InteractiveFly: GeneBrief

golden goal: Biological Overview | References

BIOLOGICAL OVERVIEW

During Drosophila visual system development, photoreceptor (R) axons choose their correct paths and targets in a step-wise fashion. R axons with different identities make specific pathfinding decisions at different stages during development. This study shows that the transmembrane protein Golden goal (Gogo), which is dynamically expressed in all R neurons and localizes predominantly to growth cones, is required in two distinct steps of R8 photoreceptor axon pathfinding: Gogo regulates axon-axon interactions and axon-target interactions in R8 photoreceptor axons. gogo loss-of-function and gain-of-function phenotypes suggest that Gogo mediates repulsive axon-axon interaction between R8 axons to maintain their proper spacing, and it promotes axon-target recognition at the temporary layer to enable R8 axons to enter their correct target columns in the medulla. From detailed structure-function experiments, it is proposed that Gogo functions as a receptor that binds an unidentified ligand through its conserved extracellular domain (Tomasi, 2008).

In the developing nervous system, a precise neuronal network is formed in a step-wise fashion through a series of recognition processes. While axons grow toward their targets, they undergo dynamic changes resulting in decisions to turn, to fasciculate or defasciculate, and to halt or extend according to the extracellular guidance cues provided by the surrounding environment over a short or long range. Extracellular cues can emanate from other axons that run in the vicinity to assemble the input connections with correct spacing and location, or derive from the target cells to attract or repel axons. Therefore, axon-axon interactions and axon-target interactions are important for reaching the target and for the selection of specific synaptic partners. Indeed, both axon-axon and axon-target interactions have been demonstrated to play critical roles in the formation of visual and olfactory circuits in flies and mammals. However, the underlying molecular mechanisms have not been sufficiently elucidated (Tomasi, 2008).

The Drosophila visual system provides an excellent model for studying the mechanisms of axon-axon and axon-target interactions and their role in axonal pathfinding. The compound eye comprises an array of some 800 ommatidia, each of which contains eight photoreceptor neurons, R1 to R8. During larval development, the R8 photoreceptor extends its axon first, followed by R1-R7. According to their neuronal types, photoreceptor (R) axons directly connect to the next order neurons in different target layers within the optic lobe. R1-R6 photoreceptor axons form synapses with target neurons in the first optic ganglion of the brain, the lamina, whereas R7 and R8 axons project through the lamina to two different layers in the second optic ganglion, the medulla. R8 terminates within the M3 layer, whereas R7 targets the deeper M6 layer. The medulla layer targeting of R7 and R8 occurs in two selection stages. In the first stage (early pupa), R8 temporarily stops at the M1 layer while maintaining a growth cone structure. The R7 growth cone extends beyond the R8 axon and temporarily stops at the M3 layer. In the second selection stage (midpupa), the R7 axon extends to its final target layer, M6, and forms a stable connection to the higher order neurons. R8 then follows the R7 tract and terminates at layer M3. R8 and R7 axons from the same ommatidium project into the same column, and each columnar axon maintains a constant distance from neighboring axons (Tomasi, 2008).

How are these complex steps of axonal path-finding decisions regulated? Recent genetic studies in the Drosophila visual system revealed roles for several receptors and cell adhesion molecules that control R axonal array establishment and target layer selection, such as the two Cadherin superfamily members, N-Cadherin (N-Cad) and Flamingo (Fmi), two receptor tyrosine phosphatases, LAR and PTP69D, and a cell adhesion molecule, Capricious (Caps). In N-cad, LAR, and PTP69D mutants, R7 axons undershoot the correct target layer M6 and terminate prematurely at layer M3, which is normally targeted by R8s. In fmi mutants, R8 axon targeting is disrupted, and R8s are frequently mistargeted to superficial levels of the medulla. Fmi has also been implicated in the regulation of axon-axon interactions, since fmi mutants show abnormal spacing between the adjacent axonal tracts. From these studies, a Cadherin-based homophilic cell adhesion, possibly controlled by the two receptor tyrosine phosphatases, has emerged as the key regulating mechanism of axon-axon and axon-target interaction in the Drosophila visual system (Tomasi, 2008).

However, since both N-Cad and Fmi are expressed on all R axon types and in multiple target layers in the optic lobe, the homophilic interaction of these two Cadherins alone cannot account for the distinct target layer selection of R7 and R8. One of the two phosphatases, LAR, which possibly modulates Cadherin interactions, is also expressed broadly in all R axons and multiple target layers. One exception is the homophilic adhesion molecule Caps, which is specifically expressed only on R8 axons and their final target layer. Loss of caps function results in R8 target layer selection defects and moreover, ectopic expression in R7 redirects R7 to the R8 target layer. This finding strongly supported the idea that the combination of homophilic adhesive interactions with additional combinatorial codes may be the key mechanism to create the specificity in layer targeting. A similar mechanism was suggested in vertebrates as the homophilic adhesion molecules, encoded by the two sidekicks genes, control layer-specific targeting of retinal neurons. However, even with the set of known molecules, the complete picture of this highly selective process still remains obscure (Tomasi, 2008).

This study investigated the functional role of the single transmembrane molecule, Golden goal (Gogo), in the Drosophila visual system. Gogo has two known conserved extracellular domains, a Tsp1 (Thromospondin1) domain and a CUB domain. Both domains are implicated in directing the migration of growing cells or growth cones in the developing nervous system: e.g., Unc-5 and class 5 Semaphorins contain Tsp1 domains, while A5 and Neuropilin have CUB domains. Gogo protein is dynamically expressed in all R neuron subclasses, where it localizes predominantly along their axons and to their growth cones. In gogo mutant third instar larvae, repulsive interactions among adjacent R8 axons are lost; and in adults, R8 axons stray before or overshoot the correct target layer in the medulla. Overexpression of Gogo in R axons redirects R8 to the superficial layer in the medulla. It is proposed that Gogo mediates repulsive axon-axon interactions between R axons to maintain their proper spacing and promotes axon-target recognition at the M1 temporary layer allowing R8 axons to enter their correct columns in the medulla. Evidence is provided that Gogo may function in R8 axons as a receptor through a heterotypic interaction with an unidentified ligand (Tomasi, 2008).

This study has shown that the single transmembrane protein Gogo, which is expressed mainly in photoreceptor cells, is required in the retina for R8 axon-axon repulsive interactions and appropriate column and synaptic target layer selection in the optic lobe of the Drosophila brain. In eye-specific gogo⁻ mosaics, R8 axons entangle each other, forming bundles both in larvae and adult. A high number of R8 axons were also observed stopping and straying at the surface of the medulla unable to enter the medulla column. It is proposed that Gogo functions in R8 as a receptor, mediating axon-axon interaction and axon-target interaction by heterotypic interaction with an unidentified ligand (Tomasi, 2008).

The autonomous function of Gogo is based on two observations: first, single isolated WT R8 axons, which are surrounded by misprojecting and stopping gogo mutant axons (reverse MARCM situation), innervate correctly the medulla. Second, in pupae, single gogo mutant R8s fail to extend their axons into the medulla column (Tomasi, 2008).

The requirement of Gogo's cytoplasmic domain is shown in rescue experiments in two different stages during development: axon-axon interaction in larvae and axon-target interaction in adults. The requirement of the cytoplasmic domain argues against a merely adhesive role as shown for adhesion molecules such as N-Cad, which does not require its cytoplasmic domain for homophilic adhesion or for its function. It is also known for repulsive guidance receptors, such as Eph receptors or Dscam, that the physical binding of extracellular domains can be achieved without their cytoplasmic domains. However, the repulsive response triggered by these receptors strictly requires the cytoplasmic domain for intracellular signal transduction. The requirement of Gogo's cytoplasmic domain also implies that cytoplasmic signaling is required for Gogo function. Both lines of evidence for the autonomous function and the requirement of the cytoplasmic domain provide a strong argument that Gogo acts as a novel receptor in axon guidance (Tomasi, 2008).

Gogo's function in axon-axon repulsive interactions becomes apparent in two findings: first, the R8 axons in small gogo mutant clones exclusively form bundles with other mutant axons, and second, the transgenic expression of Gogo in R axons of transheterozygous gogo mutants is sufficient to restore normal axonal spacing. It is proposed that Gogo prevents inappropriate adhesion or bundling among R axons through repulsive interaction (Tomasi, 2008).

The role in axon-target interaction is deduced from two observations: the cell autonomous defects of R8 axons in single gogo mutant clones, characterized by R8 failures in properly entering the medulla column and straying at the surface of the medulla, and the overexpression situation, in which R8 permanently terminates at the surface layer. This study proposes that Gogo positively regulates the adhesion within the M1 temporary layer thereby preventing R8 axons from straying to the wrong target and allowing the proper entry into medulla columns. The final targeting step might require Gogo removal during midpupal stages in order to allow axon extension into the final target layer, which is also supported by the observed antibody staining during pupal development. In the situation of Gogo overexpression, Gogo protein persists until late pupal stages, therefore resulting in R8 axons permanently anchored to the M1 layer (Tomasi, 2008).

Although the observed phenotypes might also be explained by Gogo-dependent silencing of present homophilic adhesive molecules, it is believed that Gogo acts through an as yet unidentified ligand. The possible existence of Gogo-specific ligand(s) is based on four observations. First, gogo does not promote homophilic aggregation in transfected S2 cells. Second, gogo⁻ axons preferentially form bundles with adjacent gogo mutant axons rather than with the adjacent WT axons, indicating a heterotypic repulsive interaction. Third, the gogo transheterozygous mutants can be rescued by the exclusive Gogo expression in R neurons. Axon-target interaction is therefore not dependent on Gogo expression in the brain. Forth, the N-terminal functional domain with its conserved GOGO and Tsp1 domains is strictly required. Eight cysteines that are conserved in the GOGO domain possibly form four disulfide bonds to assemble immunoglobulin-like protein interaction domains. The Tsp1 domain is able to interact with multiple cell-surface or extracellular proteins, including matrix glycoproteins and proteoglycans. Therefore, both domains show the ability for ligand binding (Tomasi, 2008).

What are the possible candidate cells serving as temporal targets of R8 axons during pupal development? It is known that the processes of lamina neurons L1-L5 innervate along the R7 and R8 axons and enter the corresponding medulla columns in early pupal stage. However, by expressing the dominant-negative form of EGFR in lamina neurons, in which the differentiation of L1-L5 is blocked, R8 and R7 still appear to show normal axonal projection at midpupal stage. Recently, it has been reported that the abnormality in axonal tiling of L1-L5 in Dscam2 mutants caused the path-finding defects of R7/R8 axons. It is possible that the abnormally guided processes of L1-L5 may result in a different outcome than the simple loss of L1-L5. Other possible candidates are some of the medulla neurons, which send their processes into the columnar structure of the M1-M6 layers in the outer medulla. Since the mechanism of differentiation and development of these neurons and their processes is largely unknown, it will be intriguing to explore the role of Gogo in possible interactions between R axons and lamina/medulla neuron processes (Tomasi, 2008).

Gogo may interact positively with cell-surface molecules, such as the protocadherin superfamily member, Fmi. Evidence in support of this idea comes from the similarity of their visual system phenotypes and their expression patterns. Both gogo and fmi mutants show loss of repulsive interaction between adjacent R8s, resulting in a bundling of R8 axons in the larval stage, and show a targeting defect in which R8 stops at the superficial layer of the medulla. It is noteworthy that Fmi seems not to serve as an ordinary adhesion molecule in this context. The similarity in the expression pattern is also striking. Both Gogo and Fmi are expressed in all of the R neurons and accumulate along the axon of newly developed young axons in the optic stalk in third instar larvae and R1-R6 axons during cartridge formation in the midpupal stage. One of the differences in the expression pattern is that Fmi is strongly expressed in lamina neurons and medulla cortical neurons, which strongly overlap with the target layers of R8 and R7s in the larva and pupal stages, while Gogo expression is hardly detectable in the proximal lamina and medulla. These similarities and differences of these two molecules might facilitate understanding the molecular code underlying this system (Tomasi, 2008).

Thus, it will be interesting to identify the relevant Gogo ligand. Strong and specific axon path-finding defects in both gogo LOF mutants and GOF transgenic flies make the Drosophila visual system an ideal model to search for a functionally relevant Gogo ligand. Another important task for the future will be elucidating the intracellular molecular mechanisms in R8 axons by which Gogo regulates R8 axon pathfinding. However, the hints to date have been limited. The cytoplasmic domain of homologs found in various species appears to have neither obvious catalytic domains or signaling modules, nor an overall conservation among the species in its primary structure. However, a short cytoplasmic motif, shared by GOGO domain orthologs, may serve as a protein-interaction domain that binds to a conserved interaction partner (Tomasi, 2008).

Elucidating exactly how Gogo regulates R axon pathfinding in Drosophila may also shed light on gogo homologs in other species. The mammalian homolog Tmtsp is the best characterized molecule of this family so far. It is expressed in endothelial cells and hematopoietic stem cells, and the level of expression gradually declines as the cells differentiate. However, no obvious neuronal expression was reported (Takayanagi, 2006). Although Tmtsp may not have a functional role in axonal pathfinding in vertebrates, it might have underlying molecular machinery analogous to Drosophila gogo in the context of cell-cell communication. In turn, the LOF study and identification of Tmtsp ligand may provide new insights in the molecular mechanism of Gogo function (Tomasi, 2008).

Golden Goal collaborates with Flamingo in conferring synaptic-layer specificity in the visual system

Neuronal connections are often organized in layers that contain synapses between neurons that have similar functions. In Drosophila, R7 and R8 photoreceptors, which detect different wavelengths, form synapses in distinct medulla layers. The mechanisms underlying the specificity of synaptic-layer selection remain unclear. This study found that Golden Goal (Gogo) and Flamingo (Fmi), two cell-surface proteins involved in photoreceptor targeting, functionally interact in R8 photoreceptor axons. The results indicate that Gogo promotes R8 photoreceptor axon adhesion to the temporary layer M1, whereas Gogo and Fmi collaborate to mediate axon targeting to the final layer M3. Structure-function analysis suggested that Gogo and Fmi interact with intracellular components through the Gogo cytoplasmic domain. Moreover, Fmi was also required in target cells for R8 photoreceptor axon targeting. It is proposed that Gogo acts as a functional partner of Fmi for R8 photoreceptor axon targeting and that the dynamic regulation of their interaction specifies synaptic-layer selection of photoreceptors (Hakeda-Suzuki, 2011).

The results suggest that the transmembrane receptor Gogo physically interacts (directly or indirectly) with the atypical cadherin Fmi in cis to cooperatively guide R8 photoreceptor axons to their correct target. However, a robust direct interaction between Gogo and Fmi could not be demonstrated by co-immunoprecipitation, bimolecular fluorescent complementation (BiFC) or proximity ligation assay. The failure in co-immunoprecipitation was probably a result of technical difficulties in solubilizing the seven-pass transmembrane Fmi and maintaining a huge complex during the procedure (Fmi is about 400 kDa). Nevertheless, a close interaction of these proteins is supported by three lines of evidence. First, ectopic expression of Gogo in wing epithelial cells was able to relocate Fmi in cis. Second, Fmi and Gogo colocalized at cell-cell contacts in cultured cells via their ectodomains. Finally, Gogo accumulation at the growth cone was strongly reduced in fmi mutant photoreceptor axons, suggesting that Fmi is at least partially required to localize and/or stabilize Gogo at the growth cone through a close association (Hakeda-Suzuki, 2011).

It has been suggested that Fmi binds homophilically in cis. This study also found that Gogo formed oligomers in cultured cells. These observations suggest that, even if Gogo and Fmi physically interact with each other, they may multimerize and form a protein cluster. Alternatively, Gogo-Gogo, Gogo-Fmi and Fmi-Fmi interactions may happen separately at distinct locations and have different functions (Hakeda-Suzuki, 2011).

Fmi controls the nervous system development broadly. It regulates axon guidance, but also synaptic target selection and dendritic field development. The phenotypic similarities and the genetic interactions of gogo and fmi in diverse aspects of neuronal development in Drosophila suggest that the collaboration of Gogo and Fmi is a general molecular mechanism (Hakeda-Suzuki, 2011).

Notably, however, in the dendrites of multi-dendritic neurons, it has been reported that the ectodomain deletion of Fmi (FmiΔN) is able to partially rescue the fmi dorsal-overgrowth phenotype in dendrites, but the cytoplasmic deletion of Fmi (FmiΔC) cannot. On the contrary, FmiΔN was not able to rescue the phenotype in R8 photoreceptor axon, but FmiΔIntra could. These observations indicate that the underlying molecular mechanisms may be different between axons and dendrites. It will be interesting to investigate the molecular mechanisms of Gogo in dendrite formation to decipher the general principles versus unique, diversified mechanisms mediated by the Gogo-Fmi interaction (Hakeda-Suzuki, 2011).

What is the function of Gogo when interacting with Fmi? Three scenarios are envisioned that are not mutually exclusive. First, Fmi homophilic adhesion properties change when it is associated with Gogo. Second, Gogo mediates intracellular signaling to transduce axon pathfinding information in the growth cone. Third, Gogo adds a specificity code to the Fmi-Fmi homotypic asymmetric interaction. To test the first scenario, a cell aggregation assay was used, mixing Fmi-expressing cells with cells co-expressing Gogo and Fmi. As the two populations of cells were equally distributed in the aggregate, Gogo seems to not have an effect on Fmi homophilic adhesion in S2 cells. The second scenario is supported by the fact that the Gogo cytoplasmic domain mediated the R7 photoreceptor co-overexpression phenotype and that the Fmi cytoplasmic domain was dispensable for R8 photoreceptor axon pathfinding. In addition, the interaction between Gogo and Fmi seems to add molecular specificity to R8 photoreceptor axons, allowing them to recognize the proper layer M3, suggesting the third scenario. Fmi seems to be the cue on the target layers, as elimination of Fmi from a population of brain cells, but not from photoreceptors, resulted in targeting defects in R8 photoreceptor axons. This suggests that Fmi-Fmi homotypic interactions take place between R8 photoreceptor axons and the target cells. However, the interaction seems to be asymmetric, as Gogo is not required in the brain for photoreceptor axon pathfinding (Hakeda-Suzuki, 2011).

Overall, it is proposed that Gogo alone promotes adherence between R8 photoreceptor axons and the M1 layer, that, at mid-pupal stages, Fmi acts antagonistically with Gogo at the M1 layer and Gogo and Fmi collaborate to mediate R8 targeting to the M3 layer, and that Fmi on the target cells mediates homophilic interaction with Fmi on R8 photoreceptor axons at the M3 layer. Fmi is detected on R8 photoreceptor axons when R8 photoreceptor axons extend their tip to the M3 layer, if Fmi protein level is reduced from surrounding neuropils, consistent with the idea that Gogo and Fmi act together to guide the M3 targeting growth cones. The above model is supported by five lines of evidence. First, overexpression of both Fmi and Gogo retargets R7 photoreceptor axons to the M3 layer. Second, removing Fmi from presumptive target cells induces R8 photoreceptor axon stopping at the M1 layer. Third, a combination of Gogo overexpression and fmi hypomorphic background induces more R8 photoreceptor axon stopping at M1 layer than is observed in each of these genotypes individually. Fourth, the Gogo overexpression phenotype is suppressed by mild fmi overexpression. Taking into account that gogo overexpression in a fmi hypomorph does not enhance the axon bundling that is typical in fmi mutant axons, it is unlikely that gogo overexpression merely has a dominant-negative effect on Fmi function. Fifth, in fmi mutants, R8 photoreceptors commonly stall at the M1 layer, whereas in gogo mutants or in the double mutants, R8 photoreceptors have a tendency to stray at the M1 layer. This difference is thought to be a result of a reduced adhesion of R8 photoreceptors to M1 layer in gogo mutants; adhesion is not impaired in fmi mutants (Hakeda-Suzuki, 2011).

The cell identity of the M3 layer that is recognized by Gogo and Fmi in R8 photoreceptor axons is not clear. fmi was knocked out almost completely from the lamina neurons. Although the R8 photoreceptor axon stopping phenotype was not completely penetrant, substantial R8 photoreceptor axon stopping was observed at the M1 layer, indicating that the lamina neurons might be the target cells in which Fmi functions as a ligand. Lamina neurons innervate into medulla layers during early pupal stages. Their processes take over R8 photoreceptor axons when R8 photoreceptor axons rest at the temporary layer, and they arborize between developing R7 and R8 photoreceptor termini. Notably, L3 lamina neurons spread their terminal processes at the M3 layer. The functional importance of L3 neurons in this context should be addressed in the future. In any case, it seems that mutual interactions between lamina neuron processes and photoreceptor axons account for the two-step targeting mechanism of R8 photoreceptor axons (Hakeda-Suzuki, 2011).

A putative tyrosine phosphorylation site of the cell surface receptor Golden goal is involved in synaptic layer selection in the visual system.

Golden goal (Gogo) is a cell surface protein that is crucial for proper synaptic layer targeting of photoreceptors (R cells) in the Drosophila visual system. In collaboration with the seven-transmembrane cadherin Flamingo (Fmi), Gogo mediates both temporary and final layer targeting of R-cell axons through its cytoplasmic activity. However, it is not known how Gogo activity is regulated. This study shows that a conserved Tyr-Tyr-Asp (YYD) tripeptide motif in the Gogo cytoplasmic domain is required for photoreceptor axon targeting. Deleting the YYD motif is sufficient to abolish Gogo function. The YYD motif is shown to be a phosphorylation site, and mutations in the YYD tripeptide impair synaptic layer targeting. Gogo phosphorylation results in axon stopping at the temporary targeting layer, and dephosphorylation is crucial for final layer targeting in collaboration with Fmi. Therefore, both temporary and final layer targeting strongly depend on the Gogo phosphorylation status. Drosophila Insulin-like receptor (DInR) has been reported to regulate the wiring of photoreceptors. This study shows that insulin signaling is a positive regulator, directly or indirectly, of YYD motif phosphorylation. These findings indicate a novel mechanism for the regulation of Gogo activity by insulin signaling-mediated phosphorylation. It is proposed that a constant phosphorylation signal is antagonized by a presumably temporal dephosphorylation signal, which creates a permissive signal that controls developmental timing in axon targeting (Mann, 2012).

The YYD tripeptide (Tyr1019-Tyr1020-Asp1021) is conserved among invertebrate and vertebrate species and has a crucial role: deleting it is sufficient to completely abolish Gogo function. YYD is a phosphorylation site and the phosphorylation status of Gogo is critical for both temporary and final layer targeting. A model is proposed in which Gogo is phosphorylated during the first targeting step. A prolonged phosphorylation during the mid-pupal stage prevents R8 axons from extending to their final layer (Mann, 2012).

The current experiments show that it is the dephosphorylated form of Gogo that is the most active. The non-phospho-Gogo is functional during R8 targeting as it could be used to rescue the gogo minus mutant. By contrast, phosphomimetic Gogo does not rescue the mutant phenotype. This mechanism resembles the molecular regulation of Robo activity in that dephosphorylated Robo shows the most activity in mediating repulsive signals during embryonic CNS axon guidance. Additionally, protein inactivation upon phosphorylation, although relatively rare, has been reported in biochemical pathways; for instance, the inactivation of the transcriptional co-activator Yorkie by Warts and others (Mann, 2012).

Gogo function was abolished by removing the YYD site (GogoδYYD) or by mimicking phosphorylation (GogoDDD). Both forms result in a very strong adhesiveness and in the stopping of growth cones at the M1 layer, suggesting the involvement of phosphorylation in the first targeting step. It was postulated that two independent pathways could be activated depending on the phosphorylation status, resulting in either normal M3 layer targeting or M1 stopping, and their activation could be mutually exclusive (Mann, 2012).

Although dephosphorylated Gogo enables the axons to leave the M1 layer, it is not clear whether in a physiological situation phosphorylation contributes to adhesiveness to the M1 layer. Phosphorylation might not entirely be necessary, as axons expressing only the dephosphorylated GogoFFD can still recognize the M1 layer. However, there might be more fine-tuning defects, such as in the location of the synapses along the R8 axons between the M1 and M3 layers (Mann, 2012).

Defects in dephosphorylation can also result in mistargeting of the M3 layer. This is supported by the fact that phospho-Gogo does not show the proper cooperation with Fmi during final layer targeting. Normally, for a proper Gogo-Fmi collaboration, colocalization is required. However, Fmi can colocalize with both phospho- and non-phospho-Gogo, suggesting that the phosphorylation status is important for signal transmission and not for the interaction with Fmi (Mann, 2012).

Another molecule known to interact with Gogo, Hts (the Drosophila homolog of adducin 1) was shown to bind to Gogo and to play a role in guiding photoreceptors (Ohler, 2011). However, this physical interaction occurs independently of the YYD site (Mann, 2012).

Further evidence for the role of Gogo phosphorylation during R8 targeting comes from the experiments in which the Gogo phosphorylation status was genetically modulated, looking for kinases and phosphatases that might modulate the Gogo phosphorylation status. The Drosophila genome contains a relatively small number of tyrosine kinases and phosphatases. This study focused on proteins that are expressed in the brain, that show a transmembrane localization and are implicated in axon guidance. Since Gogo YYD motif phosphorylation is involved in the characteristic overexpression phenotype, it was convenient to screen for the suppression or enhancement of M1 blobs when a kinase or phosphatase was co-overexpressed with gogo. From a number of genes tested (Abl, Src42A, Src64B, drl, Egfr, dinr, Lar, Ptp69D, eya) dinr was identified as a possible regulator of Gogo phosphorylation. All other tested kinases were excluded from a detailed analysis because the overexpressed genes either did not enhance/suppress the gogo gain-of-function phenotype, resulted in extensive cell death, or caused a severe axon guidance phenotype that was difficult to distinguish from a cell death phenotype (Mann, 2012).

It is striking that GogoDDD causes a much stronger adhesiveness to the M1 layer than Gogo phosphorylated by overexpressed DInR. A possible explanation is that, unlike GogoDDD, DInR-dependent phosphorylation in this case is not complete. Alternatively, there are redundant mechanisms that further modulate Gogo phosphorylation. Therefore, only a small proportion of photoreceptors stop at the M1 layer and the majority of axons form blob-like structures (Mann, 2012).

The cues that a growing axon encounters can be divided into instructive or permissive. Instructive cues usually have a restricted expression pattern and guide the axon by providing either attractive or inhibitory information to the growth cone. Permissive signals steer in response to instructive or are needed to detect and respond to extracellular guidance cues (Mann, 2012).

These findings confirm the intriguing possibility that insulin signaling modulates axon guidance. It is difficult to imagine that DInR transduces a typical guidance cue: the only known ligands for DInR, the DILPs, are secreted into the circulatory system and thus cannot provide a directional cue. Rather, insulin signaling could orchestrate the guidance signals coming from instructive directional cues. Insulin signaling could be essential for ensuring the correct wiring of the nervous system by influencing the phosphorylation of a regulator of photoreceptor axon guidance, Gogo. Gogo phosphorylation provides a signal that enhances the adhesive interaction with the M1 layer, whereas dephosphorylation could provide a permissive signal that allows the axon to leave the M1 layer and project to the M3 targeting layer (Mann, 2012).

The state of phosphorylation of a protein at any moment, and thus its activity, depends on the relative activities of the protein kinases and phosphatases that modify it. This suggests that a Gogo dephosphorylation mechanism exists. It would be rewarding to identify the phosphatase that mediates Gogo dephosphorylation and thereby constitutes an essential regulator of Gogo activity. Preliminary genetic studies of several candidates, including Lar and Ptp69D, have not revealed any genetic interaction so far. In summary, a mechanism is proposed whereby the activity of the axon guidance receptor Gogo is regulated by phosphorylation mediated by DInR and dephosphorylation mediated by an as yet unknown phosphatase. This may provide insight into how developmental timing is coordinated in neuronal circuit wiring through a phosphorylation-dephosphorylation mechanism (Mann, 2012).

Hts, the Drosophila homologue of adducin, physically interacts with the transmembrane receptor golden goal to guide photoreceptor axons

Neurons steer their axons towards their proper targets during development. Molecularly, a number of guidance receptors have been identified. The transmembrane protein Golden goal (Gogo) was reported previously to guide photoreceptor (R) axons in the Drosophila visual system. This study shows that Hts, the Drosophila homologue of Adducin, physically interacts with Gogo's cytoplasmic domain via its head-neck domain. hts null mutants show similar defects in R axon guidance as do gogo mutants. Rescue experiments suggest that the C-terminal tail but not the MARCKS homology domain of Hts is required. Overexpression of either gogo or hts causes abnormally thick swellings of R8 axons in the medulla, but if both are co-overexpressed, R8 axons appear normal and the amount of excessive Hts is reduced. The results fit with a model where Gogo both positively and negatively regulates Hts that affects the Actin-Spectrin cytoskeleton in growth cone filopodia, thereby guiding R axons (Ohler, 2011).

This study show that the axon guidance receptor Gogo physically interacts with the cytoskeletal protein Hts. The loss-of-function phenotypes of hts and gogo mutants are qualitatively very similar, albeit gogo null mutants show a phenotype slightly more severe than the hts^null mutant. This suggests that Gogo and Hts collaborate in a functional complex to guide R7 and R8 axons to their correct targets in the medulla (Ohler, 2011).

However, evidence is also shown for an antagonistic interaction between Hts and Gogo. Strong overexpression of Gogo causes abnormally thick swellings of R8 axons at layers M1 and M3. Strong overexpression of Add1 causes a different but similar phenotype leading to abnormal swellings that are restricted to layer M1. If both Gogo and Add1 are overexpressed, no abnormally thick swellings occur and R8 axons do not look different from wild type R8 axons. Moreover, in flies lacking one copy of the hts locus, the effect of excessive Gogo is enhanced. This indicates that Hts and Gogo antagonize each other and need to be in balance for the correct formation of axons. Direct evidence for an antagonistic interaction between Gogo and Hts comes from the observation that an increase in axonal Gogo protein level reduces the amount of Add1 protein in the axon. Moreover, the fact that the Add1 protein level is regulated by Gogo strongly suggests that gogo acts upstream of hts (Ohler, 2011).

How can these superficially conflicting observations be explained and reconciled? Below, a hypothetical model is discussed that explains how the Gogo-Hts complex could function to guide R axons:
Axons find their correct targets by means of the growth cone that is equipped with guidance receptors reading guidance cues provided by the growth cone's environment. The growth cone translates this guidance information into rearrangements of its cytoskeleton, which leads to directed growth of the axon. The two main components of the growth cone cytoskeleton are F-Actin (appearing as filopodia or lamellipodia) and microtubules. Within a filopodium, F-Actin is organized as parallel bundles, which requires the action of F-Actin bundling proteins like α-actinin. The barbed ends of the F-Actin bundles point distally, so that F-Actin assembly takes place at the very tip of the filopodium. This produces a force on the F-Actin bundle that moves the bundle rearwards (retrograde flow) and a force on the plasma membrane that extends the filopodium. F-Actin capping proteins influence the rate of Actin assembly, whereas the rate of retrograde F-Actin flow has been suggested to be regulated by a 'clutch' that links the cytoskeleton via transmembrane proteins to the extracellular matrix and thereby countervails the retrograde F-Actin flow (Ohler, 2011).

Adducin bundles Actin filaments and caps the barbed Actin filament ends in vitro. Assuming that its Drosophila homologue Hts serves the same molecular functions, Hts is an attractive candidate for a protein that is involved in the proper organization of filopodial F-Actin during axon guidance. To see the effect of Add1 on axonal F-Actin, medullae of flies overexpressing either Add1 or Gogo were stained with phalloidin. Compared to the control flies that have only the GMR-Gal4 driver, but no UAS-target, F-Actin seems to be stabilized in axons that feature an excessive amount of Add1 or a reduced amount of Add1 caused by excessive Gogo. In both cases, R7/R8 axons in the medulla are stained more clearly than in the control. This indicates that Add1 indeed affects the F-Actin in R axons in some way. Due to some analogies to the L1-Ankyrin system, which has been shown to function as a molecular clutch, especially a possible involvement in the regulation of retrograde F-Actin flow immediately comes to mind (Ohler, 2011).

Like Adducin, the peripheral membrane protein Ankyrin is a part of the Actin-Spectrin cytoskeleton. Ankyrin binds to the cell adhesion molecule L1, especially when L1 is homophilically bound to another L1 molecule in trans. The physical link of Ankyrin via L1 to the substratum exerts a pulling force on the filopodial F-Actin during the outgrowth of neurites, which is dependent on the binding of Ankyrin to Spectrin (Ohler, 2011).

The similar loss-of-function phenotypes of hts and Spectrin mutants indicate an intimate link between Hts and Spectrin. It is proposed that Hts links the filopodial F-Actin in a Spectrin-dependent manner via Gogo or another transmembrane protein to the substratum, which inhibits retrograde Actin flow and lets the growth cone steer straight forward towards the side where Hts is activ. When Gogo is bound to its as yet unidentified ligand, it could act as a repulsive receptor and remove Hts from the lateral filopodia, thereby assuring the proper spacing of single R8 axons. This fits with the suggested R8-R8 repulsion mediated by Gogo. Additionally, Gogo has been shown to have another adhesive function, which fits with the proposed adhesive function of Hts mediated by Gogo. This could also explain the thickening of R8 axons caused by excessive Hts, as the result of an excessive anchoring of the growth cone (Ohler, 2011).

Moreover, like Adducin, Hts may also serve as an Actin-capping protein and the abnormal swellings caused by an excess of Gogo may be the consequence of increased Actin polymerization due to an abnormally low level of Hts. When both Hts and Gogo are overexpressed, abnormally thick swellings are not observed. Although the anchoring of R8 growth cones should be abnormally strong in this situation, the Hts-antagonizing function of Gogo could also be increased and counteract this elevated adhesive force (Ohler, 2011).

An interesting finding from this work is that the MARCKS domain seems not to be required for the functions of Hts during axon guidance. Expression of both Add1, the Hts isoform including the MARCKS domain that is most closely related to mammalian Adducin, and of isoform HtsPD, an isoform lacking the MARCKS domain, in Rs rescues the defective phenotype caused by the loss of hts. This is consistent with the observation that homozygous hts^ΔG mutant flies, which have only truncated Hts protein lacking the MARCKS-related domain, do not show defects in the medulla. Both were surprising, since the MARCKS-related domain of Adducin is required for its activity in promoting association of Spectrin with Actin filaments as well as the Actin capping and Actin binding activity of Adducin in vitro. How can it be that the MARCKS domain is not required for the function of Hts during axon guidance, although the MARCKS domain has been shown to be required for the in vitro functions of Adducin including Actin binding, Actin capping, and Spectrin recruiting (Ohler, 2011)?

A possible explanation is that the function of Hts during axon guidance is indeed independent of Actin and Spectrin and that Hts serves a completely novel function here. However, for several reasons it is thought more likely that the Drosophila Hts can interact with Actin and Spectrin as the mammalian Adducin does, but does not strictly require its MARCKS domain in order to do so. There are several reasons for this assumption:
First, in the Drosophila germ line the Hts isoforms ShAdd and Ovhts are exclusively expressed, both lacking the MARCKS domain. Nevertheless, in hts mutants, the fusome, a Spectrin-based cytoskeletal structure in the germarium, is disorganized. This indicates that these MARCKS domain lacking proteins are required for the proper assembly of the Spectrin cytoskeleton (Ohler, 2011).

Second, again the similar phenotype of hts and Spectrin mutants suggest that Hts and Spectrin are functionally linked during R axon guidance. Since the hts mutant phenotype can be rescued by Hts lacking MARCKS, this indicates that the interaction with Spectrin does not require the MARCKS domain (Ohler, 2011).

Third, it has not been shown directly that mammalian Adducin binds Spectrin via its MARCKS domain. The MARCKS domain is the target of many regulatory processes. It contains phosphorylation sites for PKA and PKC and it binds calmodulin in a Ca²⁺-dependent manner. The MARCKS domain could function merely in a regulatory manner, regulating the binding of Spectrin to another part of Adducin. Indeed, the neck domain is also required for Adducin binding to Spectrin and may, therefore, contain the actual binding site (Ohler, 2011).

Although the MARCKS-related domain is dispensable for the function of Hts in R axon guidance, some part of the remaining tail domain seems to be essential, as ShAdd, the Hts isoform that does not contain the tail domain, fails to rescue flies from the defects caused by hts mutations. Moreover, ShAdd can not be detected in R axons in the medulla. It was not ascertain if this absence is due to reduced translation, degradation, impaired transport to, or efficient removal from the axon. Since ShAdd appears to be expressed and stable in Schneider cells and larval eye-brain complexes, the idea is favored that the tail domain is required for the localization of Hts to the axon. In any case, the absence of ShAdd protein from the axon could account for the failure of ShAdd to rescue the defects caused by hts (Ohler, 2011).

This work has demonstrated that the axon guidance receptor Gogo physically interacts with the cytoskeletal protein Hts to guide Drosophila R axons to their correct target in the medulla. Although there are some indications for a synergistic interaction, it was shown that Hts and Gogo also antagonize each other. This points to a highly sophisticated mechanism of the Gogo-Hts complex. Further work will be required to dissect the different aspects of this intricate complex and will shed light not only on its role in Drosophila R axon guidance, but also on the role of Hts/Adducin and, in succession, the Spectrin cytoskeleton in other systems (Ohler, 2011).

Search PubMed for articles about Drosophila Golden goal

Hakeda-Suzuki, S. et al. (2011). Golden Goal collaborates with Flamingo in conferring synaptic-layer specificity in the visual system. Nat. Neurosci. 14(3): 314-23. PubMed ID: 21317905

Mann K., et al. (2012). A putative tyrosine phosphorylation site of the cell surface receptor Golden goal is involved in synaptic layer selection in the visual system. Development Feb;139(4): 760-71. PubMed ID: 22241840

Ohler S, Hakeda-Suzuki S, Suzuki T. (2011) Hts, the Drosophila homologue of Adducin, physically interacts with the transmembrane receptor Golden goal to guide photoreceptor axons. Dev. Dyn. 240(1): 135-48. PubMed ID: 21128303

Takayanagi S, et al. (2006). Genetic marking of hematopoietic stem and endothelial cells: identification of the Tmtsp gene encoding a novel cell surface protein with the thrombospondin-1 domain. Blood 107(11): 4317-25. PubMed ID: 16455951

Tomasi, T., et al. (2008). The transmembrane protein Golden goal regulates R8 photoreceptor axon-axon and axon-target interactions. Neuron 57(5): 691-704. PubMed ID: 18341990

date revised: 12 June 2021

Home page: The Interactive Fly © 2011 Thomas Brody, Ph.D.