PP1 and PP2A phosphatases – cooperating partners in modulating retinoblastoma protein activation

Authors


V. Kolupaeva, Department of Microbiology, New York University School of Medicine, 550 First Avenue, New York, NY 10016, USA
Fax: +1 212 263 3276
Tel: +1 212 263 5331
E-mail: victoria.kolupaeva@nyumc.org

Abstract

The retinoblastoma/pocket protein family is one of the master regulators of the eukaryotic cell cycle. It includes the retinoblastoma protein (Rb) and the related p107 and p130 proteins. The importance of the Rb pathway for homeostasis and tumour suppression is evident from the fact that inactivating mutations in Rb are frequently associated with many cancers. Rbs regulate the cell cycle by controlling the activity of the E2F family of transcription factors. The activity of Rb proteins themselves is modulated by their phosphorylation status at several Ser/Thr residues: phosphorylation by cyclin-dependent kinases inactivates Rb proteins and positively influences the transcription of genes necessary for cell cycle progression. Although the mechanisms of cyclin-dependent kinase-mediated inactivation of Rb proteins are understood in great detail, our knowledge of the process that counteracts Rb phosphorylation is still quite limited. The present review focuses on the Ser/Thr phosphatases that are responsible for the dephosphorylation and thus activation of Rb proteins. Two major scenarios are considered: (a) when pocket proteins are dephosphorylated during regular cell cycle progression and (b) when rapid dephosphorylation is dictated by external stress or growth inhibitory conditions, such as oxidative stress, UV radiation or other DNA-damaging stimuli, and cell differentiation factors. It transpires that protein phosphatase 1 and protein phosphatase 2A can efficiently modulate pocket protein activity in a highly context-dependent manner and both are tightly regulated by the presence of different regulatory subunits or interacting proteins.

Abbreviations
ATRA

all-trans retinoic acid

CalA

calyculin A

CDK

cyclin-dependent kinase

CHX

cycloheximide

FGF

fibroblast growth factor

GSK-3

glycogen synthase kinase-3

I-1

inhibitor-1 of PP1

I-2

inhibitor-2 of PP1

MYPT

myosin phosphatase targeting subunit

NLS

nuclear localization signal

OA

okadaic acid

Phactr4

phosphatase and actin regulator 4

PNUTS

phosphatase 1 nuclear targeting subunit

PP

protein phosphatase

Rb

retinoblastoma protein

SV40

simian virus 40

The activity of retinoblastoma proteins is mediated by their phosphorylation status

Proper regulation of the cell cycle is vital for normal development, cell growth and differentiation. A family of Rb proteins or pocket proteins is one of the main players in this well-orchestrated process. Rb protein (pRb or Rb or Rb1) was originally identified as a tumour suppressor gene that is deleted or mutated in retinoblastomas, as well as in a wide-variety of other cancers [1,2]. Subsequently, two other proteins were isolated, p107 and p130 (also known as Rb2), which share structural and functional similarity with Rb [3–6] (Fig. 1A). Rb proteins possess a bipartite pocket structure mediating their association with many cellular proteins that are mainly transcription factors or other components of the transcriptional machinery, and that often harbour an LxCxE pocket domain-binding motif [7]. E2F transcription factors undoubtedly are the most studied partners of Rb proteins [8–11], controlling expression of many cell cycle-related genes, including cyclin E, cyclin A and CDC25C, amongst others. They form heterodimers with members of the DP family to bind DNA. Some E2F family members act as activators (E2F1–3) and some as repressors (E2F4–5). Pocket proteins associate differentially with these subgroups [12]: Rb complexes sequester E2F1–3, thereby preventing activation of cell cycle progression genes (Fig. 1B), and p107/p130 exclusively target E2F4 and E2F5, directly participating in the repression of cell cycle-related genes. The role of Rb as a tumour suppressor is well established, whereas the role of p107 and p130 in tumor suppression is less clear, despite knowledge of several examples where a loss of their activity is correlated with oncogenic transformation [13].

Figure 1.

 Pocket proteins: major cell cycle regulators. (A) Structural features of Rb family members. The pocket region of retinoblastoma proteins consists of two conserved functional domains identified as A and B pockets. This region is required for functional interaction with E2Fs. Rb protein has an additional pocket region for binding E2F1 transcription factor. Both p107 and p130 have an insertion (loop domain) in the B part of the pocket. p107 and p130 physically associate with cyclin A/CDK2 and cyclin E/CDK2 complexes through a cyclin-binding RXL motif located in the spacer domain. The N-terminal regions of p107 and p130 are also able to act as CDK inhibitors. Changes in the phosphorylation status of Rb proteins can be clearly detected by western blotting as a result of a difference in the electrophoretic mobility of hypo- and hyperphosphorylated forms. Western blots of lysates from FGF-treated rat chondrosarcoma (RCS) cells demonstrate changes in the phosphorylation status of the pocket proteins because FGF causes growth arrest of RCS cells with concomitant dephosphorylation of all members of Rb family. (B) Rb-mediated regulation of the expression of genes containing E2F sites. Binding of hypophosphorylated Rb to E2F/DP heterodimer prevents transcriptional activation. Inactivation of Rb by phosphorylation releases E2F/DP and positively influences cell cycle progression. (C) CDKs and the cell cycle. A diagram of a typical cell cycle shows G1-, S- and G2/M-phases. Approximate points when changes in Rb phosphorylation status occur are indicated by arrows. CDKs involved in progression throughout the cell cycle are indicated in accordance with their sequential activation.

The ability of retinoblastoma proteins to interact with E2F factors is modulated by their phosphorylation status during the cell cycle. In particular, the family of proline-directed cyclin-dependent kinases (CDKs), whose activity depends on an appropriate partner-cyclin, phosphorylates the pocket proteins at several Ser/Thr sites and is responsible for their inactivation. During the early G1-phase, Rb is hypophosphorylated and active as growth suppressor (Fig. 1C). In the mid G1-phase, Rb is sequentially phosphorylated by cyclin D/CDK4 complexes. Later, at the G1/S boundary, it is phosphorylated by cyclin E/CDK2 and, in the S-phase, by cyclin A/CDK2 [14–18]. Thus, Rb phosphorylation peaks at the S-phase and diminishes at the exit from mitosis. When phosphorylated, Rb proteins release E2F1–3, which positively influence cell cycle progression. Phosphorylation sites of Rb are well defined in vivo [19] and, for the majority of them, phospho-specific antibodies are available (Fig. 2, red), providing useful tools for functional studies.

Figure 2.

 Functional domains and phosphorylation sites of pocket proteins. Identified phosphorylation sites and respective kinases/phosphatases are shown for retinoblastoma proteins. Only sites that are conserved between human and rodent species are indicated, based on the previous data [19,23–25,28]. The phosphorylation sites conserved between p107 and p130 are underlined. Red residues indicate amino acids with available phospho-specific antibodies. The Rb PP1-binding site (‘RVxF’ motif) within its C-terminal domain is underlined. The p107 PP2A-binding domain is located within the spacer region in between the pocket domains. Within p130, this region also harbours a cyclin A binding motif (underlined). The identified bipartite NLS for Rb and p130 is also indicated.

Although the levels of Rb expression are barely changed throughout the cell cycle and in quiescent cells, the expression of p107 and p130 is cell cycle-dependent. In G0, p107 is present at low levels in a hypophosphorylated form. The level of p107 expression increases significantly in late G1, when the majority of p107 is hyperphosphorylated. Later, p107 phosphorylation is reversed and p107 can bind repressive E2Fs [20]. Hyper- and hypophosphorylated forms of p107 can be easily resolved by western blotting as two bands as a result of the difference in their respective electrophoretic mobilities (Fig. 1A). Regarding p130, four different forms (one nonphosphorylated and three phosphorylated) can be detected by western blotting [15] (Fig. 1A). Dephosphorylated p130 is found in G0 and early G1. Phosphorylated forms 1 and 2 of p130 can be detected in quiescent cells. When cells are progressing from G0 to G1, p130 is hyperphosphorylated to form 3, which is subsequently targeted for degradation [21], resulting in a dramatic drop in p130 levels as cells move to the S-phase [15,22].

Cyclin D/CDK4, together with cyclin E/CDK2 and cyclin A/CDK2 complexes, phosphorylates Rb proteins in a sequential manner [18,23]. In total, 16 Rb phosphorylation sites have been identified [19] (Fig. 2). Among these, phosphorylation of either T821 by cyclin A/CDK2 or T826 by cyclin D/CDK4 can disable Rb for subsequent binding of an LxCxE protein. In p107 and p130, sites that are specifically phosphorylated by cyclin D/CDK4 (Fig. 2) are required for inactivation of their growth suppressive functions [24,25]. The presence of phosphorylated p130 in quiescent and terminally differentiated cells, where CDK activity is strongly down-regulated, raised the possibility that another (i.e. not CDK) kinase is responsible for this phosphorylation [26,27]. Most of these phosphorylation sites were mapped within the loop region of p130 and glycogen synthase kinase 3 (GSK-3) was implicated in phosphorylation of p130 during quiescence [28] (Fig. 2). Phosphorylation of p130 by GSK-3 contributes to the stability of p130 but does not affect its ability to interact with E2F4 or cyclins. Of note, the loop region of p130 is absent in Rb and has no homology with the corresponding region in p107.

Although comprehensive data have accumulated about the phosphorylation of Rb proteins, our understanding of the processes that counteract their phosphorylation is still limited. The role of Ser/Thr phosphatases in regulating the activity of Rb proteins during the physiological cell cycle is not understood equally well for all members of the Rb family. Similarly, the induced dephosphorylation of Rb family proteins under different growth inhibitory stress conditions remains poorly understood. The present review summarizes our current understanding of phosphatase-mediated activation of Rb proteins during the cell cycle and under different physiological and pathological conditions.

Protein phosphatase 1 and PP2A: structural and functional centipedes regulating the cell cycle

The major cellular phosphatases that have been characterized as Rb or Rb-like protein phosphatases are the Ser/Thr phosphatases PP1 and PP2A, which together account for more than 90% of protein phosphatase activity in eukaryotes [29]. These phosphatases are represented by a limited number of catalytic subunits (PP1α, PP1β/δ and the splice variants PP1γ1 and PP1γ2 for PP1, and the Cα and Cβ isoforms for PP2A) and typically acquire their specificities by assembling a vast number of structurally different holoenzyme complexes [30,31], each characterized by their particular modes of regulation. It has been suggested that PP1 forms distinct stable complexes with as many as 650 mammalian proteins [32], which function as targeting subunits, substrates and/or inhibitors [31]. PP2A represents a family of four dimeric and > 90 heterotrimeric holoenzymes, whose function and regulation are mainly determined by the particular regulatory B-type subunit that is present within the complex [33,34]. Many additional PP2A-interacting proteins have been described, some of which act as inhibitors, activators, substrates or targeting proteins [35–36a], further contributing to the vast functional diversity of these phosphatases.

Interestingly, both PP1 and PP2A play pleiotropic roles in the mammalian cell cycle, particularly in mitosis [37–39], although the particular holoenzymes involved have not always been identified. Similarly, the mechanisms responsible for the regulation of these phosphatases during the cell cycle are poorly understood.

Recently, however, ARPP-19 and the highly-related ENSA were identified in several multicellular eukaryotes as specific mitotic inhibitors of a particular PP2A holoenzyme encompassing a B55 subunit (Bδ in Xenopus, Twins/B55 in Drosophila and B55α or δ in mammalian cell lines), the activity of which needs to be suppressed to allow mitosis [40–44]. Similarly, specific PP1 inhibitory proteins, such as inhibitor (I)-1 can regulate PP1 activity during mitosis [45]. Although Rb proteins are generally not targets of mitotic kinases, the mitotic inhibition of PP2A and PP1 is remarkably coordinated with the activation of mitotic kinases, including cyclin B/CDK1, Greatwall (Gwl or MASTL kinase) and protein kinase A, because specific phosphorylation of ENSA/ARPP-19 (by CDK1-activated Gwl) and I-1 (by protein kinase A) is mandatory for their phosphatase inhibitory activity [40,41,45]. The reversal of these phosphorylations thus represents an attractive way of activating PP2A and PP1 at the exit of mitosis.

The second important cell cycle-related PP2A regulatory mechanism is reversible phosphorylation of the C and/or regulatory B-type subunits [44,46], which either affect holoenzyme assembly or catalytic activity. Similarly, PP1 activity is regulated during the cell cycle by inhibitory T320 phosphorylation of the C subunit [47,48], a site that is conserved in all PP1 isoforms. PP1α is phosphorylated and inhibited during the M-phase and again in the late G1- through early S-phase by CDK1 [49] and CDK2 [47] kinases. Reactivation during mitotic exit is assumed to occur through autodephosphorylation [45]. Thus, cell cycle-related modulation of PP1/PP2A activity may be important for the cell cycle-dependent activation of Rb proteins.

Dephosphorylation of Rb

At mitotic exit, PP1 is a major phosphatase responsible for Rb dephosphorylation

Early studies identified a role for PP1 in cell cycle-related Rb activation

PP1 and PP2A were initially both implicated in Rb dephosphorylation, although an overwhelming amount of data have subsequently identified PP1 as the major Rb cell cycle-related phosphatase. Pulse chase experiments revealed that the changes in Rb electrophoretic mobility occurring near mitosis are a result of enzymatic dephosphorylation [50]. Two groups almost simultaneously reported the first evidence that PP1 and/or PP2A were directly responsible for this dephosphorylation [51,52]. Alberts et al. [51] used microinjection of purified PP1C or PP2A (either catalytic C subunit or AC core dimer PP2AD) into either the nucleus or the cytoplasm of cells synchronized in G1 [51]. The dephosphorylation of Rb was monitored by its resistance to extraction at the G1/S transition. Although microinjection of PP1C in either the nucleus or cytoplasm increased the amount of dephosphorylated Rb, PP2AC (and not PP2AD) generated this effect only if injected in the nucleus. However, all phosphatases (PP1C, PP2AC and PP2AD) were able to dephosphorylate CDK1-phosphorylated Rb in vitro. Ludlow et al. [52] presented data where Rb-specific mitotic phosphatase activity was higher in anaphase than in pro- or metaphase extracts. This precise timing of Rb dephosphorylation during the cell cycle prompted many researchers to look for Rb-directed phosphatase activity during mitotic exit, using protein extracts of cells released from a nocodazole block. These late mitotic cell extracts were capable of dephosphorylating Rb in vitro. Okadaic acid (OA) and protein phosphatase inhibitors I-1 and I-2 inhibited this specific Rb-directed phosphatase activity, implicating PP1 but not PP2A in the cell cycle-related Rb activation [52]. Accordingly, a peak of PP1 activity detected in early G1 could serve to maintain Rb in the dephosphorylated state [53] and the constitutively active PP1 mutant (T320A) caused cell cycle arrest in G1 in an Rb-dependent way [47].

Different roles for PP1 and PP2A in the regulation of Rb were suggested based on selective inhibition of cell cycle-related events by different phosphatase inhibitors [54]. First, treatment with tautomycin, a selective PP1 inhibitor, maintained Rb in a hyperphosphorylated state. Second, inhibition of PP2A by PP2A-specific concentrations of either OA or calyculin A (CalA) resulted in a time- and dose-dependent decrease in Rb phosphorylation, which was correlated with inhibition of CDK2 and CDK4 because the levels of CDK2 and CDK4-bound cyclins were down-regulated, and the expression of CDK inhibitors p21 and p27 went up. These results suggest that PP1 is associated with the direct dephosphorylation of Rb, whereas PP2A is positively involved in pathways regulating CDK activity [54] (Fig. 3).

Figure 3.

 Summary of Ser/Thr phosphatases involved in pocket protein dephosphorylation during cell cycle progression and under stress conditions. PP1 and PP2A regulatory subunits implicated in dephosphorylation of specific pocket proteins are indicated.  1PR48 is an N-terminally truncated form of PR70.

Direct interaction between Rb and PP1

Further support for a major role for PP1 in Rb dephosphorylation came from yeast two-hybrid studies, which, for the first time, reported a direct association between Rb and the catalytic subunit of PP1 [55]. This interaction occurs through the C-terminal region of Rb (amino acids 771 and 928) (Fig. 2) and does not involve either of the PP1 LxSxE motifs [56,57], although a fragment encompassing these motifs (amino acids 159–212 of PP1δ) was shown to still interact with Rb [57]. In addition, strong interaction with Rb was observed with a 65-amino-acid PP1δ fragment (amino acids 195–260) encompassing several residues involved in substrate binding and catalysis [57]. These studies suggested that the interaction between PP1 and Rb may be rather complex, potentially involving more than one interaction interface. In vitro binding assays demonstrated that PP1α preferentially binds the hypophosphorylated form of Rb and this association occurs from mitosis to early G1 [55,58]. However, in more sensitive binding assays using 32P-labelled Rb, the interaction with PP1 could be demonstrated with both hypo- and hyperphosphorylated forms of Rb [59]. In accordance, mutant Rb with 14 out of 16 Ser/Thr substitutions to Ala was able to bind PP1 with a similar affinity to that of wild-type Rb [60], suggesting that the phospho-acceptor sites on Rb do not regulate the interaction with PP1. The question about possible selectivity of different PP1 isoforms toward Rb was addressed using in vitro phosphatase assays and isoform-specific antibodies. Although, initially, it was suggested that the PP1β/δ isoform had higher Rb-directed phosphatase activity throughout mitosis [61,62], this preferential activity does not persist in G1 [63]. Subsequent experiments either in HeLa cells [60] or murine fetal lung epithelial cells [64] showed that Rb actually binds well to all PP1 isoforms.

The recently obtained crystal structure of the α-isoform of PP1 with the Rb peptide (amino acids 870–882) clarified many of these data [65]. First, it provided an explanation why all isoforms of PP1 are able to bind Rb because the PP1Cβ-sandwich structure and the specific side chains that contact the Rb peptide are conserved in all three mammalian PP1 isoforms. Second, it identified an enzyme docking site in the Rb C-terminus that is required for efficient PP1 activity toward Rb. This site overlaps with the known binding site for CDKs and it was suggested that PP1 competition with CDKs is sufficient to retain Rb activity and block cell cycle progression [65]. Third, the existing structure explained why substitution of 14 Rb phosphorylation sites to Ala did not change the affinity of the protein to PP1 [60]. Fourth, the Rb peptide contacts PP1 at a site that is distinct from the phosphatase active site, which explains the observation that catalytic activity of PP1 is not required for Rb–PP1 association [59]. Fifth, most PP1 interacting regulatory subunits and inhibitors contain a consensus RVxF sequence, which directly binds PP1 [31]. The structural similarities of the contacts between Rb and PP1 versus those between MYPT1 (myosin phosphatase targeting subunit 1) and PP1 indicated that the KLRF sequence at residues 874–877 of Rb likely functions as an RVxF motif, placing Rb in the group of PP1 interacting proteins with a consensus RVxF motif. Although p107 and p130 have RxL sequences in the region between two pockets that are critical for binding to cyclin A/CDK2 and cyclin E/CDK2 complexes (Fig. 1A), the phenylalanine residue in both the p107 and p130 docking motifs directly follows the leucine (RRLF), explaining why these proteins do not bind PP1 in a manner similar to Rb. According to the crystal structure, leucine forms critical contacts with PP1 when in the −2 position relative to the phenylalanine residue. Thus, the presented data indicate that PP1 directly binds Rb through a KLRF motif independently of its phosphatase activity [65].

Site-specificity of Rb dephosphorylation by PP1

Rb dephosphorylation occurs in a sequential, temporally-regulated manner and different PP1 isoforms may have different preferences for site-specific dephosphorylation [63]. This very much mirrors the sequential phosphorylation of Rb by different CDK/cyclin pairs, each of which shows a certain specificity for particular phosphorylation sites (Fig. 2). As such, many cyclin D/CDK4 sites, which are the first ones to be phosphorylated in mid-G1, are also the first ones to be dephosphorylated at mitotic exit. Interestingly, not all known phosphorylation sites of Rb are entirely dephosphorylated during the cell cycle because T821 phosphorylation was hardly reversed at all in CV1-P cells when released from a nocodazole block [63]. Accordingly, this site was not targeted by any PP1 isoform in vitro. As described below, this particular residue was shown to be rapidly dephosphorylated by PP1 under hypoxic conditions (in the same cell line) [66] and by PP2A under oxidative stress conditions (in endothelial cells) [67]. Similarly, using phospho-specific antibodies, it was shown that Rb with specific phosphorylated sites (S608, S612, S780 and T807) is capable of binding to PP1, whereas other phosphorylated forms (S249, T373, S788, S795, T811, T821 and T826) do not bind PP1 [59]. All these observations are suggestive of site-specific dephosphorylation of Rb, potentially by different PP1 isoforms or distinct PP1 holoenzyme complexes, whereas, in challenged cells, PP2A may also be involved.

The regulatory role of PP1 interacting proteins/subunits

Because Rb was found to be dephosphorylated by a subpopulation of PP1 in mitotic cells [58], the possible role of specific PP1 interactors or regulators of PP1 activity in Rb dephosphorylation was addressed.

First, a high molecular weight form of PP1 was isolated as a major Rb phosphatase from the nuclear fraction of CV-1P cells [62]. Two proteins were found to be associated with mitotic PP1, and the complex encompassing a 110-kDa protein was responsible for Rb-directed phosphatase activity. Although not formally confirmed, this 110-kDa protein may very well be the myosin targeting subunit MYPT (also called M110), which was identified subsequently in an independent study as a subunit targeting PP1C to Rb [68]. In addition, MYPT stimulated PP1 activity towards the phosphorylated C-terminal fragment of Rb, and MYPT and Rb partially co-localized in the nucleus. Interestingly, phosphorylation of MYPT on two inhibitory sites (T695 and T850), induced by CalA treatment, resulted in the inhibition of MYPT-PP1 activity [68]. A mitosis-specific phosphorylation of MYPT on three serine residues (S432, S473 and S601), probably by CDK1, has also been reported [69,70], in this case resulting in increased MYPT-PP1 activity, at least towards myosin as a substrate [69]. These studies would therefore suggest that reversible phosphorylation of MYPT might comprise a mechanism for regulating MYPT-PP1-mediated Rb phosphatase activity. More recently, another (unusual) regulatory mechanism of MYPT-PP1 was described [71]. MYPT was found to be reversibly methylated in vitro and in vivo on K442, and this modification increased MYPT stability, resulting in increased Rb dephosphorylation on S807/S811 [71]. Of note, in addition to a classical RVxF motif, MYPT harbours a second N-terminal PP1 interaction motif, denoted ‘MyPhoNE’ (myosin phosphatase N-terminal element) [32,72], potentially explaining why simultaneous binding of Rb and MYPT to PP1 is not necessarily mutually exclusive.

Second, the regulatory subunit PNUTS (Phosphatase 1 Nuclear Targeting Subunit) was implicated in the regulation of Rb-directed dephosphorylation. Although the PNUTS–PP1 binding remained unaffected throughout the cell cycle [73], PNUTS could not be detected in immune complexes of Rb or vice versa [74]. However, a correlation between PNUTS expression and phosphorylation status of Rb was demonstrated in several cancer cell lines [75]. Several Rb phosphorylation sites (S795, S780, T807/T811) were indeed sensitive to PNUTS expression and were dephosphorylated when the level of PNUTS expression was down-regulated, consistent with inhibition of PP1C activity toward Rb in vitro upon addition of PNUTS [73]. PNUTS knockdown induced apoptosis and unexpectedly resulted in dissociation of hypophosphorylated Rb from E2F1, in line with the observation that E2F1 regulates the expression of caspase proenzymes [76]. Of particular significance was the observation that this mechanism appeared only to occur in cancer and not in normal cells [75]. Because PNUTS levels were comparatively much higher in cancer cells, potentially neutralizing more PP1 in the nucleus, the effects of inhibiting PNUTS in the cancer cells might thus result in higher levels of active PP1 being available to target Rb [77].

A third insight into Rb-directed PP1-specific activity during embryonic cell cycle regulation came from an elegant study describing the humpty dumpty (humdy) mouse mutant with failure to close the neural tube and optic fissure [78]. The missense mutation that affected neural tube closure was identified within the humdy gene that encodes Phosphatase and actin regulator 4 (Phactr4). Phactr4 is a member of a family identified as PP1α-interacting proteins [79] and named after their specific PP1- and actin-binding domains. Although Phactr family proteins lack a consensus RVxF PP1-binding motif, all four members show strong conservation of their PP1- and actin-binding domains. The humdy mutation was mapped into the PP1 binding region and disrupted PP1 binding, subsequently resulting in increased T320 phosphorylation of PP1, a change in PP1 localization and eventually disruption of PP1 activity towards Rb. Humdy embryos display indeed elevated proliferation, Rb hyperphosphorylation and derepression of E2F targets. It was suggested that, normally, Phactr4 binds to PP1 and retains it in an active state because PP1 does not undergo inhibitory phosphorylation within this complex. Thus, during the cell cycle, active PP1 can shuttle to the nucleus, where it can dephosphorylate Rb. Because Phactr4 is differentially expressed during neurulation and eye development, it provides PP1 with Rb-specificity during embryonic development.

Thus, despite the fact that PP1C and Rb directly interact with strong affinity without the need for a targeting subunit, particular PP1 binding partners might still be required to regulate this Rb-specific PP1 activity, both positively and negatively, and this might be strongly dependent upon the cellular or tissue context. More detailed studies would be required to clarify this.

The role of PP2A in Rb dephosphorylation during the cell cycle

A genetic study in Drosophila melanogaster revealed normal regulation of Rb/E2F1 target genes in mutant flies in which all PP1C encoding genes were inactivated, suggesting that, in this organism, either another phosphatase is involved in the dephosphorylation of Rb or that it acts redundantly with PP1 [80]. Indeed, it was observed that when CDK activity was inhibited by addition of roscovitine, a potent and selective inhibitor of CDK1, CDK2 and CDK5, Rb dephosphorylation was rapidly induced in Jurkat cells, and this was blocked by CalA or OA [81]. Thus, it was concluded that Rb phosphorylation might in part be determined by a balance between CDK kinases and PP1/PP2A. A very similar rapid Rb phosphatase activity was unmasked upon inhibition of CDK activity by flavopiridol or upon inhibition of protein synthesis by cycloheximide (CHX) (which rapidly affected cyclin D expression and thus indirectly inhibited CDK4/cyclin D activity). In this case, dephosphorylation was delayed by ectopic expression of simian virus 40 (SV40) small t antigen, identifying the phosphatase involved as a trimeric form of PP2A [82]. Thus, it was suggested that a dynamic equilibrium between CDKs and PP2A modulates phosphorylation of Rb throughout the cell cycle: in the absence of CDK activity, the balance tilts in favour of PP2A, whereas, as long as CDKs are active, PP2A activity is overtaken. Such a model would not contradict a specific role for PP1 in Rb dephosphorylation during late mitosis and early G1, particularly because CHX-induced dephosphorylation Rb was not entirely complete [82]. Alternatively, the residual cyclin E or cyclin A/CDK2 activity might be sufficient to continue phosphorylating Rb when cyclin D/CDK4 activity has declined. In any case, this model challenges an earlier proposed model [83], according to which cell cycle-dependent Rb regulation would occur by a simultaneous ‘switch on’ of CDK and ‘switch off’ of PP1 activities in the mid–late G1-phase (resulting in Rb hyperphosphorylation), followed at an undefined point in mitosis by a ‘switch on’ of PP1 and a ‘switch off’ of CDKs (resulting in Rb dephosphorylation). Another interesting implication of the novel model would be that although dephosphorylation of Rb after the restriction point is unlikely to still affect cell cycle progression (because E2F-dependent gene expression has already generated sufficient amounts of cell cycle regulators), other functions of pocket proteins, such as their ability to regulate transcription of ribosomal and t-RNA1 [82,84], could possibly be quickly restored upon inactivation of CDKs by drugs or certain stresses [82]. The identity of the PP2A holoenzyme(s) responsible for maintaining this CDK-PP2A equilibrium towards Rb remains unclear, although a specific interaction between GST-Rb and the PP2A PR70/B”β subunit (and not the B55α subunit) has been demonstrated [85]. Thus, PP2A might functionally complement Rb-specific activity of PP1 during cell cycle progression when CDK activities are jeopardized.

Cell-cycle independent, rapid Rb dephosphorylation

Rapid dephosphorylation of pocket proteins and resulting cell cycle arrest in response to different growth inhibitory factors is well-documented in the literature. Thus, the ability of phosphatases to counteract the activity of CDKs should be responsive to stress-related changes in the intra- or extracellular environment and should be readily available during any phase of the cell cycle, with the aim of stopping cell division when circumstances are inappropriate and genomic stability is jeopardized.

Hypoxia

Mild hypoxia (1% oxygen) resulted in a 50% increase in PP1 activity toward Rb, whereas phosphorylase a activity was unaffected by hypoxic conditions [86]. Although this was not the result of increased PP1C expression [86], it was shown that PNUTS failed to associate with Rb under those conditions [73]. Consequently, T821, an Rb site that is not significantly dephosphorylated during normal cell cycling, now became dephosphorylated under hypoxic conditions [66], eventually resulting in Rb-dependent cell cycle arrest.

Oxidative stress

Rb is rapidly (within 30 min) dephosphorylated in response to oxidative stress in human umbilical vein endothelial cells in a dose-dependent manner at at least three sites (T356, T821 and T826) [67]. This was not associated with cyclin/CDK down-regulation, indicating that the activation of a phosphatase is responsible for this dephosphorylation. This phosphatase was subsequently unequivocally identified as PP2A because the expression of SV40 small t antigen, but not of its non-PP2A binding mutant, inhibited H2O2-induced Rb dephosphorylation. Accordingly, PP2A A and C subunits were detected in Rb immunoprecipitates, both in untreated and H2O2-treated cells [67]. In a subsequent study, it was demonstrated that PR70/B”β, but not the highly-related PR72/B”α2 PP2A regulatory subunit, can bind Rb in vitro and in vivo, with the latter being assessed in unstimulated cells only. PR70 overexpression, but not overexpression of the non-Rb binding PP2A subunit PR55/Bβ, was sufficient to induce dephosphorylation of ectopically expressed Rb, independently of an oxidative stimulus. In addition, shRNA-mediated knockdown of PR70/B”β precluded H2O2-induced Rb dephosphorylation and inhibition of DNA synthesis [85]. Importantly, it was demonstrated the H2O2-induced increase in intracellular Ca2+ was responsible for PP2A-PR70 activation and Rb dephosphorylation, indicating a prominent role for Ca2+ in the regulation of these particular PP2A holoenzymes. Indeed, two well-conserved canonical EF-hand motifs have been described in this B” family of PP2A subunits, which can regulate both PP2A activity and subcellular localization [87,88]. Interestingly, in plants, it was demonstrated that the interaction between the OsRBR1 and OsPP2A B” proteins (the rice orthologues of Rb and PP2A B” regulatory subunit) in the yeast two-hybrid system requires an intact pocket domain of RBR and the presence of the EF-hand domains in the regulatory B” subunit, further supporting the hypothesis that Rb dephosphorylation might be influenced by increases in intracellular Ca2+ [89].

Apoptotic stimuli: Fas ligand, colcemid and chemical DNA-damaging agents

One of the early studies suggesting an implication of PP1-mediated Rb dephosphorylation in apoptotic processes investigated Fas-induced apoptosis in Jurkat cells, which led to the dephosphorylation of Rb within 5 min [90]. OA and CalA inhibited Fas-induced apoptosis and dephosphorylation of Rb, implicating that Fas-triggered apoptotic death involves the activation of PP1. In another study, Rb immunoprecipitates from cells that accumulated DNA damage as a result of exposure to anti-cancer drugs, contained a phosphatase activity that could dephosphorylate Rb in vitro and was inhibited by CalA or OA [91]. This phosphatase was subsequently identified as PP1 and the resulting dephosphorylation of Rb converted Rb into an efficient substrate of caspase-3, leading to Rb cleavage [92]. Similalrly, PP1 activity was increased in colcemid-induced apoptotic Hela cells, and correlated with increased Rb dephosphorylation [93]. It is important to note that, in the cases described above, the levels of cyclin E- and cyclin A-associated kinase activities remained high during Rb dephosphorylation, supporting the existence of a mechanism activating a specific phosphatase. Exactly how this activation occurs remains enigmatic, with one study reporting a significant increase in absolute PP1α levels after treatment with a DNA damaging agent [92] and another study reporting an activation mechanism the the same as that in hypoxic cells (i.e. dissociation of PNUTS from PP1, followed by concominant dephosphorylation of Rb T821) [66].

Ionizing radiation

Both ionizing and UV radiation cause DNA damage, which would require an immediate response to coordinate DNA repair and interrupt cell cycle progression. Dephosphorylation of pocket proteins is one of the mechanisms for executing the latter event. Data published by Guo et al. [94] supported an ionizing radiation-induced activation of nuclear PP1 by ATM (ataxia telangiectasia mutated) kinase-dependent dephosphorylation of inhibitory T320 in Jurkat cells [94]. One interesting twist came with the fact that, in response to ionizing radiation, despite an ATM-dependent activation of PP1, Rb is temporarily dissociated from PP1 complexes [95]. Although dephosphorylation of specific Rb phospho-residues was not addressed, two different possibilities can be considered to explain the disruption of Rb–PP1 interaction. Either increased PP1 activity caused faster turnover of PP1/substrate complexes, which would interfere with the detection of Rb in PP1 immunoprecipitates, or PP1 ‘handles’ the responsibility to dephosphorylate Rb to other Ser/Thr phosphatases, such as PP2A. The latter has been implicated in UV-induced dephosphorylation of pocket proteins [96].

Dephosphorylation of the pocket proteins p107 and p130

PP2A is the major phosphatase regulating p107 and p130 during the cell cycle

Although p107, but not p130, demonstrated weak binding to PP1α in the yeast two-hybrid system, a putative PP1–p107 interaction was undetectable in co-immunoprecipitation experiments [97] because this interaction is likely weaker or more transient than the Rb–PP1 association. This is in accordance with the structural studies of the Rb–PP1 interaction [65] discussed above. Therefore, PP2A is expected to be a major phosphatase responsible for p107 and p130 dephosphorylation during the cell cycle and in challenged cells. Consistent with a role for PP2A in dephosphorylating p107/p130, the PP2A catalytic subunit was found to associate with both p130 and p107 throughout the cell cycle and in quiescent cells, with a noticeable increase as cells progress through S-phase. This interaction was not affected by CHX treatment despite a dramatic increase in dephosphorylated p107/p130 proteins [82]. Although the rapid CHX-induced dephosphorylation of p107 was complete, consistent with the fact that p107 is primarily phosphorylated by cyclin D/CDK4 complexes, p130 was incompletely dephosphorylated, potentially as a result of residual CDK2/cyclin A or E activity. Similar to Rb, p107/p130 dephosphorylation was significantly delayed by expression of SV40 small t antigen, implying that the associated PP2A is constitutively active and that one or more PP2A B-type subunit(s) is/are responsible for targeting these pocket proteins. Although the evidence is limited, different PP2A holoenzymes may indeed be involved.

First, a strong yeast two-hybrid interaction was observed between p107 and the PR59/B”γ subunit of PP2A, which was confirmed in reciprocal co-immunoprecipitations [98]. This interaction showed strong specificity because PR59/B”γ did not interact with Rb (neither in the yeast two-hybrid system, nor in co-immunoprecipitates), and p107 did not interact with PR72/B”α2. In addition, ectopic co-expression of p107 and PR59/B”γ in U2OS cells resulted in an increase in the amount of hypophosphorylated p107, whereas overexpression of PR59/B”γ did not affect Rb phosphorylation. PR59/B”γ and p107 co-localized in nuclear dots, whereas no co-localization was seen with Rb [98]. Binding of PR59/B”γ to p107 may be distinct from binding of E2F4-5 to p107 because PR59/B”γ association with p107 is more transient in nature than the p107–E2F interaction, despite the presence of an LxSxE motif within PR59/B”γ. Consistent with this, the binding of PR59/B”γ to p107 does not appear to require an intact LxCxE motif-binding ‘B pocket’ [98]. It should be noted that PR59/B”γ is a murine-specific B” isoform absent in human cells, and that its human orthologue is likely PR70/B”β based on evolutionary relationships and sequence similarities [99]. Accordingly, PR48, an N-terminally truncated form of PR70/B”β [100], could still bind to some extent to GST-p107 in vitro [101]. However, whether the PP2A-PR70 holoenzyme can mediate p107 dephosphorylation in vivo remains unknown.

Second, another PP2A subunit, B55α, was identified as the preferred interactor of p107 in vivo, and modulation of B55α expression in U2OS cells induced prominent changes in the phosphorylation of p107 [101]. Consistently, a purified B55α trimeric PP2A holoenzyme could dephosphorylate p107 in vitro. Modulation of B55α affected p130 phosphorylation to a lesser extent, in accordance with the observation that GST-p130 interacted with both B55α and PR48 at comparable levels [101]. As reported previously, GST-Rb associated with PR48 but not B55α [85]. These binding preferences may have a structural basis because B55α directly associates with the spacer region between the two pockets of p107, which is conserved in p130 but is more divergent in Rb (Fig. 2). This finding echoes interaction between Rb and PP1 because, in both cases, the phosphatase binds to the region that interacts with CDKs, suggesting an existing equilibrium/competition between kinase and phosphatase. The fact that B55α directly interacted with p107 independently of PP2A A and C subunits [101] suggests that specific regulatory subunits of PP2A target specific substrates for dephosphorylation [102,103]. In addition, it is well-established that PP2A holoenzymes encompassing a B55 subunit efficiently dephosphorylate SP/TP (CDK) sites both in vitro [104,105] and in vivo [106,107].

Cell cycle-independent, rapid p107/p130 dephosphorylation

UV irradiation

UV irradiation induces rapid (within 1 h) dephosphorylation of p107 and G1 growth arrest [96]. The observed dephosphorylation was shown to be mediated by activation of a phosphatase and was apparent in several unrelated cell lines, suggesting that UV-induced p107 dephosphorylation is of general importance. Based on the finding that p130 dephosphorylation did not correlate directly with cell cycle arrest, and that p130 represented only a minor proportion of pocket proteins in cycling cells, it was suggested that p107 is more directly involved in the initial response to UV irradiation. A similar notion of the particular importance of p107 in mediating fibroblast growth factor (FGF)-induced growth arrest in chondrocytes was also reported [108]. Modulation of PP2A composition by overexpression of PR72/B”α2, a PP2A B-type subunit unable to interact with p107 [98], prevented UV-mediated dephosphorylation of p107, potentially by competing with the p107-interacting B-type subunit for PP2AD binding. Moreover, HA-tagged PP2A C or A subunits, immunoprecipitated from U2OS cells, were able to dephosphorylate p107 in vitro, all being suggestive for a PP2A-dependent dephosphorylation mechanism. Currently, there are no conclusive data about the identity of the specific PP2A holoenzyme involved.

Oxidative stress

Concomitantly with Rb, both of the other pocket proteins, p107 and p130, also become rapidly dephosphorylated by PP2A in response to oxidative stress in HUVEC cells [67]. PP2A A and C subunits were detected in p107 immunoprecipitates, both in untreated and H2O2-treated cells, whereas the data from p130 immunoprecipitates were not conclusive as a result of cross-reactivity of the p130 antiserum with p107. SV40 small t expression prevented H2O2-induced DNA synthesis inhibition, thus over-riding the pocket protein-dependent intra-S checkpoint [67].

FGF-induced growth arrest

PP2A was also implicated in rapid FGF-induced p107 dephosphorylation in chondrocytes [108]. Whereas, in most cell types, FGF induces proliferation and protects from apoptosis, chondrocytes are distinct in their response to FGF. FGF causes growth arrest in chondrocytes and induces rapid dephosphorylation of all pocket proteins. Although p130 and Rb dephosphorylation takes a long period of time, p107 is dephosphorylated within the first hour of FGF treatment. Using chemical and pharmacological inhibition and knockdown of PP2AC, it has been shown that PP2A is responsible for FGF-induced dephosphorylation of p107. Moreover, FGF induced the association of catalytic and scaffolding subunits with p107, as well as the binding of B55α regulatory subunit (V. Kolupaeva & C. Basilico, unpublished results).

All-trans retinoic acid (ATRA) induced p130 dephosphorylation in ovarian carcinoma cells

Treatment of CAOV3 ovarian adenocarcinoma cells with ATRA leads to increased protein stability and the accumulation of p130, and subsequent G0/G1 cell cycle arrest [109]. As reported previously, dephosphorylation of p130 stabilizes the protein, whereas its phosphorylation targets it to degradation [21]. ATRA-induced dephosphorylation of p130 occurs via a PP2A-dependent mechanism, and correlates with increased expression (mRNA and protein levels) of PP2A Cα [109] and decreased tyrosine phosphorylation of PP2AC [110]. PP2AC and p130 could be co-immunoprecipitated [109] and in vitro binding assays demonstrated a direct PP2AD–p130 interaction without the need for any regulatory B-type subunit (a short N-terminal domain of PP2AC interacted with the C-terminal domain of p130) [110,111]. PP2A specifically dephosphorylates two residues (S1080 and T1097) adjacent to the nuclear localization signal (NLS) of p130, and this results in increased binding of the NLS to importins, thereby facilitating p130 transport to the nucleus [112].

These data represent interesting observations but the functional importance of PP2A-mediated p130 dephosphorylation is still unclear. Although ATRA treatment is important for PP2A-induced p130 dephosphorylation in CAOV3 cells, it has been shown that a loss or decrease of p130 expression is a frequent event in ovarian carcinoma [13,113]. This would restrict ATRA-induced p130 dephosphorylation to very few cases.

Conclusions and perspectives

The accumulated experimental data clearly support the role of PP1 in Rb dephosphorylation during the cell cycle at mitotic exit. PP1–Rb structural aspects have been clarified and several possible mechanisms of modulating Rb-directed phosphatase activity upon exit from mitosis have been suggested. All the data related to the modulation of Rb targeted PP1 activity clearly emphasize the role of PP1 as a main ‘sensor’ of cell cycle timing, leaving the passive role of a substrate to Rb. Not only is the inhibitory T320 phosphorylation of PP1 cell cycle-dependent, but also particular PP1 binding partners might further regulate its Rb-specific activity, both positively and negatively, and potentially in a cell cycle-dependent manner. Included among these PP1 interacting proteins is MYPT, whose reversible phosphorylation might comprise a mechanism for regulating MYPT–PP1-mediated Rb dephosphorylation. PNUTS is another PP1 regulatory subunit that can restrict/inhibit Rb-directed PP1 activity, although the major effects were solely observed in cancer cells. One additional regulatory subunit shown to be important for Rb-specific PP1 activity is Phactr4, whose binding to PP1 protects the phosphatase from inhibitory T320 phosphorylation and therefore maintains a pool of active PP1 ready to execute Rb-related functions.

Whereas PP1-mediated Rb dephosphorylation is mainly restricted to mitotic exit, it was proposed that PP2A is important for maintaining the phosphorylation status of all pocket proteins during cell cycle progression by counteracting CDKs. Such a model of an existing dynamic equilibrium between a given phosphatase and CDKs during the physiological cell cycle [82] becomes particularly appealing when taking into account the fact that the identified binding sites for PP1/PP2A and CDKs on Rb/p107 are overlapping. In such a set-up, the equilibrium can be easily tilted toward either phosphatase or kinase and this would allow rapid tuning of pocket protein phosphorylation according to a given cell cycle context. The association between PP2A and pocket proteins during cell cycle progression is likely mediated by different PP2A holoenzymes: it was shown that the B55α regulatory subunit of PP2A can bind p107 and p130, whereas p130 and Rb have affinity for the PR70/B”β subunit. The relevance of the reported interaction between the murine-specific PR59/B”γ subunit and p107 currently remains unclear. The observed interactions both with PP1 and PP2A usually exist throughout the cell cycle, which might indicate that activation of a phosphatase occurs within a context of pocket protein–phosphatase complexes.

It is still unclear whether PP1 plays a role in dephosphorylation of p107 and p130 during the transition through mitosis. Attempts to answer this question should consider several factors that would add an extra level of complexity to the experimental design. First, levels of p107 and p130 expression vary significantly during the cell cycle. It was shown that hyperphosphorylated p130 is rapidly targeted for degradation; therefore, there is a possibility that degradation can ‘substitute’ for dephosphorylation, although this remains to be confirmed. Second, both p107 and p130 are important for translocation of their partners E2F4 and E2F5 because these E2F family members do not possess intrinsic NLSs. Therefore the mitotic dephosphorylation of p107 and p130 is likely to take place in the cytoplasm, facilitating the formation and translocation of repressive complexes. Thus, cytoplasmic and nuclear fractionation of cell lysates at certain fixed time points of the cell cycle would be required to precisely understand the mechanism of p107/p130 dephosphorylation through mitotic progression. Third, because very few phospho-specific antibodies are available for p107 and p130, it is still very difficult to assign specific roles to certain phosphorylation sites during cell cycle progression or under stress-related conditions. It also remains unknown which phosphatase would counteract GSK-3β mediated phosphorylation of the S/T residues in the loop region of p130 that are important for stability of the protein.

The activation of phosphatases under different stress conditions often occurs in the presence of active CDK complexes. Several PP2A holoenzymes have been implicated in the rapid dephosphorylation of Rb proteins, as well as PP1 (although the latter is less frequent) (Fig.3). Therefore, the question of how a given phosphatase is activated in response to particular stress conditions [67,108] remains unanswered. Although several post-translational regulatory modifications have been described for regulatory subunits and PP2AC itself [30,114], including the regulation of particular B-type subunits by secondary messengers such as Ca2+ [87,88], the mechanism of rapid PP2A activation remains obscure, with several possible scenarios. First, the particular activation mechanism of a regulatory subunit might promote binding of PP2A to the substrate, or increase the catalytic activity of pre-existing enzyme–substrate complexes. Second, induced changes within the core PP2A dimer or PP2AC might reorganize holoenzyme assembly and affect its activity toward a specific substrate. Third, an inhibitor of Rb-specific PP2A activity might be released under certain stress conditions. The ability of Rb themselves to ‘sense’ a signal and adopt a ‘phosphatase friendly’ conformation appears unlikely because, in most cases, it was shown that pocket protein–phosphatase complexes can be formed with differentially phosphorylated or nonphosphorylated Rb proteins.

The molecular mechanisms of pocket protein dephosphorylation also remain to be clarified. It is unclear whether there is differential dephosphorylation of the many known phosphorylation sites of Rb proteins by PP2A or PP1. Another important question is whether the same phosphatase complex can perform several differential enzymatic reactions or whether dephosphorylation would rather be an ‘in-and-out’ mechanism because several phosphorylations should be reversed either by PP2A or by PP1. One of the important steps for understanding these mechanisms would be the identification of all of the PP2A or PP1 regulatory subunits involved in specific Rb dephosphorylation. This becomes feasible now when many of these regulatory subunits have been described and better/more specific tools (e.g. small interfering RNA or short hairpin RNA technologies) are available to tackle these questions. Such targeted strategies offer significant advantages over the still widely used chemical inhibitors of PP1 or PP2A, which, even when applied at concentrations that are assumed to selectively target one or the other phosphatase, undoubtedly have their limitations and may easily result in misinterpretation of the data [115].

A PP2A dimer, comprising just the scaffolding and catalytic subunit, was shown to be responsible for p130 dephosphorylation in CAOV3 ovarian carcinoma cells in only one case [109]. It was argued that the absence of any regulatory subunit to mediate interaction of p130 and PP2A would be in line with the fast kinetics of p130 dephosphorylation, although this is unlikely for two reasons. First, the timing of rapid Rb protein dephosphorylation upon UV irradiation or oxidative stress [85,96] is comparable with the observed kinetics of p130 dephosphorylation and different regulatory B-type subunits of PP2A have been implicated in dephosphorylation of Rb proteins upon UV irradiation and oxidative stress. Second, as noted previously, the presence of a regulatory subunit might be necessary to dictate PP2A substrate specificity and to avoid unregulated/nonspecific phosphatase activity, as might be expected from the dimer and as observed in vitro [35].

Finally, although the accumulated data clearly address the phosphorylation status of Rb and, to some extent, that of p107/p130, the functional importance of particular dephosphorylated forms of the pocket proteins should be addressed under different conditions. Such experiments would be required to clarify whether dephosphorylation indeed leads to the association of pocket proteins with their respective E2F factors and whether these complexes can fulfill their function as transcriptional repressors.

In conclusion, although the activity of CDKs toward Rb proteins, either during the normal cell cycle or under different stress conditions, has so far mainly been considered and studied extensively, it is clear that the action of Rb protein phosphatases (although often underestimated) is no less important, and fluctuations in PP1/PP2A activity also play an important and integral role in modulating the phosphorylation status of pocket proteins. In addition, it is clear that, although not always understood up to the molecular level, specific PP1/PP2A regulatory subunits or binding partners mediate Rb phosphatase specificity and regulation by internal or external stimuli. We therefore expect that additional putative and fascinating regulatory mechanisms of Rb-directed phosphatase activity may be discovered in the near future.

Acknowledgements

We are grateful to Lisa Dailey for critically reading the manuscript. The laboratory of VJ 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).

Ancillary