The Plexin family of transmembrane proteins appears to function as repulsive receptors for most if not all Semaphorins. Genetic and biochemical analysis in Drosophila has been used to show that the transmembrane protein Off-track (OTK) associates with Plexin A, the receptor for Sema 1a, and that OTK is a component of the repulsive signaling response to Semaphorin ligands. In vitro, OTK associates with Plexins. In vivo, mutations in the otk gene lead to phenotypes resembling those of loss-of-function mutations of either Sema1a or PlexA. The otk gene displays strong genetic interactions with Sema1a and PlexA, suggesting that OTK and Plexin A function downstream of Sema 1a (Winberg, 2001).
Immunoprecipitated human Plexins A3 and B1 copurify a number of proteins from BOSC-23 cell extracts, some of which became tyrosine phosphorylated in an in vitro kinase assay. Western blotting has indicated that this activity is not due to the presence of Met, Ron, Abl, or Src tyrosine kinases. The most prominent labeled band other than Plexins is approximately of 160 kDa (Winberg, 2001).
To identify the putative Plexin-associated protein, candidates in Drosophila were considered. Several proteins with homology to receptor-tyrosine kinases have been identified that are expressed in the CNS and could potentially interact with Plexins. However, in the cases where the loss-of-function phenotypes have been assayed, there is not a notable similarity with those described for Semaphorins or Plexins, suggesting an unrelated function (e.g., EGFR, FGFR, Derailed). In other cases, in vivo functional data are yet lacking, but some of these candidates may be considered less probable on the basis of molecular weight (e.g., Dror, Nrk). A leading remaining contender is the Drosophila protein Off-track (OTK; previously called Dtrk) (Winberg, 2001).
Based on its molecular weight, the observation that it can be tyrosine phosphorylated, and its expression on axons at the appropriate time in development to play a role in axon guidance, OTK seemed like a good candidate for possible interaction with Plexin. BLAST searches of protein databases, using either the cytoplasmic kinase or extracellular domain, indicate that the closest relatives of OTK are the chick protein KLG and its human homolog CCK4/PTK7 (Winberg, 2001).
As a first test of OTK protein function, molecular association was examined in COS cells. Epitope-tagged versions of both OTK and a variety of Drosophila and mammalian Plexins (DPlexA, PlexA3, and PlexB1) were generated and tested for expression. The cytoplasmic domains of Plexins are highly conserved, and, thus, binding relationships are likely to be conserved across phylogeny. Cells were cotransfected to express both proteins, and the formation of complexes was analyzed by immunopurification and Western blotting (Winberg, 2001).
Drosophila PlexA (HA tagged) copurifies with immunoprecipitated OTK (myc tagged). Moreover, mammalian PlexA3 and PlexB1 (VSV tagged) also copurify with immunoprecipitated OTK (myc tagged). OTK can copurify all three Plexins but not an unrelated protein, the netrin receptor DCC. In addition, OTK (myc tagged) copurifies with immunoprecipitated mammalian Plexin A3 and B1 (VSV tagged). OTK is copurified in a similar fashion with immunoprecipitated Drosophila Plexin A (HA tagged). These experiments identify OTK as a transmembrane protein that can constitutively associate with both Drosophila and mammalian Plexins in transfected cells, raising the possibility that OTK might play a role in either up- or down-regulating Plexin activity or mediating Semaphorin-Plexin signaling. To determine whether this association reflects a true functional interaction, genetic analysis was performed of OTK in Drosophila (Winberg, 2001).
A direct in vivo test of OTK function was aided by the discovery of a P element insertional mutation near the otk gene, designated EP2017. This mutant strain was obtained from the collection of the Berkeley Drosophila Genome Project and was examined for axon guidance defects in homozygous embryos. Indeed, some defects were found, but they were subtle in nature and poorly penetrant. However, the element is located upstream of the coding sequence and may not completely disrupt gene function. Attempts were made to generate complete loss-of-function otk alleles through imprecise excision of the P element (Winberg, 2001).
The EP2017 element is inserted 30 bp upstream of the 5' end of the published otk cDNA. Since the OTK transcript is ~900 bp longer than the cDNA (Pulido, 1992), it is likely that the insert is in the 5' UTR. Ten excision lines were genetically characterized; eight were homozygous lethal and two homozygous viable (the starting strain is semilethal), suggesting that otk is an essential gene. Molecular analysis indicates that the viable strains otk2 and otk8 are precise excisions. In contrast, the lethal strain otk3 carries a 3 kb deletion that extends downstream of the EP2017 element, apparently disrupting otk but not upstream genes. The otk3 lesion removes the putative translational start codon and part of the signal peptide and thus likely represents a complete loss-of-function allele. Subsequent examination of axon guidance defects has shown that otk3 and three other lethal alleles are similar to one another in the variety and severity of their phenotypes, which are more pronounced than those displayed by the original EP2017 strain. In comparison, otk2 is in the range of wild-type (Winberg, 2001).
These reagents allowed for a test of another property of the EP2017 insert. The EP series of P elements contains a UAS gene-regulatory sequence that, in combination with a GAL4 driver, permits transcription of sequences flanking the insertion site of the P element. In the present case, EP2017 is oriented such that GAL4-regulated expression yields short antisense OTK transcripts. In conjunction with elav-GAL4, one copy of EP2017 produces axon guidance abnormalities comparable with homozygous mutant otk1 or otk3 strains, suggesting that this antisense transcription from EP2017 confers a neuron-specific dominant loss-of-function phenotype (Winberg, 2001).
If OTK is important for Plexin A function, then loss-of-function mutations in otk might show guidance phenotypes similar to other mutations in the pathway. Specifically, if OTK is a positive activator or effector of Plexin A, then loss-of-function phenotypes of one should resemble loss-of-function phenotypes of the other. However, if OTK is a negative regulator of Plexin A, then the loss of OTK might lead to similar phenotypes as the overproduction of Plexin A protein. Indeed, embryos mutant for otk display axon guidance defects in the CNS and in the projections of the motor nerves, with abnormalities that are similar to those previously reported for PlexA and Sema1a loss-of-function mutants. The projections of motor neurons to their muscle targets are more obviously affected, disrupted in a way that suggests individual growth cones are not always able to defasciculate from pioneer neurons when they should. The most telling examples are provided by the dorsal projections of the segmental nerve (SN) and the ventrolateral or 'b' branch of the intersegmental nerve (ISNb) (Winberg, 2001).
The major projection of the segmental nerve, the SNa, normally extends along the body wall to a lateral position, where it divides into a dorsal and a lateral branch. The dorsal branch then extends further, dividing again and sending fine projections to innervate a group of transverse muscle fibers. In wild-type late stage 16 embryos, the dorsal SNa thus acquires a characteristic 'pitchfork' appearance. In otk loss-of-function or antisense mutants of the same age, these most dorsal growth cones remain fasciculated together in over 60% of segments and extend as a single thicker branch. This is highly similar to the aberrant SNa morphology displayed in Sema1a and PlexA loss-of-function mutant embryos. In contrast, overexpressing Plexin A causes SNa axons to defasciculate prematurely (Winberg, 2001).
The ISNb normally diverges from the main branch of the ISN in a ventral position, termed 'choice point #1'. Within the ventral muscle domain axons of the ISNb then defasciculate from one another: at choice point #2, a single axon splits off to innervate muscle fibers 6 and 7, and at choice point #3, axons either stop and innervate muscle 13 or extend further to muscle 12. By late embryonic stage 16, these growth cones have typically reached their targets and formed rudimentary synaptic contacts along the edges of these muscle fibers. In otk loss-of-function or antisense mutants, growth cones may fail to defasciculate at any of the three choice points. ISNb axons occasionally fail to exit the ISN at choice point #1, instead bypassing their muscle targets completely or else extending small aberrant projections directly from the main branch of the ISN. More often, choice point #1 is navigated correctly but then axons are unable to defasciculate at choice points #2 or #3, resulting in a thickened, stalled nerve and a failure to innervate one or more of the muscles in this domain (Winberg, 2001).
Within the CNS, additional abnormalities are observed. A subset of longitudinal axons is highlighted by monoclonal antibody labeling; in the wild-type, they form neat parallel tracks. In otk mutant embryos, these tracks are variably wavy and defasciculated and occasionally discontinuous. The incidence of 'broken' axon tracks is greater in the antisense embryos than in the loss-of-function embryos (35% versus 15%) (Winberg, 2001).
The abnormalities seen in the SNa and ISNb of embryos lacking otk are qualitatively and quantitatively highly reminiscent of those described for both Sema1a and PlexA mutants. All of these mutants also show qualitatively similar defects in the major axon tracts within the CNS, but, in the case of otk, these defects are less pronounced. Still, the strong resemblance among the phenotypes of all these mutations suggests that these three genes may all be acting in the same genetic pathway, consistent with the hypothesis that OTK positively influences Plexin A function (Winberg, 2001).
Another way to investigate whether these proteins may work together is to test for dominant genetic interactions. For most proteins, reducing gene dose to a single copy (thus reducing the protein level by 50%) produces mild or undetectable defects. However, reducing the gene dose of two different proteins may generate a phenotype if the two proteins normally function together. This 'transheterozygous' genetic test has been applied to several pairs of proteins that have also been shown to interact biochemically: Notch and Delta, Boss and Sevenless, Sema 1a and Plexin A, and Slit and Robo (Winberg, 2001).
Embryos singly and doubly heterozygous for otk and PlexA were examined and strong phenotypic effects due to the combination were observed. Embryos lacking one copy each of both otk and PlexA exhibit the same variety of SNa and ISNb defects as seen in the single homozygous mutants, to nearly the same degree of severity. This provides strong genetic support for the hypothesis that Otk and Plexin A proteins function positively together through direct contact (Winberg, 2001).
Likewise, embryos doubly heterozygous for otk and Sema1a also show phenotypic enhancement beyond additive effects of the single heterozygotes, supporting the idea of a ternary complex of Sema 1a-Plexin A-OTK proteins. However, the severity of phenotypes in the otk, Sema1a combination is somewhat less than in the others. The discrepancy may reflect a true difference between the association of OTK with Sema 1a compared to Plexin A. Alternatively, it may arise from differences in the normal expression levels of the various proteins: if Plexin A were the least abundant component under normal circumstances, then reducing the levels of the other two would be less consequential in this test (Winberg, 2001).
It has been supposed that OTK somehow affects the ability of Plexin A to mediate Sema 1a signaling. However, because all three proteins are expressed by many of the same neurons, the genetic tests above are also consistent with the possibility that OTK may interact directly with Sema 1a in cis. To verify that OTK can act genetically downstream of the signal, use was made of the GAL4 system to misexpress Sema 1a in muscles, thus offering an excess of repulsive target-derived ligand. Ectopic presentation of Sema 1a on specific muscles using UAS-Sema1a and H94-GAL4 turns these muscles into nonpermissive substrates and prevents motoneurons from innervating them correctly. The abnormal innervation of muscle 13 increases from 22% (with H94-GAL4 driver alone) to 49% (with addition of UAS-Sema1a) in this Sema1a gain-of-function experiment. This phenotype is suppressed by removing one copy of otk, reducing neuronal expression levels. Abnormal innervation of muscle 13 is reduced to 26%. It has been shown that the addition of Sema1a increases the percent abnormal from 19% to 53% and removing a single copy of PlexA reduces this frequency of abnormal innervation to 21%. Thus, removal of one copy of otk is nearly as effective in reducing the Sema1a gain-of-function as is removal of one copy of PlexA. Since neuronal OTK is sensitive to muscle-derived Sema 1a, this experiment confirms that OTK is able to act downstream of Sema 1a (Winberg, 2001).
This study has shown that Otk, a transmembrane protein of about 160 kDa, with homology to receptor tyrosine kinases, both associates with Plexins in vitro and appears to function in a Semaphorin-Plexin signaling pathway in vivo to control certain aspects of axon guidance. Biochemical data show that OTK specifically associates with Plexins in vitro. Genetic disruption of otk leads to specific defects resembling those due to lesions in either Sema 1a, a transmembrane Semaphorin that mediates axon defasciculation. These data suggest that all three proteins -- Sema 1a, Plexin A, and OTK -- may function in the same pathway. Genetic interactions suggest that OTK and Plexin A act downstream of Sema 1a. Thus, it appears that OTK and Plexin A can associate as components of a receptor complex that mediates the repulsive signaling in response to Semaphorin ligands (Winberg, 2001).
It is not known whether OTK and Plexins normally associate in vivo in growth cones or whether they might only be brought together by ligand binding. In the absence of ligand in vitro, a tight association is found between the two transmembrane proteins. If transmembrane Semaphorins, like their secreted relatives, function as dimers, then binding of Sema 1a to Plexin A might provide a mechanism for clustering receptor complexes, which by analogy might activate one or more associated kinases and lead to the phosphorylation of Plexin and OTK. Testing such speculations will have to await an appropriate system for testing ligand activation (Winberg, 2001).
Interestingly, despite its homology with receptor tyrosine kinases and the observation that immunoprecipitates of Drosophila OTK possess tyrosine kinase activity (Pulido, 1992), OTK itself is probably not an active tyrosine kinase. The OTK sequence suggests that it belongs to a family of kinase 'dead' receptors. The catalytic domain of OTK, like other members of this family, is altered in a few key conserved residues that are implicated in autophosphorylation (the conserved DFG motif substituted by YPA). Vertebrate family members bear similar alterations in the DFG motif and apparently do not have kinase activity. Modest tyrosine phosphorylation of OTK has been observed in 293T cells but no significant increase in Plexin phosphorylation has been observed upon coexpression with OTK. Thus, OTK either possesses a weak catalytic activity, which is barely detectable in the tested experimental conditions, or like other members of the CCK-4 subfamily of receptor tyrosine kinases, OTK might be kinase dead. In the latter case, some other active kinase would be expected to be present in or recruited to the OTK/Plexin complex in order to account for the observed tyrosine phosphorylation of these proteins. This situation is reminiscent of the interleukin receptors, which are heterodimers composed of a ligand binding subunit and a signal transducing subunit known as gp130. Neither subunit possesses a catalytic activity; rather, gp130 associates with the Janus kinases. Upon ligand binding, the receptors multimerize, resulting in activation of the Janus kinases and tyrosine phosphorylation of the receptor (Winberg, 2001).
Another receptor tyrosine kinase carrying mutations in conserved DFG catalytic residues, h-Ryk/d-Derailed, appears also to be kinase inactive. Nevertheless, Ryk/Derailed is crucially involved in axon guidance. Thus, at least two highly conserved receptor tyrosine kinases, both of which are members of families which are kinase dead -- OTK and Derailed -- have been shown to function in axon guidance. In the case of OTK, it functions apparently by associating with Plexins and helps to mediate their output (Winberg, 2001).
The signal transduction pathway activated by Semaphorins is beginning to be clarified. The cytoplasmic domains of Plexins do not have any obvious signal transduction motif such as a kinase or phosphatase domain. However, the cytoplasmic domains of Plexin B receptors bind directly to the Rac GTPase in a GTP-dependent manner. It has been confirmed that the cytoplasmic domain of Plexin B (PlexB) indeed binds directly to the active, GTP-bound form of the Rac GTPase and, in addition, that a different region of PlexB binds to RhoA. The genetic and biochemical evidence suggests a model whereby PlexB mediates repulsion in part by coordinately regulating two small GTPases in opposite directions: PlexB binds to RacGTP and downregulates its output by blocking its access to PAK and, at the same time, binds to and increases the output of RhoA. While the contribution of OTK to this signaling pathway has not yet been investigated, by analogy with other tyrosine-phosphorylated receptor complexes, one hypothesis to test is that a Rho exchange factor is recruited to the activated Plex/OTK complex, providing local activation of Rho (Winberg, 2001).
Prior to the identification of Plexins as Semaphorin receptors and the implication of OTK as a Plexin-associated kinase, both proteins were shown to be capable of mediating cell aggregation in vitro. These studies led to the suggestion that both Plexins and OTK might function as homophilic cell adhesion molecules. Whether either or both of them normally functions in a homophilic fashion in vivo is unknown (Winberg, 2001).
Semaphorins have come to be considered as being ligands and Plexins as their receptors. But their roles in axon guidance may not be this simple. On the one hand, some Semaphorins are transmembrane proteins with cytoplasmic domains that appear as if they might be capable of transducing signals. Thus, some Semaphorins might themselves be receptors as well as ligands. On the other hand, Plexins, which are related to Semaphorins and have extracellular Semaphorin domains, can bind to themselves. Thus, some Plexins might be both ligands and receptors. Finally, Plexins associate with OTK, which also can bind homophilically (Winberg, 2001).
The data presented it this study demonstrate a role for OTK downstream from a Semaphorin on the receiving side of a signaling event. The best evidence for this conclusion is the genetic suppression data. Removing one copy of otk suppresses a Sema 1a gain-of-function phenotype. The most parsimonious interpretation of this result is that OTK functions downstream of Sema 1a. It is not known to what degree OTK binding and function is ligand gated. Moreover, it is not known whether OTK responds directly to Semaphorins, to some other ligand, or alternatively whether it simply binds to Plexins as part of a Semaphorin signaling complex. It will be interesting in the future to determine how these different Semaphorin, Plexin, and OTK proteins associate, modulate Semaphorin-mediated signal transduction, and thus control axon guidance (Winberg, 2001).
Home page: The Interactive Fly © 1997 Thomas B. Brody, Ph.D.
The Interactive Fly resides on the
Society for Developmental Biology's Web server.