Protein tyrosine phosphatases: dual-specificity phosphatases in health and disease


R. Pulido, Centro de Investigación Príncipe Felipe, Avda Autopista del Saler, 16-3, 46013 Valencia, Spain Fax: +34 96 3289701 Tel: +34 96 3289680
R. Hooft, Merck Serono International S.A., 9, chemin des Mines, 1202 Geneva, Switzerland Fax: +41 224149771 Tel: +41 224149602


Dual-specificity phosphatases (DSPs) constitute a subfamily of protein tyrosine phosphatases (PTPs) that dephosphorylates phospho-Tyr, phospho-Ser and nonproteinaceous substrates. DSPs are involved in the regulation of both developmental and postnatal essential processes, such as early embryogenesis, placental development and immune responses. Several DSP genes are implicated in familial and sporadic human diseases, including tumor-related, neurological and muscle disorders, and cardiovascular and inflammatory diseases. This association ranges from disease-causative mutations to disease-risk-prone single-nucleotide polymorphisms, promoter methylation or gene duplication (most often in cancer). Deconvolution of the role of DSPs in disease is challenging. The enzymes’ activities are regulated at many levels and they form part of extensive, intricate networks with other signaling components. Here, we review current knowledge of the role of cysteine-based PTP-domain DSPs in health and disease, and their suitability as putative therapeutic targets for drugs is discussed.


Charcot–Marie–Tooth disease type 4B


dual-specificity phosphatase


extracellular signal-regulated kinase


fibroblast growth factor receptor




c-Jun N-terminal kinase


Lafora bodies


Lafora disease




mitogen-activated protein kinase


MAP kinase phosphatase




protein tyrosine phosphatase


tumor necrosis factor


X-linked myotubular myopathy

Mammalian class I cysteine-based dual-specificity phosphatases (DSPs) constitute a broad family of protein tyrosine phosphatases (PTPs), both in number and diversity. PTP domain class I DSPs differ from classical PTPs in that they can dephosphorylate phospho-Tyr, phospho-Ser and phospho-Thr residues, as well as nonproteinaceous substrates, including signaling lipids and complex carbohydrates. DSPs are nontransmembrane proteins (with the exception of TPIP and TPTE) that contain a single catalytic domain, and their modular structure ranges from the ‘minimal PTP core’ of the small atypical DSPs to the multimodular structure of some large myotubularins. DSPs can be clustered in several groups that include MAP-kinase phosphatases (MKPs), myotubularins (MTMs), PTEN-related phosphatases, slingshots (SSHs), PRLs, CDC14s and a heterogenous group named atypical DSPs [1]. The involvement of some of these DSPs in human disease is well established, because their genes (PTEN, MTMs, laforin) are targeted by mutations in sporadic tumors or in the germline of patients with hereditary disorders. In addition, recent findings indicate that many other DSPs may be associated with human disease because of pathological alterations in their expression patterns and/or functional activities. The growing information on DSP structure, specific inhibitory compounds and physiological function in mammalian animal models may be exploited to target-specific DSPs in particular diseases and/or subsets of patients. Here, we summarize current perspectives on the use of DSPs as therapy-related targets in human disease. We focus on DSPs whose expression and/or function is altered, or whose genes are mutated, in human disease. Also, some putative disease-relevant DSPs, as inferred from knockout mice models, are reviewed (see Table 1 and Fig. 1 for a summary). Detailed information on the biochemical and biological properties of the distinct DSP families has been provided in recent reviews [2–9].

Table 1.   Functional properties and relation with human disease of DSPs. EAE, experimental autoimmune encephalitis; RA, rheumatoid arthritis; XLMTM, X-linked myotubular myopathy.
Gene namePhosphatase nameOther namesSubstrateAssociated with disease/ target for diseasePhenotype of KO mouseChromosomal locationAccesion
DUSP1MKP-13CH134, PTPN10, erp, CL100, hVH1p38, JNK, ERKOverexpressed in many cancersIncreased innate immune response; resistant to diet-induced obesity5q35.1NM_004417
DUSP6MKP-3PYST1, rVH6ERK,p38Often suppressed in cancersExcess mortality in homozygotes; skeletal deformations linked to excess FGFR signaling12q22-q23NM_001946
DUSP10MKP-5 JNK, p38KO animals protected from EAEProtected from experimental autoimmune encephalitis1q41NM_007207
DUSP2Pac-1 ERK, p38KO animals protected from RAPartial immunodeficiency; protected from rheumatoid arthritis2q11L11329
DUSP26DUSP26MKP-8p38, JNK, ERKCancerNot available (N/A)8p12NM_024025
PTP4A1PRL-1Phosphatase of regenerating liver, PTPCAAX1UnknownCancerN/A6q12U48296
PTP4A3PRL-3PRL-RUnknownCancer; metastasisN/A8q24.3NM_032611
MTM1MTM1MyotubularinPI(3)P, PI(3,5)P2X-linked myotubular myopathy (XLMTM)Muscle myopathy; XLMTM-likeXq27.3-q28NM_000252
MTMR1MTMR1 PI(3)P, PI(3,5)P2Congenital myotonic dystrophy (cDM1)N/AXq28NM_003828
MTMR2MTMR2 PI(3)P, PI(3,5)P2Charcot–Marie–Tooth disease type 4B1 (CMT4B1); male infertilityAbnormal myelination; CMT4B-like; impaired spermatogenesis11q22NM_016156
MTMR13MTMR13SBF2, Set-binding factor 2Enzyme inactive; required for MTMR2 activityCharcot–Marie–Tooth disease type 4B2 (CMT4B2)Abnormal myelination; CMT4B-like11p15.4NM_030962
MTMR5MTMR5SBF1, Set-binding factor 1Enzyme inactiveMale infertility; cancerImpaired spermatogenesis22q13.33U93181
EPM2ALaforin Complex carbohydrates; GSK3βLafora disease (progressive myoclonus epilepsy)Neurodegeneration; Lafora disease-like6q24AF284580
PTENPTENMMAC-1, TEP1PI(3,4,5)P3CancerEarly embryonic lethality in homozygotes; high incidence of cancer and autoimmunity in heterozygotes10q23.3U92436
Figure 1.

 DSPs in human disease. (A) Putative role of class I cysteine-based DSPs in human disease. The involvement of MKP-1/DUSP1, MKP-3/DUSP6, MKP-5/DUSP10 and Pac-1/DUSP2 in human disease is mainly based on the phenotype of knockout mice. The involvement of MKP-8/DUSP26 and PRLs in cancer is based on the high expression and function of these DSPs in tumor samples and/or tumor cell lines. (B) Class I cysteine-based DPS mutated in human disease. The functional properties altered in disease of MTMs, laforin, and PTEN are indicated. In bold are shown the major diseases caused by mutations in the DSP genes.


Mitogen-activated protein-kinase phosphatases are direct inactivators of effector kinases in mitogen-activated protein kinase (MAPK) cascades, namely extracellular signal-regulated kinases (ERKs), c-Jun N-terminal kinase (JNKs), and p38s. MKPs possess an N-terminal regulatory domain that contains a kinase interaction motif and binds to MAPKs, and a catalytic DSP C-terminal domain. Binding to MAPKs through the N-terminal domain provides a mechanism of substrate specificity and allosteric activation for most MKPs. Connecting the regulatory and catalytic domains, all MKPs have a linker region of variable amino acid sequence and size. This region is of unknown function, and in the case of some MKPs contains a nuclear exclusion signal that keeps these enzymes in the cytoplasm, whereas other MKPs reside mostly in the cell nucleus [4,7,10,11].

MKP-1/DUSP1 is the earliest studied DSP. The cDNA for this gene was initially cloned because of its inducibility by serum stimulation and peroxide-induced stress in human fibroblasts, and found to be similar to Vaccinia virus H1 phosphatase [12]. It appears to protect cells against oxidative stress [13,14] and heat shock [15,16]. The gene was named MKP-1 when it was found to dephosphorylate p38 [17]. For example, Notch-induced MKP-1 dephosphorylates p38 in C2C12 cells, a process that can be prevented by MKP-1 siRNA [18], and part of dexamethasone’s anti-inflammatory activity depends on MKP-1 induction and p38 inactivation [19–22]. Interestingly, active p38 post-transcriptionally reduces MKP-1 expression in a feedback loop [23], possibly exploiting MKP-1’s rapid protein turnover [24], for MKP-1 protein expression is readily induced by proteasome inhibitors [25,26]. The substrate specificity of MKP-1 for p38 and JNK, but not ERK, is similar to MKP-5 (see below), but MKP-1 is predominantly nuclear, with translocation mediated by its N-terminal LXXLL motif [27], whereas MKP-5 is also expressed in the cytoplasm [28]. Absence of MKP-1 results in enhanced sensitivity and lethality to lipopolysaccharide (LPS) toxin [29] with significantly increased serum tumor necrosis factor (TNF)-alpha, interleukin (IL)-6, IL-12, MCP-1 and interferon-γ, but also the anti-inflammatory IL-10 [30]. It was proposed that the anti-inflammatory activity of IL-10 requires MKP-1 [31]. In addition to a role in innate immune responses, MKP-1 also affects adaptive responses, as demonstrated by enhanced severity in a collagen-induced disease model for arthritis in knockout animals [32]. Surprisingly, live bacteria (Staphylococcus aureus) caused no excess mortality in MKP-1 knockout animals [33]. The profound role of MKP-1 in LPS responses was also demonstrated by genome-wide transcriptional profiling of spleen cells from LPS-treated wild-type and knockout mice [31]. While LPS induced 608 genes in both type of animals, nearly three times as many extra genes were induced in wild-type versus knockout animals. However, it is difficult in this type of experiment to distinguish between genes that are directly downstream of MKP-1 and its substrate MAP kinases, and those that are induced in a second wave following the release of proinflammatory cytokines.

A single study looked at MKP-1 knockout animals in energy metabolism [34]. The animals showed a complex phenotype in this respect, with increased energy expenditure and resistance to obesity, yet were not protected against obesity-induced glucose intolerance when fed on diet that was rich in fat. MKP-1 has also been studied in the context of cancer. The gene is overexpressed in nonsmall cell lung cancer [35], hepatocarcinoma [36], oral cancer [37] and gastric carcinomas [38], and may prevent apoptosis in rat mesangial cells [39]. In another study for hepatocellular carcinoma, however, reduction of MKP-1 expression was correlated with poor patient outlook [40], and MKP-1 is downregulated in prostate cancer [41]. In patients with colon cancer, MKP-1 expression negatively correlates with recurrence-free intervals [42]. Moreover, resistance to the anti-tumor agent cis-platin-induced apoptosis is mediated through transcriptional upregulation of MKP-1 [43]. These results suggest that MKP-1 inhibitors may have a use in cancer.


MKP-3/DUSP6 cDNA was identified as a rat DSP that was strongly induced in NGF-1-induced PC12 cells, and whose overexpression reduces ERK-2 phosphorylation [44,45]. MKP-3 associates with its preferred substrate ERK-2 [46–48] through its N-terminal domain [49], as well as with protein kinase CK2α [50]. ERK-2 is inactivated by relatively few phosphatases: MKP-3, He-PTP-related phosphatases, DUSP3/VHR, PP2A [51,52]. MKP-3 is mainly cytoplasmic by virtue of a leucine-rich nuclear export signal in its linker domain [53]. MKP-3 transcription is induced by fibroblast growth factor receptor (FGFR) activation [54]. MKP-3 forms an expression gradient along the chick limb bud to modulate FGF-8 responses [55,56]. In zebrafish, MKP-3 control of the Ras-MAPK signaling arm downstream of the FGFR is important for axial polarity during development [57], and MKP-3 (named DMKP-3) [58] is also involved in Drosophila development [59]. Li et al. [60] showed that mice lacking MKP-3 have excess perinatal mortality and developmental defects with skeletal dwarfism, coronal craniosynostosis and deafness due to otic ossicle malformation – all likely associated with enhanced FGFR signaling.

In cancer, MKP-3 may represent a tumor suppressor whose inactivation is frequently associated with pancreatic and other cancers [61,62], mediated by promoter hypermethylation [63] or chromosomal loss [64–66]. Its expression was abrogated exclusively in invasive carcinoma cells [67]. By contrast, the gene is member of a five-gene panel whose altered expression (an increase in the case of MKP-3) is predictive of poor patient outlook in nonsmall-cell lung cancer. In this context, its inactivation of ERK-2 is believed to prevent tumor suppression and apoptosis [68]. By contrast, overexpression of MKP-3 reduces tumor growth [69].

Like many other phosphatases, MKP-3 is potentially regulated by redox control [70]. In neurons, glutamate toxicity is mediated through glutamine depletion and oxidative stress. Levinthal et al. [71] demonstrated that this effect is directly mediated through redox inactivation of MKP-3. Interestingly, overexpression of a dominant-negative MKP-3 protein resulted in accumulation of activated phospho-ERK-2, but due to its stable association with the phosphatase, ERK-2 was unable to enter the nucleus.


MKP-5/DUSP10 was cloned by virtue of its binding to p38 in a yeast two-hybrid experiment. In vitro substrates are p38 and JNK, but not ERK [72]. This MKP contains a 150-amino acid N-terminal MAPK-binding domain [73] that binds to p38 and is required for efficient p38 dephosphorylation, followed by two CDC25-like domains. Unlike many other DSPs, MKP-5 assumes a configuration that predicts constitutive enzymatic activity [74]. The gene is well-conserved among mammals and is ubiquitously expressed [75]. Like many DSPs its expression is highly regulated: it is induced by TNF-α, anisomycin and osmotic stress, but not by UV irradiation or phorbol ester [72].

Recent work has shown that calcitriol, the hormonally active form of vitamin D [76], vitamin D itself [77], curcumin, resveratrol and [6]-gingerol [78] also increase MKP-5 expression, resulting in inactivation of p38. This mode of action may partly explain the chemopreventive activity of these agents in prostate cancer, but how MKP-5 transcription is controlled by these agents remains unknown. Along these lines, MKP-5 was also identified as a tumor-suppressor gene whose promoter is hypermethylated and whose expression is reduced in mantle cell lymphoma [79]. A recent study identified a leukemia patient whose lymphoma contained a chromosomal deletion that resulted in the predicted expression of a fusion protein of MKP-5 and MEL1/PRDM16 [80]. It is currently unclear whether unchecked cell proliferation in this case was linked to lack of normal MKP-5 expression, like in mantle cell lymphoma, or to loss of normal MEL1 expression, whose deletion was observed in other lymphomas, or has a different cause.

Another study [81] focused on JNK as a key substrate for MKP-5, following the observation that MKP-5 is the best mammalian homolog for pluckered, which is the main JNK phosphatase in Drosophila. Both Th1 and Th2 cells that lack MKP-5 show enhanced JNK, but unchanged p38 or NF-κB activity. These and other results by the same authors indicate that absence of MKP-5 results in an enhanced innate immune response. Accordingly, knockout animals responded with increased TNF-α production and died more frequently from secondary LCMV infection, but in a myelin antigen-induced disease model for multiple sclerosis, knockout animals showed reduced incidence and severity for experimental autoimmune encephalitis. Interpretation of these results is not straightforward. Although p38 is a good MKP-5 substrate in vitro, p38 activation in cells was little affected by absence of MKP-5, presumably because of increased phosphorylation of JNK, whose activation is known to repress p38. Possibly, overall resistance to experimental autoimmune encephalitis is due to reduced T-cell proliferation, despite enhanced innate immune responses, although how lack of MKP-5 alters the balance of cellular responses is, at this point, not yet quite clear. Both for Pac-1 (see below) and MKP-5, the biological effects seen appear directly related to the enzymes’ catalytic activities, suggesting, a priori, that multiple sclerosis patients may benefit from enzymatic inhibition of these enzymes.


Phosphatase of activated cells 1 (Pac-1), was originally cloned by low-stringent hybridization of a genomic library [82] then shown to be a nuclear DSP that is predominantly expressed in hematopoietic tissues [83]. Evidence was provided that Pac-1 dephosphorylates and deactivates nuclear MAPKs [84], in particular ERK-2, whose association results in catalytic activation of Pac-1 [85], as is the case for MKP-1 and MKP-3. The role of Pac-1 in the immune system has not been much studied. The gene shows strong upregulation in activated human primary leukocytes [86]. In this study, Pac-1 stood out as being strongly induced in dendritic cells, macrophages, polymorphonuclear cells, umbilical cord-derived mast cells, Th1 and Th2 cells and eosinophils, following induction with lipopolysaccharide, phorbol 12-myristate 13-acetate or cross-linking of CD3 or IgE cell-surface receptors. This induction was also seen at the protein level; in addition, human rheumatoid arthritis synovium and tonsils were shown to be infiltrated with Pac-1-expressing lymphocytes. Crucially, mice homozygous for a disrupted Pac-1 gene, although developing normally, were protected from serum-induced arthritis. Macrophages and mast cells from Pac-1−/− mice displayed impaired induction of IL-6, CXCL5, COX-2, TNF-α and other inducible genes when stimulated with LPS or IgE. Western blot analysis showed that this subdued response was accompanied by reduced ERK and p38 activation (phosphorylation). By contrast, JNK phosphorylation was increased in stimulated cells that lack a functioning Pac-1 gene. It was proposed that reduced ERK phosphorylation in absence of Pac-1 is mediated by JNK activation, since a JNK inhibitor restored ERK activation in Pac-1−/− cells. Clearly, this is only part of the puzzle. It is unclear how ERK and p38 escape dephosphorylation by Pac-1 itself, nor is understood how absence of Pac-1 expression results in enhanced JNK phosphorylation. Overall, the data indicate that Pac-1 may be a good drug target for arthritis.


MKP-8/DUSP26 is a recently identified atypical DSP that localizes to the nucleus and has substrate specificity for p38 and, to lesser extent, ERK [87]. It was shown to bind heat shock transcription factor 4 in a complex with ERK, controlling transcription factor DNA binding [88]. MKP-8 is highly expressed in embryonal cancers (retinoblastoma, neuroepithelioma, and neuroblastoma) [87]. It has been known for many years that chromosomal region 8p11-12 is amplified in breast, urinary bladder, lung and ovary cancers. Yu et al. [89], focusing on a set of anaplastic thyroid cancer tumors, provided evidence that MKP-8 is a key gene in this chromosomal region, and that overexpressed MKP-8 maintains p38 in an inactive state and prevents caspase 3-mediated apoptosis. Blocking MKP-8 expression using siRNA reduces tumor cell proliferation. Together, these results clearly point to MKP-8 as a potential target for this and perhaps other types of cancer. The fact that tumoral overexpression is achieved through genomic DNA amplification strongly suggests that MKP-8 is a major driver, rather than a secondary player in tumor growth and behavior. However, no knockout animals for the gene have yet been described and the effects of systemic inhibition of MKP-8 are not yet known.


The PRL phosphatases (PRL-1, PRL-2, PRL-3) are a subfamily of small-size DSPs of wide tissue distribution, whose expression is increased in many different cancer cell lines and tumor tissues [6]. PRLs are active phosphatases that contain a C-terminal CAAX prenylation motif (an exclusive feature in the PTP family), but the nature of their physiologic substrates is unknown. Closer relatives of PRLs, both at the amino acid sequence and structural levels, are the cell-cycle regulators Cdc14 DSPs. PRLs also show high structural similarity to the small atypical DSPs and to the catalytic domain of the lipid phosphatase PTEN [90–93]. However, the large and highly hydrophobic catalytic pocket of the PRL-1 active site differs from the catalytic pockets of its related DSPs, suggesting a unique substrate specificity for PRLs. Catalytically active PRL-1 and PRL-2 facilitate progression through the cell cycle, possibly by decreasing the levels of the cyclin-dependent kinase inhibitor p21Cip1, and farnesylation of PRL-1 has been found to be important for mitosis. In mitotic cells, PRLs are located in the mitotic spindle, whereas in nonmitotic cells PRLs are associated with the plasma membrane, endoplasmic reticulum or endosome membranes [94,95]. This cell-cycle-dependent compartmentalization may control PRL access to substrates and/or regulators. The PRL-1 crystal structure revealed a homotrimeric structure in which the C-terminal protein portions are clustered, enabling a cooperative membrane binding mode for PRLs in which both the prenylation motifs and a polybasic phosphoinositide-binding C-terminal region are required. Furthermore, trimerization has been found to be essential for PRL-1 function in cells [91–93,96]. PRL-1, PRL-2, and PRL-3 are highly conserved, and the possibility that PRLs can also form heterotrimers, possibly fine-tuning PRL substrate specificity and/or activity, needs to be tested. In addition to the classical inhibitory oxidation of the catalytic cysteine, common to most PTPs, PRLs can also be inactivated by an intradisulfide bond between the catalytic cysteine and a conserved neighboring cysteine. This putative regulatory mechanism has also been described for other cysteine-based PTP-domain DSPs, such as PTEN or the Cdc14-related KAP DSP [97,98], but the implications for regulation of PRL activity in cells have not been evaluated. The involvement of PRLs in oncogenesis has been widely documented. Ectopic expression of catalytically active, farnesylated PRL-1 or PRL-3 favors many oncogenic events, including angiogenesis, cell invasion, motility and metastasis. Some of these events have been associated with the stimulation of Src, Rho and phosphatidylinositol 3-kinase signaling pathways, the latter being likely due to downregulation of PTEN [33,99–102]. Furthermore, a correlation has been found between expression of PRL-3 and the malignant or metastatic status of different tumor cell types, including breast, colon, gastric and liver carcinomas [6,103,104]. Also, forced reduction of PRL-1 or PRL-3 levels reduces cell adherence and invasion [105–107], and small-molecule PRL-3 inhibitors diminished tumor growth and invasiveness in in vitro cell systems and in mice [108,109]. Together, these findings make PRLs promising targets for anti-cancer therapy, and foster the necessity of the development of biologically active, specific PRL inhibitors.


The two groups of DSPs that are most directly linked to human disease, PTEN and MTMs, mainly function in vivo as lipid phosphatases, rather than as protein phosphatases [2,110,111]. In mammals, the MTM subfamily of DSPs is a large group of phosphatases that contain a conserved PTP domain and a PH-GRAM domain that binds phosphoinositides, flanked by other protein- or lipid-interaction regulatory domains and motifs (DENN, coiled-coil, PH, FYVE, PDZ). MTMs include active enzymes that dephosphorylate the D3 phosphate position from phosphatidylinositol 3-phosphate and phosphatidylinositol 3,5-bisphosphate signaling phosphoinositides, but the group also includes inactive enzymes which seem to function as adaptor or regulatory subunits of active MTMs. Their role in phosphoinositide metabolism make MTMs important effectors in the control of the phosphatidylinositol 3-kinase pathways that regulate membrane trafficking during endocytosis [3,9,112–114]. The prototypical myotubularin, MTM1, is encoded by a gene located in the X chromosome, whose mutation is associated in humans with X-linked myotubular myopathy (XLMTM), a congenital disease characterized by muscle weakness and hypotonia that results in most of the cases in perinatal death from respiratory failure. The hallmark of this severe pathology is the presence of disorganized skeletal muscle fibers, likely producing defects in mature skeletal muscle maintenance [115,116]. MTM1 mutations in XLMTM patients are widespread throughout the gene, and truncating mutations, as well as phosphatase loss-of-function mutations, are usually found in the more severe forms of the disease [117–119]. MTMR1 and MTMR2 genes are highly related to MTM1. MTMR1 is adjacent to MTM1 on the X chromosome, but no MTMR1 gene mutations have been associated with human disease. However, an impairment in the alternative splicing at the 5′ region (intron 2) of MTMR1 has been found in congenital myotonic dystrophy (cDM1), a pathology that shares clinical features with XLMTM [120]. The human MTMR2 gene is mutated in Charcot–Marie–Tooth disease (type 4B1; CMT4B1), an inherited infantile demyelinating neuropathy that affects to peripheral nerves, causing muscle weakness and atrophy at extremities, with severely decreased nerve conduction velocity. In mice, loss of Mtmr2 in Schwann cells is sufficient to generate a CMT4B1-like pathology [121–125]. As for MTM1 gene mutations in XLMTM, most of the MTMR2 gene mutations found in CMT4B1 patients generate truncated proteins or inactive PTP enzymes [126]. The crystal structure of the PH-GRAM plus the PTP domain of MTMR2 has been resolved [127], showing an expanded PTP domain that contains putative protein- or membrane-interaction regions (predicted SID and RID domains, respectively). Like the other PTP lipid phosphatase, PTEN, the catalytic pocket of MTMR2 is wider and deeper than the pocket of classical PTPs, and displays a surrounding electropositive surface. However, the catalytic properties of MTMR2 seem to be different than those of PTEN due to the different location of the catalytic Asp residue. Also, MTMs display resistance to oxidation that is not shared by other PTPs [128]. In good correlation with mutational–functional studies, most of the missense mutations found in MTMR2 and MTM1 are predicted to disrupt the protein core structure, and some others affect the substrate binding pocket or the active site structure [127,129].

MTMR13 is an inactive phosphatase whose gene is also mutated in CMT4B2 (clinically indistinguishable from CMT4B1) [130,131]. MTMR13 forms heterodimers with MTMR2, and regulates both MTMR2 catalytic activity and subcellular localization, which may explain the common involvement of these two MTMs in CMT4B, reinforcing the importance of MTMR2 catalytic activity in the disease [121,132]. The CMT4B-like phenotype of Mtmr13-knockout mice is similar to that displayed by Mtmr2-knockout mice [133]. However, in humans some MTMR13 mutations have been associated with glaucoma [130], suggesting the existence of MTMR2-independent functions for MTMR13. MTMR5 has been shown to stimulate cell growth by virtue of its association with nuclear SET domain-containing proteins, and MTMR5-knockout mice display male sterility due to defects in spermatogenesis [134–137], suggesting a lineage-specific cell-growth-control function for this inactive phosphatase [3,9,138]. MTMR5 is highly related to MTMR13, both at the amino acid and functional level, but no mutations have been reported for MTMR5 in human disease. MTMR5 also associates with MTMR2; in fact, heteromeric interactions between MTMs are a common feature of these proteins, and include pairs of active–inactive and active–active enzymes, indicating a complex scenario of MTM functional regulation [121,139–142]. Comprehensive studies on the mutational status of MTM genes in MTM-related diseases, as well as additional mouse MTM-knockout models (Table 1), are required to fully understand the specific functions of the different MTMs and their involvement in human disorders. Drug compounds that modulate MTM oligomerization could be of great help at this regard, as well as putative therapeutic agents to alleviate MTM-related diseases.


The gene that encodes the human phosphatase laforin, EPM2A, is mutated in about 60% of patients that present with progressive myoclonus epilepsy of Lafora type (Lafora disease; PME; LD), an inherited neurodegenerative disorder with onset in late childhood, characterized by seizures and progressive neurological dysfunctions, with death within 10 years of onset. Pathologically, LD is characterized by the presence of abnormal, poorly branched, insoluble and heavily phosphorylated glucose polymers (polyglucosans), called Lafora bodies (LB), that accumulate primarily in neurons [143–146]. Laforin is a unique DSP in mammals, both in terms of modular structure and in phosphatase specificity. Laforin contains a functionally essential carbohydrate-binding domain (CBD), followed by a PTP domain with similarity to small atypical DSPs. The carbohydrate-binding domain binds to glycogen, polyglucosans and LB in vitro, and localizes laforin to sites of glycogen–microsomal complexes in cells [147]. Laforin-knockout mice display LD-like symptoms and pathology, including the presence of LB [148]. Interestingly, transgenic mice overexpressing a dominant negative form of laforin also display LB accumulation, although these mice do not show LD-like symptoms [149]. Laforin dephosphorylates glycogen in vitro, as well as amylopectin, a weakly branched carbohydrate that is similar to the carbohydrate deposits found in LB [150,151]. This may explain the reported diminished in vitro dephosphorylation of pNPP by laforin in the presence of glucose polymers [152]. In addition, laforin dephosphorylates the inhibitory Ser9 of GSK3β, inhibiting glycogen synthesis through glycogen synthase as well as tumor growth mediated by the Wnt signaling pathway [153,154]. However, how and if laforin directly affects GSK3β is under debate [150,155]. A tumor-suppressor activity of laforin has been described in lymphomas from TCR-transgenic immunocompromised mice [154]. Nevertheless, no mutations have been described to date in the EPM2A gene associated with lymphoma patients, and no mutations were found in the EPM2A gene in a comprehensive screening for mutations at PTP genes in colorectal cancer samples [156]. It would be interesting to monitor laforin protein expression in human tumor samples. Remarkably, both types of laforin substrate specificity (carbohydrate- and GSK3β-phosphatase) relate to the direct role of laforin in glycogen storage or degradation. Optimal laforin phosphatase activity and GSK3β activation is dependent on laforin dimerization through the CBD [157]. In addition, laforin associates with proteins involved in glycogen metabolism, including glycogen synthase and the R5/PTG regulatory scaffolding subunit of PP1, as well as with HIRIP5, a protein of unknown function that it is also an in vitro substrate of laforin [150,158,159]. Laforin also associates with malin, a second causative LD gene (EPM2B, NHLRC1) [153,160]. Malin is an E3 ubiquitin ligase that promotes the degradation of laforin, and LD mutations in the malin gene block ubiquitin-dependent laforin degradation. Malin and laforin co-localize to proteasome-related protein aggregates upon cell treatment with proteasome inhibitors [161], and laforin has been found to accumulate in LB from LD patients [149]. These observations suggest that laforin loss-of-function is sufficient, but not necessary, for LD pathogenesis. The laforin–malin complex facilitates proteasome-mediated degradation of both R5/PTG and glycogen synthase in a tissue-dependent manner, which may prevent glycogen accumulation in neurons [162]. The formation of the laforin–malin complex is modulated by the metabolic-sensing AMP-activated protein kinase pathway, and AMP-activated protein kinase directly interacts and phosphorylates laforin [163]. The mutational and functional analysis of laforin reveals that some laforin mutations found in LD patients interfere with laforin dimerization, binding to malin, binding to scaffold or regulatory proteins (such as R5/PTG) or laforin subcellular localization, whereas some others cause loss of phosphatase activity and loss of binding to glycogen [94,148,158]. Thus, phosphatase-dependent and -independent activities of laforin seem to be relevant for LD pathogenesis. A scenario emerges in which a functional multimolecular protein complex, which besides laforin and malin contains glycogen and a protein synthesis and degradation machinery, is required for proper glycogen metabolism processes. Alterations in the recruitment of laforin to such complexes may lead to unregulated protein- or carbohydrate-phosphorylation that could be causative of LD. Whether modulation of laforin expression and/or activity could be feasible to target LD or laforin-related cancer diseases remains unexplored.


PTEN is a major tumor-suppressor protein that exerts negative control on the oncogenic phosphatidylinositol 3-kinase/Akt pathway by virtue of its lipid phosphatase activity towards the D3 position of the phosphoinositide second messenger phosphatidylinositol (3,4,5)-triphosphate [110]. The PTEN protein core is composed of a PTP catalytic domain followed by a lipid-binding C2-domain, which is essential for membrane binding and tumor-suppressor activity [164]. In addition, PTEN possesses N- and C-terminal regulatory regions that control its activation and subcellular localization, and modulate PTEN function in many homeostatic and metabolic cell activities, including cell growth and death, development and differentiation, cell motility and migration, cell size and morphogenesis, metabolism and energy expenditure. Regulation of PTEN function is tight and occurs at multiple levels, including redox regulation of catalysis, allosteric activation by binding to membranes or lipids, stabilization and subcellular localization-targeting by binding to PDZ-domain-containing or regulatory proteins, and post-translational modifications such as phosphorylation, acetylation or ubiquitination. Recent findings have uncovered the existence of positive and negative feedback regulatory loops of the phosphatidylinositol 3-kinase pathway exerted by phosphatidylinositol 3-kinase catalytic (PI3Kδ) and regulatory (p85α) subunits through PTEN [165,166]. PTEN shuttles between cell membranes, cytoplasm, and nucleus, and the regulated PTEN compartmentalization is associated with distinct PTEN activities, from control of cell migration and chemotaxis to modulation of gene transcription. Protein phosphatase activities towards itself or other signaling proteins (FAK, Shc, PDGFR) may be important for cell-spreading- and cell-migration-related functions, although a direct relation of PTEN protein phosphatase activity with its tumor-suppressor function has not been conclusively demonstrated. In addition to its phosphatase-dependent activities, nuclear PTEN also displays protein-interaction-dependent, and phosphatase-independent activities that result in the control of gene transcription by oncoproteins or by other tumor suppressors, such as MSP58, p53 or Rad51, involved in cell transformation, apoptosis and chromosomal stability [167–175].

The PTEN gene is mutated with very high frequency in a wide variety of sporadic human cancers [176,177], being considered, together with p53, as one of the most relevant genes in clinical oncology. PTEN absence or mutation correlates in tumor samples and cell lines with upregulation of the activity of the phosphatidylinositol 3-kinase /Akt/mTOR survival pathway, as well as with the JNK signaling pathway [178,179], and lack of functional PTEN correlates with poor outlook in human cancer. Furthermore, some anti-cancer therapies, such as anti-erbB2 therapy in breast cancer or anti-γ-secretase therapy in T-cell leukemia, depend on PTEN for efficacy [180,181]. PTEN is also mutated in the germline of a variety of related hereditary syndromes that produce developmental alterations, hamartomas, and/or tumor susceptibility, which have been collectively grouped by molecular criteria as PTEN hamartoma tumor syndromes [182]. The two major hamartoma-related PTEN hamartoma tumor syndromes are Cowden syndrome and Bannayan–Riley–Ruvalcaba syndrome, whose patients carry germline PTEN mutations in 60–85% of cases [183,184]. Cowden syndrome patients present multiple hamartomas and benign tumors, especially in mucocutaneous tissues, breast, thyroid and uterus, and present an increased risk for many types of cancer, including breast, thyroid, endometrium and skin cancer. Another manifestation of Cowden syndrome is Lhermitte–Duclos disease, characterized by hamartoma of the brain and overgrowth of hypertrophied granule cells in the cerebellum. Bannayan–Riley–Ruvalcaba syndrome patients are clinically characterized by macrocephaly, lipomas and pigmented macules of the glans penis, and show high risk for skin, gastrointestinal and brain cancer. Other PTEN hamartoma tumor syndromes include Proteus and Proteus-like syndromes, as well as autism spectrum disorder with macrocephaly. Because of the diversity in the symptoms of these disorders, the molecular diagnosis based on the PTEN mutational status is highly relevant for the follow-up and effective treatment of patients with these syndromes. PTEN mutations found in tumors and in PTEN hamartoma tumor syndromes patients distribute all along the PTEN protein core, although the catalytic motif in the PTP domain is a major mutation target. Mutations are also found in the promoter region, as well as near the PTEN gene intron–exon boundaries, whose consequences are defects in mRNA transcription or splicing, or in protein translation [175,182,185–187]. In summary, PTEN mutations associated with human cancer or PTEN hamartoma tumor syndrome result in loss of PTEN catalytic activity and/or diminished or absent PTEN protein levels, due to severe truncations and/or protein destabilization [188].

Pten-knockout mice are not viable due to early embryonic defects, and heterozigous mice are prone to tumor development and show autoimmunity [189–192]. Many distinct tissue-specific Pten-knockout mice have been generated, whose description goes beyond the scope of this review [169,193]. In general, PTEN deficiency enhances tumorigenesis in the targeted tissues of the restricted knockout mice. Of interest, PTEN deficiency in the brain did not favor neoplastic transformation, but rather caused enlargement and abnormal development of brain structures, including cortex, hippocampus and cerebellum, which resemble the symptoms of Lhermitte–Duclos disease [194,195]. Depletion of PTEN in differentiated neuronal populations in the cortex and hippocampus resulted in abnormal social interaction in mice, associated with macrocephaly and neuronal hypertrophy, which are features of some autism spectrum disorders in humans [195]. By contrast, PTEN-deficient astrocytes showed increased proliferation in astrocyte-targeted mice [196]. PTEN depletion in neural stem cells increased this pool of self-renewing cells, whereas PTEN depletion in hematopoietic stem cells caused depletion of this undifferentiated cell population, which was substituted by a leukemia stem cell population [197–199]. Pten heterozygosity, as well as adipose tissue, muscle, liver or pancreatic β-cell-specific Pten deletion in mice, resulted in hypersensitivity to insulin and/or prevented insulin resistance and diabetes [200–204]. Also, a SNP in the 5′-UTR of PTEN gene, that results in increased PTEN translation, has been found in humans associated with a Japanese population of type 2 diabetes patients [205]. This outlines the in vivo role of PTEN as a negative regulator of insulin signaling and sensitivity, and its possible role in human diabetes [206,207]. Small molecule inhibitors of PTEN have been synthesized that could be used as therapeutic agents that down-regulate PTEN activity [208]. However, sustained inhibition of PTEN in chronic diseases (such as type 2 diabetes) may be causative of cancer, making the use of PTEN inhibitors in these cases inappropriate. Alternatively, the pro-apoptotic role of PTEN in neuronal cells can be exploited therapeutically to inhibit PTEN in acute cases of neuronal damage, such as stroke or other ischemic injuries [209–211]. Electric-signal-induced wound healing is enhanced by deletion of PTEN [212], and inhibitors of PTEN are effective adjuvants in wound healing [213]. Also, in a prostate cancer mouse model, a pro-senescence activity via the p53 pathway was observed upon Pten inactivation, suggesting that interference with PTEN activity could be therapeutically beneficial in some cancers [214]. Finally, in a murine model of asthma, PTEN protein expression and activity were decreased in lung tissues, and bronchial inflammation was reduced by adenoviral PTEN overexpression [215]. Together, these findings illustrate the multiple involvement of PTEN in human disease, which seems to be cell-type and cell-compartment dependent and exceeds its role as a potent tumor suppressor. Remarkably, the hyperproliferative disorders present in heterozygous and in several tissue-specific PTEN-deficient mice, can be ameliorated using the mTOR kinase-inhibitor rapamycin [198,216–218]. In addition to drugs that interfere with the phosphatidylinositol 3-kinase/Akt/mTOR pathway, the discovery and validation of direct PTEN modulators would be of great value to elicit novel therapies, or to complement current therapies, in cancer, brain diseases and metabolic diseases.


Dual-specificity phosphatases are implicated in a variety of processes that play a role in disease, but a number of complications precludes their simple validation as drug targets. Their activity is under stringent control and they may affect multiple signal transduction partners and feedback control mechanisms. While DSPs are clearly druggable (many reports describe small-molecule inhibitors), as a rule such compounds are not totally selective and will affect subsets of enzymes with consequences that are extremely difficult to predict from mechanistic studies alone. The path forward to DSP target validation appears therefore to attempt to come to a better understanding of DSPs using genetic and genomic approaches as combined with empirical studies using inhibitors as pharmacological tools.


The authors are funded by the European Union Research Training Network MRTN-CT-2006-035830. The work in the laboratory of RP is partially funded by Ministerio de Ciencia y Tecnología (grant SAF2006-08319) and by the Instituto de Salud Carlos III (grant ISCIII-RETIC RD06/0020; Spain and Fondo Europeo de Desarrollo Regional). We thank Drs W Hendriks and P Sanz for their comments on the article.