Turtle interacts with borderless in regulating glial extension and axon ensheathment

Proper recognition between axons and glial processes is required for the establishment of axon ensheathment in the developing nervous system. Recent studies have begun to reveal molecular events underlying developmental control of axon-glia recognition. In our previous work, we showed that the transmembrane protein Borderless (Bdl) is specifically expressed in wrapping glia (WG), and is required for the extension of glial processes and the ensheathment of photoreceptor axons in the developing Drosophila visual system. The exact mechanism by which Bdl mediates axon-glia recognition, however, remains unknown. Here, we present evidence showing that Bdl interacts with the Ig transmembrane protein Turtle (Tutl). Tutl is specifically expressed in photoreceptor axons. Loss of tutl in photoreceptors, like loss of bdl in WG, disrupts glial extension and axon ensheatment. Epistasis analysis shows that Tutl interacts genetically with Bdl. Tutl interacts with Bdl in trans in cultured cells. We propose that Tutl interacts with Bdl in mediating axon-glia recognition for WG extension and axon ensheathment.


Introduction
The ensheathment of axons by glial processes is required for insulating axons for propagating action potentials [1]. The establishment of axon ensheathment during development requires proper recognition between nascent axons and glial processes. While recent studies have made progress in identifying factors important for axon ensheathment, the molecular mechanisms underlying developmental control of axon-glia recognition are still poorly defined.
Recent data from genetic dissection of the development of the Drosophila adult visual system reveals several important genes in the control of glial development and axon ensheathment. For instance, the FGF signaling pathway has been implicated in regulating glial proliferation, migration and differentiation [2]. At the third-instar larval stage, activation of the receptor tyrosine kinase Heartless (Htl) by the FGF8-like ligand Pyramus promotes the proliferation of perineurial glia (PG) and subsequent migration of PG from the optic stalk into the eye disc. PG migration also requires Gilgamesh, Hedgehog and Integrin [3,4]. Upon reaching the eye disc, Htl is then activated by the neuron-derived FGF8-like ligand Thisbe, which promotes the differentiation of PG into wrapping glia (WG) and the initiation of photoreceptor (R cell) axon ensheathment [2].
The mechanisms underlying axon-glia recognition for R-cell axon ensheathment remains unclear. In our previous study [5], we identified the immunoglobulin-like (Ig) transmembrane protein Borderless (Bdl) as a key player in regulating WG extension and R-cell axon ensheathment. The expression of Bdl undergoes dynamic changes during visual circuit development. At later stage (i.e. pupal stage), Bdl is expressed in R-cell axons, and downregulation of Bdl is required for R7 axonal tiling [6]. Whereas at earlier stage (i.e. third-instar larval stage), Bdl is exclusively expressed in WG and is specifically required for WG extension and R-cell axon ensheathment [5]. Those data leads to the speculation that Bdl mediates axon-glia recognition at the third-instar larval stage by interacting with an unknown receptor on differentiating R-cell axons.
To determine the exact mechanism by which Bdl mediates axon-glia recognition, we set out to identify R-cell axon-surface factors that interact with Bdl on WG in the developing fly visual system. We found that mutants defective in the turtle (tutl) gene encoding for a transmembrane protein, a member of the conserved IgSF9/Dasm1/Tutl subfamily of the Ig superfamily [7][8][9][10], displayed a bdl-like WG extension and axon ensheathment phenotype. Further molecular and genetic analysis support that WG-specific Bdl interacts with Tutl on R-cell axons in mediating axon-glia recognition for WG extension and axon ensheathment.

Results
Tutl is expressed in R-cell axons in the developing visual system at third-instar larval stage In our previous study [7], we showed that Tutl is present on R-cell axons and is required for specifying R7 axonal tiling at later stages (i.e. pupal stage). To determine if Tutl also plays a role in regulating visual circuit development at earlier stage (i.e. third-instar larval stage), we examined the expression pattern of Tutl in third-instar larval eye-brain complexes. At the third-instar stage, differentiating WG processes in the eye disc contact nascent R-cell axons, and subsequently migrate along R-cell axons from the eye disc through the optic stalk into the optic lobe.
In wild type, anti-Tutl staining was detected in R-cell bodies and axons (Fig. 1a, d). To confirm the specificity of anti-Tutl staining, we performed anti-Tutl staining of tutl eye-mosaic mutants in which the majority of eye tissues were homozygous for tutl mutations. Since tutl mutant clones were generated by eye-specific mitotic recombination, the tutl gene was selectively removed in R cells, but not removed in glia. We found that anti-Tutl staining was absent in tutl mutant R cells in the eye disc, and was also missing in subretial space where tutl mutant axons projected into, while R-cell differentiation remains normal in tutl eye-mosaic tissues ( Fig. 1e and  h). This result indicates that unlike Bdl that is exclusively expressed in WG [5], Tutl is specifically expressed in R-cell bodies and axons.
tutl is required for the extension of WG processes along R-cell axons in the optic lobe In our previous study [5], we showed that loss of bdl caused defects in the extension of WG processes along R-cell axons within the optic lobe. We speculated that Bdl on WG recognizes an unknown receptor on R-cell axons for promoting WG extension and axon ensheathment. Since Tutl is present on R-cell axons at the third-instar larval stage when WG initiates its contact with R-cell axons ( Fig. 1), we tested Tutl as one of the candidate receptors on R-cell axons that may be recognized by Bdl on WG.
To determine the potential role of Tutl in WG extension, we examined the effects of removing tutl on R-cell projections and WG extension. To specifically remove tutl in R cells, we performed eye-specific genetic mosaic analysis. Large clones (>90% of eye tissues) of homozygous tutl tissues were generated in the developing eye. This allows the removal of the tutl gene in R cells, but not in WG. The patterns of R-cell axonal projection and WG extension in tutl eye-mosaic mutants were then analyzed. In wild type, developing R cells in the eye disc project their axons basally into subretinal space, and then migrate posteriorly into the optic stalk ( Figs. 1d and 2a). After exiting the optic stalk, R-cell axons enter the lamina, the superficial layer of the developing optic lobe (Fig. 2h). WG processes contact the nascent R-cell axons at the subretinal space, and subsequently extend along the surface of R-cell axons from the eye disc through the optic stalk into the optic lobe (Figs. 1d and 2h).
Compared to that in wild type ( Fig. 2a and c), R-cell differentiation occurred normally in the posterior region of the eye disc in tutl eye-mosaic mutants ( Fig. 2b and d).
Like that in wild type (Fig. 2h) and bdl mutants (Fig. 2i), R-cell axons in tutl eye-mosaic mutants projected normally into the optic lobe (Fig. 2j). tutl R1-R6 growth cones established a smooth and continuous layer at the lamina termination site, their intermediate target region, while tutl R7 and R8 axons projected normally into the deeper medulla layer (Fig. 2j). However, we found that like that in bdl mutants (Fig. 2f, i), much less and shorter WG processes were observed in the lamina in tutl eye-mosaic mutants (Fig. 2g, j, k). These results indicate that tutl, like bdl, is required for the extension of WG processes in the lamina.

Eye-specific knockdown of tutl also disrupts glial extension
We also examined the effects of knocking down tutl on the extension of WG by utilizing an UAS-tutl-RNAi transgene, which was shown to effectively reduce the level of tutl in R cells in the our previous study [7]. The expression of UAS-tutl-RNAi was controlled by the eye-specific driver GMR-GAL4, which allows the expression of UAS-tutl-RNAi in R cells, but not in glia. Since Bdl is exclusively expressed in WG at this stage, we performed anti-Bdl staining to examine WG extension. Although knockdown of tutl in R cells did not affect the expression level of Bdl in WG (data not shown), the extension of WG in the lamina was severely disrupted (Fig. 3c, d, e). This result, together with that from eye-specific mosaic analysis (Fig. 2), indicates a role for tutl in R-cell axons for promoting WG extension in the developing optic lobe.
tutl interacts with bdl genetically in the control of WG extension That loss of tutl in R-cell axons phenocopies that caused by loss of bdl in WG, suggests that Tutl interacts with Bdl in mediating axon-glia recognition for promoting WG extension. To further assess this, we performed epistasis analysis in two ways.
Firstly, we tested the effects of reducing the dosage of bdl on the WG extension phenotype caused by partial loss of tutl. Weak tutl mutants (i.e. tutl k14703 /Δtutl) displayed a mild WG extension phenotype (Fig. 4a). When the dosage of bdl was reduced by 50%, we found that the WG extension phenotype was significantly enhanced (Fig. 4a).
And secondly, we examined the effects of reducing the dosage of tutl on WG extension in flies in which bdl was mildly overexpressed in WG. In our previous study [5], we showed that strong overexpression of Bdl in WG Fig. 1 Tutl is expressed in R cells in the developing visual system at third-instar larval stage. Longitudinal optic sections of wild-type (a-d) and tutl-eye mosaic (e-h) third-instar eye discs were triple-stained. Tutl immunoreactivity was detected with a rabbit anti-Tutl antibody (red). WG processes were visualized with UAS-mCD8-GFP driven by the WG-specific Mz97-GAL4 (i.e. Mz97-GFP) (green). R-cell bodies and their axons were labeled with anti-HRP (b) or MAb24B10 (f) (blue). Apical is up, and posterior is to the right. a In wild type, Tutl immunoreactivity was detected in R-cell bodies and axons in the eye epithelium as well as subretinal regions where R-cell axons migrate posteriorly towards the optic stalk. b All R-cell bodies and axons in a were visualized with anti-HRP staining. c WG processes in a were labeled with Mz97-GFP. WG processes at the subretinal region follow R-cell axons and migrate posteriorly. d Images in a-c were merged. (e-h) Large clones of homozygous tutl 23 mutant eye tissues were generated with eye-specific mitotic recombination. Thus, the tutl gene was specifically removed in mutant eye clones, but not removed in WG. e Anti-Tutl staining. Tutl immunoreactivity was absent in mutant R-cell bodies and axons. At subretinal regions where tutl mutant R-cell axons associate with WG processes, Tutl immunoreactivity was also missing. f All R-cell bodies and axons in e were labeled with MAb24B10 staining. g WG processes in e were labeled with Mz97-GFP. h Images in e-g were merged. Scale bar: 10 μm Fig. 2 Loss of tutl in R cells disrupts the extension of WG processes in the optic lobe. a-d Third-instar eye discs were double-stained with anti-Elav (green) and anti-Tutl (magenta). Anti-Elav antibody recognizes the neuronal-specific nuclear protein Elav that is expressed in all differentiating R cells. Anterior is up. a and c In wild type, differentiating R-cell nuclei were observed in the posterior region of the eye disc. b and d In tutl 23 eye-specific mosaic individuals, R-cell differentiation occurred normally. Homozygous tutl mutant tissues were identified by lack of anti-Tutl staining. e-j Third-instar eye-brain complexes were double-stained with mCD8-GFP (green) and MAb 24B10 (red). e and h In wild type, WG in the subretinal region project processes (arrow) that follow R-cell axons from the eye disc through the optic stalk (os) into the lamina (la). f and i In homozygous bdl EX2 mutants, much less and shorter WG glial processes followed R-cell axons into the lamina. g and j In tutl 23 eye-specific mosaic individuals in which tutl was specifically removed in R cells, but not in WG, a bdl-like WG extension phenotype was observed. Note that R-cell axons projected normally into the optic lobe in tutl mosaic individuals (j). k The length of individual WG extension within lamina was measured. Loss of tutl in R cells significantly affected the extension of WG processes (p < 0.001). Two-tailed t-tests were used. Error bars indicate SEM. Abbreviations: ed, eye disc; os, optic stalk; la, lamina; me, medulla. Number of individuals examined: wild type, 9; tutl eye-mosaic, 10. Scale bar: a-d, 10 μm; e-i, 20 μm disrupted WG extension by increasing WG-WG homotypic adhesion, whereas mild overexpression of Bdl in WG did not affect WG extension. We reasoned that if Tutl on R-cell axons indeed interacts with Bdl on WG, then decreasing the level of Tutl on R-cell axons may free some Tutl-interacting Bdl for Bdl-Bdl homophilic binding on opposing WG surface, thus enhancing WG-WG adhesion. Interestingly, we found that although mild Bdl overexpression did not cause a WG extension phenotype (Fig. 4b), reducing the dosage of tutl by 50% in flies with mild bdl overexpression severely disrupted WG extension (Fig. 4b).
Taken together, above results support that Tutl and Bdl interact in vivo in regulating WG extension.

Blockade of FGF8 signaling in WG severely disrupts glial extension
Previous studies by us and others show that the FGF8 receptor Heartless (Htl) is required for WG extension Fig. 3 Knockdown of tutl in R cells also disrupts glial extension. a-d WG processes and R-cell axons in third-instar eye-brain complexes were visualized with anti-Bdl (magenta) and MAb 24B10 (green), respectively. a and b In wild type, WG processes follow R-cell axons into the lamina (arrows). c and d When tutl was knocked down in R cells, but not in WG, WG extension in the lamina was disrupted. GMR-tutl-RNAi denotes that the UAS-tutl-RNAi transgene was under control of the eye-specific driver GMR-GAL4. e The length of individual WG extension within lamina was measured. Knockdown of tutl in R cells significantly disrupted WG extension (p = 2.17E-07). Two-tailed t-tests were used. Error bars indicate SEM. Number of individuals examined: wild type, 9; GMR-tutl-RNAi, 12. Scale bar: 20 μm and axonal ensheathment [2,5]. However, we did not detect any genetic interaction between bdl and htl [5]. To further evaluate the role of Htl in WG extension, we examined the effects of blocking the function of Htl specifically in WG. A dominant-negative form of Htl that has been shown previously to block the Htl activity effectively [2,11], was expressed in WG under control of the WG-specific driver Mz97-GAL4. We found that interfering with the Htl function with this dominantnegative form caused a much more severe phenotype ( Fig. 5c and d). The majority of WG processes failed to enter the optic stalk. This phenotype was never observed in tutl or bdl null mutants. Similarly, no such phenotype was observed even when the level of tutl was reduced in bdl null mutants (n = 51). This result suggests a role for Htl in WG extension that is different from that of Tutl and Bdl.

Loss of tutl disrupts the ensheathment of R-cell axons
In our previous study [5], we showed that loss of bdl disrupted WG extension and the ensheathment of R-cell axons. To determine if tutl mutations cause a similar phenotype in R-cell axonal wrapping, we performed ultrastructural analysis to examine R-cell axon ensheathment in the optic stalk at third-instar larval stage.
In wild type, R-cell axons project from the eye disc into the optic stalk where R-cell axons are wrapped by WG Fig. 4 tutl interacts genetically with bdl in the control of WG extension. a Reducing the dosage of bdl enhanced the WG extension phenotype in tutl k14703 hemizygotes. Flies in which the tutl k14703 allele was placed over a tutl deletion allele (i.e. tutl 23 ), showed a mild WG extension phenotype. This tutl k14703 hemizygote phenotype was significantly enhanced when the dosage of bdl was reduced by 50%. "****", p < 0.00001. Two-tailed t-tests were used. Number of individuals examined: tutl k14703 /Δtutl, 9; tutl k14703 /Δtutl, bdl, 12. b Reducing the dosage of tutl enhanced the WG phenotype induced by mild overexpression of bdl in WG. Mild bdl overexpression in WG (i.e. bdl mild) did not affect WG extension. However, bdl mild-overexpression flies in which the dosage of tutl was reduced by 50%, displayed a strong WG extension phenotype. ns, not significant; "***", p < 0.0001. One-way ANOVA analysis was used. Error bars indicate SEM. Number of individuals examined: wild type, 9; bdl mild; +/+, 13; tutl 23 /+, 6; bdl mild; tutl 23 /+, 15 processes (Fig. 6a). R-cell axons from the same ommatidia form a single bundle in which the central R8 axon is surrounded by R1-R7 axons. Each mature R-cell axonal bundle is ensheathed by WG processes that follow R-cell axons from the eye disc into the optic stalk. To assess the potential role of tutl in axon ensheathment, we performed eye-specific genetic mosaic analysis. Large clones (>90% of eye tissues) of homozygous tutl eye tissues were generated, which removed tutl in R cells, but not in glia. Interestingly, we found that like loss of bdl in WG [5], loss of tutl in R-cell axons caused defective entheathment of R-cell axons in the optic lobe. Compared to that in wild type, the number of bundles completely wrapped by glial processes decreased significantly in tutl eye-mosaic animals (Fig. 6b, d), a phenotype identical to loss of bdl [5]. Additionally, like loss of bdl in WG [5], loss of tutl in R-cell axons also caused some glial processes to extend into the interior region of the fascicle to contact the central R8 axon (Fig. 6c, e). We conclude that Tutl, like Bdl, is required for R-cell axonal ensheathment.

Bdl displays trans interactions with Tutl in cultured cells
That removal of tutl in R-cell axons or loss of bdl in WG causes a similar WG extension and axon ensheathment phenotype, together with the observed genetic interactions between tutl and bdl, raise the interesting possibility that Bdl on WG interacts with Tutl on R-cell axons in trans in mediating axon-glia recognition.
To test if Bdl is capable of interacting with Tutl in trans on cell surface, we performed cell-cell co-aggregation studies. Bdl and Tutl were expressed in Drosophila Schneider-2 cells (S2 cell). In our previous study [6], we showed that Bdl possesses homophilic binding activity and could induce the formation of large cell aggregates (>20 cells) (Fig. 7a and c). When two populations of cells were mixed, Bdl-RFP-transfected cells selectively formed large aggregates with each other, but not with GFP-transfected cells. Interestingly, we found that Tutl-GFP-transfected cells also formed coaggregates with Bdl-RFP-transfected cells (Fig. 7b and c). We then tested a set of Bdl deletion constructs in which certain Ig-like domains or fibronectin-type-III repeats (FN) are deleted. Deletion of both Ig-like domain 1 and 2, or deletion of Ig-like domain 1, prevented the coaggregation of Bdl-transfected cells with Tutl-transfected cells (Fig. 7c). By contrast, deletion of Ig-like domain 3 and 4, or deletion of two FN repeats, did not affect cellcell co-aggregation. These results indicate that Bdl is Fig. 5 Blockade of FGF8 signaling in WG prevents WG processes from entering the optic stalk. Third-instar eye-brain complexes were doublestained with mCD8-GFP (green) and MAb 24B10 (red). a and b Wild type. c and d In flies with the expression of a dominant-negative form of htl under control of the WG-specific driver Mz97-GAL4 (n = 35), WG extension was severely disrupted. In flies showing the most severe phenotype (~14%), the majority of WG processes failed to enter the optic stalk (arrow), a phenotype that was never observed in bdl or tutl null mutants. "os" denotes optic stalk. Scale bar: 20 μm capable of interacting with Tutl in trans, and this trans interaction requires the Ig-like domain 1 of Bdl.

Discussion
Our previous study suggests that Bdl functions as a WG-specific cell surface receptor in mediating WG extension and axon ensheathment by interacting with an unknown ligand on R-cell axons [5]. In this study, we present several lines of evidence that supports that Tutl functions as a Bdl-interacting partner on R-cell axons. Tutl is expressed in R cells at third-instar larval stage. Loss of tutl in R-cell axons, like loss of bdl in WG, disrupts WG extension and axon ensheathment. Tutl displays trans interactions with Bdl in cultured cells, and interacts genetically with Bdl in the control of WG extension and axon ensheathment in the developing visual system. We propose that Bdl on WG interacts with Tutl on R-cell axons in mediating axon-glia recognition in the developing visual system.
Our results identify a novel and important role for Tutl in mediating glial extension and axon ensheathment at third-instar larval stage. Loss of tutl disrupts WG extension and R-cell axon ensheathment, but does not affect R-cell differentiation and axonal projections.
In our previous study [7], we showed that Tutl plays an important role in specifying R7 axonal tiling at later stages (i.e. pupal stage). In addition to its role in the developing visual system, Tutl has also been shown to act in the embryonic nervous system to regulate midline crossing, dendritic morphogenesis and larval locomotion [12][13][14].
In addition to its interaction with Tutl, Bdl also possesses homophilic binding activity [6]. In our previous study [5], we showed that strong overexpression of Bdl in WG also disrupts glial extension by increasing gliaglia adhesion. Interestingly, reducing the level of tutl enhanced the phenotype in flies with mild Bdl overexpression (Fig. 5). One possible explanation is that a reduction in the level of Tutl on R-cell axons frees some Bdl on WG from the Bdl-Tutl trans interaction, and thus increases the Bdl-Bdl-mediated glia-glia adhesion leading to an enhancement of the WG extension phenotype. It appears likely that proper WG extension requires a fine balance between Bdl-Tutl-mediated axon-glia recognition and Bdl-Bdl-mediated glia-glia adhesion.
The Bdl-Tutl-dependent recognition system is unlikely to be the only axon-glia recognition system that mediates WG extension and axon ensheathment in the developing visual system. Loss of tutl, or loss of bdl, does not completely Fig. 6 Loss of tutl in R cells disrupts axon ensheathment. a-d Third-instar larval optic stalks were analyzed by electron microscopy. False coloring indicates R-cell axonal profile. a In wild type, eight R-cell axons from the same ommatidia form a single bundle, which is wrapped by WG processes within the optic stalk. Note that the central R8 axon is surrounded by other R-cell axons, and does not contact WG processes. b and c In tutl 01085 eye-specific mosaic individuals, in which tutl was mutated in R cells but not in WG, two types of defects in axonal ensheatment were observed. First, some R-cell axonal fascicles were not separated by WG processes (b). And second, some WG processes extended aberrantly within the fascicle to contact the central R8 axon (arrows) (c). d The phenotype in b was quantified. Compared to that of wild type, loss of tutl in R cells significantly increased the number of fascicles that were not separated by glial processes (*, p < 0.05). e The phenotype in c was quantified. Loss of tutl in R cells led to ectopic extension of glia processes that contact the central R8 axon (*, p < 0.05). Two-tailed t-tests were used. For each genotype, three optic stalks were analyzed by electron microscopy. Error bars indicate SEM. Scale bar: 200 nm block WG extension. Moreover, in tutl and bdl mutants, many R-cell axonal bundles are still wrapped correctly by glial processes within the optic stalk. These observations support the existence of other axon-glia recognition systems that function together with Bdl and Tutl in mediating WG extension and axon ensheathment in the developing visual system.
Recent studies show that activation of the FGF8 signaling pathway in WG is also required for WG extension and R-cell axonal ensheathment in the fly visual system [2,5]. However, several lines of evidence suggest that the FGF8 signaling pathway functions differently from Tutl and Bdl in promoting WG extension and axon ensheathment. First, blockade of FGF8 signaling in WG causes a much more severe WG extension phenotype than that in tutl or bdl null mutants (Fig. 5). Second, loss of tutl, like loss of bdl, causes some glial processes to extend into the interior region of R-cell axonal bundles to contact the central R8 axon (Fig. 6) [5], a phenotype that was not observed in mutants defective in FGF8 signaling [2]. And third, no genetic interaction between Bdl and the FGF8 receptor Htl was detected [5]. Those results raise several possibilities. For instance, the FGF signaling pathway functions in parallel to the Tutl-Bdl-mediated recognition system. Alternatively or additionally, FGF8 signaling may modulate the activity of Bdl as well as other Tutl-Bdl-independent recognition systems. Future studies will be needed to distinguish among these possibilities.
Recent studies show that Dasm1/IgSF9A and IgSF9B, the homologs of Tutl in mammals, function in the mammalian developing nervous system. IgSF9B promotes inhibitory synapse development in cultured cells [15], whereas knockout IgSF9/Dasm1 impairs inhibitory synapse development [16]. It will be of interest to determine if Dasm1/IgSF9A and IgSF9B also play a similar role in mediating axon-glia recognition in mammals. Bdl interacts with Tutl in trans in cultured cells. a Bdl-RFP-transfected S2 cells (red) formed large cell aggregates with each other, but not with cells transfected with a GFP construct (green). b Bdl-RFP-transfected S2 cells (red) formed large co-aggregates with cells transfected with a Tutl-GFP construct (green). c Among total cell aggregates (>20 cells), the percentage of cell co-aggregates were quantified. A cell co-aggregate is defined as a cell aggregate (>20 cells) in which the percentage of each cell population (i.e. GFP-or RFP-positive cells) should be at least 15%. Three independent experiments were performed. Error bars indicate SEM. "****", p < 0.0001; "***", p < 0.001. Two-tailed t-tests were used. Scale bar: 20 μm