• We are sorry, but NCBI web applications do not support your browser and may not function properly. More information
Logo of embojLink to Publisher's site
EMBO J. Feb 22, 2006; 25(4): 701–712.
Published online Feb 2, 2006. doi:  10.1038/sj.emboj.7600974
PMCID: PMC1383555

Molecular analysis of receptor protein tyrosine phosphatase μ-mediated cell adhesion

Abstract

Type IIB receptor protein tyrosine phosphatases (RPTPs) are bi-functional cell surface molecules. Their ectodomains mediate stable, homophilic, cell-adhesive interactions, whereas the intracellular catalytic regions can modulate the phosphorylation state of cadherin/catenin complexes. We describe a systematic investigation of the cell-adhesive properties of the extracellular region of RPTPμ, a prototypical type IIB RPTP. The crystal structure of a construct comprising its N-terminal MAM (meprin/A5/μ) and Ig domains was determined at 2.7 Å resolution; this assigns the MAM fold to the jelly-roll family and reveals extensive interactions between the two domains, which form a rigid structural unit. Structure-based site-directed mutagenesis, serial domain deletions and cell-adhesion assays allowed us to identify the four N-terminal domains (MAM, Ig, fibronectin type III (FNIII)-1 and FNIII-2) as a minimal functional unit. Biophysical characterization revealed at least two independent types of homophilic interaction which, taken together, suggest that there is the potential for formation of a complex and possibly ordered array of receptor molecules at cell contact sites.

Keywords: cell adhesion, MAM domain, receptor protein tyrosine phosphatase, signal transduction, X-ray crystallography

Introduction

Cell adhesion is fundamental to multicellular organisms, providing structural support and avenues for communication. Cell-adhesion molecules are specialized cell surface receptors capable of clustering at intercellular contacts via interactions, mediated by their ectodomains, which can occur in trans (between opposing cells) as well as in cis (on the same cell surface). Reversible protein phosphorylation plays a prominent role in the signalling associated with these dynamic molecular assemblies (reviewed in Perez-Moreno et al, 2003). Most cell-adhesion molecules lack intrinsic kinase or phosphatase activity, but the type IIB receptor protein tyrosine phosphatases (RPTPs) can be classified both in terms of intercellular adhesive and intracellular signalling functions (reviewed in Beltran and Bixby, 2003; Johnson and Van Vactor, 2003). One of them, RPTPρ (PTPRT), among all annotated human protein phosphatases, was recently found to have the highest occurrence of somatic mutations, affecting both the extracellular and intracellular regions, in colorectal tumours and to have the characteristics of a tumour suppressor (Wang et al, 2004). The type IIB RPTPs thus comprise a biologically very interesting family of proteins but the molecular mechanisms underlying their cell-adhesive function and the implications of these ectodomain interactions for phosphatase signalling are yet unclear.

RPTPμ is a prototypical type IIB RPTP (Gebbink et al, 1991), others being RPTPκ, RPTPρ and PCP2 (RPTPψ/π/λ). RPTPμ and RPTPκ have been shown to mediate homophilic cell adhesion (Brady-Kalnay et al, 1993; Gebbink et al, 1993; Sap et al, 1994; Cheng et al, 1997), a function probably also performed by RPTPρ and PCP2 as the four homologues have highly similar ectodomain sequences (amino-acid identities between RPTPμ and RPTPκ, RPTPρ and PCP2 are 54, 63 and 49%, respectively). These molecules display different spatial expression patterns throughout development (Fuchs et al, 1998; McAndrew et al, 1998), suggesting that they function in a strictly homophilic manner. This concurs with the absence of mixed cell aggregates in co-cultured cells expressing either RPTPμ or RPTPκ (Zondag et al, 1995). RPTPμ has been found in the nervous system where it appears to perform functions related to axon guidance within the visual system and retinal lamination (reviewed in Johnson and Van Vactor, 2003; Ensslen-Craig and Brady-Kalnay, 2004) and in tissues where tight, continuous cell–cell contacts are physiologically important, such as the vascular endothelia and the intercalated discs of cardiomyocytes (Bianchi et al, 1999; Koop et al, 2003), where it has been shown to be clustered at the intercellular boundaries (Sui et al, 2005). In this context, it has been shown that RPTPμ can interact with members of the cadherin family of cell-adhesion molecules (N, E, R and VE cadherins; Brady-Kalnay et al, 1998; Sui et al, 2005). Functionally, it is thought that such an interaction would bring the phosphatase activity of RPTPμ near the cadherin–catenin complexes, maintaining them in a dephosphorylated state and so stabilizing the intercellular contacts.

The type IIB RPTPs share the same domain structure: an ectodomain, an ~100-residue intracellular juxtamembrane region unique among RPTPs and two phosphatase domains. The ectodomain contains one MAM (meprin/A5/μ) domain (Beckmann and Bork, 1993), one Ig domain, and four fibronectin type III (FNIII) repeats. The MAM fold is to date unknown, and is also present extracellularly in a functionally diverse range of proteins such as meprins, neuropilins and zonadhesins. This domain has been shown to participate in homo-oligomerization of neuropilin-1 and -2 (Chen et al, 1998; Nakamura et al, 1998) and meprin A (Ishmael et al, 2001). For RPTPμ, the MAM and the Ig domains have been separately implicated as having key roles in homophilic adhesive (trans) interactions (Brady-Kalnay and Tonks, 1994; Zondag et al, 1995). Recently, Cismasiu et al (2004) have proposed that the homophilic binding properties of the MAM domain make a particular contribution to the lateral (cis) receptor oligomerization.

We describe here a systematic investigation of the cell-adhesive properties of the human RPTPμ extracellular region (Exμ), aiming to map further the homophilic interaction sites and to understand the relative contributions of MAM and the other constituent domains. We determined the crystal structure of a construct containing the two N-terminal domains (MAM and Ig, hereafter called MIg) at 2.7 Å resolution. This allowed the identification of a distinctive surface on the MAM domain that, as subsequent mutagenesis studies suggest, contributes directly to cell adhesion. However, a serial domain deletion analysis revealed that further interactions are required for this function, involving domains beyond MIg. Our results are consistent with a multidomain interaction model for RPTPμ and, more generally, type IIB RPTPs-mediated cell adhesion. Any adhesion-induced oligomerization of the RPTPμ molecules will affect the relative spatial arrangement of the intracellular catalytic regions and may induce specific conformational changes, thus representing a potential regulatory mechanism for downstream signalling.

Results

Crystal structure of the MIg segment of RPTPμ

Crystals of selenomethionyl (SeMet) substituted and hexahistidine-tagged MIg were grown from protein expressed in human embryonic kidney (HEK293T) cells. There is one monomer per crystallographic asymmetric unit. The protein migrates as ~40 kDa on SDS gels, ~10 kDa larger than its theoretical molecular size, a difference accounted for entirely by N-glycosylation as confirmed by PNGase F treatment (data not shown). The current model (Figure 1) is complete for residues 21–279 (residues 1–20 constitute the secretion signal sequence) with only the C-terminal His6 tag unobserved, and contains three glycosylation sites at Asn72, Asn92 and Asn131. It has been refined to Rwork/Rfree of 22.4/27.4% at 2.7 Å (Table I, representative electron density is shown in Supplementary Figure S1).

Figure 1
The MIg crystal structure and analysis of the MAM domain. (A) Ribbon diagram of MIg. The MAM-Ig linker in the MAM domain is highlighted in purple. Disulphide bonds (orange) and the N-glycosylation sites (CPK) are presented as stick models. The N- and ...
Table 1
Data collection, structure determination and refinement statistics

The overall structure of MIg is roughly L-shaped, where the principal axes of the MAM (residues 21–184) and the Ig (residues 185–279) domains form an angle of ~104° (Figure 1A). The interdomain orientation is stabilized by a short linker (residues 180–184, XPCXX) whose proline and cysteine are conserved in type IIB RPTPs (Figure 2A). Pro181 is at the turning point of the linker as it exits the main body of the MAM domain. Cys182 serves to secure the linker via a disulphide bond (Cys96–Cys182, both conserved among the MAM domains). The Ig domain is oriented edge-on against the MAM domain, with the outer strand G of the MAM domain β-sheet GDIB hydrogen bonding to strand g of the Ig domain β-sheet cfg to form a continuous sheet spanning both domains (Figure 1A). The interdomain interface contains 21 van der Waals contacts and 20 hydrogen bonds, 15 of which are main chain–main chain interactions.

Figure 2
Sequence alignments and secondary structure elements of the MAM and Ig domains. (A) Sequence alignment of the type IIB RPTPs MAM and Ig domains. Residue numbering begins with the signal sequence (not shown). Residues with more than three occurrences of ...

The MAM domain fold was heretofore unknown. A DALI (Holm and Sander, 1995) search for structural homologues reveals that it belongs to the galactose-binding domain-like fold in the SCOP database (Andreeva et al, 2004). The 10-stranded β-sandwich is of jelly-roll topology (Supplementary Figure S2) and is structurally homologous to the ephrin-binding domain of murine EphB2 receptor tyrosine kinase (PDB ID 1KGY, chain A, 181 residues) (Himanen et al, 2001) and a host of carbohydrate-binding domains of bacterial proteins, for example that of laminarinase from Thermotoga maritima (PDB ID 1GUI, chain A, 155 residues) (Boraston et al, 2002) (Figure 1B). Sequence similarities among these proteins are minimal and are confined to the structural core. The Ig domain of MIg, similar to those in many CAMs, belongs to the I-set structural class of the immunoglobulin superfamily (reviewed in Chothia and Jones, 1997). The edge of the β-sandwich between strands c and d is the least conserved region in I-set domains and compared to other structures MIg lacks strand c′ but has extended β-strands d and e (residues 229–244).

Structural comparisons of the MAM domain with the ephrin-binding domain of the EphB2 receptor and with the carbohydrate-binding domain of laminarinase provide insights into the location of putative functional sites for the MAM domains in general. The RPTPμ MAM domain, when structurally aligned with 1KGY:A, reveals 108 superposable residues with inter-Cα atom distances <3 Å, root mean square deviation (r.m.s.d.) ~1.8 Å. With 1GUI:A, it shares 103 superposable residues, r.m.s.d. ~1.9 Å. Regions (loops >6 residues) that are not superposable with the MAM structure are all part of, or demarcate, the ligand-binding sites in EphB2 and laminarinase (Figure 1B). By analogy, in the MAM domain these regions are putative protein–protein interaction sites. They can be grouped into two subsites, located on opposing faces of the β-sandwich: L1 (residues 21–31) and L2 (53–67) on one side; L3 (108–113), L4 (122–127) and L5 (133–138) on the other. Notably, L1 (stabilized by a disulphide bond composed of two cysteines conserved in MAM domains, Cys27 and Cys36) and L2 are unique to the MAM-type jelly-roll fold, with L2 displaying a particularly high degree of sequence variability (Figure 2A and B). This suggests that they might play functional roles in all MAM-containing proteins, with L2 in particular conferring binding specificity. The structure of the L1 and the L2 loops is well ordered (Supplementary Figure S1); both are anchored to the β-sandwich main body via 64% (L1) and 42% (L2) of the side-chains (Figure 1C). The surfaces formed by these two loops are relatively flat, and are approximately coplanar with each other (Figure 1B). In addition, we have identified two metal ions (most likely sodium as judged from their coordination geometry) involved in multiple interactions with residues from L1 and L2 and therefore playing a stabilizing role (Supplementary Figure S3).

The formation of a crystal lattice requires protein–protein contacts and the major contact sites in the crystal can sometimes correspond to physiologically relevant interactions. The crystal packing of MIg, however, lacks extensive intermolecular interactions and no one site involves a significant number of interatomic contacts suggestive of homophilic adhesion interactions. The largest contact site (~1000 Å2 total buried surface area) involves MAM and Ig residues in ‘elbow-to-elbow' interactions of the L-shaped MIg molecule (Figure 1A). Although this site involves the solvent-exposed side-chain of Arg211 and the main-chain of the buried Tyr273, which correspond to the cancer-linked K218T and Y280H mutations in RPTPρ (Wang et al, 2004), it appears unlikely to be, by itself, sufficient for homophilic binding. Residues corresponding to the other three RPTPρ cancer-linked mutations in the MIg structure have buried side-chains (Figures 1C and and2A);2A); their effects are likely to be because of structural perturbation rather than direct involvement in homophilic interactions.

Mapping of an adhesion-competent ectodomain segment

To characterize further the structural requirements for RPTPμ ectodomain function, we used a cell-adhesion assay based on the induced aggregation of insect Sf9 cells expressing RPTPμ domain-deletion constructs on their surface (acronyms and schematics of domain composition are shown in Figures 2C and and3).3). These constructs have their extracellular regions serially truncated from the C-terminal end, and their second cytoplasmic phosphatase domain replaced by EGFP. We combined results from the structure prediction algorithm PSIPRED/THREADER (Jones et al, 1999), sequence alignment of the type IIB RPTPs and structural knowledge of FNIII domains in general, to delineate the domain boundaries (Figure 2C and Materials and methods). A previously un-noted tail region, which is ~50 residues long and follows the FNIII-4 domain, is apparent.

Figure 3
Cell-adhesion assays. Insect Sf9 cells expressing transmembrane RPTPμ EGFP-fusion constructs were observed by fluorescence microscopy. All RPTPμ fusion constructs (except for Exμ-TM-EGFP) have JM-D1-EGFP in the intracellular region, ...

Results of the cell-adhesion assays are presented in Figure 3 (expression levels were assessed by Western blotting, see Supplementary Figure S4). Only insect cells expressing constructs with extracellular regions corresponding to Exμ, MIF3 and MIF2 formed aggregates, demonstrating that MIF2 (MAM-Ig-FNIII-1-2) is the minimal ectodomain fragment capable of inducing cell aggregation. An MIF1-x-Fc (Fc=two tandem Ig domains, dimeric; x=20-residue linker containing a 3C protease cleavage site and the IgGγ1 hinge region) fusion construct was inactive in cell adhesion, arguing against the notion that the FNIII domains 2, 3 and 4 merely play a supportive role in cell adhesion by acting as a spacer between the adhesive part of the ectodomain and the plasma membrane. Brady-Kalnay et al (1993) and Gebbink et al (1993) have shown that constructs lacking the intracellular catalytic domains (but still containing 54 juxtamembrane residues) are functional in cell-adhesion assays; therefore, this activity is contributed predominantly by the ectodomain. We can now further refine this observation because an Exμ transmembrane construct lacking all but the first 10 intracellular basic residues, which are responsible for membrane anchoring of the protein, is as active in inducing cell aggregation as its counterpart which contains the intracellular juxtamembrane region and D1 phosphatase domain (Figure 3).

Distinctive loops on the MAM domain are important for cell adhesion

Of the five loops in the MAM domain highlighted by structural and sequence comparisons two, L1 and L2, were the most distinctive features in this domain structure. To explore the role of this region, the Exμ construct was mutated in the L1 and the L2 loops (Figures 1C and and2A).2A). In L1, two acidic residues (Asp30, Glu31), conserved in type IIB RPTPs, were mutated to lysines. In the more variable L2 sequence, there is a common occurrence of multiple prolines that may provide important conformational rigidity. Two such residues in RPTPμ, Pro57 and Pro61, were mutated to alanines. All four residues have solvent exposed side-chains in the MIg crystal structure.

In parallel assays where aliquots of insect cells were infected with comparable numbers of baculoviral particles, cells expressing the L1 and the L2 mutants form smaller aggregates than wild-type Exμ. A combined L1/L2 quadruple mutant further reduces this activity, resulting in very sparse and small cell aggregates (Figure 3). Analysis of the expression levels indicated that the cells produced similar amounts of the wild-type and mutant proteins (Supplementary Figure S4). The ectodomain of the L1/L2 mutant is expressed at normal (wild type) levels by 293T cells and also is a dimer in solution at pH 8.0 (see below), confirming its overall structural integrity. These results suggest that both the L1 and L2 loops contribute to a MAM domain-mediated interaction that is necessary for cell adhesion. However, because residual binding activity was observed, even for the L1/L2 mutant, this region of the ectodomain does not appear to be the sole contributor to the adhesive interaction.

RPTPμ ectodomain forms a high-affinity, pH-dependent homodimer

We expressed our series of domain-deletion ectodomain constructs as secreted, soluble proteins in HEK293T cells, and subjected them to gel filtration analysis (chromatograms shown in Supplementary Figure S5) to evaluate their oligomeric state and structural integrity. The results are summarized in Figure 4A. At pH 8.0, both MAM and MIg are monomeric whereas MIF1 (MAM-Ig-FNIII-1) and MIF2 are exclusively dimers, and MIF3 and Exμ have apparent sizes of a tetramer. The apparent size of Exμ has been reported to be halved at pH <6 as compared to pH 8 (Cismasiu et al, 2004), therefore all C-terminal domain-deletion constructs were also chromatographed at pH 6.0. All constructs that had previously been dimeric or apparently tetrameric converted to monomers or apparent dimers, respectively, at this pH. Elongated molecules have larger sizes as estimated by gel filtration. However, analytical ultracentrifugation (data not shown) confirmed that Exμ is dimeric at pH 8.0 and monomeric at pH 6.0. These results indicate that MIF1 constitutes the minimal dimerization unit. The pH-dependent switch accords with previously reported observations based on aggregation assays (Gebbink et al, 1993). As such, it is likely that the pH-sensitive dimers are in a trans (‘head-to-head') orientation. The equilibrium pH for Exμ dimer–monomer conversion was found to be 6.2–6.3 as analysed by dynamic light scattering (Figure 4B). This pH range is close to the pKa of histidine, suggesting a role for one or more such residues in homodimer stabilization.

Figure 4
Oligomeric states of RPTPμ ectodomain constructs. (A) Schematic representation of the deletion constructs and Coomassie-stained gel showing the purified soluble constructs, and estimated Mr. The indicated number of N-glycosylation sites is based ...

Most N-terminal domain-deletion constructs, IF14t, F14t and F24t, though soluble, form nondiscrete oligomers (Figure 4A), suggesting that these constructs bear hydrophobic surfaces that cause nonspecific molecular aggregation. An exception is F34t which is monomeric, implying that this fragment does not contribute to the pH-dependent trans homodimerization. The L1/L2 mutant construct was also found to be dimeric at pH 8 and monomeric at pH 6, implying that the mutations did not disrupt the overall structure of the MAM domain sufficiently to cause nonspecific cell aggregation (Supplementary Figure S5). Thus, these results imply that Exμ contains at least two independent binding sites that contribute to the intercellular homophilic interactions. One of them (at least in part) involves the L1/L2 region, whereas the second is pH-dependent and appears to have a very high affinity (dimers on gel filtration were observed at concentrations as low as 10 nM).

Solution binding studies of the domain deletion constructs

To further investigate the contribution of individual domains to homophilic interactions, the equilibrium binding affinities of each domain-deletion mutant were measured by surface plasmon resonance (SPR). The ligands were immobilized on streptavidin-coated sensor chips via a C-terminal biotin tag to mimic their functional orientation. Ectodomain fragments (IF14t, F14t and F24t) that displayed polydisperse behaviour in solution exhibited tight, nonspecific binding to streptavidin-coated sensor chips, reflecting the nonspecific nature of the aggregates formed. These constructs were therefore not included for further analysis. All SPR measurements were performed at pH 7.4 where MAM and MIg are monomeric, whereas all the larger constructs are dimers. Therefore, the results presented here refer to pH-independent interactions. The results are summarized in Table II (binding curves and Scatchard plots are provided in Supplementary Figure S6).

Table 2
Equilibrium binding studies of RPTPμ constructs using surface plasmon resonance

Our primary set of binding experiments involved Exμ as the immobilized ligand. All soluble fragments (analytes) dissociated readily from the ligand-coated sensor chip at the end of the injection. We did not detect any binding of MIg to Exμ up to a concentration of 50 μM. The reverse experiment with Exμ binding to MIg gave similar results (data not shown). All homodimeric ectodomain constructs (MIF1, MIF2, MIF3 and Exμ) showed measurable (in the low μM range) binding to Exμ. To evaluate the effect of the L1/L2 mutations, a second set of binding experiments were carried out with wild-type and mutant ectodomains taken both as analytes and ligands (Supplementary Figure S7). The mutations affected binding to wild-type Exμ in both orientations (i.e. as analyte or ligand); while measurable, the weakness of the interactions precluded the determination of Kd values. However, even between L1/L2 mutant molecules some residual binding was apparent, consistent with the reduction, but not total loss of function, observed in the cell-adhesion assays.

In taking the analysis of the solution binding data further, there are clearly limitations regarding the interpretation of SPR experiments in the case of homophilic molecules, in particular in situations where the protein is already a dimer (possible interpretations are discussed in Supplementary data). Thus, we cannot draw definitive conclusions regarding the precise number and location of sites involved in the homophilic interactions of Exμ based on SPR alone. However, we can be confident about the order of magnitude and domains required for the pH-independent interactions. All the results converge towards the idea that the RPTPμ homophilic interactions are complex, with at least two components—a-high affinity pH-dependent interaction and the lower affinity interaction(s) (with low μM Kd values) measurable by SPR—and involve minimally three domains (MIF1) in in vitro assays while, for cell adhesion, a further domain (FNIII-2) is required.

Discussion

RPTPμ is a bi-functional molecule directly linking cell adhesion and signal transduction. However, in the absence of structural information, it is difficult to understand how this molecule, or indeed other members of the type IIB RPTPs, works. The MAM and Ig domains have long been recognized as essential for the adhesive function of RPTPμ (Brady-Kalnay and Tonks, 1994; Zondag et al, 1995; Cismasiu et al, 2004). However, these studies (based at least in part on treating the two domains as separate entities) came to rather divergent conclusions, placing differing emphasis on the MAM or Ig domain. Our crystallographic results, as well as providing the architecture of a MAM fold, reveal that MIg is a tight structural unit, with MAM and Ig domains intimately juxtaposed such that they should be viewed as interdependent for stability and function. Despite our expectations, based on the previous work in the field cited above, this two domain unit is not—in isolation—capable of mediating homophilic interactions. For our series of domain deletion constructs, biophysical characterization reveals that the addition of at least one FNIII domain to MIg (i.e. the MIF1 construct) is required for formation of stable homodimers, whereas a construct containing MIg plus two FNIII domains is the minimal functional unit in live-cell-adhesion assays. Why might the second FNIII domain be important for biological function? Three factors could contribute, either individually or in combination: (i) the threshold affinity for homodimer formation may be insufficient for cell-adhesive function (i.e. the second FNIII contributes extra affinity), (ii) adhesive function is highly sensitive to the orientation of the molecule on the cell surface resulting from addition of the second FNIII domain and (iii) additional interactions (possibly in cis) are required and these are at least in part mediated by the second FNIII domain.

We can minimally describe two types of homophilic interactions involving the RPTPμ extracellular region. The first one is very tight (Kd<10 nM), could be observed by gel filtration, dynamic light scattering and analytical ultracentrifugation and is pH-dependent, that is, can be readily and reversibly dissociated by a switch of pH from 7 to 6. This agrees with the results obtained by Cismasiu et al (2004) and effects observed in cell-aggregation assays (Gebbink et al, 1993), leading us to believe that it is a trans interaction. We could not measure its affinity by SPR because of experimental design constraints and we could not see it in the crystal structure as, as we found, it minimally requires the MAM, Ig and FNIII-1 domains. The second interaction involves the L1 and L2 loops of the MAM domain. Mutations in these loops reduce cell adhesion, although the overall domain integrity does not appear to have been disturbed. Thus, the L1/L2 region, plus the second FNIII domain (see above), both appear to contribute effects that add to a simple trans interaction mediated by the pH-dependent homodimer. Because homodimer-dependent (high affinity) and additional (low affinity) interactions have effects on cell adhesion, one could imagine two possible arrangements: both in trans or one in trans and one in cis (Figure 5A). MAM domains of neuropilin-1, -2 and meprin A have been associated with cis interactions (Chen et al, 1998; Nakamura et al, 1998; Ishmael et al, 2001) and previous results (Cismasiu et al, 2004) suggest that L2 is important for cis interactions in RPTPμ. The second (trans/cis) model therefore appears more probable.

Figure 5
Hypothetical schemes of Exμ adhesive interactions. (A) Possible arrangements of Exμ homophilic interactions that are consistent with the presence of two adhesion sites in the MIF1 region (one pH-dependent and one involving the L1/L2 loops). ...

What are the implications of a trans/cis model for the adhesive mechanisms? A primary, strong, trans interaction could play an anchoring role and initiate clustering of RPTPμ at the cell contacts, as observed by Gebbink et al (1995) and Bianchi et al (1999); the second, lower affinity interaction, in cis, could order the relative orientation of the molecules within the cluster and possibly initiate a zipper-like receptor array, in a manner previously suggested for other cell-adhesion molecules such as NCAM (Soroka et al, 2003), axonin (Kunz et al, 2002) and cadherins (Gumbiner, 2005). Adhesive events involving highly mobile cells such as leucocytes are based on multiple weak interactions. However, less dynamic cell–cell interactions may require higher affinity binding, for example as reported for NCAM (Kiselyov et al, 2005) and our data point to RPTPμ showing similar characteristics. Indeed for RPTPμ, it is possible that linear zippers could achieve a further level of organization, for example by parallel alignment, leading to a 2D (‘Velcro'-like) array, which would be entirely consistent with the large and apparently planar contact surfaces that we have observed between adherent cells (Figure 5B and C). Such an arrangement could strengthen the adhesion (through avidity effects) but would require at least one additional type of interaction between the RPTPμ molecules. Potentially, there are yet further levels of complexity to the system as RPTPμ is also thought to interact directly with cadherins (including N, E, R and VE) via the intracellular region (Brady-Kalnay et al, 1998; Hellberg et al, 2002; Sui et al, 2005). Considering the oligomerization models postulated here for RPTPμ and previously for various cadherins (Chothia and Jones, 1997; Gumbiner, 2005), more studies will be required to understand their interplay in various circumstances and the structural arrangements of the molecules in such complexes.

Taken together, all observations to date point towards a complex, multidomain, interaction model for the RPTPμ-mediated cell adhesion. The direct consequence of such a model is an ordered clustering of receptors at cell–cell contact regions, a significantly increased half-life of RPTPμ in confluent cultures (described by Gebbink et al, 1995) and, more importantly, a direct effect on signalling properties. Such an effect could be via steric modulation of the catalytic (intracellular) region, the control of subcellular localization, bringing RPTPμ into proximity with its substrates (members of the cadherin–catenin complex such as p120ctn (Zondag et al, 2000) or VE-cadherin (Sui et al, 2005)) or clustering of downstream scaffolding proteins—such as RACK1 (Mourton et al, 2001)—at intercellular contact areas.

The distribution of RPTPρ mutations from tumour isolates (targets of selection during tumorigenesis; Wang et al, 2004) provides possible insights into how in the RPTP family mis-function of cell adhesion may have direct consequences for downstream signalling. Mutations are found in both the RPTPρ extracellular and intracellular regions. The intracellular mutations are predicted to result in the loss of catalytic domains or a reduction of the catalytic activity (Wang et al, 2004). Of the extracellular mutations, five can be mapped to equivalent residues in the RPTPμ MIg structure. We predict that most of these mutations cause significant perturbations of the RPTPρ structure as they dramatically alter the character of buried side-chains (F74S, A209T, F248S and Y280H; Figures 1C and and2A).2A). As such they may exert their effects either by preventing surface expression of a misfolded RPTPρ or by disrupting a functional portion of the exposed ectodomain. However, one mutation, K218T, is at the Ig domain surface; this mutation is therefore compatible with a correctly folded structure implying that it may impact directly on a protein–protein interaction. The principle of extracellular interactions regulating intracellular signalling in RPTPs was first demonstrated by the enforced dimerization of the EGFR-CD45 hybrid receptor (Desai et al, 1993) and further confirmed by experiments with RPTPα (Jiang et al, 1999). Intra- and intermolecular interactions between the catalytic domains of many RPTPs have been identified and been shown to modulate their enzymatic activity (Blanchetot and den Hertog, 2000; Feiken et al, 2000; Aricescu et al, 2001). The crystal structure of the first catalytic domain of RPTPα revealed a self-inhibitory dimer (Bilwes et al, 1996); conversely the structural determination of the equivalent domain in RPTPμ (Hoffmann et al, 1997) revealed a different dimerization mode (one which the authors point out may be due to the crystal lattice contacts). Thus, many questions remain concerning the effect of receptor oligomerization on catalytic function in the RPTP family. Trans/cis models, which are suggestive of higher order oligomerization of the RPTP ectodomains, point to extracellular-based mechanisms that can modulate the intracellular interactions, thus potentially influencing their signalling ability.

The true importance of the type IIB RPTP family of receptor enzymes is just beginning to emerge, with several recent reports linking them to tumour suppressor functions (Wang et al, 2004; McArdle et al, 2005). It is therefore important to understand the organization and regulatory mechanisms of these signalling assemblies. Further structural and functional studies must aim to characterize the molecular architecture of RPTPμ adhesive complexes and their consequences for downstream signalling.

Materials and methods

Ectodomain constructs

The cDNA sequences encoding deletion constructs of the RPTPμ ectodomain were amplified by PCR using the phFL vector (Gebbink et al, 1993) as a template (a gift of M Gebbink, Utrecht University). They are MAM: residues 1–185; MIg: 1–279; MIF1: 1–375; MIF2: 1–481; MIF3: 1–585; Exμ: 1–742; IF14t: 189–742; F14t: 281–742; F24t: 379–742 and F34t: 482–742. The last four constructs, lacking the native secretion signal, were fused to the secretion signal of chicken RPTPσ (residues 1–28, GenBank accession number L32870). Site-directed mutagenesis was performed according to Stratagene's QuickChange protocol. All mutants have been verified by DNA sequencing.

Protein expression in HEK293T cells

A vector, pHL, a derivative of the pCAβ-EGFP plasmid containing the chicken-β-actin promoter (gift of J Gilthorpe, Kings College London), was constructed for high-level expression of secreted, C-terminally His6-tagged ectodomain constructs. Transient transfection was carried out using polyethyleneimine (Durocher et al, 2002). The cells were cultured in Dulbecco's modified Eagle's medium (DMEM) (Sigma) supplemented with 10% fetal calf serum (FCS) (Sigma), L-glutamine and nonessential amino acids (Gibco). The concentration of FCS was lowered to 2% immediately after transfection in 90% confluent cells.

To prepare SeMet-substituted MIg protein, the cells were washed once with phosphate-buffered saline just before transfection, which was carried out in Met-free DMEM (ICN) supplemented with 30 mg/l L-SeMet (Acros Organics), 2% dialyzed FCS, L-glutamine and nonessential amino acids (Gibco).

Protein purification

His6-tagged proteins were purified 3–4 days post-transfection from the media using Ni2+-affinity and gel filtration chromatography. Fc-tagged proteins were purified by protein A affinity chromatography. The proteins were cleaved on the resin with 3C protease, eluted and further purified by gel filtration. All buffers were at pH 8.0.

Cell-adhesion assays

The method was as previously described (Brady-Kalnay et al, 1993; Gebbink et al, 1993). Insect Sf9 cells were grown in suspension in Sf900II medium (Invitrogen). A shuttle vector, pFBR, was modified from pFastBac-1 (Invitrogen) for expression of transmembrane (TM), EGFP-tagged proteins, resulting in pFBR-x-TM-D1-EGFP (x=insertion site of ectodomain constructs, D1=first phosphatase domain). A pFBR-x-(3C)-Fc-TM-D1-EGFP vector was created by subcloning a cDNA fragment encoding the 3C protease cleavage site followed by the human IgGγ1 Fc region from vector pRMHA-mIL11α-(3C)-Fc (gift of JK Heath, Birmingham University). The Bac-to-Bac system (Invitrogen) was used with standard protocols to produce recombinant baculoviruses. These were titred for each construct and equal amounts were inoculated to 20 ml suspension cultures (1.4 × 106 cells/ml, 100 r.p.m.). Cells were visualized live 40–48 h post-infection using an inverted microscope equipped with epifluorescence (Nikon DIAPHOT 300) and a confocal laser scanning microscope (Bio-Rad MRC 1024).

Crystallization and crystal structure determination

Crystals of SeMet-substituted MIg-His6 were grown by sitting-drop vapour diffusion at 22°C. The drops contained 1 μl each of protein (10 mg/ml, in 100 mM Tris–HCl, pH 8.0, 250 mM NaCl) and mother liquor (100 mM tri-sodium citrate, pH 6.4, 100 mM ammonium chloride, 30% PEG 4000). The crystal was cryo-cooled (by nitrogen gas stream) in the original mother liquor. MAD data, collected on beamline BM-14 at the ESRF (Grenoble, France), were integrated and scaled using the HKL package (Otwinowski and Minor, 1997), analysed and merged using XPREP (Bruker AXS, Madison). Heavy atom location, refinement, phasing and density modification were performed using SOLVE/RESOLVE (Terwilliger, 2003). The experimental map showed clear regions of secondary structure but poor connectivities in the loop regions and ambiguous side-chain densities. Results of automatic tracing and sequence assignment done by either the programs RESOLVE, or ARP/wARP (Morris et al, 2003), alone, were not satisfactory. However, when main-chain fragments produced by cycles of ARP/wARP-REFMAC routine were fed into RESOLVE for sequence assignment in an iterative manner, ~90% of the protein sequence could be assembled in a semiautomatic fashion. The rest of the model was manually built using the program O (Jones et al, 1991) coupled with cycles of refinement by CNS (Brunger et al, 1998) and REFMAC (Murshudov et al, 1997). In the final stages of the refinement, a limited number of water molecules (picked with ARP/wARP and verified by manual inspection) were included based on decrease of Rfree. The final model (residues 21–279, 72 water molecules, three N-linked glycosylation sites and two monovalent cations that, on the basis of coordination geometry, have been modelled as Na; Harding, 2002) has an R-factor of 0.224 (Rfree=0.274) using all data between 30.0 and 2.7 Å. The stereochemical properties of the structure were assessed by PROCHECK (Laskowski et al, 1993) and showed no residues in disallowed regions of the Ramachandran plot. Final refinement statistics are given in Table I. The coordinates have been deposited in the Protein Data Bank (entry 2C9A).

Analytical gel filtration and dynamic light scattering

A Superdex 200 10/30 column standardized with molecular weight markers (Amersham Biosciences) was used. Buffers contained either 25 mM Tris–HCl, 500 mM NaCl, pH 8.0 or 25 mM MES, 500 mM NaCl, pH 6.0. Samples were equilibrated to the same pH before loading. A lower NaCl concentration (150 mM) produced the same results. Non-His6-tagged Exμ (obtained by proteolytic cleavage of the Fc fusion construct) was used for dynamic light scattering analysis.

Surface plasmon resonance

Steady-state binding experiments were carried out in duplicate at 25°C on a BIAcore 2000 instrument (BIAcore AB) using the HBS-EP buffer (25 mM HEPES, pH 7.4, 150 mM NaCl, 3.4 mM EDTA, and 0.005% surfactant P20) (BIAcore AB). All ligands and analytes were produced from HEK293T cells as secreted, soluble proteins. C-terminally biotinylated ligands were immobilized on streptavidin-coated CM5 sensor chips (BIAcore AB) via amine coupling, immobilization level ~3000–4000 response units. The intrinsic complication of self-association of the molecules, both as analytes and as ligands, has to be considered. The latter issue was addressed by immobilizing three different Exμ ligand levels on consecutive flow cells. The medium ligand level was found to be optimal (Supplementary Figure S6) and was used for the MIF1 ligand. All curve-fitting procedures were carried out with the BIAevaluation software (version 3.0, BIAcore AB).

Supplementary Material

Supplementary Information

Figure S1

Figure S2

Figure S3

Figure S4

Figure S5

Figure S6

Figure S7

Acknowledgments

We thank M Walsh and G Sutton for assistance with crystal data collection, A Kearney and A Jefferson for help with surface plasmon resonance experiments and confocal microscopy, respectively, R Gilbert for analytical ultracentrifugation analysis, M Wilson for helpful suggestions, and N Abrescia, D Stuart and S Szedlacsek for critical reading of the manuscript. This research was funded by the Medical Research Council and Cancer Research UK. EYJ is a Cancer Research UK Principal Research Fellow.

References

  • Andreeva A, Howorth D, Brenner SE, Hubbard TJ, Chothia C, Murzin AG (2004) SCOP database in 2004: refinements integrate structure and sequence family data. Nucleic Acids Res 32 (Database issue): D226–D229 [PMC free article] [PubMed]
  • Aricescu AR, Fulga TA, Cismasiu V, Goody RS, Szedlacsek SE (2001) Intramolecular interactions in protein tyrosine phosphatase RPTPmu: kinetic evidence. Biochem Biophys Res Commun 280: 319–327 [PubMed]
  • Beckmann G, Bork P (1993) An adhesive domain detected in functionally diverse receptors. Trends Biochem Sci 18: 40–41 [PubMed]
  • Beltran PJ, Bixby JL (2003) Receptor protein tyrosine phosphatases as mediators of cellular adhesion. Front Biosci 8: d87–d99 [PubMed]
  • Bianchi C, Sellke FW, Del Vecchio RL, Tonks NK, Neel BG (1999) Receptor-type protein-tyrosine phosphatase mu is expressed in specific vascular endothelial beds in vivo. Exp Cell Res 248: 329–338 [PubMed]
  • Bilwes AM, den Hertog J, Hunter T, Noel JP (1996) Structural basis for inhibition of receptor protein-tyrosine phosphatase-alpha by dimerization. Nature 382: 555–559 [PubMed]
  • Blanchetot C, den Hertog J (2000) Multiple interactions between receptor protein-tyrosine phosphatase (RPTP) alpha and membrane-distal protein-tyrosine phosphatase domains of various RPTPs. J Biol Chem 275: 12446–12452 [PubMed]
  • Boraston AB, Nurizzo D, Notenboom V, Ducros V, Rose DR, Kilburn DG, Davies GJ (2002) Differential oligosaccharide recognition by evolutionarily-related beta-1,4 and beta-1,3 glucan-binding modules. J Mol Biol 319: 1143–1156 [PubMed]
  • Brady-Kalnay SM, Flint AJ, Tonks NK (1993) Homophilic binding of PTP mu, a receptor-type protein tyrosine phosphatase, can mediate cell–cell aggregation. J Cell Biol 122: 961–972 [PMC free article] [PubMed]
  • Brady-Kalnay SM, Mourton T, Nixon JP, Pietz GE, Kinch M, Chen H, Brackenbury R, Rimm DL, Del Vecchio RL, Tonks NK (1998) Dynamic interaction of PTPmu with multiple cadherins in vivo. J Cell Biol 141: 287–296 [PMC free article] [PubMed]
  • Brady-Kalnay SM, Tonks NK (1994) Identification of the homophilic binding site of the receptor protein tyrosine phosphatase PTP mu. J Biol Chem 269: 28472–28477 [PubMed]
  • Brunger AT, Adams PD, Clore GM, DeLano WL, Gros P, Grosse-Kunstleve RW, Jiang JS, Kuszewski J, Nilges M, Pannu NS, Read RJ, Rice LM, Simonson T, Warren GL (1998) Crystallography & NMR system: a new software suite for macromolecular structure determination. Acta Crystallogr D Biol Crystallogr 54: 905–921 [PubMed]
  • Chen H, He Z, Bagri A, Tessier-Lavigne M (1998) Semaphorin-neuropilin interactions underlying sympathetic axon responses to class III semaphorins. Neuron 21: 1283–1290 [PubMed]
  • Cheng J, Wu K, Armanini M, O'Rourke N, Dowbenko D, Lasky LA (1997) A novel protein-tyrosine phosphatase related to the homotypically adhering kappa and mu receptors. J Biol Chem 272: 7264–7277 [PubMed]
  • Chothia C, Jones EY (1997) The molecular structure of cell adhesion molecules. Annu Rev Biochem 66: 823–862 [PubMed]
  • Cismasiu VB, Denes SA, Reilander H, Michel H, Szedlacsek SE (2004) The MAM (meprin/A5-protein/PTPmu) domain is a homophilic binding site promoting the lateral dimerization of receptor-like protein tyrosine phosphatase mu. J Biol Chem 279: 26922–26931 [PubMed]
  • Desai DM, Sap J, Schlessinger J, Weiss A (1993) Ligand-mediated negative regulation of a chimeric transmembrane receptor tyrosine phosphatase. Cell 73: 541–554 [PubMed]
  • Durocher Y, Perret S, Kamen A (2002) High-level and high-throughput recombinant protein production by transient transfection of suspension-growing human 293-EBNA1 cells. Nucleic Acids Res 30: E9. [PMC free article] [PubMed]
  • Ensslen-Craig SE, Brady-Kalnay SM (2004) Receptor protein tyrosine phosphatases regulate neural development and axon guidance. Dev Biol 275: 12–22 [PubMed]
  • Feiken E, van Etten I, Gebbink MF, Moolenaar WH, Zondag GC (2000) Intramolecular interactions between the juxtamembrane domain and phosphatase domains of receptor protein-tyrosine phosphatase RPTPmu. Regulation of catalytic activity. J Biol Chem 275: 15350–15356 [PubMed]
  • Fuchs M, Wang H, Ciossek T, Chen Z, Ullrich A (1998) Differential expression of MAM-subfamily protein tyrosine phosphatases during mouse development. Mech Dev 70: 91–109 [PubMed]
  • Gebbink MF, van Etten I, Hateboer G, Suijkerbuijk R, Beijersbergen RL, Geurts van Kessel A, Moolenaar WH (1991) Cloning, expression and chromosomal localization of a new putative receptor-like protein tyrosine phosphatase. FEBS Lett 290: 123–130 [PubMed]
  • Gebbink MF, Zondag GC, Koningstein GM, Feiken E, Wubbolts RW, Moolenaar WH (1995) Cell surface expression of receptor protein tyrosine phosphatase RPTP mu is regulated by cell–cell contact. J Cell Biol 131: 251–260 [PMC free article] [PubMed]
  • Gebbink MF, Zondag GC, Wubbolts RW, Beijersbergen RL, van Etten I, Moolenaar WH (1993) Cell–cell adhesion mediated by a receptor-like protein tyrosine phosphatase. J Biol Chem 268: 16101–16104 [PubMed]
  • Gumbiner BM (2005) Regulation of cadherin-mediated adhesion in morphogenesis. Nat Rev Mol Cell Biol 6: 622–634 [PubMed]
  • Harding MM (2002) Metal-ligand geometry relevant to proteins and in proteins: sodium and potassium. Acta Crystallogr D Biol Crystallogr 58: 872–874 [PubMed]
  • Hellberg CB, Burden-Gulley SM, Pietz GE, Brady-Kalnay SM (2002) Expression of the receptor protein-tyrosine phosphatase, PTPmu, restores E-cadherin-dependent adhesion in human prostate carcinoma cells. J Biol Chem 277: 11165–11173 [PubMed]
  • Himanen JP, Rajashankar KR, Lackmann M, Cowan CA, Henkemeyer M, Nikolov DB (2001) Crystal structure of an Eph receptor–ephrin complex. Nature 414: 933–938 [PubMed]
  • Hoffmann KMV, Tonks NK, Barford D (1997) The crystal structure domain 1 of receptor protein-tyrosine phosphatase μ. J Biol Chem 272: 27505–27508 [PubMed]
  • Holm L, Sander C (1995) Dali: a network tool for protein structure comparison. Trends Biochem Sci 20: 478–480 [PubMed]
  • Ishmael FT, Norcum MT, Benkovic SJ, Bond JS (2001) Multimeric structure of the secreted meprin A metalloproteinase and characterization of the functional protomer. J Biol Chem 276: 23207–23211 [PubMed]
  • Jiang G, den Hertog J, Su J, Noel J, Sap J, Hunter T (1999) Dimerization inhibits the activity of receptor-like protein-tyrosine phosphatase-alpha. Nature 401: 606–610 [PubMed]
  • Johnson KG, Van Vactor D (2003) Receptor protein tyrosine phosphatases in nervous system development. Physiol Rev 83: 1–24 [PubMed]
  • Jones DT, Tress M, Bryson K, Hadley C (1999) Successful recognition of protein folds using threading methods biased by sequence similarity and predicted secondary structure. Proteins Suppl 3: 104–111 [PubMed]
  • Jones TA, Zou JY, Cowan SW, Kjeldgaard M (1991) Improved methods for building protein models in electron density maps and the location of errors in these models. Acta Crystallogr A 47: 110–119 [PubMed]
  • Kiselyov VV, Soroka V, Berezin V, Bock E (2005) Structural biology of NCAM homophilic binding and activation of FGFR. J Neurochem 94: 1169–1179 [PubMed]
  • Koop EA, Lopes SM, Feiken E, Bluyssen HA, van der Valk M, Voest EE, Mummery CL, Moolenaar WH, Gebbink MF (2003) Receptor protein tyrosine phosphatase mu expression as a marker for endothelial cell heterogeneity; analysis of RPTPmu gene expression using LacZ knock-in mice. Int J Dev Biol 47: 345–354 [PubMed]
  • Kunz B, Lierheimer R, Rader C, Spirig M, Ziegler U, Sonderegger P (2002) Axonin-1/TAG-1 mediates cell–cell adhesion by a cis-assisted trans-interaction. J Biol Chem 277: 4551–4557 [PubMed]
  • Laskowski RA, MacArthur MW, Moss DS, Thornton JM (1993) PROCHECK: a program to check the stereochemical quality of protein structures. J Appl Crystallogr 26: 283–291
  • McAndrew PE, Frostholm A, White RA, Rotter A, Burghes AH (1998) Identification and characterization of RPTP rho, a novel RPTP mu/kappa-like receptor protein tyrosine phosphatase whose expression is restricted to the central nervous system. Brain Res Mol Brain Res 56: 9–21 [PubMed]
  • McArdle L, Rafferty MM, Satyamoorthy K, Maelandsmo GM, Dervan PA, Herlyn M, Easty DJ (2005) Microarray analysis of phosphatase gene expression in human melanoma. Br J Dermatol 152: 925–930 [PubMed]
  • Morris RJ, Perrakis A, Lamzin VS (2003) ARP/wARP and automatic interpretation of protein electron density maps. Methods Enzymol 374: 229–244 [PubMed]
  • Mourton T, Hellberg CB, Burden-Gulley SM, Hinman J, Rhee A, Brady-Kalnay SM (2001) The PTPmu protein-tyrosine phosphatase binds and recruits the scaffolding protein RACK1 to cell–cell contacts. J Biol Chem 276: 14896–14901 [PubMed]
  • Murshudov GN, Vagin AA, Dodson EJ (1997) Refinement of macromolecular structures by the maximum-likelihood method. Acta Crystallogr D Biol Crystallogr 53: 240–255 [PubMed]
  • Nakamura F, Tanaka M, Takahashi T, Kalb RG, Strittmatter SM (1998) Neuropilin-1 extracellular domains mediate semaphorin D/III-induced growth cone collapse. Neuron 21: 1093–1100 [PubMed]
  • Otwinowski Z, Minor W (1997) Processing of X-ray diffraction data collected in oscillation mode. In Methods in Enzymology, Carter Jr CW and Sweet RM (eds), Vol. 276, pp 307–326. New York: Academic Press
  • Perez-Moreno M, Jamora C, Fuchs E (2003) Sticky business: orchestrating cellular signals at adherens junctions. Cell 112: 535–548 [PubMed]
  • Sap J, Jiang YP, Friedlander D, Grumet M, Schlessinger J (1994) Receptor tyrosine phosphatase R-PTP-kappa mediates homophilic binding. Mol Cell Biol 14: 1–9 [PMC free article] [PubMed]
  • Soroka V, Kolkova K, Kastrup JS, Diederichs K, Breed J, Kiselyov VV, Poulsen FM, Larsen IK, Welte W, Berezin V, Bock E, Kasper C (2003) Structure and interactions of NCAM Ig1-2-3 suggest a novel zipper mechanism for homophilic adhesion. Structure (Cambridge) 11: 1291–1301 [PubMed]
  • Sui XF, Kiser TD, Hyun SW, Angelini DJ, Del Vecchio RL, Young BA, Hasday JD, Romer LH, Passaniti A, Tonks NK, Goldblum SE (2005) Receptor protein tyrosine phosphatase micro regulates the paracellular pathway in human lung microvascular endothelia. Am J Pathol 166: 1247–1258 [PMC free article] [PubMed]
  • Terwilliger TC (2003) SOLVE and RESOLVE: automated structure solution and density modification. Methods Enzymol 374: 22–37 [PubMed]
  • Wang Z, Shen D, Parsons DW, Bardelli A, Sager J, Szabo S, Ptak J, Silliman N, Peters BA, van der Heijden MS, Parmigiani G, Yan H, Wang TL, Riggins G, Powell SM, Willson JK, Markowitz S, Kinzler KW, Vogelstein B, Velculescu VE (2004) Mutational analysis of the tyrosine phosphatome in colorectal cancers. Science 304: 1164–1166 [PubMed]
  • Zondag GC, Koningstein GM, Jiang YP, Sap J, Moolenaar WH, Gebbink MF (1995) Homophilic interactions mediated by receptor tyrosine phosphatases mu and kappa. A critical role for the novel extracellular MAM domain. J Biol Chem 270: 14247–14250 [PubMed]
  • Zondag GC, Reynolds AB, Moolenaar WH (2000) Receptor protein-tyrosine phosphatase RPTPmu binds to and dephosphorylates the catenin p120(ctn). J Biol Chem 275: 11264–11269 [PubMed]

Articles from The EMBO Journal are provided here courtesy of The European Molecular Biology Organization
PubReader format: click here to try

Formats:

Related citations in PubMed

See reviews...See all...

Cited by other articles in PMC

See all...

Links

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...