I. INTRODUCTION

Top Previous Next References

The first evidence that phosphorylation and dephosphorylation of proteins was a critical mechanism regulating protein activity came in the 1950s with the discovery that the enzymatically active phosphorylase a and the enzymatically inactive phosphorylase b are phosphorylated and dephosphorylated forms of the same enzyme (reviewed in Ref. 83). Following this discovery, numerous proteins were shown to be activated or inactivated by serine or threonine phosphorylation, including contractile proteins, cytoskeletal proteins, and cell membrane proteins (reviewed in Ref. 83). However, serine and threonine are not the only sites for protein phosphorylation, confirmed by the identification of the first tyrosine kinases: pp60^src and the epidermal growth factor (EGF) receptor (68, 165). The subsequent identification of numerous tyrosine and serine/threonine kinases indicated that the regulation of protein phosphorylation is a common posttranslational mechanism for controlling enzymatic activity. Adding credence to this hypothesis is the observation that ~30% of all cellular proteins are substrates for protein kinases (reviewed in Ref. 35).

While the evidence implicating tyrosine kinases in specific signal transduction cascades continued to mount, it took nearly a decade after the characterization of the first tyrosine kinases before the first protein with tyrosine phosphatase activity, PTP1B (163), was identified. Almost immediately after the discovery of PTP1B, the first receptor protein tyrosine phosphatase (RPTP) was identified. Sequence analysis of CD45 showed that CD45 shares a high degree of similarity to PTP1B, and subsequent biochemical characterization of CD45 showed that it too is a tyrosine phosphatase (162). Shortly after the identification of CD45 came a series of low-stringency screens with the cytoplasmic domains of CD45 in an attempt to identify other members of the RPTP family. This rapid method for RPTP identification resulted in the initial characterization of six separate classes and more than 30 different individual RPTPs in a wide variety of vertebrate and invertebrate species (reviewed in Refs. 17, 25). Recent analysis of structural and evolutionary relationships between RPTPs has expanded the number of distinct RPTP subfamilies from six to eight (2).

	II. ENZYME CATALYSIS

Top Previous Next References

With the identification of RPTPs both in vertebrates (27, 121, 150) and in invertebrates (152, 159, 181) came the observation that within the tandem phosphatase domains exist a variety of highly conserved sequences that are required for enzyme catalysis. In vitro biochemical analysis using thiol-reducing agents suggested that cysteine may play an important role in catalyzing the dephosphorylation reaction (121, 151, 152). To examine which cysteines may be catalytically important, a highly conserved sequence in the first and second phosphatase domains [HC(X)₅RT] was the target of site-directed mutagenesis (152). In the first phosphatase domain of LAR, a cysteine to serine (C-S) mutation in this sequence resulted in the loss of 99% of catalytic activity, suggesting that not only is this cysteine a catalytically important amino acid, but also that at least 99% of catalytic activity of LAR is dependent on the first phosphatase domain (121, 152). Further evidence for this cysteine's role in enzyme catalysis came from experiments showing that labeled iodoacetate, a potent inhibitor of phosphatase activity, specifically binds to this catalytic cysteine (120).

Biochemical studies on phosphatase enzyme catalysis (35-37, 186) have outlined the general mechanism for the tyrosine dephosphorylation reaction. A substrate protein with a phosphorylated tyrosine enters the active site of the enzyme and is stabilized by an interaction between two oxygen atoms in the phosphoryl group and a conserved active site arginine (reviewed in Ref. 35). The tyrosine phosphoryl group is then transferred to the catalytic cysteine via a nucleophilic attack on the phosphate, generating a thiol phosphate intermediate (29). The dephosphorylated tyrosine is ejected from the active site following protonation by a conserved aspartic acid residue, and the thiol phosphate intermediate is hydrolyzed by a water molecule, returning the enzyme to its original state (37, 186).

The first crystal structures of tyrosine phosphatases were conducted on the cytoplasmic human PTP1B (10) and on Yop51, a secreted dual-specificity phosphatase from Yersinia (153). While these two phosphatases share only 20% amino acid identity in the catalytic domain, the crystal structures are highly similar, suggesting a shared catalytic mechanism. Shortly after the crystal structures for these non-receptor protein tyrosine phosphatases were characterized, the first crystal structure for a RPTP was solved. Examination of the crystal structure of mouse protein tyrosine phosphatase (PTP)- revealed that although the overall folding of the phosphatase domains of both receptor and cytoplasmic phosphatases is highly similar, the quaternary structure of PTP- indicated that this protein is most stable in a dimerized or multimerized state. Furthermore, PTP- has a helix-turn-helix domain that forms a structural wedge, sterically blocking the catalytic site of the opposing monomer (16). Recent studies using fluorescence resonance energy transfer have suggested that cellular PTP- exists in such a dimerized state (157). These observations have led Bilwes et al. (16) to propose a model in which RPTPs may be inactivated as dimers due to a steric hindrance of the catalytic site by the opposing monomer.

The crystal structures of the phosphatase domains of several other RPTPs have recently been solved, including the first phosphatase domains of PTP-µ (16, 64) and the tandem phosphatase domains of LAR (111). Comparative analysis of the quaternary structures of these domains suggests that while PTP- crystallizes as a dimer, the crystal structure of the first phosphatase domain of PTP-µ lacks the interaction between this helix-turn-helix domain and the active site of the first phosphatase domain on the opposing monomer (64). The phosphatase domains of LAR also appear to crystallize as monomers without any extended contact surfaces between independent molecules (111). In fact, analysis of LAR's crystal packing showed that the helix-turn-helix motif that blocks the catalytic site in PTP- crystals could not interact with the active site of the opposing monomer due to steric hindrance by the second phosphatase domain (111).

As a result of the differences between the quaternary structures of LAR, PTP-µ, and PTP- crystals, a consensus model for RPTP structure and function cannot yet be constructed. Although it is possible that the second phosphatase domain of PTP- might also sterically block the helix-turn-helix interaction with the active site of the opposing PTP- phosphatase domain, until the crystal structure of the tandem phosphatase domains of PTP- is resolved this can only be hypothesized. Thus the crystal structures for both PTP-µ and LAR suggest that the model in which an inhibitory loop sterically blocks the active site of the opposing monomer is not true for all RPTPs.

	III. COMPARATIVE ANALYSIS OF TYPE IIA RECEPTOR PROTEIN TYROSINE PHOSPHATASE SEQUENCES

Top Previous Next References

The first (membrane proximal) phosphatase domain is responsible for >99% of catalytic activity, while the second phosphatase domain appears to bind multiple downstream partners (33, 134, 175). As might be expected for enzymatically active domains, the cytoplasmic phosphatase domains are the most highly conserved domains in these proteins, containing multiple stretches of perfect conservation between different RPTP subfamilies and a high overall level of amino acid identity. This high degree of amino acid identity between the phosphatase domains would also suggest that the RPTP crystal structures are highly conserved. For the crystal structures solved to date, this appears to be the case (16, 64, 70, 111) at least at the level of tertiary protein structure.

Although the second phosphatase domain of type IIa RPTPs has not been shown to have catalytic activity on any in vitro substrates, it can be converted to a catalytically active phosphatase simply by changing two amino acids (111). When these two amino acids (KNRLVN and WPEQGVP) were changed to match the corresponding amino acids in the catalytically active first phosphatase domain (KNRYVN and WPDQGVP), the second phosphatase domain showed similar catalytic activity to the wild-type first phosphatase domain.

These studies brought up the exciting possibility that the second phosphatase domain of RPTPs may have catalytic activity on a particular class of substrates in vivo, despite being inactive on the substrates tested in vitro. In most RPTPs, the majority of catalytic amino acids are present in the second phosphatase domain. Perhaps D2 is only active on a specific substrate or requires a cofactor for catalysis. Support for this hypothesis comes from studies in Drosophila using phosphatases with a catalytically inactive D-A mutation. The loss of function phenotype of the Drosophila RPTP DPTP69D can be rescued using a transgene with a D-A mutation in the first phosphatase domain, but not by a transgene with a D-A mutation in both the first and the second phosphatase domains (51). It would appear then that catalytic activity is not only present in the second phosphatase domain, but also that this activity is required for DPTP69D function. However, catalytic activity of the second phosphatase domain is unlikely to be a general characteristic of RPTPs, as several RPTPs (including PTP- and PTP-) lack the essential catalytic cysteine in the second phosphatase domain.

While the structure of the cytoplasmic domains of RPTPs is conserved across all RPTP subfamilies, the extracellular domains are highly divergent and consist of a wide variety of different structural motifs (Fig. 1). Some RPTPs have large and complex extracellular domains, whereas others have short extracellular domains with no known function (reviewed in Refs. 66, 128). On the basis of sequence conservation and analysis of RPTP family trees, this large and diverse family of proteins has been grouped into six (17, 25) or, more recently, eight (2) subfamilies. The structure and function of the four subfamilies of RPTPs implicated in nervous system development (types IIa, IIb, III, and V) are described in section IV.

View larger version (28K): [in this window] [in a new window]   Fig. 1. Schematic diagram of the eight subfamilies of receptor protein tyrosine phosphatases (RPTPs) classified to date. Type I RPTPs contain a single fibronectin type III domain (FNIII) extracellularly and two cytoplasmic phosphatase domains. The type IIa RPTPs have large extracellular domains consisting of three NH2-terminal immunoglobulin-like (Ig) domains and eight FNIII domains. Type IIb RPTPs have an extracellular meprin-A5-PTP-µ (MAM) domain, a single Ig domain, and multiple FNIII domains. Type III RPTPs have a series of FNIII domains extracellularly but are unusual in that they only have one cytoplasmic phosphatase domain. Type IV RPTPs have the shortest extracellular domains, which are often heavily glycosylated, while type V RPTP extracellular domains have a carbonic anhydrase domain, linked to a single FNIII domain. Type VII RPTPs have one cytoplasmic phosphatase domain and a short extracellular domain, while type VIII RPTPs (thought to be catalytically inactive) contain a RDGS adhesion recognition motif. Members of the type IIa, type IIb, type III, and type IV subfamilies have been implicated in the regulation of neuronal morphogenesis. (Note that these are highly schematized diagrams and do not reflect the exact structure of each RPTP listed below, but rather a stereotyped structure for the family; for example, Dlar has 9 instead of 8 FNIII domains while DPTP99A has 2 instead of 1 phosphatase domain.)

	IV. ANALYSIS OF RECEPTOR PROTEIN TYROSINE PHOSPHATASE STRUCTURE AND FUNCTION

Top Previous Next References

The type IIa subfamily of RPTPs are cell adhesion molecule-like proteins and are the most well-characterized family of RPTPs. This family includes Drosophila LAR (Dlar) and DPTP69D, Caenorhabditis elegans PTP-3, Hirudo medicinalis HmLAR1 and HmLAR2, as well as three vertebrate homologs: LAR, PTP-, and PTP-. These RPTPs have large extracellular domains consisting of multiple immunoglobulin (Ig) domains and two to nine fibronectin type III (FNIII) domains, as well as two cytoplasmic phosphatase domains. The first suggestion that type IIa RPTPs may be involved in CNS development was the demonstration that type IIa RPTPs are expressed specifically in the CNS of Drosophila (40, 159), leech (54), and several vertebrates (72, 124, 144, 146, 149, 180) during periods of axon guidance and synapse formation.

The only type IIa RPTPs in Drosophila (Dlar and DPTP69D) are expressed exclusively in the nervous system and on subsets of CNS axons (40, 85, 159, 181). DPTP69D or Dlar mutants have partially penetrant motor axon guidance defects whereby motor axons either stop short of or grow past their normal synaptic targets (38, 39, 85), as well as more highly penetrant photoreceptor guidance defects in the visual system (30, 51, 101, 112).

In Drosophila, motor axons project laterally from the CNS in a segmentally repeated pattern. After exiting the CNS, the motor axons sort into five fascicles: the intersegmental nerve (ISN), intersegmental nerve b (ISNb), and segmental nerves a, c, and d (SNa, SNc, and SNd, respectively). When labeled using an antibody to Fasciclin II, a highly stereotyped pattern of innervation is observed for each fascicle. The most carefully studied branch is ISNb, which exits the CNS along the ISN fascicle. ISNb defasciculates from ISN, turning internally near muscle 28 to project to its ventral muscle targets (Fig. 2A). ISNb axons form synapses on muscles 6/7, 12, 13, 30, 14, and 28 (87).

View larger version (26K): [in this window] [in a new window]   Fig. 2. Motor axon and central nervous system defects observed in Drosophila RPTP mutants. The ISNb motor axon guidance defects observed in RPTP mutants are diverse. Panel I shows the wild-type pattern of innervation in which the ISNb bundle defasciculates from the ISN near muscle 28 and projects interiorly to innervate muscles 28, 14, 6/7, 30, 13, and 12, while ISN and SNa project to more dorsal targets. Panels II-V show examples of the loss-of-function phenotypes observed in Dlar or DPTP69D mutants. Panel VI shows the loss-of-function phenotype for the triple knock-out in DPTP99A, DPTP69D, and Dlar, where ISNb motor axons frequently stop short of their proper targets.

In Dlar or DPTP69D mutants, two axon pathfinding defects are observed for ISNb. In many segments, ISNb axons exhibit either a "full bypass" phenotype in which all ISNb axons fail to leave the ISN pathway and project in parallel with the ISN toward inappropriate dorsal targets, or a "partial bypass" phenotype in which only a fraction of ISNb axons innervate their proper ventral targets (85). Although phenotypically similar, Dlar and DPTP69D have a subtle but distinct difference in their bypass phenotypes; Dlar exhibits a "parallel bypass" phenotype in which ISNb follows the path of ISN as a separate fascicle (85), while DPTP69D exhibits a "fusion bypass" phenotype in which ISNb remains fasciculated with ISN (38). These observations suggest that Dlar is involved in the recognition of the ventral muscle field, while DPTP69D is required for the defasciculation of ISNb from ISN.

In addition to the bypass phenotypes described above, Dlar or DPTP69D mutant ISNb axons exhibit other misrouting defects, including "U-turn," "split-detour," and "split-stall" phenotypes (38, 39). Double mutants in both DPTP69D and Dlar have more penetrant and severe ISNb axon guidance defects, suggesting that Dlar and DPTP69D cooperate during motor axon guidance in Drosophila to allow ISNb to reach its appropriate synaptic targets (38, 39).

The axon guidance defects in Dlar or DPTP69D mutants, however, are not limited to motor neurons. In the Drosophila visual system, DPTP69D and Dlar mutants also exhibit highly penetrant defects in photoreceptor axon guidance (30, 51, 101, 112). The Drosophila compound eye consists of ~800 individual ommatidia, each containing eight photoreceptors (R1-R8). R1-R6 extend axons to the optic lamina, whereas R7 and R8 grow past the lamina to terminate in the medulla (Fig. 3A). Axons from the R cells in an individual ommatidium enter the optic lobe in a highly stereotyped spatial order; a fascicle of R1-R7 axons surrounding the R8 axon, as well as in a highly stereotyped temporal order; R8 entering first, followed by R1-R6 and finally R7. Within the lamina, R1-R6 axons remain as a single fascicle that terminates in a tight cluster until midpupal development, when growth cones extend from this bundle to reach stereotyped targets across the surface of the lamina plexus (Fig. 3B).

View larger version (27K): [in this window] [in a new window]   Fig. 3. Photoreceptor axon guidance defects are observed in Drosophila RPTP mutants. A, left: wild-type pattern of axonal connectivity in which photoreceptor (R)1-6 axons project to the lamina, while R7 and R8 project to separate layers of the medulla. Middle: photoreceptor axon guidance defects observed in DPTP69D mutants, in which R1-6 frequently bypass their normal laminar targets and project to the medulla while R7 aberrantly terminates in the R8 layer. Right: targeting defects of Dlar mutant photoreceptors, in which R7 axons fail to maintain their association with proper targets in the medulla and retract to the R8 recipient layer. This mistargeting of R7 photoreceptors to the R8 layer is also seen when N-cadherin is specifically lost in R7 photoreceptors, but when all other photoreceptors are wild type. B: defects in R1-R6 targeting within the lamina are also seen in Dlar mutants. Left: normal pattern of R1-R6 connectivity in the lamina, in which R1-R6 axons defasciculate from the central position and innervate distinct laminar targets. Right: failure of R1-R6 axons to defasciculate observed in both Dlar and N-cadherin mutants.

In DPTP69D mutant embryos, R1-R6 axons frequently fail to terminate in the lamina, and instead follow the R8 axon into the medulla (Fig. 3A) (51). In addition, nearly 50% of DPTP69D mutant R7 photoreceptor axons terminate inappropriately in the R8 recipient layer instead of projecting on to the R7 layer (112) (Fig. 3A). Using transgenes to rescue these phenotypes, Garrity et al. (51) demonstrated that the FNIII domains of DPTP69D, as well as an intact phosphatase domain, are essential for normal R1-R6 target recognition in the lamina, yet expression of DPTP69D in R7 or R8 does not retarget these axons to the lamina. These data demonstrate that DPTP69D expression is not sufficient to target photoreceptor axons to the lamina, but suggests a model by which DPTP69D acts as a permissive rather than an instructive cue, allowing the R1-R7 photoreceptors to defasciculate from the R8 axon and navigate independently to their appropriate termination zones, rather than directly initiating a signal that targets individual photoreceptor axons to their appropriate target layers (112, 145).

More recently, two elegant papers describe distinct and highly penetrant photoreceptor axon guidance defects in Dlar mutants. Instead of R1-R6 spreading out from the terminal bundle in the lamina during the midpupal stages, Dlar mutant growth cones maintain their tight association (30) and fail to reach their appropriate synaptic targets (Fig. 3B). In addition, Dlar mutants exhibit a highly penetrant R7 phenotype, where Dlar mutant R7 axons retract from the normal R7 recipient layer and inappropriately target the R8 recipient layer (30, 101). Using Dlar and DPTP69D transgenes and fusion proteins to rescue the Dlar and DPTP69D mutant phenotypes, Maurel-Zaffran et al. (101) demonstrated that although the intracellular domains of Dlar and DPTP69D are interchangeable and appear to share common signaling mechanisms [as suggested by their genetic redundancy in motor axons (39) and their ability to bind the same signaling proteins (175)], the extracellular domain of Dlar is uniquely able to recognize and interact with specific extracellular molecules required for proper R7 targeting.

The inability of Dlar mutant R7 photoreceptors to maintain their appropriate termination zone can be rescued either by expressing a full-length Dlar transgene in either R7 or R8 or by expressing the extracellular domains of Dlar in R8 (101). These data suggest that Dlar functions in both a cell-autonomous and -nonautonomous manner during R7 development, acting as both a receptor for R7 axons for which catalytic activity is required and as a ligand expressed by R8. In a similar manner, during the organization of the Drosophila follicular epithelium (12, 49), Dlar also appears to mediate cell-nonautonomous effects. Dlar mutant oocytes have severe disruptions in the organization of actin filaments. In mosaic oocytes where follicle cells lacking Dlar are surrounded by wild-type cells, the disruptions in actin filament morphology in Dlar mutant clones extends to the cells immediately surrounding these clones (12, 49). However, it is not known whether these cell-nonautonomous phenotypes are a direct or an indirect consequence of Dlar loss of function. Thus, although both studies demonstrate a cell-nonautonomous function for Dlar, it is not yet known how Dlar mediates these effects, or what extracellular proteins interact with Dlar.

In Drosophila, loss of Dlar function also results in reduced synapse size and decreased terminal branch complexity (75). Ultrastructural analysis at the Drosophila neuromuscular junction also reveals a 2.5-fold increase in active zone size in Dlar mutants, while electrophysiological characterization of Dlar mutants shows a parallel reduction in evoked, presynaptic neurotransmitter release without a change in postsynaptic sensitivity (75). Thus, not only does Dlar regulate axon guidance to the proper synaptic target, but it also plays a crucial role in the morphogenesis of the developing synapse.

Following the work on type IIa RPTPs in Drosophila, two LAR homologs were identified in leech (HmLAR1 and HmLAR2). These RPTPs are expressed at high levels on the processes and growth cones of comb cells, two unusual cells that send out ~70 parallel axonlike processes (53). To perturb the function of HmLAR homologs, Gershon et al. (53) used in vivo addition of antibodies to the extracellular domain of HmLAR2. This resulted in dramatic alterations in the projection pattern of comb cell processes, in which the normally parallel comb cell processes frequently extended filopodia that contacted neighboring processes and occasionally crossed over sibling comb cell processes (53) (Fig. 4).

View larger version (31K): [in this window] [in a new window]   Fig. 4. Type IIa RPTPs are involved in process outgrowth in the leech. Wild-type comb cells send out numerous parallel processes that do not overlap (left). Right: defects observed after inhibition of HmLAR2 function, by injecting either HmLAR2 antibodies or the extracellular domain of HmLAR2, or by focal injection of HmLAR2 RNAi. After HmLAR2 perturbations, sibling comb cell processes cross each other, and comb cell growth cones display an increased frequency of filopodial contacts. Bottom panels show higher magnification schematics of the top panels.

Recent studies have shown that tagged HmLAR2 extracellular domains not only bind to comb cell processes, but also increase comb cell process contacts (9), resulting in a phenotype similar to the addition of antibodies to the extracellular domain (53). Furthermore, injections of double-stranded RNA to knock down endogenous HmLAR2 function (RNAi) resulted in a similar phenotype to the in vivo addition of HmLAR2 antibodies or the HmLAR2 extracellular domain. These results, and the suggestion that HmLAR2 may bind homophilically in vitro (9), led the authors to propose a mechanism by which HmLAR2 regulates self-repulsion of sibling comb cell processes, promoting the outgrowth of parallel processes. Their model suggests that when sibling process filopodia contact one another, HmLAR2 homophilic binding induces filopodial retraction (reviewed in Ref. 8), maintaining parallel process outgrowth (Fig. 4).

Recent studies have demonstated that the C. elegans LAR homolog PTP-3 also plays an important role in embryonic development (61). Using RNAi and genetic loss of function of PTP-3, mild defects in epidermal morphology were observed that looked highly similar to the loss of function of the C. elegans Eph receptor VAB-1. Indeed, VAB-1 mutant phenotypes were severely enhanced by the loss of PTP-3 function (61), suggesting that these proteins (a tyrosine kinase and a tyrosine phosphatase) function synergistically during C. elegans epidermal morphogenesis. Unlike the Drosophila and leech LAR homologs, however, PTP-3 does not appear to have a CNS phenotype.

Three type IIa RPTPs have been identified in vertebrates: LAR (152), PTP- (144, 180), and PTP- (123), each of which is thought to be a vertebrate homolog of Dlar, possibly arising through gene duplication (66). Surprisingly, DPTP69D appears to have no vertebrate homologs. LAR, PTP-, and PTP- are expressed in overlapping but distinct patterns in the CNS (72, 91, 130) and appear to be concentrated in the growth cones of elongating processes (53, 144, 146, 184), suggesting that these RPTPs may play a conserved role in the development of the vertebrate nervous system.

Vertebrate type IIa RPTP proteins exhibit complex spatial and temporal expression patterns of differentially spliced transcripts (124, 144, 184, 185). Alternative splicing occurs most frequently within the FNIII domains, and although the functional significance of alternative splicing is largely unknown, differential splicing appears to regulate ligand binding specificity. For example, the fifth FNIII domain of LAR can bind the extracellular laminin-nidogen complex only if exon 13 is spliced out (116). Inclusion of this small exon, or exclusion of the fifth FNIII domain, completely blocks the ability of LAR to bind laminin. In addition, alkaline phosphatase (AP)-tagged proteins representing the two neurally expressed PTP- isoforms exhibit distinct binding patterns in the chick retinotectal system (57), suggesting that the alternatively spliced extracellular domains interact differentially with ligands in the retinotectal system.

Members of this family of RPTPs are proteolytically cleaved in the membrane-proximal part of the extracellular domain, but the cleavage products remain tightly associated. LAR, for example, has a penta-arginine cleavage site near the transmembrane domain, in which a single arginine to alanine substitution can abrogate cleavage (136). The tight association between the cleaved halves of LAR is only present in growing cells; once confluence is reached, the NH₂-terminal subunit is released, whereas the COOH-terminal subunit is degraded (183). Proteolytic processing can be triggered by the addition of a calcium iontophore and appears to depend on the function of protein kinase C (1). The functional relevance of this juxtamembrane cleavage has yet to be determined, but an appealing model proposes that binding and release of associated extracellular domains may help regulate the adhesivity of certain cell-cell contacts.

Although numerous studies have examined the spatial and temporal patterns of expression of this family of RPTPs, only recently have functional studies shed light on how these RPTPs may be involved in CNS development. The first such studies done in vertebrates addressed the roles of PTP- in axon outgrowth and guidance in the developing chick retina. Chick retinal ganglion cells (RGCs) express a variety of RPTPs, including PTP-, CRYP2, PTP-µ, PTP-, and PTP- in overlapping but distinct patterns during retinal development (91). Putative ligands for PTP-, detected using alkaline phosphatase-PTP- fusion proteins as probes, are expressed on basement membranes of the optic fiber layer in the developing retina (57), specifically on the Mueller glia endfeet which make up part of this layer (90). Blocking PTP-/ligand interactions, either at the receptor level using function-blocking antibodies or at the ligand level by adding the PTP- extracellular domain, induces a more filopodial-like growth cone and reduces the average length of RGC axons on intact basement membranes, but not on laminin or matrigel (90). These data strongly suggest that PTP- acts as a receptor for a growth-promoting cue present on the glial endfeet of the retinal basement membrane and that signaling through PTP- enhances the rate of axon elongation as well as the formation of a more lamellipodial growth cone in vitro (90, 107). The identity of at least one ligand for PTP- has recently been determined (see sect. VIA).

PTP-, like HmLAR2, is a homophilic cell adhesion molecule, but in contrast to HmLAR2, PTP- homophilic interactions serve to promote neurite outgrowth and adhesion for forebrain neurons in vitro (172). In fact, a soluble gradient of the extracellular domain of PTP- can mediate attractive turning of forebrain neurons (155), suggesting that while the extracellular domains of HmLAR2 and PTP- both serve as homophilic cell adhesion molecules, the signal transduction cascades activated by these homologous proteins generate opposing effects on outgrowth. While HmLAR2 mediates repulsive turning away from sites of homophilic interactions and prevents the fasciculation of sibling comb cell processes (9), PTP- mediates attractive responses toward sites of homophilic contact and may enhance the fasciculative outgrowth of axons expressing this receptor (172).

Possible in vivo roles for all three type IIa RPTPs (LAR, PTP-, and PTP-) were recently examined in the developing Xenopus visual system. The expression of catalytically inactive LAR, PTP-, and PTP- constructs [consisting of the cytoplasmic phosphatase domains with a cysteine to serine (C-S) mutation in the first phosphatase domain] caused specific and significant changes in the rate of axon elongation (73). Expression of C-S PTP- decreased the rate of retinal ganglion cell axon elongation, either in vivo or on retinal basement membranes in vitro. Expression of C-S PTP- caused a significant increase in the rate of axon outgrowth on retinal basement membranes, whereas expression of C-S LAR had no effect either in vitro or in vivo. Interestingly, while expression of these C-S RPTPs significantly altered the rate of retinal ganglion cell axon outgrowth, they did not appear to affect retinal ganglion cell axon guidance. Retinal ganglion cells expressing C-S LAR, C-S PTP-, or C-S PTP-, or even all three C-S RPTP constructs, were able to navigate properly from the retinal ganglion cell layer to the optic tectum (73). These studies suggest that while the rate of RGC axon extension is influenced by type IIa RPTPs, RGC axon guidance was not affected after expression of catalytically inactive RPTP constructs.

The analysis of RPTP knock-out mice has also enhanced our understanding of how type IIa RPTPs function in the developing CNS. Mice homozygous for a deletion in LAR exhibit subtle phenotypes consisting of a reduced cholinergic innervation of the dentate gyrus and a reduction in the size of basal forebrain cholinergic neurons (182). The subtle defects observed in these mice may not necessarily reflect a true loss of LAR function, because low levels of LAR expression were seen in these knock-out mice (139, 182). LAR is also required for proper peripheral nerve regeneration, because LAR knock-out mice exhibit reduced regrowth of the sciatic nerve after a lesion (167). In fact, the expression of multiple type IIa RPTPs (including LAR and PTP-) is responsive to sciatic nerve crush, although the magnitude and direction of altered expression is debatable; one study finds a 50% increase in the levels of PTP- mRNA and a 50 and 20% decrease in LAR and PTP- (62), whereas another study showed increased LAR protein expression following sciatic nerve crush (178).

Mutant mice lacking either PTP- or PTP- have more obvious neurological effects than mice lacking LAR. Mice lacking PTP- exhibit motor coordination deficits, pituitary abnormalities, delayed development, and hypomyelination (46, 170), whereas PTP- mutant mice have memory deficits and a hyperpotentiation of CA1 and CA3 hippocampal synapses (164). None of these defects is as severe as the homozygous lethal defects seen when Dlar is mutated in Drosophila, suggesting that in vertebrates the loss of one of the vertebrate LAR family members may be partially compensated for by the function of other vertebrate LAR family members (reviewed in Ref. 17). Double and triple mutant mice may help address whether the type IIa RPTPs have overlapping functions in the development of the vertebrate CNS.

One of the most surprising aspects of the LAR subfamily of RPTPs is that these homologous proteins appear to function so differently from one other. Despite sharing 66-71% amino acid identity, and interacting with common cytoplasmic effectors like liprin-, LAR, PTP-, and PTP- appear not to exhibit homologous functions. Some type IIa RPTPs appear to mediate homophilic binding while others interact with heterotypic ligands; ligand-receptor interactions for some RPTPs mediate repulsion and defasciculation and for others mediate attraction and fasciculation. As such, any detailed model outlining the function of the type IIa RPTPs, let alone the entire RPTP family, seems unlikely to be correct; rather, it appears that the type IIa RPTPs may play both overlapping and nonoverlapping roles in CNS development.

The type IIb family of RPTPs consists of at least five identified members in vertebrates: PTP-µ, PTP-, PTP-, PTP-, and PTP-, but none in Drosophila. These RPTPs are characterized based on the presence of a meprin-A5-PTP-µ (MAM) domain at the NH₂ terminus of the protein, one Ig domain, and multiple FNIII domains in the extracellular domain. Type IIb RPTPs are expressed in distinct patterns during development, with high levels of expression in the developing CNS (71, 102, 129, 179) including expression in retinal ganglion cells (26, 72, 91, 147). These RPTPs, like the type IIa RPTPs discussed earlier, are cleaved in the membrane-proximal extracellular domain and appear to maintain tight association between the cleaved fragments (71).

The most extensively studied members of this family are PTP-µ and PTP-, both of which are homophilic, neurite outgrowth-promoting, cell adhesion molecules (21, 23-25, 44, 52, 189). Homophilic binding of these RPTPs (see sect. VIB) is independent of phosphatase activity, as catalytically inactive C-S mutant RPTPs, or constructs lacking the phosphatase domains entirely, retain homophilic adhesive properties (52). PTP-µ associates with and appears to mediate its axon outgrowth-promoting effects through interactions with cadherins (22, 26). Downregulation of PTP-µ expression using antisense oligonucleotides, or expression of a catalytically inactive C-S mutant PTP-µ, can decrease retinal ganglion cell axon outgrowth on an N-cadherin substrate (26). This demonstrates that catalytic activity is essential for PTP-µ function and confirms that PTP-µ is a key regulator of N-cadherin-mediated neurite outgrowth.

The type III subfamily of RPTPs consist of four Drosophila proteins (DPTP99A, DPTP10D, DPTP52F and DPTP4E) and five vertebrate RPTPs, including PTP-, DEP1, SAP1, PTPS31, and a vertebrate gene with homologs called CRYP-2 in chick (20), GLEPP-1 in rabbits (158), RPTP-BK in rats (156), mGLEPP/mPTPRO in mice (161, 173), and PTP-U2/GLEPP1 in humans (133, 174). Although an attempt has been made to unify the nomenclature and call these homologous genes PTPRO (142), the presence of a nonhomologous human type IIb RPTP already named PTPRO (4) suggests that another name may be more well suited. Nevertheless, because we will refer exclusively to studies on the chick homolog, we will follow convention and call this gene CRYP-2/cPTPRO.

Type III RPTPs are characterized by the presence of several extracellular FNIII domains and a single cytoplasmic phosphatase domain (with the exception of PTP99A which has two cytoplasmic phosphatase domains). PTP99A, PTP52F, and PTP10D are expressed selectively in the CNS of Drosophila (131, 181), whereas PTP4E appears to be broadly expressed early in development (117). The most extensively studied vertebrate type III RPTP is CRYP-2/cPTPRO, which is expressed by retinal ganglion cells (20, 91) during periods of axonogenesis and axon guidance from the retina to the tectum.

Although no phenotype has been observed for flies lacking DPTP99A, this RPTP genetically interacts with the type IIa RPTPs, Dlar and DPTP69D. Double mutants in DPTP69D and DPTP99A exhibit more highly penetrant motor axon guidance defects than in DPTP69D alone, causing >85% of ISNb nerves to exhibit bypass, detour, or stall defects (38) (Fig. 2). The bypass phenotype observed in DPTP69D/DPTP99A double mutants is subtly different from that observed in Dlar loss of function (as described above), in that DPTP69D/DPTP99A mutants exhibit a failure to defasciculate from the ISN, while Dlar mutants frequently bypass the ventral muscle field as a separate but parallel fascicle to the ISN (85).

The motor axon guidance phenotypes for Dlar, DPTP69D, and DPTP99A mutations, as well as double and triple mutations in these genes, strongly support the hypothesis that these RPTPs do not function in isolation to recognize guidance cues at unique choice points; rather, guidance defects observed in RPTP mutants are more pleiotrophic. This might suggest a role for RPTPs in regulating the fidelity of choice point navigation; the loss of RPTP function results in an increased frequency of guidance errors in a certain set of axons, while the loss of function of multiple RPTPs enhances the frequency of guidance errors. One might hypothesize then that with each additional RPTP mutated, one would observe an increase in axon guidance errors. Although this appears to explain many of the double and triple mutant phenotypes, the guidance defects of multiple RPTP mutants are not strictly additive. Analysis of double and triple mutants in DPTP69D, DPTP99A, and Dlar showed that these RPTPs act both cooperatively and competitively during motor axon guidance in Drosophila, sometimes enhancing and sometimes suppressing the phenotypes of individual RPTP mutants (39). For example, mutations in DPTP99A suppress the ISNb bypass phenotype in Dlar but enhance the Dlar ISN stop-short phenotype (39). These data demonstrate that the function of RPTPs is not strictly redundant and illuminate the fact that genetic interactions between different RPTP subfamilies are likely to be complex.

Type III RPTPs are also involved in axon guidance at the midline of the Drosophila CNS. Although single mutations in DPTP10D and DPTP99A alone have no obvious CNS phenotype, double, triple, and quadruple RPTP mutations have severe and highly penetrant CNS defects. DPTP69D loss of function causes breaks in the third (outermost) longitudinal fascicle and also results in the inappropriate growth of longitudinal axons across the midline (154). These phenotypes are dramatically enhanced in DPTP69D/DPTP10D double mutants and are much more severe in triple or quadruple (Dlar, DPTP69D, DPTP99A, DPTP10D) RPTP mutants, in which all detectable longitudinal tracts are converted into commissural pathways (154). These data suggest that tyrosine phosphorylation controls the responsiveness of growth cones to the repulsive cues at the CNS midline in Drosophila. Adding credence to this hypothesis was the recent discovery that DPTP69D and DPTP10D interact genetically with genes known to regulate axon guidance at the Drosophila midline, such as Roundabout (Robo), Slit, and Commisureless (Comm) (79, 80, 132). From these studies, a model has emerged in which DPTP10D and DPTP69D function as positive regulators of Slit/Roundabout repulsive signaling that prevents the longitudinal tracts from crossing midline of the Drosophila CNS (154).

Mutations in DPTP52F cause multiple CNS and motor axon guidance defects. DPTP52F loss of function causes breaks in the third longitudinal fascicle as well as the failed branching of SNa (131). Surprisingly, Dlar mutations are able to rescue the DPTP52F CNS phenotype, without having an effect on the ISN or SNa phenotypes, while genetic interactions of DPTP52F with either DPTP10D or DPTP69D are entirely synergistic (131). These data provide further evidence for both competitive and cooperative interactions between the different Drosophila RPTPs and confirm that individual guidance choices are likely to involve the integration of information from several RPTP signaling pathways.

The chicken type III RPTP CRYP-2/cPTPRO is also expressed strongly in the CNS, including expression on RGCs during periods of axon outgrowth and guidance (20) as well as in the target for RGC axons, the optic tectum (91). CRYP-2/cPTPRO is alternatively spliced in the juxtamembrane region, similar to the type IIa and type IIb RPTPs. With the use of Fc-CRYP-2/cPTPRO fusion proteins, putative ligands for CRYP-2/cPTPRO were identified on numerous retinal cell types, including RGCs (142). Recently, elegant studies have shown that CRYP-2/cPTPRO functions as a repulsive molecule for RGC axons in vitro, inducing RGC growth cone collapse, inhibiting retinal neurite outgrowth, and acting as a chemorepulsive cue in a growth cone turning assay (142). Although a model is emerging suggesting that CRYP-2/cPTPRO receptor ligand interactions may be involved in the formation of the retinotectal projection, the in vivo roles of CRYP-2/cPTPRO are only beginning to be understood, because a receptor for CRYP-2/cPTPRO is not yet known.

The final family of RPTPs implicated in axon guidance and outgrowth has two identified family members (PTP-/RPTP- and PTP-) that are expressed in the developing CNS (11, 84). These RPTPs are chondroitin sulfate proteoglycans with secreted extracellular domains containing a carbonic anhydrase-like domain and a single FNIII domain (74) and are expressed at high levels in the CNS during development, including expression on cortical cells (60, 91, 114, 137, 138, 168) during periods of migration and process initiation.

In vertebrates, the secreted ECD of PTP-/RPTP- is also known as phosphacan and can bind a number of neurite outgrowth-promoting cell adhesion molecules including tenascin, axonin-1, contactin, F3, NCAM, NrCAM, and NgCAM (reviewed in Ref. 17). Interactions between phosphacan and these extracellular matrix molecules seem to promote axon outgrowth and cortical cell migration, demonstrated by the fact that the addition of soluble phosphacan can inhibit outgrowth on NgCAM, presumably by blocking the interactions of endogenous PTP-/RPTP- with its ligand(s) (105) and the observation that pleiotrophin-PTP/RPTP interactions can promote the migration of rat cortical neurons in vitro (97). However, because the vast majority of the extracellular domain of PTP/RPTP in the brain is found as soluble phosphacan, competition for ligands between phosphacan and PTP/RPTP is almost certain to exist (100). The potential in vivo role of ligand masking by phosphacan is only beginning to be addressed.

Recent studies have also implicated PTP-, a type IV RPTP, in CNS development. With the use of morpholinos to knock down PTP- expression, dramatic changes in retinal morphology were observed, including defects in retinal lamination and cell fate determination (166). The molecular mechanisms underlying these defects are, as yet, poorly understood.

	V. SIGNAL TRANSDUCTION DOWNSTREAM OF RECEPTOR PROTEIN TYROSINE PHOSPHATASES

Top Previous Next References

Unlike receptor tyrosine kinases (RTKs) which conveniently tag their substrates with phosphate groups in response to ligand binding, no simple labeling strategy and few biologically relevant ligands have been available to dissect signaling events downstream of RPTP family members. Despite these limitations, the advent of yeast interaction trap technology, PTP substrate-trapping mutations, and genetic analysis has opened the door to an emerging picture of RPTP signal transduction pathways. Although the picture is far from complete, one can now trace a path from cell surface to intracellular effectors for several RPTPs.

One simple prediction for any catalytically active RPTP is that signaling will require an intimate functional relationship with at least one protein tyrosine kinase (PTK). Indeed, genetic studies of Dlar identified the Ableson protooncogene (Abl) PTK as a regulator of the Dlar signaling pathway (175). Abl and Dlar display a potent antagonistic relationship in vivo. Reduction of Abl gene dose by half suppresses the axon guidance phenotype of multiple Dlar mutant backgrounds (175). Conversely, overexpression of Abl in postmitotic neurons results in a Dlar-like phenotype dependent on an intact kinase active site. Coexpression of wild-type Abl and Dlar reverses this effect, suggesting that Dlar is capable of dephosphorylating the relevant phosphoprotein substrates. Loss of Abl alone disrupts axon pathway formation, demonstrating that both Dlar and Abl are essential for axonal development (176). Although genetic interaction assays do not prove direct interactions, the Dlar cytoplasmic domain can recruit Abl from cellular extracts, binds to purified Abl, and can dephosphorylate the kinase in vitro (175). These findings suggest that Dlar and Abl mediate a phosphorylation-dependent switch that controls axon guidance behavior (Fig. 5).

View larger version (22K): [in this window] [in a new window]   Fig. 5. Schematic diagram representing signal transduction cascades downstream of Dlar. Dlar has been shown to interact genetically with Trio, a protein that activates the Rho family of small GTPases. Dtrio exhibits GEF activity on Drac, which can indirectly promote the formation of new actin filaments. The cytoplasmic domains of Dlar also bind to Liprin-, a regulator of synapse formation. Dlar can also directly bind and dephosphorylate Abl and Ena. Ena is a shared substrate for Dlar and Abl and appears to form a phosphorylation-dependent switch controlled by the activity of Dlar and Abl.

Although vertebrate Abl homologs have been studied extensively in the context of cancer biology, cell proliferation, and DNA damage response (reviewed in Ref. 140), relatively little is known about the role of Abl in controlling cell motility and cell shape. However, a growing body of evidence suggests that Abl and related genes also control cellular morphogenesis and the assembly of actin cytoskeleton (82, 89). But what intracellular effectors mediate this set of Abl functions? Genetic screens to unravel the Abl pathway in Drosophila identified several potential substrates, including Enabled, Disabled, and others (reviewed in Ref. 63). Because the tyrosine phosphorylation state of Enabled (Ena) is dramatically reduced in Abl mutants (55), Ena is a prime candidate. Indeed, phenotypic analysis of Ena mutants reveals a Dlar-like axon guidance phenotype (175). Like Abl, Ena binds to the Dlar (and DPTP69D) phosphatase domains, is phosphorylated by Abl, and dephosphorylated by Dlar in vitro (175), consistent with the overlap in axonal phenotypes. Indeed, it would appear that Ena serves as a molecular switch, turned on and off by the relative activities of both Abl and Dlar.

The Ena family of proteins includes three vertebrate homologs: mammalian Ena (Mena), the vasodilator stimulated phosphoprotein (VASP), and Ena-VASP-like (Evl; Ref. 56). Analysis of Mena and VASP function in mammalian cells has shown these proteins to be key regulators of actin cytoskeletal assembly and cell motility (reviewed in Ref. 88). In fibroblasts, Mena acts to promote actin assembly by regulating the average length of actin polymers, providing a means of regulating the velocity of cell movement (14, 15; Bear et al., unpublished data). Although Ena contains a number of protein interaction motifs, leading edge cell motility depends primarily on an actin-binding domain that competes with actin-capping protein to allow polymer growth (15). Consistent with this model, injection of cytochalasin D (a small molecule that binds to the barbed ends of microfilaments) at doses that mimic an increase in capping protein results in an axon guidance phenotype very similar to loss of Ena or Dlar (76). These data suggest a model where Dlar and Ena cooperate to control the rate of growth cone motility and leading edge exploration at key choice points to ensure a high fidelity in axon guidance decisions.

In addition to actin, Ena family members bind to several intracellular partners, including the actin-binding protein profilin (125). Like Abl and Ena, profilin is expressed abundantly in the developing nervous system (169, 176). Interestingly, genetic analysis of profilin in Drosophila reveals axon outgrowth defects identical to those found in Abl mutants (176), suggesting that Abl and profilin cooperate. Potent genetic interactions between Abl and profilin in vivo support this hypothesis (176). Profilin has been shown to promote and antagonize actin assembly in different contexts (reviewed in Ref. 65); however, the precise role of profilin in the Abl pathway is still a mystery. Other Abl-associated proteins are also likely to contribute to the Abl pathway, including the adenylyl cyclase-associated protein (CAP), which binds to actin monomers and inhibits polymer assembly. Although Drosophila CAP does cooperate with Abl in controlling certain axon guidance behaviors, and associates with Abl, Ena, and profilin in Drosophila cells (Wills et al., unpublished data), the role of CAP in motor axon guidance and the Dlar pathway is unknown.

Interestingly, Dlar and DPTP69D appear to share at least some of the same signaling machinery during axon guidance in Drosophila (101). The Abl pathway may represent part of this overlap, since Enabled binds directly to the cytoplasmic domain of DPTP69D (175). However, genetic interaction experiments have not been performed to test the functional relationship between DPTP69D and the Abl pathway.

Although Abl and Ena provide direct links between Dlar and cytoskeletal effectors, the pathway seems to be far more complex. Studies of the cytoplasmic domain of mammalian LAR identified an interacting protein called Trio, which contains two Dbl oncogene-homologous guanine nucleotide exchange factor (GEF) domains, the motifs that activate GTPases in the Rho subfamily (33). Characterization of the mammalian Rho family GTPases, Rac, Rho, and Cdc42, in nonneuronal cells showed that these proteins control a variety of actin-dependent cell motility behaviors (reviewed in Ref. 59). Moreover, analysis of Rho family function in Drosophila and C. elegans showed that these GTPases play crucial roles in axonal and cell migration (76, 95, 188). In particular, overexpression of dominant negative Rac1 in postmitotic neurons results in a Dlar-like axon guidance phenotype, consistent with a model where Trio acts downstream of LAR to activate Rac in motor growth cones (76). Accordingly, Rac1 and Dlar display specific, dose-sensitive genetic interactions indicative of a cooperative relationship in vivo (76).

The analysis of genetic loss of function of Rac1, Rac2, and Mtl genes in Drosophila development is only beginning to be examined. However, these alleles have demonstrated that the dominant negative Rac constructs do not necessarily recapitulate the loss-of-function allele. For example, the dominant negative isoforms of Rac have been shown to regulate planar cell polarity in both the wing and the retina (45, 47), but the triple loss of function of Rac1, Rac2, and Mtl has no similar phenotype in these tissues (58). As a result, the exact role of the Rho family of GTPases in motor axon guidance has yet to be determined.

Although embryonic lethal mutations in vertebrate Trio exist (115), axon pathfinding in these mutants has not been examined. However, the neuronal functions of this GEF protein have been extensively studied in both C. elegans and Drosophila. This analysis shows that loss of Trio function yields striking defects in axon guidance, outgrowth, and cell migration (6, 13, 93, 113, 143). These axon pathway defects are widespread throughout the nervous system, suggesting that Trio function is not limited to the Dlar pathway. However, dose-sensitive genetic interactions exist between Trio and Dlar, supporting the hypothesis that Trio mediates some of Dlar's downstream pathway (13). Further support for this model comes from the fact that Trio was independently identified as a genetic enhancer of Abl that also interacts with Ena in vivo (93).

While multiple studies implicate Rac1 downstream of Trio, the role of Rho in the pathway is controversial. Although in vitro nucleotide exchange factor assays with mammalian Trio confirm that GEF domain 2 can activate RhoA (33), parallel assays of the invertebrate genes show little if any catalytic activity in this domain (113, 143). Site-directed mutations in both GEF domains also show that retinal axons do not seem to require catalysis by GEF domain 2 (113). However, analysis of Trio function in the Drosophila brain suggests a potential overlap with Rho (5). Interestingly, in the Drosophila oocyte where Dlar regulates actin microfilament bundle polarity and helps define the shape of the developing egg, loss of Rho activity yields a phenotype highly reminiscent of Dlar loss of function (12).

In addition to intracellular proteins with catalytic activities, LAR family RPTPs recruit a group of proteins called liprins that seem to provide a scaffold linking LAR to a network of other proteins (Fig. 5). Liprins fall into related - and -subtypes that are highly conserved from worm to fly to human (75, 134, 135). Liprins are small proteins composed of NH₂-terminal coiled-coil domains and COOH-terminal domains that contain steryl alpha motif (SAM) repeats and are expressed in the developing nervous system (75, 134, 135, 187). Liprin- proteins bind directly to the second phosphatase domain of LAR family members and associate with all members of the vertebrate LAR family (LAR, PTP-, and PTP-; Ref. 124).

Although the functions of vertebrate liprins are unknown, genetic analysis in C. elegans and Drosophila have shown that liprin- plays a crucial role in nervous system function. In a screen for C. elegans lacking normal synapse structure, mutations in a liprin- ortholog (syd-2) were identified (187). In these mutants, neurotransmitter-filled synaptic vesicles fail to cluster appropriately at neuromuscular junctions (NMJs). Ultrastructural analysis reveals that the organizing centers for neurotransmitter release (active zones) are increased in size by roughly twofold (187). Behavioral defects in syd-2 mutants are consistent with a functional deficit in synaptic transmission.

In Drosophila, loss of liprin- results in a reduction in both synapse size and terminal branch complexity (75), an identical phenotype to Dlar loss of function at the synapse (75). Ultrastructural analysis at the Drosophila NMJ shows a 2.5-fold increase in active zone dimensions in both liprin- and Dlar mutants, confirming that the role of liprins has been well-conserved. Moreover, electrophysiological characterization of both liprin and Dlar mutants exhibit a parallel reduction in evoked, presynaptic neurotransmitter release without a change in postsynaptic sensitivity (75).

Although liprin- proteins associate with LAR RPTPs, their contribution to the LAR mechanism is largely unknown. In this regard, it is interesting that genetic epistasis analysis in Drosophila shows that liprin- is required for Dlar function (75). However, different models can explain this relationship. One model for liprin function is that these intracellular components deliver or recruit LAR receptors to sites of cell-cell interaction, such as focal adhesions (134). Alternatively, liprins may be scaffolding proteins that physically link LAR receptors to targets of PTP regulation. Although liprins lack catalytic domains, signaling motifs, and tyrosine phosphorylation, accumulated evidence suggests that liprins interact with a network of other proteins at the synapse, including the multiple PDZ-domain containing protein GRIP (177).

A number of studies have shown that LAR family RPTPs localize to sites of cell-cell and cell-substrate interactions in nonneuronal cells. For example, liprin- and LAR localize to focal adhesions (FAs) in fibroblasts (134), and Dlar colocalizes with Enabled and integrin receptors at FA-like junctions in epithelial cells surrounding the Drosophila oocyte (12). One additional type of cell-cell contact abundant in LAR protein is the adherens junction that forms a stable bond between neighboring cells. Here, LAR appears to associate with cadherins, the major cell adhesion molecule (CAM) responsible for the formation of adherens junctions (1). Like other CAMs, cadherin function is dependent on a linkage to the cortical actin cytoskeleton. To accomplish this, cadherins recruit members of the -catenin/Armadillo family of proteins (94). Studies in both epithelial and neuronal cell lines reveal that LAR receptors associate with catenins (86, 108), forming a protein complex likely to link LAR to different cadherins. Interestingly, tyrosine phosphorylation of -catenin plays a key role in the initiation of cell migration by increasing the free intracellular pool of the protein (108). LAR is capable of dephosphorylating -catenin, preventing the release of catenins and thus blocking the movement of epithelial cells (108). While the importance of the relationship between LAR and -catenin during axon guidance is not known, the proteins do associate in PC12 cells (86). It is known, however, that neuronal cadherins play an important role in the formation of axon pathways in the Drosophila embryo (69) and in vertebrates (18, 43, 99, 118, 126, 148). Moreover, the -catenin of Drosophila (Armadillo) appears to be important for some aspects of axonogenesis in the same system, where it also interacts with mutations in the Abl protein tyrosine kinase (94). These observations suggest that LAR receptors may modulate cadherin function through a partnership between Abl and -catenin; however, this has yet to be demonstrated.

Interestingly, the retinal phenotypes observed for Dlar loss of function are nearly indistinguishable from the phenotypes observed for N-cadherin loss of function. In both cases, R1-R6 photoreceptor axons reach the lamina, but once in the lamina fail to extend out of the ommatidial bundle (Fig. 3). In addition, R7 axons also fail to terminate in their appropriate position, and instead project to the R8 recipient layer (92). These data, together with the demonstration that vertebrate LAR associates with cadherins, provide strong circumstantial evidence that Dlar and N-cadherin function together to regulate photoreceptor axon guidance. However, subtle phenotypic differences, such as the observation that N-cadherin is required for proper topographic map formation in the lamina and medulla, while Dlar plays no role, suggest that these genes also have independent roles during photoreceptor axon guidance.

Interestingly, multiple RPTP subfamilies interact with catenin-cadherin complexes. PTP-µ has been shown to associate with cadherins and both - and -catenins in a variety of tissues, including brain (23). Both PTP-µ and PTP- can bind -catenin, and PTP- can catalyze the dephosphorylation of -catenin in vitro (50). The PTP-µ interaction appears to be mediated by direct binding of the receptor cytoplasmic domains. Consistent with a functional interplay between PTP-µ and cadherins in the developing nervous system, axon outgrowth on cadherin substrates has been shown to require PTP-µ activity in cultured neurons (26).

The type V RPTP PTP/RPTP has been shown to interact with a variety of extracellular ligands, and studies have recently begun to identify several cytoplasmic substrates. The phosphatase domains of PTP/RPTP bind to -catenin and can catalyze the dephosphorylation of -catenin in vitro (103). Interactions of the extracellular domains of PTP/RPTP with pleiotrophin cause an inhibition of PTP- catalytic activity and a concomitant increase of -catenin phosphorylation (103). This is the best experimental evidence supporting the model that ligand binding can regulate the catalytic activity of RPTPs and prompted the authors to propose a ligand-induced receptor inactivation model, in which ligand binding causes an inhibition in receptor enzymatic activity (103).

PTP/RPTP also binds to the postsynaptic density scaffold protein PSD-95/SAP90 (78), as well as the G protein-coupled receptor kinase-interactor 1/Cool-associated, tyrosine-phosphorylated 1 (GIT1/Cat-1) (77). However, the functional relevance of these interactions has yet to be uncovered.

Although relatively little is known about the intracellular partners of other RPTPs, there are a few hints. For example, biochemical experiments in Drosophila identified a substrate for DPTP10D called gp150 (160). This leucine-rich repeat bearing transmembrane glycoprotein forms a stable complex with DPTP10D in cultured Drosophila cells, where gp150 is phosphorylated by a member of the Src PTK family and can be dephosphorylated by either DPTP10D or DPTP99A (48). Interestingly, Src and a 40-kDa phosphoprotein form a complex with gp150 in this context. While gp150 is a good substrate for the v-Abl PTK in vitro, the in vivo function of gp150 remains unknown.

Interactions of RPTPs with the Src tyrosine kinase family are not limited to DPTP10D. Previous studies have demonstrated an interaction between CD45 and the tyrosine kinases Lck and Fyn (109, 110). In contrast to the notion that RPTPs could serve strictly to negatively regulate PTKs, CD45 can dephosphorylate an inhibitory COOH-terminal phosphotyrosine, thereby activating these tyrosine kinases (109, 110).

Although ample evidence suggests that RPTP signaling involves intracellular PTKs, RTKs are also likely to participate in RPTP pathways as well. One nice example comes from the analysis of the fibroblast growth factor receptor (FGFR) ortholog in C. elegans (42, 81). Genetic analysis in C. elegans identified the type II RPTP Clr-1 as an antagonist of the FGFR ortholog egl-15 (81). In this system, the Clr-1 phenotype is suppressed by loss of egl-15 and can be mimicked by overexpression of egl-15 alone (81). Because Clr-1 requires catalytic PTP activity to function during C. elegans development, it is likely that FGFR and Clr-1 share a set of functionally relevant substrate proteins. Although FGFR is known to play a role in axon outgrowth and guidance, as well as in transducing signals from other CAMs (such as N-cadherin), the identity of shared substrates for Clr-1 and the FGFR is still unknown.

Recent studies have outlined two players in a novel signaling mechanism downstream of PTP-µ. With the use of the first phosphatase domain of PTP-µ in a yeast two-hybrid screen, the receptor for activated protein kinase C (RACK1) was identified as a candidate interactor (106). This scaffolding protein has been implicated in shuttling proteins to the plasma membrane, suggesting that PTP-µ may use RACK1 to recruit other signaling proteins to sites of cell-cell contact (106). Indeed, PKC- has recently been shown to be required for neurite outgrowth downstream of PTP-µ and is also present in a protein complex with PTP-µ and RACK1 (127). Together, these data suggest that PTP-µ signals through RACK1, which recruits PKC- and results in the promotion of neurite outgrowth (127).