Target-based drug discovery has become the dominant strategy in the pharmaceutical industry1. It focuses on a disease-associated target with the goal of developing molecules modulating target activity to provide potential therapeutics for disease treatment. In this process, target identification and validation play a pivotal role in the drug discovery project's success. The number of known drug targets is approximately 300 for all approved therapeutic drugs, in which a large portion of them belongs to ion channels, G protein coupled receptors, and protein kinases2. Protein tyrosine kinases (PTKs) are a very important class of oncology drug targets with >20 small molecule kinase inhibitors already FDA-approved for various cancer treatments, and a large number of kinase inhibitors are currently under various stages of clinic trials3. PTKs and protein tyrosine phosphatases (PTPs) reversibly and coordinately control cellular protein tyrosine phosphorylation levels, which are important for nearly all cellular processes, such as growth, differentiation, migration, survival, and apoptosis4.

In Chinese philosophy, the concept of Yin-Yang is used to describe the importance of balance between two seemingly opposite or contrary forces in the natural world5, and protein tyrosine phosphorylation and dephosphorylation represent exactly such a Yin-Yang relationship. At the molecular level, the extent of tyrosine phosphorylation is precisely balanced by the actions of specific PTK(s) and PTP(s), which can either up- or downregulate downstream signaling pathways, depending on whether the phosphorylation activates or inhibits protein function4. Aberrant tyrosine phosphorylation resulting from the perturbed PTK-PTP balance can cause numerous human diseases. Indeed, excessive tyrosine phosphorylation is a hallmark of cancer. Given the great success of drug discovery targeting against PTKs and the fact that proper cellular tyrosine phosphorylation levels are controlled by the coordinated activity of kinases and phosphatases, PTPs have been suggested as next generation drug targets6,7,8. PTPs have been important topics of research in biomedical science for the past two decades, and a number of PTPs have been implicated in various human diseases, such as cancer, diabetes, autoimmune, and neurological diseases9,10,11. These findings have further strengthened the belief that targeting PTPs could be the next frontier in drug discovery.

The PTPs constitute a large family of enzymes with 107 members divided into 4 groups based on their protein sequences and functions (Figure 1)8,12. The major group is class I PTPs, which includes 99 members, and each member shares a conserved active site sequence (H/V)C(X)5R(S/T) with the cysteine as the catalytic residue in the dephosphorylation reaction. Class I is further divided into classic tyrosine-specific PTPs, and tyrosine and serine/threonine dual-specific phosphatases (DUSP), with 38 and 61 members, respectively. Classical tyrosine-specific PTPs are composed of 17 cytosolic PTPs and 21 trans-membrane receptor-like PTPs, and the DUSPs are further divided into MKP, Myotubularin, CDC14, Slingshot, PTEN, PRL, and atypical DSP subclasses. Class II PTP has only one member, the low molecular weight (LMW) PTP, and class III has three members, namely CDC25A, CDC25B, and CDC25C. Class II and class III PTPs also have catalytic cysteine residues. Class IV PTPs have 4 members with catalytic aspartic acid residues, which is in contrast to Class I, II, and III PTPs. Class IV PTPs activities also require the presence of a metal ion, suggesting they have unique evolution origins13.

Figure 1
figure 1

Human PTP classification. The gene encoding individual PTPs is shown in parenthesis after the PTP name. PTPs discussed in this review are highlighted in italic and bold font.

PowerPoint slide

Given the large number of PTPs, a summary of potential drug targets is highly important and will serve as a road map for drug discovery efforts targeting PTPs. Although reviews on similar topics are available7,14, they were published more than ten years ago, during which significant research advances have been made. Thus, this perspective aims to provide an updated view on this subject and a summary of recent research findings. We will cover well recognized and validated PTP targets in the field, including PTP1B, SHP2, LYP, CD45, FAP-1, STEP, MKP-1, PRL, LMWPTP, CDC25 (Figure 2), with a focus on their implications in human diseases.

Figure 2
figure 2

The schematic PTP structure discussed in this review.

PowerPoint slide

PTP1B in diabetes and cancer

PTP1B, encoded by the PTPN1 gene, is a ubiquitously expressed classical non-receptor PTP with 435 amino acids15. It has an N-terminal catalytic domain, two proline-rich sequences, and a C-terminal hydrophobic region (Figure 2). Biochemical and genetic studies have indicated that PTP1B is a key negative regulator of insulin and leptin signaling pathways (Figure 3A), which are important regulators of body weight, glucose homeostasis, and energy expenditure16. PTP1B downregulates insulin signaling by directly dephosphorylating insulin receptor (IR) and insulin receptor substrates (IRS)17,18, while it regulates leptin signaling by dephosphorylating activated JAK2 and STAT319,20. PTP1B antibodies and small molecule inhibitors have been shown to increase insulin-stimulated IR, IRS and STAT3 phosphorylation21,22, suggesting that PTP1B inhibition could sensitize insulin and leptin signaling pathways. Importantly, PTPN1-deficient mice exhibit enhanced insulin sensitivity and leptin hypersensitivity and have lower blood glucose and basal insulin levels. These mice were also more resistant to high-fat diet-induced weight gain23,24.

Figure 3
figure 3

The physiological/pathological signal pathways involving PTP1B (A), SHP2 (B), LYP and CD45 (C), FAP-1 (D), and STEP (E). Arrow represents positive regulation. T-bar represents dephosphorylation if it points to a phosphate group or specific pY, otherwise it represents negative regulation. Dashed lines in panel (D) represent binding interaction. See the text for regulation details.

PowerPoint slide

Decreased insulin sensitivity is a hallmark of type 2 diabetes, which accounts for 90% of diabetes cases. In this regard, PTP1B has been considered a novel drug target for type 2 diabetes and obesity. Many pharmaceutical companies, including Abbott, Novo Nordisk, AstraZeneca, Eli Lilly, Merck, Novartis, Incyte, Wyeth, and ISIS, have shown strong interest in PTP1B25,26. The combined efforts in developing PTP1B inhibitors have generated at least 4 drug candidates in clinical trials, including ertiprotafib, ISIS 113715, ISIS-PTP1BRx, and trodusquemine (Figure 4A)26. Notably, ISIS 113715 has been shown to improve glucose regulation and reduce LDL levels in type 2 diabetes patients in a phase 2 clinical trial. The more potent candidate ISIS-PTP1BRx will replace ISIS 113715 for further development, which aims to help patients whose disease is inadequately controlled by insulin and who are unresponsive to existing oral drugs.

Figure 4
figure 4

The structure of PTP1B (A), SHP2 (B), and LYP (C) inhibitors mentioned in this review.

PowerPoint slide

Given its function in dephosphorylation of receptor tyrosine kinases, which are known to induce oncogenic signalling, PTP1B has also been regarded as a potential tumor suppressor. However, recent studies have revealed that PTP1B can promote tumorigenesis27. For example, PTP1B has been shown to activate c-Src in breast cancer cell lines by dephosphorylating its negative regulatory residue Y53028 (Figure 3A). PTP1B can also activate the Ras-Raf-ERK oncogenic signaling pathway, most likely by dephosphorylating the scaffold protein p62Dok, which binds and activates p120RasGAP29 (Figure 3A). Moreover, PTP1B overexpression has been recorded in 72% of 29 human breast cancer samples examined, in comparison to healthy controls30. Its overexpression was detected in all stages of tumor development, and it correlated with ERBB2 overexpression, a frequently amplified receptor tyrosine kinase in breast cancer.

To understand the interplay between PTP1B and ERBB2 and to define PTP1B's role in breast cancer tumorigenesis, mice expressing an activated ERBB2 gene were crossed with PTPN1-null mice31,32. Compared to control mice, PTPN1-null mice had a significant delay in tumor onset and a decreased rate of lung metastasis. Delayed tumor onset and decreased metastasis were also observed after treatment with a small molecule PTP1B inhibitor. Although the mechanism underlying PTP1B deletion or inhibition in tumor development remains unclear, these data indicate that PTP1B is functionally linked to ERBB2 and plays a positive role in breast cancer tumorigenesis. Recently, PTP1B overexpression was reported in colorectal cancer tissues, and its expression correlated with tumor differentiation, tumor invasion, lymph node metastasis, and TNM stage, which suggest that PTP1B may also play a positive role in colorectal cancer33. Therefore, PTP1B inhibition may be a novel strategy to treat these major human cancers.

SHP2 in cancer

The Src homology-2 (SH2) domain-containing phosphatase 2 (SHP2) encoded by the PTPN11 gene is a 593 amino acid classical non-receptor PTP. It has two tandem N-terminal SH2 domains (N-SH2, C-SH2), a catalytic PTP domain, a C-terminal tail with two tyrosine phosphorylation sites (Y542 and Y580) and a proline-rich region (Figure 2)34,35. SHP2's N-SH2 domain blocks access of SHP2's substrates by binding to its active site pocket at resting state36. However, upon growth factor or cytokine stimulation, the N-SH2 domain preferentially binds to tyrosine-phosphorylated proteins, such as receptor tyrosine kinase or scaffold proteins, to open up the phosphatase active site for catalysis. SHP2 is a positive regulator of the growth factor-mediated Ras-Raf-ERK pathway, and its phosphatase activity is essential for Ras-Raf-ERK pathway activation35. Several presumptive mechanisms have been proposed for its positive effect on ERK activation (Figure 3B), as follows: SHP2 could dephosphorylate the RasGAP binding site on RTK and/or Gab1 to prolong Ras activation37; it could also dephosphorylate CSK binding sites on Paxillin to sequentially activate Src and Ras38; SHP2 may mediate the dephosphorylation of the negative Ras regulator Sprouty to activate the Ras-ERK signaling pathway39,40; finally, SHP2 could act as an adapter in Grb2/SOS complex recruitment, leading to Ras activation41. Moreover, SHP2 has been found to regulate PI3K-AKT, a well recognized oncogenic pathway, and SHP2 can regulate it in a ligand- and cell-dependent manner42,43. In addition, SHP2 has been indicated in JAK/STAT, JNK, and NF-κB signaling, which also have strong associations with various human cancers44.

Clinical studies have shown that SHP2 mutations broadly exist in patients with Noonan Syndrome (NS), juvenile myelomonocytic leukemia (JMML), acute myelogenous leukemia (AML) and solid tumors35,44,45,46. Not surprisingly, many mutations lie between the N-SH2 and PTP domain, disrupting their intramolecular interactions44 and leading to constitutive SHP2 activation. Specifically, SHP2 germ-line mutations are present in 50% of NS patients, and SHP2 somatic mutations are present in 35% of sporadic JMML patients. The high incidence of SHP2 mutations indicate that it is likely a causative gene in these two diseases. Indeed, the SHP2 D61G mutation in mice phenocopies human NS, exhibiting characteristics such as smaller body size, serious cardiac defects, and reduced skull length47. Mice expressing JMML-linked SHP2 mutations (D61Y, D61G) exhibit myeloproliferative disorders similar to those observed in JMML patients, including myeloid expansion, increased myeloid precursors, and granulocyte and macrophage tissue infiltration48. Human JMML characteristics include myeloid colony growth without exogenous cytokine stimulation and bone marrow cell hypersensitivity to granulocyte-macrophage colony stimulating factor (GM-CSF)49. Expression of JMML mutations D61Y and E76K in mouse bone marrow-derived hematopoietic progenitor cells promotes cell cycle progression in the presence of low level GM-CSF and cell survival in minimal media conditions50,51, supporting a positive role for SHP2 in promoting JMML. It was observed that pharmacological SHP2 inhibitors (eg, II-B0852, cryptotanshinone53, and #220-32454 in Figure 4B) blocked the proliferation of patient-derived bone marrow low density mononuclear cells52, mouse myeloid progenitor cells, and leukemic cells53,54 expressing the SHP2-activating mutation E76K. In addition, SHP2 phosphatase activity is required for oncogenic KIT-induced myeloid cell growth and survival, and genetic deletion or pharmacological inhibition of SHP2 can inhibit myeloid cell growth. SHP2 inhibitor alone or in combination with a PI3K inhibitor can prolong the survival of transplanted mice55.

A recent report has shown that SHP2 positively regulates HER2-positive and triple negative breast cancers, and SHP2 knockdown in MCF10A human breast epithelial cells blocks HER2/3-induced tumor cell invasion in a 3-D cell culture model56. Treatment with SHP2 shRNAmiRs can eradicate tumor-initiating cells and block tumor growth and metastasis in relevant xenograft mouse models56. Moreover, in the epidermal growth factor receptor (EGFR) inhibition-resistant lung cancer cell line H1975, SHP2 is required for EGF-stimulated ERK1/2 phosphorylation and cell proliferation. Additionally, SHP2 knockdown or inhibition is sufficient to reduce ERK1/2 activation and block cell growth. This SHP2 inhibitor (II-B08) also has remarkable anti-tumor activity in xenograft mice57. Taken together, these data suggest that SHP2 is a bona fide proto-oncogene and that targeting SHP2 is a promising strategy for various cancer treatments, including AML, JMML, breast cancer, and lung cancer.

LYP in autoimmune disease

The PTPN22-encoded lymphoid-specific phosphatase (LYP) is a classical 807 amino acid non-receptor PTP58. It has an N-terminal catalytic domain and a C-terminal region with 4 proline-rich motifs (P1-P4) (Figure 2). LYP protein expression is restricted to hematopoietic tissues such as thymocytes and mature B cells and T cells. LYP is a strong T cell receptor (TCR) signaling inhibitor, which mediates important immune responses59,60,61. As shown in Figure 3C, LYP dephosphorylates tyrosine residues in the activation loop of the Src family kinases LCK (Y394) and FYN (Y420), and ZAP70 (Y493), which contributes to TCR signaling activation62,63. LYP can also bind C-terminal Src kinase (CSK) through its proline rich P1 region and CSK's SH3 domain64,65,66. The LYP-CSK association might promote LYP's function in inhibiting T cell activation, as CSK can phosphorylate negative regulatory tyrosine residues in LCK (Y505) and FYN (Y531)62,63. LYP-null mice have elevated memory T cell activation and sustained TCR-induced phosphorylation of LCK (Y394) and ZAP70, although naive T cell functions remain normal67. These mice also spontaneously develop germinal centers with increased serum antibody levels. These results are consistent with cellular studies and suggest that LYP negatively regulates T cell development and function.

Importantly, a single-nucleotide polymorphism (SNP) (C1858T) in the PTPN22 gene is associated with type 1 diabetes (T1D) in many populations60,61,68,69. Autoimmune diseases are caused by abnormal immune responses against normal organs or tissues. In this case, T1D is the result of an autoimmune response of cytotoxic CD8 and CD4 positive T cells targeting insulin-producing β-cells. The C1858T LYP SNP has also been linked to other autoimmune diseases, such as rheumatoid arthritis70,71, Graves disease72,73, myasthenia gravis74, and systemic lupus erythematosus75, suggesting that LYP is a general susceptibility gene in both systemic and organ-specific autoimmune diseases.

The C1858T SNP in the PTPN22 gene encodes a mutant R620 LYP enzyme, which is 1.5 times more active than wild-type enzyme76. The mutation falls within the P1 proline-rich region, which disrupts the interaction between the P1 region and the SH3 domain, severely impairing LYP-CSK complex formation64,65,66,76. Interestingly, the R620W mutant more effectively inhibits T cell signaling compared to wild-type LYP in human T cells and Jurkat T cells76, which is contradictory to earlier notions that LYP cooperates with CSK to downregulate TCR signaling62,63. However, recent studies demonstrated that LYP forms a complex with CSK in resting T cells; upon TCR stimulation, LYP dissociates from CSK, resulting in increased LYP partition to lipid rafts, where it dephosphorylates substrates such as LCK and ZAP70 to downregulate TCR signaling77. In contrast, the R620W mutant cannot bind CSK and is therefore recruited to lipid rafts in the resting state. This uncontrolled TCR downregulation produces a gain-of-function phenotype and causes various human autoimmune diseases77. PTPN22 knock-in mice have been generated to express the analogous R619 W mutation78,79. As the mice age, they exhibit effector T cell expansion and transitional, germinal center, and B cell expansion, resulting in autoantibody development and systemic autoimmunity. These mice demonstrate the relevance of the LYP C1858T SNP in the increased risk of autoimmune diseases. Moreover, LYP pharmacological inhibitors (eg, LVT-177, Comp 480, and Comp 8b82 in Figure 4C) are very effective in promoting T cell activation77,80 and enhancing LYP-mediated signaling in thymocytes and bone marrow-derived mast cells81,82. In a passive systemic anaphylaxis mouse model, LYP inhibitors reduce calcium-mediated transcription and degranulation in mast cells and block anaphylaxis, highlighting the great therapeutic potential of LYP inhibitors in autoimmune disease treatment82.

CD45 in autoimmune disease and cancer

CD45 was studied decades ago as a cell surface glycoprotein83,84. After the identification of mammalian PTP1B, sequence comparison indicated that CD45 has two tandem PTP domains in its cytosolic tail. CD45 was then classified as a class I receptor-like PTP85. CD45 is encoded by PTPRC and has a molecular weight between 170 and 220 kDa with a highly glycosylated extracellular region, a single transmembrane domain, and two PTP domains D1 and D2 (Figure 2). The D1 PTP domain is catalytically active, but its activity requires the presence of the inactive D2 domain86. Many CD45 isoforms are generated by alternative splicing, and they primarily differ in their extracellular regions87. CD45 is expressed in the hematopoietic lineage such as T and B cells, and it is estimated to occupy up to 10% of total cell surface area, indicating its abundance in these cells88.

CD45 plays a critical role in T cell receptor and B cell receptor signaling, and Src family kinase LCK and LYN are well-characterized CD45 substrates in T cells and B cells, respectively89,90,91. LCK's kinase activity is required for TCR signaling, and it phosphorylates the TCR complex and activates ZAP70 to initiate a T cell activation cascade (Figure 3C). CD45 dephosphorylates the LCK Y505 residue, which negatively regulates LCK activity when phosphorylated, to positively regulate T cell activation. This is consistent with cellular studies in which CD45-deficient cells have increased levels of LCK Y505 phosphorylation92,93,94. CD45-null mice have T cell defects that can be rescued by an LCK Y505F mutant95. Interestingly, LCK is hyperphosphorylated at site Y394 in CD45 deficient cells89,96,97, which is a positive regulatory residue and enhances LCK kinase activity when phosphorylated. These data suggest that CD45 can also downregulate LCK activity. It has been suggested that Y505 phosphorylation affects T cell signaling at low CD45 levels, while its effect is counterbalanced by Y394 phosphorylation at higher CD45 levels98. Similarly, CD45 regulates LYN in B cells where it dephosphorylates the negative regulatory Y508 residue and the positive regulatory Y397 residue, thus controlling LYN kinase activity99,100,101. B cell development and B cell signaling responses are reduced in CD45-null mice upon ligation, including altered proportion of B cell subsets. Other reported CD45 substrates include FYN and JAK in T cells94,102, HCK and LYN in macrophages103, and LYN, HCK and FYN in dendritic cells104.

Given its important role in immune signaling pathways, CD45 has been linked to many autoimmune diseases105. For example, a SNP in CD45's exon 4 causes a C77G mutation and is associated with increased incidence of multiple sclerosis106,107,108, HIV109, autoimmune hepatitis110, and systemic sclerosis111. The C77G mutation disrupts an exonic splicing silencer, which results in high molecular weight CD45 isoform expression (eg, CD45RA, which has higher CD45 activity in cells due to decreased ability to form homodimers)112. Another SNP in exon 4 causes the C59A mutation and has also been reported to interfere with splicing in several multiple sclerosis multiplex family members, which results in CD45RA expression in memory T cells113. In contrast, a SNP in exon 6 causes the A138G mutation and has a protective effect in hepatitis B infection and autoimmune Graves' thyroiditis. This SNP promotes exon skipping, which results in increased expression of the low molecular weight isoform CD45RO114,115. In addition, the E613R mutation in mouse CD45 is thought to disrupt homodimer formation to cause aberrant CD45 activity, and its expression produces a phenotype characterized by lymphoproliferative syndrome and autoimmune diseases116. These genetic studies indicate a dose-response curve between CD45 and human autoimmune diseases, suggesting that CD45 is a good drug target for autoimmune disease treatment.

CD45 is broadly expressed and has been used in antibody-mediated therapy against hematopoietic malignancies117. Earlier studies have demonstrated that a CD45 antibody killed CD45-positive leukemic cells via antibody-dependent and host-cell-mediated cytotoxicity. However, a more general strategy is CD45-based radioimmunotherapy in which the CD45 antibody is conjugated to a radioactive agent, such that a high dose of radiation is specifically delivered to hematopoietic tissue. In a pilot study118, 131I-coupled anti-CD45 was assessed in 44 patients with high-risk acute leukemia or myelodysplasia for marrow transplantation. Seven out of 25 treated patients with acute myeloid leukemia/myelodysplastic syndrome survived disease-free for 15 to 89 months (median, 65 months) post-transplantation. More recently, 211At-coupled anti-CD45 treatment prior to hematopoietic stem cell transplantation improved the median survival of leukemic mice in a dose-dependent fashion, and 211At-coupled anti-CD45 faithfully localized to the marrow and spleen with minimal toxicity119. In addition, leukemia mouse models treated with anti-CD45 antibody-streptavidin conjugate and subsequent administration of 213Bi- or 90Y-DOTA-biotin conjugate, survived leukemia-free for more than 100 days with minimal toxicity120. These data suggest that anti-CD45-based radioimmunotherapy is a highly effective approach for hematopoietic malignancy treatment.

FAP-1 in cancer

PTPN13-encoded Fas-associated phosphatase 1 (FAP-1, also named PTP1E, PTP-BAS and PTPL1) is a 2485 amino acid classical non-receptor PTP121,122,123. It contains a KIND domain, a FERM domain, five PDZ domains, and a PTP catalytic domain (Figure 2). The FERM domain binds to phosphatidylinositol 4,5-biphosphate, which targets FAP-1 to juxtamembrane regions124, PDZ domains are responsible for protein-protein interactions125, but the KIND domain has not been well-studied. The structure of FAP-1's catalytic domain is very similar to PTP1B in overall folding126,127.

One of the most important discoveries regarding FAP-1 is that it interacts with the cytosolic portion of the FAS receptor128, a tumor necrosis factor (TNF) receptor or death receptor whose activation leads to cell apoptosis, and negatively regulates FAS-initiated apoptosis (Figure 3D). Studies have shown that FAS receptor's C-terminal tripeptide Serine-Leucine-Valine (SLV) sequence is sufficient to bind FAP-1's second PDZ domain129,130, and by doing so, FAP-1 inhibits FAS receptor export to the cell surface131. FAP-1 can also bind the p75NTR intracellular domain, another TNF receptor family member, through the second PDZ domain, and negatively regulate p75NTR-mediated NF-κB suppression and pro-apoptotic signaling132 (Figure 3D). In addition, FAP-1 binds the very C-terminus of the tumor suppressor Adenomatous Polyposis Coli (APC) through its second PDZ domain, and APC is a β-catenin-associated scaffold133. Thus, FAP-1 may indirectly modulate β-catenin tyrosine phosphorylation and regulate cell adhesion, migration, and division (Figure 3D). Other reported FAP-1 binding partners include IκBα, RhoGAP1, EphrinB1, and TRPM2125. IκBα is the only FAP-1 binding protein that is also dephosphorylated by FAP-1, suggesting that it is a putative FAP-1 substrate134,135. IκBα tyrosine phosphorylation is a key element in NF-κB activation; therefore, FAP-1 may regulate NF-κB activation. FAP-1 specifically dephosphorylates IRS-1 and blocks insulin-like growth factor-induced PI3K/AKT signaling pathway136. In addition, FAP-1 interacts with and dephosphorylates HER2 and inhibits growth factor-induced HER2 signaling137. FAP-1-deficient mice are generally healthy but have increased T cell activity such as cytokine elaboration and improved host defense against K pneumonia lung infection, as evidenced by the 35-fold less living bacteria in their lungs compared to wild-type controls138. Independent knockout studies have shown that FAP-1-null mice develop normally but have a mild deficiency in motor neuron repair, and significantly reduced retinal glial numbers in cultures from lens-lesioned knockout mice compared to wild-type controls139,140.

Because FAP-1 negatively regulates FAS-initiated cell apoptosis, it has been suggested to positively regulate tumorigenesis. FAP-1 inhibits FAS-mediated apoptosis in pancreatic adenocarcinoma141,142 and melanoma131, and FAP-1 and FAS expression highly correlate with cell survival in ovarian cancer143, colon cancer144, head/neck cancer145, hepatocellular carcinoma and hepatoblastoma146,147. Ewing's Sarcoma family of tumors (ESFT) is characterized by the formation of chimeric fusion protein EWS-FLI1, which is an oncogenic transcriptional factor that promotes tumorigenesis148. FAP-1 has been identified as a EWS-FLI1 oncogenic fusion protein transcriptional target149. FAP-1 is highly expressed in ESFT cells and patient tumor samples and is required for ESFT transformed phenotype maintenance150,151,152. Reduction of FAP-1 in ESFT cells significantly reduces monolayer and soft-agar cell growth and increased sensitivity to etoposide-induced apoptosis149.

Studies have shown that FAP-1 expression is increased in SW480 colon carcinoma cells after treatment with the chemotherapeutic agent oxaliplatin. siRNA knockdown of FAP-1 reduces cell proliferation and promotes oxaliplatin-induced cell apoptosis, suggesting that FAP-1 inhibition increases oxaliplatin efficacy in colon carcinoma treatment153. Tumor progression shares many characteristics with epithelial-to-mesenchymal transition (EMT), and cells undergoing EMT are more resistant to apoptosis154. The miR-200 family of miRNAs is a fundamental marker and powerful regulator of EMT, as it maintains the epithelial phenotype by suppressing EMT-inducing transcription factor ZEB1 and ZEB2 expression155. miR-200 family members are frequently downregulated during early malignant transformation but are subsequently upregulated in advanced stages in several human cancers156. FAP-1 has been identified as a miR-200c target, and it is responsible for the reduced sensitivity to FAS-mediated apoptosis in cells in miR-200-inhibited cells, providing a mechanism by which cancer cells with reduced miR-200 expression are less sensitive to FAS-mediated apoptosis157

In contrast, there is also evidence that FAP-1 may be a tumor suppressor. For example, reduced PTPN13 mRNA expression by promoter hypermethylation or allelic loss has been observed in gastric and hepatocellular carcinomas158,159, and FAP-1 knockdown enhances PLC5 cell proliferation159. In a large-scale mutational PTP analysis in colorectal cancers, 19 PTPN13 mutations were identified, and 7 of them were in the PTP domain, which may impair FAP-1's catalytic activity160. Nevertheless, with 8 domains, FAP-1 is among the largest intracellular PTPs, suggesting that it has multiple functions and may play positive or negative roles in a context-dependent manner. Thus, FAP-1 is a promising chemotherapy target for irradiation-resistant cancer.

STEP in neurological diseases and disorders

Striatal-enriched protein tyrosine phosphatase (STEP) is encoded by PTPN5 and is a classical and brain-specific PTP, with two alternative spliced isoforms (STEP46 and STEP61). STEP46 is cytoplasmic, while STEP61 localizes to the post-synaptic density and endoplasmic reticulum161,162,163. STEP has a phosphatase domain similar to all other PTPs and a kinase interaction motif (KIM), which is also present in HePTP and PTP-SL (Figure 2). The KIM domain allows STEP to interact with MAPKs, such as ERK and p38, and STEP dephosphorylates tyrosine residues in their activation loops to reduce ERK and p38 activation. ERK activity is significantly higher in the striatum, CA2 region of the hippocampus, and central and lateral nuclei of the amygdala in STEP-null mice. Cultured neurons from STEP-knockout mice have increased ERK phosphorylation upon synaptic stimulation compared to neurons from wild-type controls164. The p38α·STEP complex structure has been solved by NMR and small-angle X-ray scattering data, providing a molecular basis of STEP recognition of p38α165. Notably, ERK is a very important player in synaptic and neuronal plasticity and is an essential component in signaling pathways that regulate behavioral memory formation166, and p38α has been implicated in the neurological disorder pathogenesis167,168.

STEP can mediate DHPG (dihydroxyphenylglycine)-induced AMPAR (α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptor) internalization (Figure 3E), a central process in synaptic plasticity. DHPG stimulation significantly increases STEP protein expression, which is likely responsible for GluR2 (an AMPAR member) dephosphorylation and internalization. STEP-inactive mutants abolish AMPAR internalization169. In STEP knockout mice, GluR1 and GluR2 baseline synaptic expression is increased compared to wild-type littermates, and DHPG-stimulated GluR1 and GluR2 internalization is also abolished in these mice169. STEP can also regulate NMDAR (N-methyl-D-aspartate receptor) (Figure 3E), a glutamate receptor involved in synaptic plasticity and memory function. STEP blocks NMDAR exocytosis by dephosphorylating and inactivating Y420 of FYN kinase170. FYN positively regulates NMDAR exocytosis to neuronal surfaces by phosphorylating Y1472 in NR2B171. Alternatively, STEP can promote clathrin-dependent NMDAR endocytosis by directly dephosphorylating Y1472 of NR2B172,173.

STEP has been linked to many neurological diseases and disorders, including Alzheimer's disease (AD), Huntington's disease (HD), schizophrenia, fragile X syndrome (FXS), hypoxic-ischemic brain injury, depression, and alcohol use disorders174. For example, STEP protein level and activity are upregulated in the prefrontal cortex of AD patients and in the cortex of an AD mouse model. Interestingly, Aβ-enriched medium was sufficient to increase STEP expression and to decrease NR1 and NR2B surface expression in wild-type cultured cortical neurons and cortical slices, but not in STEP knockout cultures. These data indicate that STEP is required for Aβ-induced NMDAR endocytosis173,175. Schizophrenia is a mental disorder with behavioral and cognitive deficits partially due to the disruption of glutamatergic signaling. Schizophrenia patients have significantly higher STEP expression in the postmortem anterior cingulated cortex and dorsolateral prefrontal cortex, and similar observations have been made in mice treated with the psychotomimetics MK-801 and phencyclidine (PCP)176. In contrast, STEP knockout mice are less sensitive to acute and chronic PCP administration in terms of their locomotor and cognitive effects176. Furthermore, several typical and atypical antipsychotic medications for schizophrenia treatment can act through STEP, as antipsychotic treatment of mice induces protein kinase A-mediated STEP61 phosphorylation and inactivation and increased GluN1/GluN2B receptor surface expression176.

It has been suggested that STEP proteasome degradation is blocked in both AD and schizophrenia, leading to STEP accumulation and glutamate receptor internalization, and further contributing to AD and schizophrenia pathogenesis173,176. Alcohol is known to cause cognitive impairment and memory and learning disruption in alcohol abuse disorders, likely through hippocampal NMDAR inhibition. Interestingly, STEP mediates alcohol's inhibitory effects on neuronal NMDAR function. This is further supported by STEP knockout studies in which alcohol-induced NMDAR function and fear-conditioned responses were not observed. However, STEP re-introduction into neuronal cultures and slices restored ethanol-induced biochemical and electrophysiological deficits177. Taken together, STEP is a potential novel drug target for the treatment of AD, schizophrenia, alcoholism and addiction, among other neurological diseases and disorders.

MKP-1 in cancer and neurological disorders

Mitogen-activated protein kinase phosphatases (MKPs) are a group of 11 dual-specificity protein phosphatases. All MKP members are structurally similar with a PTP domain at the C-terminus, a KIM domain for MAPK recognition, and an N-terminal sequence for subcellular localization (Figure 2). As their name implies, MKPs interact with and dephosphorylate pThr and pTyr in MAPK's activation loop, including ERK, JNK, and p38, to downregulate MAPK activity178. MKP-1 was the first identified family member with 367 amino acids, and it is an immediate early gene179,180. It is ubiquitously expressed, but is expressed most highly in the heart, lungs and liver179. Its N-terminal LXXLL sequence allows exclusive MKP-1 nuclear localization181. MKP-1 preferentially dephosphorylates p38 and JNK and to a lesser extent, ERK (Figure 5A)180,182,183,184. Interestingly, MAPK activation can induce MKP-1 transcription upon growth factor stimulation or stress, creating a negative feedback regulation loop between MAPK and MKP-1185. In addition, MKP-1-null mice are viable, fertile and have no phenotypic or histologic abnormalities186.

Figure 5
figure 5

The physiological/pathological signal pathways involving MKP-1 (A) and PRL1/2/3 (B). Arrow represents positive regulation. T-bar represents dephosphorylation if it points to a phosphate group or specific pY, otherwise it represents negative regulation. See the text for regulation details.

PowerPoint slide

MKP-1 is well studied in human cancers because of its direct association with MAPK, which is a key regulator of cell proliferation, differentiation, survival, and apoptosis187,188. MKP-1 overexpression has been detected in many cancers, including colon, prostate, bladder, ovarian, breast, and NSCLC189,190,191,192,193,194. In prostate cancer, MKP-1 expression is inversely correlated with JNK activity and apoptotic marker expression190,192, and MKP-1 overexpression is associated with resistance to FAS-induced apoptosis in the prostate cancer cell line DU145195. In breast cancer cells, MKP-1 overexpression protects against chemotherapy-induced apoptosis, including doxorubicin, mechlorethamine, and paclitaxel. In contrast, siRNA knockdown of MKP-1 sensitizes cells to chemotherapy-initiated apoptosis by increasing JNK activity196. Proteasome inhibition induces MKP-1 expression by decreasing JNK activity. MKP-1 knockdown also increases cell sensitivity to proteasome inhibitors197. In a clinical study with 96 patients, MKP-1 overexpression correlates with likelihood of relapse compared to patients with normal MKP-1 levels198. In addition, MKP-1 is strongly expressed in NSCLC tumor tissues and the H-460 and H-23 cell lines. siRNA knockdown of MKP-1 resulted in a ten-fold increase in cisplatin sensitivity199. Moreover, tumors induced by MKP-1 siRNA H-460 grow slower in xenograft mouse models and show increased cisplatin susceptibility compared to parental cell control tumors199. These results suggest that MKP-1 is an important cancer target that may increase the effects of chemotherapy.

MAPK is associated with memory and learning, neuronal plasticity and development200. As a MAPK regulator, MKP-1 may thus play important roles in many brain functions. MKP1 is a transcriptional c-Jun target, and MKP-1 antagonizes JNK-dependent apoptosis in sympathetic neurons201. MKP-1 overexpression blocks JNK-mediated c-Jun phosphorylation and subsequent sympathetic neuron apoptosis, while MKP-1 knockdown enhances nerve growth factor (NGF) withdrawal-induced death. Loss of MKP-1 results in decreased numbers of superior cervical ganglion neurons at P1 during developmental sympathetic neuron death201. These results suggest that MKP-1 is part of a negative feedback loop to modulate MLK-JNK-c-Jun signaling. Neurotrophin brain-derived neurotrophic factor (BDNF) can induce MKP-1 expression to regulate outgrowth and activity-dependent remodeling of axonal arbors in vivo, which causes spatiotemporal JNK dephosphorylation and its substrates and contribute to microtubule destabilization202. Neurons from MKP-1 knockout mice cannot produce BDNF-induced axon branches. Because axonal arbor formation and maturation is strongly associated with increased synaptic connectivity and are positively regulated by BDNF, MKP-1 may play an important role in synaptogenesis. Furthermore, MKP-1 dysregulation may have deleterious effects on learning and memory, which depend on plasticity203.

MKP-1 is also a key factor in major depressive disorder (MDD) pathophysiology204. In rat and mouse models, hippocampal MKP-1 expression increases upon stress or viral-mediated gene transfer, leading to depressive behaviors. Chronic treatment with antidepressants normalizes stress-induced MKP-1 expression and depressive behaviors204. It was observed that MKP-1 knockout mice are more resistant to stress compared to wild-type littermates, and this effect is mediated by increased ERK activity, consistent with studies that MEK-ERK signaling is essential for antidepressant responses204. Moreover, MKP-1 expression is significantly increased in postmortem hippocampal tissues from 21 individuals with depression compared to samples from 18 healthy age, gender, tissue pH and postmortem interval-matched controls, indicating a direct clinical association between MKP-1 overexpression and depression204. Thus, MKP-1 may also be a novel and promising drug target for the treatment of depression and other mood disorders.

PRLs in cancer

PRLs (phosphates of regenerating liver) are a subclass of dual specific phosphatases with 3 similar members, PRL1, PRL2 and PRL3205. PRLs have an N-terminal PTP catalytic domain, a C-terminal polybasic region, and a prenylation CAAX sequence (Figure 2), which is a unique feature to this subclass and is responsible for PRL plasma or intracellular membrane localization206,207. Given their similarities, PRLs act redundantly in many biological functions (Figure 5B). For example, both PRL1 and PRL3 regulate focal adhesion contact though Src activation, although PRL1 does so by enhancing Y419 phosphorylation in the kinase activation loop208, while PRL3 downregulates the negative Src regulator CSK, leading to a decrease in Y530 autoinhibitory Src phosphorylation209. PRL2 targets p130CAS signaling to regulate Src-independent focal adhesion contact210. PRL1 and PRL3 can downregulate the p53 tumor suppressor by stabilizing MDM2 and increasing PIRH2 transcription211,212, whereas PRL2 regulation of p53 has not been elucidated. In addition, PRL1 and PRL3 can both activate and promote matrix metalloproteinase (MMP) expression, extracellular secreted proteins with important roles in tumor metastasis. PRL1 upregulates MMP2 and MMP9 through Src and ERK activation213, whereas PRL3 upregulates MMP2 and MMP7, but not MMP9, via integrin β1 and ERK pathways214.

Given their role in regulating p53 and MMPs, PRLs may play important roles in human cancer pathophysiology205,215,216. For example, PRL1 is overexpressed in lung and pancreatic cancer cell lines with increased invasive properties, which are reversed by PRL1 knockdown208,213,217,218. PRL2 is overexpressed in pancreatic, breast and lung cancer, as well as pediatric AML. PRL2 expression levels are associated with tumor progression and poor prognosis210,217,219,220. In addition, PRL3 overexpression is widely found in cancers, including colon, breast, gastric, oral, cervix, and ovarian carcinomas, multiple myelomas and AML215,221. In colon cancer, PRL3 overexpression positively correlates to poor prognosis and liver metastasis222. Studies have shown that stable PRL3 expression increases cell motility and invasiveness and induces tumor formation and metastasis in mouse models223. In contrast, PRL3 siRNA knockdown blocks cell motility and metastasis, suggesting that PRL3 is a potential cancer treatment target224,225. PRL3 can upregulate PDGFR, Eph, and integrin receptor, as shown in a proteomic analysis of PRL3-overexpressing HEK293 cells.

Phosphoproteomic data support the intracellular activation of an extensive signaling network normally governed by extracellular ligand-activated transmembrane growth factor, cytokine, and integrin receptors in PRL3-overexpressing cells226. Similarly, PRL3 induces EGFR hyperactivation and its downstream signaling pathways in multiple human cancers, contributing to cell growth, migration and tumorigenicity. PRL3-overexpressing cancer cells are highly sensitive to oncogenic EGFR signaling, as they are sensitive to EGFR inhibiting treatment227. More recent studies of genetically modified mouse models demonstrate that PRL2 is required for several developmental processes (placenta, spermatogenesis, hematopoietic stem cell self-renewal) and reveal a novel mechanistic connection between PRL2 and PTEN228,229,230. Given the strong cancer susceptibility to subtle variations in PTEN levels, PRL2's ability to repress PTEN expression qualifies it as an oncogene and a novel target for anti-cancer agents. Taken together, PRLs are of great therapeutic importance as clinical predictive biomarkers for personalized medicine and novel drug targets for high metastatic potential cancers.

LMWPTP in cancer and diabetes

Low molecular weight protein tyrosine phosphatase (LMWPTP) has a single catalytic domain (Figure 2) with an approximate molecular weight of 18 kDa. It is the sole class II PTP member, and its overall sequence and 3D structures are different from the aforementioned higher molecular weight PTPs, with the exception of the conserved PTP signature motif (P-loop). LMWPTP is well conserved across many prokaryotic and eukaryotic species, suggesting that it may play fundamental roles throughout evolution231. There are 4 human LMWPTP isoforms, and 2 are catalytically active, namely IF1 and IF2 (or LMWPTP-A and LMWPTP-B)232,233. These two isoforms have very high sequence and structure similarity, and the minor difference exists in a flexible loop flanking the active site, which is thought to determine their physiological substrate specificity234. Our knowledge of human LMWPTPs is primarily from the study of IF1.

LMWPTP regulates many receptor tyrosine kinase and growth factor-induced signaling pathways, which control cell growth, differentiation and adhesion (Figure 6A). For example, phosphorylated PDGF receptor is a well-known LMWPTP substrate dephosphorylated on Y857 in its β subunit235,236,237, which is a regulatory residue for PDGF receptor kinase activity. LMWPTP also dephosphorylates Src kinase and STAT family transcription factors, and block their PDGF receptor-induced activation238,239,240. In addition, LMWPTP can dephosphorylate and inactivate FAK241 and p190RhoGAP242, which are involved in cell-extracellular matrix adhesion. More importantly, EphA2 is a preferred LMWPTP substrate in tumor cells243,244,245. EphA2 is a receptor tyrosine kinase overexpressed in many human cancers, especially in aggressive and metastatic types243. In contrast to PDGF receptor, EphA2 phosphorylation inhibits cell growth and migration, while its dephosphorylation induces oncogenic characteristics246,247. LMWPTP overexpression in epithelial cells induces colony formation in soft agar and promotes neoplastic transformation243. LMWPTP-transfected NIH3T3 fibroblasts can induce larger fibrosarcomas in nude mice with higher proliferation activity and greatly dephosphorylated EphA2 compared to mock-transfected controls. The opposite was true in dominant negative/inactive LMWPTP-transfected cells244. Because LMWPTP is overexpressed in many tumor types, LWMPTP and EphA2 may form an oncogenic axis to promote tumor initiation, progression and metastasis.

Figure 6
figure 6

The physiological/pathological signal pathways involving LMWPTP (A) and CDC25 (B). Arrow represents positive regulation. T-bar represents dephosphorylation if it points to a phosphate group or specific pY, otherwise it represents negative regulation. See the text for regulation details.

PowerPoint slide

LMWPTP is also a key negative regulator of insulin signaling through tyrosine dephosphorylation of the insulin receptor (IR) β subunits248. A study from Eli Lilly demonstrated that decreased LMWPTP expression by a specific antisense oligonucleotide (ASO) increases IR, PI3K, and AKT phosphorylation, thus sensitizing insulin signaling249. ASO treatment resulted in decreased plasma insulin, glucose, triglyceride and cholesterol levels, with improved glucose and insulin tolerance in mice249. Interestingly, LMWPTP inhibition primarily affects liver and adipose tissues, in contrast to PTP1B, which regulates insulin signaling in liver and skeletal muscle. These data suggest that PTP1B and LMWPTP have different tissue specificity in insulin signaling regulation249. Epidemiological studies have indicated that LMWPTP levels positively associate with serum glucose and triglyceride concentrations, and the highest LMWPTP levels have been recorded in diabetic and aged patients250. In addition, clinical studies have demonstrated that genotypes with elevated LMWPTP IF1 isoforms have the highest diabetic retinopathy incidence, a common but severe diabetic complication251. These findings indicate that LMWPTP is also a novel type 2 diabetes drug target.

CDC25 in cancer

The CDC25 phosphatases are dual-specificity phosphatases comprised of 3 members, CDC25A, CDC25B, and CDC25C, encoded by genes on 3 different chromosomes252,253,254,255. CDC25A, CDC25B, and CDC25C have 423 to 566 amino acids with a conserved C-terminal PTP catalytic domain and a non-conserved N-terminal regulatory domain for various posttranslational modifications, including phosphorylation, ubiquitination, proline isomerization, and 14-3-3 sequestration256. The CDC25 family plays critical roles in cell cycle control by dephosphorylating negative regulatory residues of cyclin-dependent kinases (CDKs) (Figure 6B), which are important elements of cell cycle progression257. Specifically, CDC25A plays a significant role in the G1-S transition by activating the cyclin A and cyclin E-CDK2 complexes258,259,260; CDC25B partially activates the centrosomal cyclin B-CDK1 complex during the G2 to mitosis transition, and nuclear CDC25C further activates the complex during mitosis261,262,263.

Dysregulated cell cycle control is a hallmark of cancer cells to gain a growth advantage. Thus, the CDC25 family has gained much attention as potential cancer drug targets256,264. Studies have shown that CDC25 causes cell cycle deregulation and consequent cell genome instability, which contributes to cancer initiation and progression. For instance, CDC25A and CDC25B can transform normal mouse embryonic fibroblasts into oncogenic foci in soft agar and tumors in nude mice with oncogenic Ras or loss of RB1265. CDC25B expression enhances mammary epithelial cell proliferation and induces precocious alveolar hyperplasia266, and further increases tumor onset susceptibility upon carcinogen exposure (9,10-dimethyl-1, 2-benzanthracene, DMBA)267. CDC25A, CDC25B, and CDC25C have been found to be overexpressed in a number of human cancers, including breast, colon, gastric, lung, colorectal, pancreatic, neuroblastoma, head and neck, and non Hodgkin's lymphoma264, and their overexpression occurs at all stages of tumorigenesis, suggesting that they may play critical roles throughout tumor development. Various small molecule CDC25 inhibitors have been developed and possess anti-cancer properties. For example, the IPSEN research laboratory developed BN82685 to inhibit CDC25A, B, C with IC50 values of 250, 250, and 170 nmol/L, respectively268. This compound blocks proliferation across a large panel of human cells, as well as tumor growth in xenograft mice with the pancreas carcinoma cell line Mia PaCa-2268. Therefore, the CDC25 family of phosphatases represent novel and attractive cancer targets.

Strategies to modulate PTP activity

There are several ways to modulate PTP activity, and the most popular strategy is to design active site directed and non-hydrolyzable pTyr mimetic-based small molecules. This technique has successfully produced potent inhibitors against several PTPs7,25. However, the major challenges of this strategy are specificity due to conserved PTP structure, and cell permeability due to the positively charged PTP active sites. The cell permeability and inhibitor specificity are important considerations in PTP-based drug discovery because PTP domains are cytosolic and many PTPs may function as tumor suppressors and should not be inadvertently targeted. The recent discovery of cell permeable pTyr mimetic salicylic acid-based inhibitors have achieved appreciable success in addressing these challenges269.

Targeting allosteric PTP1B sites is a good strategy in the development of high specificity and cell permeability inhibitors270. Allosteric inhibitors are not required to have multiple negative charges and therefore may possess more favorable pharmacological properties. However, it still seems difficult to improve their potency, as allosteric sites are not as well-defined as the active site.

Another strategy is to develop mechanism-based inhibitors, which inactivate the PTP catalytic cysteine residue through oxidation or other modifications. This strategy has been employed in quinone-based CDC25 inhibitors11. However, quinone compounds may induce reactive oxygen species, which may inactivate various redox-sensitive enzymes and cause in vivo toxicity271,272. Although they seemingly possess great activities against PTPs in vitro, they are not appealing for drug development. Irreversible and selective kinase inhibitor development has generated renewed interest in covalent enzyme inhibitors273,274. This can be achieved by building a Michael acceptor into a known competitive and specific inhibitor, which then forms a covalent bond with an adjacent cysteine residue. Thus, it will be interesting to employ a similar approach to develop irreversible and specific PTP inhibitors. Unlike quinone-based PTP inhibitors, these irreversible inhibitors would not target all redox-sensitive enzymes in cells.

In addition, ASOs against PTP1B and LMWPTP have been reported249,275,276. Because they are 20-mer oligonucleotides that target PTP mRNA, they are highly specific. PTP1B and LMWPTP ASOs have been reported to increase insulin sensitivity, reduce body glucose level, and control weight gain in animal studies and clinical trials, indicating their therapeutic potential. siRNAs may be an alternative approach in targeting PTPs through similar mechanisms. Nevertheless, challenges include low target delivery efficiency and safety of long-term consumption.

Concluding remarks

Protein-tyrosine phosphorylation dysregulation is a major cause of various human diseases, and great success has been achieved in PTK target drug discovery3. PTPs represent promising next generation drug targets for decades to come. For instance, PTP1B and LMWPTP are well-known negative regulators of the insulin signaling pathway and are excellent drug targets for type 2 diabetes and insulin resistance. SHP2, PRLs, LMWPTP, CDC25, CD45, FAP-1, and MKP-1 play positive roles in tumorigenesis by distinct mechanisms, and they can be targeted to combat a number of human cancers or to sensitize cancer chemotherapy or radiation therapy. LYP and CD45 are exciting drug targets for autoimmune diseases, such as type 1 diabetes, rheumatoid arthritis, and systemic lupus erythematosus. As a brain-restricted PTP, STEP is an emerging target for neurological disease and disorder treatment, such as Alzheimer's disease, Huntington's disease, schizophrenia, and alcohol abuse. In addition, MKP-1 may be relevant in depression and may be a novel target for this common and debilitating neurological illnesses.

Although there is growing recognition of PTPs as promising therapeutic targets, there are no commercialized PTP inhibitors despite extensive pharmaceutical industry efforts. The reasons for this lack of success in PTP-based drug discovery are both historical and technological. Historically, PTP research and development has run approximately 10 years behind PTK research because PTKs were discovered earlier and because PTPs have been erroneously viewed as exclusive negative regulators of hyperactive pathways in disease. Prior to the Human Genome Project, it was commonly believed that PTPs constituted a small group of 'housekeeping' enzymes with functions too broad and conserved to be selectively therapeutically inhibited. It is now known that there are actually more PTPs (107) than PTKs (90), and accordingly, many PTPs have unique and non-redundant functions in health and disease, making them attractive drug targets.

Unfortunately, PTPs present several key challenges to traditional drug development. The PTP active site is highly conserved, so it is not trivial to produce compounds that can selectively inhibit single PTPs. This is an issue common to most target families that act upon common substrate motifs (such as pTyr for PTPs or ATP for kinases) and has been overcome in the kinase field by high throughput screening and structure-based drug design. A more serious issue in PTP-based drug discovery is that the PTP active site is highly positively charged and contains a conserved catalytic cysteine residue. Thus, high-throughput screening of large compound collections usually leads to initial hits that are strongly negatively charged or contain oxidizing groups that irreversibly react with the catalytic cysteine. Heavily charged molecules do not readily cross cell membranes and cannot be made into drugs in their native form, and chemically-reactive compounds (eg, quinones) have poor safety and selectivity profiles, making them unappealing drug candidates. Despite these challenges, a new, focused, pragmatic approach to PTP inhibitor drug discovery and development is required. Recent studies using fragment-based and structure-guided approaches to target the PTP active site and adjacent less-conserved pockets demonstrate that it is feasible to obtain PTP inhibitors with high affinity, selectivity, and excellent in vivo efficacy in animal models of oncology, diabetes/obesity, autoimmunity, and tuberculosis277. Further work will advance the lead generation paradigm and create a 'PTP-based drug discovery platform' that will ultimately broadly impact drug development of tomorrow.