Journal list menu

Volume 275, Issue 5 p. 867-882
MINIREVIEW
Free Access

Protein tyrosine phosphatases: structure–function relationships

Lydia Tabernero

Lydia Tabernero

Faculty of Life Sciences, University of Manchester, UK

Search for more papers by this author
A. Radu Aricescu

A. Radu Aricescu

Wellcome Trust Centre for Human Genetics, University of Oxford, UK

Search for more papers by this author
E. Yvonne Jones

E. Yvonne Jones

Wellcome Trust Centre for Human Genetics, University of Oxford, UK

Search for more papers by this author
Stefan E. Szedlacsek

Stefan E. Szedlacsek

Institute of Biochemistry of the Romanian Academy, Bucharest, Romania

Search for more papers by this author
First published: 08 February 2008
Citations: 113
L. Tabernero, Faculty of Life Sciences, University of Manchester, Michael Smith Building, Manchester M13 9PT, UK
Fax: +44 161275 5082
Tel: +44 1612757794
E-mail: [email protected]

Abstract

Structural analysis of protein tyrosine phosphatases (PTPs) has expanded considerably in the last several years, producing more than 200 structures in this class of enzymes (from 35 different proteins and their complexes with ligands). The small–medium size of the catalytic domain of ∼280 residues plus a very compact fold makes it amenable to cloning and overexpression in bacterial systems thus facilitating crystallographic analysis. The low molecular weight PTPs being even smaller, ∼150 residues, are also perfect targets for NMR analysis. The availability of different structures and complexes of PTPs with substrates and inhibitors has provided a wealth of information with profound effects in the way we understand their biological functions. Developments in mammalian expression technology recently led to the first crystal structure of a receptor-like PTP extracellular region. Altogether, the PTP structural work significantly advanced our knowledge regarding the architecture, regulation and substrate specificity of these enzymes. In this review, we compile the most prominent structural traits that characterize PTPs and their complexes with ligands. We discuss how the data can be used to design further functional experiments and as a basis for drug design given that many PTPs are now considered strategic therapeutic targets for human diseases such as diabetes and cancer.

Abbreviations

  • KIM
  • kinase interaction motif
  • LMW-PTP
  • low molecular weight protein tyrosine phosphatase
  • N-SH2
  • N-terminal SH2 domain
  • PTP
  • protein tyrosine phosphatase
  • RPTP
  • receptor protein tyrosine phosphatase
  • YopH
  • Yersinia PTP
  • The most significant trait of the protein tyrosine phosphatase (PTP) superfamily is conservation of the signature motif CX5R, which forms the phosphate-binding loop in the active site (known as the P-loop or PTP-loop). Despite relatively large sequence variations in the X5 segment, the conformation of the P-loop is strictly conserved and can be easily superimposed from different PTP structures, with minor deviations in the Cα tracing (< 1 Å). This structurally conserved arrangement ensures that the catalytic Cys, the nucleophile in catalysis, and the Arg, involved in phosphate binding, remain in close proximity and form a cradle to hold the phosphate group of the substrate in place for nucleophilic attack. The cysteine Sγ-atom is the nucleophile that attacks the substrate phosphorus atom leading to the cysteinyl-phosphate reaction intermediate. The arginine is involved both in substrate binding and in stabilization of the reaction intermediate [1]. Further to this, the amide groups in the P-loop point towards the interior of the cradle and form a network of hydrogen bonds to the phosphate oxygens (Fig. 1A). A conserved Ser/Thr residue in the P-loop has been proposed to play an important role in the stabilization of the thiolate group in the transition state facilitating the breakdown of the phosphoenzyme intermediate [2] (Scheme 1).

    Details are in the caption following the image

    (A) Structure of the phosphate-binding loop (P-loop). Stick representation of the consensus signature motif (CX5R) that forms the P-loop present in the active site of PTPs. The P-loop from bovine LMW-PTP (1PNT) [79] is represented and the catalytic Cys12 and Arg18 are labelled. The amide nitrogens form hydrogen-bond interactions (dotted green lines) with the phosphatase bound showing network of interactions that involve the catalytic Arg. The cradle-like conformation of the P-loop is conserved in the structures of all PTPs. (B) Structure of PTP1B (C215S mutant) in complex with phosphotyrosine (PDB entry 1PTV). Position of the substrate in the active site is illustrated by the phosphotyrosine ligand (blue). Tyr46 within the ‘KNRY’ conserved motif contributes the substrate recognition. Active-site nucleophile Cys215 (grey) (here mutated to Ser) attacks the substrate phosphorus leading to the formation of the cysteinyl-phosphate intermediate. Asp181 within the WPD-loop (cyan), here in the closed conformation, acts as a general acid donating a proton to the phenolate leaving group. (C) Binding of an allosteric inhibitor of PTP1B keeps the catalytic WPD-loop in the open conformation. Structure of PTP1B (cyan) in complex with the allosteric inhibitor 3-(3,5-dibromo-4-hydroxy-benzoyl)-2-ethyl-benzofuran-6-sulfonic acid 4-sulfamoyl-phenyl)-amide (termed ‘compound-2’ in [18]) (PDB entry 1T49) overlain on the PTP1B (C215S mutant) structure (red) in complex with a p-Tyr substrate (PDB entry 1PTV). Only the main structural elements involved in allosteric inhibition are represented. In the presence of the allosteric inhibitor (yellow), the C-terminus of PTP1B is disordered while in presence of the phosphotyrosine (green) it adopts the α-helical structure α7. Binding of allosteric inhibitor impedes the interaction between helices α3, α6 and α7, thus preventing the closure of the WPD-loop.

    Details are in the caption following the image

    General mechanism of catalysis of PTPs.

    The catalytic mechanism of PTP reaction requires the participation of a general acid and a general base. This is provided by a unique aspartic residue situated on the WPD-loop. During formation of the transition state intermediate, the catalytic Asp acts as a general acid protonating the oxygen of the leaving group in the tyrosine residue. In the second catalysis step, the same Asp functions as a general base during hydrolysis of the phospho-enzyme by accepting a proton from the attacking water and assisting in the conversion of the phospho-Cys enzyme to its resting Cys-SH state, thus regenerating the free enzyme [2,3]. Upon substrate binding, the WPD-loop closes over the active site bringing the catalytic Asp near the leaving group. An analogous Asp residue is found in the DPYY-loop of the low molecular weight protein tyrosine phosphatases (LMW-PTPs), although in this enzyme it appears to be less mobile than the WPD-loop and it adopts a fixed position near the active site. We focus our review on the tyrosine-specific PTPs with a Cys-based mechanism of catalysis (class I and class II) as described in the classification by Alonso et al. [4].

    Cytoplasmic class I PTPs

    Cytoplasmic PTPs, also called soluble or non-receptor PTPs, have a modular organization. In addition to the highly conserved catalytic domain they contain non-catalytic regions or domains that play a role in subcellular targeting, in regulation of the enzymatic activity or in recruiting specific ligands [4].

    Structural characteristics of the PTP catalytic domain

    The catalytic domain contains ∼280 amino acids that determine a specific PTP fold with several characteristic features. As illustrated by the structure of PTP1B (Fig. 1B), the first reported structure for a PTP [5], this fold is represented by a central, highly twisted β sheet composed of eight β strands forming a mixed β sheet with four parallel strands flanked by antiparallel ones. Six α helices surround the central sheet, four on one side and two on the other [6]. The active site is situated in a 9-Å deep crevice, 3 Å deeper than for dual-specificity phosphatases, thus providing selectivity for phosphotyrosine-containing protein substrates [7]. The signature motif VHCSXGXGR(T/S)G that forms the PTP-loop [8] between the C-terminus of the central β10 strand and the α4 helix is located at the bottom of the catalytic site. This loop contains the essential catalytic residues Cys215 and Arg221 in PTP1B. An essential structural component of the active site is the phosphotyrosine-recognition loop with the conserved motif KNRY (residues 43–46 in PTP1B) [7,8]. This loop determines the depth of the active site cleft and interacts, through its tyrosine residue, with the aromatic ring of the phosphotyrosine in the substrate. Another key element in catalysis is the WPD-loop (residues 179–181 in PTP1B; Fig. 1B). Remarkably, substrate binding into the active site triggers a significant movement of ∼6 Å of the essential Asp residue, simultaneous with a conformational switch of the whole WPD-loop from an ‘open’ to a ‘closed’ position [9]. In addition to phosphopeptidic substrates [7], small ligands also induce closure of the WPD loop. Structures of PTPs with small ligands like tungstate [9], sulfate [10] and phosphate [11–13] show evidence of a closed WPD-loop. Apo forms of PTPs generally have the WPD-loop in the open conformation [6]. However, there are a few notable exceptions, for example, apo-PTP1B has the WPD-loop in the closed form [14], the PTP1B complex with tungstate contains the open form of the WPD-loop [6] and similarly, in the SHP-1 complex with phosphopeptide substrates EDTLTpYADLD or PSFSEpYASVQ the WPD-loop is also in the open form [15]. These apparently contradictory findings may in part be explained by assuming that the WPD-loop fluctuates between an open and a closed position and that the binding of ligands determines a much longer residence time of the closed position [16].

    Another catalytically important structural element of the PTP fold is the Q-loop defined by the QTXXQYXF motif [8]. The glutamine residues in this motif are highly conserved and their main role is to position a water molecule in the active site, which is involved in hydrolysis of the thio-phosphate intermediate.

    Overall, the structure of cytoplasmic PTPs is highly conserved with only minor differences in the main structural core. For example, the structure of SHP-2 contains a short β sheet formed by the N-terminal βA and C-terminal βN strands, which is not encountered in PTP1B but that is present in the structures of kinase interaction motif (KIM)-containing PTPs (PTPRR, HePTP, STEP) and receptor PTPs; also, in the structure of KIM-containing PTPs there are several 310 helices which are not observed in the structures of other PTPs [11,13,17]. More prominent differences are visible at the N- and C-termini of the catalytic domains of several PTPs. Thus, PTP1B contains an additional C-terminal helix α7, not present in other PTP structures. This is a particularly important regulatory element for the catalytic activity of PTP1B as it stabilizes closure of WPD-loop through its interaction with helices α3 and α6 [18]. This idea is supported by the recent finding that most of inhibitor–resistant mutants of PTP1B are clustered on helix α7 and its surroundings [19].

    A number of reported PTP structures have an additional N-terminal helix (generically termed ‘α0’) with a relatively less-conserved sequence among PTPs, but are apparently important functionally. In SHP-1 this helix α0 is highly mobile: whereas in the peptide-bound form it is positioned far away from the catalytic core, in the ligand free-form it is rotated ∼60° and is located in the proximity of the PTP domain [15,20]. Experimental results support the idea that α0 of SHP-1 plays a cooperative role in substrate recognition [21]. PTPRR, HePTP, STEP and other KIM-containing PTPs also display an N-terminal helix α0. This is mainly stabilized by hydrophobic interactions with helix α5 and the loop following helix α2′, forming a hydrophobic cavity of ∼16 Å depth [11,13,17]. The role played by this helix seems to be related to the presence of: (a) its N-terminal residue (Thr253 in PTPRR and Thr45 in HePTP), which is specifically phosphorylated by ERK2 and p38 MAP kinases; and (b) a KIM located 15 residues upstream of α0, which is essential for the interaction with ERK2 and p38. It has been suggested that helix α0 contributes to the positioning of the KIM region for interaction with the docking groove of ERK2 and to the proper directing of Thr253 residue into the active site of ERK2 [17].

    The structure of PTPL1 provides another example of the N-terminal helix playing a specific role. Its helix α0 is located at a topologically equivalent position to helix α7 in PTP1B [12], and deletion of this helix results in an enzyme with very low activity. This finding together with the fact that helix α0 interacts with helix α3 in a similar manner as helix α7 does in PTP1B, suggests that helix α0 may be a regulatory structural element, involved in stabilization of WPD-loop in PTPL1.

    Structural determinants for substrate binding and regulation of the catalytic activity

    The catalytic domains of PTPs often exhibit relatively broad substrate specificities in vitro, although under in vivo conditions full-length PTPs have considerably more stringent specificities. This is in part due to the presence of additional regulatory domains that direct their subcellular localization and interactions with specific substrates. For example, the C-terminal segment of full-length PTP1B is responsible for anchoring PTP1B to the endoplasmic reticulum [22]. Substrate specificity may also be provided by selective substrate binding modules. The small KIM motif composed of 16 amino acids is an example of this strategy, it directs the KIM-containing PTPs to their physiological substrates, MAP kinases ERK2 or p38, which are subsequently dephosphorylated and inactivated [23]. Yersinia PTP (YopH) provides another interesting example of a substrate-targeting mechanism with multiple phosphotyrosine-binding sites. YopH is a virulence factor that dephosphorylates several focal adhesion proteins, for example, p130Cas [24] in human epithelial cells. It contains an N-terminal non-catalytic domain that binds tyrosine-phosphorylated proteins [25], but with a different fold than SH2 or PTB domains, thus representing a novel phosphotyrosine-binding domain [26]. Recently, the crystal structure of the YopH PTP domain in complex with a phosphopeptide revealed a second substrate-binding site (in addition to the active site) within the catalytic domain, that is the third phosphotyrosine binding site within the full-length YopH protein [27]. Moreover, it was proved that the two non-catalytic substrate-targeting sites co-operate in binding the p130Cas substrate, which contains 15 phosphorylation sites, thus providing efficiency and specificity for the PTP domain interaction with its specific substrate.

    The structures of three other cytoplasmic PTPs, PTP1B, TC-PTP and PTPL1, also reveal the presence of secondary phosphotyrosine-binding sites within their catalytic domains [12,28,29]. These secondary substrate-binding sites are represented by a positively charged pocket located close to the active site. In PTP1B, the secondary site is formed by Arg24, Arg254, Met258 and Gly259 and plays the important role of providing specificity for PTP1B action. This is illustrated by the fact that a physiological substrate of PTP1B, the insulin receptor kinase, contains a tandem of phosphotyrosine residues (1162 and 1163) and interacts with PTP1B in a characteristic bidentate mode: pTyr1162 is recognized and selectively dephosphorylated by the active site, whereas pTyr1163 is bound to the secondary binding site of PTP1B [30]. The interaction between the di-phosphorylated IRK peptide and the two substrate binding sites of PTP1B is highly selective, being 70-fold tighter than for the mono-phosphorylated peptides. The secondary site has also been exploited for development of inhibitors against PTP1B [31].

    The crystal structures of SHP-1 and SHP-2 reveal a more sophisticated regulatory mechanism controlled by substrate recruitment. SHP-1 and SHP-2 contain an N-terminal tandem of two SH2 domains as well as a C-terminal extension. The apo forms of both SHP-1 and SHP-2 are essentially inactive. However, their catalytic activity increases considerably upon binding of phosphopeptides by their SH2 domains [32,33]. Comparison of the structure of the auto-inhibited form of SHP-2 [34] with the structure of the N-terminal SH2 domain (N-SH2), both free and in complex with phosphopeptides, shows that N-SH2 plays the role of an allosteric switch: in the absence of a phosphopeptide ligand it binds to the PTP domain blocking closure of the WPD-loop (through the N-SH2 β sheet defined by strands βD′, βE and βF) but its phosphopeptide-binding site (defined by the loops EF and BG) is not functional; by contrast, binding of a phosphopeptide to the N-SH2 domain triggers concerted conformational transitions (involving EF loop, helix αB and β sheet βD’, βE and βF) which ultimately lead to dissociation of the N-SH2 from the PTP active site and subsequent activation. The other SH2 domain (C-SH2) seems not to be directly involved in regulation of catalytic activity and its role cannot be fully understood based on these structural data. However, it seems reasonable that C-SH2 acts in a cooperative manner with the N-SH2 domain in binding bisphosphorylated peptides. Consistent with this idea, it was proved that bisphosphorylated peptides binding both SH2 domains stimulate SHP catalytic activity 100-fold, whereas mono-phosphorylated peptides stimulate only 10-fold [35]. The structure of SHP-1 [20] supports an activation mechanism similar to that of SHP-2. The difference between the two structures consists mainly in a different orientation of the C-SH2 domain of SHP-1 and also a higher flexibility of this domain as compared to the C-SH2 of SHP-2.

    A different type of allosteric regulation of the catalytic activity was also reported for PTP1B. Benzbromarone derivatives are non-competitive inhibitors of PTP1B [18] and they bind in a novel allosteric site in PTP1B, as shown in the crystal structure of the complexes (Fig. 1C). This binding site is located ∼2 Å from the active site and is formed by helices α3 and α6. The inhibitory effect seems to result from blocking the interaction between helices α7 and α3–α6, present in the closed form of PTP1B, thus preventing closure of the WPD-loop. A truncated form of PTP1B lacking α7 is fourfold less active than the native form. This finding, together with those mentioned above, provides additional support for the particular significance of helix α7 in controlling the catalytic activity of PTP1B.

    Reversible oxidation of the catalytic cysteine is a characteristic regulatory mechanism of PTPs [36], and therefore, there was significant interest in establishing its molecular basis. Two structural analyses evidenced that a potential key intermediate in this process is a sulfenyl-amide in which the sulfur atom of active site cysteine (Cys215PTP1B) forms a covalent bridge with the amide nitrogen of the neighbouring residue (Ser216PTP1B) [37,38]. Formation of the sulfenyl-amide causes conformational modifications in the catalytic site, and has a double role: first, it protects the catalytic cysteine from irreversible oxidation to sulfonic acid and, second, it facilitates reactivation of PTP by biological reducing agents, since this is a reversible reaction. In addition, the substantial conformational changes associated with sulfenyl-amide formation, may represent a signal that the given PTP is in a temporary inactive state.

    Cytoplasmic PTPs as drug targets

    Many PTP genes in the human genome have been implicated in human diseases, leading to a special interest in selecting PTPs as drug targets [39–41]. The finding that the PTP1B knockout mouse was highly responsive to insulin and was resistant to diet-induced obesity [42] triggered intense efforts to identify specific inhibitors of PTP1B to develop drugs against type II diabetes and obesity [43]. Numerous structures of PTP1B complexes with single-site, double-site or allosteric inhibitors are now available [43] (see also the database of reported PTP structures at: http://ptp.cshl.edu or http://science.novonordisk.com/ptp). One promising drug candidate, ertiprotafib (Wyeth Research, Cambridge, MA, USA) successfully entered phase II clinical trials for treatment of type II diabetes; however, due to unsatisfactory efficacy and dose-limiting effects the trial was terminated in 2002 [39]. Despite this partial failure, trials to find new inhibitors continued and new research directions to develop efficient PTP inhibitors were identified. Structural information on PTPs has substantially contributed to the success of the fragment-based approach to identify new inhibitory compounds. Using this procedure, double-site PTP1B inhibitors were designed to bind both the active site and the second phosphotyrosine binding site of PTP1B [31,44]. The novel series of potent inhibitors exhibited sixfold selectivity over the highly homologous TCPTP and high selectivity over other phosphatases [44].

    Class I receptor-like PTPs

    Twenty-one of the 38 classical PTPs identified in the human genome [4] are type I membrane proteins and, because of their domain organization, were termed ‘receptor-like’ PTPs (RPTPs) well before any ligands had been identified [45]. Typically, an RPTP has an N-terminal extracellular region (lengths vary from ∼100 to > 1000 residues), a single transmembrane region and one or two intracellular catalytic domains, highly conserved within the family and with other classical PTPs [46]. Given this architecture, RPTPs appear ideally built to transduce signals across the plasma membrane, triggered by ligand binding to the extracellular region.

    Structure and role of the extracellular region

    The remarkable structural variety of RPTP ectodomains offers, first of all, a convenient criterion for classification [4,46]. In most cases, several types of domains are combined to produce modular arrangements: commonly used folds include meprin/A5/RPTPμ (MAM), Ig-like, fibronectin type III-like, carbonic anhydrase-like and cysteine-rich regions. In addition, alternative splicing and post-translational modifications (mainly N- and O-linked glycosylation) play important regulatory roles [46–48] and potentially contribute to a diversification of epitopes available for ligand binding. However, despite almost two decades of sustained efforts, the number of RPTP ligands identified remains surprisingly low (see review by den Hertog, Östman and Bohmer in this issue). The few notable examples are heparan sulfate proteoglycans that bind type IIa RPTPs (agrin and collagen XVIII for RPTPσ [49], syndecan and Dallylike for LAR in Drosophila [50]), the trans homophilic interactions of type IIb RPTPs RPTPμ [51,52] and RPTPκ [53], and pleiotrophin, a ligand for the type V RPTPζ [54]. A direct effect on catalytic activity has, so far, only been demonstrated for RPTPζ where pleiotrophin binding downregulates the catalytic activity; the exact mechanism of inhibition still remains unclear.

    Recently, structural work provided important insights into the ectodomain-dependent mechanisms that regulate type IIb RPTPs [55]. RPTPμ is a homophilic cell-adhesion molecule, expressed at high levels by neurons and vascular endothelia, and causing clustering when expressed on the surface of normally non-adherent cells in culture [51,52]; this activity is entirely driven by the extracellular region [56]. In confluent cell cultures, RPTPμ surface expression is significantly increased, post translationally, and the protein appears to be trapped at cell–cell contact areas, presumably via trans homophilic interactions. A crystal structure of the full-length RPTPμ ectodomain revealed an unexpectedly extended and rigid architecture, with residues from all four N-terminal domains contributing to a large adhesive interface (covering around 1630 Å2 per monomer) [55]. The RPTPμtrans dimer matches the dimensions of adherens junctions and, importantly, the length of cadherin trans dimers [57]. Moreover, cell-surface expression of RPTPμ ectodomain deletion constructs that still preserve the adhesive activity, induce intercellular spacings that correlate directly with the construct length, thus providing further evidence for existence of an extended ectodomain conformation as seen in the crystal structure. It was therefore suggested that the RPTPμ extracellular region plays a fundamental regulatory role, acting as a distance gauge and locking the phosphatase to its appropriate functional location (Fig. 2A), in proximity of the cadherin/catenin complex (one of RPTPμ physiological substrates) [55,58]. In addition to the trans interaction, the RPTPμ ectodomain can also form lateral (cis) dimers [59]. Two MAM domain loops may be involved in such interactions which, together with the trans adhesive dimers described above, may lead to the formation of 2D receptor arrays [56].

    Details are in the caption following the image

    The architecture and regulation of receptor protein tyrosine phosphatases. (A) Structural basis for ectodomain-controlled localization of RPTPμ at cell contacts. The crystal structure of a full-length RPTPμ ectodomain (rainbow coloured, PDB accession number 2V5Y) revealed a trans dimer with a rigid and extended conformation, matching in dimensions the intracellular spacings characteristic of adherens (cadherin-driven) junctions [55]. The ectodomain length and the large adhesive interface (∼1630 Å2 per monomer) are essential for controlling the subcellular localization of RPTPμ, bringing it in the proximity of its physiological substrates. Plasma membranes of two opposing cells (cell 1 and 2) are shown by gray rectangles, the full-length C-cadherin ectodomain structure is in blue (Ca2+ atoms in red, PDB accession number 1L3W). The C-terminal fibronectin type III domain of the RPTPμ extracellular region is largely disordered (indicated by dotted black oval), while the juxtamembrane and transmembrane regions are schematically shown by black lines. To give an idea of relative size, crystal structures of the RPTP LAR intracellular region (PDB accession number 1LAR, predicted to have a similar architecture with the corresponding region of type IIb RPTPs) and the complex between the β-catenin armadillo repeat region and an intracellular E-cadherin fragment (PDB accession number 1I7X) are shown in cyan and yellow/purple, respectively. Other components of the cadherin–catenin complex (and regions missing in the 1I7X structure) are indicated by the irregular green shape. For simplicity, these components are only shown in cell 2. (B) Crystal structures of RPTP intracellular regions and phosphatase regulation by sterical hindrance. The membrane proximal catalytic domains (D1) are shown in cyan, the distal ones (D2) in yellow. Catalytic sites in D1 and equivalent positions in D2 are marked by purple and green spheres, respectively. The ‘inhibitory wedge’ in D1 and the equivalent structure in D2 are shown in orange. Additional features in the CD45 D2 are the ‘acidic’ and ‘basic’ loops (largely disordered in the crystal structures 1YGR and 1YGU) are shown schematically in red and blue, respectively. In the RPTPα D1 crystal structure (PDB accession number 1YFO) the ‘inhibitory wedge’ blocks access to the catalytic site of a dyad-related molecule. In contrast, the D1 + D2 structures of LAR and CD45 are monomeric, casting doubt over the ‘wedge’ model. The recently solved structure of the RPTPγ D1 + D2 region (PDB accession number 2NLK) reveals a novel arrangement where the D1 catalytic site is blocked by the D2 of a symmetry-related molecule. The physiological significance of this arrangement remains to be determined.

    The principle of ectodomain-driven and size-controlled subcellular localization appears to apply to yet another RPTP, the type I CD45. The narrow spacings (about 15 nm) at local zones of cell contacts between T cells and antigen-presenting cells force the exclusion of CD45 (a protein with a large ectodomain) from the proximity of the MHC-TCR complex. This allows an increased phosphorylation of the TCR, a key step in signal transduction [60].

    Current efforts are directed towards the identification of additional RPTP ligands [61], as well as structural characterization of ligand–RPTP ectodomain complexes described to date. Other ectodomain-dependent processes such as shedding [62,63] and cis-oligomerization [47,64] may affect catalytic activity either indirectly (via subcellular relocation) or directly (steric contacts). Further structural and functional evidence will be required to fully understand and validate such models.

    Structure and role of the intracellular region

    With few exceptions (type VIII RPTPs that are catalytically inactive and RPTPα, where both domains have catalytic activity) the phosphatase activity of RPTPs is restricted to the membrane-proximal (and sometimes only) domain, termed D1 [46]. Its size, fold and catalytic mechanism are essentially the same as described above for cytoplasmic PTPs. Unusual features are the KIM present in type VII RPTPs (also discussed above in the context of MAP kinase phosphatases, since both PTPRR and PTPN5 can be expressed as receptor-like and cytoplasmic variants) and an ∼100 residues-long juxtamembrane region (sometimes referred as ‘cadherin-like’, although the similarity is very low) in type IIb RPTPs, a putative docking station for intracellular proteins.

    The D1 crystal structures of RPTPα [65] and RPTPμ [66] ignited a controversy still to be settled regarding the role of an N-terminal ‘wedge’ motif (a helix-turn-helix stabilized by a two-stranded β sheet, Fig. 2B). RPTPα D1 crystallized as a dimer in three different space groups, with the wedge motif of one monomer occluding the catalytic site of a dyad-related molecule. This observation received support from functional studies in both RPTPα and CD45 [46], but never replicated in any of the subsequent D1 or D1 + D2 structures. Two structural genomics initiatives, the Oxford-based Structural Genomics Consortium (http://www.sgc.ox.ac.uk/) and the New York Structural Genomics Research Consortium (http://www.nysgrc.org/) have targeted intracellular RPTP domains and, to date, elucidated most of the D1 structures: all have the wedge-like motif but the catalytic sites were never occluded.

    Four structures of full-length (D1 + D2) intracellular regions have been reported to date: LAR [67], CD45 [68], RPTPσ (NYSGC, PDB ID: 2fh7) and RPTPγ (Oxford SGC, PDB ID: 2nlk). The LAR and CD45 structures clearly exclude the wedge-based model of RPTP inhibition because of steric constraints [67,68]. The RPTPσ architecture is very similar to LAR (rmsd ∼1.53 Å over 537 equivalent Cα atoms), while RPTPγ, interestingly, reveals a considerable difference in the D2 domain orientation relative to CD45 and LAR. Examination of the RPTPγ crystal packing reveals a novel (putative) dimeric arrangement, supported by an extensive interface (∼1200 Å2 per monomer), with the β10–β11 loop of D2 appearing to block the catalytic site of D1 in a ‘head-to-tail’ arrangement (Fig. 2B). The functional relevance of this arrangement is currently being assessed (A Barr, personal communication).

    Importantly, the work on LAR intracellular region [67] revealed for the first time the structure of a D2 domain and the relative D1/D2 domain arrangement (Fig. 2B, tandem phosphatase domains are a unique feature of RPTPs). Both domains have the same tertiary fold, with an rmsd of 1.3 Å between all equivalent Cα positions. The two domains are connected by a short, four-residue linker and an extensive network of interactions stabilizes the interdomain interface. The active site architecture of the two domains is very similar to each other and to other known PTPs, yet despite the conservation of all key residues in the PTP-loop, LAR D2 exhibits < 0.001% of the enzyme activity. The structure pointed towards just two residues that might have been responsible for this low activity, one in the WPD-loop and the other in the KNRY phosphotyrosine recognition loop. Site-directed mutagenesis of these residues (Gly1779Asp and Leu1644Tyr) spectacularly restored activity to D1 levels, and the authors propose that, under physiological conditions, LAR D2 may indeed be active on specific substrates [67]. However, this cannot be extrapolated to all RPTP D2 domains: in cases such as RPTPζ and RPTPγ, for example, the catalytic Cys residues are mutated to Asp, abrogating enzymatic activity.

    The CD45 intracellular region structure [68], although maintaining the overall LAR architecture, revealed further details about the function of D2. First, the CD45 D2 is, most likely, not an active phosphatase despite preserving the catalytic Cys. The large number of substitutions in the PTP signature motif, including the essential Arg residue, results in a significantly altered shape of the active site pocket that impairs binding to the phosphoryl group [68]. Moreover, mutation of the general acid/base Asp residue in the WPD-loop to Val, and of the conserved Tyr residue in the phosphotyrosine recognition loop to Asn, explain why the CD45 D2 cannot be an active phosphatase under any conditions. Nevertheless, the CD45 D2 structure reveals features not present in any PTP domain: a 20 residue ‘acid’ loop between the β1 and β2 strands and an 11-residue ‘basic’ loop between the α3 helix and the β12 strand (Fig. 2B). Both loops (largely disordered in the crystal structures) are located in the proximity of the D1 active site and, as such, are likely to play an important role in substrate recognition and binding.

    Another important mechanism in the regulation of RPTP signalling is the reversible oxidation of the catalytic Cys residue by reactive oxygen specie) [36,46], also discussed above for cytoplasmic PTPs. Work on RPTPα, the best studied example in this respect, revealed unexpectedly that the D2 catalytic Cys (Cys723) appears to be more sensitive to oxidation than the D1 counterpart [69]. Moreover, oxidation of Cys723 results in the stabilization of observed full-length RPTPα dimers and, presumably via a conformational change, induces a relative rotation of the two molecules in the dimer that is detectable on the extracellular side of the receptor [70]. Crystallographic analysis revealed important consequences of the RPTPα D2 oxidation [71]. Cys723 forms a five-atom ring structure, termed cyclic sulfenamide, with the main chain nitrogen of the adjacent Ser724, showing similarity to the PTP1B case described above. This oxidation is associated with conformational changes in the PTP loop, although not to the extent observed in PTP1B [37,38,71], which adopts an open conformation. It is still unclear, however, how this change can cause the reactive oxygen species-mediated stabilization of RPTPα dimers and influence the relative orientation of receptor dimers. Further work, perhaps in the context of a full intracellular region or involving crystallization of a pre-oxidized RPTPα D2 (as opposed as oxidation within the crystals as performed previously [71], where the lattice may constrain movements) will be required in order to better understand the oxidation-dependent regulation of RPTPα.

    RPTPs as drug targets

    The RPTPs are prime targets for drug design, given the importance of phosphotyrosine signalling at the plasma membrane. The steady increase in understanding of their biology and specific role in various diseases (see accompanying review by Hendriks et al.) [46,72] have strongly driven both structural genomics (see above) and structure-based drug design efforts. The state of structure-based development of membrane-permeable RPTP inhibitors (a focus of the pharmaceutical industry) is difficult to gauge at the moment, given the novelty and proprietary nature of such work. Nevertheless, a recent report on RPTPβ revealed how engineering of the catalytic domain, based on its structure, has greatly helped crystallization of inhibitor complexes [73]. Considering the structural diversity of RPTP ectodomains and the potential of controlling the RPTP catalytic activity from the extracellular side via ligand interactions with the ectodomain, it is likely that the focus of structural efforts in the future will shift to this region.

    Class II PTPs (LMW-PTP)

    Low molecular weight PTPs are a family of small enzymes (18 kDa) involved in the regulation of cell growth, adhesion and cytoskeleton organization in mammalian cells. They share very low sequence homology to the rest of PTPs, except for the consensus active site motif CX5R, and a similar mechanism of catalysis where the key elements include the catalytic residues Cys12human and Arg18human in the P-loop and the general acid Asp129human in the DPYY-loop (analogous to the WPD-loop).

    Human LMW-PTP is expressed in most tissues [74] and is the predominant tyrosine phosphatase expressed in lens [75]. Two main active isoforms, A and B, have been characterized that originate by alternative splicing [76]. A third inactive isoform, C has also been reported [77,78] that lacks part of the catalytic domain. The two active isoforms differ only in the region spanning residues 40–73, also known as the ‘variable region’.

    Structure similarities albeit no sequence homology to classic PTPs

    A main difference from classic PTPs is that in LMW-PTPs, the P-loop is found at the N-terminus of the protein (residues 12–19), whereas in the rest of PTPs it is found towards the C-terminus. The overall fold of the LMW-PTP, as first revealed from X-ray crystallographic structures of the bovine and human enzymes [79,80] displays a twisted central parallel β sheet with four strands and five α helices packed on both sides, reminiscent of a classic dinucleotide-binding or Rossmann fold. This fold is similar to the one described for class I PTPs, except that it is smaller, representing the minimal size for a functional protein phosphatase enzyme. Like in other PTPs, the active site P-loop lies at the bottom of a crevice and connects the C-terminus of the β1 strand with the N-terminus of the α1 helix (Fig. 3A) Although the amino acid sequence in this loop lacks the Gly residues conserved in the canonical phosphate-binding motif, GXGXXG, it adopts a similar conformation with all the backbone amide groups oriented toward the phosphate ion (Fig. 1). This is possible because Asn15 (conserved in all LMW-PTPs) adopts a left-handed helical conformation stabilized by a hydrogen bond network with three other conserved residues, Ser19, Ser43, and His72. Thus, Asn15, Ser19, His72, and Ser43 serve structural functions that allow the active site to adopt an optimal geometry for substrate binding and transition state stabilization. Supporting evidence was provided by site-directed mutagenesis and kinetic measurements [81]. Mutations at Ser19 result in an enzyme with altered kinetic properties, impaired catalysis and changes in the pKa of the neighbouring His72. It was proposed that Ser19 acts to facilitate the ionization and orientation of Cys12 for optimal reaction as a nucleophile and as a leaving group. The Asn15 to Ala mutation appears to disrupt the hydrogen-bonding network, with an accompanying alteration of the geometry of the P-loop. The X-ray structure of the S19A mutant enzyme shows that the general conformation of the P-loop is preserved. However, changes in the loop containing His72 result in a displacement of the His72 side chain that may explain the shift in the pKa and the altered kinetic behaviour of the mutant enzyme [82].

    Details are in the caption following the image

    Structure of human LMW-PTP. (A) Labelled are the active site P-loop, the variable region V-loop and the DPYY-loop that contains the catalytic Asp residue and the phosphorylation sites Tyr131 and Tyr132 [46]. (B) Detail of the human LMW-PTPA (PDB accession code 5PNT) with a molecule of MES bound into the active site. Sticks represent the catalytic residues Cys12, Arg18, Asp129. A second Cys residue in the P-loop important in redox regulation, Cys17 is also labelled. The side chains of Tyr131 and Tyr49 stack against the ring of the MES molecule stabilizing its binding in the active site. All figures were prepared using pymol (2003, DeLano Scientific Ltd, Palo Alto, CA, USA).

    The variable region forms a long loop connecting β2 and α2 and surrounding the P-loop (Fig. 3A). Despite the sequence variation between the A and B isoforms in this region, the overall conformation is greatly conserved and contains two consecutive β turns (G. Redshaw, unpublished data). This characteristic and unique β turn tandem is also present in the structures of bacterial and yeast LMW-PTPs. The first β turn in the tandem contributes to form a deep active site cleft, analogous to the structural role of the tyrosine recognition loop (the KNRY motif) in class I PTPs.

    Structures of the A and B isoforms

    Site-directed mutagenesis, steady-state kinetics, and effector studies of the two human LMW-PTP isoenzymes (A form, HCPTPA, and B form, HCPTPB) indicated that residues 49 and 50 play important roles in determining the specificities and activity of the two enzymes [83,84]. These two residues, Trp49 and Asn50 in the B form, are located in a loop (the variable loop or V-loop) at the outer rim of the active site (Fig. 3A). Different responses of the A and B forms of the enzyme toward activators and inhibitors have also been reported [85,86]. For example, residues 49 and 50 are involved in the strong activation of the B form by guanosine and cGMP. Mutations of Trp49 to its equivalent in the A form, a tyrosine, and of Asn50 to a glutamate, result in enzymes with kinetic properties of the A form for cGMP activation [86]. The molecular basis for such differences in kinetic properties is still unclear. The differences in substrate specificity of both isoforms could be explained, in part, by the different charge distribution around the opening of the active site, where a negative patch created by Glu50 in the A form, is substituted by positive patch with Arg53 in the B form [80]. Evidence for the importance of residues at position 50, 53 and 49 in providing substrate selectivity has been also reported for the rat LMW-PTP, ACP1[87].

    Structures of LMW-PTP with different ligands have been reported. The B form was crystallized with phosphate and vanadate [79,88] and showed that both inhibitory compounds bind in the active site forming hydrogen bond interactions with the amide nitrogens in the P-loop and the catalytic Arg residue. The B form bovine enzyme (BPTP) and the human A form (HCPTPA) were also crystallized in the presence of HEPES and MES, aryl sulfonate inhibitory compounds that fill the active site with the sulfonate group bound in the P-loop [80]. In the HCPTPA structure, a molecule of MES is bound to the active site with the sulfonate group sitting in the phosphate-binding pocket and with the aryl ring stacking between Tyr131 and Tyr49 at the opening of the active site (Fig. 3B). Similarly, in the BPTP structure, the sulfonate group of a molecule of HEPES sits in the active site forming hydrogen-bond interactions with the amide groups of the P-loop and the catalytic Arg18. The six-member ring of HEPES packs parallel to the ring of Tyr131 and perpendicular to Trp49. Trp49/Tyr49 appears to act as a gatekeeper residue whose aromatic ring fluctuates between a closed position at the entrance of the active site when only small ligands (sulfate, phosphate) are present, to an open position when larger ligands like MES, HEPES are bound. Stacking interactions of Tyr131 and Trp49/Tyr49 with the aryl ring are presumably similar to those expected for a phosphotyrosine residue from a biological substrate. The structure of the yeast LMW-PTP (LTP1) with HEPES or p-nitrophenyl phosphate (pNPP) molecules bound to the active site shows again packing of the aryl rings with an aromatic side chain Trp134, analogous to Tyr131 [89].

    Aromatic residues are conserved at position 131 for all bacterial and eukaryotic LMW-PTPs, with preference for Tyr (Phe in Bacillus subtilis and Trp in Saccharomyces cerevisiae), suggesting that the stacking interactions with the substrate are important in all LMW-PTPs. On the other hand, the aromatic residues at position 49 (Trp/Tyr) found in eukarytotic LMW-PTPs, are not strictly conserved in prokaryotes. A subtype of bacterial LMW-PTPs (type II), present a large variation of residues at this position. For instance, B. subtilis YwlE has Ser42 [90] and Escherichia coli Wzb has Leu40 [91] instead of Tyr or Trp residues. These changes imply a different mechanism of substrate recognition and specificity, which could also explain why these phosphatases show a much lower binding constant towards pNPP than the eukaryotic enzymes [91]. In addition, the prokaryote type II LMW-PTPs do not have Tyr at position 132 as found in the eukaryote enzymes. This residue is critical in regulation by phosphorylation of both the enzymatic activity and the localization of the eukaryotic LMW-PTP. This suggests that the type II bacterial LMW-PTPs may have different ways of regulation.

    Structural determinants for regulation of catalytic activity in LMW-PTPs

    Oxidation of the catalytic cysteine thiol inhibits enzymatic activity of the LMW-PTP in a reversible manner [92] (as described for class I PTPs). Both NO and H2O2 can inactivate LMW-PTP in vitro. Both cysteines in the active site, Cys12 and Cys17, are essential for this reversible regulation and it was proposed that they could form a disulfide bridge, given their proximity observed in the crystal structures. This regulatory mechanism observed in vitro explains the inactivation by intracellular reactive oxygen species generated in vivo during growth factor stimulation [92]. The enzymatic activity of LMW-PTP is also regulated by phosphorylation of Tyr131 and Ty132. Phosphorylation of LMW-PTP by v-src kinase in vitro leads to enzymatic activation 1.5 to 2-fold [83]. In T cells, phosphorylation by Lck and Fyn tyrosine kinases causes activation of LMW-PTP [93] also by twofold. Further studies indicated that Tyr131 is the main regulatory site for modulation of the enzyme activity, whereas Tyr132 may serve as a docking site for Grb2 binding [94]. Tailor and co-workers also showed that LMW-PTP auto-dephosphorylates upon activation of T cells with subsequent inactivation of the enzyme representing a feedback control of its biological function. Another proposed mechanism of self-regulation for LMW-PTP implicates oligomerization of the enzyme. LMW-PTP forms dimers both in solution and in the crystal structures [82,95,96]. In the crystallographic dimer, Tyr131 and Tyr132 from one monomer insert into the active site of the second monomer. Furthermore, Tyr131 forms a hydrogen bond to Asn50 and Tyr132 forms a hydrogen bond to Asp129, the general acid in catalysis, therefore occluding access of any substrate to the active site and compromising essential catalytic residues. Since the activity of this enzyme is modulated by phosphorylation at Tyr131 and Tyr132, it was hypothesized that the structure of this dimer may provide a model for a self-regulatory mechanism of LMW-PTPs in which the enzyme can serve as its own substrate for dephosphorylation, and modulate the access of other substrates to the active site [82]. This implies that phosphorylation of Tyr131 and Ty132 is incompatible with the dimer formation, and it will favour the monomeric form of higher activity. Interestingly, NMR relaxation studies defined a second oligomerization interface that mediates the formation of tetramers [97]. This interface overlaps with the variable region, presumably important in substrate binding and specificity [80]. Evidence on the existence in vivo of two activation states for LMW-PTP has been reported [98]. LMW-PTP exists in two pools, a cytosolic non-phosphorylated form associated with the PDGFR and possibly other tyrosine kinase receptors, and a phosphorylated form, of higher catalytic activity, associated with the cytoskeleton via p190RhoGap [99]. These findings highlight the importance of a combined mechanism of regulation for this enzyme, involving phosphorylation, dimerization and substrate location.

    LMW-PTPs as a drug targets

    Human LMW-PTP is critical in the regulation of mitogenic signalling and Rho-mediated cytoskeletal rearrangements after platelet derived growth factor stimulation [98,100]. LMW-PTP is responsible for dephosphorylation and regulation of several protein tyrosine kinase receptors with subsequent downregulation of signalling pathways mediated by the epidermal growth factor receptor [101], the insulin receptor [102], the vascular endothelial growth factor receptor [103] and the fibroblast growth factor [104]. In addition, LMW-PTP has been shown to play a positive role in T-cell receptor signalling by dephosphorylation of ZAP-70 [105], and to promote cadherin-mediated cell adhesion [106]. Recent reports suggest a role for LMW-PTP in tumour onset and growth [107], probably through dephosphorylation of the EphA2 receptor [108]. The role of LMW-PTP in such important cellular processes and its oncogenic potential suggests it may be a good target for drug design. Some attention has been given to the design of small inhibitors of LMW-PTPs. The first inhibitors of the human enzyme were designed based on the binding of adenine in the active site of the S. cerevisiae LTP1 and the inhibitory effect of adenine on the human A form (HCPTP-A). Those compounds exhibit poor inhibition constants (in the mm range) although they show some selectivity between A and B form [109]. More recently, a virtual screen protocol identified a group of competitive inhibitors of the bovine LMW-PTP, some of them with activities in the low micromolar range [110]. Docking of the best inhibitors of this group shows how they exploit specific interactions and shape complementarity of a secondary site next to the active site (in the variable region) that contains Ser47. These results indicate that further development of double-site binders, as for PTP1B, could improve specificity and selectivity against this enzyme [110]. The prokaryotic LMW-PTPs were also actively screened to identify suitable inhibitors. Particular attention has been directed to the M. tuberculosis enzyme (MptpA) as a promising new target against tuberculosis [111] with some encouraging results [112].

    Concluding remarks

    Significant advances have been made on the structural front since the first PTP structures were reported more than 10 years ago. Most of the earlier work was focused on understanding mechanistic issues related to enzymatic dephosphorylation and the role of proposed catalytic residues, like Cys, Arg and the general acid Asp. A major effort was also invested in producing structures of complexes with inhibitory compounds, mostly of PTP1B. However, less information is available on the structural basis for substrate recognition and specificity by PTPs. The biological complexity of phosphorylation-dependent systems requires a better understanding of the molecular interactions of phosphatases with their different regulatory partners, rather than the traditional single protein approach. To date, complexes of PTPs with full-length biological targets remain highly elusive and thus represent the main challenge for the future. A number of important questions in RPTP signalling, that can be addressed by structural methods, remain to be answered. For example, it would be of great interest to see the crystal structure of a full RPTPα intracellular region and, hopefully, settle the ‘wedge’ dispute. Similarly, understanding the multiple interactions that can take place between ectodomains, transmembrane and intracellular regions, will require structures of full-length receptors, alone and in complex with ligands. Although this is currently at the edge of available technology, it is expected that in the near future progress in this direction will take place. Finally, RPTPs, like most cell surface receptors, never work alone. They must be part of multi-molecular complexes, involving extracellular, intracellular and other membrane proteins. Structural characterization of such assemblies represents a major target for the future.

    Acknowledgements

    We thank Alastair Barr and Gavin Redshaw for sharing information prior to publication. This work was supported in part by European Research Community Fund MRTN-CT-2006-035830.