Emerging insights into serine/threonine-specific phosphoprotein phosphatase function and selectivity

Post-translational modifications (PTMs) are covalent, mostly reversible, modifications of amino acid residues in proteins (Khoury et al., 2011; Ramazi and Zahiri, 2021). To date, more than 400 different PTMs have been identified, with the majority being applied and removed by enzymes (Khoury et al., 2011; Ramazi and Zahiri, 2021). Reversible protein phosphorylation is a binary PTM that is substantially used for signal transduction (Hunter, 1995). This modification has an effect on virtually all biological functions and changes many properties of proteins, such as their stability, binding, structure, localization and activity (Brautigan and Shenolikar, 2018). It is mediated by the counterplay of protein kinases and protein phosphatases, and can be a highly dynamic process, as demonstrated by some phosphorylated sites (p-sites) having half-lives in the order of seconds (Kleiman et al., 2011; Blazek et al., 2015; Reddy et al., 2016). To understand the nature of reversible phosphorylation, it is important to keep in mind that p-sites are not continuously modified through the kinase–phosphatase interplay. The state of a p-site depends on its required function for the cell, and can be under fairly stable kinase or phosphatase dominance (Nasa and Kettenbach, 2018; Nilsson, 2019; Kauko et al., 2020; Swartz et al., 2021). Owing to the ubiquity and dynamic nature of protein phosphorylation, this type of PTM is well suited for modulating numerous cellular processes, including transcription, DNA damage response, RNA processing, nutrient and stress responses, immunity, growth and many more (Ubersax and Ferrell, 2007; Guccione and Richard, 2019; Yang et al., 2017; Nasa et al., 2020). Consequently, dysregulated (de-)phosphorylation is a hallmark of many diseases including cancer (Zanivan et al., 2013; Vainonen et al., 2021), Alzheimer's disease (Grundke-Iqbal et al., 1986; Eidenmüller et al., 2001; Oliveira et al., 2017), cardiovascular diseases (Nicolaou and Kranias, 2009; Fischer et al., 2018), and diabetes (Meyerovitch et al., 1989; Danielsson et al., 2005). Phosphoserine (pSer) and phosphothreonine (pThr) are among the most frequently identified PTMs (Khoury et al., 2011; Ramazi and Zahiri, 2021). The pSer/pThr-specific phosphoprotein phosphatases (PPPs) dephosphorylate the majority of Ser/Thr p-sites (Li et al., 2013; Damle and Köhn, 2019). Although some of the responsible kinases are understood well enough to serve as drug targets, an understanding of the counteracting PPPs still lags behind. Nevertheless, phosphatases are by now also considered druggable targets, as recently reviewed elsewhere (Köhn, 2020; Zhang et al., 2021). The cellular and physiological functions of individual PPPs have extensively been discussed in recent reviews (Ohama, 2019; Park and Lee, 2020; Nilsson, 2019; Sager et al., 2020; Sandal et al., 2021). In this Review, we summarize recent findings regarding the mechanism of specificity, newly developed experimental approaches and phosphoproteomic insights that contribute to a better understanding of PPPs. In terms of modulator development, a recent comprehensive review summarizes small-molecule approaches in the context of cancer (Zhang et al., 2021). In addition, different reviews have covered therapeutic targeting of phosphatases (Vainonen et al., 2021; Köhn, 2020). Therefore, here, we focus on the development of non-small molecule modulators, mention new developments for the previously reviewed small-molecule modulators, and highlight the most recent developments in research tool development. As PP3 (also known as calcineurin or PP2B) has extensively been reviewed elsewhere (Creamer, 2020; Chaklader and Rothermel, 2021), it is not included here. Furthermore, we also do not consider PP7 (encoded by PPEF1 and PPEF2) because limited progress in the areas covered here has been made since a previous review was published (Brautigan and Shenolikar, 2018).

Phosphorylation of serine and threonine residues is catalyzed by more than 400 protein Ser/Thr-specific kinases (Manning et al., 2002; Li et al., 2013). The opposing dephosphorylation of the respective pSer/pThr sites is executed by protein Ser/Thr-specific phosphatases. Most of these phosphatases belong to one of three evolutionary unrelated superfamilies: (1) the phosphoprotein phosphatases-like (PPPL) family, (2) the metal-dependent protein phosphatases (PPMs), or (3) the haloacid dehalogenase (HAD) phosphatases (Damle and Köhn, 2019). More than 90% of all pSer/pThr-sites are thought to be dephosphorylated by the PPP family, a subfamily of the PPPLs, which establishes them as the major counteractors of pSer/pThr phosphorylation in eukaryotes (Brautigan and Shenolikar, 2018). This ability to regulate the vast majority of the highly complex pSer/pThr network in the cell is all the more fascinating given that this family consists only of seven phosphatases, namely PP1 (encoded by the three isoforms PPP1CA, PPP1CB and PPP1CC), PP2A (PPP2CA and PPP2CB), PP3 (PPP3CA, PPP3CB and PPP3CC), PP4 (PPP4C), PP5 (PPP5C), PP6 (PPP6C) and PP7 (PPEF1 and PPEF2). (Damle and Köhn, 2019).

Superficially, there is broad similarity between the different PPPs. The core catalytic protein is well conserved between PPPs, and there is a high overall degree of sequence identity across all eukaryotic species (Brautigan and Shenolikar, 2018). Inside their active sites, most PPPs coordinate an Fe²⁺ and a Zn²⁺ ion as essential cofactors through three highly conserved sequence stretches (-GDXHG-, -GDXVDRG- and -GNHE-) (Nasa and Kettenbach, 2018). Furthermore, most of the earliest discovered chemical inhibitors bind the catalytically active sites of several PPPs and act as pan-inhibitors with nanomolar to micromolar affinities (Zhang et al., 2021, 2013). Nevertheless, PPPs differ in the formation of their numerous holoenzymes, which are the functional entities in cells and consist of catalytic (denominated ‘c’ compared to the general holoenzyme, e.g. ‘PP1c’ and ‘PP1’) and a great variety of noncatalytic subunits (examples in Fig. 1A,B) (Brautigan, 2013; Li et al., 2013). PP5 is an exception as it is a multidomain enzyme (Fig. 1C). Regulatory subunits dictate holoenzyme activity by determining its subcellular localization, activity, binding partners, and to a large extent their substrate specificity. PPPs that act as free catalytic subunits, rather than holoenzymes within the cell, have not been reported. The assembly of catalytic subunits and PPP-specific regulatory subunits results in a highly specific biomolecular machinery, where each holoenzyme constitutes an independent signaling entity. This not only raises the complexity of PPP signaling networks and therefore complicates detailed studies (Needham et al., 2019; Fahs et al., 2016), but also leads to a similar number of distinct PPP complexes and Ser/Thr-specific kinases in eukaryotic cells. Despite the importance of holoenzyme formation by regulatory subunits, few mechanisms of PPP control by these regulatory subunits have been investigated in detail.

Technical advancements in recent years have helped establish global high-throughput approaches to better decipher PPP biology. Structural insights coupled with phospho-proteomics, or chemical molecule screens followed by computational analysis have provided novel insights. Emerging insights comprise the validation of intrinsic substrate specificity of catalytic subunits, the short linear motif (SLiM)-based interaction between catalytic and non-catalytic subunits, as well as between the holoenzymes and their interactors (Heroes et al., 2013; Hertz et al., 2016; Ueki et al., 2021; Fowle et al., 2021). The SLIM-mediated interaction is a universal mechanism through which phosphatases and kinases recognize and bind to their interactors as well as substrates (Seok, 2021). In addition, a new chemical proteomics approach that uses active site inhibitor–bait pulldowns followed by mass spectrometry (PIB-MS) read-out enables the specific identification of the PPP interactome in different conditions (Lyons et al., 2018; Nasa et al., 2020). Furthermore, new discoveries and developments of inhibitors or activators for PPPs now make it possible to distinguish certain PPPs in cellular experiments and target them specifically over other PPP members. Recent findings on interactions, mechanisms and methodologies for studying each of the different PPPs are discussed in the following sections.

Attributed to the advances of global approaches in mass spectrometry (MS), the number of identified p-sites of the human phosphoproteome has expanded exponentially over the past two decades. This has enabled scientists to uncover the extraordinary scale and complexity of the phosphoproteome (Khoury et al., 2011; Needham et al., 2019; Ochoa et al., 2019; Ramazi and Zahiri, 2021). These studies have revealed that most proteins are phosphorylated, many on multiple sites (Olsen et al., 2010; Sharma et al., 2014; Needham et al., 2019). Ochoa and colleagues generated the most complete human phosphoproteome synopsis to date. Their analysis of 112 human phosphoproteome datasets found 11,982 proteins bearing 119,809 phosphosites (Ochoa et al., 2019). Analysis of the residue-specific phosphorylation distribution through the dbPTM database unveiled that pTyr makes up only a smaller proportion (∼11%) of the total phosphorylations, whereas most of the phosphorylations are found on pThr (∼22%) and pSer (∼67%) residues (Ramazi and Zahiri, 2021). Phosphorylations on other residues exist but are scarce (Ramazi and Zahiri, 2021). This plethora of Ser and Thr phosphorylations highlights the importance of spatio-temporal regulation of PPPs for these modifications (Cohen, 2002; Khoury et al., 2011; Nilsson, 2019).

Although phosphorylation has been repeatedly found to be essential for proper function of the cell, more than 97% of identified human p-sites have no assigned function (Needham et al., 2019; Ramazi and Zahiri, 2021). Usually, substantial effort, such as thorough low-throughput experiments, is needed to uncover the individual p-site function. Thus, although we can ‘quantify’ the phosphoproteome, we are still far from understanding it. One approach to rectifying this involves classifying p-sites according to their position in an ordered region (OR) or intrinsically disordered region (IDR) of a protein. However, the p-sites of IDRs are even less characterized than the ordered sites based on follow-up experiments, because their intrinsic flexibility results in less structural information. Recent results show that ∼40% of the eukaryotic proteome consists of IDRs, and that these segments contain phosphorylation signaling hubs (Wright and Dyson, 2015), suggesting that conformational heterogeneity might be important to kinase as well as phosphatase selectivity. Furthermore, it has been observed that the phosphorylation state of IDRs correlates positively with IDR length (Fukuchi et al., 2011; Cho et al., 2020; Koike et al., 2020). Interestingly, initial studies have suggested that p-sites within ORs more often regulate enzyme activity and conformation, whereas IDRs more often contain p-sites that regulate the localization of a protein (Ochoa et al., 2019; Zarin et al., 2021).

Great advancements in recent years have led to the development of computational methods that have helped to predict ORs and IDRs within protein structures (Jumper et al., 2021; Chakravarty and Porter, 2022) and gained a deeper readout of the human phosphoproteome. However, most functions of the p-sites remain to be elucidated, and predicting and assigning them is still challenging (Kokot et al., 2022; Chaudhari et al., 2021). To change this, a detailed understanding of PPP biology and the development of global PPP approaches are essential. The first recently established global approaches to determine p-site assignment, regulation and function are listed in Table 1.

PP1 is ubiquitously expressed and is considered one of the major PPPs because it counteracts more than 100 kinases (Li et al., 2013). The catalytic subunit (PP1c) forms transient holoenzymes with one or two regulatory proteins to fulfill this comprehensive role. These regulatory proteins (>200) are known as regulatory interactors of protein phosphatase one (RIPPOs) (Bollen et al., 2010; Wu et al., 2018). The formation of these transient holoenzymes coordinates the spatiotemporal activity of PP1 and its substrate specificity (Bollen et al., 2010). Although the regulatory function of RIPPOs is undisputed, it has been speculated that PP1c itself contributes to substrate specificity of PP1 (Ragusa et al., 2010; Rogers et al., 2016). PP1c has three substrate-binding grooves – one acidic, one hydrophobic and one C-terminal groove that radiate in a Y-shaped pattern from the active site and supposedly interact with substrates for dephosphorylation (Fig. 2A) (Peti and Page, 2015; Ragusa et al., 2010; Choy et al., 2014; Peti et al., 2013). Notably, when bound to substrate-targeting RIPPOs, the active site of PP1c remains largely accessible for the substrate (Fig. 1A) (Fedoryshchak et al., 2020; Chen et al., 2015; Kwon et al., 2022). This is in contrast to PP2A–B-subunit holoenzymes where the B-subunit is in direct proximity to the active site (Xu et al., 2006, 2008). A recent crystal structure of the Phactr1–PP1c holophosphatase shows a remodeling of the hydrophobic groove, which is engaged by the short PP1 substrate peptide from the insulin receptor tyrosine kinase substrate of 53 kDa (IRSp53; also known as BAIAP2) (Fig. 2D), leaving the acidic groove available for electrostatic interactions (Fedoryshchak et al., 2020). More broadly, we recently demonstrated the intrinsic substrate specificity of PP1c by combining a phosphopeptide library dephosphorylation followed by mass spectrometry approach (PLDMS) with phosphoproteomics, which offered a range of substrate candidates that can be dephosphorylated by PP1c (Hoermann and Köhn, 2021; Hoermann et al., 2020). We validated an earlier observed preference of basophilic residues N-terminal of the p-site (Lee et al., 1999; Li et al., 2013) and reported a faster dephosphorylation of pThr compared to pSer (Hoermann et al., 2020). Our follow-up analysis demonstrated that the preference of the basophilic sequence context surrounding the p-site is enhanced when located within IDRs, which lack features for structural recognition (Fig. 2B) (Hoermann and Köhn, 2021). Thus, in IDRs, the amino acid sequence context gains importance for site targeting. Interestingly, in the case of the structured PP1 substrate translation initiation factor 2α (eIF2α), which was resolved by CryoEM together with the PP1c–PPP1R15A–G-actin complex, the substrate recruitment is mediated by PPP1R15A–G-actin (Fig. 2D), and PPP1R15A occupies the C-terminal groove of PP1c, whereas eIF2α comes close to the active pocket and does not occupy the grooves (Yan et al., 2021). Taken together, these results suggest an intrinsic catalytic subunit-derived specificity of p-sites within IDRs, whereas recognition of sites in ORs might be more RIPPO driven.

Addressing the challenge of assigning a specific holoenzyme to a substrate, Wu et al. developed a PP1c-holoenzyme-fusion based substrate trapping approach with reduced PP1 activity (Wu et al., 2018). Together with the substrate candidates from the phosphoproteomics approach (Hoermann et al., 2020), these developments offer novel possibilities to decipher the PP1 holoenzyme substratome.

Many interactions between PP1c and RIPPOs are governed by the surface features of PP1c, which are distributed as specific sites almost over its entire surface and distinguish PP1c from other PPP-type phosphatases (Fig. 2C) (Verbinnen et al., 2017; Bollen et al., 2010; Brautigan and Shenolikar, 2018). It is understood that most RIPPOs use intrinsically disordered PP1-binding domains with one or more SLiMs to dock to these particular sites on the surface of PP1c. The most common PP1-binding SLiMs are variants of the RVxF, SILK, myosine phosphatase N-terminal element (MyPhoNE) and ϕϕ (two consecutive hydrophobic residues) motifs (Terrak et al., 2004; Peti and Page, 2015; Bollen et al., 2010). A remarkable amount of RIPPOs (∼150) carry the RVxF motif (Meiselbach et al., 2006), which binds to two hydrophobic pockets on the opposite side to the active site of PP1c (Wakula et al., 2003; Egloff et al., 1997). The PP1-specific RVxF binding motif was ideal for developing chemical modulators of PP1, enabling the creation of selective PP1 modulators, so called PP1-disrupting peptides (PDPs; see Box 1) (Chatterjee et al., 2012). So far, no cleanly PP1-specific chemical inhibitor has been reported, as all also inhibit other PPPs at the higher concentrations that are commonly used in cellular applications. However, the small-molecule inhibitor Tautomycetin is fairly specific (Choy et al., 2017), and the overexpression of endogenous inhibitors of PP1, such as nuclear inhibitor of PP1 (NIPP1; also known as PPP1R8) or inhibitor-1–3 (I1, I2 and I3; also known as PPP1R1, PPP1R2 and PPP1R11) (Fig. 1A), which use SLiMs in PP1 as well, are often used to inhibit the catalytic subunit of PP1 over other PPPs in cellular experiments (Krause et al., 2018; Winkler et al., 2015).

Not all RIPPOs use SLiMs to bind to PP1. For example, to dephosphorylate RAF in the mitogen-activated protein kinase (MAPK) pathway, SHOC2 binds PP1c and MRAS through the concave surface of its leucine-rich repeat region, and only contains a cryptic RVxF motif (Fig. 1A) (Kwon et al., 2022; Hauseman et al., 2022). Similarly, the suppressor-of-Dis2-number 2 (SDS22; also known as PPP1R7) has no known SLiM and associates with PP1c through structured domains, such as helical leucine repeats, which interact with a broad surface of PP1 (Ceulemans et al., 2002; Heroes et al., 2019). SDS22 and I3 form a ternary complex with PP1, and are widely expressed and conserved RIPPOs (Cao et al., 2021; Lesage et al., 2007; Choy et al., 2019). They control multiple steps of the life cycle of PP1. Indeed, they contribute to (1) the stabilization and activation of newly translated PP1, (2) the translocation of PP1 to the nucleus, and (3) the storage of PP1 as a reserve for holoenzyme assembly (Cao et al., 2021). Preliminary evidence suggests that SDS22 and I3 might also function as scavengers of released or aged PP1 for re-use in holoenzyme assembly or proteolytic degradation, respectively (reviewed in Cao et al., 2021).

Phosphoprotein phosphatase 2A (PP2A) is the second major PPP alongside PP1. It is ubiquitously expressed and contributes up to 1% of the total cellular protein in mammalian cells (Ruediger et al., 1991). Canonically, it acts as a heterotrimeric holoenzyme consisting of the catalytic (PP2Ac) and a scaffolding (A) subunit, which form the dimeric, relatively invariant core enzyme. The core enzyme assembles with an exchangeable regulatory (B) subunit for functional phosphatase activity. So far, two multimeric non-canonical PP2A holoenzymes named STRIPAK and INTAC are described (Kück et al., 2019; Jeong et al., 2021; Zheng et al., 2020). Surprisingly, in the INTAC holoenzyme, Integrator8 (Int8; also known as INTS8) binds to the A-subunit in a B-subunit typical manner (Zheng et al., 2020). This is the first case of a PP2A holoenzyme without one of the established regulatory proteins, and the potential of integrators as a new regulatory subunit family need to be further investigated. The assembly of the variable and interchangeable B-subunit to the core enzyme results in a multitude of distinct PP2A holoenzymes (Cho and Xu, 2007). The B-subunits are subdivided into four major families – B, B′, B″ and B‴, with each comprising two to five isoforms (Fig. 3B) (Virshup and Shenolikar, 2009; Wlodarchak and Xing, 2016). They are key determinants of PP2A-substrate specificity due to their interaction with substrates. In contrast to the A- and C-subunit, isoforms of the distinct B-subunits vary strongly in their structure and function (Baharians and Schönthal, 1998; Xu et al., 2006; Slupe et al., 2011). Additionally, their expression levels alter throughout development, tissue and according to cellular requirement, whereas the dimeric core enzyme tends to be consistently and ubiquitously expressed (Ruediger et al., 1991; Seshacharyulu et al., 2013). In contrast to PP1, the structure of the PP2A holoenzyme is predominantly determined by the A and B-subunits, leaving less surface area of PP2Ac available for other interactions (Figs 1, 3A).

The limited number of assigned p-sites for a certain PP2A holoenzyme narrows our understanding of the mechanisms of holoenzyme-derived site-specific dephosphorylation. Recently, the first PP2A–B56 specific consensus binding motif, LxxIxE, was identified (Hertz et al., 2016). Structural analysis revealed that the LxxIxE SLiM engages with a well-conserved binding pocket, which is composed of a hydrophobic and a basic patch on opposite ends, and is present on all B56 isoforms (Hertz et al., 2016). Follow-up approaches exposed the requirement of a basic stretch adjacent to the LxxIxE motif on B56 interactors to enhance PP2A–B56 binding (Wang et al., 2020). After determining potential B56-binding peptide sequences using phage display libraries, more than 100 potential B56-binding partners were reported (Wu et al., 2017). Interestingly, Cip2A, a oncogenic inhibitor of PP2A, was found among them (Junttila et al., 2007; Laine et al., 2021). The idea of the LxxIxE motif as a PP2A–B56-specific inhibitor was successfully applied in a recent approach by overexpressing the SLiM as a peptide modulator (Kruse et al., 2020). Using this approach, the authors observed that PP2A–B56 assembly does not only serve substrate recruitment, but B56 binding to the catalytic subunit directly affects the sequence preference of the dephosphorylated p-site, revealing a selectivity for basophilic charged residues N-terminal of the p-site and evasion of proline-directed (P at +1) motifs (Fig. 3C; Kruse et al., 2020).

The avoidance of proline-directed motifs was also observed simultaneously in our study of the sole catalytic subunit (PP2Ac) at both the peptide and protein level (Hoermann et al., 2020). However, the basophilic selectivity of PP2A–B56 (Kruse et al., 2020) disagrees with the acidic-residue preference we noticed for PP2Ac (Hoermann et al., 2020). Furthermore, no selectivity of pThr over pSer was observed for PP2A–B56, which was detected for PP2Ac and the PP2A–B55 complex (Hoermann et al., 2020; Kruse et al., 2020). Interestingly, proline-directed motifs are targeted by PP2A–B55, but PP2A–B55 does not show a basophilic preference, resulting in significantly different p-sites being dephosphorylated by PP2A–B56 and PP2A–B55 (Fig. 3C; Kruse et al., 2020). These results show that the regulatory subunit of PP2A directly determines the p-site specificity of the holoenzyme, which results in unique patterns of kinase opposition (Kruse et al., 2020; Cundell et al., 2016; Jong et al., 2020). These findings go hand in hand with the observations in budding yeast from the Uhlmann laboratory (Godfrey et al., 2017; Touati et al., 2019). They observed that PP2A–Cdc55 (the PP2A–B55 homolog) counteracts proline-directed motifs and threonine phosphorylation while PP2A–Trs1 (the PP2A–B56 homolog) opposes basophilic p-site motifs. It is interesting to note that the respective subunits appear to structurally modify the active site and its selectivity, which raises the question as to what extent the mere catalytic subunit should be used for further selectivity experiments.

It has been known for some time that the acidic surface of PP2A–B55 binds to clusters of basic residues surrounding the phosphorylation site (Cundell et al., 2016). More recently, a binding motif for PP2A–B55 was identified on the basis of the substrate retinoblastoma-like protein 1 (p107, also known as RBL1) (Fowle et al., 2021; Jayadeva et al., 2010). Acidic residues of B55α (especially Asp197) bind to the (first arginine of the) conserved SLiM ‘[RK]Vxx[VI]R’, which is located four to 10 amino acids C-terminal of the proximal p-site (Kruse et al., 2020; Fowle et al., 2021). Interestingly, the identified sites bearing the motif validate the B55α-mediated preference of Ser/Thr-Pro-directed kinase motifs (Cundell et al., 2016; Zhao et al., 2019).

The assembly of B-subunits with the core enzyme is additionally regulated through PTMs on the unstructured C-terminal end of PP2Ac. About 70–90% of endogenous PP2A is C-terminally carboxymethylated (Lee and Pallas, 2007; Yu et al., 2001), which mediates the assembly of B55, B56 and B72 with the core dimer. Whereas the methylation appears to be crucial for B55 binding, B56 and B72 can also bind to unmethylated PP2Ac but prefer methylated forms (Yabe et al., 2018). Insufficient PP2A holoenzyme formation caused by the absence of carboxymethylation results in loss of its functionality as an important tumor suppressor (Fowle et al., 2019). Indeed, loss of carboxymethylation or removal of the C-terminal leucine residue can lead to elevated binding of striatin family B-subunits to the AC dimer, and this trimer can have oncogenic activities (Kim et al., 2020; Kurppa and Westermarck, 2020). Increased phosphorylation of Thr304 and Tyr307, as well as decreased methylation at these residues, appears to result in decreased activity of PP2A (Lee and Pallas, 2007; Janssens et al., 2008; Sangodkar et al., 2016; Löw et al., 2014). However, although the results of the C-terminus-mediated PP2A holoenzyme formation are consistent throughout the literature, the Ogris laboratory have shown that commercially available PP2A antibodies target the conserved C-terminus of the PP2A-like PPPs and do not exclusively bind to PP2Ac. Therefore, existing results need to be carefully assessed in light of the antibody issues (Frohner et al., 2020; Schüchner et al., 2020).

Owing to its prominent role in cancer (Ruvolo, 2019; Vainonen et al., 2021; Westermarck and Neel, 2020), PP2A is a popular target for small-molecule approaches (Zhang et al., 2021). One of these is the development of proximity-inducing molecules for targeted dephosphorylation (see Box 1). Moreover, small molecules have been postulated to selectively control the activity of PP2A by functioning as a molecular glue between the subunits (Morita et al., 2020; Leonard et al., 2020). However, new data reports on strong side effects and questions whether their effects can be interpreted as PP2A specific (Vit et al., 2022), resulting in previous findings now being reconsidered (Morita et al., 2022). Further studies are needed to evaluate these findings, but for now the biological conclusions drawn from utilizing these molecules should be considered in light of the available specificity data, and their use should be carefully planned in terms of conditions used.

Within the conserved PPP family, PP2A, PP4 and PP6 constitute the PP2A-like subfamily, as they are more closely related in sequence to one another than they are to the other PPPs (Andreeva and Kutuzov, 2001; Li et al., 2013). Despite their high sequence similarity, knockdown of the individual PP2A-like PPPs in C. elegans results in different phenotypes. This indicates that they fulfill separate functions despite structural similarities of the catalytic subunits (Kao et al., 2004; Han et al., 2009; Afshar et al., 2010; Martin-Granados et al., 2008). The catalytic subunit (PP4c) forms either a dimer with PPP4R1 or PPP4R4, or a trimer with PPP4R2 and PPP4R3α or PPP4Rβ (PPP4R3α/β) (Cohen et al., 2005; Gingras et al., 2005; Park and Lee, 2020). Currently, no three-dimensional structures of PP4 have been resolved, which hampers the identification of binding grooves or SLiMs as was achieved for PP1 and PP2A. Furthermore, compared to the two aforementioned major PPPs, little is known about the regulatory mechanisms of PP4 in the cell (Park and Lee, 2020). Nevertheless, PP4 has emerged as a nuclear and chromatin-associated phosphatase that is implicated in regulating numerous nuclear processes (Isono et al., 2017; Su et al., 2017; Youn et al., 2018; Wang et al., 2019a). Still, the nuclear import mechanism of PP4 remains to be elucidated, and it has not yet been determined which holoenzyme constellations of PP4 are imported into the nucleus.

Recently, binding partners of the trimeric PP4c–PPP4R2–PPP4R3α/β holoenzymes have been identified by using proteomic phage display (ProP-PD). The subunit PPP4R3 selectively recognizes binding partners containing FxxP and MxPP motifs (Lipinszki et al., 2015; Ueki et al., 2019). The binding pocket of these motifs is conserved in the N-terminal enabled/VASP homology 1 (EVH1) domain of PPP4R3. EVH1 domains are known to bind to proline-rich sequences (Ball et al., 2002). By determining the structure of the bound peptide, the phenylalanine and proline residues were demonstrated as the significant contact points of FxxP/MxPP motifs (Ueki et al., 2019). Owing to the abundance of FxxP sequences in the proteome, it is unlikely these will be the only determining factor in the PPP4R3–substrate interaction. It is expected that other as-yet-unidentified factors will contribute to this mechanism. PPP4R3 is furthermore the subunit that directly interacts with the PP4 inhibitor CT45, which reduces the activity of PP4 significantly but does not alter its interactome (Coscia et al., 2018).

Regarding a different aspect of substrate specificity, PP4c has been shown to dephosphorylate pThr faster than pSer in vitro (Ueki et al., 2019). This was also found to be the case for PP1c and PP2Ac, and concurs with the order of dephosphorylation observed in biological pathways (Hein et al., 2017; Hoermann et al., 2020). A notable exception is the above discussed PP2A:B56 holoenzyme (Kruse et al., 2020).

Post-translational modifications of PP4c and its regulatory subunits also have an impact on the activity of PP4 (Voss et al., 2013; Huang et al., 2016). Similar to the other PP2A-like phosphatases, the PP4c holoenzyme assembly is partially regulated through carboxymethylation of PP4c (Lyons et al., 2021). Of note, Lyons and colleagues nicely circumvent the above mentioned PP2A-antibody pitfalls through MS-based quantification of the PP2A-like phosphatases. To date, no PP4-specific chemical modulator has been reported, but the first endogenous inhibitors of PP4 were identified recently (Coscia et al., 2018; Han et al., 2015; Rosales et al., 2015; Park et al., 2019). Further studies of underlying mode of actions of these inhibitors might aid the development of PP4-specific chemical modulators.

Similar to other PPPs, PP5 is a ubiquitously expressed phosphatase important for a multitude of signaling pathways involved in cell cycle progression, stress responses, muscle contraction and other processes (Hinds and Sánchez, 2008; Sager et al., 2020; Krysiak et al., 2018; Neumann et al., 2021). In contrast to the other PPPs, PP5 is a multidomain phosphatase and does not form multimeric holoenzymes. The catalytic domain (PP5c) with its regulatory domain is encoded in a single gene throughout eukaryotes and forms one polypeptide (Chen et al., 2017). PP5c is N-terminally flanked by three tetratricopeptide repeat (TPR) domains, which bind to the C-terminal αJ-helix (Yang et al., 2005). Structural analysis revealed that together they maintain PP5 in an autoinhibited conformation with its active site covered by the TPR domains (Fig. 4A) (Yang et al., 2005). Interestingly, the interaction of the TPR helices with the PP5c domain resembles the mechanism of PP1c inhibition through pan-inhibitors (Yang et al., 2005; Kita et al., 2002; Maynes et al., 2001; Goldberg et al., 1995). Consequently, removal of the TPR domains or the C-terminal 13 residues potently activates the enzyme 10- to 50-fold (Yang et al., 2005; Oberoi et al., 2016). However, the TPR domain does not exclusively mediate the autoinhibition, but additionally is the primary interface for protein–protein interactions, substrate recruitment and phosphatase activation. Because of this versatility, the TPR domains of PP5 have emerged as the universal regulator of PP5.

Therefore, a significant part of research into PP5 biology focuses on its TPR domain-involved activation. PP5 can be released from its autoinhibited state through the interaction of the TPR domain with endogenous molecules, such as heat-shock proteins (HSPs), polyunsaturated fatty acids, or Ca²⁺/S100 proteins (Fig. 4B) (Yamaguchi et al., 2012; Russell et al., 1999; Haslbeck et al., 2015; Tallima and El Ridi, 2018; Hong et al., 2017). Furthermore, the allosteric PP5 small-molecule activator 2 (P5SA-2) was found to increase PP5 activity, but barely affects substrate binding (D'Arcy et al., 2019; Haslbeck et al., 2015). P5SA-2 binds to the PP5c domain at its interface with the TPR domain, inducing a conformational change of the TPR domains. This leads to an impaired compatibility of the αJ-helix with its binding groove on the TPR domain and finally in the release of autoinhibition and substrate accessibility in PP5 (Hinds and Sánchez, 2008; Haslbeck et al., 2015).

Similar to other PPPs, PP5 activity can further be modulated by PTMs on its catalytic domain. Phosphorylation of PP5 on Thr362 by the kinase CK1δ results in activation of PP5 in the absence of HSP90 in vitro (Dushukyan et al., 2017). PP5 is also subjected to multi-monoubiquitylation on Lys185 and Lys199 by the VHL complex, which leads to its degradation and serves as a control switch to shut off PP5 activity by its degradation (Dushukyan et al., 2017).

PP6 is the third member of the PP2A-like phosphatase subfamily. Nevertheless, PP6 is not dispensable since genetic deletion of the PP6 gene (PPP6C) revealed that PP6 is vital for early embryogenesis (Ye et al., 2015; Ogoh et al., 2016). Analogous to the two other PP2A-like phosphatases, PP6 forms heterotrimeric holoenzymes (Ohama, 2019). The PP6 holoenzyme is composed of one catalytic subunit PP6c, one of three Sit4-associated protein (SAPS1–3) domain-containing scaffold subunits called PPP6R1, PPP6R2 and PPP6R3, and one of the three ankyrin repeat proteins (ANKRD28, ANKRD44 and ANKRD52) that likely serves as its regulatory subunit (Stefansson et al., 2008).

As for PP4, no structure of the PP6 holoenzyme nor any subunit has been resolved yet, and only partial knowledge of the holoenzyme formation is available. It has been shown that the N-terminal region of the scaffolding subunits (SAPS domains) is sufficient for PP6c binding (Twells et al., 2001). Additionally, prior reports from mouse embryonic fibroblasts indicate that ∼75% of PP4c and PP6c are carboxymethylated, which functions as an assembly-directing PTM for PP2A and PP4 (Hwang et al., 2016; Lyons et al., 2021). However, in contrast to PP2A and PP4, loss of methylation does not affect PP6 holoenzyme formation with any of its scaffolding or regulatory subunits (Lyons et al., 2021). Furthermore, PP6 does not stably associate with protein phosphatase methyl esterase 1 (PME-1; also known as PPME1), indicating a carboxymethylation independent regulation of PP6 complex formation (Yabe et al., 2015; Lyons et al., 2021).

In further contrast to PP2Ac, the protein level of PP6c fluctuates greatly, as it is upregulated upon increased cell density in epithelial cell lines (Ohama et al., 2013). This raises the question of how its half-life is regulated. It has been suggested that the interactions of PP6Rs with the catalytic subunit is critical for its protein stability (Wengrod et al., 2015; Watanabe et al., 2018). Furthermore, PP6 has been found to be degraded by p62 (also known as SQSTM1)-dependent autophagy (Fujiwara et al., 2020); this differs from PP2A degradation, which occurs through the ubiquitin-proteasome system (Fujiwara et al., 2020; Ohama et al., 2013; Yabe et al., 2015).

Overall, PP6 remains the least-studied phosphatase of the here discussed PPPs, despite being the most expressed PPP in HeLa cells (Nagaraj et al., 2011). Of note, depending on the substrate, PP6 is even more sensitive to the commonly used PP2A inhibitor okadaic acid (Prickett and Brautigan, 2006), which is often not considered when this inhibitor is used. It is therefore possible that PP6 is in part the cause of the phenotypes induced by okadaic acid that are, however, attributed to other PPPs. In general, when using such natural toxins as PPP inhibitors, one needs to be aware that they are rarely specific (reviewed by Zhang et al., 2021; Fahs et al., 2016; Swingle et al., 2007). However, more insights into the cellular regulatory mechanisms of PP6 are needed to evaluate this potential bias. In summary, although PP6 is frequently observed to play an important role in many pathways (Ohama, 2019), very little is known about its mechanism and regulation. A better understanding of PP6 function might come from insights obtained for PP2A owing to their similarity.

The importance of each PPP for proper cellular function is indisputable. However, many characteristics of PPP function differ from other phosphatases and kinases. Thus, many mechanisms remain so far ambiguous, and specific targeting of individual PPP holoenzyme proteins is still challenging. Nonetheless, considerable improvements in the survey of the human phosphoproteome have revealed the magnitude of sites that are potentially regulated by the PPP family. Furthermore, initial large-scale approaches performed with the major PPPs, PP1 and PP2A, have helped to narrow the gap between site identification and attribution. For instance, it has been demonstrated that PP1c contains intrinsic substrate specificity, which is determined by the B-subunits for PP2A (Hoermann et al., 2020; Kruse et al., 2020). Many recent observations show that SLiM-mediated regulation is common among various aspects of PPP biology. By using SLiMs, PP1c recruits regulatory proteins, which can, but do not have to be, substrates, whereas PP2A holoenzymes (the B-subunits) bind substrate proteins. Interestingly, SLiMs described in this Review (RVxF, LxxIxE and FxxP) change their functional properties drastically upon phosphorylation within their variable positions (x) and at motif-surrounding residues (Nasa et al., 2018; Hertz et al., 2016; Ueki et al., 2019). This indicates an integral layer of regulation, which further fine tunes the SLiM-mediated regulation of PPPs. So far, our understanding of PPP biology is heavily biased towards PP1 and PP2A. Owing to the prevalence of studies addressing them and the variety of their substrates, they are typically assessed more frequently in experiments. However, this bias also has its advantages. Research on PP4 and PP6, for example, benefits greatly from new experimental designs and findings gained from research into PP2A because of their similarity to PP2A. Finally, the understanding of the TPR-driven regulation of the multidomain phosphatase PP5 has a huge potential for its specific targetability. Here, the combination of structural biology, crystallography, phosphoproteomics and chemical biology emerges as a successful approach to better understand PPPs. Thus, we anticipate many meaningful discoveries relating to PPPs in the upcoming years.

We thank the following funding agencies for supporting our work; the German Research Foundation (DFG) within the FOR 2743 (KO 4013/6-1 and HO 1518/13-1) and the CRC 1381 (Project-ID 403222702 – SFB 1381), Germany's Excellence Initiative (DFG, EXC 294 BIOSS), Germany's Excellence Strategy (DFG, CIBSS – EXC-2189 – Project ID 390939984), and the European Research Council with an ERC consolidator grant (#865119) and an ERC starting grant (#336567) to M.K.