Physiological Signaling Specificity by Protein Tyrosine Phosphatases

Matthew Soulsby, Anton M. Bennett


Protein tyrosine phosphatases (PTPs) are now recognized to be involved in a multitude of signaling events that control fundamental biological processes such as cell growth, differentiation, apoptosis, and cell movement. PTPs, which were initially thought to be less discriminating in their actions compared with their protein tyrosine kinase counterparts, are now known to regulate these various biological processes in a precise manner. This review will focus on the concept that PTPs exhibit remarkable signaling specificity through intrinsic differences between their PTP domains and through various modes of regulation that endows them with the capacity to promote unique physiological responses.

Review focusing on the concept that PTPs exhibit remarkable signaling specificity through intrinsic differences between their PTP domains and through various modes of regulation that endow them with the capacity to promote unique physiological responses.

Intracellular signaling pathways that are mediated by tyrosyl phosphorylation are controlled through the balanced and opposing actions of protein tyrosine kinases (PTKs) and protein tyrosine phosphatases (PTPs). The view that PTPs are equal contributing partners in the regulation of cellular tyrosyl phosphorylation continues to mature. However, there still remains the perception that PTPs play largely housekeeping roles. A growing body of evidence firmly dispels this perception, and it is evident that PTPs function with stringent signaling specificity, in concert with their PTK counterparts, to define specific biological outcomes. In this review, we will focus on illustrating that PTPs exhibit defined substrate specificity and subsequently regulate signaling pathways in a precise manner.

The PTP Superfamily

The superfamily of PTPs is characterized by a consensus signature motif represented by HC(X)5R, which defines the active site of these enzymes. Two classes of PTPs can be defined (FIGURE 1). The first class is composed of classical PTPs that are defined by cysteine-based phosphotyrosine specificity, and the second class constitutes the dual-specificity phosphatases (DSPs). There are 37 classical human PTP genes that contain at least one PTP domain of ~280 amino acids. The DSPs are much larger in number, and there are 65 of these genes in the human genome. The classical PTP genes can be further subdivided into transmembrane receptor PTPs (RPTPs), or non-transmembrane PTPs, which localize to the cytoplasm (2, 4, 68). There are 12 RPTPs containing tandem PTP domains with the remainder containing a single PTP domain. The membrane-proximal PTP domain (D1) and membrane-distal PTP domain (D2) serve distinct roles. In the majority of cases, the D1 PTP domain comprises the active domain, whereas in most cases, but not all, the D2 domain serves a negative regulatory role. Although still quite controversial, evidence has been put forth to suggest that RPTPs are regulated through dimerization whereby RPTP activity is inhibited upon dimerization. However, there is also evidence against such a model of RPTP regulation, and further work in this particular area is warranted. Much like the RTKs, RPTPs utilize their diverse extracellular domains to transduce intracellular signals through binding to soluble ligands, but, in addition, RPTPs utilize their extracellular domains to mediate cell-cell and cell-matrix interactions (33, 35, 45, 73). The non-transmembrane or cytoplasmic PTPs contain a single PTP domain, and their diversity is derived through noncatalytic regulatory domains that reside either at the NH2 or COOH terminus of the PTP domain (23, 68). The noncatalytic domains are critical for exerting PTP substrate specificity and include one or more of the following features: 1) regulation of PTP activity, 2) directing subcellular localization, and 3) targeting PTP protein-protein interactions. Hence, the noncatalytic regions of the non-transmembrane PTPs offer a broad level of structural diversity, and together with the PTP domain, it provides a robust mechanism for substrate selection (FIGURE 2).


The superfamily of protein tyrosine phosphatases and dual-specificity phosphatases

A: the superfamily of protein tyrosine phosphatases. The protein tyrosine phosphatase superfamily of enzymes is represented schematically, depicting receptor-like and non-transmembrane PTPs. Note the diverse extra-cellular domains of the receptor-like PTPs and non-catalytic domains of the non-transmembrane PTPs. BRO-1, BRO-1 homology; CAH, carbonic anhydrase-like; Cad, cadherin-like juxtamembrane sequence; FERM, FERM domain; FN, fibronectin type III-like domain; Gly, glycosylated; HD, histidine domain; Ig, immunoglobulin domain; KIM, kinase-interaction motif; MAM, mephrin/A5/μ domain; Pro, proline-rich; RGDS, RGDS-adhesion recognition motif; SEC14, SEC14/cellular retinaldehyde-binding protein-like; SH2, Src-homology 2. B: the dual-specificity phosphatases (DSPs) are quite diverse. Shown here are those representing the MAP kinase phosphatases (MKPs). MKPs that are shown in gray shuttle between the nucleus and cytoplasm. MK-STYK is presumed to be catalytically inactive, containing a substitution for the essential cysteine residue.


Factors that determine the specificity of PTP signaling
Several factors control a PTP’s ability to discriminate between multiple tyrosylphosphorylated targets, allowing the PTP to signal in a highly specific manner and thus generate defined physiological responses. These factors include the subcellular location of the PTP, posttranslational modifications, which regulate PTP activity, and intrinsic structural differences within the PTP domain itself. See text for details and examples.

The most diverse subgroup amongst the PTP superfamily constitutes the 65 genes encoding the DSPs (3, 9, 68). DSPs are non-transmembrane PTPs that contain a single PTP domain with an assortment of noncatalytic motifs that participate in substrate recognition. The catalytic pocket of the DSPs assumes a shallower conformation compared with the tyrosine-specific PTPs, allowing accommodation of not only phosphotyrosine but additionally phosphoserine and phosphothreonine (79). Some members of the DSPs are also capable of accommodating non-protein substrates such as phospholipids. The subgroup of DSPs that have been studied most extensively is the mitogen-activated protein kinase (MAPK) phosphatases (MKPs). The MAPKs are a subfamily of serine/threonine kinases that play major roles in regulating cellular function. The three major subgroups of the MAPKs are the extracellular signal-regulated kinases 1 and 2 (Erks), c-Jun NH2-terminal kinase (JNK), and p38 MAPK. There are 10 MKPs, all of which have the capacity to dephosphorylate the regulatory threonine and tyrosine residues on MAPK, leading to its inactivation (62). Some MKPs preferentially dephosphorylate the stress-responsive MAPKs, p38 MAPK, and JNK but not Erk, whereas others only dephosphorylate Erk (10). The selectivity of the MKPs toward the MAPKs is mediated by a noncatalytic domain on the MKPs known as the kinase interaction motif (KIM), which directs MAPK-MKP interactions (10, 52, 74). The affinity of the interaction between MAPK and the KIM dictates the selectivity of a MKP for a particular MAPK substrate (63). The KIM domain is also found in two tyrosine-specific PTPs, the striatal-enriched PTP (STEP) and the hematopoietic PTP (HePTP), both of which are capable of dephosphorylating Erk (27, 72). Again, in this instance, the KIM domain provides these tyrosine-specific PTPs with MAPK binding specificity and, hence, MAPK substrate selection.

Regulation of PTP Function

PTP function can be regulated through the control of when and where they are expressed. In addition, PTPs are posttranslationally regulated through a variety of modifications including proteolysis, phosphorylation, sumoylation, and oxidation. There has been much focus on the actions of these types of posttranslational modifications as dynamic mechanisms through which PTPs can be regulated. Recent advancements provide insight into novel modes of PTP regulation by posttranslational modifications.

Proteolysis is a common posttranslational modification that regulates PTP activity. For example, calcium is a critical initiator of protease activity, and it has been found that the calcium-activated protease, calpain, elicits cleavage of negative regulatory domains of several non-transmembrane PTPs resulting in their activation. PTP-1B, PTP-MEG, and SHP-1 are activated upon calpain-induced cleavage (23, 25, 29). In addition, several members of the receptor PTPs (e.g., LAR, RPTPσ, and RPTPγ) are subject to limited proteolysis of their extracellular domain such modifications could control receptor interactions with ligand and/or extra-cellular matrix components.

Phosphorylation is also a key regulator of PTP function. PTP-PEST, when phosphorylated on serine by protein kinase C, leads to a robust stimulation of its phosphatase activity (26). Tyrosyl phosphorylation of PTPs such as PTP-1B, SHP-1, SHP-2, and RPTPα appears to modulate the level of phosphatase activity as well as promote the assembly of protein complexes. For example, tyrosyl phosphorylation of PTP-1B has been reported to result in increased phosphatase activity (19). Whereas, tyrosyl phosphorylation of SHP-2 and RPTPα generate binding sites for protein-protein interactions with adaptor proteins such as Grb2, which functions as a platform from which to direct activation of the Ras/Erk pathway (6, 32).

PTPs are now recognized to be targets for modification by oxidation (69). Oxidation of the essential catalytic cysteine within the C(X)5R active site motif impairs the nucleophilic function of this key residue, rendering the PTP inactive. Importantly, the oxidation of the catalytic cysteine is reversible, making this modification a dynamic mode of PTP regulation (56). A number of PTPs have been found to be reversibly oxidized, and these include PTP-1B, TC-PTP, SHP-2, MKPs, and RPTPα (36, 4647, 55). The physiological outcome of reversible PTP oxidation suggests novel modes through which PTP activity and, hence, function can be discretely controlled within precise spatial locales. Two additional posttranslational modifications have recently been reported to occur on PTPs. PTP-1B was found to undergo sumoylation in response to insulin, which led to its inactivation (18). These experiments demonstrated that sumoylation of PTP-1B can function as a negative feedback mechanism in response to insulin. This was the first example that sumoylation directly modulates PTP activity. However, it remains unclear as to the physiological significance of this modification on the function of PTP-1B in metabolic regulation. Presumably, failure to appropriately sumoylate PTP-1B following insulin stimulation would be expected to increase insulin sensitivity. More recently, acetylation of MKP-1 on lysine 57 was shown to promote the association between MKP-1 and p38 MAPK (11). Deacetylase inhibitors were found to block LPS-induced p38 MAPK activation in wild-type but not in MKP-1-deficient mice, suggesting that acetylation of MKP-1 facilitates p38 MAPK dephosphorylation (11). It still remains to be established as to what the molecular basis is for this MKP-1/p38 MAPK interaction, especially in context of the established role of the KIM domain and whether acetylation is a more general mechanism of MKP/MAPK regulation.

Physiological Signaling Specificity by PTPs

Despite the tremendous advances in PTP research, there is still the perception that PTPs function non-specifically and play largely housekeeping roles. However, there is now more than ample evidence based on structural, biochemical, and genetic studies that firmly dispel this perception. The examples cited above demonstrate that posttranslational modifications and protein-protein interactions contribute to PTP specificity. Additionally, the PTP domain itself exhibits properties of intrinsic substrate specificity. From a structural standpoint, the surface of the PTP domains are quite diverse (2, 4, 5). These observations suggest that there are likely to be significant differences in the surface electrostatic potential among the PTPs that contribute to substrate selectivity (2, 4, 5). To illustrate the specificity that exists within the PTPs, and subsequently the signaling pathways that they engage, we will compare those PTPs that share very close structural similarities but yet affect distinct physiological responses.

PTP-1B and TC-PTP: close cousins but definitely not twins

The first non-transmembrane (NT1) subfamily is composed of two PTPs, PTP-1B and T-cell PTP (TC-PTP) (FIGURE 1) (3). Structurally, PTP-1B contains an NH2-terminal PTP domain and a COOH-terminus containing both an ER retention sequence and a poly-proline-rich motif. TC-PTP has two forms, one that localizes to the nucleus and another to the cytosol. Remarkably, the PTP domain of TC-PTP and PTP-1B share 71% amino acid identity (3). Yet, although TC-PTP and PTP-1B share some common targets, they are also capable of dephosphorylating distinct subsets of tyrosyl phosphorylated proteins. A number of RTKs have been identified as PTP-1B substrates, such as the epidermal growth factor receptor (EGFR), platelet-derived growth factor receptor (PDGFR), colony stimulating factor-1 receptor (CSF-1R), insulin-like growth factor-1 receptor (IGF-1R), and insulin receptor (IR) (22, 24, 31, 38, 41). Similarly, TC-PTP dephosphorylates an overlapping subset of RTK substrates that includes IR, CSF-1R, and EGFR (46, 59, 66). In contrast, TC-PTP and PTP-1B dephosphorylate similar but largely distinct sets of non-RTK substrates. For example, PTP-1B targets include the Src family kinases (SFKs) (8, 13, 43) and JAK2 (14, 28, 49, 80), whereas TC-PTP targets the SFKs (71) and JAK1 (60). These observations point to the fact that, despite their high similarity between their PTP domains, PTP-1B and TC-PTP can target different substrates.

Studies in which chimeras between PTP-1B and TC-PTP were performed provided evidence for intrinsic specificity within their PTP domains. Using PTP substrate-trapping approaches to capture tyrosyl phosphorylated PTP substrates, it has been shown that PTPs indeed interact with distinct subsets of tyrosylphosphorylated proteins (65). Interestingly, by swapping the PTP domains between PTP-1B and TC-PTP, Tiganis et al. showed that, despite the PTP domain of TC-PTP being targeted to the precise locale of endogenous PTP-1B, the TC-PTP/PTP-1B chimera was unable to form the same repertoire of PTP-substrate interactions compared with wild-type PTP-1B (66). These experiments further support the idea that the PTP domains of PTP-1B and TC-PTP exhibit unique properties that govern substrate specificity.

Insight into the physiologically important pathways that PTP-1B and TC-PTP engage in has emerged from the use of mouse genetic approaches. Mice lacking PTP-1B exhibit no obvious developmental defect but are resistant to diet-induced obesity and have enhanced insulin sensitivity due to increased levels of insulin-mediated signaling (22, 39). In contrast, mice lacking TC-PTP die shortly after birth due to severe defects in the hematopoietic system (78). These differences are likely attributed to the intrinsic PTP substrate specificity of both PTP-1B and TC-PTP in addition to their distinct tissue expression patterns and other features imparted by their noncatalytic domains. Together, these properties are likely to be sufficient to account for the unique physiological signaling effects of these closely related PTPs (FIGURE 3).


Closely related PTPs regulate vastly different physiological responses
Schematic highlighting the structural similarities and functional differences between closely related PTPs. Despite high overall homology, these enzymes generate unique physiological responses. This is due, in part, to variations in their expression pattern and subcellular localization. Intrinsic differences within the PTP domain and noncatalytic domains also contribute to their unique signaling abilities. The examples shown represent PTP-1B and TC-PTP (A) and SHP-1 and SHP-2 (B).

SHP-1 and SHP-2: opposites attract

The SH2 domain-containing phosphatases include the second non-transmembrane group (NT2) of the PTP superfamily (FIGURE 1). The SH2 domain-containing PTPs contain tandem NH2-terminal SH2 domains and a COOH-terminal PTP domain. Within the PTP domain, SHP-1 and SHP-2 share 61% sequence identity and 55% sequence identity overall (50). SHP-1 is expressed predominately in the hematopoietic compartment, whereas SHP-2 is expressed ubiquitously. The SH2 domains of SHP-1 and SHP-2 target them to specific phosphotyrosyl-containing proteins within a sequence-specific context (61). Despite the seemingly high structural similarity, SHP-1 and SHP-2 could not be more different in their function; SHP-1 negatively regulates, whereas SHP-2 is generally a positive regulator of cell signaling (FIGURE 3).

A negative signaling role for SHP-1 was demonstrated when it was discovered that mutations in SHP-1 cause the autosomal recessive disorder in mice called motheaten (me), a hematological disease in which mice exhibit a severe combined immunodeficiency syndrome (81). Indeed, the identification of several SHP-1 substrates supports the notion that SHP-1 negatively regulates hematopoietic signaling. SHP-1 has been shown to dephosphorylate the CSF-1 receptor (67) as well as intracellular signaling molecules known to positively regulate hematopoietic signaling such as the SFKs (16) and proteins that function as assembling platforms for molecules such as the paired immunoglobulin-like receptor B (PIR-B) and CD72 (67, 76). There have been cases where SHP-1 has been suggested to function as a negative regulator of cell signaling in non-hematopoietic tissues. For example, SHP-1 has been shown to dephosphorylate the carcinoembryonic-antigen-related cell adhesion molecule-1 and the insulin receptor in the liver (21). These effects are proposed to result in the impairment of glucose homeostasis in the motheaten mouse (21).

In contrast to SHP-1, SHP-2 signals positively in virtually all cases, although there have been a few select reports where it plays a negative regulatory role (82). The differential binding specificity of the SH2 domains for SHP-2 compared with SHP-1 likely explains, in part, its distinct signaling effects to that of SHP-1 (61). SHP-2 interacts, via its SH2 domains, with the EGFR and PDGFR directly and/or couples indirectly to other RTKs through binding to adaptor proteins (51). Unlike SHP-1, SHP-2 positively stimulates the Ras/Erk pathway (51), clearly indicating that SHP-1 and SHP-2 must dephosphorylate distinct sets of substrates. Moreover, the positive effects SHP-2 has on the Ras/Erk pathway has implicated it as a potential candidate in the development of cancer (12). Interestingly, domain-swapping experiments between SHP-1 and SHP-2 demonstrate that the PTP domain of SHP-1 cannot recapitulate the function of the PTP domain for SHP-2; the opposite is also the case (53). Hence, the PTP domains of SHP-1 and SHP-2 contain intrinsic specificity for substrates that define their effects on downstream signaling.

A number of SHP-2 substrates have been identified, although only a select few provide direct mechanistic insight into the basis for SHP-2’s positive signaling effects, and this area still remains quite controversial. Some of these substrates include the EGFR, whose dephosphorylation at Y992 serves to prevent p120RasGAP from being recruited to a complex to inactivate Ras (1). Our group has demonstrated a similar mechanism of SHP-2 action in muscle cells where SHP-2 prevents the inactivation of RhoA by dephosphorylating p190b RhoGAP (40). Hence, site-specific EGFR dephosphorylation or direct dephosphorylation of a RasGAP by SHP-2 promotes small GTPase activation. SHP-2 also signals positively to Erk through its ability to control SFK activation (83). Thus, in contrast to SHP-1, SHP-2 assembles with distinct protein complexes and dephosphorylates targets that are negatively regulated by tyrosyl phosphorylation. Unlike SHP-1, where disruption of its function leads to hematological disease (70), loss of SHP-2 function in mice results in embryonic lethality (58, 77). An even more striking example of the diametrically opposite functions that SHP-1 and SHP-2 display has emerged through the discovery that mutations in the human PTPN11 gene, which encodes for SHP-2, causes ~50% of Noonan syndrome cases (64). Noonan syndrome is an autosomal-dominant disorder that results in cardiovascular defects, musculoskeletal anomalies, and increased risk of leukemia (64). What is striking about the PTPN11 mutations that cause Noonan syndrome is that they map to regions of SHP-2 that result in its constitutive activation (64). Enhanced SHP-2 phosphatase activity is thought to give rise to Noonan syndrome through constitutive stimulation of the Ras/Erk pathway (7). However, it is not clear whether the Ras/Erk pathway alone or whether additional targets of SHP-2-associated Noonan syndrome mutations collaborate with Erk to manifest the entire repertoire of the disease. Regardless, what is evident is that SHP-2 plays a dramatically distinct signaling role in both physiological and pathophysiological contexts compared with SHP-1.

MKP conundrum: overlapping MAPK substrates distinct signaling fates

The MKPs represent a provocative case study for understanding the physiological signaling specificity of PTPs. Because the 10 members of the MKP family all dephosphorylate Erk, p38 MAPK, and JNK, it was initially perceived that the MKPs would have redundant functions. The first knockout phenotype of an MKP was that of MKP-1, and the initial assessment of this phenotype supported the idea of redundancy (20). Despite the fact that all the MKPs share a common set of MAPK substrates, the complexity of their regulation at the transcriptional and posttranslational level is broad enough that these enzymes show stark differences in their physiological actions (15, 17, 42, 75, 84). The emergence of genetic data now argue strongly against the initial perception of redundancy among the MKPs (Table 1), and compelling evidence for physiological signaling specificity for these enzymes has emerged.

View this table:
Table 1.

Distinct signaling and functional properties of the MKPs

MKP-1, the first MKP identified, is a nuclear localized immediate-early gene that dephosphorylates p38 MAPK and JNK comparably, and Erk to a lesser extent (54). Although initial efforts analyzing MKP-1-deficient mice revealed that there was no obvious phenotype (20), subsequent analysis of these mice in response to LPS indicated that MKP-1 functions as a major negative regulator of the innate immune response (15, 30, 57, 85). Mice lacking MKP-1 were sensitive to LPS-induced endotoxemia, and macrophages isolated from these mice exhibited dramatically enhanced p38 MAPK and JNK, but not Erk, activity in response to LPS (15, 30, 57, 85). MKP-1-deficient mice are also resistant to diet-induced obesity (75). Surprisingly, despite being lean, MKP-1-deficient mice still succumb to the effects of insulin resistance when fed a high-fat diet. These observations were reconciled by demonstrating that MKP-1 was restricted to dephosphorylation of the nuclear pool of MAPKs, whereas the cytoplasmic pool of MAPKs that influenced insulin signaling were subject to separate regulatory mechanisms and the deleterious effects of high-fat feeding (75). These results highlight an important mechanism through which the MKPs exert signaling specificity through the dephosphorylation of discrete pools of MAPKs (37) (FIGURE 4).


MAP kinase phosphatases (MKPs) coordinate compartmentalized MAP kinase dephosphorylation
MKPs shown with solid arrows are those exhibiting stronger substrate preference to a particular MAPK than those shown with the lighter broken arrow. MKPs exert specific signaling through temporal, spatial, and MAPK-selective dephosphorylation that integrate to produce defined effector responses or gene expression events.

Other MKPs that have been knocked out show quite remarkable and distinct phenotypes (Table 1). For example, MKP-3-deficient mice that survive weaning show skeletal dwarfism, cryaniosynostosis, and developmental defects, which is consistent with MKP-3 being a negative regulator of the Erk pathway downstream of FGF receptor signaling (42). However, these differences have not been observed by others (44). On the other hand, MKP-4 (DUSP9), which, like MKP-3 (DUSP6), predominantly dephosphorylates Erk (48), when deleted in mice results in embryonic lethality as a result of a placental defect (17). The knockout of MKP-5, as for MKP-1, exhibits no remarkable phenotype. However, MKP-5-deficient mice show increased JNK activity in T-cells but, surprisingly, fail to proliferate in response to anti-CD3 stimulation (84). These authors demonstrated that MKP-5-deficient mice were enhanced in their sensitivity to the innate immune response and were protected from the development of autoimmune disease (84). Hence, although MKP-5 and MKP-1 both target JNK, the physiological outcome of MKP-5 deletion suggests that MKP-5 and MKP-1 have distinct regulatory roles even within the immune system. Finally, PAC-1 (DUSP2)-deficient mice also develop normally but are protected from arthritis, similar to MKP-5-deficient mice (34). However, the signaling mechanisms of Pac-1 regulation is complex since, although there is enhanced JNK1 activity, p38 MAPK and Erk are paradoxically decreased (34). These results suggest that the enhanced activity of JNK influences p38 MAPK/Erk activities in a cross-talk type of manner (34). Collectively, the observations derived from the various MKP-deficient mouse models exemplify the physiological signaling specificity of these PTPs within the intact organism despite their remarkable selectivity toward the MAPKs. These observations suggest that there are likely to be additional levels of regulation among the MKPs that have yet to be fully recognized.

Future Perspectives

There is clearly a substantial body of data from which an unequivocal conclusion can be made to support the notion that PTPs are highly specific. The features that confer PTP specificity stem from the fact that PTPs are structurally diverse, are subject to multiple levels of regulation, and exhibit intrinsic substrate specificity within their catalytic domains. One challenge that still remains is to define the substrates that the PTPs dephosphorylate. This information will provide important insight into the fundamental mechanisms involved in PTP-mediated signaling. Nevertheless, we are now appreciating that PTPs are key players in a multitude of physiological processes. In some cases, the dysfunctional actions of PTPs underlie the molecular basis for the cause of certain human diseases. This suggests that PTPs represent an untapped source for the development of therapeutics for the treatment of human disease. The challenge will be to continue to uncover the physiological roles of the PTPs and to identify the structural determinants that dictate their specificity. With this knowledge, the future of the PTPs as coveted therapeutic targets will begin to be realized.


We apologize to our many colleagues whose contributions to this area of research was not discussed due to space limitations.

A. M. Bennett was supported by National Institute of Health Grants AR-046524, DK-075776, and DK-057751.


View Abstract