Turn and Face the Strange: ^{A New View on Phosphatases}

Abstract

Phosphorylation as a post-translational modification is critical for cellular homeostasis. Kinases and phosphatases regulate phosphorylation levels by adding or removing, respectively, a phosphate group from proteins or other biomolecules. Imbalances in phosphorylation levels are involved in a multitude of diseases. Phosphatases are often thought of as the black sheep, the strangers, of phosphorylation-mediated signal transduction, particularly when it comes to drug discovery and development. This is due to past difficulties to study them and unsuccessful attempts to target them; however, phosphatases have regained strong attention and are actively pursued now in clinical trials. By giving examples for current hot topics in phosphatase biology and for new approaches to target them, it is illustrated here how and why phosphatases made their comeback, and what is envisioned to come in the future.

Short abstract

Allostery is the key: New approaches to target phosphatases and a deeper understanding of phosphatase biology pave the way toward clinical applications of phosphatase inhibitors and activators.

Introduction

Phosphatases remove phosphate groups from biomolecules and are the important counter players of kinases, together governing the physiological phosphorylation state balance. Imbalance of the regulation of phosphorylation and dephosphorylation can contribute to the development and progression of a myriad of diseases, placing kinases and phosphatases in important positions for drug targeting. While kinases have been successfully targeted numerous times to treat diseases, especially cancer,¹ this is not the case for phosphatases.² One reason is that the initial focus for phosphatases inhibitor development was on active site inhibitors. Active site inhibitors for phosphatases generally suffer from limited selectivity and bioavailability. This is due to the strong active site conservation and the preference of the active site for negative charges that are often required to ensure tight binding.² In contrast, targeting of the ATP-binding pocket in kinases has been highly successful and resulted in several approved drugs.¹ Reasons are that negative charges are not required for binding, which improves their bioavailability, and that selectivity can be achieved to an extent that is sufficient for the treatment of cancer patients. While also a kinase inhibitor that targets the ATP-binding pocket is never completely selective,³ this has even been used as an advantage in polypharmacological applications, where it is desired that the drug targets several kinases.⁴ For a long time, phosphatases therefore have been stigmatized as “housekeeping enzymes” and “undruggable”.^2,5 Challenges to study and target them included their evolutionary diversity resulting in different dephosphorylation and regulatory mechanisms, the conserved active sites within the families, and the difficulty to assign substrates to specific phosphatases.^2,6,7 Therefore, a few years ago, I might have started an essay on phosphatase research with “Phosphatases: the final frontier. To boldly go, where no one dares going.” – but not anymore. In the last few years, interest in phosphatase biology and in phosphatases as targets for drug discovery has been strongly renewed. A recent review is indeed titled “Phosphatases start shedding their stigma of undruggability”.⁸ Why the renewed interest, and what can we expect in the coming years?

Phosphatase research has made huge progress in the past two decades after the first hype of phosphatase drug discovery led to the said stigma. The availability of new methods and new creative approaches to solving the specific challenges of phosphatase research² is paving the way to target phosphatases therapeutically. For example, a reclassification of the human phosphatome and the listing of their substrates in the human dephosphorylation database DEPOD have revealed new phosphatase relationships, have offered new insights into phosphatase-substrate networks,^6,7,9 and have triggered the extension of the reclassification to other species.¹⁰ Mass spectrometry methods such as inhibitor-bait pull-downs¹¹ or proximity-dependent biotinylation have delivered interactomes of phosphatases,¹² and structural and as well as mechanistic studies have provided a more detailed understanding of the regulation of phosphatase activity¹³⁻¹⁵ and how phosphatases recognize substrates.¹⁶ As the following examples will show, this more detailed understanding has been crucial to follow new leads for phosphatases as drug targets.

Allosteric Targeting of the Oncogene SHP2

One of the hottest topics at the moment in the world of phosphatases concerns protein tyrosine phosphatase nonreceptor type 11, ptpn11, commonly called SHP2 for SH2-domain containing phosphatase-2. SHP2 is a phosphotyrosine-specific phosphatase that acts in several cancers as a proto-oncogene. Its mutation-caused hyperactivation in cancer results in uncontrolled tumor growth.¹⁷ SHP2 serves as an essential and common node that is activated downstream of most receptor tyrosine kinases (RTKs), activating the mitogen-activated protein kinase (MAPK) pathway (Figure 1A).¹⁷ Therefore, not only is SHP2 a proto-oncogene itself, but it also triggers tumor cell proliferation and survival by integrating multiple oncogenic signaling pathways irrespective of the upstream mutations present in a particular cancer type.¹⁸ Accordingly, SHP2 inhibitors have the potential to dampen tumor growth driven by a number of different genetic aberrations. While many active site inhibitors of SHP2 have been developed and were applied successfully as chemical tools, none of them have yet made it to the clinic.^19,20 In addition, some of them have recently been shown to have severe side effects,²¹ which should be taken into consideration when using them as a tool compound to interpret data on SHP2 biology.

Figure 1.

SH2-domain containing phosphatase-2 SHP2. (A) Upon receptor activation, SHP2 is recruited in different ways to activate the MAPK pathway.³² Dotted arrows signify more than one step in between the effect. RAS: Rat sarcoma small GTPase; RAF: rapidly accelerated fibrosarcoma kinase; MEK: MAPK/ERK kinase; ERK: extracellular-signal activated kinase. (B) In the inactive state, the N-SH2 domain (green) blocks the active site of the PTP domain (orange). Upon phosphotyrosine (pY) binding of the N- and the C-SH2 (blue) domains, the closed conformation is released and the active site is accessible for substrates. The pYs can be part of a single protein, of two proteins, or of the SHP2 C-terminus, which can be phosphorylated on different tyrosines.³² (C) The inhibitor SHP099 traps SHP2 in the inactive conformation by binding at an interface between the domains. Structures in (B) and (C) from pdb entry 5EHR.¹⁸ (D) Major histocompatibility complexes (MHCs) present antigens on tumor cells that can be recognized by the T cell receptor (TCR), triggering immune cell activation. Tumor cells express PD-L1, which binds to PD1 presented on the T cell surface. Upon binding, SHP2 is recruited to PD1, leading to the dampening or abrogation of T cell activation. Cancer cells use this mechanism to evade the immune response.³³

A 2016 landmark study changed the game when Novartis introduced a new concept to inhibit SHP2.¹⁸ The key to this new concept is allostery, and the design of the inhibitor was enabled through obtaining a deeper understanding of the general SHP2 activation mechanism. SHP2 contains two SH2-domains N-terminal of the catalytically active PTP domain (Figure 1B). The N-SH2 domain binds to the PTP in the inactive state, closing the active site for substrate entry. Upon phosphotyrosine (pTyr)-containing ligand binding, the N-SH2 domain is released from the PTP domain and SHP2 is active. Often, the pTyr-containing ligands are part of RTKs such as the epidermal growth factor receptor (EGFR). In cancer, mutations that cause hyperactivation and hyperphosphorylation of RTKs hence also trigger the constant activation of SHP2. In case SHP2 itself acts as an oncogene, mutations in SHP2 cause the N-SH2 domain to disengage from the PTP domain and render SHP2 constantly active. For the new allosteric inhibitor, the authors developed a novel assay using the full-length SHP2 protein and measuring the inhibition of SHP2 pTyr-peptide-mediated activation by small molecules.¹⁸ To activate SHP2, they used a bisphosphorylated peptide derived from the insulin receptor substrate (IRS)-1 that is known to bind to both SH2 domains of SHP2.²² The activity of SHP2 was monitored using 6,8-difluoro-4-methylumbelliferyl phosphate (DiFMUP) as an unnatural substrate.¹⁸ This setup allowed them to screen a diverse library of 100 000 compounds for inhibition of SHP2 activation. In a counter screen, the authors tested whether the compounds also inhibited the SHP2-PTP domain alone, and hits from this screen were deprioritized in order to select for truly allosteric SHP2 inhibitors. The best compounds from this screen were subjected to optimization using medicinal chemistry. This led to the discovery of a selective and orally bioavailable inhibitor of SHP2, called SHP099, with an IC₅₀ of 71 nM, that binds into the cleft between the three domains in the inactive state, trapping SHP2 inactive conformation (Figure 1C). SHP099 inhibited the growth of xenograft tumors in mice.¹⁸ This landmark study was accompanied by the thorough structure–activity relationship (SAR) study carried out for the discovery of SHP099²³ and followed by the development of further allosteric inhibitors^24,25 and the structural investigation of the mechanism of inhibition and SHP2 activation by mutations in cancer.^14,26 SHP099 was used to show that SHP2 targeting could be effective upon kinase drug resistance,^27,28 and that it holds promise to be beneficial as combination therapy together with MAPK kinase (MEK) inhibitors for very difficult to therapeutically address KRAS-driven cancers.^29,30 The hope of a breakthrough that solidly establishes PTPs as drug targets in cancer, as was the case for kinases with the inhibitor Imatinib in chronic myeloid leukemia,^1,31 lies therefore in the current clinical trials involving the allosteric SHP2 inhibitor TNO155 by Novartis (Clinicaltrials.gov ID NCT03114319) and also on clinical trials of other compounds that followed shortly after TNO155: RMC-4630 by Revolution Medicines with Sanofi (Clinicaltrials.gov ID NCT03989115) as well as JAB-3068 and JAB-3312 by Jacobio Pharmaceuticals (Clinicaltrials.gov IDs NCT03565003 and NCT04045496).

Another hot topic that further spurs interest in the inhibition of SHP2 is cancer immunotherapy. This is because in immune cells SHP2 binds to the receptor programmed cell death-1 (PD1) upon PD1 activation by its ligand PD-L1, triggering downstream pathways that lead to the dampening of the immune response (Figure 1D).³³ PD-L1 can be expressed on cancer cells to block the activation of immune cells that would destroy the cancer cells. In this way, cancer cells can evade the immune response. PD1 and PD-L1 are the targets of so-called immune-checkpoint inhibitors, which are antibodies that inhibit their interaction in order to block cancer immune evasion. Inhibition of SHP2 downstream of PD1 could therefore support immune-checkpoint inhibitors in combination cancer immunotherapies.^34,35

The highly successful strategy to find allosteric SHP2 inhibitors could also be applied to other phosphatases. The obvious first choice would be SHP1, the closest homologue of SHP2 sharing 61% amino acid sequence identity¹⁸ that is also regulated through the same mechanism using two N-terminal SH2 domains.³⁶ Importantly, SHP099 showed no activity against SHP1,¹⁸ corroborating the potential of the approach for potent and selective inhibitor discovery. While SHP2 is ubiquitously expressed, SHP1 expression is more restricted and highest in hematopoietic cells.³² SHP1 regulates immune functions such as the development of immune tolerance and is a negative regulator of T cell receptor (TCR)-mediated as well as cytokine signaling and a key downstream node of several receptors in B cells.³⁶⁻³⁸ SHP1 acts downstream of different receptors than SHP2 in order to negatively regulate immune cell activity.^38,39 However, SHP1 deficiency triggers autoimmune diseases.^37,38 Nevertheless, since not all patients respond to cancer immunotherapies that target PD1 or cytotoxic T-lymphocyte associated protein 4 (CTLA-4), which is another receptor that negatively regulates the immune response, it would be attractive to not only target SHP2 but also SHP1 in cancer immunotherapy.³⁹ Temporary SHP1 inhibition could block the downregulation of the T-cell immune response, while not permanently damaging the immune system, in order to be used as support in cancer immunotherapy without causing permanent autoimmune diseases. Because SHP1 inhibitors so far have never been selective over SHP2,^39,40 the approach to screen for allosteric inhibitors appears particularly promising.

The MAPK phosphatases (MKPs) are another family of phosphatases that could be targeted allosterically using a variation of this strategy. MKPs regulate MAPK activity through MAPK dephosphorylation. MKPs carry a kinase interaction motif (KIM) that is required to recognize their specific MAPK substrate.⁴¹ While the logical consequence of the fact that elevated MAPK activity drives tumor progression would be that MKPs are tumor suppressors, this relation is not true for all MKPs. For example, elevated MKPs have been reported to reduce the pro-apoptotic effects of chemotherapeutics that often take effect through stress-activated MAPK pathways.^42,43 It therefore appears that MKPs can act also as tumor promoters, and specific inhibitors could hence be of clinical use. Some MKPs are activated upon binding to their MAPK substrate.^41,44 Accordingly, an alteration of the above-described strategy to find allosteric inhibitors for these MKPs could be to use instead of the bisphosphorylated IRS-1 peptide the corresponding recombinant MAPK (such as ERK2 for MKP-3⁴⁴). Activation of the MKP would be monitored using DiFMUP, which would allow screening for inhibitors of this activation as done with SHP2. The counter screen would include testing for inhibition of the truncated MKP without KIM, and in this way, selective allosteric inhibitors for MKPs could be discovered.

The Comeback of PTP1B

Protein tyrosine phosphatase 1B (PTP1B) is a negative regulator of insulin signaling⁴⁵ (Figure 2A), regulates Janus kinases (JAKs)/signal transducers and activators of transcription (STATs) in cytokine and leptin signaling^45,46 (Figure 2A), and is a positive regulator of human epidermal growth factor receptor 2 (HER2, ERBB2) signaling in breast tumorigenesis.⁴⁷ It was one of the first pursued phosphatase drug targets and was initially validated for diabetes and obesity.^45,48 In addition, it is also of interest as target in cancer^47,49 and in dendritic cell-based cancer immunotherapy.⁴⁶ Furthermore, it was recently suggested as a target in neurodegenerative diseases, particularly Alzheimer’s disease,⁵⁰ Rett syndrome,⁵¹ anxiety,⁵² atherosclerosis, and cardiovascular disease⁵³ including heart attack.⁵⁴ Despite its many functions, and important for drug discovery, PTP1B knockout mice were found to be healthy.^55,56 Therefore, there is a vast amount of literature on PTP1B in drug discovery, which I would like to refer to here (see refs (57−63)). Still, until now, PTP1B inhibitors have not made it to the clinic because in the past most of them have been active site inhibitors. As for active site inhibitors of SHP2 and other PTPs, this fact has caused problems regarding selectivity and bioavailability, and has led to PTPs having the stigma of “undruggability”.⁶¹ However, also for PTP1B the world has turned, and an allosteric PTP1B inhibitor, trodusquemine (MSI-1436)^47,48 (Figure 2B), was tested in clinical trials for obesity and type 2 diabetes (clinicaltrials.gov ID NCT 00606112) and for breast cancer (clinicaltrials.gov ID NCT 02524951). MSI-1436 targets full-length PTP1B (50 kDa), containing an extended C-terminal domain, 10-fold more potently than the short version of the protein (37 kDa). It binds to both the C-terminus and the PTP domain, initiating conformational changes that led to inhibition.⁴⁷ PTP1B occurs in both forms in cells, and the C-terminal extension serves a regulatory role.⁴⁵ While MSI-1436 was also potent in the treatment of cardiovascular disease,^53,54 it is charged and has limited bioavailability, and therefore an orally bioavailable analogue has been developed called DPM-1001⁴⁵ (Figure 2B). Interestingly, DPM-1001 binds copper and is more potent in its copper-bound form.⁴⁵ The detailed modes of binding of DPM-1001 to PTP1B and of copper to DPM-1001 need to be further elucidated by structural studies, and a SAR study of these compounds has not yet been published. It will be exciting to see future studies and clinical trials aiming to bring one of these molecules to clinical application.

Figure 2.

Protein tyrosine phosphatase 1B (PTP1B). (A) PTP1B dephosphorylates the insulin receptor and the insulin receptor substrate (IRS) as well as JAK2 to regulate different pathways. It is a negative regulator of insulin and leptin signaling.⁴⁷ The fat cell-derived hormone leptin binds to the leptin receptor, which is a type I cytokine receptor.⁶⁶ Other cytokine receptors and the JAK/STAT pathway regulation by PTP1B (and other PTPs) are important in immune cells.⁴⁶ (B) Structure of MSI-1436 and proposed structure of the Cu²⁺-complexed DPM-1001.⁴⁵ (C) The mechanism of action of PROTACs. E2/E3 = E2/E3 ligase; Ub = ubiquitin.

Another exciting advance comes from the world of antisense technologies. Antisense oligonucleotides (ASOs) bind to mRNA through Watson–Crick base pairing, leading to the degradation of the mRNA mediated by RNase H1.⁶⁴ In this way, PTP1B protein cannot be synthesized. In a pilot study in humans, ISIS Pharmaceuticals carried out a clinical trial successfully, observing in patients with type 2 diabetes glucose-lowering effects and improvements in leptin and insulin sensitivity.⁶⁵ Recently, the outcome of a phase II, double-blind, randomized, placebo-controlled, multicenter trial was reported.⁶⁴ The compound called IONIS-PTP-1B_Rx is a 2′-O-methoxyethyl (2′-MOE)-modified second-generation ASO. The study was carried out as an add-on therapy in overweight patients with type 2 diabetes unable to maintain glycemic control with metformin alone or in combination with sulfonylurea, both of which are standard-of-care for type 2 diabetes. Patients were given 200 mg of IONIS-PTP-1B_Rx subcutaneously once weekly for 26 weeks.⁶⁴ Overall, this trial had a positive outcome, with no safety concerns as the compound was well tolerated and showed significant weight reductions compared to placebo, improved medium-term glycemic parameters, and reduced leptin levels. The authors note that the small trial size is a limitation, and that the potency and bioavailability will have to be improved further. The initial target group of patients that this drug is aimed at should have severe insulin resistance and be obese, and should take high doses of insulin therapy. Therefore, the following studies will be dedicated to patients that are unresponsive to insulin therapy, because inhibition of PTP1B improves insulin sensitivity. In addition, IONIS-PTP-1B_Rx could potentially prevent diabetes type 2 development in obese patients by inducing weight loss.⁶⁴

These examples demonstrate that targeting PTP1B has made strong advances in the past years. Given the many disease indications that PTP1B has been shown to be involved in, these new approaches give hope that a drug for this phosphatase will become an approved therapy, and once this happens it will be a powerful drug to treat a variety of diseases. For further new approaches to target PTP1B, the reader is referred to refs (61 and 62). In addition, a promising new method for PTP1B, and for phosphatases in general, that has not yet been applied to phosphatases is the design of PROTACs, for Proteolysis TArgeting Chimeras.⁶⁷ PROTACs are bifunctional compounds that bind to the target protein and to an E3 ubiquitin ligase, leading the target protein to be ubiquitinated and degraded⁶⁷ (Figure 2C). Thus, PROTACs only require a binder and not an effective inhibitor, which avoids the issues of limited bioavailability and selectivity of active site inhibitors for phosphatases. For many phosphatases, including PTP1B, binders are available that are not necessarily very good inhibitors, and that could be converted into a PROTAC. This approach appears to be promising particularly for small phosphatases that do not contain domains for allosteric regulation and that have been notoriously difficult for inhibitor discovery such as the phosphatases of regenerating liver^68,69 (PRLs or PTP4As) and cell division cycle 25 (CDC25) phosphatases.⁷⁰ These play prominent cancer promoting roles and are therefore valuable drug targets.

Protein Phosphatase-1: Targeting of Holoenzyme Formation

Another example of strongly developing areas in phosphatase targeting is the ubiquitously expressed protein-phosphatase-1 (PP1). PP1 is one of the major serine/threonine-specific phosphatases, counteracting more than 100 kinases.⁶ It does so in a specific manner by forming holoenzymes with regulatory interactors of protein phosphatase one (RIPPOs)⁷¹ (Figure 3A), of which about 200 are known.¹³ Because in the literature often “PP1” refers indiscriminately either to these holoenzymes or to PP1 alone, here PP1 alone is referred to as “PP1c” for the catalytic subunit. These RIPPOs were previously called PP1-interacting proteins (PIPs),¹³ but were renamed to avoid confusion with phosphatidyl inositol phosphates that are also abbreviated “PIPs”, based on a discussion at the Faseb protein phosphatase conference 2018. Many RIPPOs contain a so-called RVxF-type binding motif, which is a short linear motif (SLiM)⁷² that binds to the RVxF-binding site on PP1c^73,74 (Figure 3A). This SLiM-binding site is unique to PP1c within the phosphoprotein phosphatase (PPP)-like family that PP1c belongs to and is the primary anchor point for many RIPPOs that can engage in PP1c binding through further interactions. RIPPOs are largely unstructured proteins and fold upon binding to PP1c.^13,71,74

Figure 3.

Protein phosphatase-1. (A) PP1 exists as catalytic subunit PP1c, which forms holoenzymes with RIPPOs such as Ki67 (pdb entry 5J28)⁷⁴ through different interactions. The RVxF-binding site on PP1 binds a peptide stretch, a SLiM, in the RIPPO called RVxF-type binding motif, and this is the primary anchor point of most known RIPPOs. (B) PP1-disrupting peptides (PDPs) bind PP1c through the RVxF binding site, which is far away from the active site. Therefore, the active site is not blocked through PDP binding as shown by the crystal structure of PP1c bound to PDP2 (pdb entry 4G9J).⁷³ The black arrow indicates the same point on the structure for visual orientation. (C) In diseased cardiomyocytes, hyperphosphorylation of RyR2 leads to the sarcoplasmic reticulum (SR) leaking calcium to the cytosol, causing arrhythmia and heart failure.⁷⁹ Treatment with PDP3 was shown to reduce the SR calcium leak and to lead to RyR2 dephosphorylation, making dephosphorylation of the RyR2 an attractive pharmacological approach.⁷⁹

It is highly challenging to accomplish selective modulator design for PP1c through active site targeting due to the high conservation of the active sites within the PPP-like family.^75,76 However, targeting the holoenzyme formation has enabled the creation of truly selective modulators of PP1c activity over other PPP-like family members.^73,77 To this end, peptides that bind to the RVxF-binding site were designed using alanine scans and starting from a sequence from a potently binding RIPPO (Figure 3B). These so-called “PP1-disrupting peptides” (PDPs) were optimized from a non-cell-penetrating, metabolically unstable version (PDP1⁷³) to cell-penetrating, stable versions (PDP3,⁷³ PDP-Nal⁷⁸). PDPs were shown to disrupt PP1 holoenzyme formation with nanomolar potency in vitro, liberating PP1c that can dephosphorylate nearby substrates in live cells when using the stable versions.^73,78 Excitingly, it was shown that PDP3 treatment of human diseased myocardium had an antiarrythmic effect.⁷⁹ The Ca²⁺ handling machinery that is regulated by (de)phosphorylation events is responsible for the robust, rhythmic repetition of cardiac contraction (systole) and relaxation (diastole). Within this complex machinery, phosphorylation of the ryanodine receptor type 2 (RyR2) at the sarcoplasmic reticulum (SR) by Ca/calmodulin-dependent protein kinase II (CaMKII) leads to the systolic release of Ca²⁺ (Figure 3C), and that to heart muscle contraction.^80,81 In turn, dephosphorylation of RyR2 results in diastolic removal of Ca²⁺ from the cytosol by SR Ca²⁺ ATPase (SERCA) activity, leading the muscles to relax.⁸¹ CaMKII can hyperphosphorylate RyR2 in the diseased heart, which leads to an enhanced SR diastolic Ca²⁺ leak and reduced SR Ca²⁺ load, causing arrhythmia and ultimately heart failure.^79,80 The PPP-like phosphatases PP1, PP2A, and calcineurin (PP2B) have all been linked to the modulation of RyR2 function.⁸¹ By treating human heart failure patient samples with PDP3, it was shown that liberating PP1c is sufficient to reduce RyR2 phosphorylation, resulting in potent reduction of arrhythmias and of the SR diastolic Ca²⁺ leak.⁷⁹ The effect of PDP3 treatment on RyR2 dephosphorylation was recently confirmed by an independent study in murine cardiomyocytes,⁸² and studies using PDP3 in murine cardiomyocytes stimulated with a toxin (ATX-II) to trigger a SR Ca²⁺-leak confirmed the beneficial effect.⁸³ These studies opened up the exciting prospect of the liberation of PP1c at RyR2 as a promising antiarrythmic approach in heart disease, complementary to the use of other therapies such as β-blockers.^79,83,84 However, before clinical applications can be considered, hurdles such as pharmacokinetic issues including limited uptake and stability of peptide-based therapeutics, which may require the replacement of the peptide with a peptide-mimetic or a small molecule, organ, and subcellular specificity as well as applicability to different cardiac pathologies and disease stages need to be considered.^83,84 A possible solution to achieve subcellular specificity could be the use of bidentate ligands that would bind to PP1c, like the PDPs, and to RyR2, for example ent-(+)-verticilide,⁸⁵ “gluing” PP1c and RyR2 together. A novel prodrug concept that would allow PP1c to bind only when the bidentate ligand is already localized at RyR2 would further reduce side effects. Indeed, recently the PDP1 sequence⁷³ and also a minimal RVxF-sequence required for PP1-binding⁸⁶ were fused to an AKT kinase inhibitor guiding PP1c to AKT, in order to dephosphorylate AKT leading to its deactivation.⁸⁷ The molecules were called phosphatase recruiting chimeras, PhoRCs.⁸⁷ This approach showed low dephosphorylation potency because the non-cell-penetrating, metabolically unstable PDP1⁷³ and the minimal sequence with low PP1c binding potency instead of the more potent molecule reported by Tappan and Chamberlin⁸⁶ were applied.⁸⁷ Nevertheless, this exciting proof-of-principle was successful and is expected to be more efficient when PDP3, PDP-Nal, or potent small molecule binders would be used.⁸⁷

Targeting of a SLiM-binding site was also found to be the mechanism of action of the approved drugs FK506 (Tacrolimus) and cyclosporin A (Ciclosporin) to calcineurin.^88,89 Both are important immunosuppressant medications,^88,89 with Ciclosporin being named on the World Health Organization’s List of Essential Medicines.⁹⁰ While at the time of approval their detailed mechanism of action was not clear, a few years ago it was shown that they bind to the LxVP-SLiM-binding site on calcineurin, which is also a primary binding site for substrates.^88,89 Thus, both inhibitors block substrate binding to calcineurin.⁸⁸ Intriguingly, for other members of the PPP-like family, PP2A and PP4, SLiMs were also found to be essential for substrate recognition.⁹¹⁻⁹³ Taken together, targeting the SLiM-binging sites in PPP-like phosphatases is therefore currently a very exciting and promising topic in the field.

As an approach to inhibit PP1 selectively, the design of binders of substrate-specifying RIPPOs was suggested in order to block the interaction between PP1c and the RIPPO.⁹⁴⁻⁹⁶ In this way, PP1c would be prevented from binding to a specific substrate-specifying RIPPO and would therefore not be able to recognize and dephosphorylate its substrate anymore. Therefore, only a particular function of PP1 should be compromised. These studies looked at the role of PP1 in the unfolded protein response (UPR). The UPR restores proteostasis following stresses that result in the accumulation of unfolded proteins at the endoplasmic reticulum.^95,97 One possible response is the phosphorylation of the eukaryotic initiation factor 2α (eIF2α). eIF2α is essential for translation initiation. Phosphorylation inactivates eIF2α leading to inhibition of translation initiation and reduced protein production, helping the stressed cells to deal with accumulated unfolded proteins.⁹⁸⁻¹⁰⁰ In turn, the holoenzymes PP1c:PPP1R15A (also called GADD34) and PP1c:PPP1R15B (also called CReP) dephosphorylate eIF2α, reinitiating protein translation.⁹⁸⁻¹⁰⁰

Salubrinal was the first molecule discovered to inhibit translation by blocking dephosphorylation of eIF2α.⁹⁴ While it was assumed that this happened via inhibition of the PP1c:PPPR15 holoenzymes, direct evidence was not given, also not in follow-up studies.^101,102 Following the discovery of Salubrinal, Guanabenz was observed to prolong eIF2α phosphorylation.⁹⁵ Because Guanabenz is also an α2-adrenergic receptor agonist,⁹⁵ Sephin1 was developed as a molecule that still prolonged the stress response but did not show other effects of Guanabenz.⁹⁶ Initially, data pointed into the direction that Sephin1 and Guanabenz would directly inhibit the PP1c:PPP1R15A holoenzyme by inhibiting the holoenzyme or disrupting its formation or its interaction with eIF2α.^95,96,103 However, it is now clear that under physiological buffer conditions as well as for reaction times and inhibitor concentrations that reflect the concentrations and timelines necessary for cytoprotection, the compounds neither directly inhibit the PP1c:PPP1R15A holophosphatase-mediated dephosphorylation of eIF2α,^104,105 nor directly disrupt the holoenzyme formation.^100,106 Reports on binding of the compounds to PP1c:PPP1R15A differ.^100,106 In addition, a new inhibitor called Raphin1 was reported to bind preferably to PP1c:PPP1R15B over PP1c:PPP1R15A, but it bound to the latter with a similar potency as Sephin1 did.¹⁰⁰ It was found that Raphin1 treatment leads to the proteasomal degradation of PPP1R15B.¹⁰⁰ Surprisingly, effects of Guanabenz and Sephin1 on PPP1R15A protein or mRNA levels or degradation have never been published. Furthermore, differences in findings on responses to compound treatment in PPP1R15A knockout cells and different animal models,^{95,96,99,105,107} and through which pathways of the stress response these compounds could act,^97,108 suggest that they have a variety of beneficial effects in the treatment of proteostasis diseases. For further insights and for own assessment of the data, I encourage the reader to study the primary literature on this subject.

This case stresses the importance of the thorough characterization of the mode of action of an inhibitor, in order to not mislead the conclusions. Particularly for phosphatases that still are in the process of shedding their stigma of being undruggable, proof of the mode of action is extremely important in order to not fall back into that stigma. Accordingly, as tool compounds, using Salubrinal, Sephin1, Guanabenz, and Raphin1 does not allow interpreting data concerning the direct involvement of PP1c:PPP1R15 holoenzymes. Nonetheless, in all these reports, there is the agreement that Sephin1 holds strong potential as a protective agent against cellular stress. Accordingly, a compound based on these studies called IFB-088 has entered phase 1 clinical trials (InFlectis BioScience, clinicaltrials.gov ID NCT03610334). Thus, while the mechanism of action of these compounds is still not understood, this work has led to a promising drug candidate.

Summary and Outlook

I have chosen here examples for developments of phosphatases in drug discovery that include new approaches and translational relevance, but I could not cover all developments made in diverse disease indications and concerning different phosphatases. Many more phosphatases should be mentioned here in more detail, as much exciting progress has been made in these areas as well. New creative targeting approaches are being developed and clinical trials are ongoing, with results to be expected in the next years. Therefore, I would like to refer to the references in Table 1 for further reading.

Table 1. Further Reading for New Developments Regarding the Roles of Phosphatases and New Approaches to Target Them for Drug Discovery^a.

phosphatase	context	refs
general overview	phosphatases in drug discovery	(113)
PP2A	inhibition and activation in different cancer types	(112, 114)
PP2B (PP3, calcineurin)	neurological disorders, neurodegeneration, immunosuppression, biological roles, substrates	(115, 116)
PPM1D (Wip1)	cancer, DNA damage response, immunity	(117, 118)
PTPs	various diseases, overview on current developments concerning different PTPs	(49, 61−63)
phosphatases of regenerating liver (PRL; PTP4A)1,2,3	roles and inhibition in cancer, physiological roles	(69, 119, 120)
cell division cycle-25 (CDC25) phosphatases	cell cycle, cancer, inhibition	(70, 121)

^aPlease also see references therein for further details, disease indications, clinical trials, and phosphatases not mentioned here.

As seen for PP1c:PPP1R15 holoenzyme inhibition, some difficulties and controversies remain. Another example for this is the inhibitor LB-100, which is undergoing clinical trials for future cancer therapy (clinicaltrials.gov IDs NCT03886662, NCT03027388, NCT01837667). LB-100 was reported to be a specific inhibitor of PP2A.¹⁰⁹ However, recently it was shown using not only biochemical data but also crystallography that LB-100 inhibits the active site of PP5,¹¹⁰ which is also a member of the PPP-like family and closer related to PP1 than to PP2A.^6,7,9 Clearly, LB-100 is therefore not a PP2A-specific inhibitor. Nonetheless, the kinase inhibitor Imatinib was also originally thought to be specific for BCR-ABL kinase, but discovery of other targets has enabled a broader use of this compound and its successors in cancer therapy.¹¹¹ Indeed, contrary to the expectation that a more general PPP-like family inhibitor would be highly toxic, in the initial clinical trial no safety concerns were obvious.¹⁰⁹ Moreover, PP2A plays also an important role in tumor suppression, and many efforts aim at the activation of PP2A in cancer with outstanding progress being made in the development of new approaches targeting PP2A.¹¹² Together with the above examples for PP1, this shows that particularly when it comes to PPP-like phosphatases, knowing the molecular context of the disease is highly important to decide if the phosphatase should be activated or inhibited. With new methods at hand,² our knowledge of basic phosphatase biology continues to advance enormously. Therefore, the more that personalized medicine, meaning the knowledge of the molecular background of a certain disease in a single patient, progresses into becoming clinical practice, the better we will be able to consider this in the future. For newest developments, I would like to mention here the annual phosphatase conference that takes place iteratively in Europe and in the USA, as well as the biannual phosphatase conference in Japan. The FASEB protein phosphatase conference will take place in Bend, Oregon, August 23–28, 2020, followed by the 14th International Conference on Protein Phosphatases (ICPP) in Kobe, Japan, December 10−12, 2020, and the next European phosphatase conference will happen in Athens, Greece, June 6–11, 2021.

Taken together, we have turned and faced the strange: Phosphatases are back in drug development, and are there to stay.

The author declares no competing financial interest.

Notes

^§ Title from David Bowie’s “Changes”, 1971.