Glial cells are crucial regulators of synapse formation, elimination, and plasticity. In vitro studies have begun to identify glial-derived synaptogenic factors, but neuron-glia signaling events during synapse formation in vivo remain poorly defined. The coordinated development of pre- and postsynaptic compartments at the Drosophila neuromuscular junction (NMJ) depends on a muscle-secreted retrograde signal, the TGF-β/BMP Glass bottom boat (Gbb). Muscle-derived Gbb activates the TGF-β receptors Wishful thinking (Wit) and either Saxophone (Sax) or Thick veins (Tkv) in motor neurons. This induces phosphorylation of Mad (P-Mad) in motor neurons, its translocation into the nucleus with a co-Smad, and activation of transcriptional programs controlling presynaptic bouton growth. This study shows that NMJ glia release the TGF-β ligand Maverick (Mav), which likely activates the muscle activin-type receptor Punt to potently modulate Gbb-dependent retrograde signaling and synaptic growth. Loss of glial Mav results in strikingly reduced P-Mad at NMJs, decreased Gbb transcription in muscle, and in turn reduced muscle-to-motor neuron retrograde TGF-β/BMP signaling. It is proposed that by controlling Gbb release from muscle, glial cells fine tune the ability of motor neurons to extend new synaptic boutons in correlation to muscle growth. This work identifies a novel glia-derived synaptogenic factor by which glia modulate synapse formation in vivo (Fuentes-Medel, 2012).

Given the prominent role of glia and TGF-β signaling during synaptic bouton formation at the larval neuromuscular junction (NMJ), this study sought to determine whether ligands of the TGF-β superfamily are expressed in peripheral glia at the larval stage by real-time PCR. For this analysis, total RNA was isolated from segmental nerves in which the only cell bodies were peripheral glia and no neuronal RNAs were detected. This analysis revealed the presence of several transcripts for TGF-β ligands in glia, including Myoglianin (MYO), Dawdle (Daw), and Maverick (Mav). In contrast, Activin β (Actβ) transcripts were not detected in nerves. To explore a potential involvement of glia-derived TGF-β ligands in signaling synaptic development, RNAi transgenes targeting Daw, Mav, and MYO transcripts were expressed specifically in peripheral glia using the Gal4 driver Gli-Gal4 (rl82-Gal4). Downregulating Mav and Daw, but not MYO, in NMJ glia substantially reduced NMJ size, as determined by counting the number of synaptic boutons at the third-instar larval stage. In the case of Mav, the number of branches was also reduced (the number of branches in Mav-RNAi glia is 9.13 ± 0.70 compared with 17.0 ± 0.50 in controls) (Fuentes-Medel, 2012).

A classical readout of TGF-β pathway activation is Mad phosphorylation (P-Mad). At the Drosophila larval NMJ, TGF-β activation through P-Mad detection has been documented in both motor neuron nuclei and synaptic boutons of the NMJ. At synaptic boutons, P-Mad signal is organized into discrete puncta decorating synaptic boutons. Notably, downregulating Mav in peripheral glia with two different Mav-RNAi constructs targeting different regions of the Mav transcript virtually eliminated or severely reduced P-Mad immunoreactivity at synaptic sites. In contrast, downregulating MYO had no effect. A decrease in P-Mad signal was also observed by downregulating Daw in glia, but this effect was much weaker (Fuentes-Medel, 2012).

To assess whether Mav was exclusively required in glia for activation of TGF-β signaling, Mav-RNAi was expressed either muscles or motor neurons. However, no significant change was observed in synaptic P-Mad levels as a result of these manipulations. Thus, Mav is exclusively required in glia for activation of TGF-β signaling at synaptic sites. Further evidence for such requirement was obtained by examining the effect of overexpressing Mav in glia, which resulted in an increase of P-Mad signal intensity at the NMJ. Although there was no significant increase in the number of synaptic boutons at muscles 6 and 7 in parallel with the increase in synaptic P-Mad intensity, the number of boutons at muscle 4 was significantly increased. This increase was primarily due to an increase in the number of satellite boutons (small boutons emerging from normally sized synaptic boutons), which were significantly increased at muscles 6/7 and 4 (Fuentes-Medel, 2012).

In contrast to glia, downregulating Mav in motor neurons or muscles did not significantly change NMJ size. Interestingly, in vitro studies of the synaptogenic effects of Xenopus Schwann cells (SCs) led to the proposal that SC-derived TGF-β1 could promote synaptogenesis. However, the exact source of the synaptogenic signal was not clear, and whether TGF-β1 plays a role in the intact organism remains to be determined. These observations provide direct evidence indicating that a glia-derived TGF-β ligand can influence both TGF-β signaling and synaptic growth in vivo at the NMJ (Fuentes-Medel, 2012).

The requirement for peripheral glia in TGF-β pathway activation and synaptic bouton growth led to a prediction that peripheral glia should be capable of releasing Mav. An anti-Mav peptide antibody was generated and it was found labeled bright puncta within peripheral glial cells, consistent with the detection of Mav transcript in these cells. This labeling was specific, because it was almost completely eliminated by expressing Mav-RNAi in peripheral glia. To determine whether Mav could be released by peripheral glia, transgenic flies were generated expressing a green fluorescent protein (GFP)-tagged Mav transgene, and its expression was drivedn in NMJ glia with the rl82-Gal4 driver. This transgene was likely to be functional, because it behaved just like the untagged Mav transgene. In particular, expressing the GFP-tagged transgene in glia induced a significant increase in the number of boutons and satellite boutons (the number of boutons at muscle 4 is 63.3 ± 3.7 when Mav-GFP is expressed in glia, compared with 62.1 ± 3.1 when untagged Mav is expressed in glia, and 39.5 ± 2.0 in controls; the number of satellite boutons at muscle 4 is 19.5 ± 1.5 when Mav-GFP is expressed in glia, compared with 19.8 ± 1.1 when untagged Mav is expressed in glia and 5.5 ± 0.9 in controls. As with endogenous Mav, Mav-GFP became distributed in bright GFP puncta within glia in the segmental nerves. In addition, Mav-GFP was prominent at glial extensions that interact with the NMJ, showing that Mav-GFP is efficiently transported to these glial extensions. Notably, bright GFP puncta were also observed beyond the boundary of glial extensions, indicating that Mav-GFP could be released by peripheral glial cells. Close observation of the Mav-GFP puncta outside the glial membrane extensions revealed their localization in close association with both synaptic boutons and the postsynaptic junctional region of the muscle. In contrast, expressing Mav-GFP in neurons resulted in punctate and diffuse GFP staining within synaptic boutons, but no GFP signal was observed beyond the boundary of synaptic boutons, showing that Mav-GFP is likely not released from synaptic boutons. Similarly, expressing Mav-GFP in muscles resulted in very dim GFP signal in muscle cells, but this signal did not localize to the NMJ. Thus, Mav mRNA and protein are present in peripheral glia, glia can secrete transgenic Mav, and secreted Mav associates with synaptic boutons and muscles. Previous studies demonstrated that peripheral glial membrane extensions at the NMJ are dynamic, extending and retracting processes that become associated with synaptic boutons and muscles (Fuentes-Medel, 2009). Therefore, it is possible that glial membrane extensions might directly deposit Mav as they dynamically interact with boutons or muscles. Alternatively, glia-derived Mav may be released from these processes and diffuse to relatively distant sites (Fuentes-Medel, 2012).

The finding that glia can release Mav, and that glia-derived Mav is required for TGF-β signaling at synaptic sites, raised the question as to which cells (neurons or muscles) respond to Mav. The synaptic P-Mad signal is virtually absent when Mav is downregulated in glia. However, whether the P-Mad signal is pre- or postsynaptic has been a matter of debate. One study reported that synaptic P-Mad partially colocalized with the presynaptic active zone marker Bruchpilot (BRP) while it did not colocalize with Discs-large (DLG), suggesting that P-Mad is presynaptic. However, DLG is localized at the perisynaptic region within the pre- and postsynaptic compartments, and thus it is not expected to colocalize with the postsynaptic density. Another study reported that in wit mutants, the synaptic P-Mad signal was eliminated. Given the role of Wit receptors in activating TGF-β signaling in motor neurons, it was concluded that P-Mad was presynaptic. However, whether Wit is also expressed in muscles is unknown. In a third report, the synaptic P-Mad signal was found to completely colocalize with a tagged glutamate receptor GluRIIA transgene, which suggested a postsynaptic P-Mad localization. However, a comparison with endogenous GluRs was not done. To address this issue more directly, a strong hypomorphic mad mutant, mad¹², over a mad deficiency chromosome was used and P-Mad labeling at the NMJ was examined. P-Mad immunoreactivity was completely eliminated in this mutant, providing direct evidence for the specificity of the P-Mad signal observed at the NMJ. Mad was then downregulated in either motor neurons or muscles by expressing Mad-RNAi and examined the intensity of the synaptic P-Mad signal. It was found that downregulating Mad in either neurons or muscles resulted in a significant decrease in P-Mad signal intensity, suggesting both a presynaptic and postsynaptic localization of P-Mad. Importantly, the number of synaptic boutons was significantly reduced by downregulation of Mad in either neurons or muscles, arguing strongly for a requirement for Mad function in both cell types (Fuentes-Medel, 2012).

The localization of P-Mad was examined in comparison with the endogenous localization of GluRIIA and BRP. Confirming previous reports with the GluRIIA transgene, it was found that synaptic P-Mad was always present within the boundaries of endogenous GluRIIA clusters. In contrast, only partial colocalization between BRP and P-Mad was observed, and the signals appeared juxtaposed. However, given that active zones and postsynaptic GluR clusters are apposed to each other in close proximity (<40 nm), it is noted that light microscopy alone cannot resolve this issue. Nevertheless, the finding that downregulating Mad in either muscles or neurons leads to both P-Mad reduction and NMJ growth defects is a strong indication that the signal is localized to both types of cells (Fuentes-Medel, 2012).

The finding that synaptic P-Mad can be attributed to muscles in addition to neurons, and that this signal is virtually eliminated by downregulating Mav in NMJ glia, provided compelling evidence that a TGF-β signal is activated in muscles as previously proposed. To confirm this finding, an additional reporter of TGF-β pathway activation, the transcription of daughters against dpp (Dad), an inhibitory Smad that antagonizes TGF-β signaling, was examined. Real-time PCR from larval body wall muscle RNA demonstrated that dad transcript was significantly decreased upon expression of Mav-RNAi in glia, providing additional support for a role of glia-derived Mav in activating TGF-β signaling in muscles (Fuentes-Medel, 2012).

It has been previously reported that the transcription of the Glass bottom boat (Gbb) retrograde ligand is also regulated by TGF-β pathway activation. Interestingly, downregulating Mav in glia resulted in a significant decrease in muscle gbb transcript levels. In contrast, cyclophilin control transcript levels were not affected. This observation raised the possibility that glial cells could modulate the intensity of the retrograde signal. To test this model, P-Mad levels were examined in the nuclei of motor neurons (a readout for retrograde TGF-β pathway activation). Expression of Mad-RNAi in motor neurons led to a drastic decrease in P-Mad immunoreactivity at motor neuron nuclei, demonstrating that the P-Mad signal at this site is specific. Most importantly, and consistent with the model, P-Mad immunoreactivity levels were significantly decreased in the nuclei of larval motor neurons when Mav-RNAi, but not MYO-RNAi or Daw-RNAi, was expressed in peripheral glia (Fuentes-Medel, 2012).

Finally, the levels of a TGF-β target gene in motor neurons, Trio, was also examined. Trio is a Rac-activating protein that contributes to cytoskeletal remodeling during synaptic growth. Previous studies demonstrated that upon activation of motor neuron TGF-β signaling by muscle Gbb, trio transcription is upregulated. Real-time PCR revealed that trio transcript levels were significantly reduced in RNA isolated from larval brains when Mav-RNAi was expressed in peripheral glia. In contrast, cyclophilin control transcript levels were unchanged by this manipulation. These results provide strong evidence that glia-derived Mav also modulates motor neuron TGF-β signaling. This modulation might occur through direct interaction of Mav with TGF-β receptors in motor neurons, or by regulating the levels of Gbb in muscles, leading to a change in Gbb release (Fuentes-Medel, 2012).

In support of the above model, it wass found that Mav function in NMJ development depended on Gbb. As noted, overexpressing Mav in glia results in a significant increase in the number of boutons. In contrast, no change in bouton number was observed in gbb/+ heterozygous larvae. Notably, the increase in bouton number observed upon overexpression of Mav in glia was completely suppressed in gbb/+ heterozygous larvae. These results point to a role of Gbb in mediating the response to glia-derived Mav during NMJ development (Fuentes-Medel, 2012).

Evidence was also found pointing to Punt as the likely receptor mediating the muscle response to glia-derived Mav. Downregulating Punt in muscle using two different RNAi lines targeted to different regions of Punt and the C57-Gal4 driver resulted in a substantial decrease in bouton number compared with the driver control. In addition, downregulating Punt in muscle decreased the levels of P-Mad immunoreactivity in motor neuron nuclei (Fuentes-Medel, 2012).

In summary, these studies provide direct in vivo evidence that glial cells secrete a TGF-β ligand, Mav, that influences the magnitude of Gbb-mediated muscle-to-motor neuron retrograde signaling. By controlling TGF-β signaling in both muscles and neurons, glial cells are well positioned to integrate the coordinated development of pre- and postsynaptic compartments (Fuentes-Medel, 2012).

It is interesting to note, however, that there could be partial redundancy between the two activin-type ligands, Mav and Daw, because glial knockdown of either leads to decreased NMJ growth and synaptic P-Mad, although the effect of Mav downregulation is substantially more severe. However, only glial knockdown of Mav affected levels of motor neuron nuclear P-Mad, suggesting that the pathways activated by the two ligands diverge. This divergence could result from the use of different receptor isoforms. For example, the activins Actβ and Daw can have different effects on the same tissues, as a result of differential interactions with alternative Baboon isoforms. Nevertheless, this work identifies a novel glial synaptogenic factor and provides compelling evidence for a critical role for glia in modulation of synapse assembly at the NMJ in vivo (Fuentes-Medel, 2012).

The transforming growth factor-beta (TGF-beta) superfamily of structurally related ligands regulates essential signaling pathways that control many aspects of cell behavior in organisms across the phylogenetic spectrum. This study reports the identification of maverick (mav), a gene that encodes a new member of the TGF-beta superfamily in Drosophila. Phylogenetic analysis and sequence comparison suggest that Mav cannot be easily assigned to any one sub-family, since it is equally related to BMP, activin and TGF-beta ligands. mav maps to the fourth chromosome and is expressed throughout development. In situ hybridization experiments reveal the presence of maternally derived mav transcript in precellular blastoderm embryos. Later in development, mav is expressed in a dynamic pattern in the developing gut, both in endodermal and visceral mesodermal cells. In adult females, high levels of mav mRNA are present in late stage egg chambers (Nguyen, 2000).