BIOLOGICAL OVERVIEW

The netrins are a conserved family of secreted molecules that guide neuronal growth cones in the developing nervous system. They are bifunctional, capable of attracting some axons while repelling others. Netrins act through receptors belonging to two distinct families: the Deleted in Colorectal Cancer (DCC) and UNC5 families. DCC family receptors include DCC in vertebrates and Frazzled in Drosophila. C. elegans UNC-5 and its three vertebrate homologs UNC5H1, UNC5H2, and UNC5H3/RCM are also single-pass transmembrane receptors. Their extracellular domains consist of two Ig domains and two thrombospondin type I (TSP) domains. The cytoplasmic domains of UNC5 receptors include at least three conserved regions: a ZU5 domain, a DB motif, and a carboxy-terminal death domain (DD). The predicted Drosophila Unc5 protein has the same domain organization as its worm and mammalian orthologs. In Drosophila, Netrins are expressed by midline cells of the CNS and by specific muscles in the periphery. Netrins attract commissural and motor axons expressing the DCC family receptor Frazzled. Drosophila Unc5 receptor is a repulsive Netrin receptor likely to contribute to motor axon guidance. Ectopic expression of Unc5 on CNS axons can elicit either short- or long-range repulsion from the midline. Both short- and long-range repulsion require Netrin function, but only long-range repulsion requires Frazzled (Keleman, 2001).

In C. elegans, UNC-40 (Frazzled homolog) is required primarily for ventral migrations directed toward UNC-6 (Netrin homolog), but also participates in many dorsal migrations away from UNC-6. UNC-5, in contrast, is required only for dorsal migrations away from UNC-6 (Hedgecock, 1990; Hamelin, 1993). Similarly, in vertebrates, DCC is required for the ventral growth of mammalian commissural axons in vivo (Fazeli, 1997), and the attractive Netrin response of isolated Xenopus spinal axons in vitro (Ming, 1997). Expression of UNC-5 family receptors in cultured Xenopus spinal axons switches their response to Netrin from attraction to repulsion (Hong, 1999). Like attraction, repulsion of Xenopus spinal axons also requires DCC function. Together, these data have led to a model in which DCC receptors mediate attraction, and in some cases also repulsion, while UNC5 receptors act only in repulsion (Keleman, 2001 and references therein).

The Xenopus growth cone turning assay has been exploited to investigate the molecular mechanisms of DCC and UNC5 function (Hong, 1999; Stein, 2001). Their work has shown that attraction to Netrin involves the multimerization of DCC receptors, mediated by the association of their cytoplasmic P3 domains (Stein, 2001). Repulsion, in contrast, depends on an interaction between the cytoplasmic domains of DCC and UNC5 receptors. In this case, the association is mediated by the P1 domain of DCC, which binds the DB domain of UNC5 (Hong, 1999). In both attraction and repulsion, binding of Netrin to the extracellular domain of its receptor triggers the interaction between the cytoplasmic domains (Keleman, 2001 and references therein).

A further important conclusion from these studies is that it is the cytoplasmic domain of UNC5 that specifies repulsion. This comes from the finding that a DCC-UNC5H2 chimeric receptor, in which the DCC extracellular domain is fused to the UNC5H2 cytoplasmic domain, can elicit a repulsive Netrin response that is just as strong as that induced by UNC5H2 itself (Hong, 1999). In fact, the extracellular domain of UNC5H2 is not even required: a DCC/UNC5H2 receptor complex can still form and mediate repulsion, even if one of the two proteins lacks its entire extracellular domain. When examining which parts of the UNC5H2 cytoplasmic domain are required for repulsion (Hong, 1999), the DB domain was found to be essential; the death domain was not (Keleman, 2001 and references therein).

Evidence that Drosophila Unc5 is a repulsive Netrin receptor comes from two complementary sets of observations: (1) repulsion (mediated by ectopic NetrinB) of motor axons that exit the CNS ipsilaterally via the segmental nerve root (SNa motor axons) requires Unc5 function, and conversely, (2) midline repulsion of either commissural or intersegmental interneuron-generated Ap axons forced to express Unc5 requires Netrin function (Keleman, 2001).

Within the CNS, Netrins are expressed primarily at the midline, where they act through the DCC family receptor Frazzled to help guide commissural axons toward and across the midline. If Unc5 is indeed a repulsive Netrin receptor, then misexpression of Unc5 on these commissural axons should prevent them from crossing the midline, possibly even redirecting them laterally away from the midline and out of the CNS. To test this, Unc5 was expressed in all postmitotic neurons using a UAS-Unc5 transgene and an elav-GAL4 driver. Commissures are completely lacking in such embryos. The longitudinal connectives are also more widely separated than in wild-type embryos. However, there is no detectable increase in the number of axons that extend laterally away from the midline. Pan-neural expression of Unc5 thus results in a short-range repulsive response in which commissural axons are prevented from crossing the midline, but are not driven directly away from it (Keleman, 2001).

This 'commissureless' phenotype is much stronger than the partial loss or thinning of commissures observed in either Netrin or frazzled mutant embryos, and therefore cannot simply be explained by the ability of Unc5 to sequester Netrin or Frazzled proteins into inactive complexes. Unc5 also appears to be far more potent in preventing midline crossing than the Robo receptors for the midline repellent Slit. Nine independent insertions of the UAS-Unc5 transgene were tested, and all of them give a completely penetrant commissureless phenotype with just single copies of both transgenes. In contrast, robo gain-of-function phenotypes generally require at least two copies of either or both of the UAS-robo and elav-GAL4 transgenes (Keleman, 2001).

The strong commissureless phenotype of these pan-neural Unc5 embryos provides the opportunity to examine the structural requirements for Unc5-mediated repulsion. To this end, a series of UAS constructs encoding mutant Unc5 proteins were generated, each lacking one of the conserved extracellular or intracellular domains, as well as a construct encoding an N-terminal myristoylated form of the Unc5 cytoplasmic domain. None of these mutant Unc5 transgenes, when expressed using the elav-GAL4 driver, resulted in even a weak commissureless phenotype. At least two different insertions of each transgene were tested. Expression and axonal localization of the mutant proteins were confirmed by Western blotting and in situ detection using antibodies against either an N-terminal HA epitope (for the domain deletions) or a C-terminal c-myc epitope (for the myristoylated cytoplasmic domain) (Keleman, 2001).

These results confirm the conclusions from Xenopus in vitro studies (Hong, 1999) that the DB domain is required for Unc5 repulsion, and extend them by showing a requirement also for the ZU5 domain and each of the four extracellular domains. However, the finding of a strict requirement for both the death domain and the extracellular domain of Drosophila Unc5 contrasts with the dispensibility of the corresponding domains of UNC5H2 in the Xenopus assays (Hong, 1999). Deletions were made at analogous positions of the UNC5 proteins in both sets of experiments. It therefore seems unlikely, but cannot be excluded, that the failure of these mutant Drosophila Unc5 proteins to mediate repulsion is just a trivial consequence of protein misfolding or instability. Rather, it suggests that the repulsion of commissural axons by Unc5 in Drosophila may occur through a mechanism distinct from the repulsion of spinal axons by UNC5H2 in the Xenopus assays (Keleman, 2001).

Growth cone repulsion mediated by ectopic UNC5 expression, both in C. elegans and in the Xenopus in vitro assays, has been shown to require not only Netrin function but also that of DCC family receptors (Hamelin, 1993; Colavita, 1998; Hong, 1999). It was therefore anticipated that both Netrin and frazzled function would be required to prevent midline crossing in elav-GAL4/UAS-Unc5 embryos. To test this, embryos carrying both transgenes were generated in the background of either the Netrin deficiency Df(1)NP5 or the frazzled null allelic combination fra³/fra⁴. As expected, midline repulsion by Unc5 does indeed require Netrin function. Surprisingly, however, it is independent of frazzled function. The elav-GAL4/UAS-Unc5 phenotype is just as strong in the frazzled mutant background as in the wild-type background (Keleman, 2001).

Given that Slit can bind directly to Netrin, and can also act via Robo receptors to silence Netrin attraction, might midline repulsion by Unc5 depend in any way on repulsion mediated by Slit and its Robo receptors? To test this, embryos were generated carrying both the elav-GAL4 and UAS-Unc5 transgenes, and that also were homozygous for one or more of the null alleles slit², robo¹, and robo2⁴ (Keleman, 2001).

The commissureless phenotype of pan-neural Unc5 embryos is essentially unaltered in the robo and robo2 single mutant backgrounds. The phenotype is more difficult to interpret when either slit or both robo and robo2 function is eliminated. The CNS phenotype observed in these embryos is intermediate between that of pan-neural Unc5 embryos and either slit or robo robo2 embryos. In some segments, axons are entirely collapsed at the midline as in slit or robo robo2 mutants, but in other segments axons are separated into two bundles, one on each side of the midline. This argues against a direct role for Slit in Unc5-mediated repulsion, since clearly Unc5 misexpression does have an effect in the absence of Slit. It does, however, suggest an indirect role. For example, repulsion by Netrin and Unc5 may only be effective in keeping axons away from the midline when it is added on top of the repulsive signal transduced via Slit and its Robo receptors. Another, not exclusive, possibility is that this intermediate phenotype is due to the ventral displacement of midline cells that occurs in both slit and robo robo2 mutants (Keleman, 2001).

Several recent experiments point to the modular design of axon guidance receptors, in which the extracellular domain determines the ligand specificity while the cytoplasmic domain dictates the response of the growth cone. In particular, Hong (1999) has demonstrated that a DCC-UNC5H2 chimeric receptor consisting of the extracellular domain of DCC and the cytoplasmic domain of UNC5H2 is as effective as wild-type UNC5H2 in repelling Xenopus spinal axons away from a Netrin source in vitro. This finding was tested in vivo. In addition, attempts were made to extend this result by testing the prediction that a reciprocal UNC5-DCC chimera should mediate attraction to Netrin (Keleman, 2001).

UAS transgenes were prepared encoding chimeric Fra-Unc5 and Unc5-Fra receptors, in which the cytoplasmic domains of the two Netrin receptors had been swapped immediately proximal to their transmembrane domains. To test the prediction that the cytoplasmic domain of Unc5 specifies repulsion, the CNS of embryos in which one or another of these chimeras was expressed using the elav-GAL4 driver was examined. As expected, pan-neural expression of the Fra-Unc5 chimera results in a commissureless phenotype just as strong as that observed with the full-length Unc5 receptor. Ectopic expression of Unc5-Fra has no obvious effect, as previously found to be the case also for full-length Fra (Keleman, 2001).

Does the Unc5-Fra chimera act as an attractive Netrin receptor? If so, pan-neural expression of this receptor, like that of Fra itself, should at least partially rescue the frazzled mutant phenotype. This is indeed the case. Each of two UAS-Unc5-fra transgene insertions tested almost completely rescue the frazzled null mutant. UAS-Unc5-fra rescues both the commissural and longitudinal axon defects of frazzled mutants just as efficiently as does UAS-fra. It is therefore concluded that Unc5-Fra is an attractive Netrin receptor, formally completing the demonstration that Netrin receptors are modular: the growth cone response (attraction or repulsion) is determined by the cytoplasmic domain (DCC or UNC5, respectively), irrespective of the Netrin binding extracellular domain to which it is attached (Keleman, 2001).

Experiments using the elav-GAL4 driver show that Unc5 is a potent mediator of Netrin repulsion at short range, preventing commissural axons from crossing the midline. In a final set of experiments, these observations were extended by asking how an ipsilateral interneuron -- one that does not normally cross the midline -- would respond to ectopic expression of Unc5. For this, the Ap-GAL4 driver was used. This line expresses GAL4 in three neurons (termed the Ap neurons) in each hemisegment. Their cell bodies are positioned laterally within the nerve cord, several cell diameters from the midline. One is located dorsally, the other two ventrally. All three are intersegmental interneurons. Their axons first grow toward the midline, but they do not cross it, instead turning anteriorly to continue along the medial edge of the ipsilateral longitudinal tract. In the experiments reported here, focus was placed on the behavior of the dorsal Ap neuron (Keleman, 2001).

Expression of Unc5 in this neuron has remarkable consequences. Rather than growing toward the midline, its axon now grows laterally away from the midline to exit the CNS and continue on a motor trajectory into the periphery. This phenotype is highly penetrant: 91% of dorsal Ap axons examined exited the CNS in these embryos. Thus, Unc5 can repel axons away from the midline at long range, forcing them 180° off course. All of the mutant Unc5 proteins tested in the midline crossing assay were also found to be defective in this assay (Keleman, 2001).

This long-range repulsion by Unc5 requires Netrin function, as expected. However, unlike the short-range repulsion of commissural axons at the midline, long-range repulsion of Ap axons is partially dependent on frazzled function. In frazzled mutant embryos, only 59% of Ap axons exited the CNS upon ectopic Unc5 expression. To determine whether this reflects an autonomous requirement for frazzled, its function was restored specifically in the Ap neurons by introducing a UAS-fra transgene into these embryos. The percentage of Ap axons exiting the CNS rose to 97%, demonstrating that potent long-range repulsion of Ap axons requires expression of both Unc5 and Fra in the Ap neurons themselves (Keleman, 2001).

Finally, it was asked whether Slit and its receptor Robo contribute to the long-range repulsion of Ap axons mediated by Unc5. Of the three Drosophila Robo receptors, Ap neurons express only Robo, the founding member of the Robo family. Slit repulsion could therefore be eliminated simply by removing robo function, thus avoiding the problem of the midline collapse that complicates the interpretation of Slit function in the midline crossing assay. In a robo null mutant background, Unc5 expression drives Ap axons out of the CNS just as effectively as it does in a wild-type background. Long-range repulsion mediated by Netrin and Unc5 thus operates independently of repulsion mediated by Slit and Robo (Keleman, 2001).

At this point, it is not clear whether just one or both of the Drosophila Netrins, NetA and NetB, are ligands for Unc5. Clearly, NetB alone is capable of repelling Unc5-expressing SNa motor axons (Mitchell, 1996; Winberg, 1998), while ectopic expression of NetA appears to have no effect on these axons (Winberg, 1998). Thus, while NetA and NetB are both attractants, only NetB may also be a repellent. It will be interesting to determine whether these functional interactions are reflected in the ligand specificity of the Drosophila Netrin receptors, and if so, whether such specificity depends on the unique insert domains found in each of the NetA, NetB, and Unc5 proteins (Keleman, 2001).

The expression patterns of Unc5 and the two Netrins within the developing nervous system are strikingly complementary. SN motor axons that express Unc5 avoid both midline cells and peripheral muscles that express Netrins, while commissural axons that cross the midline and motor axons that innervate Netrin-expressing muscles do not express Unc5. These observations hint at an important role for Netrin repulsion in guiding SN motor axons, first out of the CNS, and then in their choice of specific muscle targets in the periphery. However, in both Netrin-deficient and Unc5 RNAi embryos, SN motor axons still exit the CNS and choose their correct targets. Thus, repulsion by Netrins is likely to be just one of multiple guidance cues that control these pathways (Keleman, 2001).

One obvious candidate for a second repellent that drives motor axons away from the midline is Slit. However, even in embryos lacking both slit and Netrin function, many motor axons still exit the CNS. Evidently, at least one additional midline repellent for Drosophila motor axons remains to be discovered. Multiple cues also appear to repel motor axons in the vertebrate CNS. For example, trochlear motor neurons express Unc5h1 and their axons are repelled by netrin-1 in vitro, but they still project normally in netrin-1-deficient mice (Keleman, 2001).

In the periphery, a host of molecules have been identified that help direct individual motor axons to their specific muscle targets. In addition to NetA and NetB, these include Semaphorins, such as Sema2a, and Ig family cell adhesion molecules, such as FasII. Genetic analysis suggests that muscle target selection depends on a complex set of cooperative and antagonistic interactions between these and other guidance cues. In this context, the loss of just a single guidance cue generally results in only a partially penetrant targeting defect. It is therefore perhaps not so surprising to find that SN motor axons still appear to project normally when either Netrin or Unc5 function is impaired. Detecting a more subtle requirement for Netrin repulsion in muscle targeting will require the generation of loss-of-function mutations in Unc5, and the examination of embryos in which motor axons are defective in their responses to multiple cues, not just Netrins (Keleman, 2001).

A DCC-UNC5H2 chimera has been shown to act as a repulsive Netrin receptor in vitro. Likewise a Fra-Robo receptor also behaves as a repulsive Netrin receptor in vivo, while the reciprocal Robo-Fra receptor functions as an attractive Slit receptor. Together, these two studies have led to the general notion that axon guidance receptors are modular in their design -- the extracellular domain determines the ligand specificity while the cytoplasmic domain specifies the nature of the response (attraction or repulsion) (Keleman, 2001).

This idea has been tested in vivo, specifically for the attractive and repulsive Netrin receptors, Fra and Unc5. As predicted by this model, a Fra-Unc5 chimera acts as a repulsive receptor and Unc5-Fra as an attractive Netrin receptor. Both chimeras prove to be as potent as the wild-type proteins in mediating these responses. These data thus provide a clear in vivo demonstration of the modular design of the two Netrin receptors (Keleman, 2001).

Netrins can act over either a short or a long distance to repel CNS axons forced to express Unc5. Pan-neural expression of Unc5 leads to a short-range response, in which commissural axons are prevented from crossing the midline, but do not grow away from it. Expression of Unc5 in a single lateral interneuron, the dorsal Ap neuron, elicits a long-range response. Rather than growing toward the midline as it normally does, the Ap axon instead grows directly away from the midline and out of the CNS (Keleman, 2001).

Why doesn't pan-neural expression of Unc5 lead to the same long-range response, resulting in a massive exodus of CNS axons? The reason for this is not entirely clear. It may be that only very few neurons are capable of such a dramatic response, and the GAL4 line used to express Unc5 happens to be expressed in one of these. However, the only other GAL4 driver tested that is expressed in a subset of interneurons (eg-GAL4) also produces a long-range response, so it seems that this may indeed be a more general phenomenon. Another possibility is that the pan-neural and subset-specific drivers differ in the timing and/or levels of transgenic expression they provide. While this is certainly the case, it seems unlikely to account for the different responses. Introducing a second copy of the UAS-Unc5 transgene does not result in a detectable increase in the number of axons exiting the CNS with elav-GAL4, nor does the use of another, earlier, pan-neural driver, 1407-GAL4. Similarly, an extra copy of the UAS-Unc5 transgene does not alter the long-range response of Ap axons (Keleman, 2001).

It is suspected that the essential difference in these two sets of experiments is the distribution of free Netrin ligand. Normally, Netrin is likely to diffuse far from the midline, and could thus act as a long-range repellent for SN motor axons that endogenously express Unc5, or Ap axons that are forced to express Unc5. But when Unc5 is ectopically expressed at high levels on all CNS axons, Netrin diffusion may be more limited. In this case, axons might only encounter free Netrin ligand close to the midline, resulting in a short-range response (Keleman, 2001).

Whatever the reason for this difference, it has enabled the establishment of in vivo assays that clearly distinguish between short- and long-range repulsion by Netrins. This is important, as it allows the resolution of a long-standing uncertainty concerning the role of DCC proteins in UNC5-mediated repulsion. The origin of this debate goes back to the initial report of Hedgecock (1990), showing that both C. elegans unc-5 and unc-40 are required for dorsal migrations, but unc-40 to a lesser extent than unc-5. Subsequently, expression of unc-5 in lateral mechanosensory neurons was found to redirect their axons dorsally (Hamelin, 1993), and this response shows a strong requirement for unc-40 function (Colavita, 1998). unc-40 has been shown to act autonomously to mediate ventral migrations, but an autonomous role for unc-40 in dorsal migrations has not yet been demonstrated. This is a critical issue, as experiments in Drosophila have demonstrated nonautonomous guidance functions for the unc-40 homolog frazzled. It thus remains unclear what role UNC-40 actually plays in Netrin/UNC-6 repulsion in C. elegans. Further experiments in Drosophila and vertebrates have not yet clarified the situation. In Drosophila, repulsion of SNa motor axons by ectopic NetB, which has been shown to be mediated by Unc5, does not require Frazzled (Winberg, 1998). In contrast, UNC5 receptors (Hong, 1999) do require DCC function to repel Xenopus spinal axons away from a Netrin source in vitro (Keleman, 2001).

In attempting to draw a conclusion from this disparate set of results, several authors have speculated that UNC5 receptors might be able to act alone at high Netrin concentrations, but require DCC coreceptors at lower concentrations. Until now, this idea has not been directly tested. This study offers strong evidence in support of such a model: short-range repulsion of commissural axons by Netrin and Unc5 does not require Frazzled. In contrast, Frazzled is needed for potent long-range repulsion, and this is an autonomous requirement (Keleman, 2001).

Does UNC5 act alone at short range, or might it require yet another coreceptor? This remains to be addressed in future studies. It is, however, interesting to note that short-range repulsion requires the DB (DCC binding) domain of Unc5, but does not require Frazzled. The requirement for the DB domain of Unc5 thus cannot be due to its role in binding the P1 domain of Frazzled, but may instead indicate a role in binding other proteins, perhaps even another coreceptor. Another intriguing difference between UNC5-mediated repulsion in Drosophila assays and in the Xenopus assays of Hong (1999) is that the Drosophila assays show a strict requirement for the Unc5 death domain. This domain was required in both the long- and short-range assays used in this study, but not for the repulsive Netrin response of Xenopus spinal axons in vitro. Perhaps at very low Netrin concentrations, as encountered by Xenopus axons in the in vitro turning assays, UNC5 repulsion does not require the death domain and is mediated entirely through its interaction with DCC, whereas at the high Netrin concentrations likely to be encounted by Drosophila commissural axons in vivo, UNC5 repulsion requires the death domain but not DCC. At intermediate Netrin concentrations, which may operate in long-range assays, UNC5 may require both the death domain and, partially, DCC receptors. There may thus be an inverse relationship between the requirement for either DCC as a coreceptor or for the death domain of UNC5, with the relative requirements depending on the concentration of ligand. In light of this, it is interesting to note that, in several other receptors, death domains mediate self-association. The hypothesis that this may also be the case for UNC5 receptors is currently being tested. This is an attractive idea, as it would imply an appealing symmetry in the signaling modes of Netrin receptors: attraction involves the formation of DCC homodimers (mediated by their C-terminal P3 domains); short-range repulsion would involve the formation of UNC5 homodimers (mediated by their C-terminal death domains), and long-range repulsion the formation of DCC/UNC5 heterodimers (mediated by their P1 and DB domains, respectively) (Keleman, 2001).

GENE STRUCTURE

cDNA clone length - 5011

Bases in 5' UTR - 647

Exons - 8

Bases in 3' UTR - 1167

PROTEIN STRUCTURE

Amino Acids - 1072

Structural Domains

Drosophila Unc5 was initially identified by searching the EST database for sequences similar to C. elegans unc-5. One EST clone predicted to encode a protein with significant homology to UNC-5 was identified, and used to isolate full-length cDNAs from embryonic and larval-pupal cDNA libraries. These cDNAs encode a 1072 amino acid transmembrane protein most closely related in structure and sequence to C. elegans UNC-5 and its mammalian homologs. Examination of the Drosophila euchromatic genome sequence reveals that this is likely to be the only member of the unc-5 family encoded in the Drosophila genome (Keleman, 2001).

The predicted Drosophila Unc5 protein has the same domain organization as its worm and mammalian orthologs, to which it has 24%-29% amino acid identity over its entire length. The two TSP domains are the most highly conserved regions in the UNC5 proteins, with 37%-55% amino acid identity between the TSP domains of UNC5 proteins from different species. Curiously, although the TSP domains of Drosophila Unc5 also show the highest level of sequence similarity to other UNC5s, they are also unique in that they contain two small insertions: a 49 amino acid insertion within the second TSP domain and a 17 amino acid insertion between the two TSP domains. The significance of these insertions is unknown, but it is interesting to note that the two Drosophila Netrins also contain small insertions that are not found in the Netrins of other (Keleman, 2001).

date revised: 20 December 2001

Home page: The Interactive Fly © 1995, 1996 Thomas B. Brody, Ph.D.

The Interactive Fly resides on the
Society for Developmental Biology's Web server.