complex network of five important cellular PP2A regulators, phosphatase 2A phosphatase activator (PTPA), leucine carboxyl methyl transferase 1 (LCMT1), PP2A methyl esterase 1 (PME-1), α4, and target of rapamycin signaling pathway regulator-like (TIPRL)-1, most of which have been independently discovered through classical biochemical methods. Although the roles of these PP2A modulators are not always entirely understood at the molecular level, we will review them one by one, and, on the basis of the assembled data, propose an extended, partially hypothetical, biogenesis model, which might stimulate future research and lead to a more comprehensive understanding of this important process that eventually creates PP2A specificity. This is of particular relevance because improper functioning of these cellular control mechanisms may have serious consequences for the cell and contribute to human disease.

PP2A C post-translational modifications and their role in holoenzyme assembly

Reversible PP2A C phosphorylation

The PP2A C subunit is one of the most highly conserved proteins, particularly its C-terminal TPDYFL(304–309) tail [16]. A number of biochemical and structural studies have underscored that post-translational modifications of this tail can provide additional contacts with specific B-type subunits and thus contribute to or interfere with holoenzyme assembly (reviewed in [4]) (Fig. 1). Although the identities of the physiological kinases have remained obscure, phosphorylation of Thr304 [48] and Tyr307 [49] has been described, and this is, in both cases, associated with PP2A inactivation. On the basis of mutational studies, Thr304 phosphorylation might selectively inhibit holoenzyme assembly with the PR55/B class of subunits, whereas Tyr307 phosphorylation might selectively affect interaction with particular PR61/B′ subunits (reviewed in [4]), providing a theoretical basis for dynamic subunit exchanges. However, whether such dynamic changes in holoenzyme assembly occur under the physiological circumstances in which increased PP2A C phosphorylation is observed remains to be more firmly established.

Reversible PP2A C carboxymethylation

A second important post-translational modification of the conserved PP2A C tail is reversible methylation of the free C-terminus of the C-terminal Leu309. This modification is catalyzed by the S-adenosylmethionine-dependent enzyme LCMT1 [50] and reversed by the methylesterase PME-1 [51] (Fig. 1). LCMT1 deficiency in cells causes apoptotic cell death [52,53], and both LCMT1 [53] and PME-1 knockout mice [54] have lethal phenotypes, suggesting that these enzymes have indispensable functions. In agreement with this, in a cellular model of human cell transformation, PP2A C methylation was found to be crucial for PP2A tumor suppressive activity [55]. Many studies have addressed the functional consequences of PP2A C methylation for PP2A activity and holoenzyme assembly in various model systems, leading to the conclusion that this modification is an absolute prerequisite for PR55/B subunit recruitment (reviewed in [4]). By contrast, the recruitment of A, PR61/B′, PR72/B′′ and the B′′′/striatins does not strictly require PP2A C methylation [4], although holoenzyme assembly with PR61/B′ is probably facilitated by it [17,18]. Structural studies on LCMT1 have revealed a canonical S-adenosylmethionine-dependent methyltransferase domain and a unique lid domain [56–58], which makes extensive dual contacts with the PP2A C active site and the C tail. The six C-terminal residues [TPDYFL(304–309)] of the C subunit tail occupy the deep LCMT1 active site pocket of the lid domain, and this is greatly facilitated by contacts between LCMT1 and the activated PP2A active site, suggesting that only an active conformation of PP2A C can actually be efficiently methylated [58]. As such, LCMT1-mediated PP2A methylation functions to prevent promiscuous phosphatase activity (activity that is not restricted by a B-type subunit) [43,44] by ensuring efficient conversion of only active C subunit into a proper holoenzyme; at the same time, formation of inactive trimeric holoenzymes is prevented. Thus, these findings [43,58] identified an additional role for methylation in PP2A biogenesis in vivo, providing an explanation for the apparent discrepancies between the methylation independency of PR55/B and PR61/B′ holoenzyme assembly in vitro [17,19,59] and the methylation dependency of this process within cells.

Interestingly, it has been suggested that LCMT1 activity and expression might be regulated by changes in homocysteine metabolism, which itself is affected by dietary intake of folate and vitamins B6 and B12 [60]. Therefore, a deficiency in these vitamins can lead to elevated plasma levels of homocysteine, decreased PP2A methylation, and impaired holoenzyme assembly with PR55/B subunits, contributing to conditions such as Alzheimer’s disease [60–62] and, potentially, cancer [63].

Surveying PP2A activity: the interplay between PME-1 and PTPA

PTPA (encoded by PPP2R4) was originally purified and identified as an ATP/Mg2+-dependent enzymatic upregulator of phosphotyrosyl phosphatase activity of PP2A in vitro [64]. The first clue to the real in vivo function of PTPA was provided in yeast, where, following de novo synthesis of PP2A C, functional interplay with PTPA is required for full phosphoseryl/threonyl PP2A activity [44]. Consistently, in mammalian cells, knockdown of PTPA resembles a PP2A-deficient state, and, depending on the context, fully transforms human immortalized human embryonic kidney cells expressing SV40 large T, telomerase catalytic subunit and oncogenic Ras (HEK-TER cells) [55] or causes apoptosis [44]. Subsequently, it was shown in mammalian tissues that PTPA activates the phosphoseryl/threonyl phosphatase activity of a native inactive PP2A complex that can be isolated and copurifies with PME-1 [65]. Apparently, both PTPA and PME-1 are parts of a complex mechanism that controls PP2A activity and biogenesis. This was again demonstrated in yeast, where the PTPA-dependent generation of active C subunit requires a functional interaction with the A subunit and is regulated by PME-1 [43], which specifically binds with high affinity to an inactive PP2A C conformation [51,65]. Crystallographic data have suggested that the interaction between PP2A C and PME-1 promotes several structural rearrangements that result in PME-1 activation on the one hand, and in the displacement of the metal ions from the PP2A C active site on the other hand [66], suggesting that PME-1 can actually inactivate PP2A. However, the biochemical data supporting this ‘metal eviction’ hypothesis were doubtful (the inactivation of PP2A required a 15-h incubation with PME-1 at 37 °C in the presence of metal ions [66]), and are in contrast with other studies, which failed to observe direct PP2A inactivation by PME-1 [65]. Regardless of this, there are indications that this PME-1 function (the stabilization of an inactive PP2A C form) may occur independently of its esterase activity. Indeed, in a PTPA-null yeast strain, PP2A has decreased activity and is hypomethylated, owing to an increased association with PME-1, whereas the additional deletion of PME-1 in this strain restores methylation but not activity [43,44]. The observation that PP2A, bound to a catalytically inactive PME-1 mutant, is methylated but still inactive [67] also corroborates these findings. Consequently, it is probably more conceivable that, in vivo, PP2A C is actually synthesized de novo in an inactive conformation [43,44], which is stabilized by PME-1 binding, independently of its esterase function. Results from quantitative interaction proteomics studies have confirmed that this PME-1-bound fraction represents a substantial proportion (20%) of the actual PP2A pool in the cell [68]. Notably, in some cancer cells, this fraction might be significantly increased, as PME-1 overexpression correlated with advanced stages of glioblastoma and might thus serve as a progession marker in this cancer type [69].

How can PTPA then reactivate this PME-1-bound, inactive PP2A pool? It has been shown in yeast that, in the absence of PTPA, PP2A C exhibits an altered conformation characterized by severely reduced Ser/Thr but increased Tyr phosphatase activity (indicative of reduced substrate specificity), reduced protein stability, and an increased dependence on metal ions [44]. Our own studies have suggested that PTPA may induce such a conformational change by acting as a peptidyl-prolyl cis/trans isomerase with Pro190 of PP2A C as its unique target [70]. A subsequent crystallographic study further supported this hypothesis, and demonstrated that, in yeast, a PTPA mutant without detectable isomerase activity fails to reactivate the PME-1-bound inactive form of PP2A in vitro and cannot restore the rapamycin resistance phenotype of a PTPA-negative strain in vivo [71]. In contrast, another structural study highlighted the importance of a physical interaction between PP2A C and a conserved C-terminal surface patch of PTPA, which turns the PTPA–PP2A complex into a composite ATPase, the function of which was suggested to determine the PTPA-induced change of PP2A substrate specificity in vitro [72]. However, so far, there is no in vivo evidence for PTPA acting as a PP2A tyrosyl phosphatase activator. In addition, we have proposed that this ATPase activity should be considered as a consequence of rather than an initiation step in the PP2A activation process, for two reasons: (a) ATP hydrolysis is minimal at the point of maximal PP2A activation, and increases several-fold afterwards; and (b) ATPase activity can be inhibited by the phosphatase inhibitor okadaic acid, suggesting that ATP has indeed become a substrate of PTPA-activated PP2A [73]. The fact that PTPA fails to activate PP2A in the presence of nonhydrolysable ATP analogs [72,73] may not necessarily reflect a need for ATP hydrolysis as such in the activation process; it may equally well be explained by inappropriate binding of these ATP analogs to the weak ATP-binding site in PTPA [72].

The biological functions of PTPA have been extensively (and, so far, exclusively) reported in Saccharomyces cerevisiae, in which two PTPA orthologs are found, the double deletion of which is lethal in a classical wild-type W303 yeast strain under standard growth conditions [74,75]. Their specific functions in proper cell cycle progression [75,76], responses to oxidative and genotoxic stress [77,78], mitosis [79] and transcription [80–82] seem to be mediated by physical or functional interactions with different yeast PP2A-like phosphatases, including the orthologs of PP2A, protein phosphatase type 4 (PP4) and protein phosphatase type 6 (PP6) [44,78,83–85]. Whether PTPA also regulates PP4 and PP6 in higher eukaryotes remains to be determined.

Noncanonical PP2A subunits – α4 and TIPRL1

PP2A stoichiometry and noncanonical PP2A complexes

The cellular levels of PP2A C (and some other subunits [4]) are tightly controlled by poorly understood autoregulatory mechanisms. For instance, it is not possible to increase the amount of PP2A C substantially by ectopic expression, because of tight translational and transcriptional regulation of endogenous PP2A C levels [86,87]. Likewise, it is thought that monomeric PP2A C is unstable and rapidly degraded, for instance when all B-type subunits [45,46] or the A subunit [45–47] are being downregulated. On the other hand, there is evidence, both in yeast and mammals, for the existence of atypical PP2A complexes, harboring, for instance, a C subunit and a B-type subunit without any need for the A subunit [88,89].

The in vivo PP2A subunit stoichiometry has been investigated in yeast, and was initially estimated to be A/B/C ∼ 1 : 4 : 8, with Rts1/B′ being 12 times more abundant than Cdc55/B within the B-type subunit pool [13] (Table 1), suggesting the existence of nontrimeric, atypical cellular PP2A C complexes. However, in a later article reporting on a global analysis of protein expression in yeast, these numbers differed significantly, with A/B/C ∼ 17 : 9 : 10, and Rts1/B′/Cdc55/B ∼ 1 : 29 [90] (Table 1). In both studies, a similar concept was used to calculate these numbers: the genomic coding sequences of the endogenous subunits were replaced by a sequence coding for a tagged form of the protein, the expression of which was subsequently quantified through the tag. Although Ghaemmaghami et al. affixed a large TAP tag C-terminally to their proteins (which might have been an unfortunate choice in the case of PP2A C), and Gentry and Hallberg used different tags (MYC and HA), which they affixed either C-terminally or N-terminally (with similar results), the reasons for these strikingly different data remain unclear. As a consequence, we have to conclude that PP2A stoichiometries within cells are, unfortunately, still poorly understood. Nevertheless, there is a wealth of evidence for the existence of noncanonical PP2A complexes, devoid of any A subunits or B-type subunits, as we will describe in the following two sections.

Table 1. Stoichiometry of PP2A subunits and modulators. The calculated (second column) and rounded (third column) number of protein molecules per yeast cell (× 103) are indicated for a given protein (first column) according to [90]. In the fourth column, alternative and estimated numbers of molecules are given for some proteins, based on the references in the fifth column. These numbers were calculated relative to the rounded number of 10 000 molecules per cell of PP2A C as found in [90], to enable direct comparison between all numbers. NA indicates that the relevant information was not found in the list of proteins [90] or in the SGD database.
 CalculatedRoundedOther dataReferences
PPH21 (PP2A C) 5.62   
PPH22 (PP2A C) 4.11   
Total PP2A C 1010[13]
TPD3 (A)16.917 1.2[13]
CDC55 (B) 8.6  0.4[13]
RTS1 (B′) 0.3  4.6[13]
Total B-type  9 5[13]
Tap42 (α4)NA  2[91]
Tip41 (TIPRL1) 5.71 6  
RRD1 (PTPA) 4.59   
RRD2 (PTPA) 2.43   
Total PTPA  7  
PPE1 (PME-1)NA ≥ 2[68]

α4 is an important regulator of PP2A C activity and stability

Biochemical characteristics of the PP2A C–α4 interaction

Several biochemical and genetic studies have demonstrated that α4 [also called immunoglobulin-α-binding protein 1 (IGBP1), or phosphatase 2A-associated protein of 42 kDa (Tap42) in yeast] forms a stable complex directly with PP2A C, independently of the A or B-type subunits [91–93]. Several deletion/site-directed mutagenesis approaches [93–95] and structural studies [96–98] have identified the mutual α4–PP2A C interaction domains, revealing that α4 and the A subunit require charged residues in separate but overlapping surface regions to associate with PP2A C. Such a mutually exclusive binding model has been confirmed in vivo by genetic experiments in yeast, where increased association of PP2A C with Tap42 was observed in a strain with deletion of the A subunit [43,91,99]. A large-scale interactomics study [100] additionally identified members of the ATP-dependent T-complex polypeptide-1 ring complex (TRiC)/chaperonin-containing T-complex polypeptide complex (CTT) chaperone family as constituents of the PP2A C–α4 complex (Fig. 2A), potentially aiding in their proper folding. The PP2A–α4 interaction does not rely on PP2A C methylation [101–103], and may even be enhanced in its absence [102]. Additionally, interactions of α4 with the PP2A-like phosphatases PP4 and PP6 [104,105], and of Tap42 with the yeast PP2A-like phosphatases suppressor of His4 transcription 4 (Sit4), Pph3, and Ppg1, have been demonstrated [91,94]. It was estimated that ∼ 2% of cellular PP2A C is bound to Tap42, and about 10% of Tap42 is bound to PP2A C [91] (Table 1).

Figure 2.

 Noncanonical PP2A complexes with the α4 and TIPRL1 subunits. (A) Within the PP2A C–α4 complex, α4 can act as a substrate-targeting subunit that binds potential substrates (e.g. MEK3 and Mid1) through its unstructured and flexible C-terminal domain. Cytosolic chaperonins from the TRiC/CCT complex may aid in the proper folding of α4 and/or PP2A C. (B) PABP1 and EDD are two additional α4-interacting partners, involved in protein translation and degradation respectively. In these cases, it remains to be determined whether PP2A C is part of these complexes as well. (C) Mechanisms of α4-mediated regulation of PP2A C stability. Although α4 recruits the PP2A C E3 ubiquitin ligase Mid1 to the C subunit, it is at the same time well-equipped to inhibit the polyubiquitination and degradation of PP2A C: (a) it can ‘cap’ ubiquitin moieties on PP2A C through its ubiquitin-interacting motif (UIM), thereby preventing further covalent addition of ubiquitin; and (b) it is itself monoubiquitinated on a Lys within its C-terminal domain, which may interfere with the interaction between Mid1 and the E2 ubiquitin conjugating enzyme. (D) PP2A C–TIPRL1 complexes. The yeast ortholog of TIPRL1, Tip41, was originally identified as a direct interaction partner of Tap42, the yeast ortholog of α4. In mammalian cells however, TIPRL1 directly interacts with PP2A C, and was found in PP2A complexes devoid of A subunit, in part as dimeric TIPRL1–C complexes and in part as trimeric TIPRL1–C–α4 complexes.

Notably, in yeast, the Tap42–PP2A C interaction is abolished by rapamycin via a mechanism that might involve inhibition of target of rapamycin (TOR) kinase-dependent phosphorylation of Tap42 [99] and that is preceded by dissociation of Tap42–PP2A C from TOR complex 1 [106]. TOR signaling links nutrient and energy availability to cell growth, and, accordingly, nutrient deprivation, which is known to repress TOR activity, also results in Tap42–PP2A C dissociation [91]. In mammalian cells, however, this type of metabolic regulation of the α4–PP2A C complex remains controversial, with some reports demonstrating no rapamycin effect at all on α4–PP2A C binding [103,107–110], and others confirming the rapamycin sensitivity of the interaction [92,93,111,112].

PP2A C–α4 substrates and interacting proteins

The regulatory effect of α4 on PP2A C activity is ambiguous, with some reports favoring inhibition [103,104,107,113] and some activation [92,93,101,110]. This may be explained by the different substrates used in these studies, and it is therefore probably more correct to state that α4 affects the substrate specificity of PP2A C, rather than just increasing or inhibiting its activity. Likewise, the effect of PP2A C-binding inhibitors, such as okadaic acid and microcystin, on the PP2A C–α4 interaction is unclear, with some studies reporting a disruption of the complex [108,114], and some reporting no effects at all [95,109]. This might question the conclusion that binding of α4 allosterically modifies rather than occludes the PP2A C active site [115], although such an allosteric effect would be very much in line with the proposed substrate-specifying role of the combined binding of an A and a B-type subunit to PP2A C (reviewed in [14,15]). However, in contrast to the A subunit, which needs a B-type subunit to mediate direct contacts with the substrate, α4 is proposed to directly engage these phosphoprotein substrates via its structurally unordered, flexible C-terminus [96,97]. Examples of α4–PP2A C substrates that are targeted in such a way include mitogen-activated protein kinase kinase 3 (MEK3), one of the upstream kinases in cytokine-induced p38 signaling [116], and the RING-domain ubiquitin E3 ligase midline-1 (Mid1) [117–120] (Fig. 2A). Mid1 is a microtubule (MT)-associated phosphoprotein that is mutated in the developmental disease Opitz syndrome. A minority of these mutations leave the MT interaction intact but disturb the interaction with α4 (and thus PP2A C), resulting in Mid1 hyperphosphorylation and loss of its active transport along the MT [121], consistent with Mid1 being a substrate of PP2A C–α4. The majority of Opitz syndrome-associated Mid1 mutations, however, result in loss of its MT binding, while the association with α4 (and PP2A C) is preserved [122]. Notably, this coincides with increased dephosphorylation of MT-associated proteins and significantly higher PP2A C levels specifically in the MT fraction [118], suggesting that Mid1-mediated targeting of α4 to the MT might be a way of controlling the stability of MT-associated PP2A (trimers?). α4 also interacts with E3 ligase identified by differential display (EDD), a predominantly nuclear RING-domain and HECT-domain E3 ubiquitin ligase [123], and with polyA-binding protein (PABP), a cytosolic mRNA-binding protein that recruits translational factors to the polyA tails of mRNAs [123] (Fig. 2B). However, it is currently unclear whether PP2A C is also part of these α4–EDD or α4–PABP complexes, or what the function of these complexes might be. Together, all of these biochemical data suggest a dual role for α4 within PP2A C–α4: (a) as a substrate-targeting and substrate-specifying PP2A subunit, e.g. towards MEK3 and Mid1; and (b) as a potential modulator of PP2A C translation (via interaction with PABP), folding (via TRiC/CTT), and stability (via interaction with Mid1 and EDD).

A role for α4 in PP2A biogenesis

Intriguingly, Tap42 and α4 are essential proteins, and their gene knockout in yeast or mice is lethal [91,112]. Conditional gene targeting of α4 demonstrated a pivotal role for α4 in the proliferation and survival of B and T lymphocytes, embryonic fibroblasts, adipocytes, and hepatocytes [112,124,125]. In fact, a profound apoptotic phenotype was found in α4-deficient cells, mediated through the intrinsic pathway and initiated by p53 and c-Jun-dependent transcriptional mechanisms [112]. Deletion of α4 also led to impaired cell spreading and migration, correlating with reduced activation of Rac1 [126]. All of these observations suggested that α4 might play a more prominent role in PP2A biology than just targeting a particular set of PP2A substrates. A recent key study in this respect demonstrated that α4 deletion leads to progressive loss of PP4, PP6, PP2AC, and PP2A A, and subsequent loss of PP2A activity towards a wide variety of established PP2A substrates [113]. Coexpression of α4 significantly increased the protein levels of ectopically expressed PP2A C, and resulted in a substantial reduction of ubiquitinated C subunit, suggesting that α4 protects the C subunit from degradation [113]. Notably, under cellular stress conditions such as heat shock, DNA damage, or nutrient withdrawal, PP2A heterotrimers seem to be relatively unstable, and overexpression of α4 resulted in faster recovery of PP2A activity following such stress conditions, suggesting that the released C subunits may be sequestered by α4 to protect them from degradation and thus to allow more rapid reassembly into adaptive PP2A complexes that promote stress recovery and cell survival [113]. The ability of α4 to protect PP2A C from proteasomal-mediated degradation is dependent on the integrity of its PP2A C-binding domain, its C-terminal domain (known to recruit Mid1), and a unique ubiquitin-interacting motif within its N-terminal domain that might interact with ubiquitinated C subunits and prevent further covalent addition of ubiquitin [98,120] (Fig. 2C). Interestingly, α4 is monoubiquitinated in cells [120], potentially on a Lys within its C-terminal domain [127]. A small peptide encompassing this residue can inhibit E3 ubiquitin ligase activity of Mid1 in vitro, potentially through disruption of the Mid1 interaction with the E2 ubiquitin-conjugating enzyme (Fig. 2C). As Mid1 depletion leads to increased cellular PP2A C levels [128], this would suggest that another potential mechanism by which α4 might protect PP2A C from degradation is direct regulation of enzymatic Mid1 E3 ligase activity [127].

The proposed function of α4 as a preserver of PP2A C integrity [113] might seem at odds with its role as the ‘bridging’ protein that puts PP2A C in the immediate proximity of its E3 ubiquitin ligase Mid1, hence promoting PP2A degradation [118]. This could suggest that α4 might rather function as a ‘modulator’ of PP2A stability, and that this function might be under tight regulatory control, e.g. by post-translational modifications of α4 [120] or other mechanisms determining, in a given condition, which of the known α4-interacting proteins (Fig. 2A,B) will be recruited. This remains to be further established.

Aberrant α4 expression in human disease

In line with its antiapoptotic role [112,116], ectopic expression of α4 in immortalized HEK-TER cells leads to cell transformation [129]. Although this might seem to conflict with its function as a PP2A C stabilizer, this could reflect its ability to compete with the A subunit for PP2A C binding, and hence overexpression of α4 might lead to a relative decrease in tumor suppressive PP2A holoenzymes. In accordance with such an oncogenic role, α4 was found to be highly expressed in several human cancers (in 87.5% of primary hepatocellular carcinomas, 84% of primary lung cancers, and 81.8% of primary breast cancers) by downregulation of the microRNA miR-34b [129]. Likewise, α4 was found to be overexpressed in lung adenocarcinoma, and this correlated with cancer stage, suggesting that it might even serve as a biomarker for invasive pulmonary adenocarcinoma [130].

TIPRL1 – another player in PP2A biogenesis?

Phosphatase 2A-associated protein of 42 kDa-interacting protein (Tip41) was originally discovered in yeast as a Tap42 interaction partner that competed with Sit4 (the yeast ortholog of PP6) for Tap42 binding [131]. With a yeast two-hybrid strategy, Tip41 (bait) was shown to interact with all PP2A-like yeast phosphatases, including Pph21/22, Sit4, Pph3, and Ppg1 [100]. In mammalian cells, however, TIPRL1 (also called hTip41 or TIP) does not directly bind α4 but rather primarily interacts with the PP2A, PP4 or PP6 C subunits [100,109,132], and this association is insensitive to rapamycin [109,132]. In vitro and in vivo binding assays have supported this view, demonstrating that TIPRL1 does not compete with α4 for PP2A C binding, and that even a trimolecular TIPRL1–C–α4 complex can be formed and found as an endogenous complex in cells [109]. In contrast, two other studies failed to detect such a trimolecular complex in coimmunoprecipitates of ectopically coexpressed α4 and TIPRL1, potentially because the levels of endogenous C subunit might have been limiting [100,132]. Interestingly, the A subunit was not detected in isolated TIPRL1–PP2A C complexes [132], and significantly more PP2A C was found to bind TIPRL1 than α4 [132], suggesting that TIPRL1 may be part of both dimeric PP2A C–TIPRL1 and trimeric TIPRL1–C–α4 noncanonical PP2A complexes (Fig. 2D). In yeast, total Tip41 levels were estimated to be ∼ 60% of total PP2A C [90] (Table 1).

TIPRL1 is a ubiquitously expressed PP2A inhibitory protein that has been shown to inhibit free PP2A C as well as PP2AD [109,132]. Its effect on PP2A C activity was proposed to be allosteric in nature, as okadaic acid failed to dissociate the TIPRL1–PP2A C complex [109,132]. Notably, TIPRL1 may play an important role in DNA damage and repair signaling, as it regulates PP2A enzymes that oppose ATM/ATR-dependent phosphorylation events [132], and is part of a well-defined yeast PP4 holoenzyme, which is involved in regulating cisplatin sensitivity [100]. Alternatively, or additionally, we speculate that, similarly to the proposed role of α4 [113], these observations might actually link TIPRL1 to the (re)assembly of adaptive PP2A or PP2A-like complexes that promote recovery from DNA damage, implying a potential role for TIPRL1 in PP2A biogenesis or degradation after holoenzyme disassembly. The observation that, unlike PME-1, TIPRL1 can proactively inhibit active PP2A C [109,132] may be of particular relevance in this respect, because the unrestricted activity of the released C subunit could obviously pose a serious threat to the cell. In any case, all these observations probably warrant further investigation of the (unknown) TIPRL1 status in human disease, particularly in cancer.

An extended PP2A biogenesis model

The findings described above have inspired us to assemble an extended PP2A biogenesis model, based on the initial model in yeast [43] and the key role of α4 in regulating PP2A C stability [113] (Fig. 3). Although it is partly speculative, and in part it has to reconcile some contradictory observations, this model might be useful for further experimental investigation of some of the inferred interconnections. In addition, it may apply both to the de novo synthesis of PP2A holoenzymes and to their (re)assembly following cellular stress, such as DNA damage. At this point in particular, TIPRL1 may play an important role in preventing promiscuous phosphatase activity of the active C subunit released from the holoenzyme after its stress-induced disassembly. Afterwards or at the same time, α4 may play a decisive role in regulating whether the released C subunit will be degraded or recycled [113] (Fig. 3). We have also included a number of specific cytosolic chaperonins (Fig. 3), identified in some studies on the PP2A interactome, which may aid in the proper folding of individual PP2A subunits/regulators during or immediately following their translation [133,134]. In this regard, Cα, Cβ, PTPA, the striatins and all PR55/B subunits were found in complexes with members of the TRiC/CCT complex, whereas PP2A Cβ was uniquely linked to the prefoldin complex [134]. A key point in the biogenesis model is the synthesis of PP2A C subunit as an inactive protein, which is a requirement to protect the cell from any promiscuous PP2A action [43,44]. Ubiquitin-mediated degradation of unstable, free C subunit is prevented through the association with α4 [113] (see Fig. 2C for more details). As both TIPRL1 [109] and PTPA [84,85] can be found as parts of trimolecular α4–PP2A C complexes, we speculate that TIPRL1 might stabilize an inactive α4–PP2A C complex, whereas the action of PTPA might result in the generation of an active α4–PP2A C complex that is fully capable of dephosphorylating its specific set of cellular substrates. In such a scenario, TIPRL1 might thus play a similar surveillance role as PME-1 in controlling the assembly of active α4–PP2A complexes, and PTPA would be able to antagonize TIPRL1 function. Alternatively, it is possible that TIPRL1 is not directly involved at this stage of PP2A biogenesis (Fig. 3). The assembly of the classical trimeric holoenzymes would involve the appearance of the A subunit, which competes with α4 for PP2A C binding, and of PTPA, which refolds the C subunit into an active form [44], potentially through its prolyl-peptidyl cis/trans isomerase activity on Pro190 of PP2A C [70,71]. The activated PP2A dimer can now be efficiently methylated by LCMT1 [58]. PP2A C methylation subsequently allows or facilitates binding of the B-type subunits [4]. PME-1 regulates this process by binding and stabilizing the not yet activated PP2A dimeric AC form, and, at the same time, by preventing C subunit methylation [43]. Also in this case, the action of PTPA will result in PP2A C activation and the release of PME-1 [43,65], thus again allowing PP2A methylation and the completion of holoenzyme assembly (Fig. 3). Finally, we have previously reported the potential existence of PTPA-reactivatable, inactive PME-1-bound PP2A trimers [65]. On the basis of quantitative interactomics data [68], we speculate that – unlike PME-1-bound dimeric PP2A – PME-1-bound trimeric PP2A might not represent a significant PP2A fraction, but rather should be considered as a transient complex resulting from aberrations in the biogenesis process (Fig. 3).

Figure 3.

 Extended PP2A biogenesis model. This model is essentially based on three key papers [43,44,113], which have been integrated and extended, guided by the available literature on the five key PP2A modulators in this process (α4, TIPRL1, PTPA, PME-1, and LCMT1). Some parts of it are speculative (indicated by dashed lines and/or question marks). See main text for a more detailed explanation of the particular interconnections made. PFD, prefoldin.

Conclusions, perspectives, and clinical implications

Although the process is not yet entirely understood, the proposed PP2A biogenesis model obviously testifies to the exquisite controls that cells have developed to safely assemble functional and catalytically competent PP2A holoenzymes, each with their intrinsic specificities determined by the nature of the associated B-type subunit. As explained, the key premise in this model is the synthesis of PP2A C as an inactive enzyme [44], which, by itself, prevents any promiscuous PP2A activity immediately after its translation, but subsequently necessitates the existence of an activation mechanism that is tightly coupled to holoenzyme assembly [43]. This activation-and-assembly mechanism seems to occur stepwise in a tightly controlled manner, and involves at least five major PP2A regulators/binding partners, α4, the PP2A A subunit, PTPA, PME-1, and LCMT1, the single genetic knockout of all of which is lethal in mice ([8,53,54,112] and our own unpublished data), further underscoring the importance of these proteins in mammalian physiology. As explained, the roles and mechanisms of action of these modulators are highly diverse. α4 acts as a modulator of monomeric C subunit stability, the binding of which is mutually exclusive with binding of the A subunit. Hence, the A subunit should compete with α4 to be able to act as the PP2A holoenzyme scaffold. PTPA enzymatic activity is required to fold inactive PP2A C into an active conformation, potentially by a prolyl cis/trans isomerase activity. PME-1 has a dual regulatory function, by binding and stabilizing the inactive PP2A C conformation, and at the same time by preventing premature methylation through its methylesterase activity. Finally, LCMT1 carboxymethylates the C-terminal tail of activated PP2A C to allow or facilitate B-type subunit binding.

Although the stoichiometry of PP2A within cells is poorly understood and should probably be revisited, because of conflicting data (Table 1), there is evidence that a significant proportion (at least one-fifth) [68,91] of PP2A C may be bound to at least one of these regulators (as opposed to its presence in genuine PP2A trimers), suggesting a high degree of adaptability and dynamics in PP2A complex formation. This characteristic might be of particular importance under cellular stress situations, ensuring rapid adaptation and recovery. Whereas α4 may play a role in preserving the stability of PP2A C released after stress-induced holoenzyme disassembly [113], the putative involvement of the PP2A inhibitor TIPRL1 in neutralizing the promiscuous phosphatase activity of the released C subunit under these conditions is, although speculative, highly likely and should be further established. Another research line suggested by our model pertains to the potential interplay between PP2A C, PTPA, and α4, the importance of which in generating catalytically active PP2A C–α4 complexes might have been overlooked so far. Finally, the role of α4 as a modulator of PP2A stability should be further addressed, in particular to understand under which conditions it might act as a stabilizer of PP2A C, and in which conditions it might rather promote PP2A C degradation. It is possible that specific post-translational modifications of α4 or mutually exclusive binding mechanisms of the α4 interaction partners might play a decisive role in this.

Another future line of research might be to explore the potential conservation – at least partially – of this biogenesis mechanism for other PP2A-like phosphatases, in particular for PP4 and PP6. These highly homologous phosphatases also acquire their specificities by assembling several structurally different holoenzyme complexes from a single catalytic subunit (Fig. 4). For PP4 (encoded by PPP4C), five canonical subunits, which assemble into six different PP4 holoenzymes, have been identified: PPP4R1 [135,136], PPP4R2 [137], PPP4R3A (also called SMEK1) [100], PPP4R3B (also called SMEK2) [100], and PPP4R4 [134,138] (Fig. 4A). Two of these also bind TIPRL1 in substoichiometric amounts, and are functionally important in DNA damage responses [100]. It is of note that some regulatory subunits of PP2A (PR65/Aα, PR55/Bα, PR55/Bδ and PR61/B′δ) have been found in complexes with PP4 C as well [68,134,138,139]. For PP6 (encoded by PPP6C), three subunits have been identified [PPP6R1, PPP6R2, and PP6R3, also known as the Sit4-associated proteins (SAPs)], which each directly interact with PP6 C, reminiscent of the binding of the Aα and Aβ scaffolds to PP2A C [140] (Fig. 4B). In addition, three ankyrin-repeat proteins (Ankrd28, Ankrd44, and Ankrd52) have been suggested to act as B-type-like subunits of PP6 that are recruited through the SAPs [141] (Fig. 4B). Data on the estimated stoichiometries of these PP4 and PP6 subunits, if known, can be found in Table 2. As explained before, PP4 and PP6 share several common regulatory binding partners with PP2A C, in particular α4, TIPRL1, and PTPA, which have all been reported to physically and functionally interact with these phosphatases [44,78,83–85,91,94,100,104,105,109,113,132] (Fig. 4). It would thus be of interest to determine whether PP4 and PP6 are also synthesized as inactive C subunits, and whether PTPA might be able to reactivate them. For PP6, such a scenario is definitely feasible, because the catalytic activity of Sit4/PP6 was found to be decreased in yeast strains lacking the PTPA ortholog RRD1 [44,85]. A role for α4 in regulating PP4 and PP6 stability has already been fairly well established [113]. In addition, like PP2A, PP4 is known to be reversibly carboxymethylated [142], by a so far unknown methyltransferase–methylesterase couple and with unknown consequences for PP4 holoenzyme assembly or activity. Whether PP6 is also carboxymethylated in vivo remains to be determined. Although potential similarities in PP2A/PP4/PP6 biogenesis should be more firmly established, essentially through similar methodologies as the ones used in the studies on PP2A, many of these parallel observations are suggestive of common regulatory mechanisms within the PP2A-like phosphatase family.

Figure 4.

 PP4 and PP6 holoenzymes. (A) PP4. Two dimeric and two trimeric canonical PP4 complexes have been identified. The dimeric complexes consist of the catalytic C subunit and regulatory R1 or R4 subunits, and the trimeric complexes consist of the catalytic C subunit, a scaffolding R2 subunit, and a regulatory R3A or R3B subunit. In contrast to the R2 subunit, the R1 and R4 subunits show some structural similarities with the PP2A A subunits. Gene (italic) and alternative names are indicated. The noncanonical subunits α4 and TIPRL1 are found in complexes with PP4 C (and TRiC/CTT chaperonins) and in trimeric PP4 C–R2–R3A (or R3B) complexes, respectively. (B) PP6. In analogy with PP2A, it has been suggested that the composition of the PP6 holoenzymes is the result of the combinatorial assembly of the single PP6 C subunit (encoded by PPP6C), one scaffold subunit (encoded by the SAP genes PPP6R1, PPP6R2, and PPP6R3) and one regulatory subunit [encoded by the ankyrin repeat subunit (ARS) genes Ankrd28, Ankrd44, and Ankrd52]. PP6 C is also found in noncanonical complexes with α4 and TIPRL1, respectively.

Table 2. Stoichiometry of PP4 and PP6 subunits based on [90]. The numbers indicate the calculated and rounded number of protein molecules per yeast cell (× 103) and can be directly compared with the numbers in Table 1. ARS, ankyrin repeat subunit.
 [90]Rounded [90]Rounded
PPH3 (PP4 C)2.843Sit4 (PP6 C)3.974
No PP4R1 in yeast  SAP1555.966
Ybl046w (PP4R2)1.011SAP18511.211
PSY2 (PP4R3)7.017SAP190NA 
No PP4R4 in yeast  No ARS in yeast  

On the basis of the discovery of disease-associated aberrations in some of these important PP2A modulators – e.g. increased expression of PME-1 or α4 in cancer [69,129,130] or decreased PP2A methylation in Alzheimer’s disease [61,62,143] – some studies have proposed that these regulators might be potential therapeutic targets. In this respect, very recently, two structurally unrelated classes of selective covalent PME-1 inhibitors have been described [144,145], and were proposed to be of potential therapeutic relevance in cancer or neurodegeneration. Perhaps some caution about the potential detrimental consequences of using such inhibitors would be justified, given the essential, potentially esterase activity-independent regulatory role of PME-1 during PP2A biogenesis: it is, indeed, not unthinkable that such inhibitors might significantly disturb proper PP2A biogenesis, and thus have undesirable side effects. Interference with the PP2A C–α4 interaction was recently demonstrated for the antidiabetic drug metformin, a dimethylbiguanide [146]. This drug induced PP2A-mediated tau phosphatase activity in murine primary cortical neurons and brain by preventing the association of PP2A C with α4 (and hence Mid1), thereby potentially stabilizing the MT-associated PP2A pool. The proapoptotic and antitumour drug lactoferrin has also been shown to bind α4, leading to a subsequent reduction of α4-associated PP2A activity in lung adenocarcinoma cells [147]. Despite these interesting observations, the overall effects of such compounds on the PP2A system as a whole should be carefully evaluated before clinical use is considered. On the other hand, the future development of specific inhibitors of these PP2A regulators might still provide novel, alternative tools with which to specifically interfere with PP2A function(s) in cells or animal models within a fundamental research context, as many more secrets of these intriguing enzymes remain to be discovered.


Our research is currently supported by grants from the Research Foundation Flanders (G.0582.11N), Concerted Research Actions (GOA/2008/16 and GOA/2012/12), and the Belgian Science Policy (IUAP P6/28).