The receptor PTPRU is a redox sensitive pseudophosphatase

The receptor-linked protein tyrosine phosphatases (RPTPs) are key regulators of cell-cell communication through the control of cellular phosphotyrosine levels. Most human RPTPs possess an extracellular receptor domain and tandem intracellular phosphatase domains: comprising an active membrane proximal (D1) domain and an inactive distal (D2) pseudophosphatase domain. Here we demonstrate that PTPRU is unique amongst the RPTPs in possessing two pseudophosphatase domains. The PTPRU-D1 displays no detectable catalytic activity against a range of phosphorylated substrates and we show that this is due to multiple structural rearrangements that destabilise the active site pocket and block the catalytic cysteine. Upon oxidation, this cysteine forms an intramolecular disulphide bond with a vicinal “backdoor” cysteine, a process thought to reversibly inactivate related phosphatases. Importantly, despite the absence of catalytic activity, PTPRU binds substrates of related phosphatases strongly suggesting that this pseudophosphatase functions in tyrosine phosphorylation by competing with active phosphatases for the binding of substrates. Receptor-linked protein tyrosine phosphatases (RPTPs) usually contain an active membrane proximal domain and an inactive pseudophosphatase domain. Here, the authors characterize an RPTP with two pseudophosphatase domains, providing evidence that it may act as a decoy receptor for substrate sequestration.

T he human classical protein tyrosine phosphatases (PTPs) are key signalling regulators that work with kinases to finetune cellular levels of phosphotyrosine (pTyr), impacting on multiple cellular pathways including metabolism, differentiation, and adhesion [1][2][3] . PTPs do not simply function as negative regulators of tyrosine kinases to reverse protein phosphorylation, instead it is becoming clear that PTPs work in synergy with kinases to regulate complex cell signalling pathways and are important therapeutic targets in diseases such as cancer and diabetes 4,5 . The 37 classical PTPs exhibit diverse domain architectures and subcellular localizations, but all share a conserved core catalytic C-X 5 -R motif, known as the PTP loop, which includes the essential cysteine that catalyses the nucleophilic attack on the substrate phosphate group 6,7 . Less well conserved are the three additional motifs that form the PTP active site 7 . The WPD loop contains an aspartate residue that acts as a general acid/base during different steps of the catalytic cycle and assists in substrate binding. The pTyr recognition loop, typically containing the amino acid sequence KNRY, is so called because it forms the deep pocket that imparts selectivity for pTyr over smaller phosphorylated amino acids, such as serine and threonine. In addition to defining the shape of the binding pocket, the tyrosine in the pTyr recognition loop plays a crucial role in substrate orientation as its sidechain packs against the substrate pTyr phenyl ring. Finally, the Q loop positions and activates a water molecule for the hydrolysis of the phosphocysteine intermediate complex.
Despite the importance of catalysis for the function of many PTPs, there are numerous reports of non-catalytic functions [8][9][10] . Moreover, 5 of the 37 classical PTPs have been reported to be catalytically inactive against generic phospho-substrates, such as pNPP, DiFMUP and phosphopeptides. These include the nonreceptor PTPs: PTPN23 (HD-PTP) 11,12 , PTPN14 (PTPD2) and PTPN21 (PTPD1) 11,13 and the receptor PTPs: PTPRN (PTPIA2) and PTPRN2 (PTPIA2β) 14 . These PTPs contain altered sequences in their catalytic motifs and substrate binding loops that are predicted to impair catalytic activity, defining them as putative pseudoenzymes 15 . For example, PTPN23 (HDPTP) has an incomplete Q loop and serine substitution within the PTP loop, PTPN21 (PTPD1) possesses an altered WPD motif and PTPN14 (PTPD2) has a variant pTyr recognition loop 11,13 . However, it is noteworthy that activity against specific protein substrates has been reported for some of these PTPs, raising the possibility of noncanonical activation mechanisms 16,17 . Beyond the non-receptor PTPs, 12 of the 21 cell surface receptor PTPs possess highly conserved membrane distal pseudophosphatase D2 domains, which have been implicated in substrate recognition, redox sensing and enzyme regulation 3,18,19 . Interestingly, changes in the catalytic motifs can also alter substrate specificity. For example, a glutamate substitution in the WPD motif of the single PTP domain of PTPRQ (PTPS31) determines its selectivity for phosphoinositides over pTyr 20 . These examples illustrate the importance of combining functional and structural studies to characterise the catalytic properties of putative pseudophosphatases.
PTPRU is a member of the R2B receptor family, which includes PTPRK, PTPRM and PTPRT, characterised by large extracellular domains that mediate homophilic interactions and tandem intracellular PTP domains (Fig. 1a) 21 . PTPRU is expressed during development 22,23 , and is reported to function during zebrafish gastrulation 24 and chick midbrain development 25 . Interestingly, while PTPRK and PTPRT are reportedly tumour suppressors, PTPRU has been proposed to play an oncogenic role in gastric cancer and glioma cells 26,27 . Opposing reports suggest that both PTPRU overexpression and knockdown can lead to β-catenin dephosphorylation 26,28,29 . Recently, the PTPRK substrate Afadin 3 was identified as a PTPRU interactor 30 , implicating it as a cell adhesion regulator. The membrane proximal (D1) domains of PTPRK, PTPRM, and PTPRT are active tyrosine phosphatases 11 , however, the catalytic activity of PTPRU has not been determined. PTPRU possesses evolutionarily conserved sequence changes to key catalytic motifs including non-canonical WPD (WPE) and pTyr recognition loop sequences (GSRQ rather than KNRY), as well as a unique threonine in the PTP loop ( Fig. 1b and Supplementary Figs. 1 and 2). To better understand the function of PTPRU we set out to determine whether it is an active phosphatase.
Here, we demonstrate through biochemical and structural studies that PTPRU is unique amongst the receptor-linked protein tyrosine phosphatases (RPTPs) in possessing two pseudophosphatase domains. The crystal structure of the PTPRU D1 domain reveals substantial structural rearrangements to key catalytic loops such that the shape of the pTyr binding pocket is lost and the active site cysteine is occluded. Despite lacking catalytic activity, PTPRU can recruit substrates of catalytically active paralogs supporting a model where PTPRU functions as a scaffold to compete for binding to protein substrates. Thus, the levels of the different R2B RPTPs expressed at the plasma membrane will determine the local pTyr level of a subset of cell junction regulators.

Results
PTPRU is catalytically inactive. To determine the consequence of sequence variations on phosphatase activity, we expressed and purified the recombinant PTPRU D1 domain in E. coli for use in in vitro phosphatase assays. As a positive control we used the D1 domain of its closest paralog, PTPRK (Supplementary Table 1). The generic substrate 4-nitrophenyl phosphate (pNPP) was used in initial activity assays. The K m and k cat for PTPRK-D1-mediated pNPP hydrolysis were 16.16 ± 1.32 mM and 4.50 ± 0.17 s −1 , respectively, similar to previously determined kinetic parameters for the prototypic phosphatase PTP1B using this substrate (Fig. 1c) 31 . Strikingly, we were unable to detect activity for PTPRU-D1 against pNPP, even at high enzyme concentrations (up to 25 μM) or when using an extended assay duration (Fig. 1d), whereas PTPRK-D1 activity was readily detectable in the low nanomolar range (Fig. 1d). To investigate whether a cellular cofactor might be necessary for PTPRU activity, we performed dephosphorylation assays using quenched pervanadate-treated cell lysates (see "Methods"). These lysates are enriched in tyrosine phosphorylated proteins that serve as substrates for recombinant PTP domains. Again, whilst incubation with PTPRK-D1 for 16 h at 4°C resulted in visible dephosphorylation of total cellular pTyr, PTPRU-D1 showed no activity and is comparable to the inactive PTPRK D2 domain (Fig. 1e). One possible explanation for the inactivity of PTPRU could be a requirement for its D2 domain. Previous studies on other receptors have shown that interactions between the two domains can impact D1 activity 3,32 . Unfortunately, we were unable to purify the full PTPRU intracellular domain (ICD) from bacteria. Therefore, we generated N-terminal flag-tagged PTPRK and PTPRU ICD mammalian expression constructs, encompassing the juxtamembrane, D1 and D2 domains (Fig. 1f). ICDs were immunoprecipitated (IP) from transfected HEK-293T lysates and subjected to pNPP assays ( Fig. 1g and Supplementary Fig. 3a). Whilst the PTPRK-ICD, but not an inactivating cysteine mutant (C1089S), was able to hydrolyse pNPP, PTPRU and the corresponding cysteine mutant showed no activity above mock IPs after 2 h ( Fig. 1h and Supplementary Fig. 3b). Taken together, these data suggest that unlike PTPRK, the D1 domain of PTPRU has no intrinsic PTP activity.
Due to the highly divergent nature of the PTPRU pTyr recognition loop, we investigated whether PTPRU may have altered substrate specificity. PTPRU displayed no phosphatase activity against phosphoserine (pSer) or phosphothreonine (pThr; Supplementary Fig. 4a). Notably, the PTPRQ D1 domain has the same non-canonical aspartate to glutamate substitution in the WPD-loop observed in PTPRU ( Supplementary Fig. 1), and has been previously reported to be a phosphoinositide phosphatase 20 . However, PTPRU-D1 exhibited no measurable dephosphorylation of either phosphatidylinositol (PI)-4-phosphate or PI-4,5bisphosphate ( Supplementary Fig. 4b). While our investigation of protein phosphorylation was primarily focused on residues modified with an O-linked, phosphoester bond (pTyr, pSer, pThr), a recent mass spectrometry study has reported residues modified by an N-linked, phosphoramidate bond (pHis, pLys, pArg, pAsp) as having greater abundance in the cell than pTyr 33 . Indeed, human histidine phosphatases have been identified previously [34][35][36] . To test the ability of PTPRU to catalyse hydrolysis of phosphoramidate bonds we used the generic histidine phosphatase substrate imidodiphosphate (PNP; Supplementary Fig. 4c) 37 but were again unable to detect any activity ( Supplementary Fig. 4d). Together these data indicate that PTPRU does not catalyse the hydrolysis of phosphoester or phosphoramidate-based substrates.
In the absence of an identifiable substrate, we sought to determine the molecular mechanism of PTPRU-D1 inactivity.  Substitution of the WPD-loop aspartate for glutamate, as seen in PTPRU, is common amongst the D2 pseudophosphatase domains of RPTPs ( Supplementary Fig. 5a) 7 . Previously, PTP1B activity was reduced by several orders of magnitude upon mutation of the corresponding aspartate to glutamate (D181E) 38 . Similarly, mutating the WPD loop in PTPRK-D1 (D1057E) results in a~115-fold reduction in enzymatic turnover versus WT, having low residual activity with a k cat = 0.0389 s −1 (Fig. 1i). Critically, PTPRU-D1.E1053D, where the Glu has been reverted to the canonical Asp, remained inactive against cellular pTyr and pNPP (Fig. 1i, j). These data therefore suggest that an Asp to Glu substitution in the WPD loop alone is insufficient to account for complete loss of PTPRU-D1 enzymatic activity.
Structure of PTPRU-D1. In order to better understand the mechanisms underlying the lack of catalytic activity of PTPRU, we determined the structure of the D1 domain. The X-ray crystal structure of PTPRU-D1 was solved by molecular replacement using the PTPRK-D1 domain 39 (PDB ID: 2C7S) and refined to 1.72 Å resolution ( Table 1). The overall fold of PTPRU-D1 closely resembles related phosphatase domains (RMSD of 1.0 Å 2 over 237 Cα atoms with PTPRK, Fig. 2a). However, the PTPRU structure reveals several key differences that likely contribute to its catalytic inactivity. The most striking difference is the absence of an ordered pTyr recognition loop (Fig. 2a). The structure and sequence of this loop is well conserved across the PTPs ( Fig. 2b and Supplementary Fig. 5b), with the key tyrosine of the 'KNRY' motif creating an active site pocket deep enough to exclude pSer/ Thr residues, whilst favourably stacking with the phenol ring of bound pTyr substrates (Fig. 2c). In the PTPRU crystal structure residues 904-925 encompassing the pTyr recognition loop region are disordered, resulting in almost complete loss of the pocket that would normally bind the pTyr substrate.
Another striking difference observed in our PTPRU-D1 structure was the conformation of the PTP loop containing the catalytic cysteine (Fig. 2d). The conformation of this loop is highly conserved in classical PTPs (Fig. 2di), but in PTPRU-D1 this loop is arranged such that the sidechain of T1089 is in close proximity to the active site cysteine (3.0 Å between T1089 OG1 and C1085 SG atoms, Fig. 2dii). The T1089 sidechain forms a hydrogen bond with the backbone amide of R1091 capping the adjacent helix ( Supplementary Fig. 6a, b). A threonine at this position in the PTP loop is unique to the PTPRU-D1 domain and the more common hydrophobic residues at this position (Ala, Val, and Ile) are not capable of forming an equivalent interaction ( Supplementary Fig. 1). This new loop orientation blocks the catalytic cysteine and would directly interfere with pTyr binding (Fig. 2e). The combined effect of a disordered pTyr recognition loop and reorientation of the catalytic PTP loop is the loss of key structural features normally required for binding and processing of phosphorylated substrates (Fig. 2c, e).
An additional loop adjacent to the active site (C1121 to M1127) also adopts a conformation that differs from the canonical fold (Fig. 2dii lower left and Supplementary Fig. 6c). In all available structures, this loop is stabilised via hydrogen bonds between a conserved arginine and backbone carboxyl groups in this loop (R1119 in PTPRK, Supplementary Fig. 6d). Fig. 1 The PTPRU D1 domain does not dephosphorylate pTyr. a Schematic diagram of the R2B family RPTP domain structure. b Multiple sequence alignment of the 4 key PTP motifs across R2B family RPTPs, coloured by percentage identity (light-dark grey). Key variable residues in PTPRU are highlighted in yellow and essential PTP catalytic residues are marked by arrowheads. c Michaelis-Menten plot of initial rate vs. substrate (pNPP) concentration using 0.2 μM PTPRK-D1. Error bars represent ± SEM of n = 3 independent experiments. d Extended time course of pNPP dephosphorylation, monitored by absorbance at 405 nm, using low concentration (6.25 nM, orange, and 12.5 nM, red) PTPRK-D1 and high concentration (25 μM, blue) PTPRU-D1 recombinant proteins. pNPP substrate alone (black) is included as a control. e Immunoblot analysis of pervanadate-treated MCF10A lysates incubated with 0.  This arginine is completely conserved in all D1 sequences except PTPRU where it is uniquely a cysteine (C1121, Supplementary  Fig. 7). One consequence of this loss of an arginine is reorientation of the nearby methionine ( Supplementary Fig. 6c, d) and de-stabilisation of this loop. Residues C1121 to M1127 in PTPRU are not well ordered and challenging to build in a single, reliable conformation in the electron density suggesting it may adopt multiple conformations. In PTP1B (PTN1), the equivalent arginine (R254, Supplementary Fig. 7) has been suggested to form a secondary pTyr binding site via binding of a peptide containing tandem pTyr 40 . The importance of a cysteine residue in this secondary binding site, in a position resembling that of an active site cysteine, remains unclear. Consistent with the identified disorder in specific regions of the PTPRU-D1 structure, this domain has reduced thermal stability compared to PTPRK-D1, as shown by a lower global melting temperature (PTPRU-D1 = 48.2°C, PTPRK-D1 = 51.5°C; Fig. 2f).
Role of PTPRU motifs in activity and stability. To investigate the consequence of the observed structural rearrangements in the PTPRU PTP and pTyr recognition loops, we generated a series of point mutations in key residues, as well as chimeric D1 domains harbouring reciprocal substitutions of the PTPRU and PTPRK pTyr recognition loops (Fig. 3a). In dephosphorylation assays, introduction of the unique T1089 of PTPRU in to PTPRK (A1093T) was not sufficient to inactivate PTPRK-D1 (Fig. 3b). Further, removal of the active site threonine was insufficient to reactivate PTPRU-D1 and a tandem E1053D and T1089A also remained inactive (Fig. 3b). Introduction of the highly divergent  PTPRU-D1 pTyr recognition loop in to PTPRK-D1 results in loss of activity, however introduction of this loop from PTPRK in to PTPRU-D1 does not restore activity (Fig. 3b). The inability of these mutations to induce PTPRU-D1 activity was also confirmed in pNPP activity assays ( Supplementary Fig. 8). Furthermore, introduction of an E1053D mutation combined with the PTPRK pTyr recognition loop also does not result in any detectable PTPRU-D1 activity (Fig. 3b). Therefore, the inactivity of PTPRU-D1 cannot be explained simply in terms of any single change to the PTP, WPD or pTyr recognition loop.
To determine the role of the pTyr recognition loop in protein stability, we subjected WT and chimeric D1 domains to limited proteolysis with subtilisin. The PTPRU-D1 domain showed higher susceptibility to proteolytic cleavage than PTPRK-D1, as would be predicted due to the disorder of the pTyr recognition loop (Fig. 3c). Introduction of the PTPRK pTyr recognition loop into PTPRU-D1 does not confer greater resistance to proteolysis, suggesting that the PTPRK loop cannot adopt a folded conformation in the context of the PTPRU-D1 domain (Fig. 3d). Indeed, the PTPRK loop appears to further destabilise PTPRU-D1. Introduction of the PTPRU pTyr recognition loop into PTPRK-D1 does result in greater susceptibility to proteolytic cleavage, supporting that this loop is again unable to form a folded, protease-resistant conformation (Fig. 3e). These results are consistent with proteolysis using trypsin protease, confirming that any change in cleavage is not caused by an altered number of proteolytic cleavage sites introduced when generating chimeric sequences ( Supplementary Fig. 9).
Structure of oxidised PTPRU-D1. In an attempt to determine if substrate binding might induce folding of the pTyr recognition loop or rearrangement of the catalytic PTP loop, we soaked PTPRU-D1 crystals with several potential ligands including PO 4 , pTyr, and PNP. In none of the datasets collected for these crystal soaks was there any evidence of ligand binding in the active site or any induced folding of the pTyr recognition loop. However, these crystals, collected 4 weeks after the initial datasets, had clearly undergone oxidation resulting in the formation of a disulphide bridge between the highly conserved catalytic C1085 and the vicinal "backdoor" C998 cysteines ( Fig. 4ai and Supplementary Fig. 2). This alternate conformation involving disulphide bond formation with nearby cysteines has been observed for several other related phosphatases (Fig. 4aii) [41][42][43][44] . Disulphide bond formation has been proposed to protect the catalytic cysteine from oxidative damage and/or function as a redox- sensitive mechanism for reversible PTP inactivation [41][42][43][44] . The formation of this disulphide in PTPRU-D1 destabilises the conformation of adjacent residues in the PTP-loop as there is no clear density in which to model S1086-G1088 (Fig. 4b). The nearby loop (C1121-M1127) described previously as reoriented in the reduced form, is further destabilised in this oxidised structure as evidenced by a lack of electron density for this region. Nonreducing SDS-PAGE of PTPRU-D1 recombinant protein in the presence of hydrogen peroxide results in a mobility shift consistent with disulphide formation in solution, which is completely reversed under reducing conditions (Fig. 4c). While PTPRK-D1 conserves this "backdoor" cysteine, it does not undergo detectable disulphide formation under the same conditions (Fig. 4c). Thus, the catalytic cysteine of PTPRU-D1 can undergo reversible oxidation, involving intramolecular disulphide formation, identifying this domain as a redox-sensitive pseudophosphatase.
PTPRU interacts with PTPRK substrates. To understand the role of PTPRU in signalling, we exploited our previously reported observations based on D1 and D2 domain-swapping chimeras showing that the PTPRK, but not PTPRM, D2 domain was critical for recognition of Afadin 3 , a reported PTPRU interactor 30 . We generated an in vivo biotinylated chimera consisting of the active PTPRK-D1 and the PTPRU-D2 domain (Fig. 5a and Supplementary Fig. 10). We then tested the ability of chimeric proteins to bind and dephosphorylate PTPRK substrates. To probe protein-substrate interactions, we conjugated biotinylated chimeric proteins to streptavidin beads for in vitro pull-downs from pervanadate-treated cell lysates followed by immunoblotting. While the PTPRK and PTPRM substrate p120-Catenin (p120 Cat ) 3,45 could interact with all chimeras regardless of D2 domain, we found that unlike PTPRM-D2, the PTPRU-D2 domain is sufficient for binding to Afadin (Fig. 5b). Consistent with this interaction data, immunoprecipitation of tyrosine phosphorylated proteins from dephosphorylation assays confirmed the PTPRU-D2 domain as being sufficient to recruit Afadin for dephosphorylation by the active PTPRK-D1 domain (Fig. 5c). Our data suggest that in cells PTPRU will bind but not dephosphorylate PTPRK substrates.
Previously we have identified specific p120 Cat pTyr residues (Y228, Y904) which are dephosphorylated by PTPRK and PTPRM D1 domains, and are hyperphosphorylated in PTPRK-KO cells 3 . We confirmed by in-lysate dephosphorylation assays that while all PTPRK-D1-containing chimeras dephosphorylate pY228 and pY904, PTPRU-D1 cannot dephosphorylate these p120 Cat sites (Fig. 5d, e). To investigate the cellular consequence of PTPRU binding to PTPRK substrates, we generated CRISPR-Cas9 mediated PTPRU-KO MCF10A cells. As expected, we were able to observe hyperphosphorylation of p120 Cat -pY228 and -pY904 in PTPRK-KO cells vs WT (Fig. 5f). Strikingly, deleting PTPRU resulted in hypophosphorylation of both p120 Cat -pY228 and -pY904 vs WT levels (Fig. 5f, g). Taking our interaction and dephosphorylation data together, this supports a mechanism in which PTPRU can bind substrates and protect them from dephosphorylation by related phosphatases.

Discussion
The receptor PTPRU possesses several sequence variants in key catalytic motifs including a highly divergent pTyr recognition loop sequence, a unique Thr within the PTP loop and a Glu substitution in the WPD loop. Here we show that PTPRU does not exhibit detectable phosphatase activity against a range of substrates or in pTyr dephosphorylation assays with cell lysates. Our structural data identify multiple features that would disrupt both pTyr binding and catalytic activity. Despite its inactivity, we demonstrate that PTPRU can bind to key proteins previously reported as substrates for its catalytically active paralog PTPRK. This supports a role for PTPRU as a scaffold that competes with active phosphatases at the plasma membrane to locally influence tyrosine phosphorylation dynamics and potentially cell-cell adhesion.
Previous studies have shown that WPD to WPE mutations in several active phosphatases results in significant reduction in enzyme activity 13,38 . We show a similar substantial decrease in PTPRK activity following the introduction of the WPE sequence change. However, mutation of the PTPRU WPE loop to the canonical sequence (E1053D) was not sufficient to rescue any detectable phosphatase activity. Our structure of the PTPRU-D1 a b 46  reveals that there are two key structural changes within this domain that alter the pTyr binding pocket and therefore are likely to explain the lack of phosphatase activity. The first is the disordered pTyr recognition loop. The absence of electron density for this loop was unexpected as, although the PTPRU sequence is highly divergent, it does retain a conserved arginine (R918) that in related structures binds back into the main PTP fold and interacts with residues that are conserved in PTPRU-D1. Not only does loss of this ordered loop drastically alter the shape of the pTyr binding pocket, it also contributes to decreased protein stability as demonstrated by the increased susceptibility of PTPRU to proteolysis and its lower melting temperature relative to PTPRK-D1. The second key structural change in PTPRU-D1 relates to the catalytic PTP loop, which has undergone a substantial rearrangement resulting in the occlusion of the pTyr binding site, blocking the catalytic cysteine. Having identified that the sequence changes in PTPRU cause significant structural rearrangements to key loops required for enzyme activity we generated a series of point mutations and chimeric constructs to test their individual and combined effects on phosphatase activity. Using chimeric proteins where the pTyr loops of PTPRU and PTPRK were exchanged, we showed that although the PTPRU pTyr sequence was sufficient to inactivate PTPRK, the canonical pTyr loop sequence did not result in any phosphatase activity for 134 100 kDa p120 Cat -pY228 p120 Cat -pY228 p120 Cat -pY904 p120 Cat -pY904 p120 Cat p120 Cat p120 Cat -pY228 p120 Cat -pY904 p120 Cat p120 Cat the PTPRU chimera. Furthermore, the PTPRU-D1 K-pTyr chimera is even more destabilised, as shown by increased susceptibility to proteolysis, suggesting this loop cannot fold correctly to form a pTyr-binding pocket as seen for PTPRK and other PTPs. Despite lacking a pTyr-binding pocket, it remained a possibility that PTPRU could process phosphorylated substrates as it retains the catalytic cysteine. Therefore, we tested the impact on phosphatase activity of the Thr (T1089) that blocks this cysteine in the PTP loop. Interestingly, the introduction of a Thr into the equivalent position of the PTPRK-D1 does not reduce its phosphatase activity while mutation of the PTPRU Thr to Ala does not induce detectable PTPRU activity. Furthermore, combinations of multiple mutations that restore canonical PTP sequences to PTPRU including E1053D and T1089A or replacing the pTyr loop and E1053D were still unable to rescue any activity. This combination of biochemical and structural analysis demonstrates that there are multiple mechanisms contributing to the absence of phosphatase activity in PTPRU-D1. An intriguing feature of RPTP family inactive D2 pseudophosphatase domains is the retention of the catalytic cysteine residue, as seen in PTPRU (Supplementary Fig. 2). The conservation of this residue in inactive domains raises the question of whether it plays an alternative, non-catalytic role. Our structure of oxidised PTPRU-D1 demonstrates that this cysteine has the capacity to form a disulphide bond with a "backdoor" cysteine similar to that seen in the active SHP2, LYP, and PTEN phosphatases [41][42][43] . In these enzymes, the formation of a disulphide bond has been attributed to the need to protect the catalytic cysteine from irreversible oxidative damage or to allow reversible redox-sensitive inactivation 46,47 . Our observation here of a similar intramolecular disulphide in an inactive pseudophosphatase domain suggests that the proposed roles for the disulphide formation may extend beyond the modulation of enzyme activity. Importantly, oxidation of PTPRU and RPTP domains has been reported in cells 48 . Previous studies on the PTPRA D2 pseudophosphatase domain suggest oxidation can promote an intermolecular disulphide bond 18 , or a conformational change that is translated to the extracellular domain 49 . Thus, for several PTP domains catalysis is impaired yet redox sensing is preserved and warrants further investigation.
In addition to promoting a "backdoor" disulphide, PTP oxidation can induce chemical modification of the catalytic cysteine. One such modification is the formation of a sulfenyl-amide intermediate, as demonstrated for PTP1B 50,51 . This modification involves the sidechain of the catalytic cysteine forming a covalent link to the backbone nitrogen of an adjacent residue, resulting in a substantial change to the conformation of the catalytic PTP loop. Interestingly, this loop conformation is highly similar to that seen in PTPRU-D1. In PTP1B this renders the enzyme inactive but is reversible upon reduction and is proposed to be a protective intermediate during redox-regulated inhibition. This conformation of the PTPRU-D1 PTP loop is not induced by oxidation, it is instead present in the reduced form rendering the enzyme unable to bind pTyr. This suggests PTPRU has evolved to adopt an inactive conformation, even under reducing conditions. The ICDs of other members of the R2B family comprise an active membrane-proximal D1 domain and an inactive membrane-distal D2 domain. Despite the sequence divergence of PTPRU-D1 from its paralogs (Supplementary Fig. 1) and its lack of catalytic activity, this domain still possesses higher sequence identity to D1 domains than to D2 domains (69% sequence identity with R2B family D1 domains, 28% identity with R2B family D2 domains; Supplementary Table 1). Therefore, PTPRU retains a bona-fide D1 and D2 domain topology similar to that of related enzymes but with a D1 domain that has diverged to be catalytically inactive. By using chimeric ICDs containing D1 and D2 domains from PTPRU, PTPRK, and PTPRM in cell-based dephosphorylation assays we show that the D2 domain of PTPRU can recruit substrates for dephosphorylation by the active D1 domain of PTPRK. We previously showed that protein binding to the ICD of the related receptor PTPRK was phosphorylationindependent 3 , suggesting that the exclusion of pTyr by the occluded PTPRU-D1 active site would not necessarily inhibit binding of protein interactors in the context of the full PTPRU-ICD, consistent with what we observe. This ability to bind substrates that overlap with active phosphatases, combined with the lack of phosphatase activity of PTPRU suggests that the likely role of PTPRU in cells is to act as a decoy receptor that sequesters substrates protecting them from dephosphorylation. In support of this, we find that genetic deletion of PTPRU leads to a reduction in phosphorylation levels of the PTPRK substrate p120 Cat . In this way, PTPRU may modify cell signalling by altering the rate or extent of tyrosine dephosphorylation by related, active RPTP family members. The absence of phosphatase activity demonstrated here for PTPRU does not diminish its importance but highlights a new pseudophosphatase function in cell signalling.
Protein expression and purification. Escherichia coli BL21(DE3) Rosetta cells transformed with the relevant expression construct were cultured in 2X TY medium at 30°C until OD 600 = 0.6. Routinely, 1-2 mg of recombinant PTP was obtained per 1 L culture. Protein expression was induced with 1 mM isopropylthio-β-D-galactopyranoside for 18 h at 20°C. For biotinylated Avi-tag constructs, 200 μM D-biotin (Sigma Aldrich) was added at the point of induction. After a freeze-thaw cycle, bacterial pellets were resuspended in ice-cold lysis buffer (50 mM Tris, pH 7.4 [PTPRK domains]/pH 8 [PTPRU domains], 500 mM NaCl, 5% glycerol, 0.5 mM TCEP) and lysed using high-pressure cell disruption (Constant Systems Ltd). Cell lysates were clarified by centrifugation at 40,000 × g for 30 min. Cleared lysates were incubated with Ni-NTA agarose beads (Qiagen) for 1 h at 4°C. Ni-NTA beads were packed in to a 10 ml gravity-flow column and equilibrated with 10 bed volumes of purification buffer (for PTPRU constructs; 50 mM Tris-HCl, pH 8, 500 mM NaCl, 5% glycerol, 5 mM DTT, for PTPRK constructs; 50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 5% glycerol, 5 mM DTT) containing 5 mM imidazole. Ni-NTA beads were then washed with 20 bed volumes of purification buffer containing 20 mM imidazole, followed by elution in purification buffer containing 250 mM imidazole. Eluted protein was further purified by sizeexclusion chromatography on a HiLoad Superdex 200 pg 16/600 column (GE Healthcare) equilibrated in purification buffer. For crystallization-quality PTPRU-D1 domain, protein was buffer exchanged by iterative concentration/dilution in a 10 K MWCO centrifugal filter unit (Merck Millipore) against low-salt buffer (50 mM Tris-HCl, 10 mM NaCl, pH 8, 5% glycerol, 5 mM DTT) until a final NaCl concentration <15 mM. Protein was further purified by anion exchange chromatography on a MonoQ 5/50 GL column (GE Healthcare) equilibrated in low-salt buffer and bound protein was eluted by a linear 20 ml gradient against high-salt buffer (50 mM Tris-HCl, pH 8, 1 M NaCl, 5% glycerol, 5 mM DTT). Protein purity was assessed by SDS-PAGE and staining with Coomassie (Instant Blue, Expedeon).
X-ray data collection and structure solution. X-ray diffraction data were recorded at Diamond Light Source (DLS) beamlines I03 and I04. Datasets were collected at λ = 0.9795 Å. Diffraction datasets were indexed and integrated using the automated data processing pipeline available at DLS, implementing XIA2 DIALS for the reduced dataset and XIA2 3dii for the oxidized dataset 53 then scaled and merged using AIMLESS 54 . Resolution cut-off was determined by CC 1/2 > 0.5 and I/σl > 1.5. The initial structure was solved by molecular replacement using Phaser 55 , with human PTPRK-D1 39 (PDB ID: 2C7S) as a search model. Further refinements were performed using COOT 56 and phenix.refine 57 . Graphical figures of the PTPRU-D1 structure were rendered in PyMOL (Schrödinger, LLC).
pNPP phosphatase activity assay. Recombinant PTP domains were made up to 50 μL volumes in a 96-well microplate format in assay buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 5% glycerol, 5 mM DTT, 100 μg/ml BSA). Serial dilutions of pNPP substrate (0-40 mM, New England Biolabs) were performed in assay buffer. Reaction plates containing both enzyme and substrate dilutions were incubated at 30°C for 30 min prior to addition and mixing of 50 μl pNPP substrate to initiate reactions. Product formation was monitored for 15 min at 30°C by measuring absorption at 405 nm in a Spectramax M5 plate reader (Molecular Devices), followed by fitting to a 4-nitrophenol (Sigma) standard curve of known concentration. Data were fitted using linear regression in GraphPad Prism to determine initial enzymatic rates (V 0 ). V 0 values at various substrate concentrations were fitted using non-linear regression and kinetic parameters (V max and K m ) calculated from the Michaelis-Menten equation in GraphPad Prism. k cat values were calculated using the equation k cat = V max /[E T ].
For immunoprecipitation (IP), cells were transferred to ice 48-h posttransfection and washed twice with ice-cold 1X phosphate buffered saline (PBS). Cells were lysed in 700 μl ice-cold lysis buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 10% [v/v] glycerol, 1% [v/v] Triton X-100, 1 mM EDTA, 10 mM NaF, 1 mM PMSF, 1X EDTA-free protease inhibitor) on ice for 30 min, with periodic agitation. Lysates were cleared by centrifugation at 14,000 x g, 15 min at 4°C and supernatants transferred to chilled tubes. Total protein concentration was quantified by bicinchoninic acid (BCA) assay. Lysates were then adjusted to a final concentration of 5 mM DTT, to prevent air oxidation. Equal amounts of total cell lysate for each sample (~2 mg) was combined with 10 μl (20 μl of 50% slurry) of washed FLAG-M2 magnetic beads (Sigma Aldrich) in a total volume of 1 ml made up in lysis buffer and incubated for 2 h at 4°C with rotation. Beads were then collected on a magnetic stand and supernatants removed. Beads were then resuspended and washed once with 1 ml lysis buffer and four times with lysis buffer containing 500 mM NaCl.
For pNPP assays, IPs were washed once with 1 ml assay buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 5% [v/v] glycerol, 5 mM DTT), then resuspended in 500 μl assay buffer. Reactions were initiated by adding 500 μl assay buffer supplemented with 20 mM pNPP (10 mM final concentration), and reaction tubes placed horizontally on an orbital shaking platform at 30°C, 220 RPM. At each timepoint, beads were thoroughly resuspended and 100 μl samples were immediately added to 50 μl 0.58 M NaOH solution to terminate the reaction. Timepoint supernatants were transferred to a microplate. After the final timepoint, product formation was determined by measuring absorption at 405 nm in a Spectramax M5 plate reader (Molecular Devices), followed by fitting to a 4-nitrophenol (Sigma) standard curve. After assay completion, remaining beads were collected by magnet and washed in 1 ml TBS. Beads were then resuspended in 40 μl of TBS with 10 μl 5X SDS PAGE sample buffer and incubated at 95°C for 10 mins. Beads were collected by magnet and supernatants used for SDS-PAGE and immunoblot analysis.
BIOMOL green phosphatase assay. Recombinant PTP domains were made up to 20 μl volumes in a 96-well microplate format in reaction buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 5% glycerol, 5 mM DTT). Thirty microliters of 100 μM pSer, pThr, pTyr peptides (DADE-pY-LIPQQG and END-pY-INASL) or imidodiphosphate was added to protein wells to initiate reactions. Reactions were allowed to proceed for 15 min before termination by addition of 100 μl BIOMOL Green reagent (Enzo). Liberated phosphate was measured by absorbance at 360 nm in a Spectramax M5 plate reader (Molecular Devices), followed by interpolation to a standard curve of known phosphate concentration. Serial dilutions of phosphate were performed in reaction buffer using 800 μM phosphate standard (Enzo).
Phosphoinositide phosphatase activity assay. The commercially available Enzchek phosphate assay kit (Thermo Fisher Scientific) was used to measure the release of phosphate from the phosphoinositides lipids PI(4,5)P2 and PI(4)P. The kit was used according to the manufacturer's instructions. Briefly, PTPRU-D1 (3 µM) or the positive control PRL-3 (6 µM) were incubated in Enzchek reaction buffer containing 50 mm Tris-HCl pH 7.5, 2 mM MgCl2 plus 500 mM NaCl (only for PTPRU-D1) and 5 mM DTT. The reaction was initiated upon addition of each lipid substrate at a final concentration of 100 µM. The assay was conducted in triplicates at 37°C in a BioTek Synergy H1 plate reader for 45 min and the release of phosphate was monitored measuring the absorbance at 360 nm over time. For every phosphoinositide substrate, a control without enzyme for blank subtraction was also measured. The data are represented as mean +/− SD. Generation of pervanadate-treated lysates. 3 × 10 6 MCF10A cells were seeded in 10 cm dishes and cultured to complete confluence (4 days). Media was then aspirated and cells treated with 8 ml of fresh growth medium containing 1 mM sodium pervanadate for 30 min at 37°C, 5% CO 2 . Cells were then placed on ice and washed twice with ice-cold PBS. Cells were lysed in 600 μl per 10 cm dish of icecold lysis buffer ( were then incubated at 92°C for 5 min and resolved by SDS-PAGE on a 10% resolving gel. Total protein was transferred to 0.2 μm reinforced nitrocellulose membranes by wet transfer and blocked in 5% (w/v) skimmed-milk in TBS with 0.2% TWEEN-20 (TBS-T; 20 mM Tris-HCl, pH 7.6, 137 mM NaCl, 0.2% [v/v] Tween-20) for 20-60 min. Membranes were then rinsed once with TBS-T before incubation with appropriate primary antibody in 3% (w/v) BSA/TBS-T overnight at 4°C. Membranes were washed 3 × 10 min in TBS-T and then incubated with the appropriate species-specific HRP-conjugated anti-IgG secondary antibody for 1 h at RT. Following 3 × 10 min washes in TBS-T, membranes were developed using EZ-ECL solution (Geneflow) and imaged using a BioRad ChemiDoc MP imaging system (Bio-Rad). 2D-densitometry for quantification was carried out in Fiji 59 .
Limited proteolysis. All steps were performed on ice unless otherwise stated. Trypsin and subtilisin proteases were reconstituted at 1 mg/ml in proteolysis buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 10% glycerol, 1% Triton X-100, 1 mM EDTA, 5 mM DTT). Four micrograms of recombinant protein was incubated with 0, 0.625, 1.25, 2.5 μg subtilisin or 0, 1.25, 2.5 or 5 μg trypsin in a total volume of 25 μl made up in proteolysis buffer. Reactions were incubated on ice for 30 min and terminated by the addition of 10 μl 5X sample loading buffer. Samples were immediately incubated at 92°C for 5 min and resolved by SDS-PAGE on a 16% SDS-polyacrylamide gel. Proteins were visualized by Coomassie staining and gels were imaged using a ChemiDoc MP imager.
Differential scanning fluorimetry. Differential scanning fluorimetry was performed using Protein Thermal Shift Dye kit (Thermo Fisher Scientific) as per manufacturer's protocol in a ViiA-7 Real-time PCR system (Thermo Fisher Scientific). Reaction mixes consisting of 2 μg recombinant protein and 1X Protein Thermal Shift Dye were made up to a total volume of 20 μl in purification buffer. Samples were then heated on a 1°C/s gradient from 25 to 95°C and protein unfolding at each temperature monitored by measurement of fluorescence at 580/ 623 nm (excitation/emission). Fluorescent signal vs. temperature was fitted to a non-linear Boltzman-sigmoidal regression in Graphpad Prism, with the T m calculated from the inflection point of the fitted curve.
Reversible oxidation of recombinant protein. All steps were performed on ice unless otherwise stated. Ten micrograms of recombinant protein was mixed with either 0, 0.25, 1 or 2 mM H 2 O 2 in a total reaction volume of 50 μl in ice-cold oxidation buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 5% glycerol). Reactions were then incubated on ice for 30 min. Each reaction was then split equally (25 μl, 5 μg) and incubated at 95°C for 5 min with 7.5 μl 5X sample loading buffer with (reducing) or without (non-reducing) β-mercaptoethanol. Reduced and nonreduced samples were then resolved by SDS-PAGE on separate NuPage 4-12% bistris gels using 1X MOPS running buffer and visualized by staining with Coomassie. Gels imaged using a ChemiDoc MP imager.
Assessment of recombinant protein biotinylation. Ten micrograms of in vivo biotinylated recombinant protein was boiled for 5 min at 95°C in an appropriate volume of 5X sample loading buffer. Samples were cooled to RT before addition of a 3-fold molar excess of streptavidin and incubated for 5 min at RT. Protein was then resolved by SDS-PAGE on a NuPAGE 4-12% Bis-Tris gel and visualized by staining with Coomassie stain. After destaining with ddH 2 O, gels were imaged on a ChemiDoc MP imager and total levels of biotinylated protein quantified by 2Ddensitometry in Fiji 59 .
Recombinant protein pull downs. 50 μg of biotinylated Avi-tag recombinant PTP domains were bound to 167 μl of streptavidin-coated magnetic beads (New England Biolabs) made up to a total volume of 500 μl in ice-cold purification buffer (50 mM Tris, pH 7.4, 150 mM NaCl, 5% [v/v] glycerol, 5 mM DTT) at 4°C for 1.5 h with rotation. Beads were collected using a magnetic stand and washed 3 times in ice-cold purification buffer, followed by two washes with ice-cold 150 mM NaCl wash buffer (20 mM Tris-HCl, pH 7.4, 150 mM NaCl, 10% [v/v] glycerol, 1% [v/v] Triton X-100, 1 mM EDTA). Bead conjugated PTP domains were then blocked in ice-cold 5% (w/v) BSA/150 mM NaCl wash buffer at 4°C for 1 h with rotation. Pervanadate-treated cell lysates were pre-cleared by incubation with streptavidin magnetic beads (0.67 mg beads per ml of lysate) at 4°C for 1 h with rotation. Blocked PTP domains were collected on a magnetic stand and washed twice with ice-cold 150 mM NaCl wash buffer. One milligram of pre-cleared pervanadate-treated lysate (250 μl) was incubated with PTP domain-bound beads in a total volume of 1 ml 150 mM NaCl wash buffer at 4°C for 1.5 h with rotation. At 4°C, beads were collected on a magnetic stand and supernatants removed. Bead bound protein was then washed twice by resuspension in ice-cold 150 mM NaCl wash buffer, followed by one wash in ice-cold 500 mM NaCl wash buffer (20 mM Tris-HCl, pH 7.4, 500 mM NaCl, 10% [v/v], 1% [v/v] Triton X-100, 1 mM EDTA) with no resuspension. Beads were then washed twice by resuspension in ice-cold 500 mM NaCl wash buffer followed by a final wash in ice-cold TBS (20 mM Tris-HCl, pH 7.4, 137 mM NaCl). Beads were resuspended in 15 μl TBS with 25 μl 5X sample loading buffer containing 2 mM biotin and incubated at 95°C for 10 min. Beads and supernatants were stored at −20°C prior to SDS-PAGE and immunoblot analysis.
pTyr immunoprecipitation. In total, 400 μl dephosphorylation reactions (prepared as described above) were diluted to a total volume of 800 μl (0.2% [w/v] SDS final concentration) in 150 mM NaCl wash buffer (20 mM Tris-HCl, pH 7.4, 150 mM NaCl, 10% [v/v] glycerol, 1% [v/v] Triton X-100, 1 mM EDTA). Five microliters of rabbit anti-pTyr antibody (Cell Signalling Technology) was added to each sample and incubated for 3 h at 4°C with rotation. Immunoprecipitation was carried out by addition of 40 μl of washed protein G agarose beads and samples were incubated overnight (16 h) at 4°C with rotation. Beads were collected by centrifugation at 15,000 × g for 30 s at 4°C and washed five times with 1 ml of ice-cold 150 mM wash buffer. After washing, beads were resuspended in 2.5X sample loading buffer and incubated at 95°C for 10 min. Supernatants were collected and stored at −20°C prior to SDS-PAGE and immunoblot analysis.
CRISPR/Cas9 genome editing. Oligonucleotides encoding single guide RNAs (sgRNAs) targeting human PTPRU exon 1 and exon 14 were cloned into pSpCas9. (BB).mCherry and pSpCas9.(BB).eGFP respectively as previously described 60 . MCF10A cells were co-transfected with both plasmids by reverse transfection using Lipofectamine LTX/PLUS reagent as per manufacturer's instructions. Clonal celllines were established by single-cell sorting of mCherry-eGFP double positive cells by flow cytometry, 48-h post-transfection. After expansion of clones, PTPRU-KO clones were identified using immunoblot for PTPRU. sgRNA target sites were amplified from genomic DNA to confirm editing. Three independent confirmed PTPRU-KO clones were pooled to establish the final PTPRU-KO MCF10A population. PTPRK-KO MCF10A cells were generated in a previous study 3 .
Protein sequence alignments. Protein multiple sequence alignments were generated using Clustal Omega 61 and edited using Jalview 62 .
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.