PAK-kinase

Protein interactions during dorsal closure

The Rho subfamily of Ras-related small GTPases participates in a variety of cellular events including organization of the actin cytoskeleton and signaling by c-Jun N-terminal kinase and p38 kinase cascades. These functions of the Rho subfamily are likely to be required in many developmental events. A study has been performed of the participation of the Rho subfamily in dorsal closure (DC) of the Drosophila embryo, a process involving morphogenesis of the epidermis. In this study (Harding, 1999), and one published subsequently on the same problem (Ricos, 1999), a distinction is made between two types of cells at the leading edge: one type is termed 'cells flanking segment borders' and the second type is termed 'segment border cells'. The two cell types alternate along the anterior to posterior axis. Drac1, a Rho subfamily protein, is required for the presence of an actomyosin contractile apparatus believed to be driving the cell shape changes essential to DC. Expression of a dominant negative Drac1 transgene causes a loss of this contractile apparatus from the leading edge of the advancing epidermis and dorsal closure fails. It is postulated that Drac1 triggers the JNK cascade within cells flanking segment borders. Dpp is the target of the JNK cascade in the cells flanking segment borders, and Dpp is released by these flanking cells, targeting the adjacent segment border cells (Harden, 1999).

Two other Rho subfamily proteins, Dcdc42 and RhoA, as well as Ras1 are also required for dorsal closure. Dcdc42 appears to have conflicting roles during dorsal closure: establishment and/or maintenance of the leading edge cytoskeleton versus its down regulation. It is proposed that Dcdc42 is the target of Dpp, released from cells flanking segment borders. Dpp targets Dcdc42 in the adjacent segmental border cells through Dpp's receptors Thick veins and Punt (Ricos, 1999). Down regulation of the leading edge cytoskeleton may be controlled by the serine/threonine kinase PAK, a potential Dcdc42 effector. Whereas dominant negative Dcdc42 leads to a loss of Pak, constitutive active Dcdc42 induces an increase in Pak levels. Paradoxically, constitutively active Dcdc42 can also lead to a loss of leading edge cytoskeleton and leading edge Pak in some embryos. These two seemingly opposite effects of constitutively active Dcdc42 can occur in the same embryo. It is possible that Pak, activated by Dcdc42 in segmental border cells, is localized within the leading edge cytoskeleton and contributes to cytoskeletal down regulation. This breakdown is first executed transiently in the segment border cells and then permanently along the leading edge at the end of DC. Thus constitutively active Dcdc42 may be capable of elevating Pak sufficiently along the leading edge to cause premature down regulation of Pak and the cytoskeleton as a whole (Harden, 1999).

RhoA is required for the integrity of the leading edge cytoskeleton specifically in the cells flanking segment borders. Dominant negative RhoA expression leads to loss of leading edge components and a loss of anterior-posterior contraction in several cells flanking each segment border. One possible explanation for the effects of dominant negative RhoA is that RhoA is required for the assembly of the leading edge cytoskeleton in the segment border cells, perhaps downstream of Drac1. An alternative interpretation is that RhoA is functioning as a negative regulator of the leading edge cytoskeletal losses that occur in wild-type embryos, and that when Rho A function is absent, these losses become permanent instead of transient. The interactions of the various small GTPases in regulating dorsal closure reveals that there is no evidence for the hierarchy of Rho subfamily activity described in some mammalian cell types. Rather, the results suggest that while each of the p21s are required for dorsal closure, they act largely in parallel (Harden, 1999).

A model is given of the control of DC by the Drac1/JNK and Dcdc42/Dpp pathways. Drac1/JNK signaling, initiated by an as yet unknown factor, assembles cytoskeletal components (F-actin, myosin and focal complexes) and other proteins (Dpp, Puckered and Pak) in the leading edge cells and initiates the cellular migration that characterizes DC. Dpp-activated signaling controls the dynamics of epidermal migration, via Dcdc42 and the Dpp pathway, through the serine/threonine kinase Pak, which transiently downregulates the leading edge cytoskeleton at the segmental borders. Transient downregulation of the actin cytoskeleton and focal contacts near the segment border cells is likely to cause local relaxation of the anterior-posterior tension along the LE. Such transient relief of tension may then limit excessive migration of leading edge cells toward each other and prevent the bunching and shearing of epidermal segments that occurs following impairment of Dpp/Dcdc42 signaling. Segment borders cells are potential regions of highest Dpp signaling, because they are adjacent to the highest local concentrations of Dpp protein, and they have high levels of Pak protein and transcripts for the Tkv receptor. Segmental border cells are the only places where transient downregulation of the leading edge cytoskeleton is ever seen in wild-type embryos during DC. As such, it is proposed that the role of Dcdc42/Dpp signaling is the induction of Pak to downregulate the leading edge cytoskeleton at the segment borders, introducing a degree of flexibility to the leading edge during the dorsal closure process (Ricos, 1999).

Expression of dominant negative Ras causes partial loss of the leading edge cytoskeleton, and constitutively active Ras increases Pak levels at the leading edge. Thus, Ras1 has phenotypic effects similar to those of Dcdc42 and Drac1. There is mounting evidence that the Rho subfamily proteins lie downstream of Ras, contributing to the ability of Ras to cause transformation and regulate the actin cytoskeleton. The results of this study are consistent with Ras1 activating Dcdc42 and/or Drac1 during DC, although given that Ras1 expression has milder phenotypic effects than Dcdc42 or Drac1 during DC, it is not proposed that Ras1 is a chief activator of either of these p21s. The finding that constitutively active Ras1 can increase Pak levels at the leading edge is the first demonstration of Ras having an effect on the behavior of a Pak family member. Interestingly, it has recently been shown in mammals that kinase deficient PAK1 mutants can inhibit Ras transformation, indicating that PAK may be a component of Ras signaling (Tang, 1997 and Tang, 1998). Although constitutively active Ras1, like constitutively active Cdc42, elevates Pak levels at the leading edge, it does not cause the loss of leading edge components seen following constitutively active Cdc42 expression. Looking at these results in the context of the model for Pak function, it may be that constitutively active Ras1 does not increase Pak accumulation at the leading edge to a level sufficient to cause down-regulation of the cytoskeleton (Harden, 1999).

Protein interactions during axon guidance

Mammalian Paks bind Nck (Bokoch, 1996 and Galisteo, 1996). The interaction sites have been mapped to the SH3-2 domain of Nck and the N-terminal-most PXXP site in Pak. Three proteins highly related to mammalian Paks have been identified in Drosophila. Two of these, Mbt (Melzig, 1998) and DPak2 [G. Suh, unpublished observations communicated to Hing, 1999]), neither bind to Dock nor are required for R cell axon guidance. The third, called Pak, the subject of this report, contains an N-terminal PXXP site highly related to mammalian Paks, which bind to Nck (Harden, 1996). A test was performed to see whether Drosophila Pak interacts with Dock through these sites in a yeast two-hybrid assay. Indeed, Pak and Dock do interact. The DNA-binding domain of Lex A fused to full-length Dock (LexA-Dock) interacts strongly with a fusion protein containing the Gal4 activation domain and full-length Pak (GAD-Pak). The SH3-2 domain of Dock is necessary for interaction with Pak. LexA-Dock fusion proteins carrying mutations in the SH3-1 and SH3-3 domains, designed to disrupt interactions with proline-rich sequences, exhibit a similar level of interaction to that seen for LexA-Dock. Conversely, the same mutations introduced into the SH3-2 domain abolish this interaction. It has been demonstrated that the SH3-2 domain is not only necessary, but it is sufficient for binding to Pak. While Lex-SH3-2 interacts strongly, neither LexA-SH3-1 nor LexA-SH3-3 interacts with GAD-Pak (Hing, 1999).

The requirement for the N-terminal PXXP site of Pak for these interactions was demonstrated in a separate series of experiments in which LexA-Dock was tested for interactions with three different forms of the N-terminal region of Pak fused to GAD. The N-terminal region of Pak interacts strongly with LexA-Dock. Conversely, LexA-Dock does not interact with Pak N-terminal fragments containing point mutations in which arginine 14 is changed to methionine (R14M) or proline 9 is changed to leucine (P9L) (Hing, 1999).

To assess whether these proteins can associate in vivo, immunoprecipitation experiments were carried out in Drosophila S2 cells. Both Dock and Pak are endogenously expressed in these cells. S2 cell lysates were incubated with either anti-Dock antibody or a control antiserum and precipitated with protein A beads. The immunoprecipitates were analyzed on Western blots with anti-Pak antibodies. Pak was found in the anti-Dock immunoprecipates, but not in the controls (Hing, 1999).

The requirements for different domains of Pak in R cell axon guidance were assessed through analysis of EMS-induced alleles and in rescue experiments using GMR-Pak transgenes carrying specific mutations. Transgenic Paks driven by the GMR promoter were expressed at higher levels than endogenous Pak, as assessed by immunohistochemistry and Western blot analyses of eye-brain complexes from transgenic animals. Hence, the failure of Pak mutant transgenes to rescue the mutant phenotypes is not a consequence of reduced levels of protein expression. None of the mutant GMR-Pak transgenes used in these experiments induce dominant phenotypes. The analyses in this section assess the requirement of three regions of Pak in R cell axon guidance: the kinase domain, the CRIB domain, and the Dock-binding site. The kinase activity of Pak is essential for R cell axon guidance. Two EMS-induced strong loss-of-function mutations, Pak³ and Pak⁵, lead to amino acid substitutions in conserved residues in the catalytic site. The Pak³ mutation leads to the replacement of glycine at position 569 with aspartic acid. This change, G569D, is found within the DFG triplet conserved in all kinases. The change in Pak⁵ (D553N) is in a residue conserved in all Paks. These mutations are in subdomains (VI and VII) implicated in ATP binding. Both Pak³ and Pak⁵ proteins are expressed at wild-type levels, suggesting that it is the lack of kinase activity, rather than protein instability, that leads to the mutant phenotype. This requirement for kinase activity was confirmed by the observation that a Pak transgene in which the invariant lysine required for the phosphotransfer reaction was changed to alanine (GMR-Pak^K459A) also failed to rescue the Pak mutant phenotype (Hing, 1999).

Drosophila PAK-kinase binds Rac and Cdc42 proteins (Harden, 1996). Pak kinase activity is stimulated by Rac or Cdc42. This occurs through direct binding of Rac/Cdc42 in the GTP-bound form to the CRIB domain. Previous studies have demonstrated that substituting leucines for two conserved histidines in the CRIB site of mammalian Pak prevents this interaction. Similarly, it has been demonstrated in the yeast two-hybrid assay that this mutation prevents binding of Drosophila Pak to Cdc42. A Pak transgene carrying these mutations fails to rescue the Pak mutant phenotype, indicating that interaction between Pak and Cdc42/Rac is necessary for R cell growth cone guidance (Hing, 1999).

DNA sequence analysis reveals a missense mutation that changes a proline within the conserved Dock-binding site to a leucine (P9L) in Pak⁴. This substitution abolishes the interaction between Dock and Pak in the yeast two-hybrid assay. Pak⁴ protein is expressed at wild-type levels in eye-brain complexes and is localized to the optic lobe neuropil as in wild type. R cell axon guidance defects in Pak⁴/Pak¹¹ are indistinguishable from the strong loss-of-function phenotypes associated with Pak⁶/Pak¹¹ and dock null alleles. Hence, the Dock-binding site in Pak is essential for its function. This finding and the similarities between the dock and Pak mutant phenotypes provide strong evidence that direct interaction between Dock and Pak is essential for R cell axon guidance (Hing, 1999).

Slit stimulation recruits Dock and Pak to the Roundabout receptor and increases Rac activity to regulate axon repulsion at the CNS midline

Drosophila Roundabout is the founding member of a conserved family of repulsive axon guidance receptors that respond to secreted Slit proteins. Evidence is presented that the SH3-SH2 adaptor protein Dreadlocks (Dock), the p21-activated serine-threonine kinase (Pak), and the Rac1/Rac2/Mtl small GTPases can function during Robo repulsion. Loss-of-function and genetic interaction experiments suggest that limiting the function of Dock, Pak, or Rac partially disrupts Robo repulsion. In addition, Dock can directly bind to Robo's cytoplasmic domain, and the association of Dock and Robo is enhanced by stimulation with Slit. Furthermore, Slit stimulation can recruit a complex of Dock and Pak to the Robo receptor and trigger an increase in Rac1 activity. These results provide a direct physical link between the Robo receptor and an important cytoskeletal regulatory protein complex and suggest that Rac can function in both attractive and repulsive axon guidance (Fan, 2003).

Strong defects in embryonic axon guidance are observed only when both the maternal and zygotic components of dock function are removed. In these maternal minus dock mutants (dock^mat), phenotypes reminiscent of loss of robo function can often be seen. dock^mat embryos examined with an antibody that labels all axons frequently show thickening of commissural axon bundles and a commensurate reduction in the thickness of longitudinal axon bundles. Staining these embryos with an antibody that selectively labels noncrossing axons (anti-fasII) reveals a significant degree of ectopic midline crossing. These phenotypes are similar to, but considerably less severe than, those observed in robo mutants. The similarity in mutant phenotypes that is observed provides genetic support for the idea that dock could contribute to Robo repulsion (Fan, 2003).

If dock and robo function together during midline guidance, they should be coexpressed in embryonic axons. This is indeed the case. Double labeling of embryos with antibodies raised against Dock and Robo reveals substantial coexpression of the two proteins. Both Dock and Robo show enriched expression on CNS axons beginning as early as stage 12, corresponding to the time of initial axon outgrowth. At these early stages of axon growth, Dock is detected in the pCC axon, a cell known to express Robo, as revealed by double labeling with FasII. Interestingly, while Robo shows a regionally restricted expression pattern with high levels of expression on longitudinal portions of axons and low levels in commissural axons, Dock is expressed equivalently in both commissural and longitudinal axon segments. This observation raises the possibility that Dock could have additional roles in the guidance of commissural axons not shared by Robo. These observations show that Dock and Robo are both present at the right time and place to function together during midline repulsion (Fan, 2003).

Biochemical data suggests that the interaction between Dock and Robo is an SH3-dependent interaction and that the first two SH3 domains of Dock are most important for mediating Robo binding. Based on the observations that a three-protein interaction can be detected between Robo, Dock, and Pak and that Pak has been shown to interact with the SH3-2 domain of Dock, it is believed that the SH3-1 domain is the most important for Robo and Dock binding. Furthermore, Slit stimulation enhances Dock's ability to bind to Robo, suggesting a ligand-regulated SH3 domain interaction. This represents a different kind of adaptor interaction to many that have been observed previously, where Nck appears to interact with a number of tyrosine-kinase receptors through an SH2 domain/phosphotyrosine interaction. In the latter case, how ligand binding to the receptor regulates the Nck SH2 domain interaction is quite well understood. The observation that the Robo receptor shows a ligand-regulated SH3 domain interaction with Dock/Nck suggests that somehow ligand binding results in an increased availability of the SH3 binding sites in the receptor (Fan, 2003).

Perhaps the most difficult observation to explain is how reciprocal shifts in Pak levels can lead to similar consequences for Robo repulsion. Since the enhancing effects of Pak overexpression in partial loss-of-function robo backgrounds are more dramatic with the membrane-tethered form of Pak, it is tempting to speculate that in order to signal properly, turning Pak activity on and off needs to be tightly controlled. Little is known about how Pak signaling is terminated and it seems quite possible that the membrane-tethered version of pak is not as effectively regulated as the wild-type form of pak. Interestingly, in genetic backgrounds where robo signaling is specifically compromised in its output, through reduction of rac, introducing the UASPak^Myr transgene can partially suppress the midline crossing defects. Given the clear ability of alterations in pak expression to modulate midline repulsion and the observation that Slit can promote the formation of a Robo, Dock, and Pak protein complex, it is somewhat surprising that complete removal of zygotic pak does not have major consequences for embryonic axon guidance. Indeed, in the absence of clear loss-of-function phenotypes in pak mutants, it is difficult to argue unequivocally for a critical role of endogenous pak in robo function. There are a number of potential explanations for these observations including, but not limited to, maternal pak contribution and the potential redundant function of a second pak-like gene. Future experiments should address these possibilities in order to link pak more firmly to robo (Fan, 2003).

Dock has been suggested to act downstream of the Dscam axon guidance receptor during pathfinding of Bolwig's nerve, and the vertebrate homolog of Dock, Nck, has also been linked to several guidance receptors in vitro, including Eph receptors and c-Met receptors. More recently, Nck has been shown to directly interact with the cytoplasmic domain of the vertebrate attractive Netrin receptor DCC. The Nck and DCC interaction is important for DCC's function to stimulate axon extension in vitro. Together these observations raise the question of whether a similar DCC/Nck interaction occurs in Drosophila, and if so whether the interaction is important for the in vivo function of Drosophila DCC (encoded in the fly by the frazzled gene) to attract commissural axons across the midline. Interestingly, in addition to its substantial overlap in expression with the Robo receptor, Dock protein is also expressed in commissural portions of axons, as is the Frazzled receptor. While the dock mutant phenotype is most consistent with a role in midline repulsion, an additional function in attraction cannot be ruled out. In the future it will be interesting to test for genetic interactions between frazzled, dock, Rac, and pak to determine if this signaling module is also employed during midline axon attraction in Drosophila (Fan, 2003).

The implication of Dock/Nck and Rac in both DCC-mediated attraction and Robo-mediated repulsion raises the obvious question of how the specificity of attraction and repulsion is controlled and argues against a committed role of either of these signaling molecules to either one or the other type of responses. This is perhaps not too surprising, given the fact that Robo and DCC receptors themselves are intimately connected through their ability to form a heteromeric receptor complex with potentially unique signaling properties. Although it remains possible that signaling molecules or adaptors will be identified that can account for the specificity, an alternative possibility is that it is the coordinate regulation, relative activity levels, and combinatorial action of a core group of common signaling molecules that makes the difference in attraction versus repulsion (Fan, 2003).

Biochemical data support the idea that Slit stimulation of Robo can regulate the recruitment of Dock and Pak to the Robo receptor and also trigger an increase in Rac activity. Both of these events are dependent on the CC2 and CC3 sequences in Robo's cytoplasmic domain. Thus, the observations are consistent with either a Dock-dependent or a Dock-independent recruitment of Rac to Robo. Based on the known physical interactions between Dock and Pak and between Pak and Rac, it is likely that the recruitment of Rac is dependent on Dock. Alternatively, another protein interacting through CC2 and/or CC3 could function to recruit Rac in a Dock-independent fashion (Fan, 2003).

Regardless of whether the recruitment of Rac to Robo is dependent on Dock and Pak or is an independent event, the data cannot explain how Slit stimulation of Robo results in increased Rac activity. Two obvious types of molecules that are missing from the model and the protein complex are the upstream regulators of Rac, the GEF and GAP proteins. Intriguingly, in the course of a genome-wide analysis of all RhoGEFs and RhoGAPs in Drosophila, one Rac-specific GAP has been identified that when overexpressed results in phenotypes reminiscent of robo loss of function (H. Hu et al., submitted, reported in Fan, 2003). There are a number of candidate GEFs that could explain how Rac activity is upregulated by Slit activation of Robo, most notably Sos, rtGEF (pix), and Trio. It will be interesting to determine which if any of these molecules could play such a role in Robo signaling (Fan, 2003).

Protein interactions: targeting Pak to the NMJ

Mutations in rho-type guanine exchange factor (rt/GEF), also called dpix, were recovered from a large-scale screen in Drosophila for genes that control synaptic structure. dPix/rtGEF is homologous to mammalian Pix. dPix plays a major role in regulating postsynaptic structure and protein localization at the Drosophila glutamatergic neuromuscular junction. dpix mutations lead to decreased synaptic levels of the PDZ protein Discs large, the cell adhesion molecule Fas II, and the glutamate receptor subunit GluRIIA, and to a complete reduction of the serine/threonine kinase Pak and the subsynaptic reticulum. The electrophysiology of these mutant synapses is nearly normal. Many, but not all, dpix defects are mediated through dPak, a member of the family of Cdc42/Rac1-activated kinases. Direct interaction of mammalian Pix with Pak has been detected. Thus, a Rho-type GEF (Pix) and Rho-type effector kinase (Pak) regulate postsynaptic structure (Parnas, 2001).

In mammals, the Pix family contains two members: alphaPix (Cool-2) and ßPix (Cool-1). Pix has an SH3 domain, a DBL-homology GEF domain, and a pleckstrin homology domain. The Cool (for cloned-out of library)/Pix (for PAK-interactive exchange factor) proteins directly bind to members of the PAK family of serine/threonine kinases and regulate their activity. In Drosophila, dPix is localized to the PSD: dpix mutations lead to the loss of synaptic Pak kinase. Paks are a family of Cdc42/Rac1-activated serine/threonine kinases important in regulating actin-containing structures. In the fly NMJ, Pak kinase is localized to the PSD. In mammals, Pak is recruited to focal complexes in a Cdc42-, Rac1-, and Pix- dependent manner (Parnas, 2001).

Since the dpix mutation eliminates Pak kinase from the synapse, and since Pak is a downstream target of the Rac/Cdc42 pathway, which includes Pix, it was reasoned that the elimination of Pak kinase from the synapse may be responsible for the inefficient clustering of Dlg. Thus, dpak mutants were stained with antibodies for Dlg and dPix. Several combinations of dpak alleles were used. dpak¹¹ is a protein null, and dpak⁶ has a stop codon at position 113; however, there is still some synaptic expression in this allele. dpak⁴ has a missense mutation in the Dock binding domain, and Pak kinase protein levels are normal. Finally, dpak⁷ has not been characterized molecularly, but behaves genetically as a null allele. In all allelic combinations of dpak mutants, dPix levels and localization are normal. In the allelic combination dpak¹¹/dpak⁴, Dlg levels at the synapse are also normal. However, in dpak¹¹/dpak⁶, Dlg levels are somewhat lower than wild-type (reduction of 57%), and Pak kinase levels are reduced by 66%. In dpak¹¹/dpak⁷, Dlg levels are reduced to the same extent as in dpix mutants (75%), and Pak kinase is absent. Fas II levels are also reduced to the same extent as in dpix mutants (a reduction of 19.3%). Levels of GluRIIA are also reduced, although less than levels in dpix mutants (reduction of 56%). These results are consistent with Pak kinase acting as a downstream effector of dPix. Nevertheless, there are differences between dpak and dpix mutants. In dpix mutants, the synapse looks abnormal and irregular; whereas in dpak mutants, even when Dlg levels are lowered, the synapse looks normal. Also, in dpak mutants, the muscles are thin and degenerated, and the muscle nuclei are mislocalized. In dpix larvae, the muscle activity appears weaker (as assessed by larval motility) than in wild-type larvae, but they are not as affected as in dpak mutants, and muscle nuclei are localized normally. It should be noted that the ultrastructure of dpix and dpak muscles are completely normal, and muscle differentiation per se does not seem to be affected. No structural correlate could be found that would explain the weaker muscles of dpix larvae (Parnas, 2001).

In Drosophila, the PSD-95 homolog Dlg has been shown to be directly responsible for the clustering of the Shaker potassium channel and to partially control the clustering of the cell adhesion molecule Fas II to the NMJ. Many, but not all, dpix defects are mediated through Pak kinase. Thus, the data suggest a pathway for synaptic clustering from dPix to Pak kinase to Dlg to Shaker and to Fas II (Parnas, 2001).

The dpix phenotype is consistent with at least two functions at the postsynaptic compartment: targeting and stabilization of postsynaptic components. In dpix mutants, Pak kinase is completely missing from the synapse. Since Pix is known to directly interact with Pak in mammals and target it to focal complexes, the data best fit with the model in which dPix targets Pak kinase to the synapse via a direct interaction. Furthermore, overexpressing either Pak kinase or a membrane-tethered gain-of-function form of Pak kinase does not result in any accumulation of Pak kinase at the synapse. Still, it is possible that Pak kinase is targeted to the synapse via a different mechanism and fails to stabilize in dpix mutants (Parnas, 2001).

In contrast to Pak kinase, Dlg and GluRIIA are not completely eliminated from the synapse in dpix mutants, but rather, their levels are reduced. In the case of Dlg, its localization pattern is also disrupted, indicating that dPix controls the postsynaptic targeting of Dlg at least to some extent, as well as its stabilization at the synapse. The localization pattern of GluRIIA (in subsynaptic domains opposite active zones) is intact. Thus, dPix is not necessary for the synaptic targeting of GluRIIA per se, but rather, it is important for maintenance of its levels and/or stabilization (Parnas, 2001).

Rac regulates axon growth through convergent and divergent signaling pathways, and can act through the actin-depolymerizing and actin-severing protein factor cofilin encoded by twinstar

Rho GTPases are essential regulators of cytoskeletal reorganization, but how they do so during neuronal morphogenesis in vivo is poorly understood. The actin-depolymerizing and actin-severing protein factor cofilin, encoded by twinstar, is essential for axon growth in Drosophila neurons. Cofilin function in axon growth is inhibited by LIM kinase and activated by Slingshot phosphatase. Dephosphorylating cofilin appears to be the major function of Slingshot in regulating axon growth in vivo. Genetic data provide evidence that Rho or Rac/Cdc42, via effector kinases Rho-associated kinase (Rok, also named Rho kinase or ROCK), or p21-activated kinase (Pak), respectively, activate LIM kinase to inhibit axon growth. Importantly, Rac also activates a Pak-independent pathway that promotes axon growth, and different RacGEFs regulate these distinct pathways. These genetic analyses reveal convergent and divergent pathways from Rho GTPases to the cytoskeleton during axon growth in vivo and suggest that different developmental outcomes could be achieved by biases in pathway selection (Ng, 2004).

Biochemical studies have shown that LIM kinase is activated by Pak, a downstream effector kinase for Rac and Cdc42. Whether Pak activation could affect the LIMK pathway in axon growth was tested. In similar genetic interaction experiments, it was found that introducing three independently generated Drosophila Pak mutant alleles suppresses the LIMK1 overexpression phenotype. In addition, Pak overexpression also results in axon growth and guidance defects similar to those seen with LIMK1 overexpression, consistent with the model in which Pak activates LIMK1, leading to cofilin hyperphosphorylation. As predicted from this model, coexpression of active versions of cofilin with Pak results in a partial suppression of the growth and guidance defects (Ng, 2004).

Next it was determined which Rho GTPase pathway (Cdc42, Rac1/Rac2, or Rho1) regulates Drosophila Pak. It was found that loss of one copy of Cdc42 results in a strong suppression of the Pak overexpression phenotypes. Similar reduction of Rac1 and Rac2 (Rac1^J10, Rac2^Δ) results in weaker suppression. In contrast, reducing Rho1 results in a slight enhancement of the Pak overexpression phenotype. This suggests that the Pak overexpression phenotype specifically reflects the positive signaling input from Cdc42 and Rac in vivo (Ng, 2004).

Next it was tested whether mutations in the Cdc42 or Rac genes (Rac1, Rac2, Mtl, and the different combinations) would modify the LIMK1 overexpression phenotype. Reducing the Cdc42 activity leads to the suppression of the LIMK1 phenotype, suggesting that Cdc42 acts through Pak to activate LIMK1. Loss of one copy of both Rac2 and Mtl does not alter the LIMK1 phenotype. However, the LIMK1 phenotype is very sensitive to the levels of endogenous Rac1, since introducing one copy of a strong hypomorphic allele of Rac1 (Rac1^J10) significantly suppresses the LIMK1 phenotype. This suppression is further enhanced when one copy of a null allele of Rac2 is introduced. These results suggest that Rac signaling (in particular, Rac1 and Rac2) also acts to activate LIMK1. This suggestion was also supported by the overexpression experiment. Overexpression of wild-type Rac1 (UAS-Rac1 WT) results in a weak LIMK1-like phenotype. However, when Rac1 and LIMK1 are coexpressed, axon growth defects were enhanced. Together with previous biochemical studies, these results suggest that Cdc42 and Rac act via Pak to activate LIMK1 (Ng, 2004).

To verify these genetic interactions with endogenous components, whether reducing either Cdc42 or Rac activity would modify the ssh^-/- phenotype was examined. As with Rho1, reducing Cdc42 or Rac activity also partially suppresses the ssh^-/- defects. Thus, these data indicate that Rho1, Cdc42, and Rac can all act through distinct downstream kinases to activate LIMK1 (Ng, 2004).

Surprisingly, reducing Rac GTPase activity can also result in enhancing the LIMK1 overexpression phenotype. For instance, the suppression effect of Rac1^J10 Rac2^Δ/+ was reverted by heterozygosity of Mtl (Rac1^J10 Rac2^Δ Mtl^Δ/+). This enhancement is more evident when one copy of the strongest allele of Rac1 was introduced into the LIMK1 overexpression background (Rac1^J11/+). This effect is not due to dominant effects of this Rac1 allele, since in the same genetic background Rac1^J11/+ animals did not display LIMK1-like growth defects without the LIMK1 expression transgene. Introduction of another hypomorphic allele of Rac1 (Rac1^J6) together with Rac2 (Rac1^J6 Rac2^Δ/+) also strongly enhances the LIMK1 overexpression phenotype (Ng, 2004).

One interpretation of these results is that, while Rac can activate LIMK1 (via Pak), there is an alternative pathway downstream of Rac, acting antagonistically to LIMK1, that promotes axon growth. This is consistent with the finding that loss of Rac GTPase activity leads to axon growth defects in MB neurons (Ng, 2002). This pathway is likely to be Pak independent, since the Rac axon growth-promoting activity does not require direct binding to Pak (Ng, 2002), and Pak activation leads to axon growth inhibition. To further verify the existence of a Pak-independent pathway in regulating axon growth, use was made of a well-described Rac1 effector domain mutant. In mammalian fibroblasts, Rac1 activates Pak1 and, through an independent downstream pathway, promotes lamellipodia formation. A Y40C mutation in the effector binding domain of Rac1 results in the loss of Pak1 activation, but the lamellipodia promoting activity is maintained. Transgenic overexpression of Rac1 Y40C alone at levels comparable to those of wild-type Rac did not result in gross axon phenotypes. However, in contrast to wild-type Rac1 that strongly enhances LIMK1, coexpression of Rac1 Y40C strongly suppresses the LIMK1 overexpression phenotypes. These results suggest that Rac1 activates a Pak-independent pathway to counteract the effects of LIMK1 activity on axon growth (Ng, 2004).

Whether upstream activators of Rac (RacGEFs), could regulate these distinct axon growth pathways was tested. Trio encodes a RacGEF essential for axon guidance in MB neurons. Introducing one mutant copy of trio significantly suppressed the LIMK1 overexpression phenotype, suggesting that Trio acts to activate LIMK1. This was further verified by the overexpression experiments in which overexpression of wild-type Trio (UAS-trio) alone results in a mild LIMK1-like phenotype, while coexpression with LIMK1 results in a strong enhancement of the axon growth defects. Trio has two GEF domains—GEF1 is specific for Rac1, Rac2, and Mtl in vitro and in vivo, and GEF2 can activate Rho1/RhoA in vitro. Overexpression of the isolated Trio GEF1 domain (UAS-trio GEF1) in MB neurons results in severe axon growth defects, whereas overexpression of the isolated Trio GEF2 domain (UAS-trio GEF2) does not result in any gross defects. These results support the model that, in MB neurons, Trio, via its GEF1 domain, acts through Rac and Pak to activate LIMK1. These findings are also consistent with those of previous studies in which Trio acted via Rac/Pak to regulate Drosophila photoreceptor axon guidance (Ng, 2004).

In contrast to Trio, loss of one copy of still life (sif), encoding a different RacGEF, markedly enhances the LIMK1 phenotype. In overexpression experiments, UAS-sif alone does not result in gross axon defects. However, coexpression of Sif resulted in a strong suppression of the LIMK1 phenotype. These experiments suggest that Sif activates the pathway that acts antagonistically to LIMK1 (Ng, 2004).

The data suggest that Rac promotes axon growth via a pathway antagonistic to Pak and LIMK1. How does this pathway act to promote axon growth? One strong possibility is that Rac stimulates actin polymerization to promote axon growth. Therefore a number of candidate genes known to promote actin polymerization were tested. The following genetic criteria were establised by which these candidate pathways should work: (1) like Rac, loss of the candidate gene should result in axon growth defects; (2) genetic interactions with LIMK1, either through loss- or gain-of-function analyses, should show that they act antagonistically to LIMK1 (Ng, 2004).

The role of the actin nucleation factor SCAR-Arp2/3 complex was tested. Rac has been shown to promote de novo actin polymerization through interactions via SCAR and the Arp2/3 complex. Activation of the SCAR-Arp2/3 complex is required to establish cell protrusions during lamellipodia and filopodia formation, making it a good candidate pathway for promoting axon growth. This hypothesis was tested by making MARCM clones in MB neurons using null alleles of SCAR, WASp (a protein related to SCAR, also called Wsp), double mutants for SCAR and WASp, or Arpc1 (Flybase-Sop2), an essential component of the Arp2/3 complex. No axon growth defects were detected in single-cell or neuroblast clones. In addition, reduction of SCAR or Sop2 levels did not modify the LIMK1 overexpression phenotype. Furthermore, overexpression of SCAR did not suppress, but mildly enhanced, the LIMK1 overexpression phenotype. Taken together, these results suggest that the SCAR/WASp-Arp2/3 pathway does not play an essential role in axon growth of MB neurons, and it is unlikely that this pathway contributes to the Rac pathway that promotes axon growth (Ng, 2004).

Several other known regulators of actin polymerization were tested for their contribution to MB axon growth. The actin polymerization stimulators profilin (encoded by chickadee or chic) and Enabled (Ena) have been shown to be essential for axon growth and guidance in Drosophila. In addition, genetic interaction studies suggest that these proteins may be involved in Rac GTPase signaling. When assayed for MB axon growth, both ena^-/- and chic^-/- single-cell and neuroblast clones exhibited drastic axon growth defects. Interestingly, when examined at higher resolutions, neither ena nor chic axons displayed filopodia-like and lamellipodia-like protrusions at the axon termini characteristic of tsr^-/- neurons. These genetic interaction experiments found no evidence that ena or chic act antagonistically to LIMK1. In fact, Ena overexpression strongly enhances the LIMK1 overexpression phenotype. Thus, although both Ena and profilin are essential for MB axon growth, they do not appear to constitute the Rac-mediated axon growth-promoting pathway (Ng, 2004).

Finally, the Formin-class protein Diaphanous (Dia) was tested, since it has been implicated in regulating actin polymerization downstream of Rho GTPases. dia^-/- MB neurons do not exhibit axon growth defects in single-cell clones or in neuroblast clones, which exhibit strong cell proliferation defects. Interestingly, reducing Dia activity appears to suppress the LIMK1 overexpression phenotype, suggesting that Dia can act in a pathway that enhances, but does not antagonize, LIMK1 (Ng, 2004).

The leading edge during dorsal closure as a model for epithelial plasticity: Pak is required for recruitment of the Scribble complex and septate junction formation

Dorsal closure (DC) of the Drosophila embryo is a model for the study of wound healing and developmental epithelial fusions, and involves the sealing of a hole in the epidermis through the migration of the epidermal flanks over the tissue occupying the hole, the amnioserosa. During DC, the cells at the edge of the migrating epidermis extend Rac- and Cdc42-dependent actin-based lamellipodia and filopodia from their leading edge (LE), which exhibits a breakdown in apicobasal polarity as adhesions are severed with the neighbouring amnioserosa cells. Studies using mammalian cells have demonstrated that Scribble (Scrib), an important determinant of apicobasal polarity that functions in a protein complex, controls polarized cell migration through recruitment of Rac, Cdc42 and the serine/threonine kinase Pak, an effector for Rac and Cdc42, to the LE. DC and the follicular epithelium were used to study the relationship between Pak and the Scrib complex at epithelial membranes undergoing changes in apicobasal polarity and adhesion during development. It is proposed that, during DC, the LE membrane undergoes an epithelial-to-mesenchymal-like transition to initiate epithelial sheet migration, followed by a mesenchymal-to-epithelial-like transition as the epithelial sheets meet up and restore cell-cell adhesion. This latter event requires integrin-localized Pak, which recruits the Scrib complex in septate junction formation. It is concluded that there are bidirectional interactions between Pak and the Scrib complex modulating epithelial plasticity. Scrib can recruit Pak to the LE for polarized cell migration but, as migratory cells meet up, Pak can recruit the Scrib complex to restore apicobasal polarity and cell-cell adhesion (Bahri, 2010).

Some embryos lacking zygotic Pak function successfully bring the epidermal flanks together at the dorsal midline but fail to restore septate junctions and adherens junctions at the LE in the DME cells. Thus, Pak at the LE membrane of the dorsal-most epithelial cells (DME) is regulating establishment of apicobasal polarity during a mesenchymal-epithelial transition. It is suspected that Pak is acting through different routes in its regulation of adherens junction formation versus septate junction formation. This study has focused on Pak regulation of the Scrib complex in septate junction formation at the LE. The data indicate that Pak is a component of the Scrib complex at the lateral membrane. Although Pak might be associating with the Scrib complex throughout epithelia, it might only be required for recruitment of the Scrib complex in epithelia derived from a mesenchymal-like intermediate such as the follicular epithelium and the LE. With the exception of the LE, apicobasal polarity in the epidermis is determined much earlier in development with formation of the blastoderm by cellularization. The epidermis is therefore a primary epithelium that does not arise from a mesenchymal intermediate, and Pak function does not appear to be required for apicobasal polarity in primary epithelia (Bahri, 2010).

Localization of Pak at the lateral membrane in both the follicular epithelium and in the epidermis is integrin-dependent. Studies using organ culture of embryonic kidney mesenchyme and MDCK cells demonstrate a requirement for integrins in apicobasal polarity of epithelia derived from MET, and this study has shown that βPS-integrin is required for Scrib complex and septate junction protein recruitment at the LE and in the follicular epithelium. Furthermore, previous studies in the follicular epithelium and another Drosophila epithelium derived from MET, the midgut, have demonstrated a requirement for integrins in the maintenance of apicobasal polarity. It is proposed that, at the LE, the absence of the septate junction diffusion barrier allows the accumulation of integrin complexes along the lateral membrane. These lateral integrin complexes recruit Pak, around which the Scrib complex is assembled. Thus, the absence of septate junctions allows the recruitment of proteins needed for the assembly of septate junctions. The model suggests that there might be transient Pak-mediated links between integrin and the Scrib complex. Interestingly, Dlg and βPS-integrin have been shown to co-immunoprecipitate from fly head extracts, consistent with these proteins existing in a complex in the nervous system and/or in epithelia (Bahri, 2010).

The data and recent studies on the amnioserosa support the idea that septate junctions restrict the accumulation of lateral integrins. The amnioserosa is devoid of septate junction proteins such as FasIII, and this might be owing to absence in this tissue of the transcription factor Grainy head, which promotes expression of septate junction proteins. The wild-type amnioserosa has high levels of lateral βPS-integrin, but ectopic expression of Grainy head in the amnioserosa leads to an accumulation of septate junction proteins and an accompanying disruption of βPS-integrin localization. Similarly, at the completion of DC, septate junctions appear at the LE and this is accompanied by downregulation of LE lateral integrins. In pak¹⁴pak3^76A and cora¹⁴ embryos where LE septate junctions are deficient, lateral LE βPS-integrin persists (Bahri, 2010).

A recent study in mammalian cell culture indicates that Scrib recruits Pak to the LE (Nola, 2008), and this study has shown that Pak localization in the follicular epithelium is Scrib-dependent. This study of the LE at the end of DC demonstrates that the relationship between Cdc42/Rac signaling complexes and Scrib can act in the opposite direction: membrane-localized Pak recruits the Scrib complex. A bidirectional interaction between the Scrib complex and Cdc42/Rac signaling complexes, including Pak, might be a crucial regulator of events at the LE of closing epithelia during both wound healing and development in diverse systems. Scrib at the newly formed LE can lead to recruitment of the Cdc42/Rac signaling complex, allowing acquisition of mesenchymal characteristics and polarized cell migration. When the opposing epithelial flanks meet up, events can be reversed with Pak recruiting the Scrib complex to the lateral membrane, contributing to restoration of apicobasal polarity and cell adhesion at the LE during MET. The Scrib/Pak complex is believed to be a 'toggle switch', enabling the epithelial membrane to shift back and forth between a migratory state characterized by actin-based extensions and an apicobasal polarized state characterized by cell-cell adhesion (Bahri, 2010).

Neto-mediated intracellular interactions shape postsynaptic composition at the Drosophila neuromuscular junction

The molecular mechanisms controlling the subunit composition of glutamate receptors are crucial for the formation of neural circuits and for the long-term plasticity underlying learning and memory. This study use the Drosophila neuromuscular junction (NMJ) to examine how specific receptor subtypes are recruited and stabilized at synaptic locations. In flies, clustering of ionotropic glutamate receptors (iGluRs) requires Neto (Neuropillin and Tolloid-like), a highly conserved auxiliary subunit that is essential for NMJ assembly and development. Drosophila neto encodes two isoforms, Neto-α and Neto-β, with common extracellular parts and distinct cytoplasmic domains. Mutations that specifically eliminate Netoβ or its intracellular domain were generated. When Neto-β is missing or is truncated, the larval NMJs show profound changes in the subtype composition of iGluRs due to reduced synaptic accumulation of the GluRIIA subunit. Furthermore, neto-β mutant NMJs fail to accumulate p21-activated kinase (PAK), a critical postsynaptic component implicated in the synaptic stabilization of GluRIIA. Muscle expression of either Neto-α or Neto-β rescued the synaptic transmission at neto null NMJs, indicating that Neto conserved domains mediate iGluRs clustering. However, only Neto-β restored PAK synaptic accumulation at neto null NMJs. Thus, Neto engages in intracellular interactions that regulate the iGluR subtype composition by preferentially recruiting and/or stabilizing selective receptor subtypes (Ramos, 2015).

At the Drosophila NMJ, Neto enables iGluRs clustering at synaptic sites and promotes postsynaptic differentiation. This study shows that Neto-β, the major Neto isoform at the fly NMJ, plays a crucial role in controlling the distribution of specific iGluR subtypes at individual synapses. Similar to other glutamatergic synapses, the subunit composition determines the activity and plasticity of the fly NMJ. The data are consistent with a model whereby Neto-β, via its conserved domains, fulfills a significant part of Neto-dependent iGluRs clustering activities during synapse assembly. At the same time, Neto-β engages in intracellular interactions that regulate iGluR subtypes distribution by preferentially recruiting and/or stabilizing type-A receptors. In this model, Neto-β could directly associate with the GluRIIA-containing complexes and/or regulate the synaptic abundance of type-A receptors indirectly, by recruiting PSD components such as PAK. Thus, Neto-β employs multiple strategies to control which flavor of iGluR will be at the synapses and to modulate PSD composition and postsynaptic organization (Ramos, 2015).

Neto proteins have been initially characterized as auxiliary subunits that modulate the function of kainate (KA) and NMDA receptors. In vertebrates, Neto1 and Neto2 directly interact with KAR subunits and increase channel function by modulating gating properties. Since loss of KAR currents in mice lacking Neto1 and/or Neto2 exceed a reduction that could be attributed to alterations of channel gating, an additional role for Neto proteins in synaptic targeting of receptors has been proposed. The role for vertebrate Neto proteins in KAR membrane and/or synaptic targeting remains controversial and appears to be cell type-, receptor subunit-, and Neto isoform-dependent. Furthermore, the C. elegans Neto has a very small intracellular domain (24 amino acids beyond the conserved domains). This implies that 1) Neto without an intracellular domain constitutes the minimal conserved functional moiety, and 2) the divergent intracellular domains of Neto proteins may fulfill tissue and/or synapse specific modulatory functions. Indeed, Neto2 bears a class II PDZ binding motif that binds to the scaffold protein GRIP and appears to mediate KARs stabilization at selective synapses (Ramos, 2015).

In flies, Neto is an essential protein that plays active roles in synapse assembly and in the formation and maintenance of postsynaptic structures at the NMJ. The Drosophila Neto isoforms do not have PDZ binding motifs, but they use at least two different mechanisms to regulate the synaptic accumulation and subunit composition of iGluRs. First, Neto participates in extracellular interactions that enable formation of iGluR/Neto synaptic complexes; formation of stable aggregates is presumably prevented by the inhibitory prodomain of Neto. Second, the two Neto isoforms appear to facilitate the selective recruitment and/or stabilization of specific iGluR subtypes. It is speculated that Neto-β may selectively associate with and recruit type-A receptors, perhaps by engaging the C-terminal domain of GluRIIA, which is critical for the synaptic stabilization of these receptors. Aside from a possible role in the selective recruitment of iGluR subtypes, Neto-β participates in intracellular interactions that facilitate the recruitment of PAK at PSDs; in turn, PAK signals through two independent, genetically separable pathways (a) to modulate the GluRIIA synaptic abundance and (b) to facilitate formation of SSR (Ramos, 2015).

Whether Neto-β recruits PAK directly or via a larger protein complex remains to be determined. Neto-β contains an SH3 domain that may bind to the proline-rich SH3 binding domain of PAK. However, in tissue culture experiments, attempts to detect a direct interaction between PAK and Neto-β (full-length or intracellular domain) failed. PAK synaptic accumulation is completely abolished at NMJ with mutations in dPix, although not all dpix defects are mediated through PAK. Conversely, PAK together with Dreadlocks (Dock) controls the synaptic abundance of GluRIIA, while PAK and dPix regulate the Dlg distribution. The reduction of GluRIIA and Dlg synaptic abundance observed at neto-β mutant NMJs suggests that Neto-β may interact with both dPix and Dock and enable both PAK activities. In addition, Neto-β may stabilize postsynaptic type-A receptors by enhancing their binding to Coracle, which anchors GluRIIA to the postsynaptic actin cytoskeleton (Ramos, 2015).

Importantly, this study connects the complex regulatory networks that modulate the PSD composition to the Neto/iGluR clusters themselves. The Neto-β cytoplasmic domain is rich in putative protein interaction motifs, and may function as a scaffold platform to mediate multiple protein interactions that act synergistically during synapse development and homeostasis. Loss of Neto-β-mediated intracellular interactions at netoβshort NMJs reduced the GluRIIA synaptic abundance, but did not affect the GluRIIB synaptic signals. It is unlikely that the remaining cytoplasmic part of Neto-β facilitates the GluRIIB synaptic accumulation at these NMJs at the expense of GluRIIA and PAK. Instead, a model is favored whereby the synaptic stabilization of GluRIIA requires a Neto-β-dependent intracellular network. Disruption of this network diminishes GluRIIA and increases GluRIIB synaptic abundance, pending the availability of limiting subunits, GluRIIC-E and Neto. Conversely, the presence of this network ensures that at least some type-A receptors are stabilized at synaptic sites, even at Neto-deprived synapses, such as in netohypo larvae [12]. Assembly of this network does not require GluRIIA since both Neto-β and PAK accumulated normally at GluRIIA mutant NMJs. Furthermore, in the absence of Neto-β the synaptic abundance of GluRIIA can be partly restored by excess Neto-α or a δ-intracellular Neto variant, suggesting that excess iGluRs 'clustering capacity' overrides the cellular signals that shape PSD composition. What intracellular domain(s) of Neto bind to and how they are modified by post-translational modifications will be critical questions to understand how postsynaptic composition is regulated during development and homeostasis (Ramos, 2015).

The discovery of Drosophila Neto isoforms with alternative cytoplasmic domains and isoform specific activities expands the repertoire of Neto-mediated functions at glutamatergic synapses. All Neto proteins share the highly conserved CUB1, -2, LDLa and transmembrane domains that have been implicated in engaging and modulating the receptors, the central function of Neto proteins. In flies this conserved part is both required and sufficient for iGluRs clustering and NMJ development. In C. elegans the entire Neto appears to be reduced to this minimal functional unit. The only exception is a retina-specific CUB1-only Neto1 isoform with unknown function. In contrast to shared domains, the cytoplasmic domains are highly divergent among Neto proteins. This diversity might have evolved to facilitate intracellular, context specific function for Neto proteins, such as the need to couple the iGluR complexes to neuron or muscle specific scaffolds in various phyla. By engaging in different intracellular interactions, via distinct cytoplasmic domains, different Neto isoforms may undergo differential targeting and/or retention at the synapses and thus acquire isoform-specific distributions and functions within the same cell (Ramos, 2015).

Phylogenetic analyses indicate that the intracellular domains of Neto are rapidly evolving in insects. Blocks of high conservations could be clearly found in the genome of short band insect Tribolium castaneum (Coleoptera) or in Apis mellifera (Hymenoptera). However, most insects outside Diptera appear to have only one Neto isoform, more related to Neto-β. In fact, the only Neto-α isoform outside Drosophila was found in Musca domestica (unplaced genomic scaffold NCBI Reference Sequence: XM_005187241.1). Other neto loci, from Hydra to vertebrates, appear to encode Neto proteins with unique and highly divergent intracellular domains. An extreme example is the C. elegans Neto/Sol-2, with a very short cytoplasmic tail, which requires additional auxiliary subunits, Sol-1 and Stargazin, to control the function of glutamate receptors. Neto proteins appear to utilize their intracellular domains to connect to the signaling networks that regulate the distribution and subunit composition for iGluRs. Such cellular signals converge onto and are integrated by the intracellular domains of the receptors and/or by various auxiliary subunits associated with the iGluR complexes (Ramos, 2015).

Neto proteins modulate the gating behavior of KAR but also play crucial roles in the synaptic recruitment of glutamate receptors in vivo. At the fly NMJ, Neto enables iGluRs synaptic clustering and initiates synapse assembly. In addition, the intracellular domain of Neto-β recruits PSD components and triggers a cascade of events that organize postsynaptic structures and shape the composition of postsynaptic fields. The cytoplasmic domains of Neto proteins emerge as versatile signaling hubs and organizing platforms that directly control the iGluRs subunit composition and augment the previously known Neto functions in modulation of glutamatergic synapses (Ramos, 2015).

Filamin, a synaptic organizer in Drosophila, determines glutamate receptor composition and membrane growth

Filamin is a scaffolding protein that functions in many cells as an actin-crosslinker. FLN90, an isoform of the Drosophila ortholog Filamin/cheerio that lacks the actin-binding domain, is shown in this study to govern the growth of postsynaptic membrane folds and the composition of glutamate receptor clusters at the larval neuromuscular junction. Genetic and biochemical analyses reveal that FLN90 is present surrounding synaptic boutons. FLN90 is required in the muscle for localization of the kinase dPak and, downstream of dPak, for localization of the GTPase Ral and the exocyst complex to this region. Consequently, Filamin is needed for growth of the subsynaptic reticulum. In addition, in the absence of filamin, type-A glutamate receptor subunits are lacking at the postsynapse, while type-B subunits cluster correctly. Receptor composition is dependent on dPak, but independent of the Ral pathway. Thus two major aspects of synapse formation, morphological plasticity and subtype-specific receptor clustering, require postsynaptic Filamin (Lee, 2016).

Proper postsynaptic function depends on appropriate localization of receptors and signaling molecules. Scaffolds such as PSD-95/SAP90 and members of the Shank family are critical to achieving this organization. While usually without intrinsic enzymatic activity, scaffolds recruit, assemble, and stabilize receptors and protein networks through multiple protein-protein interactions: they can bind to receptors, postsynaptic signaling complexes, and the cytoskeleton at the postsynaptic density. Mutations in these proteins are associated with neuropsychiatric disorders. While understanding synapse assembly has begun, much remains to be investigated (Lee, 2016).

The Drosophila larval neuromuscular junction (NMJ) is a well-studied and genetically accessible glutamatergic synapse. Transmission is mediated by AMPA-type receptors, and several postsynaptic proteins important for its development and function have related proteins at mammalian synapses, including the PDZ-containing protein Discs-Large (DLG) and the kinase Pak. In addition, the postsynaptic membrane forms deep invaginations and folds called the subsynaptic reticulum (SSR), which are hypothesized to create subsynaptic compartments comparable to dendritic spines. Recently, we found that the SSR is a plastic structure whose growth is regulated by synaptic activity. This phenomenon may be akin to the use-dependent morphological changes, such as growth of dendritic spines, that occur postsynaptically in mammalian brain. The addition of membrane and growth of the SSR requires the exocyst complex to be recruited to the synapse by the small GTPase Ral; the SSR fails to form in ral mutant larvae. Moreover, the localization of Ral to a region surrounding synaptic boutons is likely to direct the selective addition of membrane to this domain. Ral thus provided a tractable entry point for better understanding postsynaptic assembly. The mechanism for localizing the Ral pathway, however, was unknown. The present study determined that Ral localization is dependent on cheerio, a gene encoding filamin, which is critical for proper development of the postsynapse (Lee, 2016).

Filamin is a family of highly conserved protein scaffolds with a long rod-like structure of Ig-like repeats. With over 90 identified binding partners, some of which are present also at the synapse, mammalian filamin A (FLNA) is the most abundant and commonly studied filamin. Filamin can bind actin and has received the most attention in the context of actin cytoskeletal organization. Drosophila filamin, encoded by the gene cheerio (cher), shares its domain organization and 46% identity in amino acid sequence with human FLNA. Drosophila filamin has a well-studied role in ring canal formation during oocyte development, where it recruits and organizes actin filaments. This study now shows that filamin has an essential postsynaptic role at the fly NMJ. An isoform of this scaffold protein that lacks the actin-binding domain acts via dPak to localize GluRIIA receptors and Ral; filamin thereby orchestrates both receptor composition and membrane growth at the synapse (Lee, 2016).

Filamin is a highly conserved protein whose loss of function is associated with neurodevelopmental disorders. In humans, mutations in the X-linked FLNA cause periventricular heterotopia, a disorder of cortical malformation with a wide range of clinical manifestations such as epilepsy and neuropsychiatric disturbances. Studies in rodent models have shown that abnormal filamin expression causes dendritic arborization defects in a TSC mouse and that filamin influences neuronal proliferation. Filamin is present in acetylcholine receptor clusters at the mammalian NMJ, but its function there is unknown. In lysates of the mammalian brain, filamin associates with known synaptic proteins such as Shank3, Neuroligin 3, and Kv4.2. A recent report indicated that filamin degradation promotes a transition from immature filopodia to mature dendritic spines, a phenomenon that is likely to be related to the actin-bundling properties of the long isoform of filamin. Data in the present study have uncovered a novel pathway that does not require the actin-binding domain of filamin. In this pathway, postsynaptically localized filamin, via Pak, directs two distinct effector modules governing synapse development and plasticity: (1) the Ral-exocyst pathway for activity-dependent membrane addition and (2) the composition of glutamate receptor clusters. These pathways determine key structural and physiological properties of the postsynapse (Lee, 2016).

Although loss of filamin had diverse effects on synapse assembly, they were selective. Muscle-specific knockdown or the cher^Q1415sd allele disrupted type-A but not type-B GluR localization at the postsynaptic density. Likewise, the phenotypes for muscle filamin were confined to the postsynaptic side: the presynaptic active zone protein Brp and overall architecture of the nerve endings were not altered by muscle-specific knockdown. The specificity of its effect on particular synaptic proteins, and the absence of the actin-binding region in FLN90, suggests that filamin's major mode of action here is not overall cytoskeletal organization, but rather to serve as a scaffold for particular protein-protein interactions (Lee, 2016).

Analysis of the distribution of the SSR marker Syndapin and direct examination of the subsynaptic membrane by electron microscopy revealed that formation of the SSR required filamin. Genetic analysis uncovered a sequential pathway for SSR formation from filamin to the Pak/Pix/Rac signaling complex, to Ral, to the exocyst complex and consequent membrane addition. The SSR is formed during the second half of larval life and may be an adaptation for the low input resistance of third-instar muscles. Like dendritic spines, the infoldings of the SSR create biochemically isolated compartments in the vicinity of postsynaptic receptors and may shape physiological responses, although first-order properties of the synapse, such as mini- or EPSP amplitude are little altered in mutants that lack an SSR. The formation of the SSR requires transcriptional changes driven by Wnt signaling and nuclear import, proteins that induce membrane curvature (such as Syndapin, Amphiphysin, and Past1), and Ral-driven, exocyst-dependent membrane addition. The activation of Ral by Ca2+ influx during synaptic transmission allows the SSR to grow in an activity-dependent fashion. The localization of Ral to the region surrounding the bouton appears crucial to determining the site of membrane addition because Ral localization precedes SSR formation and exocyst recruitment and because exocyst recruitment occurred selectively surrounding boutons even when Ca2+-influx occurred globally in response to calcimycin. This study has now shown that Ral localization, and consequently exocyst recruitment, membrane growth, and the presence of the SSR marker Syndapin, are all dependent on a local action of filamin at the synapse. FLN90, the filamin short isoform, localized to sites of synaptic contact and indeed surrounded the boutons just as does Ral. When this postsynaptic filamin was removed by muscle-specific filamin knockdown or the cher^Q1415sd allele, the downstream elements of the pathway, Pak, Rac, Ral, the exocyst, and Syndapin, were no longer synaptically targeted. The mislocalization is not a secondary effect of loss of the SSR but likely a direct consequence of filamin loss, as Pak and Ral can localize subsynaptically even in the absence of the SSR. Unlike the likely mode of action of nuclear signaling by Wnt, the delocalization of Ral was not a consequence of altered protein production; its expression levels did not change. Thus filamin may be viewed as orchestrating the formation of the SSR and directing it to the region surrounding synaptic boutons (Lee, 2016).

The second major feature of the filamin phenotype was the large reduction in the levels of the GluRIIA receptor subunit from the postsynaptic membranes. GluRIIA and GluRIIB differ in their electrophysiological properties and subsynaptic distribution. Because type B GluRs, which contain the IIB subunit, desensitize more rapidly than type A, the relative abundance of type A and type B GluRs is a key determinant of postsynaptic responses and changes with synapse maturation. The selective decrease in GluRIIA at filamin-null NMJs is likely a consequence of dPak mislocalization: filamin-null NMJs lack synaptic dPak, and dPak null NMJs lack synaptic GluRIIA. Moreover, the first-order electrophysiological properties at NMJs lacking filamin resembled those reported at NMJs missing dPak. In the current study, though, only the change in mEPSP frequency was statistically significant. At filamin-null NMJs, the decrease in GluRIIA is accompanied by an increase in GluRIIB, suggestive of a partial compensation that could account for the relatively normal synaptic transmission. Because the IIA and IIB subunits differ in desensitization kinetics and regulation by second messengers, functional consequences of filamin loss may become more apparent with more extensive physiological characterizations at longer time scales (Lee, 2016).

While both SSR growth and receptor composition required the kinase activity of dPak, receptor composition was independent of Ral and thus represents a distinct branch of the pathway downstream of dPak. As with Ral, the loss of GluRIIA from the synapse was due to delocalization and not a change in expression of the protein, consistent with unaltered GluRIIA transcripts in dPak null animals. Thus filamin, via dPak, alters proteins with functional significance for the synapse as well as its structural maturation (Lee, 2016).

Mammalian filamin, via its many Ig-like repeats, has known scaffold functions in submembrane cellular compartments and filamin is therefore likely also to serve as a scaffold at the fly NMJ. These epistasis data indicate that filamin recruits Ral through recruitment of a signaling complex already known to function at the fly NMJ: dPak and its partners dPix and Rac. Mammalian filamin is reported to directly interact with Ral during filopodia formation, however the details of their interaction at the fly NMJ are less clear. Because Ral localization requires filamin to recruit dPak and dPix and specifically requires the kinase activity of dPak, it is possible that either Ral or filamin need to be phosphorylated by dPak to bind one another. Mammalian filamin interacts with some components of the Pak signaling complex and is a substrate of Pak. This study has now shown that Drosophila filamin and PAK interact when coexpressed in HEK cells, and thus a direct scaffolding role for FLN90 in the recruitment of Pak and the organization of the postsynapse is likely (Lee, 2016).

The overlapping but different distributions of filamin and its downstream targets indicate that its scaffolding functions must undergo regulation by additional factors. The proteins discussed in this study take on either of two patterns at the synapse. Some, like Ral, Syndapin, and filamin itself, are what can be termed subsynaptic and, like the SSR, envelope the entire synaptic bouton. Others, like dPak and its partners and the GluRIIA proteins, are concentrated in much smaller regions, immediately opposite presynaptic active zones, where the postsynaptic density (PSD) is located. It is hypothesized that filamin interacts with additional proteins, including potentially transsynaptic adhesion proteins, that localize filamin to the subsynaptic region and also govern to which of the downstream effectors it will bind. Indeed, it appears paradoxical that dPak, though predominantly at the postsynaptic density is nonetheless required for Ral localization throughout the subsynaptic region. If dPak is needed to phosphorylate either filamin or Ral to permit Ral localization, the phosphorylations outside the PSD may be due to low levels of the dPak complex in that region; synaptic dPak was previously shown to be a relatively mobile component of the PSD (Lee, 2016).

Filamin was the first nonmuscle actin-crosslinking protein to be discovered. With an actin-binding domain at the N terminus, the long isoform of filamin and its capacity to integrate cellular signals with cytoskeletal dynamics have subsequently been the focus of the majority of the filamin literature. At the NMJ, however, this was not the case. Several lines of evidence indicated that the short FLN90 isoform of filamin, which lacks the actin-binding domain, plays an essential role in postsynaptic assembly. First, the short FLN90 isoform was the predominant and perhaps the only isoform of filamin found expressed in the muscles. Second, both endogenous and overexpressed FLN90 localized subsynaptically. Third, loss of the short isoform disrupted localization of postsynaptic components while lack of just the long isoform had little or no effect. Lastly, exogenous expression of just the short isoform in filamin null background sufficiently rescued the defect in SSR growth. The modest postsynaptic phenotypes of the cher1allele, which predominantly disrupts the long isoform, may be due to small effects of the allele on expression of the short isoform or may be an indirect consequence of the presence of the long isoform in the nerve terminals (Lee, 2016).

The existence of the short isoform has been reported in both flies and mammals and may be produced either by transcriptional regulation or calpain-mediated cleavage. The short isoform can be a transcriptional co-activator, but its functional significance and mechanisms of action have been largely elusive. The short isoform has little or no affinity for actin, but most of the known sites for other protein-protein interactions are shared by both isoforms. Thus the structure of FLN90, with nine predicted Ig repeats and likely protein-protein interactions, is consistent with a scaffolding function to localize key synaptic molecules independent of interactions with the actin cytoskeleton (Lee, 2016).

This study has introduced filamin as a major contributor to synapse development and organization. The severity of the phenotypes indicates filamin has a crucial role that is not redundant with other scaffolding proteins. The effects of filamin encompass several much-studied aspects of the Drosophila NMJ: the clustering and subunit subtype of glutamate receptors and the plastic assembly of specialized postsynaptic membrane structures. The pathways that govern these two phenomena diverge downstream of Pak kinase activity and are dependent on filamin for the proper localization of key signaling modules in the pathways. By likely acting as a scaffold protein, the short isoform of filamin may function as a link between cell surface proteins, as yet unidentified, and postsynaptic proteins with essential localizations to and functions at the synapse. Because many of the components of these pathways at the fly NMJ are also present at mammalian synapses and can interact with mammalian filamin, a parallel set of functions in CNS dendrites merits investigation (Lee, 2016).

Home page: The Interactive Fly © 1997 Thomas B. Brody, Ph.D.

The Interactive Fly resides on the
Society for Developmental Biology's Web server.