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. Here we 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 Neto-β 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.


Introduction
Ionotropic glutamate receptors (iGluRs) play major roles in excitatory transmission in the vertebrate brain and at the insect neuromuscular junction (NMJ). The synapse properties are primarily shaped by the subunit composition of the receptors, which could be further modified by RNA editing and alternative splicing [1]. Changes in the subunit composition of postsynaptic iGluRs, in particular the AMPA-type receptors, have a tremendous impact on the development and plasticity of glutamatergic synapses [2,3]. Mechanisms controlling the recruitment of selective receptor subtypes exist and they integrate signals from multiple signaling pathways and regulate synaptic trafficking via posttranslational modifications within the iGluRs intracellular domains [4,5]. In addition, several auxiliary subunits, which primarily modulate the channel properties, have been implicated in the subcellular distribution of receptors [6]. Whether and how the auxiliary subunits modulate the subunit composition of iGluRs remains unclear.
The Drosophila NMJ is a glutamatergic synapse similar to mammalian central synapses. In flies, the iGluRs are heterotetrameric complexes composed of three shared subunits, GluRIIC, GluRIID and GluRIIE, and either GluRIIA (type-A receptors) or GluRIIB (type-B) [7][8][9][10][11]. The function of the fly NMJ also requires Neto (Neuropillin and Tolloid-like), an obligatory auxiliary subunit of the iGluR complexes [12]. In the absence of Neto, or any of the shared iGluR subunits (or GluRIIA and GluRIIB together) the receptors fail to cluster at synaptic locations and the animals die as paralyzed embryos unable to develop into larval stages [12,13]. Genetic manipulation of Neto and iGluR levels indicated that Neto and the shared iGluR subunits are limiting for the synaptic localization of receptors and that GluRIIA and GluRIIB compete with each other for synaptic localization [9,14].
The type-A and type-B receptors differ in their single-channel properties, synaptic currents, regulation by second messenger, and sub-synaptic distribution [13]. The type-B channel desensitizes ten times faster than the type-A [8]. Also, the postsynaptic response to the fusion of single synaptic vesicles (the quantal size) is much reduced when only the type-B receptors are present; in fact, the dose of synaptic GluRIIA versus GluRIIB is a key determinant of quantal size. The interplay between the type-A and type-B synaptic receptors modulates synapse strength and plasticity.
The synaptic accumulation of type-A and type-B receptors at the fly NMJ is differentially regulated. Two hybrid and genetics screens identified Coracle, the Drosophila homologue of the mammalian cytoskeletal protein 4.1, as a direct binding partner for the GluRIIA cytoplasmic domain [15]. Disruptions of the actin cytoskeleton or mutations in coracle caused selective reduction in GluRIIA synaptic accumulation, suggesting that Coracle anchors the GluRIIA to the actin cytoskeleton. Mutations in the p21-activated kinase (PAK) also disrupt the localization and function of the GluRIIA. PAK co-localizes with the iGluR complexes at postsynaptic densities (PSDs) and, in conjunction with the guanine nucleotide exchange factor Pix and the adaptor protein Dreadlocks, promotes the synaptic accumulation of GluRIIA [16,17]. The synaptic accumulation of GluRIIB appears reduced in mutants of the PDZ (PSD-95/Dlg/Zona occludens-1) domain-containing scaffolding protein Discs large (Dlg) [18]. However, there is no evidence for a direct interaction between GluRIIB and Dlg, and except for GluRIIC [9], the iGluR subunits of the Drosophila NMJ lack a PDZ binding domain, suggesting that Dlg influences the synaptic abundance of GluRIIB indirectly. Also, Dlg localizes perisynaptically, at subsynaptic reticulum (SSR) [19]. In contrast, mammalian PSD-95 is a major component of the PSD, and it directly binds and modulates the synaptic targeting of iGluRs [20] and some of their interacting partners, such as the 4.1N protein [21]. The PSD-95-mediated interactions can be further regulated by post-translational modifications such as palmitoylation and phosphorylation [5].
Here we show that Neto is key in regulating the subtype composition of iGluRs at the Drosophila NMJ. We characterized a novel Neto isoform, Neto-β, which appears to be the predominant isoform at the fly NMJ. The two Drosophila isoforms, Neto-α and Neto-β, share the highly conserved extracellular and transmembrane domains, the hallmark of the Neto family of proteins [6], but have distinct cytoplasmic domains, rich in protein interaction motifs. We generated neto-β isoform specific mutants and found that the synaptic accumulation of type-A receptors is selectively impaired at neto-β mutant NMJs. Furthermore, Neto-β-deficient synapses fail to recruit PAK, a critical PSD signaling molecule required for the synaptic stabilization of type-A receptors and for the recruitment of other postsynaptic components. Muscle overexpression of either Neto-α or Neto-β isoforms could rescue the lethality and NMJ function of neto null mutants, indicating that the shared domains of Neto function to mediate iGluRs clustering. However, only Neto-β could restore the recruitment of PAK at neto null synapses. Our data demonstrate that Neto-mediated cytoplasmic interactions control the subtype composition of iGluRs and shape postsynaptic composition.

Results
Neto-β, a novel Neto isoform at the Drosophila NMJ Drosophila neto codes for two transcripts, cDNA references GH11189 and RE42119. They share the first 10 exons, encoding the extracellular and transmembrane part of Neto, but have alternative exons that generate different intracellular domains (Fig 1A). Exon 11 encodes the cytoplasmic domain of Neto-α, a 206-residue acidic domain (pI 3.88). Exons 12-14 encode the 351-residue Neto-β intracellular domain (pI 8.90). The two domains are predicted to contain multiple phosphorylation and protein interaction motifs (Neto-β shown in S1 Fig), but they show no homology with other Neto proteins. In fact, unlike the highly conserved extracellular and transmembrane domains of Neto proteins, the intracellular parts are highly divergent or even missing, such as in the C. elegans Neto (Fig 1A) [22][23][24].
Previous RNA-Seq analyses indicated that neto-α and neto-β transcripts are expressed throughout development, including larval stages. Using RT-PCR and qPCR analyses we also detected both transcripts in the third instar larval carcasses and estimated that the neto-α transcript was~4 fold less abundant than neto-β. To examine if Neto-β is present at the NMJs, we generated Neto-β isoform specific antibodies (details in Materials and Methods and S2 Fig) and compared them with previously described Neto antibodies, raised against the extracellular CUB1 domain [12]. The Neto-β positive signals co-localized with the Neto-ex puncta at control and at neto-β-rescued NMJs, but not at neto null NMJs rescued with neto-α transgenes ( Fig 1B). Importantly, muscle overexpression of either isoform could rescue the embryonic lethality and paralysis of neto null mutants. This suggests that the shared part of Neto, including the CUB1, CUB2, LDLa and the transmembrane pass, contains all the components essential for iGluRs clustering and NMJ development. Indeed, muscle expression of a Neto variant in which the intracellular domains were replaced by eGFP rescued the neto null paralyzed embryos to fertile Neto-β is a novel Neto isoform at the Drosophila NMJ. (A) Diagram of the Drosophila neto locus. Two isoforms are generated by alternative splicing, Neto-α and Neto-β. They have different cytoplasmic domains, but share highly conserved domains, CUB (complement C1r/C1s, Uegf, BMP1), LDLa (LDL receptor class a) and transmembrane (TM), with Neto proteins from vertebrates and C. elegans. (B) Confocal images of third instar larvae NMJ4 from neto null animals rescued with neto-α (neto 36 ;G14>neto α ), neto-β (neto 36 ;G14>neto β ), and a neto transgene lacking any intracellular part, netoΔ intra (neto 36 ; G14>netoΔ intra ). These Neto variants can rescue the viability and NMJ development in neto null animals. The Neto-ex signals mark all NMJs, but the Neto-β adults ( Fig 1B). The caveat of these rescue experiments is that Neto overexpression highly exceeds the endogenous levels and may obscure the contribution of individual isoforms. The function of the extracellular and transmembrane domains of Drosophila Neto will be examined elsewhere. Here we will focus on characterizing the new isoform, Neto-β, and its role during NMJ development.
We have previously demonstrated an essential role for Neto in the formation of iGluR clusters during synapse assembly and development. Likewise, Neto-β started to accumulate at the NMJ synapses at the onset of synaptogenesis (Fig 1C and 1D). In late embryo fillets, Neto-β antibodies labeled distinct synaptic puncta, which were also Neto-ex positive and co-localized with the iGluR synaptic signals (GluRIIA shown in Fig 1D). Neto-β signals remained at junctional locations throughout larval development and appeared to represent a significant fraction of the total Neto at the Drosophila NMJ (S2 Fig).

Generation of neto-β isoform specific mutants
To examine the role of Neto-β in synapse assembly and development we generated isoform specific mutants by imprecise excision of a transposable element Mi(ET1)Neto[MB07125] [25]. Several lines including precise excisions and small deletions were isolated and characterized by PCR amplification and sequencing (Fig 2A and Materials and Methods). The neto 203 allele is partly missing neto-β specific exons; neto 204 lacks the entire neto-β specific coding sequence and is likely a neto-β genetic null. No neto-β transcript was detectable in neto 204 animals, but in neto 203 larvae we found a truncated transcript, predicted to encode a short Neto-β variant (S1 Fig). For clarity we will refer to neto 203 as neto βshort , neto 204 as neto βnull , neto 36 as neto null , and neto 109 as neto hypo [12]. A diagram of the neto-β alleles and predicted Neto-β proteins is shown in Fig 2A and 2B.
The predicted Neto-β variants were detected by Western analysis in larval muscle extracts ( Fig 2C): a full-length Neto-β was detected in control, and a band of 65 kD, corresponding to the truncated Neto-β variant, was detected in neto βshort larval muscle. Interestingly, no band corresponding to Neto-α was detected suggesting that the endogenous Neto-α levels are very low, even in the absence of Neto-β. The Neto-β variants (full length and short) appeared expressed at equivalent levels relative to Tubulin. However, the synaptic Neto levels were decreased by 34% at neto βshort NMJs, suggesting that the cytoplasmic domain influences Neto distribution ( Fig 2C and S3 Fig). The Neto levels were decreased by 70% at neto βnull NMJs, consistent with the estimated 4-fold difference in the transcript levels. Further purification of Neto-β polyclonal sera against the two antigenic peptides (β1 and β2) produced two pools of antibodies ( Fig 2B and S1 Fig). As expected, only Neto-β2 antibody labeled the neto βshort synapses, whereas no Neto-β1 signals were found at neto βshort or neto βnull NMJs (Fig 2D and 2E). In contrast, clear Neto-β1 synaptic signals were detected at neto hypo NMJs. The neto hypo allele lacks the exon containing the translation initiation codon [12]. This lesion should affect both Neto isoforms.

neto-β mutants have impaired NMJ physiology and morphology
To test the functionality of NMJ synapses lacking an intact Neto-β isoform we recorded spontaneous miniature potentials (minis or mEJPs) and evoked excitatory junctional potentials (EJPs) from muscle 6 of third instar larvae. The mean frequency of mEJPs was decreased in both neto-β alleles compared to control (Fig 3A-3C and S1 Table). Mini amplitude or quantal size, the postsynaptic response to the fusion of a single vesicle, was also reduced in both neto-β alleles (to 62% of control in neto βshort , and respectively 43% in neto βnull ). We found no significant change in the resting potential and input resistance in the neto-β mutant larvae.
The reduced mEJP frequency and amplitude at neto-β mutant NMJs were similar but less severe than at Neto-or iGluR-deprived NMJs [9,12]. But while Neto-deprived NMJs have severely reduced EJPs, the amplitude of EJPs was reduced by 12-14% compared to the control in neto βshort mutant larvae and was relatively normal at neto βnull NMJs (Fig 3D-3F and S1 Table). This suggests that neto-β mutant NMJs must compensate for the decreased postsynaptic sensitivity by a compensatory increase in quantal content, the number of vesicles released in response to each action potential. We found that quantal content, estimated as ratio of average EJP amplitude to the mEJP amplitude, was increased 1.4 fold in neto βshort larvae and 2.7 fold in neto βnull compared with control ( Fig 3F). In contrast, the neto hypo or neto RNAi larvae show no presynaptic compensatory response to reduced quantal size [12,27].
These studies show that the loss of Neto-β significantly alters the number, type and/or density of postsynaptic iGluRs and elicits a homeostatic compensatory increase of presynaptic release. These changes resemble the GluRIIA loss-of-function phenotypes, and differ from Netoor iGluR-deprived NMJs [7][8][9][10][11][12]. Thus, the defects observed in neto-β mutants do not seem to originate from a general decrease in Neto levels and suggest isoform specific Neto activities at the NMJ.
NMJ morphological analyses further emphasized these specific defects. We have previously shown that Neto-deprived NMJs have longer and fewer branches, with significantly reduced number of boutons [27]. In contrast, at similarly reduced Neto levels, the neto βnull NMJs were reduced in length and had shorter branches (Fig 3G and 3H). Compared with controls, neto-β mutant NMJs had fewer but larger type Ib boutons (Fig 3I and 3J). In particular, the distal boutons had an estimated 2.4 fold volume increase (neto βshort : 18.9 μm 3 ± 2.6, n = 29, neto βnull : 20.8 μm 3 ± 2.3, n = 45, and control: 7.8 μm 3 ± 1.3, n = 29). Together these data indicate that perturbations of Neto-β have profound effects on the NMJ morphology and function that are different from a reduction in Neto levels.
neto βshort (6), and neto βnull (7) third instar larvae. Full length (black arrow) and truncated (white arrow) Neto-β are indicated. No specific signal was detected in neto βnull animals. (D-E) Confocal images of boutons at NMJ4 of third instar larvae stained with Neto-ex (red) and with either Neto-β2 (D) or Neto-β1 (E) (green). As expected, Neto-ex signals were detected in control (precise excision) and neto alleles. neto βshort NMJs show Neto-β2 signals (D) but both neto-β alleles lack Neto-β1 synaptic signals (E). In contrast, Neto-β1 puncta are present at neto hypo NMJs. Scale bars: 2 μm. doi:10.1371/journal.pgen.1005191.g002  Table. The number of NMJs examined is indicated in each bar. The mEJPs amplitude (B) and frequency (C) were reduced for both neto-β alleles. The EJPs amplitude was reduced in neto βshort compared to controls, but was normal at neto βnull NMJs (E). Both alleles had increased quantal content, indicating a presynaptic compensatory response (F). The muscle resting potential and the input resistance were not affected. (G-H) Representative confocal images at

Loss of Neto-β affects PSD composition
At the Drosophila NMJ, Neto activities are essential for trafficking and clustering of iGluRs at synaptic sites. To assess the contribution of Neto-β to these functions we examined the iGluRs distribution at neto-β mutant NMJs. For all the quantitative analyses below control/neto-β mutant sets (5-10 larvae/genotype) were processed together and imaged under the same conditions to capture and compare the synaptic signals. At the fly NMJ synapses, the sites of neurotransmitter release are marked by presynaptic specializations called T-bars, where Bruchpilot (Brp), the fly homolog of the vertebrate active zone protein ELKS, accumulates [28]. Opposite to the T-bars, the iGluRs are concentrated and stabilized at synaptic sites by a myriad of PSD-associated proteins. We found that neto-β mutant boutons contained more synaptic contacts, as visualized by Brp/GluRIIC juxtaposing puncta, and their density was increased at proximal boutons but not at the distal ones ( Fig 4A and 4B). This increase in synaptic density is consistent with the compensatory response observed at neto-β mutant NMJs ( Fig 3F). The intensity of Brp signals was largely constant, but the GluRIIC levels were diminished to 78% from control at neto βshort NMJs (n = 10) and to 39% at neto βnull NMJs (n = 10) (Fig 4A and  4C). The reduction of GluRIIC synaptic signals at neto-β mutant NMJs occurred without a change in the GluRIIC net protein indicating a defect in the synaptic localization of the receptors ( Fig 4D). Importantly, the GluRIIC synaptic signals paralleled the levels of synaptic Neto observed at these NMJs ( Fig 2D and 2E, S3A Fig), as Neto is limiting for receptors localization.
Drosophila NMJ utilizes two types of iGluRs, type-A and type-B, that differ in their subunits composition and properties [13]. The single channel current amplitude of both receptor types is identical, but the type-B channel desensitizes nearly ten times faster, thus the relative ratio of synaptic type-A/type-B receptors is a key determinant of quantal size [8]. Also, reduced levels of postsynaptic type-A receptors trigger a robust presynaptic compensatory response [7]. The physiological similarities between the GluRIIA and neto-β mutant NMJs suggest that Neto-β may primarily influence the synaptic accumulation of type-A receptors. We tested this possibility by examining the relative intensity of GluRIIA and GluRIIB signals at neto-β mutant NMJs. The neto βshort larvae had severely reduced GluRIIA synaptic levels (to 26% from control, n = 13), accompanied by no significant change in the GluRIIB levels ( Fig 4C and 4E). The neto βnull NMJs showed an even stronger reduction in the GluRIIA synaptic levels (to 18.5% from control, n = 14) and also decreased GluRIIB levels (Fig 4C and 4E). In contrast, the neto hypo NMJs showed uniformly reduced levels for all iGluR subunits examined (S3 Fig). The drastic decrease of GluR-IIA synaptic signals at neto-β mutant NMJs exceeded the reduction of GluRIIC and Neto synaptic levels observed at these NMJs, indicating a key role for the Neto-β isoform in regulating the GluRIIA synaptic abundance.
How does Neto-β regulate the selective accumulation of type-A receptors at synaptic sites? Neto-β may preferentially recruit the type-A receptors and/or influence their synaptic stabilization. If Neto isoforms promote the synaptic accumulation of selective iGluRs, and Neto-β influences the type-A receptors, the inference is that Neto-α should preferentially impact the distribution of type-B receptors. In support for this possibility, in RNAi knockdown experiments we found that Neto-α-deprived NMJs showed a dramatic loss of GluRIIB synaptic signals (by 68%) and a modest, variable increase of synaptic GluRIIA (by 18%) (Fig 4F and 4G). The efficiency of the RNAi-mediated knockdown was verified by Western blot analysis of muscle extracts from neto null larvae rescued with a V5-tagged neto-α transgene (Fig 4H and [12]). NMJ4 and NMJ6/7 (segment A3) in third instar larvae of indicated genotypes labeled with HRP. The neto-β mutant NMJs have significantly fewer and larger boutons relative to the muscle area (quantified in I-J). In particular, the volume of distal boutons increases 2.4 and respectively 2.6 fold at neto βshort and neto βnull NMJs. Error bars indicate SEM. ***; p<0.001, **; p<0.005, *; p<0.05, ns; p>0.05. Scale bars: 20 μm, 2 μm in details.
doi:10.1371/journal.pgen.1005191.g003 Lack of Neto-α-specific reagents precluded us from expanding these experiments. Nonetheless, these results raise the possibility that Neto isoforms utilize their distinct cytoplasmic domains to modulate the synaptic abundance of specific receptor subtypes. In the case of Neto-β, the last 260 C-terminal cytoplasmic residues appear to be critical for the synaptic accumulation of type-A receptors.

Neto-β recruits postsynaptic components and organizes postsynaptic structures
In Drosophila, several postsynaptic and perisynaptic proteins influencing the synaptic abundance of type-A and type-B receptors have been reported [15,16,18]. For example, PAK, a PSD component that co-localizes with the GluRIIA subunit, was shown to promote accumulation of type-A receptors at synaptic sites [16,17,29]. PAK also influences the size of the SSR, a stack of membrane folds that underlines the postsynaptic cell membrane of the type Ib boutons [16,17].
We found that PAK signals were severely reduced at neto-β mutant NMJs (Fig 5A). PAK synaptic levels dropped to 4% from control in neto βshort larvae (n = 11) and to 19% in neto βnull (n = 12). This drastic decrease exceeded the reduction in PAK signals observed at neto hypo NMJs, suggesting that the low level of Neto-β at neto hypo NMJs could partly mediate PAK stabilization at PSDs (Fig 2E and [12]). A similar, albeit less pronounced effect was observed for Dlg ( Fig 5B-5D). The Dlg junctional signals were reduced to 48% from control in neto βshort third instar larvae (n = 14) and to 27% from control in neto βnull (n = 11). Importantly, the levels of PAK and Dlg net proteins were not changed in extracts from neto-β larval muscles ( Fig 5E). Also, the accumulation of PAK and Dlg at junctional locations was completely restored in neto-β mutants by a duplication covering the neto locus (S4 Fig), indicating that the observed phenotypes are neto specific. In contrast, cysteine string protein (CSP), which labels clusters of synaptic vesicles in the vicinity of presynaptic membranes [30], and α-Spectrin, which labels the presynaptic axon and surrounds the bouton postsynaptically [31] appeared normal at netoβ mutant NMJs (S5 Fig).
Under electron microscopy, control synapses are marked by presynaptic specializations called T-bars and electron-dense membranes, corresponding to PSDs (Fig 6A-6C). At neto-β mutant boutons, serial sections revealed that the PSDs were significantly reduced in size but increased in number (Fig 6B-6D and Fig 4). The diminished PSD length (up to 2 fold in neto βnull mutants) matched the observed reduction of postsynaptic receptor fields (Fig 4). A thick stack of SSR membranes surrounds the control type Ib boutons. In contrast, the enlarged neto-β mutant boutons had diminished or even absent SSR structures, consistent with the reduced Dlg levels observed (Fig 6E-6H). In addition, neto-β mutants appeared to have misshaped T-bar structures, including double T-bar structures (Fig 6B and 6C, details). GluRIIA deprivation induces excessive recruitment of Brp at the active zones which sometimes results in active zone profiles with two or more T-bars [32]. Likewise, the excessive T-bar structures observed primarily in neto βnull mutant boutons may correlate to a homeostatic compensatory response triggered by the GluRIIA depletion (Figs 3F and 4B and 4E). accumulation of GluRIIA is severely reduced at neto-β mutant NMJs (quantified in C). In contrast, the GluRIIB synaptic signals are slightly increased at neto βshort NMJs and significantly reduced (by 31%) at neto βnull NMJs. (F) Confocal images of NMJ4 boutons in control and neto-α RNAi larvae labeled for GluRIIA (red), GluRIIB (green), and HRP (blue). The GluRIIB signals, but not GluRIIA are reduced at Neto-α-depleted NMJs (quantified in G). (H) Western blot analysis of larval muscle extracts from neto null mutants rescued by V5-tagged Neto-and carried through RNAi-mediated knockdown as indicated. The neto-RNAi appears to be more effective than the CUB1 RNAi in knocking down V5-tagged Neto-relative to the Tubulin control. Error bars indicate SEM. ***; p<0.001, **; p<0.005, *; p<0.05, ns; p>0.05. Scale bars: 2 μm. doi:10.1371/journal.pgen.1005191.g004 Neto-β, but not Neto-α, is required for PAK accumulation at PSDs Since the SSR folds are reduced in neto-β mutants, it is possible that the reduced PAK and Dlg synaptic levels are indirectly caused by the lack of proper SSR structures. To test this possibility,  we examined late embryos and first instar larvae, before the SSR develops [33]. If the SSR controls the synaptic accumulation of PAK, then PAK levels should be normal at neto-β mutant NMJs during the early stages of development. However, we found that PAK accumulated normally at the muscle attachment sites in control and neto-β mutant animals, but was severely diminished at neto-β mutant NMJs (Fig 7A). Thus, loss of synaptic PAK is not secondary to the loss of SSR and appears to be directly caused by impairments in Neto-β.
If Neto-β recruits PAK at the cell membrane, then loss of synaptic PAK in neto-β mutants should be rescued by muscle expression of PAK myr , a membrane-tethered PAK variant [34]. We found that muscle overexpression of PAK myr in neto-β mutants induced accumulation of PAK signals at perisynaptic but not synaptic sites, and could not rescue the GluRIIA synaptic accumulation (neto βshort shown in Fig 7B). This indicates that Neto-β controls the recruitment of PAK at synaptic locations. Loss of PAK did not affect the synaptic recruitment of Neto-β: while PAK signals were lost at pak null NMJs (pak 11/Df ) or in mutants that fail to localize at cell peripheries (pak 6/Df or pak 6/11 ) [35], the synaptic Neto signals remained unaffected, irrespective of the pak genetic manipulations (Fig 7C). Also, the levels of GluRIIA were 4-fold reduced in neto-β mutants (Fig 4), but only 2-fold in pak mutants [16], suggesting that Neto-β has additional, PAK-independent functions in the synaptic recruitment/stabilization of GluRIIA. Together these data indicate that Neto-β is required for the synaptic accumulation of PAK. In flies and mammals, PAK membrane localization is controlled by Pix, a Rho-type guanine nucleotide exchange factor [16,36]. Mutations in dpix led to impairments in synaptic accumulation of PAK, GluRIIA, Dlg, and other postsynaptic components. Lack of Pix antibodies precluded us to test whether Neto-β recruits dPix at PSDs. Nonetheless, while the PAK synaptic signals were completely lost in dpix mutants, Neto signals accumulated at these mutant synapses albeit they appeared somewhat reduced, likely due to defects in the synapse organization (S6 Fig). Thus, dPix is required for PAK but not for Neto recruitment of at synaptic sites.
We have previously demonstrated that the net levels of Neto are critical for synapse assembly and NMJ development [12,14]. In particular, Neto depletion has profound consequences for synapse assembly and function. The levels of synaptic Neto are reduced in neto-β mutants compared with controls (Fig 2 and S3 Fig) raising the possibility that the phenotypes observed in neto-β allelic series could be partly due to reduced Neto levels. To distinguish between isoform specific and suboptimal Neto phenotypes we manipulated the muscle expression of Netoα and Neto-β isoforms and examined their effects on the development and function of NMJ. Similar to neto-α, we found that muscle expression of neto-β transgenes rescued the viability and NMJ development of neto null mutants and induced gain-of-function phenotypes in a concentration-dependent manner (Fig 8A and [27]). Low to moderate levels of Neto-β induced formation of NMJs with relatively normal morphology and clearly defined synaptic Neto clusters, resembling the neto mutants rescued with low and moderate levels of Neto-α. Excess Neto-β appeared detrimental to the NMJ development and to the overall growth and viability of rescued animals (Fig 8A).
Intriguingly, Neto-β but not Neto-α enabled the stabilization of PAK at PSDs, over a wide range of concentrations tested (Fig 8A). Furthermore, increasing the levels of Neto-α in neto-β mutants could not restore the PAK synaptic accumulation, while introducing Neto-β effectively rescued this defect. This result cannot be explained by a difference between Neto-α and Neto-β cellular distribution, since both isoforms appear to concentrate at the NMJ. We confirmed that Neto-α and Neto-β synaptic levels correlate with the levels of Neto protein in the larval muscle, as detected by Western analysis (Fig 8B). Thus, both Neto isoforms can traffic and accumulate at synaptic locations where they mediate synapse assembly. However, only Neto-β isoform can enable PAK accumulation at PSDs. Since PAK contributes to synaptic stabilization of GluRIIA, Neto-α-rescued NMJs are expected to exhibit a reduction in the synaptic GluRIIA signals and thus a reduction in quantal size. Indeed, we found that neto null mutants rescued with low/moderate levels of Neto-α had significantly reduced mini amplitudes and reduced GluRIIA/GluRIIB ratio (Fig 8C and 8D,  S7 Fig). In contrast, neto null NMJs rescued with neto-β transgenes had relatively normal mini amplitudes, indicating normal GluRIIA/GluRIIB ratio at these synapses. The mini frequency was largely normal, except for the neto-β transgenes which showed increased mini frequency when reared at 18°C (Fig 8E). Interestingly, the evoked potentials were in the normal range in all Neto-α and Neto-β-rescued larvae tested, likely due to compensatory presynaptic response (Fig 8F-8H). These results indicate that Neto-α-rescued NMJs have deficits in the synaptic accumulation of type-A receptors. Such defects are partly obscured by excess Neto-α, likely because the conserved domains of Neto confer high iGluRs "clustering capacity" in these rescue experiments. But under normal condition, Neto is a low abundance protein and a limiting factor for iGluRs clustering in the muscle. Furthermore, Neto-β appears to be the predominant isoform at the Drosophila NMJ. Together our findings demonstrate that Neto-β, the major Neto isoform at the Drosophila NMJ, controls the subtype composition of iGluRs partly by regulating the recruitment of the PSD-associated kinase PAK.

Discussion
At the Drosophila NMJ, Neto enables iGluRs clustering at synaptic sites and promotes postsynaptic differentiation. Here we show 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. Our 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.

Neto-mediated intracellular interactions shape postsynaptic composition
Neto proteins have been initially characterized as auxiliary subunits that modulate the function of kainate (KA) and NMDA receptors [22,23]. In vertebrates, Neto1 and Neto2 directly interact with KAR subunits and increase channel function by modulating gating properties [23,37,38]. 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 [23,39,40]. Furthermore, the C. elegans Neto has a very small intracellular domain (24 amino acids beyond the conserved domains) [24]. 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 perisynaptically in the absence of an intact Neto-β (see bouton details). (C) Confocal images of NMJ4 in control (w 1118 ) and various pak heteroallelic third instar larvae labeled for Neto-ex (red), PAK (green) and HRP (blue). Lack of PAK does not impact the synaptic distribution of Neto-β. Scale bars: (A) 10 μm; (B-C) 20 μm, 2 μm in details.
doi:10.1371/journal.pgen.1005191.g007 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 [41].
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 [27]. Second, the two Neto isoforms appear to facilitate the selective recruitment and/or stabilization of specific iGluR subtypes. We speculate 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 [42,43]. 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 [17].
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, we failed to detect a direct interaction between PAK and Neto-β (full-length or intracellular domain). PAK synaptic accumulation is completely abolished at NMJ with mutations in dPix, although not all dpix defects are mediated through PAK [16]. Conversely, PAK together with Dreadlocks (Dock) controls the synaptic abundance of GluRIIA, while PAK and dPix regulate the Dlg distribution [17]. 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 [15].
doi:10.1371/journal.pgen.1005191.g008 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 neto hypo larvae [12]. Assembly of this network does not require GluRIIA since both Neto-β and PAK accumulated normally at GluRIIA mutant NMJs (Fig 8I). Furthermore, in the absence of Neto-β the synaptic abundance of GluRIIA can be partly restored by excess Neto-α or a delta-intracellular Neto variant, suggesting that excess iGluRs "clustering capacity" overrides the cellular signals that shape PSD composition [27]. 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.

Expanding the repertoire of Neto functions at glutamatergic synapses
The discovery of Drosophila Neto isoforms with alternative cytoplasmic domains and isoform specific activities expands the repertoire of Neto-mediated functions at glutamatergic synapses. On one hand, 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 [22,23,44]. 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 [24]. The only exception is a retina-specific CUB1-only Neto1 isoform with unknown function [45]. In contrast, 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. Alternatively, 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.
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 [46,47]. 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 [1,6].
Neto proteins modulate the gating behavior of KAR but also play crucial roles in the synaptic recruitment of glutamate receptors in vivo [12,40,48]. 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.

Fly stocks
To generate neto-β alleles, the Minos transposomal element Mi(ET1)Neto[MB07125] was mobilized with Minos transposase [25]. Several lines with precise excisions and imprecise excisions/ small deletions were isolated and characterized by PCR amplification over the deficiencies and DNA sequencing. The genomic fragments removed in various neto alleles were as follows: neto 203 X: 13,415,418,117, including part of the exon 13 and 14 of the predicted neto gene, and neto 204 X: 13,414,477-13,417,931, containing exons 12, 13 and 14. The UAS-neto-β lines (B lines) were generated by insertion of the neto-β cDNA (from RE42119) in pUAST vector and germline transformation (BestGene, Inc.). Similarly, for the UAS-netoΔintra (the H4 line) the Neto coding sequence M 1 -R 471 followed by a short linker (DVPALE) was placed in frame with the eGFP and cloned in pUAST. The neto-α RNAi lines were generated by insertion of a neto-α specific PCR fragment in pUAST-R57 followed by germline transformation. The PCR primers utilized were: RNAi-F (5 0 -AAGGCCTACATGGCCGGACCGGCGAACAAATGGAG GAAGACG-3 0 ) and RNAi-Rev (5 0 -AATCTAGAGGTACCTGATTTTGTGCAGGAACTTG AGG-3 0 ).

Antibody generation and protein analysis
The rabbit polyclonal anti-Neto-β antibodies were generated against two synthetic peptides: β1 (GRSHYGGLLVTSTAKQP) and β2 (LDDVSNRYYREAVPL) (21st Century Biochemicals) and separated by affinity purification. The rabbit polyclonal anti-GluRIIB and anti-GluRIIC were generated as previously described [8] against synthetic peptides ASSAKKKKKTRRIEK, and respectively QGSGSSSGSNNAGRGEKEARV (Pacific Immunology Corp).
To analyze muscle proteins, wandering third instar larvae were dissected, and all tissues except for the body wall (muscle and cuticle) were removed. The body walls were mechanically disrupted and lysed in lysis buffer (50 mM Tris-HCl, 150 mM NaCl, 1% Triton X-100, 1% deoxycholate, protease inhibitor cocktail (Roche) for 30 min on ice. The lysates were separated by SDS-PAGE on 4%-12% NuPAGE gels (Invitrogen) and transferred onto PVDF membranes (Millipore). Drosophila S2 cells were used for the production of recombinant proteins [12]. Full-length cDNA for neto-β from RE42119 was subcloned in pAcPA-based plasmids for expression in S2 cells under the actin promoter. The Neto-β truncation (neto 203 -like) was generated by looping out the deleted fragment using the QuickChange site-directed mutagenesis kit. All constructs were verified by DNA sequencing.

Immunohistology
Wandering third instar larvae were dissected as described previously in ice-cooled Ca 2+ -free HL-3 solution [49,50]. Embryos at 18h after egg laying (AEL) were dechorinated and genotyped and, after an additional incubation of 2 h at room temperature, were dissected as described previously [12]. First instar larvae were similarly dissected within 2 h from hatching. The samples were fixed in 4% paraformaldehyde (Polysciences, Inc.) for 20 min or in Bouin's fixative (Bio-Rad) for 3 min and washed in PBS containing 0.5% Triton X-100.
Samples of different genotypes were processed simultaneously and imaged under identical confocal settings in the same imaging session with a laser scanning confocal microscope (Carl-Zeiss LSM780). All images were collected as 0.2m optical sections and the z-stacks were analyzed with Imaris software (Bitplane). To detect positive puncta we used the spot finding Imaris algorithm followed by manual inspection and correction. To quantify fluorescence intensities synaptic ROI areas surrounding anti-HRP immunoreactivities were selected and the signals measured individually at NMJs (muscle 4, segment A3) from ten or more different larvae for each genotype (number of samples is indicated in the graph bar). The signal intensities were calculated relative to HRP volume and subsequently normalized to control. Boutons were counted in preparations double labeled with anti-HRP and anti-Dlg. All quantifications were performed while blinded to genotype. The numbers of samples analyzed are indicated inside the bars. Statistical analyses were performed using the Student t-test with a two-tailed distribution and a two-sample unequal variance. Error bars in all graphs indicate standard deviation ±SEM. ÃÃÃ ; p<0.001, ÃÃ ; p<0.005, Ã ; p<0.05, ns; p>0.05.

Electrophysiology
The standard larval body wall muscle preparation first developed by Jan and Jan (1976) [51] was used for electrophysiological recordings [52]. Wandering third instar larvae were dissected in physiological saline HL-3 saline [49], washed, and immersed in HL-3 containing 0.8 mM Ca 2+ or 0.3 mM Ca 2+ using a custom microscope stage system [53]. The nerve roots were cut near the exiting site of the ventral nerve cord so that the motor nerve could be picked up by a suction electrode. Intracellular recordings were made from muscle 6. Data were used when the input resistance of the muscle was >5 MO and the resting membrane potential was between -60 mV and -80 mV. The input resistance of the recording microelectrode (backfilled with 3 M KCl) ranged from 20 to 25 MO. Muscle synaptic potentials were recorded using an Axon Clamp 2B amplifier (Axon Instruments) and pClamp 10 software. Following motor nerve stimulation with a suction electrode (100 μsec, 5 V), evoked EJPs were recorded. Four to six EJPs evoked by low frequency of stimulation (0.1 Hz) were averaged. For mini recordings, TTX (1 μM) was added to prevent evoked release [49]. To calculate mEJP mean amplitudes, 50-200 events from each muscle were measured and averaged using the Mini Analysis program (Synaptosoft). Minis with a slow rise and falling time arising from neighboring electrically coupled muscle cells were excluded from analysis [54,55]. Quantal content was calculated by dividing the mean EJP by the mean mEJP after correction of EJP amplitude for nonlinear summation according to previously described methods [56,57]. Corrected EJP amplitude = E[Ln [E/(E-recorded EJP)]], where E is the difference between reversal potential and resting potential. The reversal potential used in this correction was 0 mV [57,58]. Data are presented as mean ±SEM, unless otherwise specified; EJP amplitudes and quantal contents after the nonlinear correction are shown. Student T-test was used to assess statistically significant differences among the genotypes.

Electron microscopy
Wandering third instar larvae were dissected in physiological saline HL-3 saline and fixed for 30 min on dissection plate in fixation buffer (0.1 M Sodium Cacodylate buffer, pH7.2; 2 mM MgCl 2 ; 1% glutaraldehyde; 4% paraformaldehyde). The samples were trimmed to include only the abdominal segments A2 and A3, transferred in a 1.5mL Eppendorf tube, fixed overnight at 4°C, then washed extensively with 0.1 M Sodium Cacodylate buffer with 132 mM Sucrose, pH 7.2. The samples were further processed and analyzed according to published protocols [59] at the Microscopy and Imaging Facility, NICHD.
Supporting Information S1 Fig. Sequence alignment for Neto-β cytoplasmic domain. ClustalW alignment of Neto-β cytoplasmic domains in several Drosophila species (Drosophila melanogaster, Drosophila ananassae and Drosophila pseudoobscura) reveals blocks of highly conserved sequences, including SH2 and SH3 binding domains, poly-H and poly-Q motifs, and putative phosphorylation sites (putative CaMKII phosphorylation site is indicated in red). The Neto-β truncated form (neto βshort allele) is predicted to retain the first 88 conserved intracellular residues followed by 8 additional, unrelated residues (shown in orange). The synthetic peptides used to generate the Neto-β isoform specific antibodies are marked by blue boxes. Neto-ex rat polyclonal antibodies label both Neto isoforms, while the Neto-β rabbit polyclonal antibodies can only recognize the Neto-β recombinant protein. Arrowheads point to unprocessed and processed Neto-α (white) and Neto-β (black). The apparent molecular weights are higher than predicted likely due to post-translational modifications:~100/85 kD observed for unprocessed/processed Neto-α variants (75/62 kD calculated) and~115/100 kD for Neto-β (92/77 kD calculated). (B) Representative confocal images of NMJ4 boutons in third instar larvae labeled for Neto-β (green), HRP (blue) and Neto-ex (red) (left panels) or GluRIIA (red) (right panels). Neto-β positive puncta co-localize with Neto-ex and GluRIIA signals at synaptic sites. Scale bars: 20 μm, 2 μm in details.