Differential Expression of DSCAM Guides the Patterning of Retinal Axons Along Their Path and at Their Target in the Developing Xenopus Visual System

Background The Xenopus retinotectal circuit is organized topographically, where the dorsal-ventral axis of the retina maps respectively on to the ventral-dorsal axis of the tectum; axons from the nasal-temporal axis of the retina project respectively to the caudal-rostral axis of the tectum. Studies throughout the last two decades have shown that mechanisms involving molecular recognition of proper termination domains are at work guiding topographic organization. Such studies have shown that graded distribution of molecular cues is important for topographic mapping. However, the molecular cues organizing topography along the developing optic nerve, and as retinal axons cross the chiasm and navigate towards their target in the tectum, remain unknown. Down syndrome cell adhesion molecule (DSCAM) has been characterized as a key molecule in axon guidance, making it a strong candidate involved in the topographic organization of retinal bers along the optic path. Using a combination of clearing and staining techniques we characterized DSCAM expression and the projection of and dorsal retinal bers starting from the eye, followed to the into and into the in the in Xenopus laevis tadpoles. We also assessed the effects of on the establishment of retinotopic maps through spatially and temporally targeted DSCAM knockdown on retinal ganglion cells (RGCs) with axons innervating the optic tectum. plane of the tectal neuropil independent of its effect on axon branching. We examined effects of DSCAM on the same population of RGCs (ventral) at a distinct time of development by timing the DSCAM MO knockdown by targeting the electroporation of tadpoles at different stages. We found that downregulating DSCAM expression in ventral RGCs with axons already terminating medially within the tectum caused an aberrant shift and extension of their terminal arbors away from those of dorsal RGCs, suggesting that DSCAM guides remodeling and topographic organization of arbors derived from ventral RGCs. During zebrash development (which closely resembles Xenopus development), dorsal retinal bers normally reach the optic tectum via the lateral branch, while ventral axons project via the medial branch Disrupting mechanisms dependent on RNA-binding proteins, such as Hermes, causes an aberrant shift in topographic ordering and results in lateral dorsal axons projecting ectopically into the medial branch arbor Thus, our studies indicate that DSCAM, together with other signaling molecules including Hermes, participate in the medio-lateral topographic mapping at the target.


Abstract Background
The Xenopus retinotectal circuit is organized topographically, where the dorsal-ventral axis of the retina maps respectively on to the ventral-dorsal axis of the tectum; axons from the nasal-temporal axis of the retina project respectively to the caudal-rostral axis of the tectum. Studies throughout the last two decades have shown that mechanisms involving molecular recognition of proper termination domains are at work guiding topographic organization. Such studies have shown that graded distribution of molecular cues is important for topographic mapping. However, the molecular cues organizing topography along the developing optic nerve, and as retinal axons cross the chiasm and navigate towards their target in the tectum, remain unknown. Down syndrome cell adhesion molecule (DSCAM) has been characterized as a key molecule in axon guidance, making it a strong candidate involved in the topographic organization of retinal bers along the optic path.

Methods
Using a combination of whole-brain clearing and immunohistochemistry staining techniques we characterized DSCAM expression and the projection of ventral and dorsal retinal bers starting from the eye, followed to the optic nerve into the chiasm, and into the terminal target in the optic tectum in Xenopus laevis tadpoles. We also assessed the effects of DSCAM on the establishment of retinotopic maps through spatially and temporally targeted DSCAM knockdown on retinal ganglion cells (RGCs) with axons innervating the optic tectum.

Results
Highest expression of DSCAM was localized to the ventral posterior region of the optic nerve and chiasm; this expression pattern coincides with ventral bers derived from ventral RGCs. Downregulating DSCAM levels affected the segregation and proper sorting of medial axon bers, derived from ventral RGCs, within the tectal neuropil, indicating that DSCAM plays a role in retinotopic organization.

Conclusion
These ndings together with the observation that DSCAM immunoreactivity accumulates on the primary dendrites of tectal neurons indicates that DSCAM exerts multiple roles in coordinating retinotopic order and connectivity in the developing vertebrate visual system.

Background
During embryonic eye development, connections from the retina to the brain are carefully arranged in a preserved spatial manner that creates a topographic map of the visual world. In the amphibian visual system, retinal ganglion cell (RGC) axons project to the tectum in a manner that mirrors the relative positioning of RGCs across the retina -effectively constructing a point-to-point representation of visual space in the brain (1)(2)(3). The formation of precise topographic maps requires active molecular cues guiding speci c axon targeting and establishing selective synaptic connections. For example, in the developing embryonic Xenopus visual system, dorsal retinal axons expressing high levels of Ephrin-B ligands speci cally target ventral tectal regions with high EphB receptor expression via an attractive guidance mechanism (4). Such studies demonstrate that molecular recognition of proper termination domains, often organized in matching gradient distribution, are important for topographically organizing neuronal circuits during development. Likewise, in mouse models, topographic mapping of retinal axons along the anterior-posterior axis of the superior colliculus (equivalent to the tectum in lower vertebrates) relies heavily on repulsive-mediated signaling between EphA receptors and their Ephrin-A ligands (5)(6)(7).
Disrupting the signaling gradient either by knocking out the receptor or the ligand affects topographic ordering, but not entirely (5)(6)(7). Disruption of ephrin signaling, only to a certain extent, shifts axonal bers posteriorly and others anteriorly (8). Furthermore, prior to reaching the tectum, retinal axon bers are already topographically sorted along the optic nerve where graded ephrin signaling has not been reported (9)(10)(11)(12)(13). These ndings suggest that ephrin signaling does not exclusively shape topography and that additional key molecules are involved. The molecular cues organizing topography along the developing optic nerve and as retinal axons cross the chiasm, remain unknown.
Down Syndrome Cell Adhesion Molecule (DSCAM) has been implicated in multiple aspects of neural circuit development, modulating dendrite and axon growth in both the vertebrate and invertebrate nervous systems (14). A speci c role for DSCAM in axon growth, fasciculation and guidance is supported by a number of studies (15)(16)(17)(18)(19), but whether the molecule is involved in the topographic organization of retinal bers had yet to be investigated. Multiple studies have con rmed that DSCAM is expressed by RGCs and in retinal projections along the developing mouse optic nerve (15,(20)(21)(22). Erskine and colleagues showed that DSCAM knock out disrupted the timing at which mouse retinal axons arrived at the thalamus, suggesting that DSCAM acts as a permissive signal and mediates growth-promoting interactions that help facilitate retinal axon growth towards their target (15). DSCAM was also shown to be involved in segregating contralateral retinal projections from ipsilateral bers in the dLGN (20). While these studies did not directly test DSCAM's involvement in organizing retinal topography, they indicate that DSCAM may contribute to the speci city of axonal wiring within the target. Previous work from our laboratory showed that in Xenopus, DSCAM acts as a permissive signal that facilitates axon arbor growth once RGC axons reach their target in the optic tectum. Here, we used the Xenopus tadpole visual system to further examine potential roles for DSCAM in establishing retinotopic order as axons travel towards and establish synaptic connections at their target. We observed differential DSCAM expression along the ventral and posterior regions of the optic nerve and chiasma, indicating that a subpopulation of retinal bers express DSCAM as they navigate the optic path. By tracing the projection of ventral and dorsal retinal bers as they exit the eye into the optic nerve and chiasm in xed tadpoles, and imaging in real time retinal axon arbors in the Xenopus optic tectum with altered DSCAM expression, we provide evidence that DSCAM affects the segregation and proper sorting of axon bers derived from ventral RGCs, both along the optic nerve and within the tectal neuropil, indicating that DSCAM plays a role in retinotopic Samples were transferred into a 2 mL glass vial and were blocked for 4 hrs at room temp. To visualize both DSCAM immunoreactivity and axon bundles in the tadpole's head, tissues were incubated in DSCAM rabbit polyclonal (1:500; Aviva System) and 3A10 mouse anti-neuro lament-associated protein antibody (1:500; Developmental Studies Hybridoma Bank) in 10% DMSO, 1% Triton X-100 in 1x PBST. Goat antirabbit Alexa 568 and goat anti-mouse Alexa 488 antibodies (both at 1:500; Invitrogen) were used as secondary antibodies, respectively. To reduce light scattering throughout brain and head tissues, samples were submerged in a fructose-based high-refractive index solution (1.45) at room temp overnight.
Cleared samples were imaged using a LSM780 confocal microscope (Zeiss).

Labeling Retinal Ganglion Cell Axons
To visualize retinotopic organization, ventral and dorsal RGCs axons were labeled by direct retinal electroporation following a similar protocol developed by Haas and colleagues (25). Tadpoles (27)) together with the anti-neuro lament protein antibody (Mab 3A10) that has been shown to label a subset of retinal axons (28). As observed on coronal tissue sections, DSCAM was highly expressed at the ventral region of the optic nerve (Fig 1a, b). The mean uorescence intensity of both DSCAM and 3A10 immunoreactivity were measured across the ventral-dorsal axis of the optic nerve bundle (measurements started at the ventral side of the optic nerve then continued to the dorsal side as marked by the white dotted line on Fig 1b). DSCAM immuno uorescent signal progressively decreased as measurements were obtained along the dorsal regions of the optic nerve. We plotted the mean uorescence intensity on a graph with the x-axis representing regions along the ventral to dorsal portions of the optic nerve (Fig 1d). This analysis quantitatively con rmed a high-ventral to low-dorsal graded pattern of DSCAM expression. In contrast, 3A10 labeled axons were strongly localized to the dorsal region of the optic nerve (border noted by white dotted line on Fig 1b) with lower intensity signal along the ventral side. 3A10 immuno uorescence exhibited a low-ventral to high-dorsal distribution (Fig 1d) in what appeared to be an inverse of the highventral to low-dorsal DSCAM immunoreactivity pattern. Analysis of retinal axons as they crossed the midline at the optic chiasm (white arrowhead, Fig 1c) showed that DSCAM immunoreactivity strongly localized at the ventral base of the chiasm. Here again, the intensity of DSCAM immunostaining was gradually decreased at more dorsal areas of the optic chiasm (white dotted line, Fig 1c, e). 3A10 immunoreactivity in axonal bers continued to show a low-ventral to high-dorsal uorescence intensity along the optic chiasm (Fig 1c, e).
Having encountered differential DSCAM expression along the ventrodorsal axis of the optic nerve and chiasm, we further characterized the distribution of DSCAM along the posteroanterior axis. In horizontal tissue sections, we found higher DSCAM immuno uorescence intensity speci cally at the posterior region of the optic nerve bundle (Fig 2a) and optic chiasm (white arrowhead, Fig 2b). DSCAM immuno uorescence intensity was lower at the anterior portions of the optic nerve and chiasm compared to the posterior side, indicating a high-posterior to low-anterior pattern of DSCAM expression (Fig 2 c, d). Immunostaining with the 3A10 antibody showed an inverse pattern, where the uorescence distribution of 3A10 staining appeared lower posterior and higher anterior at both the optic nerve and chiasm (Fig 2 ad). It is important to note that we did observe DSCAM expression along a number of bers stained with 3A10 (white arrows , Fig 2a), con rming that the DSCAM immunostaining identi ed RGC axon bers.
These results suggest that a subpopulation of RGC axon bers rely on DSCAM as a potential mechanism to navigate the optic nerve pathway and cross the optic chiasm.
To further determine if axonal arbors terminating and branching in the optic tectum express DSCAM in a pattern similar to the high-ventral to low-dorsal pattern found along the optic nerve and chiasm, we performed immunostaining of whole brain cleared tissues to preserve the structural layout of axonal tracts and arbors innervating the tectum. Compared to brain sectioning, brain clearing is a powerful technique that permits obtaining a novel three-dimensional perspective of any potential gradient pattern of DSCAM expression within the intact tectum (24). Cleared tissue samples of stage 45 to 46 tadpoles were immunostained with DSCAM and 3A10 antibodies (Fig 3a). Our results revealed that the optic nerve (solid white and yellow arrowhead, Fig 3a), as well as sensory and motor cranial nerves throughout the tadpole head were stained by the 3A10 antibody (Fig 3a). Even at low magni cation, we were able to observe DSCAM immunoreactivity along the optic nerve throughout its full extent up to the optic chiasm (solid white and yellow arrowhead, Fig 3a, b). At the optic tectal neuropil, strongest DSCAM immunoreactivity coincided with axon terminals labeled with the 3A10 antibody (empty white and yellow arrowheads at the tectum, Fig 3a). When we examined individual horizontal z-stacks at the level of the neuropil, the uorescence intensity of DSCAM immunoreactivity was higher along axon terminals as revealed by the 3A10 co-immunostaining (solid white and yellow arrowheads, Fig 3b) and lower at fasciculated axon bundle tracts as they enter the neuropil (empty white and yellow arrowheads, Fig 3b). To better outline DSCAM expression patterns, whole intact brain tissues were dissected, cleared, and immunostained with the DSCAM and 3A10 antibodies. When viewing the orthogonal planes of the confocal images of cleared brain tissues, we found that DSCAM immunoreactivity colocalized with the 3A10 retinal axon marker both at the x-and y-orthogonal planes (solid white and yellow arrows, Fig 3c). The mean average uorescent intensity of DSCAM expression along axon terminals were more substantially elevated compared to the DSCAM signal along axon ber tracts (Fig 3d).
It is important to note that DSCAM expression along the optic tectum is distinct compared to other brain regions of the Xenopus CNS. Noticeably, strong uniform DSCAM expression was found in the olfactory bulb and telencephalic regions of the tadpole brain (white and yellow arrowhead at OB and TEL, Fig 4a, b) where spiny neurons are located (29). A prominent pattern of DSCAM immunoreactivity was detected along axonal tracts at the lateral regions of the hindbrain (white and yellow arrowhead at the HB, Fig 4a, d). We quantitatively measured the mean uorescent intensity of DSCAM of two intact tadpole heads at various regions of the Xenopus brain. DSCAM immunoreactivity at the olfactory bulb (Fig 4a, b, e) was higher in intensity than in the telencephalic brain region, the hindbrain, and the optic tectum. Similar to the patterns observed in dissected brain tissues (Fig 3c), the intensity of DSCAM immunostaining was higher in RGC axons at their axon terminals and lower at axonal tracts entering the neuropil (Fig 4c, e). These results suggest that DSCAM is differentially expressed or transported throughout the CNS.
To determine whether the patterns of DSCAM expression in the optic nerve, chiasma and in the optic tectum correspond with differential DSCAM expression within the retina, we analyzed retinas of stage 45 tadpoles in cleared intact tissues and cryostat sections immunostained with DSCAM and the 3A10 antibodies (Fig 5a, b). As previously shown, DSCAM immunoreactivity was observed in the ganglion cell layer (GCL), inner plexiform layer (IPL) and inner nuclear layer (INL) of the Xenopus retina (27), with punctate DSCAM expression found around cell bodies within the GCL (Fig 5b). The majority of cell bodies immunostained with the 3A10 antibody localized to the GCL, adjacent to the IPL. However, as observed both in coronal sections and in cleared intact eyes, not all RGCs were immunopositive for 3A10 (Fig 5 a,  c) indicating that only a subset of RGCs express neuro lament-associated proteins recognized by the 3A10 antibody. Analysis of coronal sections of retinas co-stained with DSCAM and 3A10 antibodies showed that some axon bers exiting the eye along the optic ber layer (Fig 5d) and the optic nerve head were immunopositive for both DSCAM and 3A10 (Fig 5e; empty arrowheads), although a number of 3A10 positive bers did not express DSCAM (white arrowheads; Figs 5 d, e). These observations reveal a differential pattern of expression of DSCAM by RGC axons as they exit the eye. Moreover, different subsets of RGCs, including those that differentially express DSCAM and neuro lament-associated proteins recognized by the 3A10 antibody, appear to organize in distinct topographic order as they navigate along their pathway to their target in the optic tectum.

Dorsoventral Axon Sorting in the Xenopus Retinotectal System and DSCAM Effects on Topographic Segregation at the Optic Tectum
A graded distribution of molecular cues has largely been implicated in topographic mapping. Based on its differential distribution, it is likely that DSCAM collaborates with other guidance and cell adhesion molecules in the topographic organization of axon retinal bers at multiple points along their path (30). Indeed, analysis of a mouse model of Down syndrome showed that DSCAM regulates eye-speci c segregation of retinogeniculate projections at the target, in the dorsal lateral geniculate nucleus (20). Thus, to explore whether DSCAM is directly involved in retinotopic organization in the Xenopus optic tectum, we rst characterized the projection and ordering of ventral and dorsal retinal bers as they travel from the eye through the chiasm and into the brain (as depicted schematically in Fig 6a). A scrambled control uorescein-tagged MO (to serve as a green uorescent marker) and a control lissamine-tagged MO (red uorescent marker) were electroporated separately to label dorsal and ventral RGCs, respectively (Fig 6b, c). Our results show that ventral RGCs project axon bers that are positioned along the ventral portion of the optic nerve, while dorsal RGCs send axon bers along the dorsal region of the optic nerve (Fig 6c, d, e). As axons of both ventral and dorsal RGCs enter and cross the chiasm and turn contralaterally into the tectum, we observed a shifting of ber arrangement, with lissamine MO-labeled axon bers that were originally positioned on the ventral side of the optic nerve intermixing and positioning more dorsally in the optic chiasm (Fig 6d, f). This inverted projection was also observed for the uorescein MO-labeled axon bers that originate in the dorsal portion of the retina, shifting more ventrally (Fig 6f). A complete inverted arrangement was observed for axons as they innervate the tectum, with ventral RGC axons entering the tectum through the dorsal branch and dorsal RGC axons projecting ventrally within the tectum (Fig 6g) in agreement with previous studies (31,32). Based on these results, our immunohistochemical data indicates that speci c DSCAM expression along the ventral portion of the optic nerve would coincide with axon bers traveling on the ventral side of the optic nerve pathway prior to crossing at the tectum (Fig 1a and Fig 3b).
Analysis of axon terminals along the lateral-medial axis (as depicted schematically in Fig 7a), showed that ventral RGC axons (labeled with lissamine-tagged MO) innervate the tectum laterally, while dorsal RGC axons (labeled with Alexa 488 dextran) travel more medially, as shown for other species (9,33). Indeed, ventral and dorsal RGC axons from tadpoles injected at stage 46 and imaged 48 hrs later showed correct topographic mapping but with a consistent degree of arbor overlap (average 20 μm) as shown in Fig 7b. When retinal neurons were labeled few days later, at stage 47, and imaged 48 hrs after, medial arbors were visibly separated from lateral arbors (data not shown). This separation between lateral and medial arbors in the Xenopus tadpole is consistent with observations in zebra sh larvae at 5 days postfertilization, when the optic tectum is rst fully innervated (9). Thus these in vivo imaging studies con rm that in Xenopus, dorsal RGC axons projecting through the lateral branch initially overlap with ventral RGC axons traveling through the medial branch; then, as the tectum expands and arbors become more complex, laterally and medially projecting arbors remodel and clearly separate along the Xenopus neuropil.
To identify speci c cellular actions of DSCAM in directing retinotopy in the tectum, we targeted the population of RGCs that preferentially express DSCAM to manipulate its expression when a majority of axons have already arborized in the tectum, but when medially and laterally projecting axons still overlap. For this, we electroporated a morpholino (MO) targeting Xenopus laevis Dscam mRNA to block translation and downregulate endogenous DSCAM levels in axons of ventral RGCs in tadpoles at stage 46, while also labeling dorsal axons with Alexa 488 dextran. This strategy allowed us to manipulate and visualize the innervation patterns and topographic organization of axon arbors in the neuropil rather than interfere with axon path nding or initial axon branching (27). As shown for control tadpoles, axons derived from ventral and dorsal RGCs were correctly sorted along the medial-lateral axis (Fig 7a, b), with ventral RGC axons predominantly arborizing in the medial portion of the neuropil and dorsal RGCs axons arborizing laterally. However, 48 hours after DSCAM MO injection, ventral RGC axon arbors were positioned more medially compared to controls (Fig 7b). To quantify this effect, we measured the area occupied by the axon arbors within the tectal neuropil; total arbor width from ventral RGCs injected with DSCAM MO was compared to that from ventral RGCs in tadpoles injected with control MO. The average arbor spread of axons positioned medially in tadpoles with DSCAM MO knockdown was signi cantly larger than controls (Controls 117.5 ± 9.73 μm, n = 6; DSCAM MO 150.2 ± 7.81 μm, n = 6, p = 0.0257, Fig  7c). Dorsal RGC axons labeled with Alexa 488 dextran projecting laterally within the tectum in either control MO or DSCAM MO treated tadpoles had the same arbor width independent of ventral RGC treatment (Controls 75.03 ± 8.723 μm, n = 6; DSCAM MO 78.85 ± 2.95 μm, n = 6, p = 0.6875, not signi cant, Fig 7d). As shown above, in stage 46 tadpoles there is a degree of overlap between mediallyprojecting ventral RGC axons (red marker) and laterally-projecting dorsal RGC axons (green marker) within the tectal neuropil (Fig 7b, e). At this stage, axons from ventral RGCs with DSCAM MO knockdown showed a signi cant reduction in lateral and medial arbor overlap compared to arbors in control MO treated tadpoles (Controls 22.02 ± 3.915 μm, n = 6; DSCAM MO 9.185 ± 2.193 μm, n = 6, *p = 0.0169, Fig  7e). When measuring the entire spread of the arbors along the medial to lateral axis for each group, tadpoles treated with DSCAM MO had RGC axons that occupied a larger territory within the neuropil (Controls 117.5 ± 9.733 μm width, n = 6; DSCAM MO 150.2 ± 7.808 μm, n = 6, *p = 0.0257, Fig  7f). When expressed as percentage overlap per total arbor territory in width the reduction in overlap for ventral axons with DSCAM MO knockdown was also signi cantly different from controls (Controls 18.82 ± 2.685 %, n = 6; DSCAM MO 6.135 ± 1.482 %, n = 6, **p = 0.0019, Fig 7g). Together, these ndings suggest that changes in DSCAM expression in ventral RGC axons affect their projection patterns acting at the target where an increase in segregation of medial and lateral axons is observed in response to lowered endogenous DSCAM levels.

Dendritic Localization of DSCAM in Post-synaptic Tectal Neurons
Our previous work showed that downregulation of DSCAM expression in single RGCs interferes with axon growth and branching at the target, indicating that endogenous DSCAM acts as permissive cue that facilitates RGC axon growth. In contrast, single-cell downregulation or overexpression of DSCAM in tectal neurons showed that DSCAM acts as a restrictive cue to regulate the size and complexity of their dendritic arbors (27). Thus, in addition to RGCs, DSCAM can differentially in uence postsynaptic neurons in the Xenopus visual system. Indeed, punctate DSCAM immunoreactivity can be detected not only within the Xenopus retina but also surrounding cell bodies in the tectum as well as in the tectal neuropil in unpermeabilized tissues (Fig 8a, b). To further differentiate DSCAM expression in tectal neurons, we electroporated embryos with a GFP plasmid, and at stage 45, tadpoles with isolated or small clusters of GFP-expressing neurons were xed and immunostained for DSCAM. We found that DSCAM is not only expressed along the cell body surface of tectal neurons (white circle , Fig 8a), as previously shown (27), but is also expressed along primary dendrites and dendritic branches of tectal neurons (white arrows Fig  8a). Analysis of DSCAM immunostained tissues further revealed a unique pattern of expression, strongly labeling thin processes within the tectal neuropil (Figs 8 a, b; arrowheads). In some tadpoles, the random transfection and expression of GFP within brain neurons revealed strong DSCAM immunoreactivity along the primary processes of GFP-expressing cells that were positioned within the neuropil (Fig 8b). The identity of these cells in Xenopus is unknown, but they share similar morphology and features to tegmental projection neurons characterized in id2b transgenic zebra sh larvae that are found exclusively in the neuropil and have a prominent primary process that protrudes apically (34). Further experiments are needed to con rm their identity in Xenopus and potential roles for DSCAM in these neurons.

Discussion
Our previous studies explored cell-autonomous roles for DSCAM during the development of pre-and postsynaptic structural and functional connectivity in the developing Xenopus retinotectal circuit. We found that DSCAM primarily acts as a neuronal brake to limit and guide postsynaptic dendrite growth of tectal neurons while it also facilitates arborization of presynaptic RGC axons cell autonomously (27). In that study, we targeted ventral RGCs for our analysis since their axons are easier to visualize in vivo with confocal imaging as they project to the most dorsal part of the tectal neuropil (35). For this study, we characterized the expression of DSCAM along the ventrodorsal axis of the optic nerve, and we followed the navigation of ventral and dorsal retinal axons corresponding to this expression. We found a speci c pattern of DSCAM expression along the Xenopus optic nerve that correlated with how optic nerve bers are topographically organized. Fasciculated bundles of ventral bers derived from ventral RGCs normally navigate the optic nerve along its ventral side, which coincided with strong DSCAM expression. As bers crossed the optic chiasm, we observed that DSCAM expression was decreased. This coincides with ventral and dorsal retinal axons rearranging topographically as bers pass the chiasm and project contralaterally into the optic tract and optic tectum. DSCAM has been well characterized as a homophilic binding molecule mediating intracellular adhesion and the fasciculation of axon bundles (15,36). The site and timing of expression suggests that DSCAM is involved, to some degree, in maintaining the ventrodorsal topography of optic nerve bers and the spatial arrangement that mirrors how axons exit the optic nerve head. It is possible that through its adhesive properties and homophilic interactions, DSCAM serves to anchor ventral bers together, preventing any rearrangement or interchange with dorsal axons as bers navigate the optic pathway from the optic nerve head to the chiasm. Differential fasciculation of bers along the optic nerve may be an underlying mechanism to tra c axons in an orderly manner to the chiasm. Organized arrival of axons at the site of the chiasm would allow axons to respond to the next set of guidance cues, including ephrins, other chemoattractant cues, and neurotrophic factors (28, [37][38][39][40], which all prepare for the subsequent stage of morphological trajectory into the tectum. We showed preferential DSCAM expression on ventral RGC axons and along the posterior region of the Xenopus optic chiasm as well (Fig 2a), a nding that is in agreement with observations of DSCAM expression in the posterior region of the mouse optic chiasm (15).
In addition to mechanisms organizing the topography and spatial arrangement of axon bers, it is important to note that there are also time-based mechanisms involved that indirectly contribute to the topographic wiring of circuits. During Xenopus eye development, new retinal ganglion cells are generated at the ciliary margin located at the periphery of the eye (35,41). Older cells are pushed towards the central portion of the retina and a gradient of maturing cells is created along the retina radius. Because of the temporal pattern of early eye development, the deployment of emerging RGC axons along the optic pathway is set to a de ned temporal sequence. Dorsal retinal bers exit the eye rst, navigate the optic pathway, and reach the tectum six hours ahead of ventral retinal axons. The newer set of axon bers exiting the eye travel along the most ventral portion of the optic nerve as innervation takes place (35,42). It is possible that fasciculation of retinal bers by DSCAM indirectly modulates the pacing of "younger" ventral axons along the optic nerve -perpetuating a difference in timing at which ventral and dorsal axons reach their target sites. In our previous work using real-time imaging of RGC axons as they innervate the optic tectum, we showed that DSCAM is important in promoting the branching rate of retinal axons in vivo (27), which supports the idea that DSCAM is involved in distinct temporal aspects of RGC axon development and differentiation.
Differential timing of retinotectal projections was initially thought to be the mechanism that generates topographic mapping in the optic tectum, with the argument that pioneering dorsal bers innervate ventral areas in the tectum simply for arriving rst at the available sites. This hypothesis stated that ventral bers of the retina would later follow and would be forced to occupy the next available sites at the dorsal area of the tectum, due to the constraints of existing dorsal axons (35). Studies, however, have shown that disrupting the timing of retinal axon deployment, by heterochronic transplantation of early age RGCs into older embryos, does not seem to affect the topographic mapping formed during development, indicating that other mechanisms are at work (35). It is becoming increasingly evident, based on a number of studies, that sub-populations of RGCs employ different molecular and cellular strategies to achieve axon-target speci city (43,44). For example, sub-populations of RGCs heavily rely on repellant and attractive cues for precise axon targeting. In amphibians, populations of RGCs differentially express ephrin-Bs in a high dorsal to low ventral gradient in the retina (45,46). This gradient pattern in the retina complements EphB1 receptors expression along the Xenopus tectum which is distributed in a high ventral to low dorsal gradient. Signaling between EphB1 receptors and ephrin-B ligands have been suggested to be the underlying mechanism that attracts dorsal retinal axons into the ventral portion of the tectum (45). The work we present in this study adds DSCAM to a growing of list of molecular strategies that retinal axons use to self-organize topographically along the optic nerve and within the target.
We observed that DSCAM expression diminishes after axons cross the optic chiasm and enter the optic tract, but then reemerges gradually at the tectal neuropil where retinal axons arborize and form connections with post-synaptic tectal partners. At this spatial gap where DSCAM expression is decreased, axons bers are rearranged, most likely by pre-and post-synaptic molecular interactions mediated by Ephs/ephrin signaling, to reorient the topography of axons in an inverted manner, while also distributing their projection along the mediolateral axis in the neuropil. Here we tested speci cally whether DSCAM, plays a role in sorting the arrangement of arbors across the mediolateral plane of the tectal neuropil independent of its effect on axon branching. We examined effects of DSCAM on the same population of RGCs (ventral) at a distinct time of development by timing the DSCAM MO knockdown by targeting the electroporation of tadpoles at different stages. We found that downregulating DSCAM expression in ventral RGCs with axons already terminating medially within the tectum caused an aberrant shift and extension of their terminal arbors away from those of dorsal RGCs, suggesting that DSCAM guides remodeling and topographic organization of arbors derived from ventral RGCs. During zebra sh development (which closely resembles Xenopus development), dorsal retinal bers normally reach the optic tectum via the lateral branch, while ventral axons project via the medial branch (9). Disrupting mechanisms dependent on RNA-binding proteins, such as Hermes, causes an aberrant shift in topographic ordering and results in lateral dorsal axons projecting ectopically into the medial branch arbor (9). Thus, our studies indicate that DSCAM, together with other signaling molecules including Hermes, participate in the medio-lateral topographic mapping at the target.
During RGC axon arborization, coordinated addition and retraction of axonal branches and of dendrites of tectal neurons allows for a gradual recognition between pre-and postsynaptic partners which allows for new synaptic connections to be formed (47,48). Additionally, bi-directional communication at the molecular level is also thought to be at work facilitating synaptogenesis. For example, neurotrophins, including brain-derived neurotrophic factor (BDNF), can act as a retrograde signal to in uence presynaptic neurons, while also acting as an anterograde factor on postsynaptic cells (47,49). This type of bi-directional signaling can generally induce the development and maturation of synapses, or even modify the structure of existing synapses. Unpublished work from our lab shows that DSCAM localizes to only a sub-set of retinotectal synapses, suggesting that endogenous DSCAM, localized post-synaptically may be implicated in the stability and/or maintenance of synapses (R.A. Santos and S. Cohen-Cory, unpublished). Studies have shown that topographic arrangement of axons is also precisely organized at the synapse level. Studies both in mouse and in C. elegans indicate that graded inhibitory cues for synapse formation and maintenance are also used to restrict synapse distribution and create synapse topographic maps (50). Homophilic binding between DSCAM proteins in rodents mediates neurite adhesion, which helps facilitate precise synaptic targeting within a speci c sublamina in the retina (51). DSCAM can also functionally interact with other cell-adhesion molecules, speci cally cadherins and protocadherins, to "mask" their adhesive properties and consequently prevent neurite collision and fasciculation (52). In Aplysia, DSCAM acts trans-synaptically and in collaboration with AMPA-like receptors promotes synapse formation (53). In the developing Xenopus tadpole, visually driven Ca 2+ signals are topographically organized at the subcellular dendritic scale in tectal neurons (33).
Characterizing the spatial distribution of molecules, such as DSCAM, on both pre-and post-synaptic arbors to match their anatomical location along synapses remains open to investigation.

Conclusion
In the Xenopus, endogenous DSCAM acts at multiple levels along the visual circuit, independently modulating dendrite and axon arborization, where cell-autonomous roles vary depending on the cell type.
Our current work demonstrates that DSCAM is also involved in retinotopic organization at distinct points along the retinotectal pathway, directing the topographic organization of retinal bers as they travel along the optic nerve, in sorting and remodeling of axon arbors along the mediolateral axis within the neuropil, and in maintaining pre-and postsynaptic retinotectal arbors. The nding of selective localization of DSCAM protein in primary dendrites of a subpopulations of neurons in the neuropil opens the possibility of additional roles for DSCAM during neural circuit development. Availability of data and materials The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Competing interests
The authors declare that they have no competing interests.  immunoreactivity was measured using a 3.5 μm ROI at six positions along the posterior to anterior axis tool as for Fig 1, with ten sets of measurements made at distinct locations along the optic nerve or optic chiasm. Mean uorescence intensity is shown with the x-axis designating the optic nerve or optic chiasm from the posterior to anterior regions. (c) A high-posterior to low-anterior distribution of DSCAM uorescent signal is found along the width of the optic nerve, while 3A10 uorescent signal is higher in the anterior portion of the optic nerve. (d) As for the optic nerve, measuring DSCAM and 3A10 uorescence showed a high-posterior to low-anterior distribution of DSCAM immunpositive signals within the optic chiasm, while 3A10 uorescent signal was low posterior and increased anteriorly. Scale bars: 10 μm for a; 50 μm c.

Figure 3
Visualizing DSCAM expression in cleared Xenopus brain tissues (a, b) Whole tissue clearing followed by immunostaining was used to further characterize DSCAM expression in intact Xenopus laevis tadpoles. (a) DSCAM immunoreactivity (red, DSCAM only; right panel) along the optic nerve (solid yellow arrowhead) and within the midbrain (empty yellow arrowheads) co-localized with 3A10 antibody staining (green, 3A10 and DSCAM overlay) of optic nerve bers (solid white arrowhead) and RGC axon terminals within the tectal neuropil (empty white arrowheads). In addition to the optic nerve, the 3A10 antibody stains axonal bers in sensory and motor cranial nerves. (b) Individual confocal planes from horizontal zstacks further illustrate co-localization of DSCAM and 3A10 immunoreactivity in the midbrain neuropil (from dorsal-left to ventral-right). Stronger DSCAM immunoreactivity is observed on the dorsal-most portion of the tectum, where axon terminals extensively branch (solid white and yellow arrows). The 3A10 antibody staining also reveals RGC axon bers as they enter the midbrain more ventrally (empty arrowheads) that show weaker DSCAM immunostaining. (c) Higher magni cation confocal images of dissected brains further illustrate DSCAM and 3A10 neuro lament-associated protein co-localization RGC axon terminals identi ed by the 3A10 antibody that are also immunopositive for DSCAM (solid white and yellow arrowheads; yellow lines indicate location of x-z and y-z orthogonal planes, thickness of sample imaged was 85 µm). (d) The uorescence intensity of DSCAM and 3A10 immunostaining was measured in whole brain tissues using a 25 μm circular ROI tool and analyzed using ImageJ with ten measurements obtained across ten regions each at the level of the axon terminals and axon tracts within the midbrain neuropil. Measurements were taken from both brain hemispheres equally. Scale bars: 100 μm for a and b; 25 μm for c.

Figure 4
Comparing DSCAM immunoreactivity across CNS regions in Xenopus laevis tadpoles. (a) A maximum projection image of a cleared Xenopus tadpole illustrates the distribution of DSCAM immunoreactivity (red) across several structures within the CNS. The 3A10 anti-neuro lament antibody (green) is used to visualize axonal tracts. (b, c, d) Single plane images show speci c patterns of DSCAM expression in multiple brain structures including the olfactory bulb (OB), telencephalon (TEL), optic tectum (TEC), and hindbrain (HB). Note the absence of 3A10 staining from axonal bers in the olfactory nerve, OB and TEL.
(e) DSCAM and 3A10 immuno uorescent signals were measured in two tadpoles using a 25 μm ROI tool, with 10 measurements obtained per anatomical structure and across 10 regions for each brain structure (OB, TEL, and HB). Note that the TEC was subdivided into axon terminals or axonal tracts. Measurements were taken from both brain hemispheres equally. Fluorescence intensity for DSCAM at the OB was higher than in other brain structures. Scale bars: 200 μm for a; 100 μm b, c, d.