Dscam Guides Embryonic Axons By Netrin-Dependent And Independent Functions

Summary

Developing axons are attracted to the CNS midline by Netrin proteins and other as yet unidentified signals. Netrin signals are transduced in part by Frazzled (Fra)/DCC receptors. Genetic analysis in Drosophila indicates that additional unidentified receptors are needed to mediate the attractive response to Netrin. Analysis of Bolwig’s Nerve reveals that Netrin mutants have a similar phenotype to Down Syndrome Cell Adhesion Molecule (Dscam) mutants. Netrin and Dscam mutants display dose sensitive interactions suggesting that Dscam could act as a Netrin receptor. We show using cell overlay assays that Netrin binds to fly and vertebrate Dscam, and that Dscam binds Netrin with the same affinity as DCC. At the CNS midline, we find that Dscam and its paralog Dscam3 act redundantly to promote midline crossing. Simultaneous genetic knockout of the two Dscams and the Netrin receptor fra produces a midline crossing defect that is stronger than the removal of Netrins, suggesting that Dscams also function in a pathway parallel to Netrins. Additionally, over-expression of Dscam in axons that do not normally cross the midline is able to induce ectopic midline crossing, consistent with an attractive receptor function. Our results support the model that Dscams function as attractive receptors for Netrin and also act in parallel to Frazzled/DCC. Furthermore, the results suggest that Dscams have the ability to respond to multiple ligands and act as receptors for an unidentified midline attractive cue. These functions in axon guidance have implications for the pathogenesis of Down Syndrome.

Introduction

In bilaterally symmetric nervous systems, many axons are guided to the midline of the central nervous system (CNS) by proteins of the Netrin family (Hedgecock et al., 1990; Ishii et al., 1992; Serafini et al., 1994; Kennedy et al., 1994). In Drosophila, there are two redundant Netrin genes, NetA and NetB, and deletions of both lead to greatly reduced axon crossing of the midline (Mitchell et al., 1996; Harris et al., 1996; Brankatschk and Dickson, 2006). Nevertheless, significant numbers of axons persist in orienting towards and crossing the midline in NetA,B mutants, pointing to the existence of additional attractive midline cues (Hummel et al., 1999a; Brankatschk and Dickson, 2006). The abl non-receptor tyrosine kinase appears to participate in all midline attractive systems, because null mutants lack all commissures (Grevengoed et al., 2001), and Abl has been shown to function in Netrin signaling (Forsthoefel et al., 2005). The phenotypes of other mutants lacking midline crossing, commissureless, schizo and elav/weniger, have all been traced to effects on the Robo/Slit midline repulsion system (Onel et al., 2004; Keleman et al., 2005; Simionato et al., 2007). Therefore, despite extensive genetic screens (Seeger et al., 1993; Hummel et al., 1999b), components of the missing midline attractive system have yet to be identified and represent a major challenge for our understanding of the formation of the Drosophila ventral nerve cord.

There is also evidence for an unidentified attractive Netrin receptor. The work which identified DCC (Deleted in Colorectal Cancer) as a Netrin receptor noted that some DCC positive axons do not show any responses to Netrin, and postulated that the presence of a co-receptor might be required (Keino-Masu et al., 1996). The C. elegans DCC homologue, UNC-40, generally has mutant phenotypes less severe than UNC-6 (Netrin) mutants, suggesting the existence of a second pathway to respond to UNC-6 (Chan et al., 1996). In Drosophila, a single DCC family member, frazzled (fra), is present (Kolodziej et al., 1996). The fra (DCC) CNS phenotype is similar, but not identical to NetA,B deletions, as might be expected (Brankatschk and Dickson, 2006; Garbe and Bashaw, 2007; Garbe et al., 2007). In addition, for both migrating salivary glands and Netrin-responsive motor axons, the frazzled (fra) mutant phenotypes are of lower penetrance than those of NetA,B deletions (Kolesnikov and Beckendorf, 2005; Winberg et al., 1998; Labrador et al., 2005). Finally, two studies have provided convincing data that fra plays a non-autonomous role in axon guidance. In the embryo, the pioneer axon dMP2 has an altered trajectory in fra mutants; rescue of the mutant phenotype is not achieved by expression of fra in dMP2 alone, but requires expression by the cells encountered by the dMP2 axon (Hiramoto et al., 2000). In retinal projections, loss of axonal fra has little effect on their pathfinding, but loss of fra in the target tissue, the lamina, cause dramatic errors (Gong et al., 1999). In each case, Fra is thought to present Netrin to an unidentified receptor on the navigating axons.

The Drosophila Down Syndrome Cell Adhesion Molecule (Dscam) gene has been the focus of considerable attention due to the potential to encode 38,016 distinct protein isoforms through alternative splicing (Schmucker et al., 2000; Wojtowitz et al., 2004; Zipursky et al., 2006). However, there is minimal alternative splicing in vertebrate Dscam genes, and in the three other Drosophila Dscam genes (Yamakawa et al., 1998; Agarwala et al., 2001; Crayton et al., 2006). Given the evolutionary conservation of these molecules, this suggests that there is an important Dscam function that does not depend on molecular diversity. Genetic evidence in Drosophila also supports a diversity-independent function (Chen et al., 2006; Hattori et al., 2007). Like Drosophila Dscam, vertebrate Dscams are capable of mediating homophilic cell adhesion (Agarwala et al., 2000,2001). Knockdown of Dscam function in zebrafish leads to impaired cell movement, while perturbation of Dscam in the planarian disrupts cell migration, axon outgrowth and fasciculation (Yimlamai et al., 2005; Fusaoka et al., 2006). The diversity of these phenotypes coupled with the previously noted similarity of Dscam to other axon guidance receptors (Yamakawa et al., 1998), suggests that the primary Dscam function could be to respond to extracellular, perhaps diffusible ligands.

We found that Netrin mutants have similar phenotypes to Dscam mutants in Bolwig’s Nerve (the larval photoreceptor organ), suggesting that Dscam could function as a Netrin receptor. A physical interaction was confirmed in vitro using cell overlay assays. We also uncovered a subtle axon guidance defect in embryos mutant for one of the three additional Dscam genes in Drosophila, Dscam3 (CG31190; Millard et al., 2007). Genetic interactions between Dscam, Dscam3 and the abl tyrosine kinase indicated a role for the Dscam genes in midline crossing. Genetic interactions with the Netrin receptor, fra, suggested that the Dscams function in a parallel pathway to Netrin signaling. We favor a novel model in which Dscams are required for the transduction of several different axon guidance and cell migration cues, most probably through combinatorial association with other receptors.

Materials and Methods

Genetics

The piggybac line c02826 has two insertions (Thibault et al., 2004), one in Dscam3 and a second in CG7433, a 4-aminobutyrate (GABA) aminotransferase, which was removed by recombination. Interestingly, the parental double insert was not recovered in the F₂ generation, suggesting that the double mutant is less viable than either single mutant alone, possibly pointing to a synergistic interaction between the genes. The CG7433 insert does not appear to affect the CNS phenotypes. The f01683 Dscam3 allele is a lethal chromosome due to at least two background lethal mutations, and may only mildly disrupt Dscam3. The piggybac transposons were generated in a w⁻ isogenic background, which has significant effects on BN (Table 1) and motorneurons, so was not used as a control. fra alleles were obtained from P. Kolodziej and G. Tear. abl¹, abl⁴, Dscam alleles, and Df(1)NP5 were obtained from the Bloomington Stock Center. NetA,B^Δ was obtained from B. Dickson. In our crosses, we found the TM6 balancer increased phenotypic penetrance, and we suspect that the presence of Tb has a detrimental effect on viability, and appears to exert a maternal effect on the CNS phenotypes. Viable combinations of Dscams frequently show larval, pupal and adult phenotypes resembling Tb. To rule out Tb effects on axon guidance, all phenotypes were confirmed in the absence of balancer chromosomes by out-crossing to wild type chromosomes, and by crossing the F₁ progeny that lacked balancers.

Table 1. Quantification of axon guidance defects seen in Bolwig’s nerve.

Bolwig’s Nerve was scored for defects in axon guidance and fasciculation in homozygous and hemizygous combinations (top four rows) and in heterozygous combinations (bottom seven rows). We subdivided the phenotypes observed into mild and severe; a phenotype was severe if more than one axon bundle defasciculated from BN, if the defasciculation occurred early in the trajectory, or if the trajectory was dramatically altered from wild type. Outcrossing to wild type was critical to the trans-heterozygote analysis, as the presence of balancer chromosomes was found to enhance BN phenotypes. The NetA,B^Δ, NetA,B^NP5, Dscam^P and fra⁴ trans-heterozygous combinations were found by logical regression to increase the percentage of defects beyond purely additive effects.

Genotype	Total Counted	% Mild defects	% Severe defects	% Total defects
Oregon R (wildtype)	50	2	2	4
DscamP/P	51	15.6	11.8	27.5
NetA,BNP5/Y	52	40	50	90
NetA,BΔ/Y	52	36.5	59.6	96.2
fra3/fra4	50	16	10	26
DscamP/P fra4/4	54	28	39	67
DscamP/P fra4/4 Dscam31/1	50	12	46	58
NetA,BΔ/+	54	3.7	0	3.7
NetA,BNP5/+	50	4	2	6
DscamP/+	64	4.7	0	4.7
fra4/+	60	8.3	1.7	10
NetA,BNP5/+; fra4/+	70	14.3	15.7	30	***
DscamP +/fra4 +	40	17.5	5	22.5	***
NetA,BNP5/+ ;DscamP/+	100	15	23	38	***
NetA,BΔ/+; DscamP/+	50	10	24	34	***

^***(p<.0001)

DNA constructs

Drosophila Dscam was PCR amplified from the LP cDNA library (Berkeley Drosophila Genome Project), and cloned into pcDNA3-V5-His-Topo (Invitrogen). The Dscam splice isoform used in this work was 1-30-30-2. Human Dscam in pcDNA3 was a gift of W. Li and K.-L. Guan. The pGNET-myc chick netrin-1 and pcDNA3-HA-ratDCC constructs were gifts of M. Tessier-Lavigne. Drosophila NetB was PCR amplified from a cDNA clone (gift of G. Bashaw), and subcloned into pcDNA3-myc-His (Invitrogen). Sema3A and Sema3F expression constructs were gifts from H. Cheng.

Immunohistochemistry

Drosophila immunohistochemistry was performed as described in Patel, 1994. Mouse embryo immunohistochemistry was performed as follows: CD-1 mice were crossed and the day of the plug was designated E0.5. On the specified embryonic day embryos were dissected free of the uterus in 0.1M phosphate buffer. The embryos were then fixed in 4% paraformaldehyde in 0.1M phosphate buffer for 1 hour on ice, embedded in gelatin, frozen on a 2-methylbutane slurry and sectioned on a Leica cryotome. The sections were blocked for 1 hour in 5% goat serum in PBS (phosphate buffered saline) containing 0.1% Tween-20. Both anti-βIII-tubulin (Covance MMS-435P) and anti-hDSCAM (gift from K. Guan) were applied at 1:1000 dilutions overnight. Cy2 or Cy3-conjugated secondary antibodies (Jackson) were applied at a 1:200 dilution for 1 hour. All washes were done with PBS containing 0.1% Tween-20.

Cell overlay assays

COS-7 cells at 80% confluency were transfected with DNA expression constructs using Lipofectamine 2000 (Invitrogen), according to manufacturer's instructions (200ug DNA per 100ul). Approximately 40 hours post-transfection, the supernatant was removed from receptor expressing cells and replaced with supernatant containing epitope tagged ligand in the presence of 0.1% sodium azide. Cells were incubated at room temperature for 45 minutes before rinsing three times in 1xPBS and proceeding with antibody labeling. After rinsing, cells were fixed for 15 minutes in 4% paraformaldehyde with 0.1% Tween-20. The cells were blocked in 5% heat denatured normal goat serum in 1xPBS plus 0.1% Tween-20 for 15 minutes. After blocking, the cells were incubated with primary antibodies diluted in 5% heat denatured normal goat serum plus 1xPBS for 45 minutes (rabbit polyclonal anti-human DSCAM 1:1000 (Li and Guan, 2004); rabbit polyclonal anti-drosophila DSCAM 1:500 (Schmucker et al., 2000); anti-myc mouse monoclonal 1:200 (Calbiochem). Cells were then rinsed three times in 1xPBS plus 0.1% Tween-20. Secondary antibodies were diluted in 5% heat denatured goat serum in 1xPBS, then added to the cells and incubated for 30 minutes at room temperature (Jackson Labs Cy2 anti-rabbit 1:200 and Cy3 anti-mouse 1:200). The cells were subsequently washed in 1xPBS plus 0.1% Tween-20, followed by 1 minute incubation in 4′, 6-diamidino-2-phenylindole (DAPI) to stain the nuclei (Molecular Probes). The cells were then washed in 1XPBS and mounted in FluorSave (Calbiochem).

Binding Affinity

For binding affinity determination, COS cells were plated in 24-well plates at 50,000 cells/well and cultured in DMEM + 10% FBS + Pen/Strep. The next day, cells were transfected with 1 µg of pCDNA3.1, pCDNA3-HA-ratDCC (Hong et al., 1999) or pCDNA3-humanDSCAM (Li and Guan, 2004) using Lipofectamine 2000 (Invitrogen). One day later, the culture media was replaced with increasing concentrations of Netrin-myc protein in DMEM containing 10% FBS and 0.1% azide and incubated for 1 hour. The amount of Netrin-myc protein was quantified prior to use by western blot comparison to a standard curve of purified recombinant chicken Netrin-myc (Serafini et al., 1994). The cells were washed, fixed with 4% PFA for 5 minutes, and blocked using PBS containing 0.1% triton and 10% goat serum. Detection of the bound Netrin-myc protein was done using an anti-myc primary antibody (9E10) and a goat anti-mouse horseradish peroxydase (HRP) secondary antibody (Jackson) in PBS containing 0.1% triton and 1% goat serum. HRP activity was detected using SigmaFast OPD (o-Phenylenediamine dihydrochloride) peroxydase substrate (Sigma) and read using a microplate spectrophotometer.

Statistical Analysis

For determining the significance of trans-heterozygote combinations, the PROC GENMOD procedure in SAS software version 9.12 was used to fit a binary logistic regression model. The ESTIMATE option was used to estimate the differences between the linear combinations of the specific trans-heterozygote combinations. A chi-square test was performed to test whether the differences between the specified contrasts are statistically significant at 5% level. The SP1 neuron data were analyzed by fitting a Poisson model using the PROC GLIMMIX procedure in SAS software. Scatchard analysis was performed on four independent experiments using Prism 4 (GraphPad Software), and statistical significance was assessed using the t-test.

Results

Netrin mutants resemble Dscam mutants in Bolwig’s Nerve

The Drosophila larval photoreceptor organ, Bolwig’s Nerve (BN), extends a bundle of twelve axons to targets in the brain hemispheres, with axons making two stereotyped turns at choice points on the surface of the brain lobe (Schmucker et al., 1997,2000). Early in development, BN axons contact the optic lobe anlagen (OLA) and remain in contact with the OLA during the morphogenetic movements of head involution. After the completion of head involution, BN axon pathfinding resumes, and genetic evidence strongly suggests that axon guidance errors in BN appear not just as altered trajectories of the nerve as a whole, but also as defasciculation of the constituent axons (Schmucker et al., 1997). We examined embryos lacking both Drosophila Netrin genes (NetA,B), and found defasciculation and altered trajectories in 90–96% of embryos (Figure 1C; Table 1). The nearly complete penetrance of Netrin mutants in BN suggests they are the primary guidance cue for BN as it grows around the brain hemisphere. The closest source of Netrin expression to BN is the CNS midline and the mesoderm, suggesting that Netrins act as long-range chemoattractants for BN, much as the migrating salivary glands are guided (Kolesnikov and Beckendorf, 2005).

Figure 1. Dscam and netrin phenotypes in Bolwig’s nerve.

The axons of Bolwig’s Nerve (BN; the larval photoreceptor organ) in stage 17 embryos were visualized with monoclonal antibody 22c10, and trajectories traced digitally. The axons shown are growing along the external ventral side of the brain hemisphere. The original figure is in supplemental figure 1. (A) Wild type BN showing tightly bundled axons and growth cones (arrow) as they grow around the brain hemisphere making a characteristic turn (arrow). (B) Dscam homozygous embryo showing defasciculation of BN, and refasciculation to form a small loop (arrow). (C) NetA,B hemizygote showing extensive defasciculation of BN and refasciculation to form a large loop (arrow). (D) fra³ homozygote showing defasciculation (arrow) and separation of a growth cone from a club-like structure (arrowhead). (E) Dscam fra/+ + trans-heterozygous embryo showing defasciculation (arrow) and a club like structure (arrowhead). (F) NetA,B/+; Dscam/+ trans-heterozygote showing a large loop (arrow).

The NetA,B mutant phenotype in BN closely resembles that of Dscam mutants, which have been extensively characterized in axon guidance of BN (Schmucker et al., 2000). Phenotypes of the Netrin receptor fra in BN have not previously been described. Dscam and fra mutants both displayed similar severity of BN errors, but with much lower penetrance compared to the Netrin mutant (Table 1). Dscam fra double mutants showed an increased penetrance, but still less than that of NetA,B mutants alone. Removal of Dscam3 did not increase the penetrance of Dscam fra double mutants, suggesting additional unidentified receptors are involved in the guidance of BN. The phenotypic similarities observed for Dscam, fra and NetA,B suggested the genes might be functioning in the same pathway. To genetically test this, dosage-sensitive interactions between axon guidance genes can be informative (Bashaw et al., 2000; Schmucker et al., 2000). NetA,B heterozygotes and Dscam heterozygotes exhibit BN defects of 3.7% or 6% and 4.7% respectively (Table 1). Removing one copy of each Netrin and Dscam simultaneously (trans-heterozygote) increased the penetrance of BN defects to 34% and 38% depending on the Netrin deficiency used (Figure 1F, Table 1). This is an increase of 11.8% (NetA,B^Δ) or 13.2% (NetA,B^NP5) over the additive effects of the individual heterozygous phenotypes (p<.0001). This genetic interaction suggests that Netrins and Dscam could function together. fra and Netrins also displayed a dose-sensitive genetic interaction, showing an increase of 4.5% of defects (p<.0001). It is notable that such trans-heterozygote interactions between fra and Netrins have not been reported in the CNS. As DCC (Fra) can physically interact with other guidance receptors, we looked at Dscam fra trans-heterozygotes and found a 10.2% increase in BN defects, suggesting they act in the same pathway in vivo.

Dscams and Netrins physically interact

The individual mutant and trans-heterozygote analysis of BN strongly suggested that Dscam might be acting as a Netrin receptor. We tested whether Dscam could bind Netrins using a cell overlay assay (Keino-Masu et al, 1996). COS-7 cells were transfected with constructs expressing either Drosophila or Human Dscam genes (dDscam, hDscam). Then epitope-tagged Drosophila NetrinB (dNetB) or chick Netrin-1 (cNetrin-1; Serafini et al., 1994) proteins were incubated with the Dscam-expressing cells. Netrin proteins were detected by immunohistochemistry with antibodies directed against the myc epitope encoded by the expression vector (Figure 2). Neither Netrin bound to mock transfected COS-7 cells, but both dNetB and cNetrin-1 showed specific binding to cells expressing either dDscam or hDscam. Supernatant containing mouse Sema3A or Sema3F was used as control (Cheng et al, 2001). Neither Sema bound to dDscam, and both showed very low binding to hDscam. Cell surface expression of hDscam was confirmed with an anti-hDscam antibody (Li and Guan, 2004), and the Drosophila Dscam was detected by a V5 epitope at the carboxy terminus. Both Dscams also caused the COS-7 cells to round up and become less adherent as was previously observed for hDscam (Li and Guan, 2004). We also observed Dscam+ COS-7 cells adhering to adjacent cells, and Netrin co-localizing to the site of adhesion suggesting that the receptor and cell adhesion functions of Dscam are not mutually exclusive. Drosophila Dscam expression in COS-7 cells was greatly improved by the addition of C-terminus epitope tags, but transfection efficiency for this construct still remained low. We quantified binding of dNetB to dDscam by flow cytometry (Figure 2G); the low transfection efficiency of the dDscam construct leads to an incomplete separation of the control and experimental peaks, but clearly shows an increase in dNetB binding between control and dDscam expressing cells, confirming a physical association between dNetB and dDscam. To quantify the Netrin binding, we compared the dissociation constants (K_d) for cNetrin-1 binding to hDscam and to rDCC (Figure 2H). The Kd values were determined to be 29.1+/−2.5 nM for rDCC/netrin-myc and 35.8+/−8.6 nM for hDSCAM/netrin-myc, indicating that DCC and Dscam bind Netrin with similar affinity. We conclude that Dscams are evolutionarily conserved Netrin binding proteins.

Figure 2. Dscams and Netrins physically associate in cell overlay assays.

COS-7 cells were transfected with plasmids encoding Drosophila (A,B,C) or human (D,E,F) Dscam. The cells were incubated with medium containing secreted myc epitope tagged Drosophila Netrin B (dNetB; A,D), chick Netrin-1 (cNet-1; B,E), or chick Semaphorin A (cSemaA; C,F). After washing the cells were stained with an anti-myc antibody (red) and counterstained with DAPI (blue). (A) dNetB binding to a cell expressing dDscam. (B) cNet-1 binding to a cell expressing dDscam. (C) cSemaA does not bind to cells expressing dDscam. (D) dNetB binding to a cell expressing hDscamE. (E) cNet-1 binding to cells expressing hDscam. (F) cSemaA does not bind to cells expressing hDscam. (G) Flow cytometry quantitation of the binding of dNetB to COS-7 cells expressing dDscam (blue and green lines). Control cells not expressing dDscam and not incubated with dNetB (red line), and control cells incubated with dNetB (orange). The peaks between the controls and the samples are separate, but the shoulders overlap due to low transfection efficiency of the dDscam construct. (H) The binding of hDscam and rDCC to increasing concentrations of myc tagged Netrin were measured using a HRP conjugated secondary antibody. Scatchard analysis of the data gives dissociation constants of 29.1+/−2.5nM for rDCC/netrin-myc and 35.8+/−8.6nM for hDSCAM/netrin-myc. This difference is not statistically significant (p=0.19, t-test).

Dscams affect CNS midline axon crossing

The presence of a cytoplasmic “PYAT” motif in all four Drosophila Dscam genes first brought them to our attention, as the site is similar to the Robo CC1 “PTPYATT” Abl phosphorylation site (Bashaw et al., 2000). The non-receptor tyrosine kinase, Abl, plays an important role in axon guidance (Lanier and Gertler, 2000). Dscam protein is present on all CNS axons (Schmucker et al., 2000) and Dscam3 has a similar mRNA pattern, so we examined genetic interactions between abl, Dscam and Dscam3 for effects on CNS axons. Removal of only the zygotic component of abl leads to subtle phenotypes in CNS axons (Figure 3F). Similarly, both Dscam and Dscam3 display very subtle disorganization of the CNS axon scaffold (Figure 3D,E). However, double mutants of Dscam and zygotic abl display dramatic midline crossing defects (Figure 3H). Removal of Dscam3 and zygotic abl leads to a subtle increase in disorganization of the axon scaffold (Figure 3G; Table 2). Finally, as the Netrin receptor fra, and abl have been shown to genetically interact (Forsthoefel et al., 2005; supplemental figure 2), we examined embryos triply mutant for Dscam, fra and zygotic abl. This triple mutant has a dramatic reduction in the number of axons crossing the CNS midline (Figure 3I). A double mutant between Dscam and Dscam3 had little effect on midline crossing, although the symmetry of the axon scaffold was disrupted in an additive fashion (Figure 3J; Table 2). The strong interaction between Dscam, fra and abl prompted us to examine Dscam fra double mutants, which show a strong reduction in midline crossing (Figure 3K). Embryos lacking fra, Dscam and Dscam3 show an even stronger reduction (Figure 3L). As these double and triple mutant phenotypes are far more severe than either Netrin or fra mutants alone, they suggest that Dscams are functioning in a parallel pathway to Netrin-mediated attraction. However, this does not rule out redundancy of Dscams within multiple signaling pathways. For example, abl appears to operate in both Netrin-dependent and independent midline crossing pathways (Forsthoefel et al., 2005).

Figure 3. Dscams and abl genetically interact in CNS axon guidance.

Ventral nerve cords from stage 16 embryos stained with monoclonal antibody BP102 to visualize the CNS axon scaffold with the characteristic pattern of segmentally repeated commissures crossing the CNS midline, and linked by longitudinal tracts. Note the separation of the anterior and posterior commissures by two midline glia, which are separated by a centrally projecting axon (arrow in A). (A) Wild type CNS axon scaffold. (B) Embryos lacking NetA,B (NP5 deficiency) displaying thin and absent commissures (arrowheads), and irregular longitudinal tracts (arrow). (C) fra homozygote showing thin or absent commissures (arrowhead), and altered longitudinal tracts (arrow). (D) Dscam homozygote, which closely resembles wild type. The arrow indicates two midline glia that have not completely separated the commissures, although both are visible. The arrowhead indicates another incomplete separation. The commissures and longitudinal tracts are otherwise unaffected. (E) Dscam3 homozygote. The arrowhead indicates two midline glia that have not fully separated the commissures (subtle). (F) Embryo lacking zygotic abl function. The arrowhead indicates a failure to separate the anterior and posterior commissures. (G) Dscam3 abl homozygote, showing several segments in which the commissures have failed to separate (arrows). (H) Dscam abl homozygote, displaying severe disruption of the commissures. (I) Embryo mutant for Dscam, fra and abl showing an absence of commissures. (J) Dscam Dscam3 double mutant showing disrupted separation of commissures (arrow), and an overall irregularity to the CNS axon scaffold. (K) Dscam fra homozygote displaying greatly reduced or absent commissures, and disrupted longitudinal tracts (arrow). (L) Dscam fra Dscam3 triple mutant homozygote with very few axons crossing the midline, and significantly disrupted longitudinal axon tracts. The residual staining at the midline appears to originate from neurons whose cell bodies reside in the midline (VUMs; arrowheads).

Table 2. Quantification of midline crossing by BP102 positive commissural axons.

The anterior and posterior commissures were scored for decreased midline crossing in the homozygous or hemizygous genotypes shown. Segments were also scored for failure to correctly separate the commissures so that two midline glia were visible (“other”). The percentages for the “absent” and “thin” columns are for their relevant commissure only; the “normal” and “other” categories have been combined as a percentage for both commissures together. The number of segments scored is n. OregonR is a wild type strain. NetA,B^NP5, NetA,B^T9-B118 and NetA,B^Δ are deficiencies that delete both NetA and NetB; NetA,B^Δ deletes only the netrin genes (Brankatschk and Dickson, 2006). fra^3/4 is a protein null combination of alleles from different genetic backgrounds (Kolodziej et al., 1996). Dscam^P is a P element transposon insertion (Schmucker et al., 2000). Dscam3¹ is a piggybac transposon insertion, c02826, into Dscam3. Percentages were rounded up.

	Anterior commissure		Posterior commissure		Both commissures
Genotype	Absent	Thin	Absent	Thin	Other	Normal	n
OregonR	0%	0%	0%	0%	1%	99%	104
Ne tA,BNP5	2%	4%	8%	15%	10%	75%	131
Ne tA,BT9-B118	0%	4%	19%	20%	14%	64%	118
Ne tA,BΔ	2%	12%	2%	15%	36%	50%	110
fra3/4	2%	4%	3%	3%	8%	68%	110
fraZ965	1%	0%	4%	18%	26%	62%	114
abl4	1%	2%	1%	1%	50%	48%	110
DscamP	0%	0%	0%	0%	30%	70%	99
Dscam31	0%	0%	0%	0%	44%	56%	104
DscamP Dscam31	0%	1%	0%	0%	84%	16%	97
DscamP fra4	9%	24%	5%	39%	51%	10%	110
fra3/4 Dscam31	2%	3%	1%	9%	48%	37%	132
DscamP abl4	23%	33%	65%	26%	19%	7%	108
fra4 abl4	14%	24%	16%	24%	19%	2%	88
Dscam31 abl4	1%	0%	2%	0%	85%	13%	93
DscamP fra4	39%	51%	36%	55%	5%	5%	99
Dscam31
DscamP fra4 abl4	98%	2%	100%	0%	0%	0%	110

Analysis of identified commissural axons in Dscam mutant combinations

The anterior and posterior commissures in Drosophila are pioneered independently, in a process that is critically dependent on cell migration (Goodman and Doe, 1993). The earliest identified commissural axon known is the SP1 neuron, which crosses the midline in the anterior commissure and helps to pioneer the longitudinal connectives after crossing (Mellerick and Modica, 2002). The SP1 neuron migrates towards the midline, to occupy positions just anterior to the future anterior commissure; the SP1 axons cross the midline, and grow around the contralateral cell body to join a longitudinal pathway (Figure 4). We used the anti-Connectin antibody C1.427 to visualize the SP1 cell body and axon (Meadows et al., 1994), in various combinations of Dscam and fra (Figure 4). These results confirm those obtained with BP102 staining, in that Dscams enhance the fra commissural axon phenotype (Figure 4G). In addition, in Dscam fra Dscam3 triple mutants, most SP1 axons cross the midline, suggesting the existence of additional attractive receptors. We also analyzed the effect of the Dscam fra double mutant on egl positive commissural axons, which cross the midline midway through axonogenesis. fra mutants have no effect on the EG axon bundle (Garbe and Bashaw, 2007), but Dscam fra double mutants lead to 37% of EG axon bundles being thin or absent (n = 112 segments). We chose to focus on SP1 axons as Figure 4 clearly shows how follower axons can fasciculate with existing pathways. There are two examples of egl axons encountering a missing commissure and rerouting to the nearest intact commissure (Figure 4J,K). This emphasizes the importance of analyzing pioneer axons to understand axon guidance phenotypes. Interestingly, SP1 cell body migration defects present in fra mutants are not enhanced by Dscam (Figure 4H). However, we found that Dscam enhances fra cell migration defects in the salivary glands (Supplemental Figure 3), to a penetrance greater than that seen in netrin mutants (Kolesnikov and Beckendorf, 2005). The predominant phenotype observed was stalling of migration, or kinking of the gland reflecting growth in alternating directions. Dscams therefore appear to be important for axon guidance and cell migration in different tissues.

Figure 4. Analysis of identified commissural axons in Dscam mutant combinations.

Embryos stained with anti-Connectin antibody (A–F) or the egl-GAL4; UAStaulacZ reporter (G–I). Connectin staining highlights the SP1 cell body (arrows) and commissural axon which crosses the midline in the anterior commissure (arrowheads). All embryos are late stage 15 or early 16, with anterior to the left. (A) Wild type embryo. Note the regular position of the SP1 cell bodies (arrows) and the commissural axons crossing the midline in the anterior commissure (arrowhead). (B) fra mutant showing altered SP1 cell body position (arrow), but intact SP1 axon (arrowhead). The posterior commissure is clearly disrupted in the same segment, as is the adjacent longitudinal. (C) NetA,B mutant showing normal and disrupted SP1 cell body migration (arrows) and missing commissural axons (arrowhead). (D) Dscam mutant embryo, which is nearly identical to wild type, except for some Connectin positive axons inappropriately crossing the midline between the two commissures (asterisk). (E) Dscam fra double mutant displaying missing commissural axons (arrowhead) between the two SP1 cell bodies (arrows; one slightly out of focus). (F) Dscam fra Dscam3 triple mutant showing missing commissural axons between mispositioned SP1 cell bodies (arrows). (G,H) Graphs of SP1 axon midline crossing defects (G) and SP1 cell body migration defects (H) in mutant combinations. Abdominal segments one through seven were scored for ten embryos for each genotype. Statistical significance was assessed by fitting the data to a Poisson distribution. (G) SP1 axon defects. Dscam fra Dscam3 embryos are statistically different from fra mutant embryos (p=0.013), indicating that Dscams enhance the fra SP1 axon phenotype. Dscam fra Dscam3 embryos were statistically different from all other mutants (p<0.01), except Dscam fra. Dscam fra embryos were statistically different from OregonR, NetA,B and Dscam embryos (p<0.05), but were not found to be significantly different from fra alone. (H) Graph of SP1 cell body migration defects. The Dscam and OregonR genotypes are statistically different from all other genotypes except each other (p<0.05). There is no statistical difference between the NetA,B, fra, Dscam fra and Dscam fra Dscam3 phenotypes, indicating that Dscam and Dscam3 do not enhance the fra SP1 cell migration defect. (I) Wild type embryo showing expression of the egl-GAL4;UAS-tau-lacZ transgene. The EW axons cross the midline in the posterior commissure (arrowhead), and the EG cross in the anterior commissure (arrow). (J,K) Two focal planes of one embryo, mutant for Dscam fra. The arrows indicate an EG axon bundle, crossing the midline in an adjacent posterior commissure in response to a missing anterior commissure. The arrowheads indicate EW axons either failing to cross the midline (top) or crossing the midline in an adjacent posterior commissure in response to an absent anterior commissure. The egl-GAL4;UAS-tau-lacZ transgenes do not switch on lacZ in all egl-positive cells in each segment; we have found no evidence of altered cell fate, or missing cells in Dscam mutant combinations.

Dscam over-expression induces ectopic midline crossing

As Dscam mutants cause a failure to cross the midline in combination with other mutations, we asked whether Dscam over-expression could cause axons to be attracted to the midline inappropriately. Pan-neuronal over-expression of Dscam with either sca-GAL4 or elav-GAL4 using the UAS-GAL4 system (Brand and Perrimon, 1993) did not yield detectable phenotypes. However, over-expression of Dscam by the fushi tarazu neurogenic (ftz_ng) promoter, which drives expression in longitudinal axons that do not normally cross the midline (Lin et al., 1994), resulted in ectopic axon crossing of the midline (Figure 5B). Interestingly, the induced crossing was not suppressed by a NetA,B deficiency (Figure 5D,F), suggesting that either Dscam is not responding to Netrins from the midline, or that Dscam can respond to more than one midline-derived attractive cue. Over-expression of multiple copies of each transgene leads to more severe phenotypes (Figure 5E,H). Surprisingly, the initial projection of the pCC axon, which pioneers the medial FasII positive longitudinal affected by Dscam over-expression, was unaffected (n=220 axons, stages 12 and 13, one copy of ftz_ng-GAL4 and Dscam). The complexity of the anti-FasII pattern after stage 13 makes it difficult to monitor the pCC axon projection, so we utilized RN-GAL4 combined with a UAS-tau-lacZ reporter, which expresses specifically in pCC (and also two motor neurons; Fujioka et al., 2003). Over-expression of Dscam by RN-GAL4 had no effect on the pCC axon trajectory (n=242 axons) when assayed at stages 16/17 (Supplemental Figure 4). pCC does not express fra (Hiramoto and Hiromi, 2005), so we simultaneously over-expressed Dscam and fra using RN-GAL4, but the pCC axon did not cross the midline (n=211 axons). The results suggest that additional components will be required to generate a phenotype in pCC (see discussion).

Figure 5. Dscam misexpression phenotypes.

Stage 17 embryos stained with anti-FasII MAb 1D4 (A–E), and stage 16 embryos stained with MAb BP102 (G,H). (A) Wild type embryo in which FasII positive longitudinal fascicles do not cross the midline. (B) Embryo carrying one copy each of the ftz_ngGAL4 and UAS-Dscam1-30-30-2 transgenes showing a FasII positive fascicle crossing the midline (arrowhead), and also stalled axons (arrow). (C) Embryo mutant for NetA,B and carrying the ftz_ng-GAL4 transgene, showing a collapse of the longitudinal fascicles into one fascicle (arrow) but no ectopic midline crossing. (D) Embryo mutant for NetA,B and carrying one copy of the ftz_ng-GAL4 and UAS-Dscam1-30-30-2 transgenes, in which thick (arrowhead) and thin (arrow) fascicles can be seen ectopically crossing the midline. The staining below the thick fascicle is a combination of staining in the transverse nerve and residual staining in cell bodies above and below the plane of focus of the longitudinal axons respectively. (E) Embryo carrying two copies each (based on proportion of embryos displaying this phenotype) of the ftz_ngGAL4 and UAS-Dscam1-30-30-2 transgenes showing disrupted longitudinal fascicles (arrows) and ectopic axon crossing of the CNS midline (arrowhead). (F) Graph quantifying midline crossing by FasII positive axons in ftz_ngGAL4 UAS-Dscam1-30-30-2, NetA,B ftz_ngGAL4 UAS-Dscam1-30-30-2 and NetA,B ftz_ngGAL4 embryos. There is no statistical difference in the amount of midline crossing observed when NetA,B is deleted (Poisson regression; n=15 for each category). (G) Wild type embryo stained with MAb BP102. (H) Embryo carrying two copies of the ftz_ngGAL4 and UAS-Dscam1-30-30-2 transgenes showing disrupted separation of the anterior and posterior commissures (arrowhead), and disrupted longitudinal axon tracts (arrow).

Dscam expression on mouse commissural axons

In situ hybridization studies have previously shown that although both Dscam and the related DscamL1 mRNAs have distinct expression patterns in the developing spinal cord, both expressed in the region where commissural neuron cell bodies lie (Yamakawa et al., 1998; Agarwala et al., 2001; Barlow et al., 2002). The evolutionarily conserved binding of Netrins and Dscams suggested that Dscam might be important for midline crossing in vertebrates. To test whether Dscam could function as a Netrin receptor in vertebrates, we examined the distribution of mouse Dscam protein at the time of commissural axon growth in mice. Dscam protein was found to be present on axons known to be Netrin responsive (Shirasaki et al., 1996) in the hindbrain at E10.5, and on a subset of axons crossing the floorplate at E11.5 in the spinal cord (Figure 6). There is expression in other regions, including DRG neurons, correlating with the Dscam mRNA expression pattern. The localization of Dscam protein to commissural axons in the mouse is consistent with an evolutionarily conserved role for Dscam as a Netrin receptor.

Figure 6. DSCAM is expressed by commissural axons in the mouse CNS.

Sections of embryos were labeled with antibodies against DSCAM and neuron-specific βIII-tubulin. (A–C) DSCAM is expressed in many neurons in the developing spinal cord. The protein is also detected in commissural axons. (A) Anti-DSCAM antibody labeling of E11.5 spinal cord, including motor column (MN), and commissural axons crossing the floor plate (FP). The dorsal root ganglia (DRG) are labeled as well. (B) The pan-neuronal β-Tubulin (β-IIITub) labeling of the same section shows the location of neurons. (C) Merge of A and B. DSCAM localizes to an inner subset of the commissural fibers (yellow, arrow). (D and E) DSCAM labeling of sections of rhombomere 1 of the E10.5 hindbrain. (D) Both neuron cell bodies and axons are labeled by anti-DSCAM. (E) A close-up of the boxed region in D shows strong DSCAM signal in the hindbrain commissure. Scale Bars: 100 µm in A and 50 µm in D.

Discussion

It is known that axons are attracted to the CNS midline by Netrin proteins signaling through the DCC family of receptors. We have identified Dscam cell adhesion molecules as required for attraction to the midline in Drosophila. This was unexpected, as the role of Dscams in mediating repulsion between neurites has been well characterized (Zipursky et al., 2006). In vertebrate retina formation, Dscams have been shown to mediate attraction (Yamagata and Sanes, 2008) and repulsion (Fuerst et al., 2008), suggesting that like other axon guidance molecules, the output of Dscams may be dependent on context, such as the presence or absence of co-receptors or downstream effectors. In vitro, Netrins bind Dscams in an evolutionarily conserved manner, with a similar affinity to the known Netrin receptor, DCC. In vivo, Dscam and the netrins genetically interact in a dose-sensitive manner in BN axon guidance. These data suggest that Dscam is a receptor required to mediate attraction to Netrin diffusing from the CNS midline and mesoderm.

The dramatic reduction in number of midline crossing axons in Dscam fra and Dscam fra Dscam3 mutant combinations, with defects significantly greater than seen in NetA,B mutants, suggests that Dscam responds to additional ligands as well as Netrin. In vertebrates, sonic hedgehog (Shh) has been identified as a Netrin-1 independent chemoattractant for commissural axons (Charron et al., 2003). Drosophila hedgehog (hh) is required for the specification of the midline glia that commissural axons grow towards, but does not appear to be acting as a chemoattractant (Hummel et al., 1999b). Therefore, the missing midline chemoattractant(s) still remains to be identified, but we predict it will bind to Dscams. We favor the hypothesis that Dscams participate in both Netrin-dependent and independent pathways at the CNS midline.

A receptor role for Dscam was not entirely unexpected, as several groups had proposed the existence of a core molecular function for Dscam, independent of the diversity of protein isoforms (Wang et al., 2004; Zhan et al., 2004; Chen et al., 2006). Each group attempted to rescue Dscam mutant phenotypes by transgenic expression of a single Dscam isoform, and achieved partial rescue of their phenotype of interest. (For a discussion of the role of diversity in Dscam function, see Zipursky et al., 2006). Based on rescue of adult mechanosensory neurons, Chen et al. (2006) hypothesized that the core function depends on a receptor-ligand interaction that may not engage the variable domains of the protein. Netrin binding could fulfill this role. As Netrin binding is evolutionarily conserved, we think that Netrin reception is likely to be independent of the Dscam diversity only found in insects. The ability of Dscam to mediate homophilic cell adhesion is also evolutionarily conserved (Agarwala et al., 2000; Wojtowicz et al., 2007; Yamagata and Sanes, 2008). Our results place Dscam in the small group of cell surface molecules that can act as both receptors and mediate cell adhesion, as exemplified by the neural cell adhesion molecule NCAM (Paratcha et al., 2003).

Molecular mechanisms of Dscam function

DCC is converted from an attractive receptor for Netrin to a repulsive receptor by heterodimerization with Unc5 (reviewed in Moore et al., 2007). This suggests that Dscam could be converted to an attractive receptor by interaction with other receptors. This model is supported by the trans-heterozygous interaction between Dscam and fra in BN, which places them in the same genetic pathway. Dscam is clearly not essential for Netrin signaling, in part because there are no Dscam homologues in C. elegans. However, utilizing two different attractive Netrin receptors increases the potential complexity of responses to Netrin. Dscams may also respond to additional cues, either alone, or in combination with other receptors. The ectopic crossing induced by Dscam over-expression is significant, as fra over-expression alone is not sufficient to induce midline crossing (Dorsten et al., 2007). It is possible that Fra functions in a complex, and may require simultaneous over-expression of multiple components to generate a phenotype, as has been seen for γ-secretase activity (Edbauer et al., 2003). Interestingly, over-expression of fra lacking the entire cytoplasmic domain (fraΔC) can produce completely commissureless phenotypes, which are much stronger than fra mutant phenotypes (Garbe et al., 2007), suggesting that fraΔC could be inactivating a complex that responds to multiple attractive cues from the midline.

Organogenesis

Our phenotypic analysis of salivary glands demonstrates that Dscam can function in cell migration as well as axon guidance. Human patients with Down Syndrome (DS) frequently suffer from congenital heart disease, particularly atrioventricular septal defects (Yamakawa et al., 1998; Barlow et al., 2001). The developmental origin of these defects is probably disruption of directed growth of endocardial cushions, and Dscam is an excellent candidate to play a role in this process, especially if it can function as a receptor to guide migration. Dscam has also been associated with the mental retardation component of DS; Dscam and the related DscamL1 gene are expressed in the cortex and cerebellum (Agarwala et al., 2001; Barlow et al., 2002) suggesting that Dscam may function in neuronal and axonal migrations during brain development. Dscam may be responsible for the abnormalities in intestinal formation and the associated enteric neuron defects seen in DS patients (Yamakawa et al., 2001). It should be noted that we have been analyzing loss of function phenotypes, and DS arises due to an additional copy of chromosome 21. Three copies of Dscam may increase basal levels of cell or axon migration, or may stimulate enhanced responses to external signals in any of the processes described above. We found it quite difficult to generate over-expression phenotypes of Dscam, suggesting that other components may be required for function, and this may apply to DS patients too, limiting the tissues in which trisomy for Dscam could have an effect.

Conclusion

The evolutionary conservation of Dscams may be explained by their dual ability to act as cell adhesion molecules and as receptors. Dscams provide a new starting point for fully understanding Netrin signaling, as well as for identification of additional axon guidance cues. Dscam-associated phenotypes in DS can be reevaluated as a consequence of altered cell migration and axon guidance.