We review here recent results on the structure and function of a receptor protein tyrosine phosphatase, RPTPμ. In addition to their intercellular catalytic domains which bear the phosphatase activity, the RPTPs are cell-surface-receptor-type molecules and in many cases have large extracellular regions. What role can these extracellular regions play in function? For RPTPμ, the extracellular region is known to mediate homophilic adhesion. Sequence analysis indicates that it comprises six domains: an N-terminal MAM (meprin/A5/μ), one immunoglobulin-like domain and four fibronectin type III (FN) repeats. We have determined the crystal structure of the entire extracellular region for RPTPμ in the form of a functional adhesion dimer. The physical characteristics and dimensions of the adhesion dimer suggest a mechanism by which the location of this phosphatase can be influenced by cell–cell spacings.
The receptor protein tyrosine phosphatases
The RPTPs (receptor protein tyrosine phosphatases) are a family of cell-surface receptors numbering some 20 members in the human genome [1,2]. Each RPTP has a single membrane-spanning sequence connecting an N-terminal ectodomain to a C-terminal cytoplasmic region. The domain architecture and phosphatase activity of the cytoplasmic region provides unifying features, common to all family members (most of them have a tandem of phosphatase domains, but some RPTPs have just one). In contrast, the extracellular regions of these cell-surface glycoproteins are very varied. What is the role of the extracellular region in the function of these molecules and how might it impact on the activity of the tyrosine phosphatase catalytic domain within the cell? In the cellular context, a protein, either cytosolic or at the cell surface, is part of a very crowded population of molecules [3,4]. To fulfil their disparate functions, many such molecules will need to find their way to particular subcellular locations. Might this be the role of the extracellular region of the RPTPs, to determine the subcellular positioning of the cytoplasmic phosphatase activity? Our recent data indicate that this could indeed be a function of the extracellular region for at least one of the RPTPs, providing an example of the interplay between cell-surface adhesion and modulation of intracellular signalling . The conclusions we review here can be paraphrased as answers to the questions of how an RPTP can ‘measure’ its correct location and how, having identified this location, the RPTP can be tethered to stay there. The RPTP we have focused on is RPTPμ , one of the type IIb subfamily, and, for this particular system, the correct location is represented by regions of intercellular contact such as the adherens junctions [7,8].
The extracellular regions of the type IIb RPTPs (RPTPμ, RPTPρ, RPTPκ and RPTPλ) share a common, six-domain, architecture and are believed to mediate cellular adhesion by homodimerization [9–12]. RPTPμ is a cell-adhesion molecule involved in neural development (axon growth) and angiogenesis [7,13]. Expression is developmentally regulated within the central nervous system and the vascular network with clustering of RPTPμ molecules apparent on axonal growth cones and at endothelial cell junctions [7,14,15]. RPTPμ is believed to modulate axon growth and angiogenesis not only through homophilic adhesion, but also through interactions with members of the cadherin family of adhesion molecules [16–18]. One of the central questions for the function of RPTPμ is the nature of its interactions with the cadherins. Cadherins are themselves homophilic adhesion molecules and are central to the development and organization of many tissues within the body . Thus members of the RPTP type IIb subfamily, such as RPTPμ, may make key contributions to the balance between cellular mobility and adhesion through the modulation of cadherin function, but how?
Much is known about structure–function relationships, and the details of the catalytic mechanism, for the cytosolic regions of the RPTPs; their catalytic domains conform to the classic phosphatase architecture . An understanding of the structure–function relationship of the extracellular regions has remained rather more obscure. For the ectodomains of the type IIb RPTPs, sequence analysis predicted a MAM (meprin/A5/μ) domain at the N-terminus, followed by an Ig-like domain and four fibronectin type III (FN) domains before the single membrane-spanning helix. These six domains mediate the homophilic adhesive activity  and formed the focus of our structural studies on RPTPμ.
Structure of the MAM–Ig domains
Eukaryotic cell-based expression was essential for providing constructs derived from the RPTPμ extracellular region and suitable for structural studies. We have developed a robust and efficient protocol for the use of HEK-293T (human embryonic kidney) cells for transient expression of soluble secreted proteins . This methodology allows manipulation of the glycosylation state of a secreted protein; a functionality that is frequently essential for the fine-tuning of the homogeneity of glycoproteins to permit the growth of well-ordered protein crystals . The transient expression of His-tagged constructs allows rapid assessment of expression levels, protein stability and propensity to crystallize. An initial survey of RPTPμ constructs covered the entire extracellular region, the MAM–Ig domains and the MAM domain in isolation. Crystals from preliminary crystallization screening of the MAM–Ig construct showed immediate promise for structural studies and screen conditions were rapidly optimized to yield crystals which diffracted to 2.7 Å (1 Å = 0.1 nm) resolution . Since the topology of a MAM-fold was unknown, phasing was achieved by selenomethionine labelling in the HEK-293T expression system and collection of MAD (multiple anomalous dispersion) data at BM14, the UK MAD beamline at the European Synchrotron Radiation Facility. The crystals were grown without removal or trimming of the wild-type complex sugars, and indeed the crystallographic electron-density maps revealed sugars at three N-linked glycosylation sites. The MAM domain was revealed to consist of a β-barrel-type fold commonly referred to as a jelly-roll fold . This topology has been found in numerous human proteins, e.g. the cytokine TNF (tumour necrosis factor) . Comparison against the database of known jelly-roll fold structures scores highest similarity for the RPTPμ MAM domain and the ephrin-binding domains of the Eph receptor tyrosine kinases; an intriguing observation given that, like RPTPμ, the Eph receptors are involved in the development of the nervous system and the vasculature [25,26]. Within the MAM–Ig structure, there is a seamless interface between the β-barrels of the MAM and the Ig domains .
The minimal adhesive unit
The aim of determining the MAM–Ig structure was, of course, not only simply to reveal the architecture, but also to assess the adhesive functionality of the molecule. In this respect, it was disappointing to find that the MAM–Ig fragment was a monomer in solution as assessed using gel filtration, analytical ultracentrifugation and dynamic light scattering, even at high protein concentrations. This observation remained true in the context of the crystal packing, where the lattice showed no extensive intermolecular contact between MAM–Ig molecules that could represent an adhesive interaction. This prompted us to carry out a careful reappraisal of the biophysical properties of a series of domain-deletion constructs for the RPTPμ extracellular region . Gel filtration then indicated that the MAM, Ig and the first of the FN domains constitute the minimal fragment capable of mediating a homodimeric interaction. All constructs containing these, plus subsequent domains up to the full-length extracellular region, showed homodimerization. However, the stability of these high-affinity dimers was pH-dependent; dynamic light scattering measurements indicated that, for pH values below 6.2, dimers dissociated into monomers. The pH-controlled switch is reversible: dimers reformed on titration back to pH values above 6.2. This molecular-level behaviour matches the pH-dependency of RPTPμ-mediated cell adhesion observed in cell-aggregation assays [11,27]. We also took our series of domain-deletion constructs and tested their adhesive function in a cellular context by expression in naturally non-adherent Sf9 insect cells. For this analysis, we used ectodomain constructs containing a membrane-spanning helix followed by a GFP (green fluorescent protein) domain replacing the distal phosphatase cytoplasmic domain. Fluorescence and confocal microscopy revealed the cellular location and adhesive activity of the molecules. For transfection of a construct expressing only the GFP, the cells remained non-adherent as they did for the constructs encoding ectodomains consisting of merely the MAM domain or the MAM–Ig unit, or indeed the MAM–Ig–first FN domain. Only constructs containing the MAM–Ig and two, or more, of the FN domains were capable of inducing cell aggregation. Thus it was clear that, to relate the functional role of the extracellular region to structure, we must focus our crystallographic analyses on larger fragments, spanning from the RPTPμ N-terminus to include one or more FN domains, and, ideally, the entire extracellular region.
Structure of the RPTPμ ectodomain
We returned to our mammalian-cell-based transient expression system; however, given 12 predicted N-linked glycosylation sites in the full-length RPTPμ extracellular region, the heterogeneity introduced by the addition of wild-type complex sugars was too great to allow growth of well-ordered crystals. We therefore employed a glycosylation-deficient cell line, HEK-293S-GnTI− , which lacks N-acetylglucosaminyltransferase 1 activity, resulting in secreted proteins bearing only simple sugars of the Man5GlcNAc2 type which can be truncated back to a single GlcNAc moiety on treatment with endoglycosidase H . This deglycosylation treatment, plus maintenance of neutral pH, was essential for the growth of well-ordered crystals of the functional adhesion molecule. These crystals were capable of diffraction to 3 Å resolution and thus facilitated the structure determination of the full-length extracellular region of RPTPμ . The structure revealed an extended rigid conformation. Indeed, the molecule resembles a textbook schematic diagram showing six domains arranged contiguously as beads on a string (Figure 1A). The interfaces between each of these domains are extensive and apparently very rigid; in addition, there is an interdomain disulfide bridge between the second and third FN domains. The residues involved in these interfaces are highly conserved between members of the type IIb RPTP family (μ, ρ, κ and λ) and across species (human, mouse, dog, chicken, frog and fish). This degree of conservation strongly suggests that the rigid extended ectodomain architecture seen in the RPTPμ crystal structure has a functional role. The only exception to the rigidity of the architecture appears to occur at the fourth FN domain. For this region, the electron density of the crystal structure is very weak and indistinct, indicative of inherent flexibility. This may, in part, be the result of the proteolytic cleavage which occurs in this domain during the natural processing of the RPTPμ molecule . This cleavage does not result in shedding of the receptor from the cell surface, but may introduce some flexibility into the structure of the membrane-proximal FN domain.
Structure and function of the RPTPμ extracellular region
In addition to highlighting the interdomain interfaces, the sequence conservation between RPTP IIb family members and across species, when mapped onto the RPTPμ structure, also reveals large patches of highly conserved surface. In the crystal structure, much of this conserved surface can be seen to contribute to a trans (i.e. head-to-head) dimer interface. The extent and nature of this interface bore all the hallmarks of a physiologically relevant interaction. The interface involves some 3000 Å2 of buried surface area consisting of a mixture of interactions involving hydrophobic residues, such as tyrosine and tryptophan, as well as charge complementarity between polar, acidic and basic residues. Interestingly, these residues include among their number several histidine residues, which could, at least in part, be responsible for the observed pH-dependency of RPTPμ-mediated adhesion. As the trans dimer crystal structure appeared to be a very compelling candidate for the adhesive interaction, we sought to confirm the hypothesis by mutation of residues which played a central role in the interface. Six residues were selected for mutagenesis, of which two were hydrophobic (Trp279 and Tyr277), three were basic (Arg219, Arg220 and Arg389) and one was a histidine (His167). We again used a cell-adhesion assay based on transfection of normally non-adherent Sf9 cells with mutant constructs of RPTPμ incorporating cytoplasmic GFP tags. All of the ectodomain mutants tested lacked the adhesive activity of the wild-type ectodomain .
Does the high level of conservation at the adhesive interface indicate that interactions can occur between different members of the RPTP type IIb family? A detailed inspection of the adhesive interface argues against such cross-recognition; for example, structural mapping of the sequence conservation between human RPTPμ and RPTPκ revealed ten non-conserved residues at the dimer interface which could contribute to specific homophilic adhesion, i.e. RPTPμ adhering only to RPTPμ and not to RPTPκ. This specificity is consistent with the observations of Zondag et al.  for cells expressing either RPTPμ or RPTPκ constructs and labelled with different fluorescent dyes. The two cell populations did not intermingle, but rather red cells adhered to red cells, and green to green; in contrast, for the control experiment, cells labelled with either red or green dyes, but all expressing RPTPμ molecules, formed mottled clumps containing a mixture of red and green cells. A similar segregation between cells expressing different type IIb RPTPs can be observed in an anatomical context, e.g. in multilayered structures such as the cerebellum [14,31].
Functional implications of the adhesive dimer structure
A combination of structural and functional studies indicates that RPTPμ, and, by extrapolation, other members of the type IIB family, are very rigid extended molecules which show highly specific adhesive properties that generate trans homodimers. How might these properties relate to the role of type IIb RPTPs at the adherens junctions? As stated above, there are multiple lines of evidence to indicate a close functional interplay between RPTPμ and members of the cadherin family of adhesion molecules, and, indeed, the cadherins are important inhabitants of the adherens junctions. If we consider the structural data available for cadherins, the crystallographic analysis of the entire C-cadherin ectodomain by Boggon et al.  revealed a trans dimer capable of spanning a maximal cell–cell spacing of some 385 Å [the authors suggest a dimer orientation tilted relative to the cell surface that would result in a slightly reduced spacing consistent with that of the adherens junction observed in EM (electron microscopy) studies]. Our crystal structure of the RPTPμ ectodomain trans dimer spans 330 Å, a surprisingly close match to the reported dimensions of the adherens junctions.
It would appear that the C-cadherin and RPTPμ ectodomain trans dimers are well matched in size to each other and to the intercellular spacing appropriate for their adhesive interactions. Why might this be functionally important? If the cell–cell spacing can in some way influence the concentration of molecules at particular subcellular locations on the cell surface, this will provide a mechanism for modulation of intracellular signalling. One such mechanism, by which the extracellular cell–cell spacing can have an impact on the balance between phosphorylation and dephosphorylation in the underlying cytoplasm, has been proposed for T-cell signalling. Davis and van der Merwe  have developed the hypothesis that, at localized points of close contact between the antigen-presenting target cell and the T-cell, size-based exclusion of the RPTP CD45 may promote kinase-based signalling via the T-cell receptor. The proposed mechanism is as follows. On engagement of the T-cell receptor with an MHCI or MHCII antigen-presenting molecule, and the associated interactions of the co-receptors CD8 and CD4 and adhesion molecules such as CD2 and CD58, the distance between a T-cell and a target cell is limited to some 150 Å; the large extracellular region of the phosphatase, estimated at 280 Å by electron microscopy , cannot be accommodated within such close cell–cell spacings and CD45 is squeezed out from the interface. This depletion of phosphatase activity in the vicinity of the T-cell receptor would be conducive to Lck phosphorylation and consequent downstream signalling. Might the extracellular region of RPTPμ also be acting as a measure of cell–cell spacing? We postulated that, for RPTPμ, the properties of the ectodomain served to increase, rather than decrease, the molecular concentration at a particular intercellular spacing. At the appropriate spacing for an adherens junction, the formation of the adhesion trans dimer could lock the phosphatase activity into position, i.e. the extracellular region of RPTPμ acts as a spacer clamp .
The spacer-clamp hypothesis
We sought to test our spacer-clamp hypothesis by first assessing whether RPTPμ accumulates at adhesive interfaces in our Sf9 cell model system. Immuno-EM analysis of cryo-sections taken for cell aggregates indeed confirmed that RPTPμ became concentrated at these interfaces, consistent with the molecule becoming locked into position by formation of the trans dimer. Do these normally non-adherent cells display a distinctive cell–cell spacing defined by the RPTPμ trans dimer? And, if so, does this spacing become smaller as the length of the RPTPμ ectodomain (and hence the trans dimer) is shortened by stepwise deletion of the fourth (membrane-proximal) FN domain and then, in addition, deletion of the third FN domain (neither domain is required for cell adhesion)? We used three constructs (the full-length wild-type ectodomain, and the single- and the double-FN-domain-deletion mutants) to explore these questions. EM analysis of the resultant cell aggregate sections, using negative staining, revealed a clear correlation in cell–cell spacing with length of trans dimer. This behaviour is fully consistent with our spacer-clamp hypothesis.
The currently available structural and functional data for RPTPμ can be integrated into a model for adhesion-based regulation of subcellular phosphatase concentrations (Figure 1B). Before reaching the cell surface, RPTPμ is transported in secretory vesicles, micro-environments which are typically at pH 5.5; the acidic pH prevents formation of trans dimers. On reaching the cell surface, the RPTPμ is exposed to neutral pH values favourable to formation of the trans dimer; however, because of the rigid nature of the molecule, such trans dimers will only form at appropriate intercellular spacings. At an adherens junction, RPTPμs from opposing cells can form a trans dimer, fixing the RPTPμ at this location on the cell surface and, in so doing, locking the phosphatase activity into position. This local concentration of phosphatase activity is juxtaposed precisely with the high local concentration of the cadherins; the dimensions of the RPTPμ and cadherin cell-adhesion complexes both match the cell–cell spacing of the adherens junction. Thus phosphatase activity (RPTPμ) and substrate (the cadherin–catenin complex) are co-located in the cytoplasm, and the dephosphorylated state of the cadherin–catenin complex contributes to the stability of the adherens junction [19,35].
Questions remain open as to whether there are direct interactions between RPTPμ and cadherin ectodomains, and whether ordered arrays (involving both cis and trans adhesive interactions) form at the adherens junction, either separately for cadherin and/or RPTPμ or jointly. Is co-localization at the adherens junctions sufficient or is the precise arrangement of cadherin and RPTP orchestrated further? These questions remain to be explored in the future, as does the generality of the hypothesis that the extracellular regions of molecules such as the RPTPs are engaged in control of location as a mechanism to modulate signalling activity within the cell.
Structure and Function in Cell Adhesion: Biochemical Society Annual Symposium No. 75 held at The Palace Hotel, Manchester, U.K., 3–5 December 2007. Organized and Edited by David Garrod (Manchester, U.K.).
We thank Linda Vincent for her expert assistance in preparing this review. A.R.A. is a U.K. Medical Research Council Career Development Fellow, C.S. is a Wellcome Trust Career Development Fellow and E.Y.J. is a Cancer Research UK Principal Research Fellow. We also thank colleagues who contributed to the structural and functional characterization of RPTPμ: Kaushik Choudhuri, Veronica Chang, Weixian Lu, Simon Davis and Anton van der Merwe.