Glial cell adhesive molecule unzipped mediates axon guidance in Drosophila



Axon guidance needs help from the glial cell system during embryogenesis. In the Drosophila embryonic central nervous system (CNS), longitudinal glia (LG) have been implicated in axon guidance but the mechanism remains unclear. We identified the protein encoded by the Drosophila gene unzipped (uzip) as a novel cell adhesion molecule (CAM). Uzip expressed in Drosophila S2 cells triggered cell aggregation through homophilic binding. In the embryonic CNS, Uzip was mainly produced by the LG but was also located at axons, which is consistent with the secretion of Uzip expressed in cultured cells. Although uzip mutants displayed no axonal defect, loss of uzip enhanced the axonal defects in the mutant of N-cadherin (CadN) and the Wnt gene family member wnt5. Overexpression of uzip could rescue the phenotype in the CadNuzipD43 mutant. Thus, Uzip is a novel CAM from the LG regulating axon guidance. Developmental Dynamics, 2011. © 2010 Wiley-Liss, Inc.


During neural development, the glial cell system is required for axon guidance (Chotard and Salecker,2004). Glial ablation in the vertebrate peripheral nervous system (PNS; Gilmour et al.,2002) and in the Drosophila brain (Spindler et al.,2009) leads to abnormalities in axonal fasciculation and growth trajectories. Glia can act as guidepost cells that locate at choice points and provide guidance cues to which growth cones of axons respond. Thereby, axons are guided toward their destinations. For example, in the Drosophila embryonic ventral nerve cord (VNC)—equivalent to the vertebrate spinal cord—the midline glia are regarded as guidepost cells that control the midline crossing of commissural axons by secreting the attractant proteins Netrins and the repellent Slit (Chotard and Salecker,2004).

Another set of guidepost cells in the Drosophila VNC is the longitudinal glia (LG; Jacobs,1993; Hidalgo et al.,1995; Hidalgo and Booth,2000; Takizawa and Hotta,2001). LG are located above the longitudinal connectives of the rail-like VNC axonal trajectory and protrude sheath-like membrane structures that enwrap these connectives (Ito et al.,1995). LG are required for axonal growth, because targeted ablation of the LG results in defective axonal trajectories (Hidalgo et al.,1995; Hidalgo and Booth,2000). Some LG-produced proteins participate in regulating axon growth, such as the secreted protein Folded Gastrulation, the transcription factor Jing, and the chromatin remodeling protein DATR-X (Sun et al.,2006; Ratnaparkhi and Zinn,2007). However, because the known molecules and mechanisms are limited, how LG regulate axon growth remains elusive.

Cell adhesion molecules (CAMs) function to bind cells together. When expressed in cultured cells, CAMs trigger the formation of cell aggregates. CAMs are anchored to the plasma membrane through transmembrane domains or a glycosylphosphatidyl inositol (GPI) moiety. Some CAMs exhibit secreted forms. CAMs have been shown to function in axon guidance (Goodman,1996). However, the loss of a single CAM usually does not result in severe axonal defects. For example, the Drosophila N-cadherin (CadN) is the major cadherin in the embryonic CNS and is located on axons, yet knockout of its gene CadN does not cause global defects in axonal patterning (Iwai et al.,1997). Studies indicate that the simultaneous loss of two CAMs, or the loss of one CAM and a signaling molecule, synergistically enhances the loss-of-function axonal phenotype of the single mutants, suggesting that CAMs and signaling molecules function cooperatively in regulating axon guidance (Elkins et al.,1990; Speicher et al.,1998).

Here, we found that the protein encoded by the Drosophila gene unzipped (uzip) is a CAM. When expressed by Drosophila S2 cells, Uzip triggered aggregate formation through homophilic binding. We detected both GPI-anchored and secreted Uzip protein in cultured cells. Immunostaining showed it to be present in the LG and in longitudinal connectives. Genetic analyses and tissue-specific knockdown experiments indicated that Uzip is mainly produced by LG. Loss of uzip alone did not affect development of LG and axons, but it enhanced the axonal defects of CadN and wnt5 mutants. Overexpression of uzip could rescue the axonal phenotypes in the CadN uzipD43 mutant. Our data support the idea that Uzip is required for the development of some axons in Drosophila.


Overexpression of uzip in Peripheral Neurons Promotes Clustering of Neurons

We performed misexpression screening using a batch of EP P-element lines to cross with the elav-GAL4, UAS-mCD8::GFP (elav>mCD8::GFP) strain. In this fly, the membrane-bound green fluorescent protein (GFP) driven by elav-GAL4 is expressed in all neurons and labels neuronal positions and dendritic morphology (Lee et al.,2000). Each dorsal abdominal segment has 12 neurons in the PNS (Fig. 1A,B). Of interest, these neurons were clustered together in the progeny when the elav>mCD8::GFP fly was crossed with EP-C04101 (Fig. 1C,D). The genomic DNA adjacent to the P-element of EP-C04101 was cloned by plasmid rescue and sequenced. The P-element is located at 30 base pairs (bp) downstream of the transcription start site of the gene, uzip (Fig. 2A). Because the uzip coding sequence is located downstream of the P-element of EP-C04101, which bears a UAS sequence (Rorth,1996), overexpression of uzip might occur through the transactivation of GAL4 in the elav>mCD8::GFP; EP-C04101 strain. Western blotting and in situ hybridization of embryos confirmed this hypothesis (Supp. Fig. S1, which is available online). Therefore, overexpression of uzip in neurons results in neuronal clustering.

Figure 1.

uzip overexpression results in neuronal clustering. A–D: Immunostaining of dorsal abdominal peripheral nervous system (PNS) neurons in stage 17 embryos. A,C: Neuronal morphology was recognized by green fluorescent protein (GFP). B,D: Neuronal nuclei were stained by the anti-Elav antibody. Neurons indicated by brackets were clustered in C and D, but not in A and B. Genotypes are indicated.

Figure 2.

The genomic locus of uzip and the Uzip protein domains. A: The uzip transcript consists of five exons and the coding region of uzip is filled in. The insertion sites of the P-elements of EPC04-101, f01534, and f02444, which were used for generating uzip mutants, are shown. B: The predicted domains and modification sites in Uzip are indicated.

The uzip mutants described previously exhibited defects of embryonic dorsal closure and axonal development (Zhao et al.,1988). However, a later study proved that these defects were caused by mutation of the adjacent gene, zipper (Young et al.,1993). Thus, the biological function of uzip remains unknown.

Uzip Triggers Cell Adhesion by Homophilic Binding

The concurrence of uzip overexpression and neuronal clustering in the PNS neurons suggests that Uzip might be able to trigger cell adhesion. To verify this hypothesis, we performed aggregation assays using Drosophila S2 cells expressing GFP or Uzip tagged with His at the C-terminus (Uzip-His). We found that Uzip-, but not GFP-expressing cells were able to form aggregates (Fig. 3A,B,E). CAMs can promote cell-to-cell adhesion by means of homophilic or heterophilic interactions. To clarify which interaction occurred, we carried out an aggregation assay in which 1,1′, di-octadecyl-3,3,3′,3′,-tetramethylindo-carbocyanine perchlorate (DiI) -labeled nontransfected cells were mixed with unlabeled Uzip-transfected cells. If Uzip mediates homophilic cell-to-cell adhesion, all cells in a cluster should be DiI-negative; conversely, if Uzip mediates heterophilic cell-to-cell adhesion, the cluster should contain both DiI-positive and DiI-negative cells. We observed that the cells in clusters were all DiI-negative (Fig. 3C,D). Thus, Uzip triggers cell adhesion by homophilic binding.

Figure 3.

Uzip mediates homophilic cell-to-cell adhesion. A–D: The genes for green fluorescent protein (GFP; A,C) or Uzip-His (B,D) were transfected into Drosophila S2 cells subjected to aggregation assays. C,D: The untransfected cells were labeled with 1,1′, di-octadecyl-3,3,3′,3′,-tetramethylindo-carbocyanine perchlorate (DiI). B: Uzip-His–expressed cells formed clusters marked by the anti-His antibody. D: The cells in the clusters were exclusively DiI-negative, suggesting that Uzip triggers aggregates by means of homophilic binding. E: The different truncated forms of Uzip are indicated and the illustrated domains correspond to Fig. 2B. The quantitative evaluation of cell aggregation was induced by the expression of different truncated forms of Uzip (lower panel). The minimal cluster size was ≥10 cells. Data bars represent the mean ± SD from two independent experiments. *P < 0.05; **P < 0.01; ***P < 0.005 by Student's t-test.

Uzip Is a Conserved Insect Protein

We carried out similarity and homology searching for the Uzip protein sequence using FASTA ( Uzip has no identified domains found in the known adhesion molecules, indicating that Uzip is a newly identified CAM. Furthermore, Uzip has no homologue in organisms other than insects. For example, the ortholog in Drosophila pseudoobscura (Swiss-Port accession number is Q28Z91), exhibits 82.1% identity and 93.7% similarity to Uzip. Uzip is also conserved in other insects such as in Q7PZ21 of the African malaria mosquito, Anopheles gambiae (37.9% identity and 65.4% similarity) and in Q4PKR8 of the Western honey bee, Apis mellifera. The C-terminal residue of Q4PKR8 is not supported by any other of the published Apis mellifera expressed sequence tag (EST) sequences (Supp. Fig. S2A). Therefore, we replaced the C-terminus of Q4PKR8 with that of the EST DB778253 and termed it Q4PKR8-EST_APIME. This assembled sequence shares 28.3% identity and 58.5% similarity with Uzip.

Next, we aligned Uzip with these orthologs using ClustalW2 ( and found that the highly conserved regions are mainly located from amino acids 42 to 379 (Fig. 2B and Supp. Fig. S2B). To investigate whether this conserved region is critical for adhesive ability, we performed aggregation assays using two truncated Uzip proteins lacking the conserved regions: Uzip Δ48–232 and Uzip Δ233–376, which have lost the anterior and posterior halves of the conserved region, respectively. Aggregation assay results showed that these two constructs of Uzip have lost the ability to mediate cell-to-cell adhesion (Fig. 3E). In addition, the fragment of amino acids 401–450 shares low identity among insects, suggesting that this fragment might not be required for adhesion. Indeed, S2 cells expressing the UzipΔ401–450, in which the fragment is deleted, were still able to form aggregates (Fig. 3E). These results suggest that this highly conserved region is important for the adhesive ability of Uzip orthologs among the insects.

Characterization of Uzip

The Uzip full-length protein consists of 488 amino acids with a Serine/Threonine (Ser/Thr) rich domain at amino acids 380–400 (Fig. 2B). It has a presumptive molecular weight of 55 kDa. However, endogenous Uzip in wild-type fly extracts exhibited at least two forms: 80 and 65 kDa (Fig. 4A). This discrepancy in molecular weights could result from glycosylation, because Uzip has five N-glycosylation sites predicted by NetNGlyc 1.0 Server (Fig. 2B). We investigated the Uzip-His expressed in Drosophila S2 cells. Western blotting results revealed that the rabbit anti-Uzip antibody recognized Uzip-His in two major bands that are almost identical to that of the endogenous Uzip (Fig. 4A,B). To examine whether Uzip might be a glycoprotein, we digested it with a deglycosylation enzyme, Peptide-N-Glycosidase F (PNGase F), which can hydrolyze nearly all types of N-glycan chains (Maley et al.,1989). After PNGase F digestion, the molecular weight of Uzip-His was reduced (Fig. 4B), indicating that Uzip indeed bears N-glycosylation modifications. The appearance of multiple bands upon deglycosylation (Fig. 4B) suggests that, in addition to N-glycosylation, Uzip might be produced through other posttranslational modification and cleavage modes that we have not studied.

Figure 4.

Characterization of Uzip. A: Adult fly extracts were immunoblotted using a rabbit anti-Uzip antibody. Two major forms at 80 and 65 kDa were detected in wild-type (WT) flies. uzip23 and uzipD43 were two uzip mutant alleles showing reduced or no Uzip expression, respectively. B–D: Drosophila S2 cells were transfected with green fluorescent protein (GFP), Uzip-His, or Uzip452-NrgTM, and Western blotting assays were performed. B: The major bands of Uzip shifted to a lower molecular weight after PNGase F digestion, indicating that Uzip was modified with N-link carbohydrates. C: Uzip was detected in both cell lysates and the culture medium. One band at 65 kDa was detected by anti-Uzip but not anti-His in the culture medium. In cell lysates, one band at 80 kDa is recognized by anti-His, while two major bands at 80 and 65 kDa are recognized by anti-Uzip. Tubulin was used as a loading control. D: Uzip is a glycosylphosphatidyl inositol (GPI) -anchored protein. Uzip could be cleaved by phosphoinositide phospholipase C (PI-PLC) and secreted into the working solution (asterisks), but the transmembrane form Uzip452-NrgTM could not be cleaved by PI-PLC.

Uzip Exhibits GPI-Anchored and Secreted Forms Upon Expressed in S2 Cells

Uzip has a predicted signal peptide at its N-terminus suggesting that Uzip might be transported by a secretory pathway. To identify the form of Uzip, we used both anti-His and anti-Uzip antibodies to detect the Uzip-His expressed in S2 cells. In cell lysates, anti-His antibody recognized one band at 80 kDa, while anti-Uzip labeled two major bands at 80 and 65 kDa (Fig. 4C). The data suggest that at least two forms detected in cell lysates: C-terminus loss and C-terminus retaining forms. Surprisingly, Uzip was also detected in the culture medium. This released form is approximately 65 kDa and it was recognized only by anti-Uzip but not by anti-His (Fig. 4C), suggesting that this 65 kDa form is a secreted Uzip that lacks the C-terminus.

Besides, Uzip has a predicted transmembrane domain at amino acids 466–486 and a GPI-modification site at Asn465 (Prediction by PredGPI: (Fig. 2B). The protein sequence analysis suggests that Uzip might exist as a transmembrane or a GPI-anchored form. We tested whether Uzip exhibits GPI-anchored form by using a phosphoinositide phospholipase C (PI-PLC) cleavage assay. PI-PLC is an enzyme that cleaves the phosphoglycerol bond linking the GPI moiety (Sundler et al.,1978) and releases GPI-anchored proteins from the cell surface. We found that Uzip-His was indeed cleaved by PI-PLC and released into the medium (Fig. 4D). In addition to the full-length Uzip, we also designed a control construct Uzip452-NrgTM, in which amino acids after position 452 including the GPI-modification site were substituted by the Neuroglian transmembrane domain (NrgTM; Fremion et al.,2000). The artificial membrane-integral Uzip452-NrgTM could not be cleaved by PI-PLC (Fig. 4D). Furthermore, the predicted GPI-modification sites were also found in the Uzip orthologs of other insects (Supp. Fig. S2B). These data suggest that Uzip is a GPI-anchored protein.

Uzip has only one transcript. Uzip-His intact at its C-terminus has also been detected, suggesting the existence of a transmembrane form. However, to our understanding, GPI-anchored proteins do not exist simultaneously as transmembrane form without alternative splicing. Because extra 45 amino acids had been added at the C-terminus of Uzip while generating the Uzip-His, the C-terminus retaining form might result from interference in the cleavage of the C-terminal GPI signal. To test this possibility, we introduced full-length Uzip without the His tag into S2 cells and carried out the PI-PLC assay. As the result of Uzip-His, this nontagged Uzip was cleaved by PI-PLC and was released into the working solution (Supp. Fig. S3). We also extracted the proteins on the cell membrane using Triton X-114 treatment (Bordier,1981) and found that most nontagged Uzip detected in membrane fractions were released after PI-PLC treatment (Supp. Fig. S3). The residual Uzip in the membrane fraction could result from incomplete digestion with PI-PLC. According to these data, we speculate that most Uzip expressed on the cell membrane of S2 cells is the GPI-anchored form.

To summarize, we proved the existence of N-glycosylation modifications on Uzip and found that the Uzip-His expressed in S2 cells exhibits two forms: the GPI-anchored and the secreted forms.

Membrane Attachment Is Required for Adhesive Ability

Next, we tested whether the secreted and membrane-attached Uzip are both involved in mediating cell-to-cell adhesion. We introduced the truncated form Uzip452, which lacked the GPI-modification site but still possessed the conserved region at amino acids 42–379, into S2 cells. Uzip452 could be secreted to the culture medium (data not shown) but loss its cell adhesive ability in a cell aggregation assay (Fig. 3E). Thus, the secreted Uzip seems insufficient for cell adhesion. The data also suggest that membrane attachment might be required for adhesive ability. We verified this idea by producing the membrane-integral form, Uzip452-NrgTM. Indeed, this transmembrane form of Uzip exhibited cell adhesive ability (Fig. 3E). The data indicate that the membrane-attached Uzip is able to bind cells together.

Expression of uzip mRNA and Protein in the Embryonic VNC

We examined the expression pattern of uzip mRNA by performing whole-mount in situ hybridization in wild-type embryos (Fig. 5A) and found that uzip mRNA was expressed dominantly in the VNC as described previously (Cote et al.,1987; Zhao et al.,1988). In the dorsal region of the embryonic VNC, axons fasciculate in rail-like axonal tracts consisting of longitudinal connectives and commissures. In addition to the broad distribution in the VNC at embryonic stages 12–16, we also observed that uzip mRNA was enriched in the two most dorsal stripes identical to the LG location in the VNC (Fig. 5B), suggesting that uzip mRNA is present in the LG. Indeed, immunostaining of Uzip was detected around the LG in the loco-lacZ embryo, in which all LG nuclei are recognized by anti–β-galactosidase (anti–β-gal; Granderath et al.,1999; Fig. 5C). Uzip was also detected in a focal plane ventral to the LG. Double immunostaining using anti-Uzip antibody and anti–horseradish peroxidase (HRP), which labels axons, revealed that Uzip was found in the periphery and interior of longitudinal connectives (Fig. 5D). To further examine the subcellular localization of Uzip in the LG, we used the repo>mCD8::GFP strain, in which the plasma membranes of all lateral glia including the LG are labeled with GFP. Uzip expression was detected in the intercellular space, in the plasma membrane of the LG and in the LG sheath surrounding the axon tracts (Fig. 5E,F). The localization of uzip mRNA in the entire VNC and of Uzip protein in the LG and axons suggest that Uzip is produced at these sites. However, the enrichment of uzip mRNA in the LG suggests that this region might produce more Uzip than neurons.

Figure 5.

uzip mRNA and protein are enriched at ventral nerve cord (VNC) axons and longitudinal glia (LG). A: uzip mRNA expression of wild-type embryos was obvious in the CNS and especially enriched in the most dorsal layer of the VNC (arrowhead). B: Magnification of the dorsal layer of the VNC. The pattern of dorsally restricted uzip mRNA is identical to the LG distribution (arrowhead). C–H″: The VNC of stage 15–16 embryos were immunostained with antibodies to Uzip and to horseradish peroxidase (HRP), which marks the central nervous system (CNS) axons. C–C″: Uzip surrounded the LG, which were labeled using loco-lacZ. D–D′: Uzip was located on VNC axons (asterisks in D, D′ and F′) and wrapped the longitudinal tracts (arrowhead). E–F″: In repo>mCD8::GFP embryos, the plasma membrane of the LG was recognized by green fluorescent protein (GFP). ML: VNC midline. A cross-section of the VNC was shown (F–F″). Uzip accumulated at the LG interspaces (arrows in E–F″) and was irregularly distributed on the LG membranes (arrowheads in E–F′). Square brackets indicate the LG layer. G–H″: The cross-sections of VNC is shown. Uzip proteins were absent in the gcm (G) and uzipD43 mutants (H).

Contribution of Glia and Neurons to Uzip

To verify the role of glia in producing Uzip, we examined Uzip expression in the glial cell missing (gcm) mutant, in which glia are transformed into neurons (Jones et al.,1995). In this mutant, the Uzip signals at the focal planes of the LG and axons were both absent (Fig. 5G), suggesting that glia are indeed the main source of Uzip.

To further clarify the contribution of glia and neurons to production of Uzip, we introduced the double-stranded RNA of uzip (uzipRi) using repo-GAL4 or elav-GAL4 to knock down uzip in glia and neurons, respectively. Upon glial knockdown, the Uzip immunoactivity at LG and axons were reduced (Fig. 6A–F,J,K). In addition, the Uzip level in the whole embryo was significantly decreased (Fig. 6L). The data support the idea that LG produce Uzip and that axons contain the LG-produced Uzip. On the other hand, neuronal knockdown appeared to alter the Uzip distribution in the VNC (Fig. 6G–K). Uzip immunoactivity at LG was comparable to that of the wild-type. However, Uzip signals was increased at neurons that locate more laterally than the LG and axon tracts (Fig. 6G,J,K). Moreover, the immunoactivity at the axon tracts was also elevated (Fig. 6H,K). However, neuronal knockdown of uzip did not significantly increase the Uzip level in the whole embryo (Fig. 6L). Thus, neurons are likely to produce a small portion of Uzip, which is undetectable by antibody staining in gcm mutant but might be critical for the distribution of Uzip.

Figure 6.

Glial and neuronal knockdown of uzip alters its protein level and distribution. A–I: The ventral nerve cord (VNC) of wild-type, glial (repo>uzipRi), and neuronal (elav>uzipRi) knockdown of uzip. The stage 15 embryos were immunostained by anti-Uzip. A,D,G: The focal plane of longitudinal glia (LG). B,E,H: The focal plane of axons. C,F,I: The cross-section at red arrows are shown. A–F: The Uzip expression in LG and axons (A–C) was reduced in subsets of segments by glial knockdown (D–F). G,H: In neuronal knockdown, Uzip appeared to expand laterally at the LG focal plane (G), and the protein level was significantly elevated at axons (H). Images in all panels were collected using the same gain and contrast. J,K: Quantification analysis of the Uzip distribution at the focal plane of LG (J) and axons (K). Plots shown represent 100 individual line scans in 35 segments of wild-type, 41 of repo>uzipRi, and 28 of elav>uzipRi embryos. Arrowhead, midline. L: Western blot showing Uzip and Tubulin expression in embryo extracts. The bands were quantified by Image J. Embryonic Uzip levels (normalized to Tubulin) from four experiments are represented as arbitrary units ± SEM relative to controls. ***P < 0.005 by Student's t-test.

Generation of uzip Mutants

To generate uzip mutants, we performed experiments of imprecise excision, in which the EP-C04101 was crossed to the fly strain expressing the transposase. The consequence of P-element mobilization was determined by polymerase chain reaction (PCR) amplification and sequence analyses. The uzip23 sequence bears a 300 bp fragment of the P-element remaining at the insertion site, although the entire coding sequence of uzip is intact. The uzip23 homozygous fly expressed Uzip at a reduced level (Fig. 4A), indicating that uzip23 is a hypomorphic allele. We also deleted the uzip gene using the two P-element lines f01534 and f02444 (Fig. 2A) and the flippase recognition target/flippase recombination enzyme (FLP–FRT) -based deletion system (Parks et al.,2004). The P-element of f01534 is inserted in the second intron, upstream of the translation start site in the third exon. The P-element of f02444 is located at the sixth exon and is downstream of the translation stop codon. The P-element of these lines contains an FRT site that allows homologous recombination to occur in the presence of the FLP recombinase. The recombination between these two chromosomes can generate a deletion of the entire uzip coding sequence. We obtained uzipD43, which is homozygously viable and displayed no detectable Uzip immunoactivity in the VNC (Fig. 5H) or in adult lysates (Fig. 4A). Uzip immunoactivity was also absent in the transheterozygote of uzipD43 and Df(2R)BSC608 (data not shown), which is the deficiency uncovering uzip. These data suggest that uzipD43 is a null allele.

uzip Genetically Interacts With CadN in Regulating Axon Guidance

The localization of Uzip at the LG and axon fascicles prompted us to investigate whether uzip participates in the development of these elements. We used markers of glial nuclei, glial membranes, neurons and axons to examine the uzipD43 embryos, but found no detectable abnormalities (Supp. Fig. S4). This result did not surprise us because single mutants for many CAMs usually display no gross defects. Because CAMs cooperate with other CAMs and signaling molecules in regulating axon guidance, the simultaneous loss of two CAMs, or the loss of one CAM and one signaling molecule, results in obvious axonal defects (Elkins et al.,1990; Speicher et al.,1998). If Uzip contributes to the VNC through cooperating with other CAMs or signaling molecules, there should be a loss-of-function phenotype of uzip in the double mutant.

We performed genetic testing between uzip and CadN, which encodes the major cadherin locating at VNC axons (Iwai et al.,1997). We analyzed the axonal phenotype using immunostaining with the monoclonal antibody 1D4 against Fasciclin II (FasII), which is expressed at pioneer axons of longitudinal connectives. At embryonic stage 17, 1D4 stained three major longitudinal fascicles: medial, intermedial, and lateral. The uzip mutant exhibited the axonal fascicle as in wild-type embryos (Fig. 7A and Supp. Fig. S5A). In the CadN mutant, there were minor axonal breaks and defasciculations (Fig. 7B,K). Intriguingly, heterozygosity of uzipD43 increased the axonal defects (Fig. 7C,K), suggesting that the axonal pathway in the CadN mutant is sensitive to uzip reduction. We also found that the axonal defects were exacerbated when the levels of Uzip were reduced by using the uzip23/uzipD43 heterozygote (Fig. 7K). Moreover, in the double mutant of CadN and uzipD43, the three longitudinal pathways were broken extensively (Fig. 7D,K). The data suggest that there must be cooperation between Uzip and CadN in regulating axon guidance.

Figure 7.

uzip genetically interacts with CadN and wnt5. A–F: The three longitudinal pioneer fascicles of stage 17 embryos were immunostained by the 1D4 antibody. M, medial; I, intermedial; L, lateral. A: The uzipD43 revealed a normal Fasciclin II (FasII) axon trajectory. B: The lateral pathway was broken in CadN (arrows). C,D: The broken lateral pathway in CadN uzipD43 mutant revealed the dose-dependency of uzip. E,F: The phenotype of broken lateral pathway in wnt5 (arrows) was slightly enhanced by loss of uzip. G–J: Sema2b-τmyc embryos were immunostained with anti-Myc antibody. G: The Sema2b axon trajectory is normal in uzipD43. H: The Sema2b axons in the longitudinal pathway stalled in CadN (black arrow). I: The stalled phenotype was enhanced by the heterozygosity of uzipD43. J: Stalled axons, an abnormal projection toward the posterior (black arrowhead) and defasciculated commissures (white arrow) were found in the CadN uzipD43 mutant. K,L: The percentages of hemisegments with defective longitudinal pathways are shown as a bar graph. ***P < 0.005 using a two-tailed Fisher's exact test. The number of hemisegments counted are shown above the bars.

We used knockdown experiments to confirm the genetic interaction of uzip and CadN. Neuronal knockdown of uzip could slightly, but did not significantly, increase axonal defects of the CadN mutant (Fig. 7K). It appears to be that neuronal-produced Uzip functions in regulating the distribution of Uzip, and might contribute to axon guidance only in a low extent. In contrast, glial knockdown obviously increased the axonal defect in the CadN mutant (Fig. 7K). Therefore, upon loss of CadN, glia-produced Uzip is required for axon guidance.

Impaired Axonal Trajectories of Sema2b Neurons in the CadNuzipD43 Mutant

To investigate the axonal defect in detail, we observed the single axonal pathway labeled by Sema2b-τmyc (Rajagopalan et al.,2000). In each of the A5–A8 hemisegments, Sema2b-τmyc marks two or three Sema2b neurons located laterally in the VNC. Their axons cross the midline through the anterior commissure and then project anteriorly to reach the prior segment. Mutation of uzip did not affect this axonal patterning (Fig. 7G,L). Loss of CadN affected the Sema2b axonal pathway mildly, resulting in axons that failed to reach the prior segment (Fig. 7H,L). The defect was enhanced by heterozygosity and homozygosity of uzipD43 (Fig. 7I,J,L). In the CadN uzipD43 mutant, Sema2b axons could successfully reach the contralateral side and combined mild defasciculation in commissures (Fig. 7J). After reaching the contralateral side, some axons stalled and some axons turned posteriorly but not anteriorly (Fig. 7J). These data indicate that there is cooperation of Uzip and CadN in regulating axonal pathfinding in the longitudinal pathway.

uzip Interacts Genetically With wnt5 in Regulating Axon Guidance

The Wnt family member, Wnt5, is localized at axons and regulates axon guidance through the receptor Derailed (Drl; Yoshikawa et al.,2003; Fradkin et al.,2004), which has been reported to function in concert with the CAM Neurotactin (Speicher et al.,1998). Wnt5 is found at commissural and longitudinal axons, and is required for pathway selection of commissures and the selective defasciculation of the FasII longitudinal pioneer axons (Yoshikawa et al.,2003; Fradkin et al.,2004). Overexpression of wnt5 in the midline glia, driven by sim-GAL4 (Golembo et al.,1996), disrupts the formation of the anterior commissure (AC; Fradkin et al.,2004). Embryos with two copies of sim-GAL4 and UAS-Wnt5 caused thinner or loss of AC (46.7%, n = 150 segments). It has been hypothesized that ectopically expressed Wnt5 could repulse AC or cause overly tight fasciculation of longitudinal axons that preclude the midline crossing of AC (Fradkin et al.,2004). We performed this assay in the uzipD43 mutant to test whether Uzip is required for Wnt5-mediated axon guidance, and found that the phenotype of defected AC was not apparently affected (40%, n = 119 segments). The data suggest that Uzip is unlikely to be required for Wnt5-mediated axon guidance. However, we found that the loss of uzip exacerbates axonal defects in the wnt5 mutants. At embryonic stage 17, wnt5 mutants exhibited broken FasII fascicles mainly in the lateral pathway (lateral, 6.7%, n = 225 hemisegments; Fig. 7E). In the double mutant lacking both wnt5 and uzip, defects of the FasII axons were slightly enhanced (lateral, 13.3%, n = 162; Fig. 7F). The data suggest that Uzip might also cooperate with the signaling molecule Wnt5.

Aberrant Trajectories of Pioneer Axons in the Double Mutants

We tested whether the aforementioned axonal defects might result from the failure of axonogenesis at early stages of development. First, we examined the expression of Uzip at embryonic stage 12, when axonogenesis occurs. Uzip was detected in subsets of the LG and pioneer axons (Fig. 8A). This early expression suggests that Uzip might participate in regulating the early development of pioneer axons. At stage 12, four pioneer neurons—dMP2, MP1, vMP2, and pCC—labeled by 1D4 extend axons that form the descending MP1/dMP2 and ascending pCC/vMP2 fascicles (Lin et al.,1994; Hidalgo and Brand,1997). At stage 13, these two fascicles meet and merge to become a single fascicle (Fig. 8B). As in the single mutants of CadN and wnt5, the early trajectories of pioneer axons in the double mutants were similar to that of wild-type embryos (Fig. 8B–D and data not shown). Staining using the antibody 22C10 (Fujita et al.,1982; Seeger et al.,1993) confirmed the early pioneer axonal trajectories (data not shown). The data indicate that the interactions between uzip and CadN and between uzip and wnt5, do not contribute to the outgrowth and initial pathfinding of the pioneer axons. Later on, at stage 14 the single fascicle defasciculates to become the MP1/dMP2 outer and pCC/vMP2 inner fascicles, both of which are associated only at the segment border (Fig. 8E). As mentioned in the previous study (Iwai et al.,1997), in the CadN mutant, the MP1/dMP2 pathway was sometimes discontinuous (20.3%, n = 158 hemisegments) and the pCC/vMP2 pathway was not affected obviously (4.1%, n = 122 hemisegments). In the double mutant of CadN and uzip, the two pathways showed more defects than in the CadN mutant. They did not form normally where defasciculation is supposed to occur, but became fuzzy, thinning, or broken (MP1/dMP2: 29.2%, n = 130 hemisegments; pCC/vMP2: 50%, n = 122 hemisegments; Fig. 8E–G). The data indicate that Uzip and CadN cooperatively affect the axonal growth of pioneer axons.

Figure 8.

Expression of Uzip and axonal defects at early stages. A: Expression of Uzip in the embryonic stage 12 ventral nerve cord (VNC) of repo>mCD8::GFP embryos. During axonogenesis, Uzip was detected at the longitudinal glia (LG) labeled with green fluorescent protein (GFP; arrow) and longitudinal pioneer axons labeled with horseradish peroxidase (HRP; arrowhead). B–G: 1D4 staining of the stage 13 (B–D) and stage 14 (E–G) VNC. One of the earliest pioneers of the longitudinal pathway, the pCC axon (arrow), extended normally at stage 13 in wild-type embryos, the CadN mutant, and the double mutant of uzip and CadN. E: At stage 14, defasciculation occurs between the MP1/dMP2 outer (arrow) and the pCC/vMP2 inner (arrowhead) pathways. F: In the CadN mutant, the MP1/dMP2 pathway was partly broken (arrow) and the pCC/vMP2 pathway was barely affected. G: In the double mutant of uzip and CadN, more defects including fuzzy (square bracket), thinning axons (white arrow and arrowhead) and breaks (black arrow and arrowhead) in the two pathways were observed.

On the other hand, loss of wnt5 affects the defasciculation of the MP1/dMP2 and pCC/vMP2 pathways (Fradkin et al.,2004), resulting in a single fascicle (35.2%), thinning or breaks (MP1/dMP2: 18.5%; pCC/vMP2: 19.4%, n = 108 hemisegments). Upon simultaneous loss of uzip and wnt5, these defects were mildly enhanced: 31% of the hemisegments exhibited a single fascicle; 24.7% and 26.4% (n = 174) of the hemisegments had thinning or broken MP1/dMP2 and pCC/vMP2 pathways, respectively. These data suggest that the cooperation of Uzip and Wnt5 might slightly affect the early development of pioneer axons.

Expression of Uzip-CFP Partially Rescued the Axonal Defects of the CadN uzipD43 Mutant

We performed rescue experiments by overexpressing Uzip-CFP, a full-length Uzip tagged with a cyano fluorescent protein (CFP) at the carboxyl terminus, in the CadN uzipD43 mutant. We used repo- and elav-GAL4 to drive Uzip-CFP expression in glia and neurons, respectively. Immunostaining with anti-Uzip showed that Uzip-CFP was localized at, but not outside, the expressing cells (Fig. 9). We reason that the addition of CFP on Uzip might generate an artificial transmembrane form as the Uzip-His expressed in S2 cells and thus preclude the secretion. Importantly, both glial and neuronal expressions of Uzip-CFP partially rescued axonal phenotypes of the CadN uzipD43 mutant (Fig. 7K and Supp. Fig. S6). The neuronal-expressed Uzip-CFP was not detected at axons. The expression and the rescue result suggest that a low level of axonal Uzip-CFP, which is undetectable by antibody staining, might be functional. Thus, the data suggest that uzip is indeed responsible for the axonal phenotype, and that both Uzip attached at LG and axonal surfaces play a role for axon guidance.

Figure 9.

Expression and rescue results of Uzip–cyano fluorescent protein (CFP) in the CadN uzipD43 mutant. The stage 15 embryos were immunostained by mouse anti-Uzip (A–D), and the stage 17 embryos were immunostained by 1D4 antibody (E,F). A,C: When the Uzip-CFP transgene was driven by repo-GAL4 in the double mutant, Uzip was detected at LG, but not axons. B,D: When Uzip-CFP was neuronally overexpressed in the double mutant, Uzip largely accumulated at neuronal cell body, but could not be detected at axons. The cross sections at white arrows are shown (C,D). E,F: The Fasciclin II (FasII) axonal defects of the double mutant were partially rescued by the expression of Uzip-CFP in glia or neurons.


Uzip Is a Novel Regulator of Axon Growth Produced by LG

Guidepost cells are required for axon guidance in the nervous systems. In Drosophila embryonic CNS, the LG are such guidepost cells. However, because the known LG-produced molecules are limited, how the LG regulate axon guidance remains elusive. Our data show that Uzip is expressed by the LG and that loss of Uzip impairs the axonal trajectory, indicating that Uzip is a novel LG factor participating in regulating axon guidance.

Origin, Secretion, and Localization of Uzip

In addition to GPI-anchored Uzip, we detected secreted Uzip in S2 cells. It is unknown whether Uzip is similar to the Drosophila Fasciclin I, whose secreted form is produced by phospholipase-mediated cleavage of the GPI-anchored form (Hortsch and Goodman,1990). In the embryonic VNC, Uzip was detected at LG and axons. The Uzip signals were eliminated by a mutation of gcm, and were significantly reduced by glial knockdown of uzip. The data indicate that LG are the main source of Uzip, and that the glia-produced Uzip can be secreted to be localized at axons. Low Uzip immunoactivity outside LG and axons suggests a mechanism controlling distribution of the secreted Uzip in the VNC. Upon the neuronal knockdown of uzip, the Uzip level in whole embryos was not significantly affected. However, Uzip was increased at the neurons in the neuropile and in the axon tracts. It appears to be that Uzip, presumably secreted by glia, is spread farther than in the wild-type. The data suggest that neurons produce Uzip only in a low extent, which might be required for localization of the glia-produced Uzip. Probably, neuron-produced Uzip binds the secreted glia-produced Uzip through homophilic binding to restrict its distribution. Further investigation is needed to understand the mechanism of secretion and localization of Uzip.

Uzip Is Required for Axon Guidance

The genetic interactions between uzip and CadN and between uzip and wnt5 suggest that Uzip plays a role in regulating axon guidance through cooperation with CadN and Wnt5. One possible mechanism is that Uzip associates with CadN and Wnt5. However, several lines of evidence do not support this hypothesis. First, co-immunoprecipitation analyses revealed no associations between Uzip and CadN when they were coexpressed in S2 cells (data not shown). Second, aggregation assays showed that CadN- and Uzip-expressing cells failed to aggregate with each other (Supp. Fig. S7). In addition, our pull-down assay revealed that there were no apparent changes in the migration of Wnt5 when Uzip was coexpressed, whereas Drl-Fc pulled down Wnt5 as expected (data not shown). Thus, Uzip is not likely to associate with CadN and Wnt5.

The single mutant of uzip displays no detectable phenotype, but Uzip is responsible for the axonal defects in the double mutants. We reason that when Uzip is absent, its function is taken over by a redundant molecule, which is provided by CadN- and Wnt5-dependent axons. With the loss of CadN or Wnt5, the altered axon patterning might affect the distribution of the redundant molecules, thus making Uzip essential for axon guidance. As functional redundancy has been found between CAMs (Kristiansen et al.,2005), it will be of interest to investigate whether Uzip acts redundantly with the other CAMs in the VNC.

Uzip Might Facilitate the Adhesion at LG and Axons

Uzip triggers cell–cell adhesion in the aggregation assays, indicating Uzip is a CAM. Sequence analyses reveal that Uzip has no domains shared with known classes of CAMs. Thus, to our knowledge, Uzip is a novel CAM. Because overexpression of uzip caused clustering of PNS neurons, the adhesive ability of Uzip might exist in embryos. In addition, like the CAM FasII (Lin and Goodman,1994), overexpression of Uzip in the VNC caused mild defects (Supp. Fig. S6). It is possible that overexpression of the CAM Uzip could generate aberrant binding of cells. This could disturb the signaling and interaction at LG and axons, which do not require Uzip in the wild-type embryos, leading to axonal and glial defects. This might explain why glial cell numbers were increased by overexpression of uzip but not in uzip mutants.

In the Drosophila VNC, reciprocal interactions between neurons and glia are essential for their development. Glia are required for neuronal survival and for axon guidance (Hidalgo et al.,1995; Jones et al.,1995; Booth et al.,2000; Hidalgo and Booth,2000). On the other hand, axons provide factors controlling the specification and survival of glia (Booth et al.,2000; Thomas and van Meyel, 2007). The interdependence might explain the colocalization of the axonal and glial defects in the double mutants (Supp. Fig. S8). Overexpression of Uzip-CFP in glia and neurons, respectively, partially rescued the axonal phenotype of the CadN uzipD43 mutant, suggesting that both Uzip at glia and axons are able to regulate axon guidance. Importantly, because the Uzip-CFP appears to reside at, but not outside, the expressing cells, the rescue results also suggest a functional independence of the Uzip at LG and the Uzip at axons. The Uzip at axons, which includes the LG-secreted and neuron-produced one, might contribute to axonal development primarily. However, the Uzip at LG might regulate LG primarily and then affect axons. We propose that Uzip at LG and axonal surfaces could regulate axon guidance by promoting the LG–LG and axon–axon adhesion, respectively. Besides, the endogenous Uzip exists at both LG and axons, suggesting that Uzip might also mediate the LG–axon adhesion.

In summary, Uzip might act as the glue that facilitates the binding at LG and axons and thereby regulate their embryonic development. Further investigations of Uzip will increase the understanding of the interactions between glia and axons.


Fly Strain

A batch of transgenic flies bearing the EP P-element on the second or third chromosome was generated (unpublished data of C.T. Chien) by introducing transposase to the fly strain that possesses the EP P-element on the X chromosome (Rorth,1996). Each EP line was crossed to a fly carrying elav-GAL4 and UAS-mCD8::GFP (Lee et al.,2000). For generating transgenic flies, transformation of wild-type embryos was performed as described (Spradling,1986). The following strains were used: elav-GAL4 (Luo et al.,1994), UAS-mCD8::GFP (Lee et al.,2000), loco-lacZ, repo-GAL4, and Df(2R)BSC608 (Bloomington Drosophila Stock Center), f01534 and f02444 (Exelixis Collection at Harvard), gcm (Jones et al.,1995), CadNM19 (Iwai et al.,1997), sim-GAL4 (Golembo et al.,1996), wnt5400 and UAS-Wnt5 (Fradkin et al.,2004), and Sema2b-τmyc (Rajagopalan et al.,2000). The sympUAST-Uzip and pUAST-Uzip-CFP construct was injected into wild-type embryos following standard procedures (Rubin and Spradling,1982).

Molecular Cloning

The full-length uzip cDNA in the pOT2-SD09738 plasmid was amplified by PCR with the primers CG3533N and CG3533C-nonstop, and the fragment was cloned into pGEM-T-Easy (Invitrogen) to form pGEM-T-Easy-Uzip construct. Then the full-length uzip cDNA was digested with EcoRI and cloned into the symmetrically transcribed pUAST (sympUAST) vector (Giordano et al.,2002) predigested with EcoRI, generating the sympUAST-Uzip. To generate pUAST-Uzip-CFP construct, the full-length uzip with XhoI site was amplified from pOT2-SD09738 plasmid by PCR with the primers, CG3533N and 3′R-uzip-XhoI-nonstop. The PCR fragment was cloned into pGEM-T-Easy and confirmed by sequencing. The full-length uzip-XhoI was then subcloned into pUAST-CFP vector by EcoRI and XhoI partial digestion, termed pUAST-Uzip-CFP. To examine the consequence of P-element mobilization in uzip23, the primers uzip-5′-F300bp and uzip-3′-R300bp were used.

To express different Uzip forms in S2 cells, we cloned different uzip fragments into the expression vector pMT/V5-TOPO-His (Invitrogen). Details are described in the Supporting information. pRmHa3′-DN was used to express CadN in S2 cell (Iwai et al.,1997). All constructs were verified by sequencing, and the primers are listed in the Supporting information.

Cell Culture

The S2 cell culture followed the Drosophila Expression System (DES) manual (Invitrogen). Transient transfections were performed as in the procedure manual of the Calcium Phosphate Transfection Kit (Invitrogen). Cell aggregation assays were performed as described (Islam et al.,2003).

Deglycosylation and PI-PLC Assays, and Western Blotting

A total of 2 × 106 cells were collected and lysed in lysis buffer (50 mM Tris, pH 7.8, 150 mM NaCl, 1% NP-40, 0.1 mM phenylmethylsulfonyl fluoride, 0.2% protease inhibitor (Roche). Proteins in culture medium were concentrated by acetone precipitation, and were resuspended with 1× sodium dodecyl sulfate (SDS) loading buffer. For deglycosylation assay, 2 × 106 GFP- or Uzip-transfected cells were lysed as above description, and were digested by PNGase F (New England BioLabs) according to the manufacturer's protocol. For PI-PLC (Sigma) assay, 2 × 106 Uzip- or Uzip452-NrgTM- transfected cells were treated as previous description (Tsuda et al.,1999). The proteins in membrane fractions were extracted by the detergent Triton X-114 as description (Hortsch and Goodman,1990). To prepare tissue protein sample, 1 fly or 0.1 g dechorionated embryo were homogenized in 1× SDS loading solution. Protein samples were separated by 8% SDS-polyacrylamide gel electrophoresis (SDS-PAGE) for further analysis. We used first antibody including rabbit anti-Uzip (1:200), mouse anti-Uzip (1:100), mouse anti–penta-His (QIAGEN, 1:1,000), and mouse anti–alpha-tubulin (Developmental Studies Hybridoma Bank, DSHB, 1:1,000), and secondary antibody (Jackson Immuno Research 1:1,000) including anti-mouse HRP and anti-rabbit HRP.


The mouse anti-Uzip antibody has been raised against an extracellular fragment of 174–405 aa. To generate the immunogen, the DNA fragment encoding Uzip 174-405 aa was amplified by PCR from pOT2-SD09738 and inserted in frame to the 6xHis coding sequence in pET15b (Novagen, Madison, WI), termed pET15b-Uzip. pET15b-Uzip was expressed in Escherichia coli and purified using His-Bind Purification Kits for immunization (LTK BioLaboratories). The rabbit anti-Uzip antibody was raised against the extracellular fragment of 294–311 aa (QCB).

In Situ Hybridization and Immunostaining

Whole-mount in situ hybridization of embryos was performed as described (Tautz and Pfeifle,1989). For immunostaining, the antibodies we used including mouse anti–penta-His (QIAGEN, 1:500), mouse anti-Uzip (1:10), mouse anti-Myc (Santa Cruz, 1:100), rabbit anti-GFP (Santa Cruz, 1:200), rabbit anti–β-gal (Cappel, 1:100), and rabbit anti–HRP-TRITC (Jackson Immuno Research, 1:100). Other antibodies were obtained from the DSHB including anti-Elav (1:100), 1D4 (1:10), anti-Repo (1:100), anti-Prospero (1:100), anti-Wrapper (1:150), and anti–Even-skipped (1:100). The immunostaining protocols were performed as described (Patel,1994). Images of immunostaining were obtained with Zeiss LSM 510 confocal microscope or Zeiss Axioskop2 plus Normarski optic microscope.


All images were collected with identical laser power, gain, and contrast. For quantification of the levels of Uzip, lines were drawn perpendicular to the CNS midline of a Z projection image, and scanned along the anterior–posterior axis of embryos with identical length. The pixel intensities were plotted according to pixel number. All data were analyzed by Student's t-test.


We thank Dr. Tadashi Uemura for the fly and plasmid of CadN, Dr. Chien-Kuo Chen for providing the pUAST-CFP plasmid, and Dr. Lee G. Fradkin and Dr. Jasprina N. Noordermeer for the pull down assay. We also thank the Bloomington Stock Center for fly stocks and the Developmental Studies Hybridoma Bank for antibodies. This work was supported by grants from Academia Sinica and the National Science Council of Taiwan.